302 files changed, 48681 insertions, 45537 deletions

diff --git a/erts/emulator/Makefile.in b/erts/emulator/Makefile.in
index 1916b97a89..054692819e 100644
--- a/erts/emulator/Makefile.in
+++ b/erts/emulator/Makefile.in
@@ -1,7 +1,7 @@
 #
 # %CopyrightBegin%
 #
-# Copyright Ericsson AB 1996-2017. All Rights Reserved.
+# Copyright Ericsson AB 1996-2018. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -28,10 +28,22 @@ HIPE_ENABLED=@HIPE_ENABLED@
 DTRACE_ENABLED=@DTRACE_ENABLED@
 DTRACE_ENABLED_2STEP=@DTRACE_ENABLED_2STEP@
 USE_VM_PROBES=@USE_VM_PROBES@
+FPE=@FPE@
 LIBS = @LIBS@
 Z_LIB=@Z_LIB@
 NO_INLINE_FUNCTIONS=false
-OPCODE_TABLES = $(ERL_TOP)/lib/compiler/src/genop.tab beam/ops.tab
+OPCODE_TABLES = $(ERL_TOP)/lib/compiler/src/genop.tab \
+beam/ops.tab \
+beam/macros.tab \
+beam/instrs.tab \
+beam/arith_instrs.tab \
+beam/bif_instrs.tab \
+beam/bs_instrs.tab \
+beam/float_instrs.tab \
+beam/map_instrs.tab \
+beam/msg_instrs.tab \
+beam/select_instrs.tab \
+beam/trace_instrs.tab
 
 DEBUG_CFLAGS = @DEBUG_CFLAGS@
 CONFIGURE_CFLAGS = @CFLAGS@
@@ -51,7 +63,27 @@ ARFLAGS=rc
 OMIT_OMIT_FP=no
 TYPE_LIBS=
 
-DIRTY_SCHEDULER_SUPPORT=@DIRTY_SCHEDULER_SUPPORT@
+PROFILE_COMPILER=@PROFILE_COMPILER@
+PROFILE_MARKER=
+ifeq ($(PROFILE),generate)
+PROFILE_MARKER=_pg
+else
+ifeq ($(PROFILE),use)
+PROFILE_MARKER=_pu
+endif
+endif
+
+ifeq ($(PROFILE_COMPILER), gcc)
+PROFILE_GENERATE=-fprofile-generate
+PROFILE_USE=-fprofile-use -fprofile-correction
+PROFILE_USE_DEPS=$(OBJDIR)/%_pu.gcda
+endif
+ifeq ($(PROFILE_COMPILER), clang)
+PROFILE_GENERATE=-fprofile-instr-generate
+PROFILE_USE=-fprofile-instr-use=$(OBJDIR)/default.profdata
+PROFILE_USE_DEPS=$(OBJDIR)/default.profdata
+endif
+
 DIRTY_SCHEDULER_TEST=@DIRTY_SCHEDULER_TEST@
 
 ifeq ($(TYPE),debug)
@@ -178,31 +210,10 @@ endif
 # NOTE: When adding a new type update ERL_BUILD_TYPE_MARKER in sys/unix/sys.c
 #
 
-FLAVOR=$(DEFAULT_FLAVOR)
-
-ifeq ($(FLAVOR),plain)
-
-DS_SUPPORT=no
-DS_TEST=no
-
-FLAVOR_MARKER=
-FLAVOR_FLAGS=
-ENABLE_ALLOC_TYPE_VARS += nofrag
-M4FLAGS +=
-
-else # FLAVOR
-
-# If flavor isn't one of the above, it *is* smp flavor...
 override FLAVOR=smp
 
 FLAVOR_MARKER=.smp
-FLAVOR_FLAGS=-DERTS_SMP
-ENABLE_ALLOC_TYPE_VARS += smp nofrag
-M4FLAGS += -DERTS_SMP=1
-ifeq ($(DIRTY_SCHEDULER_SUPPORT),yes)
-THR_DEFS += -DERTS_DIRTY_SCHEDULERS
-DS_SUPPORT=yes
-
+ENABLE_ALLOC_TYPE_VARS += nofrag
 ifeq ($(DIRTY_SCHEDULER_TEST),yes)
 DS_TEST=yes
 THR_DEFS += -DERTS_DIRTY_SCHEDULERS_TEST
@@ -210,13 +221,6 @@ else # DIRTY_SCHEDULER_TEST
 DS_TEST=no
 endif # DIRTY_SCHEDULER_TEST
 
-else # DIRTY_SCHEDULER_SUPPORT
-DS_SUPPORT=no
-DS_TEST=no
-endif # DIRTY_SCHEDULER_SUPPORT
-
-endif # FLAVOR
-
 TF_MARKER=$(TYPEMARKER)$(FLAVOR_MARKER)
 
 ifeq ($(TYPE)-@HAVE_VALGRIND@,valgrind-no)
@@ -235,18 +239,9 @@ ARCH=@ARCH@
 ultrasparcCFLAGS=-Wa,-xarch=v8plusa
 ARCHCFLAGS=$($(ARCH)CFLAGS)
 
-ifdef HIPE_ENABLED
-ifeq ($(OPSYS),linux)
-ppcBEAMLDFLAGS=-Wl,-m,elf32ppc
-ppc64BEAMLDFLAGS=-Wl,-m,elf64ppc,-T,hipe/elf64ppc.x
-endif
-ifeq ($(OPSYS),darwin)
-amd64BEAMLDFLAGS=-pagezero_size 0x10000000
-endif
-HIPEBEAMLDFLAGS=$($(ARCH)BEAMLDFLAGS)
-endif
+HIPEBEAMLDFLAGS=@HIPEBEAMLDFLAGS@
 
-ERTS_ENABLE_KERNEL_POLL=@ERTS_ENABLE_KERNEL_POLL@
+ERTS_BUILD_FALLBACK_POLL=@ERTS_BUILD_FALLBACK_POLL@
 
 #
 #
@@ -425,9 +420,20 @@ ifeq ($(TARGET), win32)
 EMULATOR_EXECUTABLE = beam$(TF_MARKER).dll
 else
 EMULATOR_EXECUTABLE = beam$(TF_MARKER)
+PROFILE_EXECUTABLE = beam.prof$(TF_MARKER)
 endif
 CS_EXECUTABLE = erl_child_setup$(TYPEMARKER)
 
+ifeq ($(PROFILE), generate)
+EMULATOR_EXECUTABLE = $(PROFILE_EXECUTABLE)
+ifeq ($(PROFILE_COMPILER), gcc)
+PROFILE_LDFLAGS = -fprofile-generate
+endif
+ifeq ($(PROFILE_COMPILER), clang)
+PROFILE_LDFLAGS = -fprofile-instr-generate
+endif
+endif
+
 # ----------------------------------------------------------------------
 
 ifeq ($(ERLANG_OSTYPE), unix)
@@ -483,7 +489,6 @@ ifeq ($(TARGET),win32)
 RELEASE_INCLUDES += sys/$(ERLANG_OSTYPE)/erl_win_dyn_driver.h
 endif
 
-
 .PHONY: release_spec
 ifdef VOID_EMULATOR
 release_spec:
@@ -512,7 +517,9 @@ release_docs_spec:
 # Generated source code. Put in $(TARGET) directory
 #
 
+ifneq ($(strip $(CREATE_DIRS)),)
 _create_dirs := $(shell mkdir -p $(CREATE_DIRS))
+endif
 
 
 # has to be run after _create_dirs
@@ -548,10 +555,11 @@ DTRACE_HEADERS =
 endif
 
 ifdef HIPE_ENABLED
-OPCODE_TABLES += hipe/hipe_ops.tab
+OPCODE_TABLES += hipe/hipe_ops.tab hipe/hipe_instrs.tab
 endif
 
 $(TTF_DIR)/beam_cold.h \
+$(TTF_DIR)/beam_warm.h \
 $(TTF_DIR)/beam_hot.h \
 $(TTF_DIR)/beam_opcodes.c \
 $(TTF_DIR)/beam_opcodes.h \
@@ -561,8 +569,10 @@ $(TTF_DIR)/beam_tr_funcs.h \
 $(TTF_DIR)/OPCODES-GENERATED: $(OPCODE_TABLES) utils/beam_makeops
 	$(gen_verbose)LANG=C $(PERL) utils/beam_makeops \
 		-wordsize @EXTERNAL_WORD_SIZE@ \
+		-code-model @CODE_MODEL@ \
 		-outdir $(TTF_DIR) \
 		-DUSE_VM_PROBES=$(if $(USE_VM_PROBES),1,0) \
+		-DNO_FPE_SIGNALS=$(if $(filter unreliable,$(FPE)),1,0) \
 		-emulator $(OPCODE_TABLES) && echo $? >$(TTF_DIR)/OPCODES-GENERATED
 GENERATE += $(TTF_DIR)/OPCODES-GENERATED
 
@@ -599,7 +609,7 @@ $(HIPE_NBIF_FILES) \
 	: $(TTF_DIR)/TABLES-GENERATED
 $(TTF_DIR)/TABLES-GENERATED: $(ATOMS) $(DIRTY_BIFS) $(BIFS) utils/make_tables
 	$(gen_verbose)LANG=C $(PERL) utils/make_tables -src $(TTF_DIR) -include $(TTF_DIR)\
-		-ds $(DS_SUPPORT) -dst $(DS_TEST) -hipe $(HIPE) $(ATOMS) $(DIRTY_BIFS) $(BIFS) && echo $? >$(TTF_DIR)/TABLES-GENERATED
+		-dst $(DS_TEST) -hipe $(HIPE) $(ATOMS) $(DIRTY_BIFS) $(BIFS) && echo $? >$(TTF_DIR)/TABLES-GENERATED
 GENERATE += $(TTF_DIR)/TABLES-GENERATED
 
 $(TTF_DIR)/erl_alloc_types.h: beam/erl_alloc.types utils/make_alloc_types
@@ -626,6 +636,7 @@ GENERATE += $(TTF_DIR)/driver_tab.c
 PRELOAD_BEAM =		$(ERL_TOP)/erts/preloaded/ebin/otp_ring0.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/erts_code_purger.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/init.beam \
+			$(ERL_TOP)/erts/preloaded/ebin/prim_buffer.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/prim_eval.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/prim_inet.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/prim_file.beam \
@@ -636,7 +647,7 @@ PRELOAD_BEAM =		$(ERL_TOP)/erts/preloaded/ebin/otp_ring0.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/erts_internal.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/erl_tracer.beam \
 			$(ERL_TOP)/erts/preloaded/ebin/erts_literal_area_collector.beam \
-			$(ERL_TOP)/erts/preloaded/ebin/erts_dirty_process_code_checker.beam
+			$(ERL_TOP)/erts/preloaded/ebin/erts_dirty_process_signal_handler.beam
 
 ifeq ($(TARGET),win32)
 # On windows the preloaded objects are in a resource object.
@@ -693,17 +704,66 @@ $(OBJDIR)/beams.$(RES_EXT): $(TARGET)/beams.rc
 
 endif
 
-ifneq ($(filter tile-%,$(TARGET)),)
-$(OBJDIR)/beam_emu.o: beam/beam_emu.c
-	$(V_CC) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) \
-         $(INCLUDES) -c $< -o $@
-else
+# We disable the implicit rule of .S -> .o so that the verbose asm
+# generate is not used for compiling erts. This is only a problem on
+# old solaris make
+%.o : %.S
+
 # Usually the same as the default rule, but certain platforms (e.g. win32) mix
 # different compilers
 $(OBJDIR)/beam_emu.o: beam/beam_emu.c
 	$(V_EMU_CC) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+
+$(OBJDIR)/beam_emu.S: beam/beam_emu.c
+	$(V_EMU_CC) -S -fverbose-asm $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+
+$(OBJDIR)/%_pg.o: beam/%.c
+	$(V_CC) $(PROFILE_GENERATE) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+$(OBJDIR)/%_pu.o: beam/%.c $(PROFILE_USE_DEPS)
+	$(V_CC) $(PROFILE_USE) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+
+$(OBJDIR)/%_pu.S: beam/%.c $(PROFILE_USE_DEPS)
+	$(V_CC) -S -fverbose-asm $(PROFILE_USE) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+
+$(OBJDIR)/PROFILE: $(BINDIR)/$(PROFILE_EXECUTABLE)
+	$(V_at)echo " PROFILE ${PROFILE_EXECUTABLE}"
+	$(V_at)rm -f $(OBJDIR)/erl*.profraw
+	$(V_at)set -e; LLVM_PROFILE_FILE="$(OBJDIR)/erlc-%m.profraw" \
+          ERL_FLAGS="-emu_type prof${TYPEMARKER} +S 1" $(ERLC) -DPGO \
+	  -o $(OBJDIR) test/estone_SUITE.erl > $(OBJDIR)/PROFILE_LOG
+	$(V_at)set -e; LLVM_PROFILE_FILE="$(OBJDIR)/erl-%m.profraw" \
+	  ERL_FLAGS="-emu_type prof${TYPEMARKER} +S 1" $(ERL) -pa $(OBJDIR) \
+	  -noshell -s estone_SUITE pgo -s init stop >> $(OBJDIR)/PROFILE_LOG
+	$(V_at)touch $@
+
+.SECONDARY: $(patsubst %.o,%_pu.gcda,$(PROFILE_OBJS))
+
+$(OBJDIR)/%_pu.gcda: $(OBJDIR)/PROFILE
+	$(V_at)mv $(OBJDIR)/$*_pg.gcda $@
+	$(V_at)touch $@
+
+$(OBJDIR)/default.profdata: $(OBJDIR)/PROFILE
+	$(V_LLVM_PROFDATA) merge -output $@ $(OBJDIR)/*.profraw
+
+ifeq ($(ERTS_BUILD_FALLBACK_POLL),yes)
+# Have to treat erl_poll differently as the same .c file is used
+# twice for kernel poll builds.
+$(OBJDIR)/erl_poll.o: sys/common/erl_poll.c
+	$(V_CC) -DERTS_KERNEL_POLL_VERSION $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
+
+# Do a copy in order to make debuggers less confused
+$(TTF_DIR)/erl_poll.flbk.c: sys/common/erl_poll.c
+	$(V_at) cp $< $@
+	@touch $@
+
+$(OBJDIR)/erl_poll.flbk.o: $(TTF_DIR)/erl_poll.flbk.c
+	$(V_CC) -DERTS_NO_KERNEL_POLL_VERSION $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
 endif
 
+
+# ----------------------------------------------------------------------
+# General targets
+#
 $(OBJDIR)/%.o: beam/%.c
 	$(V_CC) $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
 
@@ -728,6 +788,9 @@ $(OBJDIR)/%.o: drivers/$(ERLANG_OSTYPE)/%.c
 $(OBJDIR)/%.o: nifs/common/%.c
 	$(V_CC) $(CFLAGS) -DLIBSCTP=$(LIBSCTP) $(INCLUDES) -Inifs/common -Inifs/$(ERLANG_OSTYPE) -c $< -o $@
 
+$(OBJDIR)/%.o: nifs/$(ERLANG_OSTYPE)/%.c
+	$(V_CC) $(CFLAGS) $(INCLUDES) -Inifs/common -Inifs/$(ERLANG_OSTYPE) -I../etc/$(ERLANG_OSTYPE) -c $< -o $@
+
 # ----------------------------------------------------------------------
 # Specials
 #
@@ -737,12 +800,6 @@ $(BINDIR)/$(CS_EXECUTABLE): $(TTF_DIR)/GENERATED $(PRELOAD_SRC) $(CS_OBJ) $(ERTS
 	$(ld_verbose)$(CS_PURIFY) $(LD) $(CS_LDFLAGS) -o $(BINDIR)/$(CS_EXECUTABLE) \
                $(CS_CFLAGS) $(COMMON_INCLUDES) $(CS_OBJ) $(CS_LIBS)
 
-$(OBJDIR)/%.kp.o: sys/common/%.c
-	$(V_CC) -DERTS_KERNEL_POLL_VERSION $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
-
-$(OBJDIR)/%.nkp.o: sys/common/%.c
-	$(V_CC) -DERTS_NO_KERNEL_POLL_VERSION $(subst -O2, $(GEN_OPT_FLGS), $(CFLAGS)) $(INCLUDES) -c $< -o $@
-
 ifeq ($(GCC),yes)
 
 $(OBJDIR)/erl_goodfit_alloc.o: beam/erl_goodfit_alloc.c
@@ -764,19 +821,20 @@ $(ERL_TOP)/lib/%.beam:
 
 INIT_OBJS = $(OBJDIR)/erl_main.o	$(PRELOAD_OBJ)
 
+PROFILE_OBJS = $(OBJDIR)/beam_emu.o	$(OBJDIR)/erl_process.o
+
 EMU_OBJS = \
-	$(OBJDIR)/beam_emu.o		$(OBJDIR)/beam_opcodes.o \
+	$(OBJDIR)/beam_opcodes.o \
 	$(OBJDIR)/beam_load.o		$(OBJDIR)/beam_bif_load.o \
 	$(OBJDIR)/beam_debug.o		$(OBJDIR)/beam_bp.o \
-	$(OBJDIR)/beam_catches.o \
-	$(OBJDIR)/code_ix.o \
+	$(OBJDIR)/beam_catches.o 	$(OBJDIR)/code_ix.o \
 	$(OBJDIR)/beam_ranges.o
 
-RUN_OBJS = \
+RUN_OBJS += \
 	$(OBJDIR)/erl_alloc.o		$(OBJDIR)/erl_mtrace.o \
 	$(OBJDIR)/erl_alloc_util.o	$(OBJDIR)/erl_goodfit_alloc.o \
 	$(OBJDIR)/erl_bestfit_alloc.o	$(OBJDIR)/erl_afit_alloc.o \
-	$(OBJDIR)/erl_instrument.o	$(OBJDIR)/erl_init.o \
+	$(OBJDIR)/erl_init.o \
 	$(OBJDIR)/erl_atom_table.o	$(OBJDIR)/erl_bif_table.o \
 	$(OBJDIR)/erl_bif_ddll.o  	$(OBJDIR)/erl_bif_guard.o \
 	$(OBJDIR)/erl_bif_info.o	$(OBJDIR)/erl_bif_op.o \
@@ -788,7 +846,7 @@ RUN_OBJS = \
 	$(OBJDIR)/utils.o		$(OBJDIR)/bif.o \
 	$(OBJDIR)/io.o 			$(OBJDIR)/erl_printf_term.o\
 	$(OBJDIR)/erl_debug.o		$(OBJDIR)/erl_md5.o \
-	$(OBJDIR)/erl_message.o		$(OBJDIR)/erl_process.o \
+	$(OBJDIR)/erl_message.o		$(OBJDIR)/erl_proc_sig_queue.o \
 	$(OBJDIR)/erl_process_dict.o	$(OBJDIR)/erl_process_lock.o \
 	$(OBJDIR)/erl_port_task.o	$(OBJDIR)/erl_arith.o \
 	$(OBJDIR)/time.o		$(OBJDIR)/erl_time_sup.o  \
@@ -802,11 +860,11 @@ RUN_OBJS = \
 	$(OBJDIR)/register.o		$(OBJDIR)/break.o \
 	$(OBJDIR)/erl_async.o		$(OBJDIR)/erl_lock_check.o \
 	$(OBJDIR)/erl_gc.o 		$(OBJDIR)/erl_lock_count.o \
-	$(OBJDIR)/erl_posix_str.o \
+	$(OBJDIR)/erl_posix_str.o	 \
 	$(OBJDIR)/erl_bits.o 		$(OBJDIR)/erl_math.o \
 	$(OBJDIR)/erl_fun.o             $(OBJDIR)/erl_bif_port.o \
 	$(OBJDIR)/erl_term.o 		$(OBJDIR)/erl_node_tables.o \
-	$(OBJDIR)/erl_monitors.o	$(OBJDIR)/erl_process_dump.o \
+	$(OBJDIR)/erl_monitor_link.o	$(OBJDIR)/erl_process_dump.o \
 	$(OBJDIR)/erl_hl_timer.o	$(OBJDIR)/erl_cpu_topology.o \
 	$(OBJDIR)/erl_drv_thread.o      $(OBJDIR)/erl_bif_chksum.o \
 	$(OBJDIR)/erl_bif_re.o		$(OBJDIR)/erl_unicode.o \
@@ -821,17 +879,17 @@ RUN_OBJS = \
 LTTNG_OBJS = $(OBJDIR)/erlang_lttng.o
 NIF_OBJS = \
 	   $(OBJDIR)/erl_tracer_nif.o \
+	   $(OBJDIR)/prim_buffer_nif.o \
+	   $(OBJDIR)/prim_file_nif.o \
 	   $(OBJDIR)/zlib_nif.o
 
 ifeq ($(TARGET),win32)
 DRV_OBJS = \
 	$(OBJDIR)/registry_drv.o \
-	$(OBJDIR)/efile_drv.o \
 	$(OBJDIR)/inet_drv.o \
 	$(OBJDIR)/ram_file_drv.o \
 	$(OBJDIR)/ttsl_drv.o
 OS_OBJS = \
-	$(OBJDIR)/win_efile.o \
 	$(OBJDIR)/win_con.o \
 	$(OBJDIR)/dll_sys.o \
 	$(OBJDIR)/driver_tab.o \
@@ -840,22 +898,23 @@ OS_OBJS = \
 	$(OBJDIR)/sys_time.o \
 	$(OBJDIR)/sys_interrupt.o \
 	$(OBJDIR)/sys_env.o \
-	$(OBJDIR)/dosmap.o
+	$(OBJDIR)/dosmap.o \
+	$(OBJDIR)/win_prim_file.o
 
 else
 OS_OBJS = \
 	$(OBJDIR)/sys.o \
 	$(OBJDIR)/sys_drivers.o \
+	$(OBJDIR)/sys_env.o \
 	$(OBJDIR)/sys_uds.o \
 	$(OBJDIR)/driver_tab.o \
-	$(OBJDIR)/unix_efile.o \
+	$(OBJDIR)/elib_memmove.o \
 	$(OBJDIR)/gzio.o \
-	$(OBJDIR)/elib_memmove.o
+	$(OBJDIR)/unix_prim_file.o
 
 	OS_OBJS += $(OBJDIR)/sys_float.o \
                    $(OBJDIR)/sys_time.o
 DRV_OBJS = \
-	$(OBJDIR)/efile_drv.o \
 	$(OBJDIR)/inet_drv.o \
 	$(OBJDIR)/ram_file_drv.o \
 	$(OBJDIR)/ttsl_drv.o
@@ -885,23 +944,20 @@ $(STATIC_NIF_LIBS) $(STATIC_DRIVER_LIBS):
 	echo "=== Leaving lib after making static libs"
 endif
 
-ifeq ($(ERTS_ENABLE_KERNEL_POLL),yes)
-OS_OBJS += 	$(OBJDIR)/erl_poll.kp.o \
-		$(OBJDIR)/erl_check_io.kp.o \
-		$(OBJDIR)/erl_poll.nkp.o \
-		$(OBJDIR)/erl_check_io.nkp.o
-else
 OS_OBJS +=	$(OBJDIR)/erl_poll.o \
-		$(OBJDIR)/erl_check_io.o
-endif
-
-OS_OBJS +=	$(OBJDIR)/erl_mseg.o \
-		$(OBJDIR)/erl_mmap.o \
+		$(OBJDIR)/erl_check_io.o \
+		$(OBJDIR)/erl_mseg.o \
+		$(OBJDIR)/erl_mmap.o  \
+		$(OBJDIR)/erl_osenv.o \
 		$(OBJDIR)/erl_$(ERLANG_OSTYPE)_sys_ddll.o \
 		$(OBJDIR)/erl_mtrace_sys_wrap.o \
 		$(OBJDIR)/erl_sys_common_misc.o \
 		$(OBJDIR)/erl_os_monotonic_time_extender.o
 
+ifeq ($(ERTS_BUILD_FALLBACK_POLL),yes)
+OS_OBJS += 	$(OBJDIR)/erl_poll.flbk.o
+endif
+
 HIPE_ARCH64_OBJS=$(OBJDIR)/hipe_bif64.o
 
 HIPE_x86_OS_OBJS=$(HIPE_x86_$(OPSYS)_OBJS)
@@ -930,21 +986,23 @@ ifdef HIPE_ENABLED
 EXTRA_BASE_OBJS += $(HIPE_OBJS)
 endif
 
-BASE_OBJS = $(EMU_OBJS) $(RUN_OBJS) $(OS_OBJS) $(EXTRA_BASE_OBJS) $(LTTNG_OBJS)
+BASE_OBJS = $(EMU_OBJS) $(RUN_OBJS) $(OS_OBJS) $(EXTRA_BASE_OBJS) \
+	$(LTTNG_OBJS) $(DRV_OBJS) $(NIF_OBJS)
 
-before_DTrace_OBJS =	$(BASE_OBJS) $(DRV_OBJS) $(NIF_OBJS)
+PROF_OBJS = $(patsubst %.o,%$(PROFILE_MARKER).o,$(PROFILE_OBJS)) $(BASE_OBJS)
+
+OBJS = $(PROF_OBJS)
 
-DTRACE_OBJS =
 ifdef DTRACE_ENABLED_2STEP
-DTRACE_OBJS = $(OBJDIR)/erlang_dtrace.o
-$(OBJDIR)/erlang_dtrace.o: $(before_DTrace_OBJS) $(TARGET)/erlang_dtrace.h
+# The $(PROFILE_MARKER) is placed in the object file name in order to
+# make sure we re-compile with the new object files for the profiled emulator
+OBJS += $(OBJDIR)/erlang$(PROFILE_MARKER)_dtrace.o
+$(OBJDIR)/erlang$(PROFILE_MARKER)_dtrace.o: $(PROF_OBJS) $(TARGET)/erlang_dtrace.h
 	dtrace -G -C -Ibeam \
 	  -s beam/erlang_dtrace.d \
-	  -o $@ $(before_DTrace_OBJS)
+	  -o $@ $(PROF_OBJS)
 endif
 
-OBJS = $(before_DTrace_OBJS) $(DTRACE_OBJS)
-
 $(INIT_OBJS): $(TTF_DIR)/GENERATED
 $(OBJS): $(TTF_DIR)/GENERATED
 
@@ -1036,8 +1094,8 @@ $(BINDIR)/$(EMULATOR_EXECUTABLE): $(INIT_OBJS) $(OBJS) $(DEPLIBS)
 
 else
 $(BINDIR)/$(EMULATOR_EXECUTABLE): $(INIT_OBJS) $(OBJS) $(DEPLIBS)
-	$(ld_verbose)$(PURIFY) $(LD) -o $(BINDIR)/$(EMULATOR_EXECUTABLE) \
-	$(HIPEBEAMLDFLAGS) $(LDFLAGS) $(DEXPORT) $(INIT_OBJS) $(OBJS) \
+	$(ld_verbose)$(PURIFY) $(LD) -o $@ \
+	$(HIPEBEAMLDFLAGS) $(PROFILE_LDFLAGS) $(LDFLAGS) $(DEXPORT) $(INIT_OBJS) $(OBJS) \
 	$(STATIC_NIF_LIBS) $(STATIC_DRIVER_LIBS) $(LIBS)
 
 endif
@@ -1056,8 +1114,7 @@ SED_REPL_O=s|^\([^:]*:\)|$$(OBJDIR)/\1|g
 SED_REPL_O_ZLIB=s|^\([^:]*:\)|$$(ZLIB_OBJDIR)/\1|g
 SED_REPL_TTF_DIR=s|$(TTF_DIR)/|$$(TTF_DIR)/|g
 SED_REPL_ERL_TOP=s|\([ 	]\)$(ERL_TOP)/|\1$$(ERL_TOP)/|g;s|^$(ERL_TOP)/|$$(ERL_TOP)/|g
-SED_REPL_POLL=s|$$(OBJDIR)/erl_poll.o|$$(OBJDIR)/erl_poll.kp.o $$(OBJDIR)/erl_poll.nkp.o|g
-SED_REPL_CHK_IO=s|$$(OBJDIR)/erl_check_io.o|$$(OBJDIR)/erl_check_io.kp.o $$(OBJDIR)/erl_check_io.nkp.o|g
+SED_REPL_POLL=s|$$(OBJDIR)/erl_poll.o|$$(OBJDIR)/erl_poll.o $$(OBJDIR)/erl_poll.flbk.o|g
 SED_REPL_TTF_COMP_FLAGS=s|\([^/]\)erl_compile_flags\.h|\1$$(TTF_DIR)/erl_compile_flags\.h|g
 
 ifeq ($(TARGET),win32)
@@ -1067,8 +1124,8 @@ else
 SED_PREFIX=
 endif
 
-ifeq ($(ERTS_ENABLE_KERNEL_POLL),yes)
-SED_SUFFIX=;$(SED_REPL_POLL);$(SED_REPL_CHK_IO)
+ifeq ($(ERTS_BUILD_FALLBACK_POLL),yes)
+SED_SUFFIX=;$(SED_REPL_POLL)
 else
 SED_SUFFIX=
 endif
@@ -1086,6 +1143,7 @@ BEAM_SRC=$(wildcard beam/*.c)
 DRV_COMMON_SRC=$(wildcard drivers/common/*.c)
 DRV_OSTYPE_SRC=$(wildcard drivers/$(ERLANG_OSTYPE)/*.c)
 NIF_COMMON_SRC=$(wildcard nifs/common/*.c)
+NIF_OSTYPE_SRC=$(wildcard nifs/$(ERLANG_OSTYPE)/*.c)
 ALL_SYS_SRC=$(wildcard sys/$(ERLANG_OSTYPE)/*.c) $(wildcard sys/common/*.c)
 # We use $(shell ls) here instead of wildcard as $(wildcard ) resolved at
 # loadtime of the makefile and at that time these files are not generated yet.
@@ -1098,7 +1156,10 @@ ifeq ($(TARGET),win32)
 
 #DEP_CC=$(EMU_CC)
 DEP_CC=$(CC)
-DEP_FLAGS=-MM  $(subst -O2,,$(CFLAGS)) $(INCLUDES) -I../etc/win32 -Idrivers/common -Idrivers/$(ERLANG_OSTYPE)
+DEP_FLAGS=-MM  $(subst -O2,,$(CFLAGS)) $(INCLUDES) -I../etc/win32 \
+	-Idrivers/common -Idrivers/$(ERLANG_OSTYPE) \
+	-Inifs/common -Inifs/$(ERLANG_OSTYPE)
+
 # ifeq (@MIXED_CYGWIN_VC@,yes)
 # VC++ used for compiling. If __GNUC__ is defined we will include
 # other headers then when compiling which will result in faulty
@@ -1118,10 +1179,16 @@ MG_FLAG=-MG
 endif
 
 DEP_CC=$(CC)
-DEP_FLAGS=-MM $(MG_FLAG) $(CFLAGS) $(INCLUDES) -Inifs/common -Idrivers/common -Idrivers/$(ERLANG_OSTYPE)
+DEP_FLAGS=-MM $(MG_FLAG) $(CFLAGS) $(INCLUDES) \
+	-Idrivers/common -Idrivers/$(ERLANG_OSTYPE) \
+	-Inifs/common -Inifs/$(ERLANG_OSTYPE)
 SYS_SRC=$(ALL_SYS_SRC)
 endif
 
+.PHONY: check_emu_registers
+check_emu_registers: $(OBJDIR)/beam_emu$(PROFILE_MARKER).S
+	utils/beam_emu_vars -vars 'c_p E HTOP FCALLS I reg' $^
+
 .PHONY: $(TARGET)/gen_git_version.mk
 $(TARGET)/gen_git_version.mk:
 # We touch beam/erl_bif.info.c if we regenerated the git version to force a
@@ -1145,6 +1212,8 @@ $(TTF_DIR)/depend.mk: $(TTF_DIR)/GENERATED $(PRELOAD_SRC)
 		| $(SED_DEPEND) >> $(TTF_DIR)/depend.mk
 	$(V_at)$(DEP_CC) $(DEP_FLAGS) $(NIF_COMMON_SRC) \
 		| $(SED_DEPEND) >> $(TTF_DIR)/depend.mk
+	$(V_at)$(DEP_CC) $(DEP_FLAGS) -I../etc/$(ERLANG_OSTYPE) $(NIF_OSTYPE_SRC) \
+		| $(SED_DEPEND) >> $(TTF_DIR)/depend.mk
 	$(V_at)$(DEP_CC) $(DEP_FLAGS) $(SYS_SRC) \
 		| $(SED_DEPEND) >> $(TTF_DIR)/depend.mk
 	$(V_at)$(DEP_CC) $(DEP_FLAGS) $(TARGET_SRC) \
diff --git a/erts/emulator/beam/arith_instrs.tab b/erts/emulator/beam/arith_instrs.tab
new file mode 100644
index 0000000000..b828e86788
--- /dev/null
+++ b/erts/emulator/beam/arith_instrs.tab
@@ -0,0 +1,396 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+OUTLINED_ARITH_2(Fail, Live, Name, BIF, Op1, Op2, Dst) {
+    Eterm result;
+    Uint live = $Live;
+    HEAVY_SWAPOUT;
+    reg[live] = $Op1;
+    reg[live+1] = $Op2;
+    result = erts_gc_$Name (c_p, reg, live);
+    HEAVY_SWAPIN;
+    ERTS_HOLE_CHECK(c_p);
+    if (ERTS_LIKELY(is_value(result))) {
+        $REFRESH_GEN_DEST();
+        $Dst = result;
+        $NEXT0();
+    }
+    $BIF_ERROR_ARITY_2($Fail, $BIF, reg[live], reg[live+1]);
+}
+
+
+i_plus := plus.fetch.execute;
+
+plus.head() {
+    Eterm PlusOp1, PlusOp2;
+}
+
+plus.fetch(Op1, Op2) {
+    PlusOp1 = $Op1;
+    PlusOp2 = $Op2;
+}
+
+plus.execute(Fail, Live, Dst) {
+    if (ERTS_LIKELY(is_both_small(PlusOp1, PlusOp2))) {
+        Sint i = signed_val(PlusOp1) + signed_val(PlusOp2);
+        if (ERTS_LIKELY(IS_SSMALL(i))) {
+            $Dst = make_small(i);
+            $NEXT0();
+        }
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, mixed_plus, BIF_splus_2, PlusOp1, PlusOp2, $Dst);
+}
+
+i_minus := minus.fetch.execute;
+
+minus.head() {
+    Eterm MinusOp1, MinusOp2;
+}
+
+minus.fetch(Op1, Op2) {
+    MinusOp1 = $Op1;
+    MinusOp2 = $Op2;
+}
+
+minus.execute(Fail, Live, Dst) {
+    if (ERTS_LIKELY(is_both_small(MinusOp1, MinusOp2))) {
+        Sint i = signed_val(MinusOp1) - signed_val(MinusOp2);
+        if (ERTS_LIKELY(IS_SSMALL(i))) {
+            $Dst = make_small(i);
+            $NEXT0();
+        }
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, mixed_minus, BIF_sminus_2, MinusOp1, MinusOp2, $Dst);
+}
+
+i_increment := increment.fetch.execute;
+
+increment.head() {
+    Eterm increment_reg_val;
+}
+
+increment.fetch(Src) {
+    increment_reg_val = $Src;
+}
+
+increment.execute(IncrementVal, Live, Dst) {
+    Eterm increment_val = $IncrementVal;
+    Uint live;
+    Eterm result;
+
+    if (ERTS_LIKELY(is_small(increment_reg_val))) {
+        Sint i = signed_val(increment_reg_val) + increment_val;
+        if (ERTS_LIKELY(IS_SSMALL(i))) {
+            $Dst = make_small(i);
+            $NEXT0();
+        }
+    }
+    live = $Live;
+    HEAVY_SWAPOUT;
+    reg[live] = increment_reg_val;
+    reg[live+1] = make_small(increment_val);
+    result = erts_gc_mixed_plus(c_p, reg, live);
+    HEAVY_SWAPIN;
+    ERTS_HOLE_CHECK(c_p);
+    if (ERTS_LIKELY(is_value(result))) {
+        $REFRESH_GEN_DEST();
+        $Dst = result;
+        $NEXT0();
+    }
+    ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
+    goto find_func_info;
+}
+
+i_times(Fail, Live, Op1, Op2, Dst) {
+    Eterm op1 = $Op1;
+    Eterm op2 = $Op2;
+    $OUTLINED_ARITH_2($Fail, $Live, mixed_times, BIF_stimes_2, op1, op2, $Dst);
+}
+
+i_m_div(Fail, Live, Op1, Op2, Dst) {
+    Eterm op1 = $Op1;
+    Eterm op2 = $Op2;
+    $OUTLINED_ARITH_2($Fail, $Live, mixed_div, BIF_div_2, op1, op2, $Dst);
+}
+
+i_int_div(Fail, Live, Op1, Op2, Dst) {
+    Eterm op1 = $Op1;
+    Eterm op2 = $Op2;
+    if (ERTS_UNLIKELY(op2 == SMALL_ZERO)) {
+        c_p->freason = BADARITH;
+        $BIF_ERROR_ARITY_2($Fail, BIF_intdiv_2, op1, op2);
+    } else if (ERTS_LIKELY(is_both_small(op1, op2))) {
+        Sint ires = signed_val(op1) / signed_val(op2);
+        if (ERTS_LIKELY(IS_SSMALL(ires))) {
+            $Dst = make_small(ires);
+            $NEXT0();
+        }
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, int_div, BIF_intdiv_2, op1, op2, $Dst);
+}
+
+i_rem := rem.fetch.execute;
+
+rem.head() {
+     Eterm RemOp1, RemOp2;
+}
+
+rem.fetch(Src1, Src2) {
+    RemOp1 = $Src1;
+    RemOp2 = $Src2;
+}
+
+rem.execute(Fail, Live, Dst) {
+    if (ERTS_UNLIKELY(RemOp2 == SMALL_ZERO)) {
+        c_p->freason = BADARITH;
+        $BIF_ERROR_ARITY_2($Fail, BIF_rem_2, RemOp1, RemOp2);
+    } else if (ERTS_LIKELY(is_both_small(RemOp1, RemOp2))) {
+        $Dst = make_small(signed_val(RemOp1) % signed_val(RemOp2));
+        $NEXT0();
+    } else {
+        $OUTLINED_ARITH_2($Fail, $Live, int_rem, BIF_rem_2, RemOp1, RemOp2, $Dst);
+    }
+}
+
+i_band := band.fetch.execute;
+
+band.head() {
+    Eterm BandOp1, BandOp2;
+}
+
+band.fetch(Src1, Src2) {
+    BandOp1 = $Src1;
+    BandOp2 = $Src2;
+}
+
+band.execute(Fail, Live, Dst) {
+    if (ERTS_LIKELY(is_both_small(BandOp1, BandOp2))) {
+        /*
+         * No need to untag -- TAG & TAG == TAG.
+         */
+        $Dst = BandOp1 & BandOp2;
+        $NEXT0();
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, band, BIF_band_2, BandOp1, BandOp2, $Dst);
+}
+
+i_bor(Fail, Live, Src1, Src2, Dst) {
+    if (ERTS_LIKELY(is_both_small($Src1, $Src2))) {
+        /*
+         * No need to untag -- TAG | TAG == TAG.
+         */
+        $Dst = $Src1 | $Src2;
+        $NEXT0();
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, bor, BIF_bor_2, $Src1, $Src2, $Dst);
+}
+
+i_bxor(Fail, Live, Src1, Src2, Dst) {
+    if (ERTS_LIKELY(is_both_small($Src1, $Src2))) {
+        /*
+         * TAG ^ TAG == 0.
+         *
+         * Therefore, we perform the XOR operation on the tagged values,
+         * and OR in the tag bits.
+         */
+        $Dst = ($Src1 ^ $Src2) | make_small(0);
+        $NEXT0();
+    }
+    $OUTLINED_ARITH_2($Fail, $Live, bxor, BIF_bxor_2, $Src1, $Src2, $Dst);
+}
+
+i_bsl := shift.setup_bsl.execute;
+i_bsr := shift.setup_bsr.execute;
+
+shift.head() {
+    Eterm Op1, Op2;
+    Sint shift_left_count;
+}
+
+shift.setup_bsr(Src1, Src2) {
+    Op1 = $Src1;
+    Op2 = $Src2;
+    shift_left_count = 0;
+    if (ERTS_LIKELY(is_small(Op2))) {
+        shift_left_count = -signed_val(Op2);
+    } else if (is_big(Op2)) {
+        /*
+         * N bsr NegativeBigNum == N bsl MAX_SMALL
+         * N bsr PositiveBigNum == N bsl MIN_SMALL
+         */
+        shift_left_count = make_small(bignum_header_is_neg(*big_val(Op2)) ?
+                                      MAX_SMALL : MIN_SMALL);
+    }
+}
+
+shift.setup_bsl(Src1, Src2) {
+    Op1 = $Src1;
+    Op2 = $Src2;
+    shift_left_count = 0;
+    if (ERTS_LIKELY(is_small(Op2))) {
+        shift_left_count = signed_val(Op2);
+    } else if (is_big(Op2)) {
+        if (bignum_header_is_neg(*big_val(Op2))) {
+            /*
+             * N bsl NegativeBigNum is either 0 or -1, depending on
+             * the sign of N. Since we don't believe this case
+             * is common, do the calculation with the minimum
+             * amount of code.
+             */
+            shift_left_count = MIN_SMALL;
+        } else if (is_integer(Op1)) {
+            /*
+             * N bsl PositiveBigNum is too large to represent.
+             */
+            shift_left_count = MAX_SMALL;
+        }
+    }
+}
+
+shift.execute(Fail, Live, Dst) {
+    Uint big_words_needed;
+
+    if (ERTS_LIKELY(is_small(Op1))) {
+        Sint int_res = signed_val(Op1);
+        if (ERTS_UNLIKELY(shift_left_count == 0 || int_res == 0)) {
+            if (ERTS_UNLIKELY(is_not_integer(Op2))) {
+                goto shift_error;
+            }
+            if (int_res == 0) {
+                $Dst = Op1;
+                $NEXT0();
+            }
+        } else if (shift_left_count < 0)  { /* Right shift */
+            Eterm bsr_res;
+            shift_left_count = -shift_left_count;
+            if (shift_left_count >= SMALL_BITS-1) {
+                bsr_res = (int_res < 0) ? SMALL_MINUS_ONE : SMALL_ZERO;
+            } else {
+                bsr_res = make_small(int_res >> shift_left_count);
+            }
+            $Dst = bsr_res;
+            $NEXT0();
+        } else if (shift_left_count < SMALL_BITS-1) { /* Left shift */
+            if ((int_res > 0 &&
+                 ((~(Uint)0 << ((SMALL_BITS-1)-shift_left_count)) & int_res) == 0) ||
+                ((~(Uint)0 << ((SMALL_BITS-1)-shift_left_count)) & ~int_res) == 0) {
+                $Dst = make_small(int_res << shift_left_count);
+                $NEXT0();
+            }
+        }
+        big_words_needed = 1;   /* big_size(small_to_big(Op1)) */
+        goto big_shift;
+    } else if (is_big(Op1)) {
+        if (shift_left_count == 0) {
+            if (is_not_integer(Op2)) {
+                goto shift_error;
+            }
+            $Dst = Op1;
+            $NEXT0();
+        }
+        big_words_needed = big_size(Op1);
+
+    big_shift:
+        if (shift_left_count > 0) {	/* Left shift. */
+            big_words_needed += (shift_left_count / D_EXP);
+        } else {	/* Right shift. */
+            if (big_words_needed <= (-shift_left_count / D_EXP)) {
+                big_words_needed = 3;       /* ??? */
+            } else {
+                big_words_needed -= (-shift_left_count / D_EXP);
+            }
+        }
+        {
+            Eterm tmp_big[2];
+            Sint big_need_size = BIG_NEED_SIZE(big_words_needed+1);
+
+            /*
+             * Slightly conservative check the size to avoid
+             * allocating huge amounts of memory for bignums that
+             * clearly would overflow the arity in the header
+             * word.
+             */
+            if (big_need_size-8 > BIG_ARITY_MAX) {
+                $SYSTEM_LIMIT($Fail);
+            }
+            $GC_TEST_PRESERVE(big_need_size+1, $Live, Op1);
+            if (is_small(Op1)) {
+                Op1 = small_to_big(signed_val(Op1), tmp_big);
+            }
+            Op1 = big_lshift(Op1, shift_left_count, HTOP);
+            if (is_big(Op1)) {
+                HTOP += bignum_header_arity(*HTOP) + 1;
+            }
+            HEAP_SPACE_VERIFIED(0);
+            if (ERTS_UNLIKELY(is_nil(Op1))) {
+                /*
+                 * This result must have been only slighty larger
+                 * than allowed since it wasn't caught by the
+                 * previous test.
+                 */
+                $SYSTEM_LIMIT($Fail);
+            }
+            ERTS_HOLE_CHECK(c_p);
+            $REFRESH_GEN_DEST();
+            $Dst = Op1;
+            $NEXT0();
+        }
+    }
+
+    /*
+     * One or more non-integer arguments.
+     */
+ shift_error:
+    c_p->freason = BADARITH;
+    if ($Fail) {
+        $FAIL($Fail);
+    } else {
+        reg[0] = Op1;
+        reg[1] = Op2;
+        SWAPOUT;
+        if (IsOpCode(I[0], i_bsl_ssjtd)) {
+            I = handle_error(c_p, I, reg, &bif_export[BIF_bsl_2]->info.mfa);
+        } else {
+            ASSERT(IsOpCode(I[0], i_bsr_ssjtd));
+            I = handle_error(c_p, I, reg, &bif_export[BIF_bsr_2]->info.mfa);
+        }
+        goto post_error_handling;
+    }
+}
+
+i_int_bnot(Fail, Src, Live, Dst) {
+    Eterm bnot_val = $Src;
+    if (ERTS_LIKELY(is_small(bnot_val))) {
+        bnot_val = make_small(~signed_val(bnot_val));
+    } else {
+        Uint live = $Live;
+        HEAVY_SWAPOUT;
+        reg[live] = bnot_val;
+        bnot_val = erts_gc_bnot(c_p, reg, live);
+        HEAVY_SWAPIN;
+        ERTS_HOLE_CHECK(c_p);
+        if (ERTS_UNLIKELY(is_nil(bnot_val))) {
+            $BIF_ERROR_ARITY_1($Fail, BIF_bnot_1, reg[live]);
+        }
+        $REFRESH_GEN_DEST();
+    }
+    $Dst = bnot_val;
+}
diff --git a/erts/emulator/beam/atom.c b/erts/emulator/beam/atom.c
index 38e02c386f..e5b7616a0d 100644
--- a/erts/emulator/beam/atom.c
+++ b/erts/emulator/beam/atom.c
@@ -34,20 +34,18 @@
 
 IndexTable erts_atom_table;	/* The index table */
 
-#include "erl_smp.h"
+static erts_rwmtx_t atom_table_lock;
 
-static erts_smp_rwmtx_t atom_table_lock;
-
-#define atom_read_lock()	erts_smp_rwmtx_rlock(&atom_table_lock)
-#define atom_read_unlock()	erts_smp_rwmtx_runlock(&atom_table_lock)
-#define atom_write_lock()	erts_smp_rwmtx_rwlock(&atom_table_lock)
-#define atom_write_unlock()	erts_smp_rwmtx_rwunlock(&atom_table_lock)
+#define atom_read_lock()	erts_rwmtx_rlock(&atom_table_lock)
+#define atom_read_unlock()	erts_rwmtx_runlock(&atom_table_lock)
+#define atom_write_lock()	erts_rwmtx_rwlock(&atom_table_lock)
+#define atom_write_unlock()	erts_rwmtx_rwunlock(&atom_table_lock)
 
 #if 0
 #define ERTS_ATOM_PUT_OPS_STAT
 #endif
 #ifdef ERTS_ATOM_PUT_OPS_STAT
-static erts_smp_atomic_t atom_put_ops;
+static erts_atomic_t atom_put_ops;
 #endif
 
 /* Functions for allocating space for the ext of atoms. We do not
@@ -76,7 +74,7 @@ void atom_info(fmtfn_t to, void *to_arg)
     index_info(to, to_arg, &erts_atom_table);
 #ifdef ERTS_ATOM_PUT_OPS_STAT
     erts_print(to, to_arg, "atom_put_ops: %ld\n",
-	       erts_smp_atomic_read_nob(&atom_put_ops));
+	       erts_atomic_read_nob(&atom_put_ops));
 #endif
 
     if (lock)
@@ -246,7 +244,7 @@ erts_atom_put_index(const byte *name, int len, ErtsAtomEncoding enc, int trunc)
     int aix;
 
 #ifdef ERTS_ATOM_PUT_OPS_STAT
-    erts_smp_atomic_inc_nob(&atom_put_ops);
+    erts_atomic_inc_nob(&atom_put_ops);
 #endif
 
     if (tlen < 0) {
@@ -359,32 +357,24 @@ am_atom_put(const char* name, int len)
 int atom_table_size(void)
 {
     int ret;
-#ifdef ERTS_SMP
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
 	atom_read_lock();
-#endif
     ret = erts_atom_table.entries;
-#ifdef ERTS_SMP
     if (lock)
 	atom_read_unlock();
-#endif
     return ret;
 }
 
 int atom_table_sz(void)
 {
     int ret;
-#ifdef ERTS_SMP
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
 	atom_read_lock();
-#endif
     ret = index_table_sz(&erts_atom_table);
-#ifdef ERTS_SMP
     if (lock)
 	atom_read_unlock();
-#endif
     return ret;
 }
 
@@ -412,19 +402,15 @@ erts_atom_get(const char *name, int len, Eterm* ap, ErtsAtomEncoding enc)
 void
 erts_atom_get_text_space_sizes(Uint *reserved, Uint *used)
 {
-#ifdef ERTS_SMP
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
 	atom_read_lock();
-#endif
     if (reserved)
 	*reserved = reserved_atom_space;
     if (used)
 	*used = atom_space;
-#ifdef ERTS_SMP
     if (lock)
 	atom_read_unlock();
-#endif
 }
 
 void
@@ -433,16 +419,16 @@ init_atom_table(void)
     HashFunctions f;
     int i;
     Atom a;
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
 
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
 #ifdef ERTS_ATOM_PUT_OPS_STAT
-    erts_smp_atomic_init_nob(&atom_put_ops, 0);
+    erts_atomic_init_nob(&atom_put_ops, 0);
 #endif
 
-    erts_smp_rwmtx_init_opt(&atom_table_lock, &rwmtx_opt, "atom_tab", NIL,
+    erts_rwmtx_init_opt(&atom_table_lock, &rwmtx_opt, "atom_tab", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
     f.hash = (H_FUN) atom_hash;
@@ -466,7 +452,7 @@ init_atom_table(void)
     /* Ordinary atoms */
     for (i = 0; erl_atom_names[i] != 0; i++) {
 	int ix;
-	a.len = strlen(erl_atom_names[i]);
+	a.len = sys_strlen(erl_atom_names[i]);
 	a.latin1_chars = a.len;
 	a.name = (byte*)erl_atom_names[i];
 	a.slot.index = i;
@@ -505,4 +491,4 @@ Uint
 erts_get_atom_limit(void)
 {
     return erts_atom_table.limit;
-}
\ No newline at end of file
+}
diff --git a/erts/emulator/beam/atom.names b/erts/emulator/beam/atom.names
index 7b12e5432d..45b7540aeb 100644
--- a/erts/emulator/beam/atom.names
+++ b/erts/emulator/beam/atom.names
@@ -1,7 +1,7 @@
 #
 # %CopyrightBegin%
 #
-# Copyright Ericsson AB 1996-2017. All Rights Reserved.
+# Copyright Ericsson AB 1996-2018. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -48,7 +48,7 @@ atom Empty=''
 #
 # Used in the Beam emulator loop. (Smaller literals usually means tighter code.)
 #
-atom fun infinity timeout normal call return
+atom infinity timeout normal call return
 atom throw error exit
 atom undefined
 
@@ -69,11 +69,9 @@ atom DOWN='DOWN'
 atom UP='UP'
 atom EXIT='EXIT'
 atom abort
-atom aborted
 atom abs_path
 atom absoluteURI
 atom ac
-atom accessor
 atom active
 atom active_tasks
 atom active_tasks_all
@@ -88,7 +86,6 @@ atom allocated_areas
 atom allocator
 atom allocator_sizes
 atom alloc_util_allocators
-atom allow_gc
 atom allow_passive_connect
 atom already_loaded
 atom amd64
@@ -108,6 +105,8 @@ atom asynchronous
 atom atom
 atom atom_used
 atom attributes
+atom auto_connect
+atom await_exit
 atom await_microstate_accounting_modifications
 atom await_port_send_result
 atom await_proc_exit
@@ -121,9 +120,7 @@ atom bag
 atom band
 atom big
 atom bif_return_trap
-atom bif_timer_server
 atom binary
-atom binary_bin_to_list_trap
 atom binary_copy_trap
 atom binary_find_trap
 atom binary_longest_prefix_trap
@@ -145,6 +142,7 @@ atom bsr
 atom bsr_anycrlf
 atom bsr_unicode
 atom build_type
+atom busy
 atom busy_dist_port
 atom busy_port
 atom call
@@ -178,7 +176,6 @@ atom convert_time_unit
 atom connect
 atom connected
 atom connection_closed
-atom cons
 atom const
 atom context_switches
 atom control
@@ -198,9 +195,6 @@ atom debug_flags
 atom decimals
 atom default
 atom delay_trap
-atom dexit
-atom depth
-atom dgroup_leader
 atom dictionary
 atom dirty_bif_exception
 atom dirty_bif_result
@@ -217,20 +211,18 @@ atom discard
 atom display_items
 atom dist
 atom dist_cmd
+atom dist_ctrl_put_data
+atom dist_data
 atom Div='/'
 atom div
-atom dlink
 atom dmonitor_node
-atom dmonitor_p
 atom DollarDollar='$$'
 atom DollarUnderscore='$_'
 atom dollar_endonly
 atom dotall
 atom driver
 atom driver_options
-atom dsend
 atom dsend_continue_trap
-atom dunlink
 atom duplicate_bag
 atom duplicated
 atom dupnames
@@ -248,20 +240,20 @@ atom Eqeq='=='
 atom erl_tracer
 atom erlang
 atom erl_signal_server
-atom ERROR='ERROR'
 atom error_handler
 atom error_logger
 atom erts_code_purger
 atom erts_debug
+atom erts_dflags
 atom erts_internal
 atom ets
 atom ETS_TRANSFER='ETS-TRANSFER'
-atom event
 atom exact_reductions
 atom exception_from
 atom exception_trace
 atom exclusive
 atom exit_status
+atom exited
 atom existing
 atom existing_processes
 atom existing_ports
@@ -283,27 +275,20 @@ atom force
 atom format_cpu_topology
 atom free
 atom fullsweep_after
-atom fullsweep_if_old_binaries
-atom fun
-atom function
 atom functions
 atom function_clause
 atom garbage_collecting
 atom garbage_collection
 atom garbage_collection_info
-atom gc_end
 atom gc_major_end
 atom gc_major_start
 atom gc_max_heap_size
 atom gc_minor_end
 atom gc_minor_start
-atom gc_start
 atom Ge='>='
 atom generational
-atom get_data
 atom get_seq_token
 atom get_tcw
-atom getenv
 atom gather_gc_info_result
 atom gather_io_bytes
 atom gather_microstate_accounting_result
@@ -345,11 +330,11 @@ atom initial_call
 atom input
 atom internal
 atom internal_error
-atom internal_status
 atom instruction_counts
 atom invalid
 atom is_constant
 atom is_seq_trace
+atom iterator
 atom io
 atom keypos
 atom kill
@@ -375,6 +360,7 @@ atom loaded
 atom load_cancelled
 atom load_failure
 atom local
+atom logger
 atom long_gc
 atom long_schedule
 atom low
@@ -390,14 +376,12 @@ atom match_spec_result
 atom max
 atom maximum
 atom max_heap_size
-atom max_tables max_processes
 atom mbuf_size
 atom md5
 atom memory
 atom memory_internal
 atom memory_types
 atom message
-atom message_binary
 atom message_queue_data
 atom message_queue_len
 atom messages
@@ -444,21 +428,18 @@ atom new_processes
 atom new_ports
 atom new_uniq
 atom newline
-atom next
 atom no
 atom nomatch
 atom none
 atom no_auto_capture
 atom noconnect
 atom noconnection
-atom nocookie
 atom node
 atom node_type
 atom nodedown
 atom nodedown_reason
 atom nodeup
 atom noeol
-atom nofile
 atom noproc
 atom normal
 atom nosuspend
@@ -466,6 +447,7 @@ atom no_float
 atom no_integer
 atom no_network
 atom no_start_optimize
+atom not_suspended
 atom not
 atom not_a_list
 atom not_loaded
@@ -480,7 +462,6 @@ atom notempty_atstart
 atom notify
 atom notsup
 atom nouse_stdio
-atom objects
 atom off_heap
 atom offset
 atom ok
@@ -550,7 +531,6 @@ atom re_run_trap
 atom read_concurrency
 atom ready_input
 atom ready_output
-atom ready_async
 atom reason
 atom receive
 atom recent_size
@@ -597,7 +577,6 @@ atom sequential_trace_token
 atom serial
 atom set
 atom set_cpu_topology
-atom set_data
 atom set_on_first_link
 atom set_on_first_spawn
 atom set_on_link
@@ -605,8 +584,6 @@ atom set_on_spawn
 atom set_seq_token
 atom set_tcw
 atom set_tcw_fake
-atom separate
-atom shared
 atom sighup
 atom sigterm
 atom sigusr1
@@ -622,7 +599,6 @@ atom sigtstp
 atom sigquit
 atom silent
 atom size
-atom sl_alloc
 atom spawn_executable
 atom spawn_driver
 atom spawned
@@ -630,7 +606,6 @@ atom ssl_tls
 atom stack_size
 atom start
 atom status
-atom static
 atom stderr_to_stdout
 atom stop
 atom stream
@@ -640,13 +615,11 @@ atom sunrm
 atom suspend
 atom suspended
 atom suspending
-atom sys_misc
 atom system
-atom system_error
+atom system_flag_scheduler_wall_time
 atom system_limit
 atom system_version
 atom system_architecture
-atom SYSTEM='SYSTEM'
 atom table
 atom term_to_binary_trap
 atom this
@@ -664,7 +637,7 @@ atom total_heap_size
 atom total_run_queue_lengths
 atom total_run_queue_lengths_all
 atom tpkt
-atom trace trace_ts traced
+atom trace traced
 atom trace_control_word
 atom trace_status
 atom tracer
@@ -673,7 +646,6 @@ atom trim
 atom trim_all
 atom try_clause
 atom true
-atom tuple
 atom type
 atom ucompile
 atom ucp
@@ -690,14 +662,11 @@ atom unblock_normal
 atom uniq
 atom unless_suspending
 atom unloaded
-atom unloading
 atom unloaded_only
 atom unload_cancelled
 atom value
-atom values
 atom version
 atom visible
-atom wait
 atom waiting
 atom wall_clock
 atom warning
@@ -708,3 +677,4 @@ atom xor
 atom x86
 atom yes
 atom yield
+atom nifs
diff --git a/erts/emulator/beam/beam_bif_load.c b/erts/emulator/beam/beam_bif_load.c
index 023ee3ef4b..a0dbd9ec7b 100644
--- a/erts/emulator/beam/beam_bif_load.c
+++ b/erts/emulator/beam/beam_bif_load.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1999-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1999-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -37,6 +37,7 @@
 #include "erl_bits.h"
 #include "erl_thr_progress.h"
 #include "erl_nfunc_sched.h"
+#include "erl_proc_sig_queue.h"
 #ifdef HIPE
 #  include "hipe_bif0.h"
 #  define IF_HIPE(X) (X)
@@ -50,7 +51,7 @@
 
 static struct {
     Eterm module;
-    erts_smp_mtx_t mtx;
+    erts_mtx_t mtx;
     Export *pending_purge_lambda;
     Eterm *sprocs;
     Eterm def_sprocs[10];
@@ -65,12 +66,9 @@ static struct {
 
 Process *erts_code_purger = NULL;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-Process *erts_dirty_process_code_checker;
-#endif
-erts_smp_atomic_t erts_copy_literal_area__;
+erts_atomic_t erts_copy_literal_area__;
 #define ERTS_SET_COPY_LITERAL_AREA(LA)			\
-    erts_smp_atomic_set_nob(&erts_copy_literal_area__,	\
+    erts_atomic_set_nob(&erts_copy_literal_area__,	\
 			    (erts_aint_t) (LA))
 Process *erts_literal_area_collector = NULL;
 
@@ -81,7 +79,7 @@ struct ErtsLiteralAreaRef_ {
 };
 
 struct {
-    erts_smp_mtx_t mtx;
+    erts_mtx_t mtx;
     ErtsLiteralAreaRef *first;
     ErtsLiteralAreaRef *last;
 } release_literal_areas;
@@ -97,7 +95,7 @@ init_purge_state(void)
 {
     purge_state.module = THE_NON_VALUE;
 
-    erts_smp_mtx_init(&purge_state.mtx, "purge_state", NIL,
+    erts_mtx_init(&purge_state.mtx, "purge_state", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
     purge_state.pending_purge_lambda =
@@ -119,12 +117,12 @@ init_purge_state(void)
 void
 erts_beam_bif_load_init(void)
 {
-    erts_smp_mtx_init(&release_literal_areas.mtx, "release_literal_areas", NIL,
+    erts_mtx_init(&release_literal_areas.mtx, "release_literal_areas", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
     release_literal_areas.first = NULL;
     release_literal_areas.last = NULL;
-    erts_smp_atomic_init_nob(&erts_copy_literal_area__,
+    erts_atomic_init_nob(&erts_copy_literal_area__,
 			     (erts_aint_t) NULL);
 
     init_purge_state();
@@ -172,8 +170,8 @@ BIF_RETTYPE code_make_stub_module_3(BIF_ALIST_3)
 			BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
     }
 
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
+    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
 
     modp = erts_get_module(mod, erts_active_code_ix());
 
@@ -197,8 +195,8 @@ BIF_RETTYPE code_make_stub_module_3(BIF_ALIST_3)
     else {
 	erts_abort_staging_code_ix();
     }
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     erts_release_code_write_permission();
     return res;
 #endif
@@ -265,7 +263,6 @@ struct m {
 };
 
 static Eterm staging_epilogue(Process* c_p, int, Eterm res, int, struct m*, int, int);
-#ifdef ERTS_SMP
 static void smp_code_ix_commiter(void*);
 
 static struct /* Protected by code_write_permission */
@@ -273,7 +270,6 @@ static struct /* Protected by code_write_permission */
     Process* stager;
     ErtsThrPrgrLaterOp lop;
 } committer_state;
-#endif
 
 static Eterm
 exception_list(Process* p, Eterm tag, struct m* mp, Sint exceptions)
@@ -401,8 +397,8 @@ finish_loading_1(BIF_ALIST_1)
 	    erts_is_default_trace_enabled() ||
 	    IF_HIPE(hipe_need_blocking(p[i].modp))) {
 	    /* tracing or hipe need thread blocking */
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	    erts_smp_thr_progress_block();
+	    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_block();
 	    is_blocking = 1;
 	    break;
 	}
@@ -465,9 +461,7 @@ static Eterm
 staging_epilogue(Process* c_p, int commit, Eterm res, int is_blocking,
 		 struct m* mods, int nmods, int free_mods)
 {    
-#ifdef ERTS_SMP
     if (is_blocking || !commit)
-#endif
     {
 	if (commit) {
 	    int i;
@@ -491,13 +485,12 @@ staging_epilogue(Process* c_p, int commit, Eterm res, int is_blocking,
 	    erts_free(ERTS_ALC_T_LOADER_TMP, mods);
 	}
 	if (is_blocking) {
-	    erts_smp_thr_progress_unblock();
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_unblock();
+	    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 	}
 	erts_release_code_write_permission();
 	return res;
     }
-#ifdef ERTS_SMP
     else {
 	ASSERT(is_value(res));
 
@@ -522,11 +515,9 @@ staging_epilogue(Process* c_p, int commit, Eterm res, int is_blocking,
 	 */
 	ERTS_BIF_YIELD_RETURN(c_p, res);
     }
-#endif
 }
 
 
-#ifdef ERTS_SMP
 static void smp_code_ix_commiter(void* null)
 {
     Process* p = committer_state.stager;
@@ -536,14 +527,13 @@ static void smp_code_ix_commiter(void* null)
     committer_state.stager = NULL;
 #endif
     erts_release_code_write_permission();
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
     if (!ERTS_PROC_IS_EXITING(p)) {
 	erts_resume(p, ERTS_PROC_LOCK_STATUS);
     }
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     erts_proc_dec_refc(p);
 }
-#endif /* ERTS_SMP */
 
 
 
@@ -613,14 +603,14 @@ badarg:
 
 BIF_RETTYPE erts_internal_check_dirty_process_code_2(BIF_ALIST_2)
 {
-#if !defined(ERTS_DIRTY_SCHEDULERS)
-    BIF_ERROR(BIF_P, EXC_NOTSUP);
-#else
+    erts_aint32_t state;
     Process *rp;
-    int reds = 0;
+    int dirty, busy, reds = 0;
     Eterm res;
 
-    if (BIF_P != erts_dirty_process_code_checker)
+    if (BIF_P != erts_dirty_process_signal_handler
+        && BIF_P != erts_dirty_process_signal_handler_high
+        && BIF_P != erts_dirty_process_signal_handler_max)
 	BIF_ERROR(BIF_P, EXC_NOTSUP);
 
     if (is_not_internal_pid(BIF_ARG_1))
@@ -629,23 +619,31 @@ BIF_RETTYPE erts_internal_check_dirty_process_code_2(BIF_ALIST_2)
     if (is_not_atom(BIF_ARG_2))
 	BIF_ERROR(BIF_P, BADARG);
 
-    rp = erts_pid2proc_not_running(BIF_P, ERTS_PROC_LOCK_MAIN,
-				   BIF_ARG_1, ERTS_PROC_LOCK_MAIN);
-    if (rp == ERTS_PROC_LOCK_BUSY)
-	ERTS_BIF_YIELD2(bif_export[BIF_erts_internal_check_dirty_process_code_2],
-			BIF_P, BIF_ARG_1, BIF_ARG_2);
+    if (BIF_ARG_1 == BIF_P->common.id)
+        BIF_RET(am_normal);
+
+    rp = erts_proc_lookup_raw(BIF_ARG_1);
     if (!rp)
-	BIF_RET(am_false);
-	
+        BIF_RET(am_false);
+
+    state = erts_atomic32_read_nob(&rp->state);
+    dirty = (state & (ERTS_PSFLG_DIRTY_RUNNING
+                      | ERTS_PSFLG_DIRTY_RUNNING_SYS));
+    if (!dirty)
+        BIF_RET(am_normal);
+
+    busy = erts_proc_trylock(rp, ERTS_PROC_LOCK_MAIN) == EBUSY;
+
+    if (busy)
+        BIF_RET(am_busy);
+
     res = erts_check_process_code(rp, BIF_ARG_2, &reds, BIF_P->fcalls);
 
-    if (BIF_P != rp)
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
 
-    ASSERT(is_value(res));
+    ASSERT(res == am_true || res == am_false);
 
     BIF_RET2(res, reds);
-#endif
 }
 
 BIF_RETTYPE delete_module_1(BIF_ALIST_1)
@@ -683,8 +681,8 @@ BIF_RETTYPE delete_module_1(BIF_ALIST_1)
 		modp->curr.num_traced_exports > 0 ||
 		IF_HIPE(hipe_need_blocking(modp))) {
 		/* tracing or hipe need to go single threaded */
-		erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-		erts_smp_thr_progress_block();
+		erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		erts_thr_progress_block();
 		is_blocking = 1;
 		if (modp->curr.num_breakpoints) {
 		    erts_clear_module_break(modp);
@@ -782,36 +780,45 @@ BIF_RETTYPE finish_after_on_load_2(BIF_ALIST_2)
     ErtsCodeIndex code_ix;
     Module* modp;
 
+    if (BIF_ARG_2 != am_false && BIF_ARG_2 != am_true) {
+	BIF_ERROR(BIF_P, BADARG);
+    }
+
     if (!erts_try_seize_code_write_permission(BIF_P)) {
 	ERTS_BIF_YIELD2(bif_export[BIF_finish_after_on_load_2],
 			BIF_P, BIF_ARG_1, BIF_ARG_2);
     }
 
-    /* ToDo: Use code_ix staging instead of thread blocking */
-
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
-
     code_ix = erts_active_code_ix();
     modp = erts_get_module(BIF_ARG_1, code_ix);
 
-    if (!modp || !modp->on_load || !modp->on_load->code_hdr) {
-    error:
-	erts_smp_thr_progress_unblock();
-        erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    if (!modp || !modp->on_load || !modp->on_load->code_hdr
+	|| !modp->on_load->code_hdr->on_load_function_ptr) {
+
 	erts_release_code_write_permission();
 	BIF_ERROR(BIF_P, BADARG);
     }
-    if (modp->on_load->code_hdr->on_load_function_ptr == NULL) {
-	goto error;
-    }
-    if (BIF_ARG_2 != am_false && BIF_ARG_2 != am_true) {
-	goto error;
-    }
 
     if (BIF_ARG_2 == am_true) {
+	struct m mods[1];
+	int is_blocking = 0;
 	int i, num_exps;
 
+	erts_start_staging_code_ix(0);
+	code_ix = erts_staging_code_ix();
+	modp = erts_get_module(BIF_ARG_1, code_ix);
+
+	ASSERT(modp && modp->on_load && modp->on_load->code_hdr
+	       && modp->on_load->code_hdr->on_load_function_ptr);
+
+        if (erts_is_default_trace_enabled()
+	    || IF_HIPE(hipe_need_blocking(modp))) {
+
+            erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+            erts_thr_progress_block();
+            is_blocking = 1;
+	}
+
 	/*
 	 * Make the code with the on_load function current.
 	 */
@@ -837,19 +844,21 @@ BIF_RETTYPE finish_after_on_load_2(BIF_ALIST_2)
 		ep->beam[1] = 0;
 	    } else {
 		if (ep->addressv[code_ix] == ep->beam &&
-		    ep->beam[0] == (BeamInstr) em_apply_bif) {
+		    BeamIsOpCode(ep->beam[0], op_apply_bif)) {
 		    continue;
 		}
-		ep->addressv[code_ix] = ep->beam;
-		ep->beam[0] = (BeamInstr) em_call_error_handler;
+                ep->addressv[code_ix] = ep->beam;
+                ep->beam[0] = BeamOpCodeAddr(op_call_error_handler);
 	    }
 	}
 	modp->curr.code_hdr->on_load_function_ptr = NULL;
-	set_default_trace_pattern(BIF_ARG_1);
-      #ifdef HIPE
-        hipe_redirect_to_module(modp);
-      #endif
-    } else if (BIF_ARG_2 == am_false) {
+
+	mods[0].modp = modp;
+	mods[0].module = BIF_ARG_1;
+        mods[0].exception = THE_NON_VALUE;
+	return staging_epilogue(BIF_P, 1, am_true, is_blocking, mods, 1, 0);
+    }
+    else if (BIF_ARG_2 == am_false) {
 	int i, num_exps;
 
 	/*
@@ -863,14 +872,12 @@ BIF_RETTYPE finish_after_on_load_2(BIF_ALIST_2)
 	    if (ep == NULL || ep->info.mfa.module != BIF_ARG_1) {
 		continue;
 	    }
-	    if (ep->beam[0] == (BeamInstr) em_apply_bif) {
+	    if (BeamIsOpCode(ep->beam[0], op_apply_bif)) {
 		continue;
 	    }
 	    ep->beam[1] = 0;
 	}
     }
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     erts_release_code_write_permission();
     BIF_RET(am_true);
 }
@@ -906,15 +913,59 @@ static void hfrag_literal_copy(Eterm **hpp, ErlOffHeap *ohp,
                                Eterm *start, Eterm *end,
                                char *lit_start, Uint lit_size);
 
+static ERTS_INLINE void
+msg_copy_literal_area(ErtsMessage *msgp, int *redsp,
+                      char *literals, Uint lit_bsize)
+{
+    ErlHeapFragment *hfrag, *hf;
+    Uint lit_sz = 0;
+
+    *redsp += 1;
+
+    if (!ERTS_SIG_IS_INTERNAL_MSG(msgp) || !msgp->data.attached)
+        return;
+
+    if (msgp->data.attached == ERTS_MSG_COMBINED_HFRAG)
+        hfrag = &msgp->hfrag;
+    else
+        hfrag = msgp->data.heap_frag;
+
+    for (hf = hfrag; hf; hf = hf->next) {
+        lit_sz += hfrag_literal_size(&hf->mem[0],
+                                     &hf->mem[hf->used_size],
+                                     literals, lit_bsize);
+        *redsp += 1;
+    }
+
+    *redsp += lit_sz / 16; /* Better value needed... */
+    if (lit_sz > 0) {
+        ErlHeapFragment *bp = new_message_buffer(lit_sz);
+        Eterm *hp = bp->mem;
+
+        for (hf = hfrag; hf; hf = hf->next) {
+            hfrag_literal_copy(&hp, &bp->off_heap,
+                               &hf->mem[0],
+                               &hf->mem[hf->used_size],
+                               literals, lit_bsize);
+            hfrag = hf;
+        }
+
+        /* link new hfrag last */
+        ASSERT(hfrag->next == NULL);
+        hfrag->next = bp;
+        bp->next = NULL;
+    }
+}
+
 Eterm
 erts_proc_copy_literal_area(Process *c_p, int *redsp, int fcalls, int gc_allowed)
 {
     ErtsLiteralArea *la;
-    ErtsMessage *msgp;
     struct erl_off_heap_header* oh;
     char *literals;
     Uint lit_bsize;
     ErlHeapFragment *hfrag;
+    ErtsMessage *mfp;
 
     la = ERTS_COPY_LITERAL_AREA();
     if (!la)
@@ -931,47 +982,14 @@ erts_proc_copy_literal_area(Process *c_p, int *redsp, int fcalls, int gc_allowed
      * any other heap than the message it self.
      */
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
-    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
-
-    for (msgp = c_p->msg.first; msgp; msgp = msgp->next) {
-	ErlHeapFragment *hf;
-	Uint lit_sz = 0;
-
-	*redsp += 1;
-
-	if (msgp->data.attached == ERTS_MSG_COMBINED_HFRAG)
-	    hfrag = &msgp->hfrag;
-	else if (is_value(ERL_MESSAGE_TERM(msgp)) && msgp->data.heap_frag)
-	    hfrag = msgp->data.heap_frag;
-	else
-	    continue; /* Content on heap or in external term format... */
-
-	for (hf = hfrag; hf; hf = hf->next) {
-	    lit_sz += hfrag_literal_size(&hf->mem[0], &hf->mem[hf->used_size],
-					literals, lit_bsize);
-	    *redsp += 1;
-	}
-
-	*redsp += lit_sz / 16; /* Better value needed... */
-	if (lit_sz > 0) {
-	    ErlHeapFragment *bp = new_message_buffer(lit_sz);
-	    Eterm *hp = bp->mem;
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+    erts_proc_sig_fetch(c_p);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
 
-	    for (hf = hfrag; hf; hf = hf->next) {
-		hfrag_literal_copy(&hp, &bp->off_heap,
-				   &hf->mem[0], &hf->mem[hf->used_size],
-				   literals, lit_bsize);
-		hfrag = hf;
-	    }
-
-	    /* link new hfrag last */
-	    ASSERT(hfrag->next == NULL);
-	    hfrag->next = bp;
-	    bp->next = NULL;
-        }
-    }
+    ERTS_FOREACH_SIG_PRIVQS(c_p, msgp, msg_copy_literal_area(msgp,
+                                                             redsp,
+                                                             literals,
+                                                             lit_bsize));
 
     if (gc_allowed) {
 	/*
@@ -1046,8 +1064,8 @@ erts_proc_copy_literal_area(Process *c_p, int *redsp, int fcalls, int gc_allowed
      *    process off heap structure.
      *  - Check for literals
      */
-    for (msgp = c_p->msg_frag; msgp; msgp = msgp->next) {
-	hfrag = erts_message_to_heap_frag(msgp);
+    for (mfp = c_p->msg_frag; mfp; mfp = mfp->next) {
+	hfrag = erts_message_to_heap_frag(mfp);
 	for (; hfrag; hfrag = hfrag->next) {
 	    Eterm *hp, *hp_end;
 
@@ -1063,10 +1081,8 @@ erts_proc_copy_literal_area(Process *c_p, int *redsp, int fcalls, int gc_allowed
 
 return_ok:
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(c_p)))
 	c_p->flags &= ~F_DIRTY_CLA;
-#endif
 
     return am_ok;
 
@@ -1081,10 +1097,8 @@ literal_gc:
     *redsp += erts_garbage_collect_literals(c_p, (Eterm *) literals, lit_bsize,
 					    oh, fcalls);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (c_p->flags & F_DIRTY_CLA)
 	return THE_NON_VALUE;
-#endif
 
     return am_ok;
 }
@@ -1314,7 +1328,6 @@ hfrag_literal_copy(Eterm **hpp, ErlOffHeap *ohp,
     }
 }
 
-#ifdef ERTS_SMP
 
 ErtsThrPrgrLaterOp later_literal_area_switch;
 
@@ -1336,13 +1349,12 @@ static void
 complete_literal_area_switch(void *literal_area)
 {
     Process *p = erts_literal_area_collector;
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
     erts_resume(p, ERTS_PROC_LOCK_STATUS);
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     if (literal_area)
 	erts_release_literal_area((ErtsLiteralArea *) literal_area);
 }
-#endif
 
 BIF_RETTYPE erts_internal_release_literal_area_switch_0(BIF_ALIST_0)
 {
@@ -1352,7 +1364,7 @@ BIF_RETTYPE erts_internal_release_literal_area_switch_0(BIF_ALIST_0)
     if (BIF_P != erts_literal_area_collector)
 	BIF_ERROR(BIF_P, EXC_NOTSUP);
 
-    erts_smp_mtx_lock(&release_literal_areas.mtx);
+    erts_mtx_lock(&release_literal_areas.mtx);
 
     la_ref = release_literal_areas.first;
     if (la_ref) {
@@ -1361,14 +1373,13 @@ BIF_RETTYPE erts_internal_release_literal_area_switch_0(BIF_ALIST_0)
 	    release_literal_areas.last = NULL;
     }
 
-    erts_smp_mtx_unlock(&release_literal_areas.mtx);
+    erts_mtx_unlock(&release_literal_areas.mtx);
 
     unused_la = ERTS_COPY_LITERAL_AREA();
 
     if (!la_ref) {
 	ERTS_SET_COPY_LITERAL_AREA(NULL);
 	if (unused_la) {
-#ifdef ERTS_SMP
 	    ErtsLaterReleasLiteralArea *lrlap;
 	    lrlap = erts_alloc(ERTS_ALC_T_RELEASE_LAREA,
 			       sizeof(ErtsLaterReleasLiteralArea));
@@ -1382,9 +1393,6 @@ BIF_RETTYPE erts_internal_release_literal_area_switch_0(BIF_ALIST_0)
 		 + ((unused_la->end
 		     - &unused_la->start[0])
 		    - 1)*(sizeof(Eterm))));
-#else
-	    erts_release_literal_area(unused_la);
-#endif
 	}
 	BIF_RET(am_false);
     }
@@ -1393,16 +1401,11 @@ BIF_RETTYPE erts_internal_release_literal_area_switch_0(BIF_ALIST_0)
 
     erts_free(ERTS_ALC_T_LITERAL_REF, la_ref);
 
-#ifdef ERTS_SMP
     erts_schedule_thr_prgr_later_op(complete_literal_area_switch,
 				    unused_la,
 				    &later_literal_area_switch);
     erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
     ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
-#else
-    erts_release_literal_area(unused_la);
-    BIF_RET(am_true);
-#endif
 
 }
 
@@ -1428,7 +1431,7 @@ erts_purge_state_add_fun(ErlFunEntry *fe)
 Export *
 erts_suspend_process_on_pending_purge_lambda(Process *c_p, ErlFunEntry* fe)
 {
-    erts_smp_mtx_lock(&purge_state.mtx);
+    erts_mtx_lock(&purge_state.mtx);
     if (purge_state.module == fe->module) {
 	/*
 	 * The process c_p is about to call a fun in the code
@@ -1454,7 +1457,7 @@ erts_suspend_process_on_pending_purge_lambda(Process *c_p, ErlFunEntry* fe)
 	erts_suspend(c_p, ERTS_PROC_LOCK_MAIN, NULL);
 	ERTS_VBUMP_ALL_REDS(c_p);
     }
-    erts_smp_mtx_unlock(&purge_state.mtx);
+    erts_mtx_unlock(&purge_state.mtx);
     return purge_state.pending_purge_lambda;
 }
 
@@ -1464,9 +1467,9 @@ finalize_purge_operation(Process *c_p, int succeded)
     Uint ix;
 
     if (c_p)
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 
-    erts_smp_mtx_lock(&purge_state.mtx);
+    erts_mtx_lock(&purge_state.mtx);
 
     ASSERT(purge_state.module != THE_NON_VALUE);
 
@@ -1482,14 +1485,14 @@ finalize_purge_operation(Process *c_p, int succeded)
 				    ERTS_PROC_LOCK_STATUS);
 	if (rp) {
 	    erts_resume(rp, ERTS_PROC_LOCK_STATUS);
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
 	}
     }
 
-    erts_smp_mtx_unlock(&purge_state.mtx);
+    erts_mtx_unlock(&purge_state.mtx);
 
     if (c_p)
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 
     if (purge_state.sprocs != &purge_state.def_sprocs[0]) {
 	erts_free(ERTS_ALC_T_PURGE_DATA, purge_state.sprocs);
@@ -1508,7 +1511,6 @@ finalize_purge_operation(Process *c_p, int succeded)
     purge_state.fe_ix = 0;
 }
 
-#ifdef ERTS_SMP
 
 static ErtsThrPrgrLaterOp purger_lop_data;
 
@@ -1516,9 +1518,9 @@ static void
 resume_purger(void *unused)
 {
     Process *p = erts_code_purger;
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
     erts_resume(p, ERTS_PROC_LOCK_STATUS);
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 }
 
 static void
@@ -1531,7 +1533,6 @@ finalize_purge_abort(void *unused)
     resume_purger(NULL);
 }
 
-#endif /* ERTS_SMP */
 
 BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 {
@@ -1590,9 +1591,9 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 	    else {
 		BeamInstr* code;
 		BeamInstr* end;
-		erts_smp_mtx_lock(&purge_state.mtx);
+		erts_mtx_lock(&purge_state.mtx);
 		purge_state.module = BIF_ARG_1;
-		erts_smp_mtx_unlock(&purge_state.mtx);
+		erts_mtx_unlock(&purge_state.mtx);
 		res = am_true;
 		code = (BeamInstr*) modp->old.code_hdr;
 		end = (BeamInstr *)((char *)code + modp->old.code_length);
@@ -1606,9 +1607,6 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 	    }
 	}
 	
-#ifndef ERTS_SMP
-	BIF_RET(res);
-#else
 	if (res != am_true)
 	    BIF_RET(res);
 	else {
@@ -1627,7 +1625,6 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 	    erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
 	    ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
 	}
-#endif
     }
 
     case am_abort: {
@@ -1641,11 +1638,6 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 
 	erts_fun_purge_abort_prepare(purge_state.funs, purge_state.fe_ix);
 
-#ifndef ERTS_SMP
-	erts_fun_purge_abort_finalize(purge_state.funs, purge_state.fe_ix);
-	finalize_purge_operation(BIF_P, 0);
-	BIF_RET(am_false);
-#else
 	/*
 	 * We need to restore the code addresses of the funs in
 	 * two stages in order to ensure that we do not get any
@@ -1661,7 +1653,6 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 					&purger_lop_data);
 	erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
 	ERTS_BIF_YIELD_RETURN(BIF_P, am_false);
-#endif
     }
 
     case am_complete: {
@@ -1713,8 +1704,8 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 		    || IF_HIPE(hipe_purge_need_blocking(modp))) {
 		    /* ToDo: Do unload nif without blocking */
 		    erts_rwunlock_old_code(code_ix);
-		    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-		    erts_smp_thr_progress_block();
+		    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		    erts_thr_progress_block();
 		    is_blocking = 1;
 		    erts_rwlock_old_code(code_ix);
 		    if (modp->old.nif) {
@@ -1752,8 +1743,8 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 	    erts_rwunlock_old_code(code_ix);
 	}
 	if (is_blocking) {
-	    erts_smp_thr_progress_unblock();
-	    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_unblock();
+	    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	}
 
 	erts_release_code_write_permission();
@@ -1766,7 +1757,7 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 			     sizeof(ErtsLiteralAreaRef));
 	    ref->literal_area = literals;
 	    ref->next = NULL;
-	    erts_smp_mtx_lock(&release_literal_areas.mtx);
+	    erts_mtx_lock(&release_literal_areas.mtx);
 	    if (release_literal_areas.last) {
 		release_literal_areas.last->next = ref;
 		release_literal_areas.last = ref;
@@ -1775,12 +1766,12 @@ BIF_RETTYPE erts_internal_purge_module_2(BIF_ALIST_2)
 		release_literal_areas.first = ref;
 		release_literal_areas.last = ref;
 	    }
-	    erts_smp_mtx_unlock(&release_literal_areas.mtx);
-	    erts_queue_message(erts_literal_area_collector,
+	    erts_mtx_unlock(&release_literal_areas.mtx);
+	    erts_queue_proc_message(BIF_P,
+                               erts_literal_area_collector,
 			       0,
 			       erts_alloc_message(0, NULL),
-			       am_copy_literals,
-			       BIF_P->common.id);
+			       am_copy_literals);
 	}
 
 	return ret;
@@ -1807,22 +1798,23 @@ delete_code(Module* modp)
 	Export *ep = export_list(i, code_ix);
         if (ep != NULL && (ep->info.mfa.module == module)) {
 	    if (ep->addressv[code_ix] == ep->beam) {
-		if (ep->beam[0] == (BeamInstr) em_apply_bif) {
+		if (BeamIsOpCode(ep->beam[0], op_apply_bif)) {
 		    continue;
 		}
-		else if (ep->beam[0] ==
-			 (BeamInstr) BeamOp(op_i_generic_breakpoint)) {
-		    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+		else if (BeamIsOpCode(ep->beam[0], op_i_generic_breakpoint)) {
+		    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 		    ASSERT(modp->curr.num_traced_exports > 0);
 		    DBG_TRACE_MFA_P(&ep->info.mfa,
 				  "export trace cleared, code_ix=%d", code_ix);
 		    erts_clear_export_break(modp, &ep->info);
 		}
-		else ASSERT(ep->beam[0] == (BeamInstr) em_call_error_handler
-			    || !erts_initialized);
-	    }
+		else {
+                    ASSERT(BeamIsOpCode(ep->beam[0], op_call_error_handler) ||
+                           !erts_initialized);
+                }
+            }
 	    ep->addressv[code_ix] = ep->beam;
-	    ep->beam[0] = (BeamInstr) em_call_error_handler;
+	    ep->beam[0] = BeamOpCodeAddr(op_call_error_handler);
 	    ep->beam[1] = 0;
 	    DBG_TRACE_MFA_P(&ep->info.mfa,
 			    "export invalidation, code_ix=%d", code_ix);
diff --git a/erts/emulator/beam/beam_bp.c b/erts/emulator/beam/beam_bp.c
index 069e7391fd..0832b3f374 100644
--- a/erts/emulator/beam/beam_bp.c
+++ b/erts/emulator/beam/beam_bp.c
@@ -46,15 +46,15 @@
 #define ReAlloc(P, SIZ)		erts_realloc(ERTS_ALC_T_BPD, (P), (SZ))
 #define Free(P)			erts_free(ERTS_ALC_T_BPD, (P))
 
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
-#  define ERTS_SMP_REQ_PROC_MAIN_LOCK(P) \
+#if defined(ERTS_ENABLE_LOCK_CHECK)
+#  define ERTS_REQ_PROC_MAIN_LOCK(P) \
       if ((P)) erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN,\
 					 __FILE__, __LINE__)
-#  define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P) \
+#  define ERTS_UNREQ_PROC_MAIN_LOCK(P) \
       if ((P)) erts_proc_lc_unrequire_lock((P), ERTS_PROC_LOCK_MAIN)
 #else
-#  define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
-#  define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
+#  define ERTS_REQ_PROC_MAIN_LOCK(P)
+#  define ERTS_UNREQ_PROC_MAIN_LOCK(P)
 #endif
 
 #define ERTS_BPF_LOCAL_TRACE       0x01
@@ -73,11 +73,9 @@ extern BeamInstr beam_return_trace[1];      /* OpCode(i_return_trace) */
 extern BeamInstr beam_exception_trace[1];   /* OpCode(i_exception_trace) */
 extern BeamInstr beam_return_time_trace[1]; /* OpCode(i_return_time_trace) */
 
-erts_smp_atomic32_t erts_active_bp_index;
-erts_smp_atomic32_t erts_staging_bp_index;
-#ifdef ERTS_DIRTY_SCHEDULERS
-erts_smp_mtx_t erts_dirty_bp_ix_mtx;
-#endif
+erts_atomic32_t erts_active_bp_index;
+erts_atomic32_t erts_staging_bp_index;
+erts_mtx_t erts_dirty_bp_ix_mtx;
 
 /*
  * Inlined helpers
@@ -94,22 +92,18 @@ acquire_bp_sched_ix(Process *c_p)
 {
     ErtsSchedulerData *esdp = erts_proc_sched_data(c_p);
     ASSERT(esdp);
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-	erts_smp_mtx_lock(&erts_dirty_bp_ix_mtx);
+	erts_mtx_lock(&erts_dirty_bp_ix_mtx);
         return (Uint32) erts_no_schedulers;
     }
-#endif
     return (Uint32) esdp->no - 1;
 }
 
 static ERTS_INLINE void
 release_bp_sched_ix(Uint32 ix)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ix == (Uint32) erts_no_schedulers)
-        erts_smp_mtx_unlock(&erts_dirty_bp_ix_mtx);
-#endif
+        erts_mtx_unlock(&erts_dirty_bp_ix_mtx);
 }
 
 
@@ -162,12 +156,10 @@ static void bp_hash_delete(bp_time_hash_t *hash);
 
 void 
 erts_bp_init(void) {
-    erts_smp_atomic32_init_nob(&erts_active_bp_index, 0);
-    erts_smp_atomic32_init_nob(&erts_staging_bp_index, 1);
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_mtx_init(&erts_dirty_bp_ix_mtx, "dirty_break_point_index", NIL,
+    erts_atomic32_init_nob(&erts_active_bp_index, 0);
+    erts_atomic32_init_nob(&erts_staging_bp_index, 1);
+    erts_mtx_init(&erts_dirty_bp_ix_mtx, "dirty_break_point_index", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
-#endif
 }
 
 
@@ -211,7 +203,7 @@ erts_bp_match_functions(BpFunctions* f, ErtsCodeMFA *mfa, int specified)
 	for (fi = 0; fi < num_functions; fi++) {
 
 	    ci = code_hdr->functions[fi];
-	    ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI));
+	    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
 	    if (erts_is_function_native(ci)) {
 		continue;
 	    }
@@ -273,11 +265,11 @@ erts_bp_match_export(BpFunctions* f, ErtsCodeMFA *mfa, int specified)
 
 	pc = ep->beam;
 	if (ep->addressv[code_ix] == pc) {
-	    if ((*pc == (BeamInstr) em_apply_bif ||
-		 *pc == (BeamInstr) em_call_error_handler)) {
-		continue;
+	    if (BeamIsOpCode(*pc, op_apply_bif) ||
+                BeamIsOpCode(*pc, op_call_error_handler)) {
+                continue;
 	    }
-	    ASSERT(*pc == (BeamInstr) BeamOp(op_i_generic_breakpoint));
+	    ASSERT(BeamIsOpCode(*pc, op_i_generic_breakpoint));
 	} else if (erts_is_function_native(erts_code_to_codeinfo(ep->addressv[code_ix]))) {
 	    continue;
 	}
@@ -306,7 +298,7 @@ erts_consolidate_bp_data(BpFunctions* f, int local)
     Uint i;
     Uint n = f->matched;
 
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
 
     for (i = 0; i < n; i++) {
 	consolidate_bp_data(fs[i].mod, fs[i].ci, local);
@@ -318,7 +310,7 @@ erts_consolidate_bif_bp_data(void)
 {
     int i;
 
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
     for (i = 0; i < BIF_SIZE; i++) {
 	Export *ep = bif_export[i];
 	consolidate_bp_data(0, &ep->info, 0);
@@ -374,8 +366,8 @@ consolidate_bp_data(Module* modp, ErtsCodeInfo *ci, int local)
 	    }
 	    ASSERT(modp->curr.num_breakpoints >= 0);
 	    ASSERT(modp->curr.num_traced_exports >= 0);
-	    ASSERT(*erts_codeinfo_to_code(ci) !=
-                   (BeamInstr) BeamOp(op_i_generic_breakpoint));
+	    ASSERT(! BeamIsOpCode(*erts_codeinfo_to_code(ci),
+                                  op_i_generic_breakpoint));
 	}
 	ci->u.gen_bp = NULL;
 	Free(g);
@@ -393,17 +385,17 @@ consolidate_bp_data(Module* modp, ErtsCodeInfo *ci, int local)
     }
     if (flags & ERTS_BPF_META_TRACE) {
 	dst->meta_tracer = src->meta_tracer;
-	erts_smp_refc_inc(&dst->meta_tracer->refc, 1);
+	erts_refc_inc(&dst->meta_tracer->refc, 1);
 	dst->meta_ms = src->meta_ms;
 	MatchSetRef(dst->meta_ms);
     }
     if (flags & ERTS_BPF_COUNT) {
 	dst->count = src->count;
-	erts_smp_refc_inc(&dst->count->refc, 1);
+	erts_refc_inc(&dst->count->refc, 1);
     }
     if (flags & ERTS_BPF_TIME_TRACE) {
 	dst->time = src->time;
-	erts_smp_refc_inc(&dst->time->refc, 1);
+	erts_refc_inc(&dst->time->refc, 1);
 	ASSERT(dst->time->hash);
     }
 }
@@ -414,8 +406,8 @@ erts_commit_staged_bp(void)
     ErtsBpIndex staging = erts_staging_bp_ix();
     ErtsBpIndex active = erts_active_bp_ix();
 
-    erts_smp_atomic32_set_nob(&erts_active_bp_index, staging);
-    erts_smp_atomic32_set_nob(&erts_staging_bp_index, active);
+    erts_atomic32_set_nob(&erts_active_bp_index, staging);
+    erts_atomic32_set_nob(&erts_staging_bp_index, active);
 }
 
 void
@@ -423,13 +415,15 @@ erts_install_breakpoints(BpFunctions* f)
 {
     Uint i;
     Uint n = f->matched;
-    BeamInstr br = (BeamInstr) BeamOp(op_i_generic_breakpoint);
+    BeamInstr br = BeamOpCodeAddr(op_i_generic_breakpoint);
 
     for (i = 0; i < n; i++) {
 	ErtsCodeInfo* ci = f->matching[i].ci;
-        BeamInstr *pc = erts_codeinfo_to_code(ci);
 	GenericBp* g = ci->u.gen_bp;
-	if (*pc != br && g) {
+        BeamInstr volatile *pc = erts_codeinfo_to_code(ci);
+        BeamInstr instr = *pc;
+
+	if (!BeamIsOpCode(instr, op_i_generic_breakpoint) && g) {
 	    Module* modp = f->matching[i].mod;
 
 	    /*
@@ -443,11 +437,16 @@ erts_install_breakpoints(BpFunctions* f)
 	    /*
 	     * The following write is not protected by any lock. We
 	     * assume that the hardware guarantees that a write of an
-	     * aligned word-size (or half-word) writes is atomic
-	     * (i.e. that other processes executing this code will not
-	     * see a half pointer).
+	     * aligned word-size writes is atomic (i.e. that other
+	     * processes executing this code will not see a half
+	     * pointer).
+             *
+             * The contents of *pc is marked 'volatile' to ensure that
+             * the compiler will do a single full-word write, and not
+             * try any fancy optimizations to write a half word.
 	     */
-	    *pc = br;
+            instr = BeamSetCodeAddr(instr, br);
+            *pc = instr;
 	    modp->curr.num_breakpoints++;
 	}
     }
@@ -468,7 +467,7 @@ static void
 uninstall_breakpoint(ErtsCodeInfo *ci)
 {
     BeamInstr *pc = erts_codeinfo_to_code(ci);
-    if (*pc == (BeamInstr) BeamOp(op_i_generic_breakpoint)) {
+    if (BeamIsOpCode(*pc, op_i_generic_breakpoint)) {
 	GenericBp* g = ci->u.gen_bp;
 	if (g->data[erts_active_bp_ix()].flags == 0) {
 	    /*
@@ -575,7 +574,7 @@ erts_clear_mtrace_bif(ErtsCodeInfo *ci)
 void
 erts_clear_debug_break(BpFunctions* f)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
     clear_break(f, ERTS_BPF_DEBUG);
 }
 
@@ -603,7 +602,7 @@ erts_clear_module_break(Module *modp) {
     Uint n;
     Uint i;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
     ASSERT(modp);
     code_hdr = modp->curr.code_hdr;
     if (!code_hdr) {
@@ -633,7 +632,7 @@ erts_clear_module_break(Module *modp) {
 void
 erts_clear_export_break(Module* modp, ErtsCodeInfo *ci)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 
     clear_function_break(ci, ERTS_BPF_ALL);
     erts_commit_staged_bp();
@@ -656,12 +655,12 @@ static void fixup_cp_before_trace(Process *c_p, int *return_to_trace)
 {
     Eterm *cpp, *E = c_p->stop;
     BeamInstr w = *c_p->cp;
-    if (w == (BeamInstr) BeamOp(op_return_trace)) {
+    if (BeamIsOpCode(w, op_return_trace)) {
         cpp = &E[2];
-    } else if (w == (BeamInstr) BeamOp(op_i_return_to_trace)) {
+    } else if (BeamIsOpCode(w, op_i_return_to_trace)) {
         *return_to_trace = 1;
         cpp = &E[0];
-    } else if (w == (BeamInstr) BeamOp(op_i_return_time_trace)) {
+    } else if (BeamIsOpCode(w, op_i_return_time_trace)) {
         cpp = &E[0];
     } else {
         cpp = NULL;
@@ -669,12 +668,12 @@ static void fixup_cp_before_trace(Process *c_p, int *return_to_trace)
     if (cpp) {
         for (;;) {
             BeamInstr w = *cp_val(*cpp);
-            if (w == (BeamInstr) BeamOp(op_return_trace)) {
+            if (BeamIsOpCode(w, op_return_trace)) {
                 cpp += 3;
-            } else if (w == (BeamInstr) BeamOp(op_i_return_to_trace)) {
+            } else if (BeamIsOpCode(w, op_i_return_to_trace)) {
                 *return_to_trace = 1;
                 cpp += 1;
-            } else if (w == (BeamInstr) BeamOp(op_i_return_time_trace)) {
+            } else if (BeamIsOpCode(w, op_i_return_time_trace)) {
                 cpp += 2;
             } else {
                 break;
@@ -693,7 +692,7 @@ erts_generic_breakpoint(Process* c_p, ErtsCodeInfo *info, Eterm* reg)
     Uint bp_flags;
     ErtsBpIndex ix = erts_active_bp_ix();
 
-    ASSERT(info->op == (BeamInstr) BeamOp(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(info->op, op_i_func_info_IaaI));
 
     g = info->u.gen_bp;
     bp = &g->data[ix];
@@ -722,12 +721,12 @@ erts_generic_breakpoint(Process* c_p, ErtsCodeInfo *info, Eterm* reg)
     if (bp_flags & ERTS_BPF_META_TRACE) {
 	ErtsTracer old_tracer, new_tracer;
 
-	old_tracer = erts_smp_atomic_read_nob(&bp->meta_tracer->tracer);
+	old_tracer = erts_atomic_read_nob(&bp->meta_tracer->tracer);
 
 	new_tracer = do_call_trace(c_p, info, reg, 1, bp->meta_ms, old_tracer);
 
 	if (!ERTS_TRACER_COMPARE(new_tracer, old_tracer)) {
-            if (old_tracer == erts_smp_atomic_cmpxchg_acqb(
+            if (old_tracer == erts_atomic_cmpxchg_acqb(
                     &bp->meta_tracer->tracer,
                     (erts_aint_t)new_tracer,
                     (erts_aint_t)old_tracer)) {
@@ -739,16 +738,16 @@ erts_generic_breakpoint(Process* c_p, ErtsCodeInfo *info, Eterm* reg)
     }
 
     if (bp_flags & ERTS_BPF_COUNT_ACTIVE) {
-	erts_smp_atomic_inc_nob(&bp->count->acount);
+	erts_atomic_inc_nob(&bp->count->acount);
     }
 
     if (bp_flags & ERTS_BPF_TIME_TRACE_ACTIVE) {
 	Eterm w;
 	erts_trace_time_call(c_p, info, bp->time);
 	w = (BeamInstr) *c_p->cp;
-	if (! (w == (BeamInstr) BeamOp(op_i_return_time_trace) ||
-	       w == (BeamInstr) BeamOp(op_return_trace) ||
-	       w == (BeamInstr) BeamOp(op_i_return_to_trace)) ) {
+	if (! (BeamIsOpCode(w, op_i_return_time_trace) ||
+	       BeamIsOpCode(w, op_return_trace) ||
+               BeamIsOpCode(w, op_i_return_to_trace)) ) {
 	    Eterm* E = c_p->stop;
 	    ASSERT(c_p->htop <= E && E <= c_p->hend);
 	    if (E - 2 < c_p->htop) {
@@ -768,7 +767,7 @@ erts_generic_breakpoint(Process* c_p, ErtsCodeInfo *info, Eterm* reg)
     }
 
     if (bp_flags & ERTS_BPF_DEBUG) {
-	return (BeamInstr) BeamOp(op_i_debug_breakpoint);
+	return BeamOpCodeAddr(op_i_debug_breakpoint);
     } else {
 	return g->orig_instr;
     }
@@ -797,7 +796,7 @@ erts_bif_trace(int bif_index, Process* p, Eterm* args, BeamInstr* I)
     Uint bp_flags = 0;
     int return_to_trace = 0;
 
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
 
     g = ep->info.u.gen_bp;
     if (g) {
@@ -823,7 +822,7 @@ erts_bif_trace(int bif_index, Process* p, Eterm* args, BeamInstr* I)
     if (bp_flags & ERTS_BPF_META_TRACE) {
 	ErtsTracer old_tracer;
 
-        meta_tracer = erts_smp_atomic_read_nob(&bp->meta_tracer->tracer);
+        meta_tracer = erts_atomic_read_nob(&bp->meta_tracer->tracer);
         old_tracer = meta_tracer;
 	flags_meta = erts_call_trace(p, &ep->info, bp->meta_ms, args,
 				     0, &meta_tracer);
@@ -831,7 +830,7 @@ erts_bif_trace(int bif_index, Process* p, Eterm* args, BeamInstr* I)
 	if (!ERTS_TRACER_COMPARE(old_tracer, meta_tracer)) {
             ErtsTracer new_tracer = erts_tracer_nil;
             erts_tracer_update(&new_tracer, meta_tracer);
-	    if (old_tracer == erts_smp_atomic_cmpxchg_acqb(
+	    if (old_tracer == erts_atomic_cmpxchg_acqb(
                     &bp->meta_tracer->tracer,
                     (erts_aint_t)new_tracer,
                     (erts_aint_t)old_tracer)) {
@@ -958,9 +957,9 @@ erts_bif_trace_epilogue(Process *p, Eterm result, int applying,
 		}
 	    }
 	    if ((flags_meta|flags) & MATCH_SET_EXCEPTION_TRACE) {
-		erts_smp_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
 		ERTS_TRACE_FLAGS(p) |= F_EXCEPTION_TRACE;
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_unlock(p, ERTS_PROC_LOCKS_ALL_MINOR);
 	    }
 	}
     } else {
@@ -983,7 +982,7 @@ erts_bif_trace_epilogue(Process *p, Eterm result, int applying,
 	    }
 	}
     }
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
     return result;
 }
 
@@ -999,9 +998,9 @@ do_call_trace(Process* c_p, ErtsCodeInfo* info, Eterm* reg,
 
     fixup_cp_before_trace(c_p, &return_to_trace);
 
-    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
     flags = erts_call_trace(c_p, info, ms, reg, local, &tracer);
-    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
 
     /* restore cp after potential fixup */
     c_p->cp = cp_save;
@@ -1041,9 +1040,9 @@ do_call_trace(Process* c_p, ErtsCodeInfo* info, Eterm* reg,
 				  the funcinfo is above i. */
 	c_p->cp = (flags & MATCH_SET_EXCEPTION_TRACE) ?
 	    beam_exception_trace : beam_return_trace;
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+	erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 	ERTS_TRACE_FLAGS(c_p) |= F_EXCEPTION_TRACE;
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
     } else
         c_p->stop = E;
     return tracer;
@@ -1060,7 +1059,7 @@ erts_trace_time_call(Process* c_p, ErtsCodeInfo *info, BpDataTime* bdt)
     Uint32 six = acquire_bp_sched_ix(c_p);
 
     ASSERT(c_p);
-    ASSERT(erts_smp_atomic32_read_acqb(&c_p->state) & (ERTS_PSFLG_RUNNING
+    ASSERT(erts_atomic32_read_acqb(&c_p->state) & (ERTS_PSFLG_RUNNING
 						       | ERTS_PSFLG_DIRTY_RUNNING));
 
     /* get previous timestamp and breakpoint
@@ -1141,7 +1140,7 @@ erts_trace_time_return(Process *p, ErtsCodeInfo *ci)
     Uint32 six = acquire_bp_sched_ix(p);
 
     ASSERT(p);
-    ASSERT(erts_smp_atomic32_read_acqb(&p->state) & (ERTS_PSFLG_RUNNING
+    ASSERT(erts_atomic32_read_acqb(&p->state) & (ERTS_PSFLG_RUNNING
 						     | ERTS_PSFLG_DIRTY_RUNNING));
 
     /* get previous timestamp and breakpoint
@@ -1223,7 +1222,7 @@ erts_is_mtrace_break(ErtsCodeInfo *ci, Binary **match_spec_ret,
 	    *match_spec_ret = bp->meta_ms;
 	}
 	if (tracer_ret) {
-            *tracer_ret = erts_smp_atomic_read_nob(&bp->meta_tracer->tracer);
+            *tracer_ret = erts_atomic_read_nob(&bp->meta_tracer->tracer);
 	}
 	return 1;
     }
@@ -1237,7 +1236,7 @@ erts_is_count_break(ErtsCodeInfo *ci, Uint *count_ret)
     
     if (bp) {
 	if (count_ret) {
-	    *count_ret = (Uint) erts_smp_atomic_read_nob(&bp->count->acount);
+	    *count_ret = (Uint) erts_atomic_read_nob(&bp->count->acount);
 	}
 	return 1;
     }
@@ -1318,7 +1317,7 @@ erts_find_local_func(ErtsCodeMFA *mfa) {
     n = (BeamInstr) code_hdr->num_functions;
     for (i = 0; i < n; ++i) {
 	ci = code_hdr->functions[i];
-	ASSERT(((BeamInstr) BeamOp(op_i_func_info_IaaI)) == ci->op);
+	ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
 	ASSERT(mfa->module == ci->mfa.module || is_nil(ci->mfa.module));
 	if (mfa->function == ci->mfa.function &&
 	    mfa->arity == ci->mfa.arity) {
@@ -1517,7 +1516,7 @@ set_function_break(ErtsCodeInfo *ci, Binary *match_spec, Uint break_flags,
     Uint common;
     ErtsBpIndex ix = erts_staging_bp_ix();
 
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
     g = ci->u.gen_bp;
     if (g == 0) {
 	int i;
@@ -1549,7 +1548,7 @@ set_function_break(ErtsCodeInfo *ci, Binary *match_spec, Uint break_flags,
 	    bp->flags &= ~ERTS_BPF_COUNT_ACTIVE;
 	} else {
 	    bp->flags |= ERTS_BPF_COUNT_ACTIVE;
-	    erts_smp_atomic_set_nob(&bp->count->acount, 0);
+	    erts_atomic_set_nob(&bp->count->acount, 0);
 	}
 	ASSERT((bp->flags & ~ERTS_BPF_ALL) == 0);
 	return;
@@ -1583,17 +1582,17 @@ set_function_break(ErtsCodeInfo *ci, Binary *match_spec, Uint break_flags,
 	MatchSetRef(match_spec);
 	bp->meta_ms = match_spec;
 	bmt = Alloc(sizeof(BpMetaTracer));
-	erts_smp_refc_init(&bmt->refc, 1);
+	erts_refc_init(&bmt->refc, 1);
         erts_tracer_update(&meta_tracer, tracer); /* copy tracer */
-	erts_smp_atomic_init_nob(&bmt->tracer, (erts_aint_t)meta_tracer);
+	erts_atomic_init_nob(&bmt->tracer, (erts_aint_t)meta_tracer);
 	bp->meta_tracer = bmt;
     } else if (break_flags & ERTS_BPF_COUNT) {
 	BpCount* bcp;
 
 	ASSERT((bp->flags & ERTS_BPF_COUNT) == 0);
 	bcp = Alloc(sizeof(BpCount));
-	erts_smp_refc_init(&bcp->refc, 1);
-	erts_smp_atomic_init_nob(&bcp->acount, 0);
+	erts_refc_init(&bcp->refc, 1);
+	erts_atomic_init_nob(&bcp->acount, 0);
 	bp->count = bcp;
     } else if (break_flags & ERTS_BPF_TIME_TRACE) {
 	BpDataTime* bdt;
@@ -1601,12 +1600,8 @@ set_function_break(ErtsCodeInfo *ci, Binary *match_spec, Uint break_flags,
 
 	ASSERT((bp->flags & ERTS_BPF_TIME_TRACE) == 0);
 	bdt = Alloc(sizeof(BpDataTime));
-	erts_smp_refc_init(&bdt->refc, 1);
-#ifdef ERTS_DIRTY_SCHEDULERS
+	erts_refc_init(&bdt->refc, 1);
 	bdt->n = erts_no_schedulers + 1;
-#else
-	bdt->n = erts_no_schedulers;
-#endif
 	bdt->hash = Alloc(sizeof(bp_time_hash_t)*(bdt->n));
 	for (i = 0; i < bdt->n; i++) {
 	    bp_hash_init(&(bdt->hash[i]), 32);
@@ -1638,7 +1633,7 @@ clear_function_break(ErtsCodeInfo *ci, Uint break_flags)
     Uint common;
     ErtsBpIndex ix = erts_staging_bp_ix();
 
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
 
     if ((g = ci->u.gen_bp) == NULL) {
 	return 1;
@@ -1671,8 +1666,8 @@ clear_function_break(ErtsCodeInfo *ci, Uint break_flags)
 static void
 bp_meta_unref(BpMetaTracer* bmt)
 {
-    if (erts_smp_refc_dectest(&bmt->refc, 0) <= 0) {
-        ErtsTracer trc = erts_smp_atomic_read_nob(&bmt->tracer);
+    if (erts_refc_dectest(&bmt->refc, 0) <= 0) {
+        ErtsTracer trc = erts_atomic_read_nob(&bmt->tracer);
         ERTS_TRACER_CLEAR(&trc);
 	Free(bmt);
     }
@@ -1681,7 +1676,7 @@ bp_meta_unref(BpMetaTracer* bmt)
 static void
 bp_count_unref(BpCount* bcp)
 {
-    if (erts_smp_refc_dectest(&bcp->refc, 0) <= 0) {
+    if (erts_refc_dectest(&bcp->refc, 0) <= 0) {
 	Free(bcp);
     }
 }
@@ -1689,7 +1684,7 @@ bp_count_unref(BpCount* bcp)
 static void
 bp_time_unref(BpDataTime* bdt)
 {
-    if (erts_smp_refc_dectest(&bdt->refc, 0) <= 0) {
+    if (erts_refc_dectest(&bdt->refc, 0) <= 0) {
 	Uint i = 0;
 	Uint j = 0;
 	Process *h_p = NULL;
@@ -1713,7 +1708,7 @@ bp_time_unref(BpDataTime* bdt)
 			    if (pbt) {
 				Free(pbt);
 			    }
-			    erts_smp_proc_unlock(h_p, ERTS_PROC_LOCK_MAIN);
+			    erts_proc_unlock(h_p, ERTS_PROC_LOCK_MAIN);
 			}
 		    }
 		}
@@ -1737,7 +1732,7 @@ check_break(ErtsCodeInfo *ci, Uint break_flags)
 {
     GenericBp* g = ci->u.gen_bp;
 
-    ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
     if (erts_is_function_native(ci)) {
 	return 0;
     }
diff --git a/erts/emulator/beam/beam_bp.h b/erts/emulator/beam/beam_bp.h
index 56fa82b912..a64765822b 100644
--- a/erts/emulator/beam/beam_bp.h
+++ b/erts/emulator/beam/beam_bp.h
@@ -41,7 +41,7 @@ typedef struct {
 typedef struct bp_data_time {     /* Call time */
     Uint n;
     bp_time_hash_t *hash;
-    erts_smp_refc_t refc;
+    erts_refc_t refc;
 } BpDataTime;
 
 typedef struct {
@@ -50,13 +50,13 @@ typedef struct {
 } process_breakpoint_time_t; /* used within psd */
 
 typedef struct {
-    erts_smp_atomic_t acount;
-    erts_smp_refc_t refc;
+    erts_atomic_t acount;
+    erts_refc_t refc;
 } BpCount;
 
 typedef struct {
-    erts_smp_atomic_t tracer;
-    erts_smp_refc_t refc;
+    erts_atomic_t tracer;
+    erts_refc_t refc;
 } BpMetaTracer;
 
 typedef struct generic_bp_data {
@@ -79,9 +79,7 @@ typedef struct generic_bp {
 #define ERTS_BP_CALL_TIME_SCHEDULE_OUT     (1)
 #define ERTS_BP_CALL_TIME_SCHEDULE_EXITING (2)
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-extern erts_smp_mtx_t erts_dirty_bp_ix_mtx;
-#endif
+extern erts_mtx_t erts_dirty_bp_ix_mtx;
 
 enum erts_break_op{
     ERTS_BREAK_NOP   =  0, /* Must be false */
@@ -173,17 +171,17 @@ ErtsCodeInfo *erts_find_local_func(ErtsCodeMFA *mfa);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
-extern erts_smp_atomic32_t erts_active_bp_index;
-extern erts_smp_atomic32_t erts_staging_bp_index;
+extern erts_atomic32_t erts_active_bp_index;
+extern erts_atomic32_t erts_staging_bp_index;
 
 ERTS_GLB_INLINE ErtsBpIndex erts_active_bp_ix(void)
 {
-    return erts_smp_atomic32_read_nob(&erts_active_bp_index);
+    return erts_atomic32_read_nob(&erts_active_bp_index);
 }
 
 ERTS_GLB_INLINE ErtsBpIndex erts_staging_bp_ix(void)
 {
-    return erts_smp_atomic32_read_nob(&erts_staging_bp_index);
+    return erts_atomic32_read_nob(&erts_staging_bp_index);
 }
 #endif
 
diff --git a/erts/emulator/beam/beam_debug.c b/erts/emulator/beam/beam_debug.c
index a2060c80de..b8a8d06315 100644
--- a/erts/emulator/beam/beam_debug.c
+++ b/erts/emulator/beam/beam_debug.c
@@ -40,6 +40,7 @@
 #include "erl_binary.h"
 #include "erl_thr_progress.h"
 #include "erl_nfunc_sched.h"
+#include "beam_catches.h"
 
 #ifdef ARCH_64
 # define HEXF "%016bpX"
@@ -53,6 +54,9 @@ void dbg_where(BeamInstr* addr, Eterm x0, Eterm* reg);
 
 static int print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr);
 static void print_bif_name(fmtfn_t to, void* to_arg, BifFunction bif);
+static BeamInstr* f_to_addr(BeamInstr* base, int op, BeamInstr* ap);
+static BeamInstr* f_to_addr_packed(BeamInstr* base, int op, Sint32* ap);
+static void print_byte_string(fmtfn_t to, void *to_arg, byte* str, Uint bytes);
 
 BIF_RETTYPE
 erts_debug_same_2(BIF_ALIST_2)
@@ -157,8 +161,8 @@ erts_debug_breakpoint_2(BIF_ALIST_2)
 	ERTS_BIF_YIELD2(bif_export[BIF_erts_debug_breakpoint_2],
 			BIF_P, BIF_ARG_1, BIF_ARG_2);
     }
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
+    erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
 
     erts_bp_match_functions(&f, &mfa, specified);
     if (boolean == am_true) {
@@ -174,8 +178,8 @@ erts_debug_breakpoint_2(BIF_ALIST_2)
     res = make_small(f.matched);
     erts_bp_free_matched_functions(&f);
 
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     erts_release_code_write_permission();
     return res;
 
@@ -197,9 +201,9 @@ void debug_dump_code(BeamInstr *I, int num)
 	erts_print(ERTS_PRINT_DSBUF, (void *) dsbufp, HEXF ": ", code_ptr);
 	instr = (BeamInstr) code_ptr[0];
 	for (i = 0; i < NUM_SPECIFIC_OPS; i++) {
-	    if (instr == (BeamInstr) BeamOp(i) && opc[i].name[0] != '\0') {
+	    if (BeamIsOpCode(instr, i) && opc[i].name[0] != '\0') {
 		code_ptr += print_op(ERTS_PRINT_DSBUF, (void *) dsbufp,
-				     i, opc[i].sz-1, code_ptr+1) + 1;
+				     i, opc[i].sz-1, code_ptr) + 1;
 		break;
 	    }
 	}
@@ -224,11 +228,11 @@ erts_debug_instructions_0(BIF_ALIST_0)
     Eterm res = NIL;
 
     for (i = 0; i < num_instructions; i++) {
-	needed += 2*strlen(opc[i].name);
+	needed += 2*sys_strlen(opc[i].name);
     }
     hp = HAlloc(BIF_P, needed);
     for (i = num_instructions-1; i >= 0; i--) {
-	Eterm s = erts_bld_string_n(&hp, 0, opc[i].name, strlen(opc[i].name));
+	Eterm s = erts_bld_string_n(&hp, 0, opc[i].name, sys_strlen(opc[i].name));
 	res = erts_bld_cons(&hp, 0, s, res);
     }
     return res;
@@ -317,9 +321,9 @@ erts_debug_disassemble_1(BIF_ALIST_1)
     erts_print(ERTS_PRINT_DSBUF, (void *) dsbufp, HEXF ": ", code_ptr);
     instr = (BeamInstr) code_ptr[0];
     for (i = 0; i < NUM_SPECIFIC_OPS; i++) {
-	if (instr == (BeamInstr) BeamOp(i) && opc[i].name[0] != '\0') {
+	if (BeamIsOpCode(instr, i) && opc[i].name[0] != '\0') {
 	    code_ptr += print_op(ERTS_PRINT_DSBUF, (void *) dsbufp,
-				 i, opc[i].sz-1, code_ptr+1) + 1;
+				 i, opc[i].sz-1, code_ptr) + 1;
 	    break;
 	}
     }
@@ -394,6 +398,7 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
     BeamInstr args[8];		/* Arguments for this instruction. */
     BeamInstr* ap;			/* Pointer to arguments. */
     BeamInstr* unpacked;		/* Unpacked arguments */
+    BeamInstr* first_arg;               /* First argument */
 
     start_prog = opc[op].pack;
 
@@ -403,8 +408,14 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	 * Avoid copying because instructions containing bignum operands
 	 * are bigger than actually declared.
 	 */
-	ap = (BeamInstr *) addr;
+        addr++;
+        ap = addr;
     } else {
+#if defined(ARCH_64) && defined(CODE_MODEL_SMALL)
+        BeamInstr instr_word = addr[0];
+#endif
+        addr++;
+
 	/*
 	 * Copy all arguments to a local buffer for the unpacking.
 	 */
@@ -420,30 +431,31 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	 * the packing program backwards and in reverse.
 	 */
 
-	prog = start_prog + strlen(start_prog);
+	prog = start_prog + sys_strlen(start_prog);
 	while (start_prog < prog) {
 	    prog--;
 	    switch (*prog) {
+	    case 'f':
 	    case 'g':
+	    case 'q':
 		*ap++ = *--sp;
 		break;
-	    case 'i':		/* Initialize packing accumulator. */
-		*ap++ = packed;
-		break;
-	    case 's':
-		*ap++ = packed & 0x3ff;
-		packed >>= 10;
+#ifdef ARCH_64
+	    case '1':		/* Tightest shift */
+		*ap++ = (packed & BEAM_TIGHTEST_MASK) << 3;
+		packed >>= BEAM_TIGHTEST_SHIFT;
 		break;
-	    case '0':		/* Tight shift */
+#endif
+	    case '2':		/* Tight shift */
 		*ap++ = packed & BEAM_TIGHT_MASK;
 		packed >>= BEAM_TIGHT_SHIFT;
 		break;
-	    case '6':		/* Shift 16 steps */
+	    case '3':		/* Loose shift */
 		*ap++ = packed & BEAM_LOOSE_MASK;
 		packed >>= BEAM_LOOSE_SHIFT;
 		break;
 #ifdef ARCH_64
-	    case 'w':		/* Shift 32 steps */
+	    case '4':		/* Shift 32 steps */
 		*ap++ = packed & BEAM_WIDE_MASK;
 		packed >>= BEAM_WIDE_SHIFT;
 		break;
@@ -454,13 +466,25 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	    case 'P':
 		packed = *--sp;
 		break;
+#if defined(ARCH_64) && defined(CODE_MODEL_SMALL)
+            case '#':       /* -1 */
+            case '$':       /* -2 */
+            case '%':       /* -3 */
+            case '&':       /* -4 */
+            case '\'':      /* -5 */
+            case '(':       /* -6 */
+                packed = (packed << BEAM_WIDE_SHIFT) | BeamExtraData(instr_word);
+		break;
+#endif
 	    default:
-		ASSERT(0);
+                erts_exit(ERTS_ERROR_EXIT, "beam_debug: invalid packing op: %c\n", *prog);
 	    }
 	}
 	ap = args;
     }
 
+    first_arg = ap;
+
     /*
      * Print the name and all operands of the instructions.
      */
@@ -489,6 +513,14 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	case 'n':		/* Nil */
 	    erts_print(to, to_arg, "[]");
 	    break;
+        case 'S':               /* Register */
+            {
+                Uint reg_type = (*ap & 1) ? 'y' : 'x';
+                Uint n = ap[0] / sizeof(Eterm);
+                erts_print(to, to_arg, "%c(%d)", reg_type, n);
+		ap++;
+                break;
+            }
 	case 's':		/* Any source (tagged constant or register) */
 	    tag = loader_tag(*ap);
 	    if (tag == LOADER_X_REG) {
@@ -522,12 +554,13 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	    ap++;
 	    break;
-	case 'I':		/* Untagged integer. */
-	case 't':
+	case 't':               /* Untagged integers */
+	case 'I':
+        case 'W':
 	    switch (op) {
-	    case op_i_gc_bif1_jIsId:
-	    case op_i_gc_bif2_jIIssd:
-	    case op_i_gc_bif3_jIIssd:
+	    case op_i_gc_bif1_jWstd:
+	    case op_i_gc_bif2_jWtssd:
+	    case op_i_gc_bif3_jWtssd:
 		{
 		    const ErtsGcBif* p;
 		    BifFunction gcf = (BifFunction) *ap;
@@ -542,37 +575,74 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 		    }
 		    break;
 		}
+	    case op_i_make_fun_Wt:
+                if (*sign == 'W') {
+                    ErlFunEntry* fe = (ErlFunEntry *) *ap;
+                    ErtsCodeMFA* cmfa = find_function_from_pc(fe->address);
+		    erts_print(to, to_arg, "%T:%T/%bpu", cmfa->module,
+                               cmfa->function, cmfa->arity);
+                } else {
+                    erts_print(to, to_arg, "%d", *ap);
+                }
+                break;
+	    case op_i_bs_match_string_xfWW:
+                if (ap - first_arg < 3) {
+                    erts_print(to, to_arg, "%d", *ap);
+                } else {
+                    Uint bits = ap[-1];
+                    Uint bytes = (bits+7)/8;
+                    byte* str = (byte *) *ap;
+                    print_byte_string(to, to_arg, str, bytes);
+                }
+                break;
+	    case op_bs_put_string_WW:
+                if (ap - first_arg == 0) {
+                    erts_print(to, to_arg, "%d", *ap);
+                } else {
+                    Uint bytes = ap[-1];
+                    byte* str = (byte *) ap[0];
+                    print_byte_string(to, to_arg, str, bytes);
+                }
+                break;
 	    default:
 		erts_print(to, to_arg, "%d", *ap);
 	    }
 	    ap++;
 	    break;
 	case 'f':		/* Destination label */
-	    {
-		ErtsCodeMFA* cmfa = find_function_from_pc((BeamInstr *)*ap);
-		if (!cmfa || erts_codemfa_to_code(cmfa) != (BeamInstr *) *ap) {
-		    erts_print(to, to_arg, "f(" HEXF ")", *ap);
-		} else {
-		    erts_print(to, to_arg, "%T:%T/%bpu", cmfa->module,
-                               cmfa->function, cmfa->arity);
-		}
-		ap++;
-	    }
-	    break;
+            switch (op) {
+            case op_catch_yf:
+                erts_print(to, to_arg, "f(" HEXF ")", catch_pc((BeamInstr)*ap));
+                break;
+            default:
+                {
+                    BeamInstr* target = f_to_addr(addr, op, ap);
+                    ErtsCodeMFA* cmfa = find_function_from_pc(target);
+                    if (!cmfa || erts_codemfa_to_code(cmfa) != target) {
+                        erts_print(to, to_arg, "f(" HEXF ")", target);
+                    } else {
+                        erts_print(to, to_arg, "%T:%T/%bpu", cmfa->module,
+                                   cmfa->function, cmfa->arity);
+                    }
+                    ap++;
+                }
+                break;
+            }
+            break;
 	case 'p':		/* Pointer (to label) */
 	    {
-		ErtsCodeMFA* cmfa = find_function_from_pc((BeamInstr *)*ap);
-		if (!cmfa || erts_codemfa_to_code(cmfa) != (BeamInstr *) *ap) {
-		    erts_print(to, to_arg, "p(" HEXF ")", *ap);
-		} else {
-		    erts_print(to, to_arg, "%T:%T/%bpu", cmfa->module,
-                               cmfa->function, cmfa->arity);
-		}
+                BeamInstr* target = f_to_addr(addr, op, ap);
+                erts_print(to, to_arg, "p(" HEXF ")", target);
 		ap++;
 	    }
 	    break;
 	case 'j':		/* Pointer (to label) */
-	    erts_print(to, to_arg, "j(" HEXF ")", *ap);
+            if (*ap == 0) {
+                erts_print(to, to_arg, "j(0)");
+            } else {
+                BeamInstr* target = f_to_addr(addr, op, ap);
+                erts_print(to, to_arg, "j(" HEXF ")", target);
+            }
 	    ap++;
 	    break;
 	case 'e':		/* Export entry */
@@ -615,12 +685,22 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 
     unpacked = ap;
     ap = addr + size;
+
+    /*
+     * In the code below, never use ap[-1], ap[-2], ...
+     * (will not work if the arguments have been packed).
+     *
+     * Instead use unpacked[-1], unpacked[-2], ...
+     */
     switch (op) {
     case op_i_select_val_lins_xfI:
     case op_i_select_val_lins_yfI:
+    case op_i_select_val_bins_xfI:
+    case op_i_select_val_bins_yfI:
 	{
-	    int n = ap[-1];
+	    int n = unpacked[-1];
 	    int ix = n;
+            Sint32* jump_tab = (Sint32 *)(ap + n);
 
 	    while (ix--) {
 		erts_print(to, to_arg, "%T ", (Eterm) ap[0]);
@@ -629,30 +709,19 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	    ix = n;
 	    while (ix--) {
-		erts_print(to, to_arg, "f(" HEXF ") ", (Eterm) ap[0]);
-		ap++;
-		size++;
-	    }
-	}
-	break;
-    case op_i_select_val_bins_xfI:
-    case op_i_select_val_bins_yfI:
-	{
-	    int n = ap[-1];
-
-	    while (n > 0) {
-		erts_print(to, to_arg, "%T f(" HEXF ") ", (Eterm) ap[0], ap[1]);
-		ap += 2;
-		size += 2;
-		n--;
+                BeamInstr* target = f_to_addr_packed(addr, op, jump_tab);
+		erts_print(to, to_arg, "f(" HEXF ") ", target);
+                jump_tab++;
 	    }
+            size += (n+1) / 2;
 	}
 	break;
     case op_i_select_tuple_arity_xfI:
     case op_i_select_tuple_arity_yfI:
         {
-            int n = ap[-1];
+            int n = unpacked[-1];
             int ix = n - 1; /* without sentinel */
+            Sint32* jump_tab = (Sint32 *)(ap + n);
 
             while (ix--) {
                 Uint arity = arityval(ap[0]);
@@ -666,39 +735,62 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
             size++;
             ix = n;
             while (ix--) {
-                erts_print(to, to_arg, "f(" HEXF ") ", ap[0]);
-                ap++;
-                size++;
+                BeamInstr* target = f_to_addr_packed(addr, op, jump_tab);
+                erts_print(to, to_arg, "f(" HEXF ") ", target);
+                jump_tab++;
+            }
+            size += (n+1) / 2;
+        }
+        break;
+    case op_i_select_val2_xfcc:
+    case op_i_select_val2_yfcc:
+    case op_i_select_tuple_arity2_xfAA:
+    case op_i_select_tuple_arity2_yfAA:
+        {
+            Sint32* jump_tab = (Sint32 *) ap;
+            BeamInstr* target;
+            int i;
+
+            for (i = 0; i < 2; i++) {
+                target = f_to_addr_packed(addr, op, jump_tab++);
+                erts_print(to, to_arg, "f(" HEXF ") ", target);
             }
+            size += 1;
         }
         break;
-    case op_i_jump_on_val_xfII:
-    case op_i_jump_on_val_yfII:
+    case op_i_jump_on_val_xfIW:
+    case op_i_jump_on_val_yfIW:
 	{
-	    int n;
-	    for (n = ap[-2]; n > 0; n--) {
-		erts_print(to, to_arg, "f(" HEXF ") ", ap[0]);
-		ap++;
-		size++;
+	    int n = unpacked[-2];
+            Sint32* jump_tab = (Sint32 *) ap;
+
+            size += (n+1) / 2;
+            while (n-- > 0) {
+                BeamInstr* target = f_to_addr_packed(addr, op, jump_tab);
+		erts_print(to, to_arg, "f(" HEXF ") ", target);
+                jump_tab++;
 	    }
 	}
 	break;
     case op_i_jump_on_val_zero_xfI:
     case op_i_jump_on_val_zero_yfI:
 	{
-	    int n;
-	    for (n = ap[-1]; n > 0; n--) {
-		erts_print(to, to_arg, "f(" HEXF ") ", ap[0]);
-		ap++;
-		size++;
+	    int n = unpacked[-1];
+            Sint32* jump_tab = (Sint32 *) ap;
+
+            size += (n+1) / 2;
+            while (n-- > 0) {
+                BeamInstr* target = f_to_addr_packed(addr, op, jump_tab);
+		erts_print(to, to_arg, "f(" HEXF ") ", target);
+                jump_tab++;
 	    }
 	}
 	break;
     case op_i_put_tuple_xI:
     case op_i_put_tuple_yI:
-    case op_new_map_dII:
-    case op_update_map_assoc_jsdII:
-    case op_update_map_exact_jsdII:
+    case op_new_map_dtI:
+    case op_update_map_assoc_sdtI:
+    case op_update_map_exact_jsdtI:
 	{
 	    int n = unpacked[-1];
 
@@ -708,7 +800,7 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 		    erts_print(to, to_arg, " x(%d)", loader_x_reg_index(ap[0]));
 		    break;
 		case LOADER_Y_REG:
-		    erts_print(to, to_arg, " x(%d)", loader_y_reg_index(ap[0]));
+		    erts_print(to, to_arg, " y(%d)", loader_y_reg_index(ap[0]) - CP_SIZE);
 		    break;
 		default:
 		    erts_print(to, to_arg, " %T", (Eterm) ap[0]);
@@ -718,6 +810,27 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 	    }
 	}
 	break;
+    case op_i_new_small_map_lit_dtq:
+        {
+            Eterm *tp = tuple_val(unpacked[-1]);
+            int n = arityval(*tp);
+
+            while (n > 0) {
+                switch (loader_tag(ap[0])) {
+                case LOADER_X_REG:
+                    erts_print(to, to_arg, " x(%d)", loader_x_reg_index(ap[0]));
+                    break;
+                case LOADER_Y_REG:
+		    erts_print(to, to_arg, " y(%d)", loader_y_reg_index(ap[0]) - CP_SIZE);
+		    break;
+                default:
+		    erts_print(to, to_arg, " %T", (Eterm) ap[0]);
+		    break;
+                }
+                ap++, size++, n--;
+            }
+        }
+        break;
     case op_i_get_map_elements_fsI:
 	{
 	    int n = unpacked[-1];
@@ -731,7 +844,7 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr)
 			erts_print(to, to_arg, " x(%d)", loader_x_reg_index(ap[0]));
 			break;
 		    case LOADER_Y_REG:
-			erts_print(to, to_arg, " y(%d)", loader_y_reg_index(ap[0]));
+			erts_print(to, to_arg, " y(%d)", loader_y_reg_index(ap[0]) - CP_SIZE);
 			break;
 		    default:
 			erts_print(to, to_arg, " %T", (Eterm) ap[0]);
@@ -766,6 +879,25 @@ static void print_bif_name(fmtfn_t to, void* to_arg, BifFunction bif)
     }
 }
 
+static BeamInstr* f_to_addr(BeamInstr* base, int op, BeamInstr* ap)
+{
+    return base - 1 + opc[op].adjust + (Sint32) *ap;
+}
+
+static BeamInstr* f_to_addr_packed(BeamInstr* base, int op, Sint32* ap)
+{
+    return base - 1 + opc[op].adjust + *ap;
+}
+
+static void print_byte_string(fmtfn_t to, void *to_arg, byte* str, Uint bytes)
+{
+    Uint i;
+
+    for (i = 0; i < bytes; i++) {
+        erts_print(to, to_arg, "%02X", str[i]);
+    }
+}
+
 /*
  * Dirty BIF testing.
  *
@@ -774,10 +906,8 @@ static void print_bif_name(fmtfn_t to, void* to_arg, BifFunction bif)
  * test suite.
  */
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 static int ms_wait(Process *c_p, Eterm etimeout, int busy);
 static int dirty_send_message(Process *c_p, Eterm to, Eterm tag);
-#endif
 static BIF_RETTYPE dirty_test(Process *c_p, Eterm type, Eterm arg1, Eterm arg2, UWord *I);
 
 /*
@@ -806,7 +936,6 @@ erts_debug_dirty_io_2(BIF_ALIST_2)
 BIF_RETTYPE
 erts_debug_dirty_3(BIF_ALIST_3)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     Eterm argv[2];
     switch (BIF_ARG_1) {
     case am_normal:
@@ -836,9 +965,6 @@ erts_debug_dirty_3(BIF_ALIST_3)
     default:
 	BIF_ERROR(BIF_P, EXC_BADARG);
     }
-#else
-	BIF_ERROR(BIF_P, EXC_UNDEF);
-#endif
 }
 
 
@@ -846,7 +972,6 @@ static BIF_RETTYPE
 dirty_test(Process *c_p, Eterm type, Eterm arg1, Eterm arg2, UWord *I)
 {
     BIF_RETTYPE ret;
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (am_scheduler == arg1) {
 	ErtsSchedulerData *esdp;
 	if (arg2 != am_type) 
@@ -1032,13 +1157,9 @@ dirty_test(Process *c_p, Eterm type, Eterm arg1, Eterm arg2, UWord *I)
     badarg:
 	ERTS_BIF_PREP_ERROR(ret, c_p, BADARG);
     }
-#else
-    ERTS_BIF_PREP_ERROR(ret, c_p, EXC_UNDEF);
-#endif
     return ret;
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static int
 dirty_send_message(Process *c_p, Eterm to, Eterm tag)
@@ -1070,12 +1191,12 @@ dirty_send_message(Process *c_p, Eterm to, Eterm tag)
     mp = erts_alloc_message_heap(rp, &rp_locks, 3, &hp, &ohp);
 
     msg = TUPLE2(hp, tag, c_p->common.id);
-    erts_queue_message(rp, rp_locks, mp, msg, c_p->common.id);
+    erts_queue_proc_message(c_p, rp, rp_locks, mp, msg);
 
     if (rp == real_c_p)
 	rp_locks &= ~c_p_locks;
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
     erts_proc_dec_refc(rp);
 
@@ -1125,13 +1246,8 @@ ms_wait(Process *c_p, Eterm etimeout, int busy)
     return 1;
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
-#ifdef ERTS_SMP
 #  define ERTS_STACK_LIMIT ((char *) ethr_get_stacklimit())
-#else
-#  define ERTS_STACK_LIMIT ((char *) erts_scheduler_stack_limit)
-#endif
 
 /*
  * The below functions is for testing of the stack
diff --git a/erts/emulator/beam/beam_emu.c b/erts/emulator/beam/beam_emu.c
index b4e6c35579..ab5920a67e 100644
--- a/erts/emulator/beam/beam_emu.c
+++ b/erts/emulator/beam/beam_emu.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -44,51 +44,47 @@
 #include "hipe_bif1.h"
 #endif
 #include "dtrace-wrapper.h"
+#include "erl_proc_sig_queue.h"
 
 /* #define HARDDEBUG 1 */
 
 #if defined(NO_JUMP_TABLE)
 #  define OpCase(OpCode)    case op_##OpCode
 #  define CountCase(OpCode) case op_count_##OpCode
-#  define OpCode(OpCode)    ((Uint*)op_##OpCode)
-#  define Goto(Rel) {Go = (int)(UWord)(Rel); goto emulator_loop;}
-#  define LabelAddr(Addr) &&##Addr
+#  define IsOpCode(InstrWord, OpCode)  (BeamCodeAddr(InstrWord) == (BeamInstr)op_##OpCode)
+#  define Goto(Rel)         {Go = BeamCodeAddr(Rel); goto emulator_loop;}
+#  define GotoPF(Rel)       Goto(Rel)
 #else
 #  define OpCase(OpCode)    lb_##OpCode
 #  define CountCase(OpCode) lb_count_##OpCode
-#  define Goto(Rel) goto *((void *)Rel)
-#  define LabelAddr(Label) &&Label
-#  define OpCode(OpCode)  (&&lb_##OpCode)
+#  define IsOpCode(InstrWord, OpCode)  (BeamCodeAddr(InstrWord) == (BeamInstr)&&lb_##OpCode)
+#  define Goto(Rel)         goto *((void *)BeamCodeAddr(Rel))
+#  define GotoPF(Rel)       goto *((void *)Rel)
+#  define LabelAddr(Label)  &&Label
 #endif
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
-#  ifdef ERTS_SMP
-#    define PROCESS_MAIN_CHK_LOCKS(P)					\
-do {									\
-    if ((P))								\
-	erts_proc_lc_chk_only_proc_main((P));				\
-    ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());		\
+#    define PROCESS_MAIN_CHK_LOCKS(P)                   \
+do {                                                    \
+    if ((P))                                            \
+	erts_proc_lc_chk_only_proc_main((P));           \
+    ERTS_LC_ASSERT(!erts_thr_progress_is_blocking());   \
 } while (0)
-#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)				\
-do {									\
-    if ((P))								\
-	erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN,		\
-				  __FILE__, __LINE__);			\
+#    define ERTS_REQ_PROC_MAIN_LOCK(P)				\
+do {                                                            \
+    if ((P))                                                    \
+	erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN,     \
+				  __FILE__, __LINE__);          \
 } while (0)
-#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)				\
+#    define ERTS_UNREQ_PROC_MAIN_LOCK(P)				\
 do {									\
     if ((P))								\
 	erts_proc_lc_unrequire_lock((P), ERTS_PROC_LOCK_MAIN);		\
 } while (0)
-#  else
-#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
-#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
-#    define PROCESS_MAIN_CHK_LOCKS(P) erts_lc_check_exact(NULL, 0)
-#  endif
 #else
 #  define PROCESS_MAIN_CHK_LOCKS(P)
-#  define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
-#  define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
+#  define ERTS_REQ_PROC_MAIN_LOCK(P)
+#  define ERTS_UNREQ_PROC_MAIN_LOCK(P)
 #endif
 
 /*
@@ -113,6 +109,8 @@ do {                                     \
 #  define CHECK_ARGS(T)
 #endif
 
+#define CHECK_ALIGNED(Dst) ASSERT((((Uint)&Dst) & (sizeof(Uint)-1)) == 0)
+
 #define GET_BIF_MODULE(p)  (p->info.mfa.module)
 #define GET_BIF_FUNCTION(p)  (p->info.mfa.function)
 #define GET_BIF_ARITY(p)  (p->info.mfa.arity)
@@ -136,11 +134,11 @@ do {                                     \
 
 /* We don't check the range if an ordinary switch is used */
 #ifdef NO_JUMP_TABLE
-#define VALID_INSTR(IP) ((UWord)(IP) < (NUMBER_OF_OPCODES*2+10))
+#  define VALID_INSTR(IP) (BeamCodeAddr(IP) < (NUMBER_OF_OPCODES*2+10))
 #else
-#define VALID_INSTR(IP) \
-   ((SWord)LabelAddr(emulator_loop) <= (SWord)(IP) && \
-    (SWord)(IP) < (SWord)LabelAddr(end_emulator_loop))
+#  define VALID_INSTR(IP) \
+    ((BeamInstr)LabelAddr(emulator_loop) <= BeamCodeAddr(IP) && \
+     BeamCodeAddr(IP) < (BeamInstr)LabelAddr(end_emulator_loop))
 #endif /* NO_JUMP_TABLE */
 
 #define SET_CP(p, ip)           \
@@ -155,50 +153,7 @@ do {                                     \
  * Register target (X or Y register).
  */
 
-#define REG_TARGET_PTR(Target) (((Target) & 1) ? &yb(Target-1) : &xb(Target))
-#define REG_TARGET(Target) (*REG_TARGET_PTR(Target))
-
-/*
- * Store a result into a register given a destination descriptor.
- */
-
-#define StoreResult(Result, DestDesc)		\
-  do {						\
-    Eterm stb_reg;				\
-    stb_reg = (DestDesc);			\
-    CHECK_TERM(Result);				\
-    REG_TARGET(stb_reg) = (Result);		\
-  } while (0)
-
-/*
- * Store a result into a register and execute the next instruction.
- * Dst points to the word with a destination descriptor, which MUST
- * be just before the next instruction.
- */
- 
-#define StoreBifResult(Dst, Result)		\
-  do {						\
-    BeamInstr* stb_next;			\
-    Eterm stb_reg;				\
-    stb_reg = Arg(Dst);				\
-    I += (Dst) + 2;				\
-    stb_next = (BeamInstr *) *I;		\
-    CHECK_TERM(Result);				\
-    REG_TARGET(stb_reg) = (Result);		\
-    Goto(stb_next);				\
-  } while (0)
-
-#define ClauseFail() goto jump_f
-
-#define SAVE_CP(X)				\
-   do {						\
-      *(X) = make_cp(c_p->cp);			\
-      c_p->cp = 0;				\
-   } while(0)
-
-#define RESTORE_CP(X)		SET_CP(c_p, (BeamInstr *) cp_val(*(X)))
-
-#define ISCATCHEND(instr) ((Eterm *) *(instr) == OpCode(catch_end_y))
+#define REG_TARGET_PTR(Target) (((Target) & 1) ? &yb((Target)-1) : &xb(Target))
 
 /*
  * Special Beam instructions.
@@ -208,11 +163,6 @@ BeamInstr beam_apply[2];
 BeamInstr beam_exit[1];
 BeamInstr beam_continue_exit[1];
 
-BeamInstr* em_call_error_handler;
-BeamInstr* em_apply_bif;
-BeamInstr* em_call_nif;
-BeamInstr* em_call_bif_e;
-
 
 /* NOTE These should be the only variables containing trace instructions.
 **      Sometimes tests are form the instruction value, and sometimes
@@ -281,163 +231,24 @@ void** beam_ops;
      HEAP_TOP((P)) = HTOP;  						\
      (P)->stop = E;  							\
      PROCESS_MAIN_CHK_LOCKS((P));					\
-     ERTS_SMP_UNREQ_PROC_MAIN_LOCK((P))
+     ERTS_UNREQ_PROC_MAIN_LOCK((P))
 
 #define db(N) (N)
+#define fb(N) ((Sint)(Sint32)(N))
+#define jb(N) ((Sint)(Sint32)(N))
 #define tb(N) (N)
-#define xb(N) (*(Eterm *) (((unsigned char *)reg) + (N)))
-#define yb(N) (*(Eterm *) (((unsigned char *)E) + (N)))
-#define fb(N) (*(double *) (((unsigned char *)&(freg[0].fd)) + (N)))
+#define xb(N) (*ADD_BYTE_OFFSET(reg, N))
+#define yb(N) (*ADD_BYTE_OFFSET(E, N))
+#define Sb(N) (*REG_TARGET_PTR(N))
+#define lb(N) (*(double *) (((unsigned char *)&(freg[0].fd)) + (N)))
 #define Qb(N) (N)
 #define Ib(N) (N)
+
 #define x(N) reg[N]
 #define y(N) E[N]
 #define r(N) x(N)
-
-/*
- * Makes sure that there are StackNeed + HeapNeed + 1 words available
- * on the combined heap/stack segment, then allocates StackNeed + 1
- * words on the stack and saves CP.
- *
- * M is number of live registers to preserve during garbage collection
- */
-
-#define AH(StackNeed, HeapNeed, M) \
-  do { \
-     int needed; \
-     needed = (StackNeed) + 1; \
-     if (E - HTOP < (needed + (HeapNeed))) { \
-           SWAPOUT; \
-           PROCESS_MAIN_CHK_LOCKS(c_p); \
-           FCALLS -= erts_garbage_collect_nobump(c_p, needed + (HeapNeed), \
-						 reg, (M), FCALLS);	\
-           ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p); \
-           PROCESS_MAIN_CHK_LOCKS(c_p); \
-           SWAPIN; \
-     } \
-     E -= needed; \
-     SAVE_CP(E); \
-  } while (0)
-
-#define Allocate(Ns, Live) AH(Ns, 0, Live)
-
-#define AllocateZero(Ns, Live)             \
- do { Eterm* ptr;                          \
-      int i = (Ns);                        \
-      AH(i, 0, Live);                      \
-      for (ptr = E + i; ptr > E; ptr--) {  \
-	 make_blank(*ptr);                 \
-     }                                     \
-  } while (0)
-
-#define AllocateHeap(Ns, Nh, Live) AH(Ns, Nh, Live)
-
-#define AllocateHeapZero(Ns, Nh, Live)     \
- do { Eterm* ptr;                          \
-      int i = (Ns);                        \
-      AH(i, Nh, Live);                     \
-      for (ptr = E + i; ptr > E; ptr--) {  \
-	 make_blank(*ptr);                 \
-     }                                     \
-  } while (0)
-
-#define AllocateInit(Ns, Live, Y) \
-  do { AH(Ns, 0, Live); make_blank(Y); } while (0)
-
-/*
- * Like the AH macro, but allocates no additional heap space.
- */
-
-#define A(StackNeed, M) AH(StackNeed, 0, M)
-
-#define D(N)             \
-     RESTORE_CP(E);      \
-     E += (N) + 1;
-
-
-
-#define TestBinVHeap(VNh, Nh, Live)                             		\
-  do {                                                          		\
-    unsigned need = (Nh);                                       		\
-    if ((E - HTOP < need) || (MSO(c_p).overhead + (VNh) >= BIN_VHEAP_SZ(c_p))) {\
-       SWAPOUT;                                                 		\
-       PROCESS_MAIN_CHK_LOCKS(c_p);                             		\
-       FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, (Live), FCALLS);	\
-       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);					\
-       PROCESS_MAIN_CHK_LOCKS(c_p);                             		\
-       SWAPIN;                                                  		\
-    }                                                           		\
-    HEAP_SPACE_VERIFIED(need);                                                  \
-  } while (0)
-
-
-
-/*
- * Check if Nh words of heap are available; if not, do a garbage collection.
- * Live is number of active argument registers to be preserved.
- */
-
-#define TestHeap(Nh, Live)                                      \
-  do {                                                          \
-    unsigned need = (Nh);                                       \
-    if (E - HTOP < need) {                                      \
-       SWAPOUT;                                                 \
-       PROCESS_MAIN_CHK_LOCKS(c_p);                             \
-       FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, (Live), FCALLS); \
-       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);			\
-       PROCESS_MAIN_CHK_LOCKS(c_p);                             \
-       SWAPIN;                                                  \
-    }                                                           \
-    HEAP_SPACE_VERIFIED(need);                             \
-  } while (0)
-
-/*
- * Check if Nh words of heap are available; if not, do a garbage collection.
- * Live is number of active argument registers to be preserved.
- * Takes special care to preserve Extra if a garbage collection occurs.
- */
-
-#define TestHeapPreserve(Nh, Live, Extra)				\
-  do {									\
-    unsigned need = (Nh);						\
-    if (E - HTOP < need) {						\
-       SWAPOUT;								\
-       reg[Live] = Extra;						\
-       PROCESS_MAIN_CHK_LOCKS(c_p);					\
-       FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, (Live)+1, FCALLS); \
-       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);				\
-       PROCESS_MAIN_CHK_LOCKS(c_p);					\
-       Extra = reg[Live];						\
-       SWAPIN;								\
-    }									\
-    HEAP_SPACE_VERIFIED(need);                                      \
-  } while (0)
-
-#define TestHeapPutList(Need, Reg)		\
-  do {						\
-     TestHeap((Need), 1);			\
-     PutList(Reg, r(0), r(0));                  \
-     CHECK_TERM(r(0));				\
-  } while (0)
-
-#define Init(N) make_blank(yb(N))
-
-#define Init2(Y1, Y2) do { make_blank(Y1); make_blank(Y2); } while (0)
-#define Init3(Y1, Y2, Y3) \
-   do { make_blank(Y1); make_blank(Y2); make_blank(Y3); } while (0)
-
-#define MakeFun(FunP, NumFree)					\
-  do {								\
-     HEAVY_SWAPOUT;						\
-     r(0) = new_fun(c_p, reg, (ErlFunEntry *) FunP, NumFree);	\
-     HEAVY_SWAPIN;						\
-  } while (0)
-
-#define PutTuple(Dst, Arity)			\
- do {						\
-   Dst = make_tuple(HTOP);			\
-   pt_arity = (Arity);				\
- } while (0)
+#define Q(N) (N*sizeof(Eterm *))
+#define l(N) (freg[N].fd)
 
 /*
  * Check that we haven't used the reductions and jump to function pointed to by
@@ -446,8 +257,8 @@ void** beam_ops;
 
 #define DispatchMacro()				\
   do {						\
-     BeamInstr* dis_next;				\
-     dis_next = (BeamInstr *) *I;			\
+     BeamInstr dis_next;                        \
+     dis_next = *I;                             \
      CHECK_ARGS(I);				\
      if (FCALLS > 0 || FCALLS > neg_o_reds) {	\
         FCALLS--;				\
@@ -455,12 +266,12 @@ void** beam_ops;
      } else {					\
 	goto context_switch;			\
      }						\
- } while (0)
+ } while (0)                                    \
 
 #define DispatchMacroFun()			\
   do {						\
-     BeamInstr* dis_next;				\
-     dis_next = (BeamInstr *) *I;			\
+     BeamInstr dis_next;                        \
+     dis_next = *I;                             \
      CHECK_ARGS(I);				\
      if (FCALLS > 0 || FCALLS > neg_o_reds) {	\
         FCALLS--;				\
@@ -470,23 +281,23 @@ void** beam_ops;
      }						\
  } while (0)
 
-#define DispatchMacrox()					\
-  do {								\
-     if (FCALLS > 0) {						\
-        Eterm* dis_next;					\
-        SET_I(((Export *) Arg(0))->addressv[erts_active_code_ix()]); \
-        dis_next = (Eterm *) *I;				\
-        FCALLS--;						\
-        CHECK_ARGS(I);						\
-        Goto(dis_next);						\
-     } else if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)		\
-		&& FCALLS > neg_o_reds) {			\
-        goto save_calls1;					\
-     } else {							\
-        SET_I(((Export *) Arg(0))->addressv[erts_active_code_ix()]); \
-        CHECK_ARGS(I);						\
-	goto context_switch;					\
-     }								\
+#define DispatchMacrox()                                                \
+  do {                                                                  \
+     if (FCALLS > 0) {                                                  \
+        BeamInstr dis_next;                                             \
+        SET_I(((Export *) Arg(0))->addressv[erts_active_code_ix()]);    \
+        dis_next = *I;                                                  \
+        FCALLS--;                                                       \
+        CHECK_ARGS(I);                                                  \
+        Goto(dis_next);                                                 \
+     } else if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)                      \
+		&& FCALLS > neg_o_reds) {                               \
+        goto save_calls1;                                               \
+     } else {                                                           \
+        SET_I(((Export *) Arg(0))->addressv[erts_active_code_ix()]);    \
+        CHECK_ARGS(I);                                                  \
+	goto context_switch;                                            \
+     }                                                                  \
  } while (0)
 
 #ifdef DEBUG
@@ -505,20 +316,7 @@ void** beam_ops;
 #  define Dispatchfun() DispatchMacroFun()
 #endif
 
-#define Self(R) R = c_p->common.id
-#define Node(R) R = erts_this_node->sysname
-
 #define Arg(N)       I[(N)+1]
-#define Next(N)                \
-    I += (N) + 1;              \
-    ASSERT(VALID_INSTR(*I));   \
-    Goto(*I)
-
-#define PreFetch(N, Dst) do { Dst = (BeamInstr *) *(I + N + 1); } while (0)
-#define NextPF(N, Dst)         \
-    I += N + 1;                \
-    ASSERT(VALID_INSTR(Dst));  \
-    Goto(Dst)
 
 #define GetR(pos, tr)				\
    do {						\
@@ -535,97 +333,20 @@ void** beam_ops;
      CHECK_TERM(tr);				\
    } while (0)
 
-#define GetArg1(N, Dst) GetR((N), Dst)
-
-#define GetArg2(N, Dst1, Dst2)     \
-   do {                            \
-     GetR(N, Dst1);                \
-     GetR((N)+1, Dst2);            \
-   } while (0)
-
-#define PutList(H, T, Dst)                      \
-  do {                                          \
-   HTOP[0] = (H); HTOP[1] = (T);                \
-   Dst = make_list(HTOP);                       \
-   HTOP += 2;                                   \
-  } while (0)
-
-#define Swap(R1, R2)				\
-  do {						\
-    Eterm V = R1;				\
-    R1 = R2;					\
-    R2 = V;					\
-  } while (0)
-
-#define SwapTemp(R1, R2, Tmp)			\
-  do {						\
-    Eterm V = R1;				\
-    R1 = R2;					\
-    R2 = Tmp = V;				\
-  } while (0)
-
-#define Move(Src, Dst) Dst = (Src)
-
-#define Move2Par(S1, D1, S2, D2)		\
-  do {						\
-      Eterm V1, V2;				\
-      V1 = (S1); V2 = (S2); D1 = V1; D2 = V2;	\
-  } while (0)
-
-#define MoveShift(Src, SD, D)			\
-  do {						\
-    Eterm V;					\
-    V = Src; D = SD; SD = V;			\
-  } while (0)
-
-#define MoveDup(Src, D1, D2)			\
-  do {						\
-    D1 = D2 = (Src);				\
-  } while (0)
-
-#define Move3(S1, D1, S2, D2, S3, D3) D1 = (S1); D2 = (S2); D3 = (S3)
-
-#define MoveWindow3(S1, S2, S3, D)		\
-  do {						\
-      Eterm xt0, xt1, xt2;			\
-      Eterm *y = &D;				\
-      xt0  = S1;				\
-      xt1  = S2;				\
-      xt2  = S3;				\
-      y[0] = xt0;				\
-      y[1] = xt1;				\
-      y[2] = xt2;				\
- } while (0)
-
-#define MoveWindow4(S1, S2, S3, S4, D)		\
-  do {						\
-      Eterm xt0, xt1, xt2, xt3;			\
-      Eterm *y = &D;				\
-      xt0  = S1;				\
-      xt1  = S2;				\
-      xt2  = S3;				\
-      xt3  = S4;				\
-      y[0] = xt0;				\
-      y[1] = xt1;				\
-      y[2] = xt2;				\
-      y[3] = xt3;				\
- } while (0)
-
-#define MoveWindow5(S1, S2, S3, S4, S5, D)	\
-  do {						\
-      Eterm xt0, xt1, xt2, xt3, xt4;		\
-      Eterm *y = &D;				\
-      xt0  = S1;				\
-      xt1  = S2;				\
-      xt2  = S3;				\
-      xt3  = S4;				\
-      xt4  = S5;				\
-      y[0] = xt0;				\
-      y[1] = xt1;				\
-      y[2] = xt2;				\
-      y[3] = xt3;				\
-      y[4] = xt4;				\
- } while (0)
+#define PUT_TERM_REG(term, desc)		\
+do {						\
+    switch (loader_tag(desc)) {			\
+    case LOADER_X_REG:				\
+	x(loader_x_reg_index(desc)) = (term);	\
+	break;					\
+    case LOADER_Y_REG:				\
+	y(loader_y_reg_index(desc)) = (term);	\
+	break;					\
+    default:					\
+	ASSERT(0);				\
+	break;					\
+    }						\
+} while(0)
 
 #define DispatchReturn                          \
 do {                                            \
@@ -640,409 +361,14 @@ do {                                            \
     }						\
 } while (0)
 
-#define MoveReturn(Src)				\
-    x(0) = (Src);				\
-    I = c_p->cp;				\
-    ASSERT(VALID_INSTR(*c_p->cp));		\
-    c_p->cp = 0;				\
-    CHECK_TERM(r(0));				\
-    DispatchReturn
-
-#define DeallocateReturn(Deallocate)       \
-  do {                                     \
-    int words_to_pop = (Deallocate);       \
-    SET_I((BeamInstr *) cp_val(*E));       \
-    E = ADD_BYTE_OFFSET(E, words_to_pop);  \
-    CHECK_TERM(r(0));                      \
-    DispatchReturn;                        \
-  } while (0)
-
-#define MoveDeallocateReturn(Src, Deallocate)	\
-    x(0) = (Src);				\
-    DeallocateReturn(Deallocate)
-
-#define MoveCall(Src, CallDest, Size)		\
-    x(0) = (Src);				\
-    SET_CP(c_p, I+Size+1);			\
-    SET_I((BeamInstr *) CallDest);		\
-    Dispatch();
-
-#define MoveCallLast(Src, CallDest, Deallocate)	\
-    x(0) = (Src);				\
-    RESTORE_CP(E);				\
-    E = ADD_BYTE_OFFSET(E, (Deallocate));	\
-    SET_I((BeamInstr *) CallDest);		\
-    Dispatch();
-
-#define MoveCallOnly(Src, CallDest)		\
-    x(0) = (Src);				\
-    SET_I((BeamInstr *) CallDest);		\
-    Dispatch();
-
-#define MoveJump(Src)				\
-     r(0) = (Src);				\
-     SET_I((BeamInstr *) Arg(0));		\
-     Goto(*I);
-
-#define GetList(Src, H, T)			\
-  do {						\
-    Eterm* tmp_ptr = list_val(Src);		\
-    Eterm hd, tl;				\
-    hd = CAR(tmp_ptr);				\
-    tl = CDR(tmp_ptr);				\
-    H = hd; T = tl;				\
-  } while (0)
-
-#define GetTupleElement(Src, Element, Dest)		\
-  do {							\
-    Eterm* src;						\
-    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
-    (Dest) = *src;					\
-  } while (0)
-
-#define GetTupleElement2(Src, Element, Dest)		\
-  do {							\
-    Eterm* src;						\
-    Eterm* dst;						\
-    Eterm E1, E2;					\
-    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
-    dst = &(Dest);					\
-    E1 = src[0];					\
-    E2 = src[1];					\
-    dst[0] = E1;					\
-    dst[1] = E2;					\
-  } while (0)
-
-#define GetTupleElement2Y(Src, Element, D1, D2)		\
-  do {							\
-    Eterm* src;						\
-    Eterm E1, E2;					\
-    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
-    E1 = src[0];					\
-    E2 = src[1];					\
-    D1 = E1;						\
-    D2 = E2;						\
-  } while (0)
-
-#define GetTupleElement3(Src, Element, Dest)		\
-  do {							\
-    Eterm* src;						\
-    Eterm* dst;						\
-    Eterm E1, E2, E3;					\
-    src = ADD_BYTE_OFFSET(tuple_val(Src), (Element));	\
-    dst = &(Dest);					\
-    E1 = src[0];					\
-    E2 = src[1];					\
-    E3 = src[2];					\
-    dst[0] = E1;					\
-    dst[1] = E2;					\
-    dst[2] = E3;					\
-  } while (0)
-
-#define EqualImmed(X, Y, Action) if (X != Y) { Action; }
-#define NotEqualImmed(X, Y, Action) if (X == Y) { Action; }
-#define EqualExact(X, Y, Action) if (!EQ(X,Y)) { Action; }
-#define NotEqualExact(X, Y, Action) if (EQ(X,Y)) { Action; }
-#define Equal(X, Y, Action) CMP_EQ_ACTION(X,Y,Action)
-#define NotEqual(X, Y, Action) CMP_NE_ACTION(X,Y,Action)
-#define IsLessThan(X, Y, Action) CMP_LT_ACTION(X,Y,Action)
-#define IsGreaterEqual(X, Y, Action) CMP_GE_ACTION(X,Y,Action)
-
-#define IsFloat(Src, Fail) if (is_not_float(Src)) { Fail; }
-
-#define IsInteger(Src, Fail) if (is_not_integer(Src)) { Fail; }
-
-#define IsNumber(X, Fail) if (is_not_integer(X) && is_not_float(X)) { Fail; }
-
-#define IsAtom(Src, Fail) if (is_not_atom(Src)) { Fail; }
-
-#define IsIntegerAllocate(Src, Need, Alive, Fail)  \
-    if (is_not_integer(Src)) { Fail; }             \
-    A(Need, Alive)
-
-#define IsNil(Src, Fail) if (is_not_nil(Src)) { Fail; }
-
-#define IsList(Src, Fail) if (is_not_list(Src) && is_not_nil(Src)) { Fail; }
-
-#define IsNonemptyList(Src, Fail) if (is_not_list(Src)) { Fail; }
-
-#define IsNonemptyListAllocate(Src, Need, Alive, Fail)  \
-    if (is_not_list(Src)) { Fail; }                     \
-    A(Need, Alive)
-
-#define IsNonemptyListTestHeap(Need, Alive, Fail)  \
-    if (is_not_list(x(0))) { Fail; }			\
-    TestHeap(Need, Alive)
-
-#define IsNonemptyListGetList(Src, H, T, Fail)	\
-    if (is_not_list(Src)) {			\
-         Fail;					\
-    } else {					\
-       Eterm* tmp_ptr = list_val(Src);		\
-       Eterm hd, tl;				\
-       hd = CAR(tmp_ptr);			\
-       tl = CDR(tmp_ptr);			\
-       H = hd; T = tl;				\
-    }
-
-#define IsTuple(X, Action) if (is_not_tuple(X)) Action
-
-#define IsArity(Pointer, Arity, Fail)		\
-    if (*tuple_val(Pointer) != (Arity)) {	\
-        Fail;					\
-    }
-
-#define IsMap(Src, Fail) if (!is_map(Src)) { Fail; }
-
-#define GetMapElement(Src, Key, Dst, Fail)	\
-  do {						\
-     Eterm _res = get_map_element(Src, Key);	\
-     if (is_non_value(_res)) {			\
-        Fail;					\
-     }						\
-     Dst = _res;				\
-  } while (0)
-
-#define GetMapElementHash(Src, Key, Hx, Dst, Fail)	\
-  do {							\
-     Eterm _res = get_map_element_hash(Src, Key, Hx);	\
-     if (is_non_value(_res)) {				\
-        Fail;						\
-     }							\
-     Dst = _res;					\
-  } while (0)
-
-#define IsFunction(X, Action)			\
-  do {						\
-     if ( !(is_any_fun(X)) ) {			\
-          Action;				\
-     }						\
-  } while (0)
-
-#define IsFunction2(F, A, Action)			\
-  do {							\
-     if (erl_is_function(c_p, F, A) != am_true ) {	\
-          Action;					\
-     }							\
-  } while (0)
-
 #ifdef DEBUG
-#define IsTupleOfArity(Src, Arityval, Fail)			\
-  do {								\
-    if (!(is_tuple(Src) && *tuple_val(Src) == Arityval)) {	\
-        Fail;							\
-    }								\
-  } while (0)
-#else
-#define IsTupleOfArity(Src, Arityval, Fail)			\
-  do {								\
-    if (!(is_boxed(Src) && *tuple_val(Src) == Arityval)) {	\
-        Fail;							\
-    }								\
-  } while (0)
-#endif
-
-#define IsTaggedTuple(Src,Arityval,Tag,Fail) \
-  do {                                       \
-    if (!(is_tuple(Src) &&                   \
-         (tuple_val(Src))[0] == Arityval &&  \
-         (tuple_val(Src))[1] == Tag)) {      \
-        Fail;                                \
-    }                                        \
-  } while (0)
-
-#define IsBoolean(X, Fail) if ((X) != am_true && (X) != am_false) { Fail; }
-
-#define IsBinary(Src, Fail) \
- if (is_not_binary(Src) || binary_bitsize(Src) != 0) { Fail; }
-
-#define IsBitstring(Src, Fail) \
-  if (is_not_binary(Src)) { Fail; }
-
-#if defined(ARCH_64)
-#define BsSafeMul(A, B, Fail, Target)		\
-   do { Uint64 _res = (A) * (B);		\
-      if (_res / B != A) { Fail; }		\
-      Target = _res;				\
-   } while (0)
+/* Better static type testing by the C compiler */
+#  define BEAM_IS_TUPLE(Src) is_tuple(Src)
 #else
-#define BsSafeMul(A, B, Fail, Target)			\
-   do { Uint64 _res = (Uint64)(A) * (Uint64)(B);	\
-      if ((_res >> (8*sizeof(Uint))) != 0) { Fail; }	\
-      Target = _res;					\
-   } while (0)
+/* Better performance */
+# define BEAM_IS_TUPLE(Src) is_boxed(Src)
 #endif
 
-#define BsGetFieldSize(Bits, Unit, Fail, Target)	\
-   do {							\
-      Sint _signed_size; Uint _uint_size;		\
-      Uint temp_bits;					\
-      if (is_small(Bits)) {				\
-        _signed_size = signed_val(Bits);		\
-         if (_signed_size < 0) { Fail; }		\
-         _uint_size = (Uint) _signed_size;		\
-      } else {						\
-         if (!term_to_Uint(Bits, &temp_bits)) { Fail; }	\
-         _uint_size = temp_bits;			\
-      }							\
-      BsSafeMul(_uint_size, Unit, Fail, Target);	\
-   } while (0)
-
-#define BsGetUncheckedFieldSize(Bits, Unit, Fail, Target)	\
-   do {								\
-      Sint _signed_size; Uint _uint_size;			\
-      Uint temp_bits;						\
-      if (is_small(Bits)) {					\
-        _signed_size = signed_val(Bits);			\
-         if (_signed_size < 0) { Fail; }			\
-         _uint_size = (Uint) _signed_size;			\
-      } else {							\
-         if (!term_to_Uint(Bits, &temp_bits)) { Fail; }		\
-         _uint_size = (Uint) temp_bits;				\
-      }								\
-      Target = _uint_size * Unit;				\
-   } while (0)
-
-#define BsGetFloat2(Ms, Live, Sz, Flags, Dst, Fail)                     \
- do {									\
-   ErlBinMatchBuffer *_mb;						\
-   Eterm _result; Sint _size;						\
-   if (!is_small(Sz) || (_size = unsigned_val(Sz)) > 64) { Fail; }	\
-   _size *= ((Flags) >> 3);						\
-   TestHeap(FLOAT_SIZE_OBJECT, Live);					\
-   _mb = ms_matchbuffer(Ms);						\
-   LIGHT_SWAPOUT;							\
-   _result = erts_bs_get_float_2(c_p, _size, (Flags), _mb);		\
-   LIGHT_SWAPIN;							\
-   HEAP_SPACE_VERIFIED(0);                                              \
-   if (is_non_value(_result)) { Fail; }					\
-   else { Dst = _result; }                                              \
- } while (0)
-
-#define BsGetBinaryImm_2(Ms, Live, Sz, Flags, Dst, Fail)	\
-  do {								\
-    ErlBinMatchBuffer *_mb;					\
-    Eterm _result;						\
-    TestHeap(heap_bin_size(ERL_ONHEAP_BIN_LIMIT), Live);	\
-    _mb = ms_matchbuffer(Ms);					\
-    LIGHT_SWAPOUT;						\
-    _result = erts_bs_get_binary_2(c_p, (Sz), (Flags), _mb);	\
-    LIGHT_SWAPIN;						\
-    HEAP_SPACE_VERIFIED(0);                                     \
-    if (is_non_value(_result)) { Fail; }			\
-    else { Dst = _result; }                                     \
-  } while (0)
-
-#define BsGetBinary_2(Ms, Live, Sz, Flags, Dst, Fail)           \
-  do {								\
-    ErlBinMatchBuffer *_mb;					\
-    Eterm _result; Uint _size;					\
-    BsGetFieldSize(Sz, ((Flags) >> 3), Fail, _size);		\
-    TestHeap(ERL_SUB_BIN_SIZE, Live);				\
-    _mb = ms_matchbuffer(Ms);					\
-    LIGHT_SWAPOUT;						\
-    _result = erts_bs_get_binary_2(c_p, _size, (Flags), _mb);	\
-    LIGHT_SWAPIN;						\
-    HEAP_SPACE_VERIFIED(0);                                     \
-    if (is_non_value(_result)) { Fail; }			\
-    else { Dst = _result; }                                     \
-  } while (0)
-
-#define BsGetBinaryAll_2(Ms, Live, Unit, Dst, Fail)	\
-  do {                                                  \
-    ErlBinMatchBuffer *_mb;                             \
-    Eterm _result;                                      \
-    TestHeap(ERL_SUB_BIN_SIZE, Live);                   \
-    _mb = ms_matchbuffer(Ms);                           \
-    if (((_mb->size - _mb->offset) % Unit) == 0) {      \
-      LIGHT_SWAPOUT;                                    \
-      _result = erts_bs_get_binary_all_2(c_p, _mb);     \
-      LIGHT_SWAPIN;                                     \
-      HEAP_SPACE_VERIFIED(0);                           \
-      ASSERT(is_value(_result));                        \
-      Dst = _result;					\
-    } else {                                            \
-	HEAP_SPACE_VERIFIED(0);                         \
-	Fail; }                                         \
- } while (0)
-
-#define BsSkipBits2(Ms, Bits, Unit, Fail)			\
- do {								\
-   ErlBinMatchBuffer *_mb;					\
-   size_t new_offset;						\
-   Uint _size;							\
-    _mb = ms_matchbuffer(Ms);					\
-   BsGetFieldSize(Bits, Unit, Fail, _size);			\
-   new_offset = _mb->offset + _size;				\
-   if (new_offset <= _mb->size) { _mb->offset = new_offset; }	\
-   else { Fail; }						\
- } while (0)
-
-#define BsSkipBitsAll2(Ms, Unit, Fail)		\
- do {						\
-    ErlBinMatchBuffer *_mb;			\
-   _mb = ms_matchbuffer(Ms);			\
-   if (((_mb->size - _mb->offset) % Unit) == 0) {_mb->offset = _mb->size; } \
-   else { Fail; }					\
- } while (0)
-
-#define BsSkipBitsImm2(Ms, Bits, Fail)				\
- do {								\
-   ErlBinMatchBuffer *_mb;					\
-   size_t new_offset;						\
-   _mb = ms_matchbuffer(Ms);					\
-   new_offset = _mb->offset + (Bits);				\
-   if (new_offset <= _mb->size) { _mb->offset = new_offset; }	\
-   else { Fail; }						\
- } while (0)
-
-#define NewBsPutIntegerImm(Sz, Flags, Src)					\
- do {									\
-    if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3((Src), (Sz), (Flags)))) { goto badarg; }	\
- } while (0)
-
-#define NewBsPutInteger(Sz, Flags, Src)						\
- do {										\
-    Sint _size;									\
-    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);		\
-    if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3((Src), _size, (Flags))))	\
-       { goto badarg; }								\
- } while (0)
-
-#define NewBsPutFloatImm(Sz, Flags, Src)					\
- do {									\
-    if (!erts_new_bs_put_float(c_p, (Src), (Sz), (Flags))) { goto badarg; }	\
- } while (0)
-
-#define NewBsPutFloat(Sz, Flags, Src)						\
- do {										\
-    Sint _size;									\
-    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);		\
-    if (!erts_new_bs_put_float(c_p, (Src), _size, (Flags))) { goto badarg; }	\
- } while (0)
-
-#define NewBsPutBinary(Sz, Flags, Src)							\
- do {											\
-    Sint _size;										\
-    BsGetUncheckedFieldSize(Sz, ((Flags) >> 3), goto badarg, _size);			\
-    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2((Src), _size))) { goto badarg; }	\
- } while (0)
-
-#define NewBsPutBinaryImm(Sz, Src)				        \
- do {							        \
-    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2((Src), (Sz)))) { goto badarg; }	\
- } while (0)
-
-#define NewBsPutBinaryAll(Src, Unit)							\
- do {											\
-    if (!erts_new_bs_put_binary_all(ERL_BITS_ARGS_2((Src), (Unit)))) { goto badarg; }	\
- } while (0)
-
-
-#define IsPort(Src, Fail) if (is_not_port(Src)) { Fail; }
-#define IsPid(Src, Fail) if (is_not_pid(Src)) { Fail; }
-#define IsRef(Src, Fail) if (is_not_ref(Src)) { Fail; }
-
 /*
  * process_main() is already huge, so we want to avoid inlining
  * into it. Especially functions that are seldom used.
@@ -1058,6 +384,7 @@ do {                                            \
  * The following functions are called directly by process_main().
  * Don't inline them.
  */
+static void init_emulator_finish(void) NOINLINE;
 static ErtsCodeMFA *ubif2mfa(void* uf) NOINLINE;
 static ErtsCodeMFA *gcbif2mfa(void* gcf) NOINLINE;
 static BeamInstr* handle_error(Process* c_p, BeamInstr* pc,
@@ -1066,20 +393,22 @@ static BeamInstr* call_error_handler(Process* p, ErtsCodeMFA* mfa,
 				     Eterm* reg, Eterm func) NOINLINE;
 static BeamInstr* fixed_apply(Process* p, Eterm* reg, Uint arity,
 			      BeamInstr *I, Uint offs) NOINLINE;
-static BeamInstr* apply(Process* p, Eterm module, Eterm function,
-			Eterm args, Eterm* reg,
-			BeamInstr *I, Uint offs) NOINLINE;
+static BeamInstr* apply(Process* p, Eterm* reg,
+                        BeamInstr *I, Uint offs) NOINLINE;
 static BeamInstr* call_fun(Process* p, int arity,
 			   Eterm* reg, Eterm args) NOINLINE;
 static BeamInstr* apply_fun(Process* p, Eterm fun,
 			    Eterm args, Eterm* reg) NOINLINE;
 static Eterm new_fun(Process* p, Eterm* reg,
 		     ErlFunEntry* fe, int num_free) NOINLINE;
-static Eterm new_map(Process* p, Eterm* reg, BeamInstr* I) NOINLINE;
-static Eterm update_map_assoc(Process* p, Eterm* reg,
-			      Eterm map, BeamInstr* I) NOINLINE;
-static Eterm update_map_exact(Process* p, Eterm* reg,
-			      Eterm map, BeamInstr* I) NOINLINE;
+static Eterm erts_gc_new_map(Process* p, Eterm* reg, Uint live,
+                             Uint n, BeamInstr* ptr) NOINLINE;
+static Eterm erts_gc_new_small_map_lit(Process* p, Eterm* reg, Eterm keys_literal,
+                               Uint live, BeamInstr* ptr) NOINLINE;
+static Eterm erts_gc_update_map_assoc(Process* p, Eterm* reg, Uint live,
+                              Uint n, BeamInstr* new_p) NOINLINE;
+static Eterm erts_gc_update_map_exact(Process* p, Eterm* reg, Uint live,
+                              Uint n, Eterm* new_p) NOINLINE;
 static Eterm get_map_element(Eterm map, Eterm key);
 static Eterm get_map_element_hash(Eterm map, Eterm key, Uint32 hx);
 
@@ -1111,6 +440,12 @@ init_emulator(void)
 #  define REG_stop asm("%l3")
 #  define REG_I asm("%l4")
 #  define REG_fcalls asm("%l5")
+#elif defined(__GNUC__) && defined(__amd64__) && !defined(DEBUG)
+#  define REG_xregs asm("%r12")
+#  define REG_htop
+#  define REG_stop asm("%r13")
+#  define REG_I asm("%rbx")
+#  define REG_fcalls asm("%r14")
 #else
 #  define REG_xregs
 #  define REG_htop
@@ -1230,6 +565,13 @@ init_emulator(void)
 #define ERTS_DBG_CHK_REDS(P, FC)
 #endif
 
+#ifdef NO_FPE_SIGNALS
+#  define ERTS_NO_FPE_CHECK_INIT ERTS_FP_CHECK_INIT
+#  define ERTS_NO_FPE_ERROR ERTS_FP_ERROR
+#else
+#  define ERTS_NO_FPE_CHECK_INIT(p)
+#  define ERTS_NO_FPE_ERROR(p, a, b)
+#endif
 
 /*
  * process_main() is called twice:
@@ -1289,12 +631,10 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
 #ifndef NO_JUMP_TABLE
     static void* opcodes[] = { DEFINE_OPCODES };
 #else
-    int Go;
+    register BeamInstr Go;
 #endif
 #endif
 
-    Eterm pt_arity;		/* Used by do_put_tuple */
-
     Uint64 start_time = 0;          /* Monitor long schedule */
     BeamInstr* start_time_i = NULL;
 
@@ -1311,7 +651,7 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
      * Note: c_p->arity must be set to reflect the number of useful terms in
      * c_p->arg_reg before calling the scheduler.
      */
-    if (!init_done) {
+    if (ERTS_UNLIKELY(!init_done)) {
        /* This should only be reached during the init phase when only the main
         * process is running. I.e. there is no race for init_done.
         */
@@ -1344,16 +684,16 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
     }
 
     PROCESS_MAIN_CHK_LOCKS(c_p);
-    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
     ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
     c_p = erts_schedule(NULL, c_p, reds_used);
     ASSERT(!(c_p->flags & F_HIPE_MODE));
     ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
     start_time = 0;
 #ifdef DEBUG
-    pid = c_p->common.id; /* Save for debugging purpouses */
+    pid = c_p->common.id; /* Save for debugging purposes */
 #endif
-    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
     PROCESS_MAIN_CHK_LOCKS(c_p);
 
     ERTS_MSACC_UPDATE_CACHE_X();
@@ -1367,7 +707,7 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
     {
 	int reds;
 	Eterm* argp;
-	BeamInstr *next;
+	BeamInstr next;
 	int i;
 
 	argp = c_p->arg_reg;
@@ -1399,7 +739,7 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
 
 	ERTS_DBG_CHK_REDS(c_p, FCALLS);
 
-	next = (BeamInstr *) *I;
+	next = *I;
 	SWAPIN;
 	ASSERT(VALID_INSTR(next));
 
@@ -1410,7 +750,7 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
             dtrace_proc_str(c_p, process_buf);
 
             if (ERTS_PROC_IS_EXITING(c_p)) {
-                strcpy(fun_buf, "<exiting>");
+                sys_strcpy(fun_buf, "<exiting>");
             } else {
                 ErtsCodeMFA *cmfa = find_function_from_pc(c_p->i);
                 if (cmfa) {
@@ -1435,1923 +775,8 @@ void process_main(Eterm * x_reg_array, FloatDef* f_reg_array)
 #ifdef NO_JUMP_TABLE
     switch (Go) {
 #endif
-#include "beam_hot.h"
-
-	{
-	    Eterm increment_reg_val;
-	    Eterm increment_val;
-	    Uint live;
-	    Eterm result;
-
-	OpCase(i_increment_rIId):
-	    increment_reg_val = x(0);
-	    I--;
-	    goto do_increment;
-
-	OpCase(i_increment_xIId):
-	    increment_reg_val = xb(Arg(0));
-	    goto do_increment;
-
-	OpCase(i_increment_yIId):
-	    increment_reg_val = yb(Arg(0));
-	    goto do_increment;
-
-	do_increment:
-	    increment_val = Arg(1);
-	    if (is_small(increment_reg_val)) {
-		Sint i = signed_val(increment_reg_val) + increment_val;
-		ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
-		if (MY_IS_SSMALL(i)) {
-		    result = make_small(i);
-		    StoreBifResult(3, result);
-		}
-	    }
 
-	    live = Arg(2);
-	    HEAVY_SWAPOUT;
-	    reg[live] = increment_reg_val;
-	    reg[live+1] = make_small(increment_val);
-	    result = erts_gc_mixed_plus(c_p, reg, live);
-	    HEAVY_SWAPIN;
-	    ERTS_HOLE_CHECK(c_p);
-	    if (is_value(result)) {
-		StoreBifResult(3, result);
-	    }
-	    ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
-	    goto find_func_info;
-	}
-
-#define DO_OUTLINED_ARITH_2(name, Op1, Op2)	\
- do {						\
-     Eterm result;				\
-     Uint live = Arg(1);			\
-						\
-     HEAVY_SWAPOUT;				\
-     reg[live] = Op1;				\
-     reg[live+1] = Op2;				\
-     result = erts_gc_##name(c_p, reg, live);	\
-     HEAVY_SWAPIN;				\
-     ERTS_HOLE_CHECK(c_p);			\
-     if (is_value(result)) {			\
-	 StoreBifResult(4, result);		\
-     }						\
-     goto lb_Cl_error;				\
- } while (0)
-
- {
-     Eterm PlusOp1, PlusOp2;
-     Eterm result;
-
- OpCase(i_plus_jIxxd):
-     PlusOp1 = xb(Arg(2));
-     PlusOp2 = xb(Arg(3));
-     goto do_plus;
-
- OpCase(i_plus_jIxyd):
-     PlusOp1 = xb(Arg(2));
-     PlusOp2 = yb(Arg(3));
-     goto do_plus;
-
- OpCase(i_plus_jIssd):
-     GetArg2(2, PlusOp1, PlusOp2);
-     goto do_plus;
-
- do_plus:
-     if (is_both_small(PlusOp1, PlusOp2)) {
-	 Sint i = signed_val(PlusOp1) + signed_val(PlusOp2);
-	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
-	 if (MY_IS_SSMALL(i)) {
-	     result = make_small(i);
-             StoreBifResult(4, result);
-	 }
-     }
-     DO_OUTLINED_ARITH_2(mixed_plus, PlusOp1, PlusOp2);
- }
-
- {
-     Eterm MinusOp1, MinusOp2;
-     Eterm result;
-
- OpCase(i_minus_jIxxd):
-     MinusOp1 = xb(Arg(2));
-     MinusOp2 = xb(Arg(3));
-     goto do_minus;
-
- OpCase(i_minus_jIssd):
-     GetArg2(2, MinusOp1, MinusOp2);
-     goto do_minus;
-
- do_minus:
-     if (is_both_small(MinusOp1, MinusOp2)) {
-	 Sint i = signed_val(MinusOp1) - signed_val(MinusOp2);
-	 ASSERT(MY_IS_SSMALL(i) == IS_SSMALL(i));
-	 if (MY_IS_SSMALL(i)) {
-	     result = make_small(i);
-             StoreBifResult(4, result);
-	 }
-     }
-     DO_OUTLINED_ARITH_2(mixed_minus, MinusOp1, MinusOp2);
- }
-
-    {
-	Eterm is_eq_exact_lit_val;
-
-    OpCase(i_is_eq_exact_literal_fxc):
-	is_eq_exact_lit_val = xb(Arg(1));
-	goto do_is_eq_exact_literal;
-
-    OpCase(i_is_eq_exact_literal_fyc):
-	is_eq_exact_lit_val = yb(Arg(1));
-	goto do_is_eq_exact_literal;
-
-    do_is_eq_exact_literal:
-	if (!eq(Arg(2), is_eq_exact_lit_val)) {
-	    ClauseFail();
-	}
-	Next(3);
-    }
-
-    {
-	Eterm is_ne_exact_lit_val;
-
-    OpCase(i_is_ne_exact_literal_fxc):
-	is_ne_exact_lit_val = xb(Arg(1));
-	goto do_is_ne_exact_literal;
-
-    OpCase(i_is_ne_exact_literal_fyc):
-	is_ne_exact_lit_val = yb(Arg(1));
-	goto do_is_ne_exact_literal;
-
-    do_is_ne_exact_literal:
-	if (eq(Arg(2), is_ne_exact_lit_val)) {
-	    ClauseFail();
-	}
-	Next(3);
-    }
-
- OpCase(i_move_call_only_fc): {
-     r(0) = Arg(1);
- }
- /* FALL THROUGH */
- OpCase(i_call_only_f): {
-     SET_I((BeamInstr *) Arg(0));
-     DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
-     Dispatch();
- }
-
- OpCase(i_move_call_last_fPc): {
-     r(0) = Arg(2);
- }
- /* FALL THROUGH */
- OpCase(i_call_last_fP): {
-     RESTORE_CP(E);
-     E = ADD_BYTE_OFFSET(E, Arg(1));
-     SET_I((BeamInstr *) Arg(0));
-     DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
-     Dispatch();
- }
-
- OpCase(i_move_call_cf): {
-     r(0) = Arg(0);
-     I++;
- }
- /* FALL THROUGH */
- OpCase(i_call_f): {
-     SET_CP(c_p, I+2);
-     SET_I((BeamInstr *) Arg(0));
-     DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
-     Dispatch();
- }
-
- OpCase(i_move_call_ext_last_ePc): {
-     r(0) = Arg(2);
- }
- /* FALL THROUGH */
- OpCase(i_call_ext_last_eP):
-    RESTORE_CP(E);
-    E = ADD_BYTE_OFFSET(E, Arg(1));
-
-    /*
-     * Note: The pointer to the export entry is never NULL; if the module
-     * is not loaded, it points to code which will invoke the error handler
-     * (see lb_call_error_handler below).
-     */
-    DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, Arg(0));
-    Dispatchx();
-
- OpCase(i_move_call_ext_ce): {
-     r(0) = Arg(0);
-     I++;
- }
- /* FALL THROUGH */
- OpCase(i_call_ext_e):
-    SET_CP(c_p, I+2);
-    DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, Arg(0));
-    Dispatchx();
-
- OpCase(i_move_call_ext_only_ec): {
-     r(0) = Arg(1);
- }
- /* FALL THROUGH */
- OpCase(i_call_ext_only_e):
-    DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, Arg(0));
-    Dispatchx();
-
- OpCase(init_y): {
-     BeamInstr *next;
-
-     PreFetch(1, next);
-     make_blank(yb(Arg(0)));
-     NextPF(1, next);
- }
-
- OpCase(i_trim_I): {
-     BeamInstr *next;
-     Uint words;
-     Uint cp;
-
-     words = Arg(0);
-     cp = E[0];
-     PreFetch(1, next);
-     E += words;
-     E[0] = cp;
-     NextPF(1, next);
- }
-
- OpCase(move_x1_c): {
-	x(1) = Arg(0);
-	Next(1);
-    }
-
- OpCase(move_x2_c): {
-	x(2) = Arg(0);
-	Next(1);
-    }
-
- OpCase(return): {
-    SET_I(c_p->cp);
-    DTRACE_RETURN_FROM_PC(c_p);
-    /*
-     * We must clear the CP to make sure that a stale value do not
-     * create a false module dependcy preventing code upgrading.
-     * It also means that we can use the CP in stack backtraces.
-     */
-    c_p->cp = 0;
-    CHECK_TERM(r(0));
-    HEAP_SPACE_VERIFIED(0);
-    DispatchReturn;
- }
-
-    /*
-     * Send is almost a standard call-BIF with two arguments, except for:
-     *    1) It cannot be traced.
-     *	  2) There is no pointer to the send_2 function stored in
-     *       the instruction.
-     */
-
- OpCase(send): {
-     BeamInstr *next;
-     Eterm result;
-
-     if (!(FCALLS > 0 || FCALLS > neg_o_reds)) {
-         /* If we have run out of reductions, we do a context
-            switch before calling the bif */
-         c_p->arity = 2;
-         c_p->current = NULL;
-         goto context_switch3;
-     }
-
-     PRE_BIF_SWAPOUT(c_p);
-     c_p->fcalls = FCALLS - 1;
-     result = erl_send(c_p, r(0), x(1));
-     PreFetch(0, next);
-     ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-     ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-     PROCESS_MAIN_CHK_LOCKS(c_p);
-     HTOP = HEAP_TOP(c_p);
-     FCALLS = c_p->fcalls;
-     if (is_value(result)) {
-	 r(0) = result;
-	 CHECK_TERM(r(0));
-	 NextPF(0, next);
-     } else if (c_p->freason == TRAP) {
-	 SET_CP(c_p, I+1);
-	 SET_I(c_p->i);
-	 SWAPIN;
-	 Dispatch();
-     }
-     goto find_func_info;
- }
-
-    {
-	Eterm element_index;
-	Eterm element_tuple;
-
-    OpCase(i_element_jxsd):
-	element_tuple = xb(Arg(1));
-	goto do_element;
-
-    OpCase(i_element_jysd):
-	element_tuple = yb(Arg(1));
-	goto do_element;
-
-    do_element:
-	GetArg1(2, element_index);
-	if (is_small(element_index) && is_tuple(element_tuple)) {
-	    Eterm* tp = tuple_val(element_tuple);
-
-	    if ((signed_val(element_index) >= 1) &&
-		(signed_val(element_index) <= arityval(*tp))) {
-		Eterm result = tp[signed_val(element_index)];
-		StoreBifResult(3, result);
-	    }
-	}
-    }
- /* Fall through */
-
- OpCase(badarg_j):
- badarg:
-    c_p->freason = BADARG;
-    goto lb_Cl_error;
-
-    {
-	Eterm fast_element_tuple;
-
-    OpCase(i_fast_element_jxId):
-     fast_element_tuple = xb(Arg(1));
-     goto do_fast_element;
-
-    OpCase(i_fast_element_jyId):
-     fast_element_tuple = yb(Arg(1));
-     goto do_fast_element;
-
-    do_fast_element:
-	if (is_tuple(fast_element_tuple)) {
-	    Eterm* tp = tuple_val(fast_element_tuple);
-	    Eterm pos = Arg(2);	/* Untagged integer >= 1 */
-	    if (pos <= arityval(*tp)) {
-		Eterm result = tp[pos];
-		StoreBifResult(3, result);
-	    }
-	}
-     goto badarg;
- }
-
- OpCase(catch_yf):
-     c_p->catches++;
-     yb(Arg(0)) = Arg(1);
-     Next(2);
-
- OpCase(catch_end_y): {
-     c_p->catches--;
-     make_blank(yb(Arg(0)));
-     if (is_non_value(r(0))) {
-         c_p->fvalue = NIL;
-	 if (x(1) == am_throw) {
-	     r(0) = x(2);
-	 } else {
-	     if (x(1) == am_error) {
-	         SWAPOUT;
-		 x(2) = add_stacktrace(c_p, x(2), x(3));
-		 SWAPIN;
-	     }
-	     /* only x(2) is included in the rootset here */
-	     if (E - HTOP < 3) {
-		 SWAPOUT;
-		 PROCESS_MAIN_CHK_LOCKS(c_p);
-		 FCALLS -= erts_garbage_collect_nobump(c_p, 3, reg+2, 1, FCALLS);
-		 ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-		 PROCESS_MAIN_CHK_LOCKS(c_p);
-		 SWAPIN;
-	     }
-	     r(0) = TUPLE2(HTOP, am_EXIT, x(2));
-	     HTOP += 3;
-	 }
-     }
-     CHECK_TERM(r(0));
-     Next(1);
- }
-
- OpCase(try_end_y): {
-     c_p->catches--;
-     make_blank(yb(Arg(0)));
-     if (is_non_value(r(0))) {
-         c_p->fvalue = NIL;
-	 r(0) = x(1);
-	 x(1) = x(2);
-	 x(2) = x(3);
-     }
-     Next(1);
- }
-
- /*
-  * Skeleton for receive statement:
-  *
-  *             recv_mark L1                     Optional
-  *             call make_ref/monitor            Optional
-  *             ...
-  *             recv_set L1                      Optional
-  *      L1:          <-------------------+
-  *                   <-----------+       |
-  *     	     	       	  |   	  |
-  *             loop_rec L2 ------+---+   |
-  *             ...               |   |   |
-  *             remove_message 	  |   |	  |
-  *             jump L3           |   |   |
-  *		...	          |   |   |
-  *		loop_rec_end L1 --+   |   |
-  *      L2:          <---------------+   |
-  *	   	wait L1  -----------------+      or wait_timeout
-  *		timeout
-  *
-  *	 L3:    Code after receive...
-  *
-  *
-  */
-
- OpCase(recv_mark_f): {
-     /*
-      * Save the current position in message buffer and the
-      * the label for the loop_rec/2 instruction for the
-      * the receive statement.
-      */
-     c_p->msg.mark = (BeamInstr *) Arg(0);
-     c_p->msg.saved_last = c_p->msg.last;
-     Next(1);
- }
-
- OpCase(i_recv_set): {
-     /*
-      * If the mark is valid (points to the loop_rec/2
-      * instruction that follows), we know that the saved
-      * position points to the first message that could
-      * possibly be matched out.
-      *
-      * If the mark is invalid, we do nothing, meaning that
-      * we will look through all messages in the message queue.
-      */
-     if (c_p->msg.mark == (BeamInstr *) (I+1)) {
-	 c_p->msg.save = c_p->msg.saved_last;
-     }
-     I++;
-     /* Fall through to the loop_rec/2 instruction */
- }
-
-    /*
-     * Pick up the next message and place it in x(0).
-     * If no message, jump to a wait or wait_timeout instruction.
-     */
- OpCase(i_loop_rec_f):
- {
-     BeamInstr *next;
-     ErtsMessage* msgp;
-
-     /*
-      * We need to disable GC while matching messages
-      * in the queue. This since messages with data outside
-      * the heap will be corrupted by a GC.
-      */
-     ASSERT(!(c_p->flags & F_DELAY_GC));
-     c_p->flags |= F_DELAY_GC;
-
- loop_rec__:
-
-     PROCESS_MAIN_CHK_LOCKS(c_p);
-
-     msgp = PEEK_MESSAGE(c_p);
-
-     if (!msgp) {
-#ifdef ERTS_SMP
-	 erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	 /* Make sure messages wont pass exit signals... */
-	 if (ERTS_PROC_PENDING_EXIT(c_p)) {
-	     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	     SWAPOUT;
-	     c_p->flags &= ~F_DELAY_GC;
-             c_p->arity = 0;
-	     goto do_schedule; /* Will be rescheduled for exit */
-	 }
-	 ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	 msgp = PEEK_MESSAGE(c_p);
-	 if (msgp)
-	     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	 else
-#endif
-	 {
-	     c_p->flags &= ~F_DELAY_GC;
-	     SET_I((BeamInstr *) Arg(0));
-	     Goto(*I);		/* Jump to a wait or wait_timeout instruction */
-	 }
-     }
-     if (is_non_value(ERL_MESSAGE_TERM(msgp))) {
-	 SWAPOUT; /* erts_decode_dist_message() may write to heap... */
-	 if (!erts_decode_dist_message(c_p, ERTS_PROC_LOCK_MAIN, msgp, 0)) {
-	     /*
-	      * A corrupt distribution message that we weren't able to decode;
-	      * remove it...
-	      */
-	     /* No swapin should be needed */
-	     ASSERT(HTOP == c_p->htop && E == c_p->stop);
-	     /* TODO: Add DTrace probe for this bad message situation? */
-	     UNLINK_MESSAGE(c_p, msgp);
-	     msgp->next = NULL;
-	     erts_cleanup_messages(msgp);
-	     goto loop_rec__;
-	 }
-	 SWAPIN;
-     }
-     PreFetch(1, next);
-     r(0) = ERL_MESSAGE_TERM(msgp);
-     NextPF(1, next);
- }
-
- /*
-  * Remove a (matched) message from the message queue.
-  */
- OpCase(remove_message): {
-     BeamInstr *next;
-     ErtsMessage* msgp;
-     PROCESS_MAIN_CHK_LOCKS(c_p);
-
-     ERTS_CHK_MBUF_SZ(c_p);
-
-     PreFetch(0, next);
-     msgp = PEEK_MESSAGE(c_p);
-
-     if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)) {
-	 save_calls(c_p, &exp_receive);
-     }
-     if (ERL_MESSAGE_TOKEN(msgp) == NIL) {
-#ifdef USE_VM_PROBES
-	 if (DT_UTAG(c_p) != NIL) {
-	     if (DT_UTAG_FLAGS(c_p) & DT_UTAG_PERMANENT) {
-		 SEQ_TRACE_TOKEN(c_p) = am_have_dt_utag;
-	     } else {
-		 DT_UTAG(c_p) = NIL;
-		 SEQ_TRACE_TOKEN(c_p) = NIL;
-	     }
-	 } else {
-#endif
-	     SEQ_TRACE_TOKEN(c_p) = NIL;
-#ifdef USE_VM_PROBES
-	 }
-	 DT_UTAG_FLAGS(c_p) &= ~DT_UTAG_SPREADING;
-#endif
-     } else if (ERL_MESSAGE_TOKEN(msgp) != am_undefined) {
-	 Eterm msg;
-	 SEQ_TRACE_TOKEN(c_p) = ERL_MESSAGE_TOKEN(msgp);
-#ifdef USE_VM_PROBES
-	 if (ERL_MESSAGE_TOKEN(msgp) == am_have_dt_utag) {
-	     if (DT_UTAG(c_p) == NIL) {
-		 DT_UTAG(c_p) = ERL_MESSAGE_DT_UTAG(msgp);
-	     }
-	     DT_UTAG_FLAGS(c_p) |= DT_UTAG_SPREADING;
-	 } else {
-#endif
-	     ASSERT(is_tuple(SEQ_TRACE_TOKEN(c_p)));
-	     ASSERT(SEQ_TRACE_TOKEN_ARITY(c_p) == 5);
-	     ASSERT(is_small(SEQ_TRACE_TOKEN_SERIAL(c_p)));
-	     ASSERT(is_small(SEQ_TRACE_TOKEN_LASTCNT(c_p)));
-	     ASSERT(is_small(SEQ_TRACE_TOKEN_FLAGS(c_p)));
-	     ASSERT(is_pid(SEQ_TRACE_TOKEN_SENDER(c_p)));
-	     c_p->seq_trace_lastcnt = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
-	     if (c_p->seq_trace_clock < unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p))) {
-		 c_p->seq_trace_clock = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
-	     }
-	     msg = ERL_MESSAGE_TERM(msgp);
-	     seq_trace_output(SEQ_TRACE_TOKEN(c_p), msg, SEQ_TRACE_RECEIVE, 
-			      c_p->common.id, c_p);
-#ifdef USE_VM_PROBES
-	 }
-#endif
-     }
-#ifdef USE_VM_PROBES
-     if (DTRACE_ENABLED(message_receive)) {
-         Eterm token2 = NIL;
-         DTRACE_CHARBUF(receiver_name, DTRACE_TERM_BUF_SIZE);
-         Sint tok_label = 0;
-         Sint tok_lastcnt = 0;
-         Sint tok_serial = 0;
-
-         dtrace_proc_str(c_p, receiver_name);
-         token2 = SEQ_TRACE_TOKEN(c_p);
-         if (have_seqtrace(token2)) {
-             tok_label = signed_val(SEQ_TRACE_T_LABEL(token2));
-             tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token2));
-             tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token2));
-         }
-         DTRACE6(message_receive,
-                 receiver_name, size_object(ERL_MESSAGE_TERM(msgp)),
-                 c_p->msg.len - 1, tok_label, tok_lastcnt, tok_serial);
-     }
-#endif
-     UNLINK_MESSAGE(c_p, msgp);
-     JOIN_MESSAGE(c_p);
-     CANCEL_TIMER(c_p);
-
-     erts_save_message_in_proc(c_p, msgp);
-     c_p->flags &= ~F_DELAY_GC;
-
-     if (ERTS_IS_GC_DESIRED_INTERNAL(c_p, HTOP, E)) {
-	 /*
-	  * We want to GC soon but we leave a few
-	  * reductions giving the message some time
-	  * to turn into garbage.
-	  */
-	 ERTS_VBUMP_LEAVE_REDS_INTERNAL(c_p, 5, FCALLS);
-     }
-
-     ERTS_DBG_CHK_REDS(c_p, FCALLS);
-     ERTS_CHK_MBUF_SZ(c_p);
-
-     ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-     PROCESS_MAIN_CHK_LOCKS(c_p);
-     NextPF(0, next);
- }
-
-    /*
-     * Advance the save pointer to the next message (the current
-     * message didn't match), then jump to the loop_rec instruction.
-     */
- OpCase(loop_rec_end_f): {
-
-     ASSERT(c_p->flags & F_DELAY_GC);
-
-     SET_I((BeamInstr *) Arg(0));
-     SAVE_MESSAGE(c_p);
-     if (FCALLS > 0 || FCALLS > neg_o_reds) {
-	 FCALLS--;
-	 goto loop_rec__;
-     }
-
-     c_p->flags &= ~F_DELAY_GC;
-     c_p->i = I;
-     SWAPOUT;
-     c_p->arity = 0;
-     c_p->current = NULL;
-     goto do_schedule;
- }
-    /*
-     * Prepare to wait for a message or a timeout, whichever occurs first.
-     *
-     * Note: In order to keep the compatibility between 32 and 64 bits
-     * emulators, only timeout values that can be represented in 32 bits
-     * (unsigned) or less are allowed.
-     */
-
-
- OpCase(i_wait_timeout_fs): {
-     erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-
-     /* Fall through */
- }
- OpCase(i_wait_timeout_locked_fs): {
-     Eterm timeout_value;
-
-     /*
-      * If we have already set the timer, we must NOT set it again.  Therefore,
-      * we must test the F_INSLPQUEUE flag as well as the F_TIMO flag.
-      */
-     if (c_p->flags & (F_INSLPQUEUE | F_TIMO)) {
-	 goto wait2;
-     }
-     GetArg1(1, timeout_value);
-     if (timeout_value != make_small(0)) {
-
-	 if (timeout_value == am_infinity)
-	     c_p->flags |= F_TIMO;
-	 else {
-	     int tres = erts_set_proc_timer_term(c_p, timeout_value);
-	     if (tres == 0) {
-		 /*
-		  * The timer routiner will set c_p->i to the value in
-		  * c_p->def_arg_reg[0].  Note that it is safe to use this
-		  * location because there are no living x registers in
-		  * a receive statement.
-		  */
-		 BeamInstr** pi = (BeamInstr**) c_p->def_arg_reg;
-		 *pi = I+3;
-	     }
-	     else { /* Wrong time */
-	     OpCase(i_wait_error_locked): {
-		     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-		     /* Fall through */
-		 }
-	     OpCase(i_wait_error): {
-		     c_p->freason = EXC_TIMEOUT_VALUE;
-		     goto find_func_info;
-		 }
-	     }
-	 }
-
-	 /*
-	  * Prepare to wait indefinitely for a new message to arrive
-	  * (or the time set above if falling through from above).
-	  *
-	  * When a new message arrives, control will be transferred
-	  * the loop_rec instruction (at label L1).  In case of
-	  * of timeout, control will be transferred to the timeout
-	  * instruction following the wait_timeout instruction.
-	  */
-
-	 OpCase(wait_locked_f):
-	 OpCase(wait_f):
-
-	 wait2: {
-#ifndef ERTS_SMP
-	     if (ERTS_PROC_IS_EXITING(c_p)) {
-		 /*
-		  * I non smp case:
-		  *
-		  * Currently executing process might be sent an exit
-		  * signal if it is traced by a port that it also is
-		  * linked to, and the port terminates during the
-		  * trace. In this case we do *not* want to clear
-		  * the active flag, which will make the process hang
-		  * in limbo forever.
-		  */
-		 SWAPOUT;
-                 c_p->arity = 0;
-		 goto do_schedule;
-	     }
-#endif
-	     c_p->i = (BeamInstr *) Arg(0); /* L1 */
-	     SWAPOUT;
-	     c_p->arity = 0;
-
-	     if (!ERTS_PTMR_IS_TIMED_OUT(c_p))
-		 erts_smp_atomic32_read_band_relb(&c_p->state,
-						  ~ERTS_PSFLG_ACTIVE);
-	     ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	     c_p->current = NULL;
-	     goto do_schedule;
-	 }
-	 OpCase(wait_unlocked_f): {
-	     erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	     goto wait2;
-	 }
-     }
-     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-     Next(2);
- }
-
- OpCase(i_wait_timeout_fI): {
-     erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
- }
-
- OpCase(i_wait_timeout_locked_fI):
- {
-     /*
-      * If we have already set the timer, we must NOT set it again.  Therefore,
-      * we must test the F_INSLPQUEUE flag as well as the F_TIMO flag.
-      */
-     if ((c_p->flags & (F_INSLPQUEUE | F_TIMO)) == 0) {
-	 BeamInstr** p = (BeamInstr **) c_p->def_arg_reg;
-	 *p = I+3;
-	 erts_set_proc_timer_uword(c_p, Arg(1));
-     }
-     goto wait2;
- }
-
-    /*
-     * A timeout has occurred.  Reset the save pointer so that the next
-     * receive statement will examine the first message first.
-     */
- OpCase(timeout_locked): {
-     erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
- }
-
- OpCase(timeout): {
-     BeamInstr *next;
-
-     PreFetch(0, next);
-     if (IS_TRACED_FL(c_p, F_TRACE_RECEIVE)) {
-	 trace_receive(c_p, am_clock_service, am_timeout, NULL);
-     }
-     if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)) {
-	 save_calls(c_p, &exp_timeout);
-     }
-     c_p->flags &= ~F_TIMO;
-     JOIN_MESSAGE(c_p);
-     NextPF(0, next);
- }
-
- {
-     Eterm select_val2;
-
- OpCase(i_select_tuple_arity2_yfAAff):
-     select_val2 = yb(Arg(0));
-     goto do_select_tuple_arity2;
-
- OpCase(i_select_tuple_arity2_xfAAff):
-     select_val2 = xb(Arg(0));
-     goto do_select_tuple_arity2;
-
- do_select_tuple_arity2:
-     if (is_not_tuple(select_val2)) {
-	 goto select_val2_fail;
-     }
-     select_val2 = *tuple_val(select_val2);
-     goto do_select_val2;
-
- OpCase(i_select_val2_yfccff):
-     select_val2 = yb(Arg(0));
-     goto do_select_val2;
-
- OpCase(i_select_val2_xfccff):
-     select_val2 = xb(Arg(0));
-     goto do_select_val2;
-
- do_select_val2:
-     if (select_val2 == Arg(2)) {
-	 I += 3;
-     } else if (select_val2 == Arg(3)) {
-	 I += 4;
-     }
-
- select_val2_fail:
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
- }
-
- {
-     Eterm select_val;
-
- OpCase(i_select_tuple_arity_xfI):
-     select_val = xb(Arg(0));
-     goto do_select_tuple_arity;
-
- OpCase(i_select_tuple_arity_yfI):
-     select_val = yb(Arg(0));
-     goto do_select_tuple_arity;
-
- do_select_tuple_arity:
-     if (is_tuple(select_val)) {
-	 select_val = *tuple_val(select_val);
-	 goto do_linear_search;
-     }
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
-
- OpCase(i_select_val_lins_xfI):
-     select_val = xb(Arg(0));
-     goto do_linear_search;
-
- OpCase(i_select_val_lins_yfI):
-     select_val = yb(Arg(0));
-     goto do_linear_search;
-
- do_linear_search: {
-     BeamInstr *vs = &Arg(3);
-     int ix = 0;
-
-     for(;;) {
-	 if (vs[ix+0] >= select_val) { ix += 0; break; }
-	 if (vs[ix+1] >= select_val) { ix += 1; break; }
-	 ix += 2;
-     }
-
-     if (vs[ix] == select_val) {
-	 I += ix + Arg(2) + 2;
-     }
-
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
-   }
-
- OpCase(i_select_val_bins_xfI):
-     select_val = xb(Arg(0));
-     goto do_binary_search;
-
- OpCase(i_select_val_bins_yfI):
-     select_val = yb(Arg(0));
-     goto do_binary_search;
-     
- do_binary_search:
- {
-     struct Pairs {
-	 BeamInstr val;
-	 BeamInstr* addr;
-     };
-     struct Pairs* low;
-     struct Pairs* high;
-     struct Pairs* mid;
-     int bdiff; /* int not long because the arrays aren't that large */
-
-     low = (struct Pairs *) &Arg(3);
-     high = low + Arg(2);
-
-     /* The pointer subtraction (high-low) below must produce
-      * a signed result, because high could be < low. That
-      * requires the compiler to insert quite a bit of code.
-      *
-      * However, high will be > low so the result will be
-      * positive. We can use that knowledge to optimise the
-      * entire sequence, from the initial comparison to the
-      * computation of mid.
-      *
-      * -- Mikael Pettersson, Acumem AB
-      *
-      * Original loop control code:
-      *
-      * while (low < high) {
-      *    mid = low + (high-low) / 2;
-      *
-      */
-     while ((bdiff = (int)((char*)high - (char*)low)) > 0) {
-	 unsigned int boffset = ((unsigned int)bdiff >> 1) & ~(sizeof(struct Pairs)-1);
-
-	 mid = (struct Pairs*)((char*)low + boffset);
-	 if (select_val < mid->val) {
-	     high = mid;
-	 } else if (select_val > mid->val) {
-	     low = mid + 1;
-	 } else {
-	     SET_I(mid->addr);
-	     Goto(*I);
-	 }
-     }
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
- }
- }
-
- {
-     Eterm jump_on_val_zero_index;
-     
- OpCase(i_jump_on_val_zero_yfI):
-     jump_on_val_zero_index = yb(Arg(0));
-     goto do_jump_on_val_zero_index;
-
- OpCase(i_jump_on_val_zero_xfI):
-     jump_on_val_zero_index = xb(Arg(0));
-     goto do_jump_on_val_zero_index;
-
- do_jump_on_val_zero_index:
-     if (is_small(jump_on_val_zero_index)) {
-	 jump_on_val_zero_index = signed_val(jump_on_val_zero_index);
-	 if (jump_on_val_zero_index < Arg(2)) {
-	     SET_I((BeamInstr *) (&Arg(3))[jump_on_val_zero_index]);
-	     Goto(*I);
-	 }
-     }
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
- }
-
- {
-     Eterm jump_on_val_index;
-
- 
- OpCase(i_jump_on_val_yfII):
-     jump_on_val_index = yb(Arg(0));
-     goto do_jump_on_val_index;
-
- OpCase(i_jump_on_val_xfII):
-     jump_on_val_index = xb(Arg(0));
-     goto do_jump_on_val_index;
-
- do_jump_on_val_index:
-     if (is_small(jump_on_val_index)) {
-	 jump_on_val_index = (Uint) (signed_val(jump_on_val_index) - Arg(3));
-	 if (jump_on_val_index < Arg(2)) {
-	     SET_I((BeamInstr *) (&Arg(4))[jump_on_val_index]);
-	     Goto(*I);
-	 }
-     }
-     SET_I((BeamInstr *) Arg(1));
-     Goto(*I);
- }
-
- do_put_tuple: {
-     Eterm* hp = HTOP;
-
-     *hp++ = make_arityval(pt_arity);
-
-     do {
-	 Eterm term = *I++;
-	 switch (loader_tag(term)) {
-	 case LOADER_X_REG:
-	     *hp++ = x(loader_x_reg_index(term));
-	     break;
-	 case LOADER_Y_REG:
-	     *hp++ = y(loader_y_reg_index(term));
-	     break;
-	 default:
-	     *hp++ = term;
-	     break;
-	 }
-     } while (--pt_arity != 0);
-     HTOP = hp;
-     Goto(*I);
- }
-
- OpCase(new_map_dII): {
-     Eterm res;
-
-     HEAVY_SWAPOUT;
-     res = new_map(c_p, reg, I-1);
-     HEAVY_SWAPIN;
-     StoreResult(res, Arg(0));
-     Next(3+Arg(2));
- }
-
-#define PUT_TERM_REG(term, desc)		\
-do {						\
-    switch (loader_tag(desc)) {			\
-    case LOADER_X_REG:				\
-	x(loader_x_reg_index(desc)) = (term);	\
-	break;					\
-    case LOADER_Y_REG:				\
-	y(loader_y_reg_index(desc)) = (term);	\
-	break;					\
-    default:					\
-	ASSERT(0);				\
-	break;					\
-    }						\
-} while(0)
-
- OpCase(i_get_map_elements_fsI): {
-    Eterm map;
-    BeamInstr *fs;
-    Uint sz, n;
-
-    GetArg1(1, map);
-
-    /* this instruction assumes Arg1 is a map,
-     * i.e. that it follows a test is_map if needed.
-     */
-
-    n  = (Uint)Arg(2) / 3;
-    fs = &Arg(3); /* pattern fields and target registers */
-
-    if (is_flatmap(map)) {
-	flatmap_t *mp;
-	Eterm *ks;
-	Eterm *vs;
-
-	mp = (flatmap_t *)flatmap_val(map);
-	sz = flatmap_get_size(mp);
-
-	if (sz == 0) {
-	    ClauseFail();
-	}
-
-	ks = flatmap_get_keys(mp);
-	vs = flatmap_get_values(mp);
-
-	while(sz) {
-	    if (EQ((Eterm) fs[0], *ks)) {
-		PUT_TERM_REG(*vs, fs[1]);
-		n--;
-		fs += 3;
-		/* no more values to fetch, we are done */
-		if (n == 0) {
-		    I = fs;
-		    Next(-1);
-		}
-	    }
-	    ks++, sz--, vs++;
-	}
-
-	ClauseFail();
-    } else {
-	const Eterm *v;
-	Uint32 hx;
-	ASSERT(is_hashmap(map));
-	while(n--) {
-	    hx = fs[2];
-	    ASSERT(hx == hashmap_make_hash((Eterm)fs[0]));
-	    if ((v = erts_hashmap_get(hx, (Eterm)fs[0], map)) == NULL) {
-		ClauseFail();
-	    }
-	    PUT_TERM_REG(*v, fs[1]);
-	    fs += 3;
-	}
-	I = fs;
-	Next(-1);
-    }
- }
-#undef PUT_TERM_REG
-
- OpCase(update_map_assoc_jsdII): {
-     Eterm res;
-     Eterm map;
-
-     GetArg1(1, map);
-     HEAVY_SWAPOUT;
-     res = update_map_assoc(c_p, reg, map, I);
-     HEAVY_SWAPIN;
-     if (is_value(res)) {
-	 StoreResult(res, Arg(2));
-	 Next(5+Arg(4));
-     } else {
-	 /*
-	  * This can only happen if the code was compiled
-	  * with the compiler in OTP 17.
-	  */
-	 c_p->freason = BADMAP;
-	 c_p->fvalue = map;
-	 goto lb_Cl_error;
-     }
- }
-
- OpCase(update_map_exact_jsdII): {
-     Eterm res;
-     Eterm map;
-
-     GetArg1(1, map);
-     HEAVY_SWAPOUT;
-     res = update_map_exact(c_p, reg, map, I);
-     HEAVY_SWAPIN;
-     if (is_value(res)) {
-	 StoreResult(res, Arg(2));
-	 Next(5+Arg(4));
-     } else {
-	 goto lb_Cl_error;
-     }
- }
-
-
-    /*
-     * All guards with zero arguments have special instructions:
-     * 	self/0
-     * 	node/0
-     *
-     * All other guard BIFs take one or two arguments.
-     */
-
-    /*
-     * Guard BIF in head.  On failure, ignore the error and jump
-     * to the code for the next clause.  We don't support tracing
-     * of guard BIFs.
-     */
-
- OpCase(bif1_fbsd):
-    {
-	ErtsBifFunc bf;
-	Eterm tmp_reg[1];
-	Eterm result;
-
-	GetArg1(2, tmp_reg[0]);
-	bf = (BifFunction) Arg(1);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, tmp_reg, I);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(3, result);
-	}
-	SET_I((BeamInstr *) Arg(0));
-	Goto(*I);
-    }
-
-    /*
-     * Guard BIF in body.  It can fail like any BIF.  No trace support.
-     */
-
- OpCase(bif1_body_bsd):
-    {
-	ErtsBifFunc bf;
-
-	Eterm tmp_reg[1];
-	Eterm result;
-
-	GetArg1(1, tmp_reg[0]);
-	bf = (ErtsBifFunc) Arg(0);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, tmp_reg, I);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(2, result);
-	}
-	reg[0] = tmp_reg[0];
-	SWAPOUT;
-	I = handle_error(c_p, I, reg, ubif2mfa((void *) bf));
-	goto post_error_handling;
-    }
-
- OpCase(i_gc_bif1_jIsId):
-    {
-	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
-	GcBifFunction bf;
-	Eterm result;
-	Uint live = (Uint) Arg(3);
-
-	GetArg1(2, x(live));
-	bf = (GcBifFunction) Arg(1);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	SWAPOUT;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, reg, live);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	SWAPIN;
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(4, result);
-	}
-	if (Arg(0) != 0) {
-	    SET_I((BeamInstr *) Arg(0));
-	    Goto(*I);
-	}
-	x(0) = x(live);
-	I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
-	goto post_error_handling;
-    }
-
- OpCase(i_gc_bif2_jIIssd): /* Note, one less parameter than the i_gc_bif1
-			    and i_gc_bif3 */
-    {
-	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
-	GcBifFunction bf;
-	Eterm result;
-	Uint live = (Uint) Arg(2);
-
-	GetArg2(3, x(live), x(live+1));
-	/*
-	 * XXX This calling convention does not make sense. 'live'
-	 * should point out the first argument, not the second
-	 * (i.e. 'live' should not be incremented below).
-	 */
-	live++;
-	bf = (GcBifFunction) Arg(1);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	SWAPOUT;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, reg, live);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	SWAPIN;
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(5, result);
-	}
-	if (Arg(0) != 0) {
-	    SET_I((BeamInstr *) Arg(0));
-	    Goto(*I);
-	}
-	live--;
-	x(0) = x(live);
-	x(1) = x(live+1);
-	I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
-	goto post_error_handling;
-    }
-
- OpCase(i_gc_bif3_jIIssd):
-    {
-	typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
-	GcBifFunction bf;
-	Eterm result;
-	Uint live = (Uint) Arg(2);
-
-	x(live) = x(SCRATCH_X_REG);
-	GetArg2(3, x(live+1), x(live+2));
-	/*
-	 * XXX This calling convention does not make sense. 'live'
-	 * should point out the first argument, not the third
-	 * (i.e. 'live' should not be incremented below).
-	 */
-	live += 2;
-	bf = (GcBifFunction) Arg(1);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	SWAPOUT;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, reg, live);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	SWAPIN;
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(5, result);
-	}
-	if (Arg(0) != 0) {
-	    SET_I((BeamInstr *) Arg(0));
-	    Goto(*I);
-	}
-	live -= 2;
-	x(0) = x(live);
-	x(1) = x(live+1);
-	x(2) = x(live+2);
-	I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
-	goto post_error_handling;
-    }
-
- /*
-  * Guards bifs and, or, xor in guards.
-  */
- OpCase(i_bif2_fbssd):
-    {
-	Eterm tmp_reg[2];
-	ErtsBifFunc bf;
-	Eterm result;
-
-	GetArg2(2, tmp_reg[0], tmp_reg[1]);
-	bf = (ErtsBifFunc) Arg(1);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS;
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, tmp_reg, I);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_HOLE_CHECK(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	if (is_value(result)) {
-	    StoreBifResult(4, result);
-	}
-	SET_I((BeamInstr *) Arg(0));
-	Goto(*I);
-    }
-
- /*
-  * Guards bifs and, or, xor, relational operators in body.
-  */
- OpCase(i_bif2_body_bssd):
-    {
-	Eterm tmp_reg[2];
-	ErtsBifFunc bf;
-	Eterm result;
-
-	GetArg2(1, tmp_reg[0], tmp_reg[1]);
-	bf = (ErtsBifFunc) Arg(0);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, tmp_reg, I);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_HOLE_CHECK(c_p);
-	if (is_value(result)) {
-	    ASSERT(!is_CP(result));
-	    StoreBifResult(3, result);
-	}
-	reg[0] = tmp_reg[0];
-	reg[1] = tmp_reg[1];
-	SWAPOUT;
-	I = handle_error(c_p, I, reg, ubif2mfa((void *) bf));
-	goto post_error_handling;
-    }
-
-    /*
-     * The most general BIF call.  The BIF may build any amount of data
-     * on the heap.  The result is always returned in r(0).
-     */
- OpCase(call_bif_e):
-    {
-	ErtsBifFunc bf;
-	Eterm result;
-	BeamInstr *next;
-	ErlHeapFragment *live_hf_end;
-        Export *export = (Export*)Arg(0);
-
-
-        if (!((FCALLS - 1) > 0 || (FCALLS-1) > neg_o_reds)) {
-            /* If we have run out of reductions, we do a context
-               switch before calling the bif */
-            c_p->arity = GET_BIF_ARITY(export);
-            c_p->current = &export->info.mfa;
-            goto context_switch3;
-        }
-
-        ERTS_MSACC_SET_BIF_STATE_CACHED_X(
-            GET_BIF_MODULE(export), GET_BIF_ADDRESS(export));
-
-	bf = GET_BIF_ADDRESS(export);
-
-	PRE_BIF_SWAPOUT(c_p);
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	c_p->fcalls = FCALLS - 1;
-	if (FCALLS <= 0) {
-	   save_calls(c_p, export);
-	}
-	PreFetch(1, next);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	live_hf_end = c_p->mbuf;
-	ERTS_CHK_MBUF_SZ(c_p);
-	result = (*bf)(c_p, reg, I);
-	ERTS_CHK_MBUF_SZ(c_p);
-	ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
-	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	ERTS_HOLE_CHECK(c_p);
-	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	if (ERTS_IS_GC_DESIRED(c_p)) {
-	    Uint arity = GET_BIF_ARITY(export);
-	    result = erts_gc_after_bif_call_lhf(c_p, live_hf_end, result, reg, arity);
-	    E = c_p->stop;
-	}
-	PROCESS_MAIN_CHK_LOCKS(c_p);
-	HTOP = HEAP_TOP(c_p);
-	FCALLS = c_p->fcalls;
-	ERTS_DBG_CHK_REDS(c_p, FCALLS);
-        /* We have to update the cache if we are enabled in order
-           to make sure no book keeping is done after we disabled
-           msacc. We don't always do this as it is quite expensive. */
-        if (ERTS_MSACC_IS_ENABLED_CACHED_X())
-            ERTS_MSACC_UPDATE_CACHE_X();
-	ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
-	if (is_value(result)) {
-	    r(0) = result;
-	    CHECK_TERM(r(0));
-	    NextPF(1, next);
-	} else if (c_p->freason == TRAP) {
-	    SET_CP(c_p, I+2);
-	    SET_I(c_p->i);
-	    SWAPIN;
-	    Dispatch();
-	}
-
-	/*
-	 * Error handling.  SWAPOUT is not needed because it was done above.
-	 */
-	ASSERT(c_p->stop == E);
-	I = handle_error(c_p, I, reg, &export->info.mfa);
-	goto post_error_handling;
-    }
-
- /*
-  * Arithmetic operations.
-  */
-
- OpCase(i_times_jIssd):
- {
-     Eterm Op1, Op2;
-     GetArg2(2, Op1, Op2);
-     DO_OUTLINED_ARITH_2(mixed_times, Op1, Op2);
- }
-
- OpCase(i_m_div_jIssd):
- {
-     Eterm Op1, Op2;
-     GetArg2(2, Op1, Op2);
-     DO_OUTLINED_ARITH_2(mixed_div, Op1, Op2);
- }
-
- OpCase(i_int_div_jIssd):
- {
-     Eterm Op1, Op2;
-
-     GetArg2(2, Op1, Op2);
-     if (Op2 == SMALL_ZERO) {
-	 goto badarith;
-     } else if (is_both_small(Op1, Op2)) {
-	 Sint ires = signed_val(Op1) / signed_val(Op2);
-	 if (MY_IS_SSMALL(ires)) {
-	     Eterm result = make_small(ires);
-	     StoreBifResult(4, result);
-	 }
-     }
-     DO_OUTLINED_ARITH_2(int_div, Op1, Op2);
- }
-
- {
-     Eterm RemOp1, RemOp2;
-
- OpCase(i_rem_jIxxd):
-     RemOp1 = xb(Arg(2));
-     RemOp2 = xb(Arg(3));
-     goto do_rem;
-
- OpCase(i_rem_jIssd):
-     GetArg2(2, RemOp1, RemOp2);
-     goto do_rem;
-
- do_rem:
-     if (RemOp2 == SMALL_ZERO) {
-         goto badarith;
-     } else if (is_both_small(RemOp1, RemOp2)) {
-         Eterm result = make_small(signed_val(RemOp1) % signed_val(RemOp2));
-         StoreBifResult(4, result);
-     } else {
-	 DO_OUTLINED_ARITH_2(int_rem, RemOp1, RemOp2);
-     }
- }
-
- {
-     Eterm BandOp1, BandOp2;
-
- OpCase(i_band_jIxcd):
-     BandOp1 = xb(Arg(2));
-     BandOp2 = Arg(3);
-     goto do_band;
-
- OpCase(i_band_jIssd):
-     GetArg2(2, BandOp1, BandOp2);
-     goto do_band;
-
- do_band:
-     if (is_both_small(BandOp1, BandOp2)) {
-         /*
-          * No need to untag -- TAG & TAG == TAG.
-          */
-         Eterm result = BandOp1 & BandOp2;
-         StoreBifResult(4, result);
-     }
-     DO_OUTLINED_ARITH_2(band, BandOp1, BandOp2);
- }
-
- /*
-  * An error occurred in an arithmetic operation or test that could
-  * appear either in a head or in a body.
-  * In a head, execution should continue at failure address in Arg(0).
-  * In a body, Arg(0) == 0 and an exception should be raised.
-  */
- lb_Cl_error: {
-     if (Arg(0) != 0) {
-	 OpCase(jump_f): {
-	 jump_f:
-	     SET_I((BeamInstr *) Arg(0));
-	     Goto(*I);
-	 }
-     }
-     ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
-     goto find_func_info;
- }
-
- OpCase(i_bor_jIssd):
- {
-     Eterm Op1, Op2;
-
-     GetArg2(2, Op1, Op2);
-     if (is_both_small(Op1, Op2)) {
-	 /*
-	  * No need to untag -- TAG | TAG == TAG.
-	  */
-	 Eterm result = Op1 | Op2;
-	 StoreBifResult(4, result);
-     }
-     DO_OUTLINED_ARITH_2(bor, Op1, Op2);
- }
-
- OpCase(i_bxor_jIssd):
- {
-     Eterm Op1, Op2;
-
-     GetArg2(2, Op1, Op2);
-     if (is_both_small(Op1, Op2)) {
-	 /*
-          * TAG ^ TAG == 0.
-          *
-          * Therefore, we perform the XOR operation on the tagged values,
-          * and OR in the tag bits.
-	  */
-	 Eterm result = (Op1 ^ Op2) | make_small(0);
-	 StoreBifResult(4, result);
-     }
-     DO_OUTLINED_ARITH_2(bxor, Op1, Op2);
- }
-
- {
-     Eterm Op1, Op2;
-     Sint i;
-     Sint ires;
-     Eterm* bigp;
-     Eterm tmp_big[2];
-
-     OpCase(i_bsr_jIssd):
-         GetArg2(2, Op1, Op2);
-	 if (is_small(Op2)) {
-	     i = -signed_val(Op2);
-	     if (is_small(Op1)) {
-		 goto small_shift;
-	     } else if (is_big(Op1)) {
-		 if (i == 0) {
-		     StoreBifResult(4, Op1);
-		 }
-		 ires = big_size(Op1);
-		 goto big_shift;
-	     }
-	 } else if (is_big(Op2)) {
-	     /*
-	      * N bsr NegativeBigNum == N bsl MAX_SMALL
-	      * N bsr PositiveBigNum == N bsl MIN_SMALL
-	      */
-	     Op2 = make_small(bignum_header_is_neg(*big_val(Op2)) ?
-				   MAX_SMALL : MIN_SMALL);
-	     goto do_bsl;
-	}
-     goto badarith;
-     
-     OpCase(i_bsl_jIssd):
-         GetArg2(2, Op1, Op2);
- do_bsl:
-	 if (is_small(Op2)) {
-	     i = signed_val(Op2);
-
-	     if (is_small(Op1)) {
-	     small_shift:
-		 ires = signed_val(Op1);
-	     
-		 if (i == 0 || ires == 0) {
-		     StoreBifResult(4, Op1);
-		 } else if (i < 0)  { /* Right shift */
-		     i = -i;
-		     if (i >= SMALL_BITS-1) {
-			 Op1 = (ires < 0) ? SMALL_MINUS_ONE : SMALL_ZERO;
-		     } else {
-			 Op1 = make_small(ires >> i);
-		     }
-		     StoreBifResult(4, Op1);
-		 } else if (i < SMALL_BITS-1) { /* Left shift */
-		     if ((ires > 0 && ((~(Uint)0 << ((SMALL_BITS-1)-i)) & ires) == 0) ||
-			 ((~(Uint)0 << ((SMALL_BITS-1)-i)) & ~ires) == 0) {
-			 Op1 = make_small(ires << i);
-			 StoreBifResult(4, Op1);
-		     }
-		 }
-		 ires = 1; /* big_size(small_to_big(Op1)) */
-
-	     big_shift:
-		 if (i > 0) {	/* Left shift. */
-		     ires += (i / D_EXP);
-		 } else {	/* Right shift. */
-		     if (ires <= (-i / D_EXP))
-			 ires = 3; /* ??? */
-		     else
-			 ires -= (-i / D_EXP);
-		 }
-		 {
-		     ires = BIG_NEED_SIZE(ires+1);
-		     /*
-		      * Slightly conservative check the size to avoid
-		      * allocating huge amounts of memory for bignums that 
-		      * clearly would overflow the arity in the header
-		      * word.
-		      */
-		     if (ires-8 > BIG_ARITY_MAX) {
-			 c_p->freason = SYSTEM_LIMIT;
-			 goto lb_Cl_error;
-		     }
-		     TestHeapPreserve(ires+1, Arg(1), Op1);
-		     if (is_small(Op1)) {
-			 Op1 = small_to_big(signed_val(Op1), tmp_big);
-		     }
-		     bigp = HTOP;
-		     Op1 = big_lshift(Op1, i, bigp);
-		     if (is_big(Op1)) {
-			 HTOP += bignum_header_arity(*HTOP) + 1;
-		     }
-		     HEAP_SPACE_VERIFIED(0);
-		     if (is_nil(Op1)) {
-			 /*
-			  * This result must have been only slight larger
-			  * than allowed since it wasn't caught by the
-			  * previous test.
-			  */
-			 c_p->freason = SYSTEM_LIMIT;
-			 goto lb_Cl_error;
-		     }
-		     ERTS_HOLE_CHECK(c_p);
-		     StoreBifResult(4, Op1);
-		 }
-	     } else if (is_big(Op1)) {
-		 if (i == 0) {
-		     StoreBifResult(4, Op1);
-		 }
-		 ires = big_size(Op1);
-		 goto big_shift;
-	     }
-	 } else if (is_big(Op2)) {
-	     if (bignum_header_is_neg(*big_val(Op2))) {
-		 /*
-		  * N bsl NegativeBigNum is either 0 or -1, depending on
-		  * the sign of N. Since we don't believe this case
-		  * is common, do the calculation with the minimum
-		  * amount of code.
-		  */
-		 Op2 = make_small(MIN_SMALL);
-		 goto do_bsl;
-	     } else if (is_small(Op1) || is_big(Op1)) {
-		 /*
-		  * N bsl PositiveBigNum is too large to represent.
-		  */
-		 c_p->freason = SYSTEM_LIMIT;
-		 goto lb_Cl_error;
-	     }
-	     /* Fall through if the left argument is not an integer. */
-	 }
-     /*
-      * One or more non-integer arguments.
-      */
-     goto badarith;
- }
-
- OpCase(i_int_bnot_jsId):
- {
-     Eterm bnot_val;
-
-     GetArg1(1, bnot_val);
-     if (is_small(bnot_val)) {
-	 bnot_val = make_small(~signed_val(bnot_val));
-     } else {
-	 Uint live = Arg(2);
-	 HEAVY_SWAPOUT;
-	 reg[live] = bnot_val;
-	 bnot_val = erts_gc_bnot(c_p, reg, live);
-	 HEAVY_SWAPIN;
-	 ERTS_HOLE_CHECK(c_p);
-	 if (is_nil(bnot_val)) {
-	     goto lb_Cl_error;
-	 }
-     }
-     StoreBifResult(3, bnot_val);
- }
-
- badarith:
-    c_p->freason = BADARITH;
-    goto lb_Cl_error;
-
- OpCase(i_apply): {
-     BeamInstr *next;
-     HEAVY_SWAPOUT;
-     next = apply(c_p, r(0), x(1), x(2), reg, NULL, 0);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, I+1);
-	 SET_I(next);
-	 Dispatch();
-     }
-     I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
-     goto post_error_handling;
- }
-
- OpCase(i_apply_last_P): {
-     BeamInstr *next;
-     HEAVY_SWAPOUT;
-     next = apply(c_p, r(0), x(1), x(2), reg, I, Arg(0));
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, (BeamInstr *) E[0]);
-	 E = ADD_BYTE_OFFSET(E, Arg(0));
-	 SET_I(next);
-	 Dispatch();
-     }
-     I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
-     goto post_error_handling;
- }
-
- OpCase(i_apply_only): {
-     BeamInstr *next;
-     HEAVY_SWAPOUT;
-     next = apply(c_p, r(0), x(1), x(2), reg, I, 0);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_I(next);
-	 Dispatch();
-     }
-     I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
-     goto post_error_handling;
- }
-
- OpCase(apply_I): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = fixed_apply(c_p, reg, Arg(0), NULL, 0);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, I+2);
-	 SET_I(next);
-	 Dispatch();
-     }
-     I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
-     goto post_error_handling;
- }
-
- OpCase(apply_last_IP): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = fixed_apply(c_p, reg, Arg(0), I, Arg(1));
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, (BeamInstr *) E[0]);
-	 E = ADD_BYTE_OFFSET(E, Arg(1));
-	 SET_I(next);
-	 Dispatch();
-     }
-     I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
-     goto post_error_handling;
- }
-
- OpCase(i_apply_fun): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = apply_fun(c_p, r(0), x(1), reg);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, I+1);
-	 SET_I(next);
-	 Dispatchfun();
-     }
-     goto find_func_info;
- }
-
- OpCase(i_apply_fun_last_P): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = apply_fun(c_p, r(0), x(1), reg);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, (BeamInstr *) E[0]);
-	 E = ADD_BYTE_OFFSET(E, Arg(0));
-	 SET_I(next);
-	 Dispatchfun();
-     }
-     goto find_func_info;
- }
-
- OpCase(i_apply_fun_only): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = apply_fun(c_p, r(0), x(1), reg);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_I(next);
-	 Dispatchfun();
-     }
-     goto find_func_info;
- }
-
- OpCase(i_call_fun_I): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = call_fun(c_p, Arg(0), reg, THE_NON_VALUE);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	 SET_CP(c_p, I+2);
-	 SET_I(next);
-	 Dispatchfun();
-     }
-     goto find_func_info;
- }
-
- OpCase(i_call_fun_last_IP): {
-     BeamInstr *next;
-
-     HEAVY_SWAPOUT;
-     next = call_fun(c_p, Arg(0), reg, THE_NON_VALUE);
-     HEAVY_SWAPIN;
-     if (next != NULL) {
-	SET_CP(c_p, (BeamInstr *) E[0]);
-	E = ADD_BYTE_OFFSET(E, Arg(1));
-	SET_I(next);
-	Dispatchfun();
-     }
-     goto find_func_info;
- }
+#include "beam_hot.h"
 
 #ifdef DEBUG
     /*
@@ -3395,7 +820,7 @@ do {						\
      Eterm* argp;
      int i;
 
-     if (erts_smp_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_EXITING) {
+     if (erts_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_EXITING) {
          c_p->i = beam_exit;
          c_p->arity = 0;
          c_p->current = NULL;
@@ -3452,64 +877,25 @@ do {						\
      goto do_schedule1;
  }
 
- OpCase(set_tuple_element_sdP): {
-     Eterm element;
-     Eterm tuple;
-     BeamInstr *next;
-     Eterm* p;
-     
-     PreFetch(3, next);
-     GetArg1(0, element);
-     tuple = REG_TARGET(Arg(1));
-     ASSERT(is_tuple(tuple));
-     p = (Eterm *) ((unsigned char *) tuple_val(tuple) + Arg(2));
-     *p = element;
-     NextPF(3, next);
- }
+#include "beam_warm.h"
 
  OpCase(normal_exit): {
      SWAPOUT;
      c_p->freason = EXC_NORMAL;
-     c_p->arity = 0;		/* In case this process will never be garbed again. */
-     ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+     c_p->arity = 0; /* In case this process will ever be garbed again. */
+     ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
      erts_do_exit_process(c_p, am_normal);
-     ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+     ERTS_REQ_PROC_MAIN_LOCK(c_p);
      goto do_schedule;
  }
 
  OpCase(continue_exit): {
-     ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+     ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
      erts_continue_exit_process(c_p);
-     ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+     ERTS_REQ_PROC_MAIN_LOCK(c_p);
      goto do_schedule;
  }
 
- OpCase(i_raise): {
-     Eterm raise_trace = x(2);
-     Eterm raise_value = x(1);
-     struct StackTrace *s;
-
-     c_p->fvalue = raise_value;
-     c_p->ftrace = raise_trace;
-     s = get_trace_from_exc(raise_trace);
-     if (s == NULL) {
-	 c_p->freason = EXC_ERROR;
-     } else {
-	 c_p->freason = PRIMARY_EXCEPTION(s->freason);
-     }
-     goto find_func_info;
- }
-
-    {
-	Eterm badmatch_val;
-
-    OpCase(badmatch_x):
-	badmatch_val = xb(Arg(0));
-	c_p->fvalue = badmatch_val;
-	c_p->freason = BADMATCH;
-    }
- /* Fall through here */
-
  find_func_info: {
      SWAPOUT;
      I = handle_error(c_p, I, reg, NULL);
@@ -3550,194 +936,6 @@ do {						\
      }
  }
 
-    {
-	Eterm nif_bif_result;
-	Eterm bif_nif_arity;
-
-    OpCase(call_nif):
-	{
-	    /*
-	     * call_nif is always first instruction in function:
-	     *
-	     * I[-3]: Module
-	     * I[-2]: Function
-	     * I[-1]: Arity
-	     * I[0]: &&call_nif
-	     * I[1]: Function pointer to NIF function
-	     * I[2]: Pointer to erl_module_nif
-	     * I[3]: Function pointer to dirty NIF
-             *
-             * This layout is determined by the NifExport struct
-	     */
-	    BifFunction vbf;
-	    ErlHeapFragment *live_hf_end;
-            ErtsCodeMFA *codemfa;
-
-	    ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_NIF);
-
-	    codemfa = erts_code_to_codemfa(I);
-
-            c_p->current = codemfa; /* current and vbf set to please handle_error */
-
-	    DTRACE_NIF_ENTRY(c_p, codemfa);
-
-	    HEAVY_SWAPOUT;
-
-	    PROCESS_MAIN_CHK_LOCKS(c_p);
-	    bif_nif_arity = codemfa->arity;
-	    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-
-	    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	    {
-		typedef Eterm NifF(struct enif_environment_t*, int argc, Eterm argv[]);
-		NifF* fp = vbf = (NifF*) I[1];
-		struct enif_environment_t env;
-#ifdef ERTS_SMP
-		ASSERT(c_p->scheduler_data);
-#endif
-		live_hf_end = c_p->mbuf;
-		ERTS_CHK_MBUF_SZ(c_p);
-		erts_pre_nif(&env, c_p, (struct erl_module_nif*)I[2], NULL);
-		nif_bif_result = (*fp)(&env, bif_nif_arity, reg);
-		if (env.exception_thrown)
-		    nif_bif_result = THE_NON_VALUE;
-		erts_post_nif(&env);
-		ERTS_CHK_MBUF_SZ(c_p);
-
-		PROCESS_MAIN_CHK_LOCKS(c_p);
-		ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-		ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
-		ASSERT(!env.exiting);
-		ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	    }
-
-	    DTRACE_NIF_RETURN(c_p, codemfa);
-	    goto apply_bif_or_nif_epilogue;
-	 
-	OpCase(apply_bif):
-	    /*
-	     * At this point, I points to the code[0] in the export entry for
-	     * the BIF:
-	     *
-	     * code[-3]: Module
-	     * code[-2]: Function
-	     * code[-1]: Arity
-	     * code[0]: &&apply_bif
-	     * code[1]: Function pointer to BIF function
-	     */
-
-            if (!((FCALLS - 1) > 0 || (FCALLS - 1) > neg_o_reds)) {
-                /* If we have run out of reductions, we do a context
-                   switch before calling the bif */
-                goto context_switch;
-            }
-
-	    codemfa = erts_code_to_codemfa(I);
-
-            ERTS_MSACC_SET_BIF_STATE_CACHED_X(codemfa->module, (BifFunction)Arg(0));
-
-
-            /* In case we apply process_info/1,2 or load_nif/1 */
-	    c_p->current = codemfa;
-	    c_p->i = I;		/* In case we apply check_process_code/2. */
-	    c_p->arity = 0;		/* To allow garbage collection on ourselves
-					 * (check_process_code/2).
-					 */
-	    DTRACE_BIF_ENTRY(c_p, codemfa);
-
-	    SWAPOUT;
-	    ERTS_DBG_CHK_REDS(c_p, FCALLS - 1);
-	    c_p->fcalls = FCALLS - 1;
-	    vbf = (BifFunction) Arg(0);
-	    PROCESS_MAIN_CHK_LOCKS(c_p);
-	    bif_nif_arity = codemfa->arity;
-            ASSERT(bif_nif_arity <= 4);
-	    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	    {
-		ErtsBifFunc bf = vbf;
-		ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-		live_hf_end = c_p->mbuf;
-		ERTS_CHK_MBUF_SZ(c_p);
-		nif_bif_result = (*bf)(c_p, reg, I);
-		ERTS_CHK_MBUF_SZ(c_p);
-		ASSERT(!ERTS_PROC_IS_EXITING(c_p) ||
-		       is_non_value(nif_bif_result));
-		ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-		PROCESS_MAIN_CHK_LOCKS(c_p);
-	    }
-            /* We have to update the cache if we are enabled in order
-               to make sure no book keeping is done after we disabled
-               msacc. We don't always do this as it is quite expensive. */
-            if (ERTS_MSACC_IS_ENABLED_CACHED_X())
-                ERTS_MSACC_UPDATE_CACHE_X();
-	    ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
-	    DTRACE_BIF_RETURN(c_p, codemfa);
-
-	apply_bif_or_nif_epilogue:
-	    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	    ERTS_HOLE_CHECK(c_p);
-	    if (ERTS_IS_GC_DESIRED(c_p)) {
-		nif_bif_result = erts_gc_after_bif_call_lhf(c_p, live_hf_end,
-							    nif_bif_result,
-							    reg, bif_nif_arity);
-	    }
-	    SWAPIN;  /* There might have been a garbage collection. */
-	    FCALLS = c_p->fcalls;
-	    ERTS_DBG_CHK_REDS(c_p, FCALLS);
-	    if (is_value(nif_bif_result)) {
-		r(0) = nif_bif_result;
-		CHECK_TERM(r(0));
-		SET_I(c_p->cp);
-		c_p->cp = 0;
-		Goto(*I);
-	    } else if (c_p->freason == TRAP) {
-		SET_I(c_p->i);
-		if (c_p->flags & F_HIBERNATE_SCHED) {
-		    c_p->flags &= ~F_HIBERNATE_SCHED;
-		    goto do_schedule;
-		}
-		Dispatch();
-	    }
-	    I = handle_error(c_p, c_p->cp, reg, c_p->current);
-	    goto post_error_handling;
-	}
-    }
-
- OpCase(i_get_sd):
-    {
-	Eterm arg;
-	Eterm result;
-
-	GetArg1(0, arg);
-	result = erts_pd_hash_get(c_p, arg);
-	StoreBifResult(1, result);
-    }
-
- OpCase(i_get_hash_cId):
-    {
-	Eterm arg;
-	Eterm result;
-
-	GetArg1(0, arg);
-	result = erts_pd_hash_get_with_hx(c_p, Arg(1), arg);
-	StoreBifResult(2, result);
-    }
-
-    {
-	Eterm case_end_val;
-
-    OpCase(case_end_x):
-	case_end_val = xb(Arg(0));
-	c_p->fvalue = case_end_val;
-	c_p->freason = EXC_CASE_CLAUSE;
-	goto find_func_info;
-    }
-
- OpCase(if_end):
-    c_p->freason = EXC_IF_CLAUSE;
-    goto find_func_info;
-
  OpCase(i_func_info_IaaI): {
      ErtsCodeInfo *ci = (ErtsCodeInfo*)I;
      c_p->freason = EXC_FUNCTION_CLAUSE;
@@ -3745,1367 +943,8 @@ do {						\
      goto handle_error;
  }
 
- OpCase(try_case_end_s):
-    {
-	Eterm try_case_end_val;
-	GetArg1(0, try_case_end_val);
-	c_p->fvalue = try_case_end_val;
-	c_p->freason = EXC_TRY_CLAUSE;
-	goto find_func_info;
-    }
-
- /*
-  * Construction of binaries using new instructions.
-  */
- {
-     Eterm new_binary;
-     Eterm num_bits_term;
-     Uint num_bits;
-     Uint alloc;
-     Uint num_bytes;
-
-     OpCase(i_bs_init_bits_heap_IIId): {
-	 num_bits = Arg(0);
-	 alloc = Arg(1);
-	 I++;
-	 goto do_bs_init_bits_known;
-     }
-     
-     OpCase(i_bs_init_bits_IId): {
-	 num_bits = Arg(0);
-	 alloc = 0;
-	 goto do_bs_init_bits_known;
-     }
-
-     OpCase(i_bs_init_bits_fail_heap_sIjId): {
-	 GetArg1(0, num_bits_term);
-	 alloc = Arg(1);
-	 I += 2;
-	 goto do_bs_init_bits;
-     }
-
-     OpCase(i_bs_init_bits_fail_yjId): {
-	 num_bits_term = yb(Arg(0));
-	 I++;
-	 alloc = 0;
-	 goto do_bs_init_bits;
-     }
-     OpCase(i_bs_init_bits_fail_xjId): {
-	 num_bits_term = xb(Arg(0));
-	 I++;
-	 alloc = 0;
-     /* FALL THROUGH */
-     }
-
-     /* num_bits_term = Term for number of bits to build (small/big)
-      * alloc = Number of words to allocate on heap
-      * Operands: Fail Live Dst
-      */
-
- do_bs_init_bits:
-     if (is_small(num_bits_term)) {
-	 Sint size = signed_val(num_bits_term);
-	 if (size < 0) {
-	     goto badarg;
-	 }
-	 num_bits = (Uint) size;
-     } else {
-	 Uint bits;
-
-	 if (!term_to_Uint(num_bits_term, &bits)) {
-	     c_p->freason = bits;
-	     goto lb_Cl_error;
-
-	 }
-	 num_bits = (Eterm) bits;
-     }
-
-     /* num_bits = Number of bits to build
-      * alloc = Number of extra words to allocate on heap
-      * Operands: NotUsed Live Dst
-      */
- do_bs_init_bits_known:
-     num_bytes = ((Uint64)num_bits+(Uint64)7) >> 3;
-     if (num_bits & 7) {
-	 alloc += ERL_SUB_BIN_SIZE;
-     }
-     if (num_bytes <= ERL_ONHEAP_BIN_LIMIT) {
-	 alloc += heap_bin_size(num_bytes);
-     } else {
-	 alloc += PROC_BIN_SIZE;
-     }
-     TestHeap(alloc, Arg(1));
-
-     /* num_bits = Number of bits to build
-      * num_bytes = Number of bytes to allocate in the binary
-      * alloc = Total number of words to allocate on heap
-      * Operands: NotUsed NotUsed Dst
-      */
-     if (num_bytes <= ERL_ONHEAP_BIN_LIMIT) {
-	 ErlHeapBin* hb;
-
-	 erts_bin_offset = 0;
-	 erts_writable_bin = 0;
-	 hb = (ErlHeapBin *) HTOP;
-	 HTOP += heap_bin_size(num_bytes);
-	 hb->thing_word = header_heap_bin(num_bytes);
-	 hb->size = num_bytes;
-	 erts_current_bin = (byte *) hb->data;
-	 new_binary = make_binary(hb);
-
-     do_bits_sub_bin:
-	 if (num_bits & 7) {
-	     ErlSubBin* sb;
-
-	     sb = (ErlSubBin *) HTOP;
-	     HTOP += ERL_SUB_BIN_SIZE;
-	     sb->thing_word = HEADER_SUB_BIN;
-	     sb->size = num_bytes - 1;
-	     sb->bitsize = num_bits & 7;
-	     sb->offs = 0;
-	     sb->bitoffs = 0;
-	     sb->is_writable = 0;
-	     sb->orig = new_binary;
-	     new_binary = make_binary(sb);
-	 }
-	 HEAP_SPACE_VERIFIED(0);
-	 StoreBifResult(2, new_binary);
-     } else {
-	 Binary* bptr;
-	 ProcBin* pb;
-
-	 erts_bin_offset = 0;
-	 erts_writable_bin = 0;
-
-	 /*
-	  * Allocate the binary struct itself.
-	  */
-	 bptr = erts_bin_nrml_alloc(num_bytes);
-	 erts_current_bin = (byte *) bptr->orig_bytes;
-
-	 /*
-	  * Now allocate the ProcBin on the heap.
-	  */
-	 pb = (ProcBin *) HTOP;
-	 HTOP += PROC_BIN_SIZE;
-	 pb->thing_word = HEADER_PROC_BIN;
-	 pb->size = num_bytes;
-	 pb->next = MSO(c_p).first;
-	 MSO(c_p).first = (struct erl_off_heap_header*) pb;
-	 pb->val = bptr;
-	 pb->bytes = (byte*) bptr->orig_bytes;
-	 pb->flags = 0;
-	 OH_OVERHEAD(&(MSO(c_p)), pb->size / sizeof(Eterm));
-	 new_binary = make_binary(pb);
-	 goto do_bits_sub_bin;
-     }
- }
-
- {
-     Eterm BsOp1, BsOp2;
-
-     OpCase(i_bs_init_fail_heap_sIjId): {
-	 GetArg1(0, BsOp1);
-	 BsOp2 = Arg(1);
-	 I += 2;
-	 goto do_bs_init;
-     }
-
-     OpCase(i_bs_init_fail_yjId): {
-	 BsOp1 = yb(Arg(0));
-	 BsOp2 = 0;
-	 I++;
-	 goto do_bs_init;
-     }
-
-     OpCase(i_bs_init_fail_xjId): {
-	 BsOp1 = xb(Arg(0));
-	 BsOp2 = 0;
-	 I++;
-     }
-	 /* FALL THROUGH */
-     do_bs_init:
-         if (is_small(BsOp1)) {
-	     Sint size = signed_val(BsOp1);
-	     if (size < 0) {
-		 goto badarg;
-	     }
-	     BsOp1 = (Eterm) size;
-	 } else {
-	     Uint bytes;
-
-	     if (!term_to_Uint(BsOp1, &bytes)) {
-		 c_p->freason = bytes;
-		 goto lb_Cl_error;
-	     }
-	     if ((bytes >> (8*sizeof(Uint)-3)) != 0) {
-		 goto system_limit;
-	     }
-	     BsOp1 = (Eterm) bytes;
-	 }
-	 if (BsOp1 <= ERL_ONHEAP_BIN_LIMIT) {
-	     goto do_heap_bin_alloc;
-	 } else {
-	     goto do_proc_bin_alloc;
-	 }
-
-
-     OpCase(i_bs_init_heap_IIId): {
-	 BsOp1 = Arg(0);
-	 BsOp2 = Arg(1);
-	 I++;
-	 goto do_proc_bin_alloc;
-     }
-
-     OpCase(i_bs_init_IId): {
-	 BsOp1 = Arg(0);
-	 BsOp2 = 0;
-     }
-     /* FALL THROUGH */
-     do_proc_bin_alloc: {
-	 Binary* bptr;
-	 ProcBin* pb;
-
-	 erts_bin_offset = 0;
-	 erts_writable_bin = 0;
-	 TestBinVHeap(BsOp1 / sizeof(Eterm),
-		      BsOp2 + PROC_BIN_SIZE + ERL_SUB_BIN_SIZE, Arg(1));
-
-	 /*
-	  * Allocate the binary struct itself.
-	  */
-	 bptr = erts_bin_nrml_alloc(BsOp1);
-	 erts_current_bin = (byte *) bptr->orig_bytes;
-
-	 /*
-	  * Now allocate the ProcBin on the heap.
-	  */
-	 pb = (ProcBin *) HTOP;
-	 HTOP += PROC_BIN_SIZE;
-	 pb->thing_word = HEADER_PROC_BIN;
-	 pb->size = BsOp1;
-	 pb->next = MSO(c_p).first;
-	 MSO(c_p).first = (struct erl_off_heap_header*) pb;
-	 pb->val = bptr;
-	 pb->bytes = (byte*) bptr->orig_bytes;
-	 pb->flags = 0;
-	 
-	 OH_OVERHEAD(&(MSO(c_p)), BsOp1 / sizeof(Eterm));
-
-	 StoreBifResult(2, make_binary(pb));
-     }
-
-     OpCase(i_bs_init_heap_bin_heap_IIId): {
-	 BsOp1 = Arg(0);
-	 BsOp2 = Arg(1);
-	 I++;
-	 goto do_heap_bin_alloc;
-     }
-
-     OpCase(i_bs_init_heap_bin_IId): {
-	 BsOp1 = Arg(0);
-	 BsOp2 = 0;
-     }
-     /* Fall through */
-     do_heap_bin_alloc:
-	 {
-	     ErlHeapBin* hb;
-	     Uint bin_need;
-
-	     bin_need = heap_bin_size(BsOp1);
-	     erts_bin_offset = 0;
-	     erts_writable_bin = 0;
-	     TestHeap(bin_need+BsOp2+ERL_SUB_BIN_SIZE, Arg(1));
-	     hb = (ErlHeapBin *) HTOP;
-	     HTOP += bin_need;
-	     hb->thing_word = header_heap_bin(BsOp1);
-	     hb->size = BsOp1;
-	     erts_current_bin = (byte *) hb->data;
-	     BsOp1 = make_binary(hb);
-	     StoreBifResult(2, BsOp1);
-	 }
- }
-
- OpCase(bs_add_jssId): {
-     Eterm Op1, Op2;
-     Uint Unit = Arg(3);
-
-     GetArg2(1, Op1, Op2);
-     if (is_both_small(Op1, Op2)) {
-	 Sint Arg1 = signed_val(Op1);
-	 Sint Arg2 = signed_val(Op2);
-
-	 if (Arg1 >= 0 && Arg2 >= 0) {
-	     BsSafeMul(Arg2, Unit, goto system_limit, Op1);
-	     Op1 += Arg1;
-
-	 store_bs_add_result:
-	     if (Op1 <= MAX_SMALL) {
-		 Op1 = make_small(Op1);
-	     } else {
-		 /*
-		  * May generate a heap fragment, but in this
-		  * particular case it is OK, since the value will be
-		  * stored into an x register (the GC will scan x
-		  * registers for references to heap fragments) and
-		  * there is no risk that value can be stored into a
-		  * location that is not scanned for heap-fragment
-		  * references (such as the heap).
-		  */
-		 SWAPOUT;
-		 Op1 = erts_make_integer(Op1, c_p);
-		 HTOP = HEAP_TOP(c_p);
-	     }
-	     StoreBifResult(4, Op1);
-	 }
-	 goto badarg;
-     } else {
-	 Uint a;
-	 Uint b;
-	 Uint c;
-
-	 /*
-	  * Now we know that one of the arguments is
-	  * not a small. We must convert both arguments
-	  * to Uints and check for errors at the same time.
-	  *
-	  * Error checking is tricky.
-	  *
-	  * If one of the arguments is not numeric or
-	  * not positive, the error reason is BADARG.
-	  *
-	  * Otherwise if both arguments are numeric,
-	  * but at least one argument does not fit in
-	  * an Uint, the reason is SYSTEM_LIMIT.
-	  */
-
-	 if (!term_to_Uint(Op1, &a)) {
-	     if (a == BADARG) {
-		 goto badarg;
-	     }
-	     if (!term_to_Uint(Op2, &b)) {
-		 c_p->freason = b;
-		 goto lb_Cl_error;
-	     }
-	     goto system_limit;
-	 } else if (!term_to_Uint(Op2, &b)) {
-	     c_p->freason = b;
-	     goto lb_Cl_error;
-	 }
-
-	 /*
-	  * The arguments are now correct and stored in a and b.
-	  */
-	 
-	 BsSafeMul(b, Unit, goto system_limit, c);
-	 Op1 = a + c;
-	 if (Op1 < a) {
-	     /*
-	      * If the result is less than one of the
-	      * arguments, there must have been an overflow.
-	      */
-	     goto system_limit;
-	 }
-	 goto store_bs_add_result;
-     }
-     /* No fallthrough */
-     ASSERT(0);
- }
-
- OpCase(bs_put_string_II):
-    {
-	BeamInstr *next;
-	PreFetch(2, next);
-	erts_new_bs_put_string(ERL_BITS_ARGS_2((byte *) Arg(1), Arg(0)));
-	NextPF(2, next);
-    }
-
- /*
-  * x(SCRATCH_X_REG);
-  * Operands: Fail ExtraHeap Live Unit Size Dst
-  */
-
- OpCase(i_bs_append_jIIIsd): {
-     Uint live = Arg(2);
-     Uint res;
-     Eterm Size;
-
-     GetArg1(4, Size);
-     HEAVY_SWAPOUT;
-     reg[live] = x(SCRATCH_X_REG);
-     res = erts_bs_append(c_p, reg, live, Size, Arg(1), Arg(3));
-     HEAVY_SWAPIN;
-     if (is_non_value(res)) {
-	 /* c_p->freason is already set (may be either BADARG or SYSTEM_LIMIT). */
-	 goto lb_Cl_error;
-     }
-     StoreBifResult(5, res);
- }
-
- /*
-  * Operands: Fail Size Src Unit Dst
-  */
- OpCase(i_bs_private_append_jIssd): {
-     Eterm res;
-     Eterm Size, Src;
-
-     GetArg2(2, Size, Src);
-     res = erts_bs_private_append(c_p, Src, Size, Arg(1));
-     if (is_non_value(res)) {
-	 /* c_p->freason is already set (may be either BADARG or SYSTEM_LIMIT). */
-	 goto lb_Cl_error;
-     }
-     StoreBifResult(4, res);
- }
-
- OpCase(bs_init_writable): {
-     HEAVY_SWAPOUT;
-     r(0) = erts_bs_init_writable(c_p, r(0));
-     HEAVY_SWAPIN;
-     Next(0);
- }
-
- /*
-  * Calculate the number of bytes needed to encode the source
-  * operarand to UTF-8. If the source operand is invalid (e.g. wrong
-  * type or range) we return a nonsense integer result (0 or 4). We
-  * can get away with that because we KNOW that bs_put_utf8 will do
-  * full error checking.
-  */
- OpCase(i_bs_utf8_size_sd): {
-     Eterm arg;
-     Eterm result;
-
-     GetArg1(0, arg);
-     if (arg < make_small(0x80UL)) {
-	 result = make_small(1);
-     } else if (arg < make_small(0x800UL)) {
-	 result = make_small(2);
-     } else if (arg < make_small(0x10000UL)) {
-	 result = make_small(3);
-     } else {
-	 result = make_small(4);
-     }
-     StoreBifResult(1, result);
- }
-
- OpCase(i_bs_put_utf8_js): {
-     Eterm arg;
-
-     GetArg1(1, arg);
-     if (!erts_bs_put_utf8(ERL_BITS_ARGS_1(arg))) {
-	 goto badarg;
-     }
-     Next(2);
- }
-
- /*
-  * Calculate the number of bytes needed to encode the source
-  * operarand to UTF-8. If the source operand is invalid (e.g. wrong
-  * type or range) we return a nonsense integer result (2 or 4). We
-  * can get away with that because we KNOW that bs_put_utf16 will do
-  * full error checking.
-  */
-
- OpCase(i_bs_utf16_size_sd): {
-     Eterm arg;
-     Eterm result = make_small(2);
-
-     GetArg1(0, arg);
-     if (arg >= make_small(0x10000UL)) {
-	 result = make_small(4);
-     }
-     StoreBifResult(1, result);
- }
-
- OpCase(bs_put_utf16_jIs): {
-     Eterm arg;
-
-     GetArg1(2, arg);
-     if (!erts_bs_put_utf16(ERL_BITS_ARGS_2(arg, Arg(1)))) {
-	 goto badarg;
-     }
-     Next(3);
- }
-
- /*
-  * Only used for validating a value about to be stored in a binary.
-  */
- OpCase(i_bs_validate_unicode_js): {
-     Eterm val;
-
-     GetArg1(1, val);
-
-     /*
-      * There is no need to untag the integer, but it IS necessary
-      * to make sure it is small (if the term is a bignum, it could
-      * slip through the test, and there is no further test that
-      * would catch it, since bit syntax construction silently masks
-      * too big numbers).
-      */
-     if (is_not_small(val) || val > make_small(0x10FFFFUL) ||
-	 (make_small(0xD800UL) <= val && val <= make_small(0xDFFFUL))) {
-	 goto badarg;
-     }
-     Next(2);
- }
-
- /*
-  * Only used for validating a value matched out. 
-  */
- OpCase(i_bs_validate_unicode_retract_jss): {
-	Eterm i; 		/* Integer to validate */
-
-	/*
-	 * There is no need to untag the integer, but it IS necessary
-	 * to make sure it is small (a bignum pointer could fall in
-	 * the valid range).
-	 */
-
-	GetArg1(1, i);
-	if (is_not_small(i) || i > make_small(0x10FFFFUL) ||
-	    (make_small(0xD800UL) <= i && i <= make_small(0xDFFFUL))) {
-	    Eterm ms;		/* Match context */
-	    ErlBinMatchBuffer* mb;
-
-	    GetArg1(2, ms);
-	    mb = ms_matchbuffer(ms);
-	    mb->offset -= 32;
-	    goto badarg;
-	}
-	Next(3);
-    }
-
- /*
-  * Matching of binaries.
-  */
-
- {
-     Eterm header;
-     BeamInstr *next;
-     Uint slots;
-     Eterm context;
-
-     do_start_match:
-	 slots = Arg(2);
-	 if (!is_boxed(context)) {
-	     ClauseFail();
-	 }
-	 PreFetch(4, next);
-	 header = *boxed_val(context);
-	 if (header_is_bin_matchstate(header)) {
-	     ErlBinMatchState* ms = (ErlBinMatchState *) boxed_val(context);
-	     Uint actual_slots = HEADER_NUM_SLOTS(header);
-	     ms->save_offset[0] = ms->mb.offset;
-	     if (actual_slots < slots) {
-		 ErlBinMatchState* dst;
-		 Uint live = Arg(1);
-		 Uint wordsneeded = ERL_BIN_MATCHSTATE_SIZE(slots);
-
-		 TestHeapPreserve(wordsneeded, live, context);
-		 ms = (ErlBinMatchState *) boxed_val(context);
-		 dst = (ErlBinMatchState *) HTOP;
-		 *dst = *ms;
-		 *HTOP = HEADER_BIN_MATCHSTATE(slots);
-		 HTOP += wordsneeded;
-		 HEAP_SPACE_VERIFIED(0);
-		 StoreResult(make_matchstate(dst), Arg(3));
-	     }
-	 } else if (is_binary_header(header)) {
-	     Eterm result;
-	     Uint live = Arg(1);
-	     Uint wordsneeded = ERL_BIN_MATCHSTATE_SIZE(slots);
-	     TestHeapPreserve(wordsneeded, live, context);
-	     HEAP_TOP(c_p) = HTOP;
-#ifdef DEBUG
-	     c_p->stop = E;	/* Needed for checking in HeapOnlyAlloc(). */
-#endif
-	     result = erts_bs_start_match_2(c_p, context, slots);
-	     HTOP = HEAP_TOP(c_p);
-	     HEAP_SPACE_VERIFIED(0);
-	     if (is_non_value(result)) {
-		 ClauseFail();
-	     } else {
-		 StoreResult(result, Arg(3));
-	     }
-	 } else {
-	     ClauseFail();
-	 }
-	 NextPF(4, next);
-
-     OpCase(i_bs_start_match2_xfIId): {
-	 context = xb(Arg(0));
-	 I++;
-	 goto do_start_match;
-     }
-     OpCase(i_bs_start_match2_yfIId): {
-	 context = yb(Arg(0));
-	 I++;
-	 goto do_start_match;
-     }
- }
-
- OpCase(bs_test_zero_tail2_fx): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     
-     PreFetch(2, next);
-     _mb = (ErlBinMatchBuffer*) ms_matchbuffer(xb(Arg(1)));
-     if (_mb->size != _mb->offset) {
-	 ClauseFail();
-     }
-     NextPF(2, next);
- }
-
- OpCase(bs_test_tail_imm2_fxI): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(3, next);
-     _mb = ms_matchbuffer(xb(Arg(1)));
-     if (_mb->size - _mb->offset != Arg(2)) {
-	 ClauseFail();
-     }
-     NextPF(3, next);
- }
-
- OpCase(bs_test_unit_fxI): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(3, next);
-     _mb = ms_matchbuffer(xb(Arg(1)));
-     if ((_mb->size - _mb->offset) % Arg(2)) {
-	 ClauseFail();
-     }
-     NextPF(3, next);
- }
-
- OpCase(bs_test_unit8_fx): {
-     BeamInstr *next;
-     ErlBinMatchBuffer *_mb;
-     PreFetch(2, next);
-     _mb = ms_matchbuffer(xb(Arg(1)));
-     if ((_mb->size - _mb->offset) & 7) {
-	 ClauseFail();
-     }
-     NextPF(2, next);
- }
-
- {
-     Eterm bs_get_integer8_context;
-
- OpCase(i_bs_get_integer_8_xfd): {
-	 ErlBinMatchBuffer *_mb;
-	 Eterm _result;
-	 bs_get_integer8_context = xb(Arg(0));
-	 I++;
-	 _mb = ms_matchbuffer(bs_get_integer8_context);
-	 if (_mb->size - _mb->offset < 8) {
-	     ClauseFail();
-	 }
-	 if (BIT_OFFSET(_mb->offset) != 0) {
-	     _result = erts_bs_get_integer_2(c_p, 8, 0, _mb);
-	 } else {
-	     _result = make_small(_mb->base[BYTE_OFFSET(_mb->offset)]);
-	     _mb->offset += 8;
-	 }
-	 StoreBifResult(1, _result);
-     }
- }
-
- {
-     Eterm bs_get_integer_16_context;
-
- OpCase(i_bs_get_integer_16_xfd):
-     bs_get_integer_16_context = xb(Arg(0));
-     I++;
-
-     {
-	 ErlBinMatchBuffer *_mb;
-	 Eterm _result;
-	 _mb = ms_matchbuffer(bs_get_integer_16_context);
-	 if (_mb->size - _mb->offset < 16) {
-	     ClauseFail(); 
-	 }
-	 if (BIT_OFFSET(_mb->offset) != 0) {
-	     _result = erts_bs_get_integer_2(c_p, 16, 0, _mb);
-	 } else {
-	     _result = make_small(get_int16(_mb->base+BYTE_OFFSET(_mb->offset)));
-	     _mb->offset += 16;
-	 }
-	 StoreBifResult(1, _result);
-     }
- }
-
- {
-     Eterm bs_get_integer_32_context;
-
- OpCase(i_bs_get_integer_32_xfId):
-     bs_get_integer_32_context = xb(Arg(0));
-     I++;
-
-     {
-	 ErlBinMatchBuffer *_mb;
-	 Uint32 _integer;
-	 Eterm _result;
-	 _mb = ms_matchbuffer(bs_get_integer_32_context);
-	 if (_mb->size - _mb->offset < 32) { ClauseFail(); }
-	 if (BIT_OFFSET(_mb->offset) != 0) {
-	     _integer = erts_bs_get_unaligned_uint32(_mb);
-	 } else {
-	     _integer = get_int32(_mb->base + _mb->offset/8);
-	 }
-	 _mb->offset += 32;
-#if !defined(ARCH_64)
-	 if (IS_USMALL(0, _integer)) {
-#endif
-	     _result = make_small(_integer);
-#if !defined(ARCH_64)
-	 } else {
-	     TestHeap(BIG_UINT_HEAP_SIZE, Arg(1));
-	     _result = uint_to_big((Uint) _integer, HTOP);
-	     HTOP += BIG_UINT_HEAP_SIZE;
-	     HEAP_SPACE_VERIFIED(0);
-	 }
-#endif
-	 StoreBifResult(2, _result);
-     }
- }
-
- {
-     Eterm Ms, Sz;
-
-     /* Operands: x(Reg) Size Live Fail Flags Dst */
- OpCase(i_bs_get_integer_imm_xIIfId): {
-	 Uint wordsneeded;
-	 Ms = xb(Arg(0));
-	 Sz = Arg(1);
-	 wordsneeded = 1+WSIZE(NBYTES(Sz));
-	 TestHeapPreserve(wordsneeded, Arg(2), Ms);
-	 I += 3;
-	 /* Operands: Fail Flags Dst */
-	 goto do_bs_get_integer_imm;
-     }
-
-     /* Operands: x(Reg) Size Fail Flags Dst */
- OpCase(i_bs_get_integer_small_imm_xIfId): {
-	 Ms = xb(Arg(0));
-	 Sz = Arg(1);
-	 I += 2;
-	 /* Operands: Fail Flags Dst */
-	 goto do_bs_get_integer_imm;
-     }
- 
-     /*
-      * Ms = match context
-      * Sz = size of field
-      * Operands: Fail Flags Dst
-      */
- do_bs_get_integer_imm: {
-	 ErlBinMatchBuffer* mb;
-	 Eterm result;
-
-	 mb = ms_matchbuffer(Ms);
-	 LIGHT_SWAPOUT;
-	 result = erts_bs_get_integer_2(c_p, Sz, Arg(1), mb);
-	 LIGHT_SWAPIN;
-	 HEAP_SPACE_VERIFIED(0);
-	 if (is_non_value(result)) {
-	     ClauseFail();
-	 }
-	 StoreBifResult(2, result);
-     }
- }
-
- /*
-  * Operands: Fail Live FlagsAndUnit Ms Sz Dst
-  */
- OpCase(i_bs_get_integer_fIIssd): {
-     Uint flags;
-     Uint size;
-     Eterm Ms;
-     Eterm Sz;
-     ErlBinMatchBuffer* mb;
-     Eterm result;
-
-     flags = Arg(2);
-     GetArg2(3, Ms, Sz);
-     BsGetFieldSize(Sz, (flags >> 3), ClauseFail(), size);
-     if (size >= SMALL_BITS) {
-	 Uint wordsneeded;
-	 /* Check bits size before potential gc.
-	  * We do not want a gc and then realize we don't need
-	  * the allocated space (i.e. if the op fails).
-	  *
-	  * Remember to re-acquire the matchbuffer after gc.
-	  */
-
-	 mb = ms_matchbuffer(Ms);
-	 if (mb->size - mb->offset < size) {
-	     ClauseFail();
-	 }
-	 wordsneeded = 1+WSIZE(NBYTES((Uint) size));
-	 TestHeapPreserve(wordsneeded, Arg(1), Ms);
-     }
-     mb = ms_matchbuffer(Ms);
-     LIGHT_SWAPOUT;
-     result = erts_bs_get_integer_2(c_p, size, flags, mb);
-     LIGHT_SWAPIN;
-     HEAP_SPACE_VERIFIED(0);
-     if (is_non_value(result)) {
-	 ClauseFail();
-     }
-     StoreBifResult(5, result);
- }
-
- {
-     Eterm get_utf8_context;
-
-     /* Operands: MatchContext Fail Dst */
- OpCase(i_bs_get_utf8_xfd): {
-	 get_utf8_context = xb(Arg(0));
-	 I++;
-     }
-
-     /*
-      * get_utf8_context = match_context
-      * Operands: Fail Dst
-      */
-
-     {
-	 Eterm result = erts_bs_get_utf8(ms_matchbuffer(get_utf8_context));
-	 if (is_non_value(result)) {
-	     ClauseFail();
-	 }
-	 StoreBifResult(1, result);
-     }
- }
-
- {
-     Eterm get_utf16_context;
-
-     /* Operands: MatchContext Fail Flags Dst */
-     OpCase(i_bs_get_utf16_xfId): {
-	 get_utf16_context = xb(Arg(0));
-	 I++;
-     }
-
-     /*
-      * get_utf16_context = match_context
-      * Operands: Fail Flags Dst
-      */
-     {
-	 Eterm result = erts_bs_get_utf16(ms_matchbuffer(get_utf16_context),
-					  Arg(1));
-	 if (is_non_value(result)) {
-	     ClauseFail();
-	 }
-	 StoreBifResult(2, result);
-     }
- }
-
- {
-     Eterm context_to_binary_context;
-     ErlBinMatchBuffer* mb;
-     ErlSubBin* sb;
-     Uint size;
-     Uint offs;
-     Uint orig;
-     Uint hole_size;
-
-     OpCase(bs_context_to_binary_x):
-	 context_to_binary_context = xb(Arg(0));
-	 I--;
-
-     if (is_boxed(context_to_binary_context) &&
-	 header_is_bin_matchstate(*boxed_val(context_to_binary_context))) {
-	 ErlBinMatchState* ms;
-	 ms = (ErlBinMatchState *) boxed_val(context_to_binary_context);
-	 mb = &ms->mb;
-	 offs = ms->save_offset[0];
-	 size = mb->size - offs;
-	 goto do_bs_get_binary_all_reuse_common;
-     }
-     Next(2);
-
-     OpCase(i_bs_get_binary_all_reuse_xfI): {
-	 context_to_binary_context = xb(Arg(0));
-	 I++;
-     }
-
-     mb = ms_matchbuffer(context_to_binary_context);
-     size = mb->size - mb->offset;
-     if (size % Arg(1) != 0) {
-	 ClauseFail();
-     }
-     offs = mb->offset;
-
- do_bs_get_binary_all_reuse_common:
-     orig = mb->orig;
-     sb = (ErlSubBin *) boxed_val(context_to_binary_context);
-     hole_size = 1 + header_arity(sb->thing_word) - ERL_SUB_BIN_SIZE;
-     sb->thing_word = HEADER_SUB_BIN;
-     sb->size = BYTE_OFFSET(size);
-     sb->bitsize = BIT_OFFSET(size);
-     sb->offs = BYTE_OFFSET(offs);
-     sb->bitoffs = BIT_OFFSET(offs);
-     sb->is_writable = 0;
-     sb->orig = orig;
-     if (hole_size) {
-	 sb[1].thing_word = make_pos_bignum_header(hole_size-1);
-     }
-     Next(2);
- }
-
- {
-     Eterm match_string_context;
-
-     OpCase(i_bs_match_string_xfII): {
-	 match_string_context = xb(Arg(0));
-	 I++;
-     }
-
-     {
-	 BeamInstr *next;
-	 byte* bytes;
-	 Uint bits;
-	 ErlBinMatchBuffer* mb;
-	 Uint offs;
-
-	 PreFetch(3, next);
-	 bits = Arg(1);
-	 bytes = (byte *) Arg(2);
-	 mb = ms_matchbuffer(match_string_context);
-	 if (mb->size - mb->offset < bits) {
-	     ClauseFail();
-	 }
-	 offs = mb->offset & 7;
-	 if (offs == 0 && (bits & 7) == 0) {
-	     if (sys_memcmp(bytes, mb->base+(mb->offset>>3), bits>>3)) {
-		 ClauseFail();
-	     }
-	 } else if (erts_cmp_bits(bytes, 0, mb->base+(mb->offset>>3), mb->offset & 7, bits)) {
-	     ClauseFail();
-	 }
-	 mb->offset += bits;
-	 NextPF(3, next);
-     }
- }
-
- OpCase(i_bs_save2_xI): {
-     BeamInstr *next;
-     ErlBinMatchState *_ms;
-     PreFetch(2, next);
-     _ms = (ErlBinMatchState*) boxed_val((Eterm) xb(Arg(0)));
-     _ms->save_offset[Arg(1)] = _ms->mb.offset;
-     NextPF(2, next);
- }
-
- OpCase(i_bs_restore2_xI): {
-     BeamInstr *next;
-     ErlBinMatchState *_ms;
-     PreFetch(2, next);
-     _ms = (ErlBinMatchState*) boxed_val((Eterm) xb(Arg(0)));
-     _ms->mb.offset = _ms->save_offset[Arg(1)];
-     NextPF(2, next);
- }
-
 #include "beam_cold.h"
 
-
- /*
-  * This instruction is probably never used (because it is combined with a
-  * a return). However, a future compiler might for some reason emit a
-  * deallocate not followed by a return, and that should work.
-  */
- OpCase(deallocate_I): {
-     BeamInstr *next;
-
-     PreFetch(1, next);
-     D(Arg(0));
-     NextPF(1, next);
- }
-
-    /*
-     * Trace and debugging support.
-     */
-
- OpCase(return_trace): {
-     ErtsCodeMFA* mfa = (ErtsCodeMFA *)(E[0]);
-     
-     SWAPOUT;		/* Needed for shared heap */
-     ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-     erts_trace_return(c_p, mfa, r(0), ERTS_TRACER_FROM_ETERM(E+1)/* tracer */);
-     ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-     SWAPIN;
-     c_p->cp = NULL;
-     SET_I((BeamInstr *) cp_val(E[2]));
-     E += 3;
-     Goto(*I);
- }
-
- OpCase(i_generic_breakpoint): {
-     BeamInstr real_I;
-     HEAVY_SWAPOUT;
-     real_I = erts_generic_breakpoint(c_p, erts_code_to_codeinfo(I), reg);
-     HEAVY_SWAPIN;
-     ASSERT(VALID_INSTR(real_I));
-     Goto(real_I);
- }
-
- OpCase(i_return_time_trace): {
-     BeamInstr *pc = (BeamInstr *) (UWord) E[0];
-     SWAPOUT;
-     erts_trace_time_return(c_p, erts_code_to_codeinfo(pc));
-     SWAPIN;
-     c_p->cp = NULL;
-     SET_I((BeamInstr *) cp_val(E[1]));
-     E += 2;
-     Goto(*I);
- }
-
- OpCase(i_return_to_trace): {
-     if (IS_TRACED_FL(c_p, F_TRACE_RETURN_TO)) {
-	 Uint *cpp = (Uint*) E;
-	 for(;;) {
-	     ASSERT(is_CP(*cpp));
-	     if (*cp_val(*cpp) == (BeamInstr) OpCode(return_trace)) {
-		 do ++cpp; while(is_not_CP(*cpp));
-		 cpp += 2;
-	     } else if (*cp_val(*cpp) == (BeamInstr) OpCode(i_return_to_trace)) {
-		 do ++cpp; while(is_not_CP(*cpp));
-	     } else break;
-	 }
-	 SWAPOUT;		/* Needed for shared heap */
-	 ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
-	 erts_trace_return_to(c_p, cp_val(*cpp));
-	 ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
-	 SWAPIN;
-     }
-     c_p->cp = NULL;
-     SET_I((BeamInstr *) cp_val(E[0]));
-     E += 1;
-     Goto(*I);
- }
-
- /*
-  * New floating point instructions.
-  */
-
- OpCase(fmove_ql): {
-     Eterm fr = Arg(1);
-     BeamInstr *next;
-
-     PreFetch(2, next);
-     GET_DOUBLE(Arg(0), *(FloatDef*)ADD_BYTE_OFFSET(freg, fr));
-     NextPF(2, next);
- }
-
- OpCase(fmove_dl): {
-     Eterm targ1;
-     Eterm fr = Arg(1);
-     BeamInstr *next;
-
-     PreFetch(2, next);
-     targ1 = REG_TARGET(Arg(0));
-     /* Arg(0) == HEADER_FLONUM */
-     GET_DOUBLE(targ1, *(FloatDef*)ADD_BYTE_OFFSET(freg, fr));
-     NextPF(2, next);
- }
-
- OpCase(fmove_ld): {
-     Eterm fr = Arg(0);
-     Eterm dest = make_float(HTOP);
-
-     PUT_DOUBLE(*(FloatDef*)ADD_BYTE_OFFSET(freg, fr), HTOP);
-     HTOP += FLOAT_SIZE_OBJECT;
-     StoreBifResult(1, dest);
- }
-
- OpCase(fconv_dl): {
-     Eterm targ1;
-     Eterm fr = Arg(1);
-     BeamInstr *next;
-
-     targ1 = REG_TARGET(Arg(0));
-     PreFetch(2, next);
-     if (is_small(targ1)) {
-	 fb(fr) = (double) signed_val(targ1);
-     } else if (is_big(targ1)) {
-	 if (big_to_double(targ1, &fb(fr)) < 0) {
-	     goto fbadarith;
-	 }
-     } else if (is_float(targ1)) {
-	 GET_DOUBLE(targ1, *(FloatDef*)ADD_BYTE_OFFSET(freg, fr));
-     } else {
-	 goto fbadarith;
-     }
-     NextPF(2, next);
- }
-
-#ifdef NO_FPE_SIGNALS
-     OpCase(fclearerror):
-     OpCase(i_fcheckerror):
-	 erts_exit(ERTS_ERROR_EXIT, "fclearerror/i_fcheckerror without fpe signals (beam_emu)");
-#  define ERTS_NO_FPE_CHECK_INIT ERTS_FP_CHECK_INIT
-#  define ERTS_NO_FPE_ERROR ERTS_FP_ERROR
-#else
-#  define ERTS_NO_FPE_CHECK_INIT(p)
-#  define ERTS_NO_FPE_ERROR(p, a, b)
-
-     OpCase(fclearerror): {
-	 BeamInstr *next;
-
-	 PreFetch(0, next);
-	 ERTS_FP_CHECK_INIT(c_p);
-	 NextPF(0, next);
-     }
-
-     OpCase(i_fcheckerror): {
-	 BeamInstr *next;
-
-	 PreFetch(0, next);
-	 ERTS_FP_ERROR(c_p, freg[0].fd, goto fbadarith);
-	 NextPF(0, next);
-     }
-#endif
-
-
- OpCase(i_fadd_lll): {
-     BeamInstr *next;
-
-     PreFetch(3, next);
-     ERTS_NO_FPE_CHECK_INIT(c_p);
-     fb(Arg(2)) = fb(Arg(0)) + fb(Arg(1));
-     ERTS_NO_FPE_ERROR(c_p, fb(Arg(2)), goto fbadarith);
-     NextPF(3, next);
- }
- OpCase(i_fsub_lll): {
-     BeamInstr *next;
-
-     PreFetch(3, next);
-     ERTS_NO_FPE_CHECK_INIT(c_p);
-     fb(Arg(2)) = fb(Arg(0)) - fb(Arg(1));
-     ERTS_NO_FPE_ERROR(c_p, fb(Arg(2)), goto fbadarith);
-     NextPF(3, next);
- }
- OpCase(i_fmul_lll): {
-     BeamInstr *next;
-
-     PreFetch(3, next);
-     ERTS_NO_FPE_CHECK_INIT(c_p);
-     fb(Arg(2)) = fb(Arg(0)) * fb(Arg(1));
-     ERTS_NO_FPE_ERROR(c_p, fb(Arg(2)), goto fbadarith);
-     NextPF(3, next);
- }
- OpCase(i_fdiv_lll): {
-     BeamInstr *next;
-
-     PreFetch(3, next);
-     ERTS_NO_FPE_CHECK_INIT(c_p);
-     fb(Arg(2)) = fb(Arg(0)) / fb(Arg(1));
-     ERTS_NO_FPE_ERROR(c_p, fb(Arg(2)), goto fbadarith);
-     NextPF(3, next);
- }
- OpCase(i_fnegate_ll): {
-     BeamInstr *next;
-
-     PreFetch(2, next);
-     ERTS_NO_FPE_CHECK_INIT(c_p);
-     fb(Arg(1)) = -fb(Arg(0));
-     ERTS_NO_FPE_ERROR(c_p, fb(Arg(1)), goto fbadarith);
-     NextPF(2, next);
-
- fbadarith:
-     c_p->freason = BADARITH;
-     goto find_func_info;
- }
-
-#ifdef HIPE
- {
-#define HIPE_MODE_SWITCH(Cmd)			\
-     SWAPOUT;					\
-     ERTS_DBG_CHK_REDS(c_p, FCALLS);		\
-     c_p->fcalls = FCALLS;			\
-     c_p->def_arg_reg[4] = -neg_o_reds;		\
-     c_p = hipe_mode_switch(c_p, Cmd, reg);     \
-     goto L_post_hipe_mode_switch
-
-     OpCase(hipe_trap_call): {
-	 /*
-	  * I[-5]: &&lb_i_func_info_IaaI
-	  * I[-4]: Native code callee (inserted by HiPE)
-	  * I[-3]: Module (tagged atom)
-	  * I[-2]: Function (tagged atom)
-	  * I[-1]: Arity (untagged integer)
-	  * I[ 0]: &&lb_hipe_trap_call
-	  * ... remainder of original BEAM code
-	  */
-         ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
-	 ASSERT(ci->op == (Uint) OpCode(i_func_info_IaaI));
-	 c_p->hipe.u.ncallee = ci->u.ncallee;
-	 ++hipe_trap_count;
-	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL | (ci->mfa.arity << 8));
-     }
-     OpCase(hipe_trap_call_closure): {
-       ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
-       ASSERT(ci->op == (Uint) OpCode(i_func_info_IaaI));
-       c_p->hipe.u.ncallee = ci->u.ncallee;
-       ++hipe_trap_count;
-       HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL_CLOSURE | (ci->mfa.arity << 8));
-     }
-     OpCase(hipe_trap_return): {
-	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RETURN);
-     }
-     OpCase(hipe_trap_throw): {
-	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_THROW);
-     }
-     OpCase(hipe_trap_resume): {
-	 HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RESUME);
-     }
-#undef HIPE_MODE_SWITCH
-
- L_post_hipe_mode_switch:
-#ifdef DEBUG
-     pid = c_p->common.id; /* may have switched process... */
-#endif
-     reg = erts_proc_sched_data(c_p)->x_reg_array;
-     freg = erts_proc_sched_data(c_p)->f_reg_array;
-     ERL_BITS_RELOAD_STATEP(c_p);
-     /* XXX: this abuse of def_arg_reg[] is horrid! */
-     neg_o_reds = -c_p->def_arg_reg[4];
-     FCALLS = c_p->fcalls;
-     SWAPIN;
-     ERTS_DBG_CHK_REDS(c_p, FCALLS);
-     switch( c_p->def_arg_reg[3] ) {
-       case HIPE_MODE_SWITCH_RES_RETURN:
-	 ASSERT(is_value(reg[0]));
-	 SET_I(c_p->cp);
-	 c_p->cp = 0;
-	 Goto(*I);
-       case HIPE_MODE_SWITCH_RES_CALL_EXPORTED:
-	 c_p->i = c_p->hipe.u.callee_exp->addressv[erts_active_code_ix()];
-	 /*fall through*/
-       case HIPE_MODE_SWITCH_RES_CALL_BEAM:
-	 SET_I(c_p->i);
-	 Dispatch();
-       case HIPE_MODE_SWITCH_RES_CALL_CLOSURE:
-	 /* This can be used to call any function value, but currently it's
-	    only used to call closures referring to unloaded modules. */
-	 {
-	     BeamInstr *next;
-
-	     next = call_fun(c_p, c_p->arity - 1, reg, THE_NON_VALUE);
-	     HEAVY_SWAPIN;
-	     if (next != NULL) {
-		 SET_I(next);
-		 Dispatchfun();
-	     }
-	     goto find_func_info;
-	 }
-       case HIPE_MODE_SWITCH_RES_THROW:
-	 c_p->cp = NULL;
-	 I = handle_error(c_p, I, reg, NULL);
-	 goto post_error_handling;
-       default:
-	 erts_exit(ERTS_ERROR_EXIT, "hipe_mode_switch: result %u\n", c_p->def_arg_reg[3]);
-     }
- }
- OpCase(hipe_call_count): {
-     /*
-      * I[-5]: &&lb_i_func_info_IaaI
-      * I[-4]: pointer to struct hipe_call_count (inserted by HiPE)
-      * I[-3]: Module (tagged atom)
-      * I[-2]: Function (tagged atom)
-      * I[-1]: Arity (untagged integer)
-      * I[ 0]: &&lb_hipe_call_count
-      * ... remainder of original BEAM code
-      */
-     ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
-     struct hipe_call_count *hcc = ci->u.hcc;
-     ASSERT(ci->op == (Uint) OpCode(i_func_info_IaaI));
-     ASSERT(hcc != NULL);
-     ASSERT(VALID_INSTR(hcc->opcode));
-     ++(hcc->count);
-     Goto(hcc->opcode);
- }
-#endif /* HIPE */
-
- OpCase(i_yield):
- {
-     /* This is safe as long as REDS_IN(c_p) is never stored 
-      * in c_p->arg_reg[0]. It is currently stored in c_p->def_arg_reg[5],
-      * which may be c_p->arg_reg[5], which is close, but no banana.
-      */
-     c_p->arg_reg[0] = am_true;
-     c_p->arity = 1; /* One living register (the 'true' return value) */
-     SWAPOUT;
-     c_p->i = I + 1; /* Next instruction */
-     c_p->current = NULL;
-     goto do_schedule;
- }
-
- OpCase(i_hibernate): {
-     HEAVY_SWAPOUT;
-     if (erts_hibernate(c_p, r(0), x(1), x(2), reg)) {
-	 FCALLS = c_p->fcalls;
-	 c_p->flags &= ~F_HIBERNATE_SCHED;
-	 goto do_schedule;
-     } else {
-	 HEAVY_SWAPIN;
-	 I = handle_error(c_p, I, reg, &bif_export[BIF_hibernate_3]->info.mfa);
-	 goto post_error_handling;
-     }
- }
-
- /* This is optimised as an instruction because
-    it has to be very very fast */
- OpCase(i_perf_counter): {
-    BeamInstr* next;
-    ErtsSysPerfCounter ts;
-    PreFetch(0, next);
-
-    ts = erts_sys_perf_counter();
-
-    if (IS_SSMALL(ts)) {
-        r(0) = make_small((Sint)ts);
-    } else {
-        TestHeap(ERTS_SINT64_HEAP_SIZE(ts),0);
-        r(0) = make_big(HTOP);
-#if defined(ARCH_32)
-        if (ts >= (((Uint64) 1) << 32)) {
-            *HTOP = make_pos_bignum_header(2);
-            BIG_DIGIT(HTOP, 0) = (Uint) (ts & ((Uint) 0xffffffff));
-            BIG_DIGIT(HTOP, 1) = (Uint) ((ts >> 32) & ((Uint) 0xffffffff));
-            HTOP += 3;
-        }
-        else
-#endif
-        {
-                *HTOP = make_pos_bignum_header(1);
-                BIG_DIGIT(HTOP, 0) = (Uint) ts;
-                HTOP += 2;
-        }
-    }
-    NextPF(0, next);
- }
-
- OpCase(i_debug_breakpoint): {
-     HEAVY_SWAPOUT;
-     I = call_error_handler(c_p, erts_code_to_codemfa(I), reg, am_breakpoint);
-     HEAVY_SWAPIN;
-     if (I) {
-	 Goto(*I);
-     }
-     goto handle_error;
- }
-
-
- OpCase(system_limit_j):
- system_limit:
-    c_p->freason = SYSTEM_LIMIT;
-    goto lb_Cl_error;
-
-
 #ifdef ERTS_OPCODE_COUNTER_SUPPORT
     DEFINE_COUNTING_LABELS;
 #endif
@@ -5128,9 +967,6 @@ do {						\
 
  init_emulator:
  {
-     int i;
-     Export* ep;
-
 #ifndef NO_JUMP_TABLE
 #ifdef ERTS_OPCODE_COUNTER_SUPPORT
 #ifdef DEBUG
@@ -5142,35 +978,8 @@ do {						\
      beam_ops = opcodes;
 #endif /* ERTS_OPCODE_COUNTER_SUPPORT */
 #endif /* NO_JUMP_TABLE */
-     
-     em_call_error_handler = OpCode(call_error_handler);
-     em_apply_bif = OpCode(apply_bif);
-     em_call_nif = OpCode(call_nif);
-     em_call_bif_e = OpCode(call_bif_e);
-
-     beam_apply[0]             = (BeamInstr) OpCode(i_apply);
-     beam_apply[1]             = (BeamInstr) OpCode(normal_exit);
-     beam_exit[0]              = (BeamInstr) OpCode(error_action_code);
-     beam_continue_exit[0]     = (BeamInstr) OpCode(continue_exit);
-     beam_return_to_trace[0]   = (BeamInstr) OpCode(i_return_to_trace);
-     beam_return_trace[0]      = (BeamInstr) OpCode(return_trace);
-     beam_exception_trace[0]   = (BeamInstr) OpCode(return_trace); /* UGLY */
-     beam_return_time_trace[0] = (BeamInstr) OpCode(i_return_time_trace);
-
-     /*
-      * Enter all BIFs into the export table.
-      */
-     for (i = 0; i < BIF_SIZE; i++) {
-	 ep = erts_export_put(bif_table[i].module,
-			      bif_table[i].name,
-			      bif_table[i].arity);
-	 bif_export[i] = ep;
-	 ep->beam[0] = (BeamInstr) OpCode(apply_bif);
-	 ep->beam[1] = (BeamInstr) bif_table[i].f;
-	 /* XXX: set func info for bifs */
-	 ep->info.op = (BeamInstr) BeamOp(op_i_func_info_IaaI);
-     }
 
+     init_emulator_finish();
      return;
  }
 #ifdef NO_JUMP_TABLE
@@ -5182,26 +991,71 @@ do {						\
 
   save_calls1:
     {
-	Eterm* dis_next;
+	BeamInstr dis_next;
 
 	save_calls(c_p, (Export *) Arg(0));
 
 	SET_I(((Export *) Arg(0))->addressv[erts_active_code_ix()]);
 
-	dis_next = (Eterm *) *I;
+	dis_next = *I;
 	FCALLS--;
 	Goto(dis_next);
     }
 }
 
 /*
+ * One-time initialization of emulator. Does not need to be
+ * in process_main().
+ */
+static void
+init_emulator_finish(void)
+{
+     int i;
+     Export* ep;
+
+#if defined(ARCH_64) && defined(CODE_MODEL_SMALL)
+     for (i = 0; i < NUMBER_OF_OPCODES; i++) {
+         BeamInstr instr = BeamOpCodeAddr(i);
+         if (instr >= (1ull << 32)) {
+             erts_exit(ERTS_ERROR_EXIT,
+                       "This run-time was supposed be compiled with all code below 2Gb,\n"
+                       "but the instruction '%s' is located at %016lx.\n",
+                       opc[i].name, instr);
+         }
+     }
+#endif
+
+     beam_apply[0]             = BeamOpCodeAddr(op_i_apply);
+     beam_apply[1]             = BeamOpCodeAddr(op_normal_exit);
+     beam_exit[0]              = BeamOpCodeAddr(op_error_action_code);
+     beam_continue_exit[0]     = BeamOpCodeAddr(op_continue_exit);
+     beam_return_to_trace[0]   = BeamOpCodeAddr(op_i_return_to_trace);
+     beam_return_trace[0]      = BeamOpCodeAddr(op_return_trace);
+     beam_exception_trace[0]   = BeamOpCodeAddr(op_return_trace); /* UGLY */
+     beam_return_time_trace[0] = BeamOpCodeAddr(op_i_return_time_trace);
+
+     /*
+      * Enter all BIFs into the export table.
+      */
+     for (i = 0; i < BIF_SIZE; i++) {
+	 ep = erts_export_put(bif_table[i].module,
+			      bif_table[i].name,
+			      bif_table[i].arity);
+	 bif_export[i] = ep;
+	 ep->beam[0] = BeamOpCodeAddr(op_apply_bif);
+	 ep->beam[1] = (BeamInstr) bif_table[i].f;
+	 /* XXX: set func info for bifs */
+	 ep->info.op = BeamOpCodeAddr(op_i_func_info_IaaI);
+     }
+}
+
+/*
  * erts_dirty_process_main() is what dirty schedulers execute. Since they handle
  * only NIF calls they do not need to be able to execute all BEAM
  * instructions.
  */
 void erts_dirty_process_main(ErtsSchedulerData *esdp)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     Process* c_p = NULL;
     ErtsMonotonicTime start_time;
 #ifdef DEBUG
@@ -5312,8 +1166,11 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
 	    reds_used = treds > INT_MAX ? INT_MAX : (int) treds;
 	}
 
+        if (c_p && ERTS_PROC_GET_PENDING_SUSPEND(c_p))
+            erts_proc_sig_handle_pending_suspend(c_p);
+
 	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+	ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
 	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
 	c_p = erts_schedule(esdp, c_p, reds_used);
 
@@ -5327,7 +1184,7 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
 #ifdef DEBUG
     pid = c_p->common.id; /* Save for debugging purposes */
 #endif
-    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
     PROCESS_MAIN_CHK_LOCKS(c_p);
 
     ASSERT(!(c_p->flags & F_HIPE_MODE));
@@ -5342,7 +1199,7 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
     else
 	c_p->fcalls = CONTEXT_REDS;
 
-    if (erts_smp_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_DIRTY_RUNNING_SYS) {
+    if (erts_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_DIRTY_RUNNING_SYS) {
 	erts_execute_dirty_system_task(c_p);
 	goto do_dirty_schedule;
     }
@@ -5376,7 +1233,7 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
             dtrace_proc_str(c_p, process_buf);
 
             if (ERTS_PROC_IS_EXITING(c_p)) {
-                strcpy(fun_buf, "<exiting>");
+                sys_strcpy(fun_buf, "<exiting>");
             } else {
                 ErtsCodeMFA *cmfa = find_function_from_pc(c_p->i);
                 if (cmfa) {
@@ -5413,21 +1270,21 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
 	c_p->current = codemfa;
 	SWAPOUT;
 	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+	ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
 
 	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
-	if (em_apply_bif == (BeamInstr *) *I) {
+	if (BeamIsOpCode(*I, op_apply_bif)) {
 	    exiting = erts_call_dirty_bif(esdp, c_p, I, reg);
 	}
 	else {
-	    ASSERT(em_call_nif == (BeamInstr *) *I);
-	    exiting = erts_call_dirty_nif(esdp, c_p, I, reg);
+	    ASSERT(BeamIsOpCode(*I, op_call_nif));
+            exiting = erts_call_dirty_nif(esdp, c_p, I, reg);
 	}
 
 	ASSERT(!(c_p->flags & F_HIBERNATE_SCHED));
 
 	PROCESS_MAIN_CHK_LOCKS(c_p);
-	ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+	ERTS_REQ_PROC_MAIN_LOCK(c_p);
 	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
 	ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
 	if (exiting)
@@ -5440,7 +1297,6 @@ void erts_dirty_process_main(ErtsSchedulerData *esdp)
 	I = c_p->i;
 	goto context_switch;
     }
-#endif /* ERTS_DIRTY_SCHEDULERS */
 }
 
 static ErtsCodeMFA *
@@ -5602,13 +1458,14 @@ handle_error(Process* c_p, BeamInstr* pc, Eterm* reg, ErtsCodeMFA *bif_mfa)
 	reg[3] = c_p->ftrace;
         if ((new_pc = next_catch(c_p, reg))) {
 	    c_p->cp = 0;	/* To avoid keeping stale references. */
+            ERTS_RECV_MARK_CLEAR(c_p); /* No longer safe to use this position */
 	    return new_pc;
 	}
 	if (c_p->catches > 0) erts_exit(ERTS_ERROR_EXIT, "Catch not found");
     }
-    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
     terminate_proc(c_p, Value);
-    ERTS_SMP_REQ_PROC_MAIN_LOCK(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
     return NULL;
 }
 
@@ -6257,8 +2114,8 @@ apply_bif_error_adjustment(Process *p, Export *ep,
      * and apply_last_IP.
      */
     if (I
-	&& ep->beam[0] == (BeamInstr) em_apply_bif
-	&& (ep == bif_export[BIF_error_1]
+	&& BeamIsOpCode(ep->beam[0], op_apply_bif)
+        && (ep == bif_export[BIF_error_1]
 	    || ep == bif_export[BIF_error_2]
 	    || ep == bif_export[BIF_exit_1]
 	    || ep == bif_export[BIF_throw_1])) {
@@ -6344,13 +2201,14 @@ apply_bif_error_adjustment(Process *p, Export *ep,
 }
 
 static BeamInstr*
-apply(
-Process* p, Eterm module, Eterm function, Eterm args, Eterm* reg,
-BeamInstr *I, Uint stack_offset)
+apply(Process* p, Eterm* reg, BeamInstr *I, Uint stack_offset)
 {
     int arity;
     Export* ep;
-    Eterm tmp, this;
+    Eterm tmp;
+    Eterm module = reg[0];
+    Eterm function = reg[1];
+    Eterm args = reg[2];
 
     /*
      * Check the arguments which should be of the form apply(Module,
@@ -6373,20 +2231,8 @@ BeamInstr *I, Uint stack_offset)
 
     while (1) {
 	Eterm m, f, a;
-	/* The module argument may be either an atom or an abstract module
-	 * (currently implemented using tuples, but this might change).
-	 */
-	this = THE_NON_VALUE;
-	if (is_not_atom(module)) {
-	    Eterm* tp;
-
-	    if (is_not_tuple(module)) goto error;
-	    tp = tuple_val(module);
-	    if (arityval(tp[0]) < 1) goto error;
-	    this = module;
-	    module = tp[1];
-	    if (is_not_atom(module)) goto error;
-	}
+
+	if (is_not_atom(module)) goto error;
 
 	if (module != am_erlang || function != am_apply)
 	    break;
@@ -6421,9 +2267,7 @@ BeamInstr *I, Uint stack_offset)
     }
     /*
      * Walk down the 3rd parameter of apply (the argument list) and copy
-     * the parameters to the x registers (reg[]). If the module argument
-     * was an abstract module, add 1 to the function arity and put the
-     * module argument in the n+1st x register as a THIS reference.
+     * the parameters to the x registers (reg[]).
      */
 
     tmp = args;
@@ -6440,9 +2284,6 @@ BeamInstr *I, Uint stack_offset)
     if (is_not_nil(tmp)) {	/* Must be well-formed list */
 	goto error;
     }
-    if (this != THE_NON_VALUE) {
-        reg[arity++] = this;
-    }
 
     /*
      * Get the index into the export table, or failing that the export
@@ -6481,22 +2322,12 @@ fixed_apply(Process* p, Eterm* reg, Uint arity,
 	return 0;
     }
 
-    /* The module argument may be either an atom or an abstract module
-     * (currently implemented using tuples, but this might change).
-     */
-    if (is_not_atom(module)) {
-	Eterm* tp;
-        if (is_not_tuple(module)) goto error;
-        tp = tuple_val(module);
-        if (arityval(tp[0]) < 1) goto error;
-        module = tp[1];
-        if (is_not_atom(module)) goto error;
-        ++arity;
-    }
+    if (is_not_atom(module)) goto error;
 
     /* Handle apply of apply/3... */
-    if (module == am_erlang && function == am_apply && arity == 3)
-	return apply(p, reg[0], reg[1], reg[2], reg, I, stack_offset);
+    if (module == am_erlang && function == am_apply && arity == 3) {
+	return apply(p, reg, I, stack_offset);
+    }
     
     /*
      * Get the index into the export table, or failing that the export
@@ -6517,27 +2348,13 @@ fixed_apply(Process* p, Eterm* reg, Uint arity,
 }
 
 int
-erts_hibernate(Process* c_p, Eterm module, Eterm function, Eterm args, Eterm* reg)
+erts_hibernate(Process* c_p, Eterm* reg)
 {
     int arity;
     Eterm tmp;
-
-#ifndef ERTS_SMP
-    if (ERTS_PROC_IS_EXITING(c_p)) {
-	/*
-	 * I non smp case:
-	 *
-	 * Currently executing process might be sent an exit
-	 * signal if it is traced by a port that it also is
-	 * linked to, and the port terminates during the
-	 * trace. In this case we do *not* want to clear
-	 * the active flag, which will make the process hang
-	 * in limbo forever. Get out of here and terminate
-	 * the process...
-	 */
-	return -1;
-    }
-#endif
+    Eterm module = reg[0];
+    Eterm function = reg[1];
+    Eterm args = reg[2];
 
     if (is_not_atom(module) || is_not_atom(function)) {
 	/*
@@ -6605,33 +2422,20 @@ erts_hibernate(Process* c_p, Eterm module, Eterm function, Eterm args, Eterm* re
      * If there are no waiting messages, garbage collect and
      * shrink the heap. 
      */
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
-    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-    if (!c_p->msg.len) {
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+    if (!erts_proc_sig_fetch(c_p)) {
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
 	c_p->fvalue = NIL;
 	PROCESS_MAIN_CHK_LOCKS(c_p);
 	erts_garbage_collect_hibernate(c_p);
 	ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
 	PROCESS_MAIN_CHK_LOCKS(c_p);
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
-#ifndef ERTS_SMP
-	if (ERTS_PROC_IS_EXITING(c_p)) {
-	    /*
-	     * See comment in the beginning of the function...
-	     *
-	     * This second test is needed since gc might be traced.
-	     */
-	    return -1;
-	}
-#else /* ERTS_SMP */
-        ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	if (!c_p->msg.len)
-#endif
-	    erts_smp_atomic32_read_band_relb(&c_p->state, ~ERTS_PSFLG_ACTIVE);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+	if (!erts_proc_sig_fetch(c_p))
+	    erts_atomic32_read_band_relb(&c_p->state, ~ERTS_PSFLG_ACTIVE);
 	ASSERT(!ERTS_PROC_IS_EXITING(c_p));
     }
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
     c_p->current = &bif_export[BIF_hibernate_3]->info.mfa;
     c_p->flags |= F_HIBERNATE_SCHED; /* Needed also when woken! */
     return 1;
@@ -6721,7 +2525,7 @@ call_fun(Process* p,		/* Current process. */
 
 		module = fe->module;
 
-		ERTS_SMP_READ_MEMORY_BARRIER;
+		ERTS_THR_READ_MEMORY_BARRIER;
 		if (fe->pend_purge_address) {
 		    /*
 		     * The system is currently trying to purge the
@@ -6852,7 +2656,7 @@ new_fun(Process* p, Eterm* reg, ErlFunEntry* fe, int num_free)
     p->htop = hp + needed;
     funp = (ErlFunThing *) hp;
     hp = funp->env;
-    erts_smp_refc_inc(&fe->refc, 2);
+    erts_refc_inc(&fe->refc, 2);
     funp->thing_word = HEADER_FUN;
     funp->next = MSO(p).first;
     MSO(p).first = (struct erl_off_heap_header*) funp;
@@ -6955,24 +2759,20 @@ do {						\
 
 
 static Eterm
-new_map(Process* p, Eterm* reg, BeamInstr* I)
+erts_gc_new_map(Process* p, Eterm* reg, Uint live, Uint n, BeamInstr* ptr)
 {
-    Uint n = Arg(3);
     Uint i;
     Uint need = n + 1 /* hdr */ + 1 /*size*/ + 1 /* ptr */ + 1 /* arity */;
     Eterm keys;
     Eterm *mhp,*thp;
     Eterm *E;
-    BeamInstr *ptr;
     flatmap_t *mp;
     ErtsHeapFactory factory;
 
-    ptr = &Arg(4);
-
     if (n > 2*MAP_SMALL_MAP_LIMIT) {
         Eterm res;
 	if (HeapWordsLeft(p) < n) {
-	    erts_garbage_collect(p, n, reg, Arg(2));
+	    erts_garbage_collect(p, n, reg, live);
 	}
 
 	mhp = p->htop;
@@ -6993,7 +2793,7 @@ new_map(Process* p, Eterm* reg, BeamInstr* I)
     }
 
     if (HeapWordsLeft(p) < need) {
-	erts_garbage_collect(p, need, reg, Arg(2));
+	erts_garbage_collect(p, need, reg, live);
     }
 
     thp    = p->htop;
@@ -7016,9 +2816,44 @@ new_map(Process* p, Eterm* reg, BeamInstr* I)
 }
 
 static Eterm
-update_map_assoc(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
+erts_gc_new_small_map_lit(Process* p, Eterm* reg, Eterm keys_literal,
+                          Uint live, BeamInstr* ptr)
+{
+    Eterm* keys = tuple_val(keys_literal);
+    Uint n = arityval(*keys);
+    Uint need = n + 1 /* hdr */ + 1 /*size*/ + 1 /* ptr */ + 1 /* arity */;
+    Uint i;
+    flatmap_t *mp;
+    Eterm *mhp;
+    Eterm *E;
+
+    ASSERT(n <= MAP_SMALL_MAP_LIMIT);
+
+    if (HeapWordsLeft(p) < need) {
+        erts_garbage_collect(p, need, reg, live);
+    }
+
+    mhp = p->htop;
+    E   = p->stop;
+
+    mp = (flatmap_t *)mhp; mhp += MAP_HEADER_FLATMAP_SZ;
+    mp->thing_word = MAP_HEADER_FLATMAP;
+    mp->size = n;
+    mp->keys = keys_literal;
+
+    for (i = 0; i < n; i++) {
+        GET_TERM(*ptr++, *mhp++);
+    }
+
+    p->htop = mhp;
+
+    return make_flatmap(mp);
+}
+
+static Eterm
+erts_gc_update_map_assoc(Process* p, Eterm* reg, Uint live,
+                         Uint n, BeamInstr* new_p)
 {
-    Uint n;
     Uint num_old;
     Uint num_updates;
     Uint need;
@@ -7028,23 +2863,18 @@ update_map_assoc(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
     Eterm* E;
     Eterm* old_keys;
     Eterm* old_vals;
-    BeamInstr* new_p;
     Eterm new_key;
     Eterm* kp;
+    Eterm map;
 
-    new_p = &Arg(5);
-    num_updates = Arg(4) / 2;
+    num_updates = n / 2;
+    map = reg[live];
 
     if (is_not_flatmap(map)) {
 	Uint32 hx;
 	Eterm val;
 
-	/* apparently the compiler does not emit is_map instructions,
-	 * bad compiler */
-
-	if (is_not_hashmap(map))
-	    return THE_NON_VALUE;
-
+	ASSERT(is_hashmap(map));
 	res = map;
 	E = p->stop;
 	while(num_updates--) {
@@ -7068,7 +2898,7 @@ update_map_assoc(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
      */
 
     if (num_old == 0) {
-	return new_map(p, reg, I+1);
+	return erts_gc_new_map(p, reg, live, n, new_p);
     }
 
     /*
@@ -7078,8 +2908,6 @@ update_map_assoc(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
 
     need = 2*(num_old+num_updates) + 1 + MAP_HEADER_FLATMAP_SZ;
     if (HeapWordsLeft(p) < need) {
-	Uint live = Arg(3);
-	reg[live] = map;
 	erts_garbage_collect(p, need, reg, live+1);
 	map      = reg[live];
 	old_mp   = (flatmap_t *)flatmap_val(map);
@@ -7226,9 +3054,8 @@ update_map_assoc(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
  */
 
 static Eterm
-update_map_exact(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
+erts_gc_update_map_exact(Process* p, Eterm* reg, Uint live, Uint n, Eterm* new_p)
 {
-    Uint n;
     Uint i;
     Uint num_old;
     Uint need;
@@ -7238,12 +3065,12 @@ update_map_exact(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
     Eterm* E;
     Eterm* old_keys;
     Eterm* old_vals;
-    BeamInstr* new_p;
     Eterm new_key;
+    Eterm map;
 
-    new_p = &Arg(5);
-    n = Arg(4) / 2;		/* Number of values to be updated */
+    n /= 2;		/* Number of values to be updated */
     ASSERT(n > 0);
+    map = reg[live];
 
     if (is_not_flatmap(map)) {
 	Uint32 hx;
@@ -7297,8 +3124,6 @@ update_map_exact(Process* p, Eterm* reg, Eterm map, BeamInstr* I)
 
     need = num_old + MAP_HEADER_FLATMAP_SZ;
     if (HeapWordsLeft(p) < need) {
-	Uint live = Arg(3);
-	reg[live] = map;
 	erts_garbage_collect(p, need, reg, live+1);
 	map      = reg[live];
 	old_mp   = (flatmap_t *)flatmap_val(map);
@@ -7380,13 +3205,16 @@ erts_is_builtin(Eterm Mod, Eterm Name, int arity)
     Export e;
     Export* ep;
 
-    if (Mod == am_erlang && Name == am_apply && arity == 3) {
-	/*
-	 * Special case. apply/3 is built-in (implemented in C),
-	 * but implemented in a different way than all other
-	 * BIFs.
-	 */
-	return 1;
+    if (Mod == am_erlang) {
+        /*
+         * Special case for built-in functions that are implemented
+         * as instructions as opposed to SNIFs.
+         */
+        if (Name == am_apply && (arity == 2 || arity == 3)) {
+            return 1;
+        } else if (Name == am_yield && arity == 0) {
+            return 1;
+        }
     }
 
     e.info.mfa.module = Mod;
@@ -7396,8 +3224,8 @@ erts_is_builtin(Eterm Mod, Eterm Name, int arity)
     if ((ep = export_get(&e)) == NULL) {
 	return 0;
     }
-    return ep->addressv[erts_active_code_ix()] == ep->beam
-	&& (ep->beam[0] == (BeamInstr) em_apply_bif);
+    return ep->addressv[erts_active_code_ix()] == ep->beam &&
+	BeamIsOpCode(ep->beam[0], op_apply_bif);
 }
 
 
diff --git a/erts/emulator/beam/beam_load.c b/erts/emulator/beam/beam_load.c
index a4e226d459..e61199a8fd 100644
--- a/erts/emulator/beam/beam_load.c
+++ b/erts/emulator/beam/beam_load.c
@@ -56,12 +56,6 @@ ErlDrvBinary* erts_gzinflate_buffer(char*, int);
 #define DEFINED   1
 #define EXPORTED  2
 
-#ifdef NO_JUMP_TABLE
-#  define BeamOpCode(Op) ((BeamInstr)(Op))
-#else
-#  define BeamOpCode(Op) ((BeamInstr)beam_ops[Op])
-#endif
-
 #if defined(WORDS_BIGENDIAN)
 # define NATIVE_ENDIAN(F)			\
   if ((F).val & BSF_NATIVE) {			\
@@ -82,16 +76,28 @@ ErlDrvBinary* erts_gzinflate_buffer(char*, int);
 #define TE_FAIL (-1)
 #define TE_SHORT_WINDOW (-2)
 
+/*
+ * Type for a reference to a label that must be patched.
+ */
+
 typedef struct {
-    Uint value;			/* Value of label (NULL if not known yet). */
-    Sint patches;		/* Index (into code buffer) to first location
-				 * which must be patched with the value of this label.
-				 */
-#ifdef ERTS_SMP
+    Uint pos;                   /* Position of label reference to patch. */
+    Uint offset;                /* Offset from patch location.  */
+    int packed;                 /* 0 (not packed), 1 (lsw), 2 (msw) */
+} LabelPatch;
+
+/*
+ * Type for a label.
+ */
+
+typedef struct {
+    Uint value;			/* Value of label (0 if not known yet). */
     Uint looprec_targeted;	/* Non-zero if this label is the target of a loop_rec
 				 * instruction.
 				 */
-#endif
+    LabelPatch* patches;       /* Array of label patches. */
+    Uint num_patches;           /* Number of patches in array. */
+    Uint num_allocated;         /* Number of allocated patches. */
 } Label;
 
 /*
@@ -228,7 +234,7 @@ typedef struct {
 
 typedef struct literal_patch LiteralPatch;
 struct literal_patch {
-    int pos;			/* Position in code */
+    Uint pos;			/* Position in code */
     LiteralPatch* next;
 };
 
@@ -308,6 +314,7 @@ typedef struct LoaderState {
     int on_load;		/* Index in the code for the on_load function
 				 * (or 0 if there is no on_load function)
 				 */
+    int otp_20_or_higher;       /* Compiled with OTP 20 or higher */
 
     /*
      * Atom table.
@@ -452,7 +459,7 @@ typedef struct LoaderState {
 
 #ifdef DEBUG
 # define GARBAGE 0xCC
-# define DEBUG_INIT_GENOP(Dst) memset(Dst, GARBAGE, sizeof(GenOp))
+# define DEBUG_INIT_GENOP(Dst) sys_memset(Dst, GARBAGE, sizeof(GenOp))
 #else
 # define DEBUG_INIT_GENOP(Dst)
 #endif
@@ -508,6 +515,7 @@ static int read_lambda_table(LoaderState* stp);
 static int read_literal_table(LoaderState* stp);
 static int read_line_table(LoaderState* stp);
 static int read_code_header(LoaderState* stp);
+static void init_label(Label* lp);
 static int load_code(LoaderState* stp);
 static GenOp* gen_element(LoaderState* stp, GenOpArg Fail, GenOpArg Index,
 			  GenOpArg Tuple, GenOpArg Dst);
@@ -538,6 +546,7 @@ static int get_tag_and_value(LoaderState* stp, Uint len_code,
 static int new_label(LoaderState* stp);
 static void new_literal_patch(LoaderState* stp, int pos);
 static void new_string_patch(LoaderState* stp, int pos);
+static int find_literal(LoaderState* stp, Eterm needle, Uint *idx);
 static Uint new_literal(LoaderState* stp, Eterm** hpp, Uint heap_size);
 static int genopargcompare(GenOpArg* a, GenOpArg* b);
 static Eterm get_module_info(Process* p, ErtsCodeIndex code_ix,
@@ -545,6 +554,7 @@ static Eterm get_module_info(Process* p, ErtsCodeIndex code_ix,
 static Eterm exported_from_module(Process* p, ErtsCodeIndex code_ix,
                                   Eterm mod);
 static Eterm functions_in_module(Process* p, BeamCodeHeader*);
+static Eterm nifs_in_module(Process* p, Eterm module);
 static Eterm attributes_for_module(Process* p, BeamCodeHeader*);
 static Eterm compilation_info_for_module(Process* p, BeamCodeHeader*);
 static Eterm md5_of_module(Process* p, BeamCodeHeader*);
@@ -741,6 +751,13 @@ erts_prepare_loading(Binary* magic, Process *c_p, Eterm group_leader,
     }
 
     /*
+     * Find out whether the code was compiled with OTP 20
+     * or higher.
+     */
+
+    stp->otp_20_or_higher = stp->chunks[UTF8_ATOM_CHUNK].size > 0;
+
+    /*
      * Load the code chunk.
      */
 
@@ -791,13 +808,8 @@ erts_finish_loading(Binary* magic, Process* c_p,
     struct erl_module_instance* inst_p;
     Uint size;
 
-    /*
-     * No other process may run since we will update the export
-     * table which is not protected by any locks.
-     */
-
-    ERTS_SMP_LC_ASSERT(erts_initialized == 0 || erts_has_code_write_permission() ||
-		       erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_initialized == 0 || erts_has_code_write_permission() ||
+		       erts_thr_progress_is_blocking());
     /*
      * Make current code for the module old and insert the new code
      * as current.  This will fail if there already exists old code
@@ -828,11 +840,10 @@ erts_finish_loading(Binary* magic, Process* c_p,
 		continue;
 	    }
 	    if (ep->addressv[code_ix] == ep->beam) {
-		if (ep->beam[0] == (BeamInstr) em_apply_bif) {
+		if (BeamIsOpCode(ep->beam[0], op_apply_bif)) {
 		    continue;
-		} else if (ep->beam[0] ==
-			   (BeamInstr) BeamOp(op_i_generic_breakpoint)) {
-		    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+		} else if (BeamIsOpCode(ep->beam[0], op_i_generic_breakpoint)) {
+		    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 		    ASSERT(mod_tab_p->curr.num_traced_exports > 0);
 		    erts_clear_export_break(mod_tab_p, &ep->info);
 		    ep->addressv[code_ix] = (BeamInstr *) ep->beam[1];
@@ -1044,6 +1055,10 @@ loader_state_dtor(Binary* magic)
         stp->codev = 0;
     }
     if (stp->labels != 0) {
+        Uint num;
+        for (num = 0; num < stp->num_labels; num++) {
+            erts_free(ERTS_ALC_T_PREPARED_CODE, (void *) stp->labels[num].patches);
+        }
 	erts_free(ERTS_ALC_T_PREPARED_CODE, (void *) stp->labels);
 	stp->labels = 0;
     }
@@ -1408,7 +1423,7 @@ load_atom_table(LoaderState* stp, ErtsAtomEncoding enc)
 	Atom* ap;
 
 	ap = atom_tab(atom_val(stp->atom[1]));
-	memcpy(sbuf, ap->name, ap->len);
+	sys_memcpy(sbuf, ap->name, ap->len);
 	sbuf[ap->len] = '\0';
 	LoadError1(stp, "module name in object code is %s", sbuf);
     }
@@ -1457,7 +1472,7 @@ load_import_table(LoaderState* stp)
 	 * the BIF function.
 	 */
 	if ((e = erts_active_export_entry(mod, func, arity)) != NULL) {
-	    if (e->beam[0] == (BeamInstr) em_apply_bif) {
+	    if (BeamIsOpCode(e->beam[0], op_apply_bif)) {
 		stp->import[i].bf = (BifFunction) e->beam[1];
 		if (func == am_load_nif && mod == am_erlang && arity == 2) {
 		    stp->may_load_nif = 1;
@@ -1527,7 +1542,7 @@ read_export_table(LoaderState* stp)
 	 * any other functions that walk through all local functions.
 	 */
 
-	if (stp->labels[n].patches >= 0) {
+	if (stp->labels[n].num_patches > 0) {
 	    LoadError3(stp, "there are local calls to the stub for "
 		       "the BIF %T:%T/%d",
 		       stp->module, func, arity);
@@ -1551,7 +1566,7 @@ is_bif(Eterm mod, Eterm func, unsigned arity)
     if (e == NULL) {
 	return 0;
     }
-    if (e->beam[0] != (BeamInstr) em_apply_bif) {
+    if (! BeamIsOpCode(e->beam[0], op_apply_bif)) {
 	return 0;
     }
     if (mod == am_erlang && func == am_apply && arity == 3) {
@@ -1873,11 +1888,7 @@ read_code_header(LoaderState* stp)
     stp->labels = (Label *) erts_alloc(ERTS_ALC_T_PREPARED_CODE,
 				       stp->num_labels * sizeof(Label));
     for (i = 0; i < stp->num_labels; i++) {
-	stp->labels[i].value = 0;
-	stp->labels[i].patches = -1;
-#ifdef ERTS_SMP
-	stp->labels[i].looprec_targeted = 0;
-#endif
+        init_label(&stp->labels[i]);
     }
 
     stp->catches = 0;
@@ -1906,12 +1917,43 @@ read_code_header(LoaderState* stp)
     
 #define TermWords(t) (((t) / (sizeof(BeamInstr)/sizeof(Eterm))) + !!((t) % (sizeof(BeamInstr)/sizeof(Eterm))))
 
+static void init_label(Label* lp)
+{
+    lp->value = 0;
+    lp->looprec_targeted = 0;
+    lp->num_patches = 0;
+    lp->num_allocated = 4;
+    lp->patches = erts_alloc(ERTS_ALC_T_PREPARED_CODE,
+                             lp->num_allocated * sizeof(LabelPatch));
+}
+
+static void
+register_label_patch(LoaderState* stp, Uint label, Uint ci, Uint offset)
+{
+    Label* lp;
+
+    ASSERT(label < stp->num_labels);
+    lp = &stp->labels[label];
+    if (lp->num_allocated <= lp->num_patches) {
+        lp->num_allocated *= 2;
+        lp->patches = erts_realloc(ERTS_ALC_T_PREPARED_CODE,
+                                   (void *) lp->patches,
+                                   lp->num_allocated * sizeof(LabelPatch));
+    }
+    lp->patches[lp->num_patches].pos = ci;
+    lp->patches[lp->num_patches].offset = offset;
+    lp->patches[lp->num_patches].packed = 0;
+    lp->num_patches++;
+    stp->codev[ci] = label;
+}
+
 static int
 load_code(LoaderState* stp)
 {
     int i;
-    int ci;
-    int last_func_start = 0;	/* Needed by nif loading and line instructions */
+    Uint ci;
+    Uint last_instr_start;       /* Needed for relative jumps */
+    Uint last_func_start = 0;	/* Needed by nif loading and line instructions */
     char* sign;
     int arg;			/* Number of current argument. */
     int num_specific;		/* Number of specific ops for current. */
@@ -1924,6 +1966,9 @@ load_code(LoaderState* stp)
     GenOp** last_op_next = NULL;
     int arity;
     int retval = 1;
+#if defined(BEAM_WIDE_SHIFT)
+    int num_trailing_f;     /* Number of extra 'f' arguments in a list */
+#endif
 
     /*
      * The size of the loaded func_info instruction is needed
@@ -2029,30 +2074,10 @@ load_code(LoaderState* stp)
 		    case 0:
 			/* Floating point number.
 			 * Not generated by the compiler in R16B and later.
+                         * (The literal pool is used instead.)
 			 */
-			{
-			    Eterm* hp;
-#if !defined(ARCH_64)
-			    Uint high, low;
-# endif
-			    last_op->a[arg].val = new_literal(stp, &hp,
-							      FLOAT_SIZE_OBJECT);
-			    hp[0] = HEADER_FLONUM;
-			    last_op->a[arg].type = TAG_q;
-#if defined(ARCH_64)
-			    GetInt(stp, 8, hp[1]);
-# else
-			    GetInt(stp, 4, high);
-			    GetInt(stp, 4, low);
-			    if (must_swap_floats) {
-				Uint t = high;
-				high = low;
-				low = t;
-			    }
-			    hp[1] = high;
-			    hp[2] = low;
-# endif
-			}
+                        LoadError0(stp, "please re-compile this module with an "
+                                   ERLANG_OTP_RELEASE " compiler");
 			break;
 		    case 1:	/* List. */
 			if (arg+1 != arity) {
@@ -2065,7 +2090,7 @@ load_code(LoaderState* stp)
 			    erts_alloc(ERTS_ALC_T_LOADER_TMP,
 				       (arity+last_op->a[arg].val)
 				       *sizeof(GenOpArg));
-			memcpy(last_op->a, last_op->def_args,
+			sys_memcpy(last_op->a, last_op->def_args,
 			       arity*sizeof(GenOpArg));
 			arity += last_op->a[arg].val;
 			break;
@@ -2287,7 +2312,8 @@ load_code(LoaderState* stp)
 
 	    stp->specific_op = specific;
 	    CodeNeed(opc[stp->specific_op].sz+16); /* Extra margin for packing */
-	    code[ci++] = BeamOpCode(stp->specific_op);
+            last_instr_start = ci + opc[stp->specific_op].adjust;
+	    code[ci++] = BeamOpCodeAddr(stp->specific_op);
 	}
 	
 	/*
@@ -2360,8 +2386,18 @@ load_code(LoaderState* stp)
 		    code[ci++] = NIL;
 		    break;
 		case TAG_q:
-		    new_literal_patch(stp, ci);
-		    code[ci++] = tmp_op->a[arg].val;
+                    {
+                        BeamInstr val = tmp_op->a[arg].val;
+                        Eterm term = stp->literals[val].term;
+                        new_literal_patch(stp, ci);
+                        code[ci++] = val;
+                        switch (loader_tag(term)) {
+                        case LOADER_X_REG:
+                        case LOADER_Y_REG:
+                            LoadError1(stp, "the term '%T' would be confused "
+                                       "with a register", term);
+                        }
+                    }
 		    break;
 		default:
 		    LoadError1(stp, "bad tag %d for general source",
@@ -2369,7 +2405,8 @@ load_code(LoaderState* stp)
 		    break;
 		}
 		break;
-	    case 'd':	/* Destination (x(0), x(N), y(N) */
+	    case 'd':	/* Destination (x(N), y(N) */
+            case 'S':   /* Source (x(N), y(N)) */
 		switch (tag) {
 		case TAG_x:
 		    code[ci++] = tmp_op->a[arg].val * sizeof(Eterm);
@@ -2383,11 +2420,29 @@ load_code(LoaderState* stp)
 		    break;
 		}
 		break;
-	    case 'I':	/* Untagged integer (or pointer). */
-		VerifyTag(stp, tag, TAG_u);
-		code[ci++] = tmp_op->a[arg].val;
-		break;
-	    case 't':	/* Small untagged integer -- can be packed. */
+	    case 't':	/* Small untagged integer (16 bits) -- can be packed. */
+	    case 'I':	/* Untagged integer (32 bits) -- can be packed.  */
+	    case 'W':	/* Untagged integer or pointer (machine word). */
+#ifdef DEBUG
+                switch (*sign) {
+                case 't':
+                    if (tmp_op->a[arg].val >> 16 != 0) {
+                        load_printf(__LINE__, stp, "value %lu of type 't' does not fit in 16 bits",
+                                    tmp_op->a[arg].val);
+                        ASSERT(0);
+                    }
+                    break;
+#ifdef ARCH_64
+                case 'I':
+                    if (tmp_op->a[arg].val >> 32 != 0) {
+                        load_printf(__LINE__, stp, "value %lu of type 'I' does not fit in 32 bits",
+                                    tmp_op->a[arg].val);
+                        ASSERT(0);
+                    }
+                    break;
+#endif
+                }
+#endif
 		VerifyTag(stp, tag, TAG_u);
 		code[ci++] = tmp_op->a[arg].val;
 		break;
@@ -2397,16 +2452,14 @@ load_code(LoaderState* stp)
 		break;
 	    case 'f':		/* Destination label */
 		VerifyTag(stp, tag_to_letter[tag], *sign);
-		code[ci] = stp->labels[tmp_op->a[arg].val].patches;
-		stp->labels[tmp_op->a[arg].val].patches = ci;
+                register_label_patch(stp, tmp_op->a[arg].val, ci, -last_instr_start);
 		ci++;
 		break;
 	    case 'j':		/* 'f' or 'p' */
 		if (tag == TAG_p) {
 		    code[ci] = 0;
 		} else if (tag == TAG_f) {
-		    code[ci] = stp->labels[tmp_op->a[arg].val].patches;
-		    stp->labels[tmp_op->a[arg].val].patches = ci;
+                    register_label_patch(stp, tmp_op->a[arg].val, ci, -last_instr_start);
 		} else {
 		    LoadError3(stp, "bad tag %d; expected %d or %d",
 			       tag, TAG_f, TAG_p);
@@ -2426,7 +2479,6 @@ load_code(LoaderState* stp)
 		    LoadError1(stp, "label %d defined more than once", last_label);
 		}
 		stp->labels[last_label].value = ci;
-		ASSERT(stp->labels[last_label].patches < ci);
 		break;
 	    case 'e':		/* Export entry */
 		VerifyTag(stp, tag, TAG_u);
@@ -2472,39 +2524,168 @@ load_code(LoaderState* stp)
 	 * The packing engine.
 	 */
 	if (opc[stp->specific_op].pack[0]) {
-	    char* prog;		/* Program for packing engine. */
-	    BeamInstr stack[8];	/* Stack. */
-	    BeamInstr* sp = stack;	/* Points to next free position. */
-	    BeamInstr packed = 0;	/* Accumulator for packed operations. */
+	    char* prog;            /* Program for packing engine. */
+	    struct pack_stack {
+                BeamInstr instr;
+                Uint* patch_pos;
+            } stack[8];            /* Stack. */
+	    struct pack_stack* sp = stack; /* Points to next free position. */
+	    BeamInstr packed = 0; /* Accumulator for packed operations. */
+            LabelPatch* packed_label = 0;
 
 	    for (prog = opc[stp->specific_op].pack; *prog; prog++) {
 		switch (*prog) {
-		case 'g':	/* Get instruction; push on stack. */
-		    *sp++ = code[--ci];
-		    break;
-		case 'i':	/* Initialize packing accumulator. */
-		    packed = code[--ci];
+		case 'g':	/* Get operand and push on stack. */
+                    ci--;
+                    sp->instr = code[ci];
+                    sp->patch_pos = 0;
+                    sp++;
+                    break;
+		case 'f':	/* Get possible 'f' operand and push on stack. */
+                    {
+                        Uint w = code[--ci];
+                        sp->instr = w;
+                        sp->patch_pos = 0;
+
+                        if (w != 0) {
+                            LabelPatch* lbl_p;
+                            int num_patches;
+                            int patch;
+
+                            ASSERT(w < stp->num_labels);
+                            lbl_p = stp->labels[w].patches;
+                            num_patches = stp->labels[w].num_patches;
+                            for (patch = num_patches - 1; patch >= 0; patch--) {
+                                if (lbl_p[patch].pos == ci) {
+                                    sp->patch_pos = &lbl_p[patch].pos;
+                                    break;
+                                }
+                            }
+                            ASSERT(sp->patch_pos);
+                        }
+                        sp++;
+                    }
+                    break;
+		case 'q':	/* Get possible 'q' operand and push on stack. */
+                    {
+                        LiteralPatch* lp;
+
+                        ci--;
+                        sp->instr = code[ci];
+                        sp->patch_pos = 0;
+
+                        for (lp = stp->literal_patches;
+                             lp && lp->pos > ci-MAX_OPARGS;
+                             lp = lp->next) {
+                            if (lp->pos == ci) {
+                                sp->patch_pos = &lp->pos;
+                                break;
+                            }
+                        }
+                        sp++;
+                    }
+                    break;
+#ifdef ARCH_64
+		case '1':	/* Tightest shift (always 10 bits) */
+                    ci--;
+                    ASSERT((code[ci] & ~0x1FF8ull) == 0); /* Fits in 10 bits */
+                    packed = (packed << BEAM_TIGHTEST_SHIFT);
+                    packed |= code[ci] >> 3;
+                    if (packed_label) {
+                        packed_label->packed++;
+                    }
 		    break;
-		case '0':	/* Tight shift */
+#endif
+		case '2':	/* Tight shift (10 or 16 bits) */
 		    packed = (packed << BEAM_TIGHT_SHIFT) | code[--ci];
+                    if (packed_label) {
+                        packed_label->packed++;
+                    }
 		    break;
-		case '6':	/* Shift 16 steps */
+		case '3':	/* Loose shift (16 bits) */
 		    packed = (packed << BEAM_LOOSE_SHIFT) | code[--ci];
+                    if (packed_label) {
+                        packed_label->packed++;
+                    }
 		    break;
 #ifdef ARCH_64
-		case 'w':	/* Shift 32 steps */
-		    packed = (packed << BEAM_WIDE_SHIFT) | code[--ci];
-		    break;
+		case '4':	/* Wide shift (32 bits) */
+                    {
+                        Uint w = code[--ci];
+
+                        if (packed_label) {
+                            packed_label->packed++;
+                        }
+
+                        /*
+                         * 'w' can handle both labels ('f' and 'j'), as well
+                         * as 'I'. Test whether this is a label.
+                         */
+
+                        if (w < stp->num_labels) {
+                            /*
+                             * Probably a label. Look for patch pointing to this
+                             * position.
+                             */
+                            LabelPatch* lp = stp->labels[w].patches;
+                            int num_patches = stp->labels[w].num_patches;
+                            int patch;
+                            for (patch = num_patches - 1; patch >= 0; patch--) {
+                                if (lp[patch].pos == ci) {
+                                    lp[patch].packed = 1;
+                                    packed_label = &lp[patch];
+                                    break;
+                                }
+                            }
+                        }
+                        packed = (packed << BEAM_WIDE_SHIFT) |
+                            (code[ci] & BEAM_WIDE_MASK);
+                    }
+                    break;
 #endif
 		case 'p':	/* Put instruction (from stack). */
-		    code[ci++] = *--sp;
+                    --sp;
+                    code[ci] = sp->instr;
+                    if (sp->patch_pos) {
+                        *sp->patch_pos = ci;
+                    }
+                    ci++;
 		    break;
-		case 'P':	/* Put packed operands. */
-		    *sp++ = packed;
+		case 'P':	/* Put packed operands (on the stack). */
+                    sp->instr = packed;
+                    sp->patch_pos = 0;
+                    if (packed_label) {
+                        sp->patch_pos = &packed_label->pos;
+                        packed_label = 0;
+                    }
+                    sp++;
 		    packed = 0;
 		    break;
+#if defined(ARCH_64) && defined(CODE_MODEL_SMALL)
+                case '#':       /* -1 */
+                case '$':       /* -2 */
+                case '%':       /* -3 */
+                case '&':       /* -4 */
+                case '\'':      /* -5 */
+                case '(':       /* -6 */
+                    /* Pack accumulator contents into instruction word. */
+                    {
+                        Sint pos = ci - (*prog - '#' + 1);
+                        /* Are the high 32 bits of the instruction word zero? */
+                        ASSERT((code[pos] & ~((1ull << BEAM_WIDE_SHIFT)-1)) == 0);
+                        code[pos] |= packed << BEAM_WIDE_SHIFT;
+                        if (packed_label) {
+                            ASSERT(packed_label->packed == 1);
+                            packed_label->pos = pos;
+                            packed_label->packed = 2;
+                            packed_label = 0;
+                        }
+                        packed >>= BEAM_WIDE_SHIFT;
+                    }
+		    break;
+#endif
 		default:
-		    ASSERT(0);
+                    erts_exit(ERTS_ERROR_EXIT, "beam_load: invalid packing op: %c\n", *prog);
 		}
 	    }
 	    ASSERT(sp == stack); /* Incorrect program? */
@@ -2514,7 +2695,17 @@ load_code(LoaderState* stp)
 	 * Load any list arguments using the primitive tags.
 	 */
 
+#if defined(BEAM_WIDE_SHIFT)
+        num_trailing_f = 0;
+#endif
 	for ( ; arg < tmp_op->arity; arg++) {
+#if defined(BEAM_WIDE_SHIFT)
+	    if (tmp_op->a[arg].type == TAG_f) {
+                num_trailing_f++;
+            } else {
+                num_trailing_f = 0;
+            }
+#endif
 	    switch (tmp_op->a[arg].type) {
 	    case TAG_i:
 		CodeNeed(1);
@@ -2528,8 +2719,7 @@ load_code(LoaderState* stp)
 		break;
 	    case TAG_f:
 		CodeNeed(1);
-		code[ci] = stp->labels[tmp_op->a[arg].val].patches;
-		stp->labels[tmp_op->a[arg].val].patches = ci;
+                register_label_patch(stp, tmp_op->a[arg].val, ci, -last_instr_start);
 		ci++;
 		break;
 	    case TAG_x:
@@ -2555,6 +2745,61 @@ load_code(LoaderState* stp)
 	    }
 	}
 
+        /*
+         * If all the extra arguments were 'f' operands,
+         * and the wordsize is 64 bits, pack two 'f' operands
+         * into each word.
+         */
+
+#if defined(BEAM_WIDE_SHIFT)
+        if (num_trailing_f >= 1) {
+            Uint src_index = ci - num_trailing_f;
+            Uint src_limit = ci;
+            Uint dst_limit = src_index + (num_trailing_f+1)/2;
+
+            ci = src_index;
+            while (ci < dst_limit) {
+                Uint w[2];
+                BeamInstr packed = 0;
+                int wi;
+
+                w[0] = code[src_index];
+                if (src_index+1 < src_limit) {
+                    w[1] = code[src_index+1];
+                } else {
+                    w[1] = 0;
+                }
+                for (wi = 0; wi < 2; wi++) {
+                    Uint lbl = w[wi];
+                    LabelPatch* lp = stp->labels[lbl].patches;
+                    int num_patches = stp->labels[lbl].num_patches;
+
+#if defined(WORDS_BIGENDIAN)
+                    packed <<= BEAM_WIDE_SHIFT;
+                    packed |= lbl & BEAM_WIDE_MASK;
+#else
+                    packed >>= BEAM_WIDE_SHIFT;
+                    packed |= lbl << BEAM_WIDE_SHIFT;
+#endif
+                    while (num_patches-- > 0) {
+                        if (lp->pos == src_index + wi) {
+                            lp->pos = ci;
+#if defined(WORDS_BIGENDIAN)
+                            lp->packed = 2 - wi;
+#else
+                            lp->packed = wi + 1;
+#endif
+                            break;
+                        }
+                        lp++;
+                    }
+                }
+                code[ci++] = packed;
+                src_index += 2;
+            }
+        }
+#endif
+
 	/*
 	 * Handle a few special cases.
 	 */
@@ -2601,17 +2846,16 @@ load_code(LoaderState* stp)
                    the size of the ops.tab i_func_info instruction is not
                    the same as FUNC_INFO_SZ */
 		ASSERT(stp->labels[last_label].value == ci - FUNC_INFO_SZ);
-		stp->hdr->functions[function_number] = (ErtsCodeInfo*) stp->labels[last_label].patches;
 		offset = function_number;
-		stp->labels[last_label].patches = offset;
+                register_label_patch(stp, last_label, offset, 0);
 		function_number++;
 		if (stp->arity > MAX_ARG) {
 		    LoadError1(stp, "too many arguments: %d", stp->arity);
 		}
 #ifdef DEBUG
-		ASSERT(stp->labels[0].patches < 0); /* Should not be referenced. */
+		ASSERT(stp->labels[0].num_patches == 0); /* Should not be referenced. */
 		for (i = 1; i < stp->num_labels; i++) {
-		    ASSERT(stp->labels[i].patches < ci);
+                    ASSERT(stp->labels[i].num_patches <= stp->labels[i].num_allocated);
 		}
 #endif
 	    }
@@ -2622,8 +2866,8 @@ load_code(LoaderState* stp)
 	    /* Remember offset for the on_load function. */
 	    stp->on_load = ci;
 	    break;
-	case op_bs_put_string_II:
-	case op_i_bs_match_string_xfII:
+	case op_bs_put_string_WW:
+	case op_i_bs_match_string_xfWW:
 	    new_string_patch(stp, ci-1);
 	    break;
 
@@ -2734,6 +2978,12 @@ load_code(LoaderState* stp)
 
 #define never(St) 0
 
+static int
+compiled_with_otp_20_or_higher(LoaderState* stp)
+{
+    return stp->otp_20_or_higher;
+}
+
 /*
  * Predicate that tests whether a jump table can be used.
  */
@@ -2873,17 +3123,18 @@ gen_element(LoaderState* stp, GenOpArg Fail, GenOpArg Index,
     op->next = NULL;
 
     if (Index.type == TAG_i && Index.val > 0 &&
+        Index.val <= ERTS_MAX_TUPLE_SIZE &&
 	(Tuple.type == TAG_x || Tuple.type == TAG_y)) {
 	op->op = genop_i_fast_element_4;
-	op->a[0] = Fail;
-	op->a[1] = Tuple;
+	op->a[0] = Tuple;
+	op->a[1] = Fail;
 	op->a[2].type = TAG_u;
 	op->a[2].val = Index.val;
 	op->a[3] = Dst;
     } else {
 	op->op = genop_i_element_4;
-	op->a[0] = Fail;
-	op->a[1] = Tuple;
+	op->a[0] = Tuple;
+	op->a[1] = Fail;
 	op->a[2] = Index;
 	op->a[3] = Dst;
     }
@@ -2963,13 +3214,14 @@ gen_get_integer2(LoaderState* stp, GenOpArg Fail, GenOpArg Ms, GenOpArg Live,
 	    op->a[0] = Ms;
 	    op->a[1] = Fail;
 	    op->a[2] = Dst;
+#ifdef ARCH_64
 	} else if (bits == 32 && (Flags.val & BSF_LITTLE) == 0) {
-	    op->op = genop_i_bs_get_integer_32_4;
-	    op->arity = 4;
+	    op->op = genop_i_bs_get_integer_32_3;
+	    op->arity = 3;
 	    op->a[0] = Ms;
 	    op->a[1] = Fail;
-	    op->a[2] = Live;
-	    op->a[3] = Dst;
+	    op->a[2] = Dst;
+#endif
 	} else {
 	generic:
 	    if (bits < SMALL_BITS) {
@@ -3104,16 +3356,6 @@ gen_get_binary2(LoaderState* stp, GenOpArg Fail, GenOpArg Ms, GenOpArg Live,
 }
 
 /*
- * Predicate to test whether a heap binary should be generated.
- */
-
-static int
-should_gen_heap_bin(LoaderState* stp, GenOpArg Src)
-{
-    return Src.val <= ERL_ONHEAP_BIN_LIMIT;
-}
-
-/*
  * Predicate to test whether a binary construction is too big.
  */
 
@@ -3385,27 +3627,14 @@ negation_is_small(LoaderState* stp, GenOpArg Int)
            IS_SSMALL(-((Sint)Int.val));
 }
 
-
-static int
-smp(LoaderState* stp)
-{
-#ifdef ERTS_SMP
-    return 1;
-#else
-    return 0;
-#endif
-}
-
 /*
  * Mark this label.
  */
 static int
 smp_mark_target_label(LoaderState* stp, GenOpArg L)
 {
-#ifdef ERTS_SMP
     ASSERT(L.type == TAG_f);
     stp->labels[L.val].looprec_targeted = 1;
-#endif
     return 1;
 }
 
@@ -3416,12 +3645,8 @@ smp_mark_target_label(LoaderState* stp, GenOpArg L)
 static int
 smp_already_locked(LoaderState* stp, GenOpArg L)
 {
-#ifdef ERTS_SMP
     ASSERT(L.type == TAG_u);
     return stp->labels[L.val].looprec_targeted;
-#else
-    return 0;
-#endif
 }
 
 /*
@@ -3435,11 +3660,11 @@ gen_literal_timeout(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
     Sint timeout;
 
     NEW_GENOP(stp, op);
-    op->op = genop_i_wait_timeout_2;
+    op->op = genop_wait_timeout_unlocked_int_2;
     op->next = NULL;
     op->arity = 2;
-    op->a[0] = Fail;
-    op->a[1].type = TAG_u;
+    op->a[0].type = TAG_u;
+    op->a[1] = Fail;
     
     if (Time.type == TAG_i && (timeout = Time.val) >= 0 &&
 #if defined(ARCH_64)
@@ -3448,7 +3673,7 @@ gen_literal_timeout(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
 	1
 #endif
 	) {
-	op->a[1].val = timeout;
+	op->a[0].val = timeout;
 #if !defined(ARCH_64)
     } else if (Time.type == TAG_q) {
 	Eterm big;
@@ -3462,7 +3687,7 @@ gen_literal_timeout(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
 	} else {
 	    Uint u;
 	    (void) term_to_Uint(big, &u);
-	    op->a[1].val = (BeamInstr) u;
+	    op->a[0].val = (BeamInstr) u;
 	}
 #endif
     } else {
@@ -3482,12 +3707,12 @@ gen_literal_timeout_locked(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
     Sint timeout;
 
     NEW_GENOP(stp, op);
-    op->op = genop_i_wait_timeout_locked_2;
+    op->op = genop_wait_timeout_locked_int_2;
     op->next = NULL;
     op->arity = 2;
-    op->a[0] = Fail;
-    op->a[1].type = TAG_u;
-    
+    op->a[0].type = TAG_u;
+    op->a[1] = Fail;
+
     if (Time.type == TAG_i && (timeout = Time.val) >= 0 &&
 #if defined(ARCH_64)
 	(timeout >> 32) == 0
@@ -3495,7 +3720,7 @@ gen_literal_timeout_locked(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
 	1
 #endif
 	) {
-	op->a[1].val = timeout;
+	op->a[0].val = timeout;
 #if !defined(ARCH_64)
     } else if (Time.type == TAG_q) {
 	Eterm big;
@@ -3509,7 +3734,7 @@ gen_literal_timeout_locked(LoaderState* stp, GenOpArg Fail, GenOpArg Time)
 	} else {
 	    Uint u;
 	    (void) term_to_Uint(big, &u);
-	    op->a[1].val = (BeamInstr) u;
+	    op->a[0].val = (BeamInstr) u;
 	}
 #endif
     } else {
@@ -3555,7 +3780,7 @@ gen_select_tuple_arity(LoaderState* stp, GenOpArg S, GenOpArg Fail,
     if (size == 2) {
 	NEW_GENOP(stp, op);
 	op->next = NULL;
-	op->op = genop_i_select_tuple_arity2_6;
+	op->op = genop_i_select_tuple_arity2_4;
 	GENOP_ARITY(op, arity - 1);
 	op->a[0] = S;
 	op->a[1] = Fail;
@@ -3845,14 +4070,13 @@ gen_select_val(LoaderState* stp, GenOpArg S, GenOpArg Fail,
     int i, j, align = 0;
 
     if (size == 2) {
-
 	/*
 	 * Use a special-cased instruction if there are only two values.
 	 */
 
 	NEW_GENOP(stp, op);
 	op->next = NULL;
-	op->op = genop_i_select_val2_6;
+	op->op = genop_i_select_val2_4;
 	GENOP_ARITY(op, arity - 1);
 	op->a[0] = S;
 	op->a[1] = Fail;
@@ -3862,47 +4086,19 @@ gen_select_val(LoaderState* stp, GenOpArg S, GenOpArg Fail,
 	op->a[5] = Rest[3];
 
 	return op;
-
-    } else if (size > 10) {
-
-	/* binary search instruction */
-
-	NEW_GENOP(stp, op);
-	op->next = NULL;
-	op->op = genop_i_select_val_bins_3;
-	GENOP_ARITY(op, arity);
-	op->a[0] = S;
-	op->a[1] = Fail;
-	op->a[2].type = TAG_u;
-	op->a[2].val = size;
-	for (i = 3; i < arity; i++) {
-	    op->a[i] = Rest[i-3];
-	}
-
-	/*
-	 * Sort the values to make them useful for a binary search.
-	 */
-
-	qsort(op->a+3, size, 2*sizeof(GenOpArg),
-		(int (*)(const void *, const void *)) genopargcompare);
-#ifdef DEBUG
-	for (i = 3; i < arity-2; i += 2) {
-	    ASSERT(op->a[i].val < op->a[i+2].val);
-	}
-#endif
-	return op;
     }
 
-    /* linear search instruction */
-
-    align = 1;
+    if (size <= 10) {
+        /* Use linear search. Reserve place for a sentinel. */
+        align = 1;
+    }
 
     arity += 2*align;
     size  += align;
 
     NEW_GENOP(stp, op);
     op->next = NULL;
-    op->op = genop_i_select_val_lins_3;
+    op->op = (align == 0) ? genop_i_select_val_bins_3 : genop_i_select_val_lins_3;
     GENOP_ARITY(op, arity);
     op->a[0] = S;
     op->a[1] = Fail;
@@ -3916,7 +4112,7 @@ gen_select_val(LoaderState* stp, GenOpArg S, GenOpArg Fail,
     }
 
     /*
-     * Sort the values to make them useful for a sentinel search
+     * Sort the values to make them useful for a binary or sentinel search.
      */
 
     qsort(tmp, size - align, 2*sizeof(GenOpArg),
@@ -3931,11 +4127,12 @@ gen_select_val(LoaderState* stp, GenOpArg S, GenOpArg Fail,
 
     erts_free(ERTS_ALC_T_LOADER_TMP, (void *) tmp);
 
-    /* add sentinel */
-
-    op->a[j].type = TAG_u;
-    op->a[j].val  = ~((BeamInstr)0);
-    op->a[j+size] = Fail;
+    if (align) {
+        /* Add sentinel for linear search. */
+        op->a[j].type = TAG_u;
+        op->a[j].val  = ~((BeamInstr)0);
+        op->a[j+size] = Fail;
+    }
 
 #ifdef DEBUG
     for (i = 0; i < size - 1; i++) {
@@ -4223,6 +4420,92 @@ literal_is_map(LoaderState* stp, GenOpArg Lit)
 }
 
 /*
+ * Predicate to test whether all of the given new small map keys are literals
+ */
+static int
+is_small_map_literal_keys(LoaderState* stp, GenOpArg Size, GenOpArg* Rest)
+{
+    if (Size.val > MAP_SMALL_MAP_LIMIT) {
+        return 0;
+    }
+
+    /*
+     * Operations with non-literals have always only one key.
+     */
+    if (Size.val != 2) {
+        return 1;
+    }
+
+    switch (Rest[0].type) {
+    case TAG_a:
+    case TAG_i:
+    case TAG_n:
+    case TAG_q:
+        return 1;
+    default:
+        return 0;
+    }
+}
+
+static GenOp*
+gen_new_small_map_lit(LoaderState* stp, GenOpArg Dst, GenOpArg Live,
+                      GenOpArg Size, GenOpArg* Rest)
+{
+    unsigned size = Size.val;
+    Uint lit;
+    unsigned i;
+    GenOp* op;
+    GenOpArg* dst;
+    Eterm* hp;
+    Eterm* tmp;
+    Eterm* thp;
+    Eterm keys;
+
+    NEW_GENOP(stp, op);
+    GENOP_ARITY(op, 3 + size/2);
+    op->next = NULL;
+    op->op = genop_i_new_small_map_lit_3;
+
+    tmp = thp = erts_alloc(ERTS_ALC_T_LOADER_TMP, (1 + size/2) * sizeof(*tmp));
+    keys = make_tuple(thp);
+    *thp++ = make_arityval(size/2);
+
+    dst = op->a+3;
+
+    for (i = 0; i < size; i += 2) {
+        switch (Rest[i].type) {
+        case TAG_a:
+            *thp++ = Rest[i].val;
+            ASSERT(is_atom(Rest[i].val));
+            break;
+        case TAG_i:
+            *thp++ = make_small(Rest[i].val);
+            break;
+        case TAG_n:
+            *thp++ = NIL;
+            break;
+        case TAG_q:
+            *thp++ = stp->literals[Rest[i].val].term;
+            break;
+        }
+        *dst++ = Rest[i + 1];
+    }
+
+    if (!find_literal(stp, keys, &lit)) {
+        lit = new_literal(stp, &hp, 1 + size/2);
+        sys_memcpy(hp, tmp, (1 + size/2) * sizeof(*tmp));
+    }
+    erts_free(ERTS_ALC_T_LOADER_TMP, tmp);
+
+    op->a[0] = Dst;
+    op->a[1] = Live;
+    op->a[2].type = TAG_q;
+    op->a[2].val = lit;
+
+    return op;
+}
+
+/*
  * Predicate to test whether the given literal is an empty map.
  */
 
@@ -4239,6 +4522,19 @@ is_empty_map(LoaderState* stp, GenOpArg Lit)
 }
 
 /*
+ * Predicate to test whether the given literal is an export.
+ */
+static int
+literal_is_export(LoaderState* stp, GenOpArg Lit)
+{
+    Eterm term;
+
+    ASSERT(Lit.type == TAG_q);
+    term = stp->literals[Lit.val].term;
+    return is_export(term);
+}
+
+/*
  * Pseudo predicate map_key_sort that will sort the Rest operand for
  * map instructions as a side effect.
  */
@@ -4633,7 +4929,9 @@ freeze_code(LoaderState* stp)
 	line_items[i] = codev + stp->ci - 1;
 
 	line_tab->fname_ptr = (Eterm*) &line_items[i + 1];
-	memcpy(line_tab->fname_ptr, stp->fname, stp->num_fnames*sizeof(Eterm));
+        if (stp->num_fnames)
+            sys_memcpy(line_tab->fname_ptr, stp->fname,
+                       stp->num_fnames*sizeof(Eterm));
 
 	line_tab->loc_size = stp->loc_size;
 	if (stp->loc_size == 2) {
@@ -4733,21 +5031,57 @@ freeze_code(LoaderState* stp)
      */
 
     for (i = 0; i < stp->num_labels; i++) {
-	Sint this_patch;
-	Sint next_patch;
+	Uint patch;
 	Uint value = stp->labels[i].value;
-	
-	if (value == 0 && stp->labels[i].patches >= 0) {
+
+	if (value == 0 && stp->labels[i].num_patches != 0) {
 	    LoadError1(stp, "label %d not resolved", i);
 	}
 	ASSERT(value < stp->ci);
-	this_patch = stp->labels[i].patches;
-	while (this_patch >= 0) {
-	    ASSERT(this_patch < stp->ci);
-	    next_patch = codev[this_patch];
-	    ASSERT(next_patch < stp->ci);
-	    codev[this_patch] = (BeamInstr) (codev + value);
-	    this_patch = next_patch;
+        for (patch = 0; patch < stp->labels[i].num_patches; patch++) {
+            LabelPatch* lp = &stp->labels[i].patches[patch];
+            Uint pos = lp->pos;
+	    ASSERT(pos < stp->ci);
+            if (pos < stp->num_functions) {
+                /*
+                 * This is the array of pointers to the beginning of
+                 * each function. The pointers must remain absolute.
+                 */
+                codev[pos] = (BeamInstr) (codev + value);
+            } else {
+#if defined(DEBUG) && defined(BEAM_WIDE_MASK)
+                Uint w;
+#endif
+                Sint32 rel = lp->offset + value;
+                switch (lp->packed) {
+                case 0:         /* Not packed */
+                    ASSERT(codev[pos] == i);
+                    codev[pos] = rel;
+                    break;
+#ifdef BEAM_WIDE_MASK
+                case 1:         /* Least significant word. */
+#ifdef DEBUG
+                    w = codev[pos] & BEAM_WIDE_MASK;
+                    /* Correct label in least significant word? */
+                    ASSERT(w == i);
+#endif
+                    codev[pos] = (codev[pos] & ~BEAM_WIDE_MASK) |
+                        (rel & BEAM_WIDE_MASK);
+                    break;
+                case 2:         /* Most significant word */
+#ifdef DEBUG
+                    w = (codev[pos] >> BEAM_WIDE_SHIFT) & BEAM_WIDE_MASK;
+                    /* Correct label in most significant word? */
+                    ASSERT(w == i);
+#endif
+                    codev[pos] = ((Uint)rel << BEAM_WIDE_SHIFT) |
+                        (codev[pos] & BEAM_WIDE_MASK);
+                    break;
+#endif
+                default:
+                    ASSERT(0);
+                }
+            }
 	}
     }
     CHKBLK(ERTS_ALC_T_CODE,code_hdr);
@@ -4790,8 +5124,11 @@ final_touch(LoaderState* stp, struct erl_module_instance* inst_p)
     catches = BEAM_CATCHES_NIL;
     while (index != 0) {
 	BeamInstr next = codev[index];
-	codev[index] = BeamOpCode(op_catch_yf);
-	catches = beam_catches_cons((BeamInstr *)codev[index+2], catches);
+        BeamInstr* abs_addr;
+	codev[index] = BeamOpCodeAddr(op_catch_yf);
+        /* We must make the address of the label absolute again. */
+        abs_addr = (BeamInstr *)codev + index + codev[index+2];
+	catches = beam_catches_cons(abs_addr, catches);
 	codev[index+2] = make_catch(catches);
 	index = next;
     }
@@ -4862,7 +5199,7 @@ final_touch(LoaderState* stp, struct erl_module_instance* inst_p)
 		/*
 		 * We are hiding a pointer into older code.
 		 */
-		erts_smp_refc_dec(&fe->refc, 1);
+		erts_refc_dec(&fe->refc, 1);
 	    }
 	    fe->address = code_ptr;
 #ifdef HIPE
@@ -5177,8 +5514,8 @@ transform_engine(LoaderState* st)
 	case TOP_store_rest_args:
 	    {
 		GENOP_ARITY(instr, instr->arity+num_rest_args);
-		memcpy(instr->a, instr->def_args, ap*sizeof(GenOpArg));
-		memcpy(instr->a+ap, rest_args, num_rest_args*sizeof(GenOpArg));
+		sys_memcpy(instr->a, instr->def_args, ap*sizeof(GenOpArg));
+		sys_memcpy(instr->a+ap, rest_args, num_rest_args*sizeof(GenOpArg));
 		ap += num_rest_args;
 	    }
 	    break;
@@ -5258,12 +5595,15 @@ get_tag_and_value(LoaderState* stp, Uint len_code,
 {
     Uint count;
     Sint val;
-    byte default_buf[128];
-    byte* bigbuf = default_buf;
+    byte default_byte_buf[128];
+    byte* byte_buf = default_byte_buf;
+    Eterm default_big_buf[128/sizeof(Eterm)];
+    Eterm* big_buf = default_big_buf;
+    Eterm tmp_big;
     byte* s;
     int i;
     int neg = 0;
-    Uint arity;
+    Uint words_needed;
     Eterm* hp;
 
     /*
@@ -5340,8 +5680,11 @@ get_tag_and_value(LoaderState* stp, Uint len_code,
 	    *result = val;
 	    return TAG_i;
 	} else {
-	    *result = new_literal(stp, &hp, BIG_UINT_HEAP_SIZE);
-	    (void) small_to_big(val, hp);
+            tmp_big = small_to_big(val, big_buf);
+            if (!find_literal(stp, tmp_big, result)) {
+                *result = new_literal(stp, &hp, BIG_UINT_HEAP_SIZE);
+                sys_memcpy(hp, big_buf, BIG_UINT_HEAP_SIZE*sizeof(Eterm));
+            }
 	    return TAG_q;
 	}
     }
@@ -5351,8 +5694,8 @@ get_tag_and_value(LoaderState* stp, Uint len_code,
      * (including margin).
      */
 
-    if (count+8 > sizeof(default_buf)) {
-	bigbuf = erts_alloc(ERTS_ALC_T_LOADER_TMP, count+8);
+    if (count+8 > sizeof(default_byte_buf)) {
+	byte_buf = erts_alloc(ERTS_ALC_T_LOADER_TMP, count+8);
     }
 
     /*
@@ -5361,20 +5704,20 @@ get_tag_and_value(LoaderState* stp, Uint len_code,
 
     GetString(stp, s, count);
     for (i = 0; i < count; i++) {
-	bigbuf[count-i-1] = *s++;
+	byte_buf[count-i-1] = *s++;
     }
 
     /*
      * Check if the number is negative, and negate it if so.
      */
 
-    if ((bigbuf[count-1] & 0x80) != 0) {
+    if ((byte_buf[count-1] & 0x80) != 0) {
 	unsigned carry = 1;
 
 	neg = 1;
 	for (i = 0; i < count; i++) {
-	    bigbuf[i] = ~bigbuf[i] + carry;
-	    carry = (bigbuf[i] == 0 && carry == 1);
+	    byte_buf[i] = ~byte_buf[i] + carry;
+	    carry = (byte_buf[i] == 0 && carry == 1);
 	}
 	ASSERT(carry == 0);
     }
@@ -5383,33 +5726,52 @@ get_tag_and_value(LoaderState* stp, Uint len_code,
      * Align to word boundary.
      */
 
-    if (bigbuf[count-1] == 0) {
+    if (byte_buf[count-1] == 0) {
 	count--;
     }
-    if (bigbuf[count-1] == 0) {
+    if (byte_buf[count-1] == 0) {
 	LoadError0(stp, "bignum not normalized");
     }
     while (count % sizeof(Eterm) != 0) {
-	bigbuf[count++] = 0;
+	byte_buf[count++] = 0;
     }
 
     /*
-     * Allocate heap space for the bignum and copy it.
+     * Convert to a bignum.
      */
 
-    arity = count/sizeof(Eterm);
-    *result = new_literal(stp, &hp, arity+1);
-    if (is_nil(bytes_to_big(bigbuf, count, neg, hp)))
-	goto load_error;
+    words_needed = count/sizeof(Eterm) + 1;
+    if (words_needed*sizeof(Eterm) > sizeof(default_big_buf)) {
+        big_buf = erts_alloc(ERTS_ALC_T_LOADER_TMP, words_needed*sizeof(Eterm));
+    }
+    tmp_big = bytes_to_big(byte_buf, count, neg, big_buf);
+    if (is_nil(tmp_big)) {
+        goto load_error;
+    }
+
+    /*
+     * Create a literal if there is no previous literal with the same value.
+     */
 
-    if (bigbuf != default_buf) {
-	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) bigbuf);
+    if (!find_literal(stp, tmp_big, result)) {
+        *result = new_literal(stp, &hp, words_needed);
+        sys_memcpy(hp, big_buf, words_needed*sizeof(Eterm));
+    }
+
+    if (byte_buf != default_byte_buf) {
+	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) byte_buf);
+    }
+    if (big_buf != default_big_buf) {
+	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) big_buf);
     }
     return TAG_q;
 
  load_error:
-    if (bigbuf != default_buf) {
-	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) bigbuf);
+    if (byte_buf != default_byte_buf) {
+	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) byte_buf);
+    }
+    if (big_buf != default_big_buf) {
+	erts_free(ERTS_ALC_T_LOADER_TMP, (void *) big_buf);
     }
     return -1;
 }
@@ -5454,8 +5816,7 @@ new_label(LoaderState* stp)
     stp->labels = (Label *) erts_realloc(ERTS_ALC_T_PREPARED_CODE,
 					 (void *) stp->labels,
 					 stp->num_labels * sizeof(Label));
-    stp->labels[num].value = 0;
-    stp->labels[num].patches = -1;
+    init_label(&stp->labels[num]);
     return num;
 }
 
@@ -5510,6 +5871,24 @@ new_literal(LoaderState* stp, Eterm** hpp, Uint heap_size)
     return stp->num_literals++;
 }
 
+static int
+find_literal(LoaderState* stp, Eterm needle, Uint *idx)
+{
+    int i;
+
+    /*
+     * The search is done backwards since the most recent literals
+     * allocated by the loader itself will be placed at the end
+     */
+    for (i = stp->num_literals - 1; i >= 0; i--) {
+        if (EQ(needle, stp->literals[i].term)) {
+            *idx = (Uint) i;
+            return 1;
+        }
+    }
+    return 0;
+}
+
 Eterm
 erts_module_info_0(Process* p, Eterm module)
 {
@@ -5589,6 +5968,8 @@ get_module_info(Process* p, ErtsCodeIndex code_ix, BeamCodeHeader* code_hdr,
 	return exported_from_module(p, code_ix, module);
     } else if (what == am_functions) {
 	return functions_in_module(p, code_hdr);
+    } else if (what == am_nifs) {
+	return nifs_in_module(p, module);
     } else if (what == am_attributes) {
 	return attributes_for_module(p, code_hdr);
     } else if (what == am_compile) {
@@ -5642,6 +6023,46 @@ functions_in_module(Process* p, /* Process whose heap to use. */
 }
 
 /*
+ * Builds a list of all NIFs in the given module:
+ *     [{Name, Arity},...]
+ */
+Eterm
+nifs_in_module(Process* p, Eterm module)
+{
+    Eterm nif_list, *hp;
+    Module *mod;
+
+    mod = erts_get_module(module, erts_active_code_ix());
+    nif_list = NIL;
+
+    if (mod->curr.nif != NULL) {
+        int func_count, func_ix;
+        ErlNifFunc *funcs;
+
+        func_count = erts_nif_get_funcs(mod->curr.nif, &funcs);
+        hp = HAlloc(p, func_count * 5);
+
+        for (func_ix = func_count - 1; func_ix >= 0; func_ix--) {
+            Eterm name, arity, pair;
+            ErlNifFunc *func;
+
+            func = &funcs[func_ix];
+
+            name = am_atom_put(func->name, sys_strlen(func->name));
+            arity = make_small(func->arity);
+
+            pair = TUPLE2(hp, name, arity);
+            hp += 3;
+
+            nif_list = CONS(hp, pair, nif_list);
+            hp += 2;
+        }
+    }
+
+    return nif_list;
+}
+
+/*
  * Returns 'true' if mod has any native compiled functions, otherwise 'false'
  */
 
@@ -5678,7 +6099,7 @@ erts_release_literal_area(ErtsLiteralArea* literal_area)
         case FUN_SUBTAG:
             {
                 ErlFunEntry* fe = ((ErlFunThing*)oh)->fe;
-                if (erts_smp_refc_dectest(&fe->refc, 0) == 0) {
+                if (erts_refc_dectest(&fe->refc, 0) == 0) {
                     erts_erase_fun_entry(fe);
                 }
                 break;
@@ -5723,9 +6144,9 @@ int
 erts_is_function_native(ErtsCodeInfo *ci)
 {
 #ifdef HIPE
-    ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI));
-    return erts_codeinfo_to_code(ci)[0] == (BeamInstr) BeamOp(op_hipe_trap_call)
-	|| erts_codeinfo_to_code(ci)[0] == (BeamInstr) BeamOp(op_hipe_trap_call_closure);
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
+    return BeamIsOpCode(erts_codeinfo_to_code(ci)[0], op_hipe_trap_call) ||
+	BeamIsOpCode(erts_codeinfo_to_code(ci)[0], op_hipe_trap_call_closure);
 #else
     return 0;
 #endif
@@ -5794,7 +6215,7 @@ exported_from_module(Process* p, /* Process whose heap to use. */
 	    Eterm tuple;
 	    
 	    if (ep->addressv[code_ix] == ep->beam &&
-		ep->beam[0] == (BeamInstr) em_call_error_handler) {
+		BeamIsOpCode(ep->beam[0], op_call_error_handler)) {
 		/* There is a call to the function, but it does not exist. */ 
 		continue;
 	    }
@@ -6064,7 +6485,7 @@ make_stub(ErtsCodeInfo* info, Eterm mod, Eterm func, Uint arity, Uint native, Be
 {
     DBG_TRACE_MFA(mod,func,arity,"make beam stub at %p", erts_codeinfo_to_code(info));
     ASSERT(WORDS_PER_FUNCTION == 6);
-    info->op = (BeamInstr) BeamOp(op_i_func_info_IaaI);
+    info->op = BeamOpCodeAddr(op_i_func_info_IaaI);
     info->u.ncallee = (void (*)(void)) native;
     info->mfa.module = mod;
     info->mfa.function = func;
@@ -6084,7 +6505,7 @@ stub_copy_info(LoaderState* stp,
     Sint decoded_size;
     Uint size = stp->chunks[chunk].size;
     if (size != 0) {
-	memcpy(info, stp->chunks[chunk].start, size);
+	sys_memcpy(info, stp->chunks[chunk].start, size);
 	*ptr_word = info;
 	decoded_size = erts_decode_ext_size(info, size);
 	if (decoded_size < 0) {
@@ -6169,7 +6590,7 @@ stub_final_touch(LoaderState* stp, ErtsCodeInfo* ci)
     for (i = 0, lp = stp->lambdas; i < n; i++, lp++) {
         ErlFunEntry* fe = stp->lambdas[i].fe;
 	if (lp->function == ci->mfa.function && lp->arity == ci->mfa.arity) {
-	    *erts_codeinfo_to_code(ci) = (Eterm) BeamOpCode(op_hipe_trap_call_closure);
+	    *erts_codeinfo_to_code(ci) = BeamOpCodeAddr(op_hipe_trap_call_closure);
             fe->address = erts_codeinfo_to_code(ci);
 	}
     }
@@ -6299,7 +6720,7 @@ patch_funentries(Eterm Patchlist)
     fe = erts_get_fun_entry(Mod, uniq, index);
     fe->native_address = (Uint *)native_address;
 
-    erts_smp_refc_dec(&fe->refc, 1);
+    erts_refc_dec(&fe->refc, 1);
 
     if (!patch(Addresses, (Uint) fe))
       return 0;
@@ -6493,7 +6914,7 @@ erts_make_stub_module(Process* p, Eterm hipe_magic_bin, Eterm Beam, Eterm Info)
 	 * as the body until we know what kind of trap we should put there.
 	 */
 	code_hdr->functions[i] = (ErtsCodeInfo*)fp;
-	op = (Eterm) BeamOpCode(op_hipe_trap_call); /* Might be changed later. */
+	op = BeamOpCodeAddr(op_hipe_trap_call); /* Might be changed later. */
 	fp = make_stub((ErtsCodeInfo*)fp, hipe_stp->module, func, arity,
                        (Uint)native_address, op);
     }
@@ -6503,7 +6924,7 @@ erts_make_stub_module(Process* p, Eterm hipe_magic_bin, Eterm Beam, Eterm Info)
      */
 
     code_hdr->functions[i] = (ErtsCodeInfo*)fp;
-    *fp++ = (BeamInstr) BeamOp(op_int_code_end);
+    *fp++ = BeamOpCodeAddr(op_int_code_end);
 
     /*
      * Copy attributes and compilation information.
@@ -6654,8 +7075,8 @@ void dbg_set_traced_mfa(const char* m, const char* f, Uint a)
 {
     unsigned i = dbg_trace_ix++;
     ASSERT(i < MFA_MAX);
-    dbg_trace_m[i] = am_atom_put(m, strlen(m));
-    dbg_trace_f[i] = am_atom_put(f, strlen(f));
+    dbg_trace_m[i] = am_atom_put(m, sys_strlen(m));
+    dbg_trace_f[i] = am_atom_put(f, sys_strlen(f));
     dbg_trace_a[i] = a;
 }
 
diff --git a/erts/emulator/beam/beam_load.h b/erts/emulator/beam/beam_load.h
index c088bdb751..156c3c45e2 100644
--- a/erts/emulator/beam/beam_load.h
+++ b/erts/emulator/beam/beam_load.h
@@ -37,11 +37,6 @@ typedef struct gen_op_entry {
 
 extern const GenOpEntry gen_opc[];
 
-extern BeamInstr beam_debug_apply[];
-extern BeamInstr* em_call_error_handler;
-extern BeamInstr* em_apply_bif;
-extern BeamInstr* em_call_nif;
-
 struct ErtsLiteralArea_;
 
 /*
@@ -111,11 +106,7 @@ typedef struct beam_code_header {
 
 }BeamCodeHeader;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 #  define BEAM_NIF_MIN_FUNC_SZ 4
-#else
-#  define BEAM_NIF_MIN_FUNC_SZ 3
-#endif
 
 void erts_release_literal_area(struct ErtsLiteralArea_* literal_area);
 int erts_is_module_native(BeamCodeHeader* code);
diff --git a/erts/emulator/beam/beam_ranges.c b/erts/emulator/beam/beam_ranges.c
index c7b17d58f3..f0c9496341 100644
--- a/erts/emulator/beam/beam_ranges.c
+++ b/erts/emulator/beam/beam_ranges.c
@@ -30,7 +30,7 @@
 
 typedef struct {
     BeamInstr* start;		/* Pointer to start of module. */
-    erts_smp_atomic_t end; /* (BeamInstr*) Points one word beyond last function in module. */
+    erts_atomic_t end; /* (BeamInstr*) Points one word beyond last function in module. */
 } Range;
 
 /*
@@ -44,7 +44,7 @@ Uint erts_dump_num_lit_areas;
 
 /* Range 'end' needs to be atomic as we purge module
     by setting end=start in active code_ix */
-#define RANGE_END(R) ((BeamInstr*)erts_smp_atomic_read_nob(&(R)->end))
+#define RANGE_END(R) ((BeamInstr*)erts_atomic_read_nob(&(R)->end))
 
 static Range* find_range(BeamInstr* pc);
 static void lookup_loc(FunctionInfo* fi, const BeamInstr* pc,
@@ -59,10 +59,10 @@ struct ranges {
     Range* modules;	       /* Sorted lists of module addresses. */
     Sint n;		       /* Number of range entries. */
     Sint allocated;	       /* Number of allocated entries. */
-    erts_smp_atomic_t mid;     /* Cached search start point */
+    erts_atomic_t mid;     /* Cached search start point */
 };
 static struct ranges r[ERTS_NUM_CODE_IX];
-static erts_smp_atomic_t mem_used;
+static erts_atomic_t mem_used;
 static Range* write_ptr;
 
 #ifdef HARD_DEBUG
@@ -100,12 +100,12 @@ erts_init_ranges(void)
 {
     Sint i;
 
-    erts_smp_atomic_init_nob(&mem_used, 0);
+    erts_atomic_init_nob(&mem_used, 0);
     for (i = 0; i < ERTS_NUM_CODE_IX; i++) {
 	r[i].modules = 0;
 	r[i].n = 0;
 	r[i].allocated = 0;
-	erts_smp_atomic_init_nob(&r[i].mid, 0);
+	erts_atomic_init_nob(&r[i].mid, 0);
     }
 
     erts_dump_num_lit_areas = 8;
@@ -122,12 +122,12 @@ erts_start_staging_ranges(int num_new)
     Sint need;
 
     if (r[dst].modules) {
-	erts_smp_atomic_add_nob(&mem_used, -r[dst].allocated);
+	erts_atomic_add_nob(&mem_used, -r[dst].allocated);
 	erts_free(ERTS_ALC_T_MODULE_REFS, r[dst].modules);
     }
 
     need = r[dst].allocated = r[src].n + num_new;
-    erts_smp_atomic_add_nob(&mem_used, need);
+    erts_atomic_add_nob(&mem_used, need);
     write_ptr = erts_alloc(ERTS_ALC_T_MODULE_REFS,
 			   need * sizeof(Range));
     r[dst].modules = write_ptr;
@@ -150,7 +150,7 @@ erts_end_staging_ranges(int commit)
 	    if (rp->start < RANGE_END(rp)) {
 		/* Only insert a module that has not been purged. */
 		write_ptr->start = rp->start;
-		erts_smp_atomic_init_nob(&write_ptr->end,
+		erts_atomic_init_nob(&write_ptr->end,
 					 (erts_aint_t)(RANGE_END(rp)));
 		write_ptr++;
 	    }
@@ -176,7 +176,7 @@ erts_end_staging_ranges(int commit)
 	}
 	r[dst].modules = mp;
 	CHECK(&r[dst]);
-	erts_smp_atomic_set_nob(&r[dst].mid,
+	erts_atomic_set_nob(&r[dst].mid,
 				(erts_aint_t) (r[dst].modules +
 					       r[dst].n / 2));
 
@@ -205,7 +205,7 @@ erts_update_ranges(BeamInstr* code, Uint size)
 	 */
 	if (r[dst].modules == NULL) {
 	    Sint need = 128;
-	    erts_smp_atomic_add_nob(&mem_used, need);
+	    erts_atomic_add_nob(&mem_used, need);
 	    r[dst].modules = erts_alloc(ERTS_ALC_T_MODULE_REFS,
 					need * sizeof(Range));
 	    r[dst].allocated = need;
@@ -215,7 +215,7 @@ erts_update_ranges(BeamInstr* code, Uint size)
 
     ASSERT(r[dst].modules);
     write_ptr->start = code;
-    erts_smp_atomic_init_nob(&(write_ptr->end),
+    erts_atomic_init_nob(&(write_ptr->end),
 			     (erts_aint_t)(((byte *)code) + size));
     write_ptr++;
 }
@@ -224,13 +224,13 @@ void
 erts_remove_from_ranges(BeamInstr* code)
 {
     Range* rp = find_range(code);
-    erts_smp_atomic_set_nob(&rp->end, (erts_aint_t)rp->start);
+    erts_atomic_set_nob(&rp->end, (erts_aint_t)rp->start);
 }
 
 UWord
 erts_ranges_sz(void)
 {
-    return erts_smp_atomic_read_nob(&mem_used) * sizeof(Range);
+    return erts_atomic_read_nob(&mem_used) * sizeof(Range);
 }
 
 /*
@@ -285,7 +285,7 @@ find_range(BeamInstr* pc)
     ErtsCodeIndex active = erts_active_code_ix();
     Range* low = r[active].modules;
     Range* high = low + r[active].n;
-    Range* mid = (Range *) erts_smp_atomic_read_nob(&r[active].mid);
+    Range* mid = (Range *) erts_atomic_read_nob(&r[active].mid);
 
     CHECK(&r[active]);
     while (low < high) {
@@ -294,7 +294,7 @@ find_range(BeamInstr* pc)
 	} else if (pc >= RANGE_END(mid)) {
 	    low = mid + 1;
 	} else {
-	    erts_smp_atomic_set_nob(&r[active].mid, (erts_aint_t) mid);
+	    erts_atomic_set_nob(&r[active].mid, (erts_aint_t) mid);
 	    return mid;
 	}
 	mid = low + (high-low) / 2;
diff --git a/erts/emulator/beam/bif.c b/erts/emulator/beam/bif.c
index d7a25adccb..97e1ee1286 100644
--- a/erts/emulator/beam/bif.c
+++ b/erts/emulator/beam/bif.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -46,21 +46,23 @@
 #include "erl_bif_unique.h"
 #include "erl_map.h"
 #include "erl_msacc.h"
+#include "erl_proc_sig_queue.h"
 
 Export *erts_await_result;
+static Export await_exit_trap;
 static Export* flush_monitor_messages_trap = NULL;
 static Export* set_cpu_topology_trap = NULL;
-static Export* await_proc_exit_trap = NULL;
 static Export* await_port_send_result_trap = NULL;
 Export* erts_format_cpu_topology_trap = NULL;
 static Export dsend_continue_trap_export;
 Export *erts_convert_time_unit_trap = NULL;
 
 static Export *await_msacc_mod_trap = NULL;
-static erts_smp_atomic32_t msacc;
+static erts_atomic32_t msacc;
 
+static Export *system_flag_scheduler_wall_time_trap;
 static Export *await_sched_wall_time_mod_trap;
-static erts_smp_atomic32_t sched_wall_time;
+static erts_atomic32_t sched_wall_time;
 
 #define DECL_AM(S) Eterm AM_ ## S = am_atom_put(#S, sizeof(#S) - 1)
 
@@ -91,85 +93,51 @@ BIF_RETTYPE spawn_3(BIF_ALIST_3)
 /* Utility to add a new link between processes p and another internal
  * process (rpid). Process p must be the currently executing process.
  */
-static int insert_internal_link(Process* p, Eterm rpid)
-{
-    Process *rp;
-    ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK;
-
-    ASSERT(is_internal_pid(rpid));
-
-#ifdef ERTS_SMP
-    if (IS_TRACED(p)
-	&& (ERTS_TRACE_FLAGS(p) & (F_TRACE_SOL|F_TRACE_SOL1))) {
-	rp_locks = ERTS_PROC_LOCKS_ALL;
-    }
-
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_LINK);
-#endif
-
-    /* get a pointer to the process struct of the linked process */
-    rp = erts_pid2proc_opt(p, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK,
-			   rpid, rp_locks,
-			   ERTS_P2P_FLG_ALLOW_OTHER_X);
-
-    if (!rp) {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	return 0;
-    }
-
-    if (p != rp) {
-	erts_add_link(&ERTS_P_LINKS(p), LINK_PID, rp->common.id);
-	erts_add_link(&ERTS_P_LINKS(rp), LINK_PID, p->common.id);
-
-	ASSERT(IS_TRACER_VALID(ERTS_TRACER(p)));
-
-	if (IS_TRACED(p)) {
-	    if (ERTS_TRACE_FLAGS(p) & (F_TRACE_SOL|F_TRACE_SOL1))  {
-		ERTS_TRACE_FLAGS(rp) |= (ERTS_TRACE_FLAGS(p) & TRACEE_FLAGS);
-                erts_tracer_replace(&rp->common, ERTS_TRACER(p));
-		if (ERTS_TRACE_FLAGS(p) & F_TRACE_SOL1) { /* maybe override */
-		    ERTS_TRACE_FLAGS(rp) &= ~(F_TRACE_SOL1 | F_TRACE_SOL);
-		    ERTS_TRACE_FLAGS(p) &= ~(F_TRACE_SOL1 | F_TRACE_SOL);
-		}
-	    }
-	}
-    }
-    if (IS_TRACED_FL(rp, F_TRACE_PROCS))
-	trace_proc(p, p == rp ? rp_locks : ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK,
-                   rp, am_getting_linked, p->common.id);
-
-    if (p == rp)
-	erts_smp_proc_unlock(p, rp_locks & ~ERTS_PROC_LOCK_MAIN);
-    else {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	erts_smp_proc_unlock(rp, rp_locks);
-    }
-
-    return 1;
-}
-
 
 /* create a link to the process */
 BIF_RETTYPE link_1(BIF_ALIST_1)
 {
-    DistEntry *dep;
-
     if (IS_TRACED_FL(BIF_P, F_TRACE_PROCS)) {
 	trace_proc(BIF_P, ERTS_PROC_LOCK_MAIN, BIF_P, am_link, BIF_ARG_1);
     }
     /* check that the pid or port which is our argument is OK */
 
     if (is_internal_pid(BIF_ARG_1)) {
-	if (insert_internal_link(BIF_P, BIF_ARG_1)) {
-	    BIF_RET(am_true);
-	}
-	else {
-	    goto res_no_proc;
-	}
+        int created;
+        ErtsLinkData *ldp;
+        ErtsLink *lnk;
+
+        if (BIF_P->common.id == BIF_ARG_1)
+            BIF_RET(am_true);
+
+        if (!erts_proc_lookup(BIF_ARG_1))
+            goto res_no_proc;
+
+        lnk = erts_link_tree_lookup_create(&ERTS_P_LINKS(BIF_P),
+                                           &created,
+                                           ERTS_LNK_TYPE_PROC,
+                                           BIF_P->common.id,
+                                           BIF_ARG_1);
+        if (!created)
+            BIF_RET(am_true);
+
+        ldp = erts_link_to_data(lnk);
+        
+
+        if (erts_proc_sig_send_link(BIF_P, BIF_ARG_1, &ldp->b))
+            BIF_RET(am_true);
+
+        erts_link_tree_delete(&ERTS_P_LINKS(BIF_P), lnk);
+        erts_link_release_both(ldp);
+        goto res_no_proc;
     }
 
     if (is_internal_port(BIF_ARG_1)) {
-	int send_link_signal = 0;
+        int created;
+        ErtsLinkData *ldp;
+        ErtsLink *lnk;
+        Eterm ref;
+        Eterm *refp;
 	Port *prt = erts_port_lookup(BIF_ARG_1,
 				     (erts_port_synchronous_ops
 				      ? ERTS_PORT_SFLGS_INVALID_DRIVER_LOOKUP
@@ -178,31 +146,31 @@ BIF_RETTYPE link_1(BIF_ALIST_1)
 	    goto res_no_proc;
 	}
 
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK);
-	
-	if (erts_add_link(&ERTS_P_LINKS(BIF_P), LINK_PID, BIF_ARG_1) >= 0)
-	    send_link_signal = 1;
-	/* else: already linked */
-
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
+        lnk = erts_link_tree_lookup_create(&ERTS_P_LINKS(BIF_P),
+                                           &created,
+                                           ERTS_LNK_TYPE_PORT,
+                                           BIF_P->common.id,
+                                           BIF_ARG_1);
+        if (!created)
+            BIF_RET(am_true);
 
-	if (send_link_signal) {
-	    Eterm ref;
-	    Eterm *refp = erts_port_synchronous_ops ? &ref : NULL;
+        ldp = erts_link_to_data(lnk);
+        refp = erts_port_synchronous_ops ? &ref : NULL;
 
-	    switch (erts_port_link(BIF_P, prt, BIF_P->common.id, refp)) {
-	    case ERTS_PORT_OP_DROPPED:
-	    case ERTS_PORT_OP_BADARG:
-		goto res_no_proc;
-	    case ERTS_PORT_OP_SCHEDULED:
-		if (refp) {
-		    ASSERT(is_internal_ordinary_ref(ref));
-		    BIF_TRAP3(await_port_send_result_trap, BIF_P, ref, am_true, am_true);
-		}
-	    default:
-		break;
-	    }
-	}
+        switch (erts_port_link(BIF_P, prt, &ldp->b, refp)) {
+        case ERTS_PORT_OP_DROPPED:
+        case ERTS_PORT_OP_BADARG:
+            erts_link_tree_delete(&ERTS_P_LINKS(BIF_P), lnk);
+            erts_link_release_both(ldp);
+            goto res_no_proc;
+        case ERTS_PORT_OP_SCHEDULED:
+            if (refp) {
+                ASSERT(is_internal_ordinary_ref(ref));
+                BIF_TRAP3(await_port_send_result_trap, BIF_P, ref, am_true, am_true);
+            }
+        default:
+            break;
+        }
 	BIF_RET(am_true);
     }
     else if (is_external_port(BIF_ARG_1)
@@ -211,342 +179,203 @@ BIF_RETTYPE link_1(BIF_ALIST_1)
     }
 
     if (is_external_pid(BIF_ARG_1)) {
+        ErtsLinkData *ldp;
+        int created;
+        DistEntry *dep;
+        ErtsLink *lnk;
+        int code;
+        ErtsDSigData dsd;
+
+        dep = external_pid_dist_entry(BIF_ARG_1);
+        if (dep == erts_this_dist_entry)
+            goto res_no_proc;
 
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK);
-
-	/* We may earn time by checking first that we're not linked already */
-	if (erts_lookup_link(ERTS_P_LINKS(BIF_P), BIF_ARG_1) != NULL) {
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-	    BIF_RET(am_true);
-	}
-	else {
-	    ErtsLink *lnk;
-	    int code;
-	    ErtsDSigData dsd;
-	    dep = external_pid_dist_entry(BIF_ARG_1);
-	    if (dep == erts_this_dist_entry) {
-		erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-		goto res_no_proc;
-	    }
-
-	    code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_DSP_RLOCK, 0);
-	    switch (code) {
-	    case ERTS_DSIG_PREP_NOT_ALIVE:
-		/* Let the dlink trap handle it */
-	    case ERTS_DSIG_PREP_NOT_CONNECTED:
-		erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-		BIF_TRAP1(dlink_trap, BIF_P, BIF_ARG_1);
-
-	    case ERTS_DSIG_PREP_CONNECTED:
-		/* We are connected. Setup link and send link signal */
-
-		erts_smp_de_links_lock(dep);
-
-		erts_add_link(&ERTS_P_LINKS(BIF_P), LINK_PID, BIF_ARG_1);
-		lnk = erts_add_or_lookup_link(&(dep->nlinks),
-					      LINK_PID,
-					      BIF_P->common.id);
-		ASSERT(lnk != NULL);
-		erts_add_link(&ERTS_LINK_ROOT(lnk), LINK_PID, BIF_ARG_1);
-
-		erts_smp_de_links_unlock(dep);
-		erts_smp_de_runlock(dep);
-		erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-
-		code = erts_dsig_send_link(&dsd, BIF_P->common.id, BIF_ARG_1);
-		if (code == ERTS_DSIG_SEND_YIELD)
-		    ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
-		BIF_RET(am_true);
-	    default:
-		ASSERT(! "Invalid dsig prepare result");
-		BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
-	    }
-	}
+        lnk = erts_link_tree_lookup_create(&ERTS_P_LINKS(BIF_P),
+                                           &created,
+                                           ERTS_LNK_TYPE_DIST_PROC,
+                                           BIF_P->common.id,
+                                           BIF_ARG_1);
+
+        if (!created)
+            BIF_RET(am_true); /* Already present... */
+
+        ldp = erts_link_to_data(lnk);
+
+        code = erts_dsig_prepare(&dsd, dep, BIF_P,
+                                 ERTS_PROC_LOCK_MAIN,
+                                 ERTS_DSP_RLOCK, 0, 1);
+        switch (code) {
+        case ERTS_DSIG_PREP_NOT_ALIVE:
+        case ERTS_DSIG_PREP_NOT_CONNECTED:
+            erts_link_set_dead_dist(&ldp->b, dep->sysname);
+            erts_proc_sig_send_link_exit(NULL, BIF_ARG_1, &ldp->b,
+                                         am_noconnection, NIL);
+            BIF_RET(am_true);
+
+        case ERTS_DSIG_PREP_PENDING:
+        case ERTS_DSIG_PREP_CONNECTED: {
+            /*
+             * We have (pending) connection.
+             * Setup link and enqueue link signal.
+             */
+#ifdef DEBUG
+            int inserted =
+#endif
+                erts_link_dist_insert(&ldp->b, dep->mld);
+            ASSERT(inserted);
+            erts_de_runlock(dep);
+
+            code = erts_dsig_send_link(&dsd, BIF_P->common.id, BIF_ARG_1);
+            if (code == ERTS_DSIG_SEND_YIELD)
+                ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
+            BIF_RET(am_true);
+            break;
+        }
+        default:
+            ERTS_ASSERT(! "Invalid dsig prepare result");
+        }
     }
 
     BIF_ERROR(BIF_P, BADARG);
 
-res_no_proc: {
-	erts_aint32_t state = erts_smp_atomic32_read_nob(&BIF_P->state);
-	if (state & ERTS_PSFLG_TRAP_EXIT) {
-	    ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN;
-	    erts_deliver_exit_message(BIF_ARG_1, BIF_P, &locks, am_noproc, NIL);
-	    erts_smp_proc_unlock(BIF_P, ~ERTS_PROC_LOCK_MAIN & locks);
-	    BIF_RET(am_true);
-	}
-	else
-	    BIF_ERROR(BIF_P, EXC_NOPROC);
+res_no_proc:
+    if (BIF_P->flags & F_TRAP_EXIT) {
+        ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN;
+        erts_deliver_exit_message(BIF_ARG_1, BIF_P, &locks, am_noproc, NIL);
+        erts_proc_unlock(BIF_P, ~ERTS_PROC_LOCK_MAIN & locks);
+        BIF_RET(am_true);
+    }
+    else {
+        /*
+         * This behaviour is *really* sad but link/1 has
+         * behaved like this for ages (and this behaviour is
+         * actually documented)... :'-(
+         *
+         * The proper behavior would have been to
+         * send calling process an exit signal..
+         */
+        BIF_ERROR(BIF_P, EXC_NOPROC);
     }
 }
 
-/* This function is allowed to return range of values handled by demonitor/1-2
- * Namely: atoms true, false, yield, internal_error, badarg or THE_NON_VALUE
- */
 static Eterm
-remote_demonitor(Process *c_p, DistEntry *dep, Eterm ref, Eterm to)
+demonitor(Process *c_p, Eterm ref, Eterm *multip)
 {
-    ErtsDSigData dsd;
-    ErtsMonitor *dmon;
-    ErtsMonitor *mon;
-    int code;
-    Eterm res = am_false;
-#ifndef ERTS_SMP
-    int stale_mon = 0;
-#endif
+    ErtsMonitor  *mon;  /* The monitor entry to delete */
 
-    ERTS_SMP_LC_ASSERT((ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK)
-		       == erts_proc_lc_my_proc_locks(c_p));
+   *multip = am_false;
 
-    code = erts_dsig_prepare(&dsd, dep, c_p, ERTS_DSP_RLOCK, 0);
-    switch (code) {
-    case ERTS_DSIG_PREP_NOT_ALIVE:
-    case ERTS_DSIG_PREP_NOT_CONNECTED:
-#ifndef ERTS_SMP
-	/* XXX Is this possible? Shouldn't this link
-	   previously have been removed if the node
-	   had previously been disconnected. */
-	ASSERT(0);
-	stale_mon = 1;
-#endif
-	/*
-	 * In the smp case this is possible if the node goes
-	 * down just before the call to demonitor.
-	 */
-	if (dep) {
-	    erts_smp_de_links_lock(dep);
-	    dmon = erts_remove_monitor(&dep->monitors, ref);
-	    erts_smp_de_links_unlock(dep);
-	    if (dmon)
-		erts_destroy_monitor(dmon);
-	}
-	mon = erts_remove_monitor(&ERTS_P_MONITORS(c_p), ref);
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_LINK);
-
-        res = am_true;
-	break;
-
-    case ERTS_DSIG_PREP_CONNECTED:
-
-	erts_smp_de_links_lock(dep);
-	mon = erts_remove_monitor(&ERTS_P_MONITORS(c_p), ref);
-	dmon = erts_remove_monitor(&dep->monitors, ref);
-	erts_smp_de_links_unlock(dep);
-	erts_smp_de_runlock(dep);
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_LINK);
-
-	if (!dmon) {
-#ifndef ERTS_SMP
-	    /* XXX How is this possible? Shouldn't this link
-	       previously have been removed when the distributed
-	       end was removed. */
-	    ASSERT(0);
-	    stale_mon = 1;
-#endif
-	    /*
-	     * This is possible when smp support is enabled.
-	     * 'DOWN' message just arrived.
-	     */
-            res = am_true;
-	}
-	else {
-	    /*
-	     * Soft (no force) send, use ->data in dist slot 
-	     * monitor list since in case of monitor name 
-	     * the atom is stored there. Yield if necessary.
-	     */
-	    code = erts_dsig_send_demonitor(&dsd,
-					    c_p->common.id, 
-					    (mon->name != NIL
-					     ? mon->name
-					     : mon->u.pid), 
-					    ref,
-					    0);
-            res = (code == ERTS_DSIG_SEND_YIELD ? am_yield : am_true);
-	    erts_destroy_monitor(dmon);
-	}
-	break;
-    default:
-	ASSERT(! "Invalid dsig prepare result");
-        return am_internal_error;
-    }
-
-#ifndef ERTS_SMP
-    if (stale_mon) {
-	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-	erts_dsprintf(dsbufp, "Stale process monitor %T to ", ref);
-	if (is_atom(to))
-	    erts_dsprintf(dsbufp, "{%T, %T}", to, dep->sysname);
-	else
-	    erts_dsprintf(dsbufp, "%T", to);
-	erts_dsprintf(dsbufp, " found\n");
-	erts_send_error_to_logger(c_p->group_leader, dsbufp);
-    }
-#endif
-
-    /*
-     * We aren't allowed to destroy 'mon' until now, since 'to'
-     * may refer into 'mon' (external pid).
-     */
-    ASSERT(mon); /* Since link lock wasn't released between
-		    lookup and remove */
-    erts_destroy_monitor(mon);
-
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
-    return res;
-}
+   if (is_not_internal_ref(ref)) {
+       if (is_external_ref(ref)
+           && (erts_this_dist_entry
+               == external_ref_dist_entry(ref))) {
+           return am_false;
+       }
+       return am_badarg; /* Not monitored by this monitor's ref */
+   }
 
-static ERTS_INLINE void
-demonitor_local_process(Process *c_p, Eterm ref, Eterm to, Eterm *res)
-{
-    Process *rp = erts_pid2proc_opt(c_p,
-                           ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK,
-                           to,
-                           ERTS_PROC_LOCK_LINK,
-                           ERTS_P2P_FLG_ALLOW_OTHER_X);
-    ErtsMonitor *mon = erts_remove_monitor(&ERTS_P_MONITORS(c_p), ref);
-
-#ifndef ERTS_SMP
-    ASSERT(mon);
-#else
-    if (!mon)
-        *res = am_false;
-    else
-#endif
-    {
-        *res = am_true;
-        erts_destroy_monitor(mon);
-    }
-    if (rp) {
-        ErtsMonitor *rmon;
-        rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), ref);
-        if (rp != c_p)
-            erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-        if (rmon != NULL)
-            erts_destroy_monitor(rmon);
-    }
-    else {
-        ERTS_SMP_ASSERT_IS_NOT_EXITING(c_p);
-    }
-}
+   mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p), ref);
+   if (!mon)
+       return am_false;
 
-static ERTS_INLINE BIF_RETTYPE
-demonitor_local_port(Process *origin, Eterm ref, Eterm target)
-{
-    BIF_RETTYPE res = am_false;
-    Port        *port = erts_port_lookup_raw(target);
+   if (!erts_monitor_is_origin(mon))
+       return am_badarg;
 
-    if (!port) {
-        BIF_ERROR(origin, BADARG);
-    }
-    erts_smp_proc_unlock(origin, ERTS_PROC_LOCK_LINK);
+   erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p), mon);
 
-    if (port) {
-        Eterm trap_ref;
-        switch (erts_port_demonitor(origin, ERTS_PORT_DEMONITOR_NORMAL,
-                                    port, ref, &trap_ref)) {
-        case ERTS_PORT_OP_DROPPED:
-        case ERTS_PORT_OP_BADARG:
-            break;
-        case ERTS_PORT_OP_SCHEDULED:
-            BIF_TRAP3(await_port_send_result_trap, origin, trap_ref,
-                      am_busy_port, am_true);
-            /* the busy_port atom will never be returned, because it cannot be
-             * returned from erts_port_(de)monitor, but just in case if in future
-             * internal API changes - you may see this atom */
-        default:
-            break;
-        }
-    }
-    else {
-        ERTS_SMP_ASSERT_IS_NOT_EXITING(origin);
-    }
-    BIF_RET(res);
-}
+   switch (mon->type) {
 
-/* Can return atom true, false, yield, internal_error, badarg or
- * THE_NON_VALUE if error occurred or trap has been set up
- */
-static
-BIF_RETTYPE demonitor(Process *c_p, Eterm ref, Eterm *multip)
-{
-   ErtsMonitor  *mon = NULL;  /* The monitor entry to delete */
-   Eterm        to = NIL;     /* Monitor link traget */
-   DistEntry    *dep = NULL;  /* Target's distribution entry */
-   int          deref_de = 0;
-   BIF_RETTYPE  res = am_false;
-   int          unlock_link = 1;
+   case ERTS_MON_TYPE_TIME_OFFSET:
+       *multip = am_true;
+       erts_demonitor_time_offset(mon);
+       return am_true;
+
+   case ERTS_MON_TYPE_PORT: {
+       Port *prt;
+       ASSERT(is_internal_port(mon->other.item));
+       prt = erts_port_lookup(mon->other.item, ERTS_PORT_SFLGS_DEAD);
+       if (!prt || erts_port_demonitor(c_p, prt, mon) == ERTS_PORT_OP_DROPPED)
+           erts_monitor_release(mon);
+       return am_true;
+   }
 
-   erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_LINK);
+   case ERTS_MON_TYPE_PROC:
+       erts_proc_sig_send_demonitor(mon);
+       return am_true;
 
-   if (is_not_internal_ref(ref)) {
-       res = am_badarg;
-       goto done; /* Cannot be this monitor's ref */
-   }
+   case ERTS_MON_TYPE_DIST_PROC: {
+       ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+       Eterm to = mon->other.item;
+       DistEntry *dep;
+       int code = ERTS_DSIG_SEND_OK;
+       int deleted;
+       ErtsDSigData dsd;
 
-   mon = erts_lookup_monitor(ERTS_P_MONITORS(c_p), ref);
-   if (!mon) {
-       goto done;
-   }
+       ASSERT(is_external_pid(to) || is_node_name_atom(to));
 
-   switch (mon->type) {
-   case MON_TIME_OFFSET:
-       *multip = am_true;
-       erts_demonitor_time_offset(ref);
-       res = am_true;
-       break;
-   case MON_ORIGIN:
-       to = mon->u.pid;
-       *multip = am_false;
-       if (is_atom(to)) {
+       if (is_external_pid(to))
+           dep = external_pid_dist_entry(to);
+       else {
            /* Monitoring a name at node to */
-           ASSERT(is_node_name_atom(to));
            dep = erts_sysname_to_connected_dist_entry(to);
            ASSERT(dep != erts_this_dist_entry);
-           if (dep)
-               deref_de = 1;
-       } else if (is_port(to)) {
-           if (port_dist_entry(to) != erts_this_dist_entry) {
-               goto badarg;
+           if (!dep) {
+               erts_monitor_release(mon);
+               return am_false;
            }
-           res = demonitor_local_port(c_p, ref, to);
-           unlock_link = 0;
-           goto done;
-       } else {
-           ASSERT(is_pid(to));
-           dep = pid_dist_entry(to);
        }
-       if (dep != erts_this_dist_entry) {
-           res = remote_demonitor(c_p, dep, ref, to);
-           /* remote_demonitor() unlocks link lock on c_p */
-           unlock_link = 0;
+
+       code = erts_dsig_prepare(&dsd, dep, c_p, ERTS_PROC_LOCK_MAIN,
+                                ERTS_DSP_RLOCK, 0, 0);
+
+       deleted = erts_monitor_dist_delete(&mdp->target);
+
+       switch (code) {
+       case ERTS_DSIG_PREP_NOT_ALIVE:
+       case ERTS_DSIG_PREP_NOT_CONNECTED:
+           /*
+            * In the smp case this is possible if the node goes
+            * down just before the call to demonitor.
+            */
+           break;
+
+       case ERTS_DSIG_PREP_PENDING:
+       case ERTS_DSIG_PREP_CONNECTED: {
+           Eterm watched;
+
+           erts_de_runlock(dep);
+
+           if (mon->flags & ERTS_ML_FLG_NAME)
+               watched = ((ErtsMonitorDataExtended *) mdp)->u.name;
+           else
+               watched = to;
+
+           /*
+            * Soft (no force) send, use ->data in dist slot 
+            * monitor list since in case of monitor name 
+            * the atom is stored there. Yield if necessary.
+            */
+           code = erts_dsig_send_demonitor(&dsd, c_p->common.id,
+                                           watched, mdp->ref, 0);
+           break;
        }
-       else { /* Local monitor */
-           if (deref_de) {
-               deref_de = 0;
-               erts_deref_dist_entry(dep);
-           }
-           dep = NULL;
-           demonitor_local_process(c_p, ref, to, &res);
+
+       default:
+           ERTS_INTERNAL_ERROR("invalid result from erts_dsig_prepare()");
+           break;
        }
-       break;
-    default /* case */ :
-badarg:
-       res = am_badarg; /* will be converted to error by caller */
-       *multip = am_false;
-       break;
-   }
-done:
 
-   if (unlock_link)
-       erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_LINK);
+       if (deleted)
+           erts_monitor_release(&mdp->target);
 
-   if (deref_de) {
-       ASSERT(dep);
-       erts_deref_dist_entry(dep);
+       erts_monitor_release(mon);
+       return code == ERTS_DSIG_SEND_YIELD ? am_yield : am_true;
    }
 
-   ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
-   BIF_RET(res);
+   default:
+       ERTS_INTERNAL_ERROR("Unexpected monitor type");
+       return am_false;
+   }
 }
 
 BIF_RETTYPE demonitor_1(BIF_ALIST_1)
@@ -554,25 +383,23 @@ BIF_RETTYPE demonitor_1(BIF_ALIST_1)
     Eterm multi;
     switch (demonitor(BIF_P, BIF_ARG_1, &multi)) {
     case am_false:
-    case am_true:       BIF_RET(am_true);
-    case THE_NON_VALUE: BIF_RET(THE_NON_VALUE);
-    case am_yield:      ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
-    case am_badarg:     BIF_ERROR(BIF_P, BADARG);
-
-    case am_internal_error:
+    case am_true:
+        BIF_RET(am_true);
+    case am_yield:
+        ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
+    case am_badarg:
     default:
-	ASSERT(! "demonitor(): internal error");
-	BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
+        BIF_ERROR(BIF_P, BADARG);
     }
 }
 
 BIF_RETTYPE demonitor_2(BIF_ALIST_2)
 {
-    BIF_RETTYPE res = am_true;
-    Eterm       multi = am_false;
-    int         info = 0;
-    int         flush = 0;
-    Eterm       list = BIF_ARG_2;
+    BIF_RETTYPE res;
+    Eterm multi = am_false;
+    int info = 0;
+    int flush = 0;
+    Eterm list = BIF_ARG_2;
 
     while (is_list(list)) {
 	Eterm* consp = list_val(list);
@@ -592,10 +419,9 @@ BIF_RETTYPE demonitor_2(BIF_ALIST_2)
     if (is_not_nil(list))
 	goto badarg;
 
+    res = am_true;
     switch (demonitor(BIF_P, BIF_ARG_1, &multi)) {
-    case THE_NON_VALUE:
-        /* If other error occurred or trap has been set up - pass through */
-        BIF_RET(THE_NON_VALUE);
+
     case am_false:
 	if (info)
 	    res = am_false;
@@ -604,20 +430,29 @@ flush_messages:
 	    BIF_TRAP3(flush_monitor_messages_trap, BIF_P,
 		      BIF_ARG_1, multi, res);
 	}
+        /* Fall through... */
+
     case am_true:
 	if (multi == am_true && flush)
 	    goto flush_messages;
 	BIF_RET(res);
+
     case am_yield:
-	ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
+        /* return true after yield... */
+        if (flush) {
+            ERTS_VBUMP_ALL_REDS(BIF_P);
+            goto flush_messages;
+        }
+        ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
+
     case am_badarg:
-badarg:
-	BIF_ERROR(BIF_P, BADARG);
-    case am_internal_error:
     default:
-	ASSERT(! "demonitor(): internal error");
-	BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
+        break;
+
     }
+
+badarg:
+    BIF_ERROR(BIF_P, BADARG);
 }
 
 /* Type must be atomic object! */
@@ -657,305 +492,212 @@ erts_queue_monitor_message(Process *p,
     erts_queue_message(p, *p_locksp, msgp, tup, am_system);
 }
 
-static Eterm
-local_pid_monitor(Process *p, Eterm target, Eterm mon_ref, int boolean)
+BIF_RETTYPE monitor_2(BIF_ALIST_2)
 {
-    Eterm         ret = mon_ref;
-    Process      *rp;
-    ErtsProcLocks p_locks = ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK;
-
-    if (target == p->common.id) {
-	return ret;
-    }
+    Eterm target = BIF_ARG_2;
+    Eterm tmp_heap[3];
+    Eterm ref, id, name;
+    ErtsMonitorData *mdp;
+
+    if (BIF_ARG_1 == am_process) {
+        DistEntry *dep;
+        int byname;
+
+        if (is_internal_pid(target)) {
+            name = NIL;
+            id = target;
+
+        local_process:
+
+            ref = erts_make_ref(BIF_P);
+            if (id != BIF_P->common.id) {
+                mdp = erts_monitor_create(ERTS_MON_TYPE_PROC,
+                                          ref, BIF_P->common.id,
+                                          id, name);
+                erts_monitor_tree_insert(&ERTS_P_MONITORS(BIF_P),
+                                         &mdp->origin);
+
+                if (!erts_proc_sig_send_monitor(&mdp->target, id))
+                    erts_proc_sig_send_monitor_down(&mdp->target,
+                                                    am_noproc);
+            }
+            BIF_RET(ref);
+        }
 
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_LINK);
-    rp = erts_pid2proc_opt(p, p_locks,
-			   target, ERTS_PROC_LOCK_LINK,
-			   ERTS_P2P_FLG_ALLOW_OTHER_X);
-    if (!rp) {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	p_locks &= ~ERTS_PROC_LOCK_LINK;
-	if (boolean)
-	    ret = am_false;
-	else
-	    erts_queue_monitor_message(p, &p_locks,
-				       mon_ref, am_process, target, am_noproc);
-    }
-    else {
-	ASSERT(rp != p);
+        if (is_atom(target)) {
+        local_named_process:
+            name = target;
+            id = erts_whereis_name_to_id(BIF_P, target);
+            if (is_internal_pid(id))
+                goto local_process;
+            target = TUPLE2(&tmp_heap[0], name,
+                            erts_this_dist_entry->sysname);
+            goto noproc;
+        }
 
-	if (boolean)
-	    ret = am_true;
+        if (is_external_pid(target)) {
+            ErtsDSigData dsd;
+            int code;
+
+            dep = external_pid_dist_entry(target);
+            if (dep == erts_this_dist_entry)
+                goto noproc;
+
+            id = target;
+            name = NIL;
+            byname = 0;
+
+        remote_process:
+
+            ref = erts_make_ref(BIF_P);
+            mdp = erts_monitor_create(ERTS_MON_TYPE_DIST_PROC, ref,
+                                      BIF_P->common.id, id, name);
+            erts_monitor_tree_insert(&ERTS_P_MONITORS(BIF_P), &mdp->origin);
+
+            code = erts_dsig_prepare(&dsd, dep,
+                                     BIF_P, ERTS_PROC_LOCK_MAIN,
+                                     ERTS_DSP_RLOCK, 0, 1);
+            switch (code) {
+            case ERTS_DSIG_PREP_NOT_ALIVE:
+            case ERTS_DSIG_PREP_NOT_CONNECTED:
+                erts_monitor_set_dead_dist(&mdp->target, dep->sysname);
+                erts_proc_sig_send_monitor_down(&mdp->target, am_noconnection);
+                code = ERTS_DSIG_SEND_OK;
+                break;
 
-	erts_add_monitor(&ERTS_P_MONITORS(p), MON_ORIGIN, mon_ref, target, NIL);
-	erts_add_monitor(&ERTS_P_MONITORS(rp), MON_TARGET, mon_ref, p->common.id, NIL);
+            case ERTS_DSIG_PREP_PENDING:
+            case ERTS_DSIG_PREP_CONNECTED: {
+#ifdef DEBUG
+                int inserted =
+#endif
 
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-    }
+                erts_monitor_dist_insert(&mdp->target, dep->mld);
+                ASSERT(inserted);
+                erts_de_runlock(dep);
 
-    erts_smp_proc_unlock(p, p_locks & ~ERTS_PROC_LOCK_MAIN);
+                code = erts_dsig_send_monitor(&dsd, BIF_P->common.id, target, ref);
+                break;
+            }
 
-    return ret;
-}
+            default:
+                ERTS_ASSERT(! "Invalid dsig prepare result");
+                code = ERTS_DSIG_SEND_OK;
+                break;
+            }
 
-static BIF_RETTYPE
-local_port_monitor(Process *origin, Eterm target)
-{
-    BIF_RETTYPE ref = erts_make_ref(origin);
-    Port *port = erts_sig_lookup_port(origin, target);
-    ErtsProcLocks p_locks = ERTS_PROC_LOCK_MAIN;
+            if (byname)
+                erts_deref_dist_entry(dep);
 
-    if (!port) {
-res_no_proc:
-        /* Send the DOWN message immediately. Ref is made on the fly because
-         * caller has never seen it yet. */
-        erts_queue_monitor_message(origin, &p_locks, ref,
-                                   am_port, target, am_noproc);
-    }
-    else {
-        switch (erts_port_monitor(origin, port, target, &ref)) {
-        case ERTS_PORT_OP_DROPPED:
-        case ERTS_PORT_OP_BADARG:
-            goto res_no_proc;
-        case ERTS_PORT_OP_SCHEDULED:
-            BIF_TRAP3(await_port_send_result_trap, origin, ref,
-                      am_busy_port, ref);
-            /* the busy_port atom will never be returned, because it cannot be
-             * returned from erts_port_monitor, but just in case if in future
-             * internal API changes - you may see this atom */
-        default:
-            break;
+            if (code == ERTS_DSIG_SEND_YIELD)
+                ERTS_BIF_YIELD_RETURN(BIF_P, ref);
+            BIF_RET(ref);
         }
-    }
-    erts_smp_proc_unlock(origin, p_locks & ~ERTS_PROC_LOCK_MAIN);
-    BIF_RET(ref);
-}
 
-/* Type = process | port :: atom(), 1st argument passed to erlang:monitor/2
- */
-static BIF_RETTYPE
-local_name_monitor(Process *self, Eterm type, Eterm target_name)
-{
-    BIF_RETTYPE ret = erts_make_ref(self);
+        if (is_tuple(target)) {
+            Eterm *tpl = tuple_val(target);
+            if (arityval(tpl[0]) != 2)
+                goto badarg;
+            if (is_not_atom(tpl[1]) || is_not_atom(tpl[2]))
+                goto badarg;
+            if (!erts_is_alive && tpl[2] != am_Noname)
+                goto badarg;
+            target = tpl[1];
+            dep = erts_find_or_insert_dist_entry(tpl[2]);
+            if (dep == erts_this_dist_entry) {
+                erts_deref_dist_entry(dep);
+                goto local_named_process;
+            }
 
-    ErtsProcLocks p_locks = ERTS_PROC_LOCK_MAIN | ERTS_PROC_LOCK_LINK;
-    Process     *proc = NULL;
-    Port        *port = NULL;
+            id = dep->sysname;
+            name = target;
+            byname = 1;
+            goto remote_process;
+        }
 
-    erts_smp_proc_lock(self, ERTS_PROC_LOCK_LINK);
+        /* badarg... */
+    }
+    else if (BIF_ARG_1 == am_port) {
+
+        if (is_internal_port(target)) {
+            Port *prt;
+            name = NIL;
+            id = target;
+        local_port:
+            ref = erts_make_ref(BIF_P);
+            mdp = erts_monitor_create(ERTS_MON_TYPE_PORT, ref,
+                                      BIF_P->common.id, id, name);
+            erts_monitor_tree_insert(&ERTS_P_MONITORS(BIF_P), &mdp->origin);
+            prt = erts_port_lookup(id, ERTS_PORT_SFLGS_INVALID_LOOKUP);
+            if (!prt || erts_port_monitor(BIF_P, prt, &mdp->target) == ERTS_PORT_OP_DROPPED)
+                erts_proc_sig_send_monitor_down(&mdp->target, am_noproc);
+            BIF_RET(ref);
+        }
 
-    erts_whereis_name(self, p_locks, target_name,
-                      &proc, ERTS_PROC_LOCK_LINK,
-                      ERTS_P2P_FLG_ALLOW_OTHER_X,
-                      &port, 0);
+        if (is_atom(target)) {
+        local_named_port:
+            name = target;
+            id = erts_whereis_name_to_id(BIF_P, target);
+            if (is_internal_port(id))
+                goto local_port;
+            target = TUPLE2(&tmp_heap[0], name,
+                            erts_this_dist_entry->sysname);
+            goto noproc;
+        }
 
-    /* If the name is not registered,
-     * or if we asked for proc and got a port,
-     * or if we asked for port and got a proc,
-     * we just send the 'DOWN' message.
-     */
-    if ((!proc && !port) ||
-        (type == am_process && port) ||
-        (type == am_port && proc)) {
-        DeclareTmpHeap(lhp,3,self);
-	Eterm item;
-        UseTmpHeap(3,self);
-
-        erts_smp_proc_unlock(self, ERTS_PROC_LOCK_LINK);
-	p_locks &= ~ERTS_PROC_LOCK_LINK;
-
-	item = TUPLE2(lhp, target_name, erts_this_dist_entry->sysname);
-        erts_queue_monitor_message(self, &p_locks,
-                                   ret,
-                                   type, /* = process|port :: atom() */
-                                   item, am_noproc);
-        UnUseTmpHeap(3,self);
-    }
-    else if (port) {
-        erts_smp_proc_unlock(self, p_locks & ~ERTS_PROC_LOCK_MAIN);
-        p_locks &= ~ERTS_PROC_LOCK_MAIN;
-
-        switch (erts_port_monitor(self, port, target_name, &ret)) {
-        case ERTS_PORT_OP_DONE:
-            return ret;
-        case ERTS_PORT_OP_SCHEDULED: { /* Scheduled a signal */
-            ASSERT(is_internal_ordinary_ref(ret));
-            BIF_TRAP3(await_port_send_result_trap, self,
-                      ret, am_true, ret);
-            /* bif_trap returns */
-        } break;
-        default:
+        if (is_external_port(target)) {
+            if (erts_this_dist_entry == external_port_dist_entry(target))
+                goto noproc;
             goto badarg;
         }
-    }
-    else if (proc != self) {
-        erts_add_monitor(&ERTS_P_MONITORS(self), MON_ORIGIN, ret,
-                         proc->common.id, target_name);
-        erts_add_monitor(&ERTS_P_MONITORS(proc), MON_TARGET, ret,
-                         self->common.id, target_name);
-        erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_LINK);
-    }
 
-    if (p_locks) {
-        erts_smp_proc_unlock(self, p_locks & ~ERTS_PROC_LOCK_MAIN);
-    }
-    BIF_RET(ret);
-badarg:
-    if (p_locks) {
-        erts_smp_proc_unlock(self, p_locks & ~ERTS_PROC_LOCK_MAIN);
-    }
-    BIF_ERROR(self, BADARG);
-}
-
-static BIF_RETTYPE
-remote_monitor(Process *p, Eterm bifarg1, Eterm bifarg2,
-	       DistEntry *dep, Eterm target, int byname)
-{
-    ErtsDSigData dsd;
-    BIF_RETTYPE ret;
-    int code;
-
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_LINK);
-    code = erts_dsig_prepare(&dsd, dep, p, ERTS_DSP_RLOCK, 0);
-    switch (code) {
-    case ERTS_DSIG_PREP_NOT_ALIVE:
-	/* Let the dmonitor_p trap handle it */
-    case ERTS_DSIG_PREP_NOT_CONNECTED:
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	ERTS_BIF_PREP_TRAP2(ret, dmonitor_p_trap, p, bifarg1, bifarg2);
-	break;
-    case ERTS_DSIG_PREP_CONNECTED:
-	if (!(dep->flags & DFLAG_DIST_MONITOR)
-	    || (byname && !(dep->flags & DFLAG_DIST_MONITOR_NAME))) {
-	    erts_smp_de_runlock(dep);
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	    ERTS_BIF_PREP_ERROR(ret, p, BADARG);
-	}
-	else {
-	    Eterm p_trgt, p_name, d_name, mon_ref;
+        if (is_tuple(target)) {
+            Eterm *tpl = tuple_val(target);
+            if (arityval(tpl[0]) != 2)
+                goto badarg;
+            if (is_not_atom(tpl[1]) || is_not_atom(tpl[2]))
+                goto badarg;
+            if (tpl[2] == erts_this_dist_entry->sysname) {
+                target = tpl[1];
+                goto local_named_port;
+            }
+        }
 
-	    mon_ref = erts_make_ref(p);
+        /* badarg... */
+    }
+    else if (BIF_ARG_1 == am_time_offset) {
 
-	    if (byname) {
-		p_trgt = dep->sysname;
-		p_name = target;
-		d_name = target;
-	    }
-	    else {
-		p_trgt = target;
-		p_name = NIL;
-		d_name = NIL;
-	    }
+        if (target != am_clock_service)
+            goto badarg;
+	ref = erts_make_ref(BIF_P);
+        mdp = erts_monitor_create(ERTS_MON_TYPE_TIME_OFFSET,
+                                  ref, BIF_P->common.id,
+                                  am_clock_service, NIL);
+        erts_monitor_tree_insert(&ERTS_P_MONITORS(BIF_P), &mdp->origin);
 
-	    erts_smp_de_links_lock(dep);
+	erts_monitor_time_offset(&mdp->target);
 
-	    erts_add_monitor(&ERTS_P_MONITORS(p), MON_ORIGIN, mon_ref, p_trgt,
-			     p_name);
-	    erts_add_monitor(&(dep->monitors), MON_TARGET, mon_ref, p->common.id,
-			     d_name);
+        BIF_RET(ref);
+    }
 
-	    erts_smp_de_links_unlock(dep);
-	    erts_smp_de_runlock(dep);
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
+badarg:
 
-	    code = erts_dsig_send_monitor(&dsd, p->common.id, target, mon_ref);
-	    if (code == ERTS_DSIG_SEND_YIELD)
-		ERTS_BIF_PREP_YIELD_RETURN(ret, p, mon_ref);
-	    else
-		ERTS_BIF_PREP_RET(ret, mon_ref);
-	}
-	break;
-    default:
-	ASSERT(! "Invalid dsig prepare result");
-	ERTS_BIF_PREP_ERROR(ret, p, EXC_INTERNAL_ERROR);
-	break;
-    }
+    BIF_ERROR(BIF_P, BADARG);
 
-    BIF_RET(ret);
-}
-	
-BIF_RETTYPE monitor_2(BIF_ALIST_2)
-{
-    Eterm target = BIF_ARG_2;
-    BIF_RETTYPE ret;
-    DistEntry  *dep = NULL; 
-    int deref_de = 0;
+noproc: {
+        ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN;
 
-    /* Only process monitors are implemented */
-    switch (BIF_ARG_1) {
-    case am_time_offset: {
-	Eterm ref;
-        if (BIF_ARG_2 != am_clock_service) {
-            goto badarg;
-        }
-	ref = erts_make_ref(BIF_P);
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK);
-	erts_add_monitor(&ERTS_P_MONITORS(BIF_P), MON_TIME_OFFSET,
-			 ref, am_clock_service, NIL);
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-	erts_monitor_time_offset(BIF_P->common.id, ref);
-	BIF_RET(ref);
-    }
-    case am_process:
-    case am_port:
-	break;
-    default:
-        goto badarg;
-    }
+        ref = erts_make_ref(BIF_P);
+        erts_queue_monitor_message(BIF_P,
+                                   &locks,
+                                   ref,
+                                   BIF_ARG_1,
+                                   target,
+                                   am_noproc);
+        if (locks != ERTS_PROC_LOCK_MAIN)
+            erts_proc_unlock(BIF_P, locks & ~ERTS_PROC_LOCK_MAIN);
 
-    if (is_internal_pid(target) && BIF_ARG_1 == am_process) {
-local_pid:
-        ret = local_pid_monitor(BIF_P, target, erts_make_ref(BIF_P), 0);
-    } else if (is_external_pid(target) && BIF_ARG_1 == am_process) {
-	dep = external_pid_dist_entry(target);
-	if (dep == erts_this_dist_entry)
-	    goto local_pid;
-	ret = remote_monitor(BIF_P, BIF_ARG_1, BIF_ARG_2, dep, target, 0);
-    } else if (is_internal_port(target) && BIF_ARG_1 == am_port) {
-local_port:
-        ret = local_port_monitor(BIF_P, target);
-    } else if (is_external_port(target) && BIF_ARG_1 == am_port) {
-        dep = external_port_dist_entry(target);
-        if (dep == erts_this_dist_entry) {
-            goto local_port;
-        }
-        goto badarg; /* No want remote port */
-    } else if (is_atom(target)) {
-        ret = local_name_monitor(BIF_P, BIF_ARG_1, target);
-    } else if (is_tuple(target)) {
-	Eterm *tp = tuple_val(target);
-	Eterm remote_node;
-	Eterm name;
-        if (arityval(*tp) != 2) {
-            goto badarg;
-        }
-	remote_node = tp[2];
-	name = tp[1];
-	if (!is_atom(remote_node) || !is_atom(name)) {
-            goto badarg;
-	}
-	if (!erts_is_alive && remote_node != am_Noname) {
-            goto badarg; /* Remote monitor from (this) undistributed node */
-	}
-	dep = erts_sysname_to_connected_dist_entry(remote_node);
-	if (dep == erts_this_dist_entry) {
-	    deref_de = 1;
-            ret = local_name_monitor(BIF_P, BIF_ARG_1, name);
-	} else {
-	    if (dep)
-		deref_de = 1;
-	    ret = remote_monitor(BIF_P, BIF_ARG_1, BIF_ARG_2, dep, name, 1);
-	}
-    } else {
-badarg:
-	ERTS_BIF_PREP_ERROR(ret, BIF_P, BADARG);
-    }
-    if (deref_de) {
-	deref_de = 0;
-	erts_deref_dist_entry(dep);
+        BIF_RET(ref);
     }
-
-    return ret;
 }
 
 /**********************************************************************/
@@ -1009,7 +751,7 @@ BIF_RETTYPE spawn_opt_1(BIF_ALIST_1)
     so.max_heap_size  = H_MAX_SIZE;
     so.max_heap_flags = H_MAX_FLAGS;
     so.priority       = PRIORITY_NORMAL;
-    so.max_gen_gcs    = (Uint16) erts_smp_atomic32_read_nob(&erts_max_gen_gcs);
+    so.max_gen_gcs    = (Uint16) erts_atomic32_read_nob(&erts_max_gen_gcs);
     so.scheduler      = 0;
 
     /*
@@ -1128,181 +870,102 @@ BIF_RETTYPE spawn_opt_1(BIF_ALIST_1)
 /* remove a link from a process */
 BIF_RETTYPE unlink_1(BIF_ALIST_1)
 {
-    Process *rp;
-    DistEntry *dep;
-    ErtsLink *l = NULL, *rl = NULL;
-    ErtsProcLocks cp_locks = ERTS_PROC_LOCK_MAIN;
-
-    /*
-     * SMP specific note concerning incoming exit signals:
-     *   We have to have at least the status lock during removal of
-     *   the link half on current process, and check for and handle
-     *   a present pending exit while the status lock is held. This
-     *   in order to ensure that we wont be exited by a link after
-     *   it has been removed.
-     *
-     *   (We also have to have the link lock, of course, in order to
-     *    be allowed to remove the link...)
-     */
-
     if (IS_TRACED_FL(BIF_P, F_TRACE_PROCS)) {
-	trace_proc(BIF_P, cp_locks, BIF_P, am_unlink, BIF_ARG_1);
+        trace_proc(BIF_P, ERTS_PROC_LOCK_MAIN,
+                   BIF_P, am_unlink, BIF_ARG_1);
     }
 
-    if (is_internal_port(BIF_ARG_1)) {
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-#ifdef ERTS_SMP
-	if (ERTS_PROC_PENDING_EXIT(BIF_P))
-	    goto handle_pending_exit;
-#endif
-
-	l = erts_remove_link(&ERTS_P_LINKS(BIF_P), BIF_ARG_1);
+    if (is_internal_pid(BIF_ARG_1)) {
+        ErtsLink *lnk = erts_link_tree_lookup(ERTS_P_LINKS(BIF_P), BIF_ARG_1);
+        if (lnk) {
+            erts_link_tree_delete(&ERTS_P_LINKS(BIF_P), lnk);
+            erts_proc_sig_send_unlink(BIF_P, lnk);
+        }
+        BIF_RET(am_true);
+    }
 
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
+    if (is_internal_port(BIF_ARG_1)) {
+        ErtsLink *lnk = erts_link_tree_lookup(ERTS_P_LINKS(BIF_P), BIF_ARG_1);
 
-	if (l) {
+	if (lnk) {
+            Eterm ref;
+            Eterm *refp = erts_port_synchronous_ops ? &ref : NULL;
+            ErtsPortOpResult res = ERTS_PORT_OP_DROPPED;
 	    Port *prt;
 
-	    erts_destroy_link(l);
+            erts_link_tree_delete(&ERTS_P_LINKS(BIF_P), lnk);
 
 	    /* Send unlink signal */
 	    prt = erts_port_lookup(BIF_ARG_1, ERTS_PORT_SFLGS_DEAD);
 	    if (prt) {
-		ErtsPortOpResult res;
-		Eterm ref;
-		Eterm *refp = erts_port_synchronous_ops ? &ref : NULL;
 #ifdef DEBUG
 		ref = NIL;
 #endif
-		res = erts_port_unlink(BIF_P, prt, BIF_P->common.id, refp);
+		res = erts_port_unlink(BIF_P, prt, lnk, refp);
 
-		if (refp && res == ERTS_PORT_OP_SCHEDULED) {
-		    ASSERT(is_internal_ordinary_ref(ref));
-		    BIF_TRAP3(await_port_send_result_trap, BIF_P, ref, am_true, am_true);
-		}
 	    }
+
+            if (res == ERTS_PORT_OP_DROPPED)
+                erts_link_release(lnk);
+            else if (refp && res == ERTS_PORT_OP_SCHEDULED) {
+                ASSERT(is_internal_ordinary_ref(ref));
+                BIF_TRAP3(await_port_send_result_trap, BIF_P, ref, am_true, am_true);
+            }
 	}
 
 	BIF_RET(am_true);
     }
-    else if (is_external_port(BIF_ARG_1)
-	     && external_port_dist_entry(BIF_ARG_1) == erts_this_dist_entry) {
-	BIF_RET(am_true);
-    }
-
-    if (is_not_pid(BIF_ARG_1))
-	BIF_ERROR(BIF_P, BADARG);
 
     if (is_external_pid(BIF_ARG_1)) {
-	ErtsDistLinkData dld;
+        ErtsLink *lnk, *dlnk;
+        ErtsLinkData *ldp;
+        DistEntry *dep;
 	int code;
 	ErtsDSigData dsd;
-	/* Blind removal, we might have trapped or anything, this leaves
-	   us in a state where monitors might be inconsistent, but the dist
-	   code should take care of it. */
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-#ifdef ERTS_SMP
-	if (ERTS_PROC_PENDING_EXIT(BIF_P))
-	    goto handle_pending_exit;
-#endif
-	l = erts_remove_link(&ERTS_P_LINKS(BIF_P), BIF_ARG_1);
-
-	erts_smp_proc_unlock(BIF_P,
-			     ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-
-	if (l)
-	    erts_destroy_link(l);
 
 	dep = external_pid_dist_entry(BIF_ARG_1);
-	if (dep == erts_this_dist_entry) {
+	if (dep == erts_this_dist_entry)
 	    BIF_RET(am_true);
-	}
 
-	code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_DSP_NO_LOCK, 0);
+        lnk = erts_link_tree_lookup(ERTS_P_LINKS(BIF_P), BIF_ARG_1);
+        if (!lnk)
+            BIF_RET(am_true);
+
+        erts_link_tree_delete(&ERTS_P_LINKS(BIF_P), lnk);
+        dlnk = erts_link_to_other(lnk, &ldp);
+
+        if (erts_link_dist_delete(dlnk))
+            erts_link_release_both(ldp);
+        else
+            erts_link_release(lnk);
+
+	code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_PROC_LOCK_MAIN,
+				 ERTS_DSP_NO_LOCK, 0, 0);
 	switch (code) {
 	case ERTS_DSIG_PREP_NOT_ALIVE:
 	case ERTS_DSIG_PREP_NOT_CONNECTED:
-#if 1
 	    BIF_RET(am_true);
-#else
-	    /*
-	     * This is how we used to do it, but the link is obviously not
-	     * active, so I see no point in setting up a connection.
-	     * /Rickard
-	     */
-	    BIF_TRAP1(dunlink_trap, BIF_P, BIF_ARG_1);
-#endif
-
+	case ERTS_DSIG_PREP_PENDING:
 	case ERTS_DSIG_PREP_CONNECTED:
-	    erts_remove_dist_link(&dld, BIF_P->common.id, BIF_ARG_1, dep);
 	    code = erts_dsig_send_unlink(&dsd, BIF_P->common.id, BIF_ARG_1);
-	    erts_destroy_dist_link(&dld);
 	    if (code == ERTS_DSIG_SEND_YIELD)
 		ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
-	    BIF_RET(am_true);
-
+            break;
 	default:
 	    ASSERT(! "Invalid dsig prepare result");
 	    BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
 	}
-    }
-
-    /* Internal pid... */
-
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-
-    cp_locks |= ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS;
 
-    /* get process struct */
-    rp = erts_pid2proc_opt(BIF_P, cp_locks,
-			   BIF_ARG_1, ERTS_PROC_LOCK_LINK,
-			   ERTS_P2P_FLG_ALLOW_OTHER_X);
-
-#ifdef ERTS_SMP
-    if (ERTS_PROC_PENDING_EXIT(BIF_P)) {
-	if (rp && rp != BIF_P)
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	goto handle_pending_exit;
-    }
-#endif
-
-    /* unlink and ignore errors */
-    l = erts_remove_link(&ERTS_P_LINKS(BIF_P), BIF_ARG_1);
-    if (l != NULL)
-	erts_destroy_link(l);
-
-    if (!rp) {
-	ERTS_SMP_ASSERT_IS_NOT_EXITING(BIF_P);
+        BIF_RET(am_true);
     }
-    else {
-	rl = erts_remove_link(&ERTS_P_LINKS(rp), BIF_P->common.id);
-	if (rl != NULL)
-	    erts_destroy_link(rl);
-
-	if (IS_TRACED_FL(rp, F_TRACE_PROCS) && rl != NULL) {
-            erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_STATUS);
-            cp_locks &= ~ERTS_PROC_LOCK_STATUS;
-	    trace_proc(BIF_P, (ERTS_PROC_LOCK_MAIN | ERTS_PROC_LOCK_LINK),
-                       rp, am_getting_unlinked, BIF_P->common.id);
-	}
 
-	if (rp != BIF_P)
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
+    if (is_external_port(BIF_ARG_1)) {
+        if (external_port_dist_entry(BIF_ARG_1) == erts_this_dist_entry)
+            BIF_RET(am_true);
+        /* Links to Remote ports not supported... */
     }
- 
-    erts_smp_proc_unlock(BIF_P, cp_locks & ~ERTS_PROC_LOCK_MAIN);
 
-    BIF_RET(am_true);
-
-#ifdef ERTS_SMP
- handle_pending_exit:
-    erts_handle_pending_exit(BIF_P, (ERTS_PROC_LOCK_MAIN
-				     | ERTS_PROC_LOCK_LINK
-				     | ERTS_PROC_LOCK_STATUS));
-    ASSERT(ERTS_PROC_IS_EXITING(BIF_P)); 
-    erts_smp_proc_unlock(BIF_P,	ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-    ERTS_BIF_EXITED(BIF_P);
-#endif
+    BIF_ERROR(BIF_P, BADARG);
 }
 
 BIF_RETTYPE hibernate_3(BIF_ALIST_3)
@@ -1314,7 +977,11 @@ BIF_RETTYPE hibernate_3(BIF_ALIST_3)
      */
     Eterm reg[3];
 
-    if (erts_hibernate(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, reg)) {
+    reg[0] = BIF_ARG_1;
+    reg[1] = BIF_ARG_2;
+    reg[2] = BIF_ARG_3;
+
+    if (erts_hibernate(BIF_P, reg)) {
         /*
          * If hibernate succeeded, TRAP. The process will be wait in a
          * hibernated state if its state is inactive (!ERTS_PSFLG_ACTIVE);
@@ -1483,6 +1150,13 @@ BIF_RETTYPE raise_3(BIF_ALIST_3)
     /* Create stacktrace and store */
     if (erts_backtrace_depth < depth) {
 	depth = erts_backtrace_depth;
+        if (depth == 0) {
+            /*
+             * For consistency with stacktraces generated
+             * automatically, always include one element.
+             */
+            depth = 1;
+        }
 	must_copy = 1;
     }
     if (must_copy) {
@@ -1543,22 +1217,69 @@ BIF_RETTYPE raise_3(BIF_ALIST_3)
     return am_badarg;
 }
 
+static BIF_RETTYPE
+erts_internal_await_exit_trap(BIF_ALIST_0)
+{
+    /*
+     * We have sent ourselves an exit signal which will
+     * terminate ourselves. Handle all signals until
+     * terminated in order to ensure that signal order
+     * is preserved. Yield if necessary.
+     */
+    erts_aint32_t state;
+    int reds = ERTS_BIF_REDS_LEFT(BIF_P);
+    (void) erts_proc_sig_handle_incoming(BIF_P, &state, &reds,
+                                         reds, !0);
+    BUMP_REDS(BIF_P, reds);
+    if (state & ERTS_PSFLG_EXITING)
+        ERTS_BIF_EXITED(BIF_P);
+
+    ERTS_BIF_YIELD0(&await_exit_trap, BIF_P);
+}
+
 /**********************************************************************/
-/* send an exit message to another process (if trapping exits) or
-   exit the other process */
+/* send an exit signal to another process */
 
-BIF_RETTYPE exit_2(BIF_ALIST_2)
+static BIF_RETTYPE send_exit_signal_bif(Process *c_p, Eterm id, Eterm reason, int exit2)
 {
-     Process *rp;
+    BIF_RETTYPE ret_val;
 
-     /*
-      * If the first argument is not a pid, or a local port it is an error.
-      */
+    /*
+     * 'id' not a process id, nor a local port id is a 'badarg' error.
+     */
 
-     if (is_internal_port(BIF_ARG_1)) {
+     if (is_internal_pid(id)) {
+         /*
+          * Preserve the very old and *very strange* behaviour
+          * of erlang:exit/2...
+          *
+          * - terminate ourselves even though exit reason
+          *   is normal (unless we trap exit)
+          * - terminate ourselves before exit/2 return
+          */
+         int exit2_suicide = (exit2
+                              && c_p->common.id == id
+                              && (reason == am_kill
+                                  || !(c_p->flags & F_TRAP_EXIT)));
+         erts_proc_sig_send_exit(c_p, c_p->common.id, id,
+                                 reason, NIL, exit2_suicide);
+         if (!exit2_suicide)
+             ERTS_BIF_PREP_RET(ret_val, am_true);
+         else {
+             erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+             erts_proc_sig_fetch(c_p);
+             erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+             ERTS_BIF_PREP_TRAP0(ret_val, &await_exit_trap, c_p);
+         }
+     }
+     else if (is_internal_port(id)) {
 	 Eterm ref, *refp;
 	 Uint32 invalid_flags;
 	 Port *prt;
+         ErtsPortOpResult res = ERTS_PORT_OP_DONE;
+#ifdef DEBUG
+         ref = NIL;
+#endif
 
 	 if (erts_port_synchronous_ops) {
 	     refp = &ref;
@@ -1569,120 +1290,88 @@ BIF_RETTYPE exit_2(BIF_ALIST_2)
 	     invalid_flags = ERTS_PORT_SFLGS_INVALID_LOOKUP;
 	 }
 
-	 prt = erts_port_lookup(BIF_ARG_1, invalid_flags);
-
-	 if (prt) {
-	     ErtsPortOpResult res;
-
-#ifdef DEBUG
-	     ref = NIL;
-#endif
-
-	     res = erts_port_exit(BIF_P, 0, prt, BIF_P->common.id, BIF_ARG_2, refp);
-
-	     ERTS_BIF_CHK_EXITED(BIF_P);
-
-	     if (refp && res == ERTS_PORT_OP_SCHEDULED) {
-		 ASSERT(is_internal_ordinary_ref(ref));
-		 BIF_TRAP3(await_port_send_result_trap, BIF_P, ref, am_true, am_true);
-	     }
-
-	 }
-
-	 BIF_RET(am_true);
+	 prt = erts_port_lookup(id, invalid_flags);
+	 if (prt)
+	     res = erts_port_exit(c_p, 0, prt, c_p->common.id, reason, refp);
+
+         if (!refp || res != ERTS_PORT_OP_SCHEDULED)
+             ERTS_BIF_PREP_RET(ret_val, am_true);
+         else {
+             ASSERT(is_internal_ordinary_ref(ref));
+             ERTS_BIF_PREP_TRAP3(ret_val, await_port_send_result_trap,
+                                 c_p, ref, am_true, am_true);
+         }
      }
-     else if(is_external_port(BIF_ARG_1)
-	     && external_port_dist_entry(BIF_ARG_1) == erts_this_dist_entry)
-	 BIF_RET(am_true);
-     
-     /*
-      * If it is a remote pid, send a signal to the remote node.
-      */
-
-     if (is_external_pid(BIF_ARG_1)) {
-	 int code;
-	 ErtsDSigData dsd;
-	 DistEntry *dep;
-
-	 dep = external_pid_dist_entry(BIF_ARG_1);
-	 if(dep == erts_this_dist_entry)
-	     BIF_RET(am_true);
-
-	 code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_DSP_NO_LOCK, 0);
-	 switch (code) {
-	 case ERTS_DSIG_PREP_NOT_ALIVE:
-	 case ERTS_DSIG_PREP_NOT_CONNECTED:
-	     BIF_TRAP2(dexit_trap, BIF_P, BIF_ARG_1, BIF_ARG_2);
-	 case ERTS_DSIG_PREP_CONNECTED:
-	     code = erts_dsig_send_exit2(&dsd, BIF_P->common.id, BIF_ARG_1, BIF_ARG_2);
-	     if (code == ERTS_DSIG_SEND_YIELD)
-		 ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
-	     BIF_RET(am_true);
-	 default:
-	     ASSERT(! "Invalid dsig prepare result");
-	     BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
-	 }
+     else if (is_external_pid(id)) {
+	 DistEntry *dep = external_pid_dist_entry(id);
+	 if (dep == erts_this_dist_entry)
+             ERTS_BIF_PREP_RET(ret_val, am_true); /* Old incarnation of this node... */
+         else {
+             int code;
+             ErtsDSigData dsd;
+
+             code = erts_dsig_prepare(&dsd, dep, c_p, ERTS_PROC_LOCK_MAIN,
+                                      ERTS_DSP_NO_LOCK, 0, 1);
+             switch (code) {
+             case ERTS_DSIG_PREP_NOT_ALIVE:
+             case ERTS_DSIG_PREP_NOT_CONNECTED:
+                 ERTS_BIF_PREP_RET(ret_val, am_true);
+                 break;
+             case ERTS_DSIG_PREP_PENDING:
+             case ERTS_DSIG_PREP_CONNECTED:
+                 code = erts_dsig_send_exit2(&dsd, c_p->common.id, id, reason);
+                 if (code == ERTS_DSIG_SEND_YIELD)
+                     ERTS_BIF_PREP_YIELD_RETURN(ret_val, c_p, am_true);
+                 else
+                     ERTS_BIF_PREP_RET(ret_val, am_true);
+                 break;
+             default:
+                 ASSERT(! "Invalid dsig prepare result");
+                 ERTS_BIF_PREP_ERROR(ret_val, c_p, EXC_INTERNAL_ERROR);
+                 break;
+             }
+         }
      }
-     else if (is_not_internal_pid(BIF_ARG_1)) {
-       BIF_ERROR(BIF_P, BADARG);
+     else if (is_external_port(id)) {
+	 DistEntry *dep = external_port_dist_entry(id);
+	 if(dep == erts_this_dist_entry)
+             ERTS_BIF_PREP_RET(ret_val, am_true); /* Old incarnation of this node... */
+         else
+             ERTS_BIF_PREP_ERROR(ret_val, c_p, BADARG);
      }
      else {
-	 /*
-	  * The pid is internal.  Verify that it refers to an existing process.
-	  */
-	 ErtsProcLocks rp_locks;
-
-	 if (BIF_ARG_1 == BIF_P->common.id) {
-	     rp_locks = ERTS_PROC_LOCKS_ALL;
-	     rp = BIF_P;
-	     erts_smp_proc_lock(rp, ERTS_PROC_LOCKS_ALL_MINOR);
-	 }
-	 else {
-	     rp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
-	     rp = erts_pid2proc(BIF_P, ERTS_PROC_LOCK_MAIN,
-				BIF_ARG_1, rp_locks);
-	     if (!rp) {
-		 BIF_RET(am_true);
-	     }
-	 }
+         /* Not an id of a process or a port... */
 
-	 /*
-	  * Send an exit signal.
-	  */
-	 erts_send_exit_signal(BIF_P,
-			       BIF_P->common.id,
-			       rp,
-			       &rp_locks,
-			       BIF_ARG_2,
-			       NIL,
-			       NULL,
-			       BIF_P == rp ? ERTS_XSIG_FLG_NO_IGN_NORMAL : 0);
-#ifdef ERTS_SMP
-	 if (rp == BIF_P)
-	     rp_locks &= ~ERTS_PROC_LOCK_MAIN;
-	 if (rp_locks)
-	     erts_smp_proc_unlock(rp, rp_locks);
-#endif
-	 /*
-	  * We may have exited ourselves and may have to take action.
-	  */
-	 ERTS_BIF_CHK_EXITED(BIF_P);
-	 BIF_RET(am_true);
+         ERTS_BIF_PREP_ERROR(ret_val, c_p, BADARG);
      }
+
+     return ret_val;
 }
 
+BIF_RETTYPE exit_2(BIF_ALIST_2)
+{
+    return send_exit_signal_bif(BIF_P, BIF_ARG_1, BIF_ARG_2, !0);
+}
+
+BIF_RETTYPE exit_signal_2(BIF_ALIST_2)
+{
+    return send_exit_signal_bif(BIF_P, BIF_ARG_1, BIF_ARG_2, 0);
+}
+
+
 /**********************************************************************/
 /* this sets some process info- trapping exits or the error handler */
 
 
 /* Handle flags common to both process_flag_2 and process_flag_3. */
-static BIF_RETTYPE process_flag_aux(Process *BIF_P,
-				    Process *rp,
-				    Eterm flag,
-				    Eterm val)
+static Eterm process_flag_aux(Process *c_p, int *redsp, Eterm flag, Eterm val)
 {
    Eterm old_value = NIL;	/* shut up warning about use before set */
    Sint i;
+
+   if (redsp)
+       *redsp = 1;
+
    if (flag == am_save_calls) {
        struct saved_calls *scb;
        if (!is_small(val))
@@ -1702,30 +1391,89 @@ static BIF_RETTYPE process_flag_aux(Process *BIF_P,
        }
 
 #ifdef HIPE
-       if (rp->flags & F_HIPE_MODE) {
-	   ASSERT(!ERTS_PROC_GET_SAVED_CALLS_BUF(rp));
-	   scb = ERTS_PROC_SET_SUSPENDED_SAVED_CALLS_BUF(rp, scb);
+       if (c_p->flags & F_HIPE_MODE) {
+	   ASSERT(!ERTS_PROC_GET_SAVED_CALLS_BUF(c_p));
+	   scb = ERTS_PROC_SET_SUSPENDED_SAVED_CALLS_BUF(c_p, scb);
        }
        else
 #endif
        {
 #ifdef HIPE
-	   ASSERT(!ERTS_PROC_GET_SUSPENDED_SAVED_CALLS_BUF(rp));
+	   ASSERT(!ERTS_PROC_GET_SUSPENDED_SAVED_CALLS_BUF(c_p));
 #endif
-	   scb = ERTS_PROC_SET_SAVED_CALLS_BUF(rp, scb);
-	   if (rp == BIF_P && ((scb && i == 0) || (!scb && i != 0))) {
-	       /* Adjust fcalls to match save calls setting... */
-	       if (i == 0)
-		   BIF_P->fcalls += CONTEXT_REDS; /* disabled it */
-	       else
-		   BIF_P->fcalls -= CONTEXT_REDS; /* enabled it */
-
-	       /*
-		* Make sure we reschedule immediately so the
-		* change take effect at once.
-		*/
-	       ERTS_VBUMP_ALL_REDS(BIF_P);
-	   }
+	   scb = ERTS_PROC_SET_SAVED_CALLS_BUF(c_p, scb);
+
+	   if (((scb && i == 0) || (!scb && i != 0))) {
+
+               /*
+                * Make sure we reschedule immediately so the
+                * change take effect at once.
+                */
+               if (!redsp) {
+                   /* Executed via BIF call.. */
+               via_bif:
+
+                   /* Adjust fcalls to match save calls setting... */
+                   if (i == 0)
+                       c_p->fcalls += CONTEXT_REDS; /* disabled it */
+                   else
+                       c_p->fcalls -= CONTEXT_REDS; /* enabled it */
+
+                   ERTS_VBUMP_ALL_REDS(c_p);
+               }
+               else {
+                   erts_aint32_t state;
+                   /*
+                    * Executed via signal handler. Try to figure
+                    * out in what context we are executing...
+                    */
+
+                   state = erts_atomic32_read_nob(&c_p->state);
+                   if (state & (ERTS_PSFLG_RUNNING_SYS
+                                | ERTS_PSFLG_DIRTY_RUNNING_SYS
+                                | ERTS_PSFLG_DIRTY_RUNNING)) {
+                       /*
+                        * We are either processing signals before
+                        * being executed or executing dirty. That
+                        * is, no need to adjust anything...
+                        */
+                       *redsp = 1;
+                   }
+                   else {
+                       ErtsSchedulerData *esdp;
+                       ASSERT(state & ERTS_PSFLG_RUNNING);
+
+                       /*
+                        * F_DELAY_GC is currently only set when
+                        * we handle signals in state running via
+                        * receive helper...
+                        */
+
+                       if (!(c_p->flags & F_DELAY_GC)) {
+                           *redsp = 1;
+                           goto via_bif;
+                       }
+
+                       /*
+                        * Executing via receive helper...
+                        *
+                        * We utilize the virtual reds counter
+                        * in order to get correct calculation
+                        * of reductions consumed when scheduling
+                        * out the process...
+                        */
+
+                       esdp = erts_get_scheduler_data();
+
+                       if (i == 0)
+                           esdp->virtual_reds += CONTEXT_REDS; /* disabled it */
+                       else
+                           esdp->virtual_reds -= CONTEXT_REDS; /* enabled it */
+
+                       *redsp = -1;
+                   }
+               }
+           }
        }
 
        if (!scb)
@@ -1735,11 +1483,12 @@ static BIF_RETTYPE process_flag_aux(Process *BIF_P,
 	   erts_free(ERTS_ALC_T_CALLS_BUF, (void *) scb);
        }
 
-       BIF_RET(old_value);
+       ASSERT(is_immed(old_value));
+       return old_value;
    }
 
  error:
-   BIF_ERROR(BIF_P, BADARG);
+   return am_badarg;
 }
 
 BIF_RETTYPE process_flag_2(BIF_ALIST_2)
@@ -1759,44 +1508,18 @@ BIF_RETTYPE process_flag_2(BIF_ALIST_2)
        BIF_RET(old_value);
    }
    else if (BIF_ARG_1 == am_trap_exit) {
-       erts_aint32_t state;
-       Uint trap_exit;
-       if (BIF_ARG_2 == am_true) {
-	   trap_exit = 1;
-       } else if (BIF_ARG_2 == am_false) {
-	   trap_exit = 0;
-       } else {
-	   goto error;
-       }
-       /*
-	* NOTE: It is important that we check for pending exit signals
-	*       and handle them before returning if trap_exit is set to
-	*       true. For more info, see implementation of
-	*       erts_send_exit_signal().
-	*/
-       erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCKS_XSIG_SEND);
-       if (trap_exit)
-	   state = erts_smp_atomic32_read_bor_mb(&BIF_P->state,
-						 ERTS_PSFLG_TRAP_EXIT);
+       old_value = (BIF_P->flags & F_TRAP_EXIT) ? am_true : am_false;
+       if (BIF_ARG_2 == am_true)
+           BIF_P->flags |= F_TRAP_EXIT;
+       else if (BIF_ARG_2 == am_false)
+           BIF_P->flags &= ~F_TRAP_EXIT;
        else
-	   state = erts_smp_atomic32_read_band_mb(&BIF_P->state,
-						  ~ERTS_PSFLG_TRAP_EXIT);
-       erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCKS_XSIG_SEND);
-
-#ifdef ERTS_SMP
-       if (state & ERTS_PSFLG_PENDING_EXIT) {
-	   erts_handle_pending_exit(BIF_P, ERTS_PROC_LOCK_MAIN);
-	   ERTS_BIF_EXITED(BIF_P);
-       }
-#endif
-
-       old_value = (state & ERTS_PSFLG_TRAP_EXIT) ? am_true : am_false;
+	   goto error;
        BIF_RET(old_value);
    }
    else if (BIF_ARG_1 == am_scheduler) {
        ErtsRunQueue *old, *new, *curr;
        Sint sched;
-       erts_aint32_t state;
 
        if (!is_small(BIF_ARG_2))
 	   goto error;
@@ -1805,25 +1528,23 @@ BIF_RETTYPE process_flag_2(BIF_ALIST_2)
 	   goto error;
 
        if (sched == 0) {
+           old = erts_bind_runq_proc(BIF_P, 0);
 	   new = NULL;
-	   state = erts_smp_atomic32_read_band_mb(&BIF_P->state,
-						  ~ERTS_PSFLG_BOUND);
        }
        else {
+           int bound = !0;
 	   new = erts_schedid2runq(sched);
-#ifdef ERTS_SMP
-	   erts_atomic_set_nob(&BIF_P->run_queue, (erts_aint_t) new);
-#endif
-	   state = erts_smp_atomic32_read_bor_mb(&BIF_P->state,
-						 ERTS_PSFLG_BOUND);
+           old = erts_set_runq_proc(BIF_P, new, &bound);
+           if (!bound)
+               old = NULL;
        }
 
+       old_value = old ? make_small(old->ix+1) : make_small(0);
+
        curr = erts_proc_sched_data(BIF_P)->run_queue;
-       old = (ERTS_PSFLG_BOUND & state) ? curr : NULL;
 
        ASSERT(!old || old == curr);
 
-       old_value = old ? make_small(old->ix+1) : make_small(0);
        if (new && new != curr)
 	   ERTS_BIF_YIELD_RETURN_X(BIF_P, old_value, am_scheduler);
        else
@@ -1895,7 +1616,7 @@ BIF_RETTYPE process_flag_2(BIF_ALIST_2)
        } else {
 	   goto error;
        }
-       erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCKS_ALL_MINOR);
+       erts_proc_lock(BIF_P, ERTS_PROC_LOCKS_ALL_MINOR);
        old_value = (ERTS_TRACE_FLAGS(BIF_P) & F_SENSITIVE
 		    ? am_true
 		    : am_false);
@@ -1904,7 +1625,7 @@ BIF_RETTYPE process_flag_2(BIF_ALIST_2)
        } else {
 	   ERTS_TRACE_FLAGS(BIF_P) &= ~F_SENSITIVE;
        }
-       erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCKS_ALL_MINOR);
+       erts_proc_unlock(BIF_P, ERTS_PROC_LOCKS_ALL_MINOR);
        /* make sure to bump all reds so that we get
           rescheduled immediately so setting takes effect */
        BIF_RET2(old_value, CONTEXT_REDS);
@@ -1936,33 +1657,73 @@ BIF_RETTYPE process_flag_2(BIF_ALIST_2)
        /* Fall through and try process_flag_aux() ... */
    }
 
-   BIF_RET(process_flag_aux(BIF_P, BIF_P, BIF_ARG_1, BIF_ARG_2));
+   old_value = process_flag_aux(BIF_P, NULL, BIF_ARG_1, BIF_ARG_2);
+   if (old_value != am_badarg)
+       BIF_RET(old_value);
  error:
    BIF_ERROR(BIF_P, BADARG);
 }
 
-BIF_RETTYPE process_flag_3(BIF_ALIST_3)
+typedef struct {
+    Eterm flag;
+    Eterm value;
+    ErlOffHeap oh;
+    Eterm heap[1];
+} ErtsProcessFlag3Args;
+
+static Eterm
+exec_process_flag_3(Process *c_p, void *arg, int *redsp, ErlHeapFragment **bpp)
 {
-   Process *rp;
-   Eterm res;
-
-#ifdef ERTS_SMP
-   rp = erts_pid2proc_not_running(BIF_P, ERTS_PROC_LOCK_MAIN,
-				  BIF_ARG_1, ERTS_PROC_LOCK_MAIN);
-   if (rp == ERTS_PROC_LOCK_BUSY)
-       ERTS_BIF_YIELD3(bif_export[BIF_process_flag_3], BIF_P,
-		       BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
-#else
-   rp = erts_proc_lookup(BIF_ARG_1);
-#endif
+    ErtsProcessFlag3Args *pf3a = arg;
+    Eterm res;
+
+    if (ERTS_PROC_IS_EXITING(c_p))
+        res = am_badarg;
+    else
+        res = process_flag_aux(c_p, redsp, pf3a->flag, pf3a->value);
+    erts_cleanup_offheap(&pf3a->oh);
+    erts_free(ERTS_ALC_T_PF3_ARGS, arg);
+    return res;
+}
+
+
+BIF_RETTYPE erts_internal_process_flag_3(BIF_ALIST_3)
+{
+   Eterm res, *hp;
+   ErlOffHeap *ohp;
+   ErtsProcessFlag3Args *pf3a;
+   Uint flag_sz, value_sz;
+
+   if (BIF_P->common.id == BIF_ARG_1) {
+       res = process_flag_aux(BIF_P, NULL, BIF_ARG_2, BIF_ARG_3);
+       BIF_RET(res);
+   }
+
+   if (is_not_internal_pid(BIF_ARG_1))
+       BIF_RET(am_badarg);
 
-   if (!rp)
-       BIF_ERROR(BIF_P, BADARG);
+   flag_sz = is_immed(BIF_ARG_2) ? 0 : size_object(BIF_ARG_2);
+   value_sz = is_immed(BIF_ARG_3) ? 0 : size_object(BIF_ARG_3);
 
-   res = process_flag_aux(BIF_P, rp, BIF_ARG_2, BIF_ARG_3);
+   pf3a = erts_alloc(ERTS_ALC_T_PF3_ARGS,
+                     sizeof(ErtsProcessFlag3Args)
+                     + sizeof(Eterm)*(flag_sz+value_sz-1));
 
-   if (rp != BIF_P)
-       erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+   ohp = &pf3a->oh;
+   ERTS_INIT_OFF_HEAP(&pf3a->oh);
+
+   hp = &pf3a->heap[0];
+
+   pf3a->flag = copy_struct(BIF_ARG_2, flag_sz, &hp, ohp);
+   pf3a->value = copy_struct(BIF_ARG_3, value_sz, &hp, ohp);
+
+   res = erts_proc_sig_send_rpc_request(BIF_P, BIF_ARG_1,
+                                        !0,
+                                        exec_process_flag_3,
+                                        (void *) pf3a);
+
+   if (is_non_value(res))
+       BIF_RET(am_badarg);
 
    return res;
 }
@@ -2034,7 +1795,7 @@ ebif_bang_2(BIF_ALIST_2)
  * Send a message to Process, Port or Registered Process.
  * Returns non-negative reduction bump or negative result code.
  */
-#define SEND_TRAP		(-1)
+#define SEND_NOCONNECT		(-1)
 #define SEND_YIELD		(-2)
 #define SEND_YIELD_RETURN	(-3)
 #define SEND_BADARG		(-4)
@@ -2050,19 +1811,22 @@ static Sint remote_send(Process *p, DistEntry *dep,
 {
     Sint res;
     int code;
-
     ASSERT(is_atom(to) || is_external_pid(to));
 
-    code = erts_dsig_prepare(&ctx->dsd, dep, p, ERTS_DSP_NO_LOCK, !ctx->suspend);
+    ctx->dep = dep;
+    code = erts_dsig_prepare(&ctx->dsd, dep, p, ERTS_PROC_LOCK_MAIN,
+			     ERTS_DSP_NO_LOCK,
+			     !ctx->suspend, ctx->connect);
     switch (code) {
     case ERTS_DSIG_PREP_NOT_ALIVE:
     case ERTS_DSIG_PREP_NOT_CONNECTED:
-	res = SEND_TRAP;
+	res = SEND_NOCONNECT;
 	break;
     case ERTS_DSIG_PREP_WOULD_SUSPEND:
 	ASSERT(!ctx->suspend);
 	res = SEND_YIELD;
 	break;
+    case ERTS_DSIG_PREP_PENDING:
     case ERTS_DSIG_PREP_CONNECTED: {
 
 	if (is_atom(to))
@@ -2198,9 +1962,6 @@ do_send(Process *p, Eterm to, Eterm msg, Eterm *refp, ErtsSendContext *ctx)
             }
 
 	    switch (erts_port_command(p, ps_flags, pt, msg, refp)) {
-	    case ERTS_PORT_OP_CALLER_EXIT:
-		/* We are exiting... */
-		return SEND_USER_ERROR;
 	    case ERTS_PORT_OP_BUSY:
 		/* Nothing has been sent */
 		if (ctx->suspend)
@@ -2239,6 +2000,7 @@ do_send(Process *p, Eterm to, Eterm msg, Eterm *refp, ErtsSendContext *ctx)
 	}
 	return ret_val;
     } else if (is_tuple(to)) { /* Remote send */
+        int deref_dep = 0;
 	int ret;
 	tp = tuple_val(to);
 	if (*tp != make_arityval(2))
@@ -2246,15 +2008,13 @@ do_send(Process *p, Eterm to, Eterm msg, Eterm *refp, ErtsSendContext *ctx)
 	if (is_not_atom(tp[1]) || is_not_atom(tp[2]))
 	    return SEND_BADARG;
 	
-	/* sysname_to_connected_dist_entry will return NULL if there
-	   is no dist_entry or the dist_entry has no port,
+	/* erts_find_dist_entry will return NULL if there is no dist_entry
 	   but remote_send() will handle that. */
 
-	dep = erts_sysname_to_connected_dist_entry(tp[2]);
+	dep = erts_find_dist_entry(tp[2]);
 
 	if (dep == erts_this_dist_entry) {
 	    Eterm id;
-	    erts_deref_dist_entry(dep);
 	    if (IS_TRACED_FL(p, F_TRACE_SEND))
 		trace_send(p, to, msg);
 	    if (ERTS_PROC_GET_SAVED_CALLS_BUF(p))
@@ -2275,15 +2035,20 @@ do_send(Process *p, Eterm to, Eterm msg, Eterm *refp, ErtsSendContext *ctx)
 	    }
 	    return 0;
 	}
+        if (dep == NULL) {
+            dep = erts_find_or_insert_dist_entry(tp[2]);
+            ASSERT(dep != erts_this_dist_entry);
+            deref_dep = 1;
+        }
+	ctx->dsd.node = tp[2];
 
 	ret = remote_send(p, dep, tp[1], to, msg, ctx);
-	if (ret != SEND_YIELD_CONTINUE) {
-	    if (dep) {
-		erts_deref_dist_entry(dep);
-	    }
-	} else {
-	    ctx->dep_to_deref = dep;
+	if (ret == SEND_YIELD_CONTINUE) {
+            erts_ref_dist_entry(ctx->dep);
+            ctx->deref_dep = 1;
 	}
+        if (deref_dep)
+            erts_deref_dist_entry(dep);
 	return ret;
     } else {
 	if (IS_TRACED_FL(p, F_TRACE_SEND))
@@ -2295,22 +2060,15 @@ do_send(Process *p, Eterm to, Eterm msg, Eterm *refp, ErtsSendContext *ctx)
     
  send_message: {
 	ErtsProcLocks rp_locks = 0;
-	Sint res;
-#ifdef ERTS_SMP
 	if (p == rp)
 	    rp_locks |= ERTS_PROC_LOCK_MAIN;
-#endif
 	/* send to local process */
-	res = erts_send_message(p, rp, &rp_locks, msg, 0);
-	if (erts_use_sender_punish)
-	    res *= 4;
-	else
-	    res = 0;
-	erts_smp_proc_unlock(rp,
+	erts_send_message(p, rp, &rp_locks, msg, 0);
+	erts_proc_unlock(rp,
 			     p == rp
 			     ? (rp_locks & ~ERTS_PROC_LOCK_MAIN)
 			     : rp_locks);
-	return res;
+	return 0;
     }
 }
 
@@ -2325,7 +2083,6 @@ BIF_RETTYPE send_3(BIF_ALIST_3)
     Eterm msg = BIF_ARG_2;
     Eterm opts = BIF_ARG_3;
 
-    int connect = !0;
     Eterm l = opts;
     Sint result;
 
@@ -2336,14 +2093,15 @@ BIF_RETTYPE send_3(BIF_ALIST_3)
     UseTmpHeap(sizeof(ErtsSendContext)/sizeof(Eterm), BIF_P);
 
     ctx->suspend = !0;
-    ctx->dep_to_deref = NULL;
+    ctx->connect = !0;
+    ctx->deref_dep = 0;
     ctx->return_term = am_ok;
     ctx->dss.reds = (Sint) (ERTS_BIF_REDS_LEFT(p) * TERM_TO_BINARY_LOOP_FACTOR);
     ctx->dss.phase = ERTS_DSIG_SEND_PHASE_INIT;
 
     while (is_list(l)) {
 	if (CAR(list_val(l)) == am_noconnect) {
-	    connect = 0;
+	    ctx->connect = 0;
 	} else if (CAR(list_val(l)) == am_nosuspend) {
 	    ctx->suspend = 0;
 	} else {
@@ -2365,8 +2123,8 @@ BIF_RETTYPE send_3(BIF_ALIST_3)
     result = do_send(p, to, msg, &ref, ctx);
     ERTS_MSACC_POP_STATE_M_X();
 
-    if (result > 0) {
-	ERTS_VBUMP_REDS(p, result);
+    if (result >= 0) {
+	ERTS_VBUMP_REDS(p, 4);
 	if (ERTS_IS_PROC_OUT_OF_REDS(p))
 	    goto yield_return;
 	ERTS_BIF_PREP_RET(retval, am_ok);
@@ -2374,15 +2132,9 @@ BIF_RETTYPE send_3(BIF_ALIST_3)
     }
 
     switch (result) {
-    case 0:
-	/* May need to yield even though we do not bump reds here... */
-	if (ERTS_IS_PROC_OUT_OF_REDS(p))
-	    goto yield_return;
-	ERTS_BIF_PREP_RET(retval, am_ok);
-	break;
-    case SEND_TRAP:
-	if (connect) {
-	    ERTS_BIF_PREP_TRAP3(retval, dsend3_trap, p, to, msg, opts);
+    case SEND_NOCONNECT:
+	if (ctx->connect) {
+	    ERTS_BIF_PREP_RET(retval, am_ok);
 	} else {
 	    ERTS_BIF_PREP_RET(retval, am_noconnect);
 	}
@@ -2486,7 +2238,8 @@ Eterm erl_send(Process *p, Eterm to, Eterm msg)
     ref = NIL;
 #endif
     ctx->suspend = !0;
-    ctx->dep_to_deref = NULL;
+    ctx->connect = !0;
+    ctx->deref_dep = 0;
     ctx->return_term = msg;
     ctx->dss.reds = (Sint) (ERTS_BIF_REDS_LEFT(p) * TERM_TO_BINARY_LOOP_FACTOR);
     ctx->dss.phase = ERTS_DSIG_SEND_PHASE_INIT;
@@ -2495,8 +2248,8 @@ Eterm erl_send(Process *p, Eterm to, Eterm msg)
 
     ERTS_MSACC_POP_STATE_M_X();
 
-    if (result > 0) {
-	ERTS_VBUMP_REDS(p, result);
+    if (result >= 0) {
+	ERTS_VBUMP_REDS(p, 4);
 	if (ERTS_IS_PROC_OUT_OF_REDS(p))
 	    goto yield_return;
 	ERTS_BIF_PREP_RET(retval, msg);
@@ -2504,15 +2257,9 @@ Eterm erl_send(Process *p, Eterm to, Eterm msg)
     }
 
     switch (result) {
-    case 0:
-	/* May need to yield even though we do not bump reds here... */
-	if (ERTS_IS_PROC_OUT_OF_REDS(p))
-	    goto yield_return;
+    case SEND_NOCONNECT:
 	ERTS_BIF_PREP_RET(retval, msg);
 	break;
-    case SEND_TRAP:
-	ERTS_BIF_PREP_TRAP2(retval, dsend2_trap, p, to, msg);
-	break;
     case SEND_YIELD:
 	ERTS_BIF_PREP_YIELD2(retval, bif_export[BIF_send_2], p, to, msg);
 	break;
@@ -3171,7 +2918,7 @@ BIF_RETTYPE list_to_integer_2(BIF_ALIST_2)
 static int do_float_to_charbuf(Process *p, Eterm efloat, Eterm list, 
 			char *fbuf, int sizeof_fbuf) {
 
-    const static int arity_two = make_arityval(2);
+    Eterm arity_two = make_arityval(2);
     int decimals = SYS_DEFAULT_FLOAT_DECIMALS;
     int compact = 0;
     enum fmt_type_ {
@@ -3535,7 +3282,7 @@ BIF_RETTYPE binary_to_float_1(BIF_ALIST_1)
     if (bit_offs)
       erts_copy_bits(bytes, bit_offs, 1, buf, 0, 1, size*8);
     else
-      memcpy(buf, bytes, size);
+      sys_memcpy(buf, bytes, size);
     
     buf[size] = '\0';
     
@@ -4010,14 +3757,14 @@ BIF_RETTYPE halt_2(BIF_ALIST_2)
 	    ERTS_BIF_YIELD2(bif_export[BIF_halt_2], BIF_P, am_undefined, am_undefined);
 	}
 	else {
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
             erts_exit(pos_int_code, "");
 	}
     }
     else if (ERTS_IS_ATOM_STR("abort", BIF_ARG_1)) {
 	VERBOSE(DEBUG_SYSTEM,
 		("System halted by BIF halt(%T, %T)\n", BIF_ARG_1, BIF_ARG_2));
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	erts_exit(ERTS_ABORT_EXIT, "");
     }
     else if (is_list(BIF_ARG_1) || BIF_ARG_1 == NIL) {
@@ -4033,7 +3780,7 @@ BIF_RETTYPE halt_2(BIF_ALIST_2)
 	halt_msg[written] = '\0';
 	VERBOSE(DEBUG_SYSTEM,
 		("System halted by BIF halt(%T, %T)\n", BIF_ARG_1, BIF_ARG_2));
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	erts_exit(ERTS_DUMP_EXIT, "%s\n", halt_msg);
     }
     else
@@ -4219,7 +3966,6 @@ BIF_RETTYPE list_to_pid_1(BIF_ALIST_1)
 	goto bad;
 
     if(dep == erts_this_dist_entry) {
-	erts_deref_dist_entry(dep);
 	BIF_RET(make_internal_pid(make_pid_data(c, b)));
     }
     else {
@@ -4239,13 +3985,10 @@ BIF_RETTYPE list_to_pid_1(BIF_ALIST_1)
       etp->data.ui[0] = make_pid_data(c, b);
 
       MSO(BIF_P).first = (struct erl_off_heap_header*) etp;
-      erts_deref_dist_entry(dep);
       BIF_RET(make_external_pid(etp));
     }
 
  bad:
-    if (dep)
-	erts_deref_dist_entry(dep);
     if (buf)
 	erts_free(ERTS_ALC_T_TMP, (void *) buf);
     BIF_ERROR(BIF_P, BADARG);
@@ -4273,12 +4016,12 @@ BIF_RETTYPE list_to_port_1(BIF_ALIST_1)
     buf[i] = '\0';		/* null terminal */
 
     cp = &buf[0];
-    if (strncmp("#Port<", cp, 6) != 0)
+    if (sys_strncmp("#Port<", cp, 6) != 0)
         goto bad;
 
-    cp += 6; /* strlen("#Port<") */
+    cp += 6; /* sys_strlen("#Port<") */
 
-    if (sscanf(cp, "%u.%u>", &n, &p) < 2)
+    if (sscanf(cp, "%u.%u>", (unsigned int*)&n, (unsigned int*)&p) < 2)
         goto bad;
 
     if (p > ERTS_MAX_PORT_NUMBER)
@@ -4290,7 +4033,6 @@ BIF_RETTYPE list_to_port_1(BIF_ALIST_1)
 	goto bad;
 
     if(dep == erts_this_dist_entry) {
-	erts_deref_dist_entry(dep);
 	BIF_RET(make_internal_port(p));
     }
     else {
@@ -4310,13 +4052,10 @@ BIF_RETTYPE list_to_port_1(BIF_ALIST_1)
       etp->data.ui[0] = p;
 
       MSO(BIF_P).first = (struct erl_off_heap_header*) etp;
-      erts_deref_dist_entry(dep);
       BIF_RET(make_external_port(etp));
     }
 
  bad:
-    if (dep)
-	erts_deref_dist_entry(dep);
     BIF_ERROR(BIF_P, BADARG);
 }
 
@@ -4436,12 +4175,9 @@ BIF_RETTYPE list_to_ref_1(BIF_ALIST_1)
       res = make_external_ref(etp);
     }
 
-    erts_deref_dist_entry(dep);
     BIF_RET(res);
 
  bad:
-    if (dep)
-	erts_deref_dist_entry(dep);
     BIF_ERROR(BIF_P, BADARG);
 }
 
@@ -4454,14 +4190,89 @@ BIF_RETTYPE group_leader_0(BIF_ALIST_0)
 }
 
 /**********************************************************************/
-/* arg1 == leader, arg2 == new member */
+/* set group leader */
 
-BIF_RETTYPE group_leader_2(BIF_ALIST_2)
+int
+erts_set_group_leader(Process *proc, Eterm new_gl)
 {
-    Process* new_member;
+    
+    erts_aint32_t state;
 
-    if (is_not_pid(BIF_ARG_1)) {
-	BIF_ERROR(BIF_P, BADARG);
+    ASSERT(is_pid(new_gl));
+
+    state = erts_atomic32_read_nob(&proc->state);
+
+    if (state & ERTS_PSFLG_EXITING)
+        return 0;
+        
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(proc));
+
+    if (!(state & ERTS_PSFLG_DIRTY_RUNNING))
+        proc->group_leader = STORE_NC_IN_PROC(proc, new_gl);
+    else {
+        ErlHeapFragment *bp;
+        Eterm *hp;
+        /*
+         * Currently executing on a dirty scheduler,
+         * so we are not allowed to write to its heap.
+         * Store group leader pid in heap fragment.
+         */
+        bp = new_message_buffer(NC_HEAP_SIZE(new_gl));
+        hp = bp->mem;
+        proc->group_leader = STORE_NC(&hp,
+                                            &proc->off_heap,
+                                            new_gl);
+        bp->next = proc->mbuf;
+        proc->mbuf = bp;
+        proc->mbuf_sz += bp->used_size;
+    }
+
+    return !0;
+}
+
+BIF_RETTYPE erts_internal_group_leader_3(BIF_ALIST_3)
+{
+    if (is_not_pid(BIF_ARG_1))
+        BIF_ERROR(BIF_P, BADARG);
+    if (is_not_internal_pid(BIF_ARG_2))
+        BIF_ERROR(BIF_P, BADARG);
+    if (is_not_internal_ref(BIF_ARG_3))
+        BIF_ERROR(BIF_P, BADARG);
+
+    erts_proc_sig_send_group_leader(BIF_P,
+                                    BIF_ARG_2,
+                                    BIF_ARG_1,
+                                    BIF_ARG_3);
+    BIF_RET(am_ok);
+}
+
+BIF_RETTYPE erts_internal_group_leader_2(BIF_ALIST_2)
+{
+    if (is_not_pid(BIF_ARG_1))
+        BIF_RET(am_badarg);
+
+    if (is_internal_pid(BIF_ARG_2)) {
+        Process *rp;
+        int res;
+
+        if (BIF_ARG_2 == BIF_P->common.id)
+            rp = BIF_P;
+        else {
+            rp = erts_try_lock_sig_free_proc(BIF_ARG_2,
+                                             ERTS_PROC_LOCK_MAIN,
+                                             NULL);
+            if (!rp)
+                BIF_RET(am_badarg);
+            if (rp == ERTS_PROC_LOCK_BUSY)
+                BIF_RET(am_false);
+        }
+
+        res = erts_set_group_leader(rp, BIF_ARG_1);
+
+        if (rp != BIF_P)
+            erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+
+        BIF_RET(res ? am_true : am_badarg);
     }
 
     if (is_external_pid(BIF_ARG_2)) {
@@ -4469,95 +4280,30 @@ BIF_RETTYPE group_leader_2(BIF_ALIST_2)
 	int code;
 	ErtsDSigData dsd;
 	dep = external_pid_dist_entry(BIF_ARG_2);
+	ERTS_ASSERT(dep);
 	if(dep == erts_this_dist_entry)
 	    BIF_ERROR(BIF_P, BADARG);
 
-	code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_DSP_NO_LOCK, 0);
+	code = erts_dsig_prepare(&dsd, dep, BIF_P, ERTS_PROC_LOCK_MAIN,
+				 ERTS_DSP_NO_LOCK, 0, 1);
 	switch (code) {
 	case ERTS_DSIG_PREP_NOT_ALIVE:
-	    BIF_RET(am_true);
 	case ERTS_DSIG_PREP_NOT_CONNECTED:
-	    BIF_TRAP2(dgroup_leader_trap, BIF_P, BIF_ARG_1, BIF_ARG_2);
+	    BIF_RET(am_true);
+	case ERTS_DSIG_PREP_PENDING:
 	case ERTS_DSIG_PREP_CONNECTED:
 	    code = erts_dsig_send_group_leader(&dsd, BIF_ARG_1, BIF_ARG_2);
 	    if (code == ERTS_DSIG_SEND_YIELD)
 		ERTS_BIF_YIELD_RETURN(BIF_P, am_true);
 	    BIF_RET(am_true);
 	default:
-	    ASSERT(! "Invalid dsig prepare result");
-	    BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
+	    ERTS_ASSERT(! "Invalid dsig prepare result");
 	}
     }
-    else if (is_internal_pid(BIF_ARG_2)) {
-	int await_x;
-	ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS;
-	new_member = erts_pid2proc_nropt(BIF_P, ERTS_PROC_LOCK_MAIN,
-					 BIF_ARG_2, locks);
-	if (!new_member)
-	    BIF_ERROR(BIF_P, BADARG);
-
-	if (new_member == ERTS_PROC_LOCK_BUSY)
-	    ERTS_BIF_YIELD2(bif_export[BIF_group_leader_2], BIF_P,
-			    BIF_ARG_1, BIF_ARG_2);
-
-	await_x = (new_member != BIF_P
-		   && ERTS_PROC_PENDING_EXIT(new_member));
-	if (!await_x) {
-	    if (is_immed(BIF_ARG_1))
-		new_member->group_leader = BIF_ARG_1;
-	    else {
-		locks &= ~ERTS_PROC_LOCK_STATUS;
-		erts_smp_proc_unlock(new_member, ERTS_PROC_LOCK_STATUS);
-		if (new_member == BIF_P
-		    || !(erts_smp_atomic32_read_nob(&new_member->state)
-			 & ERTS_PSFLG_DIRTY_RUNNING)) {
-		    new_member->group_leader = STORE_NC_IN_PROC(new_member,
-								BIF_ARG_1);
-		}
-		else {
-		    ErlHeapFragment *bp;
-		    Eterm *hp;
-		    /*
-		     * Other process executing on a dirty scheduler,
-		     * so we are not allowed to write to its heap.
-		     * Store in heap fragment.
-		     */
-
-		    bp = new_message_buffer(NC_HEAP_SIZE(BIF_ARG_1));
-		    hp = bp->mem;
-		    new_member->group_leader = STORE_NC(&hp,
-							&new_member->off_heap,
-							BIF_ARG_1);
-		    bp->next = new_member->mbuf;
-		    new_member->mbuf = bp;
-		    new_member->mbuf_sz += bp->used_size;
-		}
-	    }
-	}
-
-	if (new_member == BIF_P)
-	    locks &= ~ERTS_PROC_LOCK_MAIN;
-	if (locks)
-	    erts_smp_proc_unlock(new_member, locks);
-
-	if (await_x) {
-	    /* Wait for new_member to terminate; then badarg */
-	    Eterm args[2] = {BIF_ARG_1, BIF_ARG_2};
-	    ERTS_BIF_AWAIT_X_APPLY_TRAP(BIF_P,
-					BIF_ARG_2,
-					am_erlang,
-					am_group_leader,
-					args,
-					2);
-	}
 
-	BIF_RET(am_true);
-    }
-    else {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    BIF_RET(am_badarg);
 }
-    
+
 BIF_RETTYPE system_flag_2(BIF_ALIST_2)    
 {
     Sint n;
@@ -4565,9 +4311,6 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
     if (BIF_ARG_1 == am_multi_scheduling) {
 	if (BIF_ARG_2 == am_block || BIF_ARG_2 == am_unblock
 	    || BIF_ARG_2 == am_block_normal || BIF_ARG_2 == am_unblock_normal) {
-#ifndef ERTS_SMP
-	    BIF_RET(am_disabled);
-#else
 	    int block = (BIF_ARG_2 == am_block
 			 || BIF_ARG_2 == am_block_normal);
 	    int normal = (BIF_ARG_2 == am_block_normal
@@ -4603,15 +4346,8 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
                 BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
                 break;
             }
-#endif
 	}
     } else if (BIF_ARG_1 == am_schedulers_online) {
-#ifndef ERTS_SMP
-	if (BIF_ARG_2 != make_small(1))
-	    goto error;
-	else
-	    BIF_RET(make_small(1));
-#else
 	Sint old_no;
 	if (!is_small(BIF_ARG_2))
 	    goto error;
@@ -4635,7 +4371,6 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
 	    break;
 	}
-#endif
     } else if (BIF_ARG_1 == am_fullsweep_after) {
 	Uint16 nval;
 	Uint oval;
@@ -4643,7 +4378,7 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    goto error;
 	}
 	nval = (n > (Sint) ((Uint16) -1)) ? ((Uint16) -1) : ((Uint16) n);
-	oval = (Uint) erts_smp_atomic32_xchg_nob(&erts_max_gen_gcs,
+	oval = (Uint) erts_atomic32_xchg_nob(&erts_max_gen_gcs,
 						 (erts_aint32_t) nval);
 	BIF_RET(make_small(oval));
     } else if (BIF_ARG_1 == am_min_heap_size) {
@@ -4653,13 +4388,13 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    goto error;
 	}
 
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	H_MIN_SIZE = erts_next_heap_size(n, 0);
 
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
 	BIF_RET(make_small(oval));
     } else if (BIF_ARG_1 == am_min_bin_vheap_size) {
@@ -4669,13 +4404,13 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    goto error;
 	}
 
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	BIN_VH_MIN_SIZE = erts_next_heap_size(n, 0);
 
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
 	BIF_RET(make_small(oval));
     } else if (BIF_ARG_1 == am_max_heap_size) {
@@ -4693,14 +4428,14 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
         hp = HAlloc(BIF_P, sz);
         old_value = erts_max_heap_size_map(H_MAX_SIZE, H_MAX_FLAGS, &hp, NULL);
 
-        erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-        erts_smp_thr_progress_block();
+        erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+        erts_thr_progress_block();
 
         H_MAX_SIZE = max_heap_size;
         H_MAX_FLAGS = max_heap_flags;
 
-        erts_smp_thr_progress_unblock();
-        erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+        erts_thr_progress_unblock();
+        erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
         BIF_RET(old_value);
     } else if (BIF_ARG_1 == am_display_items) {
@@ -4753,9 +4488,8 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
         BIF_RET(ret);
     } else if (BIF_ARG_1 == make_small(1)) {
 	int i, max;
-	ErtsMessage* mp;
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	max = erts_ptab_max(&erts_proc);
 	for (i = 0; i < max; i++) {
@@ -4768,36 +4502,19 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 #endif
 		p->seq_trace_clock = 0;
 		p->seq_trace_lastcnt = 0;
-		ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p);
-		mp = p->msg.first;
-		while(mp != NULL) {
-#ifdef USE_VM_PROBES
-		    ERL_MESSAGE_TOKEN(mp) = (ERL_MESSAGE_DT_UTAG(mp) != NIL) ? am_have_dt_utag : NIL;
-#else
-		    ERL_MESSAGE_TOKEN(mp) = NIL;
-#endif
-		    mp = mp->next;
-		}
+
+                erts_proc_sig_clear_seq_trace_tokens(p);
 	    }
 	}
 
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
 	BIF_RET(am_true);
     } else if (BIF_ARG_1 == am_scheduler_wall_time) {
-	if (BIF_ARG_2 == am_true || BIF_ARG_2 == am_false) {
-	    erts_aint32_t new = BIF_ARG_2 == am_true ? 1 : 0;
-	    erts_aint32_t old = erts_smp_atomic32_xchg_nob(&sched_wall_time,
-							   new);
-	    Eterm ref = erts_sched_wall_time_request(BIF_P, 1, new, 0, 0);
-	    ASSERT(is_value(ref));
-	    BIF_TRAP2(await_sched_wall_time_mod_trap,
-		      BIF_P,
-		      ref,
-		      old ? am_true : am_false);
-	}
-#if defined(ERTS_SMP) && defined(ERTS_DIRTY_SCHEDULERS)
+	if (BIF_ARG_2 == am_true || BIF_ARG_2 == am_false)
+            BIF_TRAP1(system_flag_scheduler_wall_time_trap,
+                      BIF_P, BIF_ARG_2);
     } else if (BIF_ARG_1 == am_dirty_cpu_schedulers_online) {
 	Sint old_no;
 	if (!is_small(BIF_ARG_2))
@@ -4823,13 +4540,12 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
 	    break;
 	}
-#endif
     } else if (BIF_ARG_1 == am_time_offset
 	       && ERTS_IS_ATOM_STR("finalize", BIF_ARG_2)) {
 	ErtsTimeOffsetState res;
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	res = erts_finalize_time_offset();
-        erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+        erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	switch (res) {
 	case ERTS_TIME_OFFSET_PRELIMINARY: {
 	    DECL_AM(preliminary);
@@ -4851,7 +4567,7 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
       Eterm threads;
       if (BIF_ARG_2 == am_true || BIF_ARG_2 == am_false) {
         erts_aint32_t new = BIF_ARG_2 == am_true ? ERTS_MSACC_ENABLE : ERTS_MSACC_DISABLE;
-	erts_aint32_t old = erts_smp_atomic32_xchg_nob(&msacc, new);
+	erts_aint32_t old = erts_atomic32_xchg_nob(&msacc, new);
 	Eterm ref = erts_msacc_request(BIF_P, new, &threads);
         if (is_non_value(ref))
             BIF_RET(old ? am_true : am_false);
@@ -4862,7 +4578,7 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 		  threads);
       } else if (BIF_ARG_2 == am_reset) {
 	Eterm ref = erts_msacc_request(BIF_P, ERTS_MSACC_RESET, &threads);
-	erts_aint32_t old = erts_smp_atomic32_read_nob(&msacc);
+	erts_aint32_t old = erts_atomic32_read_nob(&msacc);
 	ASSERT(is_value(ref));
 	BIF_TRAP3(await_msacc_mod_trap,
 		  BIF_P,
@@ -4881,9 +4597,9 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
 	    what = ERTS_SCHED_STAT_MODIFY_CLEAR;
 	else
 	    goto error;
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	erts_sched_stat_modify(what);
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	BIF_RET(am_true);
     } else if (ERTS_IS_ATOM_STR("internal_cpu_topology", BIF_ARG_1)) {
 	Eterm res = erts_set_cpu_topology(BIF_P, BIF_ARG_2);
@@ -4912,6 +4628,17 @@ BIF_RETTYPE system_flag_2(BIF_ALIST_2)
     BIF_ERROR(BIF_P, BADARG);
 }
 
+BIF_RETTYPE erts_internal_scheduler_wall_time_1(BIF_ALIST_1)
+{
+    erts_aint32_t new = BIF_ARG_1 == am_true ? 1 : 0;
+    erts_aint32_t old = erts_atomic32_xchg_nob(&sched_wall_time,
+                                               new);
+    Eterm ref = erts_sched_wall_time_request(BIF_P, 1, new, 0, 0);
+    ASSERT(is_value(ref));
+    BIF_TRAP2(await_sched_wall_time_mod_trap,
+              BIF_P, ref, old ? am_true : am_false);
+}
+
 /**********************************************************************/
 
 BIF_RETTYPE phash_2(BIF_ALIST_2)
@@ -5023,7 +4750,6 @@ static BIF_RETTYPE bif_return_trap(BIF_ALIST_2)
     Eterm res = BIF_ARG_1;
 
     switch (BIF_ARG_2) {
-#ifdef ERTS_SMP
     case am_multi_scheduling: {
 	int msb = erts_is_multi_scheduling_blocked();
 	if (msb > 0)
@@ -5034,92 +4760,12 @@ static BIF_RETTYPE bif_return_trap(BIF_ALIST_2)
 	    ERTS_INTERNAL_ERROR("Unexpected multi scheduling block state");
 	break;
     }
-#endif
     default:
 	break;
     }
     BIF_RET(res);
 }
 
-/*
- * NOTE: The erts_bif_prep_await_proc_exit_*() functions are
- * tightly coupled with the implementation of erlang:await_proc_exit/3.
- * The erts_bif_prep_await_proc_exit_*() functions can safely call
- * skip_current_msgq() since they know that erlang:await_proc_exit/3
- * unconditionally will do a monitor and then unconditionally will
- * wait for the corresponding 'DOWN' message in a receive, and no other
- * receive is done before this receive. This optimization removes an
- * unnecessary scan of the currently existing message queue (which
- * can be large). If the erlang:await_proc_exit/3 implementation
- * is changed so that the above isn't true, nasty bugs in later
- * receives, etc, may appear.
- */
-
-static ERTS_INLINE int
-skip_current_msgq(Process *c_p)
-{
-    int res;
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
-    erts_proc_lc_chk_only_proc_main(c_p);
-#endif
-
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-    if (ERTS_PROC_PENDING_EXIT(c_p)) {
-	KILL_CATCHES(c_p);
-	c_p->freason = EXC_EXIT;
-	res = 0;
-    }
-    else {
-        ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	c_p->msg.save = c_p->msg.last;
-	res = 1;
-    }
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-    return res;
-}
-
-void
-erts_bif_prep_await_proc_exit_data_trap(Process *c_p, Eterm pid, Eterm ret)
-{
-    if (skip_current_msgq(c_p)) {
-	ERTS_BIF_PREP_TRAP3_NO_RET(await_proc_exit_trap, c_p, pid, am_data, ret);
-    }
-}
-
-void
-erts_bif_prep_await_proc_exit_reason_trap(Process *c_p, Eterm pid)
-{
-    if (skip_current_msgq(c_p)) {
-	ERTS_BIF_PREP_TRAP3_NO_RET(await_proc_exit_trap, c_p,
-			    pid, am_reason, am_undefined);
-    }
-}
-
-void
-erts_bif_prep_await_proc_exit_apply_trap(Process *c_p,
-					 Eterm pid,
-					 Eterm module,
-					 Eterm function,
-					 Eterm args[],
-					 int nargs)
-{
-    ASSERT(is_atom(module) && is_atom(function));
-    if (skip_current_msgq(c_p)) {
-	Eterm term;
-	Eterm *hp;
-	int i;
-
-	hp = HAlloc(c_p, 4+2*nargs);
-	term = NIL;
-	for (i = nargs-1; i >= 0; i--) {
-	    term = CONS(hp, args[i], term);
-	    hp += 2;
-	}
-	term = TUPLE3(hp, module, function, term);
-	ERTS_BIF_PREP_TRAP3_NO_RET(await_proc_exit_trap, c_p, pid, am_apply, term);
-    }
-}
-
 Export bif_return_trap_export;
 
 void erts_init_trap_export(Export* ep, Eterm m, Eterm f, Uint a,
@@ -5133,7 +4779,7 @@ void erts_init_trap_export(Export* ep, Eterm m, Eterm f, Uint a,
     ep->info.mfa.module = m;
     ep->info.mfa.function = f;
     ep->info.mfa.arity = a;
-    ep->beam[0] = (BeamInstr) em_apply_bif;
+    ep->beam[0] = BeamOpCodeAddr(op_apply_bif);
     ep->beam[1] = (BeamInstr) bif;
 }
 
@@ -5155,6 +4801,9 @@ void erts_init_bif(void)
 			  am_erts_internal, am_dsend_continue_trap, 1,
 			  dsend_continue_trap_1);
 
+    erts_init_trap_export(&await_exit_trap, am_erts_internal,
+                          am_await_exit, 0, erts_internal_await_exit_trap);
+
     flush_monitor_messages_trap = erts_export_put(am_erts_internal,
 						  am_flush_monitor_messages,
 						  3);
@@ -5169,16 +4818,17 @@ void erts_init_bif(void)
     erts_format_cpu_topology_trap = erts_export_put(am_erlang,
 						    am_format_cpu_topology,
 						    1);
-    await_proc_exit_trap = erts_export_put(am_erlang,am_await_proc_exit,3);
     await_port_send_result_trap
 	= erts_export_put(am_erts_internal, am_await_port_send_result, 3);
+    system_flag_scheduler_wall_time_trap
+        = erts_export_put(am_erts_internal, am_system_flag_scheduler_wall_time, 1);
     await_sched_wall_time_mod_trap
-        = erts_export_put(am_erlang, am_await_sched_wall_time_modifications, 2);
+        = erts_export_put(am_erts_internal, am_await_sched_wall_time_modifications, 2);
     await_msacc_mod_trap
 	= erts_export_put(am_erts_internal, am_await_microstate_accounting_modifications, 3);
 
-    erts_smp_atomic32_init_nob(&sched_wall_time, 0);
-    erts_smp_atomic32_init_nob(&msacc, ERTS_MSACC_IS_ENABLED());
+    erts_atomic32_init_nob(&sched_wall_time, 0);
+    erts_atomic32_init_nob(&msacc, ERTS_MSACC_IS_ENABLED());
 }
 
 /*
@@ -5196,15 +4846,14 @@ schedule(Process *c_p, Process *dirty_shadow_proc,
 	 Eterm module, Eterm function,
 	 int argc, Eterm *argv)
 {
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
     (void) erts_nif_export_schedule(c_p, dirty_shadow_proc,
-				    mfa, pc, (BeamInstr) em_apply_bif,
+				    mfa, pc, BeamOpCodeAddr(op_apply_bif),
 				    dfunc, ifunc,
 				    module, function,
 				    argc, argv);
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static BIF_RETTYPE dirty_bif_result(BIF_ALIST_1)
 {
@@ -5247,9 +4896,7 @@ static BIF_RETTYPE dirty_bif_exception(BIF_ALIST_2)
     BIF_ERROR(BIF_P, freason);
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
-extern BeamInstr* em_call_bif_e;
 static BIF_RETTYPE call_bif(Process *c_p, Eterm *reg, BeamInstr *I);
 
 BIF_RETTYPE
@@ -5265,15 +4912,13 @@ erts_schedule_bif(Process *proc,
     Process *c_p, *dirty_shadow_proc;
     ErtsCodeMFA *mfa;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (proc->static_flags & ERTS_STC_FLG_SHADOW_PROC) {
 	dirty_shadow_proc = proc;
 	c_p = proc->next;
 	ASSERT(c_p->common.id == dirty_shadow_proc->common.id);
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
     }
     else
-#endif
     {
 	dirty_shadow_proc = NULL;
 	c_p = proc;
@@ -5289,7 +4934,6 @@ erts_schedule_bif(Process *proc,
 	 * ibif - indirect bif
 	 */
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 	erts_aint32_t set, mask;
 	mask = (ERTS_PSFLG_DIRTY_CPU_PROC
 		| ERTS_PSFLG_DIRTY_IO_PROC);
@@ -5312,11 +4956,7 @@ erts_schedule_bif(Process *proc,
 	    break;
 	}
 
-	(void) erts_smp_atomic32_read_bset_nob(&c_p->state, mask, set);
-#else
-	dbif = call_bif;
-	ibif = bif;
-#endif
+	(void) erts_atomic32_read_bset_nob(&c_p->state, mask, set);
 
 	if (i == NULL) {
 	    ERTS_INTERNAL_ERROR("Missing instruction pointer");
@@ -5329,13 +4969,13 @@ erts_schedule_bif(Process *proc,
 	    mfa = &exp->info.mfa;
 	}
 #endif
-	else if (em_call_bif_e == (BeamInstr *) *i) {
+	else if (BeamIsOpCode(*i, op_call_bif_e)) {
 	    /* Pointer to bif export in i+1 */
 	    exp = (Export *) i[1];
 	    pc = i;
 	    mfa = &exp->info.mfa;
 	}
-	else if (em_apply_bif == (BeamInstr *) *i) {
+	else if (BeamIsOpCode(*i, op_apply_bif)) {
 	    /* Pointer to bif in i+1, and mfa in i-3 */	    
 	    pc = c_p->cp;
 	    mfa = erts_code_to_codemfa(i);
@@ -5357,7 +4997,7 @@ erts_schedule_bif(Process *proc,
     }
 
     if (dirty_shadow_proc)
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 
     return THE_NON_VALUE;
 }
@@ -5392,7 +5032,6 @@ call_bif(Process *c_p, Eterm *reg, BeamInstr *I)
     return ret;
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 int
 erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *reg)
@@ -5407,7 +5046,7 @@ erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
     erts_aint32_t state;
 
     ASSERT(!c_p->scheduler_data);
-    state = erts_smp_atomic32_read_nob(&c_p->state);
+    state = erts_atomic32_read_nob(&c_p->state);
     ASSERT((state & ERTS_PSFLG_DIRTY_RUNNING)
 	   && !(state & (ERTS_PSFLG_RUNNING|ERTS_PSFLG_RUNNING_SYS)));
     ASSERT(esdp);
@@ -5421,7 +5060,7 @@ erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
 
     bf = (ErtsBifFunc) I[1];
 
-    erts_smp_atomic32_read_band_mb(&c_p->state, ~(ERTS_PSFLG_DIRTY_CPU_PROC
+    erts_atomic32_read_band_mb(&c_p->state, ~(ERTS_PSFLG_DIRTY_CPU_PROC
 						  | ERTS_PSFLG_DIRTY_IO_PROC));
 
     dirty_shadow_proc = erts_make_dirty_shadow_proc(esdp, c_p);
@@ -5436,11 +5075,11 @@ erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
     c_p_htop = c_p->htop;
 #endif
 
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 
     result = (*bf)(dirty_shadow_proc, reg, I);
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 
     ASSERT(c_p_htop == c_p->htop);
     ASSERT(dirty_shadow_proc->static_flags & ERTS_STC_FLG_SHADOW_PROC);
@@ -5463,7 +5102,7 @@ erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
 	}
 	else if (nep->func == ERTS_SCHED_BIF_TRAP_MARKER) {
 	    /* Dirty BIF did an ordinary trap... */
-	    ASSERT(!(erts_smp_atomic32_read_nob(&c_p->state)
+	    ASSERT(!(erts_atomic32_read_nob(&c_p->state)
 		     & (ERTS_PSFLG_DIRTY_CPU_PROC|ERTS_PSFLG_DIRTY_IO_PROC)));
 	    schedule(c_p, dirty_shadow_proc, NULL, NULL,
 		     dirty_bif_trap, (void *) dirty_shadow_proc->i,
@@ -5486,7 +5125,6 @@ erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
     return exiting;
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 
 #ifdef HARDDEBUG
diff --git a/erts/emulator/beam/bif.h b/erts/emulator/beam/bif.h
index 01cca90a7a..cf9f61c0b8 100644
--- a/erts/emulator/beam/bif.h
+++ b/erts/emulator/beam/bif.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2016. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -93,7 +93,7 @@ do {									\
 
 #define BUMP_REDS(p, gc) do {			   \
      ASSERT(p);		 			   \
-     ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(p));\
+     ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(p));\
      (p)->fcalls -= (gc); 			   \
      if ((p)->fcalls < 0) { 			   \
 	if (!ERTS_PROC_GET_SAVED_CALLS_BUF((p)))   \
@@ -295,6 +295,19 @@ do {								\
     (Ret) = THE_NON_VALUE;					\
 } while (0)
 
+#define ERTS_BIF_PREP_TRAP4(Ret, Trap, Proc, A0, A1, A2, A3)	\
+do {								\
+    Eterm* reg = erts_proc_sched_data((Proc))->x_reg_array;	\
+    (Proc)->arity = 4;						\
+    reg[0] = (Eterm) (A0);					\
+    reg[1] = (Eterm) (A1);					\
+    reg[2] = (Eterm) (A2);					\
+    reg[3] = (Eterm) (A3);					\
+    (Proc)->i = (BeamInstr*) ((Trap)->addressv[erts_active_code_ix()]); \
+    (Proc)->freason = TRAP;					\
+    (Ret) = THE_NON_VALUE;					\
+} while (0)
+
 #define ERTS_BIF_PREP_TRAP3_NO_RET(Trap, Proc, A0, A1, A2)\
 do {							\
     Eterm* reg = erts_proc_sched_data((Proc))->x_reg_array;	\
@@ -306,7 +319,7 @@ do {							\
     (Proc)->freason = TRAP;				\
 } while (0)
 
-#define BIF_TRAP0(p, Trap_) do {		\
+#define BIF_TRAP0(Trap_, p) do {                  \
       (p)->arity = 0;				\
       (p)->i = (BeamInstr*) ((Trap_)->addressv[erts_active_code_ix()]);	\
       (p)->freason = TRAP;			\
@@ -343,6 +356,18 @@ do {							\
       return THE_NON_VALUE;					\
  } while(0)
 
+#define BIF_TRAP4(Trap_, p, A0, A1, A2, A3) do {		\
+      Eterm* reg = erts_proc_sched_data((p))->x_reg_array;	\
+      (p)->arity = 4;						\
+      reg[0] = (A0);						\
+      reg[1] = (A1);						\
+      reg[2] = (A2);						\
+      reg[3] = (A3);						\
+      (p)->i = (BeamInstr*) ((Trap_)->addressv[erts_active_code_ix()]); \
+      (p)->freason = TRAP;					\
+      return THE_NON_VALUE;					\
+ } while(0)
+
 #define BIF_TRAP_CODE_PTR_0(p, Code_) do {	\
       (p)->arity = 0;				\
       (p)->i = (BeamInstr*) (Code_);		\
@@ -401,10 +426,16 @@ do {									\
     ERTS_BIF_PREP_TRAP3(RET, (TRP), (P), (A0), (A1), (A2));		\
 } while (0)
 
+#define ERTS_BIF_PREP_YIELD4(RET, TRP, P, A0, A1, A2, A3)		\
+do {									\
+    ERTS_VBUMP_ALL_REDS((P));						\
+    ERTS_BIF_PREP_TRAP4(RET, (TRP), (P), (A0), (A1), (A2), (A3));       \
+} while (0)
+
 #define ERTS_BIF_YIELD0(TRP, P)						\
 do {									\
     ERTS_VBUMP_ALL_REDS((P));						\
-    BIF_TRAP0((TRP), (P));						\
+    BIF_TRAP0((TRP), (P));                                              \
 } while (0)
 
 #define ERTS_BIF_YIELD1(TRP, P, A0)					\
@@ -425,10 +456,22 @@ do {									\
     BIF_TRAP3((TRP), (P), (A0), (A1), (A2));				\
 } while (0)
 
+#define ERTS_BIF_YIELD4(TRP, P, A0, A1, A2, A3)				\
+do {									\
+    ERTS_VBUMP_ALL_REDS((P));						\
+    BIF_TRAP4((TRP), (P), (A0), (A1), (A2), (A3));                      \
+} while (0)
+
+#define ERTS_BIF_PREP_EXITED(RET, PROC)	                                \
+do {                                                                    \
+    KILL_CATCHES((PROC));                                               \
+    ERTS_BIF_PREP_ERROR((RET), (PROC), EXTAG_EXIT);                     \
+} while (0)
+
 #define ERTS_BIF_EXITED(PROC)		\
 do {					\
     KILL_CATCHES((PROC));		\
-    BIF_ERROR((PROC), EXC_EXIT);	\
+    BIF_ERROR((PROC), EXTAG_EXIT);	\
 } while (0)
 
 #define ERTS_BIF_CHK_EXITED(PROC)	\
@@ -437,71 +480,8 @@ do {					\
 	ERTS_BIF_EXITED((PROC));	\
 } while (0)
 
-/*
- * The ERTS_BIF_*_AWAIT_X_*_TRAP makros either exits the caller, or
- * sets up a trap to erlang:await_proc_exit/3.
- *
- * The caller is acquired to hold the 'main' lock on C_P. No other locks
- * are allowed to be held.
- */
-
-#define ERTS_BIF_PREP_AWAIT_X_DATA_TRAP(RET, C_P, PID, DATA)		\
-do {									\
-    erts_bif_prep_await_proc_exit_data_trap((C_P), (PID), (DATA));	\
-    (RET) = THE_NON_VALUE;						\
-} while (0)
-
-#define ERTS_BIF_PREP_AWAIT_X_REASON_TRAP(RET, C_P, PID)		\
-do {									\
-    erts_bif_prep_await_proc_exit_reason_trap((C_P), (PID));		\
-    (RET) = THE_NON_VALUE;						\
-} while (0)
-
-#define ERTS_BIF_PREP_AWAIT_X_APPLY_TRAP(RET, C_P, PID, M, F, A, AN)	\
-do {									\
-    erts_bif_prep_await_proc_exit_apply_trap((C_P), (PID),		\
-					     (M), (F), (A), (AN));	\
-    (RET) = THE_NON_VALUE;						\
-} while (0)
-
-#define ERTS_BIF_AWAIT_X_DATA_TRAP(C_P, PID, DATA)			\
-do {									\
-    erts_bif_prep_await_proc_exit_data_trap((C_P), (PID), (DATA));	\
-    return THE_NON_VALUE;						\
-} while (0)
-
-#define ERTS_BIF_AWAIT_X_REASON_TRAP(C_P, PID)				\
-do {									\
-    erts_bif_prep_await_proc_exit_reason_trap((C_P), (PID));		\
-    return THE_NON_VALUE;						\
-} while (0)
-
-#define ERTS_BIF_AWAIT_X_APPLY_TRAP(C_P, PID, M, F, A, AN)		\
-do {									\
-    erts_bif_prep_await_proc_exit_apply_trap((C_P), (PID),		\
-					     (M), (F), (A), (AN));	\
-    return THE_NON_VALUE;						\
-} while (0)
-
-void
-erts_bif_prep_await_proc_exit_data_trap(Process *c_p,
-					Eterm pid,
-					Eterm data);
-void
-erts_bif_prep_await_proc_exit_reason_trap(Process *c_p,
-					  Eterm pid);
-void
-erts_bif_prep_await_proc_exit_apply_trap(Process *c_p,
-					 Eterm pid,
-					 Eterm module,
-					 Eterm function,
-					 Eterm args[],
-					 int nargs);
-
-#ifdef ERTS_DIRTY_SCHEDULERS
 int erts_call_dirty_bif(ErtsSchedulerData *esdp, Process *c_p,
 			BeamInstr *I, Eterm *reg);
-#endif
 
 BIF_RETTYPE
 erts_schedule_bif(Process *proc,
diff --git a/erts/emulator/beam/bif.tab b/erts/emulator/beam/bif.tab
index 10ca0b5066..7548924178 100644
--- a/erts/emulator/beam/bif.tab
+++ b/erts/emulator/beam/bif.tab
@@ -1,7 +1,7 @@
 #
 # %CopyrightBegin%
 #
-# Copyright Ericsson AB 1996-2017. All Rights Reserved.
+# Copyright Ericsson AB 1996-2018. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -62,6 +62,7 @@ bif erlang:erase/0
 bif erlang:erase/1
 bif erlang:exit/1
 bif erlang:exit/2
+bif erlang:exit_signal/2
 bif erlang:external_size/1
 bif erlang:external_size/2
 gcbif erlang:float/1
@@ -73,7 +74,8 @@ bif erlang:get/0
 bif erlang:get/1
 bif erlang:get_keys/1
 bif erlang:group_leader/0
-bif erlang:group_leader/2
+bif erts_internal:group_leader/2
+bif erts_internal:group_leader/3
 bif erlang:halt/2
 bif erlang:phash/2
 bif erlang:phash2/1
@@ -123,7 +125,7 @@ bif erlang:pid_to_list/1
 bif erlang:ports/0
 bif erlang:pre_loaded/0
 bif erlang:process_flag/2
-bif erlang:process_flag/3
+bif erts_internal:process_flag/3
 bif erlang:process_info/1
 bif erlang:process_info/2
 bif erlang:processes/0
@@ -152,8 +154,11 @@ bif erlang:unregister/1
 bif erlang:whereis/1
 bif erlang:spawn_opt/1
 bif erlang:setnode/2
-bif erlang:setnode/3
-bif erlang:dist_exit/3
+bif erlang:dist_get_stat/1
+bif erlang:dist_ctrl_input_handler/2
+bif erlang:dist_ctrl_put_data/2
+bif erlang:dist_ctrl_get_data/1
+bif erlang:dist_ctrl_get_data_notification/1
 
 # Static native functions in erts_internal
 bif erts_internal:port_info/1
@@ -181,6 +186,12 @@ bif erts_internal:system_check/1
 
 bif erts_internal:release_literal_area_switch/0
 
+bif erts_internal:scheduler_wall_time/1
+
+bif erts_internal:dirty_process_handle_signals/1
+
+bif erts_internal:create_dist_channel/4
+
 # inet_db support
 bif erlang:port_set_data/2
 bif erlang:port_get_data/1
@@ -194,9 +205,9 @@ bif erlang:seq_trace/2
 bif erlang:seq_trace_info/1
 bif erlang:seq_trace_print/1
 bif erlang:seq_trace_print/2
-bif erlang:suspend_process/2
+bif erts_internal:suspend_process/2
 bif erlang:resume_process/1
-bif erlang:process_display/2
+bif erts_internal:process_display/2
 
 bif erlang:bump_reductions/1
 
@@ -256,6 +267,7 @@ bif erlang:demonitor/1
 bif erlang:demonitor/2
 
 bif erlang:is_process_alive/1
+bif erts_internal:is_process_alive/2
 
 bif erlang:error/1		error_1
 bif erlang:error/2		error_2
@@ -330,7 +342,6 @@ bif ets:internal_request_all/0
 bif ets:new/2
 bif ets:delete/1
 bif ets:delete/2
-bif ets:delete_all_objects/1
 bif ets:delete_object/2
 bif ets:first/1
 bif ets:is_compiled_ms/1
@@ -361,7 +372,6 @@ bif ets:select_count/2
 bif ets:select_reverse/1
 bif ets:select_reverse/2
 bif ets:select_reverse/3
-bif ets:select_delete/2
 bif ets:select_replace/2
 bif ets:match_spec_compile/1
 bif ets:match_spec_run_r/3
@@ -370,9 +380,10 @@ bif ets:match_spec_run_r/3
 # Bifs in os module.
 #
 
-bif os:putenv/2
-bif os:getenv/0
-bif os:getenv/1
+bif os:get_env_var/1
+bif os:set_env_var/2
+bif os:unset_env_var/1
+bif os:list_env_vars/0
 bif os:getpid/0
 bif os:timestamp/0
 bif os:system_time/0
@@ -428,13 +439,6 @@ bif erts_debug:dirty_io/2
 bif erts_debug:dirty/3
 
 #
-# Monitor testing bif's...
-#
-bif erts_debug:dump_monitors/1
-bif erts_debug:dump_links/1
-
-
-#
 # Lock counter bif's
 #
 bif erts_debug:lcnt_control/2
@@ -544,9 +548,6 @@ bif binary:last/1
 bif binary:at/2
 bif binary:part/2 binary_binary_part_2
 bif binary:part/3 binary_binary_part_3
-bif binary:bin_to_list/1
-bif binary:bin_to_list/2
-bif binary:bin_to_list/3
 bif binary:list_to_bin/1
 bif binary:copy/1
 bif binary:copy/2
@@ -613,7 +614,6 @@ bif erlang:float_to_binary/2
 bif erlang:binary_to_float/1
 
 bif io:printable_range/0
-bif os:unsetenv/1
 
 #
 # New in 17.0
@@ -623,7 +623,6 @@ bif re:inspect/2
 
 ubif erlang:is_map/1
 gcbif erlang:map_size/1
-bif maps:to_list/1
 bif maps:find/2
 bif maps:get/2
 bif maps:from_list/1
@@ -678,10 +677,24 @@ bif math:floor/1
 bif math:ceil/1
 bif math:fmod/2
 bif os:set_signal/2
-bif erts_internal:maps_to_list/2
 
 #
 # New in 20.1
 #
-
 bif erlang:iolist_to_iovec/1
+
+#
+# New in 21.0
+#
+
+bif erts_internal:get_dflags/0
+bif erts_internal:new_connection/1
+bif erts_internal:abort_connection/2
+bif erts_internal:map_next/3
+bif ets:whereis/1
+bif erts_internal:gather_alloc_histograms/1
+bif erts_internal:gather_carrier_info/1
+ubif erlang:map_get/2
+ubif erlang:is_map_key/2
+bif ets:internal_delete_all/2
+bif ets:internal_select_delete/2
diff --git a/erts/emulator/beam/bif_instrs.tab b/erts/emulator/beam/bif_instrs.tab
new file mode 100644
index 0000000000..0f074280db
--- /dev/null
+++ b/erts/emulator/beam/bif_instrs.tab
@@ -0,0 +1,547 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+// ================================================================
+// All guards with zero arguments have special instructions,
+// for example:
+//
+// 	self/0
+// 	node/0
+//
+// All other guard BIFs take one or two arguments.
+// ================================================================
+
+CALL_GUARD_BIF(BF, TmpReg, Dst) {
+    Eterm result;
+
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS;
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    ERTS_CHK_MBUF_SZ(c_p);
+    result = (*$BF)(c_p, $TmpReg, I);
+    ERTS_CHK_MBUF_SZ(c_p);
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    ERTS_HOLE_CHECK(c_p);
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    if (ERTS_LIKELY(is_value(result))) {
+        $Dst = result;
+        $NEXT0();
+    }
+}
+
+// Guard BIF in head.  On failure, ignore the error and jump
+// to the code for the next clause.  We don't support tracing
+// of guard BIFs.
+
+bif1(Fail, Bif, Src, Dst) {
+    ErtsBifFunc bf;
+    Eterm tmp_reg[1];
+
+    tmp_reg[0] = $Src;
+    bf = (BifFunction) $Bif;
+    $CALL_GUARD_BIF(bf, tmp_reg, $Dst);
+
+    $FAIL($Fail);
+}
+
+//
+// Guard BIF in body.  It can fail like any BIF.  No trace support.
+//
+
+bif1_body(Bif, Src, Dst) {
+    ErtsBifFunc bf;
+    Eterm tmp_reg[1];
+
+    tmp_reg[0] = $Src;
+    bf = (BifFunction) $Bif;
+    $CALL_GUARD_BIF(bf, tmp_reg, $Dst);
+
+    reg[0] = tmp_reg[0];
+    SWAPOUT;
+    I = handle_error(c_p, I, reg, ubif2mfa((void *) bf));
+    goto post_error_handling;
+}
+
+//
+// Guard bif in guard with two arguments ('and'/2, 'or'/2, 'xor'/2).
+//
+
+i_bif2(Fail, Bif, Src1, Src2, Dst) {
+    Eterm tmp_reg[2];
+    ErtsBifFunc bf;
+
+    tmp_reg[0] = $Src1;
+    tmp_reg[1] = $Src2;
+    bf = (ErtsBifFunc) $Bif;
+    $CALL_GUARD_BIF(bf, tmp_reg, $Dst);
+    $FAIL($Fail);
+}
+
+//
+// Guard bif in body with two arguments ('and'/2, 'or'/2, 'xor'/2).
+//
+
+i_bif2_body(Bif, Src1, Src2, Dst) {
+    Eterm tmp_reg[2];
+    ErtsBifFunc bf;
+
+    tmp_reg[0] = $Src1;
+    tmp_reg[1] = $Src2;
+    bf = (ErtsBifFunc) $Bif;
+    $CALL_GUARD_BIF(bf, tmp_reg, $Dst);
+    reg[0] = tmp_reg[0];
+    reg[1] = tmp_reg[1];
+    SWAPOUT;
+    I = handle_error(c_p, I, reg, ubif2mfa((void *) bf));
+    goto post_error_handling;
+}
+
+//
+// Garbage-collecting BIF with one argument in either guard or body.
+//
+
+i_gc_bif1(Fail, Bif, Src, Live, Dst) {
+    typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
+    GcBifFunction bf;
+    Eterm result;
+    Uint live = (Uint) $Live;
+
+    x(live) = $Src;
+    bf = (GcBifFunction) $Bif;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS;
+    SWAPOUT;
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_CHK_MBUF_SZ(c_p);
+    result = (*bf)(c_p, reg, live);
+    ERTS_CHK_MBUF_SZ(c_p);
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    SWAPIN;
+    ERTS_HOLE_CHECK(c_p);
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    if (ERTS_LIKELY(is_value(result))) {
+        $REFRESH_GEN_DEST();
+        $Dst = result;
+        $NEXT0();
+    }
+    if (ERTS_LIKELY($Fail != 0)) { /* Handle error in guard. */
+        $JUMP($Fail);
+    }
+
+    /* Handle error in body. */
+    x(0) = x(live);
+    I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
+    goto post_error_handling;
+}
+
+//
+// Garbage-collecting BIF with two arguments in either guard or body.
+//
+
+i_gc_bif2(Fail, Bif, Live, Src1, Src2, Dst) {
+    typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
+    GcBifFunction bf;
+    Eterm result;
+    Uint live = (Uint) $Live;
+
+    /*
+     * XXX This calling convention does not make sense. 'live'
+     * should point out the first argument, not the second
+     * (i.e. 'live' should not be incremented below).
+     */
+    x(live) = $Src1;
+    x(live+1) = $Src2;
+    live++;
+
+    bf = (GcBifFunction) $Bif;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS;
+    SWAPOUT;
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_CHK_MBUF_SZ(c_p);
+    result = (*bf)(c_p, reg, live);
+    ERTS_CHK_MBUF_SZ(c_p);
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    SWAPIN;
+    ERTS_HOLE_CHECK(c_p);
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    if (ERTS_LIKELY(is_value(result))) {
+        $REFRESH_GEN_DEST();
+        $Dst = result;
+        $NEXT0();
+    }
+
+    if (ERTS_LIKELY($Fail != 0)) { /* Handle error in guard. */
+        $JUMP($Fail);
+    }
+
+    /* Handle error in body. */
+    live--;
+    x(0) = x(live);
+    x(1) = x(live+1);
+    I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
+    goto post_error_handling;
+}
+
+//
+// Garbage-collecting BIF with three arguments in either guard or body.
+//
+
+i_gc_bif3(Fail, Bif, Live, Src2, Src3, Dst) {
+    typedef Eterm (*GcBifFunction)(Process*, Eterm*, Uint);
+    GcBifFunction bf;
+    Eterm result;
+    Uint live = (Uint) $Live;
+
+    /*
+     * XXX This calling convention does not make sense. 'live'
+     * should point out the first argument, not the third
+     * (i.e. 'live' should not be incremented below).
+     */
+    x(live) = x(SCRATCH_X_REG);
+    x(live+1) = $Src2;
+    x(live+2) = $Src3;
+    live += 2;
+
+    bf = (GcBifFunction) $Bif;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS;
+    SWAPOUT;
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_CHK_MBUF_SZ(c_p);
+    result = (*bf)(c_p, reg, live);
+    ERTS_CHK_MBUF_SZ(c_p);
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    SWAPIN;
+    ERTS_HOLE_CHECK(c_p);
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    if (ERTS_LIKELY(is_value(result))) {
+        $REFRESH_GEN_DEST();
+        $Dst = result;
+        $NEXT0();
+    }
+
+    /* Handle error in guard. */
+    if (ERTS_LIKELY($Fail != 0)) {
+        $JUMP($Fail);
+    }
+
+    /* Handle error in body. */
+    live -= 2;
+    x(0) = x(live);
+    x(1) = x(live+1);
+    x(2) = x(live+2);
+    I = handle_error(c_p, I, reg, gcbif2mfa((void *) bf));
+    goto post_error_handling;
+}
+
+//
+// The most general BIF call.  The BIF may build any amount of data
+// on the heap.  The result is always returned in r(0).
+//
+call_bif(Exp) {
+    ErtsBifFunc bf;
+    Eterm result;
+    ErlHeapFragment *live_hf_end;
+    Export *export = (Export*) $Exp;
+
+    if (!((FCALLS - 1) > 0 || (FCALLS-1) > neg_o_reds)) {
+        /* If we have run out of reductions, we do a context
+           switch before calling the bif */
+        c_p->arity = GET_BIF_ARITY(export);
+        c_p->current = &export->info.mfa;
+        goto context_switch3;
+    }
+
+    ERTS_MSACC_SET_BIF_STATE_CACHED_X(GET_BIF_MODULE(export),
+                                      GET_BIF_ADDRESS(export));
+
+    bf = GET_BIF_ADDRESS(export);
+
+    PRE_BIF_SWAPOUT(c_p);
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS - 1;
+    if (FCALLS <= 0) {
+        save_calls(c_p, export);
+    }
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    live_hf_end = c_p->mbuf;
+    ERTS_CHK_MBUF_SZ(c_p);
+    result = (*bf)(c_p, reg, I);
+    ERTS_CHK_MBUF_SZ(c_p);
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p) || is_non_value(result));
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    ERTS_HOLE_CHECK(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    if (ERTS_IS_GC_DESIRED(c_p)) {
+        Uint arity = GET_BIF_ARITY(export);
+        result = erts_gc_after_bif_call_lhf(c_p, live_hf_end, result,
+                                            reg, arity);
+        E = c_p->stop;
+    }
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    HTOP = HEAP_TOP(c_p);
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    /* We have to update the cache if we are enabled in order
+       to make sure no book keeping is done after we disabled
+       msacc. We don't always do this as it is quite expensive. */
+    if (ERTS_MSACC_IS_ENABLED_CACHED_X()) {
+        ERTS_MSACC_UPDATE_CACHE_X();
+    }
+    ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
+    if (ERTS_LIKELY(is_value(result))) {
+        r(0) = result;
+        CHECK_TERM(r(0));
+        $NEXT0();
+    } else if (c_p->freason == TRAP) {
+        SET_CP(c_p, I+2);
+        SET_I(c_p->i);
+        SWAPIN;
+        Dispatch();
+    }
+
+    /*
+     * Error handling.  SWAPOUT is not needed because it was done above.
+     */
+    ASSERT(c_p->stop == E);
+    I = handle_error(c_p, I, reg, &export->info.mfa);
+    goto post_error_handling;
+}
+
+//
+// Send is almost a standard call-BIF with two arguments, except for:
+//   1. It cannot be traced.
+//   2. There is no pointer to the send_2 function stored in
+//      the instruction.
+//
+
+send() {
+    Eterm result;
+
+    if (!(FCALLS > 0 || FCALLS > neg_o_reds)) {
+        /* If we have run out of reductions, we do a context
+           switch before calling the bif */
+        c_p->arity = 2;
+        c_p->current = NULL;
+        goto context_switch3;
+    }
+
+    PRE_BIF_SWAPOUT(c_p);
+    c_p->fcalls = FCALLS - 1;
+    result = erl_send(c_p, r(0), x(1));
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    HTOP = HEAP_TOP(c_p);
+    FCALLS = c_p->fcalls;
+    if (ERTS_LIKELY(is_value(result))) {
+        r(0) = result;
+        CHECK_TERM(r(0));
+    } else if (c_p->freason == TRAP) {
+        SET_CP(c_p, I+1);
+        SET_I(c_p->i);
+        SWAPIN;
+        Dispatch();
+    } else {
+        goto find_func_info;
+    }
+}
+
+call_nif := nif_bif.call_nif.epilogue;
+apply_bif := nif_bif.apply_bif.epilogue;
+
+nif_bif.head() {
+    Eterm nif_bif_result;
+    Eterm bif_nif_arity;
+    BifFunction vbf;
+    ErlHeapFragment *live_hf_end;
+    ErtsCodeMFA *codemfa;
+}
+
+nif_bif.call_nif() {
+    /*
+     * call_nif is always first instruction in function:
+     *
+     * I[-3]: Module
+     * I[-2]: Function
+     * I[-1]: Arity
+     * I[0]: &&call_nif
+     * I[1]: Function pointer to NIF function
+     * I[2]: Pointer to erl_module_nif
+     * I[3]: Function pointer to dirty NIF
+     *
+     * This layout is determined by the NifExport struct
+     */
+
+    ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_NIF);
+
+    codemfa = erts_code_to_codemfa(I);
+
+    c_p->current = codemfa; /* current and vbf set to please handle_error */
+
+    DTRACE_NIF_ENTRY(c_p, codemfa);
+
+    HEAVY_SWAPOUT;
+
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    bif_nif_arity = codemfa->arity;
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    {
+        typedef Eterm NifF(struct enif_environment_t*, int argc, Eterm argv[]);
+        NifF* fp = vbf = (NifF*) I[1];
+        struct enif_environment_t env;
+        ASSERT(c_p->scheduler_data);
+        live_hf_end = c_p->mbuf;
+        ERTS_CHK_MBUF_SZ(c_p);
+        erts_pre_nif(&env, c_p, (struct erl_module_nif*)I[2], NULL);
+
+        ASSERT((c_p->scheduler_data)->current_nif == NULL);
+        (c_p->scheduler_data)->current_nif = &env;
+
+        nif_bif_result = (*fp)(&env, bif_nif_arity, reg);
+        if (env.exception_thrown)
+            nif_bif_result = THE_NON_VALUE;
+
+        ASSERT((c_p->scheduler_data)->current_nif == &env);
+        (c_p->scheduler_data)->current_nif = NULL;
+
+        erts_post_nif(&env);
+        ERTS_CHK_MBUF_SZ(c_p);
+
+        PROCESS_MAIN_CHK_LOCKS(c_p);
+        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+        ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
+        ASSERT(!env.exiting);
+        ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    }
+
+    DTRACE_NIF_RETURN(c_p, codemfa);
+}
+
+nif_bif.apply_bif() {
+    /*
+     * At this point, I points to the code[0] in the export entry for
+     * the BIF:
+     *
+     * code[-3]: Module
+     * code[-2]: Function
+     * code[-1]: Arity
+     * code[0]: &&apply_bif
+     * code[1]: Function pointer to BIF function
+     */
+
+    if (!((FCALLS - 1) > 0 || (FCALLS - 1) > neg_o_reds)) {
+        /* If we have run out of reductions, we do a context
+           switch before calling the bif */
+        goto context_switch;
+    }
+
+    codemfa = erts_code_to_codemfa(I);
+
+    ERTS_MSACC_SET_BIF_STATE_CACHED_X(codemfa->module, (BifFunction)Arg(0));
+
+
+    /* In case we apply process_info/1,2 or load_nif/1 */
+    c_p->current = codemfa;
+    $SET_CP_I_ABS(I);     /* In case we apply check_process_code/2. */
+    c_p->arity = 0;       /* To allow garbage collection on ourselves
+                           * (check_process_code/2).
+                           */
+    DTRACE_BIF_ENTRY(c_p, codemfa);
+
+    SWAPOUT;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS - 1);
+    c_p->fcalls = FCALLS - 1;
+    vbf = (BifFunction) Arg(0);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+    bif_nif_arity = codemfa->arity;
+    ASSERT(bif_nif_arity <= 4);
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    {
+        ErtsBifFunc bf = vbf;
+        ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+        live_hf_end = c_p->mbuf;
+        ERTS_CHK_MBUF_SZ(c_p);
+        nif_bif_result = (*bf)(c_p, reg, I);
+        ERTS_CHK_MBUF_SZ(c_p);
+        ASSERT(!ERTS_PROC_IS_EXITING(c_p) ||
+               is_non_value(nif_bif_result));
+        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+        PROCESS_MAIN_CHK_LOCKS(c_p);
+    }
+    /* We have to update the cache if we are enabled in order
+       to make sure no book keeping is done after we disabled
+       msacc. We don't always do this as it is quite expensive. */
+    if (ERTS_MSACC_IS_ENABLED_CACHED_X())
+        ERTS_MSACC_UPDATE_CACHE_X();
+    ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_EMULATOR);
+    DTRACE_BIF_RETURN(c_p, codemfa);
+}
+
+nif_bif.epilogue() {
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    ERTS_HOLE_CHECK(c_p);
+    if (ERTS_IS_GC_DESIRED(c_p)) {
+        nif_bif_result = erts_gc_after_bif_call_lhf(c_p, live_hf_end,
+                                                    nif_bif_result,
+                                                    reg, bif_nif_arity);
+    }
+    SWAPIN;  /* There might have been a garbage collection. */
+    FCALLS = c_p->fcalls;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    if (ERTS_LIKELY(is_value(nif_bif_result))) {
+        r(0) = nif_bif_result;
+        CHECK_TERM(r(0));
+        SET_I(c_p->cp);
+        c_p->cp = 0;
+        Goto(*I);
+    } else if (c_p->freason == TRAP) {
+        SET_I(c_p->i);
+        if (c_p->flags & F_HIBERNATE_SCHED) {
+            c_p->flags &= ~F_HIBERNATE_SCHED;
+            goto do_schedule;
+        }
+        Dispatch();
+    }
+    I = handle_error(c_p, c_p->cp, reg, c_p->current);
+    goto post_error_handling;
+}
diff --git a/erts/emulator/beam/big.h b/erts/emulator/beam/big.h
index 48efce20e7..7556205063 100644
--- a/erts/emulator/beam/big.h
+++ b/erts/emulator/beam/big.h
@@ -70,7 +70,20 @@ typedef Uint  dsize_t;	 /* Vector size type */
 
 /* Check for small */
 #define IS_USMALL(sgn,x)  ((sgn) ? ((x) <= MAX_SMALL+1) : ((x) <= MAX_SMALL))
-#define IS_SSMALL(x)      (((x) >= MIN_SMALL) && ((x) <= MAX_SMALL))
+
+/*
+ * It seems that both clang and gcc will generate sub-optimal code
+ * for the more obvious way to write the range check:
+ *
+ *    #define IS_SSMALL(x)  (((x) >= MIN_SMALL) && ((x) <= MAX_SMALL))
+ *
+ * Note that IS_SSMALL() may be used in the 32-bit emulator with
+ * a Uint64 argument. Therefore, we must test the size of the argument
+ * to ensure that the cast does not discard the high-order 32 bits.
+ */
+#define _IS_SSMALL32(x) (((Uint32) ((((x)) >> (SMALL_BITS-1)) + 1)) < 2)
+#define _IS_SSMALL64(x) (((Uint64) ((((x)) >> (SMALL_BITS-1)) + 1)) < 2)
+#define IS_SSMALL(x) (sizeof(x) == sizeof(Uint32) ? _IS_SSMALL32(x) : _IS_SSMALL64(x))
 
 /* The heap size needed for a bignum */
 #define BIG_NEED_SIZE(x)  ((x) + 1)
diff --git a/erts/emulator/beam/binary.c b/erts/emulator/beam/binary.c
index 95d324d2c1..d53f75c279 100644
--- a/erts/emulator/beam/binary.c
+++ b/erts/emulator/beam/binary.c
@@ -113,6 +113,40 @@ new_binary(Process *p, byte *buf, Uint len)
     return build_proc_bin(&MSO(p), HAlloc(p, PROC_BIN_SIZE), bptr);
 }
 
+Eterm
+erts_heap_factory_new_binary(ErtsHeapFactory *hfact, byte *buf, Uint len,
+                             Uint reserve_size)
+{
+    Eterm *hp;
+    Binary* bptr;
+
+    if (len <= ERL_ONHEAP_BIN_LIMIT) {
+	ErlHeapBin* hb;
+        hp = erts_produce_heap(hfact, heap_bin_size(len), reserve_size);
+        hb = (ErlHeapBin *) hp;
+	hb->thing_word = header_heap_bin(len);
+	hb->size = len;
+	if (buf != NULL) {
+	    sys_memcpy(hb->data, buf, len);
+	}
+	return make_binary(hb);
+    }
+
+    /*
+     * Allocate the binary struct itself.
+     */
+    bptr = erts_bin_nrml_alloc(len);
+    if (buf != NULL) {
+	sys_memcpy(bptr->orig_bytes, buf, len);
+    }
+
+    hp = erts_produce_heap(hfact, PROC_BIN_SIZE, reserve_size);
+
+    return build_proc_bin(hfact->off_heap, hp, bptr);
+}
+
+
+
 /* 
  * When heap binary is not desired...
  */
diff --git a/erts/emulator/beam/break.c b/erts/emulator/beam/break.c
index 5ace997344..9ff52c92b8 100644
--- a/erts/emulator/beam/break.c
+++ b/erts/emulator/beam/break.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -36,9 +36,9 @@
 #include "hash.h"
 #include "atom.h"
 #include "beam_load.h"
-#include "erl_instrument.h"
 #include "erl_hl_timer.h"
 #include "erl_thr_progress.h"
+#include "erl_proc_sig_queue.h"
 
 /* Forward declarations -- should really appear somewhere else */
 static void process_killer(void);
@@ -107,36 +107,11 @@ process_killer(void)
 		if ((j = sys_get_key(0)) <= 0)
 		    erts_exit(0, "");
 		switch(j) {
-		case 'k': {
-		    ErtsProcLocks rp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
-		    erts_aint32_t state;
-		    erts_proc_inc_refc(rp);
-		    erts_smp_proc_lock(rp, rp_locks);
-		    state = erts_smp_atomic32_read_acqb(&rp->state);
-		    if (state & (ERTS_PSFLG_FREE
-				 | ERTS_PSFLG_EXITING
-				 | ERTS_PSFLG_ACTIVE
-				 | ERTS_PSFLG_ACTIVE_SYS
-				 | ERTS_PSFLG_IN_RUNQ
-				 | ERTS_PSFLG_RUNNING
-				 | ERTS_PSFLG_RUNNING_SYS
-				 | ERTS_PSFLG_DIRTY_RUNNING
-				 | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
-			erts_printf("Can only kill WAITING processes this way\n");
-		    }
-		    else {
-			(void) erts_send_exit_signal(NULL,
-						     NIL,
-						     rp,
-						     &rp_locks,
-						     am_kill,
-						     NIL,
-						     NULL,
-						     0);
-		    }
-		    erts_smp_proc_unlock(rp, rp_locks);
-		    erts_proc_dec_refc(rp);
-		}
+		case 'k':
+                    /* Send a 'kill' exit signal from init process */
+                    erts_proc_sig_send_exit(NULL, erts_init_process_id,
+                                            rp->common.id, am_kill, NIL,
+                                            0);
 		case 'n': br = 1; break;
 		case 'r': return;
 		default: return;
@@ -161,47 +136,64 @@ static void doit_print_link(ErtsLink *lnk, void *vpcontext)
 
     if (pcontext->is_first) {
 	pcontext->is_first = 0;
-	erts_print(to, to_arg, "%T", lnk->pid);
+	erts_print(to, to_arg, "%T", lnk->other.item);
     } else {
-	erts_print(to, to_arg, ", %T", lnk->pid);
+	erts_print(to, to_arg, ", %T", lnk->other.item);
     }
 }
     
 
 static void doit_print_monitor(ErtsMonitor *mon, void *vpcontext)
 {
+    ErtsMonitorData *mdp;
     PrintMonitorContext *pcontext = vpcontext;
     fmtfn_t to = pcontext->to;
     void *to_arg = pcontext->to_arg;
     char *prefix = ", ";
  
-    if (pcontext->is_first) {
-	pcontext->is_first = 0;
-	prefix = "";
-    }
-
+    mdp = erts_monitor_to_data(mon);
     switch (mon->type) {
-    case MON_ORIGIN:
-	if (is_atom(mon->u.pid)) { /* dist by name */
-	    ASSERT(is_node_name_atom(mon->u.pid));
-	    erts_print(to, to_arg, "%s{to,{%T,%T},%T}", prefix, mon->name,
-		       mon->u.pid, mon->ref);
-	} else if (is_atom(mon->name)){ /* local by name */
-	    erts_print(to, to_arg, "%s{to,{%T,%T},%T}", prefix, mon->name,
-		       erts_this_dist_entry->sysname, mon->ref);
-	} else { /* local and distributed by pid */
-	    erts_print(to, to_arg, "%s{to,%T,%T}", prefix, mon->u.pid, mon->ref);
-	}
-	break;
-    case MON_TARGET:
-	erts_print(to, to_arg, "%s{from,%T,%T}", prefix, mon->u.pid, mon->ref);
-	break;
-    case MON_NIF_TARGET: {
-        ErtsResource* rsrc = mon->u.resource;
-        erts_print(to, to_arg, "%s{from,{%T,%T},%T}", prefix, rsrc->type->module,
-                   rsrc->type->name, mon->ref);
-	break;
-    }
+    case ERTS_MON_TYPE_PROC:
+    case ERTS_MON_TYPE_PORT:
+    case ERTS_MON_TYPE_TIME_OFFSET:
+    case ERTS_MON_TYPE_DIST_PROC:
+    case ERTS_MON_TYPE_RESOURCE:
+    case ERTS_MON_TYPE_NODE:
+
+        if (pcontext->is_first) {
+            pcontext->is_first = 0;
+            prefix = "";
+        }
+
+        if (erts_monitor_is_target(mon)) {
+            if (mon->type != ERTS_MON_TYPE_RESOURCE)
+                erts_print(to, to_arg, "%s{from,%T,%T}", prefix, mon->other.item, mdp->ref);
+            else {
+                ErtsResource* rsrc = mon->other.ptr;
+                erts_print(to, to_arg, "%s{from,{%T,%T},%T}", prefix, rsrc->type->module,
+                           rsrc->type->name, mdp->ref);
+            }
+        }
+        else {
+            if (!(mon->flags & ERTS_ML_FLG_NAME))
+                erts_print(to, to_arg, "%s{to,%T,%T}", prefix, mon->other.item, mdp->ref);
+            else {
+                ErtsMonitorDataExtended *mdep = (ErtsMonitorDataExtended *) mdp;
+                Eterm node;
+                if (mdep->dist)
+                    node = mdep->dist->nodename;
+                else
+                    node = erts_this_dist_entry->sysname;
+                erts_print(to, to_arg, "%s{to,{%T,%T},%T}", prefix, mdep->u.name,
+                           node, mdp->ref);
+            }
+        }
+
+        break;
+
+    default:
+        /* ignore other monitors... */
+        break;
     }
 }
 			       
@@ -209,9 +201,9 @@ static void doit_print_monitor(ErtsMonitor *mon, void *vpcontext)
 void
 print_process_info(fmtfn_t to, void *to_arg, Process *p)
 {
-    time_t approx_started;
     int garbing = 0;
     int running = 0;
+    Sint len;
     struct saved_calls *scb;
     erts_aint32_t state;
 
@@ -221,7 +213,7 @@ print_process_info(fmtfn_t to, void *to_arg, Process *p)
     /* Display the state */
     erts_print(to, to_arg, "State: ");
 
-    state = erts_smp_atomic32_read_acqb(&p->state);
+    state = erts_atomic32_read_acqb(&p->state);
     erts_dump_process_state(to, to_arg, state);
     if (state & ERTS_PSFLG_GC) {
         garbing = 1;
@@ -258,17 +250,22 @@ print_process_info(fmtfn_t to, void *to_arg, Process *p)
     }
 
     erts_print(to, to_arg, "Spawned by: %T\n", p->parent);
-    approx_started = (time_t) p->approx_started;
-    erts_print(to, to_arg, "Started: %s", ctime(&approx_started));
-    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p);
-    erts_print(to, to_arg, "Message queue length: %d\n", p->msg.len);
+
+    erts_proc_lock(p, ERTS_PROC_LOCK_MSGQ);
+    len = erts_proc_sig_fetch(p);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_MSGQ);
+    erts_print(to, to_arg, "Message queue length: %d\n", len);
 
     /* display the message queue only if there is anything in it */
-    if (!ERTS_IS_CRASH_DUMPING && p->msg.first != NULL && !garbing) {
-	ErtsMessage* mp;
+    if (!ERTS_IS_CRASH_DUMPING && p->sig_qs.first != NULL && !garbing) {
 	erts_print(to, to_arg, "Message queue: [");
-	for (mp = p->msg.first; mp; mp = mp->next)
-	    erts_print(to, to_arg, mp->next ? "%T," : "%T", ERL_MESSAGE_TERM(mp));
+        ERTS_FOREACH_SIG_PRIVQS(
+            p, mp,
+            {
+                if (ERTS_SIG_IS_NON_MSG((ErtsSignal *) mp))
+                    erts_print(to, to_arg, mp->next ? "%T," : "%T",
+                               ERL_MESSAGE_TERM(mp));
+            });
 	erts_print(to, to_arg, "]\n");
     }
 
@@ -309,11 +306,12 @@ print_process_info(fmtfn_t to, void *to_arg, Process *p)
     }
 
     /* display the links only if there are any*/
-    if (ERTS_P_LINKS(p) || ERTS_P_MONITORS(p)) {
+    if (ERTS_P_LINKS(p) || ERTS_P_MONITORS(p) || ERTS_P_LT_MONITORS(p)) {
 	PrintMonitorContext context = {1, to, to_arg};
 	erts_print(to, to_arg,"Link list: [");
-	erts_doforall_links(ERTS_P_LINKS(p), &doit_print_link, &context);	
-	erts_doforall_monitors(ERTS_P_MONITORS(p), &doit_print_monitor, &context);
+	erts_link_tree_foreach(ERTS_P_LINKS(p), doit_print_link, &context);	
+	erts_monitor_tree_foreach(ERTS_P_MONITORS(p), doit_print_monitor, &context);
+	erts_monitor_list_foreach(ERTS_P_LT_MONITORS(p), doit_print_monitor, &context);
 	erts_print(to, to_arg,"]\n");
     }
 
@@ -352,9 +350,7 @@ print_process_info(fmtfn_t to, void *to_arg, Process *p)
 	erts_program_counter_info(to, to_arg, p);
     } else {
 	erts_print(to, to_arg, "Stack dump:\n");
-#ifdef ERTS_SMP
 	if (!garbing)
-#endif
 	    erts_stack_dump(to, to_arg, p);
     }
 
@@ -366,11 +362,9 @@ print_process_info(fmtfn_t to, void *to_arg, Process *p)
 static void
 print_garb_info(fmtfn_t to, void *to_arg, Process* p)
 {
-#ifdef ERTS_SMP
-    /* ERTS_SMP: A scheduler is probably concurrently doing gc... */
+    /* A scheduler is probably concurrently doing gc... */
     if (!ERTS_IS_CRASH_DUMPING)
       return;
-#endif
     erts_print(to, to_arg, "New heap start: %bpX\n", p->heap);
     erts_print(to, to_arg, "New heap top: %bpX\n", p->htop);
     erts_print(to, to_arg, "Stack top: %bpX\n", p->stop);
@@ -513,12 +507,12 @@ do_break(void)
     /* check if we're in console mode and, if so,
        halt immediately if break is called */
     mode = erts_read_env("ERL_CONSOLE_MODE");
-    if (mode && strcmp(mode, "window") != 0)
+    if (mode && sys_strcmp(mode, "window") != 0)
 	erts_exit(0, "");
     erts_free_read_env(mode);
 #endif /* __WIN32__ */
 
-    ASSERT(erts_smp_thr_progress_is_blocking());
+    ASSERT(erts_thr_progress_is_blocking());
 
     erts_printf("\n"
 		"BREAK: (a)bort (c)ontinue (p)roc info (i)nfo (l)oaded\n"
@@ -714,9 +708,7 @@ crash_dump_limited_writer(void* vfdp, char* buf, size_t len)
 void
 erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
 {
-#ifdef ERTS_SMP
     ErtsThrPrgrData tpd_buf; /* in case we aren't a managed thread... */
-#endif
     int fd;
     size_t envsz;
     time_t now;
@@ -736,7 +728,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     if (ERTS_SOMEONE_IS_CRASH_DUMPING)
 	return;
 
-#ifdef ERTS_SMP
     /* Order all managed threads to block, this has to be done
        first to guarantee that this is the only thread to generate
        crash dump. */
@@ -760,12 +751,9 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
 #endif
 
     /* Allow us to pass certain places without locking... */
-    erts_smp_atomic32_set_mb(&erts_writing_erl_crash_dump, 1);
-    erts_smp_tsd_set(erts_is_crash_dumping_key, (void *) 1);
+    erts_atomic32_set_mb(&erts_writing_erl_crash_dump, 1);
+    erts_tsd_set(erts_is_crash_dumping_key, (void *) 1);
 
-#else /* !ERTS_SMP */
-    erts_writing_erl_crash_dump = 1;
-#endif /* ERTS_SMP */
 
     envsz = sizeof(env);
     /* ERL_CRASH_DUMP_SECONDS not set
@@ -791,16 +779,16 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
      * - write dump until alarm or file is written completely
      */
 	
-    if (erts_sys_getenv__("ERL_CRASH_DUMP_SECONDS", env, &envsz) != 0) {
-	env_erl_crash_dump_seconds_set = 0;
-	secs = -1;
+    if (erts_sys_explicit_8bit_getenv("ERL_CRASH_DUMP_SECONDS", env, &envsz) == 1) {
+        env_erl_crash_dump_seconds_set = 1;
+        secs = atoi(env);
     } else {
-	env_erl_crash_dump_seconds_set = 1;
-	secs = atoi(env);
+        env_erl_crash_dump_seconds_set = 0;
+        secs = -1;
     }
 
     if (secs == 0) {
-	return;
+        return;
     }
 
     /* erts_sys_prepare_crash_dump returns 1 if heart port is found, otherwise 0
@@ -816,7 +804,7 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
 
     crash_dump_limit = ERTS_SINT64_MAX;
     envsz = sizeof(env);
-    if (erts_sys_getenv__("ERL_CRASH_DUMP_BYTES", env, &envsz) == 0) {
+    if (erts_sys_explicit_8bit_getenv("ERL_CRASH_DUMP_BYTES", env, &envsz) == 1) {
         Sint64 limit;
         char* endptr;
         errno = 0;
@@ -829,7 +817,7 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
         }
     }
 
-    if (erts_sys_getenv__("ERL_CRASH_DUMP",&dumpnamebuf[0],&dumpnamebufsize) != 0)
+    if (erts_sys_explicit_8bit_getenv("ERL_CRASH_DUMP",&dumpnamebuf[0],&dumpnamebufsize) != 1)
 	dumpname = "erl_crash.dump";
     else
 	dumpname = &dumpnamebuf[0];
@@ -881,7 +869,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     erts_print_nif_taints(to, to_arg);
     erts_cbprintf(to, to_arg, "Atoms: %d\n", atom_table_size());
 
-#ifdef USE_THREADS
     /* We want to note which thread it was that called erts_exit */
     if (erts_get_scheduler_data()) {
         erts_cbprintf(to, to_arg, "Calling Thread: scheduler:%d\n",
@@ -892,9 +879,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
         else
             erts_cbprintf(to, to_arg, "Calling Thread: %p\n", erts_thr_self());
     }
-#else
-    erts_cbprintf(to, to_arg, "Calling Thread: scheduler:1\n");
-#endif
 
 #if defined(ERTS_HAVE_TRY_CATCH)
 
@@ -911,7 +895,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
             erts_print_scheduler_info(to, to_arg, ERTS_SCHEDULER_IX(i)),
             erts_cbprintf(to, to_arg, "** crashed **\n"));
     }
-#ifdef ERTS_DIRTY_SCHEDULERS
     for (i = 0; i < erts_no_dirty_cpu_schedulers; i++) {
         ERTS_SYS_TRY_CATCH(
             erts_print_scheduler_info(to, to_arg, ERTS_DIRTY_CPU_SCHEDULER_IX(i)),
@@ -927,10 +910,8 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     }
     erts_cbprintf(to, to_arg, "=dirty_io_run_queue\n");
     erts_print_run_queue_info(to, to_arg, ERTS_DIRTY_IO_RUNQ);
-#endif /* ERTS_DIRTY_SCHEDULERS */
 #endif
 
-#ifdef ERTS_SMP
 
 #ifdef ERTS_SYS_SUSPEND_SIGNAL
 
@@ -952,7 +933,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
      */
     erts_thr_progress_fatal_error_wait(60000);
     /* Either worked or not... */
-#endif
 
 #ifndef ERTS_HAVE_TRY_CATCH
     /* This is safe to call here, as all schedulers are blocked */
@@ -974,20 +954,6 @@ erl_crash_dump_v(char *file, int line, char* fmt, va_list args)
     erts_cbprintf(to, to_arg, "=atoms\n");
     dump_atoms(to, to_arg);
 
-    /* Keep the instrumentation data at the end of the dump */
-    if (erts_instr_memory_map || erts_instr_stat) {
-	erts_cbprintf(to, to_arg, "=instr_data\n");
-
-	if (erts_instr_stat) {
-	    erts_cbprintf(to, to_arg, "=memory_status\n");
-	    erts_instr_dump_stat_to(to, to_arg, 0);
-	}
-	if (erts_instr_memory_map) {
-	    erts_cbprintf(to, to_arg, "=memory_map\n");
-	    erts_instr_dump_memory_map_to(to, to_arg);
-	}
-    }
-
     erts_cbprintf(to, to_arg, "=end\n");
     if (fp) {
         fclose(fp);
diff --git a/erts/emulator/beam/bs_instrs.tab b/erts/emulator/beam/bs_instrs.tab
new file mode 100644
index 0000000000..94e0000c8b
--- /dev/null
+++ b/erts/emulator/beam/bs_instrs.tab
@@ -0,0 +1,1038 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+%if ARCH_64
+BS_SAFE_MUL(A, B, Fail, Dst) {
+    Uint64 res = ($A) * ($B);
+    if (res / $B != $A) {
+        $Fail;
+    }
+    $Dst = res;
+}
+%else
+BS_SAFE_MUL(A, B, Fail, Dst) {
+    Uint64 res = (Uint64)($A) * (Uint64)($B);
+    if ((res >> (8*sizeof(Uint))) != 0) {
+        $Fail;
+    }
+    $Dst = res;
+}
+%endif
+
+BS_GET_FIELD_SIZE(Bits, Unit, Fail, Dst) {
+    Sint signed_size;
+    Uint uint_size;
+    Uint temp_bits;
+
+    if (is_small($Bits)) {
+        signed_size = signed_val($Bits);
+        if (signed_size < 0) {
+            $Fail;
+        }
+        uint_size = (Uint) signed_size;
+    } else {
+        if (!term_to_Uint($Bits, &temp_bits)) {
+            $Fail;
+        }
+        uint_size = temp_bits;
+    }
+    $BS_SAFE_MUL(uint_size, $Unit, $Fail, $Dst);
+}
+
+BS_GET_UNCHECKED_FIELD_SIZE(Bits, Unit, Fail, Dst) {
+    Sint signed_size;
+    Uint uint_size;
+    Uint temp_bits;
+
+    if (is_small($Bits)) {
+        signed_size = signed_val($Bits);
+        if (signed_size < 0) {
+            $Fail;
+        }
+        uint_size = (Uint) signed_size;
+    } else {
+        if (!term_to_Uint($Bits, &temp_bits)) {
+            $Fail;
+        }
+        uint_size = temp_bits;
+    }
+    $Dst = uint_size * $Unit;
+}
+
+TEST_BIN_VHEAP(VNh, Nh, Live) {
+    Uint need = $Nh;
+    if (E - HTOP < need || MSO(c_p).overhead + $VNh >= BIN_VHEAP_SZ(c_p)) {
+        SWAPOUT;
+        PROCESS_MAIN_CHK_LOCKS(c_p);
+        FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, $Live, FCALLS);
+        ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+        PROCESS_MAIN_CHK_LOCKS(c_p);
+        SWAPIN;
+    }
+    HEAP_SPACE_VERIFIED(need);
+}
+
+i_bs_get_binary_all2(Fail, Ms, Live, Unit, Dst) {
+    ErlBinMatchBuffer *_mb;
+    Eterm _result;
+
+    $GC_TEST(0, ERL_SUB_BIN_SIZE, $Live);
+    _mb = ms_matchbuffer($Ms);
+    if (((_mb->size - _mb->offset) % $Unit) == 0) {
+        LIGHT_SWAPOUT;
+        _result = erts_bs_get_binary_all_2(c_p, _mb);
+        LIGHT_SWAPIN;
+        HEAP_SPACE_VERIFIED(0);
+        ASSERT(is_value(_result));
+        $Dst = _result;
+    } else {
+	HEAP_SPACE_VERIFIED(0);
+	$FAIL($Fail);
+    }
+}
+
+i_bs_get_binary2(Fail, Ms, Live, Sz, Flags, Dst) {
+    ErlBinMatchBuffer *_mb;
+    Eterm _result;
+    Uint _size;
+    $BS_GET_FIELD_SIZE($Sz, (($Flags) >> 3), $FAIL($Fail), _size);
+    $GC_TEST(0, ERL_SUB_BIN_SIZE, $Live);
+    _mb = ms_matchbuffer($Ms);
+    LIGHT_SWAPOUT;
+    _result = erts_bs_get_binary_2(c_p, _size, $Flags, _mb);
+    LIGHT_SWAPIN;
+    HEAP_SPACE_VERIFIED(0);
+    if (is_non_value(_result)) {
+        $FAIL($Fail);
+    } else {
+        $Dst = _result;
+    }
+}
+
+i_bs_get_binary_imm2(Fail, Ms, Live, Sz, Flags, Dst) {
+    ErlBinMatchBuffer *_mb;
+    Eterm _result;
+    $GC_TEST(0, heap_bin_size(ERL_ONHEAP_BIN_LIMIT), $Live);
+    _mb = ms_matchbuffer($Ms);
+    LIGHT_SWAPOUT;
+    _result = erts_bs_get_binary_2(c_p, $Sz, $Flags, _mb);
+    LIGHT_SWAPIN;
+    HEAP_SPACE_VERIFIED(0);
+    if (is_non_value(_result)) {
+        $FAIL($Fail);
+    } else {
+        $Dst = _result;
+    }
+}
+
+i_bs_get_float2(Fail, Ms, Live, Sz, Flags, Dst) {
+    ErlBinMatchBuffer *_mb;
+    Eterm _result;
+    Sint _size;
+
+    if (!is_small($Sz) || (_size = unsigned_val($Sz)) > 64) {
+        $FAIL($Fail);
+    }
+    _size *= (($Flags) >> 3);
+    $GC_TEST(0, FLOAT_SIZE_OBJECT, $Live);
+    _mb = ms_matchbuffer($Ms);
+    LIGHT_SWAPOUT;
+    _result = erts_bs_get_float_2(c_p, _size, ($Flags), _mb);
+    LIGHT_SWAPIN;
+    HEAP_SPACE_VERIFIED(0);
+    if (is_non_value(_result)) {
+        $FAIL($Fail);
+    } else {
+        $Dst = _result;
+    }
+}
+
+i_bs_skip_bits2(Fail, Ms, Bits, Unit) {
+    ErlBinMatchBuffer *_mb;
+    size_t new_offset;
+    Uint _size;
+
+    _mb = ms_matchbuffer($Ms);
+    $BS_GET_FIELD_SIZE($Bits, $Unit, $FAIL($Fail), _size);
+    new_offset = _mb->offset + _size;
+    if (new_offset <= _mb->size) {
+        _mb->offset = new_offset;
+    } else {
+        $FAIL($Fail);
+    }
+}
+
+i_bs_skip_bits_all2(Fail, Ms, Unit) {
+    ErlBinMatchBuffer *_mb;
+    _mb = ms_matchbuffer($Ms);
+    if (((_mb->size - _mb->offset) % $Unit) == 0) {
+        _mb->offset = _mb->size;
+    } else {
+        $FAIL($Fail);
+    }
+}
+
+i_bs_skip_bits_imm2(Fail, Ms, Bits) {
+    ErlBinMatchBuffer *_mb;
+    size_t new_offset;
+    _mb = ms_matchbuffer($Ms);
+    new_offset = _mb->offset + ($Bits);
+    if (new_offset <= _mb->size) {
+        _mb->offset = new_offset;
+    } else {
+        $FAIL($Fail);
+    }
+}
+
+i_new_bs_put_binary(Fail, Sz, Flags, Src) {
+    Sint _size;
+    $BS_GET_UNCHECKED_FIELD_SIZE($Sz, (($Flags) >> 3), $BADARG($Fail), _size);
+    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2(($Src), _size))) {
+        $BADARG($Fail);
+    }
+}
+
+i_new_bs_put_binary_all(Fail, Src, Unit) {
+    if (!erts_new_bs_put_binary_all(ERL_BITS_ARGS_2(($Src), ($Unit)))) {
+        $BADARG($Fail);
+    }
+}
+
+i_new_bs_put_binary_imm(Fail, Sz, Src) {
+    if (!erts_new_bs_put_binary(ERL_BITS_ARGS_2(($Src), ($Sz)))) {
+        $BADARG($Fail);
+    }
+}
+
+i_new_bs_put_float(Fail, Sz, Flags, Src) {
+    Sint _size;
+    $BS_GET_UNCHECKED_FIELD_SIZE($Sz, (($Flags) >> 3), $BADARG($Fail), _size);
+    if (!erts_new_bs_put_float(c_p, ($Src), _size, ($Flags))) {
+        $BADARG($Fail);
+    }
+}
+
+i_new_bs_put_float_imm(Fail, Sz, Flags, Src) {
+    if (!erts_new_bs_put_float(c_p, ($Src), ($Sz), ($Flags))) {
+        $BADARG($Fail);
+    }
+}
+
+i_new_bs_put_integer(Fail, Sz, Flags, Src) {
+   Sint _size;
+   $BS_GET_UNCHECKED_FIELD_SIZE($Sz, (($Flags) >> 3), $BADARG($Fail), _size);
+   if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3(($Src), _size, ($Flags)))) {
+       $BADARG($Fail);
+   }
+}
+
+i_new_bs_put_integer_imm(Fail, Sz, Flags, Src) {
+    if (!erts_new_bs_put_integer(ERL_BITS_ARGS_3(($Src), ($Sz), ($Flags)))) {
+        $BADARG($Fail);
+    }
+}
+
+#
+# i_bs_init*
+#
+
+i_bs_init_fail_heap := bs_init.fail_heap.verify.execute;
+i_bs_init_fail := bs_init.fail.verify.execute;
+i_bs_init := bs_init.plain.execute;
+i_bs_init_heap := bs_init.heap.execute;
+
+bs_init.head() {
+    Eterm BsOp1;
+    Eterm BsOp2;
+}
+
+bs_init.fail_heap(Size, HeapAlloc) {
+    BsOp1 = $Size;
+    BsOp2 = $HeapAlloc;
+}
+
+bs_init.fail(Size) {
+    BsOp1 = $Size;
+    BsOp2 = 0;
+}
+
+bs_init.plain(Size) {
+    BsOp1 = $Size;
+    BsOp2 = 0;
+}
+
+bs_init.heap(Size, HeapAlloc) {
+    BsOp1 = $Size;
+    BsOp2 = $HeapAlloc;
+}
+
+bs_init.verify(Fail) {
+    if (is_small(BsOp1)) {
+        Sint size = signed_val(BsOp1);
+        if (size < 0) {
+            $BADARG($Fail);
+        }
+        BsOp1 = (Eterm) size;
+    } else {
+        Uint bytes;
+
+        if (!term_to_Uint(BsOp1, &bytes)) {
+            c_p->freason = bytes;
+            $FAIL_HEAD_OR_BODY($Fail);
+        }
+        if ((bytes >> (8*sizeof(Uint)-3)) != 0) {
+            $SYSTEM_LIMIT($Fail);
+        }
+        BsOp1 = (Eterm) bytes;
+    }
+}
+
+bs_init.execute(Live, Dst) {
+    if (BsOp1 <= ERL_ONHEAP_BIN_LIMIT) {
+        ErlHeapBin* hb;
+        Uint bin_need;
+
+        bin_need = heap_bin_size(BsOp1);
+        erts_bin_offset = 0;
+        erts_writable_bin = 0;
+        $GC_TEST(0, bin_need+BsOp2+ERL_SUB_BIN_SIZE, $Live);
+        hb = (ErlHeapBin *) HTOP;
+        HTOP += bin_need;
+        hb->thing_word = header_heap_bin(BsOp1);
+        hb->size = BsOp1;
+        erts_current_bin = (byte *) hb->data;
+        $Dst = make_binary(hb);
+    } else {
+        Binary* bptr;
+        ProcBin* pb;
+
+        erts_bin_offset = 0;
+        erts_writable_bin = 0;
+        $TEST_BIN_VHEAP(BsOp1 / sizeof(Eterm),
+                        BsOp2 + PROC_BIN_SIZE + ERL_SUB_BIN_SIZE, $Live);
+
+        /*
+         * Allocate the binary struct itself.
+         */
+        bptr = erts_bin_nrml_alloc(BsOp1);
+        erts_current_bin = (byte *) bptr->orig_bytes;
+
+        /*
+         * Now allocate the ProcBin on the heap.
+         */
+        pb = (ProcBin *) HTOP;
+        HTOP += PROC_BIN_SIZE;
+        pb->thing_word = HEADER_PROC_BIN;
+        pb->size = BsOp1;
+        pb->next = MSO(c_p).first;
+        MSO(c_p).first = (struct erl_off_heap_header*) pb;
+        pb->val = bptr;
+        pb->bytes = (byte*) bptr->orig_bytes;
+        pb->flags = 0;
+
+        OH_OVERHEAD(&(MSO(c_p)), BsOp1 / sizeof(Eterm));
+
+        $Dst = make_binary(pb);
+    }
+}
+
+#
+# i_bs_init_bits*
+#
+
+i_bs_init_bits           := bs_init_bits.plain.execute;
+i_bs_init_bits_heap      := bs_init_bits.heap.execute;
+i_bs_init_bits_fail      := bs_init_bits.fail.verify.execute;
+i_bs_init_bits_fail_heap := bs_init_bits.fail_heap.verify.execute;
+
+bs_init_bits.head() {
+    Eterm num_bits_term;
+    Uint num_bits;
+    Uint alloc;
+}
+
+bs_init_bits.plain(NumBits) {
+    num_bits = $NumBits;
+    alloc = 0;
+}
+
+bs_init_bits.heap(NumBits, Alloc) {
+    num_bits = $NumBits;
+    alloc = $Alloc;
+}
+
+bs_init_bits.fail(NumBitsTerm) {
+    num_bits_term = $NumBitsTerm;
+    alloc = 0;
+}
+
+bs_init_bits.fail_heap(NumBitsTerm, Alloc) {
+    num_bits_term = $NumBitsTerm;
+    alloc = $Alloc;
+}
+
+bs_init_bits.verify(Fail) {
+    if (is_small(num_bits_term)) {
+        Sint size = signed_val(num_bits_term);
+        if (size < 0) {
+            $BADARG($Fail);
+        }
+        num_bits = (Uint) size;
+    } else {
+        Uint bits;
+
+        if (!term_to_Uint(num_bits_term, &bits)) {
+            c_p->freason = bits;
+            $FAIL_HEAD_OR_BODY($Fail);
+        }
+        num_bits = (Uint) bits;
+    }
+}
+
+bs_init_bits.execute(Live, Dst) {
+     Eterm new_binary;
+     Uint num_bytes = ((Uint64)num_bits+(Uint64)7) >> 3;
+
+     if (num_bits & 7) {
+	 alloc += ERL_SUB_BIN_SIZE;
+     }
+     if (num_bytes <= ERL_ONHEAP_BIN_LIMIT) {
+	 alloc += heap_bin_size(num_bytes);
+     } else {
+	 alloc += PROC_BIN_SIZE;
+     }
+     $test_heap(alloc, $Live);
+
+     /* num_bits = Number of bits to build
+      * num_bytes = Number of bytes to allocate in the binary
+      * alloc = Total number of words to allocate on heap
+      * Operands: NotUsed NotUsed Dst
+      */
+     if (num_bytes <= ERL_ONHEAP_BIN_LIMIT) {
+	 ErlHeapBin* hb;
+
+	 erts_bin_offset = 0;
+	 erts_writable_bin = 0;
+	 hb = (ErlHeapBin *) HTOP;
+	 HTOP += heap_bin_size(num_bytes);
+	 hb->thing_word = header_heap_bin(num_bytes);
+	 hb->size = num_bytes;
+	 erts_current_bin = (byte *) hb->data;
+	 new_binary = make_binary(hb);
+
+     do_bits_sub_bin:
+	 if (num_bits & 7) {
+	     ErlSubBin* sb;
+
+	     sb = (ErlSubBin *) HTOP;
+	     HTOP += ERL_SUB_BIN_SIZE;
+	     sb->thing_word = HEADER_SUB_BIN;
+	     sb->size = num_bytes - 1;
+	     sb->bitsize = num_bits & 7;
+	     sb->offs = 0;
+	     sb->bitoffs = 0;
+	     sb->is_writable = 0;
+	     sb->orig = new_binary;
+	     new_binary = make_binary(sb);
+	 }
+	 HEAP_SPACE_VERIFIED(0);
+         $Dst = new_binary;
+     } else {
+	 Binary* bptr;
+	 ProcBin* pb;
+
+	 erts_bin_offset = 0;
+	 erts_writable_bin = 0;
+
+	 /*
+	  * Allocate the binary struct itself.
+	  */
+	 bptr = erts_bin_nrml_alloc(num_bytes);
+	 erts_current_bin = (byte *) bptr->orig_bytes;
+
+	 /*
+	  * Now allocate the ProcBin on the heap.
+	  */
+	 pb = (ProcBin *) HTOP;
+	 HTOP += PROC_BIN_SIZE;
+	 pb->thing_word = HEADER_PROC_BIN;
+	 pb->size = num_bytes;
+	 pb->next = MSO(c_p).first;
+	 MSO(c_p).first = (struct erl_off_heap_header*) pb;
+	 pb->val = bptr;
+	 pb->bytes = (byte*) bptr->orig_bytes;
+	 pb->flags = 0;
+	 OH_OVERHEAD(&(MSO(c_p)), pb->size / sizeof(Eterm));
+	 new_binary = make_binary(pb);
+	 goto do_bits_sub_bin;
+     }
+}
+
+bs_add(Fail, Src1, Src2, Unit, Dst) {
+    Eterm Op1 = $Src1;
+    Eterm Op2 = $Src2;
+    Uint unit = $Unit;
+
+    if (is_both_small(Op1, Op2)) {
+        Sint Arg1 = signed_val(Op1);
+        Sint Arg2 = signed_val(Op2);
+
+        if (Arg1 >= 0 && Arg2 >= 0) {
+            $BS_SAFE_MUL(Arg2, unit, $SYSTEM_LIMIT($Fail), Op1);
+            Op1 += Arg1;
+
+        store_bs_add_result:
+            if (Op1 <= MAX_SMALL) {
+                Op1 = make_small(Op1);
+            } else {
+                /*
+                 * May generate a heap fragment, but in this
+                 * particular case it is OK, since the value will be
+                 * stored into an x register (the GC will scan x
+                 * registers for references to heap fragments) and
+                 * there is no risk that value can be stored into a
+                 * location that is not scanned for heap-fragment
+                 * references (such as the heap).
+                 */
+                SWAPOUT;
+                Op1 = erts_make_integer(Op1, c_p);
+                HTOP = HEAP_TOP(c_p);
+            }
+            $Dst = Op1;
+            $NEXT0();
+        }
+        $BADARG($Fail);
+    } else {
+        Uint a;
+        Uint b;
+        Uint c;
+
+        /*
+         * Now we know that one of the arguments is
+         * not a small. We must convert both arguments
+         * to Uints and check for errors at the same time.
+         *
+         * Error checking is tricky.
+         *
+         * If one of the arguments is not numeric or
+         * not positive, the error reason is BADARG.
+         *
+         * Otherwise if both arguments are numeric,
+         * but at least one argument does not fit in
+         * an Uint, the reason is SYSTEM_LIMIT.
+         */
+
+        if (!term_to_Uint(Op1, &a)) {
+            if (a == BADARG) {
+                $BADARG($Fail);
+            }
+            if (!term_to_Uint(Op2, &b)) {
+                c_p->freason = b;
+                $FAIL_HEAD_OR_BODY($Fail);
+            }
+            $SYSTEM_LIMIT($Fail);
+        } else if (!term_to_Uint(Op2, &b)) {
+            c_p->freason = b;
+            $FAIL_HEAD_OR_BODY($Fail);
+        }
+
+        /*
+         * The arguments are now correct and stored in a and b.
+         */
+
+        $BS_SAFE_MUL(b, unit, $SYSTEM_LIMIT($Fail), c);
+        Op1 = a + c;
+        if (Op1 < a) {
+            /*
+             * If the result is less than one of the
+             * arguments, there must have been an overflow.
+             */
+            $SYSTEM_LIMIT($Fail);
+        }
+        goto store_bs_add_result;
+    }
+    /* No fallthrough */
+    ASSERT(0);
+}
+
+bs_put_string(Len, Ptr) {
+    erts_new_bs_put_string(ERL_BITS_ARGS_2((byte *) $Ptr, $Len));
+}
+
+i_bs_append(Fail, ExtraHeap, Live, Unit, Size, Dst) {
+    Uint live = $Live;
+    Uint res;
+
+    HEAVY_SWAPOUT;
+    reg[live] = x(SCRATCH_X_REG);
+    res = erts_bs_append(c_p, reg, live, $Size, $ExtraHeap, $Unit);
+    HEAVY_SWAPIN;
+    if (is_non_value(res)) {
+        /* c_p->freason is already set (to BADARG or SYSTEM_LIMIT). */
+        $FAIL_HEAD_OR_BODY($Fail);
+    }
+    $Dst = res;
+}
+
+i_bs_private_append(Fail, Unit, Size, Src, Dst) {
+    Eterm res;
+
+    res = erts_bs_private_append(c_p, $Src, $Size, $Unit);
+    if (is_non_value(res)) {
+        /* c_p->freason is already set (to BADARG or SYSTEM_LIMIT). */
+        $FAIL_HEAD_OR_BODY($Fail);
+    }
+    $Dst = res;
+}
+
+bs_init_writable() {
+    HEAVY_SWAPOUT;
+    r(0) = erts_bs_init_writable(c_p, r(0));
+    HEAVY_SWAPIN;
+}
+
+i_bs_utf8_size(Src, Dst) {
+    Eterm arg = $Src;
+    Eterm result;
+
+    /*
+     * Calculate the number of bytes needed to encode the source
+     * operand to UTF-8. If the source operand is invalid (e.g. wrong
+     * type or range) we return a nonsense integer result (0 or 4). We
+     * can get away with that because we KNOW that bs_put_utf8 will do
+     * full error checking.
+     */
+
+    if (arg < make_small(0x80UL)) {
+        result = make_small(1);
+    } else if (arg < make_small(0x800UL)) {
+        result = make_small(2);
+    } else if (arg < make_small(0x10000UL)) {
+        result = make_small(3);
+    } else {
+        result = make_small(4);
+    }
+    $Dst = result;
+}
+
+i_bs_put_utf8(Fail, Src) {
+    if (!erts_bs_put_utf8(ERL_BITS_ARGS_1($Src))) {
+        $BADARG($Fail);
+    }
+}
+
+i_bs_utf16_size(Src, Dst) {
+    Eterm arg = $Src;
+    Eterm result = make_small(2);
+
+    /*
+     * Calculate the number of bytes needed to encode the source
+     * operarand to UTF-16. If the source operand is invalid (e.g. wrong
+     * type or range) we return a nonsense integer result (2 or 4). We
+     * can get away with that because we KNOW that bs_put_utf16 will do
+     * full error checking.
+     */
+
+    if (arg >= make_small(0x10000UL)) {
+        result = make_small(4);
+    }
+    $Dst = result;
+}
+
+bs_put_utf16(Fail, Flags, Src) {
+    if (!erts_bs_put_utf16(ERL_BITS_ARGS_2($Src, $Flags))) {
+        $BADARG($Fail);
+    }
+}
+
+// Validate a value about to be stored in a binary.
+i_bs_validate_unicode(Fail, Src) {
+    Eterm val = $Src;
+
+    /*
+     * There is no need to untag the integer, but it IS necessary
+     * to make sure it is small (if the term is a bignum, it could
+     * slip through the test, and there is no further test that
+     * would catch it, since bit syntax construction silently masks
+     * too big numbers).
+     */
+    if (is_not_small(val) || val > make_small(0x10FFFFUL) ||
+        (make_small(0xD800UL) <= val && val <= make_small(0xDFFFUL))) {
+        $BADARG($Fail);
+    }
+}
+
+// Validate a value that has been matched out.
+i_bs_validate_unicode_retract(Fail, Src, Ms) {
+    /*
+     * There is no need to untag the integer, but it IS necessary
+     * to make sure it is small (a bignum pointer could fall in
+     * the valid range).
+     */
+
+    Eterm i = $Src;
+    if (is_not_small(i) || i > make_small(0x10FFFFUL) ||
+        (make_small(0xD800UL) <= i && i <= make_small(0xDFFFUL))) {
+        Eterm ms = $Ms;		/* Match context */
+        ErlBinMatchBuffer* mb;
+
+        /* Invalid value. Retract the position in the binary. */
+        mb = ms_matchbuffer(ms);
+        mb->offset -= 32;
+        $BADARG($Fail);
+    }
+}
+
+
+//
+// Matching of binaries.
+//
+
+i_bs_start_match2 := bs_start_match.fetch.execute;
+
+bs_start_match.head() {
+    Eterm context;
+}
+
+bs_start_match.fetch(Src) {
+    context = $Src;
+}
+
+bs_start_match.execute(Fail, Live, Slots, Dst) {
+    Uint slots;
+    Uint live;
+    Eterm header;
+    if (!is_boxed(context)) {
+        $FAIL($Fail);
+    }
+    header = *boxed_val(context);
+    slots = $Slots;
+    live = $Live;
+    if (header_is_bin_matchstate(header)) {
+        ErlBinMatchState* ms = (ErlBinMatchState *) boxed_val(context);
+        Uint actual_slots = HEADER_NUM_SLOTS(header);
+        ms->save_offset[0] = ms->mb.offset;
+        if (actual_slots < slots) {
+            ErlBinMatchState* dst;
+            Uint live = $Live;
+            Uint wordsneeded = ERL_BIN_MATCHSTATE_SIZE(slots);
+
+            $GC_TEST_PRESERVE(wordsneeded, live, context);
+            ms = (ErlBinMatchState *) boxed_val(context);
+            dst = (ErlBinMatchState *) HTOP;
+            *dst = *ms;
+            *HTOP = HEADER_BIN_MATCHSTATE(slots);
+            HTOP += wordsneeded;
+            HEAP_SPACE_VERIFIED(0);
+            $Dst = make_matchstate(dst);
+        }
+    } else if (is_binary_header(header)) {
+        Eterm result;
+        Uint wordsneeded = ERL_BIN_MATCHSTATE_SIZE(slots);
+        $GC_TEST_PRESERVE(wordsneeded, live, context);
+        HEAP_TOP(c_p) = HTOP;
+#ifdef DEBUG
+        c_p->stop = E;	/* Needed for checking in HeapOnlyAlloc(). */
+#endif
+        result = erts_bs_start_match_2(c_p, context, slots);
+        HTOP = HEAP_TOP(c_p);
+        HEAP_SPACE_VERIFIED(0);
+        if (is_non_value(result)) {
+            $FAIL($Fail);
+        }
+        $Dst = result;
+    } else {
+        $FAIL($Fail);
+    }
+}
+
+bs_test_zero_tail2(Fail, Ctx) {
+    ErlBinMatchBuffer *_mb;
+    _mb = (ErlBinMatchBuffer*) ms_matchbuffer($Ctx);
+    if (_mb->size != _mb->offset) {
+        $FAIL($Fail);
+    }
+}
+
+bs_test_tail_imm2(Fail, Ctx, Offset) {
+    ErlBinMatchBuffer *_mb;
+    _mb = ms_matchbuffer($Ctx);
+    if (_mb->size - _mb->offset != $Offset) {
+        $FAIL($Fail);
+    }
+}
+
+bs_test_unit(Fail, Ctx, Unit) {
+    ErlBinMatchBuffer *_mb;
+    _mb = ms_matchbuffer($Ctx);
+    if ((_mb->size - _mb->offset) % $Unit) {
+        $FAIL($Fail);
+    }
+}
+
+bs_test_unit8(Fail, Ctx) {
+    ErlBinMatchBuffer *_mb;
+    _mb = ms_matchbuffer($Ctx);
+    if ((_mb->size - _mb->offset) & 7) {
+        $FAIL($Fail);
+    }
+}
+
+i_bs_get_integer_8(Ctx, Fail, Dst) {
+    Eterm _result;
+    ErlBinMatchBuffer* _mb = ms_matchbuffer($Ctx);
+
+    if (_mb->size - _mb->offset < 8) {
+        $FAIL($Fail);
+    }
+    if (BIT_OFFSET(_mb->offset) != 0) {
+        _result = erts_bs_get_integer_2(c_p, 8, 0, _mb);
+    } else {
+        _result = make_small(_mb->base[BYTE_OFFSET(_mb->offset)]);
+        _mb->offset += 8;
+    }
+    $Dst = _result;
+}
+
+i_bs_get_integer_16(Ctx, Fail, Dst) {
+    Eterm _result;
+    ErlBinMatchBuffer* _mb = ms_matchbuffer($Ctx);
+
+    if (_mb->size - _mb->offset < 16) {
+        $FAIL($Fail);
+    }
+    if (BIT_OFFSET(_mb->offset) != 0) {
+        _result = erts_bs_get_integer_2(c_p, 16, 0, _mb);
+    } else {
+        _result = make_small(get_int16(_mb->base+BYTE_OFFSET(_mb->offset)));
+        _mb->offset += 16;
+    }
+    $Dst = _result;
+}
+
+%if ARCH_64
+i_bs_get_integer_32(Ctx, Fail, Dst) {
+    Uint32 _integer;
+    ErlBinMatchBuffer* _mb = ms_matchbuffer($Ctx);
+
+    if (_mb->size - _mb->offset < 32) {
+        $FAIL($Fail);
+    }
+    if (BIT_OFFSET(_mb->offset) != 0) {
+        _integer = erts_bs_get_unaligned_uint32(_mb);
+    } else {
+        _integer = get_int32(_mb->base + _mb->offset/8);
+    }
+    _mb->offset += 32;
+    $Dst = make_small(_integer);
+}
+%endif
+
+i_bs_get_integer_imm := bs_get_integer.fetch.execute;
+i_bs_get_integer_small_imm := bs_get_integer.fetch_small.execute;
+
+bs_get_integer.head() {
+    Eterm Ms, Sz;
+}
+
+bs_get_integer.fetch(Ctx, Size, Live) {
+    Uint wordsneeded;
+    Ms = $Ctx;
+    Sz = $Size;
+    wordsneeded = 1+WSIZE(NBYTES(Sz));
+    $GC_TEST_PRESERVE(wordsneeded, $Live, Ms);
+}
+
+bs_get_integer.fetch_small(Ctx, Size) {
+    Ms = $Ctx;
+    Sz = $Size;
+}
+
+bs_get_integer.execute(Fail, Flags, Dst) {
+    ErlBinMatchBuffer* mb;
+    Eterm result;
+
+    mb = ms_matchbuffer(Ms);
+    LIGHT_SWAPOUT;
+    result = erts_bs_get_integer_2(c_p, Sz, $Flags, mb);
+    LIGHT_SWAPIN;
+    HEAP_SPACE_VERIFIED(0);
+    if (is_non_value(result)) {
+        $FAIL($Fail);
+    }
+    $Dst = result;
+}
+
+i_bs_get_integer(Fail, Live, FlagsAndUnit, Ms, Sz, Dst) {
+    Uint flags;
+    Uint size;
+    Eterm ms;
+    ErlBinMatchBuffer* mb;
+    Eterm result;
+
+    flags = $FlagsAndUnit;
+    ms = $Ms;
+    $BS_GET_FIELD_SIZE($Sz, (flags >> 3), $FAIL($Fail), size);
+    if (size >= SMALL_BITS) {
+        Uint wordsneeded;
+        /* Check bits size before potential gc.
+         * We do not want a gc and then realize we don't need
+         * the allocated space (i.e. if the op fails).
+         *
+         * Remember to re-acquire the matchbuffer after gc.
+         */
+
+        mb = ms_matchbuffer(ms);
+        if (mb->size - mb->offset < size) {
+            $FAIL($Fail);
+        }
+        wordsneeded = 1+WSIZE(NBYTES((Uint) size));
+        $GC_TEST_PRESERVE(wordsneeded, $Live, ms);
+    }
+    mb = ms_matchbuffer(ms);
+    LIGHT_SWAPOUT;
+    result = erts_bs_get_integer_2(c_p, size, flags, mb);
+    LIGHT_SWAPIN;
+    HEAP_SPACE_VERIFIED(0);
+    if (is_non_value(result)) {
+        $FAIL($Fail);
+    }
+    $Dst = result;
+}
+
+i_bs_get_utf8(Ctx, Fail, Dst) {
+    Eterm result;
+    ErlBinMatchBuffer* mb = ms_matchbuffer($Ctx);
+
+    if (mb->size - mb->offset < 8) {
+        $FAIL($Fail);
+    }
+    if (BIT_OFFSET(mb->offset) != 0) {
+        result = erts_bs_get_utf8(mb);
+    } else {
+        byte b = mb->base[BYTE_OFFSET(mb->offset)];
+        if (b < 128) {
+            result = make_small(b);
+            mb->offset += 8;
+        } else {
+            result = erts_bs_get_utf8(mb);
+        }
+    }
+    if (is_non_value(result)) {
+        $FAIL($Fail);
+    }
+    $Dst = result;
+}
+
+i_bs_get_utf16(Ctx, Fail, Flags, Dst) {
+    ErlBinMatchBuffer* mb = ms_matchbuffer($Ctx);
+    Eterm result = erts_bs_get_utf16(mb, $Flags);
+
+    if (is_non_value(result)) {
+        $FAIL($Fail);
+    }
+    $Dst = result;
+}
+
+bs_context_to_binary := ctx_to_bin.fetch.execute;
+i_bs_get_binary_all_reuse := ctx_to_bin.fetch_bin.execute;
+
+ctx_to_bin.head() {
+    Eterm context;
+    ErlBinMatchBuffer* mb;
+    Uint size;
+    Uint offs;
+}
+
+ctx_to_bin.fetch(Src) {
+    context = $Src;
+    if (is_boxed(context) &&
+        header_is_bin_matchstate(*boxed_val(context))) {
+        ErlBinMatchState* ms;
+        ms = (ErlBinMatchState *) boxed_val(context);
+        mb = &ms->mb;
+        offs = ms->save_offset[0];
+        size = mb->size - offs;
+    } else {
+        $NEXT0();
+    }
+}
+
+ctx_to_bin.fetch_bin(Src, Fail, Unit) {
+    context = $Src;
+    mb = ms_matchbuffer(context);
+    size = mb->size - mb->offset;
+    if (size % $Unit != 0) {
+        $FAIL($Fail);
+    }
+    offs = mb->offset;
+}
+
+ctx_to_bin.execute() {
+    Uint hole_size;
+    Uint orig = mb->orig;
+    ErlSubBin* sb = (ErlSubBin *) boxed_val(context);
+    /* Since we're going to overwrite the match state with the result, an
+     * ErlBinMatchState must be at least as large as an ErlSubBin. */
+    ERTS_CT_ASSERT(sizeof(ErlSubBin) <= sizeof(ErlBinMatchState));
+    hole_size = 1 + header_arity(sb->thing_word) - ERL_SUB_BIN_SIZE;
+    sb->thing_word = HEADER_SUB_BIN;
+    sb->size = BYTE_OFFSET(size);
+    sb->bitsize = BIT_OFFSET(size);
+    sb->offs = BYTE_OFFSET(offs);
+    sb->bitoffs = BIT_OFFSET(offs);
+    sb->is_writable = 0;
+    sb->orig = orig;
+    if (hole_size) {
+        sb[1].thing_word = make_pos_bignum_header(hole_size-1);
+    }
+}
+
+i_bs_match_string(Ctx, Fail, Bits, Ptr) {
+    byte* bytes = (byte *) $Ptr;
+    Uint bits = $Bits;
+    ErlBinMatchBuffer* mb;
+    Uint offs;
+
+    mb = ms_matchbuffer($Ctx);
+    if (mb->size - mb->offset < bits) {
+        $FAIL($Fail);
+    }
+    offs = mb->offset & 7;
+    if (offs == 0 && (bits & 7) == 0) {
+        if (sys_memcmp(bytes, mb->base+(mb->offset>>3), bits>>3)) {
+            $FAIL($Fail);
+        }
+    } else if (erts_cmp_bits(bytes, 0, mb->base+(mb->offset>>3), mb->offset & 7, bits)) {
+        $FAIL($Fail);
+    }
+    mb->offset += bits;
+}
+
+i_bs_save2(Src, Slot) {
+    ErlBinMatchState* _ms = (ErlBinMatchState*) boxed_val((Eterm) $Src);
+    _ms->save_offset[$Slot] = _ms->mb.offset;
+}
+
+i_bs_restore2(Src, Slot) {
+    ErlBinMatchState* _ms = (ErlBinMatchState*) boxed_val((Eterm) $Src);
+    _ms->mb.offset = _ms->save_offset[$Slot];
+}
diff --git a/erts/emulator/beam/code_ix.c b/erts/emulator/beam/code_ix.c
index 8a3d1b20b4..34e46f5f33 100644
--- a/erts/emulator/beam/code_ix.c
+++ b/erts/emulator/beam/code_ix.c
@@ -34,8 +34,8 @@
 # define CIX_TRACE(text)
 #endif
 
-erts_smp_atomic32_t the_active_code_index;
-erts_smp_atomic32_t the_staging_code_index;
+erts_atomic32_t the_active_code_index;
+erts_atomic32_t the_staging_code_index;
 
 static Process* code_writing_process = NULL;
 struct code_write_queue_item {
@@ -43,7 +43,7 @@ struct code_write_queue_item {
     struct code_write_queue_item* next;
 };
 static struct code_write_queue_item* code_write_queue = NULL;
-static erts_smp_mtx_t code_write_permission_mtx;
+static erts_mtx_t code_write_permission_mtx;
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
 static erts_tsd_key_t has_code_write_permission;
@@ -55,9 +55,9 @@ void erts_code_ix_init(void)
      * single threaded with active and staging set both to zero.
      * Preloading is finished by a commit that will set things straight.
      */
-    erts_smp_atomic32_init_nob(&the_active_code_index, 0);
-    erts_smp_atomic32_init_nob(&the_staging_code_index, 0);
-    erts_smp_mtx_init(&code_write_permission_mtx, "code_write_permission", NIL,
+    erts_atomic32_init_nob(&the_active_code_index, 0);
+    erts_atomic32_init_nob(&the_staging_code_index, 0);
+    erts_mtx_init(&code_write_permission_mtx, "code_write_permission", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_tsd_key_create(&has_code_write_permission,
@@ -91,9 +91,9 @@ void erts_commit_staging_code_ix(void)
     /* We need to this lock as we are now making the staging export table active */
     export_staging_lock();
     ix = erts_staging_code_ix();
-    erts_smp_atomic32_set_nob(&the_active_code_index, ix);
+    erts_atomic32_set_nob(&the_active_code_index, ix);
     ix = (ix + 1) % ERTS_NUM_CODE_IX;
-    erts_smp_atomic32_set_nob(&the_staging_code_index, ix);
+    erts_atomic32_set_nob(&the_staging_code_index, ix);
     export_staging_unlock();
     erts_tracer_nif_clear();
     CIX_TRACE("activate");
@@ -115,12 +115,10 @@ void erts_abort_staging_code_ix(void)
 int erts_try_seize_code_write_permission(Process* c_p)
 {
     int success;
-#ifdef ERTS_SMP
-    ASSERT(!erts_smp_thr_progress_is_blocking()); /* to avoid deadlock */
-#endif
+    ASSERT(!erts_thr_progress_is_blocking()); /* to avoid deadlock */
     ASSERT(c_p != NULL);
 
-    erts_smp_mtx_lock(&code_write_permission_mtx);
+    erts_mtx_lock(&code_write_permission_mtx);
     success = (code_writing_process == NULL);
     if (success) {
 	code_writing_process = c_p;
@@ -138,21 +136,21 @@ int erts_try_seize_code_write_permission(Process* c_p)
 	code_write_queue = qitem;
 	erts_suspend(c_p, ERTS_PROC_LOCK_MAIN, NULL);
     }
-   erts_smp_mtx_unlock(&code_write_permission_mtx);
+   erts_mtx_unlock(&code_write_permission_mtx);
    return success;
 }
 
 void erts_release_code_write_permission(void)
 {
-    erts_smp_mtx_lock(&code_write_permission_mtx);
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    erts_mtx_lock(&code_write_permission_mtx);
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
     while (code_write_queue != NULL) { /* unleash the entire herd */
 	struct code_write_queue_item* qitem = code_write_queue;
-	erts_smp_proc_lock(qitem->p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(qitem->p, ERTS_PROC_LOCK_STATUS);
 	if (!ERTS_PROC_IS_EXITING(qitem->p)) {
 	    erts_resume(qitem->p, ERTS_PROC_LOCK_STATUS);
 	}
-	erts_smp_proc_unlock(qitem->p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(qitem->p, ERTS_PROC_LOCK_STATUS);
 	code_write_queue = qitem->next;
 	erts_proc_dec_refc(qitem->p);
 	erts_free(ERTS_ALC_T_CODE_IX_LOCK_Q, qitem);
@@ -161,7 +159,7 @@ void erts_release_code_write_permission(void)
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_tsd_set(has_code_write_permission, (void *) 0);
 #endif
-    erts_smp_mtx_unlock(&code_write_permission_mtx);
+    erts_mtx_unlock(&code_write_permission_mtx);
 }
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
diff --git a/erts/emulator/beam/code_ix.h b/erts/emulator/beam/code_ix.h
index a28b0cd36e..42976d2301 100644
--- a/erts/emulator/beam/code_ix.h
+++ b/erts/emulator/beam/code_ix.h
@@ -176,7 +176,7 @@ int erts_has_code_write_permission(void);
 ERTS_GLB_INLINE
 BeamInstr *erts_codeinfo_to_code(ErtsCodeInfo *ci)
 {
-    ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI) || !ci->op);
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI) || !ci->op);
     ASSERT_MFA(&ci->mfa);
     return (BeamInstr*)(ci + 1);
 }
@@ -185,7 +185,7 @@ ERTS_GLB_INLINE
 ErtsCodeInfo *erts_code_to_codeinfo(BeamInstr *I)
 {
     ErtsCodeInfo *ci = ((ErtsCodeInfo *)(((char *)(I)) - sizeof(ErtsCodeInfo)));
-    ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI) || !ci->op);
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI) || !ci->op);
     ASSERT_MFA(&ci->mfa);
     return ci;
 }
@@ -205,16 +205,16 @@ ErtsCodeMFA *erts_code_to_codemfa(BeamInstr *I)
     return mfa;
 }
 
-extern erts_smp_atomic32_t the_active_code_index;
-extern erts_smp_atomic32_t the_staging_code_index;
+extern erts_atomic32_t the_active_code_index;
+extern erts_atomic32_t the_staging_code_index;
 
 ERTS_GLB_INLINE ErtsCodeIndex erts_active_code_ix(void)
 {
-    return erts_smp_atomic32_read_nob(&the_active_code_index);
+    return erts_atomic32_read_nob(&the_active_code_index);
 }
 ERTS_GLB_INLINE ErtsCodeIndex erts_staging_code_ix(void)
 {
-    return erts_smp_atomic32_read_nob(&the_staging_code_index);
+    return erts_atomic32_read_nob(&the_staging_code_index);
 }
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
diff --git a/erts/emulator/beam/copy.c b/erts/emulator/beam/copy.c
index fefde256d7..7769a914db 100644
--- a/erts/emulator/beam/copy.c
+++ b/erts/emulator/beam/copy.c
@@ -611,7 +611,7 @@ Eterm copy_struct_x(Eterm obj, Uint sz, Eterm** hpp, ErlOffHeap* off_heap, Uint
     Eterm* htop;
     Eterm* hbot;
     Eterm* hp;
-    Eterm* objp;
+    Eterm* ERTS_RESTRICT objp;
     Eterm* tp;
     Eterm  res;
     Eterm  elem;
@@ -845,7 +845,7 @@ Eterm copy_struct_x(Eterm obj, Uint sz, Eterm** hpp, ErlOffHeap* off_heap, Uint
 		    funp = (ErlFunThing *) tp;
 		    funp->next = off_heap->first;
 		    off_heap->first = (struct erl_off_heap_header*) funp;
-		    erts_smp_refc_inc(&funp->fe->refc, 2);
+		    erts_refc_inc(&funp->fe->refc, 2);
 		    *argp = make_fun(tp);
 		}
 		break;
@@ -854,7 +854,7 @@ Eterm copy_struct_x(Eterm obj, Uint sz, Eterm** hpp, ErlOffHeap* off_heap, Uint
 	    case EXTERNAL_REF_SUBTAG:
 		{
 		  ExternalThing *etp = (ExternalThing *) objp;
-		  erts_smp_refc_inc(&etp->node->refc, 2);
+		  erts_refc_inc(&etp->node->refc, 2);
 		}
 	    L_off_heap_node_container_common:
 		{
@@ -1531,7 +1531,7 @@ Uint copy_shared_perform(Eterm obj, Uint size, erts_shcopy_t *info,
 		}
 		funp->next = off_heap->first;
 		off_heap->first = (struct erl_off_heap_header*) funp;
-		erts_smp_refc_inc(&funp->fe->refc, 2);
+		erts_refc_inc(&funp->fe->refc, 2);
 		goto cleanup_next;
 	    }
 	    case MAP_SUBTAG:
@@ -1658,7 +1658,7 @@ Uint copy_shared_perform(Eterm obj, Uint size, erts_shcopy_t *info,
 	    case EXTERNAL_REF_SUBTAG:
 	    {
 		ExternalThing *etp = (ExternalThing *) ptr;
-		erts_smp_refc_inc(&etp->node->refc, 2);
+		erts_refc_inc(&etp->node->refc, 2);
 	    }
 	  off_heap_node_container_common:
 	    {
@@ -1821,7 +1821,8 @@ all_clean:
  *
  * NOTE: Assumes that term is a tuple (ptr is an untagged tuple ptr).
  */
-Eterm copy_shallow(Eterm* ptr, Uint sz, Eterm** hpp, ErlOffHeap* off_heap)
+Eterm copy_shallow(Eterm* ERTS_RESTRICT ptr, Uint sz, Eterm** hpp,
+                   ErlOffHeap* off_heap)
 {
     Eterm* tp = ptr;
     Eterm* hp = *hpp;
@@ -1855,7 +1856,7 @@ Eterm copy_shallow(Eterm* ptr, Uint sz, Eterm** hpp, ErlOffHeap* off_heap)
 	    case FUN_SUBTAG:
 		{
 		    ErlFunThing* funp = (ErlFunThing *) (tp-1);
-		    erts_smp_refc_inc(&funp->fe->refc, 2);
+		    erts_refc_inc(&funp->fe->refc, 2);
 		}
 		goto off_heap_common;
 	    case EXTERNAL_PID_SUBTAG:
@@ -1863,7 +1864,7 @@ Eterm copy_shallow(Eterm* ptr, Uint sz, Eterm** hpp, ErlOffHeap* off_heap)
 	    case EXTERNAL_REF_SUBTAG:
 		{
 		    ExternalThing* etp = (ExternalThing *) (tp-1);
-		    erts_smp_refc_inc(&etp->node->refc, 2);
+		    erts_refc_inc(&etp->node->refc, 2);
 		}
 	    off_heap_common:
 		{
@@ -1985,7 +1986,7 @@ move_one_frag(Eterm** hpp, ErlHeapFragment* frag, ErlOffHeap* off_heap, int lite
 	if (is_header(val)) {
 	    struct erl_off_heap_header* hdr = (struct erl_off_heap_header*)hp;
 	    ASSERT(ptr + header_arity(val) < end);
-	    move_boxed(&ptr, val, &hp, &dummy_ref);
+	    ptr = move_boxed(ptr, val, &hp, &dummy_ref);
 	    switch (val & _HEADER_SUBTAG_MASK) {
 	    case REF_SUBTAG:
 		if (is_ordinary_ref_thing(hdr))
@@ -2002,7 +2003,7 @@ move_one_frag(Eterm** hpp, ErlHeapFragment* frag, ErlOffHeap* off_heap, int lite
 	}
 	else { /* must be a cons cell */
 	    ASSERT(ptr+1 < end);
-	    move_cons(&ptr, val, &hp, &dummy_ref);
+	    move_cons(ptr, val, &hp, &dummy_ref);
 	    ptr += 2;
 	}
     }
diff --git a/erts/emulator/beam/dist.c b/erts/emulator/beam/dist.c
index 9fddac8980..70474898b2 100644
--- a/erts/emulator/beam/dist.c
+++ b/erts/emulator/beam/dist.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -45,6 +45,7 @@
 #include "erl_binary.h"
 #include "erl_thr_progress.h"
 #include "dtrace-wrapper.h"
+#include "erl_proc_sig_queue.h"
 
 #define DIST_CTL_DEFAULT_SIZE 64
 
@@ -103,25 +104,15 @@ dist_msg_dbg(ErtsDistExternal *edep, char *what, byte *buf, int sz)
 
 
 
-#define PASS_THROUGH 'p'        /* This code should go */
-
 int erts_is_alive; /* System must be blocked on change */
 int erts_dist_buf_busy_limit;
 
 
 /* distribution trap functions */
-Export* dsend2_trap = NULL;
-Export* dsend3_trap = NULL;
-/*Export* dsend_nosuspend_trap = NULL;*/
-Export* dlink_trap = NULL;
-Export* dunlink_trap = NULL;
 Export* dmonitor_node_trap = NULL;
-Export* dgroup_leader_trap = NULL;
-Export* dexit_trap = NULL;
-Export* dmonitor_p_trap = NULL;
 
 /* local variables */
-
+static Export *dist_ctrl_put_data_trap;
 
 /* forward declarations */
 
@@ -129,9 +120,10 @@ static void clear_dist_entry(DistEntry*);
 static int dsig_send_ctl(ErtsDSigData* dsdp, Eterm ctl, int force_busy);
 static void send_nodes_mon_msgs(Process *, Eterm, Eterm, Eterm, Eterm);
 static void init_nodes_monitors(void);
+static Sint abort_connection(DistEntry* dep, Uint32 conn_id);
 
-static erts_smp_atomic_t no_caches;
-static erts_smp_atomic_t no_nodes;
+static erts_atomic_t no_caches;
+static erts_atomic_t no_nodes;
 
 struct {
     Eterm reason;
@@ -144,8 +136,8 @@ delete_cache(ErtsAtomCache *cache)
 {
     if (cache) {
 	erts_free(ERTS_ALC_T_DCACHE, (void *) cache);
-	ASSERT(erts_smp_atomic_read_nob(&no_caches) > 0);
-	erts_smp_atomic_dec_nob(&no_caches);
+	ASSERT(erts_atomic_read_nob(&no_caches) > 0);
+	erts_atomic_dec_nob(&no_caches);
     }
 }
 
@@ -156,14 +148,12 @@ create_cache(DistEntry *dep)
     int i;
     ErtsAtomCache *cp;
 
-    ERTS_SMP_LC_ASSERT(
-	is_internal_port(dep->cid)
-	&& erts_lc_is_port_locked(erts_port_lookup_raw(dep->cid)));
+    ERTS_LC_ASSERT(is_nil(dep->cid));
     ASSERT(!dep->cache);
 
     dep->cache = cp = (ErtsAtomCache*) erts_alloc(ERTS_ALC_T_DCACHE,
 						  sizeof(ErtsAtomCache));
-    erts_smp_atomic_inc_nob(&no_caches);
+    erts_atomic_inc_nob(&no_caches);
     for (i = 0; i < sizeof(cp->in_arr)/sizeof(cp->in_arr[0]); i++) {
 	cp->in_arr[i] = THE_NON_VALUE;
 	cp->out_arr[i] = THE_NON_VALUE;
@@ -172,15 +162,17 @@ create_cache(DistEntry *dep)
 
 Uint erts_dist_cache_size(void)
 {
-    return (Uint) erts_smp_atomic_read_mb(&no_caches)*sizeof(ErtsAtomCache);
+    return (Uint) erts_atomic_read_mb(&no_caches)*sizeof(ErtsAtomCache);
 }
 
 static ErtsProcList *
-get_suspended_on_de(DistEntry *dep, Uint32 unset_qflgs)
+get_suspended_on_de(DistEntry *dep, erts_aint32_t unset_qflgs)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&dep->qlock));
-    dep->qflgs &= ~unset_qflgs;
-    if (dep->qflgs & ERTS_DE_QFLG_EXIT) {
+    erts_aint32_t qflgs;
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
+    qflgs = erts_atomic32_read_band_acqb(&dep->qflgs, ~unset_qflgs);
+    qflgs &= ~unset_qflgs;
+    if (qflgs & ERTS_DE_QFLG_EXIT) {
 	/* No resume when exit has been scheduled */
 	return NULL;
     }
@@ -192,6 +184,161 @@ get_suspended_on_de(DistEntry *dep, Uint32 unset_qflgs)
     }
 }
 
+#define ERTS_MON_LNK_FIRE_LIMIT 100
+
+static void monitor_connection_down(ErtsMonitor *mon, void *unused)
+{
+    if (erts_monitor_is_origin(mon))
+        erts_proc_sig_send_demonitor(mon);
+    else
+        erts_proc_sig_send_monitor_down(mon, am_noconnection);
+}
+
+static void link_connection_down(ErtsLink *lnk, void *vdist)
+{
+    erts_proc_sig_send_link_exit(NULL, THE_NON_VALUE, lnk,
+                                 am_noconnection, NIL);
+}
+
+typedef enum {
+    ERTS_CML_CLEANUP_STATE_LINKS,
+    ERTS_CML_CLEANUP_STATE_MONITORS,
+    ERTS_CML_CLEANUP_STATE_ONAME_MONITORS,
+    ERTS_CML_CLEANUP_STATE_NODE_MONITORS
+} ErtsConMonLnkCleaupState;
+
+typedef struct {
+    ErtsConMonLnkCleaupState state;
+    ErtsMonLnkDist *dist;
+    void *yield_state;
+    int trigger_node_monitors;
+    Eterm nodename;
+    Eterm visability;
+    Eterm reason;
+    ErlOffHeap oh;
+    Eterm heap[1];
+} ErtsConMonLnkCleanup;
+
+static void
+con_monitor_link_cleanup(void *vcmlcp)
+{
+    ErtsConMonLnkCleanup *cmlcp = vcmlcp;
+    ErtsMonLnkDist *dist = cmlcp->dist;
+    ErtsSchedulerData *esdp;
+    int yield;
+
+    switch (cmlcp->state) {
+    case ERTS_CML_CLEANUP_STATE_LINKS:
+        yield = erts_link_list_foreach_delete_yielding(&dist->links,
+                                                       link_connection_down,
+                                                       NULL, &cmlcp->yield_state,
+                                                       ERTS_MON_LNK_FIRE_LIMIT);
+        if (yield)
+            break;
+
+        ASSERT(!cmlcp->yield_state);
+        cmlcp->state = ERTS_CML_CLEANUP_STATE_MONITORS;
+    case ERTS_CML_CLEANUP_STATE_MONITORS:
+        yield = erts_monitor_list_foreach_delete_yielding(&dist->monitors,
+                                                          monitor_connection_down,
+                                                          NULL, &cmlcp->yield_state,
+                                                          ERTS_MON_LNK_FIRE_LIMIT);
+        if (yield)
+            break;
+
+        ASSERT(!cmlcp->yield_state);
+        cmlcp->state = ERTS_CML_CLEANUP_STATE_ONAME_MONITORS;
+    case ERTS_CML_CLEANUP_STATE_ONAME_MONITORS:
+        yield = erts_monitor_tree_foreach_delete_yielding(&dist->orig_name_monitors,
+                                                          monitor_connection_down,
+                                                          NULL, &cmlcp->yield_state,
+                                                          ERTS_MON_LNK_FIRE_LIMIT/2);
+        if (yield)
+            break;
+
+        cmlcp->dist = NULL;
+        erts_mon_link_dist_dec_refc(dist);
+
+        ASSERT(!cmlcp->yield_state);
+        cmlcp->state = ERTS_CML_CLEANUP_STATE_NODE_MONITORS;
+    case ERTS_CML_CLEANUP_STATE_NODE_MONITORS:
+        if (cmlcp->trigger_node_monitors) {
+            send_nodes_mon_msgs(NULL,
+                                am_nodedown,
+                                cmlcp->nodename,
+                                cmlcp->visability,
+                                cmlcp->reason);
+        }
+        erts_cleanup_offheap(&cmlcp->oh);
+        erts_free(ERTS_ALC_T_CML_CLEANUP, vcmlcp);
+        return; /* done */
+    }
+
+    /* yield... */
+
+    esdp = erts_get_scheduler_data();
+    ASSERT(esdp && esdp->type == ERTS_SCHED_NORMAL);
+    erts_schedule_misc_aux_work((int) esdp->no,
+                                con_monitor_link_cleanup,
+                                (void *) cmlcp);
+}
+
+static void
+schedule_con_monitor_link_cleanup(ErtsMonLnkDist *dist,
+                                  Eterm nodename,
+                                  Eterm visability,
+                                  Eterm reason)
+{
+    if (dist || is_value(nodename)) {
+        ErtsSchedulerData *esdp;
+        ErtsConMonLnkCleanup *cmlcp;
+        Uint rsz, size;
+
+        size = sizeof(ErtsConMonLnkCleanup);
+
+        if (is_non_value(reason) || is_immed(reason)) {
+            rsz = 0;
+            size -= sizeof(Eterm);
+        }
+        else {
+            rsz = size_object(reason);
+            size += sizeof(Eterm) * (rsz - 1);
+        }
+
+        cmlcp = erts_alloc(ERTS_ALC_T_CML_CLEANUP, size);
+
+        ERTS_INIT_OFF_HEAP(&cmlcp->oh);
+
+        cmlcp->yield_state = NULL;
+        cmlcp->dist = dist;
+        if (!dist)
+            cmlcp->state = ERTS_CML_CLEANUP_STATE_NODE_MONITORS;
+        else {
+            cmlcp->state = ERTS_CML_CLEANUP_STATE_LINKS;
+            erts_mtx_lock(&dist->mtx);
+            ASSERT(dist->alive);
+            dist->alive = 0;
+            erts_mtx_unlock(&dist->mtx);
+        }
+
+        cmlcp->trigger_node_monitors = is_value(nodename);
+        cmlcp->nodename = nodename;
+        cmlcp->visability = visability;
+        if (rsz == 0)
+            cmlcp->reason = reason;
+        else {
+            Eterm *hp = &cmlcp->heap[0];
+            cmlcp->reason = copy_struct(reason, rsz, &hp, &cmlcp->oh);
+        }
+
+        esdp = erts_get_scheduler_data();
+        ASSERT(esdp && esdp->type == ERTS_SCHED_NORMAL);
+        erts_schedule_misc_aux_work((int) esdp->no,
+                                    con_monitor_link_cleanup,
+                                    (void *) cmlcp);
+    }
+}
+
 /*
 ** A full node name constists of a "n@h"
 **
@@ -243,188 +390,22 @@ int is_node_name_atom(Eterm a)
     return is_node_name((char*)atom_tab(i)->name, atom_tab(i)->len);
 }
 
-typedef struct {
-    DistEntry *dep;
-    Eterm *lhp;
-} NetExitsContext;
-
-/* 
-** This function is called when a distribution 
-** port or process terminates
-*/
-static void doit_monitor_net_exits(ErtsMonitor *mon, void *vnecp)
-{
-    Process *rp;
-    ErtsMonitor *rmon;
-    DistEntry *dep = ((NetExitsContext *) vnecp)->dep;
-    ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK;
-
-    rp = erts_pid2proc(NULL, 0, mon->u.pid, rp_locks);
-    if (!rp)
-	goto done;
-
-    if (mon->type == MON_ORIGIN) {
-	/* local pid is being monitored */
-	rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-	/* ASSERT(rmon != NULL); nope, can happen during process exit */
-	if (rmon != NULL) {
-	    erts_destroy_monitor(rmon);
-	}
-    } else {
-	DeclareTmpHeapNoproc(lhp,3);
-	Eterm watched;
-	UseTmpHeapNoproc(3);
-	ASSERT(mon->type == MON_TARGET);
-	rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-	/* ASSERT(rmon != NULL); can happen during process exit */
-	if (rmon != NULL) {
-            ASSERT(rmon->type == MON_ORIGIN);
-	    ASSERT(is_atom(rmon->name) || is_nil(rmon->name));
-	    watched = (is_atom(rmon->name)
-		       ? TUPLE2(lhp, rmon->name, dep->sysname)
-		       : rmon->u.pid);
-#ifdef ERTS_SMP
-	    rp_locks |= ERTS_PROC_LOCKS_MSG_SEND;
-	    erts_smp_proc_lock(rp, ERTS_PROC_LOCKS_MSG_SEND);
-#endif
-	    erts_queue_monitor_message(rp, &rp_locks, mon->ref, am_process, 
-				       watched, am_noconnection);
-	    erts_destroy_monitor(rmon);
-	}
-	UnUseTmpHeapNoproc(3);
-    }
-    erts_smp_proc_unlock(rp, rp_locks);
- done:
-    erts_destroy_monitor(mon);
-}
-	
-typedef struct {
-    NetExitsContext *necp;
-    ErtsLink *lnk;
-} LinkNetExitsContext;
-
-/* 
-** This is the function actually doing the job of sending exit messages
-** for links in a dist entry upon net_exit (the node goes down), NB,
-** only process links, not node monitors are handled here, 
-** they reside in a separate tree....
-*/
-static void doit_link_net_exits_sub(ErtsLink *sublnk, void *vlnecp)
-{
-    ErtsLink *lnk = ((LinkNetExitsContext *) vlnecp)->lnk; /* the local pid */
-    ErtsLink *rlnk;
-    Process *rp;
-
-    ASSERT(lnk->type == LINK_PID);
-    if (is_internal_pid(lnk->pid)) {
-	int xres;
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCKS_XSIG_SEND;
-
-	rp = erts_pid2proc(NULL, 0, lnk->pid, rp_locks);
-	if (!rp) {
-	    goto done;
-	}
-
-	rlnk = erts_remove_link(&ERTS_P_LINKS(rp), sublnk->pid);
-	xres = erts_send_exit_signal(NULL,
-				     sublnk->pid,
-				     rp,
-				     &rp_locks,
-				     am_noconnection,
-				     NIL,
-				     NULL,
-				     0);
-
-	if (rlnk) {
-	    erts_destroy_link(rlnk);
-	    if (xres >= 0 && IS_TRACED_FL(rp, F_TRACE_PROCS)) {
-		/* We didn't exit the process and it is traced */
-		trace_proc(NULL, 0, rp, am_getting_unlinked, sublnk->pid);
-	    }
-	}
-	erts_smp_proc_unlock(rp, rp_locks);
-    }
- done:
-    erts_destroy_link(sublnk);
-
-}
-    
-
-
-
-
-/* 
-** This function is called when a distribution 
-** port or process terminates, once for each link on the high level, 
-** it in turn traverses the link subtree for the specific link node...
-*/
-static void doit_link_net_exits(ErtsLink *lnk, void *vnecp)
-{
-    LinkNetExitsContext lnec = {(NetExitsContext *) vnecp, lnk};
-    ASSERT(lnk->type == LINK_PID);
-    erts_sweep_links(ERTS_LINK_ROOT(lnk), &doit_link_net_exits_sub, (void *) &lnec);
-#ifdef DEBUG
-    ERTS_LINK_ROOT(lnk) = NULL;
-#endif
-    erts_destroy_link(lnk);
-}
-
-
-static void doit_node_link_net_exits(ErtsLink *lnk, void *vnecp)
-{
-    DistEntry *dep = ((NetExitsContext *) vnecp)->dep;
-    Eterm name = dep->sysname;
-    Process *rp;
-    ErtsLink *rlnk;
-    Uint i,n;
-    ASSERT(lnk->type == LINK_NODE);
-    if (is_internal_pid(lnk->pid)) {
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK;
-	ErlOffHeap *ohp;
-	rp = erts_proc_lookup(lnk->pid);
-	if (!rp)
-	    goto done;
-	erts_smp_proc_lock(rp, rp_locks);
-        if (!ERTS_PROC_IS_EXITING(rp)) {
-            rlnk = erts_remove_link(&ERTS_P_LINKS(rp), name);
-            if (rlnk != NULL) {
-                ASSERT(is_atom(rlnk->pid) && (rlnk->type == LINK_NODE));
-                erts_destroy_link(rlnk);
-            }
-            n = ERTS_LINK_REFC(lnk);
-            for (i = 0; i < n; ++i) {
-                Eterm tup;
-                Eterm *hp;
-                ErtsMessage *msgp;
-
-                msgp = erts_alloc_message_heap(rp, &rp_locks,
-                                               3, &hp, &ohp);
-                tup = TUPLE2(hp, am_nodedown, name);
-                erts_queue_message(rp, rp_locks, msgp, tup, am_system);
-            }
-        }
-	erts_smp_proc_unlock(rp, rp_locks);
-    }
- done:
-    erts_destroy_link(lnk);
-}
-
 static void
 set_node_not_alive(void *unused)
 {
     ErlHeapFragment *bp;
     Eterm nodename = erts_this_dist_entry->sysname;
 
-    ASSERT(erts_smp_atomic_read_nob(&no_nodes) == 0);
+    ASSERT(erts_atomic_read_nob(&no_nodes) == 0);
 
-    erts_smp_thr_progress_block();
+    erts_thr_progress_block();
     erts_set_this_node(am_Noname, 0);
     erts_is_alive = 0;
     send_nodes_mon_msgs(NULL, am_nodedown, nodename, am_visible, nodedown.reason);
     nodedown.reason = NIL;
     bp = nodedown.bp;
     nodedown.bp = NULL;
-    erts_smp_thr_progress_unblock();
+    erts_thr_progress_unblock();
     if (bp)
 	free_message_buffer(bp);
 }
@@ -432,7 +413,7 @@ set_node_not_alive(void *unused)
 static ERTS_INLINE void
 dec_no_nodes(void)
 {
-    erts_aint_t no = erts_smp_atomic_dec_read_mb(&no_nodes);
+    erts_aint_t no = erts_atomic_dec_read_mb(&no_nodes);
     ASSERT(no >= 0);
     ASSERT(erts_get_scheduler_id()); /* Need to be a scheduler */
     if (no == 0)
@@ -445,12 +426,33 @@ static ERTS_INLINE void
 inc_no_nodes(void)
 {
 #ifdef DEBUG
-    erts_aint_t no = erts_smp_atomic_read_nob(&no_nodes);
+    erts_aint_t no = erts_atomic_read_nob(&no_nodes);
     ASSERT(erts_is_alive ? no > 0 : no == 0);
 #endif
-    erts_smp_atomic_inc_mb(&no_nodes);
+    erts_atomic_inc_mb(&no_nodes);
+}
+
+static void
+kill_dist_ctrl_proc(void *vpid)
+{
+    erts_proc_sig_send_exit(NULL, (Eterm) vpid, (Eterm) vpid,
+                            am_kill, NIL, 0);
+}
+
+static void
+schedule_kill_dist_ctrl_proc(Eterm pid)
+{
+    ErtsSchedulerData *esdp = erts_get_scheduler_data();
+    int sched_id = 1;
+    if (!esdp || ERTS_SCHEDULER_IS_DIRTY(esdp))
+        sched_id = 1;
+    else
+        sched_id = (int) esdp->no;
+    erts_schedule_misc_aux_work(sched_id,
+                                kill_dist_ctrl_proc,
+                                (void *) (UWord) pid);
 }
-	
+
 /*
  * proc is currently running or exiting process.
  */
@@ -460,58 +462,83 @@ int erts_do_net_exits(DistEntry *dep, Eterm reason)
 
     if (dep == erts_this_dist_entry) {  /* Net kernel has died (clean up!!) */
 	DistEntry *tdep;
-	int no_dist_port = 0;
+	int no_dist_ctrl;
+        int no_pending;
 	Eterm nd_reason = (reason == am_no_network
 			   ? am_no_network
 			   : am_net_kernel_terminated);
-	erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+        int i = 0;
+        Eterm *dist_ctrl;
+        DistEntry** pending;
+
+        ERTS_UNDEF(dist_ctrl, NULL);
+        ERTS_UNDEF(pending, NULL);
+
+	erts_rwmtx_rlock(&erts_dist_table_rwmtx);
 
-	for (tdep = erts_hidden_dist_entries; tdep; tdep = tdep->next)
-	    no_dist_port++;
-	for (tdep = erts_visible_dist_entries; tdep; tdep = tdep->next)
-	    no_dist_port++;
+        no_dist_ctrl = (erts_no_of_hidden_dist_entries +
+                        erts_no_of_visible_dist_entries);
+        no_pending = erts_no_of_pending_dist_entries;
 
 	/* KILL all port controllers */
-	if (no_dist_port == 0)
-	    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
-	else {
-	    Eterm def_buf[128];
-	    int i = 0;
-	    Eterm *dist_port;
-
-	    if (no_dist_port <= sizeof(def_buf)/sizeof(def_buf[0]))
-		dist_port = &def_buf[0];
-	    else
-		dist_port = erts_alloc(ERTS_ALC_T_TMP,
-				       sizeof(Eterm)*no_dist_port);
+	if (no_dist_ctrl) {
+            dist_ctrl = erts_alloc(ERTS_ALC_T_TMP,
+                                   sizeof(Eterm)*no_dist_ctrl);
 	    for (tdep = erts_hidden_dist_entries; tdep; tdep = tdep->next) {
-		ASSERT(is_internal_port(tdep->cid));
-		dist_port[i++] = tdep->cid;
+		ASSERT(is_internal_port(tdep->cid) || is_internal_pid(tdep->cid));
+                ASSERT(i < no_dist_ctrl);
+		dist_ctrl[i++] = tdep->cid;
 	    }
 	    for (tdep = erts_visible_dist_entries; tdep; tdep = tdep->next) {
-		ASSERT(is_internal_port(tdep->cid));
-		dist_port[i++] = tdep->cid;
+		ASSERT(is_internal_port(tdep->cid) || is_internal_pid(tdep->cid));
+                ASSERT(i < no_dist_ctrl);
+		dist_ctrl[i++] = tdep->cid;
 	    }
-	    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
-
-	    for (i = 0; i < no_dist_port; i++) {
-		Port *prt = erts_port_lookup(dist_port[i],
-					     ERTS_PORT_SFLGS_INVALID_LOOKUP);
-		if (!prt)
-		    continue;
-		ASSERT(erts_atomic32_read_nob(&prt->state)
-		       & ERTS_PORT_SFLG_DISTRIBUTION);
+            ASSERT(i == no_dist_ctrl);
+        }
+        if (no_pending) {
+            pending = erts_alloc(ERTS_ALC_T_TMP, sizeof(DistEntry*)*no_pending);
+            i = 0;
+            for (tdep = erts_pending_dist_entries; tdep; tdep = tdep->next) {
+                ASSERT(is_nil(tdep->cid));
+                ASSERT(i < no_pending);
+                pending[i++] = tdep;
+                erts_ref_dist_entry(tdep);
+            }
+            ASSERT(i == no_pending);
+        }
+        erts_rwmtx_runlock(&erts_dist_table_rwmtx);
+
+        if (no_dist_ctrl) {
+            for (i = 0; i < no_dist_ctrl; i++) {
+                if (is_internal_pid(dist_ctrl[i]))
+                    schedule_kill_dist_ctrl_proc(dist_ctrl[i]);
+                else {
+                    Port *prt = erts_port_lookup(dist_ctrl[i],
+                                                 ERTS_PORT_SFLGS_INVALID_LOOKUP);
+                    if (prt) {
+                        ASSERT(erts_atomic32_read_nob(&prt->state)
+                               & ERTS_PORT_SFLG_DISTRIBUTION);
+
+                        erts_port_exit(NULL, ERTS_PORT_SIG_FLG_FORCE_SCHED,
+                                       prt, dist_ctrl[i], nd_reason, NULL);
+                    }
+                }
+            }
+            erts_free(ERTS_ALC_T_TMP, dist_ctrl);
+        }
 
-		erts_port_exit(NULL, ERTS_PORT_SIG_FLG_FORCE_SCHED,
-			       prt, dist_port[i], nd_reason, NULL);
-	    }
+        if (no_pending) {
+            for (i = 0; i < no_pending; i++) {
+                abort_connection(pending[i], pending[i]->connection_id);
+                erts_deref_dist_entry(pending[i]);
+            }
+            erts_free(ERTS_ALC_T_TMP, pending);
+        }
 
-	    if (dist_port != &def_buf[0])
-		erts_free(ERTS_ALC_T_TMP, dist_port);
-	}
 
 	/*
-	 * When last dist port exits, node will be taken
+	 * When last dist ctrl exits, node will be taken
 	 * from alive to not alive.
 	 */
 	ASSERT(is_nil(nodedown.reason) && !nodedown.bp);
@@ -528,65 +555,53 @@ int erts_do_net_exits(DistEntry *dep, Eterm reason)
 					  &nodedown.bp->off_heap);
 	}
     }
-    else { /* Call from distribution port */
-	NetExitsContext nec = {dep};
-	ErtsLink *nlinks;
-	ErtsLink *node_links;
-	ErtsMonitor *monitors;
+    else { /* Call from distribution controller (port/process) */
+	ErtsMonLnkDist *mld;
 	Uint32 flags;
 
-	erts_smp_atomic_set_mb(&dep->dist_cmd_scheduled, 1);
-	erts_smp_de_rwlock(dep);
+	erts_atomic_set_mb(&dep->dist_cmd_scheduled, 1);
+	erts_de_rwlock(dep);
 
-	ERTS_SMP_LC_ASSERT(is_internal_port(dep->cid)
-			   && erts_lc_is_port_locked(erts_port_lookup_raw(dep->cid)));
+        if (is_internal_port(dep->cid)) {
+            ERTS_LC_ASSERT(erts_lc_is_port_locked(erts_port_lookup_raw(dep->cid)));
 
-	if (erts_port_task_is_scheduled(&dep->dist_cmd))
-	    erts_port_task_abort(&dep->dist_cmd);
+            if (erts_port_task_is_scheduled(&dep->dist_cmd))
+                erts_port_task_abort(&dep->dist_cmd);
+        }
 
-	if (dep->status & ERTS_DE_SFLG_EXITING) {
+	if (dep->state == ERTS_DE_STATE_EXITING) {
 #ifdef DEBUG
-	    erts_smp_mtx_lock(&dep->qlock);
-	    ASSERT(dep->qflgs & ERTS_DE_QFLG_EXIT);
-	    erts_smp_mtx_unlock(&dep->qlock);
+	    ASSERT(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT);
 #endif
 	}
 	else {
-	    dep->status |= ERTS_DE_SFLG_EXITING;
-	    erts_smp_mtx_lock(&dep->qlock);
-	    ASSERT(!(dep->qflgs & ERTS_DE_QFLG_EXIT));
-	    dep->qflgs |= ERTS_DE_QFLG_EXIT;
-	    erts_smp_mtx_unlock(&dep->qlock);
+	    dep->state = ERTS_DE_STATE_EXITING;
+	    erts_mtx_lock(&dep->qlock);
+	    ASSERT(!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT));
+	    erts_atomic32_read_bor_relb(&dep->qflgs, ERTS_DE_QFLG_EXIT);
+	    erts_mtx_unlock(&dep->qlock);
 	}
 
-	erts_smp_de_links_lock(dep);
-	monitors	= dep->monitors;
-        nlinks		= dep->nlinks;
-	node_links	= dep->node_links;
-	dep->monitors	= NULL;
-        dep->nlinks	= NULL;
-	dep->node_links	= NULL;
-	erts_smp_de_links_unlock(dep);
+        mld = dep->mld;
+        dep->mld = NULL;
 
 	nodename = dep->sysname;
 	flags = dep->flags;
 
 	erts_set_dist_entry_not_connected(dep);
 
-	erts_smp_de_rwunlock(dep);
+	erts_de_rwunlock(dep);
 
-	erts_sweep_monitors(monitors, &doit_monitor_net_exits, (void *) &nec);
-	erts_sweep_links(nlinks, &doit_link_net_exits, (void *) &nec);
-	erts_sweep_links(node_links, &doit_node_link_net_exits, (void *) &nec);
-
-	send_nodes_mon_msgs(NULL,
-			    am_nodedown,
-			    nodename,
-			    flags & DFLAG_PUBLISHED ? am_visible : am_hidden,
-			    reason == am_normal ? am_connection_closed : reason);
+        schedule_con_monitor_link_cleanup(mld,
+                                          nodename,
+                                          (flags & DFLAG_PUBLISHED
+                                           ? am_visible
+                                           : am_hidden),
+                                          (reason == am_normal
+                                           ? am_connection_closed
+                                           : reason));
 
 	clear_dist_entry(dep);
-
     }
 
     dec_no_nodes();
@@ -600,6 +615,14 @@ trap_function(Eterm func, int arity)
     return erts_export_put(am_erlang, func, arity);
 }
 
+/*
+ * Sync with dist_util.erl:
+ *
+ * -record(erts_dflags,
+ *         {default, mandatory, addable, rejectable, strict_order}).
+ */
+static Eterm erts_dflags_record;
+
 void init_dist(void)
 {
     init_nodes_monitors();
@@ -607,19 +630,24 @@ void init_dist(void)
     nodedown.reason = NIL;
     nodedown.bp = NULL;
 
-    erts_smp_atomic_init_nob(&no_nodes, 0);
-    erts_smp_atomic_init_nob(&no_caches, 0);
+    erts_atomic_init_nob(&no_nodes, 0);
+    erts_atomic_init_nob(&no_caches, 0);
 
     /* Lookup/Install all references to trap functions */
-    dsend2_trap = trap_function(am_dsend,2);
-    dsend3_trap = trap_function(am_dsend,3);
-    /*    dsend_nosuspend_trap = trap_function(am_dsend_nosuspend,2);*/
-    dlink_trap = trap_function(am_dlink,1);
-    dunlink_trap = trap_function(am_dunlink,1);
     dmonitor_node_trap = trap_function(am_dmonitor_node,3);
-    dgroup_leader_trap = trap_function(am_dgroup_leader,2);
-    dexit_trap = trap_function(am_dexit, 2);
-    dmonitor_p_trap = trap_function(am_dmonitor_p, 2);
+    dist_ctrl_put_data_trap = erts_export_put(am_erts_internal,
+                                              am_dist_ctrl_put_data,
+                                              2);
+    {
+        Eterm* hp = erts_alloc(ERTS_ALC_T_LITERAL, (1+6)*sizeof(Eterm));
+        erts_dflags_record = TUPLE6(hp, am_erts_dflags,
+                                    make_small(DFLAG_DIST_DEFAULT),
+                                    make_small(DFLAG_DIST_MANDATORY),
+                                    make_small(DFLAG_DIST_ADDABLE),
+                                    make_small(DFLAG_DIST_REJECTABLE),
+                                    make_small(DFLAG_DIST_STRICT_ORDER));
+        erts_set_literal_tag(&erts_dflags_record, hp, (1+6));
+    }
 }
 
 #define ErtsDistOutputBuf2Binary(OB) \
@@ -634,6 +662,7 @@ alloc_dist_obuf(Uint size)
     obuf = (ErtsDistOutputBuf *) &bin->orig_bytes[0];
 #ifdef DEBUG
     obuf->dbg_pattern = ERTS_DIST_OUTPUT_BUF_DBG_PATTERN;
+    obuf->alloc_endp = obuf->data + size;
     ASSERT(bin == ErtsDistOutputBuf2Binary(obuf));
 #endif
     return obuf;
@@ -654,26 +683,11 @@ size_obuf(ErtsDistOutputBuf *obuf)
     return bin->orig_size;
 }
 
-static void clear_dist_entry(DistEntry *dep)
+static ErtsDistOutputBuf* clear_de_out_queues(DistEntry* dep)
 {
-    Sint obufsize = 0;
-    ErtsAtomCache *cache;
-    ErtsProcList *suspendees;
     ErtsDistOutputBuf *obuf;
 
-    erts_smp_de_rwlock(dep);
-    cache = dep->cache;
-    dep->cache = NULL;
-
-#ifdef DEBUG
-    erts_smp_de_links_lock(dep);
-    ASSERT(!dep->nlinks);
-    ASSERT(!dep->node_links);
-    ASSERT(!dep->monitors);
-    erts_smp_de_links_unlock(dep);
-#endif
-
-    erts_smp_mtx_lock(&dep->qlock);
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&dep->qlock));
 
     if (!dep->out_queue.last)
 	obuf = dep->finalized_out_queue.first;
@@ -682,21 +696,24 @@ static void clear_dist_entry(DistEntry *dep)
 	obuf = dep->out_queue.first;
     }
 
+    if (dep->tmp_out_queue.first) {
+        dep->tmp_out_queue.last->next = obuf;
+        obuf = dep->tmp_out_queue.first;
+    }
+
     dep->out_queue.first = NULL;
     dep->out_queue.last = NULL;
+    dep->tmp_out_queue.first = NULL;
+    dep->tmp_out_queue.last = NULL;
     dep->finalized_out_queue.first = NULL;
     dep->finalized_out_queue.last = NULL;
-    dep->status = 0;
-    suspendees = get_suspended_on_de(dep, ERTS_DE_QFLGS_ALL);
 
-    erts_smp_mtx_unlock(&dep->qlock);
-    erts_smp_atomic_set_nob(&dep->dist_cmd_scheduled, 0);
-    dep->send = NULL;
-    erts_smp_de_rwunlock(dep);
-
-    erts_resume_processes(suspendees);
+    return obuf;
+}
 
-    delete_cache(cache);
+static void free_de_out_queues(DistEntry* dep, ErtsDistOutputBuf *obuf)
+{
+    Sint obufsize = 0;
 
     while (obuf) {
 	ErtsDistOutputBuf *fobuf;
@@ -707,13 +724,49 @@ static void clear_dist_entry(DistEntry *dep)
     }
 
     if (obufsize) {
-	erts_smp_mtx_lock(&dep->qlock);
-	ASSERT(dep->qsize >= obufsize);
-	dep->qsize -= obufsize;
-	erts_smp_mtx_unlock(&dep->qlock);
+	erts_mtx_lock(&dep->qlock);
+        ASSERT(erts_atomic_read_nob(&dep->qsize) >= obufsize);
+        erts_atomic_add_nob(&dep->qsize,
+                            (erts_aint_t) -obufsize);
+	erts_mtx_unlock(&dep->qlock);
     }
 }
 
+static void clear_dist_entry(DistEntry *dep)
+{
+    ErtsAtomCache *cache;
+    ErtsProcList *suspendees;
+    ErtsDistOutputBuf *obuf;
+
+    erts_de_rwlock(dep);
+    erts_atomic_set_nob(&dep->input_handler,
+                        (erts_aint_t) NIL);
+    cache = dep->cache;
+    dep->cache = NULL;
+
+    erts_mtx_lock(&dep->qlock);
+
+    erts_atomic64_set_nob(&dep->in, 0);
+    erts_atomic64_set_nob(&dep->out, 0);
+
+    obuf = clear_de_out_queues(dep);
+    dep->state = ERTS_DE_STATE_IDLE;
+    suspendees = get_suspended_on_de(dep, ERTS_DE_QFLGS_ALL);
+
+    erts_mtx_unlock(&dep->qlock);
+    erts_atomic_set_nob(&dep->dist_cmd_scheduled, 0);
+    dep->send = NULL;
+    erts_de_rwunlock(dep);
+
+    erts_resume_processes(suspendees);
+
+    delete_cache(cache);
+
+    free_de_out_queues(dep, obuf);
+    if (dep->transcode_ctx)
+        transcode_free_ctx(dep);
+}
+
 int erts_dsend_context_dtor(Binary* ctx_bin)
 {
     ErtsSendContext* ctx = ERTS_MAGIC_BIN_DATA(ctx_bin);
@@ -729,8 +782,8 @@ int erts_dsend_context_dtor(Binary* ctx_bin)
     if (ctx->dss.phase >= ERTS_DSIG_SEND_PHASE_ALLOC && ctx->dss.obuf) {
 	free_dist_obuf(ctx->dss.obuf);
     }
-    if (ctx->dep_to_deref)
-	erts_deref_dist_entry(ctx->dep_to_deref);
+    if (ctx->deref_dep)
+	erts_deref_dist_entry(ctx->dep);
 
     return 1;
 }
@@ -799,8 +852,7 @@ erts_dsig_send_unlink(ErtsDSigData *dsdp, Eterm local, Eterm remote)
 
 
 /* A local process that's being monitored by a remote one exits. We send:
-   {DOP_MONITOR_P_EXIT, Local pid or name, Remote pid, ref, reason},
-   which is rather sad as only the ref is needed, no pid's... */
+   {DOP_MONITOR_P_EXIT, Local pid or name, Remote pid, ref, reason} */
 int
 erts_dsig_send_m_exit(ErtsDSigData *dsdp, Eterm watcher, Eterm watched, 
 			  Eterm ref, Eterm reason)
@@ -809,17 +861,18 @@ erts_dsig_send_m_exit(ErtsDSigData *dsdp, Eterm watcher, Eterm watched,
     DeclareTmpHeapNoproc(ctl_heap,6);
     int res;
 
+    if (~dsdp->dep->flags & (DFLAG_DIST_MONITOR | DFLAG_DIST_MONITOR_NAME)) {
+        /*
+         * Receiver does not support DOP_MONITOR_P_EXIT (see dsig_send_monitor)
+         */
+        return ERTS_DSIG_SEND_OK;
+    }
+
     UseTmpHeapNoproc(6);
 
     ctl = TUPLE5(&ctl_heap[0], make_small(DOP_MONITOR_P_EXIT),
 		 watched, watcher, ref, reason);
 
-#ifdef DEBUG
-    erts_smp_de_links_lock(dsdp->dep);
-    ASSERT(!erts_lookup_monitor(dsdp->dep->monitors, ref));
-    erts_smp_de_links_unlock(dsdp->dep);
-#endif
-
     res = dsig_send_ctl(dsdp, ctl, 1);
     UnUseTmpHeapNoproc(6);
     return res;
@@ -836,6 +889,16 @@ erts_dsig_send_monitor(ErtsDSigData *dsdp, Eterm watcher, Eterm watched,
     DeclareTmpHeapNoproc(ctl_heap,5);
     int res;
 
+    if (~dsdp->dep->flags & (DFLAG_DIST_MONITOR | DFLAG_DIST_MONITOR_NAME)) {
+        /*
+         * Receiver does not support DOP_MONITOR_P.
+         * Just avoid sending it and by doing that reduce this monitor
+         * to only supervise the connection. This will work for simple c-nodes
+         * with a 1-to-1 relation between "Erlang process" and OS-process.
+         */
+        return ERTS_DSIG_SEND_OK;
+    }
+
     UseTmpHeapNoproc(5);
     ctl = TUPLE4(&ctl_heap[0],
 		 make_small(DOP_MONITOR_P),
@@ -848,8 +911,7 @@ erts_dsig_send_monitor(ErtsDSigData *dsdp, Eterm watcher, Eterm watched,
 
 /* A local process monitoring a remote one wants to stop monitoring, either 
    because of a demonitor bif call or because the local process died. We send
-   {DOP_DEMONITOR_P, Local pid, Remote pid or name, ref}, which is once again
-   rather redundant as only the ref will be needed on the other side... */
+   {DOP_DEMONITOR_P, Local pid, Remote pid or name, ref} */
 int
 erts_dsig_send_demonitor(ErtsDSigData *dsdp, Eterm watcher,
 			 Eterm watched, Eterm ref, int force)
@@ -858,6 +920,13 @@ erts_dsig_send_demonitor(ErtsDSigData *dsdp, Eterm watcher,
     DeclareTmpHeapNoproc(ctl_heap,5);
     int res;
 
+    if (~dsdp->dep->flags & (DFLAG_DIST_MONITOR | DFLAG_DIST_MONITOR_NAME)) {
+        /*
+         * Receiver does not support DOP_DEMONITOR_P (see dsig_send_monitor)
+         */
+        return ERTS_DSIG_SEND_OK;
+    }
+
     UseTmpHeapNoproc(5);
     ctl = TUPLE4(&ctl_heap[0],
 		 make_small(DOP_DEMONITOR_P),
@@ -868,6 +937,24 @@ erts_dsig_send_demonitor(ErtsDSigData *dsdp, Eterm watcher,
     return res;
 }
 
+static int can_send_seqtrace_token(ErtsSendContext* ctx, Eterm token) {
+    Eterm label;
+
+    if (ctx->dep->flags & DFLAG_BIG_SEQTRACE_LABELS) {
+        /* The other end is capable of handling arbitrary seq_trace labels. */
+        return 1;
+    }
+
+    /* The other end only tolerates smalls, but since we could potentially be
+     * talking to an old 32-bit emulator from a 64-bit one, we have to check
+     * whether the label is small on any emulator. */
+    label = SEQ_TRACE_T_LABEL(token);
+
+    return is_small(label) &&
+        signed_val(label) <= (ERTS_SINT32_MAX >> _TAG_IMMED1_SIZE) &&
+        signed_val(label) >= (ERTS_SINT32_MIN >> _TAG_IMMED1_SIZE);
+}
+
 int
 erts_dsig_send_msg(Eterm remote, Eterm message, ErtsSendContext* ctx)
 {
@@ -901,18 +988,38 @@ erts_dsig_send_msg(Eterm remote, Eterm message, ErtsSendContext* ctx)
                       "%T", remote);
         msize = size_object(message);
         if (have_seqtrace(token)) {
-            tok_label = signed_val(SEQ_TRACE_T_LABEL(token));
+            tok_label = SEQ_TRACE_T_DTRACE_LABEL(token);
             tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token));
             tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token));
         }
     }
 #endif
 
-    if (token != NIL)
-	ctl = TUPLE4(&ctx->ctl_heap[0],
-		     make_small(DOP_SEND_TT), am_Empty, remote, token);
-    else
-	ctl = TUPLE3(&ctx->ctl_heap[0], make_small(DOP_SEND), am_Empty, remote);
+    {
+        Eterm dist_op, sender_id;
+        int send_token;
+
+        send_token = (token != NIL && can_send_seqtrace_token(ctx, token));
+
+        if (ctx->dep->flags & DFLAG_SEND_SENDER) {
+            dist_op = make_small(send_token ?
+                                 DOP_SEND_SENDER_TT :
+                                 DOP_SEND_SENDER);
+            sender_id = sender->common.id;
+        } else {
+            dist_op = make_small(send_token ?
+                                 DOP_SEND_TT :
+                                 DOP_SEND);
+            sender_id = am_Empty;
+        }
+
+        if (send_token) {
+            ctl = TUPLE4(&ctx->ctl_heap[0], dist_op, sender_id, remote, token);
+        } else {
+            ctl = TUPLE3(&ctx->ctl_heap[0], dist_op, sender_id, remote);
+        }
+    }
+
     DTRACE6(message_send, sender_name, receiver_name,
             msize, tok_label, tok_lastcnt, tok_serial);
     DTRACE7(message_send_remote, sender_name, node_name, receiver_name,
@@ -958,19 +1065,20 @@ erts_dsig_send_reg_msg(Eterm remote_name, Eterm message,
                       "{%T,%s}", remote_name, node_name);
         msize = size_object(message);
         if (have_seqtrace(token)) {
-            tok_label = signed_val(SEQ_TRACE_T_LABEL(token));
+            tok_label = SEQ_TRACE_T_DTRACE_LABEL(token);
             tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token));
             tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token));
         }
     }
 #endif
 
-    if (token != NIL)
+    if (token != NIL && can_send_seqtrace_token(ctx, token))
 	ctl = TUPLE5(&ctx->ctl_heap[0], make_small(DOP_REG_SEND_TT),
 		     sender->common.id, am_Empty, remote_name, token);
     else
 	ctl = TUPLE4(&ctx->ctl_heap[0], make_small(DOP_REG_SEND),
 		     sender->common.id, am_Empty, remote_name);
+
     DTRACE6(message_send, sender_name, receiver_name,
             msize, tok_label, tok_lastcnt, tok_serial);
     DTRACE7(message_send_remote, sender_name, node_name, receiver_name,
@@ -1022,7 +1130,7 @@ erts_dsig_send_exit_tt(ErtsDSigData *dsdp, Eterm local, Eterm remote,
         erts_snprintf(reason_str, sizeof(DTRACE_CHARBUF_NAME(reason_str)),
                       "%T", reason);
         if (have_seqtrace(token)) {
-            tok_label = signed_val(SEQ_TRACE_T_LABEL(token));
+            tok_label = SEQ_TRACE_T_DTRACE_LABEL(token);
             tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token));
             tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token));
         }
@@ -1099,7 +1207,7 @@ erts_dsig_send_group_leader(ErtsDSigData *dsdp, Eterm leader, Eterm remote)
 #  define PURIFY_MSG(msg)                                                    \
     do {								     \
 	char buf__[1]; size_t bufsz__ = sizeof(buf__);			     \
-	if (erts_sys_getenv_raw("VALGRIND_LOG_XML", buf__, &bufsz__) >= 0) { \
+	if (erts_sys_explicit_8bit_getenv("VALGRIND_LOG_XML", buf__, &bufsz__) >= 0) { \
 	    VALGRIND_PRINTF_XML("<erlang_error_log>"			     \
 			    "%s, line %d: %s</erlang_error_log>\n",	     \
 			    __FILE__, __LINE__, msg);			     \
@@ -1145,26 +1253,27 @@ int erts_net_message(Port *prt,
     Sint type;
     Eterm token;
     Eterm token_size;
-    ErtsMonitor *mon;
-    ErtsLink *lnk;
     Uint tuple_arity;
     int res;
+    Uint32 connection_id;
 #ifdef ERTS_DIST_MSG_DBG
     ErlDrvSizeT orig_len = len;
 #endif
 
     UseTmpHeapNoproc(DIST_CTL_DEFAULT_SIZE);
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(!prt || erts_lc_is_port_locked(prt));
 
     if (!erts_is_alive) {
 	UnUseTmpHeapNoproc(DIST_CTL_DEFAULT_SIZE);
 	return 0;
     }
-    if (hlen != 0)
-	goto data_error;
+
+
+    ASSERT(hlen == 0);
+
     if (len == 0) {  /* HANDLE TICK !!! */
 	UnUseTmpHeapNoproc(DIST_CTL_DEFAULT_SIZE);
 	return 0;
@@ -1183,30 +1292,31 @@ int erts_net_message(Port *prt,
 	len--;
     }
 
-    if (len == 0) {
-	PURIFY_MSG("data error");
-	goto data_error;
-    }
-
-    res = erts_prepare_dist_ext(&ede, t, len, dep, dep->cache);
+    res = erts_prepare_dist_ext(&ede, t, len, dep, dep->cache, &connection_id);
 
-    if (res >= 0)
-	res = ctl_len = erts_decode_dist_ext_size(&ede);
-    else {
+    switch (res) {
+    case ERTS_PREP_DIST_EXT_CLOSED:
+        return 0; /* Connection not alive; ignore signal... */
+    case ERTS_PREP_DIST_EXT_FAILED:
 #ifdef ERTS_DIST_MSG_DBG
 	erts_fprintf(stderr, "DIST MSG DEBUG: erts_prepare_dist_ext() failed:\n");
 	bw(buf, orig_len);
 #endif
-	ctl_len = 0;
-    }
-
-    if (res < 0) {
+        goto data_error;
+    case ERTS_PREP_DIST_EXT_SUCCESS:
+	ctl_len = erts_decode_dist_ext_size(&ede);
+        if (ctl_len < 0) {
 #ifdef ERTS_DIST_MSG_DBG
-	erts_fprintf(stderr, "DIST MSG DEBUG: erts_decode_dist_ext_size(CTL) failed:\n");
-	bw(buf, orig_len);
+            erts_fprintf(stderr, "DIST MSG DEBUG: erts_decode_dist_ext_size(CTL) failed:\n");
+            bw(buf, orig_len);
 #endif
-	PURIFY_MSG("data error");
-	goto data_error;
+            PURIFY_MSG("data error");
+            goto data_error;
+        }
+        break;
+    default:
+        ERTS_INTERNAL_ERROR("Unexpected result from erts_prepare_dist_ext()");
+        break;
     }
 
     if (ctl_len > DIST_CTL_DEFAULT_SIZE) {
@@ -1237,90 +1347,92 @@ int erts_net_message(Port *prt,
     }
 
     token_size = 0;
+    token = NIL;
 
     switch (type = unsigned_val(tuple[1])) {
-    case DOP_LINK:
+    case DOP_LINK: {
+        ErtsDSigData dsd;
+        int code;
+
 	if (tuple_arity != 3) {
 	    goto invalid_message;
 	}
 	from = tuple[2];
 	to   = tuple[3];  /* local proc to link to */
 
-	if (is_not_pid(from) || is_not_pid(to)) {
-	    goto invalid_message;
-	}
+	if (is_not_external_pid(from))
+            goto invalid_message;
 
-	rp = erts_pid2proc_opt(NULL, 0,
-			       to, ERTS_PROC_LOCK_LINK,
-			       ERTS_P2P_FLG_ALLOW_OTHER_X);
-	if (!rp) {
-	    /* This is tricky (we MUST force a distributed send) */
-	    ErtsDSigData dsd;
-	    int code;
-	    code = erts_dsig_prepare(&dsd, dep, NULL, ERTS_DSP_NO_LOCK, 0);
-	    if (code == ERTS_DSIG_PREP_CONNECTED) {
-		code = erts_dsig_send_exit(&dsd, to, from, am_noproc);
-		ASSERT(code == ERTS_DSIG_SEND_OK);
-	    }
-	    break;
-	}
+        if (dep != external_pid_dist_entry(from))
+            goto invalid_message;
 
-	erts_smp_de_links_lock(dep);
-	res = erts_add_link(&ERTS_P_LINKS(rp), LINK_PID, from);
+        if (is_external_pid(to)) {
+            if (external_pid_dist_entry(to) != erts_this_dist_entry)
+                goto invalid_message;
+            /* old incarnation of node; reply noproc... */
+        }
+        else if (is_internal_pid(to)) {
+            ErtsLinkData *ldp = erts_link_create(ERTS_LNK_TYPE_DIST_PROC,
+                                                 from, to);
+            ASSERT(ldp->a.other.item == to);
+            ASSERT(eq(ldp->b.other.item, from));
+#ifdef DEBUG
+            code =
+#endif
+                erts_link_dist_insert(&ldp->a, dep->mld);
+            ASSERT(code);
 
-	if (res < 0) {
-	    /* It was already there! Lets skip the rest... */
-	    erts_smp_de_links_unlock(dep);
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	    break;
-	}
-	lnk = erts_add_or_lookup_link(&(dep->nlinks), LINK_PID, rp->common.id);
-	erts_add_link(&(ERTS_LINK_ROOT(lnk)), LINK_PID, from);
-	erts_smp_de_links_unlock(dep);
+            if (erts_proc_sig_send_link(NULL, to, &ldp->b))
+                break; /* done */
 
-	if (IS_TRACED_FL(rp, F_TRACE_PROCS))
-	    trace_proc(NULL, 0, rp, am_getting_linked, from);
+            /* Failed to send signal; cleanup and reply noproc... */
+
+#ifdef DEBUG
+            code =
+#endif
+                erts_link_dist_delete(&ldp->a);
+            ASSERT(code);
+            erts_link_release_both(ldp);
+        }
+
+        code = erts_dsig_prepare(&dsd, dep, NULL, 0, ERTS_DSP_NO_LOCK, 0, 0);
+        if (code == ERTS_DSIG_PREP_CONNECTED) {
+            code = erts_dsig_send_exit(&dsd, to, from, am_noproc);
+            ASSERT(code == ERTS_DSIG_SEND_OK);
+        }
 
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
 	break;
+    }
 
     case DOP_UNLINK: {
-	ErtsDistLinkData dld;
 	if (tuple_arity != 3) {
 	    goto invalid_message;
 	}
 	from = tuple[2];
 	to = tuple[3];
-	if (is_not_pid(from) || is_not_pid(to)) {
+	if (is_not_external_pid(from))
+	    goto invalid_message;
+        if (dep != external_pid_dist_entry(from))
 	    goto invalid_message;
-	}
-	
-	rp = erts_pid2proc_opt(NULL, 0,
-			       to, ERTS_PROC_LOCK_LINK,
-			       ERTS_P2P_FLG_ALLOW_OTHER_X);
-	if (!rp)
-	    break;
 
-	lnk = erts_remove_link(&ERTS_P_LINKS(rp), from);
+        if (is_external_pid(to)
+            && erts_this_dist_entry == external_pid_dist_entry(from))
+            break;
 
-	if (IS_TRACED_FL(rp, F_TRACE_PROCS) && lnk != NULL) {
-	    trace_proc(NULL, 0, rp, am_getting_unlinked, from);
-	}
+        if (is_not_internal_pid(to))
+            goto invalid_message;
 
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-
-	erts_remove_dist_link(&dld, to, from, dep);
-	erts_destroy_dist_link(&dld);
-	if (lnk)
-	    erts_destroy_link(lnk);
+        erts_proc_sig_send_dist_unlink(dep, from, to);
 	break;
     }
     
     case DOP_MONITOR_P: {
 	/* A remote process wants to monitor us, we get:
 	   {DOP_MONITOR_P, Remote pid, local pid or name, ref} */
-	Eterm name;
-	
+	Eterm pid, name;
+        ErtsDSigData dsd;
+        int code;
+
 	if (tuple_arity != 4) {
 	    goto invalid_message;
 	}
@@ -1329,44 +1441,58 @@ int erts_net_message(Port *prt,
 	watched = tuple[3];  /* local proc to monitor */
 	ref     = tuple[4];
 
-	if (is_not_ref(ref)) {
+        if (is_not_external_pid(watcher))
+            goto invalid_message;
+        else if (external_pid_dist_entry(watcher) != dep)
+            goto invalid_message;
+
+	if (is_not_ref(ref))
 	    goto invalid_message;
-	}
 
-	if (is_atom(watched)) {
-	    name = watched;
-	    rp = erts_whereis_process(NULL, 0,
-				      watched, ERTS_PROC_LOCK_LINK,
-				      ERTS_P2P_FLG_ALLOW_OTHER_X);
-	}
-	else {
-	    name = NIL;
-	    rp = erts_pid2proc_opt(NULL, 0,
-				   watched, ERTS_PROC_LOCK_LINK,
-				   ERTS_P2P_FLG_ALLOW_OTHER_X);
-	}
+        if (is_internal_pid(watched)) {
+            name = NIL;
+            pid = watched;
+        }
+        else if (is_atom(watched)) {
+            name = watched;
+            pid = erts_whereis_name_to_id(NULL, watched);
+            /* if port or undefined; reply noproc... */
+        }
+        else if (is_external_pid(watched)
+                 && external_pid_dist_entry(watched) == erts_this_dist_entry) {
+            name = NIL;
+            pid = am_undefined; /* old incarnation; reply noproc... */
+        }
+        else
+            goto invalid_message;
 
-	if (!rp) {
-	    ErtsDSigData dsd;
-	    int code;
-	    code = erts_dsig_prepare(&dsd, dep, NULL, ERTS_DSP_NO_LOCK, 0);
-	    if (code == ERTS_DSIG_PREP_CONNECTED) {
-		code = erts_dsig_send_m_exit(&dsd, watcher, watched, ref,
-					     am_noproc);
-		ASSERT(code == ERTS_DSIG_SEND_OK);
-	    }
-	}
-	else {
-	    if (is_atom(watched))
-		watched = rp->common.id;
-	    erts_smp_de_links_lock(dep);
-	    erts_add_monitor(&(dep->monitors), MON_ORIGIN, ref, watched, name);
-	    erts_add_monitor(&ERTS_P_MONITORS(rp), MON_TARGET, ref, watcher, name);
-	    erts_smp_de_links_unlock(dep);
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	}
+        if (is_internal_pid(pid)) {
+            ErtsMonitorData *mdp;
+            mdp = erts_monitor_create(ERTS_MON_TYPE_DIST_PROC,
+                                      ref, watcher, pid, name);
 
-	break;
+            code = erts_monitor_dist_insert(&mdp->origin, dep->mld);
+            ASSERT(code); (void)code;
+
+            if (erts_proc_sig_send_monitor(&mdp->target, pid))
+                break; /* done */
+
+            /* Failed to send to local proc; cleanup reply noproc... */
+
+            code = erts_monitor_dist_delete(&mdp->origin);
+            ASSERT(code); (void)code;
+            erts_monitor_release_both(mdp);
+
+        }
+
+        code = erts_dsig_prepare(&dsd, dep, NULL, 0, ERTS_DSP_NO_LOCK, 0, 0);
+        if (code == ERTS_DSIG_PREP_CONNECTED) {
+            code = erts_dsig_send_m_exit(&dsd, watcher, watched, ref,
+                                         am_noproc);
+            ASSERT(code == ERTS_DSIG_SEND_OK);
+        }
+
+        break;
     }
 
     case DOP_DEMONITOR_P:
@@ -1377,36 +1503,43 @@ int erts_net_message(Port *prt,
 	if (tuple_arity != 4) {
 	    goto invalid_message;
 	}
-	/* watcher = tuple[2]; */
-	/* watched = tuple[3]; May be an atom in case of monitor name */
+
+	watcher = tuple[2];
+	watched = tuple[3];
 	ref = tuple[4];
 
-	if(is_not_ref(ref)) {
+	if (is_not_ref(ref)) {
 	    goto invalid_message;
 	}
 
-	erts_smp_de_links_lock(dep);
-	mon = erts_remove_monitor(&(dep->monitors),ref);
-	erts_smp_de_links_unlock(dep);
-	/* ASSERT(mon != NULL); can happen in case of broken dist message */
-	if (mon == NULL) {
-	    break;
-	}
-	watched = mon->u.pid;
-	erts_destroy_monitor(mon);
-	rp = erts_pid2proc_opt(NULL, 0,
-			       watched, ERTS_PROC_LOCK_LINK,
-			       ERTS_P2P_FLG_ALLOW_OTHER_X);
-	if (!rp) {
-	    break;
-	}
-	mon = erts_remove_monitor(&ERTS_P_MONITORS(rp), ref);
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	ASSERT(mon != NULL);
-	if (mon == NULL) {
-	    break;
-	}
-	erts_destroy_monitor(mon);
+        if (is_not_external_pid(watcher) || external_pid_dist_entry(watcher) != dep)
+            goto invalid_message;
+
+        if (is_internal_pid(watched))
+            erts_proc_sig_send_dist_demonitor(watched, ref);
+        else if (is_external_pid(watched)
+                 && external_pid_dist_entry(watched) == erts_this_dist_entry) {
+            /* old incarnation; ignore it */
+            ;
+        }
+        else if (is_atom(watched)) {
+            ErtsMonLnkDist *mld = dep->mld;
+            ErtsMonitor *mon;
+
+            erts_mtx_lock(&mld->mtx);
+
+            mon = erts_monitor_tree_lookup(mld->orig_name_monitors, ref);
+            if (mon)
+                erts_monitor_tree_delete(&mld->orig_name_monitors, mon);
+
+            erts_mtx_unlock(&mld->mtx);
+
+            if (mon)
+                erts_proc_sig_send_demonitor(mon);
+        }
+        else
+            goto invalid_message;
+
 	break;
 
     case DOP_REG_SEND_TT:
@@ -1462,42 +1595,56 @@ int erts_net_message(Port *prt,
 
 	    erts_queue_dist_message(rp, locks, ede_copy, token, from);
 	    if (locks)
-		erts_smp_proc_unlock(rp, locks);
+		erts_proc_unlock(rp, locks);
 	}
 	break;
 
+    case DOP_SEND_SENDER_TT: {
+        Uint xsize;
     case DOP_SEND_TT:
+
 	if (tuple_arity != 4) {
 	    goto invalid_message;
 	}
-	
-	token_size = size_object(tuple[4]);
-	/* Fall through ... */
+
+	token = tuple[4];
+	token_size = size_object(token);
+        xsize = ERTS_HEAP_FRAG_SIZE(token_size);
+        goto send_common;
+
+    case DOP_SEND_SENDER:
     case DOP_SEND:
+
+        token = NIL;
+        xsize = 0;
+	if (tuple_arity != 3)
+	    goto invalid_message;
+
+    send_common:
+
 	/*
-	 * There is intentionally no testing of the cookie (it is always '')
-	 * from R9B and onwards.
+         * If DOP_SEND_SENDER or DOP_SEND_SENDER_TT element 2 contains
+         * the sender pid (i.e. DFLAG_SEND_SENDER is set); otherwise,
+         * the atom '' (empty cookie).
 	 */
+        ASSERT((type == DOP_SEND_SENDER || type == DOP_SEND_SENDER_TT)
+               ? (is_pid(tuple[2]) && (dep->flags & DFLAG_SEND_SENDER))
+               : tuple[2] == am_Empty);
+
 #ifdef ERTS_DIST_MSG_DBG
 	dist_msg_dbg(&ede, "MSG", buf, orig_len);
 #endif
-	if (type != DOP_SEND_TT && tuple_arity != 3) {
-	    goto invalid_message;
-	}
 	to = tuple[3];
 	if (is_not_pid(to)) {
 	    goto invalid_message;
 	}
 	rp = erts_proc_lookup(to);
 	if (rp) {
-	    Uint xsize = type == DOP_SEND ? 0 : ERTS_HEAP_FRAG_SIZE(token_size);
 	    ErtsProcLocks locks = 0;
 	    ErtsDistExternal *ede_copy;
 
 	    ede_copy = erts_make_dist_ext_copy(&ede, xsize);
-	    if (type == DOP_SEND) {
-		token = NIL;
-	    } else {
+	    if (is_not_nil(token)) {
 		ErlHeapFragment *heap_frag;
 		ErlOffHeap *ohp;
 		ASSERT(xsize);
@@ -1505,82 +1652,50 @@ int erts_net_message(Port *prt,
 		ERTS_INIT_HEAP_FRAG(heap_frag, token_size, token_size);
 		hp = heap_frag->mem;
 		ohp = &heap_frag->off_heap;
-		token = tuple[4];
 		token = copy_struct(token, token_size, &hp, ohp);
 	    }
 
-	    erts_queue_dist_message(rp, locks, ede_copy, token, tuple[2]);
+	    erts_queue_dist_message(rp, locks, ede_copy, token, am_Empty);
 	    if (locks)
-		erts_smp_proc_unlock(rp, locks);
+		erts_proc_unlock(rp, locks);
 	}
 	break;
+    }
 
     case DOP_MONITOR_P_EXIT: {
 	/* We are monitoring a process on the remote node which dies, we get
 	   {DOP_MONITOR_P_EXIT, Remote pid or name, Local pid, ref, reason} */
 	   
-
-	DeclareTmpHeapNoproc(lhp,3);
-	Eterm sysname;
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCKS_MSG_SEND|ERTS_PROC_LOCK_LINK;
-
 	if (tuple_arity != 5) {
 	    goto invalid_message;
 	}
 
-	/* watched = tuple[2]; */  /* remote proc which died */
-	/* watcher = tuple[3]; */
+	watched = tuple[2];  /* remote proc or name which died */
+	watcher = tuple[3];
 	ref     = tuple[4];
 	reason  = tuple[5];
 
-	if(is_not_ref(ref)) {
+	if (is_not_ref(ref))
 	    goto invalid_message;
-	}
-
-	erts_smp_de_links_lock(dep);
-	sysname = dep->sysname;
-	mon = erts_remove_monitor(&(dep->monitors), ref);
-	/*
-	 * If demonitor was performed at the same time as the
-	 * monitored process exits, monitoring side will have
-	 * removed info about monitor. In this case, do nothing
-	 * and everything will be as it should.
-	 */
-	erts_smp_de_links_unlock(dep);
-	if (mon == NULL) {
-	    break;
-	}
-	rp = erts_pid2proc(NULL, 0, mon->u.pid, rp_locks);
 
-	erts_destroy_monitor(mon);
-	if (rp == NULL) {
-	    break;
-	}
+        if (is_not_external_pid(watched) && is_not_atom(watched))
+            goto invalid_message;
 
-	mon = erts_remove_monitor(&ERTS_P_MONITORS(rp), ref);
+        if (is_not_internal_pid(watcher)) {
+            if (!is_external_pid(watcher))
+                goto invalid_message;
+            if (erts_this_dist_entry == external_pid_dist_entry(watcher))
+                break;
+            goto invalid_message;
+        }
 
-	if (mon == NULL) {
-	    erts_smp_proc_unlock(rp, rp_locks);
-	    break;
-	}
-	UseTmpHeapNoproc(3);
-	
-	watched = (is_not_nil(mon->name)
-		   ? TUPLE2(&lhp[0], mon->name, sysname)
-		   : mon->u.pid);
-	
-	erts_queue_monitor_message(rp, &rp_locks,
-				   ref, am_process, watched, reason);
-	erts_smp_proc_unlock(rp, rp_locks);
-	erts_destroy_monitor(mon);
-	UnUseTmpHeapNoproc(3);
+        erts_proc_sig_send_dist_monitor_down(dep, ref, watched,
+                                             watcher, reason);
 	break;
     }
 
     case DOP_EXIT_TT:
     case DOP_EXIT: {
-	ErtsDistLinkData dld;
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCKS_XSIG_SEND;
 	/* 'from', which 'to' is linked to, died */
 	if (type == DOP_EXIT) {
 	    if (tuple_arity != 4) {
@@ -1600,56 +1715,19 @@ int erts_net_message(Port *prt,
 	    token = tuple[4];
 	    reason = tuple[5];
 	}
-	if (is_not_pid(from) || is_not_internal_pid(to)) {
+	if (is_not_external_pid(from)
+            || dep != external_pid_dist_entry(from)
+            || is_not_internal_pid(to)) {
 	    goto invalid_message;
 	}
 
-	rp = erts_pid2proc(NULL, 0, to, rp_locks);
-	if (!rp)
-	    lnk = NULL;
-	else {
-	    lnk = erts_remove_link(&ERTS_P_LINKS(rp), from);
-
-	    /* If lnk == NULL, we have unlinked on this side, i.e.
-	     * ignore exit.
-	     */
-	    if (lnk) {
-		int xres;
-#if 0
-		/* Arndt: Maybe it should never be 'kill', but it can be,
-		   namely when a linked process does exit(kill). Until we know
-		   whether that is incorrect and what should happen instead,
-		   we leave the assertion out. */
-		ASSERT(reason != am_kill); /* should never be kill (killed) */
-#endif
-		xres = erts_send_exit_signal(NULL,
-					     from,
-					     rp,
-					     &rp_locks, 
-					     reason,
-					     token,
-					     NULL,
-					     ERTS_XSIG_FLG_IGN_KILL);
-		if (xres >= 0 && IS_TRACED_FL(rp, F_TRACE_PROCS)) {
-		    /* We didn't exit the process and it is traced */
-                    if (rp_locks & ERTS_PROC_LOCKS_XSIG_SEND) {
-                        erts_smp_proc_unlock(rp, ERTS_PROC_LOCKS_XSIG_SEND);
-                        rp_locks &= ~ERTS_PROC_LOCKS_XSIG_SEND;
-                    }
-		    trace_proc(NULL, 0, rp, am_getting_unlinked, from);
-		}
-	    }
-	    erts_smp_proc_unlock(rp, rp_locks);
-	}
-	erts_remove_dist_link(&dld, to, from, dep);
-	if (lnk)
-	    erts_destroy_link(lnk);
-	erts_destroy_dist_link(&dld);
+        erts_proc_sig_send_dist_link_exit(dep,
+                                          from, to,
+                                          reason, token);
 	break;
     }
     case DOP_EXIT2_TT:
-    case DOP_EXIT2: {
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
+    case DOP_EXIT2:
 	/* 'from' is send an exit signal to 'to' */
 	if (type == DOP_EXIT2) {
 	    if (tuple_arity != 4) {
@@ -1671,20 +1749,10 @@ int erts_net_message(Port *prt,
 	if (is_not_pid(from) || is_not_internal_pid(to)) {
 	    goto invalid_message;
 	}
-	rp = erts_pid2proc(NULL, 0, to, rp_locks);
-	if (rp) {
-	    (void) erts_send_exit_signal(NULL,
-					 from,
-					 rp,
-					 &rp_locks,
-					 reason,
-					 token,
-					 NULL,
-					 0);
-	    erts_smp_proc_unlock(rp, rp_locks);
-	}
+
+        erts_proc_sig_send_exit(NULL, from, to, reason, token, 0);
 	break;
-    }
+
     case DOP_GROUP_LEADER:
 	if (tuple_arity != 3) {
 	    goto invalid_message;
@@ -1695,11 +1763,7 @@ int erts_net_message(Port *prt,
 	    goto invalid_message;
 	}
 
-	rp = erts_pid2proc(NULL, 0, to, ERTS_PROC_LOCK_MAIN);
-	if (!rp)
-	    break;
-	rp->group_leader = STORE_NC_IN_PROC(rp, from);
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+        (void) erts_proc_sig_send_group_leader(NULL, to, from, NIL);
 	break;
 
     default: 
@@ -1711,7 +1775,7 @@ int erts_net_message(Port *prt,
 	erts_free(ERTS_ALC_T_DCTRL_BUF, (void *) ctl);
     }
     UnUseTmpHeapNoproc(DIST_CTL_DEFAULT_SIZE);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     return 0;
  invalid_message:
     {
@@ -1727,8 +1791,8 @@ decode_error:
     }
 data_error:
     UnUseTmpHeapNoproc(DIST_CTL_DEFAULT_SIZE);
-    erts_deliver_port_exit(prt, dep->cid, am_killed, 0, 1);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    erts_kill_dist_connection(dep, connection_id);
+    ERTS_CHK_NO_PROC_LOCKS;
     return -1;
 }
 
@@ -1748,6 +1812,31 @@ static int dsig_send_ctl(ErtsDSigData* dsdp, Eterm ctl, int force_busy)
     return ret;
 }
 
+static ERTS_INLINE void
+notify_dist_data(Process *c_p, Eterm pid)
+{
+    Process *rp;
+    ErtsProcLocks rp_locks;
+
+    ASSERT(erts_get_scheduler_data()
+           && !ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data()));
+    ASSERT(is_internal_pid(pid));
+
+    if (c_p && c_p->common.id == pid) {
+        rp = c_p;
+        rp_locks = ERTS_PROC_LOCK_MAIN;
+    }
+    else {
+        rp = erts_proc_lookup(pid);
+        rp_locks = 0;
+    }
+
+    if (rp) {
+        ErtsMessage *mp = erts_alloc_message(0, NULL);
+        erts_queue_message(rp, rp_locks, mp, am_dist_data, am_system);
+    }
+}
+
 int
 erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 {
@@ -1764,7 +1853,7 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 	    if (!ctx->c_p || dsdp->no_suspend)
 		ctx->force_busy = 1;
 
-	    ERTS_SMP_LC_ASSERT(!ctx->c_p
+	    ERTS_LC_ASSERT(!ctx->c_p
 			       || (ERTS_PROC_LOCK_MAIN
 				   == erts_proc_lc_my_proc_locks(ctx->c_p)));
 
@@ -1773,33 +1862,32 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 
 	    if (ctx->flags & DFLAG_DIST_HDR_ATOM_CACHE) {
 		ctx->acmp = erts_get_atom_cache_map(ctx->c_p);
-		ctx->pass_through_size = 0;
+		ctx->max_finalize_prepend = 0;
 	    }
 	    else {
 		ctx->acmp = NULL;
-		ctx->pass_through_size = 1;
+		ctx->max_finalize_prepend = 3;
 	    }
 
     #ifdef ERTS_DIST_MSG_DBG
-	    erts_fprintf(stderr, ">>%s CTL: %T\n", ctx->pass_through_size ? "P" : " ", ctx->ctl);
-        if (is_value(ctx->msg))
-            erts_fprintf(stderr, "    MSG: %T\n", ctx->msg);
+            erts_fprintf(stderr, ">> CTL: %T\n", ctx->ctl);
+            if (is_value(ctx->msg))
+                erts_fprintf(stderr, "    MSG: %T\n", ctx->msg);
     #endif
 
-	    ctx->data_size = ctx->pass_through_size;
+	    ctx->data_size = ctx->max_finalize_prepend;
 	    erts_reset_atom_cache_map(ctx->acmp);
 	    erts_encode_dist_ext_size(ctx->ctl, ctx->flags, ctx->acmp, &ctx->data_size);
 
-	    if (is_value(ctx->msg)) {
-		ctx->u.sc.wstack.wstart = NULL;
-		ctx->u.sc.flags = ctx->flags;
-		ctx->u.sc.level = 0;
-		ctx->phase = ERTS_DSIG_SEND_PHASE_MSG_SIZE;
-	    } else {
-		ctx->phase = ERTS_DSIG_SEND_PHASE_ALLOC;
-	    }
-	    break;
+	    if (is_non_value(ctx->msg)) {
+                ctx->phase = ERTS_DSIG_SEND_PHASE_ALLOC;
+                break;
+            }
+            ctx->u.sc.wstack.wstart = NULL;
+            ctx->u.sc.flags = ctx->flags;
+            ctx->u.sc.level = 0;
 
+            ctx->phase = ERTS_DSIG_SEND_PHASE_MSG_SIZE;
 	case ERTS_DSIG_SEND_PHASE_MSG_SIZE:
 	    if (erts_encode_dist_ext_size_int(ctx->msg, ctx, &ctx->data_size)) {
 		retval = ERTS_DSIG_SEND_CONTINUE;
@@ -1814,23 +1902,25 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 	    ctx->data_size += ctx->dhdr_ext_size;
 
 	    ctx->obuf = alloc_dist_obuf(ctx->data_size);
-	    ctx->obuf->ext_endp = &ctx->obuf->data[0] + ctx->pass_through_size + ctx->dhdr_ext_size;
+	    ctx->obuf->ext_endp = &ctx->obuf->data[0] + ctx->max_finalize_prepend + ctx->dhdr_ext_size;
 
 	    /* Encode internal version of dist header */
 	    ctx->obuf->extp = erts_encode_ext_dist_header_setup(ctx->obuf->ext_endp, ctx->acmp);
 	    /* Encode control message */
 	    erts_encode_dist_ext(ctx->ctl, &ctx->obuf->ext_endp, ctx->flags, ctx->acmp, NULL, NULL);
-	    if (is_value(ctx->msg)) {
-		ctx->u.ec.flags = ctx->flags;
-		ctx->u.ec.level = 0;
-		ctx->u.ec.wstack.wstart = NULL;
-		ctx->phase = ERTS_DSIG_SEND_PHASE_MSG_ENCODE;
-	    } else {
-		ctx->phase = ERTS_DSIG_SEND_PHASE_FIN;
-	    }
-	    break;
-
-	case ERTS_DSIG_SEND_PHASE_MSG_ENCODE:
+	    if (is_non_value(ctx->msg)) {
+                ctx->obuf->msg_start = NULL;
+                ctx->phase = ERTS_DSIG_SEND_PHASE_FIN;
+                break;
+            }
+            ctx->u.ec.flags = ctx->flags;
+            ctx->u.ec.hopefull_flags = 0;
+            ctx->u.ec.level = 0;
+            ctx->u.ec.wstack.wstart = NULL;
+            ctx->obuf->msg_start = ctx->obuf->ext_endp;
+
+            ctx->phase = ERTS_DSIG_SEND_PHASE_MSG_ENCODE;
+        case ERTS_DSIG_SEND_PHASE_MSG_ENCODE:
 	    if (erts_encode_dist_ext(ctx->msg, &ctx->obuf->ext_endp, ctx->flags, ctx->acmp, &ctx->u.ec, &ctx->reds)) {
 		retval = ERTS_DSIG_SEND_CONTINUE;
 		goto done;
@@ -1843,38 +1933,59 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 	    int resume = 0;
 
 	    ASSERT(ctx->obuf->extp < ctx->obuf->ext_endp);
-	    ASSERT(&ctx->obuf->data[0] <= ctx->obuf->extp - ctx->pass_through_size);
+	    ASSERT(&ctx->obuf->data[0] <= ctx->obuf->extp - ctx->max_finalize_prepend);
 	    ASSERT(ctx->obuf->ext_endp <= &ctx->obuf->data[0] + ctx->data_size);
 
 	    ctx->data_size = ctx->obuf->ext_endp - ctx->obuf->extp;
 
+            ctx->obuf->hopefull_flags = ctx->u.ec.hopefull_flags;
 	    /*
 	     * Signal encoded; now verify that the connection still exists,
 	     * and if so enqueue the signal and schedule it for send.
 	     */
 	    ctx->obuf->next = NULL;
-	    erts_smp_de_rlock(dep);
+	    erts_de_rlock(dep);
 	    cid = dep->cid;
-	    if (cid != dsdp->cid
-		|| dep->connection_id != dsdp->connection_id
-		|| dep->status & ERTS_DE_SFLG_EXITING) {
+	    if (dep->state == ERTS_DE_STATE_EXITING
+                || dep->state == ERTS_DE_STATE_IDLE
+                || dep->connection_id != dsdp->connection_id) {
 		/* Not the same connection as when we started; drop message... */
-		erts_smp_de_runlock(dep);
+		erts_de_runlock(dep);
 		free_dist_obuf(ctx->obuf);
 	    }
 	    else {
+                Sint qsize;
+                erts_aint32_t qflgs;
 		ErtsProcList *plp = NULL;
-		erts_smp_mtx_lock(&dep->qlock);
-		dep->qsize += size_obuf(ctx->obuf);
-		if (dep->qsize >= erts_dist_buf_busy_limit)
-		    dep->qflgs |= ERTS_DE_QFLG_BUSY;
-		if (!ctx->force_busy && (dep->qflgs & ERTS_DE_QFLG_BUSY)) {
-		    erts_smp_mtx_unlock(&dep->qlock);
+                Eterm notify_proc = NIL;
+                Sint obsz = size_obuf(ctx->obuf);
+
+		erts_mtx_lock(&dep->qlock);
+		qsize = erts_atomic_add_read_nob(&dep->qsize, (erts_aint_t) obsz);
+                ASSERT(qsize >= obsz);
+                qflgs = erts_atomic32_read_nob(&dep->qflgs);
+		if (!(qflgs & ERTS_DE_QFLG_BUSY) && qsize >= erts_dist_buf_busy_limit) {
+		    erts_atomic32_read_bor_relb(&dep->qflgs, ERTS_DE_QFLG_BUSY);
+                    qflgs |= ERTS_DE_QFLG_BUSY;
+                }
+                if (qsize == obsz && (qflgs & ERTS_DE_QFLG_REQ_INFO)) {
+                    /* Previously empty queue and info requested... */
+                    qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
+                                                       ~ERTS_DE_QFLG_REQ_INFO);
+                    if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
+                        notify_proc = dep->cid;
+                        ASSERT(is_internal_pid(notify_proc));
+                    }
+                    /* else: requester will send itself the message... */
+                    qflgs &= ~ERTS_DE_QFLG_REQ_INFO;
+                }
+		if (!ctx->force_busy && (qflgs & ERTS_DE_QFLG_BUSY)) {
+		    erts_mtx_unlock(&dep->qlock);
 
 		    plp = erts_proclist_create(ctx->c_p);
 		    erts_suspend(ctx->c_p, ERTS_PROC_LOCK_MAIN, NULL);
 		    suspended = 1;
-		    erts_smp_mtx_lock(&dep->qlock);
+		    erts_mtx_lock(&dep->qlock);
 		}
 
 		/* Enqueue obuf on dist entry */
@@ -1885,7 +1996,8 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 		dep->out_queue.last = ctx->obuf;
 
 		if (!ctx->force_busy) {
-		    if (!(dep->qflgs & ERTS_DE_QFLG_BUSY)) {
+                    qflgs = erts_atomic32_read_nob(&dep->qflgs);
+		    if (!(qflgs & ERTS_DE_QFLG_BUSY)) {
 			if (suspended)
 			    resume = 1; /* was busy when we started, but isn't now */
     #ifdef USE_VM_PROBES
@@ -1909,9 +2021,17 @@ erts_dsig_send(ErtsDSigData *dsdp, struct erts_dsig_send_context* ctx)
 		    }
 		}
 
-		erts_smp_mtx_unlock(&dep->qlock);
-		erts_schedule_dist_command(NULL, dep);
-		erts_smp_de_runlock(dep);
+		erts_mtx_unlock(&dep->qlock);
+		if (dep->state != ERTS_DE_STATE_PENDING) {
+                    if (is_internal_port(dep->cid))
+                        erts_schedule_dist_command(NULL, dep);
+                }
+                else {
+                    notify_proc = NIL;
+                }
+		erts_de_runlock(dep);
+                if (is_internal_pid(notify_proc))
+                    notify_dist_data(ctx->c_p, notify_proc);
 
 		if (resume) {
 		    erts_resume(ctx->c_p, ERTS_PROC_LOCK_MAIN);
@@ -1965,16 +2085,20 @@ static Uint
 dist_port_command(Port *prt, ErtsDistOutputBuf *obuf)
 {
     int fpe_was_unmasked;
-    Uint size = obuf->ext_endp - obuf->extp;
+    ErlDrvSizeT size;
+    char *bufp;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
-    if (size > (Uint) INT_MAX)
-	erts_exit(ERTS_DUMP_EXIT,
-		 "Absurdly large distribution output data buffer "
-		 "(%beu bytes) passed.\n",
-		 size);
+    if (!obuf) {
+        size = 0;
+        bufp = NULL;
+    }
+    else {
+        size = obuf->ext_endp - obuf->extp;
+        bufp = (char*) obuf->extp;
+    }
 
 #ifdef USE_VM_PROBES
     if (DTRACE_ENABLED(dist_output)) {
@@ -1989,11 +2113,10 @@ dist_port_command(Port *prt, ErtsDistOutputBuf *obuf)
                 remote_str, size);
     }
 #endif
+
     prt->caller = NIL;
     fpe_was_unmasked = erts_block_fpe();
-    (*prt->drv_ptr->output)((ErlDrvData) prt->drv_data,
-			    (char*) obuf->extp,
-			    (int) size);
+    (*prt->drv_ptr->output)((ErlDrvData) prt->drv_data, bufp, size);
     erts_unblock_fpe(fpe_was_unmasked);
     return size;
 }
@@ -2002,33 +2125,41 @@ static Uint
 dist_port_commandv(Port *prt, ErtsDistOutputBuf *obuf)
 {
     int fpe_was_unmasked;
-    Uint size = obuf->ext_endp - obuf->extp;
+    ErlDrvSizeT size;
     SysIOVec iov[2];
     ErlDrvBinary* bv[2];
     ErlIOVec eiov;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
-
-    if (size > (Uint) INT_MAX)
-	erts_exit(ERTS_DUMP_EXIT,
-		 "Absurdly large distribution output data buffer "
-		 "(%beu bytes) passed.\n",
-		 size);
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     iov[0].iov_base = NULL;
     iov[0].iov_len = 0;
     bv[0] = NULL;
 
-    iov[1].iov_base = obuf->extp;
-    iov[1].iov_len = size;
-    bv[1] = Binary2ErlDrvBinary(ErtsDistOutputBuf2Binary(obuf));
+    if (!obuf) {
+        size = 0;
+        eiov.vsize = 1;
+    }
+    else {
+        size = obuf->ext_endp - obuf->extp;
+        eiov.vsize = 2;
+
+        iov[1].iov_base = obuf->extp;
+        iov[1].iov_len = size;
+        bv[1] = Binary2ErlDrvBinary(ErtsDistOutputBuf2Binary(obuf));
+    }
 
-    eiov.vsize = 2;
     eiov.size = size;
     eiov.iov = iov;
     eiov.binv = bv;
 
+    if (size > (Uint) INT_MAX)
+	erts_exit(ERTS_DUMP_EXIT,
+		 "Absurdly large distribution output data buffer "
+		 "(%beu bytes) passed.\n",
+		 size);
+
     ASSERT(prt->drv_ptr->outputv);
 
 #ifdef USE_VM_PROBES
@@ -2062,7 +2193,6 @@ dist_port_commandv(Port *prt, ErtsDistOutputBuf *obuf)
 #endif
 
 #define ERTS_PORT_REDS_DIST_CMD_START 5
-#define ERTS_PORT_REDS_DIST_CMD_FINALIZE 3
 #define ERTS_PORT_REDS_DIST_CMD_EXIT 200
 #define ERTS_PORT_REDS_DIST_CMD_RESUMED 5
 #define ERTS_PORT_REDS_DIST_CMD_DATA(SZ) \
@@ -2071,37 +2201,36 @@ dist_port_commandv(Port *prt, ErtsDistOutputBuf *obuf)
    : ((((Sint) (SZ)) >> 10) & ((Sint) ERTS_PORT_REDS_MASK__)))
 
 int
-erts_dist_command(Port *prt, int reds_limit)
+erts_dist_command(Port *prt, int initial_reds)
 {
-    Sint reds = ERTS_PORT_REDS_DIST_CMD_START;
-    Uint32 status;
+    Sint reds = initial_reds - ERTS_PORT_REDS_DIST_CMD_START;
+    enum dist_entry_state state;
     Uint32 flags;
-    Sint obufsize = 0;
+    Sint qsize, obufsize = 0;
     ErtsDistOutputQueue oq, foq;
     DistEntry *dep = prt->dist_entry;
     Uint (*send)(Port *prt, ErtsDistOutputBuf *obuf);
     erts_aint32_t sched_flags;
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
-
-    erts_smp_refc_inc(&dep->refc, 1); /* Otherwise dist_entry might be
-				     removed if port command fails */
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
-    erts_smp_atomic_set_mb(&dep->dist_cmd_scheduled, 0);
+    erts_atomic_set_mb(&dep->dist_cmd_scheduled, 0);
 
-    erts_smp_de_rlock(dep);
+    erts_de_rlock(dep);
     flags = dep->flags;
-    status = dep->status;
+    state = dep->state;
     send = dep->send;
-    erts_smp_de_runlock(dep);
+    erts_de_runlock(dep);
 
-    if (status & ERTS_DE_SFLG_EXITING) {
+    if (state == ERTS_DE_STATE_EXITING) {
 	erts_deliver_port_exit(prt, prt->common.id, am_killed, 0, 1);
-	erts_deref_dist_entry(dep);
-	return reds + ERTS_PORT_REDS_DIST_CMD_EXIT;
+        reds -= ERTS_PORT_REDS_DIST_CMD_EXIT;
+	return initial_reds - reds;
     }
 
+    ASSERT(state != ERTS_DE_STATE_PENDING);
+
     ASSERT(send);
 
     /*
@@ -2112,42 +2241,42 @@ erts_dist_command(Port *prt, int reds_limit)
      * a mess.
      */
 
-    erts_smp_mtx_lock(&dep->qlock);
+    erts_mtx_lock(&dep->qlock);
     oq.first = dep->out_queue.first;
     oq.last = dep->out_queue.last;
     dep->out_queue.first = NULL;
     dep->out_queue.last = NULL;
-    erts_smp_mtx_unlock(&dep->qlock);
+    erts_mtx_unlock(&dep->qlock);
 
     foq.first = dep->finalized_out_queue.first;
     foq.last = dep->finalized_out_queue.last;
     dep->finalized_out_queue.first = NULL;
     dep->finalized_out_queue.last = NULL;
 
-    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
 
-    if (reds > reds_limit)
+    if (reds < 0)
 	goto preempted;
 
     if (!(sched_flags & ERTS_PTS_FLG_BUSY_PORT) && foq.first) {
 	int preempt = 0;
 	do {
-	    Uint size;
-	    ErtsDistOutputBuf *fob;
-
-	    size = (*send)(prt, foq.first);
-	    esdp->io.out += (Uint64) size;
+            Uint size;
+            ErtsDistOutputBuf *fob;
+            size = (*send)(prt, foq.first);
+            erts_atomic64_inc_nob(&dep->out);
+            esdp->io.out += (Uint64) size;
 #ifdef ERTS_RAW_DIST_MSG_DBG
-	    erts_fprintf(stderr, ">> ");
-	    bw(foq.first->extp, size);
+            erts_fprintf(stderr, ">> ");
+            bw(foq.first->extp, size);
 #endif
-	    reds += ERTS_PORT_REDS_DIST_CMD_DATA(size);
+	    reds -= ERTS_PORT_REDS_DIST_CMD_DATA(size);
 	    fob = foq.first;
 	    obufsize += size_obuf(fob);
 	    foq.first = foq.first->next;
 	    free_dist_obuf(fob);
-	    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
-	    preempt = reds > reds_limit || (sched_flags & ERTS_PTS_FLG_EXIT);
+	    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
+	    preempt = reds < 0 || (sched_flags & ERTS_PTS_FLG_EXIT);
 	    if (sched_flags & ERTS_PTS_FLG_BUSY_PORT)
 		break;
 	} while (foq.first && !preempt);
@@ -2160,79 +2289,71 @@ erts_dist_command(Port *prt, int reds_limit)
     if (sched_flags & ERTS_PTS_FLG_BUSY_PORT) {
 	if (oq.first) {
 	    ErtsDistOutputBuf *ob;
-	    int preempt;
+            ErtsDistOutputBuf *last_finalized = NULL;
 	finalize_only:
-	    preempt = 0;
 	    ob = oq.first;
 	    ASSERT(ob);
 	    do {
-		ob->extp = erts_encode_ext_dist_header_finalize(ob->extp,
-								dep->cache,
-								flags);
-		if (!(flags & DFLAG_DIST_HDR_ATOM_CACHE))
-		    *--ob->extp = PASS_THROUGH; /* Old node; 'pass through'
-						   needed */
-		ASSERT(&ob->data[0] <= ob->extp && ob->extp < ob->ext_endp);
-		reds += ERTS_PORT_REDS_DIST_CMD_FINALIZE;
-		preempt = reds > reds_limit;
-		if (preempt)
-		    break;
-		ob = ob->next;
+		reds = erts_encode_ext_dist_header_finalize(ob, dep, flags, reds);
+                if (reds < 0)
+                    break;
+                last_finalized  = ob;
+                ob = ob->next;
 	    } while (ob);
-	    /*
-	     * At least one buffer was finalized; if we got preempted,
-	     * ob points to the last buffer that we finalized.
-	     */
-	    if (foq.last)
-		foq.last->next = oq.first;
-	    else
-		foq.first = oq.first;
-	    if (!preempt) {
-		/* All buffers finalized */
-		foq.last = oq.last;
-		oq.first = oq.last = NULL;
-	    }
-	    else {
-		/* Not all buffers finalized; split oq. */
-		foq.last = ob;
-		oq.first = ob->next;
-		if (oq.first)
-		    ob->next = NULL;
-		else
-		    oq.last = NULL;
-	    }
-	    if (preempt)
-		goto preempted;
+            if (last_finalized) {
+                /*
+                 * At least one buffer was finalized; if we got preempted,
+                 * ob points to the next buffer to continue finalize.
+                 */
+                if (foq.last)
+                    foq.last->next = oq.first;
+                else
+                    foq.first = oq.first;
+                foq.last = last_finalized;
+                if (!ob) {
+                    /* All buffers finalized */
+                    ASSERT(foq.last == oq.last);
+                    ASSERT(foq.last->next == NULL);
+                    oq.first = oq.last = NULL;
+                }
+                else {
+                    /* Not all buffers finalized; split oq. */
+                    ASSERT(foq.last->next == ob);
+                    foq.last->next = NULL;
+                    oq.first = ob;
+                }
+            }
+            if (reds <= 0)
+                goto preempted;
 	}
     }
     else {
+        int de_busy;
 	int preempt = 0;
 	while (oq.first && !preempt) {
 	    ErtsDistOutputBuf *fob;
 	    Uint size;
-	    oq.first->extp
-		= erts_encode_ext_dist_header_finalize(oq.first->extp,
-						       dep->cache,
-						       flags);
-	    reds += ERTS_PORT_REDS_DIST_CMD_FINALIZE;
-	    if (!(flags & DFLAG_DIST_HDR_ATOM_CACHE))
-		*--oq.first->extp = PASS_THROUGH; /* Old node; 'pass through'
-						     needed */
+            reds = erts_encode_ext_dist_header_finalize(oq.first, dep, flags, reds);
+            if (reds < 0) {
+                preempt = 1;
+                break;
+            }
 	    ASSERT(&oq.first->data[0] <= oq.first->extp
-		   && oq.first->extp < oq.first->ext_endp);
+		   && oq.first->extp <= oq.first->ext_endp);
 	    size = (*send)(prt, oq.first);
+            erts_atomic64_inc_nob(&dep->out);
 	    esdp->io.out += (Uint64) size;
 #ifdef ERTS_RAW_DIST_MSG_DBG
-	    erts_fprintf(stderr, ">> ");
-	    bw(oq.first->extp, size);
+            erts_fprintf(stderr, ">> ");
+            bw(oq.first->extp, size);
 #endif
-	    reds += ERTS_PORT_REDS_DIST_CMD_DATA(size);
+	    reds -= ERTS_PORT_REDS_DIST_CMD_DATA(size);
 	    fob = oq.first;
 	    obufsize += size_obuf(fob);
 	    oq.first = oq.first->next;
 	    free_dist_obuf(fob);
-	    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
-	    preempt = reds > reds_limit || (sched_flags & ERTS_PTS_FLG_EXIT);
+	    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
+	    preempt = reds <= 0 || (sched_flags & ERTS_PTS_FLG_EXIT);
 	    if ((sched_flags & ERTS_PTS_FLG_BUSY_PORT) && oq.first && !preempt)
 		goto finalize_only;
 	}
@@ -2258,23 +2379,24 @@ erts_dist_command(Port *prt, int reds_limit)
 	 * dist entry in a non-busy state and resume suspended
 	 * processes.
 	 */
-	erts_smp_mtx_lock(&dep->qlock);
-	ASSERT(dep->qsize >= obufsize);
-	dep->qsize -= obufsize;
+	erts_mtx_lock(&dep->qlock);
+        de_busy = !!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_BUSY);
+        qsize = (Sint) erts_atomic_add_read_nob(&dep->qsize,
+                                                (erts_aint_t) -obufsize);
+	ASSERT(qsize >= 0);
 	obufsize = 0;
 	if (!(sched_flags & ERTS_PTS_FLG_BUSY_PORT)
-	    && (dep->qflgs & ERTS_DE_QFLG_BUSY)
-	    && dep->qsize < erts_dist_buf_busy_limit) {
+	    && de_busy && qsize < erts_dist_buf_busy_limit) {
 	    ErtsProcList *suspendees;
 	    int resumed;
 	    suspendees = get_suspended_on_de(dep, ERTS_DE_QFLG_BUSY);
-	    erts_smp_mtx_unlock(&dep->qlock);
+	    erts_mtx_unlock(&dep->qlock);
 
 	    resumed = erts_resume_processes(suspendees);
-	    reds += resumed*ERTS_PORT_REDS_DIST_CMD_RESUMED;
+	    reds -= resumed*ERTS_PORT_REDS_DIST_CMD_RESUMED;
 	}
 	else
-	    erts_smp_mtx_unlock(&dep->qlock);
+	    erts_mtx_unlock(&dep->qlock);
     }
 
     ASSERT(!oq.first && !oq.last);
@@ -2283,14 +2405,18 @@ erts_dist_command(Port *prt, int reds_limit)
 
     if (obufsize != 0) {
 	ASSERT(obufsize > 0);
-	erts_smp_mtx_lock(&dep->qlock);
-	ASSERT(dep->qsize >= obufsize);
-	dep->qsize -= obufsize;
-	erts_smp_mtx_unlock(&dep->qlock);
+	erts_mtx_lock(&dep->qlock);
+#ifdef DEBUG
+        qsize = (Sint) erts_atomic_add_read_nob(&dep->qsize,
+                                                (erts_aint_t) -obufsize);
+	ASSERT(qsize >= 0);
+#else
+        erts_atomic_add_nob(&dep->qsize, (erts_aint_t) -obufsize);
+#endif
+	erts_mtx_unlock(&dep->qlock);
     }
 
-    ASSERT(foq.first || !foq.last);
-    ASSERT(!foq.first || foq.last);
+    ASSERT(!!foq.first == !!foq.last);
     ASSERT(!dep->finalized_out_queue.first);
     ASSERT(!dep->finalized_out_queue.last);
 
@@ -2300,12 +2426,10 @@ erts_dist_command(Port *prt, int reds_limit)
     }
 
      /* Avoid wrapping reduction counter... */
-    if (reds > INT_MAX/2)
-	reds = INT_MAX/2;
+    if (reds < INT_MIN/2)
+	reds = INT_MIN/2;
 
-    erts_deref_dist_entry(dep);
-
-    return reds;
+    return initial_reds - reds;
 
  preempted:
     /*
@@ -2313,8 +2437,7 @@ erts_dist_command(Port *prt, int reds_limit)
      * since last call to driver.
      */
 
-    ASSERT(oq.first || !oq.last);
-    ASSERT(!oq.first || oq.last);
+    ASSERT(!!oq.first == !!oq.last);
 
     if (sched_flags & ERTS_PTS_FLG_EXIT) {
 	/*
@@ -2338,12 +2461,6 @@ erts_dist_command(Port *prt, int reds_limit)
 
 	foq.first = NULL;
 	foq.last = NULL;
-
-#ifdef DEBUG
-	erts_smp_mtx_lock(&dep->qlock);
-	ASSERT(dep->qsize == obufsize);
-	erts_smp_mtx_unlock(&dep->qlock);
-#endif
     }
     else {
 	if (oq.first) {
@@ -2351,14 +2468,14 @@ erts_dist_command(Port *prt, int reds_limit)
 	     * Unhandle buffers need to be put back first
 	     * in out_queue.
 	     */
-	    erts_smp_mtx_lock(&dep->qlock);
-	    dep->qsize -= obufsize;
+	    erts_mtx_lock(&dep->qlock);
+	    erts_atomic_add_nob(&dep->qsize, -obufsize);
 	    obufsize = 0;
 	    oq.last->next = dep->out_queue.first;
 	    dep->out_queue.first = oq.first;
 	    if (!dep->out_queue.last)
 		dep->out_queue.last = oq.last;
-	    erts_smp_mtx_unlock(&dep->qlock);
+	    erts_mtx_unlock(&dep->qlock);
 	}
 
 	erts_schedule_dist_command(prt, NULL);
@@ -2366,6 +2483,369 @@ erts_dist_command(Port *prt, int reds_limit)
     goto done;
 }
 
+#if 0
+
+int
+dist_data_finalize(Process *c_p, int reds_limit)
+{
+    int reds = 5;
+    DistEntry *dep = ;
+    ErtsDistOutputQueue oq, foq;
+    ErtsDistOutputBuf *ob;
+    int preempt;
+
+
+    erts_mtx_lock(&dep->qlock);
+    flags = dep->flags;
+    oq.first = dep->out_queue.first;
+    oq.last = dep->out_queue.last;
+    dep->out_queue.first = NULL;
+    dep->out_queue.last = NULL;
+    erts_mtx_unlock(&dep->qlock);
+
+    if (!oq.first) {
+        ASSERT(!oq.last);
+        oq.first = dep->tmp_out_queue.first;
+        oq.last = dep->tmp_out_queue.last;
+    }
+    else {
+        ErtsDistOutputBuf *f, *l;
+        ASSERT(oq.last);
+        if (dep->tmp_out_queue.last) {
+            dep->tmp_out_queue.last->next = oq.first;
+            oq.first = dep->tmp_out_queue.first;
+        }
+    }
+
+    if (!oq.first) {
+        /* Nothing to do... */
+        ASSERT(!oq.last);
+        return reds;
+    }
+
+    foq.first = dep->finalized_out_queue.first;
+    foq.last = dep->finalized_out_queue.last;
+
+    preempt = 0;
+    ob = oq.first;
+    ASSERT(ob);
+
+    do {
+        ob->extp = erts_encode_ext_dist_header_finalize(ob->extp,
+                                                        dep->cache,
+                                                        flags);
+        if (!(flags & DFLAG_DIST_HDR_ATOM_CACHE))
+            *--ob->extp = PASS_THROUGH; /* Old node; 'pass through'
+                                           needed */
+        ASSERT(&ob->data[0] <= ob->extp && ob->extp < ob->ext_endp);
+        reds += ERTS_PORT_REDS_DIST_CMD_FINALIZE;
+        preempt = reds > reds_limit;
+        if (preempt)
+            break;
+        ob = ob->next;
+    } while (ob);
+    /*
+     * At least one buffer was finalized; if we got preempted,
+     * ob points to the last buffer that we finalized.
+     */
+    if (foq.last)
+        foq.last->next = oq.first;
+    else
+        foq.first = oq.first;
+    if (!preempt) {
+        /* All buffers finalized */
+        foq.last = oq.last;
+        oq.first = oq.last = NULL;
+    }
+    else {
+        /* Not all buffers finalized; split oq. */
+        foq.last = ob;
+        oq.first = ob->next;
+        if (oq.first)
+            ob->next = NULL;
+        else
+            oq.last = NULL;
+    }
+
+    dep->finalized_out_queue.first = foq.first;
+    dep->finalized_out_queue.last = foq.last;
+    dep->tmp_out_queue.first = oq.first;
+    dep->tmp_out_queue.last = oq.last;
+
+    return reds;
+}
+
+#endif
+
+BIF_RETTYPE
+dist_ctrl_get_data_notification_1(BIF_ALIST_1)
+{
+    DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(BIF_P);
+    erts_aint32_t qflgs;
+    erts_aint_t qsize;
+    Eterm receiver = NIL;
+
+    if (!dep)
+        BIF_ERROR(BIF_P, EXC_NOTSUP);
+
+    if (erts_dhandle_to_dist_entry(BIF_ARG_1) != dep)
+        BIF_ERROR(BIF_P, BADARG);
+
+    /*
+     * Caller is the only one that can consume from this queue
+     * and the only one that can set the req-info flag...
+     */
+
+    erts_de_rlock(dep);
+
+    ASSERT(dep->cid == BIF_P->common.id);
+
+    qflgs = erts_atomic32_read_acqb(&dep->qflgs);
+
+    if (!(qflgs & ERTS_DE_QFLG_REQ_INFO)) {
+        qsize = erts_atomic_read_acqb(&dep->qsize);
+        ASSERT(qsize >= 0);
+        if (qsize > 0)
+            receiver = BIF_P->common.id; /* Notify ourselves... */
+        else { /* Empty queue; set req-info flag... */
+            qflgs = erts_atomic32_read_bor_mb(&dep->qflgs,
+                                                  ERTS_DE_QFLG_REQ_INFO);
+            qsize = erts_atomic_read_acqb(&dep->qsize);
+            ASSERT(qsize >= 0);
+            if (qsize > 0) {
+                qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
+                                                       ~ERTS_DE_QFLG_REQ_INFO);
+                if (qflgs & ERTS_DE_QFLG_REQ_INFO)
+                    receiver = BIF_P->common.id; /* Notify ourselves... */
+                /* else: someone else will notify us... */
+            }
+            /* else: still empty queue... */
+        }
+    }
+    /* else: Already requested... */
+
+    erts_de_runlock(dep);
+
+    if (is_internal_pid(receiver))
+        notify_dist_data(BIF_P, receiver);
+
+    BIF_RET(am_ok);
+}
+
+BIF_RETTYPE
+dist_ctrl_put_data_2(BIF_ALIST_2)
+{
+    DistEntry *dep;
+    ErlDrvSizeT size;
+    Eterm input_handler;
+
+    if (is_binary(BIF_ARG_2))
+        size = binary_size(BIF_ARG_2);
+    else if (is_nil(BIF_ARG_2))
+        size = 0;
+    else if (is_list(BIF_ARG_2))
+        BIF_TRAP2(dist_ctrl_put_data_trap,
+                  BIF_P, BIF_ARG_1, BIF_ARG_2);
+    else
+        BIF_ERROR(BIF_P, BADARG);
+
+    dep = erts_dhandle_to_dist_entry(BIF_ARG_1);
+    if (!dep)
+        BIF_ERROR(BIF_P, BADARG);
+
+    input_handler = (Eterm) erts_atomic_read_nob(&dep->input_handler);
+
+    if (input_handler != BIF_P->common.id)
+        BIF_ERROR(BIF_P, EXC_NOTSUP);
+
+    erts_atomic64_inc_nob(&dep->in);
+
+    if (size != 0) {
+        byte *data, *temp_alloc = NULL;
+
+        data = (byte *) erts_get_aligned_binary_bytes(BIF_ARG_2, &temp_alloc);
+        if (!data)
+            BIF_ERROR(BIF_P, BADARG);
+
+        erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+
+        (void) erts_net_message(NULL, dep, NULL, 0, data, size);
+        /*
+         * We ignore any decode failures. On fatal failures the
+         * connection will be taken down by killing the
+         * distribution channel controller...
+         */
+
+        erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+
+        BUMP_REDS(BIF_P, 5);
+
+        erts_free_aligned_binary_bytes(temp_alloc);
+
+    }
+
+    BIF_RET(am_ok);
+}
+
+BIF_RETTYPE
+dist_get_stat_1(BIF_ALIST_1)
+{
+    Sint64 read, write, pend;
+    Eterm res, *hp, **hpp;
+    Uint sz, *szp;
+    DistEntry *dep = erts_dhandle_to_dist_entry(BIF_ARG_1);
+
+    if (!dep)
+        BIF_ERROR(BIF_P, BADARG);
+
+    erts_de_rlock(dep);
+
+    read = (Sint64) erts_atomic64_read_nob(&dep->in);
+    write = (Sint64) erts_atomic64_read_nob(&dep->out);
+    pend = (Sint64) erts_atomic_read_nob(&dep->qsize);
+
+    erts_de_runlock(dep);
+
+    sz = 0;
+    szp = &sz;
+    hpp = NULL;
+
+    while (1) {
+        res = erts_bld_tuple(hpp, szp, 4,
+                             am_ok,
+                             erts_bld_sint64(hpp, szp, read),
+                             erts_bld_sint64(hpp, szp, write),
+                             pend ? am_true : am_false);
+        if (hpp)
+            break;
+        hp = HAlloc(BIF_P, sz);
+        hpp = &hp;
+        szp = NULL;
+    }
+
+    BIF_RET(res);
+}
+
+BIF_RETTYPE
+dist_ctrl_input_handler_2(BIF_ALIST_2)
+{
+    DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(BIF_P);
+
+    if (!dep)
+        BIF_ERROR(BIF_P, EXC_NOTSUP);
+
+    if (erts_dhandle_to_dist_entry(BIF_ARG_1) != dep)
+        BIF_ERROR(BIF_P, BADARG);
+
+    if (is_not_internal_pid(BIF_ARG_2))
+        BIF_ERROR(BIF_P, BADARG);
+
+    erts_atomic_set_nob(&dep->input_handler,
+                        (erts_aint_t) BIF_ARG_2);
+
+    BIF_RET(am_ok);
+}
+
+BIF_RETTYPE
+dist_ctrl_get_data_1(BIF_ALIST_1)
+{
+    DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(BIF_P);
+    const Sint initial_reds = ERTS_BIF_REDS_LEFT(BIF_P);
+    Sint reds = initial_reds;
+    ErtsDistOutputBuf *obuf;
+    Eterm *hp;
+    ProcBin *pb;
+    erts_aint_t qsize;
+
+    if (!dep)
+        BIF_ERROR(BIF_P, EXC_NOTSUP);
+
+    if (erts_dhandle_to_dist_entry(BIF_ARG_1) != dep)
+        BIF_ERROR(BIF_P, BADARG);
+
+    erts_de_rlock(dep);
+
+    if (dep->state == ERTS_DE_STATE_EXITING)
+        goto return_none;
+
+    ASSERT(dep->cid == BIF_P->common.id);
+
+#if 0
+    if (dep->finalized_out_queue.first) {
+        obuf = dep->finalized_out_queue.first;
+        dep->finalized_out_queue.first = obuf->next;
+        if (!obuf->next)
+            dep->finalized_out_queue.last = NULL;
+    }
+    else
+#endif
+    {
+        if (!dep->tmp_out_queue.first) {
+            ASSERT(!dep->tmp_out_queue.last);
+            ASSERT(!dep->transcode_ctx);
+            qsize = erts_atomic_read_acqb(&dep->qsize);
+            if (qsize > 0) {
+                erts_mtx_lock(&dep->qlock);
+                dep->tmp_out_queue.first = dep->out_queue.first;
+                dep->tmp_out_queue.last = dep->out_queue.last;
+                dep->out_queue.first = NULL;
+                dep->out_queue.last = NULL;
+                erts_mtx_unlock(&dep->qlock);
+            }
+        }
+
+        if (!dep->tmp_out_queue.first) {
+            ASSERT(!dep->tmp_out_queue.last);
+        return_none:
+            erts_de_runlock(dep);
+            BIF_RET(am_none);
+        }
+
+        obuf = dep->tmp_out_queue.first;
+        reds = erts_encode_ext_dist_header_finalize(obuf, dep, dep->flags, reds);
+        if (reds < 0) {
+            erts_de_runlock(dep);
+            ERTS_BIF_YIELD1(bif_export[BIF_dist_ctrl_get_data_1],
+                            BIF_P, BIF_ARG_1);
+        }
+
+        dep->tmp_out_queue.first = obuf->next;
+        if (!obuf->next)
+            dep->tmp_out_queue.last = NULL;
+    }
+
+    erts_atomic64_inc_nob(&dep->out);
+
+    erts_de_runlock(dep);
+
+    hp =  HAlloc(BIF_P, PROC_BIN_SIZE);
+    pb = (ProcBin *) (char *) hp;
+    pb->thing_word = HEADER_PROC_BIN;
+    pb->size = obuf->ext_endp - obuf->extp;
+    pb->next = MSO(BIF_P).first;
+    MSO(BIF_P).first = (struct erl_off_heap_header*) pb;
+    pb->val = ErtsDistOutputBuf2Binary(obuf);
+    pb->bytes = (byte*) obuf->extp;
+    pb->flags = 0;
+
+    qsize = erts_atomic_add_read_nob(&dep->qsize, -size_obuf(obuf));
+    ASSERT(qsize >= 0);
+
+    if (qsize < erts_dist_buf_busy_limit/2
+        && (erts_atomic32_read_acqb(&dep->qflgs) & ERTS_DE_QFLG_BUSY)) {
+        ErtsProcList *resume_procs = NULL;
+        erts_mtx_lock(&dep->qlock);
+        resume_procs = get_suspended_on_de(dep, ERTS_DE_QFLG_BUSY);
+        erts_mtx_unlock(&dep->qlock);
+        if (resume_procs) {
+            int resumed = erts_resume_processes(resume_procs);
+            reds -= resumed*ERTS_PORT_REDS_DIST_CMD_RESUMED;
+        }
+    }
+
+    BIF_RET2(make_binary(pb), (initial_reds - reds));
+}
+
 void
 erts_dist_port_not_busy(Port *prt)
 {
@@ -2385,24 +2865,33 @@ erts_dist_port_not_busy(Port *prt)
     erts_schedule_dist_command(prt, NULL);
 }
 
+static void kill_connection(DistEntry *dep)
+{
+    ERTS_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
+    ASSERT(dep->state == ERTS_DE_STATE_CONNECTED);
+
+    dep->state = ERTS_DE_STATE_EXITING;
+    erts_mtx_lock(&dep->qlock);
+    ASSERT(!(erts_atomic32_read_nob(&dep->qflgs) & ERTS_DE_QFLG_EXIT));
+    erts_atomic32_read_bor_nob(&dep->qflgs, ERTS_DE_QFLG_EXIT);
+    erts_mtx_unlock(&dep->qlock);
+
+    if (is_internal_port(dep->cid))
+        erts_schedule_dist_command(NULL, dep);
+    else if (is_internal_pid(dep->cid))
+        schedule_kill_dist_ctrl_proc(dep->cid);
+}
+
 void
 erts_kill_dist_connection(DistEntry *dep, Uint32 connection_id)
 {
-    erts_smp_de_rwlock(dep);
-    if (is_internal_port(dep->cid)
-	&& connection_id == dep->connection_id
-	&& !(dep->status & ERTS_DE_SFLG_EXITING)) {
+    erts_de_rwlock(dep);
+    if (connection_id == dep->connection_id
+        && dep->state == ERTS_DE_STATE_CONNECTED) {
 
-	dep->status |= ERTS_DE_SFLG_EXITING;
-
-	erts_smp_mtx_lock(&dep->qlock);
-	ASSERT(!(dep->qflgs & ERTS_DE_QFLG_EXIT));
-	dep->qflgs |= ERTS_DE_QFLG_EXIT;
-	erts_smp_mtx_unlock(&dep->qlock);
-
-	erts_schedule_dist_command(NULL, dep);
+        kill_connection(dep);
     }
-    erts_smp_de_rwunlock(dep);
+    erts_de_rwunlock(dep);
 }
 
 struct print_to_data {
@@ -2414,57 +2903,55 @@ static void doit_print_monitor_info(ErtsMonitor *mon, void *vptdp)
 {
     fmtfn_t to = ((struct print_to_data *) vptdp)->to;
     void *arg = ((struct print_to_data *) vptdp)->arg;
-    Process *rp;
-    ErtsMonitor *rmon;
-    rp = erts_proc_lookup(mon->u.pid);
-    if (!rp || (rmon = erts_lookup_monitor(ERTS_P_MONITORS(rp), mon->ref)) == NULL) {
-	erts_print(to, arg, "Warning, stray monitor for: %T\n", mon->u.pid);
-    } else if (mon->type == MON_ORIGIN) {
-	/* Local pid is being monitored */
+    ErtsMonitorDataExtended *mdep;
+
+    ASSERT(mon->flags & ERTS_ML_FLG_EXTENDED);
+
+    mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+
+    ASSERT(mdep->dist);
+
+    if (erts_monitor_is_origin(mon)) {
 	erts_print(to, arg, "Remotely monitored by: %T %T\n",
-		   mon->u.pid, rmon->u.pid);
-    } else {
-	erts_print(to, arg, "Remote monitoring: %T ", mon->u.pid);
-	if (is_not_atom(rmon->u.pid))
-	    erts_print(to, arg, "%T\n", rmon->u.pid);
-	else
-	    erts_print(to, arg, "{%T, %T}\n",
-		       rmon->name,
-		       rmon->u.pid); /* which in this case is the 
-				      remote system name... */
+		   mon->other.item, mdep->md.target.other.item);
+    }
+    else {
+	erts_print(to, arg, "Remote monitoring: %T ", mon->other.item);
+        if (mon->flags & ERTS_ML_FLG_NAME)
+	    erts_print(to, arg, "{%T, %T}\n", mdep->u.name, mdep->dist->nodename);
+        else
+	    erts_print(to, arg, "%T\n", mdep->md.origin.other.item);
     }
 }    
 
-static void print_monitor_info(fmtfn_t to, void *arg, ErtsMonitor *mon)
+static void print_monitor_info(fmtfn_t to, void *arg, DistEntry *dep)
 {
     struct print_to_data ptd = {to, arg};
-    erts_doforall_monitors(mon,&doit_print_monitor_info,&ptd);
-}
-
-typedef struct {
-    struct print_to_data *ptdp;
-    Eterm from;
-} PrintLinkContext;
-
-static void doit_print_link_info2(ErtsLink *lnk, void *vpplc)
-{
-    PrintLinkContext *pplc = (PrintLinkContext *) vpplc;
-    erts_print(pplc->ptdp->to, pplc->ptdp->arg, "Remote link: %T %T\n",
-	       pplc->from, lnk->pid);
+    if (dep->mld) {
+        erts_monitor_list_foreach(dep->mld->monitors,
+                                  doit_print_monitor_info,
+                                  (void *) &ptd);
+        erts_monitor_tree_foreach(dep->mld->orig_name_monitors,
+                                  doit_print_monitor_info,
+                                  (void *) &ptd);
+    }
 }
 
 static void doit_print_link_info(ErtsLink *lnk, void *vptdp)
 {
-    if (is_internal_pid(lnk->pid) && erts_proc_lookup(lnk->pid)) {
-	PrintLinkContext plc = {(struct print_to_data *) vptdp, lnk->pid};
-	erts_doforall_links(ERTS_LINK_ROOT(lnk), &doit_print_link_info2, &plc);
-    } 
+    struct print_to_data *ptdp = vptdp;
+    ErtsLink *lnk2 = erts_link_to_other(lnk, NULL);
+    erts_print(ptdp->to, ptdp->arg, "Remote link: %T %T\n",
+	       lnk2->other.item, lnk->other.item);
 }
 
-static void print_link_info(fmtfn_t to, void *arg, ErtsLink *lnk)
+static void print_link_info(fmtfn_t to, void *arg, DistEntry *dep)
 {
     struct print_to_data ptd = {to, arg};
-    erts_doforall_links(lnk, &doit_print_link_info, (void *) &ptd);
+    if (dep->mld)
+        erts_link_list_foreach(dep->mld->links,
+                               doit_print_link_info,
+                               (void *) &ptd);
 }
 
 typedef struct {
@@ -2472,23 +2959,6 @@ typedef struct {
     Eterm sysname;
 } PrintNodeLinkContext;
     
-
-static void doit_print_nodelink_info(ErtsLink *lnk, void *vpcontext)
-{
-    PrintNodeLinkContext *pcontext = vpcontext;
-
-    if (is_internal_pid(lnk->pid) && erts_proc_lookup(lnk->pid))
-	erts_print(pcontext->ptd.to, pcontext->ptd.arg,
-		   "Remote monitoring: %T %T\n", lnk->pid, pcontext->sysname);
-}
-
-static void print_nodelink_info(fmtfn_t to, void *arg, ErtsLink *lnk, Eterm sysname)
-{
-    PrintNodeLinkContext context = {{to, arg}, sysname};
-    erts_doforall_links(lnk, &doit_print_nodelink_info, &context);
-}
-
-
 static int
 info_dist_entry(fmtfn_t to, void *arg, DistEntry *dep, int visible, int connected)
 {
@@ -2517,13 +2987,10 @@ info_dist_entry(fmtfn_t to, void *arg, DistEntry *dep, int visible, int connecte
   }
 
   erts_print(to, arg, "Name: %T", dep->sysname);
-#ifdef DEBUG
-  erts_print(to, arg, " (refc=%d)", erts_smp_refc_read(&dep->refc, 0));
-#endif
   erts_print(to, arg, "\n");
   if (!connected && is_nil(dep->cid)) {
-    if (dep->nlinks) {
-      erts_print(to, arg, "Error: Got links to not connected node:%T\n",
+    if (dep->mld) {
+      erts_print(to, arg, "Error: Got links/monitors to not connected node:%T\n",
 		 dep->sysname);
     }
     return 0;
@@ -2532,9 +2999,8 @@ info_dist_entry(fmtfn_t to, void *arg, DistEntry *dep, int visible, int connecte
   erts_print(to, arg, "Controller: %T\n", dep->cid, to);
 
   erts_print_node_info(to, arg, dep->sysname, NULL, NULL);
-  print_monitor_info(to, arg, dep->monitors);
-  print_link_info(to, arg, dep->nlinks);
-  print_nodelink_info(to, arg, dep->node_links, dep->sysname);
+  print_monitor_info(to, arg, dep);
+  print_link_info(to, arg, dep);
 
   return 0;
     
@@ -2563,6 +3029,10 @@ int distribution_info(fmtfn_t to, void *arg)	/* Called by break handler */
       info_dist_entry(to, arg, dep, 0, 1);
     }
 
+    for (dep = erts_pending_dist_entries; dep; dep = dep->next) {
+        info_dist_entry(to, arg, dep, 0, 0);
+    }
+
     for (dep = erts_not_connected_dist_entries; dep; dep = dep->next) {
         if (dep != erts_this_dist_entry) {
             info_dist_entry(to, arg, dep, 0, 0);
@@ -2585,7 +3055,6 @@ int distribution_info(fmtfn_t to, void *arg)	/* Called by break handler */
 	    monitor_node  -- turn on/off node monitoring
 
             node controller only:
-            dist_exit/3       -- send exit signals from remote to local process
             dist_link/2       -- link a remote process to a local
             dist_unlink/2     -- unlink a remote from a local
 ****************************************************************************/
@@ -2595,15 +3064,6 @@ int distribution_info(fmtfn_t to, void *arg)	/* Called by break handler */
 /**********************************************************************
  ** Set the node name of current node fail if node already is set.
  ** setnode(name@host, Creation)
- ** loads functions pointer to trap_functions from module erlang.
- **    erlang:dsend/2
- **    erlang:dlink/1
- **    erlang:dunlink/1
- **    erlang:dmonitor_node/3
- **    erlang:dgroup_leader/2
- **    erlang:dexit/2
- **  -- are these needed ?
- **    dexit/1
  ***********************************************************************/
 
 BIF_RETTYPE setnode_2(BIF_ALIST_2)
@@ -2627,44 +3087,50 @@ BIF_RETTYPE setnode_2(BIF_ALIST_2)
 	goto error;
 
     /* Check that all trap functions are defined !! */
-    if (dsend2_trap->addressv[0] == NULL ||
-	dsend3_trap->addressv[0] == NULL ||
-	/*	dsend_nosuspend_trap->address == NULL ||*/
-	dlink_trap->addressv[0] == NULL ||
-	dunlink_trap->addressv[0] == NULL ||
-	dmonitor_node_trap->addressv[0] == NULL ||
-	dgroup_leader_trap->addressv[0] == NULL ||
-	dmonitor_p_trap->addressv[0] == NULL ||
-	dexit_trap->addressv[0] == NULL) {
+    if (dmonitor_node_trap->addressv[0] == NULL) {
 	goto error;
     }
 
-    net_kernel = erts_whereis_process(BIF_P, ERTS_PROC_LOCK_MAIN,
-				      am_net_kernel, ERTS_PROC_LOCK_MAIN, 0);
-    if (!net_kernel)
+    net_kernel = erts_whereis_process(BIF_P,
+                                      ERTS_PROC_LOCK_MAIN,
+				      am_net_kernel,
+                                      ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS,
+                                      0);
+    if (!net_kernel || ERTS_PROC_GET_DIST_ENTRY(net_kernel))
 	goto error;
 
     /* By setting F_DISTRIBUTION on net_kernel,
-     * do_net_exist will be called when net_kernel is terminated !! */
+     * erts_do_net_exits will be called when net_kernel is terminated !! */
     net_kernel->flags |= F_DISTRIBUTION;
 
-    if (net_kernel != BIF_P)
-	erts_smp_proc_unlock(net_kernel, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(net_kernel,
+                     (ERTS_PROC_LOCK_STATUS
+                      | ((net_kernel != BIF_P)
+                         ? ERTS_PROC_LOCK_MAIN
+                         : 0)));
 
 #ifdef DEBUG
-    erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rlock(&erts_dist_table_rwmtx);
     ASSERT(!erts_visible_dist_entries && !erts_hidden_dist_entries);
-    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_runlock(&erts_dist_table_rwmtx);
 #endif
 
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
+    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
     inc_no_nodes();
     erts_set_this_node(BIF_ARG_1, (Uint32) creation);
     erts_is_alive = 1;
     send_nodes_mon_msgs(NULL, am_nodeup, BIF_ARG_1, am_visible, NIL);
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+
+    /*
+     * Note erts_this_dist_entry is changed by erts_set_this_node(),
+     * so we *need* to use the new one after erts_set_this_node()
+     * is called.
+     */
+    erts_ref_dist_entry(erts_this_dist_entry);
+    ERTS_PROC_SET_DIST_ENTRY(net_kernel, erts_this_dist_entry);
 
     BIF_RET(am_true);
 
@@ -2672,74 +3138,80 @@ BIF_RETTYPE setnode_2(BIF_ALIST_2)
     BIF_ERROR(BIF_P, BADARG);
 }
 
-/**********************************************************************
- ** Allocate a dist entry, set node name install the connection handler
- ** setnode_3({name@host, Creation}, Cid, {Type, Version, Initial, IC, OC})
- ** Type = flag field, where the flags are specified in dist.h
- ** Version = distribution version, >= 1
- ** IC = in_cookie (ignored)
- ** OC = out_cookie (ignored)
- **
- ** Note that in distribution protocols above 1, the Initial parameter
- ** is always NIL and the cookies are always the atom '', cookies are not
- ** sent in the distribution messages but are only used in 
- ** the handshake.
- **
- ***********************************************************************/
+/*
+ * erts_internal:create_dist_channel/4 is used by
+ * erlang:setnode/3.
+ */
 
-BIF_RETTYPE setnode_3(BIF_ALIST_3)
+typedef struct {
+    DistEntry *dep;
+    Uint flags;
+    Uint version;
+} ErtsSetupConnDistCtrl;
+
+static void
+setup_connection_epiloge_rwunlock(Process *c_p, DistEntry *dep,
+                                  Eterm ctrlr, Uint flags,
+                                  Uint version);
+
+static Eterm
+setup_connection_distctrl(Process *c_p, void *arg,
+                          int *redsp, ErlHeapFragment **bpp);
+
+BIF_RETTYPE erts_internal_create_dist_channel_4(BIF_ALIST_4)
 {
     BIF_RETTYPE ret;
     Uint flags;
-    unsigned long version;
-    Eterm ic, oc;
-    Eterm *tp;
+    Uint version;
+    Eterm *hp, res_tag = THE_NON_VALUE, res = THE_NON_VALUE;
     DistEntry *dep = NULL;
+    int de_locked = 0;
     Port *pp = NULL;
 
-    /* Prepare for success */
-    ERTS_BIF_PREP_RET(ret, am_true);
-
     /*
      * Check and pick out arguments
      */
 
-    if (!is_node_name_atom(BIF_ARG_1) ||
-	is_not_internal_port(BIF_ARG_2) ||
-	(erts_this_node->sysname == am_Noname)) {
-	goto badarg;
-    }
+    /* Node name... */
+    if (!is_node_name_atom(BIF_ARG_1))
+        goto badarg;
 
-    if (!is_tuple(BIF_ARG_3))
-	goto badarg;
-    tp = tuple_val(BIF_ARG_3);
-    if (*tp++ != make_arityval(4))
-	goto badarg;
-    if (!is_small(*tp))
-	goto badarg;
-    flags = unsigned_val(*tp++);
-    if (!is_small(*tp) || (version = unsigned_val(*tp)) == 0)
-	goto badarg;
-    ic = *(++tp);
-    oc = *(++tp);
-    if (!is_atom(ic) || !is_atom(oc))
-	goto badarg;
+    /* Distribution controller... */
+    if (!is_internal_port(BIF_ARG_2) && !is_internal_pid(BIF_ARG_2))
+        goto badarg;
+
+    /* Dist flags... */
+    if (!is_small(BIF_ARG_3))
+        goto badarg;
+    flags = unsigned_val(BIF_ARG_3);
+
+    /* Version... */
+    if (!is_small(BIF_ARG_4))
+        goto badarg;
+    version = unsigned_val(BIF_ARG_4);
+
+    if (version == 0)
+        goto badarg;
 
-    /* DFLAG_EXTENDED_REFERENCES is compulsory from R9 and forward */
-    if (!(DFLAG_EXTENDED_REFERENCES & flags)) {
+    if (~flags & DFLAG_DIST_MANDATORY) {
 	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
 	erts_dsprintf(dsbufp, "%T", BIF_P->common.id);
 	if (BIF_P->common.u.alive.reg)
 	    erts_dsprintf(dsbufp, " (%T)", BIF_P->common.u.alive.reg->name);
 	erts_dsprintf(dsbufp,
 		      " attempted to enable connection to node %T "
-		      "which is not able to handle extended references.\n",
+		      "which does not support all mandatory capabilities.\n",
 		      BIF_ARG_1);
 	erts_send_error_to_logger(BIF_P->group_leader, dsbufp);
 	goto badarg;
     }
 
     /*
+     * ToDo: Should we not pass connection_id as well
+     *       to make sure it's the right connection we commit.
+     */
+
+    /*
      * Arguments seem to be in order.
      */
 
@@ -2750,90 +3222,147 @@ BIF_RETTYPE setnode_3(BIF_ALIST_3)
     else if (!dep)
 	goto system_limit; /* Should never happen!!! */
 
-    pp = erts_id2port_sflgs(BIF_ARG_2,
-			    BIF_P,
-			    ERTS_PROC_LOCK_MAIN,
-			    ERTS_PORT_SFLGS_INVALID_LOOKUP);
-    erts_smp_de_rwlock(dep);
+    erts_de_rlock(dep);
+    de_locked = -1;
+
+    if (dep->state == ERTS_DE_STATE_EXITING) {
+        /* Suspend on dist entry waiting for the exit to finish */
+        ErtsProcList *plp = erts_proclist_create(BIF_P);
+        plp->next = NULL;
+        erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
+        erts_mtx_lock(&dep->qlock);
+        erts_proclist_store_last(&dep->suspended, plp);
+        erts_mtx_unlock(&dep->qlock);
+        goto yield;
+    }
 
-    if (!pp || (erts_atomic32_read_nob(&pp->state)
-		& ERTS_PORT_SFLG_EXITING))
-	goto badarg;
+    erts_de_runlock(dep);
+    de_locked = 0;
 
-    if ((pp->drv_ptr->flags & ERL_DRV_FLAG_SOFT_BUSY) == 0)
-	goto badarg;
+    if (is_internal_pid(BIF_ARG_2)) {
+        if (BIF_P->common.id == BIF_ARG_2) {
+            ErtsSetupConnDistCtrl scdc;
 
-    if (dep->cid == BIF_ARG_2 && pp->dist_entry == dep)
-	goto done; /* Already set */
+            scdc.dep = dep;
+            scdc.flags = flags;
+            scdc.version = version;
 
-    if (dep->status & ERTS_DE_SFLG_EXITING) {
-	/* Suspend on dist entry waiting for the exit to finish */
-	ErtsProcList *plp = erts_proclist_create(BIF_P);
-	plp->next = NULL;
-	erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
-	erts_smp_mtx_lock(&dep->qlock);
-	erts_proclist_store_last(&dep->suspended, plp);
-	erts_smp_mtx_unlock(&dep->qlock);
-	goto yield;
-    }
+            res = setup_connection_distctrl(BIF_P, &scdc, NULL, NULL);
+            BUMP_REDS(BIF_P, 5);
+            dep = NULL;
 
-    ASSERT(!(dep->status & ERTS_DE_SFLG_EXITING));
+            if (res == am_badarg)
+                goto badarg;
 
-    if (pp->dist_entry || is_not_nil(dep->cid))
-	goto badarg;
+            ASSERT(is_internal_magic_ref(res));
+            res_tag = am_ok; /* Connection up */
+        }
+        else {
+            ErtsSetupConnDistCtrl *scdcp;
 
-    erts_atomic32_read_bor_nob(&pp->state, ERTS_PORT_SFLG_DISTRIBUTION);
+            scdcp = erts_alloc(ERTS_ALC_T_SETUP_CONN_ARG,
+                               sizeof(ErtsSetupConnDistCtrl));
 
-    /*
-     * Dist-ports do not use the "busy port message queue" functionality, but
-     * instead use "busy dist entry" functionality.
-     */
-    {
-	ErlDrvSizeT disable = ERL_DRV_BUSY_MSGQ_DISABLED;
-	erl_drv_busy_msgq_limits(ERTS_Port2ErlDrvPort(pp), &disable, NULL);
+            scdcp->dep = dep;
+            scdcp->flags = flags;
+            scdcp->version = version;
+
+            res = erts_proc_sig_send_rpc_request(BIF_P,
+                                                 BIF_ARG_2,
+                                                 !0,
+                                                 setup_connection_distctrl,
+                                                 (void *) scdcp);
+            if (is_non_value(res))
+                goto badarg;
+
+            dep = NULL;
+
+            ASSERT(is_internal_ordinary_ref(res));
+
+            res_tag = am_message; /* Caller need to wait for dhandle in message */
+        }
+        hp = HAlloc(BIF_P, 3);
     }
+    else {
+        int new;
+
+        pp = erts_id2port_sflgs(BIF_ARG_2,
+                                BIF_P,
+                                ERTS_PROC_LOCK_MAIN,
+                                ERTS_PORT_SFLGS_INVALID_LOOKUP);
+        erts_de_rwlock(dep);
+        de_locked = 1;
+
+        if (dep->state == ERTS_DE_STATE_EXITING)
+            goto badarg;
+
+        if (!pp || (erts_atomic32_read_nob(&pp->state)
+                    & ERTS_PORT_SFLG_EXITING))
+            goto badarg;
+
+        if ((pp->drv_ptr->flags & ERL_DRV_FLAG_SOFT_BUSY) == 0)
+            goto badarg;
+
+        if (dep->cid == BIF_ARG_2 && pp->dist_entry == dep)
+            new = 0;
+        else {
+            if (dep->state != ERTS_DE_STATE_PENDING) {
+                if (dep->state == ERTS_DE_STATE_IDLE)
+                    erts_set_dist_entry_pending(dep);
+                else
+                    goto badarg;
+            }
 
-    pp->dist_entry = dep;
+            if (pp->dist_entry || is_not_nil(dep->cid))
+                goto badarg;
 
-    dep->version = version;
-    dep->creation = 0;
+            erts_atomic32_read_bor_nob(&pp->state, ERTS_PORT_SFLG_DISTRIBUTION);
 
-    ASSERT(pp->drv_ptr->outputv || pp->drv_ptr->output);
+            pp->dist_entry = dep;
 
-#if 1
-    dep->send = (pp->drv_ptr->outputv
-		 ? dist_port_commandv
-		 : dist_port_command);
-#else
-    dep->send = dist_port_command;
-#endif
-    ASSERT(dep->send);
+            ASSERT(pp->drv_ptr->outputv || pp->drv_ptr->output);
 
-#ifdef DEBUG
-    erts_smp_mtx_lock(&dep->qlock);
-    ASSERT(dep->qsize == 0);
-    erts_smp_mtx_unlock(&dep->qlock);
-#endif
+            dep->send = (pp->drv_ptr->outputv
+                         ? dist_port_commandv
+                         : dist_port_command);
+            ASSERT(dep->send);
 
-    erts_set_dist_entry_connected(dep, BIF_ARG_2, flags);
+            /*
+             * Dist-ports do not use the "busy port message queue" functionality, but
+             * instead use "busy dist entry" functionality.
+             */
+            {
+                ErlDrvSizeT disable = ERL_DRV_BUSY_MSGQ_DISABLED;
+                erl_drv_busy_msgq_limits(ERTS_Port2ErlDrvPort(pp), &disable, NULL);
+            }
 
-    if (flags & DFLAG_DIST_HDR_ATOM_CACHE)
-	create_cache(dep);
+            setup_connection_epiloge_rwunlock(BIF_P, dep, BIF_ARG_2, flags, version);
+            de_locked = 0;
+            new = !0;
+        }
 
-    erts_smp_de_rwunlock(dep);
-    dep = NULL; /* inc of refc transferred to port (dist_entry field) */
+        hp = HAlloc(BIF_P, 3 + ERTS_MAGIC_REF_THING_SIZE);
+        res = erts_build_dhandle(&hp, &BIF_P->off_heap, dep);
+        res_tag = am_ok; /* Connection up */
+        if (new)
+            dep = NULL; /* inc of refc transferred to port (dist_entry field) */
+    }
 
-    inc_no_nodes();
+    ASSERT(is_value(res) && is_value(res_tag));
+
+    res = TUPLE2(hp, res_tag, res);
+
+    ERTS_BIF_PREP_RET(ret, res);
 
-    send_nodes_mon_msgs(BIF_P,
-			am_nodeup,
-			BIF_ARG_1,
-			flags & DFLAG_PUBLISHED ? am_visible : am_hidden,
-			NIL);
  done:
 
     if (dep && dep != erts_this_dist_entry) {
-	erts_smp_de_rwunlock(dep);
+        if (de_locked) {
+            if (de_locked > 0)
+                erts_de_rwunlock(dep);
+            else
+                erts_de_runlock(dep);
+        }
 	erts_deref_dist_entry(dep);
     }
 
@@ -2843,96 +3372,324 @@ BIF_RETTYPE setnode_3(BIF_ALIST_3)
     return ret;
 
  yield:
-    ERTS_BIF_PREP_YIELD3(ret, bif_export[BIF_setnode_3], BIF_P,
-			 BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
+    ERTS_BIF_PREP_YIELD4(ret,
+                         bif_export[BIF_erts_internal_create_dist_channel_4],
+                         BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, BIF_ARG_4);
     goto done;
 
  badarg:
-    ERTS_BIF_PREP_ERROR(ret, BIF_P, BADARG);
+    ERTS_BIF_PREP_RET(ret, am_badarg);
     goto done;
 
  system_limit:
-    ERTS_BIF_PREP_ERROR(ret, BIF_P, SYSTEM_LIMIT);
+    ERTS_BIF_PREP_RET(ret, am_system_limit);
     goto done;
 }
 
+static void
+setup_connection_epiloge_rwunlock(Process *c_p, DistEntry *dep,
+                                  Eterm ctrlr, Uint flags,
+                                  Uint version)
+{
+    Eterm notify_proc = NIL;
+    erts_aint32_t qflgs;
+
+    dep->version = version;
+    dep->creation = 0;
 
-/**********************************************************************/
-/* dist_exit(Local, Term, Remote) -> Bool */
+    ASSERT(is_internal_port(ctrlr) || is_internal_pid(ctrlr));
+    ASSERT(erts_atomic_read_nob(&dep->qsize) == 0 
+           || (dep->state == ERTS_DE_STATE_PENDING));
 
-BIF_RETTYPE dist_exit_3(BIF_ALIST_3)
+    if (flags & DFLAG_DIST_HDR_ATOM_CACHE)
+	create_cache(dep);
+
+    erts_set_dist_entry_connected(dep, ctrlr, flags);
+
+    notify_proc = NIL;
+    if (erts_atomic_read_nob(&dep->qsize)) {
+        if (is_internal_port(dep->cid)) {
+            erts_schedule_dist_command(NULL, dep);
+        }
+        else {
+            qflgs = erts_atomic32_read_nob(&dep->qflgs);
+            if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
+                qflgs = erts_atomic32_read_band_mb(&dep->qflgs,
+                                                   ~ERTS_DE_QFLG_REQ_INFO);
+                if (qflgs & ERTS_DE_QFLG_REQ_INFO) {
+                    notify_proc = dep->cid;
+                    ASSERT(is_internal_pid(notify_proc));
+                }
+            }
+        }
+    }
+
+    erts_de_rwunlock(dep);
+
+    if (is_internal_pid(notify_proc))
+        notify_dist_data(c_p, notify_proc);
+
+    inc_no_nodes();
+
+    send_nodes_mon_msgs(c_p,
+			am_nodeup,
+			dep->sysname,
+			flags & DFLAG_PUBLISHED ? am_visible : am_hidden,
+			NIL);
+}
+
+static Eterm
+setup_connection_distctrl(Process *c_p, void *arg, int *redsp, ErlHeapFragment **bpp)
 {
-    Eterm local;
-    Eterm remote;
-    DistEntry *rdep;
+    ErtsSetupConnDistCtrl *scdcp = (ErtsSetupConnDistCtrl *) arg;
+    DistEntry *dep = scdcp->dep;
+    int dep_locked = 0;
+    Eterm *hp;
+    erts_aint32_t state;
 
-    local = BIF_ARG_1;
-    remote = BIF_ARG_3;
+    if (redsp)
+        *redsp = 1;
 
-    /* Check that remote is a remote process */
-    if (is_not_external_pid(remote))
-	goto error;
+    state = erts_atomic32_read_nob(&c_p->state);
 
-    rdep = external_dist_entry(remote);
-    
-    if(rdep == erts_this_dist_entry)
-	goto error;
+    if (state & ERTS_PSFLG_EXITING)
+        goto badarg;
 
-    /* Check that local is local */
-    if (is_internal_pid(local)) {
-	Process *lp;
-	ErtsProcLocks lp_locks;
-	if (BIF_P->common.id == local) {
-	    lp_locks = ERTS_PROC_LOCKS_ALL;
-	    lp = BIF_P;
-	    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCKS_ALL_MINOR);
-	}
-	else {
-	    lp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
-	    lp = erts_pid2proc(BIF_P, ERTS_PROC_LOCK_MAIN,
-			       local, lp_locks);
-	    if (!lp) {
-		BIF_RET(am_true); /* ignore */
-	    }
-	}
-	
-	(void) erts_send_exit_signal(BIF_P,
-				     remote,
-				     lp,
-				     &lp_locks,
-				     BIF_ARG_2,
-				     NIL,
-				     NULL,
-				     0);
-#ifdef ERTS_SMP
-	if (lp == BIF_P)
-	    lp_locks &= ~ERTS_PROC_LOCK_MAIN;
-#endif
-	erts_smp_proc_unlock(lp, lp_locks);
-	if (lp == BIF_P) {
-	    erts_aint32_t state = erts_smp_atomic32_read_acqb(&BIF_P->state);
-	    /*
-	     * We may have exited current process and may have to take action.
-	     */
-	    if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT)) {
-#ifdef ERTS_SMP
-		if (state & ERTS_PSFLG_PENDING_EXIT)
-		    erts_handle_pending_exit(BIF_P, ERTS_PROC_LOCK_MAIN);
-#endif
-		ERTS_BIF_EXITED(BIF_P);
-	    }
-	}
+    erts_de_rwlock(dep);
+    dep_locked = !0;
+
+    if (dep->state == ERTS_DE_STATE_EXITING)
+        goto badarg;
+
+    if (ERTS_PROC_GET_DIST_ENTRY(c_p)) {
+        if (dep == ERTS_PROC_GET_DIST_ENTRY(c_p)
+            && (c_p->flags & F_DISTRIBUTION)
+            && dep->cid == c_p->common.id) {
+            goto connected;
+        }
+        goto badarg;
     }
-    else if (is_external_pid(local)
-	     && external_dist_entry(local) == erts_this_dist_entry) {
-	BIF_RET(am_true); /* ignore */
+
+    if (dep->state != ERTS_DE_STATE_PENDING) {
+        if (dep->state == ERTS_DE_STATE_IDLE)
+            erts_set_dist_entry_pending(dep);
+        else
+            goto badarg;
+    }
+
+    if (is_not_nil(dep->cid))
+        goto badarg;
+
+    c_p->flags |= F_DISTRIBUTION;
+    ERTS_PROC_SET_DIST_ENTRY(c_p, dep);
+
+    dep->send = NULL; /* Only for distr ports... */
+
+    if (redsp)
+        *redsp = 5;
+
+    setup_connection_epiloge_rwunlock(c_p, dep, c_p->common.id,
+                                      scdcp->flags, scdcp->version);
+connected:
+
+    /* we take over previous inc in refc of dep */
+
+    if (!bpp) /* called directly... */
+        return erts_make_dhandle(c_p, dep);
+
+    erts_free(ERTS_ALC_T_SETUP_CONN_ARG, arg);
+
+    *bpp = new_message_buffer(ERTS_MAGIC_REF_THING_SIZE);
+    hp = (*bpp)->mem;
+    return erts_build_dhandle(&hp, &(*bpp)->off_heap, dep);
+
+badarg:
+
+    if (bpp) /* not called directly */
+        erts_free(ERTS_ALC_T_SETUP_CONN_ARG, arg);
+
+    if (dep_locked)
+        erts_de_rwunlock(dep);
+
+    erts_deref_dist_entry(dep);
+
+    return am_badarg;
+}
+
+
+BIF_RETTYPE erts_internal_get_dflags_0(BIF_ALIST_0)
+{
+    return erts_dflags_record;
+}
+
+BIF_RETTYPE erts_internal_new_connection_1(BIF_ALIST_1)
+{
+    DistEntry* dep;
+    Uint32 conn_id;
+    Eterm* hp;
+    Eterm dhandle;
+
+    if (is_not_atom(BIF_ARG_1)) {
+	BIF_ERROR(BIF_P, BADARG);
+    }
+    dep = erts_find_or_insert_dist_entry(BIF_ARG_1);
+
+    if (dep == erts_this_dist_entry) {
+        erts_deref_dist_entry(dep);
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    erts_de_rwlock(dep);
+
+    switch (dep->state) {
+    case ERTS_DE_STATE_PENDING:
+    case ERTS_DE_STATE_CONNECTED:
+	conn_id = dep->connection_id;
+        break;
+    case ERTS_DE_STATE_IDLE:
+        erts_set_dist_entry_pending(dep);
+	conn_id = dep->connection_id;
+        break;
+    case ERTS_DE_STATE_EXITING:
+        conn_id = (dep->connection_id + 1) & ERTS_DIST_CON_ID_MASK;
+        break;
+    default:
+        erts_exit(ERTS_ABORT_EXIT, "Invalid dep->state (%d)\n", dep->state);
+    }
+    erts_de_rwunlock(dep);
+    hp = HAlloc(BIF_P, 3 + ERTS_MAGIC_REF_THING_SIZE);
+    dhandle = erts_build_dhandle(&hp, &BIF_P->off_heap, dep);
+    erts_deref_dist_entry(dep);
+    BIF_RET(TUPLE2(hp, make_small(conn_id), dhandle));
+}
+
+static Sint abort_connection(DistEntry* dep, Uint32 conn_id)
+{
+    erts_de_rwlock(dep);
+
+    if (dep->connection_id != conn_id)
+        ;
+    else if (dep->state == ERTS_DE_STATE_CONNECTED) {
+        kill_connection(dep);
+    }
+    else if (dep->state == ERTS_DE_STATE_PENDING) {
+        ErtsAtomCache *cache;
+        ErtsDistOutputBuf *obuf;
+        ErtsProcList *resume_procs;
+        Sint reds = 0;
+        ErtsMonLnkDist *mld;
+
+	ASSERT(is_nil(dep->cid));
+
+        mld = dep->mld;
+        dep->mld = NULL;
+
+	cache = dep->cache;
+	dep->cache = NULL;
+	erts_mtx_lock(&dep->qlock);
+        obuf = dep->out_queue.first;
+        dep->out_queue.first = NULL;
+        dep->out_queue.last = NULL;
+        ASSERT(!dep->tmp_out_queue.first);
+        ASSERT(!dep->finalized_out_queue.first);
+        resume_procs = get_suspended_on_de(dep, ERTS_DE_QFLGS_ALL);
+	erts_mtx_unlock(&dep->qlock);
+	erts_atomic_set_nob(&dep->dist_cmd_scheduled, 0);
+	dep->send = NULL;
+
+        erts_set_dist_entry_not_connected(dep);
+
+	erts_de_rwunlock(dep);
+
+        schedule_con_monitor_link_cleanup(mld, THE_NON_VALUE,
+                                          THE_NON_VALUE, THE_NON_VALUE);
+
+        if (resume_procs) {
+            int resumed = erts_resume_processes(resume_procs);
+            reds += resumed*ERTS_PORT_REDS_DIST_CMD_RESUMED;
+        }
+
+	delete_cache(cache);
+	free_de_out_queues(dep, obuf);
+
+        /*
+         * We wait to make DistEntry idle and accept new connection attempts
+         * until all is cleared and deallocated. This to get some back pressure
+         * against repeated failing connection attempts saturating all CPUs
+         * with cleanup jobs.
+         */
+        erts_de_rwlock(dep);
+        ASSERT(dep->state == ERTS_DE_STATE_EXITING);
+        dep->state = ERTS_DE_STATE_IDLE;
+        erts_de_rwunlock(dep);
+        return reds;
+    }
+    erts_de_rwunlock(dep);
+    return 0;
+}
+
+BIF_RETTYPE erts_internal_abort_connection_2(BIF_ALIST_2)
+{
+    DistEntry* dep;
+    Eterm* tp;
+
+    if (is_not_atom(BIF_ARG_1) || is_not_tuple_arity(BIF_ARG_2, 2)) {
+	BIF_ERROR(BIF_P, BADARG);
+    }
+    tp = tuple_val(BIF_ARG_2);
+    dep = erts_dhandle_to_dist_entry(tp[2]);
+    if (is_not_small(tp[1]) || dep != erts_find_dist_entry(BIF_ARG_1)
+        || dep == erts_this_dist_entry) {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    if (dep) {
+        Sint reds = abort_connection(dep, unsigned_val(tp[1]));
+        BUMP_REDS(BIF_P, reds);
     }
-    else
-	goto error;
     BIF_RET(am_true);
+}
 
- error:
-    BIF_ERROR(BIF_P, BADARG);
+int erts_auto_connect(DistEntry* dep, Process *proc, ErtsProcLocks proc_locks)
+{
+    erts_de_rwlock(dep);
+    if (dep->state != ERTS_DE_STATE_IDLE) {
+        erts_de_rwunlock(dep);
+    }
+    else {
+        Process* net_kernel;
+        ErtsProcLocks nk_locks = ERTS_PROC_LOCK_MSGQ;
+        Eterm *hp;
+        ErlOffHeap *ohp;
+        ErtsMessage *mp;
+        Eterm msg, dhandle;
+        Uint32 conn_id;
+
+        erts_set_dist_entry_pending(dep);
+        conn_id = dep->connection_id;
+        erts_de_rwunlock(dep);
+
+        net_kernel = erts_whereis_process(proc, proc_locks,
+                                          am_net_kernel, nk_locks, 0);
+        if (!net_kernel) {
+            abort_connection(dep, conn_id);
+            return 0;
+        }
+
+        /*
+         * Send {auto_connect, Node, ConnId, DHandle} to net_kernel
+         */
+        mp = erts_alloc_message_heap(net_kernel, &nk_locks,
+                                     5 + ERTS_MAGIC_REF_THING_SIZE,
+                                     &hp, &ohp);
+        dhandle = erts_build_dhandle(&hp, ohp, dep);
+        msg = TUPLE4(hp, am_auto_connect, dep->sysname, make_small(conn_id),
+                     dhandle);
+        erts_queue_proc_message(proc, net_kernel, nk_locks, mp, msg);
+        erts_proc_unlock(net_kernel, nk_locks);
+    }
+
+    return 1;
 }
 
 /**********************************************************************/
@@ -3004,13 +3761,15 @@ BIF_RETTYPE nodes_1(BIF_ALIST_1)
 
     length = 0;
 
-    erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rlock(&erts_dist_table_rwmtx);
 
     ASSERT(erts_no_of_not_connected_dist_entries > 0);
     ASSERT(erts_no_of_hidden_dist_entries >= 0);
+    ASSERT(erts_no_of_pending_dist_entries >= 0);
     ASSERT(erts_no_of_visible_dist_entries >= 0);
     if(not_connected)
-      length += (erts_no_of_not_connected_dist_entries - 1);
+      length += ((erts_no_of_not_connected_dist_entries - 1)
+                 + erts_no_of_pending_dist_entries);
     if(hidden)
       length += erts_no_of_hidden_dist_entries;
     if(visible)
@@ -3021,7 +3780,7 @@ BIF_RETTYPE nodes_1(BIF_ALIST_1)
     result = NIL;
 
     if (length == 0) {
-	erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+	erts_rwmtx_runlock(&erts_dist_table_rwmtx);
 	goto done;
     }
 
@@ -3030,13 +3789,18 @@ BIF_RETTYPE nodes_1(BIF_ALIST_1)
 #ifdef DEBUG
     endp = hp + length*2;
 #endif
-    if(not_connected)
+    if(not_connected) {
       for(dep = erts_not_connected_dist_entries; dep; dep = dep->next) {
           if (dep != erts_this_dist_entry) {
             result = CONS(hp, dep->sysname, result);
             hp += 2;
           }
+        }
+      for(dep = erts_pending_dist_entries; dep; dep = dep->next) {
+          result = CONS(hp, dep->sysname, result);
+          hp += 2;
       }
+    }
     if(hidden)
       for(dep = erts_hidden_dist_entries; dep; dep = dep->next) {
 	result = CONS(hp, dep->sysname, result);
@@ -3052,7 +3816,7 @@ BIF_RETTYPE nodes_1(BIF_ALIST_1)
 	hp += 2;
     }
     ASSERT(endp == hp);
-    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_runlock(&erts_dist_table_rwmtx);
 
 done:
     UnUseTmpHeap(2,BIF_P);
@@ -3078,76 +3842,179 @@ BIF_RETTYPE is_alive_0(BIF_ALIST_0)
 static BIF_RETTYPE
 monitor_node(Process* p, Eterm Node, Eterm Bool, Eterm Options)
 {
-    DistEntry *dep;
-    ErtsLink *lnk;
+    BIF_RETTYPE ret;
+    DistEntry *dep = NULL;
     Eterm l;
+    int async_connect = 1;
 
     for (l = Options; l != NIL && is_list(l); l = CDR(list_val(l))) {
 	Eterm t = CAR(list_val(l));
-	/* allow_passive_connect the only available option right now */
-	if (t != am_allow_passive_connect) {
+	if (t == am_allow_passive_connect) {
+	    /*
+	     * Handle this horrible feature by falling back on old synchronous
+	     * auto-connect (if needed)
+	     */
+	    async_connect = 0;
+	} else {
 	    BIF_ERROR(p, BADARG);
 	}
     }
     if (l != NIL) {
 	BIF_ERROR(p, BADARG);
     }
+    if (l != NIL)
+        goto badarg;
 
-    if (is_not_atom(Node) ||
-	((Bool != am_true) && (Bool != am_false)) ||
-	((erts_this_node->sysname == am_Noname)
-	 && (Node != erts_this_node->sysname))) {
-	BIF_ERROR(p, BADARG);
-    }
-    dep = erts_sysname_to_connected_dist_entry(Node);
-    if (!dep) {
-    do_trap:
-	BIF_TRAP3(dmonitor_node_trap, p, Node, Bool, Options);
+    if (is_not_atom(Node))
+        goto badarg;
+
+    if (erts_this_node->sysname == am_Noname && Node != am_Noname)
+	goto badarg;
+
+    switch (Bool) {
+
+    case am_false: {
+        ErtsMonitor *mon;
+        /*
+         * Before OTP-21, monitor_node(Node, false) triggered
+         * auto-connect and a 'nodedown' message if that failed.
+         * Now it's a simple no-op which feels more reasonable.
+         */
+        mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(p), Node);
+        if (mon) {
+            ErtsMonitorDataExtended *mdep;
+            ASSERT(erts_monitor_is_origin(mon));
+
+            mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+
+            ASSERT((mdep->u.refc > 0));
+            if (--mdep->u.refc == 0) {
+                if (!mdep->uptr.node_monitors)
+                    erts_monitor_tree_delete(&ERTS_P_MONITORS(p), mon);
+                else {
+                    ErtsMonitor *sub_mon;
+                    ErtsMonitorDataExtended *sub_mdep;
+                    sub_mon = erts_monitor_list_last(mdep->uptr.node_monitors);
+                    erts_monitor_list_delete(&mdep->uptr.node_monitors, sub_mon);
+                    sub_mon->flags &= ~ERTS_ML_FLG_IN_SUBTABLE;
+                    sub_mdep = ((ErtsMonitorDataExtended *)
+                                erts_monitor_to_data(sub_mon));
+                    sub_mdep->uptr.node_monitors = mdep->uptr.node_monitors;
+                    mdep->uptr.node_monitors = NULL;
+                    erts_monitor_tree_replace(&ERTS_P_MONITORS(p), mon, sub_mon);
+                }
+                if (erts_monitor_dist_delete(&mdep->md.target))
+                    erts_monitor_release_both((ErtsMonitorData *) mdep);
+                else
+                    erts_monitor_release(mon);
+            }
+        }
+        break;
     }
-    if (dep == erts_this_dist_entry)
-	goto done;
 
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_LINK);
-    erts_smp_de_rlock(dep);
-    if (ERTS_DE_IS_NOT_CONNECTED(dep)) {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
-	erts_smp_de_runlock(dep);
-	goto do_trap;
-    }
-    erts_smp_de_links_lock(dep);
-    erts_smp_de_runlock(dep);
-
-    if (Bool == am_true) {
-	ASSERT(dep->cid != NIL);
-	lnk = erts_add_or_lookup_link(&(dep->node_links), LINK_NODE, 
-				      p->common.id);
-	++ERTS_LINK_REFC(lnk);
-	lnk = erts_add_or_lookup_link(&ERTS_P_LINKS(p), LINK_NODE, Node);
-	++ERTS_LINK_REFC(lnk);
-    }
-    else  {
-	lnk = erts_lookup_link(dep->node_links, p->common.id);
-	if (lnk != NULL) {
-	    if ((--ERTS_LINK_REFC(lnk)) == 0) {
-		erts_destroy_link(erts_remove_link(&(dep->node_links), 
-						   p->common.id));
-	    }
-	}
-	lnk = erts_lookup_link(ERTS_P_LINKS(p), Node);
-	if (lnk != NULL) {
-	    if ((--ERTS_LINK_REFC(lnk)) == 0) {
-		erts_destroy_link(erts_remove_link(&ERTS_P_LINKS(p),
-						   Node));
+    case am_true: {
+        ErtsDSigData dsd;
+        dsd.node = Node;
+
+        dep = erts_find_or_insert_dist_entry(Node);
+        if (dep == erts_this_dist_entry)
+            break;
+
+        switch (erts_dsig_prepare(&dsd, dep, p,
+                                  ERTS_PROC_LOCK_MAIN,
+                                  ERTS_DSP_RLOCK, 0, async_connect)) {
+        case ERTS_DSIG_PREP_NOT_ALIVE:
+	case ERTS_DSIG_PREP_NOT_CONNECTED:
+	    /* Trap to either send 'nodedown' or do passive connection attempt */
+            goto do_trap;
+	case ERTS_DSIG_PREP_PENDING:
+	    if (!async_connect) {
+		/*
+		 * Pending connection may fail, so we must trap
+		 * to ensure passive connection attempt
+		 */
+		erts_de_runlock(dep);
+		goto do_trap;
 	    }
-	}
+	    /*fall through*/
+        case ERTS_DSIG_PREP_CONNECTED: {
+            ErtsMonitor *mon;
+            ErtsMonitorDataExtended *mdep;
+            int created;
+
+            mon = erts_monitor_tree_lookup_create(&ERTS_P_MONITORS(p),
+                                                  &created,
+                                                  ERTS_MON_TYPE_NODE,
+                                                  p->common.id,
+                                                  Node);
+            mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+            if (created) {
+#ifdef DEBUG
+                int inserted =
+#endif
+                    erts_monitor_dist_insert(&mdep->md.target, dep->mld);
+                ASSERT(inserted);
+                ASSERT(mdep->dist->connection_id == dep->connection_id);
+            }
+            else if (mdep->dist->connection_id != dep->connection_id) {
+                ErtsMonitorDataExtended *mdep2;
+                ErtsMonitor *mon2;
+#ifdef DEBUG
+                int inserted;
+#endif
+                mdep2 = ((ErtsMonitorDataExtended *)
+                         erts_monitor_create(ERTS_MON_TYPE_NODE, NIL,
+                                             p->common.id, Node, NIL));
+                mon2 = &mdep2->md.origin;
+#ifdef DEBUG
+                inserted =
+#endif
+                    erts_monitor_dist_insert(&mdep->md.target, dep->mld);
+                ASSERT(inserted);
+                ASSERT(mdep2->dist->connection_id == dep->connection_id);
+
+                mdep2->uptr.node_monitors = mdep->uptr.node_monitors;
+                mdep->uptr.node_monitors = NULL;
+                erts_monitor_tree_replace(&ERTS_P_MONITORS(p), mon, mon2);
+                erts_monitor_list_insert(&mdep2->uptr.node_monitors, mon);
+                mon->flags |= ERTS_ML_FLG_IN_SUBTABLE;
+                mdep = mdep2;
+            }
+
+            mdep->u.refc++;
+
+            break;
+        }
+
+        default:
+            ERTS_ASSERT(! "Invalid dsig prepare result");
+        }
+
+        erts_de_runlock(dep);
+
+        break;
     }
 
-    erts_smp_de_links_unlock(dep);
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
+    default:
+        goto badarg;
+    }
 
- done:
-    erts_deref_dist_entry(dep);
-    BIF_RET(am_true);
+    ERTS_BIF_PREP_RET(ret, am_true);
+
+do_return:
+
+    if (dep)
+        erts_deref_dist_entry(dep);
+
+    return ret;
+
+do_trap:
+    ERTS_BIF_PREP_TRAP3(ret, dmonitor_node_trap, p, Node, Bool, Options);
+    goto do_return;
+
+badarg:
+    ERTS_BIF_PREP_ERROR(ret, p, BADARG);
+    goto do_return;
 }
 
 BIF_RETTYPE monitor_node_3(BIF_ALIST_3)
@@ -3175,9 +4042,9 @@ BIF_RETTYPE net_kernel_dflag_unicode_io_1(BIF_ALIST_1)
     if (de == erts_this_dist_entry) {
 	BIF_RET(am_true);
     }
-    erts_smp_de_rlock(de);
+    erts_de_rlock(de);
     f = de->flags;
-    erts_smp_de_runlock(de);
+    erts_de_runlock(de);
     BIF_RET(((f & DFLAG_UNICODE_IO) ? am_true : am_false));
 }
     
@@ -3198,18 +4065,9 @@ BIF_RETTYPE net_kernel_dflag_unicode_io_1(BIF_ALIST_1)
 #define ERTS_NODES_MON_OPT_TYPES \
   (ERTS_NODES_MON_OPT_TYPE_VISIBLE|ERTS_NODES_MON_OPT_TYPE_HIDDEN)
 
-typedef struct ErtsNodesMonitor_ ErtsNodesMonitor;
-struct ErtsNodesMonitor_ {
-    ErtsNodesMonitor *prev;
-    ErtsNodesMonitor *next;
-    Process *proc;
-    Uint16 opts;
-    Uint16 no;
-};
-
-static erts_smp_mtx_t nodes_monitors_mtx;
-static ErtsNodesMonitor *nodes_monitors;
-static ErtsNodesMonitor *nodes_monitors_end;
+static erts_mtx_t nodes_monitors_mtx;
+static ErtsMonitor *nodes_monitors;
+static Uint no_nodes_monitors;
 
 /*
  * Nodes monitors are stored in a double linked list. 'nodes_monitors'
@@ -3225,112 +4083,172 @@ static ErtsNodesMonitor *nodes_monitors_end;
 static void
 init_nodes_monitors(void)
 {
-    erts_smp_mtx_init(&nodes_monitors_mtx, "nodes_monitors", NIL,
+    erts_mtx_init(&nodes_monitors_mtx, "nodes_monitors", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
     nodes_monitors = NULL;
-    nodes_monitors_end = NULL;
+    no_nodes_monitors = 0;
 }
 
-static ERTS_INLINE Uint
-nodes_mon_msg_sz(ErtsNodesMonitor *nmp, Eterm what, Eterm reason)
+Eterm
+erts_monitor_nodes(Process *c_p, Eterm on, Eterm olist)
 {
-    Uint sz;
-    if (!nmp->opts) {
-	sz = 3;
-    }
-    else {
-	sz = 0;
+    Eterm key, old_value, opts_list = olist;
+    Uint opts = (Uint) 0;
 
-	if (nmp->opts & ERTS_NODES_MON_OPT_TYPES)
-	    sz += 2 + 3;
+    ASSERT(c_p);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+
+    if (on != am_true && on != am_false)
+	return THE_NON_VALUE;
 
-	if (what == am_nodedown
-	    && (nmp->opts & ERTS_NODES_MON_OPT_DOWN_REASON)) {
-	    if (is_not_immed(reason))
-		sz += size_object(reason);
-	    sz += 2 + 3;
+    if (is_not_nil(opts_list)) {
+	int all = 0, visible = 0, hidden = 0;
+
+	while (is_list(opts_list)) {
+	    Eterm *cp = list_val(opts_list);
+	    Eterm opt = CAR(cp);
+	    opts_list = CDR(cp);
+	    if (opt == am_nodedown_reason)
+		opts |= ERTS_NODES_MON_OPT_DOWN_REASON;
+	    else if (is_tuple(opt)) {
+		Eterm* tp = tuple_val(opt);
+		if (arityval(tp[0]) != 2)
+		    return THE_NON_VALUE;
+		switch (tp[1]) {
+		case am_node_type:
+		    switch (tp[2]) {
+		    case am_visible:
+			if (hidden || all)
+			    return THE_NON_VALUE;
+			opts |= ERTS_NODES_MON_OPT_TYPE_VISIBLE;
+			visible = 1;
+			break;
+		    case am_hidden:
+			if (visible || all)
+			    return THE_NON_VALUE;
+			opts |= ERTS_NODES_MON_OPT_TYPE_HIDDEN;
+			hidden = 1;
+			break;
+		    case am_all:
+			if (visible || hidden)
+			    return THE_NON_VALUE;
+			opts |= ERTS_NODES_MON_OPT_TYPES;
+			all = 1;
+			break;
+		    default:
+			return THE_NON_VALUE;
+		    }
+		    break;
+		default:
+		    return THE_NON_VALUE;
+		}
+	    }
+	    else {
+		return THE_NON_VALUE;
+	    }
 	}
 
-	sz += 4;
+	if (is_not_nil(opts_list))
+	    return THE_NON_VALUE;
+    }
+
+    key = make_small(opts);
+
+    if (on == am_true) {
+        ErtsMonitorDataExtended *mdep;
+        ErtsMonitor *omon;
+        int created;
+        omon = erts_monitor_tree_lookup_create(&ERTS_P_MONITORS(c_p),
+                                               &created,
+                                               ERTS_MON_TYPE_NODES,
+                                               c_p->common.id,
+                                               key);
+        mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(omon);
+        if (created) {
+            erts_mtx_lock(&nodes_monitors_mtx);
+            no_nodes_monitors++;
+            erts_monitor_list_insert(&nodes_monitors, &mdep->md.target);
+            erts_mtx_unlock(&nodes_monitors_mtx);
+        }
+        old_value = mdep->u.refc;
+        mdep->u.refc++;
     }
-    return sz;
+    else {
+        ErtsMonitorDataExtended *mdep;
+        ErtsMonitor *omon;
+        omon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p), key);
+        if (!omon)
+            old_value = 0;
+        else {
+            mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(omon);
+            old_value = mdep->u.refc;
+            ASSERT(mdep->u.refc > 0);
+            erts_mtx_lock(&nodes_monitors_mtx);
+            ASSERT(no_nodes_monitors > 0);
+            no_nodes_monitors--;
+            ASSERT(erts_monitor_is_in_table(&mdep->md.target));
+            erts_monitor_list_delete(&nodes_monitors, &mdep->md.target);
+            erts_mtx_unlock(&nodes_monitors_mtx);
+            erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p), omon);
+            erts_monitor_release_both((ErtsMonitorData *) mdep);
+        }
+    }
+
+    return erts_make_integer(old_value, c_p);
 }
 
-static ERTS_INLINE void
-send_nodes_mon_msg(Process *rp,
-		   ErtsProcLocks *rp_locksp,
-		   ErtsNodesMonitor *nmp,
-		   Eterm node,
-		   Eterm what,
-		   Eterm type,
-		   Eterm reason,
-		   Uint sz)
+void
+erts_monitor_nodes_delete(ErtsMonitor *omon)
 {
-    Eterm msg;
-    Eterm *hp;
-    ErtsMessage *mp;
-    ErlOffHeap *ohp;
-#ifdef DEBUG
-    Eterm *hend;
-#endif
+    ErtsMonitorData *mdp;
 
-    mp = erts_alloc_message_heap(rp, rp_locksp, sz, &hp, &ohp);
-#ifdef DEBUG
-    hend = hp + sz;
-#endif
+    ASSERT(omon->type == ERTS_MON_TYPE_NODES);
+    ASSERT(erts_monitor_is_origin(omon));
 
-    if (!nmp->opts) {
-	msg = TUPLE2(hp, what, node);
-#ifdef DEBUG
-	hp += 3;
-#endif
-    }
-    else {
-	Eterm tup;
-	Eterm info = NIL;
+    mdp = erts_monitor_to_data(omon);
 
-	if (nmp->opts & (ERTS_NODES_MON_OPT_TYPE_VISIBLE
-			 | ERTS_NODES_MON_OPT_TYPE_HIDDEN)) {
+    erts_mtx_lock(&nodes_monitors_mtx);
+    ASSERT(erts_monitor_is_in_table(&mdp->target));
+    ASSERT(no_nodes_monitors > 0);
+    no_nodes_monitors--;
+    erts_monitor_list_delete(&nodes_monitors, &mdp->target);
+    erts_mtx_unlock(&nodes_monitors_mtx);
+    erts_monitor_release_both(mdp);
+}
 
-	    tup = TUPLE2(hp, am_node_type, type);
-	    hp += 3;
-	    info = CONS(hp, tup, info);
-	    hp += 2;
-	}
+typedef struct {
+    Eterm pid;
+    Eterm options;
+} ErtsNodesMonitorData;
 
-	if (what == am_nodedown
-	    && (nmp->opts & ERTS_NODES_MON_OPT_DOWN_REASON)) {
-	    Eterm rsn_cpy;
-	    
-	    if (is_immed(reason))
-		rsn_cpy = reason;
-	    else {
-		Eterm rsn_sz = size_object(reason);
-		rsn_cpy = copy_struct(reason, rsn_sz, &hp, ohp);
-	    }
+typedef struct {
+    ErtsNodesMonitorData *nmdp;
+    Uint i;
+} ErtsNodesMonitorContext;
 
-	    tup = TUPLE2(hp, am_nodedown_reason, rsn_cpy);
-	    hp += 3;
-	    info = CONS(hp, tup, info);
-	    hp += 2;
-	}
+static void
+save_nodes_monitor(ErtsMonitor *mon, void *vctxt)
+{
+    ErtsNodesMonitorContext *ctxt = vctxt;
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
 
-	msg = TUPLE3(hp, what, node, info);
-#ifdef DEBUG
-	hp += 4;
-#endif
-    }
+    ASSERT(erts_monitor_is_target(mon));
+    ASSERT(mon->type == ERTS_MON_TYPE_NODES);
+
+    ctxt->nmdp[ctxt->i].pid = mon->other.item;
+    ctxt->nmdp[ctxt->i].options = mdp->origin.other.item;
 
-    ASSERT(hend == hp);
-    erts_queue_message(rp, *rp_locksp, mp, msg, am_system);
+    ctxt->i++;
 }
 
 static void
 send_nodes_mon_msgs(Process *c_p, Eterm what, Eterm node, Eterm type, Eterm reason)
 {
-    ErtsNodesMonitor *nmp;
-    ErtsProcLocks rp_locks = 0; /* Init to shut up false warning */
-    Process *rp = NULL;
+    Uint opts;
+    Uint i, no, reason_size;
+    ErtsNodesMonitorData def_buf[100];
+    ErtsNodesMonitorData *nmdp = &def_buf[0];
+    ErtsNodesMonitorContext ctxt;
 
     ASSERT(is_immed(what));
     ASSERT(is_immed(node));
@@ -3351,31 +4269,44 @@ send_nodes_mon_msgs(Process *c_p, Eterm what, Eterm node, Eterm type, Eterm reas
     }
 #endif
 
-    ERTS_SMP_LC_ASSERT(!c_p
-		       || (erts_proc_lc_my_proc_locks(c_p)
-			   == ERTS_PROC_LOCK_MAIN));
-    erts_smp_mtx_lock(&nodes_monitors_mtx);
+    ctxt.i = 0;
 
-    for (nmp = nodes_monitors; nmp; nmp = nmp->next) {
-	int i;
-	Uint16 no;
-	Uint sz;
+    reason_size = is_immed(reason) ? 0 : size_object(reason);
 
-	ASSERT(nmp->proc != NULL);
+    erts_mtx_lock(&nodes_monitors_mtx);
+    if (no_nodes_monitors > sizeof(def_buf)/sizeof(def_buf[0]))
+        nmdp = erts_alloc(ERTS_ALC_T_TMP,
+                          no_nodes_monitors*sizeof(ErtsNodesMonitorData));
+    ctxt.nmdp = nmdp;
+    erts_monitor_list_foreach(nodes_monitors,
+                              save_nodes_monitor,
+                              (void *) &ctxt);
+    
+    ASSERT(ctxt.i == no_nodes_monitors);
+    no = no_nodes_monitors;
+
+    erts_mtx_unlock(&nodes_monitors_mtx);
 
-	if (!nmp->opts) {
+    for (i = 0; i < no; i++) {
+        Eterm tmp_heap[3+2+3+2+4 /* max need */];
+        Eterm *hp, msg;
+        Uint hsz;
+
+        ASSERT(is_small(nmdp[i].options));
+        opts = (Uint) signed_val(nmdp[i].options);
+	if (!opts) {
 	    if (type != am_visible)
 		continue;
 	}
 	else {
 	    switch (type) {
 	    case am_hidden:
-		if (!(nmp->opts & ERTS_NODES_MON_OPT_TYPE_HIDDEN))
+		if (!(opts & ERTS_NODES_MON_OPT_TYPE_HIDDEN))
 		    continue;
 		break;
 	    case am_visible:
-		if ((nmp->opts & ERTS_NODES_MON_OPT_TYPES)
-		    && !(nmp->opts & ERTS_NODES_MON_OPT_TYPE_VISIBLE))
+		if ((opts & ERTS_NODES_MON_OPT_TYPES)
+		    && !(opts & ERTS_NODES_MON_OPT_TYPE_VISIBLE))
 		    continue;
 		break;
 	    default:
@@ -3383,342 +4314,162 @@ send_nodes_mon_msgs(Process *c_p, Eterm what, Eterm node, Eterm type, Eterm reas
 	    }
 	}
 
-	if (rp != nmp->proc) {
-	    if (rp) {
-		if (rp == c_p)
-		    rp_locks &= ~ERTS_PROC_LOCK_MAIN;
-		erts_smp_proc_unlock(rp, rp_locks);
-	    }
-
-	    rp = nmp->proc;
-	    rp_locks = 0;
-	    if (rp == c_p)
-		rp_locks |= ERTS_PROC_LOCK_MAIN;
-	}
-
-	ASSERT(rp);
+        hsz = 0;
+        hp = &tmp_heap[0];
 
-	sz = nodes_mon_msg_sz(nmp, what, reason);
-
-	for (i = 0, no = nmp->no; i < no; i++)
-	    send_nodes_mon_msg(rp,
-			       &rp_locks,
-			       nmp,
-			       node,
-			       what,
-			       type,
-			       reason,
-			       sz);
-    }
+        if (!opts) {
+            msg = TUPLE2(hp, what, node);
+            hp += 3;
+        }
+        else {
+            Eterm tup;
+            Eterm info = NIL;
 
-    if (rp) {
-	if (rp == c_p)
-	    rp_locks &= ~ERTS_PROC_LOCK_MAIN;
-	erts_smp_proc_unlock(rp, rp_locks);
-    }
+            if (opts & (ERTS_NODES_MON_OPT_TYPE_VISIBLE
+                        | ERTS_NODES_MON_OPT_TYPE_HIDDEN)) {
 
-    erts_smp_mtx_unlock(&nodes_monitors_mtx);
-}
+                tup = TUPLE2(hp, am_node_type, type);
+                hp += 3;
+                info = CONS(hp, tup, info);
+                hp += 2;
+            }
 
-static Eterm
-insert_nodes_monitor(Process *c_p, Uint32 opts)
-{
-    Uint16 no = 1;
-    Eterm res = am_false;
-    ErtsNodesMonitor *xnmp, *nmp;
+            if (what == am_nodedown
+                && (opts & ERTS_NODES_MON_OPT_DOWN_REASON)) {
+                hsz += reason_size;
+                tup = TUPLE2(hp, am_nodedown_reason, reason);
+                hp += 3;
+                info = CONS(hp, tup, info);
+                hp += 2;
+            }
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&nodes_monitors_mtx));
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) & ERTS_PROC_LOCK_MAIN);
+            msg = TUPLE3(hp, what, node, info);
+            hp += 4;
+        }
 
-    xnmp = c_p->nodes_monitors;
-    if (xnmp) {
-	ASSERT(!xnmp->prev || xnmp->prev->proc != c_p);
+        ASSERT(hp - &tmp_heap[0] <= sizeof(tmp_heap)/sizeof(tmp_heap[0]));
 
-	while (1) {
-	    ASSERT(xnmp->proc == c_p);
-	    if (xnmp->opts == opts)
-		break;
-	    if (!xnmp->next || xnmp->next->proc != c_p)
-		break;
-	    xnmp = xnmp->next;
-	}
-	ASSERT(xnmp);
-	ASSERT(xnmp->proc == c_p);
-	ASSERT(xnmp->opts == opts
-	       || !xnmp->next
-	       || xnmp->next->proc != c_p);
-
-	if (xnmp->opts != opts)
-	    goto alloc_new;
-	else {
-	    res = am_true;
-	    no = xnmp->no++;
-	    if (!xnmp->no) {
-		/*
-		 * 'no' wrapped; transfer all prevous monitors to new
-		 * element (which will be the next element in the list)
-		 * and set this to one...
-		 */
-		xnmp->no = 1;
-		goto alloc_new;
-	    }
-	}
-    }
-    else {
-    alloc_new:
-	nmp = erts_alloc(ERTS_ALC_T_NODES_MON, sizeof(ErtsNodesMonitor));
-	nmp->proc = c_p;
-	nmp->opts = opts;
-	nmp->no = no;
-
-	if (xnmp) {
-	    ASSERT(nodes_monitors);
-	    ASSERT(c_p->nodes_monitors);
-	    nmp->next = xnmp->next;
-	    nmp->prev = xnmp;
-	    xnmp->next = nmp;
-	    if (nmp->next) {
-		ASSERT(nodes_monitors_end != xnmp);
-		ASSERT(nmp->next->prev == xnmp);
-		nmp->next->prev = nmp;
-	    }
-	    else {
-		ASSERT(nodes_monitors_end == xnmp);
-		nodes_monitors_end = nmp;
-	    }
-	}
-	else {
-	    ASSERT(!c_p->nodes_monitors);
-	    c_p->nodes_monitors = nmp;
-	    nmp->next = NULL;
-	    nmp->prev = nodes_monitors_end;
-	    if (nodes_monitors_end) {
-		ASSERT(nodes_monitors);
-		nodes_monitors_end->next = nmp;
-	    }
-	    else {
-		ASSERT(!nodes_monitors);
-		nodes_monitors = nmp;
-	    }
-	    nodes_monitors_end = nmp;
-	}
-    }
-    return res;
-}
+        hsz += hp - &tmp_heap[0];
 
-static Eterm
-remove_nodes_monitors(Process *c_p, Uint32 opts, int all)
-{
-    Eterm res = am_false;
-    ErtsNodesMonitor *nmp;
-
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&nodes_monitors_mtx));
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) & ERTS_PROC_LOCK_MAIN);
-
-    nmp = c_p->nodes_monitors;
-    ASSERT(!nmp || !nmp->prev || nmp->prev->proc != c_p);
-
-    while (nmp && nmp->proc == c_p) {
-	if (!all && nmp->opts != opts)
-	    nmp = nmp->next;
-	else { /* if (all || nmp->opts == opts) */
-	    ErtsNodesMonitor *free_nmp;
-	    res = am_true;
-	    if (nmp->prev) {
-		ASSERT(nodes_monitors != nmp);
-		nmp->prev->next = nmp->next;
-	    }
-	    else {
-		ASSERT(nodes_monitors == nmp);
-		nodes_monitors = nmp->next;
-	    }
-	    if (nmp->next) {
-		ASSERT(nodes_monitors_end != nmp);
-		nmp->next->prev = nmp->prev;
-	    }
-	    else {
-		ASSERT(nodes_monitors_end == nmp);
-		nodes_monitors_end = nmp->prev;
-	    }
-	    free_nmp = nmp;
-	    nmp = nmp->next;
-	    if (c_p->nodes_monitors == free_nmp)
-		c_p->nodes_monitors = nmp && nmp->proc == c_p ? nmp : NULL;
-	    erts_free(ERTS_ALC_T_NODES_MON, free_nmp);
-	}
+        erts_proc_sig_send_persistent_monitor_msg(ERTS_MON_TYPE_NODES,
+                                                  nmdp[i].options,
+                                                  am_system,
+                                                  nmdp[i].pid,
+                                                  msg,
+                                                  hsz);
     }
-    
-    ASSERT(!all || !c_p->nodes_monitors);
-    return res;
-}
 
-void
-erts_delete_nodes_monitors(Process *c_p, ErtsProcLocks locks)
-{
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
-    if (c_p) {
-	ErtsProcLocks might_unlock = locks & ~ERTS_PROC_LOCK_MAIN;
-	if (might_unlock)
-	    erts_proc_lc_might_unlock(c_p, might_unlock);
-    }
-#endif
-    if (erts_smp_mtx_trylock(&nodes_monitors_mtx) == EBUSY) {
-	ErtsProcLocks unlock_locks = locks & ~ERTS_PROC_LOCK_MAIN;
-	if (c_p && unlock_locks)
-	    erts_smp_proc_unlock(c_p, unlock_locks);
-	erts_smp_mtx_lock(&nodes_monitors_mtx);
-	if (c_p && unlock_locks)
-	    erts_smp_proc_lock(c_p, unlock_locks);
-    }
-    remove_nodes_monitors(c_p, 0, 1);
-    erts_smp_mtx_unlock(&nodes_monitors_mtx);
+    if (nmdp != &def_buf[0])
+        erts_free(ERTS_ALC_T_TMP, nmdp);
 }
+                           
 
-Eterm
-erts_monitor_nodes(Process *c_p, Eterm on, Eterm olist)
-{
+typedef struct {
+    Eterm **hpp;
+    Uint *szp;
     Eterm res;
-    Eterm opts_list = olist;
-    Uint16 opts = (Uint16) 0;
-
-    ASSERT(c_p);
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
-
-    if (on != am_true && on != am_false)
-	return THE_NON_VALUE;
+} ErtsNodesMonitorInfoContext;
 
-    if (is_not_nil(opts_list)) {
-	int all = 0, visible = 0, hidden = 0;
 
-	while (is_list(opts_list)) {
-	    Eterm *cp = list_val(opts_list);
-	    Eterm opt = CAR(cp);
-	    opts_list = CDR(cp);
-	    if (opt == am_nodedown_reason)
-		opts |= ERTS_NODES_MON_OPT_DOWN_REASON;
-	    else if (is_tuple(opt)) {
-		Eterm* tp = tuple_val(opt);
-		if (arityval(tp[0]) != 2)
-		    return THE_NON_VALUE;
-		switch (tp[1]) {
-		case am_node_type:
-		    switch (tp[2]) {
-		    case am_visible:
-			if (hidden || all)
-			    return THE_NON_VALUE;
-			opts |= ERTS_NODES_MON_OPT_TYPE_VISIBLE;
-			visible = 1;
-			break;
-		    case am_hidden:
-			if (visible || all)
-			    return THE_NON_VALUE;
-			opts |= ERTS_NODES_MON_OPT_TYPE_HIDDEN;
-			hidden = 1;
-			break;
-		    case am_all:
-			if (visible || hidden)
-			    return THE_NON_VALUE;
-			opts |= ERTS_NODES_MON_OPT_TYPES;
-			all = 1;
-			break;
-		    default:
-			return THE_NON_VALUE;
-		    }
-		    break;
-		default:
-		    return THE_NON_VALUE;
-		}
-	    }
-	    else {
-		return THE_NON_VALUE;
-	    }
-	}
-
-	if (is_not_nil(opts_list))
-	    return THE_NON_VALUE;
+static void
+nodes_monitor_info(ErtsMonitor *mon, void *vctxt)
+{
+    ErtsMonitorDataExtended *mdep;
+    ErtsNodesMonitorInfoContext *ctxt = vctxt;
+    Uint no, i, opts, *szp;
+    Eterm **hpp, res;
+
+    hpp = ctxt->hpp;
+    szp = ctxt->szp;
+    res = ctxt->res;
+
+    ASSERT(erts_monitor_is_target(mon));
+    ASSERT(mon->type == ERTS_MON_TYPE_NODES);
+    mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+    no = mdep->u.refc;
+
+    ASSERT(is_small(mdep->md.origin.other.item));
+    opts = (Uint) signed_val(mdep->md.origin.other.item);
+
+    for (i = 0; i < no; i++) {
+        Eterm olist = NIL;
+        if (opts & ERTS_NODES_MON_OPT_TYPES) {
+            Eterm type;
+            switch (opts & ERTS_NODES_MON_OPT_TYPES) {
+            case ERTS_NODES_MON_OPT_TYPES:        type = am_all;     break;
+            case ERTS_NODES_MON_OPT_TYPE_VISIBLE: type = am_visible; break;
+            case ERTS_NODES_MON_OPT_TYPE_HIDDEN:  type = am_hidden;  break;
+            default: erts_exit(ERTS_ABORT_EXIT, "Bad node type found\n");
+            }
+            olist = erts_bld_cons(hpp, szp, 
+                                  erts_bld_tuple(hpp, szp, 2,
+                                                 am_node_type,
+                                                 type),
+                                  olist);
+        }
+        if (opts & ERTS_NODES_MON_OPT_DOWN_REASON)
+            olist = erts_bld_cons(hpp, szp, am_nodedown_reason, olist);
+        res = erts_bld_cons(hpp, szp,
+                            erts_bld_tuple(hpp, szp, 2,
+                                           mon->other.item,
+                                           olist),
+                            res);
     }
 
-    erts_smp_mtx_lock(&nodes_monitors_mtx);
-
-    if (on == am_true)
-	res = insert_nodes_monitor(c_p, opts);
-    else
-	res = remove_nodes_monitors(c_p, opts, 0);
-
-    erts_smp_mtx_unlock(&nodes_monitors_mtx);
-
-    return res;
+    ctxt->hpp = hpp;
+    ctxt->szp = szp;
+    ctxt->res = res;
 }
 
-/*
- * Note, this function is only used for debuging.
- */
-
 Eterm
 erts_processes_monitoring_nodes(Process *c_p)
 {
-    ErtsNodesMonitor *nmp;
-    Eterm res;
+    /*
+     * Note, this function is only used for debugging.
+     */
+    ErtsNodesMonitorInfoContext ctxt;
     Eterm *hp;
-    Eterm **hpp;
     Uint sz;
-    Uint *szp;
 #ifdef DEBUG
     Eterm *hend;
 #endif
 
     ASSERT(c_p);
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
-    erts_smp_mtx_lock(&nodes_monitors_mtx);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
+
+    erts_mtx_lock(&nodes_monitors_mtx);
 
     sz = 0;
-    szp = &sz;
-    hpp = NULL;
+    ctxt.szp = &sz;
+    ctxt.hpp = NULL;
 
- bld_result:
-    res = NIL;
-
-    for (nmp = nodes_monitors_end; nmp; nmp = nmp->prev) {
-	Uint16 i;
-	for (i = 0; i < nmp->no; i++) {
-	    Eterm olist = NIL;
-	    if (nmp->opts & ERTS_NODES_MON_OPT_TYPES) {
-		Eterm type;
-		switch (nmp->opts & ERTS_NODES_MON_OPT_TYPES) {
-		case ERTS_NODES_MON_OPT_TYPES:        type = am_all;     break;
-		case ERTS_NODES_MON_OPT_TYPE_VISIBLE: type = am_visible; break;
-		case ERTS_NODES_MON_OPT_TYPE_HIDDEN:  type = am_hidden;  break;
-		default: erts_exit(ERTS_ABORT_EXIT, "Bad node type found\n");
-		}
-		olist = erts_bld_cons(hpp, szp, 
-				      erts_bld_tuple(hpp, szp, 2,
-						     am_node_type,
-						     type),
-				      olist);
-	    }
-	    if (nmp->opts & ERTS_NODES_MON_OPT_DOWN_REASON)
-		olist = erts_bld_cons(hpp, szp, am_nodedown_reason, olist);
-	    res = erts_bld_cons(hpp, szp,
-				erts_bld_tuple(hpp, szp, 2,
-					       nmp->proc->common.id,
-					       olist),
-				res);
-	}
-    }
+    while (1) {
+        ctxt.res = NIL;
+
+        erts_monitor_list_foreach(nodes_monitors,
+                                  nodes_monitor_info,
+                                  (void *) &ctxt);
+
+        if (ctxt.hpp)
+            break;
 
-    if (!hpp) {
 	hp = HAlloc(c_p, sz);
 #ifdef DEBUG
 	hend = hp + sz;
 #endif
-	hpp = &hp;
-	szp = NULL;
-	goto bld_result;
+	ctxt.hpp = &hp;
+	ctxt.szp = NULL;
     }
 
     ASSERT(hp == hend);
 
-    erts_smp_mtx_unlock(&nodes_monitors_mtx);
+    erts_mtx_unlock(&nodes_monitors_mtx);
 
-    return res;
+    erts_thr_progress_unblock();
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+
+    return ctxt.res;
 }
diff --git a/erts/emulator/beam/dist.h b/erts/emulator/beam/dist.h
index 3e17645997..dda2029a4c 100644
--- a/erts/emulator/beam/dist.h
+++ b/erts/emulator/beam/dist.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -40,10 +40,53 @@
 #define DFLAG_UNICODE_IO          0x1000
 #define DFLAG_DIST_HDR_ATOM_CACHE 0x2000
 #define DFLAG_SMALL_ATOM_TAGS     0x4000
-#define DFLAG_INTERNAL_TAGS       0x8000
+#define DFLAG_INTERNAL_TAGS       0x8000   /* used by ETS 'compressed' option */
 #define DFLAG_UTF8_ATOMS          0x10000
 #define DFLAG_MAP_TAG             0x20000
 #define DFLAG_BIG_CREATION        0x40000
+#define DFLAG_SEND_SENDER         0x80000
+#define DFLAG_BIG_SEQTRACE_LABELS 0x100000
+#define DFLAG_NO_MAGIC            0x200000 /* internal for pending connection */
+
+/* Mandatory flags for distribution */
+#define DFLAG_DIST_MANDATORY (DFLAG_EXTENDED_REFERENCES         \
+                              | DFLAG_EXTENDED_PIDS_PORTS       \
+			      | DFLAG_UTF8_ATOMS                \
+			      | DFLAG_NEW_FUN_TAGS)
+
+/*
+ * Additional optimistic flags when encoding toward pending connection.
+ * If remote node (erl_interface) does not supporting these then we may need
+ * to transcode messages enqueued before connection setup was finished.
+ */
+#define DFLAG_DIST_HOPEFULLY (DFLAG_EXPORT_PTR_TAG              \
+                              | DFLAG_BIT_BINARIES              \
+                              | DFLAG_DIST_MONITOR              \
+                              | DFLAG_DIST_MONITOR_NAME)
+
+/* Our preferred set of flags. Used for connection setup handshake */
+#define DFLAG_DIST_DEFAULT (DFLAG_DIST_MANDATORY | DFLAG_DIST_HOPEFULLY \
+                            | DFLAG_FUN_TAGS                  \
+                            | DFLAG_NEW_FLOATS                \
+                            | DFLAG_UNICODE_IO                \
+                            | DFLAG_DIST_HDR_ATOM_CACHE       \
+                            | DFLAG_SMALL_ATOM_TAGS           \
+                            | DFLAG_UTF8_ATOMS                \
+                            | DFLAG_MAP_TAG                   \
+                            | DFLAG_BIG_CREATION              \
+                            | DFLAG_SEND_SENDER               \
+                            | DFLAG_BIG_SEQTRACE_LABELS)
+
+/* Flags addable by local distr implementations */
+#define DFLAG_DIST_ADDABLE    DFLAG_DIST_DEFAULT
+
+/* Flags rejectable by local distr implementation */
+#define DFLAG_DIST_REJECTABLE (DFLAG_DIST_HDR_ATOM_CACHE         \
+                               | DFLAG_HIDDEN_ATOM_CACHE         \
+                               | DFLAG_ATOM_CACHE)
+
+/* Flags for all features needing strict order delivery */
+#define DFLAG_DIST_STRICT_ORDER DFLAG_DIST_HDR_ATOM_CACHE
 
 /* All flags that should be enabled when term_to_binary/1 is used. */
 #define TERM_TO_BINARY_DFLAGS (DFLAG_EXTENDED_REFERENCES	\
@@ -74,53 +117,37 @@
 #define DOP_DEMONITOR_P		20
 #define DOP_MONITOR_P_EXIT	21
 
+#define DOP_SEND_SENDER         22
+#define DOP_SEND_SENDER_TT      23
+
 /* distribution trap functions */
-extern Export* dsend2_trap;
-extern Export* dsend3_trap;
-extern Export* dlink_trap;
-extern Export* dunlink_trap;
 extern Export* dmonitor_node_trap;
-extern Export* dgroup_leader_trap;
-extern Export* dexit_trap;
-extern Export* dmonitor_p_trap;
 
 typedef enum {
     ERTS_DSP_NO_LOCK,
-    ERTS_DSP_RLOCK,
-    ERTS_DSP_RWLOCK
+    ERTS_DSP_RLOCK
 } ErtsDSigPrepLock;
 
 
 typedef struct {
     Process *proc;
     DistEntry *dep;
+    Eterm node;   /* used if dep == NULL */
     Eterm cid;
     Eterm connection_id;
     int no_suspend;
 } ErtsDSigData;
 
-#define ERTS_DE_IS_NOT_CONNECTED(DEP) \
-  (ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&(DEP)->rwmtx) \
-		      || erts_lc_rwmtx_is_rwlocked(&(DEP)->rwmtx)), \
-   (is_nil((DEP)->cid) || ((DEP)->status & ERTS_DE_SFLG_EXITING)))
-
-#define ERTS_DE_IS_CONNECTED(DEP) \
-  (!ERTS_DE_IS_NOT_CONNECTED((DEP)))
-
 #define ERTS_DE_BUSY_LIMIT (1024*1024)
 extern int erts_dist_buf_busy_limit;
 extern int erts_is_alive;
 
 /*
  * erts_dsig_prepare() prepares a send of a distributed signal.
- * One of the values defined below are returned. If the returned
- * value is another than ERTS_DSIG_PREP_CONNECTED, the
- * distributed signal cannot be sent before appropriate actions
- * have been taken. Appropriate actions would typically be setting
- * up the connection.
+ * One of the values defined below are returned.
  */
 
-/* Connected; signal can be sent. */
+/* Connected; signals can be enqueued and sent. */
 #define ERTS_DSIG_PREP_CONNECTED	0
 /* Not connected; connection needs to be set up. */
 #define ERTS_DSIG_PREP_NOT_CONNECTED	1
@@ -128,46 +155,80 @@ extern int erts_is_alive;
 #define ERTS_DSIG_PREP_WOULD_SUSPEND	2
 /* System not alive (distributed) */
 #define ERTS_DSIG_PREP_NOT_ALIVE	3
+/* Pending connection; signals can be enqueued */
+#define ERTS_DSIG_PREP_PENDING	        4
 
 ERTS_GLB_INLINE int erts_dsig_prepare(ErtsDSigData *,
-				      DistEntry *,
+				      DistEntry*,
 				      Process *,
+                                      ErtsProcLocks,
 				      ErtsDSigPrepLock,
+				      int,
 				      int);
 
 ERTS_GLB_INLINE
 void erts_schedule_dist_command(Port *, DistEntry *);
 
+int erts_auto_connect(DistEntry* dep, Process *proc, ErtsProcLocks proc_locks);
+
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
 ERTS_GLB_INLINE int 
 erts_dsig_prepare(ErtsDSigData *dsdp,
 		  DistEntry *dep,
 		  Process *proc,
+                  ErtsProcLocks proc_locks,
 		  ErtsDSigPrepLock dspl,
-		  int no_suspend)
+		  int no_suspend,
+		  int connect)
 {
-    int failure;
+    int res;
+
     if (!erts_is_alive)
 	return ERTS_DSIG_PREP_NOT_ALIVE;
-    if (!dep)
-	return ERTS_DSIG_PREP_NOT_CONNECTED;
-    if (dspl == ERTS_DSP_RWLOCK)
-	erts_smp_de_rwlock(dep);
-    else
-	erts_smp_de_rlock(dep);
-    if (ERTS_DE_IS_NOT_CONNECTED(dep)) {
-	failure = ERTS_DSIG_PREP_NOT_CONNECTED;
+    if (!dep) {
+        ASSERT(!connect);
+        return ERTS_DSIG_PREP_NOT_CONNECTED;
+    }
+
+#ifdef ERTS_ENABLE_LOCK_CHECK
+    if (connect) {
+        erts_proc_lc_might_unlock(proc, proc_locks);
+    }
+#endif
+
+retry:
+    erts_de_rlock(dep);
+
+    if (dep->state == ERTS_DE_STATE_CONNECTED) {
+	res = ERTS_DSIG_PREP_CONNECTED;
+    }
+    else if (dep->state == ERTS_DE_STATE_PENDING) {
+	res = ERTS_DSIG_PREP_PENDING;
+    }
+    else if (dep->state == ERTS_DE_STATE_EXITING) {
+	res = ERTS_DSIG_PREP_NOT_CONNECTED;
 	goto fail;
     }
+    else if (connect) {
+        ASSERT(dep->state == ERTS_DE_STATE_IDLE);
+        erts_de_runlock(dep);
+        if (!erts_auto_connect(dep, proc, proc_locks)) {
+            return ERTS_DSIG_PREP_NOT_ALIVE;
+        }
+	goto retry;
+    }
+    else {
+        ASSERT(dep->state == ERTS_DE_STATE_IDLE);
+	res = ERTS_DSIG_PREP_NOT_CONNECTED;
+	goto fail;
+    }
+
     if (no_suspend) {
-	failure = ERTS_DSIG_PREP_CONNECTED;
-	erts_smp_mtx_lock(&dep->qlock);
-	if (dep->qflgs & ERTS_DE_QFLG_BUSY)
-	    failure = ERTS_DSIG_PREP_WOULD_SUSPEND;
-	erts_smp_mtx_unlock(&dep->qlock);
-	if (failure == ERTS_DSIG_PREP_WOULD_SUSPEND)
+	if (erts_atomic32_read_acqb(&dep->qflgs) & ERTS_DE_QFLG_BUSY) {
+	    res = ERTS_DSIG_PREP_WOULD_SUSPEND;
 	    goto fail;
+	}
     }
     dsdp->proc = proc;
     dsdp->dep = dep;
@@ -175,16 +236,12 @@ erts_dsig_prepare(ErtsDSigData *dsdp,
     dsdp->connection_id = dep->connection_id;
     dsdp->no_suspend = no_suspend;
     if (dspl == ERTS_DSP_NO_LOCK)
-	erts_smp_de_runlock(dep);
-    return ERTS_DSIG_PREP_CONNECTED;
+	erts_de_runlock(dep);
+    return res;
 
  fail:
-    if (dspl == ERTS_DSP_RWLOCK)
-	erts_smp_de_rwunlock(dep);
-    else
-	erts_smp_de_runlock(dep);
-    return failure;
-
+    erts_de_runlock(dep);
+    return res;
 }
 
 ERTS_GLB_INLINE
@@ -194,7 +251,7 @@ void erts_schedule_dist_command(Port *prt, DistEntry *dist_entry)
     Eterm id;
 
     if (prt) {
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 	ASSERT((erts_atomic32_read_nob(&prt->state)
 		& ERTS_PORT_SFLGS_DEAD) == 0);
 	ASSERT(prt->dist_entry);
@@ -204,7 +261,7 @@ void erts_schedule_dist_command(Port *prt, DistEntry *dist_entry)
     }
     else {
 	ASSERT(dist_entry);
-	ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&dist_entry->rwmtx)
+	ERTS_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&dist_entry->rwmtx)
 			   || erts_lc_rwmtx_is_rwlocked(&dist_entry->rwmtx));
 	ASSERT(is_internal_port(dist_entry->cid));
 
@@ -212,64 +269,19 @@ void erts_schedule_dist_command(Port *prt, DistEntry *dist_entry)
 	id = dep->cid;
     }
 
-    if (!erts_smp_atomic_xchg_mb(&dep->dist_cmd_scheduled, 1))
+    if (!erts_atomic_xchg_mb(&dep->dist_cmd_scheduled, 1))
 	erts_port_task_schedule(id, &dep->dist_cmd, ERTS_PORT_TASK_DIST_CMD);
 }
 
 #endif
 
-typedef struct {
-    ErtsLink *d_lnk;
-    ErtsLink *d_sub_lnk;
-} ErtsDistLinkData;
-
-ERTS_GLB_INLINE void erts_remove_dist_link(ErtsDistLinkData *,
-					   Eterm,
-					   Eterm,
-					   DistEntry *);
-ERTS_GLB_INLINE int erts_was_dist_link_removed(ErtsDistLinkData *);
-ERTS_GLB_INLINE void erts_destroy_dist_link(ErtsDistLinkData *);
-
-#if ERTS_GLB_INLINE_INCL_FUNC_DEF
-
-ERTS_GLB_INLINE void
-erts_remove_dist_link(ErtsDistLinkData *dldp,
-		      Eterm lid,
-		      Eterm rid,
-		      DistEntry *dep)
-{
-    erts_smp_de_links_lock(dep);
-    dldp->d_lnk = erts_lookup_link(dep->nlinks, lid);
-    if (!dldp->d_lnk)
-	dldp->d_sub_lnk = NULL;
-    else {
-	dldp->d_sub_lnk = erts_remove_link(&ERTS_LINK_ROOT(dldp->d_lnk), rid);
-	dldp->d_lnk = (ERTS_LINK_ROOT(dldp->d_lnk)
-		       ? NULL
-		       : erts_remove_link(&dep->nlinks, lid));
-    }
-    erts_smp_de_links_unlock(dep);
-}
-
-ERTS_GLB_INLINE int
-erts_was_dist_link_removed(ErtsDistLinkData *dldp)
-{
-    return dldp->d_sub_lnk != NULL;
-}
-
-ERTS_GLB_INLINE void
-erts_destroy_dist_link(ErtsDistLinkData *dldp)
-{
-    if (dldp->d_lnk)
-	erts_destroy_link(dldp->d_lnk);
-    if (dldp->d_sub_lnk)
-	erts_destroy_link(dldp->d_sub_lnk);	
-}
-
+#ifdef DEBUG
+#define ERTS_DBG_CHK_NO_DIST_LNK(D, R, L) \
+    erts_dbg_chk_no_dist_proc_link((D), (R), (L))
+#else
+#define ERTS_DBG_CHK_NO_DIST_LNK(D, R, L)
 #endif
 
-
-
 /* Define for testing */
 /* #define EXTREME_TTB_TRAPPING 1 */
 
@@ -290,6 +302,7 @@ typedef struct TTBSizeContext_ {
 
 typedef struct TTBEncodeContext_ {
     Uint flags;
+    Uint hopefull_flags;
     int level;
     byte* ep;
     Eterm obj;
@@ -331,7 +344,7 @@ struct erts_dsig_send_context {
     Eterm ctl;
     Eterm msg;
     int force_busy;
-    Uint32 pass_through_size;
+    Uint32 max_finalize_prepend;
     Uint data_size, dhdr_ext_size;
     ErtsAtomCacheMap *acmp;
     ErtsDistOutputBuf *obuf;
@@ -345,10 +358,12 @@ struct erts_dsig_send_context {
 
 typedef struct {
     int suspend;
+    int connect;
 
     Eterm ctl_heap[6];
     ErtsDSigData dsd;
-    DistEntry* dep_to_deref;
+    DistEntry *dep;
+    int deref_dep;
     struct erts_dsig_send_context dss;
 
     Eterm return_term;
diff --git a/erts/emulator/beam/erl_afit_alloc.c b/erts/emulator/beam/erl_afit_alloc.c
index 4ebe37ee1d..23efe3bba4 100644
--- a/erts/emulator/beam/erl_afit_alloc.c
+++ b/erts/emulator/beam/erl_afit_alloc.c
@@ -181,7 +181,7 @@ static struct {
 
 static void ERTS_INLINE atom_init(Eterm *atom, char *name)
 {
-    *atom = am_atom_put(name, strlen(name));
+    *atom = am_atom_put(name, sys_strlen(name));
 }
 #define AM_INIT(AM) atom_init(&am.AM, #AM)
 
diff --git a/erts/emulator/beam/erl_alloc.c b/erts/emulator/beam/erl_alloc.c
index 928a91baf4..575d6ca867 100644
--- a/erts/emulator/beam/erl_alloc.c
+++ b/erts/emulator/beam/erl_alloc.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2002-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2002-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -38,9 +38,9 @@
 #include "erl_db.h"
 #include "erl_binary.h"
 #include "erl_bits.h"
-#include "erl_instrument.h"
+#include "erl_mtrace.h"
 #include "erl_mseg.h"
-#include "erl_monitors.h"
+#include "erl_monitor_link.h"
 #include "erl_hl_timer.h"
 #include "erl_cpu_topology.h"
 #include "erl_thr_queue.h"
@@ -83,14 +83,6 @@
 #define ERTS_ALC_DEFAULT_ACUL_EHEAP_ALLOC ERTS_ALC_DEFAULT_ENABLED_ACUL_EHEAP_ALLOC
 #define ERTS_ALC_DEFAULT_ACUL_LL_ALLOC ERTS_ALC_DEFAULT_ENABLED_ACUL_LL_ALLOC
 
-#ifndef ERTS_SMP
-#  undef ERTS_ALC_DEFAULT_ACUL
-#  define ERTS_ALC_DEFAULT_ACUL 0
-#  undef ERTS_ALC_DEFAULT_ACUL_EHEAP_ALLOC
-#  define ERTS_ALC_DEFAULT_ACUL_EHEAP_ALLOC 0
-#  undef ERTS_ALC_DEFAULT_ACUL_LL_ALLOC
-#  define ERTS_ALC_DEFAULT_ACUL_LL_ALLOC 0
-#endif
 
 #ifdef DEBUG
 static Uint install_debug_functions(void);
@@ -148,7 +140,7 @@ enum {
 };
 
 typedef struct {
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
     int only_sz;
     int internal;
     Uint req_sched;
@@ -210,8 +202,6 @@ typedef struct {
     int top_pad;
     AlcUInit_t alloc_util;
     struct {
-	int stat;
-	int map;
 	char *mtrace;
 	char *nodename;
     } instr;
@@ -374,16 +364,9 @@ set_default_exec_alloc_opts(struct au_init *ip)
     ip->init.util.rmbcmt	= 0;
     ip->init.util.acul		= 0;
 
-# ifdef ERTS_HAVE_EXEC_MMAPPER
-    ip->init.util.mseg_alloc    = &erts_alcu_mmapper_mseg_alloc;
-    ip->init.util.mseg_realloc  = &erts_alcu_mmapper_mseg_realloc;
-    ip->init.util.mseg_dealloc  = &erts_alcu_mmapper_mseg_dealloc;
-    ip->init.util.mseg_mmapper  = &erts_exec_mmapper;
-# else
     ip->init.util.mseg_alloc    = &erts_alcu_exec_mseg_alloc;
     ip->init.util.mseg_realloc  = &erts_alcu_exec_mseg_realloc;
     ip->init.util.mseg_dealloc  = &erts_alcu_exec_mseg_dealloc;
-# endif
 }
 #endif /* ERTS_ALC_A_EXEC */
 
@@ -393,6 +376,7 @@ set_default_temp_alloc_opts(struct au_init *ip)
     SET_DEFAULT_ALLOC_OPTS(ip);
     ip->enable			= AU_ALLOC_DEFAULT_ENABLE(1);
     ip->thr_spec		= 1;
+    ip->disable_allowed         = 0;
     ip->carrier_migration_allowed = 0;
     ip->atype			= AFIT;
     ip->init.util.name_prefix	= "temp_";
@@ -442,6 +426,7 @@ set_default_binary_alloc_opts(struct au_init *ip)
 #endif
     ip->init.util.ts 		= ERTS_ALC_MTA_BINARY;
     ip->init.util.acul		= ERTS_ALC_DEFAULT_ACUL;
+    ip->init.util.atags		= 1;
 }
 
 static void
@@ -478,6 +463,7 @@ set_default_driver_alloc_opts(struct au_init *ip)
 #endif
     ip->init.util.ts 		= ERTS_ALC_MTA_DRIVER;
     ip->init.util.acul		= ERTS_ALC_DEFAULT_ACUL;
+    ip->init.util.atags		= 1;
 }
 
 static void
@@ -515,6 +501,7 @@ set_default_test_alloc_opts(struct au_init *ip)
     ip->init.util.mmbcs 	= 0; /* Main carrier size */
     ip->init.util.ts 		= ERTS_ALC_MTA_TEST;
     ip->init.util.acul		= ERTS_ALC_DEFAULT_ACUL;
+    ip->init.util.atags		= 1;
 
     /* Use a constant minimal MBC size */
 #if ERTS_SA_MB_CARRIERS
@@ -529,7 +516,6 @@ set_default_test_alloc_opts(struct au_init *ip)
 }
 
 
-#ifdef ERTS_SMP
 
 static void
 adjust_tpref(struct au_init *ip, int no_sched)
@@ -552,7 +538,6 @@ adjust_tpref(struct au_init *ip, int no_sched)
     }
 }
 
-#endif
 
 static void handle_args(int *, char **, erts_alc_hndl_args_init_t *);
 
@@ -581,7 +566,6 @@ static void adjust_fix_alloc_sizes(UWord extra_block_size)
     if (extra_block_size && erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].enabled) {
 	int j;
 
-#ifdef ERTS_SMP
 	if (erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].thr_spec) {
 	    int i;
 	    ErtsAllocatorThrSpec_t* tspec;
@@ -597,7 +581,6 @@ static void adjust_fix_alloc_sizes(UWord extra_block_size)
 	    }
 	}
 	else
-#endif
 	{
 	    Allctr_t* allctr = erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].extra;
 	    for (j=0; j < ERTS_ALC_NO_FIXED_SIZES; ++j) {
@@ -620,7 +603,6 @@ strategy_support_carrier_migration(struct au_init *auip)
 static ERTS_INLINE void
 adjust_carrier_migration_support(struct au_init *auip)
 {
-#ifdef ERTS_SMP
     if (auip->init.util.acul) {
 	auip->thr_spec = -1; /* Need thread preferred */
 
@@ -635,9 +617,6 @@ adjust_carrier_migration_support(struct au_init *auip)
 	    auip->init.aoff.blk_order = FF_BF;
 	}
     }
-#else
-    auip->init.util.acul = 0;
-#endif
 }
 
 void
@@ -658,22 +637,18 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
 
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_PROC)]
 	= sizeof(Process);
-    fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_MONITOR_SH)]
-	= ERTS_MONITOR_SH_SIZE * sizeof(Uint);
-    fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_NLINK_SH)]
-	= ERTS_LINK_SH_SIZE * sizeof(Uint);
-    fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_DRV_EV_D_STATE)]
-	= sizeof(ErtsDrvEventDataState);
+    fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_MONITOR)]
+	= sizeof(ErtsMonitorDataHeap);
+    fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_LINK)]
+	= sizeof(ErtsLinkData);
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_DRV_SEL_D_STATE)]
 	= sizeof(ErtsDrvSelectDataState);
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_NIF_SEL_D_STATE)]
         = sizeof(ErtsNifSelectDataState);
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_MSG_REF)]
 	= sizeof(ErtsMessageRef);
-#ifdef ERTS_SMP
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_THR_Q_EL_SL)]
 	= sizeof(ErtsThrQElement_t);
-#endif
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_LL_PTIMER)]
 	= erts_timer_type_size(ERTS_ALC_T_LL_PTIMER);
     fix_type_sizes[ERTS_ALC_FIX_TYPE_IX(ERTS_ALC_T_HL_PTIMER)]
@@ -736,20 +711,6 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
 #endif
     }
 
-#ifndef ERTS_SMP
-    init.sl_alloc.thr_spec = 0;
-    init.std_alloc.thr_spec = 0;
-    init.ll_alloc.thr_spec = 0;
-    init.eheap_alloc.thr_spec = 0;
-    init.binary_alloc.thr_spec = 0;
-    init.ets_alloc.thr_spec = 0;
-    init.driver_alloc.thr_spec = 0;
-    init.fix_alloc.thr_spec = 0;
-    init.literal_alloc.thr_spec = 0;
-#ifdef ERTS_ALC_A_EXEC
-    init.exec_alloc.thr_spec = 0;
-#endif
-#endif
 
     /* Make adjustments for carrier migration support */
     init.temp_alloc.init.util.acul = 0;
@@ -800,7 +761,6 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
 #endif
     }
 
-#ifdef ERTS_SMP
     /* Only temp_alloc can use thread specific interface */
     if (init.temp_alloc.thr_spec)
 	init.temp_alloc.thr_spec = erts_no_schedulers;
@@ -819,10 +779,6 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
     adjust_tpref(&init.exec_alloc, erts_no_schedulers);
 #endif
 
-#else
-    /* No thread specific if not smp */
-    init.temp_alloc.thr_spec = 0;
-#endif
 
     /*
      * The following allocators cannot be run with afit strategy.
@@ -841,10 +797,8 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
     refuse_af_strategy(&init.exec_alloc);
 #endif
 
-#ifdef ERTS_SMP 
     if (!init.temp_alloc.thr_spec)
 	refuse_af_strategy(&init.temp_alloc);
-#endif
 
     erts_mtrace_pre_init();
 #if HAVE_ERTS_MSEG
@@ -953,7 +907,6 @@ erts_alloc_init(int *argc, char **argv, ErtsAllocInitOpts *eaiop)
 		       &test_alloc_state);
 
     erts_mtrace_install_wrapper_functions();
-    extra_block_size += erts_instr_init(init.instr.stat, init.instr.map);
 
     init_aireq_alloc();
 
@@ -1008,8 +961,6 @@ set_au_allocator(ErtsAlcType_t alctr_n, struct au_init *init, int ncpu)
 	return;
     }
 
-#ifdef USE_THREADS
-#ifdef ERTS_SMP
     if (init->thr_spec) {
 	if (init->thr_spec > 0) {
 	    af->alloc = erts_alcu_alloc_thr_spec;
@@ -1039,7 +990,6 @@ set_au_allocator(ErtsAlcType_t alctr_n, struct au_init *init, int ncpu)
 	ai->thr_spec	= tspec->size;
     }
     else
-#endif
 	if (init->init.util.ts) {
 	af->alloc = erts_alcu_alloc_ts;
 	if (init->init.util.fix_type_size)
@@ -1051,21 +1001,9 @@ set_au_allocator(ErtsAlcType_t alctr_n, struct au_init *init, int ncpu)
 	af->free = erts_alcu_free_ts;
     }
     else
-#endif
     {
-#ifdef ERTS_SMP
         erts_exit(ERTS_ABORT_EXIT, "%salloc is not thread safe\n",
                  init->init.util.name_prefix);
-#else
-	af->alloc = erts_alcu_alloc;
-	if (init->init.util.fix_type_size)
-	    af->realloc = erts_realloc_fixed_size;
-	else if (init->init.util.ramv)
-	    af->realloc = erts_alcu_realloc_mv;
-	else
-	    af->realloc = erts_alcu_realloc;
-	af->free = erts_alcu_free;
-#endif
     }
     af->extra	= NULL;
     ai->alloc_util	= 1;
@@ -1274,9 +1212,9 @@ get_bool_value(char *param_end, char** argv, int* ip)
 {
     char *param = argv[*ip]+1;
     char *value = get_value(param_end, argv, ip);
-    if (strcmp(value, "true") == 0)
+    if (sys_strcmp(value, "true") == 0)
 	return 1;
-    else if (strcmp(value, "false") == 0)
+    else if (sys_strcmp(value, "false") == 0)
 	return 0;
     else
 	bad_value(param, param_end, value);
@@ -1424,46 +1362,46 @@ handle_au_arg(struct au_init *auip,
 	}
 	else if(has_prefix("as", sub_param)) {
 	    char *alg = get_value(sub_param + 2, argv, ip);
-	    if (strcmp("bf", alg) == 0) {
+	    if (sys_strcmp("bf", alg) == 0) {
 		auip->atype = BESTFIT;
 		auip->init.bf.ao = 0;
 	    }
-	    else if (strcmp("aobf", alg) == 0) {
+	    else if (sys_strcmp("aobf", alg) == 0) {
 		auip->atype = BESTFIT;
 		auip->init.bf.ao = 1;
 	    }
-	    else if (strcmp("gf", alg) == 0) {
+	    else if (sys_strcmp("gf", alg) == 0) {
 		auip->atype = GOODFIT;
 	    }
-	    else if (strcmp("af", alg) == 0) {
+	    else if (sys_strcmp("af", alg) == 0) {
 		auip->atype = AFIT;
 	    }
-	    else if (strcmp("aoff", alg) == 0) {
+	    else if (sys_strcmp("aoff", alg) == 0) {
 		auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AOFF;
 		auip->init.aoff.blk_order = FF_AOFF;
 	    }
-	    else if (strcmp("aoffcbf", alg) == 0) {
+	    else if (sys_strcmp("aoffcbf", alg) == 0) {
 		auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AOFF;
 		auip->init.aoff.blk_order = FF_BF;
 	    }
-	    else if (strcmp("aoffcaobf", alg) == 0) {
+	    else if (sys_strcmp("aoffcaobf", alg) == 0) {
 		auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AOFF;
 		auip->init.aoff.blk_order = FF_AOBF;
 	    }
-            else if (strcmp("ageffcaoff", alg) == 0) {
+            else if (sys_strcmp("ageffcaoff", alg) == 0) {
                 auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AGEFF;
 		auip->init.aoff.blk_order = FF_AOFF;
             }
-            else if (strcmp("ageffcbf", alg) == 0) {
+            else if (sys_strcmp("ageffcbf", alg) == 0) {
                 auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AGEFF;
 		auip->init.aoff.blk_order = FF_BF;
             }
-            else if (strcmp("ageffcaobf", alg) == 0) {
+            else if (sys_strcmp("ageffcaobf", alg) == 0) {
                 auip->atype = FIRSTFIT;
 		auip->init.aoff.crr_order = FF_AGEFF;
 		auip->init.aoff.blk_order = FF_AOBF;
@@ -1473,7 +1411,9 @@ handle_au_arg(struct au_init *auip,
 	    }
 	    if (!strategy_support_carrier_migration(auip))
 		auip->init.util.acul = 0;
-	}
+	} else if (has_prefix("atags", sub_param)) {
+            auip->init.util.atags = get_bool_value(sub_param + 5, argv, ip);
+        }
 	else
 	    goto bad_switch;
 	break;
@@ -1590,8 +1530,7 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 	&init->ll_alloc,
 	&init->driver_alloc,
 	&init->fix_alloc,
-	&init->sl_alloc,
-	&init->temp_alloc
+	&init->sl_alloc
 	/* test_alloc not affected by +Mea??? or +Mu???  */
     };
     int aui_sz = (int) sizeof(aui)/sizeof(aui[0]);
@@ -1624,10 +1563,8 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 		    break;
                 case 'X':
                     if (has_prefix("scs", argv[i]+3)) {
-#ifdef ERTS_HAVE_EXEC_MMAPPER
-                        init->mseg.exec_mmap.scs =
-#endif
-                            get_mb_value(argv[i]+6, argv, &i);
+                        /* Ignore obsolete */
+                        (void) get_mb_value(argv[i]+6, argv, &i);
                     }
                     else
                         handle_au_arg(&init->exec_alloc, &argv[i][3], argv, &i, 0);
@@ -1744,7 +1681,7 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 		    }
 		    else if (has_prefix("e", param+2)) {
 			arg = get_value(param+3, argv, &i);
-			if (strcmp("true", arg) != 0)
+			if (sys_strcmp("true", arg) != 0)
 			    bad_value(param, param+3, arg);
 		    }
 		    else
@@ -1756,20 +1693,20 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 		    case 'a': {
 			int a;
 			arg = get_value(argv[i]+4, argv, &i);
-			if (strcmp("min", arg) == 0) {
+			if (sys_strcmp("min", arg) == 0) {
 			    for (a = 0; a < aui_sz; a++)
 				aui[a]->enable = 0;
 			}
-			else if (strcmp("max", arg) == 0) {
+			else if (sys_strcmp("max", arg) == 0) {
 			    for (a = 0; a < aui_sz; a++)
 				aui[a]->enable = 1;
 			}
-			else if (strcmp("config", arg) == 0) {
+			else if (sys_strcmp("config", arg) == 0) {
 			    init->erts_alloc_config = 1;
 			}
-			else if (strcmp("r9c", arg) == 0
-				 || strcmp("r10b", arg) == 0
-				 || strcmp("r11b", arg) == 0) {
+			else if (sys_strcmp("r9c", arg) == 0
+				 || sys_strcmp("r10b", arg) == 0
+				 || sys_strcmp("r11b", arg) == 0) {
 			    set_default_sl_alloc_opts(&init->sl_alloc);
 			    set_default_std_alloc_opts(&init->std_alloc);
 			    set_default_ll_alloc_opts(&init->ll_alloc);
@@ -1781,7 +1718,7 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 			    set_default_driver_alloc_opts(&init->fix_alloc);
 
 			    init->driver_alloc.enable = 0;
-			    if (strcmp("r9c", arg) == 0) {
+			    if (sys_strcmp("r9c", arg) == 0) {
 				init->sl_alloc.enable = 0;
 				init->std_alloc.enable = 0;
 				init->binary_alloc.enable = 0;
@@ -1806,24 +1743,6 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 		    break;
 		case 'i':
 		    switch (argv[i][3]) {
-		    case 's':
-			arg = get_value(argv[i]+4, argv, &i);
-			if (strcmp("true", arg) == 0)
-			    init->instr.stat = 1;
-			else if (strcmp("false", arg) == 0)
-			    init->instr.stat = 0;
-			else
-			    bad_value(param, param+3, arg);
-			break;
-		    case 'm':
-			arg = get_value(argv[i]+4, argv, &i);
-			if (strcmp("true", arg) == 0)
-			    init->instr.map = 1;
-			else if (strcmp("false", arg) == 0)
-			    init->instr.map = 0;
-			else
-			    bad_value(param, param+3, arg);
-			break;
 		    case 't':
 			init->instr.mtrace = get_value(argv[i]+4, argv, &i);
 			break;
@@ -1834,9 +1753,9 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 		case 'l':
 		    if (has_prefix("pm", param+2)) {
 			arg = get_value(argv[i]+5, argv, &i);
-			if (strcmp("all", arg) == 0)
+			if (sys_strcmp("all", arg) == 0)
 			    lock_all_physical_memory = 1;
-			else if (strcmp("no", arg) == 0)
+			else if (sys_strcmp("no", arg) == 0)
 			    lock_all_physical_memory = 0;
 			else
 			    bad_value(param, param+4, arg);
@@ -1882,12 +1801,10 @@ handle_args(int *argc, char **argv, erts_alc_hndl_args_init_t *init)
 	    case '-':
 		if (argv[i][2] == '\0') {
 		    /* End of system flags reached */
-		    if (init->instr.mtrace
-			/* || init->instr.stat
-			   || init->instr.map */) {
+		    if (init->instr.mtrace) {
 			while (i < *argc) {
-			    if(strcmp(argv[i], "-sname") == 0
-			       || strcmp(argv[i], "-name") == 0) {
+			    if(sys_strcmp(argv[i], "-sname") == 0
+			       || sys_strcmp(argv[i], "-name") == 0) {
 				if (i + 1 <*argc) {
 				    init->instr.nodename = argv[i+1];
 				    break;
@@ -1937,9 +1854,7 @@ erts_alloc_register_scheduler(void *vesdp)
     int ix = (int) esdp->no;
     int aix;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
-#endif
     for (aix = ERTS_ALC_A_MIN; aix <= ERTS_ALC_A_MAX; aix++) {
 	ErtsAllocatorThrSpec_t *tspec = &erts_allctr_thr_spec[aix];
 	esdp->alloc_data.deallctr[aix] = NULL;
@@ -1957,7 +1872,6 @@ erts_alloc_register_scheduler(void *vesdp)
     }
 }
 
-#ifdef ERTS_SMP
 void
 erts_alloc_scheduler_handle_delayed_dealloc(void *vesdp,
 					    int *need_thr_progress,
@@ -1986,12 +1900,10 @@ erts_alloc_scheduler_handle_delayed_dealloc(void *vesdp,
 	}
     }
 }
-#endif
 
 erts_aint32_t
 erts_alloc_fix_alloc_shrink(int ix, erts_aint32_t flgs)
 {
-#ifdef ERTS_SMP
     ErtsAllocatorThrSpec_t *tspec;
     tspec = &erts_allctr_thr_spec[ERTS_ALC_A_FIXED_SIZE];
     if (erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].thr_spec && tspec->enabled)
@@ -1999,11 +1911,6 @@ erts_alloc_fix_alloc_shrink(int ix, erts_aint32_t flgs)
     if (ix == 0 && erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].extra)
 	return erts_alcu_fix_alloc_shrink(
 	    erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].extra, flgs);
-#else
-    if (ix == 1 && erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].extra)
-	return erts_alcu_fix_alloc_shrink(
-	    erts_allctrs_info[ERTS_ALC_A_FIXED_SIZE].extra, flgs);
-#endif
     return 0;
 }
 
@@ -2172,7 +2079,7 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
  * NOTE! When updating this function, make sure to also update
  *       erlang:memory/[0,1] in $ERL_TOP/erts/preloaded/src/erlang.erl
  */
-#define ERTS_MEM_NEED_ALL_ALCU (!erts_instr_stat && want_tot_or_sys)
+#define ERTS_MEM_NEED_ALL_ALCU (want_tot_or_sys)
     struct {
 	int total;
 	int processes;
@@ -2183,7 +2090,6 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	int binary;
 	int code;
 	int ets;
-	int maximum;
     } want = {0};
     struct {
 	UWord total;
@@ -2195,7 +2101,6 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	UWord binary;
 	UWord code;
 	UWord ets;
-	UWord maximum;
     } size = {0};
     Eterm atoms[sizeof(size)/sizeof(UWord)];
     UWord *uintps[sizeof(size)/sizeof(UWord)];
@@ -2207,7 +2112,7 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
     int only_one_value = 0;
     ErtsAlcUFixInfo_t fi[ERTS_ALC_NO_FIXED_SIZES] = {{0,0}};
 
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 
     /* Figure out whats wanted... */
 
@@ -2248,12 +2153,6 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	want.ets = 1;
 	atoms[length] = am_ets;
 	uintps[length++] = &size.ets;
-
-	want.maximum = erts_instr_stat;
-	if (want.maximum) {
-	    atoms[length] = am_maximum;
-	    uintps[length++] = &size.maximum;
-	}
     }
     else {
 	DeclareTmpHeapNoproc(tmp_heap,2);
@@ -2335,18 +2234,6 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 		    uintps[length++] = &size.ets;
 		}
 		break;
-	    case am_maximum:
-		if (erts_instr_stat) {
-		    if (!want.maximum) {
-			want.maximum = 1;
-			atoms[length] = am_maximum;
-			uintps[length++] = &size.maximum;
-		    }
-		} else {
-		    UnUseTmpHeapNoproc(2);
-		    return am_badarg;
-		}
-		break;
 	    default:
 		UnUseTmpHeapNoproc(2);
 		return am_badarg;
@@ -2380,10 +2267,10 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 
 
     if (proc) {
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN
+	ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN
 			   == erts_proc_lc_my_proc_locks(proc));
 	/* We'll need locks early in the lock order */
-	erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
     }
 
     /* Calculate values needed... */
@@ -2445,7 +2332,6 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	}
 	tmp += erts_ptab_mem_size(&erts_proc);
 	tmp += erts_bif_timer_memory_size();
-	tmp += erts_tot_link_lh_size();
 
 	size.processes = size.processes_used = tmp;
 
@@ -2456,12 +2342,11 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	add_fix_values(&size.processes,
 		       &size.processes_used,
 		       fi,
-		       ERTS_ALC_T_MONITOR_SH);
-
+		       ERTS_ALC_T_MONITOR);
 	add_fix_values(&size.processes,
 		       &size.processes_used,
 		       fi,
-		       ERTS_ALC_T_NLINK_SH);
+		       ERTS_ALC_T_LINK);
 	add_fix_values(&size.processes,
 		       &size.processes_used,
 		       fi,
@@ -2514,14 +2399,7 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	size.ets += erts_get_ets_misc_mem_size();
     }
 
-    if (erts_instr_stat && (want_tot_or_sys || want.maximum)) {
-	if (want_tot_or_sys) {
-	    size.total = erts_instr_get_total();
-	    size.system = size.total - size.processes;
-	}
-	size.maximum = erts_instr_get_max_total();
-    }
-    else if (want_tot_or_sys) {
+    if (want_tot_or_sys) {
 	size.system = size.total - size.processes;
     }
 
@@ -2541,7 +2419,7 @@ erts_memory(fmtfn_t *print_to_p, void *print_to_arg, void *proc, Eterm earg)
 	Uint *hp;
 	Uint hsz;
 
-	erts_smp_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
 
 	if (only_one_value) {
 	    ASSERT(length == 1);
@@ -2590,27 +2468,15 @@ erts_allocated_areas(fmtfn_t *print_to_p, void *print_to_arg, void *proc)
     Uint reserved_atom_space, atom_space;
 
     if (proc) {
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN
+	ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN
 			   == erts_proc_lc_my_proc_locks(proc));
 
 	/* We'll need locks early in the lock order */
-	erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
     }
 
     i = 0;
 
-    if (erts_instr_stat) {
-	values[i].arity = 2;
-	values[i].name = "total";
-	values[i].ui[0] = erts_instr_get_total();
-	i++;
-
-	values[i].arity = 2;
-	values[i].name = "maximum";
-	values[i].ui[0] = erts_instr_get_max_total();
-	i++;
-    }
-
     values[i].arity = 2;
     values[i].name = "sys_misc";
     values[i].ui[0] = erts_sys_misc_mem_sz();
@@ -2692,11 +2558,6 @@ erts_allocated_areas(fmtfn_t *print_to_p, void *print_to_arg, void *proc)
     i++;
 
     values[i].arity = 2;
-    values[i].name = "link_lh";
-    values[i].ui[0] = erts_tot_link_lh_size();
-    i++;
-
-    values[i].arity = 2;
     values[i].name = "process_table";
     values[i].ui[0] = erts_ptab_mem_size(&erts_proc);
     i++;
@@ -2746,7 +2607,7 @@ erts_allocated_areas(fmtfn_t *print_to_p, void *print_to_arg, void *proc)
 	Uint hsz;
 	Uint *hszp;
 
-	erts_smp_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
 
 	hpp = NULL;
 	hsz = 0;
@@ -2758,7 +2619,7 @@ erts_allocated_areas(fmtfn_t *print_to_p, void *print_to_arg, void *proc)
 		Eterm atom;
 		if (hpp)
 		    atom = am_atom_put(values[i].name,
-				       (int) strlen(values[i].name));
+				       (int) sys_strlen(values[i].name));
 		else
 		    atom = am_true;
 
@@ -2834,7 +2695,7 @@ erts_allocator_info(fmtfn_t to, void *arg)
 {
     ErtsAlcType_t a;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 
     for (a = ERTS_ALC_A_MIN; a <= ERTS_ALC_A_MAX; a++) {
 	int ai;
@@ -2887,11 +2748,7 @@ erts_allocator_info(fmtfn_t to, void *arg)
 #if HAVE_ERTS_MSEG
     {
 	struct erts_mmap_info_struct emis;
-#ifdef ERTS_SMP
 	int max = (int) erts_no_schedulers;
-#else
-	int max = 0;
-#endif
 	int i;
 	for (i = 0; i <= max; i++) {
 	    erts_print(to, arg, "=allocator:mseg_alloc[%d]\n", i);
@@ -2903,10 +2760,6 @@ erts_allocator_info(fmtfn_t to, void *arg)
         erts_print(to, arg, "=allocator:erts_mmap.literal_mmap\n");
         erts_mmap_info(&erts_literal_mmapper, &to, arg, NULL, NULL, &emis);
 #endif
-#ifdef ERTS_HAVE_EXEC_MMAPPER
-        erts_print(to, arg, "=allocator:erts_mmap.exec_mmap\n");
-        erts_mmap_info(&erts_exec_mmapper, &to, arg, NULL, NULL, &emis);
-#endif
     }
 #endif
 
@@ -2914,10 +2767,7 @@ erts_allocator_info(fmtfn_t to, void *arg)
     erts_alcu_au_info_options(&to, arg, NULL, NULL);
 
     erts_print(to, arg, "=allocator:instr\n");
-    erts_print(to, arg, "option m: %s\n",
-	       erts_instr_memory_map ? "true" : "false");
-    erts_print(to, arg, "option s: %s\n",
-	       erts_instr_stat ? "true" : "false");
+
     erts_print(to, arg, "option t: %s\n",
 	       erts_mtrace_enabled ? "true" : "false");
 
@@ -2953,7 +2803,7 @@ erts_allocator_options(void *proc)
     for (a = ERTS_ALC_A_MIN; a <= ERTS_ALC_A_MAX; a++) {
 	Eterm tmp = NIL;
 	atoms[length] = am_atom_put((char *) ERTS_ALC_A2AD(a),
-				    strlen(ERTS_ALC_A2AD(a)));
+				    sys_strlen(ERTS_ALC_A2AD(a)));
 	if (erts_allctrs_info[a].enabled) {
 	    if (erts_allctrs_info[a].alloc_util) {
 		Allctr_t *allctr;
@@ -2971,20 +2821,20 @@ erts_allocator_options(void *proc)
 		Eterm as[4];
 		Eterm ts[4];
 
-		as[l] = am_atom_put("e", 1);
+		as[l] = ERTS_MAKE_AM("e");
 		ts[l++] = am_true;
 
 		switch (a) {
 		case ERTS_ALC_A_SYSTEM:
-		    as[l] = am_atom_put("m", 1);
-		    ts[l++] = am_atom_put("libc", 4);
+		    as[l] = ERTS_MAKE_AM("m");
+		    ts[l++] = ERTS_MAKE_AM("libc");
 		    if(sas.trim_threshold >= 0) {
-			as[l] = am_atom_put("tt", 2);
+			as[l] = ERTS_MAKE_AM("tt");
 			ts[l++] = erts_bld_uint(hpp, szp,
 						(Uint) sas.trim_threshold);
 		    }
 		    if(sas.top_pad >= 0) {
-			as[l] = am_atom_put("tp", 2);
+			as[l] = ERTS_MAKE_AM("tp");
 			ts[l++] = erts_bld_uint(hpp, szp, (Uint) sas.top_pad);
 		    }
 		    break;
@@ -2998,7 +2848,7 @@ erts_allocator_options(void *proc)
 
 	}
 	else {
-	    Eterm atom = am_atom_put("e", 1);
+	    Eterm atom = ERTS_MAKE_AM("e");
 	    Eterm term = am_false;
 	    tmp = erts_bld_2tup_list(hpp, szp, 1, &atom, &term);
 	}
@@ -3009,12 +2859,12 @@ erts_allocator_options(void *proc)
 
 #if HAVE_ERTS_MSEG
     if (use_mseg) {
-	atoms[length] = am_atom_put("mseg_alloc", 10);
+	atoms[length] = ERTS_MAKE_AM("mseg_alloc");
 	terms[length++] = erts_mseg_info_options(0, NULL, NULL, hpp, szp);
     }
 #endif
 
-    atoms[length] = am_atom_put("alloc_util", 10); 
+    atoms[length] = ERTS_MAKE_AM("alloc_util");
     terms[length++] = erts_alcu_au_info_options(NULL, NULL, hpp, szp);
 
 #if HAVE_ERTS_MMAP
@@ -3023,22 +2873,16 @@ erts_allocator_options(void *proc)
                                              NULL, hpp, szp);
 #endif
     {
-	Eterm o[3], v[3];
-	o[0] = am_atom_put("m", 1);
-	v[0] = erts_instr_memory_map ? am_true : am_false;
-	o[1] = am_atom_put("s", 1);
-	v[1] = erts_instr_stat ? am_true : am_false;
-	o[2] = am_atom_put("t", 1);
-	v[2] = erts_mtrace_enabled ? am_true : am_false;
-
-	atoms[length] = am_atom_put("instr", 5); 
-	terms[length++] = erts_bld_2tup_list(hpp, szp, 3, o, v);
+	Eterm o[1], v[1];
+	o[0] = ERTS_MAKE_AM("t");
+	v[0] = erts_mtrace_enabled ? am_true : am_false;
+
+	atoms[length] = ERTS_MAKE_AM("instr");
+	terms[length++] = erts_bld_2tup_list(hpp, szp, 1, o, v);
     }
 
-    atoms[length] = am_atom_put("lock_physical_memory", 20);
-    terms[length++] = (lock_all_physical_memory
-		       ? am_atom_put("all", 3)
-		       : am_atom_put("no", 2));
+    atoms[length] = ERTS_MAKE_AM("lock_physical_memory");
+    terms[length++] = (lock_all_physical_memory ? am_all : am_no);
 
     settings = erts_bld_2tup_list(hpp, szp, length, atoms, terms);
 
@@ -3047,28 +2891,25 @@ erts_allocator_options(void *proc)
     for (a = ERTS_ALC_A_MIN; a <= ERTS_ALC_A_MAX; a++) {
 	if (erts_allctrs_info[a].enabled) {
 	    terms[length++] = am_atom_put((char *) ERTS_ALC_A2AD(a),
-					  strlen(ERTS_ALC_A2AD(a)));
+					  sys_strlen(ERTS_ALC_A2AD(a)));
 	}
     }
 
 #if HAVE_ERTS_MSEG
     if (use_mseg)
-	terms[length++] = am_atom_put("mseg_alloc", 10);
+	terms[length++] = ERTS_MAKE_AM("mseg_alloc");
 #endif
 #if ERTS_HAVE_ERTS_SYS_ALIGNED_ALLOC
-    terms[length++] = am_atom_put("sys_aligned_alloc", 17);
+    terms[length++] = ERTS_MAKE_AM("sys_aligned_alloc");
 #endif
 #if defined(ARCH_64) && defined(ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION)
     terms[length++] = ERTS_MAKE_AM("literal_mmap");
 #endif
-#ifdef ERTS_HAVE_EXEC_MMAPPER
-    terms[length++] = ERTS_MAKE_AM("exec_mmap");
-#endif
     features = length ? erts_bld_list(hpp, szp, length, terms) : NIL;
 
 #if defined(__GLIBC__)
     {
-	Eterm AM_glibc = am_atom_put("glibc", 5);
+	Eterm AM_glibc = ERTS_MAKE_AM("glibc");
 	Eterm version;
 
 	version = erts_bld_cons(hpp,
@@ -3153,9 +2994,6 @@ reply_alloc_info(void *vair)
 # if defined(ARCH_64) && defined(ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION)
     struct erts_mmap_info_struct mmap_info_literal;
 # endif
-# ifdef ERTS_HAVE_EXEC_MMAPPER
-    struct erts_mmap_info_struct mmap_info_exec;
-# endif
 #endif
     int i;
     Eterm (*info_func)(Allctr_t *,
@@ -3284,17 +3122,6 @@ reply_alloc_info(void *vair)
                                             erts_bld_atom(hpp,szp,"literal_mmap"),
                                             ainfo);
 #  endif
-#  ifdef ERTS_HAVE_EXEC_MMAPPER
-                    ai_list = erts_bld_cons(hpp, szp,
-                                            ainfo, ai_list);
-                    ainfo = (air->only_sz ? NIL :
-                             erts_mmap_info(&erts_exec_mmapper, NULL, NULL,
-                                            hpp, szp, &mmap_info_exec));
-                    ainfo = erts_bld_tuple3(hpp, szp,
-                                            alloc_atom,
-                                            erts_bld_atom(hpp,szp,"exec_mmap"),
-                                            ainfo);
-#  endif
 #else  /* !HAVE_ERTS_MMAP */
                     ainfo = erts_bld_tuple2(hpp, szp, alloc_atom,
                                             am_false);
@@ -3352,7 +3179,7 @@ reply_alloc_info(void *vair)
 	    case ERTS_ALC_INFO_A_DISABLED_EXEC:
 		break;
 	    case ERTS_ALC_INFO_A_MSEG_ALLOC:
-#if HAVE_ERTS_MSEG && defined(ERTS_SMP)
+#if HAVE_ERTS_MSEG
 		alloc_atom = erts_bld_atom(hpp, szp, "mseg_alloc");
 		ainfo = erts_mseg_info(sched_id, NULL, NULL,
                                        hpp != NULL, air->only_sz, hpp, szp);
@@ -3406,10 +3233,10 @@ reply_alloc_info(void *vair)
     if (air->req_sched == sched_id)
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
  
-    erts_smp_proc_unlock(rp, rp_locks);
+    erts_proc_unlock(rp, rp_locks);
     erts_proc_dec_refc(rp);
 
-    if (erts_smp_atomic32_dec_read_nob(&air->refc) == 0) {
+    if (erts_atomic32_dec_read_nob(&air->refc) == 0) {
         erts_iref_storage_clean(&air->iref);
 	aireq_free(air);
     }
@@ -3488,18 +3315,16 @@ erts_request_alloc_info(struct process *c_p,
 
     air->allocs[airix] = ERTS_ALC_A_INVALID;
 
-    erts_smp_atomic32_init_nob(&air->refc,
+    erts_atomic32_init_nob(&air->refc,
 			       (erts_aint32_t) erts_no_schedulers);
 
     erts_proc_add_refc(c_p, (Sint) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
 					  reply_alloc_info,
 					  (void *) air);
-#endif
 
     reply_alloc_info((void *) air);
 
@@ -3567,8 +3392,8 @@ badarg:
 
 /* 
  * The allocator wrapper prelocking stuff below is about the locking order.
- * It only affects wrappers (erl_mtrace.c and erl_instrument.c) that keep locks
- * during alloc/realloc/free.
+ * It only affects wrappers (erl_mtrace.c) that keep locks during
+ * alloc/realloc/free.
  *
  * Some query functions in erl_alloc_util.c lock the allocator mutex and then
  * use erts_printf that in turn may call the sys allocator through the wrappers.
@@ -3633,35 +3458,29 @@ UWord erts_alc_test(UWord op, UWord a1, UWord a2, UWord a3)
     case 0xf:
 	switch (op) {
 	case 0xf00:
-#ifdef USE_THREADS
 	    if (((Allctr_t *) a1)->thread_safe)
 		return (UWord) erts_alcu_alloc_ts(ERTS_ALC_T_UNDEF,
 							  (void *) a1,
 							  (Uint) a2);
 	    else
-#endif
 		return (UWord) erts_alcu_alloc(ERTS_ALC_T_UNDEF,
 						       (void *) a1,
 						       (Uint) a2);
 	case 0xf01:
-#ifdef USE_THREADS
 	    if (((Allctr_t *) a1)->thread_safe)
 		return (UWord) erts_alcu_realloc_ts(ERTS_ALC_T_UNDEF,
 							    (void *) a1,
 							    (void *) a2,
 							    (Uint) a3);
 	    else
-#endif
 		return (UWord) erts_alcu_realloc(ERTS_ALC_T_UNDEF,
 							 (void *) a1,
 							 (void *) a2,
 							 (Uint) a3);
 	case 0xf02:
-#ifdef USE_THREADS
 	    if (((Allctr_t *) a1)->thread_safe)
 		erts_alcu_free_ts(ERTS_ALC_T_UNDEF, (void *) a1, (void *) a2);
 	    else
-#endif
 		erts_alcu_free(ERTS_ALC_T_UNDEF, (void *) a1, (void *) a2);
 	    return 0;
 	case 0xf03: {
@@ -3672,11 +3491,7 @@ UWord erts_alc_test(UWord op, UWord a1, UWord a2, UWord a3)
 	    init.enable = 1;
 	    init.atype = GOODFIT;
 	    init.init.util.name_prefix = (char *) a1;
-#ifdef ERTS_SMP
 	    init.init.util.ts = 1;
-#else
-            init.init.util.ts = a2 ? 1 : 0;
-#endif
 	    if ((char **) a3) {
 		char **argv = (char **) a3;
 		int i = 0;
@@ -3731,7 +3546,6 @@ UWord erts_alc_test(UWord op, UWord a1, UWord a2, UWord a3)
 	    erts_alcu_stop((Allctr_t *) a1);
 	    erts_free(ERTS_ALC_T_UNDEF, (void *) a1);
 	    break;
-#ifdef USE_THREADS
 	case 0xf05: return (UWord) 1;
 	case 0xf06: return (UWord) ((Allctr_t *) a1)->thread_safe;
 #ifdef ETHR_NO_FORKSAFETY
@@ -3801,12 +3615,7 @@ UWord erts_alc_test(UWord op, UWord a1, UWord a2, UWord a3)
 	    ethr_thr_exit((void *) a1);
 	    ERTS_ALC_TEST_ABORT;
 	    break;
-#endif /* #ifdef USE_THREADS */
-#ifdef ERTS_SMP
 	case 0xf13: return (UWord) 1;
-#else
-	case 0xf13: return (UWord) 0;
-#endif
 	case 0xf14: return (UWord) erts_alloc(ERTS_ALC_T_TEST, (Uint)a1);
 
 	case 0xf15: erts_free(ERTS_ALC_T_TEST, (void*)a1); return 0;
@@ -4012,10 +3821,8 @@ void check_allocators(void)
 	    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) erts_allctrs[i].extra;
 	    Allctr_t *allctr = real_af->extra;
 	    Carrier_t *ct;
-#ifdef USE_THREADS
 	if (allctr->thread_safe)
 	    erts_mtx_lock(&allctr->mutex);
-#endif
 
 	    if (allctr->check_mbc) {
 		for (ct = allctr->mbc_list.first; ct; ct = ct->next) {
@@ -4023,10 +3830,8 @@ void check_allocators(void)
 		    allctr->check_mbc(allctr,ct);
 		}
 	    }
-#ifdef USE_THREADS
 	if (allctr->thread_safe)
 	    erts_mtx_unlock(&allctr->mutex);
-#endif
 	}
     }
 }
@@ -4053,7 +3858,7 @@ set_memory_fence(void *ptr, Uint sz, ErtsAlcType_t n)
 
     *(ui_ptr++) = sz;
     *(ui_ptr++) = pattern;
-    memcpy((void *) (((char *) ui_ptr)+sz), (void *) &pattern, sizeof(UWord));
+    sys_memcpy((void *) (((char *) ui_ptr)+sz), (void *) &pattern, sizeof(UWord));
 
 #ifdef HARD_DEBUG
     *mblkpp = hdbg_alloc((void *) ui_ptr, sz, n);
@@ -4093,7 +3898,7 @@ check_memory_fence(void *ptr, Uint *size, ErtsAlcType_t n, int func)
 		     (UWord) ptr);
     }
 
-    memcpy((void *) &post_pattern, (void *) (((char *)ptr)+sz), sizeof(UWord));
+    sys_memcpy((void *) &post_pattern, (void *) (((char *)ptr)+sz), sizeof(UWord));
 
     if (post_pattern != MK_PATTERN(n)
 	|| pre_pattern != post_pattern) {
diff --git a/erts/emulator/beam/erl_alloc.h b/erts/emulator/beam/erl_alloc.h
index b7eb977317..578a3717d9 100644
--- a/erts/emulator/beam/erl_alloc.h
+++ b/erts/emulator/beam/erl_alloc.h
@@ -27,9 +27,7 @@
 #include "erl_thr_progress.h"
 #undef ERL_THR_PROGRESS_TSD_TYPE_ONLY
 #include "erl_alloc_util.h"
-#ifdef USE_THREADS
 #include "erl_threads.h"
-#endif
 #include "erl_mmap.h"
 
 #ifdef DEBUG
@@ -128,8 +126,10 @@ typedef struct {
     void *extra;
 } ErtsAllocatorFunctions_t;
 
-extern ErtsAllocatorFunctions_t erts_allctrs[ERTS_ALC_A_MAX+1];
-extern ErtsAllocatorInfo_t erts_allctrs_info[ERTS_ALC_A_MAX+1];
+extern ErtsAllocatorFunctions_t
+    ERTS_WRITE_UNLIKELY(erts_allctrs[ERTS_ALC_A_MAX+1]);
+extern ErtsAllocatorInfo_t
+    ERTS_WRITE_UNLIKELY(erts_allctrs_info[ERTS_ALC_A_MAX+1]);
 
 typedef struct {
     int enabled;
@@ -146,7 +146,7 @@ typedef struct ErtsAllocatorWrapper_t_ {
     void (*unlock)(void);
     struct ErtsAllocatorWrapper_t_* next;
 }ErtsAllocatorWrapper_t;
-ErtsAllocatorWrapper_t *erts_allctr_wrappers;
+extern ErtsAllocatorWrapper_t *erts_allctr_wrappers;
 extern int erts_allctr_wrapper_prelocked;
 extern erts_tsd_key_t erts_allctr_prelock_tsd_key;
 void erts_allctr_wrapper_prelock_init(ErtsAllocatorWrapper_t* wrapper);
@@ -154,12 +154,10 @@ void erts_allctr_wrapper_pre_lock(void);
 void erts_allctr_wrapper_pre_unlock(void);
 
 void erts_alloc_register_scheduler(void *vesdp);
-#ifdef ERTS_SMP
 void erts_alloc_scheduler_handle_delayed_dealloc(void *vesdp,
 						 int *need_thr_progress,
 						 ErtsThrPrgrVal *thr_prgr_p,
 						 int *more_work);
-#endif
 erts_aint32_t erts_alloc_fix_alloc_shrink(int ix, erts_aint32_t flgs);
 
 __decl_noreturn void erts_alloc_enomem(ErtsAlcType_t,Uint)		
@@ -340,37 +338,10 @@ erts_alloc_get_verify_unused_temp_alloc(Allctr_t **allctr);
   (((((SZ) - 1) / ERTS_CACHE_LINE_SIZE) + 1) * ERTS_CACHE_LINE_SIZE)
 
 #define ERTS_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)			\
-ERTS_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT,				\
-		      (void) 0, (void) 0, (void) 0)
-
-#define ERTS_SMP_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)			\
-static erts_smp_spinlock_t NAME##_lck;					\
-ERTS_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT,				\
-		      erts_smp_spinlock_init(&NAME##_lck, #NAME "_alloc_lock", NIL, \
-		          ERTS_LOCK_FLAGS_CATEGORY_ALLOCATOR),\
-		      erts_smp_spin_lock(&NAME##_lck),			\
-		      erts_smp_spin_unlock(&NAME##_lck))
-
-#ifdef ERTS_SMP
+    ERTS_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT, (void) 0, (void) 0, (void) 0)
 
 #define ERTS_TS_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)			\
-ERTS_SMP_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)
-
-#else /* !ERTS_SMP */
-
-#define ERTS_TS_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)			\
-static erts_mtx_t NAME##_lck;						\
-ERTS_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT,				\
-		      erts_mtx_init(NAME##_lck, #NAME "_alloc_lock", NIL, \
-		          ERTS_LOCK_FLAGS_CATEGORY_ALLOCATOR),\
-		      erts_mtx_lock(&NAME##_lck),			\
-		      erts_mtx_unlock(&NAME##_lck))
-
-
-#endif
-
-#define ERTS_PALLOC_IMPL(NAME, TYPE, PASZ)				\
-ERTS_PRE_ALLOC_IMPL(NAME, TYPE, PASZ, (void) 0, (void) 0, (void) 0)
+ERTS_QUALLOC_IMPL(NAME, TYPE, PASZ, ALCT)
 
 #define ERTS_TS_PALLOC_IMPL(NAME, TYPE, PASZ)				\
 static erts_spinlock_t NAME##_lck;					\
@@ -380,17 +351,10 @@ ERTS_PRE_ALLOC_IMPL(NAME, TYPE, PASZ,					\
 		    erts_spin_lock(&NAME##_lck),			\
 		    erts_spin_unlock(&NAME##_lck))
 
-#ifdef ERTS_SMP
 
-#define ERTS_SMP_PALLOC_IMPL(NAME, TYPE, PASZ)				\
+#define ERTS_PALLOC_IMPL(NAME, TYPE, PASZ)				\
   ERTS_TS_PALLOC_IMPL(NAME, TYPE, PASZ)
 
-#else /* !ERTS_SMP */
-
-#define ERTS_SMP_PALLOC_IMPL(NAME, TYPE, PASZ)				\
-  ERTS_PALLOC_IMPL(NAME, TYPE, PASZ)
-
-#endif
 
 #define ERTS_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT, ILCK, LCK, ULCK)	\
 ERTS_PRE_ALLOC_IMPL(NAME##_pre, TYPE, PASZ, ILCK, LCK, ULCK)		\
@@ -414,21 +378,11 @@ NAME##_free(TYPE *p)							\
 	erts_free(ALCT, (void *) p);					\
 }
 
-#ifdef ERTS_SMP
 #define ERTS_SCHED_PREF_PALLOC_IMPL(NAME, TYPE, PASZ)			\
   ERTS_SCHED_PREF_PRE_ALLOC_IMPL(NAME, TYPE, PASZ)
-#else
-#define ERTS_SCHED_PREF_PALLOC_IMPL(NAME, TYPE, PASZ)			\
-  ERTS_PRE_ALLOC_IMPL(NAME, TYPE, PASZ, (void) 0, (void) 0, (void) 0)
-#endif
 
-#ifdef ERTS_SMP
 #define ERTS_SCHED_PREF_AUX(NAME, TYPE, PASZ)				\
 ERTS_SCHED_PREF_PRE_ALLOC_IMPL(NAME##_pre, TYPE, PASZ)
-#else
-#define ERTS_SCHED_PREF_AUX(NAME, TYPE, PASZ)				\
-ERTS_PRE_ALLOC_IMPL(NAME##_pre, TYPE, PASZ, (void) 0, (void) 0, (void) 0)
-#endif
 
 #define ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT)	\
 ERTS_SCHED_PREF_AUX(NAME, TYPE, PASZ)					\
@@ -452,9 +406,35 @@ NAME##_free(TYPE *p)							\
 	erts_free(ALCT, (void *) p);					\
 }
 
+#define ERTS_THR_PREF_AUX(NAME, TYPE, PASZ)				\
+ERTS_THR_PREF_PRE_ALLOC_IMPL(NAME##_pre, TYPE, PASZ)
+
+#define ERTS_THR_PREF_QUICK_ALLOC_IMPL(NAME, TYPE, PASZ, ALCT)	        \
+ERTS_THR_PREF_AUX(NAME, TYPE, PASZ)					\
+static void								\
+init_##NAME##_alloc(int nthreads)					\
+{									\
+    init_##NAME##_pre_alloc(nthreads);			                \
+}									\
+static ERTS_INLINE TYPE *						\
+NAME##_alloc(void)							\
+{									\
+    TYPE *res = NAME##_pre_alloc();					\
+    if (!res)								\
+	res = erts_alloc(ALCT, sizeof(TYPE));				\
+    return res;								\
+}									\
+static ERTS_INLINE void							\
+NAME##_free(TYPE *p)							\
+{									\
+    if (!NAME##_pre_free(p))						\
+	erts_free(ALCT, (void *) p);					\
+}
+
+
 #ifdef DEBUG
 #define ERTS_PRE_ALLOC_SIZE(SZ) ((SZ) < 1000 ? (SZ)/10 + 10 : 100)
-#define ERTS_PRE_ALLOC_CLOBBER(P, T) memset((void *) (P), 0xfd, sizeof(T))
+#define ERTS_PRE_ALLOC_CLOBBER(P, T) sys_memset((void *) (P), 0xfd, sizeof(T))
 #else
 #define ERTS_PRE_ALLOC_SIZE(SZ) ((SZ) > 1 ? (SZ) : 1)
 #define ERTS_PRE_ALLOC_CLOBBER(P, T)
@@ -532,7 +512,8 @@ init_##NAME##_alloc(void)						\
 {									\
     sspa_data_##NAME##__ =						\
 	erts_sspa_create(sizeof(union erts_sspa_##NAME##__),		\
-			 ERTS_PRE_ALLOC_SIZE((PASZ)));			\
+			 ERTS_PRE_ALLOC_SIZE((PASZ)), 			\
+                         0, NULL);                                      \
 }									\
 									\
 static TYPE *								\
@@ -554,6 +535,57 @@ NAME##_free(TYPE *p)							\
 			  (char *) p);					\
 }
 
+
+#define ERTS_THR_PREF_PRE_ALLOC_IMPL(NAME, TYPE, PASZ)		        \
+union erts_sspa_##NAME##__ {						\
+    erts_sspa_blk_t next;						\
+    TYPE type;								\
+};									\
+									\
+static erts_sspa_data_t *sspa_data_##NAME##__;				\
+									\
+static void								\
+init_##NAME##_alloc(int nthreads)					\
+{									\
+    sspa_data_##NAME##__ =						\
+	erts_sspa_create(sizeof(union erts_sspa_##NAME##__),		\
+			 ERTS_PRE_ALLOC_SIZE((PASZ)),			\
+                         nthreads,                                      \
+                         #NAME);                                        \
+}									\
+                                                                        \
+void								        \
+erts_##NAME##_alloc_init_thread(void)				        \
+{									\
+    int id = erts_atomic_inc_read_nob(&sspa_data_##NAME##__->id_generator);\
+    if (id > sspa_data_##NAME##__->nthreads) {                          \
+        erts_exit(ERTS_ABORT_EXIT,                                      \
+                  "%s:%d:%s(): Too many threads for '" #NAME "'\n",     \
+                  __FILE__, __LINE__, __func__);                        \
+    }                                                                   \
+    erts_tsd_set(sspa_data_##NAME##__->tsd_key, (void*)(SWord)id);      \
+}									\
+									\
+static TYPE *								\
+NAME##_alloc(void)							\
+{									\
+    int id = (int)(SWord)erts_tsd_get(sspa_data_##NAME##__->tsd_key);   \
+    if (id == 0)                                                        \
+        return NULL;                                                    \
+    return (TYPE *) erts_sspa_alloc(sspa_data_##NAME##__,		\
+                                    id-1);		                \
+}									\
+									\
+static int								\
+NAME##_free(TYPE *p)							\
+{									\
+    int id = (int)(SWord)erts_tsd_get(sspa_data_##NAME##__->tsd_key);   \
+    return erts_sspa_free(sspa_data_##NAME##__,				\
+			  id - 1,		                        \
+			  (char *) p);					\
+}
+
+
 #ifdef DEBUG
 #define ERTS_ALC_DBG_BLK_SZ(PTR) (*(((UWord *) (PTR)) - 2))
 #endif /* #ifdef DEBUG */
diff --git a/erts/emulator/beam/erl_alloc.types b/erts/emulator/beam/erl_alloc.types
index 252bf1cc7e..5409b89bab 100644
--- a/erts/emulator/beam/erl_alloc.types
+++ b/erts/emulator/beam/erl_alloc.types
@@ -1,7 +1,7 @@
 #
 # %CopyrightBegin%
 #
-# Copyright Ericsson AB 2003-2017. All Rights Reserved.
+# Copyright Ericsson AB 2003-2018. All Rights Reserved.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -53,15 +53,6 @@
 #
 # IMPORTANT! Only use 7-bit ascii text in this file!
 
-+if smp
-+disable threads_no_smp
-+else
-+if threads
-+enable threads_no_smp
-+else
-+disable threads_no_smp
-+endif
-+endif
 
 # --- Allocator declarations -------------------------------------------------
 #
@@ -77,8 +68,6 @@
 
 allocator	SYSTEM		true		sys_alloc
 
-+if smp
-
 allocator	TEMPORARY	true		temp_alloc
 allocator	SHORT_LIVED	true		sl_alloc
 allocator	STANDARD	true		std_alloc
@@ -91,22 +80,6 @@ allocator	LITERAL 	true		literal_alloc
 allocator	EXEC     	true	        exec_alloc
 +endif
 
-+else # Non smp build
-
-allocator	TEMPORARY	false		temp_alloc
-allocator	SHORT_LIVED	false		sl_alloc
-allocator	STANDARD	false		std_alloc
-allocator	LONG_LIVED	false		ll_alloc
-allocator	EHEAP		false		eheap_alloc
-allocator	ETS		false		ets_alloc
-allocator	FIXED_SIZE	false		fix_alloc
-allocator	LITERAL 	false		literal_alloc
-+if exec_alloc
-allocator	EXEC     	false	        exec_alloc
-+endif
-
-+endif
-
 allocator	BINARY		true		binary_alloc
 allocator	DRIVER		true		driver_alloc
 
@@ -145,9 +118,6 @@ type	PORT		DRIVER		SYSTEM		port
 type	ATOM		LONG_LIVED	ATOM		atom_entry
 type	MODULE		LONG_LIVED	CODE		module_entry
 type	REG_PROC	STANDARD	PROCESSES	reg_proc
-type	LINK_LH		STANDARD	PROCESSES	link_lh
-type	SUSPEND_MON	STANDARD	PROCESSES	suspend_monitor
-type	PEND_SUSPEND	SHORT_LIVED	PROCESSES	pending_suspend
 type	PROC_LIST	SHORT_LIVED	PROCESSES	proc_list
 type	SAVED_ESTACK	SHORT_LIVED	PROCESSES	saved_estack
 type	FUN_ENTRY	LONG_LIVED	CODE		fun_entry
@@ -160,7 +130,6 @@ type	TMP_HEAP	TEMPORARY	PROCESSES	tmp_heap
 type	MSG_REF		FIXED_SIZE	PROCESSES	msg_ref
 type	MSG		EHEAP		PROCESSES	message
 type	MSGQ_CHNG	SHORT_LIVED	PROCESSES	messages_queue_change
-type	MSG_ROOTS	TEMPORARY	PROCESSES	msg_roots
 type	ROOTSET		TEMPORARY	PROCESSES	root_set
 type	LOADER_TMP	TEMPORARY	CODE		loader_tmp
 type	PREPARED_CODE	SHORT_LIVED	CODE		prepared_code
@@ -205,7 +174,6 @@ type	BPD		STANDARD	SYSTEM		bpd
 type	LINEBUF		STANDARD	SYSTEM		line_buf
 type	IOQ		STANDARD	SYSTEM		io_queue
 type	BITS_BUF	STANDARD	SYSTEM		bits_buf
-type	TMP_DIST_BUF	TEMPORARY	SYSTEM		tmp_dist_buf
 type	ASYNC_DATA	LONG_LIVED	SYSTEM		internal_async_data
 type	ESTACK		TEMPORARY	SYSTEM		estack
 type	DB_TABLE	ETS		ETS		db_tab
@@ -218,7 +186,6 @@ type	DB_MC_STK	TEMPORARY	ETS		db_mc_stack
 type	DB_MS_RUN_HEAP	SHORT_LIVED	ETS		db_match_spec_run_heap
 type	DB_MS_CMPL_HEAP	TEMPORARY	ETS		db_match_spec_cmpl_heap
 type	DB_SEG		ETS		ETS		db_segment
-type	DB_SEG_TAB	ETS		ETS		db_segment_tab
 type	DB_STK		ETS		ETS		db_stack
 type	DB_TRANS_TAB	ETS		ETS		db_trans_tab
 type	DB_SEL_LIST	ETS		ETS		db_select_list
@@ -227,7 +194,6 @@ type	DB_DMC_ERR_INFO	ETS		ETS		db_dmc_error_info
 type	DB_TERM		ETS		ETS		db_term
 type	DB_PROC_CLEANUP SHORT_LIVED	ETS		db_proc_cleanup_state
 type	ETS_ALL_REQ	SHORT_LIVED	ETS		ets_all_request
-type	INSTR_INFO	LONG_LIVED	SYSTEM		instr_info
 type	LOGGER_DSBUF	TEMPORARY	SYSTEM		logger_dsbuf
 type	TMP_DSBUF	TEMPORARY	SYSTEM		tmp_dsbuf
 type	INFO_DSBUF	SYSTEM		SYSTEM		info_dsbuf
@@ -242,7 +208,6 @@ type	PT_HNDL_LIST	SHORT_LIVED	SYSTEM		port_task_handle_list
 type	MISC_OP_LIST	SHORT_LIVED	SYSTEM		misc_op_list
 type	PORT_NAMES	SHORT_LIVED	SYSTEM		port_names
 type	PORT_DATA_LOCK	DRIVER		SYSTEM		port_data_lock
-type	NODES_MON	STANDARD	PROCESSES	nodes_monitor
 type	PTAB_LIST_DEL	SHORT_LIVED	PROCESSES	ptab_list_deleted_el
 type	PTAB_LIST_CNKI	SHORT_LIVED	PROCESSES	ptab_list_chunk_info
 type	PTAB_LIST_PIDS	SHORT_LIVED	PROCESSES	ptab_list_pids
@@ -265,10 +230,8 @@ type	CPUDATA		LONG_LIVED	SYSTEM		cpu_data
 type	TMP_CPU_IDS	SHORT_LIVED	SYSTEM		tmp_cpu_ids
 type	EXT_TERM_DATA	SHORT_LIVED	PROCESSES	external_term_data
 type	CPU_GRPS_MAP	LONG_LIVED	SYSTEM		cpu_groups_map
-type	AUX_WORK_TMO	LONG_LIVED	SYSTEM		aux_work_timeouts
 type	MISC_AUX_WORK_Q	LONG_LIVED	SYSTEM		misc_aux_work_q
 type	CODE_IX_LOCK_Q  SHORT_LIVED	SYSTEM		code_ix_lock_q
-type	PROC_INTERVAL	LONG_LIVED	SYSTEM		process_interval
 type    BUSY_CALLER_TAB	SHORT_LIVED	SYSTEM		busy_caller_table
 type    BUSY_CALLER	SHORT_LIVED	SYSTEM		busy_caller
 type	PROC_SYS_TSK	SHORT_LIVED	PROCESSES	proc_sys_task
@@ -277,77 +240,55 @@ type    NEW_TIME_OFFSET	SHORT_LIVED	SYSTEM		new_time_offset
 type	IOB_REQ		SHORT_LIVED	SYSTEM		io_bytes_request
 type    TRACER_NIF      LONG_LIVED      SYSTEM          tracer_nif
 type    TRACE_MSG_QUEUE SHORT_LIVED     SYSTEM          trace_message_queue
-type    SCHED_ASYNC_JOB SHORT_LIVED     SYSTEM          async_calls
-type    DIRTY_START     STANDARD	PROCESSES       dirty_start
 type    DIRTY_SL	SHORT_LIVED	SYSTEM		dirty_short_lived
 type    MREF_NSCHED_ENT	FIXED_SIZE	SYSTEM		nsched_magic_ref_entry
 type    MREF_ENT	STANDARD	SYSTEM		magic_ref_entry
 type	MREF_TAB_BKTS	STANDARD	SYSTEM		magic_ref_table_buckets
 type	MREF_TAB	LONG_LIVED	SYSTEM		magic_ref_table
 type	MINDIRECTION	FIXED_SIZE	SYSTEM		magic_indirection
+type	BINARY_FIND	SHORT_LIVED	PROCESSES	binary_find
 type	CRASH_DUMP	STANDARD	SYSTEM		crash_dump
+type	DIST_TRANSCODE  SHORT_LIVED	SYSTEM		dist_transcode_context
 
-+if threads_no_smp
-# Need thread safe allocs, but std_alloc and fix_alloc are not;
-# use driver_alloc which is...
-type	THR_Q_EL	DRIVER 	   	SYSTEM		thr_q_element
-type	THR_Q_EL_SL	DRIVER		SYSTEM		sl_thr_q_element
-type	MISC_AUX_WORK	DRIVER		SYSTEM		misc_aux_work
-+else
 type	THR_Q_EL	STANDARD   	SYSTEM		thr_q_element
 type	THR_Q_EL_SL	FIXED_SIZE	SYSTEM		sl_thr_q_element
 type	MISC_AUX_WORK	SHORT_LIVED	SYSTEM		misc_aux_work
-+endif
 type	THR_Q		STANDARD	SYSTEM		thr_queue
 type	THR_Q_SL	SHORT_LIVED	SYSTEM		short_lived_thr_queue
 type	THR_Q_LL	LONG_LIVED	SYSTEM		long_lived_thr_queue
 
-+if smp
 type	ASYNC		SHORT_LIVED	SYSTEM		async
 type	ZLIB		STANDARD	SYSTEM		zlib
-+else
-# sl/std_alloc is not thread safe in non smp build; therefore, we use driver_alloc
-type	ZLIB		DRIVER   	SYSTEM		zlib
-type	ASYNC		DRIVER		SYSTEM		async
-+endif
 
-+if smp
-type	PORT_LOCK	STANDARD	SYSTEM		port_lock
 type	DRIVER_LOCK	STANDARD	SYSTEM		driver_lock
 type	XPORTS_LIST	SHORT_LIVED	SYSTEM		extra_port_list
-type	PROC_LCK_WTR	LONG_LIVED	SYSTEM		proc_lock_waiter
 type	RUNQ_BLNS	LONG_LIVED	SYSTEM		run_queue_balancing
 type	THR_PRGR_IDATA	LONG_LIVED	SYSTEM		thr_prgr_internal_data
 type	THR_PRGR_DATA	LONG_LIVED	SYSTEM		thr_prgr_data
 type	T_THR_PRGR_DATA	SHORT_LIVED	SYSTEM		temp_thr_prgr_data
 type	RELEASE_LAREA	SHORT_LIVED	SYSTEM		release_literal_area
-+endif
+type	SIG_DATA	SHORT_LIVED	PROCESSES	signal_data
+type	DIST_DEMONITOR	SHORT_LIVED	PROCESSES	dist_demonitor
+type	CML_CLEANUP	SHORT_LIVED	SYSTEM		connection_ml_cleanup
+type	ML_YIELD_STATE	SHORT_LIVED	SYSTEM		monitor_link_yield_state
+type	ML_DIST		STANDARD	SYSTEM		monitor_link_dist
+type	PF3_ARGS	SHORT_LIVED	PROCESSES	process_flag_3_arguments
+type	SETUP_CONN_ARG	SHORT_LIVED	PROCESSES	setup_connection_argument
+
+type	ENVIRONMENT	SYSTEM		SYSTEM		environment
 
 #
 # Types used for special emulators
 #
 
-+if threads
-
 type	ETHR_STD	STANDARD	SYSTEM		ethread_standard
 type	ETHR_SL		SHORT_LIVED	SYSTEM		ethread_short_lived
 type	ETHR_LL		LONG_LIVED	SYSTEM		ethread_long_lived
 
-+endif
-
-+if shared_heap
-
-type	STACK		STANDARD	PROCESSES	stack
-type	ACTIVE_PROCS	STANDARD	PROCESSES	active_procs
-
-+endif
-
-+if smp
 type	SYS_MSG_Q	SHORT_LIVED	PROCESSES	system_messages_queue
 type	FP_EXCEPTION	LONG_LIVED	SYSTEM		fp_exception
 type	LL_MPATHS	LONG_LIVED	SYSTEM		ll_migration_paths
 type	SL_MPATHS	SHORT_LIVED	SYSTEM		sl_migration_paths
-+endif
 
 +if hipe
 
@@ -361,14 +302,6 @@ type	HIPE_EXEC	EXEC            CODE		hipe_code
 
 +endif
 
-
-
-+if heap_frag_elim_test
-
-type	SSB		SHORT_LIVED	PROCESSES	ssb
-
-+endif
-
 +if lcnt
 
 type    LCNT_CARRIER   STANDARD    SYSTEM  lcnt_lock_info_carrier
@@ -379,8 +312,8 @@ type    LCNT_VECTOR SHORT_LIVED SYSTEM lcnt_sample_vector
 type	DEBUG	        SHORT_LIVED	SYSTEM          debugging
 
 type	DDLL_PROCESS	STANDARD	SYSTEM		ddll_processes
-type	MONITOR_LH	STANDARD	PROCESSES	monitor_lh
-type	NLINK_LH	STANDARD	PROCESSES	nlink_lh
+type	MONITOR_EXT	STANDARD	PROCESSES	monitor_extended
+type	LINK_EXT	STANDARD	PROCESSES	link_extended
 type	CODE		LONG_LIVED	CODE		code
 type	LITERAL		LITERAL 	CODE		literal
 type	LITERAL_REF	SHORT_LIVED	CODE		literal_area_ref
@@ -388,13 +321,13 @@ type	PURGE_DATA	SHORT_LIVED	CODE		purge_data
 type	DB_HEIR_DATA	STANDARD	ETS		db_heir_data
 type	DB_MS_PSDO_PROC	LONG_LIVED	ETS		db_match_pseudo_proc
 type	SCHDLR_DATA	LONG_LIVED	SYSTEM		scheduler_data
-type  	LL_TEMP_TERM 	LONG_LIVED	SYSTEM		ll_temp_term
 
 type	NIF_TRAP_EXPORT	STANDARD	PROCESSES	nif_trap_export_entry
 type	NIF_EXP_TRACE	FIXED_SIZE	PROCESSES	nif_export_trace
 type	EXPORT		LONG_LIVED	CODE		export_entry
-type	MONITOR_SH	FIXED_SIZE	PROCESSES	monitor_sh
-type	NLINK_SH	FIXED_SIZE	PROCESSES	nlink_sh
+type	MONITOR		FIXED_SIZE	PROCESSES	monitor
+type	MONITOR_SUSPEND	STANDARD	PROCESSES	monitor_suspend
+type	LINK		FIXED_SIZE	PROCESSES	link
 type	AINFO_REQ	SHORT_LIVED	SYSTEM		alloc_info_request
 type	SCHED_WTIME_REQ	SHORT_LIVED	SYSTEM		sched_wall_time_request
 type	GC_INFO_REQ	SHORT_LIVED	SYSTEM		gc_info_request
@@ -409,34 +342,22 @@ type	SYS_CHECK_REQ	SHORT_LIVED	SYSTEM		system_check_request
 type  	TEMP_TERM 	TEMPORARY	SYSTEM		temp_term
 type	DRV_TAB		LONG_LIVED	SYSTEM		drv_tab
 type	DRV_EV_STATE	LONG_LIVED	SYSTEM		driver_event_state
-type	DRV_EV_D_STATE	FIXED_SIZE	SYSTEM		driver_event_data_state
 type	DRV_SEL_D_STATE	FIXED_SIZE	SYSTEM		driver_select_data_state
 type	NIF_SEL_D_STATE	FIXED_SIZE	SYSTEM		enif_select_data_state
-type	FD_LIST		SHORT_LIVED	SYSTEM		fd_list
-type	ACTIVE_FD_ARR	SHORT_LIVED	SYSTEM		active_fd_array
 type	POLLSET		LONG_LIVED	SYSTEM		pollset
 type	POLLSET_UPDREQ	SHORT_LIVED	SYSTEM		pollset_update_req
 type	POLL_FDS	LONG_LIVED	SYSTEM		poll_fds
-type	POLL_RES_EVS	LONG_LIVED	SYSTEM		poll_result_events
 type	FD_STATUS	LONG_LIVED	SYSTEM		fd_status
 type	SELECT_FDS	LONG_LIVED	SYSTEM		select_fds
 
 +if unix
 
 type	SYS_READ_BUF	TEMPORARY	SYSTEM		sys_read_buf
-type	FD_TAB		LONG_LIVED	SYSTEM		fd_tab
 type	FD_ENTRY_BUF	STANDARD	SYSTEM		fd_entry_buf
 type	CS_PROG_PATH	LONG_LIVED	SYSTEM		cs_prog_path
-type	ENVIRONMENT	TEMPORARY	SYSTEM		environment
-type	PUTENV_STR	SYSTEM		SYSTEM		putenv_string
-type	PRT_REP_EXIT	STANDARD	SYSTEM		port_report_exit
 type	SYS_BLOCKING	STANDARD	SYSTEM		sys_blocking
 
-+if smp
 type	SYS_WRITE_BUF	TEMPORARY	SYSTEM		sys_write_buf
-+else
-type	SYS_WRITE_BUF	BINARY		SYSTEM		sys_write_buf
-+endif
 
 +endif
 
@@ -444,10 +365,7 @@ type	SYS_WRITE_BUF	BINARY		SYSTEM		sys_write_buf
 
 type	DRV_DATA_BUF	SYSTEM		SYSTEM		drv_data_buf
 type	PRELOADED	LONG_LIVED	SYSTEM		preloaded
-type	PUTENV_STR	SYSTEM		SYSTEM		putenv_string
 type	WAITER_OBJ	LONG_LIVED	SYSTEM		waiter_object
-type	ENVIRONMENT	SYSTEM		SYSTEM		environment
-type	CON_VPRINTF_BUF	TEMPORARY	SYSTEM		con_vprintf_buf
 
 +endif
 
diff --git a/erts/emulator/beam/erl_alloc_util.c b/erts/emulator/beam/erl_alloc_util.c
index ae4ff4f453..fdf355d503 100644
--- a/erts/emulator/beam/erl_alloc_util.c
+++ b/erts/emulator/beam/erl_alloc_util.c
@@ -48,6 +48,8 @@
 #include "erl_mseg.h"
 #include "erl_threads.h"
 #include "erl_thr_progress.h"
+#include "erl_bif_unique.h"
+#include "erl_nif.h"
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
 #include "erl_lock_count.h"
@@ -139,6 +141,33 @@ MBC after deallocating first block:
 [Carrier_t|pad|Block_t 111| udata... ]
 */
 
+/* Allocation tags ...
+ *
+ * These are added to the footer of every block when enabled. Currently they
+ * consist of the allocation type and an atom identifying the allocating
+ * driver/nif (or 'system' if that can't be determined), but the format is not
+ * supposed to be set in stone.
+ *
+ * The packing scheme requires that the atom values are small enough to fit
+ * into a word with ERTS_ALC_N_BITS to spare. Users must check for overflow
+ * before MAKE_ATAG(). */
+
+typedef UWord alcu_atag_t;
+
+#define MAKE_ATAG(IdAtom, Type) \
+    (ASSERT((Type) >= ERTS_ALC_N_MIN && (Type) <= ERTS_ALC_N_MAX), \
+     ASSERT(atom_val(IdAtom) <= MAX_ATAG_ATOM_ID), \
+     (atom_val(IdAtom) << ERTS_ALC_N_BITS) | (Type))
+
+#define ATAG_ID(AT) (make_atom((AT) >> ERTS_ALC_N_BITS))
+#define ATAG_TYPE(AT) ((AT) & ERTS_ALC_N_MASK)
+
+#define MAX_ATAG_ATOM_ID (ERTS_UWORD_MAX >> ERTS_ALC_N_BITS)
+
+#define DBG_IS_VALID_ATAG(Allocator, AT) \
+    (ATAG_TYPE(AT) >= ERTS_ALC_N_MIN && \
+     ATAG_TYPE(AT) <= ERTS_ALC_N_MAX && \
+     (Allocator)->alloc_no == ERTS_ALC_T2A(ERTS_ALC_N2T(ATAG_TYPE(AT))))
 
 /* Blocks ... */
 
@@ -153,10 +182,17 @@ MBC after deallocating first block:
 #endif
 #define FBLK_FTR_SZ (sizeof(FreeBlkFtr_t))
 
+#define GET_BLK_ATAG(B) \
+    (((alcu_atag_t *) (((char *) (B)) + (BLK_SZ(B))))[-1])
+#define SET_BLK_ATAG(B, T) \
+    (((alcu_atag_t *) (((char *) (B)) + (BLK_SZ(B))))[-1] = (T))
+
+#define BLK_ATAG_SZ(AP) ((AP)->atags ? sizeof(alcu_atag_t) : 0)
+
 #define UMEMSZ2BLKSZ(AP, SZ)						\
-  (ABLK_HDR_SZ + (SZ) <= (AP)->min_block_size				\
+  (ABLK_HDR_SZ + BLK_ATAG_SZ(AP) + (SZ) <= (AP)->min_block_size		\
    ? (AP)->min_block_size						\
-   : UNIT_CEILING(ABLK_HDR_SZ + (SZ)))
+   : UNIT_CEILING(ABLK_HDR_SZ + BLK_ATAG_SZ(AP) + (SZ)))
 
 #define UMEM2BLK(P) ((Block_t *) (((char *) (P)) - ABLK_HDR_SZ))
 #define BLK2UMEM(P) ((void *)    (((char *) (P)) + ABLK_HDR_SZ))
@@ -305,9 +341,6 @@ MBC after deallocating first block:
 #  define ERTS_ALC_CPOOL_DEBUG
 #endif
 
-#ifndef ERTS_SMP
-#  undef ERTS_ALC_CPOOL_DEBUG
-#endif
 
 #ifdef ERTS_ALC_CPOOL_DEBUG
 #  define ERTS_ALC_CPOOL_ASSERT(A)				\
@@ -322,13 +355,8 @@ MBC after deallocating first block:
 #  define ERTS_ALC_CPOOL_ASSERT(A) ((void) 1)
 #endif
 
-#ifdef ERTS_SMP
 #define ERTS_ALC_IS_CPOOL_ENABLED(A)	((A)->cpool.util_limit)
-#else
-#define ERTS_ALC_IS_CPOOL_ENABLED(A)	(0)
-#endif
 
-#ifdef ERTS_SMP
 
 #define ERTS_ALC_CPOOL_MAX_DISABLE_ABANDON		1000
 #define ERTS_ALC_CPOOL_ALLOC_OP_INC			8
@@ -367,11 +395,6 @@ do {										\
     }										\
 } while (0)
 
-#else
-#define ERTS_ALC_CPOOL_ALLOC_OP(A)
-#define ERTS_ALC_CPOOL_REALLOC_OP(A)
-#define ERTS_ALC_CPOOL_FREE_OP(A)
-#endif
 
 #define ERTS_CRR_ALCTR_FLG_IN_POOL	(((erts_aint_t) 1) << 0)
 #define ERTS_CRR_ALCTR_FLG_BUSY		(((erts_aint_t) 1) << 1)
@@ -380,17 +403,10 @@ do {										\
                                  ERTS_CRR_ALCTR_FLG_BUSY |    \
                                  ERTS_CRR_ALCTR_FLG_HOMECOMING)
 
-#ifdef ERTS_SMP
 #define SBC_HEADER_SIZE	   						\
     (UNIT_CEILING(offsetof(Carrier_t, cpool)                            \
 	          + ABLK_HDR_SZ)	                                \
      - ABLK_HDR_SZ)
-#else
-#define SBC_HEADER_SIZE	   						\
-    (UNIT_CEILING(sizeof(Carrier_t)					\
-		  + ABLK_HDR_SZ)					\
-     - ABLK_HDR_SZ)
-#endif
 #define MBC_HEADER_SIZE(AP) ((AP)->mbc_header_size)
 
 
@@ -404,7 +420,7 @@ do {										\
 
 #define SET_CARRIER_HDR(C, Sz, F, AP) \
   (ASSERT(((Sz) & FLG_MASK) == 0), (C)->chdr = ((Sz) | (F)), \
-   erts_smp_atomic_init_nob(&(C)->allctr, (erts_aint_t) (AP)))
+   erts_atomic_init_nob(&(C)->allctr, (erts_aint_t) (AP)))
 
 #define BLK_TO_SBC(B) \
   ((Carrier_t *) (((char *) (B)) - SBC_HEADER_SIZE))
@@ -600,15 +616,11 @@ do {									\
     (AP)->mbcs.blocks.curr.size -= (CRR)->cpool.blocks_size;		\
 } while (0)
 
-#ifdef ERTS_SMP
 #define STAT_MBC_BLK_ALLOC_CRR(CRR, BSZ)				\
 do {									\
     (CRR)->cpool.blocks++;						\
     (CRR)->cpool.blocks_size += (BSZ);					\
 } while (0)
-#else
-#define STAT_MBC_BLK_ALLOC_CRR(CRR, BSZ) ((void) (CRR)) /* Get rid of warning */
-#endif
 
 #define STAT_MBC_BLK_ALLOC(AP, CRR, BSZ, FLGS)	       			\
 do {									\
@@ -628,7 +640,6 @@ stat_cpool_mbc_blk_free(Allctr_t *allctr,
 			Carrier_t **busy_pcrr_pp,
 			UWord blksz)
 {
-#ifdef ERTS_SMP
 
     ERTS_ALC_CPOOL_ASSERT(crr->cpool.blocks > 0);
     crr->cpool.blocks--;
@@ -653,9 +664,6 @@ stat_cpool_mbc_blk_free(Allctr_t *allctr,
 #endif
 
     return 1;
-#else
-    return 0;
-#endif
 }
 
 #define STAT_MBC_BLK_FREE(AP, CRR, BPCRRPP, BSZ, FLGS)			\
@@ -691,12 +699,7 @@ do {									\
 #endif
 
 #ifdef DEBUG
-#ifdef USE_THREADS
-# ifdef ERTS_SMP
 #  define IS_ACTUALLY_BLOCKING (erts_thr_progress_is_blocking())
-# else
-#  define IS_ACTUALLY_BLOCKING 0
-# endif
 #define ERTS_ALCU_DBG_CHK_THR_ACCESS(A)					\
 do {									\
     if (!(A)->thread_safe && !IS_ACTUALLY_BLOCKING) {                   \
@@ -705,7 +708,7 @@ do {									\
 	    (A)->debug.saved_tid = 1;					\
 	}								\
 	else {								\
-	    ERTS_SMP_LC_ASSERT(						\
+	    ERTS_LC_ASSERT(						\
 		ethr_equal_tids((A)->debug.tid, erts_thr_self()));	\
 	}								\
     }									\
@@ -713,9 +716,6 @@ do {									\
 #else
 #define ERTS_ALCU_DBG_CHK_THR_ACCESS(A)
 #endif
-#else
-#define ERTS_ALCU_DBG_CHK_THR_ACCESS(A)
-#endif
 
 static void make_name_atoms(Allctr_t *allctr);
 
@@ -724,6 +724,62 @@ static void destroy_carrier(Allctr_t *, Block_t *, Carrier_t **);
 static void mbc_free(Allctr_t *allctr, void *p, Carrier_t **busy_pcrr_pp);
 static void dealloc_block(Allctr_t *, void *, ErtsAlcFixList_t *, int);
 
+static alcu_atag_t determine_alloc_tag(Allctr_t *allocator, ErtsAlcType_t type)
+{
+    ErtsSchedulerData *esdp;
+    Eterm id;
+
+    ERTS_CT_ASSERT(_unchecked_atom_val(am_system) <= MAX_ATAG_ATOM_ID);
+    ASSERT(allocator->atags);
+
+    esdp = erts_get_scheduler_data();
+    id = am_system;
+
+    if (esdp) {
+        if (esdp->current_nif) {
+            Module *mod = erts_nif_get_module((esdp->current_nif)->mod_nif);
+
+            /* Mod can be NULL if a resource destructor allocates memory after
+             * the module has been unloaded. */
+            if (mod) {
+                id = make_atom(mod->module);
+            }
+        } else if (esdp->current_port) {
+            Port *p = esdp->current_port;
+            id = (p->drv_ptr)->name_atom;
+        }
+
+        /* We fall back to 'system' if we can't pack the driver/NIF name into
+         * the tag. This may be a bit misleading but we've made no promises
+         * that the information is complete.
+         *
+         * This can only happen on 32-bit emulators when a new driver/NIF has
+         * been loaded *after* 16 million atoms have been used, and supporting
+         * that fringe case is not worth an extra word. 64-bit emulators are
+         * unaffected since the atom cache limits atom indexes to 32 bits. */
+        if(MAX_ATOM_TABLE_SIZE > MAX_ATAG_ATOM_ID) {
+            if (atom_val(id) > MAX_ATAG_ATOM_ID) {
+                id = am_system;
+            }
+        }
+    }
+
+    return MAKE_ATAG(id, type);
+}
+
+static void set_alloc_tag(Allctr_t *allocator, void *p, alcu_atag_t tag)
+{
+    Block_t *block;
+
+    ASSERT(DBG_IS_VALID_ATAG(allocator, tag));
+    ASSERT(allocator->atags && p);
+    (void)allocator;
+
+    block = UMEM2BLK(p);
+
+    SET_BLK_ATAG(block, tag);
+}
+
 /* internal data... */
 
 #if 0
@@ -864,7 +920,7 @@ erts_alcu_literal_32_mseg_alloc(Allctr_t *allctr, Uint *size_p, Uint flags)
     Uint sz = ERTS_SUPERALIGNED_CEILING(*size_p);
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     res = erts_alcu_mseg_alloc(allctr, &sz, flags);
     if (res) {
@@ -882,7 +938,7 @@ erts_alcu_literal_32_mseg_realloc(Allctr_t *allctr, void *seg,
     Uint new_sz = ERTS_SUPERALIGNED_CEILING(*new_size_p);
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     if (seg && old_size)
         clear_literal_range(seg, old_size);
@@ -900,7 +956,7 @@ erts_alcu_literal_32_mseg_dealloc(Allctr_t *allctr, void *seg, Uint size,
 {
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     erts_alcu_mseg_dealloc(allctr, seg, size, flags);
 
@@ -910,7 +966,7 @@ erts_alcu_literal_32_mseg_dealloc(Allctr_t *allctr, void *seg, Uint size,
 #elif defined(ARCH_64) && defined(ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION)
 
 /* For allocators that have their own mmapper (super carrier),
- * like literal_alloc and exec_alloc on amd64
+ * like literal_alloc.
  */
 void*
 erts_alcu_mmapper_mseg_alloc(Allctr_t *allctr, Uint *size_p, Uint flags)
@@ -952,10 +1008,10 @@ erts_alcu_mmapper_mseg_dealloc(Allctr_t *allctr, void *seg, Uint size,
 }
 #endif /* ARCH_64 && ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION */
 
-#if defined(ERTS_ALC_A_EXEC) && !defined(ERTS_HAVE_EXEC_MMAPPER)
+#if defined(ERTS_ALC_A_EXEC)
 
 /*
- * For exec_alloc on non-amd64 that just need memory with PROT_EXEC
+ * For exec_alloc that need memory with PROT_EXEC
  */
 void*
 erts_alcu_exec_mseg_alloc(Allctr_t *allctr, Uint *size_p, Uint flags)
@@ -994,7 +1050,7 @@ erts_alcu_exec_mseg_dealloc(Allctr_t *allctr, void *seg, Uint size, Uint flags)
     ASSERT(r == 0); (void)r;
     erts_alcu_mseg_dealloc(allctr, seg, size, flags);
 }
-#endif /* ERTS_ALC_A_EXEC && !ERTS_HAVE_EXEC_MMAPPER */
+#endif /* ERTS_ALC_A_EXEC */
 
 #endif /* HAVE_ERTS_MSEG */
 
@@ -1060,7 +1116,7 @@ erts_alcu_literal_32_sys_alloc(Allctr_t *allctr, Uint* size_p, int superalign)
     Uint size = ERTS_SUPERALIGNED_CEILING(*size_p);
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     res = erts_alcu_sys_alloc(allctr, &size, 1);
     if (res) {
@@ -1078,7 +1134,7 @@ erts_alcu_literal_32_sys_realloc(Allctr_t *allctr, void *ptr, Uint* size_p, Uint
 
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     if (ptr && old_size)
         clear_literal_range(ptr, old_size);
@@ -1095,7 +1151,7 @@ erts_alcu_literal_32_sys_dealloc(Allctr_t *allctr, void *ptr, Uint size, int sup
 {
     ERTS_LC_ASSERT(allctr->alloc_no == ERTS_ALC_A_LITERAL &&
                    allctr->t == 0);
-    ERTS_SMP_LC_ASSERT(allctr->thread_safe);
+    ERTS_LC_ASSERT(allctr->thread_safe);
 
     erts_alcu_sys_dealloc(allctr, ptr, size, 1);
 
@@ -1197,8 +1253,6 @@ unlink_carrier(CarrierList_t *cl, Carrier_t *crr)
 #endif
 }
 
-#ifdef ERTS_SMP
-
 static ERTS_INLINE int is_abandoned(Carrier_t *crr)
 {
     return crr->cpool.state != ERTS_MBC_IS_HOME;
@@ -1222,18 +1276,17 @@ clear_busy_pool_carrier(Allctr_t *allctr, Carrier_t *crr)
 	max_size = (erts_aint_t) allctr->largest_fblk_in_mbc(allctr, crr);
 	erts_atomic_set_nob(&crr->cpool.max_size, max_size);
 
-        iallctr = erts_smp_atomic_read_nob(&crr->allctr);
+        iallctr = erts_atomic_read_nob(&crr->allctr);
         ERTS_ALC_CPOOL_ASSERT((iallctr & ~ERTS_CRR_ALCTR_FLG_HOMECOMING)
                               == ((erts_aint_t)allctr |
                                   ERTS_CRR_ALCTR_FLG_IN_POOL |
                                   ERTS_CRR_ALCTR_FLG_BUSY));
 
 	iallctr &= ~ERTS_CRR_ALCTR_FLG_BUSY;
-	erts_smp_atomic_set_relb(&crr->allctr, iallctr);
+	erts_atomic_set_relb(&crr->allctr, iallctr);
     }
 }
 
-#endif /* ERTS_SMP */
 
 #if 0
 #define ERTS_DBG_CHK_FIX_LIST(A, FIX, IX, B)			\
@@ -1257,12 +1310,10 @@ chk_fix_list(Allctr_t *allctr, ErtsAlcFixList_t *fix, int ix, int before)
 
 static void *mbc_alloc(Allctr_t *allctr, Uint size);
 
-#ifdef ERTS_SMP
 typedef struct {
     ErtsAllctrDDBlock_t ddblock__; /* must be first */
     ErtsAlcType_t fix_type;
 } ErtsAllctrFixDDBlock_t;
-#endif
 
 #define ERTS_ALC_FIX_NO_UNUSE (((ErtsAlcType_t) 1) << ERTS_ALC_N_BITS)
 
@@ -1273,11 +1324,9 @@ dealloc_fix_block(Allctr_t *allctr,
 		  ErtsAlcFixList_t *fix,
 		  int dec_cc_on_redirect)
 {
-#ifdef ERTS_SMP
     /* May be redirected... */
     ASSERT((type & ERTS_ALC_FIX_NO_UNUSE) == 0);
     ((ErtsAllctrFixDDBlock_t *) ptr)->fix_type = type | ERTS_ALC_FIX_NO_UNUSE;
-#endif
     dealloc_block(allctr, ptr, fix, dec_cc_on_redirect);
 }
 
@@ -1311,12 +1360,10 @@ fix_cpool_check_shrink(Allctr_t *allctr,
 	    fix->u.cpool.shrink_list = 0;
 	else {
 	    void *p;
-#ifdef ERTS_SMP
 	    if (busy_pcrr_pp) {
 		clear_busy_pool_carrier(allctr, *busy_pcrr_pp);
 		*busy_pcrr_pp = NULL;
 	    }
-#endif
 	    fix->u.cpool.shrink_list--;
 	    p = fix->list;
 	    fix->list = *((void **) p);
@@ -1409,10 +1456,8 @@ fix_cpool_alloc_shrink(Allctr_t *allctr, erts_aint32_t flgs)
     int ix, o;
     int flush = flgs == 0;
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
-#endif
 
     for (ix = 0; ix < ERTS_ALC_NO_FIXED_SIZES; ix++) {
 	ErtsAlcFixList_t *fix = &allctr->fix[ix];
@@ -1452,10 +1497,8 @@ fix_cpool_alloc_shrink(Allctr_t *allctr, erts_aint32_t flgs)
     if (all_empty)
 	sched_fix_shrink(allctr, 0);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_unlock(&allctr->mutex);
-#endif
 
     return res;
 }
@@ -1567,10 +1610,8 @@ fix_nocpool_alloc_shrink(Allctr_t *allctr, erts_aint32_t flgs)
     int ix, o;
     int flush = flgs == 0;
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
-#endif
 
     for (ix = 0; ix < ERTS_ALC_NO_FIXED_SIZES; ix++) {
 	ErtsAlcFixList_t *fix = &allctr->fix[ix];
@@ -1612,10 +1653,8 @@ fix_nocpool_alloc_shrink(Allctr_t *allctr, erts_aint32_t flgs)
     if (all_empty)
 	sched_fix_shrink(allctr, 0);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_unlock(&allctr->mutex);
-#endif
 
     return res;
 }
@@ -1641,7 +1680,6 @@ dealloc_mbc(Allctr_t *allctr, Carrier_t *crr)
     dealloc_carrier(allctr, crr, 1);
 }
 
-#ifdef ERTS_SMP
 
 static void set_new_allctr_abandon_limit(Allctr_t*);
 static void abandon_carrier(Allctr_t*, Carrier_t*);
@@ -1687,7 +1725,7 @@ get_used_allctr(Allctr_t *pref_allctr, int pref_lock, void *p, UWord *sizep,
 	crr = BLK_TO_SBC(blk);
 	if (sizep)
 	    *sizep = SBC_BLK_SZ(blk) - ABLK_HDR_SZ;  
-	iallctr = erts_smp_atomic_read_dirty(&crr->allctr);
+	iallctr = erts_atomic_read_dirty(&crr->allctr);
     }
     else {
 	crr = ABLK_TO_MBC(blk);
@@ -1695,10 +1733,10 @@ get_used_allctr(Allctr_t *pref_allctr, int pref_lock, void *p, UWord *sizep,
 	if (sizep)
 	    *sizep = MBC_ABLK_SZ(blk) - ABLK_HDR_SZ;
 	if (!ERTS_ALC_IS_CPOOL_ENABLED(pref_allctr))
-	    iallctr = erts_smp_atomic_read_dirty(&crr->allctr);
+	    iallctr = erts_atomic_read_dirty(&crr->allctr);
 	else {
 	    int locked_pref_allctr = 0;
-	    iallctr = erts_smp_atomic_read_ddrb(&crr->allctr);
+	    iallctr = erts_atomic_read_ddrb(&crr->allctr);
 
 	    if (ERTS_ALC_TS_PREF_LOCK_IF_USED == pref_lock
 		&& pref_allctr->thread_safe) {
@@ -1722,14 +1760,14 @@ get_used_allctr(Allctr_t *pref_allctr, int pref_lock, void *p, UWord *sizep,
                      * We need a mathing read barrier to guarantee a correct view
                      * of the carrier for deallocation work.
                      */
-                    act = erts_smp_atomic_cmpxchg_rb(&crr->allctr,
-                                                     iallctr|ERTS_CRR_ALCTR_FLG_BUSY,
-                                                     iallctr);
+                    act = erts_atomic_cmpxchg_rb(&crr->allctr,
+                                                 iallctr|ERTS_CRR_ALCTR_FLG_BUSY,
+                                                 iallctr);
                 }
                 else {
-                    act = erts_smp_atomic_cmpxchg_ddrb(&crr->allctr,
-                                                       iallctr|ERTS_CRR_ALCTR_FLG_BUSY,
-                                                       iallctr);
+                    act = erts_atomic_cmpxchg_ddrb(&crr->allctr,
+                                                   iallctr|ERTS_CRR_ALCTR_FLG_BUSY,
+                                                   iallctr);
                 }
 		if (act == iallctr) {
 		    *busy_pcrr_pp = crr;
@@ -2087,7 +2125,7 @@ handle_delayed_dealloc(Allctr_t *allctr,
 	    ERTS_ALC_CPOOL_ASSERT(ERTS_ALC_IS_CPOOL_ENABLED(allctr));
 	    ERTS_ALC_CPOOL_ASSERT(allctr == crr->cpool.orig_allctr);
 
-            iallctr = erts_smp_atomic_read_nob(&crr->allctr);
+            iallctr = erts_atomic_read_nob(&crr->allctr);
             ASSERT(iallctr & ERTS_CRR_ALCTR_FLG_HOMECOMING);
             while (1) {
                 if ((iallctr & (~ERTS_CRR_ALCTR_FLG_MASK |
@@ -2097,7 +2135,7 @@ handle_delayed_dealloc(Allctr_t *allctr,
                      * Carrier is home (mine and not in pool)
                      */
                     ASSERT(!(iallctr & ERTS_CRR_ALCTR_FLG_BUSY));
-                    erts_smp_atomic_set_nob(&crr->allctr, (erts_aint_t)allctr);
+                    erts_atomic_set_nob(&crr->allctr, (erts_aint_t)allctr);
                     if (IS_FREE_LAST_MBC_BLK(first_blk))
                         dealloc_my_carrier(allctr, crr);
                     else
@@ -2107,7 +2145,7 @@ handle_delayed_dealloc(Allctr_t *allctr,
                     erts_aint_t exp = iallctr;
                     erts_aint_t want = iallctr & ~ERTS_CRR_ALCTR_FLG_HOMECOMING;
 
-                    iallctr = erts_smp_atomic_cmpxchg_nob(&crr->allctr,
+                    iallctr = erts_atomic_cmpxchg_nob(&crr->allctr,
                                                           want,
                                                           exp);
                     if (iallctr != exp)
@@ -2243,7 +2281,6 @@ erts_alcu_check_delayed_dealloc(Allctr_t *allctr,
 			   thr_prgr_p,
 			   more_work);
 }
-#endif
 
 #define ERTS_ALCU_HANDLE_DD_IN_OP(Allctr, Locked)			\
     handle_delayed_dealloc((Allctr), (Locked), 1, 			\
@@ -2254,24 +2291,18 @@ dealloc_block(Allctr_t *allctr, void *ptr, ErtsAlcFixList_t *fix, int dec_cc_on_
 {
     Block_t *blk = UMEM2BLK(ptr);
 
-    ERTS_SMP_LC_ASSERT(!allctr->thread_safe
+    ERTS_LC_ASSERT(!allctr->thread_safe
 		       || erts_lc_mtx_is_locked(&allctr->mutex));
 
     if (IS_SBC_BLK(blk)) {
 	destroy_carrier(allctr, blk, NULL);
-#ifdef ERTS_SMP
 	if (fix && ERTS_ALC_IS_CPOOL_ENABLED(allctr)) {
 	    ErtsAlcType_t type = ((ErtsAllctrFixDDBlock_t *) ptr)->fix_type;
 	    if (!(type & ERTS_ALC_FIX_NO_UNUSE))
 		fix->u.cpool.used--;
 	    fix->u.cpool.allocated--;
 	}
-#endif
     }
-#ifndef ERTS_SMP
-    else
-	mbc_free(allctr, ptr, NULL);
-#else
     else if (!ERTS_ALC_IS_CPOOL_ENABLED(allctr))
 	mbc_free(allctr, ptr, NULL);
     else {
@@ -2302,7 +2333,6 @@ dealloc_block(Allctr_t *allctr, void *ptr, ErtsAlcFixList_t *fix, int dec_cc_on_
 		erts_alloc_notify_delayed_dealloc(used_allctr->ix);
 	}
     }
-#endif
 }
 
 /* Multi block carrier alloc/realloc/free ... */
@@ -2548,12 +2578,10 @@ mbc_free(Allctr_t *allctr, void *p, Carrier_t **busy_pcrr_pp)
     else {
 	(*allctr->link_free_block)(allctr, blk);
 	HARD_CHECK_BLK_CARRIER(allctr, blk);
-#ifdef ERTS_SMP
         if (busy_pcrr_pp && *busy_pcrr_pp)
             update_pooled_tree(allctr, crr, blk_sz);
         else
             check_abandon_carrier(allctr, blk, busy_pcrr_pp);
-#endif
     }
 }
 
@@ -2587,7 +2615,6 @@ mbc_realloc(Allctr_t *allctr, void *p, Uint size, Uint32 alcu_flgs,
 	return NULL;
 #else /* !MBC_REALLOC_ALWAYS_MOVES */
 
-#ifdef ERTS_SMP
     if (busy_pcrr_pp && *busy_pcrr_pp) {
         /*
          * Don't want to use carrier in pool
@@ -2601,7 +2628,6 @@ mbc_realloc(Allctr_t *allctr, void *p, Uint size, Uint32 alcu_flgs,
         mbc_free(allctr, p, busy_pcrr_pp);
         return new_p;
     }
-#endif
 
     get_blk_sz = blk_sz = UMEMSZ2BLKSZ(allctr, size);
 
@@ -2722,9 +2748,7 @@ mbc_realloc(Allctr_t *allctr, void *p, Uint size, Uint32 alcu_flgs,
 		    
 	HARD_CHECK_BLK_CARRIER(allctr, blk);
 
-#ifdef ERTS_SMP
 	check_abandon_carrier(allctr, nxt_blk, NULL);
-#endif
 
 	return p;
     }
@@ -2938,7 +2962,6 @@ mbc_realloc(Allctr_t *allctr, void *p, Uint size, Uint32 alcu_flgs,
 #endif /* !MBC_REALLOC_ALWAYS_MOVES */
 }
 
-#ifdef ERTS_SMP
 
 #define ERTS_ALC_MAX_DEALLOC_CARRIER		10
 #define ERTS_ALC_CPOOL_MAX_FETCH_INSPECT	100
@@ -3308,7 +3331,7 @@ cpool_fetch(Allctr_t *allctr, UWord size)
 
         aoff_remove_pooled_mbc(allctr, crr);
 
-        exp = erts_smp_atomic_read_nob(&crr->allctr);
+        exp = erts_atomic_read_nob(&crr->allctr);
         if (exp & ERTS_CRR_ALCTR_FLG_IN_POOL) {
             ASSERT((exp & ~ERTS_CRR_ALCTR_FLG_MASK) == (erts_aint_t)allctr);
             if (erts_atomic_read_nob(&crr->cpool.max_size) < size) {
@@ -3335,9 +3358,9 @@ cpool_fetch(Allctr_t *allctr, UWord size)
             }
 
             /* Try to fetch it... */
-            act = erts_smp_atomic_cmpxchg_mb(&crr->allctr,
-                                             exp & ~ERTS_CRR_ALCTR_FLG_IN_POOL,
-                                             exp);
+            act = erts_atomic_cmpxchg_mb(&crr->allctr,
+                                         exp & ~ERTS_CRR_ALCTR_FLG_IN_POOL,
+                                         exp);
             if (act == exp) {
                 cpool_delete(allctr, allctr, crr);
                 crr->cpool.state = ERTS_MBC_IS_HOME;
@@ -3368,7 +3391,7 @@ cpool_fetch(Allctr_t *allctr, UWord size)
         erts_aint_t iallctr;
 
         crr = ErtsContainerStruct(allctr->cpool.pooled_tree, Carrier_t, cpool.pooled);
-        iallctr = erts_smp_atomic_read_nob(&crr->allctr);
+        iallctr = erts_atomic_read_nob(&crr->allctr);
         if (iallctr & ERTS_CRR_ALCTR_FLG_IN_POOL) {
             cpool_entrance = &crr->cpool;
             break;
@@ -3431,7 +3454,7 @@ cpool_fetch(Allctr_t *allctr, UWord size)
             loop_state = HAS_SEEN_SENTINEL;
 	}
 	crr = ErtsContainerStruct(cpdp, Carrier_t, cpool);
-	exp = erts_smp_atomic_read_rb(&crr->allctr);
+	exp = erts_atomic_read_rb(&crr->allctr);
 
         if (erts_atomic_read_nob(&cpdp->max_size) < size) {
             INC_CC(allctr->cpool.stat.skip_size);
@@ -3442,7 +3465,7 @@ cpool_fetch(Allctr_t *allctr, UWord size)
             erts_aint_t want = (((erts_aint_t) allctr)
                                 | (exp & ERTS_CRR_ALCTR_FLG_HOMECOMING));
             /* Try to fetch it... */
-	    act = erts_smp_atomic_cmpxchg_mb(&crr->allctr, want, exp);
+	    act = erts_atomic_cmpxchg_mb(&crr->allctr, want, exp);
 	    if (act == exp) {
 		cpool_delete(allctr, ((Allctr_t *) (act & ~ERTS_CRR_ALCTR_FLG_MASK)), crr);
 		if (crr->cpool.orig_allctr == allctr) {
@@ -3471,7 +3494,7 @@ check_dc_list:
 	if (erts_atomic_read_nob(&crr->cpool.max_size) >= size) {
 	    Block_t* blk;
 	    unlink_carrier(&allctr->cpool.dc_list, crr);
-	    ERTS_ALC_CPOOL_ASSERT(erts_smp_atomic_read_nob(&crr->allctr)
+	    ERTS_ALC_CPOOL_ASSERT(erts_atomic_read_nob(&crr->allctr)
                                   == ((erts_aint_t) allctr));
 	    blk = MBC_TO_FIRST_BLK(allctr, crr);
 	    ASSERT(FBLK_TO_MBC(blk) == crr);
@@ -3553,7 +3576,7 @@ schedule_dealloc_carrier(Allctr_t *allctr, Carrier_t *crr)
     orig_allctr = crr->cpool.orig_allctr;
 
     if (allctr == orig_allctr) {
-        if (!(erts_smp_atomic_read_nob(&crr->allctr) & ERTS_CRR_ALCTR_FLG_HOMECOMING)) {
+        if (!(erts_atomic_read_nob(&crr->allctr) & ERTS_CRR_ALCTR_FLG_HOMECOMING)) {
             dealloc_my_carrier(allctr, crr);
         }
         /*else
@@ -3580,13 +3603,13 @@ schedule_dealloc_carrier(Allctr_t *allctr, Carrier_t *crr)
 	ERTS_ALC_CPOOL_ASSERT(IS_MBC_FIRST_ABLK(allctr, first_blk));
 	ERTS_ALC_CPOOL_ASSERT(crr == FBLK_TO_MBC(first_blk));
 	ERTS_ALC_CPOOL_ASSERT(crr == FIRST_BLK_TO_MBC(allctr, first_blk));
-	ERTS_ALC_CPOOL_ASSERT((erts_smp_atomic_read_nob(&crr->allctr)
+	ERTS_ALC_CPOOL_ASSERT((erts_atomic_read_nob(&crr->allctr)
                                & ~ERTS_CRR_ALCTR_FLG_HOMECOMING)
                               == (erts_aint_t) allctr);
 #endif
 
         iallctr = (erts_aint_t)orig_allctr | ERTS_CRR_ALCTR_FLG_HOMECOMING;
-        if (!(erts_smp_atomic_xchg_nob(&crr->allctr, iallctr)
+        if (!(erts_atomic_xchg_nob(&crr->allctr, iallctr)
               & ERTS_CRR_ALCTR_FLG_HOMECOMING)) {
             enqueue_homecoming(allctr, crr);
         }
@@ -3691,11 +3714,11 @@ abandon_carrier(Allctr_t *allctr, Carrier_t *crr)
     cpool_insert(allctr, crr);
 
 
-    iallctr = erts_smp_atomic_read_nob(&crr->allctr);
+    iallctr = erts_atomic_read_nob(&crr->allctr);
     if (allctr == crr->cpool.orig_allctr) {
         /* preserve HOMECOMING flag */
         ASSERT((iallctr & ~ERTS_CRR_ALCTR_FLG_HOMECOMING) == (erts_aint_t)allctr);
-        erts_smp_atomic_set_wb(&crr->allctr, iallctr | ERTS_CRR_ALCTR_FLG_IN_POOL);
+        erts_atomic_set_wb(&crr->allctr, iallctr | ERTS_CRR_ALCTR_FLG_IN_POOL);
         poolify_my_carrier(allctr, crr);
     }
     else {
@@ -3703,7 +3726,7 @@ abandon_carrier(Allctr_t *allctr, Carrier_t *crr)
         iallctr = ((erts_aint_t)crr->cpool.orig_allctr |
                    ERTS_CRR_ALCTR_FLG_HOMECOMING |
                    ERTS_CRR_ALCTR_FLG_IN_POOL);
-        if (!(erts_smp_atomic_xchg_wb(&crr->allctr, iallctr)
+        if (!(erts_atomic_xchg_wb(&crr->allctr, iallctr)
               & ERTS_CRR_ALCTR_FLG_HOMECOMING)) {
 
             enqueue_homecoming(allctr, crr);
@@ -3784,7 +3807,6 @@ cpool_read_stat(Allctr_t *allctr, UWord *nocp, UWord *cszp, UWord *nobp, UWord *
 }
 
 
-#endif /* ERTS_SMP */
 
 #ifdef DEBUG
 
@@ -3885,7 +3907,6 @@ create_carrier(Allctr_t *allctr, Uint umem_sz, UWord flags)
         blk_sz = UMEMSZ2BLKSZ(allctr, umem_sz);
     }
 
-#ifdef ERTS_SMP
     allctr->cpool.disable_abandon = ERTS_ALC_CPOOL_MAX_DISABLE_ABANDON;
 
     if ((flags & (CFLG_MBC|CFLG_NO_CPOOL)) == CFLG_MBC
@@ -3902,7 +3923,6 @@ create_carrier(Allctr_t *allctr, Uint umem_sz, UWord flags)
 	    return blk;
 	}
     }
-#endif
 
 #if HAVE_ERTS_MSEG
 
@@ -4032,9 +4052,7 @@ create_carrier(Allctr_t *allctr, Uint umem_sz, UWord flags)
 	    allctr->main_carrier = crr;
 	}
 
-#ifdef ERTS_SMP
 	cpool_init_carrier_data(allctr, crr);
-#endif
 
 	link_carrier(&allctr->mbc_list, crr);
 
@@ -4254,9 +4272,8 @@ destroy_carrier(Allctr_t *allctr, Block_t *blk, Carrier_t **busy_pcrr_pp)
 	}
 #endif
 
-#ifdef ERTS_SMP
 	if (busy_pcrr_pp && *busy_pcrr_pp) {
-            erts_aint_t iallctr = erts_smp_atomic_read_nob(&crr->allctr);
+            erts_aint_t iallctr = erts_atomic_read_nob(&crr->allctr);
 	    ERTS_ALC_CPOOL_ASSERT(*busy_pcrr_pp == crr);
             ERTS_ALC_CPOOL_ASSERT((iallctr & ~ERTS_CRR_ALCTR_FLG_HOMECOMING)
                                   == (((erts_aint_t) allctr)
@@ -4265,13 +4282,12 @@ destroy_carrier(Allctr_t *allctr, Block_t *blk, Carrier_t **busy_pcrr_pp)
             ERTS_ALC_CPOOL_ASSERT(allctr == crr->cpool.orig_allctr);
 
             *busy_pcrr_pp = NULL;
-	    erts_smp_atomic_set_nob(&crr->allctr,
-                                    (iallctr & ~(ERTS_CRR_ALCTR_FLG_IN_POOL |
-                                                 ERTS_CRR_ALCTR_FLG_BUSY)));
+	    erts_atomic_set_nob(&crr->allctr,
+                                (iallctr & ~(ERTS_CRR_ALCTR_FLG_IN_POOL |
+                                             ERTS_CRR_ALCTR_FLG_BUSY)));
 	    cpool_delete(allctr, allctr, crr);
 	}
-        else
-#endif
+	else
 	{
 	    unlink_carrier(&allctr->mbc_list, crr);
 #if HAVE_ERTS_MSEG
@@ -4302,11 +4318,7 @@ destroy_carrier(Allctr_t *allctr, Block_t *blk, Carrier_t **busy_pcrr_pp)
         }
 #endif
 
-#ifdef ERTS_SMP
 	schedule_dealloc_carrier(allctr, crr);
-#else
-	dealloc_mbc(allctr, crr);
-#endif
     }
 }
 
@@ -4322,6 +4334,7 @@ static struct {
     Eterm e;
     Eterm t;
     Eterm ramv;
+    Eterm atags;
 #if HAVE_ERTS_MSEG
     Eterm asbcst;
     Eterm rsbcst;
@@ -4350,7 +4363,6 @@ static struct {
     Eterm fix_types;
 
     Eterm mbcs;
-#ifdef ERTS_SMP
     Eterm mbcs_pool;
     Eterm fetch;
     Eterm fail_pooled;
@@ -4363,7 +4375,6 @@ static struct {
     Eterm skip_homecoming;
     Eterm skip_race;
     Eterm entrance_removed;
-#endif
     Eterm sbcs;
 
     Eterm sys_alloc_carriers_size;
@@ -4393,11 +4404,11 @@ static struct {
 #endif
 } am;
 
-static Eterm fix_type_atoms[ERTS_ALC_NO_FIXED_SIZES];
+static Eterm alloc_type_atoms[ERTS_ALC_N_MAX + 1];
 
 static ERTS_INLINE void atom_init(Eterm *atom, char *name)
 {
-    *atom = am_atom_put(name, strlen(name));
+    *atom = am_atom_put(name, sys_strlen(name));
 }
 #define AM_INIT(AM) atom_init(&am.AM, #AM)
 
@@ -4424,6 +4435,7 @@ init_atoms(Allctr_t *allctr)
 	AM_INIT(e);
 	AM_INIT(t);
 	AM_INIT(ramv);
+	AM_INIT(atags);
 #if HAVE_ERTS_MSEG
 	AM_INIT(asbcst);
 	AM_INIT(rsbcst);
@@ -4452,7 +4464,6 @@ init_atoms(Allctr_t *allctr)
 	AM_INIT(fix_types);
 
 	AM_INIT(mbcs);
-#ifdef ERTS_SMP
 	AM_INIT(mbcs_pool);
 	AM_INIT(fetch);
         AM_INIT(fail_pooled);
@@ -4465,7 +4476,6 @@ init_atoms(Allctr_t *allctr)
         AM_INIT(skip_homecoming);
         AM_INIT(skip_race);
         AM_INIT(entrance_removed);
-#endif
 	AM_INIT(sbcs);
 
 	AM_INIT(sys_alloc_carriers_size);
@@ -4497,12 +4507,12 @@ init_atoms(Allctr_t *allctr)
 	}
 #endif
 
-	for (ix = 0; ix < ERTS_ALC_NO_FIXED_SIZES; ix++) {
-	    ErtsAlcType_t n = ERTS_ALC_N_MIN_A_FIXED_SIZE + ix;
-	    char *name = (char *) ERTS_ALC_N2TD(n);
-	    size_t len = strlen(name);
-	    fix_type_atoms[ix] = am_atom_put(name, len);
-	}
+        for (ix = ERTS_ALC_N_MIN; ix <= ERTS_ALC_N_MAX; ix++) {
+            const char *name = ERTS_ALC_N2TD(ix);
+            size_t len = sys_strlen(name);
+
+            alloc_type_atoms[ix] = am_atom_put(name, len);
+        }
     }
     
     if (allctr && !allctr->atoms_initialized) {
@@ -4615,6 +4625,7 @@ sz_info_fix(Allctr_t *allctr,
 		ErtsAlcFixList_t *fix = &allctr->fix[ix];
 		UWord alloced = fix->type_size * fix->u.cpool.allocated;
 		UWord used = fix->type_size * fix->u.cpool.used;
+                ErtsAlcType_t n = ERTS_ALC_N_MIN_A_FIXED_SIZE + ix;
 
 		if (print_to_p) {
 		    fmtfn_t to = *print_to_p;
@@ -4622,15 +4633,14 @@ sz_info_fix(Allctr_t *allctr,
 		    erts_print(to,
 			       arg,
 			       "fix type internal: %s %bpu %bpu\n",
-			       (char *) ERTS_ALC_N2TD(ERTS_ALC_N_MIN_A_FIXED_SIZE
-						      + ix),
+			       (char *) ERTS_ALC_N2TD(n),
 			       alloced,
 			       used);
 		}
 
 		if (hpp || szp) {
 		    add_3tup(hpp, szp, &res,
-			     fix_type_atoms[ix],
+			     alloc_type_atoms[n],
 			     bld_unstable_uint(hpp, szp, alloced),
 			     bld_unstable_uint(hpp, szp, used));
 		}
@@ -4643,6 +4653,7 @@ sz_info_fix(Allctr_t *allctr,
 	    ErtsAlcFixList_t *fix = &allctr->fix[ix];
 	    UWord alloced = fix->type_size * fix->u.nocpool.allocated;
 	    UWord used = fix->type_size*fix->u.nocpool.used;
+            ErtsAlcType_t n = ERTS_ALC_N_MIN_A_FIXED_SIZE + ix;
 
 	    if (print_to_p) {
 		fmtfn_t to = *print_to_p;
@@ -4650,15 +4661,14 @@ sz_info_fix(Allctr_t *allctr,
 		erts_print(to,
 			   arg,
 			   "fix type: %s %bpu %bpu\n",
-			   (char *) ERTS_ALC_N2TD(ERTS_ALC_N_MIN_A_FIXED_SIZE
-						  + ix),
+			   (char *) ERTS_ALC_N2TD(n),
 			   alloced,
 			   used);
 	    }
 
 	    if (hpp || szp) {
 		add_3tup(hpp, szp, &res,
-			 fix_type_atoms[ix],
+			 alloc_type_atoms[n],
 			 bld_unstable_uint(hpp, szp, alloced),
 			 bld_unstable_uint(hpp, szp, used));
 	    }
@@ -4715,7 +4725,6 @@ sz_info_carriers(Allctr_t *allctr,
     return res;
 }
 
-#ifdef ERTS_SMP
 
 static Eterm
 info_cpool(Allctr_t *allctr,
@@ -4816,7 +4825,6 @@ info_cpool(Allctr_t *allctr,
     return res;
 }
 
-#endif /* ERTS_SMP */
 
 static Eterm
 info_carriers(Allctr_t *allctr,
@@ -4940,20 +4948,20 @@ make_name_atoms(Allctr_t *allctr)
     char realloc[] = "realloc";
     char free[] = "free";
     char buf[MAX_ATOM_CHARACTERS];
-    size_t prefix_len = strlen(allctr->name_prefix);
+    size_t prefix_len = sys_strlen(allctr->name_prefix);
 
     if (prefix_len > MAX_ATOM_CHARACTERS + sizeof(realloc) - 1)
 	erts_exit(ERTS_ERROR_EXIT,"Too long allocator name: %salloc\n",allctr->name_prefix);
 
-    memcpy((void *) buf, (void *) allctr->name_prefix, prefix_len);
+    sys_memcpy((void *) buf, (void *) allctr->name_prefix, prefix_len);
 
-    memcpy((void *) &buf[prefix_len], (void *) alloc, sizeof(alloc) - 1);
+    sys_memcpy((void *) &buf[prefix_len], (void *) alloc, sizeof(alloc) - 1);
     allctr->name.alloc = am_atom_put(buf, prefix_len + sizeof(alloc) - 1);
 
-    memcpy((void *) &buf[prefix_len], (void *) realloc, sizeof(realloc) - 1);
+    sys_memcpy((void *) &buf[prefix_len], (void *) realloc, sizeof(realloc) - 1);
     allctr->name.realloc = am_atom_put(buf, prefix_len + sizeof(realloc) - 1);
 
-    memcpy((void *) &buf[prefix_len], (void *) free, sizeof(free) - 1);
+    sys_memcpy((void *) &buf[prefix_len], (void *) free, sizeof(free) - 1);
     allctr->name.free = am_atom_put(buf, prefix_len + sizeof(free) - 1);
 
 }
@@ -5071,13 +5079,9 @@ info_options(Allctr_t *allctr,
 	return res;
     }
 
-#ifdef ERTS_SMP
     acul = allctr->cpool.util_limit;
     acnl = allctr->cpool.in_pool_limit;
     acfml = allctr->cpool.fblk_min_limit;
-#else
-    acul = 0; acnl = 0; acfml = 0;
-#endif
 
     if (print_to_p) {
 	char topt[21]; /* Enough for any 64-bit integer */
@@ -5090,6 +5094,7 @@ info_options(Allctr_t *allctr,
 		   "option e: true\n"
 		   "option t: %s\n"
 		   "option ramv: %s\n"
+		   "option atags: %s\n"
 		   "option sbct: %beu\n"
 #if HAVE_ERTS_MSEG
 		   "option asbcst: %bpu\n"
@@ -5108,6 +5113,7 @@ info_options(Allctr_t *allctr,
 		   "option acul: %bpu\n",
 		   topt,
 		   allctr->ramv ? "true" : "false",
+		   allctr->atags ? "true" : "false",
 		   allctr->sbc_threshold,
 #if HAVE_ERTS_MSEG
 		   allctr->mseg_opt.abs_shrink_th,
@@ -5177,6 +5183,7 @@ info_options(Allctr_t *allctr,
 		 am_sbct,
 		 bld_uint(hpp, szp, allctr->sbc_threshold));
 	add_2tup(hpp, szp, &res, am.ramv, allctr->ramv ? am_true : am_false);
+	add_2tup(hpp, szp, &res, am.atags, allctr->atags ? am_true : am_false);
 	add_2tup(hpp, szp, &res, am.t, (allctr->t ? am_true : am_false));
 	add_2tup(hpp, szp, &res, am.e, am_true);
     }
@@ -5266,19 +5273,15 @@ erts_alcu_info_options(Allctr_t *allctr,
     if (hpp || szp)
 	ensure_atoms_initialized(allctr);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe) {
 	erts_allctr_wrapper_pre_lock();
 	erts_mtx_lock(&allctr->mutex);
     }
-#endif
     res = info_options(allctr, print_to_p, print_to_arg, hpp, szp);
-#ifdef USE_THREADS
     if (allctr->thread_safe) { 
 	erts_mtx_unlock(&allctr->mutex);
 	erts_allctr_wrapper_pre_unlock();
     }
-#endif
     return res;
 }
 
@@ -5294,9 +5297,7 @@ erts_alcu_sz_info(Allctr_t *allctr,
 		  Uint *szp)
 {
     Eterm res, mbcs, sbcs, fix = THE_NON_VALUE;
-#ifdef ERTS_SMP
     Eterm mbcs_pool;
-#endif
 
     res  = THE_NON_VALUE;
 
@@ -5311,12 +5312,10 @@ erts_alcu_sz_info(Allctr_t *allctr,
     if (hpp || szp)
 	ensure_atoms_initialized(allctr);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe) {
 	erts_allctr_wrapper_pre_lock();
 	erts_mtx_lock(&allctr->mutex);
     }
-#endif
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(allctr);
 
@@ -5332,23 +5331,19 @@ erts_alcu_sz_info(Allctr_t *allctr,
 	fix = sz_info_fix(allctr, internal, print_to_p, print_to_arg, hpp, szp);
     mbcs = sz_info_carriers(allctr, &allctr->mbcs, "mbcs ", print_to_p,
 			    print_to_arg, hpp, szp);
-#ifdef ERTS_SMP
     if (ERTS_ALC_IS_CPOOL_ENABLED(allctr))
 	mbcs_pool = info_cpool(allctr, 1, "mbcs_pool ", print_to_p,
 			       print_to_arg, hpp, szp);
     else
 	mbcs_pool = THE_NON_VALUE; /* shut up annoying warning... */
-#endif
     sbcs = sz_info_carriers(allctr, &allctr->sbcs, "sbcs ", print_to_p,
 			    print_to_arg, hpp, szp);
 
     if (hpp || szp) {
 	res = NIL;
 	add_2tup(hpp, szp, &res, am.sbcs, sbcs);
-#ifdef ERTS_SMP
 	if (ERTS_ALC_IS_CPOOL_ENABLED(allctr))
 	    add_2tup(hpp, szp, &res, am.mbcs_pool, mbcs_pool);
-#endif
 	add_2tup(hpp, szp, &res, am.mbcs, mbcs);
 	add_fix_types(allctr, internal, hpp, szp, &res, fix);
     }
@@ -5359,12 +5354,10 @@ erts_alcu_sz_info(Allctr_t *allctr,
     }
 
 
-#ifdef USE_THREADS
     if (allctr->thread_safe) {
 	erts_mtx_unlock(&allctr->mutex);
 	erts_allctr_wrapper_pre_unlock();
     }
-#endif
 
     return res;
 }
@@ -5380,9 +5373,7 @@ erts_alcu_info(Allctr_t *allctr,
 	       Uint *szp)
 {
     Eterm res, sett, mbcs, sbcs, calls, fix = THE_NON_VALUE;
-#ifdef ERTS_SMP
     Eterm mbcs_pool;
-#endif
 
     res  = THE_NON_VALUE;
 
@@ -5397,12 +5388,10 @@ erts_alcu_info(Allctr_t *allctr,
     if (hpp || szp)
 	ensure_atoms_initialized(allctr);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe) {
 	erts_allctr_wrapper_pre_lock();
 	erts_mtx_lock(&allctr->mutex);
     }
-#endif
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(allctr);
 
@@ -5427,13 +5416,11 @@ erts_alcu_info(Allctr_t *allctr,
 	fix = sz_info_fix(allctr, internal, print_to_p, print_to_arg, hpp, szp);
     mbcs = info_carriers(allctr, &allctr->mbcs, "mbcs ", print_to_p,
 			 print_to_arg, hpp, szp);
-#ifdef ERTS_SMP
     if (ERTS_ALC_IS_CPOOL_ENABLED(allctr))
 	mbcs_pool = info_cpool(allctr, 0, "mbcs_pool ", print_to_p,
 			       print_to_arg, hpp, szp);
     else
 	mbcs_pool = THE_NON_VALUE; /* shut up annoying warning... */
-#endif
     sbcs = info_carriers(allctr, &allctr->sbcs, "sbcs ", print_to_p,
 			 print_to_arg, hpp, szp);
     calls = info_calls(allctr, print_to_p, print_to_arg, hpp, szp);
@@ -5443,10 +5430,8 @@ erts_alcu_info(Allctr_t *allctr,
 
 	add_2tup(hpp, szp, &res, am.calls, calls);
 	add_2tup(hpp, szp, &res, am.sbcs, sbcs);
-#ifdef ERTS_SMP
 	if (ERTS_ALC_IS_CPOOL_ENABLED(allctr))
 	    add_2tup(hpp, szp, &res, am.mbcs_pool, mbcs_pool);
-#endif
 	add_2tup(hpp, szp, &res, am.mbcs, mbcs);
 	add_fix_types(allctr, internal, hpp, szp, &res, fix);
 	add_2tup(hpp, szp, &res, am.options, sett);
@@ -5462,12 +5447,10 @@ erts_alcu_info(Allctr_t *allctr,
     }
 
 
-#ifdef USE_THREADS
     if (allctr->thread_safe) {
 	erts_mtx_unlock(&allctr->mutex);
 	erts_allctr_wrapper_pre_unlock();
     }
-#endif
 
     return res;
 }
@@ -5477,10 +5460,8 @@ void
 erts_alcu_current_size(Allctr_t *allctr, AllctrSize_t *size, ErtsAlcUFixInfo_t *fi, int fisz)
 {
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
-#endif
 
     size->carriers = allctr->mbcs.curr.norm.mseg.size;
     size->carriers += allctr->mbcs.curr.norm.sys_alloc.size;
@@ -5490,14 +5471,12 @@ erts_alcu_current_size(Allctr_t *allctr, AllctrSize_t *size, ErtsAlcUFixInfo_t *
     size->blocks = allctr->mbcs.blocks.curr.size;
     size->blocks += allctr->sbcs.blocks.curr.size;
 
-#ifdef ERTS_SMP
     if (ERTS_ALC_IS_CPOOL_ENABLED(allctr)) {
 	UWord csz, bsz;
 	cpool_read_stat(allctr, NULL, &csz, NULL, &bsz);
 	size->blocks += bsz;
 	size->carriers += csz;
     }
-#endif
 
     if (fi) {
 	int ix;
@@ -5519,25 +5498,22 @@ erts_alcu_current_size(Allctr_t *allctr, AllctrSize_t *size, ErtsAlcUFixInfo_t *
 	}
     }
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_unlock(&allctr->mutex);
-#endif
 }
 
 /* ----------------------------------------------------------------------- */
 
 static ERTS_INLINE void *
-do_erts_alcu_alloc(ErtsAlcType_t type, void *extra, Uint size)
+do_erts_alcu_alloc(ErtsAlcType_t type, Allctr_t *allctr, Uint size)
 {
-    Allctr_t *allctr = (Allctr_t *) extra; 
     void *res;
 
     ASSERT(initialized);
 
     ASSERT(allctr);
 
-    ERTS_SMP_LC_ASSERT(!allctr->thread_safe
+    ERTS_LC_ASSERT(!allctr->thread_safe
 		       || erts_lc_mtx_is_locked(&allctr->mutex));
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(allctr);
@@ -5569,41 +5545,57 @@ do_erts_alcu_alloc(ErtsAlcType_t type, void *extra, Uint size)
 
 void *erts_alcu_alloc(ErtsAlcType_t type, void *extra, Uint size)
 {
+    Allctr_t *allctr = (Allctr_t *) extra;
     void *res;
-#ifdef ERTS_SMP
+
     ASSERT(!"This is not thread safe");
-#elif defined(USE_THREADS)
-    ASSERT(erts_equal_tids(erts_main_thread, erts_thr_self()));
-#endif
-    res = do_erts_alcu_alloc(type, extra, size);
+
+    res = do_erts_alcu_alloc(type, allctr, size);
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, determine_alloc_tag(allctr, type));
+    }
+
     DEBUG_CHECK_ALIGNMENT(res);
+
     return res;
 }
 
 
-#ifdef USE_THREADS
 
 void *
 erts_alcu_alloc_ts(ErtsAlcType_t type, void *extra, Uint size)
 {
     Allctr_t *allctr = (Allctr_t *) extra;
+    alcu_atag_t tag = 0;
     void *res;
+
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     erts_mtx_lock(&allctr->mutex);
-    res = do_erts_alcu_alloc(type, extra, size);
 
-    DEBUG_CHECK_ALIGNMENT(res);
+    res = do_erts_alcu_alloc(type, allctr, size);
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
+    }
 
     erts_mtx_unlock(&allctr->mutex);
+
+    DEBUG_CHECK_ALIGNMENT(res);
+
     return res;
 }
 
-#ifdef ERTS_SMP
 
 void *
 erts_alcu_alloc_thr_spec(ErtsAlcType_t type, void *extra, Uint size)
 {
     ErtsAllocatorThrSpec_t *tspec = (ErtsAllocatorThrSpec_t *) extra;
     int ix;
+    alcu_atag_t tag = 0;
     Allctr_t *allctr;
     void *res;
 
@@ -5613,11 +5605,19 @@ erts_alcu_alloc_thr_spec(ErtsAlcType_t type, void *extra, Uint size)
 
     allctr = tspec->allctr[ix];
 
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
 
     res = do_erts_alcu_alloc(type, allctr, size);
 
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
+    }
+
     if (allctr->thread_safe)
 	erts_mtx_unlock(&allctr->mutex);
 
@@ -5630,54 +5630,55 @@ void *
 erts_alcu_alloc_thr_pref(ErtsAlcType_t type, void *extra, Uint size)
 {
     Allctr_t *pref_allctr;
+    alcu_atag_t tag = 0;
     void *res;
 
     pref_allctr = get_pref_allctr(extra);
 
+    if (pref_allctr->atags) {
+        tag = determine_alloc_tag(pref_allctr, type);
+    }
+
     if (pref_allctr->thread_safe)
 	erts_mtx_lock(&pref_allctr->mutex);
 
-#ifdef ERTS_SMP
     ASSERT(pref_allctr->dd.use);
     ERTS_ALCU_HANDLE_DD_IN_OP(pref_allctr, 1);
-#endif
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(pref_allctr);
 
     res = do_erts_alcu_alloc(type, pref_allctr, size);
 
-#ifdef ERTS_SMP
     if (!res && ERTS_ALCU_HANDLE_DD_IN_OP(pref_allctr, 1)) {
 	/* Cleaned up a bit more; try one more time... */
 	res = do_erts_alcu_alloc(type, pref_allctr, size);
     }
-#endif
+
+    if (pref_allctr->atags && res) {
+        set_alloc_tag(pref_allctr, res, tag);
+    }
 
     if (pref_allctr->thread_safe)
 	erts_mtx_unlock(&pref_allctr->mutex);
 
     DEBUG_CHECK_ALIGNMENT(res);
 
-
     return res;
 }
 
-#endif
 
-#endif
 
 /* ------------------------------------------------------------------------- */
 
 static ERTS_INLINE void
-do_erts_alcu_free(ErtsAlcType_t type, void *extra, void *p,
+do_erts_alcu_free(ErtsAlcType_t type, Allctr_t *allctr, void *p,
 		  Carrier_t **busy_pcrr_pp)
 {
-    Allctr_t *allctr = (Allctr_t *) extra; 
     ASSERT(initialized);
 
     ASSERT(allctr);
 
-    ERTS_SMP_LC_ASSERT(!allctr->thread_safe
+    ERTS_LC_ASSERT(!allctr->thread_safe
 		       || erts_lc_mtx_is_locked(&allctr->mutex));
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(allctr);
@@ -5704,21 +5705,20 @@ do_erts_alcu_free(ErtsAlcType_t type, void *extra, void *p,
 
 void erts_alcu_free(ErtsAlcType_t type, void *extra, void *p)
 {
-    do_erts_alcu_free(type, extra, p, NULL);
+    Allctr_t *allctr = (Allctr_t *) extra;
+    do_erts_alcu_free(type, allctr, p, NULL);
 }
 
-#ifdef USE_THREADS
 
 void
 erts_alcu_free_ts(ErtsAlcType_t type, void *extra, void *p)
 {
     Allctr_t *allctr = (Allctr_t *) extra;
     erts_mtx_lock(&allctr->mutex);
-    do_erts_alcu_free(type, extra, p, NULL);
+    do_erts_alcu_free(type, allctr, p, NULL);
     erts_mtx_unlock(&allctr->mutex);
 }
 
-#ifdef ERTS_SMP
 
 void
 erts_alcu_free_thr_spec(ErtsAlcType_t type, void *extra, void *p)
@@ -5769,21 +5769,18 @@ erts_alcu_free_thr_pref(ErtsAlcType_t type, void *extra, void *p)
     }
 }
 
-#endif
 
-#endif
 
 /* ------------------------------------------------------------------------- */
 
 static ERTS_INLINE void *
 do_erts_alcu_realloc(ErtsAlcType_t type,
-		     void *extra,
+		     Allctr_t *allctr,
 		     void *p,
 		     Uint size,
 		     Uint32 alcu_flgs,
 		     Carrier_t **busy_pcrr_pp)
 {
-    Allctr_t *allctr = (Allctr_t *) extra; 
     Block_t *blk;
     void *res;
 
@@ -5791,13 +5788,13 @@ do_erts_alcu_realloc(ErtsAlcType_t type,
 
     ASSERT(allctr);
 
-    ERTS_SMP_LC_ASSERT(!allctr->thread_safe
+    ERTS_LC_ASSERT(!allctr->thread_safe
 		       || erts_lc_mtx_is_locked(&allctr->mutex));
 
     ERTS_ALCU_DBG_CHK_THR_ACCESS(allctr);
 
     if (!p) {
-	res = do_erts_alcu_alloc(type, extra, size);
+	res = do_erts_alcu_alloc(type, allctr, size);
 	INC_CC(allctr->calls.this_realloc);
 	DEC_CC(allctr->calls.this_alloc);
 	return res;
@@ -5806,7 +5803,7 @@ do_erts_alcu_realloc(ErtsAlcType_t type,
 #if ALLOC_ZERO_EQ_NULL
     if (!size) {
 	ASSERT(p);
-	do_erts_alcu_free(type, extra, p, busy_pcrr_pp);
+	do_erts_alcu_free(type, allctr, p, busy_pcrr_pp);
 	INC_CC(allctr->calls.this_realloc);
 	DEC_CC(allctr->calls.this_free);
 	return NULL;
@@ -5891,19 +5888,29 @@ do_erts_alcu_realloc(ErtsAlcType_t type,
 void *
 erts_alcu_realloc(ErtsAlcType_t type, void *extra, void *p, Uint size)
 {
+    Allctr_t *allctr = (Allctr_t *)extra;
     void *res;
-    res = do_erts_alcu_realloc(type, extra, p, size, 0, NULL);
+
+    res = do_erts_alcu_realloc(type, allctr, p, size, 0, NULL);
+
     DEBUG_CHECK_ALIGNMENT(res);
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, determine_alloc_tag(allctr, type));
+    }
+
     return res;
 }
 
 void *
 erts_alcu_realloc_mv(ErtsAlcType_t type, void *extra, void *p, Uint size)
 {
+    Allctr_t *allctr = (Allctr_t *)extra;
     void *res;
-    res = do_erts_alcu_alloc(type, extra, size);
+
+    res = do_erts_alcu_alloc(type, allctr, size);
     if (!res)
-	res = erts_alcu_realloc(type, extra, p, size);
+        res = do_erts_alcu_realloc(type, allctr, p, size, 0, NULL);
     else {
 	Block_t *blk;
 	size_t cpy_size;
@@ -5913,24 +5920,42 @@ erts_alcu_realloc_mv(ErtsAlcType_t type, void *extra, void *p, Uint size)
 	if (cpy_size > size)
 	    cpy_size = size;
 	sys_memcpy(res, p, cpy_size);
-	do_erts_alcu_free(type, extra, p, NULL);
+	do_erts_alcu_free(type, allctr, p, NULL);
     }
+
     DEBUG_CHECK_ALIGNMENT(res);
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, determine_alloc_tag(allctr, type));
+    }
+
     return res;
 }
 
 
-#ifdef USE_THREADS
-
 void *
 erts_alcu_realloc_ts(ErtsAlcType_t type, void *extra, void *ptr, Uint size)
 {
     Allctr_t *allctr = (Allctr_t *) extra;
+    alcu_atag_t tag = 0;
     void *res;
+
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     erts_mtx_lock(&allctr->mutex);
-    res = do_erts_alcu_realloc(type, extra, ptr, size, 0, NULL);
+
+    res = do_erts_alcu_realloc(type, allctr, ptr, size, 0, NULL);
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
+    }
+
     erts_mtx_unlock(&allctr->mutex);
+
     DEBUG_CHECK_ALIGNMENT(res);
+
     return res;
 }
 
@@ -5938,11 +5963,17 @@ void *
 erts_alcu_realloc_mv_ts(ErtsAlcType_t type, void *extra, void *p, Uint size)
 {
     Allctr_t *allctr = (Allctr_t *) extra;
+    alcu_atag_t tag = 0;
     void *res;
+
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     erts_mtx_lock(&allctr->mutex);
-    res = do_erts_alcu_alloc(type, extra, size);
+    res = do_erts_alcu_alloc(type, allctr, size);
     if (!res)
-	res = erts_alcu_realloc_ts(type, extra, p, size);
+	res = do_erts_alcu_realloc(type, allctr, p, size, 0, NULL);
     else {
 	Block_t *blk;
 	size_t cpy_size;
@@ -5952,14 +5983,20 @@ erts_alcu_realloc_mv_ts(ErtsAlcType_t type, void *extra, void *p, Uint size)
 	if (cpy_size > size)
 	    cpy_size = size;
 	sys_memcpy(res, p, cpy_size);
-	do_erts_alcu_free(type, extra, p, NULL);
+	do_erts_alcu_free(type, allctr, p, NULL);
+    }
+
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
     }
+
     erts_mtx_unlock(&allctr->mutex);
+
     DEBUG_CHECK_ALIGNMENT(res);
+
     return res;
 }
 
-#ifdef ERTS_SMP
 
 void *
 erts_alcu_realloc_thr_spec(ErtsAlcType_t type, void *extra,
@@ -5967,6 +6004,7 @@ erts_alcu_realloc_thr_spec(ErtsAlcType_t type, void *extra,
 {
     ErtsAllocatorThrSpec_t *tspec = (ErtsAllocatorThrSpec_t *) extra;
     int ix;
+    alcu_atag_t tag = 0;
     Allctr_t *allctr;
     void *res;
 
@@ -5976,11 +6014,19 @@ erts_alcu_realloc_thr_spec(ErtsAlcType_t type, void *extra,
 
     allctr = tspec->allctr[ix];
 
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
 
     res = do_erts_alcu_realloc(type, allctr, ptr, size, 0, NULL);
 
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
+    }
+
     if (allctr->thread_safe)
 	erts_mtx_unlock(&allctr->mutex);
 
@@ -5995,6 +6041,7 @@ erts_alcu_realloc_mv_thr_spec(ErtsAlcType_t type, void *extra,
 {
     ErtsAllocatorThrSpec_t *tspec = (ErtsAllocatorThrSpec_t *) extra;
     int ix;
+    alcu_atag_t tag = 0;
     Allctr_t *allctr;
     void *res;
 
@@ -6004,14 +6051,16 @@ erts_alcu_realloc_mv_thr_spec(ErtsAlcType_t type, void *extra,
 
     allctr = tspec->allctr[ix];
 
+    if (allctr->atags) {
+        tag = determine_alloc_tag(allctr, type);
+    }
+
     if (allctr->thread_safe)
 	erts_mtx_lock(&allctr->mutex);
 
     res = do_erts_alcu_alloc(type, allctr, size);
     if (!res) {
-	if (allctr->thread_safe)
-	    erts_mtx_unlock(&allctr->mutex);
-	res = erts_alcu_realloc_thr_spec(type, allctr, ptr, size);
+        res = do_erts_alcu_realloc(type, allctr, ptr, size, 0, NULL);
     }
     else {
 	Block_t *blk;
@@ -6023,41 +6072,42 @@ erts_alcu_realloc_mv_thr_spec(ErtsAlcType_t type, void *extra,
 	    cpy_size = size;
 	sys_memcpy(res, ptr, cpy_size);
 	do_erts_alcu_free(type, allctr, ptr, NULL);
-	if (allctr->thread_safe)
-	    erts_mtx_unlock(&allctr->mutex);
     }
 
+    if (allctr->atags && res) {
+        set_alloc_tag(allctr, res, tag);
+    }
+
+    if (allctr->thread_safe)
+        erts_mtx_unlock(&allctr->mutex);
+
     DEBUG_CHECK_ALIGNMENT(res);
 
     return res;
 }
 
 static ERTS_INLINE void *
-realloc_thr_pref(ErtsAlcType_t type, void *extra, void *p, Uint size,
+realloc_thr_pref(ErtsAlcType_t type, Allctr_t *pref_allctr, void *p, Uint size,
 		 int force_move)
 {
     void *res;
-    Allctr_t *pref_allctr, *used_allctr;
+    Allctr_t *used_allctr;
     UWord old_user_size;
     Carrier_t *busy_pcrr_p;
-#ifdef ERTS_SMP
+    alcu_atag_t tag = 0;
     int retried;
-#endif
 
-    if (!p)
-	return erts_alcu_alloc_thr_pref(type, extra, size);
-
-    pref_allctr = get_pref_allctr(extra);
+    if (pref_allctr->atags) {
+        tag = determine_alloc_tag(pref_allctr, type);
+    }
 
     if (pref_allctr->thread_safe)
 	erts_mtx_lock(&pref_allctr->mutex);
 
-#ifdef ERTS_SMP
     ASSERT(pref_allctr->dd.use);
     ERTS_ALCU_HANDLE_DD_IN_OP(pref_allctr, 1);
     retried = 0;
 restart:
-#endif
 
     used_allctr = get_used_allctr(pref_allctr, ERTS_ALC_TS_PREF_LOCK_NO,
 				  p, &old_user_size, &busy_pcrr_p);
@@ -6073,13 +6123,16 @@ restart:
 				   0,
 				   &busy_pcrr_p);
 	clear_busy_pool_carrier(used_allctr, busy_pcrr_p);
-#ifdef ERTS_SMP
 	if (!res && !retried && ERTS_ALCU_HANDLE_DD_IN_OP(pref_allctr, 1)) {
 	    /* Cleaned up a bit more; try one more time... */
 	    retried = 1;
 	    goto restart;
 	}
-#endif	    
+
+        if (pref_allctr->atags && res) {
+            set_alloc_tag(pref_allctr, res, tag);
+        }
+
 	if (pref_allctr->thread_safe)
 	    erts_mtx_unlock(&pref_allctr->mutex);
     }
@@ -6088,6 +6141,9 @@ restart:
 	if (!res)
 	    goto unlock_ts_return;
 	else {
+            if (pref_allctr->atags) {
+                set_alloc_tag(pref_allctr, res, tag);
+            }
 
 	    DEBUG_CHECK_ALIGNMENT(res);
 
@@ -6118,25 +6174,37 @@ restart:
 	}
     }
 
+    DEBUG_CHECK_ALIGNMENT(res);
+
     return res;
 }
 
 void *
 erts_alcu_realloc_thr_pref(ErtsAlcType_t type, void *extra, void *p, Uint size)
 {
-    return realloc_thr_pref(type, extra, p, size, 0);
+    if (p) {
+        Allctr_t *pref_allctr = get_pref_allctr(extra);
+
+        return realloc_thr_pref(type, pref_allctr, p, size, 0);
+    }
+
+    return erts_alcu_alloc_thr_pref(type, extra, size);
 }
 
 void *
 erts_alcu_realloc_mv_thr_pref(ErtsAlcType_t type, void *extra,
 			      void *p, Uint size)
 {
-    return realloc_thr_pref(type, extra, p, size, 1);
+    if (p) {
+        Allctr_t *pref_allctr = get_pref_allctr(extra);
+
+        return realloc_thr_pref(type, pref_allctr, p, size, 1);
+    }
+
+    return erts_alcu_alloc_thr_pref(type, extra, size);
 }
 
-#endif
 
-#endif
 
 static Uint adjust_sbct(Allctr_t* allctr, Uint sbct)
 {
@@ -6188,10 +6256,8 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
     sys_memcpy((void *) &allctr->mseg_opt,
 	       (void *) &erts_mseg_default_opt,
 	       sizeof(ErtsMsegOpt_t));
-#ifdef ERTS_SMP
     if (init->tspec || init->tpref)
 	allctr->mseg_opt.sched_spec = 1;
-#endif /* ERTS_SMP */
 #endif /* HAVE_ERTS_MSEG */
 
     allctr->name_prefix			= init->name_prefix;
@@ -6219,6 +6285,7 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 	allctr->t			= 0;
 
     allctr->ramv			= init->ramv;
+    allctr->atags                       = init->atags;
     allctr->main_carrier_size		= init->mmbcs;
 
 #if HAVE_ERTS_MSEG
@@ -6249,7 +6316,6 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 	goto error;
     allctr->min_block_size		= UNIT_CEILING(allctr->min_block_size
 						       + sizeof(FreeBlkFtr_t));
-#ifdef ERTS_SMP
     if (init->tpref) {
 	Uint sz = ABLK_HDR_SZ;
 	sz += (init->fix ? 
@@ -6279,7 +6345,6 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
         allctr->cpool.in_pool_limit = 0;
         allctr->cpool.fblk_min_limit = 0;
     }
-#endif
 
     allctr->sbc_threshold = adjust_sbct(allctr, init->sbct);
 
@@ -6288,7 +6353,6 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 	allctr->mseg_opt.abs_shrink_th = ~((UWord) 0) / 100;
 #endif
 
-#ifdef USE_THREADS
     if (init->ts) {
 	allctr->thread_safe = 1;
 
@@ -6299,7 +6363,6 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 	allctr->debug.saved_tid = 0;
 #endif
     }
-#endif
 
     if(!allctr->get_free_block
        || !allctr->link_free_block
@@ -6312,14 +6375,12 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 
     if (allctr->mbc_header_size < sizeof(Carrier_t))
 	goto error;
-#ifdef ERTS_SMP
     allctr->dd.use = 0;
     if (init->tpref) {
 	allctr->dd.use = 1;
 	init_dd_queue(&allctr->dd.q);
 	allctr->dd.ix = init->ix;
     }
-#endif
     allctr->mbc_header_size = (UNIT_CEILING(allctr->mbc_header_size
 					    + ABLK_HDR_SZ)
 			       - ABLK_HDR_SZ);
@@ -6377,10 +6438,8 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 			     | CFLG_NO_CPOOL
 			     | CFLG_MAIN_CARRIER);
 	if (!blk) {
-#ifdef USE_THREADS
 	  if (allctr->thread_safe)
 	    erts_mtx_destroy(&allctr->mutex);
-#endif
 	  erts_exit(ERTS_ABORT_EXIT,
 	    "Failed to create main carrier for %salloc\n",
 	    init->name_prefix);
@@ -6400,9 +6459,7 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 	    allctr->fix[i].type_size = init->fix_type_size[i];
 	    allctr->fix[i].list_size = 0;
 	    allctr->fix[i].list = NULL;
-#ifdef ERTS_SMP
 	    ASSERT(allctr->fix[i].type_size >= sizeof(ErtsAllctrFixDDBlock_t));
-#endif
 	    if (ERTS_ALC_IS_CPOOL_ENABLED(allctr)) {
 		allctr->fix[i].u.cpool.min_list_size = 0;
 		allctr->fix[i].u.cpool.shrink_list = 0;
@@ -6422,10 +6479,8 @@ erts_alcu_start(Allctr_t *allctr, AllctrInit_t *init)
 
  error:
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_destroy(&allctr->mutex);
-#endif
 
     return 0;
 
@@ -6443,10 +6498,8 @@ erts_alcu_stop(Allctr_t *allctr)
     while (allctr->mbc_list.first)
 	destroy_carrier(allctr, MBC_TO_FIRST_BLK(allctr, allctr->mbc_list.first), NULL);
 
-#ifdef USE_THREADS
     if (allctr->thread_safe)
 	erts_mtx_destroy(&allctr->mutex);
-#endif
 
 }
 
@@ -6455,14 +6508,12 @@ erts_alcu_stop(Allctr_t *allctr)
 void
 erts_alcu_init(AlcUInit_t *init)
 {
-#ifdef ERTS_SMP
     int i;
     for (i = 0; i <= ERTS_ALC_A_MAX; i++) {
 	ErtsAlcCPoolData_t *sentinel = &carrier_pool[i].sentinel;
 	erts_atomic_init_nob(&sentinel->next, (erts_aint_t) sentinel);
 	erts_atomic_init_nob(&sentinel->prev, (erts_aint_t) sentinel);
     }
-#endif
     ERTS_CT_ASSERT(SBC_BLK_SZ_MASK == MBC_FBLK_SZ_MASK); /* see BLK_SZ */
 #if HAVE_ERTS_MSEG
     ASSERT(erts_mseg_unit_size() == ERTS_SACRR_UNIT_SZ);
@@ -6484,6 +6535,1072 @@ erts_alcu_init(AlcUInit_t *init)
     initialized = 1;
 }
 
+/* ------------------------------------------------------------------------- */
+
+/* Allocation histograms and carrier information is gathered by walking through
+ * all carriers associated with each allocator instance. This is done as
+ * aux_yield_work on the scheduler that owns each instance.
+ *
+ * Yielding is implemented by temporarily inserting a "dummy carrier" at the
+ * last position. It's permanently "busy" so it won't get picked up by someone
+ * else when in the carrier pool, and we never make the employer aware of it
+ * through callbacks so we can't accidentally allocate on it.
+ *
+ * Plain malloc/free is used to guarantee we won't allocate with the allocator
+ * we're scanning. */
+
+/* Yield between carriers once this many blocks have been processed. Note that
+ * a single carrier scan may exceed this figure. */
+#ifndef DEBUG
+    #define BLOCKSCAN_REDUCTIONS (8000)
+#else
+    #define BLOCKSCAN_REDUCTIONS (400)
+#endif
+
+/* Abort a single carrier scan after this many blocks to prevent really large
+ * MBCs from blocking forever. */
+#define BLOCKSCAN_BAILOUT_THRESHOLD (16000)
+
+typedef struct alcu_blockscan {
+    /* A per-scheduler list used when multiple scans have been queued. The
+     * current scanner will always run until completion/abort before moving on
+     * to the next. */
+    struct alcu_blockscan *scanner_queue;
+
+    Allctr_t *allocator;
+    Process *process;
+
+    int (*current_op)(struct alcu_blockscan *scanner);
+    int (*next_op)(struct alcu_blockscan *scanner);
+    int reductions;
+
+    ErtsAlcCPoolData_t *cpool_cursor;
+    CarrierList_t *current_clist;
+    Carrier_t *clist_cursor;
+    Carrier_t dummy_carrier;
+
+    /* Called if the process that started this job dies before we're done. */
+    void (*abort)(void *user_data);
+
+    /* Called on each carrier. The callback must return the number of blocks
+     * scanned to yield properly between carriers.
+     *
+     * Note that it's not possible to "yield back" into a carrier. */
+    int (*scan)(Allctr_t *, void *user_data, Carrier_t *);
+
+    /* Called when all carriers have been scanned. The callback may return
+     * non-zero to yield. */
+    int (*finish)(void *user_data);
+
+    void *user_data;
+} blockscan_t;
+
+static Carrier_t *blockscan_restore_clist_cursor(blockscan_t *state)
+{
+    Carrier_t *cursor = state->clist_cursor;
+
+    ASSERT(state->clist_cursor == (state->current_clist)->first ||
+           state->clist_cursor == &state->dummy_carrier);
+
+    if (cursor == &state->dummy_carrier) {
+        cursor = cursor->next;
+
+        unlink_carrier(state->current_clist, state->clist_cursor);
+    }
+
+    return cursor;
+}
+
+static void blockscan_save_clist_cursor(blockscan_t *state, Carrier_t *after)
+{
+    ASSERT(state->clist_cursor == (state->current_clist)->first ||
+           state->clist_cursor == &state->dummy_carrier);
+
+    state->clist_cursor = &state->dummy_carrier;
+
+    (state->clist_cursor)->next = after->next;
+    (state->clist_cursor)->prev = after;
+
+    relink_carrier(state->current_clist, state->clist_cursor);
+}
+
+static int blockscan_clist_yielding(blockscan_t *state)
+{
+    Carrier_t *cursor = blockscan_restore_clist_cursor(state);
+
+    if (ERTS_PROC_IS_EXITING(state->process)) {
+        return 0;
+    }
+
+    while (cursor) {
+        /* Skip dummy carriers inserted by another (concurrent) block scan.
+         * This can happen when scanning thread-safe allocators from multiple
+         * schedulers. */
+        if (CARRIER_SZ(cursor) > 0) {
+            int blocks_scanned = state->scan(state->allocator,
+                                             state->user_data,
+                                             cursor);
+
+            state->reductions -= blocks_scanned;
+
+            if (state->reductions <= 0) {
+                blockscan_save_clist_cursor(state, cursor);
+                return 1;
+            }
+        }
+
+        cursor = cursor->next;
+    }
+
+    return 0;
+}
+
+static ErtsAlcCPoolData_t *blockscan_restore_cpool_cursor(blockscan_t *state)
+{
+    ErtsAlcCPoolData_t *cursor;
+
+    cursor = cpool_aint2cpd(cpool_read(&(state->cpool_cursor)->next));
+
+    if (state->cpool_cursor == &state->dummy_carrier.cpool) {
+        cpool_delete(state->allocator, state->allocator, &state->dummy_carrier);
+    }
+
+    return cursor;
+}
+
+static void blockscan_save_cpool_cursor(blockscan_t *state,
+                                        ErtsAlcCPoolData_t *after)
+{
+    ErtsAlcCPoolData_t *dummy_carrier, *prev_carrier, *next_carrier;
+
+    dummy_carrier = &state->dummy_carrier.cpool;
+
+    next_carrier = cpool_aint2cpd(cpool_mod_mark(&after->next));
+    prev_carrier = cpool_aint2cpd(cpool_mod_mark(&next_carrier->prev));
+
+    cpool_init(&dummy_carrier->next, (erts_aint_t)next_carrier);
+    cpool_init(&dummy_carrier->prev, (erts_aint_t)prev_carrier);
+
+    cpool_set_mod_marked(&prev_carrier->next,
+                         (erts_aint_t)dummy_carrier,
+                         (erts_aint_t)next_carrier);
+    cpool_set_mod_marked(&next_carrier->prev,
+                         (erts_aint_t)dummy_carrier,
+                         (erts_aint_t)prev_carrier);
+
+    state->cpool_cursor = dummy_carrier;
+}
+
+static int blockscan_cpool_yielding(blockscan_t *state)
+{
+    ErtsAlcCPoolData_t *sentinel, *cursor;
+
+    sentinel = &carrier_pool[(state->allocator)->alloc_no].sentinel;
+    cursor = blockscan_restore_cpool_cursor(state);
+
+    if (ERTS_PROC_IS_EXITING(state->process)) {
+        return 0;
+    }
+
+    while (cursor != sentinel) {
+        Carrier_t *carrier;
+        erts_aint_t exp;
+
+        /* When a deallocation happens on a pooled carrier it will be routed to
+         * its owner, so the only way to be sure that it isn't modified while
+         * scanning is to skip all carriers that aren't ours. The deallocations
+         * deferred to us will get handled when we're done. */
+        while (cursor->orig_allctr != state->allocator) {
+            cursor = cpool_aint2cpd(cpool_read(&cursor->next));
+
+            if (cursor == sentinel) {
+                return 0;
+            }
+        }
+
+        carrier = ErtsContainerStruct(cursor, Carrier_t, cpool);
+        exp = erts_atomic_read_rb(&carrier->allctr);
+
+        if (exp & ERTS_CRR_ALCTR_FLG_IN_POOL) {
+            ASSERT(state->allocator == (Allctr_t*)(exp & ~ERTS_CRR_ALCTR_FLG_MASK));
+            ASSERT(!(exp & ERTS_CRR_ALCTR_FLG_BUSY));
+
+            if (erts_atomic_cmpxchg_acqb(&carrier->allctr,
+                                         exp | ERTS_CRR_ALCTR_FLG_BUSY,
+                                         exp) == exp) {
+                /* Skip dummy carriers inserted by another (concurrent) block
+                 * scan. This can happen when scanning thread-safe allocators
+                 * from multiple schedulers. */
+                if (CARRIER_SZ(carrier) > 0) {
+                    int blocks_scanned = state->scan(state->allocator,
+                                                     state->user_data,
+                                                     carrier);
+
+                    state->reductions -= blocks_scanned;
+
+                    if (state->reductions <= 0) {
+                        blockscan_save_cpool_cursor(state, cursor);
+                        erts_atomic_set_relb(&carrier->allctr, exp);
+
+                        return 1;
+                    }
+                }
+
+                erts_atomic_set_relb(&carrier->allctr, exp);
+            }
+        }
+
+        cursor = cpool_aint2cpd(cpool_read(&cursor->next));
+    }
+
+    return 0;
+}
+
+static int blockscan_yield_helper(blockscan_t *state,
+                                  int (*yielding_op)(blockscan_t*))
+{
+    /* Note that we don't check whether to abort here; only yielding_op knows
+     * whether the carrier is still in the list/pool. */
+
+    if ((state->allocator)->thread_safe) {
+        /* Locked scans have to be as short as possible. */
+        state->reductions = 1;
+
+        erts_mtx_lock(&(state->allocator)->mutex);
+    } else {
+        state->reductions = BLOCKSCAN_REDUCTIONS;
+    }
+
+    if (yielding_op(state)) {
+        state->next_op = state->current_op;
+    }
+
+    if ((state->allocator)->thread_safe) {
+        erts_mtx_unlock(&(state->allocator)->mutex);
+    }
+
+    return 1;
+}
+
+/* */
+
+static int blockscan_finish(blockscan_t *state)
+{
+    if (ERTS_PROC_IS_EXITING(state->process)) {
+        state->abort(state->user_data);
+        return 0;
+    }
+
+    state->current_op = blockscan_finish;
+
+    return state->finish(state->user_data);
+}
+
+static int blockscan_sweep_sbcs(blockscan_t *state)
+{
+    if (state->current_op != blockscan_sweep_sbcs) {
+        SET_CARRIER_HDR(&state->dummy_carrier, 0, SCH_SBC, state->allocator);
+        state->current_clist = &(state->allocator)->sbc_list;
+        state->clist_cursor = (state->current_clist)->first;
+    }
+
+    state->current_op = blockscan_sweep_sbcs;
+    state->next_op = blockscan_finish;
+
+    return blockscan_yield_helper(state, blockscan_clist_yielding);
+}
+
+static int blockscan_sweep_mbcs(blockscan_t *state)
+{
+    if (state->current_op != blockscan_sweep_mbcs) {
+        SET_CARRIER_HDR(&state->dummy_carrier, 0, SCH_MBC, state->allocator);
+        state->current_clist = &(state->allocator)->mbc_list;
+        state->clist_cursor = (state->current_clist)->first;
+    }
+
+    state->current_op = blockscan_sweep_mbcs;
+    state->next_op = blockscan_sweep_sbcs;
+
+    return blockscan_yield_helper(state, blockscan_clist_yielding);
+}
+
+static int blockscan_sweep_cpool(blockscan_t *state)
+{
+    if (state->current_op != blockscan_sweep_cpool) {
+        ErtsAlcCPoolData_t *sentinel;
+
+        SET_CARRIER_HDR(&state->dummy_carrier, 0, SCH_MBC, state->allocator);
+        sentinel = &carrier_pool[(state->allocator)->alloc_no].sentinel;
+        state->cpool_cursor = sentinel;
+    }
+
+    state->current_op = blockscan_sweep_cpool;
+    state->next_op = blockscan_sweep_mbcs;
+
+    return blockscan_yield_helper(state, blockscan_cpool_yielding);
+}
+
+static int blockscan_get_specific_allocator(int allocator_num,
+                                            int sched_id,
+                                            Allctr_t **out)
+{
+    ErtsAllocatorInfo_t *ai;
+    Allctr_t *allocator;
+
+    ASSERT(allocator_num >= ERTS_ALC_A_MIN &&
+           allocator_num <= ERTS_ALC_A_MAX);
+    ASSERT(sched_id >= 0 && sched_id <= erts_no_schedulers);
+
+    ai = &erts_allctrs_info[allocator_num];
+
+    if (!ai->enabled || !ai->alloc_util) {
+        return 0;
+    }
+
+    if (!ai->thr_spec) {
+        if (sched_id != 0) {
+            /* Only thread-specific allocators can be scanned on a specific
+             * scheduler. */
+            return 0;
+        }
+
+        allocator = (Allctr_t*)ai->extra;
+        ASSERT(allocator->thread_safe);
+    } else {
+        ErtsAllocatorThrSpec_t *tspec = (ErtsAllocatorThrSpec_t*)ai->extra;
+
+        ASSERT(sched_id < tspec->size);
+
+        allocator = tspec->allctr[sched_id];
+    }
+
+    *out = allocator;
+
+    return 1;
+}
+
+static void blockscan_sched_trampoline(void *arg)
+{
+    ErtsAlcuBlockscanYieldData *yield;
+    ErtsSchedulerData *esdp;
+    blockscan_t *scanner;
+
+    esdp = erts_get_scheduler_data();
+    scanner = (blockscan_t*)arg;
+
+    yield = ERTS_SCHED_AUX_YIELD_DATA(esdp, alcu_blockscan);
+
+    ASSERT((yield->last == NULL) == (yield->current == NULL));
+
+    if (yield->last != NULL) {
+        blockscan_t *prev_scanner = yield->last;
+
+        ASSERT(prev_scanner->scanner_queue == NULL);
+
+        prev_scanner->scanner_queue = scanner;
+    } else {
+        yield->current = scanner;
+    }
+
+    scanner->scanner_queue = NULL;
+    yield->last = scanner;
+
+    erts_notify_new_aux_yield_work(esdp);
+}
+
+static void blockscan_dispatch(blockscan_t *scanner, Process *owner,
+                               Allctr_t *allocator, int sched_id)
+{
+    ASSERT(erts_get_scheduler_id() != 0);
+
+    if (sched_id == 0) {
+        /* Global instances are always handled on the current scheduler. */
+        sched_id = ERTS_ALC_GET_THR_IX();
+        ASSERT(allocator->thread_safe);
+    }
+
+    scanner->allocator = allocator;
+    scanner->process = owner;
+
+    erts_proc_inc_refc(scanner->process);
+
+    cpool_init_carrier_data(scanner->allocator, &scanner->dummy_carrier);
+    erts_atomic_init_nob(&(scanner->dummy_carrier).allctr,
+                         (erts_aint_t)allocator | ERTS_CRR_ALCTR_FLG_BUSY);
+
+    if (ERTS_ALC_IS_CPOOL_ENABLED(scanner->allocator)) {
+        scanner->next_op = blockscan_sweep_cpool;
+    } else {
+        scanner->next_op = blockscan_sweep_mbcs;
+    }
+
+    /* Aux yield jobs can only be set up while running on the scheduler that
+     * services them, so we move there before continuing.
+     *
+     * We can't drive the scan itself through this since the scheduler will
+     * always finish *all* misc aux work in one go which makes it impossible to
+     * yield. */
+    erts_schedule_misc_aux_work(sched_id, blockscan_sched_trampoline, scanner);
+}
+
+int erts_handle_yielded_alcu_blockscan(ErtsSchedulerData *esdp,
+                                       ErtsAlcuBlockscanYieldData *yield)
+{
+    blockscan_t *scanner = yield->current;
+
+    (void)esdp;
+
+    ASSERT((yield->last == NULL) == (yield->current == NULL));
+
+    if (scanner) {
+        if (scanner->next_op(scanner)) {
+            return 1;
+        }
+
+        ASSERT(ERTS_PROC_IS_EXITING(scanner->process) ||
+               scanner->current_op == blockscan_finish);
+
+        yield->current = scanner->scanner_queue;
+
+        if (yield->current == NULL) {
+            ASSERT(scanner == yield->last);
+            yield->last = NULL;
+        }
+
+        erts_proc_dec_refc(scanner->process);
+
+        /* Plain free is intentional. */
+        free(scanner);
+
+        return yield->current != NULL;
+    }
+
+    return 0;
+}
+
+void erts_alcu_sched_spec_data_init(ErtsSchedulerData *esdp)
+{
+    ErtsAlcuBlockscanYieldData *yield;
+
+    yield = ERTS_SCHED_AUX_YIELD_DATA(esdp, alcu_blockscan);
+
+    yield->current = NULL;
+    yield->last = NULL;
+}
+
+/* ------------------------------------------------------------------------- */
+
+static ERTS_INLINE int u64_log2(Uint64 v)
+{
+    static const int log2_tab64[64] = {
+        63,  0, 58,  1, 59, 47, 53,  2,
+        60, 39, 48, 27, 54, 33, 42,  3,
+        61, 51, 37, 40, 49, 18, 28, 20,
+        55, 30, 34, 11, 43, 14, 22,  4,
+        62, 57, 46, 52, 38, 26, 32, 41,
+        50, 36, 17, 19, 29, 10, 13, 21,
+        56, 45, 25, 31, 35, 16,  9, 12,
+        44, 24, 15,  8, 23,  7,  6,  5};
+
+    v |= v >> 1;
+    v |= v >> 2;
+    v |= v >> 4;
+    v |= v >> 8;
+    v |= v >> 16;
+    v |= v >> 32;
+
+    return log2_tab64[((Uint64)((v - (v >> 1))*0x07EDD5E59A4E28C2)) >> 58];
+}
+
+/* ------------------------------------------------------------------------- */
+
+typedef struct hist_tree__ {
+    struct hist_tree__ *parent;
+    struct hist_tree__ *left;
+    struct hist_tree__ *right;
+
+    int is_red;
+
+    alcu_atag_t tag;
+    UWord histogram[1];
+} hist_tree_t;
+
+#define ERTS_RBT_PREFIX hist_tree
+#define ERTS_RBT_T hist_tree_t
+#define ERTS_RBT_KEY_T UWord
+#define ERTS_RBT_FLAGS_T int
+#define ERTS_RBT_INIT_EMPTY_TNODE(T) ((void)0)
+#define ERTS_RBT_IS_RED(T) ((T)->is_red)
+#define ERTS_RBT_SET_RED(T) ((T)->is_red = 1)
+#define ERTS_RBT_IS_BLACK(T) (!ERTS_RBT_IS_RED(T))
+#define ERTS_RBT_SET_BLACK(T) ((T)->is_red = 0)
+#define ERTS_RBT_GET_FLAGS(T) ((T)->is_red)
+#define ERTS_RBT_SET_FLAGS(T, F) ((T)->is_red = F)
+#define ERTS_RBT_GET_PARENT(T) ((T)->parent)
+#define ERTS_RBT_SET_PARENT(T, P) ((T)->parent = P)
+#define ERTS_RBT_GET_RIGHT(T) ((T)->right)
+#define ERTS_RBT_SET_RIGHT(T, R) ((T)->right = (R))
+#define ERTS_RBT_GET_LEFT(T) ((T)->left)
+#define ERTS_RBT_SET_LEFT(T, L) ((T)->left = (L))
+#define ERTS_RBT_GET_KEY(T) ((T)->tag)
+#define ERTS_RBT_IS_LT(KX, KY) (KX < KY)
+#define ERTS_RBT_IS_EQ(KX, KY) (KX == KY)
+#define ERTS_RBT_WANT_FOREACH_DESTROY_YIELDING
+#define ERTS_RBT_WANT_FOREACH_DESTROY
+#define ERTS_RBT_WANT_INSERT
+#define ERTS_RBT_WANT_LOOKUP
+#define ERTS_RBT_UNDEF
+
+#include "erl_rbtree.h"
+
+typedef struct {
+    blockscan_t common;
+
+    ErtsIRefStorage iref;
+    Process *process;
+
+    hist_tree_rbt_yield_state_t hist_tree_yield;
+    hist_tree_t *hist_tree;
+    UWord hist_count;
+
+    UWord hist_slot_start;
+    int hist_slot_count;
+
+    UWord unscanned_size;
+
+    ErtsHeapFactory msg_factory;
+    int building_result;
+    Eterm result_list;
+} gather_ahist_t;
+
+static void gather_ahist_update(gather_ahist_t *state, UWord tag, UWord size)
+{
+    hist_tree_t *hist_node;
+    UWord size_interval;
+    int hist_slot;
+
+    hist_node = hist_tree_rbt_lookup(state->hist_tree, tag);
+
+    if (hist_node == NULL) {
+        /* Plain calloc is intentional. */
+        hist_node = (hist_tree_t*)calloc(1, sizeof(hist_tree_t) +
+                                            (state->hist_slot_count - 1) *
+                                            sizeof(hist_node->histogram[0]));
+        hist_node->tag = tag;
+
+        hist_tree_rbt_insert(&state->hist_tree, hist_node);
+        state->hist_count++;
+    }
+
+    size_interval = (size / state->hist_slot_start);
+    size_interval = u64_log2(size_interval + 1);
+
+    hist_slot = MIN(size_interval, state->hist_slot_count - 1);
+
+    hist_node->histogram[hist_slot]++;
+}
+
+static int gather_ahist_scan(Allctr_t *allocator,
+                             void *user_data,
+                             Carrier_t *carrier)
+{
+    gather_ahist_t *state;
+    int blocks_scanned;
+    Block_t *block;
+
+    state = (gather_ahist_t*)user_data;
+    blocks_scanned = 1;
+
+    if (IS_SB_CARRIER(carrier)) {
+        alcu_atag_t tag;
+
+        block = SBC2BLK(allocator, carrier);
+        tag = GET_BLK_ATAG(block);
+
+        ASSERT(DBG_IS_VALID_ATAG(allocator, tag));
+
+        gather_ahist_update(state, tag, SBC_BLK_SZ(block));
+    } else {
+        UWord scanned_bytes = MBC_HEADER_SIZE(allocator);
+
+        ASSERT(IS_MB_CARRIER(carrier));
+
+        block = MBC_TO_FIRST_BLK(allocator, carrier);
+
+        while (1) {
+            UWord block_size = MBC_BLK_SZ(block);
+
+            if (IS_ALLOCED_BLK(block)) {
+                alcu_atag_t tag = GET_BLK_ATAG(block);
+
+                ASSERT(DBG_IS_VALID_ATAG(allocator, tag));
+
+                gather_ahist_update(state, tag, block_size);
+            }
+
+            scanned_bytes += block_size;
+
+            if (blocks_scanned >= BLOCKSCAN_BAILOUT_THRESHOLD) {
+                state->unscanned_size += CARRIER_SZ(carrier) - scanned_bytes;
+                break;
+            } else if (IS_LAST_BLK(block)) {
+                break;
+            }
+
+            block = NXT_BLK(block);
+            blocks_scanned++;
+        }
+    }
+
+    return blocks_scanned;
+}
+
+static void gather_ahist_append_result(hist_tree_t *node, void *arg)
+{
+    gather_ahist_t *state = (gather_ahist_t*)arg;
+
+    Eterm histogram_tuple, tag_tuple;
+
+    Eterm *hp;
+    int ix;
+
+    ASSERT(state->building_result);
+
+    hp = erts_produce_heap(&state->msg_factory, 7 + state->hist_slot_count, 0);
+
+    hp[0] = make_arityval(state->hist_slot_count);
+
+    for (ix = 0; ix < state->hist_slot_count; ix++) {
+        hp[1 + ix] = make_small(node->histogram[ix]);
+    }
+
+    histogram_tuple = make_tuple(hp);
+    hp += 1 + state->hist_slot_count;
+
+    hp[0] = make_arityval(3);
+    hp[1] = ATAG_ID(node->tag);
+    hp[2] = alloc_type_atoms[ATAG_TYPE(node->tag)];
+    hp[3] = histogram_tuple;
+
+    tag_tuple = make_tuple(hp);
+    hp += 4;
+
+    state->result_list = CONS(hp, tag_tuple, state->result_list);
+
+    /* Plain free is intentional. */
+    free(node);
+}
+
+static void gather_ahist_send(gather_ahist_t *state)
+{
+    Eterm result_tuple, unscanned_size, task_ref;
+
+    Uint term_size;
+    Eterm *hp;
+
+    ASSERT((state->result_list == NIL) ^ (state->hist_count > 0));
+    ASSERT(state->building_result);
+
+    term_size = 4 + erts_iref_storage_heap_size(&state->iref);
+    term_size += IS_USMALL(0, state->unscanned_size) ? 0 : BIG_UINT_HEAP_SIZE;
+
+    hp = erts_produce_heap(&state->msg_factory, term_size, 0);
+
+    task_ref = erts_iref_storage_make_ref(&state->iref, &hp,
+        &(state->msg_factory.message)->hfrag.off_heap, 0);
+
+    unscanned_size = bld_unstable_uint(&hp, NULL, state->unscanned_size);
+
+    hp[0] = make_arityval(3);
+    hp[1] = task_ref;
+    hp[2] = unscanned_size;
+    hp[3] = state->result_list;
+
+    result_tuple = make_tuple(hp);
+
+    erts_factory_trim_and_close(&state->msg_factory, &result_tuple, 1);
+
+    erts_queue_message(state->process, 0, state->msg_factory.message,
+                       result_tuple, am_system);
+}
+
+static int gather_ahist_finish(void *arg)
+{
+    gather_ahist_t *state = (gather_ahist_t*)arg;
+
+    if (!state->building_result) {
+        ErtsMessage *message;
+        Uint minimum_size;
+        Eterm *hp;
+
+        /* {Ref, unscanned size, [{Tag, {Histogram}} | Rest]} */
+        minimum_size = 4 + erts_iref_storage_heap_size(&state->iref) +
+                       state->hist_count * (7 + state->hist_slot_count);
+
+        message = erts_alloc_message(minimum_size, &hp);
+        erts_factory_selfcontained_message_init(&state->msg_factory,
+                                                message, hp);
+
+        ERTS_RBT_YIELD_STAT_INIT(&state->hist_tree_yield);
+
+        state->result_list = NIL;
+        state->building_result = 1;
+    }
+
+    if (hist_tree_rbt_foreach_destroy_yielding(&state->hist_tree,
+                                               &gather_ahist_append_result,
+                                               state,
+                                               &state->hist_tree_yield,
+                                               BLOCKSCAN_REDUCTIONS)) {
+        return 1;
+    }
+
+    gather_ahist_send(state);
+
+    return 0;
+}
+
+static void gather_ahist_destroy_result(hist_tree_t *node, void *arg)
+{
+    (void)arg;
+    free(node);
+}
+
+static void gather_ahist_abort(void *arg)
+{
+    gather_ahist_t *state = (gather_ahist_t*)arg;
+
+    if (state->building_result) {
+        erts_factory_undo(&state->msg_factory);
+    }
+
+    hist_tree_rbt_foreach_destroy(&state->hist_tree,
+                                  &gather_ahist_destroy_result,
+                                  NULL);
+}
+
+int erts_alcu_gather_alloc_histograms(Process *p, int allocator_num,
+                                      int sched_id, int hist_width,
+                                      UWord hist_start, Eterm ref)
+{
+    gather_ahist_t *gather_state;
+    blockscan_t *scanner;
+    Allctr_t *allocator;
+
+    ASSERT(is_internal_ref(ref));
+
+    if (!blockscan_get_specific_allocator(allocator_num,
+                                          sched_id,
+                                          &allocator)) {
+        return 0;
+    } else if (!allocator->atags) {
+        return 0;
+    }
+
+    ensure_atoms_initialized(allocator);
+
+    /* Plain calloc is intentional. */
+    gather_state = (gather_ahist_t*)calloc(1, sizeof(gather_ahist_t));
+    scanner = &gather_state->common;
+
+    scanner->abort = gather_ahist_abort;
+    scanner->scan = gather_ahist_scan;
+    scanner->finish = gather_ahist_finish;
+    scanner->user_data = gather_state;
+
+    erts_iref_storage_save(&gather_state->iref, ref);
+    gather_state->hist_slot_start = hist_start;
+    gather_state->hist_slot_count = hist_width;
+    gather_state->process = p;
+
+    blockscan_dispatch(scanner, p, allocator, sched_id);
+
+    return 1;
+}
+
+/* ------------------------------------------------------------------------- */
+
+typedef struct chist_node__ {
+    struct chist_node__ *next;
+
+    UWord carrier_size;
+    UWord unscanned_size;
+    UWord allocated_size;
+
+    /* BLOCKSCAN_BAILOUT_THRESHOLD guarantees we won't overflow this or the
+     * counters in the free block histogram. */
+    int allocated_count;
+    int flags;
+
+    int histogram[1];
+} chist_node_t;
+
+typedef struct {
+    blockscan_t common;
+
+    ErtsIRefStorage iref;
+    Process *process;
+
+    Eterm allocator_desc;
+
+    chist_node_t *info_list;
+    UWord info_count;
+
+    UWord hist_slot_start;
+    int hist_slot_count;
+
+    ErtsHeapFactory msg_factory;
+    int building_result;
+    Eterm result_list;
+} gather_cinfo_t;
+
+static int gather_cinfo_scan(Allctr_t *allocator,
+                             void *user_data,
+                             Carrier_t *carrier)
+{
+    gather_cinfo_t *state;
+    chist_node_t *node;
+    int blocks_scanned;
+    Block_t *block;
+
+    state = (gather_cinfo_t*)user_data;
+    node = calloc(1, sizeof(chist_node_t) +
+                     (state->hist_slot_count - 1) *
+                     sizeof(node->histogram[0]));
+    blocks_scanned = 1;
+
+    /* ERTS_CRR_ALCTR_FLG_BUSY is ignored since we've set it ourselves and it
+     * would be misleading to include it. */
+    node->flags = erts_atomic_read_rb(&carrier->allctr) &
+                  (ERTS_CRR_ALCTR_FLG_MASK & ~ERTS_CRR_ALCTR_FLG_BUSY);
+    node->carrier_size = CARRIER_SZ(carrier);
+
+    if (IS_SB_CARRIER(carrier)) {
+        UWord block_size;
+
+        block = SBC2BLK(allocator, carrier);
+        block_size = SBC_BLK_SZ(block);
+
+        node->allocated_size = block_size;
+        node->allocated_count = 1;
+    } else {
+        UWord scanned_bytes = MBC_HEADER_SIZE(allocator);
+
+        block = MBC_TO_FIRST_BLK(allocator, carrier);
+
+        while (1) {
+            UWord block_size = MBC_BLK_SZ(block);
+
+            scanned_bytes += block_size;
+
+            if (IS_ALLOCED_BLK(block)) {
+                node->allocated_size += block_size;
+                node->allocated_count++;
+            } else {
+                UWord size_interval;
+                int hist_slot;
+
+                size_interval = (block_size / state->hist_slot_start);
+                size_interval = u64_log2(size_interval + 1);
+
+                hist_slot = MIN(size_interval, state->hist_slot_count - 1);
+
+                node->histogram[hist_slot]++;
+            }
+
+            if (blocks_scanned >= BLOCKSCAN_BAILOUT_THRESHOLD) {
+                node->unscanned_size += CARRIER_SZ(carrier) - scanned_bytes;
+                break;
+            } else if (IS_LAST_BLK(block)) {
+                break;
+            }
+
+            block = NXT_BLK(block);
+            blocks_scanned++;
+        }
+    }
+
+    node->next = state->info_list;
+    state->info_list = node;
+    state->info_count++;
+
+    return blocks_scanned;
+}
+
+static void gather_cinfo_append_result(gather_cinfo_t *state,
+                                       chist_node_t *info)
+{
+    Eterm carrier_size, unscanned_size, allocated_size;
+    Eterm histogram_tuple, carrier_tuple;
+
+    Uint term_size;
+    Eterm *hp;
+    int ix;
+
+    ASSERT(state->building_result);
+
+    term_size = 11 + state->hist_slot_count;
+    term_size += IS_USMALL(0, info->carrier_size) ? 0 : BIG_UINT_HEAP_SIZE;
+    term_size += IS_USMALL(0, info->unscanned_size) ? 0 : BIG_UINT_HEAP_SIZE;
+    term_size += IS_USMALL(0, info->allocated_size) ? 0 : BIG_UINT_HEAP_SIZE;
+
+    hp = erts_produce_heap(&state->msg_factory, term_size, 0);
+
+    hp[0] = make_arityval(state->hist_slot_count);
+
+    for (ix = 0; ix < state->hist_slot_count; ix++) {
+        hp[1 + ix] = make_small(info->histogram[ix]);
+    }
+
+    histogram_tuple = make_tuple(hp);
+    hp += 1 + state->hist_slot_count;
+
+    carrier_size = bld_unstable_uint(&hp, NULL, info->carrier_size);
+    unscanned_size = bld_unstable_uint(&hp, NULL, info->unscanned_size);
+    allocated_size = bld_unstable_uint(&hp, NULL, info->allocated_size);
+
+    hp[0] = make_arityval(7);
+    hp[1] = state->allocator_desc;
+    hp[2] = carrier_size;
+    hp[3] = unscanned_size;
+    hp[4] = allocated_size;
+    hp[5] = make_small(info->allocated_count);
+    hp[6] = (info->flags & ERTS_CRR_ALCTR_FLG_IN_POOL) ? am_true : am_false;
+    hp[7] = histogram_tuple;
+
+    carrier_tuple = make_tuple(hp);
+    hp += 8;
+
+    state->result_list = CONS(hp, carrier_tuple, state->result_list);
+
+    free(info);
+}
+
+static void gather_cinfo_send(gather_cinfo_t *state)
+{
+    Eterm result_tuple, task_ref;
+
+    int term_size;
+    Eterm *hp;
+
+    ASSERT((state->result_list == NIL) ^ (state->info_count > 0));
+    ASSERT(state->building_result);
+
+    term_size = 3 + erts_iref_storage_heap_size(&state->iref);
+    hp = erts_produce_heap(&state->msg_factory, term_size, 0);
+
+    task_ref = erts_iref_storage_make_ref(&state->iref, &hp,
+        &(state->msg_factory.message)->hfrag.off_heap, 0);
+
+    hp[0] = make_arityval(2);
+    hp[1] = task_ref;
+    hp[2] = state->result_list;
+
+    result_tuple = make_tuple(hp);
+
+    erts_factory_trim_and_close(&state->msg_factory, &result_tuple, 1);
+
+    erts_queue_message(state->process, 0, state->msg_factory.message,
+                       result_tuple, am_system);
+}
+
+static int gather_cinfo_finish(void *arg)
+{
+    gather_cinfo_t *state = (gather_cinfo_t*)arg;
+    int reductions = BLOCKSCAN_REDUCTIONS;
+
+    if (!state->building_result) {
+        ErtsMessage *message;
+        Uint minimum_size;
+        Eterm *hp;
+
+        /* {Ref, [{Carrier size, unscanned size, allocated size,
+         *         allocated block count, {Free block histogram}} | Rest]} */
+        minimum_size = 3 + erts_iref_storage_heap_size(&state->iref) +
+                       state->info_count * (11 + state->hist_slot_count);
+
+        message = erts_alloc_message(minimum_size, &hp);
+        erts_factory_selfcontained_message_init(&state->msg_factory,
+                                                message, hp);
+
+        state->result_list = NIL;
+        state->building_result = 1;
+    }
+
+    while (state->info_list) {
+        chist_node_t *current = state->info_list;
+        state->info_list = current->next;
+
+        gather_cinfo_append_result(state, current);
+
+        if (reductions-- <= 0) {
+            return 1;
+        }
+    }
+
+    gather_cinfo_send(state);
+
+    return 0;
+}
+
+static void gather_cinfo_abort(void *arg)
+{
+    gather_cinfo_t *state = (gather_cinfo_t*)arg;
+
+    if (state->building_result) {
+        erts_factory_undo(&state->msg_factory);
+    }
+
+    while (state->info_list) {
+        chist_node_t *current = state->info_list;
+        state->info_list = current->next;
+
+        free(current);
+    }
+}
+
+int erts_alcu_gather_carrier_info(struct process *p, int allocator_num,
+                                  int sched_id, int hist_width,
+                                  UWord hist_start, Eterm ref)
+{
+    gather_cinfo_t *gather_state;
+    blockscan_t *scanner;
+
+    const char *allocator_desc;
+    Allctr_t *allocator;
+
+    ASSERT(is_internal_ref(ref));
+
+    if (!blockscan_get_specific_allocator(allocator_num,
+                                          sched_id,
+                                          &allocator)) {
+        return 0;
+    }
+
+    allocator_desc = ERTS_ALC_A2AD(allocator_num);
+
+    /* Plain calloc is intentional. */
+    gather_state = (gather_cinfo_t*)calloc(1, sizeof(gather_cinfo_t));
+    scanner = &gather_state->common;
+
+    scanner->abort = gather_cinfo_abort;
+    scanner->scan = gather_cinfo_scan;
+    scanner->finish = gather_cinfo_finish;
+    scanner->user_data = gather_state;
+
+    gather_state->allocator_desc = erts_atom_put((byte *)allocator_desc,
+                                                 sys_strlen(allocator_desc),
+                                                 ERTS_ATOM_ENC_LATIN1, 1);
+    erts_iref_storage_save(&gather_state->iref, ref);
+    gather_state->hist_slot_start = hist_start * 2;
+    gather_state->hist_slot_count = hist_width;
+    gather_state->process = p;
+
+    blockscan_dispatch(scanner, p, allocator, sched_id);
+
+    return 1;
+}
+
 
 /* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
  * NOTE:  erts_alcu_test() is only supposed to be used for testing.          *
@@ -6530,7 +7647,6 @@ erts_alcu_test(UWord op, UWord a1, UWord a2)
     case 0x01c: return (UWord) BLK_TO_MBC((Block_t*) a1);
     case 0x01d: ((Allctr_t*) a1)->add_mbc((Allctr_t*)a1, (Carrier_t*)a2); break;
     case 0x01e: ((Allctr_t*) a1)->remove_mbc((Allctr_t*)a1, (Carrier_t*)a2); break;
-#ifdef ERTS_SMP
     case 0x01f: return (UWord) sizeof(ErtsAlcCrrPool_t);
     case 0x020:
 	SET_CARRIER_HDR((Carrier_t *) a2, 0, SCH_SYS_ALLOC|SCH_MBC, (Allctr_t *) a1);
@@ -6544,14 +7660,6 @@ erts_alcu_test(UWord op, UWord a1, UWord a2)
 	return (UWord) a2;
     case 0x023: return (UWord) cpool_is_empty((Allctr_t *) a1);
     case 0x024: return (UWord) cpool_dbg_is_in_pool((Allctr_t *) a1, (Carrier_t *) a2);
-#else
-    case 0x01f: return (UWord) 0;
-    case 0x020: return (UWord) 0;
-    case 0x021: return (UWord) 0;
-    case 0x022: return (UWord) 0;
-    case 0x023: return (UWord) 0;
-    case 0x024: return (UWord) 0;
-#endif
     case 0x025: /* UMEM2BLK_TEST*/
 #ifdef DEBUG
 # ifdef HARD_DEBUG
@@ -6609,15 +7717,12 @@ erts_alcu_verify_unused(Allctr_t *allctr)
 void
 erts_alcu_verify_unused_ts(Allctr_t *allctr)
 {
-#ifdef USE_THREADS
     erts_mtx_lock(&allctr->mutex);
-#endif
     erts_alcu_verify_unused(allctr);
-#ifdef USE_THREADS
     erts_mtx_unlock(&allctr->mutex);
-#endif
 }
 
+
 #ifdef DEBUG
 int is_sbc_blk(Block_t* blk)
 {
diff --git a/erts/emulator/beam/erl_alloc_util.h b/erts/emulator/beam/erl_alloc_util.h
index 1c07086040..cbae8ce98a 100644
--- a/erts/emulator/beam/erl_alloc_util.h
+++ b/erts/emulator/beam/erl_alloc_util.h
@@ -24,10 +24,8 @@
 #define ERTS_ALCU_VSN_STR "3.0"
 
 #include "erl_alloc_types.h"
-#ifdef USE_THREADS
 #define ERL_THREADS_EMU_INTERNAL__
 #include "erl_threads.h"
-#endif
 
 #include "erl_mseg.h"
 #include "lttng-wrapper.h"
@@ -52,6 +50,7 @@ typedef struct {
     int tspec;
     int tpref;
     int ramv;
+    int atags;
     UWord sbct;
     UWord asbcst;
     UWord rsbcst;
@@ -108,6 +107,7 @@ typedef struct {
     0,			/* (bool)   tspec:  thread specific              */\
     0,			/* (bool)   tpref:  thread preferred             */\
     0,			/* (bool)   ramv:   realloc always moves         */\
+    0,			/* (bool)   atags:  tagged allocations           */\
     512*1024,		/* (bytes)  sbct:   sbc threshold                */\
     2*1024*2024,	/* (amount) asbcst: abs sbc shrink threshold     */\
     20,			/* (%)      rsbcst: rel sbc shrink threshold     */\
@@ -144,6 +144,7 @@ typedef struct {
     0,			/* (bool)   tspec:  thread specific              */\
     0,			/* (bool)   tpref:  thread preferred             */\
     0,			/* (bool)   ramv:   realloc always moves         */\
+    0,			/* (bool)   atags:  tagged allocations           */\
     64*1024,		/* (bytes)  sbct:   sbc threshold                */\
     2*1024*2024,	/* (amount) asbcst: abs sbc shrink threshold     */\
     20,			/* (%)      rsbcst: rel sbc shrink threshold     */\
@@ -168,12 +169,10 @@ void *	erts_alcu_alloc(ErtsAlcType_t, void *, Uint);
 void *	erts_alcu_realloc(ErtsAlcType_t, void *, void *, Uint);
 void *	erts_alcu_realloc_mv(ErtsAlcType_t, void *, void *, Uint);
 void	erts_alcu_free(ErtsAlcType_t, void *, void *);
-#ifdef USE_THREADS
 void *	erts_alcu_alloc_ts(ErtsAlcType_t, void *, Uint);
 void *	erts_alcu_realloc_ts(ErtsAlcType_t, void *, void *, Uint);
 void *	erts_alcu_realloc_mv_ts(ErtsAlcType_t, void *, void *, Uint);
 void	erts_alcu_free_ts(ErtsAlcType_t, void *, void *);
-#ifdef ERTS_SMP
 void *	erts_alcu_alloc_thr_spec(ErtsAlcType_t, void *, Uint);
 void *	erts_alcu_realloc_thr_spec(ErtsAlcType_t, void *, void *, Uint);
 void *	erts_alcu_realloc_mv_thr_spec(ErtsAlcType_t, void *, void *, Uint);
@@ -182,8 +181,6 @@ void *	erts_alcu_alloc_thr_pref(ErtsAlcType_t, void *, Uint);
 void *	erts_alcu_realloc_thr_pref(ErtsAlcType_t, void *, void *, Uint);
 void *	erts_alcu_realloc_mv_thr_pref(ErtsAlcType_t, void *, void *, Uint);
 void	erts_alcu_free_thr_pref(ErtsAlcType_t, void *, void *);
-#endif
-#endif
 Eterm	erts_alcu_au_info_options(fmtfn_t *, void *, Uint **, Uint *);
 Eterm	erts_alcu_info_options(Allctr_t *, fmtfn_t *, void *, Uint **, Uint *);
 Eterm	erts_alcu_sz_info(Allctr_t *, int, int, fmtfn_t *, void *, Uint **, Uint *);
@@ -191,9 +188,7 @@ Eterm	erts_alcu_info(Allctr_t *, int, int, fmtfn_t *, void *, Uint **, Uint *);
 void	erts_alcu_init(AlcUInit_t *);
 void    erts_alcu_current_size(Allctr_t *, AllctrSize_t *,
 			       ErtsAlcUFixInfo_t *, int);
-#ifdef ERTS_SMP
 void    erts_alcu_check_delayed_dealloc(Allctr_t *, int, int *, ErtsThrPrgrVal *, int *);
-#endif
 erts_aint32_t erts_alcu_fix_alloc_shrink(Allctr_t *, erts_aint32_t);
 
 #ifdef ARCH_32
@@ -213,7 +208,7 @@ void* erts_alcu_mmapper_mseg_realloc(Allctr_t*, void *seg, Uint old_size, Uint *
 void  erts_alcu_mmapper_mseg_dealloc(Allctr_t*, void *seg, Uint size, Uint flags);
 # endif
 
-# if defined(ERTS_ALC_A_EXEC) && !defined(ERTS_HAVE_EXEC_MMAPPER)
+# if defined(ERTS_ALC_A_EXEC)
 void* erts_alcu_exec_mseg_alloc(Allctr_t*, Uint *size_p, Uint flags);
 void* erts_alcu_exec_mseg_realloc(Allctr_t*, void *seg, Uint old_size, Uint *new_size_p);
 void  erts_alcu_exec_mseg_dealloc(Allctr_t*, void *seg, Uint size, Uint flags);
@@ -232,6 +227,36 @@ void erts_lcnt_update_allocator_locks(int enable);
 
 int erts_alcu_try_set_dyn_param(Allctr_t*, Eterm param, Uint value);
 
+/* Gathers per-tag allocation histograms from the given allocator number
+ * (ERTS_ALC_A_*) and scheduler id. An id of 0 means the global instance will
+ * be used.
+ *
+ * The results are sent to `p`, and it returns the number of messages to wait
+ * for. */
+int erts_alcu_gather_alloc_histograms(struct process *p, int allocator_num,
+                                      int sched_id, int hist_width,
+                                      UWord hist_start, Eterm ref);
+
+/* Gathers per-carrier info from the given allocator number (ERTS_ALC_A_*) and
+ * scheduler id. An id of 0 means the global instance will be used.
+ *
+ * The results are sent to `p`, and it returns the number of messages to wait
+ * for. */
+int erts_alcu_gather_carrier_info(struct process *p, int allocator_num,
+                                  int sched_id, int hist_width,
+                                  UWord hist_start, Eterm ref);
+
+struct alcu_blockscan;
+
+typedef struct {
+    struct alcu_blockscan *current;
+    struct alcu_blockscan *last;
+} ErtsAlcuBlockscanYieldData;
+
+int erts_handle_yielded_alcu_blockscan(struct ErtsSchedulerData_ *esdp,
+                                       ErtsAlcuBlockscanYieldData *yield);
+void erts_alcu_sched_spec_data_init(struct ErtsSchedulerData_ *esdp);
+
 #endif /* !ERL_ALLOC_UTIL__ */
 
 #if defined(GET_ERL_ALLOC_UTIL_IMPL) && !defined(ERL_ALLOC_UTIL_IMPL__)
@@ -314,7 +339,6 @@ typedef union {char c[ERTS_ALLOC_ALIGN_BYTES]; long l; double d;} Unit_t;
 
 typedef struct Carrier_t_ Carrier_t;
 
-
 typedef struct {
     UWord bhdr;
 #if !MBC_ABLK_OFFSET_BITS
@@ -390,15 +414,16 @@ struct AOFF_RBTree_t_ {
     AOFF_RBTree_t *right;
     Uint32 flags;
     Uint32 max_sz;  /* of all blocks in this sub-tree */
+    union {
+        AOFF_RBTree_t* next;   /* for best fit */
+        Sint64 birth_time;     /* for age first fit */
+    } u;
 };
-#ifdef ERTS_SMP
+
 void aoff_add_pooled_mbc(Allctr_t*, Carrier_t*);
 void aoff_remove_pooled_mbc(Allctr_t*, Carrier_t*);
 Carrier_t* aoff_lookup_pooled_mbc(Allctr_t*, Uint size);
 void erts_aoff_larger_max_size(AOFF_RBTree_t *node);
-#endif
-
-#ifdef ERTS_SMP
 
 typedef struct {
     ErtsFakeDDBlock_t homecoming_dd;
@@ -418,20 +443,16 @@ typedef struct {
     AOFF_RBTree_t pooled;  /* node in pooled_tree */
 } ErtsAlcCPoolData_t;
 
-#endif
-
 struct Carrier_t_ {
     UWord chdr;
     Carrier_t *next;
     Carrier_t *prev;
-    erts_smp_atomic_t allctr;
-#ifdef ERTS_SMP
+    erts_atomic_t allctr;
     ErtsAlcCPoolData_t cpool; /* Overwritten by block if sbc */
-#endif
 };
 
 #define ERTS_ALC_CARRIER_TO_ALLCTR(C) \
-  ((Allctr_t *) (erts_smp_atomic_read_nob(&(C)->allctr) & ~FLG_MASK))
+  ((Allctr_t *) (erts_atomic_read_nob(&(C)->allctr) & ~FLG_MASK))
 
 typedef struct {
     Carrier_t *first;
@@ -474,7 +495,6 @@ typedef struct {
     } while (0)
 #endif
 
-#ifdef ERTS_SMP
 
 typedef struct {
     ErtsAllctrDDBlock_t marker;
@@ -510,7 +530,6 @@ typedef struct {
     } head;
 } ErtsAllctrDDQueue_t;
 
-#endif
 
 typedef struct {
     size_t type_size;
@@ -533,7 +552,6 @@ typedef struct {
 } ErtsAlcFixList_t;
 
 struct Allctr_t_ {
-#ifdef ERTS_SMP
     struct {
 	/*
 	 * We want the queue at the beginning of
@@ -544,7 +562,6 @@ struct Allctr_t_ {
 	int		use;
 	int		ix;
     } dd;
-#endif
 
     /* Allocator name prefix */
     char *		name_prefix;
@@ -568,6 +585,7 @@ struct Allctr_t_ {
     /* Options */
     int			t;
     int			ramv;
+    int                 atags;
     Uint		sbc_threshold;
     Uint		sbc_move_threshold;
     Uint		mbc_move_threshold;
@@ -593,7 +611,6 @@ struct Allctr_t_ {
     /* Carriers *employed* by this allocator */
     CarrierList_t	mbc_list;
     CarrierList_t	sbc_list;
-#ifdef ERTS_SMP
     struct {
 	/* pooled_tree and dc_list contain only
            carriers *created* by this allocator */
@@ -624,7 +641,6 @@ struct Allctr_t_ {
 	    CallCounter_t       entrance_removed;
 	} stat;
     } cpool;
-#endif
 
     /* Main carrier (if there is one) */
     Carrier_t *		main_carrier;
@@ -669,7 +685,6 @@ struct Allctr_t_ {
     int			fix_shrink_scheduled;
     ErtsAlcFixList_t	*fix;
 
-#ifdef USE_THREADS
     /* Mutex for this allocator */
     erts_mtx_t		mutex;
     int			thread_safe;
@@ -678,7 +693,6 @@ struct Allctr_t_ {
 	Allctr_t	*next;
     } ts_list;
 
-#endif
 
     int			atoms_initialized;
 
@@ -701,15 +715,14 @@ struct Allctr_t_ {
     CarriersStats_t	mbcs;
     
 #ifdef DEBUG
-#ifdef USE_THREADS
     struct {
 	int saved_tid;
 	erts_tid_t tid;
     } debug;
 #endif
-#endif
 };
 
+
 int	erts_alcu_start(Allctr_t *, AllctrInit_t *);
 void	erts_alcu_stop(Allctr_t *);
 
diff --git a/erts/emulator/beam/erl_ao_firstfit_alloc.c b/erts/emulator/beam/erl_ao_firstfit_alloc.c
index 73576c0189..ebbe4af53d 100644
--- a/erts/emulator/beam/erl_ao_firstfit_alloc.c
+++ b/erts/emulator/beam/erl_ao_firstfit_alloc.c
@@ -100,22 +100,14 @@
 
 #define AOFF_BLK_SZ(B) MBC_FBLK_SZ(&(B)->hdr)
 
-/* BF block nodes keeps list of all with equal size
- */
-typedef struct {
-    AOFF_RBTree_t t;
-    AOFF_RBTree_t *next;
-}AOFF_RBTreeList_t;
-
-#define LIST_NEXT(N) (((AOFF_RBTreeList_t*) (N))->next)
-#define LIST_PREV(N) (((AOFF_RBTreeList_t*) (N))->t.parent)
+#define LIST_NEXT(N) (((AOFF_RBTree_t*)(N))->u.next)
+#define LIST_PREV(N) (((AOFF_RBTree_t*)(N))->parent)
 
 typedef struct AOFF_Carrier_t_ AOFF_Carrier_t;
 
 struct AOFF_Carrier_t_ {
     Carrier_t crr;
     AOFF_RBTree_t rbt_node;     /* My node in the carrier tree */
-    Sint64 birth_time;
     AOFF_RBTree_t* root;        /* Root of my block tree */
 };
 #define RBT_NODE_TO_MBC(PTR) ErtsContainerStruct((PTR), AOFF_Carrier_t, rbt_node)
@@ -174,7 +166,6 @@ static ERTS_INLINE void lower_max_size(AOFF_RBTree_t *node,
     else ASSERT(new_max == old_max);
 }
 
-#ifdef ERTS_SMP
 /*
  * Set possibly new larger 'max_sz' of node and propagate change toward root
  */
@@ -193,7 +184,6 @@ void erts_aoff_larger_max_size(AOFF_RBTree_t *node)
             break;
     }
 }
-#endif
 
 /* Compare nodes for both carrier and block trees */
 static ERTS_INLINE SWord cmp_blocks(enum AOFFSortOrder order,
@@ -201,9 +191,7 @@ static ERTS_INLINE SWord cmp_blocks(enum AOFFSortOrder order,
 {
     ASSERT(lhs != rhs);
     if (order == FF_AGEFF) {
-	AOFF_Carrier_t* lc = RBT_NODE_TO_MBC(lhs);
-	AOFF_Carrier_t* rc = RBT_NODE_TO_MBC(rhs);
-	Sint64 diff = lc->birth_time - rc->birth_time;
+	Sint64 diff = lhs->u.birth_time - rhs->u.birth_time;
  #ifdef ARCH_64
         if (diff)
             return diff;
@@ -298,8 +286,10 @@ erts_aoffalc_start(AOFFAllctr_t *alc,
     allctr->mbc_header_size		= sizeof(AOFF_Carrier_t);
     allctr->min_mbc_size		= MIN_MBC_SZ;
     allctr->min_mbc_first_free_size	= MIN_MBC_FIRST_FREE_SZ;
-    allctr->min_block_size = (aoffinit->blk_order == FF_BF ?
-			      sizeof(AOFF_RBTreeList_t):sizeof(AOFF_RBTree_t));
+    allctr->min_block_size = (aoffinit->blk_order == FF_BF
+                              ? (offsetof(AOFF_RBTree_t, u.next)
+                                 + ErtsSizeofMember(AOFF_RBTree_t, u.next))
+                              : offsetof(AOFF_RBTree_t, u));
 
     allctr->vsn_str			= ERTS_ALC_AOFF_ALLOC_VSN_STR;
 
@@ -867,7 +857,6 @@ rbt_search(AOFF_RBTree_t* root, Uint size)
     }
 }
 
-#ifdef ERTS_SMP
 Carrier_t* aoff_lookup_pooled_mbc(Allctr_t* allctr, Uint size)
 {
     AOFF_RBTree_t* node;
@@ -877,7 +866,6 @@ Carrier_t* aoff_lookup_pooled_mbc(Allctr_t* allctr, Uint size)
     node = rbt_search(allctr->cpool.pooled_tree, size);
     return node ? ErtsContainerStruct(node, Carrier_t, cpool.pooled) : NULL;
 }
-#endif
 
 static Block_t *
 aoff_get_free_block(Allctr_t *allctr, Uint size,
@@ -943,8 +931,11 @@ static void aoff_creating_mbc(Allctr_t *allctr, Carrier_t *carrier)
     HARD_CHECK_TREE(NULL, alc->crr_order, *root, 0);
 
     crr->rbt_node.hdr.bhdr = 0;
-    if (alc->crr_order == FF_AGEFF || IS_DEBUG)
-	crr->birth_time = get_birth_time();
+    if (alc->crr_order == FF_AGEFF || IS_DEBUG) {
+        Sint64 bt = get_birth_time();
+        crr->rbt_node.u.birth_time = bt;
+        crr->crr.cpool.pooled.u.birth_time = bt;
+    }
     rbt_insert(alc->crr_order, root, &crr->rbt_node);
 
     /* aoff_link_free_block will add free block later */
@@ -980,9 +971,9 @@ static void aoff_add_mbc(Allctr_t *allctr, Carrier_t *carrier)
     HARD_CHECK_TREE(NULL, alc->crr_order, *root, 0);
 }
 
-#ifdef ERTS_SMP
 void aoff_add_pooled_mbc(Allctr_t *allctr, Carrier_t *crr)
 {
+    AOFFAllctr_t *alc = (AOFFAllctr_t *) allctr;
     AOFF_RBTree_t **root = &allctr->cpool.pooled_tree;
 
     ASSERT(allctr == crr->cpool.orig_allctr);
@@ -990,11 +981,10 @@ void aoff_add_pooled_mbc(Allctr_t *allctr, Carrier_t *crr)
 
     /* Link carrier in address order tree
      */
-    rbt_insert(FF_AOFF, root, &crr->cpool.pooled);
+    rbt_insert(alc->crr_order, root, &crr->cpool.pooled);
 
     HARD_CHECK_TREE(NULL, 0, *root, 0);
 }
-#endif
 
 static void aoff_remove_mbc(Allctr_t *allctr, Carrier_t *carrier)
 {
@@ -1017,7 +1007,6 @@ static void aoff_remove_mbc(Allctr_t *allctr, Carrier_t *carrier)
     HARD_CHECK_TREE(NULL, alc->crr_order, *root, 0);
 }
 
-#ifdef ERTS_SMP
 void aoff_remove_pooled_mbc(Allctr_t *allctr, Carrier_t *crr)
 {
     ASSERT(allctr == crr->cpool.orig_allctr);
@@ -1034,7 +1023,6 @@ void aoff_remove_pooled_mbc(Allctr_t *allctr, Carrier_t *crr)
     HARD_CHECK_TREE(NULL, 0, allctr->cpool.pooled_tree, 0);
 
 }
-#endif
 
 
 static UWord aoff_largest_fblk_in_mbc(Allctr_t* allctr, Carrier_t* carrier)
@@ -1062,7 +1050,7 @@ static struct {
 
 static void ERTS_INLINE atom_init(Eterm *atom, const char *name)
 {
-    *atom = am_atom_put(name, strlen(name));
+    *atom = am_atom_put(name, sys_strlen(name));
 }
 #define AM_INIT(AM) atom_init(&am.AM, #AM)
 
diff --git a/erts/emulator/beam/erl_ao_firstfit_alloc.h b/erts/emulator/beam/erl_ao_firstfit_alloc.h
index 9cf4fc81a8..dad864801f 100644
--- a/erts/emulator/beam/erl_ao_firstfit_alloc.h
+++ b/erts/emulator/beam/erl_ao_firstfit_alloc.h
@@ -29,10 +29,10 @@
 typedef struct AOFFAllctr_t_ AOFFAllctr_t;
 
 enum AOFFSortOrder {
-    FF_AGEFF = 0,
+    FF_AGEFF = 0,    /* carrier trees only */
     FF_AOFF  = 1,
-    FF_AOBF  = 2,
-    FF_BF    = 3
+    FF_AOBF  = 2,    /* block trees only */
+    FF_BF    = 3     /* block trees only */
 };
 
 typedef struct {
diff --git a/erts/emulator/beam/erl_arith.c b/erts/emulator/beam/erl_arith.c
index 861532f241..b6625db0d3 100644
--- a/erts/emulator/beam/erl_arith.c
+++ b/erts/emulator/beam/erl_arith.c
@@ -114,7 +114,7 @@ BIF_RETTYPE intdiv_2(BIF_ALIST_2)
     }
     if (is_both_small(BIF_ARG_1,BIF_ARG_2)){
 	Sint ires = signed_val(BIF_ARG_1) / signed_val(BIF_ARG_2);
-	if (MY_IS_SSMALL(ires))
+	if (IS_SSMALL(ires))
 	    BIF_RET(make_small(ires));
     } 
     BIF_RET(erts_int_div(BIF_P, BIF_ARG_1, BIF_ARG_2));
@@ -276,8 +276,12 @@ shift(Process* p, Eterm arg1, Eterm arg2, int right)
 		goto do_bsl;
 	    } else if (is_small(arg1) || is_big(arg1)) {
 		/*
-		 * N bsl PositiveBigNum is too large to represent.
+		 * N bsl PositiveBigNum is too large to represent,
+                 * unless N is 0.
 		 */
+                if (arg1 == make_small(0)) {
+                    BIF_RET(arg1);
+                }
 		BIF_ERROR(p, SYSTEM_LIMIT);
 	    }
 	     /* Fall through if the left argument is not an integer. */
@@ -336,8 +340,7 @@ erts_mixed_plus(Process* p, Eterm arg1, Eterm arg2)
 		switch ((arg2 & _TAG_IMMED1_MASK) >> _TAG_PRIMARY_SIZE) {
 		case (_TAG_IMMED1_SMALL >> _TAG_PRIMARY_SIZE):
 		    ires = signed_val(arg1) + signed_val(arg2);
-		    ASSERT(MY_IS_SSMALL(ires) == IS_SSMALL(ires));
-		    if (MY_IS_SSMALL(ires)) {
+		    if (IS_SSMALL(ires)) {
 			return make_small(ires);
 		    } else {
 			hp = HAlloc(p, 2);
@@ -482,8 +485,7 @@ erts_mixed_minus(Process* p, Eterm arg1, Eterm arg2)
 		switch ((arg2 & _TAG_IMMED1_MASK) >> _TAG_PRIMARY_SIZE) {
 		case (_TAG_IMMED1_SMALL >> _TAG_PRIMARY_SIZE):
 		    ires = signed_val(arg1) - signed_val(arg2);
-		    ASSERT(MY_IS_SSMALL(ires) == IS_SSMALL(ires));
-		    if (MY_IS_SSMALL(ires)) {
+		    if (IS_SSMALL(ires)) {
 			return make_small(ires);
 		    } else {
 			hp = HAlloc(p, 2);
@@ -1177,8 +1179,7 @@ erts_gc_mixed_plus(Process* p, Eterm* reg, Uint live)
 		switch ((arg2 & _TAG_IMMED1_MASK) >> _TAG_PRIMARY_SIZE) {
 		case (_TAG_IMMED1_SMALL >> _TAG_PRIMARY_SIZE):
 		    ires = signed_val(arg1) + signed_val(arg2);
-		    ASSERT(MY_IS_SSMALL(ires) == IS_SSMALL(ires));
-		    if (MY_IS_SSMALL(ires)) {
+		    if (IS_SSMALL(ires)) {
 			return make_small(ires);
 		    } else {
 			if (ERTS_NEED_GC(p, 2)) {
@@ -1345,8 +1346,7 @@ erts_gc_mixed_minus(Process* p, Eterm* reg, Uint live)
 		switch ((arg2 & _TAG_IMMED1_MASK) >> _TAG_PRIMARY_SIZE) {
 		case (_TAG_IMMED1_SMALL >> _TAG_PRIMARY_SIZE):
 		    ires = signed_val(arg1) - signed_val(arg2);
-		    ASSERT(MY_IS_SSMALL(ires) == IS_SSMALL(ires));
-		    if (MY_IS_SSMALL(ires)) {
+		    if (IS_SSMALL(ires)) {
 			return make_small(ires);
 		    } else {
 			if (ERTS_NEED_GC(p, 2)) {
diff --git a/erts/emulator/beam/erl_async.c b/erts/emulator/beam/erl_async.c
index 9a93034fcb..3ceb2fd368 100644
--- a/erts/emulator/beam/erl_async.c
+++ b/erts/emulator/beam/erl_async.c
@@ -34,9 +34,6 @@
 
 #define ERTS_ASYNC_PRINT_JOB 0
 
-#if !defined(ERTS_SMP) && defined(USE_THREADS) && !ERTS_USE_ASYNC_READY_Q
-#  error "Need async ready queue in non-smp case"
-#endif
 
 typedef struct _erl_async {
     DE_Handle*         hndl;   /* The DE_Handle is needed when port is gone */
@@ -46,16 +43,13 @@ typedef struct _erl_async {
     ErlDrvPDL          pdl;
     void (*async_invoke)(void*);
     void (*async_free)(void*);
-#if ERTS_USE_ASYNC_READY_Q
     Uint               sched_id;
     union {
 	ErtsThrQPrepEnQ_t *prep_enq;
 	ErtsThrQFinDeQ_t   fin_deq;
     } q;
-#endif
 } ErtsAsync;
 
-#if ERTS_USE_ASYNC_READY_Q
 
 /*
  * We can do without the enqueue mutex since it isn't needed for
@@ -94,7 +88,6 @@ typedef union {
     char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsAsyncReadyQ))];
 } ErtsAlgndAsyncReadyQ;
 
-#endif /* ERTS_USE_ASYNC_READY_Q */
 
 typedef struct {
     ErtsThrQ_t thr_q;
@@ -119,12 +112,10 @@ typedef struct {
 	char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsAsyncInit))];
     } init;
     ErtsAlgndAsyncQ *queue;
-#if ERTS_USE_ASYNC_READY_Q
     ErtsAlgndAsyncReadyQ *ready_queue;
-#endif
 } ErtsAsyncData;
 
-#if defined(USE_THREADS) && defined(USE_VM_PROBES)
+#if defined(USE_VM_PROBES)
 
 /*
  * Some compilers, e.g. GCC 4.2.1 and -O3, will optimize away DTrace
@@ -140,15 +131,6 @@ int erts_async_thread_suggested_stack_size; /* Initialized by erl_init.c */
 
 static ErtsAsyncData *async;
 
-#ifndef USE_THREADS
-
-void
-erts_init_async(void)
-{
-
-}
-
-#else
 
 static void *async_main(void *);
 
@@ -158,7 +140,6 @@ async_q(int i)
     return &async->queue[i].aq;
 }
 
-#if ERTS_USE_ASYNC_READY_Q
 
 static ERTS_INLINE ErtsAsyncReadyQ *
 async_ready_q(Uint sched_id)
@@ -166,16 +147,13 @@ async_ready_q(Uint sched_id)
     return &async->ready_queue[((int)sched_id)-1].arq;
 }
 
-#endif
 
 void
 erts_init_async(void)
 {
     async = NULL;
     if (erts_async_max_threads > 0) {
-#if ERTS_USE_ASYNC_READY_Q
 	ErtsThrQInit_t qinit = ERTS_THR_Q_INIT_DEFAULT;
-#endif
 	erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER;
 	char *ptr, thr_name[16];
 	size_t tot_size = 0;
@@ -183,9 +161,7 @@ erts_init_async(void)
 
 	tot_size += ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsAsyncData));
 	tot_size += sizeof(ErtsAlgndAsyncQ)*erts_async_max_threads;
-#if ERTS_USE_ASYNC_READY_Q
 	tot_size += sizeof(ErtsAlgndAsyncReadyQ)*erts_no_schedulers;
-#endif
 
 	ptr = erts_alloc_permanent_cache_aligned(ERTS_ALC_T_ASYNC_DATA,
 						 tot_size);
@@ -202,7 +178,6 @@ erts_init_async(void)
 	async->queue = (ErtsAlgndAsyncQ *) ptr;
 	ptr += sizeof(ErtsAlgndAsyncQ)*erts_async_max_threads;
 
-#if ERTS_USE_ASYNC_READY_Q
 
 	qinit.live.queue = ERTS_THR_Q_LIVE_LONG;
 	qinit.live.objects = ERTS_THR_Q_LIVE_SHORT;
@@ -222,7 +197,6 @@ erts_init_async(void)
 	    erts_thr_q_initialize(&arq->thr_q, &qinit);
 	}
 
-#endif
 
 	/* Create async threads... */
 
@@ -253,7 +227,6 @@ erts_init_async(void)
     }
 }
 
-#if ERTS_USE_ASYNC_READY_Q
 
 void *
 erts_get_async_ready_queue(Uint sched_id)
@@ -261,7 +234,6 @@ erts_get_async_ready_queue(Uint sched_id)
     return (void *) async ? async_ready_q(sched_id) : NULL;
 }
 
-#endif
 
 static ERTS_INLINE void async_add(ErtsAsync *a, ErtsAsyncQ* q)
 {
@@ -270,10 +242,8 @@ static ERTS_INLINE void async_add(ErtsAsync *a, ErtsAsyncQ* q)
 #endif
 
     if (is_internal_port(a->port)) {
-#if ERTS_USE_ASYNC_READY_Q
 	ErtsAsyncReadyQ *arq = async_ready_q(a->sched_id);
 	a->q.prep_enq = erts_thr_q_prepare_enqueue(&arq->thr_q);
-#endif
 	/* make sure the driver will stay around */
 	if (a->hndl)
 	    erts_ddll_reference_referenced_driver(a->hndl);
@@ -309,10 +279,8 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 					erts_tse_t *tse,
 					ErtsThrQPrepEnQ_t **prep_enq)
 {
-#if ERTS_USE_ASYNC_READY_Q
     int saved_fin_deq = 0;
     ErtsThrQFinDeQ_t fin_deq;
-#endif
 #ifdef USE_VM_PROBES
     int len;
 #endif
@@ -321,12 +289,10 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 	ErtsAsync *a = (ErtsAsync *) erts_thr_q_dequeue(q);
 	if (a) {
 
-#if ERTS_USE_ASYNC_READY_Q
 	    *prep_enq = a->q.prep_enq;
 	    erts_thr_q_get_finalize_dequeue_data(q, &a->q.fin_deq);
 	    if (saved_fin_deq)
 		erts_thr_q_append_finalize_dequeue_data(&a->q.fin_deq, &fin_deq);
-#endif
 #ifdef USE_LTTNG_VM_TRACEPOINTS
             if (LTTNG_ENABLED(aio_pool_get)) {
                 lttng_decl_portbuf(port_str);
@@ -354,7 +320,6 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 
 	    erts_tse_reset(tse);
 
-#if ERTS_USE_ASYNC_READY_Q
 	chk_fin_deq:
 	    if (erts_thr_q_get_finalize_dequeue_data(q, &tmp_fin_deq)) {
 		if (!saved_fin_deq) {
@@ -364,13 +329,11 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 		erts_thr_q_append_finalize_dequeue_data(&fin_deq,
 							&tmp_fin_deq);
 	    }
-#endif
 
 	    switch (erts_thr_q_inspect(q, 1)) {
 	    case ERTS_THR_Q_DIRTY:
 		break;
 	    case ERTS_THR_Q_NEED_THR_PRGR:
-#ifdef ERTS_SMP
 	    {
 		ErtsThrPrgrVal prgr = erts_thr_q_need_thr_progress(q);
 		erts_thr_progress_wakeup(NULL, prgr);
@@ -382,17 +345,14 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 		erts_tse_wait(tse);
 		break;
 	    }
-#endif
 	    case ERTS_THR_Q_CLEAN:
 
-#if ERTS_USE_ASYNC_READY_Q
 		if (saved_fin_deq) {
 		    if (erts_thr_q_finalize_dequeue(&fin_deq))
 			goto chk_fin_deq;
 		    else
 			saved_fin_deq = 0;
 		}
-#endif
 
 		erts_tse_wait(tse);
 		break;
@@ -408,15 +368,10 @@ static ERTS_INLINE ErtsAsync *async_get(ErtsThrQ_t *q,
 
 static ERTS_INLINE void call_async_ready(ErtsAsync *a)
 {
-#if ERTS_USE_ASYNC_READY_Q
     Port *p = erts_id2port_sflgs(a->port,
 				 NULL,
 				 0,
 				 ERTS_PORT_SFLGS_INVALID_DRIVER_LOOKUP);
-#else
-    Port *p = erts_thr_id2port_sflgs(a->port,
-				     ERTS_PORT_SFLGS_INVALID_DRIVER_LOOKUP);
-#endif
     if (!p) {
 	if (a->async_free) {
             ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_PORT);
@@ -432,11 +387,7 @@ static ERTS_INLINE void call_async_ready(ErtsAsync *a)
                 ERTS_MSACC_POP_STATE();
             }
 	}
-#if ERTS_USE_ASYNC_READY_Q
 	erts_port_release(p);
-#else
-	erts_thr_port_release(p);
-#endif
     }
     if (a->pdl)
 	driver_pdl_dec_refc(a->pdl);
@@ -446,7 +397,6 @@ static ERTS_INLINE void call_async_ready(ErtsAsync *a)
 
 static ERTS_INLINE void async_reply(ErtsAsync *a, ErtsThrQPrepEnQ_t *prep_enq)
 {
-#if ERTS_USE_ASYNC_READY_Q
     ErtsAsyncReadyQ *arq;
 
 #if ERTS_ASYNC_PRINT_JOB
@@ -465,12 +415,6 @@ static ERTS_INLINE void async_reply(ErtsAsync *a, ErtsThrQPrepEnQ_t *prep_enq)
 	erts_mtx_unlock(&arq->x.data.enq_mtx);
 #endif
 
-#else /* ERTS_USE_ASYNC_READY_Q */
-
-	call_async_ready(a);
-	erts_free(ERTS_ALC_T_ASYNC, (void *) a);
-
-#endif /* ERTS_USE_ASYNC_READY_Q */
 }
 
 
@@ -486,7 +430,6 @@ static erts_tse_t *async_thread_init(ErtsAsyncQ *aq)
     erts_tse_t *tse = erts_tse_fetch();
     ERTS_DECLARE_DUMMY(Uint no);
 
-#ifdef ERTS_SMP
     ErtsThrPrgrCallbacks callbacks;
 
     callbacks.arg = (void *) tse;
@@ -495,15 +438,12 @@ static erts_tse_t *async_thread_init(ErtsAsyncQ *aq)
     callbacks.wait = NULL;
 
     erts_thr_progress_register_unmanaged_thread(&callbacks);
-#endif
 
     qinit.live.queue = ERTS_THR_Q_LIVE_LONG;
     qinit.live.objects = ERTS_THR_Q_LIVE_SHORT;
     qinit.arg = (void *) tse;
     qinit.notify = async_wakeup;
-#if ERTS_USE_ASYNC_READY_Q
     qinit.auto_finalize_dequeue = 0;
-#endif
 
     erts_thr_q_initialize(&aq->thr_q, &qinit);
 
@@ -545,12 +485,10 @@ static void *async_main(void* arg)
     return NULL;
 }
 
-#endif /* USE_THREADS */
 
 void
 erts_exit_flush_async(void)
 {
-#ifdef USE_THREADS
     int i;
     ErtsAsync a;
     a.port = NIL;
@@ -564,11 +502,8 @@ erts_exit_flush_async(void)
 	async_add(&a, async_q(i));
     for (i = 0; i < erts_async_max_threads; i++)
 	erts_thr_join(async->queue[i].aq.thr_id, NULL);
-#endif
 }
 
-#if defined(USE_THREADS) && ERTS_USE_ASYNC_READY_Q
-
 int erts_check_async_ready(void *varq)
 {
     ErtsAsyncReadyQ *arq = (ErtsAsyncReadyQ *) varq;
@@ -609,18 +544,15 @@ int erts_async_ready_clean(void *varq, void *val)
     case ERTS_THR_Q_DIRTY:
 	return ERTS_ASYNC_READY_DIRTY;
     case ERTS_THR_Q_NEED_THR_PRGR:
-#ifdef ERTS_SMP
 	*((ErtsThrPrgrVal *) val)
 	    = erts_thr_q_need_thr_progress(&arq->thr_q);
 	return ERTS_ASYNC_READY_NEED_THR_PRGR;
-#endif
     case ERTS_THR_Q_CLEAN:
 	break;
     }
     return ERTS_ASYNC_READY_CLEAN;
 }
 
-#endif
 
 /*
 ** Generate a fair async key prom an ErlDrvPort
@@ -658,28 +590,22 @@ long driver_async(ErlDrvPort ix, unsigned int* key,
     Port* prt;
     long id;
     unsigned int qix;
-#if ERTS_USE_ASYNC_READY_Q
     Uint sched_id;
     ERTS_MSACC_PUSH_STATE();
 
     sched_id = erts_get_scheduler_id();
     if (!sched_id)
 	sched_id = 1;
-#else
-    ERTS_MSACC_PUSH_STATE();
-#endif
 
     prt = erts_drvport2port(ix);
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     a = (ErtsAsync*) erts_alloc(ERTS_ALC_T_ASYNC, sizeof(ErtsAsync));
 
-#if ERTS_USE_ASYNC_READY_Q
     a->sched_id = sched_id;
-#endif
     a->hndl = (DE_Handle*)prt->drv_ptr->handle;
     a->port = prt->common.id;
     a->pdl = NULL;
@@ -709,7 +635,6 @@ long driver_async(ErlDrvPort ix, unsigned int* key,
 	    (*key % erts_async_max_threads) : 0;
 	*key = qix;
     }
-#ifdef USE_THREADS
     if (erts_async_max_threads > 0) {
 	if (prt->port_data_lock) {
 	    driver_pdl_inc_refc(prt->port_data_lock);
@@ -718,7 +643,6 @@ long driver_async(ErlDrvPort ix, unsigned int* key,
 	async_add(a, async_q(qix));
 	return id;
     }
-#endif
     
     ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_PORT);
     (*a->async_invoke)(a->async_data);
diff --git a/erts/emulator/beam/erl_async.h b/erts/emulator/beam/erl_async.h
index 4b470e7679..70ef247e0a 100644
--- a/erts/emulator/beam/erl_async.h
+++ b/erts/emulator/beam/erl_async.h
@@ -27,39 +27,12 @@ extern int erts_async_max_threads;
 #define ERTS_ASYNC_THREAD_MAX_STACK_SIZE 8192	/* Kilo words */
 extern int erts_async_thread_suggested_stack_size;
 
-
-#ifdef ERTS_SMP
-/*
- * With smp support we can choose to have, or not to
- * have an async ready queue.
- */
-#define ERTS_USE_ASYNC_READY_Q 1
-#endif
-
-#ifndef ERTS_SMP
-/* In non-smp case we *need* the async ready queue */
-#  undef ERTS_USE_ASYNC_READY_Q
-#  define ERTS_USE_ASYNC_READY_Q 1
-#endif
-
-#ifndef ERTS_USE_ASYNC_READY_Q
-#  define ERTS_USE_ASYNC_READY_Q 0
-#endif
-
-#ifndef USE_THREADS
-#  undef ERTS_USE_ASYNC_READY_Q
-#  define ERTS_USE_ASYNC_READY_Q 0
-#endif /* !USE_THREADS */
-#if ERTS_USE_ASYNC_READY_Q
 int erts_check_async_ready(void *);
 int erts_async_ready_clean(void *, void *);
 void *erts_get_async_ready_queue(Uint sched_id);
 #define ERTS_ASYNC_READY_CLEAN 0
 #define ERTS_ASYNC_READY_DIRTY 1
-#ifdef ERTS_SMP
 #define ERTS_ASYNC_READY_NEED_THR_PRGR 2
-#endif
-#endif /* ERTS_USE_ASYNC_READY_Q */
 
 void erts_init_async(void);
 void erts_exit_flush_async(void);
diff --git a/erts/emulator/beam/erl_bestfit_alloc.c b/erts/emulator/beam/erl_bestfit_alloc.c
index 6173c408e1..85fc4c3a85 100644
--- a/erts/emulator/beam/erl_bestfit_alloc.c
+++ b/erts/emulator/beam/erl_bestfit_alloc.c
@@ -875,7 +875,7 @@ static struct {
 
 static void ERTS_INLINE atom_init(Eterm *atom, char *name)
 {
-    *atom = am_atom_put(name, strlen(name));
+    *atom = am_atom_put(name, sys_strlen(name));
 }
 #define AM_INIT(AM) atom_init(&am.AM, #AM)
 
diff --git a/erts/emulator/beam/erl_bif_binary.c b/erts/emulator/beam/erl_bif_binary.c
index dcffde5777..469f6a1ea8 100644
--- a/erts/emulator/beam/erl_bif_binary.c
+++ b/erts/emulator/beam/erl_bif_binary.c
@@ -61,8 +61,6 @@ static Export binary_longest_prefix_trap_export;
 static BIF_RETTYPE binary_longest_prefix_trap(BIF_ALIST_3);
 static Export binary_longest_suffix_trap_export;
 static BIF_RETTYPE binary_longest_suffix_trap(BIF_ALIST_3);
-static Export binary_bin_to_list_trap_export;
-static BIF_RETTYPE binary_bin_to_list_trap(BIF_ALIST_3);
 static Export binary_copy_trap_export;
 static BIF_RETTYPE binary_copy_trap(BIF_ALIST_2);
 static Uint max_loop_limit;
@@ -86,10 +84,6 @@ void erts_init_bif_binary(void)
 			  am_erlang, am_binary_longest_suffix_trap, 3,
 			  &binary_longest_suffix_trap);
 
-    erts_init_trap_export(&binary_bin_to_list_trap_export,
-			  am_erlang, am_binary_bin_to_list_trap, 3,
-			  &binary_bin_to_list_trap);
-
     erts_init_trap_export(&binary_copy_trap_export,
 			  am_erlang, am_binary_copy_trap, 2,
 			  &binary_copy_trap);
@@ -171,6 +165,16 @@ static void *my_alloc(MyAllocator *my, Uint size)
 
 #define ALPHABET_SIZE 256
 
+typedef struct _findall_data {
+    Uint pos;
+    Uint len;
+#ifdef HARDDEBUG
+    Uint id;
+#endif
+    Eterm epos;
+    Eterm elen;
+} FindallData;
+
 typedef struct _ac_node {
 #ifdef HARDDEBUG
     Uint32 id;                        /* To identify h pointer targets when
@@ -208,6 +212,103 @@ typedef struct _bm_data {
     Sint badshift[ALPHABET_SIZE];
 } BMData;
 
+typedef struct _ac_find_all_state {
+    ACNode *q;
+    Uint pos;
+    Uint len;
+    Uint m;
+    Uint allocated;
+    FindallData *out;
+} ACFindAllState;
+
+typedef struct _ac_find_first_state {
+    ACNode *q;
+    Uint pos;
+    Uint len;
+    ACNode *candidate;
+    Uint candidate_start;
+} ACFindFirstState;
+
+typedef struct _bm_find_all_state {
+    Sint pos;
+    Sint len;
+    Uint m;
+    Uint allocated;
+    FindallData *out;
+} BMFindAllState;
+
+typedef struct _bm_find_first_state {
+    Sint pos;
+    Sint len;
+} BMFindFirstState;
+
+typedef enum _bf_return {
+    BF_RESTART = -3,
+    BF_NOT_FOUND,
+    BF_BADARG,
+    BF_OK
+} BFReturn;
+
+typedef struct _binary_find_all_context {
+    ErtsHeapFactory factory;
+    Eterm term;
+    Sint head;
+    Sint tail;
+    Uint end_pos;
+    Uint size;
+    FindallData *data;
+    union {
+	ACFindAllState ac;
+	BMFindAllState bm;
+    } d;
+} BinaryFindAllContext;
+
+typedef struct _binary_find_first_context {
+    Uint pos;
+    Uint len;
+    union {
+	ACFindFirstState ac;
+	BMFindFirstState bm;
+    } d;
+} BinaryFindFirstContext;
+
+typedef struct _binary_find_context BinaryFindContext;
+
+typedef struct _binary_find_search {
+    void (*init) (BinaryFindContext *);
+    BFReturn (*find) (BinaryFindContext *, byte *);
+    void (*done) (BinaryFindContext *);
+} BinaryFindSearch;
+
+typedef Eterm (*BinaryFindResult)(Process *, Eterm, BinaryFindContext **);
+
+typedef enum _binary_find_state {
+    BFSearch,
+    BFResult,
+    BFDone
+} BinaryFindState;
+
+struct _binary_find_context {
+    Eterm pat_type;
+    Eterm pat_term;
+    Binary *pat_bin;
+    Uint flags;
+    Uint hsstart;
+    Uint hsend;
+    int loop_factor;
+    int exported;
+    Uint reds;
+    BinaryFindState state;
+    Eterm trap_term;
+    BinaryFindSearch *search;
+    BinaryFindResult not_found;
+    BinaryFindResult found;
+    union {
+	BinaryFindAllContext fa;
+	BinaryFindFirstContext ff;
+    } u;
+};
+
 #ifdef HARDDEBUG
 static void dump_bm_data(BMData *bm);
 static void dump_ac_trie(ACTrie *act);
@@ -264,7 +365,7 @@ static ACTrie *create_acdata(MyAllocator *my, Uint len,
     acn->d = 0;
     acn->final = 0;
     acn->h = NULL;
-    memset(acn->g, 0, sizeof(ACNode *) * ALPHABET_SIZE);
+    sys_memset(acn->g, 0, sizeof(ACNode *) * ALPHABET_SIZE);
 #ifdef HARDDEBUG
     act->idc = 0;
     acn->id = 0;
@@ -287,7 +388,7 @@ static BMData *create_bmdata(MyAllocator *my, byte *x, Uint len,
     init_my_allocator(my, datasize, data);
     bmd = my_alloc(my, sizeof(BMData));
     bmd->x = my_alloc(my,len);
-    memcpy(bmd->x,x,len);
+    sys_memcpy(bmd->x,x,len);
     bmd->len = len;
     bmd->goodshift = my_alloc(my,sizeof(Uint) * len);
     *the_bin = mb;
@@ -332,7 +433,7 @@ static void ac_add_one_pattern(MyAllocator *my, ACTrie *act, byte *x, Uint len)
 	    nn->d = i+1;
 	    nn->h = act->root;
 	    nn->final = 0;
-	    memset(nn->g, 0, sizeof(ACNode *) * ALPHABET_SIZE);
+	    sys_memset(nn->g, 0, sizeof(ACNode *) * ALPHABET_SIZE);
 	    acn->g[x[i]] = nn;
 	    ++i;
 	    acn = nn;
@@ -414,32 +515,25 @@ static void ac_compute_failure_functions(ACTrie *act, ACNode **qbuff)
  * Basic AC finds the first end before the first start...
  *
  */
-typedef struct {
-    ACNode *q;
-    Uint pos;
-    Uint len;
-    ACNode *candidate;
-    Uint candidate_start;
-} ACFindFirstState;
-
-
-static void ac_init_find_first_match(ACFindFirstState *state, ACTrie *act, Sint startpos, Uint len)
+static void ac_init_find_first_match(BinaryFindContext *ctx)
 {
+    ACFindFirstState *state = &(ctx->u.ff.d.ac);
+    ACTrie *act = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
     state->q = act->root;
-    state->pos = startpos;
-    state->len = len;
+    state->pos = ctx->hsstart;
+    state->len = ctx->hsend;
     state->candidate = NULL;
     state->candidate_start = 0;
 }
-#define AC_OK 0
-#define AC_NOT_FOUND -1
-#define AC_RESTART -2
 
 #define AC_LOOP_FACTOR 10
 
-static int ac_find_first_match(ACFindFirstState *state, byte *haystack,
-				Uint *mpos, Uint *mlen, Uint *reductions)
+static BFReturn ac_find_first_match(BinaryFindContext *ctx, byte *haystack)
 {
+    ACFindFirstState *state = &(ctx->u.ff.d.ac);
+    Uint *mpos = &(ctx->u.ff.pos);
+    Uint *mlen = &(ctx->u.ff.len);
+    Uint *reductions = &(ctx->reds);
     ACNode *q = state->q;
     Uint i = state->pos;
     ACNode *candidate = state->candidate, *r;
@@ -455,7 +549,7 @@ static int ac_find_first_match(ACFindFirstState *state, byte *haystack,
 	    state->len = len;
 	    state->candidate = candidate;
 	    state->candidate_start = candidate_start;
-	    return AC_RESTART;
+	    return BF_RESTART;
 	}
 
 	while (q->g[haystack[i]] == NULL && q->h != q) {
@@ -485,68 +579,33 @@ static int ac_find_first_match(ACFindFirstState *state, byte *haystack,
     }
     *reductions = reds;
     if (!candidate) {
-	return AC_NOT_FOUND;
+	return BF_NOT_FOUND;
     }
 #ifdef HARDDEBUG
     dump_ac_node(candidate,0,'?');
 #endif
     *mpos = candidate_start;
     *mlen = candidate->d;
-    return AC_OK;
+    return BF_OK;
 }
 
-typedef struct _findall_data {
-    Uint pos;
-    Uint len;
-#ifdef HARDDEBUG
-    Uint id;
-#endif
-    Eterm epos;
-    Eterm elen;
-} FindallData;
-
-typedef struct {
-    ACNode *q;
-    Uint pos;
-    Uint len;
-    Uint m;
-    Uint allocated;
-    FindallData *out;
-} ACFindAllState;
-
-static void ac_init_find_all(ACFindAllState *state, ACTrie *act, Sint startpos, Uint len)
+static void ac_init_find_all(BinaryFindContext *ctx)
 {
+    ACFindAllState *state = &(ctx->u.fa.d.ac);
+    ACTrie *act = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
     state->q = act->root;
-    state->pos = startpos;
-    state->len = len;
+    state->pos = ctx->hsstart;
+    state->len = ctx->hsend;
     state->m = 0;
     state->allocated = 0;
     state->out = NULL;
 }
 
-static void ac_restore_find_all(ACFindAllState *state,
-                                const ACFindAllState *src)
-{
-    memcpy(state, src, sizeof(ACFindAllState));
-    if (state->allocated > 0) {
-	state->out = erts_alloc(ERTS_ALC_T_TMP, sizeof(FindallData) * (state->allocated));
-	memcpy(state->out, src+1, sizeof(FindallData)*state->m);
-    } else {
-	state->out = NULL;
-    }
-}
-
-static void ac_serialize_find_all(const ACFindAllState *state,
-                                  ACFindAllState *dst)
-{
-    memcpy(dst, state, sizeof(ACFindAllState));
-    memcpy(dst+1, state->out, sizeof(FindallData)*state->m);
-}
-
-static void ac_clean_find_all(ACFindAllState *state)
+static void ac_clean_find_all(BinaryFindContext *ctx)
 {
+    ACFindAllState *state = &(ctx->u.fa.d.ac);
     if (state->out != NULL) {
-	erts_free(ERTS_ALC_T_TMP, state->out);
+	erts_free(ERTS_ALC_T_BINARY_FIND, state->out);
     }
 #ifdef HARDDEBUG
     state->out = NULL;
@@ -558,9 +617,10 @@ static void ac_clean_find_all(ACFindAllState *state)
  * Differs to the find_first function in that it stores all matches and the values
  * arte returned only in the state.
  */
-static int ac_find_all_non_overlapping(ACFindAllState *state, byte *haystack,
-				       Uint *reductions)
+static BFReturn ac_find_all_non_overlapping(BinaryFindContext *ctx, byte *haystack)
 {
+    ACFindAllState *state = &(ctx->u.fa.d.ac);
+    Uint *reductions = &(ctx->reds);
     ACNode *q = state->q;
     Uint i = state->pos;
     Uint rstart;
@@ -571,7 +631,6 @@ static int ac_find_all_non_overlapping(ACFindAllState *state, byte *haystack,
     FindallData *out = state->out;
     register Uint reds = *reductions;
 
-
     while (i < len) {
 	if (--reds == 0) {
 	    state->q = q;
@@ -580,7 +639,7 @@ static int ac_find_all_non_overlapping(ACFindAllState *state, byte *haystack,
 	    state->m = m;
 	    state->allocated = allocated;
 	    state->out = out;
-	    return AC_RESTART;
+	    return BF_RESTART;
 	}
 	while (q->g[haystack[i]] == NULL && q->h != q) {
 	    q = q->h;
@@ -618,11 +677,11 @@ static int ac_find_all_non_overlapping(ACFindAllState *state, byte *haystack,
 		    if (m >= allocated) {
 			if (!allocated) {
 			    allocated = 10;
-			    out = erts_alloc(ERTS_ALC_T_TMP,
+			    out = erts_alloc(ERTS_ALC_T_BINARY_FIND,
 					     sizeof(FindallData) * allocated);
 			} else {
 			    allocated *= 2;
-			    out = erts_realloc(ERTS_ALC_T_TMP, out,
+			    out = erts_realloc(ERTS_ALC_T_BINARY_FIND, out,
 					       sizeof(FindallData) *
 					       allocated);
 			}
@@ -649,7 +708,7 @@ static int ac_find_all_non_overlapping(ACFindAllState *state, byte *haystack,
     *reductions = reds;
     state->m = m;
     state->out = out;
-    return (m == 0) ? AC_NOT_FOUND : AC_OK;
+    return (m == 0) ? BF_NOT_FOUND : BF_OK;
 }
 
 /*
@@ -736,27 +795,22 @@ static void compute_goodshifts(BMData *bmd)
     erts_free(ERTS_ALC_T_TMP, suffixes);
 }
 
-typedef struct {
-    Sint pos;
-    Sint len;
-} BMFindFirstState;
-
-#define BM_OK 0 /* used only for find_all */
-#define BM_NOT_FOUND -1
-#define BM_RESTART -2
 #define BM_LOOP_FACTOR 10 /* Should we have a higher value? */
 
-static void bm_init_find_first_match(BMFindFirstState *state, Sint startpos,
-				     Uint len)
+static void bm_init_find_first_match(BinaryFindContext *ctx)
 {
-    state->pos = startpos;
-    state->len = (Sint) len;
+    BMFindFirstState *state = &(ctx->u.ff.d.bm);
+    state->pos = ctx->hsstart;
+    state->len = ctx->hsend;
 }
 
-
-static Sint bm_find_first_match(BMFindFirstState *state, BMData *bmd,
-				byte *haystack, Uint *reductions)
+static BFReturn bm_find_first_match(BinaryFindContext *ctx, byte *haystack)
 {
+    BMFindFirstState *state = &(ctx->u.ff.d.bm);
+    BMData *bmd = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
+    Uint *mpos = &(ctx->u.ff.pos);
+    Uint *mlen = &(ctx->u.ff.len);
+    Uint *reductions = &(ctx->reds);
     Sint blen = bmd->len;
     Sint len = state->len;
     Sint *gs = bmd->goodshift;
@@ -769,61 +823,37 @@ static Sint bm_find_first_match(BMFindFirstState *state, BMData *bmd,
     while (j <= len - blen) {
 	if (--reds == 0) {
 	    state->pos = j;
-	    return BM_RESTART;
+	    return BF_RESTART;
 	}
 	for (i = blen - 1; i >= 0 && needle[i] == haystack[i + j]; --i)
 	    ;
 	if (i < 0) { /* found */
 	    *reductions = reds;
-	    return j;
+	    *mpos = (Uint) j;
+	    *mlen = (Uint) blen;
+	    return BF_OK;
 	}
 	j += MAX(gs[i],bs[haystack[i+j]] - blen + 1 + i);
     }
     *reductions = reds;
-    return BM_NOT_FOUND;
+    return BF_NOT_FOUND;
 }
 
-typedef struct {
-    Sint pos;
-    Sint len;
-    Uint m;
-    Uint allocated;
-    FindallData *out;
-} BMFindAllState;
-
-static void bm_init_find_all(BMFindAllState *state, Sint startpos, Uint len)
+static void bm_init_find_all(BinaryFindContext *ctx)
 {
-    state->pos = startpos;
-    state->len = (Sint) len;
+    BMFindAllState *state = &(ctx->u.fa.d.bm);
+    state->pos = ctx->hsstart;
+    state->len = ctx->hsend;
     state->m = 0;
     state->allocated = 0;
     state->out = NULL;
 }
 
-static void bm_restore_find_all(BMFindAllState *state,
-                                const BMFindAllState *src)
-{
-    memcpy(state, src, sizeof(BMFindAllState));
-    if (state->allocated > 0) {
-	state->out = erts_alloc(ERTS_ALC_T_TMP, sizeof(FindallData) *
-				(state->allocated));
-	memcpy(state->out, src+1, sizeof(FindallData)*state->m);
-    } else {
-	state->out = NULL;
-    }
-}
-
-static void bm_serialize_find_all(const BMFindAllState *state,
-                                  BMFindAllState *dst)
-{
-    memcpy(dst, state, sizeof(BMFindAllState));
-    memcpy(dst+1, state->out, sizeof(FindallData)*state->m);
-}
-
-static void bm_clean_find_all(BMFindAllState *state)
+static void bm_clean_find_all(BinaryFindContext *ctx)
 {
+    BMFindAllState *state = &(ctx->u.fa.d.bm);
     if (state->out != NULL) {
-	erts_free(ERTS_ALC_T_TMP, state->out);
+	erts_free(ERTS_ALC_T_BINARY_FIND, state->out);
     }
 #ifdef HARDDEBUG
     state->out = NULL;
@@ -835,10 +865,11 @@ static void bm_clean_find_all(BMFindAllState *state)
  * Differs to the find_first function in that it stores all matches and the
  * values are returned only in the state.
  */
-static Sint bm_find_all_non_overlapping(BMFindAllState *state,
-					BMData *bmd, byte *haystack,
-					Uint *reductions)
+static BFReturn bm_find_all_non_overlapping(BinaryFindContext *ctx, byte *haystack)
 {
+    BMFindAllState *state = &(ctx->u.fa.d.bm);
+    BMData *bmd = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
+    Uint *reductions = &(ctx->reds);
     Sint blen = bmd->len;
     Sint len = state->len;
     Sint *gs = bmd->goodshift;
@@ -857,7 +888,7 @@ static Sint bm_find_all_non_overlapping(BMFindAllState *state,
 	    state->m = m;
 	    state->allocated = allocated;
 	    state->out = out;
-	    return BM_RESTART;
+	    return BF_RESTART;
 	}
 	for (i = blen - 1; i >= 0 && needle[i] == haystack[i + j]; --i)
 	    ;
@@ -865,10 +896,11 @@ static Sint bm_find_all_non_overlapping(BMFindAllState *state,
 	    if (m >= allocated) {
 		if (!allocated) {
 		    allocated = 10;
-		    out = erts_alloc(ERTS_ALC_T_TMP, sizeof(FindallData) * allocated);
+		    out = erts_alloc(ERTS_ALC_T_BINARY_FIND,
+				     sizeof(FindallData) * allocated);
 		} else {
 		    allocated *= 2;
-		    out = erts_realloc(ERTS_ALC_T_TMP, out,
+		    out = erts_realloc(ERTS_ALC_T_BINARY_FIND, out,
 				       sizeof(FindallData) * allocated);
 		}
 	    }
@@ -883,7 +915,7 @@ static Sint bm_find_all_non_overlapping(BMFindAllState *state,
     state->m = m;
     state->out = out;
     *reductions = reds;
-    return (m == 0) ? BM_NOT_FOUND : BM_OK;
+    return (m == 0) ? BF_NOT_FOUND : BF_OK;
 }
 
 /*
@@ -1009,51 +1041,160 @@ BIF_RETTYPE binary_compile_pattern_1(BIF_ALIST_1)
     BIF_RET(ret);
 }
 
-#define DO_BIN_MATCH_OK 0
-#define DO_BIN_MATCH_BADARG -1
-#define DO_BIN_MATCH_RESTART -2
+#define BF_FLAG_GLOBAL		0x01
+#define BF_FLAG_SPLIT_TRIM	0x02
+#define BF_FLAG_SPLIT_TRIM_ALL	0x04
 
-#define BINARY_FIND_ALL		0x01
-#define BINARY_SPLIT_TRIM	0x02
-#define BINARY_SPLIT_TRIM_ALL	0x04
+static void bf_context_init(BinaryFindContext *ctx, BinaryFindResult not_found,
+			    BinaryFindResult single, BinaryFindResult global,
+			    Binary *pat_bin);
+static BinaryFindContext *bf_context_export(Process *p, BinaryFindContext *src);
+static int bf_context_destructor(Binary *ctx_bin);
+#ifdef HARDDEBUG
+static void bf_context_dump(BinaryFindContext *ctx);
+#endif
 
-typedef struct BinaryFindState {
-    Eterm type;
-    Uint  flags;
-    Uint  hsstart;
-    Uint  hsend;
-    Eterm (*not_found_result) (Process *, Eterm, struct BinaryFindState *);
-    Eterm (*single_result)    (Process *, Eterm, struct BinaryFindState *, Sint, Sint);
-    Eterm (*global_result)    (Process *, Eterm, struct BinaryFindState *, FindallData *, Uint);
-} BinaryFindState;
+static BinaryFindSearch bf_search_ac_global = {
+    ac_init_find_all,
+    ac_find_all_non_overlapping,
+    ac_clean_find_all
+};
+
+static BinaryFindSearch bf_search_ac_single = {
+    ac_init_find_first_match,
+    ac_find_first_match,
+    NULL
+};
+
+static BinaryFindSearch bf_search_bm_global = {
+    bm_init_find_all,
+    bm_find_all_non_overlapping,
+    bm_clean_find_all
+};
+
+static BinaryFindSearch bf_search_bm_single = {
+    bm_init_find_first_match,
+    bm_find_first_match,
+    NULL
+};
+
+static void bf_context_init(BinaryFindContext *ctx, BinaryFindResult not_found,
+			    BinaryFindResult single, BinaryFindResult global,
+			    Binary *pat_bin)
+{
+    ctx->exported = 0;
+    ctx->state = BFSearch;
+    ctx->not_found = not_found;
+    if (ctx->flags & BF_FLAG_GLOBAL) {
+	ctx->found = global;
+	if (ctx->pat_type == am_bm) {
+	    ctx->search = &bf_search_bm_global;
+	    ctx->loop_factor = BM_LOOP_FACTOR;
+	} else if (ctx->pat_type == am_ac) {
+	    ctx->search = &bf_search_ac_global;
+	    ctx->loop_factor = AC_LOOP_FACTOR;
+	}
+    } else {
+	ctx->found = single;
+	if (ctx->pat_type == am_bm) {
+	    ctx->search = &bf_search_bm_single;
+	    ctx->loop_factor = BM_LOOP_FACTOR;
+	} else if (ctx->pat_type == am_ac) {
+	    ctx->search = &bf_search_ac_single;
+	    ctx->loop_factor = AC_LOOP_FACTOR;
+	}
+    }
+    ctx->trap_term = THE_NON_VALUE;
+    ctx->pat_bin = pat_bin;
+    ctx->search->init(ctx);
+}
 
-typedef struct BinaryFindState_bignum {
-    Eterm           bignum_hdr;
-    BinaryFindState bfs;
-    union {
-	BMFindFirstState bmffs;
-	BMFindAllState   bmfas;
-	ACFindFirstState acffs;
-	ACFindAllState   acfas;
-    } data;
-} BinaryFindState_bignum;
-
-#define SIZEOF_BINARY_FIND_STATE(S) \
-	  (sizeof(BinaryFindState)+sizeof(S))
-
-#define SIZEOF_BINARY_FIND_ALL_STATE(S) \
-	  (sizeof(BinaryFindState)+sizeof(S)+(sizeof(FindallData)*(S).m))
-
-static Eterm do_match_not_found_result(Process *p, Eterm subject, BinaryFindState *bfs);
-static Eterm do_match_single_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    Sint pos, Sint len);
-static Eterm do_match_global_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    FindallData *fad, Uint fad_sz);
-static Eterm do_split_not_found_result(Process *p, Eterm subject, BinaryFindState *bfs);
-static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    Sint pos, Sint len);
-static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    FindallData *fad, Uint fad_sz);
+static BinaryFindContext *bf_context_export(Process *p, BinaryFindContext *src)
+{
+    Binary *ctx_bin;
+    BinaryFindContext *ctx;
+    Eterm *hp;
+
+    ASSERT(src->exported == 0);
+    ctx_bin = erts_create_magic_binary(sizeof(BinaryFindContext),
+				       bf_context_destructor);
+    ctx = ERTS_MAGIC_BIN_DATA(ctx_bin);
+    sys_memcpy(ctx, src, sizeof(BinaryFindContext));
+    if (ctx->pat_bin != NULL && ctx->pat_term == THE_NON_VALUE) {
+	hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE * 2);
+	ctx->pat_term = erts_mk_magic_ref(&hp, &MSO(p), ctx->pat_bin);
+    } else {
+	hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE);
+    }
+    ctx->trap_term = erts_mk_magic_ref(&hp, &MSO(p), ctx_bin);
+    ctx->exported = 1;
+    return ctx;
+}
+
+static int bf_context_destructor(Binary *ctx_bin)
+{
+    BinaryFindContext *ctx;
+
+    ctx = ERTS_MAGIC_BIN_DATA(ctx_bin);
+    if (ctx->state != BFDone) {
+	if (ctx->search->done != NULL) {
+	    ctx->search->done(ctx);
+	}
+	ctx->state = BFDone;
+    }
+    return 1;
+}
+
+#ifdef HARDDEBUG
+static void bf_context_dump(BinaryFindContext *ctx)
+{
+    if (ctx->pat_type == am_bm) {
+	BMData *bm;
+	bm = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
+	dump_bm_data(bm);
+    } else {
+	ACTrie *act;
+	act = ERTS_MAGIC_BIN_DATA(ctx->pat_bin);
+	dump_ac_trie(act);
+    }
+}
+#endif
+
+static Eterm do_match_not_found_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+static Eterm do_match_single_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+static Eterm do_match_global_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+static Eterm do_split_not_found_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindContext **ctxp);
+
+static BFReturn maybe_binary_match_compile(BinaryFindContext *ctx, Eterm arg, Binary **pat_bin)
+{
+    Eterm *tp;
+    ctx->pat_term = THE_NON_VALUE;
+    if (is_tuple(arg)) {
+	tp = tuple_val(arg);
+	if (arityval(*tp) != 2 || is_not_atom(tp[1])) {
+	    return BF_BADARG;
+	}
+	if (((tp[1] != am_bm) && (tp[1] != am_ac)) ||
+	    !is_internal_magic_ref(tp[2])) {
+	    return BF_BADARG;
+	}
+	*pat_bin = erts_magic_ref2bin(tp[2]);
+	if ((tp[1] == am_bm &&
+	     ERTS_MAGIC_BIN_DESTRUCTOR(*pat_bin) != cleanup_my_data_bm) ||
+	    (tp[1] == am_ac &&
+	     ERTS_MAGIC_BIN_DESTRUCTOR(*pat_bin) != cleanup_my_data_ac)) {
+	    *pat_bin = NULL;
+	    return BF_BADARG;
+	}
+	ctx->pat_type = tp[1];
+	ctx->pat_term = tp[2];
+    } else if (do_binary_match_compile(arg, &(ctx->pat_type), pat_bin) != 0) {
+	return BF_BADARG;
+    }
+    return BF_OK;
+}
 
 static int parse_match_opts_list(Eterm l, Eterm bin, Uint *posp, Uint *endp)
 {
@@ -1134,17 +1275,17 @@ static int parse_split_opts_list(Eterm l, Eterm bin, Uint *posp, Uint *endp, Uin
 	    Uint orig_size;
 	    if (is_atom(t)) {
 		if (t == am_global) {
-		    *optp |= BINARY_FIND_ALL;
+		    *optp |= BF_FLAG_GLOBAL;
 		    l = CDR(list_val(l));
 		    continue;
 		}
 		if (t == am_trim) {
-		    *optp |= BINARY_SPLIT_TRIM;
+		    *optp |= BF_FLAG_SPLIT_TRIM;
 		    l = CDR(list_val(l));
 		    continue;
 		}
 		if (t == am_trim_all) {
-		    *optp |= BINARY_SPLIT_TRIM_ALL;
+		    *optp |= BF_FLAG_SPLIT_TRIM_ALL;
 		    l = CDR(list_val(l));
 		    continue;
 		}
@@ -1197,266 +1338,160 @@ static int parse_split_opts_list(Eterm l, Eterm bin, Uint *posp, Uint *endp, Uin
     }
 }
 
-static int do_binary_find(Process *p, Eterm subject, BinaryFindState *bfs, Binary *bin,
-			  Eterm state_term, Eterm *res_term)
+static BFReturn do_binary_find(Process *p, Eterm subject, BinaryFindContext **ctxp,
+			       Binary *pat_bin, Binary *ctx_bin, Eterm *res_term)
 {
-    byte *bytes;
-    Uint bitoffs, bitsize;
-    byte *temp_alloc = NULL;
-    BinaryFindState_bignum *state_ptr = NULL;
+    BinaryFindContext *ctx;
+    int is_first_call;
+    Uint initial_reds;
+    BFReturn runres;
 
-    ERTS_GET_BINARY_BYTES(subject, bytes, bitoffs, bitsize);
-    if (bitsize != 0) {
-	goto badarg;
-    }
-    if (bitoffs != 0) {
-	bytes = erts_get_aligned_binary_bytes(subject, &temp_alloc);
-    }
-    if (state_term != NIL) {
-	state_ptr = (BinaryFindState_bignum *)(big_val(state_term));
-	bfs = &(state_ptr->bfs);
+    if (ctx_bin == NULL) {
+	is_first_call = 1;
+	ctx = *ctxp;
+    } else {
+	is_first_call = 0;
+	ctx = ERTS_MAGIC_BIN_DATA(ctx_bin);
+	ctx->pat_bin = pat_bin;
+	*ctxp = ctx;
     }
 
-    if (bfs->flags & BINARY_FIND_ALL) {
-	if (bfs->type == am_bm) {
-	    BMData *bm;
-	    Sint pos;
-	    BMFindAllState state;
-	    Uint reds = get_reds(p, BM_LOOP_FACTOR);
-	    Uint save_reds = reds;
+    initial_reds = ctx->reds = get_reds(p, ctx->loop_factor);
 
-	    bm = (BMData *) ERTS_MAGIC_BIN_DATA(bin);
-#ifdef HARDDEBUG
-	    dump_bm_data(bm);
-#endif
-	    if (state_term == NIL) {
-		bm_init_find_all(&state, bfs->hsstart, bfs->hsend);
-	    } else {
-		bm_restore_find_all(&state, &(state_ptr->data.bmfas));
-	    }
+    switch (ctx->state) {
+    case BFSearch: {
+	byte *bytes;
+	Uint bitoffs, bitsize;
+	byte *temp_alloc = NULL;
 
-	    pos = bm_find_all_non_overlapping(&state, bm, bytes, &reds);
-	    if (pos == BM_NOT_FOUND) {
-		*res_term = bfs->not_found_result(p, subject, bfs);
-	    } else if (pos == BM_RESTART) {
-		int x =
-		    (SIZEOF_BINARY_FIND_ALL_STATE(state) / sizeof(Eterm)) +
-		    !!(SIZEOF_BINARY_FIND_ALL_STATE(state) % sizeof(Eterm));
-#ifdef HARDDEBUG
-		erts_printf("Trap bm!\n");
-#endif
-		state_ptr = (BinaryFindState_bignum*) HAlloc(p, x+1);
-		state_ptr->bignum_hdr = make_pos_bignum_header(x);
-		memcpy(&state_ptr->bfs, bfs, sizeof(BinaryFindState));
-		bm_serialize_find_all(&state, &state_ptr->data.bmfas);
-		*res_term = make_big(&state_ptr->bignum_hdr);
-		erts_free_aligned_binary_bytes(temp_alloc);
-		bm_clean_find_all(&state);
-		return DO_BIN_MATCH_RESTART;
-	    } else {
-		*res_term = bfs->global_result(p, subject, bfs, state.out, state.m);
-	    }
-	    erts_free_aligned_binary_bytes(temp_alloc);
-	    bm_clean_find_all(&state);
-	    BUMP_REDS(p, (save_reds - reds) / BM_LOOP_FACTOR);
-	    return DO_BIN_MATCH_OK;
-	} else if (bfs->type == am_ac) {
-	    ACTrie *act;
-	    int acr;
-	    ACFindAllState state;
-	    Uint reds = get_reds(p, AC_LOOP_FACTOR);
-	    Uint save_reds = reds;
-
-	    act = (ACTrie *) ERTS_MAGIC_BIN_DATA(bin);
+	ERTS_GET_BINARY_BYTES(subject, bytes, bitoffs, bitsize);
+	if (bitsize != 0) {
+	    goto badarg;
+	}
+	if (bitoffs != 0) {
+	    bytes = erts_get_aligned_binary_bytes(subject, &temp_alloc);
+	}
 #ifdef HARDDEBUG
-	    dump_ac_trie(act);
+	bf_context_dump(ctx);
 #endif
-	    if (state_term == NIL) {
-		ac_init_find_all(&state, act, bfs->hsstart, bfs->hsend);
-	    } else {
-		ac_restore_find_all(&state, &(state_ptr->data.acfas));
-	    }
-	    acr = ac_find_all_non_overlapping(&state, bytes, &reds);
-	    if (acr == AC_NOT_FOUND) {
-		*res_term = bfs->not_found_result(p, subject, bfs);
-	    } else if (acr == AC_RESTART) {
-		int x =
-		    (SIZEOF_BINARY_FIND_ALL_STATE(state) / sizeof(Eterm)) +
-		    !!(SIZEOF_BINARY_FIND_ALL_STATE(state) % sizeof(Eterm));
+	runres = ctx->search->find(ctx, bytes);
+	if (runres == BF_NOT_FOUND) {
+	    *res_term = ctx->not_found(p, subject, &ctx);
+	    *ctxp = ctx;
+	} else if (runres == BF_RESTART) {
 #ifdef HARDDEBUG
+	    if (ctx->pat_type == am_ac) {
 		erts_printf("Trap ac!\n");
-#endif
-                state_ptr = (BinaryFindState_bignum*) HAlloc(p, x+1);
-		state_ptr->bignum_hdr = make_pos_bignum_header(x);
-		memcpy(&state_ptr->bfs, bfs, sizeof(BinaryFindState));
-		ac_serialize_find_all(&state, &state_ptr->data.acfas);
-		*res_term = make_big(&state_ptr->bignum_hdr);
-		erts_free_aligned_binary_bytes(temp_alloc);
-		ac_clean_find_all(&state);
-		return DO_BIN_MATCH_RESTART;
 	    } else {
-		*res_term = bfs->global_result(p, subject, bfs, state.out, state.m);
-	    }
-	    erts_free_aligned_binary_bytes(temp_alloc);
-	    ac_clean_find_all(&state);
-	    BUMP_REDS(p, (save_reds - reds) / AC_LOOP_FACTOR);
-	    return DO_BIN_MATCH_OK;
-	}
-    } else {
-	if (bfs->type == am_bm) {
-	    BMData *bm;
-	    Sint pos;
-	    BMFindFirstState state;
-	    Uint reds = get_reds(p, BM_LOOP_FACTOR);
-	    Uint save_reds = reds;
-
-	    bm = (BMData *) ERTS_MAGIC_BIN_DATA(bin);
-#ifdef HARDDEBUG
-	    dump_bm_data(bm);
-#endif
-	    if (state_term == NIL) {
-		bm_init_find_first_match(&state, bfs->hsstart, bfs->hsend);
-	    } else {
-		memcpy(&state, &state_ptr->data.bmffs, sizeof(BMFindFirstState));
-	    }
-
-#ifdef HARDDEBUG
-	    erts_printf("(bm) state->pos = %ld, state->len = %lu\n",state.pos,
-			state.len);
-#endif
-	    pos = bm_find_first_match(&state, bm, bytes, &reds);
-	    if (pos == BM_NOT_FOUND) {
-		*res_term = bfs->not_found_result(p, subject, bfs);
-	    } else if (pos == BM_RESTART) {
-		int x =
-		    (SIZEOF_BINARY_FIND_STATE(state) / sizeof(Eterm)) +
-		    !!(SIZEOF_BINARY_FIND_STATE(state) % sizeof(Eterm));
-#ifdef HARDDEBUG
 		erts_printf("Trap bm!\n");
+	    }
 #endif
-                state_ptr = (BinaryFindState_bignum*) HAlloc(p, x+1);
-                state_ptr->bignum_hdr = make_pos_bignum_header(x);
-		memcpy(&state_ptr->bfs, bfs, sizeof(BinaryFindState));
-		memcpy(&state_ptr->data.acffs, &state, sizeof(BMFindFirstState));
-		*res_term = make_big(&state_ptr->bignum_hdr);
-		erts_free_aligned_binary_bytes(temp_alloc);
-		return DO_BIN_MATCH_RESTART;
-	    } else {
-		*res_term = bfs->single_result(p, subject, bfs, pos, bm->len);
+	    if (is_first_call) {
+		ctx = bf_context_export(p, ctx);
+		*ctxp = ctx;
+		erts_set_gc_state(p, 0);
 	    }
 	    erts_free_aligned_binary_bytes(temp_alloc);
-	    BUMP_REDS(p, (save_reds - reds) / BM_LOOP_FACTOR);
-	    return DO_BIN_MATCH_OK;
-	} else if (bfs->type == am_ac) {
-	    ACTrie *act;
-	    Uint pos, rlen;
-	    int acr;
-	    ACFindFirstState state;
-	    Uint reds = get_reds(p, AC_LOOP_FACTOR);
-	    Uint save_reds = reds;
-
-	    act = (ACTrie *) ERTS_MAGIC_BIN_DATA(bin);
-#ifdef HARDDEBUG
-	    dump_ac_trie(act);
-#endif
-	    if (state_term == NIL) {
-		ac_init_find_first_match(&state, act, bfs->hsstart, bfs->hsend);
-	    } else {
-		memcpy(&state, &state_ptr->data.acffs, sizeof(ACFindFirstState));
+	    *res_term = THE_NON_VALUE;
+	    BUMP_ALL_REDS(p);
+	    return BF_RESTART;
+	} else {
+	    *res_term = ctx->found(p, subject, &ctx);
+	    *ctxp = ctx;
+	}
+	erts_free_aligned_binary_bytes(temp_alloc);
+	if (*res_term == THE_NON_VALUE) {
+	    if (is_first_call) {
+		erts_set_gc_state(p, 0);
 	    }
-	    acr = ac_find_first_match(&state, bytes, &pos, &rlen, &reds);
-	    if (acr == AC_NOT_FOUND) {
-		*res_term = bfs->not_found_result(p, subject, bfs);
-	    } else if (acr == AC_RESTART) {
-		int x =
-		    (SIZEOF_BINARY_FIND_STATE(state) / sizeof(Eterm)) +
-		    !!(SIZEOF_BINARY_FIND_STATE(state) % sizeof(Eterm));
-#ifdef HARDDEBUG
-		erts_printf("Trap ac!\n");
-#endif
-		state_ptr = (BinaryFindState_bignum*) HAlloc(p, x+1);
-		state_ptr->bignum_hdr = make_pos_bignum_header(x);
-		memcpy(&state_ptr->bfs, bfs, sizeof(BinaryFindState));
-		memcpy(&state_ptr->data.acffs, &state, sizeof(ACFindFirstState));
-		*res_term = make_big(&state_ptr->bignum_hdr);
-		erts_free_aligned_binary_bytes(temp_alloc);
-		return DO_BIN_MATCH_RESTART;
-	    } else {
-		*res_term = bfs->single_result(p, subject, bfs, pos, rlen);
+	    BUMP_ALL_REDS(p);
+	    return BF_RESTART;
+	}
+	if (ctx->search->done != NULL) {
+	    ctx->search->done(ctx);
+	}
+	ctx->state = BFDone;
+	if (!is_first_call) {
+	    erts_set_gc_state(p, 1);
+	}
+	BUMP_REDS(p, (initial_reds - ctx->reds) / ctx->loop_factor);
+	return BF_OK;
+    }
+    case BFResult: {
+	*res_term = ctx->found(p, subject, &ctx);
+	*ctxp = ctx;
+	if (*res_term == THE_NON_VALUE) {
+	    if (is_first_call) {
+		erts_set_gc_state(p, 0);
 	    }
-	    erts_free_aligned_binary_bytes(temp_alloc);
-	    BUMP_REDS(p, (save_reds - reds) / AC_LOOP_FACTOR);
-	    return DO_BIN_MATCH_OK;
+	    BUMP_ALL_REDS(p);
+	    return BF_RESTART;
+	}
+	if (ctx->search->done != NULL) {
+	    ctx->search->done(ctx);
+	}
+	ctx->state = BFDone;
+	if (!is_first_call) {
+	    erts_set_gc_state(p, 1);
 	}
+	BUMP_REDS(p, (initial_reds - ctx->reds) / ctx->loop_factor);
+	return BF_OK;
     }
- badarg:
-    return DO_BIN_MATCH_BADARG;
+    default:
+	ASSERT(!"Unknown state in do_binary_find");
+    }
+
+badarg:
+    if (!is_first_call) {
+	if (ctx->search->done != NULL) {
+	    ctx->search->done(ctx);
+	}
+	ctx->state = BFDone;
+	erts_set_gc_state(p, 1);
+    }
+    return BF_BADARG;
 }
 
 static BIF_RETTYPE
 binary_match(Process *p, Eterm arg1, Eterm arg2, Eterm arg3, Uint flags)
 {
-    BinaryFindState bfs;
-    Eterm *tp;
-    Binary *bin;
-    Eterm bin_term = NIL;
+    BinaryFindContext c_buff;
+    BinaryFindContext *ctx = &c_buff;
+    Binary *pat_bin;
     int runres;
     Eterm result;
 
-    if (is_not_binary(arg1)) {
+    if (is_not_binary(arg1) || binary_bitsize(arg1) != 0) {
 	goto badarg;
     }
-    bfs.flags = flags;
-    if (parse_match_opts_list(arg3, arg1, &(bfs.hsstart), &(bfs.hsend))) {
+    ctx->flags = flags;
+    if (parse_match_opts_list(arg3, arg1, &(ctx->hsstart), &(ctx->hsend))) {
 	goto badarg;
     }
-    if (bfs.hsend == 0) {
-	BIF_RET(do_match_not_found_result(p, arg1, &bfs));
+    if (ctx->hsend == 0) {
+	result = do_match_not_found_result(p, arg1, &ctx);
+	BIF_RET(result);
     }
-    if (is_tuple(arg2)) {
-	tp = tuple_val(arg2);
-	if (arityval(*tp) != 2 || is_not_atom(tp[1])) {
-	    goto badarg;
-	}
-	if (((tp[1] != am_bm) && (tp[1] != am_ac)) ||
-	    !is_internal_magic_ref(tp[2])) {
-	    goto badarg;
-	}
-	bfs.type = tp[1];
-	bin = erts_magic_ref2bin(tp[2]);
-	if (bfs.type == am_bm &&
-	    ERTS_MAGIC_BIN_DESTRUCTOR(bin) != cleanup_my_data_bm) {
-	    goto badarg;
-	}
-	if (bfs.type == am_ac &&
-	    ERTS_MAGIC_BIN_DESTRUCTOR(bin) != cleanup_my_data_ac) {
-	    goto badarg;
-	}
-	bin_term = tp[2];
-    } else if (do_binary_match_compile(arg2, &(bfs.type), &bin)) {
+    if (maybe_binary_match_compile(ctx, arg2, &pat_bin) != BF_OK) {
 	goto badarg;
     }
-    bfs.not_found_result = &do_match_not_found_result;
-    bfs.single_result = &do_match_single_result;
-    bfs.global_result = &do_match_global_result;
-    runres = do_binary_find(p, arg1, &bfs, bin, NIL, &result);
-    if (runres == DO_BIN_MATCH_RESTART && bin_term == NIL) {
-	Eterm *hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE);
-	bin_term = erts_mk_magic_ref(&hp, &MSO(p), bin);
-    } else if (bin_term == NIL) {
-	erts_bin_free(bin);
+    bf_context_init(ctx, do_match_not_found_result, do_match_single_result,
+		    do_match_global_result, pat_bin);
+    runres = do_binary_find(p, arg1, &ctx, pat_bin, NULL, &result);
+    if (runres == BF_OK && ctx->pat_term == THE_NON_VALUE) {
+	erts_bin_free(pat_bin);
     }
     switch (runres) {
-    case DO_BIN_MATCH_OK:
+    case BF_OK:
 	BIF_RET(result);
-    case DO_BIN_MATCH_RESTART:
-	BUMP_ALL_REDS(p);
-	BIF_TRAP3(&binary_find_trap_export, p, arg1, result, bin_term);
+    case BF_RESTART:
+	ASSERT(result == THE_NON_VALUE && ctx->trap_term != result && ctx->pat_term != result);
+	BIF_TRAP3(&binary_find_trap_export, p, arg1, ctx->trap_term, ctx->pat_term);
     default:
 	goto badarg;
     }
- badarg:
-    BIF_ERROR(p,BADARG);
+badarg:
+    BIF_ERROR(p, BADARG);
 }
 
 BIF_RETTYPE binary_match_2(BIF_ALIST_2)
@@ -1471,76 +1506,52 @@ BIF_RETTYPE binary_match_3(BIF_ALIST_3)
 
 BIF_RETTYPE binary_matches_2(BIF_ALIST_2)
 {
-    return binary_match(BIF_P, BIF_ARG_1, BIF_ARG_2, THE_NON_VALUE, BINARY_FIND_ALL);
+    return binary_match(BIF_P, BIF_ARG_1, BIF_ARG_2, THE_NON_VALUE, BF_FLAG_GLOBAL);
 }
 
 BIF_RETTYPE binary_matches_3(BIF_ALIST_3)
 {
-    return binary_match(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, BINARY_FIND_ALL);
+    return binary_match(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, BF_FLAG_GLOBAL);
 }
 
 static BIF_RETTYPE
 binary_split(Process *p, Eterm arg1, Eterm arg2, Eterm arg3)
 {
-    BinaryFindState bfs;
-    Eterm *tp;
-    Binary *bin;
-    Eterm bin_term = NIL;
+    BinaryFindContext c_buff;
+    BinaryFindContext *ctx = &c_buff;
+    Binary *pat_bin;
     int runres;
     Eterm result;
 
-    if (is_not_binary(arg1)) {
+    if (is_not_binary(arg1) || binary_bitsize(arg1) != 0) {
 	goto badarg;
     }
-    if (parse_split_opts_list(arg3, arg1, &(bfs.hsstart), &(bfs.hsend), &(bfs.flags))) {
+    if (parse_split_opts_list(arg3, arg1, &(ctx->hsstart), &(ctx->hsend), &(ctx->flags))) {
 	goto badarg;
     }
-    if (bfs.hsend == 0) {
-	result = do_split_not_found_result(p, arg1, &bfs);
+    if (ctx->hsend == 0) {
+	result = do_split_not_found_result(p, arg1, &ctx);
 	BIF_RET(result);
     }
-    if (is_tuple(arg2)) {
-	tp = tuple_val(arg2);
-	if (arityval(*tp) != 2 || is_not_atom(tp[1])) {
-	    goto badarg;
-	}
-	if (((tp[1] != am_bm) && (tp[1] != am_ac)) ||
-	    !is_internal_magic_ref(tp[2])) {
-	    goto badarg;
-	}
-	bfs.type = tp[1];
-	bin = erts_magic_ref2bin(tp[2]);
-	if (bfs.type == am_bm &&
-	    ERTS_MAGIC_BIN_DESTRUCTOR(bin) != cleanup_my_data_bm) {
-	    goto badarg;
-	}
-	if (bfs.type == am_ac &&
-	    ERTS_MAGIC_BIN_DESTRUCTOR(bin) != cleanup_my_data_ac) {
-	    goto badarg;
-	}
-	bin_term = tp[2];
-    } else if (do_binary_match_compile(arg2, &(bfs.type), &bin)) {
+    if (maybe_binary_match_compile(ctx, arg2, &pat_bin) != BF_OK) {
 	goto badarg;
     }
-    bfs.not_found_result = &do_split_not_found_result;
-    bfs.single_result = &do_split_single_result;
-    bfs.global_result = &do_split_global_result;
-    runres = do_binary_find(p, arg1, &bfs, bin, NIL, &result);
-    if (runres == DO_BIN_MATCH_RESTART && bin_term == NIL) {
-	Eterm *hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE);
-	bin_term = erts_mk_magic_ref(&hp, &MSO(p), bin);
-    } else if (bin_term == NIL) {
-	erts_bin_free(bin);
-    }
-    switch(runres) {
-    case DO_BIN_MATCH_OK:
+    bf_context_init(ctx, do_split_not_found_result, do_split_single_result,
+		    do_split_global_result, pat_bin);
+    runres = do_binary_find(p, arg1, &ctx, pat_bin, NULL, &result);
+    if (runres == BF_OK && ctx->pat_term == THE_NON_VALUE) {
+	erts_bin_free(pat_bin);
+    }
+    switch (runres) {
+    case BF_OK:
 	BIF_RET(result);
-    case DO_BIN_MATCH_RESTART:
-	BIF_TRAP3(&binary_find_trap_export, p, arg1, result, bin_term);
+    case BF_RESTART:
+	ASSERT(result == THE_NON_VALUE && ctx->trap_term != result && ctx->pat_term != result);
+	BIF_TRAP3(&binary_find_trap_export, p, arg1, ctx->trap_term, ctx->pat_term);
     default:
 	goto badarg;
     }
- badarg:
+badarg:
     BIF_ERROR(p, BADARG);
 }
 
@@ -1554,72 +1565,117 @@ BIF_RETTYPE binary_split_3(BIF_ALIST_3)
     return binary_split(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
 }
 
-static Eterm do_match_not_found_result(Process *p, Eterm subject, BinaryFindState *bfs)
+static Eterm do_match_not_found_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
-    if (bfs->flags & BINARY_FIND_ALL) {
+    if ((*ctxp)->flags & BF_FLAG_GLOBAL) {
 	return NIL;
     } else {
 	return am_nomatch;
     }
 }
 
-static Eterm do_match_single_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    Sint pos, Sint len)
+static Eterm do_match_single_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
+    BinaryFindContext *ctx = (*ctxp);
+    BinaryFindFirstContext *ff = &(ctx->u.ff);
     Eterm erlen;
     Eterm *hp;
     Eterm ret;
 
-    erlen = erts_make_integer((Uint)(len), p);
-    ret = erts_make_integer(pos, p);
+    erlen = erts_make_integer((Uint)(ff->len), p);
+    ret = erts_make_integer(ff->pos, p);
     hp = HAlloc(p, 3);
     ret = TUPLE2(hp, ret, erlen);
 
     return ret;
 }
 
-static Eterm do_match_global_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    FindallData *fad, Uint fad_sz)
+static Eterm do_match_global_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
-    Sint i;
+    BinaryFindContext *ctx = (*ctxp);
+    BinaryFindAllContext *fa = &(ctx->u.fa);
+    FindallData *fad;
     Eterm tpl;
-    Eterm *hp;
-    Eterm ret;
+    Sint i;
+    register Uint reds = ctx->reds;
 
-    for (i = 0; i < fad_sz; ++i) {
-	fad[i].epos = erts_make_integer(fad[i].pos, p);
-	fad[i].elen = erts_make_integer(fad[i].len, p);
+    if (ctx->state == BFSearch) {
+	if (ctx->pat_type == am_ac) {
+	    fa->data = fa->d.ac.out;
+	    fa->size = fa->d.ac.m;
+	} else {
+	    fa->data = fa->d.bm.out;
+	    fa->size = fa->d.bm.m;
+	}
+	fa->tail = fa->size - 1;
+	fa->head = 0;
+	fa->end_pos = 0;
+	fa->term = NIL;
+	if (ctx->exported == 0 && ((fa->size * 2) >= reds)) {
+	    ctx = bf_context_export(p, ctx);
+	    *ctxp = ctx;
+	    fa = &(ctx->u.fa);
+	}
+	erts_factory_proc_prealloc_init(&(fa->factory), p, fa->size * (3 + 2));
+	ctx->state = BFResult;
+    }
+
+    fad = fa->data;
+
+    if (fa->end_pos == 0) {
+	for (i = fa->head; i < fa->size; ++i) {
+	    if (--reds == 0) {
+		ASSERT(ctx->exported == 1);
+		fa->head = i;
+		ctx->reds = reds;
+		return THE_NON_VALUE;
+	    }
+	    fad[i].epos = erts_make_integer(fad[i].pos, p);
+	    fad[i].elen = erts_make_integer(fad[i].len, p);
+	}
+	fa->end_pos = 1;
+	fa->head = fa->tail;
     }
-    hp = HAlloc(p, fad_sz * (3 + 2));
-    ret = NIL;
-    for (i = fad_sz - 1; i >= 0; --i) {
-	tpl = TUPLE2(hp, fad[i].epos, fad[i].elen);
-	hp += 3;
-	ret = CONS(hp, tpl, ret);
-	hp += 2;
+
+    for (i = fa->head; i >= 0; --i) {
+	if (--reds == 0) {
+	    ASSERT(ctx->exported == 1);
+	    fa->head = i;
+	    ctx->reds = reds;
+	    return THE_NON_VALUE;
+	}
+	tpl = TUPLE2(fa->factory.hp, fad[i].epos, fad[i].elen);
+	fa->factory.hp += 3;
+	fa->term = CONS(fa->factory.hp, tpl, fa->term);
+	fa->factory.hp += 2;
     }
+    ctx->reds = reds;
+    erts_factory_close(&(fa->factory));
 
-    return ret;
+    return fa->term;
 }
 
-static Eterm do_split_not_found_result(Process *p, Eterm subject, BinaryFindState *bfs)
+static Eterm do_split_not_found_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
+    BinaryFindContext *ctx = (*ctxp);
     Eterm *hp;
     Eterm ret;
 
-    if (bfs->flags & (BINARY_SPLIT_TRIM | BINARY_SPLIT_TRIM_ALL)
+    if (ctx->flags & (BF_FLAG_SPLIT_TRIM | BF_FLAG_SPLIT_TRIM_ALL)
         && binary_size(subject) == 0) {
-        return NIL;
+	return NIL;
     }
     hp = HAlloc(p, 2);
     ret = CONS(hp, subject, NIL);
-
     return ret;
 }
 
-static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    Sint pos, Sint len)
+static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
+    BinaryFindContext *ctx = (*ctxp);
+    BinaryFindFirstContext *ff = &(ctx->u.ff);
+    Sint pos;
+    Sint len;
     size_t orig_size;
     Eterm orig;
     Uint offset;
@@ -1630,9 +1686,12 @@ static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindState *
     Eterm *hp;
     Eterm ret;
 
+    pos = ff->pos;
+    len = ff->len;
+
     orig_size = binary_size(subject);
 
-    if ((bfs->flags & (BINARY_SPLIT_TRIM | BINARY_SPLIT_TRIM_ALL)) &&
+    if ((ctx->flags & (BF_FLAG_SPLIT_TRIM | BF_FLAG_SPLIT_TRIM_ALL)) &&
 	(orig_size - pos - len) == 0) {
 	if (pos == 0) {
 	    ret = NIL;
@@ -1653,7 +1712,7 @@ static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindState *
 	    hp += 2;
 	}
     } else {
-	if ((bfs->flags & BINARY_SPLIT_TRIM_ALL) && (pos == 0)) {
+	if ((ctx->flags & BF_FLAG_SPLIT_TRIM_ALL) && (pos == 0)) {
 	    hp = HAlloc(p, 1 * (ERL_SUB_BIN_SIZE + 2));
 	    ERTS_GET_REAL_BIN(subject, orig, offset, bit_offset, bit_size);
 	    sb1 = NULL;
@@ -1691,39 +1750,60 @@ static Eterm do_split_single_result(Process *p, Eterm subject, BinaryFindState *
     return ret;
 }
 
-static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindState *bfs,
-				    FindallData *fad, Uint fad_sz)
+static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindContext **ctxp)
 {
-    size_t orig_size;
+    BinaryFindContext *ctx = (*ctxp);
+    BinaryFindAllContext *fa = &(ctx->u.fa);
+    FindallData *fad;
     Eterm orig;
+    size_t orig_size;
     Uint offset;
     Uint bit_offset;
     Uint bit_size;
     ErlSubBin *sb;
+    Uint do_trim;
     Sint i;
-    Sint tail;
-    Uint list_size;
-    Uint end_pos;
-    Uint do_trim = bfs->flags & (BINARY_SPLIT_TRIM | BINARY_SPLIT_TRIM_ALL);
-    Eterm *hp;
-    Eterm *hendp;
-    Eterm ret;
+    register Uint reds = ctx->reds;
 
-    tail = fad_sz - 1;
-    list_size = fad_sz + 1;
-    orig_size = binary_size(subject);
-    end_pos = (Uint)(orig_size);
+    if (ctx->state == BFSearch) {
+	if (ctx->pat_type == am_ac) {
+	    fa->data = fa->d.ac.out;
+	    fa->size = fa->d.ac.m;
+	} else {
+	    fa->data = fa->d.bm.out;
+	    fa->size = fa->d.bm.m;
+	}
+	fa->tail = fa->size - 1;
+	fa->head = fa->tail;
+	orig_size = binary_size(subject);
+	fa->end_pos = (Uint)(orig_size);
+	fa->term = NIL;
+	if (ctx->exported == 0 && ((fa->head + 1) >= reds)) {
+	    ctx = bf_context_export(p, ctx);
+	    *ctxp = ctx;
+	    fa = &(ctx->u.fa);
+	}
+	erts_factory_proc_prealloc_init(&(fa->factory), p, (fa->size + 1) * (ERL_SUB_BIN_SIZE + 2));
+	ctx->state = BFResult;
+    }
 
-    hp = HAlloc(p, list_size * (ERL_SUB_BIN_SIZE + 2));
-    hendp = hp + list_size * (ERL_SUB_BIN_SIZE + 2);
     ERTS_GET_REAL_BIN(subject, orig, offset, bit_offset, bit_size);
     ASSERT(bit_size == 0);
+    fad = fa->data;
+    do_trim = ctx->flags & (BF_FLAG_SPLIT_TRIM | BF_FLAG_SPLIT_TRIM_ALL);
 
-    ret = NIL;
-
-    for (i = tail; i >= 0; --i) {
-	sb = (ErlSubBin *)(hp);
-	sb->size = end_pos - (fad[i].pos + fad[i].len);
+    for (i = fa->head; i >= 0; --i) {
+	if (--reds == 0) {
+	    ASSERT(ctx->exported == 1);
+	    fa->head = i;
+	    ctx->reds = reds;
+	    if (!do_trim && (ctx->flags & BF_FLAG_SPLIT_TRIM)) {
+		ctx->flags &= ~BF_FLAG_SPLIT_TRIM;
+	    }
+	    return THE_NON_VALUE;
+	}
+	sb = (ErlSubBin *)(fa->factory.hp);
+	sb->size = fa->end_pos - (fad[i].pos + fad[i].len);
 	if (!(sb->size == 0 && do_trim)) {
 	    sb->thing_word = HEADER_SUB_BIN;
 	    sb->offs = offset + fad[i].pos + fad[i].len;
@@ -1731,15 +1811,18 @@ static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindState *
 	    sb->bitoffs = bit_offset;
 	    sb->bitsize = 0;
 	    sb->is_writable = 0;
-	    hp += ERL_SUB_BIN_SIZE;
-	    ret = CONS(hp, make_binary(sb), ret);
-	    hp += 2;
-	    do_trim &= ~BINARY_SPLIT_TRIM;
+	    fa->factory.hp += ERL_SUB_BIN_SIZE;
+	    fa->term = CONS(fa->factory.hp, make_binary(sb), fa->term);
+	    fa->factory.hp += 2;
+	    do_trim &= ~BF_FLAG_SPLIT_TRIM;
 	}
-	end_pos = fad[i].pos;
+	fa->end_pos = fad[i].pos;
     }
 
-    sb = (ErlSubBin *)(hp);
+    fa->head = i;
+    ctx->reds = reds;
+
+    sb = (ErlSubBin *)(fa->factory.hp);
     sb->size = fad[0].pos;
     if (!(sb->size == 0 && do_trim)) {
 	sb->thing_word = HEADER_SUB_BIN;
@@ -1748,26 +1831,31 @@ static Eterm do_split_global_result(Process *p, Eterm subject, BinaryFindState *
 	sb->bitoffs = bit_offset;
 	sb->bitsize = 0;
 	sb->is_writable = 0;
-	hp += ERL_SUB_BIN_SIZE;
-	ret = CONS(hp, make_binary(sb), ret);
-	hp += 2;
+	fa->factory.hp += ERL_SUB_BIN_SIZE;
+	fa->term = CONS(fa->factory.hp, make_binary(sb), fa->term);
+	fa->factory.hp += 2;
     }
-    HRelease(p, hendp, hp);
-    return ret;
+    erts_factory_close(&(fa->factory));
+
+    return fa->term;
 }
 
 static BIF_RETTYPE binary_find_trap(BIF_ALIST_3)
 {
     int runres;
     Eterm result;
-    Binary *bin = erts_magic_ref2bin(BIF_ARG_3);
-
-    runres = do_binary_find(BIF_P, BIF_ARG_1, NULL, bin, BIF_ARG_2, &result);
-    if (runres == DO_BIN_MATCH_OK) {
+    Binary *ctx_bin = erts_magic_ref2bin(BIF_ARG_2);
+    Binary *pat_bin = erts_magic_ref2bin(BIF_ARG_3);
+    BinaryFindContext *ctx = NULL;
+
+    ASSERT(ERTS_MAGIC_BIN_DESTRUCTOR(ctx_bin) == bf_context_destructor);
+    runres = do_binary_find(BIF_P, BIF_ARG_1, &ctx, pat_bin, ctx_bin, &result);
+    if (runres == BF_OK) {
+	ASSERT(result != THE_NON_VALUE);
 	BIF_RET(result);
     } else {
-	BUMP_ALL_REDS(BIF_P);
-	BIF_TRAP3(&binary_find_trap_export, BIF_P, BIF_ARG_1, result, BIF_ARG_3);
+	ASSERT(result == THE_NON_VALUE && ctx->trap_term != result && ctx->pat_term != result);
+	BIF_TRAP3(&binary_find_trap_export, BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
     }
 }
 
@@ -2176,7 +2264,7 @@ static BIF_RETTYPE do_longest_common(Process *p, Eterm list, int direction)
 	    if (cd[i].type == CL_TYPE_HEAP_NOALLOC) {
 		unsigned char *tmp = cd[i].buff;
 		cd[i].buff = erts_alloc(ERTS_ALC_T_BINARY_BUFFER, cd[i].bufflen);
-		memcpy(cd[i].buff,tmp,cd[i].bufflen);
+		sys_memcpy(cd[i].buff,tmp,cd[i].bufflen);
 		cd[i].type = CL_TYPE_HEAP;
 	    }
 	}
@@ -2346,191 +2434,6 @@ BIF_RETTYPE binary_at_2(BIF_ALIST_2)
     BIF_ERROR(BIF_P,BADARG);
 }
 
-#define BIN_TO_LIST_OK 0
-#define BIN_TO_LIST_TRAP 1
-/* No badarg, checked before call */
-
-#define BIN_TO_LIST_LOOP_FACTOR 10
-
-static int do_bin_to_list(Process *p, byte *bytes, Uint bit_offs,
-			  Uint start, Sint *lenp, Eterm *termp)
-{
-    Uint reds = get_reds(p, BIN_TO_LIST_LOOP_FACTOR); /* reds can never be 0 */
-    Uint len = *lenp;
-    Uint loops;
-    Eterm *hp;
-    Eterm term = *termp;
-    Uint n;
-
-    ASSERT(reds > 0);
-
-    loops = MIN(reds,len);
-
-    BUMP_REDS(p, loops / BIN_TO_LIST_LOOP_FACTOR);
-
-    hp = HAlloc(p,2*loops);
-    while (loops--) {
-	--len;
-	if (bit_offs) {
-	    n = ((((Uint) bytes[start+len]) << bit_offs) |
-		 (((Uint) bytes[start+len+1]) >> (8-bit_offs))) & 0xFF;
-	} else {
-	    n = bytes[start+len];
-	}
-
-	term = CONS(hp,make_small(n),term);
-	hp +=2;
-    }
-    *termp = term;
-    *lenp = len;
-    if (len) {
-	BUMP_ALL_REDS(p);
-	return BIN_TO_LIST_TRAP;
-    }
-    return BIN_TO_LIST_OK;
-}
-
-
-static BIF_RETTYPE do_trap_bin_to_list(Process *p, Eterm binary,
-				       Uint start, Sint len, Eterm sofar)
-{
-    Eterm *hp;
-    Eterm blob;
-
-    hp = HAlloc(p,3);
-    hp[0] = make_pos_bignum_header(2);
-    hp[1] = start;
-    hp[2] = (Uint) len;
-    blob = make_big(hp);
-    BIF_TRAP3(&binary_bin_to_list_trap_export, p, binary, blob, sofar);
-}
-
-static BIF_RETTYPE binary_bin_to_list_trap(BIF_ALIST_3)
-{
-    Eterm *ptr;
-    Uint start;
-    Sint len;
-    byte *bytes;
-    Uint bit_offs;
-    Uint bit_size;
-    Eterm res = BIF_ARG_3;
-
-    ptr  = big_val(BIF_ARG_2);
-    start = ptr[1];
-    len = (Sint) ptr[2];
-
-    ERTS_GET_BINARY_BYTES(BIF_ARG_1,bytes,bit_offs,bit_size);
-    if (do_bin_to_list(BIF_P, bytes, bit_offs, start, &len, &res) ==
-	BIN_TO_LIST_OK) {
-	BIF_RET(res);
-    }
-    return do_trap_bin_to_list(BIF_P,BIF_ARG_1,start,len,res);
-}
-
-static BIF_RETTYPE binary_bin_to_list_common(Process *p,
-					     Eterm bin,
-					     Eterm epos,
-					     Eterm elen)
-{
-    Uint pos;
-    Sint len;
-    size_t sz;
-    byte *bytes;
-    Uint bit_offs;
-    Uint bit_size;
-    Eterm res = NIL;
-
-    if (is_not_binary(bin)) {
-	goto badarg;
-    }
-    if (!term_to_Uint(epos, &pos)) {
-	goto badarg;
-    }
-    if (!term_to_Sint(elen, &len)) {
-	goto badarg;
-    }
-    if (len < 0) {
-	Uint lentmp = -(Uint)len;
-	/* overflow */
-	if ((Sint)lentmp < 0) {
-	    goto badarg;
-	}
-	len = lentmp;
-	if (len > pos) {
-	    goto badarg;
-	}
-	pos -= len;
-    }
-    /* overflow */
-    if ((pos + len) < pos || (len > 0 && (pos + len) == pos)) {
-	goto badarg;
-    }
-    sz = binary_size(bin);
-
-    if (pos+len > sz) {
-	goto badarg;
-    }
-    ERTS_GET_BINARY_BYTES(bin,bytes,bit_offs,bit_size);
-    if (bit_size != 0) {
-	goto badarg;
-    }
-    if(do_bin_to_list(p, bytes, bit_offs, pos, &len, &res) ==
-       BIN_TO_LIST_OK) {
-	BIF_RET(res);
-    }
-    return do_trap_bin_to_list(p,bin,pos,len,res);
-
- badarg:
-    BIF_ERROR(p,BADARG);
-}
-
-BIF_RETTYPE binary_bin_to_list_3(BIF_ALIST_3)
-{
-    return binary_bin_to_list_common(BIF_P,BIF_ARG_1,BIF_ARG_2,BIF_ARG_3);
-}
-
-BIF_RETTYPE binary_bin_to_list_2(BIF_ALIST_2)
-{
-    Eterm *tp;
-
-    if (is_not_tuple(BIF_ARG_2)) {
-	goto badarg;
-    }
-    tp = tuple_val(BIF_ARG_2);
-    if (arityval(*tp) != 2) {
-	goto badarg;
-    }
-    return binary_bin_to_list_common(BIF_P,BIF_ARG_1,tp[1],tp[2]);
- badarg:
-    BIF_ERROR(BIF_P,BADARG);
-}
-
-BIF_RETTYPE binary_bin_to_list_1(BIF_ALIST_1)
-{
-    Uint pos = 0;
-    Sint len;
-    byte *bytes;
-    Uint bit_offs;
-    Uint bit_size;
-    Eterm res = NIL;
-
-    if (is_not_binary(BIF_ARG_1)) {
-	goto badarg;
-    }
-    len = binary_size(BIF_ARG_1);
-    ERTS_GET_BINARY_BYTES(BIF_ARG_1,bytes,bit_offs,bit_size);
-    if (bit_size != 0) {
-	goto badarg;
-    }
-    if(do_bin_to_list(BIF_P, bytes, bit_offs, pos, &len, &res) ==
-       BIN_TO_LIST_OK) {
-	BIF_RET(res);
-    }
-    return do_trap_bin_to_list(BIF_P,BIF_ARG_1,pos,len,res);
- badarg:
-    BIF_ERROR(BIF_P,BADARG);
-}
-
 HIPE_WRAPPER_BIF_DISABLE_GC(binary_list_to_bin, 1)
 
 BIF_RETTYPE binary_list_to_bin_1(BIF_ALIST_1)
@@ -2653,7 +2556,7 @@ static BIF_RETTYPE do_binary_copy(Process *p, Eterm bin, Eterm en)
 						    0);
 	    if (!(cbs->temp_alloc)) { /* alignment not needed, need to copy */
 		byte *tmp = erts_alloc(ERTS_ALC_T_BINARY_BUFFER,size);
-		memcpy(tmp,cbs->source,size);
+		sys_memcpy(tmp,cbs->source,size);
 		cbs->source = tmp;
 		cbs->source_type = BC_TYPE_HEAP;
 	    } else {
@@ -2666,7 +2569,7 @@ static BIF_RETTYPE do_binary_copy(Process *p, Eterm bin, Eterm en)
 	pos = 0;
 	i = 0;
 	while (pos < reds) {
-	    memcpy(t+pos,cbs->source, size);
+	    sys_memcpy(t+pos,cbs->source, size);
 	    pos += size;
 	    ++i;
 	}
@@ -2691,7 +2594,7 @@ static BIF_RETTYPE do_binary_copy(Process *p, Eterm bin, Eterm en)
 	}
 	pos = 0;
 	while (pos < target_size) {
-	    memcpy(t+pos,source, size);
+	    sys_memcpy(t+pos,source, size);
 	    pos += size;
 	}
 	erts_free_aligned_binary_bytes(temp_alloc);
@@ -2721,7 +2624,7 @@ BIF_RETTYPE binary_copy_trap(BIF_ALIST_2)
     if ((n-1) * size >= reds) {
 	Uint i = 0;
 	while ((pos - opos) < reds) {
-	    memcpy(t+pos,cbs->source, size);
+	    sys_memcpy(t+pos,cbs->source, size);
 	    pos += size;
 	    ++i;
 	}
@@ -2731,28 +2634,21 @@ BIF_RETTYPE binary_copy_trap(BIF_ALIST_2)
 	BIF_TRAP2(&binary_copy_trap_export, BIF_P, BIF_ARG_1, BIF_ARG_2);
     } else {
 	Binary *save;
-	ProcBin* pb;
+        Eterm resbin;
 	Uint target_size = cbs->result->orig_size;
 	while (pos < target_size) {
-	    memcpy(t+pos,cbs->source, size);
+	    sys_memcpy(t+pos,cbs->source, size);
 	    pos += size;
 	}
-	save =  cbs->result;
+	save = cbs->result;
 	cbs->result = NULL;
 	cleanup_copy_bin_state(mb); /* now cbs is dead */
-	pb = (ProcBin *) HAlloc(BIF_P, PROC_BIN_SIZE);
-	pb->thing_word = HEADER_PROC_BIN;
-	pb->size = target_size;
-	pb->next = MSO(BIF_P).first;
-	MSO(BIF_P).first = (struct erl_off_heap_header*) pb;
-	pb->val = save;
-	pb->bytes = t;
-	pb->flags = 0;
-
-	OH_OVERHEAD(&(MSO(BIF_P)), target_size / sizeof(Eterm));
-	BUMP_REDS(BIF_P,(pos - opos) / BINARY_COPY_LOOP_FACTOR);
 
-	BIF_RET(make_binary(pb));
+        resbin = erts_build_proc_bin(&MSO(BIF_P),
+                                     HAlloc(BIF_P, PROC_BIN_SIZE),
+                                     save);
+        BUMP_REDS(BIF_P,(pos - opos) / BINARY_COPY_LOOP_FACTOR);
+	BIF_RET(resbin);
     }
 }
 
@@ -3133,7 +3029,7 @@ static void dump_bm_data(BMData *bm)
 static void dump_ac_node(ACNode *node, int indent, int ch) {
     int i;
     char *spaces = erts_alloc(ERTS_ALC_T_TMP, 10 * indent + 1);
-    memset(spaces,' ',10*indent);
+    sys_memset(spaces,' ',10*indent);
     spaces[10*indent] = '\0';
     erts_printf("%s-> %c\n",spaces,ch);
     erts_printf("%sId: %u\n",spaces,(unsigned) node->id);
diff --git a/erts/emulator/beam/erl_bif_chksum.c b/erts/emulator/beam/erl_bif_chksum.c
index 9417803e14..cf92687595 100644
--- a/erts/emulator/beam/erl_bif_chksum.c
+++ b/erts/emulator/beam/erl_bif_chksum.c
@@ -44,7 +44,7 @@ void erts_init_bif_chksum(void)
 {
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&chksum_md5_2_exp,
-			  am_erlang, am_atom_put("md5_trap",8), 2,
+			  am_erlang, ERTS_MAKE_AM("md5_trap"), 2,
 			  &md5_2);
 }
     
@@ -516,7 +516,7 @@ md5_2(BIF_ALIST_2)
     /* No need to check context, this function cannot be called with unaligned
        or badly sized context as it's always trapped to. */
     bytes = binary_bytes(BIF_ARG_1);
-    memcpy(&context,bytes,sizeof(MD5_CTX));
+    sys_memcpy(&context,bytes,sizeof(MD5_CTX));
     rest = do_chksum(&md5_wrap,BIF_P,BIF_ARG_2,100,(void *) &context,&res,
 		     &err);
     if (err != 0) {
@@ -564,7 +564,7 @@ md5_update_2(BIF_ALIST_2)
 	erts_free_aligned_binary_bytes(temp_alloc);
 	BIF_ERROR(BIF_P, BADARG);
     }
-    memcpy(&context,bytes,sizeof(MD5_CTX));
+    sys_memcpy(&context,bytes,sizeof(MD5_CTX));
     erts_free_aligned_binary_bytes(temp_alloc);
     rest = do_chksum(&md5_wrap,BIF_P,BIF_ARG_2,100,(void *) &context,&res,
 		     &err);
@@ -599,7 +599,7 @@ md5_final_1(BIF_ALIST_1)
 	goto error;
     }
     bin = erts_new_heap_binary(BIF_P, (byte *)NULL, 16, &result);
-    memcpy(&ctx_copy, context, sizeof(MD5_CTX));
+    sys_memcpy(&ctx_copy, context, sizeof(MD5_CTX));
     erts_free_aligned_binary_bytes(temp_alloc);
     MD5Final(result, &ctx_copy);
     BIF_RET(bin);
diff --git a/erts/emulator/beam/erl_bif_ddll.c b/erts/emulator/beam/erl_bif_ddll.c
index e9bfb39035..294bce115f 100644
--- a/erts/emulator/beam/erl_bif_ddll.c
+++ b/erts/emulator/beam/erl_bif_ddll.c
@@ -50,13 +50,6 @@
 #include "dtrace-wrapper.h"
 #include "lttng-wrapper.h"
 
-#ifdef ERTS_SMP
-#define DDLL_SMP 1
-#else
-#define DDLL_SMP 0
-#endif
-
-
 /*
  * Local types
  */
@@ -107,18 +100,18 @@ static void dereference_all_processes(DE_Handle *dh);
 static void restore_process_references(DE_Handle *dh);
 static void ddll_no_more_references(void *vdh);
 
-#define lock_drv_list() erts_smp_rwmtx_rwlock(&erts_driver_list_lock)
-#define unlock_drv_list() erts_smp_rwmtx_rwunlock(&erts_driver_list_lock)
+#define lock_drv_list() erts_rwmtx_rwlock(&erts_driver_list_lock)
+#define unlock_drv_list() erts_rwmtx_rwunlock(&erts_driver_list_lock)
 #define assert_drv_list_locked() \
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rwlocked(&erts_driver_list_lock) \
-		       || erts_smp_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&erts_driver_list_lock) \
+		       || erts_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
 #define assert_drv_list_rwlocked() \
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rwlocked(&erts_driver_list_lock))
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&erts_driver_list_lock))
 #define assert_drv_list_rlocked() \
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
 #define assert_drv_list_not_locked() \
-    ERTS_SMP_LC_ASSERT(!erts_smp_lc_rwmtx_is_rwlocked(&erts_driver_list_lock) \
-		       && !erts_smp_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
+    ERTS_LC_ASSERT(!erts_lc_rwmtx_is_rwlocked(&erts_driver_list_lock) \
+		       && !erts_lc_rwmtx_is_rlocked(&erts_driver_list_lock))
 
 
 #define FREE_PORT_FLAGS (ERTS_PORT_SFLGS_DEAD & (~ERTS_PORT_SFLG_INITIALIZING))
@@ -134,13 +127,13 @@ kill_ports_driver_unloaded(DE_Handle *dh)
 	if (!prt)
 	    continue;
 
-	ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
+	ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
 
 	state = erts_atomic32_read_nob(&prt->state);
 	if (state & FREE_PORT_FLAGS)
 	    continue;
 
-	erts_smp_port_lock(prt);
+	erts_port_lock(prt);
 
 	state = erts_atomic32_read_nob(&prt->state);
 	if (!(state & ERTS_PORT_SFLGS_DEAD) && prt->drv_ptr->handle == dh)
@@ -280,10 +273,8 @@ BIF_RETTYPE erl_ddll_try_load_3(BIF_ALIST_3)
     path[path_len++] = '/';
     sys_strcpy(path+path_len,name);
 
-#if DDLL_SMP
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
     lock_drv_list();
-#endif
     if ((drv = lookup_driver(name)) != NULL) {
 	if (drv->handle == NULL) {
 	    /* static_driver */
@@ -404,24 +395,18 @@ BIF_RETTYPE erl_ddll_try_load_3(BIF_ALIST_3)
 	erts_ddll_reference_driver(dh);
 	ASSERT(dh->status == ERL_DE_RELOAD);
 	dh->status = ERL_DE_FORCE_RELOAD;
-#if DDLL_SMP
 	unlock_drv_list();
-#endif
 	kill_ports_driver_unloaded(dh);
 	/* Dereference, eventually causing driver destruction */
-#if DDLL_SMP
 	lock_drv_list(); 
-#endif
 	erts_ddll_dereference_driver(dh);
     } 
 
-#if DDLL_SMP
     erts_ddll_reference_driver(dh);
     unlock_drv_list();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     lock_drv_list();
     erts_ddll_dereference_driver(dh);
-#endif
 
     BIF_P->flags |= F_USING_DDLL;
     if (monitor) {
@@ -432,18 +417,14 @@ BIF_RETTYPE erl_ddll_try_load_3(BIF_ALIST_3)
 	hp = HAlloc(BIF_P, 3);
 	t = TUPLE2(hp, am_ok, ok_term);
     }
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) path);
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
     BIF_RET(t);
  soft_error:
-#if DDLL_SMP
     unlock_drv_list();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
-#endif
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     if (do_build_load_error) {
 	soft_error_term = build_load_error(BIF_P, build_this_load_error);
     }
@@ -452,11 +433,11 @@ BIF_RETTYPE erl_ddll_try_load_3(BIF_ALIST_3)
     t = TUPLE2(hp, am_error, soft_error_term);
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) path);
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
     BIF_RET(t);
  error:
     assert_drv_list_not_locked();
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
     if (path != NULL) {
 	erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) path);
     }
@@ -518,7 +499,7 @@ Eterm erl_ddll_try_unload_2(BIF_ALIST_2)
     Eterm l;
     int kill_ports = 0;
 
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
     for(l = options; is_list(l); l =  CDR(list_val(l))) {
 	Eterm opt = CAR(list_val(l));
@@ -551,9 +532,7 @@ Eterm erl_ddll_try_unload_2(BIF_ALIST_2)
 	goto error;
     }
 
-#if DDLL_SMP
     lock_drv_list();
-#endif
 
     if ((drv = lookup_driver(name)) == NULL) {
 	soft_error_term = am_not_loaded;
@@ -597,7 +576,7 @@ Eterm erl_ddll_try_unload_2(BIF_ALIST_2)
 	dh->reload_full_path = dh->reload_driver_name = NULL; 
 	dh->reload_flags = 0;
     } 
-    if (erts_smp_atomic32_read_nob(&dh->port_count) > 0) {
+    if (erts_atomic32_read_nob(&dh->port_count) > 0) {
 	++kill_ports;
     }
     dh->status = ERL_DE_UNLOAD;
@@ -608,23 +587,17 @@ done:
 	/* Avoid closing the driver by referencing it */
 	erts_ddll_reference_driver(dh);
 	dh->status = ERL_DE_FORCE_UNLOAD;
-#if DDLL_SMP
 	unlock_drv_list();
-#endif
 	kill_ports_driver_unloaded(dh);
-#if DDLL_SMP
 	lock_drv_list(); 
-#endif
 	erts_ddll_dereference_driver(dh);
     } 
 
-#if DDLL_SMP
     erts_ddll_reference_driver(dh);
     unlock_drv_list();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     lock_drv_list();
     erts_ddll_dereference_driver(dh);
-#endif
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
     BIF_P->flags |= F_USING_DDLL;
     if (monitor > 0) {
@@ -638,17 +611,13 @@ done:
     if (kill_ports > 1) {
 	ERTS_BIF_CHK_EXITED(BIF_P); /* May be exited by port killing */
     }
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     BIF_RET(t);
  
 soft_error:
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     hp = HAlloc(BIF_P, 3);
     t = TUPLE2(hp, am_error, soft_error_term);
     BIF_RET(t);
@@ -658,7 +627,7 @@ soft_error:
     if (name != NULL) {
 	erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
     }
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     BIF_ERROR(BIF_P, BADARG);
 }
 
@@ -697,9 +666,7 @@ BIF_RETTYPE erl_ddll_loaded_drivers_0(BIF_ALIST_0)
     int need = 3;
     Eterm res = NIL;
     erts_driver_t *drv;
-#if DDLL_SMP
     lock_drv_list();
-#endif
     for (drv = driver_list; drv; drv = drv->next) {
 	need += sys_strlen(drv->name)*2+2;
     }
@@ -712,9 +679,7 @@ BIF_RETTYPE erl_ddll_loaded_drivers_0(BIF_ALIST_0)
     }
     res = TUPLE2(hp,am_ok,res);
     /* hp += 3 */
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     BIF_RET(res);
 }
 
@@ -736,9 +701,7 @@ BIF_RETTYPE erl_ddll_info_2(BIF_ALIST_2)
     Eterm *hp;
     int i;
     Uint filter;
-#if DDLL_SMP
     int have_lock = 0;
-#endif
 
     if ((name = pick_list_or_atom(name_term)) == NULL) {
 	goto error;
@@ -748,10 +711,8 @@ BIF_RETTYPE erl_ddll_info_2(BIF_ALIST_2)
 	goto error;
     }
 
-#if DDLL_SMP 
     lock_drv_list();
     have_lock = 1;
-#endif
     if ((drv = lookup_driver(name)) == NULL) {
 	goto error;
     }
@@ -781,7 +742,7 @@ BIF_RETTYPE erl_ddll_info_2(BIF_ALIST_2)
 	} else if (drv->handle->status == ERL_DE_PERMANENT) {
 	    res = am_permanent;
 	} else {
-	    res = make_small(erts_smp_atomic32_read_nob(&drv->handle->port_count));
+	    res = make_small(erts_atomic32_read_nob(&drv->handle->port_count));
 	}
 	goto done;
     case am_linked_in_driver:
@@ -827,9 +788,7 @@ BIF_RETTYPE erl_ddll_info_2(BIF_ALIST_2)
 	hp += 2;
     }    
  done:    
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     if (pei)
 	erts_free(ERTS_ALC_T_DDLL_TMP_BUF, pei);
     erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
@@ -838,11 +797,9 @@ BIF_RETTYPE erl_ddll_info_2(BIF_ALIST_2)
     if (name != NULL) {
 	erts_free(ERTS_ALC_T_DDLL_TMP_BUF, (void *) name);
     }
-#if DDLL_SMP
     if (have_lock) {
 	unlock_drv_list();
     }
-#endif
     BIF_ERROR(p,BADARG);
 }
 
@@ -899,13 +856,9 @@ BIF_RETTYPE erl_ddll_format_error_int_1(BIF_ALIST_1)
 	if (errdesc_to_code(code_term,&errint) != 0) {
 	    goto error;
 	}
-#if DDLL_SMP 
 	lock_drv_list();
-#endif
 	errstring = erts_ddll_error(errint);
-#if DDLL_SMP
 	unlock_drv_list();
-#endif
 	break;
     }
     if (errstring == NULL) {
@@ -968,7 +921,7 @@ Eterm erts_ddll_monitor_driver(Process *p,
 void erts_ddll_remove_monitor(Process *p, Eterm ref, ErtsProcLocks plocks)
 { 
     erts_driver_t *drv;
-    erts_smp_proc_unlock(p, plocks);
+    erts_proc_unlock(p, plocks);
     lock_drv_list();
     drv = driver_list;
     while (drv != NULL) {
@@ -993,7 +946,7 @@ void erts_ddll_remove_monitor(Process *p, Eterm ref, ErtsProcLocks plocks)
     }
  done:
     unlock_drv_list();
-    erts_smp_proc_lock(p, plocks);
+    erts_proc_lock(p, plocks);
 }
 
 /* 
@@ -1002,7 +955,7 @@ void erts_ddll_remove_monitor(Process *p, Eterm ref, ErtsProcLocks plocks)
 void erts_ddll_proc_dead(Process *p, ErtsProcLocks plocks) 
 {
     erts_driver_t *drv;
-    erts_smp_proc_unlock(p, plocks);
+    erts_proc_unlock(p, plocks);
     lock_drv_list();
     drv = driver_list;
     while (drv != NULL) {
@@ -1040,18 +993,14 @@ void erts_ddll_proc_dead(Process *p, ErtsProcLocks plocks)
 		dh->status = ERL_DE_UNLOAD;
 	    }
 	    if (!left
-		&& erts_smp_atomic32_read_nob(&drv->handle->port_count) > 0) {
+		&& erts_atomic32_read_nob(&drv->handle->port_count) > 0) {
 		if (kill_ports) {
 		    DE_Handle *dh = drv->handle;
 		    erts_ddll_reference_driver(dh);
 		    dh->status = ERL_DE_FORCE_UNLOAD;
-#if DDLL_SMP
 		    unlock_drv_list();
-#endif
 		    kill_ports_driver_unloaded(dh);
-#if DDLL_SMP
 		    lock_drv_list(); /* Needed for future list operations */
-#endif
 		    drv = drv->next; /* before allowing destruction */
 		    erts_ddll_dereference_driver(dh);
 		} else {
@@ -1065,7 +1014,7 @@ void erts_ddll_proc_dead(Process *p, ErtsProcLocks plocks)
 	}
     }
     unlock_drv_list();
-    erts_smp_proc_lock(p, plocks);
+    erts_proc_lock(p, plocks);
 }
 void erts_ddll_lock_driver(DE_Handle *dh, char *name)
 {
@@ -1093,41 +1042,41 @@ void erts_ddll_lock_driver(DE_Handle *dh, char *name)
 void erts_ddll_increment_port_count(DE_Handle *dh) 
 {
     assert_drv_list_locked();
-    erts_smp_atomic32_inc_nob(&dh->port_count);
+    erts_atomic32_inc_nob(&dh->port_count);
 }
 
 void erts_ddll_decrement_port_count(DE_Handle *dh)
 {
     assert_drv_list_locked();
 #ifdef DEBUG
-    ASSERT(erts_smp_atomic32_dec_read_nob(&dh->port_count) >= 0);
+    ASSERT(erts_atomic32_dec_read_nob(&dh->port_count) >= 0);
 #else
-    erts_smp_atomic32_dec_nob(&dh->port_count);
+    erts_atomic32_dec_nob(&dh->port_count);
 #endif
 }
 
 static void first_ddll_reference(DE_Handle *dh) 
 {
     assert_drv_list_rwlocked();
-    erts_smp_refc_init(&(dh->refc),1);
+    erts_refc_init(&(dh->refc),1);
 }
 
 void erts_ddll_reference_driver(DE_Handle *dh)
 {
     assert_drv_list_locked();
-    if (erts_smp_refc_inctest(&(dh->refc),1) == 1) {
-	erts_smp_refc_inc(&(dh->refc),2); /* add a reference for the scheduled operation */
+    if (erts_refc_inctest(&(dh->refc),1) == 1) {
+	erts_refc_inc(&(dh->refc),2); /* add a reference for the scheduled operation */
     }
 }
 
 void erts_ddll_reference_referenced_driver(DE_Handle *dh)
 {
-    erts_smp_refc_inc(&(dh->refc),2);
+    erts_refc_inc(&(dh->refc),2);
 }
 
 void erts_ddll_dereference_driver(DE_Handle *dh)
 {
-    if (erts_smp_refc_dectest(&(dh->refc),0) == 0) {
+    if (erts_refc_dectest(&(dh->refc),0) == 0) {
 	/* No lock here, but if the driver is referenced again,
 	   the scheduled deletion is added as a reference too, see above */
 	erts_schedule_misc_op(ddll_no_more_references, (void *) dh);
@@ -1150,11 +1099,11 @@ static void restore_process_references(DE_Handle *dh)
 {
     DE_ProcEntry *p;
     assert_drv_list_rwlocked();
-    ASSERT(erts_smp_refc_read(&(dh->refc),0) == 0);
+    ASSERT(erts_refc_read(&(dh->refc),0) == 0);
     for(p  = dh->procs;p != NULL; p = p->next) {
 	if (p->awaiting_status == ERL_DE_PROC_LOADED) {
 	    ASSERT(p->flags & ERL_DE_FL_DEREFERENCED);
-	    erts_smp_refc_inc(&(dh->refc),1);
+	    erts_refc_inc(&(dh->refc),1);
 	    p->flags &= ~ERL_DE_FL_DEREFERENCED;
 	}
     }
@@ -1176,9 +1125,9 @@ static void ddll_no_more_references(void *vdh)
 
     lock_drv_list();
 
-    x = erts_smp_refc_read(&(dh->refc),0);
+    x = erts_refc_read(&(dh->refc),0);
     if (x > 0) {
-	x = erts_smp_refc_dectest(&(dh->refc),0); /* delete the reference added for me */
+	x = erts_refc_dectest(&(dh->refc),0); /* delete the reference added for me */
     }
 
 
@@ -1281,10 +1230,8 @@ static Eterm notify_when_loaded(Process *p, Eterm name_term, char *name, ErtsPro
     Eterm immediate_type = NIL;
     erts_driver_t *drv;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & plocks);
-#if DDLL_SMP 
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & plocks);
     lock_drv_list();
-#endif
     if ((drv = lookup_driver(name)) == NULL) {
 	immediate_tag = am_unloaded;
 	immediate_type = am_DOWN;
@@ -1314,20 +1261,14 @@ static Eterm notify_when_loaded(Process *p, Eterm name_term, char *name, ErtsPro
     }
     p->flags |= F_USING_DDLL;
     r = add_monitor(p, drv->handle, ERL_DE_PROC_AWAIT_LOAD);
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     BIF_RET(r);
  immediate:
     r =  erts_make_ref(p);
-#if DDLL_SMP 
-    erts_smp_proc_unlock(p, plocks);
-#endif
+    erts_proc_unlock(p, plocks);
     notify_proc(p, r, name_term, immediate_type, immediate_tag, 0);
-#if DDLL_SMP
     unlock_drv_list();
-    erts_smp_proc_lock(p, plocks);
-#endif
+    erts_proc_lock(p, plocks);
     BIF_RET(r);
 }
 
@@ -1338,10 +1279,8 @@ static Eterm notify_when_unloaded(Process *p, Eterm name_term, char *name, ErtsP
     Eterm immediate_type = NIL;
     erts_driver_t *drv;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & plocks);
-#if DDLL_SMP 
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & plocks);
     lock_drv_list();
-#endif
     if ((drv = lookup_driver(name)) == NULL) {
 	immediate_tag = am_unloaded;
 	immediate_type = am_DOWN;
@@ -1355,20 +1294,14 @@ static Eterm notify_when_unloaded(Process *p, Eterm name_term, char *name, ErtsP
 
     p->flags |= F_USING_DDLL;
     r = add_monitor(p, drv->handle, flag);
-#if DDLL_SMP
     unlock_drv_list();
-#endif
     BIF_RET(r);
  immediate:
     r =  erts_make_ref(p);
-#if DDLL_SMP 
-    erts_smp_proc_unlock(p, plocks);
-#endif
+    erts_proc_unlock(p, plocks);
     notify_proc(p, r, name_term, immediate_type, immediate_tag, 0);
-#if DDLL_SMP
     unlock_drv_list();
-    erts_smp_proc_lock(p, plocks);
-#endif
+    erts_proc_lock(p, plocks);
     BIF_RET(r);
 }
 
@@ -1479,7 +1412,7 @@ static Eterm copy_ref(Eterm ref, Eterm *hp)
     ErtsORefThing *ptr;
     ASSERT(is_internal_ordinary_ref(ref));
     ptr = ordinary_ref_thing_ptr(ref);
-    memcpy(hp, ptr, sizeof(ErtsORefThing));
+    sys_memcpy(hp, ptr, sizeof(ErtsORefThing));
     return (make_internal_ref(hp));
 }
 
@@ -1521,9 +1454,9 @@ static void set_driver_reloading(DE_Handle *dh, Process *proc, char *path, char
     dh->procs = p;
     dh->status = ERL_DE_RELOAD;
     dh->reload_full_path = erts_alloc(ERTS_ALC_T_DDLL_HANDLE, sys_strlen(path) + 1);
-    strcpy(dh->reload_full_path,path);
+    sys_strcpy(dh->reload_full_path,path);
     dh->reload_driver_name = erts_alloc(ERTS_ALC_T_DDLL_HANDLE, sys_strlen(name) + 1);
-    strcpy(dh->reload_driver_name,name);
+    sys_strcpy(dh->reload_driver_name,name);
     dh->reload_flags = flags;
 }
 
@@ -1568,18 +1501,19 @@ static int do_load_driver_entry(DE_Handle *dh, char *path, char *name)
 	goto error;
     }
 
-    if (strcmp(name, dp->driver_name) != 0) {
+    if (sys_strcmp(name, dp->driver_name) != 0) {
 	res = ERL_DE_LOAD_ERROR_BAD_NAME;
 	goto error;
     }
-    erts_smp_atomic_init_nob(&(dh->refc), (erts_aint_t) 0);
-    erts_smp_atomic32_init_nob(&dh->port_count, 0);
+
+    erts_atomic_init_nob(&(dh->refc), (erts_aint_t) 0);
+    erts_atomic32_init_nob(&dh->port_count, 0);
     dh->full_path = erts_alloc(ERTS_ALC_T_DDLL_HANDLE, sys_strlen(path) + 1);
     sys_strcpy(dh->full_path, path);
     dh->flags = 0;
     dh->status = ERL_DE_OK;
 
-    if (erts_add_driver_entry(dp, dh, 1) != 0 /* io.c */) { 
+    if (erts_add_driver_entry(dp, dh, 1, 1) != 0 /* io.c */) {
 	/*
 	 * The init in the driver struct did not return 0 
 	 */
@@ -1644,8 +1578,8 @@ static int load_driver_entry(DE_Handle **dhp, char *path, char *name)
 
     dh->handle = NULL;
     dh->procs = NULL;
-    erts_smp_atomic32_init_nob(&dh->port_count, 0);
-    erts_smp_refc_init(&(dh->refc), (erts_aint_t) 0);
+    erts_atomic32_init_nob(&dh->port_count, 0);
+    erts_refc_init(&(dh->refc), (erts_aint_t) 0);
     dh->status = -1;
     dh->reload_full_path = NULL;
     dh->reload_driver_name = NULL;
@@ -1683,7 +1617,7 @@ static int reload_driver_entry(DE_Handle *dh)
     dh->reload_full_path = NULL;
     dh->reload_driver_name = NULL;
 
-    ASSERT(erts_smp_refc_read(&(dh->refc),0) == 0);
+    ASSERT(erts_refc_read(&(dh->refc),0) == 0);
     ASSERT(dh->full_path != NULL);
     erts_free(ERTS_ALC_T_DDLL_HANDLE, (void *) dh->full_path);
     dh->full_path = NULL;
@@ -1714,7 +1648,7 @@ static void notify_proc(Process *proc, Eterm ref, Eterm driver_name, Eterm type,
     ErtsMessage *mp;
     ErtsProcLocks rp_locks = 0;
     ErlOffHeap *ohp;
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     assert_drv_list_rwlocked();
     if (errcode != 0) {
@@ -1740,8 +1674,8 @@ static void notify_proc(Process *proc, Eterm ref, Eterm driver_name, Eterm type,
 	mess = TUPLE5(hp,type,r,am_driver,driver_name,tag);
     }
     erts_queue_message(proc, rp_locks, mp, mess, am_system);
-    erts_smp_proc_unlock(proc, rp_locks);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    erts_proc_unlock(proc, rp_locks);
+    ERTS_CHK_NO_PROC_LOCKS;
 }
 
 static void notify_all(DE_Handle *dh, char *name, Uint awaiting, Eterm type, Eterm tag)
@@ -1813,7 +1747,7 @@ static Eterm build_load_error(Process *p, int code)
 {
     int need = load_error_need(code);
     Eterm *hp = NULL;
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(p));
     if (need) {
 	hp = HAlloc(p,need);
     }
@@ -1866,7 +1800,7 @@ static char *pick_list_or_atom(Eterm name_term)
 	    goto error;
 	}
 	name = erts_alloc(ERTS_ALC_T_DDLL_TMP_BUF, ap->len + 1);
-	memcpy(name,ap->name,ap->len);
+	sys_memcpy(name,ap->name,ap->len);
 	name[ap->len] = '\0';
     } else {
 	if (erts_iolist_size(name_term, &name_len)) {
@@ -1937,7 +1871,7 @@ static erts_driver_t *lookup_driver(char *name)
 {
     erts_driver_t *drv;
     assert_drv_list_locked();
-    for (drv = driver_list; drv != NULL && strcmp(drv->name, name); drv = drv->next)
+    for (drv = driver_list; drv != NULL && sys_strcmp(drv->name, name); drv = drv->next)
 	;
     return drv;
 }
diff --git a/erts/emulator/beam/erl_bif_info.c b/erts/emulator/beam/erl_bif_info.c
index 4050fb6146..5789fa8e71 100644
--- a/erts/emulator/beam/erl_bif_info.c
+++ b/erts/emulator/beam/erl_bif_info.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1999-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1999-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -38,7 +38,7 @@
 #include "erl_message.h"
 #include "erl_binary.h"
 #include "erl_db.h"
-#include "erl_instrument.h"
+#include "erl_mtrace.h"
 #include "dist.h"
 #include "erl_gc.h"
 #include "erl_cpu_topology.h"
@@ -46,9 +46,12 @@
 #include "erl_thr_progress.h"
 #include "erl_bif_unique.h"
 #include "erl_map.h"
+#include "erl_check_io.h"
 #define ERTS_PTAB_WANT_DEBUG_FUNCS__
 #include "erl_ptab.h"
 #include "erl_time.h"
+#include "erl_proc_sig_queue.h"
+#include "erl_alloc_util.h"
 #ifdef HIPE
 #include "hipe_arch.h"
 #endif
@@ -71,6 +74,9 @@ static Export *gather_msacc_res_trap;
 static Export *gather_gc_info_res_trap;
 static Export *gather_system_check_res_trap;
 
+static Export *is_process_alive_trap;
+
+
 #define DECL_AM(S) Eterm AM_ ## S = am_atom_put(#S, sizeof(#S) - 1)
 
 static char otp_version[] = ERLANG_OTP_VERSION;
@@ -88,24 +94,15 @@ static char erts_system_version[] = ("Erlang/OTP " ERLANG_OTP_RELEASE
 #ifdef ARCH_64
 				     " [64-bit]"
 #endif
-#ifdef ERTS_SMP
 				     " [smp:%beu:%beu]"
-#endif
-#ifdef USE_THREADS
-#if defined(ERTS_DIRTY_SCHEDULERS) && defined(ERTS_SMP)
 				     " [ds:%beu:%beu:%beu]"
-#endif
 #if defined(ERTS_DIRTY_SCHEDULERS_TEST)
 				     " [dirty-schedulers-TEST]"
 #endif
 				     " [async-threads:%d]"
-#endif
 #ifdef HIPE
 				     " [hipe]"
 #endif	
-#ifdef ERTS_ENABLE_KERNEL_POLL
-				     " [kernel-poll:%s]"
-#endif	
 #ifdef ET_DEBUG
 #if ET_DEBUG
 				     " [type-assertions]"
@@ -159,8 +156,10 @@ static Eterm os_type_tuple;
 static Eterm os_version_tuple;
 
 static Eterm
-current_function(Process* p, Process* rp, Eterm** hpp, int full_info);
-static Eterm current_stacktrace(Process* p, Process* rp, Eterm** hpp);
+current_function(Process* p, ErtsHeapFactory *hfact, Process* rp,
+                 int full_info, Uint reserve_size, int flags);
+static Eterm current_stacktrace(ErtsHeapFactory *hfact, Process* rp,
+                                Uint reserve_size);
 
 static Eterm
 bld_bin_list(Uint **hpp, Uint *szp, ErlOffHeap* oh)
@@ -219,21 +218,27 @@ bld_magic_ref_bin_list(Uint **hpp, Uint *szp, ErlOffHeap* oh)
   make_monitor_list:
   returns a list of records..
   -record(erl_monitor, {
-            type, % MON_ORIGIN or MON_TARGET (1 or 3)
-	    ref,
+            type, % process | port | time_offset | dist_process | resource
+                  % | node | nodes | suspend
+            dir, % origin | target
+	    ref, % reference or []
 	    pid, % Process or nodename
-	    name % registered name or []
+	    extra % registered name, integer or []
           }).
 */
 
 static void do_calc_mon_size(ErtsMonitor *mon, void *vpsz)
 {
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
     Uint *psz = vpsz;
-    *psz += NC_HEAP_SIZE(mon->ref);
-    *psz += (mon->type == MON_NIF_TARGET ?
-	     erts_resource_ref_size(mon->u.resource) :
-	     (is_immed(mon->u.pid) ? 0 : NC_HEAP_SIZE(mon->u.pid)));
-    *psz += 8; /* CONS + 5-tuple */ 
+    *psz += is_immed(mdp->ref) ? 0 : NC_HEAP_SIZE(mdp->ref);
+
+    if (mon->type == ERTS_MON_TYPE_RESOURCE && erts_monitor_is_target(mon))
+        *psz += erts_resource_ref_size(mon->other.ptr);
+    else
+        *psz += is_immed(mon->other.item) ? 0 : NC_HEAP_SIZE(mon->other.item);
+
+    *psz += 9; /* CONS + 6-tuple */ 
 }
 
 typedef struct {
@@ -245,35 +250,97 @@ typedef struct {
 
 static void do_make_one_mon_element(ErtsMonitor *mon, void * vpmlc)
 {
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
     MonListContext *pmlc = vpmlc;
-    Eterm tup;
-    Eterm r = STORE_NC(&(pmlc->hp), &MSO(pmlc->p), mon->ref);
-    Eterm p = (mon->type == MON_NIF_TARGET ?
-	       erts_bld_resource_ref(&(pmlc->hp), &MSO(pmlc->p), mon->u.resource)
-	       : (is_immed(mon->u.pid) ? mon->u.pid
-		  : STORE_NC(&(pmlc->hp), &MSO(pmlc->p), mon->u.pid)));
-    tup = TUPLE5(pmlc->hp, pmlc->tag, make_small(mon->type), r, p, mon->name);
-    pmlc->hp += 6;
+    Eterm tup, t, d, r, p, x;
+
+    r = is_immed(mdp->ref) ? mdp->ref : STORE_NC(&(pmlc->hp), &MSO(pmlc->p), mdp->ref);
+    if (mon->type == ERTS_MON_TYPE_RESOURCE && erts_monitor_is_target(mon))
+        p = erts_bld_resource_ref(&(pmlc->hp), &MSO(pmlc->p), mon->other.ptr);
+    else
+        p = (is_immed(mon->other.item)
+             ? mon->other.item
+             : STORE_NC(&(pmlc->hp), &MSO(pmlc->p), mon->other.item));
+
+    if (mon->flags & ERTS_ML_FLG_NAME)
+        x = ((ErtsMonitorDataExtended *) mdp)->u.name;
+    else if (erts_monitor_is_target(mon))
+        x = NIL;
+    else if (mon->type == ERTS_MON_TYPE_NODE || mon->type == ERTS_MON_TYPE_NODES)
+        x = make_small(((ErtsMonitorDataExtended *) mdp)->u.refc);
+    else
+        x = NIL;
+
+    switch (mon->type) {
+    case ERTS_MON_TYPE_PROC:
+        t = am_process;
+        break;
+    case ERTS_MON_TYPE_PORT:
+        t = am_port;
+        break;
+    case ERTS_MON_TYPE_TIME_OFFSET:
+        t = am_time_offset;
+        break;
+    case ERTS_MON_TYPE_DIST_PROC: {
+        ERTS_DECL_AM(dist_process);
+        t = AM_dist_process;
+        break;
+    }
+    case ERTS_MON_TYPE_RESOURCE: {
+        ERTS_DECL_AM(resource);
+        t = AM_resource;
+        break;
+    }
+    case ERTS_MON_TYPE_NODE:
+        t = am_node;
+        break;
+    case ERTS_MON_TYPE_NODES: {
+        ERTS_DECL_AM(nodes);
+        t = AM_nodes;
+        break;
+    }
+    case ERTS_MON_TYPE_SUSPEND:
+        t = am_suspend;
+        break;
+    default:
+        ERTS_INTERNAL_ERROR("Unknown monitor type");
+        t = am_error;
+        break;
+    }
+    if (erts_monitor_is_target(mon)) {
+        ERTS_DECL_AM(target);
+        d = AM_target;
+    }
+    else {
+        ERTS_DECL_AM(origin);
+        d = AM_origin;
+    }
+    tup = TUPLE6(pmlc->hp, pmlc->tag, t, d, r, p, x);
+    pmlc->hp += 7;
     pmlc->res = CONS(pmlc->hp, tup, pmlc->res);
     pmlc->hp += 2;
 }
 
 static Eterm 
-make_monitor_list(Process *p, ErtsMonitor *root)
+make_monitor_list(Process *p, int tree, ErtsMonitor *root, Eterm tail)
 {
     DECL_AM(erl_monitor);
     Uint sz = 0;
     MonListContext mlc;
+    void (*foreach)(ErtsMonitor *,
+                    void (*)(ErtsMonitor *, void *),
+                    void *);
 
-    erts_doforall_monitors(root, &do_calc_mon_size, &sz);
-    if (sz == 0) {
-	return NIL;
-    }
+    foreach = tree ? erts_monitor_tree_foreach : erts_monitor_list_foreach;
+
+    (*foreach)(root, do_calc_mon_size, &sz);
+    if (sz == 0)
+	return tail;
     mlc.p = p;
     mlc.hp = HAlloc(p,sz);
-    mlc.res = NIL;
+    mlc.res = tail;
     mlc.tag = AM_erl_monitor;
-    erts_doforall_monitors(root, &do_make_one_mon_element, &mlc);
+    (*foreach)(root, do_make_one_mon_element, &mlc);
     return mlc.res;
 }
 
@@ -281,20 +348,22 @@ make_monitor_list(Process *p, ErtsMonitor *root)
   make_link_list:
   returns a list of records..
   -record(erl_link, {
-            type, % LINK_NODE or LINK_PID (1 or 3)
-	    pid, % Process or nodename
-	    targets % List of erl_link's or nil
+            type, % process | port | dist_process
+	    pid, % Process or port
+            id % (address)
           }).
 */
 
-static void do_calc_lnk_size(ErtsLink *lnk, void *vpsz)
+static void calc_lnk_size(ErtsLink *lnk, void *vpsz)
 {
     Uint *psz = vpsz;
-    *psz += is_immed(lnk->pid) ? 0 : NC_HEAP_SIZE(lnk->pid);
-    if (lnk->type != LINK_NODE && ERTS_LINK_ROOT(lnk) != NULL) { 
-	/* Node links use this pointer as ref counter... */
-	erts_doforall_links(ERTS_LINK_ROOT(lnk),&do_calc_lnk_size,vpsz);
-    }
+    Uint sz = 0;
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+
+    (void) erts_bld_uword(NULL, &sz, (UWord) ldp);
+
+    *psz += sz;
+    *psz += is_immed(lnk->other.item) ? 0 : size_object(lnk->other.item);
     *psz += 7; /* CONS + 4-tuple */ 
 }
 
@@ -305,37 +374,58 @@ typedef struct {
     Eterm tag;
 } LnkListContext;
 
-static void do_make_one_lnk_element(ErtsLink *lnk, void * vpllc)
+static void make_one_lnk_element(ErtsLink *lnk, void * vpllc)
 {
     LnkListContext *pllc = vpllc;
-    Eterm tup;
-    Eterm old_res, targets = NIL;
-    Eterm p = (is_immed(lnk->pid)
-	       ? lnk->pid
-	       : STORE_NC(&(pllc->hp), &MSO(pllc->p), lnk->pid));
-    if (lnk->type == LINK_NODE) {
-	targets = make_small(ERTS_LINK_REFC(lnk));
-    } else if (ERTS_LINK_ROOT(lnk) != NULL) {
-	old_res = pllc->res;
-	pllc->res = NIL;
-	erts_doforall_links(ERTS_LINK_ROOT(lnk),&do_make_one_lnk_element, vpllc);
-	targets = pllc->res;
-	pllc->res = old_res;
-    }
-    tup = TUPLE4(pllc->hp, pllc->tag, make_small(lnk->type), p, targets);
+    Eterm tup, t, pid, id;
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+
+    id = erts_bld_uword(&pllc->hp, NULL, (UWord) ldp);
+
+    if (is_immed(lnk->other.item))
+        pid = lnk->other.item;
+    else {
+        Uint sz = size_object(lnk->other.item);
+        pid = copy_struct(lnk->other.item, sz, &(pllc->hp), &MSO(pllc->p));
+    }
+
+    switch (lnk->type) {
+    case ERTS_LNK_TYPE_PROC:
+        t = am_process;
+        break;
+    case ERTS_LNK_TYPE_PORT:
+        t = am_port;
+        break;
+    case ERTS_LNK_TYPE_DIST_PROC: {
+        ERTS_DECL_AM(dist_process);
+        t = AM_dist_process;
+        break;
+    }
+    default:
+        ERTS_INTERNAL_ERROR("Unkown link type");
+        t = am_undefined;
+        break;
+    }
+
+    tup = TUPLE4(pllc->hp, pllc->tag, t, pid, id);
     pllc->hp += 5;
     pllc->res = CONS(pllc->hp, tup, pllc->res);
     pllc->hp += 2;
 }
 
 static Eterm 
-make_link_list(Process *p, ErtsLink *root, Eterm tail)
+make_link_list(Process *p, int tree, ErtsLink *root, Eterm tail)
 {
     DECL_AM(erl_link);
     Uint sz = 0;
     LnkListContext llc;
+    void (*foreach)(ErtsLink *,
+                    void (*)(ErtsLink *, void *),
+                    void *);
 
-    erts_doforall_links(root, &do_calc_lnk_size, &sz);
+    foreach = tree ? erts_link_tree_foreach : erts_link_list_foreach;
+
+    (*foreach)(root, calc_lnk_size, (void *) &sz);
     if (sz == 0) {
 	return tail;
     }
@@ -343,7 +433,7 @@ make_link_list(Process *p, ErtsLink *root, Eterm tail)
     llc.hp = HAlloc(p,sz);
     llc.res = tail;
     llc.tag = AM_erl_link;
-    erts_doforall_links(root, &do_make_one_lnk_element, &llc);
+    (*foreach)(root, make_one_lnk_element, (void *) &llc);
     return llc.res;
 }
 
@@ -354,14 +444,12 @@ erts_print_system_version(fmtfn_t to, void *arg, Process *c_p)
     char *rc_str = "";
     char rc_buf[100];
     char *ov = otp_version;
-#ifdef ERTS_SMP
     Uint total, online, active;
     Uint dirty_cpu, dirty_cpu_onln, dirty_io;
 
     erts_schedulers_state(&total, &online, &active,
 			  &dirty_cpu, &dirty_cpu_onln, NULL,
 			  &dirty_io, NULL);
-#endif
     for (i = 0; i < sizeof(otp_version)-4; i++) {
 	if (ov[i] == '-' && ov[i+1] == 'r' && ov[i+2] == 'c')
 	    rc = atoi(&ov[i+3]);
@@ -376,18 +464,9 @@ erts_print_system_version(fmtfn_t to, void *arg, Process *c_p)
     }
     return erts_print(to, arg, erts_system_version,
 		      rc_str
-#ifdef ERTS_SMP
 		      , total, online
-#ifdef ERTS_DIRTY_SCHEDULERS
 		      , dirty_cpu, dirty_cpu_onln, dirty_io
-#endif
-#endif
-#ifdef USE_THREADS
 		      , erts_async_max_threads
-#endif
-#ifdef ERTS_ENABLE_KERNEL_POLL
-		      , erts_use_kernel_poll ? "true" : "false"
-#endif
 	);
 }
 
@@ -400,6 +479,8 @@ typedef struct {
 	Eterm term;
 	ErtsResource* resource;
     }entity;
+    int named;
+    Uint16 type;
     Eterm node;
     /* pid is actual target being monitored, no matter pid/port or name */
     Eterm pid;
@@ -442,86 +523,107 @@ static void collect_one_link(ErtsLink *lnk, void *vmicp)
 {
     MonitorInfoCollection *micp = vmicp;
     EXTEND_MONITOR_INFOS(micp);
-    if (!(lnk->type == LINK_PID)) {
-	return;
-    }
-    micp->mi[micp->mi_i].entity.term = lnk->pid;
-    micp->sz += 2 + NC_HEAP_SIZE(lnk->pid);
+    micp->mi[micp->mi_i].entity.term = lnk->other.item;
+    micp->sz += 2 + NC_HEAP_SIZE(lnk->other.item);
     micp->mi_i++;
 } 
 
 static void collect_one_origin_monitor(ErtsMonitor *mon, void *vmicp)
 {
-    MonitorInfoCollection *micp = vmicp;
+    if (erts_monitor_is_origin(mon)) {
+        MonitorInfoCollection *micp = vmicp;
  
-    if (mon->type != MON_ORIGIN) {
-	return;
-    }
-    EXTEND_MONITOR_INFOS(micp);
-    if (is_atom(mon->u.pid)) { /* external by name */
-	micp->mi[micp->mi_i].entity.term = mon->name;
-        micp->mi[micp->mi_i].node = mon->u.pid;
-        micp->sz += 3; /* need one 2-tuple */
-    } else if (is_external_pid(mon->u.pid)) { /* external by pid */
-	micp->mi[micp->mi_i].entity.term = mon->u.pid;
-        micp->mi[micp->mi_i].node = NIL;
-        micp->sz += NC_HEAP_SIZE(mon->u.pid);
-    } else if (!is_nil(mon->name)) { /* internal by name */
-	micp->mi[micp->mi_i].entity.term = mon->name;
-        micp->mi[micp->mi_i].node = erts_this_dist_entry->sysname;
-        micp->sz += 3; /* need one 2-tuple */
-    } else { /* internal by pid */
-	micp->mi[micp->mi_i].entity.term = mon->u.pid;
-        micp->mi[micp->mi_i].node = NIL;
-	/* no additional heap space needed */
-    }
-
-    /* have always pid at hand, to assist with figuring out if its a port or
-     * a process, when we monitored by name and process_info is requested.
-     * See: erl_bif_info.c:process_info_aux section for am_monitors */
-    micp->mi[micp->mi_i].pid = mon->u.pid;
+        EXTEND_MONITOR_INFOS(micp);
+
+        micp->mi[micp->mi_i].type = mon->type;
+
+        switch (mon->type) {
+        case ERTS_MON_TYPE_PROC:
+        case ERTS_MON_TYPE_PORT:
+        case ERTS_MON_TYPE_DIST_PROC:
+        case ERTS_MON_TYPE_TIME_OFFSET:
+            if (!(mon->flags & ERTS_ML_FLG_NAME)) {
+                micp->mi[micp->mi_i].named = 0;
+                micp->mi[micp->mi_i].entity.term = mon->other.item;
+                micp->mi[micp->mi_i].node = NIL;
+                if (is_not_atom(mon->other.item))
+                    micp->sz += NC_HEAP_SIZE(mon->other.item);
+            }
+            else {
+                ErtsMonitorDataExtended *mdep;
+                micp->mi[micp->mi_i].named = !0;
+                mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+                micp->mi[micp->mi_i].entity.term = mdep->u.name;
+                if (mdep->dist)
+                    micp->mi[micp->mi_i].node = mdep->dist->nodename;
+                else
+                    micp->mi[micp->mi_i].node = erts_this_dist_entry->sysname;
+                micp->sz += 3; /* need one 2-tuple */
+            }
 
-    micp->mi_i++;
-    micp->sz += 2 + 3; /* For a cons cell and a 2-tuple */
+            /* have always pid at hand, to assist with figuring out if its a port or
+             * a process, when we monitored by name and process_info is requested.
+             * See: erl_bif_info.c:process_info_aux section for am_monitors */
+            micp->mi[micp->mi_i].pid = mon->other.item;
+
+            micp->mi_i++;
+            micp->sz += 2 + 3; /* For a cons cell and a 2-tuple */
+            break;
+        default:
+            break;
+        }
+    }
 }
 
 static void collect_one_target_monitor(ErtsMonitor *mon, void *vmicp)
 {
     MonitorInfoCollection *micp = vmicp;
  
-    if (mon->type != MON_TARGET && mon->type != MON_NIF_TARGET) {
-        return;
-    }
+    if (erts_monitor_is_target(mon)) {
 
-    EXTEND_MONITOR_INFOS(micp);
+        EXTEND_MONITOR_INFOS(micp);
   
+        micp->mi[micp->mi_i].type = mon->type;
+        micp->mi[micp->mi_i].named = !!(mon->flags & ERTS_ML_FLG_NAME);
+        switch (mon->type) {
+
+        case ERTS_MON_TYPE_PROC:
+        case ERTS_MON_TYPE_PORT:
+        case ERTS_MON_TYPE_DIST_PROC:
+
+            micp->mi[micp->mi_i].entity.term = mon->other.item;
+            micp->mi[micp->mi_i].node = NIL;
+            micp->sz += NC_HEAP_SIZE(mon->other.item);
+
+            micp->sz += 2; /* cons */;
+            micp->mi_i++;
+            break;
+
+        case ERTS_MON_TYPE_RESOURCE:
+
+            micp->mi[micp->mi_i].entity.resource = mon->other.ptr;
+            micp->mi[micp->mi_i].node = NIL;
+            micp->sz += erts_resource_ref_size(mon->other.ptr);
+
+            micp->sz += 2; /* cons */;
+            micp->mi_i++;
+            break;
+
+        default:
+            break;
+        }
 
-    if (mon->type == MON_NIF_TARGET) {
-	micp->mi[micp->mi_i].entity.resource = mon->u.resource;
-        micp->mi[micp->mi_i].node = make_small(MON_NIF_TARGET);
-        micp->sz += erts_resource_ref_size(mon->u.resource);
-    }
-    else {
-	micp->mi[micp->mi_i].entity.term = mon->u.pid;
-        micp->mi[micp->mi_i].node = NIL;
-        micp->sz += NC_HEAP_SIZE(mon->u.pid);
     }
-    micp->sz += 2; /* cons */;
-    micp->mi_i++;
 }
 
 typedef struct {
-    Process *c_p;
-    ErtsProcLocks c_p_locks;
-    ErtsSuspendMonitor **smi;
+    ErtsMonitorSuspend **smi;
     Uint smi_i;
     Uint smi_max;
-    int sz;
+    Uint sz;
 } ErtsSuspendMonitorInfoCollection;
 
-#define ERTS_INIT_SUSPEND_MONITOR_INFOS(SMIC, CP, CPL) do {		\
-    (SMIC).c_p = (CP);							\
-    (SMIC).c_p_locks = (CPL);						\
+#define ERTS_INIT_SUSPEND_MONITOR_INFOS(SMIC) do {		        \
     (SMIC).smi = NULL;							\
     (SMIC).smi_i = (SMIC).smi_max = 0;					\
     (SMIC).sz = 0;                               			\
@@ -536,10 +638,10 @@ do {									\
 				       (SMICP)->smi,			\
 				       ((SMICP)->smi_max		\
 					+ ERTS_SMI_INC)			\
-				       * sizeof(ErtsSuspendMonitor *))	\
+				       * sizeof(ErtsMonitorSuspend *))	\
 			: erts_alloc(ERTS_ALC_T_TMP,			\
 				     ERTS_SMI_INC			\
-				     * sizeof(ErtsSuspendMonitor *)));	\
+				     * sizeof(ErtsMonitorSuspend *)));	\
 	(SMICP)->smi_max += ERTS_SMI_INC;				\
     }									\
  } while (0)
@@ -552,35 +654,28 @@ do {									\
  } while (0)
 
 static void
-collect_one_suspend_monitor(ErtsSuspendMonitor *smon, void *vsmicp)
+collect_one_suspend_monitor(ErtsMonitor *mon, void *vsmicp)
 {
-    ErtsSuspendMonitorInfoCollection *smicp = vsmicp;
-    Process *suspendee = erts_pid2proc(smicp->c_p,
-				       smicp->c_p_locks,
-				       smon->pid,
-				       0);
-    if (suspendee) { /* suspendee is alive */
-	Sint a, p;
-	if (smon->active) {
-	    smon->active += smon->pending;
-	    smon->pending = 0;
-	}
+    if (mon->type == ERTS_MON_TYPE_SUSPEND) {
+        Sint count;
+        erts_aint_t mstate;
+        ErtsMonitorSuspend *msp;
+        ErtsSuspendMonitorInfoCollection *smicp;
 
-	ASSERT((smon->active && !smon->pending)
-	       || (smon->pending && !smon->active));
+        msp = (ErtsMonitorSuspend *) erts_monitor_to_data(mon);
+        smicp = vsmicp;
 
 	ERTS_EXTEND_SUSPEND_MONITOR_INFOS(smicp);
 
-	smicp->smi[smicp->smi_i] = smon;
+	smicp->smi[smicp->smi_i] = msp;
 	smicp->sz += 2 /* cons */ + 4 /* 3-tuple */;
 
-	a = (Sint) smon->active;	/* quiet compiler warnings */
-	p = (Sint) smon->pending;	/* on 64-bit machines      */
+        mstate = erts_atomic_read_nob(&msp->state);
 
-	if (!IS_SSMALL(a))
-	    smicp->sz += BIG_UINT_HEAP_SIZE;
-	if (!IS_SSMALL(p))
+        count = (Sint) (mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK);
+	if (!IS_SSMALL(count))
 	    smicp->sz += BIG_UINT_HEAP_SIZE;
+
 	smicp->smi_i++;
     }
 }
@@ -589,125 +684,222 @@ collect_one_suspend_monitor(ErtsSuspendMonitor *smon, void *vsmicp)
  * process_info/[1,2]
  */
 
-#define ERTS_PI_FAIL_TYPE_BADARG		0
-#define ERTS_PI_FAIL_TYPE_YIELD			1
-#define ERTS_PI_FAIL_TYPE_AWAIT_EXIT		2
-
-static ERTS_INLINE ErtsProcLocks
-pi_locks(Eterm info)
-{
-    switch (info) {
-    case am_status:
-    case am_priority:
-    case am_trap_exit:
-	return ERTS_PROC_LOCK_STATUS;
-    case am_links:
-    case am_monitors:
-    case am_monitored_by:
-    case am_suspending:
-	return ERTS_PROC_LOCK_LINK;
-    case am_messages:
-    case am_message_queue_len:
-    case am_total_heap_size:
-	return ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_MSGQ;
-    case am_memory:
-	return ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_MSGQ;
-    default:
-	return ERTS_PROC_LOCK_MAIN;
-    }
-}
-
 /*
  * All valid process_info arguments.
  */
-static Eterm pi_args[] = {
-    am_registered_name,
-    am_current_function,
-    am_initial_call,
-    am_status,
-    am_messages,
-    am_message_queue_len,
-    am_links,
-    am_monitors,
-    am_monitored_by,
-    am_dictionary,
-    am_trap_exit,
-    am_error_handler,
-    am_heap_size,
-    am_stack_size,
-    am_memory,
-    am_garbage_collection,
-    am_group_leader,
-    am_reductions,
-    am_priority,
-    am_trace,
-    am_binary,
-    am_sequential_trace_token,
-    am_catchlevel,
-    am_backtrace,
-    am_last_calls,
-    am_total_heap_size,
-    am_suspending,
-    am_min_heap_size,
-    am_min_bin_vheap_size,
-    am_max_heap_size,
-    am_current_location,
-    am_current_stacktrace,
-    am_message_queue_data,
-    am_garbage_collection_info,
-    am_magic_ref
+
+#define ERTS_PI_IX_REGISTERED_NAME                      0
+#define ERTS_PI_IX_CURRENT_FUNCTION                     1
+#define ERTS_PI_IX_INITIAL_CALL                         2
+#define ERTS_PI_IX_STATUS                               3
+#define ERTS_PI_IX_MESSAGES                             4
+#define ERTS_PI_IX_MESSAGE_QUEUE_LEN                    5
+#define ERTS_PI_IX_LINKS                                6
+#define ERTS_PI_IX_MONITORS                             7
+#define ERTS_PI_IX_MONITORED_BY                         8
+#define ERTS_PI_IX_DICTIONARY                           9
+#define ERTS_PI_IX_TRAP_EXIT                            10
+#define ERTS_PI_IX_ERROR_HANDLER                        11
+#define ERTS_PI_IX_HEAP_SIZE                            12
+#define ERTS_PI_IX_STACK_SIZE                           13
+#define ERTS_PI_IX_MEMORY                               14
+#define ERTS_PI_IX_GARBAGE_COLLECTION                   15
+#define ERTS_PI_IX_GROUP_LEADER                         16
+#define ERTS_PI_IX_REDUCTIONS                           17
+#define ERTS_PI_IX_PRIORITY                             18
+#define ERTS_PI_IX_TRACE                                19
+#define ERTS_PI_IX_BINARY                               20
+#define ERTS_PI_IX_SEQUENTIAL_TRACE_TOKEN               21
+#define ERTS_PI_IX_CATCHLEVEL                           22
+#define ERTS_PI_IX_BACKTRACE                            23
+#define ERTS_PI_IX_LAST_CALLS                           24
+#define ERTS_PI_IX_TOTAL_HEAP_SIZE                      25
+#define ERTS_PI_IX_SUSPENDING                           26
+#define ERTS_PI_IX_MIN_HEAP_SIZE                        27
+#define ERTS_PI_IX_MIN_BIN_VHEAP_SIZE                   28
+#define ERTS_PI_IX_MAX_HEAP_SIZE                        29
+#define ERTS_PI_IX_CURRENT_LOCATION                     30
+#define ERTS_PI_IX_CURRENT_STACKTRACE                   31
+#define ERTS_PI_IX_MESSAGE_QUEUE_DATA                   32
+#define ERTS_PI_IX_GARBAGE_COLLECTION_INFO              33
+#define ERTS_PI_IX_MAGIC_REF                            34
+#define ERTS_PI_IX_FULLSWEEP_AFTER                      35
+
+#define ERTS_PI_FLAG_SINGELTON                          (1 << 0)
+#define ERTS_PI_FLAG_ALWAYS_WRAP                        (1 << 1)
+#define ERTS_PI_FLAG_WANT_MSGS                          (1 << 2)
+#define ERTS_PI_FLAG_NEED_MSGQ_LEN                      (1 << 3)
+#define ERTS_PI_FLAG_FORCE_SIG_SEND                     (1 << 4)
+#define ERTS_PI_FLAG_REQUEST_FOR_OTHER                  (1 << 5)
+
+#define ERTS_PI_UNRESERVE(RS, SZ) \
+    (ASSERT((RS) >= (SZ)), (RS) -= (SZ))
+
+
+typedef struct {
+    Eterm name;
+    Uint reserve_size;
+    int flags;
+    ErtsProcLocks locks;
+} ErtsProcessInfoArgs;
+
+static ErtsProcessInfoArgs pi_args[] = {
+    {am_registered_name, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_current_function, 4, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_initial_call, 4, 0, ERTS_PROC_LOCK_MAIN},
+    {am_status, 0, 0, 0},
+    {am_messages, 0, ERTS_PI_FLAG_WANT_MSGS|ERTS_PI_FLAG_NEED_MSGQ_LEN|ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_message_queue_len, 0, ERTS_PI_FLAG_NEED_MSGQ_LEN, ERTS_PROC_LOCK_MAIN},
+    {am_links, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_monitors, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_monitored_by, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_dictionary, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_trap_exit, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_error_handler, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_heap_size, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_stack_size, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_memory, 0, ERTS_PI_FLAG_NEED_MSGQ_LEN|ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_garbage_collection, 3+2 + 3+2 + 3+2 + 3+2 + 3+2 + ERTS_MAX_HEAP_SIZE_MAP_SZ, 0, ERTS_PROC_LOCK_MAIN},
+    {am_group_leader, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_reductions, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_priority, 0, 0, 0},
+    {am_trace, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_binary, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_sequential_trace_token, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_catchlevel, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_backtrace, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_last_calls, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_total_heap_size, 0, ERTS_PI_FLAG_NEED_MSGQ_LEN|ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_suspending, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, 0},
+    {am_min_heap_size, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_min_bin_vheap_size, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_max_heap_size, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_current_location, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_current_stacktrace, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_message_queue_data, 0, 0, ERTS_PROC_LOCK_MAIN},
+    {am_garbage_collection_info, ERTS_PROCESS_GC_INFO_MAX_SIZE, 0, ERTS_PROC_LOCK_MAIN},
+    {am_magic_ref, 0, ERTS_PI_FLAG_FORCE_SIG_SEND, ERTS_PROC_LOCK_MAIN},
+    {am_fullsweep_after, 0, 0, ERTS_PROC_LOCK_MAIN}
 };
 
-#define ERTS_PI_ARGS ((int) (sizeof(pi_args)/sizeof(Eterm)))
+#define ERTS_PI_ARGS ((int) (sizeof(pi_args)/sizeof(pi_args[0])))
+
+#ifdef DEBUG
+#  define ERTS_PI_DEF_ARR_SZ 2
+#else
+#  define ERTS_PI_DEF_ARR_SZ ERTS_PI_ARGS
+#endif
 
 static ERTS_INLINE Eterm
 pi_ix2arg(int ix)
 {
     if (ix < 0 || ERTS_PI_ARGS <= ix)
 	return am_undefined;
-    return pi_args[ix];
+    return pi_args[ix].name;
+}
+
+static ERTS_INLINE int
+pi_ix2flags(int ix)
+{
+    if (ix < 0 || ERTS_PI_ARGS <= ix)
+	return 0;
+    return pi_args[ix].flags;
+}
+
+static ERTS_INLINE Uint
+pi_ix2rsz(int ix)
+{
+    if (ix < 0 || ERTS_PI_ARGS <= ix)
+	return 0;
+    return pi_args[ix].reserve_size;
+}
+
+static ERTS_INLINE ErtsProcLocks
+pi_ix2locks(int ix)
+{
+    if (ix < 0 || ERTS_PI_ARGS <= ix)
+	return 0;
+    return pi_args[ix].locks;
 }
 
 static ERTS_INLINE int
 pi_arg2ix(Eterm arg)
 {
     switch (arg) {
-    case am_registered_name:			return 0;
-    case am_current_function:			return 1;
-    case am_initial_call:			return 2;
-    case am_status:				return 3;
-    case am_messages:				return 4;
-    case am_message_queue_len:			return 5;
-    case am_links:				return 6;
-    case am_monitors:				return 7;
-    case am_monitored_by:			return 8;
-    case am_dictionary:				return 9;
-    case am_trap_exit:				return 10;
-    case am_error_handler:			return 11;
-    case am_heap_size:				return 12;
-    case am_stack_size:				return 13;
-    case am_memory:				return 14;
-    case am_garbage_collection:			return 15;
-    case am_group_leader:			return 16;
-    case am_reductions:				return 17;
-    case am_priority:				return 18;
-    case am_trace:				return 19;
-    case am_binary:				return 20;
-    case am_sequential_trace_token:		return 21;
-    case am_catchlevel:				return 22;
-    case am_backtrace:				return 23;
-    case am_last_calls:				return 24;
-    case am_total_heap_size:			return 25;
-    case am_suspending:				return 26;
-    case am_min_heap_size:			return 27;
-    case am_min_bin_vheap_size:			return 28;
-    case am_max_heap_size:			return 29;
-    case am_current_location:			return 30;
-    case am_current_stacktrace:			return 31;
-    case am_message_queue_data:			return 32;
-    case am_garbage_collection_info:		return 33;
-    case am_magic_ref:                          return 34;
-    default:					return -1;
+    case am_registered_name:
+        return ERTS_PI_IX_REGISTERED_NAME;
+    case am_current_function:
+        return ERTS_PI_IX_CURRENT_FUNCTION;
+    case am_initial_call:
+        return ERTS_PI_IX_INITIAL_CALL;
+    case am_status:
+        return ERTS_PI_IX_STATUS;
+    case am_messages:
+        return ERTS_PI_IX_MESSAGES;
+    case am_message_queue_len:
+        return ERTS_PI_IX_MESSAGE_QUEUE_LEN;
+    case am_links:
+        return ERTS_PI_IX_LINKS;
+    case am_monitors:
+        return ERTS_PI_IX_MONITORS;
+    case am_monitored_by:
+        return ERTS_PI_IX_MONITORED_BY;
+    case am_dictionary:
+        return ERTS_PI_IX_DICTIONARY;
+    case am_trap_exit:
+        return ERTS_PI_IX_TRAP_EXIT;
+    case am_error_handler:
+        return ERTS_PI_IX_ERROR_HANDLER;
+    case am_heap_size:
+        return ERTS_PI_IX_HEAP_SIZE;
+    case am_stack_size:
+        return ERTS_PI_IX_STACK_SIZE;
+    case am_memory:
+        return ERTS_PI_IX_MEMORY;
+    case am_garbage_collection:
+        return ERTS_PI_IX_GARBAGE_COLLECTION;
+    case am_group_leader:
+        return ERTS_PI_IX_GROUP_LEADER;
+    case am_reductions:
+        return ERTS_PI_IX_REDUCTIONS;
+    case am_priority:
+        return ERTS_PI_IX_PRIORITY;
+    case am_trace:
+        return ERTS_PI_IX_TRACE;
+    case am_binary:
+        return ERTS_PI_IX_BINARY;
+    case am_sequential_trace_token:
+        return ERTS_PI_IX_SEQUENTIAL_TRACE_TOKEN;
+    case am_catchlevel:
+        return ERTS_PI_IX_CATCHLEVEL;
+    case am_backtrace:
+        return ERTS_PI_IX_BACKTRACE;
+    case am_last_calls:
+        return ERTS_PI_IX_LAST_CALLS;
+    case am_total_heap_size:
+        return ERTS_PI_IX_TOTAL_HEAP_SIZE;
+    case am_suspending:
+        return ERTS_PI_IX_SUSPENDING;
+    case am_min_heap_size:
+        return ERTS_PI_IX_MIN_HEAP_SIZE;
+    case am_min_bin_vheap_size:
+        return ERTS_PI_IX_MIN_BIN_VHEAP_SIZE;
+    case am_max_heap_size:
+        return ERTS_PI_IX_MAX_HEAP_SIZE;
+    case am_current_location:
+        return ERTS_PI_IX_CURRENT_LOCATION;
+    case am_current_stacktrace:
+        return ERTS_PI_IX_CURRENT_STACKTRACE;
+    case am_message_queue_data:
+        return ERTS_PI_IX_MESSAGE_QUEUE_DATA;
+    case am_garbage_collection_info:
+	return ERTS_PI_IX_GARBAGE_COLLECTION_INFO;
+    case am_magic_ref:
+        return ERTS_PI_IX_MAGIC_REF;
+    case am_fullsweep_after:
+        return ERTS_PI_IX_FULLSWEEP_AFTER;
+    default:
+        return -1;
     }
 }
 
@@ -717,7 +909,6 @@ static Eterm pi_1_keys[] = {
     am_initial_call,
     am_status,
     am_message_queue_len,
-    am_messages,
     am_links,
     am_dictionary,
     am_trap_exit,
@@ -761,155 +952,41 @@ process_info_init(void)
 
 }
 
-static ERTS_INLINE Process *
-pi_pid2proc(Process *c_p, Eterm pid, ErtsProcLocks info_locks)
-{
-#ifdef ERTS_SMP
-    /*
-     * If the main lock is needed, we use erts_pid2proc_not_running()
-     * instead of erts_pid2proc() for two reasons:
-     * * Current function of pid and possibly other information will
-     *   have been updated so that process_info() is consistent with an
-     *   info-request/info-response signal model.
-     * * We avoid blocking the whole scheduler executing the
-     *   process that is calling process_info() for a long time
-     *   which will happen if pid is currently running.
-     * The caller of process_info() may have to yield if pid
-     * is currently running.
-     */
-
-    if (info_locks & ERTS_PROC_LOCK_MAIN)
-	return erts_pid2proc_not_running(c_p, ERTS_PROC_LOCK_MAIN,
-					 pid, info_locks);
-    else
-#endif
-	return erts_pid2proc(c_p, ERTS_PROC_LOCK_MAIN,
-			     pid, info_locks);
-}
-
-
-
 static BIF_RETTYPE
-process_info_aux(Process *BIF_P,
+process_info_aux(Process *c_p,
+                 ErtsHeapFactory *hfact,
 		 Process *rp,
 		 ErtsProcLocks rp_locks,
-		 Eterm rpid,
-		 Eterm item,
-		 int always_wrap);
-
-#define ERTS_PI_RES_ELEM_IX_BUF_INC 1024
-#define ERTS_PI_DEF_RES_ELEM_IX_BUF_SZ ERTS_PI_ARGS
+		 int item_ix,
+		 int flags,
+                 Uint *reserve_sizep,
+                 Uint *reds);
 
-static Eterm
-process_info_list(Process *c_p, Eterm pid, Eterm list, int always_wrap,
-		  int *fail_type)
+Eterm
+erts_process_info(Process *c_p,
+                  ErtsHeapFactory *hfact,
+                  Process *rp,
+                  ErtsProcLocks rp_locks,
+                  int *item_ix,
+                  int item_ix_len,
+                  int flags,
+                  Uint reserve_size,
+                  Uint *reds)
 {
-    int want_messages = 0;
-    int def_res_elem_ix_buf[ERTS_PI_DEF_RES_ELEM_IX_BUF_SZ];
-    int *res_elem_ix = &def_res_elem_ix_buf[0];
-    int res_elem_ix_ix = -1;
-    int res_elem_ix_sz = ERTS_PI_DEF_RES_ELEM_IX_BUF_SZ;
+    Eterm res;
     Eterm part_res[ERTS_PI_ARGS];
-    Eterm res, arg;
-    Uint *hp, *hp_end;
-    ErtsProcLocks locks = (ErtsProcLocks) 0;
-    int res_len, ix;
-    Process *rp = NULL;
+    int item_ix_ix, ix;
 
-    *fail_type = ERTS_PI_FAIL_TYPE_BADARG;
+    if (ERTS_PI_FLAG_SINGELTON & flags) {
+        ASSERT(item_ix_len == 1);
+	res = process_info_aux(c_p, hfact, rp, rp_locks, item_ix[0],
+                               flags, &reserve_size, reds);
+        return res;
+    }
 
     for (ix = 0; ix < ERTS_PI_ARGS; ix++)
 	part_res[ix] = THE_NON_VALUE;
 
-    ASSERT(is_list(list));
-
-    while (is_list(list)) {
-	Eterm* consp = list_val(list);
-
-	arg = CAR(consp);
-	ix = pi_arg2ix(arg);
-	if (ix < 0) {
-	    res = THE_NON_VALUE;
-	    goto done;
-	}
-	if (arg == am_messages)
-	    want_messages = 1;
-	locks |= pi_locks(arg);
-	res_elem_ix_ix++;
-	if (res_elem_ix_ix >= res_elem_ix_sz) {
-	    if (res_elem_ix != &def_res_elem_ix_buf[0])
-		res_elem_ix =
-		    erts_realloc(ERTS_ALC_T_TMP,
-				 res_elem_ix,
-				 sizeof(int)*(res_elem_ix_sz
-					      += ERTS_PI_RES_ELEM_IX_BUF_INC));
-	    else {
-		int new_res_elem_ix_sz = ERTS_PI_RES_ELEM_IX_BUF_INC;
-		int *new_res_elem_ix = erts_alloc(ERTS_ALC_T_TMP,
-						  sizeof(int)*new_res_elem_ix_sz);
-		sys_memcpy((void *) new_res_elem_ix,
-			   (void *) res_elem_ix,
-			   sizeof(int)*res_elem_ix_sz);
-		res_elem_ix = new_res_elem_ix;
-		res_elem_ix_sz = new_res_elem_ix_sz;
-	    }
-	}
-	res_elem_ix[res_elem_ix_ix] = ix;
-	list = CDR(consp);
-    }
-    if (is_not_nil(list)) {
-	res = THE_NON_VALUE;
-	goto done;
-    }
-
-    res_len = res_elem_ix_ix+1;
-
-    ASSERT(res_len > 0);
-
-    rp = pi_pid2proc(c_p, pid, locks|ERTS_PROC_LOCK_STATUS);
-    if (!rp) {
-	res = am_undefined;
-	goto done;
-    }
-    else if (rp == ERTS_PROC_LOCK_BUSY) {
-	rp = NULL;
-	res = THE_NON_VALUE;
-	*fail_type = ERTS_PI_FAIL_TYPE_YIELD;
-	goto done;
-    }
-    else if (c_p != rp && ERTS_PROC_PENDING_EXIT(rp)) {
-	locks |= ERTS_PROC_LOCK_STATUS;
-	res = THE_NON_VALUE;
-	*fail_type = ERTS_PI_FAIL_TYPE_AWAIT_EXIT;
-	goto done;
-    }
-    else {
-	ErtsProcLocks unlock_locks = 0;
-
-	if (c_p == rp)
-	    locks |= ERTS_PROC_LOCK_MAIN;
-
-	if (!(locks & ERTS_PROC_LOCK_STATUS))
-	    unlock_locks |= ERTS_PROC_LOCK_STATUS;
-
-	if (locks & ERTS_PROC_LOCK_MSGQ) {
-	    /*
-	     * Move in queue into private queue and
-	     * release msgq lock, enabling others to
-	     * send messages to the process while it
-	     * is being inspected...
-	     */
-	    ASSERT(locks & ERTS_PROC_LOCK_MAIN);
-	    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(rp);
-	    locks &= ~ERTS_PROC_LOCK_MSGQ;
-	    unlock_locks |= ERTS_PROC_LOCK_MSGQ;
-	}
-
-	if (unlock_locks)
-	    erts_smp_proc_unlock(rp, unlock_locks);
-
-    }
-
     /*
      * We always handle 'messages' first if it should be part
      * of the result. This since if both 'messages' and
@@ -917,28 +994,31 @@ process_info_list(Process *c_p, Eterm pid, Eterm list, int always_wrap,
      * change the result of 'message_queue_len' (in case
      * the queue contain bad distribution messages).
      */
-    if (want_messages) {
+    if (flags & ERTS_PI_FLAG_WANT_MSGS) {
 	ix = pi_arg2ix(am_messages);
 	ASSERT(part_res[ix] == THE_NON_VALUE);
-	part_res[ix] = process_info_aux(c_p, rp, locks, pid, am_messages, always_wrap);
-	ASSERT(part_res[ix] != THE_NON_VALUE);
+	res = process_info_aux(c_p, hfact, rp, rp_locks, ix,
+                               flags, &reserve_size, reds);
+	ASSERT(res != am_undefined);
+	ASSERT(res != THE_NON_VALUE);
+        part_res[ix] = res;
     }
 
-    for (; res_elem_ix_ix >= 0; res_elem_ix_ix--) {
-	ix = res_elem_ix[res_elem_ix_ix];
+    for (item_ix_ix = item_ix_len - 1; item_ix_ix >= 0; item_ix_ix--) {
+	ix = item_ix[item_ix_ix];
 	if (part_res[ix] == THE_NON_VALUE) {
-	    arg = pi_ix2arg(ix);
-	    part_res[ix] = process_info_aux(c_p, rp, locks, pid, arg, always_wrap);
-	    ASSERT(part_res[ix] != THE_NON_VALUE);
+	    res = process_info_aux(c_p, hfact, rp, rp_locks, ix,
+                                   flags, &reserve_size, reds);
+            ASSERT(res != am_undefined);
+	    ASSERT(res != THE_NON_VALUE);
+            part_res[ix] = res;
 	}
     }
 
-    hp = HAlloc(c_p, res_len*2);
-    hp_end = hp + res_len*2;
     res = NIL;
 
-    for (res_elem_ix_ix = res_len - 1; res_elem_ix_ix >= 0; res_elem_ix_ix--) {
-	ix = res_elem_ix[res_elem_ix_ix];
+    for (item_ix_ix = item_ix_len - 1; item_ix_ix >= 0; item_ix_ix--) {
+	ix = item_ix[item_ix_ix];
 	ASSERT(part_res[ix] != THE_NON_VALUE);
 	/*
 	 * If we should ignore the value of registered_name,
@@ -946,173 +1026,317 @@ process_info_list(Process *c_p, Eterm pid, Eterm list, int always_wrap,
 	 * beginning of process_info_aux().
 	 */
 	if (is_nil(part_res[ix])) {
-	    ASSERT(!always_wrap);
+	    ASSERT(!(flags & ERTS_PI_FLAG_ALWAYS_WRAP));
 	    ASSERT(pi_ix2arg(ix) == am_registered_name);
 	}
 	else {
+            Eterm *hp;
+            ERTS_PI_UNRESERVE(reserve_size, 2);
+            hp = erts_produce_heap(hfact, 2, reserve_size);
 	    res = CONS(hp, part_res[ix], res);
-	    hp += 2;
 	}
     }
 
-    if (!always_wrap) {
-	HRelease(c_p, hp_end, hp);
-    }
-
- done:
-
-    if (c_p == rp)
-	locks &= ~ERTS_PROC_LOCK_MAIN;
-    if (locks && rp)
-	erts_smp_proc_unlock(rp, locks);
-
-    if (res_elem_ix != &def_res_elem_ix_buf[0])
-	erts_free(ERTS_ALC_T_TMP, res_elem_ix);
-
     return res;
 }
 
-BIF_RETTYPE process_info_1(BIF_ALIST_1)
+static void
+pi_setup_grow(int **arr, int *def_arr, Uint *sz, int ix);
+
+static BIF_RETTYPE
+process_info_bif(Process *c_p, Eterm pid, Eterm opt, int always_wrap, int pi2)
 {
+    ErtsHeapFactory hfact;
+    int def_arr[ERTS_PI_DEF_ARR_SZ];
+    int *item_ix = &def_arr[0];
+    Process *rp = NULL;
+    erts_aint32_t state;
+    BIF_RETTYPE ret;
+    Uint reds = 0;
+    ErtsProcLocks locks = 0;
+    int flags;
+    Uint reserve_size;
+    int len;
     Eterm res;
-    int fail_type;
 
-    if (is_external_pid(BIF_ARG_1)
-	&& external_pid_dist_entry(BIF_ARG_1) == erts_this_dist_entry)
-	BIF_RET(am_undefined);
-	
-    if (is_not_internal_pid(BIF_ARG_1)) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
-
-    res = process_info_list(BIF_P, BIF_ARG_1, pi_1_keys_list, 0, &fail_type);
-    if (is_non_value(res)) {
-	switch (fail_type) {
-	case ERTS_PI_FAIL_TYPE_BADARG:
-	    BIF_ERROR(BIF_P, BADARG);
-	case ERTS_PI_FAIL_TYPE_YIELD:
-	    ERTS_BIF_YIELD1(bif_export[BIF_process_info_1], BIF_P, BIF_ARG_1);
-	case ERTS_PI_FAIL_TYPE_AWAIT_EXIT:
-	    ERTS_BIF_AWAIT_X_DATA_TRAP(BIF_P, BIF_ARG_1, am_undefined);
-	default:
-	    erts_exit(ERTS_ABORT_EXIT, "%s:%d: Internal error", __FILE__, __LINE__);
-	}
+    ERTS_CT_ASSERT(ERTS_PI_DEF_ARR_SZ > 0);
+
+    if (c_p->common.id == pid) {
+        int local_only = c_p->flags & F_LOCAL_SIGS_ONLY;
+        int sres, sreds, reds_left;
+
+        reds_left = ERTS_BIF_REDS_LEFT(c_p);
+        sreds = reds_left;
+
+        if (!local_only) {
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+            erts_proc_sig_fetch(c_p);
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+        }
+
+        sres = erts_proc_sig_handle_incoming(c_p, &state, &sreds, sreds, !0);
+
+        BUMP_REDS(c_p, (int) sreds);
+        reds_left -= sreds;
+
+        if (state & ERTS_PSFLG_EXITING) {
+            c_p->flags &= ~F_LOCAL_SIGS_ONLY;
+            goto exited;
+        }
+        if (!sres | (reds_left <= 0)) {
+            /*
+             * More signals to handle or out of reds; need
+             * to yield and continue. Prevent fetching of
+             * more signals by setting local-sigs-only flag.
+             */
+            c_p->flags |= F_LOCAL_SIGS_ONLY;
+            goto yield;
+        }
+
+        c_p->flags &= ~F_LOCAL_SIGS_ONLY;
     }
 
-    ASSERT(!(BIF_P->flags & F_P2PNR_RESCHED));
-    BIF_RET(res);
-}
+    if (is_atom(opt)) {
+	int ix = pi_arg2ix(opt);
+        item_ix[0] = ix;
+        len = 1;
+        locks = pi_ix2locks(ix);
+        reserve_size = 3 + pi_ix2rsz(ix);
+        flags = ERTS_PI_FLAG_SINGELTON;
+        flags |= pi_ix2flags(ix);
+        if (ix < 0)
+            goto badarg;
+    }
+    else {
+        Eterm list = opt;
+        Uint size = ERTS_PI_DEF_ARR_SZ;
 
+        len = 0;
+        reserve_size = 0;
+        locks = 0;
+        flags = 0;
 
-BIF_RETTYPE process_info_2(BIF_ALIST_2) 
-{
-    Eterm res;
-    Process *rp;
-    Eterm pid = BIF_ARG_1;
-    ErtsProcLocks info_locks;
-    int fail_type;
+        while (is_list(list)) {
+            Eterm *consp = list_val(list);
+            Eterm arg = CAR(consp);
+            int ix = pi_arg2ix(arg);
+            if (ix < 0)
+                goto badarg;
 
-    if (is_external_pid(pid)
-	&& external_pid_dist_entry(pid) == erts_this_dist_entry)
-	BIF_RET(am_undefined);
-	
-    if (is_not_internal_pid(pid)) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
-
-    if (is_nil(BIF_ARG_2))
-	BIF_RET(NIL);
-
-    if (is_list(BIF_ARG_2)) {
-	res = process_info_list(BIF_P, BIF_ARG_1, BIF_ARG_2, 1, &fail_type);
-	if (is_non_value(res)) {
-	    switch (fail_type) {
-	    case ERTS_PI_FAIL_TYPE_BADARG:
-		BIF_ERROR(BIF_P, BADARG);
-	    case ERTS_PI_FAIL_TYPE_YIELD:
-		ERTS_BIF_YIELD2(bif_export[BIF_process_info_2], BIF_P,
-				BIF_ARG_1, BIF_ARG_2);
-	    case ERTS_PI_FAIL_TYPE_AWAIT_EXIT:
-		ERTS_BIF_AWAIT_X_DATA_TRAP(BIF_P, BIF_ARG_1, am_undefined);
-	    default:
-		erts_exit(ERTS_ABORT_EXIT, "%s:%d: Internal error",
-			 __FILE__, __LINE__);
-	    }
-	}
-	ASSERT(!(BIF_P->flags & F_P2PNR_RESCHED));
-	BIF_RET(res);
+            if (len >= size)
+                pi_setup_grow(&item_ix, def_arr, &size, len);
+
+            item_ix[len++] = ix;
+
+            locks |= pi_ix2locks(ix);
+            flags |= pi_ix2flags(ix);
+            reserve_size += pi_ix2rsz(ix);
+            reserve_size += 3; /* 2-tuple */
+            reserve_size += 2; /* cons */
+
+            list = CDR(consp);
+        }
+
+        if (is_not_nil(list))
+            goto badarg;
     }
 
-    if (pi_arg2ix(BIF_ARG_2) < 0)
-	BIF_ERROR(BIF_P, BADARG);
+    if (is_not_internal_pid(pid)) {
+        if (is_external_pid(pid)
+            && external_pid_dist_entry(pid) == erts_this_dist_entry)
+            goto undefined;
+        goto badarg;
+    }
 
-    info_locks = pi_locks(BIF_ARG_2); 
+    if (always_wrap)
+        flags |= ERTS_PI_FLAG_ALWAYS_WRAP;
 
-    rp = pi_pid2proc(BIF_P, pid, info_locks|ERTS_PROC_LOCK_STATUS);
-    if (!rp)
-	res = am_undefined;
-    else if (rp == ERTS_PROC_LOCK_BUSY)
-	ERTS_BIF_YIELD2(bif_export[BIF_process_info_2], BIF_P,
-			BIF_ARG_1, BIF_ARG_2);
-    else if (rp != BIF_P && ERTS_PROC_PENDING_EXIT(rp)) {
-	erts_smp_proc_unlock(rp, info_locks|ERTS_PROC_LOCK_STATUS);
-	ERTS_BIF_AWAIT_X_DATA_TRAP(BIF_P, BIF_ARG_1, am_undefined);
+    if (c_p->common.id == pid) {
+        rp = c_p;
+        if (locks & ~ERTS_PROC_LOCK_MAIN)
+            erts_proc_lock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
+        locks |= ERTS_PROC_LOCK_MAIN;
     }
     else {
-	ErtsProcLocks unlock_locks = 0;
+        if (flags & ERTS_PI_FLAG_FORCE_SIG_SEND)
+            goto send_signal;
+        state = ERTS_PSFLG_RUNNING; /* fail state... */
+        rp = erts_try_lock_sig_free_proc(pid, locks, &state);
+        if (!rp)
+            goto undefined;
+        if (rp == ERTS_PROC_LOCK_BUSY) {
+            rp = NULL;
+            goto send_signal;
+        }
+        if (state & ERTS_PSFLG_EXITING) {
+            if (locks)
+                erts_proc_unlock(rp, locks);
+            locks = 0;
+            /* wait for it to terminate properly... */
+            goto send_signal;
+        }
+        if (flags & ERTS_PI_FLAG_NEED_MSGQ_LEN) {
+            ASSERT(locks & ERTS_PROC_LOCK_MAIN);
+            erts_proc_lock(rp, ERTS_PROC_LOCK_MSGQ);
+            erts_proc_sig_fetch(rp);
+            if (c_p->sig_qs.cont) {
+                erts_proc_unlock(rp, locks|ERTS_PROC_LOCK_MSGQ);
+                locks = 0;
+                goto send_signal;
+            }
+            erts_proc_unlock(rp, ERTS_PROC_LOCK_MSGQ);
+        }
+    }
 
-	if (BIF_P == rp)
-	    info_locks |= ERTS_PROC_LOCK_MAIN;
+    erts_factory_proc_init(&hfact, c_p);
 
-	if (!(info_locks & ERTS_PROC_LOCK_STATUS))
-	    unlock_locks |= ERTS_PROC_LOCK_STATUS;
+    res = erts_process_info(c_p, &hfact, rp, locks, item_ix, len,
+                            flags, reserve_size, &reds);
 
-	if (info_locks & ERTS_PROC_LOCK_MSGQ) {
-	    /*
-	     * Move in queue into private queue and
-	     * release msgq lock, enabling others to
-	     * send messages to the process while it
-	     * is being inspected...
-	     */
-	    ASSERT(info_locks & ERTS_PROC_LOCK_MAIN);
-	    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(rp);
-	    info_locks &= ~ERTS_PROC_LOCK_MSGQ;
-	    unlock_locks |= ERTS_PROC_LOCK_MSGQ;
-	}
+    erts_factory_close(&hfact);
 
-	if (unlock_locks)
-	    erts_smp_proc_unlock(rp, unlock_locks);
+    if (reds > INT_MAX/2)
+        reds = INT_MAX/2;
+    BUMP_REDS(c_p, (int) reds);
 
-	res = process_info_aux(BIF_P, rp, info_locks, pid, BIF_ARG_2, 0);
+    state = erts_atomic32_read_acqb(&rp->state);
+    if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_FREE)) {
+        if (state & ERTS_PSFLG_FREE) {
+            ASSERT(!locks);
+            goto undefined;
+        }
+        if (locks)
+            erts_proc_unlock(rp, locks);
+        locks = 0;
+        /* wait for it to terminate properly... */
+        goto send_signal;
     }
-    ASSERT(is_value(res));
 
-#ifdef ERTS_SMP
-    if (BIF_P == rp)
-	info_locks &= ~ERTS_PROC_LOCK_MAIN;
-    if (rp && info_locks)
-	erts_smp_proc_unlock(rp, info_locks);
-#endif
+    ERTS_BIF_PREP_RET(ret, res);
 
-    ASSERT(!(BIF_P->flags & F_P2PNR_RESCHED));
-    BIF_RET(res);
+done:
+
+    if (c_p == rp)
+        locks &= ~ERTS_PROC_LOCK_MAIN;
+
+    if (locks && rp)
+	erts_proc_unlock(rp, locks);
+
+    if (item_ix != def_arr)
+        erts_free(ERTS_ALC_T_TMP, item_ix);
+
+    return ret;
+
+badarg:
+    ERTS_BIF_PREP_ERROR(ret, c_p, BADARG);
+    goto done;
+
+undefined:
+    ERTS_BIF_PREP_RET(ret, am_undefined);
+    goto done;
+
+exited:
+    ERTS_BIF_PREP_EXITED(ret, c_p);
+    goto done;
+
+yield:
+    if (pi2)
+        ERTS_BIF_PREP_YIELD2(ret, bif_export[BIF_process_info_2], c_p, pid, opt);
+    else
+        ERTS_BIF_PREP_YIELD1(ret, bif_export[BIF_process_info_1], c_p, pid);
+    goto done;
+
+send_signal: {
+        Eterm ref = erts_make_ref(c_p);
+        int enqueued, need_msgq_len;
+        flags |= ERTS_PI_FLAG_REQUEST_FOR_OTHER;
+        need_msgq_len = (flags & ERTS_PI_FLAG_NEED_MSGQ_LEN);
+        /*
+         * Set receive mark so we wont have to scan the whole
+         * message queue for the result. Note caller unconditionally
+         * has to enter a receive only matching messages containing
+         * 'ref', or restore save pointer.
+         */
+        ERTS_RECV_MARK_SAVE(c_p);
+        ERTS_RECV_MARK_SET(c_p);
+        enqueued = erts_proc_sig_send_process_info_request(c_p, pid, item_ix,
+                                                           len, need_msgq_len,
+                                                           flags, reserve_size,
+                                                           ref);
+        if (!enqueued) {
+            /* Restore save pointer... */
+	    JOIN_MESSAGE(c_p);
+            goto undefined;
+        }
+        ERTS_BIF_PREP_TRAP1(ret, erts_await_result, c_p, ref);
+        goto done;
+    }
+}
+
+static void
+pi_setup_grow(int **arr, int *def_arr, Uint *sz, int ix)
+{
+    *sz = (ix+1) + ERTS_PI_DEF_ARR_SZ;
+    if (*arr != def_arr)
+        *arr = erts_realloc(ERTS_ALC_T_TMP, *arr, (*sz)*sizeof(int));
+    else {
+        int *new_arr = erts_alloc(ERTS_ALC_T_TMP, (*sz)*sizeof(int));
+        sys_memcpy((void *) new_arr, (void *) def_arr,
+                   sizeof(int)*ERTS_PI_DEF_ARR_SZ);
+        *arr = new_arr;
+    }
+}
+
+
+BIF_RETTYPE process_info_2(BIF_ALIST_2)
+{
+    return process_info_bif(BIF_P, BIF_ARG_1, BIF_ARG_2, !is_atom(BIF_ARG_2), !0);
+}
+
+BIF_RETTYPE process_info_1(BIF_ALIST_1)
+{
+    return process_info_bif(BIF_P, BIF_ARG_1, pi_1_keys_list, 0, 0);
 }
 
 Eterm
-process_info_aux(Process *BIF_P,
+process_info_aux(Process *c_p,
+                 ErtsHeapFactory *hfact,
 		 Process *rp,
 		 ErtsProcLocks rp_locks,
-		 Eterm rpid,
-		 Eterm item,
-		 int always_wrap)
+		 int item_ix,
+		 int flags,
+                 Uint *reserve_sizep,
+                 Uint *reds)
 {
     Eterm *hp;
     Eterm res = NIL;
+    Uint reserved;
+    Uint reserve_size = *reserve_sizep;
+
+#ifdef ERTS_ENABLE_LOCK_CHECK
+    ErtsProcLocks locks = erts_proc_lc_my_proc_locks(rp);
+
+    switch (item_ix) {
+    case ERTS_PI_IX_STATUS:
+    case ERTS_PI_IX_PRIORITY:
+    case ERTS_PI_IX_SUSPENDING:
+        ERTS_LC_ASSERT((locks & ~ERTS_PROC_LOCK_MAIN) == 0);
+        break;
+    default:
+        ERTS_LC_ASSERT(locks == ERTS_PROC_LOCK_MAIN);
+        break;
+    }
+#endif
+
+    reserved = pi_ix2rsz(item_ix);
+    ERTS_PI_UNRESERVE(reserve_size, reserved);
+
+    (*reds)++;
 
     ASSERT(rp);
 
     /*
-     * Q: Why this always_wrap argument?
+     * Q: Why this ERTS_PI_FLAG_ALWAYS_WRAP flag?
      *
      * A: registered_name is strange. If process has no registered name,
      *    process_info(Pid, registered_name) returns [], and
@@ -1124,41 +1348,39 @@ process_info_aux(Process *BIF_P,
      *    registered_name behaves as it should, i.e. a
      *    {registered_name, []} will appear in the resulting list.
      *
-     *    If always_wrap != 0, process_info_aux() always wrap the result
-     *    in a key two tuple. 
+     *    If ERTS_PI_FLAG_ALWAYS_WRAP is set, process_info_aux() always
+     *    wrap the result in a key two tuple. 
      */
 
-    switch (item) {
+    switch (item_ix) {
 
-    case am_registered_name:
-	if (rp->common.u.alive.reg) {
-	    hp = HAlloc(BIF_P, 3);
+    case ERTS_PI_IX_REGISTERED_NAME:
+	if (rp->common.u.alive.reg)
 	    res = rp->common.u.alive.reg->name;
-	} else {
-	    if (always_wrap) {
-		hp = HAlloc(BIF_P, 3);
+        else {
+	    if (flags & ERTS_PI_FLAG_ALWAYS_WRAP)
 		res = NIL;
-	    }
-	    else {
+	    else
 		return NIL;
-	    }
 	}
 	break;
 
-    case am_current_function:
-	res = current_function(BIF_P, rp, &hp, 0);
+    case ERTS_PI_IX_CURRENT_FUNCTION:
+	res = current_function(c_p, hfact, rp, 0,
+                               reserve_size, flags);
 	break;
 
-    case am_current_location:
-	res = current_function(BIF_P, rp, &hp, 1);
+    case ERTS_PI_IX_CURRENT_LOCATION:
+	res = current_function(c_p, hfact, rp, 1,
+                               reserve_size, flags);
 	break;
 
-    case am_current_stacktrace:
-	res = current_stacktrace(BIF_P, rp, &hp);
+    case ERTS_PI_IX_CURRENT_STACKTRACE:
+	res = current_stacktrace(hfact, rp, reserve_size);
 	break;
 
-    case am_initial_call:
-	hp = HAlloc(BIF_P, 3+4);
+    case ERTS_PI_IX_INITIAL_CALL:
+        hp = erts_produce_heap(hfact, 4, reserve_size);
 	res = TUPLE3(hp,
 		     rp->u.initial.module,
 		     rp->u.initial.function,
@@ -1166,96 +1388,130 @@ process_info_aux(Process *BIF_P,
 	hp += 4;
 	break;
 
-    case am_status:
-	res = erts_process_status(rp, rpid);
-	ASSERT(res != am_undefined);
-	hp = HAlloc(BIF_P, 3);
+    case ERTS_PI_IX_STATUS: {
+        erts_aint32_t state = erts_atomic32_read_nob(&rp->state);
+        res = erts_process_state2status(state);
+        if (res == am_running && (state & ERTS_PSFLG_RUNNING_SYS)) {
+            ASSERT(c_p == rp);
+            ASSERT(flags & ERTS_PI_FLAG_REQUEST_FOR_OTHER);
+            if (!(state & (ERTS_PSFLG_SYS_TASKS
+                           | ERTS_PSFLG_ACTIVE
+                           | ERTS_PSFLG_SIG_Q
+                           | ERTS_PSFLG_SIG_IN_Q))) {
+                /*
+                 * We are servicing a process-info request from
+                 * another process. If that other process could
+                 * have inspected our state itself, we would have
+                 * been in the 'waiting' state.
+                 */
+                res = am_waiting;
+            }
+        }
 	break;
+    }
 
-    case am_messages: {
-
-	if (rp->msg.len == 0 || ERTS_TRACE_FLAGS(rp) & F_SENSITIVE) {
-	    hp = HAlloc(BIF_P, 3);
-	} else {
+    case ERTS_PI_IX_MESSAGES: {
+        ASSERT(flags & ERTS_PI_FLAG_NEED_MSGQ_LEN);
+	if (rp->sig_qs.len == 0 || (ERTS_TRACE_FLAGS(rp) & F_SENSITIVE))
+            res = NIL;
+        else {
+            int info_on_self = !(flags & ERTS_PI_FLAG_REQUEST_FOR_OTHER);
 	    ErtsMessageInfo *mip;
-	    Sint i;
+	    Sint i, len;
 	    Uint heap_need;
-#ifdef DEBUG
-	    Eterm *hp_end;
-#endif
 
 	    mip = erts_alloc(ERTS_ALC_T_TMP,
-			     rp->msg.len*sizeof(ErtsMessageInfo));
+			     rp->sig_qs.len*sizeof(ErtsMessageInfo));
 
 	    /*
 	     * Note that message queue may shrink when calling
-	     * erts_prep_msgq_for_inspection() since it removes
+	     * erts_proc_sig_prep_msgq_for_inspection() since it removes
 	     * corrupt distribution messages.
 	     */
-	    heap_need = erts_prep_msgq_for_inspection(BIF_P, rp, rp_locks, mip);
-	    heap_need += 3; /* top 2-tuple */
-	    heap_need += rp->msg.len*2; /* Cons cells */
+	    heap_need = erts_proc_sig_prep_msgq_for_inspection(c_p, rp,
+                                                               rp_locks,
+                                                               info_on_self,
+                                                               mip);
+            len = rp->sig_qs.len;
 
-	    hp = HAlloc(BIF_P, heap_need); /* heap_need is exact */
-#ifdef DEBUG
-	    hp_end = hp + heap_need;
-#endif
+	    heap_need += len*2; /* Cons cells */
+
+            reserve_size += heap_need;
 
 	    /* Build list of messages... */
-	    for (i = rp->msg.len - 1, res = NIL; i >= 0; i--) {
+	    for (i = len - 1, res = NIL; i >= 0; i--) {
 		Eterm msg = ERL_MESSAGE_TERM(mip[i].msgp);
 		Uint sz = mip[i].size;
 
+                ERTS_PI_UNRESERVE(reserve_size, sz+2);
+                hp = erts_produce_heap(hfact, sz+2, reserve_size);
+
 		if (sz != 0)
-		    msg = copy_struct(msg, sz, &hp, &BIF_P->off_heap);
+		    msg = copy_struct(msg, sz, &hp, hfact->off_heap);
 
 		res = CONS(hp, msg, res);
 		hp += 2;
 	    }
 
-	    ASSERT(hp_end == hp + 3);
+            *reds += (Uint) len / 4;
 
 	    erts_free(ERTS_ALC_T_TMP, mip);
 	}
 	break;
     }
 
-    case am_message_queue_len:
-	hp = HAlloc(BIF_P, 3);
-	res = make_small(rp->msg.len);
+    case ERTS_PI_IX_MESSAGE_QUEUE_LEN: {
+        Sint len = rp->sig_qs.len;
+        ASSERT(flags & ERTS_PI_FLAG_NEED_MSGQ_LEN);
+        ASSERT(len >= 0);
+        if (len <= MAX_SMALL)
+            res = make_small(len);
+        else {
+            hp = erts_produce_heap(hfact, BIG_UINT_HEAP_SIZE, reserve_size);
+            res = uint_to_big((Uint) len, hp);
+        }
 	break;
+    }
 
-    case am_links: {
+    case ERTS_PI_IX_LINKS: {
 	MonitorInfoCollection mic;
 	int i;
 	Eterm item;
 
 	INIT_MONITOR_INFOS(mic);
 
-	erts_doforall_links(ERTS_P_LINKS(rp),&collect_one_link,&mic);
+	erts_link_tree_foreach(ERTS_P_LINKS(rp), collect_one_link, (void *) &mic);
 
-	hp = HAlloc(BIF_P, 3 + mic.sz);
+        reserve_size += mic.sz;
 	res = NIL;
 	for (i = 0; i < mic.mi_i; i++) {
-	    item = STORE_NC(&hp, &MSO(BIF_P), mic.mi[i].entity.term); 
+            Eterm item_src = mic.mi[i].entity.term;
+            Uint sz = NC_HEAP_SIZE(item_src) + 2;
+            ERTS_PI_UNRESERVE(reserve_size, sz);
+            hp = erts_produce_heap(hfact, sz, reserve_size);
+	    item = STORE_NC(&hp, hfact->off_heap, item_src); 
 	    res = CONS(hp, item, res);
-	    hp += 2;
 	}
+
+        *reds += (Uint) mic.mi_i / 4;
+
 	DESTROY_MONITOR_INFOS(mic);
 	break;
     }
 
-    case am_monitors: {
+    case ERTS_PI_IX_MONITORS: {
 	MonitorInfoCollection mic;
         int i;
 
 	INIT_MONITOR_INFOS(mic);
-        erts_doforall_monitors(ERTS_P_MONITORS(rp),
-                               &collect_one_origin_monitor, &mic);
-        hp = HAlloc(BIF_P, 3 + mic.sz);
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(rp),
+                                  collect_one_origin_monitor,
+                                  (void *) &mic);
+
+        reserve_size += mic.sz;
 	res = NIL;
 	for (i = 0; i < mic.mi_i; i++) {
-	    if (is_atom(mic.mi[i].entity.term)) {
+	    if (mic.mi[i].named) {
 		/* Monitor by name. 
                  * Build {process|port, {Name, Node}} and cons it.
 		 */
@@ -1267,180 +1523,228 @@ process_info_aux(Process *BIF_P,
                     || is_port(mic.mi[i].pid)
                     || is_atom(mic.mi[i].pid));
 
+                ERTS_PI_UNRESERVE(reserve_size, 3+3+2);
+                hp = erts_produce_heap(hfact, 3+3+2, reserve_size);
+
 		t1 = TUPLE2(hp, mic.mi[i].entity.term, mic.mi[i].node);
 		hp += 3;
                 t2 = TUPLE2(hp, m_type, t1);
 		hp += 3;
 		res = CONS(hp, t2, res);
-                hp += 2;
 	    }
 	    else {
-                /* Monitor by pid. Build {process|port, Pid} and cons it. */
+                /* Build {process|port|time_offset, Pid|clock_service} and cons it. */
 		Eterm t;
-		Eterm pid = STORE_NC(&hp, &MSO(BIF_P), mic.mi[i].entity.term);
+		Eterm pid;
+                Eterm m_type;
+                Eterm pid_src = mic.mi[i].entity.term;
+                Uint sz = is_atom(pid_src) ? 0 : NC_HEAP_SIZE(pid_src);
+                sz += 3 + 2;
+
+                ERTS_PI_UNRESERVE(reserve_size, sz);
+                hp = erts_produce_heap(hfact, sz, reserve_size);
+
+                pid = (is_atom(pid_src)
+                       ? pid_src
+                       : STORE_NC(&hp, hfact->off_heap, pid_src));
+
+                switch (mic.mi[i].type) {
+                case ERTS_MON_TYPE_PORT:
+                    m_type = am_port;
+                    break;
+                case ERTS_MON_TYPE_TIME_OFFSET:
+                    m_type = am_time_offset;
+                    break;
+                default:
+                    m_type = am_process;
+                    break;
+                }
 
-                Eterm m_type = is_port(mic.mi[i].pid) ? am_port : am_process;
                 ASSERT(is_pid(mic.mi[i].pid)
                     || is_port(mic.mi[i].pid));
 
                 t = TUPLE2(hp, m_type, pid);
 		hp += 3;
 		res = CONS(hp, t, res);
-                hp += 2;
 	    }
 	}
+
+        *reds += (Uint) mic.mi_i / 4;
+
         DESTROY_MONITOR_INFOS(mic);
 	break;
     }
 
-    case am_monitored_by: {
+    case ERTS_PI_IX_MONITORED_BY: {
 	MonitorInfoCollection mic;
 	int i;
 	Eterm item;
 
 	INIT_MONITOR_INFOS(mic);
-	erts_doforall_monitors(ERTS_P_MONITORS(rp),&collect_one_target_monitor,&mic);
-	hp = HAlloc(BIF_P, 3 + mic.sz);
+        erts_monitor_list_foreach(ERTS_P_LT_MONITORS(rp),
+                                  collect_one_target_monitor,
+                                  (void *) &mic);
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(rp),
+                                  collect_one_target_monitor,
+                                  (void *) &mic);
+
+        reserve_size += mic.sz;
 
 	res = NIL;
 	for (i = 0; i < mic.mi_i; ++i) {
-            if (mic.mi[i].node == make_small(MON_NIF_TARGET)) {
-                item = erts_bld_resource_ref(&hp, &MSO(BIF_P), mic.mi[i].entity.resource);
-            }
-            else {
-                item = STORE_NC(&hp, &MSO(BIF_P), mic.mi[i].entity.term);
-            }
+            Uint sz = 2;
+
+            if (mic.mi[i].type == ERTS_MON_TYPE_RESOURCE)
+                sz += erts_resource_ref_size(mic.mi[i].entity.resource);
+            else
+                sz += NC_HEAP_SIZE(mic.mi[i].entity.term);
+
+            ERTS_PI_UNRESERVE(reserve_size, sz);
+            hp = erts_produce_heap(hfact, sz, reserve_size);
+
+            if (mic.mi[i].type == ERTS_MON_TYPE_RESOURCE)
+                item = erts_bld_resource_ref(&hp,
+                                             hfact->off_heap,
+                                             mic.mi[i].entity.resource);
+            else
+                item = STORE_NC(&hp,
+                                hfact->off_heap,
+                                mic.mi[i].entity.term);
 	    res = CONS(hp, item, res);
-	    hp += 2;
 	}
+
+        *reds += (Uint) mic.mi_i / 4;
+
 	DESTROY_MONITOR_INFOS(mic);
 	break;
     }
 
-    case am_suspending: {
+    case ERTS_PI_IX_SUSPENDING: {
 	ErtsSuspendMonitorInfoCollection smic;
 	int i;
-	Eterm item;
-#ifdef DEBUG
-	Eterm *hp_end;
-#endif
 
-	ERTS_INIT_SUSPEND_MONITOR_INFOS(smic,
-					BIF_P,
-					(BIF_P == rp
-					 ? ERTS_PROC_LOCK_MAIN
-					 : 0) | ERTS_PROC_LOCK_LINK);
+	ERTS_INIT_SUSPEND_MONITOR_INFOS(smic);
+
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(rp),
+                                  collect_one_suspend_monitor,
+                                  (void *) &smic);
+
+        reserve_size += smic.sz;
 
-	erts_doforall_suspend_monitors(rp->suspend_monitors,
-				       &collect_one_suspend_monitor,
-				       &smic);
-	hp = HAlloc(BIF_P, 3 + smic.sz);
-#ifdef DEBUG
-	hp_end = hp + smic.sz;
-#endif
-	
 	res = NIL;
 	for (i = 0; i < smic.smi_i; i++) {
-	    Sint a = (Sint) smic.smi[i]->active;  /* quiet compiler warnings */
-	    Sint p = (Sint) smic.smi[i]->pending; /* on 64-bit machines...   */
-	    Eterm active;
-	    Eterm pending;
-	    if (IS_SSMALL(a))
-		active = make_small(a);
-	    else {
-		active = small_to_big(a, hp);
-		hp += BIG_UINT_HEAP_SIZE;
-	    }
-	    if (IS_SSMALL(p))
-		pending = make_small(p);
-	    else {
-		pending = small_to_big(p, hp);
-		hp += BIG_UINT_HEAP_SIZE;
-	    }
-	    item = TUPLE3(hp, smic.smi[i]->pid, active, pending);
+            ErtsMonitorSuspend *msp;
+            erts_aint_t mstate;
+	    Sint ci;
+            Eterm ct, active, pending, item;
+            Uint sz = 4 + 2;
+
+            msp = smic.smi[i];
+            mstate = erts_atomic_read_nob(&msp->state);
+
+            ci = (Sint) (mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK);
+            if (!IS_SSMALL(ci))
+                sz += BIG_UINT_HEAP_SIZE;
+
+            ERTS_PI_UNRESERVE(reserve_size, sz);
+            hp = erts_produce_heap(hfact, sz, reserve_size);
+
+            if (IS_SSMALL(ci))
+                ct = make_small(ci);
+            else {
+                ct = small_to_big(ci, hp);
+                hp += BIG_UINT_HEAP_SIZE;
+            }
+
+            if (mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE) {
+                active = ct;
+                pending = make_small(0);
+            }
+            else {
+                active = make_small(0);
+                pending = ct;
+            }
+
+            ASSERT(is_internal_pid(msp->md.origin.other.item));
+
+	    item = TUPLE3(hp, msp->md.origin.other.item, active, pending);
 	    hp += 4;
 	    res = CONS(hp, item, res);
-	    hp += 2;
 	}
 
+        *reds += (Uint) smic.smi_i / 4;
+
 	ERTS_DESTROY_SUSPEND_MONITOR_INFOS(smic);
-	ASSERT(hp == hp_end);
 
 	break;
     }
 
-    case am_dictionary:
-	if (ERTS_TRACE_FLAGS(rp) & F_SENSITIVE) {
+    case ERTS_PI_IX_DICTIONARY:
+	if (!rp->dictionary || (ERTS_TRACE_FLAGS(rp) & F_SENSITIVE)) {
 	    res = NIL;
 	} else {
-	    res = erts_dictionary_copy(BIF_P, rp->dictionary);
+            Uint num = rp->dictionary->numElements;
+	    res = erts_dictionary_copy(hfact, rp->dictionary, reserve_size);
+            *reds += (Uint) num / 4;
 	}
-	hp = HAlloc(BIF_P, 3);
+
 	break;
 
-    case am_trap_exit: {
-	erts_aint32_t state = erts_smp_atomic32_read_nob(&rp->state);
-	hp = HAlloc(BIF_P, 3);
-	if (state & ERTS_PSFLG_TRAP_EXIT)
-	    res = am_true;
-	else
-	    res = am_false;
+    case ERTS_PI_IX_TRAP_EXIT:
+        res = (rp->flags & F_TRAP_EXIT) ? am_true : am_false;
 	break;
-    }
 
-    case am_error_handler:
-	hp = HAlloc(BIF_P, 3);
-	res = erts_proc_get_error_handler(BIF_P);
+    case ERTS_PI_IX_ERROR_HANDLER:
+	res = erts_proc_get_error_handler(rp);
 	break;
 
-    case am_heap_size: {
-	Uint hsz = 3;
+    case ERTS_PI_IX_HEAP_SIZE: {
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, HEAP_SIZE(rp));
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, HEAP_SIZE(rp));
 	break;
     }
 
-    case am_fullsweep_after: {
-	Uint hsz = 3;
+    case ERTS_PI_IX_FULLSWEEP_AFTER: {
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, MAX_GEN_GCS(rp));
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, MAX_GEN_GCS(rp));
 	break;
     }
 
-    case am_min_heap_size: {
-	Uint hsz = 3;
+    case ERTS_PI_IX_MIN_HEAP_SIZE: {
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, MIN_HEAP_SIZE(rp));
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, MIN_HEAP_SIZE(rp));
 	break;
     }
 
-    case am_min_bin_vheap_size: {
-	Uint hsz = 3;
+    case ERTS_PI_IX_MIN_BIN_VHEAP_SIZE: {
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, MIN_VHEAP_SIZE(rp));
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, MIN_VHEAP_SIZE(rp));
 	break;
     }
 
-    case am_max_heap_size: {
-	Uint hsz = 3;
+    case ERTS_PI_IX_MAX_HEAP_SIZE: {
+	Uint hsz = 0;
 	(void) erts_max_heap_size_map(MAX_HEAP_SIZE_GET(rp),
                                       MAX_HEAP_SIZE_FLAGS_GET(rp),
                                       NULL, &hsz);
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_max_heap_size_map(MAX_HEAP_SIZE_GET(rp),
                                      MAX_HEAP_SIZE_FLAGS_GET(rp),
                                      &hp, NULL);
 	break;
     }
 
-    case am_total_heap_size: {
-	ErtsMessage *mp;
+    case ERTS_PI_IX_TOTAL_HEAP_SIZE: {
 	Uint total_heap_size;
-	Uint hsz = 3;
+	Uint hsz = 0;
 
 	total_heap_size = rp->heap_sz;
 	if (rp->old_hend && rp->old_heap)
@@ -1448,44 +1752,53 @@ process_info_aux(Process *BIF_P,
 
 	total_heap_size += rp->mbuf_sz;
 
-        if (rp->flags & F_ON_HEAP_MSGQ)
-            for (mp = rp->msg.first; mp; mp = mp->next)
+        if (rp->flags & F_ON_HEAP_MSGQ) {
+            ErtsMessage *mp;
+            ASSERT(flags & ERTS_PI_FLAG_NEED_MSGQ_LEN);
+            for (mp = rp->sig_qs.first; mp; mp = mp->next) {
+                ASSERT(ERTS_SIG_IS_MSG(mp));
                 if (mp->data.attached)
                     total_heap_size += erts_msg_attached_data_size(mp);
+            }
+            *reds += (Uint) rp->sig_qs.len / 4;
+        }
 
 	(void) erts_bld_uint(NULL, &hsz, total_heap_size);
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, total_heap_size);
 	break;
     }
 
-    case am_stack_size: {
+    case ERTS_PI_IX_STACK_SIZE: {
 	Uint stack_size = STACK_START(rp) - rp->stop;
-	Uint hsz = 3;
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, stack_size);
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, stack_size);
 	break;
     }
 
-    case am_memory: { /* Memory consumed in bytes */
-	Uint hsz = 3;
+    case ERTS_PI_IX_MEMORY: { /* Memory consumed in bytes */
+	Uint hsz = 0;
 	Uint size = erts_process_memory(rp, 0);
 	(void) erts_bld_uint(NULL, &hsz, size);
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, size);
+
+        ASSERT(flags & ERTS_PI_FLAG_NEED_MSGQ_LEN);
+        *reds += (Uint) rp->sig_qs.len / 4;
+
 	break;
     }
 
-    case am_garbage_collection: {
+    case ERTS_PI_IX_GARBAGE_COLLECTION: {
         DECL_AM(minor_gcs);
         Eterm t;
         Uint map_sz = 0;
 
         erts_max_heap_size_map(MAX_HEAP_SIZE_GET(rp), MAX_HEAP_SIZE_FLAGS_GET(rp), NULL, &map_sz);
 
-	hp = HAlloc(BIF_P, 3+2 + 3+2 + 3+2 + 3+2 + 3+2 + map_sz + 3);
-        /* last "3" is for outside tuple */
+        hp = erts_produce_heap(hfact, 3+2 + 3+2 + 3+2 + 3+2 + 3+2 + map_sz, reserve_size);
 
 	t = TUPLE2(hp, AM_minor_gcs, make_small(GEN_GCS(rp))); hp += 3;
 	res = CONS(hp, t, NIL); hp += 2;
@@ -1504,89 +1817,76 @@ process_info_aux(Process *BIF_P,
 	break;
     }
 
-    case am_garbage_collection_info: {
+    case ERTS_PI_IX_GARBAGE_COLLECTION_INFO: {
         Uint sz = 0, actual_sz = 0;
 
-        if (rp == BIF_P) {
-            sz += ERTS_PROCESS_GC_INFO_MAX_SIZE;
-        } else {
-            erts_process_gc_info(rp, &sz, NULL, 0, 0);
-            sz += 3;
-        }
+        erts_process_gc_info(rp, &sz, NULL, 0, 0);
 
-        hp = HAlloc(BIF_P, sz);
+        hp = erts_produce_heap(hfact, sz, reserve_size);
         res = erts_process_gc_info(rp, &actual_sz, &hp, 0, 0);
 
-        /* We may have some extra space, fill with 0 tuples */
-        if (actual_sz <= sz - 3) {
-            for (; actual_sz < sz - 3; hp++, actual_sz++)
-                hp[0] = make_arityval(0);
-        } else {
-            for (; actual_sz < sz; hp++, actual_sz++)
-                hp[0] = make_arityval(0);
-            hp = HAlloc(BIF_P, 3);
-        }
-
         break;
     }
 
-    case am_group_leader: {
+    case ERTS_PI_IX_GROUP_LEADER: {
 	int sz = NC_HEAP_SIZE(rp->group_leader);
-	hp = HAlloc(BIF_P, 3 + sz);
-	res = STORE_NC(&hp, &MSO(BIF_P), rp->group_leader);
+        hp = erts_produce_heap(hfact, sz, reserve_size);
+	res = STORE_NC(&hp, hfact->off_heap, rp->group_leader);
 	break;
     }
 
-    case am_reductions: {
-	Uint reds = rp->reds + erts_current_reductions(BIF_P, rp);
-	Uint hsz = 3;
+    case ERTS_PI_IX_REDUCTIONS: {
+	Uint reds = rp->reds + erts_current_reductions(c_p, rp);
+	Uint hsz = 0;
 	(void) erts_bld_uint(NULL, &hsz, reds);
-	hp = HAlloc(BIF_P, hsz);
+        hp = erts_produce_heap(hfact, hsz, reserve_size);
 	res = erts_bld_uint(&hp, NULL, reds);
 	break;
     }
 
-    case am_priority:
-	hp = HAlloc(BIF_P, 3);
-	res = erts_get_process_priority(rp);
+    case ERTS_PI_IX_PRIORITY: {
+        erts_aint32_t state = erts_atomic32_read_nob(&rp->state);
+        if (ERTS_PSFLG_EXITING & state)
+            return am_undefined;
+	res = erts_get_process_priority(state);
 	break;
+    }
 
-    case am_trace:
-	hp = HAlloc(BIF_P, 3);
+    case ERTS_PI_IX_TRACE:
 	res = make_small(ERTS_TRACE_FLAGS(rp) & TRACEE_FLAGS);
 	break;
 
-    case am_binary: {
-	Uint sz = 3;
+    case ERTS_PI_IX_BINARY: {
+	Uint sz = 0;
 	(void) bld_bin_list(NULL, &sz, &MSO(rp));
-	hp = HAlloc(BIF_P, sz);
+        hp = erts_produce_heap(hfact, sz, reserve_size);
 	res = bld_bin_list(&hp, NULL, &MSO(rp));
 	break;
     }
 
-    case am_sequential_trace_token:
-	res = copy_object(rp->seq_trace_token, BIF_P);
-	hp = HAlloc(BIF_P, 3);
+    case ERTS_PI_IX_SEQUENTIAL_TRACE_TOKEN: {
+        Uint sz = size_object(rp->seq_trace_token);
+        hp = erts_produce_heap(hfact, sz, reserve_size);
+        res = copy_struct(rp->seq_trace_token, sz, &hp, hfact->off_heap);
 	break;
+    }
 
-    case am_catchlevel:
-	hp = HAlloc(BIF_P, 3);
-	res = make_small(catchlevel(BIF_P));
+    case ERTS_PI_IX_CATCHLEVEL:
+	res = make_small(catchlevel(rp));
 	break;
 
-    case am_backtrace: {
+    case ERTS_PI_IX_BACKTRACE: {
 	erts_dsprintf_buf_t *dsbufp = erts_create_tmp_dsbuf(0);
 	erts_stack_dump(ERTS_PRINT_DSBUF, (void *) dsbufp, rp);
-	res = new_binary(BIF_P, (byte *) dsbufp->str, dsbufp->str_len);
+	res = erts_heap_factory_new_binary(hfact, (byte *) dsbufp->str,
+                                           dsbufp->str_len, reserve_size);
 	erts_destroy_tmp_dsbuf(dsbufp);
-	hp = HAlloc(BIF_P, 3);
 	break;
     }
 
-    case am_last_calls: {
+    case ERTS_PI_IX_LAST_CALLS: {
 	struct saved_calls *scb = ERTS_PROC_GET_SAVED_CALLS_BUF(rp);
 	if (!scb) {
-	    hp = HAlloc(BIF_P, 3);
 	    res = am_false;
 	} else {
 	    /*
@@ -1594,23 +1894,34 @@ process_info_aux(Process *BIF_P,
 	     * Might be less than that, if there are sends, receives or timeouts,
 	     * so we must do a HRelease() to avoid creating holes.
 	     */
-	    Uint needed = scb->n*(2+4) + 3;
-	    Eterm* limit;
+	    Sint needed = scb->n*(2+4);
 	    Eterm term, list;
 	    int i, j;
+            Export *exp;
+
+            reserve_size += needed;
 
-	    hp = HAlloc(BIF_P, needed);
-	    limit = hp + needed;
 	    list = NIL;
 	    for (i = 0; i < scb->n; i++) {
+                Uint sz;
 		j = scb->cur - i - 1;
 		if (j < 0)
 		    j += scb->len;
-		if (scb->ct[j] == &exp_send)
+
+                sz = 2;
+                exp = scb->ct[j];
+                if (exp != &exp_send && exp != &exp_receive && exp != &exp_timeout)
+                    sz += 4;
+
+                needed -= sz;
+                ERTS_PI_UNRESERVE(reserve_size, sz);
+                hp = erts_produce_heap(hfact, sz, reserve_size);
+
+		if (exp == &exp_send)
 		    term = am_send;
-		else if (scb->ct[j] == &exp_receive)
+		else if (exp == &exp_receive)
 		    term = am_receive;
-		else if (scb->ct[j] == &exp_timeout)
+		else if (exp == &exp_timeout)
 		    term = am_timeout;
 		else {
 		    term = TUPLE3(hp,
@@ -1620,18 +1931,18 @@ process_info_aux(Process *BIF_P,
 		    hp += 4;
 		}
 		list = CONS(hp, term, list);
-		hp += 2;
 	    }
+
+            ASSERT(needed >= 0);
+            if (needed > 0)
+                reserve_size -= needed;
+
 	    res = list;
-	    res = TUPLE2(hp, item, res);
-	    hp += 3;
-	    HRelease(BIF_P,limit,hp);
-	    return res;
 	}
 	break;
     }
 
-    case am_message_queue_data:
+    case ERTS_PI_IX_MESSAGE_QUEUE_DATA:
 	switch (rp->flags & (F_OFF_HEAP_MSGQ|F_ON_HEAP_MSGQ)) {
 	case F_OFF_HEAP_MSGQ:
 	    res = am_off_heap;
@@ -1644,14 +1955,15 @@ process_info_aux(Process *BIF_P,
 	    ERTS_INTERNAL_ERROR("Inconsistent message queue management state");
 	    break;
 	}
-	hp = HAlloc(BIF_P, 3);
 	break;
 
-    case am_magic_ref: {
-	Uint sz = 3;
+    case ERTS_PI_IX_MAGIC_REF: {
+	Uint sz = 0;
 	(void) bld_magic_ref_bin_list(NULL, &sz, &MSO(rp));
-	hp = HAlloc(BIF_P, sz);
+        hp = erts_produce_heap(hfact, sz, 0);
 	res = bld_magic_ref_bin_list(&hp, NULL, &MSO(rp));
+
+        *reds += (Uint) 10;
 	break;
     }
 
@@ -1660,12 +1972,17 @@ process_info_aux(Process *BIF_P,
 
     }
 
-    return TUPLE2(hp, item, res);
+    ERTS_PI_UNRESERVE(reserve_size, 3);
+    *reserve_sizep = reserve_size;
+    hp = erts_produce_heap(hfact, 3, reserve_size);
+
+    return TUPLE2(hp, pi_ix2arg(item_ix), res);
 }
 #undef MI_INC
 
 static Eterm
-current_function(Process* BIF_P, Process* rp, Eterm** hpp, int full_info)
+current_function(Process *c_p, ErtsHeapFactory *hfact, Process* rp,
+                 int full_info, Uint reserve_size, int flags)
 {
     Eterm* hp;
     Eterm res;
@@ -1682,7 +1999,7 @@ current_function(Process* BIF_P, Process* rp, Eterm** hpp, int full_info)
 	}
     }
 
-    if (BIF_P == rp) {
+    if (c_p == rp && !(flags & ERTS_PI_FLAG_REQUEST_FOR_OTHER)) {
 	FunctionInfo fi2;
 
 	/*
@@ -1702,24 +2019,22 @@ current_function(Process* BIF_P, Process* rp, Eterm** hpp, int full_info)
      * Return the result.
      */
     if (rp->current == NULL) {
-	hp = HAlloc(BIF_P, 3);
 	res = am_undefined;
     } else if (full_info) {
-	hp = HAlloc(BIF_P, 3+fi.needed);
-	hp = erts_build_mfa_item(&fi, hp, am_true, &res);
+        hp = erts_produce_heap(hfact, fi.needed, reserve_size);
+	erts_build_mfa_item(&fi, hp, am_true, &res);
     } else {
-	hp = HAlloc(BIF_P, 3+4);
+        hp = erts_produce_heap(hfact, 4, reserve_size);
 	res = TUPLE3(hp, rp->current->module,
 		     rp->current->function,
                      make_small(rp->current->arity));
-	hp += 4;
     }
-    *hpp = hp;
     return res;
 }
 
 static Eterm
-current_stacktrace(Process* p, Process* rp, Eterm** hpp)
+current_stacktrace(ErtsHeapFactory *hfact, Process* rp,
+                   Uint reserve_size)
 {
     Uint sz;
     struct StackTrace* s;
@@ -1728,7 +2043,7 @@ current_stacktrace(Process* p, Process* rp, Eterm** hpp)
     FunctionInfo* stkp;
     Uint heap_size;
     int i;
-    Eterm* hp = *hpp;
+    Eterm* hp;
     Eterm mfa;
     Eterm res = NIL;
 
@@ -1758,17 +2073,23 @@ current_stacktrace(Process* p, Process* rp, Eterm** hpp)
 	}
     }
 
-    hp = HAlloc(p, heap_size);
+    reserve_size += heap_size;
+
+    /*
+     * We intentionally produce heap in small chunks
+     * (for more info see process_info_aux()).
+     */
     while (stkp > stk) {
 	stkp--;
+        sz = stkp->needed + 2;
+        ERTS_PI_UNRESERVE(reserve_size, sz);
+        hp = erts_produce_heap(hfact, sz, reserve_size);
 	hp = erts_build_mfa_item(stkp, hp, am_true, &mfa);
 	res = CONS(hp, mfa, res);
-	hp += 2;
     }
 
     erts_free(ERTS_ALC_T_TMP, stk);
     erts_free(ERTS_ALC_T_TMP, s);
-    *hpp = hp;
     return res;
 }
 
@@ -1777,7 +2098,7 @@ static int check_if_xml(void)
 {
     char buf[1];
     size_t bufsz = sizeof(buf);
-    return erts_sys_getenv_raw("VALGRIND_LOG_XML", buf, &bufsz) >= 0;
+    return erts_sys_explicit_8bit_getenv("VALGRIND_LOG_XML", buf, &bufsz) >= 0;
 }
 #else
 #define check_if_xml() 0
@@ -1826,45 +2147,6 @@ info_1_tuple(Process* BIF_P,	/* Pointer to current process. */
 	    return make_small(sizeof(UWord));
 	}
 	goto badarg;
-    } else if (sel == am_allocated) {
-	if (arity == 2) {
-	    Eterm res = THE_NON_VALUE;
-	    char *buf;
-	    Sint len = is_string(*tp);
-	    if (len <= 0)
-		return res;
-	    buf = (char *) erts_alloc(ERTS_ALC_T_TMP, len+1);
-	    if (intlist_to_buf(*tp, buf, len) != len)
-		erts_exit(ERTS_ERROR_EXIT, "%s:%d: Internal error\n", __FILE__, __LINE__);
-	    buf[len] = '\0';
-	    res = erts_instr_dump_memory_map(buf) ? am_true : am_false;
-	    erts_free(ERTS_ALC_T_TMP, (void *) buf);
-	    if (is_non_value(res))
-		goto badarg;
-	    return res;
-	}
-	else if (arity == 3 && tp[0] == am_status) {
-	    if (is_atom(tp[1]))
-		return erts_instr_get_stat(BIF_P, tp[1], 1);
-	    else {
-		Eterm res = THE_NON_VALUE;
-		char *buf;
-		Sint len = is_string(tp[1]);
-		if (len <= 0)
-		    return res;
-		buf = (char *) erts_alloc(ERTS_ALC_T_TMP, len+1);
-		if (intlist_to_buf(tp[1], buf, len) != len)
-		    erts_exit(ERTS_ERROR_EXIT, "%s:%d: Internal error\n", __FILE__, __LINE__);
-		buf[len] = '\0';
-		res = erts_instr_dump_stat(buf, 1) ? am_true : am_false;
-		erts_free(ERTS_ALC_T_TMP, (void *) buf);
-		if (is_non_value(res))
-		    goto badarg;
-		return res;
-	    }
-	}
-	else
-	    goto badarg;
     } else if (sel == am_allocator) {
 	switch (arity) {
 	case 2:
@@ -2145,14 +2427,6 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	ASSERT(erts_compat_rel > 0);
 	BIF_RET(make_small(erts_compat_rel));
     } else if (BIF_ARG_1 == am_multi_scheduling) {
-#ifndef ERTS_SMP
-	BIF_RET(am_disabled);
-#else
-#ifndef ERTS_DIRTY_SCHEDULERS
-	if (erts_no_schedulers == 1)
-	    BIF_RET(am_disabled);
-	else
-#endif
 	{
 	    int msb = erts_is_multi_scheduling_blocked();
 	    BIF_RET(!msb
@@ -2161,7 +2435,6 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 		       ? am_blocked
 		       : am_blocked_normal));
 	}
-#endif
     } else if (BIF_ARG_1 == am_build_type) {
 #if defined(DEBUG)
 	ERTS_DECL_AM(debug);
@@ -2241,8 +2514,6 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
     } else if (BIF_ARG_1 == am_allocated_areas) {
 	res = erts_allocated_areas(NULL, NULL, BIF_P);
 	BIF_RET(res);
-    } else if (BIF_ARG_1 == am_allocated) {
-	BIF_RET(erts_instr_get_memory_map(BIF_P));
     } else if (BIF_ARG_1 == am_hipe_architecture) {
 #if defined(HIPE)
 	BIF_RET(hipe_arch_name);
@@ -2271,7 +2542,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	res = TUPLE2(hp, am_sequential_tracer, val);
 	BIF_RET(res);
     } else if (BIF_ARG_1 == am_garbage_collection){
-	Uint val = (Uint) erts_smp_atomic32_read_nob(&erts_max_gen_gcs);
+	Uint val = (Uint) erts_atomic32_read_nob(&erts_max_gen_gcs);
 	Eterm tup;
 	hp = HAlloc(BIF_P, 3+2 + 3+2 + 3+2 + 3+2);
 
@@ -2289,7 +2560,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 
 	BIF_RET(res);
     } else if (BIF_ARG_1 == am_fullsweep_after){
-	Uint val = (Uint) erts_smp_atomic32_read_nob(&erts_max_gen_gcs);
+	Uint val = (Uint) erts_atomic32_read_nob(&erts_max_gen_gcs);
 	hp = HAlloc(BIF_P, 3);
 	res = TUPLE2(hp, am_fullsweep_after, make_small(val));
 	BIF_RET(res);
@@ -2322,8 +2593,8 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	erts_dsprintf_buf_t *dsbufp = erts_create_info_dsbuf(0);
 
 	/* Need to be the only thread running... */
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	if (BIF_ARG_1 == am_info)
 	    info(ERTS_PRINT_DSBUF, (void *) dsbufp);
@@ -2334,8 +2605,8 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	else
 	    distribution_info(ERTS_PRINT_DSBUF, (void *) dsbufp);
 
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
 	ASSERT(dsbufp && dsbufp->str);
 	res = new_binary(BIF_P, (byte *) dsbufp->str, dsbufp->str_len);
@@ -2344,7 +2615,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
     } else if (ERTS_IS_ATOM_STR("dist_ctrl", BIF_ARG_1)) {
 	DistEntry *dep;
 	i = 0;
-        erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+        erts_rwmtx_rlock(&erts_dist_table_rwmtx);
 	for (dep = erts_visible_dist_entries; dep; dep = dep->next) 
 	    ++i;
 	for (dep = erts_hidden_dist_entries; dep; dep = dep->next)
@@ -2367,7 +2638,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	    res = CONS(hp, tpl, res);
 	    hp += 2;
 	}
-        erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+        erts_rwmtx_runlock(&erts_dist_table_rwmtx);
 	BIF_RET(res);
     } else if (BIF_ARG_1 == am_system_version) {
 	erts_dsprintf_buf_t *dsbufp = erts_create_tmp_dsbuf(0);
@@ -2383,9 +2654,6 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 			       sizeof(ERLANG_ARCHITECTURE)-1,
 			       NIL));
     } 
-    else if (BIF_ARG_1 == am_memory_types) {
-	return erts_instr_get_type_info(BIF_P);
-    }
     else if (BIF_ARG_1 == am_os_type) {
 	BIF_RET(os_type_tuple);
     }
@@ -2393,16 +2661,10 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(erts_allocator_options((void *) BIF_P));
     }
     else if (BIF_ARG_1 == am_thread_pool_size) {
-#ifdef USE_THREADS
 	extern int erts_async_max_threads;
-#endif
 	int n;
 	
-#ifdef USE_THREADS
 	n = erts_async_max_threads;
-#else
-	n = 0;
-#endif
 	BIF_RET(make_small(n));
     }
     else if (BIF_ARG_1 == am_alloc_util_allocators) {
@@ -2416,12 +2678,12 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(os_version_tuple);
     }
     else if (BIF_ARG_1 == am_version) {
-	int n = strlen(ERLANG_VERSION);
+	int n = sys_strlen(ERLANG_VERSION);
 	hp = HAlloc(BIF_P, ((sizeof ERLANG_VERSION)-1) * 2);
 	BIF_RET(buf_to_intlist(&hp, ERLANG_VERSION, n, NIL));
     }
     else if (BIF_ARG_1 == am_machine) {
-	int n = strlen(EMULATOR);
+	int n = sys_strlen(EMULATOR);
 	hp = HAlloc(BIF_P, n*2);
 	BIF_RET(buf_to_intlist(&hp, EMULATOR, n, NIL));
     }
@@ -2447,7 +2709,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	    res = erts_bld_cons(hpp, hszp,
 				erts_bld_tuple(hpp, hszp, 2,
 					       erts_atom_put((byte *)opc[i].name,
-							     strlen(opc[i].name),
+							     sys_strlen(opc[i].name),
 							     ERTS_ATOM_ENC_LATIN1,
 							     1),
 					       erts_bld_uint(hpp, hszp,
@@ -2470,7 +2732,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 #endif
 
 	BIF_RET(res);
-#endif /* #ifndef ERTS_SMP */
+#endif /* #ifndef ERTS_OPCODE_COUNTER_SUPPORT */
     } else if (BIF_ARG_1 == am_wordsize) {
 	return make_small(sizeof(Eterm));
     } else if (BIF_ARG_1 == am_endian) {
@@ -2550,11 +2812,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(res);
 #endif
     } else if (BIF_ARG_1 == am_threads) {
-#ifdef USE_THREADS
 	return am_true;
-#else
-	return am_false;
-#endif
     } else if (BIF_ARG_1 == am_creation) {
 	return make_small(erts_this_node->creation);
     } else if (BIF_ARG_1 == am_break_ignored) {
@@ -2613,11 +2871,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	hp = HAlloc(BIF_P, 2*n);
 	BIF_RET(buf_to_intlist(&hp, buf, n, NIL));
     } else if (ERTS_IS_ATOM_STR("smp_support", BIF_ARG_1)) {
-#ifdef ERTS_SMP
 	BIF_RET(am_true);
-#else
-	BIF_RET(am_false);
-#endif
     } else if (ERTS_IS_ATOM_STR("scheduler_bind_type", BIF_ARG_1)) {
 	BIF_RET(erts_bound_schedulers_term(BIF_P));
     } else if (ERTS_IS_ATOM_STR("scheduler_bindings", BIF_ARG_1)) {
@@ -2629,11 +2883,6 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	res = make_small(erts_no_schedulers);
 	BIF_RET(res);
     } else if (ERTS_IS_ATOM_STR("schedulers_state", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	Eterm *hp = HAlloc(BIF_P, 4);
-	res = TUPLE3(hp, make_small(1), make_small(1), make_small(1));
-	BIF_RET(res);
-#else
 	Eterm *hp;
 	Uint total, online, active;
 	erts_schedulers_state(&total, &online, &active,
@@ -2644,13 +2893,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 		     make_small(online),
 		     make_small(active));
 	BIF_RET(res);
-#endif
     } else if (ERTS_IS_ATOM_STR("schedulers_state", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	Eterm *hp = HAlloc(BIF_P, 4);
-	res = TUPLE3(hp, make_small(1), make_small(1), make_small(1));
-	BIF_RET(res);
-#else
 	Eterm *hp;
 	Uint total, online, active;
 	erts_schedulers_state(&total, &online, &active,
@@ -2661,19 +2904,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 		     make_small(online),
 		     make_small(active));
 	BIF_RET(res);
-#endif
     } else if (ERTS_IS_ATOM_STR("all_schedulers_state", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	Eterm *hp = HAlloc(BIF_P, 2+5);
-	res = CONS(hp+5,
-		   TUPLE4(hp,
-			  am_normal,
-			  make_small(1),
-			  make_small(1),
-			  make_small(1)),
-		   NIL);
-	BIF_RET(res);
-#else
 	Eterm *hp, tpl;
 	Uint sz, total, online, active,
 	    dirty_cpu_total, dirty_cpu_online, dirty_cpu_active,
@@ -2719,46 +2950,25 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	hp += 5;
 	res = CONS(hp, tpl, res);
 	BIF_RET(res);
-#endif
     } else if (ERTS_IS_ATOM_STR("schedulers_online", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	BIF_RET(make_small(1));
-#else
 	Uint online;
 	erts_schedulers_state(NULL, &online, NULL, NULL, NULL, NULL, NULL, NULL);
 	BIF_RET(make_small(online));
-#endif
     } else if (ERTS_IS_ATOM_STR("schedulers_active", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	BIF_RET(make_small(1));
-#else
 	Uint active;
 	erts_schedulers_state(NULL, NULL, &active, NULL, NULL, NULL, NULL, NULL);
 	BIF_RET(make_small(active));
-#endif
     } else if (ERTS_IS_ATOM_STR("dirty_cpu_schedulers", BIF_ARG_1)) {
 	Uint dirty_cpu;
-#ifdef ERTS_DIRTY_SCHEDULERS
 	erts_schedulers_state(NULL, NULL, NULL, &dirty_cpu, NULL, NULL, NULL, NULL);
-#else
-        dirty_cpu = 0;
-#endif
 	BIF_RET(make_small(dirty_cpu));
     } else if (ERTS_IS_ATOM_STR("dirty_cpu_schedulers_online", BIF_ARG_1)) {
 	Uint dirty_cpu_onln;
-#ifdef ERTS_DIRTY_SCHEDULERS
 	erts_schedulers_state(NULL, NULL, NULL, NULL, &dirty_cpu_onln, NULL, NULL, NULL);
-#else
-        dirty_cpu_onln = 0;
-#endif
 	BIF_RET(make_small(dirty_cpu_onln));
     } else if (ERTS_IS_ATOM_STR("dirty_io_schedulers", BIF_ARG_1)) {
 	Uint dirty_io;
-#ifdef ERTS_DIRTY_SCHEDULERS
 	erts_schedulers_state(NULL, NULL, NULL, NULL, NULL, NULL, &dirty_io, NULL);
-#else
-        dirty_io = 0;
-#endif
 	BIF_RET(make_small(dirty_io));
     } else if (ERTS_IS_ATOM_STR("run_queues", BIF_ARG_1)) {
 	res = make_small(erts_no_run_queues);
@@ -2780,7 +2990,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(make_small(CONTEXT_REDS));
     } else if (ERTS_IS_ATOM_STR("kernel_poll", BIF_ARG_1)) {
 #ifdef ERTS_ENABLE_KERNEL_POLL
-	BIF_RET(erts_use_kernel_poll ? am_true : am_false);
+	BIF_RET(am_true);
 #else
 	BIF_RET(am_false);
 #endif    
@@ -2805,23 +3015,15 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
     } else if (ERTS_IS_ATOM_STR("check_io", BIF_ARG_1)) {
 	BIF_RET(erts_check_io_info(BIF_P));
     } else if (ERTS_IS_ATOM_STR("multi_scheduling_blockers", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	BIF_RET(NIL);
-#else
 	if (erts_no_schedulers == 1)
 	    BIF_RET(NIL);
 	else
 	    BIF_RET(erts_multi_scheduling_blockers(BIF_P, 0));
-#endif
     } else if (ERTS_IS_ATOM_STR("normal_multi_scheduling_blockers", BIF_ARG_1)) {
-#ifndef ERTS_SMP
-	BIF_RET(NIL);
-#else
 	if (erts_no_schedulers == 1)
 	    BIF_RET(NIL);
 	else
 	    BIF_RET(erts_multi_scheduling_blockers(BIF_P, 1));
-#endif
     } else if (ERTS_IS_ATOM_STR("modified_timing_level", BIF_ARG_1)) {
 	BIF_RET(ERTS_USE_MODIFIED_TIMING()
 		? make_small(erts_modified_timing_level)
@@ -2884,12 +3086,10 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(am_false);
 #endif	
     }
-#ifdef ERTS_SMP
     else if (ERTS_IS_ATOM_STR("thread_progress", BIF_ARG_1)) {
 	erts_thr_progress_dbg_print_state();
 	BIF_RET(am_true);
     }
-#endif
     else if (BIF_ARG_1 == am_message_queue_data) {
 	switch (erts_default_spo_flags & (SPO_ON_HEAP_MSGQ|SPO_OFF_HEAP_MSGQ)) {
 	case SPO_OFF_HEAP_MSGQ:
@@ -2905,21 +3105,21 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	Uint  sz;
 	Eterm res = NIL, tup, text;
 	Eterm *hp = HAlloc(BIF_P, 3*(2 + 3) + /* three 2-tuples and three cons */
-		2*(strlen(erts_build_flags_CONFIG_H) +
-		   strlen(erts_build_flags_CFLAGS) +
-		   strlen(erts_build_flags_LDFLAGS)));
+		2*(sys_strlen(erts_build_flags_CONFIG_H) +
+		   sys_strlen(erts_build_flags_CFLAGS) +
+		   sys_strlen(erts_build_flags_LDFLAGS)));
 
-	sz   = strlen(erts_build_flags_CONFIG_H);
+	sz   = sys_strlen(erts_build_flags_CONFIG_H);
 	text = buf_to_intlist(&hp, erts_build_flags_CONFIG_H, sz, NIL);
 	tup  = TUPLE2(hp, am_config_h, text); hp += 3;
 	res  = CONS(hp, tup, res); hp += 2;
 
-	sz   = strlen(erts_build_flags_CFLAGS);
+	sz   = sys_strlen(erts_build_flags_CFLAGS);
 	text = buf_to_intlist(&hp, erts_build_flags_CFLAGS, sz, NIL);
 	tup  = TUPLE2(hp, am_cflags, text); hp += 3;
 	res  = CONS(hp, tup, res); hp += 2;
 
-	sz   = strlen(erts_build_flags_LDFLAGS);
+	sz   = sys_strlen(erts_build_flags_LDFLAGS);
 	text = buf_to_intlist(&hp, erts_build_flags_LDFLAGS, sz, NIL);
 	tup  = TUPLE2(hp, am_ldflags, text); hp += 3;
 	res  = CONS(hp, tup, res); hp += 2;
@@ -2929,6 +3129,9 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
     else if (ERTS_IS_ATOM_STR("ets_limit",BIF_ARG_1)) {
         BIF_RET(make_small(erts_db_get_max_tabs()));
     }
+    else if (ERTS_IS_ATOM_STR("ets_count",BIF_ARG_1)) {
+        BIF_RET(make_small(erts_ets_table_count()));
+    }
     else if (ERTS_IS_ATOM_STR("atom_limit",BIF_ARG_1)) {
         BIF_RET(make_small(erts_get_atom_limit()));
     }
@@ -2943,7 +3146,7 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
 	BIF_RET(am_disabled);
     }
     else if (ERTS_IS_ATOM_STR("eager_check_io",BIF_ARG_1)) {
-	BIF_RET(erts_eager_check_io ? am_true : am_false);
+	BIF_RET(am_true);
     }
     else if (ERTS_IS_ATOM_STR("literal_test",BIF_ARG_1)) {
 #ifdef ERTS_HAVE_IS_IN_LITERAL_RANGE
@@ -2963,6 +3166,16 @@ BIF_RETTYPE system_info_1(BIF_ALIST_1)
     BIF_ERROR(BIF_P, BADARG);
 }
 
+static void monitor_size(ErtsMonitor *mon, void *vsz)
+{
+    *((Uint *) vsz) = erts_monitor_size(mon);
+}
+
+static void link_size(ErtsMonitor *lnk, void *vsz)
+{
+    *((Uint *) vsz) = erts_link_size(lnk);
+}
+
 /**********************************************************************/ 
 /* Return information on ports */
 /* Info:
@@ -2981,7 +3194,7 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 {
     Eterm res = THE_NON_VALUE;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     if (item == am_id) {
 	if (hpp)
@@ -2998,7 +3211,7 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 
 	INIT_MONITOR_INFOS(mic);
 
-	erts_doforall_links(ERTS_P_LINKS(prt), &collect_one_link, &mic);
+	erts_link_tree_foreach(ERTS_P_LINKS(prt), collect_one_link, (void *) &mic);
 
 	if (szp)
 	    *szp += mic.sz;
@@ -3022,11 +3235,11 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
     else if (item == am_monitors) {
 	MonitorInfoCollection mic;
 	int i;
-	Eterm item;
 
 	INIT_MONITOR_INFOS(mic);
-        erts_doforall_monitors(ERTS_P_MONITORS(prt),
-                               &collect_one_origin_monitor, &mic);
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(prt),
+                                  collect_one_origin_monitor,
+                                  (void *) &mic);
 
 	if (szp)
 	    *szp += mic.sz;
@@ -3035,11 +3248,10 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 	    res = NIL;
 	    for (i = 0; i < mic.mi_i; i++) {
 		Eterm t;
-                Eterm m_type;
 
-                item = STORE_NC(hpp, ohp, mic.mi[i].entity.term);
-                m_type = is_port(item) ? am_port : am_process;
-                t = TUPLE2(*hpp, m_type, item);
+                ASSERT(mic.mi[i].type == ERTS_MON_TYPE_PORT);
+                ASSERT(is_internal_pid(mic.mi[i].entity.term));
+                t = TUPLE2(*hpp, am_process, mic.mi[i].entity.term);
 		*hpp += 3;
 		res = CONS(*hpp, t, res);
 		*hpp += 2;
@@ -3058,15 +3270,19 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
         Eterm item;
 
         INIT_MONITOR_INFOS(mic);
-        erts_doforall_monitors(ERTS_P_MONITORS(prt),
-                               &collect_one_target_monitor, &mic);
+        erts_monitor_list_foreach(ERTS_P_LT_MONITORS(prt),
+                                  collect_one_target_monitor,
+                                  (void *) &mic);
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(prt),
+                                  collect_one_target_monitor,
+                                  (void *) &mic);
         if (szp)
             *szp += mic.sz;
 
         if (hpp) {
             res = NIL;
             for (i = 0; i < mic.mi_i; ++i) {
-                ASSERT(mic.mi[i].node == NIL);
+                ASSERT(mic.mi[i].type != ERTS_MON_TYPE_RESOURCE);
                 item = STORE_NC(hpp, ohp, mic.mi[i].entity.term);
                 res = CONS(*hpp, item, res);
                 *hpp += 2;
@@ -3143,7 +3359,12 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 	 */
 	Uint size = 0;
 
-	erts_doforall_links(ERTS_P_LINKS(prt), &erts_one_link_size, &size);
+        erts_link_tree_foreach(ERTS_P_LINKS(prt),
+                               link_size, (void *) &size);
+        erts_monitor_tree_foreach(ERTS_P_MONITORS(prt),
+                                  monitor_size, (void *) &size);
+        erts_monitor_list_foreach(ERTS_P_LT_MONITORS(prt),
+                                  monitor_size, (void *) &size);
 
 	size += erts_port_data_size(prt);
 
@@ -3172,9 +3393,6 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
     }
     else if (ERTS_IS_ATOM_STR("locking", item)) {
 	if (hpp) {
-#ifndef ERTS_SMP
-	    res = am_false;
-#else
 	    if (erts_atomic32_read_nob(&prt->state)
 		& ERTS_PORT_SFLG_PORT_SPECIFIC_LOCK) {
 		DECL_AM(port_level);
@@ -3188,7 +3406,6 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 			 & ERL_DRV_FLAG_USE_PORT_LOCKING));
 		res = AM_driver_level;
 	    }
-#endif
 	}
 	if (szp) {
 	    res = am_true;
@@ -3201,7 +3418,7 @@ erts_bld_port_info(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, Port *prt,
 	    goto done;
 	}
 	res = ((ERTS_PTS_FLG_PARALLELISM &
-		erts_smp_atomic32_read_nob(&prt->sched.flags))
+		erts_atomic32_read_nob(&prt->sched.flags))
 	       ? am_true
 	       : am_false);
     }
@@ -3277,7 +3494,7 @@ fun_info_2(BIF_ALIST_2)
 	    }
 	    break;
 	case am_refc:
-	    val = erts_make_integer(erts_smp_atomic_read_nob(&funp->fe->refc), p);
+	    val = erts_make_integer(erts_atomic_read_nob(&funp->fe->refc), p);
 	    hp = HAlloc(p, 3);
 	    break;
 	case am_arity:
@@ -3369,69 +3586,95 @@ fun_info_mfa_1(BIF_ALIST_1)
     BIF_ERROR(p, BADARG);
 }
 
+BIF_RETTYPE erts_internal_is_process_alive_2(BIF_ALIST_2)
+{
+    if (!is_internal_pid(BIF_ARG_1) || !is_internal_ordinary_ref(BIF_ARG_2))
+        BIF_ERROR(BIF_P, BADARG);
+    erts_proc_sig_send_is_alive_request(BIF_P, BIF_ARG_1, BIF_ARG_2);
+    BIF_RET(am_ok);
+}
+
 BIF_RETTYPE is_process_alive_1(BIF_ALIST_1) 
 {
-   if(is_internal_pid(BIF_ARG_1)) {
-       Process *rp;
-
-       if (BIF_ARG_1 == BIF_P->common.id)
-	   BIF_RET(am_true);
-
-       rp = erts_proc_lookup_raw(BIF_ARG_1);
-       if (!rp) {
-	   BIF_RET(am_false);
-       }
-       else {
-	   if (erts_smp_atomic32_read_acqb(&rp->state)
-		 & (ERTS_PSFLG_PENDING_EXIT|ERTS_PSFLG_EXITING))
-	       ERTS_BIF_AWAIT_X_DATA_TRAP(BIF_P, BIF_ARG_1, am_false);
-	   else
-	       BIF_RET(am_true);
-       }
-   }
-   else if(is_external_pid(BIF_ARG_1)) {
-       if(external_pid_dist_entry(BIF_ARG_1) == erts_this_dist_entry)
+    if (is_internal_pid(BIF_ARG_1)) {
+        erts_aint32_t state;
+        Process *rp;
+
+        if (BIF_ARG_1 == BIF_P->common.id)
+            BIF_RET(am_true);
+
+        rp = erts_proc_lookup_raw(BIF_ARG_1);
+        if (!rp)
+            BIF_RET(am_false);
+
+        state = erts_atomic32_read_acqb(&rp->state);
+        if (state & (ERTS_PSFLG_EXITING
+                     | ERTS_PSFLG_SIG_Q
+                     | ERTS_PSFLG_SIG_IN_Q)) {
+            /*
+             * If in exiting state, trap out and send 'is alive'
+             * request and wait for it to complete termination.
+             *
+             * If process has signals enqueued, we need to
+             * send it an 'is alive' request via its signal
+             * queue in order to ensure that signal order is
+             * preserved (we may earlier have sent it an
+             * exit signal that has not been processed yet).
+             */
+            BIF_TRAP1(is_process_alive_trap, BIF_P, BIF_ARG_1);
+        }
+
+        BIF_RET(am_true);
+    }
+
+   if (is_external_pid(BIF_ARG_1)) {
+       if (external_pid_dist_entry(BIF_ARG_1) == erts_this_dist_entry)
 	   BIF_RET(am_false); /* A pid from an old incarnation of this node */
-       else
-	   BIF_ERROR(BIF_P, BADARG);
-   }
-   else {
-      BIF_ERROR(BIF_P, BADARG);
    }
+
+   BIF_ERROR(BIF_P, BADARG);
 }
 
-BIF_RETTYPE process_display_2(BIF_ALIST_2)
+static Eterm
+process_display(Process *c_p, void *arg, int *redsp, ErlHeapFragment **bpp)
 {
-   Process *rp;
+    if (redsp)
+        *redsp = 1;
 
-   if (BIF_ARG_2 != am_backtrace)
-       BIF_ERROR(BIF_P, BADARG);
+    if (ERTS_PROC_IS_EXITING(c_p))
+        return am_badarg;
 
-   rp = erts_pid2proc_nropt(BIF_P, ERTS_PROC_LOCK_MAIN,
-			    BIF_ARG_1, ERTS_PROC_LOCKS_ALL);
-   if(!rp) {
-       BIF_ERROR(BIF_P, BADARG);
-   }
-   if (rp == ERTS_PROC_LOCK_BUSY)
-       ERTS_BIF_YIELD2(bif_export[BIF_process_display_2], BIF_P,
-		       BIF_ARG_1, BIF_ARG_2);
-   if (rp != BIF_P && ERTS_PROC_PENDING_EXIT(rp)) {
-       Eterm args[2] = {BIF_ARG_1, BIF_ARG_2};
-       erts_smp_proc_unlock(rp, ERTS_PROC_LOCKS_ALL);
-       ERTS_BIF_AWAIT_X_APPLY_TRAP(BIF_P,
-				   BIF_ARG_1,
-				   am_erlang,
-				   am_process_display,
-				   args,
-				   2);
-   }
-   erts_stack_dump(ERTS_PRINT_STDERR, NULL, rp);
-#ifdef ERTS_SMP
-   erts_smp_proc_unlock(rp, (BIF_P == rp
-			     ? ERTS_PROC_LOCKS_ALL_MINOR
-			     : ERTS_PROC_LOCKS_ALL));
-#endif
-   BIF_RET(am_true);
+    erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_stack_dump(ERTS_PRINT_STDERR, NULL, c_p);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+
+    return am_true;
+}
+
+
+BIF_RETTYPE erts_internal_process_display_2(BIF_ALIST_2)
+{
+    Eterm res;
+
+    if (BIF_ARG_2 != am_backtrace)
+        BIF_RET(am_badarg);
+
+    if (BIF_P->common.id == BIF_ARG_1) {
+        res = process_display(BIF_P, NULL, NULL, NULL);
+        BIF_RET(res);
+    }
+
+    if (is_not_internal_pid(BIF_ARG_1))
+        BIF_RET(am_badarg);
+
+    res = erts_proc_sig_send_rpc_request(BIF_P, BIF_ARG_1,
+                                         !0,
+                                         process_display,
+                                         NULL);
+    if (is_non_value(res))
+        BIF_RET(am_badarg);
+
+    BIF_RET(res);
 }
 
 /* this is a general call which return some possibly useful information */
@@ -3608,7 +3851,7 @@ BIF_RETTYPE error_logger_warning_map_0(BIF_ALIST_0)
     BIF_RET(erts_error_logger_warnings);
 }
 
-static erts_smp_atomic_t available_internal_state;
+static erts_atomic_t available_internal_state;
 
 static int empty_magic_ref_destructor(Binary *bin)
 {
@@ -3621,7 +3864,7 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
      * NOTE: Only supposed to be used for testing, and debugging.
      */
 
-    if (!erts_smp_atomic_read_nob(&available_internal_state)) {
+    if (!erts_atomic_read_nob(&available_internal_state)) {
 	BIF_ERROR(BIF_P, EXC_UNDEF);
     }
 
@@ -3664,9 +3907,9 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 	    int no_errors;
 	    ErtsCheckIoDebugInfo ciodi = {0};
 #ifdef HAVE_ERTS_CHECK_IO_DEBUG
-	    erts_smp_proc_unlock(BIF_P,ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(BIF_P,ERTS_PROC_LOCK_MAIN);
 	    no_errors = erts_check_io_debug(&ciodi);
-	    erts_smp_proc_lock(BIF_P,ERTS_PROC_LOCK_MAIN);
+	    erts_proc_lock(BIF_P,ERTS_PROC_LOCK_MAIN);
 #else
 	    no_errors = 0;
 #endif
@@ -3681,8 +3924,8 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 						   (Uint) ciodi.no_used_fds),
 				     erts_bld_uint(hpp, szp,
 						   (Uint) ciodi.no_driver_select_structs),
-				     erts_bld_uint(hpp, szp,
-						   (Uint) ciodi.no_driver_event_structs));
+                                     erts_bld_uint(hpp, szp,
+                                                   (Uint) ciodi.no_enif_select_structs));
 		if (hpp)
 		    break;
 		hp = HAlloc(BIF_P, sz);
@@ -3697,7 +3940,7 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 	    Eterm res = NIL;
 	    Uint *hp = HAlloc(BIF_P, 2*ERTS_PI_ARGS);
 	    for (i = ERTS_PI_ARGS-1; i >= 0; i--) {
-		res = CONS(hp, pi_args[i], res);
+		res = CONS(hp, pi_args[i].name, res);
 		hp += 2;
 	    }
 	    BIF_RET(res);
@@ -3716,9 +3959,9 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 	}
 	else if (ERTS_IS_ATOM_STR("nbalance", BIF_ARG_1)) {
 	    Uint n;
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	    n = erts_debug_nbalance();
-	    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	    BIF_RET(erts_make_integer(n, BIF_P));
 	}
 	else if (ERTS_IS_ATOM_STR("available_internal_state", BIF_ARG_1)) {
@@ -3733,11 +3976,11 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 	}
 	else if (ERTS_IS_ATOM_STR("memory", BIF_ARG_1)) {
 	    Eterm res;
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	    erts_smp_thr_progress_block();
-	    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_block();
+	    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 	    res = erts_memory(NULL, NULL, BIF_P, THE_NON_VALUE);
-	    erts_smp_thr_progress_unblock();
+	    erts_thr_progress_unblock();
 	    BIF_RET(res);
 	}
         else if (ERTS_IS_ATOM_STR("mmap", BIF_ARG_1)) {
@@ -3780,7 +4023,22 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 		Eterm *hp = HAlloc(BIF_P, hsz);
 		BIF_RET(uword_to_big(size, hp));
 	    }
+        } else if (ERTS_IS_ATOM_STR("scheduler_dump", BIF_ARG_1)) {
+#if defined(ERTS_HAVE_TRY_CATCH) && defined(ERTS_SYS_SUSPEND_SIGNAL)
+            BIF_RET(am_true);
+#else
+            BIF_RET(am_false);
+#endif
+        }
+        else if (ERTS_IS_ATOM_STR("lc_graph", BIF_ARG_1)) {
+#ifdef ERTS_ENABLE_LOCK_CHECK
+            Eterm res = erts_lc_dump_graph();
+            BIF_RET(res);
+#else
+            BIF_RET(am_notsup);
+#endif
         }
+
     }
     else if (is_tuple(BIF_ARG_1)) {
 	Eterm* tp = tuple_val(BIF_ARG_1);
@@ -3793,22 +4051,59 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 		    BIF_RET(erts_process_status(NULL, tp[2]));
 		}
 	    }
+            else if (ERTS_IS_ATOM_STR("connection_id", tp[1])) {
+                DistEntry *dep;
+                Eterm *hp, res;
+                Uint con_id, hsz = 0;
+                if (!is_atom(tp[2]))
+                    BIF_ERROR(BIF_P, BADARG);
+                dep = erts_sysname_to_connected_dist_entry(tp[2]);
+                if (!dep)
+                    BIF_ERROR(BIF_P, BADARG);
+                erts_de_rlock(dep);
+                con_id = (Uint) dep->connection_id;
+                erts_de_runlock(dep);
+                (void) erts_bld_uint(NULL, &hsz, con_id);
+                hp = hsz ? HAlloc(BIF_P, hsz) : NULL;
+                res = erts_bld_uint(&hp, NULL, con_id);
+                BIF_RET(res);
+            }
 	    else if (ERTS_IS_ATOM_STR("link_list", tp[1])) {
 		/* Used by erl_link_SUITE (emulator) */
 		if(is_internal_pid(tp[2])) {
+                    erts_aint32_t state;
 		    Eterm res;
 		    Process *p;
+                    int sigs_done, local_only;
 
 		    p = erts_pid2proc(BIF_P,
 				      ERTS_PROC_LOCK_MAIN,
 				      tp[2],
-				      ERTS_PROC_LOCK_LINK);
+				      ERTS_PROC_LOCK_MAIN);
 		    if (!p) {
-			ERTS_SMP_ASSERT_IS_NOT_EXITING(BIF_P);
+			ERTS_ASSERT_IS_NOT_EXITING(BIF_P);
 			BIF_RET(am_undefined);
 		    }
-		    res = make_link_list(BIF_P, ERTS_P_LINKS(p), NIL);
-		    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
+                    
+                    local_only = 0;
+                    do {
+                        int reds = CONTEXT_REDS;
+                        sigs_done = erts_proc_sig_handle_incoming(p,
+                                                                  &state,
+                                                                  &reds,
+                                                                  CONTEXT_REDS,
+                                                                  local_only);
+                        local_only = !0;
+                    } while (!sigs_done && !(state & ERTS_PSFLG_EXITING));
+
+                    if (!(state & ERTS_PSFLG_EXITING))
+                        res = make_link_list(BIF_P, 1, ERTS_P_LINKS(p), NIL);
+                    else if (BIF_P == p)
+                        ERTS_BIF_EXITED(BIF_P);
+                    else
+                        res = am_undefined;
+                    if (BIF_P != p)
+                        erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
 		    BIF_RET(res);
 		}
 		else if(is_internal_port(tp[2])) {
@@ -3819,20 +4114,20 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 						 ERTS_PORT_SFLGS_INVALID_LOOKUP);
 		    if(!p)
 			BIF_RET(am_undefined);
-		    res = make_link_list(BIF_P, ERTS_P_LINKS(p), NIL);
+		    res = make_link_list(BIF_P, 1, ERTS_P_LINKS(p), NIL);
 		    erts_port_release(p);
 		    BIF_RET(res);
 		}
 		else if(is_node_name_atom(tp[2])) {
 		    DistEntry *dep = erts_find_dist_entry(tp[2]);
 		    if(dep) {
-			Eterm subres;
-			erts_smp_de_links_lock(dep);
-			subres = make_link_list(BIF_P, dep->nlinks, NIL);
-			subres = make_link_list(BIF_P, dep->node_links, subres);
-			erts_smp_de_links_unlock(dep);
-			erts_deref_dist_entry(dep);
-			BIF_RET(subres);
+			Eterm res = NIL;
+                        if (dep->mld) {
+                            erts_mtx_lock(&dep->mld->mtx);
+                            res = make_link_list(BIF_P, 0, dep->mld->links, NIL);
+                            erts_mtx_unlock(&dep->mld->mtx);
+                        }
+			BIF_RET(res);
 		    } else {
 			BIF_RET(am_undefined);
 		    }
@@ -3841,28 +4136,54 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 	    else if (ERTS_IS_ATOM_STR("monitor_list", tp[1])) {
 		/* Used by erl_link_SUITE (emulator) */
 		if(is_internal_pid(tp[2])) {
+                    erts_aint32_t state;
 		    Process *p;
 		    Eterm res;
+                    int sigs_done, local_only;
 
 		    p = erts_pid2proc(BIF_P,
 				      ERTS_PROC_LOCK_MAIN,
 				      tp[2],
-				      ERTS_PROC_LOCK_LINK);
+				      ERTS_PROC_LOCK_MAIN);
 		    if (!p) {
-			ERTS_SMP_ASSERT_IS_NOT_EXITING(BIF_P);
+			ERTS_ASSERT_IS_NOT_EXITING(BIF_P);
 			BIF_RET(am_undefined);
 		    }
-		    res = make_monitor_list(BIF_P, ERTS_P_MONITORS(p));
-		    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_LINK);
+                    
+                    local_only = 0;
+                    do {
+                        int reds = CONTEXT_REDS;
+                        sigs_done = erts_proc_sig_handle_incoming(p,
+                                                                  &state,
+                                                                  &reds,
+                                                                  CONTEXT_REDS,
+                                                                  local_only);
+                        local_only = !0;
+                    } while (!sigs_done && !(state & ERTS_PSFLG_EXITING));
+
+                    if (!(state & ERTS_PSFLG_EXITING)) {
+                        res = make_monitor_list(BIF_P, 1, ERTS_P_MONITORS(p), NIL);
+                        res = make_monitor_list(BIF_P, 0, ERTS_P_LT_MONITORS(p), res);
+                    }
+                    else {
+                        if (BIF_P == p)
+                            ERTS_BIF_EXITED(BIF_P);
+                        else
+                            res = am_undefined;
+                    }
+                    if (BIF_P != p)
+                        erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
 		    BIF_RET(res);
 		} else if(is_node_name_atom(tp[2])) {
 		    DistEntry *dep = erts_find_dist_entry(tp[2]);
 		    if(dep) {
-			Eterm ml;
-			erts_smp_de_links_lock(dep);
-			ml = make_monitor_list(BIF_P, dep->monitors);
-			erts_smp_de_links_unlock(dep);
-			erts_deref_dist_entry(dep);
+			Eterm ml = NIL;
+                        if (dep->mld) {
+                            erts_mtx_lock(&dep->mld->mtx);
+                            ml = make_monitor_list(BIF_P, 1, dep->mld->orig_name_monitors, NIL);
+                            ml = make_monitor_list(BIF_P, 0, dep->mld->monitors, ml);
+                            erts_mtx_unlock(&dep->mld->mtx);
+                        }
 			BIF_RET(ml);
 		    } else {
 			BIF_RET(am_undefined);
@@ -3877,22 +4198,9 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 		else {
 		    Uint cno = dist_entry_channel_no(dep);
 		    res = make_small(cno);
-		    erts_deref_dist_entry(dep);
 		}
 		BIF_RET(res);
 	    }
-	    else if (ERTS_IS_ATOM_STR("have_pending_exit", tp[1])) {
-		Process *rp = erts_pid2proc(BIF_P, ERTS_PROC_LOCK_MAIN,
-					    tp[2], ERTS_PROC_LOCK_STATUS);
-		if (!rp) {
-		    BIF_RET(am_undefined);
-		}
-		else {
-		    Eterm res = ERTS_PROC_PENDING_EXIT(rp) ? am_true : am_false;
-		    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-		    BIF_RET(res);
-		}
-	    }
 	    else if (ERTS_IS_ATOM_STR("binary_info", tp[1])) {
 		Eterm bin = tp[2];
 		if (is_binary(bin)) {
@@ -3933,21 +4241,20 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 		    BIF_RET(res);
 		}
 	    }
-	    else if (ERTS_IS_ATOM_STR("term_to_binary_no_funs", tp[1])) {
-		Uint dflags = (DFLAG_EXTENDED_REFERENCES |
-			       DFLAG_EXTENDED_PIDS_PORTS |
-			       DFLAG_BIT_BINARIES);
+	    else if (ERTS_IS_ATOM_STR("term_to_binary_tuple_fallbacks", tp[1])) {
+		Uint dflags = (TERM_TO_BINARY_DFLAGS
+                               & ~DFLAG_EXPORT_PTR_TAG
+                               & ~DFLAG_BIT_BINARIES);
 		BIF_RET(erts_term_to_binary(BIF_P, tp[2], 0, dflags));
 	    }
-	    else if (ERTS_IS_ATOM_STR("dist_port", tp[1])) {
+	    else if (ERTS_IS_ATOM_STR("dist_ctrl", tp[1])) {
 		Eterm res = am_undefined;
 		DistEntry *dep = erts_sysname_to_connected_dist_entry(tp[2]);
 		if (dep) {
-		    erts_smp_de_rlock(dep);
-		    if (is_internal_port(dep->cid))
+		    erts_de_rlock(dep);
+		    if (is_internal_port(dep->cid) || is_internal_pid(dep->cid))
 			res = dep->cid;
-		    erts_smp_de_runlock(dep);
-		    erts_deref_dist_entry(dep);
+		    erts_de_runlock(dep);
 		}
 		BIF_RET(res);
 	    }
@@ -3987,7 +4294,7 @@ BIF_RETTYPE erts_debug_get_internal_state_1(BIF_ALIST_1)
 		while (ix >= atom_table_size()) {
 		    char tmp[20];
 		    erts_snprintf(tmp, sizeof(tmp), "am%x", atom_table_size());
-		    erts_atom_put((byte *) tmp, strlen(tmp), ERTS_ATOM_ENC_LATIN1, 1);
+		    erts_atom_put((byte *) tmp, sys_strlen(tmp), ERTS_ATOM_ENC_LATIN1, 1);
 		}
 		return make_atom(ix);
 	    }
@@ -4091,7 +4398,7 @@ BIF_RETTYPE erts_internal_system_check_1(BIF_ALIST_1)
     BIF_ERROR(BIF_P, BADARG);
 }
 
-static erts_smp_atomic_t hipe_test_reschedule_flag;
+static erts_atomic_t hipe_test_reschedule_flag;
 
 #if defined(VALGRIND) && defined(__GNUC__)
 /* Force noinline for valgrind suppression */
@@ -4115,7 +4422,7 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
     if (ERTS_IS_ATOM_STR("available_internal_state", BIF_ARG_1)
 	&& (BIF_ARG_2 == am_true || BIF_ARG_2 == am_false)) {
 	erts_aint_t on = (erts_aint_t) (BIF_ARG_2 == am_true);
-	erts_aint_t prev_on = erts_smp_atomic_xchg_nob(&available_internal_state, on);
+	erts_aint_t prev_on = erts_atomic_xchg_nob(&available_internal_state, on);
 	if (on) {
 	    erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
 	    erts_dsprintf(dsbufp, "Process %T ", BIF_P->common.id);
@@ -4131,7 +4438,7 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	BIF_RET(prev_on ? am_true : am_false);
     }
 
-    if (!erts_smp_atomic_read_nob(&available_internal_state)) {
+    if (!erts_atomic_read_nob(&available_internal_state)) {
 	BIF_ERROR(BIF_P, EXC_UNDEF);
     }
 
@@ -4155,13 +4462,13 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	    Sint ms;
 	    if (term_to_Sint(BIF_ARG_2, &ms) != 0) {
 		if (ms > 0) {
-		    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		    if (block)
-			erts_smp_thr_progress_block();
+			erts_thr_progress_block();
 		    while (erts_milli_sleep((long) ms) != 0);
 		    if (block)
-			erts_smp_thr_progress_unblock();
-		    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+			erts_thr_progress_unblock();
+		    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		}
 		BIF_RET(am_true);
 	    }
@@ -4170,9 +4477,9 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	    Sint ms;
 	    if (term_to_Sint(BIF_ARG_2, &ms) != 0) {
 		if (ms > 0) {
-		    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		    while (erts_milli_sleep((long) ms) != 0);
-		    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		}
 		BIF_RET(am_true);
 	    }
@@ -4220,61 +4527,6 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	    erts_set_gc_state(BIF_P, enable);
 	    BIF_RET(res);
 	}
-	else if (ERTS_IS_ATOM_STR("send_fake_exit_signal", BIF_ARG_1)) {
-	    /* Used by signal_SUITE (emulator) */
-
-	    /* Testcases depend on the exit being received via
-	       a pending exit when the receiver is the same as
-	       the caller.  */
-	    if (is_tuple(BIF_ARG_2)) {
-		Eterm* tp = tuple_val(BIF_ARG_2);
-		if (arityval(tp[0]) == 3
-		    && (is_pid(tp[1]) || is_port(tp[1]))
-		    && is_internal_pid(tp[2])) {
-		    int xres;
-		    ErtsProcLocks rp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
-		    Process *rp = erts_pid2proc(BIF_P, ERTS_PROC_LOCK_MAIN,
-						tp[2], rp_locks);
-		    if (!rp) {
-			DECL_AM(dead);
-			BIF_RET(AM_dead);
-		    }
-
-#ifdef ERTS_SMP
-		    if (BIF_P == rp)
-			rp_locks |= ERTS_PROC_LOCK_MAIN;
-#endif
-		    xres = erts_send_exit_signal(NULL, /* NULL in order to
-							  force a pending exit
-							  when we send to our
-							  selves. */
-						 tp[1],
-						 rp,
-						 &rp_locks,
-						 tp[3],
-						 NIL,
-						 NULL,
-						 0);
-#ifdef ERTS_SMP
-		    if (BIF_P == rp)
-			rp_locks &= ~ERTS_PROC_LOCK_MAIN;
-#endif
-		    erts_smp_proc_unlock(rp, rp_locks);
-		    if (xres > 1) {
-			DECL_AM(message);
-			BIF_RET(AM_message);
-		    }
-		    else if (xres == 0) {
-			DECL_AM(unaffected);
-			BIF_RET(AM_unaffected);
-		    }
-		    else {
-			DECL_AM(exit);
-			BIF_RET(AM_exit);
-		    }
-		}
-	    }
-	}
         else if (ERTS_IS_ATOM_STR("colliding_names", BIF_ARG_1)) {
 	    /* Used by ets_SUITE (stdlib) */
 	    if (is_tuple(BIF_ARG_2)) {
@@ -4321,14 +4573,14 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	}
 	else if (ERTS_IS_ATOM_STR("hipe_test_reschedule_suspend", BIF_ARG_1)) {
 	    /* Used by hipe test suites */
-	    erts_aint_t flag = erts_smp_atomic_read_nob(&hipe_test_reschedule_flag);
+	    erts_aint_t flag = erts_atomic_read_nob(&hipe_test_reschedule_flag);
 	    if (!flag && BIF_ARG_2 != am_false) {
-		erts_smp_atomic_set_nob(&hipe_test_reschedule_flag, 1);
+		erts_atomic_set_nob(&hipe_test_reschedule_flag, 1);
 		erts_suspend(BIF_P, ERTS_PROC_LOCK_MAIN, NULL);
 		ERTS_BIF_YIELD2(bif_export[BIF_erts_debug_set_internal_state_2],
 				BIF_P, BIF_ARG_1, BIF_ARG_2);
 	    }
-	    erts_smp_atomic_set_nob(&hipe_test_reschedule_flag, !flag);
+	    erts_atomic_set_nob(&hipe_test_reschedule_flag, !flag);
 	    BIF_RET(NIL);
 	}
 	else if (ERTS_IS_ATOM_STR("hipe_test_reschedule_resume", BIF_ARG_1)) {
@@ -4339,7 +4591,7 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	    if (rp) {
 		erts_resume(rp, ERTS_PROC_LOCK_STATUS);
 		res = am_true;
-		erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
+		erts_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
 	    }
 	    BIF_RET(res);
 	}
@@ -4356,39 +4608,13 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 		BIF_RET(am_false);
 	    else {
 		Uint32 con_id;
-		erts_smp_de_rlock(dep);
+		erts_de_rlock(dep);
 		con_id = dep->connection_id;
-		erts_smp_de_runlock(dep);
+		erts_de_runlock(dep);
 		erts_kill_dist_connection(dep, con_id);
-		erts_deref_dist_entry(dep);
 		BIF_RET(am_true);
 	    }
 	}
-	else if (ERTS_IS_ATOM_STR("not_running_optimization", BIF_ARG_1)) {
-#ifdef ERTS_SMP
-	    int old_use_opt, use_opt;
-	    switch (BIF_ARG_2) {
-	    case am_true:
-		use_opt = 1;
-		break;
-	    case am_false:
-		use_opt = 0;
-		break;
-	    default:
-		BIF_ERROR(BIF_P, BADARG);
-	    }
-
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	    erts_smp_thr_progress_block();
-	    old_use_opt = !erts_disable_proc_not_running_opt;
-	    erts_disable_proc_not_running_opt = !use_opt;
-	    erts_smp_thr_progress_unblock();
-	    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
-	    BIF_RET(old_use_opt ? am_true : am_false);
-#else
-	    BIF_ERROR(BIF_P,  EXC_NOTSUP);
-#endif
-	}
 	else if (ERTS_IS_ATOM_STR("wait", BIF_ARG_1)) {
 	    if (ERTS_IS_ATOM_STR("deallocations", BIF_ARG_2)) {
 		int flag = ERTS_DEBUG_WAIT_COMPLETED_DEALLOCATIONS;
@@ -4415,9 +4641,9 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
 	    Sint64 msecs;
 	    if (term_to_Sint64(BIF_ARG_2, &msecs)) {
 		/* Negative value restore original value... */
-		erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		erts_debug_test_node_tab_delayed_delete(msecs);
-		erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+		erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 		BIF_RET(am_ok);
 	    }
 	}
@@ -4459,12 +4685,68 @@ BIF_RETTYPE erts_debug_set_internal_state_2(BIF_ALIST_2)
                                             refbin));
             }
         }
-
+        else if (ERTS_IS_ATOM_STR("ets_force_trap", BIF_ARG_1)) {
+#ifdef ETS_DBG_FORCE_TRAP
+            erts_ets_dbg_force_trap = (BIF_ARG_2 == am_true) ? 1 : 0;
+            BIF_RET(am_ok);
+#else
+            BIF_RET(am_notsup);
+#endif
+        }
     }
 
     BIF_ERROR(BIF_P, BADARG);
 }
 
+static BIF_RETTYPE
+gather_histograms_helper(Process * c_p, Eterm arg_tuple,
+                         int gather(Process *, int, int, int, UWord, Eterm))
+{
+    SWord hist_start, hist_width, sched_id;
+    int msg_count, alloc_num;
+    Eterm *args;
+
+    /* This is an internal BIF, so the error checking is mostly left to erlang
+     * code. */
+
+    ASSERT(is_tuple_arity(arg_tuple, 5));
+    args = tuple_val(arg_tuple);
+
+    for (alloc_num = ERTS_ALC_A_MIN; alloc_num <= ERTS_ALC_A_MAX; alloc_num++) {
+        if(erts_is_atom_str(ERTS_ALC_A2AD(alloc_num), args[1], 0)) {
+            break;
+        }
+    }
+
+    if (alloc_num > ERTS_ALC_A_MAX) {
+        BIF_ERROR(c_p, BADARG);
+    }
+
+    sched_id = signed_val(args[2]);
+    hist_width = signed_val(args[3]);
+    hist_start = signed_val(args[4]);
+
+    if (sched_id < 0 || sched_id > erts_no_schedulers) {
+        BIF_ERROR(c_p, BADARG);
+    }
+
+    msg_count = gather(c_p, alloc_num, sched_id, hist_width, hist_start, args[5]);
+
+    BIF_RET(make_small(msg_count));
+}
+
+BIF_RETTYPE erts_internal_gather_alloc_histograms_1(BIF_ALIST_1)
+{
+    return gather_histograms_helper(BIF_P, BIF_ARG_1,
+                                    erts_alcu_gather_alloc_histograms);
+}
+
+BIF_RETTYPE erts_internal_gather_carrier_info_1(BIF_ALIST_1)
+{
+    return gather_histograms_helper(BIF_P, BIF_ARG_1,
+                                    erts_alcu_gather_carrier_info);
+}
+
 #ifdef ERTS_ENABLE_LOCK_COUNT
 
 typedef struct {
@@ -4562,7 +4844,7 @@ static Eterm lcnt_build_lock_stats_term(Eterm **hpp, Uint *szp, erts_lcnt_lock_s
 
     file = stats->file ? stats->file : "undefined";
 
-    af    = erts_atom_put((byte *)file, strlen(file), ERTS_ATOM_ENC_LATIN1, 1);
+    af    = erts_atom_put((byte *)file, sys_strlen(file), ERTS_ATOM_ENC_LATIN1, 1);
     uil   = erts_bld_uint( hpp, szp, stats->line);
     tloc  = erts_bld_tuple(hpp, szp, 2, af, uil);
 
@@ -4608,7 +4890,7 @@ static Eterm lcnt_pretty_print_lock_id(erts_lcnt_lock_info_t *info) {
     } else if(info->flags & ERTS_LOCK_FLAGS_CATEGORY_ALLOCATOR) {
         if(is_small(id) && !sys_strcmp(info->name, "alcu_allocator")) {
             const char *name = (const char*)ERTS_ALC_A2AD(signed_val(id));
-            id = erts_atom_put((byte*)name, strlen(name), ERTS_ATOM_ENC_LATIN1, 1);
+            id = erts_atom_put((byte*)name, sys_strlen(name), ERTS_ATOM_ENC_LATIN1, 1);
         }
     }
 
@@ -4628,8 +4910,8 @@ static Eterm lcnt_build_lock_term(Eterm **hpp, Uint *szp, lcnt_sample_t *sample,
     
     lock_desc = erts_lock_flags_get_type_name(info->flags);
 
-    type  = erts_atom_put((byte*)lock_desc, strlen(lock_desc), ERTS_ATOM_ENC_LATIN1, 1);
-    name  = erts_atom_put((byte*)info->name, strlen(info->name), ERTS_ATOM_ENC_LATIN1, 1);
+    type  = erts_atom_put((byte*)lock_desc, sys_strlen(lock_desc), ERTS_ATOM_ENC_LATIN1, 1);
+    name  = erts_atom_put((byte*)info->name, sys_strlen(info->name), ERTS_ATOM_ENC_LATIN1, 1);
 
     /* Only attempt to resolve ids when actually emitting the term. This ought
      * to be safe since all immediates are the same size. */
@@ -4665,11 +4947,11 @@ static Eterm lcnt_build_result_term(Eterm **hpp, Uint *szp, erts_lcnt_time_t *du
     dtns = bld_unstable_uint64(hpp, szp, duration->ns);
     tdt  = erts_bld_tuple(hpp, szp, 2, dts, dtns);
 
-    adur = erts_atom_put((byte *)str_duration, strlen(str_duration), ERTS_ATOM_ENC_LATIN1, 1);
+    adur = erts_atom_put((byte *)str_duration, sys_strlen(str_duration), ERTS_ATOM_ENC_LATIN1, 1);
     tdur = erts_bld_tuple(hpp, szp, 2, adur, tdt);
    
     /* lock tuple */
-    aloc = erts_atom_put((byte *)str_locks, strlen(str_locks), ERTS_ATOM_ENC_LATIN1, 1);
+    aloc = erts_atom_put((byte *)str_locks, sys_strlen(str_locks), ERTS_ATOM_ENC_LATIN1, 1);
 
     for(i = 0; i < current_locks->size; i++) {
         lloc = lcnt_build_lock_term(hpp, szp, &current_locks->elements[i], lloc);
@@ -4723,7 +5005,7 @@ static Eterm lcnt_build_category_list(Eterm **hpp, Uint *szp, erts_lock_flags_t
     for(i = 0; lcnt_category_map[i].name != NULL; i++) {
         if(mask & lcnt_category_map[i].flag) {
             Eterm category = erts_atom_put((byte*)lcnt_category_map[i].name,
-                                           strlen(lcnt_category_map[i].name),
+                                           sys_strlen(lcnt_category_map[i].name),
                                            ERTS_ATOM_ENC_UTF8, 0);
 
             res = erts_bld_cons(hpp, szp, category, res);
@@ -4853,14 +5135,14 @@ BIF_RETTYPE erts_debug_lcnt_control_2(BIF_ALIST_2)
 
 static void os_info_init(void)
 {
-    Eterm type = erts_atom_put((byte *) os_type, strlen(os_type), ERTS_ATOM_ENC_LATIN1, 1);
+    Eterm type = erts_atom_put((byte *) os_type, sys_strlen(os_type), ERTS_ATOM_ENC_LATIN1, 1);
     Eterm flav;
     int major, minor, build;
     char* buf = erts_alloc(ERTS_ALC_T_TMP, 1024); /* More than enough */
     Eterm* hp;
 
     os_flavor(buf, 1024);
-    flav = erts_atom_put((byte *) buf, strlen(buf), ERTS_ATOM_ENC_LATIN1, 1);
+    flav = erts_atom_put((byte *) buf, sys_strlen(buf), ERTS_ATOM_ENC_LATIN1, 1);
     erts_free(ERTS_ALC_T_TMP, (void *) buf);
     hp = erts_alloc(ERTS_ALC_T_LITERAL, (3+4)*sizeof(Eterm));
     os_type_tuple = TUPLE2(hp, type, flav);
@@ -4878,13 +5160,13 @@ static void os_info_init(void)
 void
 erts_bif_info_init(void)
 {
-    erts_smp_atomic_init_nob(&available_internal_state, 0);
-    erts_smp_atomic_init_nob(&hipe_test_reschedule_flag, 0);
+    erts_atomic_init_nob(&available_internal_state, 0);
+    erts_atomic_init_nob(&hipe_test_reschedule_flag, 0);
 
     alloc_info_trap = erts_export_put(am_erlang, am_alloc_info, 1);
     alloc_sizes_trap = erts_export_put(am_erlang, am_alloc_sizes, 1);
     gather_sched_wall_time_res_trap
-	= erts_export_put(am_erlang, am_gather_sched_wall_time_result, 1);
+	= erts_export_put(am_erts_internal, am_gather_sched_wall_time_result, 1);
     gather_gc_info_res_trap
 	= erts_export_put(am_erlang, am_gather_gc_info_result, 1);
     gather_io_bytes_trap
@@ -4893,6 +5175,10 @@ erts_bif_info_init(void)
 	= erts_export_put(am_erts_internal, am_gather_microstate_accounting_result, 2);
     gather_system_check_res_trap
 	= erts_export_put(am_erts_internal, am_gather_system_check_result, 1);
+
+    is_process_alive_trap = erts_export_put(am_erts_internal, am_is_process_alive, 1);
+
+
     process_info_init();
     os_info_init();
 }
diff --git a/erts/emulator/beam/erl_bif_os.c b/erts/emulator/beam/erl_bif_os.c
index 5cd172c26f..ce2b27409b 100644
--- a/erts/emulator/beam/erl_bif_os.c
+++ b/erts/emulator/beam/erl_bif_os.c
@@ -36,6 +36,7 @@
 #include "big.h"
 #include "dist.h"
 #include "erl_version.h"
+#include "erl_osenv.h"
 
 /*
  * Return the pid for the Erlang process in the host OS.
@@ -65,142 +66,77 @@ BIF_RETTYPE os_getpid_0(BIF_ALIST_0)
      BIF_RET(buf_to_intlist(&hp, pid_string, n, NIL));
 }
 
-BIF_RETTYPE os_getenv_0(BIF_ALIST_0)
+static void os_getenv_foreach(Process *process, Eterm *result, Eterm key, Eterm value)
 {
-    GETENV_STATE state;
-    char *cp;
-    Eterm* hp;
-    Eterm ret;
-    Eterm str;
+    Eterm kvp_term, *hp;
 
-    init_getenv_state(&state);
+    hp = HAlloc(process, 5);
+    kvp_term = TUPLE2(hp, key, value);
+    hp += 3;
 
-    ret = NIL;
-    while ((cp = getenv_string(&state)) != NULL) {
-	str = erts_convert_native_to_filename(BIF_P,(byte *)cp);
-	hp = HAlloc(BIF_P, 2);
-	ret = CONS(hp, str, ret);
-    }
+    (*result) = CONS(hp, kvp_term, (*result));
+}
+
+BIF_RETTYPE os_list_env_vars_0(BIF_ALIST_0)
+{
+    const erts_osenv_t *global_env;
+    Eterm result = NIL;
 
-    fini_getenv_state(&state);
+    global_env = erts_sys_rlock_global_osenv();
+    erts_osenv_foreach_term(global_env, BIF_P, &result, (void*)&os_getenv_foreach);
+    erts_sys_runlock_global_osenv();
 
-    return ret;
+    return result;
 }
 
-#define STATIC_BUF_SIZE 1024
-BIF_RETTYPE os_getenv_1(BIF_ALIST_1)
+BIF_RETTYPE os_get_env_var_1(BIF_ALIST_1)
 {
-    Process* p = BIF_P;
-    Eterm str;
-    Sint len;
-    int res;
-    char *key_str, *val;
-    char buf[STATIC_BUF_SIZE];
-    size_t val_size = sizeof(buf);
-
-    key_str = erts_convert_filename_to_native(BIF_ARG_1,buf,STATIC_BUF_SIZE,
-					      ERTS_ALC_T_TMP,1,0,&len);
-
-    if (!key_str) {
-	BIF_ERROR(p, BADARG);
-    }
+    const erts_osenv_t *global_env;
+    Eterm out_term;
+    int error;
 
-    if (key_str != &buf[0])
-	val = &buf[0];
-    else {
-	/* len includes zero byte */
-	val_size -= len;
-	val = &buf[len];
-    }
-    res = erts_sys_getenv(key_str, val, &val_size);
-
-    if (res < 0) {
-    no_var:
-	str = am_false;
-    } else {
-	if (res > 0) {
-	    val = erts_alloc(ERTS_ALC_T_TMP, val_size);
-	    while (1) {
-		res = erts_sys_getenv(key_str, val, &val_size);
-		if (res == 0)
-		    break;
-		else if (res < 0)
-		    goto no_var;
-		else
-		    val = erts_realloc(ERTS_ALC_T_TMP, val, val_size);
-	    }
-	}
-	str = erts_convert_native_to_filename(p,(byte *)val);
-    }
-    if (key_str != &buf[0])
-	erts_free(ERTS_ALC_T_TMP, key_str);
-    if (val < &buf[0] || &buf[sizeof(buf)-1] < val)
-	erts_free(ERTS_ALC_T_TMP, val);
-    BIF_RET(str);
+    global_env = erts_sys_rlock_global_osenv();
+    error = erts_osenv_get_term(global_env, BIF_P, BIF_ARG_1, &out_term);
+    erts_sys_runlock_global_osenv();
+
+    if (error == 0) {
+        return am_false;
+    } else if (error < 0) {
+        BIF_ERROR(BIF_P, BADARG);
+    } 
+
+    return out_term;
 }
 
-BIF_RETTYPE os_putenv_2(BIF_ALIST_2)
+BIF_RETTYPE os_set_env_var_2(BIF_ALIST_2)
 {
-    char def_buf_key[STATIC_BUF_SIZE];
-    char def_buf_value[STATIC_BUF_SIZE];
-    char *key_buf, *value_buf;
-
-    key_buf = erts_convert_filename_to_native(BIF_ARG_1,def_buf_key,
-					      STATIC_BUF_SIZE,
-					      ERTS_ALC_T_TMP,0,0,NULL);
-    if (!key_buf) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
-    value_buf = erts_convert_filename_to_native(BIF_ARG_2,def_buf_value,
-						STATIC_BUF_SIZE,
-						ERTS_ALC_T_TMP,1,0,
-						NULL);
-    if (!value_buf) {
-	if (key_buf != def_buf_key) {
-	    erts_free(ERTS_ALC_T_TMP, key_buf);
-	}
-	BIF_ERROR(BIF_P, BADARG);
-    }
-	    
-
-    if (erts_sys_putenv(key_buf, value_buf)) {
-	if (key_buf != def_buf_key) {
-	    erts_free(ERTS_ALC_T_TMP, key_buf);
-	}
-	if (value_buf != def_buf_value) {
-	    erts_free(ERTS_ALC_T_TMP, value_buf);
-	}
-	BIF_ERROR(BIF_P, BADARG);
-    }
-    if (key_buf != def_buf_key) {
-	erts_free(ERTS_ALC_T_TMP, key_buf);
-    }
-    if (value_buf != def_buf_value) {
-	erts_free(ERTS_ALC_T_TMP, value_buf);
+    erts_osenv_t *global_env;
+    int error;
+
+    global_env = erts_sys_rwlock_global_osenv();
+    error = erts_osenv_put_term(global_env, BIF_ARG_1, BIF_ARG_2);
+    erts_sys_rwunlock_global_osenv();
+
+    if (error < 0) {
+        BIF_ERROR(BIF_P, BADARG);
     }
+
     BIF_RET(am_true);
 }
 
-BIF_RETTYPE os_unsetenv_1(BIF_ALIST_1)
+BIF_RETTYPE os_unset_env_var_1(BIF_ALIST_1)
 {
-    char *key_buf;
-    char buf[STATIC_BUF_SIZE];
+    erts_osenv_t *global_env;
+    int error;
 
-    key_buf = erts_convert_filename_to_native(BIF_ARG_1,buf,STATIC_BUF_SIZE,
-					      ERTS_ALC_T_TMP,0,0,NULL);
-    if (!key_buf) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    global_env = erts_sys_rwlock_global_osenv();
+    error = erts_osenv_unset_term(global_env, BIF_ARG_1);
+    erts_sys_rwunlock_global_osenv();
 
-    if (erts_sys_unsetenv(key_buf)) {
-	if (key_buf != buf) {
-	    erts_free(ERTS_ALC_T_TMP, key_buf);
-	}
-	BIF_ERROR(BIF_P, BADARG);
-    }
-    if (key_buf != buf) {
-	erts_free(ERTS_ALC_T_TMP, key_buf);
+    if (error < 0) {
+        BIF_ERROR(BIF_P, BADARG);
     }
+
     BIF_RET(am_true);
 }
 
diff --git a/erts/emulator/beam/erl_bif_port.c b/erts/emulator/beam/erl_bif_port.c
index d05507118e..7fe4e02782 100644
--- a/erts/emulator/beam/erl_bif_port.c
+++ b/erts/emulator/beam/erl_bif_port.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2001-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2001-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -43,9 +43,10 @@
 #include "erl_bits.h"
 #include "erl_bif_unique.h"
 #include "dtrace-wrapper.h"
+#include "erl_proc_sig_queue.h"
 
 static Port *open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump);
-static byte* convert_environment(Process* p, Eterm env);
+static int merge_global_environment(erts_osenv_t *env, Eterm key_value_pairs);
 static char **convert_args(Eterm);
 static void free_args(char **);
 
@@ -53,10 +54,13 @@ char *erts_default_arg0 = "default";
 
 BIF_RETTYPE erts_internal_open_port_2(BIF_ALIST_2)
 {
+    BIF_RETTYPE ret;
     Port *port;
     Eterm res;
     char *str;
     int err_type, err_num;
+    ErtsLinkData *ldp;
+    ErtsLink *lnk;
 
     port = open_port(BIF_P, BIF_ARG_1, BIF_ARG_2, &err_type, &err_num);
     if (!port) {
@@ -71,13 +75,23 @@ BIF_RETTYPE erts_internal_open_port_2(BIF_ALIST_2)
                 BIF_ERROR(BIF_P, EXC_INTERNAL_ERROR);
 	} else if (err_type == -2) {
 	    str = erl_errno_id(err_num);
-            res = erts_atom_put((byte *) str, strlen(str), ERTS_ATOM_ENC_LATIN1, 1);
+            res = erts_atom_put((byte *) str, sys_strlen(str), ERTS_ATOM_ENC_LATIN1, 1);
 	} else {
 	    res = am_einval;
 	}
         BIF_RET(res);
     }
 
+    ldp = erts_link_create(ERTS_LNK_TYPE_PORT, BIF_P->common.id, port->common.id);
+    ASSERT(ldp->a.other.item == port->common.id);
+    ASSERT(ldp->b.other.item == BIF_P->common.id);
+    /*
+     * This link should not already be present, but can potentially
+     * due to id wrapping...
+     */
+    lnk = erts_link_tree_lookup_insert(&ERTS_P_LINKS(BIF_P), &ldp->a);
+    erts_link_tree_insert(&ERTS_P_LINKS(port), &ldp->b);
+
     if (port->drv_ptr->flags & ERL_DRV_FLAG_USE_INIT_ACK) {
 
         /* Copied from erl_port_task.c */
@@ -86,39 +100,30 @@ BIF_RETTYPE erts_internal_open_port_2(BIF_ALIST_2)
         erts_make_ref_in_array(port->async_open_port->ref);
         port->async_open_port->to = BIF_P->common.id;
 
-        erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCKS_MSG_RECEIVE | ERTS_PROC_LOCK_LINK);
-        if (ERTS_PROC_PENDING_EXIT(BIF_P)) {
-	    /* need to exit caller instead */
-	    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCKS_MSG_RECEIVE | ERTS_PROC_LOCK_LINK);
-	    KILL_CATCHES(BIF_P);
-	    BIF_P->freason = EXC_EXIT;
-            erts_port_release(port);
-            BIF_RET(am_badarg);
-	}
-
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(BIF_P);
-	BIF_P->msg.save = BIF_P->msg.last;
-
-	erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCKS_MSG_RECEIVE);
+        /*
+         * We unconditionaly *must* do a receive on a message
+         * containing the reference after this...
+         */
+        ERTS_RECV_MARK_SAVE(BIF_P);
+        ERTS_RECV_MARK_SET(BIF_P);
 
         res = erts_proc_store_ref(BIF_P, port->async_open_port->ref);
     } else {
         res = port->common.id;
-        erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK);
     }
 
-    erts_add_link(&ERTS_P_LINKS(port), LINK_PID, BIF_P->common.id);
-    erts_add_link(&ERTS_P_LINKS(BIF_P), LINK_PID, port->common.id);
-
     if (IS_TRACED_FL(BIF_P, F_TRACE_PROCS))
-        trace_proc(BIF_P, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK, BIF_P,
+        trace_proc(BIF_P, ERTS_PROC_LOCK_MAIN, BIF_P,
                    am_link, port->common.id);
 
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
+    ERTS_BIF_PREP_RET(ret, res);
 
     erts_port_release(port);
 
-    BIF_RET(res);
+    if (lnk)
+        erts_link_release(lnk);
+
+    return ret;
 }
 
 static ERTS_INLINE Port *
@@ -195,7 +200,6 @@ BIF_RETTYPE erts_internal_port_command_3(BIF_ALIST_3)
 #endif
 
     switch (erts_port_output(BIF_P, flags, prt, prt->common.id, BIF_ARG_2, &ref)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
     case ERTS_PORT_OP_DROPPED:
  	ERTS_BIF_PREP_RET(res, am_badarg);
@@ -254,7 +258,6 @@ BIF_RETTYPE erts_internal_port_call_3(BIF_ALIST_3)
     op = (unsigned int) uint_op;
 
     switch (erts_port_call(BIF_P, prt, op, BIF_ARG_3, &retval)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_DROPPED:
     case ERTS_PORT_OP_BADARG:
 	retval = am_badarg;
@@ -271,14 +274,9 @@ BIF_RETTYPE erts_internal_port_call_3(BIF_ALIST_3)
 	break;
     }
 
-    state = erts_smp_atomic32_read_acqb(&BIF_P->state);
-    if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT)) {
-#ifdef ERTS_SMP
-	if (state & ERTS_PSFLG_PENDING_EXIT)
-	    erts_handle_pending_exit(BIF_P, ERTS_PROC_LOCK_MAIN);
-#endif
+    state = erts_atomic32_read_acqb(&BIF_P->state);
+    if (state & ERTS_PSFLG_EXITING)
 	ERTS_BIF_EXITED(BIF_P);
-    }
 
     BIF_RET(retval);
 }
@@ -304,7 +302,6 @@ BIF_RETTYPE erts_internal_port_control_3(BIF_ALIST_3)
     op = (unsigned int) uint_op;
 
     switch (erts_port_control(BIF_P, prt, op, BIF_ARG_3, &retval)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
     case ERTS_PORT_OP_DROPPED:
 	retval = am_badarg;
@@ -321,14 +318,9 @@ BIF_RETTYPE erts_internal_port_control_3(BIF_ALIST_3)
 	break;
     }
 
-    state = erts_smp_atomic32_read_acqb(&BIF_P->state);
-    if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT)) {
-#ifdef ERTS_SMP
-	if (state & ERTS_PSFLG_PENDING_EXIT)
-	    erts_handle_pending_exit(BIF_P, ERTS_PROC_LOCK_MAIN);
-#endif
+    state = erts_atomic32_read_acqb(&BIF_P->state);
+    if (state & ERTS_PSFLG_EXITING)
 	ERTS_BIF_EXITED(BIF_P);
-    }
 
     BIF_RET(retval);
 }
@@ -351,7 +343,6 @@ BIF_RETTYPE erts_internal_port_close_1(BIF_ALIST_1)
 	BIF_RET(am_badarg);
 
     switch (erts_port_exit(BIF_P, 0, prt, BIF_P->common.id, am_normal, &ref)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
     case ERTS_PORT_OP_DROPPED:
 	BIF_RET(am_badarg);
@@ -384,7 +375,6 @@ BIF_RETTYPE erts_internal_port_connect_2(BIF_ALIST_2)
 #endif
 
     switch (erts_port_connect(BIF_P, 0, prt, BIF_P->common.id, BIF_ARG_2, &ref)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
     case ERTS_PORT_OP_DROPPED:
 	BIF_RET(am_badarg);
@@ -422,7 +412,6 @@ BIF_RETTYPE erts_internal_port_info_1(BIF_ALIST_1)
     }
 
     switch (erts_port_info(BIF_P, prt, THE_NON_VALUE, &retval)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
 	BIF_RET(am_badarg);
     case ERTS_PORT_OP_DROPPED:
@@ -461,7 +450,6 @@ BIF_RETTYPE erts_internal_port_info_2(BIF_ALIST_2)
     }
 
     switch (erts_port_info(BIF_P, prt, BIF_ARG_2, &retval)) {
-    case ERTS_PORT_OP_CALLER_EXIT:
     case ERTS_PORT_OP_BADARG:
 	BIF_RET(am_badarg);
     case ERTS_PORT_OP_DROPPED:
@@ -511,39 +499,35 @@ cleanup_old_port_data(erts_aint_t data)
 	ASSERT(is_immed((Eterm) data));
     }
     else {
-#ifdef ERTS_SMP
 	ErtsPortDataHeap *pdhp = (ErtsPortDataHeap *) data;
 	size_t size;
-	ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
+	ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
 	size = sizeof(ErtsPortDataHeap) + (pdhp->hsize-1)*sizeof(Eterm);
 	erts_schedule_thr_prgr_later_cleanup_op(free_port_data_heap,
 						(void *) pdhp,
 						&pdhp->later_op,
 						size);
-#else
-	free_port_data_heap((void *) data);
-#endif
     }
 }
 
 void
 erts_init_port_data(Port *prt)
 {
-    erts_smp_atomic_init_nob(&prt->data, (erts_aint_t) am_undefined);
+    erts_atomic_init_nob(&prt->data, (erts_aint_t) am_undefined);
 }
 
 void
 erts_cleanup_port_data(Port *prt)
 {
     ASSERT(erts_atomic32_read_nob(&prt->state) & ERTS_PORT_SFLGS_INVALID_LOOKUP);
-    cleanup_old_port_data(erts_smp_atomic_xchg_nob(&prt->data,
+    cleanup_old_port_data(erts_atomic_xchg_nob(&prt->data,
 						   (erts_aint_t) NULL));
 }
 
 Uint
 erts_port_data_size(Port *prt)
 {
-    erts_aint_t data = erts_smp_atomic_read_ddrb(&prt->data);
+    erts_aint_t data = erts_atomic_read_ddrb(&prt->data);
 
     if ((data & 0x3) != 0) {
 	ASSERT(is_immed((Eterm) (UWord) data));
@@ -558,7 +542,7 @@ erts_port_data_size(Port *prt)
 ErlOffHeap *
 erts_port_data_offheap(Port *prt)
 {
-    erts_aint_t data = erts_smp_atomic_read_ddrb(&prt->data);
+    erts_aint_t data = erts_atomic_read_ddrb(&prt->data);
 
     if ((data & 0x3) != 0) {
 	ASSERT(is_immed((Eterm) (UWord) data));
@@ -603,11 +587,11 @@ BIF_RETTYPE port_set_data_2(BIF_ALIST_2)
 	ASSERT((data & 0x3) == 0);
     }
 
-    data = erts_smp_atomic_xchg_wb(&prt->data, data);
+    data = erts_atomic_xchg_wb(&prt->data, data);
 
     if (data == (erts_aint_t)NULL) {
 	/* Port terminated by racing thread */
-	data = erts_smp_atomic_xchg_wb(&prt->data, data);
+	data = erts_atomic_xchg_wb(&prt->data, data);
 	ASSERT(data != (erts_aint_t)NULL);
 	cleanup_old_port_data(data);
 	BIF_ERROR(BIF_P, BADARG);
@@ -630,7 +614,7 @@ BIF_RETTYPE port_get_data_1(BIF_ALIST_1)
     if (!prt)
         BIF_ERROR(BIF_P, BADARG);
 
-    data = erts_smp_atomic_read_ddrb(&prt->data);
+    data = erts_atomic_read_ddrb(&prt->data);
     if (data == (erts_aint_t)NULL)
         BIF_ERROR(BIF_P, BADARG);  /* Port terminated by racing thread */
 
@@ -652,7 +636,7 @@ Eterm erts_port_data_read(Port* prt)
     Eterm res;
     erts_aint_t data;
 
-    data = erts_smp_atomic_read_ddrb(&prt->data);
+    data = erts_atomic_read_ddrb(&prt->data);
     if (data == (erts_aint_t)NULL)
         return am_undefined;  /* Port terminated by racing thread */
 
@@ -680,6 +664,7 @@ Eterm erts_port_data_read(Port* prt)
 static Port *
 open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
 {
+    int merged_environment = 0;
     Sint i;
     Eterm option;
     Uint arity;
@@ -701,12 +686,13 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
     opts.read_write = 0;
     opts.hide_window = 0;
     opts.wd = NULL;
-    opts.envir = NULL;
     opts.exit_status = 0;
     opts.overlapped_io = 0; 
     opts.spawn_type = ERTS_SPAWN_ANY; 
     opts.argv = NULL;
     opts.parallelism = erts_port_parallelism;
+    erts_osenv_init(&opts.envir);
+
     linebuf = 0;
 
     *err_nump = 0;
@@ -747,11 +733,16 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
 			goto badarg;
 		    }
 		} else if (option == am_env) {
-		    byte* bytes;
-		    if ((bytes = convert_environment(p, *tp)) == NULL) {
-			goto badarg;
+		    if (merged_environment) {
+		        /* Ignore previous env option */
+		        erts_osenv_clear(&opts.envir);
+		    }
+
+		    merged_environment = 1;
+
+		    if (merge_global_environment(&opts.envir, *tp)) {
+		        goto badarg;
 		    }
-		    opts.envir = (char *) bytes;
 		} else if (option == am_args) {
 		    char **av;
 		    char **oav = opts.argv;
@@ -836,6 +827,12 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
     if((linebuf && opts.packet_bytes) || 
        (opts.redir_stderr && !opts.use_stdio)) {
 	goto badarg;
+}
+
+    /* If we lacked an env option, fill in the global environment without
+     * changes. */
+    if (!merged_environment) {
+        merge_global_environment(&opts.envir, NIL);
     }
 
     /*
@@ -946,7 +943,7 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
     }
     
 
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
 
     port = erts_open_driver(driver, p->common.id, name_buf, &opts, err_typep, err_nump);
 #ifdef USE_VM_PROBES
@@ -963,7 +960,7 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
     if (port && IS_TRACED_FL(port, F_TRACE_PORTS))
         trace_port(port, am_getting_linked, p->common.id);
 
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
 
     if (IS_TRACED_FL(p, F_TRACE_SCHED_PROCS)) {
         trace_sched(p, ERTS_PROC_LOCK_MAIN, am_in);
@@ -985,6 +982,7 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
 	erts_atomic32_read_bor_relb(&port->state, sflgs);
  
  do_return:
+    erts_osenv_clear(&opts.envir);
     if (name_buf)
 	erts_free(ERTS_ALC_T_TMP, (void *) name_buf);
     if (opts.argv) {
@@ -1004,6 +1002,45 @@ open_port(Process* p, Eterm name, Eterm settings, int *err_typep, int *err_nump)
     goto do_return;
 }
 
+/* Merges the the global environment and the given {Key, Value} list into env,
+ * unsetting all keys whose value is either 'false' or NIL. The behavior on
+ * NIL is undocumented and perhaps surprising, but the previous implementation
+ * worked in this manner. */
+static int merge_global_environment(erts_osenv_t *env, Eterm key_value_pairs) {
+    const erts_osenv_t *global_env = erts_sys_rlock_global_osenv();
+    erts_osenv_merge(env, global_env, 0);
+    erts_sys_runlock_global_osenv();
+
+    while (is_list(key_value_pairs)) {
+        Eterm *cell, *tuple;
+
+        cell = list_val(key_value_pairs);
+
+        if(!is_tuple_arity(CAR(cell), 2)) {
+            return -1;
+        }
+
+        tuple = tuple_val(CAR(cell));
+        key_value_pairs = CDR(cell);
+
+        if(is_nil(tuple[2]) || tuple[2] == am_false) {
+            if(erts_osenv_unset_term(env, tuple[1]) < 0) {
+                return -1;
+            }
+        } else {
+            if(erts_osenv_put_term(env, tuple[1], tuple[2]) < 0) {
+                return -1;
+            }
+        }
+    }
+
+    if(!is_nil(key_value_pairs)) {
+        return -1;
+    }
+
+    return 0;
+}
+
 /* Arguments can be given i unicode and as raw binaries, convert filename is used to convert */
 static char **convert_args(Eterm l)
 {
@@ -1050,75 +1087,6 @@ static void free_args(char **av)
     }
     erts_free(ERTS_ALC_T_TMP, av);
 }
-    
-
-static byte* convert_environment(Process* p, Eterm env)
-{
-    Eterm all;
-    Eterm* temp_heap;
-    Eterm* hp;
-    Uint heap_size;
-    Sint n;
-    Sint size;
-    byte* bytes;
-    int encoding = erts_get_native_filename_encoding();
-
-    if ((n = erts_list_length(env)) < 0) {
-	return NULL;
-    }
-    heap_size = 2*(5*n+1);
-    temp_heap = hp = (Eterm *) erts_alloc(ERTS_ALC_T_TMP, heap_size*sizeof(Eterm));
-    bytes = NULL;		/* Indicating error */
-
-    /*
-     * All errors below are handled by jumping to 'done', to ensure that the memory
-     * gets deallocated. Do NOT return directly from this function.
-     */
-
-    all = CONS(hp, make_small(0), NIL);
-    hp += 2;
-
-    while(is_list(env)) {
-	Eterm tmp;
-	Eterm* tp;
-
-	tmp = CAR(list_val(env));
-	if (is_not_tuple_arity(tmp, 2)) {
-	    goto done;
-	}
-	tp = tuple_val(tmp);
-	tmp = CONS(hp, make_small(0), NIL);
-	hp += 2;
-	if (tp[2] != am_false) {
-	    tmp = CONS(hp, tp[2], tmp);
-	    hp += 2;
-	}
-	tmp = CONS(hp, make_small('='), tmp);
-	hp += 2;
-	tmp = CONS(hp, tp[1], tmp);
-	hp += 2;
-	all = CONS(hp, tmp, all);
-	hp += 2;
-	env = CDR(list_val(env));
-    }
-    if (is_not_nil(env)) {
-	goto done;
-    }
-
-    if ((size = erts_native_filename_need(all,encoding)) < 0) {
-	goto done;
-    }
-
-    /*
-     * Put the result in a binary (no risk for a memory leak that way).
-     */
-    (void) erts_new_heap_binary(p, NULL, size, &bytes);
-    erts_native_filename_put(all,encoding,bytes);
-
- done:
-    erts_free(ERTS_ALC_T_TMP, temp_heap);
-    return bytes;
-}
 
 /* ------------ decode_packet() and friends: */
 
@@ -1169,7 +1137,7 @@ http_bld_string(struct packet_callback_args* pca, Uint **hpp, Uint *szp,
 		ErlHeapBin* bin = (ErlHeapBin*) *hpp;
 		bin->thing_word = header_heap_bin(len);
 		bin->size = len;
-		memcpy(bin->data, str, len);
+		sys_memcpy(bin->data, str, len);
 		*hpp += size;
 	    }
 	}
@@ -1347,9 +1315,9 @@ int ssl_tls_erl(void* arg, unsigned type, unsigned major, unsigned minor,
     Eterm bin = new_binary(pca->p, NULL, plen+len);
     byte* bin_ptr = binary_bytes(bin);
 
-    memcpy(bin_ptr+plen, buf, len);
+    sys_memcpy(bin_ptr+plen, buf, len);
     if (plen) {
-        memcpy(bin_ptr, prefix, plen);
+        sys_memcpy(bin_ptr, prefix, plen);
     }
 
     /* {ssl_tls,NIL,ContentType,{Major,Minor},Bin} */
diff --git a/erts/emulator/beam/erl_bif_re.c b/erts/emulator/beam/erl_bif_re.c
index ad124fd979..4d769c2d46 100644
--- a/erts/emulator/beam/erl_bif_re.c
+++ b/erts/emulator/beam/erl_bif_re.c
@@ -66,11 +66,7 @@ static void erts_erts_pcre_stack_free(void *ptr) {
 
 #define ERTS_PCRE_STACK_MARGIN (10*1024)
 
-#ifdef ERTS_SMP
 #  define ERTS_STACK_LIMIT ((char *) ethr_get_stacklimit())
-#else
-#  define ERTS_STACK_LIMIT ((char *) erts_scheduler_stack_limit)
-#endif
 
 static int
 stack_guard_downwards(void)
@@ -518,7 +514,7 @@ re_version_0(BIF_ALIST_0)
     Eterm ret;
     size_t version_size = 0;
     byte *version = (byte *) erts_pcre_version();
-    version_size = strlen((const char *) version);
+    version_size = sys_strlen((const char *) version);
     ret = new_binary(BIF_P, version, version_size);
     BIF_RET(ret);
 }
@@ -1002,7 +998,7 @@ build_capture(Eterm capture_spec[CAPSPEC_SIZE], const pcre *code)
 	    has_dupnames = ((options & PCRE_DUPNAMES) != 0);
 
 	    for(i=0;i<top;++i) {
-		if (last == NULL || !has_dupnames || strcmp((char *) last+2,(char *) nametable+2)) {
+		if (last == NULL || !has_dupnames || sys_strcmp((char *) last+2,(char *) nametable+2)) {
 		    ASSERT(ri->num_spec >= 0);
 		    ++(ri->num_spec);
 		    if(ri->num_spec > sallocated) {
@@ -1052,7 +1048,7 @@ build_capture(Eterm capture_spec[CAPSPEC_SIZE], const pcre *code)
 						    (tmpbsiz = ap->len + 1));
 			    }
 			}
-			memcpy(tmpb,ap->name,ap->len);
+			sys_memcpy(tmpb,ap->name,ap->len);
 			tmpb[ap->len] = '\0';
 		    } else {
 			ErlDrvSizeT slen;
@@ -1214,7 +1210,7 @@ re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3)
 	    erts_pcre_fullinfo(result, NULL, PCRE_INFO_CAPTURECOUNT, &capture_count);
 	    ovsize = 3*(capture_count+1);
 	    restart.code = erts_alloc(ERTS_ALC_T_RE_SUBJECT, code_size);
-	    memcpy(restart.code, result, code_size);
+	    sys_memcpy(restart.code, result, code_size);
 	    erts_pcre_free(result);
 	    erts_free(ERTS_ALC_T_RE_TMP_BUF, expr);
 	    /*unicode = (pflags & PARSE_FLAG_UNICODE) ? 1 : 0;*/
@@ -1256,7 +1252,7 @@ re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3)
 	    BIF_ERROR(p, BADARG);
 	}
 	restart.code = erts_alloc(ERTS_ALC_T_RE_SUBJECT, code_size);
-	memcpy(restart.code, code_tmp, code_size);
+	sys_memcpy(restart.code, code_tmp, code_size);
 	erts_free_aligned_binary_bytes(temp_alloc);
 
     }
@@ -1368,7 +1364,7 @@ handle_iolist:
 	RestartContext *restartp = ERTS_MAGIC_BIN_DATA(mbp);
 	Eterm magic_ref;
 	Eterm *hp;
-	memcpy(restartp,&restart,sizeof(RestartContext));
+	sys_memcpy(restartp,&restart,sizeof(RestartContext));
 	BUMP_ALL_REDS(p);
 	hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE);
 	magic_ref = erts_mk_magic_ref(&hp, &MSO(p), mbp);
@@ -1521,7 +1517,7 @@ re_inspect_2(BIF_ALIST_2)
     last = NULL;
     name = nametable;
     for(i=0;i<top;++i) {
-	if (last == NULL || !has_dupnames || strcmp((char *) last+2,
+	if (last == NULL || !has_dupnames || sys_strcmp((char *) last+2,
 						    (char *) name+2)) {
 	    ++num_names;
 	}
@@ -1535,9 +1531,9 @@ re_inspect_2(BIF_ALIST_2)
     name = nametable;
     j = 0;
     for(i=0;i<top;++i) {
-	if (last == NULL || !has_dupnames || strcmp((char *) last+2,
+	if (last == NULL || !has_dupnames || sys_strcmp((char *) last+2,
 						    (char *) name+2)) {
-	    tmp_vec[j++] = new_binary(BIF_P, (byte *) name+2, strlen((char *) name+2));
+	    tmp_vec[j++] = new_binary(BIF_P, (byte *) name+2, sys_strlen((char *) name+2));
 	}
 	last = name;
 	name += entrysize;
diff --git a/erts/emulator/beam/erl_bif_trace.c b/erts/emulator/beam/erl_bif_trace.c
index 45159c4392..9861483bf0 100644
--- a/erts/emulator/beam/erl_bif_trace.c
+++ b/erts/emulator/beam/erl_bif_trace.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1999-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1999-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -40,6 +40,7 @@
 #include "erl_binary.h"
 #include "erl_thr_progress.h"
 #include "erl_bif_unique.h"
+#include "erl_proc_sig_queue.h"
 
 #define DECL_AM(S) Eterm AM_ ## S = am_atom_put(#S, sizeof(#S) - 1)
 
@@ -60,10 +61,8 @@ static struct {			/* Protected by code write permission */
     int local;
     BpFunctions f;		/* Local functions */
     BpFunctions e;		/* Export entries */
-#ifdef ERTS_SMP
     Process* stager;
     ErtsThrPrgrLaterOp lop;
-#endif
 } finish_bp;
 
 static Eterm
@@ -71,9 +70,7 @@ trace_pattern(Process* p, Eterm MFA, Eterm Pattern, Eterm flaglist);
 static int
 erts_set_tracing_event_pattern(Eterm event, Binary*, int on);
 
-#ifdef ERTS_SMP
 static void smp_bp_finisher(void* arg);
-#endif
 static BIF_RETTYPE
 system_monitor(Process *p, Eterm monitor_pid, Eterm list);
 
@@ -345,7 +342,6 @@ trace_pattern(Process* p, Eterm MFA, Eterm Pattern, Eterm flaglist)
 
     ERTS_TRACER_CLEAR(&meta_tracer);
 
-#ifdef ERTS_SMP
     if (finish_bp.current >= 0) {
 	ASSERT(matches >= 0);
 	ASSERT(finish_bp.stager == NULL);
@@ -355,7 +351,6 @@ trace_pattern(Process* p, Eterm MFA, Eterm Pattern, Eterm flaglist)
 	erts_suspend(p, ERTS_PROC_LOCK_MAIN, NULL);
 	ERTS_BIF_YIELD_RETURN(p, make_small(matches));
     }
-#endif
 
     erts_release_code_write_permission();
 
@@ -367,7 +362,6 @@ trace_pattern(Process* p, Eterm MFA, Eterm Pattern, Eterm flaglist)
     }
 }
 
-#ifdef ERTS_SMP
 static void smp_bp_finisher(void* null)
 {
     if (erts_finish_breakpointing()) { /* Not done */
@@ -380,15 +374,14 @@ static void smp_bp_finisher(void* null)
 	finish_bp.stager = NULL;
 #endif
 	erts_release_code_write_permission();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	if (!ERTS_PROC_IS_EXITING(p)) {
 	    erts_resume(p, ERTS_PROC_LOCK_STATUS);
 	}
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 	erts_proc_dec_refc(p);
     }
 }
-#endif /* ERTS_SMP */
 
 void
 erts_get_default_trace_pattern(int *trace_pattern_is_on,
@@ -397,8 +390,8 @@ erts_get_default_trace_pattern(int *trace_pattern_is_on,
 			       struct trace_pattern_flags *trace_pattern_flags,
 			       ErtsTracer *meta_tracer)
 {
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission() ||
-		       erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_has_code_write_permission() ||
+		       erts_thr_progress_is_blocking());
     if (trace_pattern_is_on)
 	*trace_pattern_is_on = erts_default_trace_pattern_is_on;
     if (match_spec)
@@ -413,8 +406,8 @@ erts_get_default_trace_pattern(int *trace_pattern_is_on,
 
 int erts_is_default_trace_enabled(void)
 {
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission() ||
-		       erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_has_code_write_permission() ||
+		       erts_thr_progress_is_blocking());
     return erts_default_trace_pattern_is_on;
 }
 
@@ -543,9 +536,7 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
     int matches = 0;
     Uint mask = 0;
     int cpu_ts = 0;
-#ifdef ERTS_SMP
     int system_blocked = 0;
-#endif
 
     if (! erts_trace_flags(list, &mask, &tracer, &cpu_ts)) {
 	BIF_ERROR(p, BADARG);
@@ -620,13 +611,13 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
 	    goto error;
 
         if (start_trace(tracee_p, tracer, &tracee_p->common, on, mask)) {
-	    erts_smp_proc_unlock(tracee_p,
+	    erts_proc_unlock(tracee_p,
 				 (tracee_p == p
 				  ? ERTS_PROC_LOCKS_ALL_MINOR
 				  : ERTS_PROC_LOCKS_ALL));
 	    goto already_traced;
         }
-        erts_smp_proc_unlock(tracee_p,
+        erts_proc_unlock(tracee_p,
 			     (tracee_p == p
 			      ? ERTS_PROC_LOCKS_ALL_MINOR
 			      : ERTS_PROC_LOCKS_ALL));
@@ -652,12 +643,12 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
 			    SysTimespec tp;
 			    int i;
 			    
-			    if (sys_get_proc_cputime(start, tp) < 0)
+			    if (sys_get_cputime(start, tp) < 0)
 				goto error;
 			    start = ((SysCpuTime)tp.tv_sec * 1000000000LL) + 
 				    (SysCpuTime)tp.tv_nsec;
 			    for (i = 0; i < 100; i++)
-				sys_get_proc_cputime(stop, tp);
+				sys_get_cputime(stop, tp);
 			    stop = ((SysCpuTime)tp.tv_sec * 1000000000LL) + 
 				   (SysCpuTime)tp.tv_nsec;
 			    if (start == 0) goto error;
@@ -699,11 +690,9 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
                     mods = 1;
             }
 
-#ifdef ERTS_SMP
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-	    erts_smp_thr_progress_block();
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_block();
 	    system_blocked = 1;
-#endif
 
 	    ok = 1;
 	    if (procs || mods) {
@@ -766,12 +755,10 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
 	    goto error;
     }
 
-#ifdef ERTS_SMP
     if (system_blocked) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
     erts_release_code_write_permission();
     ERTS_TRACER_CLEAR(&tracer);
 
@@ -785,12 +772,10 @@ Eterm erts_internal_trace_3(BIF_ALIST_3)
 
     ERTS_TRACER_CLEAR(&tracer);
 
-#ifdef ERTS_SMP
     if (system_blocked) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
     erts_release_code_write_permission();
 
     BIF_ERROR(p, BADARG);
@@ -824,10 +809,129 @@ Eterm trace_info_2(BIF_ALIST_2)
 	BIF_ERROR(p, BADARG);
     }
     erts_release_code_write_permission();
+
+    if (is_internal_ref(res))
+        BIF_TRAP1(erts_await_result, BIF_P, res);
+
     BIF_RET(res);
 }
 
 static Eterm
+build_trace_flags_term(Eterm **hpp, Uint *szp, Uint trace_flags)
+{
+
+#define ERTS_TFLAG__(F, FN)                             \
+    if (trace_flags & F) {                              \
+        if (szp)                                        \
+            sz += 2;                                    \
+        if (hp) {                                       \
+            res = CONS(hp, FN, res);                    \
+            hp += 2;                                    \
+        }                                               \
+    }
+
+    Eterm res;
+    Uint sz = 0;
+    Eterm *hp;
+
+    if (hpp) {
+        hp = *hpp;
+        res = NIL;
+    }
+    else {
+        hp = NULL;
+        res = THE_NON_VALUE;
+    }
+
+    ERTS_TFLAG__(F_NOW_TS, am_timestamp);
+    ERTS_TFLAG__(F_STRICT_MON_TS, am_strict_monotonic_timestamp);
+    ERTS_TFLAG__(F_MON_TS, am_monotonic_timestamp);
+    ERTS_TFLAG__(F_TRACE_SEND, am_send);
+    ERTS_TFLAG__(F_TRACE_RECEIVE, am_receive);
+    ERTS_TFLAG__(F_TRACE_SOS, am_set_on_spawn);
+    ERTS_TFLAG__(F_TRACE_CALLS, am_call);
+    ERTS_TFLAG__(F_TRACE_PROCS, am_procs);
+    ERTS_TFLAG__(F_TRACE_SOS1, am_set_on_first_spawn);
+    ERTS_TFLAG__(F_TRACE_SOL, am_set_on_link);
+    ERTS_TFLAG__(F_TRACE_SOL1, am_set_on_first_link);
+    ERTS_TFLAG__(F_TRACE_SCHED, am_running);
+    ERTS_TFLAG__(F_TRACE_SCHED_EXIT, am_exiting);
+    ERTS_TFLAG__(F_TRACE_GC, am_garbage_collection);
+    ERTS_TFLAG__(F_TRACE_ARITY_ONLY, am_arity);
+    ERTS_TFLAG__(F_TRACE_RETURN_TO, am_return_to);
+    ERTS_TFLAG__(F_TRACE_SILENT, am_silent);
+    ERTS_TFLAG__(F_TRACE_SCHED_NO, am_scheduler_id);
+    ERTS_TFLAG__(F_TRACE_PORTS, am_ports);
+    ERTS_TFLAG__(F_TRACE_SCHED_PORTS, am_running_ports);
+    ERTS_TFLAG__(F_TRACE_SCHED_PROCS, am_running_procs);
+
+    if (szp)
+        *szp += sz;
+
+    if (hpp)
+        *hpp = hp;
+
+    return res;
+
+#undef ERTS_TFLAG__
+}
+
+static Eterm
+trace_info_tracee(Process *c_p, void *arg, int *redsp, ErlHeapFragment **bpp)
+{
+    ErlHeapFragment *bp;
+    Eterm *hp, res, key;
+    Uint sz;
+
+    *redsp = 1;
+
+    if (ERTS_PROC_IS_EXITING(c_p))
+        return am_undefined;
+
+    key = (Eterm) arg;
+    sz = 3;
+
+    if (!ERTS_TRACER_IS_NIL(ERTS_TRACER(c_p)))
+        erts_is_tracer_proc_enabled(c_p, ERTS_PROC_LOCK_MAIN,
+                                    &c_p->common);
+
+    switch (key) {
+    case am_tracer:
+
+        erts_build_tracer_to_term(NULL, NULL, &sz, ERTS_TRACER(c_p));
+        bp = new_message_buffer(sz);
+        hp = bp->mem;
+        res = erts_build_tracer_to_term(&hp, &bp->off_heap,
+                                        NULL, ERTS_TRACER(c_p));
+        if (res == am_false)
+            res = NIL;
+        break;
+
+    case am_flags:
+
+        build_trace_flags_term(NULL, &sz, ERTS_TRACE_FLAGS(c_p));
+        bp = new_message_buffer(sz);
+        hp = bp->mem;
+        res = build_trace_flags_term(&hp, NULL, ERTS_TRACE_FLAGS(c_p));
+        break;
+
+    default:
+
+        ERTS_INTERNAL_ERROR("Key not supported");
+        res = NIL;
+        bp = NULL;
+        hp = NULL;
+        break;
+    }
+
+    *redsp += 2;
+
+    res = TUPLE2(hp, key, res);
+    *bpp = bp;
+    return res;
+}
+
+static Eterm
 trace_info_pid(Process* p, Eterm pid_spec, Eterm key)
 {
     Eterm tracer;
@@ -861,24 +965,19 @@ trace_info_pid(Process* p, Eterm pid_spec, Eterm key)
         erts_port_release(tracee);
 
     } else if (is_internal_pid(pid_spec)) {
-	Process *tracee = erts_pid2proc_not_running(p, ERTS_PROC_LOCK_MAIN,
-                                                    pid_spec, ERTS_PROC_LOCK_MAIN);
-
-        if (tracee == ERTS_PROC_LOCK_BUSY)
-            ERTS_BIF_YIELD2(bif_export[BIF_trace_info_2], p, pid_spec, key);
+        Eterm ref;
 
-	if (!tracee)
-	    return am_undefined;
+        if (key != am_flags && key != am_tracer)
+            goto error;
 
-        if (!ERTS_TRACER_IS_NIL(ERTS_TRACER(tracee)))
-            erts_is_tracer_proc_enabled(tracee, ERTS_PROC_LOCK_MAIN,
-                                        &tracee->common);
+        ref = erts_proc_sig_send_rpc_request(p, pid_spec, !0,
+                                             trace_info_tracee,
+                                             (void *) key);
 
-        tracer = erts_tracer_to_term(p, ERTS_TRACER(tracee));
-        trace_flags = ERTS_TRACE_FLAGS(tracee);
+        if (is_non_value(ref))
+            return am_undefined;
 
-	if (tracee != p)
-	    erts_smp_proc_unlock(tracee, ERTS_PROC_LOCK_MAIN);
+        return ref;
     } else if (is_external_pid(pid_spec)
 	       && external_pid_dist_entry(pid_spec) == erts_this_dist_entry) {
 	    return am_undefined;
@@ -888,48 +987,16 @@ trace_info_pid(Process* p, Eterm pid_spec, Eterm key)
     }
 
     if (key == am_flags) {
-	int num_flags = 21;	/* MAXIMUM number of flags. */
-	Uint needed = 3+2*num_flags;
-	Eterm flag_list = NIL;
-	Eterm* limit;
+	Eterm flag_list;
+        Uint sz = 3;
+        Eterm *hp;
 
-#define FLAG0(flag_mask,flag) \
-  if (trace_flags & (flag_mask)) { flag_list = CONS(hp, flag, flag_list); hp += 2; } else {}
+        build_trace_flags_term(NULL, &sz, trace_flags);
+
+        hp = HAlloc(p, sz);
+
+        flag_list = build_trace_flags_term(&hp, NULL, trace_flags);
 
-#if defined(DEBUG)
-    /*
-     * Check num_flags if this assertion fires.
-     */
-#  define FLAG ASSERT(num_flags-- > 0); FLAG0
-#else
-#  define FLAG FLAG0
-#endif
-        hp = HAlloc(p, needed);
-	limit = hp+needed;
-	FLAG(F_NOW_TS, am_timestamp);
-	FLAG(F_STRICT_MON_TS, am_strict_monotonic_timestamp);
-	FLAG(F_MON_TS, am_monotonic_timestamp);
-	FLAG(F_TRACE_SEND, am_send);
-	FLAG(F_TRACE_RECEIVE, am_receive);
-	FLAG(F_TRACE_SOS, am_set_on_spawn);
-	FLAG(F_TRACE_CALLS, am_call);
-	FLAG(F_TRACE_PROCS, am_procs);
-	FLAG(F_TRACE_SOS1, am_set_on_first_spawn);
-	FLAG(F_TRACE_SOL, am_set_on_link);
-	FLAG(F_TRACE_SOL1, am_set_on_first_link);
-	FLAG(F_TRACE_SCHED, am_running);
-	FLAG(F_TRACE_SCHED_EXIT, am_exiting);
-	FLAG(F_TRACE_GC, am_garbage_collection);
-	FLAG(F_TRACE_ARITY_ONLY, am_arity);
-	FLAG(F_TRACE_RETURN_TO, am_return_to);
-	FLAG(F_TRACE_SILENT, am_silent);
-	FLAG(F_TRACE_SCHED_NO, am_scheduler_id);
-	FLAG(F_TRACE_PORTS, am_ports);
-	FLAG(F_TRACE_SCHED_PORTS, am_running_ports);
-	FLAG(F_TRACE_SCHED_PROCS, am_running_procs);
-#undef FLAG0
-#undef FLAG
-	HRelease(p,limit,hp+3);
 	return TUPLE2(hp, key, flag_list);
     } else if (key == am_tracer) {
         if (tracer == am_false)
@@ -982,12 +1049,12 @@ static int function_is_traced(Process *p,
     if ((ep = export_get(&e)) != NULL) {
 	pc = ep->beam;
 	if (ep->addressv[erts_active_code_ix()] == pc &&
-	    *pc != (BeamInstr) em_call_error_handler) {
+	    ! BeamIsOpCode(*pc, op_call_error_handler)) {
 
 	    int r = 0;
 
-	    ASSERT(*pc == (BeamInstr) em_apply_bif ||
-		   *pc == (BeamInstr) BeamOp(op_i_generic_breakpoint));
+	    ASSERT(BeamIsOpCode(*pc, op_apply_bif) ||
+		   BeamIsOpCode(*pc, op_i_generic_breakpoint));
 
 	    if (erts_is_trace_break(&ep->info, ms, 0)) {
 		return FUNC_TRACE_GLOBAL_TRACE;
@@ -1055,28 +1122,20 @@ trace_info_func(Process* p, Eterm func_spec, Eterm key)
     mfa[1] = tp[2];
     mfa[2] = signed_val(tp[3]);
 
-#ifdef ERTS_SMP
     if ( (key == am_call_time) || (key == am_all)) {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
     }
-#endif
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_mtx_lock(&erts_dirty_bp_ix_mtx);
-#endif
+    erts_mtx_lock(&erts_dirty_bp_ix_mtx);
 
 
     r = function_is_traced(p, mfa, &ms, &ms_meta, &meta, &count, &call_time);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_mtx_unlock(&erts_dirty_bp_ix_mtx);
-#endif
-#ifdef ERTS_SMP
+    erts_mtx_unlock(&erts_dirty_bp_ix_mtx);
     if ( (key == am_call_time) || (key == am_all)) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
 
     switch (r) {
     case FUNC_TRACE_NOEXIST:
@@ -1385,14 +1444,14 @@ erts_set_trace_pattern(Process*p, ErtsCodeMFA *mfa, int specified,
 	    if (ep->addressv[code_ix] != pc) {
 		fp[i].mod->curr.num_traced_exports++;
 #ifdef DEBUG
-		ep->info.op = (BeamInstr) BeamOp(op_i_func_info_IaaI);
+		ep->info.op = BeamOpCodeAddr(op_i_func_info_IaaI);
 #endif
-                ep->beam[0] = (BeamInstr) BeamOp(op_jump_f);
+                ep->beam[0] = BeamOpCodeAddr(op_trace_jump_W);
 		ep->beam[1] = (BeamInstr) ep->addressv[code_ix];
 	    }
 	    erts_set_call_trace_bif(ci, match_prog_set, 0);
 	    if (ep->addressv[code_ix] != pc) {
-		ep->beam[0] = (BeamInstr) BeamOp(op_i_generic_breakpoint);
+		ep->beam[0] = BeamOpCodeAddr(op_i_generic_breakpoint);
 	    }
 	} else if (!on && flags.breakpoint) {
 	    /* Turn off breakpoint tracing -- nothing to do here. */
@@ -1402,8 +1461,8 @@ erts_set_trace_pattern(Process*p, ErtsCodeMFA *mfa, int specified,
 	     * before turning on breakpoint tracing.
 	     */
 	    erts_clear_call_trace_bif(ci, 0);
-	    if (ep->beam[0] == (BeamInstr) BeamOp(op_i_generic_breakpoint)) {
-		ep->beam[0] = (BeamInstr) BeamOp(op_jump_f);
+	    if (BeamIsOpCode(ep->beam[0], op_i_generic_breakpoint)) {
+		ep->beam[0] = BeamOpCodeAddr(op_trace_jump_W);
 	    }
 	}
     }
@@ -1526,17 +1585,13 @@ erts_set_trace_pattern(Process*p, ErtsCodeMFA *mfa, int specified,
     finish_bp.install = on;
     finish_bp.local = flags.breakpoint;
 
-#ifdef ERTS_SMP
     if (is_blocking) {
-	ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking());
-#endif
+	ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
 	while (erts_finish_breakpointing()) {
 	    /* Empty loop body */
 	}
-#ifdef ERTS_SMP
 	finish_bp.current = -1;
     }
-#endif
 
     if (flags.breakpoint) {
 	matches += finish_bp.f.matched;
@@ -1571,11 +1626,6 @@ erts_set_tracing_event_pattern(Eterm event, Binary* match_spec, int on)
     finish_bp.f.matched = 0;
     finish_bp.f.matching = NULL;
 
-#ifndef ERTS_SMP
-    while (erts_finish_breakpointing()) {
-        /* Empty loop body */
-    }
-#endif
     return 1;
 }
 
@@ -1594,7 +1644,7 @@ consolidate_event_tracing(ErtsTracingEvent te[])
 int
 erts_finish_breakpointing(void)
 {
-    ERTS_SMP_LC_ASSERT(erts_has_code_write_permission());
+    ERTS_LC_ASSERT(erts_has_code_write_permission());
 
     /*
      * Memory barriers will be issued for all schedulers *before*
@@ -1704,7 +1754,7 @@ uninstall_exp_breakpoints(BpFunctions* f)
 	if (ep->addressv[code_ix] != ep->beam) {
 	    continue;
 	}
-	ASSERT(ep->beam[0] == (BeamInstr) BeamOp(op_jump_f));
+	ASSERT(BeamIsOpCode(ep->beam[0], op_trace_jump_W));
 	ep->addressv[code_ix] = (BeamInstr *) ep->beam[1];
     }
 }
@@ -1723,7 +1773,7 @@ clean_export_entries(BpFunctions* f)
 	if (ep->addressv[code_ix] == ep->beam) {
 	    continue;
 	}
-	if (ep->beam[0] == (BeamInstr) BeamOp(op_jump_f)) {
+	if (BeamIsOpCode(ep->beam[0], op_trace_jump_W)) {
 	    ep->beam[0] = (BeamInstr) 0;
 	    ep->beam[1] = (BeamInstr) 0;
 	}
@@ -1839,9 +1889,6 @@ Eterm erts_seq_trace(Process *p, Eterm arg1, Eterm arg2,
 	return old_value;
     }
     else if (arg1 == am_label) {
-	if (! is_small(arg2)) {
-	    return THE_NON_VALUE;
-	}
         new_seq_trace_token(p);
 	if (build_result) {
 	    old_value = SEQ_TRACE_TOKEN_LABEL(p);
@@ -2015,24 +2062,20 @@ BIF_RETTYPE seq_trace_print_2(BIF_ALIST_2)
 }
 
 void erts_system_monitor_clear(Process *c_p) {
-#ifdef ERTS_SMP
     if (c_p) {
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
     }
-#endif
     erts_set_system_monitor(NIL);
     erts_system_monitor_long_gc = 0;
     erts_system_monitor_long_schedule = 0;
     erts_system_monitor_large_heap = 0;
     erts_system_monitor_flags.busy_port = 0;
     erts_system_monitor_flags.busy_dist_port = 0;
-#ifdef ERTS_SMP
     if (c_p) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
 }
 
 
@@ -2142,8 +2185,8 @@ system_monitor(Process *p, Eterm monitor_pid, Eterm list)
 	int busy_port, busy_dist_port;
 
 	system_blocked = 1;
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	if (!erts_pid2proc(p, ERTS_PROC_LOCK_MAIN, monitor_pid, 0))
 	    goto error;
@@ -2182,16 +2225,16 @@ system_monitor(Process *p, Eterm monitor_pid, Eterm list)
 	erts_system_monitor_flags.busy_port = !!busy_port;
 	erts_system_monitor_flags.busy_dist_port = !!busy_dist_port;
 
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
 	BIF_RET(prev);
     }
 
  error:
 
     if (system_blocked) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     }
 
     BIF_ERROR(p, BADARG);
@@ -2200,23 +2243,19 @@ system_monitor(Process *p, Eterm monitor_pid, Eterm list)
 /* Begin: Trace for System Profiling */
 
 void erts_system_profile_clear(Process *c_p) {
-#ifdef ERTS_SMP
     if (c_p) {
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
     }
-#endif
     erts_set_system_profile(NIL);
     erts_system_profile_flags.scheduler = 0;
     erts_system_profile_flags.runnable_procs = 0;
     erts_system_profile_flags.runnable_ports = 0;
     erts_system_profile_flags.exclusive = 0;
-#ifdef ERTS_SMP
     if (c_p) {
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
 }
 
 static Eterm system_profile_get(Process *p) {
@@ -2278,8 +2317,8 @@ BIF_RETTYPE system_profile_2(BIF_ALIST_2)
 	int scheduler, runnable_procs, runnable_ports, exclusive;
 	system_blocked = 1;
 	
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-	erts_smp_thr_progress_block();
+	erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_block();
 
 	/* Check if valid process, no locks are taken */
 
@@ -2330,8 +2369,8 @@ BIF_RETTYPE system_profile_2(BIF_ALIST_2)
 	erts_system_profile_flags.runnable_procs = !!runnable_procs;
 	erts_system_profile_flags.exclusive = !!exclusive;
 	erts_system_profile_ts_type = ts;
-	erts_smp_thr_progress_unblock();
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	erts_thr_progress_unblock();
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
 	
 	BIF_RET(prev);
 		
@@ -2339,8 +2378,8 @@ BIF_RETTYPE system_profile_2(BIF_ALIST_2)
 
     error:
 	if (system_blocked) {
-	    erts_smp_thr_progress_unblock();
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
+	    erts_thr_progress_unblock();
+	    erts_proc_lock(p, ERTS_PROC_LOCK_MAIN);
     	}
 
     BIF_ERROR(p, BADARG);
@@ -2365,7 +2404,7 @@ typedef struct {
     Eterm ref;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Eterm target;
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
 } ErtsTraceDeliveredAll;
 
 static void
@@ -2373,31 +2412,20 @@ reply_trace_delivered_all(void *vtdarp)
 {
     ErtsTraceDeliveredAll *tdarp = (ErtsTraceDeliveredAll *) vtdarp;
 
-    if (erts_smp_atomic32_dec_read_nob(&tdarp->refc) == 0) {
+    if (erts_atomic32_dec_read_nob(&tdarp->refc) == 0) {
         Eterm ref_copy, msg;
         Process *rp = tdarp->proc;
         Eterm *hp = NULL;
         ErlOffHeap *ohp;
-#ifdef ERTS_SMP
         ErlHeapFragment *bp;
         bp = new_message_buffer(4 + NC_HEAP_SIZE(tdarp->ref));
         hp = &bp->mem[0];
         ohp = &bp->off_heap;
-#else
-        ErtsProcLocks rp_locks = 0;
-        ErtsMessage *mp;
-        mp = erts_alloc_message_heap(
-            rp, &rp_locks, 4 + NC_HEAP_SIZE(tdarp->ref), &hp, &ohp);
-#endif
 
         ref_copy = STORE_NC(&hp, ohp, tdarp->ref);
         msg = TUPLE3(hp, am_trace_delivered, tdarp->target, ref_copy);
 
-#ifdef ERTS_SMP
         erts_send_sys_msg_proc(rp->common.id, rp->common.id, msg, bp);
-#else
-        erts_queue_message(rp, rp_locks, mp, msg, am_system);
-#endif
 
 	erts_free(ERTS_ALC_T_MISC_AUX_WORK, vtdarp);
         erts_proc_dec_refc(rp);
@@ -2418,7 +2446,7 @@ trace_delivered_1(BIF_ALIST_1)
         hp = &tdarp->ref_heap[0];
         tdarp->ref = STORE_NC(&hp, NULL, ref);
         tdarp->target = BIF_ARG_1;
-        erts_smp_atomic32_init_nob(&tdarp->refc,
+        erts_atomic32_init_nob(&tdarp->refc,
                                    (erts_aint32_t) erts_no_schedulers);
         erts_proc_add_refc(BIF_P, 1);
         erts_schedule_multi_misc_aux_work(0,
diff --git a/erts/emulator/beam/erl_bif_unique.c b/erts/emulator/beam/erl_bif_unique.c
index 2f8adc87d5..19d46537f9 100644
--- a/erts/emulator/beam/erl_bif_unique.c
+++ b/erts/emulator/beam/erl_bif_unique.c
@@ -77,11 +77,9 @@ init_reference(void)
     ref_init_value += (Uint64) tv.tv_usec;
 #ifdef DEBUG
     max_thr_id = (Uint32) erts_no_schedulers;
-#ifdef ERTS_DIRTY_SCHEDULERS
     max_thr_id += (Uint32) erts_no_dirty_cpu_schedulers;
     max_thr_id += (Uint32) erts_no_dirty_io_schedulers;
 #endif
-#endif
     erts_atomic64_init_nob(&global_reference.count,
 			   (erts_aint64_t) ref_init_value);
     init_magic_ref_tables();
@@ -136,7 +134,7 @@ Eterm erts_make_ref(Process *c_p)
     Eterm* hp;
     Uint32 ref[ERTS_REF_NUMBERS];
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
 
     hp = HAlloc(c_p, ERTS_REF_THING_SIZE);
 
@@ -439,10 +437,8 @@ init_unique_integer(void)
 {
     int bits;
     unique_data.r.o.val0_max = (Uint64) erts_no_schedulers;
-#ifdef ERTS_DIRTY_SCHEDULERS
     unique_data.r.o.val0_max += (Uint64) erts_no_dirty_cpu_schedulers;
     unique_data.r.o.val0_max += (Uint64) erts_no_dirty_io_schedulers;
-#endif
     bits = erts_fit_in_bits_int64(unique_data.r.o.val0_max);
     unique_data.r.o.left_shift = bits;
     unique_data.r.o.right_shift = 64 - bits;
@@ -803,7 +799,7 @@ BIF_RETTYPE make_ref_0(BIF_ALIST_0)
     BIF_RETTYPE res;
     Eterm* hp;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(BIF_P));
 
     hp = HAlloc(BIF_P, ERTS_REF_THING_SIZE);
 
diff --git a/erts/emulator/beam/erl_bif_unique.h b/erts/emulator/beam/erl_bif_unique.h
index 9aa631fde9..6af9d4ac8c 100644
--- a/erts/emulator/beam/erl_bif_unique.h
+++ b/erts/emulator/beam/erl_bif_unique.h
@@ -307,7 +307,7 @@ erts_iref_storage_clean(ErtsIRefStorage *iref)
     if (iref->is_magic && erts_refc_dectest(&iref->u.mb->intern.refc, 0) == 0)
 	erts_ref_bin_free(iref->u.mb);
 #ifdef DEBUG
-    memset((void *) iref, 0xf, sizeof(ErtsIRefStorage));
+    sys_memset((void *) iref, 0xf, sizeof(ErtsIRefStorage));
 #endif
 }
 
@@ -342,7 +342,7 @@ erts_iref_storage_make_ref(ErtsIRefStorage *iref,
 
 #ifdef DEBUG
     if (clean_storage)
-	memset((void *) iref, 0xf, sizeof(ErtsIRefStorage));
+	sys_memset((void *) iref, 0xf, sizeof(ErtsIRefStorage));
 #endif
 
     return make_internal_ref(hp);
diff --git a/erts/emulator/beam/erl_binary.h b/erts/emulator/beam/erl_binary.h
index b036b28dbf..7dfd0c273a 100644
--- a/erts/emulator/beam/erl_binary.h
+++ b/erts/emulator/beam/erl_binary.h
@@ -146,9 +146,7 @@ typedef union {
 
 /* A "magic" binary flag */
 #define BIN_FLAG_MAGIC      1
-#define BIN_FLAG_USR1       2 /* Reserved for use by different modules too mark */
-#define BIN_FLAG_USR2       4 /*  certain binaries as special (used by ets) */
-#define BIN_FLAG_DRV        8
+#define BIN_FLAG_DRV        2
 
 #endif /* ERL_BINARY_H__TYPES__ */
 
@@ -291,7 +289,7 @@ typedef union {
      * atomics are used they might
      * differ in size.
      */
-    erts_smp_atomic_t smp_atomic_word;
+    erts_atomic_t smp_atomic_word;
     erts_atomic_t atomic_word;
 } ErtsMagicIndirectionWord;
 
@@ -326,7 +324,7 @@ ERTS_GLB_INLINE Binary *erts_create_magic_binary_x(Uint size,
 ERTS_GLB_INLINE Binary *erts_create_magic_binary(Uint size,
 						 int (*destructor)(Binary *));
 ERTS_GLB_INLINE Binary *erts_create_magic_indirection(int (*destructor)(Binary *));
-ERTS_GLB_INLINE erts_smp_atomic_t *erts_smp_binary_to_magic_indirection(Binary *bp);
+ERTS_GLB_INLINE erts_atomic_t *erts_binary_to_magic_indirection(Binary *bp);
 ERTS_GLB_INLINE erts_atomic_t *erts_binary_to_magic_indirection(Binary *bp);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
@@ -363,8 +361,6 @@ erts_free_aligned_binary_bytes(byte* buf)
 #  define CHICKEN_PAD (sizeof(void*) - 1)
 #endif
 
-/* Caller must initialize 'refc'
-*/
 ERTS_GLB_INLINE Binary *
 erts_bin_drv_alloc_fnf(Uint size)
 {
@@ -383,8 +379,6 @@ erts_bin_drv_alloc_fnf(Uint size)
     return res;
 }
 
-/* Caller must initialize 'refc'
-*/
 ERTS_GLB_INLINE Binary *
 erts_bin_drv_alloc(Uint size)
 {
@@ -401,9 +395,6 @@ erts_bin_drv_alloc(Uint size)
     return res;
 }
 
-
-/* Caller must initialize 'refc'
-*/
 ERTS_GLB_INLINE Binary *
 erts_bin_nrml_alloc(Uint size)
 {
@@ -519,16 +510,6 @@ erts_create_magic_indirection(int (*destructor)(Binary *))
                                              but word aligned */
 }
 
-ERTS_GLB_INLINE erts_smp_atomic_t *
-erts_smp_binary_to_magic_indirection(Binary *bp)
-{
-    ErtsMagicIndirectionWord *mip;
-    ASSERT(bp->intern.flags & BIN_FLAG_MAGIC);
-    ASSERT(ERTS_MAGIC_BIN_ATYPE(bp) == ERTS_ALC_T_MINDIRECTION);
-    mip = ERTS_MAGIC_BIN_UNALIGNED_DATA(bp);
-    return &mip->smp_atomic_word;
-}
-
 ERTS_GLB_INLINE erts_atomic_t *
 erts_binary_to_magic_indirection(Binary *bp)
 {
@@ -536,7 +517,7 @@ erts_binary_to_magic_indirection(Binary *bp)
     ASSERT(bp->intern.flags & BIN_FLAG_MAGIC);
     ASSERT(ERTS_MAGIC_BIN_ATYPE(bp) == ERTS_ALC_T_MINDIRECTION);
     mip = ERTS_MAGIC_BIN_UNALIGNED_DATA(bp);
-    return &mip->atomic_word;
+    return &mip->smp_atomic_word;
 }
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
diff --git a/erts/emulator/beam/erl_bits.c b/erts/emulator/beam/erl_bits.c
index 51d23a8965..3a16913473 100644
--- a/erts/emulator/beam/erl_bits.c
+++ b/erts/emulator/beam/erl_bits.c
@@ -55,30 +55,19 @@
 
 static byte get_bit(byte b, size_t a_offs); 
 
-#if defined(ERTS_SMP)
 /* the state resides in the current process' scheduler data */
-#elif defined(ERL_BITS_REENTRANT)
-/* reentrant API but with a hidden single global state, for testing only */
-struct erl_bits_state ErlBitsState_;
-#else
-/* non-reentrant API with a single global state */
-struct erl_bits_state ErlBitsState;
-#endif
 
 #define byte_buf	(ErlBitsState.byte_buf_)
 #define byte_buf_len	(ErlBitsState.byte_buf_len_)
 
-static erts_smp_atomic_t bits_bufs_size;
+static erts_atomic_t bits_bufs_size;
 
 Uint
 erts_bits_bufs_size(void)
 {
-    return (Uint) erts_smp_atomic_read_nob(&bits_bufs_size);
+    return (Uint) erts_atomic_read_nob(&bits_bufs_size);
 }
 
-#if !defined(ERTS_SMP)
-static
-#endif
 void
 erts_bits_init_state(ERL_BITS_PROTO_0)
 {
@@ -88,13 +77,11 @@ erts_bits_init_state(ERL_BITS_PROTO_0)
     erts_bin_offset = 0;
 }
 
-#if defined(ERTS_SMP)
 void
 erts_bits_destroy_state(ERL_BITS_PROTO_0)
 {
     erts_free(ERTS_ALC_T_BITS_BUF, byte_buf);
 }
-#endif
 
 void
 erts_init_bits(void)
@@ -104,13 +91,8 @@ erts_init_bits(void)
     ERTS_CT_ASSERT(offsetof(ErtsBinary,driver.binary.orig_bytes)
                 == offsetof(Binary,orig_bytes));
 
-    erts_smp_atomic_init_nob(&bits_bufs_size, 0);
-#if defined(ERTS_SMP)
+    erts_atomic_init_nob(&bits_bufs_size, 0);
     /* erl_process.c calls erts_bits_init_state() on all state instances */
-#else
-    ERL_BITS_DECLARE_STATEP;
-    erts_bits_init_state(ERL_BITS_ARGS_0);
-#endif
 }
 
 /*****************************************************************
@@ -744,7 +726,7 @@ static void
 ERTS_INLINE need_byte_buf(ERL_BITS_PROTO_1(int need))
 {
     if (byte_buf_len < need) {
-	erts_smp_atomic_add_nob(&bits_bufs_size, need - byte_buf_len);
+	erts_atomic_add_nob(&bits_bufs_size, need - byte_buf_len);
 	byte_buf_len = need;
 	byte_buf = erts_realloc(ERTS_ALC_T_BITS_BUF, byte_buf, byte_buf_len);
     }
diff --git a/erts/emulator/beam/erl_bits.h b/erts/emulator/beam/erl_bits.h
index 5da2b28a89..a3816fa820 100644
--- a/erts/emulator/beam/erl_bits.h
+++ b/erts/emulator/beam/erl_bits.h
@@ -84,31 +84,14 @@ typedef struct erl_bin_match_struct{
 #define ms_matchbuffer(_Ms) &(((ErlBinMatchState*) boxed_val(_Ms))->mb)
 
 
-#if defined(ERTS_SMP)
-#define ERL_BITS_REENTRANT
-#else
-/* uncomment to test the reentrant API in the non-SMP runtime system */
-/* #define ERL_BITS_REENTRANT */
-#endif
-
-#ifdef ERL_BITS_REENTRANT
-
 /*
  * Reentrant API with the state passed as a parameter.
  * (Except when the current Process* already is a parameter.)
  */
-#ifdef ERTS_SMP
 /* the state resides in the current process' scheduler data */
 #define ERL_BITS_DECLARE_STATEP			struct erl_bits_state *EBS
 #define ERL_BITS_RELOAD_STATEP(P)		do{EBS = &erts_proc_sched_data((P))->erl_bits_state;}while(0)
 #define ERL_BITS_DEFINE_STATEP(P)		struct erl_bits_state *EBS = &erts_proc_sched_data((P))->erl_bits_state
-#else
-/* reentrant API but with a hidden single global state, for testing only */
-extern struct erl_bits_state ErlBitsState_;
-#define ERL_BITS_DECLARE_STATEP			struct erl_bits_state *EBS = &ErlBitsState_
-#define ERL_BITS_RELOAD_STATEP(P)		do{}while(0)
-#define ERL_BITS_DEFINE_STATEP(P)		ERL_BITS_DECLARE_STATEP
-#endif
 #define ErlBitsState				(*EBS)
 
 #define ERL_BITS_PROTO_0			struct erl_bits_state *EBS
@@ -120,26 +103,6 @@ extern struct erl_bits_state ErlBitsState_;
 #define ERL_BITS_ARGS_2(ARG1,ARG2)		EBS, ARG1, ARG2
 #define ERL_BITS_ARGS_3(ARG1,ARG2,ARG3)		EBS, ARG1, ARG2, ARG3
 
-#else	/* ERL_BITS_REENTRANT */
-
-/*
- * Non-reentrant API with a single global state.
- */
-extern struct erl_bits_state ErlBitsState;
-#define ERL_BITS_DECLARE_STATEP			/*empty*/
-#define ERL_BITS_RELOAD_STATEP(P)		do{}while(0)
-#define ERL_BITS_DEFINE_STATEP(P)		/*empty*/
-
-#define ERL_BITS_PROTO_0			void
-#define ERL_BITS_PROTO_1(PARM1)			PARM1
-#define ERL_BITS_PROTO_2(PARM1,PARM2)		PARM1, PARM2
-#define ERL_BITS_PROTO_3(PARM1,PARM2,PARM3)	PARM1, PARM2, PARM3
-#define ERL_BITS_ARGS_0				/*empty*/
-#define ERL_BITS_ARGS_1(ARG1)			ARG1
-#define ERL_BITS_ARGS_2(ARG1,ARG2)		ARG1, ARG2
-#define ERL_BITS_ARGS_3(ARG1,ARG2,ARG3)		ARG1, ARG2, ARG3
-
-#endif	/* ERL_BITS_REENTRANT */
 
 #define erts_bin_offset		(ErlBitsState.erts_bin_offset_)
 #define erts_current_bin	(ErlBitsState.erts_current_bin_)
@@ -148,7 +111,7 @@ extern struct erl_bits_state ErlBitsState;
 #define copy_binary_to_buffer(DstBuffer, DstBufOffset, SrcBuffer, SrcBufferOffset, NumBits) \
   do {											    \
     if (BIT_OFFSET(DstBufOffset) == 0 && (SrcBufferOffset == 0) &&			    \
-        (BIT_OFFSET(NumBits)==0)) {							    \
+        (BIT_OFFSET(NumBits)==0) && (NumBits != 0)) {					    \
       sys_memcpy(DstBuffer+BYTE_OFFSET(DstBufOffset),					    \
 		 SrcBuffer, NBYTES(NumBits));						    \
     } else {										    \
@@ -158,10 +121,8 @@ extern struct erl_bits_state ErlBitsState;
   }  while (0)
 
 void erts_init_bits(void);	/* Initialization once. */
-#ifdef ERTS_SMP
 void erts_bits_init_state(ERL_BITS_PROTO_0);
 void erts_bits_destroy_state(ERL_BITS_PROTO_0);
-#endif
 
 
 /*
diff --git a/erts/emulator/beam/erl_cpu_topology.c b/erts/emulator/beam/erl_cpu_topology.c
index dfe49ce382..6f8d2f8c35 100644
--- a/erts/emulator/beam/erl_cpu_topology.c
+++ b/erts/emulator/beam/erl_cpu_topology.c
@@ -60,7 +60,7 @@ static int max_main_threads;
 static int reader_groups;
 
 static ErtsCpuBindData *scheduler2cpu_map;
-static erts_smp_rwmtx_t cpuinfo_rwmtx;
+static erts_rwmtx_t cpuinfo_rwmtx;
 
 typedef enum {
     ERTS_CPU_BIND_UNDEFINED,
@@ -131,13 +131,11 @@ static erts_cpu_groups_map_t *reader_groups_map;
 
 #define ERTS_MAX_CPU_TOPOLOGY_ID ((int) 0xffff)
 
-#ifdef ERTS_SMP
 static void cpu_bind_order_sort(erts_cpu_topology_t *cpudata,
 				int size,
 				ErtsCpuBindOrder bind_order,
 				int mk_seq);
 static void write_schedulers_bind_change(erts_cpu_topology_t *cpudata, int size);
-#endif
 
 static void reader_groups_callback(int, ErtsSchedulerData *, int, void *);
 static erts_cpu_groups_map_t *add_cpu_groups(int groups,
@@ -434,7 +432,6 @@ processor_order_cmp(const void *vx, const void *vy)
     return 0;
 }
 
-#ifdef ERTS_SMP
 void
 erts_sched_check_cpu_bind_prep_suspend(ErtsSchedulerData *esdp)
 {
@@ -444,7 +441,7 @@ erts_sched_check_cpu_bind_prep_suspend(ErtsSchedulerData *esdp)
     int cgcc_ix;
 
     /* Unbind from cpu */
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
     if (scheduler2cpu_map[esdp->no].bound_id >= 0
 	&& erts_unbind_from_cpu(cpuinfo) == 0) {
 	esdp->cpu_id = scheduler2cpu_map[esdp->no].bound_id = -1;
@@ -463,7 +460,7 @@ erts_sched_check_cpu_bind_prep_suspend(ErtsSchedulerData *esdp)
 	}
     }
     ASSERT(no_cpu_groups_callbacks == cgcc_ix);
-    erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
 
     for (cgcc_ix = 0; cgcc_ix < no_cpu_groups_callbacks; cgcc_ix++)
 	cgcc[cgcc_ix].callback(1,
@@ -481,7 +478,7 @@ erts_sched_check_cpu_bind_prep_suspend(ErtsSchedulerData *esdp)
 void
 erts_sched_check_cpu_bind_post_suspend(ErtsSchedulerData *esdp)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(esdp->run_queue));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(esdp->run_queue));
 
     if (esdp->no <= max_main_threads)
 	erts_thr_set_main_status(1, (int) esdp->no);
@@ -490,7 +487,6 @@ erts_sched_check_cpu_bind_post_suspend(ErtsSchedulerData *esdp)
     (void) ERTS_RUNQ_FLGS_SET(esdp->run_queue, ERTS_RUNQ_FLG_CHK_CPU_BIND);
 }
 
-#endif
 
 void
 erts_sched_check_cpu_bind(ErtsSchedulerData *esdp)
@@ -499,8 +495,8 @@ erts_sched_check_cpu_bind(ErtsSchedulerData *esdp)
     erts_cpu_groups_map_t *cgm;
     erts_cpu_groups_callback_list_t *cgcl;
     erts_cpu_groups_callback_call_t *cgcc;
-    erts_smp_runq_unlock(esdp->run_queue);
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_runq_unlock(esdp->run_queue);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
     cpu_id = scheduler2cpu_map[esdp->no].bind_id;
     if (cpu_id >= 0 && cpu_id != scheduler2cpu_map[esdp->no].bound_id) {
 	res = erts_bind_to_cpu(cpuinfo, cpu_id);
@@ -543,7 +539,7 @@ erts_sched_check_cpu_bind(ErtsSchedulerData *esdp)
     }
 
     ASSERT(no_cpu_groups_callbacks == cgcc_ix);
-    erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
 
     for (cgcc_ix = 0; cgcc_ix < no_cpu_groups_callbacks; cgcc_ix++)
 	cgcc[cgcc_ix].callback(0,
@@ -553,10 +549,9 @@ erts_sched_check_cpu_bind(ErtsSchedulerData *esdp)
 
     erts_free(ERTS_ALC_T_TMP, cgcc);
 
-    erts_smp_runq_lock(esdp->run_queue);
+    erts_runq_lock(esdp->run_queue);
 }
 
-#ifdef ERTS_SMP
 void
 erts_sched_init_check_cpu_bind(ErtsSchedulerData *esdp)
 {
@@ -565,7 +560,7 @@ erts_sched_init_check_cpu_bind(ErtsSchedulerData *esdp)
     erts_cpu_groups_callback_list_t *cgcl;
     erts_cpu_groups_callback_call_t *cgcc;
 
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
 
     cgcc = erts_alloc(ERTS_ALC_T_TMP,
 		      (no_cpu_groups_callbacks
@@ -581,7 +576,7 @@ erts_sched_init_check_cpu_bind(ErtsSchedulerData *esdp)
     }
 
     ASSERT(no_cpu_groups_callbacks == cgcc_ix);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
 
     for (cgcc_ix = 0; cgcc_ix < no_cpu_groups_callbacks; cgcc_ix++)
 	cgcc[cgcc_ix].callback(0,
@@ -594,7 +589,6 @@ erts_sched_init_check_cpu_bind(ErtsSchedulerData *esdp)
     if (esdp->no <= max_main_threads)
 	erts_thr_set_main_status(1, (int) esdp->no);
 }
-#endif
 
 static void
 write_schedulers_bind_change(erts_cpu_topology_t *cpudata, int size)
@@ -602,7 +596,7 @@ write_schedulers_bind_change(erts_cpu_topology_t *cpudata, int size)
     int s_ix = 1;
     int cpu_ix;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
 
     if (cpu_bind_order != ERTS_CPU_BIND_NONE && size) {
 
@@ -702,9 +696,9 @@ Eterm
 erts_bound_schedulers_term(Process *c_p)
 {
     ErtsCpuBindOrder order;
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     order = cpu_bind_order;
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
     return bound_schedulers_term(order);
 }
 
@@ -717,7 +711,7 @@ erts_bind_schedulers(Process *c_p, Eterm how)
     int cpudata_size;
     ErtsCpuBindOrder old_cpu_bind_order;
 
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
 
     if (erts_bind_to_cpu(cpuinfo, -1) == -ENOTSUP) {
 	if (cpu_bind_order == ERTS_CPU_BIND_NONE
@@ -773,7 +767,7 @@ erts_bind_schedulers(Process *c_p, Eterm how)
 
  done:
 
-    erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
 
     if (notify)
 	erts_sched_notify_check_cpu_bind();
@@ -793,9 +787,9 @@ erts_sched_bind_atthrcreate_child(int unbind)
 {
     int res = 0;
     if (unbind) {
-	erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+	erts_rwmtx_rlock(&cpuinfo_rwmtx);
 	res = erts_unbind_from_cpu(cpuinfo);
-	erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+	erts_rwmtx_runlock(&cpuinfo_rwmtx);
     }
     return res;
 }
@@ -812,7 +806,7 @@ erts_sched_bind_atfork_prepare(void)
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
     int unbind = esdp != NULL && erts_is_scheduler_bound(esdp);
     if (unbind)
-	erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+	erts_rwmtx_rlock(&cpuinfo_rwmtx);
     return unbind;
 }
 
@@ -820,7 +814,7 @@ int
 erts_sched_bind_atfork_child(int unbind)
 {
     if (unbind) {
-	ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&cpuinfo_rwmtx)
+	ERTS_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&cpuinfo_rwmtx)
 			   || erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
 	return erts_unbind_from_cpu(cpuinfo);
     }
@@ -831,7 +825,7 @@ void
 erts_sched_bind_atfork_parent(int unbind)
 {
     if (unbind)
-	erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+	erts_rwmtx_runlock(&cpuinfo_rwmtx);
 }
 
 Eterm
@@ -865,9 +859,9 @@ erts_fake_scheduler_bindings(Process *p, Eterm how)
 	return res;
     }
 
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     create_tmp_cpu_topology_copy(&cpudata, &cpudata_size);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
 
     if (!cpudata || fake_cpu_bind_order == ERTS_CPU_BIND_NONE)
 	ERTS_BIF_PREP_RET(res, am_false);
@@ -930,12 +924,12 @@ erts_get_schedulers_binds(Process *c_p)
     Eterm res = make_tuple(hp);
 
     *(hp++) = make_arityval(erts_no_schedulers);
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     for (ix = 1; ix <= erts_no_schedulers; ix++)
 	*(hp++) = (scheduler2cpu_map[ix].bound_id >= 0
 		   ? make_small(scheduler2cpu_map[ix].bound_id)
 		   : AM_unbound);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
     return res;
 }
 
@@ -1346,7 +1340,7 @@ erts_set_cpu_topology(Process *c_p, Eterm term)
     int cpudata_size = 0;
     Eterm res;
 
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
     res = get_cpu_topology_term(c_p, ERTS_GET_USED_CPU_TOPOLOGY);
     if (term == am_undefined) {
 	if (user_cpudata)
@@ -1367,7 +1361,7 @@ erts_set_cpu_topology(Process *c_p, Eterm term)
     }
     else if (is_not_list(term)) {
     error:
-	erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+	erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
 	res = THE_NON_VALUE;
 	goto done;
     }
@@ -1461,7 +1455,7 @@ erts_set_cpu_topology(Process *c_p, Eterm term)
 
     write_schedulers_bind_change(cpudata, cpudata_size);
 
-    erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
     erts_sched_notify_check_cpu_bind();
 
  done:
@@ -1615,7 +1609,7 @@ erts_get_cpu_topology_term(Process *c_p, Eterm which)
 {
     Eterm res;
     int type;
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     if (ERTS_IS_ATOM_STR("used", which))
 	type = ERTS_GET_USED_CPU_TOPOLOGY;
     else if (ERTS_IS_ATOM_STR("detected", which))
@@ -1628,7 +1622,7 @@ erts_get_cpu_topology_term(Process *c_p, Eterm which)
 	res = THE_NON_VALUE;
     else
 	res = get_cpu_topology_term(c_p, type);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
     return res;
 }
 
@@ -1646,9 +1640,9 @@ get_logical_processors(int *conf, int *onln, int *avail)
 void
 erts_get_logical_processors(int *conf, int *onln, int *avail)
 {
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     get_logical_processors(conf, onln, avail);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
 }
 
 void
@@ -1706,9 +1700,9 @@ erts_init_cpu_topology(void)
 {
     int ix;
 
-    erts_smp_rwmtx_init(&cpuinfo_rwmtx, "cpu_info", NIL,
+    erts_rwmtx_init(&cpuinfo_rwmtx, "cpu_info", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
 
     scheduler2cpu_map = erts_alloc(ERTS_ALC_T_CPUDATA,
 				   (sizeof(ErtsCpuBindData)
@@ -1726,13 +1720,13 @@ erts_init_cpu_topology(void)
 				       NULL);
 
     if (cpu_bind_order == ERTS_CPU_BIND_NONE)
-	erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+	erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
     else {
 	erts_cpu_topology_t *cpudata;
 	int cpudata_size;
 	create_tmp_cpu_topology_copy(&cpudata, &cpudata_size);
 	write_schedulers_bind_change(cpudata, cpudata_size);
-	erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+	erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
 	erts_sched_notify_check_cpu_bind();
 	destroy_tmp_cpu_topology_copy(cpudata);
     }
@@ -1742,7 +1736,7 @@ int
 erts_update_cpu_info(void)
 {
     int changed;
-    erts_smp_rwmtx_rwlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwlock(&cpuinfo_rwmtx);
     changed = erts_cpu_info_update(cpuinfo);
     if (changed) {
 	erts_cpu_topology_t *cpudata;
@@ -1775,7 +1769,7 @@ erts_update_cpu_info(void)
 	write_schedulers_bind_change(cpudata, cpudata_size);
 	destroy_tmp_cpu_topology_copy(cpudata);
     }
-    erts_smp_rwmtx_rwunlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rwunlock(&cpuinfo_rwmtx);
     if (changed)
 	erts_sched_notify_check_cpu_bind();
     return changed;
@@ -1792,7 +1786,7 @@ reader_groups_callback(int suspending,
 		       void *unused)
 {
     if (reader_groups && esdp->no <= max_main_threads)
-	erts_smp_rwmtx_set_reader_group(suspending ? 0 : group+1);
+	erts_rwmtx_set_reader_group(suspending ? 0 : group+1);
 }
 
 static Eterm get_cpu_groups_map(Process *c_p,
@@ -1821,9 +1815,9 @@ Eterm
 erts_get_reader_groups_map(Process *c_p)
 {
     Eterm res;
-    erts_smp_rwmtx_rlock(&cpuinfo_rwmtx);
+    erts_rwmtx_rlock(&cpuinfo_rwmtx);
     res = get_cpu_groups_map(c_p, reader_groups_map, 1);
-    erts_smp_rwmtx_runlock(&cpuinfo_rwmtx);
+    erts_rwmtx_runlock(&cpuinfo_rwmtx);
     return res;
 }
 
@@ -2203,7 +2197,7 @@ add_cpu_groups(int groups,
     erts_cpu_groups_callback_list_t *cgcl;
     erts_cpu_groups_map_t *cgm;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
 
     if (use_groups > max_main_threads)
 	use_groups = max_main_threads;
@@ -2250,7 +2244,7 @@ cpu_groups_lookup(erts_cpu_groups_map_t *map,
 {
     int start, logical, ix;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&cpuinfo_rwmtx)
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rlocked(&cpuinfo_rwmtx)
 		       || erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
 
     if (esdp->cpu_id < 0)
@@ -2278,7 +2272,7 @@ static void
 update_cpu_groups_maps(void)
 {
     erts_cpu_groups_map_t *cgm;
-    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&cpuinfo_rwmtx));
 
     for (cgm = cpu_groups_maps; cgm; cgm = cgm->next)
 	make_cpu_groups_map(cgm, 0);
diff --git a/erts/emulator/beam/erl_cpu_topology.h b/erts/emulator/beam/erl_cpu_topology.h
index c922214702..88bcad79ab 100644
--- a/erts/emulator/beam/erl_cpu_topology.h
+++ b/erts/emulator/beam/erl_cpu_topology.h
@@ -60,11 +60,9 @@ int erts_init_scheduler_bind_type_string(char *how);
 int erts_init_cpu_topology_string(char *topology_str);
 
 void erts_sched_check_cpu_bind(ErtsSchedulerData *esdp);
-#ifdef ERTS_SMP
 void erts_sched_init_check_cpu_bind(ErtsSchedulerData *esdp);
 void erts_sched_check_cpu_bind_prep_suspend(ErtsSchedulerData *esdp);
 void erts_sched_check_cpu_bind_post_suspend(ErtsSchedulerData *esdp);
-#endif
 
 int erts_update_cpu_info(void);
 
diff --git a/erts/emulator/beam/erl_db.c b/erts/emulator/beam/erl_db.c
index 68d984014f..82e31f4cab 100644
--- a/erts/emulator/beam/erl_db.c
+++ b/erts/emulator/beam/erl_db.c
@@ -44,12 +44,40 @@
 #include "erl_binary.h"
 
 
-erts_smp_atomic_t erts_ets_misc_mem_size;
+erts_atomic_t erts_ets_misc_mem_size;
 
 /*
 ** Utility macros
 */
 
+#define DB_BIF_GET_TABLE(TB, WHAT, KIND, BIF_IX) \
+        DB_GET_TABLE(TB, BIF_ARG_1, WHAT, KIND, BIF_IX, NULL, BIF_P)
+
+#define DB_TRAP_GET_TABLE(TB, TID, WHAT, KIND, BIF_EXP) \
+        DB_GET_TABLE(TB, TID, WHAT, KIND, 0, BIF_EXP, BIF_P)
+
+#define DB_GET_TABLE(TB, TID, WHAT, KIND, BIF_IX, BIF_EXP, PROC)         \
+do {                                                                     \
+    Uint freason__;                                                      \
+    if (!(TB = db_get_table(PROC, TID, WHAT, KIND, &freason__))) {       \
+        return db_bif_fail(PROC, freason__, BIF_IX, BIF_EXP);            \
+    }                                                                    \
+}while(0)
+
+static BIF_RETTYPE db_bif_fail(Process* p, Uint freason,
+                               Uint bif_ix, Export* bif_exp)
+{
+    if (freason == TRAP) {
+        if (!bif_exp)
+            bif_exp = bif_export[bif_ix];
+        p->arity = bif_exp->info.mfa.arity;
+        p->i = (BeamInstr*) bif_exp->addressv[erts_active_code_ix()];
+    }
+    p->freason = freason;
+    return THE_NON_VALUE;
+}
+
+
 /* Get a key from any table structure and a tagged object */
 #define TERM_GETKEY(tb, obj) db_getkey((tb)->common.keypos, (obj)) 
 
@@ -61,15 +89,9 @@ enum DbIterSafety {
     ITER_SAFE_LOCKED, /* Safe while table is locked, not between trap calls */
     ITER_SAFE         /* No need to fixate at all */
 };
-#ifdef ERTS_SMP
 #  define ITERATION_SAFETY(Proc,Tab) \
     ((IS_TREE_TABLE((Tab)->common.status) || ONLY_WRITER(Proc,Tab)) ? ITER_SAFE \
      : (((Tab)->common.status & DB_FINE_LOCKED) ? ITER_UNSAFE : ITER_SAFE_LOCKED))
-#else
-#  define ITERATION_SAFETY(Proc,Tab) \
-    ((IS_TREE_TABLE((Tab)->common.status) || ONLY_WRITER(Proc,Tab)) \
-     ? ITER_SAFE : ITER_SAFE_LOCKED)
-#endif
 
 #define DID_TRAP(P,Ret) (!is_value(Ret) && ((P)->freason == TRAP))
 
@@ -195,7 +217,7 @@ static void delete_sched_table(Process *c_p, DbTable *tb);
 
 static void table_dec_refc(DbTable *tb, erts_aint_t min_val)
 {
-    if (erts_smp_refc_dectest(&tb->common.refc, min_val) == 0)
+    if (erts_refc_dectest(&tb->common.refc, min_val) == 0)
 	schedule_free_dbtable(tb);
 }
 
@@ -209,21 +231,21 @@ static ERTS_INLINE void
 make_btid(DbTable *tb)
 {
     Binary *btid = erts_create_magic_indirection(db_table_tid_destructor);
-    erts_smp_atomic_t *tbref = erts_smp_binary_to_magic_indirection(btid);
-    erts_smp_atomic_init_nob(tbref, (erts_aint_t) tb);
+    erts_atomic_t *tbref = erts_binary_to_magic_indirection(btid);
+    erts_atomic_init_nob(tbref, (erts_aint_t) tb);
     tb->common.btid = btid;
     /*
      * Table and magic indirection refer eachother,
      * and table is refered once by being alive...
      */
-    erts_smp_refc_init(&tb->common.refc, 2);
+    erts_refc_init(&tb->common.refc, 2);
     erts_refc_inc(&btid->intern.refc, 1);
 }
 
 static ERTS_INLINE DbTable* btid2tab(Binary* btid)
 {
-    erts_smp_atomic_t *tbref = erts_smp_binary_to_magic_indirection(btid);
-    return (DbTable *) erts_smp_atomic_read_nob(tbref);
+    erts_atomic_t *tbref = erts_binary_to_magic_indirection(btid);
+    return (DbTable *) erts_atomic_read_nob(tbref);
 }
 
 static DbTable *
@@ -231,7 +253,7 @@ tid2tab(Eterm tid)
 {
     DbTable *tb;
     Binary *btid;
-    erts_smp_atomic_t *tbref;
+    erts_atomic_t *tbref;
     if (!is_internal_magic_ref(tid))
         return NULL;
 
@@ -239,8 +261,8 @@ tid2tab(Eterm tid)
     if (ERTS_MAGIC_BIN_DESTRUCTOR(btid) != db_table_tid_destructor)
         return NULL;
 
-    tbref = erts_smp_binary_to_magic_indirection(btid);
-    tb = (DbTable *) erts_smp_atomic_read_nob(tbref);
+    tbref = erts_binary_to_magic_indirection(btid);
+    tb = (DbTable *) erts_atomic_read_nob(tbref);
 
     ASSERT(!tb || tb->common.btid == btid);
 
@@ -250,11 +272,11 @@ tid2tab(Eterm tid)
 static ERTS_INLINE int
 is_table_alive(DbTable *tb)
 {
-    erts_smp_atomic_t *tbref;
+    erts_atomic_t *tbref;
     DbTable *rtb;
 
-    tbref = erts_smp_binary_to_magic_indirection(tb->common.btid);
-    rtb = (DbTable *) erts_smp_atomic_read_nob(tbref);
+    tbref = erts_binary_to_magic_indirection(tb->common.btid);
+    rtb = (DbTable *) erts_atomic_read_nob(tbref);
 
     ASSERT(!rtb || rtb == tb);
 
@@ -264,11 +286,7 @@ is_table_alive(DbTable *tb)
 static ERTS_INLINE int
 is_table_named(DbTable *tb)
 {
-#ifdef ERTS_SMP
     return tb->common.type & DB_NAMED_TABLE;
-#else
-    return tb->common.status & DB_NAMED_TABLE;
-#endif
 }
 
 
@@ -277,8 +295,8 @@ tid_clear(Process *c_p, DbTable *tb)
 {
     DbTable *rtb;
     Binary *btid = tb->common.btid;
-    erts_smp_atomic_t *tbref = erts_smp_binary_to_magic_indirection(btid);
-    rtb = (DbTable *) erts_smp_atomic_xchg_nob(tbref, (erts_aint_t) NULL);
+    erts_atomic_t *tbref = erts_binary_to_magic_indirection(btid);
+    rtb = (DbTable *) erts_atomic_xchg_nob(tbref, (erts_aint_t) NULL);
     ASSERT(!rtb || tb == rtb);
     if (rtb) {
         table_dec_refc(tb, 1);
@@ -293,17 +311,22 @@ make_tid(Process *c_p, DbTable *tb)
     return erts_mk_magic_ref(&hp, &c_p->off_heap, tb->common.btid);
 }
 
+Eterm
+erts_db_make_tid(Process *c_p, DbTableCommon *tb)
+{
+    return make_tid(c_p, (DbTable*)tb);
+}
+
+
 
 /* 
 ** The meta hash table of all NAMED ets tables
 */
-#ifdef ERTS_SMP
-#  define META_NAME_TAB_LOCK_CNT 16
+#  define META_NAME_TAB_LOCK_CNT 256
 union {
-    erts_smp_rwmtx_t lck;
-    byte _cache_line_alignment[64];
+    erts_rwmtx_t lck;
+    byte align[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_rwmtx_t))];
 }meta_name_tab_rwlocks[META_NAME_TAB_LOCK_CNT];
-#endif
 static struct meta_name_tab_entry {
     union {
 	Eterm name_atom;
@@ -319,13 +342,11 @@ static unsigned meta_name_tab_mask;
 
 static ERTS_INLINE
 struct meta_name_tab_entry* meta_name_tab_bucket(Eterm name, 
-						 erts_smp_rwmtx_t** lockp)
+						 erts_rwmtx_t** lockp)
 {
     unsigned bix = atom_val(name) & meta_name_tab_mask;
     struct meta_name_tab_entry* bucket = &meta_name_tab[bix];
-#ifdef ERTS_SMP
     *lockp = &meta_name_tab_rwlocks[bix % META_NAME_TAB_LOCK_CNT].lck;
-#endif
     return bucket;
 }    
 
@@ -333,8 +354,7 @@ struct meta_name_tab_entry* meta_name_tab_bucket(Eterm name,
 typedef enum {
     LCK_READ=1,     /* read only access */
     LCK_WRITE=2,    /* exclusive table write access */
-    LCK_WRITE_REC=3, /* record write access */
-    LCK_NONE=4
+    LCK_WRITE_REC=3 /* record write access */
 } db_lock_kind_t;
 
 extern DbTableMethod db_hash;
@@ -344,9 +364,6 @@ int user_requested_db_max_tabs;
 int erts_ets_realloc_always_moves;
 int erts_ets_always_compress;
 static int db_max_tabs;
-static Eterm ms_delete_all;
-static Eterm ms_delete_all_buff[8]; /* To compare with for deletion 
-				       of all objects */
 
 /* 
 ** Forward decls, static functions 
@@ -358,18 +375,19 @@ static void set_heir(Process* me, DbTable* tb, Eterm heir, UWord heir_data);
 static void free_heir_data(DbTable*);
 static SWord free_fixations_locked(Process* p, DbTable *tb);
 
+static void delete_all_objects_continue(Process* p, DbTable* tb);
 static SWord free_table_continue(Process *p, DbTable *tb, SWord reds);
 static void print_table(fmtfn_t to, void *to_arg, int show,  DbTable* tb);
-static BIF_RETTYPE ets_select_delete_1(BIF_ALIST_1);
+static BIF_RETTYPE ets_select_delete_trap_1(BIF_ALIST_1);
 static BIF_RETTYPE ets_select_count_1(BIF_ALIST_1);
 static BIF_RETTYPE ets_select_replace_1(BIF_ALIST_1);
 static BIF_RETTYPE ets_select_trap_1(BIF_ALIST_1);
 static BIF_RETTYPE ets_delete_trap(BIF_ALIST_1);
 static Eterm table_info(Process* p, DbTable* tb, Eterm What);
 
-static BIF_RETTYPE ets_select1(Process* p, Eterm arg1);
-static BIF_RETTYPE ets_select2(Process* p, Eterm arg1, Eterm arg2);
-static BIF_RETTYPE ets_select3(Process* p, Eterm arg1, Eterm arg2, Eterm arg3);
+static BIF_RETTYPE ets_select1(Process* p, int bif_ix, Eterm arg1);
+static BIF_RETTYPE ets_select2(Process* p, DbTable*, Eterm tid, Eterm ms);
+static BIF_RETTYPE ets_select3(Process* p, DbTable*, Eterm tid, Eterm ms, Sint chunk_size);
 
 
 /* 
@@ -390,16 +408,14 @@ free_dbtable(void *vtb)
 {
     DbTable *tb = (DbTable *) vtb;
 #ifdef HARDDEBUG
-	if (erts_smp_atomic_read_nob(&tb->common.memory_size) != sizeof(DbTable)) {
+	if (erts_atomic_read_nob(&tb->common.memory_size) != sizeof(DbTable)) {
 	    erts_fprintf(stderr, "ets: free_dbtable memory remain=%ld fix=%x\n",
-			 erts_smp_atomic_read_nob(&tb->common.memory_size)-sizeof(DbTable),
+			 erts_atomic_read_nob(&tb->common.memory_size)-sizeof(DbTable),
 			 tb->common.fixations);
 	}
 #endif
-#ifdef ERTS_SMP
-	erts_smp_rwmtx_destroy(&tb->common.rwlock);
-	erts_smp_mtx_destroy(&tb->common.fixlock);
-#endif
+	erts_rwmtx_destroy(&tb->common.rwlock);
+	erts_mtx_destroy(&tb->common.fixlock);
 	ASSERT(is_immed(tb->common.heir_data));
 
         if (tb->common.btid)
@@ -419,8 +435,8 @@ static void schedule_free_dbtable(DbTable* tb)
      *  	     Caller is *not* allowed to access the specialized part
      *  	     (hash or tree) of *tb after this function has returned.
      */
-    ASSERT(erts_smp_refc_read(&tb->common.refc, 0) == 0);
-    ASSERT(erts_smp_refc_read(&tb->common.fix_count, 0) == 0);
+    ASSERT(erts_refc_read(&tb->common.refc, 0) == 0);
+    ASSERT(erts_refc_read(&tb->common.fix_count, 0) == 0);
     erts_schedule_thr_prgr_later_cleanup_op(free_dbtable,
 					    (void *) tb,
 					    &tb->release.data,
@@ -434,8 +450,8 @@ save_sched_table(Process *c_p, DbTable *tb)
     DbTable *first;
 
     ASSERT(esdp);
-    esdp->ets_tables.count++;
-    erts_smp_refc_inc(&tb->common.refc, 1);
+    erts_atomic_inc_nob(&esdp->ets_tables.count);
+    erts_refc_inc(&tb->common.refc, 1);
 
     first = esdp->ets_tables.clist;
     if (!first) {
@@ -458,8 +474,8 @@ remove_sched_table(ErtsSchedulerData *esdp, DbTable *tb)
     ASSERT(erts_get_ref_numbers_thr_id(ERTS_MAGIC_BIN_REFN(tb->common.btid))
            == (Uint32) esdp->no);
 
-    ASSERT(esdp->ets_tables.count > 0);
-    esdp->ets_tables.count--;
+    ASSERT(erts_atomic_read_nob(&esdp->ets_tables.count) > 0);
+    erts_atomic_dec_nob(&esdp->ets_tables.count);
 
     eaydp = ERTS_SCHED_AUX_YIELD_DATA(esdp, ets_all);
     if (eaydp->ongoing) {
@@ -525,11 +541,11 @@ save_owned_table(Process *c_p, DbTable *tb)
 {
     DbTable *first;
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
 
     first = (DbTable*) erts_psd_get(c_p, ERTS_PSD_ETS_OWNED_TABLES);
 
-    erts_smp_refc_inc(&tb->common.refc, 1);
+    erts_refc_inc(&tb->common.refc, 1);
 
     if (!first) {
         tb->common.owned.next = tb->common.owned.prev = tb;
@@ -541,13 +557,13 @@ save_owned_table(Process *c_p, DbTable *tb)
         tb->common.owned.prev->common.owned.next = tb;
         first->common.owned.prev = tb;
     }
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
 }
 
 static ERTS_INLINE void
 delete_owned_table(Process *p, DbTable *tb)
 {
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
     if (tb->common.owned.next == tb) {
         DbTable* old;
         ASSERT(tb->common.owned.prev == tb);
@@ -570,38 +586,33 @@ delete_owned_table(Process *p, DbTable *tb)
         if (tb == first)
             erts_psd_set(p, ERTS_PSD_ETS_OWNED_TABLES, tb->common.owned.next);
     }
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 
     table_dec_refc(tb, 1);
 }
 
 static ERTS_INLINE void db_init_lock(DbTable* tb, int use_frequent_read_lock)
 {
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
     if (use_frequent_read_lock)
-	rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
+	rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
     if (erts_ets_rwmtx_spin_count >= 0)
 	rwmtx_opt.main_spincount = erts_ets_rwmtx_spin_count;
-#endif
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_init_opt(&tb->common.rwlock, &rwmtx_opt, "db_tab",
+    erts_rwmtx_init_opt(&tb->common.rwlock, &rwmtx_opt, "db_tab",
         tb->common.the_name, ERTS_LOCK_FLAGS_CATEGORY_DB);
-    erts_smp_mtx_init(&tb->common.fixlock, "db_tab_fix",
+    erts_mtx_init(&tb->common.fixlock, "db_tab_fix",
         tb->common.the_name, ERTS_LOCK_FLAGS_CATEGORY_DB);
     tb->common.is_thread_safe = !(tb->common.status & DB_FINE_LOCKED);
-#endif
 }
 
 static ERTS_INLINE void db_lock(DbTable* tb, db_lock_kind_t kind)
 {
-#ifdef ERTS_SMP
     if (tb->common.type & DB_FINE_LOCKED) {
 	if (kind == LCK_WRITE) {	   
-	    erts_smp_rwmtx_rwlock(&tb->common.rwlock);
+	    erts_rwmtx_rwlock(&tb->common.rwlock);
 	    tb->common.is_thread_safe = 1;
 	} else {	
-	    erts_smp_rwmtx_rlock(&tb->common.rwlock);
+	    erts_rwmtx_rlock(&tb->common.rwlock);
 	    ASSERT(!tb->common.is_thread_safe);
 	}
     }
@@ -610,14 +621,13 @@ static ERTS_INLINE void db_lock(DbTable* tb, db_lock_kind_t kind)
 	switch (kind) {
 	case LCK_WRITE:
 	case LCK_WRITE_REC:
-	    erts_smp_rwmtx_rwlock(&tb->common.rwlock);
+	    erts_rwmtx_rwlock(&tb->common.rwlock);
 	    break;
 	default:
-	    erts_smp_rwmtx_rlock(&tb->common.rwlock);
+	    erts_rwmtx_rlock(&tb->common.rwlock);
 	}
 	ASSERT(tb->common.is_thread_safe);
     }
-#endif
 }
 
 static ERTS_INLINE void db_unlock(DbTable* tb, db_lock_kind_t kind)
@@ -627,16 +637,15 @@ static ERTS_INLINE void db_unlock(DbTable* tb, db_lock_kind_t kind)
      * DbTable structure. That is, ONLY the SMP case is allowed
      * to follow the tb pointer!
      */
-#ifdef ERTS_SMP
     if (tb->common.type & DB_FINE_LOCKED) {
 	if (kind == LCK_WRITE) {
 	    ASSERT(tb->common.is_thread_safe);
 	    tb->common.is_thread_safe = 0;
-	    erts_smp_rwmtx_rwunlock(&tb->common.rwlock);
+	    erts_rwmtx_rwunlock(&tb->common.rwlock);
 	}
 	else {
 	    ASSERT(!tb->common.is_thread_safe);
-	    erts_smp_rwmtx_runlock(&tb->common.rwlock);
+	    erts_rwmtx_runlock(&tb->common.rwlock);
 	}
     }
     else {
@@ -644,13 +653,39 @@ static ERTS_INLINE void db_unlock(DbTable* tb, db_lock_kind_t kind)
 	switch (kind) {
 	case LCK_WRITE:
 	case LCK_WRITE_REC:
-	    erts_smp_rwmtx_rwunlock(&tb->common.rwlock);
+	    erts_rwmtx_rwunlock(&tb->common.rwlock);
 	    break;
 	default:
-	    erts_smp_rwmtx_runlock(&tb->common.rwlock);
+	    erts_rwmtx_runlock(&tb->common.rwlock);
 	}
     }
-#endif
+}
+
+static ERTS_INLINE int db_is_exclusive(DbTable* tb, db_lock_kind_t kind)
+{
+    return kind != LCK_READ && tb->common.is_thread_safe;
+}
+
+static DbTable* handle_lacking_permission(Process* p, DbTable* tb,
+                                          db_lock_kind_t kind,
+                                          Uint* freason_p)
+{
+    if (tb->common.status & DB_BUSY) {
+        if (!db_is_exclusive(tb, kind)) {
+            db_unlock(tb, kind);
+            db_lock(tb, LCK_WRITE);
+        }
+        delete_all_objects_continue(p, tb);
+        db_unlock(tb, LCK_WRITE);
+        tb = NULL;
+        *freason_p = TRAP;
+    }
+    else if (p->common.id != tb->common.owner) {
+        db_unlock(tb, kind);
+        tb = NULL;
+        *freason_p = BADARG;
+    }
+    return tb;
 }
 
 static ERTS_INLINE
@@ -658,10 +693,10 @@ DbTable* db_get_table_aux(Process *p,
 			  Eterm id,
 			  int what,
 			  db_lock_kind_t kind,
-			  int meta_already_locked)
+			  int meta_already_locked,
+                          Uint* freason_p)
 {
     DbTable *tb;
-    erts_smp_rwmtx_t *mtl = NULL;
 
     /*
      * IMPORTANT: Only scheduler threads are allowed
@@ -671,13 +706,13 @@ DbTable* db_get_table_aux(Process *p,
     ASSERT(erts_get_scheduler_data());
 
     if (is_atom(id)) {
+        erts_rwmtx_t *mtl;
 	struct meta_name_tab_entry* bucket = meta_name_tab_bucket(id,&mtl);
 	if (!meta_already_locked)
-	    erts_smp_rwmtx_rlock(mtl);
+	    erts_rwmtx_rlock(mtl);
 	else{
-	    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rlocked(mtl)
+	    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rlocked(mtl)
 			       || erts_lc_rwmtx_is_rwlocked(mtl));
-	    mtl = NULL;
 	}
         tb = NULL;
 	if (bucket->pu.tb != NULL) {
@@ -696,20 +731,29 @@ DbTable* db_get_table_aux(Process *p,
 		}
 	    }
 	}
+        if (!meta_already_locked)
+            erts_rwmtx_runlock(mtl);
     }
     else
         tb = tid2tab(id);
 
     if (tb) {
 	db_lock(tb, kind);
-	if ((tb->common.status & what) == 0
-            && p->common.id != tb->common.owner) {
-	    db_unlock(tb, kind);
-	    tb = NULL;
-	}
+#ifdef ETS_DBG_FORCE_TRAP
+        if (erts_atomic_read_nob(&tb->common.dbg_force_trap) &&
+            erts_atomic_add_read_nob(&tb->common.dbg_force_trap, 2) & 2) {
+            db_unlock(tb, kind);
+            tb = NULL;
+            *freason_p = TRAP;
+        }
+        else
+#endif
+	if (ERTS_UNLIKELY(!(tb->common.status & what)))
+            tb = handle_lacking_permission(p, tb, kind, freason_p);
     }
-    if (mtl)
-	erts_smp_rwmtx_runlock(mtl);
+    else
+        *freason_p = BADARG;
+
     return tb;
 }
 
@@ -717,20 +761,21 @@ static ERTS_INLINE
 DbTable* db_get_table(Process *p,
 		      Eterm id,
 		      int what,
-		      db_lock_kind_t kind)
+		      db_lock_kind_t kind,
+                      Uint* freason_p)
 {
-    return db_get_table_aux(p, id, what, kind, 0);
+    return db_get_table_aux(p, id, what, kind, 0, freason_p);
 }
 
 static int insert_named_tab(Eterm name_atom, DbTable* tb, int have_lock)
 {
     int ret = 0;
-    erts_smp_rwmtx_t* rwlock;
+    erts_rwmtx_t* rwlock;
     struct meta_name_tab_entry* new_entry;
     struct meta_name_tab_entry* bucket = meta_name_tab_bucket(name_atom,
 							      &rwlock);
     if (!have_lock)
-	erts_smp_rwmtx_rwlock(rwlock);
+	erts_rwmtx_rwlock(rwlock);
 
     if (bucket->pu.tb == NULL) { /* empty */
 	new_entry = bucket;
@@ -778,27 +823,25 @@ static int insert_named_tab(Eterm name_atom, DbTable* tb, int have_lock)
 
 done:
     if (!have_lock)
-	erts_smp_rwmtx_rwunlock(rwlock);
+	erts_rwmtx_rwunlock(rwlock);
     return ret;
 }
 
 static int remove_named_tab(DbTable *tb, int have_lock)
 {
     int ret = 0;
-    erts_smp_rwmtx_t* rwlock;
+    erts_rwmtx_t* rwlock;
     Eterm name_atom = tb->common.the_name;
     struct meta_name_tab_entry* bucket = meta_name_tab_bucket(name_atom,
 							      &rwlock);
     ASSERT(is_table_named(tb));
-#ifdef ERTS_SMP
-    if (!have_lock && erts_smp_rwmtx_tryrwlock(rwlock) == EBUSY) {
+    if (!have_lock && erts_rwmtx_tryrwlock(rwlock) == EBUSY) {
 	db_unlock(tb, LCK_WRITE);
-	erts_smp_rwmtx_rwlock(rwlock);
+	erts_rwmtx_rwlock(rwlock);
 	db_lock(tb, LCK_WRITE);
     }
-#endif
 
-    ERTS_SMP_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(rwlock));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(rwlock));
 
     if (bucket->pu.tb == NULL) {
 	goto done;
@@ -851,7 +894,7 @@ static int remove_named_tab(DbTable *tb, int have_lock)
 
 done:
     if (!have_lock)
-	erts_smp_rwmtx_rwunlock(rwlock);
+	erts_rwmtx_rwunlock(rwlock);
     return ret;
 }
 
@@ -860,11 +903,11 @@ done:
 */
 static ERTS_INLINE void local_fix_table(DbTable* tb)
 {
-    erts_smp_refc_inc(&tb->common.fix_count, 1);
+    erts_refc_inc(&tb->common.fix_count, 1);
 }	    
 static ERTS_INLINE void local_unfix_table(DbTable* tb)
 {	
-    if (erts_smp_refc_dectest(&tb->common.fix_count, 0) == 0) {
+    if (erts_refc_dectest(&tb->common.fix_count, 0) == 0) {
 	ASSERT(IS_HASH_TABLE(tb->common.status));
 	db_unfix_table_hash(&(tb->hash));
     }
@@ -887,9 +930,7 @@ BIF_RETTYPE ets_safe_fixtable_2(BIF_ALIST_2)
 #endif
     kind = (BIF_ARG_2 == am_true) ? LCK_READ : LCK_WRITE_REC; 
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, kind)) == NULL) {
-	    BIF_ERROR(BIF_P, BADARG);
-	}
+    DB_BIF_GET_TABLE(tb, DB_READ, kind, BIF_ets_safe_fixtable_2);
 
     if (BIF_ARG_2 == am_true) {
 	fix_table_locked(BIF_P, tb);
@@ -919,11 +960,7 @@ BIF_RETTYPE ets_first_1(BIF_ALIST_1)
 
     CHECK_TABLES();
 
-    tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ);
-
-    if (!tb) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_first_1);
 
     cret = tb->common.meth->db_first(BIF_P, tb, &ret);
 
@@ -946,11 +983,7 @@ BIF_RETTYPE ets_next_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ);
-
-    if (!tb) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_next_2);
 
     cret = tb->common.meth->db_next(BIF_P, tb, BIF_ARG_2, &ret);
 
@@ -973,11 +1006,7 @@ BIF_RETTYPE ets_last_1(BIF_ALIST_1)
 
     CHECK_TABLES();
 
-    tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ);
-
-    if (!tb) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_last_1);
 
     cret = tb->common.meth->db_last(BIF_P, tb, &ret);
 
@@ -1000,11 +1029,7 @@ BIF_RETTYPE ets_prev_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ);
-
-    if (!tb) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_prev_2);
 
     cret = tb->common.meth->db_prev(BIF_P,tb,BIF_ARG_2,&ret);
 
@@ -1022,21 +1047,15 @@ BIF_RETTYPE ets_prev_2(BIF_ALIST_2)
 BIF_RETTYPE ets_take_2(BIF_ALIST_2)
 {
     DbTable* tb;
-#ifdef DEBUG
     int cret;
-#endif
     Eterm ret;
     CHECK_TABLES();
 
-    tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC);
-    if (!tb) {
-        BIF_ERROR(BIF_P, BADARG);
-    }
-#ifdef DEBUG
-    cret =
-#endif
-        tb->common.meth->db_take(BIF_P, tb, BIF_ARG_2, &ret);
-    ASSERT(cret == DB_ERROR_NONE);
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_take_2);
+
+    cret = tb->common.meth->db_take(BIF_P, tb, BIF_ARG_2, &ret);
+
+    ASSERT(cret == DB_ERROR_NONE); (void)cret;
     db_unlock(tb, LCK_WRITE_REC);
     BIF_RET(ret);
 }
@@ -1054,9 +1073,8 @@ BIF_RETTYPE ets_update_element_3(BIF_ALIST_3)
     DeclareTmpHeap(cell,2,BIF_P);
     DbUpdateHandle handle;
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_update_element_3);
+
     UseTmpHeap(2,BIF_P);
     if (!(tb->common.status & (DB_SET | DB_ORDERED_SET))) {
 	goto bail_out;
@@ -1127,9 +1145,9 @@ bail_out:
 }
 
 static BIF_RETTYPE
-do_update_counter(Process *p, Eterm arg1, Eterm arg2, Eterm arg3, Eterm arg4)
+do_update_counter(Process *p, DbTable* tb,
+                  Eterm arg2, Eterm arg3, Eterm arg4)
 {
-    DbTable* tb;
     int cret = DB_ERROR_BADITEM;
     Eterm upop_list;
     int list_size;
@@ -1145,10 +1163,6 @@ do_update_counter(Process *p, Eterm arg1, Eterm arg2, Eterm arg3, Eterm arg4)
     Eterm* hstart;
     Eterm* hend;
 
-    if ((tb = db_get_table(p, arg1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-        BIF_ERROR(p, BADARG);
-    }
-
     UseTmpHeap(5, p);
 
     if (!(tb->common.status & (DB_SET | DB_ORDERED_SET))) {
@@ -1322,7 +1336,11 @@ bail_out:
 */
 BIF_RETTYPE ets_update_counter_3(BIF_ALIST_3)
 {
-    return do_update_counter(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, THE_NON_VALUE);
+    DbTable* tb;
+
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_update_counter_3);
+
+    return do_update_counter(BIF_P, tb, BIF_ARG_2, BIF_ARG_3, THE_NON_VALUE);
 }
 
 /*
@@ -1334,10 +1352,14 @@ BIF_RETTYPE ets_update_counter_3(BIF_ALIST_3)
 */
 BIF_RETTYPE ets_update_counter_4(BIF_ALIST_4)
 {
+    DbTable* tb;
+
     if (is_not_tuple(BIF_ARG_4)) {
         BIF_ERROR(BIF_P, BADARG);
     }
-    return do_update_counter(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, BIF_ARG_4);
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_update_counter_4);
+
+    return do_update_counter(BIF_P, tb, BIF_ARG_2, BIF_ARG_3, BIF_ARG_4);
 }
 
 
@@ -1358,9 +1380,8 @@ BIF_RETTYPE ets_insert_2(BIF_ALIST_2)
     kind = ((is_list(BIF_ARG_2) && CDR(list_val(BIF_ARG_2)) != NIL)
 	    ? LCK_WRITE : LCK_WRITE_REC);
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, kind)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, kind, BIF_ets_insert_2);
+
     if (BIF_ARG_2 == NIL) {
 	db_unlock(tb, kind);
 	BIF_RET(am_true);
@@ -1426,11 +1447,9 @@ BIF_RETTYPE ets_insert_new_2(BIF_ALIST_2)
 
 	    /* More than one object, use LCK_WRITE to keep atomicity */
 	    kind = LCK_WRITE;
-	    tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, kind);
-	    if (tb == NULL) {
-		BIF_ERROR(BIF_P, BADARG);
-	    }
-	    meth = tb->common.meth;
+            DB_BIF_GET_TABLE(tb, DB_WRITE, kind, BIF_ets_insert_new_2);
+
+            meth = tb->common.meth;
 	    for (lst = BIF_ARG_2; is_list(lst); lst = CDR(list_val(lst))) {
 		if (is_not_tuple(CAR(list_val(lst)))
 		    || (arityval(*tuple_val(CAR(list_val(lst))))
@@ -1465,9 +1484,8 @@ BIF_RETTYPE ets_insert_new_2(BIF_ALIST_2)
     /* Only one object (or NIL) 
     */
     kind = LCK_WRITE_REC;
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, kind)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, kind, BIF_ets_insert_new_2);
+
     if (BIF_ARG_2 == NIL) {
 	db_unlock(tb, kind);
 	BIF_RET(am_true);
@@ -1505,7 +1523,8 @@ BIF_RETTYPE ets_rename_2(BIF_ALIST_2)
     DbTable* tb;
     Eterm ret;
     Eterm old_name;
-    erts_smp_rwmtx_t *lck1, *lck2;
+    erts_rwmtx_t *lck1, *lck2;
+    Uint freason;
 
 #ifdef HARDDEBUG
     erts_fprintf(stderr,
@@ -1528,7 +1547,7 @@ BIF_RETTYPE ets_rename_2(BIF_ALIST_2)
 	if (lck1 == lck2)
 	    lck2 = NULL;
 	else if (lck1 > lck2) {
-	    erts_smp_rwmtx_t *tmp = lck1;
+	    erts_rwmtx_t *tmp = lck1;
 	    lck1 = lck2;
 	    lck2 = tmp;
 	}
@@ -1546,13 +1565,13 @@ BIF_RETTYPE ets_rename_2(BIF_ALIST_2)
         }
     }
 
-    erts_smp_rwmtx_rwlock(lck1);
+    erts_rwmtx_rwlock(lck1);
     if (lck2)
-	erts_smp_rwmtx_rwlock(lck2);
+	erts_rwmtx_rwlock(lck2);
 
-    tb = db_get_table_aux(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE, 1);
+    tb = db_get_table_aux(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE, 1, &freason);
     if (!tb)
-	goto badarg;
+	goto fail;
 
     if (is_table_named(tb)) {
         if (!insert_named_tab(BIF_ARG_2, tb, 1))
@@ -1568,17 +1587,22 @@ BIF_RETTYPE ets_rename_2(BIF_ALIST_2)
     tb->common.the_name = BIF_ARG_2;
 
     db_unlock(tb, LCK_WRITE);
-    erts_smp_rwmtx_rwunlock(lck1);
+    erts_rwmtx_rwunlock(lck1);
     if (lck2)
-	erts_smp_rwmtx_rwunlock(lck2);
+	erts_rwmtx_rwunlock(lck2);
     BIF_RET(ret);
- badarg:
+
+badarg:
+    freason = BADARG;
+
+fail:
     if (tb)
 	db_unlock(tb, LCK_WRITE);
-    erts_smp_rwmtx_rwunlock(lck1);
+    erts_rwmtx_rwunlock(lck1);
     if (lck2)
-	erts_smp_rwmtx_rwunlock(lck2);
-    BIF_ERROR(BIF_P, BADARG);    
+	erts_rwmtx_rwunlock(lck2);
+
+    return db_bif_fail(BIF_P, freason, BIF_ets_rename_2, NULL);
 }
 
 
@@ -1598,12 +1622,8 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
     Uint32 status;
     Sint keypos;
     int is_named, is_compressed;
-#ifdef ERTS_SMP
     int is_fine_locked, frequent_read;
-#endif
-#ifdef DEBUG
     int cret;
-#endif
     DbTableMethod* meth;
 
     if (is_not_atom(BIF_ARG_1)) {
@@ -1616,10 +1636,8 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
     status = DB_SET | DB_PROTECTED;
     keypos = 1;
     is_named = 0;
-#ifdef ERTS_SMP
     is_fine_locked = 0;
     frequent_read = 0;
-#endif
     heir = am_none;
     heir_data = (UWord) am_undefined;
     is_compressed = erts_ets_always_compress;
@@ -1647,30 +1665,18 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
 		    keypos = signed_val(tp[2]);
 		}		
 		else if (tp[1] == am_write_concurrency) {
-#ifdef ERTS_SMP
 		    if (tp[2] == am_true) {
 			is_fine_locked = 1;
 		    } else if (tp[2] == am_false) {
 			is_fine_locked = 0;
 		    } else break;
-#else
-		    if ((tp[2] != am_true) &&  (tp[2] != am_false)) {
-			break;
-		    }
-#endif
 		}
 		else if (tp[1] == am_read_concurrency) {
-#ifdef ERTS_SMP
 		    if (tp[2] == am_true) {
 			frequent_read = 1;
 		    } else if (tp[2] == am_false) {
 			frequent_read = 0;
 		    } else break;
-#else
-		    if ((tp[2] != am_true) &&  (tp[2] != am_false)) {
-			break;
-		    }
-#endif
 		    
 		}
 		else if (tp[1] == am_heir && tp[2] == am_none) {
@@ -1712,11 +1718,9 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
     }
     if (IS_HASH_TABLE(status)) {
 	meth = &db_hash;
-#ifdef ERTS_SMP
 	if (is_fine_locked && !(status & DB_PRIVATE)) {
 	    status |= DB_FINE_LOCKED;
 	}
-#endif
     }
     else if (IS_TREE_TABLE(status)) {
 	meth = &db_tree;
@@ -1725,10 +1729,8 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
 	BIF_ERROR(BIF_P, BADARG);
     }
 
-#ifdef ERTS_SMP
     if (frequent_read && !(status & DB_PRIVATE))
 	status |= DB_FREQ_READ;
-#endif
 
     /* we create table outside any table lock
      * and take the unusal cost of destroy table if it
@@ -1737,36 +1739,34 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
     {
         DbTable init_tb;
 
-	erts_smp_atomic_init_nob(&init_tb.common.memory_size, 0);
+	erts_atomic_init_nob(&init_tb.common.memory_size, 0);
 	tb = (DbTable*) erts_db_alloc(ERTS_ALC_T_DB_TABLE,
 				      &init_tb, sizeof(DbTable));
-	erts_smp_atomic_init_nob(&tb->common.memory_size,
-				 erts_smp_atomic_read_nob(&init_tb.common.memory_size));
+	erts_atomic_init_nob(&tb->common.memory_size,
+				 erts_atomic_read_nob(&init_tb.common.memory_size));
     }
 
     tb->common.meth = meth;
     tb->common.the_name = BIF_ARG_1;
     tb->common.status = status;    
-#ifdef ERTS_SMP
-    tb->common.type = status & ERTS_ETS_TABLE_TYPES;
+    tb->common.type = status;
     /* Note, 'type' is *read only* from now on... */
-#endif
-    erts_smp_refc_init(&tb->common.fix_count, 0);
+    erts_refc_init(&tb->common.fix_count, 0);
     db_init_lock(tb, status & (DB_FINE_LOCKED|DB_FREQ_READ));
     tb->common.keypos = keypos;
     tb->common.owner = BIF_P->common.id;
     set_heir(BIF_P, tb, heir, heir_data);
 
-    erts_smp_atomic_init_nob(&tb->common.nitems, 0);
+    erts_atomic_init_nob(&tb->common.nitems, 0);
 
     tb->common.fixing_procs = NULL;
     tb->common.compress = is_compressed;
-
-#ifdef DEBUG
-    cret = 
+#ifdef ETS_DBG_FORCE_TRAP
+    erts_atomic_init_nob(&tb->common.dbg_force_trap, erts_ets_dbg_force_trap);
 #endif
-	meth->db_create(BIF_P, tb);
-    ASSERT(cret == DB_ERROR_NONE);
+
+    cret = meth->db_create(BIF_P, tb);
+    ASSERT(cret == DB_ERROR_NONE); (void)cret;
 
     make_btid(tb);
 
@@ -1776,20 +1776,21 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
         ret = make_tid(BIF_P, tb);
 
     save_sched_table(BIF_P, tb);
+    save_owned_table(BIF_P, tb);
 
     if (is_named && !insert_named_tab(BIF_ARG_1, tb, 0)) {
         tid_clear(BIF_P, tb);
+        delete_owned_table(BIF_P, tb);
 
 	db_lock(tb,LCK_WRITE);
 	free_heir_data(tb);
-	tb->common.meth->db_free_table(tb);
+	tb->common.meth->db_free_empty_table(tb);
 	db_unlock(tb,LCK_WRITE);
         table_dec_refc(tb, 0);
 	BIF_ERROR(BIF_P, BADARG);
     }
     
     BIF_P->flags |= F_USING_DB; /* So we can remove tb if p dies */
-    save_owned_table(BIF_P, tb);
 
 #ifdef HARDDEBUG
     erts_fprintf(stderr,
@@ -1801,6 +1802,34 @@ BIF_RETTYPE ets_new_2(BIF_ALIST_2)
     BIF_RET(ret);
 }
 
+/*
+** Retrieves the tid() of a named ets table.
+*/
+BIF_RETTYPE ets_whereis_1(BIF_ALIST_1)
+{
+    DbTable* tb;
+    Eterm res;
+    Uint freason;
+
+    if (is_not_atom(BIF_ARG_1)) {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ, &freason)) == NULL) {
+        if (freason == BADARG)
+            BIF_RET(am_undefined);
+        else {
+            //ToDo: Could we avoid this
+            return db_bif_fail(BIF_P, freason, BIF_ets_whereis_1, NULL);
+        }
+    }
+
+    res = make_tid(BIF_P, tb);
+    db_unlock(tb, LCK_READ);
+
+    BIF_RET(res);
+}
+
 /* 
 ** The lookup BIF 
 */
@@ -1812,9 +1841,7 @@ BIF_RETTYPE ets_lookup_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_lookup_2);
 
     cret = tb->common.meth->db_get(BIF_P, tb, BIF_ARG_2, &ret);
 
@@ -1842,9 +1869,7 @@ BIF_RETTYPE ets_member_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_member_2);
 
     cret = tb->common.meth->db_member(tb, BIF_ARG_2, &ret);
 
@@ -1875,9 +1900,7 @@ BIF_RETTYPE ets_lookup_element_3(BIF_ALIST_3)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_lookup_element_3);
 
     if (is_not_small(BIF_ARG_3) || ((index = signed_val(BIF_ARG_3)) < 1)) {
 	db_unlock(tb, LCK_READ);
@@ -1915,9 +1938,7 @@ BIF_RETTYPE ets_delete_1(BIF_ALIST_1)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE, BIF_ets_delete_1);
     
     /*
      * Clear all access bits to prevent any ets operation to access the
@@ -1940,7 +1961,7 @@ BIF_RETTYPE ets_delete_1(BIF_ALIST_1)
          * Process 'rp' might be exiting, but our table lock prevents it
          * from terminating as it cannot complete erts_db_process_exiting().
          */
-        ASSERT(!(ERTS_PSFLG_FREE & erts_smp_atomic32_read_nob(&rp->state)));
+        ASSERT(!(ERTS_PSFLG_FREE & erts_atomic32_read_nob(&rp->state)));
 
         delete_owned_table(rp, tb);
         BIF_P->flags |= F_USING_DB;
@@ -1959,7 +1980,8 @@ BIF_RETTYPE ets_delete_1(BIF_ALIST_1)
     tid_clear(BIF_P, tb);
     db_unlock(tb, LCK_WRITE);
 
-    if (free_table_continue(BIF_P, tb, reds) < 0) {
+    reds = free_table_continue(BIF_P, tb, reds);
+    if (reds < 0) {
 	/*
 	 * Package the DbTable* pointer into a bignum so that it can be safely
 	 * passed through a trap. We used to pass the DbTable* pointer directly
@@ -1988,6 +2010,7 @@ BIF_RETTYPE ets_give_away_3(BIF_ALIST_3)
     Eterm to_pid = BIF_ARG_2;
     Eterm from_pid;
     DbTable* tb = NULL;
+    Uint freason;
 
     if (!is_internal_pid(to_pid)) {
 	goto badarg;
@@ -1997,10 +2020,11 @@ BIF_RETTYPE ets_give_away_3(BIF_ALIST_3)
 	goto badarg;
     }
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE)) == NULL
-	|| tb->common.owner != BIF_P->common.id) {
-	goto badarg;
-    }
+    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE, &freason)) == NULL)
+        goto fail;
+    if (tb->common.owner != BIF_P->common.id)
+        goto badarg;
+
     from_pid = tb->common.owner;
     if (to_pid == from_pid) {
 	goto badarg;  /* or should we be idempotent? return false maybe */
@@ -2014,14 +2038,17 @@ BIF_RETTYPE ets_give_away_3(BIF_ALIST_3)
     db_unlock(tb,LCK_WRITE);
     send_ets_transfer_message(BIF_P, to_proc, &to_locks,
                               tb, BIF_ARG_3);
-    erts_smp_proc_unlock(to_proc, to_locks);
+    erts_proc_unlock(to_proc, to_locks);
     UnUseTmpHeap(5,BIF_P);
     BIF_RET(am_true);
 
 badarg:
-    if (to_proc != NULL && to_proc != BIF_P) erts_smp_proc_unlock(to_proc, to_locks);
+    freason = BADARG;
+fail:
+    if (to_proc != NULL && to_proc != BIF_P) erts_proc_unlock(to_proc, to_locks);
     if (tb != NULL) db_unlock(tb, LCK_WRITE);
-    BIF_ERROR(BIF_P, BADARG);
+
+    return db_bif_fail(BIF_P, freason, BIF_ets_give_away_3, NULL);
 }
 
 BIF_RETTYPE ets_setopts_2(BIF_ALIST_2)
@@ -2072,11 +2099,13 @@ BIF_RETTYPE ets_setopts_2(BIF_ALIST_2)
 	}
     }
 
-    if (tail != NIL
-	|| (tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE)) == NULL
-	|| tb->common.owner != BIF_P->common.id) {
+    if (tail != NIL)
+        goto badarg;
+
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE, BIF_ets_setopts_2);
+
+    if (tb->common.owner != BIF_P->common.id)
 	goto badarg;
-    }
 
     if (heir_data != THE_NON_VALUE) {
 	free_heir_data(tb);
@@ -2100,23 +2129,84 @@ badarg:
 }
 
 /* 
-** BIF to erase a whole table and release all memory it holds 
-*/
-BIF_RETTYPE ets_delete_all_objects_1(BIF_ALIST_1)
+ * Common for delete_all_objects and select_delete(DeleteAll).
+ */
+BIF_RETTYPE ets_internal_delete_all_2(BIF_ALIST_2)
 {
+    SWord initial_reds = ERTS_BIF_REDS_LEFT(BIF_P);
+    SWord reds = initial_reds;
+    Eterm nitems;
     DbTable* tb;
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE, BIF_ets_internal_delete_all_2);
+
+    if (BIF_ARG_2 == am_undefined) {
+        nitems = erts_make_integer(erts_atomic_read_nob(&tb->common.nitems),
+                                   BIF_P);
 
-    tb->common.meth->db_delete_all_objects(BIF_P, tb);
+        reds = tb->common.meth->db_delete_all_objects(BIF_P, tb, reds);
+
+        ASSERT(!(tb->common.status & DB_BUSY));
+
+        if (reds < 0) {
+            /*
+             * Oboy, need to trap AND need to be atomic.
+             * Solved by cooperative trapping where every process trying to
+             * access this table (including this process) will "fail" to lookup
+             * the table and instead pitch in deleting objects
+             * (in delete_all_objects_continue) and then trap to self.
+             */
+            ASSERT((tb->common.status & (DB_PRIVATE|DB_PROTECTED|DB_PUBLIC))
+                   ==
+                   (tb->common.type & (DB_PRIVATE|DB_PROTECTED|DB_PUBLIC)));
+            tb->common.status &= ~(DB_PRIVATE|DB_PROTECTED|DB_PUBLIC);
+            tb->common.status |= DB_BUSY;
+            db_unlock(tb, LCK_WRITE);
+            BUMP_ALL_REDS(BIF_P);
+            BIF_TRAP2(bif_export[BIF_ets_internal_delete_all_2], BIF_P,
+                      BIF_ARG_1, nitems);
+        }
+        else {
+            /* Done, no trapping needed */
+            BUMP_REDS(BIF_P, (initial_reds - reds));
+        }
+
+    }
+    else {
+        /*
+         * The table lookup succeeded and second argument is nitems
+         * and not 'undefined', which means we have trapped at least once
+         * and are now done.
+         */
+        nitems = BIF_ARG_2;
+    }
 
     db_unlock(tb, LCK_WRITE);
+    BIF_RET(nitems);
+}
 
-    BIF_RET(am_true);
+static void delete_all_objects_continue(Process* p, DbTable* tb)
+{
+    SWord initial_reds = ERTS_BIF_REDS_LEFT(p);
+    SWord reds = initial_reds;
+
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&tb->common.rwlock));
+
+    if ((tb->common.status & (DB_DELETE|DB_BUSY)) != DB_BUSY)
+        return;
+
+    reds = tb->common.meth->db_delete_all_objects(p, tb, reds);
+    
+    if (reds < 0) {
+        BUMP_ALL_REDS(p);
+    }
+    else {
+        tb->common.status |= tb->common.type & (DB_PRIVATE|DB_PROTECTED|DB_PUBLIC);
+        tb->common.status &= ~DB_BUSY;
+        BUMP_REDS(p, (initial_reds - reds));
+    }
 }
 
 /* 
@@ -2132,9 +2222,7 @@ BIF_RETTYPE ets_delete_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_delete_2);
 
     cret = tb->common.meth->db_erase(tb,BIF_ARG_2,&ret);
 
@@ -2161,9 +2249,8 @@ BIF_RETTYPE ets_delete_object_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_delete_object_2);
+
     if (is_not_tuple(BIF_ARG_2) || 
 	(arityval(*tuple_val(BIF_ARG_2)) < tb->common.keypos)) {
 	db_unlock(tb, LCK_WRITE_REC);
@@ -2186,7 +2273,7 @@ BIF_RETTYPE ets_delete_object_2(BIF_ALIST_2)
 /*
 ** This is for trapping, cannot be called directly.
 */
-static BIF_RETTYPE ets_select_delete_1(BIF_ALIST_1)
+static BIF_RETTYPE ets_select_delete_trap_1(BIF_ALIST_1)
 {
     Process *p = BIF_P;
     Eterm a1 = BIF_ARG_1;
@@ -2196,15 +2283,14 @@ static BIF_RETTYPE ets_select_delete_1(BIF_ALIST_1)
     Eterm ret;
     Eterm *tptr;
     db_lock_kind_t kind = LCK_WRITE_REC;
-    
+
     CHECK_TABLES();
     ASSERT(is_tuple(a1));
     tptr = tuple_val(a1);
     ASSERT(arityval(*tptr) >= 1);
     
-    if ((tb = db_get_table(p, tptr[1], DB_WRITE, kind)) == NULL) {
-	BIF_ERROR(p,BADARG);
-    }
+    DB_TRAP_GET_TABLE(tb, tptr[1], DB_WRITE, kind,
+                      &ets_select_delete_continue_exp);
 
     cret = tb->common.meth->db_select_delete_continue(p,tb,a1,&ret);
 
@@ -2228,7 +2314,10 @@ static BIF_RETTYPE ets_select_delete_1(BIF_ALIST_1)
 }
     
 
-BIF_RETTYPE ets_select_delete_2(BIF_ALIST_2)
+/*
+ * ets:select_delete/2 without special case for "delete-all".
+ */
+BIF_RETTYPE ets_internal_select_delete_2(BIF_ALIST_2)
 {
     BIF_RETTYPE result;
     DbTable* tb;
@@ -2238,20 +2327,8 @@ BIF_RETTYPE ets_select_delete_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if(eq(BIF_ARG_2, ms_delete_all)) {
-	int nitems;
-	if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE)) == NULL) {
-	    BIF_ERROR(BIF_P, BADARG);
-	}
-	nitems = erts_smp_atomic_read_nob(&tb->common.nitems);
-	tb->common.meth->db_delete_all_objects(BIF_P, tb);
-	db_unlock(tb, LCK_WRITE);
-	BIF_RET(erts_make_integer(nitems,BIF_P));
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_internal_select_delete_2);
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
     safety = ITERATION_SAFETY(BIF_P,tb);
     if (safety == ITER_UNSAFE) {
 	local_fix_table(tb);
@@ -2294,7 +2371,7 @@ BIF_RETTYPE ets_select_delete_2(BIF_ALIST_2)
  */
 
 struct ErtsEtsAllReq_ {
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
     Process *proc;
     ErtsOIRefStorage ref;
     ErtsEtsAllReqList list[1]; /* one per scheduler */
@@ -2368,7 +2445,7 @@ ets_all_reply(ErtsSchedulerData *esdp, ErtsEtsAllReq **reqpp,
         ASSERT(!*tablepp);
 
         /* Max heap size needed... */
-        sz = esdp->ets_tables.count;
+        sz = erts_atomic_read_nob(&esdp->ets_tables.count);
         sz *= ERTS_MAGIC_REF_THING_SIZE + 2;
         sz += 3 + ERTS_REF_THING_SIZE;
         hfragp = new_message_buffer(sz);
@@ -2427,7 +2504,7 @@ ets_all_reply(ErtsSchedulerData *esdp, ErtsEtsAllReq **reqpp,
 
     erts_proc_dec_refc(reqp->proc);
 
-    if (erts_smp_atomic32_dec_read_nob(&reqp->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&reqp->refc) == 0)
         erts_free(ERTS_ALC_T_ETS_ALL_REQ, reqp);
 
     *reqpp = NULL;
@@ -2452,7 +2529,8 @@ erts_handle_yielded_ets_all_request(ErtsSchedulerData *esdp,
             if (!eaydp->queue)
                 return 0; /* All work completed! */
 
-            if (yc < ERTS_ETS_ALL_TB_YCNT_START && yc > esdp->ets_tables.count)
+            if (yc < ERTS_ETS_ALL_TB_YCNT_START &&
+                yc > erts_atomic_read_nob(&esdp->ets_tables.count))
                 return 1; /* Yield! */
 
             eaydp->ongoing = ongoing = eaydp->queue;
@@ -2515,25 +2593,22 @@ BIF_RETTYPE ets_internal_request_all_0(BIF_ALIST_0)
     Eterm ref = erts_make_ref(BIF_P);
     ErtsEtsAllReq *req = erts_alloc(ERTS_ALC_T_ETS_ALL_REQ,
                                     ERTS_ETS_ALL_REQ_SIZE);
-    erts_smp_atomic32_init_nob(&req->refc,
+    erts_atomic32_init_nob(&req->refc,
 			       (erts_aint32_t) erts_no_schedulers);
     erts_oiref_storage_save(&req->ref, ref);
     req->proc = BIF_P;
     erts_proc_add_refc(BIF_P, (Sint) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
                                           handle_ets_all_request,
 					  (void *) req);
-#endif
 
     handle_ets_all_request((void *) req);
     BIF_RET(ref);
 }
 
-
 /*
 ** db_slot(Db, Slot) -> [Items].
 */
@@ -2545,9 +2620,8 @@ BIF_RETTYPE ets_slot_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_slot_2);
+
     /* The slot number is checked in table specific code. */
     cret = tb->common.meth->db_slot(BIF_P, tb, BIF_ARG_2, &ret);
     db_unlock(tb, LCK_READ);
@@ -2567,41 +2641,53 @@ BIF_RETTYPE ets_slot_2(BIF_ALIST_2)
 
 BIF_RETTYPE ets_match_1(BIF_ALIST_1)
 {
-    return ets_select1(BIF_P, BIF_ARG_1);
+    return ets_select1(BIF_P, BIF_ets_match_1, BIF_ARG_1);
 }
 
 BIF_RETTYPE ets_match_2(BIF_ALIST_2)
 {
+    DbTable* tb;
     Eterm ms;
     DeclareTmpHeap(buff,8,BIF_P);
     Eterm *hp = buff;
     Eterm res;
 
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_match_2);
+
     UseTmpHeap(8,BIF_P);
     ms = CONS(hp, am_DollarDollar, NIL);
     hp += 2;
     ms = TUPLE3(hp, BIF_ARG_2, NIL, ms); 
     hp += 4;
     ms = CONS(hp, ms, NIL);
-    res = ets_select2(BIF_P, BIF_ARG_1, ms);
+    res = ets_select2(BIF_P, tb, BIF_ARG_1, ms);
     UnUseTmpHeap(8,BIF_P);
     return res;
 }
 
 BIF_RETTYPE ets_match_3(BIF_ALIST_3)
 {
+    DbTable* tb;
     Eterm ms;
+    Sint chunk_size;
     DeclareTmpHeap(buff,8,BIF_P);
     Eterm *hp = buff;
     Eterm res;
 
+    /* Chunk size strictly greater than 0 */
+    if (is_not_small(BIF_ARG_3) || (chunk_size = signed_val(BIF_ARG_3)) <= 0) {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_match_3);
+
     UseTmpHeap(8,BIF_P);
     ms = CONS(hp, am_DollarDollar, NIL);
     hp += 2;
     ms = TUPLE3(hp, BIF_ARG_2, NIL, ms); 
     hp += 4;
     ms = CONS(hp, ms, NIL);
-    res = ets_select3(BIF_P, BIF_ARG_1, ms, BIF_ARG_3);
+    res = ets_select3(BIF_P, tb, BIF_ARG_1, ms, chunk_size);
     UnUseTmpHeap(8,BIF_P);
     return res;
 }
@@ -2609,34 +2695,35 @@ BIF_RETTYPE ets_match_3(BIF_ALIST_3)
 
 BIF_RETTYPE ets_select_3(BIF_ALIST_3)
 {
-    return ets_select3(BIF_P, BIF_ARG_1, BIF_ARG_2, BIF_ARG_3);
+    DbTable* tb;
+    Sint chunk_size;
+
+    /* Chunk size strictly greater than 0 */
+    if (is_not_small(BIF_ARG_3) || (chunk_size = signed_val(BIF_ARG_3)) <= 0) {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_select_3);
+
+    return ets_select3(BIF_P, tb, BIF_ARG_1, BIF_ARG_2, chunk_size);
 }
 
 static BIF_RETTYPE
-ets_select3(Process* p, Eterm arg1, Eterm arg2, Eterm arg3)
+ets_select3(Process* p, DbTable* tb, Eterm tid, Eterm ms, Sint chunk_size)
 {
     BIF_RETTYPE result;
-    DbTable* tb;
     int cret;
     Eterm ret;
-    Sint chunk_size;
     enum DbIterSafety safety;
 
     CHECK_TABLES();
 
-    /* Chunk size strictly greater than 0 */
-    if (is_not_small(arg3) || (chunk_size = signed_val(arg3)) <= 0) {
-	BIF_ERROR(p, BADARG);
-    }
-    if ((tb = db_get_table(p, arg1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(p, BADARG);
-    }
     safety = ITERATION_SAFETY(p,tb);
     if (safety == ITER_UNSAFE) {
 	local_fix_table(tb);
     }
-    cret = tb->common.meth->db_select_chunk(p, tb, arg1,
-					    arg2, chunk_size,
+    cret = tb->common.meth->db_select_chunk(p, tb, tid,
+					    ms, chunk_size,
 					    0 /* not reversed */,
 					    &ret);
     if (DID_TRAP(p,ret) && safety != ITER_SAFE) {
@@ -2682,9 +2769,8 @@ static BIF_RETTYPE ets_select_trap_1(BIF_ALIST_1)
     tptr = tuple_val(a1);
     ASSERT(arityval(*tptr) >= 1);
 
-    if ((tb = db_get_table(p, tptr[1], DB_READ, kind)) == NULL) {
-	BIF_ERROR(p, BADARG);
-    }
+    DB_TRAP_GET_TABLE(tb, tptr[1], DB_READ, kind,
+                      &ets_select_continue_exp);
 
     cret = tb->common.meth->db_select_continue(p, tb, a1,
 					       &ret);
@@ -2714,10 +2800,10 @@ static BIF_RETTYPE ets_select_trap_1(BIF_ALIST_1)
 
 BIF_RETTYPE ets_select_1(BIF_ALIST_1)
 {
-    return ets_select1(BIF_P, BIF_ARG_1);
+    return ets_select1(BIF_P, BIF_ets_select_1, BIF_ARG_1);
 }
 
-static BIF_RETTYPE ets_select1(Process *p, Eterm arg1)
+static BIF_RETTYPE ets_select1(Process *p, int bif_ix, Eterm arg1)
 {
     BIF_RETTYPE result;
     DbTable* tb;
@@ -2739,10 +2825,10 @@ static BIF_RETTYPE ets_select1(Process *p, Eterm arg1)
 	BIF_ERROR(p, BADARG);
     }
     tptr = tuple_val(arg1);
-    if (arityval(*tptr) < 1 ||
-	(tb = db_get_table(p, tptr[1], DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(p, BADARG);
-    }
+    if (arityval(*tptr) < 1)
+        BIF_ERROR(p, BADARG);
+
+    DB_GET_TABLE(tb, tptr[1], DB_READ, LCK_READ, bif_ix, NULL, p);
 
     safety = ITERATION_SAFETY(p,tb);
     if (safety == ITER_UNSAFE) {
@@ -2778,33 +2864,27 @@ static BIF_RETTYPE ets_select1(Process *p, Eterm arg1)
 
 BIF_RETTYPE ets_select_2(BIF_ALIST_2)
 {
-    return ets_select2(BIF_P, BIF_ARG_1, BIF_ARG_2);
+    DbTable* tb;
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_select_2);
+    return ets_select2(BIF_P, tb, BIF_ARG_1, BIF_ARG_2);
 }
 
 static BIF_RETTYPE
-ets_select2(Process* p, Eterm arg1, Eterm arg2)
+ets_select2(Process* p, DbTable* tb, Eterm tid, Eterm ms)
 {
     BIF_RETTYPE result;
-    DbTable* tb;
     int cret;
     enum DbIterSafety safety;
     Eterm ret;
 
     CHECK_TABLES();
 
-    /*
-     * Make sure that the table exists.
-     */
-
-    if ((tb = db_get_table(p, arg1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(p, BADARG);
-    }
     safety = ITERATION_SAFETY(p,tb);
     if (safety == ITER_UNSAFE) {
 	local_fix_table(tb);
     }
 
-    cret = tb->common.meth->db_select(p, tb, arg1, arg2, 0, &ret);
+    cret = tb->common.meth->db_select(p, tb, tid, ms, 0, &ret);
 
     if (DID_TRAP(p,ret) && safety != ITER_SAFE) {
 	fix_table_locked(p, tb);
@@ -2847,9 +2927,9 @@ static BIF_RETTYPE ets_select_count_1(BIF_ALIST_1)
 
     tptr = tuple_val(a1);
     ASSERT(arityval(*tptr) >= 1);
-    if ((tb = db_get_table(p, tptr[1], DB_READ, kind)) == NULL) {
-	BIF_ERROR(p, BADARG);
-    }
+
+    DB_TRAP_GET_TABLE(tb, tptr[1], DB_READ, kind,
+                      &ets_select_count_continue_exp);
 
     cret = tb->common.meth->db_select_count_continue(p, tb, a1, &ret);
 
@@ -2884,13 +2964,9 @@ BIF_RETTYPE ets_select_count_2(BIF_ALIST_2)
     Eterm ret;
 
     CHECK_TABLES();
-    /*
-     * Make sure that the table exists.
-     */
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_select_count_2);
+
     safety = ITERATION_SAFETY(BIF_P,tb);
     if (safety == ITER_UNSAFE) {
 	local_fix_table(tb);
@@ -2940,9 +3016,8 @@ static BIF_RETTYPE ets_select_replace_1(BIF_ALIST_1)
     tptr = tuple_val(a1);
     ASSERT(arityval(*tptr) >= 1);
 
-    if ((tb = db_get_table(p, tptr[1], DB_WRITE, kind)) == NULL) {
-        BIF_ERROR(p,BADARG);
-    }
+    DB_TRAP_GET_TABLE(tb, tptr[1], DB_WRITE, kind,
+                      &ets_select_replace_continue_exp);
 
     cret = tb->common.meth->db_select_replace_continue(p,tb,a1,&ret);
 
@@ -2976,9 +3051,7 @@ BIF_RETTYPE ets_select_replace_2(BIF_ALIST_2)
 
     CHECK_TABLES();
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_WRITE, LCK_WRITE_REC)) == NULL) {
-        BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_WRITE, LCK_WRITE_REC, BIF_ets_select_replace_2);
 
     if (tb->common.status & DB_BAG) {
         /* Bag implementation presented both semantic consistency
@@ -3029,13 +3102,8 @@ BIF_RETTYPE ets_select_reverse_3(BIF_ALIST_3)
     Sint chunk_size;
 
     CHECK_TABLES();
-    /*
-     * Make sure that the table exists.
-     */
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_select_reverse_3);
     
     /* Chunk size strictly greater than 0 */
     if (is_not_small(BIF_ARG_3) || (chunk_size = signed_val(BIF_ARG_3)) <= 0) {
@@ -3073,7 +3141,7 @@ BIF_RETTYPE ets_select_reverse_3(BIF_ALIST_3)
 
 BIF_RETTYPE ets_select_reverse_1(BIF_ALIST_1)
 {
-    return ets_select1(BIF_P, BIF_ARG_1);
+    return ets_select1(BIF_P, BIF_ets_select_reverse_1, BIF_ARG_1);
 }
 
 BIF_RETTYPE ets_select_reverse_2(BIF_ALIST_2)
@@ -3085,13 +3153,9 @@ BIF_RETTYPE ets_select_reverse_2(BIF_ALIST_2)
     Eterm ret;
 
     CHECK_TABLES();
-    /*
-     * Make sure that the table exists.
-     */
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_READ, LCK_READ)) == NULL) {
-	BIF_ERROR(BIF_P, BADARG);
-    }
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_select_reverse_2);
+
     safety = ITERATION_SAFETY(BIF_P,tb);
     if (safety == ITER_UNSAFE) {
 	local_fix_table(tb);
@@ -3123,45 +3187,63 @@ BIF_RETTYPE ets_select_reverse_2(BIF_ALIST_2)
 
 
 /* 
-** ets:match_object(Continuation), ets:match_object(Table, Pattern), ets:match_object(Table,Pattern,ChunkSize) 
+** ets:match_object(Continuation)
 */
 BIF_RETTYPE ets_match_object_1(BIF_ALIST_1)
 {
-    return ets_select1(BIF_P, BIF_ARG_1);
+    return ets_select1(BIF_P, BIF_ets_match_object_1, BIF_ARG_1);
 }
 
+/*
+** ets:match_object(Table, Pattern)
+*/
 BIF_RETTYPE ets_match_object_2(BIF_ALIST_2)
 {
+    DbTable* tb;
     Eterm ms;
     DeclareTmpHeap(buff,8,BIF_P);
     Eterm *hp = buff;
     Eterm res;
 
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_match_object_2);
+
     UseTmpHeap(8,BIF_P);
     ms = CONS(hp, am_DollarUnderscore, NIL);
     hp += 2;
     ms = TUPLE3(hp, BIF_ARG_2, NIL, ms); 
     hp += 4;
     ms = CONS(hp, ms, NIL);
-    res = ets_select2(BIF_P, BIF_ARG_1, ms);
+    res = ets_select2(BIF_P, tb, BIF_ARG_1, ms);
     UnUseTmpHeap(8,BIF_P);
     return res;
 }
 
+/*
+** ets:match_object(Table,Pattern,ChunkSize)
+*/
 BIF_RETTYPE ets_match_object_3(BIF_ALIST_3)
 {
+    DbTable* tb;
+    Sint chunk_size;
     Eterm ms;
     DeclareTmpHeap(buff,8,BIF_P);
     Eterm *hp = buff;
     Eterm res;
 
+    /* Chunk size strictly greater than 0 */
+    if (is_not_small(BIF_ARG_3) || (chunk_size = signed_val(BIF_ARG_3)) <= 0) {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    DB_BIF_GET_TABLE(tb, DB_READ, LCK_READ, BIF_ets_match_object_3);
+
     UseTmpHeap(8,BIF_P);
     ms = CONS(hp, am_DollarUnderscore, NIL);
     hp += 2;
     ms = TUPLE3(hp, BIF_ARG_2, NIL, ms); 
     hp += 4;
     ms = CONS(hp, ms, NIL);
-    res = ets_select3(BIF_P, BIF_ARG_1, ms, BIF_ARG_3);
+    res = ets_select3(BIF_P, tb, BIF_ARG_1, ms, chunk_size);
     UnUseTmpHeap(8,BIF_P);
     return res;
 }
@@ -3175,22 +3257,24 @@ BIF_RETTYPE ets_info_1(BIF_ALIST_1)
     static Eterm fields[] = {am_protection, am_keypos, am_type, am_named_table,
                              am_node, am_size, am_name, am_heir, am_owner, am_memory, am_compressed,
                              am_write_concurrency,
-                             am_read_concurrency};
+                             am_read_concurrency,
+                             am_id};
     Eterm results[sizeof(fields)/sizeof(Eterm)];
     DbTable* tb;
     Eterm res;
     int i;
     Eterm* hp;
+    Uint freason;
     /*Process* rp = NULL;*/
     /* If/when we implement lockless private tables:
     Eterm owner;
     */
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ)) == NULL) {
-	if (is_atom(BIF_ARG_1) || is_ref(BIF_ARG_1)) {
+    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ, &freason)) == NULL) {
+	if (freason == BADARG && (is_atom(BIF_ARG_1) || is_ref(BIF_ARG_1)))
 	    BIF_RET(am_undefined);
-	}
-	BIF_ERROR(BIF_P, BADARG);
+	else
+            return db_bif_fail(BIF_P, freason, BIF_ets_info_1, NULL);
     }
 
     /* If/when we implement lockless private tables:
@@ -3211,7 +3295,7 @@ BIF_RETTYPE ets_info_1(BIF_ALIST_1)
 	if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ)) == NULL
 	    || tb->common.owner != owner) {
 	    if (BIF_P != rp)
-		erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+		erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
 	    if (is_atom(BIF_ARG_1) || is_small(BIF_ARG_1)) {
 		BIF_RET(am_undefined);
 	    }
@@ -3225,7 +3309,7 @@ BIF_RETTYPE ets_info_1(BIF_ALIST_1)
     db_unlock(tb, LCK_READ);
 
     /*if (rp != NULL && rp != BIF_P)
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);*/
+	erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);*/
 
     hp = HAlloc(BIF_P, 5*sizeof(fields)/sizeof(Eterm));
     res = NIL;
@@ -3247,12 +3331,13 @@ BIF_RETTYPE ets_info_2(BIF_ALIST_2)
 {
     DbTable* tb;
     Eterm ret = THE_NON_VALUE;
+    Uint freason;
 
-    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ)) == NULL) {
-	if (is_atom(BIF_ARG_1) || is_ref(BIF_ARG_1)) {
+    if ((tb = db_get_table(BIF_P, BIF_ARG_1, DB_INFO, LCK_READ, &freason)) == NULL) {
+	if (freason == BADARG && (is_atom(BIF_ARG_1) || is_ref(BIF_ARG_1)))
 	    BIF_RET(am_undefined);
-	}
-	BIF_ERROR(BIF_P, BADARG);
+        else
+            return db_bif_fail(BIF_P, freason, BIF_ets_info_2, NULL);
     }
     ret = table_info(BIF_P, tb, BIF_ARG_2);
     db_unlock(tb, LCK_READ);
@@ -3340,15 +3425,13 @@ int erts_ets_rwmtx_spin_count = -1;
 void init_db(ErtsDbSpinCount db_spin_count)
 {
     int i;
-    Eterm *hp;
     unsigned bits;
     size_t size;
 
-#ifdef ERTS_SMP
     int max_spin_count = (1 << 15) - 1; /* internal limit */
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
     switch (db_spin_count) {
     case ERTS_DB_SPNCNT_NONE:
@@ -3388,13 +3471,12 @@ void init_db(ErtsDbSpinCount db_spin_count)
 	rwmtx_opt.main_spincount = erts_ets_rwmtx_spin_count;
 
     for (i=0; i<META_NAME_TAB_LOCK_CNT; i++) {
-        erts_smp_rwmtx_init_opt(&meta_name_tab_rwlocks[i].lck, &rwmtx_opt,
+        erts_rwmtx_init_opt(&meta_name_tab_rwlocks[i].lck, &rwmtx_opt,
             "meta_name_tab", make_small(i),
             ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DB);
     }
-#endif
 
-    erts_smp_atomic_init_nob(&erts_ets_misc_mem_size, 0);
+    erts_atomic_init_nob(&erts_ets_misc_mem_size, 0);
     db_initialize_util();
 
     if (user_requested_db_max_tabs < DB_DEF_MAX_TABS)
@@ -3408,7 +3490,11 @@ void init_db(ErtsDbSpinCount db_spin_count)
 		 db_max_tabs, ((Uint)1)<<SMALL_BITS);
     }
 
-    meta_name_tab_mask = (((Uint) 1)<<(bits-1)) - 1; /* At least half the size of main tab */
+    /*
+     * We don't have ony hard limit for number of tables anymore,                                                                            .
+     * but we use 'db_max_tabs' to determine size of name hash table.
+     */
+    meta_name_tab_mask = (((Uint) 1)<<bits) - 1;
     size = sizeof(struct meta_name_tab_entry)*(meta_name_tab_mask+1);
     meta_name_tab = erts_db_alloc_nt(ERTS_ALC_T_DB_TABLES, size);
     ERTS_ETS_MISC_MEM_ADD(size);
@@ -3423,35 +3509,28 @@ void init_db(ErtsDbSpinCount db_spin_count)
 
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&ets_select_delete_continue_exp,
-			  am_ets, am_atom_put("delete_trap",11), 1,
-			  &ets_select_delete_1);
+			  am_ets, ERTS_MAKE_AM("select_delete_trap"), 1,
+			  &ets_select_delete_trap_1);
 
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&ets_select_count_continue_exp,
-			  am_ets, am_atom_put("count_trap",11), 1,
+			  am_ets, ERTS_MAKE_AM("count_trap"), 1,
 			  &ets_select_count_1);
 
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&ets_select_replace_continue_exp,
-                          am_ets, am_atom_put("replace_trap",11), 1,
+                          am_ets, ERTS_MAKE_AM("replace_trap"), 1,
                           &ets_select_replace_1);
 
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&ets_select_continue_exp,
-			  am_ets, am_atom_put("select_trap",11), 1,
+			  am_ets, ERTS_MAKE_AM("select_trap"), 1,
 			  &ets_select_trap_1);
 
     /* Non visual BIF to trap to. */
     erts_init_trap_export(&ets_delete_continue_exp,
-			  am_ets, am_atom_put("delete_trap",11), 1,
+			  am_ets, ERTS_MAKE_AM("delete_trap"), 1,
 			  &ets_delete_trap);
-
-    hp = ms_delete_all_buff;
-    ms_delete_all = CONS(hp, am_true, NIL);
-    hp += 2;
-    ms_delete_all = TUPLE3(hp,am_Underscore,NIL,ms_delete_all);
-    hp +=4;
-    ms_delete_all = CONS(hp, ms_delete_all,NIL);
 }
 
 void
@@ -3463,7 +3542,7 @@ erts_ets_sched_spec_data_init(ErtsSchedulerData *esdp)
     eaydp->tab = NULL;
     eaydp->queue = NULL;
     esdp->ets_tables.clist = NULL;
-    esdp->ets_tables.count = 0;
+    erts_atomic_init_nob(&esdp->ets_tables.count, 0);
 }
 
 
@@ -3495,14 +3574,14 @@ retry:
     
 	if (tb->common.owner != p->common.id) {
 	    if (to_proc != NULL ) {
-		erts_smp_proc_unlock(to_proc, to_locks);
+		erts_proc_unlock(to_proc, to_locks);
 	    }
 	    db_unlock(tb,LCK_WRITE);
 	    return !0; /* ok, someone already gave my table away */
 	}
 	if (tb->common.heir != to_pid) {  /* someone changed the heir */ 
 	    if (to_proc != NULL ) {
-		erts_smp_proc_unlock(to_proc, to_locks);
+		erts_proc_unlock(to_proc, to_locks);
 	    }
 	    if (to_pid == p->common.id || to_pid == am_none) {
 		return 0; /* no real heir, table still mine */
@@ -3515,7 +3594,7 @@ retry:
     }
     if (to_proc->common.u.alive.started_interval
 	!= tb->common.heir_started_interval) {
-	erts_smp_proc_unlock(to_proc, to_locks);
+	erts_proc_unlock(to_proc, to_locks);
 	return 0; /* heir dead and pid reused, table still mine */
     }
     
@@ -3532,7 +3611,7 @@ retry:
 	heir_data = tpv[1];
     }
     send_ets_transfer_message(p, to_proc, &to_locks, tb, heir_data);
-    erts_smp_proc_unlock(to_proc, to_locks);
+    erts_proc_unlock(to_proc, to_locks);
     return !0;
 }
 
@@ -3568,7 +3647,7 @@ send_ets_transfer_message(Process *c_p, Process *proc,
         hd_copy = copy_struct(heir_data, hd_sz, &hp, ohp);
     sender = c_p->common.id;
     msg = TUPLE4(hp, am_ETS_TRANSFER, tid, sender, hd_copy);
-    erts_queue_message(proc, *locks, mp, msg, sender);
+    erts_queue_proc_message(c_p, proc, *locks, mp, msg);
 }
 
 
@@ -3583,21 +3662,17 @@ static SWord proc_cleanup_fixed_table(Process* p, DbFixation* fix)
 	db_lock(tb, LCK_WRITE_REC);
 	if (!(tb->common.status & DB_DELETE)) {
 	    erts_aint_t diff;
-    #ifdef ERTS_SMP
-	    erts_smp_mtx_lock(&tb->common.fixlock);
-    #endif
+	    erts_mtx_lock(&tb->common.fixlock);
 
 	    ASSERT(fixing_procs_rbt_lookup(tb->common.fixing_procs, p));
 
 	    diff = -((erts_aint_t) fix->counter);
-	    erts_smp_refc_add(&tb->common.fix_count,diff,0);
+	    erts_refc_add(&tb->common.fix_count,diff,0);
 	    fix->counter = 0;
 
 	    fixing_procs_rbt_delete(&tb->common.fixing_procs, fix);
 
-    #ifdef ERTS_SMP
-	    erts_smp_mtx_unlock(&tb->common.fixlock);
-    #endif
+	    erts_mtx_unlock(&tb->common.fixlock);
 	    if (!IS_FIXED(tb) && IS_HASH_TABLE(tb->common.status)) {
 		work += db_unfix_table_hash(&(tb->hash));
 	    }
@@ -3654,9 +3729,9 @@ erts_db_process_exiting(Process *c_p, ErtsProcLocks c_p_locks)
 	switch (state->op) {
         case GET_OWNED_TABLE: {
             DbTable* tb;
-            erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
             tb = (DbTable*) erts_psd_get(c_p, ERTS_PSD_ETS_OWNED_TABLES);
-            erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
 
             if (!tb) {
                 /* Done with owned tables; now fixations */
@@ -3747,10 +3822,8 @@ static void fix_table_locked(Process* p, DbTable* tb)
 {
     DbFixation *fix;
 
-#ifdef ERTS_SMP
-    erts_smp_mtx_lock(&tb->common.fixlock);
-#endif
-    erts_smp_refc_inc(&tb->common.fix_count,1);
+    erts_mtx_lock(&tb->common.fixlock);
+    erts_refc_inc(&tb->common.fix_count,1);
     fix = tb->common.fixing_procs;
     if (fix == NULL) {
 	tb->common.time.monotonic
@@ -3763,9 +3836,7 @@ static void fix_table_locked(Process* p, DbTable* tb)
 	    ASSERT(fixed_tabs_find(NULL, fix));
 	    ++(fix->counter);
 
-#ifdef ERTS_SMP
-	    erts_smp_mtx_unlock(&tb->common.fixlock);
-#endif
+	    erts_mtx_unlock(&tb->common.fixlock);
 	    return;
 	}
     }
@@ -3778,9 +3849,7 @@ static void fix_table_locked(Process* p, DbTable* tb)
     fix->counter = 1;
     fixing_procs_rbt_insert(&tb->common.fixing_procs, fix);
 
-#ifdef ERTS_SMP
-    erts_smp_mtx_unlock(&tb->common.fixlock);
-#endif
+    erts_mtx_unlock(&tb->common.fixlock);
     p->flags |= F_USING_DB;
 
     fixed_tabs_insert(p, fix);
@@ -3793,20 +3862,16 @@ static void unfix_table_locked(Process* p,  DbTable* tb,
 {
     DbFixation* fix;
 
-#ifdef ERTS_SMP
-    erts_smp_mtx_lock(&tb->common.fixlock);
-#endif
+    erts_mtx_lock(&tb->common.fixlock);
     fix = fixing_procs_rbt_lookup(tb->common.fixing_procs, p);
 
     if (fix) {
-	erts_smp_refc_dec(&tb->common.fix_count,0);
+	erts_refc_dec(&tb->common.fix_count,0);
 	--(fix->counter);
 	ASSERT(fix->counter >= 0);
 	if (fix->counter == 0) {
 	    fixing_procs_rbt_delete(&tb->common.fixing_procs, fix);
-#ifdef ERTS_SMP
-	    erts_smp_mtx_unlock(&tb->common.fixlock);
-#endif
+	    erts_mtx_unlock(&tb->common.fixlock);
 	    fixed_tabs_delete(p, fix);
 
 	    erts_refc_dec(&fix->tabs.btid->intern.refc, 1);
@@ -3817,22 +3882,19 @@ static void unfix_table_locked(Process* p,  DbTable* tb,
 	    goto unlocked;
 	}
     }
-#ifdef ERTS_SMP
-    erts_smp_mtx_unlock(&tb->common.fixlock);
-#endif
+    erts_mtx_unlock(&tb->common.fixlock);
 unlocked:
 
     if (!IS_FIXED(tb) && IS_HASH_TABLE(tb->common.status)
-	&& erts_smp_atomic_read_nob(&tb->hash.fixdel) != (erts_aint_t)NULL) {
-#ifdef ERTS_SMP
+	&& erts_atomic_read_nob(&tb->hash.fixdel) != (erts_aint_t)NULL) {
 	if (*kind_p == LCK_READ && tb->common.is_thread_safe) {
 	    /* Must have write lock while purging pseudo-deleted (OTP-8166) */
-	    erts_smp_rwmtx_runlock(&tb->common.rwlock);
-	    erts_smp_rwmtx_rwlock(&tb->common.rwlock);
+	    erts_rwmtx_runlock(&tb->common.rwlock);
+	    erts_rwmtx_rwlock(&tb->common.rwlock);
 	    *kind_p = LCK_WRITE;
-	    if (tb->common.status & DB_DELETE) return;
+	    if (tb->common.status & (DB_DELETE|DB_BUSY))
+                return;
 	}
-#endif
 	db_unfix_table_hash(&(tb->hash));
     }
 }
@@ -3854,9 +3916,8 @@ static void free_fixations_op(DbFixation* fix, void* vctx)
     ASSERT(ctx->tb->common.status & DB_DELETE);
 
     diff = -((erts_aint_t) fix->counter);
-    erts_smp_refc_add(&ctx->tb->common.fix_count, diff, 0);
+    erts_refc_add(&ctx->tb->common.fix_count, diff, 0);
 
-#ifdef ERTS_SMP
     if (fix->procs.p != ctx->p) { /* Fixated by other process */
         fix->counter = 0;
 
@@ -3872,7 +3933,6 @@ static void free_fixations_op(DbFixation* fix, void* vctx)
          */
     }
     else
-#endif
     {
         fixed_tabs_delete(fix->procs.p, fix);
 
@@ -3885,7 +3945,6 @@ static void free_fixations_op(DbFixation* fix, void* vctx)
     ctx->cnt++;
 }
 
-#ifdef ERTS_SMP
 int erts_db_execute_free_fixation(Process* p, DbFixation* fix)
 {
     ASSERT(fix->counter == 0);
@@ -3897,13 +3956,12 @@ int erts_db_execute_free_fixation(Process* p, DbFixation* fix)
     ERTS_ETS_MISC_MEM_ADD(-sizeof(DbFixation));
     return 1;
 }
-#endif
 
 static SWord free_fixations_locked(Process* p, DbTable *tb)
 {
     struct free_fixations_ctx ctx;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rwlocked(&tb->common.rwlock));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&tb->common.rwlock));
 
     ctx.p = p;
     ctx.tb = tb;
@@ -3983,7 +4041,8 @@ static BIF_RETTYPE ets_delete_trap(BIF_ALIST_1)
 
     ASSERT(*ptr == make_pos_bignum_header(1));
 
-    if (free_table_continue(BIF_P, tb, reds) < 0) {
+    reds = free_table_continue(BIF_P, tb, reds);
+    if (reds < 0) {
         BUMP_ALL_REDS(BIF_P);
         BIF_TRAP1(&ets_delete_continue_exp, BIF_P, cont);
     }
@@ -4046,7 +4105,7 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
     int use_monotonic;
 
     if (What == am_size) {
-	ret = make_small(erts_smp_atomic_read_nob(&tb->common.nitems));
+	ret = make_small(erts_atomic_read_nob(&tb->common.nitems));
     } else if (What == am_type) {
 	if (tb->common.status & DB_SET)  {
 	    ret = am_set;
@@ -4059,7 +4118,7 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
 	    ret = am_bag;
 	}
     } else if (What == am_memory) {
-	Uint words = (Uint) ((erts_smp_atomic_read_nob(&tb->common.memory_size)
+	Uint words = (Uint) ((erts_atomic_read_nob(&tb->common.memory_size)
 			      + sizeof(Uint)
 			      - 1)
 			     / sizeof(Uint));
@@ -4089,14 +4148,17 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
 	ret = is_table_named(tb) ? am_true : am_false;
     } else if (What == am_compressed) {
 	ret = tb->common.compress ? am_true : am_false;
+    } else if (What == am_id) {
+        ret = make_tid(p, tb);
     }
+
     /*
      * For debugging purposes
      */
     else if (What == am_data) {
 	print_table(ERTS_PRINT_STDOUT, NULL, 1, tb);
 	ret = am_true;
-    } else if (What == am_atom_put("fixed",5)) { 
+    } else if (ERTS_IS_ATOM_STR("fixed",What)) {
 	if (IS_FIXED(tb))
 	    ret = am_true;
 	else
@@ -4105,9 +4167,7 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
 		= ERTS_IS_ATOM_STR("safe_fixed_monotonic_time",
 				   What))
 	       || ERTS_IS_ATOM_STR("safe_fixed", What)) {
-#ifdef ERTS_SMP
-	erts_smp_mtx_lock(&tb->common.fixlock);
-#endif
+	erts_mtx_lock(&tb->common.fixlock);
 	if (IS_FIXED(tb)) {
 	    Uint need;
 	    Eterm *hp;
@@ -4149,10 +4209,8 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
 	} else {
 	    ret = am_false;
 	}
-#ifdef ERTS_SMP
-	erts_smp_mtx_unlock(&tb->common.fixlock);
-#endif
-    } else if (What == am_atom_put("stats",5)) {
+	erts_mtx_unlock(&tb->common.fixlock);
+    } else if (ERTS_IS_ATOM_STR("stats",What)) {
 	if (IS_HASH_TABLE(tb->common.status)) {
 	    FloatDef f;
 	    DbHashStats stats;
@@ -4175,7 +4233,7 @@ static Eterm table_info(Process* p, DbTable* tb, Eterm What)
 	    std_dev_exp = make_float(hp);
 	    PUT_DOUBLE(f, hp);
 	    hp += FLOAT_SIZE_OBJECT;
-	    ret = TUPLE7(hp, make_small(erts_smp_atomic_read_nob(&tb->hash.nactive)),
+	    ret = TUPLE7(hp, make_small(erts_atomic_read_nob(&tb->hash.nactive)),
 			 avg, std_dev_real, std_dev_exp,
 			 make_small(stats.min_chain_len),
 			 make_small(stats.max_chain_len),
@@ -4207,9 +4265,9 @@ static void print_table(fmtfn_t to, void *to_arg, int show,  DbTable* tb)
 
     tb->common.meth->db_print(to, to_arg, show, tb);
 
-    erts_print(to, to_arg, "Objects: %d\n", (int)erts_smp_atomic_read_nob(&tb->common.nitems));
+    erts_print(to, to_arg, "Objects: %d\n", (int)erts_atomic_read_nob(&tb->common.nitems));
     erts_print(to, to_arg, "Words: %bpu\n",
-	       (Uint) ((erts_smp_atomic_read_nob(&tb->common.memory_size)
+	       (Uint) ((erts_atomic_read_nob(&tb->common.memory_size)
 			+ sizeof(Uint)
 			- 1)
 		       / sizeof(Uint)));
@@ -4249,9 +4307,9 @@ void db_info(fmtfn_t to, void *to_arg, int show)    /* Called by break handler *
 Uint
 erts_get_ets_misc_mem_size(void)
 {
-    ERTS_SMP_MEMORY_BARRIER;
+    ERTS_THR_MEMORY_BARRIER;
     /* Memory not allocated in ets_alloc */
-    return (Uint) erts_smp_atomic_read_nob(&erts_ets_misc_mem_size);
+    return (Uint) erts_atomic_read_nob(&erts_ets_misc_mem_size);
 }
 
 /* SMP Note: May only be used when system is locked */
@@ -4260,7 +4318,7 @@ erts_db_foreach_table(void (*func)(DbTable *, void *), void *arg)
 {
     int ix;
 
-    ASSERT(erts_smp_thr_progress_is_blocking());
+    ASSERT(erts_thr_progress_is_blocking());
 
     for (ix = 0; ix < erts_no_schedulers; ix++) {
         ErtsSchedulerData *esdp = ERTS_SCHEDULER_IX(ix);
@@ -4292,6 +4350,18 @@ erts_db_get_max_tabs()
     return db_max_tabs;
 }
 
+Uint erts_ets_table_count(void)
+{
+    Uint tb_count = 0;
+    Uint six;
+
+    for (six = 0; six < erts_no_schedulers; six++) {
+        ErtsSchedulerData *esdp = &erts_aligned_scheduler_data[six].esd;
+        tb_count += erts_atomic_read_nob(&esdp->ets_tables.count);
+    }
+    return tb_count;
+}
+
 /*
  * For testing of meta tables only.
  *
@@ -4312,7 +4382,7 @@ erts_ets_colliding_names(Process* p, Eterm name, Uint cnt)
             while (index >= atom_table_size()) {
                 char tmp[20];
                 erts_snprintf(tmp, sizeof(tmp), "am%x", atom_table_size());
-                erts_atom_put((byte *) tmp, strlen(tmp), ERTS_ATOM_ENC_LATIN1, 1);
+                erts_atom_put((byte *) tmp, sys_strlen(tmp), ERTS_ATOM_ENC_LATIN1, 1);
             }
             list = CONS(hp, make_atom(index), list);
             hp += 2;
@@ -4365,5 +4435,8 @@ void erts_lcnt_update_db_locks(int enable) {
     erts_schedule_multi_misc_aux_work(0, erts_no_schedulers,
         &lcnt_update_db_locks_per_sched, (void*)(UWord)enable);
 }
+#endif /* ERTS_ENABLE_LOCK_COUNT */
 
-#endif /* ERTS_ENABLE_LOCK_COUNT */
\ No newline at end of file
+#ifdef ETS_DBG_FORCE_TRAP
+erts_aint_t erts_ets_dbg_force_trap = 0;
+#endif
diff --git a/erts/emulator/beam/erl_db.h b/erts/emulator/beam/erl_db.h
index d83126b3a2..db1dec015c 100644
--- a/erts/emulator/beam/erl_db.h
+++ b/erts/emulator/beam/erl_db.h
@@ -45,7 +45,7 @@ typedef struct {
 } ErtsEtsAllYieldData;
 
 typedef struct {
-    Uint count;
+    erts_atomic_t count;
     DbTable *clist;
 } ErtsEtsTables;
 
@@ -69,6 +69,7 @@ typedef struct {
 /*TT*/
 
 Uint erts_get_ets_misc_mem_size(void);
+Uint erts_ets_table_count(void);
 
 typedef struct {
     DbTableCommon common;
@@ -93,7 +94,7 @@ union db_table {
     /*TT*/
 };
 
-#define DB_DEF_MAX_TABS 2053 /* Superseeded by environment variable 
+#define DB_DEF_MAX_TABS 8192 /* Superseeded by environment variable
 				"ERL_MAX_ETS_TABLES" */
 #define ERL_MAX_ETS_TABLES_ENV "ERL_MAX_ETS_TABLES"
 
@@ -124,16 +125,21 @@ extern Export ets_select_delete_continue_exp;
 extern Export ets_select_count_continue_exp;
 extern Export ets_select_replace_continue_exp;
 extern Export ets_select_continue_exp;
-extern erts_smp_atomic_t erts_ets_misc_mem_size;
+extern erts_atomic_t erts_ets_misc_mem_size;
 
 Eterm erts_ets_colliding_names(Process*, Eterm name, Uint cnt);
 Uint erts_db_get_max_tabs(void);
+Eterm erts_db_make_tid(Process *c_p, DbTableCommon *tb);
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
 void erts_lcnt_enable_db_lock_count(DbTable *tb, int enable);
 void erts_lcnt_update_db_locks(int enable);
 #endif
 
+#ifdef ETS_DBG_FORCE_TRAP
+extern erts_aint_t erts_ets_dbg_force_trap;
+#endif
+
 #endif /* ERL_DB_H__ */
 
 #if defined(ERTS_WANT_DB_INTERNAL__) && !defined(ERTS_HAVE_DB_INTERNAL__)
@@ -151,11 +157,11 @@ do {									\
     erts_aint_t sz__ = (((erts_aint_t) (ALLOC_SZ))			\
 			- ((erts_aint_t) (FREE_SZ)));			\
     ASSERT((TAB));							\
-    erts_smp_atomic_add_nob(&(TAB)->common.memory_size, sz__);		\
+    erts_atomic_add_nob(&(TAB)->common.memory_size, sz__);		\
 } while (0)
 
 #define ERTS_ETS_MISC_MEM_ADD(SZ) \
-  erts_smp_atomic_add_nob(&erts_ets_misc_mem_size, (SZ));
+  erts_atomic_add_nob(&erts_ets_misc_mem_size, (SZ));
 
 ERTS_GLB_INLINE void *erts_db_alloc(ErtsAlcType_t type,
 				    DbTable *tab,
@@ -292,7 +298,7 @@ erts_db_free(ErtsAlcType_t type, DbTable *tab, void *ptr, Uint size)
     ERTS_DB_ALC_MEM_UPDATE_(tab, size, 0);
 
     ASSERT(((void *) tab) != ptr
-	   || erts_smp_atomic_read_nob(&tab->common.memory_size) == 0);
+	   || erts_atomic_read_nob(&tab->common.memory_size) == 0);
 
     erts_free(type, ptr);
 }
diff --git a/erts/emulator/beam/erl_db_hash.c b/erts/emulator/beam/erl_db_hash.c
index cf928a9035..74d63325e6 100644
--- a/erts/emulator/beam/erl_db_hash.c
+++ b/erts/emulator/beam/erl_db_hash.c
@@ -21,6 +21,7 @@
 /*
 ** Implementation of unordered ETS tables.
 ** The tables are implemented as linear dynamic hash tables.
+** https://en.wikipedia.org/wiki/Linear_hashing
 */
 
 /* SMP:
@@ -109,22 +110,18 @@
 #define NSEG_2   256 /* Size of second segment table */
 #define NSEG_INC 128 /* Number of segments to grow after that */
 
-#ifdef ERTS_SMP
 #  define DB_USING_FINE_LOCKING(TB) (((TB))->common.type & DB_FINE_LOCKED)
-#else
-#  define DB_USING_FINE_LOCKING(TB) 0
-#endif
 
 #ifdef ETHR_ORDERED_READ_DEPEND
-#define SEGTAB(tb) ((struct segment**) erts_smp_atomic_read_nob(&(tb)->segtab))
+#define SEGTAB(tb) ((struct segment**) erts_atomic_read_nob(&(tb)->segtab))
 #else
 #define SEGTAB(tb)							\
     (DB_USING_FINE_LOCKING(tb)						\
-     ? ((struct segment**) erts_smp_atomic_read_ddrb(&(tb)->segtab))	\
-     : ((struct segment**) erts_smp_atomic_read_nob(&(tb)->segtab)))
+     ? ((struct segment**) erts_atomic_read_ddrb(&(tb)->segtab))	\
+     : ((struct segment**) erts_atomic_read_nob(&(tb)->segtab)))
 #endif
-#define NACTIVE(tb) ((int)erts_smp_atomic_read_nob(&(tb)->nactive))
-#define NITEMS(tb) ((int)erts_smp_atomic_read_nob(&(tb)->common.nitems))
+#define NACTIVE(tb) ((int)erts_atomic_read_nob(&(tb)->nactive))
+#define NITEMS(tb) ((int)erts_atomic_read_nob(&(tb)->common.nitems))
 
 #define SLOT_IX_TO_SEG_IX(i) (((i)+(EXT_SEGSZ-FIRST_SEGSZ)) >> EXT_SEGSZ_EXP)
 
@@ -142,31 +139,25 @@
 static ERTS_INLINE Uint hash_to_ix(DbTableHash* tb, HashValue hval)
 {
     Uint mask = (DB_USING_FINE_LOCKING(tb)
-		 ? erts_smp_atomic_read_acqb(&tb->szm)
-		 : erts_smp_atomic_read_nob(&tb->szm));
+		 ? erts_atomic_read_acqb(&tb->szm)
+		 : erts_atomic_read_nob(&tb->szm));
     Uint ix = hval & mask; 
-    if (ix >= erts_smp_atomic_read_nob(&tb->nactive)) {
+    if (ix >= erts_atomic_read_nob(&tb->nactive)) {
 	ix &= mask>>1;
-	ASSERT(ix < erts_smp_atomic_read_nob(&tb->nactive));
+	ASSERT(ix < erts_atomic_read_nob(&tb->nactive));
     }
     return ix;
 }
 
-/* Remember a slot containing a pseudo-deleted item (INVALID_HASH)
- * Return false if we got raced by unfixing thread
- * and the object should be deleted for real.
- */
-static ERTS_INLINE int add_fixed_deletion(DbTableHash* tb, int ix,
-                                          erts_aint_t fixated_by_me)
+
+static ERTS_INLINE int link_fixdel(DbTableHash* tb,
+                                   FixedDeletion* fixd,
+                                   erts_aint_t fixated_by_me)
 {
     erts_aint_t was_next;
     erts_aint_t exp_next;
-    FixedDeletion* fixd = (FixedDeletion*) erts_db_alloc(ERTS_ALC_T_DB_FIX_DEL,
-							 (DbTable *) tb,
-							 sizeof(FixedDeletion));
-    ERTS_ETS_MISC_MEM_ADD(sizeof(FixedDeletion));
-    fixd->slot = ix;
-    was_next = erts_smp_atomic_read_acqb(&tb->fixdel);
+
+    was_next = erts_atomic_read_acqb(&tb->fixdel);
     do { /* Lockless atomic insertion in linked list: */
         if (NFIXED(tb) <= fixated_by_me) {
             erts_db_free(ERTS_ALC_T_DB_FIX_DEL, (DbTable*)tb,
@@ -175,78 +166,90 @@ static ERTS_INLINE int add_fixed_deletion(DbTableHash* tb, int ix,
         }
         exp_next = was_next;
 	fixd->next = (FixedDeletion*) exp_next;
-	was_next = erts_smp_atomic_cmpxchg_mb(&tb->fixdel,
+	was_next = erts_atomic_cmpxchg_mb(&tb->fixdel,
                                               (erts_aint_t) fixd,
                                               exp_next);
     }while (was_next != exp_next);
     return 1;
 }
 
+/* Remember a slot containing a pseudo-deleted item
+ * Return false if we got raced by unfixing thread
+ * and the object should be deleted for real.
+ */
+static int add_fixed_deletion(DbTableHash* tb, int ix,
+                              erts_aint_t fixated_by_me)
+{
+    FixedDeletion* fixd = (FixedDeletion*) erts_db_alloc(ERTS_ALC_T_DB_FIX_DEL,
+							 (DbTable *) tb,
+							 sizeof(FixedDeletion));
+    ERTS_ETS_MISC_MEM_ADD(sizeof(FixedDeletion));
+    fixd->slot = ix;
+    fixd->all = 0;
+    return link_fixdel(tb, fixd, fixated_by_me);
+}
+
+
+static ERTS_INLINE int is_pseudo_deleted(HashDbTerm* p)
+{
+    return p->pseudo_deleted;
+}
 
-#define MAX_HASH 0xEFFFFFFFUL
-#define INVALID_HASH 0xFFFFFFFFUL
 
 /* optimised version of make_hash (normal case? atomic key) */
 #define MAKE_HASH(term) \
     ((is_atom(term) ? (atom_tab(atom_val(term))->slot.bucket.hvalue) : \
-      make_internal_hash(term, 0)) % MAX_HASH)
+      make_internal_hash(term, 0)) & MAX_HASH_MASK)
 
-#ifdef ERTS_SMP
 #  define DB_HASH_LOCK_MASK (DB_HASH_LOCK_CNT-1)
 #  define GET_LOCK(tb,hval) (&(tb)->locks->lck_vec[(hval) & DB_HASH_LOCK_MASK].lck)
 #  define GET_LOCK_MAYBE(tb,hval) ((tb)->common.is_thread_safe ? NULL : GET_LOCK(tb,hval))
 
 /* Fine grained read lock */
-static ERTS_INLINE erts_smp_rwmtx_t* RLOCK_HASH(DbTableHash* tb, HashValue hval)
+static ERTS_INLINE erts_rwmtx_t* RLOCK_HASH(DbTableHash* tb, HashValue hval)
 {
     if (tb->common.is_thread_safe) {
 	return NULL;
     } else {
-	erts_smp_rwmtx_t* lck = GET_LOCK(tb,hval);
+	erts_rwmtx_t* lck = GET_LOCK(tb,hval);
 	ASSERT(tb->common.type & DB_FINE_LOCKED);
-	erts_smp_rwmtx_rlock(lck);
+	erts_rwmtx_rlock(lck);
 	return lck;
     }
 }
 /* Fine grained write lock */
-static ERTS_INLINE erts_smp_rwmtx_t* WLOCK_HASH(DbTableHash* tb, HashValue hval)
+static ERTS_INLINE erts_rwmtx_t* WLOCK_HASH(DbTableHash* tb, HashValue hval)
 {
     if (tb->common.is_thread_safe) {
 	return NULL;
     } else {
-	erts_smp_rwmtx_t* lck = GET_LOCK(tb,hval);
+	erts_rwmtx_t* lck = GET_LOCK(tb,hval);
 	ASSERT(tb->common.type & DB_FINE_LOCKED);
-	erts_smp_rwmtx_rwlock(lck);
+	erts_rwmtx_rwlock(lck);
 	return lck;
     }
 }
 
-static ERTS_INLINE void RUNLOCK_HASH(erts_smp_rwmtx_t* lck)
+static ERTS_INLINE void RUNLOCK_HASH(erts_rwmtx_t* lck)
 {
     if (lck != NULL) {
-	erts_smp_rwmtx_runlock(lck);
+	erts_rwmtx_runlock(lck);
     }
 }
 
-static ERTS_INLINE void WUNLOCK_HASH(erts_smp_rwmtx_t* lck)
+static ERTS_INLINE void WUNLOCK_HASH(erts_rwmtx_t* lck)
 {
     if (lck != NULL) {
-	erts_smp_rwmtx_rwunlock(lck);
+	erts_rwmtx_rwunlock(lck);
     }
 }
-#else /* ERTS_SMP */
-# define RLOCK_HASH(tb,hval) NULL 
-# define WLOCK_HASH(tb,hval) NULL
-# define RUNLOCK_HASH(lck) ((void)lck) 
-# define WUNLOCK_HASH(lck) ((void)lck)
-#endif /* ERTS_SMP */
 
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #  define IFN_EXCL(tb,cmd) (((tb)->common.is_thread_safe) || (cmd))
-#  define IS_HASH_RLOCKED(tb,hval) IFN_EXCL(tb,erts_smp_lc_rwmtx_is_rlocked(GET_LOCK(tb,hval)))
-#  define IS_HASH_WLOCKED(tb,lck) IFN_EXCL(tb,erts_smp_lc_rwmtx_is_rwlocked(lck))
-#  define IS_TAB_WLOCKED(tb) erts_smp_lc_rwmtx_is_rwlocked(&(tb)->common.rwlock)
+#  define IS_HASH_RLOCKED(tb,hval) IFN_EXCL(tb,erts_lc_rwmtx_is_rlocked(GET_LOCK(tb,hval)))
+#  define IS_HASH_WLOCKED(tb,lck) IFN_EXCL(tb,erts_lc_rwmtx_is_rwlocked(lck))
+#  define IS_TAB_WLOCKED(tb) erts_lc_rwmtx_is_rwlocked(&(tb)->common.rwlock)
 #else
 #  define IS_HASH_RLOCKED(tb,hval) (1)
 #  define IS_HASH_WLOCKED(tb,hval) (1)
@@ -259,47 +262,44 @@ static ERTS_INLINE void WUNLOCK_HASH(erts_smp_rwmtx_t* lck)
 ** Slot READ locks updated accordingly, unlocked if EOT.
 */
 static ERTS_INLINE Sint next_slot(DbTableHash* tb, Uint ix,
-				  erts_smp_rwmtx_t** lck_ptr)
+				  erts_rwmtx_t** lck_ptr)
 {
-#ifdef ERTS_SMP
     ix += DB_HASH_LOCK_CNT;
     if (ix < NACTIVE(tb)) return ix;
     RUNLOCK_HASH(*lck_ptr);
     ix = (ix + 1) & DB_HASH_LOCK_MASK;
     if (ix != 0) *lck_ptr = RLOCK_HASH(tb,ix);
     return ix;
-#else
-    return (++ix < NACTIVE(tb)) ? ix : 0;
-#endif
 }
 /* Same as next_slot but with WRITE locking */
 static ERTS_INLINE Sint next_slot_w(DbTableHash* tb, Uint ix,
-				    erts_smp_rwmtx_t** lck_ptr)
+				    erts_rwmtx_t** lck_ptr)
 {
-#ifdef ERTS_SMP
     ix += DB_HASH_LOCK_CNT;
     if (ix < NACTIVE(tb)) return ix;
     WUNLOCK_HASH(*lck_ptr);
     ix = (ix + 1) & DB_HASH_LOCK_MASK;
     if (ix != 0) *lck_ptr = WLOCK_HASH(tb,ix);
     return ix;
-#else
-    return next_slot(tb,ix,lck_ptr);
-#endif
 }
 
 
-/*
- * Some special binary flags
- */
-#define BIN_FLAG_ALL_OBJECTS         BIN_FLAG_USR1
-
 
 static ERTS_INLINE void free_term(DbTableHash *tb, HashDbTerm* p)
 {
     db_free_term((DbTable*)tb, p, offsetof(HashDbTerm, dbterm));
 }
 
+static ERTS_INLINE void free_term_list(DbTableHash *tb, HashDbTerm* p)
+{
+    while (p) {
+        HashDbTerm* next = p->next;
+        free_term(tb, p);
+        p = next;
+    }
+}
+
+
 /*
  * Local types 
  */
@@ -309,9 +309,6 @@ struct mp_prefound {
 };
 
 struct mp_info {
-    int all_objects;		/* True if complete objects are always
-				 * returned from the match_spec (can use 
-				 * copy_shallow on the return value) */
     int something_can_match;	/* The match_spec is not "impossible" */
     int key_given;
     struct mp_prefound dlists[10];  /* Default list of "pre-found" buckets */
@@ -331,9 +328,7 @@ struct segment {
 
 /* An extended segment table */
 struct ext_segtab {
-#ifdef ERTS_SMP
     ErtsThrPrgrLaterOp lop;
-#endif
     struct segment** prev_segtab;  /* Used when table is shrinking */
     int prev_nsegs;                /* Size of prev_segtab */
     int nsegs;                     /* Size of this segtab */
@@ -347,9 +342,9 @@ static ERTS_INLINE void SET_SEGTAB(DbTableHash* tb,
 				   struct segment** segtab)
 {
     if (DB_USING_FINE_LOCKING(tb))
-	erts_smp_atomic_set_wb(&tb->segtab, (erts_aint_t) segtab);
+	erts_atomic_set_wb(&tb->segtab, (erts_aint_t) segtab);
     else
-	erts_smp_atomic_set_nob(&tb->segtab, (erts_aint_t) segtab);
+	erts_atomic_set_nob(&tb->segtab, (erts_aint_t) segtab);
 }
 
 /* Used by select_replace on analyze_pattern */
@@ -361,8 +356,8 @@ typedef int (*extra_match_validator_t)(int keypos, Eterm match, Eterm guard, Ete
 static struct ext_segtab* alloc_ext_segtab(DbTableHash* tb, unsigned seg_ix);
 static void alloc_seg(DbTableHash *tb);
 static int free_seg(DbTableHash *tb, int free_records);
-static HashDbTerm* next(DbTableHash *tb, Uint *iptr, erts_smp_rwmtx_t** lck_ptr,
-			HashDbTerm *list);
+static HashDbTerm* next_live(DbTableHash *tb, Uint *iptr, erts_rwmtx_t** lck_ptr,
+			     HashDbTerm *list);
 static HashDbTerm* search_list(DbTableHash* tb, Eterm key, 
 			       HashValue hval, HashDbTerm *list);
 static void shrink(DbTableHash* tb, int nitems);
@@ -423,7 +418,7 @@ static void db_print_hash(fmtfn_t to,
 			  void *to_arg,
 			  int show,
 			  DbTable *tbl);
-static int db_free_table_hash(DbTable *tbl);
+static int db_free_empty_table_hash(DbTable *tbl);
 
 static SWord db_free_table_continue_hash(DbTable *tbl, SWord reds);
 
@@ -432,7 +427,7 @@ static void db_foreach_offheap_hash(DbTable *,
 				    void (*)(ErlOffHeap *, void *),
 				    void *);
 
-static int db_delete_all_objects_hash(Process* p, DbTable* tbl);
+static SWord db_delete_all_objects_hash(Process* p, DbTable* tbl, SWord reds);
 #ifdef HARDDEBUG
 static void db_check_table_hash(DbTableHash *tb);
 #endif
@@ -457,7 +452,8 @@ static ERTS_INLINE void try_shrink(DbTableHash* tb)
 static ERTS_INLINE int has_live_key(DbTableHash* tb, HashDbTerm* b,
 				    Eterm key, HashValue hval)
 {
-    if (b->hvalue != hval) return 0;
+    if (b->hvalue != hval || is_pseudo_deleted(b))
+        return 0;
     else {
 	Eterm itemKey = GETKEY(tb, b->dbterm.tpl);
 	ASSERT(!is_header(itemKey));
@@ -470,7 +466,8 @@ static ERTS_INLINE int has_live_key(DbTableHash* tb, HashDbTerm* b,
 static ERTS_INLINE int has_key(DbTableHash* tb, HashDbTerm* b,
 			       Eterm key, HashValue hval)
 {
-    if (b->hvalue != hval && b->hvalue != INVALID_HASH) return 0;
+    if (b->hvalue != hval)
+        return 0;
     else {
 	Eterm itemKey = GETKEY(tb, b->dbterm.tpl);
 	ASSERT(!is_header(itemKey));
@@ -534,7 +531,7 @@ DbTableMethod db_hash =
     db_select_replace_continue_hash,
     db_take_hash,
     db_delete_all_objects_hash,
-    db_free_table_hash,
+    db_free_empty_table_hash,
     db_free_table_continue_hash,
     db_print_hash,
     db_foreach_offheap_hash,
@@ -559,7 +556,7 @@ static void restore_fixdel(DbTableHash* tb, FixedDeletion* fixdel)
 {
     /*int tries = 0;*/
     DEBUG_WAIT();
-    if (erts_smp_atomic_cmpxchg_relb(&tb->fixdel,
+    if (erts_atomic_cmpxchg_relb(&tb->fixdel,
 				     (erts_aint_t) fixdel,
 				     (erts_aint_t) NULL) != (erts_aint_t) NULL) {
 	/* Oboy, must join lists */    
@@ -568,13 +565,13 @@ static void restore_fixdel(DbTableHash* tb, FixedDeletion* fixdel)
 	erts_aint_t exp_tail;
 
 	while (last->next != NULL) last = last->next;
-	was_tail = erts_smp_atomic_read_acqb(&tb->fixdel);
+	was_tail = erts_atomic_read_acqb(&tb->fixdel);
 	do { /* Lockless atomic list insertion */
 	    exp_tail = was_tail;
 	    last->next = (FixedDeletion*) exp_tail;
 	    /*++tries;*/
 	    DEBUG_WAIT();
-	    was_tail = erts_smp_atomic_cmpxchg_relb(&tb->fixdel,
+	    was_tail = erts_atomic_cmpxchg_relb(&tb->fixdel,
 						    (erts_aint_t) fixdel,
 						    exp_tail);
 	}while (was_tail != exp_tail);
@@ -590,52 +587,62 @@ SWord db_unfix_table_hash(DbTableHash *tb)
     FixedDeletion* fixdel;
     SWord work = 0;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rwlocked(&tb->common.rwlock)
-		       || (erts_smp_lc_rwmtx_is_rlocked(&tb->common.rwlock)
-			   && !tb->common.is_thread_safe));
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&tb->common.rwlock)
+                   || (erts_lc_rwmtx_is_rlocked(&tb->common.rwlock)
+                       && !tb->common.is_thread_safe));
 restart:
-    fixdel = (FixedDeletion*) erts_smp_atomic_xchg_mb(&tb->fixdel,
-                                                      (erts_aint_t) NULL);
-    while (fixdel != NULL) {
-	FixedDeletion *fx = fixdel;
-	int ix = fx->slot;
-	HashDbTerm **bp;
-	HashDbTerm *b;
-	erts_smp_rwmtx_t* lck = WLOCK_HASH(tb,ix);
-
-	if (IS_FIXED(tb)) { /* interrupted by fixer */
-	    WUNLOCK_HASH(lck);
-	    restore_fixdel(tb,fixdel);
-	    if (!IS_FIXED(tb)) {
-		goto restart; /* unfixed again! */
-	    }
-	    return work;
-	}
-	if (ix < NACTIVE(tb)) {
-	    bp = &BUCKET(tb, ix);
-	    b = *bp;
-	    
-	    while (b != NULL) {
-		if (b->hvalue == INVALID_HASH) {
-		    *bp = b->next;
-		    free_term(tb, b);
-		    work++;
-		    b = *bp;
-		} else {
-		    bp = &b->next;
-		    b = b->next;
-		}
-	    }
-	}
-	/* else slot has been joined and purged by shrink() */
-	WUNLOCK_HASH(lck);
-	fixdel = fx->next;
-	erts_db_free(ERTS_ALC_T_DB_FIX_DEL,
-		     (DbTable *) tb,
-		     (void *) fx,
-		     sizeof(FixedDeletion));
-	ERTS_ETS_MISC_MEM_ADD(-sizeof(FixedDeletion));
-	work++;
+    fixdel = (FixedDeletion*) erts_atomic_xchg_mb(&tb->fixdel,
+                                                  (erts_aint_t) NULL);
+    while (fixdel) {
+        FixedDeletion *free_me;
+
+        do {
+            HashDbTerm **bp;
+            HashDbTerm *b;
+            HashDbTerm *free_us = NULL;
+            erts_rwmtx_t* lck;
+
+            lck = WLOCK_HASH(tb, fixdel->slot);
+
+            if (IS_FIXED(tb)) { /* interrupted by fixer */
+                WUNLOCK_HASH(lck);
+                restore_fixdel(tb,fixdel);
+                if (!IS_FIXED(tb)) {
+                    goto restart; /* unfixed again! */
+                }
+                return work;
+            }
+            if (fixdel->slot < NACTIVE(tb)) {
+                bp = &BUCKET(tb, fixdel->slot);
+                b = *bp;
+
+                while (b != NULL) {
+                    if (is_pseudo_deleted(b)) {
+                        HashDbTerm* nxt = b->next;
+                        b->next = free_us;
+                        free_us = b;
+                        work++;
+                        b = *bp = nxt;
+                    } else {
+                        bp = &b->next;
+                        b = b->next;
+                    }
+                }
+            }
+            /* else slot has been joined and purged by shrink() */
+            WUNLOCK_HASH(lck);
+            free_term_list(tb, free_us);
+
+        }while (fixdel->all && fixdel->slot-- > 0);
+
+        free_me = fixdel;
+        fixdel = fixdel->next;
+        erts_db_free(ERTS_ALC_T_DB_FIX_DEL,
+                     (DbTable *) tb,
+                     (void *) free_me,
+                     sizeof(FixedDeletion));
+        ERTS_ETS_MISC_MEM_ADD(-sizeof(FixedDeletion));
+        work++;
     }
 
     /* ToDo: Maybe try grow/shrink the table as well */
@@ -647,10 +654,10 @@ int db_create_hash(Process *p, DbTable *tbl)
 {
     DbTableHash *tb = &tbl->hash;
 
-    erts_smp_atomic_init_nob(&tb->szm, FIRST_SEGSZ_MASK);
-    erts_smp_atomic_init_nob(&tb->nactive, FIRST_SEGSZ);
-    erts_smp_atomic_init_nob(&tb->fixdel, (erts_aint_t)NULL);
-    erts_smp_atomic_init_nob(&tb->segtab, (erts_aint_t)NULL);
+    erts_atomic_init_nob(&tb->szm, FIRST_SEGSZ_MASK);
+    erts_atomic_init_nob(&tb->nactive, FIRST_SEGSZ);
+    erts_atomic_init_nob(&tb->fixdel, (erts_aint_t)NULL);
+    erts_atomic_init_nob(&tb->segtab, (erts_aint_t)NULL);
     SET_SEGTAB(tb, tb->first_segtab);
     tb->nsegs = NSEG_1;
     tb->nslots = FIRST_SEGSZ;
@@ -659,32 +666,30 @@ int db_create_hash(Process *p, DbTable *tbl)
                                                           SIZEOF_SEGMENT(FIRST_SEGSZ));
     sys_memset(tb->first_segtab[0], 0, SIZEOF_SEGMENT(FIRST_SEGSZ));
 
-#ifdef ERTS_SMP
-    erts_smp_atomic_init_nob(&tb->is_resizing, 0);
+    erts_atomic_init_nob(&tb->is_resizing, 0);
     if (tb->common.type & DB_FINE_LOCKED) {
-	erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
+	erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
 	int i;
 	if (tb->common.type & DB_FREQ_READ)
-	    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
+	    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
 	if (erts_ets_rwmtx_spin_count >= 0)
 	    rwmtx_opt.main_spincount = erts_ets_rwmtx_spin_count;
-	tb->locks = (DbTableHashFineLocks*) erts_db_alloc_fnf(ERTS_ALC_T_DB_SEG, /* Other type maybe? */ 
-							      (DbTable *) tb,
-							      sizeof(DbTableHashFineLocks));	    	    
+	tb->locks = (DbTableHashFineLocks*) erts_db_alloc(ERTS_ALC_T_DB_SEG, /* Other type maybe? */
+                                                          (DbTable *) tb,
+                                                          sizeof(DbTableHashFineLocks));
 	for (i=0; i<DB_HASH_LOCK_CNT; ++i) {
-            erts_smp_rwmtx_init_opt(&tb->locks->lck_vec[i].lck, &rwmtx_opt,
+            erts_rwmtx_init_opt(&tb->locks->lck_vec[i].lck, &rwmtx_opt,
                 "db_hash_slot", tb->common.the_name, ERTS_LOCK_FLAGS_CATEGORY_DB);
 	}
 	/* This important property is needed to guarantee the two buckets
     	 * involved in a grow/shrink operation it protected by the same lock:
 	 */
-	ASSERT(erts_smp_atomic_read_nob(&tb->nactive) % DB_HASH_LOCK_CNT == 0);
+	ASSERT(erts_atomic_read_nob(&tb->nactive) % DB_HASH_LOCK_CNT == 0);
     }
     else { /* coarse locking */
 	tb->locks = NULL;
     }
     ERTS_THR_MEMORY_BARRIER;
-#endif /* ERST_SMP */
     return DB_ERROR_NONE;
 }
 
@@ -692,22 +697,12 @@ static int db_first_hash(Process *p, DbTable *tbl, Eterm *ret)
 {
     DbTableHash *tb = &tbl->hash;
     Uint ix = 0;
-    erts_smp_rwmtx_t* lck = RLOCK_HASH(tb,ix);
+    erts_rwmtx_t* lck = RLOCK_HASH(tb,ix);
     HashDbTerm* list;
 
-    for (;;) {
-	list = BUCKET(tb,ix);
-	if (list != NULL) {
-	    if (list->hvalue == INVALID_HASH) {
-		list = next(tb,&ix,&lck,list);
-	    }
-	    break;
-	}
-	if ((ix=next_slot(tb,ix,&lck)) == 0) {
-	    list = NULL;
-	    break;
-	}
-    }
+    list = BUCKET(tb,ix);
+    list = next_live(tb, &ix, &lck, list);
+
     if (list != NULL) {
 	*ret = db_copy_key(p, tbl, &list->dbterm);
 	RUNLOCK_HASH(lck);
@@ -725,7 +720,7 @@ static int db_next_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
     HashValue hval;
     Uint ix;
     HashDbTerm* b;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
 
     hval = MAKE_HASH(key);
     lck = RLOCK_HASH(tb,hval);
@@ -744,13 +739,13 @@ static int db_next_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
     }
     /* Key found */
 
-    b = next(tb, &ix, &lck, b);
+    b = next_live(tb, &ix, &lck, b->next);
     if (tb->common.status & (DB_BAG | DB_DUPLICATE_BAG)) {
 	while (b != 0) {
 	    if (!has_live_key(tb, b, key, hval)) {
 		break;
 	    }
-	    b = next(tb, &ix, &lck, b);
+	    b = next_live(tb, &ix, &lck, b->next);
 	}
     }
     if (b == NULL) {
@@ -772,7 +767,7 @@ int db_put_hash(DbTable *tbl, Eterm obj, int key_clash_fail)
     HashDbTerm** bp;
     HashDbTerm* b;
     HashDbTerm* q;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int nitems;
     int ret = DB_ERROR_NONE;
 
@@ -797,8 +792,9 @@ int db_put_hash(DbTable *tbl, Eterm obj, int key_clash_fail)
     */
     if (tb->common.status & DB_SET) {
 	HashDbTerm* bnext = b->next;
-	if (b->hvalue == INVALID_HASH) {
-	    erts_smp_atomic_inc_nob(&tb->common.nitems);
+	if (is_pseudo_deleted(b)) {
+	    erts_atomic_inc_nob(&tb->common.nitems);
+            b->pseudo_deleted = 0;
 	}
 	else if (key_clash_fail) {
 	    ret = DB_ERROR_BADKEY;
@@ -806,14 +802,14 @@ int db_put_hash(DbTable *tbl, Eterm obj, int key_clash_fail)
 	}
 	q = replace_dbterm(tb, b, obj);
 	q->next = bnext;
-	q->hvalue = hval; /* In case of INVALID_HASH */
+	ASSERT(q->hvalue == hval);
 	*bp = q;
 	goto Ldone;
     }
     else if (key_clash_fail) { /* && (DB_BAG || DB_DUPLICATE_BAG) */
 	q = b;
 	do {
-	    if (q->hvalue != INVALID_HASH) {
+	    if (!is_pseudo_deleted(q)) {
 		ret = DB_ERROR_BADKEY;
 		goto Ldone;
 	    }
@@ -825,9 +821,10 @@ int db_put_hash(DbTable *tbl, Eterm obj, int key_clash_fail)
 	q = b;
 	do {
 	    if (db_eq(&tb->common,obj,&q->dbterm)) {
-		if (q->hvalue == INVALID_HASH) {
-		    erts_smp_atomic_inc_nob(&tb->common.nitems);
-		    q->hvalue = hval;
+		if (is_pseudo_deleted(q)) {
+		    erts_atomic_inc_nob(&tb->common.nitems);
+                    q->pseudo_deleted = 0;
+		    ASSERT(q->hvalue == hval);
 		    if (q != b) { /* must move to preserve key insertion order */
 			*qp = q->next;
 			q->next = b;
@@ -845,9 +842,10 @@ int db_put_hash(DbTable *tbl, Eterm obj, int key_clash_fail)
 Lnew:
     q = new_dbterm(tb, obj);
     q->hvalue = hval;
+    q->pseudo_deleted = 0;
     q->next = b;
     *bp = q;
-    nitems = erts_smp_atomic_inc_read_nob(&tb->common.nitems);
+    nitems = erts_atomic_inc_read_nob(&tb->common.nitems);
     WUNLOCK_HASH(lck);
     {
 	int nactive = NACTIVE(tb);       
@@ -872,7 +870,7 @@ get_term_list(Process *p, DbTableHash *tb, Eterm key, HashValue hval,
 
     if (tb->common.status & (DB_BAG | DB_DUPLICATE_BAG)) {
         while (b2 && has_key(tb, b2, key, hval)) {
-	    if (b2->hvalue != INVALID_HASH)
+	    if (!is_pseudo_deleted(b2))
 		sz += b2->dbterm.size + 2;
 
             b2 = b2->next;
@@ -891,7 +889,7 @@ int db_get_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
     HashValue hval;
     int ix;
     HashDbTerm* b;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
 
     hval = MAKE_HASH(key);
     lck = RLOCK_HASH(tb,hval);
@@ -917,7 +915,7 @@ static int db_member_hash(DbTable *tbl, Eterm key, Eterm *ret)
     HashValue hval;
     int ix;
     HashDbTerm* b1;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
 
     hval = MAKE_HASH(key);
     ix = hash_to_ix(tb, hval);
@@ -946,7 +944,7 @@ static int db_get_element_hash(Process *p, DbTable *tbl,
     HashValue hval;
     int ix;
     HashDbTerm* b1;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int retval;
     
     hval = MAKE_HASH(key);
@@ -968,7 +966,7 @@ static int db_get_element_hash(Process *p, DbTable *tbl,
 
 		while(b2 != NULL && has_key(tb,b2,key,hval)) {
 		    if (ndex > arityval(b2->dbterm.tpl[0])
-			&& b2->hvalue != INVALID_HASH) {
+			&& !is_pseudo_deleted(b2)) {
 			retval = DB_ERROR_BADITEM;
 			goto done;
 		    }
@@ -976,7 +974,7 @@ static int db_get_element_hash(Process *p, DbTable *tbl,
 		}
 		b = b1;
 		while(b != b2) {
-		    if (b->hvalue != INVALID_HASH) {
+		    if (!is_pseudo_deleted(b)) {
 			Eterm *hp;
 			Eterm copy = db_copy_element_from_ets(&tb->common, p,
 							      &b->dbterm, ndex, &hp, 2);
@@ -1011,7 +1009,8 @@ int db_erase_hash(DbTable *tbl, Eterm key, Eterm *ret)
     int ix;
     HashDbTerm** bp;
     HashDbTerm* b;
-    erts_smp_rwmtx_t* lck;
+    HashDbTerm* free_us = NULL;
+    erts_rwmtx_t* lck;
     int nitems_diff = 0;
 
     hval = MAKE_HASH(key);
@@ -1026,16 +1025,17 @@ int db_erase_hash(DbTable *tbl, Eterm key, Eterm *ret)
 	    if (nitems_diff == -1 && IS_FIXED(tb)
                 && add_fixed_deletion(tb, ix, 0)) {
 		/* Pseudo remove (no need to keep several of same key) */
-		b->hvalue = INVALID_HASH;
+		b->pseudo_deleted = 1;
 	    } else {
-		*bp = b->next;
-		free_term(tb, b);
-		b = *bp;
+		HashDbTerm* next = b->next;
+                b->next = free_us;
+		free_us = b;
+		b = *bp = next;
 		continue;
 	    }
 	}
 	else {
-	    if (nitems_diff && b->hvalue != INVALID_HASH)
+	    if (nitems_diff && !is_pseudo_deleted(b))
 		break;
 	}
 	bp = &b->next;
@@ -1043,9 +1043,10 @@ int db_erase_hash(DbTable *tbl, Eterm key, Eterm *ret)
     }
     WUNLOCK_HASH(lck);
     if (nitems_diff) {
-	erts_smp_atomic_add_nob(&tb->common.nitems, nitems_diff);
+	erts_atomic_add_nob(&tb->common.nitems, nitems_diff);
 	try_shrink(tb);
     }
+    free_term_list(tb, free_us);
     *ret = am_true;
     return DB_ERROR_NONE;
 }    
@@ -1060,7 +1061,8 @@ static int db_erase_object_hash(DbTable *tbl, Eterm object, Eterm *ret)
     int ix;
     HashDbTerm** bp;
     HashDbTerm* b;
-    erts_smp_rwmtx_t* lck;
+    HashDbTerm* free_us = NULL;
+    erts_rwmtx_t* lck;
     int nitems_diff = 0;
     int nkeys = 0;
     Eterm key;
@@ -1078,13 +1080,14 @@ static int db_erase_object_hash(DbTable *tbl, Eterm object, Eterm *ret)
 	    if (db_eq(&tb->common,object, &b->dbterm)) {
 		--nitems_diff;
 		if (nkeys==1 && IS_FIXED(tb) && add_fixed_deletion(tb,ix,0)) {
-		    b->hvalue = INVALID_HASH;        /* Pseudo remove */
+		    b->pseudo_deleted = 1;
 		    bp = &b->next;
 		    b = b->next;
 		} else {
-		    *bp = b->next;
-		    free_term(tb, b);
-		    b = *bp;
+                    HashDbTerm* next = b->next;
+		    b->next = free_us;
+                    free_us = b;
+		    b = *bp = next;
 		}
 		if (tb->common.status & (DB_DUPLICATE_BAG)) {
 		    continue;
@@ -1093,7 +1096,7 @@ static int db_erase_object_hash(DbTable *tbl, Eterm object, Eterm *ret)
 		}
 	    }
 	}
-	else if (nitems_diff && b->hvalue != INVALID_HASH) {
+	else if (nitems_diff && !is_pseudo_deleted(b)) {
 	    break;
 	}
 	bp = &b->next;
@@ -1101,9 +1104,10 @@ static int db_erase_object_hash(DbTable *tbl, Eterm object, Eterm *ret)
     }
     WUNLOCK_HASH(lck);
     if (nitems_diff) {
-	erts_smp_atomic_add_nob(&tb->common.nitems, nitems_diff);
+	erts_atomic_add_nob(&tb->common.nitems, nitems_diff);
 	try_shrink(tb);
     }
+    free_term_list(tb, free_us);
     *ret = am_true;
     return DB_ERROR_NONE;
 }    
@@ -1112,7 +1116,7 @@ static int db_erase_object_hash(DbTable *tbl, Eterm object, Eterm *ret)
 static int db_slot_hash(Process *p, DbTable *tbl, Eterm slot_term, Eterm *ret)
 {
     DbTableHash *tb = &tbl->hash;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     Sint slot;
     int retval;
     int nactive;
@@ -1139,28 +1143,19 @@ static int db_slot_hash(Process *p, DbTable *tbl, Eterm slot_term, Eterm *ret)
 
 
 /*
- * This is just here so I can take care of the return value 
- * that is to be sent during a trap (the BIF_TRAP macros explicitly returns)
- */
-static BIF_RETTYPE bif_trap1(Export *bif,
-			      Process *p, 
-			      Eterm p1) 
-{
-    BIF_TRAP1(bif, p, p1);
-}
-
-
-/*
  * Match traversal callbacks
  */
 
+typedef struct match_callbacks_t_ match_callbacks_t;
+struct match_callbacks_t_
+{
 /* Called when no match is possible.
  *      context_ptr: Pointer to context
  *      ret: Pointer to traversal function term return.
  *
  * Both the direct return value and 'ret' are used as the traversal function return values.
  */
-typedef int (*mtraversal_on_nothing_can_match_t)(void* context_ptr, Eterm* ret);
+    int (*on_nothing_can_match)(match_callbacks_t* ctx, Eterm* ret);
 
 /* Called for each match result.
  *      context_ptr: Pointer to context
@@ -1171,8 +1166,8 @@ typedef int (*mtraversal_on_nothing_can_match_t)(void* context_ptr, Eterm* ret);
  *
  * Should return 1 for successful match, 0 otherwise.
  */
-typedef int (*mtraversal_on_match_res_t)(void* context_ptr, Sint slot_ix, HashDbTerm*** current_ptr_ptr,
-                                         Eterm match_res);
+    int (*on_match_res)(match_callbacks_t* ctx, Sint slot_ix,
+                        HashDbTerm*** current_ptr_ptr, Eterm match_res);
 
 /* Called when either we've matched enough elements in this cycle or EOT was reached.
  *      context_ptr: Pointer to context
@@ -1185,8 +1180,8 @@ typedef int (*mtraversal_on_match_res_t)(void* context_ptr, Sint slot_ix, HashDb
  * Both the direct return value and 'ret' are used as the traversal function return values.
  * If *mpp is set to NULL, it won't be deallocated (useful for trapping.)
  */
-typedef int (*mtraversal_on_loop_ended_t)(void* context_ptr, Sint slot_ix, Sint got,
-                                          Sint iterations_left, Binary** mpp, Eterm* ret);
+    int (*on_loop_ended)(match_callbacks_t* ctx, Sint slot_ix, Sint got,
+                         Sint iterations_left, Binary** mpp, Eterm* ret);
 
 /* Called when it's time to trap
  *      context_ptr: Pointer to context
@@ -1198,7 +1193,11 @@ typedef int (*mtraversal_on_loop_ended_t)(void* context_ptr, Sint slot_ix, Sint
  * Both the direct return value and 'ret' are used as the traversal function return values.
  * If *mpp is set to NULL, it won't be deallocated (useful for trapping.)
  */
-typedef int (*mtraversal_on_trap_t)(void* context_ptr, Sint slot_ix, Sint got, Binary** mpp, Eterm* ret);
+    int (*on_trap)(match_callbacks_t* ctx, Sint slot_ix, Sint got, Binary** mpp,
+                   Eterm* ret);
+
+};
+
 
 /*
  * Begin hash table match traversal
@@ -1211,11 +1210,7 @@ static int match_traverse(Process* p, DbTableHash* tb,
                           Eterm** hpp,          /* Heap */
                           int lock_for_write,   /* Set to 1 if we're going to delete or
                                                    modify existing terms */
-                          mtraversal_on_nothing_can_match_t on_nothing_can_match,
-                          mtraversal_on_match_res_t on_match_res,
-                          mtraversal_on_loop_ended_t on_loop_ended,
-                          mtraversal_on_trap_t on_trap,
-                          void* context_ptr, /* State for callbacks above */
+                          match_callbacks_t* ctx,
                           Eterm* ret)
 {
     Sint slot_ix;                  /* Slot index */
@@ -1227,18 +1222,13 @@ static int match_traverse(Process* p, DbTableHash* tb,
     unsigned current_list_pos = 0; /* Prefound buckets list index */
     Eterm match_res;
     Sint got = 0;                  /* Matched terms counter */
-    erts_smp_rwmtx_t* lck;         /* Slot lock */
+    erts_rwmtx_t* lck;         /* Slot lock */
     int ret_value;
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_t* (*lock_hash_function)(DbTableHash*, HashValue)
+    erts_rwmtx_t* (*lock_hash_function)(DbTableHash*, HashValue)
         = (lock_for_write ? WLOCK_HASH : RLOCK_HASH);
-    void (*unlock_hash_function)(erts_smp_rwmtx_t*)
+    void (*unlock_hash_function)(erts_rwmtx_t*)
         = (lock_for_write ? WUNLOCK_HASH : RUNLOCK_HASH);
-#else
-    #define lock_hash_function(tb, hval) NULL
-    #define unlock_hash_function(lck) ((void)lck)
-#endif
-    Sint (*next_slot_function)(DbTableHash*, Uint, erts_smp_rwmtx_t**)
+    Sint (*next_slot_function)(DbTableHash*, Uint, erts_rwmtx_t**)
         = (lock_for_write ? next_slot_w : next_slot);
 
     if ((ret_value = analyze_pattern(tb, pattern, extra_match_validator, &mpi))
@@ -1250,14 +1240,10 @@ static int match_traverse(Process* p, DbTableHash* tb,
 
     if (!mpi.something_can_match) {
         /* Can't possibly match anything */
-        ret_value = on_nothing_can_match(context_ptr, ret);
+        ret_value = ctx->on_nothing_can_match(ctx, ret);
         goto done;
     }
 
-    if (mpi.all_objects) {
-        mpi.mp->intern.flags |= BIN_FLAG_ALL_OBJECTS;
-    }
-
     /*
      * Look for initial slot / bucket
      */
@@ -1273,7 +1259,8 @@ static int match_traverse(Process* p, DbTableHash* tb,
             }
             slot_ix = next_slot_function(tb,slot_ix,&lck);
             if (slot_ix == 0) {
-                ret_value = on_loop_ended(context_ptr, slot_ix, got, iterations_left, &mpi.mp, ret);
+                ret_value = ctx->on_loop_ended(ctx, slot_ix, got, iterations_left,
+                                               &mpi.mp, ret);
                 goto done;
             }
         }
@@ -1291,11 +1278,11 @@ static int match_traverse(Process* p, DbTableHash* tb,
      */
     for(;;) {
         if (*current_ptr != NULL) {
-            if ((*current_ptr)->hvalue != INVALID_HASH) {
-                match_res = db_match_dbterm(&tb->common, p, mpi.mp, 0,
+            if (!is_pseudo_deleted(*current_ptr)) {
+                match_res = db_match_dbterm(&tb->common, p, mpi.mp,
                                             &(*current_ptr)->dbterm, hpp, 2);
                 saved_current = *current_ptr;
-                if (on_match_res(context_ptr, slot_ix, &current_ptr, match_res)) {
+                if (ctx->on_match_res(ctx, slot_ix, &current_ptr, match_res)) {
                     ++got;
                 }
                 --iterations_left;
@@ -1309,7 +1296,7 @@ static int match_traverse(Process* p, DbTableHash* tb,
         else if (mpi.key_given) {  /* Key is bound */
             unlock_hash_function(lck);
             if (current_list_pos == mpi.num_lists) {
-                ret_value = on_loop_ended(context_ptr, -1, got, iterations_left, &mpi.mp, ret);
+                ret_value = ctx->on_loop_ended(ctx, -1, got, iterations_left, &mpi.mp, ret);
                 goto done;
             } else {
                 slot_ix = mpi.lists[current_list_pos].ix;
@@ -1334,18 +1321,18 @@ static int match_traverse(Process* p, DbTableHash* tb,
                  * Since many heap fragments will make the GC slower, trap and GC now.
                  */
                 unlock_hash_function(lck);
-                ret_value = on_trap(context_ptr, slot_ix, got, &mpi.mp, ret);
+                ret_value = ctx->on_trap(ctx, slot_ix, got, &mpi.mp, ret);
                 goto done;
             }
             current_ptr = &BUCKET(tb,slot_ix);
         }
     }
 
-    ret_value = on_loop_ended(context_ptr, slot_ix, got, iterations_left, &mpi.mp, ret);
+    ret_value = ctx->on_loop_ended(ctx, slot_ix, got, iterations_left, &mpi.mp, ret);
 
 done:
     /* We should only jump directly to this label if
-     * we've already called on_nothing_can_match / on_loop_ended / on_trap
+     * we've already called ctx->nothing_can_match / loop_ended / trap
      */
     if (mpi.mp != NULL) {
         erts_bin_free(mpi.mp);
@@ -1356,10 +1343,6 @@ done:
     }
     return ret_value;
 
-#ifndef SMP
-#undef lock_hash_function
-#undef unlock_hash_function
-#endif
 }
 
 /*
@@ -1374,30 +1357,21 @@ static int match_traverse_continue(Process* p, DbTableHash* tb,
                                    Binary** mpp,         /* Existing match program */
                                    int lock_for_write,   /* Set to 1 if we're going to delete or
                                                             modify existing terms */
-                                   mtraversal_on_match_res_t on_match_res,
-                                   mtraversal_on_loop_ended_t on_loop_ended,
-                                   mtraversal_on_trap_t on_trap,
-                                   void* context_ptr, /* For callbacks */
+                                   match_callbacks_t* ctx,
                                    Eterm* ret)
 {
-    int all_objects = (*mpp)->intern.flags & BIN_FLAG_ALL_OBJECTS;
     HashDbTerm** current_ptr;  /* Refers to either the bucket pointer or
                                        * the 'next' pointer in the previous term
                                        */
     HashDbTerm* saved_current; /* Helper to avoid double skip on match */
     Eterm match_res;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int ret_value;
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_t* (*lock_hash_function)(DbTableHash*, HashValue)
+    erts_rwmtx_t* (*lock_hash_function)(DbTableHash*, HashValue)
         = (lock_for_write ? WLOCK_HASH : RLOCK_HASH);
-    void (*unlock_hash_function)(erts_smp_rwmtx_t*)
+    void (*unlock_hash_function)(erts_rwmtx_t*)
         = (lock_for_write ? WUNLOCK_HASH : RUNLOCK_HASH);
-#else
-    #define lock_hash_function(tb, hval) NULL
-    #define unlock_hash_function(lck) ((void)lck)
-#endif
-    Sint (*next_slot_function)(DbTableHash* tb, Uint ix, erts_smp_rwmtx_t** lck_ptr)
+    Sint (*next_slot_function)(DbTableHash* tb, Uint ix, erts_rwmtx_t** lck_ptr)
         = (lock_for_write ? next_slot_w : next_slot);
 
     if (got < 0) {
@@ -1409,7 +1383,7 @@ static int match_traverse_continue(Process* p, DbTableHash* tb,
        || (chunk_size && got >= chunk_size))
     {
         /* Already got all or enough in the match_list */
-        ret_value = on_loop_ended(context_ptr, slot_ix, got, iterations_left, mpp, ret);
+        ret_value = ctx->on_loop_ended(ctx, slot_ix, got, iterations_left, mpp, ret);
         goto done;
     }
 
@@ -1427,11 +1401,11 @@ static int match_traverse_continue(Process* p, DbTableHash* tb,
     current_ptr = &BUCKET(tb,slot_ix);
     for(;;) {
         if (*current_ptr != NULL) {
-            if ((*current_ptr)->hvalue != INVALID_HASH) {
-                match_res = db_match_dbterm(&tb->common, p, *mpp, all_objects,
+            if (!is_pseudo_deleted(*current_ptr)) {
+                match_res = db_match_dbterm(&tb->common, p, *mpp,
                                             &(*current_ptr)->dbterm, hpp, 2);
                 saved_current = *current_ptr;
-                if (on_match_res(context_ptr, slot_ix, &current_ptr, match_res)) {
+                if (ctx->on_match_res(ctx, slot_ix, &current_ptr, match_res)) {
                     ++got;
                 }
                 --iterations_left;
@@ -1457,14 +1431,14 @@ static int match_traverse_continue(Process* p, DbTableHash* tb,
                  * Since many heap fragments will make the GC slower, trap and GC now.
                  */
                 unlock_hash_function(lck);
-                ret_value = on_trap(context_ptr, slot_ix, got, mpp, ret);
+                ret_value = ctx->on_trap(ctx, slot_ix, got, mpp, ret);
                 goto done;
             }
             current_ptr = &BUCKET(tb,slot_ix);
         }
     }
 
-    ret_value = on_loop_ended(context_ptr, slot_ix, got, iterations_left, mpp, ret);
+    ret_value = ctx->on_loop_ended(ctx, slot_ix, got, iterations_left, mpp, ret);
 
 done:
     /* We should only jump directly to this label if
@@ -1472,10 +1446,6 @@ done:
      */
     return ret_value;
 
-#ifndef SMP
-#undef lock_hash_function
-#undef unlock_hash_function
-#endif
 }
 
 
@@ -1485,7 +1455,7 @@ done:
  * as well as their continuation-handling counterparts.
  */
 
-static ERTS_INLINE int on_mtraversal_simple_trap(Export* trap_function,
+static ERTS_INLINE int on_simple_trap(Export* trap_function,
                                                  Process* p,
                                                  DbTableHash* tb,
                                                  Eterm tid,
@@ -1504,20 +1474,24 @@ static ERTS_INLINE int on_mtraversal_simple_trap(Export* trap_function,
 
     BUMP_ALL_REDS(p);
     if (IS_USMALL(0, got)) {
-	hp = HAlloc(p,  base_halloc_sz + 5);
+	hp = HAllocX(p,  base_halloc_sz + 5, ERTS_MAGIC_REF_THING_SIZE);
 	egot = make_small(got);
     }
     else {
-	hp = HAlloc(p, base_halloc_sz + BIG_UINT_HEAP_SIZE + 5);
+	hp = HAllocX(p, base_halloc_sz + BIG_UINT_HEAP_SIZE + 5,
+                     ERTS_MAGIC_REF_THING_SIZE);
 	egot = uint_to_big(got, hp);
 	hp += BIG_UINT_HEAP_SIZE;
     }
 
     if (is_first_trap) {
+        if (is_atom(tid))
+            tid = erts_db_make_tid(p, &tb->common);
         mpb = erts_db_make_match_prog_ref(p, *mpp, &hp);
         *mpp = NULL; /* otherwise the caller will destroy it */
     }
     else {
+        ASSERT(!is_atom(tid));
         mpb = prev_continuation_tptr[3];
     }
 
@@ -1527,16 +1501,16 @@ static ERTS_INLINE int on_mtraversal_simple_trap(Export* trap_function,
             make_small(slot_ix),
             mpb,
             egot);
-    *ret = bif_trap1(trap_function, p, continuation);
+    ERTS_BIF_PREP_TRAP1(*ret, trap_function, p, continuation);
     return DB_ERROR_NONE;
 }
 
-static ERTS_INLINE int unpack_simple_mtraversal_continuation(Eterm continuation,
-                                                             Eterm** tptr_ptr,
-                                                             Eterm* tid_ptr,
-                                                             Sint* slot_ix_p,
-                                                             Binary** mpp,
-                                                             Sint* got_p)
+static ERTS_INLINE int unpack_simple_continuation(Eterm continuation,
+                                                  Eterm** tptr_ptr,
+                                                  Eterm* tid_ptr,
+                                                  Sint* slot_ix_p,
+                                                  Binary** mpp,
+                                                  Sint* got_p)
 {
     Eterm* tptr;
     ASSERT(is_tuple(continuation));
@@ -1571,6 +1545,7 @@ static ERTS_INLINE int unpack_simple_mtraversal_continuation(Eterm continuation,
 #define MAX_SELECT_CHUNK_ITERATIONS 1000
 
 typedef struct {
+    match_callbacks_t base;
     Process* p;
     DbTableHash* tb;
     Eterm tid;
@@ -1578,77 +1553,86 @@ typedef struct {
     Sint chunk_size;
     Eterm match_list;
     Eterm* prev_continuation_tptr;
-} mtraversal_select_chunk_context_t;
+} select_chunk_context_t;
 
-static int mtraversal_select_chunk_on_nothing_can_match(void* context_ptr, Eterm* ret) {
-    mtraversal_select_chunk_context_t* sc_context_ptr = (mtraversal_select_chunk_context_t*) context_ptr;
-    *ret = (sc_context_ptr->chunk_size > 0 ? am_EOT : NIL);
+static int select_chunk_on_nothing_can_match(match_callbacks_t* ctx_base, Eterm* ret)
+{
+    select_chunk_context_t* ctx = (select_chunk_context_t*) ctx_base;
+    *ret = (ctx->chunk_size > 0 ? am_EOT : NIL);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_chunk_on_match_res(void* context_ptr, Sint slot_ix,
-                                                HashDbTerm*** current_ptr_ptr,
-                                                Eterm match_res)
+static int select_chunk_on_match_res(match_callbacks_t* ctx_base, Sint slot_ix,
+                                     HashDbTerm*** current_ptr_ptr,
+                                     Eterm match_res)
 {
-    mtraversal_select_chunk_context_t* sc_context_ptr = (mtraversal_select_chunk_context_t*) context_ptr;
+    select_chunk_context_t* ctx = (select_chunk_context_t*) ctx_base;
     if (is_value(match_res)) {
-        sc_context_ptr->match_list = CONS(sc_context_ptr->hp, match_res, sc_context_ptr->match_list);
+        ctx->match_list = CONS(ctx->hp, match_res, ctx->match_list);
         return 1;
     }
     return 0;
 }
 
-static int mtraversal_select_chunk_on_loop_ended(void* context_ptr, Sint slot_ix, Sint got,
-                                                 Sint iterations_left, Binary** mpp, Eterm* ret)
+static int select_chunk_on_loop_ended(match_callbacks_t* ctx_base,
+                                      Sint slot_ix, Sint got,
+                                      Sint iterations_left, Binary** mpp,
+                                      Eterm* ret)
 {
-    mtraversal_select_chunk_context_t* sc_context_ptr = (mtraversal_select_chunk_context_t*) context_ptr;
+    select_chunk_context_t* ctx = (select_chunk_context_t*) ctx_base;
     Eterm mpb;
 
     if (iterations_left == MAX_SELECT_CHUNK_ITERATIONS) {
         /* We didn't get to iterate a single time, which means EOT */
-        ASSERT(sc_context_ptr->match_list == NIL);
-        *ret = (sc_context_ptr->chunk_size > 0 ? am_EOT : NIL);
+        ASSERT(ctx->match_list == NIL);
+        *ret = (ctx->chunk_size > 0 ? am_EOT : NIL);
         return DB_ERROR_NONE;
     }
     else {
         ASSERT(iterations_left < MAX_SELECT_CHUNK_ITERATIONS);
-        BUMP_REDS(sc_context_ptr->p, MAX_SELECT_CHUNK_ITERATIONS - iterations_left);
-        if (sc_context_ptr->chunk_size) {
+        BUMP_REDS(ctx->p, MAX_SELECT_CHUNK_ITERATIONS - iterations_left);
+        if (ctx->chunk_size) {
             Eterm continuation;
             Eterm rest = NIL;
             Sint rest_size = 0;
 
-            if (got > sc_context_ptr->chunk_size) { /* Split list in return value and 'rest' */
-                Eterm tmp = sc_context_ptr->match_list;
-                rest = sc_context_ptr->match_list;
-                while (got-- > sc_context_ptr->chunk_size + 1) {
+            if (got > ctx->chunk_size) { /* Split list in return value and 'rest' */
+                Eterm tmp = ctx->match_list;
+                rest = ctx->match_list;
+                while (got-- > ctx->chunk_size + 1) {
                     tmp = CDR(list_val(tmp));
                     ++rest_size;
                 }
                 ++rest_size;
-                sc_context_ptr->match_list = CDR(list_val(tmp));
+                ctx->match_list = CDR(list_val(tmp));
                 CDR(list_val(tmp)) = NIL; /* Destructive, the list has never
                                              been in 'user space' */
             }
             if (rest != NIL || slot_ix >= 0) { /* Need more calls */
-                sc_context_ptr->hp = HAlloc(sc_context_ptr->p, 3 + 7 + ERTS_MAGIC_REF_THING_SIZE);
-                mpb = erts_db_make_match_prog_ref(sc_context_ptr->p, *mpp, &sc_context_ptr->hp);
+                Eterm tid = ctx->tid;
+                ctx->hp = HAllocX(ctx->p,
+                                  3 + 7 + ERTS_MAGIC_REF_THING_SIZE,
+                                  ERTS_MAGIC_REF_THING_SIZE);
+                mpb = erts_db_make_match_prog_ref(ctx->p, *mpp, &ctx->hp);
+                if (is_atom(tid))
+                    tid = erts_db_make_tid(ctx->p,
+                                           &ctx->tb->common);
                 continuation = TUPLE6(
-                        sc_context_ptr->hp,
-                        sc_context_ptr->tid,
+                        ctx->hp,
+                        tid,
                         make_small(slot_ix),
-                        make_small(sc_context_ptr->chunk_size),
+                        make_small(ctx->chunk_size),
                         mpb, rest,
                         make_small(rest_size));
                 *mpp = NULL; /* Otherwise the caller will destroy it */
-                sc_context_ptr->hp += 7;
-                *ret = TUPLE2(sc_context_ptr->hp, sc_context_ptr->match_list, continuation);
+                ctx->hp += 7;
+                *ret = TUPLE2(ctx->hp, ctx->match_list, continuation);
                 return DB_ERROR_NONE;
             } else { /* All data is exhausted */
-                if (sc_context_ptr->match_list != NIL) { /* No more data to search but still a
+                if (ctx->match_list != NIL) { /* No more data to search but still a
                                                             result to return to the caller */
-                    sc_context_ptr->hp = HAlloc(sc_context_ptr->p, 3);
-                    *ret = TUPLE2(sc_context_ptr->hp, sc_context_ptr->match_list, am_EOT);
+                    ctx->hp = HAlloc(ctx->p, 3);
+                    *ret = TUPLE2(ctx->hp, ctx->match_list, am_EOT);
                     return DB_ERROR_NONE;
                 } else { /* Reached the end of the ttable with no data to return */
                     *ret = am_EOT;
@@ -1656,78 +1640,88 @@ static int mtraversal_select_chunk_on_loop_ended(void* context_ptr, Sint slot_ix
                 }
             }
         }
-        *ret = sc_context_ptr->match_list;
+        *ret = ctx->match_list;
         return DB_ERROR_NONE;
     }
 }
 
-static int mtraversal_select_chunk_on_trap(void* context_ptr, Sint slot_ix, Sint got,
-                                           Binary** mpp, Eterm* ret)
+static int select_chunk_on_trap(match_callbacks_t* ctx_base,
+                                Sint slot_ix, Sint got,
+                                Binary** mpp, Eterm* ret)
 {
-    mtraversal_select_chunk_context_t* sc_context_ptr = (mtraversal_select_chunk_context_t*) context_ptr;
+    select_chunk_context_t* ctx = (select_chunk_context_t*) ctx_base;
     Eterm mpb;
     Eterm continuation;
     Eterm* hp;
 
-    BUMP_ALL_REDS(sc_context_ptr->p);
+    BUMP_ALL_REDS(ctx->p);
 
-    if (sc_context_ptr->prev_continuation_tptr == NULL) {
+    if (ctx->prev_continuation_tptr == NULL) {
+        Eterm tid = ctx->tid;
         /* First time we're trapping */
-        hp = HAlloc(sc_context_ptr->p, 7 + ERTS_MAGIC_REF_THING_SIZE);
-        mpb = erts_db_make_match_prog_ref(sc_context_ptr->p, *mpp, &hp);
+        hp = HAllocX(ctx->p, 7 + ERTS_MAGIC_REF_THING_SIZE,
+                     ERTS_MAGIC_REF_THING_SIZE);
+        if (is_atom(tid))
+            tid = erts_db_make_tid(ctx->p, &ctx->tb->common);
+        mpb = erts_db_make_match_prog_ref(ctx->p, *mpp, &hp);
         continuation = TUPLE6(
                 hp,
-                sc_context_ptr->tid,
+                tid,
                 make_small(slot_ix),
-                make_small(sc_context_ptr->chunk_size),
+                make_small(ctx->chunk_size),
                 mpb,
-                sc_context_ptr->match_list,
+                ctx->match_list,
                 make_small(got));
         *mpp = NULL; /* otherwise the caller will destroy it */
     }
     else {
         /* Not the first time we're trapping; reuse continuation terms */
-        hp = HAlloc(sc_context_ptr->p, 7);
+        hp = HAlloc(ctx->p, 7);
         continuation = TUPLE6(
                 hp,
-                sc_context_ptr->prev_continuation_tptr[1],
+                ctx->prev_continuation_tptr[1],
                 make_small(slot_ix),
-                sc_context_ptr->prev_continuation_tptr[3],
-                sc_context_ptr->prev_continuation_tptr[4],
-                sc_context_ptr->match_list,
+                ctx->prev_continuation_tptr[3],
+                ctx->prev_continuation_tptr[4],
+                ctx->match_list,
                 make_small(got));
     }
-    *ret = bif_trap1(&ets_select_continue_exp, sc_context_ptr->p, continuation);
+    ERTS_BIF_PREP_TRAP1(*ret, &ets_select_continue_exp, ctx->p,
+                        continuation);
     return DB_ERROR_NONE;
 }
 
-static int db_select_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern, int reverse, Eterm *ret) {
+static int db_select_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern,
+                          int reverse, Eterm *ret)
+{
     return db_select_chunk_hash(p, tbl, tid, pattern, 0, reverse, ret);
 }
 
-static int db_select_chunk_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern, Sint chunk_size,
+static int db_select_chunk_hash(Process *p, DbTable *tbl, Eterm tid,
+                                Eterm pattern, Sint chunk_size,
                                 int reverse, Eterm *ret)
 {
-    mtraversal_select_chunk_context_t sc_context;
-    sc_context.p = p;
-    sc_context.tb = &tbl->hash;
-    sc_context.tid = tid;
-    sc_context.hp = NULL;
-    sc_context.chunk_size = chunk_size;
-    sc_context.match_list = NIL;
-    sc_context.prev_continuation_tptr = NULL;
+    select_chunk_context_t ctx;
+
+    ctx.base.on_nothing_can_match = select_chunk_on_nothing_can_match;
+    ctx.base.on_match_res         = select_chunk_on_match_res;
+    ctx.base.on_loop_ended        = select_chunk_on_loop_ended;
+    ctx.base.on_trap              = select_chunk_on_trap,
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.hp = NULL;
+    ctx.chunk_size = chunk_size;
+    ctx.match_list = NIL;
+    ctx.prev_continuation_tptr = NULL;
 
     return match_traverse(
-            sc_context.p, sc_context.tb,
+            ctx.p, ctx.tb,
             pattern, NULL,
-            sc_context.chunk_size,
+            ctx.chunk_size,
             MAX_SELECT_CHUNK_ITERATIONS,
-            &sc_context.hp, 0,
-            mtraversal_select_chunk_on_nothing_can_match,
-            mtraversal_select_chunk_on_match_res,
-            mtraversal_select_chunk_on_loop_ended,
-            mtraversal_select_chunk_on_trap,
-            &sc_context, ret);
+            &ctx.hp, 0,
+            &ctx.base, ret);
 }
 
 /*
@@ -1736,47 +1730,50 @@ static int db_select_chunk_hash(Process *p, DbTable *tbl, Eterm tid, Eterm patte
  *
  */
 
-static int mtraversal_select_chunk_continue_on_loop_ended(void* context_ptr, Sint slot_ix, Sint got,
-                                                          Sint iterations_left, Binary** mpp, Eterm* ret)
+static
+int select_chunk_continue_on_loop_ended(match_callbacks_t* ctx_base,
+                                        Sint slot_ix, Sint got,
+                                        Sint iterations_left, Binary** mpp,
+                                        Eterm* ret)
 {
-    mtraversal_select_chunk_context_t* sc_context_ptr = (mtraversal_select_chunk_context_t*) context_ptr;
+    select_chunk_context_t* ctx = (select_chunk_context_t*) ctx_base;
     Eterm continuation;
     Eterm rest = NIL;
     Eterm* hp;
 
     ASSERT(iterations_left <= MAX_SELECT_CHUNK_ITERATIONS);
-    BUMP_REDS(sc_context_ptr->p, MAX_SELECT_CHUNK_ITERATIONS - iterations_left);
-    if (sc_context_ptr->chunk_size) {
+    BUMP_REDS(ctx->p, MAX_SELECT_CHUNK_ITERATIONS - iterations_left);
+    if (ctx->chunk_size) {
         Sint rest_size = 0;
-        if (got > sc_context_ptr->chunk_size) {
+        if (got > ctx->chunk_size) {
             /* Cannot write destructively here,
                the list may have
                been in user space */
-            hp = HAlloc(sc_context_ptr->p, (got - sc_context_ptr->chunk_size) * 2);
-            while (got-- > sc_context_ptr->chunk_size) {
-                rest = CONS(hp, CAR(list_val(sc_context_ptr->match_list)), rest);
+            hp = HAlloc(ctx->p, (got - ctx->chunk_size) * 2);
+            while (got-- > ctx->chunk_size) {
+                rest = CONS(hp, CAR(list_val(ctx->match_list)), rest);
                 hp += 2;
-                sc_context_ptr->match_list = CDR(list_val(sc_context_ptr->match_list));
+                ctx->match_list = CDR(list_val(ctx->match_list));
                 ++rest_size;
             }
         }
         if (rest != NIL || slot_ix >= 0) {
-            hp = HAlloc(sc_context_ptr->p, 3 + 7);
+            hp = HAlloc(ctx->p, 3 + 7);
             continuation = TUPLE6(
                     hp,
-                    sc_context_ptr->prev_continuation_tptr[1],
+                    ctx->prev_continuation_tptr[1],
                     make_small(slot_ix),
-                    sc_context_ptr->prev_continuation_tptr[3],
-                    sc_context_ptr->prev_continuation_tptr[4],
+                    ctx->prev_continuation_tptr[3],
+                    ctx->prev_continuation_tptr[4],
                     rest,
                     make_small(rest_size));
             hp += 7;
-            *ret = TUPLE2(hp, sc_context_ptr->match_list, continuation);
+            *ret = TUPLE2(hp, ctx->match_list, continuation);
             return DB_ERROR_NONE;
         } else {
-            if (sc_context_ptr->match_list != NIL) {
-                hp = HAlloc(sc_context_ptr->p, 3);
-                *ret = TUPLE2(hp, sc_context_ptr->match_list, am_EOT);
+            if (ctx->match_list != NIL) {
+                hp = HAlloc(ctx->p, 3);
+                *ret = TUPLE2(hp, ctx->match_list, am_EOT);
                 return DB_ERROR_NONE;
             } else {
                 *ret = am_EOT;
@@ -1784,15 +1781,17 @@ static int mtraversal_select_chunk_continue_on_loop_ended(void* context_ptr, Sin
             }
         }
     }
-    *ret = sc_context_ptr->match_list;
+    *ret = ctx->match_list;
     return DB_ERROR_NONE;
 }
 
 /*
  * This is called when select traps
  */
-static int db_select_continue_hash(Process* p, DbTable* tbl, Eterm continuation, Eterm* ret) {
-    mtraversal_select_chunk_context_t sc_context = {0};
+static int db_select_continue_hash(Process* p, DbTable* tbl, Eterm continuation,
+                                   Eterm* ret)
+{
+    select_chunk_context_t ctx;
     Eterm* tptr;
     Eterm tid;
     Binary* mp;
@@ -1827,21 +1826,21 @@ static int db_select_continue_hash(Process* p, DbTable* tbl, Eterm continuation,
     match_list = tptr[5];
 
     /* Proceed */
-    sc_context.p = p;
-    sc_context.tb = &tbl->hash;
-    sc_context.tid = tid;
-    sc_context.hp = NULL;
-    sc_context.chunk_size = chunk_size;
-    sc_context.match_list = match_list;
-    sc_context.prev_continuation_tptr = tptr;
+    ctx.base.on_match_res  = select_chunk_on_match_res;
+    ctx.base.on_loop_ended = select_chunk_continue_on_loop_ended;
+    ctx.base.on_trap       = select_chunk_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.hp = NULL;
+    ctx.chunk_size = chunk_size;
+    ctx.match_list = match_list;
+    ctx.prev_continuation_tptr = tptr;
 
     return match_traverse_continue(
-            sc_context.p, sc_context.tb, sc_context.chunk_size,
-            iterations_left, &sc_context.hp, slot_ix, got, &mp, 0,
-            mtraversal_select_chunk_on_match_res, /* Reuse callback */
-            mtraversal_select_chunk_continue_on_loop_ended,
-            mtraversal_select_chunk_on_trap,      /* Reuse callback */
-            &sc_context, ret);
+            ctx.p, ctx.tb, ctx.chunk_size,
+            iterations_left, &ctx.hp, slot_ix, got, &mp, 0,
+            &ctx.base, ret);
 
 badparam:
     *ret = NIL;
@@ -1860,75 +1859,83 @@ badparam:
 #define MAX_SELECT_COUNT_ITERATIONS 1000
 
 typedef struct {
+    match_callbacks_t base;
     Process* p;
     DbTableHash* tb;
     Eterm tid;
-    Eterm* hp;
     Eterm* prev_continuation_tptr;
-} mtraversal_select_count_context_t;
+} select_count_context_t;
 
-static int mtraversal_select_count_on_nothing_can_match(void* context_ptr, Eterm* ret) {
+static int select_count_on_nothing_can_match(match_callbacks_t* ctx_base,
+                                             Eterm* ret)
+{
     *ret = make_small(0);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_count_on_match_res(void* context_ptr, Sint slot_ix,
-                                                HashDbTerm*** current_ptr_ptr,
-                                                Eterm match_res)
+static int select_count_on_match_res(match_callbacks_t* ctx_base, Sint slot_ix,
+                                     HashDbTerm*** current_ptr_ptr,
+                                     Eterm match_res)
 {
     return (match_res == am_true);
 }
 
-static int mtraversal_select_count_on_loop_ended(void* context_ptr, Sint slot_ix, Sint got,
-                                                 Sint iterations_left, Binary** mpp, Eterm* ret)
+static int select_count_on_loop_ended(match_callbacks_t* ctx_base,
+                                      Sint slot_ix, Sint got,
+                                      Sint iterations_left, Binary** mpp,
+                                      Eterm* ret)
 {
-    mtraversal_select_count_context_t* scnt_context_ptr = (mtraversal_select_count_context_t*) context_ptr;
+    select_count_context_t* ctx = (select_count_context_t*) ctx_base;
     ASSERT(iterations_left <= MAX_SELECT_COUNT_ITERATIONS);
-    BUMP_REDS(scnt_context_ptr->p, MAX_SELECT_COUNT_ITERATIONS - iterations_left);
-    *ret = erts_make_integer(got, scnt_context_ptr->p);
+    BUMP_REDS(ctx->p, MAX_SELECT_COUNT_ITERATIONS - iterations_left);
+    *ret = erts_make_integer(got, ctx->p);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_count_on_trap(void* context_ptr, Sint slot_ix, Sint got,
-                                           Binary** mpp, Eterm* ret)
+static int select_count_on_trap(match_callbacks_t* ctx_base,
+                                Sint slot_ix, Sint got,
+                                Binary** mpp, Eterm* ret)
 {
-    mtraversal_select_count_context_t* scnt_context_ptr = (mtraversal_select_count_context_t*) context_ptr;
-    return on_mtraversal_simple_trap(
+    select_count_context_t* ctx = (select_count_context_t*) ctx_base;
+    return on_simple_trap(
             &ets_select_count_continue_exp,
-            scnt_context_ptr->p,
-            scnt_context_ptr->tb,
-            scnt_context_ptr->tid,
-            scnt_context_ptr->prev_continuation_tptr,
+            ctx->p,
+            ctx->tb,
+            ctx->tid,
+            ctx->prev_continuation_tptr,
             slot_ix, got, mpp, ret);
 }
 
-static int db_select_count_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern, Eterm *ret) {
-    mtraversal_select_count_context_t scnt_context = {0};
+static int db_select_count_hash(Process *p, DbTable *tbl, Eterm tid,
+                                Eterm pattern, Eterm *ret)
+{
+    select_count_context_t ctx;
     Sint iterations_left = MAX_SELECT_COUNT_ITERATIONS;
     Sint chunk_size = 0;
 
-    scnt_context.p = p;
-    scnt_context.tb = &tbl->hash;
-    scnt_context.tid = tid;
-    scnt_context.hp = NULL;
-    scnt_context.prev_continuation_tptr = NULL;
+    ctx.base.on_nothing_can_match = select_count_on_nothing_can_match;
+    ctx.base.on_match_res         = select_count_on_match_res;
+    ctx.base.on_loop_ended        = select_count_on_loop_ended;
+    ctx.base.on_trap              = select_count_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = NULL;
 
     return match_traverse(
-            scnt_context.p, scnt_context.tb,
+            ctx.p, ctx.tb,
             pattern, NULL,
             chunk_size, iterations_left, NULL, 0,
-            mtraversal_select_count_on_nothing_can_match,
-            mtraversal_select_count_on_match_res,
-            mtraversal_select_count_on_loop_ended,
-            mtraversal_select_count_on_trap,
-            &scnt_context, ret);
+            &ctx.base, ret);
 }
 
 /*
  * This is called when select_count traps
  */
-static int db_select_count_continue_hash(Process* p, DbTable* tbl, Eterm continuation, Eterm* ret) {
-    mtraversal_select_count_context_t scnt_context = {0};
+static int db_select_count_continue_hash(Process* p, DbTable* tbl,
+                                         Eterm continuation, Eterm* ret)
+{
+    select_count_context_t ctx;
     Eterm* tptr;
     Eterm tid;
     Binary* mp;
@@ -1937,25 +1944,24 @@ static int db_select_count_continue_hash(Process* p, DbTable* tbl, Eterm continu
     Sint chunk_size = 0;
     *ret = NIL;
 
-    if (unpack_simple_mtraversal_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
+    if (unpack_simple_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
         *ret = NIL;
         return DB_ERROR_BADPARAM;
     }
 
-    scnt_context.p = p;
-    scnt_context.tb = &tbl->hash;
-    scnt_context.tid = tid;
-    scnt_context.hp = NULL;
-    scnt_context.prev_continuation_tptr = tptr;
+    ctx.base.on_match_res  = select_count_on_match_res;
+    ctx.base.on_loop_ended = select_count_on_loop_ended;
+    ctx.base.on_trap       = select_count_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = tptr;
 
     return match_traverse_continue(
-            scnt_context.p, scnt_context.tb, chunk_size,
+            ctx.p, ctx.tb, chunk_size,
             MAX_SELECT_COUNT_ITERATIONS,
             NULL, slot_ix, got, &mp, 0,
-            mtraversal_select_count_on_match_res,  /* Reuse callback */
-            mtraversal_select_count_on_loop_ended, /* Reuse callback */
-            mtraversal_select_count_on_trap,       /* Reuse callback */
-            &scnt_context, ret);
+            &ctx.base, ret);
 }
 
 #undef MAX_SELECT_COUNT_ITERATIONS
@@ -1970,108 +1976,119 @@ static int db_select_count_continue_hash(Process* p, DbTable* tbl, Eterm continu
 #define MAX_SELECT_DELETE_ITERATIONS 1000
 
 typedef struct {
+    match_callbacks_t base;
     Process* p;
     DbTableHash* tb;
     Eterm tid;
-    Eterm* hp;
     Eterm* prev_continuation_tptr;
     erts_aint_t fixated_by_me;
     Uint last_pseudo_delete;
-} mtraversal_select_delete_context_t;
+    HashDbTerm* free_us;
+} select_delete_context_t;
 
-static int mtraversal_select_delete_on_nothing_can_match(void* context_ptr, Eterm* ret) {
+static int select_delete_on_nothing_can_match(match_callbacks_t* ctx_base,
+                                              Eterm* ret)
+{
     *ret = make_small(0);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_delete_on_match_res(void* context_ptr, Sint slot_ix,
-                                                 HashDbTerm*** current_ptr_ptr,
-                                                 Eterm match_res)
+static int select_delete_on_match_res(match_callbacks_t* ctx_base, Sint slot_ix,
+                                      HashDbTerm*** current_ptr_ptr,
+                                      Eterm match_res)
 {
     HashDbTerm** current_ptr = *current_ptr_ptr;
-    mtraversal_select_delete_context_t* sd_context_ptr = (mtraversal_select_delete_context_t*) context_ptr;
+    select_delete_context_t* ctx = (select_delete_context_t*) ctx_base;
     HashDbTerm* del;
     if (match_res != am_true)
         return 0;
 
-    if (NFIXED(sd_context_ptr->tb) > sd_context_ptr->fixated_by_me) { /* fixated by others? */
-        if (slot_ix != sd_context_ptr->last_pseudo_delete) {
-            if (!add_fixed_deletion(sd_context_ptr->tb, slot_ix, sd_context_ptr->fixated_by_me))
+    if (NFIXED(ctx->tb) > ctx->fixated_by_me) { /* fixated by others? */
+        if (slot_ix != ctx->last_pseudo_delete) {
+            if (!add_fixed_deletion(ctx->tb, slot_ix, ctx->fixated_by_me))
                 goto do_erase;
-            sd_context_ptr->last_pseudo_delete = slot_ix;
+            ctx->last_pseudo_delete = slot_ix;
         }
-        (*current_ptr)->hvalue = INVALID_HASH;
+        (*current_ptr)->pseudo_deleted = 1;
     }
     else {
     do_erase:
         del = *current_ptr;
         *current_ptr = (*current_ptr)->next; // replace pointer to term using next
-        free_term(sd_context_ptr->tb, del);
+        del->next = ctx->free_us;
+        ctx->free_us = del;
     }
-    erts_smp_atomic_dec_nob(&sd_context_ptr->tb->common.nitems);
+    erts_atomic_dec_nob(&ctx->tb->common.nitems);
 
     return 1;
 }
 
-static int mtraversal_select_delete_on_loop_ended(void* context_ptr, Sint slot_ix, Sint got,
-                                                  Sint iterations_left, Binary** mpp, Eterm* ret)
+static int select_delete_on_loop_ended(match_callbacks_t* ctx_base,
+                                       Sint slot_ix, Sint got,
+                                       Sint iterations_left, Binary** mpp,
+                                       Eterm* ret)
 {
-    mtraversal_select_delete_context_t* sd_context_ptr = (mtraversal_select_delete_context_t*) context_ptr;
+    select_delete_context_t* ctx = (select_delete_context_t*) ctx_base;
+    free_term_list(ctx->tb, ctx->free_us);
+    ctx->free_us = NULL;
     ASSERT(iterations_left <= MAX_SELECT_DELETE_ITERATIONS);
-    BUMP_REDS(sd_context_ptr->p, MAX_SELECT_DELETE_ITERATIONS - iterations_left);
+    BUMP_REDS(ctx->p, MAX_SELECT_DELETE_ITERATIONS - iterations_left);
     if (got) {
-	try_shrink(sd_context_ptr->tb);
+	try_shrink(ctx->tb);
     }
-    *ret = erts_make_integer(got, sd_context_ptr->p);
+    *ret = erts_make_integer(got, ctx->p);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_delete_on_trap(void* context_ptr, Sint slot_ix, Sint got,
-                                            Binary** mpp, Eterm* ret)
+static int select_delete_on_trap(match_callbacks_t* ctx_base,
+                                 Sint slot_ix, Sint got,
+                                 Binary** mpp, Eterm* ret)
 {
-    mtraversal_select_delete_context_t* sd_context_ptr = (mtraversal_select_delete_context_t*) context_ptr;
-    return on_mtraversal_simple_trap(
+    select_delete_context_t* ctx = (select_delete_context_t*) ctx_base;
+    free_term_list(ctx->tb, ctx->free_us);
+    ctx->free_us = NULL;
+    return on_simple_trap(
             &ets_select_delete_continue_exp,
-            sd_context_ptr->p,
-            sd_context_ptr->tb,
-            sd_context_ptr->tid,
-            sd_context_ptr->prev_continuation_tptr,
+            ctx->p,
+            ctx->tb,
+            ctx->tid,
+            ctx->prev_continuation_tptr,
             slot_ix, got, mpp, ret);
 }
 
-static int db_select_delete_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern, Eterm *ret) {
-    mtraversal_select_delete_context_t sd_context = {0};
+static int db_select_delete_hash(Process *p, DbTable *tbl, Eterm tid,
+                                 Eterm pattern, Eterm *ret)
+{
+    select_delete_context_t ctx;
     Sint chunk_size = 0;
 
-    sd_context.p = p;
-    sd_context.tb = &tbl->hash;
-    sd_context.tid = tid;
-    sd_context.hp = NULL;
-    sd_context.prev_continuation_tptr = NULL;
-#ifdef ERTS_SMP
-    sd_context.fixated_by_me = sd_context.tb->common.is_thread_safe ? 0 : 1; /* TODO: something nicer */
-#else
-    sd_context.fixated_by_me = 0;
-#endif
-    sd_context.last_pseudo_delete = (Uint) -1;
+    ctx.base.on_nothing_can_match = select_delete_on_nothing_can_match;
+    ctx.base.on_match_res         = select_delete_on_match_res;
+    ctx.base.on_loop_ended        = select_delete_on_loop_ended;
+    ctx.base.on_trap              = select_delete_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = NULL;
+    ctx.fixated_by_me = ctx.tb->common.is_thread_safe ? 0 : 1; /* TODO: something nicer */
+    ctx.last_pseudo_delete = (Uint) -1;
+    ctx.free_us = NULL;
 
     return match_traverse(
-            sd_context.p, sd_context.tb,
+            ctx.p, ctx.tb,
             pattern, NULL,
             chunk_size,
             MAX_SELECT_DELETE_ITERATIONS, NULL, 1,
-            mtraversal_select_delete_on_nothing_can_match,
-            mtraversal_select_delete_on_match_res,
-            mtraversal_select_delete_on_loop_ended,
-            mtraversal_select_delete_on_trap,
-            &sd_context, ret);
+            &ctx.base, ret);
 }
 
 /*
  * This is called when select_delete traps
  */
-static int db_select_delete_continue_hash(Process* p, DbTable* tbl, Eterm continuation, Eterm* ret) {
-    mtraversal_select_delete_context_t sd_context = {0};
+static int db_select_delete_continue_hash(Process* p, DbTable* tbl,
+                                          Eterm continuation, Eterm* ret)
+{
+    select_delete_context_t ctx;
     Eterm* tptr;
     Eterm tid;
     Binary* mp;
@@ -2079,27 +2096,27 @@ static int db_select_delete_continue_hash(Process* p, DbTable* tbl, Eterm contin
     Sint slot_ix;
     Sint chunk_size = 0;
 
-    if (unpack_simple_mtraversal_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
+    if (unpack_simple_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
         *ret = NIL;
         return DB_ERROR_BADPARAM;
     }
 
-    sd_context.p = p;
-    sd_context.tb = &tbl->hash;
-    sd_context.tid = tid;
-    sd_context.hp = NULL;
-    sd_context.prev_continuation_tptr = tptr;
-    sd_context.fixated_by_me = ONLY_WRITER(p, sd_context.tb) ? 0 : 1; /* TODO: something nicer */
-    sd_context.last_pseudo_delete = (Uint) -1;
+    ctx.base.on_match_res  = select_delete_on_match_res;
+    ctx.base.on_loop_ended = select_delete_on_loop_ended;
+    ctx.base.on_trap       = select_delete_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = tptr;
+    ctx.fixated_by_me = ONLY_WRITER(p, ctx.tb) ? 0 : 1; /* TODO: something nicer */
+    ctx.last_pseudo_delete = (Uint) -1;
+    ctx.free_us = NULL;
 
     return match_traverse_continue(
-            sd_context.p, sd_context.tb, chunk_size,
+            ctx.p, ctx.tb, chunk_size,
             MAX_SELECT_DELETE_ITERATIONS,
             NULL, slot_ix, got, &mp, 1,
-            mtraversal_select_delete_on_match_res,  /* Reuse callback */
-            mtraversal_select_delete_on_loop_ended, /* Reuse callback */
-            mtraversal_select_delete_on_trap,       /* Reuse callback */
-            &sd_context, ret);
+            &ctx.base, ret);
 }
 
 #undef MAX_SELECT_DELETE_ITERATIONS
@@ -2114,24 +2131,26 @@ static int db_select_delete_continue_hash(Process* p, DbTable* tbl, Eterm contin
 #define MAX_SELECT_REPLACE_ITERATIONS 1000
 
 typedef struct {
+    match_callbacks_t base;
     Process* p;
     DbTableHash* tb;
     Eterm tid;
-    Eterm* hp;
     Eterm* prev_continuation_tptr;
-} mtraversal_select_replace_context_t;
+} select_replace_context_t;
 
-static int mtraversal_select_replace_on_nothing_can_match(void* context_ptr, Eterm* ret) {
+static int select_replace_on_nothing_can_match(match_callbacks_t* ctx_base,
+                                               Eterm* ret)
+{
     *ret = make_small(0);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_replace_on_match_res(void* context_ptr, Sint slot_ix,
-                                                  HashDbTerm*** current_ptr_ptr,
-                                                  Eterm match_res)
+static int select_replace_on_match_res(match_callbacks_t* ctx_base, Sint slot_ix,
+                                       HashDbTerm*** current_ptr_ptr,
+                                       Eterm match_res)
 {
-    mtraversal_select_replace_context_t* sr_context_ptr = (mtraversal_select_replace_context_t*) context_ptr;
-    DbTableHash* tb = sr_context_ptr->tb;
+    select_replace_context_t* ctx = (select_replace_context_t*) ctx_base;
+    DbTableHash* tb = ctx->tb;
     HashDbTerm* new;
     HashDbTerm* next;
     HashValue hval;
@@ -2147,6 +2166,7 @@ static int mtraversal_select_replace_on_match_res(void* context_ptr, Sint slot_i
         new = new_dbterm(tb, match_res);
         new->next = next;
         new->hvalue = hval;
+        new->pseudo_deleted = 0;
         free_term(tb, **current_ptr_ptr);
         **current_ptr_ptr = new; /* replace 'next' pointer in previous object */
         *current_ptr_ptr = &((**current_ptr_ptr)->next); /* advance to next object */
@@ -2155,35 +2175,37 @@ static int mtraversal_select_replace_on_match_res(void* context_ptr, Sint slot_i
     return 0;
 }
 
-static int mtraversal_select_replace_on_loop_ended(void* context_ptr, Sint slot_ix, Sint got,
-                                                   Sint iterations_left, Binary** mpp, Eterm* ret)
+static int select_replace_on_loop_ended(match_callbacks_t* ctx_base, Sint slot_ix,
+                                        Sint got, Sint iterations_left,
+                                        Binary** mpp, Eterm* ret)
 {
-    mtraversal_select_replace_context_t* sr_context_ptr = (mtraversal_select_replace_context_t*) context_ptr;
+    select_replace_context_t* ctx = (select_replace_context_t*) ctx_base;
     ASSERT(iterations_left <= MAX_SELECT_REPLACE_ITERATIONS);
     /* the more objects we've replaced, the more reductions we've consumed */
-    BUMP_REDS(sr_context_ptr->p,
+    BUMP_REDS(ctx->p,
               MIN(MAX_SELECT_REPLACE_ITERATIONS * 2,
                   (MAX_SELECT_REPLACE_ITERATIONS - iterations_left) + (int)got));
-    *ret = erts_make_integer(got, sr_context_ptr->p);
+    *ret = erts_make_integer(got, ctx->p);
     return DB_ERROR_NONE;
 }
 
-static int mtraversal_select_replace_on_trap(void* context_ptr, Sint slot_ix, Sint got,
-                                             Binary** mpp, Eterm* ret)
+static int select_replace_on_trap(match_callbacks_t* ctx_base,
+                                  Sint slot_ix, Sint got,
+                                  Binary** mpp, Eterm* ret)
 {
-    mtraversal_select_replace_context_t* sr_context_ptr = (mtraversal_select_replace_context_t*) context_ptr;
-    return on_mtraversal_simple_trap(
+    select_replace_context_t* ctx = (select_replace_context_t*) ctx_base;
+    return on_simple_trap(
             &ets_select_replace_continue_exp,
-            sr_context_ptr->p,
-            sr_context_ptr->tb,
-            sr_context_ptr->tid,
-            sr_context_ptr->prev_continuation_tptr,
+            ctx->p,
+            ctx->tb,
+            ctx->tid,
+            ctx->prev_continuation_tptr,
             slot_ix, got, mpp, ret);
 }
 
 static int db_select_replace_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pattern, Eterm *ret)
 {
-    mtraversal_select_replace_context_t sr_context = {0};
+    select_replace_context_t ctx;
     Sint chunk_size = 0;
 
     /* Bag implementation presented both semantic consistency and performance issues,
@@ -2191,22 +2213,21 @@ static int db_select_replace_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pat
      */
     ASSERT(!(tbl->hash.common.status & DB_BAG));
 
-    sr_context.p = p;
-    sr_context.tb = &tbl->hash;
-    sr_context.tid = tid;
-    sr_context.hp = NULL;
-    sr_context.prev_continuation_tptr = NULL;
+    ctx.base.on_nothing_can_match = select_replace_on_nothing_can_match;
+    ctx.base.on_match_res         = select_replace_on_match_res;
+    ctx.base.on_loop_ended        = select_replace_on_loop_ended;
+    ctx.base.on_trap              = select_replace_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = NULL;
 
     return match_traverse(
-            sr_context.p, sr_context.tb,
+            ctx.p, ctx.tb,
             pattern, db_match_keeps_key,
             chunk_size,
             MAX_SELECT_REPLACE_ITERATIONS, NULL, 1,
-            mtraversal_select_replace_on_nothing_can_match,
-            mtraversal_select_replace_on_match_res,
-            mtraversal_select_replace_on_loop_ended,
-            mtraversal_select_replace_on_trap,
-            &sr_context, ret);
+            &ctx.base, ret);
 }
 
 /*
@@ -2214,7 +2235,7 @@ static int db_select_replace_hash(Process *p, DbTable *tbl, Eterm tid, Eterm pat
  */
 static int db_select_replace_continue_hash(Process* p, DbTable* tbl, Eterm continuation, Eterm* ret)
 {
-    mtraversal_select_replace_context_t sr_context = {0};
+    select_replace_context_t ctx;
     Eterm* tptr;
     Eterm tid ;
     Binary* mp;
@@ -2223,26 +2244,25 @@ static int db_select_replace_continue_hash(Process* p, DbTable* tbl, Eterm conti
     Sint chunk_size = 0;
     *ret = NIL;
 
-    if (unpack_simple_mtraversal_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
+    if (unpack_simple_continuation(continuation, &tptr, &tid, &slot_ix, &mp, &got)) {
         *ret = NIL;
         return DB_ERROR_BADPARAM;
     }
 
     /* Proceed */
-    sr_context.p = p;
-    sr_context.tb = &tbl->hash;
-    sr_context.tid = tid;
-    sr_context.hp = NULL;
-    sr_context.prev_continuation_tptr = tptr;
+    ctx.base.on_match_res  = select_replace_on_match_res;
+    ctx.base.on_loop_ended = select_replace_on_loop_ended;
+    ctx.base.on_trap       = select_replace_on_trap;
+    ctx.p = p;
+    ctx.tb = &tbl->hash;
+    ctx.tid = tid;
+    ctx.prev_continuation_tptr = tptr;
 
     return match_traverse_continue(
-            sr_context.p, sr_context.tb, chunk_size,
+            ctx.p, ctx.tb, chunk_size,
             MAX_SELECT_REPLACE_ITERATIONS,
             NULL, slot_ix, got, &mp, 1,
-            mtraversal_select_replace_on_match_res,  /* Reuse callback */
-            mtraversal_select_replace_on_loop_ended, /* Reuse callback */
-            mtraversal_select_replace_on_trap,       /* Reuse callback */
-            &sr_context, ret);
+            &ctx.base, ret);
 }
 
 
@@ -2250,8 +2270,9 @@ static int db_take_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
 {
     DbTableHash *tb = &tbl->hash;
     HashDbTerm **bp, *b;
+    HashDbTerm *free_us = NULL;
     HashValue hval = MAKE_HASH(key);
-    erts_smp_rwmtx_t *lck = WLOCK_HASH(tb, hval);
+    erts_rwmtx_t *lck = WLOCK_HASH(tb, hval);
     int ix = hash_to_ix(tb, hval);
     int nitems_diff = 0;
 
@@ -2267,12 +2288,13 @@ static int db_take_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
                     && add_fixed_deletion(tb, ix, 0)) {
                     /* Pseudo remove (no need to keep several of same key) */
                     bp = &b->next;
-                    b->hvalue = INVALID_HASH;
+                    b->pseudo_deleted = 1;
                     b = b->next;
                 } else {
-                    *bp = b->next;
-                    free_term(tb, b);
-                    b = *bp;
+                    HashDbTerm* next = b->next;
+                    b->next = free_us;
+                    free_us = b;
+                    b = *bp = next;
                 }
             }
             break;
@@ -2280,9 +2302,10 @@ static int db_take_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret)
     }
     WUNLOCK_HASH(lck);
     if (nitems_diff) {
-        erts_smp_atomic_add_nob(&tb->common.nitems, nitems_diff);
+        erts_atomic_add_nob(&tb->common.nitems, nitems_diff);
         try_shrink(tb);
     }
+    free_term_list(tb, free_us);
     return DB_ERROR_NONE;
 }
 
@@ -2296,25 +2319,52 @@ void db_initialize_hash(void)
 }
 
 
-int db_mark_all_deleted_hash(DbTable *tbl)
+static SWord db_mark_all_deleted_hash(DbTable *tbl, SWord reds)
 {
+    const int LOOPS_PER_REDUCTION  = 8;
     DbTableHash *tb = &tbl->hash;
-    HashDbTerm* list;
+    FixedDeletion* fixdel;
+    SWord loops = reds * LOOPS_PER_REDUCTION;
     int i;
 
-    ERTS_SMP_LC_ASSERT(IS_TAB_WLOCKED(tb));
+    ERTS_LC_ASSERT(IS_TAB_WLOCKED(tb));
 
-    for (i = 0; i < NACTIVE(tb); i++) {
-	if ((list = BUCKET(tb,i)) != NULL) {
-	    add_fixed_deletion(tb, i, 0);
-	    do {
-		list->hvalue = INVALID_HASH;
-		list = list->next;
-	    }while(list != NULL);
-	}
+    fixdel = (FixedDeletion*) erts_atomic_read_nob(&tb->fixdel);
+    if (fixdel && fixdel->trap) {
+        /* Continue after trap */
+        ASSERT(fixdel->all);
+        ASSERT(fixdel->slot < NACTIVE(tb));
+        i = fixdel->slot;
     }
-    erts_smp_atomic_set_nob(&tb->common.nitems, 0);    
-    return DB_ERROR_NONE;
+    else {
+        /* First call */
+        fixdel = erts_db_alloc(ERTS_ALC_T_DB_FIX_DEL,
+                               (DbTable *) tb,
+                               sizeof(FixedDeletion));
+        ERTS_ETS_MISC_MEM_ADD(sizeof(FixedDeletion));
+        link_fixdel(tb, fixdel, 0);
+        i = 0;
+    }
+
+    do {
+        HashDbTerm* b;
+	for (b = BUCKET(tb,i); b; b = b->next)
+            b->pseudo_deleted = 1;
+    } while (++i < NACTIVE(tb) && --loops > 0);
+
+    if (i < NACTIVE(tb)) {
+         /* Yield */
+        fixdel->slot = i;
+        fixdel->all = 0;
+        fixdel->trap = 1;
+        return -1;
+    }
+
+    fixdel->slot = NACTIVE(tb) - 1;
+    fixdel->all = 1;
+    fixdel->trap = 0;
+    erts_atomic_set_nob(&tb->common.nitems, 0);
+    return loops < 0 ? 0 : loops / LOOPS_PER_REDUCTION;
 }
 
 
@@ -2327,7 +2377,6 @@ static void db_print_hash(fmtfn_t to, void *to_arg, int show, DbTable *tbl)
 
     erts_print(to, to_arg, "Buckets: %d\n", NACTIVE(tb));
 
-#ifdef ERTS_SMP
     i = tbl->common.is_thread_safe;
     /* If crash dumping we set table to thread safe in order to
        avoid taking any locks */
@@ -2337,9 +2386,6 @@ static void db_print_hash(fmtfn_t to, void *to_arg, int show, DbTable *tbl)
     db_calc_stats_hash(&tbl->hash, &stats);
 
     tbl->common.is_thread_safe = i;
-#else
-    db_calc_stats_hash(&tbl->hash, &stats);
-#endif
 
     erts_print(to, to_arg, "Chain Length Avg: %f\n", stats.avg_chain_len);
     erts_print(to, to_arg, "Chain Length Max: %d\n", stats.max_chain_len);
@@ -2361,7 +2407,7 @@ static void db_print_hash(fmtfn_t to, void *to_arg, int show, DbTable *tbl)
 		continue;
 	    erts_print(to, to_arg, "%d: [", i);
 	    while(list != 0) {
-		if (list->hvalue == INVALID_HASH)
+		if (is_pseudo_deleted(list))
 		    erts_print(to, to_arg, "*");
 		if (tb->common.compress) {
 		    Eterm key = GETKEY(tb, list->dbterm.tpl);
@@ -2380,9 +2426,9 @@ static void db_print_hash(fmtfn_t to, void *to_arg, int show, DbTable *tbl)
     }
 }
 
-/* release all memory occupied by a single table */
-static int db_free_table_hash(DbTable *tbl)
+static int db_free_empty_table_hash(DbTable *tbl)
 {
+    ASSERT(NITEMS(tbl) == 0);
     while (db_free_table_continue_hash(tbl, ERTS_SWORD_MAX) < 0)
 	;
     return 0;
@@ -2391,8 +2437,8 @@ static int db_free_table_hash(DbTable *tbl)
 static SWord db_free_table_continue_hash(DbTable *tbl, SWord reds)
 {
     DbTableHash *tb = &tbl->hash;
-    FixedDeletion* fixdel = (FixedDeletion*) erts_smp_atomic_read_acqb(&tb->fixdel);
-    ERTS_SMP_LC_ASSERT(IS_TAB_WLOCKED(tb) || (tb->common.status & DB_DELETE));
+    FixedDeletion* fixdel = (FixedDeletion*) erts_atomic_read_acqb(&tb->fixdel);
+    ERTS_LC_ASSERT(IS_TAB_WLOCKED(tb) || (tb->common.status & DB_DELETE));
 
     while (fixdel != NULL) {
 	FixedDeletion *fx = fixdel;
@@ -2404,11 +2450,11 @@ static SWord db_free_table_continue_hash(DbTable *tbl, SWord reds)
 		     sizeof(FixedDeletion));
 	ERTS_ETS_MISC_MEM_ADD(-sizeof(FixedDeletion));
 	if (--reds < 0) {
-	    erts_smp_atomic_set_relb(&tb->fixdel, (erts_aint_t)fixdel);
+	    erts_atomic_set_relb(&tb->fixdel, (erts_aint_t)fixdel);
 	    return reds;		/* Not done */
 	}
     }
-    erts_smp_atomic_set_relb(&tb->fixdel, (erts_aint_t)NULL);
+    erts_atomic_set_relb(&tb->fixdel, (erts_aint_t)NULL);
 
     while(tb->nslots != 0) {
 	reds -= EXT_SEGSZ/64 + free_seg(tb, 1);
@@ -2420,7 +2466,6 @@ static SWord db_free_table_continue_hash(DbTable *tbl, SWord reds)
 	    return reds;	/* Not done */
 	}
     }
-#ifdef ERTS_SMP
     if (tb->locks != NULL) {
 	int i;
 	for (i=0; i<DB_HASH_LOCK_CNT; ++i) {
@@ -2430,8 +2475,7 @@ static SWord db_free_table_continue_hash(DbTable *tbl, SWord reds)
 		     (void*)tb->locks, sizeof(DbTableHashFineLocks));
 	tb->locks = NULL;
     }
-#endif    
-    ASSERT(erts_smp_atomic_read_nob(&tb->common.memory_size) == sizeof(DbTable));
+    ASSERT(erts_atomic_read_nob(&tb->common.memory_size) == sizeof(DbTable));
     return reds;			/* Done */
 }
 
@@ -2462,7 +2506,6 @@ static int analyze_pattern(DbTableHash *tb, Eterm pattern,
     mpi->num_lists = 0;
     mpi->key_given = 1;
     mpi->something_can_match = 0;
-    mpi->all_objects = 1;
     mpi->mp = NULL;
 
     for (lst = pattern; is_list(lst); lst = CDR(list_val(lst)))
@@ -2515,7 +2558,6 @@ static int analyze_pattern(DbTableHash *tb, Eterm pattern,
 
 	if (!is_list(body) || CDR(list_val(body)) != NIL ||
 	    CAR(list_val(body)) != am_DollarUnderscore) {
-	    mpi->all_objects = 0;
 	}
 	++i;
 	if (!(mpi->key_given)) {
@@ -2530,7 +2572,7 @@ static int analyze_pattern(DbTableHash *tb, Eterm pattern,
 		if (!db_has_variable(key)) {   /* Bound key */
 		    int ix, search_slot;
 		    HashDbTerm** bp;
-		    erts_smp_rwmtx_t* lck;
+		    erts_rwmtx_t* lck;
 		    hval = MAKE_HASH(key);
 		    lck = RLOCK_HASH(tb,hval);
 		    ix = hash_to_ix(tb, hval);
@@ -2634,14 +2676,12 @@ static void alloc_seg(DbTableHash *tb)
     tb->nslots += EXT_SEGSZ;
 }
 
-#ifdef ERTS_SMP
 static void dealloc_ext_segtab(void* lop_data)
 {
     struct ext_segtab* est = (struct ext_segtab*) lop_data;
 
     erts_free(ERTS_ALC_T_DB_SEG, est);
 }
-#endif
 
 /* Shrink table by freeing the top segment
 ** free_records: 1=free any records in segment, 0=assume segment is empty 
@@ -2680,7 +2720,6 @@ static int free_seg(DbTableHash *tb, int free_records)
             SET_SEGTAB(tb, est->prev_segtab);
             tb->nsegs = est->prev_nsegs;
 
-#ifdef ERTS_SMP
             if (!tb->common.is_thread_safe) {
                 /*
                  * Table is doing a graceful shrink operation and we must avoid
@@ -2698,7 +2737,6 @@ static int free_seg(DbTableHash *tb, int free_records)
                                                         sz);
             }
             else
-#endif
                 erts_db_free(ERTS_ALC_T_DB_SEG, (DbTable*)tb, est,
                              SIZEOF_EXT_SEGTAB(est->nsegs));
         }
@@ -2733,7 +2771,7 @@ static Eterm build_term_list(Process* p, HashDbTerm* ptr1, HashDbTerm* ptr2,
     if (!sz) {
 	ptr = ptr1;
 	while(ptr != ptr2) {
-	    if (ptr->hvalue != INVALID_HASH)
+	    if (!is_pseudo_deleted(ptr))
 		sz += ptr->dbterm.size + 2;
 	    ptr = ptr->next;
 	}
@@ -2744,7 +2782,7 @@ static Eterm build_term_list(Process* p, HashDbTerm* ptr1, HashDbTerm* ptr2,
 
     ptr = ptr1;
     while(ptr != ptr2) {
-	if (ptr->hvalue != INVALID_HASH) {
+	if (!is_pseudo_deleted(ptr)) {
 	    copy = db_copy_object_from_ets(&tb->common, &ptr->dbterm, &hp, &MSO(p));
 	    list = CONS(hp, copy, list);
 	    hp  += 2;
@@ -2759,22 +2797,18 @@ static Eterm build_term_list(Process* p, HashDbTerm* ptr1, HashDbTerm* ptr2,
 static ERTS_INLINE int
 begin_resizing(DbTableHash* tb)
 {
-#ifdef ERTS_SMP
     if (DB_USING_FINE_LOCKING(tb))
 	return !erts_atomic_xchg_acqb(&tb->is_resizing, 1);
     else
         ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&tb->common.rwlock));
-#endif
     return 1;
 }
 
 static ERTS_INLINE void
 done_resizing(DbTableHash* tb)
 {
-#ifdef ERTS_SMP
     if (DB_USING_FINE_LOCKING(tb))
 	erts_atomic_set_relb(&tb->is_resizing, 0);
-#endif
 }
 
 /* Grow table with one or more new buckets.
@@ -2785,7 +2819,7 @@ static void grow(DbTableHash* tb, int nitems)
     HashDbTerm** pnext;
     HashDbTerm** to_pnext;
     HashDbTerm* p;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int nactive;
     int from_ix, to_ix;
     int szm;
@@ -2807,7 +2841,7 @@ static void grow(DbTableHash* tb, int nitems)
         }
         ASSERT(nactive < tb->nslots);
 
-        szm = erts_smp_atomic_read_nob(&tb->szm);
+        szm = erts_atomic_read_nob(&tb->szm);
         if (nactive <= szm) {
             from_ix = nactive & (szm >> 1);
         } else {
@@ -2818,7 +2852,7 @@ static void grow(DbTableHash* tb, int nitems)
         to_ix = nactive;
 
         lck = WLOCK_HASH(tb, from_ix);
-        ERTS_SMP_ASSERT(lck == GET_LOCK_MAYBE(tb,to_ix));
+        ERTS_ASSERT(lck == GET_LOCK_MAYBE(tb,to_ix));
         /* Now a final double check (with the from_ix lock held)
          * that we did not get raced by a table fixer.
          */
@@ -2826,12 +2860,12 @@ static void grow(DbTableHash* tb, int nitems)
             WUNLOCK_HASH(lck);
             goto abort;
         }
-        erts_smp_atomic_set_nob(&tb->nactive, ++nactive);
+        erts_atomic_set_nob(&tb->nactive, ++nactive);
         if (from_ix == 0) {
             if (DB_USING_FINE_LOCKING(tb))
-                erts_smp_atomic_set_relb(&tb->szm, szm);
+                erts_atomic_set_relb(&tb->szm, szm);
             else
-                erts_smp_atomic_set_nob(&tb->szm, szm);
+                erts_atomic_set_nob(&tb->szm, szm);
         }
         done_resizing(tb);
 
@@ -2841,7 +2875,7 @@ static void grow(DbTableHash* tb, int nitems)
         p = *pnext;
         to_pnext = &BUCKET(tb, to_ix);
         while (p != NULL) {
-            if (p->hvalue == INVALID_HASH) { /* rare but possible with fine locking */
+            if (is_pseudo_deleted(p)) { /* rare but possible with fine locking */
                 *pnext = p->next;
                 free_term(tb, p);
                 p = *pnext;
@@ -2879,7 +2913,7 @@ static void shrink(DbTableHash* tb, int nitems)
     HashDbTerm** src_bp;
     HashDbTerm** dst_bp;
     HashDbTerm** bp;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int src_ix, dst_ix, low_szm;
     int nactive;
     int loop_limit = 5;
@@ -2892,13 +2926,13 @@ static void shrink(DbTableHash* tb, int nitems)
             goto abort; /* already done (race) */
         }
         src_ix = nactive - 1;
-        low_szm = erts_smp_atomic_read_nob(&tb->szm) >> 1;
+        low_szm = erts_atomic_read_nob(&tb->szm) >> 1;
         dst_ix = src_ix & low_szm;
 
         ASSERT(dst_ix < src_ix);
         ASSERT(nactive > FIRST_SEGSZ);
         lck = WLOCK_HASH(tb, dst_ix);
-        ERTS_SMP_ASSERT(lck == GET_LOCK_MAYBE(tb,src_ix));
+        ERTS_ASSERT(lck == GET_LOCK_MAYBE(tb,src_ix));
         /* Double check for racing table fixers */
         if (IS_FIXED(tb)) {
             WUNLOCK_HASH(lck);
@@ -2914,7 +2948,7 @@ static void shrink(DbTableHash* tb, int nitems)
          * as we must step through "src" anyway to purge pseudo deleted.
          */
         while(*bp != NULL) {
-            if ((*bp)->hvalue == INVALID_HASH) {
+            if (is_pseudo_deleted(*bp)) {
                 HashDbTerm* deleted = *bp;
                 *bp = deleted->next;
                 free_term(tb, deleted);
@@ -2927,9 +2961,9 @@ static void shrink(DbTableHash* tb, int nitems)
         *src_bp = NULL;
 
         nactive = src_ix;
-        erts_smp_atomic_set_nob(&tb->nactive, nactive);
+        erts_atomic_set_nob(&tb->nactive, nactive);
         if (dst_ix == 0) {
-            erts_smp_atomic_set_relb(&tb->szm, low_szm);
+            erts_atomic_set_relb(&tb->szm, low_szm);
         }
         WUNLOCK_HASH(lck);
 
@@ -2964,15 +2998,15 @@ static HashDbTerm* search_list(DbTableHash* tb, Eterm key,
 /* It return the next live object in a table, NULL if no more */
 /* In-bucket: RLOCKED */
 /* Out-bucket: RLOCKED unless NULL */
-static HashDbTerm* next(DbTableHash *tb, Uint *iptr, erts_smp_rwmtx_t** lck_ptr,
-			HashDbTerm *list)
+static HashDbTerm* next_live(DbTableHash *tb, Uint *iptr, erts_rwmtx_t** lck_ptr,
+			     HashDbTerm *list)
 {
     int i;
 
-    ERTS_SMP_LC_ASSERT(IS_HASH_RLOCKED(tb,*iptr));
+    ERTS_LC_ASSERT(IS_HASH_RLOCKED(tb,*iptr));
 
-    for (list = list->next; list != NULL; list = list->next) {
-	if (list->hvalue != INVALID_HASH)
+    for ( ; list != NULL; list = list->next) {
+	if (!is_pseudo_deleted(list))
 	    return list;        
     }
 
@@ -2981,7 +3015,7 @@ static HashDbTerm* next(DbTableHash *tb, Uint *iptr, erts_smp_rwmtx_t** lck_ptr,
 
 	list = BUCKET(tb,i);
 	while (list != NULL) {
-	    if (list->hvalue != INVALID_HASH) {
+	    if (!is_pseudo_deleted(list)) {
 		*iptr = i;
 		return list;        
 	    }
@@ -2999,7 +3033,7 @@ db_lookup_dbterm_hash(Process *p, DbTable *tbl, Eterm key, Eterm obj,
     DbTableHash *tb = &tbl->hash;
     HashValue hval;
     HashDbTerm **bp, *b;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int flags = 0;
 
     ASSERT(tb->common.status & DB_SET);
@@ -3014,7 +3048,7 @@ db_lookup_dbterm_hash(Process *p, DbTable *tbl, Eterm key, Eterm obj,
             break;
         }
         if (has_key(tb, b, key, hval)) {
-            if (b->hvalue != INVALID_HASH) {
+            if (!is_pseudo_deleted(b)) {
                 goto Ldone;
             }
             break;
@@ -3044,18 +3078,20 @@ db_lookup_dbterm_hash(Process *p, DbTable *tbl, Eterm key, Eterm obj,
             HashDbTerm *q = new_dbterm(tb, obj);
 
             q->hvalue = hval;
+            q->pseudo_deleted = 0;
             q->next = NULL;
             *bp = b = q;
             flags |= DB_INC_TRY_GROW;
         } else {
             HashDbTerm *q, *next = b->next;
 
-            ASSERT(b->hvalue == INVALID_HASH);
+            ASSERT(is_pseudo_deleted(b));
             q = replace_dbterm(tb, b, obj);
             q->next = next;
-            q->hvalue = hval;
+            ASSERT(q->hvalue == hval);
+            q->pseudo_deleted = 0;
             *bp = b = q;
-            erts_smp_atomic_inc_nob(&tb->common.nitems);
+            erts_atomic_inc_nob(&tb->common.nitems);
         }
 
         HRelease(p, hend, htop);
@@ -3081,24 +3117,24 @@ db_finalize_dbterm_hash(int cret, DbUpdateHandle* handle)
     DbTableHash *tb = &tbl->hash;
     HashDbTerm **bp = (HashDbTerm **) handle->bp;
     HashDbTerm *b = *bp;
-    erts_smp_rwmtx_t* lck = (erts_smp_rwmtx_t*) handle->lck;
+    erts_rwmtx_t* lck = (erts_rwmtx_t*) handle->lck;
     HashDbTerm* free_me = NULL;
 
-    ERTS_SMP_LC_ASSERT(IS_HASH_WLOCKED(tb, lck));  /* locked by db_lookup_dbterm_hash */
+    ERTS_LC_ASSERT(IS_HASH_WLOCKED(tb, lck));  /* locked by db_lookup_dbterm_hash */
 
     ASSERT((&b->dbterm == handle->dbterm) == !(tb->common.compress && handle->flags & DB_MUST_RESIZE));
 
     if (handle->flags & DB_NEW_OBJECT && cret != DB_ERROR_NONE) {
         if (IS_FIXED(tb) && add_fixed_deletion(tb, hash_to_ix(tb, b->hvalue),
                                                0)) {
-            b->hvalue = INVALID_HASH;
+            b->pseudo_deleted = 1;
         } else {
             *bp = b->next;
             free_me = b;
         }
 
         WUNLOCK_HASH(lck);
-        erts_smp_atomic_dec_nob(&tb->common.nitems);
+        erts_atomic_dec_nob(&tb->common.nitems);
         try_shrink(tb);
     } else {
         if (handle->flags & DB_MUST_RESIZE) {
@@ -3107,7 +3143,7 @@ db_finalize_dbterm_hash(int cret, DbUpdateHandle* handle)
         }
         if (handle->flags & DB_INC_TRY_GROW) {
             int nactive;
-            int nitems = erts_smp_atomic_inc_read_nob(&tb->common.nitems);
+            int nitems = erts_atomic_inc_read_nob(&tb->common.nitems);
             WUNLOCK_HASH(lck);
             nactive = NACTIVE(tb);
 
@@ -3128,16 +3164,19 @@ db_finalize_dbterm_hash(int cret, DbUpdateHandle* handle)
     return;
 }
 
-static int db_delete_all_objects_hash(Process* p, DbTable* tbl)
+static SWord db_delete_all_objects_hash(Process* p, DbTable* tbl, SWord reds)
 {
     if (IS_FIXED(tbl)) {
-	db_mark_all_deleted_hash(tbl);
+	reds = db_mark_all_deleted_hash(tbl, reds);
     } else {
-	db_free_table_hash(tbl);
+        reds = db_free_table_continue_hash(tbl, reds);
+        if (reds < 0)
+            return reds;
+
 	db_create_hash(p, tbl);
-	erts_smp_atomic_set_nob(&tbl->hash.common.nitems, 0);
+	erts_atomic_set_nob(&tbl->hash.common.nitems, 0);
     }
-    return 0;
+    return reds;
 }
 
 void db_foreach_offheap_hash(DbTable *tbl,
@@ -3165,7 +3204,7 @@ void db_foreach_offheap_hash(DbTable *tbl,
 void db_calc_stats_hash(DbTableHash* tb, DbHashStats* stats)
 {
     HashDbTerm* b;
-    erts_smp_rwmtx_t* lck;
+    erts_rwmtx_t* lck;
     int sum = 0;
     int sq_sum = 0;
     int kept_items = 0;
@@ -3180,7 +3219,7 @@ void db_calc_stats_hash(DbTableHash* tb, DbHashStats* stats)
 	len = 0;
 	for (b = BUCKET(tb,ix); b!=NULL; b=b->next) {
 	    len++;
-            if (b->hvalue == INVALID_HASH)
+            if (is_pseudo_deleted(b))
                 ++kept_items;
 	}
 	sum += len;
diff --git a/erts/emulator/beam/erl_db_hash.h b/erts/emulator/beam/erl_db_hash.h
index 523ed7860e..08e5b13db1 100644
--- a/erts/emulator/beam/erl_db_hash.h
+++ b/erts/emulator/beam/erl_db_hash.h
@@ -24,13 +24,26 @@
 #include "erl_db_util.h" /* DbTerm & DbTableCommon */
 
 typedef struct fixed_deletion {
-    int slot;
+    UWord slot : sizeof(UWord)*8 - 2;
+    UWord all : 1;
+    UWord trap : 1;
     struct fixed_deletion *next;
 } FixedDeletion;
 
+
+typedef Uint32 HashVal;
+
 typedef struct hash_db_term {
     struct  hash_db_term* next;  /* next bucket */
-    HashValue  hvalue;        /* stored hash value */
+#if SIZEOF_VOID_P == 4
+    Uint32 hvalue : 31;     /* stored hash value */
+    Uint32 pseudo_deleted : 1;
+# define MAX_HASH_MASK (((Uint32)1 << 31)-1)
+#elif SIZEOF_VOID_P == 8
+    Uint32 hvalue;
+    Uint32 pseudo_deleted;
+# define MAX_HASH_MASK ((Uint32)(Sint32)-1)
+#endif
     DbTerm dbterm;         /* The actual term */
 } HashDbTerm;
 
@@ -42,8 +55,8 @@ typedef struct hash_db_term {
 
 typedef struct db_table_hash_fine_locks {
     union {
-	erts_smp_rwmtx_t lck;
-	byte _cache_line_alignment[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_smp_rwmtx_t))];
+	erts_rwmtx_t lck;
+	byte _cache_line_alignment[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_rwmtx_t))];
     }lck_vec[DB_HASH_LOCK_CNT];
 } DbTableHashFineLocks;
 
@@ -51,10 +64,10 @@ typedef struct db_table_hash {
     DbTableCommon common;
 
     /* SMP: szm and nactive are write-protected by is_resizing or table write lock */
-    erts_smp_atomic_t szm;     /* current size mask. */
-    erts_smp_atomic_t nactive; /* Number of "active" slots */
+    erts_atomic_t szm;     /* current size mask. */
+    erts_atomic_t nactive; /* Number of "active" slots */
 
-    erts_smp_atomic_t segtab;  /* The segment table (struct segment**) */
+    erts_atomic_t segtab;  /* The segment table (struct segment**) */
     struct segment* first_segtab[1];
 
     /* SMP: nslots and nsegs are protected by is_resizing or table write lock */
@@ -62,11 +75,9 @@ typedef struct db_table_hash {
     int nsegs;        /* Size of segment table */
 
     /* List of slots where elements have been deleted while table was fixed */
-    erts_smp_atomic_t fixdel;  /* (FixedDeletion*) */	
-#ifdef ERTS_SMP
-    erts_smp_atomic_t is_resizing; /* grow/shrink in progress */
+    erts_atomic_t fixdel;  /* (FixedDeletion*) */
+    erts_atomic_t is_resizing; /* grow/shrink in progress */
     DbTableHashFineLocks* locks;
-#endif
 } DbTableHash;
 
 
@@ -88,9 +99,6 @@ int db_get_hash(Process *p, DbTable *tbl, Eterm key, Eterm *ret);
 
 int db_erase_hash(DbTable *tbl, Eterm key, Eterm *ret);
 
-/* not yet in method table */
-int db_mark_all_deleted_hash(DbTable *tbl);
-
 typedef struct {
     float avg_chain_len;
     float std_dev_chain_len;
diff --git a/erts/emulator/beam/erl_db_tree.c b/erts/emulator/beam/erl_db_tree.c
index 7c80e92e50..0692583dd4 100644
--- a/erts/emulator/beam/erl_db_tree.c
+++ b/erts/emulator/beam/erl_db_tree.c
@@ -50,7 +50,7 @@
 #include "erl_db_tree.h"
 
 #define GETKEY_WITH_POS(Keypos, Tplp) (*((Tplp) + Keypos))
-#define NITEMS(tb) ((int)erts_smp_atomic_read_nob(&(tb)->common.nitems))
+#define NITEMS(tb) ((int)erts_atomic_read_nob(&(tb)->common.nitems))
 
 /*
 ** A stack of this size is enough for an AVL tree with more than
@@ -91,7 +91,7 @@
 */
 static DbTreeStack* get_static_stack(DbTableTree* tb)
 {
-    if (!erts_smp_atomic_xchg_acqb(&tb->is_stack_busy, 1)) {
+    if (!erts_atomic_xchg_acqb(&tb->is_stack_busy, 1)) {
 	return &tb->static_stack;
     }
     return NULL;
@@ -103,7 +103,7 @@ static DbTreeStack* get_static_stack(DbTableTree* tb)
 static DbTreeStack* get_any_stack(DbTableTree* tb)
 {
     DbTreeStack* stack;
-    if (!erts_smp_atomic_xchg_acqb(&tb->is_stack_busy, 1)) {
+    if (!erts_atomic_xchg_acqb(&tb->is_stack_busy, 1)) {
 	return &tb->static_stack;
     }
     stack = erts_db_alloc(ERTS_ALC_T_DB_STK, (DbTable *) tb,
@@ -117,8 +117,8 @@ static DbTreeStack* get_any_stack(DbTableTree* tb)
 static void release_stack(DbTableTree* tb, DbTreeStack* stack)
 {
     if (stack == &tb->static_stack) {
-	ASSERT(erts_smp_atomic_read_nob(&tb->is_stack_busy) == 1);
-	erts_smp_atomic_set_relb(&tb->is_stack_busy, 0);
+	ASSERT(erts_atomic_read_nob(&tb->is_stack_busy) == 1);
+	erts_atomic_set_relb(&tb->is_stack_busy, 0);
     }
     else {
 	erts_db_free(ERTS_ALC_T_DB_STK, (DbTable *) tb,
@@ -170,11 +170,6 @@ static ERTS_INLINE TreeDbTerm* replace_dbterm(DbTableTree *tb, TreeDbTerm* old,
 #define DIR_END 2 
 
 /*
- * Special binary flag
- */
-#define BIN_FLAG_ALL_OBJECTS         BIN_FLAG_USR1
-
-/*
  * Number of records to delete before trapping.
  */
 #define DELETE_RECORD_LIMIT 12000
@@ -218,9 +213,6 @@ static void do_dump_tree2(DbTableTree*, int to, void *to_arg, int show,
  * functions.
  */
 struct mp_info {
-    int all_objects;		/* True if complete objects are always
-				 * returned from the match_spec (can use 
-				 * copy_shallow on the return value) */
     int something_can_match;	/* The match_spec is not "impossible" */
     int some_limitation;	/* There is some limitation on the search
 				 * area, i. e. least and/or most is set.*/
@@ -248,7 +240,6 @@ struct select_context {
     Eterm *lastobj;
     Sint32 max;
     int keypos;
-    int all_objects;
     Sint got;
     Sint chunk_size;
 };
@@ -263,7 +254,6 @@ struct select_count_context {
     Eterm *lastobj;
     Sint32 max;
     int keypos;
-    int all_objects;
     Sint got;
 };
 
@@ -293,7 +283,6 @@ struct select_replace_context {
     Eterm *lastobj;
     Sint32 max;
     int keypos;
-    int all_objects;
     Sint replaced;
 };
 
@@ -428,7 +417,7 @@ static int db_select_replace_continue_tree(Process *p, DbTable *tbl,
 static int db_take_tree(Process *, DbTable *, Eterm, Eterm *);
 static void db_print_tree(fmtfn_t to, void *to_arg,
 			  int show, DbTable *tbl);
-static int db_free_table_tree(DbTable *tbl);
+static int db_free_empty_table_tree(DbTable *tbl);
 
 static SWord db_free_table_continue_tree(DbTable *tbl, SWord);
 
@@ -436,7 +425,7 @@ static void db_foreach_offheap_tree(DbTable *,
 				    void (*)(ErlOffHeap *, void *),
 				    void *);
 
-static int db_delete_all_objects_tree(Process* p, DbTable* tbl);
+static SWord db_delete_all_objects_tree(Process* p, DbTable* tbl, SWord reds);
 
 #ifdef HARDDEBUG
 static void db_check_table_tree(DbTable *tbl);
@@ -481,7 +470,7 @@ DbTableMethod db_tree =
     db_select_replace_continue_tree,
     db_take_tree,
     db_delete_all_objects_tree,
-    db_free_table_tree,
+    db_free_empty_table_tree,
     db_free_table_continue_tree,
     db_print_tree,
     db_foreach_offheap_tree,
@@ -514,7 +503,7 @@ int db_create_tree(Process *p, DbTable *tbl)
 					   sizeof(TreeDbTerm *) * STACK_NEED);
     tb->static_stack.pos = 0;
     tb->static_stack.slot = 0;
-    erts_smp_atomic_init_nob(&tb->is_stack_busy, 0);
+    erts_atomic_init_nob(&tb->is_stack_busy, 0);
     tb->deletion = 0;
     return DB_ERROR_NONE;
 }
@@ -643,8 +632,8 @@ static int db_put_tree(DbTable *tbl, Eterm obj, int key_clash_fail)
     for (;;)
 	if (!*this) { /* Found our place */
 	    state = 1;
-	    if (erts_smp_atomic_inc_read_nob(&tb->common.nitems) >= TREE_MAX_ELEMENTS) {
-		erts_smp_atomic_dec_nob(&tb->common.nitems);
+	    if (erts_atomic_inc_read_nob(&tb->common.nitems) >= TREE_MAX_ELEMENTS) {
+		erts_atomic_dec_nob(&tb->common.nitems);
 		return DB_ERROR_SYSRES;
 	    }
 	    *this = new_dbterm(tb, obj);
@@ -992,7 +981,6 @@ static int db_select_continue_tree(Process *p,
     sc.lastobj = NULL;
     sc.max = 1000;
     sc.keypos = tb->common.keypos;
-    sc.all_objects = mp->intern.flags & BIN_FLAG_ALL_OBJECTS;
     sc.chunk_size = chunk_size;
     reverse = unsigned_val(tptr[7]);
     sc.got = signed_val(tptr[8]);
@@ -1143,7 +1131,6 @@ static int db_select_tree(Process *p, DbTable *tbl, Eterm tid,
     }
 
     sc.mp = mpi.mp;
-    sc.all_objects = mpi.all_objects;
 
     if (!mpi.got_partial && mpi.some_limitation && 
 	CMP_EQ(mpi.least,mpi.most)) {
@@ -1183,8 +1170,6 @@ static int db_select_tree(Process *p, DbTable *tbl, Eterm tid,
     sz = size_object(key);
     hp = HAlloc(p, 9 + sz + ERTS_MAGIC_REF_THING_SIZE);
     key = copy_struct(key, sz, &hp, &MSO(p));
-    if (mpi.all_objects)
-	(mpi.mp)->intern.flags |= BIN_FLAG_ALL_OBJECTS;
     mpb= erts_db_make_match_prog_ref(p,mpi.mp,&hp);
 	    
     continuation = TUPLE8
@@ -1346,7 +1331,6 @@ static int db_select_count_tree(Process *p, DbTable *tbl, Eterm tid,
     }
 
     sc.mp = mpi.mp;
-    sc.all_objects = mpi.all_objects;
 
     if (!mpi.got_partial && mpi.some_limitation && 
 	CMP_EQ(mpi.least,mpi.most)) {
@@ -1381,8 +1365,6 @@ static int db_select_count_tree(Process *p, DbTable *tbl, Eterm tid,
 	hp += BIG_UINT_HEAP_SIZE;
     }
     key = copy_struct(key, sz, &hp, &MSO(p));
-    if (mpi.all_objects)
-	(mpi.mp)->intern.flags |= BIN_FLAG_ALL_OBJECTS;
     mpb = erts_db_make_match_prog_ref(p,mpi.mp,&hp);
 	    
     continuation = TUPLE5
@@ -1449,7 +1431,6 @@ static int db_select_chunk_tree(Process *p, DbTable *tbl, Eterm tid,
     }
 
     sc.mp = mpi.mp;
-    sc.all_objects = mpi.all_objects;
 
     if (!mpi.got_partial && mpi.some_limitation && 
 	CMP_EQ(mpi.least,mpi.most)) {
@@ -1506,8 +1487,6 @@ static int db_select_chunk_tree(Process *p, DbTable *tbl, Eterm tid,
 	sz = size_object(key);
 	hp = HAlloc(p, 9 + sz + ERTS_MAGIC_REF_THING_SIZE);
 	key = copy_struct(key, sz, &hp, &MSO(p));
-	if (mpi.all_objects)
-	    (mpi.mp)->intern.flags |= BIN_FLAG_ALL_OBJECTS;
 	mpb = erts_db_make_match_prog_ref(p,mpi.mp,&hp);
 	
 	continuation = TUPLE8
@@ -1532,8 +1511,6 @@ static int db_select_chunk_tree(Process *p, DbTable *tbl, Eterm tid,
     hp = HAlloc(p, 9 + sz + ERTS_MAGIC_REF_THING_SIZE);
     key = copy_struct(key, sz, &hp, &MSO(p));
 
-    if (mpi.all_objects)
-	(mpi.mp)->intern.flags |= BIN_FLAG_ALL_OBJECTS;
     mpb = erts_db_make_match_prog_ref(p,mpi.mp,&hp);
     continuation = TUPLE8
 	(hp,
@@ -1605,7 +1582,7 @@ static int db_select_delete_continue_tree(Process *p,
     sc.max = 1000;
     sc.keypos = tb->common.keypos;
 
-    ASSERT(!erts_smp_atomic_read_nob(&tb->is_stack_busy));
+    ASSERT(!erts_atomic_read_nob(&tb->is_stack_busy));
     traverse_backwards(tb, &tb->static_stack, lastkey, &doit_select_delete, &sc);
 
     BUMP_REDS(p, 1000 - sc.max);
@@ -1882,7 +1859,6 @@ static int db_select_replace_tree(Process *p, DbTable *tbl, Eterm tid,
     }
 
     sc.mp = mpi.mp;
-    sc.all_objects = mpi.all_objects;
 
     stack = get_static_stack(tb);
     if (!mpi.got_partial && mpi.some_limitation &&
@@ -1928,8 +1904,6 @@ static int db_select_replace_tree(Process *p, DbTable *tbl, Eterm tid,
         hp += BIG_UINT_HEAP_SIZE;
     }
     key = copy_struct(key, sz, &hp, &MSO(p));
-    if (mpi.all_objects)
-        (mpi.mp)->intern.flags |= BIN_FLAG_ALL_OBJECTS;
     mpb = erts_db_make_match_prog_ref(p,mpi.mp,&hp);
 
     continuation = TUPLE5
@@ -1995,8 +1969,9 @@ static void db_print_tree(fmtfn_t to, void *to_arg,
 }
 
 /* release all memory occupied by a single table */
-static int db_free_table_tree(DbTable *tbl)
+static int db_free_empty_table_tree(DbTable *tbl)
 {
+    ASSERT(tbl->tree.root == NULL);
     while (db_free_table_continue_tree(tbl, ERTS_SWORD_MAX) < 0)
 	;
     return 1;
@@ -2017,18 +1992,20 @@ static SWord db_free_table_continue_tree(DbTable *tbl, SWord reds)
 		     (DbTable *) tb,
 		     (void *) tb->static_stack.array,
 		     sizeof(TreeDbTerm *) * STACK_NEED);
-	ASSERT(erts_smp_atomic_read_nob(&tb->common.memory_size)
+	ASSERT(erts_atomic_read_nob(&tb->common.memory_size)
 	       == sizeof(DbTable));
     }
     return reds;
 }
 
-static int db_delete_all_objects_tree(Process* p, DbTable* tbl)
+static SWord db_delete_all_objects_tree(Process* p, DbTable* tbl, SWord reds)
 {
-    db_free_table_tree(tbl);
+    reds = db_free_table_continue_tree(tbl, reds);
+    if (reds < 0)
+        return reds;
     db_create_tree(p, tbl);
-    erts_smp_atomic_set_nob(&tbl->tree.common.nitems, 0);
-    return 0;
+    erts_atomic_set_nob(&tbl->tree.common.nitems, 0);
+    return reds;
 }
 
 static void do_db_tree_foreach_offheap(TreeDbTerm *,
@@ -2107,7 +2084,7 @@ static TreeDbTerm *linkout_tree(DbTableTree *tb, Eterm key) {
 		tstack[tpos++] = this;
 		state = delsub(this);
 	    }
-	    erts_smp_atomic_dec_nob(&tb->common.nitems);
+	    erts_atomic_dec_nob(&tb->common.nitems);
 	    break;
 	}
     }
@@ -2174,7 +2151,7 @@ static TreeDbTerm *linkout_object_tree(DbTableTree *tb,
 		tstack[tpos++] = this;
 		state = delsub(this);
 	    }
-	    erts_smp_atomic_dec_nob(&tb->common.nitems);
+	    erts_atomic_dec_nob(&tb->common.nitems);
 	    break;
 	}
     }
@@ -2214,7 +2191,6 @@ static int analyze_pattern(DbTableTree *tb, Eterm pattern,
     mpi->got_partial = 0;
     mpi->something_can_match = 0;
     mpi->mp = NULL;
-    mpi->all_objects = 1;
     mpi->save_term = NULL;
 
     for (lst = pattern; is_list(lst); lst = CDR(list_val(lst)))
@@ -2264,7 +2240,6 @@ static int analyze_pattern(DbTableTree *tb, Eterm pattern,
 
 	if (!is_list(body) || CDR(list_val(body)) != NIL ||
 	    CAR(list_val(body)) != am_DollarUnderscore) {
-	    mpi->all_objects = 0;
 	}
 	++i;
 
@@ -3339,8 +3314,7 @@ static int doit_select(DbTableTree *tb, TreeDbTerm *this, void *ptr,
 			   GETKEY_WITH_POS(sc->keypos, this->dbterm.tpl)) > 0))) {
 	return 0;
     }
-    ret = db_match_dbterm(&tb->common,sc->p,sc->mp,sc->all_objects,
-			  &this->dbterm, &hp, 2);
+    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, &this->dbterm, &hp, 2);
     if (is_value(ret)) {
 	sc->accum = CONS(hp, ret, sc->accum);
     }
@@ -3371,8 +3345,7 @@ static int doit_select_count(DbTableTree *tb, TreeDbTerm *this, void *ptr,
 			  GETKEY_WITH_POS(sc->keypos, this->dbterm.tpl)) > 0)) {
 	return 0;
     }
-    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, 0,
-			  &this->dbterm, NULL, 0);
+    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, &this->dbterm, NULL, 0);
     if (ret == am_true) {
 	++(sc->got);
     }
@@ -3401,8 +3374,7 @@ static int doit_select_chunk(DbTableTree *tb, TreeDbTerm *this, void *ptr,
 	return 0;
     }
 
-    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, sc->all_objects,
-			  &this->dbterm, &hp, 2);
+    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, &this->dbterm, &hp, 2);
     if (is_value(ret)) {
 	++(sc->got);
 	sc->accum = CONS(hp, ret, sc->accum);
@@ -3437,8 +3409,7 @@ static int doit_select_delete(DbTableTree *tb, TreeDbTerm *this, void *ptr,
 	cmp_partly_bound(sc->end_condition, 
 			 GETKEY_WITH_POS(sc->keypos, this->dbterm.tpl)) > 0)
 	return 0;
-    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, 0,
-			  &this->dbterm, NULL, 0);
+    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, &this->dbterm, NULL, 0);
     if (ret == am_true) {
 	key = GETKEY(sc->tb, this->dbterm.tpl);
 	linkout_tree(sc->tb, key);
@@ -3465,8 +3436,7 @@ static int doit_select_replace(DbTableTree *tb, TreeDbTerm **this, void *ptr,
 			  GETKEY_WITH_POS(sc->keypos, (*this)->dbterm.tpl)) > 0)) {
 	return 0;
     }
-    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, 0,
-			  &(*this)->dbterm, NULL, 0);
+    ret = db_match_dbterm(&tb->common, sc->p, sc->mp, &(*this)->dbterm, NULL, 0);
 
     if (is_value(ret)) {
         TreeDbTerm* new;
diff --git a/erts/emulator/beam/erl_db_tree.h b/erts/emulator/beam/erl_db_tree.h
index 72749ead1e..dc1b93d410 100644
--- a/erts/emulator/beam/erl_db_tree.h
+++ b/erts/emulator/beam/erl_db_tree.h
@@ -41,7 +41,7 @@ typedef struct db_table_tree {
     /* Tree-specific fields */
     TreeDbTerm *root;         /* The tree root */
     Uint deletion;		/* Being deleted */
-    erts_smp_atomic_t is_stack_busy;
+    erts_atomic_t is_stack_busy;
     DbTreeStack static_stack;
 } DbTableTree;
 
diff --git a/erts/emulator/beam/erl_db_util.c b/erts/emulator/beam/erl_db_util.c
index 13eacaa8a9..f1d47326b4 100644
--- a/erts/emulator/beam/erl_db_util.c
+++ b/erts/emulator/beam/erl_db_util.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1998-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1998-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -38,6 +38,7 @@
 #include "erl_binary.h"
 #include "erl_map.h"
 #include "erl_thr_progress.h"
+#include "erl_proc_sig_queue.h"
 
 #include "erl_db_util.h"
 
@@ -74,12 +75,35 @@ DBIF_TABLE_GUARD | DBIF_TABLE_BODY | DBIF_TRACE_GUARD | DBIF_TRACE_BODY
 typedef struct DMC_STACK_TYPE(Type) {		\
     int pos;					\
     int siz;					\
-    Type def[DMC_DEFAULT_SIZE];		        \
+    int bytes;                                  \
     Type *data;					\
+    Type def[DMC_DEFAULT_SIZE];		        \
 } DMC_STACK_TYPE(Type)
+
+
+typedef int Dummy;
+DMC_DECLARE_STACK_TYPE(Dummy);
+
+static void dmc_stack_grow(DMC_Dummy_stack* s)
+{
+    int was_bytes = s->bytes;
+    s->siz *= 2;
+    s->bytes *= 2;
+    if (s->data == s->def) {
+        s->data = erts_alloc(ERTS_ALC_T_DB_MC_STK, s->bytes);
+        sys_memcpy(s->data, s->def, was_bytes);
+    }
+    else {
+        s->data = erts_realloc(ERTS_ALC_T_DB_MC_STK, s->data, s->bytes);
+    }
+}
     
-#define DMC_INIT_STACK(Name) \
-     (Name).pos = 0; (Name).siz = DMC_DEFAULT_SIZE; (Name).data = (Name).def
+#define DMC_INIT_STACK(Name) do {       \
+    (Name).pos = 0;                     \
+    (Name).siz = DMC_DEFAULT_SIZE;      \
+    (Name).bytes = sizeof((Name).def);  \
+    (Name).data = (Name).def;           \
+} while (0)
 
 #define DMC_STACK_DATA(Name) (Name).data
 
@@ -87,21 +111,19 @@ typedef struct DMC_STACK_TYPE(Type) {		\
 
 #define DMC_PUSH(On, What)						\
 do {									\
-    if ((On).pos >= (On).siz) {						\
-	(On).siz *= 2;							\
-	(On).data							\
-	    = (((On).def == (On).data)					\
-	       ? memcpy(erts_alloc(ERTS_ALC_T_DB_MC_STK,		\
-				   (On).siz*sizeof(*((On).data))),	\
-			(On).def,					\
-			DMC_DEFAULT_SIZE*sizeof(*((On).data)))		\
-	       : erts_realloc(ERTS_ALC_T_DB_MC_STK,			\
-			      (void *) (On).data,			\
-			      (On).siz*sizeof(*((On).data))));		\
-    }									\
+    if ((On).pos >= (On).siz)  						\
+        dmc_stack_grow((DMC_Dummy_stack*)&(On));                        \
     (On).data[(On).pos++] = What;					\
 } while (0)
 
+#define DMC_PUSH2(On, A, B)						\
+do {									\
+    if ((On).pos+1 >= (On).siz)  					\
+        dmc_stack_grow((DMC_Dummy_stack*)&(On));                        \
+    (On).data[(On).pos++] = A;					        \
+    (On).data[(On).pos++] = B;					        \
+} while (0)
+
 #define DMC_POP(From) (From).data[--(From).pos]
 
 #define DMC_TOP(From) (From).data[(From).pos - 1]
@@ -155,6 +177,7 @@ set_tracee_flags(Process *tracee_p, ErtsTracer tracer,
 	   : am_false);
     erts_tracer_replace(&tracee_p->common, tracer);
     ERTS_TRACE_FLAGS(tracee_p) = flags;
+
     return ret;
 }
 /*
@@ -170,7 +193,7 @@ static Eterm
 set_match_trace(Process *tracee_p, Eterm fail_term, ErtsTracer tracer,
 		Uint d_flags, Uint e_flags) {
 
-    ERTS_SMP_LC_ASSERT(
+    ERTS_LC_ASSERT(
         ERTS_PROC_LOCKS_ALL == erts_proc_lc_my_proc_locks(tracee_p)
         || erts_thr_progress_is_blocking());
 
@@ -361,11 +384,7 @@ typedef struct {
 } ErtsMatchPseudoProcess;
 
 
-#ifdef ERTS_SMP
-static erts_smp_tsd_key_t match_pseudo_process_key;
-#else
-static ErtsMatchPseudoProcess *match_pseudo_process;
-#endif
+static erts_tsd_key_t match_pseudo_process_key;
 
 static ERTS_INLINE void
 cleanup_match_pseudo_process(ErtsMatchPseudoProcess *mpsp, int keep_heap)
@@ -414,32 +433,27 @@ static ERTS_INLINE ErtsMatchPseudoProcess *
 get_match_pseudo_process(Process *c_p, Uint heap_size)
 {
     ErtsMatchPseudoProcess *mpsp;
-#ifdef ERTS_SMP
     ErtsSchedulerData *esdp;
 
     esdp = c_p ? c_p->scheduler_data : erts_get_scheduler_data();
 
     mpsp = esdp ? esdp->match_pseudo_process :
-        (ErtsMatchPseudoProcess*) erts_smp_tsd_get(match_pseudo_process_key);
+        (ErtsMatchPseudoProcess*) erts_tsd_get(match_pseudo_process_key);
 
     if (mpsp) {
-        ASSERT(mpsp == erts_smp_tsd_get(match_pseudo_process_key));
+        ASSERT(mpsp == erts_tsd_get(match_pseudo_process_key));
         ASSERT(mpsp->process.scheduler_data == esdp);
 	cleanup_match_pseudo_process(mpsp, 0);
     }
     else {
-	ASSERT(erts_smp_tsd_get(match_pseudo_process_key) == NULL);
+	ASSERT(erts_tsd_get(match_pseudo_process_key) == NULL);
 	mpsp = create_match_pseudo_process();
         if (esdp) {
             esdp->match_pseudo_process = (void *) mpsp;
         }
         mpsp->process.scheduler_data = esdp;
-	erts_smp_tsd_set(match_pseudo_process_key, (void *) mpsp);
+	erts_tsd_set(match_pseudo_process_key, (void *) mpsp);
     }
-#else
-    mpsp = match_pseudo_process;
-    cleanup_match_pseudo_process(mpsp, 0);
-#endif
     if (heap_size > ERTS_DEFAULT_MS_HEAP_SIZE*sizeof(Eterm)) {
 	mpsp->u.heap = (Eterm*) erts_alloc(ERTS_ALC_T_DB_MS_RUN_HEAP, heap_size);
     }
@@ -449,31 +463,25 @@ get_match_pseudo_process(Process *c_p, Uint heap_size)
     return mpsp;
 }
 
-#ifdef ERTS_SMP
 static void
 destroy_match_pseudo_process(void)
 {
     ErtsMatchPseudoProcess *mpsp;
-    mpsp = (ErtsMatchPseudoProcess *)erts_smp_tsd_get(match_pseudo_process_key);
+    mpsp = (ErtsMatchPseudoProcess *)erts_tsd_get(match_pseudo_process_key);
     if (mpsp) {
 	cleanup_match_pseudo_process(mpsp, 0);
 	erts_free(ERTS_ALC_T_DB_MS_PSDO_PROC, (void *) mpsp);
-	erts_smp_tsd_set(match_pseudo_process_key, (void *) NULL);
+	erts_tsd_set(match_pseudo_process_key, (void *) NULL);
     }
 }
-#endif
 
 static
 void
 match_pseudo_process_init(void)
 {
-#ifdef ERTS_SMP
-    erts_smp_tsd_key_create(&match_pseudo_process_key,
+    erts_tsd_key_create(&match_pseudo_process_key,
 			    "erts_match_pseudo_process_key");
-    erts_smp_install_exit_handler(destroy_match_pseudo_process);
-#else
-    match_pseudo_process = create_match_pseudo_process();
-#endif
+    erts_thr_install_exit_handler(destroy_match_pseudo_process);
 }
 
 void
@@ -484,7 +492,7 @@ erts_match_set_release_result(Process* c_p)
 
 /* The trace control word. */
 
-static erts_smp_atomic32_t trace_control_word;
+static erts_atomic32_t trace_control_word;
 
 /* This needs to be here, before the bif table... */
 
@@ -630,6 +638,18 @@ static DMCGuardBif guard_tab[] =
 	DBIF_ALL
     },
     {
+        am_map_get,
+        &map_get_2,
+        2,
+        DBIF_ALL
+    },
+    {
+        am_is_map_key,
+        &is_map_key_2,
+        2,
+        DBIF_ALL
+    },
+    {
 	am_bit_size,
 	&bit_size_1,
 	1,
@@ -923,7 +943,7 @@ static void db_free_tmp_uncompressed(DbTerm* obj);
 */
 BIF_RETTYPE db_get_trace_control_word(Process *p)
 {
-    Uint32 tcw = (Uint32) erts_smp_atomic32_read_acqb(&trace_control_word);
+    Uint32 tcw = (Uint32) erts_atomic32_read_acqb(&trace_control_word);
     BIF_RET(erts_make_integer((Uint) tcw, p));
 }
 
@@ -941,7 +961,7 @@ BIF_RETTYPE db_set_trace_control_word(Process *p, Eterm new)
     if (val != ((Uint32)val))
 	BIF_ERROR(p, BADARG);
 
-    old_tcw = (Uint32) erts_smp_atomic32_xchg_relb(&trace_control_word,
+    old_tcw = (Uint32) erts_atomic32_xchg_relb(&trace_control_word,
 						   (erts_aint32_t) val);
     BIF_RET(erts_make_integer((Uint) old_tcw, p));
 }
@@ -1466,7 +1486,7 @@ void db_initialize_util(void){
 	  sizeof(DMCGuardBif), 
 	  (int (*)(const void *, const void *)) &cmp_guard_bif);
     match_pseudo_process_init();
-    erts_smp_atomic32_init_nob(&trace_control_word, 0);
+    erts_atomic32_init_nob(&trace_control_word, 0);
 }
 
 
@@ -1533,7 +1553,7 @@ restart:
 	 context.current_match < num_progs; 
 	 ++context.current_match) { /* This loop is long, 
 				       too long */
-	memset(heap.vars, 0, heap.size * sizeof(*heap.vars));
+	sys_memset(heap.vars, 0, heap.size * sizeof(*heap.vars));
 	t = context.matchexpr[context.current_match];
 	context.stack_used = 0;
 	structure_checked = 0;
@@ -1552,8 +1572,7 @@ restart:
                 if (is_flatmap(t)) {
                     num_iters = flatmap_get_size(flatmap_val(t));
                     if (!structure_checked) {
-                        DMC_PUSH(text, matchMap);
-                        DMC_PUSH(text, num_iters);
+                        DMC_PUSH2(text, matchMap, num_iters);
                     }
                     structure_checked = 0;
                     for (i = 0; i < num_iters; ++i) {
@@ -1573,8 +1592,7 @@ restart:
                             }
                             goto error;
                         }
-                        DMC_PUSH(text, matchKey);
-                        DMC_PUSH(text, dmc_private_copy(&context, key));
+                        DMC_PUSH2(text, matchKey, dmc_private_copy(&context, key));
                         {
                             int old_stack = ++(context.stack_used);
                             Eterm value = flatmap_get_values(flatmap_val(t))[i];
@@ -1602,8 +1620,7 @@ restart:
                     Eterm *kv;
                     num_iters = hashmap_size(t);
                     if (!structure_checked) {
-                        DMC_PUSH(text, matchMap);
-                        DMC_PUSH(text, num_iters);
+                        DMC_PUSH2(text, matchMap, num_iters);
                     }
                     structure_checked = 0;
 
@@ -1629,8 +1646,7 @@ restart:
                             DESTROY_WSTACK(wstack);
                             goto error;
                         }
-                        DMC_PUSH(text, matchKey);
-                        DMC_PUSH(text, dmc_private_copy(&context, key));
+                        DMC_PUSH2(text, matchKey, dmc_private_copy(&context, key));
                         {
                             int old_stack = ++(context.stack_used);
                             res = dmc_one_term(&context, &heap, &stack, &text,
@@ -1660,8 +1676,7 @@ restart:
 		num_iters = arityval(*tuple_val(t));
 		if (!structure_checked) { /* i.e. we did not 
 					     pop it */
-		    DMC_PUSH(text,matchTuple);
-		    DMC_PUSH(text,num_iters);
+		    DMC_PUSH2(text, matchTuple, num_iters);
 		}
 		structure_checked = 0;
 		for (i = 1; i <= num_iters; ++i) {
@@ -2155,7 +2170,7 @@ restart:
 	case matchEqFloat:
 	    if (!is_float(*ep))
 		FAIL();
-	    if (memcmp(float_val(*ep) + 1, pc, sizeof(double)))
+	    if (sys_memcmp(float_val(*ep) + 1, pc, sizeof(double)))
 		FAIL();
 	    pc += TermWords(2);
 	    ++ep;
@@ -2505,32 +2520,27 @@ restart:
             if (have_no_seqtrace(SEQ_TRACE_TOKEN(c_p)))
 		*esp++ = NIL;
 	    else {
-		Eterm sender = SEQ_TRACE_TOKEN_SENDER(c_p);
-		Uint sender_sz = is_immed(sender) ? 0 : size_object(sender);
-		ehp = HAllocX(build_proc, 6 + sender_sz, HEAP_XTRA);
-		if (sender_sz) {
-		    sender = copy_struct(sender, sender_sz, &ehp, &MSO(build_proc));
-		}
- 		*esp++ = make_tuple(ehp);
- 		ehp[0] = make_arityval(5);
- 		ehp[1] = SEQ_TRACE_TOKEN_FLAGS(c_p);
- 		ehp[2] = SEQ_TRACE_TOKEN_LABEL(c_p);
- 		ehp[3] = SEQ_TRACE_TOKEN_SERIAL(c_p);
-		ehp[4] = sender;
- 		ehp[5] = SEQ_TRACE_TOKEN_LASTCNT(c_p);
-		ASSERT(SEQ_TRACE_TOKEN_ARITY(c_p) == 5);
-		ASSERT(is_immed(ehp[1]));
-		ASSERT(is_immed(ehp[2]));
-		ASSERT(is_immed(ehp[3]));
-		ASSERT(is_immed(ehp[5]));
-	    } 
+                Eterm token;
+                Uint token_sz;
+
+                ASSERT(SEQ_TRACE_TOKEN_ARITY(c_p) == 5);
+                ASSERT(is_immed(SEQ_TRACE_TOKEN_FLAGS(c_p)));
+                ASSERT(is_immed(SEQ_TRACE_TOKEN_SERIAL(c_p)));
+                ASSERT(is_immed(SEQ_TRACE_TOKEN_LASTCNT(c_p)));
+
+                token = SEQ_TRACE_TOKEN(c_p);
+                token_sz = size_object(token);
+
+                ehp = HAllocX(build_proc, token_sz, HEAP_XTRA);
+                *esp++ = copy_struct(token, token_sz, &ehp, &MSO(build_proc));
+	    }
 	    break;
         case matchEnableTrace:
             ASSERT(c_p == self);
 	    if ( (n = erts_trace_flag2bit(esp[-1]))) {
-                erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		set_tracee_flags(c_p, ERTS_TRACER(c_p), 0, n);
-                erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		esp[-1] = am_true;
 	    } else {
 		esp[-1] = FAIL_TERM;
@@ -2545,9 +2555,9 @@ restart:
 		    /* Always take over the tracer of the current process */
 		    set_tracee_flags(tmpp, ERTS_TRACER(c_p), 0, n);
                     if (tmpp == c_p)
-                        erts_smp_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL_MINOR);
+                        erts_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL_MINOR);
                     else
-                        erts_smp_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
+                        erts_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
                     esp[-1] = am_true;
 		}
 	    }
@@ -2555,9 +2565,9 @@ restart:
         case matchDisableTrace:
             ASSERT(c_p == self);
 	    if ( (n = erts_trace_flag2bit(esp[-1]))) {
-                erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		set_tracee_flags(c_p, ERTS_TRACER(c_p), n, 0);
-                erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		esp[-1] = am_true;
 	    } else {
 		esp[-1] = FAIL_TERM;
@@ -2572,9 +2582,9 @@ restart:
 		    /* Always take over the tracer of the current process */
 		    set_tracee_flags(tmpp, ERTS_TRACER(c_p), n, 0);
                     if (tmpp == c_p)
-                        erts_smp_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL_MINOR);
+                        erts_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL_MINOR);
                     else
-                        erts_smp_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
+                        erts_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
                     esp[-1] = am_true;
 		}
 	    }
@@ -2598,14 +2608,14 @@ restart:
 	    if (in_flags & ERTS_PAM_IGNORE_TRACE_SILENT)
 	      break;
 	    if (*esp == am_true) {
-		erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		ERTS_TRACE_FLAGS(c_p) |= F_TRACE_SILENT;
-		erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 	    }
 	    else if (*esp == am_false) {
-		erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		ERTS_TRACE_FLAGS(c_p) &= ~F_TRACE_SILENT;
-		erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 	    }
 	    break;
         case matchTrace2:
@@ -2634,10 +2644,10 @@ restart:
                     ERTS_TRACER_CLEAR(&tracer);
 		    break;
 		}
-		erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		(--esp)[-1] = set_match_trace(c_p, FAIL_TERM, tracer,
 					      d_flags, e_flags);
-		erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
                 ERTS_TRACER_CLEAR(&tracer);
 	    }
 	    break;
@@ -2667,13 +2677,13 @@ restart:
 		if (tmpp == c_p) {
 		    (--esp)[-1] = set_match_trace(c_p, FAIL_TERM, tracer,
 						  d_flags, e_flags);
-		    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+		    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 		} else {
-		    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+		    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 		    (--esp)[-1] = set_match_trace(tmpp, FAIL_TERM, tracer,
 						  d_flags, e_flags);
-		    erts_smp_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
-		    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+		    erts_proc_unlock(tmpp, ERTS_PROC_LOCKS_ALL);
+		    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 		}
                 ERTS_TRACER_CLEAR(&tracer);
 	    }
@@ -2757,12 +2767,10 @@ Eterm db_format_dmc_err_info(Process *p, DMCErrInfo *ei)
 	if (vnum >= 0)
 	    erts_snprintf(buff,sizeof(buff)+20,tmp->error_string, vnum);
 	else
-	    strcpy(buff,tmp->error_string);
-	sl = strlen(buff);
+	    sys_strcpy(buff,tmp->error_string);
+	sl = sys_strlen(buff);
 	shp = HAlloc(p, sl * 2 + 5);
-	sev = (tmp->severity == dmcWarning) ? 
-	    am_atom_put("warning",7) :
-	    am_error;
+	sev = (tmp->severity == dmcWarning) ? am_warning : am_error;
 	tlist = buf_to_intlist(&shp, buff, sl, NIL);
 	tpl = TUPLE2(shp, sev, tlist);
 	shp += 3;
@@ -3277,7 +3285,7 @@ void db_cleanup_offheap_comp(DbTerm* obj)
 	    break;
 	case FUN_SUBTAG:
 	    ASSERT(u.pb != &tmp);
-	    if (erts_smp_refc_dectest(&u.fun->fe->refc, 0) == 0) {
+	    if (erts_refc_dectest(&u.fun->fe->refc, 0) == 0) {
 		erts_erase_fun_entry(u.fun->fe);
 	    }
 	    break;
@@ -3550,20 +3558,17 @@ static DMCRet dmc_one_term(DMCContext *context,
 		return retRestart;
 	    }
 	    if (heap->vars[n].is_bound) {
-		DMC_PUSH(*text,matchCmp);
-		DMC_PUSH(*text,n);
+		DMC_PUSH2(*text, matchCmp, n);
 	    } else { /* Not bound, bind! */
 		if (n >= heap->vars_used)
 		    heap->vars_used = n + 1;
-		DMC_PUSH(*text,matchBind);
-		DMC_PUSH(*text,n);
+		DMC_PUSH2(*text, matchBind, n);
 		heap->vars[n].is_bound = 1;
 	    }
 	} else if (c == am_Underscore) {
 	    DMC_PUSH(*text, matchSkip);
 	} else { /* Any immediate value */
-	    DMC_PUSH(*text, matchEq);
-	    DMC_PUSH(*text, (Uint) c);
+	    DMC_PUSH2(*text, matchEq, (Uint) c);
 	}
 	break;
     case TAG_PRIMARY_LIST:
@@ -3576,9 +3581,8 @@ static DMCRet dmc_one_term(DMCContext *context,
 	switch ((hdr & _TAG_HEADER_MASK) >> _TAG_PRIMARY_SIZE) {
 	case (_TAG_HEADER_ARITYVAL >> _TAG_PRIMARY_SIZE):    
 	    n = arityval(*tuple_val(c));
-	    DMC_PUSH(*text, matchPushT);
+	    DMC_PUSH2(*text, matchPushT, n);
 	    ++(context->stack_used);
-	    DMC_PUSH(*text, n);
 	    DMC_PUSH(*stack, c);
 	    break;
         case (_TAG_HEADER_MAP >> _TAG_PRIMARY_SIZE):
@@ -3586,9 +3590,8 @@ static DMCRet dmc_one_term(DMCContext *context,
                 n = flatmap_get_size(flatmap_val(c));
             else
                 n = hashmap_size(c);
-            DMC_PUSH(*text, matchPushM);
+            DMC_PUSH2(*text, matchPushM, n);
             ++(context->stack_used);
-            DMC_PUSH(*text, n);
             DMC_PUSH(*stack, c);
             break;
 	case (_TAG_HEADER_REF >> _TAG_PRIMARY_SIZE):
@@ -3613,8 +3616,7 @@ static DMCRet dmc_one_term(DMCContext *context,
 	    break;
 	}
 	case (_TAG_HEADER_FLOAT >> _TAG_PRIMARY_SIZE):
-	    DMC_PUSH(*text,matchEqFloat);
-	    DMC_PUSH(*text, (Uint) float_val(c)[1]);
+	    DMC_PUSH2(*text, matchEqFloat, (Uint) float_val(c)[1]);
 #ifdef ARCH_64
 	    DMC_PUSH(*text, (Uint) 0);
 #else
@@ -3622,8 +3624,7 @@ static DMCRet dmc_one_term(DMCContext *context,
 #endif
 	    break;
 	default: /* BINARY, FUN, VECTOR, or EXTERNAL */
-	    DMC_PUSH(*text, matchEqBin);
-            DMC_PUSH(*text, dmc_private_copy(context, c));
+	    DMC_PUSH2(*text, matchEqBin, dmc_private_copy(context, c));
 	    break;
 	}
 	break;
@@ -3673,8 +3674,7 @@ static void do_emit_constant(DMCContext *context, DMC_STACK_TYPE(UWord) *text,
 	    emb->next = context->save;
 	    context->save = emb;
 	}
-	DMC_PUSH(*text,matchPushC);
-	DMC_PUSH(*text,(Uint) tmp);
+	DMC_PUSH2(*text, matchPushC, (Uint)tmp);
 	if (++context->stack_used > context->stack_need)
 	    context->stack_need = context->stack_used;
 }
@@ -3812,8 +3812,7 @@ dmc_tuple(DMCContext *context, DMCHeap *heap, DMC_STACK_TYPE(UWord) *text,
         *constant = 1;
         return retOk;
     }
-    DMC_PUSH(*text, matchMkTuple);
-    DMC_PUSH(*text, nelems);
+    DMC_PUSH2(*text, matchMkTuple, nelems);
     context->stack_used -= (nelems - 1);
     *constant = 0;
     return retOk;
@@ -3840,13 +3839,11 @@ dmc_map(DMCContext *context, DMCHeap *heap, DMC_STACK_TYPE(UWord) *text,
             *constant = 1;
             return retOk;
         }
-        DMC_PUSH(*text, matchPushC);
-        DMC_PUSH(*text, dmc_private_copy(context, m->keys));
+        DMC_PUSH2(*text, matchPushC, dmc_private_copy(context, m->keys));
         if (++context->stack_used > context->stack_need) {
             context->stack_need = context->stack_used;
         }
-        DMC_PUSH(*text, matchMkFlatMap);
-        DMC_PUSH(*text, nelems);
+        DMC_PUSH2(*text, matchMkFlatMap, nelems);
         context->stack_used -= nelems;
         *constant = 0;
         return retOk;
@@ -3898,8 +3895,7 @@ dmc_map(DMCContext *context, DMCHeap *heap, DMC_STACK_TYPE(UWord) *text,
                 do_emit_constant(context, text, CDR(kv));
             }
         }
-        DMC_PUSH(*text, matchMkHashMap);
-        DMC_PUSH(*text, nelems);
+        DMC_PUSH2(*text, matchMkHashMap, nelems);
         context->stack_used -= nelems;
         DESTROY_WSTACK(wstack);
         return retOk;
@@ -5059,7 +5055,7 @@ static int match_compact(ErlHeapFragment *expr, DMCErrInfo *err_info)
 	    ASSERT(j < x);
 	    erts_snprintf(buff+1, sizeof(buff) - 1, "%u", (unsigned) j);
 	    /* Yes, writing directly into terms, they ARE off heap */
-	    *p = erts_atom_put((byte *) buff, strlen(buff),
+	    *p = erts_atom_put((byte *) buff, sys_strlen(buff),
 			       ERTS_ATOM_ENC_LATIN1, 1);
 	}
 	++p;
@@ -5182,7 +5178,7 @@ BIF_RETTYPE match_spec_test_3(BIF_ALIST_3)
 {
     Eterm res;
 #ifdef DMC_DEBUG
-    if (BIF_ARG_3 == am_atom_put("dis",3)) {
+    if (BIF_ARG_3 == ERTS_MAKE_AM("dis")) {
 	test_disassemble_next = 1;
 	BIF_RET(am_true);
     } else
@@ -5293,7 +5289,7 @@ static Eterm match_spec_test(Process *p, Eterm against, Eterm spec, int trace)
 	    erts_free(ERTS_ALC_T_DB_TMP, arr);
 	}
 	erts_bin_free(mps);
-	ret = TUPLE4(hp, am_atom_put("ok",2), res, flg, lint_res);
+	ret = TUPLE4(hp, am_ok, res, flg, lint_res);
     }
     return ret;
 }
@@ -5335,7 +5331,7 @@ void db_free_tmp_uncompressed(DbTerm* obj)
 }
 
 Eterm db_match_dbterm(DbTableCommon* tb, Process* c_p, Binary* bprog,
-			     int all, DbTerm* obj, Eterm** hpp, Uint extra)
+                      DbTerm* obj, Eterm** hpp, Uint extra)
 {
     enum erts_pam_run_flags flags;
     Uint32 dummy;
@@ -5520,7 +5516,7 @@ void db_match_dis(Binary *bp)
 	    ++t;
 	    {
 		double num;
-		memcpy(&num,t,sizeof(double));
+		sys_memcpy(&num,t,sizeof(double));
 		t += TermWords(2);
 		erts_printf("EqFloat\t%f\n", num);
 	    }
@@ -5751,5 +5747,3 @@ void db_match_dis(Binary *bp)
 }
 
 #endif /* DMC_DEBUG */
-
-
diff --git a/erts/emulator/beam/erl_db_util.h b/erts/emulator/beam/erl_db_util.h
index 7ce104a84c..73d242449e 100644
--- a/erts/emulator/beam/erl_db_util.h
+++ b/erts/emulator/beam/erl_db_util.h
@@ -32,6 +32,7 @@
 ** DMC_DEBUG does NOT need DEBUG, but DEBUG needs DMC_DEBUG
 */
 #define DMC_DEBUG 1
+#define ETS_DBG_FORCE_TRAP 1
 #endif
 
 /*
@@ -180,10 +181,9 @@ typedef struct db_table_method
             Eterm* ret);
     int (*db_take)(Process *, DbTable *, Eterm, Eterm *);
 
-    int (*db_delete_all_objects)(Process* p,
-				 DbTable* db /* [in out] */ );
+    SWord (*db_delete_all_objects)(Process* p, DbTable* db, SWord reds);
 
-    int (*db_free_table)(DbTable* db /* [in out] */ );
+    int (*db_free_empty_table)(DbTable* db);
     SWord (*db_free_table_continue)(DbTable* db, SWord reds);
     
     void (*db_print)(fmtfn_t to,
@@ -240,16 +240,14 @@ typedef struct {
  */
 
 typedef struct db_table_common {
-    erts_smp_refc_t refc;     /* reference count of table struct */
-    erts_smp_refc_t fix_count;/* fixation counter */
+    erts_refc_t refc;     /* reference count of table struct */
+    erts_refc_t fix_count;/* fixation counter */
     DbTableList all;
     DbTableList owned;
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_t rwlock;  /* rw lock on table */
-    erts_smp_mtx_t fixlock;   /* Protects fixing_procs and time */
+    erts_rwmtx_t rwlock;  /* rw lock on table */
+    erts_mtx_t fixlock;   /* Protects fixing_procs and time */
     int is_thread_safe;       /* No fine locking inside table needed */
     Uint32 type;              /* table type, *read only* after creation */
-#endif
     Eterm owner;              /* Pid of the creator */
     Eterm heir;               /* Pid of the heir */
     UWord heir_data;          /* To send in ETS-TRANSFER (is_immed or (DbTerm*) */
@@ -257,8 +255,8 @@ typedef struct db_table_common {
     Eterm the_name;           /* an atom */
     Binary *btid;
     DbTableMethod* meth;      /* table methods */
-    erts_smp_atomic_t nitems; /* Total number of items in table */
-    erts_smp_atomic_t memory_size;/* Total memory size. NOTE: in bytes! */
+    erts_atomic_t nitems; /* Total number of items in table */
+    erts_atomic_t memory_size;/* Total memory size. NOTE: in bytes! */
     struct {                  /* Last fixation time */
 	ErtsMonotonicTime monotonic;
 	ErtsMonotonicTime offset;
@@ -269,6 +267,10 @@ typedef struct db_table_common {
     Uint32 status;            /* bit masks defined  below */
     int keypos;               /* defaults to 1 */
     int compress;
+
+#ifdef ETS_DBG_FORCE_TRAP
+    erts_atomic_t dbg_force_trap;  /* &1 force enabled, &2 trap this call */
+#endif
 } DbTableCommon;
 
 /* These are status bit patterns */
@@ -283,15 +285,13 @@ typedef struct db_table_common {
 #define DB_FINE_LOCKED   (1 << 8) /* write_concurrency */
 #define DB_FREQ_READ     (1 << 9) /* read_concurrency */
 #define DB_NAMED_TABLE   (1 << 10)
-
-#define ERTS_ETS_TABLE_TYPES (DB_BAG|DB_SET|DB_DUPLICATE_BAG|DB_ORDERED_SET\
-                              |DB_FINE_LOCKED|DB_FREQ_READ|DB_NAMED_TABLE)
+#define DB_BUSY          (1 << 11)
 
 #define IS_HASH_TABLE(Status) (!!((Status) & \
 				  (DB_BAG | DB_SET | DB_DUPLICATE_BAG)))
 #define IS_TREE_TABLE(Status) (!!((Status) & \
 				  DB_ORDERED_SET))
-#define NFIXED(T) (erts_smp_refc_read(&(T)->common.fix_count,0))
+#define NFIXED(T) (erts_refc_read(&(T)->common.fix_count,0))
 #define IS_FIXED(T) (NFIXED(T) != 0) 
 
 /*
@@ -471,7 +471,7 @@ Binary *db_match_compile(Eterm *matchexpr, Eterm *guards,
 /* Returns newly allocated MatchProg binary with refc == 0*/
 
 Eterm db_match_dbterm(DbTableCommon* tb, Process* c_p, Binary* bprog,
-		      int all, DbTerm* obj, Eterm** hpp, Uint extra);
+		      DbTerm* obj, Eterm** hpp, Uint extra);
 
 Eterm db_prog_match(Process *p, Process *self,
                     Binary *prog, Eterm term,
diff --git a/erts/emulator/beam/erl_debug.c b/erts/emulator/beam/erl_debug.c
index bf8244564a..db78378257 100644
--- a/erts/emulator/beam/erl_debug.c
+++ b/erts/emulator/beam/erl_debug.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1998-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1998-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -404,15 +404,16 @@ void verify_process(Process *p)
         erts_exit(ERTS_ERROR_EXIT,"Wild pointer found in " name " of %T!\n",p->common.id); }
 
 
-    ErtsMessage* mp = p->msg.first;
-
     VERBOSE(DEBUG_MEMORY,("Verify process: %T...\n",p->common.id));
 
-    while (mp != NULL) {
-        VERIFY_ETERM("message term",ERL_MESSAGE_TERM(mp));
-        VERIFY_ETERM("message token",ERL_MESSAGE_TOKEN(mp));
-        mp = mp->next;
-    }
+    ERTS_FOREACH_SIG_PRIVQS(
+        p, mp,
+        {
+            if (ERTS_SIG_IS_MSG(mp)) {
+                VERIFY_ETERM("message term",ERL_MESSAGE_TERM(mp));
+                VERIFY_ETERM("message token",ERL_MESSAGE_TOKEN(mp));
+            }
+        });
 
     erts_check_stack(p);
     erts_check_heap(p);
@@ -532,16 +533,15 @@ static void print_process_memory(Process *p)
     erts_printf("-- %-*s ---%s-%s-%s-%s--\n",
                 PTR_SIZE, "PCB", dashes, dashes, dashes, dashes);
 
-    if (p->msg.first != NULL) {
-        ErtsMessage* mp;
-        erts_printf("  Message Queue:\n");
-        mp = p->msg.first;
-        while (mp != NULL) {
-            erts_printf("| 0x%0*lx | 0x%0*lx |\n",PTR_SIZE,
-                        ERL_MESSAGE_TERM(mp),PTR_SIZE,ERL_MESSAGE_TOKEN(mp));
-            mp = mp->next;
-        }
-    }
+
+    erts_printf("  Message Queue:\n");
+    ERTS_FOREACH_SIG_PRIVQS(
+        p, mp,
+        {
+            if (ERTS_SIG_IS_MSG(mp))
+                erts_printf("| 0x%0*lx | 0x%0*lx |\n",PTR_SIZE,
+                            ERL_MESSAGE_TERM(mp),PTR_SIZE,ERL_MESSAGE_TOKEN(mp));
+        });
 
     if (p->dictionary != NULL) {
         int n = ERTS_PD_SIZE(p->dictionary);
diff --git a/erts/emulator/beam/erl_driver.h b/erts/emulator/beam/erl_driver.h
index 5ad616fec3..d5379a40d5 100644
--- a/erts/emulator/beam/erl_driver.h
+++ b/erts/emulator/beam/erl_driver.h
@@ -148,14 +148,6 @@ typedef struct _erl_drv_event* ErlDrvEvent; /* An event to be selected on. */
 typedef struct _erl_drv_port* ErlDrvPort; /* A port descriptor. */
 typedef struct _erl_drv_port* ErlDrvThreadData; /* Thread data. */
 
-#if !defined(__WIN32__) && !defined(_WIN32) && !defined(_WIN32_) && !defined(USE_SELECT)
-struct erl_drv_event_data {
-    short events;
-    short revents;
-};
-#endif
-typedef struct erl_drv_event_data *ErlDrvEventData; /* Event data */
-
 typedef struct {
     unsigned long megasecs;
     unsigned long secs;
@@ -270,10 +262,7 @@ typedef struct erl_drv_entry {
 			 unsigned int *flags); /* Works mostly like 'control',
 						  a synchronous
 						  call into the driver. */
-    void (*event)(ErlDrvData drv_data, ErlDrvEvent event,
-		  ErlDrvEventData event_data);
-                                /* Called when an event selected by 
-				   driver_event() has occurred */
+    void (*unused_event_callback)(void);
     int extended_marker;	/* ERL_DRV_EXTENDED_MARKER */
     int major_version;		/* ERL_DRV_EXTENDED_MAJOR_VERSION */
     int minor_version;		/* ERL_DRV_EXTENDED_MINOR_VERSION */
@@ -340,8 +329,6 @@ EXTERN void erl_drv_busy_msgq_limits(ErlDrvPort port,
 				     ErlDrvSizeT *high);
 
 EXTERN int driver_select(ErlDrvPort port, ErlDrvEvent event, int mode, int on);
-EXTERN int driver_event(ErlDrvPort port, ErlDrvEvent event, 
-			ErlDrvEventData event_data);
 
 EXTERN int driver_output(ErlDrvPort port, char *buf, ErlDrvSizeT len);
 EXTERN int driver_output2(ErlDrvPort port, char *hbuf, ErlDrvSizeT hlen,
diff --git a/erts/emulator/beam/erl_drv_thread.c b/erts/emulator/beam/erl_drv_thread.c
index 742c428f2a..4cf42fce57 100644
--- a/erts/emulator/beam/erl_drv_thread.c
+++ b/erts/emulator/beam/erl_drv_thread.c
@@ -50,7 +50,6 @@ fatal_error(int err, char *func)
 #define ERL_DRV_TSD_EXTRA 10
 #define ERL_DRV_INVALID_TSD_KEY INT_MAX
 
-#ifdef USE_THREADS
 
 struct ErlDrvMutex_ {
     ethr_mutex mtx;
@@ -85,10 +84,6 @@ struct ErlDrvTid_ {
 
 static ethr_tsd_key tid_key;
 
-#else /* USE_THREADS */
-static Uint tsd_len;
-static void **tsd;
-#endif
 
 static ErlDrvTSDKey next_tsd_key;
 static ErlDrvTSDKey max_used_tsd_key;
@@ -97,7 +92,6 @@ static char **used_tsd_keys;
 static erts_mtx_t tsd_mtx;
 static char *no_name;
 
-#ifdef USE_THREADS
 
 static void
 thread_exit_handler(void)
@@ -122,21 +116,15 @@ erl_drv_thread_wrapper(void *vdtid)
     return (*dtid->func)(dtid->arg);
 }
 
-#endif
 
 void erl_drv_thr_init(void)
 {
     int i;
-#ifdef USE_THREADS
     int res = ethr_tsd_key_create(&tid_key,"erts_tid_key");
     if (res == 0)
 	res = ethr_install_exit_handler(thread_exit_handler);
     if (res != 0)
 	fatal_error(res, "erl_drv_thr_init()");
-#else
-    tsd_len = 0;
-    tsd = NULL;
-#endif
 
     no_name = "unknown";
     next_tsd_key = 0;
@@ -153,13 +141,12 @@ void erl_drv_thr_init(void)
 /*
  * These functions implement the driver thread interface in erl_driver.h.
  * NOTE: Only use this interface from drivers. From within the emulator use
- * either the erl_threads.h, the erl_smp.h or the ethread.h interface.
+ * either the erl_threads.h or the ethread.h interface.
  */
 
 ErlDrvMutex *
 erl_drv_mutex_create(char *name)
 {
-#ifdef USE_THREADS
     ErlDrvMutex *dmtx = erts_alloc_fnf(ERTS_ALC_T_DRV_MTX,
 				       (sizeof(ErlDrvMutex)
 					+ (name ? sys_strlen(name) + 1 : 0)));
@@ -182,15 +169,11 @@ erl_drv_mutex_create(char *name)
 #endif
     }
     return dmtx;
-#else
-    return (ErlDrvMutex *) NULL;
-#endif
 }
 
 void
 erl_drv_mutex_destroy(ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     int res;
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_uninstall(&dmtx->lcnt);
@@ -199,24 +182,18 @@ erl_drv_mutex_destroy(ErlDrvMutex *dmtx)
     if (res != 0)
 	fatal_error(res, "erl_drv_mutex_destroy()");
     erts_free(ERTS_ALC_T_DRV_MTX, (void *) dmtx);
-#endif
 }
 
 
 char *
 erl_drv_mutex_name(ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     return dmtx ? dmtx->name : NULL;
-#else
-    return NULL;
-#endif
 }
 
 int
 erl_drv_mutex_trylock(ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     int res;
     if (!dmtx)
 	fatal_error(EINVAL, "erl_drv_mutex_trylock()");
@@ -225,22 +202,17 @@ erl_drv_mutex_trylock(ErlDrvMutex *dmtx)
     erts_lcnt_trylock(&dmtx->lcnt, res);
 #endif
     return res;
-#else
-    return 0;
-#endif
 }
 
 void
 erl_drv_mutex_lock(ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     if (!dmtx)
 	fatal_error(EINVAL, "erl_drv_mutex_lock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock(&dmtx->lcnt);
 #endif
     ethr_mutex_lock(&dmtx->mtx);
-#endif
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post(&dmtx->lcnt);
 #endif
@@ -249,20 +221,17 @@ erl_drv_mutex_lock(ErlDrvMutex *dmtx)
 void
 erl_drv_mutex_unlock(ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     if (!dmtx)
 	fatal_error(EINVAL, "erl_drv_mutex_unlock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_unlock(&dmtx->lcnt);
 #endif
     ethr_mutex_unlock(&dmtx->mtx);
-#endif
 }
 
 ErlDrvCond *
 erl_drv_cond_create(char *name)
 {
-#ifdef USE_THREADS
     ErlDrvCond *dcnd = erts_alloc_fnf(ERTS_ALC_T_DRV_CND,
 				      (sizeof(ErlDrvCond)
 				       + (name ? sys_strlen(name) + 1 : 0)));
@@ -281,57 +250,43 @@ erl_drv_cond_create(char *name)
 	}
     }
     return dcnd;
-#else
-    return (ErlDrvCond *) NULL;
-#endif
 }
 
 void
 erl_drv_cond_destroy(ErlDrvCond *dcnd)
 {
-#ifdef USE_THREADS
     int res = dcnd ? ethr_cond_destroy(&dcnd->cnd) : EINVAL;
     if (res != 0)
 	fatal_error(res, "erl_drv_cond_destroy()");
     erts_free(ERTS_ALC_T_DRV_CND, (void *) dcnd);
-#endif
 }
 
 char *
 erl_drv_cond_name(ErlDrvCond *dcnd)
 {
-#ifdef USE_THREADS
     return dcnd ? dcnd->name : NULL;
-#else
-    return NULL;
-#endif
 }
 
 void
 erl_drv_cond_signal(ErlDrvCond *dcnd)
 {
-#ifdef USE_THREADS
     if (!dcnd)
 	fatal_error(EINVAL, "erl_drv_cond_signal()");
     ethr_cond_signal(&dcnd->cnd);
-#endif
 }
 
 void
 erl_drv_cond_broadcast(ErlDrvCond *dcnd)
 {
-#ifdef USE_THREADS
     if (!dcnd)
 	fatal_error(EINVAL, "erl_drv_cond_broadcast()");
     ethr_cond_broadcast(&dcnd->cnd);
-#endif
 }
 
 
 void
 erl_drv_cond_wait(ErlDrvCond *dcnd, ErlDrvMutex *dmtx)
 {
-#ifdef USE_THREADS
     if (!dcnd || !dmtx) {
 	fatal_error(EINVAL, "erl_drv_cond_wait()");
     }
@@ -348,13 +303,11 @@ erl_drv_cond_wait(ErlDrvCond *dcnd, ErlDrvMutex *dmtx)
 	    break;
         }
     }
-#endif
 }
 
 ErlDrvRWLock *
 erl_drv_rwlock_create(char *name)
 {
-#ifdef USE_THREADS
     ErlDrvRWLock *drwlck = erts_alloc_fnf(ERTS_ALC_T_DRV_RWLCK,
 					  (sizeof(ErlDrvRWLock)
 					   + (name ? sys_strlen(name) + 1 : 0)));
@@ -375,15 +328,11 @@ erl_drv_rwlock_create(char *name)
 #endif
     }
     return drwlck;
-#else
-    return (ErlDrvRWLock *) NULL;
-#endif
 }
 
 void
 erl_drv_rwlock_destroy(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     int res;
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_uninstall(&drwlck->lcnt);
@@ -392,23 +341,17 @@ erl_drv_rwlock_destroy(ErlDrvRWLock *drwlck)
     if (res != 0)
 	fatal_error(res, "erl_drv_rwlock_destroy()");
     erts_free(ERTS_ALC_T_DRV_RWLCK, (void *) drwlck);
-#endif
 }
 
 char *
 erl_drv_rwlock_name(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     return drwlck ? drwlck->name : NULL;
-#else
-    return NULL;
-#endif
 }
 
 int
 erl_drv_rwlock_tryrlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     int res;
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_tryrlock()");
@@ -417,15 +360,11 @@ erl_drv_rwlock_tryrlock(ErlDrvRWLock *drwlck)
     erts_lcnt_trylock_opt(&drwlck->lcnt, res, ERTS_LOCK_OPTIONS_READ);
 #endif
     return res;
-#else
-    return 0;
-#endif
 }
 
 void
 erl_drv_rwlock_rlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_rlock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
@@ -435,26 +374,22 @@ erl_drv_rwlock_rlock(ErlDrvRWLock *drwlck)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post(&drwlck->lcnt);
 #endif
-#endif
 }
 
 void
 erl_drv_rwlock_runlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_runlock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_unlock_opt(&drwlck->lcnt, ERTS_LOCK_OPTIONS_READ);
 #endif
     ethr_rwmutex_runlock(&drwlck->rwmtx);
-#endif
 }
 
 int
 erl_drv_rwlock_tryrwlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     int res;
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_tryrwlock()");
@@ -463,15 +398,11 @@ erl_drv_rwlock_tryrwlock(ErlDrvRWLock *drwlck)
     erts_lcnt_trylock_opt(&drwlck->lcnt, res, ERTS_LOCK_OPTIONS_RDWR);
 #endif
     return res;
-#else
-    return 0;
-#endif
 }
 
 void
 erl_drv_rwlock_rwlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_rwlock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
@@ -481,20 +412,17 @@ erl_drv_rwlock_rwlock(ErlDrvRWLock *drwlck)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post(&drwlck->lcnt);
 #endif
-#endif
 }
 
 void
 erl_drv_rwlock_rwunlock(ErlDrvRWLock *drwlck)
 {
-#ifdef USE_THREADS
     if (!drwlck)
 	fatal_error(EINVAL, "erl_drv_rwlock_rwunlock()");
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_unlock_opt(&drwlck->lcnt, ERTS_LOCK_OPTIONS_RDWR);
 #endif
     ethr_rwmutex_rwunlock(&drwlck->rwmtx);
-#endif
 }
 
 int
@@ -511,7 +439,7 @@ erl_drv_tsd_key_create(char *name, ErlDrvTSDKey *key)
 	name_copy = no_name;
     else {
 	name_copy = erts_alloc_fnf(ERTS_ALC_T_DRV_TSD,
-				   sizeof(char)*(strlen(name) + 1));
+				   sizeof(char)*(sys_strlen(name) + 1));
 	if (!name_copy) {
 	    *key = -1;
 	    return ENOMEM;
@@ -588,20 +516,13 @@ erl_drv_tsd_key_destroy(ErlDrvTSDKey key)
 }
 
 
-#ifdef USE_THREADS
 #define ERL_DRV_TSD__ (dtid->tsd)
 #define ERL_DRV_TSD_LEN__ (dtid->tsd_len)
-#else
-#define ERL_DRV_TSD__ (tsd)
-#define ERL_DRV_TSD_LEN__ (tsd_len)
-#endif
 
 void
 erl_drv_tsd_set(ErlDrvTSDKey key, void *data)
 {
-#ifdef USE_THREADS
     struct ErlDrvTid_ *dtid = (struct ErlDrvTid_ *) erl_drv_thread_self();
-#endif
 
     if (key < 0 || max_used_tsd_key < key || !used_tsd_keys[key])
 	fatal_error(EINVAL, "erl_drv_tsd_set()");
@@ -629,15 +550,11 @@ erl_drv_tsd_set(ErlDrvTSDKey key, void *data)
 void *
 erl_drv_tsd_get(ErlDrvTSDKey key)
 {
-#ifdef USE_THREADS
     struct ErlDrvTid_ *dtid = ethr_tsd_get(tid_key);
-#endif
     if (key < 0 || max_used_tsd_key < key || !used_tsd_keys[key])
 	fatal_error(EINVAL, "erl_drv_tsd_get()");
-#ifdef USE_THREADS
     if (!dtid)
 	return NULL;
-#endif
     if (ERL_DRV_TSD_LEN__ <= key)
 	return NULL;
     return ERL_DRV_TSD__[key];
@@ -672,7 +589,6 @@ erl_drv_thread_create(char *name,
 		      void* arg,
 		      ErlDrvThreadOpts *opts)
 {
-#ifdef USE_THREADS
     int res;
     struct ErlDrvTid_ *dtid;
     ethr_thr_opts ethr_opts = ETHR_THR_OPTS_DEFAULT_INITER;
@@ -714,27 +630,19 @@ erl_drv_thread_create(char *name,
 
     *tid = (ErlDrvTid) dtid;
     return 0;
-#else
-    return ENOTSUP;
-#endif
 }
 
 char *
 erl_drv_thread_name(ErlDrvTid tid)
 {
-#ifdef USE_THREADS
     struct ErlDrvTid_ *dtid = (struct ErlDrvTid_ *) tid;
     return dtid ? dtid->name : NULL;
-#else
-    return NULL;
-#endif
 }
 
 
 ErlDrvTid
 erl_drv_thread_self(void)
 {
-#ifdef USE_THREADS
     struct ErlDrvTid_ *dtid = ethr_tsd_get(tid_key);
     if (!dtid) {
 	int res;
@@ -753,15 +661,11 @@ erl_drv_thread_self(void)
 	    fatal_error(res, "erl_drv_thread_self()");
     }
     return (ErlDrvTid) dtid;
-#else
-    return (ErlDrvTid) NULL;
-#endif
 }
 
 int
 erl_drv_equal_tids(ErlDrvTid tid1, ErlDrvTid tid2)
 {   
-#ifdef USE_THREADS
     int res;
     struct ErlDrvTid_ *dtid1 = (struct ErlDrvTid_ *) tid1;
     struct ErlDrvTid_ *dtid2 = (struct ErlDrvTid_ *) tid2;
@@ -775,28 +679,22 @@ erl_drv_equal_tids(ErlDrvTid tid1, ErlDrvTid tid2)
 	   : !ethr_equal_tids(dtid1->tid, dtid2->tid));
 
     return res;
-#else
-    return 1;
-#endif
 }
 
 void
 erl_drv_thread_exit(void *res)
 {
-#ifdef USE_THREADS
     struct ErlDrvTid_ *dtid = ethr_tsd_get(tid_key);
     if (dtid && dtid->drv_thr) {
 	ethr_thr_exit(res);
 	fatal_error(0, "erl_drv_thread_exit()");
     }
-#endif
     fatal_error(EACCES, "erl_drv_thread_exit()");
 }
 
 int
 erl_drv_thread_join(ErlDrvTid tid, void **respp)
 {
-#ifdef USE_THREADS
     int res;
     struct ErlDrvTid_ *dtid = (struct ErlDrvTid_ *) tid;
 
@@ -809,12 +707,9 @@ erl_drv_thread_join(ErlDrvTid tid, void **respp)
     if (res == 0)
 	erts_free(ERTS_ALC_T_DRV_TID, dtid);
     return res;
-#else
-    return ENOTSUP;
-#endif
 }
 
-#if defined(__DARWIN__) && defined(USE_THREADS) && defined(ERTS_SMP)
+#if defined(__DARWIN__)
 extern int erts_darwin_main_thread_pipe[2];
 extern int erts_darwin_main_thread_result_pipe[2];
 
@@ -855,7 +750,7 @@ erl_drv_steal_main_thread(char *name,
 			   + (name ? sys_strlen(name) + 1 : 0)));
     if (!dtid)
 	return ENOMEM;
-    memset(dtid,0,sizeof(ErlDrvTid_));
+    sys_memset(dtid,0,sizeof(ErlDrvTid_));
     dtid->tid = (void * ) -1;
     dtid->drv_thr = 1;
     dtid->func = func;
@@ -868,8 +763,8 @@ erl_drv_steal_main_thread(char *name,
     *tid = NULL;
     /* Ignore options and name... */
     
-    memcpy(buff,&func,sizeof(void* (*)(void*)));
-    memcpy(buff + sizeof(void* (*)(void*)),&arg,sizeof(void *));
+    sys_memcpy(buff,&func,sizeof(void* (*)(void*)));
+    sys_memcpy(buff + sizeof(void* (*)(void*)),&arg,sizeof(void *));
     write(erts_darwin_main_thread_pipe[1],buff,buff_sz);
     return 0;
 }
diff --git a/erts/emulator/beam/erl_fun.c b/erts/emulator/beam/erl_fun.c
index 535f677bb3..9c866250bb 100644
--- a/erts/emulator/beam/erl_fun.c
+++ b/erts/emulator/beam/erl_fun.c
@@ -30,17 +30,16 @@
 
 static Hash erts_fun_table;
 
-#include "erl_smp.h"
 #ifdef HIPE
 # include "hipe_mode_switch.h"
 #endif
 
-static erts_smp_rwmtx_t erts_fun_table_lock;
+static erts_rwmtx_t erts_fun_table_lock;
 
-#define erts_fun_read_lock()	erts_smp_rwmtx_rlock(&erts_fun_table_lock)
-#define erts_fun_read_unlock()	erts_smp_rwmtx_runlock(&erts_fun_table_lock)
-#define erts_fun_write_lock()	erts_smp_rwmtx_rwlock(&erts_fun_table_lock)
-#define erts_fun_write_unlock()	erts_smp_rwmtx_rwunlock(&erts_fun_table_lock)
+#define erts_fun_read_lock()	erts_rwmtx_rlock(&erts_fun_table_lock)
+#define erts_fun_read_unlock()	erts_rwmtx_runlock(&erts_fun_table_lock)
+#define erts_fun_write_lock()	erts_rwmtx_rwlock(&erts_fun_table_lock)
+#define erts_fun_write_unlock()	erts_rwmtx_rwunlock(&erts_fun_table_lock)
 
 static HashValue fun_hash(ErlFunEntry* obj);
 static int fun_cmp(ErlFunEntry* obj1, ErlFunEntry* obj2);
@@ -59,11 +58,11 @@ void
 erts_init_fun_table(void)
 {
     HashFunctions f;
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&erts_fun_table_lock, &rwmtx_opt, "fun_tab", NIL,
+    erts_rwmtx_init_opt(&erts_fun_table_lock, &rwmtx_opt, "fun_tab", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
     f.hash = (H_FUN) fun_hash;
@@ -114,9 +113,9 @@ erts_put_fun_entry(Eterm mod, int uniq, int index)
     fe = (ErlFunEntry *) hash_put(&erts_fun_table, (void*) &template);
     sys_memset(fe->uniq, 0, sizeof(fe->uniq));
     fe->index = 0;
-    refc = erts_smp_refc_inctest(&fe->refc, 0);
+    refc = erts_refc_inctest(&fe->refc, 0);
     if (refc < 2) /* New or pending delete */
-	erts_smp_refc_inc(&fe->refc, 1);
+	erts_refc_inc(&fe->refc, 1);
     erts_fun_write_unlock();
     return fe;
 }
@@ -138,9 +137,9 @@ erts_put_fun_entry2(Eterm mod, int old_uniq, int old_index,
     sys_memcpy(fe->uniq, uniq, sizeof(fe->uniq));
     fe->index = index;
     fe->arity = arity;
-    refc = erts_smp_refc_inctest(&fe->refc, 0);
+    refc = erts_refc_inctest(&fe->refc, 0);
     if (refc < 2) /* New or pending delete */
-	erts_smp_refc_inc(&fe->refc, 1);
+	erts_refc_inc(&fe->refc, 1);
     erts_fun_write_unlock();
     return fe;
 }
@@ -165,9 +164,9 @@ erts_get_fun_entry(Eterm mod, int uniq, int index)
     erts_fun_read_lock();
     ret = (ErlFunEntry *) hash_get(&erts_fun_table, (void*) &template);
     if (ret) {
-	erts_aint_t refc = erts_smp_refc_inctest(&ret->refc, 1);
+	erts_aint_t refc = erts_refc_inctest(&ret->refc, 1);
 	if (refc < 2) /* Pending delete */
-	    erts_smp_refc_inc(&ret->refc, 1);
+	    erts_refc_inc(&ret->refc, 1);
     }
     erts_fun_read_unlock();
     return ret;
@@ -183,13 +182,11 @@ void
 erts_erase_fun_entry(ErlFunEntry* fe)
 {
     erts_fun_write_lock();
-#ifdef ERTS_SMP
     /*
      * We have to check refc again since someone might have looked up
      * the fun entry and incremented refc after last check.
      */
-    if (erts_smp_refc_dectest(&fe->refc, -1) <= 0)
-#endif
+    if (erts_refc_dectest(&fe->refc, -1) <= 0)
     {
 	if (fe->address != unloaded_fun)
 	    erts_exit(ERTS_ERROR_EXIT,
@@ -221,7 +218,7 @@ erts_fun_purge_prepare(BeamInstr* start, BeamInstr* end)
 
 	    if (start <= addr && addr < end) {
 		fe->pend_purge_address = addr;
-		ERTS_SMP_WRITE_MEMORY_BARRIER;
+		ERTS_THR_WRITE_MEMORY_BARRIER;
 		fe->address = unloaded_fun;
 #ifdef HIPE
                 fe->pend_purge_native_address = fe->native_address;
@@ -275,10 +272,10 @@ erts_fun_purge_complete(ErlFunEntry **funs, Uint no)
 #ifdef HIPE
         fe->pend_purge_native_address = NULL;
 #endif
-	if (erts_smp_refc_dectest(&fe->refc, 0) == 0)
+	if (erts_refc_dectest(&fe->refc, 0) == 0)
 	    erts_erase_fun_entry(fe);
     }
-    ERTS_SMP_WRITE_MEMORY_BARRIER;
+    ERTS_THR_WRITE_MEMORY_BARRIER;
 }
 
 void
@@ -307,7 +304,7 @@ erts_dump_fun_entries(fmtfn_t to, void *to_arg)
 #ifdef HIPE
 	    erts_print(to, to_arg, "Native_address: %p\n", fe->native_address);
 #endif
-	    erts_print(to, to_arg, "Refc: %ld\n", erts_smp_refc_read(&fe->refc, 1));
+	    erts_print(to, to_arg, "Refc: %ld\n", erts_refc_read(&fe->refc, 1));
 	    b = b->next;
 	}
     }
@@ -338,7 +335,7 @@ fun_alloc(ErlFunEntry* template)
     obj->old_uniq = template->old_uniq;
     obj->old_index = template->old_index;
     obj->module = template->module;
-    erts_smp_refc_init(&obj->refc, -1);
+    erts_refc_init(&obj->refc, -1);
     obj->address = unloaded_fun;
     obj->pend_purge_address = NULL;
 #ifdef HIPE
diff --git a/erts/emulator/beam/erl_fun.h b/erts/emulator/beam/erl_fun.h
index 289d0d0b28..fb2901d866 100644
--- a/erts/emulator/beam/erl_fun.h
+++ b/erts/emulator/beam/erl_fun.h
@@ -21,7 +21,7 @@
 #ifndef __ERLFUNTABLE_H__
 #define __ERLFUNTABLE_H__
 
-#include "erl_smp.h"
+#include "erl_threads.h"
 
 /*
  * Fun entry.
@@ -42,7 +42,7 @@ typedef struct erl_fun_entry {
 
     Uint arity;			/* The arity of the fun. */
     Eterm module;		/* Tagged atom for module. */
-    erts_smp_refc_t refc;		/* Reference count: One for code + one for each
+    erts_refc_t refc;		/* Reference count: One for code + one for each
 				   fun object in each process. */
     BeamInstr *pend_purge_address; /* address stored during a pending purge */
 #ifdef HIPE
diff --git a/erts/emulator/beam/erl_gc.c b/erts/emulator/beam/erl_gc.c
index 154b9b8dac..a65dbbf42b 100644
--- a/erts/emulator/beam/erl_gc.c
+++ b/erts/emulator/beam/erl_gc.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2002-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2002-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -43,6 +43,7 @@
 #include "erl_bif_unique.h"
 #include "dist.h"
 #include "erl_nfunc_sched.h"
+#include "erl_proc_sig_queue.h"
 
 #define ERTS_INACT_WR_PB_LEAVE_MUCH_LIMIT 1
 #define ERTS_INACT_WR_PB_LEAVE_MUCH_PERCENTAGE 20
@@ -154,7 +155,6 @@ static void offset_rootset(Process *p, Sint offs, char* area, Uint area_size,
 			   Eterm* objv, int nobj);
 static void offset_off_heap(Process* p, Sint offs, char* area, Uint area_size);
 static void offset_mqueue(Process *p, Sint offs, char* area, Uint area_size);
-static void move_msgq_to_heap(Process *p);
 static int reached_max_heap_size(Process *p, Uint total_heap_size,
                                  Uint extra_heap_size, Uint extra_old_heap_size);
 static void init_gc_info(ErtsGCInfo *gcip);
@@ -181,15 +181,28 @@ typedef struct {
     Eterm ref;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Uint req_sched;
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
 } ErtsGCInfoReq;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 static struct {
     erts_mtx_t mtx;
     ErtsGCInfo info;
 } dirty_gc;
-#endif
+
+static void move_msgs_to_heap(Process *c_p)
+{
+    Uint64 pre_oh, post_oh;
+
+    pre_oh = c_p->off_heap.overhead;
+    erts_proc_sig_move_msgs_to_heap(c_p);
+    post_oh = c_p->off_heap.overhead;
+
+    if (pre_oh != post_oh) {
+	/* Got new binaries; update bin vheap size... */
+        c_p->bin_vheap_sz = next_vheap_size(c_p, post_oh,
+                                            c_p->bin_vheap_sz);
+    }
+}
 
 static ERTS_INLINE int
 gc_cost(Uint gc_moved_live_words, Uint resize_moved_words)
@@ -209,6 +222,7 @@ ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(gcireq,
                                  ErtsGCInfoReq,
                                  5,
                                  ERTS_ALC_T_GC_INFO_REQ)
+
 /*
  * Initialize GC global data.
  */
@@ -274,11 +288,9 @@ erts_init_gc(void)
       init_gc_info(&esdp->gc_info);
     }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_mtx_init(&dirty_gc.mtx, "dirty_gc_info", NIL,
+    erts_mtx_init(&dirty_gc.mtx, "dirty_gc_info", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
     init_gc_info(&dirty_gc.info);
-#endif
 
     init_gcireq_alloc();
 }
@@ -342,7 +354,7 @@ erts_heap_sizes(Process* p)
 
     for (i = num_heap_sizes-1; i >= 0; i--) {
 	n += 2;
-	if (!MY_IS_SSMALL(heap_sizes[i])) {
+	if (!IS_SSMALL(heap_sizes[i])) {
 	    big += BIG_UINT_HEAP_SIZE;
 	}
     }
@@ -357,7 +369,7 @@ erts_heap_sizes(Process* p)
 	Eterm num;
 	Sint sz = heap_sizes[i];
 
-	if (MY_IS_SSMALL(sz)) {
+	if (IS_SSMALL(sz)) {
 	    num = make_small(sz);
 	} else {
 	    num = uint_to_big(sz, bigp);
@@ -401,10 +413,16 @@ erts_gc_after_bif_call_lhf(Process* p, ErlHeapFragment *live_hf_end,
 {
     int cost;
 
-    if (p->flags & F_HIBERNATE_SCHED) {
+    if (p->flags & (F_HIBERNATE_SCHED|F_HIPE_RECV_LOCKED)) {
 	/*
 	 * We just hibernated. We do *not* want to mess
 	 * up the hibernation by an ordinary GC...
+         *
+         * OR
+         *
+         * We left a receive in HiPE with message
+         * queue lock locked, and we do not want to
+         * do a GC with message queue locked...
 	 */
 	return result;
     }
@@ -482,12 +500,10 @@ delay_garbage_collection(Process *p, ErlHeapFragment *live_hf_end, int need, int
     }
 
     if (need == 0) {
-#ifdef ERTS_DIRTY_SCHEDULERS
         if (p->flags & (F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC)) {
             ASSERT(!ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(p)));
             goto force_reschedule;
         }
-#endif
 	return 1;
     }
     /*
@@ -542,9 +558,7 @@ delay_garbage_collection(Process *p, ErlHeapFragment *live_hf_end, int need, int
     p->heap_hfrag = hfrag;
 #endif
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 force_reschedule:
-#endif
 
     /* Make sure that we do a proper GC as soon as possible... */
     p->flags |= F_FORCE_GC;
@@ -573,20 +587,26 @@ young_gen_usage(Process *p)
     hsz = p->mbuf_sz;
 
     if (p->flags & F_ON_HEAP_MSGQ) {
-	ErtsMessage *mp;
-	for (mp = p->msg.first; mp; mp = mp->next) {
-	    /*
-	     * We leave not yet decoded distribution messages
-	     * as they are in the queue since it is not
-	     * possible to determine a maximum size until
-	     * actual decoding. However, we use their estimated
-	     * size when calculating need, and by this making
-	     * it more likely that they will fit on the heap
-	     * when actually decoded.
-	     */
-	    if (mp->data.attached)
-		hsz += erts_msg_attached_data_size(mp);
-	}
+        ERTS_FOREACH_SIG_PRIVQS(
+            p, mp,
+            {
+                /*
+                 * We leave not yet decoded distribution messages
+                 * as they are in the queue since it is not
+                 * possible to determine a maximum size until
+                 * actual decoding. However, we use their estimated
+                 * size when calculating need, and by this making
+                 * it more likely that they will fit on the heap
+                 * when actually decoded.
+                 *
+                 * We however ignore off heap messages...
+                 */
+                if (ERTS_SIG_IS_MSG(mp)
+                    && mp->data.attached
+                    && mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
+                    hsz += erts_msg_attached_data_size(mp);
+                }
+            });
     }
 
     hsz += p->htop - p->heap;
@@ -617,7 +637,6 @@ young_gen_usage(Process *p)
 	}							\
     } while (0)
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static ERTS_INLINE void
 check_for_possibly_long_gc(Process *p, Uint ygen_usage)
@@ -630,7 +649,7 @@ check_for_possibly_long_gc(Process *p, Uint ygen_usage)
     sz = ygen_usage;
     sz += p->hend - p->stop;
     if (p->flags & F_ON_HEAP_MSGQ)
-        sz += p->msg.len;
+        sz += erts_proc_sig_privqs_len(p);
     if (major)
 	sz += p->old_htop - p->old_heap;
 
@@ -641,7 +660,6 @@ check_for_possibly_long_gc(Process *p, Uint ygen_usage)
     }
 }
 
-#endif
 
 /*
  * Garbage collect a process.
@@ -673,24 +691,20 @@ garbage_collect(Process* p, ErlHeapFragment *live_hf_end,
 
     ASSERT(CONTEXT_REDS - ERTS_REDS_LEFT(p, fcalls) >= esdp->virtual_reds);
 
-    state = erts_smp_atomic32_read_nob(&p->state);
+    state = erts_atomic32_read_nob(&p->state);
 
     if ((p->flags & (F_DISABLE_GC|F_DELAY_GC)) || state & ERTS_PSFLG_EXITING) {
-#ifdef ERTS_DIRTY_SCHEDULERS
     delay_gc_before_start:
-#endif
 	return delay_garbage_collection(p, live_hf_end, need, fcalls);
     }
 
     ygen_usage = max_young_gen_usage ? max_young_gen_usage : young_gen_usage(p);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
 	check_for_possibly_long_gc(p, ygen_usage);
 	if (p->flags & (F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC))
 	    goto delay_gc_before_start;
     }
-#endif
 
     if (p->abandoned_heap)
 	live_hf_end = ERTS_INVALID_HFRAG_PTR;
@@ -699,7 +713,7 @@ garbage_collect(Process* p, ErlHeapFragment *live_hf_end,
 
     ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_GC);
 
-    erts_smp_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
+    erts_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
     if (erts_system_monitor_long_gc != 0)
 	start_time = erts_get_monotonic_time(esdp);
 
@@ -732,14 +746,12 @@ garbage_collect(Process* p, ErlHeapFragment *live_hf_end,
             if (IS_TRACED_FL(p, F_TRACE_GC)) {
                 trace_gc(p, am_gc_minor_end, reclaimed_now, THE_NON_VALUE);
             }
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
 		p->flags |= F_NEED_FULLSWEEP;
 		check_for_possibly_long_gc(p, ygen_usage);
 		if (p->flags & F_DIRTY_MAJOR_GC)
 		    goto delay_gc_after_start;
 	    }
-#endif
             goto do_major_collection;
         }
         if (ERTS_SCHEDULER_IS_DIRTY(esdp))
@@ -777,20 +789,14 @@ do_major_collection:
        long_gc/large_gc triggers when this happens as process was
        killed before a GC could be done. */
     if (reds == -2) {
-        ErtsProcLocks locks = ERTS_PROC_LOCKS_ALL;
         int res;
 
-        erts_smp_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
-        erts_send_exit_signal(p, p->common.id, p, &locks,
-                              am_kill, NIL, NULL, 0);
-        erts_smp_proc_unlock(p, locks & ERTS_PROC_LOCKS_ALL_MINOR);
+        erts_set_self_exiting(p, am_killed);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     delay_gc_after_start:
-#endif
         /* erts_send_exit_signal looks for ERTS_PSFLG_GC, so
            we have to remove it after the signal is sent */
-        erts_smp_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
+        erts_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
 
         /* We have to make sure that we have space for need on the heap */
         res = delay_garbage_collection(p, live_hf_end, need, fcalls);
@@ -798,7 +804,7 @@ do_major_collection:
         return res;
     }
 
-    erts_smp_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
+    erts_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
 
     if (IS_TRACED_FL(p, F_TRACE_GC)) {
         trace_gc(p, gc_trace_end_tag, reclaimed_now, THE_NON_VALUE);
@@ -822,7 +828,6 @@ do_major_collection:
 	    monitor_large_heap(p);
     }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_SCHEDULER_IS_DIRTY(esdp)) {
 	erts_mtx_lock(&dirty_gc.mtx);
 	dirty_gc.info.garbage_cols++;
@@ -830,7 +835,6 @@ do_major_collection:
 	erts_mtx_unlock(&dirty_gc.mtx);
     }
     else
-#endif
     {
 	esdp->gc_info.garbage_cols++;
 	esdp->gc_info.reclaimed += reclaimed_now;
@@ -908,7 +912,6 @@ garbage_collect_hibernate(Process* p, int check_long_gc)
     if (p->flags & F_DISABLE_GC)
 	ERTS_INTERNAL_ERROR("GC disabled");
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(p)))
         p->flags &= ~(F_DIRTY_GC_HIBERNATE|F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC);
     else if (check_long_gc) {
@@ -921,11 +924,10 @@ garbage_collect_hibernate(Process* p, int check_long_gc)
 	}
 	p->flags = flags;
     }
-#endif
     /*
      * Preliminaries.
      */
-    erts_smp_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
+    erts_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
     ErtsGcQuickSanityCheck(p);
     ASSERT(p->stop == p->hend);	/* Stack must be empty. */
 
@@ -1017,7 +1019,7 @@ garbage_collect_hibernate(Process* p, int check_long_gc)
 
     p->flags |= F_HIBERNATED;
 
-    erts_smp_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
+    erts_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
 
     reds = gc_cost(actual_size, actual_size);
     return reds;
@@ -1112,7 +1114,6 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
 
     p->flags |= F_NEED_FULLSWEEP;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(p)))
 	p->flags &= ~F_DIRTY_CLA;
     else {
@@ -1128,7 +1129,6 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
 	    return 10;
 	}
     }
-#endif
 
     reds = (Sint64) garbage_collect(p, ERTS_INVALID_HFRAG_PTR, 0,
 				    p->arg_reg, p->arity, fcalls,
@@ -1139,7 +1139,7 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
     /*
      * Set GC state.
      */
-    erts_smp_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
+    erts_atomic32_read_bor_nob(&p->state, ERTS_PSFLG_GC);
 
     /*
      * Just did a major collection (which has discarded the old heap),
@@ -1187,7 +1187,7 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
 
 	roots++;
 
-        while (g_sz--) {
+        for ( ; g_sz--; g_ptr++) {
             Eterm gval = *g_ptr;
 
             switch (primary_tag(gval)) {
@@ -1196,26 +1196,21 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
 		val = *ptr;
                 if (IS_MOVED_BOXED(val)) {
 		    ASSERT(is_boxed(val));
-                    *g_ptr++ = val;
+                    *g_ptr = val;
 		} else if (ErtsInArea(ptr, area, area_size)) {
-                    move_boxed(&ptr,val,&old_htop,g_ptr++);
-		} else {
-		    g_ptr++;
+                    move_boxed(ptr,val,&old_htop,g_ptr);
 		}
 		break;
 	    case TAG_PRIMARY_LIST:
                 ptr = list_val(gval);
                 val = *ptr;
                 if (IS_MOVED_CONS(val)) { /* Moved */
-                    *g_ptr++ = ptr[1];
+                    *g_ptr = ptr[1];
 		} else if (ErtsInArea(ptr, area, area_size)) {
-                    move_cons(&ptr,val,&old_htop,g_ptr++);
-                } else {
-		    g_ptr++;
-		}
+                    move_cons(ptr,val,&old_htop,g_ptr);
+                }
 		break;
 	    default:
-                g_ptr++;
 		break;
 	    }
 	}
@@ -1289,7 +1284,7 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
                     ExternalThing *etp;
                     ASSERT(is_external_header(ptr->thing_word));
                     etp = (ExternalThing *) ptr;
-                    erts_smp_refc_inc(&etp->node->refc, 1);
+                    erts_refc_inc(&etp->node->refc, 1);
                     break;
                 }
             }
@@ -1311,7 +1306,7 @@ erts_garbage_collect_literals(Process* p, Eterm* literals,
     /*
      * Restore status.
      */
-    erts_smp_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
+    erts_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_GC);
 
     reds += (Sint64) gc_cost((p->htop - p->heap) + byte_lit_size/sizeof(Uint), 0);
 
@@ -1404,7 +1399,7 @@ minor_collection(Process* p, ErlHeapFragment *live_hf_end,
 		 new_sz, objv, nobj);
 
 	if (p->flags & F_ON_HEAP_MSGQ)
-	    move_msgq_to_heap(p);
+	    move_msgs_to_heap(p);
 
 	new_mature = p->old_htop - prev_old_htop;
 
@@ -1527,7 +1522,7 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
         Uint g_sz = roots->sz;
 
 	roots++;
-        while (g_sz--) {
+        for ( ; g_sz--; g_ptr++) {
             gval = *g_ptr;
 
             switch (primary_tag(gval)) {
@@ -1537,14 +1532,12 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
                 val = *ptr;
                 if (IS_MOVED_BOXED(val)) {
 		    ASSERT(is_boxed(val));
-                    *g_ptr++ = val;
+                    *g_ptr = val;
                 } else if (ErtsInArea(ptr, mature, mature_size)) {
-                    move_boxed(&ptr,val,&old_htop,g_ptr++);
+                    move_boxed(ptr,val,&old_htop,g_ptr);
                 } else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
-                    move_boxed(&ptr,val,&n_htop,g_ptr++);
-                } else {
-		    g_ptr++;
-		}
+                    move_boxed(ptr,val,&n_htop,g_ptr);
+                }
                 break;
 	    }
 
@@ -1552,19 +1545,15 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
                 ptr = list_val(gval);
                 val = *ptr;
                 if (IS_MOVED_CONS(val)) { /* Moved */
-                    *g_ptr++ = ptr[1];
+                    *g_ptr = ptr[1];
                 } else if (ErtsInArea(ptr, mature, mature_size)) {
-                    move_cons(&ptr,val,&old_htop,g_ptr++);
+                    move_cons(ptr,val,&old_htop,g_ptr);
                 } else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
-                    move_cons(&ptr,val,&n_htop,g_ptr++);
-                } else {
-		    g_ptr++;
-		}
+                    move_cons(ptr,val,&n_htop,g_ptr);
+                }
 		break;
 	    }
-
 	    default:
-                g_ptr++;
 		break;
             }
         }
@@ -1598,9 +1587,9 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
 		    ASSERT(is_boxed(val));
 		    *n_hp++ = val;
 		} else if (ErtsInArea(ptr, mature, mature_size)) {
-		    move_boxed(&ptr,val,&old_htop,n_hp++);
+		    move_boxed(ptr,val,&old_htop,n_hp++);
 		} else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
-		    move_boxed(&ptr,val,&n_htop,n_hp++);
+		    move_boxed(ptr,val,&n_htop,n_hp++);
 		} else {
 		    n_hp++;
 		}
@@ -1612,9 +1601,9 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
 		if (IS_MOVED_CONS(val)) {
 		    *n_hp++ = ptr[1];
 		} else if (ErtsInArea(ptr, mature, mature_size)) {
-		    move_cons(&ptr,val,&old_htop,n_hp++);
+		    move_cons(ptr,val,&old_htop,n_hp++);
 		} else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
-		    move_cons(&ptr,val,&n_htop,n_hp++);
+		    move_cons(ptr,val,&n_htop,n_hp++);
 		} else {
 		    n_hp++;
 		}
@@ -1634,10 +1623,10 @@ do_minor(Process *p, ErlHeapFragment *live_hf_end,
 			    *origptr = val;
 			    mb->base = binary_bytes(val);
 			} else if (ErtsInArea(ptr, mature, mature_size)) {
-			    move_boxed(&ptr,val,&old_htop,origptr);
+			    move_boxed(ptr,val,&old_htop,origptr);
 			    mb->base = binary_bytes(mb->orig);
 			} else if (ErtsInYoungGen(*origptr, ptr, oh, oh_size)) {
-			    move_boxed(&ptr,val,&n_htop,origptr);
+			    move_boxed(ptr,val,&n_htop,origptr);
 			    mb->base = binary_bytes(mb->orig);
 			}
 		    }
@@ -1795,7 +1784,7 @@ major_collection(Process* p, ErlHeapFragment *live_hf_end,
     HIGH_WATER(p) = HEAP_TOP(p);
 
     if (p->flags & F_ON_HEAP_MSGQ)
-	move_msgq_to_heap(p);
+	move_msgs_to_heap(p);
 
     ErtsGcQuickSanityCheck(p);
 
@@ -1854,7 +1843,7 @@ full_sweep_heaps(Process *p,
 	Eterm g_sz = roots->sz;
 
 	roots++;
-	while (g_sz--) {
+	for ( ; g_sz--; g_ptr++) {
 	    Eterm* ptr;
 	    Eterm val;
 	    Eterm gval = *g_ptr;
@@ -1866,32 +1855,26 @@ full_sweep_heaps(Process *p,
 		val = *ptr;
 		if (IS_MOVED_BOXED(val)) {
 		    ASSERT(is_boxed(val));
-		    *g_ptr++ = val;
+		    *g_ptr = val;
 		} else if (!erts_is_literal(gval, ptr)) {
-		    move_boxed(&ptr,val,&n_htop,g_ptr++);
-		} else {
-		    g_ptr++;
+		    move_boxed(ptr,val,&n_htop,g_ptr);
 		}
-		continue;
+                break;
 	    }
 
 	    case TAG_PRIMARY_LIST: {
 		ptr = list_val(gval);
 		val = *ptr;
 		if (IS_MOVED_CONS(val)) {
-		    *g_ptr++ = ptr[1];
+		    *g_ptr = ptr[1];
 		} else if (!erts_is_literal(gval, ptr)) {
-		    move_cons(&ptr,val,&n_htop,g_ptr++);
-		} else {
-		    g_ptr++;
+		    move_cons(ptr,val,&n_htop,g_ptr);
 		}
-		continue;
+                break;
 	    }
 
-	    default: {
-		g_ptr++;
-		continue;
-	    }
+            default:
+                break;
 	    }
 	}
     }
@@ -2170,7 +2153,7 @@ sweep(Eterm *n_hp, Eterm *n_htop,
 		ASSERT(is_boxed(val));
 		*n_hp++ = val;
 	    } else if (ERTS_IS_IN_SWEEP_AREA(gval, ptr)) {
-		move_boxed(&ptr,val,&n_htop,n_hp++);
+		move_boxed(ptr,val,&n_htop,n_hp++);
 	    } else {
 		n_hp++;
 	    }
@@ -2182,7 +2165,7 @@ sweep(Eterm *n_hp, Eterm *n_htop,
 	    if (IS_MOVED_CONS(val)) {
 		*n_hp++ = ptr[1];
 	    } else if (ERTS_IS_IN_SWEEP_AREA(gval, ptr)) {
-		move_cons(&ptr,val,&n_htop,n_hp++);
+		move_cons(ptr,val,&n_htop,n_hp++);
 	    } else {
 		n_hp++;
 	    }
@@ -2203,7 +2186,7 @@ sweep(Eterm *n_hp, Eterm *n_htop,
 			*origptr = val;
 			mb->base = binary_bytes(*origptr);
 		    } else if (ERTS_IS_IN_SWEEP_AREA(*origptr, ptr)) {
-			move_boxed(&ptr,val,&n_htop,origptr);
+			move_boxed(ptr,val,&n_htop,origptr);
 			mb->base = binary_bytes(*origptr);
 		    }
 		}
@@ -2266,7 +2249,7 @@ sweep_literals_to_old_heap(Eterm* heap_ptr, Eterm* heap_end, Eterm* htop,
 		ASSERT(is_boxed(val));
 		*heap_ptr++ = val;
 	    } else if (ErtsInArea(ptr, src, src_size)) {
-		move_boxed(&ptr,val,&htop,heap_ptr++);
+		move_boxed(ptr,val,&htop,heap_ptr++);
 	    } else {
 		heap_ptr++;
 	    }
@@ -2278,7 +2261,7 @@ sweep_literals_to_old_heap(Eterm* heap_ptr, Eterm* heap_end, Eterm* htop,
 	    if (IS_MOVED_CONS(val)) {
 		*heap_ptr++ = ptr[1];
 	    } else if (ErtsInArea(ptr, src, src_size)) {
-		move_cons(&ptr,val,&htop,heap_ptr++);
+		move_cons(ptr,val,&htop,heap_ptr++);
 	    } else {
 		heap_ptr++;
 	    }
@@ -2299,7 +2282,7 @@ sweep_literals_to_old_heap(Eterm* heap_ptr, Eterm* heap_end, Eterm* htop,
 			*origptr = val;
 			mb->base = binary_bytes(*origptr);
 		    } else if (ErtsInArea(ptr, src, src_size)) {
-			move_boxed(&ptr,val,&htop,origptr);
+			move_boxed(ptr,val,&htop,origptr);
 			mb->base = binary_bytes(*origptr);
 		    }
 		}
@@ -2332,11 +2315,11 @@ move_one_area(Eterm* n_htop, char* src, Uint src_size)
 	ASSERT(val != ERTS_HOLE_MARKER);
 	if (is_header(val)) {
 	    ASSERT(ptr + header_arity(val) < end);
-	    move_boxed(&ptr, val, &n_htop, &dummy_ref);
+	    ptr = move_boxed(ptr, val, &n_htop, &dummy_ref);
 	}
 	else { /* must be a cons cell */
 	    ASSERT(ptr+1 < end);
-	    move_cons(&ptr, val, &n_htop, &dummy_ref);
+	    move_cons(ptr, val, &n_htop, &dummy_ref);
 	    ptr += 2;
 	}
     }
@@ -2373,9 +2356,9 @@ collect_live_heap_frags(Process* p, ErlHeapFragment *live_hf_end, Eterm* n_htop)
     return n_htop;
 }
 
-static ERTS_INLINE void
-copy_one_frag(Eterm** hpp, ErlOffHeap* off_heap,
-	      ErlHeapFragment *bp, Eterm *refs, int nrefs)
+void
+erts_copy_one_frag(Eterm** hpp, ErlOffHeap* off_heap,
+                   ErlHeapFragment *bp, Eterm *refs, int nrefs)
 {
     Uint sz;
     int i;
@@ -2481,61 +2464,6 @@ copy_one_frag(Eterm** hpp, ErlOffHeap* off_heap,
     bp->off_heap.first = NULL;
 }
 
-static void
-move_msgq_to_heap(Process *p)
-{
-    
-    ErtsMessage **mpp = &p->msg.first;
-    Uint64 pre_oh = MSO(p).overhead;
-
-    while (*mpp) {
-	ErtsMessage *mp = *mpp;
-
-	if (mp->data.attached) {
-	    ErlHeapFragment *bp;
-
-	    /*
-	     * We leave not yet decoded distribution messages
-	     * as they are in the queue since it is not
-	     * possible to determine a maximum size until
-	     * actual decoding...
-	     */
-	    if (is_value(ERL_MESSAGE_TERM(mp))) {
-
-                bp = erts_message_to_heap_frag(mp);
-
-		if (bp->next)
-		    erts_move_multi_frags(&p->htop, &p->off_heap, bp,
-					  mp->m, ERL_MESSAGE_REF_ARRAY_SZ, 0);
-		else
-		    copy_one_frag(&p->htop, &p->off_heap, bp,
-				  mp->m, ERL_MESSAGE_REF_ARRAY_SZ);
-
-		if (mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
-		    mp->data.heap_frag = NULL;
-		    free_message_buffer(bp);
-		}
-		else {
-		    ErtsMessage *new_mp = erts_alloc_message(0, NULL);
-		    sys_memcpy((void *) new_mp->m, (void *) mp->m,
-			       sizeof(Eterm)*ERL_MESSAGE_REF_ARRAY_SZ);
-		    erts_msgq_replace_msg_ref(&p->msg, new_mp, mpp);
-		    mp->next = NULL;
-		    erts_cleanup_messages(mp);
-		    mp = new_mp;
-		}
-	    }
-	}
-
-	mpp = &(*mpp)->next;
-    }
-
-    if (pre_oh != MSO(p).overhead) {
-	/* Got new binaries; update vheap size... */
-	BIN_VHEAP_SZ(p) = next_vheap_size(p, MSO(p).overhead, BIN_VHEAP_SZ(p));
-    }
-}
-
 static Uint
 setup_rootset(Process *p, Eterm *objv, int nobj, Rootset *rootset)
 {
@@ -2620,15 +2548,17 @@ setup_rootset(Process *p, Eterm *objv, int nobj, Rootset *rootset)
 	break;
     case F_OFF_HEAP_MSGQ_CHNG:
     case 0: {
+        Sint len;
 	/*
 	 * We do not have off heap message queue enabled, i.e. we
-	 * need to add message queue to rootset...
+	 * need to add signal queues to rootset...
 	 */
-	ErtsMessage *mp;
+
+        len = erts_proc_sig_privqs_len(p);
 
 	/* Ensure large enough rootset... */
-	if (n + p->msg.len > rootset->size) {
-	    Uint new_size = n + p->msg.len;
+	if (n + len > rootset->size) {
+	    Uint new_size = n + len;
 	    ERTS_GC_ASSERT(roots == rootset->def);
 	    roots = erts_alloc(ERTS_ALC_T_ROOTSET,
 			       new_size*sizeof(Roots));
@@ -2636,18 +2566,19 @@ setup_rootset(Process *p, Eterm *objv, int nobj, Rootset *rootset)
 	    rootset->size = new_size;
 	}
 
-	for (mp = p->msg.first; mp; mp = mp->next) {
-
-	    if (!mp->data.attached) {
-		/*
-		 * Message may refer data on heap;
-		 * add it to rootset...
-		 */
-		roots[n].v = mp->m;
-		roots[n].sz = ERL_MESSAGE_REF_ARRAY_SZ;
-		n++;
-	    }
-	}
+        ERTS_FOREACH_SIG_PRIVQS(
+            p, mp,
+            {
+                if (ERTS_SIG_IS_INTERNAL_MSG(mp) && !mp->data.attached) {
+                    /*
+                     * Message may refer data on heap;
+                     * add it to rootset...
+                     */
+                    roots[n].v = mp->m;
+                    roots[n].sz = ERL_MESSAGE_REF_ARRAY_SZ;
+                    n++;
+                }
+            });
 	break;
     }
     }
@@ -2949,7 +2880,7 @@ sweep_off_heap(Process *p, int fullsweep)
 	    case FUN_SUBTAG:
 		{
 		    ErlFunEntry* fe = ((ErlFunThing*)ptr)->fe;
-		    if (erts_smp_refc_dectest(&fe->refc, 0) == 0) {
+		    if (erts_refc_dectest(&fe->refc, 0) == 0) {
 			erts_erase_fun_entry(fe);
 		    }
 		    break;
@@ -3158,51 +3089,59 @@ offset_off_heap(Process* p, Sint offs, char* area, Uint area_size)
     }
 }
 
+#ifndef USE_VM_PROBES
+#define ERTS_OFFSET_DT_UTAG(MP, A, ASZ, O)
+#else
+#define ERTS_OFFSET_DT_UTAG(MP, A, ASZ, O)                                      \
+    do {                                                                        \
+        Eterm utag__ = ERL_MESSAGE_DT_UTAG((MP));                               \
+        if (is_boxed(utag__) && ErtsInArea(ptr_val(utag__), (A), (ASZ))) {      \
+            ERL_MESSAGE_DT_UTAG((MP)) = offset_ptr(utag__, (O));                \
+        }                                                                       \
+    } while (0)
+#endif
+
+static ERTS_INLINE void
+offset_message(ErtsMessage *mp, Sint offs, char* area, Uint area_size)
+{
+    Eterm mesg = ERL_MESSAGE_TERM(mp);
+    if (ERTS_SIG_IS_MSG_TAG(mesg)) {
+        if (ERTS_SIG_IS_INTERNAL_MSG_TAG(mesg)) {
+            switch (primary_tag(mesg)) {
+            case TAG_PRIMARY_LIST:
+            case TAG_PRIMARY_BOXED:
+                if (ErtsInArea(ptr_val(mesg), area, area_size)) {
+                    ERL_MESSAGE_TERM(mp) = offset_ptr(mesg, offs);
+                }
+                break;
+            }
+        }
+        mesg = ERL_MESSAGE_TOKEN(mp);
+        if (is_boxed(mesg) && ErtsInArea(ptr_val(mesg), area, area_size)) {
+            ERL_MESSAGE_TOKEN(mp) = offset_ptr(mesg, offs);
+        }
+
+        ERTS_OFFSET_DT_UTAG(mp, area, area_size, offs);
+	
+        ASSERT((is_nil(ERL_MESSAGE_TOKEN(mp)) ||
+                is_tuple(ERL_MESSAGE_TOKEN(mp)) ||
+                is_atom(ERL_MESSAGE_TOKEN(mp))));
+    }
+}
+
 /*
  * Offset pointers in message queue.
  */
 static void
 offset_mqueue(Process *p, Sint offs, char* area, Uint area_size)
 {
-    ErtsMessage* mp = p->msg.first;
-
-    if ((p->flags & (F_OFF_HEAP_MSGQ|F_OFF_HEAP_MSGQ_CHNG)) != F_OFF_HEAP_MSGQ) {
-
-	while (mp != NULL) {
-	    Eterm mesg = ERL_MESSAGE_TERM(mp);
-	    if (is_value(mesg)) {
-		switch (primary_tag(mesg)) {
-		case TAG_PRIMARY_LIST:
-		case TAG_PRIMARY_BOXED:
-		    if (ErtsInArea(ptr_val(mesg), area, area_size)) {
-			ERL_MESSAGE_TERM(mp) = offset_ptr(mesg, offs);
-		    }
-		    break;
-		}
-	    }
-	    mesg = ERL_MESSAGE_TOKEN(mp);
-	    if (is_boxed(mesg) && ErtsInArea(ptr_val(mesg), area, area_size)) {
-		ERL_MESSAGE_TOKEN(mp) = offset_ptr(mesg, offs);
-	    }
-#ifdef USE_VM_PROBES
-	    mesg = ERL_MESSAGE_DT_UTAG(mp);
-	    if (is_boxed(mesg) && ErtsInArea(ptr_val(mesg), area, area_size)) {
-		ERL_MESSAGE_DT_UTAG(mp) = offset_ptr(mesg, offs);
-	    }
-#endif	
-	
-	    ASSERT((is_nil(ERL_MESSAGE_TOKEN(mp)) ||
-		    is_tuple(ERL_MESSAGE_TOKEN(mp)) ||
-		    is_atom(ERL_MESSAGE_TOKEN(mp))));
-	    mp = mp->next;
-	}
-
-    }
+    if ((p->flags & (F_OFF_HEAP_MSGQ|F_OFF_HEAP_MSGQ_CHNG)) != F_OFF_HEAP_MSGQ)
+        ERTS_FOREACH_SIG_PRIVQS(p, mp, offset_message(mp, offs, area, area_size));
 }
 
 static void ERTS_INLINE
 offset_one_rootset(Process *p, Sint offs, char* area, Uint area_size,
-	       Eterm* objv, int nobj)
+                   Eterm* objv, int nobj)
 {
     Eterm *v;
     Uint sz;
@@ -3265,7 +3204,6 @@ reply_gc_info(void *vgcirp)
 
     reclaimed = esdp->gc_info.reclaimed;
     garbage_cols = esdp->gc_info.garbage_cols;
-#ifdef ERTS_DIRTY_SCHEDULERS
     /*
      * Add dirty schedulers info on requesting
      * schedulers info
@@ -3276,7 +3214,6 @@ reply_gc_info(void *vgcirp)
 	garbage_cols += dirty_gc.info.garbage_cols;
 	erts_mtx_unlock(&dirty_gc.mtx);
     }
-#endif
 
     sz = 0;
     hpp = NULL;
@@ -3309,16 +3246,15 @@ reply_gc_info(void *vgcirp)
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
  
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
     erts_proc_dec_refc(rp);
 
-    if (erts_smp_atomic32_dec_read_nob(&gcirp->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&gcirp->refc) == 0)
 	gcireq_free(vgcirp);
 }
 
-void erts_sub_binary_to_heap_binary(Eterm **pp, Eterm **hpp, Eterm *orig) {
-    Eterm *ptr = *pp;
+Eterm* erts_sub_binary_to_heap_binary(Eterm *ptr, Eterm **hpp, Eterm *orig) {
     Eterm *htop = *hpp;
     Eterm gval;
     ErlSubBin *sb = (ErlSubBin *)ptr;
@@ -3346,7 +3282,7 @@ void erts_sub_binary_to_heap_binary(Eterm **pp, Eterm **hpp, Eterm *orig) {
     htop += heap_bin_size(sb->size);
 
     *hpp = htop;
-    *pp  = ptr;
+    return ptr;
 }
 
 
@@ -3365,18 +3301,16 @@ erts_gc_info_request(Process *c_p)
     gcirp->proc = c_p;
     gcirp->ref = STORE_NC(&hp, NULL, ref);
     gcirp->req_sched = esdp->no;
-    erts_smp_atomic32_init_nob(&gcirp->refc,
+    erts_atomic32_init_nob(&gcirp->refc,
 			       (erts_aint32_t) erts_no_schedulers);
 
     erts_proc_add_refc(c_p, (Sint) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
 					  reply_gc_info,
 					  (void *) gcirp);
-#endif
 
     reply_gc_info((void *) gcirp);
 
@@ -3424,7 +3358,6 @@ erts_process_gc_info(Process *p, Uint *sizep, Eterm **hpp,
     };
 
     Eterm res = THE_NON_VALUE;
-    ErtsMessage *mp;
 
     ERTS_CT_ASSERT(sizeof(values)/sizeof(*values) == sizeof(tags)/sizeof(*tags));
     ERTS_CT_ASSERT(sizeof(values)/sizeof(*values) == ERTS_PROCESS_GC_INFO_MAX_TERMS);
@@ -3443,9 +3376,15 @@ erts_process_gc_info(Process *p, Uint *sizep, Eterm **hpp,
            be the same as adding old_heap_block_size + heap_block_size
            + mbuf_size.
         */
-        for (mp = p->msg.first; mp; mp = mp->next)
-            if (mp->data.attached)
-                values[2] += erts_msg_attached_data_size(mp);
+        ERTS_FOREACH_SIG_PRIVQS(
+            p, mp,
+            {
+                if (ERTS_SIG_IS_MSG(mp)
+                    && mp->data.attached
+                    && mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
+                    values[2] += erts_msg_attached_data_size(mp);
+                }
+            });
     }
 
     res = erts_bld_atom_uword_2tup_list(hpp,
@@ -3519,7 +3458,7 @@ erts_max_heap_size_map(Sint max_heap_size, Uint max_heap_flags,
                        Eterm **hpp, Uint *sz)
 {
     if (!hpp) {
-        *sz += (2*3 + 1 + MAP_HEADER_FLATMAP_SZ);
+        *sz += ERTS_MAX_HEAP_SIZE_MAP_SZ;
         return THE_NON_VALUE;
     } else {
         Eterm *hp = *hpp;
@@ -3663,12 +3602,12 @@ erts_check_off_heap2(Process *p, Eterm *htop)
 	    refc = erts_refc_read(&u.pb->val->intern.refc, 1);
 	    break;
 	case FUN_SUBTAG:
-	    refc = erts_smp_refc_read(&u.fun->fe->refc, 1);
+	    refc = erts_refc_read(&u.fun->fe->refc, 1);
 	    break;
 	case EXTERNAL_PID_SUBTAG:
 	case EXTERNAL_PORT_SUBTAG:
 	case EXTERNAL_REF_SUBTAG:
-	    refc = erts_smp_refc_read(&u.ext->node->refc, 1);
+	    refc = erts_refc_read(&u.ext->node->refc, 1);
 	    break;
 	case REF_SUBTAG:
 	    ASSERT(is_magic_ref_thing(u.hdr));
diff --git a/erts/emulator/beam/erl_gc.h b/erts/emulator/beam/erl_gc.h
index 6a529b8443..b9b1ed728c 100644
--- a/erts/emulator/beam/erl_gc.h
+++ b/erts/emulator/beam/erl_gc.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2007-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2007-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -33,14 +33,15 @@
 
 #define IS_MOVED_BOXED(x)	(!is_header((x)))
 #define IS_MOVED_CONS(x)	(is_non_value((x)))
-void erts_sub_binary_to_heap_binary(Eterm **pp, Eterm **hpp, Eterm *orig);
+Eterm* erts_sub_binary_to_heap_binary(Eterm *ptr, Eterm **hpp, Eterm *orig);
 
-ERTS_GLB_INLINE void move_cons(Eterm **pp, Eterm car, Eterm **hpp, Eterm *orig);
+ERTS_GLB_INLINE void move_cons(Eterm *ERTS_RESTRICT ptr, Eterm car, Eterm **hpp,
+                               Eterm *orig);
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
-ERTS_GLB_INLINE void move_cons(Eterm **pp, Eterm car, Eterm **hpp, Eterm *orig)
+ERTS_GLB_INLINE void move_cons(Eterm *ERTS_RESTRICT ptr, Eterm car, Eterm **hpp,
+                               Eterm *orig)
 {
-    Eterm *ptr  = *pp;
-    Eterm *htop = *hpp;
+    Eterm *ERTS_RESTRICT htop = *hpp;
     Eterm gval;
 
     htop[0] = car;               /* copy car */
@@ -53,14 +54,15 @@ ERTS_GLB_INLINE void move_cons(Eterm **pp, Eterm car, Eterm **hpp, Eterm *orig)
 }
 #endif
 
-ERTS_GLB_INLINE void move_boxed(Eterm **pp, Eterm hdr, Eterm **hpp, Eterm *orig);
+ERTS_GLB_INLINE Eterm* move_boxed(Eterm *ERTS_RESTRICT ptr, Eterm hdr, Eterm **hpp,
+                                  Eterm *orig);
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
-ERTS_GLB_INLINE void move_boxed(Eterm **pp, Eterm hdr, Eterm **hpp, Eterm *orig)
+ERTS_GLB_INLINE Eterm* move_boxed(Eterm *ERTS_RESTRICT ptr, Eterm hdr, Eterm **hpp,
+                                  Eterm *orig)
 {
     Eterm gval;
     Sint nelts;
-    Eterm *ptr = *pp;
-    Eterm *htop = *hpp;
+    Eterm *ERTS_RESTRICT htop = *hpp;
 
     ASSERT(is_header(hdr));
     nelts = header_arity(hdr);
@@ -71,8 +73,7 @@ ERTS_GLB_INLINE void move_boxed(Eterm **pp, Eterm hdr, Eterm **hpp, Eterm *orig)
             /* convert sub-binary to heap-binary if applicable */
             if (sb->bitsize == 0 && sb->bitoffs == 0 &&
                 sb->is_writable == 0 && sb->size <= sizeof(Eterm) * 3) {
-                erts_sub_binary_to_heap_binary(pp, hpp, orig);
-                return;
+                return erts_sub_binary_to_heap_binary(ptr, hpp, orig);
             }
         }
         nelts++;
@@ -90,7 +91,7 @@ ERTS_GLB_INLINE void move_boxed(Eterm **pp, Eterm hdr, Eterm **hpp, Eterm *orig)
     while (nelts--) *htop++ = *ptr++;
 
     *hpp = htop;
-    *pp  = ptr;
+    return ptr;
 }
 #endif
 
@@ -153,6 +154,8 @@ typedef struct {
   Uint64 garbage_cols;
 } ErtsGCInfo;
 
+#define ERTS_MAX_HEAP_SIZE_MAP_SZ (2*3 + 1 + MAP_HEADER_FLATMAP_SZ)
+
 #define ERTS_PROCESS_GC_INFO_MAX_TERMS (11)  /* number of elements in process_gc_info*/
 #define ERTS_PROCESS_GC_INFO_MAX_SIZE                                   \
     (ERTS_PROCESS_GC_INFO_MAX_TERMS * (2/*cons*/ + 3/*2-tuple*/ + BIG_UINT_HEAP_SIZE))
@@ -180,6 +183,8 @@ void erts_free_heap_frags(struct process* p);
 Eterm erts_max_heap_size_map(Sint, Uint, Eterm **, Uint *);
 int erts_max_heap_size(Eterm, Uint *, Uint *);
 void erts_deallocate_young_generation(Process *c_p);
+void erts_copy_one_frag(Eterm** hpp, ErlOffHeap* off_heap,
+                        ErlHeapFragment *bp, Eterm *refs, int nrefs);
 #if defined(DEBUG) || defined(ERTS_OFFHEAP_DEBUG)
 int erts_dbg_within_proc(Eterm *ptr, Process *p, Eterm* real_htop);
 #endif
diff --git a/erts/emulator/beam/erl_goodfit_alloc.c b/erts/emulator/beam/erl_goodfit_alloc.c
index a38f6c7daf..e3ba67f0af 100644
--- a/erts/emulator/beam/erl_goodfit_alloc.c
+++ b/erts/emulator/beam/erl_goodfit_alloc.c
@@ -508,7 +508,7 @@ static struct {
 
 static void ERTS_INLINE atom_init(Eterm *atom, char *name)
 {
-    *atom = am_atom_put(name, strlen(name));
+    *atom = am_atom_put(name, sys_strlen(name));
 }
 #define AM_INIT(AM) atom_init(&am.AM, #AM)
 
diff --git a/erts/emulator/beam/erl_hl_timer.c b/erts/emulator/beam/erl_hl_timer.c
index 6e5cc7b801..6ec6f8065e 100644
--- a/erts/emulator/beam/erl_hl_timer.c
+++ b/erts/emulator/beam/erl_hl_timer.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2015-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2015-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -38,6 +38,7 @@
 #define ERTS_WANT_TIMER_WHEEL_API
 #include "erl_time.h"
 #include "erl_hl_timer.h"
+#include "erl_proc_sig_queue.h"
 #ifdef ERTS_MAGIC_REF_BIF_TIMERS
 #include "erl_binary.h"
 #endif
@@ -96,13 +97,6 @@ typedef enum {
 
 #define ERTS_BIF_TIMER_SHORT_TIME 5000
 
-#ifdef ERTS_SMP
-#  define ERTS_HLT_SMP_MEMBAR_LoadLoad_LoadStore \
-    ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore)
-#else
-#  define ERTS_HLT_SMP_MEMBAR_LoadLoad_LoadStore
-#endif
-
 /* Bit 0 to 10 contains scheduler id (see mask below) */
 #define ERTS_TMR_ROFLG_HLT		(((Uint32) 1) << 11)
 #define ERTS_TMR_ROFLG_BIF_TMR		(((Uint32) 1) << 12)
@@ -159,7 +153,7 @@ typedef struct {
 
 typedef struct {
     Uint32 roflgs;
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
     union {
 	void *arg;
 	erts_atomic_t next;
@@ -200,7 +194,7 @@ struct ErtsBifTimer_ {
         ErtsTWTimer twt;
     } type;
     struct {
-        erts_smp_atomic32_t state;
+        erts_atomic32_t state;
 #ifdef ERTS_MAGIC_REF_BIF_TIMERS
         ErtsMagicBinary *mbin;
         ErtsHLTimerList proc_list;
@@ -269,7 +263,6 @@ typedef struct {
     erts_atomic_t last;
 } ErtsHLTCncldTmrQTail;
 
-#ifdef ERTS_SMP
 
 typedef struct {
     /*
@@ -301,7 +294,6 @@ typedef struct {
     } head;
 } ErtsHLTCncldTmrQ;
 
-#endif /* ERTS_SMP */
 
 typedef struct {
     ErtsHLTimer *root;
@@ -309,9 +301,7 @@ typedef struct {
 } ErtsYieldingTimeoutState;
 
 struct ErtsHLTimerService_ {
-#ifdef ERTS_SMP
     ErtsHLTCncldTmrQ canceled_queue;
-#endif
     ErtsHLTimer *time_tree;
 #ifndef ERTS_MAGIC_REF_BIF_TIMERS
     ErtsBifTimer *btm_tree;
@@ -720,9 +710,7 @@ proc_btm_list_foreach_destroy_yielding(ErtsBifTimer **list,
 
 #endif /* !ERTS_MAGIC_REF_BIF_TIMERS */
 
-#ifdef ERTS_SMP
 static void init_canceled_queue(ErtsHLTCncldTmrQ *cq);
-#endif
 
 void
 erts_hl_timer_init(void)
@@ -747,9 +735,7 @@ erts_create_timer_service(void)
     srv->yield = init_yield;
     erts_twheel_init_timer(&srv->service_timer);
 
-#ifdef ERTS_SMP
     init_canceled_queue(&srv->canceled_queue);
-#endif
 
     return srv;
 }
@@ -791,13 +777,13 @@ get_time_left(ErtsSchedulerData *esdp, ErtsMonotonicTime timeout_pos)
 static ERTS_INLINE int
 proc_timeout_common(Process *proc, void *tmr)
 {
-    if (tmr == (void *) erts_smp_atomic_cmpxchg_mb(&proc->common.timer,
+    if (tmr == (void *) erts_atomic_cmpxchg_mb(&proc->common.timer,
 						   ERTS_PTMR_TIMEDOUT,
 						   (erts_aint_t) tmr)) {
 	erts_aint32_t state;
-	erts_smp_proc_lock(proc, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	state = erts_smp_atomic32_read_acqb(&proc->state);
-	erts_smp_proc_unlock(proc, ERTS_PROC_LOCKS_MSG_RECEIVE);
+	erts_proc_lock(proc, ERTS_PROC_LOCKS_MSG_RECEIVE);
+	state = erts_atomic32_read_acqb(&proc->state);
+	erts_proc_unlock(proc, ERTS_PROC_LOCKS_MSG_RECEIVE);
 	if (!(state & (ERTS_PSFLG_ACTIVE|ERTS_PSFLG_EXITING)))
 	    erts_schedule_process(proc, state, 0);
 	return 1;
@@ -808,7 +794,7 @@ proc_timeout_common(Process *proc, void *tmr)
 static ERTS_INLINE int
 port_timeout_common(Port *port, void *tmr)
 {
-    if (tmr == (void *) erts_smp_atomic_cmpxchg_mb(&port->common.timer,
+    if (tmr == (void *) erts_atomic_cmpxchg_mb(&port->common.timer,
 						   ERTS_PTMR_TIMEDOUT,
 						   (erts_aint_t) tmr)) {
 	erts_port_task_schedule(port->common.id,
@@ -821,24 +807,24 @@ port_timeout_common(Port *port, void *tmr)
 
 #ifdef ERTS_MAGIC_REF_BIF_TIMERS
 
-static erts_smp_atomic_t *
+static erts_atomic_t *
 mbin_to_btmref__(ErtsMagicBinary *mbin)
 {
-    return erts_smp_binary_to_magic_indirection((Binary *) mbin);
+    return erts_binary_to_magic_indirection((Binary *) mbin);
 }
 
 static ERTS_INLINE void
 magic_binary_init(ErtsMagicBinary *mbin, ErtsBifTimer *tmr)
 {
-    erts_smp_atomic_t *aptr = mbin_to_btmref__(mbin);
-    erts_smp_atomic_init_nob(aptr, (erts_aint_t) tmr);
+    erts_atomic_t *aptr = mbin_to_btmref__(mbin);
+    erts_atomic_init_nob(aptr, (erts_aint_t) tmr);
 }
 
 static ERTS_INLINE ErtsBifTimer *
 magic_binary_to_btm(ErtsMagicBinary *mbin)
 {
-    erts_smp_atomic_t *aptr = mbin_to_btmref__(mbin);
-    ErtsBifTimer *tmr = (ErtsBifTimer *) erts_smp_atomic_read_nob(aptr);
+    erts_atomic_t *aptr = mbin_to_btmref__(mbin);
+    ErtsBifTimer *tmr = (ErtsBifTimer *) erts_atomic_read_nob(aptr);
     ERTS_HLT_ASSERT(!tmr || tmr->btm.mbin == mbin);
     return tmr;
 }
@@ -884,7 +870,7 @@ init_btm_specifics(ErtsSchedulerData *esdp,
     btm_rbt_insert(&esdp->timer_service->btm_tree, tmr);
 #endif
 
-    erts_smp_atomic32_init_nob(&tmr->btm.state, ERTS_TMR_STATE_ACTIVE);
+    erts_atomic32_init_nob(&tmr->btm.state, ERTS_TMR_STATE_ACTIVE);
     return refc; /* refc from magic binary... */
 }
 
@@ -917,10 +903,10 @@ timer_pre_dec_refc(ErtsTimer *tmr)
 {
 #ifdef ERTS_HLT_DEBUG
     erts_aint_t refc;
-    refc = erts_smp_atomic32_dec_read_nob(&tmr->head.refc);
+    refc = erts_atomic32_dec_read_nob(&tmr->head.refc);
     ERTS_HLT_ASSERT(refc > 0);
 #else
-    erts_smp_atomic32_dec_nob(&tmr->head.refc);
+    erts_atomic32_dec_nob(&tmr->head.refc);
 #endif
 }
 
@@ -969,8 +955,8 @@ schedule_tw_timer_destroy(ErtsTWTimer *tmr)
 static ERTS_INLINE void
 tw_timer_dec_refc(ErtsTWTimer *tmr)
 {
-    if (erts_smp_atomic32_dec_read_relb(&tmr->head.refc) == 0) {
-	ERTS_HLT_SMP_MEMBAR_LoadLoad_LoadStore;
+    if (erts_atomic32_dec_read_relb(&tmr->head.refc) == 0) {
+        ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
 	schedule_tw_timer_destroy(tmr);
     }
 }
@@ -1114,7 +1100,7 @@ create_tw_timer(ErtsSchedulerData *esdp,
 	return NULL;
     }
 
-    erts_smp_atomic32_init_nob(&tmr->head.refc, refc);
+    erts_atomic32_init_nob(&tmr->head.refc, refc);
 
     erts_twheel_set_timer(esdp->timer_wheel,
 			  &tmr->u.tw_tmr,
@@ -1147,7 +1133,7 @@ schedule_hl_timer_destroy(ErtsHLTimer *tmr, Uint32 roflgs)
      * at once...
      */
 
-    ERTS_HLT_ASSERT(erts_smp_atomic32_read_nob(&tmr->head.refc) == 0);
+    ERTS_HLT_ASSERT(erts_atomic32_read_nob(&tmr->head.refc) == 0);
 
     if (roflgs & ERTS_TMR_ROFLG_REG_NAME) {
 	ERTS_HLT_ASSERT(is_atom(tmr->head.receiver.name));
@@ -1179,14 +1165,13 @@ schedule_hl_timer_destroy(ErtsHLTimer *tmr, Uint32 roflgs)
 static ERTS_INLINE void
 hl_timer_dec_refc(ErtsHLTimer *tmr, Uint32 roflgs)
 {
-    if (erts_smp_atomic32_dec_read_relb(&tmr->head.refc) == 0) {
-	ERTS_HLT_SMP_MEMBAR_LoadLoad_LoadStore;
+    if (erts_atomic32_dec_read_relb(&tmr->head.refc) == 0) {
+        ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
 	schedule_hl_timer_destroy(tmr, roflgs);
     }
 }
 
 static void hlt_service_timeout(void *vesdp);
-#ifdef ERTS_SMP
 static void handle_canceled_queue(ErtsSchedulerData *esdp,
 				  ErtsHLTCncldTmrQ *cq,
 				  int use_limit,
@@ -1194,12 +1179,11 @@ static void handle_canceled_queue(ErtsSchedulerData *esdp,
 				  int *need_thr_progress,
 				  ErtsThrPrgrVal *thr_prgr_p,
 				  int *need_more_work);
-#endif
 
 static ERTS_INLINE void
 check_canceled_queue(ErtsSchedulerData *esdp, ErtsHLTimerService *srv)
 {
-#if defined(ERTS_SMP) && ERTS_TMR_CHECK_CANCEL_ON_CREATE
+#if ERTS_TMR_CHECK_CANCEL_ON_CREATE
     ErtsHLTCncldTmrQ *cq = &srv->canceled_queue;
     if (cq->head.first != cq->head.unref_end)
 	handle_canceled_queue(esdp, cq, 1,
@@ -1219,14 +1203,14 @@ bif_timer_ref_destructor(Binary *unused)
 static ERTS_INLINE void
 btm_clear_magic_binary(ErtsBifTimer *tmr)
 {
-    erts_smp_atomic_t *aptr = mbin_to_btmref__(tmr->btm.mbin);
+    erts_atomic_t *aptr = mbin_to_btmref__(tmr->btm.mbin);
     Uint32 roflgs = tmr->type.head.roflgs;
 #ifdef ERTS_HLT_DEBUG
-    erts_aint_t tval = erts_smp_atomic_xchg_nob(aptr,
+    erts_aint_t tval = erts_atomic_xchg_nob(aptr,
                                                 (erts_aint_t) NULL);
     ERTS_HLT_ASSERT(tval == (erts_aint_t) tmr);
 #else
-    erts_smp_atomic_set_nob(aptr, (erts_aint_t) NULL);
+    erts_atomic_set_nob(aptr, (erts_aint_t) NULL);
 #endif
     if (roflgs & ERTS_TMR_ROFLG_HLT)
         hl_timer_dec_refc(&tmr->type.hlt, roflgs);
@@ -1246,7 +1230,7 @@ bif_timer_timeout(ErtsHLTimerService *srv,
     ERTS_HLT_ASSERT(tmr->type.head.roflgs == roflgs);    
     ERTS_HLT_ASSERT(roflgs & ERTS_TMR_ROFLG_BIF_TMR);
 
-    state = erts_smp_atomic32_cmpxchg_acqb(&tmr->btm.state,
+    state = erts_atomic32_cmpxchg_acqb(&tmr->btm.state,
 					   ERTS_TMR_STATE_TIMED_OUT,
 					   ERTS_TMR_STATE_ACTIVE);
 
@@ -1279,7 +1263,7 @@ bif_timer_timeout(ErtsHLTimerService *srv,
             tmr->btm.bp = NULL;
             erts_queue_message(proc, 0, mp, tmr->btm.message,
                                am_clock_service);
-            erts_smp_proc_lock(proc, ERTS_PROC_LOCK_BTM);
+            erts_proc_lock(proc, ERTS_PROC_LOCK_BTM);
             /* If the process is exiting do not disturb the cleanup... */
             if (!ERTS_PROC_IS_EXITING(proc)) {
 #ifdef ERTS_MAGIC_REF_BIF_TIMERS
@@ -1295,7 +1279,7 @@ bif_timer_timeout(ErtsHLTimerService *srv,
                 }
 #endif
             }
-            erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
+            erts_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
             if (dec_refc)
                 timer_pre_dec_refc((ErtsTimer *) tmr);
         }
@@ -1432,7 +1416,7 @@ create_hl_timer(ErtsSchedulerData *esdp,
     }
 
     tmr->head.roflgs = roflgs;
-    erts_smp_atomic32_init_nob(&tmr->head.refc, refc);
+    erts_atomic32_init_nob(&tmr->head.refc, refc);
 
     if (!srv->next_timeout
 	|| tmr->timeout < srv->next_timeout->timeout) {
@@ -1664,7 +1648,6 @@ cleanup_sched_local_canceled_timer(ErtsSchedulerData *esdp,
     }
 }
 
-#ifdef ERTS_SMP
 
 static void
 init_canceled_queue(ErtsHLTCncldTmrQ *cq)
@@ -1794,7 +1777,7 @@ cq_check_incoming(ErtsSchedulerData *esdp, ErtsHLTCncldTmrQ *cq)
 	cq->head.next.thr_progress_reached = 1;
 	/* Move unreferenced end pointer forward... */
 
-	ERTS_HLT_SMP_MEMBAR_LoadLoad_LoadStore;
+        ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
 
 	cq->head.unref_end = cq->head.next.unref_end;
 
@@ -1887,31 +1870,24 @@ erts_handle_canceled_timers(void *vesdp,
 			  need_more_work);
 }
 
-#endif /* ERTS_SMP */
 
 static void
 queue_canceled_timer(ErtsSchedulerData *esdp, int rsched_id, ErtsTimer *tmr)
 {
-#ifdef ERTS_SMP
     ErtsHLTCncldTmrQ *cq;
     cq = &ERTS_SCHEDULER_IX(rsched_id-1)->timer_service->canceled_queue;
     if (cq_enqueue(cq, tmr, rsched_id - (int) esdp->no))
 	erts_notify_canceled_timer(esdp, rsched_id);
-#else
-    ERTS_INTERNAL_ERROR("Unexpected enqueue of canceled timer");
-#endif
 }
 
 static void
 continue_cancel_ptimer(ErtsSchedulerData *esdp, ErtsTimer *tmr)
 {
-#ifdef ERTS_SMP
     Uint32 sid = (tmr->head.roflgs & ERTS_TMR_ROFLG_SID_MASK);
 
     if (esdp->no != sid)
 	queue_canceled_timer(esdp, sid, tmr);
     else
-#endif
 	cleanup_sched_local_canceled_timer(esdp, tmr);
 }
 
@@ -1997,7 +1973,7 @@ setup_bif_timer(Process *c_p, int twheel, ErtsMonotonicTime timeout_pos,
 #else
 	    proc_btm_rbt_insert(&proc->bif_timers, tmr);
 #endif
-	    erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
+	    erts_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
             tmr->type.head.receiver.proc = proc;
 	}
     }
@@ -2018,7 +1994,7 @@ cancel_bif_timer(ErtsBifTimer *tmr)
     Uint32 roflgs;
     int res;
 
-    state = erts_smp_atomic32_cmpxchg_acqb(&tmr->btm.state,
+    state = erts_atomic32_cmpxchg_acqb(&tmr->btm.state,
 					   ERTS_TMR_STATE_CANCELED,
 					   ERTS_TMR_STATE_ACTIVE);
     if (state != ERTS_TMR_STATE_ACTIVE)
@@ -2040,7 +2016,7 @@ cancel_bif_timer(ErtsBifTimer *tmr)
         proc = tmr->type.head.receiver.proc;
 	ERTS_HLT_ASSERT(!(tmr->type.head.roflgs & ERTS_TMR_ROFLG_REG_NAME));
 
-	erts_smp_proc_lock(proc, ERTS_PROC_LOCK_BTM);
+	erts_proc_lock(proc, ERTS_PROC_LOCK_BTM);
 	/*
 	 * If process is exiting, let it clean up
 	 * the btm tree by itself (it may be in
@@ -2059,7 +2035,7 @@ cancel_bif_timer(ErtsBifTimer *tmr)
 	    res = 1;
 	}
 #endif
-	erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
+	erts_proc_unlock(proc, ERTS_PROC_LOCK_BTM);
     }
 
     return res;
@@ -2082,7 +2058,7 @@ access_btm(ErtsBifTimer *tmr, Uint32 sid, ErtsSchedulerData *esdp, int cancel)
                : erts_tweel_read_timeout(&tmr->type.twt.u.tw_tmr));
 
     if (!cancel) {
-        erts_aint32_t state = erts_smp_atomic32_read_acqb(&tmr->btm.state);
+        erts_aint32_t state = erts_atomic32_read_acqb(&tmr->btm.state);
         if (state == ERTS_TMR_STATE_ACTIVE)
             return get_time_left(esdp, timeout);
         return -1;
@@ -2176,7 +2152,7 @@ send_async_info(Process *proc, ErtsProcLocks initial_locks,
 
     locks &= ~initial_locks;
     if (locks)
-        erts_smp_proc_unlock(proc, locks);
+        erts_proc_unlock(proc, locks);
 
     return am_ok;
 }
@@ -2262,7 +2238,7 @@ send_sync_info(Process *proc, ErtsProcLocks initial_locks,
 
     locks &= ~initial_locks;
     if (locks)
-        erts_smp_proc_unlock(proc, locks);
+        erts_proc_unlock(proc, locks);
 
     return am_ok;
 }
@@ -2376,9 +2352,9 @@ try_access_sched_remote_btm(ErtsSchedulerData *esdp,
      * Check if the timer is aimed at current
      * process...
      */
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_BTM);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_BTM);
     tmr = proc_btm_rbt_lookup(c_p->bif_timers, trefn);
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_BTM);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_BTM);
     if (!tmr)
 	return 0;
 
@@ -2419,7 +2395,7 @@ no_timer_result(Process *c_p, Eterm tref, int cancel, int async, int info)
     erts_queue_message(c_p, locks, mp, msg, am_clock_service);
     locks &= ~ERTS_PROC_LOCK_MAIN;
     if (locks)
-        erts_smp_proc_unlock(c_p, locks);
+        erts_proc_unlock(c_p, locks);
     return am_ok;
 }
 
@@ -2495,28 +2471,21 @@ access_bif_timer(Process *c_p, Eterm tref, int cancel, int async, int info)
 	    req->rrefn[1] = rrefn[1];
 	    req->rrefn[2] = rrefn[2];
 
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-
-	    if (ERTS_PROC_PENDING_EXIT(c_p))
-		ERTS_VBUMP_ALL_REDS(c_p);
-	    else {
-		/*
-		 * Caller needs to wait for a message containing
-		 * the ref that we just created. No such message
-		 * can exist in callers message queue at this time.
-		 * We therefore move the save pointer of the
-		 * callers message queue to the end of the queue.
-		 *
-		 * NOTE: It is of vital importance that the caller
-		 *       immediately do a receive unconditionaly
-		 *       waiting for the message with the reference;
-		 *       otherwise, next receive will *not* work
-		 *       as expected!
-		 */
-		ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-		c_p->msg.save = c_p->msg.last;
-	    }
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+            /*
+             * Caller needs to wait for a message containing
+             * the ref that we just created. No such message
+             * can exist in callers message queue at this time.
+             * We therefore move the save pointer of the
+             * callers message queue to the end of the queue.
+             *
+             * NOTE: It is of vital importance that the caller
+             *       immediately do a receive unconditionaly
+             *       waiting for the message with the reference;
+             *       otherwise, next receive will *not* work
+             *       as expected!
+             */
+            ERTS_RECV_MARK_SAVE(c_p);
+            ERTS_RECV_MARK_SET(c_p);
 
 	    ERTS_BIF_PREP_TRAP1(ret, erts_await_result, c_p, rref);
 	}
@@ -2606,7 +2575,7 @@ exit_cancel_bif_timer(ErtsBifTimer *tmr, void *vesdp)
     erts_aint_t state;
     int is_hlt;
 
-    state = erts_smp_atomic32_cmpxchg_acqb(&tmr->btm.state,
+    state = erts_atomic32_cmpxchg_acqb(&tmr->btm.state,
 					   ERTS_TMR_STATE_CANCELED,
 					   ERTS_TMR_STATE_ACTIVE);
 
@@ -2992,7 +2961,7 @@ set_proc_timer_common(Process *c_p, ErtsSchedulerData *esdp, Sint64 tmo,
                                     ERTS_TMR_PROC, (void *) c_p,
                                     c_p->common.id, THE_NON_VALUE,
                                     NULL, NULL, NULL);
-	erts_smp_atomic_set_relb(&c_p->common.timer, (erts_aint_t) tmr);
+	erts_atomic_set_relb(&c_p->common.timer, (erts_aint_t) tmr);
     }
 }
 
@@ -3003,7 +2972,7 @@ erts_set_proc_timer_term(Process *c_p, Eterm etmo)
     ErtsMonotonicTime tmo, timeout_pos;
     int short_time, tres;
 
-    ERTS_HLT_ASSERT(erts_smp_atomic_read_nob(&c_p->common.timer)
+    ERTS_HLT_ASSERT(erts_atomic_read_nob(&c_p->common.timer)
 		    == ERTS_PTMR_NONE);
 
     tres = parse_timeout_pos(esdp, etmo, &tmo, 0,
@@ -3023,7 +2992,7 @@ erts_set_proc_timer_uword(Process *c_p, UWord tmo)
 {
     ErtsSchedulerData *esdp = erts_proc_sched_data(c_p);
 
-    ERTS_HLT_ASSERT(erts_smp_atomic_read_nob(&c_p->common.timer)
+    ERTS_HLT_ASSERT(erts_atomic_read_nob(&c_p->common.timer)
 		    == ERTS_PTMR_NONE);
 
 #ifndef ARCH_32
@@ -3046,13 +3015,13 @@ void
 erts_cancel_proc_timer(Process *c_p)
 {
     erts_aint_t tval;
-    tval = erts_smp_atomic_xchg_acqb(&c_p->common.timer,
+    tval = erts_atomic_xchg_acqb(&c_p->common.timer,
 				     ERTS_PTMR_NONE);
     c_p->flags &= ~(F_INSLPQUEUE|F_TIMO);
     if (tval == ERTS_PTMR_NONE)
 	return;
     if (tval == ERTS_PTMR_TIMEDOUT) {
-	erts_smp_atomic_set_nob(&c_p->common.timer, ERTS_PTMR_NONE);
+	erts_atomic_set_nob(&c_p->common.timer, ERTS_PTMR_NONE);
 	return;
     }
     continue_cancel_ptimer(erts_proc_sched_data(c_p),
@@ -3067,7 +3036,7 @@ erts_set_port_timer(Port *c_prt, Sint64 tmo)
     ErtsMonotonicTime timeout_pos;
     ErtsCreateTimerFunc create_timer;
 
-    if (erts_smp_atomic_read_nob(&c_prt->common.timer) != ERTS_PTMR_NONE)
+    if (erts_atomic_read_nob(&c_prt->common.timer) != ERTS_PTMR_NONE)
 	erts_cancel_port_timer(c_prt);
 
     check_canceled_queue(esdp, esdp->timer_service);
@@ -3080,14 +3049,14 @@ erts_set_port_timer(Port *c_prt, Sint64 tmo)
     tmr = (void *) create_timer(esdp, timeout_pos, 0, ERTS_TMR_PORT,
                                 (void *) c_prt, c_prt->common.id,
                                 THE_NON_VALUE, NULL, NULL, NULL);
-    erts_smp_atomic_set_relb(&c_prt->common.timer, (erts_aint_t) tmr);
+    erts_atomic_set_relb(&c_prt->common.timer, (erts_aint_t) tmr);
 }
 
 void
 erts_cancel_port_timer(Port *c_prt)
 {
     erts_aint_t tval;
-    tval = erts_smp_atomic_xchg_acqb(&c_prt->common.timer,
+    tval = erts_atomic_xchg_acqb(&c_prt->common.timer,
 				     ERTS_PTMR_NONE);
     if (tval == ERTS_PTMR_NONE)
 	return;
@@ -3095,7 +3064,7 @@ erts_cancel_port_timer(Port *c_prt)
 	while (!erts_port_task_is_scheduled(&c_prt->timeout_task))
 	    erts_thr_yield();
 	erts_port_task_abort(&c_prt->timeout_task);
-	erts_smp_atomic_set_nob(&c_prt->common.timer, ERTS_PTMR_NONE);
+	erts_atomic_set_nob(&c_prt->common.timer, ERTS_PTMR_NONE);
 	return;
     }
     continue_cancel_ptimer(erts_get_scheduler_data(),
@@ -3109,7 +3078,7 @@ erts_read_port_timer(Port *c_prt)
     erts_aint_t itmr;
     ErtsMonotonicTime timeout_pos;
 
-    itmr = erts_smp_atomic_read_acqb(&c_prt->common.timer);
+    itmr = erts_atomic_read_acqb(&c_prt->common.timer);
     if (itmr == ERTS_PTMR_NONE)
 	return (Sint64) -1;
     if (itmr == ERTS_PTMR_TIMEDOUT)
@@ -3246,7 +3215,7 @@ debug_btm_foreach(ErtsBifTimer *tmr, void *vbtmfd)
     if (!(tmr->type.head.roflgs & ERTS_TMR_ROFLG_BIF_TMR))
         return;
 #endif
-    if (erts_smp_atomic32_read_nob(&tmr->btm.state) == ERTS_TMR_STATE_ACTIVE) {
+    if (erts_atomic32_read_nob(&tmr->btm.state) == ERTS_TMR_STATE_ACTIVE) {
 	ErtsBTMForeachDebug *btmfd = (ErtsBTMForeachDebug *) vbtmfd;
         Eterm id = ((tmr->type.head.roflgs & ERTS_TMR_ROFLG_REG_NAME)
                     ? tmr->type.head.receiver.name
@@ -3284,7 +3253,7 @@ erts_debug_bif_timer_foreach(void (*func)(Eterm,
     btmfd.func = func;
     btmfd.arg = arg;
 
-    if (!erts_smp_thr_progress_is_blocking())
+    if (!erts_thr_progress_is_blocking())
 	ERTS_INTERNAL_ERROR("Not blocking thread progress");
 
     for (six = 0; six < erts_no_schedulers; six++) {
@@ -3375,7 +3344,7 @@ erts_debug_callback_timer_foreach(void (*tclbk)(void *),
     dfct.func = func;
     dfct.arg = arg;
 
-    if (!erts_smp_thr_progress_is_blocking())
+    if (!erts_thr_progress_is_blocking())
 	ERTS_INTERNAL_ERROR("Not blocking thread progress");
 
     for (six = 0; six < erts_no_schedulers; six++) {
diff --git a/erts/emulator/beam/erl_hl_timer.h b/erts/emulator/beam/erl_hl_timer.h
index ff31f04cb9..e6f5e8b67d 100644
--- a/erts/emulator/beam/erl_hl_timer.h
+++ b/erts/emulator/beam/erl_hl_timer.h
@@ -36,16 +36,16 @@ typedef struct ErtsHLTimerService_ ErtsHLTimerService;
 #define ERTS_PTMR_TIMEDOUT (ERTS_PTMR_NONE + ((erts_aint_t) 1))
 
 #define ERTS_PTMR_INIT(P) \
-    erts_smp_atomic_init_nob(&(P)->common.timer, ERTS_PTMR_NONE)
+    erts_atomic_init_nob(&(P)->common.timer, ERTS_PTMR_NONE)
 #define ERTS_PTMR_IS_SET(P) \
-    (ERTS_PTMR_NONE != erts_smp_atomic_read_nob(&(P)->common.timer))
+    (ERTS_PTMR_NONE != erts_atomic_read_nob(&(P)->common.timer))
 #define ERTS_PTMR_IS_TIMED_OUT(P) \
-    (ERTS_PTMR_TIMEDOUT == erts_smp_atomic_read_nob(&(P)->common.timer))
+    (ERTS_PTMR_TIMEDOUT == erts_atomic_read_nob(&(P)->common.timer))
 
 #define ERTS_PTMR_CLEAR(P)					\
     do {							\
 	ASSERT(ERTS_PTMR_IS_TIMED_OUT((P)));			\
-	erts_smp_atomic_set_nob(&(P)->common.timer,		\
+	erts_atomic_set_nob(&(P)->common.timer,		\
 				ERTS_PTMR_NONE);		\
     } while (0)
 
@@ -63,13 +63,11 @@ void erts_hl_timer_init(void);
 void erts_start_timer_callback(ErtsMonotonicTime,
 			       void (*)(void *),
 			       void *);
-#ifdef ERTS_SMP
 void
 erts_handle_canceled_timers(void *vesdp,
 			    int *need_thr_progress,
 			    ErtsThrPrgrVal *thr_prgr_p,
 			    int *need_more_work);
-#endif
 
 Uint erts_bif_timer_memory_size(void);
 void erts_print_bif_timer_info(fmtfn_t to, void *to_arg);
diff --git a/erts/emulator/beam/erl_init.c b/erts/emulator/beam/erl_init.c
index 5206d7564f..57c6c10c7f 100644
--- a/erts/emulator/beam/erl_init.c
+++ b/erts/emulator/beam/erl_init.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1997-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1997-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -37,7 +37,7 @@
 #include "erl_mseg.h"
 #include "erl_threads.h"
 #include "erl_hl_timer.h"
-#include "erl_instrument.h"
+#include "erl_mtrace.h"
 #include "erl_printf_term.h"
 #include "erl_misc_utils.h"
 #include "packet_parser.h"
@@ -49,6 +49,9 @@
 #include "erl_bif_unique.h"
 #define ERTS_WANT_TIMER_WHEEL_API
 #include "erl_time.h"
+#include "erl_check_io.h"
+#include "erl_osenv.h"
+#include "erl_proc_sig_queue.h"
 
 #ifdef HIPE
 #include "hipe_mode_switch.h"	/* for hipe_mode_switch_init() */
@@ -59,7 +62,7 @@
 #  include <sys/resource.h>
 #endif
 
-#define ERTS_DEFAULT_NO_ASYNC_THREADS	10
+#define ERTS_DEFAULT_NO_ASYNC_THREADS	1
 
 #define ERTS_DEFAULT_SCHED_STACK_SIZE   128
 #define ERTS_DEFAULT_DCPU_SCHED_STACK_SIZE 40
@@ -69,23 +72,17 @@
  * The variables below (prefixed with etp_) are for erts/etc/unix/etp-commands
  * only. Do not remove even though they aren't used elsewhere in the emulator!
  */
-#ifdef ERTS_SMP
 const int etp_smp_compiled = 1;
-#else
-const int etp_smp_compiled = 0;
-#endif
-#ifdef USE_THREADS
 const int etp_thread_compiled = 1;
-#else
-const int etp_thread_compiled = 0;
-#endif
 const char etp_erts_version[] = ERLANG_VERSION;
 const char etp_otp_release[] = ERLANG_OTP_RELEASE;
 const char etp_compile_date[] = ERLANG_COMPILE_DATE;
 const char etp_arch[] = ERLANG_ARCHITECTURE;
 #ifdef ERTS_ENABLE_KERNEL_POLL
+const int erts_use_kernel_poll = 1;
 const int etp_kernel_poll_support = 1;
 #else
+const int erts_use_kernel_poll = 0;
 const int etp_kernel_poll_support = 0;
 #endif
 #if defined(ARCH_64)
@@ -131,6 +128,8 @@ const Eterm etp_hole_marker = 0;
 
 static int modified_sched_thread_suggested_stack_size = 0;
 
+Eterm erts_init_process_id;
+
 /*
  * Note about VxWorks: All variables must be initialized by executable code,
  * not by an initializer. Otherwise a new instance of the emulator will
@@ -156,20 +155,10 @@ static void erl_init(int ncpu,
 
 static erts_atomic_t exiting;
 
-#ifdef ERTS_SMP
-erts_smp_atomic32_t erts_writing_erl_crash_dump;
+erts_atomic32_t erts_writing_erl_crash_dump;
 erts_tsd_key_t erts_is_crash_dumping_key;
-#else
-volatile int erts_writing_erl_crash_dump = 0;
-#endif
 int erts_initialized = 0;
 
-#if defined(USE_THREADS) && !defined(ERTS_SMP)
-erts_tid_t erts_main_thread;
-#endif
-
-int erts_use_sender_punish;
-
 /*
  * Configurable parameters.
  */
@@ -185,7 +174,7 @@ int erts_backtrace_depth;	/* How many functions to show in a backtrace
 				 * in error codes.
 				 */
 
-erts_smp_atomic32_t erts_max_gen_gcs;
+erts_atomic32_t erts_max_gen_gcs;
 
 Eterm erts_error_logger_warnings; /* What to map warning logs to, am_error, 
 				     am_info or am_warning, am_error is 
@@ -195,11 +184,9 @@ int erts_compat_rel;
 
 static int no_schedulers;
 static int no_schedulers_online;
-#ifdef ERTS_DIRTY_SCHEDULERS
 static int no_dirty_cpu_schedulers;
 static int no_dirty_cpu_schedulers_online;
 static int no_dirty_io_schedulers;
-#endif
 
 #ifdef DEBUG
 Uint32 verbose;             /* See erl_debug.h for information about verbose */
@@ -220,16 +207,16 @@ int erts_no_line_info = 0;	/* -L: Don't load line information */
  */
 
 ErtsModifiedTimings erts_modified_timings[] = {
-    /* 0 */	{make_small(0), CONTEXT_REDS, INPUT_REDUCTIONS},
-    /* 1 */	{make_small(0), (3*CONTEXT_REDS)/4, 2*INPUT_REDUCTIONS},
-    /* 2 */	{make_small(0), CONTEXT_REDS/2, INPUT_REDUCTIONS/2},
-    /* 3 */	{make_small(0), (7*CONTEXT_REDS)/8, 3*INPUT_REDUCTIONS},
-    /* 4 */	{make_small(0), CONTEXT_REDS/3, 3*INPUT_REDUCTIONS},
-    /* 5 */	{make_small(0), (10*CONTEXT_REDS)/11, INPUT_REDUCTIONS/2},
-    /* 6 */	{make_small(1), CONTEXT_REDS/4, 2*INPUT_REDUCTIONS},
-    /* 7 */	{make_small(1), (5*CONTEXT_REDS)/7, INPUT_REDUCTIONS/3},
-    /* 8 */	{make_small(10), CONTEXT_REDS/5, 3*INPUT_REDUCTIONS},
-    /* 9 */	{make_small(10), (6*CONTEXT_REDS)/7, INPUT_REDUCTIONS/4}
+    /* 0 */	{make_small(0), CONTEXT_REDS},
+    /* 1 */	{make_small(0), (3*CONTEXT_REDS)/4},
+    /* 2 */	{make_small(0), CONTEXT_REDS/2},
+    /* 3 */	{make_small(0), (7*CONTEXT_REDS)/8},
+    /* 4 */	{make_small(0), CONTEXT_REDS/3},
+    /* 5 */	{make_small(0), (10*CONTEXT_REDS)/11},
+    /* 6 */	{make_small(1), CONTEXT_REDS/4},
+    /* 7 */	{make_small(1), (5*CONTEXT_REDS)/7},
+    /* 8 */	{make_small(10), CONTEXT_REDS/5},
+    /* 9 */	{make_small(10), (6*CONTEXT_REDS)/7}
 };
 
 #define ERTS_MODIFIED_TIMING_LEVELS \
@@ -255,7 +242,7 @@ progname(char *fullname)
 {
     int i;
     
-    i = strlen(fullname);
+    i = sys_strlen(fullname);
     while (i >= 0) {
 	if ((fullname[i] != '/') && (fullname[i] != '\\')) 
 	    i--;
@@ -321,18 +308,18 @@ erl_init(int ncpu,
 	 int node_tab_delete_delay,
 	 ErtsDbSpinCount db_spin_count)
 {
+    erts_monitor_link_init();
+    erts_proc_sig_queue_init();
     erts_bif_unique_init();
-    erts_init_monitors();
     erts_init_time(time_correction, time_warp_mode);
     erts_init_sys_common_misc();
     erts_init_process(ncpu, proc_tab_sz, legacy_proc_tab);
     erts_init_scheduling(no_schedulers,
-			 no_schedulers_online
-#ifdef ERTS_DIRTY_SCHEDULERS
-			 , no_dirty_cpu_schedulers,
+			 no_schedulers_online,
+                         erts_no_poll_threads,
+			 no_dirty_cpu_schedulers,
 			 no_dirty_cpu_schedulers_online,
 			 no_dirty_io_schedulers
-#endif
 			 );
     erts_late_init_time_sup();
     erts_init_cpu_topology(); /* Must be after init_scheduling */
@@ -410,7 +397,7 @@ erl_first_process_otp(char* modname, void* code, unsigned size, int argc, char**
      */
 
     erts_init_empty_process(&parent);
-    erts_smp_proc_lock(&parent, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(&parent, ERTS_PROC_LOCK_MAIN);
     hp = HAlloc(&parent, argc*2 + 4);
     args = NIL;
     for (i = argc-1; i >= 0; i--) {
@@ -425,13 +412,13 @@ erl_first_process_otp(char* modname, void* code, unsigned size, int argc, char**
 
     so.flags = erts_default_spo_flags|SPO_SYSTEM_PROC;
     res = erl_create_process(&parent, start_mod, am_start, args, &so);
-    erts_smp_proc_unlock(&parent, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(&parent, ERTS_PROC_LOCK_MAIN);
     erts_cleanup_empty_process(&parent);
     return res;
 }
 
 static Eterm
-erl_system_process_otp(Eterm parent_pid, char* modname, int off_heap_msgq)
+erl_system_process_otp(Eterm parent_pid, char* modname, int off_heap_msgq, int prio)
 {
     Eterm start_mod;
     Process* parent;
@@ -446,11 +433,18 @@ erl_system_process_otp(Eterm parent_pid, char* modname, int off_heap_msgq)
 
     parent = erts_pid2proc(NULL, 0, parent_pid, ERTS_PROC_LOCK_MAIN);
 
-    so.flags = erts_default_spo_flags|SPO_SYSTEM_PROC;
+    so.flags = erts_default_spo_flags|SPO_SYSTEM_PROC|SPO_USE_ARGS;
     if (off_heap_msgq)
         so.flags |= SPO_OFF_HEAP_MSGQ;
+    so.min_heap_size    = H_MIN_SIZE;
+    so.min_vheap_size   = BIN_VH_MIN_SIZE;
+    so.max_heap_size    = H_MAX_SIZE;
+    so.max_heap_flags   = H_MAX_FLAGS;
+    so.priority         = prio;
+    so.max_gen_gcs      = (Uint16) erts_atomic32_read_nob(&erts_max_gen_gcs);
+    so.scheduler        = 0;
     res = erl_create_process(parent, start_mod, am_start, NIL, &so);
-    erts_smp_proc_unlock(parent, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(parent, ERTS_PROC_LOCK_MAIN);
     return res;
 }
 
@@ -577,9 +571,19 @@ void erts_usage(void)
     erts_fprintf(stderr, "-hmqd  val     set default message queue data flag for processes,\n");
     erts_fprintf(stderr, "               valid values are: off_heap | on_heap\n");
 
+    erts_fprintf(stderr, "-IOp number    set number of pollsets to be used to poll for I/O,\n");
+    erts_fprintf(stderr, "               This value has to be equal or smaller than the\n");
+    erts_fprintf(stderr, "               number of poll threads. If the current platform\n");
+    erts_fprintf(stderr, "               does not support concurrent update of pollsets\n");
+    erts_fprintf(stderr, "               this value is ignored.\n");
+    erts_fprintf(stderr, "-IOt number    set number of threads to be used to poll for I/O\n");
+    erts_fprintf(stderr, "-IOPp number   set number of pollsets as a percentage of the\n");
+    erts_fprintf(stderr, "               number of poll threads.");
+    erts_fprintf(stderr, "-IOPt number   set number of threads to be used to poll for I/O\n");
+    erts_fprintf(stderr, "               as a percentage of the number of schedulers.");
+
     /*    erts_fprintf(stderr, "-i module  set the boot module (default init)\n"); */
 
-    erts_fprintf(stderr, "-K boolean     enable or disable kernel poll\n");
     erts_fprintf(stderr, "-n[s|a|d]      Control behavior of signals to ports\n");
     erts_fprintf(stderr, "               Note that this flag is deprecated!\n");
     erts_fprintf(stderr, "-M<X> <Y>      memory allocator switches,\n");
@@ -601,6 +605,10 @@ void erts_usage(void)
     erts_fprintf(stderr, "-stbt type     u|ns|ts|ps|s|nnts|nnps|tnnps|db\n");
     erts_fprintf(stderr, "-sbwt val      set scheduler busy wait threshold, valid values are:\n");
     erts_fprintf(stderr, "               none|very_short|short|medium|long|very_long.\n");
+    erts_fprintf(stderr, "-sbwtdcpu val  set dirty CPU scheduler busy wait threshold, valid values are:\n");
+    erts_fprintf(stderr, "               none|very_short|short|medium|long|very_long.\n");
+    erts_fprintf(stderr, "-sbwtdio val   set dirty IO scheduler busy wait threshold, valid values are:\n");
+    erts_fprintf(stderr, "               none|very_short|short|medium|long|very_long.\n");
     erts_fprintf(stderr, "-scl bool      enable/disable compaction of scheduler load,\n");
     erts_fprintf(stderr, "               see the erl(1) documentation for more info.\n");
     erts_fprintf(stderr, "-sct cput      set cpu topology,\n");
@@ -619,12 +627,15 @@ void erts_usage(void)
     erts_fprintf(stderr, "               very_lazy|lazy|medium|eager|very_eager.\n");
     erts_fprintf(stderr, "-swt val       set scheduler wakeup threshold, valid values are:\n");
     erts_fprintf(stderr, "               very_low|low|medium|high|very_high.\n");
+    erts_fprintf(stderr, "-swtdcpu val   set dirty CPU scheduler wakeup threshold, valid values are:\n");
+    erts_fprintf(stderr, "               very_low|low|medium|high|very_high.\n");
+    erts_fprintf(stderr, "-swtdio val    set dirty IO scheduler wakeup threshold, valid values are:\n");
+    erts_fprintf(stderr, "               very_low|low|medium|high|very_high.\n");
     erts_fprintf(stderr, "-sss size      suggested stack size in kilo words for scheduler threads,\n");
     erts_fprintf(stderr, "               valid range is [%d-%d] (default %d)\n",
 		 ERTS_SCHED_THREAD_MIN_STACK_SIZE,
 		 ERTS_SCHED_THREAD_MAX_STACK_SIZE,
                  ERTS_DEFAULT_SCHED_STACK_SIZE);
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_fprintf(stderr, "-sssdcpu size  suggested stack size in kilo words for dirty CPU scheduler\n");
     erts_fprintf(stderr, "               threads, valid range is [%d-%d] (default %d)\n",
 		 ERTS_SCHED_THREAD_MIN_STACK_SIZE,
@@ -635,7 +646,6 @@ void erts_usage(void)
 		 ERTS_SCHED_THREAD_MIN_STACK_SIZE,
 		 ERTS_SCHED_THREAD_MAX_STACK_SIZE,
                  ERTS_DEFAULT_DIO_SCHED_STACK_SIZE);
-#endif
     erts_fprintf(stderr, "-spp Bool      set port parallelism scheduling hint\n");
     erts_fprintf(stderr, "-S n1:n2       set number of schedulers (n1), and number of\n");
     erts_fprintf(stderr, "               schedulers online (n2), maximum for both\n");
@@ -644,7 +654,6 @@ void erts_usage(void)
     erts_fprintf(stderr, "-SP p1:p2      specify schedulers (p1) and schedulers online (p2)\n");
     erts_fprintf(stderr, "               as percentages of logical processors configured and logical\n");
     erts_fprintf(stderr, "               processors available, respectively\n");
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_fprintf(stderr, "-SDcpu n1:n2   set number of dirty CPU schedulers (n1), and number of\n");
     erts_fprintf(stderr, "               dirty CPU schedulers online (n2), valid range for both\n");
     erts_fprintf(stderr, "               numbers is [1-%d], and n2 must be less than or equal to n1\n",
@@ -654,7 +663,6 @@ void erts_usage(void)
     erts_fprintf(stderr, "               and logical processors available, respectively\n");
     erts_fprintf(stderr, "-SDio n        set number of dirty I/O schedulers, valid range is [0-%d]\n",
 		 ERTS_MAX_NO_OF_DIRTY_IO_SCHEDULERS);
-#endif
     erts_fprintf(stderr, "-t size        set the maximum number of atoms the emulator can handle\n");
     erts_fprintf(stderr, "               valid range is [%d-%d]\n",
 		 MIN_ATOM_TABLE_SIZE, MAX_ATOM_TABLE_SIZE);
@@ -682,7 +690,6 @@ void erts_usage(void)
     erts_exit(1, "");
 }
 
-#ifdef USE_THREADS
 /*
  * allocators for thread lib
  */
@@ -724,7 +731,6 @@ static void ethr_ll_free(void *ptr)
     erts_free(ERTS_ALC_T_ETHR_LL, ptr);
 }
 
-#endif
 
 static int
 early_init(int *argc, char **argv) /*
@@ -742,22 +748,16 @@ early_init(int *argc, char **argv) /*
     int schdlrs_percentage = 100;
     int schdlrs_onln_percentage = 100;
     int max_main_threads;
-#ifdef ERTS_DIRTY_SCHEDULERS
     int dirty_cpu_scheds;
     int dirty_cpu_scheds_online;
     int dirty_cpu_scheds_pctg = 100;
     int dirty_cpu_scheds_onln_pctg = 100;
     int dirty_io_scheds;
-#endif
     int max_reader_groups;
     int reader_groups;
     char envbuf[21]; /* enough for any 64-bit integer */
     size_t envbufsz;
 
-#if defined(USE_THREADS) && !defined(ERTS_SMP)
-    erts_main_thread = erts_thr_self();
-#endif
-
     erts_save_emu_args(*argc, argv);
 
     erts_sched_compact_load = 1;
@@ -775,17 +775,10 @@ early_init(int *argc, char **argv) /*
 
     erts_initialized = 0;
 
-    erts_use_sender_punish = 1;
-
     erts_pre_early_init_cpu_topology(&max_reader_groups,
 				     &ncpu,
 				     &ncpuonln,
 				     &ncpuavail);
-#ifndef ERTS_SMP
-    ncpu = 1;
-    ncpuonln = 1;
-    ncpuavail = 1;
-#endif
 
     ignore_break = 0;
     replace_intr = 0;
@@ -797,18 +790,12 @@ early_init(int *argc, char **argv) /*
 
     erts_sys_pre_init();
     erts_atomic_init_nob(&exiting, 0);
-#ifdef ERTS_SMP
     erts_thr_progress_pre_init();
-#endif
 
-#ifdef ERTS_SMP
-    erts_smp_atomic32_init_nob(&erts_writing_erl_crash_dump, 0L);
+    erts_atomic32_init_nob(&erts_writing_erl_crash_dump, 0L);
     erts_tsd_key_create(&erts_is_crash_dumping_key,"erts_is_crash_dumping_key");
-#else
-    erts_writing_erl_crash_dump = 0;
-#endif
 
-    erts_smp_atomic32_init_nob(&erts_max_gen_gcs,
+    erts_atomic32_init_nob(&erts_max_gen_gcs,
 			       (erts_aint32_t) ((Uint16) -1));
 
     erts_pre_init_process();
@@ -825,16 +812,15 @@ early_init(int *argc, char **argv) /*
     schdlrs = no_schedulers;
     schdlrs_onln = no_schedulers_online;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     dirty_cpu_scheds = no_schedulers;
     dirty_cpu_scheds_online = no_schedulers_online;
     dirty_io_scheds = 10;
-#endif
 
     envbufsz = sizeof(envbuf);
 
-    /* erts_sys_getenv(_raw)() not initialized yet; need erts_sys_getenv__() */
-    if (erts_sys_getenv__("ERL_THREAD_POOL_SIZE", envbuf, &envbufsz) == 0)
+    /* erts_osenv hasn't been initialized yet, so we need to fall back to
+     * erts_sys_explicit_host_getenv() */
+    if (erts_sys_explicit_host_getenv("ERL_THREAD_POOL_SIZE", envbuf, &envbufsz) == 1)
 	erts_async_max_threads = atoi(envbuf);
     else
 	erts_async_max_threads = ERTS_DEFAULT_NO_ASYNC_THREADS;
@@ -844,7 +830,7 @@ early_init(int *argc, char **argv) /*
     if (argc && argv) {
 	int i = 1;
 	while (i < *argc) {
-	    if (strcmp(argv[i], "--") == 0) { /* end of emulator options */
+	    if (sys_strcmp(argv[i], "--") == 0) { /* end of emulator options */
 		i++;
 		break;
 	    }
@@ -882,6 +868,7 @@ early_init(int *argc, char **argv) /*
 		    }
 		    break;
 		}
+
 		case 'S' :
 		    if (argv[i][2] == 'P') {
 			int ptot, ponln;
@@ -922,11 +909,10 @@ early_init(int *argc, char **argv) /*
                                 ("using %d:%d scheduler percentages\n",
                                  schdlrs_percentage, schdlrs_onln_percentage));
                     }
-#ifdef ERTS_DIRTY_SCHEDULERS
 		    else if (argv[i][2] == 'D') {
 			char *arg;
 			char *type = argv[i]+3;
-			if (strncmp(type, "Pcpu", 4) == 0) {
+			if (sys_strncmp(type, "Pcpu", 4) == 0) {
 			    int ptot, ponln;
 			    arg = get_arg(argv[i]+7, argv[i+1], &i);
 			    switch (sscanf(arg, "%d:%d", &ptot, &ponln)) {
@@ -963,7 +949,7 @@ early_init(int *argc, char **argv) /*
 			    VERBOSE(DEBUG_SYSTEM,
 				    ("using %d:%d dirty CPU scheduler percentages\n",
 				     dirty_cpu_scheds_pctg, dirty_cpu_scheds_onln_pctg));
-			} else if (strncmp(type, "cpu", 3) == 0) {
+			} else if (sys_strncmp(type, "cpu", 3) == 0) {
 			    int tot, onln;
 			    arg = get_arg(argv[i]+6, argv[i+1], &i);
 			    switch (sscanf(arg, "%d:%d", &tot, &onln)) {
@@ -1014,7 +1000,7 @@ early_init(int *argc, char **argv) /*
 			    }
 			    VERBOSE(DEBUG_SYSTEM,
 				    ("using %d:%d dirty CPU scheduler(s)\n", tot, onln));
-			} else if (strncmp(type, "io", 2) == 0) {
+			} else if (sys_strncmp(type, "io", 2) == 0) {
 			    arg = get_arg(argv[i]+5, argv[i+1], &i);
 			    dirty_io_scheds = atoi(arg);
 			    if (dirty_io_scheds < 0 ||
@@ -1034,7 +1020,6 @@ early_init(int *argc, char **argv) /*
 			    break;
 			}
 		    }
-#endif
 		    else {
 			int tot, onln;
 			char *arg = get_arg(argv[i]+2, argv[i+1], &i);
@@ -1093,7 +1078,6 @@ early_init(int *argc, char **argv) /*
 	    i++;
 	}
 
-#ifdef ERTS_SMP
 	/* apply any scheduler percentages */
 	if (schdlrs_percentage != 100 || schdlrs_onln_percentage != 100) {
 	    schdlrs = schdlrs * schdlrs_percentage / 100;
@@ -1117,12 +1101,6 @@ early_init(int *argc, char **argv) /*
 		erts_usage();
 	    }
 	}
-#else
-	/* Silence gcc warnings */
-	(void)schdlrs_percentage;
-	(void)schdlrs_onln_percentage;
-#endif
-#ifdef ERTS_DIRTY_SCHEDULERS
 	/* apply any dirty scheduler precentages */
 	if (dirty_cpu_scheds_pctg != 100 || dirty_cpu_scheds_onln_pctg != 100) {
 	    dirty_cpu_scheds = dirty_cpu_scheds * dirty_cpu_scheds_pctg / 100;
@@ -1136,33 +1114,25 @@ early_init(int *argc, char **argv) /*
 	    dirty_cpu_scheds_online = schdlrs_onln;
 	if (dirty_cpu_scheds_online < 1)
 	    dirty_cpu_scheds_online = 1;
-#endif
     }
 
-#ifndef USE_THREADS
-    erts_async_max_threads = 0;
-#endif
 
-#ifdef ERTS_SMP
     no_schedulers = schdlrs;
     no_schedulers_online = schdlrs_onln;
 
     erts_no_schedulers = (Uint) no_schedulers;
-#else
-    erts_no_schedulers = 1;
-#endif
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_no_dirty_cpu_schedulers = no_dirty_cpu_schedulers = dirty_cpu_scheds;
     no_dirty_cpu_schedulers_online = dirty_cpu_scheds_online;
     erts_no_dirty_io_schedulers = no_dirty_io_schedulers = dirty_io_scheds;
-#endif
     erts_early_init_scheduling(no_schedulers);
 
     alloc_opts.ncpu = ncpu;
     erts_alloc_init(argc, argv, &alloc_opts); /* Handles (and removes)
 						 -M flags. */
     /* Require allocators */
-#ifdef ERTS_SMP
+
+    erts_init_check_io(argc, argv);
+
     /*
      * Thread progress management:
      *
@@ -1170,22 +1140,18 @@ early_init(int *argc, char **argv) /*
      * ** Scheduler threads (see erl_process.c)
      * ** Aux thread (see erl_process.c)
      * ** Sys message dispatcher thread (see erl_trace.c)
+     * ** IO Poll threads (see erl_check_io.c)
      *
      * * Unmanaged threads that need to register:
      * ** Async threads (see erl_async.c)
      * ** Dirty scheduler threads
      */
     erts_thr_progress_init(no_schedulers,
-			   no_schedulers+2,
-#ifndef ERTS_DIRTY_SCHEDULERS
-			   erts_async_max_threads
-#else
+			   no_schedulers+2+erts_no_poll_threads,
 			   erts_async_max_threads +
 			   erts_no_dirty_cpu_schedulers +
 			   erts_no_dirty_io_schedulers
-#endif
 			   );
-#endif
     erts_thr_q_init();
     erts_init_utils();
     erts_early_init_cpu_topology(no_schedulers,
@@ -1193,7 +1159,6 @@ early_init(int *argc, char **argv) /*
 				 max_reader_groups,
 				 &reader_groups);
 
-#ifdef USE_THREADS
     {
 	erts_thr_late_init_data_t elid = ERTS_THR_LATE_INIT_DATA_DEF_INITER;
 	elid.mem.std.alloc = ethr_std_alloc;
@@ -1210,7 +1175,6 @@ early_init(int *argc, char **argv) /*
 
 	erts_thr_late_init(&elid);
     }
-#endif
     erts_msacc_early_init();
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
@@ -1237,40 +1201,6 @@ early_init(int *argc, char **argv) /*
     return ncpu;
 }
 
-#ifndef ERTS_SMP
-
-void *erts_scheduler_stack_limit;
-
-
-static void set_main_stack_size(void)
-{
-    char c;
-    UWord stacksize;
-# if HAVE_DECL_GETRLIMIT && HAVE_DECL_SETRLIMIT && HAVE_DECL_RLIMIT_STACK
-    struct rlimit rl;
-    int bytes;
-    stacksize = erts_sched_thread_suggested_stack_size * sizeof(Uint) * 1024;
-    /* Add some extra pages... neede by some systems... */
-    bytes = (int) stacksize + 3*erts_sys_get_page_size();
-    if (getrlimit(RLIMIT_STACK, &rl) != 0 ||
-        (rl.rlim_cur = bytes, setrlimit(RLIMIT_STACK, &rl) != 0)) {
-        erts_fprintf(stderr, "failed to set stack size for scheduler "
-                     "thread to %d bytes\n", bytes);
-        erts_usage();
-    }
-# else
-    if (modified_sched_thread_suggested_stack_size) {
-	erts_fprintf(stderr, "no OS support for dynamic stack size limit\n");
-	erts_usage();
-    }
-    /* Be conservative and hope it is not more than 64 kWords... */
-    stacksize = 64*1024*sizeof(void *);
-# endif
-
-    erts_scheduler_stack_limit = erts_calc_stacklimit(&c, stacksize);
-}
-
-#endif
 
 void
 erl_start(int argc, char **argv)
@@ -1290,26 +1220,25 @@ erl_start(int argc, char **argv)
     ErtsTimeWarpMode time_warp_mode;
     int node_tab_delete_delay = ERTS_NODE_TAB_DELAY_GC_DEFAULT;
     ErtsDbSpinCount db_spin_count = ERTS_DB_SPNCNT_NORMAL;
-    Eterm otp_ring0_pid;
 
     set_default_time_adj(&time_correction,
 			 &time_warp_mode);
 
     envbufsz = sizeof(envbuf);
-    if (erts_sys_getenv_raw(ERL_MAX_ETS_TABLES_ENV, envbuf, &envbufsz) == 0)
+    if (erts_sys_explicit_8bit_getenv(ERL_MAX_ETS_TABLES_ENV, envbuf, &envbufsz) == 1)
 	user_requested_db_max_tabs = atoi(envbuf);
     else
 	user_requested_db_max_tabs = 0;
 
     envbufsz = sizeof(envbuf);
-    if (erts_sys_getenv_raw("ERL_FULLSWEEP_AFTER", envbuf, &envbufsz) == 0) {
+    if (erts_sys_explicit_8bit_getenv("ERL_FULLSWEEP_AFTER", envbuf, &envbufsz) == 1) {
 	Uint16 max_gen_gcs = atoi(envbuf);
-	erts_smp_atomic32_set_nob(&erts_max_gen_gcs,
+	erts_atomic32_set_nob(&erts_max_gen_gcs,
 				  (erts_aint32_t) max_gen_gcs);
     }
 
     envbufsz = sizeof(envbuf);
-    if (erts_sys_getenv_raw("ERL_MAX_PORTS", envbuf, &envbufsz) == 0) {
+    if (erts_sys_explicit_8bit_getenv("ERL_MAX_PORTS", envbuf, &envbufsz) == 1) {
 	port_tab_sz = atoi(envbuf);
 	port_tab_sz_ignore_files = 1;
     }
@@ -1319,10 +1248,8 @@ erl_start(int argc, char **argv)
      * a lot of stack.
      */
     erts_sched_thread_suggested_stack_size = ERTS_DEFAULT_SCHED_STACK_SIZE;
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_dcpu_sched_thread_suggested_stack_size = ERTS_DEFAULT_DCPU_SCHED_STACK_SIZE;
     erts_dio_sched_thread_suggested_stack_size = ERTS_DEFAULT_DIO_SCHED_STACK_SIZE;
-#endif
 
 #ifdef DEBUG
     verbose = DEBUG_DEFAULT;
@@ -1334,7 +1261,7 @@ erl_start(int argc, char **argv)
 	if (argv[i][0] != '-') {
 	    erts_usage();
 	}
-	if (strcmp(argv[i], "--") == 0) { /* end of emulator options */
+	if (sys_strcmp(argv[i], "--") == 0) { /* end of emulator options */
 	    i++;
 	    break;
 	}
@@ -1362,12 +1289,12 @@ erl_start(int argc, char **argv)
                     ("using display items %d\n",display_items));
 	    break;
 	case 'p':
-	    if (!strncmp(argv[i],"-pc",3)) {
+	    if (!sys_strncmp(argv[i],"-pc",3)) {
 		int printable_chars = ERL_PRINTABLE_CHARACTERS_LATIN1;
 		arg = get_arg(argv[i]+3, argv[i+1], &i);
-		if (!strcmp(arg,"unicode")) {
+		if (!sys_strcmp(arg,"unicode")) {
 		    printable_chars = ERL_PRINTABLE_CHARACTERS_UNICODE;
-		} else if (strcmp(arg,"latin1")) {
+		} else if (sys_strcmp(arg,"latin1")) {
 		    erts_fprintf(stderr, "bad range of printable "
 				 "characters: %s\n", arg);
 		    erts_usage();
@@ -1379,7 +1306,7 @@ erl_start(int argc, char **argv)
 		erts_usage();
 	    }
 	case 'f':
-	    if (!strncmp(argv[i],"-fn",3)) {
+	    if (!sys_strncmp(argv[i],"-fn",3)) {
 		int warning_type =  ERL_FILENAME_WARNING_WARNING;
 		arg = get_arg(argv[i]+3, argv[i+1], &i);
 		switch (*arg) {
@@ -1490,12 +1417,8 @@ erl_start(int argc, char **argv)
 #ifdef DEBUG
 		strcat(tmp, ",DEBUG");
 #endif
-#ifdef ERTS_SMP
 		strcat(tmp, ",SMP");
-#endif
-#ifdef USE_THREADS
 		strcat(tmp, ",ASYNC_THREADS");
-#endif
 #ifdef HIPE
 		strcat(tmp, ",HIPE");
 #endif
@@ -1566,9 +1489,9 @@ erl_start(int argc, char **argv)
 		}
             } else if (has_prefix("maxk", sub_param)) {
 		arg = get_arg(sub_param+4, argv[i+1], &i);
-		if (strcmp(arg,"true") == 0) {
+		if (sys_strcmp(arg,"true") == 0) {
                     H_MAX_FLAGS |= MAX_HEAP_SIZE_KILL;
-                } else if (strcmp(arg,"false") == 0) {
+                } else if (sys_strcmp(arg,"false") == 0) {
                     H_MAX_FLAGS &= ~MAX_HEAP_SIZE_KILL;
                 } else {
 		    erts_fprintf(stderr, "bad max heap kill %s\n", arg);
@@ -1577,9 +1500,9 @@ erl_start(int argc, char **argv)
 		VERBOSE(DEBUG_SYSTEM, ("using max heap kill %d\n", H_MAX_FLAGS));
             } else if (has_prefix("maxel", sub_param)) {
 		arg = get_arg(sub_param+5, argv[i+1], &i);
-		if (strcmp(arg,"true") == 0) {
+		if (sys_strcmp(arg,"true") == 0) {
                     H_MAX_FLAGS |= MAX_HEAP_SIZE_LOG;
-                } else if (strcmp(arg,"false") == 0) {
+                } else if (sys_strcmp(arg,"false") == 0) {
                     H_MAX_FLAGS &= ~MAX_HEAP_SIZE_LOG;
                 } else {
 		    erts_fprintf(stderr, "bad max heap error logger %s\n", arg);
@@ -1671,16 +1594,6 @@ erl_start(int argc, char **argv)
 	    have_break_handler = 0;
 	  break;
 
-	case 'K':
-	    /* If kernel poll support is present,
-	       erl_sys_args() will remove the K parameter
-	       and value */
-	    get_arg(argv[i]+2, argv[i+1], &i);
-	    erts_fprintf(stderr,
-		       "kernel-poll not supported; \"K\" parameter ignored\n",
-		       arg);
-	    break;
-
 	case 'n':
 	    arg = get_arg(argv[i]+2, argv[i+1], &i);
 	    switch (arg[0]) {
@@ -1707,7 +1620,7 @@ erl_start(int argc, char **argv)
 
 	case 'P': /* set maximum number of processes */
 	    arg = get_arg(argv[i]+2, argv[i+1], &i);
-	    if (strcmp(arg, "legacy") == 0)
+	    if (sys_strcmp(arg, "legacy") == 0)
 		legacy_proc_tab = 1;
 	    else {
 		errno = 0;
@@ -1723,7 +1636,7 @@ erl_start(int argc, char **argv)
 
 	case 'Q': /* set maximum number of ports */
 	    arg = get_arg(argv[i]+2, argv[i+1], &i);
-	    if (strcmp(arg, "legacy") == 0)
+	    if (sys_strcmp(arg, "legacy") == 0)
 		legacy_port_tab = 1;
 	    else {
 		errno = 0;
@@ -1741,11 +1654,11 @@ erl_start(int argc, char **argv)
 	case 'S' : /* Was handled in early_init() just read past it */
 	    if (argv[i][2] == 'D') {
 		char* type = argv[i]+3;
-		if (strcmp(type, "Pcpu") == 0)
+		if (sys_strcmp(type, "Pcpu") == 0)
 		    (void) get_arg(argv[i]+7, argv[i+1], &i);
-		if (strcmp(type, "cpu") == 0)
+		if (sys_strcmp(type, "cpu") == 0)
 		    (void) get_arg(argv[i]+6, argv[i+1], &i);
-		else if (strcmp(type, "io") == 0)
+		else if (sys_strcmp(type, "io") == 0)
 		    (void) get_arg(argv[i]+5, argv[i+1], &i);
 	    } else if (argv[i][2] == 'P')
 		(void) get_arg(argv[i]+3, argv[i+1], &i);
@@ -1782,15 +1695,41 @@ erl_start(int argc, char **argv)
 		    erts_usage();
 		}
 	    }
+	    else if (has_prefix("bwtdcpu", sub_param)) {
+                arg = get_arg(sub_param + 7, argv[i+1], &i);
+
+		if (erts_sched_set_busy_wait_threshold(ERTS_SCHED_DIRTY_CPU, arg) != 0) {
+		    erts_fprintf(stderr, "bad dirty CPU scheduler busy wait threshold: %s\n",
+				 arg);
+		    erts_usage();
+		}
+
+		VERBOSE(DEBUG_SYSTEM,
+			("dirty CPU scheduler wakeup threshold: %s\n", arg));
+	    }
+	    else if (has_prefix("bwtdio", sub_param)) {
+                arg = get_arg(sub_param + 6, argv[i+1], &i);
+
+		if (erts_sched_set_busy_wait_threshold(ERTS_SCHED_DIRTY_IO, arg) != 0) {
+		    erts_fprintf(stderr, "bad dirty IO scheduler busy wait threshold: %s\n",
+				 arg);
+		    erts_usage();
+		}
+
+		VERBOSE(DEBUG_SYSTEM,
+			("dirty IO scheduler wakeup threshold: %s\n", arg));
+	    }
 	    else if (has_prefix("bwt", sub_param)) {
-		arg = get_arg(sub_param+3, argv[i+1], &i);
-		if (erts_sched_set_busy_wait_threshold(arg) != 0) {
+                arg = get_arg(sub_param + 3, argv[i+1], &i);
+
+		if (erts_sched_set_busy_wait_threshold(ERTS_SCHED_NORMAL, arg) != 0) {
 		    erts_fprintf(stderr, "bad scheduler busy wait threshold: %s\n",
 				 arg);
 		    erts_usage();
 		}
+
 		VERBOSE(DEBUG_SYSTEM,
-			("scheduler wakup threshold: %s\n", arg));
+			("scheduler wakeup threshold: %s\n", arg));
 	    }
 	    else if (has_prefix("cl", sub_param)) {
 		arg = get_arg(sub_param+2, argv[i+1], &i);
@@ -1850,22 +1789,10 @@ erl_start(int argc, char **argv)
 		    erts_usage();
 		}
 	    }
-	    else if (has_prefix("ecio", sub_param)) {
-		arg = get_arg(sub_param+4, argv[i+1], &i);
-#ifndef __OSE__
-		if (sys_strcmp("true", arg) == 0)
-		    erts_eager_check_io = 1;
-		else
-#endif
-		if (sys_strcmp("false", arg) == 0)
-		    erts_eager_check_io = 0;
-		else {
-		    erts_fprintf(stderr,
-				 "bad schedule eager check I/O value '%s'\n",
-				 arg);
-		    erts_usage();
-		}
-	    }
+            else if (has_prefix("ecio", sub_param)) {
+                /* ignore argument, eager check io no longer used */
+                arg = get_arg(sub_param+4, argv[i+1], &i);
+            }
 	    else if (has_prefix("pp", sub_param)) {
 		arg = get_arg(sub_param+2, argv[i+1], &i);
 		if (sys_strcmp(arg, "true") == 0)
@@ -1879,8 +1806,6 @@ erl_start(int argc, char **argv)
 		    erts_usage();
 		}
 	    }
-	    else if (sys_strcmp("nsp", sub_param) == 0)
-		erts_use_sender_punish = 0;
 	    else if (has_prefix("tbt", sub_param)) {
 		arg = get_arg(sub_param+3, argv[i+1], &i);
 		res = erts_init_scheduler_bind_type_string(arg);
@@ -1921,9 +1846,29 @@ erl_start(int argc, char **argv)
 		VERBOSE(DEBUG_SYSTEM,
 			("scheduler wake cleanup threshold: %s\n", arg));
 	    }
+	    else if (has_prefix("wtdcpu", sub_param)) {
+		arg = get_arg(sub_param+6, argv[i+1], &i);
+		if (erts_sched_set_wakeup_other_threshold(ERTS_SCHED_DIRTY_CPU, arg) != 0) {
+		    erts_fprintf(stderr, "dirty CPU scheduler wakeup threshold: %s\n",
+				 arg);
+		    erts_usage();
+		}
+		VERBOSE(DEBUG_SYSTEM,
+			("dirty CPU scheduler wakeup threshold: %s\n", arg));
+	    }
+	    else if (has_prefix("wtdio", sub_param)) {
+		arg = get_arg(sub_param+5, argv[i+1], &i);
+		if (erts_sched_set_wakeup_other_threshold(ERTS_SCHED_DIRTY_IO, arg) != 0) {
+		    erts_fprintf(stderr, "dirty IO scheduler wakeup threshold: %s\n",
+				 arg);
+		    erts_usage();
+		}
+		VERBOSE(DEBUG_SYSTEM,
+			("dirty IO scheduler wakeup threshold: %s\n", arg));
+	    }
 	    else if (has_prefix("wt", sub_param)) {
 		arg = get_arg(sub_param+2, argv[i+1], &i);
-		if (erts_sched_set_wakeup_other_thresold(arg) != 0) {
+		if (erts_sched_set_wakeup_other_threshold(ERTS_SCHED_NORMAL, arg) != 0) {
 		    erts_fprintf(stderr, "scheduler wakeup threshold: %s\n",
 				 arg);
 		    erts_usage();
@@ -1933,7 +1878,7 @@ erl_start(int argc, char **argv)
 	    }
 	    else if (has_prefix("ws", sub_param)) {
 		arg = get_arg(sub_param+2, argv[i+1], &i);
-		if (erts_sched_set_wakeup_other_type(arg) != 0) {
+		if (erts_sched_set_wakeup_other_type(ERTS_SCHED_NORMAL, arg) != 0) {
 		    erts_fprintf(stderr, "scheduler wakeup strategy: %s\n",
 				 arg);
 		    erts_usage();
@@ -1941,7 +1886,6 @@ erl_start(int argc, char **argv)
 		VERBOSE(DEBUG_SYSTEM,
 			("scheduler wakeup threshold: %s\n", arg));
 	    }
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    else if (has_prefix("ssdcpu", sub_param)) {
 		/* suggested stack size (Kilo Words) for dirty CPU scheduler threads */
 		arg = get_arg(sub_param+6, argv[i+1], &i);
@@ -1976,7 +1920,6 @@ erl_start(int argc, char **argv)
 			("suggested dirty IO scheduler thread stack size %d kilo words\n",
 			 erts_dio_sched_thread_suggested_stack_size));
 	    }
-#endif
 	    else if (has_prefix("ss", sub_param)) {
 		/* suggested stack size (Kilo Words) for scheduler threads */
 		arg = get_arg(sub_param+2, argv[i+1], &i);
@@ -2007,9 +1950,7 @@ erl_start(int argc, char **argv)
 				 arg);
 		    erts_usage();
 		}
-#ifdef ERTS_SMP
 		erts_runq_supervision_interval = val;
-#endif
 	    }
 	    else {
 		erts_fprintf(stderr, "bad scheduling option %s\n", argv[i]);
@@ -2293,12 +2234,10 @@ erl_start(int argc, char **argv)
 
     if (erts_sched_thread_suggested_stack_size < ERTS_SCHED_THREAD_MIN_STACK_SIZE)
         erts_sched_thread_suggested_stack_size = ERTS_SCHED_THREAD_MIN_STACK_SIZE;
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (erts_dcpu_sched_thread_suggested_stack_size < ERTS_SCHED_THREAD_MIN_STACK_SIZE)
         erts_dcpu_sched_thread_suggested_stack_size = ERTS_SCHED_THREAD_MIN_STACK_SIZE;
     if (erts_dio_sched_thread_suggested_stack_size < ERTS_SCHED_THREAD_MIN_STACK_SIZE)
         erts_dio_sched_thread_suggested_stack_size = ERTS_SCHED_THREAD_MIN_STACK_SIZE;
-#endif
 
     erl_init(ncpu,
 	     proc_tab_sz,
@@ -2317,8 +2256,8 @@ erl_start(int argc, char **argv)
 
     erts_initialized = 1;
 
-    otp_ring0_pid = erl_first_process_otp("otp_ring0", NULL, 0,
-					  boot_argc, boot_argv);
+    erts_init_process_id = erl_first_process_otp("otp_ring0", NULL, 0,
+                                                 boot_argc, boot_argv);
 
     {
 	/*
@@ -2328,14 +2267,18 @@ erl_start(int argc, char **argv)
 	 */
 	Eterm pid;
 
-	pid = erl_system_process_otp(otp_ring0_pid, "erts_code_purger", !0);
+	pid = erl_system_process_otp(erts_init_process_id,
+                                     "erts_code_purger", !0,
+                                     PRIORITY_NORMAL);
 	erts_code_purger
 	    = (Process *) erts_ptab_pix2intptr_ddrb(&erts_proc,
 						    internal_pid_index(pid));
 	ASSERT(erts_code_purger && erts_code_purger->common.id == pid);
 	erts_proc_inc_refc(erts_code_purger); 
 
-	pid = erl_system_process_otp(otp_ring0_pid, "erts_literal_area_collector", !0);
+	pid = erl_system_process_otp(erts_init_process_id,
+                                     "erts_literal_area_collector",
+                                     !0, PRIORITY_NORMAL);
 	erts_literal_area_collector
 	    = (Process *) erts_ptab_pix2intptr_ddrb(&erts_proc,
 						    internal_pid_index(pid));
@@ -2343,19 +2286,37 @@ erl_start(int argc, char **argv)
 	       && erts_literal_area_collector->common.id == pid);
 	erts_proc_inc_refc(erts_literal_area_collector);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-	pid = erl_system_process_otp(otp_ring0_pid, "erts_dirty_process_code_checker", !0);
-	erts_dirty_process_code_checker
+	pid = erl_system_process_otp(erts_init_process_id,
+                                     "erts_dirty_process_signal_handler",
+                                     !0, PRIORITY_NORMAL);
+        erts_dirty_process_signal_handler
 	    = (Process *) erts_ptab_pix2intptr_ddrb(&erts_proc,
 						    internal_pid_index(pid));
-	ASSERT(erts_dirty_process_code_checker
-	       && erts_dirty_process_code_checker->common.id == pid);
-	erts_proc_inc_refc(erts_dirty_process_code_checker);
-#endif
-
+	ASSERT(erts_dirty_process_signal_handler
+	       && erts_dirty_process_signal_handler->common.id == pid);
+	erts_proc_inc_refc(erts_dirty_process_signal_handler);
+
+	pid = erl_system_process_otp(erts_init_process_id,
+                                     "erts_dirty_process_signal_handler",
+                                     !0, PRIORITY_HIGH);
+        erts_dirty_process_signal_handler_high
+	    = (Process *) erts_ptab_pix2intptr_ddrb(&erts_proc,
+						    internal_pid_index(pid));
+	ASSERT(erts_dirty_process_signal_handler_high
+	       && erts_dirty_process_signal_handler_high->common.id == pid);
+	erts_proc_inc_refc(erts_dirty_process_signal_handler_high);
+
+	pid = erl_system_process_otp(erts_init_process_id,
+                                     "erts_dirty_process_signal_handler",
+                                     !0, PRIORITY_MAX);
+        erts_dirty_process_signal_handler_max
+	    = (Process *) erts_ptab_pix2intptr_ddrb(&erts_proc,
+						    internal_pid_index(pid));
+	ASSERT(erts_dirty_process_signal_handler_max
+	       && erts_dirty_process_signal_handler_max->common.id == pid);
+	erts_proc_inc_refc(erts_dirty_process_signal_handler_max);
     }
 
-#ifdef ERTS_SMP
     erts_start_schedulers();
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
@@ -2364,31 +2325,9 @@ erl_start(int argc, char **argv)
 
     /* Let system specific code decide what to do with the main thread... */
     erts_sys_main_thread(); /* May or may not return! */
-#else
-    {
-	ErtsSchedulerData *esdp = erts_get_scheduler_data();
-        erts_msacc_init_thread("scheduler", 1, 1);
-	erts_thr_set_main_status(1, 1);
-#if ERTS_USE_ASYNC_READY_Q
-	esdp->aux_work_data.async_ready.queue
-	    = erts_get_async_ready_queue(1);
-#endif
-	set_main_stack_size();
-	erts_sched_init_time_sup(esdp);
-        erts_ets_sched_spec_data_init(esdp);
-        erts_aux_work_timeout_late_init(esdp);
-
-#ifdef ERTS_ENABLE_LOCK_COUNT
-        erts_lcnt_post_startup();
-#endif
-
-	process_main(esdp->x_reg_array, esdp->f_reg_array);
-    }
-#endif
 }
 
 
-#ifdef USE_THREADS
 
 __decl_noreturn void erts_thr_fatal_error(int err, char *what)
 {
@@ -2402,7 +2341,6 @@ __decl_noreturn void erts_thr_fatal_error(int err, char *what)
     abort();
 }
 
-#endif
 
 static void
 system_cleanup(int flush_async)
@@ -2415,7 +2353,6 @@ system_cleanup(int flush_async)
 	 * Another thread is currently exiting the system;
 	 * wait for it to do its job.
 	 */
-#ifdef ERTS_SMP
 	if (erts_thr_progress_is_managed_thread()) {
 	    /*
 	     * The exiting thread might be waiting for
@@ -2424,7 +2361,6 @@ system_cleanup(int flush_async)
 	    erts_thr_progress_active(NULL, 0);
 	    erts_thr_progress_prepare_wait(NULL);
 	}
-#endif
 	/* Wait forever... */
 	while (1)
 	    erts_milli_sleep(10000000);
@@ -2439,17 +2375,12 @@ system_cleanup(int flush_async)
 
     if (!flush_async
 	|| !erts_initialized
-#if defined(USE_THREADS) && !defined(ERTS_SMP)
-	|| !erts_equal_tids(erts_main_thread, erts_thr_self())
-#endif
 	)
 	return;
 
-#ifdef ERTS_SMP
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_check_exact(NULL, 0);
 #endif
-#endif
 
     erts_exit_flush_async();
 }
diff --git a/erts/emulator/beam/erl_instrument.c b/erts/emulator/beam/erl_instrument.c
deleted file mode 100644
index 634509f880..0000000000
--- a/erts/emulator/beam/erl_instrument.c
+++ /dev/null
@@ -1,1257 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 2003-2016. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-
-#include "global.h"
-#include "big.h"
-#include "erl_instrument.h"
-#include "erl_threads.h"
-
-typedef union { long l; double d; } Align_t;
-
-typedef struct {
-    Uint size;
-#ifdef VALGRIND
-    void* valgrind_leak_suppressor;
-#endif
-    Align_t mem[1];
-} StatBlock_t;
-
-#define STAT_BLOCK_HEADER_SIZE (sizeof(StatBlock_t) - sizeof(Align_t))
-
-typedef struct MapStatBlock_t_ MapStatBlock_t;
-struct MapStatBlock_t_ {
-    Uint size;
-    ErtsAlcType_t type_no;
-    Eterm pid;
-    MapStatBlock_t *prev;
-    MapStatBlock_t *next;
-    Align_t mem[1];
-};
-
-#define MAP_STAT_BLOCK_HEADER_SIZE (sizeof(MapStatBlock_t) - sizeof(Align_t))
-
-typedef struct {
-    Uint size;
-    Uint max_size;
-    Uint max_size_ever;
-
-    Uint blocks;
-    Uint max_blocks;
-    Uint max_blocks_ever;
-} Stat_t;
-
-static erts_mtx_t instr_mutex;
-static erts_mtx_t instr_x_mutex;
-
-int erts_instr_memory_map;
-int erts_instr_stat;
-
-static ErtsAllocatorFunctions_t real_allctrs[ERTS_ALC_A_MAX+1];
-
-struct stats_ {
-    Stat_t tot;
-    Stat_t a[ERTS_ALC_A_MAX+1];
-    Stat_t *ap[ERTS_ALC_A_MAX+1];
-    Stat_t c[ERTS_ALC_C_MAX+1];
-    Stat_t n[ERTS_ALC_N_MAX+1];
-};
-
-static struct stats_ *stats;
-
-static MapStatBlock_t *mem_anchor;
-
-static Eterm *am_tot;
-static Eterm *am_n;
-static Eterm *am_a;
-static Eterm *am_c;
-
-static int atoms_initialized;
-
-static struct {
-    Eterm total;
-    Eterm allocators;
-    Eterm classes;
-    Eterm types;
-    Eterm sizes;
-    Eterm blocks;
-    Eterm instr_hdr;
-#ifdef DEBUG
-    Eterm end_of_atoms;
-#endif
-} am;
-
-static void ERTS_INLINE atom_init(Eterm *atom, const char *name)
-{
-    *atom = am_atom_put((char *) name, strlen(name));
-}
-#define AM_INIT(AM) atom_init(&am.AM, #AM)
-
-static void
-init_atoms(void)
-{
-#ifdef DEBUG
-    Eterm *atom;
-    for (atom = (Eterm *) &am; atom <= &am.end_of_atoms; atom++) {
-	*atom = THE_NON_VALUE;
-    }
-#endif
-
-    AM_INIT(total);
-    AM_INIT(allocators);
-    AM_INIT(classes);
-    AM_INIT(types);
-    AM_INIT(sizes);
-    AM_INIT(blocks);
-    AM_INIT(instr_hdr);
-
-#ifdef DEBUG
-    for (atom = (Eterm *) &am; atom < &am.end_of_atoms; atom++) {
-	ASSERT(*atom != THE_NON_VALUE);
-    }
-#endif
-
-    atoms_initialized = 1;
-}
-
-#undef AM_INIT
-
-static void
-init_am_tot(void)
-{
-    am_tot = (Eterm *) erts_alloc(ERTS_ALC_T_INSTR_INFO,
-				  sizeof(Eterm));
-    atom_init(am_tot, "total");
-}
-
-
-static void
-init_am_n(void)
-{
-    int i;
-    am_n = (Eterm *) erts_alloc(ERTS_ALC_T_INSTR_INFO,
-				(ERTS_ALC_N_MAX+1)*sizeof(Eterm));
-
-    for (i = ERTS_ALC_N_MIN; i <= ERTS_ALC_N_MAX; i++) {
-	atom_init(&am_n[i], ERTS_ALC_N2TD(i));
-    }
-
-}
-
-static void
-init_am_c(void)
-{
-    int i;
-    am_c = (Eterm *) erts_alloc(ERTS_ALC_T_INSTR_INFO,
-				(ERTS_ALC_C_MAX+1)*sizeof(Eterm));
-
-    for (i = ERTS_ALC_C_MIN; i <= ERTS_ALC_C_MAX; i++) {
-	atom_init(&am_c[i], ERTS_ALC_C2CD(i));
-    }
-
-}
-
-static void
-init_am_a(void)
-{
-    int i;
-    am_a = (Eterm *) erts_alloc(ERTS_ALC_T_INSTR_INFO,
-				(ERTS_ALC_A_MAX+1)*sizeof(Eterm));
-
-    for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	atom_init(&am_a[i], ERTS_ALC_A2AD(i));
-    }
-
-}
-
-static ERTS_INLINE void
-stat_upd_alloc(ErtsAlcType_t n, Uint size)
-{
-    ErtsAlcType_t t = ERTS_ALC_N2T(n);
-    ErtsAlcType_t a = ERTS_ALC_T2A(t);
-    ErtsAlcType_t c = ERTS_ALC_T2C(t);
-
-    stats->ap[a]->size += size;
-    if (stats->ap[a]->max_size < stats->ap[a]->size)
-	stats->ap[a]->max_size = stats->ap[a]->size;
-
-    stats->c[c].size += size;
-    if (stats->c[c].max_size < stats->c[c].size)
-	stats->c[c].max_size = stats->c[c].size;
-
-    stats->n[n].size += size;
-    if (stats->n[n].max_size < stats->n[n].size)
-	stats->n[n].max_size = stats->n[n].size;
-
-    stats->tot.size += size;
-    if (stats->tot.max_size < stats->tot.size)
-	stats->tot.max_size = stats->tot.size;
-
-    stats->ap[a]->blocks++;
-    if (stats->ap[a]->max_blocks < stats->ap[a]->blocks)
-	stats->ap[a]->max_blocks = stats->ap[a]->blocks;
-
-    stats->c[c].blocks++;
-    if (stats->c[c].max_blocks < stats->c[c].blocks)
-	stats->c[c].max_blocks = stats->c[c].blocks;
-
-    stats->n[n].blocks++;
-    if (stats->n[n].max_blocks < stats->n[n].blocks)
-	stats->n[n].max_blocks = stats->n[n].blocks;
-
-    stats->tot.blocks++;
-    if (stats->tot.max_blocks < stats->tot.blocks)
-	stats->tot.max_blocks = stats->tot.blocks;
-
-}
-
-
-static ERTS_INLINE void
-stat_upd_free(ErtsAlcType_t n, Uint size)
-{
-    ErtsAlcType_t t = ERTS_ALC_N2T(n);
-    ErtsAlcType_t a = ERTS_ALC_T2A(t);
-    ErtsAlcType_t c = ERTS_ALC_T2C(t);
-
-    ASSERT(stats->ap[a]->size >= size);
-    stats->ap[a]->size -= size;
-
-    ASSERT(stats->c[c].size >= size);
-    stats->c[c].size -= size;
-
-    ASSERT(stats->n[n].size >= size);
-    stats->n[n].size -= size;
-
-    ASSERT(stats->tot.size >= size);
-    stats->tot.size -= size;
-
-    ASSERT(stats->ap[a]->blocks > 0);
-    stats->ap[a]->blocks--;
-
-    ASSERT(stats->c[c].blocks > 0);
-    stats->c[c].blocks--;
-
-    ASSERT(stats->n[n].blocks > 0);
-    stats->n[n].blocks--;
-
-    ASSERT(stats->tot.blocks > 0);
-    stats->tot.blocks--;
-
-}
-
-
-static ERTS_INLINE void
-stat_upd_realloc(ErtsAlcType_t n, Uint size, Uint old_size)
-{
-    if (old_size)
-	stat_upd_free(n, old_size);
-    stat_upd_alloc(n, size);
-}
-
-/*
- * stat instrumentation callback functions
- */
-
-static void stat_pre_lock(void)
-{
-    erts_mtx_lock(&instr_mutex);
-}
-
-static void stat_pre_unlock(void)
-{
-    erts_mtx_unlock(&instr_mutex);
-}
-
-static ErtsAllocatorWrapper_t instr_wrapper;
-
-static void *
-stat_alloc(ErtsAlcType_t n, void *extra, Uint size)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    Uint ssize;
-    void *res;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    ssize = size + STAT_BLOCK_HEADER_SIZE;
-    res = (*real_af->alloc)(n, real_af->extra, ssize);
-    if (res) {
-	stat_upd_alloc(n, size);
-	((StatBlock_t *) res)->size = size;
-#ifdef VALGRIND
-	/* Suppress "possibly leaks" by storing an actual dummy pointer
-	   to the _start_ of the allocated block.*/
-	((StatBlock_t *) res)->valgrind_leak_suppressor = res;
-#endif
-	res = (void *) ((StatBlock_t *) res)->mem;
-    }
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-    }
-
-    return res;
-}
-
-static void *
-stat_realloc(ErtsAlcType_t n, void *extra, void *ptr, Uint size)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    Uint old_size;
-    Uint ssize;
-    void *sptr;
-    void *res;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    if (ptr) {
-	sptr = (void *) (((char *) ptr) - STAT_BLOCK_HEADER_SIZE);
-	old_size = ((StatBlock_t *) sptr)->size;
-    }
-    else {
-	sptr = NULL;
-	old_size = 0;
-    }
-
-    ssize = size + STAT_BLOCK_HEADER_SIZE;
-    res = (*real_af->realloc)(n, real_af->extra, sptr, ssize);
-    if (res) {
-	stat_upd_realloc(n, size, old_size);
-	((StatBlock_t *) res)->size = size;
-#ifdef VALGRIND
-	((StatBlock_t *) res)->valgrind_leak_suppressor = res;
-#endif
-	res = (void *) ((StatBlock_t *) res)->mem;
-    }
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-    }
-
-    return res;
-}
-
-static void
-stat_free(ErtsAlcType_t n, void *extra, void *ptr)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    void *sptr;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    if (ptr) {
-	sptr = (void *) (((char *) ptr) - STAT_BLOCK_HEADER_SIZE);
-	stat_upd_free(n, ((StatBlock_t *) sptr)->size);
-    }
-    else {
-	sptr = NULL;
-    }
-
-    (*real_af->free)(n, real_af->extra, sptr);
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-    }
-
-}
-
-/*
- * map stat instrumentation callback functions
- */
-
-static void map_stat_pre_lock(void)
-{
-    erts_mtx_lock(&instr_x_mutex);
-    erts_mtx_lock(&instr_mutex);
-}
-
-static void map_stat_pre_unlock(void)
-{
-    erts_mtx_unlock(&instr_mutex);
-    erts_mtx_unlock(&instr_x_mutex);
-}
-
-static void *
-map_stat_alloc(ErtsAlcType_t n, void *extra, Uint size)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    Uint msize;
-    void *res;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    msize = size + MAP_STAT_BLOCK_HEADER_SIZE;
-    res = (*real_af->alloc)(n, real_af->extra, msize);
-    if (res) {
-	MapStatBlock_t *mb = (MapStatBlock_t *) res;
-	stat_upd_alloc(n, size);
-
-	mb->size = size;
-	mb->type_no = n;
-	mb->pid = erts_get_current_pid();
-
-	mb->prev = NULL;
-	mb->next = mem_anchor;
-	if (mem_anchor)
-	    mem_anchor->prev = mb;
-	mem_anchor = mb;
-
-	res = (void *) mb->mem;
-    }
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-    }
-
-    return res;
-}
-
-static void *
-map_stat_realloc(ErtsAlcType_t n, void *extra, void *ptr, Uint size)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    Uint old_size;
-    Uint msize;
-    void *mptr;
-    void *res;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_x_mutex);
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    if (ptr) {
-	mptr = (void *) (((char *) ptr) - MAP_STAT_BLOCK_HEADER_SIZE);
-	old_size = ((MapStatBlock_t *) mptr)->size;
-    }
-    else {
-	mptr = NULL;
-	old_size = 0;
-    }
-
-    msize = size + MAP_STAT_BLOCK_HEADER_SIZE;
-    res = (*real_af->realloc)(n, real_af->extra, mptr, msize);
-    if (res) {
-	MapStatBlock_t *mb = (MapStatBlock_t *) res;
-
-	mb->size = size;
-	mb->type_no = n;
-	mb->pid = erts_get_current_pid();
-
-	stat_upd_realloc(n, size, old_size);
-
-	if (mptr != res) {
-
-	    if (mptr) {
-		if (mb->prev)
-		    mb->prev->next = mb;
-		else {
-		    ASSERT(mem_anchor == (MapStatBlock_t *) mptr);
-		    mem_anchor = mb;
-		}
-		if (mb->next)
-		    mb->next->prev = mb;
-	    }
-	    else {
-		mb->prev = NULL;
-		mb->next = mem_anchor;
-		if (mem_anchor)
-		    mem_anchor->prev = mb;
-		mem_anchor = mb;
-	    }
-
-	}
-
-	res = (void *) mb->mem;
-    }
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-	erts_mtx_unlock(&instr_x_mutex);
-    }
-
-    return res;
-}
-
-static void
-map_stat_free(ErtsAlcType_t n, void *extra, void *ptr)
-{
-    ErtsAllocatorFunctions_t *real_af = (ErtsAllocatorFunctions_t *) extra;
-    void *mptr;
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_lock(&instr_x_mutex);
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    if (ptr) {
-	MapStatBlock_t *mb;
-
-	mptr = (void *) (((char *) ptr) - MAP_STAT_BLOCK_HEADER_SIZE);
-	mb = (MapStatBlock_t *) mptr;
-
-	stat_upd_free(n, mb->size);
-
-	if (mb->prev)
-	    mb->prev->next = mb->next;
-	else
-	    mem_anchor = mb->next;
-	if (mb->next)
-	    mb->next->prev = mb->prev;
-    }
-    else {
-	mptr = NULL;
-    }
-
-    (*real_af->free)(n, real_af->extra, mptr);
-
-    if (!erts_is_allctr_wrapper_prelocked()) {
-	erts_mtx_unlock(&instr_mutex);
-	erts_mtx_unlock(&instr_x_mutex);
-    }
-
-}
-
-static void dump_memory_map_to_stream(fmtfn_t to, void* to_arg)
-{
-    ErtsAlcType_t n;
-    MapStatBlock_t *bp;
-    int lock = !ERTS_IS_CRASH_DUMPING;
-    if (lock) {
-	ASSERT(!erts_is_allctr_wrapper_prelocked());
-	erts_mtx_lock(&instr_mutex);
-    }
-
-    /* Write header */
-
-    erts_cbprintf(to, to_arg,
-	    "{instr_hdr,\n"
-	    " %lu,\n"
-	    " %lu,\n"
-	    " {",
-	    (unsigned long) ERTS_INSTR_VSN,
-	    (unsigned long) MAP_STAT_BLOCK_HEADER_SIZE);
-
-#if ERTS_ALC_N_MIN != 1
-#error ERTS_ALC_N_MIN is not 1
-#endif
-
-    for (n = ERTS_ALC_N_MIN; n <= ERTS_ALC_N_MAX; n++) {
-	ErtsAlcType_t t = ERTS_ALC_N2T(n);
-	ErtsAlcType_t a = ERTS_ALC_T2A(t);
-	ErtsAlcType_t c = ERTS_ALC_T2C(t);
-	const char *astr;
-
-	if (erts_allctrs_info[a].enabled)
-	    astr = ERTS_ALC_A2AD(a);
-	else
-	    astr = ERTS_ALC_A2AD(ERTS_ALC_A_SYSTEM);
-
-	erts_cbprintf(to, to_arg,
-		"%s{%s,%s,%s}%s",
-		(n == ERTS_ALC_N_MIN) ? "" : "  ",
-		ERTS_ALC_N2TD(n),
-		astr,
-		ERTS_ALC_C2CD(c),
-		(n == ERTS_ALC_N_MAX) ? "" : ",\n");
-    }
-
-    erts_cbprintf(to, to_arg, "}}.\n");
-
-    /* Write memory data */
-    for (bp = mem_anchor; bp; bp = bp->next) {
-	if (is_internal_pid(bp->pid))
-	    erts_cbprintf(to, to_arg,
-		    "{%lu, %lu, %lu, {%lu,%lu,%lu}}.\n",
-		    (UWord) bp->type_no,
-		    (UWord) bp->mem,
-		    (UWord) bp->size,
-		    (UWord) pid_channel_no(bp->pid),
-		    (UWord) pid_number(bp->pid),
-		    (UWord) pid_serial(bp->pid));
-	else
-	    erts_cbprintf(to, to_arg,
-		    "{%lu, %lu, %lu, undefined}.\n",
-		    (UWord) bp->type_no,
-		    (UWord) bp->mem,
-		    (UWord) bp->size);
-    }
-
-    if (lock)
-	erts_mtx_unlock(&instr_mutex);
-}
-
-int erts_instr_dump_memory_map_to(fmtfn_t to, void* to_arg)
-{
-    if (!erts_instr_memory_map)
-	return 0;
-
-    dump_memory_map_to_stream(to, to_arg);
-    return 1;
-}
-
-int erts_instr_dump_memory_map(const char *name)
-{
-    int fd;
-
-    if (!erts_instr_memory_map)
-	return 0;
-
-    fd = open(name, O_WRONLY | O_CREAT | O_TRUNC, 0640);
-    if (fd < 0)
-	return 0;
-
-    dump_memory_map_to_stream(erts_write_fd, (void*)&fd);
-
-    close(fd);
-    return 1;
-}
-
-Eterm erts_instr_get_memory_map(Process *proc)
-{
-    MapStatBlock_t *org_mem_anchor;
-    Eterm hdr_tuple, md_list, res;
-    Eterm *hp;
-    Uint hsz;
-    MapStatBlock_t *bp;
-#ifdef DEBUG
-    Eterm *end_hp;
-#endif
-    
-    if (!erts_instr_memory_map)
-	return am_false;
-
-    if (!atoms_initialized)
-	init_atoms();
-    if (!am_n)
-	init_am_n();
-    if (!am_c)
-	init_am_c();
-    if (!am_a)
-	init_am_a();
-
-    erts_mtx_lock(&instr_x_mutex);
-    erts_mtx_lock(&instr_mutex);
-
-    /* Header size */
-    hsz = 5 + 1 + (ERTS_ALC_N_MAX+1-ERTS_ALC_N_MIN)*(1 + 4);
-
-    /* Memory data list */
-    for (bp = mem_anchor; bp; bp = bp->next) {
-	if (is_internal_pid(bp->pid)) {
-#if (_PID_NUM_SIZE - 1 > MAX_SMALL)
-	    if (internal_pid_number(bp->pid) > MAX_SMALL)
-		hsz += BIG_UINT_HEAP_SIZE;
-#endif
-#if (_PID_SER_SIZE - 1 > MAX_SMALL)
-	    if (internal_pid_serial(bp->pid) > MAX_SMALL)
-		hsz += BIG_UINT_HEAP_SIZE;
-#endif
-	    hsz += 4;
-	}
-
-	if ((UWord) bp->mem > MAX_SMALL)
-	    hsz += BIG_UINT_HEAP_SIZE;
-	if (bp->size > MAX_SMALL)
-	    hsz += BIG_UINT_HEAP_SIZE;
-
-	hsz += 5 + 2;
-    }
-
-    hsz += 3; /* Root tuple */
-
-    org_mem_anchor = mem_anchor;
-    mem_anchor = NULL;
-
-    erts_mtx_unlock(&instr_mutex);
-
-    hp = HAlloc(proc, hsz); /* May end up calling map_stat_alloc() */
-
-    erts_mtx_lock(&instr_mutex);
-
-#ifdef DEBUG
-    end_hp = hp + hsz;
-#endif
-
-    {	/* Build header */
-	ErtsAlcType_t n;
-	Eterm type_map;
-	Uint *hp2 = hp;
-#ifdef DEBUG
-	Uint *hp2_end;
-#endif
-
-	hp += (ERTS_ALC_N_MAX + 1 - ERTS_ALC_N_MIN)*4;
-
-#ifdef DEBUG
-	hp2_end = hp;
-#endif
-
-	type_map = make_tuple(hp);
-	*(hp++) = make_arityval(ERTS_ALC_N_MAX + 1 - ERTS_ALC_N_MIN);
-
-	for (n = ERTS_ALC_N_MIN; n <= ERTS_ALC_N_MAX; n++) {
-	    ErtsAlcType_t t = ERTS_ALC_N2T(n);
-	    ErtsAlcType_t a = ERTS_ALC_T2A(t);
-	    ErtsAlcType_t c = ERTS_ALC_T2C(t);
-
-	    if (!erts_allctrs_info[a].enabled)
-		a = ERTS_ALC_A_SYSTEM;
-
-	    *(hp++) = TUPLE3(hp2, am_n[n], am_a[a], am_c[c]);
-	    hp2 += 4;
-	}
-
-	ASSERT(hp2 == hp2_end);
-
-	hdr_tuple = TUPLE4(hp,
-			   am.instr_hdr,
-			   make_small(ERTS_INSTR_VSN),
-			   make_small(MAP_STAT_BLOCK_HEADER_SIZE),
-			   type_map);
-
-	hp += 5;
-    }
-
-    /* Build memory data list */
-
-    for (md_list = NIL, bp = org_mem_anchor; bp; bp = bp->next) {
-	Eterm tuple;
-	Eterm type;
-	Eterm ptr;
-	Eterm size;
-	Eterm pid;
-
-	if (is_not_internal_pid(bp->pid))
-	    pid = am_undefined;
-	else {
-	    Eterm c;
-	    Eterm n;
-	    Eterm s;
-
-#if (ERST_INTERNAL_CHANNEL_NO > MAX_SMALL)
-#error Oversized internal channel number
-#endif
-	    c = make_small(ERST_INTERNAL_CHANNEL_NO);
-
-#if (_PID_NUM_SIZE - 1 > MAX_SMALL)
-	    if (internal_pid_number(bp->pid) > MAX_SMALL) {
-		n = uint_to_big(internal_pid_number(bp->pid), hp);
-		hp += BIG_UINT_HEAP_SIZE;
-	    }
-	    else
-#endif
-		n = make_small(internal_pid_number(bp->pid));
-
-#if (_PID_SER_SIZE - 1 > MAX_SMALL)
-	    if (internal_pid_serial(bp->pid) > MAX_SMALL) {
-		s = uint_to_big(internal_pid_serial(bp->pid), hp);
-		hp += BIG_UINT_HEAP_SIZE;
-	    }
-	    else
-#endif
-		s = make_small(internal_pid_serial(bp->pid));
-	    pid = TUPLE3(hp, c, n, s);
-	    hp += 4;
-	}
-
-
-#if ERTS_ALC_N_MAX > MAX_SMALL
-#error Oversized memory type number
-#endif
-	type = make_small(bp->type_no);
-
-	if ((UWord) bp->mem > MAX_SMALL) {
-	    ptr = uint_to_big((UWord) bp->mem, hp);
-	    hp += BIG_UINT_HEAP_SIZE;
-	}
-	else
-	    ptr = make_small((UWord) bp->mem);
-
-	if (bp->size > MAX_SMALL) {
-	    size = uint_to_big(bp->size, hp);
-	    hp += BIG_UINT_HEAP_SIZE;
-	}
-	else
-	    size = make_small(bp->size);
-
-	tuple = TUPLE4(hp, type, ptr, size, pid);
-	hp += 5;
-
-	md_list = CONS(hp, tuple, md_list);
-	hp += 2;
-    }
-
-    res = TUPLE2(hp, hdr_tuple, md_list);
-    
-    ASSERT(hp + 3 == end_hp);
-
-    if (mem_anchor) {
-	for (bp = mem_anchor; bp->next; bp = bp->next)
-	    ;
-	ASSERT(org_mem_anchor);
-	org_mem_anchor->prev = bp; 
-	bp->next = org_mem_anchor;
-    }
-    else {
-	mem_anchor = org_mem_anchor;
-    }
-
-    erts_mtx_unlock(&instr_mutex);
-    erts_mtx_unlock(&instr_x_mutex);
-
-    return res;
-}
-
-static ERTS_INLINE void
-begin_new_max_period(Stat_t *stat, int min, int max)
-{
-    int i;
-    for (i = min; i <= max; i++) {
-	stat[i].max_size = stat[i].size;
-	stat[i].max_blocks = stat[i].blocks;
-    }
-}
-
-static ERTS_INLINE void
-update_max_ever_values(Stat_t *stat, int min, int max)
-{
-    int i;
-    for (i = min; i <= max; i++) {
-	if (stat[i].max_size_ever < stat[i].max_size)
-	    stat[i].max_size_ever = stat[i].max_size;
-	if (stat[i].max_blocks_ever < stat[i].max_blocks)
-	    stat[i].max_blocks_ever = stat[i].max_blocks;
-    }
-}
-
-#define bld_string	erts_bld_string
-#define bld_tuple	erts_bld_tuple
-#define bld_tuplev	erts_bld_tuplev
-#define bld_list	erts_bld_list
-#define bld_2tup_list	erts_bld_2tup_list
-#define bld_uint	erts_bld_uint
-
-Eterm
-erts_instr_get_stat(Process *proc, Eterm what, int begin_max_period)
-{
-    int i, len, max, min, allctr;
-    Eterm *names, *values, res;
-    Uint arr_size, stat_size, hsz, *hszp, *hp, **hpp;
-    Stat_t *stat_src, *stat;
-
-    if (!erts_instr_stat)
-	return am_false;
-
-    if (!atoms_initialized)
-	init_atoms();
-
-    if (what == am.total) {
-	min		= 0;
-	max		= 0;
-	allctr		= 0;
-	stat_size	= sizeof(Stat_t);
-	stat_src	= &stats->tot;
-	if (!am_tot)
-	    init_am_tot();
-	names		= am_tot;
-    }
-    else if (what == am.allocators) {
-	min		= ERTS_ALC_A_MIN;
-	max		= ERTS_ALC_A_MAX;
-	allctr		= 1;
-	stat_size	= sizeof(Stat_t)*(ERTS_ALC_A_MAX+1);
-	stat_src	= stats->a;
-	if (!am_a)
-	    init_am_a();
-	names		= am_a;
-    }
-    else if (what == am.classes) {
-	min		= ERTS_ALC_C_MIN;
-	max		= ERTS_ALC_C_MAX;
-	allctr		= 0;
-	stat_size	= sizeof(Stat_t)*(ERTS_ALC_C_MAX+1);
-	stat_src	= stats->c;
-	if (!am_c)
-	    init_am_c();
-	names		= &am_c[ERTS_ALC_C_MIN];
-    }
-    else if (what == am.types) {
-	min		= ERTS_ALC_N_MIN;
-	max		= ERTS_ALC_N_MAX;
-	allctr		= 0;
-	stat_size	= sizeof(Stat_t)*(ERTS_ALC_N_MAX+1);
-	stat_src	= stats->n;
-	if (!am_n)
-	    init_am_n();
-	names		= &am_n[ERTS_ALC_N_MIN];
-    }
-    else {
-	return THE_NON_VALUE;
-    }
-
-    stat = (Stat_t *) erts_alloc(ERTS_ALC_T_TMP, stat_size);
-
-    arr_size = (max - min + 1)*sizeof(Eterm);
-
-    if (allctr)
-	names = (Eterm *) erts_alloc(ERTS_ALC_T_TMP, arr_size);
-
-    values = (Eterm *) erts_alloc(ERTS_ALC_T_TMP, arr_size);
-
-    erts_mtx_lock(&instr_mutex);
-
-    update_max_ever_values(stat_src, min, max);
-
-    sys_memcpy((void *) stat, (void *) stat_src, stat_size);
-
-    if (begin_max_period)
-	begin_new_max_period(stat_src, min, max);
-
-    erts_mtx_unlock(&instr_mutex);
-
-    hsz = 0;
-    hszp = &hsz;
-    hpp = NULL;
-
- restart_bld:
-
-    len = 0;
-    for (i = min; i <= max; i++) {
-	if (!allctr || erts_allctrs_info[i].enabled) {
-	    Eterm s[2];
-
-	    if (allctr)
-		names[len] = am_a[i];
-	    
-	    s[0] = bld_tuple(hpp, hszp, 4,
-			     am.sizes,
-			     bld_uint(hpp, hszp, stat[i].size),
-			     bld_uint(hpp, hszp, stat[i].max_size),
-			     bld_uint(hpp, hszp, stat[i].max_size_ever));
-
-	    s[1] = bld_tuple(hpp, hszp, 4,
-			     am.blocks,
-			     bld_uint(hpp, hszp, stat[i].blocks),
-			     bld_uint(hpp, hszp, stat[i].max_blocks),
-			     bld_uint(hpp, hszp, stat[i].max_blocks_ever));
-
-	    values[len] = bld_list(hpp, hszp, 2, s);
-	    
-	    len++;
-	}
-    }
-
-    res = bld_2tup_list(hpp, hszp, len, names, values);
-
-    if (!hpp) {
-	hp = HAlloc(proc, hsz);
-	hszp = NULL;
-	hpp = &hp;
-	goto restart_bld;
-    }
-
-    erts_free(ERTS_ALC_T_TMP, (void *) stat);
-    erts_free(ERTS_ALC_T_TMP, (void *) values);
-    if (allctr)
-	erts_free(ERTS_ALC_T_TMP, (void *) names);
-
-    return res;
-}
-
-static void
-dump_stat_to_stream(fmtfn_t to, void* to_arg, int begin_max_period)
-{
-    ErtsAlcType_t i, a_max, a_min;
-
-    erts_mtx_lock(&instr_mutex);
-
-    erts_cbprintf(to, to_arg,
-	    "{instr_vsn,%lu}.\n",
-	    (unsigned long) ERTS_INSTR_VSN);
-    
-    update_max_ever_values(&stats->tot, 0, 0);
-
-    erts_cbprintf(to, to_arg,
-	    "{total,[{total,[{sizes,%lu,%lu,%lu},{blocks,%lu,%lu,%lu}]}]}.\n",
-	    (UWord) stats->tot.size,
-	    (UWord) stats->tot.max_size,
-	    (UWord) stats->tot.max_size_ever,
-	    (UWord) stats->tot.blocks,
-	    (UWord) stats->tot.max_blocks,
-	    (UWord) stats->tot.max_blocks_ever);
-
-    a_max = 0;
-    a_min = ~0;
-    for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	if (erts_allctrs_info[i].enabled) {
-	    if (a_min > i)
-		a_min = i;
-	    if (a_max < i)
-		a_max = i;
-	}
-    }
-
-    ASSERT(ERTS_ALC_A_MIN <= a_min && a_min <= ERTS_ALC_A_MAX);
-    ASSERT(ERTS_ALC_A_MIN <= a_max && a_max <= ERTS_ALC_A_MAX);
-    ASSERT(a_min <= a_max);
-
-    update_max_ever_values(stats->a, a_min, a_max);
-
-    for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	if (erts_allctrs_info[i].enabled) {
-	    erts_cbprintf(to, to_arg,
-		    "%s{%s,[{sizes,%lu,%lu,%lu},{blocks,%lu,%lu,%lu}]}%s",
-		    i == a_min ? "{allocators,\n [" : "  ",
-		    ERTS_ALC_A2AD(i),
-		    (UWord) stats->a[i].size,
-		    (UWord) stats->a[i].max_size,
-		    (UWord) stats->a[i].max_size_ever,
-		    (UWord) stats->a[i].blocks,
-		    (UWord) stats->a[i].max_blocks,
-		    (UWord) stats->a[i].max_blocks_ever,
-		    i == a_max ? "]}.\n" : ",\n");
-	}
-    }
-
-    update_max_ever_values(stats->c, ERTS_ALC_C_MIN, ERTS_ALC_C_MAX);
-
-    for (i = ERTS_ALC_C_MIN; i <= ERTS_ALC_C_MAX; i++) {
-	erts_cbprintf(to, to_arg,
-		"%s{%s,[{sizes,%lu,%lu,%lu},{blocks,%lu,%lu,%lu}]}%s",
-		i == ERTS_ALC_C_MIN ? "{classes,\n [" : "  ",
-		ERTS_ALC_C2CD(i),
-		(UWord) stats->c[i].size,
-		(UWord) stats->c[i].max_size,
-		(UWord) stats->c[i].max_size_ever,
-		(UWord) stats->c[i].blocks,
-		(UWord) stats->c[i].max_blocks,
-		(UWord) stats->c[i].max_blocks_ever,
-		i == ERTS_ALC_C_MAX ? "]}.\n" :  ",\n" );
-    }
-
-    update_max_ever_values(stats->n, ERTS_ALC_N_MIN, ERTS_ALC_N_MAX);
-
-    for (i = ERTS_ALC_N_MIN; i <= ERTS_ALC_N_MAX; i++) {
-	erts_cbprintf(to, to_arg,
-		"%s{%s,[{sizes,%lu,%lu,%lu},{blocks,%lu,%lu,%lu}]}%s",
-		i == ERTS_ALC_N_MIN ? "{types,\n [" : "  ",
-		ERTS_ALC_N2TD(i),
-		(UWord) stats->n[i].size,
-		(UWord) stats->n[i].max_size,
-		(UWord) stats->n[i].max_size_ever,
-		(UWord) stats->n[i].blocks,
-		(UWord) stats->n[i].max_blocks,
-		(UWord) stats->n[i].max_blocks_ever,
-		i == ERTS_ALC_N_MAX ? "]}.\n" :  ",\n" );
-    }
-
-    if (begin_max_period) {
-	begin_new_max_period(&stats->tot, 0, 0);
-	begin_new_max_period(stats->a, a_min, a_max);
-	begin_new_max_period(stats->c, ERTS_ALC_C_MIN, ERTS_ALC_C_MAX);
-	begin_new_max_period(stats->n, ERTS_ALC_N_MIN, ERTS_ALC_N_MAX);
-    }
-
-    erts_mtx_unlock(&instr_mutex);
-
-}
-
-int erts_instr_dump_stat_to(fmtfn_t to, void* to_arg, int begin_max_period)
-{
-    if (!erts_instr_stat)
-	return 0;
-
-    dump_stat_to_stream(to, to_arg, begin_max_period);
-    return 1;
-}
-
-int erts_instr_dump_stat(const char *name, int begin_max_period)
-{
-    int fd;
-
-    if (!erts_instr_stat)
-	return 0;
-
-    fd = open(name, O_WRONLY | O_CREAT | O_TRUNC,0640);
-    if (fd < 0)
-	return 0;
-
-    dump_stat_to_stream(erts_write_fd, (void*)&fd, begin_max_period);
-
-    close(fd);
-    return 1;
-}
-
-
-Uint
-erts_instr_get_total(void)
-{
-    return erts_instr_stat ? stats->tot.size : 0;
-}
-
-Uint
-erts_instr_get_max_total(void)
-{
-    if (erts_instr_stat) {
-	update_max_ever_values(&stats->tot, 0, 0);
-	return stats->tot.max_size_ever;
-    }
-    return 0;
-}
-
-Eterm
-erts_instr_get_type_info(Process *proc)
-{
-    Eterm res, *tpls;
-    Uint hsz, *hszp, *hp, **hpp;
-    ErtsAlcType_t n;
-
-    if (!am_n)
-	init_am_n();
-    if (!am_a)
-	init_am_a();
-    if (!am_c)
-	init_am_c();
-
-    tpls = (Eterm *) erts_alloc(ERTS_ALC_T_TMP,
-				(ERTS_ALC_N_MAX-ERTS_ALC_N_MIN+1)
-				* sizeof(Eterm));
-    hsz = 0;
-    hszp = &hsz;
-    hpp = NULL;
-
- restart_bld:
-
-#if ERTS_ALC_N_MIN != 1
-#error ERTS_ALC_N_MIN is not 1
-#endif
-
-    for (n = ERTS_ALC_N_MIN; n <= ERTS_ALC_N_MAX; n++) {
-	ErtsAlcType_t t = ERTS_ALC_N2T(n);
-	ErtsAlcType_t a = ERTS_ALC_T2A(t);
-	ErtsAlcType_t c = ERTS_ALC_T2C(t);
-
-	if (!erts_allctrs_info[a].enabled)
-	    a = ERTS_ALC_A_SYSTEM;
-
-	tpls[n - ERTS_ALC_N_MIN]
-	    = bld_tuple(hpp, hszp, 3, am_n[n], am_a[a], am_c[c]);
-    }
-
-    res = bld_tuplev(hpp, hszp, ERTS_ALC_N_MAX-ERTS_ALC_N_MIN+1, tpls);
-    
-    if (!hpp) {
-	hp = HAlloc(proc, hsz);
-	hszp = NULL;
-	hpp = &hp;
-	goto restart_bld;
-    }
-
-    erts_free(ERTS_ALC_T_TMP, tpls);
-
-    return res;
-}
-
-Uint
-erts_instr_init(int stat, int map_stat)
-{
-    Uint extra_sz;
-    int i;
-
-    am_tot = NULL;
-    am_n = NULL;
-    am_c = NULL;
-    am_a = NULL;
-
-    erts_instr_memory_map = 0;
-    erts_instr_stat = 0;
-    atoms_initialized = 0;
-
-    if (!stat && !map_stat)
-	return 0;
-
-    stats = erts_alloc(ERTS_ALC_T_INSTR_INFO, sizeof(struct stats_));
-
-    erts_mtx_init(&instr_mutex, "instr", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
-
-    mem_anchor = NULL;
-
-    /* Install instrumentation functions */
-    ERTS_CT_ASSERT(sizeof(erts_allctrs) == sizeof(real_allctrs));
-
-    sys_memcpy((void *)real_allctrs,(void *)erts_allctrs,sizeof(erts_allctrs));
-
-    sys_memzero((void *) &stats->tot, sizeof(Stat_t));
-    sys_memzero((void *) stats->a, sizeof(Stat_t)*(ERTS_ALC_A_MAX+1));
-    sys_memzero((void *) stats->c, sizeof(Stat_t)*(ERTS_ALC_C_MAX+1));
-    sys_memzero((void *) stats->n, sizeof(Stat_t)*(ERTS_ALC_N_MAX+1));
-
-    for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	if (erts_allctrs_info[i].enabled)
-	    stats->ap[i] = &stats->a[i];
-	else
-	    stats->ap[i] = &stats->a[ERTS_ALC_A_SYSTEM];
-    }
-
-    if (map_stat) {
-
-	erts_mtx_init(&instr_x_mutex, "instr_x", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
-
-	erts_instr_memory_map = 1;
-	erts_instr_stat = 1;
-	for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	    erts_allctrs[i].alloc	= map_stat_alloc;
-	    erts_allctrs[i].realloc	= map_stat_realloc;
-	    erts_allctrs[i].free	= map_stat_free;
-	    erts_allctrs[i].extra	= (void *) &real_allctrs[i];
-	}
-	instr_wrapper.lock = map_stat_pre_lock;
-	instr_wrapper.unlock = map_stat_pre_unlock;
-	extra_sz = MAP_STAT_BLOCK_HEADER_SIZE;
-    }
-    else {
-	erts_instr_stat = 1;
-	for (i = ERTS_ALC_A_MIN; i <= ERTS_ALC_A_MAX; i++) {
-	    erts_allctrs[i].alloc	= stat_alloc;
-	    erts_allctrs[i].realloc	= stat_realloc;
-	    erts_allctrs[i].free	= stat_free;
-	    erts_allctrs[i].extra	= (void *) &real_allctrs[i];
-	}
-	instr_wrapper.lock = stat_pre_lock;
-	instr_wrapper.unlock = stat_pre_unlock;
-	extra_sz = STAT_BLOCK_HEADER_SIZE;
-    }
-    erts_allctr_wrapper_prelock_init(&instr_wrapper);
-    return extra_sz;
-}
-
diff --git a/erts/emulator/beam/erl_instrument.h b/erts/emulator/beam/erl_instrument.h
deleted file mode 100644
index 351172b2fa..0000000000
--- a/erts/emulator/beam/erl_instrument.h
+++ /dev/null
@@ -1,42 +0,0 @@
-/*
- * %CopyrightBegin%
- * 
- * Copyright Ericsson AB 2003-2016. All Rights Reserved.
- * 
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * 
- * %CopyrightEnd%
- */
-
-#ifndef ERL_INSTRUMENT_H__
-#define ERL_INSTRUMENT_H__
-
-#include "erl_mtrace.h"
-
-#define ERTS_INSTR_VSN 2
-
-extern int erts_instr_memory_map;
-extern int erts_instr_stat;
-
-Uint  erts_instr_init(int stat, int map_stat);
-int   erts_instr_dump_memory_map_to(fmtfn_t to, void* to_arg);
-int   erts_instr_dump_memory_map(const char *name);
-Eterm erts_instr_get_memory_map(Process *process);
-int   erts_instr_dump_stat_to(fmtfn_t to, void* to_arg, int begin_max_period);
-int   erts_instr_dump_stat(const char *name, int begin_max_period);
-Eterm erts_instr_get_stat(Process *proc, Eterm what, int begin_max_period);
-Eterm erts_instr_get_type_info(Process *proc);
-Uint  erts_instr_get_total(void);
-Uint  erts_instr_get_max_total(void);
-
-#endif
diff --git a/erts/emulator/beam/erl_io_queue.c b/erts/emulator/beam/erl_io_queue.c
index eb156cc578..d779d1031a 100644
--- a/erts/emulator/beam/erl_io_queue.c
+++ b/erts/emulator/beam/erl_io_queue.c
@@ -658,7 +658,7 @@ io_list_vec_count(Eterm obj, Uint *v_size,
 int
 erts_ioq_iolist_vec_len(Eterm obj, int* vsize, Uint* csize,
                         Uint* pvsize, Uint* pcsize,
-                        Uint* total_size, Uint blimit)
+                        size_t* total_size, Uint blimit)
 {
     DECLARE_ESTACK(s);
     Eterm* objp;
@@ -669,7 +669,7 @@ erts_ioq_iolist_vec_len(Eterm obj, int* vsize, Uint* csize,
     Uint p_v_size = 0;
     Uint p_c_size = 0;
     Uint p_in_clist = 0;
-    Uint total;
+    size_t total;
 
     goto L_jump_start;  /* avoid a push */
 
@@ -816,24 +816,15 @@ static Eterm iol2v_make_sub_bin(iol2v_state_t *state, Eterm bin_term,
 }
 
 static Eterm iol2v_promote_acc(iol2v_state_t *state) {
-    ProcBin *pb;
-
-    state->acc = erts_bin_realloc(state->acc, state->acc_size);
-
-    pb = (ProcBin*)HAlloc(state->process, PROC_BIN_SIZE);
-    pb->thing_word = HEADER_PROC_BIN;
-    pb->size = state->acc_size;
-    pb->val = state->acc;
-    pb->bytes = (byte*)(state->acc)->orig_bytes;
-    pb->flags = 0;
-    pb->next = MSO(state->process).first;
-    OH_OVERHEAD(&(MSO(state->process)), pb->size / sizeof(Eterm));
-    MSO(state->process).first = (struct erl_off_heap_header*)pb;
+    Eterm bin;
 
+    bin = erts_build_proc_bin(&MSO(state->process),
+                              HAlloc(state->process, PROC_BIN_SIZE),
+                              erts_bin_realloc(state->acc, state->acc_size));
     state->acc_size = 0;
     state->acc = NULL;
 
-    return make_binary(pb);
+    return bin;
 }
 
 /* Destructively enqueues a term to the result list, saving us the hassle of
diff --git a/erts/emulator/beam/erl_io_queue.h b/erts/emulator/beam/erl_io_queue.h
index 51abe99510..7d0fe6751c 100644
--- a/erts/emulator/beam/erl_io_queue.h
+++ b/erts/emulator/beam/erl_io_queue.h
@@ -103,7 +103,7 @@ Uint erts_ioq_sizeq(ErtsIOQueue *q);
 
 int erts_ioq_iolist_vec_len(Eterm obj, int* vsize, Uint* csize,
                             Uint* pvsize, Uint* pcsize,
-                            Uint* total_size, Uint blimit);
+                            size_t* total_size, Uint blimit);
 int erts_ioq_iolist_to_vec(Eterm obj, SysIOVec* iov,
                            ErtsIOQBinary** binv, ErtsIOQBinary* cbin,
                            Uint bin_limit, int driver_binary);
@@ -111,7 +111,7 @@ int erts_ioq_iolist_to_vec(Eterm obj, SysIOVec* iov,
 ERTS_GLB_INLINE
 int erts_ioq_iodata_vec_len(Eterm obj, int* vsize, Uint* csize,
                             Uint* pvsize, Uint* pcsize,
-                            Uint* total_size, Uint blimit);
+                            size_t* total_size, Uint blimit);
 ERTS_GLB_INLINE
 int erts_ioq_iodata_to_vec(Eterm obj, SysIOVec* iov,
                            ErtsIOQBinary** binv, ErtsIOQBinary* cbin,
@@ -123,7 +123,7 @@ int erts_ioq_iodata_to_vec(Eterm obj, SysIOVec* iov,
 ERTS_GLB_INLINE
 int erts_ioq_iodata_vec_len(Eterm obj, int* vsize, Uint* csize,
                             Uint* pvsize, Uint* pcsize,
-                            Uint* total_size, Uint blimit) {
+                            size_t* total_size, Uint blimit) {
   if (is_binary(obj)) {
     /* We optimize for when we get a procbin without a bit-offset
      * that fits in one iov slot
diff --git a/erts/emulator/beam/erl_lock_check.c b/erts/emulator/beam/erl_lock_check.c
index 189c88ac4a..463ae898a3 100644
--- a/erts/emulator/beam/erl_lock_check.c
+++ b/erts/emulator/beam/erl_lock_check.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2005-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2005-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -41,6 +41,7 @@
 #include "erl_lock_check.h"
 #include "erl_term.h"
 #include "erl_threads.h"
+#include "erl_atom_table.h"
 
 typedef struct {
     char *name;
@@ -75,13 +76,10 @@ static erts_lc_lock_order_t erts_lock_order[] = {
      *						 if only one lock use
      *						 the lock name)"
      */
-#ifdef ERTS_SMP
+    {   "NO LOCK",                              NULL                    },
     {	"driver_lock",				"driver_name"		},
     {	"port_lock",				"port_id"		},
-#endif
     {	"port_data_lock",			"address"		},
-#ifdef ERTS_SMP
-    {	"bif_timers",				NULL			},
     {	"reg_tab",				NULL			},
     {	"proc_main",				"pid"			},
     {   "old_code",                             "address"               },
@@ -89,47 +87,37 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"hipe_mfait_lock",			NULL			},
 #endif
     {	"nodes_monitors",			NULL			},
-#ifdef ERTS_SMP
+    {	"meta_name_tab",	         	"address"		},
+    {	"db_tab",				"address"		},
+    {	"db_tab_fix",				"address"		},
+    {	"db_hash_slot",				"address"		},
     {	"resource_monitors",			"address"	        },
-#endif
     {   "driver_list",                          NULL                    },
-    {	"proc_link",				"pid"			},
     {	"proc_msgq",				"pid"			},
     {	"proc_btm",				"pid"			},
     {	"dist_entry",				"address"		},
     {	"dist_entry_links",			"address"		},
     {   "code_write_permission",                NULL                    },
     {	"purge_state",		      		NULL			},
-    {	"meta_name_tab",	         	"address"		},
-    {	"db_tab",				"address"		},
     {	"proc_status",				"pid"			},
     {	"proc_trace",				"pid"			},
-    {	"db_tab_fix",				"address"		},
-    {	"db_hash_slot",				"address"		},
     {	"node_table",				NULL			},
     {	"dist_table",				NULL			},
     {	"sys_tracers",				NULL			},
-    {	"module_tab",				NULL			},
     {	"export_tab",				NULL			},
     {	"fun_tab",				NULL			},
     {	"environ",				NULL			},
     {	"release_literal_areas",		NULL			},
-#endif
-    {	"efile_drv",				"address"		},
     {	"drv_ev_state_grow",			NULL,   		},
     {	"drv_ev_state",				"address"		},
     {	"safe_hash",				"address"		},
-    {   "pollset_rm_list",                      NULL                    },
-    {   "removed_fd_pre_alloc_lock",            "address"               },
     {   "state_prealloc",                       NULL                    },
     {	"schdlr_sspnd",				NULL			},
     {	"migration_info_update",		NULL			},
     {	"run_queue",				"address"		},
-#ifdef ERTS_DIRTY_SCHEDULERS
     {   "dirty_run_queue_sleep_list",		"address"		},
     {	"dirty_gc_info",			NULL			},
     {	"dirty_break_point_index",		NULL			},
-#endif
     {	"process_table",			NULL			},
     {	"cpu_info",				NULL			},
     {	"pollset",				"address"		},
@@ -144,54 +132,31 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"async_enq_mtx",			NULL			},
     {   "msacc_list_mutex",                     NULL                    },
     {   "msacc_unmanaged_mutex",                NULL                    },
-#ifdef ERTS_SMP
     {	"atom_tab",				NULL			},
-    {	"misc_op_list_pre_alloc_lock",		"address"		},
-    {	"message_pre_alloc_lock",		"address"		},
-    {	"ptimer_pre_alloc_lock",		"address",		},
-    {	"btm_pre_alloc_lock",			NULL,			},
     {	"dist_entry_out_queue",			"address"		},
     {	"port_sched_lock",			"port_id"		},
     {	"sys_msg_q", 				NULL			},
     {	"tracer_mtx", 				NULL			},
     {   "port_table",                           NULL                    },
-#endif
     {	"magic_ref_table",			"address"		},
     {	"mtrace_op",				NULL			},
     {	"instr_x",				NULL			},
     {	"instr",				NULL			},
-#ifdef ERTS_SMP
-    {	"pollsets_lock",			NULL			},
-#endif
     {	"alcu_allocator",			"index"			},
     {	"mseg",					NULL			},
-#ifdef ERTS_SMP
-    {	"port_task_pre_alloc_lock",		"address"		},
-    {	"proclist_pre_alloc_lock",		"address"		},
-    {	"xports_list_pre_alloc_lock",		"address"		},
-    {	"inet_buffer_stack_lock",		NULL			},
-    {	"system_block",				NULL			},
     {	"get_time",				NULL			},
     {	"get_corrected_time",			NULL			},
     {	"runtime",				NULL			},
-    {	"breakpoints",				NULL			},
     {	"pix_lock",				"address"		},
-    {	"run_queues_lists",			NULL			},
     {	"sched_stat",				NULL			},
-#endif
     {	"async_init_mtx",			NULL			},
 #ifdef __WIN32__
 #ifdef DEBUG
     {   "save_ops_lock",                        NULL                    },
 #endif
 #endif
-#ifdef	USE_VM_PROBES
-    {   "efile_drv dtrace mutex",               NULL                    },
-#endif
     {	"mtrace_buf",				NULL			},
-#ifdef ERTS_SMP
     {	"os_monotonic_time",			NULL			},
-#endif
     {	"erts_alloc_hard_debug",		NULL			},
     {	"hard_dbg_mseg",		        NULL	                },
     {	"erts_mmap",				NULL			}
@@ -216,10 +181,10 @@ static const char *rw_op_str(erts_lock_options_t options)
     return erts_lock_options_get_short_desc(options);
 }
 
-typedef struct erts_lc_locked_lock_t_ erts_lc_locked_lock_t;
-struct erts_lc_locked_lock_t_ {
-    erts_lc_locked_lock_t *next;
-    erts_lc_locked_lock_t *prev;
+typedef struct lc_locked_lock_t_ lc_locked_lock_t;
+struct lc_locked_lock_t_ {
+    lc_locked_lock_t *next;
+    lc_locked_lock_t *prev;
     UWord extra;
     Sint16 id;
     char *file;
@@ -229,209 +194,224 @@ struct erts_lc_locked_lock_t_ {
 };
 
 typedef struct {
-    erts_lc_locked_lock_t *first;
-    erts_lc_locked_lock_t *last;
-} erts_lc_locked_lock_list_t;
+    lc_locked_lock_t *first;
+    lc_locked_lock_t *last;
+} lc_locked_lock_list_t;
+
+typedef union lc_free_block_t_ lc_free_block_t;
+union lc_free_block_t_ {
+    lc_free_block_t *next;
+    lc_locked_lock_t lock;
+};
+
+typedef struct {
+    /*
+     * m[X][Y] & 1 if we locked X directly after Y was locked.
+     * m[X][Y] & 2 if we locked X indirectly after Y was locked.
+     * m[X][0] = 1 if we locked X when nothing else was locked.
+     * m[0][] is unused as it would represent locking "NO LOCK"
+     */
+    char m[ERTS_LOCK_ORDER_SIZE][ERTS_LOCK_ORDER_SIZE];
+
+} lc_matrix_t;
+
+static lc_matrix_t tot_lc_matrix;
 
-typedef struct erts_lc_locked_locks_t_ erts_lc_locked_locks_t;
-struct erts_lc_locked_locks_t_ {
+#define ERTS_LC_FB_CHUNK_SIZE 10
+
+typedef struct lc_alloc_chunk_t_ lc_alloc_chunk_t;
+struct lc_alloc_chunk_t_ {
+    lc_alloc_chunk_t* next;
+    lc_free_block_t array[ERTS_LC_FB_CHUNK_SIZE];
+};
+
+typedef struct lc_thread_t_ lc_thread_t;
+struct lc_thread_t_ {
     char *thread_name;
     int emu_thread;
     erts_tid_t tid;
-    erts_lc_locked_locks_t *next;
-    erts_lc_locked_locks_t *prev;
-    erts_lc_locked_lock_list_t locked;
-    erts_lc_locked_lock_list_t required;
-};
-
-typedef union erts_lc_free_block_t_ erts_lc_free_block_t;
-union erts_lc_free_block_t_ {
-    erts_lc_free_block_t *next;
-    erts_lc_locked_lock_t lock;
+    lc_thread_t *next;
+    lc_thread_t *prev;
+    lc_locked_lock_list_t locked;
+    lc_locked_lock_list_t required;
+    lc_free_block_t *free_blocks;
+    lc_alloc_chunk_t *chunks;
+    lc_matrix_t matrix;
 };
 
 static ethr_tsd_key locks_key;
 
-static erts_lc_locked_locks_t *erts_locked_locks = NULL;
-
-static erts_lc_free_block_t *free_blocks = NULL;
-
-#ifdef ERTS_LC_STATIC_ALLOC
-#define ERTS_LC_FB_CHUNK_SIZE 10000
-#else
-#define ERTS_LC_FB_CHUNK_SIZE 10
-#endif
-
-static ethr_spinlock_t free_blocks_lock;
+static lc_thread_t *lc_threads = NULL;
+static ethr_spinlock_t lc_threads_lock;
 
 static ERTS_INLINE void
-lc_lock(void)
+lc_lock_threads(void)
 {
-    ethr_spin_lock(&free_blocks_lock);
+    ethr_spin_lock(&lc_threads_lock);
 }
 
 static ERTS_INLINE void
-lc_unlock(void)
+lc_unlock_threads(void)
 {
-    ethr_spin_unlock(&free_blocks_lock);
+    ethr_spin_unlock(&lc_threads_lock);
 }
 
-static ERTS_INLINE void lc_free(void *p)
+static ERTS_INLINE void lc_free(lc_thread_t* thr, lc_locked_lock_t *p)
 {
-    erts_lc_free_block_t *fb = (erts_lc_free_block_t *) p;
+    lc_free_block_t *fb = (lc_free_block_t *) p;
 #ifdef DEBUG
-    memset((void *) p, 0xdf, sizeof(erts_lc_free_block_t));
+    sys_memset((void *) p, 0xdf, sizeof(lc_free_block_t));
 #endif
-    lc_lock();
-    fb->next = free_blocks;
-    free_blocks = fb;
-    lc_unlock();   
-}
-
-#ifdef ERTS_LC_STATIC_ALLOC
-
-static void *lc_core_alloc(void)
-{
-    lc_unlock();
-    ERTS_INTERNAL_ERROR("Lock checker out of memory!\n");
+    fb->next = thr->free_blocks;
+    thr->free_blocks = fb;
 }
 
-#else
-
-static void *lc_core_alloc(void)
+static lc_locked_lock_t *lc_core_alloc(lc_thread_t* thr)
 {
     int i;
-    erts_lc_free_block_t *fbs;
-    lc_unlock();
-    fbs = (erts_lc_free_block_t *) malloc(sizeof(erts_lc_free_block_t)
-					  * ERTS_LC_FB_CHUNK_SIZE);
-    if (!fbs) {
+    lc_alloc_chunk_t* chunk;
+    lc_free_block_t* fbs;
+    chunk = (lc_alloc_chunk_t*) malloc(sizeof(lc_alloc_chunk_t));
+    if (!chunk) {
         ERTS_INTERNAL_ERROR("Lock checker failed to allocate memory!");
     }
+    chunk->next = thr->chunks;
+    thr->chunks = chunk;
+
+    fbs = chunk->array;
     for (i = 1; i < ERTS_LC_FB_CHUNK_SIZE - 1; i++) {
 #ifdef DEBUG
-	memset((void *) &fbs[i], 0xdf, sizeof(erts_lc_free_block_t));
+	sys_memset((void *) &fbs[i], 0xdf, sizeof(lc_free_block_t));
 #endif
 	fbs[i].next = &fbs[i+1];
     }
 #ifdef DEBUG
-    memset((void *) &fbs[ERTS_LC_FB_CHUNK_SIZE-1],
-	   0xdf, sizeof(erts_lc_free_block_t));
+    sys_memset((void *) &fbs[ERTS_LC_FB_CHUNK_SIZE-1],
+	   0xdf, sizeof(lc_free_block_t));
 #endif
-    lc_lock();
-    fbs[ERTS_LC_FB_CHUNK_SIZE-1].next = free_blocks;
-    free_blocks = &fbs[1];
-    return (void *) &fbs[0];
+    fbs[ERTS_LC_FB_CHUNK_SIZE-1].next = thr->free_blocks;
+    thr->free_blocks = &fbs[1];
+    return &fbs[0].lock;
 }
 
-#endif
-
-static ERTS_INLINE void *lc_alloc(void)
+static ERTS_INLINE lc_locked_lock_t *lc_alloc(lc_thread_t* thr)
 {
-    void *res;
-    lc_lock();
-    if (!free_blocks)
-	res = lc_core_alloc();
+    lc_locked_lock_t *res;
+    if (!thr->free_blocks)
+	res = lc_core_alloc(thr);
     else {
-	res = (void *) free_blocks;
-	free_blocks = free_blocks->next;
+	res = &thr->free_blocks->lock;
+	thr->free_blocks = thr->free_blocks->next;
     }
-    lc_unlock();
     return res;
 }
 
 
-static erts_lc_locked_locks_t *
-create_locked_locks(char *thread_name)
+static lc_thread_t *
+create_thread_data(char *thread_name)
 {
-    erts_lc_locked_locks_t *l_lcks = malloc(sizeof(erts_lc_locked_locks_t));
-    if (!l_lcks)
+    lc_thread_t *thr = malloc(sizeof(lc_thread_t));
+    if (!thr)
 	ERTS_INTERNAL_ERROR("Lock checker failed to allocate memory!");
 
-    l_lcks->thread_name = strdup(thread_name ? thread_name : "unknown");
-    if (!l_lcks->thread_name)
+    thr->thread_name = strdup(thread_name ? thread_name : "unknown");
+    if (!thr->thread_name)
 	ERTS_INTERNAL_ERROR("Lock checker failed to allocate memory!");
 
-    l_lcks->emu_thread = 0;
-    l_lcks->tid = erts_thr_self();
-    l_lcks->required.first = NULL;
-    l_lcks->required.last = NULL;
-    l_lcks->locked.first = NULL;
-    l_lcks->locked.last = NULL;
-    l_lcks->prev = NULL;
-    lc_lock();
-    l_lcks->next = erts_locked_locks;
-    if (erts_locked_locks)
-	erts_locked_locks->prev = l_lcks;
-    erts_locked_locks = l_lcks;
-    lc_unlock();
-    erts_tsd_set(locks_key, (void *) l_lcks);
-    return l_lcks;
-}
+    thr->emu_thread = 0;
+    thr->tid = erts_thr_self();
+    thr->required.first = NULL;
+    thr->required.last = NULL;
+    thr->locked.first = NULL;
+    thr->locked.last = NULL;
+    thr->prev = NULL;
+    thr->free_blocks = NULL;
+    thr->chunks = NULL;
+    sys_memzero(&thr->matrix, sizeof(thr->matrix));
+
+    lc_lock_threads();
+    thr->next = lc_threads;
+    if (lc_threads)
+	lc_threads->prev = thr;
+    lc_threads = thr;
+    lc_unlock_threads();
+    erts_tsd_set(locks_key, (void *) thr);
+    return thr;
+}
+
+static void collect_matrix(lc_matrix_t*);
 
 static void
-destroy_locked_locks(erts_lc_locked_locks_t *l_lcks)
-{
-    ASSERT(l_lcks->thread_name);
-    free((void *) l_lcks->thread_name);
-    ASSERT(l_lcks->required.first == NULL);
-    ASSERT(l_lcks->required.last == NULL);
-    ASSERT(l_lcks->locked.first == NULL);
-    ASSERT(l_lcks->locked.last == NULL);
-
-    lc_lock();
-    if (l_lcks->prev)
-	l_lcks->prev->next = l_lcks->next;
+destroy_thread_data(lc_thread_t *thr)
+{
+    ASSERT(thr->thread_name);
+    free((void *) thr->thread_name);
+    ASSERT(thr->required.first == NULL);
+    ASSERT(thr->required.last == NULL);
+    ASSERT(thr->locked.first == NULL);
+    ASSERT(thr->locked.last == NULL);
+
+    lc_lock_threads();
+    if (thr->prev)
+	thr->prev->next = thr->next;
     else {
-	ASSERT(erts_locked_locks == l_lcks);
-	erts_locked_locks = l_lcks->next;
+	ASSERT(lc_threads == thr);
+	lc_threads = thr->next;
     }
+    if (thr->next)
+	thr->next->prev = thr->prev;
+
+    collect_matrix(&thr->matrix);
 
-    if (l_lcks->next)
-	l_lcks->next->prev = l_lcks->prev;
-    lc_unlock();
+    lc_unlock_threads();
 
-    free((void *) l_lcks);
+    while (thr->chunks) {
+        lc_alloc_chunk_t* free_me = thr->chunks;
+        thr->chunks = thr->chunks->next;
+        free(free_me);
+    }
 
+    free((void *) thr);
 }
 
-static ERTS_INLINE erts_lc_locked_locks_t *
+static ERTS_INLINE lc_thread_t *
 get_my_locked_locks(void)
 {
     return erts_tsd_get(locks_key);
 }
 
-static ERTS_INLINE erts_lc_locked_locks_t *
+static ERTS_INLINE lc_thread_t *
 make_my_locked_locks(void)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    if (l_lcks)
-	return l_lcks;
+    lc_thread_t *thr = get_my_locked_locks();
+    if (thr)
+	return thr;
     else
-	return create_locked_locks(NULL);
+	return create_thread_data(NULL);
 }
 
-static ERTS_INLINE erts_lc_locked_lock_t *
-new_locked_lock(erts_lc_lock_t *lck, erts_lock_options_t options,
+static ERTS_INLINE lc_locked_lock_t *
+new_locked_lock(lc_thread_t* thr,
+                erts_lc_lock_t *lck, erts_lock_options_t options,
 		char *file, unsigned int line)
 {
-    erts_lc_locked_lock_t *l_lck = (erts_lc_locked_lock_t *) lc_alloc();
-    l_lck->next = NULL;
-    l_lck->prev = NULL;
-    l_lck->id = lck->id;
-    l_lck->extra = lck->extra;
-    l_lck->file = file;
-    l_lck->line = line;
-    l_lck->flags = lck->flags;
-    l_lck->taken_options = options;
-    return l_lck;
+    lc_locked_lock_t *ll = lc_alloc(thr);
+    ll->next = NULL;
+    ll->prev = NULL;
+    ll->id = lck->id;
+    ll->extra = lck->extra;
+    ll->file = file;
+    ll->line = line;
+    ll->flags = lck->flags;
+    ll->taken_options = options;
+    return ll;
 }
 
 static void
 raw_print_lock(char *prefix, Sint16 id, Wterm extra, erts_lock_flags_t flags,
 	       char* file, unsigned int line, char *suffix)
 {
-    char *lname = (0 <= id && id < ERTS_LOCK_ORDER_SIZE
+    char *lname = (1 <= id && id < ERTS_LOCK_ORDER_SIZE
 		   ? erts_lock_order[id].name
 		   : "unknown");
     erts_fprintf(stderr,"%s'%s:",prefix,lname);
@@ -461,20 +441,20 @@ print_lock(char *prefix, erts_lc_lock_t *lck, char *suffix)
 }
 
 static void
-print_curr_locks(erts_lc_locked_locks_t *l_lcks)
+print_curr_locks(lc_thread_t *thr)
 {
-    erts_lc_locked_lock_t *l_lck;
-    if (!l_lcks || !l_lcks->locked.first)
+    lc_locked_lock_t *ll;
+    if (!thr || !thr->locked.first)
 	erts_fprintf(stderr,
 		     "Currently no locks are locked by the %s thread.\n",
-		     l_lcks->thread_name);
+		     thr->thread_name);
     else {
 	erts_fprintf(stderr,
 		     "Currently these locks are locked by the %s thread:\n",
-		     l_lcks->thread_name);
-	for (l_lck = l_lcks->locked.first; l_lck; l_lck = l_lck->next)
-	  raw_print_lock("  ", l_lck->id, l_lck->extra, l_lck->flags,
-			 l_lck->file, l_lck->line, "\n");
+		     thr->thread_name);
+	for (ll = thr->locked.first; ll; ll = ll->next)
+	  raw_print_lock("  ", ll->id, ll->extra, ll->flags,
+			 ll->file, ll->line, "\n");
     }
 }
 
@@ -503,55 +483,55 @@ uninitialized_lock(void)
 }
 
 static void
-lock_twice(char *prefix, erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck,
+lock_twice(char *prefix, lc_thread_t *thr, erts_lc_lock_t *lck,
 	   erts_lock_options_t options)
 {
     erts_fprintf(stderr, "%s (%s)", prefix, rw_op_str(options));
     print_lock(" ", lck, " lock which is already locked by thread!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-unlock_op_mismatch(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck,
+unlock_op_mismatch(lc_thread_t *thr, erts_lc_lock_t *lck,
 		   erts_lock_options_t options)
 {
     erts_fprintf(stderr, "Unlocking (%s) ", rw_op_str(options));
     print_lock("", lck, " lock which mismatch previous lock operation!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-unlock_of_not_locked(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+unlock_of_not_locked(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Unlocking ", lck, " lock which is not locked by thread!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-lock_order_violation(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+lock_order_violation(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Lock order violation occured when locking ", lck, "!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     print_lock_order();
     lc_abort();
 }
 
 static void
-type_order_violation(char *op, erts_lc_locked_locks_t *l_lcks,
+type_order_violation(char *op, lc_thread_t *thr,
 		     erts_lc_lock_t *lck)
 {
     erts_fprintf(stderr, "Lock type order violation occured when ");
     print_lock(op, lck, "!\n");
-    ASSERT(l_lcks);
-    print_curr_locks(l_lcks);
+    ASSERT(thr);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-lock_mismatch(erts_lc_locked_locks_t *l_lcks, int exact,
+lock_mismatch(lc_thread_t *thr, int exact,
 	      int failed_have, erts_lc_lock_t *have, int have_len,
 	      int failed_have_not, erts_lc_lock_t *have_not, int have_not_len)
 {
@@ -598,39 +578,39 @@ lock_mismatch(erts_lc_locked_locks_t *l_lcks, int exact,
 		print_lock2("  ", have_not[i].id, have_not[i].extra, 0, "\n");
 	}
     }
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-unlock_of_required_lock(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+unlock_of_required_lock(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Unlocking required ", lck, " lock!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-unrequire_of_not_required_lock(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+unrequire_of_not_required_lock(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Unrequire on ", lck, " lock not required!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-require_twice(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+require_twice(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Require on ", lck, " lock already required!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
 static void
-required_not_locked(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
+required_not_locked(lc_thread_t *thr, erts_lc_lock_t *lck)
 {
     print_lock("Required ", lck, " lock not locked!\n");
-    print_curr_locks(l_lcks);
+    print_curr_locks(thr);
     lc_abort();
 }
 
@@ -638,15 +618,15 @@ required_not_locked(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
 static void
 thread_exit_handler(void)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    if (l_lcks) {
-	if (l_lcks->locked.first) {
+    lc_thread_t *thr = get_my_locked_locks();
+    if (thr) {
+	if (thr->locked.first) {
 	    erts_fprintf(stderr,
 			 "Thread exiting while having locked locks!\n");
-	    print_curr_locks(l_lcks);
+	    print_curr_locks(thr);
 	    lc_abort();
 	}
-	destroy_locked_locks(l_lcks);
+	destroy_thread_data(thr);
 	/* erts_tsd_set(locks_key, NULL); */
     }
 }
@@ -664,24 +644,24 @@ lc_abort(void)
 void
 erts_lc_set_thread_name(char *thread_name)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    if (!l_lcks)
-	l_lcks = create_locked_locks(thread_name);
+    lc_thread_t *thr = get_my_locked_locks();
+    if (!thr)
+	thr = create_thread_data(thread_name);
     else {
-	ASSERT(l_lcks->thread_name);
-	free((void *) l_lcks->thread_name);
-	l_lcks->thread_name = strdup(thread_name ? thread_name : "unknown");
-	if (!l_lcks->thread_name)
+	ASSERT(thr->thread_name);
+	free((void *) thr->thread_name);
+	thr->thread_name = strdup(thread_name ? thread_name : "unknown");
+	if (!thr->thread_name)
 	    ERTS_INTERNAL_ERROR("strdup failed");
     }
-    l_lcks->emu_thread = 1;
+    thr->emu_thread = 1;
 }
 
 int
 erts_lc_is_emu_thr(void)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    return l_lcks->emu_thread;
+    lc_thread_t *thr = get_my_locked_locks();
+    return thr->emu_thread;
 }
 
 int
@@ -716,7 +696,7 @@ erts_lc_get_lock_order_id(char *name)
 	erts_fprintf(stderr, "Missing lock name\n");
     else {
 	for (i = 0; i < ERTS_LOCK_ORDER_SIZE; i++)
-	    if (strcmp(erts_lock_order[i].name, name) == 0)
+	    if (sys_strcmp(erts_lock_order[i].name, name) == 0)
 		return i;
 	erts_fprintf(stderr,
 		     "Lock name '%s' missing in lock order "
@@ -727,7 +707,7 @@ erts_lc_get_lock_order_id(char *name)
     return (Sint16) -1;
 }
 
-static int compare_locked_by_id(erts_lc_locked_lock_t *locked_lock, erts_lc_lock_t *comparand)
+static int compare_locked_by_id(lc_locked_lock_t *locked_lock, erts_lc_lock_t *comparand)
 {
     if(locked_lock->id < comparand->id) {
         return -1;
@@ -738,7 +718,7 @@ static int compare_locked_by_id(erts_lc_locked_lock_t *locked_lock, erts_lc_lock
     return 0;
 }
 
-static int compare_locked_by_id_extra(erts_lc_locked_lock_t *locked_lock, erts_lc_lock_t *comparand)
+static int compare_locked_by_id_extra(lc_locked_lock_t *locked_lock, erts_lc_lock_t *comparand)
 {
     int order = compare_locked_by_id(locked_lock, comparand);
 
@@ -753,18 +733,18 @@ static int compare_locked_by_id_extra(erts_lc_locked_lock_t *locked_lock, erts_l
     return 0;
 }
 
-typedef int (*locked_compare_func)(erts_lc_locked_lock_t *, erts_lc_lock_t *);
+typedef int (*locked_compare_func)(lc_locked_lock_t *, erts_lc_lock_t *);
 
 /* Searches through a list of taken locks, bailing when it hits an entry whose
  * order relative to the search template is the opposite of the one at the
  * start of the search. (*closest_neighbor) is either set to the exact match,
  * or the one closest to it in the sort order. */
 static int search_locked_list(locked_compare_func compare,
-                              erts_lc_locked_lock_t *locked_locks,
+                              lc_locked_lock_t *locked_locks,
                               erts_lc_lock_t *search_template,
-                              erts_lc_locked_lock_t **closest_neighbor)
+                              lc_locked_lock_t **closest_neighbor)
 {
-    erts_lc_locked_lock_t *iterator = locked_locks;
+    lc_locked_lock_t *iterator = locked_locks;
 
     (*closest_neighbor) = iterator;
 
@@ -800,9 +780,9 @@ static int search_locked_list(locked_compare_func compare,
 /* Searches for a lock in the given list that matches search_template, and sets
  * (*locked_locks) to the closest lock in the sort order. */
 static int
-find_lock(erts_lc_locked_lock_t **locked_locks, erts_lc_lock_t *search_template)
+find_lock(lc_locked_lock_t **locked_locks, erts_lc_lock_t *search_template)
 {
-    erts_lc_locked_lock_t *closest_neighbor;
+    lc_locked_lock_t *closest_neighbor;
     int found_lock;
 
     found_lock = search_locked_list(compare_locked_by_id_extra,
@@ -831,9 +811,9 @@ find_lock(erts_lc_locked_lock_t **locked_locks, erts_lc_lock_t *search_template)
 /* Searches for a lock in the given list by id, and sets (*locked_locks) to the
  * closest lock in the sort order. */
 static int
-find_id(erts_lc_locked_lock_t **locked_locks, Sint16 id)
+find_id(lc_locked_lock_t **locked_locks, Sint16 id)
 {
-    erts_lc_locked_lock_t *closest_neighbor;
+    lc_locked_lock_t *closest_neighbor;
     erts_lc_lock_t search_template;
     int found_lock;
 
@@ -852,34 +832,34 @@ find_id(erts_lc_locked_lock_t **locked_locks, Sint16 id)
 void
 erts_lc_have_locks(int *resv, erts_lc_lock_t *locks, int len)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
+    lc_thread_t *thr = get_my_locked_locks();
     int i;
 
-    if (!l_lcks) {
+    if (!thr) {
 	for (i = 0; i < len; i++)
 	    resv[i] = 0;
     }
     else {
-	erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
+	lc_locked_lock_t *ll = thr->locked.first;
 	for (i = 0; i < len; i++)
-	    resv[i] = find_lock(&l_lck, &locks[i]);
+	    resv[i] = find_lock(&ll, &locks[i]);
     }
 }
 
 void
 erts_lc_have_lock_ids(int *resv, int *ids, int len)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
+    lc_thread_t *thr = get_my_locked_locks();
     int i;
 
-    if (!l_lcks) {
+    if (!thr) {
 	for (i = 0; i < len; i++)
 	    resv[i] = 0;
     }
     else {
-	erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
+	lc_locked_lock_t *ll = thr->locked.first;
 	for (i = 0; i < len; i++)
-	    resv[i] = find_id(&l_lck, ids[i]);
+	    resv[i] = find_id(&ll, ids[i]);
     }
 }
 
@@ -888,27 +868,27 @@ erts_lc_check(erts_lc_lock_t *have, int have_len,
 	      erts_lc_lock_t *have_not, int have_not_len)
 {
     int i;
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    erts_lc_locked_lock_t *l_lck;
+    lc_thread_t *thr = get_my_locked_locks();
+    lc_locked_lock_t *ll;
     
     if (have && have_len > 0) {
-	if (!l_lcks)
+	if (!thr)
 	    lock_mismatch(NULL, 0,
 			  -1, have, have_len,
 			  -1, have_not, have_not_len);
-	l_lck = l_lcks->locked.first;
+	ll = thr->locked.first;
 	for (i = 0; i < have_len; i++) {
-	    if (!find_lock(&l_lck, &have[i]))
-		lock_mismatch(l_lcks, 0,
+	    if (!find_lock(&ll, &have[i]))
+		lock_mismatch(thr, 0,
 			      i, have, have_len,
 			      -1, have_not, have_not_len);
 	}
     }
-    if (have_not && have_not_len > 0 && l_lcks) {
-	l_lck = l_lcks->locked.first;
+    if (have_not && have_not_len > 0 && thr) {
+	ll = thr->locked.first;
 	for (i = 0; i < have_not_len; i++) {
-	    if (find_lock(&l_lck, &have_not[i]))
-		lock_mismatch(l_lcks, 0,
+	    if (find_lock(&ll, &have_not[i]))
+		lock_mismatch(thr, 0,
 			      -1, have, have_len,
 			      i, have_not, have_not_len);
 	}
@@ -918,8 +898,8 @@ erts_lc_check(erts_lc_lock_t *have, int have_len,
 void
 erts_lc_check_exact(erts_lc_lock_t *have, int have_len)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    if (!l_lcks) {
+    lc_thread_t *thr = get_my_locked_locks();
+    if (!thr) {
 	if (have && have_len > 0)
 	    lock_mismatch(NULL, 1,
 			  -1, have, have_len,
@@ -927,17 +907,17 @@ erts_lc_check_exact(erts_lc_lock_t *have, int have_len)
     }
     else {
 	int i;
-	erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
+	lc_locked_lock_t *ll = thr->locked.first;
 	for (i = 0; i < have_len; i++) {
-	    if (!find_lock(&l_lck, &have[i]))
-		lock_mismatch(l_lcks, 1,
+	    if (!find_lock(&ll, &have[i]))
+		lock_mismatch(thr, 1,
 			      i, have, have_len,
 			      -1, NULL, 0);
 	}
-	for (i = 0, l_lck = l_lcks->locked.first; l_lck; l_lck = l_lck->next)
+	for (i = 0, ll = thr->locked.first; ll; ll = ll->next)
 	    i++;
 	if (i != have_len)
-	    lock_mismatch(l_lcks, 1,
+	    lock_mismatch(thr, 1,
 			  -1, have, have_len,
 			  -1, NULL, 0);
     }
@@ -946,16 +926,16 @@ erts_lc_check_exact(erts_lc_lock_t *have, int have_len)
 void
 erts_lc_check_no_locked_of_type(erts_lock_flags_t type)
 {
-    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
-    if (l_lcks) {
-	erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
-	for (l_lck = l_lcks->locked.first; l_lck; l_lck = l_lck->next) {
-	    if ((l_lck->flags & ERTS_LOCK_FLAGS_MASK_TYPE) == type) {
+    lc_thread_t *thr = get_my_locked_locks();
+    if (thr) {
+	lc_locked_lock_t *ll = thr->locked.first;
+	for (ll = thr->locked.first; ll; ll = ll->next) {
+	    if ((ll->flags & ERTS_LOCK_FLAGS_MASK_TYPE) == type) {
 		erts_fprintf(stderr,
 			     "Locked lock of type %s found which isn't "
 			     "allowed here!\n",
-			     erts_lock_flags_get_type_name(l_lck->flags));
-		print_curr_locks(l_lcks);
+			     erts_lock_flags_get_type_name(ll->flags));
+		print_curr_locks(thr);
 		lc_abort();
 	    }
 	}
@@ -973,7 +953,7 @@ erts_lc_trylock_force_busy_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
      * This in order to make sure that caller can handle
      * the situation without causing a lock order violation.
      */
-    erts_lc_locked_locks_t *l_lcks;
+    lc_thread_t *thr;
 
     if (lck->inited != ERTS_LC_INITITALIZED)
 	uninitialized_lock();
@@ -981,25 +961,25 @@ erts_lc_trylock_force_busy_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
     if (lck->id < 0)
 	return 0;
 
-    l_lcks = get_my_locked_locks();
+    thr = get_my_locked_locks();
 
-    if (!l_lcks || !l_lcks->locked.first) {
-	ASSERT(!l_lcks || !l_lcks->locked.last);
+    if (!thr || !thr->locked.first) {
+	ASSERT(!thr || !thr->locked.last);
 	return 0;
     }
     else {
-	erts_lc_locked_lock_t *tl_lck;
+	lc_locked_lock_t *tl_lck;
 
-	ASSERT(l_lcks->locked.last);
+	ASSERT(thr->locked.last);
 
 #if 0 /* Ok when trylocking I guess... */
-	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, l_lcks->locked.last->flags))
-	    type_order_violation("trylocking ", l_lcks, lck);
+	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, thr->locked.last->flags))
+	    type_order_violation("trylocking ", thr, lck);
 #endif
 
-	if (l_lcks->locked.last->id < lck->id
-	    || (l_lcks->locked.last->id == lck->id
-		&& l_lcks->locked.last->extra < lck->extra))
+	if (thr->locked.last->id < lck->id
+	    || (thr->locked.last->id == lck->id
+		&& thr->locked.last->extra < lck->extra))
 	    return 0;
 
 	/*
@@ -1008,11 +988,11 @@ erts_lc_trylock_force_busy_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
 
 
 	/* Check that we are not trying to lock this lock twice */
-	for (tl_lck = l_lcks->locked.last; tl_lck; tl_lck = tl_lck->prev) {
+	for (tl_lck = thr->locked.last; tl_lck; tl_lck = tl_lck->prev) {
 	    if (tl_lck->id < lck->id
 		|| (tl_lck->id == lck->id && tl_lck->extra <= lck->extra)) {
 		if (tl_lck->id == lck->id && tl_lck->extra == lck->extra)
-		    lock_twice("Trylocking", l_lcks, lck, options);
+		    lock_twice("Trylocking", thr, lck, options);
 		break;
 	    }
 	}
@@ -1037,8 +1017,8 @@ erts_lc_trylock_force_busy_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
 void erts_lc_trylock_flg_x(int locked, erts_lc_lock_t *lck, erts_lock_options_t options,
 			   char *file, unsigned int line)
 {
-    erts_lc_locked_locks_t *l_lcks;
-    erts_lc_locked_lock_t *l_lck;
+    lc_thread_t *thr;
+    lc_locked_lock_t *ll;
 
     if (lck->inited != ERTS_LC_INITITALIZED)
 	uninitialized_lock();
@@ -1046,43 +1026,43 @@ void erts_lc_trylock_flg_x(int locked, erts_lc_lock_t *lck, erts_lock_options_t
     if (lck->id < 0)
 	return;
 
-    l_lcks = make_my_locked_locks();
-    l_lck = locked ? new_locked_lock(lck, options, file, line) : NULL;
+    thr = make_my_locked_locks();
+    ll = locked ? new_locked_lock(thr, lck, options, file, line) : NULL;
 
-    if (!l_lcks->locked.last) {
-	ASSERT(!l_lcks->locked.first);
+    if (!thr->locked.last) {
+	ASSERT(!thr->locked.first);
 	if (locked)
-	    l_lcks->locked.first = l_lcks->locked.last = l_lck;
+	    thr->locked.first = thr->locked.last = ll;
     }
     else {
-	erts_lc_locked_lock_t *tl_lck;
+	lc_locked_lock_t *tl_lck;
 #if 0 /* Ok when trylocking I guess... */
-	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, l_lcks->locked.last->flags))
-	    type_order_violation("trylocking ", l_lcks, lck);
+	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, thr->locked.last->flags))
+	    type_order_violation("trylocking ", thr, lck);
 #endif
 
-	for (tl_lck = l_lcks->locked.last; tl_lck; tl_lck = tl_lck->prev) {
+	for (tl_lck = thr->locked.last; tl_lck; tl_lck = tl_lck->prev) {
 	    if (tl_lck->id < lck->id
 		|| (tl_lck->id == lck->id && tl_lck->extra <= lck->extra)) {
 		if (tl_lck->id == lck->id && tl_lck->extra == lck->extra)
-		    lock_twice("Trylocking", l_lcks, lck, options);
+		    lock_twice("Trylocking", thr, lck, options);
 		if (locked) {
-		    l_lck->next = tl_lck->next;
-		    l_lck->prev = tl_lck;
+		    ll->next = tl_lck->next;
+		    ll->prev = tl_lck;
 		    if (tl_lck->next)
-			tl_lck->next->prev = l_lck;
+			tl_lck->next->prev = ll;
 		    else
-			l_lcks->locked.last = l_lck;
-		    tl_lck->next = l_lck;
+			thr->locked.last = ll;
+		    tl_lck->next = ll;
 		}
 		return;
 	    }
 	}
 
 	if (locked) {
-	    l_lck->next = l_lcks->locked.first;
-	    l_lcks->locked.first->prev = l_lck;
-	    l_lcks->locked.first = l_lck;
+	    ll->next = thr->locked.first;
+	    thr->locked.first->prev = ll;
+	    thr->locked.first = ll;
 	}
     }
 
@@ -1091,83 +1071,83 @@ void erts_lc_trylock_flg_x(int locked, erts_lc_lock_t *lck, erts_lock_options_t
 void erts_lc_require_lock_flg(erts_lc_lock_t *lck, erts_lock_options_t options,
 			      char *file, unsigned int line)
 {
-    erts_lc_locked_locks_t *l_lcks = make_my_locked_locks();
-    erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
-    if (!find_lock(&l_lck, lck))
-	required_not_locked(l_lcks, lck);
-    l_lck = new_locked_lock(lck, options, file, line);
-    if (!l_lcks->required.last) {
-	ASSERT(!l_lcks->required.first);
-	l_lck->next = l_lck->prev = NULL;
-	l_lcks->required.first = l_lcks->required.last = l_lck;
+    lc_thread_t *thr = make_my_locked_locks();
+    lc_locked_lock_t *ll = thr->locked.first;
+    if (!find_lock(&ll, lck))
+	required_not_locked(thr, lck);
+    ll = new_locked_lock(thr, lck, options, file, line);
+    if (!thr->required.last) {
+	ASSERT(!thr->required.first);
+	ll->next = ll->prev = NULL;
+	thr->required.first = thr->required.last = ll;
     }
     else {
-	erts_lc_locked_lock_t *l_lck2;
-	ASSERT(l_lcks->required.first);
-	for (l_lck2 = l_lcks->required.last;
+	lc_locked_lock_t *l_lck2;
+	ASSERT(thr->required.first);
+	for (l_lck2 = thr->required.last;
 	     l_lck2;
 	     l_lck2 = l_lck2->prev) {
 	    if (l_lck2->id < lck->id
 		|| (l_lck2->id == lck->id && l_lck2->extra < lck->extra))
 		break;
 	    else if (l_lck2->id == lck->id && l_lck2->extra == lck->extra)
-		require_twice(l_lcks, lck);
+		require_twice(thr, lck);
 	}
 	if (!l_lck2) {
-	    l_lck->next = l_lcks->required.first;
-	    l_lck->prev = NULL;
-	    l_lcks->required.first->prev = l_lck;
-	    l_lcks->required.first = l_lck;
+	    ll->next = thr->required.first;
+	    ll->prev = NULL;
+	    thr->required.first->prev = ll;
+	    thr->required.first = ll;
 	}
 	else {
-	    l_lck->next = l_lck2->next;
-	    if (l_lck->next) {
-		ASSERT(l_lcks->required.last != l_lck2);
-		l_lck->next->prev = l_lck;
+	    ll->next = l_lck2->next;
+	    if (ll->next) {
+		ASSERT(thr->required.last != l_lck2);
+		ll->next->prev = ll;
 	    }
 	    else {
-		ASSERT(l_lcks->required.last == l_lck2);
-		l_lcks->required.last = l_lck;
+		ASSERT(thr->required.last == l_lck2);
+		thr->required.last = ll;
 	    }
-	    l_lck->prev = l_lck2;
-	    l_lck2->next = l_lck;		
+	    ll->prev = l_lck2;
+	    l_lck2->next = ll;
 	}
     }
 }
 
 void erts_lc_unrequire_lock_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
 {
-    erts_lc_locked_locks_t *l_lcks = make_my_locked_locks();
-    erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
-    if (!find_lock(&l_lck, lck))
-	required_not_locked(l_lcks, lck);
-    l_lck = l_lcks->required.first;
-    if (!find_lock(&l_lck, lck))
-	unrequire_of_not_required_lock(l_lcks, lck);
-    if (l_lck->prev) {
-	ASSERT(l_lcks->required.first != l_lck);
-	l_lck->prev->next = l_lck->next;
+    lc_thread_t *thr = make_my_locked_locks();
+    lc_locked_lock_t *ll = thr->locked.first;
+    if (!find_lock(&ll, lck))
+	required_not_locked(thr, lck);
+    ll = thr->required.first;
+    if (!find_lock(&ll, lck))
+	unrequire_of_not_required_lock(thr, lck);
+    if (ll->prev) {
+	ASSERT(thr->required.first != ll);
+	ll->prev->next = ll->next;
     }
     else {
-	ASSERT(l_lcks->required.first == l_lck);
-	l_lcks->required.first = l_lck->next;
+	ASSERT(thr->required.first == ll);
+	thr->required.first = ll->next;
     }
-    if (l_lck->next) {
-	ASSERT(l_lcks->required.last != l_lck);
-	l_lck->next->prev = l_lck->prev;
+    if (ll->next) {
+	ASSERT(thr->required.last != ll);
+	ll->next->prev = ll->prev;
     }
     else {
-	ASSERT(l_lcks->required.last == l_lck);
-	l_lcks->required.last = l_lck->prev;
+	ASSERT(thr->required.last == ll);
+	thr->required.last = ll->prev;
     }
-    lc_free((void *) l_lck);
+    lc_free(thr, ll);
 }
 
 void erts_lc_lock_flg_x(erts_lc_lock_t *lck, erts_lock_options_t options,
 			char *file, unsigned int line)
 {
-    erts_lc_locked_locks_t *l_lcks;
-    erts_lc_locked_lock_t *l_lck;
+    lc_thread_t *thr;
+    lc_locked_lock_t *new_ll;
 
     if (lck->inited != ERTS_LC_INITITALIZED)
 	uninitialized_lock();
@@ -1175,32 +1155,45 @@ void erts_lc_lock_flg_x(erts_lc_lock_t *lck, erts_lock_options_t options,
     if (lck->id < 0)
 	return;
 
-    l_lcks = make_my_locked_locks();
-    l_lck = new_locked_lock(lck, options, file, line);
+    thr = make_my_locked_locks();
+    new_ll = new_locked_lock(thr, lck, options, file, line);
 
-    if (!l_lcks->locked.last) {
-	ASSERT(!l_lcks->locked.first);
-	l_lcks->locked.last = l_lcks->locked.first = l_lck;
+    if (!thr->locked.last) {
+	ASSERT(!thr->locked.first);
+	thr->locked.last = thr->locked.first = new_ll;
+        ASSERT(0 < lck->id && lck->id < ERTS_LOCK_ORDER_SIZE);
+        thr->matrix.m[lck->id][0] = 1;
     }
-    else if (l_lcks->locked.last->id < lck->id
-	     || (l_lcks->locked.last->id == lck->id
-		 && l_lcks->locked.last->extra < lck->extra)) {
-	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, l_lcks->locked.last->flags))
-	    type_order_violation("locking ", l_lcks, lck);
-	l_lck->prev = l_lcks->locked.last;
-	l_lcks->locked.last->next = l_lck;
-	l_lcks->locked.last = l_lck;
+    else if (thr->locked.last->id < lck->id
+	     || (thr->locked.last->id == lck->id
+		 && thr->locked.last->extra < lck->extra)) {
+        lc_locked_lock_t* ll;
+	if (LOCK_IS_TYPE_ORDER_VIOLATION(lck->flags, thr->locked.last->flags)) {
+	    type_order_violation("locking ", thr, lck);
+        }
+
+        ASSERT(0 < lck->id && lck->id < ERTS_LOCK_ORDER_SIZE);
+        ll = thr->locked.last;
+        thr->matrix.m[lck->id][ll->id] |= 1;
+        for (ll = ll->prev; ll; ll = ll->prev) {
+            ASSERT(0 < ll->id && ll->id < ERTS_LOCK_ORDER_SIZE);
+            thr->matrix.m[lck->id][ll->id] |= 2;
+        }
+
+	new_ll->prev = thr->locked.last;
+	thr->locked.last->next = new_ll;
+	thr->locked.last = new_ll;
     }
-    else if (l_lcks->locked.last->id == lck->id && l_lcks->locked.last->extra == lck->extra)
-	lock_twice("Locking", l_lcks, lck, options);
+    else if (thr->locked.last->id == lck->id && thr->locked.last->extra == lck->extra)
+	lock_twice("Locking", thr, lck, options);
     else
-	lock_order_violation(l_lcks, lck);
+	lock_order_violation(thr, lck);
 }
 
 void erts_lc_unlock_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
 {
-    erts_lc_locked_locks_t *l_lcks;
-    erts_lc_locked_lock_t *l_lck;
+    lc_thread_t *thr;
+    lc_locked_lock_t *ll;
 
     if (lck->inited != ERTS_LC_INITITALIZED)
 	uninitialized_lock();
@@ -1208,38 +1201,38 @@ void erts_lc_unlock_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
     if (lck->id < 0)
 	return;
 
-    l_lcks = get_my_locked_locks();
+    thr = get_my_locked_locks();
 
-    if (l_lcks) {
-	l_lck = l_lcks->required.first;
-	if (find_lock(&l_lck, lck))
-	    unlock_of_required_lock(l_lcks, lck);
+    if (thr) {
+	ll = thr->required.first;
+	if (find_lock(&ll, lck))
+	    unlock_of_required_lock(thr, lck);
     }
 
-    for (l_lck = l_lcks ? l_lcks->locked.last : NULL; l_lck; l_lck = l_lck->prev) {
-	if (l_lck->id == lck->id && l_lck->extra == lck->extra) {
-	    if ((l_lck->taken_options & ERTS_LOCK_OPTIONS_RDWR) != options)
-		unlock_op_mismatch(l_lcks, lck, options);
-	    if (l_lck->prev)
-		l_lck->prev->next = l_lck->next;
+    for (ll = thr ? thr->locked.last : NULL; ll; ll = ll->prev) {
+	if (ll->id == lck->id && ll->extra == lck->extra) {
+	    if ((ll->taken_options & ERTS_LOCK_OPTIONS_RDWR) != options)
+		unlock_op_mismatch(thr, lck, options);
+	    if (ll->prev)
+		ll->prev->next = ll->next;
 	    else
-		l_lcks->locked.first = l_lck->next;
-	    if (l_lck->next)
-		l_lck->next->prev = l_lck->prev;
+		thr->locked.first = ll->next;
+	    if (ll->next)
+		ll->next->prev = ll->prev;
 	    else
-		l_lcks->locked.last = l_lck->prev;
-	    lc_free((void *) l_lck);
+		thr->locked.last = ll->prev;
+	    lc_free(thr, ll);
 	    return;
 	}
     }
     
-    unlock_of_not_locked(l_lcks, lck);
+    unlock_of_not_locked(thr, lck);
 }
 
 void erts_lc_might_unlock_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
 {
-    erts_lc_locked_locks_t *l_lcks;
-    erts_lc_locked_lock_t *l_lck;
+    lc_thread_t *thr;
+    lc_locked_lock_t *ll;
 
     if (lck->inited != ERTS_LC_INITITALIZED)
 	uninitialized_lock();
@@ -1247,17 +1240,17 @@ void erts_lc_might_unlock_flg(erts_lc_lock_t *lck, erts_lock_options_t options)
     if (lck->id < 0)
 	return;
 
-    l_lcks = get_my_locked_locks();
+    thr = get_my_locked_locks();
 
-    if (l_lcks) {
-	l_lck = l_lcks->required.first;
-	if (find_lock(&l_lck, lck))
-	    unlock_of_required_lock(l_lcks, lck);
+    if (thr) {
+	ll = thr->required.first;
+	if (find_lock(&ll, lck))
+	    unlock_of_required_lock(thr, lck);
     }
 
-    l_lck = l_lcks->locked.first;
-    if (!find_lock(&l_lck, lck))
-	unlock_of_not_locked(l_lcks, lck);
+    ll = thr->locked.first;
+    if (!find_lock(&ll, lck))
+	unlock_of_not_locked(thr, lck);
 }
 
 int
@@ -1338,26 +1331,7 @@ erts_lc_destroy_lock(erts_lc_lock_t *lck)
 void
 erts_lc_init(void)
 {
-#ifdef ERTS_LC_STATIC_ALLOC
-    int i;
-    static erts_lc_free_block_t fbs[ERTS_LC_FB_CHUNK_SIZE];
-    for (i = 0; i < ERTS_LC_FB_CHUNK_SIZE - 1; i++) {
-#ifdef DEBUG
-	memset((void *) &fbs[i], 0xdf, sizeof(erts_lc_free_block_t));
-#endif
-	fbs[i].next = &fbs[i+1];
-    }
-#ifdef DEBUG
-    memset((void *) &fbs[ERTS_LC_FB_CHUNK_SIZE-1],
-	   0xdf, sizeof(erts_lc_free_block_t));
-#endif
-    fbs[ERTS_LC_FB_CHUNK_SIZE-1].next = NULL;
-    free_blocks = &fbs[0]; 
-#else /* #ifdef ERTS_LC_STATIC_ALLOC */
-    free_blocks = NULL;
-#endif /* #ifdef ERTS_LC_STATIC_ALLOC */
-
-    if (ethr_spinlock_init(&free_blocks_lock) != 0)
+    if (ethr_spinlock_init(&lc_threads_lock) != 0)
 	ERTS_INTERNAL_ERROR("spinlock_init failed");
 
     erts_tsd_key_create(&locks_key,"erts_lock_check_key");
@@ -1379,5 +1353,76 @@ erts_lc_pll(void)
     print_curr_locks(get_my_locked_locks());
 }
 
+static void collect_matrix(lc_matrix_t* matrix)
+{
+    int i, j;
+    for (i = 1; i < ERTS_LOCK_ORDER_SIZE; i++) {
+        for (j = 0; j <= i; j++) {
+            tot_lc_matrix.m[i][j] |= matrix->m[i][j];
+        }
+#ifdef DEBUG
+        for ( ; j < ERTS_LOCK_ORDER_SIZE; j++) {
+            ASSERT(matrix->m[i][j] == 0);
+        }
+#endif
+    }
+}
+
+Eterm
+erts_lc_dump_graph(void)
+{
+    const char* basename = "lc_graph.";
+    char filename[40];
+    lc_matrix_t* tot = &tot_lc_matrix;
+    lc_thread_t* thr;
+    int i, j, name_max = 0;
+    FILE* ff;
+
+    lc_lock_threads();
+    for (thr = lc_threads; thr; thr = thr->next) {
+        collect_matrix(&thr->matrix);
+    }
+    lc_unlock_threads();
+
+    sys_strcpy(filename, basename);
+    sys_get_pid(filename + strlen(basename),
+                sizeof(filename) - strlen(basename));
+    ff = fopen(filename, "w");
+    if (!ff)
+        return am_error;
+
+    for (i = 1; i < ERTS_LOCK_ORDER_SIZE; i++) {
+        int len = strlen(erts_lock_order[i].name);
+        if (name_max < len)
+            name_max = len;
+    }
+    fputs("%This file was generated by erts_debug:lc_graph()\n\n", ff);
+    fputs("%{ThisLockName, ThisLockId, LockedDirectlyBeforeThis, LockedIndirectlyBeforeThis}\n", ff);
+    fprintf(ff, "[{%*s, %2d}", name_max, "\"NO LOCK\"", 0);
+    for (i = 1; i < ERTS_LOCK_ORDER_SIZE; i++) {
+        char* delim = "";
+        fprintf(ff, ",\n {%*s, %2d, [", name_max, erts_lock_order[i].name, i);
+        for (j = 0; j < ERTS_LOCK_ORDER_SIZE; j++) {
+            if (tot->m[i][j] & 1) {
+                fprintf(ff, "%s%d", delim, j);
+                delim = ",";
+            }
+        }
+        fprintf(ff, "], [");
+        delim = "";
+        for (j = 0; j < ERTS_LOCK_ORDER_SIZE; j++) {
+            if (tot->m[i][j] == 2) {
+                fprintf(ff, "%s%d", delim, j);
+                delim = ",";
+            }
+        }
+        fputs("]}", ff);
+    }
+    fputs("].", ff);
+    fclose(ff);
+    erts_fprintf(stderr, "Created file '%s' in current working directory\n",
+                 filename);
+    return am_ok;
+}
 
 #endif /* #ifdef ERTS_ENABLE_LOCK_CHECK */
diff --git a/erts/emulator/beam/erl_lock_check.h b/erts/emulator/beam/erl_lock_check.h
index 8c754a8dfa..138bc810bd 100644
--- a/erts/emulator/beam/erl_lock_check.h
+++ b/erts/emulator/beam/erl_lock_check.h
@@ -94,15 +94,11 @@ void erts_lc_unrequire_lock(erts_lc_lock_t *lck);
 
 int erts_lc_is_emu_thr(void);
 
+Eterm erts_lc_dump_graph(void);
+
 #define ERTS_LC_ASSERT(A) \
     ((void) (((A) || ERTS_SOMEONE_IS_CRASH_DUMPING) ? 1 : erts_lc_assert_failed(__FILE__, __LINE__, #A)))
-#ifdef ERTS_SMP
-#define ERTS_SMP_LC_ASSERT(A) ERTS_LC_ASSERT(A)
-#else
-#define ERTS_SMP_LC_ASSERT(A) ((void) 1)
-#endif
 #else /* #ifdef ERTS_ENABLE_LOCK_CHECK */
-#define ERTS_SMP_LC_ASSERT(A) ((void) 1)
 #define ERTS_LC_ASSERT(A) ((void) 1)
 #endif /* #ifdef ERTS_ENABLE_LOCK_CHECK */
 
diff --git a/erts/emulator/beam/erl_lock_count.c b/erts/emulator/beam/erl_lock_count.c
index d2e8f47d59..1ae6076b12 100644
--- a/erts/emulator/beam/erl_lock_count.c
+++ b/erts/emulator/beam/erl_lock_count.c
@@ -554,16 +554,6 @@ erts_lock_flags_t erts_lcnt_get_category_mask() {
     return lcnt_category_mask__;
 }
 
-#ifdef ERTS_ENABLE_KERNEL_POLL
-/* erl_poll/erl_check_io only exports one of these variants at a time, and we
- * may need to use either one depending on emulator startup flags. */
-void erts_lcnt_update_pollset_locks_nkp(int);
-void erts_lcnt_update_pollset_locks_kp(int);
-
-void erts_lcnt_update_cio_locks_nkp(int);
-void erts_lcnt_update_cio_locks_kp(int);
-#endif
-
 void erts_lcnt_set_category_mask(erts_lock_flags_t mask) {
     erts_lock_flags_t changed_categories;
 
@@ -590,19 +580,7 @@ void erts_lcnt_set_category_mask(erts_lock_flags_t mask) {
     }
 
     if(changed_categories & ERTS_LOCK_FLAGS_CATEGORY_IO) {
-#ifdef ERTS_ENABLE_KERNEL_POLL
-        if(erts_use_kernel_poll) {
-            erts_lcnt_update_pollset_locks_kp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
-            erts_lcnt_update_cio_locks_kp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
-        } else {
-            erts_lcnt_update_pollset_locks_nkp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
-            erts_lcnt_update_cio_locks_nkp(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
-        }
-#else
-        erts_lcnt_update_pollset_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
         erts_lcnt_update_cio_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
-#endif
-
         erts_lcnt_update_driver_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
         erts_lcnt_update_port_locks(mask & ERTS_LOCK_FLAGS_CATEGORY_IO);
     }
diff --git a/erts/emulator/beam/erl_map.c b/erts/emulator/beam/erl_map.c
index f0c54e05f7..48154b5d0f 100644
--- a/erts/emulator/beam/erl_map.c
+++ b/erts/emulator/beam/erl_map.c
@@ -43,7 +43,9 @@
  *
  * DONE:
  * - erlang:is_map/1
+ * - erlang:is_map_key/2
  * - erlang:map_size/1
+ * - erlang:map_get/2
  *
  * - maps:find/2
  * - maps:from_list/1
@@ -91,7 +93,6 @@ static BIF_RETTYPE hashmap_merge(Process *p, Eterm nodeA, Eterm nodeB, int swap_
 static Export hashmap_merge_trap_export;
 static BIF_RETTYPE maps_merge_trap_1(BIF_ALIST_1);
 static Uint hashmap_subtree_size(Eterm node);
-static Eterm hashmap_to_list(Process *p, Eterm map, Sint n);
 static Eterm hashmap_keys(Process *p, Eterm map);
 static Eterm hashmap_values(Process *p, Eterm map);
 static Eterm hashmap_delete(Process *p, Uint32 hx, Eterm key, Eterm node, Eterm *value);
@@ -139,80 +140,6 @@ BIF_RETTYPE map_size_1(BIF_ALIST_1) {
     BIF_ERROR(BIF_P, BADMAP);
 }
 
-/* maps:to_list/1 */
-
-BIF_RETTYPE maps_to_list_1(BIF_ALIST_1) {
-    if (is_flatmap(BIF_ARG_1)) {
-	Uint n;
-	Eterm* hp;
-	Eterm *ks,*vs, res, tup;
-	flatmap_t *mp = (flatmap_t*)flatmap_val(BIF_ARG_1);
-
-	ks  = flatmap_get_keys(mp);
-	vs  = flatmap_get_values(mp);
-	n   = flatmap_get_size(mp);
-	hp  = HAlloc(BIF_P, (2 + 3) * n);
-	res = NIL;
-
-	while(n--) {
-	    tup = TUPLE2(hp, ks[n], vs[n]); hp += 3;
-	    res = CONS(hp, tup, res); hp += 2;
-	}
-
-	BIF_RET(res);
-    } else if (is_hashmap(BIF_ARG_1)) {
-	return hashmap_to_list(BIF_P, BIF_ARG_1, -1);
-    }
-
-    BIF_P->fvalue = BIF_ARG_1;
-    BIF_ERROR(BIF_P, BADMAP);
-}
-
-/* erts_internal:maps_to_list/2
- *
- * This function should be removed once iterators are in place.
- * Never document it.
- * Never encourage its usage.
- *
- * A negative value in ARG 2 means the entire map.
- */
-
-BIF_RETTYPE erts_internal_maps_to_list_2(BIF_ALIST_2) {
-    Sint m;
-    if (term_to_Sint(BIF_ARG_2, &m)) {
-        if (is_flatmap(BIF_ARG_1)) {
-            Uint n;
-            Eterm* hp;
-            Eterm *ks,*vs, res, tup;
-            flatmap_t *mp = (flatmap_t*)flatmap_val(BIF_ARG_1);
-
-            ks  = flatmap_get_keys(mp);
-            vs  = flatmap_get_values(mp);
-            n   = flatmap_get_size(mp);
-
-            if (m >= 0) {
-                n = m < n ? m : n;
-            }
-
-            hp  = HAlloc(BIF_P, (2 + 3) * n);
-            res = NIL;
-
-            while(n--) {
-                tup = TUPLE2(hp, ks[n], vs[n]); hp += 3;
-                res = CONS(hp, tup, res); hp += 2;
-            }
-
-            BIF_RET(res);
-        } else if (is_hashmap(BIF_ARG_1)) {
-            return hashmap_to_list(BIF_P, BIF_ARG_1, m);
-        }
-        BIF_P->fvalue = BIF_ARG_1;
-        BIF_ERROR(BIF_P, BADMAP);
-    }
-    BIF_ERROR(BIF_P, BADARG);
-}
-
-
 /* maps:find/2
  * return value if key *matches* a key in the map
  */
@@ -277,7 +204,7 @@ BIF_RETTYPE maps_find_2(BIF_ALIST_2) {
     BIF_ERROR(BIF_P, BADMAP);
 }
 
-/* maps:get/2
+/* maps:get/2 and erlang:map_get/2
  * return value if key *matches* a key in the map
  * exception badkey if none matches
  */
@@ -298,6 +225,10 @@ BIF_RETTYPE maps_get_2(BIF_ALIST_2) {
     BIF_ERROR(BIF_P, BADMAP);
 }
 
+BIF_RETTYPE map_get_2(BIF_ALIST_2) {
+    BIF_RET(maps_get_2(BIF_CALL_ARGS));
+}
+
 /* maps:from_list/1
  * List may be unsorted [{K,V}]
  */
@@ -565,7 +496,9 @@ Eterm erts_map_from_ks_and_vs(ErtsHeapFactory *factory, Eterm *ks0, Eterm *vs0,
         sys_memcpy(ks, ks0, n * sizeof(Eterm));
         sys_memcpy(vs, vs0, n * sizeof(Eterm));
 
-        erts_validate_and_sort_flatmap(mp);
+        if (!erts_validate_and_sort_flatmap(mp)) {
+            return THE_NON_VALUE;
+        }
 
         return make_flatmap(mp);
     } else {
@@ -987,7 +920,7 @@ static int hxnodecmp(hxnode_t *a, hxnode_t *b) {
 	return -1;
 }
 
-/* maps:is_key/2 */
+/* maps:is_key/2 and erlang:is_map_key/2 */
 
 BIF_RETTYPE maps_is_key_2(BIF_ALIST_2) {
     if (is_map(BIF_ARG_2)) {
@@ -997,6 +930,10 @@ BIF_RETTYPE maps_is_key_2(BIF_ALIST_2) {
     BIF_ERROR(BIF_P, BADMAP);
 }
 
+BIF_RETTYPE is_map_key_2(BIF_ALIST_2) {
+    BIF_RET(maps_is_key_2(BIF_CALL_ARGS));
+}
+
 /* maps:keys/1 */
 
 BIF_RETTYPE maps_keys_1(BIF_ALIST_1) {
@@ -1962,45 +1899,31 @@ BIF_RETTYPE maps_values_1(BIF_ALIST_1) {
     BIF_ERROR(BIF_P, BADMAP);
 }
 
-static Eterm hashmap_to_list(Process *p, Eterm node, Sint m) {
-    DECLARE_WSTACK(stack);
-    Eterm *hp, *kv;
-    Eterm tup, res = NIL;
-    Uint n = hashmap_size(node);
-
-    if (m >= 0) {
-        n = m < n ? m : n;
-    }
-
-    hp  = HAlloc(p, n * (2 + 3));
-    hashmap_iterator_init(&stack, node, 0);
-    while (n--) {
-        kv  = hashmap_iterator_next(&stack);
-        ASSERT(kv != NULL);
-	tup = TUPLE2(hp, CAR(kv), CDR(kv));
-	hp += 3;
-	res = CONS(hp, tup, res);
-	hp += 2;
-    }
-    DESTROY_WSTACK(stack);
-    return res;
-}
-
-void hashmap_iterator_init(ErtsWStack* s, Eterm node, int reverse) {
-    Eterm hdr = *hashmap_val(node);
+static ERTS_INLINE
+Uint hashmap_node_size(Eterm hdr, Eterm **nodep)
+{
     Uint sz;
 
     switch(hdr & _HEADER_MAP_SUBTAG_MASK) {
     case HAMT_SUBTAG_HEAD_ARRAY:
 	sz = 16;
+        if (nodep) ++*nodep;
 	break;
     case HAMT_SUBTAG_HEAD_BITMAP:
+        if (nodep) ++*nodep;
     case HAMT_SUBTAG_NODE_BITMAP:
         sz = hashmap_bitcount(MAP_HEADER_VAL(hdr));
+        ASSERT(sz < 17);
 	break;
     default:
 	erts_exit(ERTS_ABORT_EXIT, "bad header");
     }
+    return sz;
+}
+
+void hashmap_iterator_init(ErtsWStack* s, Eterm node, int reverse) {
+    Eterm hdr = *hashmap_val(node);
+    Uint sz = hashmap_node_size(hdr, NULL);
 
     WSTACK_PUSH3((*s), (UWord)THE_NON_VALUE,  /* end marker */
 		 (UWord)(!reverse ? 0 : sz+1),
@@ -2024,20 +1947,7 @@ Eterm* hashmap_iterator_next(ErtsWStack* s) {
 	    ptr = boxed_val(node);
 	    hdr = *ptr;
 	    ASSERT(is_header(hdr));
-	    switch(hdr & _HEADER_MAP_SUBTAG_MASK) {
-	    case HAMT_SUBTAG_HEAD_ARRAY:
-		ptr++;
-		sz = 16;
-		break;
-	    case HAMT_SUBTAG_HEAD_BITMAP:
-		ptr++;
-	    case HAMT_SUBTAG_NODE_BITMAP:
-		sz = hashmap_bitcount(MAP_HEADER_VAL(hdr));
-		ASSERT(sz < 17);
-		break;
-	    default:
-		erts_exit(ERTS_ABORT_EXIT, "bad header");
-	    }
+            sz = hashmap_node_size(hdr, &ptr);
 
 	    idx++;
 
@@ -2074,20 +1984,7 @@ Eterm* hashmap_iterator_prev(ErtsWStack* s) {
 	    ptr = boxed_val(node);
 	    hdr = *ptr;
 	    ASSERT(is_header(hdr));
-	    switch(hdr & _HEADER_MAP_SUBTAG_MASK) {
-	    case HAMT_SUBTAG_HEAD_ARRAY:
-		ptr++;
-		sz = 16;
-		break;
-	    case HAMT_SUBTAG_HEAD_BITMAP:
-		ptr++;
-	    case HAMT_SUBTAG_NODE_BITMAP:
-		sz = hashmap_bitcount(MAP_HEADER_VAL(hdr));
-		ASSERT(sz < 17);
-		break;
-	    default:
-		erts_exit(ERTS_ERROR_EXIT, "bad header");
-	    }
+            sz = hashmap_node_size(hdr, &ptr);
 
             if (idx > sz)
 		idx = sz;
@@ -3061,6 +2958,379 @@ static Eterm hashmap_bld_tuple_uint(Uint **hpp, Uint *szp, Uint n, Uint nums[])
 }
 
 
+/**
+ * In hashmap the Path is a bit pattern that describes
+ * which slot we should traverse in each hashmap node.
+ * Since each hashmap node can only be up to 16 elements
+ * large we use 4 bits per level in the path.
+ *
+ * So a Path with value 0x110 will first get the 0:th
+ * slot in the head node, and then the 1:st slot in the
+ * resulting node and then finally the 1:st slot in the
+ * node beneath. If that slot is not a leaf, then the path
+ * continues down the 0:th slot until it finds a leaf.
+ *
+ * Once the leaf has been found, the return value is created
+ * by traversing the tree using the the stack that was built
+ * when searching for the first leaf to return.
+ *
+ * The index can become a bignum, which complicates the code
+ * a bit. However it should be very rare that this happens
+ * even on a 32bit system as you would need a tree of depth
+ * 7 or more.
+ *
+ * If the number of elements remaining in the map is greater
+ * than how many we want to return, we build a new Path, using
+ * the stack, that points to the next leaf.
+ *
+ * The third argument to this function controls how the data
+ * is returned.
+ *
+ * iterator: The key-value associations are to be used by
+ *           maps:iterator. The return has this format:
+ *             {K1,V1,{K2,V2,none | [Path | Map]}}
+ *           this makes the maps:next function very simple
+ *           and performant.
+ *
+ * list(): The key-value associations are to be used by
+ *         maps:to_list. The return has this format:
+ *             [Path, Map | [{K1,V1},{K2,V2} | BIF_ARG_3]]
+ *                 or if no more associations remain
+ *             [{K1,V1},{K2,V2} | BIF_ARG_3]
+ */
+
+#define PATH_ELEM_SIZE 4
+#define PATH_ELEM_MASK 0xf
+#define PATH_ELEM(PATH) ((PATH) & PATH_ELEM_MASK)
+#define PATH_ELEMS_PER_DIGIT (sizeof(ErtsDigit) * 8 / PATH_ELEM_SIZE)
+
+BIF_RETTYPE erts_internal_map_next_3(BIF_ALIST_3) {
+
+    Eterm path, map;
+    enum { iterator, list } type;
+
+    path = BIF_ARG_1;
+    map  = BIF_ARG_2;
+
+    if (!is_map(map))
+        BIF_ERROR(BIF_P, BADARG);
+
+    if (BIF_ARG_3 == am_iterator) {
+        type = iterator;
+    } else if (is_nil(BIF_ARG_3) || is_list(BIF_ARG_3)) {
+        type = list;
+    } else {
+        BIF_ERROR(BIF_P, BADARG);
+    }
+
+    if (is_flatmap(map)) {
+        Uint n;
+	Eterm *ks,*vs, res, *hp;
+	flatmap_t *mp = (flatmap_t*)flatmap_val(map);
+
+	ks  = flatmap_get_keys(mp);
+	vs  = flatmap_get_values(mp);
+	n   = flatmap_get_size(mp);
+
+        if (!is_small(BIF_ARG_1) || n < unsigned_val(BIF_ARG_1))
+            BIF_ERROR(BIF_P, BADARG);
+
+        if (type == iterator) {
+            hp  = HAlloc(BIF_P, 4 * n);
+            res = am_none;
+
+            while(n--) {
+                res = TUPLE3(hp, ks[n], vs[n], res); hp += 4;
+            }
+        } else {
+            hp  = HAlloc(BIF_P, (2 + 3) * n);
+            res = BIF_ARG_3;
+
+            while(n--) {
+                Eterm tup = TUPLE2(hp, ks[n], vs[n]); hp += 3;
+                res = CONS(hp, tup, res); hp += 2;
+            }
+        }
+
+	BIF_RET(res);
+    } else {
+        Uint curr_path;
+        Uint path_length = 0;
+        Uint *path_rest = NULL;
+        int i, elems, orig_elems;
+        Eterm node = map, res, *patch_ptr = NULL, *hp;
+
+        /* A stack WSTACK is used when traversing the hashmap.
+         * It contains: node, idx, sz, ptr
+         *
+         * `node` is not really needed, but it is very nice to
+         * have when debugging.
+         *
+         * `idx` always points to the next un-explored entry in
+         * a node. If there are no more un-explored entries,
+         * `idx` is equal to `sz`.
+         *
+         * `sz` is the number of elements in the node.
+         *
+         * `ptr` is a pointer to where the elements of the node begins.
+         */
+        DECLARE_WSTACK(stack);
+
+        ASSERT(is_hashmap(node));
+
+/* How many elements we return in one call depends on the number of reductions
+ * that the process has left to run. In debug we return fewer elements to test
+ * the Path implementation better.
+ *
+ * Also, when the path is 0 (i.e. for the first call) we limit the number of
+ * elements to MAP_SMALL_MAP_LIMIT in order to not use a huge amount of heap
+ * when only the first X associations in the hashmap was needed.
+ */
+#if defined(DEBUG)
+#define FCALLS_ELEMS(BIF_P) ((BIF_P->fcalls / 4) & 0xF)
+#else
+#define FCALLS_ELEMS(BIF_P) (BIF_P->fcalls / 4)
+#endif
+
+        if (MAX(FCALLS_ELEMS(BIF_P), 1) < hashmap_size(map))
+            elems = MAX(FCALLS_ELEMS(BIF_P), 1);
+        else
+            elems = hashmap_size(map);
+
+#undef FCALLS_ELEMS
+
+        if (is_small(path)) {
+            curr_path = unsigned_val(path);
+
+            if (curr_path == 0 && elems > MAP_SMALL_MAP_LIMIT) {
+                elems = MAP_SMALL_MAP_LIMIT;
+            }
+        } else if (is_big(path)) {
+            Eterm *big = big_val(path);
+            if (bignum_header_is_neg(*big))
+                BIF_ERROR(BIF_P, BADARG);
+            path_length = BIG_ARITY(big) - 1;
+            curr_path = BIG_DIGIT(big,  0);
+            path_rest = BIG_V(big) + 1;
+        } else {
+            BIF_ERROR(BIF_P, BADARG);
+        }
+
+        if (type == iterator) {
+            /*
+             * Iterator uses the format {K1, V1, {K2, V2, {K3, V3, [Path | Map]}}},
+             * so each element is 4 words large.
+             * To make iteration order independent of input reductions
+             * the KV-pairs are here built in DESTRUCTIVE non-reverse order.
+             */
+            hp = HAlloc(BIF_P, 4 * elems);
+        } else {
+            /*
+             * List used the format [Path, Map, {K3,V3}, {K2,V2}, {K1,V1} | BIF_ARG_3],
+             * so each element is 2+3 words large.
+             * To make list order independent of input reductions
+             * the KV-pairs are here built in FUNCTIONAL reverse order
+             * as this is how the list as a whole is constructed.
+             */
+            hp = HAlloc(BIF_P, (2 + 3) * elems);
+        }
+
+        orig_elems = elems;
+
+        /* First we look for the leaf to start at using the
+           path given. While doing so, we push each map node
+           and the index onto the stack to use later. */
+        for (i = 1; ; i++) {
+            Eterm *ptr = hashmap_val(node),
+                hdr = *ptr++;
+            Uint sz;
+
+            sz = hashmap_node_size(hdr, &ptr);
+
+            if (PATH_ELEM(curr_path) >= sz)
+                goto badarg;
+
+            WSTACK_PUSH4(stack, node, PATH_ELEM(curr_path)+1, sz, (UWord)ptr);
+
+            /* We have found a leaf, return it and the next X elements */
+            if (is_list(ptr[PATH_ELEM(curr_path)])) {
+                Eterm *lst = list_val(ptr[PATH_ELEM(curr_path)]);
+                if (type == iterator) {
+                    res = make_tuple(hp);
+                    hp[0] = make_arityval(3);
+                    hp[1] = CAR(lst);
+                    hp[2] = CDR(lst);
+                    patch_ptr = &hp[3];
+                    hp += 4;
+                } else {
+                    Eterm tup = TUPLE2(hp, CAR(lst), CDR(lst)); hp += 3;
+                    res = CONS(hp, tup, BIF_ARG_3); hp += 2;
+                }
+                elems--;
+                break;
+            }
+
+            node = ptr[PATH_ELEM(curr_path)];
+
+            curr_path >>= PATH_ELEM_SIZE;
+
+            if (i == PATH_ELEMS_PER_DIGIT) {
+                /* Switch to next bignum word if available,
+                   otherwise just follow 0 path */
+                i = 0;
+                if (path_length) {
+                    curr_path = *path_rest;
+                    path_length--;
+                    path_rest++;
+                } else {
+                    curr_path = 0;
+                }
+            }
+        }
+
+        /* We traverse the hashmap and return at most `elems` elements */
+        while(1) {
+            Eterm *ptr = (Eterm*)WSTACK_POP(stack);
+            Uint sz = (Uint)WSTACK_POP(stack);
+            Uint idx = (Uint)WSTACK_POP(stack);
+            Eterm node = (Eterm)WSTACK_POP(stack);
+
+            while (idx < sz && elems != 0 && is_list(ptr[idx])) {
+                Eterm *lst = list_val(ptr[idx]);
+                if (type == iterator) {
+                    *patch_ptr = make_tuple(hp);
+                    hp[0] = make_arityval(3);
+                    hp[1] = CAR(lst);
+                    hp[2] = CDR(lst);
+                    patch_ptr = &hp[3];
+                    hp += 4;
+                } else {
+                    Eterm tup = TUPLE2(hp, CAR(lst), CDR(lst)); hp += 3;
+                    res = CONS(hp, tup, res); hp += 2;
+                }
+                elems--;
+                idx++;
+            }
+
+            if (elems == 0) {
+                if (idx < sz) {
+                    /* There are more elements in this node to explore */
+                    WSTACK_PUSH4(stack, node, idx+1, sz, (UWord)ptr);
+                } else {
+                    /* pop stack to find the next value */
+                    while (!WSTACK_ISEMPTY(stack)) {
+                        Eterm *ptr = (Eterm*)WSTACK_POP(stack);
+                        Uint sz = (Uint)WSTACK_POP(stack);
+                        Uint idx = (Uint)WSTACK_POP(stack);
+                        Eterm node = (Eterm)WSTACK_POP(stack);
+                        if (idx < sz) {
+                            WSTACK_PUSH4(stack, node, idx+1, sz, (UWord)ptr);
+                            break;
+                        }
+                    }
+                }
+                break;
+            } else {
+                if (idx < sz) {
+                    Eterm hdr;
+                    /* Push next idx in current node */
+                    WSTACK_PUSH4(stack, node, idx+1, sz, (UWord)ptr);
+
+                    /* Push first idx in child node */
+                    node = ptr[idx];
+                    ptr = hashmap_val(ptr[idx]);
+                    hdr = *ptr++;
+                    sz = hashmap_node_size(hdr, &ptr);
+                    WSTACK_PUSH4(stack, node, 0, sz, (UWord)ptr);
+                }
+            }
+
+            /* There are no more element in the hashmap */
+            if (WSTACK_ISEMPTY(stack)) {
+                break;
+            }
+
+        }
+
+        if (!WSTACK_ISEMPTY(stack)) {
+            Uint depth = WSTACK_COUNT(stack) / 4 + 1;
+            /* +1 because we already have the first element in curr_path */
+            Eterm *path_digits = NULL;
+            Uint curr_path = 0;
+
+            /* If the path cannot fit in a small, we allocate a bignum */
+            if (depth >= PATH_ELEMS_PER_DIGIT) {
+                /* We need multiple ErtsDigit's to represent the path */
+                int big_size = BIG_NEED_FOR_BITS(depth * PATH_ELEM_SIZE);
+                hp = HAlloc(BIF_P, big_size);
+                hp[0] = make_pos_bignum_header(big_size - BIG_NEED_SIZE(0));
+                path_digits = hp + big_size - 1;
+            }
+
+
+            /* Pop the stack to create the complete path to the next leaf */
+            while(!WSTACK_ISEMPTY(stack)) {
+                Uint idx;
+
+                (void)WSTACK_POP(stack);
+                (void)WSTACK_POP(stack);
+                idx = (Uint)WSTACK_POP(stack)-1;
+                /* idx - 1 because idx in the stack is pointing to
+                   the next element to fetch. */
+                (void)WSTACK_POP(stack);
+
+                depth--;
+                if (depth % PATH_ELEMS_PER_DIGIT == 0) {
+                    /* Switch to next bignum element */
+                    path_digits[0] = curr_path;
+                    path_digits--;
+                    curr_path = 0;
+                }
+
+                curr_path <<= PATH_ELEM_SIZE;
+                curr_path |= idx;
+            }
+
+            if (path_digits) {
+                path_digits[0] = curr_path;
+                path = make_big(hp);
+            } else {
+                /* The Uint could be too large for a small */
+                path = erts_make_integer(curr_path, BIF_P);
+            }
+
+            if (type == iterator) {
+                hp = HAlloc(BIF_P, 2);
+                *patch_ptr = CONS(hp, path, map); hp += 2;
+            } else {
+                hp = HAlloc(BIF_P, 4);
+                res = CONS(hp, map, res); hp += 2;
+                res = CONS(hp, path, res); hp += 2;
+            }
+        } else {
+            if (type == iterator) {
+                *patch_ptr = am_none;
+                HRelease(BIF_P, hp + 4 * elems, hp);
+            } else {
+                HRelease(BIF_P, hp + (2+3) * elems, hp);
+            }
+        }
+        BIF_P->fcalls -= 4 * (orig_elems - elems);
+        DESTROY_WSTACK(stack);
+        BIF_RET(res);
+
+    badarg:
+        if (type == iterator) {
+            HRelease(BIF_P, hp + 4 * elems, hp);
+        } else {
+            HRelease(BIF_P, hp + (2+3) * elems, hp);
+        }
+        BIF_P->fcalls -= 4 * (orig_elems - elems);
+        DESTROY_WSTACK(stack);
+        BIF_ERROR(BIF_P, BADARG);
+    }
+}
+
 /* implementation of builtin emulations */
 
 #if !ERTS_AT_LEAST_GCC_VSN__(3, 4, 0)
diff --git a/erts/emulator/beam/erl_map.h b/erts/emulator/beam/erl_map.h
index c3ccf80b85..718d400e22 100644
--- a/erts/emulator/beam/erl_map.h
+++ b/erts/emulator/beam/erl_map.h
@@ -64,7 +64,6 @@ typedef struct flatmap_s {
 #define hashmap_shift_hash(Heap,Hx,Lvl,Key) \
     (((++(Lvl)) & 7) ? (Hx) >> 4 : hashmap_make_hash(CONS(Heap, make_small((Lvl)>>3), Key)))
 
-
 /* erl_term.h stuff */
 #define flatmap_get_values(x)        (((Eterm *)(x)) + sizeof(flatmap_t)/sizeof(Eterm))
 #define flatmap_get_keys(x)          (((Eterm *)tuple_val(((flatmap_t *)(x))->keys)) + 1)
diff --git a/erts/emulator/beam/erl_message.c b/erts/emulator/beam/erl_message.c
index b30c4a49d7..507cc989d2 100644
--- a/erts/emulator/beam/erl_message.c
+++ b/erts/emulator/beam/erl_message.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1997-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1997-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -33,6 +33,7 @@
 #include "erl_binary.h"
 #include "dtrace-wrapper.h"
 #include "beam_bp.h"
+#include "erl_proc_sig_queue.h"
 
 ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(message_ref,
 				 ErtsMessageRef,
@@ -170,7 +171,7 @@ erts_cleanup_offheap(ErlOffHeap *offheap)
             erts_bin_release(u.pb->val);
 	    break;
 	case FUN_SUBTAG:
-	    if (erts_smp_refc_dectest(&u.fun->fe->refc, 0) == 0) {
+	    if (erts_refc_dectest(&u.fun->fe->refc, 0) == 0) {
 		erts_erase_fun_entry(u.fun->fe);		    
 	    }
 	    break;
@@ -207,7 +208,7 @@ erts_cleanup_messages(ErtsMessage *msgp)
     while (mp) {
 	ErtsMessage *fmp;
 	ErlHeapFragment *bp;
-	if (is_non_value(ERL_MESSAGE_TERM(mp))) {
+	if (ERTS_SIG_IS_EXTERNAL_MSG(mp)) {
 	    if (is_not_immed(ERL_MESSAGE_TOKEN(mp))) {
 		bp = (ErlHeapFragment *) mp->data.dist_ext->ext_endp;
 		erts_cleanup_offheap(&bp->off_heap);
@@ -215,10 +216,13 @@ erts_cleanup_messages(ErtsMessage *msgp)
 	    if (mp->data.dist_ext)
 		erts_free_dist_ext_copy(mp->data.dist_ext);
 	}
-	else {
-	    if (mp->data.attached != ERTS_MSG_COMBINED_HFRAG)
+        else {
+	    if (ERTS_SIG_IS_INTERNAL_MSG(mp)
+                && mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
 		bp = mp->data.heap_frag;
+            }
 	    else {
+                mp->data.attached = ERTS_MSG_COMBINED_HFRAG;
 		bp = mp->hfrag.next;
 		erts_cleanup_offheap(&mp->hfrag.off_heap);
 	    }
@@ -260,20 +264,14 @@ erts_queue_dist_message(Process *rcvr,
                         Eterm from)
 {
     ErtsMessage* mp;
-#ifdef USE_VM_PROBES
-    Sint tok_label = 0;
-    Sint tok_lastcnt = 0;
-    Sint tok_serial = 0;
-#endif
-#ifdef ERTS_SMP
     erts_aint_t state;
-#endif
 
-    ERTS_SMP_LC_ASSERT(rcvr_locks == erts_proc_lc_my_proc_locks(rcvr));
+    ERTS_LC_ASSERT(rcvr_locks == erts_proc_lc_my_proc_locks(rcvr));
 
     mp = erts_alloc_message(0, NULL);
     mp->data.dist_ext = dist_ext;
 
+    ERL_MESSAGE_FROM(mp) = dist_ext->dep->sysname;
     ERL_MESSAGE_TERM(mp) = THE_NON_VALUE;
 #ifdef USE_VM_PROBES
     ERL_MESSAGE_DT_UTAG(mp) = NIL;
@@ -283,249 +281,169 @@ erts_queue_dist_message(Process *rcvr,
 #endif
 	ERL_MESSAGE_TOKEN(mp) = token;
 
-#ifdef ERTS_SMP
     if (!(rcvr_locks & ERTS_PROC_LOCK_MSGQ)) {
-	if (erts_smp_proc_trylock(rcvr, ERTS_PROC_LOCK_MSGQ) == EBUSY) {
+	if (erts_proc_trylock(rcvr, ERTS_PROC_LOCK_MSGQ) == EBUSY) {
 	    ErtsProcLocks need_locks = ERTS_PROC_LOCK_MSGQ;
             ErtsProcLocks unlocks =
                 rcvr_locks & ERTS_PROC_LOCKS_HIGHER_THAN(ERTS_PROC_LOCK_MSGQ);
 	    if (unlocks) {
-		erts_smp_proc_unlock(rcvr, unlocks);
+		erts_proc_unlock(rcvr, unlocks);
 		need_locks |= unlocks;
 	    }
-	    erts_smp_proc_lock(rcvr, need_locks);
+	    erts_proc_lock(rcvr, need_locks);
 	}
     }
 
-    state = erts_smp_atomic32_read_acqb(&rcvr->state);
-    if (state & (ERTS_PSFLG_PENDING_EXIT|ERTS_PSFLG_EXITING)) {
+
+    state = erts_atomic32_read_acqb(&rcvr->state);
+    if (state & ERTS_PSFLG_EXITING) {
 	if (!(rcvr_locks & ERTS_PROC_LOCK_MSGQ))
-	    erts_smp_proc_unlock(rcvr, ERTS_PROC_LOCK_MSGQ);
+	    erts_proc_unlock(rcvr, ERTS_PROC_LOCK_MSGQ);
 	/* Drop message if receiver is exiting or has a pending exit ... */
 	erts_cleanup_messages(mp);
     }
-    else
-#endif
-    if (IS_TRACED_FL(rcvr, F_TRACE_RECEIVE)) {
-        if (from == am_Empty)
-            from = dist_ext->dep->sysname;
-
-	/* Ahh... need to decode it in order to trace it... */
-	if (!(rcvr_locks & ERTS_PROC_LOCK_MSGQ))
-	    erts_smp_proc_unlock(rcvr, ERTS_PROC_LOCK_MSGQ);
-	if (!erts_decode_dist_message(rcvr, rcvr_locks, mp, 0))
-	    erts_free_message(mp);
-	else {
-	    Eterm msg = ERL_MESSAGE_TERM(mp);
-	    token = ERL_MESSAGE_TOKEN(mp);
-#ifdef USE_VM_PROBES
-	    if (DTRACE_ENABLED(message_queued)) {
-		DTRACE_CHARBUF(receiver_name, DTRACE_TERM_BUF_SIZE);
-
-		dtrace_proc_str(rcvr, receiver_name);
-                if (have_seqtrace(token)) {
-		    tok_label = signed_val(SEQ_TRACE_T_LABEL(token));
-		    tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token));
-		    tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token));
-		}
-		DTRACE6(message_queued,
-			receiver_name, size_object(msg), rcvr->msg.len,
-			tok_label, tok_lastcnt, tok_serial);
-	    }
-#endif
-	    erts_queue_message(rcvr, rcvr_locks, mp, msg, from);
-	}
-    }
     else {
-	/* Enqueue message on external format */
+	LINK_MESSAGE(rcvr, mp);
 
-#ifdef USE_VM_PROBES
-        if (DTRACE_ENABLED(message_queued)) {
-            DTRACE_CHARBUF(receiver_name, DTRACE_TERM_BUF_SIZE);
-
-            dtrace_proc_str(rcvr, receiver_name);
-            if (have_seqtrace(token)) {
-                tok_label = signed_val(SEQ_TRACE_T_LABEL(token));
-                tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token));
-                tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token));
-            }
-            /*
-             * TODO: We don't know the real size of the external message here.
-             *       -1 will appear to a D script as 4294967295.
-             */
-            DTRACE6(message_queued, receiver_name, -1, rcvr->msg.len + 1,
-                    tok_label, tok_lastcnt, tok_serial);
-        }
-#endif
-
-	LINK_MESSAGE(rcvr, mp, &mp->next, 1);
+        if (rcvr_locks & ERTS_PROC_LOCK_MAIN)
+            erts_proc_sig_fetch(rcvr);
 
 	if (!(rcvr_locks & ERTS_PROC_LOCK_MSGQ))
-	    erts_smp_proc_unlock(rcvr, ERTS_PROC_LOCK_MSGQ);
+	    erts_proc_unlock(rcvr, ERTS_PROC_LOCK_MSGQ);
 
-	erts_proc_notify_new_message(rcvr,
-#ifdef ERTS_SMP
-				     rcvr_locks
-#else
-				     0
-#endif
-	    );
+	erts_proc_notify_new_message(rcvr, rcvr_locks);
     }
 }
 
 /* Add messages last in message queue */
-static Sint
+static void
 queue_messages(Process* receiver,
-               erts_aint32_t *receiver_state,
                ErtsProcLocks receiver_locks,
                ErtsMessage* first,
                ErtsMessage** last,
-               Uint len,
-               Eterm from)
+               Uint len)
 {
-    ErtsTracingEvent* te;
-    Sint res;
     int locked_msgq = 0;
     erts_aint32_t state;
 
-    ASSERT(is_value(ERL_MESSAGE_TERM(first)));
-    ASSERT(ERL_MESSAGE_TOKEN(first) == am_undefined ||
-           ERL_MESSAGE_TOKEN(first) == NIL ||
-           is_tuple(ERL_MESSAGE_TOKEN(first)));
-
-#ifdef ERTS_SMP
-#ifdef ERTS_ENABLE_LOCK_CHECK
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(receiver) < ERTS_PROC_LOCK_MSGQ ||
-                       receiver_locks == erts_proc_lc_my_proc_locks(receiver));
+#ifdef DEBUG
+    {
+        ErtsMessage* fmsg = ERTS_SIG_IS_MSG(first) ? first : first->next;
+        ASSERT(fmsg);
+        ASSERT(is_value(ERL_MESSAGE_TERM(fmsg)));
+        ASSERT(is_value(ERL_MESSAGE_FROM(fmsg)));
+        ASSERT(ERL_MESSAGE_TOKEN(fmsg) == am_undefined ||
+               ERL_MESSAGE_TOKEN(fmsg) == NIL ||
+               is_tuple(ERL_MESSAGE_TOKEN(fmsg)));
+    }
 #endif
 
-    if (!(receiver_locks & ERTS_PROC_LOCK_MSGQ)) {
-	if (erts_smp_proc_trylock(receiver, ERTS_PROC_LOCK_MSGQ) == EBUSY) {
-            ErtsProcLocks need_locks;
+    ERTS_LC_ASSERT((erts_proc_lc_my_proc_locks(receiver) & ERTS_PROC_LOCK_MSGQ)
+                   == (receiver_locks & ERTS_PROC_LOCK_MSGQ));
 
-	    if (receiver_state)
-		state = *receiver_state;
-	    else
-		state = erts_smp_atomic32_read_nob(&receiver->state);
-	    if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT))
-		goto exiting;
-
-            need_locks = receiver_locks & ERTS_PROC_LOCKS_HIGHER_THAN(ERTS_PROC_LOCK_MSGQ);
-	    if (need_locks) {
-		erts_smp_proc_unlock(receiver, need_locks);
-	    }
-            need_locks |= ERTS_PROC_LOCK_MSGQ;
-	    erts_smp_proc_lock(receiver, need_locks);
-	}
+    if (!(receiver_locks & ERTS_PROC_LOCK_MSGQ)) {
+        erts_proc_lock(receiver, ERTS_PROC_LOCK_MSGQ);
 	locked_msgq = 1;
     }
 
-#endif
-
-    state = erts_smp_atomic32_read_nob(&receiver->state);
+    state = erts_atomic32_read_nob(&receiver->state);
 
-    if (state & (ERTS_PSFLG_PENDING_EXIT|ERTS_PSFLG_EXITING)) {
-#ifdef ERTS_SMP
-    exiting:
-#endif
+    if (state & ERTS_PSFLG_EXITING) {
 	/* Drop message if receiver is exiting or has a pending exit... */
 	if (locked_msgq)
-	    erts_smp_proc_unlock(receiver, ERTS_PROC_LOCK_MSGQ);
+            erts_proc_unlock(receiver, ERTS_PROC_LOCK_MSGQ);
+        if (ERTS_SIG_IS_NON_MSG(first)) {
+            ErtsSchedulerData* esdp = erts_get_scheduler_data();
+            ASSERT(esdp);
+            ASSERT(!esdp->pending_signal.sig);
+            esdp->pending_signal.sig = (ErtsSignal*) first;
+            esdp->pending_signal.to = receiver->common.id;
+            first = first->next;
+        }
 	erts_cleanup_messages(first);
-	return 0;
+        return;
     }
 
-    res = receiver->msg.len;
-#ifdef ERTS_SMP
-    if (receiver_locks & ERTS_PROC_LOCK_MAIN) {
-	/*
-	 * We move 'in queue' to 'private queue' and place
-	 * message at the end of 'private queue' in order
-	 * to ensure that the 'in queue' doesn't contain
-	 * references into the heap. By ensuring this,
-	 * we don't need to include the 'in queue' in
-	 * the root set when garbage collecting.
-	 */
-	res += receiver->msg_inq.len;
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(receiver);
-        LINK_MESSAGE_PRIVQ(receiver, first, last, len);
+    if (last == &first->next) {
+        ASSERT(len == 1);
+        LINK_MESSAGE(receiver, first);
     }
-    else
-#endif
-    {
-	LINK_MESSAGE(receiver, first, last, len);
+    else {
+        erts_enqueue_signals(receiver, first, last, NULL, len, state);
     }
 
-    if (IS_TRACED_FL(receiver, F_TRACE_RECEIVE)
-        && (te = &erts_receive_tracing[erts_active_bp_ix()],
-            te->on)) {
-
-        ErtsMessage *msg = first;
-
-#ifdef USE_VM_PROBES
-        if (DTRACE_ENABLED(message_queued)) {
-            DTRACE_CHARBUF(receiver_name, DTRACE_TERM_BUF_SIZE);
-            Sint tok_label = 0;
-            Sint tok_lastcnt = 0;
-            Sint tok_serial = 0;
-            Eterm seq_trace_token = ERL_MESSAGE_TOKEN(msg);
-
-            dtrace_proc_str(receiver, receiver_name);
-            if (seq_trace_token != NIL && is_tuple(seq_trace_token)) {
-                tok_label = signed_val(SEQ_TRACE_T_LABEL(seq_trace_token));
-                tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(seq_trace_token));
-                tok_serial = signed_val(SEQ_TRACE_T_SERIAL(seq_trace_token));
-            }
-            DTRACE6(message_queued,
-                    receiver_name, size_object(ERL_MESSAGE_TERM(msg)),
-                    receiver->msg.len,
-                    tok_label, tok_lastcnt, tok_serial);
-        }
-#endif
-        while (msg) {
-            trace_receive(receiver, from, ERL_MESSAGE_TERM(msg), te);
-            msg = msg->next;
-        }
+    if (receiver_locks & ERTS_PROC_LOCK_MAIN)
+        erts_proc_sig_fetch(receiver);
 
-    }
     if (locked_msgq) {
-	erts_smp_proc_unlock(receiver, ERTS_PROC_LOCK_MSGQ);
+	erts_proc_unlock(receiver, ERTS_PROC_LOCK_MSGQ);
     }
 
-#ifdef ERTS_SMP
-    erts_proc_notify_new_message(receiver, receiver_locks);
-#else
-    erts_proc_notify_new_message(receiver, 0);
-#endif
-    return res;
+    if (last == &first->next)
+        erts_proc_notify_new_message(receiver, receiver_locks);
+    else
+        erts_proc_notify_new_sig(receiver, state, ERTS_PSFLG_ACTIVE);
 }
 
-static Sint
-queue_message(Process* receiver,
-              erts_aint32_t *receiver_state,
-              ErtsProcLocks receiver_locks,
-              ErtsMessage* mp, Eterm msg, Eterm from)
+static ERTS_INLINE
+ErtsMessage* prepend_pending_sig_maybe(Process* sender, Process* receiver,
+                                       ErtsMessage* mp)
 {
-    ERL_MESSAGE_TERM(mp) = msg;
-    return queue_messages(receiver, receiver_state, receiver_locks,
-                          mp, &mp->next, 1, from);
+    ErtsSchedulerData* esdp = sender->scheduler_data;
+    ErtsSignal* pend_sig;
+
+    if (!esdp || esdp->pending_signal.to != receiver->common.id)
+        return mp;
+   
+     pend_sig = esdp->pending_signal.sig;
+
+     ASSERT(esdp->pending_signal.dbg_from == sender);
+     esdp->pending_signal.sig = NULL;
+     esdp->pending_signal.to = THE_NON_VALUE;
+     pend_sig->common.next = mp;
+     pend_sig->common.specific.next = NULL;
+     return (ErtsMessage*) pend_sig;
 }
 
-Sint
+/**
+ *
+ * @brief Send one message from *NOT* a local process.
+ *
+ */
+void
 erts_queue_message(Process* receiver, ErtsProcLocks receiver_locks,
                    ErtsMessage* mp, Eterm msg, Eterm from)
 {
-    return queue_message(receiver, NULL, receiver_locks, mp, msg, from);
+    ASSERT(is_not_internal_pid(from));
+    ERL_MESSAGE_TERM(mp) = msg;
+    ERL_MESSAGE_FROM(mp) = from;
+    queue_messages(receiver, receiver_locks, mp, &mp->next, 1);
 }
 
+/**
+ * @brief Send one message from a local process.
+ */
+void
+erts_queue_proc_message(Process* sender,
+                        Process* receiver, ErtsProcLocks receiver_locks,
+                        ErtsMessage* mp, Eterm msg)
+{
+    ERL_MESSAGE_TERM(mp) = msg;
+    ERL_MESSAGE_FROM(mp) = sender->common.id;
+    queue_messages(receiver, receiver_locks,
+                   prepend_pending_sig_maybe(sender, receiver, mp),
+                   &mp->next, 1);
+}
 
-Sint
-erts_queue_messages(Process* receiver, ErtsProcLocks receiver_locks,
-                    ErtsMessage* first, ErtsMessage** last, Uint len,
-                    Eterm from)
+
+void
+erts_queue_proc_messages(Process* sender,
+                         Process* receiver, ErtsProcLocks receiver_locks,
+                         ErtsMessage* first, ErtsMessage** last, Uint len)
 {
-    return queue_messages(receiver, NULL, receiver_locks,
-                          first, last, len, from);
+    queue_messages(receiver, receiver_locks,
+                   prepend_pending_sig_maybe(sender, receiver, first),
+                   last, len);
 }
 
 void
@@ -594,9 +512,7 @@ erts_try_alloc_message_on_heap(Process *pp,
 			       ErlOffHeap **ohpp,
 			       int *on_heap_p)
 {
-#ifdef ERTS_SMP
     int locked_main = 0;
-#endif
     ErtsMessage *mp;
 
     ASSERT(!(*psp & ERTS_PSFLG_OFF_HEAP_MSGQ));
@@ -604,15 +520,9 @@ erts_try_alloc_message_on_heap(Process *pp,
     if ((*psp) & ERTS_PSFLGS_VOLATILE_HEAP)
 	goto in_message_fragment;
     else if (
-#if defined(ERTS_SMP)
 	*plp & ERTS_PROC_LOCK_MAIN
-#else
-	pp
-#endif
 	) {
-#ifdef ERTS_SMP
     try_on_heap:
-#endif
 	if (((*psp) & ERTS_PSFLGS_VOLATILE_HEAP)
 	    || (pp->flags & F_DISABLE_GC)
 	    || HEAP_LIMIT(pp) - HEAP_TOP(pp) <= sz) {
@@ -620,12 +530,10 @@ erts_try_alloc_message_on_heap(Process *pp,
 	     * The heap is either potentially in an inconsistent
 	     * state, or not large enough.
 	     */
-#ifdef ERTS_SMP
 	    if (locked_main) {
 		*plp &= ~ERTS_PROC_LOCK_MAIN;
-		erts_smp_proc_unlock(pp, ERTS_PROC_LOCK_MAIN);
+		erts_proc_unlock(pp, ERTS_PROC_LOCK_MAIN);
 	    }
-#endif
 	    goto in_message_fragment;
 	}
 
@@ -636,17 +544,15 @@ erts_try_alloc_message_on_heap(Process *pp,
 	mp->data.attached = NULL;
 	*on_heap_p = !0;
     }
-#ifdef ERTS_SMP
-    else if (pp && erts_smp_proc_trylock(pp, ERTS_PROC_LOCK_MAIN) == 0) {
+    else if (pp && erts_proc_trylock(pp, ERTS_PROC_LOCK_MAIN) == 0) {
 	locked_main = 1;
-	*psp = erts_smp_atomic32_read_nob(&pp->state);
+	*psp = erts_atomic32_read_nob(&pp->state);
 	*plp |= ERTS_PROC_LOCK_MAIN;
 	goto try_on_heap;
     }
-#endif
     else {
     in_message_fragment:
-	if (!((*psp) & ERTS_PSFLG_ON_HEAP_MSGQ)) {
+	if ((*psp) & ERTS_PSFLG_OFF_HEAP_MSGQ) {
 	    mp = erts_alloc_message(sz, hpp);
 	    *ohpp = sz == 0 ? NULL : &mp->hfrag.off_heap;
 	}
@@ -674,7 +580,7 @@ erts_try_alloc_message_on_heap(Process *pp,
  * Send a local message when sender & receiver processes are known.
  */
 
-Sint
+void
 erts_send_message(Process* sender,
 		  Process* receiver,
 		  ErtsProcLocks *receiver_locks,
@@ -685,7 +591,6 @@ erts_send_message(Process* sender,
     ErtsMessage* mp;
     ErlOffHeap *ohp;
     Eterm token = NIL;
-    Sint res = 0;
 #ifdef USE_VM_PROBES
     DTRACE_CHARBUF(sender_name, 64);
     DTRACE_CHARBUF(receiver_name, 64);
@@ -712,7 +617,7 @@ erts_send_message(Process* sender,
     }
 #endif
 
-    receiver_state = erts_smp_atomic32_read_nob(&receiver->state);
+    receiver_state = erts_atomic32_read_nob(&receiver->state);
 
     if (SEQ_TRACE_TOKEN(sender) != NIL && !(flags & ERTS_SND_FLG_NO_SEQ_TRACE)) {
         Eterm* hp;
@@ -731,7 +636,8 @@ erts_send_message(Process* sender,
 	    seq_trace_update_send(sender);
 	    seq_trace_output(stoken, message, SEQ_TRACE_SEND,
 			     receiver->common.id, sender);
-	    seq_trace_size = 6; /* TUPLE5 */
+
+	    seq_trace_size = size_object(stoken);
 	}
 #ifdef USE_VM_PROBES
         if (DT_UTAG_FLAGS(sender) & DT_UTAG_SPREADING) {
@@ -780,7 +686,7 @@ erts_send_message(Process* sender,
 	}
         if (DTRACE_ENABLED(message_send)) {
             if (have_seqtrace(stoken)) {
-		tok_label = signed_val(SEQ_TRACE_T_LABEL(stoken));
+                tok_label = SEQ_TRACE_T_DTRACE_LABEL(stoken);
 		tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(stoken));
 		tok_serial = signed_val(SEQ_TRACE_T_SERIAL(stoken));
 	    }
@@ -827,13 +733,8 @@ erts_send_message(Process* sender,
 #ifdef USE_VM_PROBES
     ERL_MESSAGE_DT_UTAG(mp) = utag;
 #endif
-    res = queue_message(receiver,
-			&receiver_state,
-			*receiver_locks,
-			mp, message,
-                        sender->common.id);
 
-    return res;
+    erts_queue_proc_message(sender, receiver, *receiver_locks, mp, message);
 }
 
 
@@ -952,70 +853,77 @@ Sint
 erts_move_messages_off_heap(Process *c_p)
 {
     int reds = 1;
+    int i;
+    ErtsMessage *msgq[] = {c_p->sig_qs.first, c_p->sig_qs.cont};
     /*
      * Move all messages off heap. This *only* occurs when the
      * process had off heap message disabled and just enabled
      * it...
      */
-    ErtsMessage *mp;
 
-    reds += c_p->msg.len / 10;
+    reds += erts_proc_sig_privqs_len(c_p) / 10;
 
-    ASSERT(erts_smp_atomic32_read_nob(&c_p->state)
+    ASSERT(erts_atomic32_read_nob(&c_p->state)
 	   & ERTS_PSFLG_OFF_HEAP_MSGQ);
     ASSERT(c_p->flags & F_OFF_HEAP_MSGQ_CHNG);
 
-    for (mp = c_p->msg.first; mp; mp = mp->next) {
-	Uint msg_sz, token_sz;
+    for (i = 0; i < sizeof(msgq)/sizeof(msgq[0]); i++) {
+        ErtsMessage *mp;
+        for (mp = msgq[i]; mp; mp = mp->next) {
+            Uint msg_sz, token_sz;
 #ifdef USE_VM_PROBES
-	Uint utag_sz;
+            Uint utag_sz;
 #endif
-	Eterm *hp;
-	ErlHeapFragment *hfrag;
+            Eterm *hp;
+            ErlHeapFragment *hfrag;
+
+            if (!ERTS_SIG_IS_INTERNAL_MSG(mp))
+                continue;
 
-	if (mp->data.attached)
-	    continue;
+            if (mp->data.attached)
+                continue;
 
-	if (is_immed(ERL_MESSAGE_TERM(mp))
+            if (is_immed(ERL_MESSAGE_TERM(mp))
 #ifdef USE_VM_PROBES
-	    && is_immed(ERL_MESSAGE_DT_UTAG(mp))
+                && is_immed(ERL_MESSAGE_DT_UTAG(mp))
 #endif
-	    && is_not_immed(ERL_MESSAGE_TOKEN(mp)))
-	    continue;
+                && is_not_immed(ERL_MESSAGE_TOKEN(mp)))
+                continue;
 
-	/*
-	 * The message refers into the heap. Copy the message
-	 * from the heap into a heap fragment and attach
-	 * it to the message...
-	 */
-	msg_sz = size_object(ERL_MESSAGE_TERM(mp));
+            /*
+             * The message refers into the heap. Copy the message
+             * from the heap into a heap fragment and attach
+             * it to the message...
+             */
+            msg_sz = size_object(ERL_MESSAGE_TERM(mp));
 #ifdef USE_VM_PROBES
-	utag_sz = size_object(ERL_MESSAGE_DT_UTAG(mp));
+            utag_sz = size_object(ERL_MESSAGE_DT_UTAG(mp));
 #endif
-	token_sz = size_object(ERL_MESSAGE_TOKEN(mp));
+            token_sz = size_object(ERL_MESSAGE_TOKEN(mp));
 
-	hfrag = new_message_buffer(msg_sz
+            hfrag = new_message_buffer(msg_sz
 #ifdef USE_VM_PROBES
-				   + utag_sz
+                                       + utag_sz
 #endif
-				   + token_sz);
-	hp = hfrag->mem;
-	if (is_not_immed(ERL_MESSAGE_TERM(mp)))
-	    ERL_MESSAGE_TERM(mp) = copy_struct(ERL_MESSAGE_TERM(mp),
-                                               msg_sz, &hp,
-                                               &hfrag->off_heap);
-	if (is_not_immed(ERL_MESSAGE_TOKEN(mp)))
-	    ERL_MESSAGE_TOKEN(mp) = copy_struct(ERL_MESSAGE_TOKEN(mp),
-						token_sz, &hp,
-						&hfrag->off_heap);
+                                       + token_sz);
+            hp = hfrag->mem;
+            if (is_not_immed(ERL_MESSAGE_TERM(mp)))
+                ERL_MESSAGE_TERM(mp) = copy_struct(ERL_MESSAGE_TERM(mp),
+                                                   msg_sz, &hp,
+                                                   &hfrag->off_heap);
+            if (is_not_immed(ERL_MESSAGE_TOKEN(mp)))
+                ERL_MESSAGE_TOKEN(mp) = copy_struct(ERL_MESSAGE_TOKEN(mp),
+                                                    token_sz, &hp,
+                                                    &hfrag->off_heap);
 #ifdef USE_VM_PROBES
-	if (is_not_immed(ERL_MESSAGE_DT_UTAG(mp)))
-	    ERL_MESSAGE_DT_UTAG(mp) = copy_struct(ERL_MESSAGE_DT_UTAG(mp),
-						  utag_sz, &hp,
-						  &hfrag->off_heap);
+            if (is_not_immed(ERL_MESSAGE_DT_UTAG(mp)))
+                ERL_MESSAGE_DT_UTAG(mp) = copy_struct(ERL_MESSAGE_DT_UTAG(mp),
+                                                      utag_sz, &hp,
+                                                      &hfrag->off_heap);
 #endif
-	mp->data.heap_frag = hfrag;
-	reds += 1;
+            mp->data.heap_frag = hfrag;
+            reds += 1;
+        }
     }
 
     return reds;
@@ -1026,9 +934,9 @@ erts_complete_off_heap_message_queue_change(Process *c_p)
 {
     int reds = 1;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
     ASSERT(c_p->flags & F_OFF_HEAP_MSGQ_CHNG);
-    ASSERT(erts_smp_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_OFF_HEAP_MSGQ);
+    ASSERT(erts_atomic32_read_nob(&c_p->state) & ERTS_PSFLG_OFF_HEAP_MSGQ);
 
     /*
      * This job was first initiated when the process changed to off heap
@@ -1040,13 +948,13 @@ erts_complete_off_heap_message_queue_change(Process *c_p)
      */
 
     if (!(c_p->flags & F_OFF_HEAP_MSGQ))
-	erts_smp_atomic32_read_band_nob(&c_p->state,
+	erts_atomic32_read_band_nob(&c_p->state,
 					~ERTS_PSFLG_OFF_HEAP_MSGQ);
     else {
 	reds += 2;
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+        erts_proc_sig_fetch(c_p);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
 	reds += erts_move_messages_off_heap(c_p);
     }
     c_p->flags &= ~F_OFF_HEAP_MSGQ_CHNG;
@@ -1083,16 +991,16 @@ erts_change_message_queue_management(Process *c_p, Eterm new_state)
 
 #ifdef DEBUG
     if (c_p->flags & F_OFF_HEAP_MSGQ) {
-	ASSERT(erts_smp_atomic32_read_nob(&c_p->state)
+	ASSERT(erts_atomic32_read_nob(&c_p->state)
 	       & ERTS_PSFLG_OFF_HEAP_MSGQ);
     }
     else {
 	if (c_p->flags & F_OFF_HEAP_MSGQ_CHNG) {
-	    ASSERT(erts_smp_atomic32_read_nob(&c_p->state)
+	    ASSERT(erts_atomic32_read_nob(&c_p->state)
 		   & ERTS_PSFLG_OFF_HEAP_MSGQ);
 	}
 	else {
-	    ASSERT(!(erts_smp_atomic32_read_nob(&c_p->state)
+	    ASSERT(!(erts_atomic32_read_nob(&c_p->state)
 		     & ERTS_PSFLG_OFF_HEAP_MSGQ));
 	}
     }
@@ -1109,8 +1017,6 @@ erts_change_message_queue_management(Process *c_p, Eterm new_state)
 	case am_on_heap:
 	    c_p->flags |= F_ON_HEAP_MSGQ;
 	    c_p->flags &= ~F_OFF_HEAP_MSGQ;
-	    erts_smp_atomic32_read_bor_nob(&c_p->state,
-					   ERTS_PSFLG_ON_HEAP_MSGQ);
 	    /*
 	     * We are not allowed to clear ERTS_PSFLG_OFF_HEAP_MSGQ
 	     * if a off heap change is ongoing. It will be adjusted
@@ -1118,7 +1024,7 @@ erts_change_message_queue_management(Process *c_p, Eterm new_state)
 	     */
 	    if (!(c_p->flags & F_OFF_HEAP_MSGQ_CHNG)) {
 		/* Safe to clear ERTS_PSFLG_OFF_HEAP_MSGQ... */
-		erts_smp_atomic32_read_band_nob(&c_p->state,
+		erts_atomic32_read_band_nob(&c_p->state,
 						~ERTS_PSFLG_OFF_HEAP_MSGQ);
 	    }
 	    break;
@@ -1136,8 +1042,6 @@ erts_change_message_queue_management(Process *c_p, Eterm new_state)
 	    break;
 	case am_off_heap:
 	    c_p->flags &= ~F_ON_HEAP_MSGQ;
-	    erts_smp_atomic32_read_band_nob(&c_p->state,
-					    ~ERTS_PSFLG_ON_HEAP_MSGQ);
 	    goto change_to_off_heap;
 	default:
 	    res = THE_NON_VALUE; /* badarg */
@@ -1171,7 +1075,7 @@ change_to_off_heap:
 	 * change has completed, GC does not need to inspect
 	 * the message queue at all.
 	 */
-	erts_smp_atomic32_read_bor_nob(&c_p->state,
+	erts_atomic32_read_bor_nob(&c_p->state,
 				       ERTS_PSFLG_OFF_HEAP_MSGQ);
 	c_p->flags |= F_OFF_HEAP_MSGQ_CHNG;
 	cohmq = erts_alloc(ERTS_ALC_T_MSGQ_CHNG,
@@ -1259,139 +1163,6 @@ erts_decode_dist_message(Process *proc, ErtsProcLocks proc_locks,
     return 1;
 }
 
-/*
- * ERTS_INSPECT_MSGQ_KEEP_OH_MSGS == 0 will move off heap messages
- * into the heap of the inspected process if off_heap_message_queue
- * is false when process_info(_, messages) is called. That is, the
- * following GC will have more data in the rootset compared to the
- * scenario when process_info(_, messages) had not been called.
- *
- * ERTS_INSPECT_MSGQ_KEEP_OH_MSGS != 0 will keep off heap messages
- * off heap when process_info(_, messages) is called regardless of
- * the off_heap_message_queue setting of the process. That is, it
- * will change the following execution of the process as little as
- * possible.
- */
-#define ERTS_INSPECT_MSGQ_KEEP_OH_MSGS 1
-
-Uint
-erts_prep_msgq_for_inspection(Process *c_p, Process *rp,
-			      ErtsProcLocks rp_locks, ErtsMessageInfo *mip)
-{
-    Uint tot_heap_size;
-    ErtsMessage* mp;
-    Sint i;
-    int self_on_heap;
-    
-    /*
-     * Prepare the message queue for inspection
-     * by process_info().
-     *
-     *
-     * - Decode all messages on external format
-     * - Remove all corrupt dist messages from queue
-     * - Save pointer to, and heap size need of each
-     *   message in the mip array.
-     * - Return total heap size need for all messages
-     *   that needs to be copied.
-     *
-     * If ERTS_INSPECT_MSGQ_KEEP_OH_MSGS == 0:
-     * - In case off heap messages is disabled and
-     *   we are inspecting our own queue, move all
-     *   off heap data into the heap.
-     */
-
-    self_on_heap = c_p == rp && !(c_p->flags & F_OFF_HEAP_MSGQ);
-
-    tot_heap_size = 0;
-    i = 0;
-    mp = rp->msg.first;
-    while (mp) {
-	Eterm msg = ERL_MESSAGE_TERM(mp);
-
-	mip[i].size = 0;
-
-	if (is_non_value(msg)) {
-	    /* Dist message on external format; decode it... */
-	    if (mp->data.attached)
-		erts_decode_dist_message(rp, rp_locks, mp,
-					 ERTS_INSPECT_MSGQ_KEEP_OH_MSGS);
-
-	    msg = ERL_MESSAGE_TERM(mp);
-
-	    if (is_non_value(msg)) {
-		ErtsMessage **mpp;
-		ErtsMessage *bad_mp = mp;
-		/*
-		 * Bad distribution message; remove
-		 * it from the queue...
-		 */
-		ASSERT(!mp->data.attached);
-
-		mpp = i == 0 ? &rp->msg.first : &mip[i-1].msgp->next;
-
-		ASSERT(*mpp == bad_mp);
-
-		erts_msgq_update_internal_pointers(&rp->msg, mpp, &bad_mp->next);
-
-		mp = mp->next;
-		*mpp = mp;
-		rp->msg.len--;
-		bad_mp->next = NULL;
-		erts_cleanup_messages(bad_mp);
-		continue;
-	    }
-	}
-
-	ASSERT(is_value(msg));
-
-#if ERTS_INSPECT_MSGQ_KEEP_OH_MSGS
-	if (is_not_immed(msg) && (!self_on_heap || mp->data.attached)) {
-	    Uint sz = size_object(msg);
-	    mip[i].size = sz;
-	    tot_heap_size += sz;
-	}
-#else
-	if (self_on_heap) {
-	    if (mp->data.attached) {
-		ErtsMessage *tmp = NULL;
-		if (mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
-		    erts_link_mbuf_to_proc(rp, mp->data.heap_frag);
-		    mp->data.attached = NULL;
-		}
-		else {
-		    /*
-		     * Need to replace the message reference since
-		     * we will get references to the message data
-		     * from the heap...
-		     */
-		    ErtsMessage **mpp;
-		    tmp = erts_alloc_message(0, NULL);
-		    sys_memcpy((void *) tmp->m, (void *) mp->m,
-			       sizeof(Eterm)*ERL_MESSAGE_REF_ARRAY_SZ); 
-		    mpp = i == 0 ? &rp->msg.first : &mip[i-1].msgp->next;
-		    erts_msgq_replace_msg_ref(&rp->msg, tmp, mpp);
-		    erts_save_message_in_proc(rp, mp);
-		    mp = tmp;
-		}
-	    }
-	}
-	else if (is_not_immed(msg)) {
-	    Uint sz = size_object(msg);
-	    mip[i].size = sz;
-	    tot_heap_size += sz;
-	}
-
-#endif
-
-	mip[i].msgp = mp;
-	i++;
-	mp = mp->next;
-    }
-
-    return tot_heap_size;
-}
-
 void erts_factory_proc_init(ErtsHeapFactory* factory,
 			    Process* p)
 {
@@ -1452,7 +1223,7 @@ erts_factory_message_create(ErtsHeapFactory* factory,
     int on_heap;
     erts_aint32_t state;
 
-    state = proc ? erts_smp_atomic32_read_nob(&proc->state) : 0;
+    state = proc ? erts_atomic32_read_nob(&proc->state) : 0;
 
     if (state & ERTS_PSFLG_OFF_HEAP_MSGQ) {
 	msgp = erts_alloc_message(sz, &hp);
@@ -1467,7 +1238,7 @@ erts_factory_message_create(ErtsHeapFactory* factory,
     }
 
     if (on_heap) {
-	ERTS_SMP_ASSERT(*proc_locksp & ERTS_PROC_LOCK_MAIN);
+	ERTS_ASSERT(*proc_locksp & ERTS_PROC_LOCK_MAIN);
 	ASSERT(ohp == &proc->off_heap);
 	factory->mode = FACTORY_HALLOC;
 	factory->p = proc;
@@ -1581,32 +1352,18 @@ void erts_factory_dummy_init(ErtsHeapFactory* factory)
     factory->mode = FACTORY_CLOSED;
 }
 
-static void reserve_heap(ErtsHeapFactory*, Uint need, Uint xtra);
-
-Eterm* erts_produce_heap(ErtsHeapFactory* factory, Uint need, Uint xtra)
-{
-    Eterm* res;
-
-    ASSERT((unsigned int)factory->mode > (unsigned int)FACTORY_CLOSED);
-    if (factory->hp + need > factory->hp_end) {
-	reserve_heap(factory, need, xtra);
-    }
-    res = factory->hp;
-    factory->hp += need;
-    return res;
-}
-
 Eterm* erts_reserve_heap(ErtsHeapFactory* factory, Uint need)
 {
     ASSERT((unsigned int)factory->mode > (unsigned int)FACTORY_CLOSED);
     if (factory->hp + need > factory->hp_end) {
-	reserve_heap(factory, need, 200);
+	erts_reserve_heap__(factory, need, 200);
     }
     return factory->hp;
 }
 
-static void reserve_heap(ErtsHeapFactory* factory, Uint need, Uint xtra)
+void erts_reserve_heap__(ErtsHeapFactory* factory, Uint need, Uint xtra)
 {
+    /* internal... */
     ErlHeapFragment* bp;
 
     switch (factory->mode) {
@@ -1616,7 +1373,9 @@ static void reserve_heap(ErtsHeapFactory* factory, Uint need, Uint xtra)
 	factory->hp_end = factory->hp + need;
 	return;
 
-    case FACTORY_MESSAGE:
+    case FACTORY_MESSAGE: {
+        int replace_oh;
+        int replace_msg_hfrag;
 	if (!factory->heap_frags) {
 	    ASSERT(factory->message->data.attached == ERTS_MSG_COMBINED_HFRAG);
 	    bp = &factory->message->hfrag;
@@ -1628,25 +1387,45 @@ static void reserve_heap(ErtsHeapFactory* factory, Uint need, Uint xtra)
 	    bp = factory->heap_frags;
 	}
 
+        replace_oh = 0;
+        replace_msg_hfrag = 0;
+
         if (bp) {
-	    ASSERT(factory->hp > bp->mem);
+	    ASSERT(factory->hp >= bp->mem);
 	    ASSERT(factory->hp <= factory->hp_end);
 	    ASSERT(factory->hp_end == bp->mem + bp->alloc_size);
 
 	    bp->used_size = factory->hp - bp->mem;
+            if (!bp->used_size && factory->heap_frags) {
+                factory->heap_frags = bp->next;
+                bp->next = NULL;
+                ASSERT(!bp->off_heap.first);
+                if (factory->off_heap == &bp->off_heap)
+                    replace_oh = !0;
+                if (factory->message && factory->message->data.heap_frag == bp)
+                    replace_msg_hfrag = !0;
+                free_message_buffer(bp);
+            }
         }
 	bp = (ErlHeapFragment*) ERTS_HEAP_ALLOC(factory->alloc_type,
 						ERTS_HEAP_FRAG_SIZE(need+xtra));
 	bp->next = factory->heap_frags;
 	factory->heap_frags = bp;
 	bp->alloc_size = need + xtra;
-	bp->used_size = need;
+	bp->used_size = need + xtra;
 	bp->off_heap.first = NULL;
 	bp->off_heap.overhead = 0;
-
+        if (replace_oh) {
+            factory->off_heap = &bp->off_heap;
+            factory->off_heap_saved.first = factory->off_heap->first;
+            factory->off_heap_saved.overhead = factory->off_heap->overhead;
+        }
+        if (replace_msg_hfrag)
+            factory->message->data.heap_frag = bp;
 	factory->hp     = bp->mem;
 	factory->hp_end = bp->mem + bp->alloc_size;
 	return;
+    }
 
     case FACTORY_STATIC:
     case FACTORY_CLOSED:
@@ -1729,9 +1508,11 @@ void erts_factory_trim_and_close(ErtsHeapFactory* factory,
         if (bp->next == NULL) {
             Uint used_sz = factory->hp - bp->mem;
             ASSERT(used_sz <= bp->alloc_size);
-	    if (used_sz > 0)
-		bp = erts_resize_message_buffer(bp, used_sz,
-						brefs, brefs_size);
+	    if (used_sz > 0) {
+                if (used_sz != bp->alloc_size)
+                    bp = erts_resize_message_buffer(bp, used_sz,
+                                                    brefs, brefs_size);
+            }
 	    else {
 		free_message_buffer(bp);
 		bp = NULL;
diff --git a/erts/emulator/beam/erl_message.h b/erts/emulator/beam/erl_message.h
index 42ed14e69c..d120111634 100644
--- a/erts/emulator/beam/erl_message.h
+++ b/erts/emulator/beam/erl_message.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1997-2016. All Rights Reserved.
+ * Copyright Ericsson AB 1997-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -21,6 +21,11 @@
 #ifndef __ERL_MESSAGE_H__
 #define __ERL_MESSAGE_H__
 
+#include "sys.h"
+#define ERTS_PROC_SIG_QUEUE_TYPE_ONLY
+#include "erl_proc_sig_queue.h"
+#undef ERTS_PROC_SIG_QUEUE_TYPE_ONLY
+
 struct proc_bin;
 struct external_thing_;
 
@@ -84,12 +89,33 @@ void erts_factory_static_init(ErtsHeapFactory*, Eterm* hp, Uint size, ErlOffHeap
 void erts_factory_tmp_init(ErtsHeapFactory*, Eterm* hp, Uint size, Uint32 atype);
 void erts_factory_dummy_init(ErtsHeapFactory*);
 
-Eterm* erts_produce_heap(ErtsHeapFactory*, Uint need, Uint xtra);
+ERTS_GLB_INLINE Eterm* erts_produce_heap(ErtsHeapFactory*, Uint need, Uint xtra);
+
 Eterm* erts_reserve_heap(ErtsHeapFactory*, Uint need);
 void erts_factory_close(ErtsHeapFactory*);
 void erts_factory_trim_and_close(ErtsHeapFactory*,Eterm *brefs, Uint brefs_size);
 void erts_factory_undo(ErtsHeapFactory*);
 
+void erts_reserve_heap__(ErtsHeapFactory*, Uint need, Uint xtra); /* internal */
+
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE Eterm *
+erts_produce_heap(ErtsHeapFactory* factory, Uint need, Uint xtra)
+{
+    Eterm* res;
+
+    ASSERT((unsigned int)factory->mode > (unsigned int)FACTORY_CLOSED);
+    if (factory->hp + need > factory->hp_end) {
+	erts_reserve_heap__(factory, need, xtra);
+    }
+    res = factory->hp;
+    factory->hp += need;
+    return res;
+}
+
+#endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
 #ifdef CHECK_FOR_HOLES
 # define ERTS_FACTORY_HOLE_CHECK(f) do {    \
         /*if ((f)->p) erts_check_for_holes((f)->p);*/ \
@@ -118,15 +144,16 @@ struct erl_heap_fragment {
 };
 
 /* m[0] = message, m[1] = seq trace token */
-#define ERL_MESSAGE_REF_ARRAY_SZ 2
+#define ERL_MESSAGE_REF_ARRAY_SZ 3
 #define ERL_MESSAGE_TERM(mp) ((mp)->m[0])
 #define ERL_MESSAGE_TOKEN(mp) ((mp)->m[1])
+#define ERL_MESSAGE_FROM(mp) ((mp)->m[2])
 
 #ifdef USE_VM_PROBES
 /* m[2] = dynamic trace user tag */
 #undef ERL_MESSAGE_REF_ARRAY_SZ
-#define ERL_MESSAGE_REF_ARRAY_SZ 3
-#define ERL_MESSAGE_DT_UTAG(mp) ((mp)->m[2])
+#define ERL_MESSAGE_REF_ARRAY_SZ 4
+#define ERL_MESSAGE_DT_UTAG(mp) ((mp)->m[3])
 #else
 #endif
 
@@ -157,30 +184,124 @@ struct erl_mesg {
     ErlHeapFragment hfrag;
 };
 
+/*
+ * The ErtsMessage struct is only one special type
+ * of signal. All signal structs have a common
+ * begining and can be differentiated by looking
+ * at the ErtsSignal 'common.tag' field.
+ *
+ * - An ordinary message will have a value
+ * - A distribution message that has not been
+ *   decoded yet will have the non-value.
+ * - Other signals will have an external pid
+ *   header tag. In order to differentiate
+ *   between those signals one needs to look
+ *   at the arity part of the header (see
+ *   erts_proc_sig_queue.h).
+ */
+
+#define ERTS_SIG_IS_NON_MSG_TAG(Tag) \
+    (is_external_pid_header((Tag)))
+
+#define ERTS_SIG_IS_NON_MSG(Sig) \
+    ERTS_SIG_IS_NON_MSG_TAG(((ErtsSignal *) (Sig))->common.tag)
+
+#define ERTS_SIG_IS_INTERNAL_MSG_TAG(Tag) \
+    (!is_header((Tag)))
+#define ERTS_SIG_IS_INTERNAL_MSG(Sig) \
+    ERTS_SIG_IS_INTERNAL_MSG_TAG(((ErtsSignal *) (Sig))->common.tag)
+
+#define ERTS_SIG_IS_EXTERNAL_MSG_TAG(Tag) \
+    ((Tag) == THE_NON_VALUE)
+#define ERTS_SIG_IS_EXTERNAL_MSG(Sig) \
+    ERTS_SIG_IS_EXTERNAL_MSG_TAG(((ErtsSignal *) (Sig))->common.tag)
+
+#define ERTS_SIG_IS_MSG_TAG(Tag) \
+    (!ERTS_SIG_IS_NON_MSG_TAG(Tag))
+#define ERTS_SIG_IS_MSG(Sig) \
+    ERTS_SIG_IS_MSG_TAG(((ErtsSignal *) (Sig))->common.tag)
+
+typedef union {
+    ErtsSignalCommon common;
+    ErtsMessageRef msg;
+} ErtsSignal;
+
+typedef struct {
+    /* pointers to next pointers pointing to... */
+    ErtsMessage **next; /* ... next (non-message) signal */
+    ErtsMessage **last; /* ... last (non-message) signal */
+} ErtsMsgQNMSigs;
+
 /* Size of default message buffer (erl_message.c) */
 #define ERL_MESSAGE_BUF_SZ 500
 
 typedef struct {
-    ErtsMessage* first;
-    ErtsMessage** last;  /* point to the last next pointer */
-    ErtsMessage** save;
-    Sint len;            /* queue length */
-
     /*
-     * The following two fields are used by the recv_mark/1 and
-     * recv_set/1 instructions.
+     * ** The signal queues private to a process. **
+     *
+     * These are:
+     * - an inner queue which only consists of
+     *   message signals
+     * - a middle queue which contains a mixture
+     *   of message and non-message signals
+     *
+     * When the process isn't processing signals in
+     * erts_proc_sig_handle_incoming():
+     * - the message queue corresponds to the inner
+     *   queue. Messages in the middle queue (and
+     *   in the outer queue) are in transit and
+     *   have NOT been received yet!
+     *
+     * When the process is processing signals in
+     * erts_proc_sig_handle_incoming():
+     * - the message queue corresponds to the inner
+     *   queue plus the head of the middle queue up
+     *   to the signal currently being processed.
+     *   Any messages further back in the middle queue
+     *   (and in the outer queue) are still in transit
+     *   and have NOT been received yet!
+     *
+     * In the general case the 'len' field of this
+     * structure does NOT correspond to the message
+     * queue length. When the process is inspected
+     * via process info it does however correspond
+     * to the message queue length, but this is a
+     * special case!
+     *
+     * When no process-info request is in transit to
+     * the process the 'len' field corresponds to
+     * the total amount of messages in inner and
+     * middle queues (which does NOT correspond to
+     * the message queue length). When process-info
+     * requests are in transit to the process, the
+     * usage of the 'len' field changes and is used
+     * as an offset which even might be negative.
      */
-    BeamInstr* mark;		/* address to rec_loop/2 instruction */
-    ErtsMessage** saved_last;	/* saved last pointer */
-} ErlMessageQueue;
 
-#ifdef ERTS_SMP
+    /* inner queue */
+    ErtsMessage *first;
+    ErtsMessage **last;  /* point to the last next pointer */
+    ErtsMessage **save;
+
+    /* middle queue */
+    ErtsMessage *cont;
+    ErtsMessage **cont_last;
+    ErtsMsgQNMSigs nmsigs;
+
+    /* Common for inner and middle queue */
+    ErtsMessage **saved_last;	/* saved last pointer */
+    Sint len; /* NOT message queue length (see above) */
+} ErtsSignalPrivQueues;
 
 typedef struct {
     ErtsMessage* first;
     ErtsMessage** last;  /* point to the last next pointer */
-    Sint len;            /* queue length */
-} ErlMessageInQueue;
+    Sint len;            /* number of messages in queue */
+    ErtsMsgQNMSigs nmsigs;
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+    int may_contain_heap_terms;
+#endif
+} ErtsSignalInQueue;
 
 typedef struct erl_trace_message_queue__ {
     struct erl_trace_message_queue__ *next; /* point to the next receiver */
@@ -190,10 +311,45 @@ typedef struct erl_trace_message_queue__ {
     Sint len;            /* queue length */
 } ErlTraceMessageQueue;
 
-#endif
+#define ERTS_RECV_MARK_SAVE(P)                                          \
+    do {                                                                \
+        erts_proc_lock((P), ERTS_PROC_LOCK_MSGQ);                       \
+        erts_proc_sig_fetch((P));                                       \
+        erts_proc_unlock((P), ERTS_PROC_LOCK_MSGQ);                     \
+        if ((P)->sig_qs.cont) {                                         \
+            (P)->sig_qs.saved_last = (P)->sig_qs.cont_last;             \
+            (P)->flags |= F_DEFERRED_SAVED_LAST;                        \
+        }                                                               \
+        else {                                                          \
+            (P)->sig_qs.saved_last = (P)->sig_qs.last;                  \
+            (P)->flags &= ~F_DEFERRED_SAVED_LAST;                       \
+        }                                                               \
+    } while (0)
+
+#define ERTS_RECV_MARK_SET(P)                                           \
+    do {                                                                \
+        if ((P)->sig_qs.saved_last) {                                   \
+            if ((P)->flags & F_DEFERRED_SAVED_LAST) {                   \
+                /* Points to middle queue; use end of inner */          \
+                (P)->sig_qs.save = (P)->sig_qs.last;                    \
+                ASSERT(!PEEK_MESSAGE((P)));                             \
+            }                                                           \
+            else {                                                      \
+                /* Points to inner queue; safe to use */                \
+                (P)->sig_qs.save = (P)->sig_qs.saved_last;              \
+            }                                                           \
+        }                                                               \
+    } while (0)
+
+#define ERTS_RECV_MARK_CLEAR(P)                                         \
+    do {                                                                \
+        (P)->sig_qs.saved_last = NULL;                                  \
+        (P)->flags &= ~F_DEFERRED_SAVED_LAST;                           \
+    } while (0)
+
 
 /* Get "current" message */
-#define PEEK_MESSAGE(p)  (*(p)->msg.save)
+#define PEEK_MESSAGE(p)  (*(p)->sig_qs.save)
 
 #ifdef USE_VM_PROBES
 #define LINK_MESSAGE_DTAG(mp, dt) ERL_MESSAGE_DT_UTAG(mp) = dt
@@ -201,65 +357,53 @@ typedef struct erl_trace_message_queue__ {
 #define LINK_MESSAGE_DTAG(mp, dt)
 #endif
 
-#define LINK_MESSAGE_IMPL(p, first_msg, last_msg, num_msgs, where) do { \
-        *(p)->where.last = (first_msg);                                 \
-        (p)->where.last = (last_msg);                                   \
-        (p)->where.len += (num_msgs);                                   \
-    } while(0)
-
-#ifdef ERTS_SMP
-
-/* Add message last in private message queue */
-#define LINK_MESSAGE_PRIVQ(p, first_msg, last_msg, len)                 \
-    do {                                                                \
-        LINK_MESSAGE_IMPL(p, first_msg, last_msg, len, msg);            \
-    } while (0)
-
-/* Add message last_msg in message queue */
-#define LINK_MESSAGE(p, first_msg, last_msg, len)       \
-    LINK_MESSAGE_IMPL(p, first_msg, last_msg, len, msg_inq)
-
-#define ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p)                   \
-    do {                                                \
-        if (p->msg_inq.first) {                         \
-            *p->msg.last = p->msg_inq.first;            \
-            p->msg.last = p->msg_inq.last;              \
-            p->msg.len += p->msg_inq.len;               \
-            p->msg_inq.first = NULL;                    \
-            p->msg_inq.last = &p->msg_inq.first;        \
-            p->msg_inq.len = 0;                         \
-        }                                               \
-    } while (0)
-
+#ifdef USE_VM_PROBES
+#  define ERTS_MSG_RECV_TRACED(P)                                       \
+    ((ERTS_TRACE_FLAGS((P)) & F_TRACE_RECEIVE)                          \
+     || DTRACE_ENABLED(message_queued))
 #else
+#  define ERTS_MSG_RECV_TRACED(P)                                       \
+    (ERTS_TRACE_FLAGS((P)) & F_TRACE_RECEIVE)
 
-#define ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p)
+#endif
 
-/* Add message last_msg in message queue */
-#define LINK_MESSAGE(p, first_msg, last_msg, len)                       \
+/* Add one message last in message queue */
+#define LINK_MESSAGE(p, msg) \
     do {                                                                \
-        LINK_MESSAGE_IMPL(p, first_msg, last_msg, len, msg);            \
+        ASSERT(ERTS_SIG_IS_MSG(msg));                                   \
+        ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE__((p), "before");               \
+        *(p)->sig_inq.last = (msg);                                     \
+        (p)->sig_inq.last = &(msg)->next;                               \
+        (p)->sig_inq.len++;                                             \
+        ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE__((p), "before");               \
     } while(0)
 
-#endif
-
 /* Unlink current message */
-#define UNLINK_MESSAGE(p,msgp) do { \
-     ErtsMessage* __mp = (msgp)->next; \
-     *(p)->msg.save = __mp; \
-     (p)->msg.len--; \
-     if (__mp == NULL) \
-         (p)->msg.last = (p)->msg.save; \
-     (p)->msg.mark = 0; \
-} while(0)
+#define UNLINK_MESSAGE(p,msgp)                                          \
+    do {                                                                \
+        ErtsMessage *mp__ = (msgp)->next;                               \
+        ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE__((p), 0, "before");          \
+        *(p)->sig_qs.save = mp__;                                       \
+        (p)->sig_qs.len--;                                              \
+        if (mp__ == NULL)                                               \
+            (p)->sig_qs.last = (p)->sig_qs.save;                        \
+        ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE__((p), 0, "after");           \
+    } while(0)
 
-/* Reset message save point (after receive match) */
-#define JOIN_MESSAGE(p) \
-     (p)->msg.save = &(p)->msg.first
+/*
+ * Reset message save point (after receive match).
+ * Also invalidate the saved position since it may no
+ * longer be safe to use.
+ */
+#define JOIN_MESSAGE(p)                                                 \
+   do {                                                                 \
+       (p)->sig_qs.save = &(p)->sig_qs.first;                           \
+       ERTS_RECV_MARK_CLEAR((p));                                       \
+   } while(0)
 
 /* Save current message */
 #define SAVE_MESSAGE(p) \
-     (p)->msg.save = &(*(p)->msg.save)->next
+     (p)->sig_qs.save = &(*(p)->sig_qs.save)->next
 
 #define ERTS_SND_FLG_NO_SEQ_TRACE		(((unsigned) 1) << 0)
 
@@ -286,6 +430,7 @@ typedef struct erl_trace_message_queue__ {
         (MP)->next = NULL;                              \
         ERL_MESSAGE_TERM(MP) = THE_NON_VALUE;           \
         ERL_MESSAGE_TOKEN(MP) = NIL;                    \
+        ERL_MESSAGE_FROM(MP) = NIL;                     \
         ERL_MESSAGE_DT_UTAG_INIT(MP);                   \
         MP->data.attached = NULL;                       \
     } while (0)
@@ -296,11 +441,12 @@ ErlHeapFragment* erts_resize_message_buffer(ErlHeapFragment *, Uint,
 					    Eterm *, Uint);
 void free_message_buffer(ErlHeapFragment *);
 void erts_queue_dist_message(Process*, ErtsProcLocks, ErtsDistExternal *, Eterm, Eterm);
-Sint erts_queue_message(Process*, ErtsProcLocks,ErtsMessage*, Eterm, Eterm);
-Sint erts_queue_messages(Process*, ErtsProcLocks,
-                         ErtsMessage*, ErtsMessage**, Uint, Eterm);
+void erts_queue_message(Process*, ErtsProcLocks,ErtsMessage*, Eterm, Eterm);
+void erts_queue_proc_message(Process* from,Process* to, ErtsProcLocks,ErtsMessage*, Eterm);
+void erts_queue_proc_messages(Process* from, Process* to, ErtsProcLocks,
+                              ErtsMessage*, ErtsMessage**, Uint);
 void erts_deliver_exit_message(Eterm, Process*, ErtsProcLocks *, Eterm, Eterm);
-Sint erts_send_message(Process*, Process*, ErtsProcLocks*, Eterm, unsigned);
+void erts_send_message(Process*, Process*, ErtsProcLocks*, Eterm, unsigned);
 void erts_link_mbuf_to_proc(Process *proc, ErlHeapFragment *bp);
 
 Uint erts_msg_attached_data_size_aux(ErtsMessage *msg);
@@ -315,16 +461,6 @@ int erts_decode_dist_message(Process *, ErtsProcLocks, ErtsMessage *, int);
 
 void erts_cleanup_messages(ErtsMessage *mp);
 
-typedef struct {
-    Uint size;
-    ErtsMessage *msgp;
-} ErtsMessageInfo;
-
-Uint erts_prep_msgq_for_inspection(Process *c_p,
-				   Process *rp,
-				   ErtsProcLocks rp_locks,
-				   ErtsMessageInfo *mip);
-
 void *erts_alloc_message_ref(void);
 void erts_free_message_ref(void *);
 
@@ -366,12 +502,6 @@ ERTS_GLB_FORCE_INLINE ErtsMessage *erts_shrink_message(ErtsMessage *mp, Uint sz,
 ERTS_GLB_FORCE_INLINE void erts_free_message(ErtsMessage *mp);
 ERTS_GLB_INLINE Uint erts_used_frag_sz(const ErlHeapFragment*);
 ERTS_GLB_INLINE Uint erts_msg_attached_data_size(ErtsMessage *msg);
-ERTS_GLB_INLINE void erts_msgq_update_internal_pointers(ErlMessageQueue *msgq,
-							ErtsMessage **newpp,
-							ErtsMessage **oldpp);
-ERTS_GLB_INLINE void erts_msgq_replace_msg_ref(ErlMessageQueue *msgq,
-					       ErtsMessage *newp,
-					       ErtsMessage **oldpp);
 
 #define ERTS_MSG_COMBINED_HFRAG ((void *) 0x1)
 
@@ -475,28 +605,6 @@ ERTS_GLB_INLINE Uint erts_msg_attached_data_size(ErtsMessage *msg)
     }
 }
 
-ERTS_GLB_INLINE void
-erts_msgq_update_internal_pointers(ErlMessageQueue *msgq,
-				   ErtsMessage **newpp,
-				   ErtsMessage **oldpp)
-{
-    if (msgq->save == oldpp)
-	msgq->save = newpp;
-    if (msgq->last == oldpp)
-	msgq->last = newpp;
-    if (msgq->saved_last == oldpp)
-	msgq->saved_last = newpp;
-}
-
-ERTS_GLB_INLINE void
-erts_msgq_replace_msg_ref(ErlMessageQueue *msgq, ErtsMessage *newp, ErtsMessage **oldpp)
-{
-    ErtsMessage *oldp = *oldpp;
-    newp->next = oldp->next;
-    erts_msgq_update_internal_pointers(msgq, &newp->next, &oldp->next);
-    *oldpp = newp;
-}
-
 #endif
 
 Uint erts_mbuf_size(Process *p);
@@ -510,4 +618,17 @@ Uint erts_mbuf_size(Process *p);
 #  define ERTS_CHK_MBUF_SZ(P) ((void) 1)
 #endif
 
+#define ERTS_FOREACH_SIG_PRIVQS(PROC, MVAR, CODE)                       \
+    do {                                                                \
+        int i__;                                                        \
+        ErtsMessage *msgs__[] = {(PROC)->sig_qs.first,                  \
+                                 (PROC)->sig_qs.cont};                  \
+        for (i__ = 0; i__ < sizeof(msgs__)/sizeof(msgs__[0]); i__++) {  \
+            ErtsMessage *MVAR;                                          \
+            for (MVAR = msgs__[i__]; MVAR; MVAR = MVAR->next) {         \
+                CODE;                                                   \
+            }                                                           \
+        }                                                               \
+    } while (0)
+
 #endif
diff --git a/erts/emulator/beam/erl_monitor_link.c b/erts/emulator/beam/erl_monitor_link.c
new file mode 100644
index 0000000000..48d9bd4ca5
--- /dev/null
+++ b/erts/emulator/beam/erl_monitor_link.c
@@ -0,0 +1,1326 @@
+/*
+ * %CopyrightBegin%
+ * 
+ * Copyright Ericsson AB 2018. All Rights Reserved.
+ * 
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * 
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description:	Monitor and link implementation.
+ *
+ * Author: 	Rickard Green
+ */
+
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#include "sys.h"
+#include <stddef.h>
+#include "global.h"
+#include "erl_node_tables.h"
+#include "erl_monitor_link.h"
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Red-black tree implementation used for monitors and links.                *
+\*                                                                           */
+
+static ERTS_INLINE Eterm
+ml_get_key(ErtsMonLnkNode *mln)
+{
+    char *ptr = (char *) mln;
+    ptr += mln->key_offset;
+
+#ifdef ERTS_ML_DEBUG
+    switch (mln->type) {
+    case ERTS_MON_TYPE_PROC:
+    case ERTS_MON_TYPE_PORT:
+    case ERTS_MON_TYPE_DIST_PROC:
+    case ERTS_MON_TYPE_TIME_OFFSET:
+    case ERTS_MON_TYPE_RESOURCE: {
+        ErtsMonitorData *mdp = erts_monitor_to_data(mln);
+        ERTS_ML_ASSERT(&mdp->ref == (Eterm *) ptr);
+        break;
+    }
+    case ERTS_LNK_TYPE_PROC:
+    case ERTS_LNK_TYPE_DIST_PROC:
+    case ERTS_LNK_TYPE_PORT:
+    case ERTS_MON_TYPE_NODE:
+    case ERTS_MON_TYPE_NODES:
+    case ERTS_MON_TYPE_SUSPEND:
+        ERTS_ML_ASSERT(&mln->other.item == (Eterm *) ptr);
+        break;
+    default:
+        ERTS_INTERNAL_ERROR("Invalid type");
+        break;
+    }
+#endif
+
+    return *((Eterm *) ptr);
+}
+
+/*
+ * Comparison functions for valid *and only valid* keys. That
+ * is, if the set of valid keys is changed, this function needs
+ * to be updated...
+ *
+ * Note that this function does *not* order terms according to
+ * term order, and that the order may vary between emulator
+ * restarts...
+ */
+static int
+ml_cmp_keys(Eterm key1, Eterm key2)
+{
+    /*
+     * A monitor key is either an internal pid, a (nodename) atom,
+     * a small (monitor nodes bitmask), an ordinary internal ref,
+     * or an external ref.
+     *
+     * Order between monitors with keys of different types:
+     *  internal pid < atom < small < internal ref < external ref
+     *
+     * A link key is either a pid, an internal port
+     *
+     * Order between links with keys of different types:
+     *  internal pid < internal port < external pid
+     *
+     */
+    ERTS_ML_ASSERT(is_internal_pid(key1)
+                   || is_internal_port(key1)
+                   || is_atom(key1)
+                   || is_small(key1)
+                   || is_external_pid(key1)
+                   || is_internal_ordinary_ref(key1)
+                   || is_external_ref(key1));
+
+    ERTS_ML_ASSERT(is_internal_pid(key2)
+                   || is_internal_port(key2)
+                   || is_atom(key2)
+                   || is_small(key2)
+                   || is_external_pid(key2)
+                   || is_internal_ordinary_ref(key2)
+                   || is_external_ref(key2));
+
+    if (is_immed(key1)) {
+        int key1_tag, key2_tag;
+
+        ERTS_ML_ASSERT(is_internal_pid(key1)
+                       || is_internal_port(key1)
+                       || is_atom(key1)
+                       || is_small(key1));
+
+        if (key1 == key2)
+            return 0;
+
+        if (is_boxed(key2))
+            return -1;
+
+        ERTS_ML_ASSERT(is_internal_pid(key2)
+                       || is_internal_port(key2)
+                       || is_atom(key2)
+                       || is_small(key2));
+
+        key1_tag = (int) (key1 & _TAG_IMMED1_MASK);
+        key2_tag = (int) (key2 & _TAG_IMMED1_MASK);
+
+        if (key1_tag != key2_tag)
+            return key1_tag - key2_tag;
+
+        ASSERT((is_atom(key1) && is_atom(key2))
+               || (is_small(key1) && is_small(key2))
+               || (is_internal_pid(key1) && is_internal_pid(key2))
+               || (is_internal_port(key1) && is_internal_port(key2)));
+
+        return key1 < key2 ? -1 : 1;
+    }
+    else {
+        Eterm *w1, hdr1;
+
+        ERTS_ML_ASSERT(is_boxed(key1));
+
+        w1 = boxed_val(key1);
+        hdr1 = *w1;
+
+        if ((hdr1 & _TAG_HEADER_MASK) == _TAG_HEADER_REF) {
+            Eterm *w2;
+
+            if (!is_internal_ref(key2))
+                return is_immed(key2) ? 1 : -1;
+
+            w2 = internal_ref_val(key2);
+
+            ERTS_ML_ASSERT(w1[0] == ERTS_REF_THING_HEADER);
+            ERTS_ML_ASSERT(w2[0] == ERTS_REF_THING_HEADER);
+
+            return sys_memcmp((void *) &w1[1], (void *) &w2[1],
+                              (ERTS_REF_THING_SIZE - 1)*sizeof(Eterm));
+        }
+
+        ERTS_ML_ASSERT(is_external(key1));
+
+        if (is_not_external(key2))
+            return 1;
+        else {
+            Uint ndw1, ndw2;
+            ExternalThing *et1, *et2;
+            ErlNode *n1, *n2;
+
+            ERTS_ML_ASSERT((is_external_ref(key1) && is_external_ref(key2))
+                           || (is_external_pid(key1) && is_external_pid(key2)));
+
+            et1 = (ExternalThing *) w1;
+            et2 = (ExternalThing *) external_val(key2);
+
+            n1 = et1->node;
+            n2 = et2->node;
+
+            if (n1 != n2) {
+                if (n1->sysname != n2->sysname)
+                    return n1->sysname < n2->sysname ? -1 : 1;
+                ASSERT(n1->creation != n2->creation);
+                return n1->creation < n2->creation ? -1 : 1;
+            }
+
+            ndw1 = external_thing_data_words(et1);
+            ndw2 = external_thing_data_words(et1);
+            if (ndw1 != ndw2)
+                return ndw1 < ndw2 ? -1 : 1;
+
+            return sys_memcmp((void *) &et1->data.ui[0],
+                              (void *) &et2->data.ui[0],
+                              ndw1*sizeof(Eterm));
+        }
+    }
+}
+
+#define ERTS_ML_TPFLG_RED               (((UWord) 1) << 0)
+
+#define ERTS_ML_TPFLGS_MASK             (ERTS_ML_TPFLG_RED)
+
+#define ERTS_RBT_PREFIX mon_lnk
+#define ERTS_RBT_T ErtsMonLnkNode
+#define ERTS_RBT_KEY_T Eterm
+#define ERTS_RBT_FLAGS_T UWord
+#define ERTS_RBT_INIT_EMPTY_TNODE(T)					\
+    do {								\
+	(T)->node.tree.parent = (UWord) NULL;                           \
+	(T)->node.tree.right = NULL;                                    \
+	(T)->node.tree.left = NULL;                                     \
+    } while (0)
+#define ERTS_RBT_IS_RED(T)						\
+    (!!((T)->node.tree.parent & ERTS_ML_TPFLG_RED))
+#define ERTS_RBT_SET_RED(T)						\
+    ((T)->node.tree.parent |= ERTS_ML_TPFLG_RED)
+#define ERTS_RBT_IS_BLACK(T)						\
+    (!ERTS_RBT_IS_RED((T)))
+#define ERTS_RBT_SET_BLACK(T)						\
+    ((T)->node.tree.parent &= ~ERTS_ML_TPFLG_RED)
+#define ERTS_RBT_GET_FLAGS(T)						\
+    ((T)->node.tree.parent & ERTS_ML_TPFLGS_MASK)
+#define ERTS_RBT_SET_FLAGS(T, F)					\
+    do {								\
+	ERTS_ML_ASSERT((((UWord) (F)) & ~ERTS_ML_TPFLGS_MASK) == 0);	\
+	(T)->node.tree.parent &= ~ERTS_ML_TPFLGS_MASK;                 \
+	(T)->node.tree.parent |= (F);                                   \
+    } while (0)
+#define ERTS_RBT_GET_PARENT(T)						\
+    ((ERTS_RBT_T *) ((T)->node.tree.parent & ~ERTS_ML_TPFLGS_MASK))
+#define ERTS_RBT_SET_PARENT(T, P)					\
+    do {								\
+	ERTS_ML_ASSERT((((UWord) (P)) & ERTS_ML_TPFLGS_MASK) == 0);	\
+	(T)->node.tree.parent &= ERTS_ML_TPFLGS_MASK;                  \
+	(T)->node.tree.parent |= (UWord) (P);                           \
+    } while (0)
+#define ERTS_RBT_GET_RIGHT(T) ((T)->node.tree.right)
+#define ERTS_RBT_SET_RIGHT(T, R) ((T)->node.tree.right = (R))
+#define ERTS_RBT_GET_LEFT(T) ((T)->node.tree.left)
+#define ERTS_RBT_SET_LEFT(T, L) ((T)->node.tree.left = (L))
+#define ERTS_RBT_GET_KEY(T) (ml_get_key((T)))
+#define ERTS_RBT_CMP_KEYS(KX, KY) (ml_cmp_keys((KX), (KY)))
+#define ERTS_RBT_WANT_DELETE
+#define ERTS_RBT_WANT_LOOKUP
+#define ERTS_RBT_WANT_LOOKUP_INSERT
+#define ERTS_RBT_WANT_LOOKUP_CREATE
+#define ERTS_RBT_WANT_INSERT
+#define ERTS_RBT_WANT_REPLACE
+#define ERTS_RBT_WANT_FOREACH
+#define ERTS_RBT_WANT_FOREACH_YIELDING
+#define ERTS_RBT_WANT_FOREACH_DESTROY
+#define ERTS_RBT_WANT_FOREACH_DESTROY_YIELDING
+#define ERTS_RBT_UNDEF
+
+#include "erl_rbtree.h"
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Tree Operations                                                           *
+\*                                                                           */
+
+static ErtsMonLnkNode *
+ml_rbt_lookup(ErtsMonLnkNode *root, Eterm ref)
+{
+    ErtsMonLnkNode *ml = mon_lnk_rbt_lookup(root, ref);
+    ASSERT(!ml || (ml->flags & ERTS_ML_FLG_IN_TABLE));
+    return ml;
+}
+
+static ErtsMonLnkNode *
+ml_rbt_lookup_insert(ErtsMonLnkNode **root, ErtsMonLnkNode *ml)
+{
+    ErtsMonLnkNode *res;
+    ERTS_ML_ASSERT(!(ml->flags & ERTS_ML_FLG_IN_TABLE));
+    res = mon_lnk_rbt_lookup_insert(root, ml);
+    if (!res)
+        ml->flags |= ERTS_ML_FLG_IN_TABLE;
+    return res;
+}
+
+static ErtsMonLnkNode *
+ml_rbt_lookup_create(ErtsMonLnkNode ** root, Eterm key,
+                     ErtsMonLnkNode *(*create)(Eterm, void *),
+                     void *carg, int *created)
+{
+    ErtsMonLnkNode *ml;
+    ml = mon_lnk_rbt_lookup_create(root, key, create, carg, created);
+    if (*created)
+        ml->flags |= ERTS_ML_FLG_IN_TABLE;
+    return ml;
+}
+
+static void
+ml_rbt_insert(ErtsMonLnkNode **root, ErtsMonLnkNode *ml)
+{
+#ifdef ERTS_ML_DEBUG
+    ErtsMonLnkNode *res;
+    ERTS_ML_ASSERT(!(ml->flags & ERTS_ML_FLG_IN_TABLE));
+    res = ml_rbt_lookup(*root, ml_get_key(ml));
+    ERTS_ML_ASSERT(res == NULL);
+#endif
+
+    mon_lnk_rbt_insert(root, ml);
+    ml->flags |= ERTS_ML_FLG_IN_TABLE;
+}
+
+static void
+ml_rbt_replace(ErtsMonLnkNode **root, ErtsMonLnkNode *old, ErtsMonLnkNode *new)
+{
+    ERTS_ML_ASSERT(old->flags & ERTS_ML_FLG_IN_TABLE);
+    ERTS_ML_ASSERT(!(new->flags & ERTS_ML_FLG_IN_TABLE));
+
+    mon_lnk_rbt_replace(root, old, new);
+    old->flags &= ~ERTS_ML_FLG_IN_TABLE;
+    new->flags |= ERTS_ML_FLG_IN_TABLE;
+}
+
+static void
+ml_rbt_delete(ErtsMonLnkNode **root, ErtsMonLnkNode *ml)
+{
+    ERTS_ML_ASSERT(ml->flags & ERTS_ML_FLG_IN_TABLE);
+    mon_lnk_rbt_delete(root, ml);
+    ml->flags &= ~ERTS_ML_FLG_IN_TABLE;
+}
+
+
+static void
+ml_rbt_foreach(ErtsMonLnkNode *root,
+               void (*func)(ErtsMonLnkNode *, void *),
+               void *arg)
+{
+    mon_lnk_rbt_foreach(root, func, arg);
+}
+
+typedef struct {
+    ErtsMonLnkNode *root;
+    mon_lnk_rbt_yield_state_t rbt_ystate;
+} ErtsMonLnkYieldState;
+
+static int
+ml_rbt_foreach_yielding(ErtsMonLnkNode *root,
+                        void (*func)(ErtsMonLnkNode *, void *),
+                        void *arg,
+                        void **vyspp,
+                        Sint limit)
+{
+    int res;
+    ErtsMonLnkYieldState ys = {root, ERTS_RBT_YIELD_STAT_INITER};
+    ErtsMonLnkYieldState *ysp;
+
+    ysp = (ErtsMonLnkYieldState *) *vyspp;
+    if (!ysp)
+	ysp = &ys;
+    res = mon_lnk_rbt_foreach_yielding(ysp->root, func, arg,
+                                       &ysp->rbt_ystate, limit);
+    if (res == 0) {
+	if (ysp != &ys)
+	    erts_free(ERTS_ALC_T_ML_YIELD_STATE, ysp);
+	*vyspp = NULL;
+    }
+    else {
+
+	if (ysp == &ys) {
+	    ysp = erts_alloc(ERTS_ALC_T_ML_YIELD_STATE,
+			     sizeof(ErtsMonLnkYieldState));
+	    sys_memcpy((void *) ysp, (void *) &ys,
+		       sizeof(ErtsMonLnkYieldState));
+	}
+
+	*vyspp = (void *) ysp;
+    }
+
+    return res;
+}
+
+typedef struct {
+    void (*func)(ErtsMonLnkNode *, void *);
+    void *arg;
+} ErtsMonLnkForeachDeleteContext;
+
+static void
+rbt_wrap_foreach_delete(ErtsMonLnkNode *ml, void *vctxt)
+{
+    ErtsMonLnkForeachDeleteContext *ctxt = vctxt;
+    ERTS_ML_ASSERT(ml->flags & ERTS_ML_FLG_IN_TABLE);
+    ml->flags &= ~ERTS_ML_FLG_IN_TABLE;
+    ctxt->func(ml, ctxt->arg);
+}
+
+static void
+ml_rbt_foreach_delete(ErtsMonLnkNode **root,
+                      void (*func)(ErtsMonLnkNode *, void *),
+                      void *arg)
+{
+    ErtsMonLnkForeachDeleteContext ctxt;
+    ctxt.func = func;
+    ctxt.arg = arg;
+    mon_lnk_rbt_foreach_destroy(root,
+                                rbt_wrap_foreach_delete,
+                                (void *) &ctxt);
+}
+
+static int
+ml_rbt_foreach_delete_yielding(ErtsMonLnkNode **root,
+                               void (*func)(ErtsMonLnkNode *, void *),
+                               void *arg,
+                               void **vyspp,
+                               Sint limit)
+{
+    int res;
+    ErtsMonLnkYieldState ys = {*root, ERTS_RBT_YIELD_STAT_INITER};
+    ErtsMonLnkYieldState *ysp;
+    ErtsMonLnkForeachDeleteContext ctxt;
+    ctxt.func = func;
+    ctxt.arg = arg;
+
+    ysp = (ErtsMonLnkYieldState *) *vyspp;
+    if (!ysp) {
+        *root = NULL;
+	ysp = &ys;
+    }
+    res = mon_lnk_rbt_foreach_destroy_yielding(&ysp->root,
+                                               rbt_wrap_foreach_delete,
+                                               (void *) &ctxt,
+                                               &ysp->rbt_ystate,
+                                               limit);
+    if (res == 0) {
+	if (ysp != &ys)
+	    erts_free(ERTS_ALC_T_ML_YIELD_STATE, ysp);
+	*vyspp = NULL;
+    }
+    else {
+
+	if (ysp == &ys) {
+	    ysp = erts_alloc(ERTS_ALC_T_ML_YIELD_STATE,
+			     sizeof(ErtsMonLnkYieldState));
+	    sys_memcpy((void *) ysp, (void *) &ys,
+		       sizeof(ErtsMonLnkYieldState));
+	}
+
+	*vyspp = (void *) ysp;
+    }
+
+    return res;
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * List Operations                                                           *
+\*                                                                           */
+
+static int
+ml_dl_list_foreach_yielding(ErtsMonLnkNode *list,
+                            void (*func)(ErtsMonLnkNode *, void *),
+                            void *arg,
+                            void **vyspp,
+                            Sint limit)
+{
+    Sint cnt = 0;
+    ErtsMonLnkNode *ml = (ErtsMonLnkNode *) *vyspp;
+
+    ERTS_ML_ASSERT(!ml || list);
+
+    if (!ml)
+        ml = list;
+
+    if (ml) {
+        do {
+            ERTS_ML_ASSERT(ml->flags & ERTS_ML_FLG_IN_TABLE);
+            func(ml, arg);
+            ml = ml->node.list.next;
+            cnt++;
+        } while (ml != list && cnt < limit);
+        if (ml != list) {
+            *vyspp = (void *) ml;
+            return 1; /* yield */
+        }
+    }
+
+    *vyspp = NULL;
+    return 0; /* done */
+}
+
+static int
+ml_dl_list_foreach_delete_yielding(ErtsMonLnkNode **list,
+                                   void (*func)(ErtsMonLnkNode *, void *),
+                                   void *arg,
+                                   void **vyspp,
+                                   Sint limit)
+{
+    Sint cnt = 0;
+    ErtsMonLnkNode *first = *list;
+    ErtsMonLnkNode *ml = (ErtsMonLnkNode *) *vyspp;
+
+    ERTS_ML_ASSERT(!ml || first);
+
+    if (!ml)
+        ml = first;
+
+    if (ml) {
+        do {
+            ErtsMonLnkNode *next = ml->node.list.next;
+            ERTS_ML_ASSERT(ml->flags & ERTS_ML_FLG_IN_TABLE);
+            ml->flags &= ~ERTS_ML_FLG_IN_TABLE;
+            func(ml, arg);
+            ml = next;
+            cnt++;
+        } while (ml != first && cnt < limit);
+        if (ml != first) {
+            *vyspp = (void *) ml;
+            return 1; /* yield */
+        }
+    }
+
+    *vyspp = NULL;
+    *list = NULL;
+    return 0; /* done */
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Misc                                                                      *
+\*                                                                           */
+
+ErtsMonLnkDist *
+erts_mon_link_dist_create(Eterm nodename)
+{
+    ErtsMonLnkDist *mld = erts_alloc(ERTS_ALC_T_ML_DIST,
+                                     sizeof(ErtsMonLnkDist));
+    mld->nodename = nodename;
+    mld->connection_id = ~((Uint32) 0);
+    erts_atomic_init_nob(&mld->refc, 1);
+    erts_mtx_init(&mld->mtx, "dist_entry_links", nodename,
+                  ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
+    mld->alive = !0;
+    mld->links = NULL;
+    mld->monitors = NULL;
+    mld->orig_name_monitors = NULL;
+    return mld;
+}
+
+void
+erts_mon_link_dist_destroy__(ErtsMonLnkDist *mld)
+{
+    ERTS_ML_ASSERT(erts_atomic_read_nob(&mld->refc) == 0);
+    ERTS_ML_ASSERT(!mld->alive);
+    ERTS_ML_ASSERT(!mld->links);
+    ERTS_ML_ASSERT(!mld->monitors);
+    ERTS_ML_ASSERT(!mld->orig_name_monitors);
+
+    erts_mtx_destroy(&mld->mtx);
+    erts_free(ERTS_ALC_T_ML_DIST, mld);
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Monitor Operations                                                        *
+\*                                                                           */
+
+#ifdef ERTS_ML_DEBUG
+size_t erts_monitor_origin_offset;
+size_t erts_monitor_origin_key_offset;
+size_t erts_monitor_target_offset;
+size_t erts_monitor_target_key_offset;
+size_t erts_monitor_node_key_offset;
+#endif
+
+static ERTS_INLINE void
+monitor_init(void)
+{
+#ifdef ERTS_ML_DEBUG
+    erts_monitor_origin_offset = offsetof(ErtsMonitorData, origin);
+    erts_monitor_origin_key_offset = offsetof(ErtsMonitorData, ref);
+    ASSERT(erts_monitor_origin_key_offset >= erts_monitor_origin_offset);
+    erts_monitor_origin_key_offset -= erts_monitor_origin_offset;
+    erts_monitor_target_offset = offsetof(ErtsMonitorData, target);
+    erts_monitor_target_key_offset = offsetof(ErtsMonitorData, ref);
+    ASSERT(erts_monitor_target_key_offset >= erts_monitor_target_offset);
+    erts_monitor_target_key_offset -= erts_monitor_target_offset;
+    erts_monitor_node_key_offset = offsetof(ErtsMonitor, other.item);
+#endif
+}
+
+ErtsMonitor *
+erts_monitor_tree_lookup(ErtsMonitor *root, Eterm key)
+{
+    ERTS_ML_ASSERT(is_internal_ordinary_ref(key)
+                   || is_external_ref(key)
+                   || is_atom(key)
+                   || is_small(key)
+                   || is_internal_pid(key));
+    return (ErtsMonitor *) ml_rbt_lookup((ErtsMonLnkNode *) root, key);
+}
+
+ErtsMonitor *
+erts_monotor_tree_lookup_insert(ErtsMonitor **root, ErtsMonitor *mon)
+{
+    return (ErtsMonitor *) ml_rbt_lookup_insert((ErtsMonLnkNode **) root,
+                                                (ErtsMonLnkNode *) mon);
+}
+
+typedef struct {
+    Uint16 type;
+    Eterm origin;
+} ErtsMonitorCreateCtxt;
+
+static ErtsMonLnkNode *
+create_monitor(Eterm target, void *vcctxt)
+{
+    ErtsMonitorCreateCtxt *cctxt = vcctxt;
+    ErtsMonitorData *mdp = erts_monitor_create(cctxt->type,
+                                               NIL,
+                                               cctxt->origin,
+                                               target,
+                                               NIL);
+    ERTS_ML_ASSERT(ml_cmp_keys(ml_get_key(&mdp->origin), target) == 0);
+    return (ErtsMonLnkNode *) &mdp->origin;
+}
+
+ErtsMonitor *
+erts_monitor_tree_lookup_create(ErtsMonitor **root, int *created, Uint16 type,
+                                Eterm origin, Eterm target)
+{
+    ErtsMonitor *res;
+    ErtsMonitorCreateCtxt cctxt = {type, origin};
+
+    ERTS_ML_ASSERT(type == ERTS_MON_TYPE_NODE
+                   || type == ERTS_MON_TYPE_NODES
+                   || type == ERTS_MON_TYPE_SUSPEND);
+
+    res = (ErtsMonitor *) ml_rbt_lookup_create((ErtsMonLnkNode **) root,
+                                               target, create_monitor,
+                                               (void *) &cctxt,
+                                               created);
+
+    ERTS_ML_ASSERT(res && erts_monitor_is_origin(res));
+
+    return res;
+}
+
+void
+erts_monitor_tree_insert(ErtsMonitor **root, ErtsMonitor *mon)
+{
+    ml_rbt_insert((ErtsMonLnkNode **) root, (ErtsMonLnkNode *) mon);
+}
+
+void
+erts_monitor_tree_replace(ErtsMonitor **root, ErtsMonitor *old, ErtsMonitor *new)
+{
+    ml_rbt_replace((ErtsMonLnkNode **) root,
+                   (ErtsMonLnkNode *) old,
+                   (ErtsMonLnkNode *) new);
+}
+
+void
+erts_monitor_tree_delete(ErtsMonitor **root, ErtsMonitor *mon)
+{
+    ml_rbt_delete((ErtsMonLnkNode **) root, (ErtsMonLnkNode *) mon);
+}
+
+void
+erts_monitor_tree_foreach(ErtsMonitor *root,
+                          void (*func)(ErtsMonitor *, void *),
+                          void *arg)
+{
+    ml_rbt_foreach((ErtsMonLnkNode *) root,
+                   (void (*)(ErtsMonLnkNode*, void*)) func,
+                   arg);
+}
+
+int
+erts_monitor_tree_foreach_yielding(ErtsMonitor *root,
+                                   void (*func)(ErtsMonitor *, void *),
+                                   void *arg,
+                                   void **vyspp,
+                                   Sint limit)
+{
+    return ml_rbt_foreach_yielding((ErtsMonLnkNode *) root,
+                                   (void (*)(ErtsMonLnkNode*, void*)) func,
+                                   arg, vyspp, limit);
+}
+
+void
+erts_monitor_tree_foreach_delete(ErtsMonitor **root,
+                                 void (*func)(ErtsMonitor *, void *),
+                                 void *arg)
+{
+    ml_rbt_foreach_delete((ErtsMonLnkNode **) root,
+                          (void (*)(ErtsMonLnkNode*, void*)) func,
+                          arg);
+}
+
+int
+erts_monitor_tree_foreach_delete_yielding(ErtsMonitor **root,
+                                          void (*func)(ErtsMonitor *, void *),
+                                          void *arg,
+                                          void **vyspp,
+                                          Sint limit)
+{
+    return ml_rbt_foreach_delete_yielding((ErtsMonLnkNode **) root,
+                                          (void (*)(ErtsMonLnkNode*, void*)) func,
+                                          arg, vyspp, limit);
+}
+
+void
+erts_monitor_list_foreach(ErtsMonitor *list,
+                          void (*func)(ErtsMonitor *, void *),
+                          void *arg)
+{
+    void *ystate = NULL;
+    while (ml_dl_list_foreach_yielding((ErtsMonLnkNode *) list,
+                                       (void (*)(ErtsMonLnkNode *, void *)) func,
+                                       arg, &ystate, (Sint) INT_MAX));
+}
+
+int
+erts_monitor_list_foreach_yielding(ErtsMonitor *list,
+                                   void (*func)(ErtsMonitor *, void *),
+                                   void *arg,
+                                   void **vyspp,
+                                   Sint limit)
+{
+    return ml_dl_list_foreach_yielding((ErtsMonLnkNode *) list,
+                                       (void (*)(ErtsMonLnkNode *, void *)) func,
+                                       arg, vyspp, limit);
+}
+
+void
+erts_monitor_list_foreach_delete(ErtsMonitor **list,
+                                 void (*func)(ErtsMonitor *, void *),
+                                 void *arg)
+{
+    void *ystate = NULL;
+    while (ml_dl_list_foreach_delete_yielding((ErtsMonLnkNode **) list,
+                                              (void (*)(ErtsMonLnkNode*, void*)) func,
+                                              arg, &ystate, (Sint) INT_MAX));
+}
+
+int
+erts_monitor_list_foreach_delete_yielding(ErtsMonitor **list,
+                                          void (*func)(ErtsMonitor *, void *),
+                                          void *arg,
+                                          void **vyspp,
+                                          Sint limit)
+{
+    return ml_dl_list_foreach_delete_yielding((ErtsMonLnkNode **) list,
+                                              (void (*)(ErtsMonLnkNode*, void*)) func,
+                                              arg, vyspp, limit);
+}
+
+ErtsMonitorData *
+erts_monitor_create(Uint16 type, Eterm ref, Eterm orgn, Eterm trgt, Eterm name)
+{
+    ErtsMonitorData *mdp;
+
+    switch (type) {
+    case ERTS_MON_TYPE_PROC:
+    case ERTS_MON_TYPE_PORT:
+        if (is_nil(name)) {
+            ErtsMonitorDataHeap *mdhp;
+            ErtsORefThing *ortp;
+
+        case ERTS_MON_TYPE_TIME_OFFSET:
+
+            ERTS_ML_ASSERT(is_nil(name));
+            ERTS_ML_ASSERT(is_immed(orgn) && is_immed(trgt));
+            ERTS_ML_ASSERT(is_internal_ordinary_ref(ref));
+        
+            mdhp = erts_alloc(ERTS_ALC_T_MONITOR, sizeof(ErtsMonitorDataHeap));
+            mdp = &mdhp->md;
+            ERTS_ML_ASSERT(((void *) mdp) == ((void *) mdhp));
+
+            ortp = (ErtsORefThing *) (char *) internal_ref_val(ref);
+            mdhp->oref_thing = *ortp;
+            mdp->ref = make_internal_ref(&mdhp->oref_thing.header);
+
+            mdp->origin.other.item = trgt;
+            mdp->origin.offset = (Uint16) offsetof(ErtsMonitorData, origin);
+            mdp->origin.key_offset = (Uint16) offsetof(ErtsMonitorData, ref);
+            ERTS_ML_ASSERT(mdp->origin.key_offset >= mdp->origin.offset);
+            mdp->origin.key_offset -= mdp->origin.offset;
+            mdp->origin.flags = (Uint16) 0;
+            mdp->origin.type = type;
+
+            mdp->target.other.item = orgn;
+            mdp->target.offset = (Uint16) offsetof(ErtsMonitorData, target);
+            mdp->target.key_offset = (Uint16) offsetof(ErtsMonitorData, ref);
+            ERTS_ML_ASSERT(mdp->target.key_offset >= mdp->target.offset);
+            mdp->target.key_offset -= mdp->target.offset;
+            mdp->target.flags = ERTS_ML_FLG_TARGET;
+            mdp->target.type = type;
+            break;
+        }
+    case ERTS_MON_TYPE_DIST_PROC:
+    case ERTS_MON_TYPE_RESOURCE:
+    case ERTS_MON_TYPE_NODE:
+    case ERTS_MON_TYPE_NODES: {
+        ErtsMonitorDataExtended *mdep;
+        Uint size = sizeof(ErtsMonitorDataExtended) - sizeof(Eterm);
+        Uint rsz, osz, tsz;
+        Eterm *hp;
+        ErlOffHeap oh;
+        Uint16 name_flag = is_nil(name) ? ((Uint16) 0) : ERTS_ML_FLG_NAME;
+
+        rsz = is_immed(ref) ? 0 : size_object(ref);
+        tsz = is_immed(trgt) ? 0 : size_object(trgt);
+        if (type == ERTS_MON_TYPE_RESOURCE)
+            osz = 0;
+        else
+            osz = is_immed(orgn) ? 0 : size_object(orgn);
+
+        size += (rsz + osz + tsz) * sizeof(Eterm);
+
+        mdep = erts_alloc(ERTS_ALC_T_MONITOR_EXT, size);
+
+        ERTS_INIT_OFF_HEAP(&oh);
+
+        hp = &mdep->heap[0];
+
+        mdp = &mdep->md;
+        ERTS_ML_ASSERT(((void *) mdp) == ((void *) mdep));
+
+        mdp->ref = rsz ? copy_struct(ref, rsz, &hp, &oh) : ref;
+
+        mdp->origin.other.item = tsz ? copy_struct(trgt, tsz, &hp, &oh) : trgt;
+        mdp->origin.offset = (Uint16) offsetof(ErtsMonitorData, origin);
+        mdp->origin.flags = ERTS_ML_FLG_EXTENDED|name_flag;
+        mdp->origin.type = type;
+
+        if (type == ERTS_MON_TYPE_RESOURCE)
+            mdp->target.other.ptr = (void *) orgn;
+        else
+            mdp->target.other.item = osz ? copy_struct(orgn, osz, &hp, &oh) : orgn;
+        mdp->target.offset = (Uint16) offsetof(ErtsMonitorData, target);
+        mdp->target.flags = ERTS_ML_FLG_TARGET|ERTS_ML_FLG_EXTENDED|name_flag;
+        mdp->target.type = type;
+
+        if (type == ERTS_MON_TYPE_NODE || type == ERTS_MON_TYPE_NODES) {
+            mdep->u.refc = 0;
+            mdp->origin.key_offset = (Uint16) offsetof(ErtsMonitor, other.item);
+            mdp->target.key_offset = (Uint16) offsetof(ErtsMonitor, other.item);
+            ERTS_ML_ASSERT(!oh.first);
+            mdep->uptr.node_monitors = NULL;
+        }
+        else {
+            mdep->u.name = name;
+
+            mdp->origin.key_offset = (Uint16) offsetof(ErtsMonitorData, ref);
+            ERTS_ML_ASSERT(mdp->origin.key_offset >= mdp->origin.offset);
+            mdp->origin.key_offset -= mdp->origin.offset;
+
+            mdp->target.key_offset = (Uint16) offsetof(ErtsMonitorData, ref);
+            ERTS_ML_ASSERT(mdp->target.key_offset >= mdp->target.offset);
+            mdp->target.key_offset -= mdp->target.offset;
+
+            mdep->uptr.ohhp = oh.first;
+        }
+        mdep->dist = NULL;
+        break;
+    }
+    case ERTS_MON_TYPE_SUSPEND: {
+        ErtsMonitorSuspend *msp;
+
+        ERTS_ML_ASSERT(is_nil(name));
+        ERTS_ML_ASSERT(is_nil(ref));
+        ERTS_ML_ASSERT(is_internal_pid(orgn) && is_internal_pid(trgt));
+
+        msp = erts_alloc(ERTS_ALC_T_MONITOR_SUSPEND,
+                         sizeof(ErtsMonitorSuspend));
+        mdp = &msp->md;
+        ERTS_ML_ASSERT(((void *) mdp) == ((void *) msp));
+
+        mdp->ref = NIL;
+
+        mdp->origin.other.item = trgt;
+        mdp->origin.offset = (Uint16) offsetof(ErtsMonitorData, origin);
+        mdp->origin.key_offset = (Uint16) offsetof(ErtsMonitor, other.item);
+        ERTS_ML_ASSERT(mdp->origin.key_offset >= mdp->origin.offset);
+        mdp->origin.flags = (Uint16) ERTS_ML_FLG_EXTENDED;
+        mdp->origin.type = type;
+
+        mdp->target.other.item = orgn;
+        mdp->target.offset = (Uint16) offsetof(ErtsMonitorData, target);
+        mdp->target.key_offset = (Uint16) offsetof(ErtsMonitor, other.item);
+        mdp->target.flags = ERTS_ML_FLG_TARGET|ERTS_ML_FLG_EXTENDED;
+        mdp->target.type = type;
+
+        msp->next = NULL;
+        erts_atomic_init_relb(&msp->state, 0);
+
+        break;
+    }
+    default:
+        ERTS_INTERNAL_ERROR("Invalid monitor type");
+        mdp = NULL;
+        break;
+    }
+
+    erts_atomic32_init_nob(&mdp->refc, 2);
+
+    return mdp;
+}
+
+/*
+ * erts_monitor_destroy__() should only be called from
+ * erts_monitor_release() or erts_monitor_release_both().
+ */
+void
+erts_monitor_destroy__(ErtsMonitorData *mdp)
+{
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&mdp->refc) == 0);
+    ERTS_ML_ASSERT(!(mdp->origin.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(!(mdp->target.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT((mdp->origin.flags & ERTS_ML_FLGS_SAME)
+                   == (mdp->target.flags & ERTS_ML_FLGS_SAME));
+
+    if (!(mdp->origin.flags & ERTS_ML_FLG_EXTENDED))
+        erts_free(ERTS_ALC_T_MONITOR, mdp);
+    else if (mdp->origin.type == ERTS_MON_TYPE_SUSPEND)
+        erts_free(ERTS_ALC_T_MONITOR_SUSPEND, mdp);
+    else {
+        ErtsMonitorDataExtended *mdep = (ErtsMonitorDataExtended *) mdp;
+        ErlOffHeap oh;
+        if (mdp->origin.type == ERTS_MON_TYPE_NODE)
+            ERTS_ML_ASSERT(!mdep->uptr.node_monitors);
+        else if (mdep->uptr.ohhp) {
+            ERTS_INIT_OFF_HEAP(&oh);
+            oh.first = mdep->uptr.ohhp;
+            erts_cleanup_offheap(&oh);
+        }
+        if (mdep->dist)
+            erts_mon_link_dist_dec_refc(mdep->dist);
+        erts_free(ERTS_ALC_T_MONITOR_EXT, mdp);
+    }
+}
+
+void
+erts_monitor_set_dead_dist(ErtsMonitor *mon, Eterm nodename)
+{
+    ErtsMonitorDataExtended *mdep;
+    mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+
+    ERTS_ML_ASSERT(mon->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(mon->type == ERTS_MON_TYPE_DIST_PROC);
+    ERTS_ML_ASSERT(!mdep->dist);
+
+    mdep->dist = erts_mon_link_dist_create(nodename);
+    mdep->dist->alive = 0;
+}
+
+Uint
+erts_monitor_size(ErtsMonitor *mon)
+{
+    Uint size, refc;
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+
+    if (!(mon->flags & ERTS_ML_FLG_EXTENDED))
+        size = sizeof(ErtsMonitorDataHeap);
+    else if (mon->type == ERTS_MON_TYPE_SUSPEND)
+        size = sizeof(ErtsMonitorSuspend);
+    else {
+        ErtsMonitorDataExtended *mdep;
+        Uint hsz = 0;
+
+        mdep = (ErtsMonitorDataExtended *) mdp;
+
+        if (mon->type != ERTS_MON_TYPE_NODE
+            && mon->type != ERTS_MON_TYPE_NODES) {
+            if (!is_immed(mdep->md.ref))
+                hsz += NC_HEAP_SIZE(mdep->md.ref);
+            if (mon->type == ERTS_MON_TYPE_DIST_PROC) {
+                if (!is_immed(mdep->md.origin.other.item))
+                    hsz += NC_HEAP_SIZE(mdep->md.origin.other.item);
+                if (!is_immed(mdep->md.target.other.item))
+                    hsz += NC_HEAP_SIZE(mdep->md.target.other.item);
+            }
+        }
+        size = sizeof(ErtsMonitorDataExtended) + (hsz - 1)*sizeof(Eterm);
+    }
+    
+    refc = (Uint) erts_atomic32_read_nob(&mdp->refc);
+    ASSERT(refc > 0);
+
+    return size / refc;
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Link Operations                                                           *
+ *                                                                           *
+\*                                                                           */
+
+#ifdef ERTS_ML_DEBUG
+size_t erts_link_a_offset;
+size_t erts_link_b_offset;
+size_t erts_link_key_offset;
+#endif
+
+static ERTS_INLINE void
+link_init(void)
+{
+#ifdef ERTS_ML_DEBUG
+    erts_link_a_offset = offsetof(ErtsLinkData, a);
+    erts_link_b_offset = offsetof(ErtsLinkData, b);
+    erts_link_key_offset = offsetof(ErtsLink, other.item);
+#endif
+}
+
+ErtsLink *
+erts_link_tree_lookup(ErtsLink *root, Eterm key)
+{
+    ASSERT(is_pid(key) || is_internal_port(key));
+    return (ErtsLink *) ml_rbt_lookup((ErtsMonLnkNode *) root, key);
+}
+
+ErtsLink *
+erts_link_tree_lookup_insert(ErtsLink **root, ErtsLink *lnk)
+{
+    return (ErtsLink *) ml_rbt_lookup_insert((ErtsMonLnkNode **) root,
+                                             (ErtsMonLnkNode *) lnk);
+}
+
+typedef struct {
+    Uint16 type;
+    Eterm a;
+} ErtsLinkCreateCtxt;
+
+static ErtsMonLnkNode *
+create_link(Eterm b, void *vcctxt)
+{
+    ErtsLinkCreateCtxt *cctxt = vcctxt;
+    ErtsLinkData *ldp = erts_link_create(cctxt->type, cctxt->a, b);
+    ERTS_ML_ASSERT(ml_cmp_keys(ldp->a.other.item, b) == 0);
+    return (ErtsMonLnkNode *) &ldp->a;
+}
+
+ErtsLink *erts_link_tree_lookup_create(ErtsLink **root, int *created,
+                                       Uint16 type, Eterm this,
+                                       Eterm other)
+{
+    ErtsLinkCreateCtxt cctxt;
+    cctxt.type = type;
+    cctxt.a = this;
+    return (ErtsLink *) ml_rbt_lookup_create((ErtsMonLnkNode **) root,
+                                             other, create_link,
+                                             (void *) &cctxt,
+                                             created);
+}
+
+void
+erts_link_tree_insert(ErtsLink **root, ErtsLink *lnk)
+{
+    ml_rbt_insert((ErtsMonLnkNode **) root, (ErtsMonLnkNode *) lnk);
+}
+
+void
+erts_link_tree_replace(ErtsLink **root, ErtsLink *old, ErtsLink *new)
+{
+    ml_rbt_replace((ErtsMonLnkNode **) root,
+                   (ErtsMonLnkNode *) old,
+                   (ErtsMonLnkNode *) new);
+}
+
+void
+erts_link_tree_delete(ErtsLink **root, ErtsLink *lnk)
+{
+    ml_rbt_delete((ErtsMonLnkNode **) root, (ErtsMonLnkNode *) lnk);
+}
+
+void
+erts_link_tree_foreach(ErtsLink *root,
+                       void (*func)(ErtsLink *, void *),
+                       void *arg)
+{
+    ml_rbt_foreach((ErtsMonLnkNode *) root,
+                   (void (*)(ErtsMonLnkNode*, void*)) func,
+                   arg);
+
+}
+
+int
+erts_link_tree_foreach_yielding(ErtsLink *root,
+                                void (*func)(ErtsLink *, void *),
+                                void *arg,
+                                void **vyspp,
+                                Sint limit)
+{
+    return ml_rbt_foreach_yielding((ErtsMonLnkNode *) root,
+                                   (void (*)(ErtsMonLnkNode*, void*)) func,
+                                   arg, vyspp, limit);
+}
+
+void
+erts_link_tree_foreach_delete(ErtsLink **root,
+                              void (*func)(ErtsLink *, void *),
+                              void *arg)
+{
+    ml_rbt_foreach_delete((ErtsMonLnkNode **) root,
+                          (void (*)(ErtsMonLnkNode*, void*)) func,
+                          arg);
+}
+
+int
+erts_link_tree_foreach_delete_yielding(ErtsLink **root,
+                                       void (*func)(ErtsLink *, void *),
+                                       void *arg,
+                                       void **vyspp,
+                                       Sint limit)
+{
+    return ml_rbt_foreach_delete_yielding((ErtsMonLnkNode **) root,
+                                          (void (*)(ErtsMonLnkNode*, void*)) func,
+                                          arg, vyspp, limit);
+}
+
+void
+erts_link_list_foreach(ErtsLink *list,
+                       void (*func)(ErtsLink *, void *),
+                       void *arg)
+{
+    void *ystate = NULL;
+    while (ml_dl_list_foreach_yielding((ErtsMonLnkNode *) list,
+                                       (void (*)(ErtsMonLnkNode *, void *)) func,
+                                       arg, &ystate, (Sint) INT_MAX));
+}
+
+int
+erts_link_list_foreach_yielding(ErtsLink *list,
+                                void (*func)(ErtsLink *, void *),
+                                void *arg,
+                                void **vyspp,
+                                Sint limit)
+{
+    return ml_dl_list_foreach_yielding((ErtsMonLnkNode *) list,
+                                       (void (*)(ErtsMonLnkNode *, void *)) func,
+                                       arg, vyspp, limit);
+}
+
+void
+erts_link_list_foreach_delete(ErtsLink **list,
+                               void (*func)(ErtsLink *, void *),
+                               void *arg)
+{
+    void *ystate = NULL;
+    while (ml_dl_list_foreach_delete_yielding((ErtsMonLnkNode **) list,
+                                              (void (*)(ErtsMonLnkNode*, void*)) func,
+                                              arg, &ystate, (Sint) INT_MAX));
+}
+
+int
+erts_link_list_foreach_delete_yielding(ErtsLink **list,
+                                       void (*func)(ErtsLink *, void *),
+                                       void *arg,
+                                       void **vyspp,
+                                       Sint limit)
+{
+    return ml_dl_list_foreach_delete_yielding((ErtsMonLnkNode **) list,
+                                              (void (*)(ErtsMonLnkNode*, void*)) func,
+                                              arg, vyspp, limit);
+}
+
+ErtsLinkData *
+erts_link_create(Uint16 type, Eterm a, Eterm b)
+{
+    ErtsLinkData *ldp;
+
+#ifdef ERTS_ML_DEBUG
+    switch (type) {
+    case ERTS_LNK_TYPE_PROC:
+        ERTS_ML_ASSERT(is_internal_pid(a) && is_internal_pid(a));
+        break;
+    case ERTS_LNK_TYPE_PORT:
+        ERTS_ML_ASSERT(is_internal_pid(a) || is_internal_pid(b));
+        ERTS_ML_ASSERT(is_internal_port(a) || is_internal_port(b));
+        break;
+    case ERTS_LNK_TYPE_DIST_PROC:
+        ERTS_ML_ASSERT(is_internal_pid(a) || is_internal_pid(b));
+        ERTS_ML_ASSERT(is_external_pid(a) || is_external_pid(b));
+        break;
+    default:
+        ERTS_INTERNAL_ERROR("Invalid link type");
+        break;
+    }
+#endif
+
+    if (type != ERTS_LNK_TYPE_DIST_PROC) {
+        ldp = erts_alloc(ERTS_ALC_T_LINK, sizeof(ErtsLinkData));
+
+        ldp->a.other.item = b;
+        ldp->a.flags = (Uint16) 0;
+
+        ldp->b.other.item = a;
+        ldp->b.flags = (Uint16) 0;
+    }
+    else {
+        ErtsLinkDataExtended *ldep;
+        Uint size, hsz;
+        Eterm *hp;
+        ErlOffHeap oh;
+
+        if (is_internal_pid(a))
+            hsz = NC_HEAP_SIZE(b);
+        else 
+            hsz = NC_HEAP_SIZE(a);
+        ERTS_ML_ASSERT(hsz > 0);
+
+        size = sizeof(ErtsLinkDataExtended) - sizeof(Eterm);
+        size += hsz*sizeof(Eterm);
+
+        ldp = erts_alloc(ERTS_ALC_T_LINK_EXT, size);
+
+        ldp->a.flags = ERTS_ML_FLG_EXTENDED;
+        ldp->b.flags = ERTS_ML_FLG_EXTENDED;
+
+        ldep = (ErtsLinkDataExtended *) ldp;
+        hp = &ldep->heap[0];
+
+        ERTS_INIT_OFF_HEAP(&oh);
+
+        if (is_internal_pid(a)) {
+            ldp->a.other.item = STORE_NC(&hp, &oh, b);
+            ldp->b.other.item = a;
+        }
+        else {
+            ldp->a.other.item = b;
+            ldp->b.other.item = STORE_NC(&hp, &oh, a);
+        }
+
+        ldep->ohhp = oh.first;
+        ldep->dist = NULL;
+    }
+
+    erts_atomic32_init_nob(&ldp->refc, 2);
+
+    ldp->a.key_offset = (Uint16) offsetof(ErtsLink, other.item);
+    ldp->a.offset = (Uint16) offsetof(ErtsLinkData, a);
+    ldp->a.type = type;
+
+    ldp->b.key_offset = (Uint16) offsetof(ErtsLink, other.item);
+    ldp->b.offset = (Uint16) offsetof(ErtsLinkData, b);
+    ldp->b.type = type;
+
+    return ldp;
+}
+
+/*
+ * erts_link_destroy__() should only be called from
+ * erts_link_release() or erts_link_release_both().
+ */
+void
+erts_link_destroy__(ErtsLinkData *ldp)
+{
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&ldp->refc) == 0);
+    ERTS_ML_ASSERT(!(ldp->a.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(!(ldp->b.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT((ldp->a.flags & ERTS_ML_FLGS_SAME)
+                   == (ldp->b.flags & ERTS_ML_FLGS_SAME));
+
+    if (!(ldp->a.flags & ERTS_ML_FLG_EXTENDED))
+        erts_free(ERTS_ALC_T_LINK, ldp);
+    else {
+        ErtsLinkDataExtended *ldep = (ErtsLinkDataExtended *) ldp;
+        ErlOffHeap oh;
+        if (ldep->ohhp) {
+            ERTS_INIT_OFF_HEAP(&oh);
+            oh.first = ldep->ohhp;
+            erts_cleanup_offheap(&oh);
+        }
+        if (ldep->dist)
+            erts_mon_link_dist_dec_refc(ldep->dist);
+        erts_free(ERTS_ALC_T_LINK_EXT, ldep);
+    }
+}
+
+void
+erts_link_set_dead_dist(ErtsLink *lnk, Eterm nodename)
+{
+    ErtsLinkDataExtended *ldep;
+    ldep = (ErtsLinkDataExtended *) erts_link_to_data(lnk);
+
+    ERTS_ML_ASSERT(lnk->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(lnk->type == ERTS_LNK_TYPE_DIST_PROC);
+    ERTS_ML_ASSERT(!ldep->dist);
+
+    ldep->dist = erts_mon_link_dist_create(nodename);
+    ldep->dist->alive = 0;
+}
+
+Uint
+erts_link_size(ErtsLink *lnk)
+{
+    Uint size, refc;
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+
+    if (!(lnk->flags & ERTS_ML_FLG_EXTENDED))
+        size = sizeof(ErtsLinkData);
+    else {
+        ErtsLinkDataExtended *ldep = (ErtsLinkDataExtended *) ldp;
+
+        ASSERT(lnk->type == ERTS_LNK_TYPE_DIST_PROC);
+        ASSERT(is_external_pid_header(ldep->heap[0]));
+
+        size = sizeof(ErtsLinkDataExtended);
+        size += thing_arityval(ldep->heap[0])*sizeof(Eterm);
+    }
+    
+    refc = (Uint) erts_atomic32_read_nob(&ldp->refc);
+    ASSERT(refc > 0);
+
+    return size / refc;
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Misc                                                                      *
+\*                                                                           */
+
+void
+erts_monitor_link_init(void)
+{
+    monitor_init();
+    link_init();
+}
diff --git a/erts/emulator/beam/erl_monitor_link.h b/erts/emulator/beam/erl_monitor_link.h
new file mode 100644
index 0000000000..9ff8aa509a
--- /dev/null
+++ b/erts/emulator/beam/erl_monitor_link.h
@@ -0,0 +1,2349 @@
+/*
+ * %CopyrightBegin%
+ * 
+ * Copyright Ericsson AB 2018. All Rights Reserved.
+ * 
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * 
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description:	Monitor and link implementation.
+ *
+ *              === Monitors ==================================================
+ *
+ *              The monitor data structure contains:
+ *              - an 'origin' part that should be inserted in a data structure
+ *                of the origin entity, i.e., the entity monitoring another
+ *                entity
+ *              - a 'target' part that should be inserted in a data structure
+ *                of the target entity, i.e., the entity being monitored by
+ *                another entity
+ *              - a shared part that contains information shared between both
+ *                origin and target entities
+ *
+ *              That is, the two halves of the monitor as well as shared data
+ *              are allocated in one single continuous memory block. The
+ *              origin and target parts can separately each be inserted in
+ *              either a (red-black) tree, a (circular double linked) list, or
+ *              in a process signal queue.
+ *
+ *              Each process and port contains:
+ *              - a monitor list for local target monitors that is accessed
+ *                via the ERTS_P_LT_MONITORS() macro, and
+ *              - a monitor tree for other monitors that is accessed via the
+ *                ERTS_P_MONITORS() macro
+ *
+ *              These fields of processes/ports are protected by the main lock
+ *              of the process/port. These are only intended to be accessed by
+ *              the process/port itself. When setting up or tearing down a
+ *              monitor one should *only* operate on the monitor tree/list of
+ *              the currently executing process/port and send signals to the
+ *              other involved process/port so it can modify its own monitor
+ *              tree/list by itself (see erl_proc_sig_queue.h). One should
+ *              absolutely *not* acquire the lock of the other involved
+ *              process/port and operate on its monitor tree/list directly.
+ *
+ *              Each dist entry contains a monitor/link dist structure that
+ *              contains:
+ *              - a monitor tree for origin named monitors that is accessed via
+ *                the field 'orig_name_monitors', and
+ *              - a monitor list for other monitors that is accessed via the
+ *                'monitors' field.
+ *              Monitors in these fields contain information about all monitors
+ *              over this specific connection.
+ *
+ *              The fields of the dist structure are protected by a mutex in
+ *              the same dist structure. Operations on these fields are
+ *              normally performed by the locally involved process only,
+ *              except when a connection is taken down. However in the case
+ *              of distributed named monitors that originates from another
+ *              node this is not possible. That is this operation is also
+ *              performed from another context that the locally involved
+ *              process.
+ *
+ *              Access to monitor trees are performed using the
+ *              erts_monitor_tree_* functions below. Access to monitor lists
+ *              are performed using the erts_monitor_list_* functions below.
+ *
+ *
+ *              The different monitor types:
+ *
+ *              --- ERTS_MON_TYPE_PROC ----------------------------------------
+ *
+ *              A local process (origin) monitors another local process
+ *              (target).
+ *
+ *              Origin:
+ *                      Other Item:     Target process identifier
+ *              Target:
+ *                      Other Item:     Origin process identifier
+ *              Shared:
+ *                      Key:            Reference
+ *                      Name:           Name (atom) if by name
+ *
+ *              Valid keys are only ordinary internal references.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list for local targets on the target process.
+ *
+ *              --- ERTS_MON_TYPE_PORT ----------------------------------------
+ *
+ *              A local process (origin) monitors a local port (target), or a
+ *              local port (origin) monitors a local process (target).
+ *
+ *              Origin:
+ *                      Other Item:     Target process/port identifier
+ *              Target:
+ *                      Other Item:     Origin process/port identifier
+ *              Shared:
+ *                      Key:            Reference
+ *                      Name:           Name (atom) if by name
+ *
+ *              Valid keys are only ordinary internal references.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process/port and target part of the monitor is stored
+ *              in monitor list for local targets on the target process/port.
+ *
+ *
+ *              --- ERTS_MON_TYPE_TIME_OFFSET ---------------------------------
+ *
+ *              A local process (origin) monitors time offset (target)
+ *
+ *              Origin:
+ *                      Other Item:     clock_service
+ *              Target:
+ *                      Other Item:     Origin process identifier
+ *              Shared:
+ *                      Key:            Reference
+ *
+ *              Valid keys are only ordinary internal references.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list referred by the variable 'time_offset_monitors'
+ *              (see erl_time_sup.c).
+ *
+ *
+ *              --- ERTS_MON_TYPE_DIST_PROC -----------------------------------
+ *
+ *              A local process (origin) monitors a remote process (target).
+ *              Origin node on local process and target node on dist entry.
+ *
+ *              Origin:
+ *                      Other Item:     Remote process identifier/Node name
+ *                                      if by name
+ *              Target:
+ *                      Other Item:     Local process identifier
+ *              Shared:
+ *                      Key:            Reference
+ *                      Name:           Name (atom) if by name
+ *                      Dist:           Pointer to dist structure
+ *
+ *              Valid keys are only ordinary internal references.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list referred by 'monitors' field of the dist
+ *              structure.
+ *
+ *
+ *              A remote process (origin) monitors a local process (target).
+ *              Origin node on dist entry and target node on local process.
+ *
+ *              Origin:
+ *                      Other Item:     Local process identifier
+ *              Target:
+ *                      Other Item:     Remote process identifier
+ *              Shared:
+ *                      Key:            Reference
+ *                      Name:           Name (atom) if by name
+ *
+ *              Valid keys are only external references.
+ *
+ *              If monitor by name, the origin part of the monitor is stored
+ *              in the monitor tree referred by 'orig_name_monitors' field in
+ *              dist structure; otherwise in the monitor list referred by
+ *              'monitors' field in dist structure. The target part of the
+ *              monitor is stored in the monitor tree of the local target
+ *              process.
+ *
+ *
+ *              --- ERTS_MON_TYPE_RESOURCE ------------------------------------
+ *
+ *              A NIF resource (origin) monitors a process (target).
+ *
+ *              Origin:
+ *                      Other Item:     Target process identifier
+ *              Target:
+ *                      Other Ptr:      Pointer to resource
+ *              Shared:
+ *                      Key:            Reference
+ *
+ *              Valid keys are only ordinary internal references.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin resource (see erl_nif.c) and target part of the
+ *              monitor is stored in monitor list for local targets on the
+ *              target process.
+ *
+ *              --- ERTS_MON_TYPE_NODE ----------------------------------------
+ *
+ *              A local process (origin) monitors a distribution connection
+ *              (target) via erlang:monitor_node().
+ *
+ *              Origin:
+ *                      Other Item:     Node name (atom)
+ *                      Key:            Node name
+ *              Target:
+ *                      Other Item:     Origin process identifier
+ *                      Key:            Origin process identifier
+ *              Shared:
+ *                      Refc:           Number of invocations
+ *
+ *              Valid keys are only node-name atoms and internal process
+ *              identifiers.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list referred by 'monitors' field of the dist
+ *              structure.
+ *
+ *              --- ERTS_MON_TYPE_NODES ---------------------------------------
+ *
+ *              A local process (origin) monitors all connections (target),
+ *              via net_kernel:monitor_nodes().
+ *
+ *              Origin:
+ *                      Other Item:     Bit mask (small)
+ *                      Key:            Bit mask
+ *              Target:
+ *                      Other Item:     Origin process identifier
+ *                      Key:            Origin process identifier
+ *              Shared:
+ *                      Refc:           Number of invocations
+ *
+ *              Valid keys are only small integers and internal process
+ *              identifiers.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list referred by the variable 'nodes_monitors' (see
+ *              dist.c).
+ *
+ *              --- ERTS_MON_TYPE_SUSPEND -------------------------------------
+ *
+ *              Suspend monitor. A local process (origin) suspends another
+ *              local process (target).
+ *
+ *              Origin:
+ *                      Other Item:     Process identifier of suspendee
+ *                                      (target)
+ *                      Key:            Process identifier of suspendee
+ *                                      (target)
+ *              Target:
+ *                      Other Item:     Process identifier of suspender
+ *                                      (origin)
+ *                      Key:            Process identifier of suspender
+ *                                      (origin)
+ *              Shared:
+ *                      Next:           Pointer to another suspend monitor
+ *                      State:          Number of suspends and a flag
+ *                                      indicating if the suspend is
+ *                                      active or not.
+ *
+ *              Origin part of the monitor is stored in the monitor tree of
+ *              origin process and target part of the monitor is stored in
+ *              monitor list for local targets on the target process.
+ *
+ *
+ *
+ *              === Links =====================================================
+ *
+ *              The link data structure contains:
+ *              - an 'a' part that should be inserted in a data structure of
+ *                one entity and contains the identifier of the other involved
+ *                entity (b)
+ *              - a 'b' part that should be inserted in a data structure of
+ *                the other involved entity and contains the identifier of the
+ *                other involved entity (a)
+ *              - shared part that contains information shared between both
+ *                involved entities
+ *
+ *              That is, the two halves of the link as well as shared data
+ *              are allocated in one single continuous memory block. The 'a'
+ *              and the 'b' parts can separately each be inserted in either
+ *              a (red-black) tree, a (circular double linked) list, or in a
+ *              process signal queue.
+ *
+ *              Each process and port contains:
+ *              - a link tree for links that is accessed via the
+ *                ERTS_P_LINKS() macro
+ *
+ *              This field of processes/ports is protected by the main lock of
+ *              the process/port. It is only intended to be accessed by the
+ *              process/port itself. When setting up or tearing down a link
+ *              one should *only* operate on the link tree of the currently
+ *              executing process/port and send signals to the other involved
+ *              process/port so it can modify its own monitor tree by itself
+ *              (see erl_proc_sig_queue.h). One should absolutely *not*
+ *              acquire the lock of the other involved process/port and
+ *              operate on its link tree directly.
+ *
+ *              Each dist entry contains a monitor/link dist structure that
+ *              contains:
+ *              - a link list for links via the 'links' field.
+ *              Links in this field contain information about all links over
+ *              this specific connection.
+ *
+ *              The fields of the dist structure are protected by a mutex in
+ *              the same dist structure. Operation on the 'links' fields are
+ *              normally performed by the locally involved process only,
+ *              except when a connection is taken down.
+ *
+ *              Access to link trees are performed using the erts_link_tree_*
+ *              functions below. Access to link lists are performed using the
+ *              erts_link_list_* functions below.
+ *
+ *              There can only be one link between the same pair of
+ *              processes/ports. Since a link can be simultaneously initiated
+ *              from both ends we always save the link data structure with the
+ *              lowest address if multiple links should appear between the
+ *              same pair of processes/ports.
+ *
+ *
+ *              The different link types:
+ *
+ *              --- ERTS_LNK_TYPE_PROC -----------------------------------------
+ *
+ *              A link between a local process A and a local process B.
+ *
+ *              A:
+ *                      Other Item:     B process identifier
+ *                      Key:            B process identifier
+ *              B:
+ *                      Other Item:     A process identifier
+ *                      Key:            A process identifier
+ *
+ *              Valid keys are only internal process identifiers.
+ *
+ *              'A' part of the link stored in the link tree of process A and
+ *              'B' part of the link is stored in link tree of process B.
+ *
+ *              --- ERTS_LNK_TYPE_PORT -----------------------------------------
+ *
+ *              A link between a local process/port A and a local process/port
+ *              B.
+ *
+ *              A:
+ *                      Other Item:     B process/port identifier
+ *                      Key:            B process/port identifier
+ *              B:
+ *                      Other Item:     A process/port identifier
+ *                      Key:            A process/port identifier
+ *
+ *              Valid keys are internal process identifiers and internal port
+ *              identifiers.
+ *
+ *              'A' part of the link stored in the link tree of process/port
+ *              A and 'B' part of the link is stored in link tree of
+ *              process/port B.
+ *
+ *              --- ERTS_LNK_TYPE_DIST_PROC ------------------------------------
+ *
+ *              A link between a local process and a remote process. Either of
+ *              the processes can be used as A or B.
+ *
+ *              A:
+ *                      Other Item:     B process identifier
+ *                      Key:            B process identifier
+ *              B:
+ *                      Other Item:     A process identifier
+ *                      Key:            A process identifier
+ *              Shared:
+ *                      Dist:           Pointer to dist structure
+ *
+ *              Valid keys are internal and external process identifiers.
+ *
+ *              The part of the link with a remote pid as "other item" is
+ *              stored in the link tree of the local process. The part of
+ *              the link with a local pid as "other item" is stored in the
+ *              links list of the dist structure.
+ *
+ *              ===============================================================
+ *
+ * Author: 	Rickard Green
+ *
+ */
+
+#ifndef ERL_MONITOR_LINK_H__
+#define ERL_MONITOR_LINK_H__
+
+#define ERTS_PROC_SIG_QUEUE_TYPE_ONLY
+#include "erl_proc_sig_queue.h"
+#undef ERTS_PROC_SIG_QUEUE_TYPE_ONLY
+
+#if defined(DEBUG) || 0
+#  define ERTS_ML_DEBUG
+#else
+#  undef ERTS_ML_DEBUG
+#endif
+
+#ifdef ERTS_ML_DEBUG
+#  define ERTS_ML_ASSERT ERTS_ASSERT
+#else
+#  define ERTS_ML_ASSERT(E) ((void) 1)
+#endif
+
+#define ERTS_MON_TYPE_MAX               ((Uint16) 7)
+
+#define ERTS_MON_TYPE_PROC              ((Uint16) 0)
+#define ERTS_MON_TYPE_PORT              ((Uint16) 1)
+#define ERTS_MON_TYPE_TIME_OFFSET       ((Uint16) 2)
+#define ERTS_MON_TYPE_DIST_PROC         ((Uint16) 3)
+#define ERTS_MON_TYPE_RESOURCE          ((Uint16) 4)
+#define ERTS_MON_TYPE_NODE              ((Uint16) 5)
+#define ERTS_MON_TYPE_NODES             ((Uint16) 6)
+#define ERTS_MON_TYPE_SUSPEND           ERTS_MON_TYPE_MAX
+
+#define ERTS_MON_LNK_TYPE_MAX           (ERTS_MON_TYPE_MAX + ((Uint16) 3))
+#define ERTS_LNK_TYPE_MAX               ERTS_MON_LNK_TYPE_MAX
+
+#define ERTS_LNK_TYPE_PROC              (ERTS_MON_TYPE_MAX + ((Uint16) 1))
+#define ERTS_LNK_TYPE_PORT              (ERTS_MON_TYPE_MAX + ((Uint16) 2))
+#define ERTS_LNK_TYPE_DIST_PROC         ERTS_LNK_TYPE_MAX
+
+#define ERTS_ML_FLG_TARGET              (((Uint16) 1) << 0)
+#define ERTS_ML_FLG_IN_TABLE            (((Uint16) 1) << 1)
+#define ERTS_ML_FLG_IN_SUBTABLE         (((Uint16) 1) << 2)
+#define ERTS_ML_FLG_NAME                (((Uint16) 1) << 3)
+#define ERTS_ML_FLG_EXTENDED            (((Uint16) 1) << 4)
+
+#define ERTS_ML_FLG_DBG_VISITED         (((Uint16) 1) << 15)
+
+/* Flags that should be the same on both monitor/link halves */
+#define ERTS_ML_FLGS_SAME \
+    (ERTS_ML_FLG_EXTENDED|ERTS_ML_FLG_NAME)
+
+typedef struct ErtsMonLnkNode__ ErtsMonLnkNode;
+
+typedef struct {
+    UWord parent; /* Parent ptr and flags... */
+    ErtsMonLnkNode *right;
+    ErtsMonLnkNode *left;
+} ErtsMonLnkTreeNode;
+
+typedef struct {
+    ErtsMonLnkNode *next;
+    ErtsMonLnkNode *prev;
+} ErtsMonLnkListNode;
+
+struct ErtsMonLnkNode__ {
+    union {
+        ErtsSignalCommon signal;
+        ErtsMonLnkTreeNode tree;
+        ErtsMonLnkListNode list;
+    } node;
+    union {
+        Eterm item;
+        void *ptr;
+    } other;
+    Uint16 offset; /* offset from monitor/link data to this structure (node) */
+    Uint16 key_offset; /* offset from this structure (node) to key */
+    Uint16 flags;
+    Uint16 type;
+};
+
+typedef struct {
+    Eterm nodename;
+    Uint32 connection_id;
+    erts_atomic_t refc;
+    erts_mtx_t mtx;
+    int alive;
+    ErtsMonLnkNode *links; /* Link double linked circular list */
+    ErtsMonLnkNode *monitors; /* Monitor double linked circular list */
+    ErtsMonLnkNode *orig_name_monitors; /* Origin named monitors
+                                           read-black tree */
+} ErtsMonLnkDist;
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Misc                                                                      *
+\*                                                                           */
+
+/**
+ *
+ * @brief Initialize monitor/link implementation
+ *
+ */
+void erts_monitor_link_init(void);
+
+/**
+ *
+ * @brief Create monitor/link dist structure to attach to dist entry
+ *
+ * Create dist structure containing monitor and link containers. This
+ * structure is to be attached to a connected dist entry.
+ *
+ * @param[in]     nodename      Node name as an atom
+ *
+ * @returns                     Pointer to dist structure
+ *
+ */
+ErtsMonLnkDist *erts_mon_link_dist_create(Eterm nodename);
+
+/**
+ *
+ * @brief Increase reference count of monitor/link dist structure
+ *
+ * @param[in]     mld           Pointer to dist structure
+ *
+ */
+ERTS_GLB_INLINE void erts_mon_link_dist_inc_refc(ErtsMonLnkDist *mld);
+
+/**
+ *
+ * @brief Decrease reference count of monitor/link dist structure
+ *
+ * @param[in]     mld           Pointer to dist structure
+ *
+ */
+ERTS_GLB_INLINE void erts_mon_link_dist_dec_refc(ErtsMonLnkDist *mld);
+
+/* internal functions... */
+ERTS_GLB_INLINE void erts_ml_dl_list_insert__(ErtsMonLnkNode **list,
+                                              ErtsMonLnkNode *ml);
+ERTS_GLB_INLINE void erts_ml_dl_list_delete__(ErtsMonLnkNode **list,
+                                              ErtsMonLnkNode *ml);
+ERTS_GLB_INLINE ErtsMonLnkNode *erts_ml_dl_list_first__(ErtsMonLnkNode *list);
+ERTS_GLB_INLINE ErtsMonLnkNode *erts_ml_dl_list_last__(ErtsMonLnkNode *list);
+void erts_mon_link_dist_destroy__(ErtsMonLnkDist *mld);
+ERTS_GLB_INLINE void *erts_ml_node_to_main_struct__(ErtsMonLnkNode *mln);
+
+/* implementations for globally inlined misc functions... */
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE void
+erts_mon_link_dist_inc_refc(ErtsMonLnkDist *mld)
+{
+    ERTS_ML_ASSERT(erts_atomic_read_nob(&mld->refc) > 0);
+    erts_atomic_inc_nob(&mld->refc);
+}
+
+ERTS_GLB_INLINE void
+erts_mon_link_dist_dec_refc(ErtsMonLnkDist *mld)
+{
+    ERTS_ML_ASSERT(erts_atomic_read_nob(&mld->refc) > 0);
+    if (erts_atomic_dec_read_nob(&mld->refc) == 0)
+        erts_mon_link_dist_destroy__(mld);
+}
+
+ERTS_GLB_INLINE void *
+erts_ml_node_to_main_struct__(ErtsMonLnkNode *mln)
+{
+    return (void *) (((char *) mln) - ((size_t) mln->offset));
+}
+
+ERTS_GLB_INLINE void
+erts_ml_dl_list_insert__(ErtsMonLnkNode **list, ErtsMonLnkNode *ml)
+{
+    ErtsMonLnkNode *first = *list;
+    ERTS_ML_ASSERT(!(ml->flags & ERTS_ML_FLG_IN_TABLE));
+    if (!first) {
+        ml->node.list.next = ml->node.list.prev = ml;
+        *list = ml;
+    }
+    else {
+        ERTS_ML_ASSERT(first->node.list.prev->node.list.next == first);
+        ERTS_ML_ASSERT(first->node.list.next->node.list.prev == first);
+        ml->node.list.next = first;
+        ml->node.list.prev = first->node.list.prev;
+        first->node.list.prev = ml;
+        ml->node.list.prev->node.list.next = ml;
+    }
+    ml->flags |= ERTS_ML_FLG_IN_TABLE;
+}
+
+ERTS_GLB_INLINE void
+erts_ml_dl_list_delete__(ErtsMonLnkNode **list, ErtsMonLnkNode *ml)
+{
+    ERTS_ML_ASSERT(ml->flags & ERTS_ML_FLG_IN_TABLE);
+    if (ml->node.list.next == ml) {
+        ERTS_ML_ASSERT(ml->node.list.prev == ml);
+        ERTS_ML_ASSERT(*list == ml);
+
+        *list = NULL;
+    }
+    else {
+        ERTS_ML_ASSERT(ml->node.list.prev->node.list.next == ml);
+        ERTS_ML_ASSERT(ml->node.list.prev != ml);
+        ERTS_ML_ASSERT(ml->node.list.next->node.list.prev == ml);
+        ERTS_ML_ASSERT(ml->node.list.next != ml);
+
+        if (*list == ml)
+            *list = ml->node.list.next;
+        ml->node.list.prev->node.list.next = ml->node.list.next;
+        ml->node.list.next->node.list.prev = ml->node.list.prev;
+    }
+    ml->flags &= ~ERTS_ML_FLG_IN_TABLE;
+}
+
+ERTS_GLB_INLINE ErtsMonLnkNode *
+erts_ml_dl_list_first__(ErtsMonLnkNode *list)
+{
+    return list;
+}
+
+ERTS_GLB_INLINE ErtsMonLnkNode *
+erts_ml_dl_list_last__(ErtsMonLnkNode *list)
+{
+    if (!list)
+        return NULL;
+    return list->node.list.prev;
+}
+
+#endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Monitor Operations                                                        *
+\*                                                                           */
+
+
+typedef struct ErtsMonLnkNode__ ErtsMonitor;
+
+typedef struct {
+    ErtsMonitor origin;
+    ErtsMonitor target;
+    Eterm ref;
+    erts_atomic32_t refc;
+} ErtsMonitorData;
+
+typedef struct {
+    ErtsMonitorData md;
+    ErtsORefThing oref_thing;
+} ErtsMonitorDataHeap;
+
+typedef struct ErtsMonitorDataExtended__ ErtsMonitorDataExtended;
+
+struct ErtsMonitorDataExtended__ {
+    ErtsMonitorData md;
+    union {
+        Eterm name;
+        Uint refc;
+    } u;
+    union {
+        struct erl_off_heap_header *ohhp;
+        ErtsMonitor *node_monitors;
+    } uptr;
+    ErtsMonLnkDist *dist;
+    Eterm heap[1]; /* heap start... */
+};
+
+typedef struct ErtsMonitorSuspend__ ErtsMonitorSuspend;
+
+struct ErtsMonitorSuspend__ {
+    ErtsMonitorData md; /* origin = suspender; target = suspendee */
+    ErtsMonitorSuspend *next;
+    erts_atomic_t state;
+};
+#define ERTS_MSUSPEND_STATE_FLG_ACTIVE ((erts_aint_t) (((Uint) 1) << (sizeof(Uint)*8 - 1)))
+#define ERTS_MSUSPEND_STATE_COUNTER_MASK (~ERTS_MSUSPEND_STATE_FLG_ACTIVE)
+
+/* 
+ * --- Monitor tree operations ---
+ */
+
+/**
+ *
+ * @brief Lookup a monitor in a monitor tree
+ *
+ *
+ * @param[in]     root          Pointer to root of monitor tree
+ *
+ * @param[in]     key           Key of monitor to lookup
+ *
+ * @returns                     Pointer to a monitor with the
+ *                              key 'key', or NULL if no such
+ *                              monitor was found
+ *
+ */
+ErtsMonitor *erts_monitor_tree_lookup(ErtsMonitor *root, Eterm key);
+
+/**
+ *
+ * @brief Lookup or insert a monitor in a monitor tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is not part of any tree or list
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     mon           Monitor to insert if no monitor
+ *                              with the same key already exists
+ *
+ * @returns                     Pointer to a monitor with the
+ *                              key 'key'. If no monitor with the key
+ *                              'key' was found and 'mon' was inserted
+ *                              'mon' is returned.
+ *
+ */
+ErtsMonitor *erts_monotor_tree_lookup_insert(ErtsMonitor **root,
+                                             ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Lookup or create a node or a nodes monitor in a monitor tree.
+ *
+ * Looks up an origin monitor with the key 'target' in the monitor tree.
+ * If it is not found, creates a monitor and returns a pointer to the
+ * origin monitor.
+ *
+ * When the funcion is called it is assumed that:
+ * - no target monitors with the key 'target' exists in the tree.
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[out]    created       Pointer to integer. The integer is set to
+ *                              a non-zero value if no monitor with key
+ *                              'target' was found, and a new monitor
+ *                              was created. If a monitor was found, it
+ *                              is set to zero.
+ *
+ * @param[in]     type          ERTS_MON_TYPE_NODE | ERTS_MON_TYPE_NODES
+ *
+ * @param[in]     origin        The key of the origin
+ *
+ * @param[in]     target        The key of the target
+ *
+ */
+ErtsMonitor *erts_monitor_tree_lookup_create(ErtsMonitor **root, int *created,
+                                             Uint16 type, Eterm origin,
+                                             Eterm target);
+
+/**
+ *
+ * @brief Insert a monitor in a monitor tree
+ *
+ * When the funcion is called it is assumed that:
+ * - no monitors with the same key that 'mon' exist in the tree
+ * - 'mon' is not part of any list of tree
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     mon           Monitor to insert.
+ *
+ */
+void erts_monitor_tree_insert(ErtsMonitor **root, ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Replace a monitor in a monitor tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'old' monitor and 'new' monitor have exactly the same key
+ * - 'old' monitor is part of the tree
+ * - 'new' monitor is not part of any tree or list
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     old           Monitor to remove from the tree
+ *
+ * @param[in]     new           Monitor to insert into the tree
+ *
+ */
+void erts_monitor_tree_replace(ErtsMonitor **root, ErtsMonitor *old,
+                               ErtsMonitor *new);
+
+/**
+ *
+ * @brief Delete a monitor from a monitor tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is part of the tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     mon           Monitor to remove from the tree
+ *
+ */
+void erts_monitor_tree_delete(ErtsMonitor **root, ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Call a function for each monitor in a monitor tree
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * @param[in]     root          Pointer to root of monitor tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_monitor_tree_foreach(ErtsMonitor *root,
+                               void (*func)(ErtsMonitor *, void *),
+                               void *arg);
+
+/**
+ *
+ * @brief Call a function for each monitor in a monitor tree. Yield
+ *        if lots of monitors exist.
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetedly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]     root          Pointer to root of monitor tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of monitors to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all monitors has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_monitor_tree_foreach_yielding(ErtsMonitor *root,
+                                       void (*func)(ErtsMonitor *, void *),
+                                       void *arg,
+                                       void **vyspp,
+                                       Sint limit);
+
+/**
+ *
+ * @brief Delete all monitors from a monitor tree and call a function for
+ *        each monitor
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_monitor_tree_foreach_delete(ErtsMonitor **root,
+                                      void (*func)(ErtsMonitor *, void *),
+                                      void *arg);
+
+/**
+ *
+ * @brief Delete all monitors from a monitor tree and call a function for
+ *        each monitor
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetededly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of monitor tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of monitors to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all monitors has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_monitor_tree_foreach_delete_yielding(ErtsMonitor **root,
+                                              void (*func)(ErtsMonitor *, void *),
+                                              void *arg,
+                                              void **vyspp,
+                                              Sint limit);
+
+/*
+ * --- Monitor list operations --
+ */
+
+/**
+ *
+ * @brief Insert a monitor in a monitor list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is not part of any list or tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to monitor list
+ *
+ * @param[in]     mon           Monitor to insert
+ *
+ */
+ERTS_GLB_INLINE void erts_monitor_list_insert(ErtsMonitor **list, ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Delete a monitor from a monitor list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is part of the list
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to monitor list
+ *
+ * @param[in]     mon           Monitor to remove from the list
+ *
+ */
+ERTS_GLB_INLINE void erts_monitor_list_delete(ErtsMonitor **list, ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Get a pointer to first monitor in a monitor list
+ *
+ * The monitor will still remain in the list after the return
+ *
+ * @param[in] list              Pointer to monitor list
+ *
+ * @returns                     Pointer to first monitor in list if
+ *                              list is not empty. If list is empty
+ *                              NULL is returned.
+ *
+ */
+ERTS_GLB_INLINE ErtsMonitor *erts_monitor_list_first(ErtsMonitor *list);
+
+/**
+ *
+ * @brief Get a pointer to last monitor in a monitor list
+ *
+ * The monitor will still remain in the list after the return
+ *
+ * @param[in] list              Pointer to monitor list
+ *
+ * @returns                     Pointer to last monitor in list if
+ *                              list is not empty. If list is empty
+ *                              NULL is returned.
+ *
+ */
+ERTS_GLB_INLINE ErtsMonitor *erts_monitor_list_last(ErtsMonitor *list);
+
+/**
+ *
+ * @brief Call a function for each monitor in a monitor list
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'list'.
+ *
+ * @param[in]     list          Pointer to root of monitor list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_monitor_list_foreach(ErtsMonitor *list,
+                               void (*func)(ErtsMonitor *, void *),
+                               void *arg);
+
+/**
+ *
+ * @brief Call a function for each monitor in a monitor list. Yield
+ *        if lots of monitors exist.
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetedly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]     list          Pointer to monitor list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of monitors to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all monitors has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_monitor_list_foreach_yielding(ErtsMonitor *list,
+                                       void (*func)(ErtsMonitor *, void *),
+                                       void *arg,
+                                       void **vyspp,
+                                       Sint limit);
+
+/**
+ *
+ * @brief Delete all monitors from a monitor list and call a function for
+ *        each monitor
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * @param[in,out] list          Pointer to pointer to monitor list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_monitor_list_foreach_delete(ErtsMonitor **list,
+                                      void (*func)(ErtsMonitor *, void *),
+                                      void *arg);
+
+/**
+ *
+ * @brief Delete all monitors from a monitor list and call a function for
+ *        each monitor
+ *
+ * The funcion 'func' will be called with a pointer to a monitor
+ * as first argument and 'arg' as second argument for each monitor
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetededly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to monitor list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of monitors to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all monitors has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_monitor_list_foreach_delete_yielding(ErtsMonitor **list,
+                                              void (*func)(ErtsMonitor *, void *),
+                                              void *arg,
+                                              void **vyspp,
+                                              Sint limit);
+
+/*
+ * --- Misc monitor operations ---
+ */
+
+/**
+ *
+ * @brief Create a monitor
+ *
+ * Can create all types of monitors
+ *
+ * When the funcion is called it is assumed that:
+ * - 'ref' is an internal ordinary reference if type is ERTS_MON_TYPE_PROC,
+ *   ERTS_MON_TYPE_PORT, ERTS_MON_TYPE_TIME_OFFSET, or ERTS_MON_TYPE_RESOURCE
+ * - 'ref' is NIL if type is ERTS_MON_TYPE_NODE, ERTS_MON_TYPE_NODES, or
+ *   ERTS_MON_TYPE_SUSPEND
+ * - 'ref' is and ordinary internal reference or an external reference if
+ *   type is ERTS_MON_TYPE_DIST_PROC
+ * - 'name' is an atom or NIL if type is ERTS_MON_TYPE_PROC,
+ *   ERTS_MON_TYPE_PORT, or ERTS_MON_TYPE_DIST_PROC
+ * - 'name is NIL if type is ERTS_MON_TYPE_TIME_OFFSET, ERTS_MON_TYPE_RESOURCE,
+ *   ERTS_MON_TYPE_NODE, ERTS_MON_TYPE_NODES, or ERTS_MON_TYPE_SUSPEND
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]     type          ERTS_MON_TYPE_PROC, ERTS_MON_TYPE_PORT,
+ *                              ERTS_MON_TYPE_TIME_OFFSET, ERTS_MON_TYPE_DIST_PROC,
+ *                              ERTS_MON_TYPE_RESOURCE, ERTS_MON_TYPE_NODE,
+ *                              ERTS_MON_TYPE_NODES, or ERTS_MON_TYPE_SUSPEND
+ *
+ * @param[in]     ref           A reference or NIL depending on type
+ *
+ * @param[in]     origin        The key of the origin
+ *
+ * @param[in]     target        The key of the target
+ *
+ * @param[in]     name          An atom (the name) or NIL depending on type
+ *
+ * @returns                     A pointer to monitor data structure
+ *
+ */
+ErtsMonitorData *erts_monitor_create(Uint16 type, Eterm ref, Eterm origin,
+                                     Eterm target, Eterm name);
+
+/**
+ *
+ * @brief Get pointer to monitor data structure
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ * @returns                     Pointer to monitor data structure
+ *
+ */
+ERTS_GLB_INLINE ErtsMonitorData *erts_monitor_to_data(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Check if monitor is a target monitor
+ *
+ * @param[in]    mon            Pointer to monitor to check
+ *
+ * @returns                     A non-zero value if target monitor;
+ *                              otherwise zero
+ *
+ */
+ERTS_GLB_INLINE int erts_monitor_is_target(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Check if monitor is an origin monitor
+ *
+ * @param[in]    mon            Pointer to monitor to check
+ *
+ * @returns                     A non-zero value if origin monitor;
+ *                              otherwise zero
+ *
+ */
+ERTS_GLB_INLINE int erts_monitor_is_origin(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Check if monitor is in tree or list
+ *
+ * @param[in]    mon            Pointer to monitor to check
+ *
+ * @returns                     A non-zero value if in tree or list;
+ *                              otherwise zero
+ *
+ */
+ERTS_GLB_INLINE int erts_monitor_is_in_table(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Release monitor
+ *
+ * When both the origin and the target part of the monitor have
+ * been released the monitor structure will be deallocated.
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is not part of any list or tree
+ * - 'mon' is not referred to by any other structures
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ */
+ERTS_GLB_INLINE void erts_monitor_release(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Release both target and origin monitor structures simultaneously
+ *
+ * Release both the origin and target parts of the monitor
+ * simultaneously and deallocate the structure.
+ *
+ * When the funcion is called it is assumed that:
+ * - Neither the origin part nor the target part of the monitor
+ *   are not part of any list or tree
+ * - Neither the origin part nor the target part of the monitor
+ *   are referred to by any other structures
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]    mdp            Pointer to monitor data structure
+ *
+ */
+ERTS_GLB_INLINE void erts_monitor_release_both(ErtsMonitorData *mdp);
+
+/**
+ *
+ * @brief Insert monitor in dist monitor tree or list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor is not part of any list or tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ * @param[in]    dist           Pointer to dist structure
+ *
+ * @returns                     A non-zero value if inserted;
+ *                              otherwise, zero. The monitor
+ *                              is not inserted if the dist
+ *                              structure has been set in a
+ *                              dead state.
+ *
+ */
+ERTS_GLB_INLINE int erts_monitor_dist_insert(ErtsMonitor *mon, ErtsMonLnkDist *dist);
+
+/**
+ *
+ * @brief Delete monitor from dist monitor tree or list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' monitor earler has been inserted into 'dist'
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ * @param[in]    dist           Pointer to dist structure
+ *
+ * @returns                     A non-zero value if deleted;
+ *                              otherwise, zero. The monitor
+ *                              is not deleted if the dist
+ *                              structure has been set in a
+ *                              dead state or if it has already
+ *                              been deleted.
+ *
+ */
+ERTS_GLB_INLINE int erts_monitor_dist_delete(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Set dead dist structure on monitor
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ * @param[in]    nodename       Name of remote node
+ *
+ */
+void
+erts_monitor_set_dead_dist(ErtsMonitor *mon, Eterm nodename);
+
+/**
+ *
+ * @brief Get charged size of monitor
+ *
+ * If the other side of the monitor has been released, the
+ * whole size of the monitor data structure is returned; otherwise,
+ * half of the size is returned.
+ *
+ * When the funcion is called it is assumed that:
+ * - 'mon' has not been released
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    mon            Pointer to monitor
+ *
+ * @returns                     Charged size in bytes
+ *
+ */
+Uint erts_monitor_size(ErtsMonitor *mon);
+
+
+/* internal function... */
+void erts_monitor_destroy__(ErtsMonitorData *mdp);
+
+/* implementations for globally inlined monitor functions... */
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE int
+erts_monitor_is_target(ErtsMonitor *mon)
+{
+    return !!(mon->flags & ERTS_ML_FLG_TARGET);
+}
+
+ERTS_GLB_INLINE int
+erts_monitor_is_origin(ErtsMonitor *mon)
+{
+    return !(mon->flags & ERTS_ML_FLG_TARGET);
+}
+
+ERTS_GLB_INLINE int
+erts_monitor_is_in_table(ErtsMonitor *mon)
+{
+    return !!(mon->flags & ERTS_ML_FLG_IN_TABLE);
+}
+
+ERTS_GLB_INLINE void
+erts_monitor_list_insert(ErtsMonitor **list, ErtsMonitor *mon)
+{
+    erts_ml_dl_list_insert__((ErtsMonLnkNode **) list, (ErtsMonLnkNode *) mon);
+}
+
+ERTS_GLB_INLINE void
+erts_monitor_list_delete(ErtsMonitor **list, ErtsMonitor *mon)
+{
+    erts_ml_dl_list_delete__((ErtsMonLnkNode **) list, (ErtsMonLnkNode *) mon);
+}
+
+ERTS_GLB_INLINE ErtsMonitor *
+erts_monitor_list_first(ErtsMonitor *list)
+{
+    return (ErtsMonitor *) erts_ml_dl_list_first__((ErtsMonLnkNode *) list);
+}
+
+ERTS_GLB_INLINE ErtsMonitor *
+erts_monitor_list_last(ErtsMonitor *list)
+{
+    return (ErtsMonitor *) erts_ml_dl_list_last__((ErtsMonLnkNode *) list);
+}
+
+#ifdef ERTS_ML_DEBUG
+extern size_t erts_monitor_origin_offset;
+extern size_t erts_monitor_origin_key_offset;
+extern size_t erts_monitor_target_offset;
+extern size_t erts_monitor_target_key_offset;
+extern size_t erts_monitor_node_key_offset;
+#endif
+
+ERTS_GLB_INLINE ErtsMonitorData *
+erts_monitor_to_data(ErtsMonitor *mon)
+{
+    ErtsMonitorData *mdp = erts_ml_node_to_main_struct__((ErtsMonLnkNode *) mon);
+
+#ifdef ERTS_ML_DEBUG
+    ERTS_ML_ASSERT(!(mdp->origin.flags & ERTS_ML_FLG_TARGET));
+    ERTS_ML_ASSERT(erts_monitor_origin_offset == (size_t) mdp->origin.offset);
+    ERTS_ML_ASSERT(!!(mdp->target.flags & ERTS_ML_FLG_TARGET));
+    ERTS_ML_ASSERT(erts_monitor_target_offset == (size_t) mdp->target.offset);
+    if (mon->type == ERTS_MON_TYPE_NODE || mon->type == ERTS_MON_TYPE_NODES
+        || mon->type == ERTS_MON_TYPE_SUSPEND) {
+        ERTS_ML_ASSERT(erts_monitor_node_key_offset == (size_t) mdp->origin.key_offset);
+        ERTS_ML_ASSERT(erts_monitor_node_key_offset == (size_t) mdp->target.key_offset);
+    }
+    else {
+        ERTS_ML_ASSERT(erts_monitor_origin_key_offset == (size_t) mdp->origin.key_offset);
+        ERTS_ML_ASSERT(erts_monitor_target_key_offset == (size_t) mdp->target.key_offset);
+    }
+#endif
+
+    return mdp;
+}
+
+ERTS_GLB_INLINE void
+erts_monitor_release(ErtsMonitor *mon)
+{
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+    ERTS_ML_ASSERT(!(mon->flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&mdp->refc) > 0);
+
+    if (erts_atomic32_dec_read_nob(&mdp->refc) == 0)
+        erts_monitor_destroy__(mdp);
+}
+
+ERTS_GLB_INLINE void
+erts_monitor_release_both(ErtsMonitorData *mdp)
+{
+    ERTS_ML_ASSERT((mdp->origin.flags & ERTS_ML_FLGS_SAME)
+                   == (mdp->target.flags & ERTS_ML_FLGS_SAME));
+    ERTS_ML_ASSERT(!(mdp->origin.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(!(mdp->target.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&mdp->refc) >= 2);
+
+    if (erts_atomic32_add_read_nob(&mdp->refc, (erts_aint32_t) -2) == 0)
+        erts_monitor_destroy__(mdp);
+}
+
+ERTS_GLB_INLINE int
+erts_monitor_dist_insert(ErtsMonitor *mon, ErtsMonLnkDist *dist)
+{
+    ErtsMonitorDataExtended *mdep;
+    int insert;
+
+    ERTS_ML_ASSERT(mon->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(mon->type == ERTS_MON_TYPE_DIST_PROC
+                   || mon->type == ERTS_MON_TYPE_NODE);
+
+    mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+
+    ERTS_ML_ASSERT(!mdep->dist);
+    ERTS_ML_ASSERT(dist);
+    mdep->dist = dist;
+
+    erts_mon_link_dist_inc_refc(dist);
+
+    erts_mtx_lock(&dist->mtx);
+
+    insert = dist->alive;
+    if (insert) {
+        if ((mon->flags & (ERTS_ML_FLG_NAME
+                           | ERTS_ML_FLG_TARGET)) == ERTS_ML_FLG_NAME)
+            erts_monitor_tree_insert(&dist->orig_name_monitors, mon);
+        else
+            erts_monitor_list_insert(&dist->monitors, mon);
+    }
+
+    erts_mtx_unlock(&dist->mtx);
+
+    return insert;
+}
+
+ERTS_GLB_INLINE int
+erts_monitor_dist_delete(ErtsMonitor *mon)
+{
+    ErtsMonitorDataExtended *mdep;
+    ErtsMonLnkDist *dist;
+    Uint16 flags;
+    int delete;
+
+    ERTS_ML_ASSERT(mon->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(mon->type == ERTS_MON_TYPE_DIST_PROC
+                   || mon->type == ERTS_MON_TYPE_NODE);
+
+    mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+    dist = mdep->dist;
+    ERTS_ML_ASSERT(dist);
+
+    erts_mtx_lock(&dist->mtx);
+
+    flags = mon->flags;
+    delete = !!dist->alive & !!(flags & ERTS_ML_FLG_IN_TABLE);
+    if (delete) {
+        if ((flags & (ERTS_ML_FLG_NAME
+                      | ERTS_ML_FLG_TARGET)) == ERTS_ML_FLG_NAME)
+            erts_monitor_tree_delete(&dist->orig_name_monitors, mon);
+        else
+            erts_monitor_list_delete(&dist->monitors, mon);
+    }
+
+    erts_mtx_unlock(&dist->mtx);
+
+    return delete;
+}
+
+
+#endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
+/* suspend monitors... */
+ErtsMonitorSuspend *erts_monitor_suspend_create(Eterm pid);
+ErtsMonitorSuspend *erts_monitor_suspend_tree_lookup_create(ErtsMonitor **root,
+                                                            int *created,
+                                                            Eterm pid);
+void erts_monitor_suspend_destroy(ErtsMonitorSuspend *msp);
+
+ERTS_GLB_INLINE ErtsMonitorSuspend *erts_monitor_suspend(ErtsMonitor *mon);
+
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE ErtsMonitorSuspend *erts_monitor_suspend(ErtsMonitor *mon)
+{
+    ERTS_ML_ASSERT(!mon || mon->type == ERTS_MON_TYPE_SUSPEND);
+    return (ErtsMonitorSuspend *) mon;
+}
+
+#endif
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\
+ * Link Operations                                                           *
+\*                                                                           */
+
+typedef struct ErtsMonLnkNode__ ErtsLink;
+
+typedef struct {
+    ErtsLink a;
+    ErtsLink b;
+    erts_atomic32_t refc;
+} ErtsLinkData;
+
+typedef struct {
+    ErtsLinkData ld;
+    struct erl_off_heap_header *ohhp;
+    ErtsMonLnkDist *dist;
+    Eterm heap[1]; /* heap start... */
+} ErtsLinkDataExtended;
+
+/*
+ * --- Link tree operations ---
+ */
+
+/**
+ *
+ * @brief Lookup a link in a link tree
+ *
+ *
+ * @param[in]     root          Pointer to root of link tree
+ *
+ * @param[in]     key           Key of link to lookup
+ *
+ * @returns                     Pointer to a link with the
+ *                              key 'key', or NULL if no such
+ *                              link was found
+ *
+ */
+ErtsLink *erts_link_tree_lookup(ErtsLink *root, Eterm item);
+
+/**
+ *
+ * @brief Lookup or insert a link in a link tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is not part of any tree or list
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     lnk           Link to insert if no link
+ *                              with the same key already exists
+ *
+ * @returns                     Pointer to a link with the
+ *                              key 'key'. If no link with the key
+ *                              'key' was found and 'lnk' was inserted
+ *                              'lnk' is returned.
+ *
+ */
+ErtsLink *erts_link_tree_lookup_insert(ErtsLink **root, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Lookup or create a link in a link tree.
+ *
+ * Looks up a link with the key 'other' in the link tree. If it is not
+ * found, creates and insert a link with the key 'other'.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[out]    created       Pointer to integer. The integer is set to
+ *                              a non-zero value if no link with key
+ *                              'other' was found, and a new link
+ *                              was created. If a link was found, it
+ *                              is set to zero.
+ *
+ * @param[in]     type          Type of link
+ *
+ * @param[in]     this          Id of this entity
+ *
+ * @param[in]     other         Id of other entity
+ *
+ */
+ErtsLink *erts_link_tree_lookup_create(ErtsLink **root, int *created,
+                                       Uint16 type, Eterm this, Eterm other);
+
+/**
+ *
+ * @brief Insert a link in a link tree
+ *
+ * When the funcion is called it is assumed that:
+ * - no links with the same key that 'lnk' exist in the tree
+ * - 'lnk' is not part of any list of tree
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     lnk           Link to insert.
+ *
+ */
+void erts_link_tree_insert(ErtsLink **root, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Replace a link in a link tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'old' link and 'new' link have exactly the same key
+ * - 'old' link is part of the tree
+ * - 'new' link is not part of any tree or list
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     old           Link to remove from the tree
+ *
+ * @param[in]     new           Link to insert into the tree
+ *
+ */
+void erts_link_tree_replace(ErtsLink **root, ErtsLink *old, ErtsLink *new);
+
+/**
+ *
+ * @brief Replace a link in a link tree if key already exist based on adress
+ *
+ * Inserts the link 'lnk' in the tree if no link with the same key
+ * already exists in tree. If a link with the same key exists in
+ * the tree and 'lnk' has a lower address than the link in the
+ * tree, the existing link in the tree is replaced by 'lnk'.
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is not part of any tree or list
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     lnk           Link to insert into the tree
+ *
+ * @returns                     A pointer to the link that is not part
+ *                              of the tree after this operation.
+ *
+ */
+ERTS_GLB_INLINE ErtsLink *erts_link_tree_insert_addr_replace(ErtsLink **root,
+                                                             ErtsLink *lnk);
+
+/**
+ *
+ * @brief Delete a link from a link tree
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is part of the tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     lnk           Link to remove from the tree
+ *
+ */
+void erts_link_tree_delete(ErtsLink **root, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Delete a link from a link tree based on key
+ *
+ * If link 'lnk' is in the tree, 'lnk' is deleted from the tree.
+ * If link 'lnk' is not in the tree, another link with the same
+ * key as 'lnk' is deleted from the tree if such a link exist.
+ *
+ * When the funcion is called it is assumed that:
+ * - if 'lnk' link is part of a tree or list, it is part of this tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     lnk           Link to remove from the tree
+ *
+ * @returns                     A pointer to the link that was deleted
+ *                              from the tree, or NULL in case no link
+ *                              was deleted from the tree
+ *
+ */
+ERTS_GLB_INLINE ErtsLink *erts_link_tree_key_delete(ErtsLink **root, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Call a function for each link in a link tree
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * @param[in]     root          Pointer to root of link tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_link_tree_foreach(ErtsLink *root,
+                            void (*func)(ErtsLink *, void *),
+                            void *arg);
+
+/**
+ *
+ * @brief Call a function for each link in a link tree. Yield if lots
+ *        of links exist.
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetedly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]     root          Pointer to root of link tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of links to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all links has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_link_tree_foreach_yielding(ErtsLink *root,
+                                    void (*func)(ErtsLink *, void *),
+                                    void *arg,
+                                    void **vyspp,
+                                    Sint limit);
+
+/**
+ *
+ * @brief Delete all links from a link tree and call a function for
+ *        each link
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_link_tree_foreach_delete(ErtsLink **root,
+                                   void (*func)(ErtsLink *, void *),
+                                   void *arg);
+
+/**
+ *
+ * @brief Delete all links from a link tree and call a function for
+ *        each link
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetededly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] root          Pointer to pointer to root of link tree
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of links to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all links has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_link_tree_foreach_delete_yielding(ErtsLink **root,
+                                           void (*func)(ErtsLink *, void *),
+                                           void *arg,
+                                           void **vyspp,
+                                           Sint limit);
+
+/* 
+ * --- Link list operations ---
+ */
+
+/**
+ *
+ * @brief Insert a link in a link list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is not part of any list or tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to link list
+ *
+ * @param[in]     lnk           Link to insert
+ *
+ */
+ERTS_GLB_INLINE void erts_link_list_insert(ErtsLink **list, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Delete a link from a link list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is part of the list
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to link list
+ *
+ * @param[in]     lnk           Link to remove from the list
+ *
+ */
+ERTS_GLB_INLINE void erts_link_list_delete(ErtsLink **list, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Get a pointer to first link in a link list
+ *
+ * The link will still remain in the list after the return
+ *
+ * @param[in] list              Pointer to link list
+ *
+ * @returns                     Pointer to first link in list if
+ *                              list is not empty. If list is empty
+ *                              NULL is returned.
+ *
+ */
+ERTS_GLB_INLINE ErtsLink *erts_link_list_first(ErtsLink *list);
+
+/**
+ *
+ * @brief Get a pointer to last link in a link list
+ *
+ * The link will still remain in the list after the return
+ *
+ * @param[in] list              Pointer to link list
+ *
+ * @returns                     Pointer to last link in list if
+ *                              list is not empty. If list is empty
+ *                              NULL is returned.
+ *
+ */
+ERTS_GLB_INLINE ErtsLink *erts_link_list_last(ErtsLink *list);
+
+/**
+ *
+ * @brief Call a function for each link in a link list
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'list'.
+ *
+ * @param[in]     list          Pointer to root of link list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_link_list_foreach(ErtsLink *list,
+                            void (*func)(ErtsLink *, void *),
+                            void *arg);
+
+/**
+ *
+ * @brief Call a function for each link in a link list. Yield
+ *        if lots of links exist.
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetedly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first call
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]     list          Pointer to link list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of links to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all links has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_link_list_foreach_yielding(ErtsLink *list,
+                                    void (*func)(ErtsLink *, void *),
+                                    void *arg,
+                                    void **vyspp,
+                                    Sint limit);
+
+/**
+ *
+ * @brief Delete all links from a link list and call a function for
+ *        each link
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * @param[in,out] list          Pointer to pointer to link list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ */
+void erts_link_list_foreach_delete(ErtsLink **list,
+                                   void (*func)(ErtsLink *, void *),
+                                   void *arg);
+
+/**
+ *
+ * @brief Delete all links from a link list and call a function for
+ *        each link
+ *
+ * The funcion 'func' will be called with a pointer to a link
+ * as first argument and 'arg' as second argument for each link
+ * in the tree referred to by 'root'.
+ *
+ * It is assumed that:
+ * - *yspp equals NULL on first call
+ * - this function is repetededly called with *yspp set
+ *   as set when previous call returned until a non-zero
+ *   value is returned.
+ * - no modifications are made on the tree between first
+ *   and the call that returns a non-zero value
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in,out] list          Pointer to pointer to link list
+ *
+ * @param[in]     func          Pointer to function to call
+ *
+ * @param[in]     arg           Argument to pass as second argument in
+ *                              calls to 'func'
+ *
+ * @param[in,out] vyspp         Pointer to a pointer to an internal state
+ *                              used by this function. At initial call
+ *                              *yspp should be NULL. When done *yspp
+ *                              will be NULL.
+ *
+ * @param[in]     limit         Maximum amount of links to process
+ *                              before yielding.
+ *
+ * @returns                     A non-zero value when all links has been
+ *                              processed, and zero when more work is needed.
+ *
+ */
+int erts_link_list_foreach_delete_yielding(ErtsLink **list,
+                                           void (*func)(ErtsLink *, void *),
+                                           void *arg,
+                                           void **vyspp,
+                                           Sint limit);
+
+/*
+ * --- Misc link operations ---
+ */
+
+/**
+ *
+ * @brief Create a link
+ *
+ * Can create all types of links
+ *
+ * When the funcion is called it is assumed that:
+ * - 'ref' is an internal ordinary reference if type is ERTS_MON_TYPE_PROC,
+ *   ERTS_MON_TYPE_PORT, ERTS_MON_TYPE_TIME_OFFSET, or ERTS_MON_TYPE_RESOURCE
+ * - 'ref' is NIL if type is ERTS_MON_TYPE_NODE or ERTS_MON_TYPE_NODES
+ * - 'ref' is and ordinary internal reference or an external reference if
+ *   type is ERTS_MON_TYPE_DIST_PROC
+ * - 'name' is an atom or NIL if type is ERTS_MON_TYPE_PROC,
+ *   ERTS_MON_TYPE_PORT, or ERTS_MON_TYPE_DIST_PROC
+ * - 'name is NIL if type is ERTS_MON_TYPE_TIME_OFFSET, ERTS_MON_TYPE_RESOURCE,
+ *   ERTS_MON_TYPE_NODE, or ERTS_MON_TYPE_NODES
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]     type          ERTS_MON_TYPE_PROC, ERTS_MON_TYPE_PORT,
+ *                              ERTS_MON_TYPE_TIME_OFFSET, ERTS_MON_TYPE_DIST_PROC,
+ *                              ERTS_MON_TYPE_RESOURCE, ERTS_MON_TYPE_NODE,
+ *                              or ERTS_MON_TYPE_NODES
+ *
+ * @param[in]     a             The key of entity a. Link structure a will
+ *                              have field other.item set to 'b'.
+ *
+ * @param[in]     b             The key of entity b. Link structure b will
+ *                              have field other.item set to 'a'.
+ *
+ */
+ErtsLinkData *erts_link_create(Uint16 type, Eterm a, Eterm b);
+
+/**
+ *
+ * @brief Get pointer to link data structure
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @returns                     Pointer to link data structure
+ *
+ */
+ERTS_GLB_INLINE ErtsLinkData *erts_link_to_data(ErtsLink *lnk);
+
+/**
+ *
+ * @brief Get pointer to the other link structure part of the link
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @param[out]   ldpp           Pointer to pointer to link data structure,
+ *                              if a non-NULL value is passed in the call
+ *
+ * @returns                     Pointer to other link
+ *
+ */
+ERTS_GLB_INLINE ErtsLink *erts_link_to_other(ErtsLink *lnk, ErtsLinkData **ldpp);
+
+/**
+ *
+ * @brief Check if link is in tree or list
+ *
+ * @param[in]    lnk            Pointer to lnk to check
+ *
+ * @returns                     A non-zero value if in tree or list;
+ *                              otherwise zero
+ *
+ */
+ERTS_GLB_INLINE int erts_link_is_in_table(ErtsLink *lnk);
+
+/**
+ *
+ * @brief Release link
+ *
+ * When both link halves part of the link have been released the link
+ * structure will be deallocated.
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is not part of any list or tree
+ * - 'lnk' is not referred to by any other structures
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ */
+ERTS_GLB_INLINE void erts_link_release(ErtsLink *lnk);
+
+/**
+ *
+ * @brief Release both link halves of a link simultaneously
+ *
+ * Release both halves of a link simultaneously and deallocate
+ * the structure.
+ *
+ * When the funcion is called it is assumed that:
+ * - Neither of the parts of the link are part of any list or tree
+ * - Neither of the parts of the link or the link data structure
+ *   are referred to by any other structures
+ * If the above are not true, bad things will happen.
+ *
+ * @param[in]    mdp            Pointer to link data structure
+ *
+ */
+ERTS_GLB_INLINE void erts_link_release_both(ErtsLinkData *ldp);
+
+/**
+ *
+ * @brief Insert link in dist link list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link is not part of any list or tree
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @param[in]    dist           Pointer to dist structure
+ *
+ * @returns                     A non-zero value if inserted;
+ *                              otherwise, zero. The link
+ *                              is not inserted if the dist
+ *                              structure has been set in a
+ *                              dead state.
+ *
+ */
+ERTS_GLB_INLINE int erts_link_dist_insert(ErtsLink *lnk, ErtsMonLnkDist *dist);
+
+/**
+ *
+ * @brief Delete link from dist link list
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' link earler has been inserted into 'dist'
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @param[in]    dist           Pointer to dist structure
+ *
+ * @returns                     A non-zero value if deleted;
+ *                              otherwise, zero. The link
+ *                              is not deleted if the dist
+ *                              structure has been set in a
+ *                              dead state or if it has already
+ *                              been deleted.
+ *
+ */
+ERTS_GLB_INLINE int erts_link_dist_delete(ErtsLink *lnk);
+
+/**
+ *
+ * @brief Set dead dist structure on link
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @param[in]    nodename       Name of remote node
+ *
+ */
+void
+erts_link_set_dead_dist(ErtsLink *lnk, Eterm nodename);
+
+/**
+ *
+ * @brief Get charged size of link
+ *
+ * If the other side of the link has been released, the
+ * whole size of the link data structure is returned; otherwise,
+ * half of the size is returned.
+ *
+ * When the funcion is called it is assumed that:
+ * - 'lnk' has not been released
+ * If the above is not true, bad things will happen.
+ *
+ * @param[in]    lnk            Pointer to link
+ *
+ * @returns                     Charged size in bytes
+ *
+ */
+Uint erts_link_size(ErtsLink *lnk);
+
+/* internal function... */
+void erts_link_destroy__(ErtsLinkData *ldp);
+
+/* implementations for globally inlined link functions... */
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+#ifdef ERTS_ML_DEBUG
+extern size_t erts_link_a_offset;
+extern size_t erts_link_b_offset;
+extern size_t erts_link_key_offset;
+#endif
+
+ERTS_GLB_INLINE ErtsLinkData *
+erts_link_to_data(ErtsLink *lnk)
+{
+    ErtsLinkData *ldp = erts_ml_node_to_main_struct__((ErtsMonLnkNode *) lnk);
+
+#ifdef ERTS_ML_DEBUG
+    ERTS_ML_ASSERT(erts_link_a_offset == (size_t) ldp->a.offset);
+    ERTS_ML_ASSERT(erts_link_key_offset == (size_t) ldp->a.key_offset);
+    ERTS_ML_ASSERT(erts_link_b_offset == (size_t) ldp->b.offset);
+    ERTS_ML_ASSERT(erts_link_key_offset == (size_t) ldp->b.key_offset);
+#endif
+
+    return ldp;
+}
+
+ERTS_GLB_INLINE ErtsLink *
+erts_link_to_other(ErtsLink *lnk, ErtsLinkData **ldpp)
+{
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+    if (ldpp)
+        *ldpp = ldp;
+    return lnk == &ldp->a ? &ldp->b : &ldp->a;
+}
+
+ERTS_GLB_INLINE int
+erts_link_is_in_table(ErtsLink *lnk)
+{
+    return !!(lnk->flags & ERTS_ML_FLG_IN_TABLE);
+}
+
+ERTS_GLB_INLINE void
+erts_link_list_insert(ErtsLink **list, ErtsLink *lnk)
+{
+    erts_ml_dl_list_insert__((ErtsMonLnkNode **) list, (ErtsMonLnkNode *) lnk);
+}
+
+ERTS_GLB_INLINE void
+erts_link_list_delete(ErtsLink **list, ErtsLink *lnk)
+{
+    erts_ml_dl_list_delete__((ErtsMonLnkNode **) list, (ErtsMonLnkNode *) lnk);
+}
+
+ERTS_GLB_INLINE ErtsLink *
+erts_link_list_first(ErtsLink *list)
+{
+    return (ErtsLink *) erts_ml_dl_list_first__((ErtsMonLnkNode *) list);
+}
+
+ERTS_GLB_INLINE ErtsLink *
+erts_link_list_last(ErtsLink *list)
+{
+    return (ErtsLink *) erts_ml_dl_list_last__((ErtsMonLnkNode *) list);
+}
+
+ERTS_GLB_INLINE void
+erts_link_release(ErtsLink *lnk)
+{
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+    ERTS_ML_ASSERT(!(lnk->flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&ldp->refc) > 0);
+    if (erts_atomic32_dec_read_nob(&ldp->refc) == 0)
+        erts_link_destroy__(ldp);
+}
+
+ERTS_GLB_INLINE void
+erts_link_release_both(ErtsLinkData *ldp)
+{
+    ERTS_ML_ASSERT(!(ldp->a.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(!(ldp->b.flags & ERTS_ML_FLG_IN_TABLE));
+    ERTS_ML_ASSERT(erts_atomic32_read_nob(&ldp->refc) >= 2);
+    if (erts_atomic32_add_read_nob(&ldp->refc, (erts_aint32_t) -2) == 0)
+        erts_link_destroy__(ldp);
+}
+
+ERTS_GLB_INLINE ErtsLink *
+erts_link_tree_insert_addr_replace(ErtsLink **root, ErtsLink *lnk)
+{
+    ErtsLink *lnk2 = erts_link_tree_lookup_insert(root, lnk);
+    if (!lnk2)
+        return NULL;
+    if (lnk2 < lnk)
+        return lnk;
+    erts_link_tree_replace(root, lnk2, lnk);
+    return lnk2;
+}
+
+ERTS_GLB_INLINE ErtsLink *
+erts_link_tree_key_delete(ErtsLink **root, ErtsLink *lnk)
+{
+    ErtsLink *dlnk;
+    if (erts_link_is_in_table(lnk))
+        dlnk = lnk;
+    else
+        dlnk = erts_link_tree_lookup(*root, lnk->other.item);
+    if (dlnk)
+        erts_link_tree_delete(root, dlnk);
+    return dlnk;
+}
+
+ERTS_GLB_INLINE int
+erts_link_dist_insert(ErtsLink *lnk, ErtsMonLnkDist *dist)
+{
+    ErtsLinkDataExtended *ldep;
+    int insert;
+
+    ERTS_ML_ASSERT(lnk->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(lnk->type == ERTS_LNK_TYPE_DIST_PROC);
+
+    ldep = (ErtsLinkDataExtended *) erts_link_to_data(lnk);
+
+    ERTS_ML_ASSERT(!ldep->dist);
+    ERTS_ML_ASSERT(dist);
+    ldep->dist = dist;
+
+    erts_mon_link_dist_inc_refc(dist);
+
+    erts_mtx_lock(&dist->mtx);
+
+    insert = dist->alive;
+    if (insert)
+        erts_link_list_insert(&dist->links, lnk);
+
+    erts_mtx_unlock(&dist->mtx);
+
+    return insert;
+}
+
+ERTS_GLB_INLINE int
+erts_link_dist_delete(ErtsLink *lnk)
+{
+    ErtsLinkDataExtended *ldep;
+    ErtsMonLnkDist *dist;
+    int delete;
+
+    ERTS_ML_ASSERT(lnk->flags & ERTS_ML_FLG_EXTENDED);
+    ERTS_ML_ASSERT(lnk->type == ERTS_LNK_TYPE_DIST_PROC);
+
+    ldep = (ErtsLinkDataExtended *) erts_link_to_data(lnk);
+    dist = ldep->dist;
+    if (!dist)
+        return -1;
+
+    erts_mtx_lock(&dist->mtx);
+
+    delete = !!dist->alive & !!(lnk->flags & ERTS_ML_FLG_IN_TABLE);
+    if (delete)
+        erts_link_list_delete(&dist->links, lnk);
+
+    erts_mtx_unlock(&dist->mtx);
+
+    return delete;
+}
+
+
+#endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
+#endif /* ERL_MONITOR_LINK_H__ */
diff --git a/erts/emulator/beam/erl_monitors.c b/erts/emulator/beam/erl_monitors.c
deleted file mode 100644
index 3994800ba7..0000000000
--- a/erts/emulator/beam/erl_monitors.c
+++ /dev/null
@@ -1,1075 +0,0 @@
-/*
- * %CopyrightBegin%
- * 
- * Copyright Ericsson AB 2004-2017. All Rights Reserved.
- * 
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * 
- * %CopyrightEnd%
- */
-
-/**************************************************************************
- * Monitors and links data structure manipulation.
- * Monitors and links are organized as AVL trees with the reference as 
- * key in the monitor case and the pid of the linked process as key in the 
- * link case. Lookups the order of the references is somewhat special. Local 
- * references are strictly smaller than remote references and are sorted 
- * by inlined comparison functionality. Remote references are handled by the
- * usual cmp function.
- * Each Monitor is tagged with different tags depending on which end of the 
- * monitor it is.
- * A monitor is removed either explicitly by reference or all monitors are 
- * removed when the process exits. No need to access the monitor by pid.
- **************************************************************************/ 
-
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-
-#include "sys.h"
-#include "erl_vm.h"
-#include "global.h"
-#include "erl_process.h"
-#include "error.h"
-#include "erl_db.h"
-#include "bif.h"
-#include "big.h"
-#include "erl_monitors.h"
-#include "erl_bif_unique.h"
-
-#define STACK_NEED 50
-#define MAX_MONITORS 0xFFFFFFFFUL
-
-#define DIR_LEFT 0
-#define DIR_RIGHT 1
-#define DIR_END 2 
-
-static erts_smp_atomic_t tot_link_lh_size;
-
-/* Implements the sort order in monitor trees, which is different from 
-   the ordinary term order.
-   No short local ref's should ever exist (the ref is created by the bif's 
-   in runtime), therefore:
-   All local ref's are less than external ref's
-   Local ref's are inline-compared,
-   External ref's are compared by cmp */
-
-#if 0   
-#define CMP_MON_REF(Ref1,Ref2) \
-cmp((Ref1),(Ref2)) /* XXX, the inline comparision yet to be done */
-#else
-#define CMP_MON_REF(Ref1,Ref2) cmp_mon_ref((Ref1),(Ref2))
-#endif
-
-static ERTS_INLINE int cmp_mon_ref(Eterm ref1, Eterm ref2) 
-{
-    Eterm *b1, *b2;
-
-
-    b1 = boxed_val(ref1);
-    b2 = boxed_val(ref2);
-    if (is_ref_thing_header(*b1)) {
-	if (is_ref_thing_header(*b2)) {
-	    Uint32 *num1, *num2;
-	    if (is_ordinary_ref_thing(b1)) {
-		ErtsORefThing *rtp = (ErtsORefThing *) b1;
-		num1 = rtp->num;
-	    }
-	    else {
-		ErtsMRefThing *mrtp = (ErtsMRefThing *) b1;
-		num1 = mrtp->mb->refn;
-	    }
-	    if (is_ordinary_ref_thing(b2)) {
-		ErtsORefThing *rtp = (ErtsORefThing *) b2;
-		num2 = rtp->num;
-	    }
-	    else {
-		ErtsMRefThing *mrtp = (ErtsMRefThing *) b2;
-		num2 = mrtp->mb->refn;
-	    }
-	    return erts_internal_ref_number_cmp(num1, num2);
-	}
-	return -1;
-    }
-    if (is_ref_thing_header(*b2)) {
-	return 1;
-    }
-    return CMP(ref1,ref2);
-}
-	    
-#define CP_LINK_VAL(To, Hp, From)				\
-do {								\
-    if (is_immed(From))						\
-	(To) = (From);						\
-    else {							\
-	Uint i__;						\
-	Uint len__;						\
-	ASSERT((Hp));						\
-	ASSERT(is_internal_ordinary_ref((From))                 \
-               || is_external((From)));                         \
-	(To) = make_boxed((Hp));				\
-	len__ = thing_arityval(*boxed_val((From))) + 1;		\
-	for(i__ = 0; i__ < len__; i__++)			\
-	    (*((Hp)++)) = boxed_val((From))[i__];		\
-	if (is_external((To))) {				\
-	    external_thing_ptr((To))->next = NULL;		\
-	    erts_smp_refc_inc(&(external_thing_ptr((To))->node->refc), 2);\
-	}							\
-    }								\
-} while (0)
-
-static ErtsMonitor *create_monitor(Uint type, Eterm ref, UWord entity, Eterm name)
-{
-     Uint mon_size = ERTS_MONITOR_SIZE;
-     ErtsMonitor *n;
-     Eterm *hp;
-
-     mon_size += NC_HEAP_SIZE(ref);
-     if (type != MON_NIF_TARGET && is_not_immed(entity)) {
-	 mon_size += NC_HEAP_SIZE(entity);
-     }
-
-     if (mon_size <= ERTS_MONITOR_SH_SIZE) {
-	 n = (ErtsMonitor *) erts_alloc(ERTS_ALC_T_MONITOR_SH,
-					mon_size*sizeof(Uint));
-     } else {
-	 n = (ErtsMonitor *) erts_alloc(ERTS_ALC_T_MONITOR_LH,
-					mon_size*sizeof(Uint));
-	 erts_smp_atomic_add_nob(&tot_link_lh_size, mon_size*sizeof(Uint));
-     } 
-     hp = n->heap;
-
-
-     n->left = n->right = NULL; /* Always the same initial value*/
-     n->type = (Uint16) type;
-     n->balance = 0;            /* Always the same initial value */
-     n->name = name; /* atom() or [] */
-     CP_LINK_VAL(n->ref, hp, ref); /*XXX Unnecessary check, never immediate*/
-     if (type == MON_NIF_TARGET)
-         n->u.resource = (ErtsResource*)entity;
-     else
-         CP_LINK_VAL(n->u.pid, hp, (Eterm)entity);
-
-     return n;
-}
-
-static ErtsLink *create_link(Uint type, Eterm pid)
-{
-     Uint lnk_size = ERTS_LINK_SIZE;
-     ErtsLink *n;
-     Eterm *hp;
-
-     if (is_not_immed(pid)) {
-	 lnk_size += NC_HEAP_SIZE(pid);
-     }
-
-     if (lnk_size <= ERTS_LINK_SH_SIZE) {
-	 n = (ErtsLink *) erts_alloc(ERTS_ALC_T_NLINK_SH,
-				     lnk_size*sizeof(Uint));
-     } else {
-	 n = (ErtsLink *) erts_alloc(ERTS_ALC_T_NLINK_LH,
-				     lnk_size*sizeof(Uint));
-	 erts_smp_atomic_add_nob(&tot_link_lh_size, lnk_size*sizeof(Uint));
-     } 
-     hp = n->heap;
-
-
-     n->left = n->right = NULL; /* Always the same initial value*/
-     n->type = (Uint16) type;
-     n->balance = 0;            /* Always the same initial value */
-     if (n->type == LINK_NODE) {
-	 ERTS_LINK_REFC(n) = 0;
-     } else {
-	 ERTS_LINK_ROOT(n) = NULL; 
-     }
-     CP_LINK_VAL(n->pid, hp, pid);
-
-     return n;
-}
-
-#undef CP_LINK_VAL
-
-static ErtsSuspendMonitor *create_suspend_monitor(Eterm pid)
-{
-    ErtsSuspendMonitor *smon = erts_alloc(ERTS_ALC_T_SUSPEND_MON,
-					  sizeof(ErtsSuspendMonitor));
-    smon->left = smon->right = NULL; /* Always the same initial value */
-    smon->balance = 0;               /* Always the same initial value */
-    smon->pending = 0;
-    smon->active = 0;
-    smon->pid = pid;
-    return smon;
-}
-
-void
-erts_init_monitors(void)
-{
-    erts_smp_atomic_init_nob(&tot_link_lh_size, 0);
-}
-
-Uint
-erts_tot_link_lh_size(void)
-{
-    return (Uint) erts_smp_atomic_read_nob(&tot_link_lh_size);
-}
-
-void erts_destroy_monitor(ErtsMonitor *mon)
-{
-    Uint mon_size = ERTS_MONITOR_SIZE;
-    ErlNode *node;
-
-    ASSERT(is_not_immed(mon->ref));
-    mon_size +=  NC_HEAP_SIZE(mon->ref);
-    if (is_external(mon->ref)) {
-	node = external_thing_ptr(mon->ref)->node;
-	erts_deref_node_entry(node);
-    }
-    if (mon->type != MON_NIF_TARGET && is_not_immed(mon->u.pid)) {
-	mon_size += NC_HEAP_SIZE(mon->u.pid);
-	if (is_external(mon->u.pid)) {
-	    node = external_thing_ptr(mon->u.pid)->node;
-	    erts_deref_node_entry(node);
-	}
-    }
-    if (mon_size <= ERTS_MONITOR_SH_SIZE) {
-	erts_free(ERTS_ALC_T_MONITOR_SH, (void *) mon);
-    } else {
-	erts_free(ERTS_ALC_T_MONITOR_LH, (void *) mon);
-	erts_smp_atomic_add_nob(&tot_link_lh_size, -1*mon_size*sizeof(Uint));
-    }
-}
-    
-void erts_destroy_link(ErtsLink *lnk)
-{
-    Uint lnk_size = ERTS_LINK_SIZE;
-    ErlNode *node;
-
-    ASSERT(lnk->type == LINK_NODE || ERTS_LINK_ROOT(lnk) == NULL);
-
-    if (is_not_immed(lnk->pid)) {
-	lnk_size += NC_HEAP_SIZE(lnk->pid);
-	if (is_external(lnk->pid)) {
-	    node = external_thing_ptr(lnk->pid)->node;
-	    erts_deref_node_entry(node);
-	}
-    }
-    if (lnk_size <= ERTS_LINK_SH_SIZE) {
-	erts_free(ERTS_ALC_T_NLINK_SH, (void *) lnk);
-    } else {
-	erts_free(ERTS_ALC_T_NLINK_LH, (void *) lnk);
-	erts_smp_atomic_add_nob(&tot_link_lh_size, -1*lnk_size*sizeof(Uint));
-    }
-}
-
-void erts_destroy_suspend_monitor(ErtsSuspendMonitor *smon)
-{
-    erts_free(ERTS_ALC_T_SUSPEND_MON, smon);
-}
-     
-static void insertion_rotation(int dstack[], int dpos, 
-			       void *tstack[], int tpos, 
-			       int state) {
-    
-    ErtsMonitorOrLink **this;
-    ErtsMonitorOrLink *p1, *p2, *p;
-    int dir;
-
-    while (state && ( dir = dstack[--dpos] ) != DIR_END) {
-	this = tstack[--tpos];
-	p = *this;
-	if (dir == DIR_LEFT) {
-	    switch (p->balance) {
-	    case 1:
-		p->balance = 0;
-		state = 0;
-		break;
-	    case 0:
-		p->balance = -1;
-		break;
-	    case -1: /* The icky case */
-		p1 = p->left;
-		if (p1->balance == -1) { /* Single LL rotation */
-		    p->left = p1->right;
-		    p1->right = p;
-		    p->balance = 0;
-		    (*this) = p1;
-		} else { /* Double RR rotation */
-		    p2 = p1->right;
-		    p1->right = p2->left;
-		    p2->left = p1;
-		    p->left = p2->right;
-		    p2->right = p;
-		    p->balance = (p2->balance == -1) ? +1 : 0;
-		    p1->balance = (p2->balance == 1) ? -1 : 0;
-		    (*this) = p2;
-		}
-		(*this)->balance = 0;
-		state = 0;
-		break;
-	    }
-	} else { /* dir == DIR_RIGHT */
-	    switch (p->balance) {
-	    case -1:
-		p->balance = 0;
-		state = 0;
-		break;
-	    case 0:
-		p->balance = 1;
-		break;
-	    case 1:
-		p1 = p->right;
-		if (p1->balance == 1) { /* Single RR rotation */
-		    p->right = p1->left;
-		    p1->left = p;
-		    p->balance = 0;
-		    (*this) = p1;
-		} else { /* Double RL rotation */
-		    p2 = p1->left;
-		    p1->left = p2->right;
-		    p2->right = p1;
-		    p->right = p2->left;
-		    p2->left = p;
-		    p->balance = (p2->balance == 1) ? -1 : 0;
-		    p1->balance = (p2->balance == -1) ? 1 : 0;
-		    (*this) = p2;
-		}
-		(*this)->balance = 0; 
-		state = 0;
-		break;
-	    }
-	}
-    }
-}
-
-void erts_add_monitor(ErtsMonitor **root, Uint type, Eterm ref, UWord entity, 
-		      Eterm name)
-{
-    void *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsMonitor **this = root;
-    Sint c;
-
-    ASSERT(is_internal_ordinary_ref(ref) || is_external_ref(ref));
-  
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Found our place */
-	    state = 1;
-	    *this = create_monitor(type,ref,entity,name);
-	    break;
-	} else if ((c = CMP_MON_REF(ref,(*this)->ref)) < 0) { 
-	    /* go left */
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key is an error for monitors */
-	    erts_exit(ERTS_ERROR_EXIT,"Insertion of already present monitor!");
-	    break;
-	}
-    }
-    insertion_rotation(dstack, dpos, tstack, tpos, state);
-}
-
-
-/* Returns 0 if OK, < 0 if already present */
-int erts_add_link(ErtsLink **root, Uint type, Eterm pid)
-{
-    void *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsLink **this = root;
-    Sint c;
-  
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Found our place */
-	    state = 1;
-	    *this = create_link(type,pid);
-	    break;
-	} else if ((c = CMP(pid,(*this)->pid)) < 0) {
-	    /* go left */
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key is an error for monitors */
-	    return -1;
-	}
-    }
-    insertion_rotation(dstack, dpos, tstack, tpos, state);
-    return 0;
-}
-
-ErtsSuspendMonitor *
-erts_add_or_lookup_suspend_monitor(ErtsSuspendMonitor **root, Eterm pid)
-{
-    void *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsSuspendMonitor **this = root;
-    ErtsSuspendMonitor *res;
-    Sint c;
-  
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Found our place */
-	    state = 1;
-	    res = *this = create_suspend_monitor(pid);
-	    break;
-	} else if ((c = CMP(pid,(*this)->pid)) < 0) {
-	    /* go left */
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Already here... */
-	    ASSERT((*this)->pid == pid);
-	    return *this;
-	}
-    }
-    insertion_rotation(dstack, dpos, tstack, tpos, state);
-    return res;
-}
-
-
-/* Returns the new or old link structure */
-ErtsLink *erts_add_or_lookup_link(ErtsLink **root, Uint type, Eterm pid)
-{
-    void *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsLink **this = root;
-    Sint c;
-    ErtsLink *ret = NULL;
-
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Found our place */
-	    state = 1;
-	    *this = create_link(type,pid);
-	    ret = *this;
-	    break;
-	} else if ((c = CMP(pid,(*this)->pid)) < 0) {
-	    /* go left */
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key is an error for monitors */
-	    return *this;
-	}
-    }
-    insertion_rotation(dstack, dpos, tstack, tpos, state);
-    return ret;
-}
-
-
-/*
- * Deletion helpers
- */
-static int balance_left(ErtsMonitorOrLink **this) 
-{
-    ErtsMonitorOrLink *p, *p1, *p2;
-    int b1, b2, h = 1;
-    
-    p = *this;
-    switch (p->balance) {
-    case -1:
-	p->balance = 0;
-	break;
-    case 0:
-	p->balance = 1;
-	h = 0;
-	break;
-    case 1:
-	p1 = p->right;
-	b1 = p1->balance;
-	if (b1 >= 0) { /* Single RR rotation */
-	    p->right = p1->left;
-	    p1->left = p;
-	    if (b1 == 0) {
-		p->balance = 1;
-		p1->balance = -1;
-		h = 0;
-	    } else {
-		p->balance = p1->balance = 0;
-	    }
-	    (*this) = p1;
-	} else { /* Double RL rotation */
-	    p2 = p1->left;
-	    b2 = p2->balance;
-	    p1->left = p2->right;
-	    p2->right = p1;
-	    p->right = p2->left;
-	    p2->left = p;
-	    p->balance = (b2 == 1) ? -1 : 0;
-	    p1->balance = (b2 == -1) ? 1 : 0;
-	    p2->balance = 0;
-	    (*this) = p2;
-	}
-	break;
-    }
-    return h;
-}
-
-static int balance_right(ErtsMonitorOrLink **this) 
-{
-    ErtsMonitorOrLink *p, *p1, *p2;
-    int b1, b2, h = 1;
-    
-    p = *this;
-    switch (p->balance) {
-    case 1:
-	p->balance = 0;
-	break;
-    case 0:
-	p->balance = -1;
-	h = 0;
-	break;
-    case -1:
-	p1 = p->left;
-	b1 = p1->balance;
-	if (b1 <= 0) { /* Single LL rotation */
-	    p->left = p1->right;
-	    p1->right = p;
-	    if (b1 == 0) {
-		p->balance = -1;
-		p1->balance = 1;
-		h = 0;
-	    } else {
-		p->balance = p1->balance = 0;
-	    }
-	    (*this) = p1;
-	} else { /* Double LR rotation */
-	    p2 = p1->right;
-	    b2 = p2->balance;
-	    p1->right = p2->left;
-	    p2->left = p1;
-	    p->left = p2->right;
-	    p2->right = p;
-	    p->balance = (b2 == -1) ? 1 : 0;
-	    p1->balance = (b2 == 1) ? -1 : 0;
-	    p2->balance = 0;
-	    (*this) = p2;
-	}
-    }
-    return h;
-}
-
-static int delsub(ErtsMonitorOrLink **this) 
-{
-    ErtsMonitorOrLink **tstack[STACK_NEED];
-    int tpos = 0;
-    ErtsMonitorOrLink *q = (*this);
-    ErtsMonitorOrLink **r = &(q->left);
-    int h;
-
-    /*
-     * Walk down the tree to the right and search 
-     * for a void right child, pick that child out
-     * and return it to be put in the deleted 
-     * object's place.
-     */
-    
-    while ((*r)->right != NULL) {
-	tstack[tpos++] = r;
-	r = &((*r)->right);
-    }
-    *this = *r;
-    *r = (*r)->left;
-    (*this)->left = q->left;
-    (*this)->right = q->right;
-    (*this)->balance = q->balance;
-    tstack[0] = &((*this)->left);
-    h = 1;
-    while (tpos && h) {
-	r = tstack[--tpos];
-	h = balance_right(r);
-    }
-    return h;
-}
-
-ErtsMonitor *erts_remove_monitor(ErtsMonitor **root, Eterm ref) 
-{
-    ErtsMonitor **tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsMonitor **this = root;
-    Sint c;
-    int dir;
-    ErtsMonitor *q = NULL;
-
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Failure */
-	    return NULL;
-	} else if ((c = CMP_MON_REF(ref,(*this)->ref)) < 0) { 
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key, found the one to delete */
-	    q = (*this);
-	    if (q->right == NULL) {
-		(*this) = q->left;
-		state = 1;
-	    } else if (q->left == NULL) {
-		(*this) = q->right;
-		state = 1;
-	    } else {
-		dstack[dpos++] = DIR_LEFT;
-		tstack[tpos++] = this;
-		state = delsub((ErtsMonitorOrLink **) this);
-	    }
-	    break;
-	}
-    }
-    while (state && ( dir = dstack[--dpos] ) != DIR_END) {
-	this = tstack[--tpos];
-	if (dir == DIR_LEFT) {
-	    state = balance_left((ErtsMonitorOrLink **) this);
-	} else {
-	    state = balance_right((ErtsMonitorOrLink **) this);
-	}
-    }
-    return q;
-}
-
-ErtsLink *erts_remove_link(ErtsLink **root, Eterm pid) 
-{
-    ErtsLink **tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsLink **this = root;
-    Sint c;
-    int dir;
-    ErtsLink *q = NULL;
-
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Failure */
-	    return NULL;
-	} else if ((c = CMP(pid,(*this)->pid)) < 0) {
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key, found the one to delete */
-	    q = (*this);
-	    if (q->right == NULL) {
-		(*this) = q->left;
-		state = 1;
-	    } else if (q->left == NULL) {
-		(*this) = q->right;
-		state = 1;
-	    } else {
-		dstack[dpos++] = DIR_LEFT;
-		tstack[tpos++] = this;
-		state = delsub((ErtsMonitorOrLink **) this);
-	    }
-	    break;
-	}
-    }
-    while (state && ( dir = dstack[--dpos] ) != DIR_END) {
-	this = tstack[--tpos];
-	if (dir == DIR_LEFT) {
-	    state = balance_left((ErtsMonitorOrLink **) this);
-	} else {
-	    state = balance_right((ErtsMonitorOrLink **) this);
-	}
-    }
-    return q;
-}
-
-void
-erts_delete_suspend_monitor(ErtsSuspendMonitor **root, Eterm pid)
-{
-    ErtsSuspendMonitor **tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int state = 0;
-    ErtsSuspendMonitor **this = root;
-    Sint c;
-    int dir;
-    ErtsSuspendMonitor *q = NULL;
-
-    dstack[0] = DIR_END;
-    for (;;) {
-	if (!*this) { /* Nothing found */
-	    return;
-	} else if ((c = CMP(pid,(*this)->pid)) < 0) {
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->left);
-	} else if (c > 0) { /* go right */
-	    dstack[dpos++] = DIR_RIGHT;
-	    tstack[tpos++] = this;
-	    this = &((*this)->right);
-	} else { /* Equal key, found the one to delete */
-	    q = (*this);
-	    ASSERT(q->pid == pid);
-	    if (q->right == NULL) {
-		(*this) = q->left;
-		state = 1;
-	    } else if (q->left == NULL) {
-		(*this) = q->right;
-		state = 1;
-	    } else {
-		dstack[dpos++] = DIR_LEFT;
-		tstack[tpos++] = this;
-		state = delsub((ErtsMonitorOrLink **) this);
-	    }
-	    erts_destroy_suspend_monitor(q);
-	    break;
-	}
-    }
-    while (state && ( dir = dstack[--dpos] ) != DIR_END) {
-	this = tstack[--tpos];
-	if (dir == DIR_LEFT) {
-	    state = balance_left((ErtsMonitorOrLink **) this);
-	} else {
-	    state = balance_right((ErtsMonitorOrLink **) this);
-	}
-    }
-}
-
-ErtsMonitor *erts_lookup_monitor(ErtsMonitor *root, Eterm ref) 
-{
-    Sint c;
-
-    for (;;) {
-	if (root == NULL || (c = CMP_MON_REF(ref,root->ref)) == 0) {
-	    return root;
-	} else if (c < 0) { 
-	    root = root->left;
-	} else { /* c > 0 */ 
-	    root = root->right;
-	} 
-    }
-}
-
-ErtsLink *erts_lookup_link(ErtsLink *root, Eterm pid) 
-{
-    Sint c;
-
-    for (;;) {
-	if (root == NULL || (c = CMP(pid,root->pid)) == 0) {
-	    return root;
-	} else if (c < 0) { 
-	    root = root->left;
-	} else { /* c > 0 */ 
-	    root = root->right;
-	} 
-    }
-}
-
-ErtsSuspendMonitor *
-erts_lookup_suspend_monitor(ErtsSuspendMonitor *root, Eterm pid) 
-{
-    Sint c;
-
-    for (;;) {
-	if (root == NULL || (c = CMP(pid,root->pid)) == 0) {
-	    return root;
-	} else if (c < 0) { 
-	    root = root->left;
-	} else { /* c > 0 */ 
-	    root = root->right;
-	} 
-    }
-}
-
-void erts_sweep_monitors(ErtsMonitor *root, 
-			 void (*doit)(ErtsMonitor *, void *),
-			 void *context) 
-{
-    ErtsMonitor *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int dir;
-    
-    dstack[0] = DIR_END;
-
-    for (;;) {
-	if (root == NULL) {
-	    if ((dir = dstack[dpos-1]) == DIR_END) {
-		return;
-	    }
-	    if (dir == DIR_LEFT) {
-		/* Still has DIR_RIGHT to do */
-		dstack[dpos-1] = DIR_RIGHT;
-		root = (tstack[tpos-1])->right;
-	    } else {
-		/* stacktop is an object to be deleted */
-		(*doit)(tstack[--tpos],context); /* expeted to do the 
-						    deletion */
-		--dpos;
-		root = NULL;
-	    }
-	} else {
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = root;
-	    root = root->left;
-	}
-    }
-}
-
-void erts_sweep_links(ErtsLink *root, 
-		      void (*doit)(ErtsLink *, void *),
-		      void *context) 
-{
-    ErtsLink *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int dir;
-    
-    dstack[0] = DIR_END;
-
-    for (;;) {
-	if (root == NULL) {
-	    if ((dir = dstack[dpos-1]) == DIR_END) {
-		return;
-	    }
-	    if (dir == DIR_LEFT) {
-		/* Still has DIR_RIGHT to do */
-		dstack[dpos-1] = DIR_RIGHT;
-		root = (tstack[tpos-1])->right;
-	    } else {
-		/* stacktop is an object to be deleted */
-		(*doit)(tstack[--tpos],context); /* expeted to do the 
-						    deletion */
-		--dpos;
-		root = NULL;
-	    }
-	} else {
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = root;
-	    root = root->left;
-	}
-    }
-}
-
-void erts_sweep_suspend_monitors(ErtsSuspendMonitor *root,
-				 void (*doit)(ErtsSuspendMonitor *, void *),
-				 void *context)
-{
-    ErtsSuspendMonitor *tstack[STACK_NEED];
-    int tpos = 0;
-    int dstack[STACK_NEED+1];
-    int dpos = 1;
-    int dir;
-    
-    dstack[0] = DIR_END;
-
-    for (;;) {
-	if (root == NULL) {
-	    if ((dir = dstack[dpos-1]) == DIR_END) {
-		return;
-	    }
-	    if (dir == DIR_LEFT) {
-		/* Still has DIR_RIGHT to do */
-		dstack[dpos-1] = DIR_RIGHT;
-		root = (tstack[tpos-1])->right;
-	    } else {
-		/* stacktop is an object to be deleted */
-		(*doit)(tstack[--tpos],context); /* expeted to do the 
-						    deletion */
-		--dpos;
-		root = NULL;
-	    }
-	} else {
-	    dstack[dpos++] = DIR_LEFT;
-	    tstack[tpos++] = root;
-	    root = root->left;
-	}
-    }
-}
-				
-
-/* Debug BIF, always present, but undocumented... */
-	    
-static void erts_dump_monitors(ErtsMonitor *root, int indent)
-{
-    if (root == NULL)
-	return;
-    erts_dump_monitors(root->right,indent+2);
-    erts_printf("%*s[%b16d:%b16u:%T:%T", indent, "", root->balance,
-		root->type, root->ref, root->name);
-    if (root->type == MON_NIF_TARGET)
-	erts_printf(":%p]\n", root->u.resource);
-    else
-	erts_printf(":%T]\n", root->u.pid);
-    erts_dump_monitors(root->left,indent+2);
-}
-
-static void erts_dump_links_aux(ErtsLink *root, int indent,
-				erts_dsprintf_buf_t *dsbufp)
-{
-    if (root == NULL)
-	return;
-    erts_dump_links_aux(root->right, indent+2, dsbufp);
-    dsbufp->str_len = 0;
-    erts_dsprintf(dsbufp, "%*s[%b16d:%b16u:%T:%p]", indent, "",
-		  root->balance, root->type, root->pid, ERTS_LINK_ROOT(root));
-    if (ERTS_LINK_ROOT(root) != NULL) {
-	ErtsLink *sub = ERTS_LINK_ROOT(root);
-	int len = dsbufp->str_len;
-	erts_dump_links_aux(sub->right, indent+len+5, dsbufp);
-	erts_dsprintf(dsbufp, "-> %*s[%b16d:%b16u:%T:%p]", indent, "",
-		      sub->balance, sub->type, sub->pid, ERTS_LINK_ROOT(sub));
-	erts_printf("%s\n", dsbufp->str);
-	erts_dump_links_aux(sub->left, indent+len+5, dsbufp);
-    } else {
-	erts_printf("%s\n", dsbufp->str);
-    }
-    erts_dump_links_aux(root->left, indent+2, dsbufp);
-}
-
-static void erts_dump_links(ErtsLink *root, int indent)
-{
-    erts_dsprintf_buf_t *dsbufp = erts_create_tmp_dsbuf(0);
-    erts_dump_links_aux(root, indent, dsbufp);
-    erts_destroy_tmp_dsbuf(dsbufp);
-}
-
-Eterm erts_debug_dump_monitors_1(BIF_ALIST_1)
-{
-    Process *p = BIF_P;
-    Eterm pid = BIF_ARG_1;
-    Process *rp;
-    DistEntry *dep;
-    rp = erts_pid2proc(p, ERTS_PROC_LOCK_MAIN, pid, ERTS_PROC_LOCK_LINK);
-    if (!rp) {
-	ERTS_SMP_ASSERT_IS_NOT_EXITING(p);
-	if (is_atom(pid) && is_node_name_atom(pid) &&
-	    (dep = erts_find_dist_entry(pid)) != NULL) {
-	    erts_printf("Dumping dist monitors-------------------\n");
-	    erts_smp_de_links_lock(dep);
-	    erts_dump_monitors(dep->monitors,0);
-	    erts_smp_de_links_unlock(dep);
-	    erts_printf("Monitors dumped-------------------------\n");
-	    erts_deref_dist_entry(dep);
-	    BIF_RET(am_true);
-	} else {
-	    BIF_ERROR(p,BADARG);
-	}
-    } else {
-	erts_printf("Dumping pid monitors--------------------\n");
-	erts_dump_monitors(ERTS_P_MONITORS(rp),0);
-	erts_printf("Monitors dumped-------------------------\n");
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	BIF_RET(am_true);
-    }
-}
-
-Eterm erts_debug_dump_links_1(BIF_ALIST_1)
-{
-    Process *p = BIF_P;
-    Eterm pid = BIF_ARG_1;
-    Process *rp;
-    DistEntry *dep;
-    if (is_internal_port(pid)) {
-	Port *rport = erts_id2port_sflgs(pid,
-					 p,
-					 ERTS_PROC_LOCK_MAIN,
-					 ERTS_PORT_SFLGS_INVALID_LOOKUP);
-	if (rport) {
-	    erts_printf("Dumping port links----------------------\n");
-	    erts_dump_links(ERTS_P_LINKS(rport), 0);
-	    erts_printf("Links dumped----------------------------\n");
-	    erts_port_release(rport);
-	    BIF_RET(am_true);
-	} else {
-	    BIF_ERROR(p,BADARG);
-	}
-    } else {
-	rp = erts_pid2proc(p, ERTS_PROC_LOCK_MAIN, pid, ERTS_PROC_LOCK_LINK);
-	if (!rp) {
-	    ERTS_SMP_ASSERT_IS_NOT_EXITING(p);
-	    if (is_atom(pid) && is_node_name_atom(pid) &&
-		(dep = erts_find_dist_entry(pid)) != NULL) {
-		erts_printf("Dumping dist links----------------------\n");
-		erts_smp_de_links_lock(dep);
-		erts_dump_links(dep->nlinks,0);
-		erts_smp_de_links_unlock(dep);
-		erts_printf("Links dumped----------------------------\n");
-		erts_deref_dist_entry(dep);
-		BIF_RET(am_true);
-	    } else {
-		BIF_ERROR(p,BADARG);
-	    }
-
-	} else {
-	    erts_printf("Dumping pid links-----------------------\n");
-	    erts_dump_links(ERTS_P_LINKS(rp), 0);
-	    erts_printf("Links dumped----------------------------\n");
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	    BIF_RET(am_true);
-	}
-    }
-}
-
-void erts_one_link_size(ErtsLink *lnk, void *vpu)
-{
-    Uint *pu = vpu;
-    *pu += ERTS_LINK_SIZE*sizeof(Uint);
-    if(is_not_immed(lnk->pid))
-	*pu += NC_HEAP_SIZE(lnk->pid)*sizeof(Uint);
-    if (lnk->type != LINK_NODE && ERTS_LINK_ROOT(lnk) != NULL) {
-	erts_doforall_links(ERTS_LINK_ROOT(lnk),&erts_one_link_size,vpu);
-    }
-}
-void erts_one_mon_size(ErtsMonitor *mon, void *vpu)
-{
-    Uint *pu = vpu;
-    *pu += ERTS_MONITOR_SIZE*sizeof(Uint);
-    if(mon->type != MON_NIF_TARGET && is_not_immed(mon->u.pid))
-	*pu += NC_HEAP_SIZE(mon->u.pid)*sizeof(Uint);
-    if(is_not_immed(mon->ref))
-	*pu += NC_HEAP_SIZE(mon->ref)*sizeof(Uint);
-}
diff --git a/erts/emulator/beam/erl_monitors.h b/erts/emulator/beam/erl_monitors.h
deleted file mode 100644
index 1cacecd7e9..0000000000
--- a/erts/emulator/beam/erl_monitors.h
+++ /dev/null
@@ -1,187 +0,0 @@
-/*
- * %CopyrightBegin%
- * 
- * Copyright Ericsson AB 2004-2017. All Rights Reserved.
- * 
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- * 
- * %CopyrightEnd%
- */
-
-/**********************************************************************
- * Header for monitors and links data structures.
- * Monitors are kept in an AVL tree and the data structures for
- * the four different types of monitors are like this:
- **********************************************************************
- * Local monitor by pid/port: 
- * (Ref is always same in all involved data structures)
- **********************************************************************
- *        Process/Port X     Process Y
- *       +-------------+    +-------------+
- * Type: | MON_ORIGIN  |    | MON_TARGET  |
- *       +-------------+    +-------------+
- * Pid:  | Pid(Y)      |    | Pid/Port(X) |
- *       +-------------+    +-------------+
- * Name: | []          |    | []          |
- *       +-------------+    +-------------+
- **********************************************************************
- * Local monitor by name: (Ref is always same in all involved data structures)
- **********************************************************************
- *        Process X          Process Y (name foo)
- *       +-------------+    +-------------+
- * Type: | MON_ORIGIN  |    | MON_TARGET  |
- *       +-------------+    +-------------+
- * Pid:  | Pid(Y)      |    | Pid(X)      |
- *       +-------------+    +-------------+
- * Name: | Atom(foo)   |    | Atom(foo)   |
- *       +-------------+    +-------------+
- **********************************************************************
- * Remote monitor by pid: (Ref is always same in all involved data structures)
- **********************************************************************
- *                    Node A              |           Node B
- *       ---------------------------------+----------------------------------
- *        Process X (@A)   Distentry @A       Distentry @B     Process Y (@B)
- *                         for node B         for node A 
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Type: | MON_ORIGIN  |  | MON_TARGET  |    | MON_ORIGIN  |  | MON_TARGET  |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Pid:  | Pid(Y)      |  | Pid(X)      |    | Pid(Y)      |  | Pid(X)      |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Name: | []          |  | []          |    | []          |  | []          |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- **********************************************************************
- * Remote monitor by name: (Ref is always same in all involved data structures)
- **********************************************************************
- *                    Node A              |           Node B
- *       ---------------------------------+----------------------------------
- *        Process X (@A)   Distentry @A       Distentry @B     Process Y (@B)
- *                         for node B         for node A       (name foo)
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Type: | MON_ORIGIN  |  | MON_TARGET  |    | MON_ORIGIN  |  | MON_TARGET  |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Pid:  | Atom(node B)|  | Pid(X)      |    | Pid(Y)      |  | Pid(X)      |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * Name: | Atom(foo)   |  | Atom(foo)   |    | Atom(foo)   |  | Atom(foo)   |
- *       +-------------+  +-------------+    +-------------+  +-------------+
- * The reason for the node atom in X->pid is that we don't know the actual
- * pid of the monitored process on the other node when setting the monitor
- * (which is done asyncronously).
- **********************************************************************/
-#ifndef _ERL_MONITORS_H
-#define _ERL_MONITORS_H
-
-/* Type tags for monitors */
-#define MON_ORIGIN 1
-#define MON_TARGET 2
-#define MON_NIF_TARGET 3
-#define MON_TIME_OFFSET 4
-
-/* Type tags for links */
-#define LINK_PID 1   /* ...Or port */
-#define LINK_NODE 3  /* "Node monitor" */  
-
-/* Size of a monitor without heap, in words (fixalloc) */
-#define ERTS_MONITOR_SIZE ((sizeof(ErtsMonitor) - sizeof(Uint))/sizeof(Uint))
-#define ERTS_MONITOR_SH_SIZE (ERTS_MONITOR_SIZE + ERTS_REF_THING_SIZE)
-#define ERTS_LINK_SIZE ((sizeof(ErtsLink) - sizeof(Uint))/sizeof(Uint))
-#define ERTS_LINK_SH_SIZE ERTS_LINK_SIZE /* Size of fix-alloced links */
-
-/* ErtsMonitor and ErtsLink *need* to begin in a similar way as 
-   ErtsMonitorOrLink */
-typedef struct erts_monitor_or_link {
-    struct erts_monitor_or_link *left, *right;
-    Sint16 balance;    
-} ErtsMonitorOrLink;
-
-typedef struct erts_monitor {
-    struct erts_monitor *left, *right; 
-    Sint16 balance;
-    Uint16 type;  /* MON_ORIGIN | MON_TARGET | MON_NIF_TARGET | MON_TIME_OFFSET */
-    Eterm ref;
-    union {
-        Eterm pid;  /* In case of distributed named monitor, this is the
-		     * nodename atom in MON_ORIGIN process, otherwise a pid or,
-		     * in case of a MON_TARGET, a port
-                     */
-        struct ErtsResource_* resource; /* MON_NIF_TARGET */
-    }u;
-    Eterm name;   /* When monitoring a named process: atom() else [] */
-    Uint heap[1]; /* Larger in reality */
-} ErtsMonitor;
-
-typedef struct erts_link {
-    struct erts_link *left, *right;
-    Sint16 balance;
-    Uint16 type;             /* LINK_PID | LINK_NODE */
-    Eterm pid;               /* When node monitor, 
-				the node atom is here instead */
-    union {
-	struct erts_link *root;  /* Used only in dist entries */
-	Uint refc;
-    } shared;
-    Uint heap[1];            /* Larger in reality */
-} ErtsLink;   
-
-typedef struct erts_suspend_monitor {
-    struct erts_suspend_monitor *left, *right;
-    Sint16 balance;
-
-    int pending;
-    int active;
-    Eterm pid;
-} ErtsSuspendMonitor;
-
-#define ERTS_LINK_ROOT(Linkp) ((Linkp)->shared.root)
-#define ERTS_LINK_REFC(Linkp) ((Linkp)->shared.refc) 
-
-Uint erts_tot_link_lh_size(void);
-
-
-/* Prototypes */
-void erts_destroy_monitor(ErtsMonitor *mon);
-void erts_add_monitor(ErtsMonitor **root, Uint type, Eterm ref, UWord entity,
-		      Eterm name);
-ErtsMonitor *erts_remove_monitor(ErtsMonitor **root, Eterm ref);
-ErtsMonitor *erts_lookup_monitor(ErtsMonitor *root, Eterm ref);
-void erts_sweep_monitors(ErtsMonitor *root, 
-			 void (*doit)(ErtsMonitor *, void *),
-			 void *context);
-
-void erts_destroy_link(ErtsLink *lnk);
-/* Returns 0 if OK, < 0 if already present */
-int erts_add_link(ErtsLink **root, Uint type, Eterm pid);
-ErtsLink *erts_add_or_lookup_link(ErtsLink **root, Uint type, Eterm pid);
-ErtsLink *erts_remove_link(ErtsLink **root, Eterm pid);
-ErtsLink *erts_lookup_link(ErtsLink *root, Eterm pid);
-void erts_sweep_links(ErtsLink *root, 
-		      void (*doit)(ErtsLink *, void *),
-		      void *context);
-
-void erts_destroy_suspend_monitor(ErtsSuspendMonitor *sproc);
-void erts_sweep_suspend_monitors(ErtsSuspendMonitor *root,
-				 void (*doit)(ErtsSuspendMonitor *, void *),
-				 void *context);
-ErtsSuspendMonitor *erts_add_or_lookup_suspend_monitor(ErtsSuspendMonitor **root,
-						       Eterm pid);
-ErtsSuspendMonitor *erts_lookup_suspend_monitor(ErtsSuspendMonitor *root,
-						Eterm pid);
-void erts_delete_suspend_monitor(ErtsSuspendMonitor **root, Eterm pid);
-void erts_init_monitors(void);
-void erts_one_link_size(ErtsLink *lnk, void *vpu);
-void erts_one_mon_size(ErtsMonitor *mon, void *vpu);
-
-#define erts_doforall_monitors erts_sweep_monitors
-#define erts_doforall_links erts_sweep_links
-#define erts_doforall_suspend_monitors erts_sweep_suspend_monitors
-
-#endif /* _ERL_MONITORS_H */
diff --git a/erts/emulator/beam/erl_msacc.c b/erts/emulator/beam/erl_msacc.c
index 6c477be615..d13d6080e1 100644
--- a/erts/emulator/beam/erl_msacc.c
+++ b/erts/emulator/beam/erl_msacc.c
@@ -48,11 +48,7 @@ static Eterm erts_msacc_gather_stats(ErtsMsAcc *msacc, ErtsHeapFactory *factory)
 static void erts_msacc_reset(ErtsMsAcc *msacc);
 static ErtsMsAcc* get_msacc(void);
 
-#ifdef USE_THREADS
 erts_tsd_key_t ERTS_WRITE_UNLIKELY(erts_msacc_key);
-#else
-ErtsMsAcc *ERTS_WRITE_UNLIKELY(erts_msacc) = NULL;
-#endif
 #ifndef ERTS_MSACC_ALWAYS_ON
 int ERTS_WRITE_UNLIKELY(erts_msacc_enabled);
 #endif
@@ -60,10 +56,8 @@ int ERTS_WRITE_UNLIKELY(erts_msacc_enabled);
 static Eterm *erts_msacc_state_atoms = NULL;
 static erts_rwmtx_t msacc_mutex;
 static ErtsMsAcc *msacc_managed = NULL;
-#ifdef USE_THREADS
 static ErtsMsAcc *msacc_unmanaged = NULL;
 static Uint msacc_unmanaged_count = 0;
-#endif
 
 #if ERTS_MSACC_STATE_COUNT < MAP_SMALL_MAP_LIMIT
 #define DEFAULT_MSACC_MSG_SIZE (3 + 1 + ERTS_MSACC_STATE_COUNT * 2 + 3 + ERTS_REF_THING_SIZE)
@@ -78,11 +72,7 @@ void erts_msacc_early_init(void) {
 #endif
     erts_rwmtx_init(&msacc_mutex, "msacc_list_mutex", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
-#ifdef USE_THREADS
     erts_tsd_key_create(&erts_msacc_key,"erts_msacc_key");
-#else
-    erts_msacc = NULL;
-#endif
 }
 
 void erts_msacc_init(void) {
@@ -91,7 +81,7 @@ void erts_msacc_init(void) {
                                         sizeof(Eterm)*ERTS_MSACC_STATE_COUNT);
     for (i = 0; i < ERTS_MSACC_STATE_COUNT; i++) {
         erts_msacc_state_atoms[i] = am_atom_put(erts_msacc_states[i],
-                                                strlen(erts_msacc_states[i]));
+                                                sys_strlen(erts_msacc_states[i]));
     }
 }
 
@@ -107,7 +97,6 @@ void erts_msacc_init_thread(char *type, int id, int managed) {
     msacc->tid = erts_thr_self();
     msacc->perf_counter = 0;
 
-#ifdef USE_THREADS
     erts_rwmtx_rwlock(&msacc_mutex);
     if (!managed) {
         erts_mtx_init(&msacc->mtx, "msacc_unmanaged_mutex", NIL,
@@ -121,9 +110,6 @@ void erts_msacc_init_thread(char *type, int id, int managed) {
         msacc_managed = msacc;
     }
     erts_rwmtx_rwunlock(&msacc_mutex);
-#else
-    msacc_managed = msacc;
-#endif
 
     erts_msacc_reset(msacc);
 
@@ -205,7 +191,7 @@ Eterm erts_msacc_gather_stats(ErtsMsAcc *msacc, ErtsHeapFactory *factory) {
     map->keys = key;
     hp[0] = state_map;
     hp[1] = msacc->id;
-    hp[2] = am_atom_put(msacc->type,strlen(msacc->type));
+    hp[2] = am_atom_put(msacc->type,sys_strlen(msacc->type));
 
     return make_flatmap(map);
 }
@@ -216,7 +202,7 @@ typedef struct {
     Eterm ref;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Uint req_sched;
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
 } ErtsMSAccReq;
 
 static ErtsMsAcc* get_msacc(void) {
@@ -267,7 +253,7 @@ static void send_reply(ErtsMsAcc *msacc, ErtsMSAccReq *msaccrp) {
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
 
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
 }
 
@@ -303,7 +289,7 @@ reply_msacc(void *vmsaccrp)
 
     erts_proc_dec_refc(msaccrp->proc);
 
-    if (erts_smp_atomic32_dec_read_nob(&msaccrp->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&msaccrp->refc) == 0)
       erts_free(ERTS_ALC_T_MSACC, vmsaccrp);
 }
 
@@ -370,14 +356,10 @@ erts_msacc_request(Process *c_p, int action, Eterm *threads)
     msaccrp->ref = STORE_NC(&hp, NULL, ref);
     msaccrp->req_sched = esdp->no;
 
-#ifdef ERTS_SMP
     *threads = erts_no_schedulers;
     *threads += 1; /* aux thread */
-#else
-    *threads = 1;
-#endif
 
-    erts_smp_atomic32_init_nob(&msaccrp->refc,(erts_aint32_t)*threads);
+    erts_atomic32_init_nob(&msaccrp->refc,(erts_aint32_t)*threads);
 
     erts_proc_add_refc(c_p, *threads);
 
@@ -386,12 +368,9 @@ erts_msacc_request(Process *c_p, int action, Eterm *threads)
                                           erts_no_schedulers,
                                           reply_msacc,
                                           (void *) msaccrp);
-#ifdef ERTS_SMP
     /* aux thread */
     erts_schedule_misc_aux_work(0, reply_msacc, (void *) msaccrp);
-#endif
 
-#ifdef USE_THREADS
     /* Manage unmanaged threads */
     switch (action) {
     case ERTS_MSACC_GATHER: {
@@ -468,7 +447,6 @@ erts_msacc_request(Process *c_p, int action, Eterm *threads)
     default: { ASSERT(0); }
     }
 
-#endif
 
     *threads = make_small(*threads);
 
diff --git a/erts/emulator/beam/erl_msacc.h b/erts/emulator/beam/erl_msacc.h
index 3f6273f214..895b1ae319 100644
--- a/erts/emulator/beam/erl_msacc.h
+++ b/erts/emulator/beam/erl_msacc.h
@@ -159,25 +159,16 @@ struct erl_msacc_t_ {
 
 #ifdef ERTS_ENABLE_MSACC
 
-#ifdef USE_THREADS
-extern erts_tsd_key_t erts_msacc_key;
-#else
-extern ErtsMsAcc *erts_msacc;
-#endif
+extern erts_tsd_key_t ERTS_WRITE_UNLIKELY(erts_msacc_key);
 
 #ifdef ERTS_MSACC_ALWAYS_ON
 #define erts_msacc_enabled 1
 #else
-extern int erts_msacc_enabled;
+extern int ERTS_WRITE_UNLIKELY(erts_msacc_enabled);
 #endif
 
-#ifdef USE_THREADS
 #define ERTS_MSACC_TSD_GET() erts_tsd_get(erts_msacc_key)
 #define ERTS_MSACC_TSD_SET(tsd) erts_tsd_set(erts_msacc_key,tsd)
-#else
-#define ERTS_MSACC_TSD_GET() erts_msacc
-#define ERTS_MSACC_TSD_SET(tsd) erts_msacc = tsd
-#endif
 
 void erts_msacc_early_init(void);
 void erts_msacc_init(void);
@@ -341,8 +332,8 @@ ERTS_GLB_INLINE
 void erts_msacc_set_state_um__(ErtsMsAcc *msacc, Uint new_state, int increment) {
     if (ERTS_UNLIKELY(msacc->unmanaged)) {
         erts_mtx_lock(&msacc->mtx);
-        msacc->state = new_state;
         if (ERTS_LIKELY(!msacc->perf_counter)) {
+            msacc->state = new_state;
             erts_mtx_unlock(&msacc->mtx);
             return;
         }
diff --git a/erts/emulator/beam/erl_mtrace.c b/erts/emulator/beam/erl_mtrace.c
index f2a660f085..2807b443a1 100644
--- a/erts/emulator/beam/erl_mtrace.c
+++ b/erts/emulator/beam/erl_mtrace.c
@@ -314,7 +314,7 @@ disable_trace(int error, char *reason, int eno)
 	erts_fprintf(stderr, "%s: %s\n", mt_dis, reason);
     else {
 	eno_str = erl_errno_id(eno);
-	if (strcmp(eno_str, "unknown") == 0)
+	if (sys_strcmp(eno_str, "unknown") == 0)
 	    erts_fprintf(stderr, "%s: %s: %d\n", mt_dis, reason, eno);
 	else
 	    erts_fprintf(stderr, "%s: %s: %s\n", mt_dis, reason, eno_str);
@@ -401,9 +401,9 @@ write_trace_header(char *nodename, char *pid, char *hostname)
     PUT_UI32(tracep, ERTS_MT_MAJOR_VSN);
     PUT_UI32(tracep, ERTS_MT_MINOR_VSN);
 
-    n_len = strlen(nodename);
-    h_len = strlen(hostname);
-    p_len = strlen(pid);
+    n_len = sys_strlen(nodename);
+    h_len = sys_strlen(hostname);
+    p_len = sys_strlen(pid);
     hdr_prolog_len = (2*UI32_SZ
 		      + 3*UI16_SZ
 		      + 3*UI32_SZ
@@ -436,15 +436,15 @@ write_trace_header(char *nodename, char *pid, char *hostname)
     PUT_UI32(tracep, start_time.usec);
 
     PUT_UI8(tracep, (byte) n_len);
-    memcpy((void *) tracep, (void *) nodename, n_len);
+    sys_memcpy((void *) tracep, (void *) nodename, n_len);
     tracep += n_len;
 
     PUT_UI8(tracep, (byte) h_len);
-    memcpy((void *) tracep, (void *) hostname, h_len);
+    sys_memcpy((void *) tracep, (void *) hostname, h_len);
     tracep += h_len;
 
     PUT_UI8(tracep, (byte) p_len);
-    memcpy((void *) tracep, (void *) pid, p_len);
+    sys_memcpy((void *) tracep, (void *) pid, p_len);
     tracep += p_len;
 
     ASSERT(startp + hdr_prolog_len == tracep);
@@ -472,7 +472,7 @@ write_trace_header(char *nodename, char *pid, char *hostname)
 
 	str = ERTS_ALC_A2AD(i);
 	ASSERT(str);
-	str_len = strlen(str);
+	str_len = sys_strlen(str);
 	if (str_len >= (1 << 8)) {
 	    disable_trace(1, "Excessively large allocator string", 0);
 	    return 0;
@@ -493,7 +493,7 @@ write_trace_header(char *nodename, char *pid, char *hostname)
 	PUT_UI16(tracep, aflags);
 	PUT_UI16(tracep, (Uint16) i);
 	PUT_UI8( tracep, (byte) str_len);
-	memcpy((void *) tracep, (void *) str, str_len);
+	sys_memcpy((void *) tracep, (void *) str, str_len);
 	tracep += str_len;
 	if (erts_allctrs_info[i].alloc_util) {
 	    PUT_UI8(tracep, 2);
@@ -519,7 +519,7 @@ write_trace_header(char *nodename, char *pid, char *hostname)
 	str = ERTS_ALC_N2TD(i);
 	ASSERT(str);
 
-	str_len = strlen(str);
+	str_len = sys_strlen(str);
 	if (str_len >= (1 << 8)) {
 	    disable_trace(1, "Excessively large type string", 0);
 	    return 0;
@@ -543,7 +543,7 @@ write_trace_header(char *nodename, char *pid, char *hostname)
 	PUT_UI16(tracep, nflags);
 	PUT_UI16(tracep, (Uint16) i);
 	PUT_UI8(tracep, (byte) str_len);
-	memcpy((void *) tracep, (void *) str, str_len);
+	sys_memcpy((void *) tracep, (void *) str, str_len);
 	tracep += str_len;
 	PUT_UI16(tracep, no);
 	ASSERT(startp + entry_sz == tracep);
diff --git a/erts/emulator/beam/erl_nfunc_sched.c b/erts/emulator/beam/erl_nfunc_sched.c
index 1bebc1eda4..f97e86bf95 100644
--- a/erts/emulator/beam/erl_nfunc_sched.c
+++ b/erts/emulator/beam/erl_nfunc_sched.c
@@ -113,7 +113,7 @@ erts_nif_export_schedule(Process *c_p, Process *dirty_shadow_proc,
     NifExport* nep;
     int i;
 
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
 		       & ERTS_PROC_LOCK_MAIN);
 
     if (dirty_shadow_proc) {
diff --git a/erts/emulator/beam/erl_nfunc_sched.h b/erts/emulator/beam/erl_nfunc_sched.h
index 55a3a6dbf6..b8a4e4ebc3 100644
--- a/erts/emulator/beam/erl_nfunc_sched.h
+++ b/erts/emulator/beam/erl_nfunc_sched.h
@@ -144,9 +144,9 @@ ERTS_GLB_INLINE void
 erts_nif_export_restore(Process *c_p, NifExport *ep, Eterm result)
 {
     ASSERT(!ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data()));
-    ERTS_SMP_LC_ASSERT(!(c_p->static_flags
+    ERTS_LC_ASSERT(!(c_p->static_flags
 			 & ERTS_STC_FLG_SHADOW_PROC));
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
 		       & ERTS_PROC_LOCK_MAIN);
 
     c_p->current = ep->current;
@@ -193,7 +193,6 @@ erts_nif_export_check_save_trace(Process *c_p, Eterm result,
 ERTS_GLB_INLINE Process *
 erts_proc_shadow2real(Process *c_p)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (c_p->static_flags & ERTS_STC_FLG_SHADOW_PROC) {
 	Process *real_c_p = c_p->next;
 	ASSERT(ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data()));
@@ -201,7 +200,6 @@ erts_proc_shadow2real(Process *c_p)
 	return real_c_p;
     }
     ASSERT(!ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data()));
-#endif
     return c_p;
 }
 
@@ -213,11 +211,10 @@ erts_proc_shadow2real(Process *c_p)
 #define ERTS_NFUNC_SCHED_INTERNALS__
 
 #define ERTS_I_BEAM_OP_TO_NIF_EXPORT(I)					\
-    (ASSERT(BeamOp(op_apply_bif) == (BeamInstr *) (*(I))		\
-	    || BeamOp(op_call_nif) == (BeamInstr *) (*(I))),		\
+    (ASSERT(BeamIsOpCode(*(I), op_apply_bif) ||                         \
+            BeamIsOpCode(*(I), op_call_nif)),                           \
      ((NifExport *) (((char *) (I)) - offsetof(NifExport, exp.beam[0]))))
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 #include "erl_message.h"
 #include <stddef.h>
@@ -235,7 +232,7 @@ erts_flush_dirty_shadow_proc(Process *sproc)
     Process *c_p = sproc->next;
 
     ASSERT(sproc->common.id == c_p->common.id);
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
 		       & ERTS_PROC_LOCK_MAIN);
 
     ASSERT(c_p->stop == sproc->stop);
@@ -283,7 +280,7 @@ erts_cache_dirty_shadow_proc(Process *sproc)
     Process *c_p = sproc->next;
     ASSERT(c_p);
     ASSERT(sproc->common.id == c_p->common.id);
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
 		       & ERTS_PROC_LOCK_MAIN);
 
     sproc->htop = c_p->htop;
@@ -311,7 +308,7 @@ erts_make_dirty_shadow_proc(ErtsSchedulerData *esdp, Process *c_p)
     sproc = esdp->dirty_shadow_process;
     ASSERT(sproc);
     ASSERT(sproc->static_flags & ERTS_STC_FLG_SHADOW_PROC);
-    ASSERT(erts_smp_atomic32_read_nob(&sproc->state)
+    ASSERT(erts_atomic32_read_nob(&sproc->state)
 	   == (ERTS_PSFLG_ACTIVE
 	       | ERTS_PSFLG_DIRTY_RUNNING
 	       | ERTS_PSFLG_PROXY));
@@ -326,7 +323,6 @@ erts_make_dirty_shadow_proc(ErtsSchedulerData *esdp, Process *c_p)
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 #endif /* defined(ERTS_WANT_NFUNC_SCHED_INTERNALS__) && !defined(ERTS_NFUNC_SCHED_INTERNALS__) */
 
diff --git a/erts/emulator/beam/erl_nif.c b/erts/emulator/beam/erl_nif.c
index b96db5d054..0fbf0eb03a 100644
--- a/erts/emulator/beam/erl_nif.c
+++ b/erts/emulator/beam/erl_nif.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2009-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2009-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -57,6 +57,7 @@
 #include "erl_bif_unique.h"
 #include "erl_utils.h"
 #include "erl_io_queue.h"
+#include "erl_proc_sig_queue.h"
 #undef ERTS_WANT_NFUNC_SCHED_INTERNALS__
 #define ERTS_WANT_NFUNC_SCHED_INTERNALS__
 #include "erl_nfunc_sched.h"
@@ -138,7 +139,7 @@ execution_state(ErlNifEnv *env, Process **c_pp, int *schedp)
 	    Process *c_p = env->proc;
 
 	    if (!(c_p->static_flags & ERTS_STC_FLG_SHADOW_PROC)) {
-		ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
+		ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p)
 				   & ERTS_PROC_LOCK_MAIN);
 	    }
 	    else {
@@ -220,7 +221,7 @@ void erts_pre_nif(ErlNifEnv* env, Process* p, struct erl_module_nif* mod_nif,
 	ASSERT(esdp);
 
 	if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-	    erts_aint32_t state = erts_smp_atomic32_read_nob(&p->state);
+	    erts_aint32_t state = erts_atomic32_read_nob(&p->state);
 
 	    ASSERT(p->scheduler_data == esdp);
 	    ASSERT((state & (ERTS_PSFLG_RUNNING
@@ -237,9 +238,11 @@ static void cache_env(ErlNifEnv* env);
 static void full_flush_env(ErlNifEnv *env);
 static void flush_env(ErlNifEnv* env);
 
-/* Temporary object header, auto-deallocated when NIF returns
- * or when independent environment is cleared.
- */
+/* Temporary object header, auto-deallocated when NIF returns or when
+ * independent environment is cleared.
+ *
+ * The payload can be accessed with &tmp_obj_ptr[1] but keep in mind that its
+ * first element must not require greater alignment than `next`. */
 struct enif_tmp_obj_t {
     struct enif_tmp_obj_t* next;
     void (*dtor)(struct enif_tmp_obj_t*);
@@ -256,6 +259,46 @@ static ERTS_INLINE void free_tmp_objs(ErlNifEnv* env)
     }
 }
 
+/* Whether the given environment is bound to a process and will be cleaned up
+ * when the NIF returns. It's safe to use temp_alloc for objects in
+ * env->tmp_obj_list when this is true. */
+static ERTS_INLINE int is_proc_bound(ErlNifEnv *env)
+{
+    return env->mod_nif != NULL;
+}
+
+/* Allocates and attaches an object to the given environment, running its
+ * destructor when the environment is cleared. To avoid temporary variables the
+ * address of the allocated object is returned instead of the enif_tmp_obj_t.
+ *
+ * The destructor *must* call `erts_free(tmp_obj->allocator, tmp_obj)` to free
+ * the object. If the destructor needs to refer to the allocated object its
+ * address will be &tmp_obj[1]. */
+static ERTS_INLINE void *alloc_tmp_obj(ErlNifEnv *env, size_t size,
+                                       void (*dtor)(struct enif_tmp_obj_t*)) {
+    struct enif_tmp_obj_t *tmp_obj;
+    ErtsAlcType_t allocator;
+
+    allocator = is_proc_bound(env) ? ERTS_ALC_T_TMP : ERTS_ALC_T_NIF;
+
+    tmp_obj = erts_alloc(allocator, sizeof(struct enif_tmp_obj_t) + MAX(1, size));
+
+    tmp_obj->next = env->tmp_obj_list;
+    tmp_obj->allocator = allocator;
+    tmp_obj->dtor = dtor;
+
+    env->tmp_obj_list = tmp_obj;
+
+    return (void*)&tmp_obj[1];
+}
+
+/* Generic destructor for objects allocated through alloc_tmp_obj that don't
+ * care about their payload. */
+static void tmp_alloc_dtor(struct enif_tmp_obj_t *tmp_obj)
+{
+    erts_free(tmp_obj->allocator, tmp_obj);
+}
+
 void erts_post_nif(ErlNifEnv* env)
 {
     erts_unblock_fpe(env->fpe_was_unmasked);
@@ -287,12 +330,12 @@ schedule(ErlNifEnv* env, NativeFunPtr direct_fp, NativeFunPtr indirect_fp,
     else
 	dirty_shadow_proc = env->proc;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & erts_proc_lc_my_proc_locks(c_p));
 
     ep = erts_nif_export_schedule(c_p, dirty_shadow_proc,
 				  c_p->current,
 				  c_p->cp,
-				  (BeamInstr) em_call_nif,
+				  BeamOpCodeAddr(op_call_nif),
 				  direct_fp, indirect_fp,
 				  mod, func_name,
 				  argc, (const Eterm *) argv);
@@ -304,7 +347,6 @@ schedule(ErlNifEnv* env, NativeFunPtr direct_fp, NativeFunPtr indirect_fp,
     return (ERL_NIF_TERM) THE_NON_VALUE;
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static ERL_NIF_TERM dirty_nif_finalizer(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]);
 static ERL_NIF_TERM dirty_nif_exception(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]);
@@ -320,7 +362,7 @@ erts_call_dirty_nif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
     ErlNifEnv env;
     ERL_NIF_TERM result;
 #ifdef DEBUG
-    erts_aint32_t state = erts_smp_atomic32_read_nob(&c_p->state);
+    erts_aint32_t state = erts_atomic32_read_nob(&c_p->state);
 
     ASSERT(nep == ERTS_PROC_GET_NIF_TRAP_EXPORT(c_p));
 
@@ -343,14 +385,20 @@ erts_call_dirty_nif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
 
     ASSERT(ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(c_p)));
 
-    erts_smp_atomic32_read_band_mb(&c_p->state, ~(ERTS_PSFLG_DIRTY_CPU_PROC
+    erts_atomic32_read_band_mb(&c_p->state, ~(ERTS_PSFLG_DIRTY_CPU_PROC
 						   | ERTS_PSFLG_DIRTY_IO_PROC));
 
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+    ASSERT(esdp->current_nif == NULL);
+    esdp->current_nif = &env;
+
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 
     result = (*dirty_nif)(&env, codemfa->arity, argv); /* Call dirty NIF */
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+
+    ASSERT(esdp->current_nif == &env);
+    esdp->current_nif = NULL;
 
     ASSERT(env.proc->static_flags & ERTS_STC_FLG_SHADOW_PROC);
     ASSERT(env.proc->next == c_p);
@@ -394,24 +442,19 @@ erts_call_dirty_nif(ErtsSchedulerData *esdp, Process *c_p, BeamInstr *I, Eterm *
     return exiting;
 }
 
-#endif
 
 static void full_flush_env(ErlNifEnv* env)
 {
     flush_env(env);
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (env->proc->static_flags & ERTS_STC_FLG_SHADOW_PROC)
 	/* Dirty nif call using shadow process struct */
 	erts_flush_dirty_shadow_proc(env->proc);
-#endif
 }
 
 static void full_cache_env(ErlNifEnv* env)
 {    
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (env->proc->static_flags & ERTS_STC_FLG_SHADOW_PROC)
 	erts_cache_dirty_shadow_proc(env->proc);
-#endif
     cache_env(env);
 }
 
@@ -452,6 +495,7 @@ static void cache_env(ErlNifEnv* env)
 	env->hp_end = env->heap_frag->mem + env->heap_frag->alloc_size;
     }
 }
+
 void* enif_priv_data(ErlNifEnv* env)
 {
     return env->mod_nif->priv_data;
@@ -492,7 +536,7 @@ setup_nif_env(struct enif_msg_environment_t* msg_env,
     msg_env->env.tmp_obj_list = NULL;
     msg_env->env.proc = &msg_env->phony_proc;
     msg_env->env.exception_thrown = 0;
-    memset(&msg_env->phony_proc, 0, sizeof(Process));
+    sys_memset(&msg_env->phony_proc, 0, sizeof(Process));
     HEAP_START(&msg_env->phony_proc) = phony_heap;
     HEAP_TOP(&msg_env->phony_proc) = phony_heap;
     HEAP_LIMIT(&msg_env->phony_proc) = phony_heap;
@@ -566,10 +610,9 @@ void enif_clear_env(ErlNifEnv* env)
     ASSERT(!is_offheap(&MSO(p)));
 }
 
-#ifdef ERTS_SMP
 #ifdef DEBUG
 static int enif_send_delay = 0;
-#define ERTS_FORCE_ENIF_SEND_DELAY() (enif_send_delay++ % 2 == 0)
+#define ERTS_FORCE_ENIF_SEND_DELAY() (enif_send_delay++ % 32 == 0)
 #else
 #ifdef ERTS_PROC_LOCK_OWN_IMPL
 #define ERTS_FORCE_ENIF_SEND_DELAY() 0
@@ -592,9 +635,9 @@ int erts_flush_trace_messages(Process *c_p, ErtsProcLocks c_p_locks)
 
     /* Only one thread at a time is allowed to flush trace messages,
        so we require the main lock to be held when doing the flush */
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_TRACE);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_TRACE);
 
     msgq = c_p->trace_msg_q;
 
@@ -613,7 +656,7 @@ int erts_flush_trace_messages(Process *c_p, ErtsProcLocks c_p_locks)
         msgq->first = NULL;
         msgq->last = &msgq->first;
         msgq->len = 0;
-        erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_TRACE);
+        erts_proc_unlock(c_p, ERTS_PROC_LOCK_TRACE);
 
         ASSERT(len != 0);
 
@@ -622,17 +665,17 @@ int erts_flush_trace_messages(Process *c_p, ErtsProcLocks c_p_locks)
             rp_locks = 0;
             if (rp->common.id == c_p->common.id)
                 rp_locks = c_p_locks;
-            erts_queue_messages(rp, rp_locks, first, last, len, c_p->common.id);
+            erts_queue_proc_messages(c_p, rp, rp_locks, first, last, len);
             if (rp->common.id == c_p->common.id)
                 rp_locks &= ~c_p_locks;
             if (rp_locks)
-                erts_smp_proc_unlock(rp, rp_locks);
+                erts_proc_unlock(rp, rp_locks);
             reds += len;
         } else {
             erts_cleanup_messages(first);
         }
         reds += 1;
-        erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_TRACE);
+        erts_proc_lock(c_p, ERTS_PROC_LOCK_TRACE);
         msgq = msgq->next;
     } while (msgq);
 
@@ -649,36 +692,27 @@ int erts_flush_trace_messages(Process *c_p, ErtsProcLocks c_p_locks)
     }
 
 error:
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_TRACE);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_TRACE);
 
     return reds;
 }
 
-#endif
 
 int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 	      ErlNifEnv* msg_env, ERL_NIF_TERM msg)
 {
     struct enif_msg_environment_t* menv = (struct enif_msg_environment_t*)msg_env;
     ErtsProcLocks rp_locks = 0;
-#ifdef ERTS_SMP
     ErtsProcLocks lc_locks = 0;
-#endif
     Process* rp;
     Process* c_p;
     ErtsMessage *mp;
+    Eterm from;
     Eterm receiver = to_pid->pid;
     int scheduler;
 
     execution_state(env, &c_p, &scheduler);
 
-#ifndef ERTS_SMP
-    if (!scheduler) {
-	erts_exit(ERTS_ABORT_EXIT,
-		  "enif_send: called from non-scheduler thread on non-SMP VM");
-	return 0;
-   }
-#endif
 
     if (scheduler > 0) { /* Normal scheduler */
 	rp = erts_proc_lookup(receiver);
@@ -692,7 +726,7 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 		return 0;
 
 	    if (env->proc->static_flags & ERTS_STC_FLG_SHADOW_PROC) {
-		erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+		erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 	    }
 	}
 
@@ -701,7 +735,7 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 			       ERTS_P2P_FLG_INC_REFC);
 	if (!rp) {
 	    if (c_p && (env->proc->static_flags & ERTS_STC_FLG_SHADOW_PROC))
-		erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+		erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 	    return 0;
 	}
     }
@@ -736,7 +770,7 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 	    full_cache_env(env);
 	}
 	else {
-	    erts_aint_t state = erts_smp_atomic32_read_nob(&rp->state);
+	    erts_aint_t state = erts_atomic32_read_nob(&rp->state);
 	    if (state & ERTS_PSFLG_OFF_HEAP_MSGQ) {
 		mp = erts_alloc_message(sz, &hp);
 		ohp = sz == 0 ? NULL : &mp->hfrag.off_heap;
@@ -752,7 +786,7 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
         msg = copy_struct_litopt(msg, sz, &hp, ohp, &litarea);
     }
 
-    ERL_MESSAGE_TERM(mp) = msg;
+    from = c_p ? c_p->common.id : am_undefined;
 
     if (!env || !env->tracee) {
 
@@ -762,7 +796,6 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 	    full_cache_env(env);
 	}
     }
-#ifdef ERTS_SMP
     else {
         /* This clause is taken when the nif is called in the context
            of a traced process. We do not know which locks we have
@@ -772,8 +805,7 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
         ErlTraceMessageQueue *msgq;
         Process *t_p = env->tracee;
 
-
-        erts_smp_proc_lock(t_p, ERTS_PROC_LOCK_TRACE);
+        erts_proc_lock(t_p, ERTS_PROC_LOCK_TRACE);
 
         msgq = t_p->trace_msg_q;
 
@@ -790,7 +822,10 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
 #endif
         if (ERTS_FORCE_ENIF_SEND_DELAY() || msgq ||
             rp_locks & ERTS_PROC_LOCK_MSGQ ||
-            erts_smp_proc_trylock(rp, ERTS_PROC_LOCK_MSGQ) == EBUSY) {
+            erts_proc_trylock(rp, ERTS_PROC_LOCK_MSGQ) == EBUSY) {
+
+            ERL_MESSAGE_TERM(mp) = msg;
+            ERL_MESSAGE_FROM(mp) = from;
 
             if (!msgq) {
                 msgq = erts_alloc(ERTS_ALC_T_TRACE_MSG_QUEUE,
@@ -804,36 +839,35 @@ int enif_send(ErlNifEnv* env, const ErlNifPid* to_pid,
                 msgq->next = t_p->trace_msg_q;
                 t_p->trace_msg_q = msgq;
 
-                erts_smp_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
+                erts_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
 
 		erts_schedule_flush_trace_messages(t_p, 0);
             } else {
                 msgq->len++;
                 *msgq->last = mp;
                 msgq->last = &mp->next;
-                erts_smp_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
+                erts_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
             }
             goto done;
         } else {
-            erts_smp_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
+            erts_proc_unlock(t_p, ERTS_PROC_LOCK_TRACE);
             rp_locks &= ~ERTS_PROC_LOCK_TRACE;
             rp_locks |= ERTS_PROC_LOCK_MSGQ;
         }
     }
-#endif /* ERTS_SMP */
 
-    erts_queue_message(rp, rp_locks, mp, msg,
-                       c_p ? c_p->common.id : am_undefined);
+    if (c_p)
+        erts_queue_proc_message(c_p, rp, rp_locks, mp, msg);
+    else
+        erts_queue_message(rp, rp_locks, mp, msg, from);
 
-#ifdef ERTS_SMP
 done:
     if (c_p == rp)
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
     if (rp_locks & ~lc_locks)
-	erts_smp_proc_unlock(rp, rp_locks & ~lc_locks);
+	erts_proc_unlock(rp, rp_locks & ~lc_locks);
     if (c_p && (env->proc->static_flags & ERTS_STC_FLG_SHADOW_PROC))
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
-#endif
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
     if (scheduler <= 0)
 	erts_proc_dec_refc(rp);
 
@@ -863,15 +897,9 @@ enif_port_command(ErlNifEnv *env, const ErlNifPort* to_port,
     if (scheduler > 0)
 	prt = erts_port_lookup(to_port->port_id, iflags);
     else {
-#ifdef ERTS_SMP
 	if (ERTS_PROC_IS_EXITING(c_p))
 	    return 0;
 	prt = erts_thr_port_lookup(to_port->port_id, iflags);
-#else
-        erts_exit(ERTS_ABORT_EXIT,
-		  "enif_port_command: called from non-scheduler "
-                  "thread on non-SMP VM");
-#endif
     }
 
     if (!prt)
@@ -1046,11 +1074,6 @@ int enif_is_number(ErlNifEnv* env, ERL_NIF_TERM term)
     return is_number(term);
 }
 
-static ERTS_INLINE int is_proc_bound(ErlNifEnv* env)
-{
-    return env->mod_nif != NULL;
-}
-
 static void aligned_binary_dtor(struct enif_tmp_obj_t* obj)
 {
     erts_free_aligned_binary_bytes_extra((byte*)obj, obj->allocator);
@@ -1085,22 +1108,14 @@ int enif_inspect_binary(ErlNifEnv* env, Eterm bin_term, ErlNifBinary* bin)
 	u.tmp->dtor = &aligned_binary_dtor;
 	env->tmp_obj_list = u.tmp;
     }
-    bin->bin_term = bin_term;
     bin->size = binary_size(bin_term);
     bin->ref_bin = NULL;
     ADD_READONLY_CHECK(env, bin->data, bin->size);
     return 1;
 }
 
-static void tmp_alloc_dtor(struct enif_tmp_obj_t* obj)
-{
-    erts_free(obj->allocator,  obj);
-}
-
 int enif_inspect_iolist_as_binary(ErlNifEnv* env, Eterm term, ErlNifBinary* bin)
 {
-    struct enif_tmp_obj_t* tobj;
-    ErtsAlcType_t allocator;
     ErlDrvSizeT sz;
     if (is_binary(term)) {
 	return enif_inspect_binary(env,term,bin);
@@ -1108,7 +1123,6 @@ int enif_inspect_iolist_as_binary(ErlNifEnv* env, Eterm term, ErlNifBinary* bin)
     if (is_nil(term)) {
 	bin->data = (unsigned char*) &bin->data; /* dummy non-NULL */
 	bin->size = 0;
-	bin->bin_term = THE_NON_VALUE;
 	bin->ref_bin = NULL;
 	return 1;
     }
@@ -1116,16 +1130,8 @@ int enif_inspect_iolist_as_binary(ErlNifEnv* env, Eterm term, ErlNifBinary* bin)
 	return 0;
     }
 
-    allocator = is_proc_bound(env) ? ERTS_ALC_T_TMP : ERTS_ALC_T_NIF;
-    tobj = erts_alloc(allocator, sz + sizeof(struct enif_tmp_obj_t));
-    tobj->allocator = allocator;
-    tobj->next = env->tmp_obj_list;
-    tobj->dtor = &tmp_alloc_dtor;
-    env->tmp_obj_list = tobj;
-
-    bin->data = (unsigned char*) &tobj[1]; 
+    bin->data = alloc_tmp_obj(env, sz, &tmp_alloc_dtor);
     bin->size = sz;
-    bin->bin_term = THE_NON_VALUE;
     bin->ref_bin = NULL;
     erts_iolist_to_buf(term, (char*) bin->data, sz);
     ADD_READONLY_CHECK(env, bin->data, bin->size); 
@@ -1143,7 +1149,6 @@ int enif_alloc_binary(size_t size, ErlNifBinary* bin)
 
     bin->size = size;
     bin->data = (unsigned char*) refbin->orig_bytes;
-    bin->bin_term = THE_NON_VALUE;
     bin->ref_bin = refbin;
     return 1;
 }
@@ -1177,12 +1182,10 @@ void enif_release_binary(ErlNifBinary* bin)
 {
     if (bin->ref_bin != NULL) {
 	Binary* refbin = bin->ref_bin;
-	ASSERT(bin->bin_term == THE_NON_VALUE);
         erts_bin_release(refbin);
     }
 #ifdef DEBUG
     bin->data = NULL;
-    bin->bin_term = THE_NON_VALUE;
     bin->ref_bin = NULL;
 #endif
 }
@@ -1230,11 +1233,12 @@ size_t enif_binary_to_term(ErlNifEnv *dst_env,
     Sint size;
     ErtsHeapFactory factory;
     byte *bp = (byte*) data;
+    Uint32 flags = 0;
 
-    ERTS_CT_ASSERT(ERL_NIF_BIN2TERM_SAFE == ERTS_DIST_EXT_BTT_SAFE);
-
-    if (opts & ~ERL_NIF_BIN2TERM_SAFE) {
-        return 0;
+    switch ((Uint32)opts) {
+    case 0: break;
+    case ERL_NIF_BIN2TERM_SAFE: flags = ERTS_DIST_EXT_BTT_SAFE; break;
+    default: return 0;
     }
     if ((size = erts_decode_ext_size(bp, data_sz)) < 0)
         return 0;
@@ -1246,13 +1250,15 @@ size_t enif_binary_to_term(ErlNifEnv *dst_env,
         erts_factory_dummy_init(&factory);
     }
 
-    *term = erts_decode_ext(&factory, &bp, (Uint32)opts);
+    *term = erts_decode_ext(&factory, &bp, flags);
 
     if (is_non_value(*term)) {
         return 0;
     }
-    erts_factory_close(&factory);
-    cache_env(dst_env);
+    if (size > 0) {
+        erts_factory_close(&factory);
+        cache_env(dst_env);
+    }
 
     ASSERT(bp > data);
     return bp - data;
@@ -1338,39 +1344,51 @@ int enif_get_string(ErlNifEnv *env, ERL_NIF_TERM list, char* buf, unsigned len,
 
 Eterm enif_make_binary(ErlNifEnv* env, ErlNifBinary* bin)
 {
-    if (bin->bin_term != THE_NON_VALUE) {
-	return bin->bin_term;
-    }
-    else if (bin->ref_bin != NULL) {
-	Binary* bptr = bin->ref_bin;
-	ProcBin* pb;
-	Eterm bin_term;
-	
-	/* !! Copy-paste from new_binary() !! */
-	pb = (ProcBin *) alloc_heap(env, PROC_BIN_SIZE);
-	pb->thing_word = HEADER_PROC_BIN;
-	pb->size = bptr->orig_size;
-	pb->next = MSO(env->proc).first;
-	MSO(env->proc).first = (struct erl_off_heap_header*) pb;
-	pb->val = bptr;
-	pb->bytes = (byte*) bptr->orig_bytes;
-	pb->flags = 0;
-	
-	OH_OVERHEAD(&(MSO(env->proc)), pb->size / sizeof(Eterm));
-	bin_term = make_binary(pb);	
-	if (erts_refc_read(&bptr->intern.refc, 1) == 1) {
-	    /* Total ownership transfer */
-	    bin->ref_bin = NULL;
-	    bin->bin_term = bin_term;
-	}
-	return bin_term;
-    }
-    else {
-	flush_env(env);
-	bin->bin_term = new_binary(env->proc, bin->data, bin->size);
-	cache_env(env);
-	return bin->bin_term;
+    Eterm bin_term;
+
+    if (bin->ref_bin != NULL) {
+        Binary* binary = bin->ref_bin;
+
+        /* If the binary is smaller than the heap binary limit we'll return a
+         * heap binary to reduce the number of small refc binaries in the
+         * system. We can't simply release the refc binary right away however;
+         * the documentation states that the binary should be considered
+         * read-only from this point on, which implies that it should still be
+         * readable.
+         *
+         * We could keep it alive until we return by adding it to the temporary
+         * object list, but that requires an off-heap allocation which is
+         * potentially quite slow, so we create a dummy ProcBin instead and
+         * rely on the next minor GC to get rid of it. */
+        if (bin->size <= ERL_ONHEAP_BIN_LIMIT) {
+            ErlHeapBin* hb;
+
+            hb = (ErlHeapBin*)alloc_heap(env, heap_bin_size(bin->size));
+            hb->thing_word = header_heap_bin(bin->size);
+            hb->size = bin->size;
+
+            sys_memcpy(hb->data, bin->data, bin->size);
+
+            erts_build_proc_bin(&MSO(env->proc),
+                                alloc_heap(env, PROC_BIN_SIZE),
+                                binary);
+
+            bin_term = make_binary(hb);
+        } else {
+            bin_term = erts_build_proc_bin(&MSO(env->proc),
+                                           alloc_heap(env, PROC_BIN_SIZE),
+                                           binary);
+        }
+
+        /* Our (possibly shared) ownership has been transferred to the term. */
+        bin->ref_bin = NULL;
+    } else {
+        flush_env(env);
+        bin_term = new_binary(env->proc, bin->data, bin->size);
+        cache_env(env);
     }
+
+    return bin_term;
 }
 
 Eterm enif_make_sub_binary(ErlNifEnv* env, ERL_NIF_TERM bin_term,
@@ -1854,18 +1872,11 @@ int enif_is_process_alive(ErlNifEnv* env, ErlNifPid *proc)
     if (scheduler > 0)
 	return !!erts_proc_lookup(proc->pid);
     else {
-#ifdef ERTS_SMP
 	Process* rp = erts_pid2proc_opt(NULL, 0, proc->pid, 0,
 					ERTS_P2P_FLG_INC_REFC);
 	if (rp)
 	    erts_proc_dec_refc(rp);
 	return !!rp;
-#else
-	erts_exit(ERTS_ABORT_EXIT, "enif_is_process_alive: "
-		  "called from non-scheduler thread "
-                  "in non-smp emulator");
-	return 0;
-#endif
     }
 }
 
@@ -1881,17 +1892,10 @@ int enif_is_port_alive(ErlNifEnv *env, ErlNifPort *port)
     if (scheduler > 0)
 	return !!erts_port_lookup(port->port_id, iflags);
     else {
-#ifdef ERTS_SMP
 	Port *prt = erts_thr_port_lookup(port->port_id, iflags);
 	if (prt)
 	    erts_port_dec_refc(prt);
 	return !!prt;
-#else
-	erts_exit(ERTS_ABORT_EXIT, "enif_is_port_alive: "
-		  "called from non-scheduler thread "
-                  "in non-smp emulator");
-	return 0;
-#endif
     }
 }
 
@@ -1973,6 +1977,12 @@ ErlNifTid enif_thread_self(void) { return erl_drv_thread_self(); }
 int enif_equal_tids(ErlNifTid tid1, ErlNifTid tid2) { return erl_drv_equal_tids(tid1,tid2); }
 void enif_thread_exit(void *resp) { erl_drv_thread_exit(resp); }
 int enif_thread_join(ErlNifTid tid, void **respp) { return erl_drv_thread_join(tid,respp); }
+
+char* enif_mutex_name(ErlNifMutex *mtx) {return erl_drv_mutex_name(mtx); }
+char* enif_cond_name(ErlNifCond *cnd) { return erl_drv_cond_name(cnd); }
+char* enif_rwlock_name(ErlNifRWLock* rwlck) { return erl_drv_rwlock_name(rwlck); }
+char* enif_thread_name(ErlNifTid tid) { return erl_drv_thread_name(tid); }
+
 int enif_getenv(const char *key, char *value, size_t *value_size) { return erl_drv_getenv(key, value, value_size); }
 
 ErlNifTime enif_monotonic_time(ErlNifTimeUnit time_unit)
@@ -1995,16 +2005,21 @@ enif_convert_time_unit(ErlNifTime val,
 						    (int) to);
 }
 
-int enif_fprintf(void* filep, const char* format, ...) 
+int enif_fprintf(FILE* filep, const char* format, ...)
 { 
     int ret;
     va_list arglist;
     va_start(arglist, format);
-    ret = erts_vfprintf((FILE*)filep, format, arglist);
+    ret = erts_vfprintf(filep, format, arglist);
     va_end(arglist);
     return ret;
 }    
 
+int enif_vfprintf(FILE* filep, const char *format, va_list ap)
+{
+    return erts_vfprintf(filep, format, ap);
+}
+
 int enif_snprintf(char *buffer, size_t size, const char* format, ...) 
 { 
     int ret;
@@ -2015,6 +2030,12 @@ int enif_snprintf(char *buffer, size_t size, const char* format, ...)
     return ret;
 }
 
+int enif_vsnprintf(char* buffer, size_t size, const char *format, va_list ap)
+{
+    return erts_vsnprintf(buffer, size, format, ap);
+}
+
+
 /***********************************************************
  **       Memory managed (GC'ed) "resource" objects       **
  ***********************************************************/
@@ -2099,7 +2120,7 @@ ErlNifResourceType* open_resource_type(ErlNifEnv* env,
     ErlNifResourceFlags op = flags;
     Eterm module_am, name_am;
 
-    ASSERT(erts_smp_thr_progress_is_blocking());
+    ASSERT(erts_thr_progress_is_blocking());
     module_am = make_atom(env->mod_nif->mod->module);
     name_am = enif_make_atom(env, name_str);
 
@@ -2216,106 +2237,62 @@ static void rollback_opened_resource_types(void)
     }
 }
 
-struct destroy_monitor_ctx
-{
-    ErtsResource* resource;
-    int exiting_procs;
-    int scheduler;
-};
-
-static void destroy_one_monitor(ErtsMonitor* mon, void* context)
-{
-    struct destroy_monitor_ctx* ctx = (struct destroy_monitor_ctx*) context;
-    Process* rp;
-    ErtsMonitor *rmon = NULL;
-    int is_exiting;
-
-    ASSERT(mon->type == MON_ORIGIN);
-    ASSERT(is_internal_pid(mon->u.pid));
-    ASSERT(is_internal_ref(mon->ref));
-
-    if (ctx->scheduler > 0) { /* Normal scheduler */
-        rp = erts_proc_lookup(mon->u.pid);
-    }
-    else {
-#ifdef ERTS_SMP
-        rp = erts_proc_lookup_inc_refc(mon->u.pid);
+#ifdef ARCH_64
+#  define ERTS_RESOURCE_DYING_FLAG (((Uint) 1) << 63)
 #else
-        ASSERT(!"nif monitor destruction in non-scheduler thread");
-        rp = NULL;
+#  define ERTS_RESOURCE_DYING_FLAG (((Uint) 1) << 31)
 #endif
-    }
+#define ERTS_RESOURCE_REFC_MASK (~ERTS_RESOURCE_DYING_FLAG)
 
-    if (!rp) {
-        is_exiting = 1;
-    }
-    if (rp) {
-        erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-        if (ERTS_PROC_IS_EXITING(rp)) {
-            is_exiting = 1;
-        } else {
-            rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-            ASSERT(rmon);
-            is_exiting = 0;
-        }
-        erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-#ifdef ERTS_SMP
-        if (ctx->scheduler <= 0)
-            erts_proc_dec_refc(rp);
-#endif
-    }
-    if (is_exiting) {
-        ctx->resource->monitors->pending_failed_fire++;
-    }
-
-    /* ToDo: Delay destruction after monitor_locks */
-    if (rmon) {
-        ASSERT(rmon->type == MON_NIF_TARGET);
-        ASSERT(rmon->u.resource == ctx->resource);
-        erts_destroy_monitor(rmon);
-    }
-    erts_destroy_monitor(mon);
+static ERTS_INLINE void
+rmon_set_dying(ErtsResourceMonitors *rms)
+{
+    rms->refc |= ERTS_RESOURCE_DYING_FLAG;
 }
 
-static void destroy_all_monitors(ErtsMonitor* monitors, ErtsResource* resource)
+static ERTS_INLINE int
+rmon_is_dying(ErtsResourceMonitors *rms)
 {
-    struct destroy_monitor_ctx ctx;
-
-    execution_state(NULL, NULL, &ctx.scheduler);
-
-    ctx.resource = resource;
-    erts_sweep_monitors(monitors, &destroy_one_monitor, &ctx);
+    return !!(rms->refc & ERTS_RESOURCE_DYING_FLAG);
 }
 
+static ERTS_INLINE void
+rmon_refc_inc(ErtsResourceMonitors *rms)
+{
+    rms->refc++;
+}
 
-#ifdef ERTS_SMP
-#  define NIF_RESOURCE_DTOR &nif_resource_dtor
-#else
-#  define NIF_RESOURCE_DTOR &nosmp_nif_resource_dtor_prologue
+static ERTS_INLINE Uint
+rmon_refc_dec_read(ErtsResourceMonitors *rms)
+{
+    Uint res;
+    ASSERT((rms->refc & ERTS_RESOURCE_REFC_MASK) != 0);
+    res = --rms->refc;
+    return res & ERTS_RESOURCE_REFC_MASK;
+}
 
-/*
- * NO-SMP: Always run resource destructor on scheduler thread 
- *         as we may have to remove process monitors.
- */
-static int nif_resource_dtor(Binary*);
+static ERTS_INLINE void
+rmon_refc_dec(ErtsResourceMonitors *rms)
+{
+    ASSERT((rms->refc & ERTS_RESOURCE_REFC_MASK) != 0);
+    --rms->refc;
+}
 
-static void nosmp_nif_resource_dtor_scheduled(void* vbin)
+static ERTS_INLINE Uint
+rmon_refc_read(ErtsResourceMonitors *rms)
 {
-    erts_bin_free((Binary*)vbin);
+    return rms->refc & ERTS_RESOURCE_REFC_MASK;
 }
 
-static int nosmp_nif_resource_dtor_prologue(Binary* bin)
+static void dtor_demonitor(ErtsMonitor* mon, void* context)
 {
-    if (is_scheduler()) {
-        return nif_resource_dtor(bin);
-    }
-    else {
-        erts_schedule_misc_aux_work(1, nosmp_nif_resource_dtor_scheduled, bin);
-        return 0; /* do not free */
-    }
+    ASSERT(erts_monitor_is_origin(mon));
+    ASSERT(is_internal_pid(mon->other.item));
+
+    erts_proc_sig_send_demonitor(mon);
 }
 
-#endif /* !ERTS_SMP */
+#  define NIF_RESOURCE_DTOR &nif_resource_dtor
 
 static int nif_resource_dtor(Binary* bin)
 {
@@ -2325,35 +2302,37 @@ static int nif_resource_dtor(Binary* bin)
 
     if (resource->monitors) {
         ErtsResourceMonitors* rm = resource->monitors;
+        int kill;
+        ErtsMonitor *root;
+        Uint refc;
 
         ASSERT(type->down);
-        erts_smp_mtx_lock(&rm->lock);
+        erts_mtx_lock(&rm->lock);
         ASSERT(erts_refc_read(&bin->intern.refc, 0) == 0);
-        if (rm->root) {
-            ASSERT(!rm->is_dying);
-            destroy_all_monitors(rm->root, resource);
+        kill = !rmon_is_dying(rm);
+        if (kill) {
+            rmon_set_dying(rm);
+            root = rm->root;
             rm->root = NULL;
         }
-        if (rm->pending_failed_fire) {
-            /*
-             * Resource death struggle prolonged to serve exiting process(es).
-             * Destructor will be called again when last exiting process
-             * tries to fire its MON_NIF_TARGET monitor (and fails).
-             *
-             * This resource is doomed. It has no "real" references and
-             * should get not get called upon to do anything except the
-             * final destructor call.
-             *
-             * We keep refc at 0 and use a separate counter for exiting
-             * processes to avoid resource getting revived by "dec_term".
-             */
-            ASSERT(!rm->is_dying);
-            rm->is_dying = 1;
-            erts_smp_mtx_unlock(&rm->lock);
-            return 0;
-        }
-        erts_smp_mtx_unlock(&rm->lock);
-        erts_smp_mtx_destroy(&rm->lock);
+        refc = rmon_refc_read(rm);
+        erts_mtx_unlock(&rm->lock);
+
+        if (kill)
+            erts_monitor_tree_foreach_delete(&root,
+                                             dtor_demonitor,
+                                             NULL);
+
+        /*
+         * If resource->monitors->refc != 0 there are
+         * outstanding references to the resource from
+         * monitors that has not been removed yet.
+         * nif_resource_dtor() will be called again this
+         * reference count reach zero.
+         */
+        if (refc != 0)
+            return 0; /* we'll be back... */
+        erts_mtx_destroy(&rm->lock);
     }
 
     if (type->dtor != NULL) {
@@ -2382,54 +2361,82 @@ void erts_resource_stop(ErtsResource* resource, ErlNifEvent e,
     post_nif_noproc(&msg_env);
 }
 
-void erts_fire_nif_monitor(ErtsResource* resource, Eterm pid, Eterm ref)
+void erts_nif_demonitored(ErtsResource* resource)
 {
-    ErtsMonitor* rmon;
+    ErtsResourceMonitors* rmp = resource->monitors;
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
+    int free_me;
+
+    ASSERT(rmp);
+    ASSERT(resource->type->down);
+
+    erts_mtx_lock(&rmp->lock);
+    free_me = ((rmon_refc_dec_read(rmp) == 0) & !!rmon_is_dying(rmp));
+    erts_mtx_unlock(&rmp->lock);
+
+    if (free_me)
+        erts_bin_free(&bin->binary);
+}
+
+void erts_fire_nif_monitor(ErtsMonitor *tmon)
+{
+    ErtsResource* resource;
+    ErtsMonitorData *mdp;
+    ErtsMonitor *omon;
+    ErtsBinary* bin;
     struct enif_msg_environment_t msg_env;
     ErlNifPid nif_pid;
     ErlNifMonitor nif_monitor;
-    ErtsResourceMonitors* rmp = resource->monitors;
+    ErtsResourceMonitors* rmp;
+    Uint mrefc, brefc;
+    int active, is_dying;
+
+    ASSERT(tmon->type == ERTS_MON_TYPE_RESOURCE);
+    ASSERT(erts_monitor_is_target(tmon));
+
+    resource = tmon->other.ptr;
+    bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
+    rmp = resource->monitors;
+
+    mdp = erts_monitor_to_data(tmon);
+    omon = &mdp->origin;
 
     ASSERT(rmp);
     ASSERT(resource->type->down);
 
-    erts_smp_mtx_lock(&rmp->lock);
-    rmon = erts_remove_monitor(&rmp->root, ref);
-    if (!rmon) {
-        int free_me = (--rmp->pending_failed_fire == 0) && rmp->is_dying;
-        ASSERT(rmp->pending_failed_fire >= 0);
-        erts_smp_mtx_unlock(&rmp->lock);
+    erts_mtx_lock(&rmp->lock);
 
-        if (free_me) {
-            ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) == 0);
-            erts_bin_free(&bin->binary);
-        }
-        return;
+    mrefc = rmon_refc_dec_read(rmp);
+    is_dying = rmon_is_dying(rmp);
+    active = !is_dying && erts_monitor_is_in_table(omon);
+
+    if (active) {
+        erts_monitor_tree_delete(&rmp->root, omon);
+        brefc = (Uint) erts_refc_inc_unless(&bin->binary.intern.refc, 0, 0);
     }
-    ASSERT(!rmp->is_dying);
-    if (erts_refc_inc_unless(&bin->binary.intern.refc, 0, 0) == 0) {
-        /*
-         * Racing resource destruction. 
-         * To avoid a more complex refc-dance with destructing thread
-         * we avoid calling 'down' and just silently remove the monitor.
-         * This can happen even for non smp as destructor calls may be scheduled.
-         */
-        erts_smp_mtx_unlock(&rmp->lock);
+
+    erts_mtx_unlock(&rmp->lock);
+
+    if (!active) {
+        ASSERT(!is_dying || erts_refc_read(&bin->binary.intern.refc, 0) == 0);
+        if (is_dying && mrefc == 0)
+            erts_bin_free(&bin->binary);
+        erts_monitor_release(tmon);
     }
     else {
-        erts_smp_mtx_unlock(&rmp->lock);
-
-        ASSERT(rmon->u.pid == pid);
-        erts_ref_to_driver_monitor(ref, &nif_monitor);
-        nif_pid.pid = pid;
-        pre_nif_noproc(&msg_env, resource->type->owner, NULL);
-        resource->type->down(&msg_env.env, resource->data, &nif_pid, &nif_monitor);
-        post_nif_noproc(&msg_env);
+        if (brefc > 0) {
+            ASSERT(is_internal_pid(omon->other.item));
+            erts_ref_to_driver_monitor(mdp->ref, &nif_monitor);
+            nif_pid.pid = omon->other.item;
+            pre_nif_noproc(&msg_env, resource->type->owner, NULL);
+            resource->type->down(&msg_env.env, resource->data, &nif_pid, &nif_monitor);
+            post_nif_noproc(&msg_env);
+
+            erts_bin_release(&bin->binary);
+        }
 
-        erts_bin_release(&bin->binary);
+        erts_monitor_release_both(mdp);
     }
-    erts_destroy_monitor(rmon);
 }
 
 void* enif_alloc_resource(ErlNifResourceType* type, size_t data_sz)
@@ -2463,11 +2470,10 @@ void* enif_alloc_resource(ErlNifResourceType* type, size_t data_sz)
     erts_refc_inc(&resource->type->refc, 2);
     if (type->down) {
         resource->monitors = (ErtsResourceMonitors*) (resource->data + monitors_offs);
-        erts_smp_mtx_init(&resource->monitors->lock, "resource_monitors", NIL,
+        erts_mtx_init(&resource->monitors->lock, "resource_monitors", NIL,
             ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
         resource->monitors->root = NULL;
-        resource->monitors->pending_failed_fire = 0;
-        resource->monitors->is_dying = 0;
+        resource->monitors->refc = 0;
         resource->monitors->user_data_sz = data_sz;
     }
     else {
@@ -2482,7 +2488,7 @@ void enif_release_resource(void* obj)
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
 
     ASSERT(ERTS_MAGIC_BIN_DESTRUCTOR(bin) == NIF_RESOURCE_DTOR);
-    ASSERT(!(resource->monitors && resource->monitors->is_dying));
+    ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) != 0);
 #ifdef DEBUG
     erts_refc_dec(&resource->nif_refc, 0);
 #endif
@@ -2495,7 +2501,7 @@ void enif_keep_resource(void* obj)
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
 
     ASSERT(ERTS_MAGIC_BIN_DESTRUCTOR(bin) == NIF_RESOURCE_DTOR);
-    ASSERT(!(resource->monitors && resource->monitors->is_dying));
+    ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) != 0);
 #ifdef DEBUG
     erts_refc_inc(&resource->nif_refc, 1);
 #endif
@@ -2505,7 +2511,7 @@ void enif_keep_resource(void* obj)
 Eterm erts_bld_resource_ref(Eterm** hpp, ErlOffHeap* oh, ErtsResource* resource)
 {
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
-    ASSERT(!(resource->monitors && resource->monitors->is_dying));
+    ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) != 0);
     return erts_mk_magic_ref(hpp, oh, &bin->binary);
 }
 
@@ -2514,7 +2520,7 @@ ERL_NIF_TERM enif_make_resource(ErlNifEnv* env, void* obj)
     ErtsResource* resource = DATA_TO_RESOURCE(obj);
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(resource);
     Eterm* hp = alloc_heap(env, ERTS_MAGIC_REF_THING_SIZE);
-    ASSERT(!(resource->monitors && resource->monitors->is_dying));
+    ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) != 0);
     return erts_mk_magic_ref(&hp, &MSO(env->proc), &bin->binary);
 }
 
@@ -2662,7 +2668,6 @@ nif_export_restore(Process *c_p, NifExport *ep, Eterm res)
 }
 
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 /*
  * Finalize a dirty NIF call. This function is scheduled to cause the VM to
@@ -2732,7 +2737,7 @@ schedule_dirty_nif(ErlNifEnv* env, int flags, NativeFunPtr fp,
 
     execution_state(env, &proc, NULL);
 
-    (void) erts_smp_atomic32_read_bset_nob(&proc->state,
+    (void) erts_atomic32_read_bset_nob(&proc->state,
 					   (ERTS_PSFLG_DIRTY_CPU_PROC
 					    | ERTS_PSFLG_DIRTY_IO_PROC),
 					   (flags == ERL_NIF_DIRTY_JOB_CPU_BOUND
@@ -2770,7 +2775,7 @@ static_schedule_dirty_nif(ErlNifEnv* env, erts_aint32_t dirty_psflg,
     ASSERT(is_atom(mod) && is_atom(func));
     ASSERT(fp);
 
-    (void) erts_smp_atomic32_read_bset_nob(&proc->state,
+    (void) erts_atomic32_read_bset_nob(&proc->state,
 					   (ERTS_PSFLG_DIRTY_CPU_PROC
 					    | ERTS_PSFLG_DIRTY_IO_PROC),
 					   dirty_psflg);
@@ -2790,7 +2795,6 @@ static_schedule_dirty_cpu_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[
     return static_schedule_dirty_nif(env, ERTS_PSFLG_DIRTY_CPU_PROC, argc, argv);
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 /*
  * NIF execution wrapper used by enif_schedule_nif() for regular NIFs. It
@@ -2864,24 +2868,20 @@ enif_schedule_nif(ErlNifEnv* env, const char* fun_name, int flags,
     if (scheduler <= 0) {
 	if (scheduler == 0)
 	    enif_make_badarg(env);
-	erts_smp_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
     }
 
     if (flags == 0)
 	result = schedule(env, execute_nif, fp, proc->current->module,
 			  fun_name_atom, argc, argv);
     else if (!(flags & ~(ERL_NIF_DIRTY_JOB_IO_BOUND|ERL_NIF_DIRTY_JOB_CPU_BOUND))) {
-#ifdef ERTS_DIRTY_SCHEDULERS
 	result = schedule_dirty_nif(env, flags, fp, fun_name_atom, argc, argv);
-#else
-        result = enif_raise_exception(env, am_notsup);
-#endif
     }
     else
 	result = enif_make_badarg(env);
 
     if (scheduler < 0)
-	erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
 
     return result;
 }
@@ -2897,12 +2897,10 @@ enif_thread_type(void)
     switch (esdp->type) {
     case ERTS_SCHED_NORMAL:
 	return ERL_NIF_THR_NORMAL_SCHEDULER;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
 	return ERL_NIF_THR_DIRTY_CPU_SCHEDULER;
     case ERTS_SCHED_DIRTY_IO:
         return ERL_NIF_THR_DIRTY_IO_SCHEDULER;
-#endif
     default:
         ERTS_INTERNAL_ERROR("Invalid scheduler type");
 	return -1;
@@ -2947,6 +2945,44 @@ ERL_NIF_TERM enif_make_new_map(ErlNifEnv* env)
     return make_flatmap(mp);
 }
 
+int enif_make_map_from_arrays(ErlNifEnv *env,
+                              ERL_NIF_TERM keys[],
+                              ERL_NIF_TERM values[],
+                              size_t cnt,
+                              ERL_NIF_TERM *map_out)
+{
+    ErtsHeapFactory factory;
+    int succeeded;
+
+#ifdef ERTS_NIF_ASSERT_IN_ENV
+    size_t index = 0;
+
+    while (index < cnt) {
+        ASSERT_IN_ENV(env, keys[index], index, "key");
+        ASSERT_IN_ENV(env, values[index], index, "value");
+        index++;
+    }
+#endif
+
+    flush_env(env);
+
+    erts_factory_proc_prealloc_init(&factory, env->proc,
+        cnt * 2 + MAP_HEADER_FLATMAP_SZ + 1);
+
+    (*map_out) = erts_map_from_ks_and_vs(&factory, keys, values, cnt);
+    succeeded = (*map_out) != THE_NON_VALUE;
+
+    if (!succeeded) {
+        erts_factory_undo(&factory);
+    }
+
+    erts_factory_close(&factory);
+
+    cache_env(env);
+
+    return succeeded;
+}
+
 int enif_make_map_put(ErlNifEnv* env,
 	              Eterm map_in,
 		      Eterm key,
@@ -3202,143 +3238,88 @@ int enif_map_iterator_get_pair(ErlNifEnv *env,
 int enif_monitor_process(ErlNifEnv* env, void* obj, const ErlNifPid* target_pid,
                          ErlNifMonitor* monitor)
 {
-    int scheduler;
     ErtsResource* rsrc = DATA_TO_RESOURCE(obj);
-    Process *rp;
     Eterm tmp[ERTS_REF_THING_SIZE];
     Eterm ref;
-    int retval;
+    ErtsResourceMonitors *rm;
+    ErtsMonitorData *mdp;
 
     ASSERT(ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(rsrc)->magic_binary.destructor
            == NIF_RESOURCE_DTOR);
-    ASSERT(!(rsrc->monitors && rsrc->monitors->is_dying));
+    ASSERT(erts_refc_read(&ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(rsrc)->binary.intern.refc, 0) != 0);
     ASSERT(!rsrc->monitors == !rsrc->type->down);
 
-
-    if (!rsrc->monitors) {
+    rm = rsrc->monitors;
+    if (!rm) {
         ASSERT(!rsrc->type->down);
         return -1;
     }
     ASSERT(rsrc->type->down);
 
-    execution_state(env, NULL, &scheduler);
+    ref = erts_make_ref_in_buffer(tmp);
 
-#ifdef ERTS_SMP
-    if (scheduler > 0) /* Normal scheduler */
-        rp = erts_proc_lookup_raw(target_pid->pid);
-    else
-        rp = erts_proc_lookup_raw_inc_refc(target_pid->pid);
-#else
-    if (scheduler <= 0) {
-        erts_exit(ERTS_ABORT_EXIT, "enif_monitor_process: called from "
-                  "non-scheduler thread on non-SMP VM");
-    }
-    rp = erts_proc_lookup(target_pid->pid);
+    mdp = erts_monitor_create(ERTS_MON_TYPE_RESOURCE, ref,
+                              (Eterm) rsrc, target_pid->pid, NIL);
+    erts_mtx_lock(&rm->lock);
+    ASSERT(!rmon_is_dying(rm));
+    erts_monitor_tree_insert(&rm->root, &mdp->origin);
+    rmon_refc_inc(rm);
+    erts_mtx_unlock(&rm->lock);
+
+    if (!erts_proc_sig_send_monitor(&mdp->target, target_pid->pid)) {
+        /* Failed to send monitor signal; cleanup... */
+#ifdef DEBUG
+        ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(rsrc);
 #endif
 
-    if (!rp)
-        return 1;
+        erts_mtx_lock(&rm->lock);
+        ASSERT(!rmon_is_dying(rm));
+        erts_monitor_tree_delete(&rm->root, &mdp->origin);
+        rmon_refc_dec(rm);
+        ASSERT(erts_refc_read(&bin->binary.intern.refc, 1) != 0);
+        erts_mtx_unlock(&rm->lock);
+        erts_monitor_release_both(mdp);
 
-    ref = erts_make_ref_in_buffer(tmp);
-
-    erts_smp_mtx_lock(&rsrc->monitors->lock);
-    erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-    if (ERTS_PSFLG_FREE & erts_smp_atomic32_read_nob(&rp->state)) {
-        retval = 1;
-    }
-    else {
-        erts_add_monitor(&rsrc->monitors->root, MON_ORIGIN, ref, rp->common.id, NIL);
-        erts_add_monitor(&ERTS_P_MONITORS(rp), MON_NIF_TARGET, ref, (UWord)rsrc, NIL);
-        retval = 0;
+        return 1;
     }
-    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-    erts_smp_mtx_unlock(&rsrc->monitors->lock);
 
-#ifdef ERTS_SMP
-    if (scheduler <= 0)
-        erts_proc_dec_refc(rp);
-#endif
     if (monitor)
         erts_ref_to_driver_monitor(ref,monitor);
 
-    return retval;
+    return 0;
 }
 
 int enif_demonitor_process(ErlNifEnv* env, void* obj, const ErlNifMonitor* monitor)
 {
-    int scheduler;
     ErtsResource* rsrc = DATA_TO_RESOURCE(obj);
 #ifdef DEBUG
     ErtsBinary* bin = ERTS_MAGIC_BIN_FROM_UNALIGNED_DATA(rsrc);
 #endif
-    Process *rp;
+    ErtsResourceMonitors *rm;
     ErtsMonitor *mon;
-    ErtsMonitor *rmon = NULL;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Eterm ref;
-    int is_exiting;
 
     ASSERT(bin->magic_binary.destructor == NIF_RESOURCE_DTOR);
-    ASSERT(!(rsrc->monitors && rsrc->monitors->is_dying));
-
-    execution_state(env, NULL, &scheduler);
+    ASSERT(erts_refc_read(&bin->binary.intern.refc, 0) != 0);
 
     ref = erts_driver_monitor_to_ref(ref_heap, monitor);
 
-    erts_smp_mtx_lock(&rsrc->monitors->lock);
-    mon = erts_remove_monitor(&rsrc->monitors->root, ref);
+    rm = rsrc->monitors;
+    erts_mtx_lock(&rm->lock);
+    ASSERT(!rmon_is_dying(rm));
+    mon = erts_monitor_tree_lookup(rm->root, ref);
+    if (mon)
+        erts_monitor_tree_delete(&rm->root, mon);
+    erts_mtx_unlock(&rm->lock);
 
-    if (mon == NULL) {
-        erts_smp_mtx_unlock(&rsrc->monitors->lock);
+    if (!mon)
         return 1;
-    }
-
-    ASSERT(mon->type == MON_ORIGIN);
-    ASSERT(is_internal_pid(mon->u.pid));
 
-#ifdef ERTS_SMP
-    if (scheduler > 0) /* Normal scheduler */
-        rp = erts_proc_lookup(mon->u.pid);
-    else
-        rp = erts_proc_lookup_inc_refc(mon->u.pid);
-#else
-    if (scheduler <= 0) {
-        erts_exit(ERTS_ABORT_EXIT, "enif_demonitor_process: called from "
-                  "non-scheduler thread on non-SMP VM");
-    }
-    rp = erts_proc_lookup(mon->u.pid);
-#endif
+    ASSERT(erts_monitor_is_origin(mon));
+    ASSERT(is_internal_pid(mon->other.item));
 
-    if (!rp) {
-        is_exiting = 1;
-    }
-    else {
-        erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-        if (ERTS_PROC_IS_EXITING(rp)) {
-            is_exiting = 1;
-        } else {
-            rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), ref);
-            ASSERT(rmon);
-            is_exiting = 0;
-        }
-        erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-
-#ifdef ERTS_SMP
-        if (scheduler <= 0)
-            erts_proc_dec_refc(rp);
-#endif
-    }
-    if (is_exiting) {
-        rsrc->monitors->pending_failed_fire++;
-    }
-    erts_smp_mtx_unlock(&rsrc->monitors->lock);
-
-    if (rmon) {
-        ASSERT(rmon->type == MON_NIF_TARGET);
-        ASSERT(rmon->u.resource == rsrc);
-        erts_destroy_monitor(rmon);
-    }
-    erts_destroy_monitor(mon);
+    erts_proc_sig_send_demonitor(mon);
 
     return 0;
 }
@@ -3492,7 +3473,6 @@ static void marshal_iovec_binary(Eterm binary, ErlNifBinary *copy_buffer,
     parent_header = binary_val(parent_binary);
 
     result->size = binary_size(binary);
-    result->bin_term = binary;
 
     if (thing_subtag(*parent_header) == REFC_BINARY_SUBTAG) {
         ProcBin *pb = (ProcBin*)parent_header;
@@ -3520,6 +3500,7 @@ static void marshal_iovec_binary(Eterm binary, ErlNifBinary *copy_buffer,
      * and reference that instead. */
 
     if (result->ref_bin == NULL || bit_offset != 0) {
+        ASSERT(copy_buffer->ref_bin != NULL && copy_buffer->data != NULL);
         ASSERT(result->size <= (copy_buffer->size - *copy_offset));
 
         if (bit_offset == 0) {
@@ -3541,8 +3522,8 @@ static void marshal_iovec_binary(Eterm binary, ErlNifBinary *copy_buffer,
 static int fill_iovec_with_slice(ErlNifEnv *env,
                                  iovec_slice_t *slice,
                                  ErlNifIOVec *iovec) {
+    ErlNifBinary copy_buffer = {0};
     UWord copy_offset, iovec_idx;
-    ErlNifBinary copy_buffer;
     Eterm sublist_iterator;
 
     /* Set up a common refc binary for all on-heap and unaligned binaries. */
@@ -3550,11 +3531,8 @@ static int fill_iovec_with_slice(ErlNifEnv *env,
         if (!enif_alloc_binary(slice->copied_size, &copy_buffer)) {
             return 0;
         }
-    } else {
-#ifdef DEBUG
-        copy_buffer.data = NULL;
-        copy_buffer.size = 0;
-#endif
+
+        ASSERT(copy_buffer.ref_bin != NULL);
     }
 
     sublist_iterator = slice->sublist_start;
@@ -3604,9 +3582,11 @@ static int fill_iovec_with_slice(ErlNifEnv *env,
         }
     } else {
         if (slice->copied_size > 0) {
-            /* Attach the binary to our environment and let the GC take care of
-             * it after returning. */
-            enif_make_binary(env, &copy_buffer);
+            /* Attach the binary to our environment and let the next minor GC
+             * get rid of it. This is slightly faster than using the tmp object
+             * list since it avoids off-heap allocations. */
+            erts_build_proc_bin(&MSO(env->proc),
+                alloc_heap(env, PROC_BIN_SIZE), copy_buffer.ref_bin);
         }
     }
 
@@ -3632,19 +3612,14 @@ static int create_iovec_from_slice(ErlNifEnv *env,
         alloc_size = binv_offset;
         alloc_size += slice->iovec_len * sizeof(Binary*);
 
-        /* If we have an environment we'll attach the allocated data to it. The
-         * GC will take care of releasing it later on. */
+        /* When the user passes an environment, we attach the iovec to it so
+         * the user won't have to bother managing it (similar to
+         * enif_inspect_binary). It'll disappear once the environment is
+         * cleaned up. */
         if (env != NULL) {
-            ErlNifBinary gc_bin;
-
-            if (!enif_alloc_binary(alloc_size, &gc_bin)) {
-                return 0;
-            }
-
-            alloc_base = (char*)gc_bin.data;
-            enif_make_binary(env, &gc_bin);
+            alloc_base = alloc_tmp_obj(env, alloc_size, &tmp_alloc_dtor);
         } else {
-            alloc_base = enif_alloc(alloc_size);
+            alloc_base = erts_alloc(ERTS_ALC_T_NIF, alloc_size);
         }
 
         iovec = (ErlNifIOVec*)alloc_base;
@@ -3658,7 +3633,7 @@ static int create_iovec_from_slice(ErlNifEnv *env,
 
     if(!fill_iovec_with_slice(env, slice, iovec)) {
         if (env == NULL && !(iovec->flags & ERL_NIF_IOVEC_FLAGS_PREALLOC)) {
-            enif_free(iovec);
+            erts_free(ERTS_ALC_T_NIF, iovec);
         }
 
         return 0;
@@ -3725,6 +3700,55 @@ int enif_ioq_deq(ErlNifIOQueue *q, size_t elems, size_t *size)
     return 1;
 }
 
+int enif_ioq_peek_head(ErlNifEnv *env, ErlNifIOQueue *q, size_t *size, ERL_NIF_TERM *bin_term) {
+    SysIOVec *iov_entry;
+    Binary *ref_bin;
+
+    if (q->size == 0) {
+        return 0;
+    }
+
+    ASSERT(q->b_head != q->b_tail && q->v_head != q->v_tail);
+
+    ref_bin = &q->b_head[0]->nif;
+    iov_entry = &q->v_head[0];
+
+    if (size != NULL) {
+        *size = iov_entry->iov_len;
+    }
+
+    if (iov_entry->iov_len > ERL_ONHEAP_BIN_LIMIT) {
+        ProcBin *pb = (ProcBin*)alloc_heap(env, PROC_BIN_SIZE);
+
+        pb->thing_word = HEADER_PROC_BIN;
+        pb->next = MSO(env->proc).first;
+        pb->val = ref_bin;
+        pb->flags = 0;
+
+        ASSERT((byte*)iov_entry->iov_base >= (byte*)ref_bin->orig_bytes);
+        ASSERT(iov_entry->iov_len <= ref_bin->orig_size);
+
+        pb->bytes = (byte*)iov_entry->iov_base;
+        pb->size = iov_entry->iov_len;
+
+        MSO(env->proc).first = (struct erl_off_heap_header*) pb;
+        OH_OVERHEAD(&(MSO(env->proc)), pb->size / sizeof(Eterm));
+
+        erts_refc_inc(&ref_bin->intern.refc, 2);
+        *bin_term = make_binary(pb);
+    } else {
+        ErlHeapBin* hb = (ErlHeapBin*)alloc_heap(env, heap_bin_size(iov_entry->iov_len));
+
+        hb->thing_word = header_heap_bin(iov_entry->iov_len);
+        hb->size = iov_entry->iov_len;
+
+        sys_memcpy(hb->data, iov_entry->iov_base, iov_entry->iov_len);
+        *bin_term = make_binary(hb);
+    }
+
+    return 1;
+}
+
 SysIOVec *enif_ioq_peek(ErlNifIOQueue *q, int *iovlen)
 {
     return erts_ioq_peekq(q, iovlen);
@@ -3741,7 +3765,7 @@ static ErtsCodeInfo** get_func_pp(BeamCodeHeader* mod_code, Eterm f_atom, unsign
     int j;
     for (j = 0; j < n; ++j) {
 	ErtsCodeInfo* ci = mod_code->functions[j];
-	ASSERT(ci->op == (BeamInstr) BeamOp(op_i_func_info_IaaI));
+	ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
 	if (f_atom == ci->mfa.function
 	    && arity == ci->mfa.arity) {
 	    return mod_code->functions+j;
@@ -3755,16 +3779,26 @@ static Eterm mkatom(const char *str)
     return am_atom_put(str, sys_strlen(str));
 }
 
-static struct tainted_module_t
+struct tainted_module_t
 {
     struct tainted_module_t* next;
     Eterm module_atom;
-}*first_tainted_module = NULL;
+};
 
-static void add_taint(Eterm mod_atom)
+erts_atomic_t first_taint; /* struct tainted_module_t* */
+
+void erts_add_taint(Eterm mod_atom)
 {
-    struct tainted_module_t* t;
-    for (t=first_tainted_module ; t!=NULL; t=t->next) {
+#ifdef ERTS_ENABLE_LOCK_CHECK
+    extern erts_rwmtx_t erts_driver_list_lock; /* Mutex for driver list */
+#endif
+    struct tainted_module_t *first, *t;
+
+    ERTS_LC_ASSERT(erts_lc_rwmtx_is_rwlocked(&erts_driver_list_lock)
+                   || erts_thr_progress_is_blocking());
+
+    first = (struct tainted_module_t*) erts_atomic_read_nob(&first_taint);
+    for (t=first ; t; t=t->next) {
 	if (t->module_atom == mod_atom) {
 	    return;
 	}
@@ -3772,22 +3806,24 @@ static void add_taint(Eterm mod_atom)
     t = erts_alloc_fnf(ERTS_ALC_T_TAINT, sizeof(*t));
     if (t != NULL) {
 	t->module_atom = mod_atom;
-	t->next = first_tainted_module;
-	first_tainted_module = t;
+	t->next = first;
+        erts_atomic_set_nob(&first_taint, (erts_aint_t)t);
     }
 }
 
 Eterm erts_nif_taints(Process* p)
 {
-    struct tainted_module_t* t;
+    struct tainted_module_t *first, *t;
     unsigned cnt = 0;
     Eterm list = NIL;
     Eterm* hp;
-    for (t=first_tainted_module ; t!=NULL; t=t->next) {
+
+    first = (struct tainted_module_t*) erts_atomic_read_nob(&first_taint);
+    for (t=first ; t!=NULL; t=t->next) {
 	cnt++;
     }
     hp = HAlloc(p,cnt*2);
-    for (t=first_tainted_module ; t!=NULL; t=t->next) {
+    for (t=first ; t!=NULL; t=t->next) {
 	list = CONS(hp, t->module_atom, list);
 	hp += 2;
     }
@@ -3796,9 +3832,11 @@ Eterm erts_nif_taints(Process* p)
 
 void erts_print_nif_taints(fmtfn_t to, void* to_arg)
 {
-    struct tainted_module_t* t;
+    struct tainted_module_t *t;
     const char* delim = "";
-    for (t=first_tainted_module ; t!=NULL; t=t->next) {
+
+    t = (struct tainted_module_t*) erts_atomic_read_nob(&first_taint);
+    for ( ; t; t = t->next) {
 	const Atom* atom = atom_tab(atom_val(t->module_atom));
 	erts_cbprintf(to,to_arg,"%s%.*s", delim, atom->len, atom->name);
 	delim = ",";
@@ -3885,6 +3923,11 @@ static struct erl_module_nif* create_lib(const ErlNifEntry* src)
     } else {
         dst->sizeof_ErlNifResourceTypeInit = 0;
     }
+    if (AT_LEAST_VERSION(src, 2, 14)) {
+        dst->min_erts = src->min_erts;
+    } else {
+        dst->min_erts = "erts-?";
+    }
     return lib;
 };
 
@@ -3907,6 +3950,7 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
     ErtsSysDdllError errdesc = ERTS_SYS_DDLL_ERROR_INIT;
     Eterm ret = am_ok;
     int veto;
+    int taint = 1;
     struct erl_module_nif* lib = NULL;
     struct erl_module_instance* this_mi;
     struct erl_module_instance* prev_mi;
@@ -3937,8 +3981,8 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
     }
 
     /* Block system (is this the right place to do it?) */
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
+    erts_proc_unlock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
 
     /* Find calling module */
     ASSERT(BIF_P->current != NULL);
@@ -3953,10 +3997,15 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
     ASSERT(module_p != NULL);
 
     mod_atomp = atom_tab(atom_val(mod_atom));
-    init_func = erts_static_nif_get_nif_init((char*)mod_atomp->name, mod_atomp->len);
-    if (init_func != NULL)
-      handle = init_func;
-
+    {
+        ErtsStaticNifEntry* sne;
+        sne = erts_static_nif_get_nif_init((char*)mod_atomp->name, mod_atomp->len);
+        if (sne != NULL) {
+            init_func = sne->nif_init;
+            handle = init_func;
+            taint = sne->taint;
+        }
+    }
     this_mi = &module_p->curr;
     prev_mi = &module_p->old;
     if (in_area(caller, module_p->old.code_hdr, module_p->old.code_length)) {
@@ -3993,18 +4042,24 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
 			      " function: '%s'", errdesc.str);
 	
     }
-    else if ((add_taint(mod_atom),
+    else if ((taint ? erts_add_taint(mod_atom) : 0,
 	      (entry = erts_sys_ddll_call_nif_init(init_func)) == NULL)) {
 	ret = load_nif_error(BIF_P, bad_lib, "Library init-call unsuccessful");
     }
+    else if (entry->major > ERL_NIF_MAJOR_VERSION
+             || (entry->major == ERL_NIF_MAJOR_VERSION
+                 && entry->minor > ERL_NIF_MINOR_VERSION)) {
+        char* fmt = "That '%T' NIF library needs %s or newer. Either try to"
+            " recompile the NIF lib or use a newer erts runtime.";
+        ret = load_nif_error(BIF_P, bad_lib, fmt, mod_atom, entry->min_erts);
+    }
     else if (entry->major < ERL_NIF_MIN_REQUIRED_MAJOR_VERSION_ON_LOAD
-	     || (ERL_NIF_MAJOR_VERSION < entry->major
-		 || (ERL_NIF_MAJOR_VERSION == entry->major
-		     && ERL_NIF_MINOR_VERSION < entry->minor))
 	     || (entry->major==2 && entry->minor == 5)) { /* experimental maps */
 	
-	ret = load_nif_error(BIF_P, bad_lib, "Library version (%d.%d) not compatible (with %d.%d).",
-			     entry->major, entry->minor, ERL_NIF_MAJOR_VERSION, ERL_NIF_MINOR_VERSION);
+        char* fmt = "That old NIF library (%d.%d) is not compatible with this "
+            "erts runtime (%d.%d). Try recompile the NIF lib.";
+        ret = load_nif_error(BIF_P, bad_lib, fmt, entry->major, entry->minor,
+                             ERL_NIF_MAJOR_VERSION, ERL_NIF_MINOR_VERSION);
     }   
     else if (AT_LEAST_VERSION(entry, 2, 1)
 	     && sys_strcmp(entry->vm_variant, ERL_NIF_VM_VARIANT) != 0) {
@@ -4043,14 +4098,9 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
 		 * dirty scheduler support, treat a non-zero flags field as
 		 * a load error.
 		 */
-#ifdef ERTS_DIRTY_SCHEDULERS
 		if (f->flags != ERL_NIF_DIRTY_JOB_IO_BOUND && f->flags != ERL_NIF_DIRTY_JOB_CPU_BOUND)
 		    ret = load_nif_error(BIF_P, bad_lib, "Illegal flags field value %d for NIF %T:%s/%u",
 					 f->flags, mod_atom, f->name, f->arity);
-#else
-		ret = load_nif_error(BIF_P, bad_lib, "NIF %T:%s/%u requires a runtime with dirty scheduler support.",
-				     mod_atom, f->name, f->arity);
-#endif
 	    }
 	    else if (erts_codeinfo_to_code(ci_pp[1]) - erts_codeinfo_to_code(ci_pp[0])
                      < BEAM_NIF_MIN_FUNC_SZ)
@@ -4115,15 +4165,13 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
             code_ptr = erts_codeinfo_to_code(ci);
 
 	    if (ci->u.gen_bp == NULL) {
-		code_ptr[0] = (BeamInstr) BeamOp(op_call_nif);
+		code_ptr[0] = BeamOpCodeAddr(op_call_nif);
 	    }
 	    else { /* Function traced, patch the original instruction word */
 		GenericBp* g = ci->u.gen_bp;
-		ASSERT(code_ptr[0] ==
-		       (BeamInstr) BeamOp(op_i_generic_breakpoint));
-		g->orig_instr = (BeamInstr) BeamOp(op_call_nif);
+		ASSERT(BeamIsOpCode(code_ptr[0], op_i_generic_breakpoint));
+		g->orig_instr = BeamOpCodeAddr(op_call_nif);
 	    }
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    if (f->flags) {
 		code_ptr[3] = (BeamInstr) f->fptr;
 		code_ptr[1] = (f->flags == ERL_NIF_DIRTY_JOB_IO_BOUND) ?
@@ -4131,7 +4179,6 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
 		    (BeamInstr) static_schedule_dirty_cpu_nif;
 	    }
 	    else
-#endif
 		code_ptr[1] = (BeamInstr) f->fptr;
 	    code_ptr[2] = (BeamInstr) lib;
 	}
@@ -4149,8 +4196,8 @@ BIF_RETTYPE load_nif_2(BIF_ALIST_2)
 	erts_sys_ddll_free_error(&errdesc);
     }
 
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
     erts_release_code_write_permission();
     erts_free(ERTS_ALC_T_TMP, lib_name);
 
@@ -4163,7 +4210,7 @@ erts_unload_nif(struct erl_module_nif* lib)
 {
     ErlNifResourceType* rt;
     ErlNifResourceType* next;
-    ASSERT(erts_smp_thr_progress_is_blocking());
+    ASSERT(erts_thr_progress_is_blocking());
     ASSERT(lib != NULL);
     ASSERT(lib->mod != NULL);
 
@@ -4222,6 +4269,10 @@ int erts_nif_get_funcs(struct erl_module_nif* mod,
     return mod->entry.num_of_funcs;
 }
 
+Module *erts_nif_get_module(struct erl_module_nif *nif_mod) {
+    return nif_mod->mod;
+}
+
 Eterm erts_nif_call_function(Process *p, Process *tracee,
                              struct erl_module_nif* mod,
                              ErlNifFunc *fun, int argc, Eterm *argv)
@@ -4235,8 +4286,8 @@ Eterm erts_nif_call_function(Process *p, Process *tracee,
             break;
     ASSERT(i < mod->entry.num_of_funcs);
     if (p)
-        ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(p) & ERTS_PROC_LOCK_MAIN
-                           || erts_smp_thr_progress_is_blocking());
+        ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(p) & ERTS_PROC_LOCK_MAIN
+                           || erts_thr_progress_is_blocking());
 #endif
     if (p) {
         /* This is almost a normal nif call like in beam_emu,
@@ -4307,34 +4358,31 @@ static unsigned calc_checksum(unsigned char* ptr, unsigned size);
 
 struct readonly_check_t
 {
-    struct enif_tmp_obj_t hdr;
     unsigned char* ptr;
     unsigned size;
     unsigned checksum;
 };
 static void add_readonly_check(ErlNifEnv* env, unsigned char* ptr, unsigned sz)
 {
-    ErtsAlcType_t allocator = is_proc_bound(env) ? ERTS_ALC_T_TMP : ERTS_ALC_T_NIF;
-    struct readonly_check_t* obj = erts_alloc(allocator, 
-					      sizeof(struct readonly_check_t));
-    obj->hdr.allocator = allocator;
-    obj->hdr.next = env->tmp_obj_list;
-    env->tmp_obj_list = &obj->hdr;
-    obj->hdr.dtor = &readonly_check_dtor;
+    struct readonly_check_t* obj;
+
+    obj = alloc_tmp_obj(env, sizeof(struct readonly_check_t),
+        &readonly_check_dtor);
+
     obj->ptr = ptr;
     obj->size = sz;
-    obj->checksum = calc_checksum(ptr, sz);    
+    obj->checksum = calc_checksum(ptr, sz);
 }
-static void readonly_check_dtor(struct enif_tmp_obj_t* o)
+static void readonly_check_dtor(struct enif_tmp_obj_t* tmp_obj)
 {
-    struct readonly_check_t* obj = (struct readonly_check_t*) o;
-    unsigned chksum = calc_checksum(obj->ptr, obj->size);
-    if (chksum != obj->checksum) { 
+    struct readonly_check_t* ro_check = (struct readonly_check_t*)&tmp_obj[1];
+    unsigned chksum = calc_checksum(ro_check->ptr, ro_check->size);
+    if (chksum != ro_check->checksum) { 
 	fprintf(stderr, "\r\nReadonly data written by NIF, checksums differ"
-		" %x != %x\r\nABORTING\r\n", chksum, obj->checksum);
+		" %x != %x\r\nABORTING\r\n", chksum, ro_check->checksum);
 	abort();
     }
-    erts_free(obj->hdr.allocator,  obj);
+    erts_free(tmp_obj->allocator, tmp_obj);
 }
 static unsigned calc_checksum(unsigned char* ptr, unsigned size)
 {
@@ -4409,7 +4457,7 @@ static void get_string_maybe(ErlNifEnv *env, const ERL_NIF_TERM term,
         str_bin.size > bufsiz) {
         *ptr = NULL;
     } else {
-        memcpy(buf, (char *) str_bin.data, str_bin.size);
+        sys_memcpy(buf, (char *) str_bin.data, str_bin.size);
         buf[str_bin.size] = '\0';
         *ptr = buf;
     }
@@ -4426,7 +4474,7 @@ ERL_NIF_TERM erl_nif_user_trace_s1(ErlNifEnv* env, int argc,
 	    message_bin.size > MESSAGE_BUFSIZ) {
 	    return am_badarg;
 	}
-	memcpy(messagebuf, (char *) message_bin.data, message_bin.size);
+	sys_memcpy(messagebuf, (char *) message_bin.data, message_bin.size);
         messagebuf[message_bin.size] = '\0';
 	DTRACE1(user_trace_s1, messagebuf);
 	return am_true;
diff --git a/erts/emulator/beam/erl_nif.h b/erts/emulator/beam/erl_nif.h
index 7fb447e4a8..1906da732b 100644
--- a/erts/emulator/beam/erl_nif.h
+++ b/erts/emulator/beam/erl_nif.h
@@ -52,9 +52,12 @@
 ** 2.11: 19.0 enif_snprintf 
 ** 2.12: 20.0 add enif_select, enif_open_resource_type_x
 ** 2.13: 20.1 add enif_ioq
+** 2.14: 21.0 add enif_ioq_peek_head, enif_(mutex|cond|rwlock|thread)_name
+**                enif_vfprintf, enif_vsnprintf, enif_make_map_from_arrays
 */
 #define ERL_NIF_MAJOR_VERSION 2
-#define ERL_NIF_MINOR_VERSION 13
+#define ERL_NIF_MINOR_VERSION 14
+#define ERL_NIF_MIN_ERTS_VERSION "erts-10.0 (OTP-21)"
 
 /*
  * The emulator will refuse to load a nif-lib with a major version
@@ -69,6 +72,8 @@
 #define ERL_NIF_MIN_REQUIRED_MAJOR_VERSION_ON_LOAD 2
 
 #include <stdlib.h>
+#include <stdio.h>
+#include <stdarg.h>
 
 #ifdef __cplusplus
 extern "C" {
@@ -127,6 +132,9 @@ typedef struct enif_entry_t
 
     /* Added in 2.12 */
     size_t sizeof_ErlNifResourceTypeInit;
+
+    /* Added in 2.14 */
+    const char* min_erts;
 }ErlNifEntry;
 
 
@@ -136,8 +144,9 @@ typedef struct
     unsigned char* data;
 
     /* Internals (avert your eyes) */
-    ERL_NIF_TERM bin_term;
     void* ref_bin;
+    /* for future additions to be ABI compatible (same struct size) */
+    void* __spare__[2];
 }ErlNifBinary;
 
 #if (defined(__WIN32__) || defined(_WIN32) || defined(_WIN32_))
@@ -348,7 +357,8 @@ ERL_NIF_INIT_DECL(NAME)			\
 	LOAD, RELOAD, UPGRADE, UNLOAD,	\
 	ERL_NIF_VM_VARIANT,		\
         1,                              \
-        sizeof(ErlNifResourceTypeInit)  \
+        sizeof(ErlNifResourceTypeInit), \
+        ERL_NIF_MIN_ERTS_VERSION        \
     };                                  \
     ERL_NIF_INIT_BODY;                  \
     return &entry;			\
diff --git a/erts/emulator/beam/erl_nif_api_funcs.h b/erts/emulator/beam/erl_nif_api_funcs.h
index 9e573307d8..61f8fcf6ed 100644
--- a/erts/emulator/beam/erl_nif_api_funcs.h
+++ b/erts/emulator/beam/erl_nif_api_funcs.h
@@ -92,7 +92,7 @@ ERL_NIF_API_FUNC_DECL(int,enif_thread_join,(ErlNifTid, void **respp));
 
 ERL_NIF_API_FUNC_DECL(void*,enif_realloc,(void* ptr, size_t size));
 ERL_NIF_API_FUNC_DECL(void,enif_system_info,(ErlNifSysInfo *sip, size_t si_size));
-ERL_NIF_API_FUNC_DECL(int,enif_fprintf,(void/* FILE* */ *filep, const char *format, ...));
+ERL_NIF_API_FUNC_DECL(int,enif_fprintf,(FILE* filep, const char *format, ...));
 ERL_NIF_API_FUNC_DECL(int,enif_inspect_iolist_as_binary,(ErlNifEnv*, ERL_NIF_TERM term, ErlNifBinary* bin));
 ERL_NIF_API_FUNC_DECL(ERL_NIF_TERM,enif_make_sub_binary,(ErlNifEnv*, ERL_NIF_TERM bin_term, size_t pos, size_t size));
 ERL_NIF_API_FUNC_DECL(int,enif_get_string,(ErlNifEnv*, ERL_NIF_TERM list, char* buf, unsigned len, ErlNifCharEncoding));
@@ -198,6 +198,17 @@ ERL_NIF_API_FUNC_DECL(SysIOVec*,enif_ioq_peek,(ErlNifIOQueue *q, int *iovlen));
 ERL_NIF_API_FUNC_DECL(int,enif_inspect_iovec,(ErlNifEnv *env, size_t max_length, ERL_NIF_TERM iovec_term, ERL_NIF_TERM *tail, ErlNifIOVec **iovec));
 ERL_NIF_API_FUNC_DECL(void,enif_free_iovec,(ErlNifIOVec *iov));
 
+ERL_NIF_API_FUNC_DECL(int,enif_ioq_peek_head,(ErlNifEnv *env, ErlNifIOQueue *q, size_t *size, ERL_NIF_TERM *head));
+
+ERL_NIF_API_FUNC_DECL(char*,enif_mutex_name,(ErlNifMutex*));
+ERL_NIF_API_FUNC_DECL(char*,enif_cond_name,(ErlNifCond*));
+ERL_NIF_API_FUNC_DECL(char*,enif_rwlock_name,(ErlNifRWLock*));
+ERL_NIF_API_FUNC_DECL(char*,enif_thread_name,(ErlNifTid));
+
+ERL_NIF_API_FUNC_DECL(int,enif_vfprintf,(FILE*, const char *fmt, va_list));
+ERL_NIF_API_FUNC_DECL(int,enif_vsnprintf,(char*, size_t, const char *fmt, va_list));
+
+ERL_NIF_API_FUNC_DECL(int,enif_make_map_from_arrays,(ErlNifEnv *env, ERL_NIF_TERM keys[], ERL_NIF_TERM values[], size_t cnt, ERL_NIF_TERM *map_out));
 
 /*
 ** ADD NEW ENTRIES HERE (before this comment) !!!
@@ -373,6 +384,14 @@ ERL_NIF_API_FUNC_DECL(void,enif_free_iovec,(ErlNifIOVec *iov));
 #  define enif_ioq_peek ERL_NIF_API_FUNC_MACRO(enif_ioq_peek)
 #  define enif_inspect_iovec ERL_NIF_API_FUNC_MACRO(enif_inspect_iovec)
 #  define enif_free_iovec ERL_NIF_API_FUNC_MACRO(enif_free_iovec)
+#  define enif_ioq_peek_head ERL_NIF_API_FUNC_MACRO(enif_ioq_peek_head)
+#  define enif_mutex_name ERL_NIF_API_FUNC_MACRO(enif_mutex_name)
+#  define enif_cond_name ERL_NIF_API_FUNC_MACRO(enif_cond_name)
+#  define enif_rwlock_name ERL_NIF_API_FUNC_MACRO(enif_rwlock_name)
+#  define enif_thread_name ERL_NIF_API_FUNC_MACRO(enif_thread_name)
+#  define enif_vfprintf ERL_NIF_API_FUNC_MACRO(enif_vfprintf)
+#  define enif_vsnprintf ERL_NIF_API_FUNC_MACRO(enif_vsnprintf)
+#  define enif_make_map_from_arrays ERL_NIF_API_FUNC_MACRO(enif_make_map_from_arrays)
 
 /*
 ** ADD NEW ENTRIES HERE (before this comment)
diff --git a/erts/emulator/beam/erl_node_container_utils.h b/erts/emulator/beam/erl_node_container_utils.h
index 6ec428e282..99e938266b 100644
--- a/erts/emulator/beam/erl_node_container_utils.h
+++ b/erts/emulator/beam/erl_node_container_utils.h
@@ -318,5 +318,3 @@ extern ErtsPTab erts_port;
 #define is_not_ref(x)			(!is_ref(x))
 
 #endif
-
-
diff --git a/erts/emulator/beam/erl_node_tables.c b/erts/emulator/beam/erl_node_tables.c
index deadf435e9..1f147011a8 100644
--- a/erts/emulator/beam/erl_node_tables.c
+++ b/erts/emulator/beam/erl_node_tables.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2001-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2001-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -29,17 +29,22 @@
 #include "error.h"
 #include "erl_thr_progress.h"
 #include "dtrace-wrapper.h"
+#include "erl_binary.h"
+#include "erl_bif_unique.h"
+#include "erl_proc_sig_queue.h"
 
 Hash erts_dist_table;
 Hash erts_node_table;
-erts_smp_rwmtx_t erts_dist_table_rwmtx;
-erts_smp_rwmtx_t erts_node_table_rwmtx;
+erts_rwmtx_t erts_dist_table_rwmtx;
+erts_rwmtx_t erts_node_table_rwmtx;
 
 DistEntry *erts_hidden_dist_entries;
 DistEntry *erts_visible_dist_entries;
+DistEntry *erts_pending_dist_entries;
 DistEntry *erts_not_connected_dist_entries; /* including erts_this_dist_entry */
 Sint erts_no_of_hidden_dist_entries;
 Sint erts_no_of_visible_dist_entries;
+Sint erts_no_of_pending_dist_entries;
 Sint erts_no_of_not_connected_dist_entries; /* including erts_this_dist_entry */
 
 DistEntry *erts_this_dist_entry;
@@ -57,6 +62,60 @@ static ErtsMonotonicTime node_tab_delete_delay;
 
 /* -- The distribution table ---------------------------------------------- */
 
+#define ErtsBin2DistEntry(B) \
+    ((DistEntry *) ERTS_MAGIC_BIN_DATA((B)))
+#define ErtsDistEntry2Bin(DEP) \
+    ((Binary *) ERTS_MAGIC_BIN_FROM_DATA((DEP)))
+
+static ERTS_INLINE erts_aint_t
+de_refc_read(DistEntry *dep, erts_aint_t min)
+{
+    return erts_refc_read(&ErtsDistEntry2Bin(dep)->intern.refc, min);
+}
+
+static ERTS_INLINE erts_aint_t
+de_refc_inc_read(DistEntry *dep, erts_aint_t min)
+{
+    return erts_refc_inctest(&ErtsDistEntry2Bin(dep)->intern.refc, min);
+}
+
+static ERTS_INLINE void
+de_refc_inc(DistEntry *dep, erts_aint_t min)
+{
+    erts_refc_inc(&ErtsDistEntry2Bin(dep)->intern.refc, min);
+}
+
+static ERTS_INLINE void
+de_refc_dec(DistEntry *dep, erts_aint_t min)
+{
+#ifdef DEBUG
+    (void) erts_refc_read(&ErtsDistEntry2Bin(dep)->intern.refc, min+1);
+#endif
+    erts_bin_release(ErtsDistEntry2Bin(dep));
+}
+
+static ERTS_INLINE erts_aint_t
+de_refc_dec_read(DistEntry *dep, erts_aint_t min)
+{
+    return erts_refc_dectest(&ErtsDistEntry2Bin(dep)->intern.refc, min);
+}
+
+void
+erts_ref_dist_entry(DistEntry *dep)
+{
+    ASSERT(dep);
+    if (de_refc_inc_read(dep, 1) == 1) {
+        de_refc_inc(dep, 2);  /* Pending delete */
+    }
+}
+
+void
+erts_deref_dist_entry(DistEntry *dep)
+{
+    ASSERT(dep);
+    de_refc_dec(dep, 0);
+}
+
 #ifdef DEBUG
 static int
 is_in_de_list(DistEntry *dep, DistEntry *dep_list)
@@ -85,48 +144,59 @@ dist_table_cmp(void *dep1, void *dep2)
 static void*
 dist_table_alloc(void *dep_tmpl)
 {
+    erts_aint_t refc;
     Eterm sysname;
+    Binary *bin;
     DistEntry *dep;
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
 
     sysname = ((DistEntry *) dep_tmpl)->sysname;
-    dep = (DistEntry *) erts_alloc(ERTS_ALC_T_DIST_ENTRY, sizeof(DistEntry));
+
+    bin = erts_create_magic_binary_x(sizeof(DistEntry),
+                                     erts_dist_entry_destructor,
+                                     ERTS_ALC_T_DIST_ENTRY,
+                                     0);
+    dep = ErtsBin2DistEntry(bin);
 
     dist_entries++;
 
+    refc = de_refc_dec_read(dep, -1);
+    ASSERT(refc == -1); (void)refc;
+
     dep->prev				= NULL;
-    erts_smp_refc_init(&dep->refc, -1);
-    erts_smp_rwmtx_init_opt(&dep->rwmtx, &rwmtx_opt, "dist_entry", sysname,
+    erts_rwmtx_init_opt(&dep->rwmtx, &rwmtx_opt, "dist_entry", sysname,
         ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
     dep->sysname			= sysname;
     dep->cid				= NIL;
+    erts_atomic_init_nob(&dep->input_handler, (erts_aint_t) NIL);
     dep->connection_id			= 0;
-    dep->status				= 0;
+    dep->state				= ERTS_DE_STATE_IDLE;
     dep->flags				= 0;
     dep->version			= 0;
 
-    erts_smp_mtx_init(&dep->lnk_mtx, "dist_entry_links", sysname,
-        ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
-    dep->node_links			= NULL;
-    dep->nlinks				= NULL;
-    dep->monitors			= NULL;
+    dep->mld                            = NULL;
 
-    erts_smp_mtx_init(&dep->qlock, "dist_entry_out_queue", sysname,
+    erts_mtx_init(&dep->qlock, "dist_entry_out_queue", sysname,
         ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
-    dep->qflgs				= 0;
-    dep->qsize				= 0;
+    erts_atomic32_init_nob(&dep->qflgs, 0);
+    erts_atomic_init_nob(&dep->qsize, 0);
+    erts_atomic64_init_nob(&dep->in, 0);
+    erts_atomic64_init_nob(&dep->out, 0);
     dep->out_queue.first		= NULL;
     dep->out_queue.last			= NULL;
     dep->suspended			= NULL;
 
+    dep->tmp_out_queue.first    	= NULL;
+    dep->tmp_out_queue.last     	= NULL;
     dep->finalized_out_queue.first	= NULL;
     dep->finalized_out_queue.last	= NULL;
 
-    erts_smp_atomic_init_nob(&dep->dist_cmd_scheduled, 0);
+    erts_atomic_init_nob(&dep->dist_cmd_scheduled, 0);
     erts_port_task_handle_init(&dep->dist_cmd);
     dep->send				= NULL;
     dep->cache				= NULL;
+    dep->transcode_ctx                  = NULL;
 
     /* Link in */
 
@@ -149,10 +219,10 @@ dist_table_free(void *vdep)
 {
     DistEntry *dep = (DistEntry *) vdep;
 
+    ASSERT(de_refc_read(dep, -1) == -1);
+    ASSERT(dep->state == ERTS_DE_STATE_IDLE);
     ASSERT(is_nil(dep->cid));
-    ASSERT(dep->nlinks == NULL);
-    ASSERT(dep->node_links == NULL);
-    ASSERT(dep->monitors == NULL);
+    ASSERT(dep->mld == NULL);
 
     /* Link out */
 
@@ -174,14 +244,13 @@ dist_table_free(void *vdep)
     erts_no_of_not_connected_dist_entries--;
 
     ASSERT(!dep->cache);
-    erts_smp_rwmtx_destroy(&dep->rwmtx);
-    erts_smp_mtx_destroy(&dep->lnk_mtx);
-    erts_smp_mtx_destroy(&dep->qlock);
+    erts_rwmtx_destroy(&dep->rwmtx);
+    erts_mtx_destroy(&dep->qlock);
 
 #ifdef DEBUG
     sys_memset(vdep, 0x77, sizeof(DistEntry));
 #endif
-    erts_free(ERTS_ALC_T_DIST_ENTRY, (void *) dep);
+    erts_bin_free(ErtsDistEntry2Bin(dep));
 
     ASSERT(dist_entries > 0);
     dist_entries--;
@@ -193,25 +262,58 @@ erts_dist_table_info(fmtfn_t to, void *to_arg)
 {
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
-	erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+	erts_rwmtx_rlock(&erts_dist_table_rwmtx);
     hash_info(to, to_arg, &erts_dist_table);
     if (lock)
-	erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+	erts_rwmtx_runlock(&erts_dist_table_rwmtx);
+}
+
+static ERTS_INLINE DistEntry *find_dist_entry(Eterm sysname,
+                                              int inc_refc,
+                                              int connected_only)
+{
+    DistEntry *res;
+    DistEntry de;
+    de.sysname = sysname;
+    erts_rwmtx_rlock(&erts_dist_table_rwmtx);
+    res = hash_get(&erts_dist_table, (void *) &de);
+    if (res) {
+        if (connected_only && is_nil(res->cid))
+            res = NULL;
+        else {
+            int pend_delete;
+            erts_aint_t refc;
+            if (inc_refc) {
+                refc = de_refc_inc_read(res, 1);
+                pend_delete = refc < 2;
+            }
+            else {
+                refc = de_refc_read(res, 0);
+                pend_delete = refc < 1;
+            }
+            if (pend_delete) /* Pending delete */
+                de_refc_inc(res, 1);
+        }
+    }
+    erts_rwmtx_runlock(&erts_dist_table_rwmtx);
+    return res;
 }
 
 DistEntry *
 erts_channel_no_to_dist_entry(Uint cno)
 {
+    /*
+     * Does NOT increase reference count!
+     */
+
 /*
  * For this node (and previous incarnations of this node),
  * ERST_INTERNAL_CHANNEL_NO (will always be 0 I guess) is used as
  * channel no. For other nodes, the atom index of the atom corresponding
  * to the node name is used as channel no.
  */
-    if(cno == ERST_INTERNAL_CHANNEL_NO) {
-	erts_smp_refc_inc(&erts_this_dist_entry->refc, 2);
+    if (cno == ERST_INTERNAL_CHANNEL_NO)
 	return erts_this_dist_entry;
-    }
 
     if((cno > MAX_ATOM_INDEX)
        || (cno >= atom_table_size())
@@ -220,118 +322,174 @@ erts_channel_no_to_dist_entry(Uint cno)
         
     /* cno is a valid atom index; find corresponding dist entry (if there
        is one) */
-    return erts_find_dist_entry(make_atom(cno));
+    return find_dist_entry(make_atom(cno), 0, 0);
 }
 
-
 DistEntry *
 erts_sysname_to_connected_dist_entry(Eterm sysname)
 {
-    DistEntry de;
-    DistEntry *res_dep;
-    de.sysname = sysname;
-  
-    if(erts_this_dist_entry->sysname == sysname) {
-	erts_smp_refc_inc(&erts_this_dist_entry->refc, 2);
+    /*
+     * Does NOT increase reference count!
+     */
+    if(erts_this_dist_entry->sysname == sysname)
 	return erts_this_dist_entry;
-    }
-
-    erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
-    res_dep = (DistEntry *) hash_get(&erts_dist_table, (void *) &de);
-    if (res_dep) {
-	erts_aint_t refc = erts_smp_refc_inctest(&res_dep->refc, 1);
-	if (refc < 2) /* Pending delete */
-	    erts_smp_refc_inc(&res_dep->refc, 1);
-    }
-    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
-    if (res_dep) {
-	int deref;
-	erts_smp_rwmtx_rlock(&res_dep->rwmtx);
-	deref = is_nil(res_dep->cid);
-	erts_smp_rwmtx_runlock(&res_dep->rwmtx);
-	if (deref) {
-	    erts_deref_dist_entry(res_dep);
-	    res_dep = NULL;
-	}
-    }
-    return res_dep;
+    return find_dist_entry(sysname, 0, 1);
 }
 
 DistEntry *erts_find_or_insert_dist_entry(Eterm sysname)
 {
+    /*
+     * This function DOES increase reference count!
+     */
     DistEntry *res;
     DistEntry de;
     erts_aint_t refc;
-    res = erts_find_dist_entry(sysname);
+    res = find_dist_entry(sysname, 1, 0);
     if (res)
 	return res;
     de.sysname = sysname;
-    erts_smp_rwmtx_rwlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rwlock(&erts_dist_table_rwmtx);
     res = hash_put(&erts_dist_table, (void *) &de);
-    refc = erts_smp_refc_inctest(&res->refc, 0);
+    refc = de_refc_inc_read(res, 0);
     if (refc < 2) /* New or pending delete */
-	erts_smp_refc_inc(&res->refc, 1);
-    erts_smp_rwmtx_rwunlock(&erts_dist_table_rwmtx);
+        de_refc_inc(res, 1);
+    erts_rwmtx_rwunlock(&erts_dist_table_rwmtx);
     return res;
 }
 
 DistEntry *erts_find_dist_entry(Eterm sysname)
 {
-    DistEntry *res;
-    DistEntry de;
-    de.sysname = sysname;
-    erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
-    res = hash_get(&erts_dist_table, (void *) &de);
-    if (res) {
-	erts_aint_t refc = erts_smp_refc_inctest(&res->refc, 1);
-	if (refc < 2) /* Pending delete */
-	    erts_smp_refc_inc(&res->refc, 1);
+    /*
+     * Does NOT increase reference count!
+     */
+    return find_dist_entry(sysname, 0, 0);
+}
+
+DistEntry *
+erts_dhandle_to_dist_entry(Eterm dhandle)
+{
+    Binary *bin;
+    if (!is_internal_magic_ref(dhandle))
+        return NULL;
+    bin = erts_magic_ref2bin(dhandle);
+    if (ERTS_MAGIC_BIN_DESTRUCTOR(bin) != erts_dist_entry_destructor)
+        return NULL;
+    return ErtsBin2DistEntry(bin);
+}
+
+Eterm
+erts_build_dhandle(Eterm **hpp, ErlOffHeap* ohp, DistEntry *dep)
+{
+    Binary *bin = ErtsDistEntry2Bin(dep);
+    ASSERT(bin);
+    ASSERT(ERTS_MAGIC_BIN_DESTRUCTOR(bin) == erts_dist_entry_destructor);
+    return erts_mk_magic_ref(hpp, ohp, bin);
+}
+
+Eterm
+erts_make_dhandle(Process *c_p, DistEntry *dep)
+{
+    Eterm *hp = HAlloc(c_p, ERTS_MAGIC_REF_THING_SIZE);
+    return erts_build_dhandle(&hp, &c_p->off_heap, dep);
+}
+
+static void start_timer_delete_dist_entry(void *vdep);
+static void prepare_try_delete_dist_entry(void *vdep);
+static void try_delete_dist_entry(DistEntry*);
+
+static void schedule_delete_dist_entry(DistEntry* dep)
+{
+    /*
+     * Here we need thread progress to wait for other threads, that may have
+     * done lookup without refc++, to do either refc++ or drop their refs.
+     *
+     * Note that timeouts do not guarantee thread progress.
+     */
+    erts_schedule_thr_prgr_later_op(start_timer_delete_dist_entry,
+                                    dep, &dep->later_op);
+}
+
+static void
+start_timer_delete_dist_entry(void *vdep)
+{
+    if (node_tab_delete_delay == 0) {
+        prepare_try_delete_dist_entry(vdep);
+    }
+    else {
+        ASSERT(node_tab_delete_delay > 0);
+        erts_start_timer_callback(node_tab_delete_delay,
+                                  prepare_try_delete_dist_entry,
+                                  vdep);
     }
-    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
-    return res;
 }
 
-static void try_delete_dist_entry(void *vdep)
+static void
+prepare_try_delete_dist_entry(void *vdep)
+{
+    DistEntry *dep = vdep;
+    erts_aint_t refc;
+
+    /*
+     * Time has passed since we decremented refc to zero and DistEntry may
+     * have been revived. Do a fast check without table lock first.
+     */
+    refc = de_refc_read(dep, 0);
+    if (refc == 0) {
+        try_delete_dist_entry(dep);
+    }
+    else {
+        /*
+         * Someone has done lookup and done refc++ for us.
+         */
+        refc = de_refc_dec_read(dep, 0);
+        if (refc == 0)
+            schedule_delete_dist_entry(dep);
+    }
+}
+
+static void try_delete_dist_entry(DistEntry* dep)
 {
-    DistEntry *dep = (DistEntry *) vdep;
     erts_aint_t refc;
 
-    erts_smp_rwmtx_rwlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rwlock(&erts_dist_table_rwmtx);
     /*
      * Another thread might have looked up this dist entry after
      * we decided to delete it (refc became zero). If so, the other
-     * thread incremented refc twice. Once for the new reference
-     * and once for this thread.
+     * thread incremented refc one extra step for this thread.
      *
-     * If refc reach -1, no one has used the entry since we
-     * set up the timer. Delete the entry.
+     * If refc reach -1, no one has done lookup and no one can do lookup
+     * as we have table lock. Delete the entry.
      *
-     * If refc reach 0, the entry is currently not in use
-     * but has been used since we set up the timer. Set up a
-     * new timer.
+     * If refc reach 0, someone raced us and either
+     *  (1) did lookup with own refc++ and already released it again
+     *  (2) did lookup without own refc++
+     * Schedule new delete operation.
      *
      * If refc > 0, the entry is in use. Keep the entry.
      */
-    refc = erts_smp_refc_dectest(&dep->refc, -1);
+    refc = de_refc_dec_read(dep, -1);
     if (refc == -1)
 	(void) hash_erase(&erts_dist_table, (void *) dep);
-    erts_smp_rwmtx_rwunlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rwunlock(&erts_dist_table_rwmtx);
 
-    if (refc == 0)
-	erts_schedule_delete_dist_entry(dep);
+    if (refc == 0) {
+        schedule_delete_dist_entry(dep);
+    }
 }
 
-void erts_schedule_delete_dist_entry(DistEntry *dep)
+int erts_dist_entry_destructor(Binary *bin)
 {
-    ASSERT(dep != erts_this_dist_entry);
-    if (dep != erts_this_dist_entry) {
-	if (node_tab_delete_delay == 0)
-	    try_delete_dist_entry((void *) dep);
-	else if (node_tab_delete_delay > 0)
-	    erts_start_timer_callback(node_tab_delete_delay,
-				      try_delete_dist_entry,
-				      (void *) dep);
-    }
+    DistEntry *dep = ErtsBin2DistEntry(bin);
+    erts_aint_t refc;
+
+    refc = de_refc_read(dep, -1);
+
+    if (refc == -1)
+        return 1; /* Allow deallocation of structure... */
+
+    schedule_delete_dist_entry(dep);
+
+    return 0;
 }
 
 Uint
@@ -346,7 +504,7 @@ erts_dist_table_size(void)
     int lock = !ERTS_IS_CRASH_DUMPING;
 
     if (lock)
-	erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+	erts_rwmtx_rlock(&erts_dist_table_rwmtx);
 #ifdef DEBUG
     hash_get_info(&hi, &erts_dist_table);
     ASSERT(dist_entries == hi.objs);
@@ -360,12 +518,17 @@ erts_dist_table_size(void)
 	i++;
     ASSERT(i == erts_no_of_hidden_dist_entries);
     i = 0;
+    for(dep = erts_pending_dist_entries; dep; dep = dep->next)
+        i++;
+    ASSERT(i == erts_no_of_pending_dist_entries);
+    i = 0;
     for(dep = erts_not_connected_dist_entries; dep; dep = dep->next)
 	i++;
     ASSERT(i == erts_no_of_not_connected_dist_entries);
 
     ASSERT(dist_entries == (erts_no_of_visible_dist_entries
 			    + erts_no_of_hidden_dist_entries
+                            + erts_no_of_pending_dist_entries
 			    + erts_no_of_not_connected_dist_entries));
 #endif
 
@@ -373,50 +536,53 @@ erts_dist_table_size(void)
 	   + dist_entries*sizeof(DistEntry)
 	   + erts_dist_cache_size());
     if (lock)
-	erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+	erts_rwmtx_runlock(&erts_dist_table_rwmtx);
     return res;
 }
 
 void
 erts_set_dist_entry_not_connected(DistEntry *dep)
 {
-    ERTS_SMP_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
-    erts_smp_rwmtx_rwlock(&erts_dist_table_rwmtx);
+    DistEntry** head;
 
-    ASSERT(dep != erts_this_dist_entry);
-    ASSERT(is_internal_port(dep->cid));
+    ERTS_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
+    erts_rwmtx_rwlock(&erts_dist_table_rwmtx);
 
-    if(dep->flags & DFLAG_PUBLISHED) {
-	if(dep->prev) {
-	    ASSERT(is_in_de_list(dep, erts_visible_dist_entries));
-	    dep->prev->next = dep->next;
-	}
-	else {
-	    ASSERT(erts_visible_dist_entries == dep);
-	    erts_visible_dist_entries = dep->next;
-	}
+    ASSERT(dep != erts_this_dist_entry);
 
-	ASSERT(erts_no_of_visible_dist_entries > 0);
-	erts_no_of_visible_dist_entries--;
+    if (dep->state == ERTS_DE_STATE_PENDING) {
+        ASSERT(is_nil(dep->cid));
+        ASSERT(erts_no_of_pending_dist_entries > 0);
+        erts_no_of_pending_dist_entries--;
+        head = &erts_pending_dist_entries;
     }
     else {
-	if(dep->prev) {
-	    ASSERT(is_in_de_list(dep, erts_hidden_dist_entries));
-	    dep->prev->next = dep->next;
-	}
-	else {
-	    ASSERT(erts_hidden_dist_entries == dep);
-	    erts_hidden_dist_entries = dep->next;
-	}
-
-	ASSERT(erts_no_of_hidden_dist_entries > 0);
-	erts_no_of_hidden_dist_entries--;
+        ASSERT(dep->state != ERTS_DE_STATE_IDLE);
+        ASSERT(is_internal_port(dep->cid) || is_internal_pid(dep->cid));
+        if (dep->flags & DFLAG_PUBLISHED) {
+            ASSERT(erts_no_of_visible_dist_entries > 0);
+            erts_no_of_visible_dist_entries--;
+            head = &erts_visible_dist_entries;
+        }
+        else {
+            ASSERT(erts_no_of_hidden_dist_entries > 0);
+            erts_no_of_hidden_dist_entries--;
+            head = &erts_hidden_dist_entries;
+        }
     }
 
+    if(dep->prev) {
+        ASSERT(is_in_de_list(dep, *head));
+        dep->prev->next = dep->next;
+    }
+    else {
+        ASSERT(*head == dep);
+        *head = dep->next;
+    }
     if(dep->next)
 	dep->next->prev = dep->prev;
 
-    dep->status &= ~ERTS_DE_SFLG_CONNECTED;
+    dep->state = ERTS_DE_STATE_EXITING;
     dep->flags = 0;
     dep->prev = NULL;
     dep->cid = NIL;
@@ -428,41 +594,98 @@ erts_set_dist_entry_not_connected(DistEntry *dep)
     }
     erts_not_connected_dist_entries = dep;
     erts_no_of_not_connected_dist_entries++;
-    erts_smp_rwmtx_rwunlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rwunlock(&erts_dist_table_rwmtx);
+}
+
+void
+erts_set_dist_entry_pending(DistEntry *dep)
+{
+    ErtsMonLnkDist *mld = erts_mon_link_dist_create(dep->sysname);
+    ERTS_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
+    
+    erts_rwmtx_rwlock(&erts_dist_table_rwmtx);
+
+    ASSERT(dep != erts_this_dist_entry);
+    ASSERT(dep->state == ERTS_DE_STATE_IDLE);
+    ASSERT(is_nil(dep->cid));
+
+    if(dep->prev) {
+        ASSERT(is_in_de_list(dep, erts_not_connected_dist_entries));
+        dep->prev->next = dep->next;
+    }
+    else {
+        ASSERT(dep == erts_not_connected_dist_entries);
+        erts_not_connected_dist_entries = dep->next;
+    }
+
+    if(dep->next)
+	dep->next->prev = dep->prev;
+
+    erts_no_of_not_connected_dist_entries--;
+
+    dep->state = ERTS_DE_STATE_PENDING;
+    dep->flags = (DFLAG_DIST_MANDATORY | DFLAG_DIST_HOPEFULLY | DFLAG_NO_MAGIC);
+    dep->connection_id = (dep->connection_id + 1) & ERTS_DIST_CON_ID_MASK;
+
+    ASSERT(!dep->mld);
+    mld->connection_id = dep->connection_id;
+    dep->mld = mld;
+
+    dep->prev = NULL;
+    dep->next = erts_pending_dist_entries;
+    if(erts_pending_dist_entries) {
+	ASSERT(erts_pending_dist_entries->prev == NULL);
+	erts_pending_dist_entries->prev = dep;
+    }
+    erts_pending_dist_entries = dep;
+    erts_no_of_pending_dist_entries++;
+    erts_rwmtx_rwunlock(&erts_dist_table_rwmtx);
 }
 
 void
 erts_set_dist_entry_connected(DistEntry *dep, Eterm cid, Uint flags)
 {
-    ERTS_SMP_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
-    erts_smp_rwmtx_rwlock(&erts_dist_table_rwmtx);
+    erts_aint32_t set_qflgs;
+
+    ASSERT(dep->mld);
+
+    ERTS_LC_ASSERT(erts_lc_is_de_rwlocked(dep));
+    erts_rwmtx_rwlock(&erts_dist_table_rwmtx);
 
     ASSERT(dep != erts_this_dist_entry);
     ASSERT(is_nil(dep->cid));
-    ASSERT(is_internal_port(cid));
+    ASSERT(dep->state == ERTS_DE_STATE_PENDING);
+    ASSERT(is_internal_port(cid) || is_internal_pid(cid));
 
     if(dep->prev) {
-	ASSERT(is_in_de_list(dep, erts_not_connected_dist_entries));
+	ASSERT(is_in_de_list(dep, erts_pending_dist_entries));
 	dep->prev->next = dep->next;
     }
     else {
-	ASSERT(erts_not_connected_dist_entries == dep);
-	erts_not_connected_dist_entries = dep->next;
+	ASSERT(erts_pending_dist_entries == dep);
+	erts_pending_dist_entries = dep->next;
     }
 
     if(dep->next)
 	dep->next->prev = dep->prev;
 
-    ASSERT(erts_no_of_not_connected_dist_entries > 0);
-    erts_no_of_not_connected_dist_entries--;
+    ASSERT(erts_no_of_pending_dist_entries > 0);
+    erts_no_of_pending_dist_entries--;
 
-    dep->status |= ERTS_DE_SFLG_CONNECTED;
-    dep->flags = flags;
+    dep->state = ERTS_DE_STATE_CONNECTED;
+    dep->flags = flags & ~DFLAG_NO_MAGIC;
     dep->cid = cid;
-    dep->connection_id++;
-    dep->connection_id &= ERTS_DIST_EXT_CON_ID_MASK;
+    erts_atomic_set_nob(&dep->input_handler,
+                            (erts_aint_t) cid);
+
     dep->prev = NULL;
 
+    erts_atomic64_set_nob(&dep->in, 0);
+    erts_atomic64_set_nob(&dep->out, 0);
+    set_qflgs = (is_internal_port(cid) ?
+                 ERTS_DE_QFLG_PORT_CTRL : ERTS_DE_QFLG_PROC_CTRL);
+    erts_atomic32_read_bor_nob(&dep->qflgs, set_qflgs);
+
     if(flags & DFLAG_PUBLISHED) {
 	dep->next = erts_visible_dist_entries;
 	if(erts_visible_dist_entries) {
@@ -481,7 +704,7 @@ erts_set_dist_entry_connected(DistEntry *dep, Eterm cid, Uint flags)
 	erts_hidden_dist_entries = dep;
 	erts_no_of_hidden_dist_entries++;
     }
-    erts_smp_rwmtx_rwunlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rwunlock(&erts_dist_table_rwmtx);
 }
 
 /* -- Node table --------------------------------------------------------- */
@@ -519,7 +742,7 @@ node_table_alloc(void *venp_tmpl)
 
     node_entries++;
 
-    erts_smp_refc_init(&enp->refc, -1);
+    erts_refc_init(&enp->refc, -1);
     enp->creation = ((ErlNode *) venp_tmpl)->creation;
     enp->sysname = ((ErlNode *) venp_tmpl)->sysname;
     enp->dist_entry = erts_find_or_insert_dist_entry(((ErlNode *) venp_tmpl)->sysname);
@@ -532,7 +755,7 @@ node_table_free(void *venp)
 {
     ErlNode *enp = (ErlNode *) venp;
 
-    ERTS_SMP_LC_ASSERT(enp != erts_this_node || erts_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(enp != erts_this_node || erts_thr_progress_is_blocking());
 
     erts_deref_dist_entry(enp->dist_entry);
 #ifdef DEBUG
@@ -553,14 +776,14 @@ erts_node_table_size(void)
 #endif
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
-	erts_smp_rwmtx_rlock(&erts_node_table_rwmtx);
+	erts_rwmtx_rlock(&erts_node_table_rwmtx);
 #ifdef DEBUG
     hash_get_info(&hi, &erts_node_table);
     ASSERT(node_entries == hi.objs);
 #endif
     res = hash_table_sz(&erts_node_table) + node_entries*sizeof(ErlNode);
     if (lock)
-	erts_smp_rwmtx_runlock(&erts_node_table_rwmtx);
+	erts_rwmtx_runlock(&erts_node_table_rwmtx);
     return res;
 }
 
@@ -569,10 +792,10 @@ erts_node_table_info(fmtfn_t to, void *to_arg)
 {
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
-	erts_smp_rwmtx_rlock(&erts_node_table_rwmtx);
+	erts_rwmtx_rlock(&erts_node_table_rwmtx);
     hash_info(to, to_arg, &erts_node_table);
     if (lock)
-	erts_smp_rwmtx_runlock(&erts_node_table_rwmtx);
+	erts_rwmtx_runlock(&erts_node_table_rwmtx);
 }
 
 
@@ -583,26 +806,26 @@ ErlNode *erts_find_or_insert_node(Eterm sysname, Uint32 creation)
     ne.sysname = sysname;
     ne.creation = creation;
 
-    erts_smp_rwmtx_rlock(&erts_node_table_rwmtx);
+    erts_rwmtx_rlock(&erts_node_table_rwmtx);
     res = hash_get(&erts_node_table, (void *) &ne);
     if (res && res != erts_this_node) {
-	erts_aint_t refc = erts_smp_refc_inctest(&res->refc, 0);
+	erts_aint_t refc = erts_refc_inctest(&res->refc, 0);
 	if (refc < 2) /* New or pending delete */
-	    erts_smp_refc_inc(&res->refc, 1);
+	    erts_refc_inc(&res->refc, 1);
     }
-    erts_smp_rwmtx_runlock(&erts_node_table_rwmtx);
+    erts_rwmtx_runlock(&erts_node_table_rwmtx);
     if (res)
 	return res;
 
-    erts_smp_rwmtx_rwlock(&erts_node_table_rwmtx);
+    erts_rwmtx_rwlock(&erts_node_table_rwmtx);
     res = hash_put(&erts_node_table, (void *) &ne);
     ASSERT(res);
     if (res != erts_this_node) {
-	erts_aint_t refc = erts_smp_refc_inctest(&res->refc, 0);
+	erts_aint_t refc = erts_refc_inctest(&res->refc, 0);
 	if (refc < 2) /* New or pending delete */
-	    erts_smp_refc_inc(&res->refc, 1);
+	    erts_refc_inc(&res->refc, 1);
     }
-    erts_smp_rwmtx_rwunlock(&erts_node_table_rwmtx);
+    erts_rwmtx_rwunlock(&erts_node_table_rwmtx);
     return res;
 }
 
@@ -611,7 +834,7 @@ static void try_delete_node(void *venp)
     ErlNode *enp = (ErlNode *) venp;
     erts_aint_t refc;
 
-    erts_smp_rwmtx_rwlock(&erts_node_table_rwmtx);
+    erts_rwmtx_rwlock(&erts_node_table_rwmtx);
     /*
      * Another thread might have looked up this node after we
      * decided to delete it (refc became zero). If so, the other
@@ -627,10 +850,10 @@ static void try_delete_node(void *venp)
      *
      * If refc > 0, the entry is in use. Keep the entry.
      */
-    refc = erts_smp_refc_dectest(&enp->refc, -1);
+    refc = erts_refc_dectest(&enp->refc, -1);
     if (refc == -1)
 	(void) hash_erase(&erts_node_table, (void *) enp);
-    erts_smp_rwmtx_rwunlock(&erts_node_table_rwmtx);
+    erts_rwmtx_rwunlock(&erts_node_table_rwmtx);
 
     if (refc == 0)
 	erts_schedule_delete_node(enp);
@@ -673,7 +896,7 @@ static void print_node(void *venp, void *vpndp)
 	erts_print(pndp->to, pndp->to_arg, " %d", enp->creation);
 #ifdef DEBUG
 	erts_print(pndp->to, pndp->to_arg, " (refc=%ld)",
-		   erts_smp_refc_read(&enp->refc, 0));
+		   erts_refc_read(&enp->refc, 0));
 #endif
 	pndp->no_sysname++;
     }
@@ -696,13 +919,13 @@ void erts_print_node_info(fmtfn_t to,
     pnd.no_total = 0;
 
     if (lock)
-	erts_smp_rwmtx_rlock(&erts_node_table_rwmtx);
+	erts_rwmtx_rlock(&erts_node_table_rwmtx);
     hash_foreach(&erts_node_table, print_node, (void *) &pnd);
     if (pnd.no_sysname != 0) {
 	erts_print(to, to_arg, "\n");
     }
     if (lock)
-	erts_smp_rwmtx_runlock(&erts_node_table_rwmtx);
+	erts_rwmtx_runlock(&erts_node_table_rwmtx);
 
     if(no_sysname)
 	*no_sysname = pnd.no_sysname;
@@ -715,20 +938,19 @@ void erts_print_node_info(fmtfn_t to,
 void
 erts_set_this_node(Eterm sysname, Uint creation)
 {
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_blocking());
-    ASSERT(erts_smp_refc_read(&erts_this_dist_entry->refc, 2));
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
+    ASSERT(2 <= de_refc_read(erts_this_dist_entry, 2));
 
-    if (erts_smp_refc_dectest(&erts_this_node->refc, 0) == 0)
+    if (erts_refc_dectest(&erts_this_node->refc, 0) == 0)
         try_delete_node(erts_this_node);
 
-    if (erts_smp_refc_dectest(&erts_this_dist_entry->refc, 0) == 0)
-        try_delete_dist_entry(erts_this_dist_entry);
+    erts_deref_dist_entry(erts_this_dist_entry);
 
     erts_this_node = NULL; /* to make sure refc is bumped for this node */
     erts_this_node = erts_find_or_insert_node(sysname, creation);
     erts_this_dist_entry = erts_this_node->dist_entry;
 
-    erts_smp_refc_inc(&erts_this_dist_entry->refc, 2);
+    erts_ref_dist_entry(erts_this_dist_entry);
 
     erts_this_node_sysname = erts_this_node_sysname_BUFFER;
     erts_snprintf(erts_this_node_sysname, sizeof(erts_this_node_sysname_BUFFER),
@@ -747,7 +969,7 @@ erts_delayed_node_table_gc(void)
 
 void erts_init_node_tables(int dd_sec)
 {
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
     HashFunctions f;
     ErlNode node_tmpl;
 
@@ -758,12 +980,12 @@ void erts_init_node_tables(int dd_sec)
 
     orig_node_tab_delete_delay = node_tab_delete_delay;
 
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&erts_node_table_rwmtx, &rwmtx_opt, "node_table", NIL,
+    erts_rwmtx_init_opt(&erts_node_table_rwmtx, &rwmtx_opt, "node_table", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
-    erts_smp_rwmtx_init_opt(&erts_dist_table_rwmtx, &rwmtx_opt, "dist_table", NIL,
+    erts_rwmtx_init_opt(&erts_dist_table_rwmtx, &rwmtx_opt, "dist_table", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
 
     f.hash       = (H_FUN)		dist_table_hash;
@@ -783,23 +1005,25 @@ void erts_init_node_tables(int dd_sec)
 
     erts_hidden_dist_entries				= NULL;
     erts_visible_dist_entries				= NULL;
+    erts_pending_dist_entries			        = NULL;
     erts_not_connected_dist_entries			= NULL;
     erts_no_of_hidden_dist_entries			= 0;
     erts_no_of_visible_dist_entries			= 0;
+    erts_no_of_pending_dist_entries			= 0;
     erts_no_of_not_connected_dist_entries		= 0;
 
     node_tmpl.sysname = am_Noname;
     node_tmpl.creation = 0;
     erts_this_node = hash_put(&erts_node_table, &node_tmpl);
      /* +1 for erts_this_node */
-    erts_smp_refc_init(&erts_this_node->refc, 1);
+    erts_refc_init(&erts_this_node->refc, 1);
 
     ASSERT(erts_this_node->dist_entry != NULL);
     erts_this_dist_entry = erts_this_node->dist_entry;
     /* +1 for erts_this_dist_entry */
-    /* +1 for erts_this_node->dist_entry */
-    erts_smp_refc_init(&erts_this_dist_entry->refc, 2);
+    erts_ref_dist_entry(erts_this_dist_entry);
 
+    ASSERT(2 == de_refc_read(erts_this_dist_entry, 2));
 
     erts_this_node_sysname = erts_this_node_sysname_BUFFER;
     erts_snprintf(erts_this_node_sysname, sizeof(erts_this_node_sysname_BUFFER),
@@ -808,18 +1032,16 @@ void erts_init_node_tables(int dd_sec)
     references_atoms_need_init = 1;
 }
 
-#ifdef ERTS_SMP
 #ifdef ERTS_ENABLE_LOCK_CHECK
 int erts_lc_is_de_rwlocked(DistEntry *dep)
 {
-    return erts_smp_lc_rwmtx_is_rwlocked(&dep->rwmtx);
+    return erts_lc_rwmtx_is_rwlocked(&dep->rwmtx);
 }
 int erts_lc_is_de_rlocked(DistEntry *dep)
 {
-    return erts_smp_lc_rwmtx_is_rlocked(&dep->rwmtx);
+    return erts_lc_rwmtx_is_rlocked(&dep->rwmtx);
 }
 #endif
-#endif
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
 
@@ -829,22 +1051,24 @@ static void erts_lcnt_enable_dist_lock_count(void *dep_raw, void *enable) {
     if(enable) {
         erts_lcnt_install_new_lock_info(&dep->rwmtx.lcnt, "dist_entry", dep->sysname, 
             ERTS_LOCK_TYPE_RWMUTEX | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
-        erts_lcnt_install_new_lock_info(&dep->lnk_mtx.lcnt, "dist_entry_links", dep->sysname,
-            ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
         erts_lcnt_install_new_lock_info(&dep->qlock.lcnt, "dist_entry_out_queue", dep->sysname,
             ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
+        if (dep->mld)
+            erts_lcnt_install_new_lock_info(&dep->mld->mtx.lcnt, "dist_entry_links", dep->sysname,
+                                            ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_DISTRIBUTION);
     } else {
         erts_lcnt_uninstall(&dep->rwmtx.lcnt);
-        erts_lcnt_uninstall(&dep->lnk_mtx.lcnt);
         erts_lcnt_uninstall(&dep->qlock.lcnt);
+        if (dep->mld)
+            erts_lcnt_uninstall(&dep->mld->mtx.lcnt);
     }
 }
 
 void erts_lcnt_update_distribution_locks(int enable) {
-    erts_smp_rwmtx_rlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_rlock(&erts_dist_table_rwmtx);
     hash_foreach(&erts_dist_table, erts_lcnt_enable_dist_lock_count,
         (void*)(UWord)enable);
-    erts_smp_rwmtx_runlock(&erts_dist_table_rwmtx);
+    erts_rwmtx_runlock(&erts_dist_table_rwmtx);
 }
 #endif
 
@@ -878,6 +1102,8 @@ static Eterm AM_node_references;
 static Eterm AM_system;
 static Eterm AM_timer;
 static Eterm AM_delayed_delete_timer;
+static Eterm AM_thread_progress_delete_timer;
+static Eterm AM_signal;
 
 static void setup_reference_table(void);
 static Eterm reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp);
@@ -899,6 +1125,7 @@ typedef struct node_referrer_ {
     int bin_ref;
     int timer_ref;
     int system_ref;
+    int signal_ref;
     Eterm id;
     Uint id_heap[ID_HEAP_SIZE];
     ErlOffHeap off_heap;
@@ -912,9 +1139,11 @@ typedef struct {
 typedef struct dist_referrer_ {
     struct dist_referrer_ *next;
     int heap_ref;
+    int ets_ref;
     int node_ref;
     int ctrl_ref;
     int system_ref;
+    int signal_ref;
     Eterm id;
     Uint creation;
     Uint id_heap[ID_HEAP_SIZE];
@@ -946,8 +1175,8 @@ erts_get_node_and_dist_references(struct process *proc)
     Uint *endp;
 #endif
 
-    erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
-    erts_smp_thr_progress_block();
+    erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_block();
     /* No need to lock any thing since we are alone... */
 
     if (references_atoms_need_init) {
@@ -967,6 +1196,8 @@ erts_get_node_and_dist_references(struct process *proc)
 	INIT_AM(timer);
 	INIT_AM(system);
 	INIT_AM(delayed_delete_timer);
+        INIT_AM(thread_progress_delete_timer);
+	INIT_AM(signal);
 	references_atoms_need_init = 0;
     }
 
@@ -989,8 +1220,8 @@ erts_get_node_and_dist_references(struct process *proc)
 
     delete_reference_table();
 
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(proc, ERTS_PROC_LOCK_MAIN);
     return res;
 }
 
@@ -1003,6 +1234,7 @@ erts_get_node_and_dist_references(struct process *proc)
 #define MONITOR_REF 7
 #define TIMER_REF 8
 #define SYSTEM_REF 9
+#define SIGNAL_REF 10
 
 #define INC_TAB_SZ 10
 
@@ -1029,20 +1261,24 @@ insert_dist_referrer(ReferredDist *referred_dist,
 	else {
 	    Uint *hp = &drp->id_heap[0];
 	    ASSERT(is_tuple(id));
-	    drp->id = copy_struct(id, size_object(id), &hp, NULL);
+            drp->id = copy_struct(id, size_object(id), &hp, NULL);
 	}
 	drp->creation = creation;
 	drp->heap_ref = 0;
+        drp->ets_ref = 0;
 	drp->node_ref = 0;
 	drp->ctrl_ref = 0;
 	drp->system_ref = 0;
+        drp->signal_ref = 0;
     }
 
     switch (type) {
     case NODE_REF:	drp->node_ref++;	break;
     case CTRL_REF:	drp->ctrl_ref++;	break;
     case HEAP_REF:	drp->heap_ref++;	break;
+    case ETS_REF:	drp->ets_ref++;	        break;
     case SYSTEM_REF:	drp->system_ref++;	break;
+    case SIGNAL_REF:	drp->signal_ref++;	break;
     default:		ASSERT(0);
     }
 }
@@ -1096,6 +1332,7 @@ insert_node_referrer(ReferredNode *referred_node, int type, Eterm id)
 	nrp->bin_ref = 0;
 	nrp->timer_ref = 0;
 	nrp->system_ref = 0;
+	nrp->signal_ref = 0;
     }
 
     switch (type) {
@@ -1106,6 +1343,7 @@ insert_node_referrer(ReferredNode *referred_node, int type, Eterm id)
     case MONITOR_REF:   nrp->monitor_ref++;     break;
     case TIMER_REF:	nrp->timer_ref++;	break;
     case SYSTEM_REF:	nrp->system_ref++;	break;
+    case SIGNAL_REF:	nrp->signal_ref++;	break;
     default:		ASSERT(0);
     }
 }
@@ -1150,6 +1388,15 @@ insert_offheap2(ErlOffHeap *oh, void *arg)
     insert_offheap(oh, a->type, a->id);
 }
 
+#define ErtsIsDistEntryBinary(Bin)                                      \
+    (((Bin)->intern.flags & BIN_FLAG_MAGIC)                             \
+     && ERTS_MAGIC_BIN_DESTRUCTOR((Bin)) == erts_dist_entry_destructor)
+
+#define IsSendCtxBinary(Bin)                                            \
+    (((Bin)->intern.flags & BIN_FLAG_MAGIC)                             \
+     && ERTS_MAGIC_BIN_DESTRUCTOR((Bin)) == erts_dsend_context_dtor)
+
+
 static void
 insert_offheap(ErlOffHeap *oh, int type, Eterm id)
 {
@@ -1160,7 +1407,10 @@ insert_offheap(ErlOffHeap *oh, int type, Eterm id)
     for (u.hdr = oh->first; u.hdr; u.hdr = u.hdr->next) {
 	switch (thing_subtag(u.hdr->thing_word)) {
 	case REF_SUBTAG:
-	    if(IsMatchProgBinary(u.mref->mb)) {
+            if (ErtsIsDistEntryBinary(u.mref->mb))
+                insert_dist_entry(ErtsBin2DistEntry(u.mref->mb),
+                                  type, id, 0);
+	    else if(IsMatchProgBinary(u.mref->mb)) {
 		InsertedBin *ib;
 		int insert_bin = 1;
 		for (ib = inserted_bins; ib; ib = ib->next)
@@ -1183,7 +1433,12 @@ insert_offheap(ErlOffHeap *oh, int type, Eterm id)
 		    inserted_bins = nib;
 		    UnUseTmpHeapNoproc(BIG_UINT_HEAP_SIZE);
 		}
-	    }		
+	    }
+            else if (IsSendCtxBinary(u.mref->mb)) {
+                ErtsSendContext* ctx = ERTS_MAGIC_BIN_DATA(u.mref->mb);
+                if (ctx->deref_dep)
+                    insert_dist_entry(ctx->dep, type, id, 0);
+            }
 	    break;
 	case REFC_BINARY_SUBTAG:
 	case FUN_SUBTAG:
@@ -1196,49 +1451,145 @@ insert_offheap(ErlOffHeap *oh, int type, Eterm id)
     }
 }
 
-static void doit_insert_monitor(ErtsMonitor *monitor, void *p)
+static void insert_monitor_data(ErtsMonitor *mon, int type, Eterm id)
 {
-    Eterm *idp = p;
-    if(monitor->type != MON_NIF_TARGET && is_external(monitor->u.pid))
-	insert_node(external_thing_ptr(monitor->u.pid)->node, MONITOR_REF, *idp);
-    if(is_external(monitor->ref))
-	insert_node(external_thing_ptr(monitor->ref)->node, MONITOR_REF, *idp);
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+    if ((mdp->origin.flags & (ERTS_ML_FLG_DBG_VISITED
+                              | ERTS_ML_FLG_EXTENDED)) == ERTS_ML_FLG_EXTENDED) {
+        if (mon->type != ERTS_MON_TYPE_NODE) {
+            ErtsMonitorDataExtended *mdep = (ErtsMonitorDataExtended *) mdp;
+            ASSERT(mon->flags & ERTS_ML_FLG_EXTENDED);
+            if (mdep->uptr.ohhp) {
+                ErlOffHeap oh;
+                ERTS_INIT_OFF_HEAP(&oh);
+                oh.first = mdep->uptr.ohhp;
+                insert_offheap(&oh, type, id);
+            }
+        }
+    }
+    mdp->origin.flags |= ERTS_ML_FLG_DBG_VISITED;
 }
 
-static void doit_insert_link(ErtsLink *lnk, void *p)
+static void insert_monitor(ErtsMonitor *mon, void *idp)
 {
-    Eterm *idp = p;
-    if(is_external(lnk->pid))
-	insert_node(external_thing_ptr(lnk->pid)->node, LINK_REF, 
-		    *idp);
+    Eterm id = *((Eterm *) idp);
+    insert_monitor_data(mon, MONITOR_REF, id);
 }
 
+static void clear_visited_monitor(ErtsMonitor *mon, void *p)
+{
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+    mdp->origin.flags &= ~ERTS_ML_FLG_DBG_VISITED;
+}
 
 static void
-insert_monitors(ErtsMonitor *monitors, Eterm id)
+insert_p_monitors(ErtsPTabElementCommon *p)
 {
-    erts_doforall_monitors(monitors,&doit_insert_monitor,&id);
+    Eterm id = p->id;
+    erts_monitor_tree_foreach(p->u.alive.monitors,
+                              insert_monitor,
+                              (void *) &id);
+    erts_monitor_list_foreach(p->u.alive.lt_monitors,
+                              insert_monitor,
+                              (void *) &id);
 }
 
 static void
-insert_links(ErtsLink *lnk, Eterm id)
+insert_dist_monitors(DistEntry *dep)
 {
-    erts_doforall_links(lnk,&doit_insert_link,&id);
+    if (dep->mld) {
+        erts_monitor_list_foreach(dep->mld->monitors,
+                                  insert_monitor,
+                                  (void *) &dep->sysname);
+        erts_monitor_tree_foreach(dep->mld->orig_name_monitors,
+                                  insert_monitor,
+                                  (void *) &dep->sysname);
+    }
 }
 
-static void doit_insert_link2(ErtsLink *lnk, void *p)
+static void
+clear_visited_p_monitors(ErtsPTabElementCommon *p)
 {
-    Eterm *idp = p;
-    if(is_external(lnk->pid))
-	insert_node(external_thing_ptr(lnk->pid)->node, LINK_REF, 
-		    *idp);
-    insert_links(ERTS_LINK_ROOT(lnk), *idp);
+    erts_monitor_tree_foreach(p->u.alive.monitors,
+                              clear_visited_monitor,
+                              NULL);
+    erts_monitor_list_foreach(p->u.alive.lt_monitors,
+                              clear_visited_monitor,
+                              NULL);
 }
 
 static void
-insert_links2(ErtsLink *lnk, Eterm id)
+clear_visited_dist_monitors(DistEntry *dep)
 {
-    erts_doforall_links(lnk,&doit_insert_link2,&id);
+    if (dep->mld) {
+        erts_monitor_list_foreach(dep->mld->monitors,
+                                  clear_visited_monitor,
+                                  NULL);
+        erts_monitor_tree_foreach(dep->mld->orig_name_monitors,
+                                  clear_visited_monitor,
+                                  NULL);
+    }
+}
+
+static void insert_link_data(ErtsLink *lnk, int type, Eterm id)
+{
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+    if ((ldp->a.flags & (ERTS_ML_FLG_DBG_VISITED
+                         | ERTS_ML_FLG_EXTENDED)) == ERTS_ML_FLG_EXTENDED) {
+        ErtsLinkDataExtended *ldep = (ErtsLinkDataExtended *) ldp;
+        if (ldep->ohhp) {
+            ErlOffHeap oh;
+            ERTS_INIT_OFF_HEAP(&oh);
+            oh.first = ldep->ohhp;
+            insert_offheap(&oh, type, id);
+        }
+    }
+    ldp->a.flags |= ERTS_ML_FLG_DBG_VISITED;
+}
+
+static void insert_link(ErtsLink *lnk, void *idp)
+{
+    Eterm id = *((Eterm *) idp);
+    insert_link_data(lnk, LINK_REF, id);
+}
+
+static void clear_visited_link(ErtsLink *lnk, void *p)
+{
+    ErtsLinkData *ldp = erts_link_to_data(lnk);
+    ldp->a.flags &= ~ERTS_ML_FLG_DBG_VISITED;
+}
+
+static void
+insert_p_links(ErtsPTabElementCommon *p)
+{
+    Eterm id = p->id;
+    erts_link_tree_foreach(p->u.alive.links, insert_link, (void *) &id);
+}
+
+static void
+insert_dist_links(DistEntry *dep)
+{
+    if (dep->mld)
+        erts_link_list_foreach(dep->mld->links,
+                               insert_link,
+                               (void *) &dep->sysname);
+}
+
+static void
+clear_visited_p_links(ErtsPTabElementCommon *p)
+{
+    erts_link_tree_foreach(p->u.alive.links,
+                           clear_visited_link,
+                           NULL);
+}
+
+static void
+clear_visited_dist_links(DistEntry *dep)
+{
+    if (dep->mld)
+        erts_link_list_foreach(dep->mld->links,
+                               clear_visited_link,
+                               NULL);
 }
 
 static void
@@ -1292,25 +1643,32 @@ init_referred_dist(void *dist, void *unused)
     no_referred_dists++;
 }
 
-#ifdef ERTS_SMP
 static void
 insert_sys_msg(Eterm from, Eterm to, Eterm msg, ErlHeapFragment *bp)
 {
     insert_offheap(&bp->off_heap, HEAP_REF, to);
 }
-#endif
 
 static void
 insert_delayed_delete_node(void *state,
 			   ErtsMonotonicTime timeout_pos,
 			   void *vnp)
 {
-    DeclareTmpHeapNoproc(heap,3);
-    UseTmpHeapNoproc(3);
+    Eterm heap[3];
     insert_node((ErlNode *) vnp,
 		SYSTEM_REF,
 		TUPLE2(&heap[0], AM_system, AM_delayed_delete_timer));
-    UnUseTmpHeapNoproc(3);
+}
+
+static void
+insert_thr_prgr_delete_dist_entry(void *arg, ErtsThrPrgrVal thr_prgr, void *vdep)
+{
+    DistEntry *dep = vdep;
+    Eterm heap[3];
+    insert_dist_entry(dep,
+		      SYSTEM_REF,
+		      TUPLE2(&heap[0], AM_system, AM_thread_progress_delete_timer),
+		      0);
 }
 
 static void
@@ -1318,13 +1676,66 @@ insert_delayed_delete_dist_entry(void *state,
 				 ErtsMonotonicTime timeout_pos,
 				 void *vdep)
 {
-    DeclareTmpHeapNoproc(heap,3);
-    UseTmpHeapNoproc(3);
-    insert_dist_entry((DistEntry *) vdep,
+    DistEntry *dep = vdep;
+    Eterm heap[3];
+    insert_dist_entry(dep,
 		      SYSTEM_REF,
 		      TUPLE2(&heap[0], AM_system, AM_delayed_delete_timer),
 		      0);
-    UnUseTmpHeapNoproc(3);
+}
+
+static void
+insert_message(ErtsMessage *msg, int type, Process *proc)
+{
+    ErlHeapFragment *heap_frag = NULL;
+
+    ASSERT(ERTS_SIG_IS_MSG(msg));
+    if (msg->data.attached) {
+        if (msg->data.attached == ERTS_MSG_COMBINED_HFRAG)
+            heap_frag = &msg->hfrag;
+        else if (ERTS_SIG_IS_INTERNAL_MSG(msg))
+            heap_frag = msg->data.heap_frag;
+        else {
+            if (msg->data.dist_ext->dep)
+                insert_dist_entry(msg->data.dist_ext->dep,
+                                  type, proc->common.id, 0);
+            if (is_not_nil(ERL_MESSAGE_TOKEN(msg)))
+                heap_frag = erts_dist_ext_trailer(msg->data.dist_ext);
+        }
+    }
+    while (heap_frag) {
+        insert_offheap(&(heap_frag->off_heap),
+                       type,
+                       proc->common.id);
+        heap_frag = heap_frag->next;
+    }
+}
+
+static void
+insert_sig_msg(ErtsMessage *msg, void *arg)
+{
+    insert_message(msg, SIGNAL_REF, (Process *) arg);
+}
+
+static void
+insert_sig_offheap(ErlOffHeap *ohp, void *arg)
+{
+    Process *proc = arg;
+    insert_offheap(ohp, SIGNAL_REF, proc->common.id);
+}
+
+static void
+insert_sig_monitor(ErtsMonitor *mon, void *arg)
+{
+    Process *proc = arg;
+    insert_monitor_data(mon, SIGNAL_REF, proc->common.id);
+}
+
+static void
+insert_sig_link(ErtsLink *lnk, void *arg)
+{
+    Process *proc = arg;
+    insert_link_data(lnk, SIGNAL_REF, proc->common.id);
 }
 
 static void
@@ -1358,9 +1769,12 @@ setup_reference_table(void)
     erts_debug_callback_timer_foreach(try_delete_node,
 				      insert_delayed_delete_node,
 				      NULL);
-    erts_debug_callback_timer_foreach(try_delete_dist_entry,
+    erts_debug_callback_timer_foreach(prepare_try_delete_dist_entry,
 				      insert_delayed_delete_dist_entry,
 				      NULL);
+    erts_debug_later_op_foreach(start_timer_delete_dist_entry,
+                                insert_thr_prgr_delete_dist_entry,
+                                NULL);
 
     UseTmpHeapNoproc(3);
     insert_node(erts_this_node,
@@ -1379,12 +1793,7 @@ setup_reference_table(void)
 	Process *proc = erts_pix2proc(i);
 	if (proc) {
 	    int mli;
-	    ErtsMessage *msg_list[] = {
-		proc->msg.first,
-#ifdef ERTS_SMP
-		proc->msg_inq.first,
-#endif
-		proc->msg_frag};
+	    ErtsMessage *msg_list[] = {proc->msg_frag};
 
 	    /* Insert Heap */
 	    insert_offheap(&(proc->off_heap),
@@ -1399,40 +1808,36 @@ setup_reference_table(void)
 	    /* Insert msg buffers */
 	    for (mli = 0; mli < sizeof(msg_list)/sizeof(msg_list[0]); mli++) {
 		ErtsMessage *msg;
-		for (msg = msg_list[mli]; msg; msg = msg->next) {
-		    ErlHeapFragment *heap_frag = NULL;
-		    if (msg->data.attached) {
-			if (msg->data.attached == ERTS_MSG_COMBINED_HFRAG)
-			    heap_frag = &msg->hfrag;
-			else if (is_value(ERL_MESSAGE_TERM(msg)))
-			    heap_frag = msg->data.heap_frag;
-			else {
-			    if (msg->data.dist_ext->dep)
-				insert_dist_entry(msg->data.dist_ext->dep,
-						  HEAP_REF, proc->common.id, 0);
-			    if (is_not_nil(ERL_MESSAGE_TOKEN(msg)))
-				heap_frag = erts_dist_ext_trailer(msg->data.dist_ext);
-			}
-		    }
-		    while (heap_frag) {
-			insert_offheap(&(heap_frag->off_heap),
-				       HEAP_REF,
-				       proc->common.id);
-			heap_frag = heap_frag->next;
-		    }
-		}
+		for (msg = msg_list[mli]; msg; msg = msg->next)
+                    insert_message(msg, HEAP_REF, proc);
 	    }
+
+            /* Insert signal queue */
+            erts_proc_sig_debug_foreach_sig(proc,
+                                            insert_sig_msg,
+                                            insert_sig_offheap,
+                                            insert_sig_monitor,
+                                            insert_sig_link,
+                                            (void *) proc);
+
 	    /* Insert links */
-	    if (ERTS_P_LINKS(proc))
-		insert_links(ERTS_P_LINKS(proc), proc->common.id);
-	    if (ERTS_P_MONITORS(proc))
-		insert_monitors(ERTS_P_MONITORS(proc), proc->common.id);
+            insert_p_links(&proc->common);
+
+            /* Insert monitors */
+            insert_p_monitors(&proc->common);
+
+            {
+                DistEntry *dep = ERTS_PROC_GET_DIST_ENTRY(proc);
+                if (dep)
+                    insert_dist_entry(dep,
+                                      CTRL_REF,
+                                      proc->common.id,
+                                      0);
+            }
 	}
     }
     
-#ifdef ERTS_SMP
     erts_foreach_sys_msg_in_q(insert_sys_msg);
-#endif
 
     /* Insert all ports */
     max = erts_ptab_max(&erts_port);
@@ -1450,11 +1855,9 @@ setup_reference_table(void)
 	    continue;
 
 	/* Insert links */
-	if (ERTS_P_LINKS(prt))
-	    insert_links(ERTS_P_LINKS(prt), prt->common.id);
+        insert_p_links(&prt->common);
 	/* Insert monitors */
-	if (ERTS_P_MONITORS(prt))
-	    insert_monitors(ERTS_P_MONITORS(prt), prt->common.id);
+        insert_p_monitors(&prt->common);
 	/* Insert port data */
 	ohp = erts_port_data_offheap(prt);
 	if (ohp)
@@ -1503,32 +1906,25 @@ setup_reference_table(void)
     /* Insert all dist links */
 
     for(dep = erts_visible_dist_entries; dep; dep = dep->next) {
-	if(dep->nlinks)
-	    insert_links2(dep->nlinks, dep->sysname);
-	if(dep->node_links)
-	    insert_links(dep->node_links, dep->sysname);
-	if(dep->monitors)
-	    insert_monitors(dep->monitors, dep->sysname);
+        insert_dist_links(dep);
+        insert_dist_monitors(dep);
     }
 
     for(dep = erts_hidden_dist_entries; dep; dep = dep->next) {
-	if(dep->nlinks)
-	    insert_links2(dep->nlinks, dep->sysname);
-	if(dep->node_links)
-	    insert_links(dep->node_links, dep->sysname);
-	if(dep->monitors)
-	    insert_monitors(dep->monitors, dep->sysname);
+        insert_dist_links(dep);
+        insert_dist_monitors(dep);
+    }
+
+    for(dep = erts_pending_dist_entries; dep; dep = dep->next) {
+        insert_dist_links(dep);
+        insert_dist_monitors(dep);
     }
 
     /* Not connected dist entries should not have any links,
        but inspect them anyway */
     for(dep = erts_not_connected_dist_entries; dep; dep = dep->next) {
-	if(dep->nlinks)
-	    insert_links2(dep->nlinks, dep->sysname);
-	if(dep->node_links)
-	    insert_links(dep->node_links, dep->sysname);
-	if(dep->monitors)
-	    insert_monitors(dep->monitors, dep->sysname);
+        insert_dist_links(dep);
+        insert_dist_monitors(dep);
     }
 
     /* Insert all ets tables */
@@ -1613,6 +2009,10 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 		tup = MK_2TUP(AM_system, MK_UINT(nrp->system_ref));
 		nrl = MK_CONS(tup, nrl);
 	    }
+	    if(nrp->signal_ref) {
+		tup = MK_2TUP(AM_signal, MK_UINT(nrp->signal_ref));
+		nrl = MK_CONS(tup, nrl);
+	    }
 
 	    nrid = nrp->id;
 	    if (!IS_CONST(nrp->id)) {
@@ -1636,24 +2036,25 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 	    }
 	    else if(is_internal_port(nrid)) {
 		ASSERT(!nrp->heap_ref && !nrp->ets_ref && !nrp->bin_ref
-		       && !nrp->timer_ref && !nrp->system_ref);
+		       && !nrp->timer_ref && !nrp->system_ref && !nrp->signal_ref);
 		tup = MK_2TUP(AM_port, nrid);
 	    }
 	    else if(nrp->ets_ref) {
 		ASSERT(!nrp->heap_ref && !nrp->link_ref && 
 		       !nrp->monitor_ref && !nrp->bin_ref
-		       && !nrp->timer_ref && !nrp->system_ref);
+		       && !nrp->timer_ref && !nrp->system_ref && !nrp->signal_ref);
 		tup = MK_2TUP(AM_ets, nrid);
 	    }
 	    else if(nrp->bin_ref) {
 		ASSERT(is_small(nrid) || is_big(nrid));
 		ASSERT(!nrp->heap_ref && !nrp->ets_ref && !nrp->link_ref && 
 		       !nrp->monitor_ref && !nrp->timer_ref
-		       && !nrp->system_ref);
+		       && !nrp->system_ref && !nrp->signal_ref);
 		tup = MK_2TUP(AM_match_spec, nrid);
 	    }
 	    else  {
-		ASSERT(!nrp->heap_ref && !nrp->ets_ref && !nrp->bin_ref);
+		ASSERT(!nrp->heap_ref && !nrp->ets_ref && !nrp->bin_ref
+                       && !nrp->signal_ref);
 		ASSERT(is_atom(nrid));
 		tup = MK_2TUP(AM_dist, nrid);
 	    }
@@ -1666,7 +2067,7 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 
 	tup = MK_2TUP(referred_nodes[i].node->sysname,
 		      MK_UINT(referred_nodes[i].node->creation));
-	tup = MK_3TUP(tup, MK_UINT(erts_smp_refc_read(&referred_nodes[i].node->refc, 0)), nril);
+	tup = MK_3TUP(tup, MK_UINT(erts_refc_read(&referred_nodes[i].node->refc, 0)), nril);
 	nl = MK_CONS(tup, nl);
     }
 
@@ -1689,29 +2090,44 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 		tup = MK_2TUP(AM_heap, MK_UINT(drp->heap_ref));
 		drl = MK_CONS(tup, drl);
 	    }
+            if(drp->ets_ref) {
+                tup = MK_2TUP(AM_ets, MK_UINT(drp->ets_ref));
+                drl = MK_CONS(tup, drl);
+            }
 	    if(drp->system_ref) {
 		tup = MK_2TUP(AM_system, MK_UINT(drp->system_ref));
 		drl = MK_CONS(tup, drl);
 	    }
+	    if(drp->signal_ref) {
+		tup = MK_2TUP(AM_signal, MK_UINT(drp->signal_ref));
+		drl = MK_CONS(tup, drl);
+	    }
 
 	    if (is_internal_pid(drp->id)) {
 		ASSERT(!drp->node_ref);
 		tup = MK_2TUP(AM_process, drp->id);
 	    }
 	    else if(is_internal_port(drp->id)) {
-		ASSERT(drp->ctrl_ref && !drp->node_ref);
+		ASSERT(drp->ctrl_ref && !drp->node_ref && !drp->signal_ref);
 		tup = MK_2TUP(AM_port, drp->id);
 	    }
 	    else if (is_tuple(drp->id)) {
 		Eterm *t;
 		ASSERT(drp->system_ref && !drp->node_ref
-		       && !drp->ctrl_ref && !drp->heap_ref);
+		       && !drp->ctrl_ref && !drp->heap_ref && !drp->ets_ref
+                       && !drp->signal_ref);
 		t = tuple_val(drp->id);
 		ASSERT(2 == arityval(t[0]));
 		tup = MK_2TUP(t[1], t[2]);
 	    }
+	    else if (drp->ets_ref) {
+                ASSERT(!drp->heap_ref && !drp->node_ref &&
+                       !drp->ctrl_ref && !drp->system_ref
+                       && !drp->signal_ref);
+                tup = MK_2TUP(AM_ets, drp->id);
+            }
 	    else {
-		ASSERT(!drp->ctrl_ref && drp->node_ref);
+		ASSERT(!drp->ctrl_ref && drp->node_ref && !drp->signal_ref);
 		ASSERT(is_atom(drp->id));
 		tup = MK_2TUP(drp->id, MK_UINT(drp->creation));
 		tup = MK_2TUP(AM_node, tup);
@@ -1727,7 +2143,7 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 
 	/* DistList = [{Dist, Refc, ReferenceIdList}] */
 	tup = MK_3TUP(referred_dists[i].dist->sysname,
-		      MK_UINT(erts_smp_refc_read(&referred_dists[i].dist->refc, 0)),
+		      MK_UINT(de_refc_read(referred_dists[i].dist, 0)),
 		      dril);
 	dl = MK_CONS(tup, dl);
     }
@@ -1746,10 +2162,21 @@ reference_table_term(Uint **hpp, ErlOffHeap *ohp, Uint *szp)
 
 }
 
+static void noop_sig_msg(ErtsMessage *msg, void *arg)
+{
+
+}
+
+static void noop_sig_offheap(ErlOffHeap *oh, void *arg)
+{
+
+}
+
 static void
 delete_reference_table(void)
 {
-    Uint i;
+    DistEntry *dep;
+    int i, max;
     for(i = 0; i < no_referred_nodes; i++) {
 	NodeReferrer *nrp;
 	NodeReferrer *tnrp;
@@ -1781,17 +2208,71 @@ delete_reference_table(void)
 	inserted_bins = inserted_bins->next;
 	erts_free(ERTS_ALC_T_NC_TMP, (void *)ib);
     }
+
+    /* Cleanup... */
+
+    max = erts_ptab_max(&erts_proc);
+    for (i = 0; i < max; i++) {
+	Process *proc = erts_pix2proc(i);
+	if (proc) {
+            clear_visited_p_links(&proc->common);
+            clear_visited_p_monitors(&proc->common);
+            erts_proc_sig_debug_foreach_sig(proc,
+                                            noop_sig_msg,
+                                            noop_sig_offheap,
+                                            clear_visited_monitor,
+                                            clear_visited_link,
+                                            (void *) proc);
+        }
+    }
+
+    max = erts_ptab_max(&erts_port);
+    for (i = 0; i < max; i++) {
+	erts_aint32_t state;
+	Port *prt;
+
+	prt = erts_pix2port(i);
+	if (!prt)
+	    continue;
+
+	state = erts_atomic32_read_nob(&prt->state);
+	if (state & ERTS_PORT_SFLGS_DEAD)
+	    continue;
+
+        clear_visited_p_links(&prt->common);
+        clear_visited_p_monitors(&prt->common);
+    }
+
+    for(dep = erts_visible_dist_entries; dep; dep = dep->next) {
+        clear_visited_dist_links(dep);
+        clear_visited_dist_monitors(dep);
+    }
+
+    for(dep = erts_hidden_dist_entries; dep; dep = dep->next) {
+        clear_visited_dist_links(dep);
+        clear_visited_dist_monitors(dep);
+    }
+
+    for(dep = erts_pending_dist_entries; dep; dep = dep->next) {
+        clear_visited_dist_links(dep);
+        clear_visited_dist_monitors(dep);
+    }
+
+    for(dep = erts_not_connected_dist_entries; dep; dep = dep->next) {
+        clear_visited_dist_links(dep);
+        clear_visited_dist_monitors(dep);
+    }
 }
 
 void
 erts_debug_test_node_tab_delayed_delete(Sint64 millisecs)
 {
-    erts_smp_thr_progress_block();
+    erts_thr_progress_block();
 
     if (millisecs < 0)
 	node_tab_delete_delay = orig_node_tab_delete_delay;
     else
 	node_tab_delete_delay = millisecs;
 
-    erts_smp_thr_progress_unblock();
+    erts_thr_progress_unblock();
 }
diff --git a/erts/emulator/beam/erl_node_tables.h b/erts/emulator/beam/erl_node_tables.h
index 91bcb4fce1..9a792b10b1 100644
--- a/erts/emulator/beam/erl_node_tables.h
+++ b/erts/emulator/beam/erl_node_tables.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2001-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2001-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -18,7 +18,17 @@
  * %CopyrightEnd%
  */
 
-#ifndef ERL_NODE_TABLES_H__
+#ifndef ERL_NODE_TABLES_BASIC__
+#define ERL_NODE_TABLES_BASIC__
+
+typedef struct dist_entry_ DistEntry;
+typedef struct ErtsDistOutputBuf_ ErtsDistOutputBuf;
+void erts_ref_dist_entry(DistEntry *dep);
+void erts_deref_dist_entry(DistEntry *dep);
+
+#endif
+
+#if !defined(ERL_NODE_TABLES_BASIC_ONLY) && !defined(ERL_NODE_TABLES_H__)
 #define ERL_NODE_TABLES_H__
 
 /*
@@ -42,12 +52,14 @@
 #include "sys.h"
 #include "hash.h"
 #include "erl_alloc.h"
-#include "erl_process.h"
-#include "erl_monitors.h"
-#include "erl_smp.h"
 #define ERTS_PORT_TASK_ONLY_BASIC_TYPES__
 #include "erl_port_task.h"
 #undef ERTS_PORT_TASK_ONLY_BASIC_TYPES__
+#include "erl_process.h"
+#define ERTS_BINARY_TYPES_ONLY__
+#include "erl_binary.h"
+#undef ERTS_BINARY_TYPES_ONLY__
+#include "erl_monitor_link.h"
 
 #define ERTS_NODE_TAB_DELAY_GC_DEFAULT (60)
 #define ERTS_NODE_TAB_DELAY_GC_MAX (100*1000*1000)
@@ -55,17 +67,24 @@
  
 #define ERST_INTERNAL_CHANNEL_NO 0
 
-#define ERTS_DE_SFLG_CONNECTED			(((Uint32) 1) <<  0)
-#define ERTS_DE_SFLG_EXITING			(((Uint32) 1) <<  1)
-
-#define ERTS_DE_SFLGS_ALL			(ERTS_DE_SFLG_CONNECTED \
-						 | ERTS_DE_SFLG_EXITING)
+enum dist_entry_state {
+    ERTS_DE_STATE_IDLE,
+    ERTS_DE_STATE_PENDING,
+    ERTS_DE_STATE_CONNECTED,
+    ERTS_DE_STATE_EXITING
+};
 
-#define ERTS_DE_QFLG_BUSY			(((Uint32) 1) <<  0)
-#define ERTS_DE_QFLG_EXIT			(((Uint32) 1) <<  1)
+#define ERTS_DE_QFLG_BUSY			(((erts_aint32_t) 1) <<  0)
+#define ERTS_DE_QFLG_EXIT			(((erts_aint32_t) 1) <<  1)
+#define ERTS_DE_QFLG_REQ_INFO			(((erts_aint32_t) 1) <<  2)
+#define ERTS_DE_QFLG_PORT_CTRL                  (((erts_aint32_t) 1) <<  3)
+#define ERTS_DE_QFLG_PROC_CTRL                  (((erts_aint32_t) 1) <<  4)
 
 #define ERTS_DE_QFLGS_ALL			(ERTS_DE_QFLG_BUSY \
-						 | ERTS_DE_QFLG_EXIT)
+						 | ERTS_DE_QFLG_EXIT \
+                                                 | ERTS_DE_QFLG_REQ_INFO \
+                                                 | ERTS_DE_QFLG_PORT_CTRL \
+                                                 | ERTS_DE_QFLG_PROC_CTRL)
 
 #if defined(ARCH_64)
 #define ERTS_DIST_OUTPUT_BUF_DBG_PATTERN ((Uint) 0xf713f713f713f713UL)
@@ -73,14 +92,16 @@
 #define ERTS_DIST_OUTPUT_BUF_DBG_PATTERN ((Uint) 0xf713f713)
 #endif
 
-typedef struct ErtsDistOutputBuf_ ErtsDistOutputBuf;
 struct ErtsDistOutputBuf_ {
 #ifdef DEBUG
     Uint dbg_pattern;
+    byte *alloc_endp;
 #endif
     ErtsDistOutputBuf *next;
+    Uint hopefull_flags;
     byte *extp;
     byte *ext_endp;
+    byte *msg_start;
     byte data[1];
 };
 
@@ -101,55 +122,50 @@ struct ErtsProcList_;
  *   unlock mutexes with higher numbers before mutexes with higher numbers.
  */
 
-struct erl_link;
-
-typedef struct dist_entry_ {
+struct dist_entry_ {
     HashBucket hash_bucket;     /* Hash bucket */
     struct dist_entry_ *next;	/* Next entry in dist_table (not sorted) */
     struct dist_entry_ *prev;	/* Previous entry in dist_table (not sorted) */
-    erts_smp_refc_t refc;		/* Reference count */
 
-    erts_smp_rwmtx_t rwmtx;     /* Protects all fields below until lck_mtx. */
+    erts_rwmtx_t rwmtx;         /* Protects all fields below until lck_mtx. */
     Eterm sysname;		/* name@host atom for efficiency */
     Uint32 creation;		/* creation of connected node */
-    Eterm cid;			/* connection handler (pid or port), NIL == free */
+    erts_atomic_t input_handler; /* Input handler */
+    Eterm cid;			/* connection handler (pid or port),
+                                   NIL == free */
     Uint32 connection_id;	/* Connection id incremented on connect */
-    Uint32 status;		/* Slot status, like exiting reserved etc */
+    enum dist_entry_state state;
     Uint32 flags;		/* Distribution flags, like hidden, 
 				   atom cache etc. */
     unsigned long version;	/* Protocol version */
 
+    ErtsMonLnkDist *mld;        /* Monitors and links */
 
-    erts_smp_mtx_t lnk_mtx;     /* Protects node_links, nlinks, and
-				   monitors. */
-    ErtsLink *node_links;       /* In a dist entry, node links are kept 
-				   in a separate tree, while they are 
-				   colocted with the ordinary link tree
-				   for processes. It's not due to confusion,
-				   it's because the link tree for the dist 
-				   entry is in two levels, see erl_monitors.h 
-				*/
-    ErtsLink *nlinks;           /* Link tree with subtrees */
-    ErtsMonitor *monitors;      /* Monitor tree */
-
-    erts_smp_mtx_t qlock;       /* Protects qflgs and out_queue */
-    Uint32 qflgs;
-    Sint qsize;
+    erts_mtx_t qlock;           /* Protects qflgs and out_queue */
+    erts_atomic32_t qflgs;
+    erts_atomic_t qsize;
+    erts_atomic64_t in;
+    erts_atomic64_t out;
     ErtsDistOutputQueue out_queue;
     struct ErtsProcList_ *suspended;
 
+    ErtsDistOutputQueue tmp_out_queue;
     ErtsDistOutputQueue finalized_out_queue;
-    erts_smp_atomic_t dist_cmd_scheduled;
+    erts_atomic_t dist_cmd_scheduled;
     ErtsPortTaskHandle dist_cmd;
 
     Uint (*send)(Port *prt, ErtsDistOutputBuf *obuf);
 
     struct cache* cache;	/* The atom cache */
-} DistEntry;
+
+    ErtsThrPrgrLaterOp later_op;
+
+    struct transcode_context* transcode_ctx;
+};
 
 typedef struct erl_node_ {
   HashBucket hash_bucket;	/* Hash bucket */
-  erts_smp_refc_t refc;		/* Reference count */
+  erts_refc_t refc;		/* Reference count */
   Eterm	sysname;		/* name@host atom for efficiency */
   Uint32 creation;		/* Creation */
   DistEntry *dist_entry;	/* Corresponding dist entry */
@@ -158,14 +174,16 @@ typedef struct erl_node_ {
 
 extern Hash erts_dist_table;
 extern Hash erts_node_table;
-extern erts_smp_rwmtx_t erts_dist_table_rwmtx;
-extern erts_smp_rwmtx_t erts_node_table_rwmtx;
+extern erts_rwmtx_t erts_dist_table_rwmtx;
+extern erts_rwmtx_t erts_node_table_rwmtx;
 
 extern DistEntry *erts_hidden_dist_entries;
 extern DistEntry *erts_visible_dist_entries;
+extern DistEntry *erts_pending_dist_entries;
 extern DistEntry *erts_not_connected_dist_entries;
 extern Sint erts_no_of_hidden_dist_entries;
 extern Sint erts_no_of_visible_dist_entries;
+extern Sint erts_no_of_pending_dist_entries;
 extern Sint erts_no_of_not_connected_dist_entries;
 
 extern DistEntry *erts_this_dist_entry;
@@ -181,6 +199,7 @@ void erts_schedule_delete_dist_entry(DistEntry *);
 Uint erts_dist_table_size(void);
 void erts_dist_table_info(fmtfn_t, void *);
 void erts_set_dist_entry_not_connected(DistEntry *);
+void erts_set_dist_entry_pending(DistEntry *);
 void erts_set_dist_entry_connected(DistEntry *, Eterm, Uint);
 ErlNode *erts_find_or_insert_node(Eterm, Uint32);
 void erts_schedule_delete_node(ErlNode *);
@@ -190,80 +209,58 @@ void erts_init_node_tables(int);
 void erts_node_table_info(fmtfn_t, void *);
 void erts_print_node_info(fmtfn_t, void *, Eterm, int*, int*);
 Eterm erts_get_node_and_dist_references(struct process *);
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 int erts_lc_is_de_rwlocked(DistEntry *);
 int erts_lc_is_de_rlocked(DistEntry *);
 #endif
+int erts_dist_entry_destructor(Binary *bin);
+DistEntry *erts_dhandle_to_dist_entry(Eterm dhandle);
+Eterm erts_build_dhandle(Eterm **hpp, ErlOffHeap*, DistEntry*);
+Eterm erts_make_dhandle(Process *c_p, DistEntry *dep);
 
-#ifdef ERTS_ENABLE_LOCK_COUNT
-void erts_lcnt_update_distribution_locks(int enable);
-#endif
-
-ERTS_GLB_INLINE void erts_deref_dist_entry(DistEntry *dep);
 ERTS_GLB_INLINE void erts_deref_node_entry(ErlNode *np);
-ERTS_GLB_INLINE void erts_smp_de_rlock(DistEntry *dep);
-ERTS_GLB_INLINE void erts_smp_de_runlock(DistEntry *dep);
-ERTS_GLB_INLINE void erts_smp_de_rwlock(DistEntry *dep);
-ERTS_GLB_INLINE void erts_smp_de_rwunlock(DistEntry *dep);
-ERTS_GLB_INLINE void erts_smp_de_links_lock(DistEntry *dep);
-ERTS_GLB_INLINE void erts_smp_de_links_unlock(DistEntry *dep);
+ERTS_GLB_INLINE void erts_de_rlock(DistEntry *dep);
+ERTS_GLB_INLINE void erts_de_runlock(DistEntry *dep);
+ERTS_GLB_INLINE void erts_de_rwlock(DistEntry *dep);
+ERTS_GLB_INLINE void erts_de_rwunlock(DistEntry *dep);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
 ERTS_GLB_INLINE void
-erts_deref_dist_entry(DistEntry *dep)
-{
-    ASSERT(dep);
-    if (erts_smp_refc_dectest(&dep->refc, 0) == 0)
-	erts_schedule_delete_dist_entry(dep);
-}
-
-ERTS_GLB_INLINE void
 erts_deref_node_entry(ErlNode *np)
 {
     ASSERT(np);
-    if (erts_smp_refc_dectest(&np->refc, 0) == 0)
+    if (erts_refc_dectest(&np->refc, 0) == 0)
 	erts_schedule_delete_node(np);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_de_rlock(DistEntry *dep)
+erts_de_rlock(DistEntry *dep)
 {
-    erts_smp_rwmtx_rlock(&dep->rwmtx);
+    erts_rwmtx_rlock(&dep->rwmtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_de_runlock(DistEntry *dep)
+erts_de_runlock(DistEntry *dep)
 {
-    erts_smp_rwmtx_runlock(&dep->rwmtx);
+    erts_rwmtx_runlock(&dep->rwmtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_de_rwlock(DistEntry *dep)
+erts_de_rwlock(DistEntry *dep)
 {
-    erts_smp_rwmtx_rwlock(&dep->rwmtx);
+    erts_rwmtx_rwlock(&dep->rwmtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_de_rwunlock(DistEntry *dep)
+erts_de_rwunlock(DistEntry *dep)
 {
-    erts_smp_rwmtx_rwunlock(&dep->rwmtx);
-}
-
-ERTS_GLB_INLINE void
-erts_smp_de_links_lock(DistEntry *dep)
-{
-    erts_smp_mtx_lock(&dep->lnk_mtx);
-}
-
-ERTS_GLB_INLINE void
-erts_smp_de_links_unlock(DistEntry *dep)
-{
-    erts_smp_mtx_unlock(&dep->lnk_mtx);
+    erts_rwmtx_rwunlock(&dep->rwmtx);
 }
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
 void erts_debug_test_node_tab_delayed_delete(Sint64 millisecs);
+void erts_lcnt_update_distribution_locks(int enable);
 
 #endif
diff --git a/erts/emulator/beam/erl_port.h b/erts/emulator/beam/erl_port.h
index c30bbbca06..9b52b648e5 100644
--- a/erts/emulator/beam/erl_port.h
+++ b/erts/emulator/beam/erl_port.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2012-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2012-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -119,9 +119,7 @@ typedef struct {
     void *data[ERTS_PRTSD_SIZE];
 } ErtsPrtSD;
 
-#ifdef ERTS_SMP
 typedef struct ErtsXPortsList_ ErtsXPortsList;
-#endif
 
 /*
  * Port locking:
@@ -146,17 +144,12 @@ struct _erl_drv_port {
 
     ErtsPortTaskSched sched;
     ErtsPortTaskHandle timeout_task;
-#ifdef ERTS_SMP
     erts_mtx_t *lock;
     ErtsXPortsList *xports;
-    erts_smp_atomic_t run_queue;
-#else
-    erts_atomic32_t refc;
-    int cleanup;
-#endif
+    erts_atomic_t run_queue;
     erts_atomic_t connected;	/* A connected process */
     Eterm caller;		/* Current caller. */
-    erts_smp_atomic_t data;	/* Data associated with port. */
+    erts_atomic_t data;	/* Data associated with port. */
     Uint bytes_in;		/* Number of bytes read */
     Uint bytes_out;		/* Number of bytes written */
 
@@ -173,7 +166,7 @@ struct _erl_drv_port {
     int control_flags;		 /* Flags for port_control()  */
     ErlDrvPDL port_data_lock;
 
-    erts_smp_atomic_t psd;	 /* Port specific data */
+    erts_atomic_t psd;	 /* Port specific data */
     int reds; /* Only used while executing driver callbacks */
 
     struct {
@@ -203,31 +196,53 @@ struct erl_drv_port_data_lock {
     Port *prt;
 };
 
+ERTS_GLB_INLINE void erts_init_runq_port(Port *prt, ErtsRunQueue *runq);
+ERTS_GLB_INLINE void erts_set_runq_port(Port *prt, ErtsRunQueue *runq);
+ERTS_GLB_INLINE ErtsRunQueue *erts_get_runq_port(Port *prt);
 ERTS_GLB_INLINE ErtsRunQueue *erts_port_runq(Port *prt);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
+ERTS_GLB_INLINE void
+erts_init_runq_port(Port *prt, ErtsRunQueue *runq)
+{
+    if (!runq)
+        ERTS_INTERNAL_ERROR("Missing run-queue");
+    erts_atomic_init_nob(&prt->run_queue, (erts_aint_t) runq);
+}
+
+ERTS_GLB_INLINE void
+erts_set_runq_port(Port *prt, ErtsRunQueue *runq)
+{
+    if (!runq)
+        ERTS_INTERNAL_ERROR("Missing run-queue");
+    erts_atomic_set_nob(&prt->run_queue, (erts_aint_t) runq);
+}
+
+ERTS_GLB_INLINE ErtsRunQueue *
+erts_get_runq_port(Port *prt)
+{
+    ErtsRunQueue *runq;
+    runq = (ErtsRunQueue *) erts_atomic_read_nob(&prt->run_queue);
+    if (!runq)
+        ERTS_INTERNAL_ERROR("Missing run-queue");
+    return runq;
+}
+
+
 ERTS_GLB_INLINE ErtsRunQueue *
 erts_port_runq(Port *prt)
 {
-#ifdef ERTS_SMP
     ErtsRunQueue *rq1, *rq2;
-    rq1 = (ErtsRunQueue *) erts_smp_atomic_read_nob(&prt->run_queue);
-    if (!rq1)
-	return NULL;
+    rq1 = erts_get_runq_port(prt);
     while (1) {
-	erts_smp_runq_lock(rq1);
-	rq2 = (ErtsRunQueue *) erts_smp_atomic_read_nob(&prt->run_queue);
+	erts_runq_lock(rq1);
+	rq2 = erts_get_runq_port(prt);
 	if (rq1 == rq2)
 	    return rq1;
-	erts_smp_runq_unlock(rq1);
+	erts_runq_unlock(rq1);
 	rq1 = rq2;
-	if (!rq1)
-	    return NULL;
     }
-#else
-    return ERTS_RUNQ_IX(0);
-#endif
 }
 
 #endif
@@ -241,10 +256,10 @@ ERTS_GLB_INLINE void *erts_prtsd_set(Port *p, int ix, void *new);
 ERTS_GLB_INLINE void *
 erts_prtsd_get(Port *prt, int ix)
 {
-    ErtsPrtSD *psd = (ErtsPrtSD *) erts_smp_atomic_read_nob(&prt->psd);
+    ErtsPrtSD *psd = (ErtsPrtSD *) erts_atomic_read_nob(&prt->psd);
     if (!psd)
 	return NULL;
-    ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
+    ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
     return psd->data[ix];
 }
 
@@ -255,16 +270,14 @@ erts_prtsd_set(Port *prt, int ix, void *data)
     void *old;
     int i;
 
-    psd = (ErtsPrtSD *) erts_smp_atomic_read_nob(&prt->psd);
+    psd = (ErtsPrtSD *) erts_atomic_read_nob(&prt->psd);
 
     if (psd) {
-#ifdef ERTS_SMP
 #ifdef ETHR_ORDERED_READ_DEPEND
 	ETHR_MEMBAR(ETHR_LoadStore|ETHR_StoreStore);
 #else
 	ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore|ETHR_StoreStore);
 #endif
-#endif
 	old = psd->data[ix];
 	psd->data[ix] = data;
 	return old;
@@ -276,7 +289,7 @@ erts_prtsd_set(Port *prt, int ix, void *data)
     new_psd = erts_alloc(ERTS_ALC_T_PRTSD, sizeof(ErtsPrtSD));
     for (i = 0; i < ERTS_PRTSD_SIZE; i++)
 	new_psd->data[i] = NULL;
-    psd = (ErtsPrtSD *) erts_smp_atomic_cmpxchg_mb(&prt->psd,
+    psd = (ErtsPrtSD *) erts_atomic_cmpxchg_mb(&prt->psd,
 						   (erts_aint_t) new_psd,
 						   (erts_aint_t) NULL);
     if (psd)
@@ -360,15 +373,10 @@ Eterm erts_request_io_bytes(Process *c_p);
 
 void print_port_info(Port *, fmtfn_t, void *);
 void erts_port_free(Port *);
-#ifndef ERTS_SMP
-void erts_port_cleanup(Port *);
-#endif
-void erts_fire_port_monitor(Port *prt, Eterm ref);
-#ifdef ERTS_SMP
+void erts_fire_port_monitor(Port *prt, ErtsMonitor *tmon);
 int erts_port_handle_xports(Port *);
-#endif
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 int erts_lc_is_port_locked(Port *);
 #endif
 
@@ -377,9 +385,9 @@ ERTS_GLB_INLINE void erts_port_dec_refc(Port *prt);
 ERTS_GLB_INLINE void erts_port_add_refc(Port *prt, Sint32 add_refc);
 ERTS_GLB_INLINE Sint erts_port_read_refc(Port *prt);
 
-ERTS_GLB_INLINE int erts_smp_port_trylock(Port *prt);
-ERTS_GLB_INLINE void erts_smp_port_lock(Port *prt);
-ERTS_GLB_INLINE void erts_smp_port_unlock(Port *prt);
+ERTS_GLB_INLINE int erts_port_trylock(Port *prt);
+ERTS_GLB_INLINE void erts_port_lock(Port *prt);
+ERTS_GLB_INLINE void erts_port_unlock(Port *prt);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
@@ -408,35 +416,27 @@ ERTS_GLB_INLINE Sint erts_port_read_refc(Port *prt)
 }
 
 ERTS_GLB_INLINE int
-erts_smp_port_trylock(Port *prt)
+erts_port_trylock(Port *prt)
 {
-#ifdef ERTS_SMP
     /* *Need* to be a managed thread */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
     return erts_mtx_trylock(prt->lock);
-#else
-    return 0;
-#endif
 }
 
 ERTS_GLB_INLINE void
-erts_smp_port_lock(Port *prt)
+erts_port_lock(Port *prt)
 {
-#ifdef ERTS_SMP
     /* *Need* to be a managed thread */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
     erts_mtx_lock(prt->lock);
-#endif
 }
 
 ERTS_GLB_INLINE void
-erts_smp_port_unlock(Port *prt)
+erts_port_unlock(Port *prt)
 {
-#ifdef ERTS_SMP
     /* *Need* to be a managed thread */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
     erts_mtx_unlock(prt->lock);
-#endif
 }
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
@@ -467,9 +467,7 @@ extern const Port erts_invalid_port;
 int erts_is_port_ioq_empty(Port *);
 void erts_terminate_port(Port *);
 
-#ifdef ERTS_SMP
 Port *erts_de2port(DistEntry *, Process *, ErtsProcLocks);
-#endif
 
 ERTS_GLB_INLINE Port *erts_pix2port(int);
 ERTS_GLB_INLINE Port *erts_port_lookup_raw(Eterm);
@@ -477,11 +475,9 @@ ERTS_GLB_INLINE Port *erts_port_lookup(Eterm, Uint32);
 ERTS_GLB_INLINE Port*erts_id2port(Eterm id);
 ERTS_GLB_INLINE Port *erts_id2port_sflgs(Eterm, Process *, ErtsProcLocks, Uint32);
 ERTS_GLB_INLINE void erts_port_release(Port *);
-#ifdef ERTS_SMP
 ERTS_GLB_INLINE Port *erts_thr_port_lookup(Eterm id, Uint32 invalid_sflgs);
 ERTS_GLB_INLINE Port *erts_thr_id2port_sflgs(Eterm id, Uint32 invalid_sflgs);
 ERTS_GLB_INLINE void erts_thr_port_release(Port *prt);
-#endif
 ERTS_GLB_INLINE Port *erts_thr_drvport2port(ErlDrvPort, int);
 ERTS_GLB_INLINE Port *erts_drvport2port_state(ErlDrvPort, erts_aint32_t *);
 ERTS_GLB_INLINE Eterm erts_drvport2id(ErlDrvPort);
@@ -489,6 +485,8 @@ ERTS_GLB_INLINE Uint32 erts_portid2status(Eterm);
 ERTS_GLB_INLINE int erts_is_port_alive(Eterm);
 ERTS_GLB_INLINE int erts_is_valid_tracer_port(Eterm);
 ERTS_GLB_INLINE int erts_port_driver_callback_epilogue(Port *, erts_aint32_t *);
+ERTS_GLB_INLINE Port *erts_get_current_port(void);
+ERTS_GLB_INLINE Eterm erts_get_current_port_id(void);
 
 #define erts_drvport2port(Prt) erts_drvport2port_state((Prt), NULL)
 
@@ -507,7 +505,7 @@ erts_port_lookup_raw(Eterm id)
 {
     Port *prt;
 
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_lc_is_delaying());
+    ERTS_LC_ASSERT(erts_thr_progress_lc_is_delaying());
 
     if (is_not_internal_port(id))
 	return NULL;
@@ -536,7 +534,7 @@ erts_id2port(Eterm id)
     Port *prt;
 
     /* Only allowed to be called from managed threads */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
 
     if (is_not_internal_port(id))
 	return NULL;
@@ -547,10 +545,10 @@ erts_id2port(Eterm id)
     if (!prt || prt->common.id != id)
 	return NULL;
 
-    erts_smp_port_lock(prt);
+    erts_port_lock(prt);
     state = erts_atomic32_read_nob(&prt->state);
     if (state & ERTS_PORT_SFLGS_INVALID_LOOKUP) {
-	erts_smp_port_unlock(prt);
+	erts_port_unlock(prt);
 	return NULL;
     }
 
@@ -563,14 +561,12 @@ erts_id2port_sflgs(Eterm id,
 		   Process *c_p, ErtsProcLocks c_p_locks,
 		   Uint32 invalid_sflgs)
 {
-#ifdef ERTS_SMP
     int no_proc_locks = !c_p || !c_p_locks;
-#endif
     erts_aint32_t state;
     Port *prt;
 
     /* Only allowed to be called from managed threads */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
 
     if (is_not_internal_port(id))
 	return NULL;
@@ -581,21 +577,17 @@ erts_id2port_sflgs(Eterm id,
     if (!prt || prt->common.id != id)
 	return NULL;
 
-#ifdef ERTS_SMP
     if (no_proc_locks)
-	erts_smp_port_lock(prt);
-    else if (erts_smp_port_trylock(prt) == EBUSY) {
+	erts_port_lock(prt);
+    else if (erts_port_trylock(prt) == EBUSY) {
 	/* Unlock process locks, and acquire locks in lock order... */
-	erts_smp_proc_unlock(c_p, c_p_locks);
-	erts_smp_port_lock(prt);
-	erts_smp_proc_lock(c_p, c_p_locks);
+	erts_proc_unlock(c_p, c_p_locks);
+	erts_port_lock(prt);
+	erts_proc_lock(c_p, c_p_locks);
     }
-#endif
     state = erts_atomic32_read_nob(&prt->state);
     if (state & invalid_sflgs) {
-#ifdef ERTS_SMP
-	erts_smp_port_unlock(prt);
-#endif
+	erts_port_unlock(prt);
 	return NULL;
     }
 
@@ -606,18 +598,10 @@ ERTS_GLB_INLINE void
 erts_port_release(Port *prt)
 {
     /* Only allowed to be called from managed threads */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
-#ifdef ERTS_SMP
-    erts_smp_port_unlock(prt);
-#else
-    if (prt->cleanup) {
-	prt->cleanup = 0;
-	erts_port_cleanup(prt);
-    }
-#endif
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    erts_port_unlock(prt);
 }
 
-#ifdef ERTS_SMP
 /*
  * erts_thr_id2port_sflgs() and erts_port_dec_refc(prt) can
  * be used by unmanaged threads in the SMP case.
@@ -703,13 +687,10 @@ ERTS_GLB_INLINE void
 erts_thr_port_release(Port *prt)
 {
     erts_mtx_unlock(prt->lock);
-#ifdef ERTS_SMP
     if (!erts_thr_progress_is_managed_thread())
 	erts_port_dec_refc(prt);
-#endif
 }
 
-#endif
 
 ERTS_GLB_INLINE Port *
 erts_thr_drvport2port(ErlDrvPort drvport, int lock_pdl)
@@ -725,7 +706,7 @@ erts_thr_drvport2port(ErlDrvPort drvport, int lock_pdl)
 #ifdef ERTS_ENABLE_LOCK_CHECK
     if (!ERTS_IS_CRASH_DUMPING) {
 	if (erts_lc_is_emu_thr()) {
-	    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 	    ERTS_LC_ASSERT(!prt->port_data_lock
 			   || erts_lc_mtx_is_locked(&prt->port_data_lock->mtx));
 	}
@@ -754,7 +735,7 @@ erts_drvport2port_state(ErlDrvPort drvport, erts_aint32_t *statep)
 //    ERTS_LC_ASSERT(erts_lc_is_emu_thr());
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return ERTS_INVALID_ERL_DRV_PORT;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt)
 		       || ERTS_IS_CRASH_DUMPING);
     /* 
      * This state check is only needed since a driver callback
@@ -811,23 +792,21 @@ erts_port_driver_callback_epilogue(Port *prt, erts_aint32_t *statep)
     int reds = 0;
     erts_aint32_t state;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     state = erts_atomic32_read_nob(&prt->state);
     if ((state & ERTS_PORT_SFLG_CLOSING) && erts_is_port_ioq_empty(prt)) {
 	reds += ERTS_PORT_REDS_TERMINATE;
 	erts_terminate_port(prt);
 	state = erts_atomic32_read_nob(&prt->state);
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     }
 
-#ifdef ERTS_SMP
     if (prt->xports) {
 	reds += erts_port_handle_xports(prt);
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 	ASSERT(!prt->xports);
     }
-#endif
 
     if (statep)
 	*statep = state;
@@ -835,6 +814,20 @@ erts_port_driver_callback_epilogue(Port *prt, erts_aint32_t *statep)
     return reds;
 }
 
+ERTS_GLB_INLINE
+Port *erts_get_current_port(void)
+{
+    ErtsSchedulerData *esdp = erts_get_scheduler_data();
+    return esdp ? esdp->current_port : NULL;
+}
+
+ERTS_GLB_INLINE
+Eterm erts_get_current_port_id(void)
+{
+    Port *port = erts_get_current_port();
+    return port ? port->common.id : THE_NON_VALUE;
+}
+
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
 void erts_port_resume_procs(Port *);
@@ -910,20 +903,20 @@ struct ErtsProc2PortSigData_ {
 	    Eterm item;
 	} info;
 	struct {
-	    Eterm port;
-	    Eterm to;
+            Eterm port_id;
+	    ErtsLink *lnk;
 	} link;
 	struct {
-	    Eterm from;
+            Eterm port_id;
+	    ErtsLink *lnk;
 	} unlink;
         struct {
-            Eterm origin;   /* who receives monitor event, pid */
-            Eterm name;     /* either name for named monitor, or port id */
+            Eterm port_id;
+            ErtsMonitor *mon;
         } monitor;
         struct {
-            Eterm origin;   /* who is at the other end of the monitor, pid */
-            Eterm name;     /* port id */
-            Uint32 ref[ERTS_MAX_REF_NUMBERS]; /* box contents of a ref */
+            Eterm port_id;
+            ErtsMonitor *mon;
         } demonitor;
     } u;
 } ;
@@ -969,7 +962,6 @@ typedef int (*ErtsProc2PortSigCallback)(Port *,
 
 typedef enum {
     ERTS_PORT_OP_BADARG,
-    ERTS_PORT_OP_CALLER_EXIT,
     ERTS_PORT_OP_BUSY,
     ERTS_PORT_OP_BUSY_SCHEDULED,
     ERTS_PORT_OP_SCHEDULED,
@@ -1005,32 +997,13 @@ ErtsPortOpResult erts_port_command(Process *, int, Port *, Eterm, Eterm *);
 ErtsPortOpResult erts_port_output(Process *, int, Port *, Eterm, Eterm, Eterm *);
 ErtsPortOpResult erts_port_exit(Process *, int, Port *, Eterm, Eterm, Eterm *);
 ErtsPortOpResult erts_port_connect(Process *, int, Port *, Eterm, Eterm, Eterm *);
-ErtsPortOpResult erts_port_link(Process *, Port *, Eterm, Eterm *);
-ErtsPortOpResult erts_port_unlink(Process *, Port *, Eterm, Eterm *);
+ErtsPortOpResult erts_port_link(Process *, Port *, ErtsLink *, Eterm *);
+ErtsPortOpResult erts_port_unlink(Process *, Port *, ErtsLink *, Eterm *);
 ErtsPortOpResult erts_port_control(Process *, Port *, unsigned int, Eterm, Eterm *);
 ErtsPortOpResult erts_port_call(Process *, Port *, unsigned int, Eterm, Eterm *);
 ErtsPortOpResult erts_port_info(Process *, Port *, Eterm, Eterm *);
-
-/* Creates monitor between Origin and Target. Ref must be initialized to
- * a reference (ref may be rewritten to be used to serve additionally as a
- * signal id). Name is atom if user monitors port by name or NIL */
-ErtsPortOpResult erts_port_monitor(Process *origin, Port *target, Eterm name,
-                                   Eterm *ref);
-
-typedef enum {
-    /* Normal demonitor rules apply with locking and reductions bump */
-    ERTS_PORT_DEMONITOR_NORMAL = 1,
-    /* Relaxed demonitor rules when process is about to die, which means that
-     * pid lookup won't work, locks won't work, no reductions bump. */
-    ERTS_PORT_DEMONITOR_ORIGIN_ON_DEATHBED = 2,
-} ErtsDemonitorMode;
-
-/* Removes monitor between origin and target, identified by ref.
- * origin_is_dying can be 0 (false, normal locking rules and reductions bump
- * apply) or 1 (true, in case when we avoid origin locking) */
-ErtsPortOpResult erts_port_demonitor(Process *origin, ErtsDemonitorMode mode,
-                                    Port *target, Eterm ref,
-                                    Eterm *trap_ref);
+ErtsPortOpResult erts_port_monitor(Process *, Port *, ErtsMonitor *);
+ErtsPortOpResult erts_port_demonitor(Process *, Port *, ErtsMonitor *);
 /* defined in erl_bif_port.c */
 Port *erts_sig_lookup_port(Process *c_p, Eterm id_or_name);
 
diff --git a/erts/emulator/beam/erl_port_task.c b/erts/emulator/beam/erl_port_task.c
index 4d7a86398a..3953a4c2e9 100644
--- a/erts/emulator/beam/erl_port_task.c
+++ b/erts/emulator/beam/erl_port_task.c
@@ -36,6 +36,7 @@
 #include "erl_check_io.h"
 #include "dtrace-wrapper.h"
 #include "lttng-wrapper.h"
+#include "erl_check_io.h"
 #include <stdarg.h>
 
 /*
@@ -83,15 +84,12 @@ static void chk_task_queues(Port *pp, ErtsPortTask *execq, int processing_busy_q
 #define LTTNG_DRIVER(TRACEPOINT, PP) do {} while(0)
 #endif
 
-#define ERTS_SMP_LC_VERIFY_RQ(RQ, PP)					\
-    do {								\
-	ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));		\
-	ERTS_SMP_LC_ASSERT((RQ) == ((ErtsRunQueue *)			\
-				    erts_smp_atomic_read_nob(&(PP)->run_queue))); \
+#define ERTS_LC_VERIFY_RQ(RQ, PP)				\
+    do {							\
+	ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));		\
+	ERTS_LC_ASSERT((RQ) == erts_get_runq_port((PP)));       \
     } while (0)
 
-erts_smp_atomic_t erts_port_task_outstanding_io_tasks;
-
 #define ERTS_PT_STATE_SCHEDULED		0
 #define ERTS_PT_STATE_ABORTED		1
 #define ERTS_PT_STATE_EXECUTING		2
@@ -99,7 +97,6 @@ erts_smp_atomic_t erts_port_task_outstanding_io_tasks;
 typedef union {
     struct { /* I/O tasks */
 	ErlDrvEvent event;
-	ErlDrvEventData event_data;
     } io;
     struct {
 	ErtsProc2PortSigCallback callback;
@@ -108,7 +105,7 @@ typedef union {
 } ErtsPortTaskTypeData;
 
 struct ErtsPortTask_ {
-    erts_smp_atomic32_t state;
+    erts_atomic32_t state;
     ErtsPortTaskType type;
     union {
 	struct {
@@ -126,9 +123,7 @@ struct ErtsPortTaskHandleList_ {
     ErtsPortTaskHandle handle;
     union {
 	ErtsPortTaskHandleList *next;
-#ifdef ERTS_SMP
 	ErtsThrPrgrLaterOp release;
-#endif
     } u;
 };
 
@@ -151,35 +146,29 @@ static void begin_port_cleanup(Port *pp,
 			       ErtsPortTask **execq,
 			       int *processing_busy_q_p);
 
-ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(port_task,
-				 ErtsPortTask,
-				 1000,
-				 ERTS_ALC_T_PORT_TASK)
+ERTS_THR_PREF_QUICK_ALLOC_IMPL(port_task,
+                               ErtsPortTask,
+                               1000,
+                               ERTS_ALC_T_PORT_TASK)
 
 ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(busy_caller_table,
 				 ErtsPortTaskBusyCallerTable,
 				 50,
 				 ERTS_ALC_T_BUSY_CALLER_TAB)
 
-#ifdef ERTS_SMP
 static void
 call_port_task_free(void *vptp)
 {
     port_task_free((ErtsPortTask *) vptp);
 }
-#endif
 
 static ERTS_INLINE void
 schedule_port_task_free(ErtsPortTask *ptp)
 {
-#ifdef ERTS_SMP
     erts_schedule_thr_prgr_later_cleanup_op(call_port_task_free,
 					    (void *) ptp,
 					    &ptp->u.release,
 					    sizeof(ErtsPortTask));
-#else
-    port_task_free(ptp);
-#endif
 }
 
 static ERTS_INLINE ErtsPortTask *
@@ -199,7 +188,7 @@ p2p_sig_data_init(ErtsPortTask *ptp)
 
     ptp->type = ERTS_PORT_TASK_PROC_SIG;
     ptp->u.alive.flags = ERTS_PT_FLG_SIG_DEP;
-    erts_smp_atomic32_init_nob(&ptp->state, ERTS_PT_STATE_SCHEDULED);
+    erts_atomic32_init_nob(&ptp->state, ERTS_PT_STATE_SCHEDULED);
 
     ASSERT(ptp == p2p_sig_data_to_task(&ptp->u.alive.td.psig.data));
 
@@ -290,7 +279,7 @@ popped_from_busy_queue(Port *pp, ErtsPortTask *ptp, int last)
 #ifdef DEBUG
 	    erts_aint32_t flags =
 #endif
-		erts_smp_atomic32_read_band_nob(
+		erts_atomic32_read_band_nob(
 		    &pp->sched.flags,
 		    ~ERTS_PTS_FLG_HAVE_BUSY_TASKS);
 	    ASSERT(flags & ERTS_PTS_FLG_HAVE_BUSY_TASKS);
@@ -337,7 +326,7 @@ busy_wait_move_to_busy_queue(Port *pp, ErtsPortTask *ptp)
 #ifdef DEBUG
 	flags = 
 #endif
-	    erts_smp_atomic32_read_bor_nob(&pp->sched.flags,
+	    erts_atomic32_read_bor_nob(&pp->sched.flags,
 					   ERTS_PTS_FLG_HAVE_BUSY_TASKS);
 	ASSERT(!(flags & ERTS_PTS_FLG_HAVE_BUSY_TASKS));
 
@@ -477,7 +466,7 @@ no_sig_dep_move_from_busyq(Port *pp)
 	int bix;
 	erts_aint32_t flags =
 #endif
-	    erts_smp_atomic32_read_band_nob(
+	    erts_atomic32_read_band_nob(
 		&pp->sched.flags,
 		~ERTS_PTS_FLG_HAVE_BUSY_TASKS);
 	ASSERT(flags & ERTS_PTS_FLG_HAVE_BUSY_TASKS);
@@ -510,11 +499,11 @@ chk_task_queues(Port *pp, ErtsPortTask *execq, int processing_busy_queue)
     if (!first) {
 	ASSERT(!tabp);
 	ASSERT(!pp->sched.taskq.local.busy.last);
-	ASSERT(!(erts_smp_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_HAVE_BUSY_TASKS));
+	ASSERT(!(erts_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_HAVE_BUSY_TASKS));
 	return;
     }
 
-    ASSERT(erts_smp_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_HAVE_BUSY_TASKS);
+    ASSERT(erts_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_HAVE_BUSY_TASKS);
     ASSERT(tabp);
 
     tot_count = 0;
@@ -570,13 +559,13 @@ chk_task_queues(Port *pp, ErtsPortTask *execq, int processing_busy_queue)
 static ERTS_INLINE void
 reset_port_task_handle(ErtsPortTaskHandle *pthp)
 {
-    erts_smp_atomic_set_relb(pthp, (erts_aint_t) NULL);
+    erts_atomic_set_relb(pthp, (erts_aint_t) NULL);
 }
 
 static ERTS_INLINE ErtsPortTask *
 handle2task(ErtsPortTaskHandle *pthp)
 {
-    return (ErtsPortTask *) erts_smp_atomic_read_acqb(pthp);
+    return (ErtsPortTask *) erts_atomic_read_acqb(pthp);
 }
 
 static ERTS_INLINE void
@@ -593,8 +582,9 @@ reset_executed_io_task_handle(ErtsPortTask *ptp)
 {
     if (ptp->u.alive.handle) {
 	ASSERT(ptp == handle2task(ptp->u.alive.handle));
-	erts_io_notify_port_task_executed(ptp->u.alive.handle);
-	reset_port_task_handle(ptp->u.alive.handle);
+        /* The port task handle is reset inside task_executed */
+	erts_io_notify_port_task_executed(ptp->type, ptp->u.alive.handle,
+                                          reset_port_task_handle);
     }
 }
 
@@ -603,7 +593,7 @@ set_handle(ErtsPortTask *ptp, ErtsPortTaskHandle *pthp)
 {
     ptp->u.alive.handle = pthp;
     if (pthp) {
-	erts_smp_atomic_set_relb(pthp, (erts_aint_t) ptp);
+	erts_atomic_set_relb(pthp, (erts_aint_t) ptp);
 	ASSERT(ptp == handle2task(ptp->u.alive.handle));
     }
 }
@@ -617,7 +607,7 @@ set_tmp_handle(ErtsPortTask *ptp, ErtsPortTaskHandle *pthp)
 	 * IMPORTANT! Task either need to be aborted, or task handle
 	 * need to be detached before thread progress has been made.
 	 */
-	erts_smp_atomic_set_relb(pthp, (erts_aint_t) ptp);
+	erts_atomic_set_relb(pthp, (erts_aint_t) ptp);
     }
 }
 
@@ -635,20 +625,20 @@ check_unset_busy_port_q(Port *pp,
     int resume_procs = 0;
 
     ASSERT(bpq);
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
 
     erts_port_task_sched_lock(&pp->sched);
-    qsize = (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->size);
-    low = (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->low);
+    qsize = (ErlDrvSizeT) erts_atomic_read_nob(&bpq->size);
+    low = (ErlDrvSizeT) erts_atomic_read_nob(&bpq->low);
     if (qsize < low) {
 	erts_aint32_t mask = ~(ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q
 			       | ERTS_PTS_FLG_BUSY_PORT_Q);
-	flags = erts_smp_atomic32_read_band_relb(&pp->sched.flags, mask);
+	flags = erts_atomic32_read_band_relb(&pp->sched.flags, mask);
 	if ((flags & ERTS_PTS_FLGS_BUSY) == ERTS_PTS_FLG_BUSY_PORT_Q)
 	    resume_procs = 1;
     }
     else if (flags & ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q) {
-	flags = erts_smp_atomic32_read_band_relb(&pp->sched.flags,
+	flags = erts_atomic32_read_band_relb(&pp->sched.flags,
 						 ~ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q);
 	flags &= ~ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q;
     }
@@ -673,16 +663,16 @@ aborted_proc2port_data(Port *pp, ErlDrvSizeT size)
 
     bpq = pp->sched.taskq.bpq;
 
-    qsz = (ErlDrvSizeT) erts_smp_atomic_add_read_acqb(&bpq->size,
+    qsz = (ErlDrvSizeT) erts_atomic_add_read_acqb(&bpq->size,
 						      (erts_aint_t) -size);
     ASSERT(qsz + size > qsz);
-    flags = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    flags = erts_atomic32_read_nob(&pp->sched.flags);
     ASSERT(pp->sched.taskq.bpq);
     if ((flags & (ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q
 		  | ERTS_PTS_FLG_BUSY_PORT_Q)) != ERTS_PTS_FLG_BUSY_PORT_Q)
 	return;
-    if (qsz < (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->low))
-	erts_smp_atomic32_read_bor_nob(&pp->sched.flags,
+    if (qsz < (ErlDrvSizeT) erts_atomic_read_nob(&bpq->low))
+	erts_atomic32_read_bor_nob(&pp->sched.flags,
 				       ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q);
 }
 
@@ -700,13 +690,13 @@ dequeued_proc2port_data(Port *pp, ErlDrvSizeT size)
 
     bpq = pp->sched.taskq.bpq;
 
-    qsz = (ErlDrvSizeT) erts_smp_atomic_add_read_acqb(&bpq->size,
+    qsz = (ErlDrvSizeT) erts_atomic_add_read_acqb(&bpq->size,
 						      (erts_aint_t) -size);
     ASSERT(qsz + size > qsz);
-    flags = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    flags = erts_atomic32_read_nob(&pp->sched.flags);
     if (!(flags & ERTS_PTS_FLG_BUSY_PORT_Q))
 	return;
-    if (qsz < (ErlDrvSizeT) erts_smp_atomic_read_acqb(&bpq->low))
+    if (qsz < (ErlDrvSizeT) erts_atomic_read_acqb(&bpq->low))
 	check_unset_busy_port_q(pp, flags, bpq);
 }
 
@@ -719,19 +709,19 @@ enqueue_proc2port_data(Port *pp,
     if (sigdp && bpq) {
 	ErlDrvSizeT size = erts_proc2port_sig_command_data_size(sigdp);
 	if (size) {
-	    erts_aint_t asize = erts_smp_atomic_add_read_acqb(&bpq->size,
+	    erts_aint_t asize = erts_atomic_add_read_acqb(&bpq->size,
 							      (erts_aint_t) size);
 	    ErlDrvSizeT qsz = (ErlDrvSizeT) asize;
 
 	    ASSERT(qsz - size < qsz);
 
 	    if (!(flags & ERTS_PTS_FLG_BUSY_PORT_Q) && qsz > bpq->high) {
-		flags = erts_smp_atomic32_read_bor_acqb(&pp->sched.flags,
+		flags = erts_atomic32_read_bor_acqb(&pp->sched.flags,
 							ERTS_PTS_FLG_BUSY_PORT_Q);
 		flags |= ERTS_PTS_FLG_BUSY_PORT_Q;
-		qsz = (ErlDrvSizeT) erts_smp_atomic_read_acqb(&bpq->size);
-		if (qsz < (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->low)) {
-		    flags = (erts_smp_atomic32_read_bor_relb(
+		qsz = (ErlDrvSizeT) erts_atomic_read_acqb(&bpq->size);
+		if (qsz < (ErlDrvSizeT) erts_atomic_read_nob(&bpq->low)) {
+		    flags = (erts_atomic32_read_bor_relb(
 				 &pp->sched.flags,
 				 ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q));
 		    flags |= ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q;
@@ -779,18 +769,18 @@ erl_drv_busy_msgq_limits(ErlDrvPort dport, ErlDrvSizeT *lowp, ErlDrvSizeT *highp
 	erts_aint32_t flags;
 	pp->sched.taskq.bpq = NULL;
 	flags = ~(ERTS_PTS_FLG_BUSY_PORT_Q|ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q);
-	flags = erts_smp_atomic32_read_band_acqb(&pp->sched.flags, flags);
+	flags = erts_atomic32_read_band_acqb(&pp->sched.flags, flags);
 	if ((flags & ERTS_PTS_FLGS_BUSY) == ERTS_PTS_FLG_BUSY_PORT_Q)
 	    resume_procs = 1;
     }
     else {
 
 	if (!low)
-	    low = (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->low);
+	    low = (ErlDrvSizeT) erts_atomic_read_nob(&bpq->low);
 	else {
 	    if (bpq->high < low)
 		bpq->high = low;
-	    erts_smp_atomic_set_relb(&bpq->low, (erts_aint_t) low);
+	    erts_atomic_set_relb(&bpq->low, (erts_aint_t) low);
 	    written = 1;
 	}
     
@@ -799,19 +789,19 @@ erl_drv_busy_msgq_limits(ErlDrvPort dport, ErlDrvSizeT *lowp, ErlDrvSizeT *highp
 	else {
 	    if (low > high) {
 		low = high;
-		erts_smp_atomic_set_relb(&bpq->low, (erts_aint_t) low);
+		erts_atomic_set_relb(&bpq->low, (erts_aint_t) low);
 	    }
 	    bpq->high = high;
 	    written = 1;
 	}
 
 	if (written) {
-	    ErlDrvSizeT size = (ErlDrvSizeT) erts_smp_atomic_read_nob(&bpq->size);
+	    ErlDrvSizeT size = (ErlDrvSizeT) erts_atomic_read_nob(&bpq->size);
 	    if (size > high)
-		erts_smp_atomic32_read_bor_relb(&pp->sched.flags,
+		erts_atomic32_read_bor_relb(&pp->sched.flags,
 						ERTS_PTS_FLG_BUSY_PORT_Q);
 	    else if (size < low)
-		erts_smp_atomic32_read_bor_relb(&pp->sched.flags,
+		erts_atomic32_read_bor_relb(&pp->sched.flags,
 						ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q);
 	}
     }
@@ -830,25 +820,19 @@ erl_drv_busy_msgq_limits(ErlDrvPort dport, ErlDrvSizeT *lowp, ErlDrvSizeT *highp
  * No-suspend handles.
  */
 
-#ifdef ERTS_SMP
 static void
 free_port_task_handle_list(void *vpthlp)
 {
     erts_free(ERTS_ALC_T_PT_HNDL_LIST, vpthlp);
 }
-#endif
 
 static void
 schedule_port_task_handle_list_free(ErtsPortTaskHandleList *pthlp)
 {
-#ifdef ERTS_SMP
     erts_schedule_thr_prgr_later_cleanup_op(free_port_task_handle_list,
 					    (void *) pthlp,
 					    &pthlp->u.release,
 					    sizeof(ErtsPortTaskHandleList));
-#else
-    erts_free(ERTS_ALC_T_PT_HNDL_LIST, pthlp);
-#endif
 }
 
 static ERTS_INLINE void
@@ -891,7 +875,7 @@ get_free_nosuspend_handles(Port *pp)
 {
     ErtsPortTaskHandleList *nshp, *last_nshp = NULL;
 
-    ERTS_SMP_LC_ASSERT(erts_port_task_sched_lock_is_locked(&pp->sched));
+    ERTS_LC_ASSERT(erts_port_task_sched_lock_is_locked(&pp->sched));
 
     nshp = pp->sched.taskq.local.busy.nosuspend;
 
@@ -907,7 +891,7 @@ get_free_nosuspend_handles(Port *pp)
 	pp->sched.taskq.local.busy.nosuspend = last_nshp->u.next;
 	last_nshp->u.next = NULL;
 	if (!pp->sched.taskq.local.busy.nosuspend)
-	    erts_smp_atomic32_read_band_nob(&pp->sched.flags,
+	    erts_atomic32_read_band_nob(&pp->sched.flags,
 					    ~ERTS_PTS_FLG_HAVE_NS_TASKS);
     }
     return nshp;
@@ -930,7 +914,7 @@ free_nosuspend_handles(ErtsPortTaskHandleList *free_nshp)
 static ERTS_INLINE void
 enqueue_port(ErtsRunQueue *runq, Port *pp)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
     pp->sched.next = NULL;
     if (runq->ports.end) {
 	ASSERT(runq->ports.start);
@@ -944,19 +928,17 @@ enqueue_port(ErtsRunQueue *runq, Port *pp)
     runq->ports.end = pp;
     ASSERT(runq->ports.start && runq->ports.end);
 
-    erts_smp_inc_runq_len(runq, &runq->ports.info, ERTS_PORT_PRIO_LEVEL);
+    erts_inc_runq_len(runq, &runq->ports.info, ERTS_PORT_PRIO_LEVEL);
 
-#ifdef ERTS_SMP
     if (ERTS_RUNQ_FLGS_GET_NOB(runq) & ERTS_RUNQ_FLG_HALTING)
 	erts_non_empty_runq(runq);
-#endif
 }
 
 static ERTS_INLINE Port *
 pop_port(ErtsRunQueue *runq)
 {
     Port *pp = runq->ports.start;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
     if (!pp) {
 	ASSERT(!runq->ports.end);
     }
@@ -966,7 +948,7 @@ pop_port(ErtsRunQueue *runq)
 	    ASSERT(runq->ports.end == pp);
 	    runq->ports.end = NULL;
 	}
-	erts_smp_dec_runq_len(runq, &runq->ports.info, ERTS_PORT_PRIO_LEVEL);
+	erts_dec_runq_len(runq, &runq->ports.info, ERTS_PORT_PRIO_LEVEL);
     }
 
     ASSERT(runq->ports.start || !runq->ports.end);
@@ -993,7 +975,7 @@ enqueue_task(Port *pp,
     if (ns_pthlp)
 	fail_flags |= ERTS_PTS_FLG_BUSY_PORT;
     erts_port_task_sched_lock(&pp->sched);
-    flags = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    flags = erts_atomic32_read_nob(&pp->sched.flags);
     if (flags & fail_flags)
 	res = 0;
     else {
@@ -1024,7 +1006,7 @@ enqueue_task(Port *pp,
 static ERTS_INLINE void
 prepare_exec(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 {
-    erts_aint32_t act = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    erts_aint32_t act = erts_atomic32_read_nob(&pp->sched.flags);
 
     if (!pp->sched.taskq.local.busy.first || (act & ERTS_PTS_FLG_BUSY_PORT)) {
 	*execqp = pp->sched.taskq.local.first;
@@ -1045,7 +1027,7 @@ prepare_exec(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 	new &= ~ERTS_PTS_FLG_IN_RUNQ;
 	new |= ERTS_PTS_FLG_EXEC;
 
-	act = erts_smp_atomic32_cmpxchg_nob(&pp->sched.flags, new, exp);
+	act = erts_atomic32_cmpxchg_nob(&pp->sched.flags, new, exp);
 
 	ASSERT(act & ERTS_PTS_FLG_IN_RUNQ);
 
@@ -1072,7 +1054,7 @@ finalize_exec(Port *pp, ErtsPortTask **execq, int processing_busy_q)
 
     *execq = NULL;
 
-    act = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    act = erts_atomic32_read_nob(&pp->sched.flags);
     if (act & ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q)
 	act = check_unset_busy_port_q(pp, act, pp->sched.taskq.bpq);
 
@@ -1089,7 +1071,7 @@ finalize_exec(Port *pp, ErtsPortTask **execq, int processing_busy_q)
 	if (act & ERTS_PTS_FLG_HAVE_TASKS)
 	    new |= ERTS_PTS_FLG_IN_RUNQ;
 
-	act = erts_smp_atomic32_cmpxchg_relb(&pp->sched.flags, new, exp);
+	act = erts_atomic32_cmpxchg_relb(&pp->sched.flags, new, exp);
 
 	ERTS_LC_ASSERT(!(act & ERTS_PTS_FLG_IN_RUNQ));
 	ERTS_LC_ASSERT(!(act & ERTS_PTS_FLG_EXEC_IMM));
@@ -1115,7 +1097,7 @@ finalize_exec(Port *pp, ErtsPortTask **execq, int processing_busy_q)
 static ERTS_INLINE erts_aint32_t
 select_queue_for_exec(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 {
-    erts_aint32_t flags = erts_smp_atomic32_read_nob(&pp->sched.flags);
+    erts_aint32_t flags = erts_atomic32_read_nob(&pp->sched.flags);
 
     if (flags & ERTS_PTS_FLG_CHK_UNSET_BUSY_PORT_Q)
 	flags = check_unset_busy_port_q(pp, flags, pp->sched.taskq.bpq);
@@ -1225,7 +1207,7 @@ fetch_in_queue(Port *pp, ErtsPortTask **execqp)
     if (ptp)
 	*execqp = ptp->u.alive.next;
     else
-	erts_smp_atomic32_read_band_nob(&pp->sched.flags,
+	erts_atomic32_read_band_nob(&pp->sched.flags,
 					~ERTS_PTS_FLG_HAVE_TASKS);
 
     
@@ -1288,7 +1270,7 @@ erl_drv_consume_timeslice(ErlDrvPort dprt, int percent)
 void
 erts_port_task_tmp_handle_detach(ErtsPortTaskHandle *pthp)
 {
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_lc_is_delaying());
+    ERTS_LC_ASSERT(erts_thr_progress_lc_is_delaying());
     reset_port_task_handle(pthp);
 }
 
@@ -1301,9 +1283,7 @@ erts_port_task_abort(ErtsPortTaskHandle *pthp)
 {
     int res;
     ErtsPortTask *ptp;
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
     ptp = handle2task(pthp);
     if (!ptp)
@@ -1313,41 +1293,25 @@ erts_port_task_abort(ErtsPortTaskHandle *pthp)
 
 #ifdef DEBUG
 	ErtsPortTaskHandle *saved_pthp = ptp->u.alive.handle;
-	ERTS_SMP_READ_MEMORY_BARRIER;
-	old_state = erts_smp_atomic32_read_nob(&ptp->state);
+	ERTS_THR_READ_MEMORY_BARRIER;
+	old_state = erts_atomic32_read_nob(&ptp->state);
 	if (old_state == ERTS_PT_STATE_SCHEDULED) {
 	    ASSERT(!saved_pthp || saved_pthp == pthp);
 	}
 #endif
 
-	old_state = erts_smp_atomic32_cmpxchg_nob(&ptp->state,
+	old_state = erts_atomic32_cmpxchg_nob(&ptp->state,
 						  ERTS_PT_STATE_ABORTED,
 						  ERTS_PT_STATE_SCHEDULED);
 	if (old_state != ERTS_PT_STATE_SCHEDULED)
 	    res = - 1; /* Task already aborted, executing, or executed */
 	else {
-
 	    reset_port_task_handle(pthp);
-
-	    switch (ptp->type) {
-	    case ERTS_PORT_TASK_INPUT:
-	    case ERTS_PORT_TASK_OUTPUT:
-	    case ERTS_PORT_TASK_EVENT:
-		ASSERT(erts_smp_atomic_read_nob(
-			   &erts_port_task_outstanding_io_tasks) > 0);
-		erts_smp_atomic_dec_relb(&erts_port_task_outstanding_io_tasks);
-		break;
-	    default:
-		break;
-	    }
-
 	    res = 0;
 	}
     }
 
-#ifdef ERTS_SMP
     erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 
     return res;
 }
@@ -1356,12 +1320,10 @@ void
 erts_port_task_abort_nosuspend_tasks(Port *pp)
 {
     ErtsPortTaskHandleList *abort_list;
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl = ERTS_THR_PRGR_DHANDLE_INVALID;
-#endif
 
     erts_port_task_sched_lock(&pp->sched);
-    erts_smp_atomic32_read_band_nob(&pp->sched.flags,
+    erts_atomic32_read_band_nob(&pp->sched.flags,
 				    ~ERTS_PTS_FLG_HAVE_NS_TASKS);
     abort_list = pp->sched.taskq.local.busy.nosuspend;
     pp->sched.taskq.local.busy.nosuspend = NULL;
@@ -1381,40 +1343,34 @@ erts_port_task_abort_nosuspend_tasks(Port *pp)
 	pthlp = abort_list;
 	abort_list = pthlp->u.next;
 
-#ifdef ERTS_SMP
 	if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	    dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
 	pthp = &pthlp->handle;
 	ptp = handle2task(pthp);
 	if (!ptp) {
-#ifdef ERTS_SMP
 	    if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 		erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 	    schedule_port_task_handle_list_free(pthlp);
 	    continue;
 	}
 
 #ifdef DEBUG
 	saved_pthp = ptp->u.alive.handle;
-	ERTS_SMP_READ_MEMORY_BARRIER;
-	old_state = erts_smp_atomic32_read_nob(&ptp->state);
+	ERTS_THR_READ_MEMORY_BARRIER;
+	old_state = erts_atomic32_read_nob(&ptp->state);
 	if (old_state == ERTS_PT_STATE_SCHEDULED) {
 	    ASSERT(saved_pthp == pthp);
 	}
 #endif
 
-	old_state = erts_smp_atomic32_cmpxchg_nob(&ptp->state,
+	old_state = erts_atomic32_cmpxchg_nob(&ptp->state,
 						  ERTS_PT_STATE_ABORTED,
 						  ERTS_PT_STATE_SCHEDULED);
 	if (old_state != ERTS_PT_STATE_SCHEDULED) {
 	    /* Task already aborted, executing, or executed */
-#ifdef ERTS_SMP
 	    if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 		erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 	    schedule_port_task_handle_list_free(pthlp);
 	    continue;
 	}
@@ -1424,10 +1380,8 @@ erts_port_task_abort_nosuspend_tasks(Port *pp)
 	type = ptp->type;
 	td = ptp->u.alive.td;
 
-#ifdef ERTS_SMP
 	if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	    erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 	schedule_port_task_handle_list_free(pthlp);
 
 	abort_nosuspend_task(pp, type, &td, pp->sched.taskq.bpq != NULL);
@@ -1446,10 +1400,8 @@ erts_port_task_schedule(Eterm id,
 {
     ErtsProc2PortSigData *sigdp = NULL;
     ErtsPortTaskHandleList *ns_pthlp = NULL;
-#ifdef ERTS_SMP
     ErtsRunQueue *xrunq;
     ErtsThrPrgrDelayHandle dhndl;
-#endif
     ErtsRunQueue *runq;
     Port *pp;
     ErtsPortTask *ptp = NULL;
@@ -1460,19 +1412,15 @@ erts_port_task_schedule(Eterm id,
 
     ASSERT(is_internal_port(id));
 
-#ifdef ERTS_SMP
     dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
     pp = erts_port_lookup_raw(id);
 
-#ifdef ERTS_SMP
     if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED) {
 	if (pp)
 	    erts_port_inc_refc(pp);
 	erts_thr_progress_unmanaged_continue(dhndl);
     }
-#endif
 
     if (type != ERTS_PORT_TASK_PROC_SIG) {
         if (!pp)
@@ -1483,7 +1431,7 @@ erts_port_task_schedule(Eterm id,
 	ptp->type = type;
 	ptp->u.alive.flags = 0;
 
-	erts_smp_atomic32_init_nob(&ptp->state, ERTS_PT_STATE_SCHEDULED);
+	erts_atomic32_init_nob(&ptp->state, ERTS_PT_STATE_SCHEDULED);
 
 	set_handle(ptp, pthp);
     }
@@ -1495,16 +1443,6 @@ erts_port_task_schedule(Eterm id,
 	va_start(argp, type);
 	ptp->u.alive.td.io.event = va_arg(argp, ErlDrvEvent);
 	va_end(argp);
-	erts_smp_atomic_inc_relb(&erts_port_task_outstanding_io_tasks);
-	break;
-    }
-    case ERTS_PORT_TASK_EVENT: {
-	va_list argp;
-	va_start(argp, type);
-	ptp->u.alive.td.io.event = va_arg(argp, ErlDrvEvent);
-	ptp->u.alive.td.io.event_data = va_arg(argp, ErlDrvEventData);
-	va_end(argp);
-	erts_smp_atomic_inc_relb(&erts_port_task_outstanding_io_tasks);
 	break;
     }
     case ERTS_PORT_TASK_PROC_SIG: {
@@ -1559,7 +1497,7 @@ erts_port_task_schedule(Eterm id,
 	if (!(act & (ERTS_PTS_FLG_IN_RUNQ|ERTS_PTS_FLG_EXEC)))
 	    new |= ERTS_PTS_FLG_IN_RUNQ;
 
-	act = erts_smp_atomic32_cmpxchg_relb(&pp->sched.flags, new, exp);
+	act = erts_atomic32_cmpxchg_relb(&pp->sched.flags, new, exp);
 
 	if (exp == act) {
 	    if (!(act & (ERTS_PTS_FLG_IN_RUNQ|ERTS_PTS_FLG_EXEC)))
@@ -1581,47 +1519,37 @@ erts_port_task_schedule(Eterm id,
     /* Enqueue port on run-queue */
 
     runq = erts_port_runq(pp);
-    if (!runq)
-	ERTS_INTERNAL_ERROR("Missing run-queue");
 
-#ifdef ERTS_SMP
     xrunq = erts_check_emigration_need(runq, ERTS_PORT_PRIO_LEVEL);
-    ERTS_SMP_LC_ASSERT(runq != xrunq);
-    ERTS_SMP_LC_VERIFY_RQ(runq, pp);
+    ERTS_LC_ASSERT(runq != xrunq);
+    ERTS_LC_VERIFY_RQ(runq, pp);
     if (xrunq) {
 	/* Emigrate port ... */
-	erts_smp_atomic_set_nob(&pp->run_queue, (erts_aint_t) xrunq);
-	erts_smp_runq_unlock(runq);
+        erts_set_runq_port(pp, xrunq);
+	erts_runq_unlock(runq);
 	runq = erts_port_runq(pp);
-	if (!runq)
-	    ERTS_INTERNAL_ERROR("Missing run-queue");
     }
-#endif
 
     enqueue_port(runq, pp);
 
-    erts_smp_runq_unlock(runq);
+    erts_runq_unlock(runq);
 
-    erts_smp_notify_inc_runq(runq);
+    erts_notify_inc_runq(runq);
 
 done:
 
     if (prof_runnable_ports)
 	erts_port_task_sched_unlock(&pp->sched);
 
-#ifdef ERTS_SMP
     if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	erts_port_dec_refc(pp);
-#endif
 
     return 0;
 
 abort_nosuspend:
 
-#ifdef ERTS_SMP
     if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	erts_port_dec_refc(pp);
-#endif
 
     abort_nosuspend_task(pp, ptp->type, &ptp->u.alive.td, 0);
 
@@ -1635,14 +1563,13 @@ abort_nosuspend:
 
 fail:
 
-#ifdef ERTS_SMP
     if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	erts_port_dec_refc(pp);
-#endif
 
     if (ptp) {
-        abort_signal_task(pp, ERTS_PROC2PORT_SIG_ABORT,
-                          ptp->type, &ptp->u.alive.td, 0);
+        if (ptp->type == ERTS_PORT_TASK_PROC_SIG)
+            abort_signal_task(pp, ERTS_PROC2PORT_SIG_ABORT,
+                              ptp->type, &ptp->u.alive.td, 0);
 	port_task_free(ptp);
     }
 
@@ -1658,14 +1585,12 @@ erts_port_task_free_port(Port *pp)
     erts_aint32_t flags;
     ErtsRunQueue *runq;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
     ASSERT(!(erts_atomic32_read_nob(&pp->state) & ERTS_PORT_SFLGS_DEAD));
 
     runq = erts_port_runq(pp);
-    if (!runq)
-	ERTS_INTERNAL_ERROR("Missing run-queue");
     erts_port_task_sched_lock(&pp->sched);
-    flags = erts_smp_atomic32_read_bor_relb(&pp->sched.flags,
+    flags = erts_atomic32_read_bor_relb(&pp->sched.flags,
 					    ERTS_PTS_FLG_EXIT);
     erts_port_task_sched_unlock(&pp->sched);
     erts_atomic32_read_bset_relb(&pp->state,
@@ -1675,7 +1600,7 @@ erts_port_task_free_port(Port *pp)
 				  | ERTS_PORT_SFLG_FREE),
 				 ERTS_PORT_SFLG_FREE);
 
-    erts_smp_runq_unlock(runq);
+    erts_runq_unlock(runq);
 
     if (!(flags & (ERTS_PTS_FLG_IN_RUNQ|ERTS_PTS_FLG_EXEC)))
 	begin_port_cleanup(pp, NULL, NULL);
@@ -1688,13 +1613,12 @@ erts_port_task_free_port(Port *pp)
  * scheduling of processes. Run-queue lock should be held by caller.
  */
 
-int
+void
 erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 {
     Port *pp;
     ErtsPortTask *execq;
     int processing_busy_q;
-    int res = 0;
     int vreds = 0;
     int reds = 0;
     erts_aint_t io_tasks_executed = 0;
@@ -1705,17 +1629,16 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
     ErtsSchedulerData *esdp = runq->scheduler;
     ERTS_MSACC_PUSH_STATE_M();
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
 
     pp = pop_port(runq);
     if (!pp) {
-	res = 0;
 	goto done;
     }
 
-    ERTS_SMP_LC_VERIFY_RQ(runq, pp);
+    ERTS_LC_VERIFY_RQ(runq, pp);
 
-    erts_smp_runq_unlock(runq);
+    erts_runq_unlock(runq);
 
     *curr_port_pp = pp;
     
@@ -1723,19 +1646,19 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 	Uint old = ERTS_PORT_SCHED_ID(pp, esdp->no);
 	int migrated = old && old != esdp->no;
 
-	erts_smp_spin_lock(&erts_sched_stat.lock);
+	erts_spin_lock(&erts_sched_stat.lock);
 	erts_sched_stat.prio[ERTS_PORT_PRIO_LEVEL].total_executed++;
 	erts_sched_stat.prio[ERTS_PORT_PRIO_LEVEL].executed++;
 	if (migrated) {
 	    erts_sched_stat.prio[ERTS_PORT_PRIO_LEVEL].total_migrated++;
 	    erts_sched_stat.prio[ERTS_PORT_PRIO_LEVEL].migrated++;
 	}
-	erts_smp_spin_unlock(&erts_sched_stat.lock);
+	erts_spin_unlock(&erts_sched_stat.lock);
     }
 
     prepare_exec(pp, &execq, &processing_busy_q);
 
-    erts_smp_port_lock(pp);
+    erts_port_lock(pp);
 
     /* trace port scheduling, in */
     if (IS_TRACED_FL(pp, F_TRACE_SCHED_PORTS)) {
@@ -1757,7 +1680,7 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 	if (!ptp)
 	    break;
 
-	task_state = erts_smp_atomic32_cmpxchg_nob(&ptp->state,
+	task_state = erts_atomic32_cmpxchg_nob(&ptp->state,
 						   ERTS_PT_STATE_EXECUTING,
 						   ERTS_PT_STATE_SCHEDULED);
 	if (task_state != ERTS_PT_STATE_SCHEDULED) {
@@ -1769,8 +1692,8 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 	    start_time = erts_timestamp_millis();
 	}
 
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
-	ERTS_SMP_CHK_NO_PROC_LOCKS;
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
+	ERTS_CHK_NO_PROC_LOCKS;
 	ASSERT(pp->drv_ptr);
 
 	switch (ptp->type) {
@@ -1812,17 +1735,6 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 	    reset_executed_io_task_handle(ptp);
 	    io_tasks_executed++;
 	    break;
-	case ERTS_PORT_TASK_EVENT:
-	    reds = ERTS_PORT_REDS_EVENT;
-	    ASSERT((state & ERTS_PORT_SFLGS_DEAD) == 0);
-            DTRACE_DRIVER(driver_event, pp);
-            LTTNG_DRIVER(driver_event, pp);
-	    (*pp->drv_ptr->event)((ErlDrvData) pp->drv_data,
-				  ptp->u.alive.td.io.event,
-				  ptp->u.alive.td.io.event_data);
-	    reset_executed_io_task_handle(ptp);
-	    io_tasks_executed++;
-	    break;
 	case ERTS_PORT_TASK_PROC_SIG: {
 	    ErtsProc2PortSigData *sigdp = &ptp->u.alive.td.psig.data;
 	    reset_handle(ptp);
@@ -1887,17 +1799,7 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
     erts_unblock_fpe(fpe_was_unmasked);
     ERTS_MSACC_POP_STATE_M();
 
-
-    if (io_tasks_executed) {
-	ASSERT(erts_smp_atomic_read_nob(&erts_port_task_outstanding_io_tasks)
-	       >= io_tasks_executed);
-	erts_smp_atomic_add_relb(&erts_port_task_outstanding_io_tasks,
-				 -1*io_tasks_executed);
-    }
-
-#ifdef ERTS_SMP
-    ASSERT(runq == (ErtsRunQueue *) erts_smp_atomic_read_nob(&pp->run_queue));
-#endif
+    ASSERT(runq == erts_get_runq_port(pp));
 
     active = finalize_exec(pp, &execq, processing_busy_q);
 
@@ -1907,54 +1809,42 @@ erts_port_task_execute(ErtsRunQueue *runq, Port **curr_port_pp)
 
     *curr_port_pp = NULL;
 
-    erts_smp_runq_lock(runq);
+    erts_runq_lock(runq);
  
     if (active) {
-#ifdef ERTS_SMP
 	ErtsRunQueue *xrunq;
-#endif
 
 	ASSERT(!(erts_atomic32_read_nob(&pp->state) & ERTS_PORT_SFLGS_DEAD));
 
-#ifdef ERTS_SMP
 	xrunq = erts_check_emigration_need(runq, ERTS_PORT_PRIO_LEVEL);
-	ERTS_SMP_LC_ASSERT(runq != xrunq);
-	ERTS_SMP_LC_VERIFY_RQ(runq, pp);
+	ERTS_LC_ASSERT(runq != xrunq);
+	ERTS_LC_VERIFY_RQ(runq, pp);
 	if (!xrunq) {
-#endif
 	    enqueue_port(runq, pp);
 	    /* No need to notify ourselves about inc in runq. */
-#ifdef ERTS_SMP
 	}
 	else {
 	    /* Emigrate port... */
-	    erts_smp_atomic_set_nob(&pp->run_queue, (erts_aint_t) xrunq);
-	    erts_smp_runq_unlock(runq);
+            erts_set_runq_port(pp, xrunq);
+	    erts_runq_unlock(runq);
 
 	    xrunq = erts_port_runq(pp);
-	    ASSERT(xrunq);
 	    enqueue_port(xrunq, pp);
-	    erts_smp_runq_unlock(xrunq);
-	    erts_smp_notify_inc_runq(xrunq);
+	    erts_runq_unlock(xrunq);
+	    erts_notify_inc_runq(xrunq);
 
-	    erts_smp_runq_lock(runq);
+	    erts_runq_lock(runq);
 	}
-#endif
     }
 
  done:
-    res = (erts_smp_atomic_read_nob(&erts_port_task_outstanding_io_tasks)
-	   != (erts_aint_t) 0);
 
     runq->scheduler->reductions += reds;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
     ERTS_PORT_REDUCTIONS_EXECUTED(esdp, runq, reds);
-
-    return res;
 }
 
-#ifdef ERTS_SMP
 static void
 release_port(void *vport)
 {
@@ -1970,7 +1860,6 @@ schedule_release_port(void *vport) {
 				  &pp->common.u.release);
 }
 
-#endif
 
 static void
 begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
@@ -1981,7 +1870,7 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
     ErtsPortTaskHandleList *free_nshp = NULL;
     ErtsProcList *plp;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
 
     /*
      * Abort remaining tasks...
@@ -2054,11 +1943,11 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 	    qs[i] = ptp->u.alive.next;
 
 	    /* Normal case here is aborted tasks... */
-	    state = erts_smp_atomic32_read_nob(&ptp->state);
+	    state = erts_atomic32_read_nob(&ptp->state);
 	    if (state == ERTS_PT_STATE_ABORTED)
 		goto aborted_port_task;
 
-	    state = erts_smp_atomic32_cmpxchg_nob(&ptp->state,
+	    state = erts_atomic32_cmpxchg_nob(&ptp->state,
 						  ERTS_PT_STATE_EXECUTING,
 						  ERTS_PT_STATE_SCHEDULED);
 	    if (state != ERTS_PT_STATE_SCHEDULED) {
@@ -2085,13 +1974,6 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 				      DO_WRITE,
 				      1);
 		break;
-	    case ERTS_PORT_TASK_EVENT:
-		erts_stale_drv_select(pp->common.id,
-				      ERTS_Port2ErlDrvPort(pp),
-				      ptp->u.alive.td.io.event,
-				      0,
-				      1);
-		break;
 	    case ERTS_PORT_TASK_DIST_CMD:
 		break;
 	    case ERTS_PORT_TASK_PROC_SIG: {
@@ -2122,7 +2004,7 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 	}
     }
 
-    erts_smp_atomic32_read_band_nob(&pp->sched.flags,
+    erts_atomic32_read_band_nob(&pp->sched.flags,
 				    ~(ERTS_PTS_FLG_HAVE_BUSY_TASKS
 				      |ERTS_PTS_FLG_HAVE_TASKS
 				      |ERTS_PTS_FLGS_BUSY));
@@ -2164,7 +2046,6 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
     /*
      * Schedule cleanup of port structure...
      */
-#ifdef ERTS_SMP
     /* We might not be a scheduler, eg. traceing to port we are sys_msg_dispatcher */
     if (!erts_get_scheduler_data()) {
       erts_schedule_misc_aux_work(1, schedule_release_port, (void*)pp);
@@ -2174,19 +2055,15 @@ begin_port_cleanup(Port *pp, ErtsPortTask **execqp, int *processing_busy_q_p)
 				      (void *) pp,
 				      &pp->common.u.release);
     }
-#else
-    pp->cleanup = 1;
-#endif
 }
 
-#ifdef ERTS_SMP
 
 void
 erts_enqueue_port(ErtsRunQueue *rq, Port *pp)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
-    ASSERT(rq == (ErtsRunQueue *) erts_smp_atomic_read_nob(&pp->run_queue));
-    ASSERT(erts_smp_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_IN_RUNQ);
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
+    ASSERT(rq == erts_get_runq_port(pp));
+    ASSERT(erts_atomic32_read_nob(&pp->sched.flags) & ERTS_PTS_FLG_IN_RUNQ);
     enqueue_port(rq, pp);
 }
 
@@ -2194,16 +2071,14 @@ Port *
 erts_dequeue_port(ErtsRunQueue *rq)
 {
     Port *pp;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
     pp = pop_port(rq);
-    ASSERT(!pp
-	   || rq == (ErtsRunQueue *) erts_smp_atomic_read_nob(&pp->run_queue));
-    ASSERT(!pp || (erts_smp_atomic32_read_nob(&pp->sched.flags)
+    ASSERT(!pp || rq == erts_get_runq_port(pp));
+    ASSERT(!pp || (erts_atomic32_read_nob(&pp->sched.flags)
 		   & ERTS_PTS_FLG_IN_RUNQ));
     return pp;
 }
 
-#endif
 
 /*
  * Initialize the module.
@@ -2211,8 +2086,7 @@ erts_dequeue_port(ErtsRunQueue *rq)
 void
 erts_port_task_init(void)
 {
-    erts_smp_atomic_init_nob(&erts_port_task_outstanding_io_tasks,
-			     (erts_aint_t) 0);
-    init_port_task_alloc();
+    init_port_task_alloc(erts_no_schedulers + erts_no_poll_threads
+                         + 1); /* aux_thread */
     init_busy_caller_table_alloc();
 }
diff --git a/erts/emulator/beam/erl_port_task.h b/erts/emulator/beam/erl_port_task.h
index 39f403b443..ae78a7d8a3 100644
--- a/erts/emulator/beam/erl_port_task.h
+++ b/erts/emulator/beam/erl_port_task.h
@@ -27,11 +27,11 @@
 #ifndef ERTS_PORT_TASK_H_BASIC_TYPES__
 #define ERTS_PORT_TASK_H_BASIC_TYPES__
 #include "erl_sys_driver.h"
-#include "erl_smp.h"
+#include "erl_threads.h"
 #define ERL_PORT_GET_PORT_TYPE_ONLY__
 #include "erl_port.h"
 #undef ERL_PORT_GET_PORT_TYPE_ONLY__
-typedef erts_smp_atomic_t ErtsPortTaskHandle;
+typedef erts_atomic_t ErtsPortTaskHandle;
 #endif
 
 #ifndef ERTS_PORT_TASK_ONLY_BASIC_TYPES__
@@ -56,17 +56,11 @@ typedef erts_smp_atomic_t ErtsPortTaskHandle;
 typedef enum {
     ERTS_PORT_TASK_INPUT,
     ERTS_PORT_TASK_OUTPUT,
-    ERTS_PORT_TASK_EVENT,
     ERTS_PORT_TASK_TIMEOUT,
     ERTS_PORT_TASK_DIST_CMD,
     ERTS_PORT_TASK_PROC_SIG
 } ErtsPortTaskType;
 
-#ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
-/* NOTE: Do not access any of the exported variables directly */
-extern erts_smp_atomic_t erts_port_task_outstanding_io_tasks;
-#endif
-
 #define ERTS_PTS_FLG_IN_RUNQ			(((erts_aint32_t) 1) <<  0)
 #define ERTS_PTS_FLG_EXEC			(((erts_aint32_t) 1) <<  1)
 #define ERTS_PTS_FLG_HAVE_TASKS			(((erts_aint32_t) 1) <<  2)
@@ -98,8 +92,8 @@ extern erts_smp_atomic_t erts_port_task_outstanding_io_tasks;
 
 typedef struct {
     ErlDrvSizeT high;
-    erts_smp_atomic_t low;
-    erts_smp_atomic_t size;
+    erts_atomic_t low;
+    erts_atomic_t size;
 }  ErtsPortTaskBusyPortQ;
 
 typedef struct ErtsPortTask_ ErtsPortTask;
@@ -124,10 +118,8 @@ typedef struct {
 	} in;
 	ErtsPortTaskBusyPortQ *bpq;
     } taskq;
-    erts_smp_atomic32_t flags;
-#ifdef ERTS_SMP
+    erts_atomic32_t flags;
     erts_mtx_t mtx;
-#endif
 } ErtsPortTaskSched;
 
 ERTS_GLB_INLINE void erts_port_task_handle_init(ErtsPortTaskHandle *pthp);
@@ -142,22 +134,18 @@ ERTS_GLB_INLINE void erts_port_task_sched_unlock(ErtsPortTaskSched *ptsp);
 ERTS_GLB_INLINE int erts_port_task_sched_lock_is_locked(ErtsPortTaskSched *ptsp);
 ERTS_GLB_INLINE void erts_port_task_sched_enter_exiting_state(ErtsPortTaskSched *ptsp);
 
-#ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
-ERTS_GLB_INLINE int erts_port_task_have_outstanding_io_tasks(void);
-#endif
-
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
 ERTS_GLB_INLINE void
 erts_port_task_handle_init(ErtsPortTaskHandle *pthp)
 {
-    erts_smp_atomic_init_nob(pthp, (erts_aint_t) NULL);
+    erts_atomic_init_nob(pthp, (erts_aint_t) NULL);
 }
 
 ERTS_GLB_INLINE int
 erts_port_task_is_scheduled(ErtsPortTaskHandle *pthp)
 {
-    return ((void *) erts_smp_atomic_read_acqb(pthp)) != NULL;
+    return ((void *) erts_atomic_read_acqb(pthp)) != NULL;
 }
 
 ERTS_GLB_INLINE void erts_port_task_pre_init_sched(ErtsPortTaskSched *ptsp,
@@ -165,9 +153,9 @@ ERTS_GLB_INLINE void erts_port_task_pre_init_sched(ErtsPortTaskSched *ptsp,
 {
     if (bpq) {
 	erts_aint_t low = (erts_aint_t) ERTS_PORT_TASK_DEFAULT_BUSY_PORT_Q_LOW;
-	erts_smp_atomic_init_nob(&bpq->low, low);
+	erts_atomic_init_nob(&bpq->low, low);
  	bpq->high = (ErlDrvSizeT) ERTS_PORT_TASK_DEFAULT_BUSY_PORT_Q_HIGH;
-	erts_smp_atomic_init_nob(&bpq->size, (erts_aint_t) 0);
+	erts_atomic_init_nob(&bpq->size, (erts_aint_t) 0);
     }
     ptsp->taskq.bpq = bpq;
 }
@@ -175,9 +163,7 @@ ERTS_GLB_INLINE void erts_port_task_pre_init_sched(ErtsPortTaskSched *ptsp,
 ERTS_GLB_INLINE void
 erts_port_task_init_sched(ErtsPortTaskSched *ptsp, Eterm instr_id)
 {
-#ifdef ERTS_SMP
     char *lock_str = "port_sched_lock";
-#endif
     ptsp->next = NULL;
     ptsp->taskq.local.busy.first = NULL;
     ptsp->taskq.local.busy.last = NULL;
@@ -186,32 +172,26 @@ erts_port_task_init_sched(ErtsPortTaskSched *ptsp, Eterm instr_id)
     ptsp->taskq.local.first = NULL;
     ptsp->taskq.in.first = NULL;
     ptsp->taskq.in.last = NULL;
-    erts_smp_atomic32_init_nob(&ptsp->flags, 0);
-#ifdef ERTS_SMP
+    erts_atomic32_init_nob(&ptsp->flags, 0);
     erts_mtx_init(&ptsp->mtx, lock_str, instr_id, ERTS_LOCK_FLAGS_CATEGORY_IO);
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_port_task_sched_lock(ErtsPortTaskSched *ptsp)
 {
-#ifdef ERTS_SMP
     erts_mtx_lock(&ptsp->mtx);
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_port_task_sched_unlock(ErtsPortTaskSched *ptsp)
 {
-#ifdef ERTS_SMP
     erts_mtx_unlock(&ptsp->mtx);
-#endif
 }
 
 ERTS_GLB_INLINE int
 erts_port_task_sched_lock_is_locked(ErtsPortTaskSched *ptsp)
 {
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     return erts_lc_mtx_is_locked(&ptsp->mtx);
 #else
     return 0;
@@ -222,35 +202,25 @@ erts_port_task_sched_lock_is_locked(ErtsPortTaskSched *ptsp)
 ERTS_GLB_INLINE void
 erts_port_task_fini_sched(ErtsPortTaskSched *ptsp)
 {
-#ifdef ERTS_SMP
     erts_mtx_destroy(&ptsp->mtx);
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_port_task_sched_enter_exiting_state(ErtsPortTaskSched *ptsp)
 {
-    erts_smp_atomic32_read_bor_nob(&ptsp->flags, ERTS_PTS_FLG_EXITING);
-}
-
-#ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
-
-ERTS_GLB_INLINE int
-erts_port_task_have_outstanding_io_tasks(void)
-{
-    return (erts_smp_atomic_read_acqb(&erts_port_task_outstanding_io_tasks)
-	    != 0);
+    erts_atomic32_read_bor_nob(&ptsp->flags, ERTS_PTS_FLG_EXITING);
 }
 
-#endif /* ERTS_INCLUDE_SCHEDULER_INTERNALS */
-
 #endif
 
 #ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
-int erts_port_task_execute(ErtsRunQueue *, Port **);
+void erts_port_task_execute(ErtsRunQueue *, Port **);
 void erts_port_task_init(void);
 #endif
 
+/* generated for 'port_task' quick allocator */
+void erts_port_task_pre_alloc_init_thread(void);
+
 void erts_port_task_tmp_handle_detach(ErtsPortTaskHandle *);
 int erts_port_task_abort(ErtsPortTaskHandle *);
 
@@ -265,10 +235,8 @@ ErtsProc2PortSigData *erts_port_task_alloc_p2p_sig_data(void);
 ErtsProc2PortSigData *erts_port_task_alloc_p2p_sig_data_extra(size_t extra, void **extra_ptr);
 void erts_port_task_free_p2p_sig_data(ErtsProc2PortSigData *sigdp);
 
-#ifdef ERTS_SMP
 void erts_enqueue_port(ErtsRunQueue *rq, Port *pp);
 Port *erts_dequeue_port(ErtsRunQueue *rq);
-#endif
 #undef ERTS_INCLUDE_SCHEDULER_INTERNALS
 #endif /* ERL_PORT_TASK_H__ */
 #endif /* ERTS_PORT_TASK_ONLY_BASIC_TYPES__ */
diff --git a/erts/emulator/beam/erl_posix_str.c b/erts/emulator/beam/erl_posix_str.c
index deb7e3e173..7b3e640d3f 100644
--- a/erts/emulator/beam/erl_posix_str.c
+++ b/erts/emulator/beam/erl_posix_str.c
@@ -156,6 +156,9 @@ erl_errno_id(error)
 #ifdef EFAULT
     case EFAULT: return "efault";
 #endif
+#ifdef EFTYPE
+    case EFTYPE: return "eftype";
+#endif
 #ifdef EFBIG
     case EFBIG: return "efbig";
 #endif
@@ -351,6 +354,9 @@ erl_errno_id(error)
 #if defined(EOPNOTSUPP) && (!defined(ENOTSUP) || (EOPNOTSUPP != ENOTSUP))
     case EOPNOTSUPP: return "eopnotsupp";
 #endif
+#ifdef EOVERFLOW
+    case EOVERFLOW: return "eoverflow";
+#endif
 #ifdef EPERM
     case EPERM: return "eperm";
 #endif
diff --git a/erts/emulator/beam/erl_printf_term.c b/erts/emulator/beam/erl_printf_term.c
index e6f8460164..910f241a3a 100644
--- a/erts/emulator/beam/erl_printf_term.c
+++ b/erts/emulator/beam/erl_printf_term.c
@@ -532,14 +532,13 @@ print_term(fmtfn_t fn, void* arg, Eterm obj, long *dcount) {
 		Atom* module = atom_tab(atom_val(ep->info.mfa.module));
 		Atom* name = atom_tab(atom_val(ep->info.mfa.function));
 
-		PRINT_STRING(res, fn, arg, "#Fun<");
+		PRINT_STRING(res, fn, arg, "fun ");
 		PRINT_BUF(res, fn, arg, module->name, module->len);
-		PRINT_CHAR(res, fn, arg, '.');
+		PRINT_CHAR(res, fn, arg, ':');
 		PRINT_BUF(res, fn, arg, name->name, name->len);
-		PRINT_CHAR(res, fn, arg, '.');
+		PRINT_CHAR(res, fn, arg, '/');
 		PRINT_SWORD(res, fn, arg, 'd', 0, 1,
 			    (ErlPfSWord) ep->info.mfa.arity);
-		PRINT_CHAR(res, fn, arg, '>');
 	    }
 	    break;
 	case FUN_DEF:
diff --git a/erts/emulator/beam/erl_proc_sig_queue.c b/erts/emulator/beam/erl_proc_sig_queue.c
new file mode 100644
index 0000000000..1aa390d94d
--- /dev/null
+++ b/erts/emulator/beam/erl_proc_sig_queue.c
@@ -0,0 +1,4825 @@
+/*
+ * %CopyrightBegin%
+ * 
+ * Copyright Ericsson AB 2018. All Rights Reserved.
+ * 
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * 
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description:	Process signal queue implementation.
+ *
+ * Author: 	Rickard Green
+ */
+
+
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#include "sys.h"
+#include "global.h"
+#include "dist.h"
+#include "erl_process.h"
+#include "erl_port_task.h"
+#include "erl_trace.h"
+#include "beam_bp.h"
+#include "big.h"
+#include "erl_gc.h"
+#include "bif.h"
+#include "erl_bif_unique.h"
+#include "erl_proc_sig_queue.h"
+#include "dtrace-wrapper.h"
+
+#define ERTS_SIG_REDS_CNT_FACTOR 4
+#define ERTS_PROC_SIG_TRACE_COUNT_LIMIT 200
+
+/*
+ * Note that not all signal are handled using this functionality!
+ */
+
+#define ERTS_SIG_Q_OP_MAX 13
+
+#define ERTS_SIG_Q_OP_EXIT                      0
+#define ERTS_SIG_Q_OP_EXIT_LINKED               1
+#define ERTS_SIG_Q_OP_MONITOR_DOWN              2
+#define ERTS_SIG_Q_OP_MONITOR                   3
+#define ERTS_SIG_Q_OP_DEMONITOR                 4
+#define ERTS_SIG_Q_OP_LINK                      5
+#define ERTS_SIG_Q_OP_UNLINK                    6
+#define ERTS_SIG_Q_OP_GROUP_LEADER              7
+#define ERTS_SIG_Q_OP_TRACE_CHANGE_STATE        8
+#define ERTS_SIG_Q_OP_PERSISTENT_MON_MSG        9
+#define ERTS_SIG_Q_OP_IS_ALIVE                  10
+#define ERTS_SIG_Q_OP_PROCESS_INFO              11
+#define ERTS_SIG_Q_OP_SYNC_SUSPEND              12
+#define ERTS_SIG_Q_OP_RPC                       ERTS_SIG_Q_OP_MAX
+
+#define ERTS_SIG_Q_TYPE_MAX (ERTS_MON_LNK_TYPE_MAX + 5)
+
+#define ERTS_SIG_Q_TYPE_UNDEFINED \
+    (ERTS_MON_LNK_TYPE_MAX + 1)
+#define ERTS_SIG_Q_TYPE_DIST_LINK \
+    (ERTS_MON_LNK_TYPE_MAX + 2)
+#define ERTS_SIG_Q_TYPE_GEN_EXIT \
+    (ERTS_MON_LNK_TYPE_MAX + 3)
+#define ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR \
+    (ERTS_MON_LNK_TYPE_MAX + 4)
+#define ERTS_SIG_Q_TYPE_ADJUST_TRACE_INFO \
+    ERTS_SIG_Q_TYPE_MAX
+
+Process *ERTS_WRITE_UNLIKELY(erts_dirty_process_signal_handler);
+Process *ERTS_WRITE_UNLIKELY(erts_dirty_process_signal_handler_high);
+Process *ERTS_WRITE_UNLIKELY(erts_dirty_process_signal_handler_max);
+
+void
+erts_proc_sig_queue_init(void)
+{
+    ERTS_CT_ASSERT(ERTS_SIG_Q_OP_MASK > ERTS_SIG_Q_OP_MAX);
+    ERTS_CT_ASSERT(ERTS_SIG_Q_OP_MSGQ_LEN_OFFS_MARK > ERTS_SIG_Q_OP_MAX);
+    ERTS_CT_ASSERT(ERTS_SIG_Q_TYPE_MASK >= ERTS_SIG_Q_TYPE_MAX);
+}
+
+typedef struct {
+    int active;
+    int procs;
+    struct {
+        int active;
+#if defined(USE_VM_PROBES)
+        int vm_probes;
+        char receiver_name[DTRACE_TERM_BUF_SIZE];
+#endif
+        int receive_trace;
+        int bp_ix;
+        ErtsMessage **next;
+        ErtsTracingEvent *event;
+    } messages;
+} ErtsSigRecvTracing;
+
+typedef struct {
+    Eterm message;
+    Eterm from;
+    Eterm reason;
+    union {
+        Eterm ref;
+        int normal_kills;
+    } u;
+} ErtsExitSignalData;
+
+typedef struct {
+    Eterm message;
+    Eterm key;
+} ErtsPersistMonMsg;
+
+typedef struct {
+    ErtsSignalCommon common;
+    Eterm local; /* internal pid (immediate) */
+    Eterm remote; /* external pid (heap for it follow) */
+    Eterm heap[EXTERNAL_THING_HEAD_SIZE + 1];
+} ErtsSigDistLinkOp;
+
+typedef struct {
+    ErtsSignalCommon common;
+    Uint flags_on;
+    Uint flags_off;
+    Eterm tracer;
+} ErtsSigTraceInfo;
+
+#define ERTS_SIG_GL_FLG_ACTIVE          (((erts_aint_t) 1) << 0)
+#define ERTS_SIG_GL_FLG_RECEIVER        (((erts_aint_t) 1) << 1)
+#define ERTS_SIG_GL_FLG_SENDER          (((erts_aint_t) 1) << 2)
+
+typedef struct {
+    ErtsSignalCommon common;
+    erts_atomic_t flags;
+    Eterm group_leader;
+    Eterm reply_to;
+    Eterm ref;
+    ErlOffHeap oh;
+    Eterm heap[1];
+} ErtsSigGroupLeader;
+
+typedef struct {
+    Eterm message;
+    Eterm requester;
+} ErtsIsAliveRequest;
+
+typedef struct {
+    Eterm message;
+    Eterm requester;
+    int async;
+} ErtsSyncSuspendRequest;
+
+typedef struct {
+    ErtsMonitorSuspend *mon;
+    ErtsMessage *sync;
+} ErtsProcSigPendingSuspend;
+
+typedef struct {
+    ErtsSignalCommon common;
+    Sint refc;
+    Sint delayed_len;
+    Sint len_offset;
+} ErtsProcSigMsgQLenOffsetMarker;
+
+typedef struct {
+    ErtsSignalCommon common;
+    ErtsProcSigMsgQLenOffsetMarker marker;
+    Sint msgq_len_offset;
+    Eterm requester;
+    Eterm ref;
+    ErtsORefThing oref_thing;
+    Uint reserve_size;
+    Uint len;
+    int flags;
+    int item_ix[1]; /* of len size in reality... */
+} ErtsProcessInfoSig;
+
+#define ERTS_PROC_SIG_PI_MSGQ_LEN_IGNORE        ((Sint) -1)
+#define ERTS_PROC_SIG_PI_MSGQ_LEN_SYNC          ((Sint) -2)
+
+typedef struct {
+    ErtsSignalCommon common;
+    Eterm requester;
+    Eterm (*func)(Process *, void *, int *, ErlHeapFragment **);
+    void *arg;
+    Eterm ref;
+    ErtsORefThing oref_thing;
+} ErtsProcSigRPC;
+
+static int handle_msg_tracing(Process *c_p,
+                              ErtsSigRecvTracing *tracing,
+                              ErtsMessage ***next_nm_sig);
+static int handle_trace_change_state(Process *c_p,
+                                     ErtsSigRecvTracing *tracing,
+                                     Uint16 type,
+                                     ErtsMessage *sig,
+                                     ErtsMessage ***next_nm_sig);
+static void getting_unlinked(Process *c_p, Eterm unlinker);
+static void getting_linked(Process *c_p, Eterm linker);
+static void group_leader_reply(Process *c_p, Eterm to,
+                               Eterm ref, int success);
+static int stretch_limit(Process *c_p, ErtsSigRecvTracing *tp,
+                         int abs_lim, int *limp);
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+#define ERTS_PROC_SIG_HDBG_PRIV_CHKQ(P, T, NMN)                 \
+    do {                                                        \
+        ErtsMessage **nm_next__ = *(NMN);                       \
+        ErtsMessage **nm_last__ = (P)->sig_qs.nmsigs.last;      \
+        if (!nm_next__ || !*nm_next__) {                        \
+            nm_next__ = NULL;                                   \
+            nm_last__ = NULL;                                   \
+        }                                                       \
+        proc_sig_hdbg_check_queue((P),                          \
+                                  1,                            \
+                                  &(P)->sig_qs.cont,            \
+                                  (P)->sig_qs.cont_last,        \
+                                  nm_next__,                    \
+                                  nm_last__,                    \
+                                  (T),                          \
+                                  NULL,                         \
+                                  ERTS_PSFLG_FREE);             \
+    } while (0);
+static Sint
+proc_sig_hdbg_check_queue(Process *c_p,
+                          int privq,
+                          ErtsMessage **sig_next,
+                          ErtsMessage **sig_last,
+                          ErtsMessage **sig_nm_next,
+                          ErtsMessage **sig_nm_last,
+                          ErtsSigRecvTracing *tracing,
+                          int *found_saved_last_p,
+                          erts_aint32_t sig_psflg);
+#else
+#define ERTS_PROC_SIG_HDBG_PRIV_CHKQ(P, T, NMN)
+#endif
+
+typedef struct {
+    ErtsSignalCommon common;
+    Eterm ref;
+    Eterm heap[1];
+} ErtsSigDistProcDemonitor;
+
+static void
+destroy_dist_proc_demonitor(ErtsSigDistProcDemonitor *dmon)
+{
+    Eterm ref = dmon->ref;
+    if (is_external(ref)) {
+        ExternalThing *etp = external_thing_ptr(ref);
+        erts_deref_node_entry(etp->node);
+    }
+    erts_free(ERTS_ALC_T_DIST_DEMONITOR, dmon);
+}
+
+static ERTS_INLINE ErtsSigDistLinkOp *
+make_sig_dist_link_op(int op, Eterm local, Eterm remote)
+{
+    Eterm *hp;
+    ErlOffHeap oh = {0};
+    ErtsSigDistLinkOp *sdlnk = erts_alloc(ERTS_ALC_T_SIG_DATA,
+                                          sizeof(ErtsSigDistLinkOp));
+    ASSERT(is_internal_pid(local));
+    ASSERT(is_external_pid(remote));
+
+    hp = &sdlnk->heap[0];
+
+    sdlnk->common.tag = ERTS_PROC_SIG_MAKE_TAG(op,
+                                               ERTS_SIG_Q_TYPE_DIST_LINK,
+                                               0);
+    sdlnk->local = local;
+    sdlnk->remote = STORE_NC(&hp, &oh, remote);
+
+    ASSERT(&sdlnk->heap[0] < hp);
+    ASSERT(hp <= &sdlnk->heap[0] + sizeof(sdlnk->heap)/sizeof(sdlnk->heap[0]));
+    ASSERT(boxed_val(sdlnk->remote) == &sdlnk->heap[0]);
+
+    return sdlnk;
+}
+
+static ERTS_INLINE void
+destroy_sig_dist_link_op(ErtsSigDistLinkOp *sdlnk)
+{
+    ASSERT(is_external_pid(sdlnk->remote));
+    ASSERT(boxed_val(sdlnk->remote) == &sdlnk->heap[0]);
+    erts_deref_node_entry(((ExternalThing *) &sdlnk->heap[0])->node);
+    erts_free(ERTS_ALC_T_SIG_DATA, sdlnk);
+}
+
+static ERTS_INLINE ErtsExitSignalData *
+get_exit_signal_data(ErtsMessage *xsig)
+{
+    ASSERT(ERTS_SIG_IS_NON_MSG(xsig));
+    ASSERT((ERTS_PROC_SIG_OP(((ErtsSignal *) xsig)->common.tag)
+            == ERTS_SIG_Q_OP_EXIT)
+           || (ERTS_PROC_SIG_OP(((ErtsSignal *) xsig)->common.tag)
+                == ERTS_SIG_Q_OP_EXIT_LINKED)
+           || (ERTS_PROC_SIG_OP(((ErtsSignal *) xsig)->common.tag)
+               == ERTS_SIG_Q_OP_MONITOR_DOWN));
+    ASSERT(xsig->hfrag.alloc_size > xsig->hfrag.used_size);
+    ASSERT((xsig->hfrag.alloc_size - xsig->hfrag.used_size)*sizeof(UWord)
+           >= sizeof(ErtsExitSignalData));
+    return (ErtsExitSignalData *) (char *) (&xsig->hfrag.mem[0]
+                                            + xsig->hfrag.used_size);
+}
+
+static ERTS_INLINE void
+destroy_trace_info(ErtsSigTraceInfo *ti)
+{
+    if (is_value(ti->tracer))
+        erts_tracer_update(&ti->tracer, NIL);
+    erts_free(ERTS_ALC_T_SIG_DATA, ti);
+}
+
+static void
+destroy_sig_group_leader(ErtsSigGroupLeader *sgl)
+{
+    erts_cleanup_offheap(&sgl->oh);
+    erts_free(ERTS_ALC_T_SIG_DATA, sgl);
+}
+
+static ERTS_INLINE void
+sig_enqueue_trace(Process *c_p, ErtsMessage **sigp, int op,
+                  Process *rp, ErtsMessage ***last_next)
+{
+    switch (op) {
+    case ERTS_SIG_Q_OP_LINK:
+        if (c_p
+            && ((!!IS_TRACED(c_p))
+                & (ERTS_TRACE_FLAGS(c_p) & (F_TRACE_SOL
+                                            | F_TRACE_SOL1)))) {
+            ErtsSigTraceInfo *ti;
+            Eterm tag;
+            /*
+             * Set on link enabled.
+             *
+             * Prepend a trace-change-state signal before the
+             * link signal...
+             */
+            tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_TRACE_CHANGE_STATE,
+                                         ERTS_SIG_Q_TYPE_ADJUST_TRACE_INFO,
+                                         0);
+            ti = erts_alloc(ERTS_ALC_T_SIG_DATA, sizeof(ErtsSigTraceInfo));
+            ti->common.next = *sigp;
+            ti->common.specific.next = &ti->common.next;
+            ti->common.tag = tag;
+            ti->flags_on = ERTS_TRACE_FLAGS(c_p) & TRACEE_FLAGS;
+            if (!(ti->flags_on & F_TRACE_SOL1))
+                ti->flags_off = 0;
+            else {
+                ti->flags_off = F_TRACE_SOL1|F_TRACE_SOL;
+                erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                ERTS_TRACE_FLAGS(c_p) &= ~(F_TRACE_SOL1|F_TRACE_SOL);
+                erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+            }
+            erts_tracer_update(&ti->tracer, ERTS_TRACER(c_p));
+            *sigp = (ErtsMessage *) ti;
+            if (!*last_next || *last_next == sigp)
+                *last_next = &ti->common.next;
+        }
+        break;
+
+#ifdef USE_VM_PROBES
+    case ERTS_SIG_Q_OP_EXIT:
+    case ERTS_SIG_Q_OP_EXIT_LINKED:
+
+        if (DTRACE_ENABLED(process_exit_signal)) {
+            ErtsMessage* sig = *sigp;
+            Uint16 type = ERTS_PROC_SIG_TYPE(((ErtsSignal *) sig)->common.tag);
+            Eterm reason, from;
+
+            if (type == ERTS_SIG_Q_TYPE_GEN_EXIT) {
+                ErtsExitSignalData *xsigd = get_exit_signal_data(sig);
+                reason = xsigd->reason;
+                from = xsigd->from;
+            }
+            else {
+                ErtsLink *lnk = (ErtsLink *) sig, *olnk;
+
+                ASSERT(type == ERTS_LNK_TYPE_PROC
+                       || type == ERTS_LNK_TYPE_PORT
+                       || type == ERTS_LNK_TYPE_DIST_PROC);
+
+                olnk = erts_link_to_other(lnk, NULL);
+                reason = lnk->other.item;
+                from = olnk->other.item;
+            }
+
+            if (is_pid(from)) {
+
+                DTRACE_CHARBUF(sender_str, DTRACE_TERM_BUF_SIZE);
+                DTRACE_CHARBUF(receiver_str, DTRACE_TERM_BUF_SIZE);
+                DTRACE_CHARBUF(reason_buf, DTRACE_TERM_BUF_SIZE);
+
+                if (reason == am_kill) {
+                    reason = am_killed;
+                }
+
+                dtrace_pid_str(from, sender_str);
+                dtrace_proc_str(rp, receiver_str);
+                erts_snprintf(reason_buf, sizeof(DTRACE_CHARBUF_NAME(reason_buf)) - 1, "%T", reason);
+                DTRACE3(process_exit_signal, sender_str, receiver_str, reason_buf);
+            }
+        }
+        break;
+
+#endif
+
+    default:
+        break;
+    }
+}
+
+static void
+sig_enqueue_trace_cleanup(ErtsMessage *first, ErtsSignal *sig)
+{
+    ErtsMessage *tmp;
+
+    /* The usual case; no tracing signals... */
+    if (sig == (ErtsSignal *) first) {
+        ASSERT(sig->common.next == NULL);
+        return;
+    }
+
+    /* Got trace signals to clean up... */
+
+    tmp = first;
+
+    while (tmp) {
+        ErtsMessage *tmp_free = tmp;
+        tmp = tmp->next;
+        if (sig != (ErtsSignal *) tmp_free) {
+            switch (ERTS_PROC_SIG_OP(((ErtsSignal *) tmp_free)->common.tag)) {
+            case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE:
+                destroy_trace_info((ErtsSigTraceInfo *) tmp_free);
+                break;
+            case ERTS_SIG_Q_OP_MONITOR:
+                break; /* ignore flushed pending signal */
+            default:
+                ERTS_INTERNAL_ERROR("Unexpected signal op");
+                break;
+            }
+        }
+    }
+}
+
+#ifdef DEBUG
+static int dbg_count_nmsigs(ErtsMessage *first)
+{
+    ErtsMessage *sig;
+    int cnt = 0;
+
+    for (sig = first; sig; sig = sig->next) {
+        if (ERTS_SIG_IS_NON_MSG(sig))
+            ++cnt;
+    }
+    return cnt;
+}
+#endif
+
+static ERTS_INLINE erts_aint32_t
+enqueue_signals(Process *rp, ErtsMessage *first,
+                ErtsMessage **last, ErtsMessage **last_next,
+                Uint num_msgs,
+                erts_aint32_t in_state)
+{
+    erts_aint32_t state = in_state;
+    ErtsMessage **this = rp->sig_inq.last;
+
+    ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE(rp);
+
+    ASSERT(!*this);
+    *this = first;
+    rp->sig_inq.last = last;
+
+    if (!rp->sig_inq.nmsigs.next) {
+        ASSERT(!rp->sig_inq.nmsigs.last);
+        if (ERTS_SIG_IS_NON_MSG(first)) {
+            rp->sig_inq.nmsigs.next = this;
+        }
+        else if (last_next) {
+            ASSERT(first->next && ERTS_SIG_IS_NON_MSG(first->next));
+            rp->sig_inq.nmsigs.next = &first->next;
+        }
+        else
+            goto no_nmsig;
+
+        state = erts_atomic32_read_bor_nob(&rp->state,
+                                           ERTS_PSFLG_SIG_IN_Q);
+    no_nmsig:
+        ASSERT(!(state & ERTS_PSFLG_SIG_IN_Q));
+    }
+    else {
+        ErtsSignal *sig;
+        ASSERT(rp->sig_inq.nmsigs.last);
+
+        sig = (ErtsSignal *) *rp->sig_inq.nmsigs.last;
+
+        ASSERT(sig && !sig->common.specific.next);
+        ASSERT(state & ERTS_PSFLG_SIG_IN_Q);
+        if (ERTS_SIG_IS_NON_MSG(first)) {
+            sig->common.specific.next = this;
+        }
+        else if (last_next) {
+            ASSERT(first->next && ERTS_SIG_IS_NON_MSG(first->next));
+            sig->common.specific.next = &first->next;
+        }
+    }
+
+    if (last_next) {
+        ASSERT(dbg_count_nmsigs(first) >= 2);
+        rp->sig_inq.nmsigs.last = last_next;
+    }
+    else if (ERTS_SIG_IS_NON_MSG(first)) {
+        ASSERT(dbg_count_nmsigs(first) == 1);
+        rp->sig_inq.nmsigs.last = this;
+    }
+    else
+        ASSERT(dbg_count_nmsigs(first) == 0);
+
+    rp->sig_inq.len += num_msgs;
+
+    ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE(rp);
+
+    return state;
+}
+
+erts_aint32_t erts_enqueue_signals(Process *rp, ErtsMessage *first,
+                                   ErtsMessage **last, ErtsMessage **last_next,
+                                   Uint num_msgs,
+                                   erts_aint32_t in_state)
+{
+    return enqueue_signals(rp, first, last, last_next, num_msgs, in_state);
+}
+
+void
+erts_make_dirty_proc_handled(Eterm pid,
+                             erts_aint32_t state,
+                             erts_aint32_t prio)
+{
+    Eterm *hp;
+    ErtsMessage *mp;
+    Process *sig_handler;
+
+    ASSERT(state & (ERTS_PSFLG_DIRTY_RUNNING |
+                    ERTS_PSFLG_DIRTY_RUNNING_SYS));
+
+    if (prio < 0)
+        prio = (int) ERTS_PSFLGS_GET_USR_PRIO(state);
+
+    switch (prio) {
+    case PRIORITY_MAX:
+        sig_handler = erts_dirty_process_signal_handler_max;
+        break;
+    case PRIORITY_HIGH:
+        sig_handler = erts_dirty_process_signal_handler_high;
+        break;
+    default:
+        sig_handler = erts_dirty_process_signal_handler;
+        break;
+    }
+
+    /* Make sure signals are handled... */
+    mp = erts_alloc_message(0, &hp);
+    erts_queue_message(sig_handler, 0, mp, pid, am_system);
+}
+
+static void
+check_push_msgq_len_offs_marker(Process *rp, ErtsSignal *sig);
+
+
+static int
+proc_queue_signal(Process *c_p, Eterm pid, ErtsSignal *sig, int op)
+{
+    int res;
+    Process *rp;
+    ErtsMessage *first, *last, **last_next, **sigp;
+    ErtsSchedulerData *esdp = erts_get_scheduler_data();
+    int is_normal_sched = !!esdp && esdp->type == ERTS_SCHED_NORMAL;
+    erts_aint32_t state;
+    ErtsSignal *pend_sig;
+
+    if (is_normal_sched) {
+        pend_sig = esdp->pending_signal.sig;
+        if (op == ERTS_SIG_Q_OP_MONITOR
+            && ((ErtsMonitor*)sig)->type == ERTS_MON_TYPE_PROC) {
+
+            if (!pend_sig) {
+                esdp->pending_signal.sig = sig;
+                esdp->pending_signal.to = pid;
+#ifdef DEBUG
+                esdp->pending_signal.dbg_from = esdp->current_process;
+#endif
+                return 1;
+            }
+            ASSERT(esdp->pending_signal.dbg_from == esdp->current_process);
+            if (pend_sig != sig) {
+                /* Switch them and send previously pending signal instead */
+                Eterm pend_to = esdp->pending_signal.to;
+                esdp->pending_signal.sig = sig;
+                esdp->pending_signal.to = pid;
+                sig = pend_sig;
+                pid = pend_to;
+            }
+            else {
+                /* Caller wants to flush pending signal */
+                ASSERT(pid == esdp->pending_signal.to);
+                esdp->pending_signal.sig = NULL;
+                esdp->pending_signal.to = THE_NON_VALUE;
+#ifdef DEBUG
+                esdp->pending_signal.dbg_from = NULL;
+#endif
+                pend_sig = NULL;
+            }
+            rp = erts_proc_lookup_raw(pid);
+            if (!rp) {
+                erts_proc_sig_send_monitor_down((ErtsMonitor*)sig, am_noproc);
+                return 1;
+            }
+        }
+        else if (pend_sig && pid == esdp->pending_signal.to) {
+            /* Flush pending signal to maintain signal order */
+            esdp->pending_signal.sig = NULL;
+            esdp->pending_signal.to = THE_NON_VALUE;
+
+            rp = erts_proc_lookup_raw(pid);
+            if (!rp) {
+                erts_proc_sig_send_monitor_down((ErtsMonitor*)pend_sig, am_noproc);
+                return 0;
+            }
+
+            /* Prepend pending signal */
+            pend_sig->common.next = (ErtsMessage*) sig;
+            pend_sig->common.specific.next = &pend_sig->common.next;
+            first = (ErtsMessage*) pend_sig;
+            last = (ErtsMessage*) sig;
+            sigp = last_next = &pend_sig->common.next;
+            goto first_last_done; 
+        }
+        else {
+            pend_sig = NULL;
+            rp = erts_proc_lookup_raw(pid);
+            if (!rp)
+                return 0;
+        }
+    }
+    else {
+        rp = erts_proc_lookup_raw_inc_refc(pid);
+        if (!rp)
+            return 0;
+        pend_sig = NULL;
+    }
+
+    first = last = (ErtsMessage *) sig;
+    last_next = NULL;
+    sigp = &first;
+
+first_last_done:
+    sig->common.specific.next = NULL;
+
+    /* may add signals before sig */
+    sig_enqueue_trace(c_p, sigp, op, rp, &last_next);
+
+    last->next = NULL;
+
+    erts_proc_lock(rp, ERTS_PROC_LOCK_MSGQ);
+
+    state = erts_atomic32_read_nob(&rp->state);
+
+    if (ERTS_PSFLG_FREE & state)
+        res = 0;
+    else {
+        state = enqueue_signals(rp, first, &last->next, last_next, 0, state);
+        if (ERTS_UNLIKELY(op == ERTS_SIG_Q_OP_PROCESS_INFO))
+            check_push_msgq_len_offs_marker(rp, sig);
+        res = !0;
+    }
+
+    erts_proc_unlock(rp, ERTS_PROC_LOCK_MSGQ);
+
+    if (res == 0) {
+        sig_enqueue_trace_cleanup(first, sig);
+        if (pend_sig) {
+            erts_proc_sig_send_monitor_down((ErtsMonitor*)pend_sig, am_noproc);
+            if (sig == pend_sig) {
+                /* We did a switch, callers signal is now pending (still ok) */
+                ASSERT(esdp->pending_signal.sig);
+                res = 1;
+            }
+        }
+    }
+    else
+        erts_proc_notify_new_sig(rp, state, 0);
+
+    if (!is_normal_sched)
+        erts_proc_dec_refc(rp);
+
+    return res;
+}
+
+void erts_proc_sig_send_pending(ErtsSchedulerData* esdp)
+{
+    ErtsSignal* sig = esdp->pending_signal.sig;
+    int op;
+
+    ASSERT(esdp && esdp->type == ERTS_SCHED_NORMAL);
+    ASSERT(sig);
+    ASSERT(is_internal_pid(esdp->pending_signal.to));
+
+    op = ERTS_SIG_Q_OP_MONITOR;
+    ASSERT(op == ERTS_PROC_SIG_OP(sig->common.tag));
+
+    if (!proc_queue_signal(NULL, esdp->pending_signal.to, sig, op)) {
+        ErtsMonitor* mon = (ErtsMonitor*)sig;
+        erts_proc_sig_send_monitor_down(mon, am_noproc);
+    }
+}
+
+static int
+maybe_elevate_sig_handling_prio(Process *c_p, Eterm other)
+{
+    /*
+     * returns:
+     *  > 0 -> elevated prio; process alive or exiting
+     *  < 0 -> no elevation needed; process alive or exiting
+     *    0 -> process terminated (free)
+     */
+    int res;
+    Process *rp;
+    erts_aint32_t state, my_prio, other_prio;
+
+    rp = erts_proc_lookup_raw(other);
+    if (!rp)
+        res = 0;
+    else {
+        res = -1;
+        state = erts_atomic32_read_nob(&c_p->state);
+        my_prio = ERTS_PSFLGS_GET_USR_PRIO(state);
+
+        state = erts_atomic32_read_nob(&rp->state);
+        other_prio = ERTS_PSFLGS_GET_USR_PRIO(state);
+
+        if (other_prio > my_prio) {
+            /* Others prio is lower than mine; elevate it... */
+            res = !!erts_sig_prio(other, my_prio);
+            if (res) {
+                /* ensure handled if dirty executing... */
+                state = erts_atomic32_read_nob(&rp->state);
+                if (state & (ERTS_PSFLG_DIRTY_RUNNING
+                             | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
+                    erts_make_dirty_proc_handled(other, state, my_prio);
+                }
+            }
+        }
+    }
+    return res;
+}
+
+void
+erts_proc_sig_fetch__(Process *proc)
+{
+    ASSERT(proc->sig_inq.first);
+
+    if (!proc->sig_inq.nmsigs.next) {
+        ASSERT(!(ERTS_PSFLG_SIG_IN_Q
+                 & erts_atomic32_read_nob(&proc->state)));
+        ASSERT(!proc->sig_inq.nmsigs.last);
+
+        if (proc->sig_qs.cont || ERTS_MSG_RECV_TRACED(proc)) {
+            *proc->sig_qs.cont_last = proc->sig_inq.first;
+            proc->sig_qs.cont_last = proc->sig_inq.last;
+        }
+        else {
+            *proc->sig_qs.last = proc->sig_inq.first;
+            proc->sig_qs.last = proc->sig_inq.last;
+        }
+    }
+    else {
+        erts_aint32_t s;
+        ASSERT(proc->sig_inq.nmsigs.last);
+         if (!proc->sig_qs.nmsigs.last) {
+            ASSERT(!proc->sig_qs.nmsigs.next);
+            if (proc->sig_inq.nmsigs.next == &proc->sig_inq.first)
+                proc->sig_qs.nmsigs.next = proc->sig_qs.cont_last;
+            else
+                proc->sig_qs.nmsigs.next = proc->sig_inq.nmsigs.next;
+
+            s = erts_atomic32_read_bset_nob(&proc->state,
+                                            (ERTS_PSFLG_SIG_Q
+                                             | ERTS_PSFLG_SIG_IN_Q),
+                                            ERTS_PSFLG_SIG_Q);
+
+            ASSERT((s & (ERTS_PSFLG_SIG_Q|ERTS_PSFLG_SIG_IN_Q))
+                   == ERTS_PSFLG_SIG_IN_Q); (void)s;
+        }
+        else {
+            ErtsSignal *sig;
+            ASSERT(proc->sig_qs.nmsigs.next);
+            sig = ((ErtsSignal *) *proc->sig_qs.nmsigs.last);
+            ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+            ASSERT(!sig->common.specific.next);
+            if (proc->sig_inq.nmsigs.next == &proc->sig_inq.first)
+                sig->common.specific.next = proc->sig_qs.cont_last;
+            else
+                sig->common.specific.next = proc->sig_inq.nmsigs.next;
+
+            s = erts_atomic32_read_band_nob(&proc->state,
+                                            ~ERTS_PSFLG_SIG_IN_Q);
+
+            ASSERT((s & (ERTS_PSFLG_SIG_Q|ERTS_PSFLG_SIG_IN_Q))
+                   == (ERTS_PSFLG_SIG_Q|ERTS_PSFLG_SIG_IN_Q)); (void)s;
+        }
+        if (proc->sig_inq.nmsigs.last == &proc->sig_inq.first)
+            proc->sig_qs.nmsigs.last = proc->sig_qs.cont_last;
+        else
+            proc->sig_qs.nmsigs.last = proc->sig_inq.nmsigs.last;
+        proc->sig_inq.nmsigs.next = NULL;
+        proc->sig_inq.nmsigs.last = NULL;
+
+        *proc->sig_qs.cont_last = proc->sig_inq.first;
+        proc->sig_qs.cont_last = proc->sig_inq.last;
+    }
+
+    proc->sig_qs.len += proc->sig_inq.len;
+
+    proc->sig_inq.first = NULL;
+    proc->sig_inq.last = &proc->sig_inq.first;
+    proc->sig_inq.len = 0;
+
+}
+
+Sint
+erts_proc_sig_fetch_msgq_len_offs__(Process *proc)
+{
+    ErtsProcSigMsgQLenOffsetMarker *marker
+        = (ErtsProcSigMsgQLenOffsetMarker *) proc->sig_inq.first;
+
+    ASSERT(marker->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK);
+
+    if (marker->common.next) {
+        Sint len;
+
+        proc->flags |= F_DELAYED_PSIGQS_LEN;
+
+        /*
+         * Prevent update of sig_qs.len in fetch. These
+         * updates are done via process-info signal(s)
+         * instead...
+         */
+        len = proc->sig_inq.len;
+        marker->delayed_len += len;
+        marker->len_offset -= len;
+        proc->sig_inq.len = 0;
+
+        /*
+         * Temorarily remove marker during fetch...
+         */
+
+        proc->sig_inq.first = marker->common.next;
+        if (proc->sig_inq.last == &marker->common.next)
+            proc->sig_inq.last = &proc->sig_inq.first;
+        if (proc->sig_inq.nmsigs.next == &marker->common.next)
+            proc->sig_inq.nmsigs.next = &proc->sig_inq.first;
+        if (proc->sig_inq.nmsigs.last == &marker->common.next)
+            proc->sig_inq.nmsigs.last = &proc->sig_inq.first;
+
+        erts_proc_sig_fetch__(proc);
+
+        marker->common.next = NULL;
+        proc->sig_inq.first = (ErtsMessage *) marker;
+        proc->sig_inq.last = &marker->common.next;
+
+    }
+
+    return marker->delayed_len;
+}
+
+static ERTS_INLINE Sint
+proc_sig_privqs_len(Process *c_p, int have_qlock)
+{
+    Sint res = c_p->sig_qs.len;
+
+    ERTS_LC_ASSERT(!have_qlock
+                   ? (ERTS_PROC_LOCK_MAIN
+                      == erts_proc_lc_my_proc_locks(c_p))
+                   : ((ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_MAIN)
+                      == ((ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_MAIN)
+                          & erts_proc_lc_my_proc_locks(c_p))));
+
+    if (c_p->flags & F_DELAYED_PSIGQS_LEN) {
+        ErtsProcSigMsgQLenOffsetMarker *marker;
+
+        if (!have_qlock)
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+
+        marker = (ErtsProcSigMsgQLenOffsetMarker *) c_p->sig_inq.first;
+        ASSERT(marker);
+        ASSERT(marker->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK);
+
+        res += marker->delayed_len;
+
+        if (!have_qlock)
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+    }
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+    {
+        Sint len = 0;
+        ERTS_FOREACH_SIG_PRIVQS(
+            c_p, mp,
+            {
+                if (ERTS_SIG_IS_MSG(mp))
+                    len++;
+            });
+        ERTS_ASSERT(res == len);
+    }
+#endif
+
+    return res;
+}
+
+Sint
+erts_proc_sig_privqs_len(Process *c_p)
+{
+    return proc_sig_privqs_len(c_p, 0);
+}
+
+static void do_seq_trace_output(Eterm to, Eterm token, Eterm msg);
+
+static void
+send_gen_exit_signal(Process *c_p, Eterm from_tag,
+                     Eterm from, Eterm to,
+                     Sint16 op, Eterm reason, Eterm ref,
+                     Eterm token, int normal_kills)
+{
+    ErtsExitSignalData *xsigd;
+    Eterm *hp, *start_hp, s_reason, s_ref, s_message, s_token, s_from;
+    ErtsMessage *mp;
+    ErlHeapFragment *hfrag;
+    ErlOffHeap *ohp;
+    Uint hsz, from_sz, reason_sz, ref_sz, token_sz;
+    int seq_trace;
+#ifdef USE_VM_PROBES
+    Eterm s_utag, utag;
+    Uint utag_sz;
+#endif
+
+    ASSERT(is_immed(from_tag));
+
+    hsz = sizeof(ErtsExitSignalData)/sizeof(Uint);
+
+    seq_trace = c_p && have_seqtrace(token);
+    if (seq_trace)
+        seq_trace_update_send(c_p);
+
+#ifdef USE_VM_PROBES
+    utag_sz = 0;
+    utag = NIL;
+    if (c_p && token != NIL && (DT_UTAG_FLAGS(c_p) & DT_UTAG_SPREADING)) {
+        utag_sz = size_object(DT_UTAG(c_p));
+        utag = DT_UTAG(c_p);
+    }
+    else if (token == am_have_dt_utag) {
+        token = NIL;
+    }
+    hsz += utag_sz;
+#endif
+
+    token_sz = is_immed(token) ? 0 : size_object(token);
+    hsz += token_sz;
+
+    from_sz = is_immed(from) ? 0 : size_object(from);
+    hsz += from_sz;
+
+    reason_sz = is_immed(reason) ? 0 : size_object(reason);
+    hsz += reason_sz;
+
+    switch (op) {
+    case ERTS_SIG_Q_OP_EXIT:
+    case ERTS_SIG_Q_OP_EXIT_LINKED: {
+        /* {'EXIT', From, Reason} */
+        hsz += 4; /* 3-tuple */
+        ref_sz = 0;
+        break;
+    }
+    case ERTS_SIG_Q_OP_MONITOR_DOWN: {
+        /* {'DOWN', Ref, process, From, Reason} */
+        hsz += 6; /* 5-tuple */
+        ref_sz = NC_HEAP_SIZE(ref);
+        hsz += ref_sz;
+        break;
+    }
+    default:
+        ERTS_INTERNAL_ERROR("Invalid exit signal op");
+        break;
+    }
+
+    /*
+     * Allocate message combined with heap fragment...
+     */
+    mp = erts_alloc_message(hsz, &hp);
+    hfrag = &mp->hfrag;
+    mp->next = NULL;
+    ohp = &hfrag->off_heap;
+    start_hp = hp;
+
+    s_token = (is_immed(token)
+                ? token
+                : copy_struct(token, token_sz, &hp, ohp));
+
+    s_reason = (is_immed(reason)
+                ? reason
+                : copy_struct(reason, reason_sz, &hp, ohp));
+
+    s_from = (is_immed(from)
+              ? from
+              : copy_struct(from, from_sz, &hp, ohp));
+
+    if (!ref_sz)
+        s_ref = NIL;
+    else
+        s_ref = STORE_NC(&hp, ohp, ref);
+
+    switch (op) {
+    case ERTS_SIG_Q_OP_EXIT:
+    case ERTS_SIG_Q_OP_EXIT_LINKED:
+        /* {'EXIT', From, Reason} */
+        s_message = TUPLE3(hp, am_EXIT, s_from, s_reason);
+        hp += 4;
+        break;
+    case ERTS_SIG_Q_OP_MONITOR_DOWN:
+        /* {'DOWN', Ref, process, From, Reason} */
+        s_message = TUPLE5(hp, am_DOWN, s_ref, am_process, s_from, s_reason);
+        hp += 6;
+        break;
+    }
+
+#ifdef USE_VM_PROBES
+    s_utag = (is_immed(utag)
+              ? utag
+              : copy_struct(utag, utag_sz, &hp, ohp));
+    ERL_MESSAGE_DT_UTAG(mp) = s_utag;
+#endif
+
+    ERL_MESSAGE_TERM(mp) = ERTS_PROC_SIG_MAKE_TAG(op,
+                                                  ERTS_SIG_Q_TYPE_GEN_EXIT,
+                                                  0);
+    ERL_MESSAGE_TOKEN(mp) = s_token;
+    ERL_MESSAGE_FROM(mp) = from_tag; /* immediate... */
+
+    hfrag->used_size = hp - start_hp;
+
+    xsigd = (ErtsExitSignalData *) (char *) hp;
+
+    xsigd->message = s_message;
+    xsigd->from = s_from;
+    xsigd->reason = s_reason;
+    if (is_nil(s_ref))
+        xsigd->u.normal_kills = normal_kills;
+    else {
+        ASSERT(is_ref(s_ref));
+        xsigd->u.ref = s_ref;
+    }
+
+    if (seq_trace)
+        do_seq_trace_output(to, s_token, s_message);
+
+    if (!proc_queue_signal(c_p, to, (ErtsSignal *) mp, op)) {
+        mp->next = NULL;
+        erts_cleanup_messages(mp);
+    }
+}
+
+static void
+do_seq_trace_output(Eterm to, Eterm token, Eterm msg)
+{
+    /*
+     * We could do this when enqueuing the signal and avoid some
+     * locking. However, the enqueuing code would then always
+     * have the penalty of this seq-tracing code which we do not
+     * want...
+     */
+    ErtsSchedulerData *esdp = erts_get_scheduler_data();
+    int is_normal_sched = !!esdp && esdp->type == ERTS_SCHED_NORMAL;
+    Process *rp;
+
+    if (is_normal_sched)
+        rp = erts_proc_lookup_raw(to);
+    else
+        rp = erts_proc_lookup_raw_inc_refc(to);
+
+    if (rp) {
+        erts_proc_lock(rp, ERTS_PROC_LOCK_MSGQ);
+
+        if (!ERTS_PROC_IS_EXITING(rp))
+            seq_trace_output(token, msg, SEQ_TRACE_SEND, to, rp);
+
+        erts_proc_unlock(rp, ERTS_PROC_LOCK_MSGQ);
+
+        if (!is_normal_sched)
+            erts_proc_dec_refc(rp);
+    }
+}
+
+void
+erts_proc_sig_send_persistent_monitor_msg(Uint16 type, Eterm key,
+                                          Eterm from, Eterm to,
+                                          Eterm msg, Uint msg_sz)
+{
+    ErtsPersistMonMsg *prst_mon;
+    ErtsMessage *mp;
+    ErlHeapFragment *hfrag;
+    Eterm *hp, *start_hp, message;
+    ErlOffHeap *ohp;
+    Uint hsz = sizeof(ErtsPersistMonMsg) + msg_sz;
+
+    /*
+     * Allocate message combined with heap fragment...
+     */
+    mp = erts_alloc_message(hsz, &hp);
+    hfrag = &mp->hfrag;
+    mp->next = NULL;
+    ohp = &hfrag->off_heap;
+    start_hp = hp;
+
+    ASSERT(msg_sz == size_object(msg));
+    message = copy_struct(msg, msg_sz, &hp, ohp);
+    hfrag->used_size = hp - start_hp;
+
+    prst_mon = (ErtsPersistMonMsg *) (char *) hp;
+    prst_mon->message = message;
+
+    switch (type) {
+    case ERTS_MON_TYPE_NODES:
+        ASSERT(is_small(key));
+        prst_mon->key = key;
+        break;
+
+    case ERTS_MON_TYPE_TIME_OFFSET:
+        ASSERT(is_internal_ref(key));
+        ASSERT(is_tuple_arity(message, 5));
+
+        prst_mon->key = tuple_val(message)[2];
+
+        ASSERT(eq(prst_mon->key, key));
+        break;
+
+    default:
+        ERTS_INTERNAL_ERROR("Invalid persistent monitor type");
+        prst_mon->key = key;
+        break;
+    }
+
+    ASSERT(is_immed(from));
+
+    ERL_MESSAGE_TERM(mp) = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_PERSISTENT_MON_MSG,
+                                                  type, 0);
+    ERL_MESSAGE_FROM(mp) = from;
+    ERL_MESSAGE_TOKEN(mp) = NIL;
+#ifdef USE_VM_PROBES
+    ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+
+    if (!proc_queue_signal(NULL, to, (ErtsSignal *) mp,
+                           ERTS_SIG_Q_OP_PERSISTENT_MON_MSG)) {
+        mp->next = NULL;
+        erts_cleanup_messages(mp);
+    }
+}
+
+static ERTS_INLINE Eterm
+get_persist_mon_msg(ErtsMessage *sig, Eterm *msg)
+{
+    ErtsPersistMonMsg *prst_mon;
+    prst_mon = ((ErtsPersistMonMsg *)
+                (char *) (&sig->hfrag.mem[0]
+                          + sig->hfrag.used_size));
+    *msg = prst_mon->message;
+    return prst_mon->key;
+}
+
+void
+erts_proc_sig_send_exit(Process *c_p, Eterm from, Eterm to,
+                        Eterm reason, Eterm token,
+                        int normal_kills)
+{
+    Eterm from_tag;
+    ASSERT(!c_p || c_p->common.id == from);
+    if (is_immed(from)) {
+        ASSERT(is_internal_pid(from) || is_internal_port(from));
+        from_tag = from;
+    }
+    else {
+        DistEntry *dep;
+        ASSERT(is_external_pid(from));
+        dep = external_pid_dist_entry(from);
+        from_tag = dep->sysname;
+    }
+    send_gen_exit_signal(c_p, from_tag, from, to, ERTS_SIG_Q_OP_EXIT,
+                         reason, NIL, token, normal_kills);
+}
+
+void
+erts_proc_sig_send_link_exit(Process *c_p, Eterm from, ErtsLink *lnk,
+                             Eterm reason, Eterm token)
+{
+    Eterm to;
+    ASSERT(!c_p || c_p->common.id == from);
+    ASSERT(lnk);
+    to = lnk->other.item;
+    if (is_not_immed(reason) || is_not_nil(token)) {
+        ASSERT(is_internal_pid(from) || is_internal_port(from));
+        send_gen_exit_signal(c_p, from, from, to, ERTS_SIG_Q_OP_EXIT_LINKED,
+                             reason, NIL, token, 0);
+    }
+    else {
+        /* Pass signal using old link structure... */
+        ErtsSignal *sig = (ErtsSignal *) lnk;
+        lnk->other.item = reason; /* pass reason via this other.item */
+        sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_EXIT_LINKED,
+                                                 lnk->type, 0);
+        if (proc_queue_signal(c_p, to, sig, ERTS_SIG_Q_OP_EXIT_LINKED))
+            return; /* receiver will destroy lnk structure */
+    }
+    if (lnk)
+        erts_link_release(lnk);
+}
+
+int
+erts_proc_sig_send_link(Process *c_p, Eterm to, ErtsLink *lnk)
+{
+    ErtsSignal *sig;
+    Uint16 type = lnk->type;
+
+    ASSERT(!c_p || c_p->common.id == lnk->other.item);
+    ASSERT(lnk);
+    ASSERT(is_internal_pid(to));
+
+    sig = (ErtsSignal *) lnk;
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_LINK,
+                                             type, 0);
+
+    return proc_queue_signal(c_p, to, sig, ERTS_SIG_Q_OP_LINK);
+}
+
+void
+erts_proc_sig_send_unlink(Process *c_p, ErtsLink *lnk)
+{
+    ErtsSignal *sig;
+    Eterm to;
+
+    ASSERT(lnk);
+
+    sig = (ErtsSignal *) lnk;
+    to = lnk->other.item;
+
+    ASSERT(is_internal_pid(to));
+
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_UNLINK,
+                                             lnk->type, 0);
+
+    if (!proc_queue_signal(c_p, to, sig, ERTS_SIG_Q_OP_UNLINK))
+        erts_link_release(lnk);
+}
+
+void
+erts_proc_sig_send_dist_link_exit(DistEntry *dep,
+                                  Eterm from, Eterm to,
+                                  Eterm reason, Eterm token)
+{
+    send_gen_exit_signal(NULL, dep->sysname, from, to, ERTS_SIG_Q_OP_EXIT_LINKED,
+                         reason, NIL, token, 0);
+}
+
+void
+erts_proc_sig_send_dist_unlink(DistEntry *dep, Eterm from, Eterm to)
+{
+    ErtsSignal *sig;
+
+    ASSERT(is_internal_pid(to));
+    ASSERT(is_external_pid(from));
+    ASSERT(dep == external_pid_dist_entry(from));
+
+    sig = (ErtsSignal *) make_sig_dist_link_op(ERTS_SIG_Q_OP_UNLINK,
+                                               to, from);
+
+    if (!proc_queue_signal(NULL, to, sig, ERTS_SIG_Q_OP_UNLINK))
+        destroy_sig_dist_link_op((ErtsSigDistLinkOp *) sig);
+}
+
+void
+erts_proc_sig_send_dist_monitor_down(DistEntry *dep, Eterm ref,
+                                     Eterm from, Eterm to,
+                                     Eterm reason)
+{
+    Eterm monitored, heap[3];
+    if (is_atom(from)) 
+        monitored = TUPLE2(&heap[0], from, dep->sysname);
+    else
+        monitored = from;
+    send_gen_exit_signal(NULL, dep->sysname, monitored,
+                         to, ERTS_SIG_Q_OP_MONITOR_DOWN,
+                         reason, ref, NIL, 0);    
+}
+
+void
+erts_proc_sig_send_monitor_down(ErtsMonitor *mon, Eterm reason)
+{
+    Eterm to;
+
+    ASSERT(erts_monitor_is_target(mon));
+    ASSERT(!erts_monitor_is_in_table(mon));
+
+    to = mon->other.item;
+    ASSERT(is_internal_pid(to));
+
+    if (is_immed(reason)) {
+        /* Pass signal using old monitor structure... */
+        ErtsSignal *sig;
+
+    send_using_monitor_struct:
+
+        mon->other.item = reason; /* Pass immed reason via other.item... */
+        sig = (ErtsSignal *) mon;
+        sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_MONITOR_DOWN,
+                                                 mon->type, 0);
+        if (proc_queue_signal(NULL, to, sig, ERTS_SIG_Q_OP_MONITOR_DOWN))
+            return; /* receiver will destroy mon structure */
+    }
+    else {
+        ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+        Eterm from_tag, monitored, heap[3];
+
+        if (mon->type == ERTS_MON_TYPE_SUSPEND) {
+            /*
+             * Set reason to 'undefined', since exit
+             * reason is not used for suspend monitors,
+             * and send using monitor structure. This
+             * since we don't want to trigger
+             * unnecessary memory allocation etc...
+             */
+            reason = am_undefined;
+            goto send_using_monitor_struct;
+        }
+
+        if (!(mon->flags & ERTS_ML_FLG_NAME)) {
+            from_tag = monitored = mdp->origin.other.item;
+            if (is_external_pid(from_tag)) {
+                DistEntry *dep = external_pid_dist_entry(from_tag);
+                from_tag = dep->sysname;
+            }
+        }
+        else {
+            ErtsMonitorDataExtended *mdep;
+            Eterm name, node;
+            mdep = (ErtsMonitorDataExtended *) mdp;
+            name = mdep->u.name;
+            ASSERT(is_atom(name));
+            if (mdep->dist) {
+                node = mdep->dist->nodename;
+                from_tag = node;
+            }
+            else {
+                node = erts_this_dist_entry->sysname;
+                from_tag = mdp->origin.other.item;
+            }
+            ASSERT(is_internal_port(from_tag)
+                   || is_internal_pid(from_tag)
+                   || is_atom(from_tag));
+            monitored = TUPLE2(&heap[0], name, node);
+        }    
+        send_gen_exit_signal(NULL, from_tag, monitored,
+                             to, ERTS_SIG_Q_OP_MONITOR_DOWN,
+                             reason, mdp->ref, NIL, 0);
+    }
+    erts_monitor_release(mon);
+}
+
+void
+erts_proc_sig_send_dist_demonitor(Eterm to, Eterm ref)
+{
+    ErtsSigDistProcDemonitor *dmon;
+    ErtsSignal *sig;
+    Eterm *hp;
+    ErlOffHeap oh;
+    size_t size;
+
+    ERTS_INIT_OFF_HEAP(&oh);
+
+    ASSERT(is_internal_pid(to));
+
+    size = sizeof(ErtsSigDistProcDemonitor) - sizeof(Eterm);
+    ASSERT(is_ref(ref));
+    size += NC_HEAP_SIZE(ref)*sizeof(Eterm);
+
+    dmon = erts_alloc(ERTS_ALC_T_DIST_DEMONITOR, size);
+
+    hp = &dmon->heap[0];
+    dmon->ref = STORE_NC(&hp, &oh, ref);
+    sig = (ErtsSignal *) dmon;
+
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_DEMONITOR,
+                                             ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR,
+                                             0);
+
+    if (!proc_queue_signal(NULL, to, sig, ERTS_SIG_Q_OP_DEMONITOR))
+        destroy_dist_proc_demonitor(dmon);
+}
+
+void
+erts_proc_sig_send_demonitor(ErtsMonitor *mon)
+{
+    ErtsSignal *sig = (ErtsSignal *) mon;
+    Uint16 type = mon->type;
+    Eterm to = mon->other.item;
+
+    ASSERT(is_internal_pid(to));
+    ASSERT(erts_monitor_is_origin(mon));
+    ASSERT(!erts_monitor_is_in_table(mon));
+
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_DEMONITOR,
+                                             type, 0);
+    
+    if (!proc_queue_signal(NULL, to, sig, ERTS_SIG_Q_OP_DEMONITOR))
+        erts_monitor_release(mon);
+}
+
+int
+erts_proc_sig_send_monitor(ErtsMonitor *mon, Eterm to)
+{
+    ErtsSignal *sig = (ErtsSignal *) mon;
+    Uint16 type = mon->type;
+
+    ASSERT(is_internal_pid(to) || to == am_undefined);
+    ASSERT(erts_monitor_is_target(mon));
+
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_MONITOR,
+                                             type, 0);
+    
+    return proc_queue_signal(NULL, to, sig, ERTS_SIG_Q_OP_MONITOR);
+}
+
+void
+erts_proc_sig_send_trace_change(Eterm to, Uint on, Uint off, Eterm tracer)
+{
+    ErtsSigTraceInfo *ti;
+    Eterm tag;
+
+    ti = erts_alloc(ERTS_ALC_T_SIG_DATA, sizeof(ErtsSigTraceInfo));
+    tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_TRACE_CHANGE_STATE,
+                                 ERTS_SIG_Q_TYPE_ADJUST_TRACE_INFO,
+                                 0);
+
+    ti->common.tag = tag;
+    ti->flags_off = off;
+    ti->flags_on = on;
+    ti->tracer = NIL;
+    if (is_not_nil(tracer))
+        erts_tracer_update(&ti->tracer, tracer);
+
+    if (!proc_queue_signal(NULL, to, (ErtsSignal *) ti,
+                           ERTS_SIG_Q_OP_TRACE_CHANGE_STATE))
+        destroy_trace_info(ti);
+}
+
+void
+erts_proc_sig_send_group_leader(Process *c_p, Eterm to, Eterm gl, Eterm ref)
+{
+    int res;
+    ErtsSigGroupLeader *sgl;
+    Eterm *hp;
+    Uint gl_sz, ref_sz, size;
+    erts_aint_t init_flags = ERTS_SIG_GL_FLG_ACTIVE|ERTS_SIG_GL_FLG_RECEIVER;
+    if (c_p)
+        init_flags |= ERTS_SIG_GL_FLG_SENDER;
+
+    ASSERT(c_p ? is_internal_ref(ref) : ref == NIL);
+
+    gl_sz = is_immed(gl) ? 0 : size_object(gl);
+    ref_sz = is_immed(ref) ? 0 : size_object(ref);
+    
+    size = sizeof(ErtsSigGroupLeader);
+
+    size += (gl_sz + ref_sz - 1) * sizeof(Eterm);
+
+    sgl = erts_alloc(ERTS_ALC_T_SIG_DATA, size);
+
+    erts_atomic_init_nob(&sgl->flags, init_flags);
+
+    ERTS_INIT_OFF_HEAP(&sgl->oh);
+
+    hp = &sgl->heap[0];
+
+    sgl->group_leader = is_immed(gl) ? gl : copy_struct(gl, gl_sz, &hp, &sgl->oh);
+    sgl->reply_to = c_p ? c_p->common.id : NIL;
+    sgl->ref = is_immed(ref) ? ref : copy_struct(ref, ref_sz, &hp, &sgl->oh);
+
+    sgl->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_GROUP_LEADER,
+                                             ERTS_SIG_Q_TYPE_UNDEFINED,
+                                             0);
+
+    res = proc_queue_signal(c_p, to, (ErtsSignal *) sgl,
+                            ERTS_SIG_Q_OP_GROUP_LEADER);
+
+    if (!res)
+        destroy_sig_group_leader(sgl);
+    else if (c_p) {
+        erts_aint_t flags, rm_flags = ERTS_SIG_GL_FLG_SENDER;
+        int prio_res = maybe_elevate_sig_handling_prio(c_p, to);
+        if (!prio_res)
+            rm_flags |= ERTS_SIG_GL_FLG_ACTIVE;
+        flags = erts_atomic_read_band_nob(&sgl->flags, ~rm_flags);
+        if (!prio_res && (flags & ERTS_SIG_GL_FLG_ACTIVE))
+            res = 0; /* We deactivated signal... */
+        if ((flags & ~rm_flags) == 0)
+            destroy_sig_group_leader(sgl);
+    }
+
+    if (!res && c_p)
+        group_leader_reply(c_p, c_p->common.id, ref, 0);
+}
+
+void
+erts_proc_sig_send_is_alive_request(Process *c_p, Eterm to, Eterm ref)
+{
+    ErlHeapFragment *hfrag;
+    Uint hsz;
+    Eterm *hp, *start_hp, ref_cpy, msg;
+    ErlOffHeap *ohp;
+    ErtsMessage *mp;
+    ErtsIsAliveRequest *alive_req;
+
+    ASSERT(is_internal_ordinary_ref(ref));
+
+    hsz = ERTS_REF_THING_SIZE + 3 + sizeof(ErtsIsAliveRequest)/sizeof(Eterm);
+
+    mp = erts_alloc_message(hsz, &hp);
+    hfrag = &mp->hfrag;
+    mp->next = NULL;
+    ohp = &hfrag->off_heap;
+    start_hp = hp;
+
+    ref_cpy = STORE_NC(&hp, ohp, ref);
+    msg = TUPLE2(hp, ref_cpy, am_false); /* default res 'false' */
+    hp += 3;
+
+    hfrag->used_size = hp - start_hp;
+
+    alive_req = (ErtsIsAliveRequest *) (char *) hp;
+    alive_req->message = msg;
+    alive_req->requester = c_p->common.id;
+
+    ERL_MESSAGE_TERM(mp) = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_IS_ALIVE,
+                                                  ERTS_SIG_Q_TYPE_UNDEFINED,
+                                                  0);
+    ERL_MESSAGE_TOKEN(mp) = NIL;
+    ERL_MESSAGE_FROM(mp) = am_system;
+#ifdef USE_VM_PROBES
+    ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+
+    if (proc_queue_signal(c_p, to, (ErtsSignal *) mp, ERTS_SIG_Q_OP_IS_ALIVE))
+        (void) maybe_elevate_sig_handling_prio(c_p, to);
+    else {
+        /* It wasn't alive; reply to ourselves... */
+        mp->next = NULL;
+        mp->data.attached = ERTS_MSG_COMBINED_HFRAG;
+        erts_queue_message(c_p, ERTS_PROC_LOCK_MAIN, mp, msg, am_system);
+    }
+}
+
+int
+erts_proc_sig_send_process_info_request(Process *c_p,
+                                        Eterm to,
+                                        int *item_ix,
+                                        int len,
+                                        int need_msgq_len,
+                                        int flags,
+                                        Uint reserve_size,
+                                        Eterm ref)
+{
+    Uint size = sizeof(ErtsProcessInfoSig) + (len - 1) * sizeof(int);
+    ErtsProcessInfoSig *pis = erts_alloc(ERTS_ALC_T_SIG_DATA, size);
+    int res;
+
+    ASSERT(c_p);
+    ASSERT(item_ix);
+    ASSERT(len > 0);
+    ASSERT(is_internal_ordinary_ref(ref));
+
+    pis->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_PROCESS_INFO,
+                                             0, 0);
+
+    if (!need_msgq_len)
+        pis->msgq_len_offset = ERTS_PROC_SIG_PI_MSGQ_LEN_IGNORE;
+    else {
+        pis->msgq_len_offset = ERTS_PROC_SIG_PI_MSGQ_LEN_SYNC;
+        pis->marker.common.next = NULL;
+        pis->marker.common.specific.next = NULL;
+        pis->marker.common.tag = ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK;
+        pis->marker.refc = 0;
+        pis->marker.delayed_len = 0;
+        pis->marker.len_offset = 0;
+    }
+    pis->requester = c_p->common.id;
+    sys_memcpy((void *) &pis->oref_thing,
+               (void *) internal_ref_val(ref),
+               sizeof(ErtsORefThing));
+    pis->ref = make_internal_ref((char *) &pis->oref_thing);
+    pis->reserve_size = reserve_size;
+    pis->len = len;
+    pis->flags = flags;
+    sys_memcpy((void *) &pis->item_ix[0],
+               (void *) item_ix,
+               sizeof(int)*len);
+    res = proc_queue_signal(c_p, to, (ErtsSignal *) pis,
+                            ERTS_SIG_Q_OP_PROCESS_INFO);
+    if (res)
+        (void) maybe_elevate_sig_handling_prio(c_p, to);
+    else
+        erts_free(ERTS_ALC_T_SIG_DATA, pis);
+    return res;
+}
+
+void
+erts_proc_sig_send_sync_suspend(Process *c_p, Eterm to, Eterm tag, Eterm reply)
+{
+    ErlHeapFragment *hfrag;
+    Uint hsz, tag_sz;
+    Eterm *hp, *start_hp, tag_cpy, msg, default_reply;
+    ErlOffHeap *ohp;
+    ErtsMessage *mp;
+    ErtsSyncSuspendRequest *ssusp;
+    int async_suspend;
+
+    tag_sz = size_object(tag);
+
+    hsz = 3 + tag_sz + sizeof(ErtsSyncSuspendRequest)/sizeof(Eterm);
+
+    mp = erts_alloc_message(hsz, &hp);
+    hfrag = &mp->hfrag;
+    mp->next = NULL;
+    ohp = &hfrag->off_heap;
+    start_hp = hp;
+
+    tag_cpy = copy_struct(tag, tag_sz, &hp, ohp);
+
+    async_suspend = is_non_value(reply);
+    default_reply = async_suspend ? am_suspended : reply;
+
+    msg = TUPLE2(hp, tag_cpy, default_reply);
+    hp += 3;
+
+    hfrag->used_size = hp - start_hp;
+
+    ssusp = (ErtsSyncSuspendRequest *) (char *) hp;
+    ssusp->message = msg;
+    ssusp->requester = c_p->common.id;
+    ssusp->async = async_suspend;
+
+    ERL_MESSAGE_TERM(mp) = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_SYNC_SUSPEND,
+                                                  ERTS_SIG_Q_TYPE_UNDEFINED,
+                                                  0);
+    ERL_MESSAGE_TOKEN(mp) = NIL;
+    ERL_MESSAGE_FROM(mp) = am_system;
+#ifdef USE_VM_PROBES
+    ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+
+    if (proc_queue_signal(c_p, to, (ErtsSignal *) mp, ERTS_SIG_Q_OP_SYNC_SUSPEND))
+        (void) maybe_elevate_sig_handling_prio(c_p, to);
+    else {
+        Eterm *tp;
+        /* It wasn't alive; reply to ourselves... */
+        mp->next = NULL;
+        mp->data.attached = ERTS_MSG_COMBINED_HFRAG;
+        tp = tuple_val(msg);
+        tp[2] = async_suspend ? am_badarg : am_exited;
+        erts_queue_message(c_p, ERTS_PROC_LOCK_MAIN,
+                           mp, msg, am_system);
+    }
+}
+
+Eterm
+erts_proc_sig_send_rpc_request(Process *c_p,
+                               Eterm to,
+                               int reply,
+                               Eterm (*func)(Process *, void *, int *, ErlHeapFragment **),
+                               void *arg)
+{
+    Eterm res;
+    ErtsProcSigRPC *sig = erts_alloc(ERTS_ALC_T_SIG_DATA,
+                                     sizeof(ErtsProcSigRPC));
+    sig->common.tag = ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_RPC,
+                                             ERTS_SIG_Q_TYPE_UNDEFINED,
+                                             0);
+    sig->requester = reply ? c_p->common.id : NIL;
+    sig->func = func;
+    sig->arg = arg;
+
+    if (!reply) {
+        res = am_ok;
+        sig->ref = am_ok;
+    }
+    else {
+        res = erts_make_ref(c_p);
+
+        sys_memcpy((void *) &sig->oref_thing,
+                   (void *) internal_ref_val(res),
+                   sizeof(ErtsORefThing));
+
+        sig->ref = make_internal_ref(&sig->oref_thing);
+
+        ERTS_RECV_MARK_SAVE(c_p);
+        ERTS_RECV_MARK_SET(c_p);
+    }
+
+    if (proc_queue_signal(c_p, to, (ErtsSignal *) sig, ERTS_SIG_Q_OP_RPC))
+        (void) maybe_elevate_sig_handling_prio(c_p, to);
+    else {
+        erts_free(ERTS_ALC_T_SIG_DATA, sig);
+        res = THE_NON_VALUE;
+        if (reply)
+            JOIN_MESSAGE(c_p);
+    }
+
+    return res;
+}
+
+static int
+handle_rpc(Process *c_p, ErtsProcSigRPC *rpc, int cnt, int limit, int *yieldp)
+{
+    Process *rp;
+    ErlHeapFragment *bp = NULL;
+    Eterm res;
+    Uint hsz;
+    int reds, out_cnt;
+
+    /*
+     * reds in:
+     *  Reductions left.
+     *
+     * reds out:
+     *  Absolute value of reds out equals consumed
+     *  amount of reds. If a negative value, force
+     *  a yield.
+     */
+
+    reds = (limit - cnt) / ERTS_SIG_REDS_CNT_FACTOR;
+    if (reds <= 0)
+        reds = 1;
+
+    res = (*rpc->func)(c_p, rpc->arg, &reds, &bp);
+
+    if (reds < 0) {
+        /* Force yield... */
+        *yieldp = !0;
+        reds *= -1;
+    }
+
+    out_cnt = reds*ERTS_SIG_REDS_CNT_FACTOR;
+
+    hsz = 3 + sizeof(ErtsORefThing)/sizeof(Eterm);
+
+    rp = erts_proc_lookup(rpc->requester);
+    if (!rp) {
+        if (bp)
+            free_message_buffer(bp);
+    }
+    else {
+        Eterm *hp, msg, ref;
+        ErtsMessage *mp = erts_alloc_message(hsz, &hp);
+        
+        sys_memcpy((void *) hp, (void *) &rpc->oref_thing,
+                   sizeof(rpc->oref_thing));
+
+        ref = make_internal_ref(hp);
+        hp += sizeof(rpc->oref_thing)/sizeof(Eterm);
+        msg = TUPLE2(hp, ref, res);
+
+        mp->hfrag.next = bp;
+
+        erts_queue_proc_message(c_p, rp, 0, mp, msg);
+    }
+
+    erts_free(ERTS_ALC_T_SIG_DATA, rpc);
+
+    return out_cnt;
+}
+
+static void
+is_alive_response(Process *c_p, ErtsMessage *mp, int is_alive)
+{
+    /*
+     * Sender prepared the message for us. Just patch
+     * the result if necessary. The default prepared
+     * result is 'false'.
+     */
+    Process *rp;
+    ErtsIsAliveRequest *alive_req;
+
+    alive_req = (ErtsIsAliveRequest *) (char *) (&mp->hfrag.mem[0]
+                                                 + mp->hfrag.used_size);
+
+
+    ASSERT(ERTS_SIG_IS_NON_MSG(mp));
+    ASSERT(ERTS_PROC_SIG_OP(((ErtsSignal *) mp)->common.tag)
+           == ERTS_SIG_Q_OP_IS_ALIVE);
+    ASSERT(mp->hfrag.alloc_size > mp->hfrag.used_size);
+    ASSERT((mp->hfrag.alloc_size - mp->hfrag.used_size)*sizeof(UWord)
+           >= sizeof(ErtsIsAliveRequest));
+    ASSERT(is_internal_pid(alive_req->requester));
+    ASSERT(alive_req->requester != c_p->common.id);
+    ASSERT(is_tuple_arity(alive_req->message, 2));
+    ASSERT(is_internal_ordinary_ref(tuple_val(alive_req->message)[1]));
+    ASSERT(tuple_val(alive_req->message)[2] == am_false);
+
+    ERL_MESSAGE_TERM(mp) = alive_req->message;
+    mp->data.attached = ERTS_MSG_COMBINED_HFRAG;
+    mp->next = NULL;
+
+    rp = erts_proc_lookup(alive_req->requester);
+    if (!rp)
+        erts_cleanup_messages(mp);
+    else {
+        if (is_alive) { /* patch result... */
+            Eterm *tp = tuple_val(alive_req->message);
+            tp[2] = am_true;
+        }
+        erts_queue_message(rp, 0, mp, alive_req->message, am_system);
+    }
+}
+
+static ERTS_INLINE void
+adjust_tracing_state(Process *c_p, ErtsSigRecvTracing *tracing, int setup)
+{
+    if (!IS_TRACED(c_p) || (ERTS_TRACE_FLAGS(c_p) & F_SENSITIVE)) {
+        tracing->messages.active = 0;
+        tracing->messages.receive_trace = 0;
+        tracing->messages.event = NULL;
+        tracing->messages.next = NULL;
+        tracing->procs = 0;
+        tracing->active = 0;
+    }
+    else {
+        Uint flgs = ERTS_TRACE_FLAGS(c_p);
+        int procs_trace = !!(flgs & F_TRACE_PROCS);
+        int recv_trace = !!(flgs & F_TRACE_RECEIVE);
+        /* procs tracing enabled? */
+
+        tracing->procs = procs_trace;
+
+        /* message receive tracing enabled? */
+        tracing->messages.receive_trace = recv_trace;
+        if (!recv_trace)
+            tracing->messages.event = NULL;
+        else {
+            if (tracing->messages.bp_ix < 0)
+                tracing->messages.bp_ix = erts_active_bp_ix();
+            tracing->messages.event = &erts_receive_tracing[tracing->messages.bp_ix];
+        }
+        if (setup) {
+            if (recv_trace)
+                tracing->messages.next = &c_p->sig_qs.cont;
+            else
+                tracing->messages.next = NULL;
+        }
+        tracing->messages.active = recv_trace;
+        tracing->active = recv_trace | procs_trace;
+    }
+
+#if defined(USE_VM_PROBES)
+    /* vm probe message_queued enabled? */
+
+    tracing->messages.vm_probes = DTRACE_ENABLED(message_queued);
+    if (tracing->messages.vm_probes) {
+        dtrace_proc_str(c_p, tracing->messages.receiver_name);
+        tracing->messages.active = !0;
+        tracing->active = !0;
+        if (setup && !tracing->messages.next)
+            tracing->messages.next = &c_p->sig_qs.cont;
+    }
+
+#endif
+}
+
+static ERTS_INLINE void
+setup_tracing_state(Process *c_p, ErtsSigRecvTracing *tracing)
+{
+    tracing->messages.bp_ix = -1;
+    adjust_tracing_state(c_p, tracing, !0);
+}
+
+static ERTS_INLINE void
+remove_iq_sig(Process *c_p, ErtsMessage *sig, ErtsMessage **next_sig)
+{
+    /*
+     * Remove signal from inner queue.
+     */
+    ASSERT(c_p->sig_qs.cont_last != &sig->next);
+    ASSERT(c_p->sig_qs.nmsigs.next != &sig->next);
+    ASSERT(c_p->sig_qs.nmsigs.last != &sig->next);
+
+    if (c_p->sig_qs.save == &sig->next)
+        c_p->sig_qs.save = next_sig;
+    if (c_p->sig_qs.last == &sig->next)
+        c_p->sig_qs.last = next_sig;
+    if (c_p->sig_qs.saved_last == &sig->next)
+        c_p->sig_qs.saved_last = next_sig;
+
+    *next_sig = sig->next;
+}
+
+static ERTS_INLINE void
+remove_mq_sig(Process *c_p, ErtsMessage *sig,
+           ErtsMessage **next_sig, ErtsMessage ***next_nm_sig)
+{
+    /*
+     * Remove signal from middle queue.
+     */
+    ASSERT(c_p->sig_qs.save != &sig->next);
+    ASSERT(c_p->sig_qs.last != &sig->next);
+
+    if (c_p->sig_qs.cont_last == &sig->next)
+	c_p->sig_qs.cont_last = next_sig;
+    if (c_p->sig_qs.saved_last == &sig->next)
+        c_p->sig_qs.saved_last = next_sig;
+    if (*next_nm_sig == &sig->next)
+        *next_nm_sig = next_sig;
+    if (c_p->sig_qs.nmsigs.last == &sig->next)
+        c_p->sig_qs.nmsigs.last = next_sig;
+
+    *next_sig = sig->next;
+}
+
+static ERTS_INLINE void
+remove_nm_sig(Process *c_p, ErtsMessage *sig, ErtsMessage ***next_nm_sig)
+{
+    ErtsMessage **next_sig = *next_nm_sig;
+    ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+    ASSERT(*next_sig == sig);
+    *next_nm_sig = ((ErtsSignal *) sig)->common.specific.next;
+    remove_mq_sig(c_p, sig, next_sig, next_nm_sig);
+}
+
+static ERTS_INLINE void
+convert_to_msg(Process *c_p, ErtsMessage *sig, ErtsMessage *msg,
+               ErtsMessage ***next_nm_sig)
+{
+    ErtsMessage **next_sig = *next_nm_sig;
+    ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+    *next_nm_sig = ((ErtsSignal *) sig)->common.specific.next;
+    c_p->sig_qs.len++;
+    *next_sig = msg;
+    remove_mq_sig(c_p, sig, &msg->next, next_nm_sig);
+}
+
+static ERTS_INLINE void
+convert_to_msgs(Process *c_p, ErtsMessage *sig, Uint no_msgs,
+                ErtsMessage *first_msg, ErtsMessage *last_msg,
+                ErtsMessage ***next_nm_sig)
+{
+    ErtsMessage **next_sig = *next_nm_sig;
+    ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+    *next_nm_sig = ((ErtsSignal *) sig)->common.specific.next;
+    c_p->sig_qs.len += no_msgs;
+    *next_sig = first_msg;
+    remove_mq_sig(c_p, sig, &last_msg->next, next_nm_sig);
+}
+
+static ERTS_INLINE void
+insert_messages(Process *c_p, ErtsMessage **next, ErtsMessage *first,
+                ErtsMessage *last, Uint no_msgs, ErtsMessage ***next_nm_sig)
+{
+    last->next = *next;
+    if (c_p->sig_qs.cont_last == next)
+	c_p->sig_qs.cont_last = &last->next;
+    if (*next_nm_sig == next)
+        *next_nm_sig = &last->next;
+    if (c_p->sig_qs.nmsigs.last == next)
+        c_p->sig_qs.nmsigs.last = &last->next;
+    c_p->sig_qs.len += no_msgs;
+    *next = first;
+}
+
+static ERTS_INLINE void
+remove_mq_m_sig(Process *c_p, ErtsMessage *sig, ErtsMessage **next_sig, ErtsMessage ***next_nm_sig)
+{
+    /* Removing message... */
+    ASSERT(!ERTS_SIG_IS_NON_MSG(sig));
+    c_p->sig_qs.len--;
+    remove_mq_sig(c_p, sig, next_sig, next_nm_sig);
+}
+
+static ERTS_INLINE void
+remove_iq_m_sig(Process *c_p, ErtsMessage *sig, ErtsMessage **next_sig)
+{
+    /* Removing message... */
+    ASSERT(!ERTS_SIG_IS_NON_MSG(sig));
+    c_p->sig_qs.len--;
+    remove_iq_sig(c_p, sig, next_sig);
+}
+
+static ERTS_INLINE void
+convert_prepared_sig_to_msg(Process *c_p, ErtsMessage *sig, Eterm msg,
+                            ErtsMessage ***next_nm_sig)
+{
+    /*
+     * Everything is already there except for the reference to
+     * the message and the combined hfrag marker that needs to be
+     * restored...
+     */
+    *next_nm_sig = ((ErtsSignal *) sig)->common.specific.next;
+    sig->data.attached = ERTS_MSG_COMBINED_HFRAG;
+    ERL_MESSAGE_TERM(sig) = msg;
+    c_p->sig_qs.len++;
+}
+
+static ERTS_INLINE int
+handle_exit_signal(Process *c_p, ErtsSigRecvTracing *tracing,
+                   ErtsMessage *sig, ErtsMessage ***next_nm_sig,
+                   int *exited)
+{
+    ErtsMessage *conv_msg = NULL;
+    ErtsExitSignalData *xsigd = NULL;
+    ErtsLinkData *ldp = NULL; /* Avoid erroneous warning... */
+    ErtsLink *dlnk = NULL; /* Avoid erroneous warning... */
+    Eterm tag = ((ErtsSignal *) sig)->common.tag;
+    Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+    int op = ERTS_PROC_SIG_OP(tag);
+    int destroy = 0;
+    int ignore = 0;
+    int save = 0;
+    int exit = 0;
+    int cnt = 1;
+    Eterm reason;
+    Eterm from;
+
+    if (type == ERTS_SIG_Q_TYPE_GEN_EXIT) {
+        xsigd = get_exit_signal_data(sig);
+        from = xsigd->from;
+        reason = xsigd->reason;
+        if (op != ERTS_SIG_Q_OP_EXIT_LINKED)
+            ignore = 0;
+        else {
+            ErtsLink *llnk = erts_link_tree_lookup(ERTS_P_LINKS(c_p), from);
+            if (!llnk) {
+                /* Link no longer active; ignore... */
+                ignore = !0;
+                destroy = !0;
+            }
+            else {
+                ignore = 0;
+                erts_link_tree_delete(&ERTS_P_LINKS(c_p), llnk);
+                if (llnk->type != ERTS_LNK_TYPE_DIST_PROC)
+                    erts_link_release(llnk);
+                else {
+                    dlnk = erts_link_to_other(llnk, &ldp);
+                    if (erts_link_dist_delete(dlnk))
+                        erts_link_release_both(ldp);
+                    else
+                        erts_link_release(llnk);
+                }
+            }
+        }
+
+        if (!ignore) {
+
+            if ((op != ERTS_SIG_Q_OP_EXIT || reason != am_kill)
+                && (c_p->flags & F_TRAP_EXIT)) {
+                convert_prepared_sig_to_msg(c_p, sig,
+                                            xsigd->message, next_nm_sig);
+                conv_msg = sig;
+            }
+            else if (reason == am_normal && !xsigd->u.normal_kills) {
+                /* Ignore it... */
+                destroy = !0;
+                ignore = !0;
+            }
+            else {
+                /* Terminate... */
+                save = !0;
+                exit = !0;
+                if (op == ERTS_SIG_Q_OP_EXIT && reason == am_kill)
+                    reason = am_killed;
+            }
+        }
+    }
+    else { /* Link exit */
+        ErtsLink *slnk = (ErtsLink *) sig;
+        ErtsLink *llnk = erts_link_to_other(slnk, &ldp);
+
+        ASSERT(type == ERTS_LNK_TYPE_PROC
+               || type == ERTS_LNK_TYPE_PORT
+               || type == ERTS_LNK_TYPE_DIST_PROC);
+
+        from = llnk->other.item;
+        reason = slnk->other.item; /* reason in other.item ... */
+        ASSERT(is_pid(from) || is_internal_port(from));
+        ASSERT(is_immed(reason));
+        ASSERT(op == ERTS_SIG_Q_OP_EXIT_LINKED);
+        dlnk = erts_link_tree_key_delete(&ERTS_P_LINKS(c_p), llnk);
+        if (!dlnk) {
+            ignore = !0; /* Link no longer active; ignore... */
+            ldp = NULL;
+        }
+        else {
+            Eterm pid;
+            ErtsMessage *mp;
+            ErtsProcLocks locks;
+            Uint hsz;
+            Eterm *hp;
+            ErlOffHeap *ohp;
+            ignore = 0;
+            if (dlnk == llnk)
+                dlnk = NULL;
+            else
+                ldp = NULL;
+
+            ASSERT(is_immed(reason));
+
+            if (!(c_p->flags & F_TRAP_EXIT)) {
+                if (reason == am_normal)
+                    ignore = !0; /* Ignore it... */
+                else
+                    exit = !0; /* Terminate... */
+            }
+            else {
+
+                /*
+                 * Create and EXIT message and replace
+                 * the original signal with the message...
+                 */
+
+                locks = ERTS_PROC_LOCK_MAIN;
+
+                hsz = 4 + NC_HEAP_SIZE(from);
+
+                mp = erts_alloc_message_heap(c_p, &locks, hsz, &hp, &ohp);
+
+                if (locks != ERTS_PROC_LOCK_MAIN)
+                    erts_proc_unlock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
+
+                pid = STORE_NC(&hp, ohp, from);
+
+                ERL_MESSAGE_TERM(mp) = TUPLE3(hp, am_EXIT, pid, reason);
+                ERL_MESSAGE_TOKEN(mp) = NIL;
+#ifdef USE_VM_PROBES
+                ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+                if (is_immed(pid))
+                    ERL_MESSAGE_FROM(mp) = pid;
+                else {
+                    DistEntry *dep;
+                    ASSERT(is_external_pid(pid));
+                    dep = external_pid_dist_entry(pid);
+                    ERL_MESSAGE_FROM(mp) = dep->sysname;
+                }
+
+                /* Replace original signal with the exit message... */
+                convert_to_msg(c_p, sig, mp, next_nm_sig);
+
+                cnt += 4;
+
+                conv_msg = mp;
+            }
+        }
+        destroy = !0;
+    }
+
+    if (ignore|exit) {
+        remove_nm_sig(c_p, sig, next_nm_sig);
+        if (exit) {
+            if (save) {
+                sig->data.attached = ERTS_MSG_COMBINED_HFRAG;
+                ERL_MESSAGE_TERM(sig) = xsigd->message;
+                erts_save_message_in_proc(c_p, sig);
+            }
+            /* Exit process... */
+            erts_set_self_exiting(c_p, reason);
+
+            cnt++;
+        }
+    }
+
+    if (!exit) {
+        if (conv_msg)
+            erts_proc_notify_new_message(c_p, ERTS_PROC_LOCK_MAIN);
+        if (op == ERTS_SIG_Q_OP_EXIT_LINKED && tracing->procs)
+            getting_unlinked(c_p, from);
+    }
+
+    if (destroy) {
+        cnt++;
+        if (type == ERTS_SIG_Q_TYPE_GEN_EXIT) {
+            sig->next = NULL;
+            erts_cleanup_messages(sig);
+        }
+        else {
+            if (ldp)
+                erts_link_release_both(ldp);
+            else {
+                if (dlnk)
+                    erts_link_release(dlnk);
+                erts_link_release((ErtsLink *) sig);
+            }
+        }
+    }
+
+    *exited = exit;
+
+    return cnt;
+}
+
+static ERTS_INLINE int
+convert_prepared_down_message(Process *c_p, ErtsMessage *sig,
+                              Eterm msg, ErtsMessage ***next_nm_sig)
+{
+    convert_prepared_sig_to_msg(c_p, sig, msg, next_nm_sig);
+    erts_proc_notify_new_message(c_p, ERTS_PROC_LOCK_MAIN);
+    return 1;
+}
+
+static int
+convert_to_down_message(Process *c_p,
+                        ErtsMessage *sig,
+                        ErtsMonitorData *mdp,
+                        Uint16 mon_type,
+                        ErtsMessage ***next_nm_sig)
+{
+    /*
+     * Create a 'DOWN' message and replace the signal
+     * with it...
+     */
+    int cnt = 0;
+    Eterm node = am_undefined;
+    ErtsMessage *mp;
+    ErtsProcLocks locks;
+    Uint hsz;
+    Eterm *hp, ref, from, type, reason;
+    ErlOffHeap *ohp;
+
+    ASSERT(mdp);
+    ASSERT((mdp->origin.flags & ERTS_ML_FLGS_SAME)
+           == (mdp->target.flags & ERTS_ML_FLGS_SAME));
+
+    hsz = 6; /* 5-tuple */
+
+    if (mdp->origin.flags & ERTS_ML_FLG_NAME)
+        hsz += 3;  /* reg name 2-tuple */
+    else {
+        ASSERT(is_pid(mdp->origin.other.item)
+               || is_internal_port(mdp->origin.other.item));
+        hsz += NC_HEAP_SIZE(mdp->origin.other.item);
+    }
+
+    ASSERT(is_ref(mdp->ref));
+    hsz += NC_HEAP_SIZE(mdp->ref);
+
+    locks = ERTS_PROC_LOCK_MAIN;
+
+    /* reason is mdp->target.other.item */
+    reason = mdp->target.other.item;
+    ASSERT(is_immed(reason));
+
+    mp = erts_alloc_message_heap(c_p, &locks, hsz, &hp, &ohp);
+
+    if (locks != ERTS_PROC_LOCK_MAIN)
+        erts_proc_unlock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
+
+    cnt += 4;
+
+    ref = STORE_NC(&hp, ohp, mdp->ref);
+
+    if (!(mdp->origin.flags & ERTS_ML_FLG_NAME)) {
+        from = STORE_NC(&hp, ohp, mdp->origin.other.item);
+    }
+    else {
+        ErtsMonitorDataExtended *mdep;
+        ASSERT(mdp->origin.flags & ERTS_ML_FLG_EXTENDED);
+        mdep = (ErtsMonitorDataExtended *) mdp;
+        ASSERT(is_atom(mdep->u.name));
+        if (mdep->dist)
+            node = mdep->dist->nodename;
+        else
+            node = erts_this_dist_entry->sysname;
+        from = TUPLE2(hp, mdep->u.name, node);
+        hp += 3;
+    }
+
+    ASSERT(mdp->origin.type == mon_type);
+    switch (mon_type) {
+    case ERTS_MON_TYPE_PORT:
+        type = am_port;
+        if (mdp->origin.other.item == am_undefined) {
+            /* failed by name... */
+            ERL_MESSAGE_FROM(mp) = am_system;
+        }
+        else {
+            ASSERT(is_internal_port(mdp->origin.other.item));
+            ERL_MESSAGE_FROM(mp) = mdp->origin.other.item;
+        }
+        break;
+    case ERTS_MON_TYPE_PROC:
+        type = am_process;
+        if (mdp->origin.other.item == am_undefined) {
+            /* failed by name... */
+            ERL_MESSAGE_FROM(mp) = am_system;
+        }
+        else {
+            ASSERT(is_internal_pid(mdp->origin.other.item));
+            ERL_MESSAGE_FROM(mp) = mdp->origin.other.item;
+        }
+        break;
+    case ERTS_MON_TYPE_DIST_PROC:
+        type = am_process;
+        if (node == am_undefined) {
+            ErtsMonitorDataExtended *mdep;
+            ASSERT(mdp->origin.flags & ERTS_ML_FLG_EXTENDED);
+            mdep = (ErtsMonitorDataExtended *) mdp;
+            ASSERT(mdep->dist);
+            node = mdep->dist->nodename;
+        }
+        ASSERT(is_atom(node) && node != am_undefined);
+        ERL_MESSAGE_FROM(mp) = node;
+        break;
+    default:
+        ERTS_INTERNAL_ERROR("Unexpected monitor type");
+        type = am_undefined;
+        ERL_MESSAGE_FROM(mp) = am_undefined;
+        break;
+    }
+
+    ERL_MESSAGE_TERM(mp) = TUPLE5(hp, am_DOWN, ref,
+                                  type, from, reason);
+    hp += 6;
+
+    ERL_MESSAGE_TOKEN(mp) = NIL;
+#ifdef USE_VM_PROBES
+    ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+    /* Replace original signal with the exit message... */
+    convert_to_msg(c_p, sig, mp, next_nm_sig);
+
+    cnt += 4;
+
+    erts_proc_notify_new_message(c_p, ERTS_PROC_LOCK_MAIN);
+
+    return cnt;
+}
+
+static ERTS_INLINE int
+convert_to_nodedown_messages(Process *c_p,
+                             ErtsMessage *sig,
+                             ErtsMonitorData *mdp,
+                             ErtsMessage ***next_nm_sig)
+{
+    int cnt = 1;
+    Uint n;
+    ErtsMonitorDataExtended *mdep = (ErtsMonitorDataExtended *) mdp;
+
+    ASSERT((mdp->origin.flags & ERTS_ML_FLGS_SAME)
+           == (mdp->target.flags & ERTS_ML_FLGS_SAME));
+    ASSERT(mdp->origin.flags & ERTS_ML_FLG_EXTENDED);
+
+    n = mdep->u.refc;
+
+    if (n == 0)
+        remove_nm_sig(c_p, sig, next_nm_sig);
+    else {
+        Uint i;
+        ErtsMessage *nd_first = NULL;
+        ErtsMessage *nd_last = NULL;
+        ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN;
+        Eterm node = mdep->dist->nodename;
+
+        ASSERT(is_atom(node));
+        ASSERT(n > 0);
+
+        for (i = 0; i < n; i++) {
+            ErtsMessage *mp;
+            ErlOffHeap *ohp;
+            Eterm *hp;
+
+            mp = erts_alloc_message_heap(c_p, &locks, 3, &hp, &ohp);
+
+            ERL_MESSAGE_TERM(mp) = TUPLE2(hp, am_nodedown, node);
+            ERL_MESSAGE_FROM(mp) = am_system;
+            ERL_MESSAGE_TOKEN(mp) = NIL;
+#ifdef USE_VM_PROBES
+            ERL_MESSAGE_DT_UTAG(mp) = NIL;
+#endif
+            mp->next = nd_first;
+            nd_first = mp;
+            if (!nd_last)
+                nd_last = mp;
+            cnt++;
+        }
+
+        if (locks != ERTS_PROC_LOCK_MAIN)
+            erts_proc_unlock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
+
+        /* Replace signal with 'nodedown' messages */
+        convert_to_msgs(c_p, sig, n, nd_first, nd_last, next_nm_sig);
+
+        erts_proc_notify_new_message(c_p, ERTS_PROC_LOCK_MAIN);
+    }
+    return cnt;
+}
+
+static int 
+handle_nodedown(Process *c_p,
+                ErtsMessage *sig,
+                ErtsMonitorData *mdp,
+                ErtsMessage ***next_nm_sig)
+{
+    ErtsMonitorDataExtended *mdep = (ErtsMonitorDataExtended *) mdp;
+    ErtsMonitor *omon = &mdp->origin;
+    int not_in_subtab = !(omon->flags & ERTS_ML_FLG_IN_SUBTABLE);
+    int cnt = 1;
+
+    ASSERT(erts_monitor_is_in_table(omon));
+
+    if (not_in_subtab & !mdep->uptr.node_monitors)
+        erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p), omon);
+    else if (not_in_subtab) {
+        ErtsMonitor *sub_mon;
+        ErtsMonitorDataExtended *sub_mdep;
+        sub_mon = erts_monitor_list_last(mdep->uptr.node_monitors);
+        ASSERT(sub_mon);
+        erts_monitor_list_delete(&mdep->uptr.node_monitors, sub_mon);
+        sub_mon->flags &= ~ERTS_ML_FLG_IN_SUBTABLE;
+        sub_mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(sub_mon);
+        ASSERT(!sub_mdep->uptr.node_monitors);
+        sub_mdep->uptr.node_monitors = mdep->uptr.node_monitors;
+        mdep->uptr.node_monitors = NULL;
+        erts_monitor_tree_replace(&ERTS_P_MONITORS(c_p), omon, sub_mon);
+        cnt += 2;
+    }
+    else {
+        ErtsMonitorDataExtended *top_mdep;
+        ErtsMonitor *top_mon;
+        ASSERT(is_atom(omon->other.item));
+        ASSERT(!mdep->uptr.node_monitors);
+        top_mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p),
+                                           omon->other.item);
+        ASSERT(top_mon);
+        top_mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(top_mon);
+        ASSERT(top_mdep->uptr.node_monitors);
+        erts_monitor_list_delete(&top_mdep->uptr.node_monitors, omon);
+        omon->flags &= ~ERTS_ML_FLG_IN_SUBTABLE;
+        cnt += 3;
+    }
+
+    return cnt + convert_to_nodedown_messages(c_p, sig, mdp, next_nm_sig);
+}
+
+static void
+handle_persistent_mon_msg(Process *c_p, Uint16 type,
+                          ErtsMonitor *mon, ErtsMessage *sig,
+                          Eterm msg, ErtsMessage ***next_nm_sig)
+{
+    convert_prepared_sig_to_msg(c_p, sig, msg, next_nm_sig);
+
+    switch (type) {
+
+    case ERTS_MON_TYPE_TIME_OFFSET:
+        ASSERT(mon->type == ERTS_MON_TYPE_TIME_OFFSET);
+        break;
+
+    case ERTS_MON_TYPE_NODES: {
+        ErtsMonitorDataExtended *mdep;
+        Uint n;
+        ASSERT(mon->type == ERTS_MON_TYPE_NODES);
+        mdep = (ErtsMonitorDataExtended *) erts_monitor_to_data(mon);
+        ERTS_ML_ASSERT(mdep->u.refc > 0);
+        n = mdep->u.refc;
+        n--;
+        if (n > 0) {
+            ErtsProcLocks locks = ERTS_PROC_LOCK_MAIN;
+            ErtsMessage *first = NULL, *prev, *last;
+            Uint hsz = size_object(msg);
+            Uint i;
+
+            for (i = 0; i < n; i++) {
+                Eterm *hp;
+                ErlOffHeap *ohp;
+
+                last = erts_alloc_message_heap(c_p, &locks, hsz, &hp, &ohp);
+
+                if (!first)
+                    first = last;
+                else
+                    prev->next = last;
+                prev = last;
+
+                ERL_MESSAGE_TERM(last) = copy_struct(msg, hsz, &hp, ohp);
+
+#ifdef USE_VM_PROBES
+                ASSERT(is_immed(ERL_MESSAGE_DT_UTAG(sig)));
+                ERL_MESSAGE_DT_UTAG(last) = ERL_MESSAGE_DT_UTAG(sig);
+#endif
+                ASSERT(is_immed(ERL_MESSAGE_TOKEN(sig)));
+                ERL_MESSAGE_TOKEN(last) = ERL_MESSAGE_TOKEN(sig);
+                ASSERT(is_immed(ERL_MESSAGE_FROM(sig)));
+                ERL_MESSAGE_FROM(last) = ERL_MESSAGE_FROM(sig);
+
+            }
+            if (locks != ERTS_PROC_LOCK_MAIN)
+                erts_proc_unlock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
+            insert_messages(c_p, &sig->next, first, last, n, next_nm_sig);
+        }
+        break;
+    }
+
+    default:
+        ERTS_INTERNAL_ERROR("Invalid type");
+        break;
+    }
+
+    erts_proc_notify_new_message(c_p, ERTS_PROC_LOCK_MAIN);
+}
+
+static void
+group_leader_reply(Process *c_p, Eterm to, Eterm ref, int success)
+{
+    Process *rp = erts_proc_lookup(to);
+
+    if (rp) {
+        ErtsProcLocks locks;
+        Uint sz;
+        Eterm *hp, msg, ref_cpy, result;
+        ErlOffHeap *ohp;
+        ErtsMessage *mp;
+
+        ASSERT(is_internal_ref(ref));
+
+        locks = c_p == rp ? ERTS_PROC_LOCK_MAIN : 0;
+        sz = size_object(ref);
+
+        mp = erts_alloc_message_heap(rp, &locks, sz+3,
+                                     &hp, &ohp);
+
+        ref_cpy = copy_struct(ref, sz, &hp, ohp);
+        result = success ? am_true : am_badarg;
+        msg = TUPLE2(hp, ref_cpy, result);
+
+        erts_queue_message(rp, locks, mp, msg, am_system);
+
+        if (c_p == rp)
+            locks &= ~ERTS_PROC_LOCK_MAIN;
+
+        if (locks)
+            erts_proc_unlock(rp, locks);
+    }
+}
+
+static void
+handle_group_leader(Process *c_p, ErtsSigGroupLeader *sgl)
+{
+    erts_aint_t flags;
+
+    flags = erts_atomic_read_band_nob(&sgl->flags, ~ERTS_SIG_GL_FLG_ACTIVE);
+    if (flags & ERTS_SIG_GL_FLG_ACTIVE) {
+        int res = erts_set_group_leader(c_p, sgl->group_leader);
+        if (is_internal_pid(sgl->reply_to))
+            group_leader_reply(c_p, sgl->reply_to, sgl->ref, res);
+    }
+
+    flags = erts_atomic_read_band_nob(&sgl->flags, ~ERTS_SIG_GL_FLG_RECEIVER);
+    if ((flags & ~ERTS_SIG_GL_FLG_RECEIVER) == 0)
+        destroy_sig_group_leader(sgl);
+}
+
+static void
+check_push_msgq_len_offs_marker(Process *rp, ErtsSignal *sig)
+{
+    ErtsProcessInfoSig *pisig = (ErtsProcessInfoSig *) sig;
+
+    ASSERT(ERTS_PROC_SIG_OP(sig->common.tag) == ERTS_SIG_Q_OP_PROCESS_INFO);
+
+    if (pisig->msgq_len_offset == ERTS_PROC_SIG_PI_MSGQ_LEN_SYNC) {
+        ErtsProcSigMsgQLenOffsetMarker *mrkr;
+        Sint len, msgq_len_offset;
+        ErtsMessage *first = rp->sig_inq.first;
+        ASSERT(first);
+        if (((ErtsSignal *) first)->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK)
+            mrkr = (ErtsProcSigMsgQLenOffsetMarker *) first;
+        else {
+            mrkr = &pisig->marker;
+
+            ASSERT(mrkr->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK);
+
+            mrkr->common.next = first;
+            ASSERT(rp->sig_inq.last != &rp->sig_inq.first);
+            if (rp->sig_inq.nmsigs.next == &rp->sig_inq.first)
+                rp->sig_inq.nmsigs.next = &mrkr->common.next;
+            if (rp->sig_inq.nmsigs.last == &rp->sig_inq.first)
+                rp->sig_inq.nmsigs.last = &mrkr->common.next;
+            rp->sig_inq.first = (ErtsMessage *) mrkr;
+        }
+
+        len = rp->sig_inq.len;
+        msgq_len_offset = len - mrkr->len_offset;
+
+        mrkr->len_offset = len;
+        mrkr->refc++;
+
+        pisig->msgq_len_offset = msgq_len_offset;
+
+#ifdef DEBUG
+        /* save pointer to used marker... */
+        pisig->marker.common.specific.attachment = (void *) mrkr;
+#endif
+
+    }
+}
+
+static void
+destroy_process_info_request(Process *c_p, ErtsProcessInfoSig *pisig)
+{
+    int dealloc_pisig = !0;
+
+    if (pisig->msgq_len_offset != ERTS_PROC_SIG_PI_MSGQ_LEN_IGNORE) {
+        Sint refc;
+        int dealloc_marker = 0;
+        ErtsProcSigMsgQLenOffsetMarker *marker;
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+        Sint delayed_len;
+#endif
+
+        ASSERT(pisig->msgq_len_offset >= 0);
+
+        erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+        marker = (ErtsProcSigMsgQLenOffsetMarker *) c_p->sig_inq.first;
+        ASSERT(marker);
+        ASSERT(marker->refc > 0);
+        ASSERT(pisig->marker.common.specific.attachment == (void *) marker);
+
+        marker->delayed_len -= pisig->msgq_len_offset;
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+        delayed_len = marker->delayed_len;
+#endif
+
+        refc = --marker->refc;
+        if (refc) {
+            if (marker == &pisig->marker) {
+                /* Another signal using our marker... */
+                dealloc_pisig = 0;
+            }
+        }
+        else {
+            /* Marker unused; remove it... */
+            ASSERT(marker->delayed_len + marker->len_offset == 0);
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+            delayed_len += marker->len_offset;
+#endif
+            if (marker != &pisig->marker)
+                dealloc_marker = !0; /* used another signals marker... */
+            c_p->sig_inq.first = marker->common.next;
+            if (c_p->sig_inq.last == &marker->common.next)
+                c_p->sig_inq.last = &c_p->sig_inq.first;
+            if (c_p->sig_inq.nmsigs.next == &marker->common.next)
+                c_p->sig_inq.nmsigs.next = &c_p->sig_inq.first;
+            if (c_p->sig_inq.nmsigs.last == &marker->common.next)
+                c_p->sig_inq.nmsigs.last = &c_p->sig_inq.first;
+        }
+        erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+
+        if (!refc) {
+            c_p->flags &= ~F_DELAYED_PSIGQS_LEN;
+            /* Adjust msg len of inner+middle queue */
+            ASSERT(marker->len_offset <= 0);
+            c_p->sig_qs.len -= marker->len_offset;
+
+            ASSERT(c_p->sig_qs.len >= 0);
+        }
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+        {
+            Sint len = 0;
+            ERTS_FOREACH_SIG_PRIVQS(
+                c_p, mp,
+                {
+                    if (ERTS_SIG_IS_MSG(mp))
+                        len++;
+                });
+            ERTS_ASSERT(c_p->sig_qs.len + delayed_len == len);
+        }
+#endif
+
+
+        if (dealloc_marker) {
+            ErtsProcessInfoSig *pisig2
+                = (ErtsProcessInfoSig *) (((char *) marker)
+                                          - offsetof(ErtsProcessInfoSig,
+                                                     marker));
+            erts_free(ERTS_ALC_T_SIG_DATA, pisig2);
+        }
+    }
+
+    if (dealloc_pisig)
+        erts_free(ERTS_ALC_T_SIG_DATA, pisig);
+}
+
+static int
+handle_process_info(Process *c_p, ErtsSigRecvTracing *tracing,
+                    ErtsMessage *sig, ErtsMessage ***next_nm_sig,
+                    int is_alive)
+{
+    ErtsProcessInfoSig *pisig = (ErtsProcessInfoSig *) sig;
+    Uint reds = 0;
+    Process *rp;
+
+    if (pisig->msgq_len_offset != ERTS_PROC_SIG_PI_MSGQ_LEN_IGNORE) {
+        /*
+         * Request requires message queue data to be updated
+         * before inspection...
+         */
+
+        ASSERT(pisig->msgq_len_offset >= 0);
+        /*
+         * Update sig_qs.len to reflect the length
+         * of the message queue...
+         */
+        c_p->sig_qs.len += pisig->msgq_len_offset;
+
+        if (is_alive) {
+            /*
+             * Move messages part of message queue into inner
+             * signal queue...
+             */
+            ASSERT(tracing);
+
+            if (*next_nm_sig != &c_p->sig_qs.cont) {
+                if (*next_nm_sig == tracing->messages.next)
+                    tracing->messages.next = &c_p->sig_qs.cont;
+                *c_p->sig_qs.last = c_p->sig_qs.cont;
+                c_p->sig_qs.last = *next_nm_sig;
+
+                c_p->sig_qs.cont = **next_nm_sig;
+                if (c_p->sig_qs.nmsigs.last == *next_nm_sig)
+                    c_p->sig_qs.nmsigs.last = &c_p->sig_qs.cont;
+                *next_nm_sig = &c_p->sig_qs.cont;
+                *c_p->sig_qs.last = NULL;
+            }
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+            {
+                Sint len;
+                ErtsMessage *mp;
+                for (mp = c_p->sig_qs.first, len = 0; mp; mp = mp->next) {
+                    ERTS_ASSERT(ERTS_SIG_IS_MSG(mp));
+                    len++;
+                }
+                ERTS_ASSERT(c_p->sig_qs.len == len);
+            }
+#endif
+        }
+    }
+    if (is_alive) {
+        if (!pisig->common.specific.next) {
+            /*
+             * No more signals in middle queue...
+             *
+             * Process-info 'status' needs sig-q
+             * process flag to be updated in order
+             * to show accurate result...
+             */
+            erts_atomic32_read_band_nob(&c_p->state,
+                                        ~ERTS_PSFLG_SIG_Q);
+        }
+        remove_nm_sig(c_p, sig, next_nm_sig);
+    }
+
+    rp = erts_proc_lookup(pisig->requester);
+    ASSERT(c_p != rp);
+    if (rp) {
+        Eterm msg, res, ref, *hp;
+        ErtsProcLocks locks = 0;
+        ErtsHeapFactory hfact;
+        ErtsMessage *mp;
+        Uint reserve_size = 3 + sizeof(pisig->oref_thing)/sizeof(Eterm);
+
+        if (!is_alive) {
+            ErlOffHeap *ohp;
+            mp = erts_alloc_message_heap(rp, &locks, reserve_size, &hp, &ohp);
+            res = am_undefined;
+        }
+        else {
+            ErlHeapFragment *hfrag;
+
+            reserve_size += pisig->reserve_size;
+
+            mp = erts_alloc_message(0, NULL);
+            hfrag = new_message_buffer(reserve_size);
+            mp->data.heap_frag = hfrag;
+            erts_factory_selfcontained_message_init(&hfact, mp, &hfrag->mem[0]);
+
+            res = erts_process_info(c_p, &hfact, c_p, ERTS_PROC_LOCK_MAIN,
+                                    pisig->item_ix, pisig->len,
+                                    pisig->flags, reserve_size, &reds);
+
+            hp = erts_produce_heap(&hfact,
+                                   3 + sizeof(pisig->oref_thing)/sizeof(Eterm),
+                                   0);
+        }
+
+        sys_memcpy((void *) hp, (void *) &pisig->oref_thing,
+                   sizeof(pisig->oref_thing));
+        ref = make_internal_ref(hp);
+        hp += sizeof(pisig->oref_thing)/sizeof(Eterm);
+
+        msg = TUPLE2(hp, ref, res);
+
+        if (is_alive)
+            erts_factory_trim_and_close(&hfact, &msg, 1);
+
+        erts_queue_proc_message(c_p, rp, locks, mp, msg);
+
+        if (!is_alive && locks)
+            erts_proc_unlock(rp, locks);
+    }
+
+    destroy_process_info_request(c_p, pisig);
+
+    if (reds > INT_MAX/8)
+        reds = INT_MAX/8;
+
+    return ((int) reds)*4 + 8;
+}
+
+static void
+handle_suspend(Process *c_p, ErtsMonitor *mon, int *yieldp)
+{
+    erts_aint32_t state = erts_atomic32_read_nob(&c_p->state);
+
+    ASSERT(mon->type == ERTS_MON_TYPE_SUSPEND);
+
+    if (!(state & ERTS_PSFLG_DIRTY_RUNNING)) {
+        ErtsMonitorSuspend *msp;
+        erts_aint_t mstate;
+
+        msp = (ErtsMonitorSuspend *) erts_monitor_to_data(mon);
+        mstate = erts_atomic_read_bor_acqb(&msp->state,
+                                           ERTS_MSUSPEND_STATE_FLG_ACTIVE);
+        ASSERT(!(mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE)); (void) mstate;
+        erts_suspend(c_p, ERTS_PROC_LOCK_MAIN, NULL);
+        *yieldp = !0;
+    }
+    else {
+        /* Executing dirty; delay suspend... */
+        ErtsProcSigPendingSuspend *psusp;
+        ErtsMonitorSuspend *msp;
+
+        psusp = ERTS_PROC_GET_PENDING_SUSPEND(c_p);
+        if (!psusp) {
+            psusp = erts_alloc(ERTS_ALC_T_SIG_DATA,
+                               sizeof(ErtsProcSigPendingSuspend));
+            psusp->mon = NULL;
+            psusp->sync = NULL;
+            ERTS_PROC_SET_PENDING_SUSPEND(c_p, (void *) psusp);
+        }
+
+        msp = (ErtsMonitorSuspend *) erts_monitor_to_data(mon);
+
+        msp->next = psusp->mon;
+        psusp->mon = msp;
+
+        erts_atomic32_inc_nob(&msp->md.refc);
+    }
+}
+
+static void
+sync_suspend_reply(Process *c_p, ErtsMessage *mp, erts_aint32_t state)
+{
+    /*
+     * Sender prepared the message for us. Just patch
+     * the result if necessary. The default prepared
+     * result is 'false'.
+     */
+    Process *rp;
+    ErtsSyncSuspendRequest *ssusp;
+
+    ssusp = (ErtsSyncSuspendRequest *) (char *) (&mp->hfrag.mem[0]
+                                                 + mp->hfrag.used_size);
+
+    ASSERT(ERTS_SIG_IS_NON_MSG(mp));
+    ASSERT(ERTS_PROC_SIG_OP(((ErtsSignal *) mp)->common.tag)
+           == ERTS_SIG_Q_OP_SYNC_SUSPEND);
+    ASSERT(mp->hfrag.alloc_size > mp->hfrag.used_size);
+    ASSERT((mp->hfrag.alloc_size - mp->hfrag.used_size)*sizeof(UWord)
+           >= sizeof(ErtsSyncSuspendRequest));
+    ASSERT(is_internal_pid(ssusp->requester));
+    ASSERT(ssusp->requester != c_p->common.id);
+    ASSERT(is_tuple_arity(ssusp->message, 2));
+    ASSERT(is_immed(tuple_val(ssusp->message)[2]));
+
+    ERL_MESSAGE_TERM(mp) = ssusp->message;
+    mp->data.attached = ERTS_MSG_COMBINED_HFRAG;
+    mp->next = NULL;
+
+    rp = erts_proc_lookup(ssusp->requester);
+    if (!rp)
+        erts_cleanup_messages(mp);
+    else {
+        if ((state & (ERTS_PSFLG_EXITING
+                      | ERTS_PSFLG_SUSPENDED)) != ERTS_PSFLG_SUSPENDED) {
+            /* Not suspended -> patch result... */
+            if (state & ERTS_PSFLG_EXITING) {
+                Eterm *tp = tuple_val(ssusp->message);
+                tp[2] = ssusp->async ? am_exited : am_badarg;
+            }
+            else {
+                Eterm *tp = tuple_val(ssusp->message);
+                ASSERT(!(state & ERTS_PSFLG_SUSPENDED));
+                tp[2] = ssusp->async ? am_not_suspended : am_internal_error;
+            }
+        }
+        erts_queue_proc_message(c_p, rp, 0, mp, ssusp->message);
+    }
+}
+
+static void
+handle_sync_suspend(Process *c_p, ErtsMessage *mp)
+{
+    ErtsProcSigPendingSuspend *psusp;
+
+    psusp = (ErtsProcSigPendingSuspend *) ERTS_PROC_GET_PENDING_SUSPEND(c_p);
+    if (!psusp)
+        sync_suspend_reply(c_p, mp, erts_atomic32_read_nob(&c_p->state));
+    else {
+        mp->next = psusp->sync;
+        psusp->sync = mp;
+    }
+}
+
+void
+erts_proc_sig_handle_pending_suspend(Process *c_p)
+{
+    ErtsMonitorSuspend *msp;
+    ErtsMessage *sync;
+    ErtsProcSigPendingSuspend *psusp;
+    erts_aint32_t state = erts_atomic32_read_nob(&c_p->state);
+
+    psusp = (ErtsProcSigPendingSuspend *) ERTS_PROC_GET_PENDING_SUSPEND(c_p);
+
+    msp = psusp->mon;
+    
+    while (msp) {
+        ErtsMonitorSuspend *next_msp = msp->next;
+        msp->next = NULL;
+        if (!(state & ERTS_PSFLG_EXITING)
+            && erts_monitor_is_in_table(&msp->md.target)) {
+            erts_aint_t mstate;
+
+            mstate = erts_atomic_read_bor_acqb(&msp->state,
+                                               ERTS_MSUSPEND_STATE_FLG_ACTIVE);
+            ASSERT(!(mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE)); (void) mstate;
+            erts_suspend(c_p, ERTS_PROC_LOCK_MAIN, NULL);
+        }
+
+        erts_monitor_release(&msp->md.target);
+
+        msp = next_msp;
+    }
+
+    sync = psusp->sync;
+
+    while (sync) {
+        ErtsMessage *next_sync = sync->next;
+        sync->next = NULL;
+        sync_suspend_reply(c_p, sync, state);
+        sync = next_sync;
+    }
+
+    erts_free(ERTS_ALC_T_SIG_DATA, psusp);
+
+    ERTS_PROC_SET_PENDING_SUSPEND(c_p, NULL);
+}
+
+/*
+ * Called in order to handle incoming signals.
+ */
+
+int
+erts_proc_sig_handle_incoming(Process *c_p, erts_aint32_t *statep,
+                              int *redsp, int max_reds, int local_only)
+{
+    Eterm tag;
+    erts_aint32_t state;
+    int yield, cnt, limit, abs_lim, msg_tracing;
+    ErtsMessage *sig, ***next_nm_sig;
+    ErtsSigRecvTracing tracing;
+
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(c_p, 0);
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
+
+    if (local_only)
+        state = -1; /* can never be a valid state... */
+    else {
+        state = erts_atomic32_read_nob(&c_p->state);
+        if (ERTS_PSFLG_SIG_IN_Q & state) {
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+            erts_proc_sig_fetch(c_p);
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+        }
+    }
+
+    limit = *redsp;
+    *redsp = 0;
+    yield = 0;
+
+    if (!c_p->sig_qs.cont) {
+        if (state == -1)
+            *statep = erts_atomic32_read_nob(&c_p->state);
+        else
+            *statep = state;
+        return !0;
+    }
+
+    next_nm_sig = &c_p->sig_qs.nmsigs.next;
+
+    setup_tracing_state(c_p, &tracing);
+    msg_tracing = tracing.messages.active;
+
+    limit *= ERTS_SIG_REDS_CNT_FACTOR;
+    abs_lim = ERTS_SIG_REDS_CNT_FACTOR*max_reds;
+    if (limit > abs_lim)
+        limit = abs_lim;
+
+    cnt = 0;
+
+    do {
+
+        if (msg_tracing) {
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            if (handle_msg_tracing(c_p, &tracing, next_nm_sig) != 0) {
+                ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+                break; /* tracing limit or end... */
+            }
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+        }
+
+        if (!*next_nm_sig)
+            break;
+
+        sig = **next_nm_sig;
+
+        ASSERT(sig);
+        ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+        tag = ((ErtsSignal *) sig)->common.tag;
+ 
+        switch (ERTS_PROC_SIG_OP(tag)) {
+
+        case ERTS_SIG_Q_OP_EXIT:
+        case ERTS_SIG_Q_OP_EXIT_LINKED: {
+            int exited;
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            cnt += handle_exit_signal(c_p, &tracing, sig,
+                                      next_nm_sig, &exited);
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            if (exited)
+                goto stop; /* terminated by signal */
+            /* ignored or converted to exit message... */
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_MONITOR_DOWN: {
+            Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+            ErtsExitSignalData *xsigd = NULL;
+            ErtsMonitorData *mdp = NULL;
+            ErtsMonitor *omon = NULL, *tmon = NULL;
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            switch (type) {
+            case ERTS_MON_TYPE_DIST_PROC:
+            case ERTS_MON_TYPE_PROC:
+            case ERTS_MON_TYPE_PORT:
+                tmon = (ErtsMonitor *) sig;
+                ASSERT(erts_monitor_is_target(tmon));
+                ASSERT(!erts_monitor_is_in_table(tmon));
+                mdp = erts_monitor_to_data(tmon);
+                if (erts_monitor_is_in_table(&mdp->origin)) {
+                    omon = &mdp->origin;
+                    erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p),
+                                             omon);
+                    cnt += convert_to_down_message(c_p, sig, mdp,
+                                                   type, next_nm_sig);
+                }
+                break;
+            case ERTS_SIG_Q_TYPE_GEN_EXIT:
+                xsigd = get_exit_signal_data(sig);
+                omon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p),
+                                                xsigd->u.ref);
+                if (omon) {
+                    ASSERT(erts_monitor_is_origin(omon));
+                    if (omon->type == ERTS_MON_TYPE_DIST_PROC) {
+                        mdp = erts_monitor_to_data(omon);
+                        if (erts_monitor_dist_delete(&mdp->target))
+                            tmon = &mdp->target;
+                    }
+                    erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p),
+                                             omon);
+                    cnt += convert_prepared_down_message(c_p, sig,
+                                                         xsigd->message,
+                                                         next_nm_sig);
+                }
+                break;
+            case ERTS_MON_TYPE_NODE:
+                tmon = (ErtsMonitor *) sig;
+                ASSERT(erts_monitor_is_target(tmon));
+                ASSERT(!erts_monitor_is_in_table(tmon));
+                mdp = erts_monitor_to_data(tmon);
+                if (erts_monitor_is_in_table(&mdp->origin)) {
+                    omon = &mdp->origin;
+                    cnt += handle_nodedown(c_p, sig, mdp, next_nm_sig);
+                }
+                break;
+            case ERTS_MON_TYPE_SUSPEND:
+                tmon = (ErtsMonitor *) sig;
+                ASSERT(erts_monitor_is_target(tmon));
+                ASSERT(!erts_monitor_is_in_table(tmon));
+                mdp = erts_monitor_to_data(tmon);
+                if (erts_monitor_is_in_table(&mdp->origin)) {
+                    erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p),
+                                             &mdp->origin);
+                    omon = &mdp->origin;
+                }
+                remove_nm_sig(c_p, sig, next_nm_sig);
+                break;
+            default:
+                ERTS_INTERNAL_ERROR("invalid monitor type");
+                break;
+            }
+
+            if (omon) {
+                if (tmon)
+                    erts_monitor_release_both(mdp);
+                else
+                    erts_monitor_release(omon);
+            }
+            else {
+                remove_nm_sig(c_p, sig, next_nm_sig);
+                if (xsigd) {
+                    sig->next = NULL;
+                    erts_cleanup_messages(sig);
+                }
+                if (tmon)
+                    erts_monitor_release(tmon);
+            }
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_PERSISTENT_MON_MSG: {
+            Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+            ErtsMonitor *mon;
+            Eterm msg;
+            Eterm key;
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            key = get_persist_mon_msg(sig, &msg);
+
+            cnt++;
+            mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p), key);
+            if (mon) {
+                ASSERT(erts_monitor_is_origin(mon));
+                handle_persistent_mon_msg(c_p, type, mon, sig,
+                                          msg, next_nm_sig);
+            }
+            else {
+                cnt++;
+                remove_nm_sig(c_p, sig, next_nm_sig);
+                sig->next = NULL;
+                erts_cleanup_messages(sig);
+            }
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_MONITOR: {
+            ErtsMonitor *mon = (ErtsMonitor *) sig;
+
+            ASSERT(erts_monitor_is_target(mon));
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            remove_nm_sig(c_p, sig, next_nm_sig);
+
+            if (mon->type == ERTS_MON_TYPE_DIST_PROC)
+                erts_monitor_tree_insert(&ERTS_P_MONITORS(c_p), mon);
+            else {
+                erts_monitor_list_insert(&ERTS_P_LT_MONITORS(c_p), mon);
+                if (mon->type == ERTS_MON_TYPE_SUSPEND)
+                    handle_suspend(c_p, mon, &yield);
+            }
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            cnt += 2;
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_DEMONITOR: {
+            Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            remove_nm_sig(c_p, sig, next_nm_sig);
+
+            if (type == ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR) {
+                ErtsMonitor *tmon;
+                ErtsSigDistProcDemonitor *dmon;
+                dmon = (ErtsSigDistProcDemonitor *) sig;
+                tmon = erts_monitor_tree_lookup(ERTS_P_MONITORS(c_p), dmon->ref);
+                destroy_dist_proc_demonitor(dmon);
+                cnt++;
+                if (tmon) {
+                    ErtsMonitorData *mdp = erts_monitor_to_data(tmon);
+                    ASSERT(erts_monitor_is_target(tmon));
+                    erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p), tmon);
+                    if (!erts_monitor_dist_delete(&mdp->origin))
+                        erts_monitor_release(tmon);
+                    else
+                        erts_monitor_release_both(mdp);
+                    cnt += 2;
+                }
+            }
+            else {
+                ErtsMonitor *omon = (ErtsMonitor *) sig;
+                ErtsMonitorData *mdp = erts_monitor_to_data(omon);
+                ASSERT(omon->type == type);
+                ASSERT(erts_monitor_is_origin(omon));
+                ASSERT(!erts_monitor_is_in_table(omon));
+                if (!erts_monitor_is_in_table(&mdp->target))
+                    erts_monitor_release(omon);
+                else {
+                    ErtsMonitor *tmon = &mdp->target;
+                    ASSERT(tmon->type == type);
+                    if (type == ERTS_MON_TYPE_DIST_PROC)
+                        erts_monitor_tree_delete(&ERTS_P_MONITORS(c_p), tmon);
+                    else {
+                        erts_monitor_list_delete(&ERTS_P_LT_MONITORS(c_p), tmon);
+                        switch (type) {
+                        case ERTS_MON_TYPE_RESOURCE:
+                            erts_nif_demonitored((ErtsResource *) tmon->other.ptr);
+                            cnt++;
+                            break;
+                        case ERTS_MON_TYPE_SUSPEND:
+                            erts_resume(c_p, ERTS_PROC_LOCK_MAIN);
+                            break;
+                        default:
+                            break;
+                        }
+                    }
+                    erts_monitor_release_both(mdp);
+                    cnt++;
+                }
+                cnt++;
+            }
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_LINK: {
+            ErtsLink *rlnk, *lnk = (ErtsLink *) sig;
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            rlnk = erts_link_tree_insert_addr_replace(&ERTS_P_LINKS(c_p),
+                                                      lnk);
+            if (!rlnk) {
+                if (tracing.procs)
+                    getting_linked(c_p, lnk->other.item);
+            }
+            else {
+                if (rlnk->type != ERTS_LNK_TYPE_DIST_PROC)
+                    erts_link_release(rlnk);
+                else {
+                    ErtsLinkData *ldp;
+                    ErtsLink *dlnk = erts_link_to_other(rlnk, &ldp);
+                    if (erts_link_dist_delete(dlnk))
+                        erts_link_release_both(ldp);
+                    else
+                        erts_link_release(rlnk);
+                }
+            }
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_UNLINK: {
+            Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+            ErtsLinkData *ldp;
+            ErtsLink *llnk;
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            if (type == ERTS_SIG_Q_TYPE_DIST_LINK) {
+                ErtsSigDistLinkOp *sdlnk = (ErtsSigDistLinkOp *) sig;
+                ASSERT(type == ERTS_SIG_Q_TYPE_DIST_LINK);
+                ASSERT(is_external_pid(sdlnk->remote));
+                llnk = erts_link_tree_lookup(ERTS_P_LINKS(c_p), sdlnk->remote);
+                if (llnk) {
+                    ErtsLink *dlnk = erts_link_to_other(llnk, &ldp);
+                    erts_link_tree_delete(&ERTS_P_LINKS(c_p), llnk);
+                    if (erts_link_dist_delete(dlnk))
+                        erts_link_release_both(ldp);
+                    else
+                        erts_link_release(llnk);
+                    cnt += 8;
+                    if (tracing.procs)
+                        getting_unlinked(c_p, sdlnk->remote);
+                }
+                destroy_sig_dist_link_op(sdlnk);
+                cnt++;
+            }
+            else {
+                ErtsLinkData *ldp;
+                ErtsLink *dlnk, *slnk;
+                slnk = (ErtsLink *) sig;
+                llnk = erts_link_to_other(slnk, &ldp);
+                dlnk = erts_link_tree_key_delete(&ERTS_P_LINKS(c_p), llnk);
+                if (!dlnk)
+                    erts_link_release(slnk);
+                else {
+                    if (tracing.procs)
+                        getting_unlinked(c_p, llnk->other.item);
+                    if (dlnk == llnk)
+                        erts_link_release_both(ldp);
+                    else {
+                        erts_link_release(slnk);
+                        erts_link_release(dlnk);
+                    }
+                }
+                cnt += 2;
+            }
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_GROUP_LEADER: {
+            ErtsSigGroupLeader *sgl = (ErtsSigGroupLeader *) sig;
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            handle_group_leader(c_p, sgl);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_IS_ALIVE:
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            is_alive_response(c_p, sig, !0);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+
+        case ERTS_SIG_Q_OP_PROCESS_INFO:
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            handle_process_info(c_p, &tracing, sig, next_nm_sig, !0);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+
+        case ERTS_SIG_Q_OP_SYNC_SUSPEND:
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            handle_sync_suspend(c_p, sig);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+
+        case ERTS_SIG_Q_OP_RPC:
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            remove_nm_sig(c_p, sig, next_nm_sig);
+            cnt += handle_rpc(c_p, (ErtsProcSigRPC *) sig, cnt,
+                              limit, &yield);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            break;
+
+        case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE: {
+            Uint16 type = ERTS_PROC_SIG_TYPE(tag);
+
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            msg_tracing = handle_trace_change_state(c_p, &tracing,
+                                                    type, sig,
+                                                    next_nm_sig);
+            ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, &tracing, next_nm_sig);
+            
+            break;
+        }
+
+        default:
+            ERTS_INTERNAL_ERROR("Unknown signal");
+            break;
+        }
+
+        cnt++;
+
+    } while (cnt <= limit || stretch_limit(c_p, &tracing, abs_lim, &limit));
+
+stop: {
+        int deferred_save, deferred_saved_last, res;
+
+        deferred_saved_last = !!(c_p->flags & F_DEFERRED_SAVED_LAST);
+        deferred_save = 0;
+
+        if (!deferred_saved_last)
+            deferred_save = 0;
+        else {
+            if (c_p->sig_qs.saved_last == &c_p->sig_qs.cont) {
+                c_p->sig_qs.saved_last = c_p->sig_qs.last;
+                c_p->flags &= ~F_DEFERRED_SAVED_LAST;
+                deferred_saved_last = deferred_save = 0;
+            }
+            else {
+                if (c_p->sig_qs.save == c_p->sig_qs.last)
+                    deferred_save = !0;
+                else
+                    deferred_save = 0;
+            }
+        }
+
+        ASSERT(c_p->sig_qs.saved_last != &c_p->sig_qs.cont);
+
+        if (ERTS_UNLIKELY(msg_tracing != 0)) {
+            /*
+             * All messages that has been traced should
+             * be moved to inner queue. Next signal in
+             * middle queue should either be next message
+             * to trace or next non-message signal.
+             */
+            ASSERT(tracing.messages.next);
+            if (*next_nm_sig) {
+                if (*next_nm_sig == tracing.messages.next)
+                    *next_nm_sig = &c_p->sig_qs.cont;
+                if (c_p->sig_qs.nmsigs.last == tracing.messages.next)
+                    c_p->sig_qs.nmsigs.last = &c_p->sig_qs.cont;
+                *statep = erts_atomic32_read_nob(&c_p->state);
+            }
+            else {
+                ASSERT(!c_p->sig_qs.nmsigs.next);
+                c_p->sig_qs.nmsigs.last = NULL;
+                state = erts_atomic32_read_band_nob(&c_p->state,
+                                                    ~ERTS_PSFLG_SIG_Q);
+                state &= ~ERTS_PSFLG_SIG_Q;
+                *statep = state;
+            }
+
+            if (tracing.messages.next != &c_p->sig_qs.cont) {
+                *c_p->sig_qs.last = c_p->sig_qs.cont;
+                c_p->sig_qs.last = tracing.messages.next;
+
+                c_p->sig_qs.cont = *tracing.messages.next;
+                if (!c_p->sig_qs.cont)
+                    c_p->sig_qs.cont_last = &c_p->sig_qs.cont;
+                *c_p->sig_qs.last = NULL;
+            }
+
+            res = !c_p->sig_qs.cont;
+        }
+        else if (*next_nm_sig) {
+            /*
+             * All messages prior to next non-message
+             * signal should be moved to inner queue.
+             * Next non-message signal to handle should
+             * be first in middle queue.
+             */
+            ASSERT(**next_nm_sig);
+            if (*next_nm_sig != &c_p->sig_qs.cont) {
+                *c_p->sig_qs.last = c_p->sig_qs.cont;
+                c_p->sig_qs.last = *next_nm_sig;
+
+                c_p->sig_qs.cont = **next_nm_sig;
+                if (c_p->sig_qs.nmsigs.last == *next_nm_sig)
+                    c_p->sig_qs.nmsigs.last = &c_p->sig_qs.cont;
+                *next_nm_sig = &c_p->sig_qs.cont;
+                *c_p->sig_qs.last = NULL;
+            }
+
+            ASSERT(c_p->sig_qs.cont);
+
+            *statep = erts_atomic32_read_nob(&c_p->state);
+
+            res = 0;
+        }
+        else {
+            /*
+             * All non-message signals handled. All
+             * messages should be moved to inner queue.
+             * Middle queue should be empty.
+             */
+            ASSERT(!c_p->sig_qs.nmsigs.next);
+            c_p->sig_qs.nmsigs.last = NULL;
+
+            if (c_p->sig_qs.cont_last != &c_p->sig_qs.cont) {
+                ASSERT(!*c_p->sig_qs.last);
+                *c_p->sig_qs.last = c_p->sig_qs.cont;
+                c_p->sig_qs.last = c_p->sig_qs.cont_last;
+                ASSERT(!*c_p->sig_qs.last);
+
+                c_p->sig_qs.cont_last = &c_p->sig_qs.cont;
+                c_p->sig_qs.cont = NULL;
+            }
+
+            ASSERT(!c_p->sig_qs.cont);
+
+            state = erts_atomic32_read_band_nob(&c_p->state,
+                                                ~ERTS_PSFLG_SIG_Q);
+            state &= ~ERTS_PSFLG_SIG_Q;
+            *statep = state;
+            res = !0;
+        }
+
+        if (deferred_saved_last
+            && (c_p->sig_qs.saved_last == &c_p->sig_qs.cont)) {
+            c_p->sig_qs.saved_last = c_p->sig_qs.last;
+            c_p->flags &= ~F_DEFERRED_SAVED_LAST;
+            if (deferred_save)
+                c_p->sig_qs.save = c_p->sig_qs.saved_last;
+        }
+        else if (!res) {
+            if (deferred_save) {
+                c_p->sig_qs.save = c_p->sig_qs.last;
+                ASSERT(!PEEK_MESSAGE(c_p));
+            }
+        }
+        else {
+            c_p->flags &= ~F_DEFERRED_SAVED_LAST;
+            if (deferred_save)
+                c_p->sig_qs.save = c_p->sig_qs.saved_last;
+        }
+
+        ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(c_p, 0);
+
+        *redsp = cnt/4 + 1;
+
+        if (yield) {
+            int vreds = max_reds - *redsp;
+            if (vreds > 0) {
+                ErtsSchedulerData *esdp = erts_get_scheduler_data();
+                esdp->virtual_reds += vreds;
+            }
+            *redsp = max_reds;
+        }
+
+        return res;
+    }
+}
+
+static int
+stretch_limit(Process *c_p, ErtsSigRecvTracing *tp,
+              int abs_lim, int *limp)
+{
+    int lim;
+    /*
+     * Stretch limit up to a maximum of 'abs_lim' if
+     * there currently are no messages available to
+     * inspect by 'receive' and it might be possible
+     * to get messages available by processing
+     * signals (or trace messages).
+     */
+
+    lim = *limp;
+    ASSERT(abs_lim >= lim);
+    if (abs_lim == lim)
+        return 0;
+
+    if (!(c_p->flags & F_DEFERRED_SAVED_LAST)) {
+        ErtsSignal *sig;
+
+        if (PEEK_MESSAGE(c_p))
+            return 0;
+        sig = (ErtsSignal *) c_p->sig_qs.cont;
+        if (!sig)
+            return 0; /* No signals to process available... */
+        if (ERTS_SIG_IS_MSG(sig) && tp->messages.next != &c_p->sig_qs.cont)
+            return 0;
+    }
+
+    lim += ERTS_SIG_REDS_CNT_FACTOR*100;
+    if (lim > abs_lim)
+        lim = abs_lim;
+    *limp = lim;
+    return !0;
+}
+
+
+int
+erts_proc_sig_handle_exit(Process *c_p, int *redsp)
+{
+    int cnt, limit;
+    ErtsMessage *sig, ***next_nm_sig;
+
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(c_p, 0);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+
+    ASSERT(!(ERTS_PSFLG_SIG_IN_Q & erts_atomic32_read_nob(&c_p->state)));
+
+    limit = *redsp;
+    limit *= ERTS_SIG_REDS_CNT_FACTOR;
+
+    *redsp = 1;
+
+    next_nm_sig = &c_p->sig_qs.nmsigs.next;
+
+    if (!*next_nm_sig) {
+        ASSERT(!c_p->sig_qs.nmsigs.last);
+        return !0; /* done... */
+    }
+
+    cnt = 0;
+
+    ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, NULL, next_nm_sig);
+
+    do {
+        Eterm tag;
+        Uint16 type;
+        int op;
+
+        sig = **next_nm_sig;
+
+        ASSERT(sig);
+        ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+        tag = ((ErtsSignal *) sig)->common.tag;
+        type = ERTS_PROC_SIG_TYPE(tag);
+        op = ERTS_PROC_SIG_OP(tag);
+
+        remove_nm_sig(c_p, sig, next_nm_sig);
+ 
+        ERTS_PROC_SIG_HDBG_PRIV_CHKQ(c_p, NULL, next_nm_sig);
+
+        cnt++;
+
+        switch (op) {
+
+        case ERTS_SIG_Q_OP_EXIT:
+        case ERTS_SIG_Q_OP_EXIT_LINKED:
+        case ERTS_SIG_Q_OP_MONITOR_DOWN:
+            switch (type) {
+            case ERTS_SIG_Q_TYPE_GEN_EXIT:
+                sig->next = NULL;
+                erts_cleanup_messages(sig);
+                break;
+            case ERTS_LNK_TYPE_PORT:
+            case ERTS_LNK_TYPE_PROC:
+            case ERTS_LNK_TYPE_DIST_PROC:
+                erts_link_release((ErtsLink *) sig);
+                break;
+            case ERTS_MON_TYPE_PORT:
+            case ERTS_MON_TYPE_PROC:
+            case ERTS_MON_TYPE_DIST_PROC:
+            case ERTS_MON_TYPE_NODE:
+            case ERTS_MON_TYPE_NODES:
+            case ERTS_MON_TYPE_SUSPEND:
+                erts_monitor_release((ErtsMonitor *) sig);
+                break;
+            default:
+                ERTS_INTERNAL_ERROR("Unexpected sig type");
+                break;
+            }
+            break;
+
+        case ERTS_SIG_Q_OP_PERSISTENT_MON_MSG:
+            sig->next = NULL;
+            erts_cleanup_messages(sig);
+            break;
+
+        case ERTS_SIG_Q_OP_MONITOR: {
+            ErtsProcExitContext pectxt = {c_p, am_noproc};
+            erts_proc_exit_handle_monitor((ErtsMonitor *) sig,
+                                          (void *) &pectxt);
+            cnt += 4;
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_DEMONITOR:
+            if (type == ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR)
+                destroy_dist_proc_demonitor((ErtsSigDistProcDemonitor *) sig);
+            else
+                erts_monitor_release((ErtsMonitor *) sig);
+            break;
+
+        case ERTS_SIG_Q_OP_LINK: {
+            ErtsProcExitContext pectxt = {c_p, am_noproc};
+            erts_proc_exit_handle_link((ErtsLink *) sig, (void *) &pectxt);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_UNLINK:
+            if (type == ERTS_SIG_Q_TYPE_DIST_LINK)
+                destroy_sig_dist_link_op((ErtsSigDistLinkOp *) sig);
+            else
+                erts_link_release((ErtsLink *) sig);
+            break;
+
+        case ERTS_SIG_Q_OP_GROUP_LEADER: {
+            ErtsSigGroupLeader *sgl = (ErtsSigGroupLeader *) sig;
+            handle_group_leader(c_p, sgl);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_IS_ALIVE:
+            is_alive_response(c_p, sig, 0);
+            break;
+
+        case ERTS_SIG_Q_OP_PROCESS_INFO:
+            handle_process_info(c_p, NULL, sig, next_nm_sig, 0);
+            break;
+
+        case ERTS_SIG_Q_OP_SYNC_SUSPEND:
+            handle_sync_suspend(c_p, sig);
+            break;
+
+        case ERTS_SIG_Q_OP_RPC: {
+            int yield = 0;
+            handle_rpc(c_p, (ErtsProcSigRPC *) sig,
+                       cnt, limit, &yield);
+            break;
+        }
+
+        case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE:
+            destroy_trace_info((ErtsSigTraceInfo *) sig);
+            break;
+
+        default:
+            ERTS_INTERNAL_ERROR("Unknown signal");
+            break;
+        }
+
+    } while (cnt >= limit && *next_nm_sig);
+
+    *redsp += cnt / ERTS_SIG_REDS_CNT_FACTOR;
+
+    if (*next_nm_sig)
+        return 0;
+
+    ASSERT(!c_p->sig_qs.nmsigs.next);
+    c_p->sig_qs.nmsigs.last = NULL;
+    (void) erts_atomic32_read_band_nob(&c_p->state,
+                                       ~ERTS_PSFLG_SIG_Q);
+    return !0;
+}
+
+#ifdef USE_VM_PROBES
+#  define ERTS_CLEAR_SEQ_TOKEN(MP)                                      \
+    ERL_MESSAGE_TOKEN((MP)) = ((ERL_MESSAGE_DT_UTAG((MP)) != NIL)       \
+                               ? am_have_dt_utag : NIL)
+#else
+#  define ERTS_CLEAR_SEQ_TOKEN(MP)                                      \
+    ERL_MESSAGE_TOKEN((MP)) = NIL
+#endif
+
+static ERTS_INLINE void
+clear_seq_trace_token(ErtsMessage *sig)
+{
+    if (ERTS_SIG_IS_MSG((ErtsSignal *) sig))
+        ERTS_CLEAR_SEQ_TOKEN(sig);
+    else {
+        Uint tag;
+        Uint16 op, type;
+
+        tag = ((ErtsSignal *) sig)->common.tag;
+        type = ERTS_PROC_SIG_TYPE(tag);
+        op = ERTS_PROC_SIG_OP(tag);
+
+        switch (op) {
+
+        case ERTS_SIG_Q_OP_EXIT:
+        case ERTS_SIG_Q_OP_EXIT_LINKED:
+        case ERTS_SIG_Q_OP_MONITOR_DOWN:
+            switch (type) {
+            case ERTS_SIG_Q_TYPE_GEN_EXIT:
+                ERTS_CLEAR_SEQ_TOKEN(sig);
+                break;
+            case ERTS_LNK_TYPE_PORT:
+            case ERTS_LNK_TYPE_PROC:
+            case ERTS_LNK_TYPE_DIST_PROC:
+            case ERTS_MON_TYPE_PORT:
+            case ERTS_MON_TYPE_PROC:
+            case ERTS_MON_TYPE_DIST_PROC:
+            case ERTS_MON_TYPE_NODE:
+                break;
+            default:
+                ERTS_INTERNAL_ERROR("Unexpected sig type");
+                break;
+            }
+            break;
+
+        case ERTS_SIG_Q_OP_PERSISTENT_MON_MSG:
+            ERTS_CLEAR_SEQ_TOKEN(sig);
+            break;
+
+        case ERTS_SIG_Q_OP_MONITOR:
+        case ERTS_SIG_Q_OP_DEMONITOR:
+        case ERTS_SIG_Q_OP_LINK:
+        case ERTS_SIG_Q_OP_UNLINK:
+        case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE:
+            break;
+
+        default:
+            ERTS_INTERNAL_ERROR("Unknown signal");
+            break;
+        }
+    }
+}
+
+void
+erts_proc_sig_clear_seq_trace_tokens(Process *c_p)
+{
+    ASSERT(erts_thr_progress_is_blocking());
+    erts_proc_sig_fetch(c_p);
+    ERTS_FOREACH_SIG_PRIVQS(c_p, sig, clear_seq_trace_token(sig));
+}
+
+Uint
+erts_proc_sig_signal_size(ErtsSignal *sig)
+{
+    Eterm tag;
+    Uint16 type;
+    int op;
+    Uint size = 0;
+
+    ASSERT(sig);
+    ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+    tag = sig->common.tag;
+    type = ERTS_PROC_SIG_TYPE(tag);
+    op = ERTS_PROC_SIG_OP(tag);
+
+    switch (op) {
+    case ERTS_SIG_Q_OP_EXIT:
+    case ERTS_SIG_Q_OP_EXIT_LINKED:
+    case ERTS_SIG_Q_OP_MONITOR_DOWN:
+        switch (type) {
+        case ERTS_SIG_Q_TYPE_GEN_EXIT:
+            size = ((ErtsMessage *) sig)->hfrag.alloc_size;
+            size *= sizeof(Eterm);
+            size += sizeof(ErtsMessage) - sizeof(Eterm);
+            break;
+        case ERTS_LNK_TYPE_PORT:
+        case ERTS_LNK_TYPE_PROC:
+        case ERTS_LNK_TYPE_DIST_PROC:
+            size = erts_link_size((ErtsLink *) sig);
+            break;
+        case ERTS_MON_TYPE_PORT:
+        case ERTS_MON_TYPE_PROC:
+        case ERTS_MON_TYPE_DIST_PROC:
+        case ERTS_MON_TYPE_NODE:
+            size = erts_monitor_size((ErtsMonitor *) sig);
+        default:
+            ERTS_INTERNAL_ERROR("Unexpected sig type");
+            break;
+        }
+        break;
+
+    case ERTS_SIG_Q_OP_SYNC_SUSPEND:
+    case ERTS_SIG_Q_OP_PERSISTENT_MON_MSG:
+    case ERTS_SIG_Q_OP_IS_ALIVE:
+        size = ((ErtsMessage *) sig)->hfrag.alloc_size;
+        size *= sizeof(Eterm);
+        size += sizeof(ErtsMessage) - sizeof(Eterm);
+        break;
+
+    case ERTS_SIG_Q_OP_DEMONITOR:
+        if (type == ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR) {
+            size = NC_HEAP_SIZE(((ErtsSigDistProcDemonitor *) sig)->ref);
+            size--;
+            size *= sizeof(Eterm);
+            size += sizeof(ErtsSigDistProcDemonitor);
+            break;
+        }
+        /* Fall through... */
+
+    case ERTS_SIG_Q_OP_MONITOR:
+        size = erts_monitor_size((ErtsMonitor *) sig);
+        break;
+
+    case ERTS_SIG_Q_OP_UNLINK:
+        if (type == ERTS_SIG_Q_TYPE_DIST_LINK) {
+            size = NC_HEAP_SIZE(((ErtsSigDistLinkOp *) sig)->remote);
+            size--;
+            size *= sizeof(Eterm);
+            size += sizeof(ErtsSigDistLinkOp);
+            break;
+        }
+        /* Fall through... */
+
+    case ERTS_SIG_Q_OP_LINK:
+        size = erts_link_size((ErtsLink *) sig);
+        break;
+
+    case ERTS_SIG_Q_OP_GROUP_LEADER: {
+        ErtsSigGroupLeader *sgl = (ErtsSigGroupLeader *) sig;
+        size = size_object(sgl->group_leader);
+        size += size_object(sgl->ref);
+        size *= sizeof(Eterm);
+        size += sizeof(ErtsSigGroupLeader) - sizeof(Eterm);
+        break;
+    }
+
+    case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE:
+        size = sizeof(ErtsSigTraceInfo);
+        break;
+
+    case ERTS_SIG_Q_OP_PROCESS_INFO: {
+        ErtsProcessInfoSig *pisig = (ErtsProcessInfoSig *) sig;
+        size = sizeof(ErtsProcessInfoSig);
+        size += (pisig->len - 1) * sizeof(int);
+        break;
+    }
+
+    case ERTS_SIG_Q_OP_RPC:
+        size = sizeof(ErtsProcSigRPC);
+        break;
+
+    default:
+        ERTS_INTERNAL_ERROR("Unknown signal");
+        break;
+    }
+
+    return size;
+}
+
+int
+erts_proc_sig_receive_helper(Process *c_p,
+                             int fcalls,
+                             int neg_o_reds,
+                             ErtsMessage **msgpp,
+                             int *get_outp)
+{
+    ErtsMessage *msgp;
+    int reds, consumed_reds, left_reds, max_reds;
+
+    /*
+     * Called from the loop-rec instruction when receive
+     * has reached end of inner (private) queue. This function
+     * tries to move more messages into the inner queue
+     * for the receive to handle. This by, processing the
+     * middle (private) queue and/or moving signals from
+     * the outer (public) queue into the middle queue.
+     *
+     * If this function succeeds in making more messages
+     * available in the inner queue, *msgpp points to next
+     * message. If *msgpp is set to NULL when:
+     * -- process became exiting. *get_outp is set to a
+     *    value greater than zero.
+     * -- process needs to yield. *get_outp is set to a
+     *    value less than zero.
+     * -- no more messages exist in any of the queues.
+     *    *get_outp is set to zero and the message queue
+     *    lock remains locked. This so the process can
+     *    make its way to the appropriate wait instruction
+     *    without racing with new incoming messages.
+     */
+
+    ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    ASSERT(!*msgpp);
+
+    left_reds = fcalls - neg_o_reds;
+    consumed_reds = 0;
+
+    while (!0) {
+        erts_aint32_t state;
+
+        if (!c_p->sig_qs.cont) {
+
+            consumed_reds += 4;
+            left_reds -= 4;
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ);
+            erts_proc_sig_fetch(c_p);
+            /*
+             * Messages may have been moved directly to
+             * inner queue...
+             */
+            msgp = PEEK_MESSAGE(c_p);
+            if (msgp) {
+                erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+                *get_outp = 0;
+                *msgpp = msgp;
+                return consumed_reds;
+            }
+
+            if (!c_p->sig_qs.cont) {
+                /*
+                 * No messages! Return with message queue
+                 * locked and let the process continue
+                 * to wait instruction...
+                 */
+                *get_outp = 0;
+                *msgpp = NULL;
+
+                return consumed_reds;
+            }
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ);
+
+            if (left_reds <= 0)
+                break; /* Yield */
+
+            /* handle newly arrived signals... */
+        }
+
+        reds = ERTS_SIG_HANDLE_REDS_MAX_PREFERED;
+#ifdef DEBUG
+        /* test that it works also with very few reds */
+        max_reds = left_reds;
+        if (reds > left_reds)
+            reds = left_reds;
+#else
+        /* At least work preferred amount of reds... */
+        max_reds = left_reds;
+        if (max_reds < reds)
+            max_reds = reds;
+#endif
+        (void) erts_proc_sig_handle_incoming(c_p, &state, &reds,
+                                             max_reds, !0);
+        consumed_reds += reds;
+        left_reds -= reds;
+
+        /* we may have exited or suspended by an incoming signal... */
+
+        if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_SUSPENDED)) {
+            if (state & ERTS_PSFLG_SUSPENDED)
+                break; /* Yield */
+
+            /*
+             * Process need to schedule out in order
+             * to terminate. Prepare this a bit...
+             */
+            ASSERT(state & ERTS_PSFLG_EXITING);
+            ASSERT(c_p->flags & F_DELAY_GC);
+
+            c_p->flags &= ~F_DELAY_GC;
+            c_p->arity = 0;
+            c_p->current = NULL;
+
+            *get_outp = 1;
+            *msgpp = NULL;
+
+            return consumed_reds;
+        }
+
+        msgp = PEEK_MESSAGE(c_p);
+        if (msgp) {
+            *get_outp = 0;
+            *msgpp = msgp;
+            return consumed_reds;
+        }
+
+        if (left_reds <= 0)
+            break; /* yield */
+
+        ASSERT(!c_p->sig_qs.cont);
+        /* Go fetch again... */
+    }
+    
+    /* Yield... */
+
+    *get_outp = -1;
+    *msgpp = NULL;
+
+    ASSERT(consumed_reds >= (fcalls - neg_o_reds));
+    return consumed_reds;
+}
+
+static int
+handle_trace_change_state(Process *c_p,
+                          ErtsSigRecvTracing *tracing,
+                          Uint16 type,
+                          ErtsMessage *sig,
+                          ErtsMessage ***next_nm_sig)
+{
+    ErtsSigTraceInfo *trace_info = (ErtsSigTraceInfo *) sig;
+    ErtsMessage **next = *next_nm_sig;
+    int msgs_active, old_msgs_active = !!tracing->messages.active;
+
+    ASSERT(sig == *next);
+
+    erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+
+    ERTS_TRACE_FLAGS(c_p) |= trace_info->flags_on;
+    ERTS_TRACE_FLAGS(c_p) &= ~trace_info->flags_off;
+    if (is_value(trace_info->tracer))
+        erts_tracer_replace(&c_p->common, trace_info->tracer);
+
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+
+    remove_nm_sig(c_p, sig, next_nm_sig);
+    destroy_trace_info(trace_info);
+    /*
+     * Adjust tracing state according to modifications made by
+     * the trace info signal...
+     */
+    adjust_tracing_state(c_p, tracing, 0);
+    msgs_active = !!tracing->messages.active;
+
+    if (old_msgs_active ^ msgs_active) {
+        if (msgs_active) {
+            ASSERT(!tracing->messages.next);
+            tracing->messages.next = next;
+        }
+        else {
+            ASSERT(tracing->messages.next);
+            tracing->messages.next = NULL;
+        }
+    }
+
+    ASSERT(!msgs_active || tracing->messages.next);
+
+    return msgs_active;
+}
+
+static void
+getting_unlinked(Process *c_p, Eterm unlinker)
+{
+    trace_proc(c_p, ERTS_PROC_LOCK_MAIN, c_p,
+               am_getting_unlinked, unlinker);
+}
+
+static void
+getting_linked(Process *c_p, Eterm linker)
+{
+    trace_proc(c_p, ERTS_PROC_LOCK_MAIN, c_p,
+               am_getting_linked, linker);
+}
+
+static ERTS_INLINE void
+handle_message_enqueued_tracing(Process *c_p,
+                                ErtsSigRecvTracing *tracing,
+                                ErtsMessage *msg)
+{
+    ASSERT(ERTS_SIG_IS_INTERNAL_MSG(msg));
+
+#if defined(USE_VM_PROBES)
+    if (tracing->messages.vm_probes && DTRACE_ENABLED(message_queued)) {
+        Sint tok_label = 0;
+        Sint tok_lastcnt = 0;
+        Sint tok_serial = 0;
+        Sint len = erts_proc_sig_privqs_len(c_p);
+        Eterm seq_trace_token = ERL_MESSAGE_TOKEN(msg);
+
+        if (seq_trace_token != NIL && is_tuple(seq_trace_token)) {
+            tok_label = SEQ_TRACE_T_DTRACE_LABEL(seq_trace_token);
+            tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(seq_trace_token));
+            tok_serial = signed_val(SEQ_TRACE_T_SERIAL(seq_trace_token));
+        }
+        /* Message intentionally not passed... */
+        DTRACE6(message_queued,
+                tracing->messages.receiver_name,
+                size_object(ERL_MESSAGE_TERM(msg)),
+                len, /* This is NOT message queue len, but its something... */
+                tok_label, tok_lastcnt, tok_serial);
+    }
+#endif
+
+    if (tracing->messages.receive_trace && tracing->messages.event->on) {
+        ASSERT(IS_TRACED(c_p));
+        trace_receive(c_p,
+                      ERL_MESSAGE_FROM(msg),
+                      ERL_MESSAGE_TERM(msg),
+                      tracing->messages.event);
+    }
+}
+
+static int
+handle_msg_tracing(Process *c_p, ErtsSigRecvTracing *tracing,
+                   ErtsMessage ***next_nm_sig)
+{
+    ErtsMessage **next_sig, *sig;
+    int cnt = 0, limit = ERTS_PROC_SIG_TRACE_COUNT_LIMIT;
+
+    ASSERT(tracing->messages.next);
+    next_sig = tracing->messages.next;
+    sig = *next_sig;
+
+    /*
+     * Receive tracing active. Handle all messages
+     * until next non-message signal...
+     */
+
+    while (sig && ERTS_SIG_IS_MSG(sig)) {
+        if (cnt > limit) {
+            tracing->messages.next = next_sig;
+            return -1; /* Yield... */
+        }
+        if (ERTS_SIG_IS_EXTERNAL_MSG(sig)) {
+            cnt++;
+            if (!erts_decode_dist_message(c_p, ERTS_PROC_LOCK_MAIN,
+                                          sig, 0)) {
+                /* Bad dist message; remove it... */
+                remove_mq_m_sig(c_p, sig, next_sig, next_nm_sig);
+                sig = *next_sig;
+                continue;
+            }
+        }
+        handle_message_enqueued_tracing(c_p, tracing, sig);
+        cnt++;
+
+        next_sig = &(*next_sig)->next;
+        sig = *next_sig;
+    }
+
+    tracing->messages.next = next_sig;
+
+    if (!sig) {
+        ASSERT(!*next_nm_sig);
+        return 1; /* end... */
+    }
+
+    ASSERT(*next_nm_sig);
+    ASSERT(**next_nm_sig == sig);
+
+    /* Next signal a non-message signal... */
+    ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+    /*
+     * Return and handle the non-message signal...
+     */
+
+    return 0;
+}
+
+Uint
+erts_proc_sig_prep_msgq_for_inspection(Process *c_p,
+                                       Process *rp,
+                                       ErtsProcLocks rp_locks,
+                                       int info_on_self,
+                                       ErtsMessageInfo *mip)
+{
+    Uint tot_heap_size;
+    ErtsMessage *mp, **mpp;
+    Sint i;
+    int self_on_heap;
+
+    /*
+     * Prepare the message queue (inner signal queue)
+     * for inspection by process_info().
+     *
+     * - Decode all messages on external format
+     * - Remove all corrupt dist messages from queue
+     * - Save pointer to, and heap size need of each
+     *   message in the mip array.
+     * - Return total heap size need for all messages
+     *   that needs to be copied.
+     *
+     */
+
+    ASSERT(!info_on_self || c_p == rp);
+
+    self_on_heap = info_on_self && !(c_p->flags & F_OFF_HEAP_MSGQ);
+
+    tot_heap_size = 0;
+    i = 0;
+    mpp = &rp->sig_qs.first;
+    mp = rp->sig_qs.first;
+    while (mp) {
+	Eterm msg = ERL_MESSAGE_TERM(mp);
+
+	mip[i].size = 0;
+
+	if (ERTS_SIG_IS_EXTERNAL_MSG(mp)) {
+	    /* decode it... */
+	    if (mp->data.attached)
+		erts_decode_dist_message(rp, rp_locks, mp, !0);
+
+	    msg = ERL_MESSAGE_TERM(mp);
+
+	    if (is_non_value(msg)) {
+		ErtsMessage *bad_mp = mp;
+		/*
+		 * Bad distribution message; remove
+		 * it from the queue...
+		 */
+		ASSERT(!mp->data.attached);
+
+		ASSERT(*mpp == bad_mp);
+
+                remove_iq_m_sig(rp, mp, mpp);
+
+                mp = *mpp;
+
+		bad_mp->next = NULL;
+		erts_cleanup_messages(bad_mp);
+		continue;
+	    }
+	}
+
+	ASSERT(is_value(msg));
+
+	if (is_not_immed(msg) && (!self_on_heap || mp->data.attached)) {
+	    Uint sz = size_object(msg);
+	    mip[i].size = sz;
+	    tot_heap_size += sz;
+	}
+
+	mip[i].msgp = mp;
+	i++;
+        mpp = &mp->next;
+	mp = mp->next;
+    }
+
+    ASSERT(c_p->sig_qs.len == i);
+
+    return tot_heap_size;
+}
+
+static ERTS_INLINE void
+move_msg_to_heap(Process *c_p, ErtsMessage *mp)
+{
+    /*
+     * We leave not yet decoded distribution messages
+     * as they are in the queue since it is not
+     * possible to determine a maximum size until
+     * actual decoding...
+     *
+     * We also leave combined messages as they are...
+     */
+    if (ERTS_SIG_IS_INTERNAL_MSG(mp)
+        && mp->data.attached
+        && mp->data.attached != ERTS_MSG_COMBINED_HFRAG) {
+        ErlHeapFragment *bp;
+
+        bp = erts_message_to_heap_frag(mp);
+
+        if (bp->next)
+            erts_move_multi_frags(&c_p->htop, &c_p->off_heap, bp,
+                                  mp->m, ERL_MESSAGE_REF_ARRAY_SZ, 0);
+        else
+            erts_copy_one_frag(&c_p->htop, &c_p->off_heap, bp,
+                               mp->m, ERL_MESSAGE_REF_ARRAY_SZ);
+
+        mp->data.heap_frag = NULL;
+        free_message_buffer(bp);
+    }
+}
+
+void
+erts_proc_sig_move_msgs_to_heap(Process *c_p)
+{
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(c_p, 0);
+
+    ERTS_FOREACH_SIG_PRIVQS(c_p, sig, move_msg_to_heap(c_p, sig));
+
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(c_p, 0);
+}
+
+
+BIF_RETTYPE
+erts_internal_dirty_process_handle_signals_1(BIF_ALIST_1)
+{
+    erts_aint32_t state, dirty, noproc;
+    int busy;
+    Process *rp;
+
+    if (BIF_P != erts_dirty_process_signal_handler
+        && BIF_P != erts_dirty_process_signal_handler_high
+        && BIF_P != erts_dirty_process_signal_handler_max)
+        BIF_ERROR(BIF_P, EXC_NOTSUP);
+
+    if (is_not_internal_pid(BIF_ARG_1))
+        BIF_RET(am_false);
+
+    rp = erts_proc_lookup_raw(BIF_ARG_1);
+    if (!rp)
+        BIF_RET(am_noproc);
+
+    state = erts_atomic32_read_nob(&rp->state);
+    dirty = (state & (ERTS_PSFLG_DIRTY_RUNNING
+                      | ERTS_PSFLG_DIRTY_RUNNING_SYS));
+    if (!dirty)
+        BIF_RET(am_normal);
+
+    busy = erts_proc_trylock(rp, ERTS_PROC_LOCK_MAIN) == EBUSY;
+
+    state = erts_atomic32_read_mb(&rp->state);
+    noproc = (state & ERTS_PSFLG_FREE);
+    dirty = (state & (ERTS_PSFLG_DIRTY_RUNNING
+                      | ERTS_PSFLG_DIRTY_RUNNING_SYS));
+
+    if (busy) {
+        if (noproc)
+            BIF_RET(am_noproc);
+        if (dirty)
+            BIF_RET(am_more); /* try again... */
+        BIF_RET(am_normal); /* will handle signals itself... */
+    }
+    else {
+        erts_aint32_t state;
+        int done;
+        Eterm res = am_false;
+        int reds = 0;
+
+        if (noproc)
+            res = am_noproc;
+        else if (!dirty)
+            res = am_normal; /* will handle signals itself... */
+        else {
+            reds = ERTS_BIF_REDS_LEFT(BIF_P);
+            done = erts_proc_sig_handle_incoming(rp, &state, &reds,
+                                                 reds, 0);
+            if (done || (state & ERTS_PSFLG_EXITING))
+                res = am_ok;
+            else
+                res = am_more;
+        }
+
+        erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN);
+
+        if (reds)
+            BUMP_REDS(BIF_P, reds);
+
+        BIF_RET(res);
+    }
+}
+
+static void
+debug_foreach_sig_heap_frags(ErlHeapFragment *hfrag,
+                             void (*oh_func)(ErlOffHeap *, void *),
+                             void *arg)
+{
+    ErlHeapFragment *hf = hfrag;
+    while (hf) {
+        oh_func(&(hf->off_heap), arg);
+        hf = hf->next;
+    }
+}
+
+static void
+debug_foreach_sig_fake_oh(Eterm term,
+                          void (*oh_func)(ErlOffHeap *, void *),
+                          void *arg)
+{
+    if (is_external(term)) {
+        ErlOffHeap oh;
+        oh.overhead = 0;
+        oh.first = ((struct erl_off_heap_header *)
+                    (char *) external_thing_ptr(term));
+        ASSERT(!oh.first->next);
+        oh_func(&oh, arg);
+    }
+
+}
+
+void
+erts_proc_sig_debug_foreach_sig(Process *c_p,
+                                void (*msg_func)(ErtsMessage *, void *),
+                                void (*oh_func)(ErlOffHeap *, void *),
+                                void (*mon_func)(ErtsMonitor *, void *),
+                                void (*lnk_func)(ErtsLink *, void *),
+                                void *arg)
+{
+    ErtsMessage *queue[] = {c_p->sig_qs.first, c_p->sig_qs.cont, c_p->sig_inq.first};
+    int qix;
+
+    for (qix = 0; qix < sizeof(queue)/sizeof(queue[0]); qix++) {
+        ErtsMessage *sig;
+        for (sig = queue[qix]; sig; sig = sig->next) {
+            
+            if (ERTS_SIG_IS_MSG(sig))
+                msg_func(sig, arg);
+            else {
+                Eterm tag;
+                Uint16 type;
+                int op;
+
+                ASSERT(sig);
+                ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+                tag = ((ErtsSignal *) sig)->common.tag;
+                type = ERTS_PROC_SIG_TYPE(tag);
+                op = ERTS_PROC_SIG_OP(tag);
+
+                switch (op) {
+
+                case ERTS_SIG_Q_OP_EXIT:
+                case ERTS_SIG_Q_OP_EXIT_LINKED:
+                case ERTS_SIG_Q_OP_MONITOR_DOWN:
+                    switch (type) {
+                    case ERTS_SIG_Q_TYPE_GEN_EXIT:
+                        debug_foreach_sig_heap_frags(&sig->hfrag, oh_func, arg);
+                        break;
+                    case ERTS_LNK_TYPE_PORT:
+                    case ERTS_LNK_TYPE_PROC:
+                    case ERTS_LNK_TYPE_DIST_PROC:
+                        lnk_func((ErtsLink *) sig, arg);
+                        break;
+                    case ERTS_MON_TYPE_PORT:
+                    case ERTS_MON_TYPE_PROC:
+                    case ERTS_MON_TYPE_DIST_PROC:
+                    case ERTS_MON_TYPE_NODE:
+                        mon_func((ErtsMonitor *) sig, arg);
+                        break;
+                    default:
+                        ERTS_INTERNAL_ERROR("Unexpected sig type");
+                        break;
+                    }
+                    break;
+
+                case ERTS_SIG_Q_OP_PERSISTENT_MON_MSG:
+                    debug_foreach_sig_heap_frags(&sig->hfrag, oh_func, arg);
+                    break;
+
+                case ERTS_SIG_Q_OP_DEMONITOR:
+                    if (type == ERTS_SIG_Q_TYPE_DIST_PROC_DEMONITOR) {
+                        debug_foreach_sig_fake_oh(((ErtsSigDistProcDemonitor *) sig)->ref,
+                                                  oh_func, arg);
+                        break;
+                    }
+                    /* Fall through... */
+
+                case ERTS_SIG_Q_OP_MONITOR:
+                    mon_func((ErtsMonitor *) sig, arg);
+                    break;
+
+                case ERTS_SIG_Q_OP_UNLINK:
+                    if (type == ERTS_SIG_Q_TYPE_DIST_LINK) {
+                        debug_foreach_sig_fake_oh(((ErtsSigDistLinkOp *) sig)->remote,
+                                                  oh_func, arg);
+                        break;
+                    }
+                    /* Fall through... */
+
+                case ERTS_SIG_Q_OP_LINK:
+                    lnk_func((ErtsLink *) sig, arg);
+                    break;
+
+                case ERTS_SIG_Q_OP_GROUP_LEADER: {
+                    ErtsSigGroupLeader *sgl = (ErtsSigGroupLeader *) sig;
+                    oh_func(&sgl->oh, arg);
+                    break;
+                }
+
+                case ERTS_SIG_Q_OP_IS_ALIVE:
+                case ERTS_SIG_Q_OP_TRACE_CHANGE_STATE:
+                case ERTS_SIG_Q_OP_PROCESS_INFO:
+                    break;
+
+                default:
+                    ERTS_INTERNAL_ERROR("Unknown signal");
+                    break;
+                }
+
+            }
+        }
+    }
+}
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+
+static void
+chk_eterm(Process *c_p, int privq, ErtsMessage *mp, Eterm term)
+{
+    ErlHeapFragment *bp;
+    Eterm *ptr = NULL;
+
+    switch (primary_tag(term)) {
+    case TAG_PRIMARY_IMMED1:
+        return;
+    case TAG_PRIMARY_LIST:
+        ptr = list_val(term);
+        ERTS_ASSERT(!is_header(CAR(ptr)));
+        ERTS_ASSERT(!is_header(CDR(ptr)));
+        break;
+    case TAG_PRIMARY_BOXED:
+        ptr = boxed_val(term);
+        ERTS_ASSERT(is_header(*ptr));
+        break;
+    case TAG_PRIMARY_HEADER:
+    default:
+        ERTS_INTERNAL_ERROR("Not valid term");
+        break;
+    }
+
+    if (erts_is_literal(term, ptr))
+        return;
+
+    for (bp = erts_message_to_heap_frag(mp); bp; bp = bp->next) {
+        if (bp->mem <= ptr && ptr < bp->mem + bp->used_size)
+            return;
+    }
+
+    ASSERT(erts_dbg_within_proc(ptr, c_p, NULL));
+}
+
+static Sint
+proc_sig_hdbg_check_queue(Process *proc,
+                          int privq,
+                          ErtsMessage **sig_next,
+                          ErtsMessage **sig_last,
+                          ErtsMessage **sig_nm_next,
+                          ErtsMessage **sig_nm_last,
+                          ErtsSigRecvTracing *tracing,
+                          int *found_saved_last_p,
+                          erts_aint32_t sig_psflg)
+{
+    ErtsMessage **next, *sig, **nm_next, **nm_last;
+    int last_nm_sig_found, nm_sigs = 0, found_next_trace = 0,
+        found_save = 0, last_sig_found = 0, found_saved_last = 0;
+    Sint msg_len = 0;
+    ErtsMessage **next_trace = tracing ? tracing->messages.next : NULL;
+    ErtsMessage **save = proc->sig_qs.save;
+    ErtsMessage **saved_last = proc->sig_qs.saved_last;
+
+    if (!privq) {
+        ErtsSignal *sig = (ErtsSignal *) *sig_next;
+        if (sig->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK) {
+
+        }
+    }
+
+    nm_next = sig_nm_next;
+    nm_last = sig_nm_last;
+    next = sig_next;
+    sig = *sig_next;
+
+    last_nm_sig_found = !nm_last;
+    if (last_nm_sig_found)
+        ERTS_ASSERT(!nm_next);
+    else
+        ERTS_ASSERT(nm_next);
+
+    while (1) {
+        ErtsSignal *nm_sig;
+
+        if (next == sig_last) {
+            ASSERT(!*next);
+            last_sig_found = 1;
+        }
+
+        if (next == save)
+            found_save = 1;
+
+        if (next == saved_last)
+            found_saved_last = 1;
+
+        if (next == next_trace) {
+            found_next_trace = 1;
+            ERTS_ASSERT(nm_sigs == 0);
+        }
+
+        while (sig && ERTS_SIG_IS_MSG(sig)) {
+            int i;
+            if (ERTS_SIG_IS_EXTERNAL_MSG(sig))
+                i = 1;
+            else
+                i = 0;
+            for (; i < ERL_MESSAGE_REF_ARRAY_SZ; i++)
+                chk_eterm(proc, privq, sig, sig->m[i]);
+
+            msg_len++;
+            next = &sig->next;
+            sig = sig->next;
+
+            if (next == sig_last) {
+                ASSERT(!*next);
+                last_sig_found = 1;
+            }
+
+            if (next == save)
+                found_save = 1;
+
+            if (next == saved_last)
+                found_saved_last = 1;
+
+            if (next == next_trace) {
+                found_next_trace = 1;
+                ERTS_ASSERT(nm_sigs == 0);
+            }
+        }
+
+        if (!sig)
+            break;
+
+        nm_sig = (ErtsSignal *) sig;
+
+        if (nm_sig->common.tag == ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK) {
+            ERTS_ASSERT(!privq);
+            ERTS_ASSERT(sig == *sig_next);
+        }
+        else {
+            nm_sigs++;
+
+            ERTS_ASSERT(!last_nm_sig_found);
+            ERTS_ASSERT(ERTS_SIG_IS_NON_MSG(sig));
+
+            ERTS_ASSERT(nm_next == next);
+
+            if (nm_last == next) {
+                ASSERT(!nm_sig->common.specific.next);
+                last_nm_sig_found = 1;
+            }
+
+            nm_next = nm_sig->common.specific.next;
+
+        }
+
+        next = &nm_sig->common.next;
+        sig = nm_sig->common.next;
+
+    }
+
+    if (!privq) {
+        /* outer queue */
+        ERTS_ASSERT(!found_save);
+        ERTS_ASSERT(!found_saved_last);
+    }
+    else if (privq > 0) {
+        /* middle queue */
+        ERTS_ASSERT(!next_trace || found_next_trace);
+        ERTS_ASSERT(!found_save);
+        if (!found_saved_last_p) {
+            ERTS_ASSERT(!found_saved_last
+                        || (proc->flags & F_DEFERRED_SAVED_LAST));
+        }
+        else {
+            if (*found_saved_last_p) {
+                ERTS_ASSERT(!found_saved_last);
+                ERTS_ASSERT(!(proc->flags & F_DEFERRED_SAVED_LAST));
+            }
+            else if (saved_last) {
+                ERTS_ASSERT(found_saved_last);
+                ERTS_ASSERT(proc->flags & F_DEFERRED_SAVED_LAST);
+            }
+            *found_saved_last_p |= found_saved_last;
+        }
+    }
+    else {
+        /* inner queue */
+        ERTS_ASSERT(!found_next_trace);
+        ERTS_ASSERT(nm_sigs == 0);
+        ERTS_ASSERT(found_save);
+        ERTS_ASSERT(!saved_last
+                    || (found_saved_last
+                        || (proc->flags & F_DEFERRED_SAVED_LAST)));
+        if (found_saved_last_p)
+            *found_saved_last_p |= found_saved_last;
+    }
+
+    ERTS_ASSERT(last_nm_sig_found);
+    ERTS_ASSERT(last_sig_found);
+
+    if (sig_psflg != ERTS_PSFLG_FREE) {
+        erts_aint32_t state = erts_atomic32_read_nob(&proc->state);
+        ERTS_ASSERT(nm_sigs ? !!(state & sig_psflg) : !(state & sig_psflg));
+    }
+
+    return msg_len;
+}
+
+void
+erts_proc_sig_hdbg_check_priv_queue(Process *p, int qlock, char *what, char *file, int line)
+{
+    int found_saved_last = 0;
+    Sint len, len1, len2;
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking()
+                   || ERTS_PROC_IS_EXITING(p)
+                   || (ERTS_PROC_LOCK_MAIN
+                       & erts_proc_lc_my_proc_locks(p)));
+    len1 = proc_sig_hdbg_check_queue(p,
+                                     -1,
+                                     &p->sig_qs.first,
+                                     p->sig_qs.last,
+                                     NULL,
+                                     NULL,
+                                     NULL,
+                                     &found_saved_last,
+                                     ERTS_PSFLG_FREE);
+    len2 = proc_sig_hdbg_check_queue(p,
+                                     1,
+                                     &p->sig_qs.cont,
+                                     p->sig_qs.cont_last,
+                                     p->sig_qs.nmsigs.next,
+                                     p->sig_qs.nmsigs.last,
+                                     NULL,
+                                     &found_saved_last,
+                                     ERTS_PSFLG_SIG_Q);
+    if (p->sig_qs.saved_last)
+        ERTS_ASSERT(found_saved_last);
+    len = proc_sig_privqs_len(p, qlock);
+    ERTS_ASSERT(len == len1 + len2);
+}
+
+void
+erts_proc_sig_hdbg_check_in_queue(Process *p, char *what, char *file, int line)
+{
+    Sint len;
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking()
+                   || ERTS_PROC_IS_EXITING(p)
+                   || (ERTS_PROC_LOCK_MSGQ
+                       & erts_proc_lc_my_proc_locks(p)));
+    len = proc_sig_hdbg_check_queue(p,
+                                    0,
+                                    &p->sig_inq.first,
+                                    p->sig_inq.last,
+                                    p->sig_inq.nmsigs.next,
+                                    p->sig_inq.nmsigs.last,
+                                    NULL,
+                                    NULL,
+                                    ERTS_PSFLG_SIG_IN_Q);
+    ASSERT(p->sig_inq.len == len); (void)len;
+}
+
+#endif /* ERTS_PROC_SIG_HARD_DEBUG */
diff --git a/erts/emulator/beam/erl_proc_sig_queue.h b/erts/emulator/beam/erl_proc_sig_queue.h
new file mode 100644
index 0000000000..3fc2d06b2d
--- /dev/null
+++ b/erts/emulator/beam/erl_proc_sig_queue.h
@@ -0,0 +1,1051 @@
+/*
+ * %CopyrightBegin%
+ * 
+ * Copyright Ericsson AB 2018. All Rights Reserved.
+ * 
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ * 
+ * %CopyrightEnd%
+ */
+
+/*
+ * Description:	Process signal queue implementation.
+ *
+ *              Currently the following signals are handled:
+ *              - Messages
+ *              - Exit
+ *              - Monitor
+ *              - Demonitor
+ *              - Monitor down
+ *              - Persistent monitor message
+ *              - Link
+ *              - Unlink
+ *              - Group leader
+ *              - Is process alive
+ *              - Process info request
+ *              - Suspend request (monitor of suspend type)
+ *              - Resume request (demonitor of suspend type)
+ *              - Suspend cleanup (monitor down of suspend type)
+ *              - Sync suspend
+ *              - RPC request
+ *              - Trace change
+ *
+ *              The signal queue consists of three parts:
+ *              - Outer queue (sig_inq field in process struct)
+ *              - Middle queue (sig_qs field in process struct)
+ *              - Inner queue (sig_qs field in process struct)
+ *
+ *              Incoming signals are placed in the outer queue
+ *              by other processes, ports, or by the runtime system
+ *              itself. This queue is protected by the msgq process
+ *              lock and may be accessed by any other entity. While
+ *              a signal is located in the outer queue, it is still
+ *              in transit between sender and receiver.
+ *
+ *              The middle and the inner queues are private to the
+ *              receiving process and can only be accessed while
+ *              holding the main process lock. The signal changes
+ *              from being in transit to being received while in
+ *              the middle queue. Non-message signals are handled
+ *              immediately upon reception while message signals
+ *              are moved into the inner queue.
+ *
+ *              In the outer and middle queues both message signals
+ *              and non-message signals are mixed. Signals in these
+ *              queues are referenced using two single linked lists.
+ *              One single linked list that go through all signals
+ *              in the queue and another single linked list that
+ *              goes through only non-message signals. The list
+ *              through the non-message signals is used for fast
+ *              access to these signals in the middle queue, since
+ *              these should be handled immediately upon reception.
+ *
+ *              The inner queue consists only of one single linked
+ *              list through the message signals. A receive
+ *              expression can only operate on messages once they
+ *              have entered the inner queue.
+ *
+ * Author: 	Rickard Green
+ */
+
+#ifndef ERTS_PROC_SIG_QUEUE_H_TYPE__
+#define ERTS_PROC_SIG_QUEUE_H_TYPE__
+
+#if 0
+#  define ERTS_PROC_SIG_HARD_DEBUG
+#endif
+#if 0
+#  define ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+#endif
+
+struct erl_mesg;
+
+typedef struct {
+    struct erl_mesg *next;
+    union {
+        struct erl_mesg **next;
+        void *attachment;
+    } specific;
+    Eterm tag;
+} ErtsSignalCommon;
+
+#define ERTS_SIG_HANDLE_REDS_MAX_PREFERED (CONTEXT_REDS/40)
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+#  define ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE(P)       \
+    ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE__((P), "")
+#  define ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(P, QL) \
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE__((P), (QL), "")
+#  define ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE__(P, What)   \
+    erts_proc_sig_hdbg_check_in_queue((P), (What), __FILE__, __LINE__)
+#  define ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE__(P, QL, What)              \
+    erts_proc_sig_hdbg_check_priv_queue((P), (QL), (What), __FILE__, __LINE__)
+struct process;
+void erts_proc_sig_hdbg_check_priv_queue(struct process *c_p, int qlock,
+                                         char *what, char *file, int line);
+void erts_proc_sig_hdbg_check_in_queue(struct process *c_p, char *what,
+                                       char *file, int line);
+#else
+#  define ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE(P)
+#  define ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(P, QL)
+#  define ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE__(P, What)
+#define ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE__(P, QL, What)
+#endif
+
+#endif
+
+#if !defined(ERTS_PROC_SIG_QUEUE_H__) && !defined(ERTS_PROC_SIG_QUEUE_TYPE_ONLY)
+#define ERTS_PROC_SIG_QUEUE_H__
+
+#define ERTS_SIG_Q_OP_BITS      8                      
+#define ERTS_SIG_Q_OP_SHIFT     0
+#define ERTS_SIG_Q_OP_MASK      ((1 << ERTS_SIG_Q_OP_BITS) - 1)
+
+#define ERTS_SIG_Q_TYPE_BITS    8
+#define ERTS_SIG_Q_TYPE_SHIFT   ERTS_SIG_Q_OP_BITS
+#define ERTS_SIG_Q_TYPE_MASK    ((1 << ERTS_SIG_Q_TYPE_BITS) - 1)
+
+#define ERTS_SIG_Q_NON_X_BITS__ (_HEADER_ARITY_OFFS \
+                                 + ERTS_SIG_Q_OP_BITS \
+                                 + ERTS_SIG_Q_TYPE_BITS)
+
+#define ERTS_SIG_Q_XTRA_BITS    (32 - ERTS_SIG_Q_NON_X_BITS__)
+#define ERTS_SIG_Q_XTRA_SHIFT   (ERTS_SIG_Q_OP_BITS \
+                                 + ERTS_SIG_Q_TYPE_BITS)
+#define ERTS_SIG_Q_XTRA_MASK    ((1 << ERTS_SIG_Q_XTRA_BITS) - 1)
+
+
+#define ERTS_PROC_SIG_OP(Tag) \
+    ((int) (_unchecked_thing_arityval((Tag)) \
+            >> ERTS_SIG_Q_OP_SHIFT) & ERTS_SIG_Q_OP_MASK)
+
+#define ERTS_PROC_SIG_TYPE(Tag) \
+    ((Uint16) (_unchecked_thing_arityval((Tag)) \
+               >> ERTS_SIG_Q_TYPE_SHIFT) & ERTS_SIG_Q_TYPE_MASK)
+
+#define ERTS_PROC_SIG_XTRA(Tag) \
+    ((Uint32) (_unchecked_thing_arityval((Tag)) \
+               >> ERTS_SIG_Q_XTRA_SHIFT) & ERTS_SIG_Q_XTRA_MASK)
+
+#define ERTS_PROC_SIG_MAKE_TAG(Op, Type, Xtra)                  \
+    (ASSERT(0 <= (Xtra) && (Xtra) <= ERTS_SIG_Q_XTRA_MASK),     \
+     _make_header((((Type) & ERTS_SIG_Q_TYPE_MASK)              \
+                   << ERTS_SIG_Q_TYPE_SHIFT)                    \
+                  | (((Op) & ERTS_SIG_Q_OP_MASK)                \
+                     << ERTS_SIG_Q_OP_SHIFT)                    \
+                  | (((Xtra) & ERTS_SIG_Q_XTRA_MASK)            \
+                     << ERTS_SIG_Q_XTRA_SHIFT),                 \
+                  _TAG_HEADER_EXTERNAL_PID))
+
+
+/*
+ * ERTS_SIG_Q_OP_MSGQ_LEN_OFFS_MARK is not an actual
+ * operation. We keep it at the top of the OP range,
+ * larger than ERTS_SIG_Q_OP_MAX.
+ */
+#define ERTS_SIG_Q_OP_MSGQ_LEN_OFFS_MARK ERTS_SIG_Q_OP_MASK
+
+#define ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK \
+    ERTS_PROC_SIG_MAKE_TAG(ERTS_SIG_Q_OP_MSGQ_LEN_OFFS_MARK,0,0)
+
+struct dist_entry_;
+
+/*
+ * Send operations of currently supported process signals follow...
+ */
+
+/**
+ *
+ * @brief Send an exit signal to a process.
+ *
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     to            Identifier of local process
+ *                              to send signal to.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ * @param[in]     token         Seq trace token.
+ *
+ * @param[in]     normal_kills  If non-zero, also normal exit
+ *                              reason will kill the receiver
+ *                              if it is not trapping exit.
+ *
+ */
+void
+erts_proc_sig_send_exit(Process *c_p, Eterm from, Eterm to,
+                        Eterm reason, Eterm token, int normal_kills);
+
+/**
+ *
+ * @brief Send an exit signal due to broken link to a process.
+ *
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     lnk           Pointer to link structure
+ *                              from the sending side. It
+ *                              should contain information
+ *                              about receiver.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ * @param[in]     token         Seq trace token.
+ *
+ */
+void
+erts_proc_sig_send_link_exit(Process *c_p, Eterm from, ErtsLink *lnk,
+                             Eterm reason, Eterm token);
+
+/**
+ *
+ * @brief Send an link signal to a process.
+ *
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     lnk           Pointer to link structure to
+ *                              insert on receiver side.
+ *
+ * @return                      A non-zero value if
+ *                              signal was successfully
+ *                              sent. If a zero, value
+ *                              the signal was not sent
+ *                              due to the receiver not
+ *                              existing. The sender
+ *                              needs to deallocate the
+ *                              link structure.
+ *
+ */
+int
+erts_proc_sig_send_link(Process *c_p, Eterm to, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Send an unlink signal to a process.
+ *
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     lnk           Pointer to link structure from
+ *                              the sending side. It should
+ *                              contain information about
+ *                              receiver.
+ */
+void
+erts_proc_sig_send_unlink(Process *c_p, ErtsLink *lnk);
+
+/**
+ *
+ * @brief Send an exit signal due to broken link to a process.
+ *
+ * This function is used instead of erts_proc_sig_send_link_exit()
+ * when the signal arrives via the distribution and
+ * no link structure is available.
+ *
+ * @param[in]     dep           Distribution entry of channel
+ *                              that the signal arrived on.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ * @param[in]     token         Seq trace token.
+ *
+ */
+void
+erts_proc_sig_send_dist_link_exit(struct dist_entry_ *dep,
+                                  Eterm from, Eterm to,
+                                  Eterm reason, Eterm token);
+
+/**
+ *
+ * @brief Send an unlink signal to a process.
+ *
+ * This function is used instead of erts_proc_sig_send_unlink()
+ * when the signal arrives via the distribution and
+ * no link structure is available.
+ *
+ * @param[in]     dep           Distribution entry of channel
+ *                              that the signal arrived on.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ */
+void
+erts_proc_sig_send_dist_unlink(struct dist_entry_ *dep,
+                               Eterm from, Eterm to);
+
+/**
+ *
+ * @brief Send a monitor down signal to a process.
+ *
+ * @param[in]     mon           Pointer to target monitor
+ *                              structure from the sending
+ *                              side. It should contain
+ *                              information about receiver.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ */
+void
+erts_proc_sig_send_monitor_down(ErtsMonitor *mon, Eterm reason);
+
+/**
+ *
+ * @brief Send a demonitor signal to a process.
+ *
+ * @param[in]     mon           Pointer to origin monitor
+ *                              structure from the sending
+ *                              side. It should contain
+ *                              information about receiver.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ */
+void
+erts_proc_sig_send_demonitor(ErtsMonitor *mon);
+
+/**
+ *
+ * @brief Send a monitor signal to a process.
+ *
+ * @param[in]     mon           Pointer to target monitor
+ *                              structure to insert on
+ *                              receiver side.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @return                      A non-zero value if
+ *                              signal was successfully
+ *                              sent. If a zero, value
+ *                              the signal was not sent
+ *                              due to the receiver not
+ *                              existing. The sender
+ *                              needs to deallocate the
+ *                              monitor structure.
+ *
+ */
+int
+erts_proc_sig_send_monitor(ErtsMonitor *mon, Eterm to);
+
+/**
+ *
+ * @brief Send a monitor down signal to a process.
+ *
+ * This function is used instead of erts_proc_sig_send_monitor_down()
+ * when the signal arrives via the distribution and
+ * no link structure is available.
+ *
+ * @param[in]     dep           Pointer to distribution entry
+ *                              of channel that the signal
+ *                              arrived on.
+ *
+ * @param[in]     ref           Reference identifying the monitor.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     reason        Exit reason.
+ *
+ */
+void
+erts_proc_sig_send_dist_monitor_down(DistEntry *dep, Eterm ref,
+                                     Eterm from, Eterm to,
+                                     Eterm reason);
+
+/**
+ *
+ * @brief Send a demonitor signal to a process.
+ *
+ * This function is used instead of erts_proc_sig_send_demonitor()
+ * when the signal arrives via the distribution and
+ * no monitor structure is available.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     ref           Reference identifying the monitor.
+ *
+ */
+void
+erts_proc_sig_send_dist_demonitor(Eterm to, Eterm ref);
+
+/**
+ *
+ * @brief Send a persistent monitor triggered signal to a process.
+ *
+ * Used by monitors that are not auto disabled such as for
+ * example 'time_offset' monitors.
+ *
+ * @param[in]     type          Monitor type.
+ *
+ * @param[in]     key           Monitor key.
+ *
+ * @param[in]     from          Identifier of sender.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     msg           Message template.
+ *
+ * @param[in]     msg_sz        Heap size of message template.
+ *
+ */
+void
+erts_proc_sig_send_persistent_monitor_msg(Uint16 type, Eterm key,
+                                          Eterm from, Eterm to,
+                                          Eterm msg, Uint msg_sz);
+
+/**
+ *
+ * @brief Send a trace change signal to a process.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     on            Trace flags to enable.
+ *
+ * @param[in]     off           Trace flags to disable.
+ *
+ * @param[in]     tracer        Tracer to set. If the non-value,
+ *                              tracer will not be changed.
+ *
+ */
+void
+erts_proc_sig_send_trace_change(Eterm to, Uint on, Uint off,
+                                Eterm tracer);
+
+/**
+ *
+ * @brief Send a group leader signal to a process.
+ *
+ * Set group-leader of receiving process. If sent locally,
+ * a response message '{Ref, Result}' is sent to the original
+ * sender when performed where Ref is the reference passed
+ * as 'ref' argument, and Result is either 'true' or 'badarg'.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *                              NULL if signal arrived via
+ *                              distribution.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     gl            Identifier of new group leader.
+ *
+ * @param[in]     ref           Reference to use in response
+ *                              message to locally sending
+ *                              process (i.e., c_p when c_p
+ *                              is non-null).
+ *
+ */
+void
+erts_proc_sig_send_group_leader(Process *c_p, Eterm to, Eterm gl,
+                                Eterm ref);
+
+/**
+ *
+ * @brief Send an 'is process alive' signal to a process.
+ *
+ * A response message '{Ref, Result}' is sent to the
+ * sender when performed where Ref is the reference passed
+ * as 'ref' argument, and Result is either 'true' or 'false'.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *                              NULL if signal arrived via
+ *                              distribution.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     ref           Reference to use in response
+ *                              message to the sending
+ *                              process (i.e., c_p).
+ *
+ */
+void
+erts_proc_sig_send_is_alive_request(Process *c_p, Eterm to,
+                                    Eterm ref);
+
+/**
+ *
+ * @brief Send a 'process info request' signal to a process.
+ *
+ * A response message '{Ref, Result}' is sent to the
+ * sender when performed where Ref is the reference passed
+ * as 'ref' argument, and Result corresponds to return result
+ * from erlang:process_info/[1,2].
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *                              NULL if signal arrived via
+ *                              distribution.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     item_ix       Info index array to pass to
+ *                              erts_process_info()
+ *
+ * @param[in]     len           Lenght of info index array
+ *
+ * @param[in]     need_msgq_len Non-zero if message queue
+ *                              length is needed; otherwise,
+ *                              zero. If non-zero, sig_qs.len
+ *                              will be set to correspond
+ *                              to the message queue length
+ *                              before call to
+ *                              erts_process_info()
+ *
+ * @param[in]     flags         Flags to pass to
+ *                              erts_process_info()
+ *
+ * @param[in]     reserve_size  Heap size that is known to
+ *                              be needed. May not be correct
+ *                              though.
+ *
+ * @param[in]     ref           Reference to use in response
+ *                              message to the sending
+ *                              process (i.e., c_p).
+ *
+ */
+int
+erts_proc_sig_send_process_info_request(Process *c_p,
+                                        Eterm to,
+                                        int *item_ix,
+                                        int len,
+                                        int need_msgq_len,
+                                        int flags,
+                                        Uint reserve_size,
+                                        Eterm ref);
+
+/**
+ *
+ * @brief Send a 'sync suspend' signal to a process.
+ *
+ * A response message '{Tag, Reply}' is sent to the
+ * sender when performed where Tag is the term passed
+ * as 'tag' argument. Reply is either 'suspended',
+ * 'not_suspended', 'exited' if the operation is
+ * asynchronous; otherwise, the 'reply' argument or
+ * 'badarg' if process terminated.
+ *
+ * This signal does *not* change the suspend state, only
+ * reads and reply the state. This signal is typically
+ * sent after a suspend request (monitor of suspend type)
+ * signal has been sent to the process in order to get a
+ * response when the suspend monitor has been processed.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     to            Identifier of receiver.
+ *
+ * @param[in]     tag           Tag to use in response
+ *                              message to the sending
+ *                              process (i.e., c_p).
+ *
+ * @param[in]     reply         Reply to send if this
+ *                              is a synchronous operation;
+ *                              otherwise, THE_NON_VALUE.
+ */
+void
+erts_proc_sig_send_sync_suspend(Process *c_p, Eterm to,
+                                Eterm tag, Eterm reply);
+
+/**
+ *
+ * @brief Send an 'rpc' signal to a process.
+ *
+ * The function 'func' will be executed in the
+ * context of the receiving process. A response
+ * message '{Ref, Result}' is sent to the sender
+ * when 'func' has been called. 'Ref' is the reference
+ * returned by this function and 'Result' is the
+ * term returned by 'func'. If the return value of
+ * 'func' is not an immediate term, 'func' has to
+ * allocate a heap fragment where the result is stored
+ * and update the the heap fragment pointer pointer
+ * passed as third argument to point to it.
+ *
+ * If this function returns a reference, 'func' will
+ * be called in the context of the receiver. However,
+ * note that this might happen when the receiver is in
+ * an exiting state. The caller of this function
+ * *unconditionally* has to enter a receive that match
+ * on the returned reference in all clauses as next
+ * receive; otherwise, bad things will happen!
+ *
+ * If THE_NON_VALUE is returned, the receiver did not
+ * exist. The signal was not sent, and no specific
+ * receive has to be entered by the caller.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]     to            Identifier of receiver process.
+ *
+ * @param[in]     reply         Non-zero if a reply is wanted.
+ *
+ * @param[in]     func          Function to execute in the
+ *                              context of the receiver.
+ *                              First argument will be a
+ *                              pointer to the process struct
+ *                              of the receiver process.
+ *                              Second argument will be 'arg'
+ *                              (see below). Third argument
+ *                              will be a pointer to a pointer
+ *                              to a heap fragment for storage
+ *                              of result returned from 'func'
+ *                              (i.e. an 'out' parameter).
+ *
+ * @param[in]     arg           Void pointer to argument
+ *                              to pass as second argument
+ *                              in call of 'func'.
+ *
+ * @returns                     If the request was sent,
+ *                              an internal ordinary
+ *                              reference; otherwise,
+ *                              THE_NON_VALUE (non-existing
+ *                              receiver).
+ */
+Eterm
+erts_proc_sig_send_rpc_request(Process *c_p,
+                               Eterm to,
+                               int reply,
+                               Eterm (*func)(Process *, void *, int *, ErlHeapFragment **),
+                               void *arg);
+
+/*
+ * End of send operations of currently supported process signals.
+ */
+
+
+/**
+ *
+ * @brief Handle incoming signals.
+ *
+ * Called by an ordinary scheduler in order to handle incoming
+ * signals for a process. The work is done on the middle part
+ * of the signal queue. The maximum amount of signals handled
+ * is limited by the amount of reductions given when calling.
+ * Note that a reduction does not necessarily map to a signal.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[out]    statep        Pointer to process state after
+ *                              signal handling. May not be NULL.
+ *
+ * @param[in,out] redsp         Pointer to an integer containing
+ *                              reductions. On input, the amount
+ *                              of preferred reductions to be
+ *                              used by the call. On output, the
+ *                              amount of reductions consumed.
+ *
+ * @param[in]     max_reds      Absolute maximum of reductions
+ *                              to use. If the process cannot
+ *                              make progress after the preferred
+ *                              amount of reductions has been
+ *                              consumed, signal handling may
+ *                              proceed up to a maximum of
+ *                              'max_reds' in order to make
+ *                              the process able to proceed
+ *                              with other tasks after handling
+ *                              has finished.
+ *
+ * @param[in]     local_only    If is zero, new signals may be
+ *                              fetched from the outer queue and
+ *                              put in the middle queue before
+ *                              signal handling is performed. If
+ *                              non-zero, no new signals will be
+ *                              fetched before handling begins.
+ *
+ * @return                      Returns a non-zero value, when
+ *                              no more signals to handle in the
+ *                              middle queue remain. A zero
+ *                              return value means that there
+ *                              remains signals in the middle
+ *                              queue.
+ */
+int
+erts_proc_sig_handle_incoming(Process *c_p, erts_aint32_t *statep,
+                              int *redsp, int max_reds,
+                              int local_only);
+
+/**
+ *
+ * @brief Handle remaining signals for an exiting process
+ *
+ * Called as part of termination of a process. It will handle
+ * remaining signals.
+ *
+ * @param[in]     c_p           Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in,out] redsp         Pointer to an integer containing
+ *                              reductions. On input, the amount
+ *                              of maximum reductions to be
+ *                              used by the call. On output, the
+ *                              amount of reductions consumed.
+ *
+ * @return                      Returns a non-zero value, when
+ *                              no more signals to handle in the
+ *                              middle queue remain. A zero
+ *                              return value means that there
+ *                              remains signals in the middle
+ *                              queue.
+ */
+int
+erts_proc_sig_handle_exit(Process *c_p, int *redsp);
+
+/**
+ *
+ * @brief Helper for loop_rec instruction.
+ *
+ * This function should only be called from the loop_rec
+ * instruction (or equivalents). It is called when loop_rec
+ * reach the end of the inner queue (which is the only
+ * part of the signal queue that receive is allowed to
+ * operate on). When called, this function tries to make
+ * more messages available in the inner queue. This by
+ * fetching signals from the outer queue to the middle
+ * queue and/or processing signals in the middle queue.
+ *
+ * @param[in]   c_p             Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]   fcalls          Content of FCALLS in
+ *                              process_main()
+ *
+ * @param[in]   neg_o_reds      Content of neg_o_reds in
+ *                              process_main()
+ *
+ * @param[out]  msgpp           Pointer to pointer to next
+ *                              available message to process.
+ *                              If *msgpp == NULL, no more
+ *                              messages are available.
+ *
+ * @param[out]  get_outp        Pointer to an integer
+ *                              indicating how to respond
+ *                              if no more messages are
+ *                              available (msgpp). If integer
+ *                              is set to zero, loop_rec
+ *                              should jump to an appropriate
+ *                              wait instruction. If zero,
+ *                              the message queue lock remain
+ *                              locked since the test for
+ *                              more messages was done.
+ *                              If the integer is set to a
+ *                              value larger that zero, the
+ *                              process exited. If the integer
+ *                              is set to a value less than
+ *                              zero, the process is required
+ *                              to yield.
+ *
+ *
+ * @return                      The amount of reductions
+ *                              consumed.
+ *
+ */
+int
+erts_proc_sig_receive_helper(Process *c_p, int fcalls,
+                             int neg_o_reds, ErtsMessage **msgpp,
+                             int *get_outp);
+
+/**
+ *
+ * @brief Fetch signals from the outer queue
+ *
+ * Fetches signals from outer queue and places them in the
+ * middle queue ready for signal handling. If the middle
+ * queue is empty, only message signals were present in the
+ * outer queue, and no receive tracing has been enabled on
+ * the process, the middle queue is bypassed and messages
+ * are delivered directly to the inner queue instead.
+ *
+ * @param[in]   c_p             Pointer to process struct of
+ *                              currently executing process.
+ * @returns                     Amount of message signals in
+ *                              inner plus middle signal
+ *                              queues after fetch completed
+ *                              (NOT the message queue
+ *                              length).
+ */
+ERTS_GLB_INLINE Sint erts_proc_sig_fetch(Process *p);
+
+/**
+ *
+ * @brief Get amount of messages in private queues
+ *
+ * @param[in]   c_p             Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @returns                     Amount of message signals in
+ *                              inner plus middle signal
+ *                              queues after fetch completed
+ *                              (NOT the message queue
+ *                              length).
+ */
+Sint
+erts_proc_sig_privqs_len(Process *c_p);
+
+
+/**
+ * @brief Enqueue list of signals on process.
+ *
+ * Message queue must be locked on receiving process.
+ *
+ * @param rp                Receiving process.
+ * @param first             First signal in list.
+ * @param last              Last signal in list.
+ * @param last_next         Pointer to next-pointer to last non-message signal
+ *                          or NULL if no non-message signal after 'first'.
+ * @param msg_cnt           Number of message signals in list.
+ * @param in_state          'state' of rp.
+ *
+ * @return                  'state' of rp.
+ */
+erts_aint32_t
+erts_enqueue_signals(Process *rp, ErtsMessage *first,
+                     ErtsMessage **last, ErtsMessage **last_next,
+                     Uint msg_cnt,
+                     erts_aint32_t in_state);
+
+/**
+ *
+ * @brief Flush pending signal.
+ *
+ */
+void
+erts_proc_sig_send_pending(ErtsSchedulerData* esdp);
+
+/**
+ *
+ * @brief Schedule process to handle enqueued signal(s).
+ *
+ * @param rp                Receiving process.
+ * @param state             'state' of rp.
+ * @param enable_flag       Additional state flags to enable, like
+ *                          ERTS_PSFLG_ACTIVE if message has been enqueued.
+ */
+ERTS_GLB_INLINE void erts_proc_notify_new_sig(Process* rp, erts_aint32_t state,
+                                              erts_aint32_t enable_flag);
+
+void erts_make_dirty_proc_handled(Eterm pid, erts_aint32_t state,
+                                  erts_aint32_t prio);
+
+
+typedef struct {
+    Uint size;
+    ErtsMessage *msgp;
+} ErtsMessageInfo;
+
+/**
+ *
+ * @brief Prepare signal queue for inspection by process_info()
+ *
+ *
+ * @param[in]   c_p             Pointer to process struct of
+ *                              currently executing process.
+ *
+ * @param[in]   rp              Pointer to process struct of
+ *                              process to inspect.
+ *
+ * @param[in]   rp_locks        Process locks held on 'rp'.
+ *
+ * @param[in]   info_on_self    Integer set to non-zero value
+ *                              if caller is inspecting itself;
+ *                              otherwise, zero.
+ *
+ * @param[in]   mip             Pointer to array of
+ *                              ErtsMessageInfo structures.
+ */
+Uint erts_proc_sig_prep_msgq_for_inspection(Process *c_p,
+                                            Process *rp,
+                                            ErtsProcLocks rp_locks,
+                                            int info_on_self,
+                                            ErtsMessageInfo *mip);
+
+/**
+ *
+ * @brief Move message data of messages in private queues to heap
+ *
+ * Move message data of messages in private queues to the heap.
+ * This is part of GC of processes that uses on-heap message
+ * data.
+ *
+ * @param[in]   c_p             Pointer to process struct of
+ *                              currently executing process.
+ *
+ */
+void erts_proc_sig_move_msgs_to_heap(Process *c_p);
+
+/**
+ *
+ * @brief Size of signal in bytes.
+ *
+ * @param[in]   sig             Signal to inspect.
+ *
+ */
+Uint erts_proc_sig_signal_size(ErtsSignal *sig);
+
+
+/**
+ *
+ * @brief Clear seq trace tokens on all signals
+ *
+ * Assumes thread progress has been blocked!
+ *
+ * @param[in]   c_p             Pointer to process
+ *
+ */
+void
+erts_proc_sig_clear_seq_trace_tokens(Process *c_p);
+
+/**
+ *
+ * @brief Handle pending suspend requests
+ *
+ * Should be called by processes when they stop
+ * execution on a dirty scheduler if they have
+ * pending suspend requests (i.e. when
+ * ERTS_PROC_GET_PENDING_SUSPEND(c_p) != NULL).
+ *
+ * @param[in]   c_p             Pointer to executing
+ *                              process
+ */
+void
+erts_proc_sig_handle_pending_suspend(Process *c_p);
+
+/**
+ * @brief Initialize this functionality
+ */
+void erts_proc_sig_queue_init(void);
+
+void
+erts_proc_sig_debug_foreach_sig(Process *c_p,
+                                void (*msg_func)(ErtsMessage *, void *),
+                                void (*oh_func)(ErlOffHeap *, void *),
+                                void (*mon_func)(ErtsMonitor *, void *),
+                                void (*lnk_func)(ErtsLink *, void *),
+                                void *arg);
+
+extern Process *erts_dirty_process_signal_handler;
+extern Process *erts_dirty_process_signal_handler_high;
+extern Process *erts_dirty_process_signal_handler_max;
+
+void erts_proc_sig_fetch__(Process *proc);
+Sint erts_proc_sig_fetch_msgq_len_offs__(Process *proc);
+
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE Sint
+erts_proc_sig_fetch(Process *proc)
+{
+    Sint res = 0;
+    ErtsSignal *sig;
+
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking()
+                   || ERTS_PROC_IS_EXITING(proc)
+                   || ((erts_proc_lc_my_proc_locks(proc)
+                        & (ERTS_PROC_LOCK_MAIN
+                           | ERTS_PROC_LOCK_MSGQ))
+                       == (ERTS_PROC_LOCK_MAIN
+                           | ERTS_PROC_LOCK_MSGQ)));
+
+    ERTS_HDBG_CHECK_SIGNAL_IN_QUEUE(proc);
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(proc, !0);
+
+    sig = (ErtsSignal *) proc->sig_inq.first;
+    if (sig) {
+        if (ERTS_LIKELY(sig->common.tag != ERTS_PROC_SIG_MSGQ_LEN_OFFS_MARK))
+            erts_proc_sig_fetch__(proc);
+        else
+            res = erts_proc_sig_fetch_msgq_len_offs__(proc);
+    }
+
+    res += proc->sig_qs.len;
+
+    ERTS_HDBG_CHECK_SIGNAL_PRIV_QUEUE(proc, !0);
+
+#ifdef ERTS_PROC_SIG_HARD_DEBUG_SIGQ_MSG_LEN
+    {
+        Sint len = 0;
+        ERTS_FOREACH_SIG_PRIVQS(
+            proc, mp,
+            {
+                if (ERTS_SIG_IS_MSG(mp))
+                    len++;
+            });
+        ERTS_ASSERT(res == len);
+    }
+#endif
+
+    return res;
+}
+
+ERTS_GLB_INLINE void
+erts_proc_notify_new_sig(Process* rp, erts_aint32_t state,
+                         erts_aint32_t enable_flag)
+{
+    if (~(state & (ERTS_PSFLG_EXITING
+                   | ERTS_PSFLG_ACTIVE_SYS
+                   | ERTS_PSFLG_SIG_IN_Q))
+        | (~state & enable_flag)) {
+        /* Schedule process... */
+        state = erts_proc_sys_schedule(rp, state, enable_flag);
+    }
+
+    if (state & (ERTS_PSFLG_DIRTY_RUNNING
+                 | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
+        erts_make_dirty_proc_handled(rp->common.id, state, -1);
+    }
+}
+
+#endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
+#endif /* ERTS_PROC_SIG_QUEUE_H__ */
diff --git a/erts/emulator/beam/erl_process.c b/erts/emulator/beam/erl_process.c
index 9b22f024b0..8253ec4f40 100644
--- a/erts/emulator/beam/erl_process.c
+++ b/erts/emulator/beam/erl_process.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -24,6 +24,8 @@
 #  include "config.h"
 #endif
 
+#define ERTS_WANT_BREAK_HANDLING
+
 #include <stddef.h> /* offsetof() */
 #include "sys.h"
 #include "erl_vm.h"
@@ -34,7 +36,6 @@
 #include "erl_db.h"
 #include "dist.h"
 #include "beam_catches.h"
-#include "erl_instrument.h"
 #include "erl_threads.h"
 #include "erl_binary.h"
 #include "beam_bp.h"
@@ -49,6 +50,9 @@
 #define ERTS_WANT_TIMER_WHEEL_API
 #include "erl_time.h"
 #include "erl_nfunc_sched.h"
+#include "erl_check_io.h"
+#include "erl_poll.h"
+#include "erl_proc_sig_queue.h"
 
 #define ERTS_CHECK_TIME_REDS CONTEXT_REDS
 #define ERTS_DELAYED_WAKEUP_INFINITY (~(Uint64) 0)
@@ -127,43 +131,11 @@ runq_got_work_to_execute_flags(Uint32 flags)
     return !ERTS_IS_RUNQ_EMPTY_FLGS(flags);
 }
 
-#ifdef ERTS_SMP
 static ERTS_INLINE int
 runq_got_work_to_execute(ErtsRunQueue *rq)
 {
     return runq_got_work_to_execute_flags(ERTS_RUNQ_FLGS_GET_NOB(rq));
 }
-#endif
-
-#undef RUNQ_READ_RQ
-#undef RUNQ_SET_RQ
-#define RUNQ_READ_RQ(X) ((ErtsRunQueue *) erts_smp_atomic_read_nob((X)))
-#define RUNQ_SET_RQ(X, RQ) erts_smp_atomic_set_nob((X), (erts_aint_t) (RQ))
-
-#ifdef DEBUG
-#  if defined(ARCH_64)
-#    define ERTS_DBG_SET_INVALID_RUNQP(RQP, N) \
-    (RUNQ_SET_RQ((RQP), (0xdeadbeefdead0003LL | ((N) << 4)))
-#  define ERTS_DBG_VERIFY_VALID_RUNQP(RQP) \
-do { \
-    ASSERT((RQP) != NULL); \
-    ASSERT(((((Uint) (RQP)) & ((Uint) 0x3))) == ((Uint) 0)); \
-    ASSERT((((Uint) (RQP)) & ~((Uint) 0xffff)) != ((Uint) 0xdeadbeefdead0000LL));\
-} while (0)
-#  else
-#    define ERTS_DBG_SET_INVALID_RUNQP(RQP, N) \
-    (RUNQ_SET_RQ((RQP), (0xdead0003 | ((N) << 4))))
-#  define ERTS_DBG_VERIFY_VALID_RUNQP(RQP) \
-do { \
-    ASSERT((RQP) != NULL); \
-    ASSERT(((((UWord) (RQP)) & ((UWord) 1))) == ((UWord) 0)); \
-    ASSERT((((UWord) (RQP)) & ~((UWord) 0xffff)) != ((UWord) 0xdead0000)); \
-} while (0)
-#  endif
-#else
-#  define ERTS_DBG_SET_INVALID_RUNQP(RQP, N)
-#  define ERTS_DBG_VERIFY_VALID_RUNQP(RQP)
-#endif
 
 const Process erts_invalid_process = {{ERTS_INVALID_PID}};
 
@@ -172,7 +144,6 @@ extern BeamInstr beam_exit[];
 extern BeamInstr beam_continue_exit[];
 
 int ERTS_WRITE_UNLIKELY(erts_default_spo_flags) = SPO_ON_HEAP_MSGQ;
-int ERTS_WRITE_UNLIKELY(erts_eager_check_io) = 1;
 int ERTS_WRITE_UNLIKELY(erts_sched_compact_load);
 int ERTS_WRITE_UNLIKELY(erts_sched_balance_util) = 0;
 Uint ERTS_WRITE_UNLIKELY(erts_no_schedulers);
@@ -194,57 +165,57 @@ static UWord thr_prgr_later_cleanup_op_threshold = ERTS_THR_PRGR_LATER_CLEANUP_O
 ErtsPTab erts_proc erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
 int erts_sched_thread_suggested_stack_size = -1;
-#ifdef ERTS_DIRTY_SCHEDULERS
 int erts_dcpu_sched_thread_suggested_stack_size = -1;
 int erts_dio_sched_thread_suggested_stack_size = -1;
-#endif
 #ifdef ERTS_ENABLE_LOCK_CHECK
 ErtsLcPSDLocks erts_psd_required_locks[ERTS_PSD_SIZE];
 #endif
 
-static struct {
+typedef struct {
     int aux_work;
     int tse;
     int sys_schedule;
-} sched_busy_wait;
+} ErtsBusyWaitParams;
 
-#ifdef ERTS_SMP
-int erts_disable_proc_not_running_opt;
+static ErtsBusyWaitParams sched_busy_wait_params[ERTS_SCHED_TYPE_LAST + 1];
+
+static ERTS_INLINE ErtsBusyWaitParams *
+sched_get_busy_wait_params(ErtsSchedulerData *esdp)
+{
+    return &sched_busy_wait_params[esdp->type];
+}
 
 static ErtsAuxWorkData *aux_thread_aux_work_data;
+static ErtsAuxWorkData *poll_thread_aux_work_data;
 
 #define ERTS_SCHDLR_SSPND_CHNG_NMSB		(((erts_aint32_t) 1) << 0)
 #define ERTS_SCHDLR_SSPND_CHNG_MSB		(((erts_aint32_t) 1) << 1)
 #define ERTS_SCHDLR_SSPND_CHNG_ONLN		(((erts_aint32_t) 1) << 2)
 #define ERTS_SCHDLR_SSPND_CHNG_DCPU_ONLN	(((erts_aint32_t) 1) << 3)
 
-typedef struct {
+typedef struct ErtsMultiSchedulingBlock_ {
     int ongoing;
     ErtsProcList *blckrs;
     ErtsProcList *chngq;
 } ErtsMultiSchedulingBlock;
 
-typedef struct {
+typedef struct ErtsSchedTypeCounters_ {
     Uint32 normal;
-#ifdef ERTS_DIRTY_SCHEDULERS
     Uint32 dirty_cpu;
     Uint32 dirty_io;
-#endif
 } ErtsSchedTypeCounters;
 
-static struct {
-    erts_smp_mtx_t mtx;
+static struct ErtsSchedSuspend_ {
+    erts_mtx_t mtx;
     ErtsSchedTypeCounters online;
     ErtsSchedTypeCounters curr_online;
     ErtsSchedTypeCounters active;
-    erts_smp_atomic32_t changing;
+    erts_atomic32_t changing;
     ErtsProcList *chngq;
     Eterm changer;
     ErtsMultiSchedulingBlock nmsb; /* Normal multi Scheduling Block */
     ErtsMultiSchedulingBlock msb; /* Multi Scheduling Block */
-#ifdef ERTS_DIRTY_SCHEDULERS
     ErtsSchedType last_msb_dirty_type;
-#endif
 } schdlr_sspnd;
 
 static void init_scheduler_suspend(void);
@@ -253,10 +224,8 @@ static ERTS_INLINE Uint32
 schdlr_sspnd_eq_nscheds(ErtsSchedTypeCounters *val1p, ErtsSchedTypeCounters *val2p)
 {
     int res = val1p->normal == val2p->normal;
-#ifdef ERTS_DIRTY_SCHEDULERS
     res &= val1p->dirty_cpu == val2p->dirty_cpu;
     res &= val1p->dirty_io == val2p->dirty_io;
-#endif
     return res;
 }
 
@@ -267,16 +236,10 @@ schdlr_sspnd_get_nscheds(ErtsSchedTypeCounters *valp,
     switch (type) {
     case ERTS_SCHED_NORMAL:
         return valp->normal;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
         return valp->dirty_cpu;
     case ERTS_SCHED_DIRTY_IO:
         return valp->dirty_io;
-#else
-    case ERTS_SCHED_DIRTY_CPU:
-    case ERTS_SCHED_DIRTY_IO:
-        return 0;
-#endif
     default:
 	ERTS_INTERNAL_ERROR("Invalid scheduler type");
 	return 0;
@@ -288,10 +251,8 @@ static ERTS_INLINE Uint32
 schdlr_sspnd_get_nscheds_tot(ErtsSchedTypeCounters *valp)
 {
     Uint32 res = valp->normal;
-#ifdef ERTS_DIRTY_SCHEDULERS
     res += valp->dirty_cpu;
     res += valp->dirty_io;
-#endif
     return res;
 }
 #endif
@@ -306,14 +267,12 @@ schdlr_sspnd_dec_nscheds(ErtsSchedTypeCounters *valp,
     case ERTS_SCHED_NORMAL:
         valp->normal--;
 	break;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
         valp->dirty_cpu--;
 	break;
     case ERTS_SCHED_DIRTY_IO:
         valp->dirty_io--;
 	break;
-#endif
     default:
 	ERTS_INTERNAL_ERROR("Invalid scheduler type");
     }
@@ -327,14 +286,12 @@ schdlr_sspnd_inc_nscheds(ErtsSchedTypeCounters *valp,
     case ERTS_SCHED_NORMAL:
         valp->normal++;
 	break;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
         valp->dirty_cpu++;
 	break;
     case ERTS_SCHED_DIRTY_IO:
         valp->dirty_io++;
 	break;
-#endif
     default:
 	ERTS_INTERNAL_ERROR("Invalid scheduler type");
     }
@@ -348,25 +305,23 @@ schdlr_sspnd_set_nscheds(ErtsSchedTypeCounters *valp,
     case ERTS_SCHED_NORMAL:
         valp->normal = no;
 	break;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
         valp->dirty_cpu = no;
 	break;
     case ERTS_SCHED_DIRTY_IO:
         valp->dirty_io = no;
 	break;
-#endif
     default:
 	ERTS_INTERNAL_ERROR("Invalid scheduler type");
     }
 }
 
 static struct {
-    erts_smp_mtx_t update_mtx;
-    erts_smp_atomic32_t no_runqs;
+    erts_mtx_t update_mtx;
+    erts_atomic32_t no_runqs;
     int last_active_runqs;
     int forced_check_balance;
-    erts_smp_atomic32_t checking_balance;
+    erts_atomic32_t checking_balance;
     int halftime;
     int full_reds_history_index;
     struct {
@@ -384,51 +339,38 @@ do {							\
     balance_info.prev_rise.reds = (REDS);		\
 } while (0)
 
-#endif
 
 erts_sched_stat_t erts_sched_stat;
 
-#ifdef USE_THREADS
 static erts_tsd_key_t ERTS_WRITE_UNLIKELY(sched_data_key);
-#endif
-
-static erts_smp_atomic32_t function_calls;
 
-#ifdef ERTS_SMP
-static erts_smp_atomic32_t doing_sys_schedule;
-static erts_smp_atomic32_t no_empty_run_queues;
+static erts_atomic32_t no_empty_run_queues;
 long erts_runq_supervision_interval = 0;
 static ethr_event runq_supervision_event;
 static erts_tid_t runq_supervisor_tid;
 static erts_atomic_t runq_supervisor_sleeping;
-#else /* !ERTS_SMP */
-ErtsSchedulerData *erts_scheduler_data;
-#endif
 
 ErtsAlignedRunQueue * ERTS_WRITE_UNLIKELY(erts_aligned_run_queues);
 Uint ERTS_WRITE_UNLIKELY(erts_no_run_queues);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 struct {
     union {
-        erts_smp_atomic32_t active;
+        erts_atomic32_t active;
         char align__[ERTS_CACHE_LINE_SIZE];
     } cpu;
     union {
-        erts_smp_atomic32_t active;
+        erts_atomic32_t active;
         char align__[ERTS_CACHE_LINE_SIZE];
     } io;
 } dirty_count erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
-#endif
 
 static ERTS_INLINE void
 dirty_active(ErtsSchedulerData *esdp, erts_aint32_t add)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_aint32_t val;
-    erts_smp_atomic32_t *ap;
+    erts_atomic32_t *ap;
     switch (esdp->type) {
     case ERTS_SCHED_DIRTY_CPU:
         ap = &dirty_count.cpu.active;
@@ -446,23 +388,20 @@ dirty_active(ErtsSchedulerData *esdp, erts_aint32_t add)
      * All updates done under run-queue lock, so
      * no inc or dec needed...
      */
-    ERTS_SMP_ASSERT(erts_smp_lc_runq_is_locked(esdp->run_queue));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(esdp->run_queue));
 
-    val = erts_smp_atomic32_read_nob(ap);
+    val = erts_atomic32_read_nob(ap);
     val += add;
-    erts_smp_atomic32_set_nob(ap, val);
-#endif
+    erts_atomic32_set_nob(ap, val);
 }
 
 ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_scheduler_data);
-#ifdef ERTS_DIRTY_SCHEDULERS
 ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_dirty_cpu_scheduler_data);
 ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_dirty_io_scheduler_data);
 typedef union {
     Process dsp;
     char align[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(Process))];
 } ErtsAlignedDirtyShadowProcess;
-#endif
 
 typedef union {
     ErtsSchedulerSleepInfo ssi;
@@ -470,12 +409,9 @@ typedef union {
 } ErtsAlignedSchedulerSleepInfo;
 
 static ErtsAlignedSchedulerSleepInfo *aligned_sched_sleep_info;
-#ifdef ERTS_DIRTY_SCHEDULERS
-#ifdef ERTS_SMP
 static ErtsAlignedSchedulerSleepInfo *aligned_dirty_cpu_sched_sleep_info;
 static ErtsAlignedSchedulerSleepInfo *aligned_dirty_io_sched_sleep_info;
-#endif
-#endif
+static ErtsAlignedSchedulerSleepInfo *aligned_poll_thread_sleep_info;
 
 static Uint last_reductions;
 static Uint last_exact_reductions;
@@ -501,7 +437,8 @@ typedef enum {
     ERTS_PSTT_CLA,	/* Copy Literal Area */
     ERTS_PSTT_COHMQ,    /* Change off heap message queue */
     ERTS_PSTT_FTMQ,     /* Flush trace msg queue */
-    ERTS_PSTT_ETS_FREE_FIXATION
+    ERTS_PSTT_ETS_FREE_FIXATION,
+    ERTS_PSTT_PRIO_SIG  /* Elevate prio on signal management */
 } ErtsProcSysTaskType;
 
 #define ERTS_MAX_PROC_SYS_TASK_ARGS 2
@@ -543,11 +480,14 @@ ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(proclist,
 				 200,
 				 ERTS_ALC_T_PROC_LIST)
 
+#define ERTS_POLL_THREAD_SLEEP_INFO_IX(IX)                              \
+        (ASSERT(0 <= ((int) (IX))                                       \
+                && ((int) (IX)) < ((int) erts_no_poll_threads)),        \
+         &aligned_poll_thread_sleep_info[(IX)].ssi)
 #define ERTS_SCHED_SLEEP_INFO_IX(IX)					\
-    (ASSERT(-1 <= ((int) (IX))					        \
-		 && ((int) (IX)) < ((int) erts_no_schedulers)),		\
+    (ASSERT(((int)-1) <= ((int) (IX))                 \
+            && ((int) (IX)) < ((int) erts_no_schedulers)),		\
      &aligned_sched_sleep_info[(IX)].ssi)
-#ifdef ERTS_DIRTY_SCHEDULERS
 #define ERTS_DIRTY_CPU_SCHED_SLEEP_INFO_IX(IX)				\
     (ASSERT(0 <= ((int) (IX))					        \
 	    && ((int) (IX)) < ((int) erts_no_dirty_cpu_schedulers)),	\
@@ -556,7 +496,6 @@ ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(proclist,
     (ASSERT(0 <= ((int) (IX))					        \
 	    && ((int) (IX)) < ((int) erts_no_dirty_io_schedulers)),	\
      &aligned_dirty_io_sched_sleep_info[(IX)].ssi)
-#endif
 
 #define ERTS_FOREACH_RUNQ(RQVAR, DO)					\
 do {									\
@@ -564,9 +503,9 @@ do {									\
     int ix__;								\
     for (ix__ = 0; ix__ < erts_no_run_queues; ix__++) {			\
 	RQVAR = ERTS_RUNQ_IX(ix__);					\
-	erts_smp_runq_lock(RQVAR);					\
+	erts_runq_lock(RQVAR);					\
 	{ DO; }								\
-	erts_smp_runq_unlock(RQVAR);					\
+	erts_runq_unlock(RQVAR);					\
     }									\
 } while (0)
 
@@ -576,12 +515,12 @@ do {									\
     int ix__;								\
     int online__ = (int) schdlr_sspnd_get_nscheds(&schdlr_sspnd.online,	\
 						  ERTS_SCHED_NORMAL);	\
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&schdlr_sspnd.mtx));	\
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&schdlr_sspnd.mtx));	\
     for (ix__ = 0; ix__ < online__; ix__++) {				\
 	RQVAR = ERTS_RUNQ_IX(ix__);					\
-	erts_smp_runq_lock(RQVAR);					\
+	erts_runq_lock(RQVAR);					\
 	{ DO; }								\
-	erts_smp_runq_unlock(RQVAR);					\
+	erts_runq_unlock(RQVAR);					\
     }									\
 } while (0)
 
@@ -592,12 +531,12 @@ do {									\
     int ix__;								\
     for (ix__ = 0; ix__ < nrqs; ix__++) {                               \
 	RQVAR = ERTS_RUNQ_IX(ix__);					\
-	erts_smp_runq_lock(RQVAR);					\
+	erts_runq_lock(RQVAR);					\
 	{ DO; }								\
     }									\
     { DOX; }								\
     for (ix__ = 0; ix__ < nrqs; ix__++)                                 \
-	erts_smp_runq_unlock(ERTS_RUNQ_IX(ix__));			\
+	erts_runq_unlock(ERTS_RUNQ_IX(ix__));			\
 } while (0)
 
 #define ERTS_ATOMIC_FOREACH_RUNQ(RQVAR, DO)                             \
@@ -638,11 +577,8 @@ dbg_chk_aux_work_val(erts_aint32_t value)
     valid |= ERTS_SSI_AUX_WORK_MISC;
     valid |= ERTS_SSI_AUX_WORK_FIX_ALLOC_LOWER_LIM;
     valid |= ERTS_SSI_AUX_WORK_FIX_ALLOC_DEALLOC;
-#if ERTS_USE_ASYNC_READY_Q
     valid |= ERTS_SSI_AUX_WORK_ASYNC_READY;
     valid |= ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN;
-#endif
-#ifdef ERTS_SMP
     valid |= ERTS_SSI_AUX_WORK_DELAYED_AW_WAKEUP;
     valid |= ERTS_SSI_AUX_WORK_MISC_THR_PRGR;
     valid |= ERTS_SSI_AUX_WORK_DD;
@@ -650,8 +586,6 @@ dbg_chk_aux_work_val(erts_aint32_t value)
     valid |= ERTS_SSI_AUX_WORK_CNCLD_TMRS;
     valid |= ERTS_SSI_AUX_WORK_CNCLD_TMRS_THR_PRGR;
     valid |= ERTS_SSI_AUX_WORK_THR_PRGR_LATER_OP;
-    valid |= ERTS_SSI_AUX_WORK_PENDING_EXITERS;
-#endif
 #if HAVE_ERTS_MSEG
     valid |= ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK;
 #endif
@@ -673,16 +607,13 @@ dbg_chk_aux_work_val(erts_aint32_t value)
 #define ERTS_DBG_CHK_SSI_AUX_WORK(SSI)
 #endif
 
-#ifdef ERTS_SMP
-static void do_handle_pending_exiters(ErtsProcList *);
 static void wake_scheduler(ErtsRunQueue *rq);
-#endif
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 int
-erts_smp_lc_runq_is_locked(ErtsRunQueue *runq)
+erts_lc_runq_is_locked(ErtsRunQueue *runq)
 {
-    return erts_smp_lc_mtx_is_locked(&runq->mtx);
+    return erts_lc_mtx_is_locked(&runq->mtx);
 }
 #endif
 
@@ -690,13 +621,13 @@ erts_smp_lc_runq_is_locked(ErtsRunQueue *runq)
 static ERTS_INLINE Uint64
 ensure_later_proc_interval(Uint64 interval)
 {
-    return erts_smp_ensure_later_interval_nob(erts_ptab_interval(&erts_proc), interval);
+    return erts_ensure_later_interval_nob(erts_ptab_interval(&erts_proc), interval);
 }
 
 Uint64
 erts_get_proc_interval(void)
 {
-    return erts_smp_current_interval_nob(erts_ptab_interval(&erts_proc));
+    return erts_current_interval_nob(erts_ptab_interval(&erts_proc));
 }
 
 Uint64
@@ -708,15 +639,13 @@ erts_ensure_later_proc_interval(Uint64 interval)
 Uint64
 erts_step_proc_interval(void)
 {
-    return erts_smp_step_interval_nob(erts_ptab_interval(&erts_proc));
+    return erts_step_interval_nob(erts_ptab_interval(&erts_proc));
 }
 
 void
 erts_pre_init_process(void)
 {
-#ifdef USE_THREADS
     erts_tsd_key_create(&sched_data_key, "erts_sched_data_key");
-#endif
 
     erts_aux_work_flag_descr[ERTS_SSI_AUX_WORK_DELAYED_AW_WAKEUP_IX]
 	= "DELAYED_AW_WAKEUP";
@@ -742,8 +671,6 @@ erts_pre_init_process(void)
 	= "MISC_THR_PRGR";
     erts_aux_work_flag_descr[ERTS_SSI_AUX_WORK_MISC_IX]
 	= "MISC";
-    erts_aux_work_flag_descr[ERTS_SSI_AUX_WORK_PENDING_EXITERS_IX]
-	= "PENDING_EXITERS";
     erts_aux_work_flag_descr[ERTS_SSI_AUX_WORK_SET_TMO_IX]
 	= "SET_TMO";
     erts_aux_work_flag_descr[ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK_IX]
@@ -796,6 +723,16 @@ erts_pre_init_process(void)
         = ERTS_PSD_ETS_FIXED_TABLES_GET_LOCKS;
     erts_psd_required_locks[ERTS_PSD_ETS_FIXED_TABLES].set_locks
         = ERTS_PSD_ETS_FIXED_TABLES_SET_LOCKS;
+
+    erts_psd_required_locks[ERTS_PSD_DIST_ENTRY].get_locks
+        = ERTS_PSD_DIST_ENTRY_GET_LOCKS;
+    erts_psd_required_locks[ERTS_PSD_DIST_ENTRY].set_locks
+        = ERTS_PSD_DIST_ENTRY_SET_LOCKS;
+
+    erts_psd_required_locks[ERTS_PSD_PENDING_SUSPEND].get_locks
+        = ERTS_PSD_PENDING_SUSPEND_GET_LOCKS;
+    erts_psd_required_locks[ERTS_PSD_PENDING_SUSPEND].set_locks
+        = ERTS_PSD_PENDING_SUSPEND_SET_LOCKS;
 #endif
 }
 
@@ -810,10 +747,7 @@ void
 erts_init_process(int ncpu, int proc_tab_size, int legacy_proc_tab)
 {
 
-#ifdef ERTS_SMP
-    erts_disable_proc_not_running_opt = 0;
     erts_init_proc_lock(ncpu);
-#endif
 
     init_proclist_alloc();
 
@@ -825,11 +759,7 @@ erts_init_process(int ncpu, int proc_tab_size, int legacy_proc_tab)
 			 sizeof(Process),
 			 "process_table",
 			 legacy_proc_tab,
-#ifdef ERTS_SMP
 			 1
-#else
-			 0
-#endif
 	);
 
     last_reductions = 0;
@@ -841,7 +771,7 @@ erts_late_init_process(void)
 {
     int ix;
 
-    erts_smp_spinlock_init(&erts_sched_stat.lock, "sched_stat", NIL,
+    erts_spinlock_init(&erts_sched_stat.lock, "sched_stat", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_SCHEDULER);
 
     for (ix = 0; ix < ERTS_NO_PRIO_LEVELS; ix++) {
@@ -883,7 +813,6 @@ erts_late_init_process(void)
 static void
 init_sched_wall_time(ErtsSchedulerData *esdp, Uint64 time_stamp)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (esdp->type != ERTS_SCHED_NORMAL) {
         erts_atomic32_init_nob(&esdp->sched_wall_time.u.mod, 0);
         esdp->sched_wall_time.enabled = 1;
@@ -892,7 +821,6 @@ init_sched_wall_time(ErtsSchedulerData *esdp, Uint64 time_stamp)
         esdp->sched_wall_time.working.start = ERTS_SCHED_WTIME_IDLE;
     }
     else
-#endif
     {
         esdp->sched_wall_time.u.need = erts_sched_balance_util;
         esdp->sched_wall_time.enabled = 0;
@@ -1041,14 +969,14 @@ erts_get_sched_util(ErtsRunQueue *rq, int initially_locked, int short_interval)
 	if (!locked) {
 	    if (++try >= ERTS_GET_AVG_MAX_UNLOCKED_TRY) {
 		/* Writer will eventually block on runq-lock */
-		erts_smp_runq_lock(rq);
+		erts_runq_lock(rq);
 		locked = 1;
 	    }
 	}
     }
 
     if (!initially_locked && locked)
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
 
     now = sched_wall_time_ts();
     worktime = calc_sched_worktime(is_working, now, last, interval, old_worktime);
@@ -1090,7 +1018,6 @@ init_runq_sched_util(ErtsRunQueueSchedUtil *rqsu, int enabled)
 
 #endif /* ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT */
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 typedef struct {
     Uint64 working;
@@ -1143,9 +1070,7 @@ read_dirty_sched_wall_time(ErtsSchedulerData *esdp, ErtsDirtySchedWallTime *info
         info->working = info->total;
 }
 
-#endif
 
-#ifdef ERTS_SMP
 
 static void
 dirty_sched_wall_time_change(ErtsSchedulerData *esdp, int working)
@@ -1193,16 +1118,13 @@ dirty_sched_wall_time_change(ErtsSchedulerData *esdp, int working)
     mod++;
     erts_atomic32_set_nob(&esdp->sched_wall_time.u.mod, mod);
 
-#if 0
     if (!working) {
-        ERTS_MSACC_SET_STATE_M_X(ERTS_MSACC_STATE_BUSY_WAIT);
+        ERTS_MSACC_SET_STATE_X(ERTS_MSACC_STATE_BUSY_WAIT);
     } else {
-        ERTS_MSACC_SET_STATE_M_X(ERTS_MSACC_STATE_OTHER);
+        ERTS_MSACC_SET_STATE_X(ERTS_MSACC_STATE_OTHER);
     }
-#endif
 }
 
-#endif /* ERTS_SMP */
 
 static void
 sched_wall_time_change(ErtsSchedulerData *esdp, int working)
@@ -1247,11 +1169,9 @@ typedef struct {
     Eterm ref;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Uint req_sched;
-    erts_smp_atomic32_t refc;
-#ifdef ERTS_DIRTY_SCHEDULERS
+    erts_atomic32_t refc;
     int want_dirty_cpu;
     int want_dirty_io;
-#endif
 } ErtsSchedWallTimeReq;
 
 typedef struct {
@@ -1259,7 +1179,7 @@ typedef struct {
     Eterm ref;
     Eterm ref_heap[ERTS_REF_THING_SIZE];
     Uint req_sched;
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
 } ErtsSystemCheckReq;
 
 
@@ -1291,10 +1211,8 @@ reply_sched_wall_time(void *vswtrp)
     ErlOffHeap *ohp = NULL;
     ErtsMessage *mp = NULL;
 
-    ASSERT(esdp);
-#ifdef ERTS_DIRTY_SCHEDULERS
-    ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
-#endif
+    ASSERT(esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp));
+
     if (swtrp->set) {
 	if (!swtrp->enable && esdp->sched_wall_time.enabled) {
 	    esdp->sched_wall_time.u.need = erts_sched_balance_util;
@@ -1324,7 +1242,6 @@ reply_sched_wall_time(void *vswtrp)
     hpp = NULL;
     szp = &sz;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (esdp->sched_wall_time.enabled
         && swtrp->req_sched == esdp->no
         && (swtrp->want_dirty_cpu || swtrp->want_dirty_io)) {
@@ -1406,7 +1323,6 @@ reply_sched_wall_time(void *vswtrp)
         erts_free(ERTS_ALC_T_TMP, dswt);
     }
     else
-#endif
     {
         /* Reply with info about this scheduler only... */
 
@@ -1443,11 +1359,11 @@ reply_sched_wall_time(void *vswtrp)
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
  
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
     erts_proc_dec_refc(rp);
 
-    if (erts_smp_atomic32_dec_read_nob(&swtrp->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&swtrp->refc) == 0)
 	swtreq_free(vswtrp);
 }
 
@@ -1460,11 +1376,10 @@ erts_sched_wall_time_request(Process *c_p, int set, int enable,
     ErtsSchedWallTimeReq *swtrp;
     Eterm *hp;
 
+    ASSERT(esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp));
+
     if (!set && !esdp->sched_wall_time.enabled)
 	return THE_NON_VALUE;
-#ifdef ERTS_DIRTY_SCHEDULERS
-    ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
-#endif
 
     swtrp = swtreq_alloc();
     ref = erts_make_ref(c_p);
@@ -1475,22 +1390,18 @@ erts_sched_wall_time_request(Process *c_p, int set, int enable,
     swtrp->proc = c_p;
     swtrp->ref = STORE_NC(&hp, NULL, ref);
     swtrp->req_sched = esdp->no;
-#ifdef ERTS_DIRTY_SCHEDULERS
     swtrp->want_dirty_cpu = want_dirty_cpu;
     swtrp->want_dirty_io = want_dirty_io;
-#endif
-    erts_smp_atomic32_init_nob(&swtrp->refc,
+    erts_atomic32_init_nob(&swtrp->refc,
 			       (erts_aint32_t) erts_no_schedulers);
 
     erts_proc_add_refc(c_p, (Sint32) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
 					  reply_sched_wall_time,
 					  (void *) swtrp);
-#endif
 
     reply_sched_wall_time((void *) swtrp);
 
@@ -1511,10 +1422,7 @@ reply_system_check(void *vscrp)
     ErlOffHeap *ohp = NULL;
     ErtsMessage *mp = NULL;
 
-    ASSERT(esdp);
-#ifdef ERTS_DIRTY_SCHEDULERS
-    ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
-#endif
+    ASSERT(esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp));
 
     sz = ERTS_REF_THING_SIZE;
     mp = erts_alloc_message_heap(rp, &rp_locks, sz, &hp, &ohp);
@@ -1527,11 +1435,11 @@ reply_system_check(void *vscrp)
 	rp_locks &= ~ERTS_PROC_LOCK_MAIN;
 
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
     erts_proc_dec_refc(rp);
 
-    if (erts_smp_atomic32_dec_read_nob(&scrp->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&scrp->refc) == 0)
 	screq_free(vscrp);
 }
 
@@ -1549,17 +1457,15 @@ Eterm erts_system_check_request(Process *c_p) {
     scrp->proc = c_p;
     scrp->ref = STORE_NC(&hp, NULL, ref);
     scrp->req_sched = esdp->no;
-    erts_smp_atomic32_init_nob(&scrp->refc, (erts_aint32_t) erts_no_schedulers);
+    erts_atomic32_init_nob(&scrp->refc, (erts_aint32_t) erts_no_schedulers);
 
     erts_proc_add_refc(c_p, (Sint) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
 					  reply_system_check,
 					  (void *) scrp);
-#endif
 
     reply_system_check((void *) scrp);
 
@@ -1634,7 +1540,7 @@ erts_psd_set_init(Process *p, int ix, void *data)
     for (i = 0; i < ERTS_PSD_SIZE; i++)
 	new_psd->data[i] = NULL;
 
-    psd = (ErtsPSD *) erts_smp_atomic_cmpxchg_mb(&p->psd,
+    psd = (ErtsPSD *) erts_atomic_cmpxchg_mb(&p->psd,
 						 (erts_aint_t) new_psd,
 						 (erts_aint_t) NULL);
     if (psd)
@@ -1646,21 +1552,21 @@ erts_psd_set_init(Process *p, int ix, void *data)
     return old;
 }
 
-#ifdef ERTS_SMP
 
 void
-erts_sched_finish_poke(ErtsSchedulerSleepInfo *ssi, erts_aint32_t flags)
+erts_sched_finish_poke(ErtsSchedulerSleepInfo *ssi,
+                       erts_aint32_t flags)
 {
     switch (flags & ERTS_SSI_FLGS_SLEEP_TYPE) {
     case ERTS_SSI_FLG_POLL_SLEEPING:
-	erts_sys_schedule_interrupt(1);
+	erts_check_io_interrupt(ssi->psi, 1);
 	break;
     case ERTS_SSI_FLG_POLL_SLEEPING|ERTS_SSI_FLG_TSE_SLEEPING:
 	/*
 	 * Thread progress blocking while poll sleeping; need
 	 * to signal on both...
 	 */
-	erts_sys_schedule_interrupt(1);
+	erts_check_io_interrupt(ssi->psi, 1);
 	/* fall through */
     case ERTS_SSI_FLG_TSE_SLEEPING:
 	erts_tse_set(ssi->event);
@@ -1674,7 +1580,6 @@ erts_sched_finish_poke(ErtsSchedulerSleepInfo *ssi, erts_aint32_t flags)
     }
 }
 
-#endif
 
 static ERTS_INLINE void
 set_aux_work_flags_wakeup_nob(ErtsSchedulerSleepInfo *ssi,
@@ -1690,11 +1595,7 @@ set_aux_work_flags_wakeup_nob(ErtsSchedulerSleepInfo *ssi,
 	old_flgs = erts_atomic32_read_bor_nob(&ssi->aux_work, flgs);
 
 	if ((old_flgs & flgs) != flgs) {
-#ifdef ERTS_SMP
 	    erts_sched_poke(ssi);
-#else
-	    erts_sys_schedule_interrupt(1);
-#endif
 	}
     }
 }
@@ -1710,11 +1611,7 @@ set_aux_work_flags_wakeup_relb(ErtsSchedulerSleepInfo *ssi,
     old_flgs = erts_atomic32_read_bor_relb(&ssi->aux_work, flgs);
 
     if ((old_flgs & flgs) != flgs) {
-#ifdef ERTS_SMP
 	erts_sched_poke(ssi);
-#else
-	erts_sys_schedule_interrupt(1);
-#endif
     }
 }
 
@@ -1730,7 +1627,6 @@ unset_aux_work_flags(ErtsSchedulerSleepInfo *ssi, erts_aint32_t flgs)
     return erts_atomic32_read_band_nob(&ssi->aux_work, ~flgs);
 }
 
-#ifdef ERTS_SMP
 
 static ERTS_INLINE void
 haw_chk_later_cleanup_op_wakeup(ErtsAuxWorkData *awdp, ErtsThrPrgrVal val)
@@ -1800,9 +1696,9 @@ static ERTS_INLINE void
 haw_thr_prgr_current_check_progress(ErtsAuxWorkData *awdp)
 {
     ErtsThrPrgrVal current = awdp->current_thr_prgr;
-#ifdef ERTS_DIRTY_SCHEDULERS
+
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     if (current != ERTS_THR_PRGR_INVALID
 	&& !erts_thr_progress_equal(current, erts_thr_progress_current())) {
 	/*
@@ -1819,9 +1715,7 @@ handle_delayed_aux_work_wakeup(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, in
 {
     int jix, max_jix;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
 
     ASSERT(awdp->delayed_wakeup.next != ERTS_DELAYED_WAKEUP_INFINITY);
 
@@ -1879,7 +1773,6 @@ schedule_aux_work_wakeup(ErtsAuxWorkData *awdp,
     }
 }
 
-#endif
 
 typedef struct erts_misc_aux_work_t_ erts_misc_aux_work_t;
 struct erts_misc_aux_work_t_ {
@@ -1920,11 +1813,7 @@ init_misc_aux_work(void)
 					   sizeof(erts_algnd_misc_aux_work_q_t)
 					   * (erts_no_schedulers+1));
 
-#ifdef ERTS_SMP
     ix = 0; /* aux_thread + schedulers */
-#else
-    ix = 1; /* scheduler only */
-#endif
 
     for (; ix <= erts_no_schedulers; ix++) {
 	qinit.arg = (void *) ERTS_SCHED_SLEEP_INFO_IX(ix-1);
@@ -1942,10 +1831,8 @@ misc_aux_work_clean(ErtsThrQ_t *q,
 	set_aux_work_flags(awdp->ssi, ERTS_SSI_AUX_WORK_MISC);
 	return aux_work | ERTS_SSI_AUX_WORK_MISC;
     case ERTS_THR_Q_NEED_THR_PRGR:
-#ifdef ERTS_SMP
 	set_aux_work_flags(awdp->ssi, ERTS_SSI_AUX_WORK_MISC_THR_PRGR);
 	haw_thr_prgr_soft_wakeup(awdp, erts_thr_q_need_thr_progress(q));
-#endif
     case ERTS_THR_Q_CLEAN:
 	break;
     }
@@ -1971,16 +1858,14 @@ handle_misc_aux_work(ErtsAuxWorkData *awdp,
     return misc_aux_work_clean(q, awdp, aux_work & ~ERTS_SSI_AUX_WORK_MISC);
 }
 
-#ifdef ERTS_SMP
 
 static ERTS_INLINE erts_aint32_t
 handle_misc_aux_work_thr_prgr(ErtsAuxWorkData *awdp,
 			      erts_aint32_t aux_work,
 			      int waiting)
 {
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     if (!erts_thr_progress_has_reached_this(haw_thr_prgr_current(awdp),
 					    awdp->misc.thr_prgr))
 	return aux_work & ~ERTS_SSI_AUX_WORK_MISC_THR_PRGR;
@@ -1992,7 +1877,6 @@ handle_misc_aux_work_thr_prgr(ErtsAuxWorkData *awdp,
 			       aux_work & ~ERTS_SSI_AUX_WORK_MISC_THR_PRGR);
 }
 
-#endif
 
 static ERTS_INLINE void
 schedule_misc_aux_work(int sched_id,
@@ -2002,11 +1886,7 @@ schedule_misc_aux_work(int sched_id,
     ErtsThrQ_t *q;
     erts_misc_aux_work_t *mawp;
 
-#ifdef ERTS_SMP
     ASSERT(0 <= sched_id && sched_id <= erts_no_schedulers);
-#else
-    ASSERT(sched_id == 1);
-#endif
 
     q = &misc_aux_work_queues[sched_id].q;
     mawp = misc_aux_work_alloc();
@@ -2050,7 +1930,6 @@ erts_schedule_multi_misc_aux_work(int ignore_self,
    }
 }
 
-#if ERTS_USE_ASYNC_READY_Q
 
 void
 erts_notify_check_async_ready_queue(void *vno)
@@ -2066,9 +1945,9 @@ handle_async_ready(ErtsAuxWorkData *awdp,
 		   int waiting)
 {
     ErtsSchedulerSleepInfo *ssi = awdp->ssi;
-#ifdef ERTS_DIRTY_SCHEDULERS
+
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     unset_aux_work_flags(ssi, ERTS_SSI_AUX_WORK_ASYNC_READY);
     if (erts_check_async_ready(awdp->async_ready.queue)) {
 	if (set_aux_work_flags(ssi, ERTS_SSI_AUX_WORK_ASYNC_READY)
@@ -2078,9 +1957,7 @@ handle_async_ready(ErtsAuxWorkData *awdp,
 	}
 	return aux_work;
     }
-#ifdef ERTS_SMP
     awdp->async_ready.need_thr_prgr = 0;
-#endif
     set_aux_work_flags(ssi, ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN);
     return ((aux_work & ~ERTS_SSI_AUX_WORK_ASYNC_READY)
 	    | ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN);
@@ -2093,10 +1970,8 @@ handle_async_ready_clean(ErtsAuxWorkData *awdp,
 {
     void *thr_prgr_p;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
-#ifdef ERTS_SMP
+
     if (awdp->async_ready.need_thr_prgr
 	&& !erts_thr_progress_has_reached_this(haw_thr_prgr_current(awdp),
 					       awdp->async_ready.thr_prgr)) {
@@ -2105,26 +1980,20 @@ handle_async_ready_clean(ErtsAuxWorkData *awdp,
 
     awdp->async_ready.need_thr_prgr = 0;
     thr_prgr_p = (void *) &awdp->async_ready.thr_prgr;
-#else
-    thr_prgr_p = NULL;
-#endif
 
     switch (erts_async_ready_clean(awdp->async_ready.queue, thr_prgr_p)) {
     case ERTS_ASYNC_READY_CLEAN:
 	unset_aux_work_flags(awdp->ssi, ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN);
 	return aux_work & ~ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN;
-#ifdef ERTS_SMP
     case ERTS_ASYNC_READY_NEED_THR_PRGR:
 	haw_thr_prgr_soft_wakeup(awdp, awdp->async_ready.thr_prgr);
 	awdp->async_ready.need_thr_prgr = 1;
 	return aux_work & ~ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN;
-#endif
     default:
 	return aux_work;
     }
 }
 
-#endif /* ERTS_USE_ASYNC_READY_Q */
 
 
 static ERTS_INLINE erts_aint32_t
@@ -2133,9 +2002,8 @@ handle_fix_alloc(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
     ErtsSchedulerSleepInfo *ssi = awdp->ssi;
     erts_aint32_t res;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     unset_aux_work_flags(ssi, (ERTS_SSI_AUX_WORK_FIX_ALLOC_LOWER_LIM
 			       | ERTS_SSI_AUX_WORK_FIX_ALLOC_DEALLOC));
     aux_work &= ~(ERTS_SSI_AUX_WORK_FIX_ALLOC_LOWER_LIM
@@ -2149,7 +2017,6 @@ handle_fix_alloc(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
     return aux_work;
 }
 
-#ifdef ERTS_SMP
 
 void
 erts_alloc_notify_delayed_dealloc(int ix)
@@ -2183,9 +2050,9 @@ handle_delayed_dealloc(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waitin
     ErtsThrPrgrVal wakeup = ERTS_THR_PRGR_INVALID;
     int more_work = 0;
     ERTS_MSACC_PUSH_STATE_M_X();
-#ifdef ERTS_DIRTY_SCHEDULERS
+
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     unset_aux_work_flags(ssi, ERTS_SSI_AUX_WORK_DD);
     ERTS_MSACC_SET_STATE_CACHED_M_X(ERTS_MSACC_STATE_ALLOC);
     erts_alloc_scheduler_handle_delayed_dealloc((void *) awdp->esdp,
@@ -2222,9 +2089,8 @@ handle_delayed_dealloc_thr_prgr(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, i
     ErtsThrPrgrVal wakeup = ERTS_THR_PRGR_INVALID;
     ErtsThrPrgrVal current = haw_thr_prgr_current(awdp);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     if (!erts_thr_progress_has_reached_this(current, awdp->dd.thr_prgr))
 	return aux_work & ~ERTS_SSI_AUX_WORK_DD_THR_PRGR;
 
@@ -2281,9 +2147,8 @@ handle_canceled_timers(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waitin
     ErtsThrPrgrVal wakeup = ERTS_THR_PRGR_INVALID;
     int more_work = 0;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     unset_aux_work_flags(ssi, ERTS_SSI_AUX_WORK_CNCLD_TMRS);
     erts_handle_canceled_timers((void *) awdp->esdp,
 				&need_thr_progress,
@@ -2317,9 +2182,8 @@ handle_canceled_timers_thr_prgr(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, i
     ErtsThrPrgrVal wakeup = ERTS_THR_PRGR_INVALID;
     ErtsThrPrgrVal current = haw_thr_prgr_current(awdp);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     if (!erts_thr_progress_has_reached_this(current, awdp->cncld_tmrs.thr_prgr))
 	return aux_work & ~ERTS_SSI_AUX_WORK_CNCLD_TMRS_THR_PRGR;
 
@@ -2362,9 +2226,8 @@ handle_thr_prgr_later_op(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int wait
     int lops;
     ErtsThrPrgrVal current = haw_thr_prgr_current(awdp);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
+
     for (lops = 0; lops < ERTS_MAX_THR_PRGR_LATER_OPS; lops++) {
 	ErtsThrPrgrLaterOp *lop = awdp->later_op.first;
 
@@ -2394,7 +2257,7 @@ enqueue_later_op(ErtsSchedulerData *esdp,
 		 ErtsThrPrgrLaterOp *lop)
 {
     ErtsThrPrgrVal later = erts_thr_progress_later(esdp);
-    ASSERT(esdp);
+    ASSERT(esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp));
 
     lop->func = later_func;
     lop->data = later_data;
@@ -2410,20 +2273,15 @@ enqueue_later_op(ErtsSchedulerData *esdp,
     return later;
 }
 
-#endif /* ERTS_SMP */
 
 void
 erts_schedule_thr_prgr_later_op(void (*later_func)(void *),
 				void *later_data,
 				ErtsThrPrgrLaterOp *lop)
 {
-#ifndef ERTS_SMP
-    later_func(later_data);
-#else
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
     ErtsThrPrgrVal later = enqueue_later_op(esdp, later_func, later_data, lop);
     haw_thr_prgr_wakeup(&esdp->aux_work_data, later);
-#endif
 }
 
 void
@@ -2432,13 +2290,9 @@ erts_schedule_thr_prgr_later_cleanup_op(void (*later_func)(void *),
 					ErtsThrPrgrLaterOp *lop,
 					UWord size)
 {
-#ifndef ERTS_SMP
-    later_func(later_data);
-#else
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
     ErtsThrPrgrVal later = enqueue_later_op(esdp, later_func, later_data, lop);
     haw_thr_prgr_later_cleanup_op_wakeup(&esdp->aux_work_data, later, size);
-#endif
 }
 
 static ERTS_INLINE erts_aint32_t
@@ -2447,9 +2301,7 @@ handle_debug_wait_completed(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int w
     ErtsSchedulerSleepInfo *ssi = awdp->ssi;
     erts_aint32_t saved_aux_work, flags;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#endif
 
     flags = awdp->debug.wait_completed.flags;
 
@@ -2490,11 +2342,7 @@ setup_thr_debug_wait_completed(void *vproc)
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
     ErtsAuxWorkData *awdp;
     erts_aint32_t wait_flags, aux_work_flags;
-#ifdef ERTS_SMP
     awdp = esdp ? &esdp->aux_work_data : aux_thread_aux_work_data;
-#else
-    awdp = &esdp->aux_work_data;
-#endif
 
     wait_flags = 0;
     aux_work_flags = ERTS_SSI_AUX_WORK_DEBUG_WAIT_COMPLETED;
@@ -2503,18 +2351,14 @@ setup_thr_debug_wait_completed(void *vproc)
 	erts_alloc_fix_alloc_shrink(awdp->sched_id, 0);
 	wait_flags |= (ERTS_SSI_AUX_WORK_DD
 		       | ERTS_SSI_AUX_WORK_DD_THR_PRGR);
-#ifdef ERTS_SMP
 	aux_work_flags |= ERTS_SSI_AUX_WORK_DD;
-#endif
     }
 
     if (debug_wait_completed_flags & ERTS_DEBUG_WAIT_COMPLETED_TIMER_CANCELLATIONS) {
 	wait_flags |= (ERTS_SSI_AUX_WORK_CNCLD_TMRS
 		       | ERTS_SSI_AUX_WORK_CNCLD_TMRS_THR_PRGR);
-#ifdef ERTS_SMP
 	if (awdp->esdp && !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp))
 	    aux_work_flags |= ERTS_SSI_AUX_WORK_CNCLD_TMRS;
-#endif
     }
 
     set_aux_work_flags_wakeup_nob(awdp->ssi, aux_work_flags);
@@ -2533,21 +2377,17 @@ static void later_thr_debug_wait_completed(void *vlop)
 {
     struct debug_lop *lop = vlop;
     erts_aint32_t count = (erts_aint32_t) erts_no_schedulers;
-#ifdef ERTS_SMP
     count += 1; /* aux thread */
-#endif
     if (erts_atomic32_dec_read_mb(&debug_wait_completed_count) == count) {
         /* scheduler threads */
         erts_schedule_multi_misc_aux_work(0,
                                           erts_no_schedulers,
                                           setup_thr_debug_wait_completed,
                                           lop->proc);
-#ifdef ERTS_SMP
         /* aux_thread */
         erts_schedule_misc_aux_work(0,
                                     setup_thr_debug_wait_completed,
                                     lop->proc);
-#endif
     }
     erts_free(ERTS_ALC_T_DEBUG, lop);
 }
@@ -2568,9 +2408,7 @@ erts_debug_wait_completed(Process *c_p, int flags)
 {
     /* Only one process at a time can do this */
     erts_aint32_t count = (erts_aint32_t) (2*erts_no_schedulers);
-#ifdef ERTS_SMP
     count += 1; /* aux thread */
-#endif
     if (0 == erts_atomic32_cmpxchg_mb(&debug_wait_completed_count,
 				      count,
 				      0)) {
@@ -2599,7 +2437,7 @@ notify_reap_ports_relb(void)
     }
 }
 
-erts_smp_atomic32_t erts_halt_progress;
+erts_atomic32_t erts_halt_progress;
 int erts_halt_code;
 
 static ERTS_INLINE erts_aint32_t
@@ -2608,9 +2446,9 @@ handle_reap_ports(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
     unset_aux_work_flags(awdp->ssi, ERTS_SSI_AUX_WORK_REAP_PORTS);
     ERTS_RUNQ_FLGS_SET(awdp->esdp->run_queue, ERTS_RUNQ_FLG_HALTING);
 
-    if (erts_smp_atomic32_dec_read_acqb(&erts_halt_progress) == 0) {
+    if (erts_atomic32_dec_read_acqb(&erts_halt_progress) == 0) {
 	int i, max = erts_ptab_max(&erts_port);
-	erts_smp_atomic32_set_nob(&erts_halt_progress, 1);
+	erts_atomic32_set_nob(&erts_halt_progress, 1);
 	for (i = 0; i < max; i++) {
 	    erts_aint32_t state;
 	    Port *prt = erts_pix2port(i);
@@ -2623,21 +2461,21 @@ handle_reap_ports(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
 
 	    /* We need to set the halt flag - get the port lock */
 
-	    erts_smp_port_lock(prt);
+	    erts_port_lock(prt);
 
 	    state = erts_atomic32_read_nob(&prt->state);
 	    if (!(state & (ERTS_PORT_SFLGS_INVALID_DRIVER_LOOKUP
 			   | ERTS_PORT_SFLG_HALT))) {
 		state = erts_atomic32_read_bor_relb(&prt->state,
 						    ERTS_PORT_SFLG_HALT);
-		erts_smp_atomic32_inc_nob(&erts_halt_progress);
+		erts_atomic32_inc_nob(&erts_halt_progress);
 		if (!(state & (ERTS_PORT_SFLG_EXITING|ERTS_PORT_SFLG_CLOSING)))
 		    erts_deliver_port_exit(prt, prt->common.id, am_killed, 0, 1);
 	    }
 
 	    erts_port_release(prt);
 	}
-	if (erts_smp_atomic32_dec_read_nob(&erts_halt_progress) == 0) {
+	if (erts_atomic32_dec_read_nob(&erts_halt_progress) == 0) {
 	    erts_flush_async_exit(erts_halt_code, "");
 	}
     }
@@ -2671,6 +2509,8 @@ handle_yield(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
 
     yield |= erts_handle_yielded_ets_all_request(awdp->esdp,
                                                  &awdp->yield.ets_all);
+    yield |= erts_handle_yielded_alcu_blockscan(awdp->esdp,
+                                                &awdp->yield.alcu_blockscan);
 
     /*
      * Other yielding operations...
@@ -2698,29 +2538,6 @@ handle_mseg_cache_check(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiti
 
 #endif
 
-#ifdef ERTS_SMP
-
-static ERTS_INLINE erts_aint32_t
-handle_pending_exiters(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
-{
-    ErtsProcList *pnd_xtrs;
-    ErtsRunQueue *rq;
-
-    rq = awdp->esdp->run_queue;
-    unset_aux_work_flags(awdp->ssi, ERTS_SSI_AUX_WORK_PENDING_EXITERS);
-
-    erts_smp_runq_lock(rq);
-    pnd_xtrs = rq->procs.pending_exiters;
-    rq->procs.pending_exiters = NULL;
-    erts_smp_runq_unlock(rq);
-
-    if (erts_proclist_fetch(&pnd_xtrs, NULL))
-	do_handle_pending_exiters(pnd_xtrs);
-
-    return aux_work & ~ERTS_SSI_AUX_WORK_PENDING_EXITERS;
-}
-
-#endif
 
 static ERTS_INLINE erts_aint32_t
 handle_setup_aux_work_timer(ErtsAuxWorkData *awdp, erts_aint32_t aux_work, int waiting)
@@ -2752,9 +2569,7 @@ handle_aux_work(ErtsAuxWorkData *awdp, erts_aint32_t orig_aux_work, int waiting)
     ERTS_MSACC_PUSH_AND_SET_STATE_M(ERTS_MSACC_STATE_AUX);
 
     ASSERT(!awdp->esdp || !ERTS_SCHEDULER_IS_DIRTY(awdp->esdp));
-#ifdef ERTS_SMP
     haw_thr_prgr_current_reset(awdp);
-#endif
 
     ERTS_DBG_CHK_AUX_WORK_VAL(aux_work);
     ASSERT(aux_work);
@@ -2773,7 +2588,6 @@ handle_aux_work(ErtsAuxWorkData *awdp, erts_aint32_t orig_aux_work, int waiting)
      * Keep ERTS_SSI_AUX_WORK flags in expected frequency order relative
      * eachother. Most frequent first.
      */
-#ifdef ERTS_SMP
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_DELAYED_AW_WAKEUP,
 		    handle_delayed_aux_work_wakeup);
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_DD,
@@ -2781,13 +2595,11 @@ handle_aux_work(ErtsAuxWorkData *awdp, erts_aint32_t orig_aux_work, int waiting)
     /* DD must be before DD_THR_PRGR */
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_DD_THR_PRGR,
 		    handle_delayed_dealloc_thr_prgr);
-#endif
 
     HANDLE_AUX_WORK((ERTS_SSI_AUX_WORK_FIX_ALLOC_LOWER_LIM
 		     | ERTS_SSI_AUX_WORK_FIX_ALLOC_DEALLOC),
 		    handle_fix_alloc);
 
-#ifdef ERTS_SMP
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_THR_PRGR_LATER_OP,
 		    handle_thr_prgr_later_op);
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_CNCLD_TMRS,
@@ -2795,29 +2607,19 @@ handle_aux_work(ErtsAuxWorkData *awdp, erts_aint32_t orig_aux_work, int waiting)
     /* CNCLD_TMRS must be before CNCLD_TMRS_THR_PRGR */
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_CNCLD_TMRS_THR_PRGR,
 		    handle_canceled_timers_thr_prgr);
-#endif
 
-#if ERTS_USE_ASYNC_READY_Q
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_ASYNC_READY,
 		    handle_async_ready);
     /* ASYNC_READY must be before ASYNC_READY_CLEAN */
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN,
 		    handle_async_ready_clean);
-#endif
 
-#ifdef ERTS_SMP
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_MISC_THR_PRGR,
 		    handle_misc_aux_work_thr_prgr);
-#endif
     /* MISC_THR_PRGR must be before MISC */
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_MISC,
 		    handle_misc_aux_work);
 
-#ifdef ERTS_SMP
-    HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_PENDING_EXITERS,
-		    handle_pending_exiters);
-#endif
-
     HANDLE_AUX_WORK(ERTS_SSI_AUX_WORK_SET_TMO,
 		    handle_setup_aux_work_timer);
 
@@ -2842,10 +2644,8 @@ handle_aux_work(ErtsAuxWorkData *awdp, erts_aint32_t orig_aux_work, int waiting)
 
     ERTS_DBG_CHK_AUX_WORK_VAL(aux_work);
 
-#ifdef ERTS_SMP
     if (waiting && !aux_work)
 	haw_thr_prgr_current_check_progress(awdp);
-#endif
 
     ERTS_MSACC_UPDATE_CACHE();
     ERTS_MSACC_POP_STATE_M();
@@ -2944,11 +2744,7 @@ aux_work_timeout(void *vesdp)
     ASSERT(esdp == (ErtsSchedulerData *) vesdp);
 #endif
 
-#ifdef ERTS_SMP
     i = 0;
-#else
-    i = 1;
-#endif
 
     for (; i <= erts_no_schedulers; i++) {
 	erts_aint32_t type;
@@ -2982,9 +2778,6 @@ erts_set_aux_work_timeout(int ix, erts_aint32_t type, int enable)
 {
     erts_aint32_t old, refc;
 
-#ifndef ERTS_SMP
-    ix = 1;
-#endif
 
     ERTS_DBG_CHK_AUX_WORK_VAL(type);
     ERTS_DBG_CHK_AUX_WORK_VAL(erts_atomic32_read_nob(&aux_work_tmo->type[ix]));
@@ -3011,36 +2804,6 @@ erts_set_aux_work_timeout(int ix, erts_aint32_t type, int enable)
     return old;
 }
 
-
-
-static ERTS_INLINE void
-sched_waiting_sys(Uint no, ErtsRunQueue *rq)
-{
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
-    ASSERT(rq->waiting >= 0);
-    (void) ERTS_RUNQ_FLGS_SET(rq, (ERTS_RUNQ_FLG_OUT_OF_WORK
-				   | ERTS_RUNQ_FLG_HALFTIME_OUT_OF_WORK));
-    rq->waiting++;
-    rq->waiting *= -1;
-    rq->woken = 0;
-    if (erts_system_profile_flags.scheduler)
-	profile_scheduler(make_small(no), am_inactive);
-}
-
-static ERTS_INLINE void
-sched_active_sys(Uint no, ErtsRunQueue *rq)
-{
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
-#ifdef ERTS_DIRTY_SCHEDULERS
-    ASSERT(!ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
-#endif
-    ASSERT(rq->waiting < 0);
-    rq->waiting *= -1;
-    rq->waiting--;
-    if (erts_system_profile_flags.scheduler)
-	profile_scheduler(make_small(no), am_active);
-}
-
 Uint
 erts_active_schedulers(void)
 {
@@ -3051,64 +2814,10 @@ erts_active_schedulers(void)
     return as;
 }
 
-#ifdef ERTS_SMP
-
-static ERTS_INLINE void
-clear_sys_scheduling(void)
-{
-    erts_smp_atomic32_set_mb(&doing_sys_schedule, 0);
-}
-
-static ERTS_INLINE int
-try_set_sys_scheduling(void)
-{
-    return 0 == erts_smp_atomic32_cmpxchg_acqb(&doing_sys_schedule, 1, 0);
-}
-
-#endif
-
-static ERTS_INLINE int
-prepare_for_sys_schedule(int non_blocking)
-{
-    if (non_blocking && erts_eager_check_io) {
-#ifdef ERTS_SMP
-	return try_set_sys_scheduling();
-#else
-	return 1;
-#endif
-    }
-    else {
-#ifdef ERTS_SMP
-	while (!erts_port_task_have_outstanding_io_tasks()
-	       && try_set_sys_scheduling()) {
-	    if (!erts_port_task_have_outstanding_io_tasks())
-		return 1;
-	    clear_sys_scheduling();
-	}
-	return 0;
-#else
-	return !erts_port_task_have_outstanding_io_tasks();
-#endif
-    }
-}
-
-#ifdef ERTS_SMP
-
-static ERTS_INLINE void
-sched_change_waiting_sys_to_waiting(Uint no, ErtsRunQueue *rq)
-{
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
-#ifdef ERTS_DIRTY_SCHEDULERS
-    ASSERT(!ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
-#endif
-    ASSERT(rq->waiting < 0);
-    rq->waiting *= -1;
-}
-
 static ERTS_INLINE void
 sched_waiting(Uint no, ErtsRunQueue *rq)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
     (void) ERTS_RUNQ_FLGS_SET(rq, (ERTS_RUNQ_FLG_OUT_OF_WORK
 				   | ERTS_RUNQ_FLG_HALFTIME_OUT_OF_WORK));
     if (rq->waiting < 0)
@@ -3123,7 +2832,7 @@ sched_waiting(Uint no, ErtsRunQueue *rq)
 static ERTS_INLINE void
 sched_active(Uint no, ErtsRunQueue *rq)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
     if (rq->waiting < 0)
 	rq->waiting++;
     else
@@ -3137,7 +2846,7 @@ empty_runq_aux(ErtsRunQueue *rq, Uint32 old_flags)
 {
     if (!ERTS_RUNQ_IX_IS_DIRTY(rq->ix) && old_flags & ERTS_RUNQ_FLG_NONEMPTY) {
 #ifdef DEBUG
-	erts_aint32_t empty = erts_smp_atomic32_read_nob(&no_empty_run_queues);
+	erts_aint32_t empty = erts_atomic32_read_nob(&no_empty_run_queues);
 	/*
 	 * For a short period of time no_empty_run_queues may have
 	 * been increased twice for a specific run queue.
@@ -3145,9 +2854,9 @@ empty_runq_aux(ErtsRunQueue *rq, Uint32 old_flags)
 	ASSERT(0 <= empty && empty < 2*erts_no_run_queues);
 #endif
 	if (!erts_runq_supervision_interval)
-	    erts_smp_atomic32_inc_relb(&no_empty_run_queues);
+	    erts_atomic32_inc_relb(&no_empty_run_queues);
 	else {
-	    erts_smp_atomic32_inc_mb(&no_empty_run_queues);
+	    erts_atomic32_inc_mb(&no_empty_run_queues);
 	    if (erts_atomic_read_nob(&runq_supervisor_sleeping))
 		ethr_event_set(&runq_supervision_event);
 	}
@@ -3177,7 +2886,7 @@ non_empty_runq(ErtsRunQueue *rq)
     Uint32 old_flags = ERTS_RUNQ_FLGS_SET(rq, ERTS_RUNQ_FLG_NONEMPTY);
     if (!ERTS_RUNQ_IX_IS_DIRTY(rq->ix) && (!(old_flags & ERTS_RUNQ_FLG_NONEMPTY))) {
 #ifdef DEBUG
-	erts_aint32_t empty = erts_smp_atomic32_read_nob(&no_empty_run_queues);
+	erts_aint32_t empty = erts_atomic32_read_nob(&no_empty_run_queues);
 	/*
 	 * For a short period of time no_empty_run_queues may have
 	 * been increased twice for a specific run queue.
@@ -3185,10 +2894,10 @@ non_empty_runq(ErtsRunQueue *rq)
 	ASSERT(0 < empty && empty <= 2*erts_no_run_queues);
 #endif
 	if (!erts_runq_supervision_interval)
-	    erts_smp_atomic32_dec_relb(&no_empty_run_queues);
+	    erts_atomic32_dec_relb(&no_empty_run_queues);
 	else {
 	    erts_aint32_t no;
-	    no = erts_smp_atomic32_dec_read_mb(&no_empty_run_queues);
+	    no = erts_atomic32_dec_read_mb(&no_empty_run_queues);
 	    if (no > 0 && erts_atomic_read_nob(&runq_supervisor_sleeping))
 		ethr_event_set(&runq_supervision_event);
 	}
@@ -3217,7 +2926,7 @@ sched_prep_spin_wait(ErtsSchedulerSleepInfo *ssi)
     do {
         nflgs = (xflgs & ERTS_SSI_FLG_MSB_EXEC);
         nflgs |= ERTS_SSI_FLG_SLEEPING|ERTS_SSI_FLG_WAITING;
-	oflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return nflgs;
 	xflgs = oflgs;
@@ -3234,7 +2943,7 @@ sched_prep_cont_spin_wait(ErtsSchedulerSleepInfo *ssi)
     erts_aint32_t xflgs = ERTS_SSI_FLG_WAITING;
 
     do {
-	oflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return nflgs;
 	xflgs = oflgs;
@@ -3251,7 +2960,7 @@ sched_spin_wait(ErtsSchedulerSleepInfo *ssi, int spincount)
     erts_aint32_t flgs;
 
     do {
-	flgs = erts_smp_atomic32_read_acqb(&ssi->flags);
+	flgs = erts_atomic32_read_acqb(&ssi->flags);
 	if ((flgs & (ERTS_SSI_FLG_SLEEPING|ERTS_SSI_FLG_WAITING))
 	    != (ERTS_SSI_FLG_SLEEPING|ERTS_SSI_FLG_WAITING)) {
 	    break;
@@ -3276,11 +2985,11 @@ sched_set_sleeptype(ErtsSchedulerSleepInfo *ssi, erts_aint32_t sleep_type)
 	erts_tse_reset(ssi->event);
     else {
 	ASSERT(sleep_type == ERTS_SSI_FLG_POLL_SLEEPING);
-	erts_sys_schedule_interrupt(0);
+        erts_check_io_interrupt(ssi->psi, 0);
     }
 
     while (1) {
-	oflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return nflgs;
 	if ((oflgs & (ERTS_SSI_FLG_SLEEPING|ERTS_SSI_FLG_WAITING))
@@ -3307,7 +3016,7 @@ static void
 thr_prgr_prep_wait(void *vssi)
 {
     ErtsSchedulerSleepInfo *ssi = (ErtsSchedulerSleepInfo *) vssi;
-    erts_smp_atomic32_read_bor_acqb(&ssi->flags,
+    erts_atomic32_read_bor_acqb(&ssi->flags,
 				    ERTS_SSI_FLG_SLEEPING);
 }
 
@@ -3322,7 +3031,7 @@ thr_prgr_wait(void *vssi)
     while (1) {
 	erts_aint32_t aflgs, nflgs;
 	nflgs = xflgs | ERTS_SSI_FLG_TSE_SLEEPING;
-	aflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	aflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (aflgs == xflgs) {
 	    erts_tse_wait(ssi->event);
 	    break;
@@ -3337,13 +3046,19 @@ static void
 thr_prgr_fin_wait(void *vssi)
 {
     ErtsSchedulerSleepInfo *ssi = (ErtsSchedulerSleepInfo *) vssi;
-    erts_smp_atomic32_read_band_nob(&ssi->flags,
+    erts_atomic32_read_band_nob(&ssi->flags,
 				    ~(ERTS_SSI_FLG_SLEEPING
 				      | ERTS_SSI_FLG_TSE_SLEEPING));
 }
 
 static void init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp);
 
+void
+erts_aux_thread_poke()
+{
+    erts_sched_poke(ERTS_SCHED_SLEEP_INFO_IX(-1));
+}
+
 static void *
 aux_thread(void *unused)
 {
@@ -3352,6 +3067,7 @@ aux_thread(void *unused)
     erts_aint32_t aux_work;
     ErtsThrPrgrCallbacks callbacks;
     int thr_prgr_active = 1;
+    ERTS_MSACC_DECLARE_CACHE();
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
     {
@@ -3360,6 +3076,7 @@ aux_thread(void *unused)
     }
 #endif
 
+    erts_port_task_pre_alloc_init_thread();
     ssi->event = erts_tse_fetch();
 
     erts_msacc_init_thread("aux", 1, 1);
@@ -3374,9 +3091,14 @@ aux_thread(void *unused)
     init_aux_work_data(awdp, NULL, NULL);
     awdp->ssi = ssi;
 
+#if ERTS_POLL_USE_FALLBACK
+    ssi->psi = erts_create_pollset_thread(-1);
+#endif
 
     sched_prep_spin_wait(ssi);
 
+    ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER);
+
     while (1) {
 	erts_aint32_t flgs;
 
@@ -3385,30 +3107,54 @@ aux_thread(void *unused)
 	    if (!thr_prgr_active)
 		erts_thr_progress_active(NULL, thr_prgr_active = 1);
 	    aux_work = handle_aux_work(awdp, aux_work, 1);
+            ERTS_MSACC_UPDATE_CACHE();
 	    if (aux_work && erts_thr_progress_update(NULL))
 		erts_thr_progress_leader_update(NULL);
 	}
 
 	if (!aux_work) {
+
+#ifdef ERTS_BREAK_REQUESTED
+            if (ERTS_BREAK_REQUESTED)
+                erts_do_break_handling();
+#endif
+
 	    if (thr_prgr_active)
 		erts_thr_progress_active(NULL, thr_prgr_active = 0);
-	    erts_thr_progress_prepare_wait(NULL);
+
+#if ERTS_POLL_USE_FALLBACK
 
 	    flgs = sched_spin_wait(ssi, 0);
 
 	    if (flgs & ERTS_SSI_FLG_SLEEPING) {
 		ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+		flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_POLL_SLEEPING);
+		if (flgs & ERTS_SSI_FLG_SLEEPING) {
+		    ASSERT(flgs & ERTS_SSI_FLG_POLL_SLEEPING);
+		    ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+                    erts_check_io(ssi->psi);
+		}
+            }
+#else
+            erts_thr_progress_prepare_wait(NULL);
+
+	    flgs = sched_spin_wait(ssi, 0);
+
+	    if (flgs & ERTS_SSI_FLG_SLEEPING) {
 		flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_TSE_SLEEPING);
 		if (flgs & ERTS_SSI_FLG_SLEEPING) {
-		    int res;
+                    int res;
 		    ASSERT(flgs & ERTS_SSI_FLG_TSE_SLEEPING);
 		    ASSERT(flgs & ERTS_SSI_FLG_WAITING);
-		    do {
-			res = erts_tse_wait(ssi->event);
-		    } while (res == EINTR);
+                    ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
+                    do {
+                        res = erts_tse_wait(ssi->event);
+                    } while (res == EINTR);
+                    ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER);
 		}
-	    }
-	    erts_thr_progress_finalize_wait(NULL);
+            }
+            erts_thr_progress_finalize_wait(NULL);
+#endif
 	}
 
 	flgs = sched_prep_spin_wait(ssi);
@@ -3416,9 +3162,79 @@ aux_thread(void *unused)
     return NULL;
 }
 
-static void suspend_scheduler(ErtsSchedulerData *esdp);
+static void *
+poll_thread(void *arg)
+{
+    int id = (int)(UWord)arg;
+    ErtsAuxWorkData *awdp = poll_thread_aux_work_data+id;
+    ErtsSchedulerSleepInfo *ssi = ERTS_POLL_THREAD_SLEEP_INFO_IX(id);
+    erts_aint32_t aux_work;
+    ErtsThrPrgrCallbacks callbacks;
+    int thr_prgr_active = 1;
+    struct erts_poll_thread *psi = erts_create_pollset_thread(id);
+    ERTS_MSACC_DECLARE_CACHE();
+
+#ifdef ERTS_ENABLE_LOCK_CHECK
+    {
+	char buf[] = "poll_thread";
+	erts_lc_set_thread_name(buf);
+    }
+#endif
+
+    erts_port_task_pre_alloc_init_thread();
+    ssi->event = erts_tse_fetch();
+
+    erts_msacc_init_thread("poll", id, 0);
+
+    callbacks.arg = (void *) ssi;
+    callbacks.wakeup = thr_prgr_wakeup;
+    callbacks.prepare_wait = thr_prgr_prep_wait;
+    callbacks.wait = thr_prgr_wait;
+    callbacks.finalize_wait = thr_prgr_fin_wait;
+
+    erts_thr_progress_register_managed_thread(NULL, &callbacks, 0);
+    init_aux_work_data(awdp, NULL, NULL);
+    awdp->ssi = ssi;
+    ssi->psi = psi;
+
+    sched_prep_spin_wait(ssi);
+
+    ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER);
+
+    while (1) {
+	erts_aint32_t flgs;
+
+	aux_work = erts_atomic32_read_acqb(&ssi->aux_work);
+	if (aux_work) {
+	    if (!thr_prgr_active)
+		erts_thr_progress_active(NULL, thr_prgr_active = 1);
+	    aux_work = handle_aux_work(awdp, aux_work, 1);
+            ERTS_MSACC_UPDATE_CACHE();
+	    if (aux_work && erts_thr_progress_update(NULL))
+		erts_thr_progress_leader_update(NULL);
+	}
 
-#endif /* ERTS_SMP */
+	if (!aux_work) {
+	    if (thr_prgr_active)
+		erts_thr_progress_active(NULL, thr_prgr_active = 0);
+
+	    flgs = sched_spin_wait(ssi, 0);
+
+	    if (flgs & ERTS_SSI_FLG_SLEEPING) {
+		ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+		flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_POLL_SLEEPING);
+		if (flgs & ERTS_SSI_FLG_SLEEPING) {
+		    ASSERT(flgs & ERTS_SSI_FLG_POLL_SLEEPING);
+		    ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+                    erts_check_io(psi);
+		}
+	    }
+	}
+
+	flgs = sched_prep_spin_wait(ssi);
+    }
+    return NULL;
+}
 
 static void
 scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq)
@@ -3427,386 +3243,168 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq)
     ErtsSchedulerSleepInfo *ssi = esdp->ssi;
     int spincount;
     erts_aint32_t aux_work = 0;
-#ifdef ERTS_SMP
     int thr_prgr_active = 1;
     erts_aint32_t flgs;
-#endif
     ERTS_MSACC_PUSH_STATE();
-#ifdef ERTS_SMP
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_RUNQ_IX_IS_DIRTY(rq->ix))
-	erts_smp_spin_lock(&rq->sleepers.lock);
-#endif
+	erts_spin_lock(&rq->sleepers.lock);
     flgs = sched_prep_spin_wait(ssi);
     if (flgs & ERTS_SSI_FLG_SUSPENDED) {
 	/* Go suspend instead... */
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (ERTS_RUNQ_IX_IS_DIRTY(rq->ix))
-	    erts_smp_spin_unlock(&rq->sleepers.lock);
-#endif
+	    erts_spin_unlock(&rq->sleepers.lock);
 	return;
     }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (ERTS_RUNQ_IX_IS_DIRTY(rq->ix)) {
 	ssi->prev = NULL;
 	ssi->next = rq->sleepers.list;
 	if (rq->sleepers.list)
 	    rq->sleepers.list->prev = ssi;
 	rq->sleepers.list = ssi;
-	erts_smp_spin_unlock(&rq->sleepers.lock);
+	erts_spin_unlock(&rq->sleepers.lock);
         dirty_active(esdp, -1);
     }
-#endif
-
-    /*
-     * If all schedulers are waiting, one of them *should*
-     * be waiting in erl_sys_schedule()
-     */
-
-    if (ERTS_SCHEDULER_IS_DIRTY(esdp) || !prepare_for_sys_schedule(0)) {
-
-	sched_waiting(esdp->no, rq);
 
-	erts_smp_runq_unlock(rq);
+    sched_waiting(esdp->no, rq);
 
-	spincount = sched_busy_wait.tse;
+    erts_runq_unlock(rq);
 
-    tse_wait:
+    spincount = sched_get_busy_wait_params(esdp)->tse;
 
-	if (ERTS_SCHEDULER_IS_DIRTY(esdp))
-	    dirty_sched_wall_time_change(esdp, working = 0);
-        else if (thr_prgr_active != working)
-	    sched_wall_time_change(esdp, working = thr_prgr_active);
-
-	while (1) {
-	    ErtsMonotonicTime current_time = 0;
-
-	    aux_work = erts_atomic32_read_acqb(&ssi->aux_work);
-	    if (aux_work && !ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-		if (!thr_prgr_active) {
-		    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-		    sched_wall_time_change(esdp, 1);
-		}
-		aux_work = handle_aux_work(&esdp->aux_work_data, aux_work, 1);
-                ERTS_MSACC_UPDATE_CACHE();
-		if (aux_work && erts_thr_progress_update(esdp))
-		    erts_thr_progress_leader_update(esdp);
-	    }
-
-	    if (aux_work) {
-		if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-		    flgs = erts_smp_atomic32_read_acqb(&ssi->flags);
-		    current_time = erts_get_monotonic_time(esdp);
-		    if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref)) {
-			if (!thr_prgr_active) {
-			    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-			    sched_wall_time_change(esdp, 1);
-			}
-			erts_bump_timers(esdp->timer_wheel, current_time);
-		    }
-		}
-	    }
-	    else {
-		ErtsMonotonicTime timeout_time;
-		int do_timeout = 0;
-		if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-		    timeout_time = erts_check_next_timeout_time(esdp);
-		    current_time = erts_get_monotonic_time(esdp);
-		    do_timeout = (current_time >= timeout_time);
-		} else {
-		    current_time = 0;
-		    timeout_time = ERTS_MONOTONIC_TIME_MAX;
-		}
-		if (do_timeout) {
-		    if (!thr_prgr_active) {
-			erts_thr_progress_active(esdp, thr_prgr_active = 1);
-			sched_wall_time_change(esdp, 1);
-		    }
-		}
-		else {
-		    if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-			if (thr_prgr_active) {
-			    erts_thr_progress_active(esdp, thr_prgr_active = 0);
-			    sched_wall_time_change(esdp, 0);
-			}
-			erts_thr_progress_prepare_wait(esdp);
-		    }
+    if (ERTS_SCHEDULER_IS_DIRTY(esdp))
+        dirty_sched_wall_time_change(esdp, working = 0);
+    else if (thr_prgr_active != working)
+        sched_wall_time_change(esdp, working = thr_prgr_active);
 
-		    flgs = sched_spin_wait(ssi, spincount);
-		    if (flgs & ERTS_SSI_FLG_SLEEPING) {
-			ASSERT(flgs & ERTS_SSI_FLG_WAITING);
-			flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_TSE_SLEEPING);
-			if (flgs & ERTS_SSI_FLG_SLEEPING) {
-			    int res;
-			    ASSERT(flgs & ERTS_SSI_FLG_TSE_SLEEPING);
-			    ASSERT(flgs & ERTS_SSI_FLG_WAITING);
-			    current_time = ERTS_SCHEDULER_IS_DIRTY(esdp) ? 0 :
-				erts_get_monotonic_time(esdp);
-			    do {
-				Sint64 timeout;
-				if (current_time >= timeout_time)
-				    break;
-				if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-				    timeout = ERTS_MONOTONIC_TO_NSEC(timeout_time
-								     - current_time
-								     - 1) + 1;
-				} else
-				    timeout = -1;
-                                ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
-				res = erts_tse_twait(ssi->event, timeout);
-                                ERTS_MSACC_POP_STATE();
-				current_time = ERTS_SCHEDULER_IS_DIRTY(esdp) ? 0 :
-				    erts_get_monotonic_time(esdp);
-			    } while (res == EINTR);
-			}
-		    }
-		    if (!ERTS_SCHEDULER_IS_DIRTY(esdp))
-			erts_thr_progress_finalize_wait(esdp);
-		}
-		if (!ERTS_SCHEDULER_IS_DIRTY(esdp) && current_time >= timeout_time)
-		    erts_bump_timers(esdp->timer_wheel, current_time);
-	    }
+    while (1) {
+        ErtsMonotonicTime current_time = 0;
 
-	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
-		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
-		break;
-	    }
+        aux_work = erts_atomic32_read_acqb(&ssi->aux_work);
+        if (aux_work && !ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+            if (!thr_prgr_active) {
+                erts_thr_progress_active(esdp, thr_prgr_active = 1);
+                sched_wall_time_change(esdp, 1);
+            }
+            aux_work = handle_aux_work(&esdp->aux_work_data, aux_work, 1);
+            ERTS_MSACC_UPDATE_CACHE();
+            if (aux_work && erts_thr_progress_update(esdp))
+                erts_thr_progress_leader_update(esdp);
+        }
 
-	    flgs = sched_prep_cont_spin_wait(ssi);
-	    spincount = sched_busy_wait.aux_work;
+        if (aux_work) {
+            if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+                flgs = erts_atomic32_read_acqb(&ssi->flags);
+                current_time = erts_get_monotonic_time(esdp);
+                if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref)) {
+                    if (!thr_prgr_active) {
+                        erts_thr_progress_active(esdp, thr_prgr_active = 1);
+                        sched_wall_time_change(esdp, 1);
+                    }
+                    erts_bump_timers(esdp->timer_wheel, current_time);
+                }
+            }
+        }
+        else {
+            ErtsMonotonicTime timeout_time;
+            int do_timeout = 0;
+            if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+                timeout_time = erts_check_next_timeout_time(esdp);
+                current_time = erts_get_monotonic_time(esdp);
+                do_timeout = (current_time >= timeout_time);
+            } else {
+                current_time = 0;
+                timeout_time = ERTS_MONOTONIC_TIME_MAX;
+            }
+            if (do_timeout) {
+                if (!thr_prgr_active) {
+                    erts_thr_progress_active(esdp, thr_prgr_active = 1);
+                    sched_wall_time_change(esdp, 1);
+                }
+            }
+            else {
+                if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+                    if (thr_prgr_active) {
+                        erts_thr_progress_active(esdp, thr_prgr_active = 0);
+                        sched_wall_time_change(esdp, 0);
+                    }
+                    erts_thr_progress_prepare_wait(esdp);
+                }
 
-	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
-		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
-		break;
-	    }
+                flgs = sched_spin_wait(ssi, spincount);
+                if (flgs & ERTS_SSI_FLG_SLEEPING) {
+                    ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+                    flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_TSE_SLEEPING);
+                    if (flgs & ERTS_SSI_FLG_SLEEPING) {
+                        int res;
+                        ASSERT(flgs & ERTS_SSI_FLG_TSE_SLEEPING);
+                        ASSERT(flgs & ERTS_SSI_FLG_WAITING);
+                        current_time = ERTS_SCHEDULER_IS_DIRTY(esdp) ? 0 :
+                            erts_get_monotonic_time(esdp);
+                        do {
+                            Sint64 timeout;
+                            if (current_time >= timeout_time)
+                                break;
+                            if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
+                                timeout = ERTS_MONOTONIC_TO_NSEC(timeout_time
+                                                                 - current_time
+                                                                 - 1) + 1;
+                            } else
+                                timeout = -1;
+                            ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
+                            res = erts_tse_twait(ssi->event, timeout);
+                            ERTS_MSACC_POP_STATE();
+                            current_time = ERTS_SCHEDULER_IS_DIRTY(esdp) ? 0 :
+                                erts_get_monotonic_time(esdp);
+                        } while (res == EINTR);
+                    }
+                }
+                if (!ERTS_SCHEDULER_IS_DIRTY(esdp))
+                    erts_thr_progress_finalize_wait(esdp);
+            }
+            if (!ERTS_SCHEDULER_IS_DIRTY(esdp) && current_time >= timeout_time)
+                erts_bump_timers(esdp->timer_wheel, current_time);
+        }
 
-	}
+        if (!(flgs & ERTS_SSI_FLG_WAITING)) {
+            ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
+            break;
+        }
 
-	if (flgs & ~(ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC))
-	    erts_smp_atomic32_read_band_nob(&ssi->flags,
-                                            (ERTS_SSI_FLG_SUSPENDED
-                                             | ERTS_SSI_FLG_MSB_EXEC));
+        flgs = sched_prep_cont_spin_wait(ssi);
+        spincount = sched_get_busy_wait_params(esdp)->aux_work;
 
-	if (ERTS_SCHEDULER_IS_DIRTY(esdp))
-	    dirty_sched_wall_time_change(esdp, working = 1);
-        else if (!thr_prgr_active) {
-	    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-	    sched_wall_time_change(esdp, 1);
+        if (!(flgs & ERTS_SSI_FLG_WAITING)) {
+            ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
+            break;
         }
 
-	erts_smp_runq_lock(rq);
-	sched_active(esdp->no, rq);
-
     }
-    else
-#endif
-    {
-
-	erts_smp_atomic32_set_relb(&function_calls, 0);
-	*fcalls = 0;
-
-#ifdef ERTS_DIRTY_SCHEDULERS
-	ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
-#endif
-	sched_waiting_sys(esdp->no, rq);
-
-	erts_smp_runq_unlock(rq);
-
-	ASSERT(working);
-	sched_wall_time_change(esdp, working = 0);
-
-	spincount = sched_busy_wait.sys_schedule;
-	if (spincount == 0)
-	    goto sys_aux_work;
-
-	while (spincount-- > 0) {
-	    ErtsMonotonicTime current_time;
-
-	sys_poll_aux_work:
-
-	    if (working)
-		sched_wall_time_change(esdp, working = 0);
-
-	    ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_CHECK_IO);
-
-	    ASSERT(!erts_port_task_have_outstanding_io_tasks());
-            LTTNG2(scheduler_poll, esdp->no, 1);
-	    erl_sys_schedule(1); /* Might give us something to do */
-
-	    ERTS_MSACC_POP_STATE_M();
-
-	    if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-		current_time = erts_get_monotonic_time(esdp);
-		if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref))
-		    erts_bump_timers(esdp->timer_wheel, current_time);
-	    }
-
-	sys_aux_work:
-#ifndef ERTS_SMP
-	    erts_sys_schedule_interrupt(0);
-#endif
-
-	    aux_work = erts_atomic32_read_acqb(&ssi->aux_work);
-	    if (aux_work && !ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-		if (!working)
-		    sched_wall_time_change(esdp, working = 1);
-#ifdef ERTS_SMP
-		if (!thr_prgr_active)
-		    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-#endif
-		aux_work = handle_aux_work(&esdp->aux_work_data, aux_work, 1);
-                ERTS_MSACC_UPDATE_CACHE();
-#ifdef ERTS_SMP
-		if (aux_work && erts_thr_progress_update(esdp))
-		    erts_thr_progress_leader_update(esdp);
-#endif
-	    }
-
-#ifndef ERTS_SMP
-	    if (erts_smp_atomic32_read_dirty(&rq->len) != 0 || rq->misc.start)
-		goto sys_woken;
-#else
-	    flgs = erts_smp_atomic32_read_acqb(&ssi->flags);
-	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
-		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
-		goto sys_woken;
-	    }
-
-	    /*
-	     * If we got new I/O tasks we aren't allowed to
-	     * call erl_sys_schedule() until it is handled.
-	     */
-	    if (erts_port_task_have_outstanding_io_tasks()) {
-		clear_sys_scheduling();
-		/*
-		 * Got to check that we still got I/O tasks; otherwise
-		 * we have to continue checking for I/O...
-		 */
-		if (!prepare_for_sys_schedule(0)) {
-		    spincount *= ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT;
-		    goto tse_wait;
-		}
-	    }
-#endif
-	}
-
-	erts_smp_runq_lock(rq);
-
-#ifdef ERTS_SMP
-	/*
-	 * If we got new I/O tasks we aren't allowed to
-	 * sleep in erl_sys_schedule().
-	 */
-	if (erts_port_task_have_outstanding_io_tasks()) {
-	    clear_sys_scheduling();
 
-	    /*
-	     * Got to check that we still got I/O tasks; otherwise
-	     * we have to wait in erl_sys_schedule() after all...
-	     */
-	    if (!prepare_for_sys_schedule(0)) {
-		/*
-		 * Not allowed to wait in erl_sys_schedule;
-		 * do tse wait instead...
-		 */
-		sched_change_waiting_sys_to_waiting(esdp->no, rq);
-		erts_smp_runq_unlock(rq);
-		spincount = 0;
-		goto tse_wait;
-	    }
-	}
-#endif
-	if (aux_work) {
-	    erts_smp_runq_unlock(rq);
-	    goto sys_poll_aux_work;
-	}
-#ifdef ERTS_SMP
-	flgs = sched_set_sleeptype(ssi, ERTS_SSI_FLG_POLL_SLEEPING);
-	if (!(flgs & ERTS_SSI_FLG_SLEEPING)) {
-	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
-		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
-		goto sys_locked_woken;
-	    }
-	    erts_smp_runq_unlock(rq);
-	    flgs = sched_prep_cont_spin_wait(ssi);
-	    if (!(flgs & ERTS_SSI_FLG_WAITING)) {
-		ASSERT(!(flgs & ERTS_SSI_FLG_SLEEPING));
-		goto sys_woken;
-	    }
-	    ASSERT(!erts_port_task_have_outstanding_io_tasks());
-	    goto sys_poll_aux_work;
-	}
-
-	ASSERT(flgs & ERTS_SSI_FLG_POLL_SLEEPING);
-	ASSERT(flgs & ERTS_SSI_FLG_WAITING);
-#endif
-
-	erts_smp_runq_unlock(rq);
-
-	if (working)
-	    sched_wall_time_change(esdp, working = 0);
-
-#ifdef ERTS_SMP
-	if (thr_prgr_active)
-	    erts_thr_progress_active(esdp, thr_prgr_active = 0);
-#endif
-
-	ASSERT(!erts_port_task_have_outstanding_io_tasks());
+    if (flgs & ~(ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC))
+        erts_atomic32_read_band_nob(&ssi->flags,
+                                        (ERTS_SSI_FLG_SUSPENDED
+                                         | ERTS_SSI_FLG_MSB_EXEC));
 
-	ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_CHECK_IO);
-        LTTNG2(scheduler_poll, esdp->no, 0);
-
-	erl_sys_schedule(0);
-
-	ERTS_MSACC_POP_STATE();
-
-	if (!ERTS_SCHEDULER_IS_DIRTY(esdp)) {
-	    ErtsMonotonicTime current_time = erts_get_monotonic_time(esdp);
-	    if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref))
-		erts_bump_timers(esdp->timer_wheel, current_time);
-	}
-
-#ifndef ERTS_SMP
-	if (erts_smp_atomic32_read_dirty(&rq->len) == 0 && !rq->misc.start)
-	    goto sys_aux_work;
-    sys_woken:
-#else
-	flgs = sched_prep_cont_spin_wait(ssi);
-	if (flgs & ERTS_SSI_FLG_WAITING)
-	    goto sys_aux_work;
-
-    sys_woken:
-	if (!thr_prgr_active)
-	    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-	erts_smp_runq_lock(rq);
-    sys_locked_woken:
-	if (!thr_prgr_active) {
-	    erts_smp_runq_unlock(rq);
-	    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-	    erts_smp_runq_lock(rq);
-	}
-	clear_sys_scheduling();
-	if (flgs & ~(ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC))
-	    erts_smp_atomic32_read_band_nob(&ssi->flags,
-                                            (ERTS_SSI_FLG_SUSPENDED
-                                             | ERTS_SSI_FLG_MSB_EXEC));
-#endif
-	if (!working)
-	    sched_wall_time_change(esdp, working = 1);
-	sched_active_sys(esdp->no, rq);
+    if (ERTS_SCHEDULER_IS_DIRTY(esdp))
+        dirty_sched_wall_time_change(esdp, working = 1);
+    else if (!thr_prgr_active) {
+        erts_thr_progress_active(esdp, thr_prgr_active = 1);
+        sched_wall_time_change(esdp, 1);
     }
 
+    erts_runq_lock(rq);
+    sched_active(esdp->no, rq);
+
     if (ERTS_SCHEDULER_IS_DIRTY(esdp))
         dirty_active(esdp, 1);
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 }
 
-#ifdef ERTS_SMP
 
 static ERTS_INLINE erts_aint32_t
 ssi_flags_set_wake(ErtsSchedulerSleepInfo *ssi)
@@ -3816,7 +3414,7 @@ ssi_flags_set_wake(ErtsSchedulerSleepInfo *ssi)
     erts_aint32_t nflgs = 0;
     erts_aint32_t xflgs = ERTS_SSI_FLG_SLEEPING|ERTS_SSI_FLG_WAITING;
     while (1) {
-	oflgs = erts_smp_atomic32_cmpxchg_relb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_relb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return oflgs;
 	nflgs = oflgs & (ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC);
@@ -3830,13 +3428,12 @@ ssi_wake(ErtsSchedulerSleepInfo *ssi)
     erts_sched_finish_poke(ssi, ssi_flags_set_wake(ssi));
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static void
 dcpu_sched_ix_suspend_wake(Uint ix)
 {
     ErtsSchedulerSleepInfo* ssi = ERTS_DIRTY_CPU_SCHED_SLEEP_INFO_IX(ix);
-    erts_smp_atomic32_read_bor_nob(&ssi->flags, ERTS_SSI_FLG_SUSPENDED);
+    erts_atomic32_read_bor_nob(&ssi->flags, ERTS_SSI_FLG_SUSPENDED);
     ssi_wake(ssi);
 }
 
@@ -3844,7 +3441,7 @@ static void
 dio_sched_ix_suspend_wake(Uint ix)
 {
     ErtsSchedulerSleepInfo* ssi = ERTS_DIRTY_IO_SCHED_SLEEP_INFO_IX(ix);
-    erts_smp_atomic32_read_bor_nob(&ssi->flags, ERTS_SSI_FLG_SUSPENDED);
+    erts_atomic32_read_bor_nob(&ssi->flags, ERTS_SSI_FLG_SUSPENDED);
     ssi_wake(ssi);
 }
 
@@ -3862,7 +3459,6 @@ dio_sched_ix_wake(Uint ix)
 }
 #endif
 
-#endif
 
 static void
 wake_scheduler(ErtsRunQueue *rq)
@@ -3875,13 +3471,12 @@ wake_scheduler(ErtsRunQueue *rq)
      * so all code *should* handle this without having
      * the lock on the run queue.
      */
-    ERTS_SMP_LC_ASSERT(!erts_smp_lc_runq_is_locked(rq)
+    ERTS_LC_ASSERT(!erts_lc_runq_is_locked(rq)
 		       || ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
 
     ssi_wake(rq->scheduler->ssi);
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 static void
 wake_dirty_schedulers(ErtsRunQueue *rq, int one)
 {
@@ -3891,10 +3486,10 @@ wake_dirty_schedulers(ErtsRunQueue *rq, int one)
     ASSERT(ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
 
     sl = &rq->sleepers;
-    erts_smp_spin_lock(&sl->lock);
+    erts_spin_lock(&sl->lock);
     ssi = sl->list;
     if (!ssi) {
-	erts_smp_spin_unlock(&sl->lock);
+	erts_spin_unlock(&sl->lock);
 	if (one)
 	    wake_scheduler(rq);
     } else if (one) {
@@ -3908,14 +3503,14 @@ wake_dirty_schedulers(ErtsRunQueue *rq, int one)
 	if (ssi->next)
 	    ssi->next->prev = ssi->prev;
 
-	erts_smp_spin_unlock(&sl->lock);
+	erts_spin_unlock(&sl->lock);
 
 	ERTS_THR_MEMORY_BARRIER;
 	flgs = ssi_flags_set_wake(ssi);
 	erts_sched_finish_poke(ssi, flgs);
     } else {
 	sl->list = NULL;
-	erts_smp_spin_unlock(&sl->lock);
+	erts_spin_unlock(&sl->lock);
 
 	ERTS_THR_MEMORY_BARRIER;
 	do {
@@ -3932,7 +3527,6 @@ wake_dirty_scheduler(ErtsRunQueue *rq)
     wake_dirty_schedulers(rq, 1);
 }
 
-#endif
 
 #define ERTS_NO_USED_RUNQS_SHIFT 16
 #define ERTS_NO_RUNQS_MASK 0xffffU
@@ -3946,13 +3540,13 @@ init_no_runqs(int active, int used)
 {
     erts_aint32_t no_runqs = (erts_aint32_t) (active & ERTS_NO_RUNQS_MASK);
     no_runqs |= (erts_aint32_t) ((used & ERTS_NO_RUNQS_MASK) << ERTS_NO_USED_RUNQS_SHIFT);
-    erts_smp_atomic32_init_nob(&balance_info.no_runqs, no_runqs);
+    erts_atomic32_init_nob(&balance_info.no_runqs, no_runqs);
 }
 
 static ERTS_INLINE void
 get_no_runqs(int *active, int *used)
 {
-    erts_aint32_t no_runqs = erts_smp_atomic32_read_nob(&balance_info.no_runqs);
+    erts_aint32_t no_runqs = erts_atomic32_read_nob(&balance_info.no_runqs);
     if (active)
 	*active = (int) (no_runqs & ERTS_NO_RUNQS_MASK);
     if (used)
@@ -3962,12 +3556,12 @@ get_no_runqs(int *active, int *used)
 static ERTS_INLINE void
 set_no_used_runqs(int used)
 {
-    erts_aint32_t exp = erts_smp_atomic32_read_nob(&balance_info.no_runqs);
+    erts_aint32_t exp = erts_atomic32_read_nob(&balance_info.no_runqs);
     while (1) {
 	erts_aint32_t act, new;
 	new = (used & ERTS_NO_RUNQS_MASK) << ERTS_NO_USED_RUNQS_SHIFT;
 	new |= exp & ERTS_NO_RUNQS_MASK;
-	act = erts_smp_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
+	act = erts_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
 	if (act == exp)
 	    break;
 	exp = act;
@@ -3977,14 +3571,14 @@ set_no_used_runqs(int used)
 static ERTS_INLINE void
 set_no_active_runqs(int active)
 {
-    erts_aint32_t exp = erts_smp_atomic32_read_nob(&balance_info.no_runqs);
+    erts_aint32_t exp = erts_atomic32_read_nob(&balance_info.no_runqs);
     while (1) {
 	erts_aint32_t act, new;
 	if ((exp & ERTS_NO_RUNQS_MASK) == active)
 	    break;
 	new = exp & (ERTS_NO_RUNQS_MASK << ERTS_NO_USED_RUNQS_SHIFT);
 	new |= active & ERTS_NO_RUNQS_MASK;
-	act = erts_smp_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
+	act = erts_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
 	if (act == exp)
 	    break;
 	exp = act;
@@ -3994,14 +3588,14 @@ set_no_active_runqs(int active)
 static ERTS_INLINE int
 try_inc_no_active_runqs(int active)
 {
-    erts_aint32_t exp = erts_smp_atomic32_read_nob(&balance_info.no_runqs);
+    erts_aint32_t exp = erts_atomic32_read_nob(&balance_info.no_runqs);
     if (((exp >> ERTS_NO_USED_RUNQS_SHIFT) & ERTS_NO_RUNQS_MASK) < active)
 	return 0;
     if ((exp & ERTS_NO_RUNQS_MASK) + 1 == active) {
 	erts_aint32_t new, act;
 	new = exp & (ERTS_NO_RUNQS_MASK << ERTS_NO_USED_RUNQS_SHIFT);
 	new |= active & ERTS_NO_RUNQS_MASK;
-	act = erts_smp_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
+	act = erts_atomic32_cmpxchg_nob(&balance_info.no_runqs, new, exp);
 	if (act == exp)
 	    return 1;
     }
@@ -4063,25 +3657,20 @@ wake_scheduler_on_empty_runq(ErtsRunQueue *crq)
     }
 }
 
-#endif /* ERTS_SMP */
 
 static ERTS_INLINE void
 smp_notify_inc_runq(ErtsRunQueue *runq)
 {
-#ifdef ERTS_SMP
     if (runq) {
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (ERTS_RUNQ_IX_IS_DIRTY(runq->ix))
 	    wake_dirty_scheduler(runq);
 	else
-#endif
 	    wake_scheduler(runq);
     }
-#endif
 }
 
 void
-erts_smp_notify_inc_runq(ErtsRunQueue *runq)
+erts_notify_inc_runq(ErtsRunQueue *runq)
 {
     smp_notify_inc_runq(runq);
 }
@@ -4089,16 +3678,12 @@ erts_smp_notify_inc_runq(ErtsRunQueue *runq)
 void
 erts_sched_notify_check_cpu_bind(void)
 {
-#ifdef ERTS_SMP
     int ix;
     for (ix = 0; ix < erts_no_run_queues; ix++) {
 	ErtsRunQueue *rq = ERTS_RUNQ_IX(ix);
 	(void) ERTS_RUNQ_FLGS_SET(rq, ERTS_RUNQ_FLG_CHK_CPU_BIND);
 	wake_scheduler(rq);
     }
-#else
-    erts_sched_check_cpu_bind(erts_get_scheduler_data());
-#endif
 }
 
 
@@ -4107,9 +3692,9 @@ enqueue_process(ErtsRunQueue *runq, int prio, Process *p)
 {
     ErtsRunPrioQueue *rpq;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
 
-    erts_smp_inc_runq_len(runq, &runq->procs.prio_info[prio], prio);
+    erts_inc_runq_len(runq, &runq->procs.prio_info[prio], prio);
 
     if (prio == PRIORITY_LOW) {
 	p->schedule_count = RESCHEDULE_LOW;
@@ -4137,7 +3722,7 @@ unqueue_process(ErtsRunQueue *runq,
 		Process *prev_proc,
 		Process *proc)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
 
     if (prev_proc)
 	prev_proc->next = proc->next;
@@ -4149,7 +3734,7 @@ unqueue_process(ErtsRunQueue *runq,
     if (!rpq->first)
 	rpq->last = NULL;
 
-    erts_smp_dec_runq_len(runq, rqi, prio);
+    erts_dec_runq_len(runq, rqi, prio);
 }
 
 
@@ -4162,7 +3747,7 @@ dequeue_process(ErtsRunQueue *runq, int prio_q, erts_aint32_t *statep)
     ErtsRunQueueInfo *rqi;
     Process *p;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(runq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(runq));
 
     ASSERT(PRIORITY_NORMAL == prio_q
 	   || PRIORITY_HIGH == prio_q
@@ -4173,9 +3758,9 @@ dequeue_process(ErtsRunQueue *runq, int prio_q, erts_aint32_t *statep)
     if (!p)
 	return NULL;
 
-    ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
+    ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
 
-    state = erts_smp_atomic32_read_nob(&p->state);
+    state = erts_atomic32_read_nob(&p->state);
     if (statep)
 	*statep = state;
 
@@ -4208,11 +3793,10 @@ check_requeue_process(ErtsRunQueue *rq, int prio_q)
 static ERTS_INLINE void
 free_proxy_proc(Process *proxy)
 {
-    ASSERT(erts_smp_atomic32_read_nob(&proxy->state) & ERTS_PSFLG_PROXY);
+    ASSERT(erts_atomic32_read_nob(&proxy->state) & ERTS_PSFLG_PROXY);
     erts_free(ERTS_ALC_T_PROC, proxy);
 }
 
-#ifdef ERTS_SMP
 
 static ErtsRunQueue *
 check_immigration_need(ErtsRunQueue *c_rq, ErtsMigrationPath *mp, int prio)
@@ -4265,7 +3849,7 @@ static void
 immigrate(ErtsRunQueue *c_rq, ErtsMigrationPath *mp)
 {
     Uint32 iflags, iflag;
-    erts_smp_runq_unlock(c_rq);
+    erts_runq_unlock(c_rq);
 
     ASSERT(erts_thr_progress_is_managed_thread());
 
@@ -4306,26 +3890,21 @@ immigrate(ErtsRunQueue *c_rq, ErtsMigrationPath *mp)
 
 	rq = check_immigration_need(c_rq, mp, prio);
 	if (rq) {
-	    erts_smp_runq_lock(rq);
+	    erts_runq_lock(rq);
 	    if (prio == ERTS_PORT_PRIO_LEVEL) {
 		Port *prt;
 		prt = erts_dequeue_port(rq);
 		if (prt)
-		    RUNQ_SET_RQ(&prt->run_queue, c_rq);
-		erts_smp_runq_unlock(rq);
+                    erts_set_runq_port(prt, c_rq);
+		erts_runq_unlock(rq);
 		if (prt) {
-		    /* port might terminate while we have no lock... */
 		    rq = erts_port_runq(prt);
-		    if (rq) {
-			if (rq != c_rq)
-			    erts_exit(ERTS_ABORT_EXIT,
-				     "%s:%d:%s(): Internal error",
-				     __FILE__, __LINE__, __func__);
-			erts_enqueue_port(c_rq, prt);
-			if (!iflag)
-			    return; /* done */
-			erts_smp_runq_unlock(c_rq);
-		    }
+                    if (rq != c_rq)
+                        ERTS_INTERNAL_ERROR("Unexpected run-queue");
+                    erts_enqueue_port(c_rq, prt);
+                    if (!iflag)
+                        return; /* done */
+                    erts_runq_unlock(c_rq);
 		}
 	    }
 	    else {
@@ -4338,38 +3917,37 @@ immigrate(ErtsRunQueue *c_rq, ErtsMigrationPath *mp)
 
 		while (proc) {
 		    erts_aint32_t state;
-		    state = erts_smp_atomic32_read_acqb(&proc->state);
-		    if (!(ERTS_PSFLG_BOUND & state)
-			&& (prio == (int) ERTS_PSFLGS_GET_PRQ_PRIO(state))) {
+		    state = erts_atomic32_read_acqb(&proc->state);
+		    if (prio == (int) ERTS_PSFLGS_GET_PRQ_PRIO(state)
+                        && erts_try_change_runq_proc(proc, c_rq)) {
 			ErtsRunQueueInfo *rqi = &rq->procs.prio_info[prio];
 			unqueue_process(rq, rpq, rqi, prio, prev_proc, proc);
-			erts_smp_runq_unlock(rq);
-			RUNQ_SET_RQ(&proc->run_queue, c_rq);
+			erts_runq_unlock(rq);
 			rq_locked = 0;
 
-			erts_smp_runq_lock(c_rq);
+			erts_runq_lock(c_rq);
 			enqueue_process(c_rq, prio, proc);
 			if (!iflag)
 			    return; /* done */
-			erts_smp_runq_unlock(c_rq);
+			erts_runq_unlock(c_rq);
 			break;
 		    }
 		    prev_proc = proc;
 		    proc = proc->next;
 		}
 		if (rq_locked)
-		    erts_smp_runq_unlock(rq);
+		    erts_runq_unlock(rq);
 	    }
 	}
     }
 
-    erts_smp_runq_lock(c_rq);
+    erts_runq_lock(c_rq);
 }
 
 static ERTS_INLINE void
 suspend_run_queue(ErtsRunQueue *rq)
 {
-    erts_smp_atomic32_read_bor_nob(&rq->scheduler->ssi->flags,
+    erts_atomic32_read_bor_nob(&rq->scheduler->ssi->flags,
 				   ERTS_SSI_FLG_SUSPENDED);
     (void) ERTS_RUNQ_FLGS_SET(rq, ERTS_RUNQ_FLG_SUSPENDED);
 
@@ -4386,7 +3964,7 @@ resume_run_queue(ErtsRunQueue *rq)
 
     ASSERT(!ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
 
-    erts_smp_runq_lock(rq);
+    erts_runq_lock(rq);
 
     oflgs = ERTS_RUNQ_FLGS_READ_BSET(rq,
 				     (ERTS_RUNQ_FLG_OUT_OF_WORK
@@ -4401,19 +3979,19 @@ resume_run_queue(ErtsRunQueue *rq)
 
 	rq->check_balance_reds = ERTS_RUNQ_CALL_CHECK_BALANCE_REDS;
 	for (pix = 0; pix < ERTS_NO_PROC_PRIO_LEVELS; pix++) {
-	    len = erts_smp_atomic32_read_dirty(&rq->procs.prio_info[pix].len);
+	    len = erts_atomic32_read_dirty(&rq->procs.prio_info[pix].len);
 	    rq->procs.prio_info[pix].max_len = len;
 	    rq->procs.prio_info[pix].reds = 0;
 	}
-	len = erts_smp_atomic32_read_dirty(&rq->ports.info.len);
+	len = erts_atomic32_read_dirty(&rq->ports.info.len);
 	rq->ports.info.max_len = len;
 	rq->ports.info.reds = 0;
-	len = erts_smp_atomic32_read_dirty(&rq->len);
+	len = erts_atomic32_read_dirty(&rq->len);
 	rq->max_len = len;
 
     }
 
-    erts_smp_runq_unlock(rq);
+    erts_runq_unlock(rq);
 
     nrml_sched_ix_resume_wake(rq->ix);
 }
@@ -4428,18 +4006,17 @@ schedule_bound_processes(ErtsRunQueue *rq,
 			 ErtsStuckBoundProcesses *sbpp)
 {
     Process *proc, *next;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
     proc = sbpp->first;
     while (proc) {
-	erts_aint32_t state = erts_smp_atomic32_read_acqb(&proc->state);
+	erts_aint32_t state = erts_atomic32_read_acqb(&proc->state);
 	next = proc->next;
 	enqueue_process(rq, (int) ERTS_PSFLGS_GET_PRQ_PRIO(state), proc);
 	proc = next;
     }
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static ERTS_INLINE void
 clear_proc_dirty_queue_bit(Process *p, ErtsRunQueue *rq, int prio_bit)
@@ -4459,11 +4036,10 @@ clear_proc_dirty_queue_bit(Process *p, ErtsRunQueue *rq, int prio_bit)
 #else
 	(void)
 #endif
-	erts_smp_atomic32_read_band_mb(&p->dirty_state, ~qb);
+	erts_atomic32_read_band_mb(&p->dirty_state, ~qb);
     ASSERT(old & qb);
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 
 static void
@@ -4475,7 +4051,7 @@ evacuate_run_queue(ErtsRunQueue *rq,
     ErtsMigrationPaths *mps;
     ErtsMigrationPath *mp;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
     (void) ERTS_RUNQ_FLGS_UNSET(rq, ERTS_RUNQ_FLG_PROTECTED);
 
@@ -4498,9 +4074,9 @@ evacuate_run_queue(ErtsRunQueue *rq,
 	rq->misc.start = NULL;
 	rq->misc.end = NULL;
         ERTS_RUNQ_FLGS_UNSET_NOB(rq, ERTS_RUNQ_FLG_MISC_OP);
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
 
-	erts_smp_runq_lock(to_rq);
+	erts_runq_lock(to_rq);
 	if (to_rq->misc.end)
 	    to_rq->misc.end->next = start;
 	else
@@ -4510,9 +4086,9 @@ evacuate_run_queue(ErtsRunQueue *rq,
 
 	non_empty_runq(to_rq);
 
-	erts_smp_runq_unlock(to_rq);
+	erts_runq_unlock(to_rq);
 	smp_notify_inc_runq(to_rq);
-	erts_smp_runq_lock(to_rq);
+	erts_runq_lock(to_rq);
     }
 
     if (rq->ports.start) {
@@ -4527,22 +4103,14 @@ evacuate_run_queue(ErtsRunQueue *rq,
 	while (prt) {
 	    ErtsRunQueue *prt_rq;
 	    prt = erts_dequeue_port(rq);
-	    RUNQ_SET_RQ(&prt->run_queue, to_rq);
-	    erts_smp_runq_unlock(rq);
-	    /*
-	     * The port might terminate while
-	     * we have no lock on it...
-	     */
+            erts_set_runq_port(prt, to_rq);
+	    erts_runq_unlock(rq);
 	    prt_rq = erts_port_runq(prt);
-	    if (prt_rq) {
-		if (prt_rq != to_rq)
-		    erts_exit(ERTS_ABORT_EXIT,
-			     "%s:%d:%s() internal error\n",
-			     __FILE__, __LINE__, __func__);
-		erts_enqueue_port(to_rq, prt);
-		erts_smp_runq_unlock(to_rq);
-	    }
-	    erts_smp_runq_lock(rq);
+            if (prt_rq != to_rq)
+                ERTS_INTERNAL_ERROR("Unexpected run-queue");
+            erts_enqueue_port(to_rq, prt);
+            erts_runq_unlock(to_rq);
+	    erts_runq_lock(rq);
 	    prt = rq->ports.start;
 	}
 	smp_notify_inc_runq(to_rq);
@@ -4552,8 +4120,6 @@ evacuate_run_queue(ErtsRunQueue *rq,
     for (prio_q = 0; prio_q < ERTS_NO_PROC_PRIO_QUEUES; prio_q++) {
 	erts_aint32_t state;
 	Process *proc;
-	int notify = 0;
-	to_rq = NULL;
 
         if (!mp->prio[prio_q].runq)
             return;
@@ -4564,14 +4130,13 @@ evacuate_run_queue(ErtsRunQueue *rq,
 	while (proc) {
 	    Process *real_proc;
 	    int prio;
-	    erts_aint32_t max_qbit, qbit, real_state;
+	    erts_aint32_t max_qbit, qbit;
 
 	    prio = ERTS_PSFLGS_GET_PRQ_PRIO(state);
 	    qbit = ((erts_aint32_t) 1) << prio;
 
 	    if (!(state & ERTS_PSFLG_PROXY)) {
 		real_proc = proc;
-		real_state = state;
 	    }
 	    else {
 		real_proc = erts_proc_lookup_raw(proc->common.id);
@@ -4579,7 +4144,6 @@ evacuate_run_queue(ErtsRunQueue *rq,
 		    free_proxy_proc(proc);
 		    goto handle_next_proc;
 		}
-		real_state =  erts_smp_atomic32_read_acqb(&real_proc->state);
 	    }
 
 	    max_qbit = (state >> ERTS_PSFLGS_IN_PRQ_MASK_OFFSET);
@@ -4605,7 +4169,7 @@ evacuate_run_queue(ErtsRunQueue *rq,
 #else
 			(void)
 #endif
-			erts_smp_atomic32_read_band_mb(&proc->state,
+			erts_atomic32_read_band_mb(&proc->state,
 						       ~clr_bits);
 			ASSERT((old & clr_bits) == clr_bits);
 
@@ -4614,7 +4178,13 @@ evacuate_run_queue(ErtsRunQueue *rq,
 		goto handle_next_proc;
 	    }
 
-	    if (ERTS_PSFLG_BOUND & real_state) {
+            prio = (int) ERTS_PSFLGS_GET_PRQ_PRIO(state);
+            to_rq = mp->prio[prio].runq;
+
+            if (!to_rq)
+                goto handle_next_proc;
+
+	    if (!erts_try_change_runq_proc(proc, to_rq)) {
 		/* Bound processes get stuck here... */
 		proc->next = NULL;
 		if (sbpp->last)
@@ -4624,25 +4194,21 @@ evacuate_run_queue(ErtsRunQueue *rq,
 		sbpp->last = proc;
 	    }
 	    else {
-		int prio = (int) ERTS_PSFLGS_GET_PRQ_PRIO(state);
-		erts_smp_runq_unlock(rq);
+		erts_runq_unlock(rq);
 
-                to_rq = mp->prio[prio].runq;
-		RUNQ_SET_RQ(&proc->run_queue, to_rq);
-
-		erts_smp_runq_lock(to_rq);
+		erts_runq_lock(to_rq);
 		enqueue_process(to_rq, prio, proc);
-		erts_smp_runq_unlock(to_rq);
-		notify = 1;
+		erts_runq_unlock(to_rq);
+
+                smp_notify_inc_runq(to_rq);
 
-		erts_smp_runq_lock(rq);
+		erts_runq_lock(rq);
 	    }
 
 	handle_next_proc:
 	    proc = dequeue_process(rq, prio_q, &state);
 	}
-	if (notify)
-	    smp_notify_inc_runq(to_rq);
+
     }
 }
 
@@ -4654,13 +4220,13 @@ try_steal_task_from_victim(ErtsRunQueue *rq, int *rq_lockedp, ErtsRunQueue *vrq,
     ErtsRunPrioQueue *rpq;
 
     if (*rq_lockedp) {
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
 	*rq_lockedp = 0;
     }
 
-    ERTS_SMP_LC_ASSERT(!erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(!erts_lc_runq_is_locked(rq));
 
-    erts_smp_runq_lock(vrq);
+    erts_runq_lock(vrq);
 
     if (ERTS_RUNQ_FLGS_GET_NOB(rq) & ERTS_RUNQ_FLG_HALTING)
 	goto no_procs;
@@ -4696,16 +4262,15 @@ try_steal_task_from_victim(ErtsRunQueue *rq, int *rq_lockedp, ErtsRunQueue *vrq,
 	proc = rpq->first;
 
 	while (proc) {
-	    erts_aint32_t state = erts_smp_atomic32_read_acqb(&proc->state);
-	    if (!(ERTS_PSFLG_BOUND & state)) {
+	    if (erts_try_change_runq_proc(proc, rq)) {
+                erts_aint32_t state = erts_atomic32_read_acqb(&proc->state);
 		/* Steal process */
 		int prio = (int) ERTS_PSFLGS_GET_PRQ_PRIO(state);
 		ErtsRunQueueInfo *rqi = &vrq->procs.prio_info[prio];
 		unqueue_process(vrq, rpq, rqi, prio, prev_proc, proc);
-		erts_smp_runq_unlock(vrq);
-		RUNQ_SET_RQ(&proc->run_queue, rq);
+		erts_runq_unlock(vrq);
 
-		erts_smp_runq_lock(rq);
+		erts_runq_lock(rq);
 		*rq_lockedp = 1;
 		enqueue_process(rq, prio, proc);
 		return !0;
@@ -4719,7 +4284,7 @@ try_steal_task_from_victim(ErtsRunQueue *rq, int *rq_lockedp, ErtsRunQueue *vrq,
 
 no_procs:
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(vrq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(vrq));
 
     /*
      * Check for a runnable port to steal...
@@ -4728,29 +4293,17 @@ no_procs:
     if (vrq->ports.start) {
 	ErtsRunQueue *prt_rq;
 	Port *prt = erts_dequeue_port(vrq);
-	RUNQ_SET_RQ(&prt->run_queue, rq);
-	erts_smp_runq_unlock(vrq);
-
-	/*
-	 * The port might terminate while
-	 * we have no lock on it...
-	 */
-
+	erts_set_runq_port(prt, rq);
+	erts_runq_unlock(vrq);
 	prt_rq = erts_port_runq(prt);
-	if (!prt_rq)
-	    return 0;
-	else {
-	    if (prt_rq != rq)
-		erts_exit(ERTS_ABORT_EXIT,
-			 "%s:%d:%s() internal error\n",
-			 __FILE__, __LINE__, __func__);
-	    *rq_lockedp = 1;
-	    erts_enqueue_port(rq, prt);
-	    return !0;
-	}
+        if (prt_rq != rq)
+            ERTS_INTERNAL_ERROR("Unexpected run-queue");
+        *rq_lockedp = 1;
+        erts_enqueue_port(rq, prt);
+        return !0;
     }
 
-    erts_smp_runq_unlock(vrq);
+    erts_runq_unlock(vrq);
 
     return 0;
 }
@@ -4782,7 +4335,7 @@ try_steal_task(ErtsRunQueue *rq)
     res = 0;
     rq_locked = 1;
 
-    ERTS_SMP_LC_CHK_RUNQ_LOCK(rq, rq_locked);
+    ERTS_LC_CHK_RUNQ_LOCK(rq, rq_locked);
 
     get_no_runqs(&active_rqs, &blnc_rqs);
 
@@ -4795,7 +4348,7 @@ try_steal_task(ErtsRunQueue *rq)
 	if (active_rqs < blnc_rqs) {
 	    int no = blnc_rqs - active_rqs;
 	    int stop_ix = vix = active_rqs + rq->ix % no;
-	    while (erts_smp_atomic32_read_acqb(&no_empty_run_queues) < blnc_rqs) {
+	    while (erts_atomic32_read_acqb(&no_empty_run_queues) < blnc_rqs) {
 		res = check_possible_steal_victim(rq, &rq_locked, vix);
 		if (res)
 		    goto done;
@@ -4810,7 +4363,7 @@ try_steal_task(ErtsRunQueue *rq)
 	vix = rq->ix;
 
 	/* ... then try to steal a job from another active queue... */
-	while (erts_smp_atomic32_read_acqb(&no_empty_run_queues) < blnc_rqs) {
+	while (erts_atomic32_read_acqb(&no_empty_run_queues) < blnc_rqs) {
 	    vix++;
 	    if (vix >= active_rqs)
 		vix = 0;
@@ -4827,7 +4380,7 @@ try_steal_task(ErtsRunQueue *rq)
  done:
 
     if (!rq_locked)
-	erts_smp_runq_lock(rq);
+	erts_runq_lock(rq);
 
     if (res)
         return res;
@@ -4953,7 +4506,7 @@ alloc_mpaths(void)
 {
     void *block;
     ErtsMigrationPaths *res;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&balance_info.update_mtx));
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&balance_info.update_mtx));
 
     res = mpaths.freelist;
     if (res) {
@@ -4976,7 +4529,7 @@ retire_mpaths(ErtsMigrationPaths *mps)
 {
     ErtsThrPrgrVal current;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&balance_info.update_mtx));
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&balance_info.update_mtx));
 
     current = erts_thr_progress_current();
 
@@ -5022,7 +4575,7 @@ check_balance(ErtsRunQueue *c_rq)
     int sched_util_balancing;
 #endif
 
-    if (erts_smp_atomic32_xchg_nob(&balance_info.checking_balance, 1)) {
+    if (erts_atomic32_xchg_nob(&balance_info.checking_balance, 1)) {
 	c_rq->check_balance_reds = INT_MAX;
 	return;
     }
@@ -5030,15 +4583,15 @@ check_balance(ErtsRunQueue *c_rq)
     get_no_runqs(NULL, &blnc_no_rqs);
     if (blnc_no_rqs == 1) {
 	c_rq->check_balance_reds = INT_MAX;
-	erts_smp_atomic32_set_nob(&balance_info.checking_balance, 0);
+	erts_atomic32_set_nob(&balance_info.checking_balance, 0);
 	return;
     }
 
-    erts_smp_runq_unlock(c_rq);
+    erts_runq_unlock(c_rq);
 
     if (balance_info.halftime) {	
 	balance_info.halftime = 0;
-	erts_smp_atomic32_set_nob(&balance_info.checking_balance, 0);
+	erts_atomic32_set_nob(&balance_info.checking_balance, 0);
 	ERTS_FOREACH_RUNQ(rq,
 	{
 	    if (rq->waiting)
@@ -5048,7 +4601,7 @@ check_balance(ErtsRunQueue *c_rq)
 	    rq->check_balance_reds = ERTS_RUNQ_CALL_CHECK_BALANCE_REDS;
 	});
 
-	erts_smp_runq_lock(c_rq);
+	erts_runq_lock(c_rq);
 	return;
     }
 
@@ -5061,7 +4614,7 @@ check_balance(ErtsRunQueue *c_rq)
      * is manipulated. Such updates of the migration information
      * might clash with balancing.
      */
-    erts_smp_mtx_lock(&balance_info.update_mtx);
+    erts_mtx_lock(&balance_info.update_mtx);
 
     forced = balance_info.forced_check_balance;
     balance_info.forced_check_balance = 0;
@@ -5069,10 +4622,10 @@ check_balance(ErtsRunQueue *c_rq)
     get_no_runqs(&current_active, &blnc_no_rqs);
 
     if (blnc_no_rqs == 1) {
-	erts_smp_mtx_unlock(&balance_info.update_mtx);
-	erts_smp_runq_lock(c_rq);
+	erts_mtx_unlock(&balance_info.update_mtx);
+	erts_runq_lock(c_rq);
 	c_rq->check_balance_reds = INT_MAX;
-	erts_smp_atomic32_set_nob(&balance_info.checking_balance, 0);
+	erts_atomic32_set_nob(&balance_info.checking_balance, 0);
 	return;
     }
 
@@ -5088,7 +4641,7 @@ check_balance(ErtsRunQueue *c_rq)
     /* Read balance information for all run queues */
     for (qix = 0; qix < blnc_no_rqs; qix++) {
 	ErtsRunQueue *rq = ERTS_RUNQ_IX(qix);
-	erts_smp_runq_lock(rq);
+	erts_runq_lock(rq);
 
 	run_queue_info[qix].flags = ERTS_RUNQ_FLGS_GET_NOB(rq);
 	for (pix = 0; pix < ERTS_NO_PROC_PRIO_LEVELS; pix++) {
@@ -5116,7 +4669,7 @@ check_balance(ErtsRunQueue *c_rq)
 	    run_queue_info[qix].sched_util = erts_get_sched_util(rq, 1, 0);
 #endif
 
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
     }
 
     full_scheds = 0;
@@ -5555,7 +5108,7 @@ erts_fprintf(stderr, "--------------------------------\n");
 	Uint32 flags = run_queue_info[qix].flags;
 	ErtsRunQueue *rq = ERTS_RUNQ_IX(qix);
 
-	erts_smp_runq_lock(rq);
+	erts_runq_lock(rq);
 	ASSERT(!(flags & ERTS_RUNQ_FLG_OUT_OF_WORK));
 	if (rq->waiting)
 	    flags |= ERTS_RUNQ_FLG_OUT_OF_WORK;
@@ -5570,27 +5123,27 @@ erts_fprintf(stderr, "--------------------------------\n");
 	rq->out_of_work_count = 0;
 	(void) ERTS_RUNQ_FLGS_READ_BSET(rq, ERTS_RUNQ_FLGS_MIGRATION_INFO, flags);
 
-	rq->max_len = erts_smp_atomic32_read_dirty(&rq->len);
+	rq->max_len = erts_atomic32_read_dirty(&rq->len);
 	for (pix = 0; pix < ERTS_NO_PRIO_LEVELS; pix++) {
 	    ErtsRunQueueInfo *rqi;
 	    rqi = (pix == ERTS_PORT_PRIO_LEVEL
 		   ? &rq->ports.info
 		   : &rq->procs.prio_info[pix]);
-	    erts_smp_reset_max_len(rq, rqi);
+	    erts_reset_max_len(rq, rqi);
 	    rqi->reds = 0;
 	}
 
 	rq->check_balance_reds = ERTS_RUNQ_CALL_CHECK_BALANCE_REDS;
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
     }
 
-    erts_smp_atomic32_set_nob(&balance_info.checking_balance, 0);
+    erts_atomic32_set_nob(&balance_info.checking_balance, 0);
 
     balance_info.n++;
     retire_mpaths(old_mpaths);
-    erts_smp_mtx_unlock(&balance_info.update_mtx);
+    erts_mtx_unlock(&balance_info.update_mtx);
 
-    erts_smp_runq_lock(c_rq);
+    erts_runq_lock(c_rq);
 }
 
 static void
@@ -5598,7 +5151,7 @@ change_no_used_runqs(int used)
 {
     ErtsMigrationPaths *new_mpaths, *old_mpaths;
     int qix;
-    erts_smp_mtx_lock(&balance_info.update_mtx);
+    erts_mtx_lock(&balance_info.update_mtx);
     set_no_used_runqs(used);
 
     old_mpaths = erts_get_migration_paths_managed();
@@ -5645,28 +5198,23 @@ change_no_used_runqs(int used)
     /* Make sure that we balance soon... */
     balance_info.forced_check_balance = 1;
 
-    erts_smp_mtx_unlock(&balance_info.update_mtx);
+    erts_mtx_unlock(&balance_info.update_mtx);
 
-    erts_smp_runq_lock(ERTS_RUNQ_IX(0));
+    erts_runq_lock(ERTS_RUNQ_IX(0));
     ERTS_RUNQ_IX(0)->check_balance_reds = 0;
-    erts_smp_runq_unlock(ERTS_RUNQ_IX(0));
+    erts_runq_unlock(ERTS_RUNQ_IX(0));
 }
 
 
-#endif /* #ifdef ERTS_SMP */
 
 Uint
 erts_debug_nbalance(void)
 {
-#ifdef ERTS_SMP
     Uint n;
-    erts_smp_mtx_lock(&balance_info.update_mtx);
+    erts_mtx_lock(&balance_info.update_mtx);
     n = balance_info.n;
-    erts_smp_mtx_unlock(&balance_info.update_mtx);
+    erts_mtx_unlock(&balance_info.update_mtx);
     return n;
-#else
-    return 0;
-#endif
 }
 
 /* Wakeup other schedulers */
@@ -5712,26 +5260,36 @@ typedef enum {
 #define ERTS_WAKEUP_OTHER_DEC_LEGACY 10
 #define ERTS_WAKEUP_OTHER_FIXED_INC_LEGACY (CONTEXT_REDS/10)
 
-#ifdef ERTS_SMP
 
-static struct {
+typedef struct {
     ErtsSchedWakeupOtherThreshold threshold;
     ErtsSchedWakeupOtherType type;
     int limit;
     int dec_shift;
     int dec_mask;
     void (*check)(ErtsRunQueue *rq, Uint32 flags);
-} wakeup_other;
+} ErtsWakeupOtherParams;
+
+static ErtsWakeupOtherParams sched_wakeup_other_params[ERTS_SCHED_TYPE_LAST + 1];
+
+static ERTS_INLINE ErtsWakeupOtherParams *
+runq_get_wakeup_other_params(ErtsRunQueue *rq)
+{
+    ErtsSchedulerData *esdp = rq->scheduler;
+    return &sched_wakeup_other_params[esdp->type];
+}
 
 static void
 wakeup_other_check(ErtsRunQueue *rq, Uint32 flags)
 {
+    ErtsWakeupOtherParams *wo_params = runq_get_wakeup_other_params(rq);
     int wo_reds = rq->wakeup_other_reds;
+
     if (wo_reds) {
-	int left_len = erts_smp_atomic32_read_dirty(&rq->len) - 1;
+	int left_len = erts_atomic32_read_dirty(&rq->len) - 1;
 	if (left_len < 1) {
-	    int wo_reduce = wo_reds << wakeup_other.dec_shift;
-	    wo_reduce &= wakeup_other.dec_mask;
+	    int wo_reduce = wo_reds << wo_params->dec_shift;
+	    wo_reduce &= wo_params->dec_mask;
 	    rq->wakeup_other -= wo_reduce;
 	    if (rq->wakeup_other < 0)
 		rq->wakeup_other = 0;
@@ -5739,17 +5297,15 @@ wakeup_other_check(ErtsRunQueue *rq, Uint32 flags)
 	else {
 	    rq->wakeup_other += (left_len*wo_reds
 				 + ERTS_WAKEUP_OTHER_FIXED_INC);
-	    if (rq->wakeup_other > wakeup_other.limit) {
-#ifdef ERTS_DIRTY_SCHEDULERS
+	    if (rq->wakeup_other > wo_params->limit) {
 		if (ERTS_RUNQ_IX_IS_DIRTY(rq->ix)) {
 		    if (rq->waiting) {
 			wake_dirty_scheduler(rq);
 		    }
 		} else
-#endif
 		{
 		    int empty_rqs =
-			erts_smp_atomic32_read_acqb(&no_empty_run_queues);
+			erts_atomic32_read_acqb(&no_empty_run_queues);
 		    if (flags & ERTS_RUNQ_FLG_PROTECTED)
 			(void) ERTS_RUNQ_FLGS_UNSET(rq, ERTS_RUNQ_FLG_PROTECTED);
 		    if (empty_rqs != 0)
@@ -5763,56 +5319,58 @@ wakeup_other_check(ErtsRunQueue *rq, Uint32 flags)
 }
 
 static void
-wakeup_other_set_limit(void)
+wakeup_other_set_limit(ErtsWakeupOtherParams *params)
 {
-    switch (wakeup_other.threshold) {
+    switch (params->threshold) {
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_HIGH:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_HIGH;
-	wakeup_other.dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_VERY_HIGH;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_HIGH;
+	params->dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_VERY_HIGH;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_HIGH:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_HIGH;
-	wakeup_other.dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_HIGH;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_HIGH;
+	params->dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_HIGH;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_MEDIUM:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_MEDIUM;
-	wakeup_other.dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_MEDIUM;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_MEDIUM;
+	params->dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_MEDIUM;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_LOW:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_LOW;
-	wakeup_other.dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_LOW;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_LOW;
+	params->dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_LOW;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_LOW:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_LOW;
-	wakeup_other.dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_VERY_LOW;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_LOW;
+	params->dec_shift = ERTS_WAKEUP_OTHER_DEC_SHIFT_VERY_LOW;
 	break;
     }
-    if (wakeup_other.dec_shift < 0)
-	wakeup_other.dec_mask = (1 << (sizeof(wakeup_other.dec_mask)*8
-				       + wakeup_other.dec_shift)) - 1;
+
+    if (params->dec_shift < 0)
+	params->dec_mask = (1 << (sizeof(params->dec_mask)*8
+				  + params->dec_shift)) - 1;
     else {
-	wakeup_other.dec_mask = 0;
-	wakeup_other.dec_mask = ~wakeup_other.dec_mask;
+	params->dec_mask = 0;
+	params->dec_mask = ~params->dec_mask;
     }
 }
 
 static void
 wakeup_other_check_legacy(ErtsRunQueue *rq, Uint32 flags)
 {
+    ErtsWakeupOtherParams *wo_params = runq_get_wakeup_other_params(rq);
     int wo_reds = rq->wakeup_other_reds;
     if (wo_reds) {
-	erts_aint32_t len = erts_smp_atomic32_read_dirty(&rq->len);
+	erts_aint32_t len = erts_atomic32_read_dirty(&rq->len);
 	if (len < 2) {
 	    rq->wakeup_other -= ERTS_WAKEUP_OTHER_DEC_LEGACY*wo_reds;
 	    if (rq->wakeup_other < 0)
 		rq->wakeup_other = 0;
 	}
-	else if (rq->wakeup_other < wakeup_other.limit)
+	else if (rq->wakeup_other < wo_params->limit)
 	    rq->wakeup_other += len*wo_reds + ERTS_WAKEUP_OTHER_FIXED_INC_LEGACY;
 	else {
 	    if (flags & ERTS_RUNQ_FLG_PROTECTED)
 		(void) ERTS_RUNQ_FLGS_UNSET(rq, ERTS_RUNQ_FLG_PROTECTED);
-	    if (erts_smp_atomic32_read_acqb(&no_empty_run_queues) != 0) {
+	    if (erts_atomic32_read_acqb(&no_empty_run_queues) != 0) {
 		wake_scheduler_on_empty_runq(rq);
 		rq->wakeup_other = 0;
 	    }
@@ -5823,23 +5381,23 @@ wakeup_other_check_legacy(ErtsRunQueue *rq, Uint32 flags)
 }
 
 static void
-wakeup_other_set_limit_legacy(void)
+wakeup_other_set_limit_legacy(ErtsWakeupOtherParams *params)
 {
-    switch (wakeup_other.threshold) {
+    switch (params->threshold) {
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_HIGH:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_HIGH_LEGACY;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_HIGH_LEGACY;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_HIGH:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_HIGH_LEGACY;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_HIGH_LEGACY;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_MEDIUM:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_MEDIUM_LEGACY;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_MEDIUM_LEGACY;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_LOW:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_LOW_LEGACY;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_LOW_LEGACY;
 	break;
     case ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_LOW:
-	wakeup_other.limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_LOW_LEGACY;
+	params->limit = ERTS_WAKEUP_OTHER_LIMIT_VERY_LOW_LEGACY;
 	break;
     }
 }
@@ -5847,15 +5405,21 @@ wakeup_other_set_limit_legacy(void)
 static void
 set_wakeup_other_data(void)
 {
-    switch (wakeup_other.type) {
-    case ERTS_SCHED_WAKEUP_OTHER_TYPE_DEFAULT:
-	wakeup_other.check = wakeup_other_check;
-	wakeup_other_set_limit();
-	break;
-    case ERTS_SCHED_WAKEUP_OTHER_TYPE_LEGACY:
-	wakeup_other.check = wakeup_other_check_legacy;
-	wakeup_other_set_limit_legacy();
-	break;
+    ErtsSchedType type;
+
+    for (type = ERTS_SCHED_TYPE_FIRST; type <= ERTS_SCHED_TYPE_LAST; type++) {
+        ErtsWakeupOtherParams *params = &sched_wakeup_other_params[type];
+
+        switch (params->type) {
+        case ERTS_SCHED_WAKEUP_OTHER_TYPE_DEFAULT:
+            params->check = wakeup_other_check;
+            wakeup_other_set_limit(params);
+            break;
+        case ERTS_SCHED_WAKEUP_OTHER_TYPE_LEGACY:
+            params->check = wakeup_other_check_legacy;
+            wakeup_other_set_limit_legacy(params);
+            break;
+        }
     }
 }
 
@@ -5863,7 +5427,7 @@ static int
 no_runqs_to_supervise(void)
 {
     int used;
-    erts_aint32_t nerq = erts_smp_atomic32_read_acqb(&no_empty_run_queues);
+    erts_aint32_t nerq = erts_atomic32_read_acqb(&no_empty_run_queues);
     if (nerq <= 0)
 	return 0;
     get_no_runqs(NULL, &used);
@@ -5896,89 +5460,78 @@ runq_supervisor(void *unused)
 	for (ix = 0; ix < no_rqs; ix++) {
 	    ErtsRunQueue *rq = ERTS_RUNQ_IX(ix);
 	    if (ERTS_RUNQ_FLGS_GET(rq) & ERTS_RUNQ_FLG_NONEMPTY) {
-		erts_smp_runq_lock(rq);
-		if (erts_smp_atomic32_read_dirty(&rq->len) != 0)
+		erts_runq_lock(rq);
+		if (erts_atomic32_read_dirty(&rq->len) != 0)
 		    wake_scheduler_on_empty_runq(rq); /* forced wakeup... */
-		erts_smp_runq_unlock(rq);
+		erts_runq_unlock(rq);
 	    }
 	}
     }
     return NULL;
 }
 
-#endif
 
 void
 erts_early_init_scheduling(int no_schedulers)
 {
+    ErtsSchedType type;
+
     aux_work_timeout_early_init(no_schedulers);
-#ifdef ERTS_SMP
-    wakeup_other.threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_MEDIUM;
-    wakeup_other.type = ERTS_SCHED_WAKEUP_OTHER_TYPE_DEFAULT;
-#endif
-    sched_busy_wait.sys_schedule = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM;
-    sched_busy_wait.tse = (ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM
-			   * ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT);
-    sched_busy_wait.aux_work = (ERTS_SCHED_SYS_SLEEP_SPINCOUNT_MEDIUM
-				* ERTS_SCHED_AUX_WORK_SLEEP_SPINCOUNT_FACT_MEDIUM);
+
+    for (type = ERTS_SCHED_TYPE_FIRST; type <= ERTS_SCHED_TYPE_LAST; type++) {
+        erts_sched_set_wakeup_other_threshold(type, "medium");
+        erts_sched_set_wakeup_other_type(type, "default");
+
+        erts_sched_set_busy_wait_threshold(type, "medium");
+    }
+
+    erts_sched_set_busy_wait_threshold(ERTS_SCHED_DIRTY_CPU, "short");
+    erts_sched_set_busy_wait_threshold(ERTS_SCHED_DIRTY_IO, "short");
 }
 
 int
-erts_sched_set_wakeup_other_thresold(char *str)
-{
-#ifdef ERTS_SMP
-    ErtsSchedWakeupOtherThreshold threshold;
-    if (sys_strcmp(str, "very_high") == 0)
-	threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_HIGH;
-    else if (sys_strcmp(str, "high") == 0)
-	threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_HIGH;
-    else if (sys_strcmp(str, "medium") == 0)
-	threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_MEDIUM;
-    else if (sys_strcmp(str, "low") == 0)
-	threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_LOW;
-    else if (sys_strcmp(str, "very_low") == 0)
-	threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_LOW;
-    else
-	return EINVAL;
-    wakeup_other.threshold = threshold;
-    set_wakeup_other_data();
+erts_sched_set_wakeup_other_threshold(ErtsSchedType sched_type, char *str)
+{
+    ErtsWakeupOtherParams *params = &sched_wakeup_other_params[sched_type];
+
+    if (sys_strcmp(str, "very_high") == 0) {
+        params->threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_HIGH;
+    } else if (sys_strcmp(str, "high") == 0) {
+        params->threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_HIGH;
+    } else if (sys_strcmp(str, "medium") == 0) {
+        params->threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_MEDIUM;
+    } else if (sys_strcmp(str, "low") == 0) {
+        params->threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_LOW;
+    } else if (sys_strcmp(str, "very_low") == 0) {
+        params->threshold = ERTS_SCHED_WAKEUP_OTHER_THRESHOLD_VERY_LOW;
+    } else {
+        return EINVAL;
+    }
+
     return 0;
-#else
-    if (sys_strcmp(str, "very_high") == 0 || sys_strcmp(str, "high") == 0 ||
-	sys_strcmp(str, "medium") == 0 || sys_strcmp(str, "low") == 0 ||
-	sys_strcmp(str, "very_low") == 0) {
-	return 0;
-    } 
-    return EINVAL;
-#endif
 }
 
 int
-erts_sched_set_wakeup_other_type(char *str)
+erts_sched_set_wakeup_other_type(ErtsSchedType sched_type, char *str)
 {
-#ifdef ERTS_SMP
-    ErtsSchedWakeupOtherType type;
-    if (sys_strcmp(str, "default") == 0)
-	type = ERTS_SCHED_WAKEUP_OTHER_TYPE_DEFAULT;
-    else if (sys_strcmp(str, "legacy") == 0)
-	type = ERTS_SCHED_WAKEUP_OTHER_TYPE_LEGACY;
-    else
-	return EINVAL;
-    wakeup_other.type = type;
+    ErtsWakeupOtherParams *params = &sched_wakeup_other_params[sched_type];
+
+    if (sys_strcmp(str, "default") == 0) {
+        params->type = ERTS_SCHED_WAKEUP_OTHER_TYPE_DEFAULT;
+    } else if (sys_strcmp(str, "legacy") == 0) {
+        params->type = ERTS_SCHED_WAKEUP_OTHER_TYPE_LEGACY;
+    } else {
+        return EINVAL;
+    }
+
     return 0;
-#else
-    if (sys_strcmp(str, "default") == 0 || sys_strcmp(str, "legacy") == 0) {
-	return 0;
-    } 
-    return EINVAL;
-#endif
 }
 
 int
-erts_sched_set_busy_wait_threshold(char *str)
+erts_sched_set_busy_wait_threshold(ErtsSchedType sched_type, char *str)
 {
-    int sys_sched;
-    int aux_work_fact;
+    ErtsBusyWaitParams *params = &sched_busy_wait_params[sched_type];
+    int aux_work_fact, sys_sched;
 
     if (sys_strcmp(str, "very_long") == 0) {
 	sys_sched = ERTS_SCHED_SYS_SLEEP_SPINCOUNT_VERY_LONG;
@@ -6008,9 +5561,9 @@ erts_sched_set_busy_wait_threshold(char *str)
 	return EINVAL;
     }
 
-    sched_busy_wait.sys_schedule = sys_sched;
-    sched_busy_wait.tse = sys_sched*ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT;
-    sched_busy_wait.aux_work = sys_sched*aux_work_fact;
+    params->sys_schedule = sys_sched;
+    params->tse = sys_sched * ERTS_SCHED_TSE_SLEEP_SPINCOUNT_FACT;
+    params->aux_work = sys_sched * aux_work_fact;
 
     return 0;
 }
@@ -6042,7 +5595,6 @@ init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp)
         case ERTS_SCHED_NORMAL:
             id = (int) esdp->no;
             break;
-#ifdef ERTS_DIRTY_SCHEDULERS
         case ERTS_SCHED_DIRTY_CPU:
             id = (int) erts_no_schedulers;
             id += (int) esdp->dirty_no;
@@ -6052,7 +5604,6 @@ init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp)
             id += (int) erts_no_dirty_cpu_schedulers;
             id += (int) esdp->dirty_no;
             break;
-#endif
         default:
             ERTS_INTERNAL_ERROR("Invalid scheduler type");
             break;
@@ -6062,7 +5613,6 @@ init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp)
     awdp->sched_id = id;
     awdp->esdp = esdp;
     awdp->ssi = esdp ? esdp->ssi : NULL;
-#ifdef ERTS_SMP
     awdp->latest_wakeup = ERTS_THR_PRGR_VAL_FIRST;
     awdp->misc.thr_prgr = ERTS_THR_PRGR_VAL_WAITING;
     awdp->dd.thr_prgr = ERTS_THR_PRGR_VAL_WAITING;
@@ -6071,15 +5621,9 @@ init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp)
     awdp->later_op.size = 0;
     awdp->later_op.first = NULL;
     awdp->later_op.last = NULL;
-#endif
-#ifdef ERTS_USE_ASYNC_READY_Q
-#ifdef ERTS_SMP
     awdp->async_ready.need_thr_prgr = 0;
     awdp->async_ready.thr_prgr = ERTS_THR_PRGR_VAL_WAITING;
-#endif
     awdp->async_ready.queue = NULL;
-#endif
-#ifdef ERTS_SMP
     awdp->delayed_wakeup.next = ERTS_DELAYED_WAKEUP_INFINITY;
     if (!dawwp) {
 	awdp->delayed_wakeup.job = NULL;
@@ -6095,7 +5639,6 @@ init_aux_work_data(ErtsAuxWorkData *awdp, ErtsSchedulerData *esdp, char *dawwp)
 	for (i = 0; i <= erts_no_schedulers; i++)
 	    awdp->delayed_wakeup.sched2jix[i] = -1;
     }
-#endif
     awdp->debug.wait_completed.flags = 0;
     awdp->debug.wait_completed.callback = NULL;
     awdp->debug.wait_completed.arg = NULL;
@@ -6110,11 +5653,9 @@ init_scheduler_data(ErtsSchedulerData* esdp, int num,
                     Uint64 time_stamp)
 {
     esdp->timer_wheel = NULL;
-#ifdef ERTS_SMP
     erts_bits_init_state(&esdp->erl_bits_state);
     esdp->match_pseudo_process = NULL;
     esdp->free_process = NULL;
-#endif
     esdp->x_reg_array =
 	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_BEAM_REGISTER,
 					   ERTS_X_REGS_ALLOCATED *
@@ -6122,7 +5663,6 @@ init_scheduler_data(ErtsSchedulerData* esdp, int num,
     esdp->f_reg_array =
 	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_BEAM_REGISTER,
 					   MAX_REG * sizeof(FloatDef));
-#ifdef ERTS_DIRTY_SCHEDULERS
     esdp->run_queue = runq;
     if (ERTS_RUNQ_IX_IS_DIRTY(runq->ix)) {
 	esdp->no = 0;
@@ -6150,22 +5690,18 @@ init_scheduler_data(ErtsSchedulerData* esdp, int num,
     esdp->dirty_shadow_process = shadow_proc;
     if (shadow_proc) {
 	erts_init_empty_process(shadow_proc);
-	erts_smp_atomic32_init_nob(&shadow_proc->state,
+	erts_atomic32_init_nob(&shadow_proc->state,
 				   (ERTS_PSFLG_ACTIVE
 				    | ERTS_PSFLG_DIRTY_RUNNING
 				    | ERTS_PSFLG_PROXY));
 	shadow_proc->static_flags = ERTS_STC_FLG_SHADOW_PROC;
     }
-#else
-    runq->scheduler = esdp;
-    esdp->run_queue = runq;
-    esdp->no = (Uint) num;
-    esdp->type = ERTS_SCHED_NORMAL;
-#endif
 
+    ssi->esdp = esdp;
     esdp->ssi = ssi;
     esdp->current_process = NULL;
     esdp->current_port = NULL;
+    esdp->current_nif = NULL;
 
     esdp->virtual_reds = 0;
     esdp->cpu_id = -1;
@@ -6181,11 +5717,15 @@ init_scheduler_data(ErtsSchedulerData* esdp, int num,
     esdp->io.out = (Uint64) 0;
     esdp->io.in = (Uint64) 0;
 
+    esdp->pending_signal.sig = NULL;
+    esdp->pending_signal.to = THE_NON_VALUE;
+#ifdef DEBUG
+    esdp->pending_signal.dbg_from = NULL;
+#endif
+
     if (daww_ptr) {
 	init_aux_work_data(&esdp->aux_work_data, esdp, *daww_ptr);
-#ifdef ERTS_SMP
 	*daww_ptr += daww_sz;
-#endif
     }
 
     esdp->reductions = 0;
@@ -6195,27 +5735,21 @@ init_scheduler_data(ErtsSchedulerData* esdp, int num,
 }
 
 void
-erts_init_scheduling(int no_schedulers, int no_schedulers_online
-#ifdef ERTS_DIRTY_SCHEDULERS
-		     , int no_dirty_cpu_schedulers, int no_dirty_cpu_schedulers_online,
+erts_init_scheduling(int no_schedulers, int no_schedulers_online, int no_poll_threads,
+		     int no_dirty_cpu_schedulers, int no_dirty_cpu_schedulers_online,
 		     int no_dirty_io_schedulers
-#endif
 		     )
 {
     int ix, n, no_ssi, tot_rqs;
     char *daww_ptr;
     size_t daww_sz;
     size_t size_runqs;
-#ifdef ERTS_SMP
     erts_aint32_t set_schdlr_sspnd_change_flags;
-#endif
 
     init_misc_op_list_alloc();
     init_proc_sys_task_queues_alloc();
 
-#ifdef ERTS_SMP
     set_wakeup_other_data();
-#endif
 
 #if ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT
     if (erts_sched_balance_util)
@@ -6225,12 +5759,11 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
     ASSERT(no_schedulers_online <= no_schedulers);
     ASSERT(no_schedulers_online >= 1);
     ASSERT(no_schedulers >= 1);
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(no_dirty_cpu_schedulers <= no_schedulers);
     ASSERT(no_dirty_cpu_schedulers >= 1);
     ASSERT(no_dirty_cpu_schedulers_online <= no_schedulers_online);
     ASSERT(no_dirty_cpu_schedulers_online >= 1);
-#endif
+    ASSERT(erts_no_poll_threads == no_poll_threads);
 
     /* Create and initialize run queues */
 
@@ -6239,9 +5772,7 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
     size_runqs = sizeof(ErtsAlignedRunQueue) * tot_rqs;
     erts_aligned_run_queues =
 	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_RUNQS, size_runqs);
-#ifdef ERTS_SMP
-    erts_smp_atomic32_init_nob(&no_empty_run_queues, 0);
-#endif
+    erts_atomic32_init_nob(&no_empty_run_queues, 0);
 
     erts_no_run_queues = n;
 
@@ -6255,20 +5786,16 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	 * id if the esdp->no <-> ix+1 mapping change.
 	 */
 
-	erts_smp_mtx_init(&rq->mtx, "run_queue", make_small(ix + 1),
+	erts_mtx_init(&rq->mtx, "run_queue", make_small(ix + 1),
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_SCHEDULER);
-	erts_smp_cnd_init(&rq->cnd);
+	erts_cnd_init(&rq->cnd);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-#ifdef ERTS_SMP
         if (ERTS_RUNQ_IX_IS_DIRTY(ix)) {
-            erts_smp_spinlock_init(&rq->sleepers.lock, "dirty_run_queue_sleep_list",
+            erts_spinlock_init(&rq->sleepers.lock, "dirty_run_queue_sleep_list",
                 make_small(ix + 1),
                 ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_SCHEDULER);
         }
 	rq->sleepers.list = NULL;
-#endif
-#endif
 
 	rq->waiting = 0;
 	rq->woken = 0;
@@ -6281,16 +5808,15 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	}
 	rq->out_of_work_count = 0;
 	rq->max_len = 0;
-	erts_smp_atomic32_set_nob(&rq->len, 0);
+	erts_atomic32_set_nob(&rq->len, 0);
 	rq->wakeup_other = 0;
 	rq->wakeup_other_reds = 0;
 
-	rq->procs.pending_exiters = NULL;
 	rq->procs.context_switches = 0;
 	rq->procs.reductions = 0;
 
 	for (pix = 0; pix < ERTS_NO_PROC_PRIO_LEVELS; pix++) {
-	    erts_smp_atomic32_init_nob(&rq->procs.prio_info[pix].len, 0);
+	    erts_atomic32_init_nob(&rq->procs.prio_info[pix].len, 0);
 	    rq->procs.prio_info[pix].max_len = 0;
 	    rq->procs.prio_info[pix].reds = 0;
 	    if (pix < ERTS_NO_PROC_PRIO_LEVELS - 1) {
@@ -6302,7 +5828,7 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	rq->misc.start = NULL;
 	rq->misc.end = NULL;
 
-	erts_smp_atomic32_init_nob(&rq->ports.info.len, 0);
+	erts_atomic32_init_nob(&rq->ports.info.len, 0);
 	rq->ports.info.max_len = 0;
 	rq->ports.info.reds = 0;
 	rq->ports.start = NULL;
@@ -6314,7 +5840,6 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 
     }
 
-#ifdef ERTS_SMP
 
     if (erts_no_run_queues != 1) {
 	run_queue_info = erts_alloc(ERTS_ALC_T_RUNQ_BLNS,
@@ -6325,52 +5850,42 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 					* erts_no_run_queues));
     }
 
-#endif
 
     n = (int) no_schedulers;
     erts_no_schedulers = n;
     erts_no_total_schedulers = n;
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_no_dirty_cpu_schedulers = no_dirty_cpu_schedulers;
     erts_no_total_schedulers += no_dirty_cpu_schedulers;
     erts_no_dirty_io_schedulers = no_dirty_io_schedulers;
     erts_no_total_schedulers += no_dirty_io_schedulers;
-#endif
 
     /* Create and initialize scheduler sleep info */
-#ifdef ERTS_SMP
-    no_ssi = n+1;
-#else
-    no_ssi = 1;
-#endif
+    no_ssi = n + 1 /* aux thread */;
     aligned_sched_sleep_info =
 	erts_alloc_permanent_cache_aligned(
 	    ERTS_ALC_T_SCHDLR_SLP_INFO,
 	    no_ssi*sizeof(ErtsAlignedSchedulerSleepInfo));
     for (ix = 0; ix < no_ssi; ix++) {
 	ErtsSchedulerSleepInfo *ssi = &aligned_sched_sleep_info[ix].ssi;
-#ifdef ERTS_SMP
 #if 0 /* no need to initialize these... */
 	ssi->next = NULL;
 	ssi->prev = NULL;
 #endif
-	erts_smp_atomic32_init_nob(&ssi->flags, 0);
+        ssi->esdp = NULL;
+	erts_atomic32_init_nob(&ssi->flags, 0);
 	ssi->event = NULL; /* initialized in sched_thread_func */
-#endif
 	erts_atomic32_init_nob(&ssi->aux_work, 0);
     }
 
-#ifdef ERTS_SMP
-    aligned_sched_sleep_info++;
+    aligned_sched_sleep_info += 1 /* aux thread */;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     aligned_dirty_cpu_sched_sleep_info =
 	erts_alloc_permanent_cache_aligned(
 	    ERTS_ALC_T_SCHDLR_SLP_INFO,
 	    no_dirty_cpu_schedulers*sizeof(ErtsAlignedSchedulerSleepInfo));
     for (ix = 0; ix < no_dirty_cpu_schedulers; ix++) {
 	ErtsSchedulerSleepInfo *ssi = &aligned_dirty_cpu_sched_sleep_info[ix].ssi;
-	erts_smp_atomic32_init_nob(&ssi->flags, 0);
+	erts_atomic32_init_nob(&ssi->flags, 0);
 	ssi->event = NULL; /* initialized in sched_dirty_cpu_thread_func */
 	erts_atomic32_init_nob(&ssi->aux_work, 0);
     }
@@ -6380,24 +5895,29 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	    no_dirty_io_schedulers*sizeof(ErtsAlignedSchedulerSleepInfo));
     for (ix = 0; ix < no_dirty_io_schedulers; ix++) {
 	ErtsSchedulerSleepInfo *ssi = &aligned_dirty_io_sched_sleep_info[ix].ssi;
-	erts_smp_atomic32_init_nob(&ssi->flags, 0);
+	erts_atomic32_init_nob(&ssi->flags, 0);
 	ssi->event = NULL; /* initialized in sched_dirty_io_thread_func */
 	erts_atomic32_init_nob(&ssi->aux_work, 0);
     }
-#endif
-#endif
+
+    aligned_poll_thread_sleep_info =
+        erts_alloc_permanent_cache_aligned(
+	    ERTS_ALC_T_SCHDLR_SLP_INFO,
+	    no_poll_threads*sizeof(ErtsAlignedSchedulerSleepInfo));
+    for (ix = 0; ix < no_poll_threads; ix++) {
+        ErtsSchedulerSleepInfo *ssi = &aligned_poll_thread_sleep_info[ix].ssi;
+        ssi->esdp = NULL;
+	erts_atomic32_init_nob(&ssi->flags, 0);
+	ssi->event = NULL; /* initialized in poll_thread */
+	erts_atomic32_init_nob(&ssi->aux_work, 0);
+    }
 
     /* Create and initialize scheduler specific data */
 
-#ifdef ERTS_SMP
     daww_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE((sizeof(ErtsDelayedAuxWorkWakeupJob)
 					      + sizeof(int))*(n+1));
     daww_ptr = erts_alloc_permanent_cache_aligned(ERTS_ALC_T_SCHDLR_DATA,
 						  daww_sz*n);
-#else
-    daww_sz = 0;
-    daww_ptr = NULL;
-#endif
 
     erts_aligned_scheduler_data = 
 	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_SCHDLR_DATA,
@@ -6410,7 +5930,6 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 			    NULL, 0);
     }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     {
         Uint64 ts = sched_wall_time_ts();
 	int dirty_scheds = no_dirty_cpu_schedulers + no_dirty_io_schedulers;
@@ -6441,7 +5960,6 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 				&adsp[adspix++].dsp, ts);
 	}
     }
-#endif
 
     init_misc_aux_work();
     init_swtreq_alloc();
@@ -6450,20 +5968,22 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
     erts_atomic32_init_nob(&debug_wait_completed_count, 0); /* debug only */
     debug_wait_completed_flags = 0;
 
-#ifdef ERTS_SMP
-
     aux_thread_aux_work_data =
 	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_SCHDLR_DATA,
 					   sizeof(ErtsAuxWorkData));
 
+    poll_thread_aux_work_data =
+	erts_alloc_permanent_cache_aligned(ERTS_ALC_T_SCHDLR_DATA,
+					   no_poll_threads * sizeof(ErtsAuxWorkData));
+
     init_no_runqs(no_schedulers_online, no_schedulers_online);
     balance_info.last_active_runqs = no_schedulers;
-    erts_smp_mtx_init(&balance_info.update_mtx, "migration_info_update", NIL,
+    erts_mtx_init(&balance_info.update_mtx, "migration_info_update", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_SCHEDULER);
     balance_info.forced_check_balance = 0;
     balance_info.halftime = 1;
     balance_info.full_reds_history_index = 0;
-    erts_smp_atomic32_init_nob(&balance_info.checking_balance, 0);
+    erts_atomic32_init_nob(&balance_info.checking_balance, 0);
     balance_info.prev_rise.active_runqs = 0;
     balance_info.prev_rise.max_len = 0;
     balance_info.prev_rise.reds = 0;
@@ -6494,7 +6014,6 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	    suspend_run_queue(ERTS_RUNQ_IX(ix));
     }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
     schdlr_sspnd_set_nscheds(&schdlr_sspnd.online,
 			     ERTS_SCHED_DIRTY_CPU,
@@ -6511,7 +6030,7 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 	set_schdlr_sspnd_change_flags |= ERTS_SCHDLR_SSPND_CHNG_DCPU_ONLN;
 	for (ix = no_dirty_cpu_schedulers_online; ix < no_dirty_cpu_schedulers; ix++) {
 	    ErtsSchedulerData* esdp = ERTS_DIRTY_CPU_SCHEDULER_IX(ix);
-	    erts_smp_atomic32_read_bor_nob(&esdp->ssi->flags, ERTS_SSI_FLG_SUSPENDED);
+	    erts_atomic32_read_bor_nob(&esdp->ssi->flags, ERTS_SSI_FLG_SUSPENDED);
 	}
     }
 
@@ -6525,54 +6044,26 @@ erts_init_scheduling(int no_schedulers, int no_schedulers_online
 			     ERTS_SCHED_DIRTY_IO,
 			     no_dirty_io_schedulers);
 
-    erts_smp_atomic32_init_nob(&dirty_count.cpu.active,
+    erts_atomic32_init_nob(&dirty_count.cpu.active,
                                (erts_aint32_t) no_dirty_cpu_schedulers);
-    erts_smp_atomic32_init_nob(&dirty_count.io.active,
+    erts_atomic32_init_nob(&dirty_count.io.active,
                                (erts_aint32_t) no_dirty_io_schedulers);
 
-#endif
 
     if (set_schdlr_sspnd_change_flags)
-	erts_smp_atomic32_set_nob(&schdlr_sspnd.changing,
+	erts_atomic32_set_nob(&schdlr_sspnd.changing,
 				  set_schdlr_sspnd_change_flags);
 
-    erts_smp_atomic32_init_nob(&doing_sys_schedule, 0);
-
     init_misc_aux_work();
 
-#else /* !ERTS_SMP */
-    {
-	ErtsSchedulerData *esdp;
-	esdp = ERTS_SCHEDULER_IX(0);
-	erts_scheduler_data = esdp;
-#ifdef USE_THREADS
-	erts_tsd_set(sched_data_key, (void *) esdp);
-#endif
-    }
-    erts_no_dirty_cpu_schedulers = 0;
-    erts_no_dirty_io_schedulers = 0;
-#endif
-
-    erts_smp_atomic32_init_nob(&function_calls, 0);
 
     /* init port tasks */
     erts_port_task_init();
 
-#ifndef ERTS_SMP
-#ifdef ERTS_DO_VERIFY_UNUSED_TEMP_ALLOC
-    erts_scheduler_data->verify_unused_temp_alloc
-	= erts_alloc_get_verify_unused_temp_alloc(
-	    &erts_scheduler_data->verify_unused_temp_alloc_data);
-    ERTS_VERIFY_UNUSED_TEMP_ALLOC(NULL);
-#endif
-#endif
 
-    erts_smp_atomic32_init_relb(&erts_halt_progress, -1);
+    erts_atomic32_init_relb(&erts_halt_progress, -1);
     erts_halt_code = 0;
 
-#if !defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    erts_lc_set_thread_name("scheduler 1");
-#endif
 
 }
 
@@ -6585,7 +6076,6 @@ erts_schedid2runq(Uint id)
     return ERTS_RUNQ_IX(ix);
 }
 
-#ifdef USE_THREADS
 
 ErtsSchedulerData *
 erts_get_scheduler_data(void)
@@ -6593,16 +6083,14 @@ erts_get_scheduler_data(void)
     return (ErtsSchedulerData *) erts_tsd_get(sched_data_key);
 }
 
-#endif
 
 static Process *
 make_proxy_proc(Process *prev_proxy, Process *proc, erts_aint32_t prio)
 {
     erts_aint32_t state;
     Process *proxy;
-#ifdef ERTS_SMP
-    ErtsRunQueue *rq = RUNQ_READ_RQ(&proc->run_queue);
-#endif
+    int bound;
+    ErtsRunQueue *rq = erts_get_runq_proc(proc, &bound);
 
     state = (ERTS_PSFLG_PROXY
 	     | ERTS_PSFLG_IN_RUNQ
@@ -6613,11 +6101,9 @@ make_proxy_proc(Process *prev_proxy, Process *proc, erts_aint32_t prio)
 
     if (prev_proxy) {
 	proxy = prev_proxy;
-	ASSERT(erts_smp_atomic32_read_nob(&proxy->state) & ERTS_PSFLG_PROXY);
-	erts_smp_atomic32_set_nob(&proxy->state, state);
-#ifdef ERTS_SMP
-	RUNQ_SET_RQ(&proc->run_queue, rq);
-#endif
+	ASSERT(erts_atomic32_read_nob(&proxy->state) & ERTS_PSFLG_PROXY);
+	erts_atomic32_set_nob(&proxy->state, state);
+        (void) erts_set_runq_proc(proxy, rq, &bound);
     }
     else {
 	proxy = erts_alloc(ERTS_ALC_T_PROC, sizeof(Process));
@@ -6629,11 +6115,8 @@ make_proxy_proc(Process *prev_proxy, Process *proc, erts_aint32_t prio)
 		ui32[i] = (Uint32) 0xdeadbeef;
 	}
 #endif
-	erts_smp_atomic32_init_nob(&proxy->state, state);
-#ifdef ERTS_SMP
-	erts_smp_atomic_init_nob(&proxy->run_queue,
-				 erts_smp_atomic_read_nob(&proc->run_queue));
-#endif
+	erts_atomic32_init_nob(&proxy->state, state);
+        erts_init_runq_proc(proxy, rq, bound);
     }
 
     proxy->common.id = proc->common.id;
@@ -6646,7 +6129,6 @@ make_proxy_proc(Process *prev_proxy, Process *proc, erts_aint32_t prio)
 #define ERTS_ENQUEUE_DIRTY_CPU_QUEUE 2
 #define ERTS_ENQUEUE_DIRTY_IO_QUEUE 3
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static int
 check_dirty_enqueue_in_prio_queue(Process *c_p,
@@ -6677,7 +6159,7 @@ check_dirty_enqueue_in_prio_queue(Process *c_p,
     if ((*newp) & ERTS_PSFLG_ACTIVE_SYS)
 	return ERTS_ENQUEUE_NORMAL_QUEUE;
 
-    dact = erts_smp_atomic32_read_mb(&c_p->dirty_state);
+    dact = erts_atomic32_read_mb(&c_p->dirty_state);
     if (actual & (ERTS_PSFLG_DIRTY_ACTIVE_SYS
 		  | ERTS_PSFLG_DIRTY_CPU_PROC)) {
 	max_qbit = ((dact >> ERTS_PDSFLGS_IN_CPU_PRQ_MASK_OFFSET)
@@ -6722,7 +6204,7 @@ fin_dirty_enq_s_change(Process *p,
     erts_aint32_t qbit = 1 << enq_prio;
     qbit <<= qmask_offset;
 
-    if (qbit & erts_smp_atomic32_read_bor_mb(&p->dirty_state, qbit)) {
+    if (qbit & erts_atomic32_read_bor_mb(&p->dirty_state, qbit)) {
 	/* Already enqueue by someone else... */
 	if (pstruct_reserved) {
 	    /* We reserved process struct for enqueue; clear it... */
@@ -6731,7 +6213,7 @@ fin_dirty_enq_s_change(Process *p,
 #else
 		(void)
 #endif
-		erts_smp_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_IN_RUNQ);
+		erts_atomic32_read_band_nob(&p->state, ~ERTS_PSFLG_IN_RUNQ);
 	    ASSERT(old & ERTS_PSFLG_IN_RUNQ);
 	}
 	return 0;
@@ -6740,7 +6222,6 @@ fin_dirty_enq_s_change(Process *p,
     return !0;
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 static ERTS_INLINE int
 check_enqueue_in_prio_queue(Process *c_p,
@@ -6755,14 +6236,12 @@ check_enqueue_in_prio_queue(Process *c_p,
 
     *prq_prio_p = aprio;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (actual & ERTS_PSFLGS_DIRTY_WORK) {
 	int res = check_dirty_enqueue_in_prio_queue(c_p, newp, actual,
 						    aprio, qbit);
 	if (res != ERTS_ENQUEUE_NORMAL_QUEUE)
 	    return res;
     }
-#endif
 
     max_qbit = (actual >> ERTS_PSFLGS_IN_PRQ_MASK_OFFSET) & ERTS_PSFLGS_QMASK;
     max_qbit |= 1 << ERTS_PSFLGS_QMASK_BITS;
@@ -6805,7 +6284,6 @@ select_enqueue_run_queue(int enqueue, int enq_prio, Process *p, erts_aint32_t st
 
 	return NULL;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
     case ERTS_ENQUEUE_DIRTY_CPU_QUEUE:
     case -ERTS_ENQUEUE_DIRTY_CPU_QUEUE:
@@ -6826,25 +6304,25 @@ select_enqueue_run_queue(int enqueue, int enq_prio, Process *p, erts_aint32_t st
 
 	return NULL;
 
-#endif
 
     default: {
 	ErtsRunQueue* runq;
+        int bound;
 
 	ASSERT(enqueue == ERTS_ENQUEUE_NORMAL_QUEUE
 	       || enqueue == -ERTS_ENQUEUE_NORMAL_QUEUE);
 
-	runq = erts_get_runq_proc(p);
+	runq = erts_get_runq_proc(p, &bound);
 
-#ifdef ERTS_SMP
-	if (!(ERTS_PSFLG_BOUND & state)) {
+	if (!bound) {
 	    ErtsRunQueue *new_runq = erts_check_emigration_need(runq, enq_prio);
-	    if (new_runq) {
-		RUNQ_SET_RQ(&p->run_queue, new_runq);
-		runq = new_runq;
-	    }
+            if (new_runq) {
+                if (erts_try_change_runq_proc(p, new_runq))
+                    runq = new_runq;
+                else
+                    runq = erts_get_runq_proc(p, NULL);
+            }
 	}
-#endif
 
 	ASSERT(runq);
 
@@ -6868,12 +6346,36 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
     int enqueue; /* < 0 -> use proxy */
     ErtsRunQueue* runq;
 
+    ASSERT(!(state & (ERTS_PSFLG_DIRTY_IO_PROC
+                      |ERTS_PSFLG_DIRTY_CPU_PROC))
+           || (BeamIsOpCode(*p->i, op_call_nif)
+               || BeamIsOpCode(*p->i, op_apply_bif)));
+
+    a = state;
+
+    /* Clear activ-sys if needed... */
+    while (1) {
+        n = e = a;
+        if (a & ERTS_PSFLG_ACTIVE_SYS) {
+            if (a & (ERTS_PSFLG_SIG_Q
+                     | ERTS_PSFLG_SIG_IN_Q
+                     | ERTS_PSFLG_SYS_TASKS))
+                break;
+            /* Clear active-sys */
+            n &= ~ERTS_PSFLG_ACTIVE_SYS;
+        }
+        a = erts_atomic32_cmpxchg_nob(&p->state, n, e);
+        if (a == e) {
+            a = n;
+            break;
+        }
+    }
+
     if (!is_normal_sched)
 	running_flgs = ERTS_PSFLG_DIRTY_RUNNING|ERTS_PSFLG_DIRTY_RUNNING_SYS;
     else {
 	running_flgs = ERTS_PSFLG_RUNNING|ERTS_PSFLG_RUNNING_SYS;
-#ifdef ERTS_DIRTY_SCHEDULERS
-        if (state & ERTS_PSFLG_DIRTY_ACTIVE_SYS
+        if ((a & ERTS_PSFLG_DIRTY_ACTIVE_SYS)
             && (p->flags & (F_DELAY_GC|F_DISABLE_GC))) {
             /*
              * Delay dirty GC; will be enabled automatically
@@ -6886,15 +6388,12 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
              */
             ASSERT(!(p->flags & (F_DIRTY_CLA | F_DIRTY_GC_HIBERNATE)));
 
-            state = erts_smp_atomic32_read_band_nob(&p->state,
-                                                    ~ERTS_PSFLG_DIRTY_ACTIVE_SYS);
-            state &= ~ERTS_PSFLG_DIRTY_ACTIVE_SYS;
+            a = erts_atomic32_read_band_nob(&p->state,
+                                            ~ERTS_PSFLG_DIRTY_ACTIVE_SYS);
+            a &= ~ERTS_PSFLG_DIRTY_ACTIVE_SYS;
         }
-#endif
     }
 
-    a = state;
-
     while (1) {
 	n = e = a;
 
@@ -6902,12 +6401,17 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
 
 	enqueue = ERTS_ENQUEUE_NOT;
 
+        ASSERT(((a & (ERTS_PSFLG_EXITING|ERTS_PSFLG_FREE))
+                != ERTS_PSFLG_EXITING)
+               || ((a & (ERTS_PSFLG_ACTIVE|ERTS_PSFLG_SUSPENDED))
+                   == ERTS_PSFLG_ACTIVE));
+
 	n &= ~running_flgs;
 	if ((a & (ERTS_PSFLG_ACTIVE_SYS|ERTS_PSFLG_DIRTY_ACTIVE_SYS))
 	    || (a & (ERTS_PSFLG_ACTIVE|ERTS_PSFLG_SUSPENDED)) == ERTS_PSFLG_ACTIVE) {
 	    enqueue = check_enqueue_in_prio_queue(p, &enq_prio, &n, a);
 	}
-	a = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
+	a = erts_atomic32_cmpxchg_mb(&p->state, n, e);
 	if (a == e)
 	    break;
     }
@@ -6919,7 +6423,7 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
 	if (erts_system_profile_flags.runnable_procs) {
 
 	    /* Status lock prevents out of order "runnable proc" trace msgs */
-	    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
+	    ERTS_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
 
 	    if (!(a & (ERTS_PSFLG_ACTIVE_SYS|ERTS_PSFLG_DIRTY_ACTIVE_SYS))
 		&& (!(a & ERTS_PSFLG_ACTIVE) || (a & ERTS_PSFLG_SUSPENDED))) {
@@ -6931,15 +6435,10 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
 	if (proxy)
 	    free_proxy_proc(proxy);
 
-	erts_smp_runq_lock(c_rq);
+	erts_runq_lock(c_rq);
 
-#if !defined(ERTS_SMP)
-        /* Decrement refc if process struct is free... */
-        return !!(n & ERTS_PSFLG_FREE);
-#else
         /* Decrement refc if scheduled out from dirty scheduler... */
 	return !is_normal_sched;
-#endif
     }
     else {
 	Process* sched_p;
@@ -6958,7 +6457,7 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
 
 	ASSERT(runq);
 
-	erts_smp_runq_lock(runq);
+	erts_runq_lock(runq);
 
         if (is_normal_sched && sched_p == p && ERTS_RUNQ_IX_IS_DIRTY(runq->ix))
             erts_proc_inc_refc(p); /* Needs to be done before enqueue_process() */
@@ -6969,11 +6468,11 @@ schedule_out_process(ErtsRunQueue *c_rq, erts_aint32_t state, Process *p,
 	if (runq == c_rq)
 	    return 0;
 
-	erts_smp_runq_unlock(runq);
+	erts_runq_unlock(runq);
 
 	smp_notify_inc_runq(runq);
 
-	erts_smp_runq_lock(c_rq);
+	erts_runq_lock(c_rq);
 
         /*
          * Decrement refc if process is scheduled out by a
@@ -7021,12 +6520,12 @@ add2runq(int enqueue, erts_aint32_t prio,
 	    sched_p = make_proxy_proc(pxy, proc, prio);
 	}
 
-	erts_smp_runq_lock(runq);
+	erts_runq_lock(runq);
 
 	/* Enqueue the process */
 	enqueue_process(runq, (int) prio, sched_p);
 
-	erts_smp_runq_unlock(runq);
+	erts_runq_unlock(runq);
 	smp_notify_inc_runq(runq);
     }
 }
@@ -7051,7 +6550,7 @@ change_proc_schedule_state(Process *p,
     unsigned int lock_status = (prof_runnable_procs
 				&& !(locks & ERTS_PROC_LOCK_STATUS));
 
-    ERTS_SMP_LC_ASSERT(locks == erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(locks == erts_proc_lc_my_proc_locks(p));
 
     ASSERT(!(a & ERTS_PSFLG_PROXY));
     ASSERT((clear_state_flags & (ERTS_PSFLG_RUNNING
@@ -7066,7 +6565,7 @@ change_proc_schedule_state(Process *p,
 			       | ERTS_PSFLG_ACTIVE_SYS)) == 0);
 
     if (lock_status)
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 
     while (1) {
 	erts_aint32_t e;
@@ -7090,11 +6589,9 @@ change_proc_schedule_state(Process *p,
 		  | ERTS_PSFLG_DIRTY_RUNNING_SYS
 		  | ERTS_PSFLG_IN_RUNQ
 		  | ERTS_PSFLG_ACTIVE)) == ERTS_PSFLG_ACTIVE
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    || (n & (ERTS_PSFLG_RUNNING
 		     | ERTS_PSFLG_RUNNING_SYS
 		     | ERTS_PSFLG_EXITING)) == ERTS_PSFLG_EXITING
-#endif
 	    ) {
 	    /*
 	     * Active and seemingly need to be enqueued, but
@@ -7104,7 +6601,7 @@ change_proc_schedule_state(Process *p,
 	    enqueue = check_enqueue_in_prio_queue(p, enq_prio_p, &n, a);
 	}
 
-	a = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
+	a = erts_atomic32_cmpxchg_mb(&p->state, n, e);
 	if (a == e)
 	    break;
 	if (enqueue == ERTS_ENQUEUE_NOT && n == a)
@@ -7117,23 +6614,23 @@ change_proc_schedule_state(Process *p,
 
 	if (((n & (ERTS_PSFLG_SUSPENDED
 		   | ERTS_PSFLG_ACTIVE)) == ERTS_PSFLG_ACTIVE)
-	    && (!(a & (ERTS_PSFLG_ACTIVE_SYS
-		       | ERTS_PSFLG_RUNNING
-		       | ERTS_PSFLG_RUNNING_SYS
-		       | ERTS_PSFLG_DIRTY_RUNNING
-		       | ERTS_PSFLG_DIRTY_RUNNING_SYS)
-		  && (!(a & ERTS_PSFLG_ACTIVE)
-		      || (a & ERTS_PSFLG_SUSPENDED))))) {
+            & ((a & (ERTS_PSFLG_SUSPENDED
+                     | ERTS_PSFLG_ACTIVE)) != ERTS_PSFLG_ACTIVE)
+	    & !(a & (ERTS_PSFLG_ACTIVE_SYS
+                     | ERTS_PSFLG_RUNNING
+                     | ERTS_PSFLG_RUNNING_SYS
+                     | ERTS_PSFLG_DIRTY_RUNNING
+                     | ERTS_PSFLG_DIRTY_RUNNING_SYS))) {
 	    /* We activated a prevously inactive process */
 	    profile_runnable_proc(p, am_active);
 	}
 
 	if (lock_status)
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     }
 
 
-    *statep = a;
+    *statep = n;
 
     return enqueue;
 }
@@ -7158,6 +6655,95 @@ erts_schedule_process(Process *p, erts_aint32_t state, ErtsProcLocks locks)
     schedule_process(p, state, locks);
 }
 
+static ERTS_INLINE erts_aint32_t
+active_sys_enqueue(Process *p, erts_aint32_t state,
+                   erts_aint32_t enable_flags, int status_locked)
+{
+    /*
+     * This function may or may not be called with status locke held.
+     * It always returns without the status lock held!
+     */
+    unsigned int prof_runnable_procs = erts_system_profile_flags.runnable_procs;
+    erts_aint32_t n, a = state, enq_prio = -1;
+    int slocked = status_locked;
+    int enqueue; /* < 0 -> use proxy */
+
+    /* Status lock prevents out of order "runnable proc" trace msgs */
+    ERTS_LC_ASSERT(slocked || !(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p)));
+    ERTS_LC_ASSERT(!slocked || (ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p)));
+
+    if (!prof_runnable_procs) {
+        if (slocked) {
+            erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+            slocked = 0;
+        }
+    }
+    else {
+        if (!slocked) {
+            erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+            slocked = !0;
+        }
+    }
+
+    ASSERT(!(state & ERTS_PSFLG_PROXY));
+
+    while (1) {
+	erts_aint32_t e;
+	n = e = a;
+
+	if (a & ERTS_PSFLG_FREE)
+	    goto cleanup; /* We don't want to schedule free processes... */
+
+	enqueue = ERTS_ENQUEUE_NOT;
+        n |= enable_flags;
+	n |= ERTS_PSFLG_ACTIVE_SYS;
+	if (!(a & (ERTS_PSFLG_RUNNING
+		   | ERTS_PSFLG_RUNNING_SYS
+		   | ERTS_PSFLG_DIRTY_RUNNING
+		   | ERTS_PSFLG_DIRTY_RUNNING_SYS)))
+	    enqueue = check_enqueue_in_prio_queue(p, &enq_prio, &n, a);
+	a = erts_atomic32_cmpxchg_mb(&p->state, n, e);
+	if (a == e)
+	    break;
+	if (a == n && enqueue == ERTS_ENQUEUE_NOT)
+	    goto cleanup;
+    }
+
+    if (prof_runnable_procs) {
+
+	if (!(a & (ERTS_PSFLG_ACTIVE_SYS
+		   | ERTS_PSFLG_RUNNING
+		   | ERTS_PSFLG_RUNNING_SYS
+		   | ERTS_PSFLG_DIRTY_RUNNING
+		   | ERTS_PSFLG_DIRTY_RUNNING_SYS))
+	    && (!(a & ERTS_PSFLG_ACTIVE) || (a & ERTS_PSFLG_SUSPENDED))) {
+	    /* We activated a prevously inactive process */
+	    profile_runnable_proc(p, am_active);
+	}
+
+	erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	slocked = 0;
+    }
+
+    add2runq(enqueue, enq_prio, p, n, NULL);
+
+cleanup:
+
+    if (slocked)
+	erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+
+    ERTS_LC_ASSERT(!(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p)));
+
+    return n;
+}
+
+erts_aint32_t
+erts_proc_sys_schedule(Process *p, erts_aint32_t state, erts_aint32_t enable_flag)
+{
+    /* We are not allowed to call this function with status lock held... */
+    return active_sys_enqueue(p, state, enable_flag, 0);
+}
+
 static int
 schedule_process_sys_task(Process *p, erts_aint32_t prio, ErtsProcSysTask *st,
 			  erts_aint32_t *fail_state_p)
@@ -7165,19 +6751,16 @@ schedule_process_sys_task(Process *p, erts_aint32_t prio, ErtsProcSysTask *st,
     int res;
     int locked;
     ErtsProcSysTaskQs *stqs, *free_stqs;
-    erts_aint32_t fail_state, state, a, n, enq_prio;
-    int enqueue; /* < 0 -> use proxy */
-    unsigned int prof_runnable_procs;
+    erts_aint32_t fail_state, state;
 
     fail_state = *fail_state_p;
 
     res = 1; /* prepare for success */
     st->next = st->prev = st; /* Prep for empty prio queue */
-    state = erts_smp_atomic32_read_nob(&p->state);
-    prof_runnable_procs = erts_system_profile_flags.runnable_procs;
+    state = erts_atomic32_read_nob(&p->state);
     locked = 0;
     free_stqs = NULL;
-    if (state & ERTS_PSFLG_ACTIVE_SYS)
+    if (state & ERTS_PSFLG_SYS_TASKS)
 	stqs = NULL;
     else {
     alloc_qs:
@@ -7193,10 +6776,11 @@ schedule_process_sys_task(Process *p, erts_aint32_t prio, ErtsProcSysTask *st,
 
     if (!locked) {
 	locked = 1;
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 
-	state = erts_smp_atomic32_read_nob(&p->state);
+	state = erts_atomic32_read_nob(&p->state);
 	if (state & fail_state) {
+            erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 	    *fail_state_p = (state & fail_state);
 	    free_stqs = stqs;
 	    res = 0;
@@ -7239,74 +6823,19 @@ schedule_process_sys_task(Process *p, erts_aint32_t prio, ErtsProcSysTask *st,
 	    n = e = a;
 	    n &= ~ERTS_PSFLGS_ACT_PRIO_MASK;
 	    n |= (prio << ERTS_PSFLGS_ACT_PRIO_OFFSET);
-	    a = erts_smp_atomic32_cmpxchg_nob(&p->state, n, e);    
+	    a = erts_atomic32_cmpxchg_nob(&p->state, n, e);
 	} while (a != e);
 	state = n;
     }
 
-
-    a = state;
-    enq_prio = -1;
-
-    /* Status lock prevents out of order "runnable proc" trace msgs */
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
-
-    if (!prof_runnable_procs) {
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
-	locked = 0;
-    }
-
-    ASSERT(!(state & ERTS_PSFLG_PROXY));
-
-    while (1) {
-	erts_aint32_t e;
-	n = e = a;
-
-	if (a & ERTS_PSFLG_FREE)
-	    goto cleanup; /* We don't want to schedule free processes... */
-
-	enqueue = ERTS_ENQUEUE_NOT;
-	n |= ERTS_PSFLG_ACTIVE_SYS;
-	if (!(a & (ERTS_PSFLG_RUNNING
-		   | ERTS_PSFLG_RUNNING_SYS
-		   | ERTS_PSFLG_DIRTY_RUNNING
-		   | ERTS_PSFLG_DIRTY_RUNNING_SYS)))
-	    enqueue = check_enqueue_in_prio_queue(p, &enq_prio, &n, a);
-	a = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
-	if (a == e)
-	    break;
-	if (a == n && enqueue == ERTS_ENQUEUE_NOT)
-	    goto cleanup;
-    }
-
-    if (prof_runnable_procs) {
-
-	if (!(a & (ERTS_PSFLG_ACTIVE_SYS
-		   | ERTS_PSFLG_RUNNING
-		   | ERTS_PSFLG_RUNNING_SYS
-		   | ERTS_PSFLG_DIRTY_RUNNING
-		   | ERTS_PSFLG_DIRTY_RUNNING_SYS))
-	    && (!(a & ERTS_PSFLG_ACTIVE) || (a & ERTS_PSFLG_SUSPENDED))) {
-	    /* We activated a prevously inactive process */
-	    profile_runnable_proc(p, am_active);
-	}
-
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
-	locked = 0;
-    }
-
-    add2runq(enqueue, enq_prio, p, n, NULL);
+    /* active_sys_enqueue() always return with status lock unlocked */
+    (void) active_sys_enqueue(p, state, ERTS_PSFLG_SYS_TASKS, locked);
 
 cleanup:
 
-    if (locked)
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
-
     if (free_stqs)
 	proc_sys_task_queues_free(free_stqs);
 
-    ERTS_SMP_LC_ASSERT(!(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p)));
-
     return res;
 }
 
@@ -7315,15 +6844,15 @@ suspend_process(Process *c_p, Process *p)
 {
     erts_aint32_t state;
     int suspended = 0;
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
 
-    state = erts_smp_atomic32_read_acqb(&p->state);
+    state = erts_atomic32_read_acqb(&p->state);
 
     if ((state & ERTS_PSFLG_SUSPENDED))
 	suspended = -1;
     else {
 	if (c_p == p) {
-	    state = erts_smp_atomic32_read_bor_relb(&p->state,
+	    state = erts_atomic32_read_bor_relb(&p->state,
 						    ERTS_PSFLG_SUSPENDED);
 	    ASSERT(state & (ERTS_PSFLG_RUNNING
 			    | ERTS_PSFLG_RUNNING_SYS
@@ -7339,7 +6868,7 @@ suspend_process(Process *c_p, Process *p)
 
 		n = e = state;
 		n |= ERTS_PSFLG_SUSPENDED;
-		state = erts_smp_atomic32_cmpxchg_relb(&p->state, n, e);
+		state = erts_atomic32_cmpxchg_relb(&p->state, n, e);
 		if (state == e) {
 		    suspended = 1;
 		    break;
@@ -7384,14 +6913,14 @@ resume_process(Process *p, ErtsProcLocks locks)
     erts_aint32_t state, enq_prio = -1;
     int enqueue;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
 
     ASSERT(p->rcount > 0);
 
     if (--p->rcount > 0)  /* multiple suspend */
 	return;
 
-    state = erts_smp_atomic32_read_nob(&p->state);
+    state = erts_atomic32_read_nob(&p->state);
     enqueue = change_proc_schedule_state(p,
 					 ERTS_PSFLG_SUSPENDED,
 					 0,
@@ -7401,7 +6930,6 @@ resume_process(Process *p, ErtsProcLocks locks)
     add2runq(enqueue, enq_prio, p, state, NULL);
 }
 
-#ifdef ERTS_SMP
 
 static ERTS_INLINE void
 sched_resume_wake__(ErtsSchedulerSleepInfo *ssi)
@@ -7412,7 +6940,7 @@ sched_resume_wake__(ErtsSchedulerSleepInfo *ssi)
 			   | ERTS_SSI_FLG_SUSPENDED);
     erts_aint32_t oflgs;
     do {
-	oflgs = erts_smp_atomic32_cmpxchg_relb(&ssi->flags, 0, xflgs);
+	oflgs = erts_atomic32_cmpxchg_relb(&ssi->flags, 0, xflgs);
 	if (oflgs == xflgs) {
 	    erts_sched_finish_poke(ssi, oflgs);
 	    break;
@@ -7427,7 +6955,6 @@ nrml_sched_ix_resume_wake(Uint ix)
     sched_resume_wake__(ERTS_SCHED_SLEEP_INFO_IX(ix));
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static void
 dcpu_sched_ix_resume_wake(Uint ix)
@@ -7441,7 +6968,6 @@ dio_sched_ix_resume_wake(Uint ix)
     sched_resume_wake__(ERTS_DIRTY_IO_SCHED_SLEEP_INFO_IX(ix));
 }
 
-#endif
 
 static erts_aint32_t
 sched_prep_spin_suspended(ErtsSchedulerSleepInfo *ssi, erts_aint32_t xpct)
@@ -7453,7 +6979,7 @@ sched_prep_spin_suspended(ErtsSchedulerSleepInfo *ssi, erts_aint32_t xpct)
     erts_aint32_t xflgs = xpct;
 
     do {
-	oflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return nflgs;
 	xflgs = oflgs;
@@ -7470,7 +6996,7 @@ sched_spin_suspended(ErtsSchedulerSleepInfo *ssi, int spincount)
     erts_aint32_t flgs;
 
     do {
-	flgs = erts_smp_atomic32_read_acqb(&ssi->flags);
+	flgs = erts_atomic32_read_acqb(&ssi->flags);
 	if ((flgs & (ERTS_SSI_FLG_SLEEPING
 		     | ERTS_SSI_FLG_WAITING
 		     | ERTS_SSI_FLG_SUSPENDED))
@@ -7489,21 +7015,23 @@ sched_spin_suspended(ErtsSchedulerSleepInfo *ssi, int spincount)
 }
 
 static erts_aint32_t
-sched_set_suspended_sleeptype(ErtsSchedulerSleepInfo *ssi)
+sched_set_suspended_sleeptype(ErtsSchedulerSleepInfo *ssi,
+                              erts_aint32_t sleep_type)
 {
     erts_aint32_t oflgs;
-    erts_aint32_t nflgs = (ERTS_SSI_FLG_SLEEPING
-			   | ERTS_SSI_FLG_TSE_SLEEPING
-			   | ERTS_SSI_FLG_WAITING
-			   | ERTS_SSI_FLG_SUSPENDED);
+    erts_aint32_t nflgs = ((ERTS_SSI_FLG_SLEEPING
+                            | ERTS_SSI_FLG_WAITING
+                            | ERTS_SSI_FLG_SUSPENDED)
+                           | sleep_type);
     erts_aint32_t xflgs = (ERTS_SSI_FLG_SLEEPING
 			   | ERTS_SSI_FLG_WAITING
 			   | ERTS_SSI_FLG_SUSPENDED);
 
+    ASSERT(sleep_type == ERTS_SSI_FLG_TSE_SLEEPING);
     erts_tse_reset(ssi->event);
 
     while (1) {
-	oflgs = erts_smp_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
+	oflgs = erts_atomic32_cmpxchg_acqb(&ssi->flags, nflgs, xflgs);
 	if (oflgs == xflgs)
 	    return nflgs;
 	if ((oflgs & (ERTS_SSI_FLG_SLEEPING
@@ -7521,12 +7049,11 @@ sched_set_suspended_sleeptype(ErtsSchedulerSleepInfo *ssi)
 static void
 init_scheduler_suspend(void)
 {
-    erts_smp_mtx_init(&schdlr_sspnd.mtx, "schdlr_sspnd", NIL,
+    erts_mtx_init(&schdlr_sspnd.mtx, "schdlr_sspnd", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_SCHEDULER);
     schdlr_sspnd.online.normal = 1;
     schdlr_sspnd.curr_online.normal = 1;
     schdlr_sspnd.active.normal = 1;
-#ifdef ERTS_DIRTY_SCHEDULERS
     schdlr_sspnd.online.dirty_cpu = 0;
     schdlr_sspnd.curr_online.dirty_cpu = 0;
     schdlr_sspnd.active.dirty_cpu = 0;
@@ -7534,8 +7061,7 @@ init_scheduler_suspend(void)
     schdlr_sspnd.curr_online.dirty_io = 0;
     schdlr_sspnd.active.dirty_io = 0;
     schdlr_sspnd.last_msb_dirty_type = ERTS_SCHED_DIRTY_IO;
-#endif
-    erts_smp_atomic32_init_nob(&schdlr_sspnd.changing, 0);
+    erts_atomic32_init_nob(&schdlr_sspnd.changing, 0);
     schdlr_sspnd.chngq = NULL;
     schdlr_sspnd.changer = am_false;
     schdlr_sspnd.nmsb.ongoing = 0;
@@ -7566,7 +7092,7 @@ schdlr_sspnd_resume_proc(ErtsSchedType sched_type, Eterm pid)
                            : 0));
     if (p) {
 	resume_process(p, ERTS_PROC_LOCK_STATUS);
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
         if (sched_type != ERTS_SCHED_NORMAL)
             erts_proc_dec_refc(p);
     }
@@ -7595,7 +7121,6 @@ schdlr_sspnd_resume_procs(ErtsSchedType sched_type,
     }
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 static ERTS_INLINE int
 have_dirty_work(void)
@@ -7755,17 +7280,6 @@ msb_scheduler_type_switch(ErtsSchedType sched_type,
     if (exec_type != ERTS_SCHED_NORMAL)
         schdlr_sspnd.last_msb_dirty_type = exec_type;
     else {
-        erts_aint32_t calls;
-        /*
-         * Going back to normal scheduler after
-         * dirty execution; make sure it will check
-         * for I/O...
-         */
-        if (ERTS_USE_MODIFIED_TIMING())
-            calls = ERTS_MODIFIED_TIMING_INPUT_REDS + 1;
-        else
-            calls = INPUT_REDUCTIONS + 1;
-        erts_smp_atomic32_set_nob(&function_calls, calls);
 
         if ((nrml_prio == ERTS_MSB_NONE_PRIO_BIT)
             & ((dcpu_prio != ERTS_MSB_NONE_PRIO_BIT)
@@ -7800,7 +7314,7 @@ msb_scheduler_type_switch(ErtsSchedType sched_type,
 #else
     (void)
 #endif
-        erts_smp_atomic32_read_bset_mb(&esdp->ssi->flags,
+        erts_atomic32_read_bset_mb(&esdp->ssi->flags,
                                        (ERTS_SSI_FLG_SUSPENDED
                                         | ERTS_SSI_FLG_MSB_EXEC),
                                        ERTS_SSI_FLG_SUSPENDED);
@@ -7827,7 +7341,7 @@ msb_scheduler_type_switch(ErtsSchedType sched_type,
 #else
     (void)
 #endif
-        erts_smp_atomic32_read_bset_mb(&exec_rq->scheduler->ssi->flags,
+        erts_atomic32_read_bset_mb(&exec_rq->scheduler->ssi->flags,
                                        (ERTS_SSI_FLG_SUSPENDED
                                         | ERTS_SSI_FLG_MSB_EXEC),
                                        ERTS_SSI_FLG_MSB_EXEC);
@@ -7839,7 +7353,34 @@ msb_scheduler_type_switch(ErtsSchedType sched_type,
 
 }
 
-#endif
+static ERTS_INLINE void
+suspend_normal_scheduler_sleep(ErtsSchedulerData *esdp)
+{
+    ErtsSchedulerSleepInfo *ssi = esdp->ssi;
+    erts_aint32_t flgs = sched_spin_suspended(ssi,
+                                              ERTS_SCHED_SUSPEND_SLEEP_SPINCOUNT);
+    if (flgs == (ERTS_SSI_FLG_SLEEPING
+                 | ERTS_SSI_FLG_WAITING
+                 | ERTS_SSI_FLG_SUSPENDED)) {
+        flgs = sched_set_suspended_sleeptype(ssi, ERTS_SSI_FLG_TSE_SLEEPING);
+        if (flgs == (ERTS_SSI_FLG_SLEEPING
+                     | ERTS_SSI_FLG_TSE_SLEEPING
+                     | ERTS_SSI_FLG_WAITING
+                     | ERTS_SSI_FLG_SUSPENDED)) {
+            int res;
+
+            do {
+                res = erts_tse_wait(ssi->event);
+            } while (res == EINTR);
+        }
+    }
+}
+
+static ERTS_INLINE void
+suspend_dirty_scheduler_sleep(ErtsSchedulerData *esdp)
+{
+    suspend_normal_scheduler_sleep(esdp);
+}
 
 static void
 suspend_scheduler(ErtsSchedulerData *esdp)
@@ -7867,14 +7408,6 @@ suspend_scheduler(ErtsSchedulerData *esdp)
      */
 
 
-#if !defined(ERTS_DIRTY_SCHEDULERS)
-
-    sched_type = ERTS_SCHED_NORMAL;
-    online_flag = ERTS_SCHDLR_SSPND_CHNG_ONLN;
-    no = esdp->no;
-    ASSERT(no != 1);
-
-#else
 
     sched_type = esdp->type;
     switch (sched_type) {
@@ -7895,25 +7428,24 @@ suspend_scheduler(ErtsSchedulerData *esdp)
         return;
     }
 
-    if (erts_smp_atomic32_read_nob(&ssi->flags) & ERTS_SSI_FLG_MSB_EXEC) {
+    if (erts_atomic32_read_nob(&ssi->flags) & ERTS_SSI_FLG_MSB_EXEC) {
         ASSERT(no == 1);
         if (!msb_scheduler_type_switch(sched_type, esdp, no))
             return;
         /* Suspend and let scheduler 1 of another type execute... */
     }
 
-#endif
 
     if (sched_type != ERTS_SCHED_NORMAL) {
         dirty_active(esdp, -1);
-	erts_smp_runq_unlock(esdp->run_queue);
+	erts_runq_unlock(esdp->run_queue);
         dirty_sched_wall_time_change(esdp, 0);
     }
     else {
         if (no != 1)
             evacuate_run_queue(esdp->run_queue, &sbp);
 
-	erts_smp_runq_unlock(esdp->run_queue);
+	erts_runq_unlock(esdp->run_queue);
 
 	erts_sched_check_cpu_bind_prep_suspend(esdp);
 
@@ -7921,14 +7453,14 @@ suspend_scheduler(ErtsSchedulerData *esdp)
 	    profile_scheduler(make_small(esdp->no), am_inactive);
     }
 
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
 
     flgs = sched_prep_spin_suspended(ssi, ERTS_SSI_FLG_SUSPENDED);
     if (flgs & ERTS_SSI_FLG_SUSPENDED) {
 
 	schdlr_sspnd_dec_nscheds(&schdlr_sspnd.active, sched_type);
 
-	changing = erts_smp_atomic32_read_nob(&schdlr_sspnd.changing);
+	changing = erts_atomic32_read_nob(&schdlr_sspnd.changing);
 
 	while (1) {
 
@@ -7960,7 +7492,7 @@ suspend_scheduler(ErtsSchedulerData *esdp)
 
 		    if (clr_flg) {
 			ErtsProcList *plp, *end_plp;
-			changing = erts_smp_atomic32_read_band_nob(&schdlr_sspnd.changing,
+			changing = erts_atomic32_read_band_nob(&schdlr_sspnd.changing,
 								   ~clr_flg);
 			changing &= ~clr_flg;
 			(void) erts_proclist_fetch(&msb[i]->chngq, &end_plp);
@@ -8006,7 +7538,7 @@ suspend_scheduler(ErtsSchedulerData *esdp)
 			== schdlr_sspnd_get_nscheds(&schdlr_sspnd.curr_online,
 						    sched_type))) {
 		    ErtsProcList *plp;
-		    changing = erts_smp_atomic32_read_band_nob(&schdlr_sspnd.changing,
+		    changing = erts_atomic32_read_band_nob(&schdlr_sspnd.changing,
 							       ~online_flag);
 		    changing &= ~online_flag;
 		    if (sched_type == ERTS_SCHED_NORMAL) {
@@ -8028,23 +7560,24 @@ suspend_scheduler(ErtsSchedulerData *esdp)
 	    }
 
 	    if (curr_online) {
-		flgs = erts_smp_atomic32_read_acqb(&ssi->flags);
+		flgs = erts_atomic32_read_acqb(&ssi->flags);
 		if (!(flgs & ERTS_SSI_FLG_SUSPENDED))
 		    break;
 	    }
-	    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+	    erts_mtx_unlock(&schdlr_sspnd.mtx);
 
 	    schdlr_sspnd_resume_procs(sched_type, &resume);
 
 	    while (1) {
-		ErtsMonotonicTime current_time;
-		erts_aint32_t flgs;
-
 		if (sched_type != ERTS_SCHED_NORMAL)
-		    aux_work = 0;
-		else {
+                    suspend_dirty_scheduler_sleep(esdp);
+		else
+                {
+                    ErtsMonotonicTime current_time, timeout_time;
                     int evacuate = no == 1 ? 0 : !ERTS_EMPTY_RUNQ(esdp->run_queue);
 
+		    ASSERT(sched_type == ERTS_SCHED_NORMAL);
+
 		    aux_work = erts_atomic32_read_acqb(&ssi->aux_work);
 
 		    if (aux_work|evacuate) {
@@ -8060,123 +7593,62 @@ suspend_scheduler(ErtsSchedulerData *esdp)
 			if (aux_work && erts_thr_progress_update(esdp))
 			    erts_thr_progress_leader_update(esdp);
 			if (evacuate) {
-			    erts_smp_runq_lock(esdp->run_queue);
+			    erts_runq_lock(esdp->run_queue);
 			    evacuate_run_queue(esdp->run_queue, &sbp);
-			    erts_smp_runq_unlock(esdp->run_queue);
+			    erts_runq_unlock(esdp->run_queue);
 			}
 		    }
 
-		}
 
-		if (aux_work) {
-		    ASSERT(sched_type == ERTS_SCHED_NORMAL);
-		    current_time = erts_get_monotonic_time(esdp);
-		    if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref)) {
-			if (!thr_prgr_active) {
-			    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-			    sched_wall_time_change(esdp, 1);
-			}
-			erts_bump_timers(esdp->timer_wheel, current_time);
-		    }
-		}
-		else {
-		    ErtsMonotonicTime timeout_time;
-		    int do_timeout;
+                    if (aux_work)
+                        timeout_time = erts_next_timeout_time(esdp->next_tmo_ref);
+                    else
+                        timeout_time = erts_check_next_timeout_time(esdp);
 
-		    if (sched_type == ERTS_SCHED_NORMAL) {
-			timeout_time = erts_check_next_timeout_time(esdp);
-			current_time = erts_get_monotonic_time(esdp);
-			do_timeout = (current_time >= timeout_time);
-		    }
-		    else {
-			timeout_time = ERTS_MONOTONIC_TIME_MAX;
-			current_time = 0;
-			do_timeout = 0;
-		    }
+                    current_time = erts_get_monotonic_time(esdp);
 
-		    if (do_timeout) {
-			ASSERT(sched_type == ERTS_SCHED_NORMAL);
-			if (!thr_prgr_active) {
-			    erts_thr_progress_active(esdp, thr_prgr_active = 1);
-			    sched_wall_time_change(esdp, 1);
-			}
-		    }
-		    else {
-			if (sched_type == ERTS_SCHED_NORMAL) {
-			    if (thr_prgr_active) {
-				erts_thr_progress_active(esdp, thr_prgr_active = 0);
-				sched_wall_time_change(esdp, 0);
-			    }
-			    erts_thr_progress_prepare_wait(esdp);
-			}
-			flgs = sched_spin_suspended(ssi,
-						    ERTS_SCHED_SUSPEND_SLEEP_SPINCOUNT);
-			if (flgs == (ERTS_SSI_FLG_SLEEPING
-				     | ERTS_SSI_FLG_WAITING
-				     | ERTS_SSI_FLG_SUSPENDED)) {
-			    flgs = sched_set_suspended_sleeptype(ssi);
-			    if (flgs == (ERTS_SSI_FLG_SLEEPING
-					 | ERTS_SSI_FLG_TSE_SLEEPING
-					 | ERTS_SSI_FLG_WAITING
-					 | ERTS_SSI_FLG_SUSPENDED)) {
-				int res;
-
-				if (sched_type == ERTS_SCHED_NORMAL)
-				    current_time = erts_get_monotonic_time(esdp);
-				else
-				    current_time = 0;
-
-				do {
-				    Sint64 timeout;
-				    if (current_time >= timeout_time)
-					break;
-				    if (sched_type != ERTS_SCHED_NORMAL)
-					timeout = -1;
-				    else
-					timeout = ERTS_MONOTONIC_TO_NSEC(timeout_time
-									 - current_time
-									 - 1) + 1;
-				    res = erts_tse_twait(ssi->event, timeout);
-
-				    if (sched_type == ERTS_SCHED_NORMAL)
-					current_time = erts_get_monotonic_time(esdp);
-				    else
-					current_time = 0;
-
-				} while (res == EINTR);
-			    }
-			}
-			if (sched_type == ERTS_SCHED_NORMAL)
-			    erts_thr_progress_finalize_wait(esdp);
-		    }
+                    if (!aux_work && current_time < timeout_time) {
+                        /* go to sleep... */
+                        if (thr_prgr_active) {
+                            erts_thr_progress_active(esdp, thr_prgr_active = 0);
+                            sched_wall_time_change(esdp, 0);
+                        }
+                        erts_thr_progress_prepare_wait(NULL);
+                        suspend_normal_scheduler_sleep(esdp);
+                        erts_thr_progress_finalize_wait(NULL);
+                        current_time = erts_get_monotonic_time(esdp);
+                    }
 
-		    if (current_time >= timeout_time) {
-			ASSERT(sched_type == ERTS_SCHED_NORMAL);
-			erts_bump_timers(esdp->timer_wheel, current_time);
-		    }
-		}
+                    if (current_time >= timeout_time) {
+                        if (!thr_prgr_active) {
+                            erts_thr_progress_active(esdp, thr_prgr_active = 1);
+                            sched_wall_time_change(esdp, 1);
+                        }
+                        erts_bump_timers(esdp->timer_wheel, current_time);
+                    }
+                }
 
 		flgs = sched_prep_spin_suspended(ssi, (ERTS_SSI_FLG_WAITING
 						       | ERTS_SSI_FLG_SUSPENDED));
 		if (!(flgs & ERTS_SSI_FLG_SUSPENDED))
 		    break;
-		changing = erts_smp_atomic32_read_nob(&schdlr_sspnd.changing);
+		changing = erts_atomic32_read_nob(&schdlr_sspnd.changing);
 		if (changing)
 		    break;
 	    }
 
-	    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
-	    changing = erts_smp_atomic32_read_nob(&schdlr_sspnd.changing);
+	    erts_mtx_lock(&schdlr_sspnd.mtx);
+	    changing = erts_atomic32_read_nob(&schdlr_sspnd.changing);
 	}
 
 	schdlr_sspnd_inc_nscheds(&schdlr_sspnd.active, sched_type);
-	changing = erts_smp_atomic32_read_nob(&schdlr_sspnd.changing);
+	changing = erts_atomic32_read_nob(&schdlr_sspnd.changing);
         if (changing) {
             if ((changing & ERTS_SCHDLR_SSPND_CHNG_MSB)
                 && !schdlr_sspnd.msb.ongoing
                 && schdlr_sspnd_eq_nscheds(&schdlr_sspnd.online,
                                            &schdlr_sspnd.active)) {
-                erts_smp_atomic32_read_band_nob(&schdlr_sspnd.changing,
+                erts_atomic32_read_band_nob(&schdlr_sspnd.changing,
                                                 ~ERTS_SCHDLR_SSPND_CHNG_MSB);
             }
             if ((changing & ERTS_SCHDLR_SSPND_CHNG_NMSB)
@@ -8185,14 +7657,14 @@ suspend_scheduler(ErtsSchedulerData *esdp)
                                              ERTS_SCHED_NORMAL)
                     == schdlr_sspnd_get_nscheds(&schdlr_sspnd.active,
                                                 ERTS_SCHED_NORMAL))) {
-                erts_smp_atomic32_read_band_nob(&schdlr_sspnd.changing,
+                erts_atomic32_read_band_nob(&schdlr_sspnd.changing,
                                                 ~ERTS_SCHDLR_SSPND_CHNG_NMSB);
             }
         }
 	ASSERT(no <= schdlr_sspnd_get_nscheds(&schdlr_sspnd.online, sched_type));
     }
 
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
 
     schdlr_sspnd_resume_procs(sched_type, &resume);
 
@@ -8211,7 +7683,7 @@ suspend_scheduler(ErtsSchedulerData *esdp)
         }
     }
 
-    erts_smp_runq_lock(esdp->run_queue);
+    erts_runq_lock(esdp->run_queue);
     non_empty_runq(esdp->run_queue);
 
     if (sched_type != ERTS_SCHED_NORMAL)
@@ -8235,7 +7707,7 @@ erts_schedulers_state(Uint *total,
 {
     if (active || online || dirty_cpu_online
 	|| dirty_cpu_active || dirty_io_active) {
-	erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+	erts_mtx_lock(&schdlr_sspnd.mtx);
 	if (active)
 	    *active = schdlr_sspnd_get_nscheds(&schdlr_sspnd.active,
 					       ERTS_SCHED_NORMAL);
@@ -8251,7 +7723,7 @@ erts_schedulers_state(Uint *total,
 	if (dirty_io_active)
 	    *dirty_io_active = schdlr_sspnd_get_nscheds(&schdlr_sspnd.active,
 							ERTS_SCHED_DIRTY_IO);
-	erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+	erts_mtx_unlock(&schdlr_sspnd.mtx);
     }
 
     if (total)
@@ -8267,7 +7739,7 @@ abort_sched_onln_chng_waitq(Process *p)
 {
     Eterm resume = NIL;
 
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
 
 #ifdef DEBUG
     {
@@ -8317,7 +7789,7 @@ abort_sched_onln_chng_waitq(Process *p)
 	}
     }
 
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
 
     if (is_internal_pid(resume))
 	schdlr_sspnd_resume_proc(ERTS_SCHED_NORMAL, resume);
@@ -8334,11 +7806,7 @@ erts_set_schedulers_online(Process *p,
     erts_aint32_t changing = 0, change_flags;
     int online, increase;
     ErtsProcList *plp;
-#ifdef ERTS_DIRTY_SCHEDULERS
     int dirty_no, change_dirty, dirty_online;
-#else
-    ASSERT(!dirty_only);
-#endif
 
     if (new_no < 1)
 	return ERTS_SCHDLR_SSPND_EINVAL;
@@ -8347,11 +7815,9 @@ erts_set_schedulers_online(Process *p,
     else if (erts_no_schedulers < new_no)
 	return ERTS_SCHDLR_SSPND_EINVAL;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (dirty_only)
 	resume_proc = 0;
     else
-#endif
     {
 	resume_proc = 1;
 	/*
@@ -8360,23 +7826,21 @@ erts_set_schedulers_online(Process *p,
 	 * race...
 	 */
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	suspend_process(p, p);
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     }
 
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
 
     change_flags = 0;
     have_unlocked_plocks = 0;
     no = (int) new_no;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (!dirty_only)
-#endif
     {
-	changing = erts_smp_atomic32_read_nob(&schdlr_sspnd.changing);
+	changing = erts_atomic32_read_nob(&schdlr_sspnd.changing);
 	if (changing & ERTS_SCHDLR_SSPND_CHNG_ONLN) {
 	enqueue_wait:
 	    p->flags |= F_SCHDLR_ONLN_WAITQ;
@@ -8402,12 +7866,6 @@ erts_set_schedulers_online(Process *p,
 
     *old_no = online = schdlr_sspnd_get_nscheds(&schdlr_sspnd.online,
 						ERTS_SCHED_NORMAL);
-#ifndef ERTS_DIRTY_SCHEDULERS
-    if (no == online) {
-	res = ERTS_SCHDLR_SSPND_DONE;
-	goto done;
-    }
-#else /* ERTS_DIRTY_SCHEDULERS */
     dirty_online = schdlr_sspnd_get_nscheds(&schdlr_sspnd.online,
 					    ERTS_SCHED_DIRTY_CPU);
     if (dirty_only)
@@ -8459,7 +7917,6 @@ erts_set_schedulers_online(Process *p,
     if (dirty_only)
 	increase = (dirty_no > dirty_online);
     else
-#endif /* ERTS_DIRTY_SCHEDULERS */
     {
 	change_flags |= ERTS_SCHDLR_SSPND_CHNG_ONLN;
 	schdlr_sspnd_set_nscheds(&schdlr_sspnd.online,
@@ -8468,12 +7925,11 @@ erts_set_schedulers_online(Process *p,
 	increase = (no > online);
     }
 
-    erts_smp_atomic32_read_bor_nob(&schdlr_sspnd.changing, change_flags);
+    erts_atomic32_read_bor_nob(&schdlr_sspnd.changing, change_flags);
 
     res = ERTS_SCHDLR_SSPND_DONE;
     if (increase) {
 	int ix;
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (change_dirty) {
 	    ErtsSchedulerSleepInfo* ssi;
 	    if (schdlr_sspnd.msb.ongoing) {
@@ -8487,7 +7943,6 @@ erts_set_schedulers_online(Process *p,
 	    }
 	}
 	if (!dirty_only)
-#endif
 	{
 	    if (schdlr_sspnd.msb.ongoing|schdlr_sspnd.nmsb.ongoing) {
 		for (ix = online; ix < no; ix++)
@@ -8496,7 +7951,7 @@ erts_set_schedulers_online(Process *p,
 	    else {
 		if (plocks) {
 		    have_unlocked_plocks = 1;
-		    erts_smp_proc_unlock(p, plocks);
+		    erts_proc_unlock(p, plocks);
 		}
 		change_no_used_runqs(no);
 
@@ -8509,7 +7964,6 @@ erts_set_schedulers_online(Process *p,
 	}
     }
     else /* if decrease */ {
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (change_dirty) {
 	    if (schdlr_sspnd.msb.ongoing) {
 		for (ix = dirty_no; ix < dirty_online; ix++)
@@ -8527,7 +7981,6 @@ erts_set_schedulers_online(Process *p,
 	    }
 	}
 	if (!dirty_only)
-#endif
 	{
 	    if (schdlr_sspnd.msb.ongoing|schdlr_sspnd.nmsb.ongoing) {
 		for (ix = no; ix < online; ix++)
@@ -8536,7 +7989,7 @@ erts_set_schedulers_online(Process *p,
 	    else {
 		if (plocks) {
 		    have_unlocked_plocks = 1;
-		    erts_smp_proc_unlock(p, plocks);
+		    erts_proc_unlock(p, plocks);
 		}
 
 		change_no_used_runqs(no);
@@ -8565,17 +8018,17 @@ done:
 	   <= schdlr_sspnd_get_nscheds(&schdlr_sspnd.online,
 				       ERTS_SCHED_NORMAL));
 
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
 
     if (have_unlocked_plocks)
-	erts_smp_proc_lock(p, plocks);
+	erts_proc_lock(p, plocks);
 
     if (resume_proc) {
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	resume_process(p, plocks|ERTS_PROC_LOCK_STATUS);
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     }
 
     return res;
@@ -8613,13 +8066,13 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
     else {
 	resume_proc = 1;
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	suspend_process(p, p);
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     }
 
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
     if (on) { /* ------ BLOCK ------ */
 	if (msbp->chngq) {
 	    ASSERT(msbp->ongoing);
@@ -8649,12 +8102,12 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
 	    p->flags |= have_blckd_flg;
 	    if (plocks) {
 		have_unlocked_plocks = 1;
-		erts_smp_proc_unlock(p, plocks);
+		erts_proc_unlock(p, plocks);
 	    }
 	    ASSERT(!msbp->ongoing);
 	    msbp->ongoing = 1;
 
-            erts_smp_atomic32_read_bor_nob(&schdlr_sspnd.changing,
+            erts_atomic32_read_bor_nob(&schdlr_sspnd.changing,
                                            chng_flg);
             change_no_used_runqs(1);
             for (ix = 1; ix < erts_no_run_queues; ix++)
@@ -8665,21 +8118,19 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
                 wake_scheduler(rq);
             }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
             if (!normal) {
                 ERTS_RUNQ_FLGS_SET_NOB(ERTS_RUNQ_IX(0), ERTS_RUNQ_FLG_MSB_EXEC);
-                erts_smp_atomic32_read_bor_nob(&ERTS_RUNQ_IX(0)->scheduler->ssi->flags,
+                erts_atomic32_read_bor_nob(&ERTS_RUNQ_IX(0)->scheduler->ssi->flags,
                                                ERTS_SSI_FLG_MSB_EXEC);
                 for (ix = 0; ix < erts_no_dirty_cpu_schedulers; ix++)
                     dcpu_sched_ix_suspend_wake(ix);
                 for (ix = 0; ix < erts_no_dirty_io_schedulers; ix++)
                     dio_sched_ix_suspend_wake(ix);
             }
-#endif
 
         wait_until_msb:
 
-            ASSERT(chng_flg & erts_smp_atomic32_read_nob(&schdlr_sspnd.changing));
+            ASSERT(chng_flg & erts_atomic32_read_nob(&schdlr_sspnd.changing));
 
             plp = proclist_create(p);
             erts_proclist_store_last(&msbp->chngq, plp);
@@ -8720,14 +8171,14 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
 	}
 	if (!msbp->blckrs && !msbp->chngq) {
 	    int online;
-	    erts_smp_atomic32_read_bor_nob(&schdlr_sspnd.changing,
+	    erts_atomic32_read_bor_nob(&schdlr_sspnd.changing,
 					   chng_flg);
 	    p->flags &= ~have_blckd_flg;
 	    msbp->ongoing = 0;
             if (!(schdlr_sspnd.msb.ongoing|schdlr_sspnd.nmsb.ongoing)) {
                 if (plocks) {
 		    have_unlocked_plocks = 1;
-		    erts_smp_proc_unlock(p, plocks);
+		    erts_proc_unlock(p, plocks);
 		}
 
                 online = (int) schdlr_sspnd_get_nscheds(&schdlr_sspnd.online,
@@ -8741,7 +8192,6 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
 		for (ix = online; ix < erts_no_run_queues; ix++)
 		    suspend_run_queue(ERTS_RUNQ_IX(ix));
 	    }
-#ifdef ERTS_DIRTY_SCHEDULERS
             if (!schdlr_sspnd.msb.ongoing) {
                 /* Get rid of msb-exec flag in run-queue of scheduler 1 */
                 resume_run_queue(ERTS_RUNQ_IX(0));
@@ -8752,7 +8202,6 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
 		for (ix = 0; ix < erts_no_dirty_io_schedulers; ix++)
 		    dio_sched_ix_resume_wake(ix);
 	    }
-#endif
 	}
 
     unblock_res:
@@ -8764,17 +8213,17 @@ erts_block_multi_scheduling(Process *p, ErtsProcLocks plocks, int on, int normal
 	    res = ERTS_SCHDLR_SSPND_DONE;
     }
 
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
 
     if (have_unlocked_plocks)
-	erts_smp_proc_lock(p, plocks);
+	erts_proc_lock(p, plocks);
 
     if (resume_proc) {
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	resume_process(p, plocks|ERTS_PROC_LOCK_STATUS);
 	if (!(plocks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
     }
 
     return res;
@@ -8784,14 +8233,14 @@ int
 erts_is_multi_scheduling_blocked(void)
 {
     int res;
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
     if (schdlr_sspnd.msb.blckrs)
 	res = 1;
     else if (schdlr_sspnd.nmsb.blckrs)
 	res = -1;
     else
 	res = 0;
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
     return res;
 }
 
@@ -8803,7 +8252,7 @@ erts_multi_scheduling_blockers(Process *p, int normal)
 
     msbp = normal ? &schdlr_sspnd.nmsb : &schdlr_sspnd.msb;
 
-    erts_smp_mtx_lock(&schdlr_sspnd.mtx);
+    erts_mtx_lock(&schdlr_sspnd.mtx);
     if (!erts_proclist_is_empty(msbp->blckrs)) {
 	Eterm *hp, *hp_end;
 	ErtsProcList *plp1, *plp2;
@@ -8831,7 +8280,7 @@ erts_multi_scheduling_blockers(Process *p, int normal)
 	}
 	HRelease(p, hp_end, hp);
     }
-    erts_smp_mtx_unlock(&schdlr_sspnd.mtx);
+    erts_mtx_unlock(&schdlr_sspnd.mtx);
     return res;
 }
 
@@ -8841,10 +8290,9 @@ sched_thread_func(void *vesdp)
     ErtsThrPrgrCallbacks callbacks;
     ErtsSchedulerData *esdp = vesdp;
     Uint no = esdp->no;
-#ifdef ERTS_SMP
     erts_tse_t *tse;
-#endif
 
+    erts_port_task_pre_alloc_init_thread();
     erts_sched_init_time_sup(esdp);
 
     if (no == 1)
@@ -8853,7 +8301,6 @@ sched_thread_func(void *vesdp)
     (void) ERTS_RUNQ_FLGS_SET_NOB(esdp->run_queue,
 				  ERTS_RUNQ_FLG_EXEC);
 
-#ifdef ERTS_SMP
     tse = erts_tse_fetch();
     erts_tse_prepare_timed(tse);
     ERTS_SCHED_SLEEP_INFO_IX(no - 1)->event = tse;
@@ -8867,7 +8314,6 @@ sched_thread_func(void *vesdp)
 
     erts_thr_progress_register_managed_thread(esdp, &callbacks, 0);
     erts_alloc_register_scheduler(vesdp);
-#endif
 #ifdef ERTS_ENABLE_LOCK_CHECK
     {
 	char buf[31];
@@ -8876,18 +8322,14 @@ sched_thread_func(void *vesdp)
     }
 #endif
     erts_tsd_set(sched_data_key, vesdp);
-#ifdef ERTS_SMP
 #if HAVE_ERTS_MSEG
     erts_mseg_late_init();
 #endif
-#if ERTS_USE_ASYNC_READY_Q
     esdp->aux_work_data.async_ready.queue = erts_get_async_ready_queue(no);
-#endif
 
     erts_sched_init_check_cpu_bind(esdp);
 
     erts_proc_lock_prepare_proc_lock_waiter();
-#endif
 
 #ifdef HIPE
     hipe_thread_signal_init();
@@ -8901,6 +8343,7 @@ sched_thread_func(void *vesdp)
     ERTS_VERIFY_UNUSED_TEMP_ALLOC(NULL);
 #endif
 
+    erts_alcu_sched_spec_data_init(esdp);
     erts_ets_sched_spec_data_init(esdp);
 
     process_main(esdp->x_reg_array, esdp->f_reg_array);
@@ -8912,8 +8355,6 @@ sched_thread_func(void *vesdp)
     return NULL;
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-#ifdef ERTS_SMP
 static void*
 sched_dirty_cpu_thread_func(void *vesdp)
 {
@@ -8943,9 +8384,7 @@ sched_dirty_cpu_thread_func(void *vesdp)
     }
 #endif
     erts_tsd_set(sched_data_key, vesdp);
-#if ERTS_USE_ASYNC_READY_Q
     esdp->aux_work_data.async_ready.queue = NULL;
-#endif
 
     erts_proc_lock_prepare_proc_lock_waiter();
 
@@ -8991,9 +8430,7 @@ sched_dirty_io_thread_func(void *vesdp)
     }
 #endif
     erts_tsd_set(sched_data_key, vesdp);
-#if ERTS_USE_ASYNC_READY_Q
     esdp->aux_work_data.async_ready.queue = NULL;
-#endif
 
     erts_proc_lock_prepare_proc_lock_waiter();
 
@@ -9009,40 +8446,38 @@ sched_dirty_io_thread_func(void *vesdp)
 	     no);
     return NULL;
 }
-#endif
-#endif
-
-static ethr_tid aux_tid;
 
 void
 erts_start_schedulers(void)
 {
+    ethr_tid tid;
     int res = 0;
     Uint actual;
     Uint wanted = erts_no_schedulers;
     Uint wanted_no_schedulers = erts_no_schedulers;
     char name[16];
     ethr_thr_opts opts = ETHR_THR_OPTS_DEFAULT_INITER;
+    int ix;
 
     opts.detached = 1;
 
     opts.name = name;
 
-#ifdef ERTS_SMP
     if (erts_runq_supervision_interval) {
 	opts.suggested_stack_size = 16;
         erts_snprintf(opts.name, 16, "runq_supervisor");
 	erts_atomic_init_nob(&runq_supervisor_sleeping, 0);
 	if (0 != ethr_event_init(&runq_supervision_event))
 	    erts_exit(ERTS_ERROR_EXIT, "Failed to create run-queue supervision event\n");
-	if (0 != ethr_thr_create(&runq_supervisor_tid,
-				 runq_supervisor,
-				 NULL,
-				 &opts))
-	    erts_exit(ERTS_ERROR_EXIT, "Failed to create run-queue supervision thread\n");
+        res = ethr_thr_create(&runq_supervisor_tid,
+                              runq_supervisor,
+                              NULL,
+                              &opts);
+	if (0 != res)
+	    erts_exit(ERTS_ERROR_EXIT, "Failed to create run-queue supervision thread, "
+                      "error = %d\n", res);
 
     }
-#endif
 
     opts.suggested_stack_size = erts_sched_thread_suggested_stack_size;
 
@@ -9068,17 +8503,14 @@ erts_start_schedulers(void)
     }
     erts_no_schedulers = actual;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-#ifdef ERTS_SMP
     {
-	int ix;
 	for (ix = 0; ix < erts_no_dirty_cpu_schedulers; ix++) {
 	    ErtsSchedulerData *esdp = ERTS_DIRTY_CPU_SCHEDULER_IX(ix);
 	    erts_snprintf(opts.name, 16, "%d_dirty_cpu_scheduler", ix + 1);
             opts.suggested_stack_size = erts_dcpu_sched_thread_suggested_stack_size;
 	    res = ethr_thr_create(&esdp->tid,sched_dirty_cpu_thread_func,(void*)esdp,&opts);
 	    if (res != 0)
-		erts_exit(ERTS_ERROR_EXIT, "Failed to create dirty cpu scheduler thread %d\n", ix);
+		erts_exit(ERTS_ERROR_EXIT, "Failed to create dirty cpu scheduler thread %d, error = %d\n", ix, res);
 	}
 	for (ix = 0; ix < erts_no_dirty_io_schedulers; ix++) {
 	    ErtsSchedulerData *esdp = ERTS_DIRTY_IO_SCHEDULER_IX(ix);
@@ -9086,19 +8518,25 @@ erts_start_schedulers(void)
             opts.suggested_stack_size = erts_dio_sched_thread_suggested_stack_size;
 	    res = ethr_thr_create(&esdp->tid,sched_dirty_io_thread_func,(void*)esdp,&opts);
 	    if (res != 0)
-		erts_exit(ERTS_ERROR_EXIT, "Failed to create dirty io scheduler thread %d\n", ix);
+		erts_exit(ERTS_ERROR_EXIT, "Failed to create dirty io scheduler thread %d, error = %d\n", ix, res);
 	}
     }
-#endif
-#endif
 
     ERTS_THR_MEMORY_BARRIER;
 
     erts_snprintf(opts.name, 16, "aux");
 
-    res = ethr_thr_create(&aux_tid, aux_thread, NULL, &opts);
+    res = ethr_thr_create(&tid, aux_thread, NULL, &opts);
     if (res != 0)
-	erts_exit(ERTS_ERROR_EXIT, "Failed to create aux thread\n");
+	erts_exit(ERTS_ERROR_EXIT, "Failed to create aux thread, error = %d\n", res);
+
+    for (ix = 0; ix < erts_no_poll_threads; ix++) {
+        erts_snprintf(opts.name, 16, "%d_poller", ix);
+
+        res = ethr_thr_create(&tid, poll_thread, (void*)(UWord)ix, &opts);
+        if (res != 0)
+            erts_exit(ERTS_ERROR_EXIT, "Failed to create poll thread\n");
+    }
 
     if (actual < 1)
 	erts_exit(ERTS_ERROR_EXIT,
@@ -9117,416 +8555,22 @@ erts_start_schedulers(void)
     }
 }
 
-#endif /* ERTS_SMP */
-
-#ifdef ERTS_SMP
-
-static void
-add_pend_suspend(Process *suspendee,
-		 Eterm originator_pid,
-		 void (*handle_func)(Process *,
-				     ErtsProcLocks,
-				     int,
-				     Eterm))
-{
-    ErtsPendingSuspend *psp = erts_alloc(ERTS_ALC_T_PEND_SUSPEND,
-					 sizeof(ErtsPendingSuspend));
-    psp->next = NULL;
-#ifdef DEBUG
-#if defined(ARCH_64)
-    psp->end = (ErtsPendingSuspend *) 0xdeaddeaddeaddead;
-#else
-    psp->end = (ErtsPendingSuspend *) 0xdeaddead;
-#endif
-#endif
-    psp->pid = originator_pid;
-    psp->handle_func = handle_func;
-
-    if (suspendee->pending_suspenders)
-	suspendee->pending_suspenders->end->next = psp;
-    else
-	suspendee->pending_suspenders = psp;
-    suspendee->pending_suspenders->end = psp;
-}
-
-static void
-handle_pending_suspend(Process *p, ErtsProcLocks p_locks)
-{
-    ErtsPendingSuspend *psp;
-    int is_alive = !ERTS_PROC_IS_EXITING(p);
-
-    ERTS_SMP_LC_ASSERT(p_locks & ERTS_PROC_LOCK_STATUS);
-
-    /*
-     * New pending suspenders might appear while we are processing
-     * (since we may release the status lock on p while processing).
-     */
-    while (p->pending_suspenders) {
-	psp = p->pending_suspenders;
-	p->pending_suspenders = NULL;
-	while (psp) {
-	    ErtsPendingSuspend *free_psp;
-	    (*psp->handle_func)(p, p_locks, is_alive, psp->pid);
-	    free_psp = psp;
-	    psp = psp->next;
-	    erts_free(ERTS_ALC_T_PEND_SUSPEND, (void *) free_psp);
-	}
-    }
-    
-}
-
-static ERTS_INLINE void
-cancel_suspend_of_suspendee(Process *p, ErtsProcLocks p_locks)
-{
-    if (is_not_nil(p->suspendee)) {
-	Process *rp;
-	if (!(p_locks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
-	rp = erts_pid2proc(p, p_locks|ERTS_PROC_LOCK_STATUS,
-			   p->suspendee, ERTS_PROC_LOCK_STATUS);
-	if (rp) {
-	    erts_resume(rp, ERTS_PROC_LOCK_STATUS);
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-	}
-	if (!(p_locks & ERTS_PROC_LOCK_STATUS))
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
-	p->suspendee = NIL;
-    }
-}
-
-static void
-handle_pend_sync_suspend(Process *suspendee,
-			 ErtsProcLocks suspendee_locks,
-			 int suspendee_alive,
-			 Eterm suspender_pid)
-{
-    Process *suspender;
-
-    ERTS_SMP_LC_ASSERT(suspendee_locks & ERTS_PROC_LOCK_STATUS);
-
-    suspender = erts_pid2proc(suspendee,
-			      suspendee_locks,
-			      suspender_pid,
-			      ERTS_PROC_LOCK_STATUS);
-    if (suspender) {
-	ASSERT(is_nil(suspender->suspendee));
-	if (suspendee_alive) {
-	    erts_suspend(suspendee, suspendee_locks, NULL);
-	    suspender->suspendee = suspendee->common.id;
-	}
-	/* suspender is suspended waiting for suspendee to suspend;
-	   resume suspender */
-	ASSERT(suspendee != suspender);
-	resume_process(suspender, ERTS_PROC_LOCK_STATUS);
-	erts_smp_proc_unlock(suspender, ERTS_PROC_LOCK_STATUS);
-    }
-}
-
-static Process *
-pid2proc_not_running(Process *c_p, ErtsProcLocks c_p_locks,
-		     Eterm pid, ErtsProcLocks pid_locks, int suspend)
-{
-    Process *rp;
-    int unlock_c_p_status;
-
-    ERTS_SMP_LC_ASSERT(c_p_locks == erts_proc_lc_my_proc_locks(c_p));
-
-    ERTS_SMP_LC_ASSERT(c_p_locks & ERTS_PROC_LOCK_MAIN);
-    ERTS_SMP_LC_ASSERT(pid_locks & (ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS));
-
-    if (c_p->common.id == pid)
-	return erts_pid2proc(c_p, c_p_locks, pid, pid_locks);
-
-    if (c_p_locks & ERTS_PROC_LOCK_STATUS)
-	unlock_c_p_status = 0;
-    else {
-	unlock_c_p_status = 1;
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
-    }
-
-    if (c_p->suspendee == pid) {
-	/* Process previously suspended by c_p (below)... */
-	ErtsProcLocks rp_locks = pid_locks|ERTS_PROC_LOCK_STATUS;
-	rp = erts_pid2proc(c_p, c_p_locks|ERTS_PROC_LOCK_STATUS, pid, rp_locks);
-	c_p->suspendee = NIL;
-	ASSERT(c_p->flags & F_P2PNR_RESCHED);
-	c_p->flags &= ~F_P2PNR_RESCHED;
-	if (!suspend && rp)
-	    resume_process(rp, rp_locks);
-    }
-    else {
-	rp = erts_pid2proc(c_p, c_p_locks|ERTS_PROC_LOCK_STATUS,
-			   pid, ERTS_PROC_LOCK_STATUS);
-
-	if (!rp) {
-	    c_p->flags &= ~F_P2PNR_RESCHED;
-	    goto done;
-	}
-
-	ASSERT(!(c_p->flags & F_P2PNR_RESCHED));
-
-	/*
-	 * Suspend the other process in order to prevent
-	 * it from being selected for normal execution.
-	 * This will however not prevent it from being
-	 * selected for execution of a system task. If
-	 * it is selected for execution of a system task
-	 * we might be blocked for quite a while if the
-	 * try-lock below fails. That is, there is room
-	 * for improvement here...
-	 */
-
-	if (!suspend_process(c_p, rp)) {
-	    /* Other process running */
-
-	    ASSERT((ERTS_PSFLG_RUNNING | ERTS_PSFLG_DIRTY_RUNNING)
-		   & erts_smp_atomic32_read_nob(&rp->state));
-
-#ifdef ERTS_DIRTY_SCHEDULERS
-	    if (!suspend
-		&& (erts_smp_atomic32_read_nob(&rp->state)
-		    & ERTS_PSFLG_DIRTY_RUNNING)) {
-		ErtsProcLocks need_locks = pid_locks & ~ERTS_PROC_LOCK_STATUS;
-		if (need_locks && erts_smp_proc_trylock(rp, need_locks) == EBUSY) {
-		    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-		    rp = erts_pid2proc(c_p, c_p_locks|ERTS_PROC_LOCK_STATUS,
-				       pid, pid_locks|ERTS_PROC_LOCK_STATUS);
-		}
-		goto done;
-	    }
-#endif
-
-	running:
-
-	    /*
-	     * If we got pending suspenders and suspend ourselves waiting
-	     * to suspend another process we might deadlock.
-	     * In this case we have to yield, be suspended by
-	     * someone else and then do it all over again.
-	     */
-	    if (!c_p->pending_suspenders) {
-		/* Mark rp pending for suspend by c_p */
-		add_pend_suspend(rp, c_p->common.id, handle_pend_sync_suspend);
-		ASSERT(is_nil(c_p->suspendee));
-
-		/* Suspend c_p; when rp is suspended c_p will be resumed. */
-		suspend_process(c_p, c_p);
-		c_p->flags |= F_P2PNR_RESCHED;
-	    }
-	    /* Yield (caller is assumed to yield immediately in bif). */
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-	    rp = ERTS_PROC_LOCK_BUSY;
-	}
-	else {
-	    ErtsProcLocks need_locks = pid_locks & ~ERTS_PROC_LOCK_STATUS;
-	    if (need_locks && erts_smp_proc_trylock(rp, need_locks) == EBUSY) {
-		if ((ERTS_PSFLG_RUNNING_SYS|ERTS_PSFLG_DIRTY_RUNNING_SYS)
-		    & erts_smp_atomic32_read_nob(&rp->state)) {
-		    /* Executing system task... */
-		    resume_process(rp, ERTS_PROC_LOCK_STATUS);
-		    goto running;
-		}
-		erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-		/*
-		 * If we are unlucky, the process just got selected for
-		 * execution of a system task. In this case we may be
-		 * blocked here for quite a while... Execution of system
-		 * tasks are fortunately quite rare events. We try to
-		 * avoid this by checking if it is in a state executing
-		 * system tasks (above), but it will not prevent all
-		 * scenarios for a long block here...
-		 */
-		rp = erts_pid2proc(c_p, c_p_locks|ERTS_PROC_LOCK_STATUS,
-				   pid, pid_locks|ERTS_PROC_LOCK_STATUS);
-		if (!rp)
-		    goto done;
-	    }
-
-	    /*
-	     * The previous suspend has prevented the process
-	     * from being selected for normal execution regardless
-	     * of locks held or not held on it...
-	     */
-#ifdef DEBUG
-	    {
-		erts_aint32_t state;
-		state = erts_smp_atomic32_read_nob(&rp->state);
-		ASSERT((state & ERTS_PSFLG_PENDING_EXIT)
-		       || !(state & ERTS_PSFLG_RUNNING));
-	    }
-#endif
-
-	    if (!suspend)
-		resume_process(rp, pid_locks|ERTS_PROC_LOCK_STATUS);
-	}
-    }
-
- done:
-	
-    if (rp && rp != ERTS_PROC_LOCK_BUSY && !(pid_locks & ERTS_PROC_LOCK_STATUS))
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_STATUS);
-    if (unlock_c_p_status)
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
-    return rp;
-}
-
-
-/*
- * Like erts_pid2proc() but:
- *
- * * At least ERTS_PROC_LOCK_MAIN have to be held on c_p.
- * * At least ERTS_PROC_LOCK_MAIN have to be taken on pid.
- * * It also waits for proc to be in a state != running and garbing.
- * * If ERTS_PROC_LOCK_BUSY is returned, the calling process has to
- *   yield (ERTS_BIF_YIELD[0-3]()). c_p might in this case have been
- *   suspended.
- */
-Process *
-erts_pid2proc_not_running(Process *c_p, ErtsProcLocks c_p_locks,
-			  Eterm pid, ErtsProcLocks pid_locks)
-{
-    return pid2proc_not_running(c_p, c_p_locks, pid, pid_locks, 0);
-}
-
-/*
- * erts_pid2proc_nropt() is normally the same as
- * erts_pid2proc_not_running(). However it is only
- * to be used when 'not running' is a pure optimization,
- * not a requirement.
- */
-
-Process *
-erts_pid2proc_nropt(Process *c_p, ErtsProcLocks c_p_locks,
-		    Eterm pid, ErtsProcLocks pid_locks)
-{
-    if (erts_disable_proc_not_running_opt)
-	return erts_pid2proc(c_p, c_p_locks, pid, pid_locks);
-    else
-	return erts_pid2proc_not_running(c_p, c_p_locks, pid, pid_locks);
-}
-
-static ERTS_INLINE int
-do_bif_suspend_process(Process *c_p,
-		       ErtsSuspendMonitor *smon,
-		       Process *suspendee)
-{
-    ASSERT(suspendee);
-    ASSERT(!ERTS_PROC_IS_EXITING(suspendee));
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS
-		       & erts_proc_lc_my_proc_locks(suspendee));
-    if (smon) {
-	if (!smon->active) {
-	    if (!suspend_process(c_p, suspendee))
-		return 0;
-	}
-	smon->active += smon->pending;
-	ASSERT(smon->active);
-	smon->pending = 0;
-	return 1;
-    }
-    return 0;
-}
-
-static void
-handle_pend_bif_sync_suspend(Process *suspendee,
-			     ErtsProcLocks suspendee_locks,
-			     int suspendee_alive,
-			     Eterm suspender_pid)
-{
-    Process *suspender;
-
-    ERTS_SMP_LC_ASSERT(suspendee_locks & ERTS_PROC_LOCK_STATUS);
-
-    suspender = erts_pid2proc(suspendee,
-			      suspendee_locks,
-			      suspender_pid,
-			      ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-    if (suspender) {
-	ASSERT(is_nil(suspender->suspendee));
-	if (!suspendee_alive)
-	    erts_delete_suspend_monitor(&suspender->suspend_monitors,
-					suspendee->common.id);
-	else {
-#ifdef DEBUG
-	    int res;
-#endif
-	    ErtsSuspendMonitor *smon;
-	    smon = erts_lookup_suspend_monitor(suspender->suspend_monitors,
-					       suspendee->common.id);
-#ifdef DEBUG
-	    res =
-#endif
-		do_bif_suspend_process(suspendee, smon, suspendee);
-	    ASSERT(!smon || res != 0);
-	    suspender->suspendee = suspendee->common.id;
-	}
-	/* suspender is suspended waiting for suspendee to suspend;
-	   resume suspender */
-	ASSERT(suspender != suspendee);
-	resume_process(suspender, ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-	erts_smp_proc_unlock(suspender,
-			     ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS);
-    }
-}
-
-static void
-handle_pend_bif_async_suspend(Process *suspendee,
-			      ErtsProcLocks suspendee_locks,
-			      int suspendee_alive,
-			      Eterm suspender_pid)
-{
-
-    Process *suspender;
-
-    ERTS_SMP_LC_ASSERT(suspendee_locks & ERTS_PROC_LOCK_STATUS);
-
-    suspender = erts_pid2proc(suspendee,
-			      suspendee_locks,
-			      suspender_pid,
-			      ERTS_PROC_LOCK_LINK);
-    if (suspender) {
-	ASSERT(is_nil(suspender->suspendee));
-	if (!suspendee_alive)
-	    erts_delete_suspend_monitor(&suspender->suspend_monitors,
-					suspendee->common.id);
-	else {
-#ifdef DEBUG
-	    int res;
-#endif
-	    ErtsSuspendMonitor *smon;
-	    smon = erts_lookup_suspend_monitor(suspender->suspend_monitors,
-					       suspendee->common.id);
-#ifdef DEBUG
-	    res =
-#endif
-	    do_bif_suspend_process(suspendee, smon, suspendee);
-	    ASSERT(!smon || res != 0);
-	}
-	erts_smp_proc_unlock(suspender, ERTS_PROC_LOCK_LINK);
-    }
-}
-
-#endif /* ERTS_SMP */
-
-/*
- * The erlang:suspend_process/2 BIF
- */
-
 BIF_RETTYPE
-suspend_process_2(BIF_ALIST_2)
+erts_internal_suspend_process_2(BIF_ALIST_2)
 {
     Eterm res;
-    Process* suspendee = NULL;
-    ErtsSuspendMonitor *smon;
-    ErtsProcLocks xlocks = (ErtsProcLocks) 0;
-
-    /* Options and default values: */
-    int asynchronous = 0;
+    Eterm reply_tag = THE_NON_VALUE;
+    Eterm reply_res = THE_NON_VALUE;
+    int suspend;
+    int sync = 0;
+    int async = 0;
     int unless_suspending = 0;
-
+    erts_aint_t mstate;
+    ErtsMonitorSuspend *msp;
+    ErtsMonitorData *mdp;
 
     if (BIF_P->common.id == BIF_ARG_1)
-	goto badarg; /* We are not allowed to suspend ourselves */
+	BIF_RET(am_badarg); /* We are not allowed to suspend ourselves */
 
     if (is_not_nil(BIF_ARG_2)) {
 	/* Parse option list */
@@ -9540,213 +8584,128 @@ suspend_process_2(BIF_ALIST_2)
 		unless_suspending = 1;
 		break;
 	    case am_asynchronous:
-		asynchronous = 1;
+		async = 1;
 		break;
-	    default:
-		goto badarg;
+	    default: {
+                if (is_tuple_arity(arg, 2)) {
+                    Eterm *tp = tuple_val(arg);
+                    if (tp[1] == am_asynchronous) {
+                        async = 1;
+                        reply_tag = tp[2];
+                        break;
+                    }
+                }
+                BIF_RET(am_badarg);
 	    }
+            }
 	    arg = CDR(lp);
-	}
+        }
 	if (is_not_nil(arg))
-	    goto badarg;
+            BIF_RET(am_badarg);
     }
 
-    xlocks = ERTS_PROC_LOCK_LINK | (asynchronous
-				    ? (ErtsProcLocks) 0
-				    : ERTS_PROC_LOCK_STATUS);
+    if (!unless_suspending) {
+        ErtsMonitor *mon;
+        mon = erts_monitor_tree_lookup_create(&ERTS_P_MONITORS(BIF_P),
+                                              &suspend,
+                                              ERTS_MON_TYPE_SUSPEND,
+                                              BIF_P->common.id,
+                                              BIF_ARG_1);
+        ASSERT(mon->other.item == BIF_ARG_1);
 
-    erts_smp_proc_lock(BIF_P, xlocks);
+        mdp = erts_monitor_to_data(mon);
+        msp = (ErtsMonitorSuspend *) mdp;
 
-    suspendee = erts_pid2proc(BIF_P,
-			      ERTS_PROC_LOCK_MAIN|xlocks,
-			      BIF_ARG_1,
-			      ERTS_PROC_LOCK_STATUS);
-    if (!suspendee)
-	goto no_suspendee;
-
-    smon = erts_add_or_lookup_suspend_monitor(&BIF_P->suspend_monitors,
-					      BIF_ARG_1);
-#ifndef ERTS_SMP /* no ERTS_SMP */
-
-    /* This is really a piece of cake without SMP support... */
-    if (!smon->active) {
-	erts_smp_atomic32_read_bor_nob(&suspendee->state, ERTS_PSFLG_SUSPENDED);
-	suspend_process(BIF_P, suspendee);
-	smon->active++;
-	res = am_true;
+        mstate = erts_atomic_inc_read_relb(&msp->state);
+        ASSERT(suspend || (mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK) > 1);
+        sync = !async & !suspend & !(mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE);
+        suspend = !!suspend; /* ensure 0|1 */
+        res = am_true;
     }
-    else if (unless_suspending)
-	res = am_false;
-    else if (smon->active == INT_MAX)
-	goto system_limit;
     else {
-	smon->active++;
-	res = am_true;
+        ErtsMonitor *mon;
+        mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(BIF_P),
+                                       BIF_ARG_1);
+        if (mon) {
+            ASSERT(mon->type == ERTS_MON_TYPE_SUSPEND);
+            mdp = erts_monitor_to_data(mon);
+            msp = (ErtsMonitorSuspend *) mdp;
+            mstate = erts_atomic_read_nob(&msp->state);
+            ASSERT((mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK) > 0);
+            mdp = NULL;
+            sync = !async & !(mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE);
+            suspend = 0;
+            res = am_false;
+        }
+        else {
+            mdp = erts_monitor_create(ERTS_MON_TYPE_SUSPEND, NIL,
+                                      BIF_P->common.id,
+                                      BIF_ARG_1, NIL);
+            mon = &mdp->origin;
+            erts_monitor_tree_insert(&ERTS_P_MONITORS(BIF_P), mon);
+            msp = (ErtsMonitorSuspend *) mdp;
+            mstate = erts_atomic_inc_read_relb(&msp->state);
+            ASSERT(!(mstate & ERTS_MSUSPEND_STATE_FLG_ACTIVE));
+            suspend = !0;
+            res = am_true;
+        }
     }
 
-#else /* ERTS_SMP */
-
-    /* ... but a little trickier with SMP support ... */
-
-    if (asynchronous) {
-	/* --- Asynchronous suspend begin ---------------------------------- */
-
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_LINK
-			   & erts_proc_lc_my_proc_locks(BIF_P));
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS
-			   == erts_proc_lc_my_proc_locks(suspendee));
-
-	if (smon->active) {
-	    smon->active += smon->pending;
-	    smon->pending = 0;
-	    if (unless_suspending)
-		res = am_false;
-	    else if (smon->active == INT_MAX)
-		goto system_limit;
-	    else {
-		smon->active++;
-		res = am_true;
-	    }
-	    /* done */
-	}
-	else {
-	    /* We havn't got any active suspends on the suspendee */
-	    if (smon->pending && unless_suspending)
-		res = am_false;
-	    else {
-		if (smon->pending == INT_MAX)
-		    goto system_limit;
-
-		smon->pending++;
-
-		if (!do_bif_suspend_process(BIF_P, smon, suspendee))
-		    add_pend_suspend(suspendee,
-				     BIF_P->common.id,
-				     handle_pend_bif_async_suspend);
-
-		res = am_true;
-	    }
-	    /* done */
-	}
-	/* --- Asynchronous suspend end ------------------------------------ */
-    }
-    else /* if (!asynchronous) */ {
-	/* --- Synchronous suspend begin ----------------------------------- */
-
-	ERTS_SMP_LC_ASSERT(((ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS)
-			    & erts_proc_lc_my_proc_locks(BIF_P))
-			   == (ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCK_STATUS));
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS
-			   == erts_proc_lc_my_proc_locks(suspendee));
-
-	if (BIF_P->suspendee == BIF_ARG_1) {
-	    /* We are back after a yield and the suspendee
-	       has been suspended on behalf of us. */
-	    ASSERT(smon->active >= 1);
-	    BIF_P->suspendee = NIL;
-	    res = (!unless_suspending || smon->active == 1
-		   ? am_true
-		   : am_false);
-	    /* done */
-	}
-	else if (smon->active) {
-	    if (unless_suspending)
-		res = am_false;
-	    else {
-		smon->active++;
-		res = am_true;
-	    }
-	    /* done */
-	}
-	else {
-	    /* We haven't got any active suspends on the suspendee */
-
-	    /*
-	     * If we have pending suspenders and suspend ourselves waiting
-	     * to suspend another process, or suspend another process
-	     * we might deadlock. In this case we have to yield,
-	     * be suspended by someone else, and then do it all over again.
-	     */
-	    if (BIF_P->pending_suspenders)
-		goto yield;
-
-	    if (!unless_suspending && smon->pending == INT_MAX)
-		goto system_limit;
-	    if (!unless_suspending || smon->pending == 0)
-		smon->pending++;
-
-	    if (do_bif_suspend_process(BIF_P, smon, suspendee)) {
-		res = (!unless_suspending || smon->active == 1
-		       ? am_true
-		       : am_false);
-		/* done */
-	    }
-	    else {
-		/* Mark suspendee pending for suspend by BIF_P */
-		add_pend_suspend(suspendee,
-				 BIF_P->common.id,
-				 handle_pend_bif_sync_suspend);
-
-		ASSERT(is_nil(BIF_P->suspendee));
-
-		/*
-		 * Suspend BIF_P; when suspendee is suspended, BIF_P
-		 * will be resumed and this BIF will be called again.
-		 * This time with BIF_P->suspendee == BIF_ARG_1 (see
-		 * above).
-		 */
-		suspend_process(BIF_P, BIF_P);
-		goto yield;
-	    }
-	}
-	/* --- Synchronous suspend end ------------------------------------- */
+    if (suspend) {
+        erts_aint32_t state;
+        Process *rp;
+        int send_sig = 0;
+
+        /* fail state... */
+        state = (ERTS_PSFLG_EXITING
+                 | ERTS_PSFLG_RUNNING
+                 | ERTS_PSFLG_RUNNING_SYS
+                 | ERTS_PSFLG_DIRTY_RUNNING
+                 | ERTS_PSFLG_DIRTY_RUNNING_SYS);
+
+        rp = erts_try_lock_sig_free_proc(BIF_ARG_1,
+                                         ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS,
+                                         &state);
+        if (!rp)
+            goto noproc;
+        if (rp == ERTS_PROC_LOCK_BUSY)
+            send_sig = !0;
+        else {
+            send_sig = !suspend_process(BIF_P, rp);
+            if (!send_sig) {
+                erts_monitor_list_insert(&ERTS_P_LT_MONITORS(rp), &mdp->target);
+                erts_atomic_read_bor_relb(&msp->state,
+                                          ERTS_MSUSPEND_STATE_FLG_ACTIVE);
+            }
+            erts_proc_unlock(rp, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS);
+        }
+        if (send_sig) {
+            if (erts_proc_sig_send_monitor(&mdp->target, BIF_ARG_1))
+                sync = !async;
+            else {
+            noproc:
+                erts_monitor_tree_delete(&ERTS_P_MONITORS(BIF_P), &mdp->origin);
+                erts_monitor_release_both(mdp);
+                if (!async)
+                    res = am_badarg;
+            }
+        }
     }
 
-#endif /* ERTS_SMP */
-#ifdef DEBUG
-    {
-	erts_aint32_t state = erts_smp_atomic32_read_acqb(&suspendee->state);
-	ASSERT((state & ERTS_PSFLG_SUSPENDED)
-	       || (asynchronous && smon->pending));
-	ASSERT((state & ERTS_PSFLG_SUSPENDED)
-	       || !smon->active);
+    if (sync) {
+        ASSERT(is_non_value(reply_tag));
+        reply_res = res;
+        reply_tag = res = erts_make_ref(BIF_P);
+        ERTS_RECV_MARK_SAVE(BIF_P);
+        ERTS_RECV_MARK_SET(BIF_P);
     }
-#endif
-
-    erts_smp_proc_unlock(suspendee, ERTS_PROC_LOCK_STATUS);
-    erts_smp_proc_unlock(BIF_P, xlocks);
-    BIF_RET(res);
-
- system_limit:
-    ERTS_BIF_PREP_ERROR(res, BIF_P, SYSTEM_LIMIT);
-    goto do_return;
-
- no_suspendee:
-#ifdef ERTS_SMP
-    BIF_P->suspendee = NIL;
-#endif
-    erts_delete_suspend_monitor(&BIF_P->suspend_monitors, BIF_ARG_1);
-
- badarg:
-    ERTS_BIF_PREP_ERROR(res, BIF_P, BADARG);
-#ifdef ERTS_SMP
-    goto do_return;
-
- yield:
-    ERTS_BIF_PREP_YIELD2(res, bif_export[BIF_suspend_process_2],
-			 BIF_P, BIF_ARG_1, BIF_ARG_2);
-#endif
 
- do_return:
-    if (suspendee)
-	erts_smp_proc_unlock(suspendee, ERTS_PROC_LOCK_STATUS);
-    if (xlocks)
-	erts_smp_proc_unlock(BIF_P, xlocks);
-    return res;
+    if (is_value(reply_tag))
+        erts_proc_sig_send_sync_suspend(BIF_P, BIF_ARG_1, reply_tag, reply_res);
 
+    BIF_RET(res);
 }
 
-
 /*
  * The erlang:resume_process/1 BIF
  */
@@ -9754,80 +8713,33 @@ suspend_process_2(BIF_ALIST_2)
 BIF_RETTYPE
 resume_process_1(BIF_ALIST_1)
 {
-    ErtsSuspendMonitor *smon;
-    Process *suspendee;
-    int is_active;
+    ErtsMonitor *mon;
+    ErtsMonitorSuspend *msp;
+    erts_aint_t mstate;
  
     if (BIF_P->common.id == BIF_ARG_1)
 	BIF_ERROR(BIF_P, BADARG);
 
-    erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_LINK);
-    smon = erts_lookup_suspend_monitor(BIF_P->suspend_monitors, BIF_ARG_1);
-
-    if (!smon) {
+    mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(BIF_P),
+                                   BIF_ARG_1);
+    if (!mon) {
 	/* No previous suspend or dead suspendee */
-	goto error;
-    }
-    else if (smon->pending) {
-	smon->pending--;
-	ASSERT(smon->pending >= 0);
-	if (smon->active) {
-	    smon->active += smon->pending;
-	    smon->pending = 0;
-	}
-	is_active = smon->active;
-    }
-    else if (smon->active) {
-	smon->active--;
-	ASSERT(smon->pending >= 0);
-	is_active = 1;
-    }
-    else {
-	/* No previous suspend or dead suspendee */
-	goto error;
+        BIF_ERROR(BIF_P, BADARG);
     }
 
-    if (smon->active || smon->pending || !is_active) {
-	/* Leave the suspendee as it is; just verify that it is still alive */
-	suspendee = erts_pid2proc(BIF_P,
-				  ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK,
-				  BIF_ARG_1,
-				  0);
-	if (!suspendee)
-	    goto no_suspendee;
+    ASSERT(mon->type == ERTS_MON_TYPE_SUSPEND);
+    msp = (ErtsMonitorSuspend *) erts_monitor_to_data(mon);
 
-    }
-    else {
-	/* Resume */
-	suspendee = erts_pid2proc(BIF_P,
-				  ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_LINK,
-				  BIF_ARG_1,
-				  ERTS_PROC_LOCK_STATUS);
-	if (!suspendee)
-	    goto no_suspendee;
+    mstate = erts_atomic_dec_read_relb(&msp->state);
 
-	ASSERT(ERTS_PSFLG_SUSPENDED
-	       & erts_smp_atomic32_read_nob(&suspendee->state));
-	ASSERT(BIF_P != suspendee);
-	resume_process(suspendee, ERTS_PROC_LOCK_STATUS);
+    ASSERT((mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK) >= 0);
 
-	erts_smp_proc_unlock(suspendee, ERTS_PROC_LOCK_STATUS);
+    if ((mstate & ERTS_MSUSPEND_STATE_COUNTER_MASK) == 0) {
+        erts_monitor_tree_delete(&ERTS_P_MONITORS(BIF_P), mon);
+        erts_proc_sig_send_demonitor(mon);
     }
 
-    if (!smon->active && !smon->pending)
-	erts_delete_suspend_monitor(&BIF_P->suspend_monitors, BIF_ARG_1);
-
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-
     BIF_RET(am_true);
-
- no_suspendee:
-    /* cleanup */
-    erts_delete_suspend_monitor(&BIF_P->suspend_monitors, BIF_ARG_1);
-
- error:
-    erts_smp_proc_unlock(BIF_P, ERTS_PROC_LOCK_LINK);
-    BIF_ERROR(BIF_P, BADARG);
 }
 
 BIF_RETTYPE
@@ -9835,18 +8747,16 @@ erts_internal_is_process_executing_dirty_1(BIF_ALIST_1)
 {
     if (is_not_internal_pid(BIF_ARG_1))
 	BIF_ERROR(BIF_P, BADARG);
-#ifdef ERTS_DIRTY_SCHEDULERS
     else {
 	Process *rp = erts_proc_lookup(BIF_ARG_1);
 	if (rp) {
-	    erts_aint32_t state = erts_smp_atomic32_read_nob(&rp->state);
+	    erts_aint32_t state = erts_atomic32_read_nob(&rp->state);
 	    if (state & (ERTS_PSFLG_DIRTY_RUNNING
 			 |ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
 		BIF_RET(am_true);
 	    }
 	}
     }
-#endif
     BIF_RET(am_false);
 }
 
@@ -9856,28 +8766,26 @@ run_queues_len_aux(ErtsRunQueue *rq, Uint *tot_len, Uint *qlen, int *ip, int inc
     Sint rq_len;
 
     if (locked)
-        rq_len = (Sint) erts_smp_atomic32_read_dirty(&rq->len);
+        rq_len = (Sint) erts_atomic32_read_dirty(&rq->len);
     else
-        rq_len = (Sint) erts_smp_atomic32_read_nob(&rq->len);
+        rq_len = (Sint) erts_atomic32_read_nob(&rq->len);
     ASSERT(rq_len >= 0);
 
     if (incl_active_sched) {
-#ifdef ERTS_DIRTY_SCHEDULERS
         if (ERTS_RUNQ_IX_IS_DIRTY(rq->ix)) {
             erts_aint32_t dcnt;
             if (ERTS_RUNQ_IS_DIRTY_CPU_RUNQ(rq)) {
-                dcnt = erts_smp_atomic32_read_nob(&dirty_count.cpu.active);
+                dcnt = erts_atomic32_read_nob(&dirty_count.cpu.active);
                 ASSERT(0 <= dcnt && dcnt <= erts_no_dirty_cpu_schedulers);
             }
             else {
                 ASSERT(ERTS_RUNQ_IS_DIRTY_IO_RUNQ(rq));
-                dcnt = erts_smp_atomic32_read_nob(&dirty_count.io.active);
+                dcnt = erts_atomic32_read_nob(&dirty_count.io.active);
                 ASSERT(0 <= dcnt && dcnt <= erts_no_dirty_io_schedulers);
             }
             rq_len += (Sint) dcnt;
         }
         else
-#endif
         {
             if (ERTS_RUNQ_FLGS_GET_NOB(rq) & ERTS_RUNQ_FLG_EXEC)
                 rq_len++;
@@ -9896,12 +8804,10 @@ erts_run_queues_len(Uint *qlen, int atomic_queues_read, int incl_active_sched,
     Uint len = 0;
     int no_rqs = erts_no_run_queues;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (incl_dirty_io)
         no_rqs += ERTS_NUM_DIRTY_RUNQS;
     else
         no_rqs += ERTS_NUM_DIRTY_CPU_RUNQS;
-#endif
 
     if (atomic_queues_read) {
         ERTS_ATOMIC_FOREACH_RUNQ_X(rq, no_rqs,
@@ -9955,27 +8861,25 @@ erts_process_status(Process *rp, Eterm rpid)
     Process *p = rp ? rp : erts_proc_lookup_raw(rpid);
 
     if (p) {
-	erts_aint32_t state = erts_smp_atomic32_read_acqb(&p->state);
+	erts_aint32_t state = erts_atomic32_read_acqb(&p->state);
 	res = erts_process_state2status(state);
     }
-#ifdef ERTS_SMP
     else {
 	int i;
 	ErtsSchedulerData *esdp;
 
 	for (i = 0; i < erts_no_schedulers; i++) {
 	    esdp = ERTS_SCHEDULER_IX(i);
-	    erts_smp_runq_lock(esdp->run_queue);
+	    erts_runq_lock(esdp->run_queue);
 	    if (esdp->free_process
 		&& esdp->free_process->common.id == rpid) {
 		res = am_free;
-		erts_smp_runq_unlock(esdp->run_queue);
+		erts_runq_unlock(esdp->run_queue);
 		break;
 	    }
-	    erts_smp_runq_unlock(esdp->run_queue);
+	    erts_runq_unlock(esdp->run_queue);
 	}
     }
-#endif
     return res;
 }
 
@@ -9991,9 +8895,9 @@ erts_suspend(Process* c_p, ErtsProcLocks c_p_locks, Port *busy_port)
     int suspend;
 
     ASSERT(c_p == erts_get_current_process());
-    ERTS_SMP_LC_ASSERT(c_p_locks == erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(c_p_locks == erts_proc_lc_my_proc_locks(c_p));
     if (!(c_p_locks & ERTS_PROC_LOCK_STATUS))
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
 
     if (busy_port)
 	suspend = erts_save_suspend_process_on_port(busy_port, c_p);
@@ -10009,7 +8913,7 @@ erts_suspend(Process* c_p, ErtsProcLocks c_p_locks, Port *busy_port)
     }
 
     if (!(c_p_locks & ERTS_PROC_LOCK_STATUS))
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
 
     if (suspend && busy_port && erts_system_monitor_flags.busy_port)
 	monitor_generic(c_p, am_busy_port, busy_port->common.id);
@@ -10018,12 +8922,12 @@ erts_suspend(Process* c_p, ErtsProcLocks c_p_locks, Port *busy_port)
 void
 erts_resume(Process* process, ErtsProcLocks process_locks)
 {
-    ERTS_SMP_LC_ASSERT(process_locks == erts_proc_lc_my_proc_locks(process));
+    ERTS_LC_ASSERT(process_locks == erts_proc_lc_my_proc_locks(process));
     if (!(process_locks & ERTS_PROC_LOCK_STATUS))
-	erts_smp_proc_lock(process, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(process, ERTS_PROC_LOCK_STATUS);
     resume_process(process, process_locks|ERTS_PROC_LOCK_STATUS);
     if (!(process_locks & ERTS_PROC_LOCK_STATUS))
-	erts_smp_proc_unlock(process, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(process, ERTS_PROC_LOCK_STATUS);
 }
 
 int
@@ -10043,7 +8947,7 @@ erts_resume_processes(ErtsProcList *list)
 		resume_process(proc, ERTS_PROC_LOCK_STATUS);
 		nresumed++;
 	    }
-	    erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(proc, ERTS_PROC_LOCK_STATUS);
 	}
 	fplp = plp;
 	plp = plp->next;
@@ -10053,9 +8957,8 @@ erts_resume_processes(ErtsProcList *list)
 }
 
 Eterm
-erts_get_process_priority(Process *p)
+erts_get_process_priority(erts_aint32_t state)
 {
-    erts_aint32_t state = erts_smp_atomic32_read_nob(&p->state);
     switch (ERTS_PSFLGS_GET_USR_PRIO(state)) {
     case PRIORITY_MAX:		return am_max;
     case PRIORITY_HIGH:		return am_high;
@@ -10078,7 +8981,7 @@ erts_set_process_priority(Process *p, Eterm value)
     default:		return THE_NON_VALUE;				break;
     }
 
-    a = erts_smp_atomic32_read_nob(&p->state);
+    a = erts_atomic32_read_nob(&p->state);
     if (nprio == ERTS_PSFLGS_GET_USR_PRIO(a))
 	oprio = nprio;
     else {
@@ -10086,7 +8989,7 @@ erts_set_process_priority(Process *p, Eterm value)
 	erts_aint32_t e, n, aprio;
 
 	if (a & ERTS_PSFLG_ACTIVE_SYS) {
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	    slocked = 1;
 	}
 
@@ -10100,12 +9003,15 @@ erts_set_process_priority(Process *p, Eterm value)
 		int max_qbit;
 
 		if (!slocked) {
-		    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+		    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 		    slocked = 1;
 		}
 
 		max_qbit = 0;
-		if (a & ERTS_PSFLG_ACTIVE_SYS)
+                ASSERT((a & ERTS_PSFLG_SYS_TASKS)
+                       ? !!p->sys_task_qs
+                       : !p->sys_task_qs);
+		if (a & ERTS_PSFLG_SYS_TASKS)
 		    max_qbit |= p->sys_task_qs->qmask;
 		if (a & ERTS_PSFLG_DELAYED_SYS) {
 		    ErtsProcSysTaskQs *qs;
@@ -10128,8 +9034,8 @@ erts_set_process_priority(Process *p, Eterm value)
 		    aprio = PRIORITY_LOW;
 		    break;
 		default:
-		    ERTS_INTERNAL_ERROR("Invalid qmask");
-		    aprio = -1;
+                    aprio = nprio;
+                    break;
 		}
 
 		if (aprio > nprio) /* low value -> high prio */
@@ -10141,11 +9047,11 @@ erts_set_process_priority(Process *p, Eterm value)
 	    n |= ((nprio << ERTS_PSFLGS_USR_PRIO_OFFSET)
 		  | (aprio << ERTS_PSFLGS_ACT_PRIO_OFFSET));
 
-	    a = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
+	    a = erts_atomic32_cmpxchg_mb(&p->state, n, e);
 	} while (a != e);
 
         if (slocked)
-            erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+            erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 
     }
 
@@ -10179,6 +9085,17 @@ scheduler_gc_proc(Process *c_p, int reds_left)
     return reds;
 }
 
+static void
+unlock_lock_rq(int pre_free, void *vrq)
+{
+    ErtsRunQueue *rq = vrq;
+    if (pre_free)
+        erts_runq_unlock(rq);
+    else
+        erts_runq_lock(rq);
+}
+
+
 /*
  * schedule() is called from BEAM (process_main()) or HiPE
  * (hipe_mode_switch()) when the current process is to be
@@ -10203,7 +9120,6 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
     ErtsRunQueue *rq;
     int context_reds;
     int fcalls;
-    int input_reductions;
     int actual_reds;
     int reds;
     Uint32 flags;
@@ -10223,21 +9139,18 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 
     if (ERTS_USE_MODIFIED_TIMING()) {
 	context_reds = ERTS_MODIFIED_TIMING_CONTEXT_REDS;
-	input_reductions = ERTS_MODIFIED_TIMING_INPUT_REDS;
     }
     else {
 	context_reds = CONTEXT_REDS;
-	input_reductions = INPUT_REDUCTIONS;
     }
 
-    ERTS_SMP_LC_ASSERT(ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data())
+    ERTS_LC_ASSERT(ERTS_SCHEDULER_IS_DIRTY(erts_get_scheduler_data())
 		       || !erts_thr_progress_is_blocking());
 
     /*
      * Clean up after the process being scheduled out.
      */
     if (!p) {	/* NULL in the very first schedule() call */
-#ifdef ERTS_DIRTY_SCHEDULERS
 	is_normal_sched = !esdp;
 	if (is_normal_sched) {
 	    esdp = erts_get_scheduler_data();
@@ -10246,57 +9159,47 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	else {
 	    ASSERT(ERTS_SCHEDULER_IS_DIRTY(esdp));
 	}
-#else
-	esdp = erts_get_scheduler_data();
-	is_normal_sched = 1;
-#endif
 	rq = erts_get_runq_current(esdp);
 	ASSERT(esdp);
-	fcalls = (int) erts_smp_atomic32_read_acqb(&function_calls);
 	actual_reds = reds = 0;
-	erts_smp_runq_lock(rq);
+	erts_runq_lock(rq);
     } else {
-#ifdef ERTS_SMP
-#ifdef ERTS_DIRTY_SCHEDULERS
 	is_normal_sched = !esdp;
 	if (is_normal_sched) {
-	    esdp = p->scheduler_data;
+            esdp = p->scheduler_data;
 	    ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
+
+            if (esdp->pending_signal.sig) {
+                ASSERT(esdp->pending_signal.dbg_from == p);
+                erts_proc_sig_send_pending(esdp);
+            }
 	}
 	else {
 	    ASSERT(ERTS_SCHEDULER_IS_DIRTY(esdp));
 	}
-#else
-	esdp = p->scheduler_data;
-	is_normal_sched = 1;
-#endif
 	ASSERT(esdp->current_process == p
 	       || esdp->free_process == p);
-#else
-	esdp = erts_scheduler_data;
-	ASSERT(esdp->current_process == p);
-	is_normal_sched = 1;
-#endif
 
-    sched_out_proc:
-
-	ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
 
 	reds = actual_reds = calls - esdp->virtual_reds;
 
+    internal_sched_out_proc:
+
+	ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
+        ASSERT(p->scheduler_data || ERTS_SCHEDULER_IS_DIRTY(esdp));
+
 	ASSERT(actual_reds >= 0);
 	if (reds < ERTS_PROC_MIN_CONTEXT_SWITCH_REDS_COST)
 	    reds = ERTS_PROC_MIN_CONTEXT_SWITCH_REDS_COST;
 	esdp->virtual_reds = 0;
 
-	fcalls = (int) erts_smp_atomic32_add_read_acqb(&function_calls, reds);
 	ASSERT(esdp && esdp == erts_get_scheduler_data());
 
 	rq = erts_get_runq_current(esdp);
 
 	p->reds += actual_reds;
 
-	state = erts_smp_atomic32_read_nob(&p->state);
+	state = erts_atomic32_read_nob(&p->state);
 
 	if (IS_TRACED(p)) {
 	    if (IS_TRACED_FL(p, F_TRACE_CALLS) && !(state & ERTS_PSFLG_FREE))
@@ -10315,29 +9218,16 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	    }
 	}
 
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+	erts_proc_lock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
 
-#ifdef ERTS_SMP
         if (p->trace_msg_q) {
-	    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+	    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
             erts_schedule_flush_trace_messages(p, 1);
-	    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+	    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
 	}
-#endif
 
         /* have to re-read state after taking lock */
-        state = erts_smp_atomic32_read_nob(&p->state);
-
-#ifdef ERTS_SMP
-	if (is_normal_sched && (state & ERTS_PSFLG_PENDING_EXIT))
-	    erts_handle_pending_exit(p, (ERTS_PROC_LOCK_MAIN
-					 | ERTS_PROC_LOCK_TRACE
-					 | ERTS_PROC_LOCK_STATUS));
-	if (p->pending_suspenders) 
-	    handle_pending_suspend(p, (ERTS_PROC_LOCK_MAIN
-				       | ERTS_PROC_LOCK_TRACE
-				       | ERTS_PROC_LOCK_STATUS));
-#endif
+        state = erts_atomic32_read_nob(&p->state);
 
 	esdp->reductions += reds;
 
@@ -10355,18 +9245,15 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                                           actual_reds);
 
             esdp->current_process = NULL;
-#ifdef ERTS_SMP
             if (is_normal_sched)
                 p->scheduler_data = NULL;
-#endif
 
-            erts_smp_proc_unlock(p, (ERTS_PROC_LOCK_MAIN
+            erts_proc_unlock(p, (ERTS_PROC_LOCK_MAIN
                                      | ERTS_PROC_LOCK_STATUS
                                      | ERTS_PROC_LOCK_TRACE));
 
             ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER);
 
-#ifdef ERTS_SMP
             if (state & ERTS_PSFLG_FREE) {
                 if (!is_normal_sched) {
                     ASSERT(p->flags & F_DELAYED_DEL_PROC);
@@ -10376,44 +9263,42 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                     esdp->free_process = NULL;
                 }
             }
-#endif
 
             if (dec_refc)
-                erts_proc_dec_refc(p);
+                erts_proc_dec_refc_free_func(p,
+                                             unlock_lock_rq,
+                                             (void *) rq);
         }
 
-#ifdef ERTS_SMP
 	ASSERT(!esdp->free_process);
-#endif
 	ASSERT(!esdp->current_process);
 
-	ERTS_SMP_CHK_NO_PROC_LOCKS;
+	ERTS_CHK_NO_PROC_LOCKS;
 
-	if (is_normal_sched) {
-	    if (esdp->check_time_reds >= ERTS_CHECK_TIME_REDS)
-		(void) erts_get_monotonic_time(esdp);
-
-	    if (esdp->last_monotonic_time >= erts_next_timeout_time(esdp->next_tmo_ref)) {
-		erts_smp_runq_unlock(rq);
-		erts_bump_timers(esdp->timer_wheel, esdp->last_monotonic_time);
-		erts_smp_runq_lock(rq);
-	    }
-	}
     }
 
-    ERTS_SMP_LC_ASSERT(!is_normal_sched || !erts_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(!is_normal_sched || !erts_thr_progress_is_blocking());
 
  check_activities_to_run: {
 	erts_aint32_t psflg_running, psflg_running_sys;
-#ifdef ERTS_SMP
 	ErtsMigrationPaths *mps;
 	ErtsMigrationPath *mp;
 
 	if (is_normal_sched) {
+
+	    if (esdp->check_time_reds >= ERTS_CHECK_TIME_REDS)
+		(void) erts_get_monotonic_time(esdp);
+
+	    if (esdp->last_monotonic_time >= erts_next_timeout_time(esdp->next_tmo_ref)) {
+		erts_runq_unlock(rq);
+		erts_bump_timers(esdp->timer_wheel, esdp->last_monotonic_time);
+		erts_runq_lock(rq);
+	    }
+
 	    if (rq->check_balance_reds <= 0)
 		check_balance(rq);
 
-	    ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
+	    ERTS_LC_ASSERT(!erts_thr_progress_is_blocking());
 
 	    mps = erts_get_migration_paths_managed();
 	    mp = &mps->mpath[rq->ix];
@@ -10422,14 +9307,14 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 		immigrate(rq, mp);
 	}
 
-	ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+	ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
     continue_check_activities_to_run:
 	flags = ERTS_RUNQ_FLGS_GET_NOB(rq);
     continue_check_activities_to_run_known_flags:
 	ASSERT(!is_normal_sched || (flags & ERTS_RUNQ_FLG_NONEMPTY));
 
 	if (!is_normal_sched) {
-	    if (erts_smp_atomic32_read_acqb(&esdp->ssi->flags)
+	    if (erts_atomic32_read_acqb(&esdp->ssi->flags)
 		& (ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC)) {
 		suspend_scheduler(esdp);
 	    }
@@ -10459,26 +9344,18 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	    leader_update = erts_thr_progress_update(esdp);
 	    aux_work = erts_atomic32_read_acqb(&esdp->ssi->aux_work);
 	    if (aux_work | leader_update) {
-		erts_smp_runq_unlock(rq);
+		erts_runq_unlock(rq);
 		if (leader_update)
 		    erts_thr_progress_leader_update(esdp);
 		if (aux_work)
 		    handle_aux_work(&esdp->aux_work_data, aux_work, 0);
-		erts_smp_runq_lock(rq);
+		erts_runq_lock(rq);
 	    }
 
-	    ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
+	    ERTS_LC_ASSERT(!erts_thr_progress_is_blocking());
 	}
-	ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+	ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
-#else /* ERTS_SMP */
-	{
-	    erts_aint32_t aux_work;
-	    aux_work = erts_atomic32_read_acqb(&esdp->ssi->aux_work);
-	    if (aux_work)
-		handle_aux_work(&esdp->aux_work_data, aux_work, 0);
-	}
-#endif /* ERTS_SMP */
 
 	flags = ERTS_RUNQ_FLGS_GET_NOB(rq);
 
@@ -10489,8 +9366,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	}
 	else if (!runq_got_work_to_execute_flags(flags)) {
 	    /* Prepare for scheduler wait */
-#ifdef ERTS_SMP
-	    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+	    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
 	    rq->wakeup_other = 0;
 	    rq->wakeup_other_reds = 0;
@@ -10504,7 +9380,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                 ASSERT(!runq_got_work_to_execute(rq));
 		if (!is_normal_sched) {
                     /* Dirty scheduler */
-                    if (erts_smp_atomic32_read_acqb(&esdp->ssi->flags)
+                    if (erts_atomic32_read_acqb(&esdp->ssi->flags)
                         & (ERTS_SSI_FLG_SUSPENDED|ERTS_SSI_FLG_MSB_EXEC)) {
                         /* Go suspend... */
                         goto continue_check_activities_to_run_known_flags;
@@ -10520,7 +9396,6 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                      */
                     flags = ERTS_RUNQ_FLGS_GET_NOB(rq);
                     if ((flags & ERTS_RUNQ_FLG_SUSPENDED)
-#ifdef ERTS_DIRTY_SCHEDULERS
                         /* If multi scheduling block and we have
                          * dirty work, suspend and let dirty
                          * scheduler handle work... */
@@ -10528,7 +9403,6 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                                        | ERTS_RUNQ_FLG_MSB_EXEC))
                               == ERTS_RUNQ_FLG_MSB_EXEC))
                             && have_dirty_work())
-#endif
                         ) {
                         non_empty_runq(rq);
                         flags |= ERTS_RUNQ_FLG_NONEMPTY;
@@ -10540,62 +9414,21 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 		}
 		empty_runq(rq);
 	    }
-#endif
 
 	    (void) ERTS_RUNQ_FLGS_UNSET(rq, ERTS_RUNQ_FLG_EXEC);
 	    scheduler_wait(&fcalls, esdp, rq);
 	    flags = ERTS_RUNQ_FLGS_SET_NOB(rq, ERTS_RUNQ_FLG_EXEC);
 	    flags |= ERTS_RUNQ_FLG_EXEC;
             ERTS_MSACC_UPDATE_CACHE();
-#ifdef ERTS_SMP
 	    non_empty_runq(rq);
-#endif
-
-	    goto check_activities_to_run;
-	}
-	else if (is_normal_sched
-		 && (fcalls > input_reductions
-		     && prepare_for_sys_schedule(!0))) {
-	    ErtsMonotonicTime current_time;
-	    /*
-	     * Schedule system-level activities.
-	     */
-
-	    ERTS_MSACC_PUSH_STATE_CACHED_M();
-
-	    erts_smp_atomic32_set_relb(&function_calls, 0);
-	    fcalls = 0;
-
-#if 0 /* Not needed since we wont wait in sys schedule */
-	    erts_sys_schedule_interrupt(0);
-#endif
-	    erts_smp_runq_unlock(rq);
-
-            ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_CHECK_IO);
-            LTTNG2(scheduler_poll, esdp->no, 1);
-
-	    erl_sys_schedule(1);
-	    ERTS_MSACC_POP_STATE_M();
-
-	    current_time = erts_get_monotonic_time(esdp);
-	    if (current_time >= erts_next_timeout_time(esdp->next_tmo_ref))
-		erts_bump_timers(esdp->timer_wheel, current_time);
 
-#ifdef ERTS_SMP
-	    erts_smp_runq_lock(rq);
-	    clear_sys_scheduling();
-	    goto continue_check_activities_to_run;
-#else
 	    goto check_activities_to_run;
-#endif
 	}
 
         if (flags & ERTS_RUNQ_FLG_MISC_OP)
 	    exec_misc_ops(rq);
 
-#ifdef ERTS_SMP
-	wakeup_other.check(rq, flags);
-#endif
+        runq_get_wakeup_other_params(rq)->check(rq, flags);
 
 	/*
 	 * Find a new port to run.
@@ -10604,29 +9437,9 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
         flags = ERTS_RUNQ_FLGS_GET_NOB(rq);
 
 	if (flags & PORT_BIT) {
-	    int have_outstanding_io;
-	    have_outstanding_io = erts_port_task_execute(rq, &esdp->current_port);
-	    if ((!erts_eager_check_io
-		 && have_outstanding_io
-		 && fcalls > 2*input_reductions)
-		|| (flags & ERTS_RUNQ_FLG_HALTING)) {
-		/*
-		 * If we have performed more than 2*INPUT_REDUCTIONS since
-		 * last call to erl_sys_schedule() and we still haven't
-		 * handled all I/O tasks we stop running processes and
-		 * focus completely on ports.
-		 *
-		 * One could argue that this is a strange behavior. The
-		 * reason for doing it this way is that it is similar
-		 * to the behavior before port tasks were introduced.
-		 * We don't want to change the behavior too much, at
-		 * least not at the time of writing. This behavior
-		 * might change in the future.
-		 *
-		 * /rickard
-		 */
-		goto check_activities_to_run;
-	    }
+	    erts_port_task_execute(rq, &esdp->current_port);
+            if (flags & ERTS_RUNQ_FLG_HALTING)
+                goto check_activities_to_run;
 	}
 
 	/*
@@ -10693,13 +9506,11 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 		    proxy_p = NULL;
 		    goto pick_next_process;
 		}
-		state = erts_smp_atomic32_read_nob(&p->state);
+		state = erts_atomic32_read_nob(&p->state);
 	    }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    if (!is_normal_sched)
 		clear_proc_dirty_queue_bit(p, rq, qbit);
-#endif
 
 	    while (1) {
 		erts_aint32_t exp, new;
@@ -10717,7 +9528,6 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 					    | ERTS_PSFLG_DIRTY_RUNNING
 					    | ERTS_PSFLG_DIRTY_RUNNING_SYS
 					    | ERTS_PSFLG_FREE)))
-#ifdef ERTS_DIRTY_SCHEDULERS
 				| (((state & (ERTS_PSFLG_RUNNING
 
 					      | ERTS_PSFLG_FREE
@@ -10726,20 +9536,15 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 					      | ERTS_PSFLG_EXITING))
 				    == ERTS_PSFLG_EXITING)
 				   & (!!is_normal_sched))
-#endif
 				   )
 			       & ((state & (ERTS_PSFLG_SUSPENDED
 					    | ERTS_PSFLG_EXITING
 					    | ERTS_PSFLG_FREE
-					    | ERTS_PSFLG_PENDING_EXIT
 					    | ERTS_PSFLG_ACTIVE_SYS
 					    | ERTS_PSFLG_DIRTY_ACTIVE_SYS))
 				  != ERTS_PSFLG_SUSPENDED)
-#ifdef ERTS_DIRTY_SCHEDULERS
-                               & (!(state & (ERTS_PSFLG_EXITING
-                                             | ERTS_PSFLG_PENDING_EXIT))
+                               & (!(state & ERTS_PSFLG_EXITING)
                                   | (!!is_normal_sched))
-#endif
                     );
 
 		if (run_process) {
@@ -10749,7 +9554,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 		    else
 			new |= psflg_running;
 		}
-		state = erts_smp_atomic32_cmpxchg_relb(&p->state, new, exp);
+		state = erts_atomic32_cmpxchg_relb(&p->state, new, exp);
 		if (state == exp) {
 		    if (!run_process) {
 			if (proxy_p) {
@@ -10776,22 +9581,21 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	    calls = 0;
 	    reds = context_reds;
 
-	    erts_smp_runq_unlock(rq);
+	    erts_runq_unlock(rq);
 
 	}
 
         ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_EMULATOR);
 
-#ifdef ERTS_SMP
 
 	if (flags & ERTS_RUNQ_FLG_PROTECTED)
 	    (void) ERTS_RUNQ_FLGS_UNSET(rq, ERTS_RUNQ_FLG_PROTECTED);
 
-	ERTS_SMP_CHK_NO_PROC_LOCKS;
+	ERTS_CHK_NO_PROC_LOCKS;
 
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(p, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS);
 
-	state = erts_smp_atomic32_read_nob(&p->state);
+	state = erts_atomic32_read_nob(&p->state);
 
 	if (erts_sched_stat.enabled) {
 	    int prio;
@@ -10802,36 +9606,35 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 
 	    prio = (int) ERTS_PSFLGS_GET_USR_PRIO(state);
 
-	    erts_smp_spin_lock(&erts_sched_stat.lock);
+	    erts_spin_lock(&erts_sched_stat.lock);
 	    erts_sched_stat.prio[prio].total_executed++;
 	    erts_sched_stat.prio[prio].executed++;
 	    if (migrated) {
 		erts_sched_stat.prio[prio].total_migrated++;
 		erts_sched_stat.prio[prio].migrated++;
 	    }
-	    erts_smp_spin_unlock(&erts_sched_stat.lock);
+	    erts_spin_unlock(&erts_sched_stat.lock);
 	}
 
-	state = erts_smp_atomic32_read_nob(&p->state);
+	state = erts_atomic32_read_nob(&p->state);
 
-#ifndef ERTS_DIRTY_SCHEDULERS
-	ASSERT(!p->scheduler_data);
-	p->scheduler_data = esdp;
-#else /* ERTS_DIRTY_SCHEDULERS */
 	if (is_normal_sched) {
+            ASSERT(!p->scheduler_data);
+	    p->scheduler_data = esdp;
 	    if ((!!(state & ERTS_PSFLGS_DIRTY_WORK))
 		& (!(state & ERTS_PSFLG_ACTIVE_SYS))) {
 		/* Migrate to dirty scheduler... */
 	    sunlock_sched_out_proc:
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+		erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 		goto sched_out_proc;
 	    }
-	    ASSERT(!p->scheduler_data);
-	    p->scheduler_data = esdp;
 	}
 	else {
+            if (!(state & ERTS_PSFLGS_DIRTY_WORK)) {
+                /* Dirty work completed... */
+		goto sunlock_sched_out_proc;
+            }
 	    if (state & (ERTS_PSFLG_ACTIVE_SYS
-			 | ERTS_PSFLG_PENDING_EXIT
 			 | ERTS_PSFLG_EXITING)) {
 		/*
 		 * IMPORTANT! We need to take care of
@@ -10853,17 +9656,8 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 		   : (rq == ERTS_DIRTY_IO_RUNQ
 		      && (state & ERTS_PSFLG_DIRTY_IO_PROC)));	    
 	}
-#endif
-
-	if (state & ERTS_PSFLG_PENDING_EXIT) {
-	    erts_handle_pending_exit(p,
-				     ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_STATUS);
-	    state = erts_smp_atomic32_read_nob(&p->state);
-	}
-
-#endif /* ERTS_SMP */
 
-	erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 
         /* Clear tracer if it has been removed */
 	if (IS_TRACED(p) && erts_is_tracer_proc_enabled(
@@ -10884,26 +9678,48 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	}
 
         if (is_normal_sched) {
-
             if (state & ERTS_PSFLG_RUNNING_SYS) {
-                /*
-                 * GC is normally never delayed when a process
-                 * is scheduled out, but might be when executing
-                 * hand written beam assembly in
-                 * prim_eval:'receive'. If GC is delayed we are
-                 * not allowed to execute system tasks.
-                 */
-                if (!(p->flags & F_DELAY_GC)) {
-                    int cost = execute_sys_tasks(p, &state, reds);
-                    calls += cost;
-                    reds -= cost;
-                    if (reds <= 0)
-                        goto sched_out_proc;
-#ifdef ERTS_DIRTY_SCHEDULERS
-                    if (state & ERTS_PSFLGS_DIRTY_WORK)
-                        goto sched_out_proc;
-#endif
+                if (state & (ERTS_PSFLG_SIG_Q|ERTS_PSFLG_SIG_IN_Q)) {
+                    int local_only = (!!(p->flags & F_LOCAL_SIGS_ONLY)
+                                      & !(state & ERTS_PSFLG_SUSPENDED));
+                    if (!local_only | !!(state & ERTS_PSFLG_SIG_Q)) {
+                        int sig_reds;
+                        /*
+                         * If we have dirty work scheduled we allow
+                         * usage of all reductions since we need to
+                         * handle all signals before doing dirty
+                         * work...
+                         */
+                        if (state & ERTS_PSFLGS_DIRTY_WORK)
+                            sig_reds = reds;
+                        else
+                            sig_reds = ERTS_SIG_HANDLE_REDS_MAX_PREFERED;
+                        (void) erts_proc_sig_handle_incoming(p,
+                                                             &state,
+                                                             &sig_reds,
+                                                             sig_reds,
+                                                             local_only);
+                        reds -= sig_reds;
+                    }
                 }
+                if ((state & (ERTS_PSFLG_SYS_TASKS
+                              | ERTS_PSFLG_EXITING)) == ERTS_PSFLG_SYS_TASKS) {
+                    /*
+                     * GC is normally never delayed when a process
+                     * is scheduled out, but might be when executing
+                     * hand written beam assembly in
+                     * prim_eval:'receive'. If GC is delayed we are
+                     * not allowed to execute system tasks.
+                     */
+                    if (!(p->flags & F_DELAY_GC)) {
+                        int cost = execute_sys_tasks(p, &state, reds);
+                        calls += cost;
+                        reds -= cost;
+                    }
+                }
+
+                if (reds <= 0 || (state & ERTS_PSFLGS_DIRTY_WORK))
+                    goto sched_out_proc;
 
                 ASSERT(state & psflg_running_sys);
                 ASSERT(!(state & psflg_running));
@@ -10913,7 +9729,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 
                     if (((state & (ERTS_PSFLG_SUSPENDED
                                    | ERTS_PSFLG_ACTIVE)) != ERTS_PSFLG_ACTIVE)
-                        && !(state & ERTS_PSFLG_EXITING)) {
+                        & !(state & ERTS_PSFLG_EXITING)) {
                         goto sched_out_proc;
                     }
 
@@ -10921,7 +9737,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                     n &= ~psflg_running_sys;
                     n |= psflg_running;
 
-                    state = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
+                    state = erts_atomic32_cmpxchg_mb(&p->state, n, e);
                     if (state == e) {
                         state = n;
                         break;
@@ -10940,10 +9756,8 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
                     reds -= cost;
                     if (reds <= 0)
                         goto sched_out_proc;
-#ifdef ERTS_DIRTY_SCHEDULERS
                     if (p->flags & (F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC))
                         goto sched_out_proc;
-#endif
                 }
             }
         }
@@ -10954,25 +9768,24 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	}
 
 	p->fcalls = reds;
-	    
-	ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
-
-	/* Never run a suspended process */
-#ifdef DEBUG
-        {
-            erts_aint32_t dstate = erts_smp_atomic32_read_nob(&p->state);
-            ASSERT(!(ERTS_PSFLG_SUSPENDED & dstate)
-                   || (ERTS_PSFLG_DIRTY_RUNNING_SYS & dstate));
+        if (reds != context_reds) {
+            actual_reds = context_reds - reds - esdp->virtual_reds;
+            ASSERT(actual_reds >= 0);
+            esdp->virtual_reds = 0;
+            p->reds += actual_reds;
+            ERTS_PROC_REDUCTIONS_EXECUTED(esdp, rq,
+                                          (int) ERTS_PSFLGS_GET_USR_PRIO(state),
+                                          reds,
+                                          actual_reds);
         }
-#endif
+
+	ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
 
 	ASSERT(erts_proc_read_refc(p) > 0);
 
 	if (!(state & ERTS_PSFLG_EXITING) && ERTS_PTMR_IS_TIMED_OUT(p)) {
 	    BeamInstr** pi;
-#ifdef ERTS_SMP
 	    ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
-#endif
 	    pi = (BeamInstr **) p->def_arg_reg;
 	    p->i = *pi;
 	    p->flags &= ~F_INSLPQUEUE;
@@ -10980,7 +9793,49 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls)
 	    ERTS_PTMR_CLEAR(p);
 	}
 
+        /* if exiting, we *shall* exit... */
+        ASSERT(!(state & ERTS_PSFLG_EXITING)
+               || p->i == (BeamInstr *) beam_exit
+               || p->i == (BeamInstr *) beam_continue_exit);
+
+#ifdef DEBUG
+        if (is_normal_sched) {
+	    if (state & ERTS_PSFLGS_DIRTY_WORK)
+                ERTS_INTERNAL_ERROR("Executing dirty code on normal scheduler");
+        }
+        else {
+	    if (!(state & ERTS_PSFLGS_DIRTY_WORK)) {
+                if (esdp->type == ERTS_SCHED_DIRTY_CPU)
+                    ERTS_INTERNAL_ERROR("Executing normal code on dirty CPU scheduler");
+                else if (esdp->type == ERTS_SCHED_DIRTY_IO)
+                    ERTS_INTERNAL_ERROR("Executing normal code on dirty IO scheduler");
+                else
+                    ERTS_INTERNAL_ERROR("Executing normal code on dirty UNKNOWN scheduler");
+            }
+        }
+        {
+            erts_aint32_t dstate = erts_atomic32_read_nob(&p->state);
+
+            /* Never run a suspended process */
+            ASSERT(!(ERTS_PSFLG_SUSPENDED & dstate)
+                   || (ERTS_PSFLG_DIRTY_RUNNING_SYS & dstate));
+
+            /* Do not execute on the wrong type of scheduler... */
+            ASSERT(is_normal_sched
+                   ? !(dstate & ERTS_PSFLGS_DIRTY_WORK)
+                   : !!(dstate & ERTS_PSFLGS_DIRTY_WORK));
+        }
+#endif
+
 	return p;
+
+    sched_out_proc:
+        actual_reds = context_reds;
+        actual_reds -= reds;
+        actual_reds -= esdp->virtual_reds;
+        reds = actual_reds;
+        goto internal_sched_out_proc;
+
     }
 }
 
@@ -10989,13 +9844,11 @@ notify_sys_task_executed(Process *c_p, ErtsProcSysTask *st,
 			 Eterm st_result, int normal_sched)
 {
     Process *rp;
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (!normal_sched)
 	rp = erts_pid2proc_opt(c_p, ERTS_PROC_LOCK_MAIN,
 			       st->requester, 0,
 			       ERTS_P2P_FLG_INC_REFC);
     else
-#endif
 	rp = erts_proc_lookup(st->requester);
     if (rp) {
 	ErtsProcLocks rp_locks;
@@ -11037,18 +9890,16 @@ notify_sys_task_executed(Process *c_p, ErtsProcSysTask *st,
 	ASSERT(hp_start + hsz == hp);
 #endif
 
-	erts_queue_message(rp, rp_locks, mp, msg, c_p->common.id);
+	erts_queue_proc_message(c_p, rp, rp_locks, mp, msg);
 
 	if (c_p == rp)
 	    rp_locks &= ~ERTS_PROC_LOCK_MAIN;
 
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (!normal_sched)
 	    erts_proc_dec_refc(rp);
-#endif
     }
 
     erts_cleanup_offheap(&st->off_heap);
@@ -11067,7 +9918,7 @@ fetch_sys_task(Process *c_p, erts_aint32_t state, int *qmaskp, int *priop)
 
     *priop = -1; /* Shut up annoying erroneous warning */
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
 
     if (!c_p->sys_task_qs) {
 	qmask = 0;
@@ -11183,17 +10034,17 @@ fetch_sys_task(Process *c_p, erts_aint32_t state, int *qmaskp, int *priop)
 	    n |= (prio << ERTS_PSFLGS_ACT_PRIO_OFFSET);
 
 	    if (!qmask)
-		n &= ~ERTS_PSFLG_ACTIVE_SYS;
+		n &= ~ERTS_PSFLG_SYS_TASKS;
 
 	    if (a == n)
 		break;
-	    a = erts_smp_atomic32_cmpxchg_nob(&c_p->state, n, e);
+	    a = erts_atomic32_cmpxchg_nob(&c_p->state, n, e);
 	} while (a != e);
     }
 
 done:
 
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
 
     if (unused_qs)
 	proc_sys_task_queues_free(unused_qs);
@@ -11203,10 +10054,10 @@ done:
     return st;
 }
 
+
+static void exit_permanent_prio_elevation(Process *c_p, erts_aint32_t state);
 static void save_gc_task(Process *c_p, ErtsProcSysTask *st, int prio);
-#ifdef ERTS_DIRTY_SCHEDULERS
 static void save_dirty_task(Process *c_p, ErtsProcSysTask *st);
-#endif
 
 static int
 execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
@@ -11217,7 +10068,7 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
     int qmask = 0;
 
     ASSERT(!ERTS_SCHEDULER_IS_DIRTY(erts_proc_sched_data(c_p)));
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
 
     do {
 	ErtsProcSysTaskType type;
@@ -11225,14 +10076,8 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
 	int st_prio;
 	Eterm st_res;
 
-	if (state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT)) {
-#ifdef ERTS_SMP
-	    if (state & ERTS_PSFLG_PENDING_EXIT)
-		erts_handle_pending_exit(c_p, ERTS_PROC_LOCK_MAIN);
-#endif
-	    ASSERT(ERTS_PROC_IS_EXITING(c_p));
+	if (state & ERTS_PSFLG_EXITING)
 	    break;
-	}
 
 	st = fetch_sys_task(c_p, state, &qmask, &st_prio);
 	if (!st)
@@ -11256,13 +10101,11 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
                         FLAGS(c_p) |= F_NEED_FULLSWEEP;
                     }
 		    reds -= scheduler_gc_proc(c_p, reds);
-#ifdef ERTS_DIRTY_SCHEDULERS
 		    if (c_p->flags & (F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC)) {
 			save_dirty_task(c_p, st);
 			st = NULL;
 			break;
 		    }
-#endif
 		    if (type == ERTS_PSTT_GC_MAJOR)
 			minor_gc = major_gc = 1;
 		    else
@@ -11304,13 +10147,11 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
 						 fcalls, do_gc);
             reds -= cla_reds;
 	    if (is_non_value(st_res)) {
-#ifdef ERTS_DIRTY_SCHEDULERS
 		if (c_p->flags & F_DIRTY_CLA) {
 		    save_dirty_task(c_p, st);
 		    st = NULL;
 		    break;
 		}
-#endif
 		/* Needed gc, but gc was disabled */
 		save_gc_task(c_p, st, st_prio);
 		st = NULL;
@@ -11324,18 +10165,51 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
 	    reds -= erts_complete_off_heap_message_queue_change(c_p);
 	    st_res = am_true;
 	    break;
-#ifdef ERTS_SMP
         case ERTS_PSTT_FTMQ:
 	    reds -= erts_flush_trace_messages(c_p, ERTS_PROC_LOCK_MAIN);
 	    st_res = am_true;
 	    break;
-#endif
-#ifdef ERTS_SMP
         case ERTS_PSTT_ETS_FREE_FIXATION:
 	    reds -= erts_db_execute_free_fixation(c_p, (DbFixation*)st->arg[0]);
 	    st_res = am_true;
 	    break;
-#endif
+        case ERTS_PSTT_PRIO_SIG: {
+            erts_aint32_t fail_state, state;
+            int local_only, sig_res, sig_reds = reds;
+	    st_res = am_false;
+
+            if (st->arg[0] == am_true)
+                local_only = !0;
+            else
+                local_only = 0;
+
+            sig_reds = reds;
+            sig_res = erts_proc_sig_handle_incoming(c_p, &state, &sig_reds,
+                                                    reds, local_only);
+            reds -= sig_reds;
+
+            if (state & ERTS_PSFLG_EXITING) {
+                exit_permanent_prio_elevation(c_p, state);
+                break;
+            }
+
+            if (sig_res)
+                break;
+
+            st->arg[0] = am_true;
+
+            fail_state = ERTS_PSFLG_EXITING;
+
+            if (schedule_process_sys_task(c_p, st_prio, st, &fail_state)) {
+                /* Successfully rescheduled task... */
+                st = NULL;
+            }
+            else {
+                state = erts_atomic32_read_nob(&c_p->state);                         
+                exit_permanent_prio_elevation(c_p, state);
+            }
+            break;
+        }
 	default:
 	    ERTS_INTERNAL_ERROR("Invalid process sys task type");
 	    st_res = am_false;
@@ -11344,7 +10218,7 @@ execute_sys_tasks(Process *c_p, erts_aint32_t *statep, int in_reds)
 	if (st)
 	    reds += notify_sys_task_executed(c_p, st, st_res, 1);
 
-	state = erts_smp_atomic32_read_acqb(&c_p->state);
+	state = erts_atomic32_read_acqb(&c_p->state);
     } while (qmask && reds > 0);
 
     *statep = state;
@@ -11365,20 +10239,18 @@ cleanup_sys_tasks(Process *c_p, erts_aint32_t in_state, int in_reds)
                              * are dirty tasks.
                              */
 
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
 
     do {
 	ErtsProcSysTask *st;
 	Eterm st_res;
 	int st_prio;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 	if (c_p->dirty_sys_tasks) {
 	    st = c_p->dirty_sys_tasks;
 	    c_p->dirty_sys_tasks = st->next;
 	}
 	else
-#endif
 	{
 	    st = fetch_sys_task(c_p, state, &qmask, &st_prio);
 	    if (!st)
@@ -11386,6 +10258,10 @@ cleanup_sys_tasks(Process *c_p, erts_aint32_t in_state, int in_reds)
 	}
 
 	switch (st->type) {
+        case ERTS_PSTT_PRIO_SIG:
+            state = erts_atomic32_read_nob(&c_p->state);                         
+            exit_permanent_prio_elevation(c_p, state);
+            /* fall through... */
         case ERTS_PSTT_GC_MAJOR:
         case ERTS_PSTT_GC_MINOR:
 	case ERTS_PSTT_CPC:
@@ -11396,12 +10272,10 @@ cleanup_sys_tasks(Process *c_p, erts_aint32_t in_state, int in_reds)
 	case ERTS_PSTT_CLA:
 	    st_res = am_ok;
 	    break;
-#ifdef ERTS_SMP
         case ERTS_PSTT_FTMQ:
 	    reds -= erts_flush_trace_messages(c_p, ERTS_PROC_LOCK_MAIN);
 	    st_res = am_true;
 	    break;
-#endif
 	default:
 	    ERTS_INTERNAL_ERROR("Invalid process sys task type");
 	    st_res = am_false;
@@ -11410,13 +10284,41 @@ cleanup_sys_tasks(Process *c_p, erts_aint32_t in_state, int in_reds)
 
 	reds += notify_sys_task_executed(c_p, st, st_res, 1);
 
-	state = erts_smp_atomic32_read_acqb(&c_p->state);
+	state = erts_atomic32_read_acqb(&c_p->state);
     } while (qmask && reds < max_reds);
 
     return reds;
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
+static void
+exit_permanent_prio_elevation(Process *c_p, erts_aint32_t state)
+{
+    erts_aint32_t a;
+    /*
+     * we are about to terminate; permanently elevate
+     * prio in order to ensure high prio signal
+     * handling...
+     */
+    a = state;
+    while (1) {
+        erts_aint32_t aprio, uprio, n, e;
+        ASSERT(a & ERTS_PSFLG_EXITING);
+        aprio = ERTS_PSFLGS_GET_ACT_PRIO(a);
+        uprio = ERTS_PSFLGS_GET_USR_PRIO(a);
+        if (aprio >= uprio)
+            break; /* user prio >= actual prio */
+        /*
+         * actual prio is higher than user prio; raise
+         * user prio to actual prio...
+         */
+        n = e = a;
+        n &= ~ERTS_PSFLGS_USR_PRIO_MASK;
+        n |= aprio << ERTS_PSFLGS_USR_PRIO_OFFSET;
+        a = erts_atomic32_cmpxchg_mb(&c_p->state, n, e);
+        if (a == e)
+            break;
+    }
+}
 
 void
 erts_execute_dirty_system_task(Process *c_p)
@@ -11445,19 +10347,18 @@ erts_execute_dirty_system_task(Process *c_p)
     }
 
     if (c_p->flags & F_DIRTY_GC_HIBERNATE) {
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	if (c_p->msg.len)
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+	if (erts_proc_sig_fetch(c_p))
             c_p->flags &= ~F_DIRTY_GC_HIBERNATE; /* operation aborted... */
         else {
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+	    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
 	    c_p->fvalue = NIL;
 	    erts_garbage_collect_hibernate(c_p);
 	    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+	    erts_proc_lock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
 	    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
 	}
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MSGQ|ERTS_PROC_LOCK_STATUS);
     }
 
     if (c_p->flags & (F_DIRTY_MAJOR_GC|F_DIRTY_MINOR_GC)) {
@@ -11501,7 +10402,7 @@ erts_execute_dirty_system_task(Process *c_p)
 
     }
 
-    erts_smp_atomic32_read_band_relb(&c_p->state, ~ERTS_PSFLG_DIRTY_ACTIVE_SYS);
+    erts_atomic32_read_band_relb(&c_p->state, ~ERTS_PSFLG_DIRTY_ACTIVE_SYS);
 }
 
 static BIF_RETTYPE
@@ -11519,7 +10420,7 @@ dispatch_system_task(Process *c_p, erts_aint_t fail_state,
 
     switch (st->type) {
     case ERTS_PSTT_CPC:
-	rp = erts_dirty_process_code_checker;
+	rp = erts_dirty_process_signal_handler;
 	ASSERT(fail_state & (ERTS_PSFLG_DIRTY_RUNNING
 			     | ERTS_PSFLG_DIRTY_RUNNING_SYS));
 	if (c_p == rp) {
@@ -11551,15 +10452,14 @@ dispatch_system_task(Process *c_p, erts_aint_t fail_state,
     msg = copy_struct(operation, osz, &hp, ohp);
     msg = TUPLE4(hp, st->requester, target, prio, msg);
 
-    erts_queue_message(rp, rp_locks, mp, msg, st->requester);
+    erts_queue_message(rp, rp_locks, mp, msg, am_system);
 
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
 
     return ret;
 }
 
-#endif
 
 static BIF_RETTYPE
 request_system_task(Process *c_p, Eterm requester, Eterm target,
@@ -11664,7 +10564,6 @@ request_system_task(Process *c_p, Eterm requester, Eterm target,
 	st->type = ERTS_PSTT_CPC;
 	if (!rp)
 	    goto noproc;
-#ifdef ERTS_DIRTY_SCHEDULERS
 	/*
 	 * If the process should start executing dirty
 	 * code it is important that this task is
@@ -11672,7 +10571,6 @@ request_system_task(Process *c_p, Eterm requester, Eterm target,
 	 */
 	fail_state |= (ERTS_PSFLG_DIRTY_RUNNING
 		       | ERTS_PSFLG_DIRTY_RUNNING_SYS);
-#endif
 	break;
 
     case am_copy_literals:
@@ -11694,14 +10592,12 @@ request_system_task(Process *c_p, Eterm requester, Eterm target,
 	noproc:
 	    failure = noproc_res;
 	}
-#ifdef ERTS_DIRTY_SCHEDULERS
 	else if (fail_state & (ERTS_PSFLG_DIRTY_RUNNING
 			       | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
 	    ret = dispatch_system_task(c_p, fail_state, st,
 				       target, priority, operation);
 	    goto cleanup_return;
 	}
-#endif
 	else {
 	    ERTS_INTERNAL_ERROR("Unknown failure schedule_process_sys_task()");
 	    failure = am_internal_error;
@@ -11717,9 +10613,7 @@ badarg:
 
     ERTS_BIF_PREP_ERROR(ret, c_p, BADARG);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 cleanup_return:
-#endif
 
     if (st) {
 	erts_cleanup_offheap(&st->off_heap);
@@ -11743,13 +10637,15 @@ erts_internal_request_system_task_4(BIF_ALIST_4)
 			       BIF_ARG_2, BIF_ARG_3, BIF_ARG_4);
 }
 
-static void
-erts_schedule_generic_sys_task(Eterm pid, ErtsProcSysTaskType type, void* arg)
+static int
+schedule_generic_sys_task(Eterm pid, ErtsProcSysTaskType type,
+                          int prio, Eterm arg0, Eterm arg1)
 {
-    Process *rp = erts_proc_lookup(pid);
+    int res = 0;
+    Process *rp = erts_proc_lookup_raw(pid);
     if (rp) {
 	ErtsProcSysTask *st;
-	erts_aint32_t state, fail_state;
+	erts_aint32_t st_prio, fail_state;
 
 	st = erts_alloc(ERTS_ALC_T_PROC_SYS_TSK,
 			ERTS_PROC_SYS_TASK_SIZE(0));
@@ -11758,32 +10654,46 @@ erts_schedule_generic_sys_task(Eterm pid, ErtsProcSysTaskType type, void* arg)
 	st->reply_tag = NIL;
 	st->req_id = NIL;
 	st->req_id_sz = 0;
-        st->arg[0] = (Eterm)arg;
+        st->arg[0] = arg0;
+        st->arg[1] = arg1;
 	ERTS_INIT_OFF_HEAP(&st->off_heap);
-	state = erts_smp_atomic32_read_nob(&rp->state);
-
-	fail_state = ERTS_PSFLG_EXITING;
 
-	if (!schedule_process_sys_task(rp, ERTS_PSFLGS_GET_USR_PRIO(state),
-				       st, &fail_state))
+        if (prio >= 0) {
+            st_prio = (erts_aint32_t) prio;
+            fail_state = ERTS_PSFLG_FREE;
+        }
+        else {
+            erts_aint32_t state = erts_atomic32_read_nob(&rp->state);
+            st_prio = ERTS_PSFLGS_GET_USR_PRIO(state);
+            fail_state = ERTS_PSFLG_EXITING;
+        }
+        res = schedule_process_sys_task(rp, st_prio, st, &fail_state);
+	if (!res)
 	    erts_free(ERTS_ALC_T_PROC_SYS_TSK, st);
     }
+    return res;
 }
 
-
 void
 erts_schedule_complete_off_heap_message_queue_change(Eterm pid)
 {
-    erts_schedule_generic_sys_task(pid, ERTS_PSTT_COHMQ, NULL);
+    schedule_generic_sys_task(pid, ERTS_PSTT_COHMQ,
+                              -1, NIL, NIL);
 }
 
 void
 erts_schedule_ets_free_fixation(Eterm pid, DbFixation* fix)
 {
-    erts_schedule_generic_sys_task(pid, ERTS_PSTT_ETS_FREE_FIXATION, fix);
+    schedule_generic_sys_task(pid, ERTS_PSTT_ETS_FREE_FIXATION,
+                              -1, (Eterm) fix, NIL);
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
+int
+erts_sig_prio(Eterm pid, int prio)
+{
+    return schedule_generic_sys_task(pid, ERTS_PSTT_PRIO_SIG,
+                                     prio, am_false, NIL);
+}
 
 static void
 flush_dirty_trace_messages(void *vpid)
@@ -11800,45 +10710,35 @@ flush_dirty_trace_messages(void *vpid)
     proc = erts_pid2proc_opt(NULL, 0, pid, ERTS_PROC_LOCK_MAIN, 0);
     if (proc) {
 	(void) erts_flush_trace_messages(proc, ERTS_PROC_LOCK_MAIN);
-        erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+        erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
     }
 }
 
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 void
 erts_schedule_flush_trace_messages(Process *proc, int force_on_proc)
 {
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl;
-#endif
     Eterm pid = proc->common.id;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     erts_aint32_t state;
 
     if (!force_on_proc) {
-	state = erts_smp_atomic32_read_nob(&proc->state);
+	state = erts_atomic32_read_nob(&proc->state);
 	if (state & (ERTS_PSFLG_DIRTY_RUNNING
 		     | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
 	    goto sched_flush_dirty;
 	}
     }
-#endif
 
-#ifdef ERTS_SMP
     dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
-    erts_schedule_generic_sys_task(pid, ERTS_PSTT_FTMQ, NULL);
+    schedule_generic_sys_task(pid, ERTS_PSTT_FTMQ, -1, NIL, NIL);
 
-#ifdef ERTS_SMP
     erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (!force_on_proc) {
-	state = erts_smp_atomic32_read_mb(&proc->state);
+	state = erts_atomic32_read_mb(&proc->state);
 	if (state & (ERTS_PSFLG_DIRTY_RUNNING
 		     | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
 	    void *vargp;
@@ -11866,7 +10766,6 @@ erts_schedule_flush_trace_messages(Process *proc, int force_on_proc)
 	    erts_schedule_misc_aux_work(1, flush_dirty_trace_messages, vargp);
 	}
     }
-#endif
 }
 
 static void
@@ -11875,7 +10774,7 @@ save_gc_task(Process *c_p, ErtsProcSysTask *st, int prio)
     erts_aint32_t state;
     ErtsProcSysTaskQs *qs;
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
 
     qs = ERTS_PROC_GET_DELAYED_GC_TASK_QS(c_p);
     if (!qs) {
@@ -11905,7 +10804,7 @@ save_gc_task(Process *c_p, ErtsProcSysTask *st, int prio)
 	}
     }
 
-    state = erts_smp_atomic32_read_nob(&c_p->state);
+    state = erts_atomic32_read_nob(&c_p->state);
     ASSERT((ERTS_PSFLG_RUNNING
 	    | ERTS_PSFLG_RUNNING_SYS
 	    | ERTS_PSFLG_DIRTY_RUNNING
@@ -11921,20 +10820,18 @@ save_gc_task(Process *c_p, ErtsProcSysTask *st, int prio)
 	    n &= ~ERTS_PSFLGS_ACT_PRIO_MASK;
 	    n |= prio << ERTS_PSFLGS_ACT_PRIO_OFFSET;
 	}
-	state = erts_smp_atomic32_cmpxchg_relb(&c_p->state, n, e);
+	state = erts_atomic32_cmpxchg_relb(&c_p->state, n, e);
 	if (state == e)
 	    break;
     }
 }
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 static void
 save_dirty_task(Process *c_p, ErtsProcSysTask *st)
 {
     st->next = c_p->dirty_sys_tasks;
     c_p->dirty_sys_tasks = st;
 }
-#endif
 
 int
 erts_set_gc_state(Process *c_p, int enable)
@@ -11942,8 +10839,8 @@ erts_set_gc_state(Process *c_p, int enable)
     ErtsProcSysTaskQs *dgc_tsk_qs;
     ASSERT(c_p == erts_get_current_process());
     ASSERT((ERTS_PSFLG_RUNNING|ERTS_PSFLG_RUNNING_SYS)
-	   & erts_smp_atomic32_read_nob(&c_p->state));
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
+	   & erts_atomic32_read_nob(&c_p->state));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(c_p));
 
     if (!enable) {
 	c_p->flags |= F_DISABLE_GC;
@@ -11958,7 +10855,7 @@ erts_set_gc_state(Process *c_p, int enable)
 
     /* Move delayed gc tasks into sys tasks queues. */
 
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_STATUS);
 
     if (!c_p->sys_task_qs) {
 	c_p->sys_task_qs = dgc_tsk_qs;
@@ -12031,10 +10928,12 @@ erts_set_gc_state(Process *c_p, int enable)
 	erts_aint32_t aprio, state =
 #endif
 
-	    erts_smp_atomic32_read_bset_nob(&c_p->state,
-					    (ERTS_PSFLG_DELAYED_SYS
-					     | ERTS_PSFLG_ACTIVE_SYS),
-					    ERTS_PSFLG_ACTIVE_SYS);
+	    erts_atomic32_read_bset_nob(&c_p->state,
+                                        (ERTS_PSFLG_DELAYED_SYS
+                                         | ERTS_PSFLG_ACTIVE_SYS
+                                         | ERTS_PSFLG_SYS_TASKS),
+                                        (ERTS_PSFLG_ACTIVE_SYS
+                                         | ERTS_PSFLG_SYS_TASKS));
 
 #ifdef DEBUG
 	ASSERT(state & ERTS_PSFLG_DELAYED_SYS);
@@ -12045,7 +10944,7 @@ erts_set_gc_state(Process *c_p, int enable)
     }
 #endif
 
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_STATUS);
 
     (void) ERTS_PROC_SET_DELAYED_GC_TASK_QS(c_p, NULL);
 
@@ -12061,24 +10960,24 @@ erts_sched_stat_modify(int what)
     int ix;
     switch (what) {
     case ERTS_SCHED_STAT_MODIFY_ENABLE:
-	erts_smp_thr_progress_block();
+	erts_thr_progress_block();
 	erts_sched_stat.enabled = 1;
-	erts_smp_thr_progress_unblock();
+	erts_thr_progress_unblock();
 	break;
     case ERTS_SCHED_STAT_MODIFY_DISABLE:
-	erts_smp_thr_progress_block();
+	erts_thr_progress_block();
 	erts_sched_stat.enabled = 0;
-	erts_smp_thr_progress_unblock();
+	erts_thr_progress_unblock();
 	break;
     case ERTS_SCHED_STAT_MODIFY_CLEAR:
-	erts_smp_spin_lock(&erts_sched_stat.lock);
+	erts_spin_lock(&erts_sched_stat.lock);
 	for (ix = 0; ix < ERTS_NO_PRIO_LEVELS; ix++) {
 	    erts_sched_stat.prio[ix].total_executed = 0;
 	    erts_sched_stat.prio[ix].executed = 0;
 	    erts_sched_stat.prio[ix].total_migrated = 0;
 	    erts_sched_stat.prio[ix].migrated = 0;
 	}
-	erts_smp_spin_unlock(&erts_sched_stat.lock);
+	erts_spin_unlock(&erts_sched_stat.lock);
 	break;
     }
 }
@@ -12092,7 +10991,7 @@ erts_sched_stat_term(Process *p, int total)
     Uint executed[ERTS_NO_PRIO_LEVELS];
     Uint migrated[ERTS_NO_PRIO_LEVELS];
 
-    erts_smp_spin_lock(&erts_sched_stat.lock);
+    erts_spin_lock(&erts_sched_stat.lock);
     if (total) {
 	int i;
 	for (i = 0; i < ERTS_NO_PRIO_LEVELS; i++) {
@@ -12111,7 +11010,7 @@ erts_sched_stat_term(Process *p, int total)
 	    erts_sched_stat.prio[i].migrated = 0;
 	}
     }
-    erts_smp_spin_unlock(&erts_sched_stat.lock);
+    erts_spin_unlock(&erts_sched_stat.lock);
 
     sz = 0;
     (void) erts_bld_atom_2uint_3tup_list(NULL, &sz, ERTS_NO_PRIO_LEVELS,
@@ -12131,7 +11030,6 @@ erts_schedule_misc_op(void (*func)(void *), void *arg)
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
     ErtsRunQueue *rq = esdp ? esdp->run_queue : ERTS_RUNQ_IX(0);
     ErtsMiscOpList *molp = misc_op_list_alloc();
-#ifdef ERTS_SMP
     ErtsMigrationPaths *mpaths = erts_get_migration_paths();
 
     if (!mpaths)
@@ -12141,9 +11039,8 @@ erts_schedule_misc_op(void (*func)(void *), void *arg)
 	if (erq)
 	    rq = erq;
     }
-#endif
 
-    erts_smp_runq_lock(rq);
+    erts_runq_lock(rq);
 
     molp->next = NULL;
     molp->func = func;
@@ -12154,13 +11051,11 @@ erts_schedule_misc_op(void (*func)(void *), void *arg)
 	rq->misc.start = molp;
     rq->misc.end = molp;
 
-#ifdef ERTS_SMP
     non_empty_runq(rq);
-#endif
 
     ERTS_RUNQ_FLGS_SET_NOB(rq, ERTS_RUNQ_FLG_MISC_OP);
 
-    erts_smp_runq_unlock(rq);
+    erts_runq_unlock(rq);
 
     smp_notify_inc_runq(rq);
 }
@@ -12193,7 +11088,7 @@ exec_misc_ops(ErtsRunQueue *rq)
     if (!rq->misc.start)
         ERTS_RUNQ_FLGS_UNSET_NOB(rq, ERTS_RUNQ_FLG_MISC_OP);
 
-    erts_smp_runq_unlock(rq);
+    erts_runq_unlock(rq);
 
     while (molp) {
 	tmp_molp = molp;
@@ -12202,7 +11097,7 @@ exec_misc_ops(ErtsRunQueue *rq)
 	misc_op_list_free(tmp_molp);
     }
 
-    erts_smp_runq_lock(rq);
+    erts_runq_lock(rq);
 }
 
 Uint
@@ -12233,12 +11128,12 @@ erts_get_exact_total_reductions(Process *c_p, Uint *redsp, Uint *diffp)
 {
     Uint reds = erts_current_reductions(c_p, c_p);
     int ix;
-    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
     /*
      * Wait for other schedulers to schedule out their processes
      * and update 'reductions'.
      */
-    erts_smp_thr_progress_block();
+    erts_thr_progress_block();
     for (reds = 0, ix = 0; ix < erts_no_run_queues; ix++)
 	reds += ERTS_RUNQ_IX(ix)->procs.reductions;
     if (redsp)
@@ -12246,8 +11141,8 @@ erts_get_exact_total_reductions(Process *c_p, Uint *redsp, Uint *diffp)
     if (diffp)
 	*diffp = reds - last_exact_reductions;
     last_exact_reductions = reds;
-    erts_smp_thr_progress_unblock();
-    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+    erts_thr_progress_unblock();
+    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 }
 
 static void delete_process(Process* p);
@@ -12255,10 +11150,8 @@ static void delete_process(Process* p);
 void
 erts_free_proc(Process *p)
 {
-#ifdef ERTS_SMP
     erts_proc_lock_fin(p);
-#endif
-    ASSERT(erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_FREE);
+    ASSERT(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_FREE);
     ASSERT(0 == erts_proc_read_refc(p));
     if (p->flags & F_DELAYED_DEL_PROC)
 	delete_process(p);
@@ -12269,6 +11162,7 @@ typedef struct {
     Process *proc;
     erts_aint32_t state;
     ErtsRunQueue *run_queue;
+    int bound;
 } ErtsEarlyProcInit;
 
 static void early_init_process_struct(void *varg, Eterm data)
@@ -12277,17 +11171,12 @@ static void early_init_process_struct(void *varg, Eterm data)
     Process *proc = arg->proc;
 
     proc->common.id = make_internal_pid(data);
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_atomic32_init_nob(&proc->dirty_state, 0);
+    erts_atomic32_init_nob(&proc->dirty_state, 0);
     proc->dirty_sys_tasks = NULL;
-#endif
-    erts_smp_atomic32_init_relb(&proc->state, arg->state);
-
-#ifdef ERTS_SMP
-    RUNQ_SET_RQ(&proc->run_queue, arg->run_queue);
+    erts_init_runq_proc(proc, arg->run_queue, arg->bound);
+    erts_atomic32_init_relb(&proc->state, arg->state);
 
     erts_proc_lock_init(proc); /* All locks locked */
-#endif
 
 }
 
@@ -12295,7 +11184,7 @@ static void early_init_process_struct(void *varg, Eterm data)
 ** Allocate process and find out where to place next process.
 */
 static Process*
-alloc_process(ErtsRunQueue *rq, erts_aint32_t state)
+alloc_process(ErtsRunQueue *rq, int bound, erts_aint32_t state)
 {
     ErtsEarlyProcInit init_arg;
     Process *p;
@@ -12304,9 +11193,12 @@ alloc_process(ErtsRunQueue *rq, erts_aint32_t state)
     if (!p)
 	return NULL;
 
+    ASSERT(rq);
+
     init_arg.proc = (Process *) p;
-    init_arg.run_queue = rq;
     init_arg.state = state;
+    init_arg.run_queue = rq;
+    init_arg.bound = bound;
 
     ERTS_CT_ASSERT(offsetof(Process,common) == 0);
 
@@ -12322,7 +11214,6 @@ alloc_process(ErtsRunQueue *rq, erts_aint32_t state)
 
     ASSERT(internal_pid_serial(p->common.id) <= ERTS_MAX_PID_SERIAL);
     
-    p->approx_started = erts_get_approx_time();
     p->rcount = 0;
     p->heap = NULL;
 
@@ -12341,6 +11232,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 		   Eterm args,	/* Arguments for function (must be well-formed list). */
 		   ErlSpawnOpts* so) /* Options for spawn. */
 {
+    int bound = 0;
     Uint flags = 0;
     ErtsRunQueue *rq = NULL;
     Process *p;
@@ -12360,7 +11252,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
     INITIALIZE_LITERAL_PURGE_AREA(litarea);
 #endif
 
-    erts_smp_proc_lock(parent, ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_proc_lock(parent, ERTS_PROC_LOCKS_ALL_MINOR);
 
     /*
      * Check for errors.
@@ -12377,7 +11269,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 	    ASSERT(0 <= ix && ix < erts_no_run_queues);
 	    rq = ERTS_RUNQ_IX(ix);
 	    /* Unsupported feature... */
-	    state |= ERTS_PSFLG_BOUND;
+            bound = !0;
 	}
 	prio = (erts_aint32_t) so->priority;
     }
@@ -12390,17 +11282,16 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 	flags |= F_OFF_HEAP_MSGQ;
     }
     else if (so->flags & SPO_ON_HEAP_MSGQ) {
-	state |= ERTS_PSFLG_ON_HEAP_MSGQ;
 	flags |= F_ON_HEAP_MSGQ;
     }
 
     ASSERT((flags & F_ON_HEAP_MSGQ) || (flags & F_OFF_HEAP_MSGQ));
 
     if (!rq)
-	rq = erts_get_runq_proc(parent);
+	rq = erts_get_runq_proc(parent, NULL);
 
-    p = alloc_process(rq, state); /* All proc locks are locked by this thread
-				     on success */
+    p = alloc_process(rq, bound, state); /* All proc locks are locked by this thread
+                                            on success */
     if (!p) {
 	erts_send_error_to_logger_str(parent->group_leader,
 				      "Too many processes\n");
@@ -12408,11 +11299,6 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 	goto error;
     }
 
-    ASSERT((erts_smp_atomic32_read_nob(&p->state)
-	    & ERTS_PSFLG_ON_HEAP_MSGQ)
-	   || (erts_smp_atomic32_read_nob(&p->state)
-	       & ERTS_PSFLG_OFF_HEAP_MSGQ));
-
 #ifdef SHCOPY_SPAWN
     arg_size = copy_shared_calculate(args, &info);
 #else
@@ -12436,7 +11322,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 	p->min_vheap_size = BIN_VH_MIN_SIZE;
         MAX_HEAP_SIZE_SET(p, H_MAX_SIZE);
         MAX_HEAP_SIZE_FLAGS_SET(p, H_MAX_FLAGS);
-	p->max_gen_gcs    = (Uint16) erts_smp_atomic32_read_nob(&erts_max_gen_gcs);
+	p->max_gen_gcs    = (Uint16) erts_atomic32_read_nob(&erts_max_gen_gcs);
     }
     p->schedule_count = 0;
     ASSERT(p->min_heap_size == erts_next_heap_size(p->min_heap_size, 0));
@@ -12462,9 +11348,6 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 
 #ifdef HIPE
     hipe_init_process(&p->hipe);
-#ifdef ERTS_SMP
-    hipe_init_process_smp(&p->hipe_smp);
-#endif
 #endif
     p->heap = (Eterm *) ERTS_HEAP_ALLOC(ERTS_ALC_T_HEAP, sizeof(Eterm)*sz);
     p->old_hend = p->old_htop = p->old_heap = NULL;
@@ -12512,8 +11395,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
     p->common.u.alive.reg = NULL;
     ERTS_P_LINKS(p) = NULL;
     ERTS_P_MONITORS(p) = NULL;
-    p->nodes_monitors = NULL;
-    p->suspend_monitors = NULL;
+    ERTS_P_LT_MONITORS(p) = NULL;
 
     ASSERT(is_pid(parent->group_leader));
 
@@ -12529,21 +11411,28 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 
     erts_get_default_proc_tracing(&ERTS_TRACE_FLAGS(p), &ERTS_TRACER(p));
 
-    p->msg.first = NULL;
-    p->msg.last = &p->msg.first;
-    p->msg.save = &p->msg.first;
-    p->msg.saved_last = &p->msg.first;
-    p->msg.len = 0;
-#ifdef ERTS_SMP
-    p->msg_inq.first = NULL;
-    p->msg_inq.last = &p->msg_inq.first;
-    p->msg_inq.len = 0;
+    p->sig_qs.first = NULL;
+    p->sig_qs.last = &p->sig_qs.first;
+    p->sig_qs.cont = NULL;
+    p->sig_qs.cont_last = &p->sig_qs.cont;
+    p->sig_qs.save = &p->sig_qs.first;
+    p->sig_qs.saved_last = NULL;
+    p->sig_qs.len = 0;
+    p->sig_qs.nmsigs.next = NULL;
+    p->sig_qs.nmsigs.last = NULL;
+    p->sig_inq.first = NULL;
+    p->sig_inq.last = &p->sig_inq.first;
+    p->sig_inq.len = 0;
+    p->sig_inq.nmsigs.next = NULL;
+    p->sig_inq.nmsigs.last = NULL;
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+    p->sig_inq.may_contain_heap_terms = 0;
 #endif
     p->bif_timers = NULL;
     p->mbuf = NULL;
     p->msg_frag = NULL;
     p->mbuf_sz = 0;
-    erts_smp_atomic_init_nob(&p->psd, (erts_aint_t) NULL);
+    erts_atomic_init_nob(&p->psd, (erts_aint_t) NULL);
     p->dictionary = NULL;
     p->seq_trace_lastcnt = 0;
     p->seq_trace_clock = 0;
@@ -12561,14 +11450,8 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
     p->last_old_htop = NULL;
 #endif
 
-#ifdef ERTS_SMP
     p->trace_msg_q = NULL;
     p->scheduler_data = NULL;
-    p->suspendee = NIL;
-    p->pending_suspenders = NULL;
-    p->pending_exit.reason = THE_NON_VALUE;
-    p->pending_exit.bp = NULL;
-#endif
 
 #if !defined(NO_FPE_SIGNALS) || defined(HIPE)
     p->fp_exception = 0;
@@ -12596,8 +11479,8 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
         }
         if (ARE_TRACE_FLAGS_ON(parent, F_TRACE_PROCS)) {
             locks &= ~(ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
-            erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
-            erts_smp_proc_unlock(parent, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+            erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+            erts_proc_unlock(parent, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
             trace_proc_spawn(parent, am_spawn, p->common.id, mod, func, args);
             if (so->flags & SPO_LINK)
                 trace_proc(parent, locks, parent, am_link, p->common.id);
@@ -12609,8 +11492,8 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
               == (ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE)) {
             /* This happens when parent was not traced, but child is */
             locks &= ~(ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
-            erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
-            erts_smp_proc_unlock(parent, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+            erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
+            erts_proc_unlock(parent, ERTS_PROC_LOCK_STATUS|ERTS_PROC_LOCK_TRACE);
         }
         trace_proc_spawn(p, am_spawned, parent->common.id, mod, func, args);
         if (so->flags & SPO_LINK)
@@ -12622,34 +11505,37 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
      */
 
     if (so->flags & SPO_LINK) {
-#ifdef DEBUG
-	int ret;
-#endif
-#ifdef DEBUG
-	ret = erts_add_link(&ERTS_P_LINKS(parent),  LINK_PID, p->common.id);
-	ASSERT(ret == 0);
-	ret = erts_add_link(&ERTS_P_LINKS(p), LINK_PID, parent->common.id);
-	ASSERT(ret == 0);
-#else	
-	erts_add_link(&ERTS_P_LINKS(parent), LINK_PID, p->common.id);
-	erts_add_link(&ERTS_P_LINKS(p), LINK_PID, parent->common.id);
-#endif
-
+        ErtsLink *lnk;
+        ErtsLinkData *ldp = erts_link_create(ERTS_LNK_TYPE_PROC,
+                                             parent->common.id,
+                                             p->common.id);
+        lnk = erts_link_tree_lookup_insert(&ERTS_P_LINKS(parent), &ldp->a);
+        if (lnk) {
+            /*
+             * This should more or less never happen, but could
+             * potentially happen if pid:s wrap...
+             */
+            erts_link_release(lnk);
+        }
+        erts_link_tree_insert(&ERTS_P_LINKS(p), &ldp->b);
     }
 
     /*
      * Test whether this process should be initially monitored by its parent.
      */
     if (so->flags & SPO_MONITOR) {
-	Eterm mref;
-
-	mref = erts_make_ref(parent);
-	erts_add_monitor(&ERTS_P_MONITORS(parent), MON_ORIGIN, mref, p->common.id, NIL);
-	erts_add_monitor(&ERTS_P_MONITORS(p), MON_TARGET, mref, parent->common.id, NIL);
+	Eterm mref = erts_make_ref(parent);
+        ErtsMonitorData *mdp = erts_monitor_create(ERTS_MON_TYPE_PROC,
+                                                   mref,
+                                                   parent->common.id,
+                                                   p->common.id,
+                                                   NIL);
+        erts_monitor_tree_insert(&ERTS_P_MONITORS(parent), &mdp->origin);
+        erts_monitor_list_insert(&ERTS_P_LT_MONITORS(p), &mdp->target);
 	so->mref = mref;
     }
 
-    erts_smp_proc_unlock(p, locks);
+    erts_proc_unlock(p, locks);
 
     res = p->common.id;
 
@@ -12657,7 +11543,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
      * Schedule process for execution.
      */
 
-    erts_smp_proc_unlock(parent, locks & ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_proc_unlock(parent, locks & ERTS_PROC_LOCKS_ALL_MINOR);
 
     schedule_process(p, state, 0);
 
@@ -12677,7 +11563,7 @@ erl_create_process(Process* parent, /* Parent of process (default group leader).
 
  error:
 
-    erts_smp_proc_unlock(parent, locks & ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_proc_unlock(parent, locks & ERTS_PROC_LOCKS_ALL_MINOR);
 
     return res;
 }
@@ -12728,15 +11614,27 @@ void erts_init_empty_process(Process *p)
     p->mbuf = NULL;
     p->msg_frag = NULL;
     p->mbuf_sz = 0;
-    erts_smp_atomic_init_nob(&p->psd, (erts_aint_t) NULL);
+    erts_atomic_init_nob(&p->psd, (erts_aint_t) NULL);
     ERTS_P_MONITORS(p) = NULL;
+    ERTS_P_LT_MONITORS(p) = NULL;
     ERTS_P_LINKS(p) = NULL;         /* List of links */
-    p->nodes_monitors = NULL;
-    p->suspend_monitors = NULL;
-    p->msg.first = NULL;
-    p->msg.last = &p->msg.first;
-    p->msg.save = &p->msg.first;
-    p->msg.len = 0;
+    p->sig_qs.first = NULL;
+    p->sig_qs.last = &p->sig_qs.first;
+    p->sig_qs.cont = NULL;
+    p->sig_qs.cont_last = &p->sig_qs.cont;
+    p->sig_qs.save = &p->sig_qs.first;
+    p->sig_qs.saved_last = NULL;
+    p->sig_qs.len = 0;
+    p->sig_qs.nmsigs.next = NULL;
+    p->sig_qs.nmsigs.last = NULL;
+    p->sig_inq.first = NULL;
+    p->sig_inq.last = &p->sig_inq.first;
+    p->sig_inq.len = 0;
+    p->sig_inq.nmsigs.next = NULL;
+    p->sig_inq.nmsigs.last = NULL;
+#ifdef ERTS_PROC_SIG_HARD_DEBUG
+    p->sig_inq.may_contain_heap_terms = 0;
+#endif
     p->bif_timers = NULL;
     p->dictionary = NULL;
     p->seq_trace_clock = 0;
@@ -12764,16 +11662,12 @@ void erts_init_empty_process(Process *p)
     p->def_arg_reg[5] = 0;
 
     p->parent = NIL;
-    p->approx_started = 0;
     p->static_flags = 0;
 
     p->common.u.alive.started_interval = 0;
 
 #ifdef HIPE
     hipe_init_process(&p->hipe);
-#ifdef ERTS_SMP
-    hipe_init_process_smp(&p->hipe_smp);
-#endif
 #endif
 
     INIT_HOLE_CHECK(p);
@@ -12781,25 +11675,14 @@ void erts_init_empty_process(Process *p)
     p->last_old_htop = NULL;
 #endif
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_atomic32_init_nob(&p->dirty_state, 0);
+    erts_atomic32_init_nob(&p->dirty_state, 0);
     p->dirty_sys_tasks = NULL;
-#endif
-    erts_smp_atomic32_init_nob(&p->state, (erts_aint32_t) PRIORITY_NORMAL);
+    erts_atomic32_init_nob(&p->state, (erts_aint32_t) PRIORITY_NORMAL);
 
-#ifdef ERTS_SMP
     p->scheduler_data = NULL;
-    p->msg_inq.first = NULL;
-    p->msg_inq.last = &p->msg_inq.first;
-    p->msg_inq.len = 0;
-    p->suspendee = NIL;
-    p->pending_suspenders = NULL;
-    p->pending_exit.reason = THE_NON_VALUE;
-    p->pending_exit.bp = NULL;
     erts_proc_lock_init(p);
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_ALL);
-    RUNQ_SET_RQ(&p->run_queue, ERTS_RUNQ_IX(0));
-#endif
+    erts_proc_unlock(p, ERTS_PROC_LOCKS_ALL);
+    erts_init_runq_proc(p, ERTS_RUNQ_IX(0), 0);
 
 #if !defined(NO_FPE_SIGNALS) || defined(HIPE)
     p->fp_exception = 0;
@@ -12832,11 +11715,10 @@ erts_debug_verify_clean_empty_process(Process* p)
     ASSERT(p->old_heap == NULL);
 
     ASSERT(ERTS_P_MONITORS(p) == NULL);
+    ASSERT(ERTS_P_LT_MONITORS(p) == NULL);
     ASSERT(ERTS_P_LINKS(p) == NULL);
-    ASSERT(p->nodes_monitors == NULL);
-    ASSERT(p->suspend_monitors == NULL);
-    ASSERT(p->msg.first == NULL);
-    ASSERT(p->msg.len == 0);
+    ASSERT(p->sig_qs.first == NULL);
+    ASSERT(p->sig_qs.len == 0);
     ASSERT(p->bif_timers == NULL);
     ASSERT(p->dictionary == NULL);
     ASSERT(p->catches == 0);
@@ -12846,14 +11728,8 @@ erts_debug_verify_clean_empty_process(Process* p)
 
     ASSERT(p->parent == NIL);
 
-#ifdef ERTS_SMP
-    ASSERT(p->msg_inq.first == NULL);
-    ASSERT(p->msg_inq.len == 0);
-    ASSERT(p->suspendee == NIL);
-    ASSERT(p->pending_suspenders == NULL);
-    ASSERT(p->pending_exit.reason == THE_NON_VALUE);
-    ASSERT(p->pending_exit.bp == NULL);
-#endif
+    ASSERT(p->sig_inq.first == NULL);
+    ASSERT(p->sig_inq.len == 0);
 
     /* Thing that erts_cleanup_empty_process() cleans up */
 
@@ -12878,9 +11754,7 @@ erts_cleanup_empty_process(Process* p)
 	free_message_buffer(p->mbuf);
 	p->mbuf = NULL;
     }
-#ifdef ERTS_SMP
     erts_proc_lock_fin(p);
-#endif
 #ifdef DEBUG
     erts_debug_verify_clean_empty_process(p);
 #endif
@@ -12912,10 +11786,10 @@ delete_process(Process* p)
 
     /* Cleanup psd */
 
-    psd = (ErtsPSD *) erts_smp_atomic_read_nob(&p->psd);
+    psd = (ErtsPSD *) erts_atomic_read_nob(&p->psd);
 
     if (psd) {
-	erts_smp_atomic_set_nob(&p->psd, (erts_aint_t) NULL); /* Reduction counting depends on this... */
+	erts_atomic_set_nob(&p->psd, (erts_aint_t) NULL); /* Reduction counting depends on this... */
 	erts_free(ERTS_ALC_T_PSD, psd);
     }
 
@@ -12956,848 +11830,313 @@ delete_process(Process* p)
     erts_erase_dicts(p);
 
     /* free all pending messages */
-    erts_cleanup_messages(p->msg.first);
-    p->msg.first = NULL;
-
-    ASSERT(!p->nodes_monitors);
-    ASSERT(!p->suspend_monitors);
+    erts_cleanup_messages(p->sig_qs.first);
+    p->sig_qs.first = NULL;
+    erts_cleanup_messages(p->sig_qs.cont);
+    p->sig_qs.cont = NULL;
 
     p->fvalue = NIL;
 }
 
 static ERTS_INLINE void
-set_proc_exiting(Process *p,
-		 erts_aint32_t in_state,
-		 Eterm reason,
-		 ErlHeapFragment *bp)
+set_self_exiting(Process *c_p, Eterm reason, int *enqueue,
+                 erts_aint32_t *prio, erts_aint32_t *state)
 {
-    erts_aint32_t state = in_state, enq_prio = -1;
-    int enqueue;
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(p) == ERTS_PROC_LOCKS_ALL);
+    erts_aint32_t st, enq_prio = -1;
+    int enq;
 
-    enqueue = change_proc_schedule_state(p,
-					 (ERTS_PSFLG_SUSPENDED
-					  | ERTS_PSFLG_PENDING_EXIT
-					  | ERTS_PSFLGS_DIRTY_WORK),
-					 ERTS_PSFLG_EXITING|ERTS_PSFLG_ACTIVE,
-					 &state,
-					 &enq_prio,
-					 ERTS_PROC_LOCKS_ALL);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCKS_ALL);
 
-    p->fvalue = reason;
-    if (bp)
-	erts_link_mbuf_to_proc(p, bp);
-    /*
-     * We used to set freason to EXC_EXIT here, but there is no need to
-     * save the stack trace since this process irreversibly is going to
-     * exit.
-     */
-    p->freason = EXTAG_EXIT;
-    KILL_CATCHES(p);
-    p->i = (BeamInstr *) beam_exit;
+    c_p->fvalue = reason;
 
-#ifndef ERTS_SMP
-    if (state & (ERTS_PSFLG_RUNNING|ERTS_PSFLG_RUNNING_SYS)
-        && !(state & ERTS_PSFLG_GC)) {
-	/*
-	 * I non smp case:
-	 *
-	 * Currently executing process might be sent an exit
-	 * signal if it is traced by a port that it also is
-	 * linked to, and the port terminates during the
-	 * trace. In this case we want schedule out the
-	 * process as quickly as possible in order to detect
-	 * the event as fast as possible.
-	 */
-	ERTS_VBUMP_ALL_REDS(p);
-    }
-#endif
-
-    add2runq(enqueue, enq_prio, p, state, NULL);
-}
-
-static ERTS_INLINE erts_aint32_t
-set_proc_self_exiting(Process *c_p)
-{
-#ifdef DEBUG
-    int enqueue;
-#endif
-    erts_aint32_t state, enq_prio = -1;
-
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCKS_ALL);
+    st = erts_atomic32_read_nob(&c_p->state);
+    ASSERT(enqueue || (st & (ERTS_PSFLG_RUNNING
+                             |ERTS_PSFLG_RUNNING_SYS
+                             | ERTS_PSFLG_DIRTY_RUNNING
+                             | ERTS_PSFLG_DIRTY_RUNNING_SYS)));
 
-    state = erts_smp_atomic32_read_nob(&c_p->state);
-    ASSERT(state & (ERTS_PSFLG_RUNNING
-		    |ERTS_PSFLG_RUNNING_SYS
-		    | ERTS_PSFLG_DIRTY_RUNNING
-		    | ERTS_PSFLG_DIRTY_RUNNING_SYS));
+    enq = change_proc_schedule_state(c_p,
+                                     (ERTS_PSFLG_SUSPENDED
+                                      | ERTS_PSFLGS_DIRTY_WORK),
+                                     ERTS_PSFLG_EXITING|ERTS_PSFLG_ACTIVE,
+                                     &st,
+                                     &enq_prio,
+                                     ERTS_PROC_LOCKS_ALL);
 
-#ifdef DEBUG
-    enqueue =
-#endif
-	change_proc_schedule_state(c_p,
-				   ERTS_PSFLG_SUSPENDED|ERTS_PSFLG_PENDING_EXIT,
-				   ERTS_PSFLG_EXITING|ERTS_PSFLG_ACTIVE,
-				   &state,
-				   &enq_prio,
-				   ERTS_PROC_LOCKS_ALL);
-
-    ASSERT(!enqueue);
-    return state;
+    ASSERT(enqueue || !enq);
+    if (enqueue)
+        *enqueue = enq;
+    if (prio)
+        *prio = enq_prio;
+    if (state)
+        *state = st;
 }
 
-#ifdef ERTS_SMP
-
 void
-erts_handle_pending_exit(Process *c_p, ErtsProcLocks locks)
-{
-    ErtsProcLocks xlocks;
-    ASSERT(is_value(c_p->pending_exit.reason));
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == locks);
-    ERTS_SMP_LC_ASSERT(locks & ERTS_PROC_LOCK_MAIN);
-    ERTS_SMP_LC_ASSERT(!((ERTS_PSFLG_EXITING|ERTS_PSFLG_FREE)
-			 & erts_smp_atomic32_read_nob(&c_p->state)));
-
-    /* Ensure that all locks on c_p are locked before proceeding... */
-    if (locks == ERTS_PROC_LOCKS_ALL)
-	xlocks = 0;
-    else {
-	xlocks = ~locks & ERTS_PROC_LOCKS_ALL;
-	if (erts_smp_proc_trylock(c_p, xlocks) == EBUSY) {
-	    erts_smp_proc_unlock(c_p, locks & ~ERTS_PROC_LOCK_MAIN);
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
-	}
-    }
-
-    set_proc_exiting(c_p,
-		     erts_smp_atomic32_read_acqb(&c_p->state),
-		     c_p->pending_exit.reason,
-		     c_p->pending_exit.bp);
-    c_p->pending_exit.reason = THE_NON_VALUE;
-    c_p->pending_exit.bp = NULL;
-
-    if (xlocks)
-	erts_smp_proc_unlock(c_p, xlocks);
-}
-
-static void save_pending_exiter(Process *p, ErtsProcList *plp);
-
-static void
-do_handle_pending_exiters(ErtsProcList *pnd_xtrs)
-{
-    /* 'list' is expected to have been fetched (i.e. not a ring anymore) */
-    ErtsProcList *plp = pnd_xtrs;
-
-    while (plp) {
-	ErtsProcList *next_plp = plp->next;
-	Process *p = erts_proc_lookup(plp->pid);
-	if (p) {
-	    erts_aint32_t state;
-	    /*
-	     * If the process is running on a normal scheduler, the
-	     * pending exit will soon be detected and handled by the
-	     * scheduler running the process (at schedule in/out).
-	     */
-	    if (erts_smp_proc_trylock(p, ERTS_PROC_LOCKS_ALL) != EBUSY) {
-		if (erts_proclist_same(plp, p)) {
-		    state = erts_smp_atomic32_read_acqb(&p->state);
-		    if (!(state & (ERTS_PSFLG_RUNNING
-				   | ERTS_PSFLG_RUNNING_SYS
-				   | ERTS_PSFLG_EXITING))) {
-			ASSERT(state & ERTS_PSFLG_PENDING_EXIT);
-			erts_handle_pending_exit(p, ERTS_PROC_LOCKS_ALL);
-		    }
-		}
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_ALL);
-	    }
-	    else {
-		erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
-		if (erts_proclist_same(plp, p)) {
-		    state = erts_smp_atomic32_read_acqb(&p->state);
-		    if (!(state & (ERTS_PSFLG_RUNNING
-				   | ERTS_PSFLG_RUNNING_SYS
-				   | ERTS_PSFLG_EXITING))) {
-			/*
-			 * Save process and try to acquire all
-			 * locks at a later time...
-			 */
-			save_pending_exiter(p, plp);
-			plp = NULL;
-		    }
-		}
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
-	    }
-	}
-	if (plp)
-	    proclist_destroy(plp);
-	plp = next_plp;
-    }
-}
-
-static void
-save_pending_exiter(Process *p, ErtsProcList *plp)
+erts_set_self_exiting(Process *c_p, Eterm reason)
 {
-    ErtsSchedulerSleepInfo *ssi;
-    ErtsRunQueue *rq;
-
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_STATUS & erts_proc_lc_my_proc_locks(p));
-
-    rq = RUNQ_READ_RQ(&p->run_queue);
-    ASSERT(rq && !ERTS_RUNQ_IX_IS_DIRTY(rq->ix));
-
-    if (!plp)
-	plp = proclist_create(p);
-
-    erts_smp_runq_lock(rq);
-
-    erts_proclist_store_last(&rq->procs.pending_exiters, plp);
-
-    non_empty_runq(rq);
-
-    ssi = rq->scheduler->ssi;
+    int enqueue;
+    erts_aint32_t enq_prio, state;
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == ERTS_PROC_LOCK_MAIN);
 
-    erts_smp_runq_unlock(rq);
+    erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
 
-    set_aux_work_flags_wakeup_nob(ssi, ERTS_SSI_AUX_WORK_PENDING_EXITERS);
-}
+    set_self_exiting(c_p, reason, &enqueue, &enq_prio, &state);
+    c_p->freason = EXTAG_EXIT;
+    KILL_CATCHES(c_p);
+    c_p->i = (BeamInstr *) beam_exit;
 
-#endif
+    /* Always active when exiting... */
+    ASSERT(state & ERTS_PSFLG_ACTIVE);
 
-/*
- * This function delivers an EXIT message to a process
- * which is trapping EXITs.
- */
-
-static ERTS_INLINE void
-send_exit_message(Process *to, ErtsProcLocks *to_locksp,
-		  Eterm exit_term, Uint term_size, Eterm token)
-{
-    ErtsMessage *mp;
-    ErlOffHeap *ohp;
-    Eterm* hp;
-    Eterm mess;
-#ifdef SHCOPY_SEND
-    erts_shcopy_t info;
-#endif
+    /*
+     * If we are terminating a process that currently
+     * is executing on a dirty scheduler. It *should*
+     * be scheduled on a normal scheduler...
+     */
+    ASSERT(!(state & (ERTS_PSFLG_DIRTY_RUNNING
+                      | ERTS_PSFLG_DIRTY_RUNNING_SYS))
+           || enqueue == ERTS_ENQUEUE_NORMAL_QUEUE
+           || enqueue == -ERTS_ENQUEUE_NORMAL_QUEUE);
 
-    if (!have_seqtrace(token)) {
-#ifdef SHCOPY_SEND
-        INITIALIZE_SHCOPY(info);
-        term_size = copy_shared_calculate(exit_term, &info);
-	mp = erts_alloc_message_heap(to, to_locksp, term_size, &hp, &ohp);
-        mess = copy_shared_perform(exit_term, term_size, &info, &hp, ohp);
-        DESTROY_SHCOPY(info);
-#else
-	mp = erts_alloc_message_heap(to, to_locksp, term_size, &hp, &ohp);
-	mess = copy_struct(exit_term, term_size, &hp, ohp);
-#endif
-	erts_queue_message(to, *to_locksp, mp, mess, am_system);
-    } else {
-	Eterm temp_token;
-	Uint sz_token;
-
-	ASSERT(is_tuple(token));
-	sz_token = size_object(token);
-#ifdef SHCOPY_SEND
-        INITIALIZE_SHCOPY(info);
-        term_size = copy_shared_calculate(exit_term, &info);
-	mp = erts_alloc_message_heap(to, to_locksp, term_size+sz_token, &hp, &ohp);
-        mess = copy_shared_perform(exit_term, term_size, &info, &hp, ohp);
-        DESTROY_SHCOPY(info);
-#else
-	mp = erts_alloc_message_heap(to, to_locksp, term_size+sz_token, &hp, &ohp);
-	mess = copy_struct(exit_term, term_size, &hp, ohp);
-#endif
-	/* the trace token must in this case be updated by the caller */
-	seq_trace_output(token, mess, SEQ_TRACE_SEND, to->common.id, to);
-	temp_token = copy_struct(token, sz_token, &hp, ohp);
-        ERL_MESSAGE_TOKEN(mp) = temp_token;
-	erts_queue_message(to, *to_locksp, mp, mess, am_system);
-    }
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+    if (enqueue)
+        add2runq(enqueue, enq_prio, c_p, state, NULL);
 }
 
-/*
- *
- * *** Exit signal behavior ***
- *
- * Exit signals are asynchronous (truly asynchronous in the
- * SMP emulator). When the signal is received the receiver receives an
- * 'EXIT' message if it is trapping exits; otherwise, it will either
- * ignore the signal if the exit reason is normal, or go into an
- * exiting state (ERTS_PSFLG_EXITING). When a process has gone into the
- * exiting state it will not execute any more Erlang code, but it might
- * take a while before it actually exits. The exit signal is being
- * received when the 'EXIT' message is put in the message queue, the
- * signal is dropped, or when it changes state into exiting. The time it
- * is in the exiting state before actually exiting is undefined (it
- * might take a really long time under certain conditions). The
- * receiver of the exit signal does not break links or trigger monitors
- * until it actually exits.
- *
- * Exit signals and other signals, e.g. messages, have to be received
- * by a receiver in the same order as sent by a sender.
- *
- *
- *
- * Exit signal implementation in the SMP emulator:
- *
- * If the receiver is trapping exits, the signal is transformed
- * into an 'EXIT' message and sent as a normal message, if the
- * reason is normal the signal is dropped; otherwise, the process
- * is determined to be exited. The interesting case is when the
- * process is to be exited and this is what is described below.
- *
- * If it is possible, the receiver is set in the exiting state straight
- * away and we are done; otherwise, the sender places the exit reason
- * in the pending_exit field of the process struct and if necessary
- * adds the receiver to the run queue. It is typically not possible
- * to set a scheduled process or a process which we cannot get all locks
- * on without releasing locks on it in an exiting state straight away.
- *
- * The receiver will poll the pending_exit field when it reach certain
- * places during it's execution. When it discovers the pending exit
- * it will change state into the exiting state. If the receiver wasn't
- * scheduled when the pending exit was set, the first scheduler that
- * schedules a new process will set the receiving process in the exiting
- * state just before it schedules next process.
- * 
- * When the exit signal is placed in the pending_exit field, the signal
- * is considered as being in transit on the Erlang level. The signal is
- * actually in some kind of semi transit state, since we have already
- * determined how it should be received. It will exit the process no
- * matter what if it is received (the process may exit by itself before
- * reception of the exit signal). The signal is received when it is
- * discovered in the pending_exit field by the receiver.
- *
- * The receiver have to poll the pending_exit field at least before:
- * - moving messages from the message in queue to the private message
- *   queue. This in order to preserve signal order.
- * - unlink. Otherwise the process might get exited on a link that
- *   have been removed.
- * - changing the trap_exit flag to true. This in order to simplify the
- *   implementation; otherwise, we would have to transform the signal
- *   into an 'EXIT' message when setting the trap_exit flag to true. We
- *   would also have to maintain a queue of exit signals in transit.
- * - being scheduled in or out.
- */
-
-static ERTS_INLINE int
-send_exit_signal(Process *c_p,		/* current process if and only
-					   if reason is stored on it */
-		 Eterm from,		/* Id of sender of signal */
-		 Process *rp,		/* receiving process */
-		 ErtsProcLocks *rp_locks,/* current locks on receiver */
-		 Eterm reason,		/* exit reason */
-		 Eterm exit_tuple,	/* Prebuild exit tuple
-					   or THE_NON_VALUE */
-		 Uint exit_tuple_sz,	/* Size of prebuilt exit tuple
-					   (if exit_tuple != THE_NON_VALUE) */
-		 Eterm token,		/* token */
-		 Process *token_update, /* token updater */
-		 Uint32 flags		/* flags */
-    )		
-{
-    erts_aint32_t state = erts_smp_atomic32_read_nob(&rp->state);
-    Eterm rsn = reason == am_kill ? am_killed : reason;
-
-    ERTS_SMP_LC_ASSERT(*rp_locks == erts_proc_lc_my_proc_locks(rp));
-    ERTS_SMP_LC_ASSERT((*rp_locks & ERTS_PROC_LOCKS_XSIG_SEND)
-		       == ERTS_PROC_LOCKS_XSIG_SEND);
-
-    ASSERT(reason != THE_NON_VALUE);
-
-#ifdef USE_VM_PROBES
-    if(DTRACE_ENABLED(process_exit_signal) && is_pid(from)) {
-        DTRACE_CHARBUF(sender_str, DTRACE_TERM_BUF_SIZE);
-        DTRACE_CHARBUF(receiver_str, DTRACE_TERM_BUF_SIZE);
-        DTRACE_CHARBUF(reason_buf, DTRACE_TERM_BUF_SIZE);
-
-        dtrace_pid_str(from, sender_str);
-        dtrace_proc_str(rp, receiver_str);
-        erts_snprintf(reason_buf, sizeof(DTRACE_CHARBUF_NAME(reason_buf)) - 1, "%T", reason);
-        DTRACE3(process_exit_signal, sender_str, receiver_str, reason_buf);
+void
+erts_proc_exit_handle_monitor(ErtsMonitor *mon, void *vctxt)
+{
+    Process *c_p = ((ErtsProcExitContext *) vctxt)->c_p;
+    Eterm reason = ((ErtsProcExitContext *) vctxt)->reason;
+    ErtsMonitorData *mdp = NULL;
+
+    if (erts_monitor_is_target(mon)) {
+        /* We are being watched... */
+        switch (mon->type) {
+        case ERTS_MON_TYPE_SUSPEND:
+        case ERTS_MON_TYPE_PROC:
+            erts_proc_sig_send_monitor_down(mon, reason);
+            mon = NULL;
+            break;
+        case ERTS_MON_TYPE_PORT: {
+	    Port *prt;
+            ASSERT(is_internal_port(mon->other.item));
+            erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+            prt = erts_id2port(mon->other.item);
+	    if (prt) {
+                erts_fire_port_monitor(prt, mon);
+                erts_port_release(prt);
+                mon = NULL;
+            }
+            erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+            break;
+        }
+        case ERTS_MON_TYPE_RESOURCE:
+            erts_fire_nif_monitor(mon);
+            mon = NULL;
+            break;
+        case ERTS_MON_TYPE_DIST_PROC: {
+            ErtsMonLnkDist *dist;
+            DistEntry *dep;
+            ErtsDSigData dsd;
+            int code;
+            Eterm watcher;
+            Eterm watched;
+
+            mdp = erts_monitor_to_data(mon);
+
+            if (mon->flags & ERTS_ML_FLG_NAME)
+                watched = ((ErtsMonitorDataExtended *) mdp)->u.name;
+            else
+                watched = c_p->common.id;
+            ASSERT(is_internal_pid(watched) || is_atom(watched));
+
+            watcher = mon->other.item;
+	    ASSERT(is_external_pid(watcher));    
+	    dep = external_pid_dist_entry(watcher);
+            ASSERT(dep);
+            dist = ((ErtsMonitorDataExtended *) mdp)->dist;
+            ASSERT(dist);
+            code = erts_dsig_prepare(&dsd, dep, NULL, 0,
+                                     ERTS_DSP_NO_LOCK, 0, 0);
+            switch (code) {
+            case ERTS_DSIG_PREP_CONNECTED:
+            case ERTS_DSIG_PREP_PENDING:
+                if (dist->connection_id == dsd.connection_id) {
+                    code = erts_dsig_send_m_exit(&dsd,
+                                                 watcher,
+                                                 watched,
+                                                 mdp->ref,
+                                                 reason);
+                    ASSERT(code == ERTS_DSIG_SEND_OK);
+                }
+            default:
+                break;
+            }
+            if (!erts_monitor_dist_delete(&mdp->origin))
+                mdp = NULL;
+            break;
+        }
+        default:
+            ERTS_INTERNAL_ERROR("Invalid target monitor type");
+            break;
+        }
     }
-#endif
-
-    if ((state & ERTS_PSFLG_TRAP_EXIT)
-	&& (reason != am_kill || (flags & ERTS_XSIG_FLG_IGN_KILL))) {
-        /* have to release the status and trace lock in order to send the exit message */
-        erts_smp_proc_unlock(rp, *rp_locks & (ERTS_PROC_LOCKS_XSIG_SEND|ERTS_PROC_LOCK_TRACE));
-        *rp_locks &= ~(ERTS_PROC_LOCKS_XSIG_SEND|ERTS_PROC_LOCK_TRACE);
-        if (have_seqtrace(token) && token_update)
-	    seq_trace_update_send(token_update);
-	if (is_value(exit_tuple))
-	    send_exit_message(rp, rp_locks, exit_tuple, exit_tuple_sz, token);
-	else
-	    erts_deliver_exit_message(from, rp, rp_locks, rsn, token);
-	return 1; /* Receiver will get a message */
-    }
-    else if (reason != am_normal || (flags & ERTS_XSIG_FLG_NO_IGN_NORMAL)) {
-#ifdef ERTS_SMP
-	if (!(state & (ERTS_PSFLG_EXITING|ERTS_PSFLG_PENDING_EXIT))) {
-	    ASSERT(!rp->pending_exit.bp);
-
-	    if (rp == c_p && (*rp_locks & ERTS_PROC_LOCK_MAIN)) {
-		/* Ensure that all locks on c_p are locked before
-		   proceeding... */
-		if (*rp_locks != ERTS_PROC_LOCKS_ALL) {
-		    ErtsProcLocks need_locks = (~(*rp_locks)
-						& ERTS_PROC_LOCKS_ALL);
-		    if (erts_smp_proc_trylock(c_p, need_locks) == EBUSY) {
-			erts_smp_proc_unlock(c_p,
-					     *rp_locks & ~ERTS_PROC_LOCK_MAIN);
-			erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
-		    }
-		    *rp_locks = ERTS_PROC_LOCKS_ALL;
-		}
-		set_proc_exiting(c_p, state, rsn, NULL);
-	    }
-	    else if (!(state & (ERTS_PSFLG_RUNNING
-                                | ERTS_PSFLG_RUNNING_SYS
-                                | ERTS_PSFLG_DIRTY_RUNNING_SYS))) {
-		/* Process not running ... */
-		ErtsProcLocks need_locks = ~(*rp_locks) & ERTS_PROC_LOCKS_ALL;
-		ErlHeapFragment *bp = NULL;
-		Eterm rsn_cpy;
-		if (need_locks
-		    && erts_smp_proc_trylock(rp, need_locks) == EBUSY) {
-		    /* ... but we havn't got all locks on it ... */
-		    save_pending_exiter(rp, NULL); 
-		    /*
-		     * The pending exit will be discovered when next
-		     * process is scheduled in
-		     */
-                    goto set_pending_exit;
-		}
-                /* ...and we have all locks on it... */
-                *rp_locks = ERTS_PROC_LOCKS_ALL;
-
-		state = erts_smp_atomic32_read_nob(&rp->state);
-
-		if (is_immed(rsn))
-		    rsn_cpy = rsn;
-		else {
-		    Eterm *hp;
-		    ErlOffHeap *ohp;
-		    Uint rsn_sz = size_object(rsn);
-#ifdef ERTS_DIRTY_SCHEDULERS
-		    if (state & ERTS_PSFLG_DIRTY_RUNNING) {
-			bp = new_message_buffer(rsn_sz);
-			ohp = &bp->off_heap;
-			hp = &bp->mem[0];
-		    }
-		    else
-#endif
-		    {
-			hp = HAlloc(rp, rsn_sz);
-			ohp = &rp->off_heap;
-		    }
-		    rsn_cpy = copy_struct(rsn, rsn_sz, &hp, ohp);
-		}
-
-		set_proc_exiting(rp, state, rsn_cpy, bp);
-	    }
-	    else { /* Process running... */
- 
-		/*
-		 * The pending exit will be discovered when the process
-		 * is scheduled out if not discovered earlier.
-		 */
-
-	    set_pending_exit:
-		if (is_immed(rsn)) {
-		    rp->pending_exit.reason = rsn;
-		}
-		else {
-		    Eterm *hp;
-		    Uint sz = size_object(rsn);
-		    ErlHeapFragment *bp = new_message_buffer(sz);
-
-		    hp = &bp->mem[0];
-		    rp->pending_exit.reason = copy_struct(rsn,
-							  sz,
-							  &hp,
-							  &bp->off_heap);
-		    rp->pending_exit.bp = bp;
-		}
-
-		/*
-		 * If no dirty work has been scheduled, pending exit will
-                 * be discovered when the process is scheduled. If dirty work
-                 * has been scheduled, we may need to add it to a normal run
-                 * queue...
-		 */
-#ifndef ERTS_DIRTY_SCHEDULERS
-                (void) erts_smp_atomic32_read_bor_relb(&rp->state,
-                                                       ERTS_PSFLG_PENDING_EXIT);
-#else
-                {
-                    erts_aint32_t a = erts_smp_atomic32_read_nob(&rp->state);
-                    while (1) {
-                        erts_aint32_t n, e;
-                        int dwork;
-                        n = e = a;
-                        n |= ERTS_PSFLG_PENDING_EXIT;
-                        dwork = !!(n & ERTS_PSFLGS_DIRTY_WORK);
-                        n &= ~ERTS_PSFLGS_DIRTY_WORK;
-                        a = erts_smp_atomic32_cmpxchg_mb(&rp->state, n, e);
-                        if (a == e) {
-                            if (dwork)
-                                erts_schedule_process(rp, n, *rp_locks);
-                            break;
-                        }
-                    }
+    else { /* Origin monitor */
+        /* We are watching someone else... */
+        switch (mon->type) {
+        case ERTS_MON_TYPE_SUSPEND:
+        case ERTS_MON_TYPE_PROC:
+            erts_proc_sig_send_demonitor(mon);
+            mon = NULL;
+            break;
+        case ERTS_MON_TYPE_TIME_OFFSET:
+            erts_demonitor_time_offset(mon);
+            mon = NULL;
+            break;
+        case ERTS_MON_TYPE_NODE:
+            mdp = erts_monitor_to_data(mon);
+            if (!erts_monitor_dist_delete(&mdp->target))
+                mdp = NULL;
+            break;
+        case ERTS_MON_TYPE_NODES:
+            erts_monitor_nodes_delete(mon);
+            mon = NULL;
+            break;
+        case ERTS_MON_TYPE_PORT: {
+            Port *prt;
+            ASSERT(is_internal_port(mon->other.item));
+            prt = erts_port_lookup_raw(mon->other.item);
+            if (prt) {
+                if (erts_port_demonitor(c_p, prt, mon) != ERTS_PORT_OP_DROPPED)
+                    mon = NULL;
+            }
+            break;
+        }
+        case ERTS_MON_TYPE_DIST_PROC: {
+            ErtsMonLnkDist *dist;
+            DistEntry *dep;
+            ErtsDSigData dsd;
+            int code;
+            Eterm watched;
+
+            mdp = erts_monitor_to_data(mon);
+            dist = ((ErtsMonitorDataExtended *) mdp)->dist;
+            ASSERT(dist);
+            if (mon->flags & ERTS_ML_FLG_NAME) {
+                watched = ((ErtsMonitorDataExtended *) mdp)->u.name;
+                ASSERT(is_atom(watched));
+                dep = erts_sysname_to_connected_dist_entry(dist->nodename);
+            }
+            else {
+                watched = mon->other.item;
+                ASSERT(is_external_pid(watched));
+		dep = external_pid_dist_entry(watched);
+            }
+            code = erts_dsig_prepare(&dsd, dep, NULL, 0,
+                                     ERTS_DSP_NO_LOCK, 0, 0);
+            switch (code) {
+            case ERTS_DSIG_PREP_CONNECTED:
+            case ERTS_DSIG_PREP_PENDING:
+                if (dist->connection_id == dsd.connection_id) {
+                    code = erts_dsig_send_demonitor(&dsd,
+                                                    c_p->common.id,
+                                                    watched,
+                                                    mdp->ref,
+                                                    1);
+                    ASSERT(code == ERTS_DSIG_SEND_OK);
                 }
-#endif
-	    }
-	}
-	/* else:
-	 *
-	 *    The receiver already has a pending exit (or is exiting)
-	 *    so we drop this signal.
-	 *
-	 *    NOTE: dropping this exit signal is based on the assumption
-	 *          that the receiver *will* exit; either on the pending
-	 *          exit or by itself before seeing the pending exit.
-	 */
-#else /* !ERTS_SMP */
-	erts_aint32_t state = erts_smp_atomic32_read_nob(&rp->state);
-	if (!(state & ERTS_PSFLG_EXITING)) {
-	    set_proc_exiting(rp,
-			     state,
-			     (is_immed(rsn) || c_p == rp
-			      ? rsn
-			      : copy_object(rsn, rp)),
-			     NULL);
-	}
-#endif
-	return -1; /* Receiver will exit */
+            default:
+                break;
+            }
+            if (!erts_monitor_dist_delete(&mdp->target))
+                mdp = NULL;
+            break;
+        }
+        default:
+            ERTS_INTERNAL_ERROR("Invalid origin monitor type");
+            break;
+        }
     }
 
-    return 0; /* Receiver unaffected */
+    if (mdp)
+        erts_monitor_release_both(mdp);
+    else if (mon)
+        erts_monitor_release(mon);
 }
 
-
-int
-erts_send_exit_signal(Process *c_p,
-		      Eterm from,
-		      Process *rp,
-		      ErtsProcLocks *rp_locks,
-		      Eterm reason,
-		      Eterm token,
-		      Process *token_update,
-		      Uint32 flags)
-{
-    return send_exit_signal(c_p,
-			    from,
-			    rp,
-			    rp_locks,
-			    reason,
-			    THE_NON_VALUE,
-			    0,
-			    token,
-			    token_update,
-			    flags);
-}
-
-typedef struct {
-    Eterm reason;
-    Process *p;
-} ExitMonitorContext;
-
-static void doit_exit_monitor(ErtsMonitor *mon, void *vpcontext)
-{
-    ExitMonitorContext *pcontext = vpcontext;
-    DistEntry *dep;
-    ErtsMonitor *rmon;
-
-    switch (mon->type) {
-    case MON_ORIGIN:
-	/* We are monitoring someone else, we need to demonitor that one.. */
-	if (is_atom(mon->u.pid)) { /* remote by name */
-	    ASSERT(is_node_name_atom(mon->u.pid));
-	    dep = erts_sysname_to_connected_dist_entry(mon->u.pid);
-	    if (dep) {
-		erts_smp_de_links_lock(dep);
-		rmon = erts_remove_monitor(&(dep->monitors), mon->ref);
-		erts_smp_de_links_unlock(dep);
-		if (rmon) {
-		    ErtsDSigData dsd;
-		    int code = erts_dsig_prepare(&dsd, dep, NULL,
-						 ERTS_DSP_NO_LOCK, 0);
-		    if (code == ERTS_DSIG_PREP_CONNECTED) {
-			code = erts_dsig_send_demonitor(&dsd,
-							rmon->u.pid,
-							mon->name,
-							mon->ref,
-							1);
-			ASSERT(code == ERTS_DSIG_SEND_OK);
-		    }
-		    erts_destroy_monitor(rmon);
-		}
-		erts_deref_dist_entry(dep);
-	    }
-	} else {
-            ASSERT(is_pid(mon->u.pid) || is_port(mon->u.pid));
-            /* if is local by pid or name */
-            if (is_internal_pid(mon->u.pid)) {
-                Process *rp = erts_pid2proc(NULL, 0, mon->u.pid, ERTS_PROC_LOCK_LINK);
-		if (!rp) {
-		    goto done;
-		}
-		rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-		erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-		if (rmon == NULL) {
-		    goto done;
-		}
-		erts_destroy_monitor(rmon);
-            } else if (is_internal_port(mon->u.pid)) {
-                /* Is a local port */
-                Port *prt = erts_port_lookup_raw(mon->u.pid);
-                if (!prt) {
-                    goto done;
-                }
-                erts_port_demonitor(pcontext->p,
-                                    ERTS_PORT_DEMONITOR_ORIGIN_ON_DEATHBED,
-                                    prt, mon->ref, NULL);
-            } else { /* remote by pid */
-		ASSERT(is_external_pid(mon->u.pid));
-		dep = external_pid_dist_entry(mon->u.pid);
-		ASSERT(dep != NULL);
-		if (dep) {
-		    erts_smp_de_links_lock(dep);
-		    rmon = erts_remove_monitor(&(dep->monitors), mon->ref);
-		    erts_smp_de_links_unlock(dep);
-		    if (rmon) {
-			ErtsDSigData dsd;
-			int code = erts_dsig_prepare(&dsd, dep, NULL,
-						     ERTS_DSP_NO_LOCK, 0);
-			if (code == ERTS_DSIG_PREP_CONNECTED) {
-			    code = erts_dsig_send_demonitor(&dsd,
-							    rmon->u.pid,
-							    mon->u.pid,
-							    mon->ref,
-							    1);
-			    ASSERT(code == ERTS_DSIG_SEND_OK);
-			}
-			erts_destroy_monitor(rmon);
-		    }
-		}
-	    }
-	}
-	break;
-    case MON_TARGET:
-	ASSERT(is_pid(mon->u.pid) || is_internal_port(mon->u.pid));
-	if (is_internal_port(mon->u.pid)) {
-	    Port *prt = erts_id2port(mon->u.pid);
-	    if (prt == NULL) {
-		goto done;
-	    }
-	    erts_fire_port_monitor(prt, mon->ref);
-	    erts_port_release(prt); 
-	} else if (is_internal_pid(mon->u.pid)) {/* local by name or pid */
-	    Eterm watched;
-            Process *rp;
-	    DeclareTmpHeapNoproc(lhp,3);
-	    ErtsProcLocks rp_locks = (ERTS_PROC_LOCK_LINK
-				      | ERTS_PROC_LOCKS_MSG_SEND);
-	    rp = erts_pid2proc(NULL, 0, mon->u.pid, rp_locks);
-	    if (rp == NULL) {
-		goto done;
-	    }
-	    UseTmpHeapNoproc(3);
-	    rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-	    if (rmon) {
-		erts_destroy_monitor(rmon);
-		watched = (is_atom(mon->name)
-			   ? TUPLE2(lhp, mon->name, 
-				    erts_this_dist_entry->sysname)
-			   : pcontext->p->common.id);
-		erts_queue_monitor_message(rp, &rp_locks, mon->ref, am_process, 
-					   watched, pcontext->reason);
-	    }
-	    UnUseTmpHeapNoproc(3);
-	    /* else: demonitor while we exited, i.e. do nothing... */
-	    erts_smp_proc_unlock(rp, rp_locks);
-	} else { /* external by pid or name */
-	    ASSERT(is_external_pid(mon->u.pid));    
-	    dep = external_pid_dist_entry(mon->u.pid);
-	    ASSERT(dep != NULL);
-	    if (dep) {
-		erts_smp_de_links_lock(dep);
-		rmon = erts_remove_monitor(&(dep->monitors), mon->ref);
-		erts_smp_de_links_unlock(dep);
-		if (rmon) {
-		    ErtsDSigData dsd;
-		    int code = erts_dsig_prepare(&dsd, dep, NULL,
-						 ERTS_DSP_NO_LOCK, 0);
-		    if (code == ERTS_DSIG_PREP_CONNECTED) {
-			code = erts_dsig_send_m_exit(&dsd,
-						     mon->u.pid,
-						     (rmon->name != NIL
-						      ? rmon->name
-						      : rmon->u.pid),
-						     mon->ref,
-						     pcontext->reason);
-			ASSERT(code == ERTS_DSIG_SEND_OK);
-		    }
-		    erts_destroy_monitor(rmon);
-		}
-	    }
-	}
-	break;
-    case MON_NIF_TARGET:
-        erts_fire_nif_monitor(mon->u.resource,
-                              pcontext->p->common.id,
-                              mon->ref);
+void
+erts_proc_exit_handle_link(ErtsLink *lnk, void *vctxt)
+{
+    Process *c_p = ((ErtsProcExitContext *) vctxt)->c_p;
+    Eterm reason = ((ErtsProcExitContext *) vctxt)->reason;
+    ErtsLinkData *ldp = NULL;
+
+    switch (lnk->type) {
+    case ERTS_LNK_TYPE_PROC:
+        ASSERT(is_internal_pid(lnk->other.item));
+        erts_proc_sig_send_link_exit(c_p, c_p->common.id, lnk,
+                                     reason, SEQ_TRACE_TOKEN(c_p));
+        lnk = NULL;
+        break;
+    case ERTS_LNK_TYPE_PORT: {
+        Port *prt;
+        ASSERT(is_internal_port(lnk->other.item));
+        prt = erts_port_lookup(lnk->other.item,
+                               ERTS_PORT_SFLGS_INVALID_LOOKUP);
+        if (prt)
+            erts_port_exit(NULL,
+                           (ERTS_PORT_SIG_FLG_FORCE_SCHED
+                            | ERTS_PORT_SIG_FLG_BROKEN_LINK),
+                           prt,
+                           c_p->common.id,
+                           reason,
+                           NULL);
+        break;
+    }
+    case ERTS_LNK_TYPE_DIST_PROC: {
+        DistEntry *dep;
+        ErtsMonLnkDist *dist;
+        ErtsLink *dlnk;
+        ErtsDSigData dsd;
+        int code;
+
+        dlnk = erts_link_to_other(lnk, &ldp);
+        dist = ((ErtsLinkDataExtended *) ldp)->dist;
+
+        ASSERT(is_external_pid(lnk->other.item));
+        dep = external_pid_dist_entry(lnk->other.item);
+
+        ASSERT(dep != erts_this_dist_entry);
+
+        if (!erts_link_dist_delete(dlnk))
+            ldp = NULL;
+
+        code = erts_dsig_prepare(&dsd, dep, c_p, 0, ERTS_DSP_NO_LOCK, 0, 0);
+        switch (code) {
+        case ERTS_DSIG_PREP_CONNECTED:
+        case ERTS_DSIG_PREP_PENDING:
+            if (dist->connection_id == dsd.connection_id) {
+                code = erts_dsig_send_exit_tt(&dsd,
+                                              c_p->common.id,
+                                              lnk->other.item,
+                                              reason,
+                                              SEQ_TRACE_TOKEN(c_p));
+                ASSERT(code == ERTS_DSIG_SEND_OK);
+            }
+        }
         break;
-    case MON_TIME_OFFSET:
-	erts_demonitor_time_offset(mon->ref);
-	break;
-    default:
-	ERTS_INTERNAL_ERROR("Invalid monitor type");
     }
- done:
-    /* As the monitors are previously removed from the process, 
-       distribution operations will not cause monitors to disappear,
-       we can safely delete it. */
-       
-    erts_destroy_monitor(mon);
-}
-
-typedef struct {
-    Process *p;
-    Eterm reason;
-    Eterm exit_tuple;
-    Uint exit_tuple_sz;
-} ExitLinkContext;
-
-static void doit_exit_link(ErtsLink *lnk, void *vpcontext)
-{
-    ExitLinkContext *pcontext = vpcontext;
-    /* Unpack context, it's readonly */
-    Process *p = pcontext->p;
-    Eterm reason = pcontext->reason;
-    Eterm exit_tuple = pcontext->exit_tuple;
-    Uint exit_tuple_sz = pcontext->exit_tuple_sz;
-    Eterm item = lnk->pid;
-    ErtsLink *rlnk;
-    DistEntry *dep;
-    Process *rp;
-
-
-    switch(lnk->type) {
-    case LINK_PID:
-	if(is_internal_port(item)) {
-	    Port *prt = erts_port_lookup(item, ERTS_PORT_SFLGS_INVALID_LOOKUP);
-	    if (prt)
-		erts_port_exit(NULL,
-			       (ERTS_PORT_SIG_FLG_FORCE_SCHED
-				| ERTS_PORT_SIG_FLG_BROKEN_LINK),
-			       prt,
-			       p->common.id,
-			       reason,
-			       NULL);
-	}
-	else if(is_external_port(item)) {
-	    erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-	    erts_dsprintf(dsbufp,
-			  "Erroneous link between %T and external port %T "
-			  "found\n",
-			  p->common.id,
-			  item);
-	    erts_send_error_to_logger_nogl(dsbufp);
-	    ASSERT(0); /* It isn't possible to setup such a link... */
-	}
-	else if (is_internal_pid(item)) {
-	    ErtsProcLocks rp_locks = (ERTS_PROC_LOCK_LINK
-				      | ERTS_PROC_LOCKS_XSIG_SEND);
-	    rp = erts_pid2proc(NULL, 0, item, rp_locks);
-	    if (rp) {
-		rlnk = erts_remove_link(&ERTS_P_LINKS(rp), p->common.id);
-		/* If rlnk == NULL, we got unlinked while exiting,
-		   i.e., do nothing... */
-		if (rlnk) {
-		    int xres;
-		    erts_destroy_link(rlnk);
-		    xres = send_exit_signal(NULL,
-					    p->common.id,
-					    rp,
-					    &rp_locks, 
-					    reason,
-					    exit_tuple,
-					    exit_tuple_sz,
-					    SEQ_TRACE_TOKEN(p),
-					    p,
-					    ERTS_XSIG_FLG_IGN_KILL);
-		    if (xres >= 0 && IS_TRACED_FL(rp, F_TRACE_PROCS)) {
-			/* We didn't exit the process and it is traced */
-			if (IS_TRACED_FL(rp, F_TRACE_PROCS)) {
-                            if (rp_locks & ERTS_PROC_LOCKS_XSIG_SEND) {
-                                erts_smp_proc_unlock(rp, ERTS_PROC_LOCKS_XSIG_SEND);
-                                rp_locks &= ~ERTS_PROC_LOCKS_XSIG_SEND;
-                            }
-			    trace_proc(NULL, 0, rp, am_getting_unlinked, p->common.id);
-			}
-		    }
-		}
-		ASSERT(rp != p);
-		erts_smp_proc_unlock(rp, rp_locks);
-	    }
-	}
-	else if (is_external_pid(item)) {
-	    dep = external_pid_dist_entry(item);
-	    if(dep != erts_this_dist_entry) {
-		ErtsDSigData dsd;
-		int code;
-		ErtsDistLinkData dld;
-		erts_remove_dist_link(&dld, p->common.id, item, dep);
-		erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
-		code = erts_dsig_prepare(&dsd, dep, p, ERTS_DSP_NO_LOCK, 0);
-		if (code == ERTS_DSIG_PREP_CONNECTED) {
-		    code = erts_dsig_send_exit_tt(&dsd, p->common.id, item,
-						  reason, SEQ_TRACE_TOKEN(p));
-		    ASSERT(code == ERTS_DSIG_SEND_OK);
-		}
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
-		erts_destroy_dist_link(&dld);
-	    }
-	}
-	break;
-    case LINK_NODE:
-	ASSERT(is_node_name_atom(item));
-	dep = erts_sysname_to_connected_dist_entry(item);
-	if(dep) {
-	    /* dist entries have node links in a separate structure to 
-	       avoid confusion */
-	    erts_smp_de_links_lock(dep);
-	    rlnk = erts_remove_link(&(dep->node_links), p->common.id);
-	    erts_smp_de_links_unlock(dep);
-	    if (rlnk)
-		erts_destroy_link(rlnk);
-	    erts_deref_dist_entry(dep);
-	}
-	break;
-	
     default:
-	erts_exit(ERTS_ERROR_EXIT, "bad type in link list\n");
-	break;
+        ERTS_INTERNAL_ERROR("Unexpected link type");
+        break;
     }
-    erts_destroy_link(lnk);
-}
 
-static void
-resume_suspend_monitor(ErtsSuspendMonitor *smon, void *vc_p)
-{
-    Process *suspendee = erts_pid2proc((Process *) vc_p, ERTS_PROC_LOCK_MAIN,
-				       smon->pid, ERTS_PROC_LOCK_STATUS);
-    if (suspendee) {
-	ASSERT(suspendee != vc_p);
-	if (smon->active)
-	    resume_process(suspendee, ERTS_PROC_LOCK_STATUS);
-	erts_smp_proc_unlock(suspendee, ERTS_PROC_LOCK_STATUS);
-    }
-    erts_destroy_suspend_monitor(smon);
+    if (ldp)
+        erts_link_release_both(ldp);
+    else if (lnk)
+        erts_link_release(lnk);
 }
 
 /* this function fishishes a process and propagates exit messages - called
@@ -13806,8 +12145,6 @@ void
 erts_do_exit_process(Process* p, Eterm reason)
 {
     p->arity = 0;		/* No live registers */
-    p->fvalue = reason;
-    
 
 #ifdef USE_VM_PROBES
     if (DTRACE_ENABLED(process_exit)) {
@@ -13824,31 +12161,14 @@ erts_do_exit_process(Process* p, Eterm reason)
 	erts_exit(ERTS_DUMP_EXIT, "System process %T terminated: %T\n",
                  p->common.id, reason);
 
-#ifdef ERTS_SMP
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
     /* By locking all locks (main lock is already locked) when going
        to exiting state (ERTS_PSFLG_EXITING), it is enough to take any lock when
        looking up a process (erts_pid2proc()) to prevent the looked up
        process from exiting until the lock has been released. */
-    erts_smp_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
-#endif
-
-#ifndef ERTS_SMP
-    set_proc_self_exiting(p);
-#else
-    if (ERTS_PSFLG_PENDING_EXIT & set_proc_self_exiting(p)) {
-	/* Process exited before pending exit was received... */
-	p->pending_exit.reason = THE_NON_VALUE;
-	if (p->pending_exit.bp) {
-	    free_message_buffer(p->pending_exit.bp);
-	    p->pending_exit.bp = NULL;
-	}
-    }
+    erts_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
 
-    cancel_suspend_of_suspendee(p, ERTS_PROC_LOCKS_ALL); 
-
-    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p);
-#endif
+    set_self_exiting(p, reason, NULL, NULL, NULL);
 
     if (IS_TRACED(p)) {
 	if (IS_TRACED_FL(p, F_TRACE_CALLS))
@@ -13867,7 +12187,7 @@ erts_do_exit_process(Process* p, Eterm reason)
 	ASSERT(erts_proc_read_refc(p) > 0);
     }
 
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_proc_unlock(p, ERTS_PROC_LOCKS_ALL_MINOR);
 
     if (IS_TRACED_FL(p,F_TRACE_PROCS))
         trace_proc(p, ERTS_PROC_LOCK_MAIN, p, am_exit, reason);
@@ -13885,19 +12205,21 @@ erts_do_exit_process(Process* p, Eterm reason)
 void
 erts_continue_exit_process(Process *p)
 {
-    ErtsLink* lnk;
-    ErtsMonitor *mon;
+    ErtsLink *links;
+    ErtsMonitor *monitors;
+    ErtsMonitor *lt_monitors;
     ErtsProcLocks curr_locks = ERTS_PROC_LOCK_MAIN;
     Eterm reason = p->fvalue;
-    DistEntry *dep;
+    DistEntry *dep = NULL;
     erts_aint32_t state;
     int delay_del_proc = 0;
+    ErtsProcExitContext pectxt;
 
 #ifdef DEBUG
     int yield_allowed = 1;
 #endif
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(p));
 
     ASSERT(ERTS_PROC_IS_EXITING(p));
 
@@ -13911,7 +12233,6 @@ erts_continue_exit_process(Process *p)
 	p->bif_timers = NULL;
     }
 
-#ifdef ERTS_SMP
     if (p->flags & F_SCHDLR_ONLN_WAITQ)
 	abort_sched_onln_chng_waitq(p);
 
@@ -13955,7 +12276,6 @@ erts_continue_exit_process(Process *p)
 		     __FILE__, __LINE__, (int) ssr);
 	}
     }
-#endif
 
     if (p->flags & F_USING_DB) {
 	if (erts_db_process_exiting(p, ERTS_PROC_LOCK_MAIN))
@@ -13964,24 +12284,17 @@ erts_continue_exit_process(Process *p)
     }
 
     erts_set_gc_state(p, 1);
-    state = erts_smp_atomic32_read_acqb(&p->state);
-    if (state & ERTS_PSFLG_ACTIVE_SYS
-#ifdef ERTS_DIRTY_SCHEDULERS
-        || p->dirty_sys_tasks
-#endif
-        ) {
+    state = erts_atomic32_read_acqb(&p->state);
+    if ((state & ERTS_PSFLG_SYS_TASKS) || p->dirty_sys_tasks) {
 	if (cleanup_sys_tasks(p, state, CONTEXT_REDS) >= CONTEXT_REDS/2)
 	    goto yield;
     }
 
 #ifdef DEBUG
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
-    ASSERT(p->sys_task_qs == NULL);
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
     ASSERT(ERTS_PROC_GET_DELAYED_GC_TASK_QS(p) == NULL);
-#ifdef ERTS_DIRTY_SCHEDULERS
     ASSERT(p->dirty_sys_tasks == NULL);
-#endif
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
 #endif
 
     if (p->flags & F_USING_DDLL) {
@@ -13989,19 +12302,6 @@ erts_continue_exit_process(Process *p)
 	p->flags &= ~F_USING_DDLL;
     }
 
-    if (p->nodes_monitors) {
-	erts_delete_nodes_monitors(p, ERTS_PROC_LOCK_MAIN);
-	p->nodes_monitors = NULL;
-    }
-	
-
-    if (p->suspend_monitors) {
-	erts_sweep_suspend_monitors(p->suspend_monitors,
-				    resume_suspend_monitor,
-				    p);
-	p->suspend_monitors = NULL;
-    }
-
     /*
      * The registered name *should* be the last "erlang resource" to
      * cleanup.
@@ -14014,7 +12314,7 @@ erts_continue_exit_process(Process *p)
     if (IS_TRACED_FL(p, F_TRACE_SCHED_EXIT))
         trace_sched(p, curr_locks, am_out_exited);
 
-    erts_smp_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
+    erts_proc_lock(p, ERTS_PROC_LOCKS_ALL_MINOR);
     curr_locks = ERTS_PROC_LOCKS_ALL;
 
     /*
@@ -14029,31 +12329,28 @@ erts_continue_exit_process(Process *p)
      * Note! The monitor and link fields will be overwritten 
      * by erts_ptab_delete_element() below.
      */
-    mon = ERTS_P_MONITORS(p);
-    lnk = ERTS_P_LINKS(p);
+    links = ERTS_P_LINKS(p);
+    monitors = ERTS_P_MONITORS(p);
+    lt_monitors = ERTS_P_LT_MONITORS(p);
 
     {
 	/* Do *not* use erts_get_runq_proc() */
 	ErtsRunQueue *rq;
 	rq = erts_get_runq_current(erts_proc_sched_data(p));
 
-	erts_smp_runq_lock(rq);
+	erts_runq_lock(rq);
 
-#ifdef ERTS_SMP
 	ASSERT(p->scheduler_data);
 	ASSERT(p->scheduler_data->current_process == p);
 	ASSERT(p->scheduler_data->free_process == NULL);
 
 	p->scheduler_data->current_process = NULL;
 	p->scheduler_data->free_process = p;
-#else
-	erts_proc_inc_refc(p); /* Decremented in schedule() */
-#endif
 
 	/* Time of death! */
 	erts_ptab_delete_element(&erts_proc, &p->common);
 
-	erts_smp_runq_unlock(rq);
+	erts_runq_unlock(rq);
     }
 
     /*
@@ -14065,23 +12362,24 @@ erts_continue_exit_process(Process *p)
 
     {
 	/* Inactivate and notify free */
-	erts_aint32_t n, e, a = erts_smp_atomic32_read_nob(&p->state);
+	erts_aint32_t n, e, a = erts_atomic32_read_nob(&p->state);
 	int refc_inced = 0;
 	while (1) {
 	    n = e = a;
 	    ASSERT(a & ERTS_PSFLG_EXITING);
 	    n |= ERTS_PSFLG_FREE;
-	    n &= ~ERTS_PSFLG_ACTIVE;
+	    n &= ~(ERTS_PSFLG_ACTIVE
+                   | ERTS_PSFLG_ACTIVE_SYS
+                   | ERTS_PSFLG_DIRTY_ACTIVE_SYS);
 	    if ((n & ERTS_PSFLG_IN_RUNQ) && !refc_inced) {
 		erts_proc_inc_refc(p);
 		refc_inced = 1;
 	    }
-	    a = erts_smp_atomic32_cmpxchg_mb(&p->state, n, e);
+	    a = erts_atomic32_cmpxchg_mb(&p->state, n, e);
 	    if (a == e)
 		break;
 	}
 
-#ifdef ERTS_DIRTY_SCHEDULERS
         if (a & (ERTS_PSFLG_DIRTY_RUNNING
                  | ERTS_PSFLG_DIRTY_RUNNING_SYS)) {
 	    p->flags |= F_DELAYED_DEL_PROC;
@@ -14091,55 +12389,89 @@ erts_continue_exit_process(Process *p)
              * when done with the process...
 	     */
 	}
-#endif
 
 	if (refc_inced && !(n & ERTS_PSFLG_IN_RUNQ))
 	    erts_proc_dec_refc(p);
     }
-    
-    dep = (p->flags & F_DISTRIBUTION) ? erts_this_dist_entry : NULL;
 
-    erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_ALL);
+    dep = ((p->flags & F_DISTRIBUTION)
+           ? ERTS_PROC_SET_DIST_ENTRY(p, NULL)
+           : NULL);
 
-    if (dep) {
-	erts_do_net_exits(dep, reason);
-    }
 
     /*
-     * Pre-build the EXIT tuple if there are any links.
+     * It might show up signal prio elevation tasks until we
+     * have entered free state. Cleanup such tasks now.
      */
-    if (lnk) {
-	DeclareTmpHeap(tmp_heap,4,p);
-	Eterm exit_tuple;
-	Uint exit_tuple_sz;
-	Eterm* hp;
+    state = erts_atomic32_read_acqb(&p->state);
+    if (!(state & ERTS_PSFLG_SYS_TASKS))
+        erts_proc_unlock(p, ERTS_PROC_LOCKS_ALL);
+    else {
+        erts_proc_unlock(p, ERTS_PROC_LOCKS_ALL_MINOR);
+
+        do {
+            (void) cleanup_sys_tasks(p, state, CONTEXT_REDS);
+            state = erts_atomic32_read_acqb(&p->state);
+        } while (state & ERTS_PSFLG_SYS_TASKS);
+        
+        erts_proc_unlock(p, ERTS_PROC_LOCK_MAIN);
+    }
+
+#ifdef DEBUG
+    erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+    ASSERT(p->sys_task_qs == NULL);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_STATUS);
+#endif
+
+    if (dep) {
+        erts_do_net_exits(dep, (reason == am_kill) ? am_killed : reason);
+        erts_deref_dist_entry(dep);
+    }
 
-	UseTmpHeap(4,p);
-	hp = &tmp_heap[0];
+    pectxt.c_p = p;
+    pectxt.reason = reason;
 
-	exit_tuple = TUPLE3(hp, am_EXIT, p->common.id, reason);
+    erts_proc_lock(p, ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_MSGQ);
 
-	exit_tuple_sz = size_object(exit_tuple);
+    erts_proc_sig_fetch(p);
 
-	{
-	    ExitLinkContext context = {p, reason, exit_tuple, exit_tuple_sz};
-	    erts_sweep_links(lnk, &doit_exit_link, &context);
-	}
-	UnUseTmpHeap(4,p);
+    erts_proc_unlock(p, ERTS_PROC_LOCK_MSGQ);
+
+    if (links) {
+        erts_link_tree_foreach_delete(&links,
+                                      erts_proc_exit_handle_link,
+                                      (void *) &pectxt);
+        ASSERT(!links);
     }
 
-    {
-	ExitMonitorContext context = {reason, p};
-	erts_sweep_monitors(mon,&doit_exit_monitor,&context); /* Allocates TmpHeap, but we
-								 have none here */
+    if (monitors) {
+        erts_monitor_tree_foreach_delete(&monitors,
+                                         erts_proc_exit_handle_monitor,
+                                         (void *) &pectxt);
+        ASSERT(!monitors);
+    }
+
+    if (lt_monitors) {
+        erts_monitor_list_foreach_delete(&lt_monitors,
+                                         erts_proc_exit_handle_monitor,
+                                         (void *) &pectxt);
+        ASSERT(!lt_monitors);
+    }
+
+    /*
+     * erts_proc_sig_handle_exit() implements yielding.
+     * However, this function cannot handle it yet... loop
+     * until done...
+     */
+    while (!0) {
+        int reds = CONTEXT_REDS;
+        if (erts_proc_sig_handle_exit(p, &reds))
+            break;
     }
 
-    erts_smp_proc_lock(p, ERTS_PROC_LOCK_MAIN);
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(p);
 
-#ifdef ERTS_SMP
     erts_flush_trace_messages(p, ERTS_PROC_LOCK_MAIN);
-#endif
 
     ERTS_TRACER_CLEAR(&ERTS_TRACER(p));
 
@@ -14154,24 +12486,78 @@ erts_continue_exit_process(Process *p)
     ASSERT(yield_allowed);
 #endif
 
-    ERTS_SMP_LC_ASSERT(curr_locks == erts_proc_lc_my_proc_locks(p));
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN & curr_locks);
+    ERTS_LC_ASSERT(curr_locks == erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN & curr_locks);
 
     p->i = (BeamInstr *) beam_continue_exit;
 
     if (!(curr_locks & ERTS_PROC_LOCK_STATUS)) {
-	erts_smp_proc_lock(p, ERTS_PROC_LOCK_STATUS);
+	erts_proc_lock(p, ERTS_PROC_LOCK_STATUS);
 	curr_locks |= ERTS_PROC_LOCK_STATUS;
     }
 
     if (curr_locks != ERTS_PROC_LOCK_MAIN)
-	erts_smp_proc_unlock(p, ~ERTS_PROC_LOCK_MAIN & curr_locks);
+	erts_proc_unlock(p, ~ERTS_PROC_LOCK_MAIN & curr_locks);
 
-    ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(p));
+    ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN == erts_proc_lc_my_proc_locks(p));
 
     BUMP_ALL_REDS(p);
 }
 
+Process *
+erts_try_lock_sig_free_proc(Eterm pid, ErtsProcLocks locks,
+                            erts_aint32_t *statep)
+{
+    Process *rp = erts_proc_lookup_raw(pid);
+    erts_aint32_t fail_state = ERTS_PSFLG_SIG_IN_Q|ERTS_PSFLG_SIG_Q;
+    erts_aint32_t state;
+    ErtsProcLocks tmp_locks = ERTS_PROC_LOCK_MAIN|ERTS_PROC_LOCK_MSGQ;
+
+    tmp_locks |= locks;
+    if (statep)
+        fail_state |= *statep;
+
+    if (!rp) {
+        if (statep)
+            *statep = ERTS_PSFLG_EXITING|ERTS_PSFLG_FREE;
+        return NULL;
+    }
+
+    ERTS_LC_ASSERT(!erts_proc_lc_my_proc_locks(rp));
+
+    state = erts_atomic32_read_nob(&rp->state);
+    if (statep)
+        *statep = state;
+
+    if (state & ERTS_PSFLG_FREE)
+        return NULL;
+
+    if (state & fail_state)
+        return ERTS_PROC_LOCK_BUSY;
+
+    if (erts_proc_trylock(rp, tmp_locks) == EBUSY)
+        return ERTS_PROC_LOCK_BUSY;
+
+    state = erts_atomic32_read_nob(&rp->state);
+    if (statep)
+        *statep = state;
+
+    if ((state & fail_state)
+        || rp->sig_inq.first
+        || rp->sig_qs.cont) {
+        erts_proc_unlock(rp, tmp_locks);
+        if (state & ERTS_PSFLG_FREE)
+            return NULL;
+        else
+            return ERTS_PROC_LOCK_BUSY;
+    }
+
+    if (tmp_locks != locks)
+        erts_proc_unlock(rp, tmp_locks & ~locks);
+
+    return rp;
+}
+
 /*
  * Stack dump functions follow.
  */
@@ -14203,7 +12589,7 @@ erts_program_counter_info(fmtfn_t to, void *to_arg, Process *p)
     erts_print(to, to_arg, "CP: %p (", p->cp);
     print_function_from_pc(to, to_arg, p->cp);
     erts_print(to, to_arg, ")\n");
-    state = erts_smp_atomic32_read_acqb(&p->state);
+    state = erts_atomic32_read_acqb(&p->state);
     if (!(state & (ERTS_PSFLG_RUNNING
 		   | ERTS_PSFLG_RUNNING_SYS
 		   | ERTS_PSFLG_GC))) {
@@ -14288,7 +12674,6 @@ erts_print_scheduler_info(fmtfn_t to, void *to_arg, ErtsSchedulerData *esdp)
     case ERTS_SCHED_NORMAL:
         erts_print(to, to_arg, "=scheduler:%u\n", esdp->no);
         break;
-#ifdef ERTS_DIRTY_SCHEDULERS
     case ERTS_SCHED_DIRTY_CPU:
         erts_print(to, to_arg, "=dirty_cpu_scheduler:%u\n",
                    (esdp->dirty_no + erts_no_schedulers));
@@ -14297,14 +12682,12 @@ erts_print_scheduler_info(fmtfn_t to, void *to_arg, ErtsSchedulerData *esdp)
         erts_print(to, to_arg, "=dirty_io_scheduler:%u\n",
                    (esdp->dirty_no + erts_no_schedulers + erts_no_dirty_cpu_schedulers));
         break;
-#endif
     default:
         erts_print(to, to_arg, "=unknown_scheduler_type:%u\n", esdp->type);
         break;
     }
 
-#ifdef ERTS_SMP
-    flg = erts_smp_atomic32_read_dirty(&esdp->ssi->flags);
+    flg = erts_atomic32_read_dirty(&esdp->ssi->flags);
     erts_print(to, to_arg, "Scheduler Sleep Info Flags: ");
     for (i = 0; i < ERTS_SSI_FLGS_MAX && flg; i++) {
         erts_aint32_t chk = (1 << i);
@@ -14331,7 +12714,6 @@ erts_print_scheduler_info(fmtfn_t to, void *to_arg, ErtsSchedulerData *esdp)
         }
     }
     erts_print(to, to_arg, "\n");
-#endif
 
     flg = erts_atomic32_read_dirty(&esdp->ssi->aux_work);
     erts_print(to, to_arg, "Scheduler Sleep Info Aux Work: ");
@@ -14388,12 +12770,12 @@ void erts_print_run_queue_info(fmtfn_t to, void *to_arg,
             break;
         }
         erts_print(to, to_arg, "Length: %d\n",
-                   erts_smp_atomic32_read_dirty(&run_queue->procs.prio_info[i].len));
+                   erts_atomic32_read_dirty(&run_queue->procs.prio_info[i].len));
     }
     erts_print(to, to_arg, "Run Queue Port Length: %d\n",
-               erts_smp_atomic32_read_dirty(&run_queue->ports.info.len));
+               erts_atomic32_read_dirty(&run_queue->ports.info.len));
 
-    flg = erts_smp_atomic32_read_dirty(&run_queue->flags);
+    flg = erts_atomic32_read_dirty(&run_queue->flags);
     erts_print(to, to_arg, "Run Queue Flags: ");
     for (i = 0; i < ERTS_RUNQ_FLG_MAX && flg; i++) {
         erts_aint32_t chk = (1 << i);
@@ -14471,7 +12853,7 @@ static void print_current_process_info(fmtfn_t to, void *to_arg,
 
     erts_print(to, to_arg, "Current Process: ");
     if (esdp->current_process && !(ERTS_TRACE_FLAGS(p) & F_SENSITIVE)) {
-	flg = erts_smp_atomic32_read_dirty(&p->state);
+	flg = erts_atomic32_read_dirty(&p->state);
 	erts_print(to, to_arg, "%T\n", p->common.id);
 
 	erts_print(to, to_arg, "Current Process State: ");
@@ -14521,19 +12903,17 @@ static void print_current_process_info(fmtfn_t to, void *to_arg,
  */
 void erts_halt(int code)
 {
-    if (-1 == erts_smp_atomic32_cmpxchg_acqb(&erts_halt_progress,
+    if (-1 == erts_atomic32_cmpxchg_acqb(&erts_halt_progress,
 					     erts_no_schedulers,
 					     -1)) {
-#ifdef ERTS_DIRTY_SCHEDULERS
         ERTS_RUNQ_FLGS_SET(ERTS_DIRTY_CPU_RUNQ, ERTS_RUNQ_FLG_HALTING);
         ERTS_RUNQ_FLGS_SET(ERTS_DIRTY_IO_RUNQ, ERTS_RUNQ_FLG_HALTING);
-#endif
 	erts_halt_code = code;
 	notify_reap_ports_relb();
     }
 }
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 int
 erts_dbg_check_halloc_lock(Process *p)
 {
@@ -14553,3 +12933,24 @@ erts_dbg_check_halloc_lock(Process *p)
     return 0;
 }
 #endif
+
+void
+erts_debug_later_op_foreach(void (*callback)(void*),
+                            void (*func)(void *, ErtsThrPrgrVal, void *),
+                            void *arg)
+{
+    int six;
+    if (!erts_thr_progress_is_blocking())
+	ERTS_INTERNAL_ERROR("Not blocking thread progress");
+
+    for (six = 0; six < erts_no_schedulers; six++) {
+        ErtsSchedulerData *esdp = &erts_aligned_scheduler_data[six].esd;
+	ErtsThrPrgrLaterOp *lop = esdp->aux_work_data.later_op.first;
+
+        while (lop) {
+            if (lop->func == callback)
+                func(arg, lop->later, lop->data);
+            lop = lop->next;
+        }
+    }
+}
diff --git a/erts/emulator/beam/erl_process.h b/erts/emulator/beam/erl_process.h
index 5cac939771..a60e117bab 100644
--- a/erts/emulator/beam/erl_process.h
+++ b/erts/emulator/beam/erl_process.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -47,12 +47,11 @@ typedef struct process Process;
 #include "erl_port.h"
 #undef ERL_PORT_GET_PORT_TYPE_ONLY__
 #include "erl_vm.h"
-#include "erl_smp.h"
 #include "erl_message.h"
 #include "erl_process_dict.h"
 #include "erl_node_container_utils.h"
 #include "erl_node_tables.h"
-#include "erl_monitors.h"
+#include "erl_monitor_link.h"
 #include "erl_hl_timer.h"
 #include "erl_time.h"
 #include "erl_atom_table.h"
@@ -76,8 +75,6 @@ typedef struct process Process;
 #include "erl_thr_progress.h"
 #undef ERL_THR_PROGRESS_TSD_TYPE_ONLY
 
-struct ErtsNodesMonitor_;
-
 #define ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT_OPT	0
 #define ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT		0
 
@@ -106,27 +103,20 @@ struct saved_calls {
 };
 
 extern Export exp_send, exp_receive, exp_timeout;
-extern int erts_eager_check_io;
-extern int erts_sched_compact_load;
-extern int erts_sched_balance_util;
-extern Uint erts_no_schedulers;
-extern Uint erts_no_total_schedulers;
-#ifdef ERTS_DIRTY_SCHEDULERS
-extern Uint erts_no_dirty_cpu_schedulers;
-extern Uint erts_no_dirty_io_schedulers;
-#endif
-extern Uint erts_no_run_queues;
+extern int ERTS_WRITE_UNLIKELY(erts_sched_compact_load);
+extern int ERTS_WRITE_UNLIKELY(erts_sched_balance_util);
+extern Uint ERTS_WRITE_UNLIKELY(erts_no_schedulers);
+extern Uint ERTS_WRITE_UNLIKELY(erts_no_total_schedulers);
+extern Uint ERTS_WRITE_UNLIKELY(erts_no_dirty_cpu_schedulers);
+extern Uint ERTS_WRITE_UNLIKELY(erts_no_dirty_io_schedulers);
+extern Uint ERTS_WRITE_UNLIKELY(erts_no_run_queues);
 extern int erts_sched_thread_suggested_stack_size;
-#ifdef ERTS_DIRTY_SCHEDULERS
 extern int erts_dcpu_sched_thread_suggested_stack_size;
 extern int erts_dio_sched_thread_suggested_stack_size;
-#endif
 #define ERTS_SCHED_THREAD_MIN_STACK_SIZE 20	/* Kilo words */
 #define ERTS_SCHED_THREAD_MAX_STACK_SIZE 8192	/* Kilo words */
 
-#ifdef ERTS_SMP
 #include "erl_bits.h"
-#endif
 
 /* process priorities */
 #define PRIORITY_MAX          0
@@ -224,34 +214,37 @@ extern int erts_dio_sched_thread_suggested_stack_size;
   ((FLGS) &= ~ERTS_RUNQ_FLG_EVACUATE((PRIO)))
 
 #define ERTS_RUNQ_FLGS_INIT(RQ, INIT)					\
-    erts_smp_atomic32_init_nob(&(RQ)->flags, (erts_aint32_t) (INIT))
+    erts_atomic32_init_nob(&(RQ)->flags, (erts_aint32_t) (INIT))
 #define ERTS_RUNQ_FLGS_SET(RQ, FLGS)					\
-    ((Uint32) erts_smp_atomic32_read_bor_relb(&(RQ)->flags,		\
+    ((Uint32) erts_atomic32_read_bor_relb(&(RQ)->flags,		\
 					      (erts_aint32_t) (FLGS)))
 #define ERTS_RUNQ_FLGS_SET_NOB(RQ, FLGS)				\
-    ((Uint32) erts_smp_atomic32_read_bor_nob(&(RQ)->flags,		\
+    ((Uint32) erts_atomic32_read_bor_nob(&(RQ)->flags,		\
 					     (erts_aint32_t) (FLGS)))
 #define ERTS_RUNQ_FLGS_BSET(RQ, MSK, FLGS)				\
-    ((Uint32) erts_smp_atomic32_read_bset_relb(&(RQ)->flags,		\
+    ((Uint32) erts_atomic32_read_bset_relb(&(RQ)->flags,		\
 					       (erts_aint32_t) (MSK),	\
 					       (erts_aint32_t) (FLGS)))
 #define ERTS_RUNQ_FLGS_UNSET(RQ, FLGS)					\
-    ((Uint32) erts_smp_atomic32_read_band_relb(&(RQ)->flags,		\
+    ((Uint32) erts_atomic32_read_band_relb(&(RQ)->flags,		\
 					       (erts_aint32_t) ~(FLGS)))
 #define ERTS_RUNQ_FLGS_UNSET_NOB(RQ, FLGS)					\
-    ((Uint32) erts_smp_atomic32_read_band_nob(&(RQ)->flags,		\
+    ((Uint32) erts_atomic32_read_band_nob(&(RQ)->flags,		\
 					      (erts_aint32_t) ~(FLGS)))
 #define ERTS_RUNQ_FLGS_GET(RQ)						\
-    ((Uint32) erts_smp_atomic32_read_acqb(&(RQ)->flags))
+    ((Uint32) erts_atomic32_read_acqb(&(RQ)->flags))
 #define ERTS_RUNQ_FLGS_GET_NOB(RQ)					\
-    ((Uint32) erts_smp_atomic32_read_nob(&(RQ)->flags))
+    ((Uint32) erts_atomic32_read_nob(&(RQ)->flags))
 #define ERTS_RUNQ_FLGS_GET_MB(RQ)					\
-    ((Uint32) erts_smp_atomic32_read_mb(&(RQ)->flags))
+    ((Uint32) erts_atomic32_read_mb(&(RQ)->flags))
 #define ERTS_RUNQ_FLGS_READ_BSET(RQ, MSK, FLGS)		  		\
-    ((Uint32) erts_smp_atomic32_read_bset_relb(&(RQ)->flags, 		\
+    ((Uint32) erts_atomic32_read_bset_relb(&(RQ)->flags, 		\
 					       (erts_aint32_t) (MSK),	\
 					       (erts_aint32_t) (FLGS)))
 
+#define ERTS_RUNQ_POINTER_MASK  (~((erts_aint_t) 3))
+#define ERTS_RUNQ_BOUND_FLAG    ((erts_aint_t) 1)
+
 typedef enum {
     ERTS_SCHDLR_SSPND_DONE_MSCHED_BLOCKED,
     ERTS_SCHDLR_SSPND_DONE_NMSCHED_BLOCKED,
@@ -316,7 +309,6 @@ typedef enum {
     ERTS_SSI_AUX_WORK_ASYNC_READY_CLEAN_IX,
     ERTS_SSI_AUX_WORK_MISC_THR_PRGR_IX,
     ERTS_SSI_AUX_WORK_MISC_IX,
-    ERTS_SSI_AUX_WORK_PENDING_EXITERS_IX,
     ERTS_SSI_AUX_WORK_SET_TMO_IX,
     ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK_IX,
     ERTS_SSI_AUX_WORK_YIELD_IX,
@@ -350,8 +342,6 @@ typedef enum {
     (((erts_aint32_t) 1) << ERTS_SSI_AUX_WORK_MISC_THR_PRGR_IX)
 #define ERTS_SSI_AUX_WORK_MISC \
     (((erts_aint32_t) 1) << ERTS_SSI_AUX_WORK_MISC_IX)
-#define ERTS_SSI_AUX_WORK_PENDING_EXITERS \
-    (((erts_aint32_t) 1) << ERTS_SSI_AUX_WORK_PENDING_EXITERS_IX)
 #define ERTS_SSI_AUX_WORK_SET_TMO \
     (((erts_aint32_t) 1) << ERTS_SSI_AUX_WORK_SET_TMO_IX)
 #define ERTS_SSI_AUX_WORK_MSEG_CACHE_CHECK \
@@ -365,20 +355,18 @@ typedef enum {
 
 typedef struct ErtsSchedulerSleepInfo_ ErtsSchedulerSleepInfo;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 typedef struct {
-    erts_smp_spinlock_t lock;
+    erts_spinlock_t lock;
     ErtsSchedulerSleepInfo *list;
 } ErtsSchedulerSleepList;
-#endif
 
 struct ErtsSchedulerSleepInfo_ {
-#ifdef ERTS_SMP
+    struct ErtsSchedulerData_ *esdp;
     ErtsSchedulerSleepInfo *next;
     ErtsSchedulerSleepInfo *prev;
-    erts_smp_atomic32_t flags;
+    erts_atomic32_t flags;
     erts_tse_t *event;
-#endif
+    struct erts_poll_thread *psi;
     erts_atomic32_t aux_work;
 };
 
@@ -412,9 +400,12 @@ typedef struct {
 } ErtsRunPrioQueue;
 
 typedef enum {
-    ERTS_SCHED_NORMAL,
-    ERTS_SCHED_DIRTY_CPU,
-    ERTS_SCHED_DIRTY_IO
+    ERTS_SCHED_NORMAL = 0,
+    ERTS_SCHED_DIRTY_CPU = 1,
+    ERTS_SCHED_DIRTY_IO = 2,
+
+    ERTS_SCHED_TYPE_FIRST = ERTS_SCHED_NORMAL,
+    ERTS_SCHED_TYPE_LAST = ERTS_SCHED_DIRTY_IO
 } ErtsSchedType;
 
 typedef struct ErtsSchedulerData_ ErtsSchedulerData;
@@ -422,7 +413,7 @@ typedef struct ErtsSchedulerData_ ErtsSchedulerData;
 typedef struct ErtsRunQueue_ ErtsRunQueue;
 
 typedef struct {
-    erts_smp_atomic32_t len;
+    erts_atomic32_t len;
     erts_aint32_t max_len;
     int reds;
 } ErtsRunQueueInfo;
@@ -433,7 +424,6 @@ typedef struct {
 #  define ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT_OPT 1
 #endif
 
-#ifdef ERTS_SMP
 
 #undef ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT
 #define ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT ERTS_HAVE_SCHED_UTIL_BALANCING_SUPPORT_OPT
@@ -476,35 +466,29 @@ struct ErtsMigrationPaths_ {
     ErtsMigrationPath mpath[1];
 };
 
-#endif /* ERTS_SMP */
 
 struct ErtsRunQueue_ {
     int ix;
 
-    erts_smp_mtx_t mtx;
-    erts_smp_cnd_t cnd;
+    erts_mtx_t mtx;
+    erts_cnd_t cnd;
 
-#ifdef ERTS_DIRTY_SCHEDULERS
-#ifdef ERTS_SMP
     ErtsSchedulerSleepList sleepers;
-#endif
-#endif
 
     ErtsSchedulerData *scheduler;
     int waiting; /* < 0 in sys schedule; > 0 on cnd variable */
     int woken;
-    erts_smp_atomic32_t flags;
+    erts_atomic32_t flags;
     int check_balance_reds;
     int full_reds_history_sum;
     int full_reds_history[ERTS_FULL_REDS_HISTORY_SIZE];
     int out_of_work_count;
     erts_aint32_t max_len;
-    erts_smp_atomic32_t len;
+    erts_atomic32_t len;
     int wakeup_other;
     int wakeup_other_reds;
 
     struct {
-	ErtsProcList *pending_exiters;
 	Uint context_switches;
 	Uint reductions;
 
@@ -518,7 +502,7 @@ struct ErtsRunQueue_ {
     struct {
 	ErtsMiscOpList *start;
 	ErtsMiscOpList *end;
-	erts_smp_atomic_t evac_runq;
+	erts_atomic_t evac_runq;
     } misc;
 
     struct {
@@ -531,16 +515,14 @@ struct ErtsRunQueue_ {
 #endif
 };
 
-#ifdef ERTS_SMP
 extern long erts_runq_supervision_interval;
-#endif
 
 typedef union {
     ErtsRunQueue runq;
     char align[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsRunQueue))];
 } ErtsAlignedRunQueue;
 
-extern ErtsAlignedRunQueue *erts_aligned_run_queues;
+extern ErtsAlignedRunQueue * ERTS_WRITE_UNLIKELY(erts_aligned_run_queues);
 
 #define ERTS_PROC_REDUCTIONS_EXECUTED(SD, RQ, PRIO, REDS, AREDS)\
 do {								\
@@ -581,17 +563,12 @@ typedef struct {
     int sched_id;
     ErtsSchedulerData *esdp;
     ErtsSchedulerSleepInfo *ssi;
-#ifdef ERTS_SMP
     ErtsThrPrgrVal current_thr_prgr;
     ErtsThrPrgrVal latest_wakeup;
-#endif
     struct {
 	int ix;
-#ifdef ERTS_SMP
 	ErtsThrPrgrVal thr_prgr;
-#endif
     } misc;
-#ifdef ERTS_SMP
     struct {
 	ErtsThrPrgrVal thr_prgr;
     } dd;
@@ -604,25 +581,19 @@ typedef struct {
 	ErtsThrPrgrLaterOp *first;
 	ErtsThrPrgrLaterOp *last;
     } later_op;
-#endif
-#ifdef ERTS_USE_ASYNC_READY_Q
     struct {
-#ifdef ERTS_SMP
 	int need_thr_prgr;
 	ErtsThrPrgrVal thr_prgr;
-#endif
 	void *queue;
     } async_ready;
-#endif
-#ifdef ERTS_SMP
     struct {
 	Uint64 next;
 	int *sched2jix;
 	int jix;
 	ErtsDelayedAuxWorkWakeupJob *job;
     } delayed_wakeup;
-#endif
     struct {
+        ErtsAlcuBlockscanYieldData alcu_blockscan;
         ErtsEtsAllYieldData ets_all;
         /* Other yielding operations... */
     } yield;
@@ -639,13 +610,11 @@ typedef struct {
     (&(ESDP)->aux_work_data.yield.NAME)
 void erts_notify_new_aux_yield_work(ErtsSchedulerData *esdp);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 typedef enum {
     ERTS_DIRTY_CPU_SCHEDULER,
     ERTS_DIRTY_IO_SCHEDULER
 } ErtsDirtySchedulerType;
 
-#endif
 
 struct ErtsSchedulerData_ {
     /*
@@ -659,21 +628,18 @@ struct ErtsSchedulerData_ {
     ErtsTimerWheel *timer_wheel;
     ErtsNextTimeoutRef next_tmo_ref;
     ErtsHLTimerService *timer_service;
-#ifdef ERTS_SMP
     ethr_tid tid;		/* Thread id */
     struct erl_bits_state erl_bits_state; /* erl_bits.c state */
     void *match_pseudo_process; /* erl_db_util.c:db_prog_match() */
     Process *free_process;
     ErtsThrPrgrData thr_progress_data;
-#endif
     ErtsSchedulerSleepInfo *ssi;
     Process *current_process;
     ErtsSchedType type;
     Uint no;			/* Scheduler number for normal schedulers */
-#ifdef ERTS_DIRTY_SCHEDULERS
     Uint dirty_no;  /* Scheduler number for dirty schedulers */
+    struct enif_environment_t *current_nif;
     Process *dirty_shadow_process;
-#endif
     Port *current_port;
     ErtsRunQueue *run_queue;
     int virtual_reds;
@@ -694,6 +660,13 @@ struct ErtsSchedulerData_ {
 	Uint64 out;
 	Uint64 in;
     } io;
+    struct {
+        ErtsSignal* sig;
+        Eterm to;
+#ifdef DEBUG
+	Process* dbg_from;
+#endif
+    } pending_signal;
 
     Uint64 reductions;
     ErtsSchedWallTime sched_wall_time;
@@ -711,26 +684,24 @@ typedef union {
     char align[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(ErtsSchedulerData))];
 } ErtsAlignedSchedulerData;
 
-extern ErtsAlignedSchedulerData *erts_aligned_scheduler_data;
-#ifdef ERTS_DIRTY_SCHEDULERS
-extern ErtsAlignedSchedulerData *erts_aligned_dirty_cpu_scheduler_data;
-extern ErtsAlignedSchedulerData *erts_aligned_dirty_io_scheduler_data;
-#endif
+extern ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_scheduler_data);
+extern ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_dirty_cpu_scheduler_data);
+extern ErtsAlignedSchedulerData * ERTS_WRITE_UNLIKELY(erts_aligned_dirty_io_scheduler_data);
 
-#ifndef ERTS_SMP
-extern ErtsSchedulerData *erts_scheduler_data;
-#endif
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-int erts_smp_lc_runq_is_locked(ErtsRunQueue *);
+#if defined(ERTS_ENABLE_LOCK_CHECK)
+int erts_lc_runq_is_locked(ErtsRunQueue *);
 #endif
 
+void
+erts_debug_later_op_foreach(void (*callback)(void*),
+                            void (*func)(void *, ErtsThrPrgrVal, void *),
+                            void *arg);
+
 #ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
 
-#ifdef ERTS_SMP
 void erts_empty_runq(ErtsRunQueue *rq);
 void erts_non_empty_runq(ErtsRunQueue *rq);
-#endif
 
 
 /*
@@ -738,86 +709,84 @@ void erts_non_empty_runq(ErtsRunQueue *rq);
  * other threads peek at values without run queue lock.
  */
 
-ERTS_GLB_INLINE void erts_smp_inc_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio);
-ERTS_GLB_INLINE void erts_smp_dec_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio);
-ERTS_GLB_INLINE void erts_smp_reset_max_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi);
+ERTS_GLB_INLINE void erts_inc_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio);
+ERTS_GLB_INLINE void erts_dec_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio);
+ERTS_GLB_INLINE void erts_reset_max_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
 ERTS_GLB_INLINE void
-erts_smp_inc_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio)
+erts_inc_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio)
 {
     erts_aint32_t len;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
-    len = erts_smp_atomic32_read_dirty(&rq->len);
+    len = erts_atomic32_read_dirty(&rq->len);
 
-#ifdef ERTS_SMP
     if (len == 0)
 	erts_non_empty_runq(rq);
-#endif
     len++;
     if (rq->max_len < len)
 	rq->max_len = len;
     ASSERT(len > 0);
-    erts_smp_atomic32_set_nob(&rq->len, len);
+    erts_atomic32_set_nob(&rq->len, len);
 
-    len = erts_smp_atomic32_read_dirty(&rqi->len);
+    len = erts_atomic32_read_dirty(&rqi->len);
     ASSERT(len >= 0);
     if (len == 0) {
-	ASSERT((erts_smp_atomic32_read_nob(&rq->flags)
+	ASSERT((erts_atomic32_read_nob(&rq->flags)
 		& ((erts_aint32_t) (1 << prio))) == 0);
-	erts_smp_atomic32_read_bor_nob(&rq->flags,
+	erts_atomic32_read_bor_nob(&rq->flags,
 				       (erts_aint32_t) (1 << prio));
     }
     len++;
     if (rqi->max_len < len)
 	rqi->max_len = len;
 
-    erts_smp_atomic32_set_relb(&rqi->len, len);
+    erts_atomic32_set_relb(&rqi->len, len);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_dec_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio)
+erts_dec_runq_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi, int prio)
 {
     erts_aint32_t len;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
-    len = erts_smp_atomic32_read_dirty(&rq->len);
+    len = erts_atomic32_read_dirty(&rq->len);
     len--;
     ASSERT(len >= 0);
-    erts_smp_atomic32_set_nob(&rq->len, len);
+    erts_atomic32_set_nob(&rq->len, len);
 
-    len = erts_smp_atomic32_read_dirty(&rqi->len);
+    len = erts_atomic32_read_dirty(&rqi->len);
     len--;
     ASSERT(len >= 0);
     if (len == 0) {
-	ASSERT((erts_smp_atomic32_read_nob(&rq->flags)
+	ASSERT((erts_atomic32_read_nob(&rq->flags)
 		& ((erts_aint32_t) (1 << prio))));
-	erts_smp_atomic32_read_band_nob(&rq->flags,
+	erts_atomic32_read_band_nob(&rq->flags,
 					~((erts_aint32_t) (1 << prio)));
     }
-    erts_smp_atomic32_set_relb(&rqi->len, len);
+    erts_atomic32_set_relb(&rqi->len, len);
 
 }
 
 ERTS_GLB_INLINE void
-erts_smp_reset_max_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi)
+erts_reset_max_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi)
 {
     erts_aint32_t len;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked(rq));
+    ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq));
 
-    len = erts_smp_atomic32_read_dirty(&rqi->len);
+    len = erts_atomic32_read_dirty(&rqi->len);
     ASSERT(rqi->max_len >= len);
     rqi->max_len = len;
 }
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#define RUNQ_READ_LEN(X) erts_smp_atomic32_read_nob((X))
+#define RUNQ_READ_LEN(X) erts_atomic32_read_nob((X))
 
 #endif /* ERTS_INCLUDE_SCHEDULER_INTERNALS */
 
@@ -835,14 +804,16 @@ erts_smp_reset_max_len(ErtsRunQueue *rq, ErtsRunQueueInfo *rqi)
 #define ERTS_PSD_NIF_TRAP_EXPORT		5
 #define ERTS_PSD_ETS_OWNED_TABLES               6
 #define ERTS_PSD_ETS_FIXED_TABLES               7
-#define ERTS_PSD_SUSPENDED_SAVED_CALLS_BUF	8
+#define ERTS_PSD_DIST_ENTRY	                8
+#define ERTS_PSD_PENDING_SUSPEND                9
+#define ERTS_PSD_SUSPENDED_SAVED_CALLS_BUF	10 /* keep last... */
 
-#define ERTS_PSD_SIZE				9
+#define ERTS_PSD_SIZE				11
 
 #if !defined(HIPE)
 #  undef ERTS_PSD_SUSPENDED_SAVED_CALLS_BUF
 #  undef ERTS_PSD_SIZE
-#  define ERTS_PSD_SIZE 8
+#  define ERTS_PSD_SIZE 10
 #endif
 
 typedef struct {
@@ -876,6 +847,12 @@ typedef struct {
 #define ERTS_PSD_ETS_FIXED_TABLES_GET_LOCKS ERTS_PROC_LOCK_MAIN
 #define ERTS_PSD_ETS_FIXED_TABLES_SET_LOCKS ERTS_PROC_LOCK_MAIN
 
+#define ERTS_PSD_DIST_ENTRY_GET_LOCKS ERTS_PROC_LOCK_MAIN
+#define ERTS_PSD_DIST_ENTRY_SET_LOCKS ERTS_PROC_LOCK_MAIN
+
+#define ERTS_PSD_PENDING_SUSPEND_GET_LOCKS ERTS_PROC_LOCK_MAIN
+#define ERTS_PSD_PENDING_SUSPEND_SET_LOCKS ERTS_PROC_LOCK_MAIN
+
 typedef struct {
     ErtsProcLocks get_locks;
     ErtsProcLocks set_locks;
@@ -890,7 +867,7 @@ extern ErtsLcPSDLocks erts_psd_required_locks[ERTS_PSD_SIZE];
 #define ERTS_SCHED_STAT_MODIFY_CLEAR		3
 
 typedef struct {
-    erts_smp_spinlock_t lock;
+    erts_spinlock_t lock;
     int enabled;
     struct {
 	Eterm name;
@@ -911,22 +888,6 @@ typedef struct {
 typedef struct ErtsProcSysTask_ ErtsProcSysTask;
 typedef struct ErtsProcSysTaskQs_ ErtsProcSysTaskQs;
 
-#ifdef ERTS_SMP
-
-typedef struct ErtsPendingSuspend_ ErtsPendingSuspend;
-struct ErtsPendingSuspend_ {
-    ErtsPendingSuspend *next;
-    ErtsPendingSuspend *end;
-    Eterm pid;
-    void (*handle_func)(Process *suspendee,
-			ErtsProcLocks suspendee_locks,
-			int suspendee_alive,
-			Eterm pid);
-};
-
-#endif
-
-
 /* Defines to ease the change of memory architecture */
 #  define HEAP_START(p)     (p)->heap
 #  define HEAP_TOP(p)       (p)->htop
@@ -1021,14 +982,7 @@ struct process {
 
     Process *next;		/* Pointer to next process in run queue */
 
-    struct ErtsNodesMonitor_ *nodes_monitors;
-
-    ErtsSuspendMonitor *suspend_monitors; /* Processes suspended by
-					     this process via
-					     erlang:suspend_process/1 */
-
-    ErlMessageQueue msg;	/* Message queue */
-
+    ErtsSignalPrivQueues sig_qs; /* Signal queues */
     ErtsBifTimers *bif_timers;	/* Bif timers aiming at this process */
 
     ProcDict  *dictionary;       /* Process dictionary, may be NULL */
@@ -1057,7 +1011,6 @@ struct process {
      * Information mainly for post-mortem use (erl crash dump).
      */
     Eterm parent;		/* Pid of process that created this process. */
-    erts_approx_time_t approx_started; /* Time when started. */
 
     Uint32 static_flags;        /* Flags that do *not* change */
 
@@ -1076,35 +1029,23 @@ struct process {
     ErlHeapFragment* live_hf_end;
     ErtsMessage *msg_frag;	/* Pointer to message fragment list */
     Uint mbuf_sz;		/* Total size of heap fragments and message fragments */
-    erts_smp_atomic_t psd;		/* Rarely used process specific data */
+    erts_atomic_t psd;		/* Rarely used process specific data */
 
     Uint64 bin_vheap_sz;	/* Virtual heap block size for binaries */
     Uint64 bin_old_vheap_sz;	/* Virtual old heap block size for binaries */
     Uint64 bin_old_vheap;	/* Virtual old heap size for binaries */
 
     ErtsProcSysTaskQs *sys_task_qs;
-#ifdef ERTS_DIRTY_SCHEDULERS
     ErtsProcSysTask *dirty_sys_tasks;
-#endif
 
-    erts_smp_atomic32_t state;  /* Process state flags (see ERTS_PSFLG_*) */
-#ifdef ERTS_DIRTY_SCHEDULERS
-    erts_smp_atomic32_t dirty_state; /* Process dirty state flags (see ERTS_PDSFLG_*) */
-#endif
+    erts_atomic32_t state;  /* Process state flags (see ERTS_PSFLG_*) */
+    erts_atomic32_t dirty_state; /* Process dirty state flags (see ERTS_PDSFLG_*) */
 
-#ifdef ERTS_SMP
-    ErlMessageInQueue msg_inq;
+    ErtsSignalInQueue sig_inq;
     ErlTraceMessageQueue *trace_msg_q;
-    ErtsPendExit pending_exit;
     erts_proc_lock_t lock;
     ErtsSchedulerData *scheduler_data;
-    Eterm suspendee;
-    ErtsPendingSuspend *pending_suspenders;
-    erts_smp_atomic_t run_queue;
-#ifdef HIPE
-    struct hipe_process_state_smp hipe_smp;
-#endif
-#endif
+    erts_atomic_t run_queue;
 
 #ifdef CHECK_FOR_HOLES
     Eterm* last_htop;		/* No need to scan the heap below this point. */
@@ -1130,6 +1071,7 @@ struct process {
 #endif
 };
 
+extern Eterm erts_init_process_id; /* pid of init process */
 extern const Process erts_invalid_process;
 
 #ifdef CHECK_FOR_HOLES
@@ -1206,20 +1148,20 @@ void erts_check_for_holes(Process* p);
 #define ERTS_PSFLG_IN_PRQ_LOW 		ERTS_PSFLG_BIT(3)
 #define ERTS_PSFLG_FREE			ERTS_PSFLG_BIT(4)
 #define ERTS_PSFLG_EXITING		ERTS_PSFLG_BIT(5)
-#define ERTS_PSFLG_PENDING_EXIT		ERTS_PSFLG_BIT(6)
+#define ERTS_PSFLG_UNUSED		ERTS_PSFLG_BIT(6)
 #define ERTS_PSFLG_ACTIVE		ERTS_PSFLG_BIT(7)
 #define ERTS_PSFLG_IN_RUNQ		ERTS_PSFLG_BIT(8)
 #define ERTS_PSFLG_RUNNING		ERTS_PSFLG_BIT(9)
 #define ERTS_PSFLG_SUSPENDED		ERTS_PSFLG_BIT(10)
 #define ERTS_PSFLG_GC			ERTS_PSFLG_BIT(11)
-#define ERTS_PSFLG_BOUND		ERTS_PSFLG_BIT(12)
-#define ERTS_PSFLG_TRAP_EXIT		ERTS_PSFLG_BIT(13)
+#define ERTS_PSFLG_SYS_TASKS		ERTS_PSFLG_BIT(12)
+#define ERTS_PSFLG_SIG_IN_Q		ERTS_PSFLG_BIT(13)
 #define ERTS_PSFLG_ACTIVE_SYS		ERTS_PSFLG_BIT(14)
 #define ERTS_PSFLG_RUNNING_SYS		ERTS_PSFLG_BIT(15)
 #define ERTS_PSFLG_PROXY		ERTS_PSFLG_BIT(16)
 #define ERTS_PSFLG_DELAYED_SYS		ERTS_PSFLG_BIT(17)
 #define ERTS_PSFLG_OFF_HEAP_MSGQ	ERTS_PSFLG_BIT(18)
-#define ERTS_PSFLG_ON_HEAP_MSGQ		ERTS_PSFLG_BIT(19)
+#define ERTS_PSFLG_SIG_Q		ERTS_PSFLG_BIT(19)
 #define ERTS_PSFLG_DIRTY_CPU_PROC	ERTS_PSFLG_BIT(20)
 #define ERTS_PSFLG_DIRTY_IO_PROC	ERTS_PSFLG_BIT(21)
 #define ERTS_PSFLG_DIRTY_ACTIVE_SYS	ERTS_PSFLG_BIT(22)
@@ -1238,7 +1180,6 @@ void erts_check_for_holes(Process* p);
 					 | ERTS_PSFLG_IN_PRQ_LOW)
 
 #define ERTS_PSFLGS_VOLATILE_HEAP	(ERTS_PSFLG_EXITING		\
-					 | ERTS_PSFLG_PENDING_EXIT	\
 					 | ERTS_PSFLG_DIRTY_RUNNING	\
 					 | ERTS_PSFLG_DIRTY_RUNNING_SYS)
 
@@ -1249,7 +1190,6 @@ void erts_check_for_holes(Process* p);
 #define ERTS_PSFLGS_GET_PRQ_PRIO(PSFLGS) \
     (((PSFLGS) >> ERTS_PSFLGS_PRQ_PRIO_OFFSET) & ERTS_PSFLGS_PRIO_MASK)
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 
 /*
  * Flags in the dirty_state field.
@@ -1276,7 +1216,6 @@ void erts_check_for_holes(Process* p);
 					 | ERTS_PDSFLG_IN_CPU_PRQ_HIGH	\
 					 | ERTS_PDSFLG_IN_CPU_PRQ_NORMAL\
 					 | ERTS_PDSFLG_IN_CPU_PRQ_LOW)
-#endif
 
 
 /*
@@ -1305,7 +1244,24 @@ void erts_check_for_holes(Process* p);
 #define SEQ_TRACE_T_SENDER(token)   (*(tuple_val(token) + 4))
 #define SEQ_TRACE_T_LASTCNT(token)  (*(tuple_val(token) + 5))
 
+#ifdef USE_VM_PROBES
+/* The dtrace probe for seq_trace only supports 'int' labels, so we represent
+ * all values that won't fit into a 32-bit signed integer as ERTS_SINT32_MIN
+ * (bigints, tuples, etc). */
+
+#define SEQ_TRACE_T_DTRACE_LABEL(token) \
+    DTRACE_SEQ_TRACE_LABEL__(SEQ_TRACE_T_LABEL(token))
+
+#define DTRACE_SEQ_TRACE_LABEL__(label_term) \
+    (is_small((label_term)) ? \
+        ((signed_val((label_term)) <= ERTS_SINT32_MAX && \
+          signed_val((label_term)) >= ERTS_SINT32_MIN) ? \
+             signed_val((label_term)) : ERTS_SINT32_MIN) \
+        : ERTS_SINT32_MIN)
+#endif
+
 /*
+
  * Possible flags for the flags field in ErlSpawnOpts below.
  */
 
@@ -1316,7 +1272,7 @@ void erts_check_for_holes(Process* p);
 #define SPO_OFF_HEAP_MSGQ 16
 #define SPO_ON_HEAP_MSGQ 32
 
-extern int erts_default_spo_flags;
+extern int ERTS_WRITE_UNLIKELY(erts_default_spo_flags);
 
 /*
  * The following struct contains options for a process to be spawned.
@@ -1367,15 +1323,15 @@ Eterm* erts_heap_alloc(Process* p, Uint need, Uint xtra);
 Eterm* erts_set_hole_marker(Eterm* ptr, Uint sz);
 #endif
 
-extern erts_smp_rwmtx_t erts_cpu_bind_rwmtx;
+extern erts_rwmtx_t erts_cpu_bind_rwmtx;
 /* If any of the erts_system_monitor_* variables are set (enabled),
 ** erts_system_monitor must be != NIL, to allow testing on just
 ** the erts_system_monitor_* variables.
 */
-extern Eterm erts_system_monitor;
-extern Uint erts_system_monitor_long_gc;
-extern Uint erts_system_monitor_long_schedule;
-extern Uint erts_system_monitor_large_heap;
+extern Eterm ERTS_WRITE_UNLIKELY(erts_system_monitor);
+extern Uint ERTS_WRITE_UNLIKELY(erts_system_monitor_long_gc);
+extern Uint ERTS_WRITE_UNLIKELY(erts_system_monitor_long_schedule);
+extern Uint ERTS_WRITE_UNLIKELY(erts_system_monitor_large_heap);
 struct erts_system_monitor_flags_t {
 	 unsigned int busy_port : 1;
     unsigned int busy_dist_port : 1;
@@ -1406,7 +1362,7 @@ extern int erts_system_profile_ts_type;
 #define F_DISTRIBUTION       (1 <<  6) /* Process used in distribution */
 #define F_USING_DDLL         (1 <<  7) /* Process has used the DDLL interface */
 #define F_HAVE_BLCKD_MSCHED  (1 <<  8) /* Process has blocked multi-scheduling */
-#define F_P2PNR_RESCHED      (1 <<  9) /* Process has been rescheduled via erts_pid2proc_not_running() */
+#define F_UNUSED             (1 <<  9)
 #define F_FORCE_GC           (1 << 10) /* Force gc at process in-scheduling */
 #define F_DISABLE_GC         (1 << 11) /* Disable GC (see below) */
 #define F_OFF_HEAP_MSGQ      (1 << 12) /* Off heap msg queue */
@@ -1423,6 +1379,12 @@ extern int erts_system_profile_ts_type;
 #define F_DIRTY_MAJOR_GC     (1 << 23) /* Dirty major GC scheduled */
 #define F_DIRTY_MINOR_GC     (1 << 24) /* Dirty minor GC scheduled */
 #define F_HIBERNATED         (1 << 25) /* Hibernated */
+#define F_LOCAL_SIGS_ONLY    (1 << 26) /* Handle privq sigs only */
+#define F_TRAP_EXIT          (1 << 27) /* Trapping exit */
+#define F_DEFERRED_SAVED_LAST (1 << 28) /* Deferred sig_qs.saved_last */
+#define F_DELAYED_PSIGQS_LEN (1 << 29) /* Delayed update of sig_qs.len */
+#define F_HIPE_RECV_LOCKED   (1 << 30) /* HiPE message queue locked */
+#define F_HIPE_RECV_YIELD    (1 << 31) /* HiPE receive yield */
 
 /*
  * F_DISABLE_GC and F_DELAY_GC are similar. Both will prevent
@@ -1488,7 +1450,7 @@ extern int erts_system_profile_ts_type;
 		     | F_TRACE_ARITY_ONLY | F_TRACE_RETURN_TO \
                      | F_TRACE_SILENT | F_TRACE_SCHED_PROCS | F_TRACE_PORTS \
 		     | F_TRACE_SCHED_PORTS | F_TRACE_SCHED_NO \
-		     | F_TRACE_SCHED_EXIT)
+		     | F_TRACE_SCHED_EXIT )
 
 
 #define ERTS_TRACEE_MODIFIER_FLAGS \
@@ -1501,6 +1463,14 @@ extern int erts_system_profile_ts_type;
 
 #define SEQ_TRACE_FLAG(N)        (1 << (ERTS_TRACE_TS_TYPE_BITS + (N)))
 
+#define ERTS_SIG_ENABLE_TRACE_FLAGS \
+    ( F_TRACE_RECEIVE | F_TRACE_PROCS)
+
+/*
+ * F_TRACE_RECEIVE is always enabled/disable via signaling.
+ * F_TRACE_PROCS enable/disable F_TRACE_PROCS_SIG via signaling.
+ */
+
 /* Sequential trace flags */
 
 /* SEQ_TRACE_TIMESTAMP_MASK is a bit-field */
@@ -1527,10 +1497,6 @@ extern int erts_system_profile_ts_type;
 #define DT_UTAG_FLAGS(P)  ((P)->dt_utag_flags) 
 #endif
 
-/* Option flags to erts_send_exit_signal() */
-#define ERTS_XSIG_FLG_IGN_KILL		(((Uint32) 1) << 0)
-#define ERTS_XSIG_FLG_NO_IGN_NORMAL	(((Uint32) 1) << 1)
-
 #define CANCEL_TIMER(P)					\
     do {						\
 	if ((P)->flags & (F_INSLPQUEUE|F_TIMO)) {	\
@@ -1541,20 +1507,14 @@ extern int erts_system_profile_ts_type;
 	}						\
     } while (0)
 
-#if defined(ERTS_DIRTY_SCHEDULERS) && defined(ERTS_SMP)
 #define ERTS_NUM_DIRTY_CPU_RUNQS 1
 #define ERTS_NUM_DIRTY_IO_RUNQS 1
-#else
-#define ERTS_NUM_DIRTY_CPU_RUNQS 0
-#define ERTS_NUM_DIRTY_IO_RUNQS 0
-#endif
 
 #define ERTS_NUM_DIRTY_RUNQS (ERTS_NUM_DIRTY_CPU_RUNQS+ERTS_NUM_DIRTY_IO_RUNQS)
 
 #define ERTS_RUNQ_IX(IX)						\
   (ASSERT(0 <= (IX) && (IX) < erts_no_run_queues+ERTS_NUM_DIRTY_RUNQS), \
    &erts_aligned_run_queues[(IX)].runq)
-#ifdef ERTS_DIRTY_SCHEDULERS
 #define ERTS_RUNQ_IX_IS_DIRTY(IX)					\
   (ASSERT(0 <= (IX) && (IX) < erts_no_run_queues+ERTS_NUM_DIRTY_RUNQS), \
    (erts_no_run_queues <= (IX)))
@@ -1565,13 +1525,9 @@ extern int erts_system_profile_ts_type;
 #define ERTS_DIRTY_IO_RUNQ  (&erts_aligned_run_queues[erts_no_run_queues+1].runq)
 #define ERTS_RUNQ_IS_DIRTY_CPU_RUNQ(RQ) ((RQ) == ERTS_DIRTY_CPU_RUNQ)
 #define ERTS_RUNQ_IS_DIRTY_IO_RUNQ(RQ) ((RQ) == ERTS_DIRTY_IO_RUNQ)
-#else
-#define ERTS_RUNQ_IX_IS_DIRTY(IX) 0
-#endif
 #define ERTS_SCHEDULER_IX(IX)						\
   (ASSERT(0 <= (IX) && (IX) < erts_no_schedulers),			\
    &erts_aligned_scheduler_data[(IX)].esd)
-#ifdef ERTS_DIRTY_SCHEDULERS
 #define ERTS_DIRTY_CPU_SCHEDULER_IX(IX)					\
   (ASSERT(0 <= (IX) && (IX) < erts_no_dirty_cpu_schedulers),		\
    &erts_aligned_dirty_cpu_scheduler_data[(IX)].esd)
@@ -1584,24 +1540,12 @@ extern int erts_system_profile_ts_type;
     ((ESDP)->type == ERTS_SCHED_DIRTY_CPU)
 #define ERTS_SCHEDULER_IS_DIRTY_IO(ESDP)				\
     ((ESDP)->type == ERTS_SCHED_DIRTY_IO)
-#else /* !ERTS_DIRTY_SCHEDULERS */
-#define ERTS_RUNQ_IX_IS_DIRTY(IX) 0
-#define ERTS_SCHEDULER_IS_DIRTY(ESDP) 0
-#define ERTS_SCHEDULER_IS_DIRTY_CPU(ESDP) 0
-#define ERTS_SCHEDULER_IS_DIRTY_IO(ESDP) 0
-#endif
 
 void erts_pre_init_process(void);
 void erts_late_init_process(void);
 void erts_early_init_scheduling(int);
-void erts_init_scheduling(int, int
-#ifdef ERTS_DIRTY_SCHEDULERS
-			  , int, int, int
-#endif
-			  );
-#ifdef ERTS_DIRTY_SCHEDULERS
+void erts_init_scheduling(int, int, int, int, int, int);
 void erts_execute_dirty_system_task(Process *c_p);
-#endif
 int erts_set_gc_state(Process *c_p, int enable);
 Eterm erts_sched_wall_time_request(Process *c_p, int set, int enable,
                                    int dirty_cpu, int want_dirty_io);
@@ -1782,9 +1726,9 @@ ERTS_GLB_INLINE int erts_proclist_is_last(ErtsProcList *list,
 
 #endif
 
-int erts_sched_set_wakeup_other_thresold(char *str);
-int erts_sched_set_wakeup_other_type(char *str);
-int erts_sched_set_busy_wait_threshold(char *str);
+int erts_sched_set_wakeup_other_threshold(ErtsSchedType sched_type, char *str);
+int erts_sched_set_wakeup_other_type(ErtsSchedType sched_type, char *str);
+int erts_sched_set_busy_wait_threshold(ErtsSchedType sched_type, char *str);
 int erts_sched_set_wake_cleanup_threshold(char *);
 
 void erts_schedule_thr_prgr_later_op(void (*)(void *),
@@ -1799,15 +1743,13 @@ struct db_fixation;
 void erts_schedule_ets_free_fixation(Eterm pid, struct db_fixation*);
 void erts_schedule_flush_trace_messages(Process *proc, int force_on_proc);
 int erts_flush_trace_messages(Process *c_p, ErtsProcLocks locks);
+int erts_sig_prio(Eterm pid, int prio);
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 int erts_dbg_check_halloc_lock(Process *p);
 #endif
-#if defined(ERTS_SMP) || defined(ERTS_DIRTY_SCHEDULERS)
 void
 erts_schedulers_state(Uint *, Uint *, Uint *, Uint *, Uint *, Uint *, Uint *, Uint *);
-#endif
-#ifdef ERTS_SMP
 ErtsSchedSuspendResult
 erts_set_schedulers_online(Process *p,
 			   ErtsProcLocks plocks,
@@ -1822,14 +1764,9 @@ void erts_start_schedulers(void);
 void erts_alloc_notify_delayed_dealloc(int);
 void erts_alloc_ensure_handle_delayed_dealloc_call(int);
 void erts_notify_canceled_timer(ErtsSchedulerData *, int);
-#endif
-#if ERTS_USE_ASYNC_READY_Q
 void erts_notify_check_async_ready_queue(void *);
-#endif
-#ifdef ERTS_SMP
 void erts_notify_code_ix_activation(Process* p, ErtsThrPrgrVal later);
 void erts_notify_finish_breakpointing(Process* p);
-#endif
 void erts_schedule_misc_aux_work(int sched_id,
 				 void (*func)(void *),
 				 void *arg);
@@ -1855,8 +1792,10 @@ ErtsRunQueue *erts_schedid2runq(Uint);
 Process *erts_schedule(ErtsSchedulerData *, Process*, int);
 void erts_schedule_misc_op(void (*)(void *), void *);
 Eterm erl_create_process(Process*, Eterm, Eterm, Eterm, ErlSpawnOpts*);
+void erts_set_self_exiting(Process *, Eterm);
 void erts_do_exit_process(Process*, Eterm);
 void erts_continue_exit_process(Process *);
+void erts_proc_exit_link(Process *, ErtsLink *, Uint16, Eterm, Eterm);
 /* Begin System profile */
 Uint erts_runnable_process_count(void);
 /* End System profile */
@@ -1872,8 +1811,19 @@ void erts_print_scheduler_info(fmtfn_t to, void *to_arg, ErtsSchedulerData *esdp
 void erts_print_run_queue_info(fmtfn_t, void *to_arg, ErtsRunQueue*);
 void erts_dump_extended_process_state(fmtfn_t to, void *to_arg, erts_aint32_t psflg);
 void erts_dump_process_state(fmtfn_t to, void *to_arg, erts_aint32_t psflg);
+Eterm erts_process_info(Process *c_p, ErtsHeapFactory *hfact,
+                        Process *rp, ErtsProcLocks rp_locks,
+                        int *item_ix, int item_ix_len,
+                        int flags, Uint reserve_size, Uint *reds);
 
-Eterm erts_get_process_priority(Process *p);
+typedef struct {
+    Process *c_p;
+    Eterm reason;
+} ErtsProcExitContext;
+void erts_proc_exit_handle_monitor(ErtsMonitor *mon, void *vctxt);
+void erts_proc_exit_handle_link(ErtsLink *lnk, void *vctxt);
+
+Eterm erts_get_process_priority(erts_aint32_t state);
 Eterm erts_set_process_priority(Process *p, Eterm prio);
 
 Uint erts_get_total_context_switches(void);
@@ -1891,22 +1841,6 @@ void erts_suspend(Process*, ErtsProcLocks, Port*);
 void erts_resume(Process*, ErtsProcLocks);
 int erts_resume_processes(ErtsProcList *);
 
-int erts_send_exit_signal(Process *,
-			  Eterm,
-			  Process *,
-			  ErtsProcLocks *,
-			  Eterm,
-			  Eterm,
-			  Process *,
-			  Uint32);
-#ifdef ERTS_SMP
-void erts_handle_pending_exit(Process *, ErtsProcLocks);
-#define ERTS_PROC_PENDING_EXIT(P) \
-    (ERTS_PSFLG_PENDING_EXIT & erts_smp_atomic32_read_acqb(&(P)->state))
-#else
-#define ERTS_PROC_PENDING_EXIT(P) 0
-#endif
-
 void erts_deep_process_dump(fmtfn_t, void *);
 
 Eterm erts_get_reader_groups_map(Process *c_p);
@@ -1919,7 +1853,7 @@ Uint erts_debug_nbalance(void);
 
 int erts_debug_wait_completed(Process *c_p, int flags);
 
-Uint erts_process_memory(Process *c_p, int incl_msg_inq);
+Uint erts_process_memory(Process *c_p, int include_sigs_in_transit);
 
 #ifdef ERTS_DO_VERIFY_UNUSED_TEMP_ALLOC
 #  define ERTS_VERIFY_UNUSED_TEMP_ALLOC(P)					\
@@ -1935,21 +1869,11 @@ do {										\
 #  define ERTS_VERIFY_UNUSED_TEMP_ALLOC(ESDP)
 #endif
 
-#if defined(ERTS_SMP) || defined(USE_THREADS)
 ErtsSchedulerData *erts_get_scheduler_data(void);
-#else
-ERTS_GLB_INLINE ErtsSchedulerData *erts_get_scheduler_data(void);
-#if ERTS_GLB_INLINE_INCL_FUNC_DEF
-
-ERTS_GLB_INLINE
-ErtsSchedulerData *erts_get_scheduler_data(void)
-{
-    return erts_scheduler_data;
-}
-#endif
-#endif
 
 void erts_schedule_process(Process *, erts_aint32_t, ErtsProcLocks);
+erts_aint32_t erts_proc_sys_schedule(Process *p, erts_aint32_t state,
+                                     erts_aint32_t enable_flag);
 
 ERTS_GLB_INLINE void erts_proc_notify_new_message(Process *p, ErtsProcLocks locks);
 ERTS_GLB_INLINE void erts_schedule_dirty_sys_execution(Process *c_p);
@@ -1959,7 +1883,7 @@ ERTS_GLB_INLINE void
 erts_proc_notify_new_message(Process *p, ErtsProcLocks locks)
 {
     /* No barrier needed, due to msg lock */
-    erts_aint32_t state = erts_smp_atomic32_read_nob(&p->state);
+    erts_aint32_t state = erts_atomic32_read_nob(&p->state);
     if (!(state & ERTS_PSFLG_ACTIVE))
 	erts_schedule_process(p, state, locks);
 }
@@ -1969,7 +1893,7 @@ erts_schedule_dirty_sys_execution(Process *c_p)
 {
     erts_aint32_t a, n, e;
 
-    a = erts_smp_atomic32_read_nob(&c_p->state);
+    a = erts_atomic32_read_nob(&c_p->state);
 
     /*
      * Only a currently executing process schedules
@@ -1981,11 +1905,10 @@ erts_schedule_dirty_sys_execution(Process *c_p)
     /* Don't set dirty-active-sys if we are about to exit... */
 
     while (!(a & (ERTS_PSFLG_DIRTY_ACTIVE_SYS
-                  | ERTS_PSFLG_EXITING
-                  | ERTS_PSFLG_PENDING_EXIT))) {
+                  | ERTS_PSFLG_EXITING))) {
         e = a;
         n = a | ERTS_PSFLG_DIRTY_ACTIVE_SYS;
-        a = erts_smp_atomic32_cmpxchg_mb(&c_p->state, n, e);
+        a = erts_atomic32_cmpxchg_mb(&c_p->state, n, e);
         if (a == e)
             break; /* dirty-active-sys set */
     }
@@ -1993,21 +1916,21 @@ erts_schedule_dirty_sys_execution(Process *c_p)
 
 #endif
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 
 #define ERTS_PROCESS_LOCK_ONLY_LOCK_CHECK_PROTO__
 #include "erl_process_lock.h"
 #undef ERTS_PROCESS_LOCK_ONLY_LOCK_CHECK_PROTO__
 
-#define ERTS_SMP_LC_CHK_RUNQ_LOCK(RQ, L)				\
+#define ERTS_LC_CHK_RUNQ_LOCK(RQ, L)				\
 do {									\
     if ((L))								\
-	ERTS_SMP_LC_ASSERT(erts_smp_lc_runq_is_locked((RQ)));		\
+	ERTS_LC_ASSERT(erts_lc_runq_is_locked((RQ)));		\
     else								\
-	ERTS_SMP_LC_ASSERT(!erts_smp_lc_runq_is_locked((RQ)));		\
+	ERTS_LC_ASSERT(!erts_lc_runq_is_locked((RQ)));		\
 } while (0)
 #else
-#define ERTS_SMP_LC_CHK_RUNQ_LOCK(RQ, L)
+#define ERTS_LC_CHK_RUNQ_LOCK(RQ, L)
 #endif
 
 void *erts_psd_set_init(Process *p, int ix, void *data);
@@ -2023,22 +1946,22 @@ ERTS_GLB_INLINE void *
 erts_psd_get(Process *p, int ix)
 {
     ErtsPSD *psd;
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     ErtsProcLocks locks = erts_proc_lc_my_proc_locks(p);
     if (ERTS_LC_PSD_ANY_LOCK == erts_psd_required_locks[ix].get_locks)
-	ERTS_SMP_LC_ASSERT(locks || erts_thr_progress_is_blocking());
+	ERTS_LC_ASSERT(locks || erts_thr_progress_is_blocking());
     else {
 	locks &= erts_psd_required_locks[ix].get_locks;
-	ERTS_SMP_LC_ASSERT(erts_psd_required_locks[ix].get_locks == locks
+	ERTS_LC_ASSERT(erts_psd_required_locks[ix].get_locks == locks
 			   || erts_thr_progress_is_blocking());
     }
 #endif
 
-    psd = (ErtsPSD *) erts_smp_atomic_read_nob(&p->psd);
+    psd = (ErtsPSD *) erts_atomic_read_nob(&p->psd);
     ASSERT(0 <= ix && ix < ERTS_PSD_SIZE);
     if (!psd)
 	return NULL;
-    ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
+    ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER;
     return psd->data[ix];
 }
 
@@ -2046,30 +1969,28 @@ ERTS_GLB_INLINE void *
 erts_psd_set(Process *p, int ix, void *data)
 {
     ErtsPSD *psd;
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     ErtsProcLocks locks = erts_proc_lc_my_proc_locks(p);
-    erts_aint32_t state = state = erts_smp_atomic32_read_nob(&p->state);
+    erts_aint32_t state = state = erts_atomic32_read_nob(&p->state);
     if (!(state & ERTS_PSFLG_FREE)) {
 	if (ERTS_LC_PSD_ANY_LOCK == erts_psd_required_locks[ix].set_locks)
-	    ERTS_SMP_LC_ASSERT(locks || erts_thr_progress_is_blocking());
+	    ERTS_LC_ASSERT(locks || erts_thr_progress_is_blocking());
 	else {
 	    locks &= erts_psd_required_locks[ix].set_locks;
-	    ERTS_SMP_LC_ASSERT(erts_psd_required_locks[ix].set_locks == locks
+	    ERTS_LC_ASSERT(erts_psd_required_locks[ix].set_locks == locks
 			       || erts_thr_progress_is_blocking());
 	}
     }
 #endif
-    psd = (ErtsPSD *) erts_smp_atomic_read_nob(&p->psd);
+    psd = (ErtsPSD *) erts_atomic_read_nob(&p->psd);
     ASSERT(0 <= ix && ix < ERTS_PSD_SIZE);
     if (psd) {
 	void *old;
-#ifdef ERTS_SMP
 #ifdef ETHR_ORDERED_READ_DEPEND
 	ETHR_MEMBAR(ETHR_LoadStore|ETHR_StoreStore);
 #else
 	ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore|ETHR_StoreStore);
 #endif
-#endif
 	old = psd->data[ix];
 	psd->data[ix] = data;
 	return old;
@@ -2106,6 +2027,16 @@ erts_psd_set(Process *p, int ix, void *data)
 #define ERTS_PROC_SET_NIF_TRAP_EXPORT(P, NTE) \
     erts_psd_set((P), ERTS_PSD_NIF_TRAP_EXPORT, (void *) (NTE))
 
+#define ERTS_PROC_GET_DIST_ENTRY(P) \
+    ((DistEntry *) erts_psd_get((P), ERTS_PSD_DIST_ENTRY))
+#define ERTS_PROC_SET_DIST_ENTRY(P, DE) \
+    ((DistEntry *) erts_psd_set((P), ERTS_PSD_DIST_ENTRY, (void *) (DE)))
+
+#define ERTS_PROC_GET_PENDING_SUSPEND(P) \
+    ((void *) erts_psd_get((P), ERTS_PSD_PENDING_SUSPEND))
+#define ERTS_PROC_SET_PENDING_SUSPEND(P, PS) \
+    ((void *) erts_psd_set((P), ERTS_PSD_PENDING_SUSPEND, (void *) (PS)))
+
 #ifdef HIPE
 #define ERTS_PROC_GET_SUSPENDED_SAVED_CALLS_BUF(P) \
   ((struct saved_calls *) erts_psd_get((P), ERTS_PSD_SUSPENDED_SAVED_CALLS_BUF))
@@ -2149,7 +2080,6 @@ erts_proc_set_error_handler(Process *p, Eterm handler)
 
 #ifdef ERTS_INCLUDE_SCHEDULER_INTERNALS
 
-#ifdef ERTS_SMP
 
 #include "erl_thr_progress.h"
 
@@ -2251,7 +2181,6 @@ erts_check_emigration_need(ErtsRunQueue *c_rq, int prio)
 
 #endif
 
-#endif
 
 #endif
 
@@ -2260,15 +2189,20 @@ ERTS_GLB_INLINE int erts_is_scheduler_bound(ErtsSchedulerData *esdp);
 ERTS_GLB_INLINE Process *erts_get_current_process(void);
 ERTS_GLB_INLINE Eterm erts_get_current_pid(void);
 ERTS_GLB_INLINE Uint erts_get_scheduler_id(void);
-ERTS_GLB_INLINE ErtsRunQueue *erts_get_runq_proc(Process *p);
+ERTS_GLB_INLINE void erts_init_runq_proc(Process *p, ErtsRunQueue *rq, int bnd);
+ERTS_GLB_INLINE ErtsRunQueue *erts_set_runq_proc(Process *p, ErtsRunQueue *rq, int *boundp);
+ERTS_GLB_INLINE int erts_try_change_runq_proc(Process *p, ErtsRunQueue *rq);
+ERTS_GLB_INLINE ErtsRunQueue *erts_bind_runq_proc(Process *p, int bind);
+ERTS_GLB_INLINE int erts_proc_runq_is_bound(Process *p);
+ERTS_GLB_INLINE ErtsRunQueue *erts_get_runq_proc(Process *p, int *boundp);
 ERTS_GLB_INLINE ErtsRunQueue *erts_get_runq_current(ErtsSchedulerData *esdp);
-ERTS_GLB_INLINE void erts_smp_runq_lock(ErtsRunQueue *rq);
-ERTS_GLB_INLINE int erts_smp_runq_trylock(ErtsRunQueue *rq);
-ERTS_GLB_INLINE void erts_smp_runq_unlock(ErtsRunQueue *rq);
-ERTS_GLB_INLINE void erts_smp_xrunq_lock(ErtsRunQueue *rq, ErtsRunQueue *xrq);
-ERTS_GLB_INLINE void erts_smp_xrunq_unlock(ErtsRunQueue *rq, ErtsRunQueue *xrq);
-ERTS_GLB_INLINE void erts_smp_runqs_lock(ErtsRunQueue *rq1, ErtsRunQueue *rq2);
-ERTS_GLB_INLINE void erts_smp_runqs_unlock(ErtsRunQueue *rq1, ErtsRunQueue *rq2);
+ERTS_GLB_INLINE void erts_runq_lock(ErtsRunQueue *rq);
+ERTS_GLB_INLINE int erts_runq_trylock(ErtsRunQueue *rq);
+ERTS_GLB_INLINE void erts_runq_unlock(ErtsRunQueue *rq);
+ERTS_GLB_INLINE void erts_xrunq_lock(ErtsRunQueue *rq, ErtsRunQueue *xrq);
+ERTS_GLB_INLINE void erts_xrunq_unlock(ErtsRunQueue *rq, ErtsRunQueue *xrq);
+ERTS_GLB_INLINE void erts_runqs_lock(ErtsRunQueue *rq1, ErtsRunQueue *rq2);
+ERTS_GLB_INLINE void erts_runqs_unlock(ErtsRunQueue *rq1, ErtsRunQueue *rq2);
 
 ERTS_GLB_INLINE ErtsMessage *erts_alloc_message_heap_state(Process *pp,
 							   erts_aint32_t *psp,
@@ -2293,11 +2227,7 @@ ErtsSchedulerData *erts_proc_sched_data(Process *c_p)
 {
     ErtsSchedulerData *esdp;
     ASSERT(c_p);
-#if !defined(ERTS_SMP)
-    esdp = erts_get_scheduler_data();
-#else
     esdp = c_p->scheduler_data;
-#  if defined(ERTS_DIRTY_SCHEDULERS)
     if (esdp) {
 	ASSERT(esdp == erts_get_scheduler_data());
 	ASSERT(!ERTS_SCHEDULER_IS_DIRTY(esdp));
@@ -2307,8 +2237,6 @@ ErtsSchedulerData *erts_proc_sched_data(Process *c_p)
 	ASSERT(esdp);
 	ASSERT(ERTS_SCHEDULER_IS_DIRTY(esdp));
     }
-#  endif
-#endif
     ASSERT(esdp);
     return esdp;
 }
@@ -2339,124 +2267,229 @@ Eterm erts_get_current_pid(void)
 ERTS_GLB_INLINE
 Uint erts_get_scheduler_id(void)
 {
-#ifdef ERTS_SMP
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
-#ifdef ERTS_DIRTY_SCHEDULERS
     if (esdp && ERTS_SCHEDULER_IS_DIRTY(esdp))
 	return 0;
     else
-#endif
 	return esdp ? esdp->no : (Uint) 0;
-#else
-    return erts_get_scheduler_data() ? (Uint) 1 : (Uint) 0;
-#endif
 }
 
+/**
+ * Init run-queue of process.
+ *
+ * @param p[in,out]     Process
+ * @param rq[in]        Run-queue that process will be assigned to
+ * @param bnd[in,out]   If non-zero binds process to run-queue.
+ */
+
+ERTS_GLB_INLINE void
+erts_init_runq_proc(Process *p, ErtsRunQueue *rq, int bnd)
+{
+    erts_aint_t rqint = (erts_aint_t) rq;
+    if (bnd)
+        rqint |= ERTS_RUNQ_BOUND_FLAG;
+    erts_atomic_init_nob(&p->run_queue, rqint);
+}
+
+/**
+ * Forcibly set run-queue of process.
+ *
+ * @param p[in,out]     Process
+ * @param rq[in]        Run-queue that process will be assigned to
+ * @param bndp[in,out]  Pointer to integer. On input non-zero
+ *                      value causes the process to be bound to
+ *                      the run-queue. On output, indicating
+ *                      wether process previously was bound or
+ *                      not.
+ * @return              Previous run-queue.
+ */
+
 ERTS_GLB_INLINE ErtsRunQueue *
-erts_get_runq_proc(Process *p)
+erts_set_runq_proc(Process *p, ErtsRunQueue *rq, int *bndp)
 {
-#ifdef ERTS_SMP
-    ASSERT(ERTS_AINT_NULL != erts_atomic_read_nob(&p->run_queue));
-    return (ErtsRunQueue *) erts_atomic_read_nob(&p->run_queue);
-#else
-    return ERTS_RUNQ_IX(0);
-#endif
+    erts_aint_t rqint = (erts_aint_t) rq;
+    ASSERT(bndp);
+    ASSERT(rq);
+    if (*bndp)
+        rqint |= ERTS_RUNQ_BOUND_FLAG;
+    rqint = erts_atomic_xchg_nob(&p->run_queue, rqint);
+    *bndp = (int) (rqint & ERTS_RUNQ_BOUND_FLAG);
+    return (ErtsRunQueue *) (rqint & ERTS_RUNQ_POINTER_MASK);
+}
+
+/**
+ * Try to change run-queue assignment of a process.
+ *
+ * @param p[in,out]     Process
+ * @param rq[int]       Run-queue that process will be assigned to
+ * @return              Non-zero if the run-queue assignment was
+ *                      successfully changed.
+ */
+
+ERTS_GLB_INLINE int
+erts_try_change_runq_proc(Process *p, ErtsRunQueue *rq)
+{
+    erts_aint_t old_rqint, new_rqint;
+
+    ASSERT(rq);
+
+    new_rqint = (erts_aint_t) rq;
+    old_rqint = (erts_aint_t) erts_atomic_read_nob(&p->run_queue);
+    while (1) {
+        erts_aint_t act_rqint;
+
+        if (old_rqint & ERTS_RUNQ_BOUND_FLAG)
+            return 0;
+
+        act_rqint = erts_atomic_cmpxchg_nob(&p->run_queue,
+                                            new_rqint,
+                                            old_rqint);
+        if (act_rqint == old_rqint)
+            return !0;
+    }
+}
+
+/**
+ *
+ * Bind or unbind process to/from currently used run-queue.
+ *
+ * @param p             Process
+ * @param bind          Bind if non-zero; otherwise unbind
+ * @return              Pointer to previously bound run-queue,
+ *                      or NULL if previously unbound
+ */
+
+ERTS_GLB_INLINE ErtsRunQueue *
+erts_bind_runq_proc(Process *p, int bind)
+{
+    erts_aint_t rqint;
+    if (bind)
+        rqint = erts_atomic_read_bor_nob(&p->run_queue,
+                                         ERTS_RUNQ_BOUND_FLAG);
+    else
+        rqint = erts_atomic_read_band_nob(&p->run_queue,
+                                          ~ERTS_RUNQ_BOUND_FLAG);
+    if (rqint & ERTS_RUNQ_BOUND_FLAG)
+        return (ErtsRunQueue *) (rqint & ERTS_RUNQ_POINTER_MASK);
+    else
+        return NULL;
+}
+
+/**
+ * Determine wether a process is bound to a run-queue or not.
+ *
+ * @return              Returns a non-zero value if bound,
+ *                      and zero of not bound.
+ */
+
+ERTS_GLB_INLINE int
+erts_proc_runq_is_bound(Process *p)
+{
+    erts_aint_t rqint = erts_atomic_read_nob(&p->run_queue);
+    return (int) (rqint & ERTS_RUNQ_BOUND_FLAG);
+}
+
+/**
+ * Set run-queue of process.
+ *
+ * @param p[in,out]     Process
+ * @param bndp[out]     Pointer to integer. If non-NULL pointer,
+ *                      the integer will be set to a non-zero
+ *                      value if the process is bound to the
+ *                      run-queue.
+ * @return              Pointer to the normal run-queue that
+ *                      the process currently is assigend to.
+ *                      A process is always assigned to a
+ *                      normal run-queue.
+ */
+
+ERTS_GLB_INLINE ErtsRunQueue *
+erts_get_runq_proc(Process *p, int *bndp)
+{
+    erts_aint_t rqint = erts_atomic_read_nob(&p->run_queue);
+    ErtsRunQueue *rq;
+    if (bndp)
+        *bndp = (int) (rqint & ERTS_RUNQ_BOUND_FLAG);
+    rqint &= ERTS_RUNQ_POINTER_MASK;
+    rq = (ErtsRunQueue *) rqint;
+    ASSERT(rq);
+    return rq;
 }
 
 ERTS_GLB_INLINE ErtsRunQueue *
 erts_get_runq_current(ErtsSchedulerData *esdp)
 {
     ASSERT(!esdp || esdp == erts_get_scheduler_data());
-#ifdef ERTS_SMP
     if (!esdp)
 	esdp = erts_get_scheduler_data();
     return esdp->run_queue;
-#else
-    return ERTS_RUNQ_IX(0);
-#endif
 }
 
 ERTS_GLB_INLINE void
-erts_smp_runq_lock(ErtsRunQueue *rq)
+erts_runq_lock(ErtsRunQueue *rq)
 {
-#ifdef ERTS_SMP
-    erts_smp_mtx_lock(&rq->mtx);
-#endif
+    erts_mtx_lock(&rq->mtx);
 }
 
 ERTS_GLB_INLINE int
-erts_smp_runq_trylock(ErtsRunQueue *rq)
+erts_runq_trylock(ErtsRunQueue *rq)
 {
-#ifdef ERTS_SMP
-    return erts_smp_mtx_trylock(&rq->mtx);
-#else
-    return 0;
-#endif
+    return erts_mtx_trylock(&rq->mtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_runq_unlock(ErtsRunQueue *rq)
+erts_runq_unlock(ErtsRunQueue *rq)
 {
-#ifdef ERTS_SMP
-    erts_smp_mtx_unlock(&rq->mtx);
-#endif
+    erts_mtx_unlock(&rq->mtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_xrunq_lock(ErtsRunQueue *rq, ErtsRunQueue *xrq)
+erts_xrunq_lock(ErtsRunQueue *rq, ErtsRunQueue *xrq)
 {
-#ifdef ERTS_SMP
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(&rq->mtx));
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&rq->mtx));
     if (xrq != rq) {
-	if (erts_smp_mtx_trylock(&xrq->mtx) == EBUSY) {
+	if (erts_mtx_trylock(&xrq->mtx) == EBUSY) {
 	    if (rq < xrq)
-		erts_smp_mtx_lock(&xrq->mtx);
+		erts_mtx_lock(&xrq->mtx);
 	    else {
-		erts_smp_mtx_unlock(&rq->mtx);
-		erts_smp_mtx_lock(&xrq->mtx);
-		erts_smp_mtx_lock(&rq->mtx);
+		erts_mtx_unlock(&rq->mtx);
+		erts_mtx_lock(&xrq->mtx);
+		erts_mtx_lock(&rq->mtx);
 	    }
 	}
     }
-#endif
 }
 
 ERTS_GLB_INLINE void
-erts_smp_xrunq_unlock(ErtsRunQueue *rq, ErtsRunQueue *xrq)
+erts_xrunq_unlock(ErtsRunQueue *rq, ErtsRunQueue *xrq)
 {
-#ifdef ERTS_SMP
     if (xrq != rq)
-	erts_smp_mtx_unlock(&xrq->mtx);
-#endif
+	erts_mtx_unlock(&xrq->mtx);
 }
 
 ERTS_GLB_INLINE void
-erts_smp_runqs_lock(ErtsRunQueue *rq1, ErtsRunQueue *rq2)
+erts_runqs_lock(ErtsRunQueue *rq1, ErtsRunQueue *rq2)
 {
-#ifdef ERTS_SMP
     ASSERT(rq1 && rq2);
     if (rq1 == rq2)
-	erts_smp_mtx_lock(&rq1->mtx);
+	erts_mtx_lock(&rq1->mtx);
     else if (rq1 < rq2) {
-	erts_smp_mtx_lock(&rq1->mtx);
-	erts_smp_mtx_lock(&rq2->mtx);	
+	erts_mtx_lock(&rq1->mtx);
+	erts_mtx_lock(&rq2->mtx);
     }
     else {
-	erts_smp_mtx_lock(&rq2->mtx);
-	erts_smp_mtx_lock(&rq1->mtx);	
+	erts_mtx_lock(&rq2->mtx);
+	erts_mtx_lock(&rq1->mtx);
     }
-#endif
 }
 
 ERTS_GLB_INLINE void
-erts_smp_runqs_unlock(ErtsRunQueue *rq1, ErtsRunQueue *rq2)
+erts_runqs_unlock(ErtsRunQueue *rq1, ErtsRunQueue *rq2)
 {
-#ifdef ERTS_SMP
     ASSERT(rq1 && rq2);
-    erts_smp_mtx_unlock(&rq1->mtx);
+    erts_mtx_unlock(&rq1->mtx);
     if (rq1 != rq2)
-	erts_smp_mtx_unlock(&rq2->mtx);
-#endif
+	erts_mtx_unlock(&rq2->mtx);
 }
 
 ERTS_GLB_INLINE ErtsMessage *
@@ -2488,7 +2521,7 @@ erts_alloc_message_heap(Process *pp,
 			Eterm **hpp,
 			ErlOffHeap **ohpp)
 {
-    erts_aint32_t state = pp ? erts_smp_atomic32_read_nob(&pp->state) : 0;
+    erts_aint32_t state = pp ? erts_atomic32_read_nob(&pp->state) : 0;
     return erts_alloc_message_heap_state(pp, &state, plp, sz, hpp, ohpp);
 }
 
@@ -2502,7 +2535,7 @@ erts_shrink_message_heap(ErtsMessage **msgpp, Process *pp,
 	*msgpp = erts_shrink_message(*msgpp, used_hp - start_hp,
 				     brefs, brefs_size);
     else if (!(*msgpp)->data.attached) {
-	ERTS_SMP_LC_ASSERT(ERTS_PROC_LOCK_MAIN
+	ERTS_LC_ASSERT(ERTS_PROC_LOCK_MAIN
 			   & erts_proc_lc_my_proc_locks(pp));
 	HRelease(pp, end_hp, used_hp);
     }
@@ -2564,50 +2597,34 @@ ERTS_TIME2REDS_IMPL__(ErtsMonotonicTime start, ErtsMonotonicTime end)
 }
 #endif
 
-#ifdef ERTS_SMP
-
-Process *erts_pid2proc_not_running(Process *,
-				   ErtsProcLocks,
-				   Eterm,
-				   ErtsProcLocks);
-Process *erts_pid2proc_nropt(Process *c_p,
-			     ErtsProcLocks c_p_locks,
-			     Eterm pid,
-			     ErtsProcLocks pid_locks);
-extern int erts_disable_proc_not_running_opt;
+Process *erts_try_lock_sig_free_proc(Eterm pid,
+                                     ErtsProcLocks locks,
+                                     erts_aint32_t *statep);
 
 #ifdef DEBUG
-#define ERTS_SMP_ASSERT_IS_NOT_EXITING(P) \
+#define ERTS_ASSERT_IS_NOT_EXITING(P) \
     do { ASSERT(!ERTS_PROC_IS_EXITING((P))); } while (0)
 #else
-#define ERTS_SMP_ASSERT_IS_NOT_EXITING(P)
+#define ERTS_ASSERT_IS_NOT_EXITING(P)
 #endif
 
-#else /* !ERTS_SMP */
-
-#define ERTS_SMP_ASSERT_IS_NOT_EXITING(P)
-
-#define erts_pid2proc_not_running erts_pid2proc
-#define erts_pid2proc_nropt erts_pid2proc
-
-#endif
 
 #define ERTS_PROC_IS_EXITING(P) \
-    (ERTS_PSFLG_EXITING & erts_smp_atomic32_read_acqb(&(P)->state))
+    (ERTS_PSFLG_EXITING & erts_atomic32_read_acqb(&(P)->state))
 
 
 /* Minimum NUMBER of processes for a small system to start */
 #define ERTS_MIN_PROCESSES		1024
-#if defined(ERTS_SMP) && ERTS_MIN_PROCESSES < ERTS_NO_OF_PIX_LOCKS
+#if ERTS_MIN_PROCESSES < ERTS_NO_OF_PIX_LOCKS
 #undef ERTS_MIN_PROCESSES
 #define ERTS_MIN_PROCESSES		ERTS_NO_OF_PIX_LOCKS
 #endif
 
-void erts_smp_notify_inc_runq(ErtsRunQueue *runq);
+void erts_notify_inc_runq(ErtsRunQueue *runq);
 
-#ifdef ERTS_SMP
 void erts_sched_finish_poke(ErtsSchedulerSleepInfo *, erts_aint32_t);
 ERTS_GLB_INLINE void erts_sched_poke(ErtsSchedulerSleepInfo *ssi);
+void erts_aux_thread_poke(void);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
@@ -2616,9 +2633,9 @@ erts_sched_poke(ErtsSchedulerSleepInfo *ssi)
 {
     erts_aint32_t flags;
     ERTS_THR_MEMORY_BARRIER;
-    flags = erts_smp_atomic32_read_nob(&ssi->flags);
+    flags = erts_atomic32_read_nob(&ssi->flags);
     if (flags & ERTS_SSI_FLG_SLEEPING) {
-	flags = erts_smp_atomic32_read_band_nob(&ssi->flags, ~ERTS_SSI_FLGS_SLEEP);
+	flags = erts_atomic32_read_band_nob(&ssi->flags, ~ERTS_SSI_FLGS_SLEEP);
 	erts_sched_finish_poke(ssi, flags);
     }
 }
@@ -2626,7 +2643,6 @@ erts_sched_poke(ErtsSchedulerSleepInfo *ssi)
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#endif /* #ifdef ERTS_SMP */
 
 #include "erl_process_lock.h"
 
@@ -2636,5 +2652,5 @@ erts_sched_poke(ErtsSchedulerSleepInfo *ssi)
 
 
 void erts_halt(int code);
-extern erts_smp_atomic32_t erts_halt_progress;
+extern erts_atomic32_t erts_halt_progress;
 extern int erts_halt_code;
diff --git a/erts/emulator/beam/erl_process_dict.c b/erts/emulator/beam/erl_process_dict.c
index 3c80f0e0f6..38be3938cd 100644
--- a/erts/emulator/beam/erl_process_dict.c
+++ b/erts/emulator/beam/erl_process_dict.c
@@ -79,6 +79,8 @@
 /* Array access macro */ 
 #define ARRAY_GET(PDict, Index) (ASSERT((Index) < (PDict)->arraySize), \
 				 (PDict)->data[Index])
+#define ARRAY_GET_PTR(PDict, Index) (ASSERT((Index) < (PDict)->arraySize), \
+				     &(PDict)->data[Index])
 #define ARRAY_PUT(PDict, Index, Val) (ASSERT((Index) < (PDict)->arraySize), \
                                       (PDict)->data[Index] = (Val))
 
@@ -92,7 +94,7 @@ static void pd_hash_erase_all(Process *p);
 static Eterm pd_hash_get_with_hval(Process *p, Eterm bucket, Eterm id);
 static Eterm pd_hash_get_keys(Process *p, Eterm value);
 static Eterm pd_hash_get_all_keys(Process *p, ProcDict *pd);
-static Eterm pd_hash_get_all(Process *p, ProcDict *pd);
+static Eterm pd_hash_get_all(Process *p, ProcDict *pd, int keep_dict);
 static Eterm pd_hash_put(Process *p, Eterm id, Eterm value);
 
 static void shrink(Process *p, Eterm* ret); 
@@ -237,39 +239,70 @@ erts_erase_dicts(Process *p)
 /* 
  * Called from process_info/1,2.
  */
-Eterm erts_dictionary_copy(Process *p, ProcDict *pd) 
+Eterm erts_dictionary_copy(ErtsHeapFactory *hfact, ProcDict *pd, Uint reserve_size) 
 {
-    Eterm* hp;
-    Eterm* heap_start;
-    Eterm res = NIL;
-    Eterm tmp, tmp2;
+    Eterm res;
     unsigned int i, num;
+    Uint *sz;
+    Uint szi, rsz = reserve_size;
 
-    if (pd == NULL) {
-	return res;
-    }
+    if (pd == NULL)
+	return NIL;
 
     PD_CHECK(pd);
     num = HASH_RANGE(pd);
-    heap_start = hp = (Eterm *) erts_alloc(ERTS_ALC_T_TMP,
-					   sizeof(Eterm) * pd->numElements * 2);
-    for (i = 0; i < num; ++i) {
-	tmp = ARRAY_GET(pd, i);
+    sz = (Uint *) erts_alloc(ERTS_ALC_T_TMP, sizeof(Uint) * pd->numElements);
+
+    for (i = 0, szi = 0; i < num; ++i) {
+	Eterm tmp = ARRAY_GET(pd, i);
 	if (is_boxed(tmp)) {
+            Uint size;
 	    ASSERT(is_tuple(tmp));
-	    res = CONS(hp, tmp, res);
-	    hp += 2;
-	} else if (is_list(tmp)) {
+            size = size_object(tmp) + 2;
+            sz[szi++] = size;
+            rsz += size;
+	}
+        else if (is_list(tmp)) {
 	    while (tmp != NIL) {
-		tmp2 = TCAR(tmp);
-		res = CONS(hp, tmp2, res);
-		hp += 2;
+                Uint size = size_object(TCAR(tmp)) + 2;
+                sz[szi++] = size;
+                rsz += size;
+    		tmp = TCDR(tmp);
+	    }
+	}
+    }
+
+    res = NIL;
+
+    for (i = 0, szi = 0; i < num; ++i) {
+	Eterm tmp = ARRAY_GET(pd, i);
+	if (is_boxed(tmp)) {
+            Uint size;
+            Eterm el, *hp;
+	    ASSERT(is_tuple(tmp));
+            size = sz[szi++];
+            rsz -= size;
+            hp = erts_produce_heap(hfact, size, rsz);
+            el = copy_struct(tmp, size-2, &hp, hfact->off_heap);
+	    res = CONS(hp, el, res);
+	}
+        else if (is_list(tmp)) {
+	    while (tmp != NIL) {
+                Uint size = sz[szi++];
+                Eterm el, *hp;
+                rsz -= size;
+                hp = erts_produce_heap(hfact, size, rsz);
+                el = copy_struct(TCAR(tmp), size-2, &hp, hfact->off_heap);
+		res = CONS(hp, el, res);
 		tmp = TCDR(tmp);
 	    }
 	}
     }
-    res = copy_object(res, p);
-    erts_free(ERTS_ALC_T_TMP, (void *) heap_start);
+
+    ASSERT(rsz == reserve_size);
+
+    erts_free(ERTS_ALC_T_TMP, sz);
+
     return res;
 }
 
@@ -281,7 +314,7 @@ BIF_RETTYPE get_0(BIF_ALIST_0)
 {
     Eterm ret;
     PD_CHECK(BIF_P->dictionary);
-    ret = pd_hash_get_all(BIF_P, BIF_P->dictionary);
+    ret = pd_hash_get_all(BIF_P, BIF_P->dictionary, 1);
     PD_CHECK(BIF_P->dictionary);
     BIF_RET(ret);
 }
@@ -329,7 +362,7 @@ BIF_RETTYPE erase_0(BIF_ALIST_0)
 {
     Eterm ret;
     PD_CHECK(BIF_P->dictionary);
-    ret = pd_hash_get_all(BIF_P, BIF_P->dictionary);
+    ret = pd_hash_get_all(BIF_P, BIF_P->dictionary, 0);
     pd_hash_erase_all(BIF_P);
     PD_CHECK(BIF_P->dictionary);
     BIF_RET(ret);
@@ -541,29 +574,46 @@ static Eterm pd_hash_get_keys(Process *p, Eterm value)
 	
 
 static Eterm
-pd_hash_get_all(Process *p, ProcDict *pd)
+pd_hash_get_all(Process *p, ProcDict *pd, int keep_dict)
 {
     Eterm* hp;
+    Eterm* tp;
     Eterm res = NIL;
     Eterm tmp, tmp2;
     unsigned int i;
     unsigned int num;
+    Uint need;
 
     if (pd == NULL) {
 	return res;
     }
     num = HASH_RANGE(pd);
-    hp = HAlloc(p, pd->numElements * 2);
-    
+
+    /*
+     * If this is not erase/0, then must copy all key-value tuples
+     * as they may be mutated by put/2.
+     */
+    need = pd->numElements * (keep_dict ? 2+3 : 2);
+    hp = HAlloc(p, need);
+
     for (i = 0; i < num; ++i) {
 	tmp = ARRAY_GET(pd, i);
 	if (is_boxed(tmp)) {
-	    ASSERT(is_tuple(tmp));
+            if (keep_dict) {
+                tp = tuple_val(tmp);
+                tmp = TUPLE2(hp, tp[1], tp[2]);
+                hp += 3;
+            }
 	    res = CONS(hp, tmp, res);
 	    hp += 2;
 	} else if (is_list(tmp)) {
 	    while (tmp != NIL) {
 		tmp2 = TCAR(tmp);
+                if (keep_dict) {
+                    tp = tuple_val(tmp2);
+                    tmp2 = TUPLE2(hp, tp[1], tp[2]);
+                    hp += 3;
+                }
 		res = CONS(hp, tmp2, res);
 		hp += 2;
 		tmp = TCDR(tmp);
@@ -577,11 +627,14 @@ static Eterm pd_hash_put(Process *p, Eterm id, Eterm value)
 { 
     unsigned int hval;
     Eterm *hp;
+    Eterm *tp;
+    Eterm *bucket;
     Eterm tpl;
     Eterm old;
+    Eterm old_val = am_undefined;
     Eterm tmp;
     int needed;
-    int i = 0;
+    int new_key = 1;
 #ifdef DEBUG
     Eterm *hp_limit;
 #endif
@@ -595,7 +648,8 @@ static Eterm pd_hash_put(Process *p, Eterm id, Eterm value)
         p->dictionary->numElements = 0;
     }	
     hval = pd_hash_value(p->dictionary, id);
-    old = ARRAY_GET(p->dictionary, hval);
+    bucket = ARRAY_GET_PTR(p->dictionary, hval);
+    old = *bucket;
 
     /*
      * Calculate the number of heap words needed and garbage
@@ -603,32 +657,49 @@ static Eterm pd_hash_put(Process *p, Eterm id, Eterm value)
      */
     needed = 3;			/* {Key,Value} tuple */
     if (is_boxed(old)) {
-	/*
-	 * We don't want to compare keys twice, so we'll always
-	 * reserve the space for two CONS cells.
-	 */
-	needed += 2+2;
+        ASSERT(is_tuple(old));
+        tp = tuple_val(old);
+        if (EQ(tp[1], id)) {
+            old_val = tp[2];
+            if (is_immed(value)) {
+                tp[2] = value;     /* DESTRUCTIVE HEAP ASSIGNMENT */
+                return old_val;
+            }
+            new_key = 0;
+        }
+        else {
+            needed += 2+2;
+        }
     } else if (is_list(old)) {
-	i = 0;
-	for (tmp = old; tmp != NIL && !EQ(tuple_val(TCAR(tmp))[1], id); tmp = TCDR(tmp)) {
-	    ++i;
-	}
-	if (is_nil(tmp)) {
-	    i = -1;
-	    needed += 2;
-	} else {
-	    needed += 2*(i+1);
+        Eterm* prev_cdr = bucket;
+
+        needed += 2;
+	for (tmp = old; tmp != NIL; prev_cdr = &TCDR(tmp), tmp = *prev_cdr) {
+            tp = tuple_val(TCAR(tmp));
+            if (EQ(tp[1], id)) {
+                old_val = tp[2];
+                if (is_immed(value)) {
+                    tp[2] = value;     /* DESTRUCTIVE HEAP ASSIGNMENT */
+                    return old_val;
+                }
+                new_key = 0;
+                /* Unlink old {Key,Value} from list */
+                *prev_cdr = TCDR(tmp);  /* maybe DESTRUCTIVE HEAP ASSIGNMENT */
+                break;
+            }
 	}
     }
     if (HeapWordsLeft(p) < needed) {
 	Eterm root[3];
 	root[0] = id;
 	root[1] = value;
-	root[2] = old;
+        root[2] = old_val;
 	erts_garbage_collect(p, needed, root, 3);
 	id = root[0];
 	value = root[1];
-	old = root[2];
+        old_val = root[2];
+        ASSERT(bucket == ARRAY_GET_PTR(p->dictionary, hval));
+        old = *bucket;
     }
 #ifdef DEBUG
     hp_limit = p->htop + needed;
@@ -644,66 +715,29 @@ static Eterm pd_hash_put(Process *p, Eterm id, Eterm value)
      * Update the dictionary.
      */
     if (is_nil(old)) {
-	ARRAY_PUT(p->dictionary, hval, tpl);
-	++(p->dictionary->numElements);
+	*bucket = tpl;
     } else if (is_boxed(old)) {
 	ASSERT(is_tuple(old));
-	if (EQ(tuple_val(old)[1],id)) {
-	    ARRAY_PUT(p->dictionary, hval, tpl);
-	    return tuple_val(old)[2];
+	if (!new_key) {
+            ASSERT(EQ(tuple_val(old)[1],id));
+	    *bucket = tpl;
+	    return old_val;
 	} else {
 	    hp = HeapOnlyAlloc(p, 4);
 	    tmp = CONS(hp, old, NIL);
 	    hp += 2;
-	    ++(p->dictionary->numElements);
-	    ARRAY_PUT(p->dictionary, hval, CONS(hp, tpl, tmp));
+	    *bucket = CONS(hp, tpl, tmp);
 	    hp += 2;
 	    ASSERT(hp <= hp_limit);
 	}
     } else if (is_list(old)) {
-	if (i == -1) {
-	    /*
-	     * New key. Simply prepend the tuple to the beginning of the list.
-	     */
-	    hp = HeapOnlyAlloc(p, 2);
-	    ARRAY_PUT(p->dictionary, hval, CONS(hp, tpl, old));
-	    hp += 2;
-	    ASSERT(hp <= hp_limit);
-	    ++(p->dictionary->numElements);
-	} else {
-	    /*
-	     * i = Number of CDRs to skip to reach the changed element in the list.
-	     *
-	     * Replace old value in list. To avoid pointers from the old generation
-	     * to the new, we must rebuild the list from the beginning up to and
-	     * including the changed element.
-	     */
-	    Eterm nlist;
-	    int j;
-
-	    hp = HeapOnlyAlloc(p, (i+1)*2);
-	    
-	    /* Find the list element to change. */
-	    for (j = 0, nlist = old; j < i; j++, nlist = TCDR(nlist)) {
-		;
-	    }
-	    ASSERT(EQ(tuple_val(TCAR(nlist))[1], id));
-	    nlist = TCDR(nlist); /* Unchanged part of list. */
-
-	    /* Rebuild list before the updated element. */
-	    for (tmp = old; i-- > 0; tmp = TCDR(tmp)) {
-		nlist = CONS(hp, TCAR(tmp), nlist);
-		hp += 2;
-	    }
-	    ASSERT(EQ(tuple_val(TCAR(tmp))[1], id));
-
-	    /* Put the updated element first in the new list. */
-	    nlist = CONS(hp, tpl, nlist);
-	    hp += 2;
-	    ASSERT(hp <= hp_limit);
-	    ARRAY_PUT(p->dictionary, hval, nlist);
-	    return tuple_val(TCAR(tmp))[2];
-	}
+        /*
+	 * Simply prepend the tuple to the beginning of the list.
+	 */
+	hp = HeapOnlyAlloc(p, 2);
+        *bucket = CONS(hp, tpl, *bucket);
+	hp += 2;
+	ASSERT(hp <= hp_limit);
     } else {
 #ifdef DEBUG
 	erts_fprintf(stderr,
@@ -714,10 +748,13 @@ static Eterm pd_hash_put(Process *p, Eterm id, Eterm value)
 
 	erts_exit(ERTS_ERROR_EXIT, "Damaged process dictionary found during put/2.");
     }
+
+    p->dictionary->numElements += new_key;
+
     if (HASH_RANGE(p->dictionary) <= p->dictionary->numElements) {
 	grow(p);
     }
-    return am_undefined;
+    return old_val;
 }
 
 /*
diff --git a/erts/emulator/beam/erl_process_dict.h b/erts/emulator/beam/erl_process_dict.h
index ab58f3c239..b89b387f5a 100644
--- a/erts/emulator/beam/erl_process_dict.h
+++ b/erts/emulator/beam/erl_process_dict.h
@@ -40,7 +40,7 @@ void erts_erase_dicts(struct process *p);
 void erts_dictionary_dump(fmtfn_t to, void *to_arg, ProcDict *pd);
 void erts_deep_dictionary_dump(fmtfn_t to, void *to_arg,
 			       ProcDict* pd, void (*cb)(fmtfn_t, void *, Eterm obj));
-Eterm erts_dictionary_copy(struct process *p, ProcDict *pd);
+Eterm erts_dictionary_copy(ErtsHeapFactory *hfact, ProcDict *pd, Uint reserve_size);
 
 Eterm erts_pd_hash_get(struct process *p, Eterm id);
 Uint32 erts_pd_make_hx(Eterm key);
diff --git a/erts/emulator/beam/erl_process_dump.c b/erts/emulator/beam/erl_process_dump.c
index 00fe30dfcb..243db4c734 100644
--- a/erts/emulator/beam/erl_process_dump.c
+++ b/erts/emulator/beam/erl_process_dump.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2003-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2003-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -35,6 +35,7 @@
 #include "erl_map.h"
 #define ERTS_WANT_EXTERNAL_TAGS
 #include "external.h"
+#include "erl_proc_sig_queue.h"
 
 #define PTR_FMT "%bpX"
 #define ETERM_FMT "%beX"
@@ -77,7 +78,7 @@ erts_deep_process_dump(fmtfn_t to, void *to_arg)
     for (i = 0; i < max; i++) {
 	Process *p = erts_pix2proc(i);
 	if (p && p->i != ENULL) {
-	    erts_aint32_t state = erts_smp_atomic32_read_acqb(&p->state);
+	    erts_aint32_t state = erts_atomic32_read_acqb(&p->state);
 	    if (state & ERTS_PSFLG_EXITING)
                 continue;
             if (state & ERTS_PSFLG_GC) {
@@ -96,72 +97,133 @@ erts_deep_process_dump(fmtfn_t to, void *to_arg)
     dump_binaries(to, to_arg, all_binaries);
 }
 
-Uint erts_process_memory(Process *p, int incl_msg_inq) {
-  ErtsMessage *mp;
-  Uint size = 0;
-  struct saved_calls *scb;
-  size += sizeof(Process);
-
-  if (incl_msg_inq)
-      ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p);
+static void
+monitor_size(ErtsMonitor *mon, void *vsize)
+{
+    *((Uint *) vsize) += erts_monitor_size(mon);
+}
 
-  erts_doforall_links(ERTS_P_LINKS(p), &erts_one_link_size, &size);
-  erts_doforall_monitors(ERTS_P_MONITORS(p), &erts_one_mon_size, &size);
-  size += (p->heap_sz + p->mbuf_sz) * sizeof(Eterm);
-  if (p->abandoned_heap)
-      size += (p->hend - p->heap) * sizeof(Eterm);
-  if (p->old_hend && p->old_heap)
-    size += (p->old_hend - p->old_heap) * sizeof(Eterm);
+static void
+link_size(ErtsMonitor *lnk, void *vsize)
+{
+    *((Uint *) vsize) += erts_link_size(lnk);
+}
 
+Uint erts_process_memory(Process *p, int include_sigs_in_transit)
+{
+    Uint size = 0;
+    struct saved_calls *scb;
+
+    size += sizeof(Process);
+
+    erts_link_tree_foreach(ERTS_P_LINKS(p),
+                           link_size, (void *) &size);
+    erts_monitor_tree_foreach(ERTS_P_MONITORS(p),
+                              monitor_size, (void *) &size);
+    erts_monitor_list_foreach(ERTS_P_LT_MONITORS(p),
+                              monitor_size, (void *) &size);
+    size += (p->heap_sz + p->mbuf_sz) * sizeof(Eterm);
+    if (p->abandoned_heap)
+        size += (p->hend - p->heap) * sizeof(Eterm);
+    if (p->old_hend && p->old_heap)
+        size += (p->old_hend - p->old_heap) * sizeof(Eterm);
+
+    if (!include_sigs_in_transit) {
+        /*
+         * Size of message queue!
+         *
+         * Note that this assumes that any part of message
+         * queue located in middle queue have been moved
+         * into the inner queue prior to this call.
+         * process_info() management ensures this is done-
+         */
+        ErtsMessage *mp;
+        for (mp = p->sig_qs.first; mp; mp = mp->next) {
+            ASSERT(ERTS_SIG_IS_MSG((ErtsSignal *) mp));
+            size += sizeof(ErtsMessage);
+            if (mp->data.attached)
+                size += erts_msg_attached_data_size(mp) * sizeof(Eterm);
+        }
+    }
+    else {
+        /*
+         * Size of message queue plus size of all signals
+         * in transit to the process!
+         */
+        erts_proc_lock(p, ERTS_PROC_LOCK_MSGQ);
+        erts_proc_sig_fetch(p);
+        erts_proc_unlock(p, ERTS_PROC_LOCK_MSGQ);
+
+        ERTS_FOREACH_SIG_PRIVQS(
+            p, mp,
+            {
+                size += sizeof(ErtsMessage);
+                if (ERTS_SIG_IS_NON_MSG((ErtsSignal *) mp))
+                    size += erts_proc_sig_signal_size((ErtsSignal *) mp);
+                else if (mp->data.attached)
+                    size += erts_msg_attached_data_size(mp) * sizeof(Eterm);
+            });
+    }
 
-  size += p->msg.len * sizeof(ErtsMessage);
+    if (p->arg_reg != p->def_arg_reg) {
+        size += p->arity * sizeof(p->arg_reg[0]);
+    }
 
-  for (mp = p->msg.first; mp; mp = mp->next)
-    if (mp->data.attached)
-      size += erts_msg_attached_data_size(mp)*sizeof(Eterm);
+    if (erts_atomic_read_nob(&p->psd) != (erts_aint_t) NULL)
+        size += sizeof(ErtsPSD);
 
-  if (p->arg_reg != p->def_arg_reg) {
-    size += p->arity * sizeof(p->arg_reg[0]);
-  }
+    scb = ERTS_PROC_GET_SAVED_CALLS_BUF(p);
+    if (scb) {
+        size += (sizeof(struct saved_calls)
+                 + (scb->len-1) * sizeof(scb->ct[0]));
+    }
 
-  if (erts_smp_atomic_read_nob(&p->psd) != (erts_aint_t) NULL)
-    size += sizeof(ErtsPSD);
+    size += erts_dicts_mem_size(p);
+    return size;
+}
 
-  scb = ERTS_PROC_GET_SAVED_CALLS_BUF(p);
-  if (scb) {
-    size += (sizeof(struct saved_calls)
-	     + (scb->len-1) * sizeof(scb->ct[0]));
-  }
+static ERTS_INLINE void
+dump_msg(fmtfn_t to, void *to_arg, ErtsMessage *mp)
+{
+    if (ERTS_SIG_IS_MSG((ErtsSignal *) mp)) {
+        Eterm mesg = ERL_MESSAGE_TERM(mp);
+        if (is_value(mesg))
+            dump_element(to, to_arg, mesg);
+        else
+            dump_dist_ext(to, to_arg, mp->data.dist_ext);
+        mesg = ERL_MESSAGE_TOKEN(mp);
+        erts_print(to, to_arg, ":");
+        dump_element(to, to_arg, mesg);
+        erts_print(to, to_arg, "\n");
+    }
+}
 
-  size += erts_dicts_mem_size(p);
-  return size;
+static ERTS_INLINE void
+heap_dump_msg(fmtfn_t to, void *to_arg, ErtsMessage *mp)
+{
+    if (ERTS_SIG_IS_MSG((ErtsSignal *) mp)) {
+        Eterm mesg = ERL_MESSAGE_TERM(mp);
+        if (is_value(mesg))
+            heap_dump(to, to_arg, mesg);
+        mesg = ERL_MESSAGE_TOKEN(mp);
+        heap_dump(to, to_arg, mesg);
+    }
 }
 
 static void
 dump_process_info(fmtfn_t to, void *to_arg, Process *p)
 {
     Eterm* sp;
-    ErtsMessage* mp;
     int yreg = -1;
 
     if (ERTS_TRACE_FLAGS(p) & F_SENSITIVE)
         return;
 
-    ERTS_SMP_MSGQ_MV_INQ2PRIVQ(p);
+    erts_proc_sig_fetch(p);
 
-    if (p->msg.first) {
+    if (p->sig_qs.first || p->sig_qs.cont) {
 	erts_print(to, to_arg, "=proc_messages:%T\n", p->common.id);
-	for (mp = p->msg.first; mp != NULL; mp = mp->next) {
-	    Eterm mesg = ERL_MESSAGE_TERM(mp);
-	    if (is_value(mesg))
-		dump_element(to, to_arg, mesg);
-	    else
-		dump_dist_ext(to, to_arg, mp->data.dist_ext);
-	    mesg = ERL_MESSAGE_TOKEN(mp);
-	    erts_print(to, to_arg, ":");
-	    dump_element(to, to_arg, mesg);
-	    erts_print(to, to_arg, "\n");
-	}
+        ERTS_FOREACH_SIG_PRIVQS(p, mp, dump_msg(to, to_arg, mp));
     }
 
     if (p->dictionary) {
@@ -183,13 +245,10 @@ dump_process_info(fmtfn_t to, void *to_arg, Process *p)
             heap_dump(to, to_arg, term);
         }
     }
-    for (mp = p->msg.first; mp != NULL; mp = mp->next) {
-        Eterm mesg = ERL_MESSAGE_TERM(mp);
-        if (is_value(mesg))
-            heap_dump(to, to_arg, mesg);
-        mesg = ERL_MESSAGE_TOKEN(mp);
-        heap_dump(to, to_arg, mesg);
-    }
+
+    if (p->sig_qs.first || p->sig_qs.cont)
+        ERTS_FOREACH_SIG_PRIVQS(p, mp, heap_dump_msg(to, to_arg, mp));
+
     if (p->dictionary) {
         erts_deep_dictionary_dump(to, to_arg, p->dictionary, heap_dump);
     }
@@ -904,6 +963,9 @@ dump_module_literals(fmtfn_t to, void *to_arg, ErtsLiteralArea* lit_area)
                     }
                     erts_putc(to, to_arg, '\n');
                 }
+            } else if (is_export_header(w)) {
+                dump_externally(to, to_arg, term);
+                erts_putc(to, to_arg, '\n');
             }
             size = 1 + header_arity(w);
             switch (w & _HEADER_SUBTAG_MASK) {
@@ -997,8 +1059,8 @@ erts_dump_extended_process_state(fmtfn_t to, void *to_arg, erts_aint32_t psflg)
                 erts_print(to, to_arg, "FREE"); break;
             case ERTS_PSFLG_EXITING:
                 erts_print(to, to_arg, "EXITING"); break;
-            case ERTS_PSFLG_PENDING_EXIT:
-                erts_print(to, to_arg, "PENDING_EXIT"); break;
+            case ERTS_PSFLG_UNUSED:
+                erts_print(to, to_arg, "UNUSED"); break;
             case ERTS_PSFLG_ACTIVE:
                 erts_print(to, to_arg, "ACTIVE"); break;
             case ERTS_PSFLG_IN_RUNQ:
@@ -1009,10 +1071,10 @@ erts_dump_extended_process_state(fmtfn_t to, void *to_arg, erts_aint32_t psflg)
                 erts_print(to, to_arg, "SUSPENDED"); break;
             case ERTS_PSFLG_GC:
                 erts_print(to, to_arg, "GC"); break;
-            case ERTS_PSFLG_BOUND:
-                erts_print(to, to_arg, "BOUND"); break;
-            case ERTS_PSFLG_TRAP_EXIT:
-                erts_print(to, to_arg, "TRAP_EXIT"); break;
+            case ERTS_PSFLG_SYS_TASKS:
+                erts_print(to, to_arg, "SYS_TASKS"); break;
+            case ERTS_PSFLG_SIG_IN_Q:
+                erts_print(to, to_arg, "SIG_IN_Q"); break;
             case ERTS_PSFLG_ACTIVE_SYS:
                 erts_print(to, to_arg, "ACTIVE_SYS"); break;
             case ERTS_PSFLG_RUNNING_SYS:
@@ -1023,8 +1085,8 @@ erts_dump_extended_process_state(fmtfn_t to, void *to_arg, erts_aint32_t psflg)
                 erts_print(to, to_arg, "DELAYED_SYS"); break;
             case ERTS_PSFLG_OFF_HEAP_MSGQ:
                 erts_print(to, to_arg, "OFF_HEAP_MSGQ"); break;
-            case ERTS_PSFLG_ON_HEAP_MSGQ:
-                erts_print(to, to_arg, "ON_HEAP_MSGQ"); break;
+            case ERTS_PSFLG_SIG_Q:
+                erts_print(to, to_arg, "SIG_Q"); break;
             case ERTS_PSFLG_DIRTY_CPU_PROC:
                 erts_print(to, to_arg, "DIRTY_CPU_PROC"); break;
             case ERTS_PSFLG_DIRTY_IO_PROC:
diff --git a/erts/emulator/beam/erl_process_lock.c b/erts/emulator/beam/erl_process_lock.c
index ff124d5ba7..44c7892040 100644
--- a/erts/emulator/beam/erl_process_lock.c
+++ b/erts/emulator/beam/erl_process_lock.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2007-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2007-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -56,9 +56,9 @@
  *     Note that wait flags may be read without the pix lock, but
  *   it is important that wait flags only are modified when the pix
  *   lock is held.
- *     This implementation assumes that erts_smp_atomic_or_retold()
+ *     This implementation assumes that erts_atomic_or_retold()
  *   provides necessary memorybarriers for a lock operation, and that
- *   erts_smp_atomic_and_retold() provides necessary memorybarriers
+ *   erts_atomic_and_retold() provides necessary memorybarriers
  *   for an unlock operation.
  */
 
@@ -69,7 +69,6 @@
 #include "erl_process.h"
 #include "erl_thr_progress.h"
 
-#ifdef ERTS_SMP
 
 #if ERTS_PROC_LOCK_OWN_IMPL
 
@@ -102,7 +101,6 @@ static void cleanup_tse(void);
 #ifdef ERTS_ENABLE_LOCK_CHECK
 static struct {
     Sint16 proc_lock_main;
-    Sint16 proc_lock_link;
     Sint16 proc_lock_msgq;
     Sint16 proc_lock_btm;
     Sint16 proc_lock_status;
@@ -141,7 +139,6 @@ erts_init_proc_lock(int cpus)
 #endif
 #ifdef ERTS_ENABLE_LOCK_CHECK
     lc_id.proc_lock_main	= erts_lc_get_lock_order_id("proc_main");
-    lc_id.proc_lock_link	= erts_lc_get_lock_order_id("proc_link");
     lc_id.proc_lock_msgq	= erts_lc_get_lock_order_id("proc_msgq");
     lc_id.proc_lock_btm		= erts_lc_get_lock_order_id("proc_btm");
     lc_id.proc_lock_status	= erts_lc_get_lock_order_id("proc_status");
@@ -464,7 +461,7 @@ wait_for_locks(Process *p,
 }
 
 /*
- * erts_proc_lock_failed() is called when erts_smp_proc_lock()
+ * erts_proc_lock_failed() is called when erts_proc_lock()
  * wasn't able to lock all locks. We may need to transfer locks
  * to waiters and wait for our turn on locks.
  *
@@ -543,7 +540,7 @@ erts_proc_lock_failed(Process *p,
 }
 
 /*
- * erts_proc_unlock_failed() is called when erts_smp_proc_unlock()
+ * erts_proc_unlock_failed() is called when erts_proc_unlock()
  * wasn't able to unlock all locks. We may need to transfer locks
  * to waiters.
  */
@@ -709,7 +706,7 @@ proc_safelock(int is_managed,
 		refc1 = 1;
 		erts_proc_inc_refc(p1);
 	    }
-	    erts_smp_proc_unlock(p1, unlock_locks);
+	    erts_proc_unlock(p1, unlock_locks);
 	}
 	unlock_locks = unlock_mask & have_locks2;
 	if (unlock_locks) {
@@ -719,7 +716,7 @@ proc_safelock(int is_managed,
 		refc2 = 1;
 		erts_proc_inc_refc(p2);
 	    }
-	    erts_smp_proc_unlock(p2, unlock_locks);
+	    erts_proc_unlock(p2, unlock_locks);
 	}
     }
 
@@ -750,7 +747,7 @@ proc_safelock(int is_managed,
 	    if (need_locks2 & lock)
 		lock_no--;
 	    locks = need_locks1 & lock_mask;
-	    erts_smp_proc_lock(p1, locks);
+	    erts_proc_lock(p1, locks);
 	    have_locks1 |= locks;
 	    need_locks1 &= ~locks;
 	}
@@ -761,7 +758,7 @@ proc_safelock(int is_managed,
 		lock = (1 << ++lock_no);
 	    }
 	    locks = need_locks2 & lock_mask;
-	    erts_smp_proc_lock(p2, locks);
+	    erts_proc_lock(p2, locks);
 	    have_locks2 |= locks;
 	    need_locks2 &= ~locks;
 	}
@@ -898,7 +895,7 @@ erts_pid2proc_opt(Process *c_p,
 #endif /* ERTS_PROC_LOCK_OWN_IMPL */
 	    {
 		/* Try a quick trylock to grab all the locks we need. */
-		busy = (int) erts_smp_proc_raw_trylock__(proc, need_locks);
+		busy = (int) erts_proc_raw_trylock__(proc, need_locks);
 
 #if ERTS_PROC_LOCK_OWN_IMPL && defined(ERTS_ENABLE_LOCK_CHECK)
 		erts_proc_lc_trylock(proc, need_locks, !busy, __FILE__,__LINE__);
@@ -976,7 +973,7 @@ erts_pid2proc_opt(Process *c_p,
 	    : (proc
 	       != (Process *) erts_ptab_pix2intptr_nob(&erts_proc, pix)))) {
 
-	erts_smp_proc_unlock(proc, need_locks);
+	erts_proc_unlock(proc, need_locks);
 
 	if (flags & ERTS_P2P_FLG_INC_REFC)
 	    dec_refc_proc = proc;
@@ -1002,11 +999,9 @@ static ERTS_INLINE
 Process *proc_lookup_inc_refc(Eterm pid, int allow_exit)
 {
     Process *proc;
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl;
 
     dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
     proc = erts_proc_lookup_raw(pid);
     if (proc) {
@@ -1016,9 +1011,7 @@ Process *proc_lookup_inc_refc(Eterm pid, int allow_exit)
             erts_proc_inc_refc(proc);
     }
 
-#ifdef ERTS_SMP
     erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 
     return proc;
 }
@@ -1042,7 +1035,7 @@ erts_proc_lock_init(Process *p)
 #if ERTS_PROC_LOCK_OWN_IMPL
     /* We always start with all locks locked */
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
-    erts_smp_atomic32_init_nob(&p->lock.flags,
+    erts_atomic32_init_nob(&p->lock.flags,
 			       (erts_aint32_t) ERTS_PROC_LOCKS_ALL);
 #else
     p->lock.flags = ERTS_PROC_LOCKS_ALL;
@@ -1060,12 +1053,6 @@ erts_proc_lock_init(Process *p)
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_trylock(1, &p->lock.main.lc);
 #endif
-    erts_mtx_init(&p->lock.link, "proc_link", p->common.id,
-        ERTS_LOCK_FLAGS_CATEGORY_PROCESS);
-    ethr_mutex_lock(&p->lock.link.mtx);
-#ifdef ERTS_ENABLE_LOCK_CHECK
-    erts_lc_trylock(1, &p->lock.link.lc);
-#endif
     erts_mtx_init(&p->lock.msgq, "proc_msgq", p->common.id,
         ERTS_LOCK_FLAGS_CATEGORY_PROCESS);
     ethr_mutex_lock(&p->lock.msgq.mtx);
@@ -1093,7 +1080,7 @@ erts_proc_lock_init(Process *p)
 #endif
 #ifdef ERTS_PROC_LOCK_DEBUG
     for (i = 0; i <= ERTS_PROC_LOCK_MAX_BIT; i++)
-	erts_smp_atomic32_init_nob(&p->lock.locked[i], (erts_aint32_t) 1);
+	erts_atomic32_init_nob(&p->lock.locked[i], (erts_aint32_t) 1);
 #endif
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_proc_lock_init(p);
@@ -1107,13 +1094,12 @@ erts_proc_lock_fin(Process *p)
 {
 #if ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     erts_mtx_destroy(&p->lock.main);
-    erts_mtx_destroy(&p->lock.link);
     erts_mtx_destroy(&p->lock.msgq);
     erts_mtx_destroy(&p->lock.btm);
     erts_mtx_destroy(&p->lock.status);
     erts_mtx_destroy(&p->lock.trace);
 #endif
-#if defined(ERTS_ENABLE_LOCK_COUNT) && defined(ERTS_SMP)
+#if defined(ERTS_ENABLE_LOCK_COUNT)
     erts_lcnt_proc_lock_destroy(p);
 #endif
 }
@@ -1149,8 +1135,6 @@ void erts_lcnt_enable_proc_lock_count(Process *proc, int enable) {
 
         erts_lcnt_init_lock_info_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN,
             "proc_main", proc->common.id, ERTS_LOCK_TYPE_PROCLOCK);
-        erts_lcnt_init_lock_info_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK,
-            "proc_link", proc->common.id, ERTS_LOCK_TYPE_PROCLOCK);
         erts_lcnt_init_lock_info_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ,
             "proc_msgq", proc->common.id, ERTS_LOCK_TYPE_PROCLOCK);
         erts_lcnt_init_lock_info_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_BTM,
@@ -1205,10 +1189,6 @@ erts_proc_lc_lock(Process *p, ErtsProcLocks locks, char *file, unsigned int line
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_lock_x(&lck,file,line);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_lock_x(&lck,file,line);
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_lock_x(&lck,file,line);
@@ -1238,10 +1218,6 @@ erts_proc_lc_trylock(Process *p, ErtsProcLocks locks, int locked,
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_trylock_x(locked, &lck, file, line);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_trylock_x(locked, &lck, file, line);
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_trylock_x(locked, &lck, file, line);
@@ -1282,10 +1258,6 @@ erts_proc_lc_unlock(Process *p, ErtsProcLocks locks)
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_unlock(&lck);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_unlock(&lck);
-    }
     if (locks & ERTS_PROC_LOCK_MAIN) {
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_unlock(&lck);
@@ -1317,10 +1289,6 @@ erts_proc_lc_might_unlock(Process *p, ErtsProcLocks locks)
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_might_unlock(&lck);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_might_unlock(&lck);
-    }
     if (locks & ERTS_PROC_LOCK_MAIN) {
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_might_unlock(&lck);
@@ -1328,8 +1296,6 @@ erts_proc_lc_might_unlock(Process *p, ErtsProcLocks locks)
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_lc_might_unlock(&p->lock.main.lc);
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_lc_might_unlock(&p->lock.link.lc);
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_lc_might_unlock(&p->lock.msgq.lc);
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -1353,10 +1319,6 @@ erts_proc_lc_require_lock(Process *p, ErtsProcLocks locks, char *file,
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_require_lock(&lck, file, line);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_require_lock(&lck, file, line);
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_require_lock(&lck, file, line);
@@ -1376,8 +1338,6 @@ erts_proc_lc_require_lock(Process *p, ErtsProcLocks locks, char *file,
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_lc_require_lock(&p->lock.main.lc, file, line);
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_lc_require_lock(&p->lock.link.lc, file, line);
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_lc_require_lock(&p->lock.msgq.lc, file, line);
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -1412,10 +1372,6 @@ erts_proc_lc_unrequire_lock(Process *p, ErtsProcLocks locks)
 	lck.id = lc_id.proc_lock_msgq;
 	erts_lc_unrequire_lock(&lck);
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	lck.id = lc_id.proc_lock_link;
-	erts_lc_unrequire_lock(&lck);
-    }
     if (locks & ERTS_PROC_LOCK_MAIN) {
 	lck.id = lc_id.proc_lock_main;
 	erts_lc_unrequire_lock(&lck);
@@ -1423,8 +1379,6 @@ erts_proc_lc_unrequire_lock(Process *p, ErtsProcLocks locks)
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_lc_unrequire_lock(&p->lock.main.lc);
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_lc_unrequire_lock(&p->lock.link.lc);
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_lc_unrequire_lock(&p->lock.msgq.lc);
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -1448,8 +1402,6 @@ erts_proc_lc_trylock_force_busy(Process *p, ErtsProcLocks locks)
 
 	if (locks & ERTS_PROC_LOCK_MAIN)
 	    lck.id = lc_id.proc_lock_main;
-	else if (locks & ERTS_PROC_LOCK_LINK)
-	    lck.id = lc_id.proc_lock_link;
 	else if (locks & ERTS_PROC_LOCK_MSGQ)
 	    lck.id = lc_id.proc_lock_msgq;
 	else if (locks & ERTS_PROC_LOCK_BTM)
@@ -1492,10 +1444,6 @@ void erts_proc_lc_chk_only_proc(Process *p, ErtsProcLocks locks)
 	have_locks[have_locks_len].id = lc_id.proc_lock_main;
 	have_locks[have_locks_len++].extra = p->common.id;
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	have_locks[have_locks_len].id = lc_id.proc_lock_link;
-	have_locks[have_locks_len++].extra = p->common.id;
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	have_locks[have_locks_len].id = lc_id.proc_lock_msgq;
 	have_locks[have_locks_len++].extra = p->common.id;
@@ -1516,8 +1464,6 @@ void erts_proc_lc_chk_only_proc(Process *p, ErtsProcLocks locks)
     erts_lc_lock_t have_locks[6];
     if (locks & ERTS_PROC_LOCK_MAIN)
 	have_locks[have_locks_len++] = p->lock.main.lc;
-    if (locks & ERTS_PROC_LOCK_LINK)
-	have_locks[have_locks_len++] = p->lock.link.lc;
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	have_locks[have_locks_len++] = p->lock.msgq.lc;
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -1545,10 +1491,6 @@ erts_proc_lc_chk_have_proc_locks(Process *p, ErtsProcLocks locks)
 	have_locks[have_locks_len].id = lc_id.proc_lock_main;
 	have_locks[have_locks_len++].extra = p->common.id;
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	have_locks[have_locks_len].id = lc_id.proc_lock_link;
-	have_locks[have_locks_len++].extra = p->common.id;
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	have_locks[have_locks_len].id = lc_id.proc_lock_msgq;
 	have_locks[have_locks_len++].extra = p->common.id;
@@ -1569,8 +1511,6 @@ erts_proc_lc_chk_have_proc_locks(Process *p, ErtsProcLocks locks)
     erts_lc_lock_t have_locks[6];
     if (locks & ERTS_PROC_LOCK_MAIN)
 	have_locks[have_locks_len++] = p->lock.main.lc;
-    if (locks & ERTS_PROC_LOCK_LINK)
-	have_locks[have_locks_len++] = p->lock.link.lc;
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	have_locks[have_locks_len++] = p->lock.msgq.lc;
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -1608,14 +1548,6 @@ erts_proc_lc_chk_proc_locks(Process *p, ErtsProcLocks locks)
 	have_not_locks[have_not_locks_len].id = lc_id.proc_lock_main;
 	have_not_locks[have_not_locks_len++].extra = p->common.id;
     }
-    if (locks & ERTS_PROC_LOCK_LINK) {
-	have_locks[have_locks_len].id = lc_id.proc_lock_link;
-	have_locks[have_locks_len++].extra = p->common.id;
-    }
-    else {
-	have_not_locks[have_not_locks_len].id = lc_id.proc_lock_link;
-	have_not_locks[have_not_locks_len++].extra = p->common.id;
-    }
     if (locks & ERTS_PROC_LOCK_MSGQ) {
 	have_locks[have_locks_len].id = lc_id.proc_lock_msgq;
 	have_locks[have_locks_len++].extra = p->common.id;
@@ -1656,10 +1588,6 @@ erts_proc_lc_chk_proc_locks(Process *p, ErtsProcLocks locks)
 	have_locks[have_locks_len++] = p->lock.main.lc;
     else
 	have_not_locks[have_not_locks_len++] = p->lock.main.lc;
-    if (locks & ERTS_PROC_LOCK_LINK)
-	have_locks[have_locks_len++] = p->lock.link.lc;
-    else
-	have_not_locks[have_not_locks_len++] = p->lock.link.lc;
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	have_locks[have_locks_len++] = p->lock.msgq.lc;
     else
@@ -1685,13 +1613,10 @@ erts_proc_lc_chk_proc_locks(Process *p, ErtsProcLocks locks)
 ErtsProcLocks
 erts_proc_lc_my_proc_locks(Process *p)
 {
-    int resv[6];
+    int resv[5];
     ErtsProcLocks res = 0;
 #if ERTS_PROC_LOCK_OWN_IMPL
-    erts_lc_lock_t locks[6] = {ERTS_LC_LOCK_INIT(lc_id.proc_lock_main,
-						 p->common.id,
-						 ERTS_LOCK_TYPE_PROCLOCK),
-			       ERTS_LC_LOCK_INIT(lc_id.proc_lock_link,
+    erts_lc_lock_t locks[5] = {ERTS_LC_LOCK_INIT(lc_id.proc_lock_main,
 						 p->common.id,
 						 ERTS_LOCK_TYPE_PROCLOCK),
 			       ERTS_LC_LOCK_INIT(lc_id.proc_lock_msgq,
@@ -1707,26 +1632,23 @@ erts_proc_lc_my_proc_locks(Process *p)
 						 p->common.id,
 						 ERTS_LOCK_TYPE_PROCLOCK)};
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
-    erts_lc_lock_t locks[6] = {p->lock.main.lc,
-			       p->lock.link.lc,
+    erts_lc_lock_t locks[5] = {p->lock.main.lc,
 			       p->lock.msgq.lc,
 			       p->lock.btm.lc,
 			       p->lock.status.lc,
 			       p->lock.trace.lc};
 #endif
 
-    erts_lc_have_locks(resv, locks, 6);
+    erts_lc_have_locks(resv, locks, 5);
     if (resv[0])
 	res |= ERTS_PROC_LOCK_MAIN;
     if (resv[1])
-	res |= ERTS_PROC_LOCK_LINK;
-    if (resv[2])
 	res |= ERTS_PROC_LOCK_MSGQ;
-    if (resv[3])
+    if (resv[2])
 	res |= ERTS_PROC_LOCK_BTM;
-    if (resv[4])
+    if (resv[3])
 	res |= ERTS_PROC_LOCK_STATUS;
-    if (resv[5])
+    if (resv[4])
 	res |= ERTS_PROC_LOCK_TRACE;
 
     return res;
@@ -1735,15 +1657,14 @@ erts_proc_lc_my_proc_locks(Process *p)
 void
 erts_proc_lc_chk_no_proc_locks(char *file, int line)
 {
-    int resv[6];
-    int ids[6] = {lc_id.proc_lock_main,
-		  lc_id.proc_lock_link,
+    int resv[5];
+    int ids[5] = {lc_id.proc_lock_main,
 		  lc_id.proc_lock_msgq,
 		  lc_id.proc_lock_btm,
 		  lc_id.proc_lock_status,
 		  lc_id.proc_lock_trace};
-    erts_lc_have_lock_ids(resv, ids, 6);
-    if (!ERTS_IS_CRASH_DUMPING && (resv[0] || resv[1] || resv[2] || resv[3] || resv[4] || resv[5])) {
+    erts_lc_have_lock_ids(resv, ids, 5);
+    if (!ERTS_IS_CRASH_DUMPING && (resv[0] || resv[1] || resv[2] || resv[3] || resv[4])) {
 	erts_lc_fail("%s:%d: Thread has process locks locked when expected "
 		     "not to have any process locks locked",
 		     file, line);
@@ -1785,4 +1706,3 @@ check_queue(erts_proc_lock_t *lck)
 }
 #endif
 
-#endif /* ERTS_SMP */
diff --git a/erts/emulator/beam/erl_process_lock.h b/erts/emulator/beam/erl_process_lock.h
index 023ba4d4ae..bd38eca4dc 100644
--- a/erts/emulator/beam/erl_process_lock.h
+++ b/erts/emulator/beam/erl_process_lock.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2007-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2007-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -36,7 +36,7 @@
 #include "erl_lock_count.h"
 #endif
 
-#include "erl_smp.h"
+#include "erl_threads.h"
 
 #if defined(VALGRIND) || defined(ETHR_DISABLE_NATIVE_IMPLS)
 #  define ERTS_PROC_LOCK_OWN_IMPL 0
@@ -66,14 +66,14 @@
 
 #endif
 
-#define ERTS_PROC_LOCK_MAX_BIT 5
+#define ERTS_PROC_LOCK_MAX_BIT 4
 
 typedef erts_aint32_t ErtsProcLocks;
 
 typedef struct erts_proc_lock_t_ {
 #if ERTS_PROC_LOCK_OWN_IMPL
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
-    erts_smp_atomic32_t flags;
+    erts_atomic32_t flags;
 #else
     ErtsProcLocks flags;
 #endif
@@ -82,19 +82,17 @@ typedef struct erts_proc_lock_t_ {
     /* Each erts_mtx_t has its own lock counter ^ */
 
     #define ERTS_LCNT_PROCLOCK_IDX_MAIN 0
-    #define ERTS_LCNT_PROCLOCK_IDX_LINK 1
-    #define ERTS_LCNT_PROCLOCK_IDX_MSGQ 2
-    #define ERTS_LCNT_PROCLOCK_IDX_BTM 3
-    #define ERTS_LCNT_PROCLOCK_IDX_STATUS 4
-    #define ERTS_LCNT_PROCLOCK_IDX_TRACE 5
+    #define ERTS_LCNT_PROCLOCK_IDX_MSGQ 1
+    #define ERTS_LCNT_PROCLOCK_IDX_BTM 2
+    #define ERTS_LCNT_PROCLOCK_IDX_STATUS 3
+    #define ERTS_LCNT_PROCLOCK_IDX_TRACE 4
 
-    #define ERTS_LCNT_PROCLOCK_COUNT 6
+    #define ERTS_LCNT_PROCLOCK_COUNT 5
 
     erts_lcnt_ref_t lcnt_carrier;
 #endif
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     erts_mtx_t main;
-    erts_mtx_t link;
     erts_mtx_t msgq;
     erts_mtx_t btm;
     erts_mtx_t status;
@@ -103,7 +101,7 @@ typedef struct erts_proc_lock_t_ {
 #  error "no implementation"
 #endif
 #ifdef ERTS_PROC_LOCK_DEBUG
-    erts_smp_atomic32_t locked[ERTS_PROC_LOCK_MAX_BIT+1];
+    erts_atomic32_t locked[ERTS_PROC_LOCK_MAX_BIT+1];
 #endif
 } erts_proc_lock_t;
 
@@ -118,27 +116,18 @@ typedef struct erts_proc_lock_t_ {
 #define ERTS_PROC_LOCK_MAIN		(((ErtsProcLocks) 1) << 0)
 
 /*
- * Link lock:
- *   Protects the following fields in the process structure:
- *   * nlinks
- *   * monitors
- *   * suspend_monitors
- */
-#define ERTS_PROC_LOCK_LINK		(((ErtsProcLocks) 1) << 1)
-
-/*
  * Message queue lock:
  *   Protects the following fields in the process structure:
  *   * msg_inq
  */
-#define ERTS_PROC_LOCK_MSGQ		(((ErtsProcLocks) 1) << 2)
+#define ERTS_PROC_LOCK_MSGQ		(((ErtsProcLocks) 1) << 1)
 
 /*
  * Bif timer lock:
  *   Protects the following fields in the process structure:
  *   * bif_timers
  */
-#define ERTS_PROC_LOCK_BTM		(((ErtsProcLocks) 1) << 3)
+#define ERTS_PROC_LOCK_BTM		(((ErtsProcLocks) 1) << 2)
 
 /*
  * Status lock:
@@ -148,7 +137,7 @@ typedef struct erts_proc_lock_t_ {
  *   * sys_tasks
  *   * ...
  */
-#define ERTS_PROC_LOCK_STATUS		(((ErtsProcLocks) 1) << 4)
+#define ERTS_PROC_LOCK_STATUS		(((ErtsProcLocks) 1) << 3)
 
 /*
  * Trace message lock:
@@ -243,11 +232,11 @@ typedef struct erts_proc_lock_t_ {
 /* Lock counter implemetation */
 
 #ifdef ERTS_ENABLE_LOCK_POSITION
-#define erts_smp_proc_lock__(P,I,L) erts_smp_proc_lock_x__(P,I,L,__FILE__,__LINE__)
-#define erts_smp_proc_lock(P,L) erts_smp_proc_lock_x(P,L,__FILE__,__LINE__)
+#define erts_proc_lock__(P,I,L) erts_proc_lock_x__(P,I,L,__FILE__,__LINE__)
+#define erts_proc_lock(P,L) erts_proc_lock_x(P,L,__FILE__,__LINE__)
 #endif
 
-#if defined(ERTS_SMP) && defined (ERTS_ENABLE_LOCK_COUNT)
+#if defined (ERTS_ENABLE_LOCK_COUNT)
 
 void erts_lcnt_proc_lock_init(Process *p);
 void erts_lcnt_proc_lock_destroy(Process *p);
@@ -277,9 +266,6 @@ void erts_lcnt_proc_lock(erts_proc_lock_t *lock, ErtsProcLocks locks) {
         if (locks & ERTS_PROC_LOCK_MAIN) {
             erts_lcnt_lock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN);
         }
-        if (locks & ERTS_PROC_LOCK_LINK) {
-            erts_lcnt_lock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK);
-        }
         if (locks & ERTS_PROC_LOCK_MSGQ) {
             erts_lcnt_lock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ);
         }
@@ -307,9 +293,6 @@ void erts_lcnt_proc_lock_post_x(erts_proc_lock_t *lock, ErtsProcLocks locks,
         if (locks & ERTS_PROC_LOCK_MAIN) {
             erts_lcnt_lock_post_x_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN, file, line);
         }
-        if (locks & ERTS_PROC_LOCK_LINK) {
-            erts_lcnt_lock_post_x_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK, file, line);
-        }
         if (locks & ERTS_PROC_LOCK_MSGQ) {
             erts_lcnt_lock_post_x_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ, file, line);
         }
@@ -336,9 +319,6 @@ void erts_lcnt_proc_lock_unacquire(erts_proc_lock_t *lock, ErtsProcLocks locks)
         if (locks & ERTS_PROC_LOCK_MAIN) {
             erts_lcnt_lock_unacquire_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN);
         }
-        if (locks & ERTS_PROC_LOCK_LINK) {
-            erts_lcnt_lock_unacquire_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK);
-        }
         if (locks & ERTS_PROC_LOCK_MSGQ) {
             erts_lcnt_lock_unacquire_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ);
         }
@@ -365,9 +345,6 @@ void erts_lcnt_proc_unlock(erts_proc_lock_t *lock, ErtsProcLocks locks) {
         if (locks & ERTS_PROC_LOCK_MAIN) {
             erts_lcnt_unlock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN);
         }
-        if (locks & ERTS_PROC_LOCK_LINK) {
-            erts_lcnt_unlock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK);
-        }
         if (locks & ERTS_PROC_LOCK_MSGQ) {
             erts_lcnt_unlock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ);
         }
@@ -394,9 +371,6 @@ void erts_lcnt_proc_trylock(erts_proc_lock_t *lock, ErtsProcLocks locks, int res
         if (locks & ERTS_PROC_LOCK_MAIN) {
             erts_lcnt_trylock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MAIN, res);
         }
-        if (locks & ERTS_PROC_LOCK_LINK) {
-            erts_lcnt_trylock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_LINK, res);
-        }
         if (locks & ERTS_PROC_LOCK_MSGQ) {
             erts_lcnt_trylock_idx(carrier, ERTS_LCNT_PROCLOCK_IDX_MSGQ, res);
         }
@@ -421,10 +395,10 @@ void erts_lcnt_proc_trylock(erts_proc_lock_t *lock, ErtsProcLocks locks, int res
 
 /* --- Process lock checking ----------------------------------------------- */
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-#define ERTS_SMP_CHK_NO_PROC_LOCKS \
+#if defined(ERTS_ENABLE_LOCK_CHECK)
+#define ERTS_CHK_NO_PROC_LOCKS \
   erts_proc_lc_chk_no_proc_locks(__FILE__, __LINE__)
-#define ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(P) \
+#define ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(P) \
   erts_proc_lc_chk_only_proc_main((P))
 void erts_proc_lc_lock(Process *p, ErtsProcLocks locks,
 		       char *file, unsigned int line);
@@ -443,8 +417,8 @@ void erts_proc_lc_require_lock(Process *p, ErtsProcLocks locks,
 			       char* file, unsigned int line);
 void erts_proc_lc_unrequire_lock(Process *p, ErtsProcLocks locks);
 #else
-#define ERTS_SMP_CHK_NO_PROC_LOCKS
-#define ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(P)
+#define ERTS_CHK_NO_PROC_LOCKS
+#define ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(P)
 #endif
 
 #endif /* #ifndef ERTS_PROC_LOCK_LOCK_CHECK__ */
@@ -455,7 +429,6 @@ void erts_proc_lc_unrequire_lock(Process *p, ErtsProcLocks locks);
 #ifndef ERTS_PROCESS_LOCK_H__
 #define ERTS_PROCESS_LOCK_H__
 
-#ifdef ERTS_SMP
 
 typedef struct {
     union {
@@ -472,21 +445,21 @@ typedef struct {
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
 
 #define ERTS_PROC_LOCK_FLGS_BAND_(L, MSK) \
-  ((ErtsProcLocks) erts_smp_atomic32_read_band_nob(&(L)->flags, \
+  ((ErtsProcLocks) erts_atomic32_read_band_nob(&(L)->flags, \
 						   (erts_aint32_t) (MSK)))
 #define ERTS_PROC_LOCK_FLGS_BOR_ACQB_(L, MSK) \
-  ((ErtsProcLocks) erts_smp_atomic32_read_bor_acqb(&(L)->flags, \
+  ((ErtsProcLocks) erts_atomic32_read_bor_acqb(&(L)->flags, \
 						   (erts_aint32_t) (MSK)))
 #define ERTS_PROC_LOCK_FLGS_CMPXCHG_ACQB_(L, NEW, EXPECTED) \
-  ((ErtsProcLocks) erts_smp_atomic32_cmpxchg_acqb(&(L)->flags, \
+  ((ErtsProcLocks) erts_atomic32_cmpxchg_acqb(&(L)->flags, \
 						  (erts_aint32_t) (NEW), \
 						  (erts_aint32_t) (EXPECTED)))
 #define ERTS_PROC_LOCK_FLGS_CMPXCHG_RELB_(L, NEW, EXPECTED) \
-  ((ErtsProcLocks) erts_smp_atomic32_cmpxchg_relb(&(L)->flags, \
+  ((ErtsProcLocks) erts_atomic32_cmpxchg_relb(&(L)->flags, \
 						  (erts_aint32_t) (NEW), \
 						  (erts_aint32_t) (EXPECTED)))
 #define ERTS_PROC_LOCK_FLGS_READ_(L) \
-  ((ErtsProcLocks) erts_smp_atomic32_read_nob(&(L)->flags))
+  ((ErtsProcLocks) erts_atomic32_read_nob(&(L)->flags))
 
 #else /* no opt atomic ops */
 
@@ -557,22 +530,22 @@ ERTS_GLB_INLINE void erts_pix_lock(erts_pix_lock_t *);
 ERTS_GLB_INLINE void erts_pix_unlock(erts_pix_lock_t *);
 ERTS_GLB_INLINE int erts_lc_pix_lock_is_locked(erts_pix_lock_t *);
 
-ERTS_GLB_INLINE ErtsProcLocks erts_smp_proc_raw_trylock__(Process *p,
+ERTS_GLB_INLINE ErtsProcLocks erts_proc_raw_trylock__(Process *p,
 							  ErtsProcLocks locks);
 #ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE void erts_smp_proc_lock_x__(Process *,
+ERTS_GLB_INLINE void erts_proc_lock_x__(Process *,
 					    erts_pix_lock_t *,
 					    ErtsProcLocks,
 					    char *file, unsigned int line);
 #else
-ERTS_GLB_INLINE void erts_smp_proc_lock__(Process *,
+ERTS_GLB_INLINE void erts_proc_lock__(Process *,
 					  erts_pix_lock_t *,
 					  ErtsProcLocks);
 #endif
-ERTS_GLB_INLINE void erts_smp_proc_unlock__(Process *,
+ERTS_GLB_INLINE void erts_proc_unlock__(Process *,
 					    erts_pix_lock_t *,
 					    ErtsProcLocks);
-ERTS_GLB_INLINE int erts_smp_proc_trylock__(Process *,
+ERTS_GLB_INLINE int erts_proc_trylock__(Process *,
 					    erts_pix_lock_t *,
 					    ErtsProcLocks);
 
@@ -600,7 +573,7 @@ ERTS_GLB_INLINE int erts_lc_pix_lock_is_locked(erts_pix_lock_t *pixlck)
 }
 
 /*
- * Helper function for erts_smp_proc_lock__ and erts_smp_proc_trylock__.
+ * Helper function for erts_proc_lock__ and erts_proc_trylock__.
  *
  * Attempts to grab all of 'locks' simultaneously.
  *
@@ -613,7 +586,7 @@ ERTS_GLB_INLINE int erts_lc_pix_lock_is_locked(erts_pix_lock_t *pixlck)
  * Does not release the pix lock.
  */
 ERTS_GLB_INLINE ErtsProcLocks
-erts_smp_proc_raw_trylock__(Process *p, ErtsProcLocks locks)
+erts_proc_raw_trylock__(Process *p, ErtsProcLocks locks)
 {
 #if ERTS_PROC_LOCK_OWN_IMPL
     ErtsProcLocks expct_lflgs = 0;
@@ -641,9 +614,6 @@ erts_smp_proc_raw_trylock__(Process *p, ErtsProcLocks locks)
     if (locks & ERTS_PROC_LOCK_MAIN)
 	if (erts_mtx_trylock(&p->lock.main) == EBUSY)
 	    goto busy_main;
-    if (locks & ERTS_PROC_LOCK_LINK)
-	if (erts_mtx_trylock(&p->lock.link) == EBUSY)
-	    goto busy_link;
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	if (erts_mtx_trylock(&p->lock.msgq) == EBUSY)
 	    goto busy_msgq;
@@ -669,9 +639,6 @@ busy_btm:
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_mtx_unlock(&p->lock.msgq);
 busy_msgq:
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_mtx_unlock(&p->lock.link);
-busy_link:
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_mtx_unlock(&p->lock.main);
 busy_main:
@@ -682,12 +649,12 @@ busy_main:
 
 ERTS_GLB_INLINE void
 #ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_proc_lock_x__(Process *p,
+erts_proc_lock_x__(Process *p,
 		     erts_pix_lock_t *pix_lck,
 		     ErtsProcLocks locks,
 		     char *file, unsigned int line)
 #else
-erts_smp_proc_lock__(Process *p,
+erts_proc_lock__(Process *p,
 		     erts_pix_lock_t *pix_lck,
 		     ErtsProcLocks locks)
 #endif
@@ -709,7 +676,7 @@ erts_smp_proc_lock__(Process *p,
     erts_proc_lc_lock(p, locks, file, line);
 #endif
 
-    old_lflgs = erts_smp_proc_raw_trylock__(p, locks);
+    old_lflgs = erts_proc_raw_trylock__(p, locks);
 
     if (old_lflgs != 0) {
 	/*
@@ -742,8 +709,6 @@ erts_smp_proc_lock__(Process *p,
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_mtx_lock(&p->lock.main);
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_mtx_lock(&p->lock.link);
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_mtx_lock(&p->lock.msgq);
     if (locks & ERTS_PROC_LOCK_BTM)
@@ -761,7 +726,7 @@ erts_smp_proc_lock__(Process *p,
 }
 
 ERTS_GLB_INLINE void
-erts_smp_proc_unlock__(Process *p,
+erts_proc_unlock__(Process *p,
 		       erts_pix_lock_t *pix_lck,
 		       ErtsProcLocks locks)
 {
@@ -845,8 +810,6 @@ erts_smp_proc_unlock__(Process *p,
 	erts_mtx_unlock(&p->lock.btm);
     if (locks & ERTS_PROC_LOCK_MSGQ)
 	erts_mtx_unlock(&p->lock.msgq);
-    if (locks & ERTS_PROC_LOCK_LINK)
-	erts_mtx_unlock(&p->lock.link);
     if (locks & ERTS_PROC_LOCK_MAIN)
 	erts_mtx_unlock(&p->lock.main);
 #endif
@@ -854,7 +817,7 @@ erts_smp_proc_unlock__(Process *p,
 }
 
 ERTS_GLB_INLINE int
-erts_smp_proc_trylock__(Process *p,
+erts_proc_trylock__(Process *p,
 			erts_pix_lock_t *pix_lck,
 			ErtsProcLocks locks)
 {
@@ -875,7 +838,7 @@ erts_smp_proc_trylock__(Process *p,
 	erts_pix_lock(pix_lck);
 #endif
 
-	if (erts_smp_proc_raw_trylock__(p, locks) != 0) {
+	if (erts_proc_raw_trylock__(p, locks) != 0) {
 	    /* Didn't get all locks... */
 	    res = EBUSY;
 
@@ -912,7 +875,7 @@ erts_smp_proc_trylock__(Process *p,
     return res;
 
 #elif ERTS_PROC_LOCK_RAW_MUTEX_IMPL
-    if (erts_smp_proc_raw_trylock__(p, locks) != 0)
+    if (erts_proc_raw_trylock__(p, locks) != 0)
 	return EBUSY;
     else {
 #ifdef ERTS_PROC_LOCK_DEBUG
@@ -933,11 +896,11 @@ erts_proc_lock_op_debug(Process *p, ErtsProcLocks locks, int locked)
 	if (locks & lock) {
 	    erts_aint32_t lock_count;
 	    if (locked) {
-		lock_count = erts_smp_atomic32_inc_read_nob(&p->lock.locked[i]);
+		lock_count = erts_atomic32_inc_read_nob(&p->lock.locked[i]);
 		ERTS_LC_ASSERT(lock_count == 1);
 	    }
 	    else {
-		lock_count = erts_smp_atomic32_dec_read_nob(&p->lock.locked[i]);
+		lock_count = erts_atomic32_dec_read_nob(&p->lock.locked[i]);
 		ERTS_LC_ASSERT(lock_count == 0);
 	    }
 	}
@@ -947,18 +910,20 @@ erts_proc_lock_op_debug(Process *p, ErtsProcLocks locks, int locked)
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#endif /* ERTS_SMP */
 
 #ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE void erts_smp_proc_lock_x(Process *, ErtsProcLocks, char *file, unsigned int line);
+ERTS_GLB_INLINE void erts_proc_lock_x(Process *, ErtsProcLocks, char *file, unsigned int line);
 #else
-ERTS_GLB_INLINE void erts_smp_proc_lock(Process *, ErtsProcLocks);
+ERTS_GLB_INLINE void erts_proc_lock(Process *, ErtsProcLocks);
 #endif
-ERTS_GLB_INLINE void erts_smp_proc_unlock(Process *, ErtsProcLocks);
-ERTS_GLB_INLINE int erts_smp_proc_trylock(Process *, ErtsProcLocks);
+ERTS_GLB_INLINE void erts_proc_unlock(Process *, ErtsProcLocks);
+ERTS_GLB_INLINE int erts_proc_trylock(Process *, ErtsProcLocks);
 
 ERTS_GLB_INLINE void erts_proc_inc_refc(Process *);
 ERTS_GLB_INLINE void erts_proc_dec_refc(Process *);
+ERTS_GLB_INLINE void erts_proc_dec_refc_free_func(Process *p,
+                                                  void (*func)(int, void *),
+                                                  void *arg);
 ERTS_GLB_INLINE void erts_proc_add_refc(Process *, Sint);
 ERTS_GLB_INLINE Sint erts_proc_read_refc(Process *);
 
@@ -966,94 +931,91 @@ ERTS_GLB_INLINE Sint erts_proc_read_refc(Process *);
 
 ERTS_GLB_INLINE void
 #ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_proc_lock_x(Process *p, ErtsProcLocks locks, char *file, unsigned int line)
+erts_proc_lock_x(Process *p, ErtsProcLocks locks, char *file, unsigned int line)
 #else
-erts_smp_proc_lock(Process *p, ErtsProcLocks locks)
+erts_proc_lock(Process *p, ErtsProcLocks locks)
 #endif 
 {
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_smp_proc_lock_x__(p,
+#if defined(ERTS_ENABLE_LOCK_POSITION)
+    erts_proc_lock_x__(p,
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
 			 NULL,
 #else
 			 ERTS_PID2PIXLOCK(p->common.id),
 #endif /*ERTS_PROC_LOCK_ATOMIC_IMPL*/
 			 locks, file, line);
-#elif defined(ERTS_SMP)
-    erts_smp_proc_lock__(p,
+#else
+    erts_proc_lock__(p,
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
 			 NULL,
 #else
 			 ERTS_PID2PIXLOCK(p->common.id),
 #endif /*ERTS_PROC_LOCK_ATOMIC_IMPL*/
 			 locks);
-#endif /*ERTS_SMP*/
+#endif /*ERTS_ENABLE_LOCK_POSITION*/
 }
 
 ERTS_GLB_INLINE void
-erts_smp_proc_unlock(Process *p, ErtsProcLocks locks)
+erts_proc_unlock(Process *p, ErtsProcLocks locks)
 {
-#ifdef ERTS_SMP
-    erts_smp_proc_unlock__(p,
+    erts_proc_unlock__(p,
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
 			   NULL,
 #else
 			   ERTS_PID2PIXLOCK(p->common.id),
 #endif
 			   locks);
-#endif
 }
 
 ERTS_GLB_INLINE int
-erts_smp_proc_trylock(Process *p, ErtsProcLocks locks)
+erts_proc_trylock(Process *p, ErtsProcLocks locks)
 {
-#ifndef ERTS_SMP
-    return 0;
-#else
-    return erts_smp_proc_trylock__(p,
+    return erts_proc_trylock__(p,
 #if ERTS_PROC_LOCK_ATOMIC_IMPL
 				   NULL,
 #else
 				   ERTS_PID2PIXLOCK(p->common.id),
 #endif
 				   locks);
-#endif
 }
 
 ERTS_GLB_INLINE void erts_proc_inc_refc(Process *p)
 {
-    ASSERT(!(erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
-#ifdef ERTS_SMP
+    ASSERT(!(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
     erts_ptab_atmc_inc_refc(&p->common);
-#else
-    erts_ptab_inc_refc(&p->common);
-#endif
 }
 
 ERTS_GLB_INLINE void erts_proc_dec_refc(Process *p)
 {
     Sint referred;
-    ASSERT(!(erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
-#ifdef ERTS_SMP
+    ASSERT(!(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
+    referred = erts_ptab_atmc_dec_test_refc(&p->common);
+    if (!referred) {
+	ASSERT(ERTS_PROC_IS_EXITING(p));
+	erts_free_proc(p);
+    }
+}
+
+ERTS_GLB_INLINE void erts_proc_dec_refc_free_func(Process *p,
+                                                  void (*func)(int, void *),
+                                                  void *arg)
+{
+    Sint referred;
+    ASSERT(!(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
     referred = erts_ptab_atmc_dec_test_refc(&p->common);
-#else
-    referred = erts_ptab_dec_test_refc(&p->common);
-#endif
     if (!referred) {
 	ASSERT(ERTS_PROC_IS_EXITING(p));
+        (*func)(!0, arg);
 	erts_free_proc(p);
+        (*func)(0, arg);
     }
 }
 
 ERTS_GLB_INLINE void erts_proc_add_refc(Process *p, Sint add_refc)
 {
     Sint referred;
-    ASSERT(!(erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
-#ifdef ERTS_SMP
+    ASSERT(!(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
     referred = erts_ptab_atmc_add_test_refc(&p->common, add_refc);
-#else
-    referred = erts_ptab_add_test_refc(&p->common, add_refc);
-#endif
     if (!referred) {
 	ASSERT(ERTS_PROC_IS_EXITING(p));
 	erts_free_proc(p);
@@ -1062,17 +1024,12 @@ ERTS_GLB_INLINE void erts_proc_add_refc(Process *p, Sint add_refc)
 
 ERTS_GLB_INLINE Sint erts_proc_read_refc(Process *p)
 {
-    ASSERT(!(erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
-#ifdef ERTS_SMP
+    ASSERT(!(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_PROXY));
     return erts_ptab_atmc_read_refc(&p->common);
-#else
-    return erts_ptab_read_refc(&p->common);
-#endif
 }
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#ifdef ERTS_SMP
 void erts_proc_lock_init(Process *);
 void erts_proc_lock_fin(Process *);
 void erts_proc_safelock(Process *a_proc,
@@ -1081,7 +1038,6 @@ void erts_proc_safelock(Process *a_proc,
 			Process *b_proc,
 			ErtsProcLocks b_have_locks,
 			ErtsProcLocks b_need_locks);
-#endif
 
 /*
  * --- Process table lookup ------------------------------------------------
@@ -1113,9 +1069,6 @@ ERTS_GLB_INLINE Process *erts_pix2proc(int ix);
 ERTS_GLB_INLINE Process *erts_proc_lookup_raw(Eterm pid);
 ERTS_GLB_INLINE Process *erts_proc_lookup(Eterm pid);
 
-#ifndef ERTS_SMP
-ERTS_GLB_INLINE
-#endif
 Process *erts_pid2proc_opt(Process *, ErtsProcLocks, Eterm, ErtsProcLocks, int);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
@@ -1132,7 +1085,7 @@ ERTS_GLB_INLINE Process *erts_proc_lookup_raw(Eterm pid)
 {
     Process *proc;
 
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_lc_is_delaying());
+    ERTS_LC_ASSERT(erts_thr_progress_lc_is_delaying());
 
     if (is_not_internal_pid(pid))
 	return NULL;
@@ -1152,25 +1105,6 @@ ERTS_GLB_INLINE Process *erts_proc_lookup(Eterm pid)
     return proc;
 }
 
-#ifndef ERTS_SMP
-ERTS_GLB_INLINE Process *
-erts_pid2proc_opt(Process *c_p_unused,
-		  ErtsProcLocks c_p_have_locks_unused,
-		  Eterm pid,
-		  ErtsProcLocks pid_need_locks_unused,
-		  int flags)
-{
-    Process *proc = erts_proc_lookup_raw(pid);
-    if (!proc)
-	return NULL;
-    if (!(flags & ERTS_P2P_FLG_ALLOW_OTHER_X)
-	&& ERTS_PROC_IS_EXITING(proc))
-	return NULL;
-    if (flags & ERTS_P2P_FLG_INC_REFC)
-	erts_proc_inc_refc(proc);
-    return proc;
-}
-#endif /* !ERTS_SMP */
 
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
diff --git a/erts/emulator/beam/erl_ptab.c b/erts/emulator/beam/erl_ptab.c
index b3bcb3af3f..38c095fb4a 100644
--- a/erts/emulator/beam/erl_ptab.c
+++ b/erts/emulator/beam/erl_ptab.c
@@ -284,31 +284,31 @@ struct ErtsPTabListBifData_ {
 static ERTS_INLINE void
 last_data_init_nob(ErtsPTab *ptab, Uint64 val)
 {
-    erts_smp_atomic64_init_nob(&ptab->vola.tile.last_data, (erts_aint64_t) val);
+    erts_atomic64_init_nob(&ptab->vola.tile.last_data, (erts_aint64_t) val);
 }
 
 static ERTS_INLINE void
 last_data_set_relb(ErtsPTab *ptab, Uint64 val)
 {
-    erts_smp_atomic64_set_relb(&ptab->vola.tile.last_data, (erts_aint64_t) val);
+    erts_atomic64_set_relb(&ptab->vola.tile.last_data, (erts_aint64_t) val);
 }
 
 static ERTS_INLINE Uint64
 last_data_read_nob(ErtsPTab *ptab)
 {
-    return (Uint64) erts_smp_atomic64_read_nob(&ptab->vola.tile.last_data);
+    return (Uint64) erts_atomic64_read_nob(&ptab->vola.tile.last_data);
 }
 
 static ERTS_INLINE Uint64
 last_data_read_acqb(ErtsPTab *ptab)
 {
-    return (Uint64) erts_smp_atomic64_read_acqb(&ptab->vola.tile.last_data);
+    return (Uint64) erts_atomic64_read_acqb(&ptab->vola.tile.last_data);
 }
 
 static ERTS_INLINE Uint64
 last_data_cmpxchg_relb(ErtsPTab *ptab, Uint64 new, Uint64 exp)
 {
-    return (Uint64) erts_smp_atomic64_cmpxchg_relb(&ptab->vola.tile.last_data,
+    return (Uint64) erts_atomic64_cmpxchg_relb(&ptab->vola.tile.last_data,
 						   (erts_aint64_t) new,
 						   (erts_aint64_t) exp);
 }
@@ -346,9 +346,9 @@ ix_to_free_id_data_ix(ErtsPTab *ptab, Uint32 ix)
 UWord
 erts_ptab_mem_size(ErtsPTab *ptab)
 {
-    UWord size = ptab->r.o.max*sizeof(erts_smp_atomic_t);
+    UWord size = ptab->r.o.max*sizeof(erts_atomic_t);
     if (ptab->r.o.free_id_data)
-	size += ptab->r.o.max*sizeof(erts_smp_atomic32_t);
+	size += ptab->r.o.max*sizeof(erts_atomic32_t);
     return size;
 }
 
@@ -367,14 +367,14 @@ erts_ptab_init_table(ErtsPTab *ptab,
     size_t tab_sz, alloc_sz;
     Uint32 bits, cl, cli, ix, ix_per_cache_line, tab_cache_lines; 
     char *tab_end;
-    erts_smp_atomic_t *tab_entry;
-    erts_smp_rwmtx_opt_t rwmtx_opts = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opts.type = ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
-    rwmtx_opts.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_atomic_t *tab_entry;
+    erts_rwmtx_opt_t rwmtx_opts = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opts.type = ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
+    rwmtx_opts.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&ptab->list.data.rwmtx, &rwmtx_opts, name, NIL,
+    erts_rwmtx_init_opt(&ptab->list.data.rwmtx, &rwmtx_opts, name, NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
-    erts_smp_atomic32_init_nob(&ptab->vola.tile.count, 0);
+    erts_atomic32_init_nob(&ptab->vola.tile.count, 0);
     last_data_init_nob(ptab, ~((Uint64) 0));
 
     /* A size that is a power of 2 is to prefer performance wise */
@@ -388,20 +388,20 @@ erts_ptab_init_table(ErtsPTab *ptab,
     ptab->r.o.element_size = element_size;
     ptab->r.o.max = size;
 
-    tab_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_smp_atomic_t));
+    tab_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_atomic_t));
     alloc_sz = tab_sz;
     if (!legacy)
-	alloc_sz += ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_smp_atomic32_t));
+	alloc_sz += ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_atomic32_t));
     ptab->r.o.tab = erts_alloc_permanent_cache_aligned(atype, alloc_sz);
     tab_end = ((char *) ptab->r.o.tab) + tab_sz;
     tab_entry = ptab->r.o.tab;
     while (tab_end > ((char *) tab_entry)) {
-	erts_smp_atomic_init_nob(tab_entry, ERTS_AINT_NULL);
+	erts_atomic_init_nob(tab_entry, ERTS_AINT_NULL);
 	tab_entry++;
     }
 
     tab_cache_lines = tab_sz/ERTS_CACHE_LINE_SIZE;
-    ix_per_cache_line = (ERTS_CACHE_LINE_SIZE/sizeof(erts_smp_atomic_t));
+    ix_per_cache_line = (ERTS_CACHE_LINE_SIZE/sizeof(erts_atomic_t));
     ASSERT((ptab->r.o.max & (ptab->r.o.max - 1)) == 0); /* power of 2 */
     ASSERT((ix_per_cache_line & (ix_per_cache_line - 1)) == 0); /* power of 2 */
     ASSERT((tab_cache_lines & (tab_cache_lines - 1)) == 0); /* power of 2 */
@@ -429,11 +429,11 @@ erts_ptab_init_table(ErtsPTab *ptab,
     }
     else {
 
-	tab_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_smp_atomic32_t));
-	ptab->r.o.free_id_data = (erts_smp_atomic32_t *) tab_end;
+	tab_sz = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(size*sizeof(erts_atomic32_t));
+	ptab->r.o.free_id_data = (erts_atomic32_t *) tab_end;
 
 	tab_cache_lines = tab_sz/ERTS_CACHE_LINE_SIZE;
-	ix_per_cache_line = (ERTS_CACHE_LINE_SIZE/sizeof(erts_smp_atomic32_t));
+	ix_per_cache_line = (ERTS_CACHE_LINE_SIZE/sizeof(erts_atomic32_t));
 
 	ptab->r.o.dix_cl_mask = tab_cache_lines-1;
 	ptab->r.o.dix_cl_shift = erts_fit_in_bits_int32(ix_per_cache_line-1);
@@ -448,19 +448,19 @@ erts_ptab_init_table(ErtsPTab *ptab,
 	ix = 0;
 	for (cl = 0; cl < tab_cache_lines; cl++) {
 	    for (cli = 0; cli < ix_per_cache_line; cli++) {
-		erts_smp_atomic32_init_nob(&ptab->r.o.free_id_data[ix],
+		erts_atomic32_init_nob(&ptab->r.o.free_id_data[ix],
 					   cli*tab_cache_lines+cl);
-		ASSERT(erts_smp_atomic32_read_nob(&ptab->r.o.free_id_data[ix]) != ptab->r.o.invalid_data);
+		ASSERT(erts_atomic32_read_nob(&ptab->r.o.free_id_data[ix]) != ptab->r.o.invalid_data);
 		ix++;
 	    }
 	}
 
-	erts_smp_atomic32_init_nob(&ptab->vola.tile.aid_ix, -1);
-	erts_smp_atomic32_init_nob(&ptab->vola.tile.fid_ix, -1);
+	erts_atomic32_init_nob(&ptab->vola.tile.aid_ix, -1);
+	erts_atomic32_init_nob(&ptab->vola.tile.fid_ix, -1);
 
     }
 
-    erts_smp_interval_init(&ptab->list.data.interval);
+    erts_interval_init(&ptab->list.data.interval);
     ptab->list.data.deleted.start = NULL;
     ptab->list.data.deleted.end = NULL;
     ptab->list.data.chunks = (((ptab->r.o.max - 1)
@@ -480,9 +480,9 @@ erts_ptab_init_table(ErtsPTab *ptab,
 	 * have ERTS_PTAB_MAX_SIZE-1 valid elements in the table while
 	 * still having a table size of the power of 2.
 	 */
-	erts_smp_atomic32_inc_nob(&ptab->vola.tile.count);
+	erts_atomic32_inc_nob(&ptab->vola.tile.count);
 	pix = erts_ptab_data2pix(ptab, ptab->r.o.invalid_data);
-	erts_smp_atomic_set_relb(&ptab->r.o.tab[pix],
+	erts_atomic_set_relb(&ptab->r.o.tab[pix],
 				 (erts_aint_t) ptab->r.o.invalid_element);
     }
 
@@ -506,12 +506,12 @@ erts_ptab_new_element(ErtsPTab *ptab,
 
     erts_ptab_rlock(ptab);
 
-    count = erts_smp_atomic32_inc_read_acqb(&ptab->vola.tile.count);
+    count = erts_atomic32_inc_read_acqb(&ptab->vola.tile.count);
     if (count > ptab->r.o.max) {
 	while (1) {
 	    erts_aint32_t act_count;
 
-	    act_count = erts_smp_atomic32_cmpxchg_relb(&ptab->vola.tile.count,
+	    act_count = erts_atomic32_cmpxchg_relb(&ptab->vola.tile.count,
 						       count-1,
 						       count);
 	    if (act_count == count) {
@@ -525,14 +525,14 @@ erts_ptab_new_element(ErtsPTab *ptab,
     }
 
     ptab_el->u.alive.started_interval
-	= erts_smp_current_interval_nob(erts_ptab_interval(ptab));
+	= erts_current_interval_nob(erts_ptab_interval(ptab));
 
     if (ptab->r.o.free_id_data) {
 	do {
-	    ix = (Uint32) erts_smp_atomic32_inc_read_acqb(&ptab->vola.tile.aid_ix);
+	    ix = (Uint32) erts_atomic32_inc_read_acqb(&ptab->vola.tile.aid_ix);
 	    ix = ix_to_free_id_data_ix(ptab, ix);
 
-	    data = erts_smp_atomic32_xchg_nob(&ptab->r.o.free_id_data[ix],
+	    data = erts_atomic32_xchg_nob(&ptab->r.o.free_id_data[ix],
 					      (erts_aint32_t)ptab->r.o.invalid_data);
 	}while ((Eterm)data == ptab->r.o.invalid_data);
 
@@ -546,10 +546,10 @@ erts_ptab_new_element(ErtsPTab *ptab,
 	pix = erts_ptab_data2pix(ptab, (Eterm) data);
 
 #ifdef DEBUG
-	ASSERT(ERTS_AINT_NULL == erts_smp_atomic_xchg_relb(&ptab->r.o.tab[pix],
+	ASSERT(ERTS_AINT_NULL == erts_atomic_xchg_relb(&ptab->r.o.tab[pix],
 							   (erts_aint_t) ptab_el));
 #else
-	erts_smp_atomic_set_relb(&ptab->r.o.tab[pix], (erts_aint_t) ptab_el);
+	erts_atomic_set_relb(&ptab->r.o.tab[pix], (erts_aint_t) ptab_el);
 #endif
 
 	erts_ptab_runlock(ptab);
@@ -563,7 +563,7 @@ erts_ptab_new_element(ErtsPTab *ptab,
     restart:
 
 	ptab_el->u.alive.started_interval
-	    = erts_smp_current_interval_nob(erts_ptab_interval(ptab));
+	    = erts_current_interval_nob(erts_ptab_interval(ptab));
 
 	ld = last_data_read_acqb(ptab);
 
@@ -571,10 +571,10 @@ erts_ptab_new_element(ErtsPTab *ptab,
 	while (1) {
 	    ld++;
 	    pix = erts_ptab_data2pix(ptab, ERTS_PTAB_LastData2EtermData(ld));
-	    if (erts_smp_atomic_read_nob(&ptab->r.o.tab[pix])
+	    if (erts_atomic_read_nob(&ptab->r.o.tab[pix])
 		== ERTS_AINT_NULL) {
 		erts_aint_t val;
-		val = erts_smp_atomic_cmpxchg_relb(&ptab->r.o.tab[pix],
+		val = erts_atomic_cmpxchg_relb(&ptab->r.o.tab[pix],
 						   invalid,	
 						   ERTS_AINT_NULL);
 
@@ -621,10 +621,10 @@ erts_ptab_new_element(ErtsPTab *ptab,
 
 	/* Move into slot reserved */
 #ifdef DEBUG
-	ASSERT(invalid == erts_smp_atomic_xchg_relb(&ptab->r.o.tab[pix],
+	ASSERT(invalid == erts_atomic_xchg_relb(&ptab->r.o.tab[pix],
 						(erts_aint_t) ptab_el));
 #else
-	erts_smp_atomic_set_relb(&ptab->r.o.tab[pix], (erts_aint_t) ptab_el);
+	erts_atomic_set_relb(&ptab->r.o.tab[pix], (erts_aint_t) ptab_el);
 #endif
 
 	if (rlocked)
@@ -644,7 +644,7 @@ save_deleted_element(ErtsPTab *ptab, ErtsPTabElementCommon *ptab_el)
 					       sizeof(ErtsPTabDeletedElement));
     ERTS_PTAB_LIST_ASSERT(ptab->list.data.deleted.start
 			  && ptab->list.data.deleted.end);
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_ptab_is_rwlocked(ptab));
+    ERTS_LC_ASSERT(erts_lc_ptab_is_rwlocked(ptab));
 
     ERTS_PTAB_LIST_DBG_CHK_DEL_LIST(ptab);
 
@@ -654,7 +654,7 @@ save_deleted_element(ErtsPTab *ptab, ErtsPTabElementCommon *ptab_el)
     ptdep->u.element.id = ptab_el->id;
     ptdep->u.element.inserted = ptab_el->u.alive.started_interval;
     ptdep->u.element.deleted =
-	erts_smp_current_interval_nob(erts_ptab_interval(ptab));
+	erts_current_interval_nob(erts_ptab_interval(ptab));
 
     ptab->list.data.deleted.end->next = ptdep;
     ptab->list.data.deleted.end = ptdep;
@@ -678,7 +678,7 @@ erts_ptab_delete_element(ErtsPTab *ptab,
     pix = erts_ptab_id2pix(ptab, ptab_el->id);
 
     /* *Need* to be an managed thread */
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_managed_thread());
+    ERTS_LC_ASSERT(erts_thr_progress_is_managed_thread());
 
     erts_ptab_rlock(ptab);
     maybe_save = ptab->list.data.deleted.end != NULL;
@@ -687,7 +687,7 @@ erts_ptab_delete_element(ErtsPTab *ptab,
 	erts_ptab_rwlock(ptab);
     }
 
-    erts_smp_atomic_set_relb(&ptab->r.o.tab[pix], ERTS_AINT_NULL);
+    erts_atomic_set_relb(&ptab->r.o.tab[pix], ERTS_AINT_NULL);
 
     if (ptab->r.o.free_id_data) {
 	Uint32 prev_data;
@@ -703,17 +703,17 @@ erts_ptab_delete_element(ErtsPTab *ptab,
 	ASSERT(pix == erts_ptab_data2pix(ptab, data));
 
 	do { 
-	    ix = (Uint32) erts_smp_atomic32_inc_read_relb(&ptab->vola.tile.fid_ix);
+	    ix = (Uint32) erts_atomic32_inc_read_relb(&ptab->vola.tile.fid_ix);
 	    ix = ix_to_free_id_data_ix(ptab, ix);
     
-	    prev_data = erts_smp_atomic32_cmpxchg_nob(&ptab->r.o.free_id_data[ix],
+	    prev_data = erts_atomic32_cmpxchg_nob(&ptab->r.o.free_id_data[ix],
 						      data,
 						      ptab->r.o.invalid_data);
 	}while ((Eterm)prev_data != ptab->r.o.invalid_data);
     }
 
-    ASSERT(erts_smp_atomic32_read_nob(&ptab->vola.tile.count) > 0);
-    erts_smp_atomic32_dec_relb(&ptab->vola.tile.count);
+    ASSERT(erts_atomic32_read_nob(&ptab->vola.tile.count) > 0);
+    erts_atomic32_dec_relb(&ptab->vola.tile.count);
 
     if (!maybe_save)
 	erts_ptab_runlock(ptab);
@@ -927,7 +927,7 @@ ptab_list_bif_engine(Process *c_p, Eterm *res_accp, Binary *mbp)
 				 sizeof(ErtsPTabDeletedElement));
 		ptlbdp->bif_invocation->ix = -1;
 		ptlbdp->bif_invocation->u.bif_invocation.interval
-		    = erts_smp_step_interval_nob(erts_ptab_interval(ptab));
+		    = erts_step_interval_nob(erts_ptab_interval(ptab));
 		ERTS_PTAB_LIST_DBG_CHK_DEL_LIST(ptab);
 
 		ptlbdp->bif_invocation->next = NULL;
@@ -968,12 +968,12 @@ ptab_list_bif_engine(Process *c_p, Eterm *res_accp, Binary *mbp)
 		locked = 1;
 	    }
 
-	    ERTS_SMP_LC_ASSERT(erts_smp_lc_ptab_is_rwlocked(ptab));
+	    ERTS_LC_ASSERT(erts_lc_ptab_is_rwlocked(ptab));
 	    ERTS_PTAB_LIST_DBG_TRACE(p->common.id, insp_table);
 
 	    if (cix != 0)
 		ptlbdp->chunk[cix].interval
-		    = erts_smp_step_interval_nob(erts_ptab_interval(ptab));
+		    = erts_step_interval_nob(erts_ptab_interval(ptab));
 	    else if (ptlbdp->bif_invocation)
 		ptlbdp->chunk[0].interval = *invocation_interval_p;
 	    /* else: interval is irrelevant */
@@ -1331,18 +1331,18 @@ static void assert_ptab_consistency(ErtsPTab *ptab)
 	int null_slots = 0;
 	
 	for (ix=0; ix < ptab->r.o.max; ix++) {
-	    if (erts_smp_atomic32_read_nob(&ptab->r.o.free_id_data[ix]) != ptab->r.o.invalid_data) {
+	    if (erts_atomic32_read_nob(&ptab->r.o.free_id_data[ix]) != ptab->r.o.invalid_data) {
 		++free_pids;
-		data = erts_smp_atomic32_read_nob(&ptab->r.o.free_id_data[ix]);
+		data = erts_atomic32_read_nob(&ptab->r.o.free_id_data[ix]);
 		pix = erts_ptab_data2pix(ptab, (Eterm) data);
 		ASSERT(erts_ptab_pix2intptr_nob(ptab, pix) == ERTS_AINT_NULL);
 	    }
-	    if (erts_smp_atomic_read_nob(&ptab->r.o.tab[ix]) == ERTS_AINT_NULL) {
+	    if (erts_atomic_read_nob(&ptab->r.o.tab[ix]) == ERTS_AINT_NULL) {
 		++null_slots;
 	    }
 	}	
 	ASSERT(free_pids == null_slots);
-	ASSERT(free_pids == ptab->r.o.max - erts_smp_atomic32_read_nob(&ptab->vola.tile.count));
+	ASSERT(free_pids == ptab->r.o.max - erts_atomic32_read_nob(&ptab->vola.tile.count));
     }
 #endif
 }
@@ -1366,7 +1366,7 @@ erts_ptab_test_next_id(ErtsPTab *ptab, int set, Uint next)
 	    Uint32 i, max_ix, num, stop_id_ix;
 	    max_ix = ptab->r.o.max - 1;
 	    num = next;
-	    id_ix = (Uint32) erts_smp_atomic32_read_nob(&ptab->vola.tile.aid_ix);
+	    id_ix = (Uint32) erts_atomic32_read_nob(&ptab->vola.tile.aid_ix);
 
 	    for (i=0; i <= max_ix; ++i) {
 		Uint32 pix;
@@ -1380,26 +1380,26 @@ erts_ptab_test_next_id(ErtsPTab *ptab, int set, Uint next)
 		if (ERTS_AINT_NULL == erts_ptab_pix2intptr_nob(ptab, pix)) {
 		    ++id_ix;
 		    dix = ix_to_free_id_data_ix(ptab, id_ix);
-		    erts_smp_atomic32_set_nob(&ptab->r.o.free_id_data[dix], num);
+		    erts_atomic32_set_nob(&ptab->r.o.free_id_data[dix], num);
 		    ASSERT(pix == erts_ptab_data2pix(ptab, num));
 		}
 	    }
-	    erts_smp_atomic32_set_nob(&ptab->vola.tile.fid_ix, id_ix);
+	    erts_atomic32_set_nob(&ptab->vola.tile.fid_ix, id_ix);
 
 	    /* Write invalid_data in rest of free_id_data[]: */
-	    stop_id_ix = (1 + erts_smp_atomic32_read_nob(&ptab->vola.tile.aid_ix)) & max_ix;
+	    stop_id_ix = (1 + erts_atomic32_read_nob(&ptab->vola.tile.aid_ix)) & max_ix;
 	    while (1) {
 		id_ix = (id_ix+1) & max_ix;
 		if (id_ix == stop_id_ix)
 		    break;
 		dix = ix_to_free_id_data_ix(ptab, id_ix);
-		erts_smp_atomic32_set_nob(&ptab->r.o.free_id_data[dix],
+		erts_atomic32_set_nob(&ptab->r.o.free_id_data[dix],
 					  ptab->r.o.invalid_data);
 	    }
 	}
-	id_ix = (Uint32) erts_smp_atomic32_read_nob(&ptab->vola.tile.aid_ix) + 1;
+	id_ix = (Uint32) erts_atomic32_read_nob(&ptab->vola.tile.aid_ix) + 1;
 	dix = ix_to_free_id_data_ix(ptab, id_ix);
-	res = (Sint) erts_smp_atomic32_read_nob(&ptab->r.o.free_id_data[dix]);
+	res = (Sint) erts_atomic32_read_nob(&ptab->r.o.free_id_data[dix]);
     }
     else {
 	/* Deprecated legacy algorithm... */
@@ -1616,11 +1616,11 @@ debug_ptab_list_verify_all_pids(ErtsPTabListBifData *ptlbdp)
 static void
 debug_ptab_list_check_del_list(ErtsPTab *ptab)
 {
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_ptab_is_rwlocked(ptab));
+    ERTS_LC_ASSERT(erts_lc_ptab_is_rwlocked(ptab));
     if (!ptab->list.data.deleted.start)
 	ERTS_PTAB_LIST_ASSERT(!ptab->list.data.deleted.end);
     else {
-	Uint64 curr_interval = erts_smp_current_interval_nob(erts_ptab_interval(ptab));
+	Uint64 curr_interval = erts_current_interval_nob(erts_ptab_interval(ptab));
 	Uint64 *prev_x_interval_p = NULL;
 	ErtsPTabDeletedElement *ptdep;
 
diff --git a/erts/emulator/beam/erl_ptab.h b/erts/emulator/beam/erl_ptab.h
index fecfd96ab0..94f0247492 100644
--- a/erts/emulator/beam/erl_ptab.h
+++ b/erts/emulator/beam/erl_ptab.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2012-2016. All Rights Reserved.
+ * Copyright Ericsson AB 2012-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -35,7 +35,7 @@
 #include "erl_thr_progress.h"
 #undef ERL_THR_PROGRESS_TSD_TYPE_ONLY
 #include "erl_alloc.h"
-#include "erl_monitors.h"
+#include "erl_monitor_link.h"
 
 #define ERTS_TRACER(P)          ((P)->common.tracer)
 #define ERTS_TRACER_MODULE(T) 	(CAR(list_val(T)))
@@ -44,6 +44,7 @@
 
 #define ERTS_P_LINKS(P)		((P)->common.u.alive.links)
 #define ERTS_P_MONITORS(P)	((P)->common.u.alive.monitors)
+#define ERTS_P_LT_MONITORS(P)	((P)->common.u.alive.lt_monitors)
 
 #define IS_TRACED(p) \
     (ERTS_TRACER(p) != NIL)
@@ -60,7 +61,7 @@ typedef struct {
     } refc;
     ErtsTracer tracer;
     Uint trace_flags;
-    erts_smp_atomic_t timer;
+    erts_atomic_t timer;
     union {
 	/* --- While being alive --- */
 	struct {
@@ -68,6 +69,7 @@ typedef struct {
 	    struct reg_proc *reg;
 	    ErtsLink *links;
 	    ErtsMonitor *monitors;
+            ErtsMonitor *lt_monitors;
 	} alive;
 
 	/* --- While being released --- */
@@ -78,7 +80,7 @@ typedef struct {
 typedef struct ErtsPTabDeletedElement_ ErtsPTabDeletedElement;
 
 typedef struct {
-    erts_smp_rwmtx_t rwmtx;
+    erts_rwmtx_t rwmtx;
     erts_interval_t interval;
     struct {
 	ErtsPTabDeletedElement *start;
@@ -88,15 +90,15 @@ typedef struct {
 } ErtsPTabListData;
 
 typedef struct {
-    erts_smp_atomic64_t last_data;
-    erts_smp_atomic32_t count;
-    erts_smp_atomic32_t aid_ix;
-    erts_smp_atomic32_t fid_ix;
+    erts_atomic64_t last_data;
+    erts_atomic32_t count;
+    erts_atomic32_t aid_ix;
+    erts_atomic32_t fid_ix;
 } ErtsPTabVolatileData;
 
 typedef struct {
-    erts_smp_atomic_t *tab;
-    erts_smp_atomic32_t *free_id_data;
+    erts_atomic_t *tab;
+    erts_atomic32_t *free_id_data;
     Uint32 max;
     Uint32 pix_mask;
     Uint32 pix_cl_mask;
@@ -223,8 +225,8 @@ ERTS_GLB_INLINE void erts_ptab_runlock(ErtsPTab *ptab);
 ERTS_GLB_INLINE void erts_ptab_rwlock(ErtsPTab *ptab);
 ERTS_GLB_INLINE int erts_ptab_tryrwlock(ErtsPTab *ptab);
 ERTS_GLB_INLINE void erts_ptab_rwunlock(ErtsPTab *ptab);
-ERTS_GLB_INLINE int erts_smp_lc_ptab_is_rlocked(ErtsPTab *ptab);
-ERTS_GLB_INLINE int erts_smp_lc_ptab_is_rwlocked(ErtsPTab *ptab);
+ERTS_GLB_INLINE int erts_lc_ptab_is_rlocked(ErtsPTab *ptab);
+ERTS_GLB_INLINE int erts_lc_ptab_is_rwlocked(ErtsPTab *ptab);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
@@ -245,7 +247,7 @@ ERTS_GLB_INLINE int
 erts_ptab_count(ErtsPTab *ptab)
 {
     int max = ptab->r.o.max;
-    erts_aint32_t res = erts_smp_atomic32_read_nob(&ptab->vola.tile.count);
+    erts_aint32_t res = erts_atomic32_read_nob(&ptab->vola.tile.count);
     if (max == ERTS_PTAB_MAX_SIZE) {
 	max--;
 	res--;
@@ -352,25 +354,25 @@ erts_ptab_id2data(ErtsPTab *ptab, Eterm id)
 ERTS_GLB_INLINE erts_aint_t erts_ptab_pix2intptr_nob(ErtsPTab *ptab, int ix)
 {
     ASSERT(0 <= ix && ix < ptab->r.o.max);
-    return erts_smp_atomic_read_nob(&ptab->r.o.tab[ix]);
+    return erts_atomic_read_nob(&ptab->r.o.tab[ix]);
 }
 
 ERTS_GLB_INLINE erts_aint_t erts_ptab_pix2intptr_ddrb(ErtsPTab *ptab, int ix)
 {
     ASSERT(0 <= ix && ix < ptab->r.o.max);
-    return erts_smp_atomic_read_ddrb(&ptab->r.o.tab[ix]);
+    return erts_atomic_read_ddrb(&ptab->r.o.tab[ix]);
 }
 
 ERTS_GLB_INLINE erts_aint_t erts_ptab_pix2intptr_rb(ErtsPTab *ptab, int ix)
 {
     ASSERT(0 <= ix && ix < ptab->r.o.max);
-    return erts_smp_atomic_read_rb(&ptab->r.o.tab[ix]);
+    return erts_atomic_read_rb(&ptab->r.o.tab[ix]);
 }
 
 ERTS_GLB_INLINE erts_aint_t erts_ptab_pix2intptr_acqb(ErtsPTab *ptab, int ix)
 {
     ASSERT(0 <= ix && ix < ptab->r.o.max);
-    return erts_smp_atomic_read_acqb(&ptab->r.o.tab[ix]);
+    return erts_atomic_read_acqb(&ptab->r.o.tab[ix]);
 }
 
 ERTS_GLB_INLINE void erts_ptab_atmc_inc_refc(ErtsPTabElementCommon *ptab_el)
@@ -386,11 +388,9 @@ ERTS_GLB_INLINE void erts_ptab_atmc_inc_refc(ErtsPTabElementCommon *ptab_el)
 ERTS_GLB_INLINE Sint erts_ptab_atmc_dec_test_refc(ErtsPTabElementCommon *ptab_el)
 {
     erts_aint_t refc = erts_atomic_dec_read_relb(&ptab_el->refc.atmc);
-    ERTS_SMP_LC_ASSERT(refc >= 0);
-#ifdef ERTS_SMP
+    ERTS_LC_ASSERT(refc >= 0);
     if (refc == 0)
 	ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
-#endif
     return (Sint) refc;
 }
 
@@ -399,7 +399,7 @@ ERTS_GLB_INLINE Sint erts_ptab_atmc_add_test_refc(ErtsPTabElementCommon *ptab_el
 {
     erts_aint_t refc = erts_atomic_add_read_mb(&ptab_el->refc.atmc,
 					       (erts_aint_t) add_refc);
-    ERTS_SMP_LC_ASSERT(refc >= 0);
+    ERTS_LC_ASSERT(refc >= 0);
     return (Sint) refc;
 }
 
@@ -417,7 +417,7 @@ ERTS_GLB_INLINE void erts_ptab_inc_refc(ErtsPTabElementCommon *ptab_el)
 ERTS_GLB_INLINE Sint erts_ptab_dec_test_refc(ErtsPTabElementCommon *ptab_el)
 {
     Sint refc = --ptab_el->refc.sint;
-    ERTS_SMP_LC_ASSERT(refc >= 0);
+    ERTS_LC_ASSERT(refc >= 0);
     return refc;
 }
 
@@ -425,7 +425,7 @@ ERTS_GLB_INLINE Sint erts_ptab_add_test_refc(ErtsPTabElementCommon *ptab_el,
 					     Sint add_refc)
 {
     ptab_el->refc.sint += add_refc;
-    ERTS_SMP_LC_ASSERT(ptab_el->refc.sint >= 0);
+    ERTS_LC_ASSERT(ptab_el->refc.sint >= 0);
     return (Sint) ptab_el->refc.sint;
 }
 
@@ -436,42 +436,42 @@ ERTS_GLB_INLINE Sint erts_ptab_read_refc(ErtsPTabElementCommon *ptab_el)
 
 ERTS_GLB_INLINE void erts_ptab_rlock(ErtsPTab *ptab)
 {
-    erts_smp_rwmtx_rlock(&ptab->list.data.rwmtx);
+    erts_rwmtx_rlock(&ptab->list.data.rwmtx);
 }
 
 ERTS_GLB_INLINE int erts_ptab_tryrlock(ErtsPTab *ptab)
 {
-    return erts_smp_rwmtx_tryrlock(&ptab->list.data.rwmtx);
+    return erts_rwmtx_tryrlock(&ptab->list.data.rwmtx);
 }
 
 ERTS_GLB_INLINE void erts_ptab_runlock(ErtsPTab *ptab)
 {
-    erts_smp_rwmtx_runlock(&ptab->list.data.rwmtx);
+    erts_rwmtx_runlock(&ptab->list.data.rwmtx);
 }
 
 ERTS_GLB_INLINE void erts_ptab_rwlock(ErtsPTab *ptab)
 {
-    erts_smp_rwmtx_rwlock(&ptab->list.data.rwmtx);
+    erts_rwmtx_rwlock(&ptab->list.data.rwmtx);
 }
 
 ERTS_GLB_INLINE int erts_ptab_tryrwlock(ErtsPTab *ptab)
 {
-    return erts_smp_rwmtx_tryrwlock(&ptab->list.data.rwmtx);
+    return erts_rwmtx_tryrwlock(&ptab->list.data.rwmtx);
 }
 
 ERTS_GLB_INLINE void erts_ptab_rwunlock(ErtsPTab *ptab)
 {
-    erts_smp_rwmtx_rwunlock(&ptab->list.data.rwmtx);
+    erts_rwmtx_rwunlock(&ptab->list.data.rwmtx);
 }
 
-ERTS_GLB_INLINE int erts_smp_lc_ptab_is_rlocked(ErtsPTab *ptab)
+ERTS_GLB_INLINE int erts_lc_ptab_is_rlocked(ErtsPTab *ptab)
 {
-    return erts_smp_lc_rwmtx_is_rlocked(&ptab->list.data.rwmtx);
+    return erts_lc_rwmtx_is_rlocked(&ptab->list.data.rwmtx);
 }
 
-ERTS_GLB_INLINE int erts_smp_lc_ptab_is_rwlocked(ErtsPTab *ptab)
+ERTS_GLB_INLINE int erts_lc_ptab_is_rwlocked(ErtsPTab *ptab)
 {
-    return erts_smp_lc_rwmtx_is_rwlocked(&ptab->list.data.rwmtx);
+    return erts_lc_rwmtx_is_rwlocked(&ptab->list.data.rwmtx);
 }
 
 #endif
diff --git a/erts/emulator/beam/erl_rbtree.h b/erts/emulator/beam/erl_rbtree.h
index e59d6900b0..e50abf5cec 100644
--- a/erts/emulator/beam/erl_rbtree.h
+++ b/erts/emulator/beam/erl_rbtree.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2015-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2015-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -50,8 +50,14 @@
  * - ERTS_RBT_GET_LEFT(T) - Get left child node.
  * - ERTS_RBT_SET_LEFT(T, L) - Set left child node.
  * - ERTS_RBT_GET_KEY(T) - Get key of node.
- * - ERTS_RBT_IS_LT(KX, KY) - Is key KX less than key KY?
- * - ERTS_RBT_IS_EQ(KX, KY) - Is key KX equal to key KY?
+ * Either:
+ *   - ERTS_RBT_CMP_KEYS(KX, KY) - Compare keys...
+ * or:
+ *   - ERTS_RBT_IS_LT(KX, KY) - Is key KX less than key KY?
+ *   - ERTS_RBT_IS_EQ(KX, KY) - Is key KX equal to key KY?
+ *
+ * If ERTS_RBT_CMP_KEYS is defined ERTS_RBT_IS_LT and
+ * ERTS_RBT_IS_EQ will be redefined using ERTS_RBT_CMP_KEYS
  *
  * Optional defines:
  *
@@ -337,6 +343,15 @@
  *         Should only be used for debuging.
  */
 
+#ifdef ERTS_RBT_CMP_KEYS
+
+#  undef ERTS_RBT_IS_LT
+#  define ERTS_RBT_IS_LT(KX, KY) (ERTS_RBT_CMP_KEYS((KX), (KY)) < 0)
+
+#  undef ERTS_RBT_IS_EQ
+#  define ERTS_RBT_IS_EQ(KX, KY) (ERTS_RBT_CMP_KEYS((KX), (KY)) == 0)
+
+#endif
 
 /*
  * Check that we have all mandatory defines
@@ -396,6 +411,16 @@
 #  error Missing definition of ERTS_RBT_IS_EQ
 #endif
 
+#undef ERTS_RBT_IS_GT__
+#ifdef ERTS_RBT_CMP_KEYS
+#  define ERTS_RBT_IS_GT__(KX, KY) \
+    (ERTS_RBT_CMP_KEYS((KX), (KY)) > 0)
+#else
+#  define ERTS_RBT_IS_GT__(KX, KY) \
+    (!ERTS_RBT_IS_LT((KX), (KY)) && !ERTS_RBT_IS_EQ((KX), (KY)))
+
+#endif
+
 #if defined(ERTS_RBT_HARD_DEBUG) || defined(DEBUG)
 #  ifndef ERTS_RBT_DEBUG
 #    define ERTS_RBT_DEBUG 1
@@ -1007,19 +1032,30 @@ ERTS_RBT_FUNC__(insert_aux__)(ERTS_RBT_T **root, ERTS_RBT_T *n, int lookup)
 	ERTS_RBT_T *p, *x = *root;
 
 	while (1) {
-	    ERTS_RBT_KEY_T kx;
+	    ERTS_RBT_KEY_T kx = ERTS_RBT_GET_KEY(x);
 	    ERTS_RBT_T *c;
+            int kres;
+#ifdef ERTS_RBT_CMP_KEYS
+            int kcmp = ERTS_RBT_CMP_KEYS(kn, kx);
+            kres = kcmp == 0;
+#else
+            kres = ERTS_RBT_IS_EQ(kn, kx);
+#endif
 
-	    kx = ERTS_RBT_GET_KEY(x);
-
-	    if (lookup && ERTS_RBT_IS_EQ(kn, kx)) {
+	    if (lookup && kres) {
 
 		ERTS_RBT_HDBG_CHECK_TREE__(*root, NULL);
 
 		return x;
 	    }
 
-	    if (ERTS_RBT_IS_LT(kn, kx)) {
+#ifdef ERTS_RBT_CMP_KEYS
+            kres = kcmp < 0;
+#else
+            kres = ERTS_RBT_IS_LT(kn, kx);
+#endif
+
+	    if (kres) {
 		c = ERTS_RBT_GET_LEFT(x);
 		if (!c) {
 		    ERTS_RBT_SET_PARENT(n, x);
@@ -1075,6 +1111,101 @@ ERTS_RBT_FUNC__(insert)(ERTS_RBT_T **root, ERTS_RBT_T *n)
 
 #endif /* ERTS_RBT_WANT_INSERT */
 
+#ifdef ERTS_RBT_WANT_LOOKUP_CREATE
+static ERTS_INLINE ERTS_RBT_T *
+ERTS_RBT_FUNC__(lookup_create)(ERTS_RBT_T **root,
+                               ERTS_RBT_KEY_T kn,
+                               ERTS_RBT_T *(*create)(ERTS_RBT_KEY_T, void *),
+                               void *arg,
+                               int *created)
+{
+    ERTS_RBT_T *n;
+    ERTS_RBT_HDBG_CHECK_TREE__(*root, NULL);
+
+    if (!*root) {
+        n = (*create)(kn, arg);
+        ERTS_RBT_INIT_EMPTY_TNODE(n);
+        ERTS_RBT_ASSERT(ERTS_RBT_IS_EQ(ERTS_RBT_GET_KEY(n), kn));
+	ERTS_RBT_SET_BLACK(n);
+	*root = n;
+        *created = !0;
+#ifdef ERTS_RBT_UPDATE_ATTACHED_DATA_CHGROOT
+	ERTS_RBT_UPDATE_ATTACHED_DATA_CHGROOT(NULL, n);
+#endif
+    }
+    else {
+	ERTS_RBT_T *p, *x = *root;
+
+	while (1) {
+	    ERTS_RBT_KEY_T kx = ERTS_RBT_GET_KEY(x);
+	    ERTS_RBT_T *c;
+            int kres;
+#ifdef ERTS_RBT_CMP_KEYS
+            int kcmp = ERTS_RBT_CMP_KEYS(kn, kx);
+            kres = kcmp == 0;
+#else
+            kres = ERTS_RBT_IS_EQ(kn, kx);
+#endif
+
+	    if (kres) {
+
+		ERTS_RBT_HDBG_CHECK_TREE__(*root, NULL);
+
+                *created = 0;
+		return x;
+	    }
+
+#ifdef ERTS_RBT_CMP_KEYS
+            kres = kcmp < 0;
+#else
+            kres = ERTS_RBT_IS_LT(kn, kx);
+#endif
+
+	    if (kres) {
+		c = ERTS_RBT_GET_LEFT(x);
+		if (!c) {
+                    n = (*create)(kn, arg);
+                    ERTS_RBT_INIT_EMPTY_TNODE(n);
+                    ERTS_RBT_ASSERT(ERTS_RBT_IS_EQ(ERTS_RBT_GET_KEY(n), kn));
+                    *created = !0;
+		    ERTS_RBT_SET_PARENT(n, x);
+		    ERTS_RBT_SET_LEFT(x, n);
+		    p = x;
+		    break;
+		}
+	    }
+	    else {
+		c = ERTS_RBT_GET_RIGHT(x);
+		if (!c) {
+                    n = (*create)(kn, arg);
+                    ERTS_RBT_INIT_EMPTY_TNODE(n);
+                    ERTS_RBT_ASSERT(ERTS_RBT_IS_EQ(ERTS_RBT_GET_KEY(n), kn));
+                    *created = !0;
+		    ERTS_RBT_SET_PARENT(n, x);
+		    ERTS_RBT_SET_RIGHT(x, n);
+		    p = x;
+		    break;
+		}
+	    }
+
+	    x = c;
+	}
+
+        ERTS_RBT_ASSERT(ERTS_RBT_IS_EQ(ERTS_RBT_GET_KEY(n), kn));
+	ERTS_RBT_ASSERT(p);
+
+	ERTS_RBT_SET_RED(n);
+	if (ERTS_RBT_IS_RED(p))
+	    ERTS_RBT_FUNC__(insert_fixup__)(root, n);
+    }
+
+    ERTS_RBT_HDBG_CHECK_TREE__(*root, n);
+
+    return n;
+}
+
+#endif /* ERTS_RBT_WANT_LOOKUP_CREATE */
+
 #ifdef ERTS_RBT_WANT_LOOKUP
 
 static ERTS_RBT_API_INLINE__ ERTS_RBT_T *
@@ -1088,11 +1219,24 @@ ERTS_RBT_FUNC__(lookup)(ERTS_RBT_T *root, ERTS_RBT_KEY_T key)
     while (1) {
 	ERTS_RBT_KEY_T kx = ERTS_RBT_GET_KEY(x);
 	ERTS_RBT_T *c;
+        int kres;
+#ifdef ERTS_RBT_CMP_KEYS
+        int kcmp = ERTS_RBT_CMP_KEYS(key, kx);
+        kres = kcmp == 0;
+#else
+        kres = ERTS_RBT_IS_EQ(key, kx);
+#endif
 
-	if (ERTS_RBT_IS_EQ(key, kx))
+	if (kres)
 	    return x;
 
-	if (ERTS_RBT_IS_LT(key, kx)) {
+#ifdef ERTS_RBT_CMP_KEYS
+        kres = kcmp < 0;
+#else
+        kres = ERTS_RBT_IS_LT(key, kx);
+#endif
+
+	if (kres) {
 	    c = ERTS_RBT_GET_LEFT(x);
 	    if (!c)
 		return NULL;
@@ -1426,14 +1570,14 @@ ERTS_RBT_FUNC__(foreach_large)(ERTS_RBT_T *root,
 
 #ifdef ERTS_RBT_WANT_FOREACH_YIELDING
 
-static ERTS_RBT_API_INLINE__ void
+static ERTS_RBT_API_INLINE__ int
 ERTS_RBT_FUNC__(foreach_yielding)(ERTS_RBT_T *root,
 				  void (*op)(ERTS_RBT_T *, void *),
 				  void *arg,
 				  ERTS_RBT_YIELD_STATE_T__ *ystate,
 				  Sint ylimit)
 {
-    (void) ERTS_RBT_FUNC__(foreach_unordered__)(*root, 0, op, arg,
+    return ERTS_RBT_FUNC__(foreach_unordered__)(&root, 0, op, arg,
 						1, ystate, ylimit);
 }
 
@@ -1630,8 +1774,7 @@ ERTS_RBT_FUNC__(hdbg_check_tree)(ERTS_RBT_T *root, ERTS_RBT_T *n)
 		kx = ERTS_RBT_GET_KEY(x);
 		kc = ERTS_RBT_GET_KEY(c);
 
-		ERTS_RBT_ASSERT(ERTS_RBT_IS_LT(kc, kx)
-				|| ERTS_RBT_IS_EQ(kc, kx));
+                ERTS_RBT_ASSERT(!ERTS_RBT_IS_GT__(kc, kx));
 
 		x = c;
 	    }
@@ -1649,8 +1792,8 @@ ERTS_RBT_FUNC__(hdbg_check_tree)(ERTS_RBT_T *root, ERTS_RBT_T *n)
 	    kx = ERTS_RBT_GET_KEY(x);
 	    kc = ERTS_RBT_GET_KEY(c);
 
-	    ERTS_RBT_ASSERT(ERTS_RBT_IS_LT(kx, kc)
-			    || ERTS_RBT_IS_EQ(kx, kc));
+            ERTS_RBT_ASSERT(!ERTS_RBT_IS_GT__(kx, kc));
+
 	    x = c;
 	}
 
@@ -1672,8 +1815,8 @@ ERTS_RBT_FUNC__(hdbg_check_tree)(ERTS_RBT_T *root, ERTS_RBT_T *n)
 		    kx = ERTS_RBT_GET_KEY(x);
 		    kc = ERTS_RBT_GET_KEY(c);
 
-		    ERTS_RBT_ASSERT(ERTS_RBT_IS_LT(kx, kc)
-				    || ERTS_RBT_IS_EQ(kx, kc));
+                    ERTS_RBT_ASSERT(!ERTS_RBT_IS_GT__(kx, kc));
+
 		    /* Go down tree of x's sibling... */
 		    x = c;
 		    break;
@@ -1707,6 +1850,7 @@ ERTS_RBT_FUNC__(hdbg_check_tree)(ERTS_RBT_T *root, ERTS_RBT_T *n)
 #undef ERTS_RBT_NEED_FOREACH_ORDERED__
 #undef ERTS_RBT_NEED_HDBG_CHECK_TREE__
 #undef ERTS_RBT_HDBG_CHECK_TREE__
+#undef ERTS_RBT_IS_GT__
 
 #ifdef ERTS_RBT_UNDEF
 #  undef ERTS_RBT_PREFIX
@@ -1727,6 +1871,7 @@ ERTS_RBT_FUNC__(hdbg_check_tree)(ERTS_RBT_T *root, ERTS_RBT_T *n)
 #  undef ERTS_RBT_GET_LEFT
 #  undef ERTS_RBT_SET_LEFT
 #  undef ERTS_RBT_GET_KEY
+#  undef ERTS_RBT_CMP_KEYS
 #  undef ERTS_RBT_IS_LT
 #  undef ERTS_RBT_IS_EQ
 #  undef ERTS_RBT_UNDEF
diff --git a/erts/emulator/beam/erl_sched_spec_pre_alloc.c b/erts/emulator/beam/erl_sched_spec_pre_alloc.c
index 96238318c9..4a6e02281a 100644
--- a/erts/emulator/beam/erl_sched_spec_pre_alloc.c
+++ b/erts/emulator/beam/erl_sched_spec_pre_alloc.c
@@ -32,13 +32,12 @@
 #  include "config.h"
 #endif
 
-#ifdef ERTS_SMP
 
 #include "erl_process.h"
 #include "erl_thr_progress.h"
 
 erts_sspa_data_t *
-erts_sspa_create(size_t blk_sz, int pa_size)
+erts_sspa_create(size_t blk_sz, int pa_size, int nthreads, const char* name)
 {
     erts_sspa_data_t *data;
     size_t tot_size;
@@ -48,23 +47,40 @@ erts_sspa_create(size_t blk_sz, int pa_size)
     int cix;
     int no_blocks = pa_size;
     int no_blocks_per_chunk;
+    size_t aligned_blk_sz;
 
-    if (erts_no_schedulers == 1)
+#if !defined(ERTS_STRUCTURE_ALIGNED_ALLOC)
+    /* Force 64-bit alignment... */
+    aligned_blk_sz = ((blk_sz - 1) / 8) * 8 + 8;
+#else
+    /* Alignment of structure is enough... */
+    aligned_blk_sz = blk_sz;
+#endif
+
+    if (!name) { /* schedulers only variant */
+        ASSERT(!nthreads);
+        nthreads = erts_no_schedulers;
+    }
+    else {
+        ASSERT(nthreads > 0);
+    }
+
+    if (nthreads == 1)
 	no_blocks_per_chunk = no_blocks;
     else {
 	int extra = (no_blocks - 1)/4 + 1;
 	if (extra == 0)
 	    extra = 1;
 	no_blocks_per_chunk = no_blocks;
-	no_blocks_per_chunk += extra*erts_no_schedulers;
-	no_blocks_per_chunk /= erts_no_schedulers;
+	no_blocks_per_chunk += extra * nthreads;
+	no_blocks_per_chunk /= nthreads;
     }
-    no_blocks = no_blocks_per_chunk * erts_no_schedulers;
+    no_blocks = no_blocks_per_chunk * nthreads;
     chunk_mem_size = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_sspa_chunk_header_t));
-    chunk_mem_size += blk_sz * no_blocks_per_chunk;
+    chunk_mem_size += aligned_blk_sz * no_blocks_per_chunk;
     chunk_mem_size = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(chunk_mem_size);
     tot_size = ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_sspa_data_t));
-    tot_size += chunk_mem_size*erts_no_schedulers;
+    tot_size += chunk_mem_size * nthreads;
 
     p = erts_alloc_permanent_cache_aligned(ERTS_ALC_T_PRE_ALLOC_DATA, tot_size);
     data = (erts_sspa_data_t *) p;
@@ -73,10 +89,16 @@ erts_sspa_create(size_t blk_sz, int pa_size)
 
     data->chunks_mem_size = chunk_mem_size;
     data->start = chunk_start;
-    data->end = chunk_start + chunk_mem_size*erts_no_schedulers;
+    data->end = chunk_start + chunk_mem_size * nthreads;
+    data->nthreads = nthreads;
+
+    if (name) { /* thread variant */
+        erts_tsd_key_create(&data->tsd_key, (char*)name);
+        erts_atomic_init_nob(&data->id_generator, 0);
+    }
 
     /* Initialize all chunks */
-    for (cix = 0; cix < erts_no_schedulers; cix++) {
+    for (cix = 0; cix < nthreads; cix++) {
 	erts_sspa_chunk_t *chnk = erts_sspa_cix2chunk(data, cix);
 	erts_sspa_chunk_header_t *chdr = &chnk->aligned.header;
 	erts_sspa_blk_t *blk;
@@ -102,7 +124,7 @@ erts_sspa_create(size_t blk_sz, int pa_size)
 	blk = (erts_sspa_blk_t *) p;
 	for (i = 0; i < no_blocks_per_chunk; i++) {
 	    blk = (erts_sspa_blk_t *) p;
-	    p += blk_sz;
+	    p += aligned_blk_sz;
 	    blk->next_ptr = (erts_sspa_blk_t *) p;
 	}
 
@@ -325,4 +347,3 @@ erts_sspa_process_remote_frees(erts_sspa_chunk_header_t *chdr,
     return res;
 }
 
-#endif /* ERTS_SMP */
diff --git a/erts/emulator/beam/erl_sched_spec_pre_alloc.h b/erts/emulator/beam/erl_sched_spec_pre_alloc.h
index 7808d7d438..d232db0e69 100644
--- a/erts/emulator/beam/erl_sched_spec_pre_alloc.h
+++ b/erts/emulator/beam/erl_sched_spec_pre_alloc.h
@@ -31,7 +31,6 @@
 #ifndef ERTS_SCHED_SPEC_PRE_ALLOC_H__
 #define ERTS_SCHED_SPEC_PRE_ALLOC_H__
 
-#ifdef ERTS_SMP
 
 #undef ERL_THR_PROGRESS_TSD_TYPE_ONLY
 #define ERL_THR_PROGRESS_TSD_TYPE_ONLY
@@ -60,6 +59,11 @@ typedef struct {
     char *start;
     char *end;
     int chunks_mem_size;
+    int nthreads;
+
+    /* Used only by thread variant: */
+    erts_tsd_key_t tsd_key;
+    erts_atomic_t id_generator;
 } erts_sspa_data_t;
 
 typedef union erts_sspa_blk_t_ erts_sspa_blk_t;
@@ -141,7 +145,9 @@ check_local_list(erts_sspa_chunk_header_t *chdr)
 #endif
 
 erts_sspa_data_t *erts_sspa_create(size_t blk_sz,
-				   int pa_size);
+				   int pa_size,
+                                   int nthreads,
+                                   const char* name);
 void erts_sspa_remote_free(erts_sspa_chunk_header_t *chdr,
 			   erts_sspa_blk_t *blk,
 			   int cinit);
@@ -159,7 +165,7 @@ ERTS_GLB_INLINE int erts_sspa_free(erts_sspa_data_t *data, int cix, char *blk);
 ERTS_GLB_INLINE erts_sspa_chunk_t *
 erts_sspa_cix2chunk(erts_sspa_data_t *data, int cix)
 {
-    ASSERT(0 <= cix && cix < erts_no_schedulers);
+    ASSERT(0 <= cix && cix < data->nthreads);
     return (erts_sspa_chunk_t *) (data->start + cix*data->chunks_mem_size);
 }
 
@@ -172,7 +178,7 @@ erts_sspa_ptr2cix(erts_sspa_data_t *data, void *ptr)
 	return -1;
     diff = ((char *) ptr) - data->start;
     cix = (int) diff / data->chunks_mem_size;
-    ASSERT(0 <= cix && cix < erts_no_schedulers);
+    ASSERT(0 <= cix && cix < data->nthreads);
     return cix;
 }
 
@@ -236,6 +242,5 @@ erts_sspa_free(erts_sspa_data_t *data, int cix, char *cblk)
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-#endif /* ERTS_SMP */
 
 #endif /* ERTS_SCHED_SPEC_PRE_ALLOC_H__ */
diff --git a/erts/emulator/beam/erl_smp.h b/erts/emulator/beam/erl_smp.h
deleted file mode 100644
index 696bdbdaf1..0000000000
--- a/erts/emulator/beam/erl_smp.h
+++ /dev/null
@@ -1,1585 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 2005-2017. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-/*
- * SMP interface to ethread library.
- * This is essentially "sed s/erts_/erts_smp_/g < erl_threads.h > erl_smp.h",
- * plus changes to NOP operations when ERTS_SMP is disabled.
- * Author: Mikael Pettersson
- */
-#ifndef ERL_SMP_H
-#define ERL_SMP_H
-#include "erl_threads.h"
-
-#ifdef ERTS_ENABLE_LOCK_POSITION
-#define erts_smp_mtx_lock(L) erts_smp_mtx_lock_x(L, __FILE__, __LINE__)
-#define erts_smp_mtx_trylock(L) erts_smp_mtx_trylock_x(L, __FILE__, __LINE__)
-#define erts_smp_spin_lock(L) erts_smp_spin_lock_x(L, __FILE__, __LINE__)
-#define erts_smp_rwmtx_tryrlock(L) erts_smp_rwmtx_tryrlock_x(L, __FILE__, __LINE__)
-#define erts_smp_rwmtx_rlock(L) erts_smp_rwmtx_rlock_x(L, __FILE__, __LINE__)
-#define erts_smp_rwmtx_tryrwlock(L) erts_smp_rwmtx_tryrwlock_x(L, __FILE__, __LINE__)
-#define erts_smp_rwmtx_rwlock(L) erts_smp_rwmtx_rwlock_x(L, __FILE__, __LINE__)
-#define erts_smp_read_lock(L) erts_smp_read_lock_x(L, __FILE__, __LINE__)
-#define erts_smp_write_lock(L) erts_smp_write_lock_x(L, __FILE__, __LINE__)
-#endif
-
-
-#ifdef ERTS_SMP
-#define ERTS_SMP_THR_OPTS_DEFAULT_INITER ERTS_THR_OPTS_DEFAULT_INITER
-typedef erts_thr_opts_t erts_smp_thr_opts_t;
-typedef erts_thr_init_data_t erts_smp_thr_init_data_t;
-typedef erts_tid_t erts_smp_tid_t;
-typedef erts_mtx_t erts_smp_mtx_t;
-typedef erts_cnd_t erts_smp_cnd_t;
-#define ERTS_SMP_RWMTX_OPT_DEFAULT_INITER ERTS_RWMTX_OPT_DEFAULT_INITER
-#define ERTS_SMP_RWMTX_TYPE_NORMAL ERTS_RWMTX_TYPE_NORMAL
-#define ERTS_SMP_RWMTX_TYPE_FREQUENT_READ ERTS_RWMTX_TYPE_FREQUENT_READ
-#define ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ \
-  ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ
-#define ERTS_SMP_RWMTX_LONG_LIVED ERTS_RWMTX_LONG_LIVED
-#define ERTS_SMP_RWMTX_SHORT_LIVED ERTS_RWMTX_SHORT_LIVED
-#define ERTS_SMP_RWMTX_UNKNOWN_LIVED ERTS_RWMTX_UNKNOWN_LIVED
-typedef erts_rwmtx_opt_t erts_smp_rwmtx_opt_t;
-typedef erts_rwmtx_t erts_smp_rwmtx_t;
-typedef erts_tsd_key_t erts_smp_tsd_key_t;
-#define erts_smp_dw_atomic_t erts_dw_atomic_t
-#define erts_smp_atomic_t erts_atomic_t
-#define erts_smp_atomic32_t erts_atomic32_t
-#define erts_smp_atomic64_t erts_atomic64_t
-typedef erts_spinlock_t erts_smp_spinlock_t;
-typedef erts_rwlock_t erts_smp_rwlock_t;
-void erts_thr_fatal_error(int, char *); /* implemented in erl_init.c */
-
-#define ERTS_SMP_MEMORY_BARRIER ERTS_THR_MEMORY_BARRIER
-#define ERTS_SMP_WRITE_MEMORY_BARRIER ERTS_THR_WRITE_MEMORY_BARRIER
-#define ERTS_SMP_READ_MEMORY_BARRIER ERTS_THR_READ_MEMORY_BARRIER
-#define ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER
-
-#else /* #ifdef ERTS_SMP */
-
-#define ERTS_SMP_THR_OPTS_DEFAULT_INITER {0}
-typedef int erts_smp_thr_opts_t;
-typedef int erts_smp_thr_init_data_t;
-typedef int erts_smp_tid_t;
-typedef int erts_smp_mtx_t;
-typedef int erts_smp_cnd_t;
-#define ERTS_SMP_RWMTX_OPT_DEFAULT_INITER {0}
-#define ERTS_SMP_RWMTX_TYPE_NORMAL 0
-#define ERTS_SMP_RWMTX_TYPE_FREQUENT_READ 0
-#define ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ 0
-#define ERTS_SMP_RWMTX_LONG_LIVED 0
-#define ERTS_SMP_RWMTX_SHORT_LIVED 0
-#define ERTS_SMP_RWMTX_UNKNOWN_LIVED 0
-typedef struct {
-    char type;
-    char lived;
-    int main_spincount;
-    int aux_spincount;
-} erts_smp_rwmtx_opt_t;
-typedef int erts_smp_rwmtx_t;
-typedef int erts_smp_tsd_key_t;
-#define erts_smp_dw_atomic_t erts_no_dw_atomic_t
-#define erts_smp_atomic_t erts_no_atomic_t
-#define erts_smp_atomic32_t erts_no_atomic32_t
-#define erts_smp_atomic64_t erts_no_atomic64_t
-#if __GNUC__ > 2
-typedef struct { } erts_smp_spinlock_t;
-typedef struct { } erts_smp_rwlock_t;
-#else
-typedef struct { int gcc_is_buggy; } erts_smp_spinlock_t;
-typedef struct { int gcc_is_buggy; } erts_smp_rwlock_t;
-#endif
-
-#define ERTS_SMP_MEMORY_BARRIER
-#define ERTS_SMP_WRITE_MEMORY_BARRIER
-#define ERTS_SMP_READ_MEMORY_BARRIER
-#define ERTS_SMP_DATA_DEPENDENCY_READ_MEMORY_BARRIER
-
-#endif /* #ifdef ERTS_SMP */
-
-ERTS_GLB_INLINE void erts_smp_thr_init(erts_smp_thr_init_data_t *id);
-ERTS_GLB_INLINE void erts_smp_thr_create(erts_smp_tid_t *tid,
-					 void * (*func)(void *),
-					 void *arg,
-					 erts_smp_thr_opts_t *opts);
-ERTS_GLB_INLINE void erts_smp_thr_join(erts_smp_tid_t tid, void **thr_res);
-ERTS_GLB_INLINE void erts_smp_thr_detach(erts_smp_tid_t tid);
-ERTS_GLB_INLINE void erts_smp_thr_exit(void *res);
-ERTS_GLB_INLINE void erts_smp_install_exit_handler(void (*exit_handler)(void));
-ERTS_GLB_INLINE erts_smp_tid_t erts_smp_thr_self(void);
-ERTS_GLB_INLINE int erts_smp_equal_tids(erts_smp_tid_t x, erts_smp_tid_t y);
-#ifdef ERTS_HAVE_REC_MTX_INIT
-#define ERTS_SMP_HAVE_REC_MTX_INIT 1
-ERTS_GLB_INLINE void erts_smp_rec_mtx_init(erts_smp_mtx_t *mtx);
-#endif
-ERTS_GLB_INLINE void erts_smp_mtx_init(erts_smp_mtx_t *mtx,
-                                       char *name,
-                                       Eterm extra,
-                                       erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_mtx_init_locked(erts_smp_mtx_t *mtx,
-                                               char *name,
-                                               Eterm extra,
-                                               erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_mtx_destroy(erts_smp_mtx_t *mtx);
-#ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE int erts_smp_mtx_trylock_x(erts_smp_mtx_t *mtx, char *file, unsigned int line);
-ERTS_GLB_INLINE void erts_smp_mtx_lock_x(erts_smp_mtx_t *mtx, char *file, unsigned int line);
-#else
-ERTS_GLB_INLINE int erts_smp_mtx_trylock(erts_smp_mtx_t *mtx);
-ERTS_GLB_INLINE void erts_smp_mtx_lock(erts_smp_mtx_t *mtx);
-#endif
-ERTS_GLB_INLINE void erts_smp_mtx_unlock(erts_smp_mtx_t *mtx);
-ERTS_GLB_INLINE int erts_smp_lc_mtx_is_locked(erts_smp_mtx_t *mtx);
-ERTS_GLB_INLINE void erts_smp_cnd_init(erts_smp_cnd_t *cnd);
-ERTS_GLB_INLINE void erts_smp_cnd_destroy(erts_smp_cnd_t *cnd);
-ERTS_GLB_INLINE void erts_smp_cnd_wait(erts_smp_cnd_t *cnd,
-					   erts_smp_mtx_t *mtx);
-ERTS_GLB_INLINE void erts_smp_cnd_signal(erts_smp_cnd_t *cnd);
-ERTS_GLB_INLINE void erts_smp_cnd_broadcast(erts_smp_cnd_t *cnd);
-ERTS_GLB_INLINE void erts_smp_rwmtx_set_reader_group(int no);
-ERTS_GLB_INLINE void erts_smp_rwmtx_init_opt(erts_smp_rwmtx_t *rwmtx,
-                                             erts_smp_rwmtx_opt_t *opt,
-                                             char *name,
-                                             Eterm extra,
-                                             erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_rwmtx_init(erts_smp_rwmtx_t *rwmtx,
-                                         char *name,
-                                         Eterm extra,
-                                         erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_rwmtx_destroy(erts_smp_rwmtx_t *rwmtx);
-#ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE int erts_smp_rwmtx_tryrlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line);
-ERTS_GLB_INLINE void erts_smp_rwmtx_rlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line);
-ERTS_GLB_INLINE void erts_smp_rwmtx_rwlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line);
-ERTS_GLB_INLINE int erts_smp_rwmtx_tryrwlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line);
-#else
-ERTS_GLB_INLINE int erts_smp_rwmtx_tryrlock(erts_smp_rwmtx_t *rwmtx);
-ERTS_GLB_INLINE void erts_smp_rwmtx_rlock(erts_smp_rwmtx_t *rwmtx);
-ERTS_GLB_INLINE void erts_smp_rwmtx_rwlock(erts_smp_rwmtx_t *rwmtx);
-ERTS_GLB_INLINE int erts_smp_rwmtx_tryrwlock(erts_smp_rwmtx_t *rwmtx);
-#endif
-ERTS_GLB_INLINE void erts_smp_rwmtx_runlock(erts_smp_rwmtx_t *rwmtx);
-ERTS_GLB_INLINE void erts_smp_rwmtx_rwunlock(erts_smp_rwmtx_t *rwmtx);
-ERTS_GLB_INLINE int erts_smp_lc_rwmtx_is_rlocked(erts_smp_rwmtx_t *mtx);
-ERTS_GLB_INLINE int erts_smp_lc_rwmtx_is_rwlocked(erts_smp_rwmtx_t *mtx);
-ERTS_GLB_INLINE void erts_smp_spinlock_init(erts_smp_spinlock_t *lock,
-                                            char *name,
-                                            Eterm extra,
-                                            erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_spinlock_destroy(erts_smp_spinlock_t *lock);
-ERTS_GLB_INLINE void erts_smp_spin_unlock(erts_smp_spinlock_t *lock);
-#ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE void erts_smp_spin_lock_x(erts_smp_spinlock_t *lock, char *file, unsigned int line);
-#else
-ERTS_GLB_INLINE void erts_smp_spin_lock(erts_smp_spinlock_t *lock);
-#endif
-ERTS_GLB_INLINE int erts_smp_lc_spinlock_is_locked(erts_smp_spinlock_t *lock);
-ERTS_GLB_INLINE void erts_smp_rwlock_init(erts_smp_rwlock_t *lock,
-                                          char *name,
-                                          Eterm extra,
-                                          erts_lock_flags_t flags);
-ERTS_GLB_INLINE void erts_smp_rwlock_destroy(erts_smp_rwlock_t *lock);
-ERTS_GLB_INLINE void erts_smp_read_unlock(erts_smp_rwlock_t *lock);
-#ifdef ERTS_ENABLE_LOCK_POSITION
-ERTS_GLB_INLINE void erts_smp_read_lock_x(erts_smp_rwlock_t *lock, char *file, unsigned int line);
-ERTS_GLB_INLINE void erts_smp_write_lock_x(erts_smp_rwlock_t *lock, char *file, unsigned int line);
-#else
-ERTS_GLB_INLINE void erts_smp_read_lock(erts_smp_rwlock_t *lock);
-ERTS_GLB_INLINE void erts_smp_write_lock(erts_smp_rwlock_t *lock);
-#endif
-ERTS_GLB_INLINE void erts_smp_write_unlock(erts_smp_rwlock_t *lock);
-ERTS_GLB_INLINE int erts_smp_lc_rwlock_is_rlocked(erts_smp_rwlock_t *lock);
-ERTS_GLB_INLINE int erts_smp_lc_rwlock_is_rwlocked(erts_smp_rwlock_t *lock);
-ERTS_GLB_INLINE void erts_smp_tsd_key_create(erts_smp_tsd_key_t *keyp,
-					     char *keyname);
-ERTS_GLB_INLINE void erts_smp_tsd_key_delete(erts_smp_tsd_key_t key);
-ERTS_GLB_INLINE void erts_smp_tsd_set(erts_smp_tsd_key_t key, void *value);
-ERTS_GLB_INLINE void * erts_smp_tsd_get(erts_smp_tsd_key_t key);
-
-#ifdef ERTS_THR_HAVE_SIG_FUNCS
-#define ERTS_SMP_THR_HAVE_SIG_FUNCS 1
-ERTS_GLB_INLINE void erts_smp_thr_sigmask(int how,
-					  const sigset_t *set,
-					  sigset_t *oset);
-ERTS_GLB_INLINE void erts_smp_thr_sigwait(const sigset_t *set, int *sig);
-#endif /* #ifdef ERTS_THR_HAVE_SIG_FUNCS */
-
-/*
- * See "Documentation of atomics and memory barriers" at the top
- * of erl_threads.h for info on atomics.
- */
-
-#ifdef ERTS_SMP
-
-/* Double word size atomics */
-
-#define erts_smp_dw_atomic_init_nob erts_dw_atomic_init_nob
-#define erts_smp_dw_atomic_set_nob erts_dw_atomic_set_nob
-#define erts_smp_dw_atomic_read_nob erts_dw_atomic_read_nob
-#define erts_smp_dw_atomic_cmpxchg_nob erts_dw_atomic_cmpxchg_nob
-
-#define erts_smp_dw_atomic_init_mb erts_dw_atomic_init_mb
-#define erts_smp_dw_atomic_set_mb erts_dw_atomic_set_mb
-#define erts_smp_dw_atomic_read_mb erts_dw_atomic_read_mb
-#define erts_smp_dw_atomic_cmpxchg_mb erts_dw_atomic_cmpxchg_mb
-
-#define erts_smp_dw_atomic_init_acqb erts_dw_atomic_init_acqb
-#define erts_smp_dw_atomic_set_acqb erts_dw_atomic_set_acqb
-#define erts_smp_dw_atomic_read_acqb erts_dw_atomic_read_acqb
-#define erts_smp_dw_atomic_cmpxchg_acqb erts_dw_atomic_cmpxchg_acqb
-
-#define erts_smp_dw_atomic_init_relb erts_dw_atomic_init_relb
-#define erts_smp_dw_atomic_set_relb erts_dw_atomic_set_relb
-#define erts_smp_dw_atomic_read_relb erts_dw_atomic_read_relb
-#define erts_smp_dw_atomic_cmpxchg_relb erts_dw_atomic_cmpxchg_relb
-
-#define erts_smp_dw_atomic_init_ddrb erts_dw_atomic_init_ddrb
-#define erts_smp_dw_atomic_set_ddrb erts_dw_atomic_set_ddrb
-#define erts_smp_dw_atomic_read_ddrb erts_dw_atomic_read_ddrb
-#define erts_smp_dw_atomic_cmpxchg_ddrb erts_dw_atomic_cmpxchg_ddrb
-
-#define erts_smp_dw_atomic_init_rb erts_dw_atomic_init_rb
-#define erts_smp_dw_atomic_set_rb erts_dw_atomic_set_rb
-#define erts_smp_dw_atomic_read_rb erts_dw_atomic_read_rb
-#define erts_smp_dw_atomic_cmpxchg_rb erts_dw_atomic_cmpxchg_rb
-
-#define erts_smp_dw_atomic_init_wb erts_dw_atomic_init_wb
-#define erts_smp_dw_atomic_set_wb erts_dw_atomic_set_wb
-#define erts_smp_dw_atomic_read_wb erts_dw_atomic_read_wb
-#define erts_smp_dw_atomic_cmpxchg_wb erts_dw_atomic_cmpxchg_wb
-
-#define erts_smp_dw_atomic_set_dirty erts_dw_atomic_set_dirty
-#define erts_smp_dw_atomic_read_dirty erts_dw_atomic_read_dirty
-
-/* Word size atomics */
-
-#define erts_smp_atomic_init_nob erts_atomic_init_nob
-#define erts_smp_atomic_set_nob erts_atomic_set_nob
-#define erts_smp_atomic_read_nob erts_atomic_read_nob
-#define erts_smp_atomic_inc_read_nob erts_atomic_inc_read_nob
-#define erts_smp_atomic_dec_read_nob erts_atomic_dec_read_nob
-#define erts_smp_atomic_inc_nob erts_atomic_inc_nob
-#define erts_smp_atomic_dec_nob erts_atomic_dec_nob
-#define erts_smp_atomic_add_read_nob erts_atomic_add_read_nob
-#define erts_smp_atomic_add_nob erts_atomic_add_nob
-#define erts_smp_atomic_read_bor_nob erts_atomic_read_bor_nob
-#define erts_smp_atomic_read_band_nob erts_atomic_read_band_nob
-#define erts_smp_atomic_xchg_nob erts_atomic_xchg_nob
-#define erts_smp_atomic_cmpxchg_nob erts_atomic_cmpxchg_nob
-#define erts_smp_atomic_read_bset_nob erts_atomic_read_bset_nob
-
-#define erts_smp_atomic_init_mb erts_atomic_init_mb
-#define erts_smp_atomic_set_mb erts_atomic_set_mb
-#define erts_smp_atomic_read_mb erts_atomic_read_mb
-#define erts_smp_atomic_inc_read_mb erts_atomic_inc_read_mb
-#define erts_smp_atomic_dec_read_mb erts_atomic_dec_read_mb
-#define erts_smp_atomic_inc_mb erts_atomic_inc_mb
-#define erts_smp_atomic_dec_mb erts_atomic_dec_mb
-#define erts_smp_atomic_add_read_mb erts_atomic_add_read_mb
-#define erts_smp_atomic_add_mb erts_atomic_add_mb
-#define erts_smp_atomic_read_bor_mb erts_atomic_read_bor_mb
-#define erts_smp_atomic_read_band_mb erts_atomic_read_band_mb
-#define erts_smp_atomic_xchg_mb erts_atomic_xchg_mb
-#define erts_smp_atomic_cmpxchg_mb erts_atomic_cmpxchg_mb
-#define erts_smp_atomic_read_bset_mb erts_atomic_read_bset_mb
-
-#define erts_smp_atomic_init_acqb erts_atomic_init_acqb
-#define erts_smp_atomic_set_acqb erts_atomic_set_acqb
-#define erts_smp_atomic_read_acqb erts_atomic_read_acqb
-#define erts_smp_atomic_inc_read_acqb erts_atomic_inc_read_acqb
-#define erts_smp_atomic_dec_read_acqb erts_atomic_dec_read_acqb
-#define erts_smp_atomic_inc_acqb erts_atomic_inc_acqb
-#define erts_smp_atomic_dec_acqb erts_atomic_dec_acqb
-#define erts_smp_atomic_add_read_acqb erts_atomic_add_read_acqb
-#define erts_smp_atomic_add_acqb erts_atomic_add_acqb
-#define erts_smp_atomic_read_bor_acqb erts_atomic_read_bor_acqb
-#define erts_smp_atomic_read_band_acqb erts_atomic_read_band_acqb
-#define erts_smp_atomic_xchg_acqb erts_atomic_xchg_acqb
-#define erts_smp_atomic_cmpxchg_acqb erts_atomic_cmpxchg_acqb
-#define erts_smp_atomic_read_bset_acqb erts_atomic_read_bset_acqb
-
-#define erts_smp_atomic_init_relb erts_atomic_init_relb
-#define erts_smp_atomic_set_relb erts_atomic_set_relb
-#define erts_smp_atomic_read_relb erts_atomic_read_relb
-#define erts_smp_atomic_inc_read_relb erts_atomic_inc_read_relb
-#define erts_smp_atomic_dec_read_relb erts_atomic_dec_read_relb
-#define erts_smp_atomic_inc_relb erts_atomic_inc_relb
-#define erts_smp_atomic_dec_relb erts_atomic_dec_relb
-#define erts_smp_atomic_add_read_relb erts_atomic_add_read_relb
-#define erts_smp_atomic_add_relb erts_atomic_add_relb
-#define erts_smp_atomic_read_bor_relb erts_atomic_read_bor_relb
-#define erts_smp_atomic_read_band_relb erts_atomic_read_band_relb
-#define erts_smp_atomic_xchg_relb erts_atomic_xchg_relb
-#define erts_smp_atomic_cmpxchg_relb erts_atomic_cmpxchg_relb
-#define erts_smp_atomic_read_bset_relb erts_atomic_read_bset_relb
-
-#define erts_smp_atomic_init_ddrb erts_atomic_init_ddrb
-#define erts_smp_atomic_set_ddrb erts_atomic_set_ddrb
-#define erts_smp_atomic_read_ddrb erts_atomic_read_ddrb
-#define erts_smp_atomic_inc_read_ddrb erts_atomic_inc_read_ddrb
-#define erts_smp_atomic_dec_read_ddrb erts_atomic_dec_read_ddrb
-#define erts_smp_atomic_inc_ddrb erts_atomic_inc_ddrb
-#define erts_smp_atomic_dec_ddrb erts_atomic_dec_ddrb
-#define erts_smp_atomic_add_read_ddrb erts_atomic_add_read_ddrb
-#define erts_smp_atomic_add_ddrb erts_atomic_add_ddrb
-#define erts_smp_atomic_read_bor_ddrb erts_atomic_read_bor_ddrb
-#define erts_smp_atomic_read_band_ddrb erts_atomic_read_band_ddrb
-#define erts_smp_atomic_xchg_ddrb erts_atomic_xchg_ddrb
-#define erts_smp_atomic_cmpxchg_ddrb erts_atomic_cmpxchg_ddrb
-#define erts_smp_atomic_read_bset_ddrb erts_atomic_read_bset_ddrb
-
-#define erts_smp_atomic_init_rb erts_atomic_init_rb
-#define erts_smp_atomic_set_rb erts_atomic_set_rb
-#define erts_smp_atomic_read_rb erts_atomic_read_rb
-#define erts_smp_atomic_inc_read_rb erts_atomic_inc_read_rb
-#define erts_smp_atomic_dec_read_rb erts_atomic_dec_read_rb
-#define erts_smp_atomic_inc_rb erts_atomic_inc_rb
-#define erts_smp_atomic_dec_rb erts_atomic_dec_rb
-#define erts_smp_atomic_add_read_rb erts_atomic_add_read_rb
-#define erts_smp_atomic_add_rb erts_atomic_add_rb
-#define erts_smp_atomic_read_bor_rb erts_atomic_read_bor_rb
-#define erts_smp_atomic_read_band_rb erts_atomic_read_band_rb
-#define erts_smp_atomic_xchg_rb erts_atomic_xchg_rb
-#define erts_smp_atomic_cmpxchg_rb erts_atomic_cmpxchg_rb
-#define erts_smp_atomic_read_bset_rb erts_atomic_read_bset_rb
-
-#define erts_smp_atomic_init_wb erts_atomic_init_wb
-#define erts_smp_atomic_set_wb erts_atomic_set_wb
-#define erts_smp_atomic_read_wb erts_atomic_read_wb
-#define erts_smp_atomic_inc_read_wb erts_atomic_inc_read_wb
-#define erts_smp_atomic_dec_read_wb erts_atomic_dec_read_wb
-#define erts_smp_atomic_inc_wb erts_atomic_inc_wb
-#define erts_smp_atomic_dec_wb erts_atomic_dec_wb
-#define erts_smp_atomic_add_read_wb erts_atomic_add_read_wb
-#define erts_smp_atomic_add_wb erts_atomic_add_wb
-#define erts_smp_atomic_read_bor_wb erts_atomic_read_bor_wb
-#define erts_smp_atomic_read_band_wb erts_atomic_read_band_wb
-#define erts_smp_atomic_xchg_wb erts_atomic_xchg_wb
-#define erts_smp_atomic_cmpxchg_wb erts_atomic_cmpxchg_wb
-#define erts_smp_atomic_read_bset_wb erts_atomic_read_bset_wb
-
-#define erts_smp_atomic_set_dirty erts_atomic_set_dirty
-#define erts_smp_atomic_read_dirty erts_atomic_read_dirty
-
-/* 32-bit atomics */
-
-#define erts_smp_atomic32_init_nob erts_atomic32_init_nob
-#define erts_smp_atomic32_set_nob erts_atomic32_set_nob
-#define erts_smp_atomic32_read_nob erts_atomic32_read_nob
-#define erts_smp_atomic32_inc_read_nob erts_atomic32_inc_read_nob
-#define erts_smp_atomic32_dec_read_nob erts_atomic32_dec_read_nob
-#define erts_smp_atomic32_inc_nob erts_atomic32_inc_nob
-#define erts_smp_atomic32_dec_nob erts_atomic32_dec_nob
-#define erts_smp_atomic32_add_read_nob erts_atomic32_add_read_nob
-#define erts_smp_atomic32_add_nob erts_atomic32_add_nob
-#define erts_smp_atomic32_read_bor_nob erts_atomic32_read_bor_nob
-#define erts_smp_atomic32_read_band_nob erts_atomic32_read_band_nob
-#define erts_smp_atomic32_xchg_nob erts_atomic32_xchg_nob
-#define erts_smp_atomic32_cmpxchg_nob erts_atomic32_cmpxchg_nob
-#define erts_smp_atomic32_read_bset_nob erts_atomic32_read_bset_nob
-
-#define erts_smp_atomic32_init_mb erts_atomic32_init_mb
-#define erts_smp_atomic32_set_mb erts_atomic32_set_mb
-#define erts_smp_atomic32_read_mb erts_atomic32_read_mb
-#define erts_smp_atomic32_inc_read_mb erts_atomic32_inc_read_mb
-#define erts_smp_atomic32_dec_read_mb erts_atomic32_dec_read_mb
-#define erts_smp_atomic32_inc_mb erts_atomic32_inc_mb
-#define erts_smp_atomic32_dec_mb erts_atomic32_dec_mb
-#define erts_smp_atomic32_add_read_mb erts_atomic32_add_read_mb
-#define erts_smp_atomic32_add_mb erts_atomic32_add_mb
-#define erts_smp_atomic32_read_bor_mb erts_atomic32_read_bor_mb
-#define erts_smp_atomic32_read_band_mb erts_atomic32_read_band_mb
-#define erts_smp_atomic32_xchg_mb erts_atomic32_xchg_mb
-#define erts_smp_atomic32_cmpxchg_mb erts_atomic32_cmpxchg_mb
-#define erts_smp_atomic32_read_bset_mb erts_atomic32_read_bset_mb
-
-#define erts_smp_atomic32_init_acqb erts_atomic32_init_acqb
-#define erts_smp_atomic32_set_acqb erts_atomic32_set_acqb
-#define erts_smp_atomic32_read_acqb erts_atomic32_read_acqb
-#define erts_smp_atomic32_inc_read_acqb erts_atomic32_inc_read_acqb
-#define erts_smp_atomic32_dec_read_acqb erts_atomic32_dec_read_acqb
-#define erts_smp_atomic32_inc_acqb erts_atomic32_inc_acqb
-#define erts_smp_atomic32_dec_acqb erts_atomic32_dec_acqb
-#define erts_smp_atomic32_add_read_acqb erts_atomic32_add_read_acqb
-#define erts_smp_atomic32_add_acqb erts_atomic32_add_acqb
-#define erts_smp_atomic32_read_bor_acqb erts_atomic32_read_bor_acqb
-#define erts_smp_atomic32_read_band_acqb erts_atomic32_read_band_acqb
-#define erts_smp_atomic32_xchg_acqb erts_atomic32_xchg_acqb
-#define erts_smp_atomic32_cmpxchg_acqb erts_atomic32_cmpxchg_acqb
-#define erts_smp_atomic32_read_bset_acqb erts_atomic32_read_bset_acqb
-
-#define erts_smp_atomic32_init_relb erts_atomic32_init_relb
-#define erts_smp_atomic32_set_relb erts_atomic32_set_relb
-#define erts_smp_atomic32_read_relb erts_atomic32_read_relb
-#define erts_smp_atomic32_inc_read_relb erts_atomic32_inc_read_relb
-#define erts_smp_atomic32_dec_read_relb erts_atomic32_dec_read_relb
-#define erts_smp_atomic32_inc_relb erts_atomic32_inc_relb
-#define erts_smp_atomic32_dec_relb erts_atomic32_dec_relb
-#define erts_smp_atomic32_add_read_relb erts_atomic32_add_read_relb
-#define erts_smp_atomic32_add_relb erts_atomic32_add_relb
-#define erts_smp_atomic32_read_bor_relb erts_atomic32_read_bor_relb
-#define erts_smp_atomic32_read_band_relb erts_atomic32_read_band_relb
-#define erts_smp_atomic32_xchg_relb erts_atomic32_xchg_relb
-#define erts_smp_atomic32_cmpxchg_relb erts_atomic32_cmpxchg_relb
-#define erts_smp_atomic32_read_bset_relb erts_atomic32_read_bset_relb
-
-#define erts_smp_atomic32_init_ddrb erts_atomic32_init_ddrb
-#define erts_smp_atomic32_set_ddrb erts_atomic32_set_ddrb
-#define erts_smp_atomic32_read_ddrb erts_atomic32_read_ddrb
-#define erts_smp_atomic32_inc_read_ddrb erts_atomic32_inc_read_ddrb
-#define erts_smp_atomic32_dec_read_ddrb erts_atomic32_dec_read_ddrb
-#define erts_smp_atomic32_inc_ddrb erts_atomic32_inc_ddrb
-#define erts_smp_atomic32_dec_ddrb erts_atomic32_dec_ddrb
-#define erts_smp_atomic32_add_read_ddrb erts_atomic32_add_read_ddrb
-#define erts_smp_atomic32_add_ddrb erts_atomic32_add_ddrb
-#define erts_smp_atomic32_read_bor_ddrb erts_atomic32_read_bor_ddrb
-#define erts_smp_atomic32_read_band_ddrb erts_atomic32_read_band_ddrb
-#define erts_smp_atomic32_xchg_ddrb erts_atomic32_xchg_ddrb
-#define erts_smp_atomic32_cmpxchg_ddrb erts_atomic32_cmpxchg_ddrb
-#define erts_smp_atomic32_read_bset_ddrb erts_atomic32_read_bset_ddrb
-
-#define erts_smp_atomic32_init_rb erts_atomic32_init_rb
-#define erts_smp_atomic32_set_rb erts_atomic32_set_rb
-#define erts_smp_atomic32_read_rb erts_atomic32_read_rb
-#define erts_smp_atomic32_inc_read_rb erts_atomic32_inc_read_rb
-#define erts_smp_atomic32_dec_read_rb erts_atomic32_dec_read_rb
-#define erts_smp_atomic32_inc_rb erts_atomic32_inc_rb
-#define erts_smp_atomic32_dec_rb erts_atomic32_dec_rb
-#define erts_smp_atomic32_add_read_rb erts_atomic32_add_read_rb
-#define erts_smp_atomic32_add_rb erts_atomic32_add_rb
-#define erts_smp_atomic32_read_bor_rb erts_atomic32_read_bor_rb
-#define erts_smp_atomic32_read_band_rb erts_atomic32_read_band_rb
-#define erts_smp_atomic32_xchg_rb erts_atomic32_xchg_rb
-#define erts_smp_atomic32_cmpxchg_rb erts_atomic32_cmpxchg_rb
-#define erts_smp_atomic32_read_bset_rb erts_atomic32_read_bset_rb
-
-#define erts_smp_atomic32_init_wb erts_atomic32_init_wb
-#define erts_smp_atomic32_set_wb erts_atomic32_set_wb
-#define erts_smp_atomic32_read_wb erts_atomic32_read_wb
-#define erts_smp_atomic32_inc_read_wb erts_atomic32_inc_read_wb
-#define erts_smp_atomic32_dec_read_wb erts_atomic32_dec_read_wb
-#define erts_smp_atomic32_inc_wb erts_atomic32_inc_wb
-#define erts_smp_atomic32_dec_wb erts_atomic32_dec_wb
-#define erts_smp_atomic32_add_read_wb erts_atomic32_add_read_wb
-#define erts_smp_atomic32_add_wb erts_atomic32_add_wb
-#define erts_smp_atomic32_read_bor_wb erts_atomic32_read_bor_wb
-#define erts_smp_atomic32_read_band_wb erts_atomic32_read_band_wb
-#define erts_smp_atomic32_xchg_wb erts_atomic32_xchg_wb
-#define erts_smp_atomic32_cmpxchg_wb erts_atomic32_cmpxchg_wb
-#define erts_smp_atomic32_read_bset_wb erts_atomic32_read_bset_wb
-
-#define erts_smp_atomic32_set_dirty erts_atomic32_set_dirty
-#define erts_smp_atomic32_read_dirty erts_atomic32_read_dirty
-
-/* 64-bit atomics */
-
-#define erts_smp_atomic64_init_nob erts_atomic64_init_nob
-#define erts_smp_atomic64_set_nob erts_atomic64_set_nob
-#define erts_smp_atomic64_read_nob erts_atomic64_read_nob
-#define erts_smp_atomic64_inc_read_nob erts_atomic64_inc_read_nob
-#define erts_smp_atomic64_dec_read_nob erts_atomic64_dec_read_nob
-#define erts_smp_atomic64_inc_nob erts_atomic64_inc_nob
-#define erts_smp_atomic64_dec_nob erts_atomic64_dec_nob
-#define erts_smp_atomic64_add_read_nob erts_atomic64_add_read_nob
-#define erts_smp_atomic64_add_nob erts_atomic64_add_nob
-#define erts_smp_atomic64_read_bor_nob erts_atomic64_read_bor_nob
-#define erts_smp_atomic64_read_band_nob erts_atomic64_read_band_nob
-#define erts_smp_atomic64_xchg_nob erts_atomic64_xchg_nob
-#define erts_smp_atomic64_cmpxchg_nob erts_atomic64_cmpxchg_nob
-#define erts_smp_atomic64_read_bset_nob erts_atomic64_read_bset_nob
-
-#define erts_smp_atomic64_init_mb erts_atomic64_init_mb
-#define erts_smp_atomic64_set_mb erts_atomic64_set_mb
-#define erts_smp_atomic64_read_mb erts_atomic64_read_mb
-#define erts_smp_atomic64_inc_read_mb erts_atomic64_inc_read_mb
-#define erts_smp_atomic64_dec_read_mb erts_atomic64_dec_read_mb
-#define erts_smp_atomic64_inc_mb erts_atomic64_inc_mb
-#define erts_smp_atomic64_dec_mb erts_atomic64_dec_mb
-#define erts_smp_atomic64_add_read_mb erts_atomic64_add_read_mb
-#define erts_smp_atomic64_add_mb erts_atomic64_add_mb
-#define erts_smp_atomic64_read_bor_mb erts_atomic64_read_bor_mb
-#define erts_smp_atomic64_read_band_mb erts_atomic64_read_band_mb
-#define erts_smp_atomic64_xchg_mb erts_atomic64_xchg_mb
-#define erts_smp_atomic64_cmpxchg_mb erts_atomic64_cmpxchg_mb
-#define erts_smp_atomic64_read_bset_mb erts_atomic64_read_bset_mb
-
-#define erts_smp_atomic64_init_acqb erts_atomic64_init_acqb
-#define erts_smp_atomic64_set_acqb erts_atomic64_set_acqb
-#define erts_smp_atomic64_read_acqb erts_atomic64_read_acqb
-#define erts_smp_atomic64_inc_read_acqb erts_atomic64_inc_read_acqb
-#define erts_smp_atomic64_dec_read_acqb erts_atomic64_dec_read_acqb
-#define erts_smp_atomic64_inc_acqb erts_atomic64_inc_acqb
-#define erts_smp_atomic64_dec_acqb erts_atomic64_dec_acqb
-#define erts_smp_atomic64_add_read_acqb erts_atomic64_add_read_acqb
-#define erts_smp_atomic64_add_acqb erts_atomic64_add_acqb
-#define erts_smp_atomic64_read_bor_acqb erts_atomic64_read_bor_acqb
-#define erts_smp_atomic64_read_band_acqb erts_atomic64_read_band_acqb
-#define erts_smp_atomic64_xchg_acqb erts_atomic64_xchg_acqb
-#define erts_smp_atomic64_cmpxchg_acqb erts_atomic64_cmpxchg_acqb
-#define erts_smp_atomic64_read_bset_acqb erts_atomic64_read_bset_acqb
-
-#define erts_smp_atomic64_init_relb erts_atomic64_init_relb
-#define erts_smp_atomic64_set_relb erts_atomic64_set_relb
-#define erts_smp_atomic64_read_relb erts_atomic64_read_relb
-#define erts_smp_atomic64_inc_read_relb erts_atomic64_inc_read_relb
-#define erts_smp_atomic64_dec_read_relb erts_atomic64_dec_read_relb
-#define erts_smp_atomic64_inc_relb erts_atomic64_inc_relb
-#define erts_smp_atomic64_dec_relb erts_atomic64_dec_relb
-#define erts_smp_atomic64_add_read_relb erts_atomic64_add_read_relb
-#define erts_smp_atomic64_add_relb erts_atomic64_add_relb
-#define erts_smp_atomic64_read_bor_relb erts_atomic64_read_bor_relb
-#define erts_smp_atomic64_read_band_relb erts_atomic64_read_band_relb
-#define erts_smp_atomic64_xchg_relb erts_atomic64_xchg_relb
-#define erts_smp_atomic64_cmpxchg_relb erts_atomic64_cmpxchg_relb
-#define erts_smp_atomic64_read_bset_relb erts_atomic64_read_bset_relb
-
-#define erts_smp_atomic64_init_ddrb erts_atomic64_init_ddrb
-#define erts_smp_atomic64_set_ddrb erts_atomic64_set_ddrb
-#define erts_smp_atomic64_read_ddrb erts_atomic64_read_ddrb
-#define erts_smp_atomic64_inc_read_ddrb erts_atomic64_inc_read_ddrb
-#define erts_smp_atomic64_dec_read_ddrb erts_atomic64_dec_read_ddrb
-#define erts_smp_atomic64_inc_ddrb erts_atomic64_inc_ddrb
-#define erts_smp_atomic64_dec_ddrb erts_atomic64_dec_ddrb
-#define erts_smp_atomic64_add_read_ddrb erts_atomic64_add_read_ddrb
-#define erts_smp_atomic64_add_ddrb erts_atomic64_add_ddrb
-#define erts_smp_atomic64_read_bor_ddrb erts_atomic64_read_bor_ddrb
-#define erts_smp_atomic64_read_band_ddrb erts_atomic64_read_band_ddrb
-#define erts_smp_atomic64_xchg_ddrb erts_atomic64_xchg_ddrb
-#define erts_smp_atomic64_cmpxchg_ddrb erts_atomic64_cmpxchg_ddrb
-#define erts_smp_atomic64_read_bset_ddrb erts_atomic64_read_bset_ddrb
-
-#define erts_smp_atomic64_init_rb erts_atomic64_init_rb
-#define erts_smp_atomic64_set_rb erts_atomic64_set_rb
-#define erts_smp_atomic64_read_rb erts_atomic64_read_rb
-#define erts_smp_atomic64_inc_read_rb erts_atomic64_inc_read_rb
-#define erts_smp_atomic64_dec_read_rb erts_atomic64_dec_read_rb
-#define erts_smp_atomic64_inc_rb erts_atomic64_inc_rb
-#define erts_smp_atomic64_dec_rb erts_atomic64_dec_rb
-#define erts_smp_atomic64_add_read_rb erts_atomic64_add_read_rb
-#define erts_smp_atomic64_add_rb erts_atomic64_add_rb
-#define erts_smp_atomic64_read_bor_rb erts_atomic64_read_bor_rb
-#define erts_smp_atomic64_read_band_rb erts_atomic64_read_band_rb
-#define erts_smp_atomic64_xchg_rb erts_atomic64_xchg_rb
-#define erts_smp_atomic64_cmpxchg_rb erts_atomic64_cmpxchg_rb
-#define erts_smp_atomic64_read_bset_rb erts_atomic64_read_bset_rb
-
-#define erts_smp_atomic64_init_wb erts_atomic64_init_wb
-#define erts_smp_atomic64_set_wb erts_atomic64_set_wb
-#define erts_smp_atomic64_read_wb erts_atomic64_read_wb
-#define erts_smp_atomic64_inc_read_wb erts_atomic64_inc_read_wb
-#define erts_smp_atomic64_dec_read_wb erts_atomic64_dec_read_wb
-#define erts_smp_atomic64_inc_wb erts_atomic64_inc_wb
-#define erts_smp_atomic64_dec_wb erts_atomic64_dec_wb
-#define erts_smp_atomic64_add_read_wb erts_atomic64_add_read_wb
-#define erts_smp_atomic64_add_wb erts_atomic64_add_wb
-#define erts_smp_atomic64_read_bor_wb erts_atomic64_read_bor_wb
-#define erts_smp_atomic64_read_band_wb erts_atomic64_read_band_wb
-#define erts_smp_atomic64_xchg_wb erts_atomic64_xchg_wb
-#define erts_smp_atomic64_cmpxchg_wb erts_atomic64_cmpxchg_wb
-#define erts_smp_atomic64_read_bset_wb erts_atomic64_read_bset_wb
-
-#define erts_smp_atomic64_set_dirty erts_atomic64_set_dirty
-#define erts_smp_atomic64_read_dirty erts_atomic64_read_dirty
-
-#else /* !ERTS_SMP */
-
-/* Double word size atomics */
-
-#define erts_smp_dw_atomic_init_nob erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_set_nob erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_nob erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_nob erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_mb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_mb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_mb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_mb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_acqb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_acqb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_acqb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_acqb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_relb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_relb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_relb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_relb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_ddrb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_ddrb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_ddrb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_ddrb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_rb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_rb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_rb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_rb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_init_wb erts_no_dw_atomic_init
-#define erts_smp_dw_atomic_set_wb erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_wb erts_no_dw_atomic_read
-#define erts_smp_dw_atomic_cmpxchg_wb erts_no_dw_atomic_cmpxchg
-
-#define erts_smp_dw_atomic_set_dirty erts_no_dw_atomic_set
-#define erts_smp_dw_atomic_read_dirty erts_no_dw_atomic_read
-
-/* Word size atomics */
-
-#define erts_smp_atomic_init_nob erts_no_atomic_set
-#define erts_smp_atomic_set_nob erts_no_atomic_set
-#define erts_smp_atomic_read_nob erts_no_atomic_read
-#define erts_smp_atomic_inc_read_nob erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_nob erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_nob erts_no_atomic_inc
-#define erts_smp_atomic_dec_nob erts_no_atomic_dec
-#define erts_smp_atomic_add_read_nob erts_no_atomic_add_read
-#define erts_smp_atomic_add_nob erts_no_atomic_add
-#define erts_smp_atomic_read_bor_nob erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_nob erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_nob erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_nob erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_nob erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_mb erts_no_atomic_set
-#define erts_smp_atomic_set_mb erts_no_atomic_set
-#define erts_smp_atomic_read_mb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_mb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_mb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_mb erts_no_atomic_inc
-#define erts_smp_atomic_dec_mb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_mb erts_no_atomic_add_read
-#define erts_smp_atomic_add_mb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_mb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_mb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_mb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_mb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_mb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_acqb erts_no_atomic_set
-#define erts_smp_atomic_set_acqb erts_no_atomic_set
-#define erts_smp_atomic_read_acqb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_acqb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_acqb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_acqb erts_no_atomic_inc
-#define erts_smp_atomic_dec_acqb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_acqb erts_no_atomic_add_read
-#define erts_smp_atomic_add_acqb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_acqb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_acqb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_acqb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_acqb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_acqb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_relb erts_no_atomic_set
-#define erts_smp_atomic_set_relb erts_no_atomic_set
-#define erts_smp_atomic_read_relb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_relb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_relb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_relb erts_no_atomic_inc
-#define erts_smp_atomic_dec_relb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_relb erts_no_atomic_add_read
-#define erts_smp_atomic_add_relb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_relb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_relb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_relb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_relb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_relb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_ddrb erts_no_atomic_set
-#define erts_smp_atomic_set_ddrb erts_no_atomic_set
-#define erts_smp_atomic_read_ddrb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_ddrb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_ddrb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_ddrb erts_no_atomic_inc
-#define erts_smp_atomic_dec_ddrb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_ddrb erts_no_atomic_add_read
-#define erts_smp_atomic_add_ddrb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_ddrb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_ddrb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_ddrb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_ddrb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_ddrb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_rb erts_no_atomic_set
-#define erts_smp_atomic_set_rb erts_no_atomic_set
-#define erts_smp_atomic_read_rb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_rb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_rb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_rb erts_no_atomic_inc
-#define erts_smp_atomic_dec_rb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_rb erts_no_atomic_add_read
-#define erts_smp_atomic_add_rb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_rb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_rb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_rb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_rb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_rb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_init_wb erts_no_atomic_set
-#define erts_smp_atomic_set_wb erts_no_atomic_set
-#define erts_smp_atomic_read_wb erts_no_atomic_read
-#define erts_smp_atomic_inc_read_wb erts_no_atomic_inc_read
-#define erts_smp_atomic_dec_read_wb erts_no_atomic_dec_read
-#define erts_smp_atomic_inc_wb erts_no_atomic_inc
-#define erts_smp_atomic_dec_wb erts_no_atomic_dec
-#define erts_smp_atomic_add_read_wb erts_no_atomic_add_read
-#define erts_smp_atomic_add_wb erts_no_atomic_add
-#define erts_smp_atomic_read_bor_wb erts_no_atomic_read_bor
-#define erts_smp_atomic_read_band_wb erts_no_atomic_read_band
-#define erts_smp_atomic_xchg_wb erts_no_atomic_xchg
-#define erts_smp_atomic_cmpxchg_wb erts_no_atomic_cmpxchg
-#define erts_smp_atomic_read_bset_wb erts_no_atomic_read_bset
-
-#define erts_smp_atomic_set_dirty erts_no_atomic_set
-#define erts_smp_atomic_read_dirty erts_no_atomic_read
-
-/* 32-bit atomics */
-
-#define erts_smp_atomic32_init_nob erts_no_atomic32_set
-#define erts_smp_atomic32_set_nob erts_no_atomic32_set
-#define erts_smp_atomic32_read_nob erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_nob erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_nob erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_nob erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_nob erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_nob erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_nob erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_nob erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_nob erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_nob erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_nob erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_nob erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_mb erts_no_atomic32_set
-#define erts_smp_atomic32_set_mb erts_no_atomic32_set
-#define erts_smp_atomic32_read_mb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_mb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_mb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_mb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_mb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_mb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_mb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_mb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_mb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_mb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_mb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_mb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_acqb erts_no_atomic32_set
-#define erts_smp_atomic32_set_acqb erts_no_atomic32_set
-#define erts_smp_atomic32_read_acqb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_acqb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_acqb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_acqb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_acqb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_acqb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_acqb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_acqb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_acqb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_acqb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_acqb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_acqb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_relb erts_no_atomic32_set
-#define erts_smp_atomic32_set_relb erts_no_atomic32_set
-#define erts_smp_atomic32_read_relb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_relb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_relb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_relb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_relb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_relb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_relb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_relb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_relb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_relb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_relb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_relb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_ddrb erts_no_atomic32_set
-#define erts_smp_atomic32_set_ddrb erts_no_atomic32_set
-#define erts_smp_atomic32_read_ddrb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_ddrb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_ddrb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_ddrb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_ddrb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_ddrb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_ddrb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_ddrb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_ddrb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_ddrb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_ddrb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_ddrb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_rb erts_no_atomic32_set
-#define erts_smp_atomic32_set_rb erts_no_atomic32_set
-#define erts_smp_atomic32_read_rb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_rb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_rb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_rb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_rb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_rb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_rb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_rb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_rb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_rb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_rb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_rb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_init_wb erts_no_atomic32_set
-#define erts_smp_atomic32_set_wb erts_no_atomic32_set
-#define erts_smp_atomic32_read_wb erts_no_atomic32_read
-#define erts_smp_atomic32_inc_read_wb erts_no_atomic32_inc_read
-#define erts_smp_atomic32_dec_read_wb erts_no_atomic32_dec_read
-#define erts_smp_atomic32_inc_wb erts_no_atomic32_inc
-#define erts_smp_atomic32_dec_wb erts_no_atomic32_dec
-#define erts_smp_atomic32_add_read_wb erts_no_atomic32_add_read
-#define erts_smp_atomic32_add_wb erts_no_atomic32_add
-#define erts_smp_atomic32_read_bor_wb erts_no_atomic32_read_bor
-#define erts_smp_atomic32_read_band_wb erts_no_atomic32_read_band
-#define erts_smp_atomic32_xchg_wb erts_no_atomic32_xchg
-#define erts_smp_atomic32_cmpxchg_wb erts_no_atomic32_cmpxchg
-#define erts_smp_atomic32_read_bset_wb erts_no_atomic32_read_bset
-
-#define erts_smp_atomic32_set_dirty erts_no_atomic32_set
-#define erts_smp_atomic32_read_dirty erts_no_atomic32_read
-
-/* 64-bit atomics */
-
-#define erts_smp_atomic64_init_nob erts_no_atomic64_set
-#define erts_smp_atomic64_set_nob erts_no_atomic64_set
-#define erts_smp_atomic64_read_nob erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_nob erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_nob erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_nob erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_nob erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_nob erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_nob erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_nob erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_nob erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_nob erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_nob erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_nob erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_mb erts_no_atomic64_set
-#define erts_smp_atomic64_set_mb erts_no_atomic64_set
-#define erts_smp_atomic64_read_mb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_mb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_mb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_mb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_mb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_mb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_mb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_mb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_mb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_mb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_mb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_mb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_acqb erts_no_atomic64_set
-#define erts_smp_atomic64_set_acqb erts_no_atomic64_set
-#define erts_smp_atomic64_read_acqb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_acqb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_acqb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_acqb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_acqb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_acqb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_acqb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_acqb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_acqb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_acqb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_acqb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_acqb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_relb erts_no_atomic64_set
-#define erts_smp_atomic64_set_relb erts_no_atomic64_set
-#define erts_smp_atomic64_read_relb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_relb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_relb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_relb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_relb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_relb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_relb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_relb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_relb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_relb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_relb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_relb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_ddrb erts_no_atomic64_set
-#define erts_smp_atomic64_set_ddrb erts_no_atomic64_set
-#define erts_smp_atomic64_read_ddrb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_ddrb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_ddrb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_ddrb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_ddrb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_ddrb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_ddrb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_ddrb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_ddrb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_ddrb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_ddrb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_ddrb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_rb erts_no_atomic64_set
-#define erts_smp_atomic64_set_rb erts_no_atomic64_set
-#define erts_smp_atomic64_read_rb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_rb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_rb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_rb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_rb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_rb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_rb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_rb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_rb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_rb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_rb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_rb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_init_wb erts_no_atomic64_set
-#define erts_smp_atomic64_set_wb erts_no_atomic64_set
-#define erts_smp_atomic64_read_wb erts_no_atomic64_read
-#define erts_smp_atomic64_inc_read_wb erts_no_atomic64_inc_read
-#define erts_smp_atomic64_dec_read_wb erts_no_atomic64_dec_read
-#define erts_smp_atomic64_inc_wb erts_no_atomic64_inc
-#define erts_smp_atomic64_dec_wb erts_no_atomic64_dec
-#define erts_smp_atomic64_add_read_wb erts_no_atomic64_add_read
-#define erts_smp_atomic64_add_wb erts_no_atomic64_add
-#define erts_smp_atomic64_read_bor_wb erts_no_atomic64_read_bor
-#define erts_smp_atomic64_read_band_wb erts_no_atomic64_read_band
-#define erts_smp_atomic64_xchg_wb erts_no_atomic64_xchg
-#define erts_smp_atomic64_cmpxchg_wb erts_no_atomic64_cmpxchg
-#define erts_smp_atomic64_read_bset_wb erts_no_atomic64_read_bset
-
-#define erts_smp_atomic64_set_dirty erts_no_atomic64_set
-#define erts_smp_atomic64_read_dirty erts_no_atomic64_read
-
-#endif /* !ERTS_SMP */
-
-#if ERTS_GLB_INLINE_INCL_FUNC_DEF
-
-ERTS_GLB_INLINE void
-erts_smp_thr_init(erts_smp_thr_init_data_t *id)
-{
-#ifdef ERTS_SMP
-    erts_thr_init(id);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_thr_create(erts_smp_tid_t *tid, void * (*func)(void *), void *arg,
-		    erts_smp_thr_opts_t *opts)
-{
-#ifdef ERTS_SMP
-    erts_thr_create(tid, func, arg, opts);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_thr_join(erts_smp_tid_t tid, void **thr_res)
-{
-#ifdef ERTS_SMP
-    erts_thr_join(tid, thr_res);
-#endif
-}
-
-
-ERTS_GLB_INLINE void
-erts_smp_thr_detach(erts_smp_tid_t tid)
-{
-#ifdef ERTS_SMP
-    erts_thr_detach(tid);
-#endif
-}
-
-
-ERTS_GLB_INLINE void
-erts_smp_thr_exit(void *res)
-{
-#ifdef ERTS_SMP
-    erts_thr_exit(res);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_install_exit_handler(void (*exit_handler)(void))
-{
-#ifdef ERTS_SMP
-    erts_thr_install_exit_handler(exit_handler);
-#endif
-}
-
-ERTS_GLB_INLINE erts_smp_tid_t
-erts_smp_thr_self(void)
-{
-#ifdef ERTS_SMP
-    return erts_thr_self();
-#else
-    return 0;
-#endif
-}
-
-
-ERTS_GLB_INLINE int
-erts_smp_equal_tids(erts_smp_tid_t x, erts_smp_tid_t y)
-{
-#ifdef ERTS_SMP
-    return erts_equal_tids(x, y);
-#else
-    return 1;
-#endif
-}
-
-
-#ifdef ERTS_HAVE_REC_MTX_INIT
-ERTS_GLB_INLINE void
-erts_smp_rec_mtx_init(erts_smp_mtx_t *mtx)
-{
-#ifdef ERTS_SMP
-    erts_rec_mtx_init(mtx);
-#endif
-}
-#endif
-
-ERTS_GLB_INLINE void
-erts_smp_mtx_init(erts_smp_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_mtx_init(mtx, name, extra, flags);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_mtx_init_locked(erts_smp_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_mtx_init_locked(mtx, name, extra, flags);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_mtx_destroy(erts_smp_mtx_t *mtx)
-{
-#ifdef ERTS_SMP
-    erts_mtx_destroy(mtx);
-#endif
-}
-
-ERTS_GLB_INLINE int
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_mtx_trylock_x(erts_smp_mtx_t *mtx, char *file, unsigned int line)
-#else
-erts_smp_mtx_trylock(erts_smp_mtx_t *mtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    return erts_mtx_trylock_x(mtx,file,line);
-#elif defined(ERTS_SMP)
-    return erts_mtx_trylock(mtx);
-#else
-    return 0;
-#endif
-
-}
-
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_mtx_lock_x(erts_smp_mtx_t *mtx, char *file, unsigned int line)
-#else
-erts_smp_mtx_lock(erts_smp_mtx_t *mtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_mtx_lock_x(mtx, file, line);
-#elif defined(ERTS_SMP)
-    erts_mtx_lock(mtx);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_mtx_unlock(erts_smp_mtx_t *mtx)
-{
-#ifdef ERTS_SMP
-    erts_mtx_unlock(mtx);
-#endif
-}
-
-ERTS_GLB_INLINE int
-erts_smp_lc_mtx_is_locked(erts_smp_mtx_t *mtx)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_mtx_is_locked(mtx);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_cnd_init(erts_smp_cnd_t *cnd)
-{
-#ifdef ERTS_SMP
-    erts_cnd_init(cnd);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_cnd_destroy(erts_smp_cnd_t *cnd)
-{
-#ifdef ERTS_SMP
-    erts_cnd_destroy(cnd);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_cnd_wait(erts_smp_cnd_t *cnd, erts_smp_mtx_t *mtx)
-{
-#ifdef ERTS_SMP
-    erts_cnd_wait(cnd, mtx);
-#endif
-}
-
-/*
- * IMPORTANT note about erts_smp_cnd_signal() and erts_smp_cnd_broadcast()
- *
- * POSIX allow a call to `pthread_cond_signal' or `pthread_cond_broadcast'
- * even though the associated mutex/mutexes isn't/aren't locked by the
- * caller. Our implementation do not allow that in order to avoid a
- * performance penalty. That is, all associated mutexes *need* to be
- * locked by the caller of erts_smp_cnd_signal()/erts_smp_cnd_broadcast()!
- */
-
-ERTS_GLB_INLINE void
-erts_smp_cnd_signal(erts_smp_cnd_t *cnd)
-{
-#ifdef ERTS_SMP
-    erts_cnd_signal(cnd);
-#endif
-}
-
-
-ERTS_GLB_INLINE void
-erts_smp_cnd_broadcast(erts_smp_cnd_t *cnd)
-{
-#ifdef ERTS_SMP
-    erts_cnd_broadcast(cnd);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_set_reader_group(int no)
-{
-#ifdef ERTS_SMP
-    erts_rwmtx_set_reader_group(no);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_init(erts_smp_rwmtx_t *rwmtx,
-                    char *name,
-                    Eterm extra,
-                    erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_smp_rwmtx_init_opt(rwmtx, NULL, name, extra, flags);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_init_opt(erts_smp_rwmtx_t *rwmtx,
-                        erts_smp_rwmtx_opt_t *opt,
-                        char *name,
-                        Eterm extra,
-                        erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_rwmtx_init_opt(rwmtx, opt, name, extra, flags);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_destroy(erts_smp_rwmtx_t *rwmtx)
-{
-#ifdef ERTS_SMP
-    erts_rwmtx_destroy(rwmtx);
-#endif
-}
-
-ERTS_GLB_INLINE int
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_rwmtx_tryrlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line)
-#else
-erts_smp_rwmtx_tryrlock(erts_smp_rwmtx_t *rwmtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    return erts_rwmtx_tryrlock_x(rwmtx, file, line);
-#elif defined(ERTS_SMP)
-    return erts_rwmtx_tryrlock(rwmtx);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_rwmtx_rlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line)
-#else
-erts_smp_rwmtx_rlock(erts_smp_rwmtx_t *rwmtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_rwmtx_rlock_x(rwmtx, file, line);
-#elif defined(ERTS_SMP)
-    erts_rwmtx_rlock(rwmtx);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_runlock(erts_smp_rwmtx_t *rwmtx)
-{
-#ifdef ERTS_SMP
-    erts_rwmtx_runlock(rwmtx);
-#endif
-}
-
-
-ERTS_GLB_INLINE int
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_rwmtx_tryrwlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line)
-#else
-erts_smp_rwmtx_tryrwlock(erts_smp_rwmtx_t *rwmtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    return erts_rwmtx_tryrwlock_x(rwmtx, file, line);
-#elif defined(ERTS_SMP)
-    return erts_rwmtx_tryrwlock(rwmtx);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_rwmtx_rwlock_x(erts_smp_rwmtx_t *rwmtx, char *file, unsigned int line)
-#else
-erts_smp_rwmtx_rwlock(erts_smp_rwmtx_t *rwmtx)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_rwmtx_rwlock_x(rwmtx, file, line);
-#elif defined(ERTS_SMP)
-    erts_rwmtx_rwlock(rwmtx);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_rwunlock(erts_smp_rwmtx_t *rwmtx)
-{
-#ifdef ERTS_SMP
-    erts_rwmtx_rwunlock(rwmtx);
-#endif
-}
-
-#if 0 /* The following rwmtx function names are
-	 reserved for potential future use. */ 
-
-/* Try upgrade from r-locked state to rw-locked state */
-ERTS_GLB_INLINE int
-erts_smp_rwmtx_trywlock(erts_smp_rwmtx_t *rwmtx)
-{
-    return 0;
-}
-
-/* Upgrade from r-locked state to rw-locked state */
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_wlock(erts_smp_rwmtx_t *rwmtx)
-{
-
-}
-
-/* Downgrade from rw-locked state to r-locked state */
-ERTS_GLB_INLINE void
-erts_smp_rwmtx_wunlock(erts_smp_rwmtx_t *rwmtx)
-{
-
-}
-
-#endif
-
-ERTS_GLB_INLINE int
-erts_smp_lc_rwmtx_is_rlocked(erts_smp_rwmtx_t *mtx)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_rwmtx_is_rlocked(mtx);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE int
-erts_smp_lc_rwmtx_is_rwlocked(erts_smp_rwmtx_t *mtx)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_rwmtx_is_rwlocked(mtx);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_spinlock_init(erts_smp_spinlock_t *lock, char *name, Eterm extra, erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_spinlock_init(lock, name, extra, flags);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_spinlock_destroy(erts_smp_spinlock_t *lock)
-{
-#ifdef ERTS_SMP
-    erts_spinlock_destroy(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_spin_unlock(erts_smp_spinlock_t *lock)
-{
-#ifdef ERTS_SMP
-    erts_spin_unlock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_spin_lock_x(erts_smp_spinlock_t *lock, char *file, unsigned int line)
-#else
-erts_smp_spin_lock(erts_smp_spinlock_t *lock)
-#endif
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_spin_lock_x(lock, file, line);
-#elif defined(ERTS_SMP)
-    erts_spin_lock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE int
-erts_smp_lc_spinlock_is_locked(erts_smp_spinlock_t *lock)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_spinlock_is_locked(lock);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwlock_init(erts_smp_rwlock_t *lock, char *name, Eterm extra, erts_lock_flags_t flags)
-{
-#ifdef ERTS_SMP
-    erts_rwlock_init(lock, name, extra, flags);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_rwlock_destroy(erts_smp_rwlock_t *lock)
-{
-#ifdef ERTS_SMP
-    erts_rwlock_destroy(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_read_unlock(erts_smp_rwlock_t *lock)
-{
-#ifdef ERTS_SMP
-    erts_read_unlock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_read_lock_x(erts_smp_rwlock_t *lock, char *file, unsigned int line)
-#else
-erts_smp_read_lock(erts_smp_rwlock_t *lock)
-#endif 
-{
-#if defined(ERTS_ENABLE_LOCK_POSITION) && defined(ERTS_SMP)
-    erts_read_lock_x(lock, file, line);
-#elif defined(ERTS_SMP)
-    erts_read_lock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_write_unlock(erts_smp_rwlock_t *lock)
-{
-#ifdef ERTS_SMP
-    erts_write_unlock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE void
-#ifdef ERTS_ENABLE_LOCK_POSITION
-erts_smp_write_lock_x(erts_smp_rwlock_t *lock, char *file, unsigned int line)
-#else
-erts_smp_write_lock(erts_smp_rwlock_t *lock)
-#endif 
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_POSITION)
-    erts_write_lock_x(lock, file, line);
-#elif defined(ERTS_SMP)
-    erts_write_lock(lock);
-#else
-    (void)lock;
-#endif
-}
-
-ERTS_GLB_INLINE int
-erts_smp_lc_rwlock_is_rlocked(erts_smp_rwlock_t *lock)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_rwlock_is_rlocked(lock);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE int
-erts_smp_lc_rwlock_is_rwlocked(erts_smp_rwlock_t *lock)
-{
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    return erts_lc_rwlock_is_rwlocked(lock);
-#else
-    return 0;
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_tsd_key_create(erts_smp_tsd_key_t *keyp, char* keyname)
-{
-#ifdef ERTS_SMP
-    erts_tsd_key_create(keyp,keyname);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_tsd_key_delete(erts_smp_tsd_key_t key)
-{
-#ifdef ERTS_SMP
-    erts_tsd_key_delete(key);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_tsd_set(erts_smp_tsd_key_t key, void *value)
-{
-#ifdef ERTS_SMP
-    erts_tsd_set(key, value);
-#endif
-}
-
-ERTS_GLB_INLINE void *
-erts_smp_tsd_get(erts_smp_tsd_key_t key)
-{
-#ifdef ERTS_SMP
-    return erts_tsd_get(key);
-#else
-    return NULL;
-#endif
-}
-
-#ifdef ERTS_THR_HAVE_SIG_FUNCS
-#define ERTS_SMP_THR_HAVE_SIG_FUNCS 1
-
-ERTS_GLB_INLINE void
-erts_smp_thr_sigmask(int how, const sigset_t *set, sigset_t *oset)
-{
-#ifdef ERTS_SMP
-    erts_thr_sigmask(how, set, oset);
-#endif
-}
-
-ERTS_GLB_INLINE void
-erts_smp_thr_sigwait(const sigset_t *set, int *sig)
-{
-#ifdef ERTS_SMP
-    erts_thr_sigwait(set, sig);
-#endif
-}
-
-#endif /* #ifdef ERTS_THR_HAVE_SIG_FUNCS */
-
-#endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
-
-#endif /* ERL_SMP_H */
-
-#ifdef ERTS_UNDEF_DEPRECATED_ATOMICS
-
-/* Deprecated functions to replace */
-
-#undef erts_smp_atomic_init
-#undef erts_smp_atomic_set
-#undef erts_smp_atomic_read
-#undef erts_smp_atomic_inctest
-#undef erts_smp_atomic_dectest
-#undef erts_smp_atomic_inc
-#undef erts_smp_atomic_dec
-#undef erts_smp_atomic_addtest
-#undef erts_smp_atomic_add
-#undef erts_smp_atomic_xchg
-#undef erts_smp_atomic_cmpxchg
-#undef erts_smp_atomic_bor
-#undef erts_smp_atomic_band
-
-#undef erts_smp_atomic32_init
-#undef erts_smp_atomic32_set
-#undef erts_smp_atomic32_read
-#undef erts_smp_atomic32_inctest
-#undef erts_smp_atomic32_dectest
-#undef erts_smp_atomic32_inc
-#undef erts_smp_atomic32_dec
-#undef erts_smp_atomic32_addtest
-#undef erts_smp_atomic32_add
-#undef erts_smp_atomic32_xchg
-#undef erts_smp_atomic32_cmpxchg
-#undef erts_smp_atomic32_bor
-#undef erts_smp_atomic32_band
-
-#endif
diff --git a/erts/emulator/beam/erl_term.h b/erts/emulator/beam/erl_term.h
index 842802f8d9..bddf403b0a 100644
--- a/erts/emulator/beam/erl_term.h
+++ b/erts/emulator/beam/erl_term.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2000-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2000-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -55,9 +55,6 @@ struct erl_node_; /* Declared in erl_node_tables.h */
 #if defined(ARCH_64)
 #  define TAG_PTR_MASK__	0x7
 #  if !defined(ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION)
-#    ifdef HIPE
-#      error Hipe on 64-bit needs a real mmap as it does not support the literal tag
-#    endif
 #    define TAG_LITERAL_PTR	0x4
 #  else
 #    undef TAG_LITERAL_PTR
@@ -270,7 +267,6 @@ _ET_DECLARE_CHECKED(Eterm*,list_val,Wterm)
 #define is_byte(x)	(((x) & ((~(Uint)0 << (_TAG_IMMED1_SIZE+8)) + _TAG_IMMED1_MASK)) == _TAG_IMMED1_SMALL)
 #define is_valid_bit_size(x) (((Sint)(x)) >= 0 && ((x) & 0x7F) == _TAG_IMMED1_SMALL)
 #define is_not_valid_bit_size(x) (!is_valid_bit_size((x)))
-#define MY_IS_SSMALL(x) (((Uint) ((((x)) >> (SMALL_BITS-1)) + 1)) < 2)
 #define _unchecked_unsigned_val(x)	((x) >> _TAG_IMMED1_SIZE)
 _ET_DECLARE_CHECKED(Uint,unsigned_val,Eterm)
 #define unsigned_val(x)	_ET_APPLY(unsigned_val,(x))
@@ -317,7 +313,8 @@ _ET_DECLARE_CHECKED(Uint,header_arity,Eterm)
 #define MAX_ARITYVAL            ((((Uint)1) << 24) - 1)
 #define ERTS_MAX_TUPLE_SIZE     MAX_ARITYVAL
 
-#define make_arityval(sz)	_make_header((sz),_TAG_HEADER_ARITYVAL)
+#define make_arityval(sz)	(ASSERT((sz) <= MAX_ARITYVAL), \
+                                 _make_header((sz),_TAG_HEADER_ARITYVAL))
 #define is_arity_value(x)	(((x) & _TAG_HEADER_MASK) == _TAG_HEADER_ARITYVAL)
 #define is_sane_arity_value(x)	((((x) & _TAG_HEADER_MASK) == _TAG_HEADER_ARITYVAL) && \
 				 (((x) >> _HEADER_ARITY_OFFS) <= MAX_ARITYVAL))
@@ -348,6 +345,9 @@ _ET_DECLARE_CHECKED(Uint,thing_subtag,Eterm)
  *
  * To help find code which makes unwarranted assumptions about zero,
  * we now use a non-zero bit-pattern in debug mode.
+ *
+ * In order to be able to differentiata against values, the non-value
+ * needs to be tagged as a header of some sort.
  */
 #if ET_DEBUG
 # ifdef HIPE
@@ -358,7 +358,7 @@ _ET_DECLARE_CHECKED(Uint,thing_subtag,Eterm)
 #  define THE_NON_VALUE	_make_header(0,_TAG_HEADER_FLOAT)
 # endif
 #else
-#define THE_NON_VALUE	(0)
+#define THE_NON_VALUE	(TAG_PRIMARY_HEADER)
 #endif
 #define is_non_value(x)	((x) == THE_NON_VALUE)
 #define is_value(x)	((x) != THE_NON_VALUE)
@@ -863,7 +863,7 @@ do {									\
   ((ErtsMRefThing *) (Hp))->mb = (Binp);				\
   ((ErtsMRefThing *) (Hp))->next = (Ohp)->first;			\
   (Ohp)->first = (struct erl_off_heap_header*) (Hp);			\
-  ASSERT(erts_is_ref_numbers_magic(&(Binp)->refn));			\
+  ASSERT(erts_is_ref_numbers_magic((Binp)->refn));			\
 } while (0)
 
 #endif /* ARCH_32 */
@@ -1184,6 +1184,54 @@ _ET_DECLARE_CHECKED(struct erl_node_*,external_ref_node,Eterm)
 #define is_map(x)              (is_boxed((x)) && is_map_header(*boxed_val(x)))
 #define is_not_map(x)          (!is_map(x))
 
+#define MAP_HEADER(hp, sz, keys)                \
+    ((hp)[0] = MAP_HEADER_FLATMAP,              \
+     (hp)[1] = sz,                              \
+     (hp)[2] = keys)
+
+#define MAP_SZ(sz) (MAP_HEADER_FLATMAP_SZ + 2*sz + 1)
+
+#define MAP0_SZ MAP_SZ(0)
+#define MAP1_SZ MAP_SZ(1)
+#define MAP2_SZ MAP_SZ(2)
+#define MAP3_SZ MAP_SZ(3)
+#define MAP4_SZ MAP_SZ(4)
+#define MAP5_SZ MAP_SZ(5)
+#define MAP0(hp)                                                \
+    (MAP_HEADER(hp, 0, TUPLE0(hp+MAP_HEADER_FLATMAP_SZ)),       \
+     make_flatmap(hp))
+#define MAP1(hp, k1, v1)                                                \
+    (MAP_HEADER(hp, 1, TUPLE1(hp+1+MAP_HEADER_FLATMAP_SZ, k1)),         \
+     (hp)[MAP_HEADER_FLATMAP_SZ+0] = v1,                                \
+     make_flatmap(hp))
+#define MAP2(hp, k1, v1, k2, v2)                                        \
+    (MAP_HEADER(hp, 2, TUPLE2(hp+2+MAP_HEADER_FLATMAP_SZ, k1, k2)),     \
+     (hp)[MAP_HEADER_FLATMAP_SZ+0] = v1,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+1] = v2,                                \
+     make_flatmap(hp))
+#define MAP3(hp, k1, v1, k2, v2, k3, v3)                \
+    (MAP_HEADER(hp, 3, TUPLE3(hp+3+MAP_HEADER_FLATMAP_SZ, k1, k2, k3)), \
+     (hp)[MAP_HEADER_FLATMAP_SZ+0] = v1,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+1] = v2,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+2] = v3,                                \
+     make_flatmap(hp))
+#define MAP4(hp, k1, v1, k2, v2, k3, v3, k4, v4)                \
+    (MAP_HEADER(hp, 4, TUPLE4(hp+4+MAP_HEADER_FLATMAP_SZ, k1, k2, k3, k4)), \
+     (hp)[MAP_HEADER_FLATMAP_SZ+0] = v1,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+1] = v2,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+2] = v3,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+3] = v4,                                \
+     make_flatmap(hp))
+#define MAP5(hp, k1, v1, k2, v2, k3, v3, k4, v4, k5, v5)                \
+    (MAP_HEADER(hp, 5, TUPLE5(hp+5+MAP_HEADER_FLATMAP_SZ, k1, k2, k3, k4, k5)), \
+     (hp)[MAP_HEADER_FLATMAP_SZ+0] = v1,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+1] = v2,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+2] = v3,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+3] = v4,                                \
+     (hp)[MAP_HEADER_FLATMAP_SZ+4] = v5,                                \
+     make_flatmap(hp))
+
+
 /* number tests */
 
 #define is_integer(x)		(is_small(x) || is_big(x))
diff --git a/erts/emulator/beam/erl_thr_progress.c b/erts/emulator/beam/erl_thr_progress.c
index 2a9f276e02..96824dc06e 100644
--- a/erts/emulator/beam/erl_thr_progress.c
+++ b/erts/emulator/beam/erl_thr_progress.c
@@ -80,7 +80,6 @@
 #include "erl_thr_progress.h"
 #include "global.h"
 
-#ifdef ERTS_SMP
 
 #define ERTS_THR_PRGR_DBG_CHK_WAKEUP_REQUEST_VALUE 0
 
@@ -1513,4 +1512,3 @@ void erts_thr_progress_dbg_print_state(void)
 
 }
 
-#endif
diff --git a/erts/emulator/beam/erl_thr_progress.h b/erts/emulator/beam/erl_thr_progress.h
index b2894ba1fe..8c029bcf99 100644
--- a/erts/emulator/beam/erl_thr_progress.h
+++ b/erts/emulator/beam/erl_thr_progress.h
@@ -28,23 +28,11 @@
  * Author: 	Rickard Green
  */
 
-#if !defined(ERL_THR_PROGRESS_H__TSD_TYPE__)
+#ifndef ERL_THR_PROGRESS_H__TSD_TYPE__
 #define ERL_THR_PROGRESS_H__TSD_TYPE__
 
 #include "sys.h"
 
-#ifndef ERTS_SMP
-
-#define erts_smp_thr_progress_block() ((void) 0)
-#define erts_smp_thr_progress_unblock() ((void) 0)
-#define erts_smp_thr_progress_is_blocking() 1
-
-#else /* ERTS_SMP */
-
-#define erts_smp_thr_progress_block erts_thr_progress_block
-#define erts_smp_thr_progress_unblock erts_thr_progress_unblock
-#define erts_smp_thr_progress_is_blocking erts_thr_progress_is_blocking
-
 void erts_thr_progress_block(void);
 void erts_thr_progress_unblock(void);
 int erts_thr_progress_is_blocking(void);
@@ -87,13 +75,10 @@ typedef struct {
 int erts_thr_progress_fatal_error_block(ErtsThrPrgrData *tmp_tpd_bufp);
 void erts_thr_progress_fatal_error_wait(SWord timeout);
 
-#endif /* ERTS_SMP */
 
 typedef struct ErtsThrPrgrLaterOp_ ErtsThrPrgrLaterOp;
 struct ErtsThrPrgrLaterOp_ {
-#ifdef ERTS_SMP
     ErtsThrPrgrVal later;
-#endif
     void (*func)(void *);
     void *data;
     ErtsThrPrgrLaterOp *next;
@@ -107,7 +92,6 @@ struct ErtsThrPrgrLaterOp_ {
 #include "erl_threads.h"
 #include "erl_process.h"
 
-#ifdef ERTS_SMP
 
 /* ERTS_THR_PRGR_VAL_FIRST should only be used when initializing... */
 #define ERTS_THR_PRGR_VAL_FIRST ((ErtsThrPrgrVal) 0)
@@ -324,6 +308,5 @@ erts_thr_progress_has_reached(ErtsThrPrgrVal val)
 
 #endif
 
-#endif /* ERTS_SMP */
 
 #endif
diff --git a/erts/emulator/beam/erl_thr_queue.c b/erts/emulator/beam/erl_thr_queue.c
index f56d0828dd..548c2768e5 100644
--- a/erts/emulator/beam/erl_thr_queue.c
+++ b/erts/emulator/beam/erl_thr_queue.c
@@ -87,32 +87,10 @@
 
 #define ERTS_THR_Q_MAX_FINI_DEQ_OPS 50
 
-#ifdef ERTS_SMP
 ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(sl_element,
 				 ErtsThrQElement_t,
 				 1000,
 				 ERTS_ALC_T_THR_Q_EL_SL)
-#else
-
-static void
-init_sl_element_alloc(void)
-{
-}
-
-static ErtsThrQElement_t *
-sl_element_alloc(void)
-{
-    return erts_alloc(ERTS_ALC_T_THR_Q_EL_SL,
-		      sizeof(ErtsThrQElement_t));
-}
-
-static void
-sl_element_free(ErtsThrQElement_t *p)
-{
-    erts_free(ERTS_ALC_T_THR_Q_EL_SL, p);
-}
-
-#endif
 
 #define ErtsThrQDirtyReadEl(A) \
     ((ErtsThrQElement_t *) erts_atomic_read_dirty((A)))
@@ -135,14 +113,6 @@ static void noop_callback(void *arg) { }
 void
 erts_thr_q_initialize(ErtsThrQ_t *q, ErtsThrQInit_t *qi)
 {
-#ifndef USE_THREADS
-    q->init = *qi;
-    if (!q->init.notify)
-	q->init.notify = noop_callback;
-    q->first = NULL;
-    q->last = NULL;
-    q->q.blk = NULL;
-#else
     erts_atomic_init_nob(&q->tail.data.marker.next, ERTS_AINT_NULL);
     q->tail.data.marker.data.ptr = NULL;
     erts_atomic_init_nob(&q->tail.data.last,
@@ -164,10 +134,8 @@ erts_thr_q_initialize(ErtsThrQ_t *q, ErtsThrQInit_t *qi)
     q->head.deq_fini.automatic = qi->auto_finalize_dequeue;
     q->head.deq_fini.start = NULL;
     q->head.deq_fini.end = NULL;
-#ifdef ERTS_SMP
     q->head.next.thr_progress = erts_thr_progress_current();
     q->head.next.thr_progress_reached = 1;
-#endif
     q->head.next.um_refc_ix = 1;
     q->head.next.unref_end = &q->tail.data.marker;
     q->head.used_marker = 1;
@@ -176,15 +144,12 @@ erts_thr_q_initialize(ErtsThrQ_t *q, ErtsThrQInit_t *qi)
     q->q.finalizing = 0;
     q->q.live = qi->live.queue;
     q->q.blk = NULL;
-#endif
 }
 
 ErtsThrQCleanState_t
 erts_thr_q_finalize(ErtsThrQ_t *q)
 {
-#ifdef USE_THREADS
     q->q.finalizing = 1;
-#endif
     while (erts_thr_q_dequeue(q));
     return erts_thr_q_clean(q);
 }
@@ -229,7 +194,6 @@ erts_thr_q_destroy(ErtsThrQ_t *q)
     return erts_thr_q_finalize(q);
 }
 
-#ifdef USE_THREADS
 
 static void
 destroy(ErtsThrQ_t *q)
@@ -249,7 +213,6 @@ destroy(ErtsThrQ_t *q)
     erts_free(atype, q->q.blk);
 }
 
-#endif
 
 static ERTS_INLINE ErtsThrQElement_t *
 element_live_alloc(ErtsThrQLive_t live)
@@ -267,11 +230,7 @@ static ERTS_INLINE ErtsThrQElement_t *
 element_alloc(ErtsThrQ_t *q)
 {
     ErtsThrQLive_t live;
-#ifdef USE_THREADS
     live = q->tail.data.live;
-#else
-    live = q->init.live.objects;
-#endif
     return element_live_alloc(live);
 }
 
@@ -291,15 +250,10 @@ static ERTS_INLINE void
 element_free(ErtsThrQ_t *q, ErtsThrQElement_t *el)
 {
     ErtsThrQLive_t live;
-#ifdef USE_THREADS
     live = q->head.live;
-#else
-    live = q->init.live.objects;
-#endif
     element_live_free(live, el);
 }
 
-#ifdef USE_THREADS
 
 static ERTS_INLINE ErtsThrQElement_t *
 enqueue_managed(ErtsThrQ_t *q, ErtsThrQElement_t *this)
@@ -423,11 +377,9 @@ clean(ErtsThrQ_t *q, int max_ops, int do_notify)
 	return ERTS_THR_Q_CLEAN;
     }
 
-#ifdef ERTS_SMP
     if (q->head.next.thr_progress_reached
 	|| erts_thr_progress_has_reached(q->head.next.thr_progress)) {
 	q->head.next.thr_progress_reached = 1;
-#endif
 	um_refc_ix = q->head.next.um_refc_ix;
 	if (erts_atomic_read_acqb(&q->tail.data.um_refc[um_refc_ix]) == 0) {
 	    /* Move unreferenced end pointer forward... */
@@ -439,23 +391,17 @@ clean(ErtsThrQ_t *q, int max_ops, int do_notify)
 		ilast = (erts_aint_t) enqueue_marker(q, NULL);
 
 	    if (q->head.unref_end == (ErtsThrQElement_t *) ilast)
-		ERTS_SMP_MEMORY_BARRIER;
+		ERTS_THR_MEMORY_BARRIER;
 	    else {
 		q->head.next.unref_end = (ErtsThrQElement_t *) ilast;
-#ifdef ERTS_SMP
 		q->head.next.thr_progress = erts_thr_progress_later(NULL);
-#endif
 		erts_atomic32_set_relb(&q->tail.data.um_refc_ix,
 				       um_refc_ix);
 		q->head.next.um_refc_ix = um_refc_ix == 0 ? 1 : 0;
-#ifdef ERTS_SMP
 		q->head.next.thr_progress_reached = 0;
-#endif
 	    }
 	}
-#ifdef ERTS_SMP
     }
-#endif
 
     head = ErtsThrQDirtyReadEl(&q->head.head);
     if (q->head.first == head) {
@@ -489,9 +435,7 @@ clean(ErtsThrQ_t *q, int max_ops, int do_notify)
 
 check_thr_progress:
 
-#ifdef ERTS_SMP
     if (q->head.next.thr_progress_reached)
-#endif
     {
 	int um_refc_ix = q->head.next.um_refc_ix;
 	if (erts_atomic_read_acqb(&q->tail.data.um_refc[um_refc_ix]) == 0) {
@@ -505,24 +449,16 @@ check_thr_progress:
     return ERTS_THR_Q_NEED_THR_PRGR;
 }
 
-#endif
 
 ErtsThrQCleanState_t
 erts_thr_q_clean(ErtsThrQ_t *q)
 {
-#ifdef USE_THREADS
     return clean(q, ERTS_THR_Q_MAX_SCHED_CLEAN_OPS, 0);
-#else
-    return ERTS_THR_Q_CLEAN;
-#endif
 }
 
 ErtsThrQCleanState_t
 erts_thr_q_inspect(ErtsThrQ_t *q, int ensure_empty)
 {
-#ifndef USE_THREADS
-    return ERTS_THR_Q_CLEAN;
-#else
     ErtsThrQElement_t *head = ErtsThrQDirtyReadEl(&q->head.head);
     if (ensure_empty) {
 	erts_aint_t inext;
@@ -553,39 +489,21 @@ erts_thr_q_inspect(ErtsThrQ_t *q, int ensure_empty)
     if (q->head.first != q->head.unref_end)
 	return ERTS_THR_Q_DIRTY;
 
-#ifdef ERTS_SMP
     if (q->head.next.thr_progress_reached)
-#endif
     {
 	int um_refc_ix = q->head.next.um_refc_ix;
 	if (erts_atomic_read_acqb(&q->tail.data.um_refc[um_refc_ix]) == 0)
 	    return ERTS_THR_Q_DIRTY;
     }
     return ERTS_THR_Q_NEED_THR_PRGR;
-#endif
 }
 
 static void
 enqueue(ErtsThrQ_t *q, void *data, ErtsThrQElement_t *this)
 {
-#ifndef USE_THREADS
-    ASSERT(data);
-
-    this->next = NULL;
-    this->data.ptr = data;
-
-    if (q->last)
-	q->last->next = this;
-    else {
-	q->first = q->last = this;
-	q->init.notify(q->init.arg);
-    }
-#else
     int notify;
     int um_refc_ix = 0;
-#ifdef ERTS_SMP
     int unmanaged_thread;
-#endif
 
 #if ERTS_THR_Q_DBG_CHK_DATA
     if (!data)
@@ -596,10 +514,8 @@ enqueue(ErtsThrQ_t *q, void *data, ErtsThrQElement_t *this)
 
     this->data.ptr = data;
 
-#ifdef ERTS_SMP
     unmanaged_thread = !erts_thr_progress_is_managed_thread();
     if (unmanaged_thread)
-#endif
     {
 	um_refc_ix = erts_atomic32_read_acqb(&q->tail.data.um_refc_ix);
 	while (1) {
@@ -616,9 +532,7 @@ enqueue(ErtsThrQ_t *q, void *data, ErtsThrQElement_t *this)
     notify = this == enqueue_managed(q, this);
 	
 
-#ifdef ERTS_SMP
     if (unmanaged_thread)
-#endif
     {
 	if (notify)
 	    erts_atomic_dec_relb(&q->tail.data.um_refc[um_refc_ix]);
@@ -627,7 +541,6 @@ enqueue(ErtsThrQ_t *q, void *data, ErtsThrQElement_t *this)
     }
     if (notify)
 	q->tail.data.notify(q->tail.data.arg);
-#endif
 }
 
 void
@@ -645,9 +558,6 @@ erts_thr_q_prepare_enqueue(ErtsThrQ_t *q)
 int
 erts_thr_q_get_finalize_dequeue_data(ErtsThrQ_t *q, ErtsThrQFinDeQ_t *fdp)
 {
-#ifndef USE_THREADS
-    return 0;
-#else
 #ifdef DEBUG
     if (!q->head.deq_fini.start) {
 	ASSERT(!q->head.deq_fini.end);
@@ -670,14 +580,12 @@ erts_thr_q_get_finalize_dequeue_data(ErtsThrQ_t *q, ErtsThrQFinDeQ_t *fdp)
     q->head.deq_fini.start = NULL;
     q->head.deq_fini.end = NULL;
     return fdp->start != NULL;
-#endif
 }
 
 void
 erts_thr_q_append_finalize_dequeue_data(ErtsThrQFinDeQ_t *fdp0,
 					ErtsThrQFinDeQ_t *fdp1)
 {
-#ifdef USE_THREADS
     if (fdp1->start) {
 	if (fdp0->end)
 	    ErtsThrQDirtySetEl(&fdp0->end->next, fdp1->start);
@@ -685,13 +593,11 @@ erts_thr_q_append_finalize_dequeue_data(ErtsThrQFinDeQ_t *fdp0,
 	    fdp0->start = fdp1->start;
 	fdp0->end = fdp1->end;
     }
-#endif
 }
 
 
 int erts_thr_q_finalize_dequeue(ErtsThrQFinDeQ_t *state)
 {
-#ifdef USE_THREADS
     ErtsThrQElement_t *start = state->start;
     if (start) {
 	ErtsThrQLive_t live;
@@ -710,17 +616,14 @@ int erts_thr_q_finalize_dequeue(ErtsThrQFinDeQ_t *state)
 	    return 1; /* More to do */
 	state->end = NULL;
     }
-#endif
     return 0;
 }
 
 void
 erts_thr_q_finalize_dequeue_state_init(ErtsThrQFinDeQ_t *state)
 {
-#ifdef USE_THREADS
     state->start = NULL;
     state->end = NULL;
-#endif
 }
 
 
@@ -734,22 +637,6 @@ erts_thr_q_enqueue_prepared(ErtsThrQ_t *q, void *data, ErtsThrQPrepEnQ_t *prep)
 void *
 erts_thr_q_dequeue(ErtsThrQ_t *q)
 {
-#ifndef USE_THREADS
-    void *res;
-    ErtsThrQElement_t *tmp;
-
-    if (!q->first)
-	return NULL;
-    tmp = q->first;
-    res = tmp->data.ptr;
-    q->first = tmp->next;
-    if (!q->first)
-	q->last = NULL;
-
-    element_free(q, tmp);
-
-    return res;
-#else
     ErtsThrQElement_t *head;
     erts_aint_t inext;
     void *res;
@@ -778,7 +665,6 @@ erts_thr_q_dequeue(ErtsThrQ_t *q)
 	   ? ERTS_THR_Q_MAX_DEQUEUE_CLEAN_OPS
 	   : ERTS_THR_Q_MAX_SCHED_CLEAN_OPS), 1);
     return res;
-#endif
 }
 
 #ifdef USE_LTTNG_VM_TRACEPOINTS
@@ -786,14 +672,6 @@ int
 erts_thr_q_length_dirty(ErtsThrQ_t *q)
 {
     int n = 0;
-#ifndef USE_THREADS
-    void *res;
-    ErtsThrQElement_t *tmp;
-
-    for (tmp = q->first; tmp != NULL; tmp = tmp->next) {
-        n++;
-    }
-#else
     ErtsThrQElement_t *e;
     erts_aint_t inext;
 
@@ -808,7 +686,6 @@ erts_thr_q_length_dirty(ErtsThrQ_t *q)
         }
         inext = erts_atomic_read_acqb(&e->next);
     }
-#endif
     return n;
 }
 #endif
diff --git a/erts/emulator/beam/erl_thr_queue.h b/erts/emulator/beam/erl_thr_queue.h
index 705a67af4c..163a25318d 100644
--- a/erts/emulator/beam/erl_thr_queue.h
+++ b/erts/emulator/beam/erl_thr_queue.h
@@ -78,11 +78,7 @@ typedef struct ErtsThrQElement_t_ ErtsThrQElement_t;
 typedef struct ErtsThrQElement_t ErtsThrQPrepEnQ_t;
 
 struct ErtsThrQElement_t_ {
-#ifdef USE_THREADS
     erts_atomic_t next;
-#else
-    ErtsThrQElement_t *next;
-#endif
     union {
 	erts_atomic_t atmc;
 	void *ptr;
@@ -100,7 +96,6 @@ typedef enum {
     ERTS_THR_Q_DIRTY,
 } ErtsThrQCleanState_t;
 
-#ifdef USE_THREADS
 
 typedef struct {
     ErtsThrQElement_t marker;
@@ -108,9 +103,7 @@ typedef struct {
     erts_atomic_t um_refc[2];
     erts_atomic32_t um_refc_ix;
     ErtsThrQLive_t live;
-#ifdef ERTS_SMP
     erts_atomic32_t thr_prgr_clean_scheduled;
-#endif
     void *arg;
     void (*notify)(void *);
 } ErtsThrQTail_t;
@@ -141,10 +134,8 @@ struct ErtsThrQ_t_ {
 	    ErtsThrQElement_t *end;
 	} deq_fini;
 	struct {
-#ifdef ERTS_SMP
 	    ErtsThrPrgrVal thr_progress;
 	    int thr_progress_reached;
-#endif
 	    int um_refc_ix;
 	    ErtsThrQElement_t *unref_end;
 	} next;
@@ -159,18 +150,6 @@ struct ErtsThrQ_t_ {
     } q;
 };
 
-#else /* !USE_THREADS */
-
-struct ErtsThrQ_t_ {
-    ErtsThrQInit_t init;
-    ErtsThrQElement_t *first;
-    ErtsThrQElement_t *last;
-    struct {
-	void *blk;
-    } q;
-};
-
-#endif
 
 void erts_thr_q_init(void);
 void erts_thr_q_initialize(ErtsThrQ_t *, ErtsThrQInit_t *);
@@ -194,19 +173,15 @@ void erts_thr_q_finalize_dequeue_state_init(ErtsThrQFinDeQ_t *);
 int erts_thr_q_length_dirty(ErtsThrQ_t *);
 #endif
 
-#ifdef ERTS_SMP
 ERTS_GLB_INLINE ErtsThrPrgrVal erts_thr_q_need_thr_progress(ErtsThrQ_t *q);
-#endif
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
-#ifdef ERTS_SMP
 ERTS_GLB_INLINE ErtsThrPrgrVal
 erts_thr_q_need_thr_progress(ErtsThrQ_t *q)
 {
     return q->head.next.thr_progress;
 }
-#endif
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
diff --git a/erts/emulator/beam/erl_threads.h b/erts/emulator/beam/erl_threads.h
index 8b5c17d739..aedceb6fc2 100644
--- a/erts/emulator/beam/erl_threads.h
+++ b/erts/emulator/beam/erl_threads.h
@@ -45,11 +45,6 @@
  *      Data dependency read barrier. Orders *only* loads
  *      according to data dependency across the barrier.
  *
- * If thread support has been disabled, these barriers will become no-ops.
- *
- * If the prefix ERTS_THR_ is replaced with ERTS_SMP_, the barriers will
- * be enabled only in the SMP enabled runtime system.
- *
  * --- Atomic operations ---
  *
  * Atomics operations exist for 32-bit, word size, and double word size
@@ -86,20 +81,6 @@
  *          barrier. Load in atomic operation is ordered
  *          before the barrier.
  *
- * If thread support has been disabled, these functions are mapped to
- * functions that performs the same operation, but aren't atomic
- * and don't imply any memory barriers.
- *
- * If the atomic operations are prefixed with erts_smp_ instead of only
- * erts_ the atomic operations will only be atomic in the SMP enabled
- * runtime system, and will be mapped to non-atomic operations without
- * memory barriers in the runtime system without SMP support. Atomic
- * operations with erts_smp_ prefix should use the atomic types
- * erts_smp_atomic32_t, erts_smp_atomic_t, and erts_smp_dw_atomic_t
- * instead of erts_atomic32_t, erts_atomic_t, and erts_dw_atomic_t. The
- * integer data types erts_aint32_t, erts_aint_t, and erts_dw_atomic_t
- * are the same.
- *
  * --- 32-bit atomic operations ---
  *
  * The following 32-bit atomic operations exist. <B> should be
@@ -262,7 +243,6 @@
 #include "erl_lock_flags.h"
 #include "erl_term.h"
 
-#ifdef USE_THREADS
 
 #define ETHR_TRY_INLINE_FUNCS
 #include "ethread.h"
@@ -405,76 +385,6 @@ __decl_noreturn void  __noreturn erts_thr_fatal_error(int, char *);
 #  define ERTS_HAVE_REC_MTX_INIT	ETHR_HAVE_ETHR_REC_MUTEX_INIT
 #endif
 
-#else /* #ifdef USE_THREADS */
-
-#define ERTS_THR_MEMORY_BARRIER
-#define ERTS_THR_WRITE_MEMORY_BARRIER
-#define ERTS_THR_READ_MEMORY_BARRIER
-#define ERTS_THR_DATA_DEPENDENCY_READ_MEMORY_BARRIER
-
-#define ERTS_THR_OPTS_DEFAULT_INITER 0
-typedef int erts_thr_opts_t;
-typedef int erts_thr_init_data_t;
-typedef int erts_thr_late_init_data_t;
-typedef int erts_tid_t;
-typedef int erts_mtx_t;
-typedef int erts_cnd_t;
-#define ERTS_RWMTX_OPT_DEFAULT_INITER {0}
-#define ERTS_RWMTX_TYPE_NORMAL 0
-#define ERTS_RWMTX_TYPE_FREQUENT_READ 0
-#define ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ 0
-#define ERTS_RWMTX_LONG_LIVED 0
-#define ERTS_RWMTX_SHORT_LIVED 0
-#define ERTS_RWMTX_UNKNOWN_LIVED 0
-typedef struct {
-    char type;
-    char lived;
-    int main_spincount;
-    int aux_spincount;
-} erts_rwmtx_opt_t;
-typedef int erts_rwmtx_t;
-typedef int erts_tsd_key_t;
-typedef int erts_tse_t;
-
-typedef struct { SWord sint[2]; } erts_dw_aint_t;
-typedef SWord erts_aint_t;
-typedef Sint32 erts_aint32_t;
-typedef Sint64 erts_aint64_t;
-
-#define erts_dw_atomic_t erts_dw_aint_t
-#define erts_atomic_t erts_aint_t
-#define erts_atomic32_t erts_aint32_t
-#define erts_atomic64_t erts_aint64_t
-
-#if __GNUC__ > 2
-typedef struct { } erts_spinlock_t;
-typedef struct { } erts_rwlock_t;
-#else
-typedef struct { int gcc_is_buggy; } erts_spinlock_t;
-typedef struct { int gcc_is_buggy; } erts_rwlock_t;
-#endif
-
-#ifdef WORDS_BIGENDIAN
-#define ERTS_DW_AINT_LOW_WORD 1
-#define ERTS_DW_AINT_HIGH_WORD 0
-#else
-#define ERTS_DW_AINT_LOW_WORD 0
-#define ERTS_DW_AINT_HIGH_WORD 1
-#endif
-
-#define ERTS_MTX_INITER			0
-#define ERTS_CND_INITER			0
-#define ERTS_THR_INIT_DATA_DEF_INITER	0
-
-#define ERTS_HAVE_REC_MTX_INIT		1
-
-#endif /* #ifdef USE_THREADS */
-
-#define erts_no_dw_atomic_t erts_dw_aint_t
-#define erts_no_atomic_t erts_aint_t
-#define erts_no_atomic32_t erts_aint32_t
-#define erts_no_atomic64_t erts_aint64_t
-
 #define ERTS_AINT_NULL ((erts_aint_t) NULL)
 
 #define ERTS_AINT_T_MAX (~(((erts_aint_t) 1) << (sizeof(erts_aint_t)*8-1)))
@@ -544,79 +454,6 @@ ERTS_GLB_INLINE void erts_rwmtx_runlock(erts_rwmtx_t *rwmtx);
 ERTS_GLB_INLINE void erts_rwmtx_rwunlock(erts_rwmtx_t *rwmtx);
 ERTS_GLB_INLINE int erts_lc_rwmtx_is_rlocked(erts_rwmtx_t *mtx);
 ERTS_GLB_INLINE int erts_lc_rwmtx_is_rwlocked(erts_rwmtx_t *mtx);
-
-ERTS_GLB_INLINE void erts_no_dw_atomic_set(erts_no_dw_atomic_t *var, erts_no_dw_atomic_t *val);
-ERTS_GLB_INLINE void erts_no_dw_atomic_read(erts_no_dw_atomic_t *var, erts_no_dw_atomic_t *val);
-ERTS_GLB_INLINE int erts_no_dw_atomic_cmpxchg(erts_no_dw_atomic_t *var,
-					      erts_no_dw_atomic_t *val,
-					      erts_no_dw_atomic_t *old_val);
-ERTS_GLB_INLINE void erts_no_atomic_set(erts_no_atomic_t *var, erts_aint_t i);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_read(erts_no_atomic_t *var);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_inc_read(erts_no_atomic_t *incp);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_dec_read(erts_no_atomic_t *decp);
-ERTS_GLB_INLINE void erts_no_atomic_inc(erts_no_atomic_t *incp);
-ERTS_GLB_INLINE void erts_no_atomic_dec(erts_no_atomic_t *decp);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_add_read(erts_no_atomic_t *addp,
-						    erts_aint_t i);
-ERTS_GLB_INLINE void erts_no_atomic_add(erts_no_atomic_t *addp, erts_aint_t i);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_read_bor(erts_no_atomic_t *var,
-						    erts_aint_t mask);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_read_band(erts_no_atomic_t *var,
-						     erts_aint_t mask);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_xchg(erts_no_atomic_t *xchgp,
-						erts_aint_t new);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_cmpxchg(erts_no_atomic_t *xchgp,
-						   erts_aint_t new,
-						   erts_aint_t expected);
-ERTS_GLB_INLINE erts_aint_t erts_no_atomic_read_bset(erts_no_atomic_t *var,
-						     erts_aint_t mask,
-						     erts_aint_t set);
-ERTS_GLB_INLINE void erts_no_atomic32_set(erts_no_atomic32_t *var,
-					  erts_aint32_t i);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_read(erts_no_atomic32_t *var);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_inc_read(erts_no_atomic32_t *incp);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_dec_read(erts_no_atomic32_t *decp);
-ERTS_GLB_INLINE void erts_no_atomic32_inc(erts_no_atomic32_t *incp);
-ERTS_GLB_INLINE void erts_no_atomic32_dec(erts_no_atomic32_t *decp);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_add_read(erts_no_atomic32_t *addp,
-							erts_aint32_t i);
-ERTS_GLB_INLINE void erts_no_atomic32_add(erts_no_atomic32_t *addp,
-					  erts_aint32_t i);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_read_bor(erts_no_atomic32_t *var,
-							erts_aint32_t mask);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_read_band(erts_no_atomic32_t *var,
-							 erts_aint32_t mask);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_xchg(erts_no_atomic32_t *xchgp,
-						    erts_aint32_t new);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_cmpxchg(erts_no_atomic32_t *xchgp,
-						       erts_aint32_t new,
-						       erts_aint32_t expected);
-ERTS_GLB_INLINE erts_aint32_t erts_no_atomic32_read_bset(erts_no_atomic32_t *var,
-							 erts_aint32_t mask,
-							 erts_aint32_t set);
-ERTS_GLB_INLINE void erts_no_atomic64_set(erts_no_atomic64_t *var,
-					  erts_aint64_t i);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_read(erts_no_atomic64_t *var);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_inc_read(erts_no_atomic64_t *incp);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_dec_read(erts_no_atomic64_t *decp);
-ERTS_GLB_INLINE void erts_no_atomic64_inc(erts_no_atomic64_t *incp);
-ERTS_GLB_INLINE void erts_no_atomic64_dec(erts_no_atomic64_t *decp);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_add_read(erts_no_atomic64_t *addp,
-							erts_aint64_t i);
-ERTS_GLB_INLINE void erts_no_atomic64_add(erts_no_atomic64_t *addp,
-					  erts_aint64_t i);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_read_bor(erts_no_atomic64_t *var,
-							erts_aint64_t mask);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_read_band(erts_no_atomic64_t *var,
-							 erts_aint64_t mask);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_xchg(erts_no_atomic64_t *xchgp,
-						    erts_aint64_t new);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_cmpxchg(erts_no_atomic64_t *xchgp,
-						       erts_aint64_t new,
-						       erts_aint64_t expected);
-ERTS_GLB_INLINE erts_aint64_t erts_no_atomic64_read_bset(erts_no_atomic64_t *var,
-							 erts_aint64_t mask,
-							 erts_aint64_t set);
 ERTS_GLB_INLINE void erts_spinlock_init(erts_spinlock_t *lock,
                                         char *name,
                                         Eterm extra,
@@ -670,13 +507,10 @@ ERTS_GLB_INLINE void erts_thr_sigmask(int how, const sigset_t *set,
 				      sigset_t *oset);
 ERTS_GLB_INLINE void erts_thr_sigwait(const sigset_t *set, int *sig);
 
-#ifdef USE_THREADS
 ERTS_GLB_INLINE void erts_thr_kill(erts_tid_t tid, int sig);
-#endif
 
 #endif /* #ifdef HAVE_ETHR_SIG_FUNCS */
 
-#ifdef USE_THREADS
 
 ERTS_GLB_INLINE erts_aint_t
 erts_atomic_read_bset_nob(erts_atomic_t *var,
@@ -1684,379 +1518,6 @@ erts_atomic64_read_dirty(erts_atomic64_t *var)
 
 #endif /* ARCH_32 */
 
-#else /* !USE_THREADS */
-
-/* Double word size atomics */
-
-#define erts_dw_atomic_init_nob erts_no_dw_atomic_set
-#define erts_dw_atomic_set_nob erts_no_dw_atomic_set
-#define erts_dw_atomic_read_nob erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_nob erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_mb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_mb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_mb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_mb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_acqb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_acqb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_acqb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_acqb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_relb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_relb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_relb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_relb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_ddrb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_ddrb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_ddrb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_ddrb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_rb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_rb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_rb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_rb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_init_wb erts_no_dw_atomic_init
-#define erts_dw_atomic_set_wb erts_no_dw_atomic_set
-#define erts_dw_atomic_read_wb erts_no_dw_atomic_read
-#define erts_dw_atomic_cmpxchg_wb erts_no_dw_atomic_cmpxchg
-
-#define erts_dw_atomic_set_dirty erts_no_dw_atomic_set
-#define erts_dw_atomic_read_dirty erts_no_dw_atomic_read
-
-/* Word size atomics */
-
-#define erts_atomic_init_nob erts_no_atomic_set
-#define erts_atomic_set_nob erts_no_atomic_set
-#define erts_atomic_read_nob erts_no_atomic_read
-#define erts_atomic_inc_read_nob erts_no_atomic_inc_read
-#define erts_atomic_dec_read_nob erts_no_atomic_dec_read
-#define erts_atomic_inc_nob erts_no_atomic_inc
-#define erts_atomic_dec_nob erts_no_atomic_dec
-#define erts_atomic_add_read_nob erts_no_atomic_add_read
-#define erts_atomic_add_nob erts_no_atomic_add
-#define erts_atomic_read_bor_nob erts_no_atomic_read_bor
-#define erts_atomic_read_band_nob erts_no_atomic_read_band
-#define erts_atomic_xchg_nob erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_nob erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_nob erts_no_atomic_read_bset
-
-#define erts_atomic_init_mb erts_no_atomic_set
-#define erts_atomic_set_mb erts_no_atomic_set
-#define erts_atomic_read_mb erts_no_atomic_read
-#define erts_atomic_inc_read_mb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_mb erts_no_atomic_dec_read
-#define erts_atomic_inc_mb erts_no_atomic_inc
-#define erts_atomic_dec_mb erts_no_atomic_dec
-#define erts_atomic_add_read_mb erts_no_atomic_add_read
-#define erts_atomic_add_mb erts_no_atomic_add
-#define erts_atomic_read_bor_mb erts_no_atomic_read_bor
-#define erts_atomic_read_band_mb erts_no_atomic_read_band
-#define erts_atomic_xchg_mb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_mb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_mb erts_no_atomic_read_bset
-
-#define erts_atomic_init_acqb erts_no_atomic_set
-#define erts_atomic_set_acqb erts_no_atomic_set
-#define erts_atomic_read_acqb erts_no_atomic_read
-#define erts_atomic_inc_read_acqb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_acqb erts_no_atomic_dec_read
-#define erts_atomic_inc_acqb erts_no_atomic_inc
-#define erts_atomic_dec_acqb erts_no_atomic_dec
-#define erts_atomic_add_read_acqb erts_no_atomic_add_read
-#define erts_atomic_add_acqb erts_no_atomic_add
-#define erts_atomic_read_bor_acqb erts_no_atomic_read_bor
-#define erts_atomic_read_band_acqb erts_no_atomic_read_band
-#define erts_atomic_xchg_acqb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_acqb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_acqb erts_no_atomic_read_bset
-
-#define erts_atomic_init_relb erts_no_atomic_set
-#define erts_atomic_set_relb erts_no_atomic_set
-#define erts_atomic_read_relb erts_no_atomic_read
-#define erts_atomic_inc_read_relb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_relb erts_no_atomic_dec_read
-#define erts_atomic_inc_relb erts_no_atomic_inc
-#define erts_atomic_dec_relb erts_no_atomic_dec
-#define erts_atomic_add_read_relb erts_no_atomic_add_read
-#define erts_atomic_add_relb erts_no_atomic_add
-#define erts_atomic_read_bor_relb erts_no_atomic_read_bor
-#define erts_atomic_read_band_relb erts_no_atomic_read_band
-#define erts_atomic_xchg_relb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_relb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_relb erts_no_atomic_read_bset
-
-#define erts_atomic_init_ddrb erts_no_atomic_set
-#define erts_atomic_set_ddrb erts_no_atomic_set
-#define erts_atomic_read_ddrb erts_no_atomic_read
-#define erts_atomic_inc_read_ddrb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_ddrb erts_no_atomic_dec_read
-#define erts_atomic_inc_ddrb erts_no_atomic_inc
-#define erts_atomic_dec_ddrb erts_no_atomic_dec
-#define erts_atomic_add_read_ddrb erts_no_atomic_add_read
-#define erts_atomic_add_ddrb erts_no_atomic_add
-#define erts_atomic_read_bor_ddrb erts_no_atomic_read_bor
-#define erts_atomic_read_band_ddrb erts_no_atomic_read_band
-#define erts_atomic_xchg_ddrb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_ddrb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_ddrb erts_no_atomic_read_bset
-
-#define erts_atomic_init_rb erts_no_atomic_set
-#define erts_atomic_set_rb erts_no_atomic_set
-#define erts_atomic_read_rb erts_no_atomic_read
-#define erts_atomic_inc_read_rb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_rb erts_no_atomic_dec_read
-#define erts_atomic_inc_rb erts_no_atomic_inc
-#define erts_atomic_dec_rb erts_no_atomic_dec
-#define erts_atomic_add_read_rb erts_no_atomic_add_read
-#define erts_atomic_add_rb erts_no_atomic_add
-#define erts_atomic_read_bor_rb erts_no_atomic_read_bor
-#define erts_atomic_read_band_rb erts_no_atomic_read_band
-#define erts_atomic_xchg_rb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_rb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_rb erts_no_atomic_read_bset
-
-#define erts_atomic_init_wb erts_no_atomic_set
-#define erts_atomic_set_wb erts_no_atomic_set
-#define erts_atomic_read_wb erts_no_atomic_read
-#define erts_atomic_inc_read_wb erts_no_atomic_inc_read
-#define erts_atomic_dec_read_wb erts_no_atomic_dec_read
-#define erts_atomic_inc_wb erts_no_atomic_inc
-#define erts_atomic_dec_wb erts_no_atomic_dec
-#define erts_atomic_add_read_wb erts_no_atomic_add_read
-#define erts_atomic_add_wb erts_no_atomic_add
-#define erts_atomic_read_bor_wb erts_no_atomic_read_bor
-#define erts_atomic_read_band_wb erts_no_atomic_read_band
-#define erts_atomic_xchg_wb erts_no_atomic_xchg
-#define erts_atomic_cmpxchg_wb erts_no_atomic_cmpxchg
-#define erts_atomic_read_bset_wb erts_no_atomic_read_bset
-
-#define erts_atomic_set_dirty erts_no_atomic_set
-#define erts_atomic_read_dirty erts_no_atomic_read
-
-/* 32-bit atomics */
-
-#define erts_atomic32_init_nob erts_no_atomic32_set
-#define erts_atomic32_set_nob erts_no_atomic32_set
-#define erts_atomic32_read_nob erts_no_atomic32_read
-#define erts_atomic32_inc_read_nob erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_nob erts_no_atomic32_dec_read
-#define erts_atomic32_inc_nob erts_no_atomic32_inc
-#define erts_atomic32_dec_nob erts_no_atomic32_dec
-#define erts_atomic32_add_read_nob erts_no_atomic32_add_read
-#define erts_atomic32_add_nob erts_no_atomic32_add
-#define erts_atomic32_read_bor_nob erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_nob erts_no_atomic32_read_band
-#define erts_atomic32_xchg_nob erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_nob erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_nob erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_mb erts_no_atomic32_set
-#define erts_atomic32_set_mb erts_no_atomic32_set
-#define erts_atomic32_read_mb erts_no_atomic32_read
-#define erts_atomic32_inc_read_mb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_mb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_mb erts_no_atomic32_inc
-#define erts_atomic32_dec_mb erts_no_atomic32_dec
-#define erts_atomic32_add_read_mb erts_no_atomic32_add_read
-#define erts_atomic32_add_mb erts_no_atomic32_add
-#define erts_atomic32_read_bor_mb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_mb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_mb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_mb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_mb erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_acqb erts_no_atomic32_set
-#define erts_atomic32_set_acqb erts_no_atomic32_set
-#define erts_atomic32_read_acqb erts_no_atomic32_read
-#define erts_atomic32_inc_read_acqb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_acqb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_acqb erts_no_atomic32_inc
-#define erts_atomic32_dec_acqb erts_no_atomic32_dec
-#define erts_atomic32_add_read_acqb erts_no_atomic32_add_read
-#define erts_atomic32_add_acqb erts_no_atomic32_add
-#define erts_atomic32_read_bor_acqb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_acqb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_acqb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_acqb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_acqb erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_relb erts_no_atomic32_set
-#define erts_atomic32_set_relb erts_no_atomic32_set
-#define erts_atomic32_read_relb erts_no_atomic32_read
-#define erts_atomic32_inc_read_relb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_relb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_relb erts_no_atomic32_inc
-#define erts_atomic32_dec_relb erts_no_atomic32_dec
-#define erts_atomic32_add_read_relb erts_no_atomic32_add_read
-#define erts_atomic32_add_relb erts_no_atomic32_add
-#define erts_atomic32_read_bor_relb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_relb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_relb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_relb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_relb erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_ddrb erts_no_atomic32_set
-#define erts_atomic32_set_ddrb erts_no_atomic32_set
-#define erts_atomic32_read_ddrb erts_no_atomic32_read
-#define erts_atomic32_inc_read_ddrb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_ddrb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_ddrb erts_no_atomic32_inc
-#define erts_atomic32_dec_ddrb erts_no_atomic32_dec
-#define erts_atomic32_add_read_ddrb erts_no_atomic32_add_read
-#define erts_atomic32_add_ddrb erts_no_atomic32_add
-#define erts_atomic32_read_bor_ddrb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_ddrb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_ddrb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_ddrb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_ddrb erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_rb erts_no_atomic32_set
-#define erts_atomic32_set_rb erts_no_atomic32_set
-#define erts_atomic32_read_rb erts_no_atomic32_read
-#define erts_atomic32_inc_read_rb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_rb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_rb erts_no_atomic32_inc
-#define erts_atomic32_dec_rb erts_no_atomic32_dec
-#define erts_atomic32_add_read_rb erts_no_atomic32_add_read
-#define erts_atomic32_add_rb erts_no_atomic32_add
-#define erts_atomic32_read_bor_rb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_rb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_rb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_rb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_rb erts_no_atomic32_read_bset
-
-#define erts_atomic32_init_wb erts_no_atomic32_set
-#define erts_atomic32_set_wb erts_no_atomic32_set
-#define erts_atomic32_read_wb erts_no_atomic32_read
-#define erts_atomic32_inc_read_wb erts_no_atomic32_inc_read
-#define erts_atomic32_dec_read_wb erts_no_atomic32_dec_read
-#define erts_atomic32_inc_wb erts_no_atomic32_inc
-#define erts_atomic32_dec_wb erts_no_atomic32_dec
-#define erts_atomic32_add_read_wb erts_no_atomic32_add_read
-#define erts_atomic32_add_wb erts_no_atomic32_add
-#define erts_atomic32_read_bor_wb erts_no_atomic32_read_bor
-#define erts_atomic32_read_band_wb erts_no_atomic32_read_band
-#define erts_atomic32_xchg_wb erts_no_atomic32_xchg
-#define erts_atomic32_cmpxchg_wb erts_no_atomic32_cmpxchg
-#define erts_atomic32_read_bset_wb erts_no_atomic32_read_bset
-
-#define erts_atomic32_set_dirty erts_no_atomic32_set
-#define erts_atomic32_read_dirty erts_no_atomic32_read
-
-/* 64-bit atomics */
-
-#define erts_atomic64_init_nob erts_no_atomic64_set
-#define erts_atomic64_set_nob erts_no_atomic64_set
-#define erts_atomic64_read_nob erts_no_atomic64_read
-#define erts_atomic64_inc_read_nob erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_nob erts_no_atomic64_dec_read
-#define erts_atomic64_inc_nob erts_no_atomic64_inc
-#define erts_atomic64_dec_nob erts_no_atomic64_dec
-#define erts_atomic64_add_read_nob erts_no_atomic64_add_read
-#define erts_atomic64_add_nob erts_no_atomic64_add
-#define erts_atomic64_read_bor_nob erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_nob erts_no_atomic64_read_band
-#define erts_atomic64_xchg_nob erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_nob erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_nob erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_mb erts_no_atomic64_set
-#define erts_atomic64_set_mb erts_no_atomic64_set
-#define erts_atomic64_read_mb erts_no_atomic64_read
-#define erts_atomic64_inc_read_mb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_mb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_mb erts_no_atomic64_inc
-#define erts_atomic64_dec_mb erts_no_atomic64_dec
-#define erts_atomic64_add_read_mb erts_no_atomic64_add_read
-#define erts_atomic64_add_mb erts_no_atomic64_add
-#define erts_atomic64_read_bor_mb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_mb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_mb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_mb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_mb erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_acqb erts_no_atomic64_set
-#define erts_atomic64_set_acqb erts_no_atomic64_set
-#define erts_atomic64_read_acqb erts_no_atomic64_read
-#define erts_atomic64_inc_read_acqb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_acqb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_acqb erts_no_atomic64_inc
-#define erts_atomic64_dec_acqb erts_no_atomic64_dec
-#define erts_atomic64_add_read_acqb erts_no_atomic64_add_read
-#define erts_atomic64_add_acqb erts_no_atomic64_add
-#define erts_atomic64_read_bor_acqb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_acqb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_acqb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_acqb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_acqb erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_relb erts_no_atomic64_set
-#define erts_atomic64_set_relb erts_no_atomic64_set
-#define erts_atomic64_read_relb erts_no_atomic64_read
-#define erts_atomic64_inc_read_relb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_relb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_relb erts_no_atomic64_inc
-#define erts_atomic64_dec_relb erts_no_atomic64_dec
-#define erts_atomic64_add_read_relb erts_no_atomic64_add_read
-#define erts_atomic64_add_relb erts_no_atomic64_add
-#define erts_atomic64_read_bor_relb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_relb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_relb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_relb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_relb erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_ddrb erts_no_atomic64_set
-#define erts_atomic64_set_ddrb erts_no_atomic64_set
-#define erts_atomic64_read_ddrb erts_no_atomic64_read
-#define erts_atomic64_inc_read_ddrb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_ddrb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_ddrb erts_no_atomic64_inc
-#define erts_atomic64_dec_ddrb erts_no_atomic64_dec
-#define erts_atomic64_add_read_ddrb erts_no_atomic64_add_read
-#define erts_atomic64_add_ddrb erts_no_atomic64_add
-#define erts_atomic64_read_bor_ddrb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_ddrb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_ddrb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_ddrb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_ddrb erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_rb erts_no_atomic64_set
-#define erts_atomic64_set_rb erts_no_atomic64_set
-#define erts_atomic64_read_rb erts_no_atomic64_read
-#define erts_atomic64_inc_read_rb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_rb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_rb erts_no_atomic64_inc
-#define erts_atomic64_dec_rb erts_no_atomic64_dec
-#define erts_atomic64_add_read_rb erts_no_atomic64_add_read
-#define erts_atomic64_add_rb erts_no_atomic64_add
-#define erts_atomic64_read_bor_rb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_rb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_rb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_rb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_rb erts_no_atomic64_read_bset
-
-#define erts_atomic64_init_wb erts_no_atomic64_set
-#define erts_atomic64_set_wb erts_no_atomic64_set
-#define erts_atomic64_read_wb erts_no_atomic64_read
-#define erts_atomic64_inc_read_wb erts_no_atomic64_inc_read
-#define erts_atomic64_dec_read_wb erts_no_atomic64_dec_read
-#define erts_atomic64_inc_wb erts_no_atomic64_inc
-#define erts_atomic64_dec_wb erts_no_atomic64_dec
-#define erts_atomic64_add_read_wb erts_no_atomic64_add_read
-#define erts_atomic64_add_wb erts_no_atomic64_add
-#define erts_atomic64_read_bor_wb erts_no_atomic64_read_bor
-#define erts_atomic64_read_band_wb erts_no_atomic64_read_band
-#define erts_atomic64_xchg_wb erts_no_atomic64_xchg
-#define erts_atomic64_cmpxchg_wb erts_no_atomic64_cmpxchg
-#define erts_atomic64_read_bset_wb erts_no_atomic64_read_bset
-
-#define erts_atomic64_set_dirty erts_no_atomic64_set
-#define erts_atomic64_read_dirty erts_no_atomic64_read
-
-#endif /* !USE_THREADS */
 
 #include "erl_msacc.h"
 
@@ -2065,110 +1526,83 @@ erts_atomic64_read_dirty(erts_atomic64_t *var)
 ERTS_GLB_INLINE void
 erts_thr_init(erts_thr_init_data_t *id)
 {
-#ifdef USE_THREADS
     int res = ethr_init(id);
     if (res)
 	erts_thr_fatal_error(res, "initialize thread library");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_late_init(erts_thr_late_init_data_t *id)
 {
-#ifdef USE_THREADS
     int res = ethr_late_init(id);
     if (res)
 	erts_thr_fatal_error(res, "complete initialization of thread library");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_create(erts_tid_t *tid, void * (*func)(void *), void *arg,
 		erts_thr_opts_t *opts)
 {
-#ifdef USE_THREADS
     int res = ethr_thr_create(tid, func, arg, opts);
     if (res)
 	erts_thr_fatal_error(res, "create thread");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_join(erts_tid_t tid, void **thr_res)
 {
-#ifdef USE_THREADS
     int res = ethr_thr_join(tid, thr_res);
     if (res)
 	erts_thr_fatal_error(res, "join thread");
-#endif
 }
 
 
 ERTS_GLB_INLINE void
 erts_thr_detach(erts_tid_t tid)
 {
-#ifdef USE_THREADS
     int res = ethr_thr_detach(tid);
     if (res)
 	erts_thr_fatal_error(res, "detach thread");
-#endif
 }
 
 
 ERTS_GLB_INLINE void
 erts_thr_exit(void *res)
 {
-#ifdef USE_THREADS
     ethr_thr_exit(res);
     erts_thr_fatal_error(0, "terminate thread");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_install_exit_handler(void (*exit_handler)(void))
 {
-#ifdef USE_THREADS
     int res = ethr_install_exit_handler(exit_handler);
     if (res != 0)
 	erts_thr_fatal_error(res, "install thread exit handler");
-#endif
 }
 
 ERTS_GLB_INLINE erts_tid_t
 erts_thr_self(void)
 {
-#ifdef USE_THREADS
     return ethr_self();
-#else
-    return 0;
-#endif
 }
 
 ERTS_GLB_INLINE int
 erts_thr_getname(erts_tid_t tid, char *buf, size_t len)
 {
-#ifdef USE_THREADS
     return ethr_getname(tid, buf, len);
-#else
-    return -1;
-#endif
 }
 
 
 ERTS_GLB_INLINE int
 erts_equal_tids(erts_tid_t x, erts_tid_t y)
 {
-#ifdef USE_THREADS
     return ethr_equal_tids(x, y);
-#else
-    return 1;
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_mtx_init(erts_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t flags)
 {
-#ifdef USE_THREADS
     int res = ethr_mutex_init(&mtx->mtx);
     if (res) {
         erts_thr_fatal_error(res, "initialize mutex");
@@ -2185,13 +1619,11 @@ erts_mtx_init(erts_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t flags)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_init_ref_x(&mtx->lcnt, name, extra, flags);
 #endif
-#endif /* USE_THREADS */
 }
 
 ERTS_GLB_INLINE void
 erts_mtx_init_locked(erts_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t flags)
 {
-#ifdef USE_THREADS
     erts_mtx_init(mtx, name, extra, flags);
 
     ethr_mutex_lock(&mtx->mtx);
@@ -2201,13 +1633,11 @@ erts_mtx_init_locked(erts_mtx_t *mtx, char *name, Eterm extra, erts_lock_flags_t
     #ifdef ERTS_ENABLE_LOCK_COUNT
         erts_lcnt_trylock(&mtx->lcnt, 1);
     #endif
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_mtx_destroy(erts_mtx_t *mtx)
 {
-#ifdef USE_THREADS
     int res;
 
     ASSERT(!(mtx->flags & ERTS_LOCK_FLAGS_PROPERTY_STATIC));
@@ -2230,7 +1660,6 @@ erts_mtx_destroy(erts_mtx_t *mtx)
 #endif
 	    erts_thr_fatal_error(res, "destroy mutex");
     }
-#endif
 }
 
 ERTS_GLB_INLINE int
@@ -2240,7 +1669,6 @@ erts_mtx_trylock_x(erts_mtx_t *mtx, char *file, unsigned int line)
 erts_mtx_trylock(erts_mtx_t *mtx)
 #endif
 {
-#ifdef USE_THREADS
     int res;
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
@@ -2262,9 +1690,6 @@ erts_mtx_trylock(erts_mtx_t *mtx)
     erts_lcnt_trylock(&mtx->lcnt, res);
 #endif    
     return res;
-#else
-    return 0;
-#endif
 
 }
 
@@ -2275,7 +1700,6 @@ erts_mtx_lock_x(erts_mtx_t *mtx, char *file, unsigned int line)
 erts_mtx_lock(erts_mtx_t *mtx)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_x(&mtx->lc, file, line);
@@ -2290,13 +1714,11 @@ erts_mtx_lock(erts_mtx_t *mtx)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&mtx->lcnt, file, line);
 #endif
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_mtx_unlock(erts_mtx_t *mtx)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock(&mtx->lc);
 #endif
@@ -2304,13 +1726,12 @@ erts_mtx_unlock(erts_mtx_t *mtx)
     erts_lcnt_unlock(&mtx->lcnt);
 #endif
     ethr_mutex_unlock(&mtx->mtx);
-#endif
 }
 
 ERTS_GLB_INLINE int
 erts_lc_mtx_is_locked(erts_mtx_t *mtx)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = mtx->lc;
     lc.flags = ERTS_LOCK_FLAGS_TYPE_MUTEX;
@@ -2325,17 +1746,14 @@ erts_lc_mtx_is_locked(erts_mtx_t *mtx)
 ERTS_GLB_INLINE void
 erts_cnd_init(erts_cnd_t *cnd)
 {
-#ifdef USE_THREADS
     int res = ethr_cond_init(cnd);
     if (res)
 	erts_thr_fatal_error(res, "initialize condition variable");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_cnd_destroy(erts_cnd_t *cnd)
 {
-#ifdef USE_THREADS
     int res = ethr_cond_destroy(cnd);
     if (res != 0) {
 #ifdef ERTS_THR_HAVE_BUSY_DESTROY_BUG
@@ -2348,13 +1766,11 @@ erts_cnd_destroy(erts_cnd_t *cnd)
 #endif
 	    erts_thr_fatal_error(res, "destroy condition variable");
     }
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_cnd_wait(erts_cnd_t *cnd, erts_mtx_t *mtx)
 {
-#ifdef USE_THREADS
     int res;
     ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_SLEEP);
 #ifdef ERTS_ENABLE_LOCK_CHECK
@@ -2376,7 +1792,6 @@ erts_cnd_wait(erts_cnd_t *cnd, erts_mtx_t *mtx)
     if (res != 0 && res != EINTR)
 	erts_thr_fatal_error(res, "wait on condition variable");
     ERTS_MSACC_POP_STATE();
-#endif
 }
 
 /*
@@ -2392,18 +1807,14 @@ erts_cnd_wait(erts_cnd_t *cnd, erts_mtx_t *mtx)
 ERTS_GLB_INLINE void
 erts_cnd_signal(erts_cnd_t *cnd)
 {
-#ifdef USE_THREADS
     ethr_cond_signal(cnd);
-#endif
 }
 
 
 ERTS_GLB_INLINE void
 erts_cnd_broadcast(erts_cnd_t *cnd)
 {
-#ifdef USE_THREADS
     ethr_cond_broadcast(cnd);
-#endif
 }
 
 /* rwmutex */
@@ -2411,7 +1822,6 @@ erts_cnd_broadcast(erts_cnd_t *cnd)
 ERTS_GLB_INLINE void
 erts_rwmtx_set_reader_group(int no)
 {
-#ifdef USE_THREADS
     int res;
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_check_no_locked_of_type(ERTS_LOCK_TYPE_RWMUTEX);
@@ -2419,13 +1829,11 @@ erts_rwmtx_set_reader_group(int no)
     res = ethr_rwmutex_set_reader_group(no);
     if (res != 0)
 	erts_thr_fatal_error(res, "set reader group");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_rwmtx_init_opt(erts_rwmtx_t *rwmtx, erts_rwmtx_opt_t *opt,
                     char *name, Eterm extra, erts_lock_flags_t flags) {
-#ifdef USE_THREADS
     int res = ethr_rwmutex_init_opt(&rwmtx->rwmtx, opt);
     if (res != 0) {
         erts_thr_fatal_error(res, "initialize rwmutex");
@@ -2442,7 +1850,6 @@ erts_rwmtx_init_opt(erts_rwmtx_t *rwmtx, erts_rwmtx_opt_t *opt,
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_init_ref_x(&rwmtx->lcnt, name, extra, flags);
 #endif
-#endif /* USE_THREADS */
 }
 
 ERTS_GLB_INLINE void
@@ -2454,7 +1861,6 @@ erts_rwmtx_init(erts_rwmtx_t *rwmtx, char *name, Eterm extra,
 ERTS_GLB_INLINE void
 erts_rwmtx_destroy(erts_rwmtx_t *rwmtx)
 {
-#ifdef USE_THREADS
     int res;
 
     ASSERT(!(rwmtx->flags & ERTS_LOCK_FLAGS_PROPERTY_STATIC));
@@ -2477,7 +1883,6 @@ erts_rwmtx_destroy(erts_rwmtx_t *rwmtx)
 #endif
 	    erts_thr_fatal_error(res, "destroy rwmutex");
     }
-#endif
 }
 
 ERTS_GLB_INLINE int
@@ -2487,7 +1892,6 @@ erts_rwmtx_tryrlock_x(erts_rwmtx_t *rwmtx, char *file, unsigned int line)
 erts_rwmtx_tryrlock(erts_rwmtx_t *rwmtx)
 #endif
 {
-#ifdef USE_THREADS
     int res;
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
@@ -2510,9 +1914,6 @@ erts_rwmtx_tryrlock(erts_rwmtx_t *rwmtx)
 #endif
     
     return res;
-#else
-    return 0;
-#endif
 }
 
 ERTS_GLB_INLINE void
@@ -2522,7 +1923,6 @@ erts_rwmtx_rlock_x(erts_rwmtx_t *rwmtx, char *file, unsigned int line)
 erts_rwmtx_rlock(erts_rwmtx_t *rwmtx)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_flg_x(&rwmtx->lc, ERTS_LOCK_OPTIONS_READ,file,line);
@@ -2537,13 +1937,11 @@ erts_rwmtx_rlock(erts_rwmtx_t *rwmtx)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&rwmtx->lcnt, file, line);
 #endif
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_rwmtx_runlock(erts_rwmtx_t *rwmtx)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock_flg(&rwmtx->lc, ERTS_LOCK_OPTIONS_READ);
 #endif
@@ -2551,7 +1949,6 @@ erts_rwmtx_runlock(erts_rwmtx_t *rwmtx)
     erts_lcnt_unlock_opt(&rwmtx->lcnt, ERTS_LOCK_OPTIONS_READ);
 #endif
     ethr_rwmutex_runlock(&rwmtx->rwmtx);
-#endif
 }
 
 
@@ -2562,7 +1959,6 @@ erts_rwmtx_tryrwlock_x(erts_rwmtx_t *rwmtx, char *file, unsigned int line)
 erts_rwmtx_tryrwlock(erts_rwmtx_t *rwmtx)
 #endif
 {
-#ifdef USE_THREADS
     int res;
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
@@ -2585,9 +1981,6 @@ erts_rwmtx_tryrwlock(erts_rwmtx_t *rwmtx)
 #endif
     
     return res;
-#else
-    return 0;
-#endif
 }
 
 ERTS_GLB_INLINE void
@@ -2597,7 +1990,6 @@ erts_rwmtx_rwlock_x(erts_rwmtx_t *rwmtx, char *file, unsigned int line)
 erts_rwmtx_rwlock(erts_rwmtx_t *rwmtx)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_flg_x(&rwmtx->lc, ERTS_LOCK_OPTIONS_RDWR,file,line);
@@ -2612,13 +2004,11 @@ erts_rwmtx_rwlock(erts_rwmtx_t *rwmtx)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&rwmtx->lcnt, file, line);
 #endif
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_rwmtx_rwunlock(erts_rwmtx_t *rwmtx)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock_flg(&rwmtx->lc, ERTS_LOCK_OPTIONS_RDWR);
 #endif
@@ -2626,7 +2016,6 @@ erts_rwmtx_rwunlock(erts_rwmtx_t *rwmtx)
     erts_lcnt_unlock_opt(&rwmtx->lcnt, ERTS_LOCK_OPTIONS_RDWR);
 #endif
     ethr_rwmutex_rwunlock(&rwmtx->rwmtx);
-#endif
 }
 
 #if 0 /* The following rwmtx function names are
@@ -2658,7 +2047,7 @@ erts_rwmtx_wunlock(erts_rwmtx_t *rwmtx)
 ERTS_GLB_INLINE int
 erts_lc_rwmtx_is_rlocked(erts_rwmtx_t *mtx)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = mtx->lc;
     lc.flags = ERTS_LOCK_TYPE_RWMUTEX;
@@ -2673,7 +2062,7 @@ erts_lc_rwmtx_is_rlocked(erts_rwmtx_t *mtx)
 ERTS_GLB_INLINE int
 erts_lc_rwmtx_is_rwlocked(erts_rwmtx_t *mtx)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = mtx->lc;
     lc.flags = ERTS_LOCK_TYPE_RWMUTEX;
@@ -2685,334 +2074,11 @@ erts_lc_rwmtx_is_rwlocked(erts_rwmtx_t *mtx)
 #endif
 }
 
-/* No atomic ops */
-
-ERTS_GLB_INLINE void
-erts_no_dw_atomic_set(erts_no_dw_atomic_t *var, erts_no_dw_atomic_t *val)
-{
-    var->sint[0] = val->sint[0];
-    var->sint[1] = val->sint[1];
-}
-
-ERTS_GLB_INLINE void
-erts_no_dw_atomic_read(erts_no_dw_atomic_t *var, erts_no_dw_atomic_t *val)
-{
-    val->sint[0] = var->sint[0];
-    val->sint[1] = var->sint[1];
-}
-
-ERTS_GLB_INLINE int erts_no_dw_atomic_cmpxchg(erts_no_dw_atomic_t *var,
-					      erts_no_dw_atomic_t *new_val,
-					      erts_no_dw_atomic_t *old_val)
-{
-    if (var->sint[0] != old_val->sint[0] || var->sint[1] != old_val->sint[1]) {
-	erts_no_dw_atomic_read(var, old_val);
-	return 0;
-    }
-    else {
-	erts_no_dw_atomic_set(var, new_val);
-	return !0;
-    }
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic_set(erts_no_atomic_t *var, erts_aint_t i)
-{
-    *var = i;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_read(erts_no_atomic_t *var)
-{
-    return *var;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_inc_read(erts_no_atomic_t *incp)
-{
-    return ++(*incp);
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_dec_read(erts_no_atomic_t *decp)
-{
-    return --(*decp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic_inc(erts_no_atomic_t *incp)
-{
-    ++(*incp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic_dec(erts_no_atomic_t *decp)
-{
-    --(*decp);
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_add_read(erts_no_atomic_t *addp, erts_aint_t i)
-{
-    return *addp += i;
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic_add(erts_no_atomic_t *addp, erts_aint_t i)
-{
-    *addp += i;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_read_bor(erts_no_atomic_t *var, erts_aint_t mask)
-{
-    erts_aint_t old;
-    old = *var;
-    *var |= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_read_band(erts_no_atomic_t *var, erts_aint_t mask)
-{
-    erts_aint_t old;
-    old = *var;
-    *var &= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_xchg(erts_no_atomic_t *xchgp, erts_aint_t new)
-{
-    erts_aint_t old = *xchgp;
-    *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_cmpxchg(erts_no_atomic_t *xchgp,
-		       erts_aint_t new,
-		       erts_aint_t expected)
-{
-    erts_aint_t old = *xchgp;
-    if (old == expected)
-        *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint_t
-erts_no_atomic_read_bset(erts_no_atomic_t *var,
-			 erts_aint_t mask,
-			 erts_aint_t set)
-{
-    erts_aint_t old = *var;
-    *var &= ~mask;
-    *var |= (mask & set);
-    return old;
-}
-
-/* atomic32 */
-
-ERTS_GLB_INLINE void
-erts_no_atomic32_set(erts_no_atomic32_t *var, erts_aint32_t i)
-{
-    *var = i;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_read(erts_no_atomic32_t *var)
-{
-    return *var;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_inc_read(erts_no_atomic32_t *incp)
-{
-    return ++(*incp);
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_dec_read(erts_no_atomic32_t *decp)
-{
-    return --(*decp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic32_inc(erts_no_atomic32_t *incp)
-{
-    ++(*incp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic32_dec(erts_no_atomic32_t *decp)
-{
-    --(*decp);
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_add_read(erts_no_atomic32_t *addp, erts_aint32_t i)
-{
-    return *addp += i;
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic32_add(erts_no_atomic32_t *addp, erts_aint32_t i)
-{
-    *addp += i;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_read_bor(erts_no_atomic32_t *var, erts_aint32_t mask)
-{
-    erts_aint32_t old;
-    old = *var;
-    *var |= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_read_band(erts_no_atomic32_t *var, erts_aint32_t mask)
-{
-    erts_aint32_t old;
-    old = *var;
-    *var &= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_xchg(erts_no_atomic32_t *xchgp, erts_aint32_t new)
-{
-    erts_aint32_t old = *xchgp;
-    *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_cmpxchg(erts_no_atomic32_t *xchgp,
-			 erts_aint32_t new,
-			 erts_aint32_t expected)
-{
-    erts_aint32_t old = *xchgp;
-    if (old == expected)
-        *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint32_t
-erts_no_atomic32_read_bset(erts_no_atomic32_t *var,
-			   erts_aint32_t mask,
-			   erts_aint32_t set)
-{
-    erts_aint32_t old = *var;
-    *var &= ~mask;
-    *var |= (mask & set);
-    return old;
-}
-
-/* atomic64 */
-
-ERTS_GLB_INLINE void
-erts_no_atomic64_set(erts_no_atomic64_t *var, erts_aint64_t i)
-{
-    *var = i;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_read(erts_no_atomic64_t *var)
-{
-    return *var;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_inc_read(erts_no_atomic64_t *incp)
-{
-    return ++(*incp);
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_dec_read(erts_no_atomic64_t *decp)
-{
-    return --(*decp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic64_inc(erts_no_atomic64_t *incp)
-{
-    ++(*incp);
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic64_dec(erts_no_atomic64_t *decp)
-{
-    --(*decp);
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_add_read(erts_no_atomic64_t *addp, erts_aint64_t i)
-{
-    return *addp += i;
-}
-
-ERTS_GLB_INLINE void
-erts_no_atomic64_add(erts_no_atomic64_t *addp, erts_aint64_t i)
-{
-    *addp += i;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_read_bor(erts_no_atomic64_t *var, erts_aint64_t mask)
-{
-    erts_aint64_t old;
-    old = *var;
-    *var |= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_read_band(erts_no_atomic64_t *var, erts_aint64_t mask)
-{
-    erts_aint64_t old;
-    old = *var;
-    *var &= mask;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_xchg(erts_no_atomic64_t *xchgp, erts_aint64_t new)
-{
-    erts_aint64_t old = *xchgp;
-    *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_cmpxchg(erts_no_atomic64_t *xchgp,
-			 erts_aint64_t new,
-			 erts_aint64_t expected)
-{
-    erts_aint64_t old = *xchgp;
-    if (old == expected)
-        *xchgp = new;
-    return old;
-}
-
-ERTS_GLB_INLINE erts_aint64_t
-erts_no_atomic64_read_bset(erts_no_atomic64_t *var,
-			   erts_aint64_t mask,
-			   erts_aint64_t set)
-{
-    erts_aint64_t old = *var;
-    *var &= ~mask;
-    *var |= (mask & set);
-    return old;
-}
-
 /* spinlock */
 
 ERTS_GLB_INLINE void
 erts_spinlock_init(erts_spinlock_t *lock, char *name, Eterm extra, erts_lock_flags_t flags)
 {
-#ifdef USE_THREADS
     int res = ethr_spinlock_init(&lock->slck);
     if (res) {
         erts_thr_fatal_error(res, "init spinlock");
@@ -3029,13 +2095,11 @@ erts_spinlock_init(erts_spinlock_t *lock, char *name, Eterm extra, erts_lock_fla
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_init_ref_x(&lock->lcnt, name, extra, flags);
 #endif
-#endif /* USE_THREADS */
 }
 
 ERTS_GLB_INLINE void
 erts_spinlock_destroy(erts_spinlock_t *lock)
 {
-#ifdef USE_THREADS
     int res;
 
     ASSERT(!(lock->flags & ERTS_LOCK_FLAGS_PROPERTY_STATIC));
@@ -3058,15 +2122,11 @@ erts_spinlock_destroy(erts_spinlock_t *lock)
 #endif
 	    erts_thr_fatal_error(res, "destroy rwlock");
     }
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_spin_unlock(erts_spinlock_t *lock)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock(&lock->lc);
 #endif
@@ -3074,9 +2134,6 @@ erts_spin_unlock(erts_spinlock_t *lock)
     erts_lcnt_unlock(&lock->lcnt);
 #endif
     ethr_spin_unlock(&lock->slck);
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
@@ -3086,7 +2143,6 @@ erts_spin_lock_x(erts_spinlock_t *lock, char *file, unsigned int line)
 erts_spin_lock(erts_spinlock_t *lock)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_x(&lock->lc,file,line);
@@ -3101,15 +2157,12 @@ erts_spin_lock(erts_spinlock_t *lock)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&lock->lcnt, file, line);
 #endif
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE int
 erts_lc_spinlock_is_locked(erts_spinlock_t *lock)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = lock->lc;
     lc.flags = ERTS_LOCK_TYPE_SPINLOCK;
@@ -3126,7 +2179,6 @@ erts_lc_spinlock_is_locked(erts_spinlock_t *lock)
 ERTS_GLB_INLINE void
 erts_rwlock_init(erts_rwlock_t *lock, char *name, Eterm extra, erts_lock_flags_t flags)
 {
-#ifdef USE_THREADS
     int res = ethr_rwlock_init(&lock->rwlck);
     if (res) {
         erts_thr_fatal_error(res, "init rwlock");
@@ -3143,13 +2195,11 @@ erts_rwlock_init(erts_rwlock_t *lock, char *name, Eterm extra, erts_lock_flags_t
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_init_ref_x(&lock->lcnt, name, extra, flags);
 #endif
-#endif /* USE_THREADS */
 }
 
 ERTS_GLB_INLINE void
 erts_rwlock_destroy(erts_rwlock_t *lock)
 {
-#ifdef USE_THREADS
     int res;
 
     ASSERT(!(lock->flags & ERTS_LOCK_FLAGS_PROPERTY_STATIC));
@@ -3172,15 +2222,11 @@ erts_rwlock_destroy(erts_rwlock_t *lock)
 #endif
 	    erts_thr_fatal_error(res, "destroy rwlock");
     }
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_read_unlock(erts_rwlock_t *lock)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock_flg(&lock->lc, ERTS_LOCK_OPTIONS_READ);
 #endif
@@ -3188,9 +2234,6 @@ erts_read_unlock(erts_rwlock_t *lock)
     erts_lcnt_unlock_opt(&lock->lcnt, ERTS_LOCK_OPTIONS_READ);
 #endif
     ethr_read_unlock(&lock->rwlck);
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
@@ -3200,7 +2243,6 @@ erts_read_lock_x(erts_rwlock_t *lock, char *file, unsigned int line)
 erts_read_lock(erts_rwlock_t *lock)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_flg_x(&lock->lc, ERTS_LOCK_OPTIONS_READ,file,line);
@@ -3215,15 +2257,11 @@ erts_read_lock(erts_rwlock_t *lock)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&lock->lcnt, file, line);
 #endif
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_write_unlock(erts_rwlock_t *lock)
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_unlock_flg(&lock->lc, ERTS_LOCK_OPTIONS_RDWR);
 #endif
@@ -3231,9 +2269,6 @@ erts_write_unlock(erts_rwlock_t *lock)
     erts_lcnt_unlock_opt(&lock->lcnt, ERTS_LOCK_OPTIONS_RDWR);
 #endif
     ethr_write_unlock(&lock->rwlck);
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE void
@@ -3243,7 +2278,6 @@ erts_write_lock_x(erts_rwlock_t *lock, char *file, unsigned int line)
 erts_write_lock(erts_rwlock_t *lock)
 #endif
 {
-#ifdef USE_THREADS
 #ifdef ERTS_ENABLE_LOCK_CHECK
 #ifdef ERTS_ENABLE_LOCK_POSITION
     erts_lc_lock_flg_x(&lock->lc, ERTS_LOCK_OPTIONS_RDWR,file,line);
@@ -3258,15 +2292,12 @@ erts_write_lock(erts_rwlock_t *lock)
 #ifdef ERTS_ENABLE_LOCK_COUNT
     erts_lcnt_lock_post_x(&lock->lcnt, file, line);
 #endif
-#else
-    (void)lock;
-#endif
 }
 
 ERTS_GLB_INLINE int
 erts_lc_rwlock_is_rlocked(erts_rwlock_t *lock)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = lock->lc;
     lc.flags = ERTS_LOCK_TYPE_RWSPINLOCK;
@@ -3281,7 +2312,7 @@ erts_lc_rwlock_is_rlocked(erts_rwlock_t *lock)
 ERTS_GLB_INLINE int
 erts_lc_rwlock_is_rwlocked(erts_rwlock_t *lock)
 {
-#if defined(USE_THREADS) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     int res;
     erts_lc_lock_t lc = lock->lc;
     lc.flags = ERTS_LOCK_TYPE_RWSPINLOCK;
@@ -3296,125 +2327,90 @@ erts_lc_rwlock_is_rwlocked(erts_rwlock_t *lock)
 ERTS_GLB_INLINE void
 erts_tsd_key_create(erts_tsd_key_t *keyp, char *keyname)
 {
-#ifdef USE_THREADS
     int res = ethr_tsd_key_create(keyp, keyname);
     if (res)
 	erts_thr_fatal_error(res, "create thread specific data key");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_tsd_key_delete(erts_tsd_key_t key)
 {
-#ifdef USE_THREADS
     int res = ethr_tsd_key_delete(key);
     if (res)
 	erts_thr_fatal_error(res, "delete thread specific data key");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_tsd_set(erts_tsd_key_t key, void *value)
 {
-#ifdef USE_THREADS
     int res = ethr_tsd_set(key, value);
     if (res)
 	erts_thr_fatal_error(res, "set thread specific data");
-#endif
 }
 
 ERTS_GLB_INLINE void *
 erts_tsd_get(erts_tsd_key_t key)
 {
-#ifdef USE_THREADS
     return ethr_tsd_get(key);
-#else
-    return NULL;
-#endif
 }
 
 ERTS_GLB_INLINE erts_tse_t *erts_tse_fetch(void)
 {
-#ifdef USE_THREADS
     return (erts_tse_t *) ethr_get_ts_event();
-#else
-    return (erts_tse_t *) NULL;
-#endif
 }
 
 ERTS_GLB_INLINE void erts_tse_return(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     ethr_leave_ts_event(ep);
-#endif
 }
 
 ERTS_GLB_INLINE void erts_tse_prepare_timed(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     int res = ethr_event_prepare_timed(&((ethr_ts_event *) ep)->event);
     if (res != 0)
 	erts_thr_fatal_error(res, "prepare timed");
-#endif
 }
 
 ERTS_GLB_INLINE void erts_tse_set(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     ethr_event_set(&((ethr_ts_event *) ep)->event);
-#endif
 }
 
 ERTS_GLB_INLINE void erts_tse_reset(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     ethr_event_reset(&((ethr_ts_event *) ep)->event);
-#endif
 }
 
 ERTS_GLB_INLINE int erts_tse_wait(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     int res;
     ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_SLEEP);
     res = ethr_event_wait(&((ethr_ts_event *) ep)->event);
     ERTS_MSACC_POP_STATE();
     return res;
-#else
-    return ENOTSUP;
-#endif
 }
 
 ERTS_GLB_INLINE int erts_tse_swait(erts_tse_t *ep, int spincount)
 {
-#ifdef USE_THREADS
     int res;
     ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_SLEEP);
     res = ethr_event_swait(&((ethr_ts_event *) ep)->event, spincount);
     ERTS_MSACC_POP_STATE();
     return res;
-#else
-    return ENOTSUP;
-#endif
 }
 
 ERTS_GLB_INLINE int erts_tse_twait(erts_tse_t *ep, Sint64 tmo)
 {
-#ifdef USE_THREADS
     int res;
     ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_SLEEP);
     res = ethr_event_twait(&((ethr_ts_event *) ep)->event,
                            (ethr_sint64_t) tmo);
     ERTS_MSACC_POP_STATE();
     return res;
-#else
-    return ENOTSUP;
-#endif
 }
 
 ERTS_GLB_INLINE int erts_tse_stwait(erts_tse_t *ep, int spincount, Sint64 tmo)
 {
-#ifdef USE_THREADS
     int res;
     ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_SLEEP);
     res = ethr_event_stwait(&((ethr_ts_event *) ep)->event,
@@ -3422,49 +2418,34 @@ ERTS_GLB_INLINE int erts_tse_stwait(erts_tse_t *ep, int spincount, Sint64 tmo)
                             (ethr_sint64_t) tmo);
     ERTS_MSACC_POP_STATE();
     return res;
-#else
-    return ENOTSUP;
-#endif
 }
 
 ERTS_GLB_INLINE int erts_tse_is_tmp(erts_tse_t *ep)
 {
-#ifdef USE_THREADS
     return (ep->iflgs & ETHR_TS_EV_TMP) == ETHR_TS_EV_TMP;
-#else
-    return 0;
-#endif
 }
 
 ERTS_GLB_INLINE void erts_thr_set_main_status(int on, int no)
 {
-#ifdef USE_THREADS
     int res = ethr_set_main_thr_status(on, no);
     if (res != 0)
 	erts_thr_fatal_error(res, "set thread main status");
-#endif
 }
 
 ERTS_GLB_INLINE int erts_thr_get_main_status(void)
 {
-#ifdef USE_THREADS
     int main_status;
     int res = ethr_get_main_thr_status(&main_status);
     if (res != 0)
 	erts_thr_fatal_error(res, "get thread main status");
     return main_status;
-#else
-    return 1;
-#endif
 }
 
 ERTS_GLB_INLINE void erts_thr_yield(void)
 {
-#ifdef USE_THREADS
     int res = ETHR_YIELD();
     if (res != 0)
 	erts_thr_fatal_error(res, "yield");
-#endif    
 }
 
 
@@ -3472,34 +2453,28 @@ ERTS_GLB_INLINE void erts_thr_yield(void)
 
 ERTS_GLB_INLINE void
 erts_thr_kill(erts_tid_t tid, int sig) {
-#ifdef USE_THREADS
   int res = ethr_kill((ethr_tid)tid, sig);
   if (res)
     erts_thr_fatal_error(res, "killing thread");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_sigmask(int how, const sigset_t *set, sigset_t *oset)
 {
-#ifdef USE_THREADS
     int res = ethr_sigmask(how, set, oset);
     if (res)
 	erts_thr_fatal_error(res, "get or set signal mask");
-#endif
 }
 
 ERTS_GLB_INLINE void
 erts_thr_sigwait(const sigset_t *set, int *sig)
 {
-#ifdef USE_THREADS
     int res;
     do {
 	res = ethr_sigwait(set, sig);
     } while (res == EINTR);
     if (res)
 	erts_thr_fatal_error(res, "to wait for signal");
-#endif
 }
 
 #endif /* #ifdef HAVE_ETHR_SIG_FUNCS */
@@ -3507,37 +2482,3 @@ erts_thr_sigwait(const sigset_t *set, int *sig)
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
 #endif /* #ifndef ERL_THREAD_H__ */
-
-#ifdef ERTS_UNDEF_DEPRECATED_ATOMICS
-
-/* Deprecated functions to replace */
-
-#undef erts_atomic_init
-#undef erts_atomic_set
-#undef erts_atomic_read
-#undef erts_atomic_inctest
-#undef erts_atomic_dectest
-#undef erts_atomic_inc
-#undef erts_atomic_dec
-#undef erts_atomic_addtest
-#undef erts_atomic_add
-#undef erts_atomic_xchg
-#undef erts_atomic_cmpxchg
-#undef erts_atomic_bor
-#undef erts_atomic_band
-
-#undef erts_atomic32_init
-#undef erts_atomic32_set
-#undef erts_atomic32_read
-#undef erts_atomic32_inctest
-#undef erts_atomic32_dectest
-#undef erts_atomic32_inc
-#undef erts_atomic32_dec
-#undef erts_atomic32_addtest
-#undef erts_atomic32_add
-#undef erts_atomic32_xchg
-#undef erts_atomic32_cmpxchg
-#undef erts_atomic32_bor
-#undef erts_atomic32_band
-
-#endif
diff --git a/erts/emulator/beam/erl_time.h b/erts/emulator/beam/erl_time.h
index 27164d50a0..968f21fd51 100644
--- a/erts/emulator/beam/erl_time.h
+++ b/erts/emulator/beam/erl_time.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2006-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2006-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -21,6 +21,8 @@
 #ifndef ERL_TIME_H__
 #define ERL_TIME_H__
 
+#include "erl_monitor_link.h"
+
 #if 0
 #  define ERTS_TW_DEBUG
 #endif
@@ -79,8 +81,8 @@ typedef ErtsMonotonicTime * ErtsNextTimeoutRef;
 extern SysTimeval erts_first_emu_time;
 
 
-void erts_monitor_time_offset(Eterm id, Eterm ref);
-int erts_demonitor_time_offset(Eterm ref);
+void erts_monitor_time_offset(ErtsMonitor *mon);
+void erts_demonitor_time_offset(ErtsMonitor *mon);
 
 int erts_init_time_sup(int, ErtsTimeWarpMode);
 void erts_late_init_time_sup(void);
@@ -107,9 +109,6 @@ void erts_p_slpq(void);
 void erts_get_now_cpu(Uint* megasec, Uint* sec, Uint* microsec);
 #endif
 
-typedef UWord erts_approx_time_t;
-erts_approx_time_t erts_get_approx_time(void);
-
 int erts_has_time_correction(void);
 int erts_check_time_adj_support(int time_correction,
 				ErtsTimeWarpMode time_warp_mode);
diff --git a/erts/emulator/beam/erl_time_sup.c b/erts/emulator/beam/erl_time_sup.c
index 979c03fd43..29c698e34f 100644
--- a/erts/emulator/beam/erl_time_sup.c
+++ b/erts/emulator/beam/erl_time_sup.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1999-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1999-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -35,12 +35,13 @@
 #include "erl_time.h"
 #include "erl_driver.h"
 #include "erl_nif.h"
+#include "erl_proc_sig_queue.h"
  
-static erts_smp_mtx_t erts_get_time_mtx;
+static erts_mtx_t erts_get_time_mtx;
 
  /* used by erts_runtime_elapsed_both */
 typedef struct {
-    erts_smp_mtx_t mtx;
+    erts_mtx_t mtx;
     ErtsMonotonicTime user;
     ErtsMonotonicTime sys;
 } ErtsRunTimePrevData;
@@ -51,13 +52,13 @@ static union {
 } runtime_prev erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
 static union {
-    erts_smp_atomic64_t time;
-    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_smp_atomic64_t))];
+    erts_atomic64_t time;
+    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_atomic64_t))];
 } wall_clock_prev erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
 static union {
-    erts_smp_atomic64_t time;
-    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_smp_atomic64_t))];
+    erts_atomic64_t time;
+    char align__[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_atomic64_t))];
 } now_prev erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
 static ErtsMonitor *time_offset_monitors = NULL;
@@ -157,7 +158,7 @@ typedef struct {
 struct time_sup_infrequently_changed__ {
 #ifdef ERTS_HAVE_OS_MONOTONIC_TIME_SUPPORT
     struct {
-	erts_smp_rwmtx_t rwmtx;
+	erts_rwmtx_t rwmtx;
 	ErtsTWheelTimer timer;
 	ErtsMonotonicCorrectionData cdata;
     } parmon;
@@ -165,9 +166,9 @@ struct time_sup_infrequently_changed__ {
 #endif
     ErtsSystemTime sinit;
     ErtsMonotonicTime not_corrected_moffset;
-    erts_smp_atomic64_t offset;
+    erts_atomic64_t offset;
     ErtsMonotonicTime shadow_offset;
-    erts_smp_atomic32_t preliminary_offset;
+    erts_atomic32_t preliminary_offset;
 };
 
 struct time_sup_frequently_changed__ {
@@ -191,33 +192,22 @@ static struct {
 
 ErtsTimeSupData erts_time_sup__ erts_align_attribute(ERTS_CACHE_LINE_SIZE);
 
-/*
- * erts_get_approx_time() returns an *approximate* time
- * in seconds. NOTE that this time may jump backwards!!!
- */
-erts_approx_time_t
-erts_get_approx_time(void)
-{
-    ErtsSystemTime stime = erts_os_system_time();
-    return (erts_approx_time_t) ERTS_MONOTONIC_TO_SEC(stime);
-}
-
 static ERTS_INLINE void
 init_time_offset(ErtsMonotonicTime offset)
 {
-    erts_smp_atomic64_init_nob(&time_sup.inf.c.offset, (erts_aint64_t) offset);
+    erts_atomic64_init_nob(&time_sup.inf.c.offset, (erts_aint64_t) offset);
 }
 
 static ERTS_INLINE void
 set_time_offset(ErtsMonotonicTime offset)
 {
-    erts_smp_atomic64_set_relb(&time_sup.inf.c.offset, (erts_aint64_t) offset);
+    erts_atomic64_set_relb(&time_sup.inf.c.offset, (erts_aint64_t) offset);
 }
 
 static ERTS_INLINE ErtsMonotonicTime
 get_time_offset(void)
 {
-    return (ErtsMonotonicTime) erts_smp_atomic64_read_acqb(&time_sup.inf.c.offset);
+    return (ErtsMonotonicTime) erts_atomic64_read_acqb(&time_sup.inf.c.offset);
 }
 
 static ERTS_INLINE void
@@ -298,7 +288,7 @@ read_corrected_time(int os_drift_corrected)
     ErtsMonotonicTime os_mtime;
     ErtsMonotonicCorrectionInstance ci;
 
-    erts_smp_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
 
     os_mtime = erts_os_monotonic_time();
 
@@ -311,7 +301,7 @@ read_corrected_time(int os_drift_corrected)
 	ci = time_sup.inf.c.parmon.cdata.insts.prev;
     }
 
-    erts_smp_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
 
     return calc_corrected_erl_mtime(os_mtime, &ci, NULL,
 				    os_drift_corrected);
@@ -389,13 +379,13 @@ check_time_correction(void *vesdp)
     int os_drift_corrected = time_sup.r.o.os_corrected_monotonic_time;
     int set_new_correction = 0, begin_short_intervals = 0;
 
-    erts_smp_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
 
     erts_os_times(&os_mtime, &os_stime);
 
     ci = time_sup.inf.c.parmon.cdata.insts.curr;
 
-    erts_smp_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
 
     if (os_mtime < ci.os_mtime)
 	erts_exit(ERTS_ABORT_EXIT,
@@ -410,7 +400,7 @@ check_time_correction(void *vesdp)
 
     if (time_sup.inf.c.shadow_offset) {
 	ERTS_TIME_ASSERT(time_sup.r.o.warp_mode == ERTS_SINGLE_TIME_WARP_MODE);
-	if (erts_smp_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
+	if (erts_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
 	    sdiff += time_sup.inf.c.shadow_offset;
 	else
 	    time_sup.inf.c.shadow_offset = 0;
@@ -433,7 +423,7 @@ check_time_correction(void *vesdp)
 	}
     }
     else if ((time_sup.r.o.warp_mode == ERTS_SINGLE_TIME_WARP_MODE
-	      && erts_smp_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
+	      && erts_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
 	     && (sdiff < -2*time_sup.r.o.adj.small_diff
 		 || 2*time_sup.r.o.adj.small_diff < sdiff)) {
 	/*
@@ -658,7 +648,7 @@ check_time_correction(void *vesdp)
 #endif
 
     if (set_new_correction) {
-	erts_smp_rwmtx_rwlock(&time_sup.inf.c.parmon.rwmtx);
+	erts_rwmtx_rwlock(&time_sup.inf.c.parmon.rwmtx);
 
 	os_mtime = erts_os_monotonic_time();
 
@@ -686,7 +676,7 @@ check_time_correction(void *vesdp)
 	time_sup.inf.c.parmon.cdata.insts.curr.os_mtime = os_mtime;
 	time_sup.inf.c.parmon.cdata.insts.curr.correction = new_correction;
 
-	erts_smp_rwmtx_rwunlock(&time_sup.inf.c.parmon.rwmtx);
+	erts_rwmtx_rwunlock(&time_sup.inf.c.parmon.rwmtx);
     }
 
     if (!esdp)
@@ -804,13 +794,13 @@ finalize_corrected_time_offset(ErtsSystemTime *stimep)
     ErtsMonotonicCorrectionInstance ci;
     int os_drift_corrected = time_sup.r.o.os_corrected_monotonic_time;
 
-    erts_smp_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_rlock(&time_sup.inf.c.parmon.rwmtx);
 
     erts_os_times(&os_mtime, stimep);
 
     ci = time_sup.inf.c.parmon.cdata.insts.curr;
 
-    erts_smp_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
+    erts_rwmtx_runlock(&time_sup.inf.c.parmon.rwmtx);
 
     if (os_mtime < ci.os_mtime)
 	erts_exit(ERTS_ABORT_EXIT,
@@ -863,7 +853,7 @@ static ErtsMonotonicTime get_not_corrected_time(void)
 {
     ErtsMonotonicTime stime, mtime;
 
-    erts_smp_mtx_lock(&erts_get_time_mtx);
+    erts_mtx_lock(&erts_get_time_mtx);
 
     stime = erts_os_system_time();
 
@@ -889,7 +879,7 @@ static ErtsMonotonicTime get_not_corrected_time(void)
 
     ASSERT(stime == mtime + time_sup.inf.c.not_corrected_moffset);
 
-    erts_smp_mtx_unlock(&erts_get_time_mtx);
+    erts_mtx_unlock(&erts_get_time_mtx);
 
     return mtime;
 }
@@ -971,9 +961,9 @@ erts_init_time_sup(int time_correction, ErtsTimeWarpMode time_warp_mode)
 
     ASSERT(ERTS_MONOTONIC_TIME_MIN < ERTS_MONOTONIC_TIME_MAX);
 
-    erts_smp_mtx_init(&erts_get_time_mtx, "get_time", NIL,
+    erts_mtx_init(&erts_get_time_mtx, "get_time", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
-    erts_smp_mtx_init(&runtime_prev.data.mtx, "runtime", NIL,
+    erts_mtx_init(&runtime_prev.data.mtx, "runtime", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
     runtime_prev.data.user = 0;
     runtime_prev.data.sys = 0;
@@ -982,9 +972,9 @@ erts_init_time_sup(int time_correction, ErtsTimeWarpMode time_warp_mode)
     time_sup.r.o.warp_mode = time_warp_mode;
  
     if (time_warp_mode == ERTS_SINGLE_TIME_WARP_MODE)
-	erts_smp_atomic32_init_nob(&time_sup.inf.c.preliminary_offset, 1);
+	erts_atomic32_init_nob(&time_sup.inf.c.preliminary_offset, 1);
     else
-	erts_smp_atomic32_init_nob(&time_sup.inf.c.preliminary_offset, 0);
+	erts_atomic32_init_nob(&time_sup.inf.c.preliminary_offset, 0);
     time_sup.inf.c.shadow_offset = 0;
 
 #if !ERTS_COMPILE_TIME_MONOTONIC_TIME_UNIT
@@ -1128,7 +1118,7 @@ erts_init_time_sup(int time_correction, ErtsTimeWarpMode time_warp_mode)
 
     if (time_sup.r.o.correction) {
 	ErtsMonotonicCorrectionData *cdatap;
-	erts_smp_rwmtx_opt_t rwmtx_opts = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
+	erts_rwmtx_opt_t rwmtx_opts = ERTS_RWMTX_OPT_DEFAULT_INITER;
 	ErtsMonotonicTime offset;
 	erts_os_times(&time_sup.inf.c.minit,
 		      &time_sup.inf.c.sinit);
@@ -1138,10 +1128,10 @@ erts_init_time_sup(int time_correction, ErtsTimeWarpMode time_warp_mode)
 	offset -= ERTS_MONOTONIC_BEGIN;
 	init_time_offset(offset);
 
-	rwmtx_opts.type = ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
-	rwmtx_opts.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+	rwmtx_opts.type = ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
+	rwmtx_opts.lived = ERTS_RWMTX_LONG_LIVED;
 
-        erts_smp_rwmtx_init_opt(&time_sup.inf.c.parmon.rwmtx, &rwmtx_opts,
+        erts_rwmtx_init_opt(&time_sup.inf.c.parmon.rwmtx, &rwmtx_opts,
             "get_corrected_time", NIL,
             ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
@@ -1176,10 +1166,10 @@ erts_init_time_sup(int time_correction, ErtsTimeWarpMode time_warp_mode)
 	time_sup.f.c.last_not_corrected_time = 0;
     }
 
-    erts_smp_atomic64_init_nob(&wall_clock_prev.time,
-                               (erts_aint64_t) 0);
+    erts_atomic64_init_nob(&wall_clock_prev.time,
+                           (erts_aint64_t) 0);
 
-    erts_smp_atomic64_init_nob(
+    erts_atomic64_init_nob(
         &now_prev.time,
         (erts_aint64_t) ERTS_MONOTONIC_TO_USEC(get_time_offset()));
 
@@ -1223,7 +1213,7 @@ ErtsTimeOffsetState erts_time_offset_state(void)
     case ERTS_NO_TIME_WARP_MODE:
 	return ERTS_TIME_OFFSET_FINAL;
     case ERTS_SINGLE_TIME_WARP_MODE:
-	if (erts_smp_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
+	if (erts_atomic32_read_nob(&time_sup.inf.c.preliminary_offset))
 	    return ERTS_TIME_OFFSET_PRELIMINARY;
 	return ERTS_TIME_OFFSET_FINAL;
     case ERTS_MULTI_TIME_WARP_MODE:
@@ -1256,9 +1246,9 @@ erts_finalize_time_offset(void)
     case ERTS_SINGLE_TIME_WARP_MODE: {
 	ErtsTimeOffsetState res = ERTS_TIME_OFFSET_FINAL;
 
-	erts_smp_mtx_lock(&erts_get_time_mtx);
+	erts_mtx_lock(&erts_get_time_mtx);
 
-	if (erts_smp_atomic32_read_nob(&time_sup.inf.c.preliminary_offset)) {
+	if (erts_atomic32_read_nob(&time_sup.inf.c.preliminary_offset)) {
 	    ErtsMonotonicTime mtime, new_offset;
 
 #ifdef ERTS_HAVE_OS_MONOTONIC_TIME_SUPPORT
@@ -1295,11 +1285,11 @@ erts_finalize_time_offset(void)
 	    set_time_offset(new_offset);
 	    schedule_send_time_offset_changed_notifications(new_offset);
 
-	    erts_smp_atomic32_set_nob(&time_sup.inf.c.preliminary_offset, 0);
+	    erts_atomic32_set_nob(&time_sup.inf.c.preliminary_offset, 0);
 	    res = ERTS_TIME_OFFSET_PRELIMINARY;
 	}
 
-	erts_smp_mtx_unlock(&erts_get_time_mtx);
+	erts_mtx_unlock(&erts_get_time_mtx);
 
 	return res;
     }
@@ -1358,14 +1348,14 @@ erts_runtime_elapsed_both(ErtsMonotonicTime *ms_user, ErtsMonotonicTime *ms_sys,
 
     if (ms_user_diff || ms_sys_diff) {
  
-        erts_smp_mtx_lock(&runtime_prev.data.mtx);
+        erts_mtx_lock(&runtime_prev.data.mtx);
 
         prev_user = runtime_prev.data.user;
         prev_sys = runtime_prev.data.sys;
         runtime_prev.data.user = user;
         runtime_prev.data.sys = sys;
 
-        erts_smp_mtx_unlock(&runtime_prev.data.mtx);
+        erts_mtx_unlock(&runtime_prev.data.mtx);
 
         if (ms_user_diff)
             *ms_user_diff = user - prev_user;
@@ -1394,8 +1384,8 @@ erts_wall_clock_elapsed_both(ErtsMonotonicTime *ms_total, ErtsMonotonicTime *ms_
         ErtsMonotonicTime prev;
 
         prev = ((ErtsMonotonicTime)
-                erts_smp_atomic64_xchg_mb(&wall_clock_prev.time,
-                                          (erts_aint64_t) elapsed));
+                erts_atomic64_xchg_mb(&wall_clock_prev.time,
+                                      (erts_aint64_t) elapsed));
 
         *ms_diff = elapsed - prev;
     }
@@ -1784,15 +1774,15 @@ get_now(Uint* megasec, Uint* sec, Uint* microsec)
     now = ERTS_MONOTONIC_TO_USEC(mtime + time_offset);
 
     /* Make sure now time is later than last time */
-    prev = erts_smp_atomic64_read_nob(&now_prev.time);
+    prev = erts_atomic64_read_nob(&now_prev.time);
     while (1) {
         ErtsMonotonicTime act;
         if (now <= prev)
             now = prev + 1;
         act = ((ErtsMonotonicTime)
-               erts_smp_atomic64_cmpxchg_mb(&now_prev.time,
-                                            (erts_aint64_t) now,
-                                            (erts_aint64_t) prev));
+               erts_atomic64_cmpxchg_mb(&now_prev.time,
+                                        (erts_aint64_t) now,
+                                        (erts_aint64_t) prev));
         if (act == prev)
             break;
         prev = act;
@@ -1870,7 +1860,7 @@ void erts_get_now_cpu(Uint* megasec, Uint* sec, Uint* microsec) {
   SysCpuTime t;
   SysTimespec tp;
 
-  sys_get_proc_cputime(t, tp);
+  sys_get_cputime(t, tp);
   *microsec = (Uint)(tp.tv_nsec / 1000);
   t = (tp.tv_sec / 1000000);
   *megasec = (Uint)(t % 1000000);
@@ -1881,36 +1871,33 @@ void erts_get_now_cpu(Uint* megasec, Uint* sec, Uint* microsec) {
 #include "big.h"
 
 void
-erts_monitor_time_offset(Eterm id, Eterm ref)
+erts_monitor_time_offset(ErtsMonitor *mon)
 {
-    erts_smp_mtx_lock(&erts_get_time_mtx);
-    erts_add_monitor(&time_offset_monitors, MON_TIME_OFFSET, ref, id, NIL);
+    erts_mtx_lock(&erts_get_time_mtx);
+    erts_monitor_list_insert(&time_offset_monitors, mon);
     no_time_offset_monitors++;
-    erts_smp_mtx_unlock(&erts_get_time_mtx);
+    erts_mtx_unlock(&erts_get_time_mtx);
 }
 
-int
-erts_demonitor_time_offset(Eterm ref)
-{
-    int res;
-    ErtsMonitor *mon;
-    ASSERT(is_internal_ref(ref));
-    erts_smp_mtx_lock(&erts_get_time_mtx);
-    if (is_internal_ordinary_ref(ref))
-        mon = erts_remove_monitor(&time_offset_monitors, ref);
-    else
-        mon = NULL;
-    if (!mon)
-	res = 0;
-    else {
-	ASSERT(no_time_offset_monitors > 0);
-	no_time_offset_monitors--;
-	res = 1;
-    }
-    erts_smp_mtx_unlock(&erts_get_time_mtx);
-    if (res)
-	erts_destroy_monitor(mon);
-    return res;
+void
+erts_demonitor_time_offset(ErtsMonitor *mon)
+{
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+    ASSERT(erts_monitor_is_origin(mon));
+    ASSERT(mon->type == ERTS_MON_TYPE_TIME_OFFSET);
+
+    erts_mtx_lock(&erts_get_time_mtx);
+
+    ASSERT(erts_monitor_is_in_table(&mdp->target));
+
+    erts_monitor_list_delete(&time_offset_monitors, &mdp->target);
+
+    ASSERT(no_time_offset_monitors > 0);
+    no_time_offset_monitors--;
+
+    erts_mtx_unlock(&erts_get_time_mtx);
+
+    erts_monitor_release_both(mdp);
 }
 
 typedef struct {
@@ -1928,17 +1915,19 @@ static void
 save_time_offset_monitor(ErtsMonitor *mon, void *vcntxt)
 {
     ErtsTimeOffsetMonitorContext *cntxt;
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
     Eterm *from_hp, *to_hp;
     Uint mix;
     int hix;
 
     cntxt = (ErtsTimeOffsetMonitorContext *) vcntxt;
     mix = (cntxt->ix)++;
-    cntxt->to_mon_info[mix].pid = mon->u.pid;
+    ASSERT(is_internal_pid(mon->other.item));
+    cntxt->to_mon_info[mix].pid = mon->other.item;
     to_hp = &cntxt->to_mon_info[mix].heap[0];
 
-    ASSERT(is_internal_ordinary_ref(mon->ref));
-    from_hp = internal_ref_val(mon->ref);
+    ASSERT(is_internal_ordinary_ref(mdp->ref));
+    from_hp = internal_ref_val(mdp->ref);
     ASSERT(thing_arityval(*from_hp) + 1 == ERTS_REF_THING_SIZE);
 
     for (hix = 0; hix < ERTS_REF_THING_SIZE; hix++)
@@ -1965,7 +1954,7 @@ send_time_offset_changed_notifications(void *new_offsetp)
 #endif
     new_offset -= ERTS_MONOTONIC_OFFSET_NATIVE;
 
-    erts_smp_mtx_lock(&erts_get_time_mtx);
+    erts_mtx_lock(&erts_get_time_mtx);
 
     no_monitors = no_time_offset_monitors;
     if (no_monitors) {
@@ -1983,14 +1972,14 @@ send_time_offset_changed_notifications(void *new_offsetp)
 	cntxt.ix = 0;
 	cntxt.to_mon_info = to_mon_info;
 
-	erts_doforall_monitors(time_offset_monitors,
-			       save_time_offset_monitor,
-			       &cntxt);
+        erts_monitor_list_foreach(time_offset_monitors,
+                                  save_time_offset_monitor,
+                                  &cntxt);
 
 	ASSERT(cntxt.ix == no_monitors);
     }
 
-    erts_smp_mtx_unlock(&erts_get_time_mtx);
+    erts_mtx_unlock(&erts_get_time_mtx);
 
     if (no_monitors) {
 	Eterm *hp, *patch_refp, new_offset_term, message_template;
@@ -2019,26 +2008,14 @@ send_time_offset_changed_notifications(void *new_offsetp)
 	ASSERT(*patch_refp == THE_NON_VALUE);
 
 	for (mix = 0; mix < no_monitors; mix++) {
-	    Process *rp = erts_proc_lookup(to_mon_info[mix].pid);
-	    if (rp) {
-		Eterm ref = to_mon_info[mix].ref;
-		ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK;
-		erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-		if (erts_lookup_monitor(ERTS_P_MONITORS(rp), ref)) {
-		    ErtsMessage *mp;
-		    ErlOffHeap *ohp;
-		    Eterm message;
-
-		    mp = erts_alloc_message_heap(rp, &rp_locks,
-						 hsz, &hp, &ohp);
-		    *patch_refp = ref;
-		    ASSERT(hsz == size_object(message_template));
-		    message = copy_struct(message_template, hsz, &hp, ohp);
-		    erts_queue_message(rp, rp_locks, mp, message, am_clock_service);
-		}
-		erts_smp_proc_unlock(rp, rp_locks);
-	    }
-	}
+            *patch_refp = to_mon_info[mix].ref;
+            erts_proc_sig_send_persistent_monitor_msg(ERTS_MON_TYPE_TIME_OFFSET,
+                                                      *patch_refp,
+                                                      am_clock_service,
+                                                      to_mon_info[mix].pid,
+                                                      message_template,
+                                                      hsz);
+        }
 
 	erts_free(ERTS_ALC_T_TMP, tmp);
     }
@@ -2227,6 +2204,8 @@ time_unit_conversion(Process *c_p, Eterm term, ErtsMonotonicTime val, ErtsMonoto
 	ERTS_BIF_PREP_RET(ret, make_time_val(c_p, result));
 	break;
 #endif
+    case am_perf_counter:
+	goto trap_to_erlang_code;
     default: {
 	Eterm value, native_res;
 #ifndef ARCH_64
diff --git a/erts/emulator/beam/erl_trace.c b/erts/emulator/beam/erl_trace.c
index 9fda4ff6ff..f4161b14f2 100644
--- a/erts/emulator/beam/erl_trace.c
+++ b/erts/emulator/beam/erl_trace.c
@@ -77,8 +77,8 @@ static Eterm system_profile;
 int erts_cpu_timestamp;
 #endif
 
-static erts_smp_mtx_t smq_mtx;
-static erts_smp_rwmtx_t sys_trace_rwmtx;
+static erts_mtx_t smq_mtx;
+static erts_rwmtx_t sys_trace_rwmtx;
 
 enum ErtsSysMsgType {
     SYS_MSG_TYPE_UNDEFINED,
@@ -237,7 +237,6 @@ write_timestamp(ErtsTraceTimeStamp *tsp, Eterm **hpp)
     }
 }
 
-#ifdef ERTS_SMP
 
 static ERTS_INLINE Uint
 patch_ts_size(int ts_type)
@@ -257,7 +256,6 @@ patch_ts_size(int ts_type)
 	return 0;
     }
 }
-#endif /* ERTS_SMP */
 
 /*
  * Write a timestamp. The timestamp MUST be the last
@@ -298,18 +296,11 @@ write_ts(int ts_type, Eterm *hp, ErlHeapFragment *bp, Process *tracer)
     if (shrink) {
 	if (bp)
 	    bp->used_size -= shrink;
-#ifndef ERTS_SMP
-	else if (tracer) {
-	    Eterm *endp = ts_hp + shrink;
-	    HRelease(tracer, endp, ts_hp);
-	}
-#endif
     }
 
     return res;
 }
 
-#ifdef ERTS_SMP
 static void enqueue_sys_msg_unlocked(enum ErtsSysMsgType type,
 				     Eterm from,
 				     Eterm to,
@@ -321,7 +312,6 @@ static void enqueue_sys_msg(enum ErtsSysMsgType type,
 			    Eterm msg,
 			    ErlHeapFragment *bp);
 static void init_sys_msg_dispatcher(void);
-#endif
 
 static void init_tracer_nif(void);
 static int tracer_cmp_fun(void*, void*);
@@ -332,11 +322,11 @@ static void tracer_free_fun(void*);
 typedef struct ErtsTracerNif_ ErtsTracerNif;
 
 void erts_init_trace(void) {
-    erts_smp_rwmtx_opt_t rwmtx_opts = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opts.type = ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
-    rwmtx_opts.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t rwmtx_opts = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opts.type = ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
+    rwmtx_opts.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&sys_trace_rwmtx, &rwmtx_opts, "sys_tracers", NIL,
+    erts_rwmtx_init_opt(&sys_trace_rwmtx, &rwmtx_opts, "sys_tracers", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
 
 #ifdef HAVE_ERTS_NOW_CPU
@@ -350,9 +340,7 @@ void erts_init_trace(void) {
     default_port_trace_flags = F_INITIAL_TRACE_FLAGS;
     default_port_tracer = erts_tracer_nil;
     system_seq_tracer = erts_tracer_nil;
-#ifdef ERTS_SMP
     init_sys_msg_dispatcher();
-#endif
     init_tracer_nif();
 }
 
@@ -412,43 +400,35 @@ static Uint active_sched;
 void
 erts_system_profile_setup_active_schedulers(void)
 {
-    ERTS_SMP_LC_ASSERT(erts_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking());
     active_sched = erts_active_schedulers();
 }
 
 static void
 exiting_reset(Eterm exiting)
 {
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     if (exiting == system_monitor) {
-#ifdef ERTS_SMP
 	system_monitor = NIL;
 	/* Let the trace message dispatcher clear flags, etc */
-#else
-	erts_system_monitor_clear(NULL);
-#endif
     }
     if (exiting == system_profile) {
-#ifdef ERTS_SMP
 	system_profile = NIL;
 	/* Let the trace message dispatcher clear flags, etc */
-#else
-	erts_system_profile_clear(NULL);
-#endif
     }
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
 }
 
 void
 erts_trace_check_exiting(Eterm exiting)
 {
     int reset = 0;
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     if (exiting == system_monitor)
 	reset = 1;
     else if (exiting == system_profile)
 	reset = 1;
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
     if (reset)
 	exiting_reset(exiting);
 }
@@ -468,7 +448,7 @@ erts_set_system_seq_tracer(Process *c_p, ErtsProcLocks c_p_locks, ErtsTracer new
         }
     }
 
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     old = system_seq_tracer;
     system_seq_tracer = erts_tracer_nil;
     erts_tracer_update(&system_seq_tracer, new);
@@ -476,7 +456,7 @@ erts_set_system_seq_tracer(Process *c_p, ErtsProcLocks c_p_locks, ErtsTracer new
 #ifdef DEBUG_PRINTOUTS
     erts_fprintf(stderr, "set seq tracer new=%T old=%T\n", new, old);
 #endif
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
     return old;
 }
 
@@ -484,12 +464,12 @@ ErtsTracer
 erts_get_system_seq_tracer(void)
 {
     ErtsTracer st;
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     st = system_seq_tracer;
 #ifdef DEBUG_PRINTOUTS
     erts_fprintf(stderr, "get seq tracer %T\n", st);
 #endif
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
 
     if (st != erts_tracer_nil &&
         call_enabled_tracer(st, NULL, TRACE_FUN_ENABLED,
@@ -522,8 +502,8 @@ get_default_tracing(Uint *flagsp, ErtsTracer *tracerp,
             ErtsTracer curr_default_tracer = *default_tracer;
             if (tracerp) {
                 /* we only have a rlock, so we have to unlock and then rwlock */
-                erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
-                erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+                erts_rwmtx_runlock(&sys_trace_rwmtx);
+                erts_rwmtx_rwlock(&sys_trace_rwmtx);
             }
             /* check if someone else changed default tracer
                while we got the write lock, if so we don't do
@@ -533,8 +513,8 @@ get_default_tracing(Uint *flagsp, ErtsTracer *tracerp,
                 ERTS_TRACER_CLEAR(default_tracer);
             }
             if (tracerp) {
-                erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
-                erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+                erts_rwmtx_rwunlock(&sys_trace_rwmtx);
+                erts_rwmtx_rlock(&sys_trace_rwmtx);
             }
         }
         }
@@ -567,98 +547,84 @@ void
 erts_change_default_proc_tracing(int setflags, Uint flagsp,
                                  const ErtsTracer tracer)
 {
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     erts_change_default_tracing(
         setflags, flagsp, tracer,
         &default_proc_trace_flags,
         &default_proc_tracer);
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
 }
 
 void
 erts_change_default_port_tracing(int setflags, Uint flagsp,
                                  const ErtsTracer tracer)
 {
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     erts_change_default_tracing(
         setflags, flagsp, tracer,
         &default_port_trace_flags,
         &default_port_tracer);
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
 }
 
 void
 erts_get_default_proc_tracing(Uint *flagsp, ErtsTracer *tracerp)
 {
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     *tracerp = erts_tracer_nil; /* initialize */
     get_default_tracing(
         flagsp, tracerp,
         &default_proc_trace_flags,
         &default_proc_tracer);
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
 }
 
 void
 erts_get_default_port_tracing(Uint *flagsp, ErtsTracer *tracerp)
 {
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     *tracerp = erts_tracer_nil; /* initialize */
     get_default_tracing(
         flagsp, tracerp,
         &default_port_trace_flags,
         &default_port_tracer);
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
 }
 
 void
 erts_set_system_monitor(Eterm monitor)
 {
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     system_monitor = monitor;
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
 }
 
 Eterm
 erts_get_system_monitor(void)
 {
     Eterm monitor;
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     monitor = system_monitor;
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
     return monitor;
 }
 
 /* Performance monitoring */
 void erts_set_system_profile(Eterm profile) {
-    erts_smp_rwmtx_rwlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwlock(&sys_trace_rwmtx);
     system_profile = profile;
-    erts_smp_rwmtx_rwunlock(&sys_trace_rwmtx);
+    erts_rwmtx_rwunlock(&sys_trace_rwmtx);
 }
 
 Eterm
 erts_get_system_profile(void) {
     Eterm profile;
-    erts_smp_rwmtx_rlock(&sys_trace_rwmtx);
+    erts_rwmtx_rlock(&sys_trace_rwmtx);
     profile = system_profile;
-    erts_smp_rwmtx_runlock(&sys_trace_rwmtx);
+    erts_rwmtx_runlock(&sys_trace_rwmtx);
     return profile;
 }
 
-
-#ifdef HAVE_ERTS_NOW_CPU
-#  define GET_NOW(m, s, u) \
-do { \
-    if (erts_cpu_timestamp) \
-	erts_get_now_cpu(m, s, u); \
-    else \
-	get_now(m, s, u); \
-} while (0)
-#else
-#  define GET_NOW(m, s, u) do {get_now(m, s, u);} while (0)
-#endif
-
-
 static void
 write_sys_msg_to_port(Eterm unused_to,
 		      Port* trace_port,
@@ -679,71 +645,11 @@ write_sys_msg_to_port(Eterm unused_to,
 	erts_exit(ERTS_ERROR_EXIT, "Internal error in do_send_to_port: %d\n", ptr-buffer);
     }
 
-#ifndef ERTS_SMP
-    if (!INVALID_TRACER_PORT(trace_port, trace_port->common.id))
-#endif
 	erts_raw_port_command(trace_port, buffer, ptr-buffer);
 
     erts_free(ERTS_ALC_T_TMP, (void *) buffer);
 }
 
-#ifndef ERTS_SMP
-/* Profile send
- * Checks if profiler is port or process
- * Eterm msg is local, need copying.
- */
-
-static void 
-profile_send(Eterm from, Eterm message) {
-    Uint sz = 0;
-    Uint *hp = NULL;
-    Eterm msg = NIL;
-    Process *profile_p = NULL;
-
-    Eterm profiler = erts_get_system_profile();
-
-    /* do not profile profiler pid */
-    if (from == profiler) return;
-
-    if (is_internal_port(profiler)) {
-    	Port *profiler_port = NULL;
-
-	/* not smp */
-
-	profiler_port = erts_id2port_sflgs(profiler,
-					   NULL,
-					   0,
-					   ERTS_PORT_SFLGS_INVALID_TRACER_LOOKUP);
-	if (profiler_port) {
-	    write_sys_msg_to_port(profiler,
-                                  profiler_port,
-                                  NIL, /* or current process->common.id */
-                                  SYS_MSG_TYPE_SYSPROF,
-                                  message);
-	    erts_port_release(profiler_port);
-	}
-    	
-    } else {
-	ErtsMessage *mp;
-	ASSERT(is_internal_pid(profiler));
-        
-	profile_p = erts_proc_lookup(profiler);
-
-	if (!profile_p)
-	    return;
-
-	sz = size_object(message);
-	mp = erts_alloc_message(sz, &hp);
-	if (sz == 0)
-	    msg = message;
-	else
-	    msg = copy_struct(message, sz, &hp, &mp->hfrag.off_heap);
-
-        erts_queue_message(profile_p, 0, mp, msg, from);
-    }
-}
-
-#endif
 
 static void
 trace_sched_aux(Process *p, ErtsProcLocks locks, Eterm what)
@@ -815,9 +721,7 @@ trace_send(Process *p, Eterm to, Eterm msg)
     ErtsTracerNif *tnif = NULL;
     ErtsTracingEvent* te;
     Eterm pam_result;
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl;
-#endif
 
     ASSERT(ARE_TRACE_FLAGS_ON(p, F_TRACE_SEND));
 
@@ -842,9 +746,7 @@ trace_send(Process *p, Eterm to, Eterm msg)
     } else
         pam_result = am_true;
 
-#ifdef ERTS_SMP
     dhndl = erts_thr_progress_unmanaged_delay();
-#endif
 
     if (is_internal_pid(to)) {
 	if (!erts_proc_lookup(to))
@@ -862,9 +764,7 @@ trace_send(Process *p, Eterm to, Eterm msg)
                            operation, msg, to, pam_result);
     }
 
-#ifdef ERTS_SMP
     erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 
     erts_match_set_release_result_trace(p, pam_result);
 }
@@ -1179,7 +1079,7 @@ erts_call_trace(Process* p, ErtsCodeInfo *info, Binary *match_spec,
     Eterm transformed_args[MAX_ARG];
     ErtsTracer pre_ms_tracer = erts_tracer_nil;
 
-    ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(p) & ERTS_PROC_LOCK_MAIN);
+    ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(p) & ERTS_PROC_LOCK_MAIN);
 
     ASSERT(tracer);
     if (ERTS_TRACER_COMPARE(*tracer, erts_tracer_true)) {
@@ -1467,21 +1367,11 @@ monitor_long_schedule_proc(Process *p, ErtsCodeMFA *in_fp,
 {
     ErlHeapFragment *bp;
     ErlOffHeap *off_heap;
-#ifndef ERTS_SMP
-    Process *monitor_p;
-#endif
     Uint hsz;
     Eterm *hp, list, in_mfa = am_undefined, out_mfa = am_undefined;
     Eterm in_tpl, out_tpl, tmo_tpl, tmo, msg;
  
 
-#ifndef ERTS_SMP
-    ASSERT(is_internal_pid(system_monitor));
-    monitor_p = erts_proc_lookup(system_monitor);
-    if (!monitor_p || p == monitor_p) {
-	return;
-    }
-#endif
     /* 
      * Size: {monitor, pid, long_schedule, [{timeout, T}, {in, {M,F,A}},{out,{M,F,A}}]} ->
      * 5 (top tuple of 4), (3 (elements) * 2 (cons)) + 3 (timeout tuple of 2) + size of Timeout +
@@ -1517,36 +1407,18 @@ monitor_long_schedule_proc(Process *p, ErtsCodeMFA *in_fp,
     hp += 2;
     msg = TUPLE4(hp, am_monitor, p->common.id, am_long_schedule, list);
     hp += 5;
-#ifdef ERTS_SMP
     enqueue_sys_msg(SYS_MSG_TYPE_SYSMON, p->common.id, NIL, msg, bp);
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(monitor_p, 0, mp, msg, am_system);
-    }
-#endif
 }
 void 
 monitor_long_schedule_port(Port *pp, ErtsPortTaskType type, Uint time)
 {
     ErlHeapFragment *bp;
     ErlOffHeap *off_heap;
-#ifndef ERTS_SMP
-    Process *monitor_p;
-#endif
     Uint hsz;
     Eterm *hp, list, op;
     Eterm op_tpl, tmo_tpl, tmo, msg;
  
 
-#ifndef ERTS_SMP
-    ASSERT(is_internal_pid(system_monitor));
-    monitor_p = erts_proc_lookup(system_monitor);
-    if (!monitor_p) {
-	return;
-    }
-#endif
     /* 
      * Size: {monitor, port, long_schedule, [{timeout, T}, {op, Operation}]} ->
      * 5 (top tuple of 4), (2 (elements) * 2 (cons)) + 3 (timeout tuple of 2) 
@@ -1563,7 +1435,6 @@ monitor_long_schedule_port(Port *pp, ErtsPortTaskType type, Uint time)
     case ERTS_PORT_TASK_TIMEOUT: op = am_timeout; break;
     case ERTS_PORT_TASK_INPUT: op = am_input; break;
     case ERTS_PORT_TASK_OUTPUT: op = am_output; break;
-    case ERTS_PORT_TASK_EVENT: op = am_event; break;
     case ERTS_PORT_TASK_DIST_CMD: op = am_dist_cmd; break;
     default: op = am_undefined; break;
     }
@@ -1582,24 +1453,13 @@ monitor_long_schedule_port(Port *pp, ErtsPortTaskType type, Uint time)
     hp += 2;
     msg = TUPLE4(hp, am_monitor, pp->common.id, am_long_schedule, list);
     hp += 5;
-#ifdef ERTS_SMP
     enqueue_sys_msg(SYS_MSG_TYPE_SYSMON, pp->common.id, NIL, msg, bp);
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(monitor_p, 0, mp, msg, am_system);
-    }
-#endif
 }
 
 void
 monitor_long_gc(Process *p, Uint time) {
     ErlHeapFragment *bp;
     ErlOffHeap *off_heap;
-#ifndef ERTS_SMP
-    Process *monitor_p;
-#endif
     Uint hsz;
     Eterm *hp, list, msg;
     Eterm tags[] = {
@@ -1624,12 +1484,6 @@ monitor_long_gc(Process *p, Uint time) {
     Eterm *hp_end;
 #endif
 	
-#ifndef ERTS_SMP
-    ASSERT(is_internal_pid(system_monitor));
-    monitor_p = erts_proc_lookup(system_monitor);
-    if (!monitor_p || p == monitor_p)
-	return;
-#endif
 
     hsz = 0;
     (void) erts_bld_atom_uword_2tup_list(NULL,
@@ -1657,24 +1511,13 @@ monitor_long_gc(Process *p, Uint time) {
     ASSERT(hp == hp_end);
 #endif
 
-#ifdef ERTS_SMP
     enqueue_sys_msg(SYS_MSG_TYPE_SYSMON, p->common.id, NIL, msg, bp);
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(monitor_p, 0, mp, msg, am_system);
-    }
-#endif
 }
 
 void
 monitor_large_heap(Process *p) {
     ErlHeapFragment *bp;
     ErlOffHeap *off_heap;
-#ifndef ERTS_SMP
-    Process *monitor_p;
-#endif
     Uint hsz;
     Eterm *hp, list, msg;
     Eterm tags[] = {
@@ -1698,13 +1541,6 @@ monitor_large_heap(Process *p) {
 #endif
 
 
-#ifndef ERTS_SMP 
-    ASSERT(is_internal_pid(system_monitor));
-    monitor_p = erts_proc_lookup(system_monitor);
-    if (!monitor_p || p == monitor_p) {
-	return;
-    }
-#endif
 
     hsz = 0;
     (void) erts_bld_atom_uword_2tup_list(NULL,
@@ -1732,47 +1568,22 @@ monitor_large_heap(Process *p) {
     ASSERT(hp == hp_end);
 #endif
 
-#ifdef ERTS_SMP
     enqueue_sys_msg(SYS_MSG_TYPE_SYSMON, p->common.id, NIL, msg, bp);
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(monitor_p, 0, mp, msg, am_system);
-    }
-#endif
 }
 
 void
 monitor_generic(Process *p, Eterm type, Eterm spec) {
     ErlHeapFragment *bp;
     ErlOffHeap *off_heap;
-#ifndef ERTS_SMP
-    Process *monitor_p;
-#endif
     Eterm *hp, msg;
 
-#ifndef ERTS_SMP
-    ASSERT(is_internal_pid(system_monitor));
-    monitor_p = erts_proc_lookup(system_monitor);
-    if (!monitor_p || p == monitor_p)
-	return;
-#endif
 
     hp = ERTS_ALLOC_SYSMSG_HEAP(5, &bp, &off_heap, monitor_p);
 
     msg = TUPLE4(hp, am_monitor, p->common.id, type, spec); 
     hp += 5;
 
-#ifdef ERTS_SMP
     enqueue_sys_msg(SYS_MSG_TYPE_SYSMON, p->common.id, NIL, msg, bp);
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(monitor_p, 0, mp, msg, am_system);
-    }
-#endif
 
 }
 
@@ -1785,21 +1596,14 @@ profile_scheduler(Eterm scheduler_id, Eterm state) {
     Eterm *hp, msg;
     ErlHeapFragment *bp = NULL;
 
-#ifndef ERTS_SMP
-#define LOCAL_HEAP_SIZE (7 + ERTS_TRACE_PATCH_TS_MAX_SIZE)
-    DeclareTmpHeapNoproc(local_heap,LOCAL_HEAP_SIZE);
-    UseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-    hp = local_heap;
-#else
     Uint hsz;
 
     hsz = 7 + patch_ts_size(erts_system_profile_ts_type)-1;
 	
     bp = new_message_buffer(hsz);
     hp = bp->mem;
-#endif
 
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
 
     switch (state) {
     case am_active:
@@ -1821,14 +1625,8 @@ profile_scheduler(Eterm scheduler_id, Eterm state) {
     /* Write timestamp in element 6 of the 'msg' tuple */
     hp[-1] = write_ts(erts_system_profile_ts_type, hp, bp, NULL);
 
-#ifndef ERTS_SMP
-    profile_send(NIL, msg);
-    UnUseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-#undef LOCAL_HEAP_SIZE
-#else
     enqueue_sys_msg_unlocked(SYS_MSG_TYPE_SYSPROF, NIL, NIL, msg, bp);
-#endif
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 
 }
 
@@ -1837,7 +1635,7 @@ profile_scheduler(Eterm scheduler_id, Eterm state) {
 void
 trace_port_open(Port *p, Eterm calling_pid, Eterm drv_name) {
     ErtsTracerNif *tnif = NULL;
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &p->common, &tnif, TRACE_FUN_E_PORTS, am_open))
         send_to_tracer_nif(NULL, &p->common, p->common.id, tnif, TRACE_FUN_T_PORTS,
                 am_open, calling_pid, drv_name, am_true);
@@ -1854,9 +1652,9 @@ void
 trace_port(Port *t_p, Eterm what, Eterm data) {
 
     ErtsTracerNif *tnif = NULL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(t_p)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(t_p)
 		       || erts_thr_progress_is_blocking());
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &t_p->common, &tnif, TRACE_FUN_E_PORTS, what))
         send_to_tracer_nif(NULL, &t_p->common, t_p->common.id, tnif, TRACE_FUN_T_PORTS,
                            what, data, THE_NON_VALUE, am_true);
@@ -1899,9 +1697,9 @@ void
 trace_port_receive(Port *t_p, Eterm caller, Eterm what, ...)
 {
     ErtsTracerNif *tnif = NULL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(t_p)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(t_p)
 		       || erts_thr_progress_is_blocking());
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &t_p->common, &tnif, TRACE_FUN_E_RECEIVE, am_receive)) {
         /* We can use a stack heap here, as the nif is called in the
            context of a port */
@@ -2016,9 +1814,9 @@ trace_port_send(Port *t_p, Eterm receiver, Eterm msg, int exists)
 {
     ErtsTracerNif *tnif = NULL;
     Eterm op = exists ? am_send : am_send_to_non_existing_process;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(t_p)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(t_p)
 		       || erts_thr_progress_is_blocking());
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &t_p->common, &tnif, TRACE_FUN_E_SEND, op))
         send_to_tracer_nif(NULL, &t_p->common, t_p->common.id, tnif, TRACE_FUN_T_SEND,
                            op, msg, receiver, am_true);
@@ -2027,9 +1825,9 @@ trace_port_send(Port *t_p, Eterm receiver, Eterm msg, int exists)
 void trace_port_send_binary(Port *t_p, Eterm to, Eterm what, char *bin, Sint sz)
 {
     ErtsTracerNif *tnif = NULL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(t_p)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(t_p)
 		       || erts_thr_progress_is_blocking());
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &t_p->common, &tnif, TRACE_FUN_E_SEND, am_send)) {
         Eterm msg;
         Binary* bptr = NULL;
@@ -2075,9 +1873,9 @@ trace_sched_ports(Port *p, Eterm what) {
 void
 trace_sched_ports_where(Port *t_p, Eterm what, Eterm where) {
     ErtsTracerNif *tnif = NULL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(t_p)
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(t_p)
 		       || erts_thr_progress_is_blocking());
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (is_tracer_enabled(NULL, 0, &t_p->common, &tnif, TRACE_FUN_E_SCHED_PORT, what))
         send_to_tracer_nif(NULL, &t_p->common, t_p->common.id,
                            tnif, TRACE_FUN_T_SCHED_PORT,
@@ -2092,24 +1890,14 @@ profile_runnable_port(Port *p, Eterm status) {
     ErlHeapFragment *bp = NULL;
     Eterm count = make_small(0);
 
-#ifndef ERTS_SMP
-#define LOCAL_HEAP_SIZE (6 + ERTS_TRACE_PATCH_TS_MAX_SIZE)
-
-    DeclareTmpHeapNoproc(local_heap,LOCAL_HEAP_SIZE);
-    UseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-
-    hp = local_heap;
-    
-#else
     Uint hsz;
 
     hsz = 6 + patch_ts_size(erts_system_profile_ts_type)-1;
 
     bp = new_message_buffer(hsz);
     hp = bp->mem;
-#endif
 
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
 
     msg = TUPLE5(hp, am_profile, p->common.id, status, count,
 		 NIL /* Will be overwritten by timestamp */);
@@ -2118,14 +1906,8 @@ profile_runnable_port(Port *p, Eterm status) {
     /* Write timestamp in element 5 of the 'msg' tuple */
     hp[-1] = write_ts(erts_system_profile_ts_type, hp, bp, NULL);
 
-#ifndef ERTS_SMP
-    profile_send(p->common.id, msg);
-    UnUseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-#undef LOCAL_HEAP_SIZE
-#else
     enqueue_sys_msg_unlocked(SYS_MSG_TYPE_SYSPROF, p->common.id, NIL, msg, bp);
-#endif
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 /* Process profiling */
@@ -2136,23 +1918,13 @@ profile_runnable_proc(Process *p, Eterm status){
     ErlHeapFragment *bp = NULL;
     ErtsCodeMFA *cmfa = NULL;
 
-#ifndef ERTS_SMP
-#define LOCAL_HEAP_SIZE (4 + 6 + ERTS_TRACE_PATCH_TS_MAX_SIZE)
-    DeclareTmpHeapNoproc(local_heap,LOCAL_HEAP_SIZE);
-    UseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-
-    hp = local_heap;
-#else
     ErtsThrPrgrDelayHandle dhndl;
     Uint hsz = 4 + 6 + patch_ts_size(erts_system_profile_ts_type)-1;
-#endif
     /* Assumptions:
      * We possibly don't have the MAIN_LOCK for the process p here.
      * We assume that we can read from p->current and p->i atomically
      */
-#ifdef ERTS_SMP
     dhndl = erts_thr_progress_unmanaged_delay(); /* suspend purge operations */
-#endif
 
     if (!ERTS_PROC_IS_EXITING(p)) {
         if (p->current) {
@@ -2162,14 +1934,12 @@ profile_runnable_proc(Process *p, Eterm status){
         }
     }
 
-#ifdef ERTS_SMP
     if (!cmfa) {
 	hsz -= 4;
     }
 
     bp = new_message_buffer(hsz);
     hp = bp->mem;
-#endif
 
     if (cmfa) {
 	where = TUPLE3(hp, cmfa->module, cmfa->function,
@@ -2179,11 +1949,9 @@ profile_runnable_proc(Process *p, Eterm status){
 	where = make_small(0);
     }
 
-#ifdef ERTS_SMP
     erts_thr_progress_unmanaged_continue(dhndl);
-#endif
 	
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
 
     msg = TUPLE5(hp, am_profile, p->common.id, status, where,
 		 NIL /* Will be overwritten by timestamp */);
@@ -2192,20 +1960,13 @@ profile_runnable_proc(Process *p, Eterm status){
     /* Write timestamp in element 5 of the 'msg' tuple */
     hp[-1] = write_ts(erts_system_profile_ts_type, hp, bp, NULL);
 
-#ifndef ERTS_SMP
-    profile_send(p->common.id, msg);
-    UnUseTmpHeapNoproc(LOCAL_HEAP_SIZE);
-#undef LOCAL_HEAP_SIZE
-#else
     enqueue_sys_msg_unlocked(SYS_MSG_TYPE_SYSPROF, p->common.id, NIL, msg, bp);
-#endif
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 /* End system_profile tracing */
 
 
 
-#ifdef ERTS_SMP
 
 typedef struct ErtsSysMsgQ_ ErtsSysMsgQ;
 struct  ErtsSysMsgQ_ {
@@ -2251,7 +2012,7 @@ enqueue_sys_msg_unlocked(enum ErtsSysMsgType type,
 	sys_message_queue = smqp;
     }
     sys_message_queue_end = smqp;
-    erts_smp_cnd_signal(&smq_cnd);
+    erts_cnd_signal(&smq_cnd);
 }
 
 static void
@@ -2261,15 +2022,15 @@ enqueue_sys_msg(enum ErtsSysMsgType type,
 		Eterm msg,
 		ErlHeapFragment *bp)
 {
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     enqueue_sys_msg_unlocked(type, from, to, msg, bp);
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 void
 erts_queue_error_logger_message(Eterm from, Eterm msg, ErlHeapFragment *bp)
 {
-    enqueue_sys_msg(SYS_MSG_TYPE_ERRLGR, from, am_error_logger, msg, bp);
+    enqueue_sys_msg(SYS_MSG_TYPE_ERRLGR, from, am_logger, msg, bp);
 }
 
 void
@@ -2315,10 +2076,10 @@ sys_msg_disp_failure(ErtsSysMsgQ *smqp, Eterm receiver)
 	    && !erts_system_monitor_flags.busy_port
 	    && !erts_system_monitor_flags.busy_dist_port)
 	    break; /* Everything is disabled */
-	erts_smp_thr_progress_block();
+	erts_thr_progress_block();
 	if (system_monitor == receiver || receiver == NIL)
 	    erts_system_monitor_clear(NULL);
-	erts_smp_thr_progress_unblock();
+	erts_thr_progress_unblock();
 	break;
 	 case SYS_MSG_TYPE_SYSPROF:
 	if (receiver == NIL
@@ -2328,20 +2089,20 @@ sys_msg_disp_failure(ErtsSysMsgQ *smqp, Eterm receiver)
 	    && !erts_system_profile_flags.scheduler)
 		 break;
 	/* Block system to clear flags */
-	erts_smp_thr_progress_block();
+	erts_thr_progress_block();
 	if (system_profile == receiver || receiver == NIL) { 
 		erts_system_profile_clear(NULL);
 	}
-	erts_smp_thr_progress_unblock();
+	erts_thr_progress_unblock();
 	break;
     case SYS_MSG_TYPE_ERRLGR: {
-	char *no_elgger = "(no error logger present)";
+	char *no_elgger = "(no logger present)";
 	Eterm *tp;
 	Eterm tag;
 	if (is_not_tuple(smqp->msg)) {
 	unexpected_elmsg:
 	    erts_fprintf(stderr,
-			 "%s unexpected error logger message: %T\n",
+			 "%s unexpected logger message: %T\n",
 			 no_elgger,
 			 smqp->msg);
 	}
@@ -2377,38 +2138,38 @@ static void
 sys_msg_dispatcher_wakeup(void *vwait_p)
 {
     int *wait_p = (int *) vwait_p;
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     *wait_p = 0;
-    erts_smp_cnd_signal(&smq_cnd);
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_cnd_signal(&smq_cnd);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 static void
 sys_msg_dispatcher_prep_wait(void *vwait_p)
 {
     int *wait_p = (int *) vwait_p;
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     *wait_p = 1;
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 static void
 sys_msg_dispatcher_fin_wait(void *vwait_p)
 {
     int *wait_p = (int *) vwait_p;
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     *wait_p = 0;
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 static void
 sys_msg_dispatcher_wait(void *vwait_p)
 {
     int *wait_p = (int *) vwait_p;
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     while (*wait_p)
-	erts_smp_cnd_wait(&smq_cnd, &smq_mtx);
-    erts_smp_mtx_unlock(&smq_mtx);
+	erts_cnd_wait(&smq_cnd, &smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 static void *
@@ -2434,9 +2195,9 @@ sys_msg_dispatcher_func(void *unused)
 	int end_wait = 0;
 	ErtsSysMsgQ *smqp;
 
-	ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
+	ERTS_LC_ASSERT(!erts_thr_progress_is_blocking());
 
-	erts_smp_mtx_lock(&smq_mtx);
+	erts_mtx_lock(&smq_mtx);
 
 	/* Free previously used queue ... */
 	while (local_sys_message_queue) {
@@ -2447,21 +2208,21 @@ sys_msg_dispatcher_func(void *unused)
 
 	/* Fetch current trace message queue ... */
 	if (!sys_message_queue) {
-	    erts_smp_mtx_unlock(&smq_mtx);
+	    erts_mtx_unlock(&smq_mtx);
 	    end_wait = 1;
 	    erts_thr_progress_active(NULL, 0);
 	    erts_thr_progress_prepare_wait(NULL);
-	    erts_smp_mtx_lock(&smq_mtx);
+	    erts_mtx_lock(&smq_mtx);
 	}
 
 	while (!sys_message_queue)
-	    erts_smp_cnd_wait(&smq_cnd, &smq_mtx);
+	    erts_cnd_wait(&smq_cnd, &smq_mtx);
 
 	local_sys_message_queue = sys_message_queue;
 	sys_message_queue = NULL;
 	sys_message_queue_end = NULL;
 
-	erts_smp_mtx_unlock(&smq_mtx);
+	erts_mtx_unlock(&smq_mtx);
 
 	if (end_wait) {
 	    erts_thr_progress_finalize_wait(NULL);
@@ -2509,7 +2270,7 @@ sys_msg_dispatcher_func(void *unused)
 		}
 		break;
 	    case SYS_MSG_TYPE_ERRLGR:
-		receiver = am_error_logger;
+		receiver = am_logger;
 		break;
 	    default:
 		receiver = NIL;
@@ -2535,10 +2296,10 @@ sys_msg_dispatcher_func(void *unused)
 #ifdef DEBUG_PRINTOUTS
 		    erts_fprintf(stderr, "delivered\n");
 #endif
-		    erts_smp_proc_unlock(proc, proc_locks);
+		    erts_proc_unlock(proc, proc_locks);
 		}
 	    }
-	    else if (receiver == am_error_logger) {
+	    else if (receiver == am_logger) {
 		proc = erts_whereis_process(NULL,0,receiver,proc_locks,0);
 		if (!proc)
 		    goto failure;
@@ -2573,7 +2334,7 @@ sys_msg_dispatcher_func(void *unused)
 		sys_msg_disp_failure(smqp, receiver);
 	    drop_sys_msg:
 		if (proc)
-		    erts_smp_proc_unlock(proc, proc_locks);
+		    erts_proc_unlock(proc, proc_locks);
 		if (smqp->bp)
 		    free_message_buffer(smqp->bp);
 #ifdef DEBUG_PRINTOUTS
@@ -2593,7 +2354,7 @@ erts_foreach_sys_msg_in_q(void (*func)(Eterm,
 				       ErlHeapFragment *))
 {
     ErtsSysMsgQ *sm;
-    erts_smp_mtx_lock(&smq_mtx);
+    erts_mtx_lock(&smq_mtx);
     for (sm = sys_message_queue; sm; sm = sm->next) {
 	Eterm to;
 	switch (sm->type) {
@@ -2604,7 +2365,7 @@ erts_foreach_sys_msg_in_q(void (*func)(Eterm,
 	    to = erts_get_system_profile();
 	    break;
 	case SYS_MSG_TYPE_ERRLGR:
-	    to = am_error_logger;
+	    to = am_logger;
 	    break;
 	default:
 	    to = NIL;
@@ -2612,29 +2373,28 @@ erts_foreach_sys_msg_in_q(void (*func)(Eterm,
 	}
 	(*func)(sm->from, to, sm->msg, sm->bp);
     }
-    erts_smp_mtx_unlock(&smq_mtx);
+    erts_mtx_unlock(&smq_mtx);
 }
 
 
 static void
 init_sys_msg_dispatcher(void)
 {
-    erts_smp_thr_opts_t thr_opts = ERTS_SMP_THR_OPTS_DEFAULT_INITER;
+    erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER;
     thr_opts.detached = 1;
     thr_opts.name = "sys_msg_dispatcher";
     init_smq_element_alloc();
     sys_message_queue = NULL;
     sys_message_queue_end = NULL;
-    erts_smp_cnd_init(&smq_cnd);
-    erts_smp_mtx_init(&smq_mtx, "sys_msg_q", NIL,
+    erts_cnd_init(&smq_cnd);
+    erts_mtx_init(&smq_mtx, "sys_msg_q", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
-    erts_smp_thr_create(&sys_msg_dispatcher_tid,
+    erts_thr_create(&sys_msg_dispatcher_tid,
 			sys_msg_dispatcher_func,
 			NULL,
 			&thr_opts);
 }
 
-#endif
 
 #include "erl_nif.h"
 
@@ -2730,7 +2490,7 @@ static void init_tracer_template(ErtsTracerNif *tnif) {
 }
 
 static Hash *tracer_hash = NULL;
-static erts_smp_rwmtx_t tracer_mtx;
+static erts_rwmtx_t tracer_mtx;
 
 static ErtsTracerNif *
 load_tracer_nif(const ErtsTracer tracer)
@@ -2759,7 +2519,7 @@ load_tracer_nif(const ErtsTracer tracer)
 
     for(i = 0; i < num_of_funcs; i++) {
         for (j = 0; j < NIF_TRACER_TYPES; j++) {
-            if (strcmp(tracers[j].name, funcs[i].name) == 0 && tracers[j].arity == funcs[i].arity) {
+            if (sys_strcmp(tracers[j].name, funcs[i].name) == 0 && tracers[j].arity == funcs[i].arity) {
                 tracers[j].cb = &(funcs[i]);
                 break;
             }
@@ -2770,9 +2530,9 @@ load_tracer_nif(const ErtsTracer tracer)
         return NULL;
     }
 
-    erts_smp_rwmtx_rwlock(&tracer_mtx);
+    erts_rwmtx_rwlock(&tracer_mtx);
     tnif = hash_put(tracer_hash, &tnif_tmpl);
-    erts_smp_rwmtx_rwunlock(&tracer_mtx);
+    erts_rwmtx_rwunlock(&tracer_mtx);
 
     return tnif;
 }
@@ -2783,14 +2543,14 @@ lookup_tracer_nif(const ErtsTracer tracer)
     ErtsTracerNif tnif_tmpl;
     ErtsTracerNif *tnif;
     tnif_tmpl.module = ERTS_TRACER_MODULE(tracer);
-    erts_smp_rwmtx_rlock(&tracer_mtx);
+    erts_rwmtx_rlock(&tracer_mtx);
     if ((tnif = hash_get(tracer_hash, &tnif_tmpl)) == NULL) {
-        erts_smp_rwmtx_runlock(&tracer_mtx);
+        erts_rwmtx_runlock(&tracer_mtx);
         tnif = load_tracer_nif(tracer);
         ASSERT(!tnif || tnif->nif_mod);
         return tnif;
     }
-    erts_smp_rwmtx_runlock(&tracer_mtx);
+    erts_rwmtx_runlock(&tracer_mtx);
     ASSERT(tnif->nif_mod);
     return tnif;
 }
@@ -2819,7 +2579,7 @@ erts_term_to_tracer(Eterm prefix, Eterm t)
                     state = tp[3];
                 }
             } else {
-                if (arityval(tp[0]) == 2 && is_atom(tp[2])) {
+                if (arityval(tp[0]) == 2 && is_atom(tp[1])) {
                     module = tp[1];
                     state = tp[2];
                 }
@@ -2855,6 +2615,38 @@ erts_tracer_to_term(Process *p, ErtsTracer tracer)
     }
 }
 
+Eterm
+erts_build_tracer_to_term(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, ErtsTracer tracer)
+{
+    Eterm res;
+    Eterm state;
+    Uint sz;
+
+    if (ERTS_TRACER_IS_NIL(tracer))
+        return am_false;
+
+    state = ERTS_TRACER_STATE(tracer);
+    sz = is_immed(state) ? 0 : size_object(state);
+
+    if (szp)
+        *szp += sz;
+
+    if (hpp)
+        res = is_immed(state) ? state : copy_struct(state, sz, hpp, ohp);
+    else
+        res = THE_NON_VALUE;
+
+    if (ERTS_TRACER_MODULE(tracer) != am_erl_tracer) {
+        if (szp)
+            *szp += 3;
+        if (hpp) {
+            res = TUPLE2(*hpp, ERTS_TRACER_MODULE(tracer), res);
+            *hpp += 3;
+        }
+    }
+
+    return res;
+}
 
 static ERTS_INLINE int
 send_to_tracer_nif_raw(Process *c_p, Process *tracee,
@@ -2928,17 +2720,17 @@ send_to_tracer_nif(Process *c_p, ErtsPTabElementCommon *t_p,
                    Eterm t_p_id, ErtsTracerNif *tnif, enum ErtsTracerOpt topt,
                    Eterm tag, Eterm msg, Eterm extra, Eterm pam_result)
 {
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     if (c_p) {
         /* We have to hold the main lock of the currently executing process */
         erts_proc_lc_chk_have_proc_locks(c_p, ERTS_PROC_LOCK_MAIN);
     }
     if (is_internal_pid(t_p->id)) {
         /* We have to have at least one lock */
-        ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks((Process*)t_p) & ERTS_PROC_LOCKS_ALL);
+        ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks((Process*)t_p) & ERTS_PROC_LOCKS_ALL);
     } else {
         ASSERT(is_internal_port(t_p->id));
-        ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked((Port*)t_p));
+        ERTS_LC_ASSERT(erts_lc_is_port_locked((Port*)t_p));
     }
 #endif
 
@@ -2981,17 +2773,17 @@ is_tracer_enabled(Process* c_p, ErtsProcLocks c_p_locks,
                   enum ErtsTracerOpt topt, Eterm tag) {
     Eterm nif_result;
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     if (c_p)
-        ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == c_p_locks
+        ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) == c_p_locks
                            || erts_thr_progress_is_blocking());
     if (is_internal_pid(t_p->id)) {
         /* We have to have at least one lock */
-        ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks((Process*)t_p) & ERTS_PROC_LOCKS_ALL
+        ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks((Process*)t_p) & ERTS_PROC_LOCKS_ALL
                            || erts_thr_progress_is_blocking());
     } else {
         ASSERT(is_internal_port(t_p->id));
-        ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked((Port*)t_p)
+        ERTS_LC_ASSERT(erts_lc_is_port_locked((Port*)t_p)
                            || erts_thr_progress_is_blocking());
     }
 #endif
@@ -3015,12 +2807,12 @@ is_tracer_enabled(Process* c_p, ErtsProcLocks c_p_locks,
         ErtsProcLocks c_p_xlocks = 0;
         if (esdp && !ERTS_SCHEDULER_IS_DIRTY(esdp)) {
             if (is_internal_pid(t_p->id)) {
-                ERTS_SMP_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) & ERTS_PROC_LOCK_MAIN);
+                ERTS_LC_ASSERT(erts_proc_lc_my_proc_locks(c_p) & ERTS_PROC_LOCK_MAIN);
                 if (c_p_locks != ERTS_PROC_LOCKS_ALL) {
                     c_p_xlocks = ~c_p_locks & ERTS_PROC_LOCKS_ALL;
-                    if (erts_smp_proc_trylock(c_p, c_p_xlocks) == EBUSY) {
-                        erts_smp_proc_unlock(c_p, c_p_locks & ~ERTS_PROC_LOCK_MAIN);
-                        erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
+                    if (erts_proc_trylock(c_p, c_p_xlocks) == EBUSY) {
+                        erts_proc_unlock(c_p, c_p_locks & ~ERTS_PROC_LOCK_MAIN);
+                        erts_proc_lock(c_p, ERTS_PROC_LOCKS_ALL_MINOR);
                     }
                 }
             }
@@ -3029,7 +2821,7 @@ is_tracer_enabled(Process* c_p, ErtsProcLocks c_p_locks,
             t_p->trace_flags &= ~TRACEE_FLAGS;
 
             if (c_p_xlocks)
-                erts_smp_proc_unlock(c_p, c_p_xlocks);
+                erts_proc_unlock(c_p, c_p_xlocks);
         }
     }
 
@@ -3070,7 +2862,7 @@ int erts_is_tracer_proc_enabled_send(Process* c_p, ErtsProcLocks c_p_locks,
 
 void erts_tracer_replace(ErtsPTabElementCommon *t_p, const ErtsTracer tracer)
 {
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     if (is_internal_pid(t_p->id) && !erts_thr_progress_is_blocking()) {
         erts_proc_lc_chk_have_proc_locks((Process*)t_p, ERTS_PROC_LOCKS_ALL);
     } else if (is_internal_port(t_p->id)) {
@@ -3188,11 +2980,11 @@ erts_tracer_update(ErtsTracer *tracer, const ErtsTracer new_tracer)
 
 static void init_tracer_nif()
 {
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&tracer_mtx, &rwmtx_opt, "tracer_mtx", NIL,
+    erts_rwmtx_init_opt(&tracer_mtx, &rwmtx_opt, "tracer_mtx", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_DEBUG);
 
     erts_tracer_nif_clear();
@@ -3202,7 +2994,7 @@ static void init_tracer_nif()
 int erts_tracer_nif_clear()
 {
 
-    erts_smp_rwmtx_rlock(&tracer_mtx);
+    erts_rwmtx_rlock(&tracer_mtx);
     if (!tracer_hash || tracer_hash->nobjs) {
 
         HashFunctions hf;
@@ -3214,19 +3006,19 @@ int erts_tracer_nif_clear()
         hf.meta_free = (HMFREE_FUN) erts_free;
         hf.meta_print = (HMPRINT_FUN) erts_print;
 
-        erts_smp_rwmtx_runlock(&tracer_mtx);
-        erts_smp_rwmtx_rwlock(&tracer_mtx);
+        erts_rwmtx_runlock(&tracer_mtx);
+        erts_rwmtx_rwlock(&tracer_mtx);
 
         if (tracer_hash)
             hash_delete(tracer_hash);
 
         tracer_hash = hash_new(ERTS_ALC_T_TRACER_NIF, "tracer_hash", 10, hf);
 
-        erts_smp_rwmtx_rwunlock(&tracer_mtx);
+        erts_rwmtx_rwunlock(&tracer_mtx);
         return 1;
     }
 
-    erts_smp_rwmtx_runlock(&tracer_mtx);
+    erts_rwmtx_runlock(&tracer_mtx);
     return 0;
 }
 
@@ -3245,7 +3037,7 @@ static void *tracer_alloc_fun(void* tmpl)
     ErtsTracerNif *obj = erts_alloc(ERTS_ALC_T_TRACER_NIF,
                                     sizeof(ErtsTracerNif) +
                                     sizeof(ErtsThrPrgrLaterOp));
-    memcpy(obj, tmpl, sizeof(*obj));
+    sys_memcpy(obj, tmpl, sizeof(*obj));
     return obj;
 }
 
diff --git a/erts/emulator/beam/erl_trace.h b/erts/emulator/beam/erl_trace.h
index 01fe1e5e23..3228e19809 100644
--- a/erts/emulator/beam/erl_trace.h
+++ b/erts/emulator/beam/erl_trace.h
@@ -87,7 +87,6 @@ void erts_set_system_monitor(Eterm monitor);
 Eterm erts_get_system_monitor(void);
 int erts_is_tracer_valid(Process* p);
 
-#ifdef ERTS_SMP
 void erts_check_my_tracer_proc(Process *);
 void erts_block_sys_msg_dispatcher(void);
 void erts_release_sys_msg_dispatcher(void);
@@ -97,7 +96,6 @@ void erts_foreach_sys_msg_in_q(void (*func)(Eterm,
 					    ErlHeapFragment *));
 void erts_queue_error_logger_message(Eterm, Eterm, ErlHeapFragment *);
 void erts_send_sys_msg_proc(Eterm, Eterm, Eterm, ErlHeapFragment *);
-#endif
 
 void trace_send(Process*, Eterm, Eterm);
 void trace_receive(Process*, Eterm, Eterm, ErtsTracingEvent*);
@@ -149,16 +147,12 @@ erts_bif_trace_epilogue(Process *p, Eterm result, int applying,
 			Uint32 flags_meta, BeamInstr* I,
 			ErtsTracer meta_tracer);
 
-#ifdef ERTS_SMP
 void erts_send_pending_trace_msgs(ErtsSchedulerData *esdp);
-#define ERTS_SMP_CHK_PEND_TRACE_MSGS(ESDP)				\
+#define ERTS_CHK_PEND_TRACE_MSGS(ESDP)				\
 do {									\
     if ((ESDP)->pending_trace_msgs)					\
 	erts_send_pending_trace_msgs((ESDP));				\
 } while (0)
-#else
-#define ERTS_SMP_CHK_PEND_TRACE_MSGS(ESDP)
-#endif
 
 #define seq_trace_output(token, msg, type, receiver, process) \
 seq_trace_output_generic((token), (msg), (type), (receiver), (process), NIL)
@@ -204,6 +198,8 @@ int erts_is_tracer_proc_enabled_send(Process* c_p, ErtsProcLocks c_p_locks,
                                      ErtsPTabElementCommon *t_p);
 int erts_is_tracer_enabled(const ErtsTracer tracer, ErtsPTabElementCommon *t_p);
 Eterm erts_tracer_to_term(Process *p, ErtsTracer tracer);
+Eterm erts_build_tracer_to_term(Eterm **hpp, ErlOffHeap *ohp, Uint *szp, ErtsTracer tracer);
+
 ErtsTracer erts_term_to_tracer(Eterm prefix, Eterm term);
 void erts_tracer_replace(ErtsPTabElementCommon *t_p,
                          const ErtsTracer new_tracer);
diff --git a/erts/emulator/beam/erl_unicode.c b/erts/emulator/beam/erl_unicode.c
index c5fc6f9815..d225916ac5 100644
--- a/erts/emulator/beam/erl_unicode.c
+++ b/erts/emulator/beam/erl_unicode.c
@@ -79,23 +79,23 @@ void erts_init_unicode(void)
     max_loop_limit = CONTEXT_REDS * LOOP_FACTOR;
     /* Non visual BIFs to trap to. */
     erts_init_trap_export(&characters_to_utf8_trap_exp,
-			  am_erlang, am_atom_put("characters_to_utf8_trap",23), 3,
+			  am_erlang, ERTS_MAKE_AM("characters_to_utf8_trap"), 3,
 			  &characters_to_utf8_trap);
 
     erts_init_trap_export(&characters_to_list_trap_1_exp,
-			  am_erlang, am_atom_put("characters_to_list_trap_1",25), 3,
+			  am_erlang, ERTS_MAKE_AM("characters_to_list_trap_1"), 3,
 			  &characters_to_list_trap_1);
 
     erts_init_trap_export(&characters_to_list_trap_2_exp,
-			  am_erlang, am_atom_put("characters_to_list_trap_2",25), 3,
+			  am_erlang, ERTS_MAKE_AM("characters_to_list_trap_2"), 3,
 			  &characters_to_list_trap_2);
 
     erts_init_trap_export(&characters_to_list_trap_3_exp,
-			  am_erlang, am_atom_put("characters_to_list_trap_3",25), 3,
+			  am_erlang, ERTS_MAKE_AM("characters_to_list_trap_3"), 3,
 			  &characters_to_list_trap_3);
 
     erts_init_trap_export(&characters_to_list_trap_4_exp,
-			  am_erlang, am_atom_put("characters_to_list_trap_4",25), 1,
+			  am_erlang, ERTS_MAKE_AM("characters_to_list_trap_4"), 1,
 			  &characters_to_list_trap_4);
 
     c_to_b_int_trap_exportp =  erts_export_put(am_unicode,am_characters_to_binary_int,2);
@@ -144,7 +144,7 @@ static Eterm make_magic_bin_for_restart(Process *p, RestartContext *rc)
 					   cleanup_restart_context_bin);
     RestartContext *restartp = ERTS_MAGIC_BIN_DATA(mbp);
     Eterm *hp;
-    memcpy(restartp,rc,sizeof(RestartContext));
+    sys_memcpy(restartp,rc,sizeof(RestartContext));
     hp = HAlloc(p, ERTS_MAGIC_REF_THING_SIZE);
     return erts_mk_magic_ref(&hp, &MSO(p), mbp);
 }
@@ -254,12 +254,12 @@ static Uint copy_utf8_bin(byte *target, byte *source, Uint size,
 	ASSERT(need > 0);
 	ASSERT(from_source > 0);
 	if (size < from_source) {
-	    memcpy(leftover + (*num_leftovers), source, size);
+	    sys_memcpy(leftover + (*num_leftovers), source, size);
 	    *num_leftovers += size;
 	    return 0;
 	}
 	/* leftover has room for four bytes (see bif) */
-	memcpy(leftover + (*num_leftovers),source,from_source);
+	sys_memcpy(leftover + (*num_leftovers),source,from_source);
 	c = copy_utf8_bin(target, leftover, need, NULL, NULL, &tmp_err_pos, characters);
 	if (tmp_err_pos != 0) {
 	    *err_pos = source;
@@ -291,7 +291,7 @@ static Uint copy_utf8_bin(byte *target, byte *source, Uint size,
 	    size -= 2; copied += 2;
 	} else if (((*source) & ((byte) 0xF0)) == 0xE0) {
 	    if (leftover && size < 3) {
-		memcpy(leftover, source, (int) size);
+		sys_memcpy(leftover, source, (int) size);
 		*num_leftovers = (int) size;
 		break;
 	    }
@@ -313,7 +313,7 @@ static Uint copy_utf8_bin(byte *target, byte *source, Uint size,
 	    size -= 3; copied += 3;
 	} else if (((*source) & ((byte) 0xF8)) == 0xF0) {
 	    if (leftover && size < 4) {
-		memcpy(leftover, source, (int) size);
+		sys_memcpy(leftover, source, (int) size);
 		*num_leftovers = (int) size;
 		break;
 	    }
@@ -2070,7 +2070,7 @@ char *erts_convert_filename_to_encoding(Eterm name, char *statbuf, size_t statbu
 	is_list(name) || 
 	(allow_empty && is_nil(name))) {
 	Sint need;
-	if ((need = erts_native_filename_need(name,encoding)) < 0) {
+	if ((need = erts_native_filename_need(name, encoding)) < 0) {
 	    return NULL;
 	}
 	if (encoding == ERL_FILENAME_WIN_WCHAR) {
@@ -2108,7 +2108,7 @@ char *erts_convert_filename_to_encoding(Eterm name, char *statbuf, size_t statbu
 	    } else {
 		name_buf = statbuf;
 	    }
-	    memcpy(name_buf,bytes,size);
+	    sys_memcpy(name_buf,bytes,size);
 	    name_buf[size]=0;
 	} else {
             name_buf = erts_convert_filename_to_wchar(bytes, size,
@@ -2165,18 +2165,9 @@ char* erts_convert_filename_to_wchar(byte* bytes, Uint size,
     return name_buf;
 }
 
-
-static int filename_len_16bit(byte *str) 
-{
-    byte *p = str;
-    while(*p != '\0' || p[1] != '\0') {
-	p += 2;
-    }
-    return (p - str);
-}
-Eterm erts_convert_native_to_filename(Process *p, byte *bytes)
+Eterm erts_convert_native_to_filename(Process *p, size_t size, byte *bytes)
 {
-    Uint size,num_chars;
+    Uint num_chars;
     Eterm *hp;
     byte *err_pos;
     Uint num_built; /* characters */
@@ -2190,7 +2181,6 @@ Eterm erts_convert_native_to_filename(Process *p, byte *bytes)
     case ERL_FILENAME_UTF8_MAC:
 	mac = 1;
     case ERL_FILENAME_UTF8:
-	size = strlen((char *) bytes);
 	if (size == 0)
 	    return NIL;
 	if (erts_analyze_utf8(bytes,size,&err_pos,&num_chars,NULL) != ERTS_UTF8_OK) {
@@ -2205,7 +2195,6 @@ Eterm erts_convert_native_to_filename(Process *p, byte *bytes)
 	} 
 	return ret;
     case ERL_FILENAME_WIN_WCHAR:
-	size=filename_len_16bit(bytes);
 	if ((size % 2) != 0) { /* Panic fixup to avoid crashing the emulator */
 	    size--;
 	    hp = HAlloc(p, size+2);
@@ -2228,13 +2217,12 @@ Eterm erts_convert_native_to_filename(Process *p, byte *bytes)
 	goto noconvert;
     }
  noconvert:
-    size = strlen((char *) bytes);
     hp = HAlloc(p, 2 * size);
     return erts_bin_bytes_to_list(NIL, hp, bytes, size, 0);
 }
 
 
-Sint erts_native_filename_need(Eterm ioterm, int encoding) 
+Sint erts_native_filename_need(Eterm ioterm, int encoding)
 {
     Eterm *objp;
     Eterm obj;
@@ -2276,6 +2264,20 @@ Sint erts_native_filename_need(Eterm ioterm, int encoding)
 	default:
 	    need = -1;
 	}
+        /*
+         * Do not allow null in
+         * the middle of filenames
+         */
+        if (need > 0) {
+            byte *name = ap->name;
+            int len = ap->len;
+            for (i = 0; i < len; i++) {
+                if (name[i] == 0) {
+                    need = -1;
+                    break;
+                }
+            }
+        }
 	DESTROY_ESTACK(stack);
 	return need;
     }
@@ -2306,6 +2308,14 @@ L_Again:   /* Restart with sublist, old listend was pushed on stack */
 		if (is_small(obj)) { /* Always small */
 		    for(;;) {
 			Uint x = unsigned_val(obj);
+                        /*
+                         * Do not allow null in
+                         * the middle of filenames
+                         */
+                        if (x == 0) {
+                            DESTROY_ESTACK(stack);
+                            return ((Sint) -1);
+                        }
 			switch (encoding) {
 			case ERL_FILENAME_LATIN1:
 			    if (x > 255) {
@@ -2579,6 +2589,38 @@ void erts_copy_utf8_to_utf16_little(byte *target, byte *bytes, int num_chars)
 }
 
 /*
+ * *** Requirements on Raw Filename Format ***
+ *
+ * These requirements are due to the 'filename' module
+ * in stdlib. This since it is documented that it
+ * should be able to operate on raw filenames as well
+ * as ordinary filenames.
+ *
+ * A raw filename *must* be a byte sequence where:
+ * 1. Codepoints 0-127 (7-bit ascii) *must* be encoded
+ *    as a byte with the corresponding value. That is,
+ *    the most significant bit in the byte encoding the
+ *    codepoint is never set.
+ * 2. Codepoints greater than 127 *must* be encoded
+ *    with the most significant bit set in *every* byte
+ *    encoding it.
+ *
+ * Latin1 and UTF-8 meet these requirements while
+ * UTF-16 and UTF-32 don't.
+ *
+ * On Windows filenames are natively stored as malformed
+ * UTF-16LE (lonely surrogates may appear). A more correct
+ * description than UTF-16 would be an array of 16-bit
+ * words... In order to meet the requirements of the
+ * raw file format we convert the malformed UTF-16LE to
+ * malformed UTF-8 which meet the requirements.
+ *
+ * Note that these requirements are today only OTP
+ * internal (erts-stdlib internal) requirements that
+ * could be changed.
+ */
+
+/*
  * This internal bif converts a filename to whatever format is suitable for the file driver
  * It also adds zero termination so that prim_file needn't bother with the character encoding
  * of the file driver 
@@ -2589,6 +2631,12 @@ BIF_RETTYPE prim_file_internal_name2native_1(BIF_ALIST_1)
     Sint need;
     Eterm bin_term;
     byte* bin_p;
+
+    /*
+     * See comment on "Requirements on Raw Filename Format"
+     * above.
+     */
+
     /* Prim file explicitly does not allow atoms, although we could 
        very well cope with it. Instead of letting 'file' handle them,
        it would probably be more efficient to handle them here. Subject to 
@@ -2606,10 +2654,16 @@ BIF_RETTYPE prim_file_internal_name2native_1(BIF_ALIST_1)
 	size = binary_size(BIF_ARG_1);
 	bytes = erts_get_aligned_binary_bytes(BIF_ARG_1, &temp_alloc);
 	if (encoding != ERL_FILENAME_WIN_WCHAR) {
+            Uint i;
 	    /*Add 0 termination only*/
 	    bin_term = new_binary(BIF_P, NULL, size+1);
 	    bin_p = binary_bytes(bin_term);
-	    memcpy(bin_p,bytes,size);
+            for (i = 0; i < size; i++) {
+                /* Don't allow null in the middle of filenames... */
+                if (bytes[i] == 0)
+                    goto bin_name_error;
+                bin_p[i] = bytes[i];
+            }
 	    bin_p[size]=0;
 	    erts_free_aligned_binary_bytes(temp_alloc);
 	    BIF_RET(bin_term);
@@ -2623,6 +2677,9 @@ BIF_RETTYPE prim_file_internal_name2native_1(BIF_ALIST_1)
 	    bin_term = new_binary(BIF_P, 0, (size+1)*2);
 	    bin_p = binary_bytes(bin_term);
 	    while (size--) {
+                /* Don't allow null in the middle of filenames... */
+                if (*bytes == 0)
+                    goto bin_name_error;
 		*bin_p++ = *bytes++;
 		*bin_p++ = 0;
 	    }
@@ -2640,11 +2697,14 @@ BIF_RETTYPE prim_file_internal_name2native_1(BIF_ALIST_1)
 	bin_p[num_chars*2+1] = 0;
 	erts_free_aligned_binary_bytes(temp_alloc);
 	BIF_RET(bin_term);
+    bin_name_error:
+        erts_free_aligned_binary_bytes(temp_alloc);
+        BIF_ERROR(BIF_P,BADARG);
     } /* binary */   
 	    
 
-    if ((need = erts_native_filename_need(BIF_ARG_1,encoding)) < 0) {
-	BIF_ERROR(BIF_P,BADARG);
+    if ((need = erts_native_filename_need(BIF_ARG_1, encoding)) < 0) {
+        BIF_ERROR(BIF_P,BADARG);
     }
     if (encoding == ERL_FILENAME_WIN_WCHAR) {
 	need += 2;
@@ -2678,6 +2738,11 @@ BIF_RETTYPE prim_file_internal_native2name_1(BIF_ALIST_1)
     Eterm ret;
     int mac = 0;
 
+    /*
+     * See comment on "Requirements on Raw Filename Format"
+     * above.
+     */
+
     if (is_not_binary(BIF_ARG_1)) {
 	BIF_ERROR(BIF_P,BADARG);
     }
diff --git a/erts/emulator/beam/erl_utils.h b/erts/emulator/beam/erl_utils.h
index 3d28b05752..e4087e0ac8 100644
--- a/erts/emulator/beam/erl_utils.h
+++ b/erts/emulator/beam/erl_utils.h
@@ -22,15 +22,11 @@
 #define ERL_UTILS_H__
 
 #include "sys.h"
-#include "erl_smp.h"
 #include "erl_printf.h"
 
 struct process;
 
 typedef struct {
-#ifdef DEBUG
-    int smp_api;
-#endif
     union {
 	Uint64 not_atomic;
 	erts_atomic64_t atomic;
@@ -38,70 +34,25 @@ typedef struct {
 } erts_interval_t;
 
 void erts_interval_init(erts_interval_t *);
-void erts_smp_interval_init(erts_interval_t *);
 Uint64 erts_step_interval_nob(erts_interval_t *);
 Uint64 erts_step_interval_relb(erts_interval_t *);
-Uint64 erts_smp_step_interval_nob(erts_interval_t *);
-Uint64 erts_smp_step_interval_relb(erts_interval_t *);
 Uint64 erts_ensure_later_interval_nob(erts_interval_t *, Uint64);
 Uint64 erts_ensure_later_interval_acqb(erts_interval_t *, Uint64);
-Uint64 erts_smp_ensure_later_interval_nob(erts_interval_t *, Uint64);
-Uint64 erts_smp_ensure_later_interval_acqb(erts_interval_t *, Uint64);
-ERTS_GLB_INLINE Uint64 erts_current_interval_nob__(erts_interval_t *);
-ERTS_GLB_INLINE Uint64 erts_current_interval_acqb__(erts_interval_t *);
 ERTS_GLB_INLINE Uint64 erts_current_interval_nob(erts_interval_t *);
 ERTS_GLB_INLINE Uint64 erts_current_interval_acqb(erts_interval_t *);
-ERTS_GLB_INLINE Uint64 erts_smp_current_interval_nob(erts_interval_t *);
-ERTS_GLB_INLINE Uint64 erts_smp_current_interval_acqb(erts_interval_t *);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
 ERTS_GLB_INLINE Uint64
-erts_current_interval_nob__(erts_interval_t *icp)
-{
-    return (Uint64) erts_atomic64_read_nob(&icp->counter.atomic);
-}
-
-ERTS_GLB_INLINE Uint64
-erts_current_interval_acqb__(erts_interval_t *icp)
-{
-    return (Uint64) erts_atomic64_read_acqb(&icp->counter.atomic);
-}
-
-ERTS_GLB_INLINE Uint64
 erts_current_interval_nob(erts_interval_t *icp)
 {
-    ASSERT(!icp->smp_api);
-    return erts_current_interval_nob__(icp);
+    return (Uint64) erts_atomic64_read_nob(&icp->counter.atomic);
 }
 
 ERTS_GLB_INLINE Uint64
 erts_current_interval_acqb(erts_interval_t *icp)
 {
-    ASSERT(!icp->smp_api);
-    return erts_current_interval_acqb__(icp);
-}
-
-ERTS_GLB_INLINE Uint64
-erts_smp_current_interval_nob(erts_interval_t *icp)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
-    return erts_current_interval_nob__(icp);
-#else
-    return icp->counter.not_atomic;
-#endif
-}
-
-ERTS_GLB_INLINE Uint64
-erts_smp_current_interval_acqb(erts_interval_t *icp)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
-    return erts_current_interval_acqb__(icp);
-#else
-    return icp->counter.not_atomic;
-#endif
+    return (Uint64) erts_atomic64_read_acqb(&icp->counter.atomic);
 }
 
 #endif /* ERTS_GLB_INLINE_INCL_FUNC_DEF */
@@ -140,7 +91,7 @@ Eterm erts_bld_tuple(Uint **hpp, Uint *szp, Uint arity, ...);
 #define erts_bld_tuple5(H,S,E1,E2,E3,E4,E5) erts_bld_tuple(H,S,5,E1,E2,E3,E4,E5)
 Eterm erts_bld_tuplev(Uint **hpp, Uint *szp, Uint arity, Eterm terms[]);
 Eterm erts_bld_string_n(Uint **hpp, Uint *szp, const char *str, Sint len);
-#define erts_bld_string(hpp,szp,str) erts_bld_string_n(hpp,szp,str,strlen(str))
+#define erts_bld_string(hpp,szp,str) erts_bld_string_n(hpp,szp,str,sys_strlen(str))
 Eterm erts_bld_list(Uint **hpp, Uint *szp, Sint length, Eterm terms[]);
 Eterm erts_bld_2tup_list(Uint **hpp, Uint *szp,
 			 Sint length, Eterm terms1[], Uint terms2[]);
diff --git a/erts/emulator/beam/erl_vm.h b/erts/emulator/beam/erl_vm.h
index 0b8d78c469..f8391fb665 100644
--- a/erts/emulator/beam/erl_vm.h
+++ b/erts/emulator/beam/erl_vm.h
@@ -46,8 +46,6 @@
  */
 #define ERTS_X_REGS_ALLOCATED (MAX_REG+3)
 
-#define INPUT_REDUCTIONS (2 * CONTEXT_REDS)
-
 #define H_DEFAULT_SIZE  233        /* default (heap + stack) min size */
 #define VH_DEFAULT_SIZE  32768     /* default virtual (bin) heap min size (words) */
 #define H_DEFAULT_MAX_SIZE 0       /* default max heap size is off */
@@ -55,7 +53,7 @@
 #define CP_SIZE 1
 
 #define ErtsHAllocLockCheck(P) \
-  ERTS_SMP_LC_ASSERT(erts_dbg_check_halloc_lock((P)))
+  ERTS_LC_ASSERT(erts_dbg_check_halloc_lock((P)))
 
 
 #ifdef DEBUG
@@ -71,7 +69,7 @@
 #  ifdef CHECK_FOR_HOLES
 #    define INIT_HEAP_MEM(p,sz) erts_set_hole_marker(HEAP_TOP(p), (sz))
 #  else
-#    define INIT_HEAP_MEM(p,sz) memset(HEAP_TOP(p),0x01,(sz)*sizeof(Eterm*))
+#    define INIT_HEAP_MEM(p,sz) sys_memset(HEAP_TOP(p),0x01,(sz)*sizeof(Eterm*))
 #  endif
 #else
 #  define INIT_HEAP_MEM(p,sz) ((void)0)
@@ -102,9 +100,11 @@
   if ((ptr) == (endp)) {					\
      ;								\
   } else if (HEAP_START(p) <= (ptr) && (ptr) < HEAP_TOP(p)) {	\
+     ASSERT(HEAP_TOP(p) == (endp));                             \
      HEAP_TOP(p) = (ptr);					\
   } else {							\
-     erts_heap_frag_shrink(p, ptr);					\
+     ASSERT(MBUF(p)->mem + MBUF(p)->used_size == (endp));       \
+     erts_heap_frag_shrink(p, ptr);                             \
   }
 
 #define HeapWordsLeft(p) (HEAP_LIMIT(p) - HEAP_TOP(p))
@@ -157,6 +157,7 @@ typedef struct op_entry {
    Uint32 mask[3];		/* Signature mask. */
    unsigned involves_r;		/* Needs special attention when matching. */
    int sz;			/* Number of loaded words. */
+   int adjust;                  /* Adjustment for start of instruction. */
    char* pack;			/* Instructions for packing engine. */
    char* sign;			/* Signature string. */
 } OpEntry;
@@ -199,11 +200,24 @@ extern int erts_pd_initial_size;/* Initial Process dictionary table size */
 
 #include "erl_term.h"
 
-#ifdef NO_JUMP_TABLE 
-#define BeamOp(Op) (Op)
+#if defined(NO_JUMP_TABLE)
+#  define BeamOpsAreInitialized() (1)
+#  define BeamOpCodeAddr(OpCode) ((BeamInstr)(OpCode))
 #else
 extern void** beam_ops;
-#define BeamOp(Op) beam_ops[(Op)]
+#  define BeamOpsAreInitialized() (beam_ops != 0)
+#  define BeamOpCodeAddr(OpCode) ((BeamInstr)beam_ops[(OpCode)])
 #endif
 
+#if defined(ARCH_64) && defined(CODE_MODEL_SMALL)
+#  define BeamCodeAddr(InstrWord) ((BeamInstr)(Uint32)(InstrWord))
+#  define BeamSetCodeAddr(InstrWord, Addr) (((InstrWord) & ~((1ull << 32)-1)) | (Addr))
+#  define BeamExtraData(InstrWord) ((InstrWord) >> 32)
+#else
+#  define BeamCodeAddr(InstrWord) ((BeamInstr)(InstrWord))
+#  define BeamSetCodeAddr(InstrWord, Addr) (Addr)
+#endif
+
+#define BeamIsOpCode(InstrWord, OpCode) (BeamCodeAddr(InstrWord) == BeamOpCodeAddr(OpCode))
+
 #endif	/* __ERL_VM_H__ */
diff --git a/erts/emulator/beam/erlang_dtrace.d b/erts/emulator/beam/erlang_dtrace.d
index 237889e0f5..c47a37eb62 100644
--- a/erts/emulator/beam/erlang_dtrace.d
+++ b/erts/emulator/beam/erlang_dtrace.d
@@ -55,7 +55,8 @@ provider erlang {
      * @param sender the PID (string form) of the sender
      * @param receiver the PID (string form) of the receiver
      * @param size the size of the message being delivered (words)
-     * @param token_label for the sender's sequential trace token
+     * @param token_label for the sender's sequential trace token. This will be
+     *        INT_MIN if the label does not fit into a 32-bit integer.
      * @param token_previous count for the sender's sequential trace token
      * @param token_current count for the sender's sequential trace token
      */
@@ -73,7 +74,8 @@ provider erlang {
      * @param node_name the Erlang node name (string form) of the receiver
      * @param receiver the PID/name (string form) of the receiver
      * @param size the size of the message being delivered (words)
-     * @param token_label for the sender's sequential trace token
+     * @param token_label for the sender's sequential trace token. This will be
+     *        INT_MIN if the label does not fit into a 32-bit integer.
      * @param token_previous count for the sender's sequential trace token
      * @param token_current count for the sender's sequential trace token
      */
@@ -98,7 +100,8 @@ provider erlang {
      * @param receiver the PID (string form) of the receiver
      * @param size the size of the message being delivered (words)
      * @param queue_len length of the queue of the receiving process
-     * @param token_label for the sender's sequential trace token
+     * @param token_label for the sender's sequential trace token. This will be
+     *        INT_MIN if the label does not fit into a 32-bit integer.
      * @param token_previous count for the sender's sequential trace token
      * @param token_current count for the sender's sequential trace token
      */
@@ -122,7 +125,8 @@ provider erlang {
      * @param receiver the PID (string form) of the receiver
      * @param size the size of the message being delivered (words)
      * @param queue_len length of the queue of the receiving process
-     * @param token_label for the sender's sequential trace token
+     * @param token_label for the sender's sequential trace token. This will be
+     *        INT_MIN if the label does not fit into a 32-bit integer.
      * @param token_previous count for the sender's sequential trace token
      * @param token_current count for the sender's sequential trace token
      */
@@ -273,7 +277,8 @@ provider erlang {
      * @param node_name the Erlang node name (string form) of the receiver
      * @param receiver the PID (string form) of the process receiving EXIT signal
      * @param reason the reason for the exit (may be truncated)
-     * @param token_label for the sender's sequential trace token
+     * @param token_label for the sender's sequential trace token. This will be
+     *        INT_MIN if the label does not fit into a 32-bit integer.
      * @param token_previous count for the sender's sequential trace token
      * @param token_current count for the sender's sequential trace token
      */
@@ -634,72 +639,6 @@ provider erlang {
      */
     probe aio_pool__get(char *, int);
 
-    /* Probes for efile_drv.c */
-
-    /**
-     * Entry into the efile_drv.c file I/O driver
-     *
-     * For a list of command numbers used by this driver, see the section
-     * "Guide to efile_drv.c probe arguments" in ../../../HOWTO/DTRACE.md.
-     * That section also contains explanation of the various integer and
-     * string arguments that may be present when any particular probe fires.
-     *
-     * NOTE: Not all Linux platforms (using SystemTap) can support
-     *       arguments beyond arg9.
-     *
-     *
-     * TODO: Adding the port string, args[10], is a pain.  Making that
-     *       port string available to all the other efile_drv.c probes
-     *       will be more pain.  Is the pain worth it?  If yes, then
-     *       add them everywhere else and grit our teeth.  If no, then
-     *       rip it out.
-     *
-     * @param thread-id number of the scheduler Pthread                   arg0
-     * @param tag number: {thread-id, tag} uniquely names a driver operation
-     * @param user-tag string                                             arg2
-     * @param command number                                              arg3
-     * @param string argument 1                                           arg4
-     * @param string argument 2                                           arg5
-     * @param integer argument 1                                          arg6
-     * @param integer argument 2                                          arg7
-     * @param integer argument 3                                          arg8
-     * @param integer argument 4                                          arg9
-     * @param port the port ID of the busy port                       args[10]
-     */
-    probe efile_drv__entry(int, int, char *, int, char *, char *,
-                           int64_t, int64_t, int64_t, int64_t, char *);
-
-    /**
-     * Entry into the driver's internal work function.  Computation here
-     * is performed by a async worker pool Pthread.
-     *
-     * @param thread-id number
-     * @param tag number
-     * @param command number
-     */
-    probe efile_drv__int_entry(int, int, int);
-
-    /**
-     * Return from the driver's internal work function.
-     *
-     * @param thread-id number
-     * @param tag number
-     * @param command number
-     */
-    probe efile_drv__int_return(int, int, int);
-
-    /**
-     * Return from the efile_drv.c file I/O driver
-     *
-     * @param thread-id number                                            arg0
-     * @param tag number                                                  arg1
-     * @param user-tag string                                             arg2
-     * @param command number                                              arg3
-     * @param Success? 1 is success, 0 is failure                         arg4
-     * @param If failure, the errno of the error.                         arg5
-     */
-    probe efile_drv__return(int, int, char *, int, int, int);
-
 
 /*
  * The set of probes called by the erlang tracer nif backend. In order
diff --git a/erts/emulator/beam/erlang_lttng.h b/erts/emulator/beam/erlang_lttng.h
index 4e869671f7..feb05f4f4c 100644
--- a/erts/emulator/beam/erlang_lttng.h
+++ b/erts/emulator/beam/erlang_lttng.h
@@ -159,21 +159,6 @@ TRACEPOINT_EVENT(
 
 TRACEPOINT_EVENT(
     org_erlang_otp,
-    driver_event,
-    TP_ARGS(
-        char*, pid,
-        char*, port,
-        char*, driver
-    ),
-    TP_FIELDS(
-        ctf_string(pid, pid)
-        ctf_string(port, port)
-        ctf_string(driver, driver)
-    )
-)
-
-TRACEPOINT_EVENT(
-    org_erlang_otp,
     driver_timeout,
     TP_ARGS(
         char*, pid,
diff --git a/erts/emulator/beam/export.c b/erts/emulator/beam/export.c
index 828c833ffc..946ffeffb8 100644
--- a/erts/emulator/beam/export.c
+++ b/erts/emulator/beam/export.c
@@ -41,16 +41,13 @@
 
 static IndexTable export_tables[ERTS_NUM_CODE_IX];  /* Active not locked */
 
-static erts_smp_atomic_t total_entries_bytes;
-
-#include "erl_smp.h"
+static erts_atomic_t total_entries_bytes;
 
 /* This lock protects the staging export table from concurrent access
  * AND it protects the staging table from becoming active.
  */
-erts_smp_mtx_t export_staging_lock;
+erts_mtx_t export_staging_lock;
 
-extern BeamInstr* em_call_error_handler;
 extern BeamInstr* em_call_traced_function;
 
 struct export_entry
@@ -85,17 +82,13 @@ static struct export_blob* entry_to_blob(struct export_entry* ee)
 void
 export_info(fmtfn_t to, void *to_arg)
 {
-#ifdef ERTS_SMP
     int lock = !ERTS_IS_CRASH_DUMPING;
     if (lock)
 	export_staging_lock();
-#endif
     index_info(to, to_arg, &export_tables[erts_active_code_ix()]);
     hash_info(to, to_arg, &export_tables[erts_staging_code_ix()].htable);
-#ifdef ERTS_SMP
     if (lock)
 	export_staging_unlock();
-#endif
 }
 
 
@@ -129,14 +122,17 @@ export_alloc(struct export_entry* tmpl_e)
 	Export* obj;
 
 	blob = (struct export_blob*) erts_alloc(ERTS_ALC_T_EXPORT, sizeof(*blob));
-	erts_smp_atomic_add_nob(&total_entries_bytes, sizeof(*blob));
+	erts_atomic_add_nob(&total_entries_bytes, sizeof(*blob));
 	obj = &blob->exp;
 	obj->info.op =  0;
 	obj->info.u.gen_bp = NULL;
 	obj->info.mfa.module = tmpl->info.mfa.module;
 	obj->info.mfa.function = tmpl->info.mfa.function;
 	obj->info.mfa.arity = tmpl->info.mfa.arity;
-	obj->beam[0] = (BeamInstr) em_call_error_handler;
+        obj->beam[0] = 0;
+        if (BeamOpsAreInitialized()) {
+            obj->beam[0] = BeamOpCodeAddr(op_call_error_handler);
+        }
 	obj->beam[1] = 0;
 
 	for (ix=0; ix<ERTS_NUM_CODE_IX; ix++) {
@@ -173,7 +169,7 @@ export_free(struct export_entry* obj)
     }
     DBG_TRACE_MFA_P(&blob->exp.info.mfa, "export blob deallocation at %p", &blob->exp);
     erts_free(ERTS_ALC_T_EXPORT, blob);
-    erts_smp_atomic_add_nob(&total_entries_bytes, -sizeof(*blob));
+    erts_atomic_add_nob(&total_entries_bytes, -sizeof(*blob));
 }
 
 void
@@ -182,9 +178,9 @@ init_export_table(void)
     HashFunctions f;
     int i;
 
-    erts_smp_mtx_init(&export_staging_lock, "export_tab", NIL,
+    erts_mtx_init(&export_staging_lock, "export_tab", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
-    erts_smp_atomic_init_nob(&total_entries_bytes, 0);
+    erts_atomic_init_nob(&total_entries_bytes, 0);
 
     f.hash = (H_FUN) export_hash;
     f.cmp  = (HCMP_FUN) export_cmp;
@@ -273,7 +269,7 @@ erts_find_function(Eterm m, Eterm f, unsigned int a, ErtsCodeIndex code_ix)
     ee = hash_get(&export_tables[code_ix].htable, init_template(&templ, m, f, a));
     if (ee == NULL ||
 	(ee->ep->addressv[code_ix] == ee->ep->beam &&
-	 ee->ep->beam[0] != (BeamInstr) BeamOp(op_i_generic_breakpoint))) {
+	 ! BeamIsOpCode(ee->ep->beam[0], op_i_generic_breakpoint))) {
 	return NULL;
     }
     return ee->ep;
@@ -373,7 +369,7 @@ int export_table_sz(void)
 }
 int export_entries_sz(void)
 {
-    return erts_smp_atomic_read_nob(&total_entries_bytes);
+    return erts_atomic_read_nob(&total_entries_bytes);
 }
 Export *export_get(Export *e)
 {
diff --git a/erts/emulator/beam/export.h b/erts/emulator/beam/export.h
index 7c812b306c..194e514b12 100644
--- a/erts/emulator/beam/export.h
+++ b/erts/emulator/beam/export.h
@@ -66,14 +66,14 @@ Export *export_get(Export*);
 void export_start_staging(void);
 void export_end_staging(int commit);
 
-extern erts_smp_mtx_t export_staging_lock;
-#define export_staging_lock()	erts_smp_mtx_lock(&export_staging_lock)
-#define export_staging_unlock()	erts_smp_mtx_unlock(&export_staging_lock)
+extern erts_mtx_t export_staging_lock;
+#define export_staging_lock()	erts_mtx_lock(&export_staging_lock)
+#define export_staging_unlock()	erts_mtx_unlock(&export_staging_lock)
 
 #include "beam_load.h" /* For em_* extern declarations */ 
 #define ExportIsBuiltIn(EntryPtr) 			\
 (((EntryPtr)->addressv[erts_active_code_ix()] == (EntryPtr)->beam) && \
- ((EntryPtr)->beam[0] == (BeamInstr) em_apply_bif))
+ (BeamIsOpCode((EntryPtr)->beam[0], op_apply_bif)))
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
diff --git a/erts/emulator/beam/external.c b/erts/emulator/beam/external.c
index c0a3838d42..fb42969a8f 100644
--- a/erts/emulator/beam/external.c
+++ b/erts/emulator/beam/external.c
@@ -122,8 +122,9 @@ static int encode_size_struct_int(struct TTBSizeContext_*, ErtsAtomCacheMap *acm
 
 static Export binary_to_term_trap_export;
 static BIF_RETTYPE binary_to_term_trap_1(BIF_ALIST_1);
-static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binary* context_b,
-				      Export *bif, Eterm arg0, Eterm arg1);
+static Sint transcode_dist_obuf(ErtsDistOutputBuf*, DistEntry*, Uint32 dflags, Sint reds);
+
+
 
 void erts_init_external(void) {
     erts_init_trap_export(&term_to_binary_trap_export,
@@ -222,23 +223,10 @@ erts_destroy_atom_cache_map(ErtsAtomCacheMap *acmp)
 static ERTS_INLINE void
 insert_acache_map(ErtsAtomCacheMap *acmp, Eterm atom, Uint32 dflags)
 {
-    /*
-     * If the receiver do not understand utf8 atoms
-     * and this atom cannot be represented in latin1,
-     * we are not allowed to cache it.
-     *
-     * In this case all atoms are assumed to have
-     * latin1 encoding in the cache. By refusing it
-     * in the cache we will instead encode it using
-     * ATOM_UTF8_EXT/SMALL_ATOM_UTF8_EXT which the
-     * receiver do not recognize and tear down the
-     * connection.
-     */
-    if (acmp && acmp->sz < ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES
-	&& ((dflags & DFLAG_UTF8_ATOMS)
-	    || atom_tab(atom_val(atom))->latin1_chars >= 0)) {
+    if (acmp && acmp->sz < ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES) {
 	int ix;
 	ASSERT(acmp->hdr_sz < 0);
+        ASSERT(dflags & DFLAG_UTF8_ATOMS);
 	ix = atom2cix(atom);
 	if (acmp->cache[ix].iix < 0) {
 	    acmp->cache[ix].iix = acmp->sz;
@@ -258,9 +246,7 @@ get_iix_acache_map(ErtsAtomCacheMap *acmp, Eterm atom, Uint32 dflags)
 	ASSERT(is_atom(atom));
 	ix = atom2cix(atom);
 	if (acmp->cache[ix].iix < 0) {
-	    ASSERT(acmp->sz == ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES
-		   || (!(dflags & DFLAG_UTF8_ATOMS)
-		       && atom_tab(atom_val(atom))->latin1_chars < 0));
+	    ASSERT(acmp->sz == ERTS_MAX_INTERNAL_ATOM_CACHE_ENTRIES);
 	    return -1;
 	}
 	else {
@@ -274,7 +260,6 @@ void
 erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp, Uint32 dflags)
 {
     if (acmp) {
-	int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
 	int long_atoms = 0; /* !0 if one or more atoms are longer than 255. */
 	int i;
 	int sz;
@@ -285,6 +270,7 @@ erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp, Uint32 dflags)
 	    + 1 /* number of internal cache entries */
 	    ;
 	int min_sz;
+        ASSERT(dflags & DFLAG_UTF8_ATOMS);
 	ASSERT(acmp->hdr_sz < 0);
 	/* Make sure cache update instructions fit */
 	min_sz = fix_sz+(2+4)*acmp->sz;
@@ -296,7 +282,7 @@ erts_finalize_atom_cache_map(ErtsAtomCacheMap *acmp, Uint32 dflags)
 	    atom = acmp->cache[acmp->cix[i]].atom;
 	    ASSERT(is_atom(atom));
 	    a = atom_tab(atom_val(atom));
-	    len = (int) (utf8_atoms ? a->len : a->latin1_chars);
+	    len = (int) a->len;
 	    ASSERT(len >= 0);
 	    if (!long_atoms && len > 255)
 		long_atoms = 1;
@@ -366,18 +352,77 @@ byte *erts_encode_ext_dist_header_setup(byte *ctl_ext, ErtsAtomCacheMap *acmp)
     }
 }
 
-byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache, Uint32 dflags)
+
+#define PASS_THROUGH 'p'
+
+Sint erts_encode_ext_dist_header_finalize(ErtsDistOutputBuf* ob,
+                                          DistEntry* dep,
+                                          Uint32 dflags,
+                                          Sint reds)
 {
     byte *ip;
     byte instr_buf[(2+4)*ERTS_ATOM_CACHE_SIZE];
     int ci, sz;
     byte dist_hdr_flags;
     int long_atoms;
-    int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
-    register byte *ep = ext;
-    ASSERT(ep[0] == VERSION_MAGIC);
-    if (ep[1] != DIST_HEADER)
-	return ext;
+    register byte *ep = ob->extp;
+    ASSERT(dflags & DFLAG_UTF8_ATOMS);
+
+    /*
+     * The buffer can have different layouts at this point depending on
+     * what was known when encoded:
+     *
+     * Pending connection: CtrlTerm [, MsgTerm]
+     * With atom cache   : VERSION_MAGIC, DIST_HEADER, ..., CtrlTerm [, MsgTerm]
+     * No atom cache     : VERSION_MAGIC, CtrlTerm [, VERSION_MAGIC, MsgTerm]
+     */
+
+    if (ep[0] != VERSION_MAGIC || dep->transcode_ctx) {
+        /*
+         * Was encoded without atom cache toward pending connection.
+         */
+        ASSERT(ep[0] == SMALL_TUPLE_EXT || ep[0] == LARGE_TUPLE_EXT);
+
+        if (~dflags & (DFLAG_DIST_MONITOR | DFLAG_DIST_MONITOR_NAME)
+            && ep[0] == SMALL_TUPLE_EXT
+            && ep[1] == 4
+            && ep[2] == SMALL_INTEGER_EXT
+            && (ep[3] == DOP_MONITOR_P ||
+                ep[3] == DOP_MONITOR_P_EXIT ||
+                ep[3] == DOP_DEMONITOR_P)) {
+            /*
+             * Receiver does not support process monitoring.
+             * Suppress monitor control msg (see erts_dsig_send_monitor)
+             * by converting it to an empty (tick) packet.
+             */
+            ob->ext_endp = ob->extp;
+            return reds;
+        }
+        if (~dflags & (DFLAG_BIT_BINARIES | DFLAG_EXPORT_PTR_TAG
+                       | DFLAG_DIST_HDR_ATOM_CACHE)) {
+            reds = transcode_dist_obuf(ob, dep, dflags, reds);
+            if (reds < 0)
+                return reds;
+            ep = ob->extp;
+        }
+        if (dflags & DFLAG_DIST_HDR_ATOM_CACHE) {
+            /*
+             * Encoding was done without atom caching but receiver expects
+             * a dist header, so we prepend an empty one.
+             */
+            *--ep = 0; /* NumberOfAtomCacheRefs */
+            *--ep = DIST_HEADER;
+            *--ep = VERSION_MAGIC;
+        }
+        goto done;
+    }
+    else if (ep[1] != DIST_HEADER) {
+        ASSERT(ep[1] == SMALL_TUPLE_EXT || ep[1] == LARGE_TUPLE_EXT);
+        ASSERT(!(dflags & DFLAG_DIST_HDR_ATOM_CACHE));
+        /* Node without atom cache, 'pass through' needed */
+        *--ep = PASS_THROUGH;
+        goto done;
+    }
 
     dist_hdr_flags = ep[2];
     long_atoms = ERTS_DIST_HDR_LONG_ATOMS_FLG & ((int) dist_hdr_flags);
@@ -405,6 +450,7 @@ byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache, Uint
 			 / sizeof(Uint32))+1];
 	register Uint32 flgs;
 	int iix, flgs_bytes, flgs_buf_ix, used_half_bytes;
+        ErtsAtomCache* cache = dep->cache;
 #ifdef DEBUG
 	int tot_used_half_bytes;
 #endif
@@ -447,17 +493,9 @@ byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache, Uint
 		Atom *a;
 		cache->out_arr[cix] = atom;
 		a = atom_tab(atom_val(atom));
-		if (utf8_atoms) {
-		    sz = a->len;
-		    ep -= sz;
-		    sys_memcpy((void *) ep, (void *) a->name, sz);
-		}
-		else {
-		    ASSERT(0 <= a->latin1_chars && a->latin1_chars <= MAX_ATOM_CHARACTERS);
-		    ep -= a->latin1_chars;
-		    sz = erts_utf8_to_latin1(ep, a->name, a->len);
-		    ASSERT(a->latin1_chars == sz);
-		}
+                sz = a->len;
+                ep -= sz;
+                sys_memcpy((void *) ep, (void *) a->name, sz);
 		if (long_atoms) {
 		    ep -= 2;
 		    put_int16(sz, ep);
@@ -504,12 +542,16 @@ byte *erts_encode_ext_dist_header_finalize(byte *ext, ErtsAtomCache *cache, Uint
 		break;
 	    }
 	}
+        reds -= 3; /*was ERTS_PORT_REDS_DIST_CMD_FINALIZE*/
     }
     --ep;
     put_int8(ci, ep);
     *--ep = DIST_HEADER;
     *--ep = VERSION_MAGIC;
-    return ep;
+done:
+    ob->extp = ep;
+    ASSERT(&ob->data[0] <= ob->extp && ob->extp < ob->ext_endp);
+    return reds < 0 ? 0 : reds;
 }
 
 int erts_encode_dist_ext_size(Eterm term, Uint32 flags, ErtsAtomCacheMap *acmp,
@@ -520,7 +562,7 @@ int erts_encode_dist_ext_size(Eterm term, Uint32 flags, ErtsAtomCacheMap *acmp,
 	return -1;
     } else {
 #ifndef ERTS_DEBUG_USE_DIST_SEP
-	if (!(flags & DFLAG_DIST_HDR_ATOM_CACHE))
+	if (!(flags & (DFLAG_DIST_HDR_ATOM_CACHE | DFLAG_NO_MAGIC)))
 #endif
 	    sz++ /* VERSION_MAGIC */;
 
@@ -536,7 +578,7 @@ int erts_encode_dist_ext_size_int(Eterm term, struct erts_dsig_send_context* ctx
 	return -1;
     } else {
 #ifndef ERTS_DEBUG_USE_DIST_SEP
-	if (!(ctx->flags & DFLAG_DIST_HDR_ATOM_CACHE))
+	if (!(ctx->flags & (DFLAG_DIST_HDR_ATOM_CACHE | DFLAG_NO_MAGIC)))
 #endif
 	    sz++ /* VERSION_MAGIC */;
 
@@ -568,7 +610,7 @@ int erts_encode_dist_ext(Eterm term, byte **ext, Uint32 flags, ErtsAtomCacheMap
 {
     if (!ctx || !ctx->wstack.wstart) {
     #ifndef ERTS_DEBUG_USE_DIST_SEP
-	if (!(flags & DFLAG_DIST_HDR_ATOM_CACHE))
+	if (!(flags & (DFLAG_DIST_HDR_ATOM_CACHE | DFLAG_NO_MAGIC)))
     #endif
 	    *(*ext)++ = VERSION_MAGIC;
     }
@@ -616,7 +658,7 @@ erts_make_dist_ext_copy(ErtsDistExternal *edep, Uint xsize)
     sys_memcpy((void *) ep, (void *) edep, dist_ext_sz);
     ep += dist_ext_sz;
     if (new_edep->dep)
-	erts_smp_refc_inc(&new_edep->dep->refc, 1);
+        erts_ref_dist_entry(new_edep->dep);
     new_edep->extp = ep;
     new_edep->ext_endp = ep + ext_sz;
     new_edep->heap_size = -1;
@@ -629,7 +671,8 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 		      byte *ext,
 		      Uint size,
 		      DistEntry *dep,
-		      ErtsAtomCache *cache)
+		      ErtsAtomCache *cache,
+                      Uint32 *connection_id)
 {
 #undef ERTS_EXT_FAIL
 #undef ERTS_EXT_HDR_FAIL
@@ -642,7 +685,7 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 #endif
 
     register byte *ep = ext;
-    int utf8_atoms = (int) (dep->flags & DFLAG_UTF8_ATOMS);
+    ASSERT(dep->flags & DFLAG_UTF8_ATOMS);
 
     edep->heap_size = -1;
     edep->ext_endp = ext+size;
@@ -650,32 +693,34 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
     if (size < 2)
 	ERTS_EXT_FAIL;
 
+    if (!dep)
+        ERTS_INTERNAL_ERROR("Invalid use");
+
     if (ep[0] != VERSION_MAGIC) {
 	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-	if (dep)
-	    erts_dsprintf(dsbufp,
-			  "** Got message from incompatible erlang on "
-			  "channel %d\n",
-			  dist_entry_channel_no(dep));
-	else
-	    erts_dsprintf(dsbufp,
-			  "** Attempt to convert old incompatible "
-			  "binary %d\n",
-			  *ep);
+        erts_dsprintf(dsbufp,
+                      "** Got message from incompatible erlang on "
+                      "channel %d\n",
+                      dist_entry_channel_no(dep));
 	erts_send_error_to_logger_nogl(dsbufp);
 	ERTS_EXT_FAIL;
     }
 
     edep->flags = 0;
     edep->dep = dep;
-    if (dep) {
-	erts_smp_de_rlock(dep);
-	if (dep->flags & DFLAG_DIST_HDR_ATOM_CACHE)
-	    edep->flags |= ERTS_DIST_EXT_DFLAG_HDR;
-	    
-	edep->flags |= (dep->connection_id & ERTS_DIST_EXT_CON_ID_MASK);
-	erts_smp_de_runlock(dep);
+
+    erts_de_rlock(dep);
+
+    if (dep->state != ERTS_DE_STATE_CONNECTED &&
+	dep->state != ERTS_DE_STATE_PENDING) {
+        erts_de_runlock(dep);
+        return ERTS_PREP_DIST_EXT_CLOSED;
     }
+    if (dep->flags & DFLAG_DIST_HDR_ATOM_CACHE)
+        edep->flags |= ERTS_DIST_EXT_DFLAG_HDR;
+
+    *connection_id = dep->connection_id;
+    edep->connection_id = dep->connection_id;
 
     if (ep[1] != DIST_HEADER) {
 	if (edep->flags & ERTS_DIST_EXT_DFLAG_HDR)
@@ -802,9 +847,7 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 		    CHKSIZE(len);
 		    atom = erts_atom_put((byte *) ep,
 					 len,
-					 (utf8_atoms
-					  ? ERTS_ATOM_ENC_UTF8
-					  : ERTS_ATOM_ENC_LATIN1),
+                                         ERTS_ATOM_ENC_UTF8,
 					 0);
 		    if (is_non_value(atom))
 			ERTS_EXT_HDR_FAIL;
@@ -835,14 +878,15 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 	ERTS_EXT_FAIL;
 #endif
 
-    return 0;
+    erts_de_runlock(dep);
+
+    return ERTS_PREP_DIST_EXT_SUCCESS;
 
 #undef CHKSIZE
 #undef ERTS_EXT_FAIL
 #undef ERTS_EXT_HDR_FAIL
 
- bad_hdr:
-    if (dep) {
+ bad_hdr: {
 	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
 	erts_dsprintf(dsbufp,
 		      "%T got a corrupted distribution header from %T "
@@ -855,10 +899,11 @@ erts_prepare_dist_ext(ErtsDistExternal *edep,
 	erts_dsprintf(dsbufp, ">>");
 	erts_send_warning_to_logger_nogl(dsbufp);
     }
- fail:
-    if (dep)
-	erts_kill_dist_connection(dep, dep->connection_id);
-    return -1;
+ fail: {
+	erts_de_runlock(dep);
+	erts_kill_dist_connection(dep, *connection_id);
+    }
+    return ERTS_PREP_DIST_EXT_FAILED;
 }
 
 static void
@@ -889,7 +934,7 @@ bad_dist_ext(ErtsDistExternal *edep)
 		erts_dsprintf(dsbufp, ", %d=%T", i, edep->attab.atom[i]);
 	}
 	erts_send_warning_to_logger_nogl(dsbufp);
-	erts_kill_dist_connection(dep, ERTS_DIST_EXT_CON_ID(edep));
+	erts_kill_dist_connection(dep, edep->connection_id);
     }
 }
 
@@ -1214,7 +1259,8 @@ typedef struct B2TContext_t {
     ErtsBinary2TermState b2ts;
     Uint32 flags;
     SWord reds;
-    Eterm trap_bin;
+    Uint used_bytes; /* In: boolean, Out: bytes */
+    Eterm trap_bin;  /* THE_NON_VALUE if not exported */
     Export *bif;
     Eterm arg[2];
     enum B2TState state;
@@ -1308,6 +1354,11 @@ binary2term_prepare(ErtsBinary2TermState *state, byte *data, Sint data_size,
 
 	    ctx->u.uc.dbytes = state->extp;
 	    ctx->u.uc.dleft = dest_len;
+            if (ctx->used_bytes) {
+                ASSERT(ctx->used_bytes == 1);
+                 /* to be subtracted by stream.avail_in when done */
+                ctx->used_bytes = data_size;
+            }
 	    ctx->state = B2TUncompressChunk;
             *ctxp = ctx;
         }
@@ -1410,13 +1461,15 @@ static int b2t_context_destructor(Binary *context_bin)
     return 1;
 }
 
+static BIF_RETTYPE binary_to_term_int(Process*, Eterm bin, B2TContext*);
+
+
 static BIF_RETTYPE binary_to_term_trap_1(BIF_ALIST_1)
 {
     Binary *context_bin = erts_magic_ref2bin(BIF_ARG_1);
     ASSERT(ERTS_MAGIC_BIN_DESTRUCTOR(context_bin) == b2t_context_destructor);
 
-    return binary_to_term_int(BIF_P, 0, THE_NON_VALUE, context_bin, NULL,
-			      THE_NON_VALUE, THE_NON_VALUE);
+    return binary_to_term_int(BIF_P, THE_NON_VALUE, ERTS_MAGIC_BIN_DATA(context_bin));
 }
 
 
@@ -1442,6 +1495,8 @@ static B2TContext* b2t_export_context(Process* p, B2TContext* src)
                                                  b2t_context_destructor);
     B2TContext* ctx = ERTS_MAGIC_BIN_DATA(context_b);
     Eterm* hp;
+
+    ASSERT(is_non_value(src->trap_bin));
     sys_memcpy(ctx, src, sizeof(B2TContext));
     if (ctx->state >= B2TDecode && ctx->u.dc.next == &src->u.dc.res) {
         ctx->u.dc.next = &ctx->u.dc.res;
@@ -1451,8 +1506,7 @@ static B2TContext* b2t_export_context(Process* p, B2TContext* src)
     return ctx;
 }
 
-static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binary* context_b,
-				      Export *bif_init, Eterm arg0, Eterm arg1)
+static BIF_RETTYPE binary_to_term_int(Process* p, Eterm bin, B2TContext *ctx)
 {
     BIF_RETTYPE ret_val;
 #ifdef EXTREME_B2T_TRAPPING
@@ -1460,25 +1514,17 @@ static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binar
 #else
     SWord initial_reds = (Uint)(ERTS_BIF_REDS_LEFT(p) * B2T_BYTES_PER_REDUCTION);
 #endif
-    B2TContext c_buff;
-    B2TContext *ctx;
     int is_first_call;
 
-    if (context_b == NULL) {
+    if (is_value(bin)) {
 	/* Setup enough to get started */
         is_first_call = 1;
-        ctx = &c_buff;
 	ctx->state = B2TPrepare;
         ctx->aligned_alloc = NULL;
-        ctx->flags = flags;
-	ctx->bif = bif_init;
-	ctx->arg[0] = arg0;
-	ctx->arg[1] = arg1;
-        IF_DEBUG(ctx->trap_bin = THE_NON_VALUE;)
     } else {
-        is_first_call = 0;
-	ctx = ERTS_MAGIC_BIN_DATA(context_b);
+        ASSERT(is_value(ctx->trap_bin));
         ASSERT(ctx->state != B2TPrepare);
+        is_first_call = 0;
     }
     ctx->reds = initial_reds;
 
@@ -1522,6 +1568,10 @@ static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binar
                      && zret == Z_STREAM_END
                      && ctx->u.uc.dleft == 0) {
                 ctx->reds -= chunk;
+                if (ctx->used_bytes) {
+                    ASSERT(ctx->used_bytes > 5 + ctx->u.uc.stream.avail_in);
+                    ctx->used_bytes -= ctx->u.uc.stream.avail_in;
+                }
                 ctx->state = B2TSizeInit;
             }
             else {
@@ -1540,11 +1590,11 @@ static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binar
             break;
 
         case B2TDecodeInit:
-            if (ctx == &c_buff && ctx->b2ts.extsize > ctx->reds) {
+            if (is_non_value(ctx->trap_bin) && ctx->b2ts.extsize > ctx->reds) {
                 /* dec_term will maybe trap, allocate space for magic bin
                    before result term to make it easy to trim with HRelease.
                  */
-                ctx = b2t_export_context(p, &c_buff);
+                ctx = b2t_export_context(p, ctx);
             }
             ctx->u.dc.ep = ctx->b2ts.extp;
             ctx->u.dc.res = (Eterm) (UWord) NULL;
@@ -1587,6 +1637,25 @@ static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binar
 	    return ret_val;
 
         case B2TDone:
+            if (ctx->used_bytes) {
+                Eterm *hp;
+                Eterm used;
+                if (!ctx->b2ts.exttmp) {
+                    ASSERT(ctx->used_bytes == 1);
+                    ctx->used_bytes = (ctx->u.dc.ep - ctx->b2ts.extp
+                                       +1); /* VERSION_MAGIC */
+                }
+                if (IS_USMALL(0, ctx->used_bytes)) {
+                    hp = erts_produce_heap(&ctx->u.dc.factory, 3, 0);
+                    used = make_small(ctx->used_bytes);
+                }
+                else {
+                    hp = erts_produce_heap(&ctx->u.dc.factory, 3+BIG_UINT_HEAP_SIZE, 0);
+                    used = uint_to_big(ctx->used_bytes, hp);
+                    hp += BIG_UINT_HEAP_SIZE;
+                }
+                ctx->u.dc.res = TUPLE2(hp, ctx->u.dc.res, used);
+            }
             b2t_destroy_context(ctx);
 
             if (ctx->u.dc.factory.hp > ctx->u.dc.factory.hp_end) {
@@ -1607,11 +1676,10 @@ static BIF_RETTYPE binary_to_term_int(Process* p, Uint32 flags, Eterm bin, Binar
         }
     }while (ctx->reds > 0 || ctx->state >= B2TDone);
 
-    if (ctx == &c_buff) {
-        ASSERT(ctx->trap_bin == THE_NON_VALUE);
-        ctx = b2t_export_context(p, &c_buff);
+    if (is_non_value(ctx->trap_bin)) {
+        ctx = b2t_export_context(p, ctx);
+        ASSERT(is_value(ctx->trap_bin));
     }
-    ASSERT(ctx->trap_bin != THE_NON_VALUE);
 
     if (is_first_call) {
         erts_set_gc_state(p, 0);
@@ -1628,23 +1696,35 @@ HIPE_WRAPPER_BIF_DISABLE_GC(binary_to_term, 1)
 
 BIF_RETTYPE binary_to_term_1(BIF_ALIST_1)
 {
-    return binary_to_term_int(BIF_P, 0, BIF_ARG_1, NULL, bif_export[BIF_binary_to_term_1],
-			      BIF_ARG_1, THE_NON_VALUE);
+    B2TContext ctx;
+
+    ctx.flags = 0;
+    ctx.used_bytes = 0;
+    ctx.trap_bin = THE_NON_VALUE;
+    ctx.bif = bif_export[BIF_binary_to_term_1];
+    ctx.arg[0] = BIF_ARG_1;
+    ctx.arg[1] = THE_NON_VALUE;
+    return binary_to_term_int(BIF_P, BIF_ARG_1, &ctx);
 }
 
 HIPE_WRAPPER_BIF_DISABLE_GC(binary_to_term, 2)
 
 BIF_RETTYPE binary_to_term_2(BIF_ALIST_2)
 {
+    B2TContext ctx;
     Eterm opts;
     Eterm opt;
-    Uint32 flags = 0;
 
+    ctx.flags = 0;
+    ctx.used_bytes = 0;
     opts = BIF_ARG_2;
     while (is_list(opts)) {
         opt = CAR(list_val(opts));
         if (opt == am_safe) {
-            flags |= ERTS_DIST_EXT_BTT_SAFE;
+            ctx.flags |= ERTS_DIST_EXT_BTT_SAFE;
+        }
+        else if (opt == am_used) {
+            ctx.used_bytes = 1;
         }
 	else {
             goto error;
@@ -1655,8 +1735,11 @@ BIF_RETTYPE binary_to_term_2(BIF_ALIST_2)
     if (is_not_nil(opts))
         goto error;
 
-    return binary_to_term_int(BIF_P, flags, BIF_ARG_1, NULL, bif_export[BIF_binary_to_term_2],
-			      BIF_ARG_1, BIF_ARG_2);
+    ctx.trap_bin = THE_NON_VALUE;
+    ctx.bif = bif_export[BIF_binary_to_term_2];
+    ctx.arg[0] = BIF_ARG_1;
+    ctx.arg[1] = BIF_ARG_2;
+    return binary_to_term_int(BIF_P, BIF_ARG_1, &ctx);
 
 error:
     BIF_ERROR(BIF_P, BADARG);
@@ -1870,7 +1953,7 @@ static Eterm erts_term_to_binary_int(Process* p, Eterm Term, int level, Uint fla
 	    context_b = erts_create_magic_binary(sizeof(TTBContext),    \
                                                  ttb_context_destructor);   \
 	    context =  ERTS_MAGIC_BIN_DATA(context_b);			\
-	    memcpy(context,&c_buff,sizeof(TTBContext));			\
+	    sys_memcpy(context,&c_buff,sizeof(TTBContext));			\
 	}								\
     } while (0)
 
@@ -1949,23 +2032,14 @@ static Eterm erts_term_to_binary_int(Process* p, Eterm Term, int level, Uint fla
 		level = context->s.ec.level;
 		BUMP_REDS(p, (initial_reds - reds) / TERM_TO_BINARY_LOOP_FACTOR);
 		if (level == 0 || real_size < 6) { /* We are done */
-		    ProcBin* pb;
 		return_normal:
 		    context->s.ec.result_bin = NULL;
 		    context->alive = 0;
-		    pb = (ProcBin *) HAlloc(p, PROC_BIN_SIZE);
-		    pb->thing_word = HEADER_PROC_BIN;
-		    pb->size = real_size;
-		    pb->next = MSO(p).first;
-		    MSO(p).first = (struct erl_off_heap_header*)pb;
-		    pb->val = result_bin;
-		    pb->bytes = (byte*) result_bin->orig_bytes;
-		    pb->flags = 0;
-		    OH_OVERHEAD(&(MSO(p)), pb->size / sizeof(Eterm));
 		    if (context_b && erts_refc_read(&context_b->intern.refc,0) == 0) {
 			erts_bin_free(context_b);
 		    }
-		    return make_binary(pb);
+		    return erts_build_proc_bin(&MSO(p), HAlloc(p, PROC_BIN_SIZE),
+                                               result_bin);
 		}
 		/* Continue with compression... */
 		/* To make absolutely sure that zlib does not barf on a reallocated context, 
@@ -2022,16 +2096,7 @@ static Eterm erts_term_to_binary_int(Process* p, Eterm Term, int level, Uint fla
 			result_bin = erts_bin_realloc(context->s.cc.destination_bin,
 						      context->s.cc.dest_len+6);
 			context->s.cc.destination_bin = NULL;
-			pb = (ProcBin *) HAlloc(p, PROC_BIN_SIZE);
-			pb->thing_word = HEADER_PROC_BIN;
-			pb->size = context->s.cc.dest_len+6;
-			pb->next = MSO(p).first;
-			MSO(p).first = (struct erl_off_heap_header*)pb;
-			pb->val = result_bin;
 			ASSERT(erts_refc_read(&result_bin->intern.refc, 1));
-			pb->bytes = (byte*) result_bin->orig_bytes;
-			pb->flags = 0;
-			OH_OVERHEAD(&(MSO(p)), pb->size / sizeof(Eterm));
 			erts_bin_free(context->s.cc.result_bin);
 			context->s.cc.result_bin = NULL;
 			context->alive = 0;
@@ -2039,7 +2104,9 @@ static Eterm erts_term_to_binary_int(Process* p, Eterm Term, int level, Uint fla
 			if (context_b && erts_refc_read(&context_b->intern.refc,0) == 0) {
 			    erts_bin_free(context_b);
 			}
-			return make_binary(pb);
+			return erts_build_proc_bin(&MSO(p),
+						   HAlloc(p, PROC_BIN_SIZE),
+                                                   result_bin);
 		    }
 		default: /* Compression error, revert to uncompressed binary (still in 
 			    context) */
@@ -2093,7 +2160,7 @@ enc_atom(ErtsAtomCacheMap *acmp, Eterm atom, byte *ep, Uint32 dflags)
 {
     int iix;
     int len;
-    int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
+    const int utf8_atoms = (int) (dflags & DFLAG_UTF8_ATOMS);
 
     ASSERT(is_atom(atom));
 
@@ -2488,8 +2555,6 @@ enc_term_int(TTBEncodeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj, byte* ep,
 	    break;
 	}
 
-    L_jump_start:
-
 	if (ctx && --r <= 0) {
 	    *reds = 0;
 	    ctx->obj = obj;
@@ -2497,6 +2562,8 @@ enc_term_int(TTBEncodeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj, byte* ep,
 	    WSTACK_SAVE(s, &ctx->wstack);
 	    return -1;
 	}
+
+    L_jump_start:
 	switch(tag_val_def(obj)) {
 	case NIL_DEF:
 	    *ep++ = NIL_EXT;
@@ -2806,6 +2873,8 @@ enc_term_int(TTBEncodeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj, byte* ep,
 		    ep[j] = 0;	/* Zero unused bits at end of binary */
 		    data_dst = ep;
 		    ep += j + 1;
+                    if (ctx)
+                        ctx->hopefull_flags |= DFLAG_BIT_BINARIES;
 		} else {
 		    /*
 		     * Bit-level binary, but the receiver doesn't support it.
@@ -2841,6 +2910,8 @@ enc_term_int(TTBEncodeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj, byte* ep,
 		    ep = enc_atom(acmp, exp->info.mfa.function, ep, dflags);
 		    ep = enc_term(acmp, make_small(exp->info.mfa.arity),
                                   ep, dflags, off_heap);
+                    if (ctx)
+                        ctx->hopefull_flags |= DFLAG_EXPORT_PTR_TAG;
 		} else {
 		    /* Tag, arity */
 		    *ep++ = SMALL_TUPLE_EXT;
@@ -2861,62 +2932,27 @@ enc_term_int(TTBEncodeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj, byte* ep,
 		ErlFunThing* funp = (ErlFunThing *) fun_val(obj);
 		int ei;
 
-		if ((dflags & DFLAG_NEW_FUN_TAGS) != 0) {
-		    *ep++ = NEW_FUN_EXT;
-		    WSTACK_PUSH2(s, ENC_PATCH_FUN_SIZE,
-				 (UWord) ep); /* Position for patching in size */
-		    ep += 4;
-		    *ep = funp->arity;
-		    ep += 1;
-		    sys_memcpy(ep, funp->fe->uniq, 16);
-		    ep += 16;
-		    put_int32(funp->fe->index, ep);
-		    ep += 4;
-		    put_int32(funp->num_free, ep);
-		    ep += 4;
-		    ep = enc_atom(acmp, funp->fe->module, ep, dflags);
-		    ep = enc_term(acmp, make_small(funp->fe->old_index), ep, dflags, off_heap);
-		    ep = enc_term(acmp, make_small(funp->fe->old_uniq), ep, dflags, off_heap);
-		    ep = enc_pid(acmp, funp->creator, ep, dflags);
-		} else {
-		    /*
-		     * Communicating with an obsolete erl_interface or
-		     * jinterface node. Convert the fun to a tuple to
-		     * avoid crasching.
-		     */
-		
-		    /* Tag, arity */
-		    *ep++ = SMALL_TUPLE_EXT;
-		    put_int8(5, ep);
-		    ep += 1;
-		
-		    /* 'fun' */
-		    ep = enc_atom(acmp, am_fun, ep, dflags);
-		
-		    /* Module name */
-		    ep = enc_atom(acmp, funp->fe->module, ep, dflags);
-		
-		    /* Index, Uniq */
-		    *ep++ = INTEGER_EXT;
-		    put_int32(funp->fe->old_index, ep);
-		    ep += 4;
-		    *ep++ = INTEGER_EXT;
-		    put_int32(funp->fe->old_uniq, ep);
-		    ep += 4;
-		
-		    /* Environment sub-tuple arity */
-		    ASSERT(funp->num_free < MAX_ARG);
-		    *ep++ = SMALL_TUPLE_EXT;
-		    put_int8(funp->num_free, ep);
-		    ep += 1;
-		}
-		for (ei = funp->num_free-1; ei > 0; ei--) {
+		ASSERT(dflags & DFLAG_NEW_FUN_TAGS);
+                *ep++ = NEW_FUN_EXT;
+                WSTACK_PUSH2(s, ENC_PATCH_FUN_SIZE,
+                             (UWord) ep); /* Position for patching in size */
+                ep += 4;
+                *ep = funp->arity;
+                ep += 1;
+                sys_memcpy(ep, funp->fe->uniq, 16);
+                ep += 16;
+                put_int32(funp->fe->index, ep);
+                ep += 4;
+                put_int32(funp->num_free, ep);
+                ep += 4;
+                ep = enc_atom(acmp, funp->fe->module, ep, dflags);
+                ep = enc_term(acmp, make_small(funp->fe->old_index), ep, dflags, off_heap);
+                ep = enc_term(acmp, make_small(funp->fe->old_uniq), ep, dflags, off_heap);
+                ep = enc_pid(acmp, funp->creator, ep, dflags);
+
+		for (ei = funp->num_free-1; ei >= 0; ei--) {
 		    WSTACK_PUSH2(s, ENC_TERM, (UWord) funp->env[ei]);
 		}
-		if (funp->num_free != 0) {
-		    obj = funp->env[0];
-		    goto L_jump_start;
-		}
 	    }
 	    break;
 	}
@@ -3109,7 +3145,7 @@ dec_term(ErtsDistExternal *edep,
 #if defined(ARCH_64)
 		*objp = make_small(sn);
 #else
-		if (MY_IS_SSMALL(sn)) {
+		if (IS_SSMALL(sn)) {
 		    *objp = make_small(sn);
 		} else {
 		    *objp = small_to_big(sn, hp);
@@ -3254,6 +3290,7 @@ dec_term_atom_common:
             n--;
 	    if (ctx) {
                 if (reds < n) {
+                    ASSERT(reds > 0);
 		    ctx->state = B2TDecodeList;
 		    ctx->u.dc.remaining_n = n - reds;
 		    n = reds;
@@ -3535,18 +3572,9 @@ dec_term_atom_common:
 		    *objp = make_binary(hb);
 		} else {
 		    Binary* dbin = erts_bin_nrml_alloc(n);
-		    ProcBin* pb;
-		    pb = (ProcBin *) hp;
+
+		    *objp = erts_build_proc_bin(factory->off_heap, hp, dbin);
 		    hp += PROC_BIN_SIZE;
-		    pb->thing_word = HEADER_PROC_BIN;
-		    pb->size = n;
-		    pb->next = factory->off_heap->first;
-		    factory->off_heap->first = (struct erl_off_heap_header*)pb;
-		    OH_OVERHEAD(factory->off_heap, pb->size / sizeof(Eterm));
-		    pb->val = dbin;
-		    pb->bytes = (byte*) dbin->orig_bytes;
-		    pb->flags = 0;
-		    *objp = make_binary(pb);
                     if (ctx) {
                         int n_limit = reds * B2T_MEMCPY_FACTOR;
                         if (n > n_limit) {
@@ -3586,18 +3614,9 @@ dec_term_atom_common:
                     ep += n;
 		} else {
 		    Binary* dbin = erts_bin_nrml_alloc(n);
-		    ProcBin* pb;
+		    Uint n_copy = n;
 
-		    pb = (ProcBin *) hp;
-		    pb->thing_word = HEADER_PROC_BIN;
-		    pb->size = n;
-		    pb->next = factory->off_heap->first;
-		    factory->off_heap->first = (struct erl_off_heap_header*)pb;
-		    OH_OVERHEAD(factory->off_heap, pb->size / sizeof(Eterm));
-		    pb->val = dbin;
-		    pb->bytes = (byte*) dbin->orig_bytes;
-		    pb->flags = 0;
-		    bin = make_binary(pb);
+		    bin = erts_build_proc_bin(factory->off_heap, hp, dbin);
 		    hp += PROC_BIN_SIZE;
                     if (ctx) {
                         int n_limit = reds * B2T_MEMCPY_FACTOR;
@@ -3605,15 +3624,15 @@ dec_term_atom_common:
                             ctx->state = B2TDecodeBinary;
                             ctx->u.dc.remaining_n = n - n_limit;
                             ctx->u.dc.remaining_bytes = dbin->orig_bytes + n_limit;
-                            n = n_limit;
+                            n_copy = n_limit;
                             reds = 0;
                         }
-                        else
+                        else {
                             reds -= n / B2T_MEMCPY_FACTOR;
+			}
                     }
-                    sys_memcpy(dbin->orig_bytes, ep, n);
-                    ep += n;
-                    n = pb->size;
+                    sys_memcpy(dbin->orig_bytes, ep, n_copy);
+                    ep += n_copy;
                 }
 
 		if (bitsize == 8 || n == 0) {
@@ -3994,6 +4013,7 @@ dec_term_atom_common:
     if (ctx) {
         ctx->state = B2TDone;
 	ctx->reds = reds;
+        ctx->u.dc.ep = ep;
     }
 
     return ep;
@@ -4263,24 +4283,12 @@ encode_size_struct_int(TTBSizeContext* ctx, ErtsAtomCacheMap *acmp, Eterm obj,
 	    {
 		ErlFunThing* funp = (ErlFunThing *) fun_val(obj);
 		
-		if ((dflags & DFLAG_NEW_FUN_TAGS) != 0) {
-		    result += 20+1+1+4;	/* New ID + Tag */
-		    result += 4; /* Length field (number of free variables */
-		    result += encode_size_struct2(acmp, funp->creator, dflags);
-		    result += encode_size_struct2(acmp, funp->fe->module, dflags);
-		    result += 2 * (1+4);	/* Index, Uniq */
-		} else {
-		    /*
-		     * Size when fun is mapped to a tuple.
-		     */
-		    result += 1 + 1; /* Tuple tag, arity */
-		    result += 1 + 1 + 2 +
-			atom_tab(atom_val(am_fun))->len; /* 'fun' */
-		    result += 1 + 1 + 2 +
-			atom_tab(atom_val(funp->fe->module))->len; /* Module name */
-		    result += 2 * (1 + 4); /* Index + Uniq */
-		    result += 1 + (funp->num_free < 0x100 ? 1 : 4);
-		}
+                ASSERT(dflags & DFLAG_NEW_FUN_TAGS);
+                result += 20+1+1+4;	/* New ID + Tag */
+                result += 4; /* Length field (number of free variables */
+                result += encode_size_struct2(acmp, funp->creator, dflags);
+                result += encode_size_struct2(acmp, funp->fe->module, dflags);
+                result += 2 * (1+4);	/* Index, Uniq */
 		if (funp->num_free > 1) {
 		    WSTACK_PUSH2(s, (UWord) (funp->env + 1),
 				    (UWord) TERM_ARRAY_OP(funp->num_free-1));
@@ -4368,7 +4376,7 @@ decoded_size(byte *ep, byte* endp, int internal_tags, B2TContext* ctx)
         }
     }
     else
-        reds = 0; /* not used but compiler warns anyway */
+        ERTS_UNDEF(reds, 0);
 
     heap_size = 0;
     terms = 1;
@@ -4678,3 +4686,182 @@ error:
 #undef SKIP2
 #undef CHKSIZE
 }
+
+
+struct transcode_context {
+    enum {
+        TRANSCODE_DEC_MSG_SIZE,
+        TRANSCODE_DEC_MSG,
+        TRANSCODE_ENC_CTL,
+        TRANSCODE_ENC_MSG
+    }state;
+    Eterm ctl_term;
+    Eterm* ctl_heap;
+    ErtsHeapFactory ctl_factory;
+    Eterm* msg_heap;
+    B2TContext b2t;
+    TTBEncodeContext ttb;
+#ifdef DEBUG
+    ErtsDistOutputBuf* dbg_ob;
+#endif
+};
+
+void transcode_free_ctx(DistEntry* dep)
+{
+    struct transcode_context* ctx = dep->transcode_ctx;
+
+    erts_factory_close(&ctx->ctl_factory);
+    erts_free(ERTS_ALC_T_DIST_TRANSCODE, ctx->ctl_heap);
+
+    if (ctx->msg_heap) {
+        erts_factory_close(&ctx->b2t.u.dc.factory);
+        erts_free(ERTS_ALC_T_DIST_TRANSCODE, ctx->msg_heap);
+    }
+    erts_free(ERTS_ALC_T_DIST_TRANSCODE, ctx);
+    dep->transcode_ctx = NULL;
+}
+
+Sint transcode_dist_obuf(ErtsDistOutputBuf* ob,
+                         DistEntry* dep,
+                         Uint32 dflags,
+                         Sint reds)
+{
+    Sint hsz;
+    byte* decp;
+    const int have_msg = !!ob->msg_start;
+    int i;
+    struct transcode_context* ctx = dep->transcode_ctx;
+
+    if (!ctx) { /* first call for 'ob' */
+	ASSERT(!(ob->hopefull_flags & ~(Uint)(DFLAG_BIT_BINARIES |
+					      DFLAG_EXPORT_PTR_TAG)));
+        if (~dflags & ob->hopefull_flags) {
+            /*
+             * Receiver does not support bitstrings and/or export funs
+             * and output buffer contains such message tags (hopefull_flags).
+             * Must transcode control and message terms to use tuple fallbacks.
+             */
+            ctx = erts_alloc(ERTS_ALC_T_DIST_TRANSCODE, sizeof(struct transcode_context));
+            dep->transcode_ctx = ctx;
+        #ifdef DEBUG
+            ctx->dbg_ob = ob;
+        #endif
+
+            hsz = decoded_size(ob->extp, ob->ext_endp, 0, NULL);
+            ctx->ctl_heap = erts_alloc(ERTS_ALC_T_DIST_TRANSCODE, hsz*sizeof(Eterm));
+            erts_factory_tmp_init(&ctx->ctl_factory, ctx->ctl_heap, hsz, ERTS_ALC_T_DIST_TRANSCODE);
+            ctx->msg_heap = NULL;
+
+            decp = dec_term(NULL, &ctx->ctl_factory, ob->extp, &ctx->ctl_term, NULL);
+            if (have_msg) {
+                ASSERT(decp == ob->msg_start); (void)decp;
+                ctx->b2t.u.sc.ep = NULL;
+                ctx->b2t.state = B2TSize;
+                ctx->b2t.aligned_alloc = NULL;
+                ctx->b2t.b2ts.exttmp = 0;
+                ctx->state = TRANSCODE_DEC_MSG_SIZE;
+            }
+            else {
+                ASSERT(decp == ob->ext_endp);
+                ctx->state = TRANSCODE_ENC_CTL;
+            }
+        }
+        else if (!(dflags & DFLAG_DIST_HDR_ATOM_CACHE)) {
+            /*
+             * No need for full transcoding, but primitive receiver (erl_/jinterface)
+             * expects VERSION_MAGIC before both control and message terms.
+             */
+            if (ob->msg_start) {
+                Sint ctl_bytes = ob->msg_start - ob->extp;
+                ASSERT(ob->extp < ob->msg_start && ob->msg_start < ob->ext_endp);
+                /* Move control term back 1 byte to make room */
+                sys_memmove(ob->extp-1, ob->extp, ctl_bytes);
+                *--(ob->msg_start) = VERSION_MAGIC;
+                --(ob->extp);
+                reds -= ctl_bytes / (B2T_BYTES_PER_REDUCTION * B2T_MEMCPY_FACTOR);
+            }
+            *--(ob->extp) = VERSION_MAGIC;
+            goto done;
+        }
+        else
+            goto done;
+    }
+    else {  /* continue after yield */
+        ASSERT(ctx->dbg_ob == ob);
+    }
+    ctx->b2t.reds = reds * B2T_BYTES_PER_REDUCTION;
+
+    switch (ctx->state) {
+    case TRANSCODE_DEC_MSG_SIZE:
+        hsz = decoded_size(ob->msg_start, ob->ext_endp, 0, &ctx->b2t);
+        if (ctx->b2t.state == B2TSize) {
+            return -1;
+        }
+        ASSERT(ctx->b2t.state == B2TDecodeInit);
+        ctx->msg_heap = erts_alloc(ERTS_ALC_T_DIST_TRANSCODE, hsz*sizeof(Eterm));
+        ctx->b2t.u.dc.ep = ob->msg_start;
+        ctx->b2t.u.dc.res = (Eterm) NULL;
+        ctx->b2t.u.dc.next = &ctx->b2t.u.dc.res;
+        erts_factory_tmp_init(&ctx->b2t.u.dc.factory,
+                              ctx->msg_heap, hsz, ERTS_ALC_T_DIST_TRANSCODE);
+        ctx->b2t.u.dc.flat_maps.wstart = NULL;
+        ctx->b2t.u.dc.hamt_array.pstart = NULL;
+        ctx->b2t.state = B2TDecode;
+
+        ctx->state = TRANSCODE_DEC_MSG;
+    case TRANSCODE_DEC_MSG:
+        if (ctx->b2t.reds <= 0)
+            ctx->b2t.reds = 1;
+        decp = dec_term(NULL, NULL, NULL, NULL, &ctx->b2t);
+        if (ctx->b2t.state < B2TDone) {
+            return -1;
+        }
+        ASSERT(ctx->b2t.state == B2TDone);
+        ASSERT(decp && decp <= ob->ext_endp);
+        reds = ctx->b2t.reds / B2T_BYTES_PER_REDUCTION;
+        b2t_destroy_context(&ctx->b2t);
+
+        ctx->state = TRANSCODE_ENC_CTL;
+    case TRANSCODE_ENC_CTL:
+        if (!(dflags & DFLAG_DIST_HDR_ATOM_CACHE)) {
+            ASSERT(!(dflags & DFLAG_NO_MAGIC));
+            ob->extp -= 2; /* VERSION_MAGIC x 2 */
+        }
+        ob->ext_endp = ob->extp;
+        i = erts_encode_dist_ext(ctx->ctl_term, &ob->ext_endp, dflags,
+                                 NULL, NULL, NULL);
+        ASSERT(i == 0); (void)i;
+        ASSERT(ob->ext_endp <= ob->alloc_endp);
+
+        if (!have_msg) {
+            break;
+        }
+        ob->msg_start = ob->ext_endp;
+        ctx->ttb.wstack.wstart = NULL;
+        ctx->ttb.flags = dflags;
+        ctx->ttb.hopefull_flags = 0;
+        ctx->ttb.level = 0;
+
+        ctx->state = TRANSCODE_ENC_MSG;
+    case TRANSCODE_ENC_MSG:
+        reds *= TERM_TO_BINARY_LOOP_FACTOR;
+        if (erts_encode_dist_ext(ctx->b2t.u.dc.res, &ob->ext_endp, dflags, NULL,
+                                 &ctx->ttb, &reds)) {
+            return -1;
+        }
+        reds /= TERM_TO_BINARY_LOOP_FACTOR;
+
+        ASSERT(ob->ext_endp <= ob->alloc_endp);
+        ASSERT(!ctx->ttb.hopefull_flags);
+    }
+    transcode_free_ctx(dep);
+
+done:
+    if (!(dflags & DFLAG_DIST_HDR_ATOM_CACHE))
+        *--(ob->extp) = PASS_THROUGH;
+
+    if (reds < 0)
+        reds = 0;
+
+    return reds;
+}
diff --git a/erts/emulator/beam/external.h b/erts/emulator/beam/external.h
index f00426cc16..bbd9b4bad2 100644
--- a/erts/emulator/beam/external.h
+++ b/erts/emulator/beam/external.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2016. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -83,7 +83,10 @@
 #ifndef ERL_EXTERNAL_H__
 #define ERL_EXTERNAL_H__
 
+#define ERL_NODE_TABLES_BASIC_ONLY
 #include "erl_node_tables.h"
+#undef ERL_NODE_TABLES_BASIC_ONLY
+#include "erl_alloc.h"
 
 #define ERTS_ATOM_CACHE_SIZE 2048
 
@@ -109,35 +112,26 @@ typedef struct {
 } ErtsAtomTranslationTable;
 
 /*
- * These flags are tagged onto the high bits of a connection ID and stored in
- * the ErtsDistExternal structure's flags field.  They are used to indicate
- * various bits of state necessary to decode binaries in a variety of
- * scenarios. The mask ERTS_DIST_EXT_CON_ID_MASK is used later to separate the
- * connection ID from the flags. Be careful to ensure that the mask does not
- * overlap any of the bits used for flags, or ERTS will leak flags bits into
- * connection IDs and leak connection ID bits into the flags.
+ * These flags are stored in the ErtsDistExternal structure's flags field.
+ * They are used to indicate various bits of state necessary to decode binaries
+ * in a variety of scenarios.
  */
-#define ERTS_DIST_EXT_DFLAG_HDR      ((Uint32) 0x80000000)
-#define ERTS_DIST_EXT_ATOM_TRANS_TAB ((Uint32) 0x40000000)
-#define ERTS_DIST_EXT_BTT_SAFE       ((Uint32) 0x20000000)
-#define ERTS_DIST_EXT_CON_ID_MASK    ((Uint32) 0x1fffffff)
+#define ERTS_DIST_EXT_DFLAG_HDR      ((Uint32) 0x1)
+#define ERTS_DIST_EXT_ATOM_TRANS_TAB ((Uint32) 0x2)
+#define ERTS_DIST_EXT_BTT_SAFE       ((Uint32) 0x4)
+
+#define ERTS_DIST_CON_ID_MASK ((Uint32) 0x00ffffff) /* also in net_kernel.erl */
 
-#define ERTS_DIST_EXT_CON_ID(DIST_EXTP) \
-  ((DIST_EXTP)->flags & ERTS_DIST_EXT_CON_ID_MASK)
 typedef struct {
     DistEntry *dep;
     byte *extp;
     byte *ext_endp;
     Sint heap_size;
+    Uint32 connection_id;
     Uint32 flags;
     ErtsAtomTranslationTable attab;
 } ErtsDistExternal;
 
-typedef struct {
-    int have_header;
-    int cache_entries;
-} ErtsDistHeaderPeek;
-
 #define ERTS_DIST_EXT_SIZE(EDEP) \
   (sizeof(ErtsDistExternal) \
    - (((EDEP)->flags & ERTS_DIST_EXT_ATOM_TRANS_TAB) \
@@ -163,7 +157,7 @@ void erts_finalize_atom_cache_map(ErtsAtomCacheMap *, Uint32);
 
 Uint erts_encode_ext_dist_header_size(ErtsAtomCacheMap *);
 byte *erts_encode_ext_dist_header_setup(byte *, ErtsAtomCacheMap *);
-byte *erts_encode_ext_dist_header_finalize(byte *, ErtsAtomCache *, Uint32);
+Sint erts_encode_ext_dist_header_finalize(ErtsDistOutputBuf*, DistEntry *, Uint32 dflags, Sint reds);
 struct erts_dsig_send_context;
 int erts_encode_dist_ext_size(Eterm, Uint32, ErtsAtomCacheMap*, Uint* szp);
 int erts_encode_dist_ext_size_int(Eterm term, struct erts_dsig_send_context* ctx, Uint* szp);
@@ -177,16 +171,18 @@ Uint erts_encode_ext_size_ets(Eterm);
 void erts_encode_ext(Eterm, byte **);
 byte* erts_encode_ext_ets(Eterm, byte *, struct erl_off_heap_header** ext_off_heap);
 
-#ifdef ERTS_WANT_EXTERNAL_TAGS
-ERTS_GLB_INLINE void erts_peek_dist_header(ErtsDistHeaderPeek *, byte *, Uint);
-#endif
 ERTS_GLB_INLINE void erts_free_dist_ext_copy(ErtsDistExternal *);
 ERTS_GLB_INLINE void *erts_dist_ext_trailer(ErtsDistExternal *);
 ErtsDistExternal *erts_make_dist_ext_copy(ErtsDistExternal *, Uint);
 void *erts_dist_ext_trailer(ErtsDistExternal *);
 void erts_destroy_dist_ext_copy(ErtsDistExternal *);
+
+#define ERTS_PREP_DIST_EXT_FAILED       (-1)
+#define ERTS_PREP_DIST_EXT_SUCCESS      (0)
+#define ERTS_PREP_DIST_EXT_CLOSED       (1)
+
 int erts_prepare_dist_ext(ErtsDistExternal *, byte *, Uint,
-			  DistEntry *, ErtsAtomCache *);
+			  DistEntry *, ErtsAtomCache *, Uint32 *);
 Sint erts_decode_dist_ext_size(ErtsDistExternal *);
 Eterm erts_decode_dist_ext(ErtsHeapFactory* factory, ErtsDistExternal *);
 
@@ -202,23 +198,9 @@ void erts_binary2term_abort(ErtsBinary2TermState *);
 Eterm erts_binary2term_create(ErtsBinary2TermState *, ErtsHeapFactory*);
 int erts_debug_max_atom_out_cache_index(void);
 int erts_debug_atom_to_out_cache_index(Eterm);
-
+void transcode_free_ctx(DistEntry* dep);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
-#ifdef ERTS_WANT_EXTERNAL_TAGS
-ERTS_GLB_INLINE void
-erts_peek_dist_header(ErtsDistHeaderPeek *dhpp, byte *ext, Uint sz)
-{
-    if (ext[0] == VERSION_MAGIC
-	|| ext[1] != DIST_HEADER
-	|| sz < (1+1+1))
-	dhpp->have_header = 0;
-    else {
-	dhpp->have_header = 1;
-	dhpp->cache_entries = (int) get_int8(&ext[2]);
-    }
-}
-#endif
 
 ERTS_GLB_INLINE void
 erts_free_dist_ext_copy(ErtsDistExternal *edep)
diff --git a/erts/emulator/beam/float_instrs.tab b/erts/emulator/beam/float_instrs.tab
new file mode 100644
index 0000000000..3d4db77892
--- /dev/null
+++ b/erts/emulator/beam/float_instrs.tab
@@ -0,0 +1,88 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+LOAD_DOUBLE(Src, Dst) {
+    GET_DOUBLE($Src, *(FloatDef *) &$Dst);
+}
+
+fload(Reg, Dst) {
+    $LOAD_DOUBLE($Reg, $Dst);
+}
+
+fstore(Float, Dst) {
+    PUT_DOUBLE(*((FloatDef *) &$Float), HTOP);
+    $Dst = make_float(HTOP);
+    HTOP += FLOAT_SIZE_OBJECT;
+}
+
+fconv(Src, Dst) {
+    Eterm src = $Src;
+
+    if (is_small(src)) {
+        $Dst = (double) signed_val(src);
+    } else if (is_big(src)) {
+        if (big_to_double(src, &$Dst) < 0) {
+            $BADARITH0();
+        }
+    } else if (is_float(src)) {
+        $LOAD_DOUBLE(src, $Dst);
+    } else {
+        $BADARITH0();
+    }
+}
+
+FLOAT_OP(Src1, OP, Src2, Dst) {
+    ERTS_NO_FPE_CHECK_INIT(c_p);
+    $Dst = $Src1 $OP $Src2;
+    ERTS_NO_FPE_ERROR(c_p, $Dst, $BADARITH0());
+}
+
+i_fadd(Src1, Src2, Dst) {
+    $FLOAT_OP($Src1, +, $Src2, $Dst);
+}
+
+i_fsub(Src1, Src2, Dst) {
+    $FLOAT_OP($Src1, -, $Src2, $Dst);
+}
+
+i_fmul(Src1, Src2, Dst) {
+    $FLOAT_OP($Src1, *, $Src2, $Dst);
+}
+
+i_fdiv(Src1, Src2, Dst) {
+    $FLOAT_OP($Src1, /, $Src2, $Dst);
+}
+
+i_fnegate(Src, Dst) {
+    ERTS_NO_FPE_CHECK_INIT(c_p);
+    $Dst = -$Src;
+    ERTS_NO_FPE_ERROR(c_p, $Dst, $BADARITH0());
+}
+
+%unless NO_FPE_SIGNALS
+fclearerror() {
+    ERTS_FP_CHECK_INIT(c_p);
+}
+
+i_fcheckerror() {
+    ERTS_FP_ERROR(c_p, freg[0].fd, $BADARITH0());
+}
+%endif
diff --git a/erts/emulator/beam/global.h b/erts/emulator/beam/global.h
index 87777d14e9..2cf268162d 100644
--- a/erts/emulator/beam/global.h
+++ b/erts/emulator/beam/global.h
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -85,11 +85,9 @@ struct enif_resource_type_t
 
 typedef struct
 {
-    erts_smp_mtx_t lock;
+    erts_mtx_t lock;
     ErtsMonitor* root;
-    int pending_failed_fire;
-    int is_dying;
-
+    Uint refc;
     size_t user_data_sz;
 } ErtsResourceMonitors;
 
@@ -116,22 +114,23 @@ extern void erts_pre_nif(struct enif_environment_t*, Process*,
 			 struct erl_module_nif*, Process* tracee);
 extern void erts_post_nif(struct enif_environment_t* env);
 extern void erts_resource_stop(ErtsResource*, ErlNifEvent, int is_direct_call);
-void erts_fire_nif_monitor(ErtsResource*, Eterm pid, Eterm ref);
+void erts_fire_nif_monitor(ErtsMonitor *tmon);
+void erts_nif_demonitored(ErtsResource* resource);
+extern void erts_add_taint(Eterm mod_atom);
 extern Eterm erts_nif_taints(Process* p);
 extern void erts_print_nif_taints(fmtfn_t to, void* to_arg);
 void erts_unload_nif(struct erl_module_nif* nif);
 extern void erl_nif_init(void);
 extern int erts_nif_get_funcs(struct erl_module_nif*,
                               struct enif_func_t **funcs);
+extern Module *erts_nif_get_module(struct erl_module_nif*);
 extern Eterm erts_nif_call_function(Process *p, Process *tracee,
                                     struct erl_module_nif*,
                                     struct enif_func_t *,
                                     int argc, Eterm *argv);
 
-#ifdef ERTS_DIRTY_SCHEDULERS
 int erts_call_dirty_nif(ErtsSchedulerData *esdp, Process *c_p,
 			BeamInstr *I, Eterm *reg);
-#endif /* ERTS_DIRTY_SCHEDULERS */
 
 
 /* Driver handle (wrapper for old plain handle) */
@@ -183,9 +182,9 @@ typedef struct {
     void         *handle;             /* Handle for DLL or SO (for dyn. drivers). */
     DE_ProcEntry *procs;              /* List of pids that have loaded this driver,
 				         or that wait for it to change state */
-    erts_smp_refc_t  refc;                /* Number of ports/processes having
+    erts_refc_t  refc;                /* Number of ports/processes having
 					 references to the driver */
-    erts_smp_atomic32_t port_count;   /* Number of ports using the driver */
+    erts_atomic32_t port_count;   /* Number of ports using the driver */
     Uint         flags;               /* ERL_DE_FL_KILL_PORTS */
     int          status;              /* ERL_DE_xxx */
     char         *full_path;          /* Full path of the driver */
@@ -202,6 +201,7 @@ typedef struct {
 struct erts_driver_t_ {
     erts_driver_t *next;
     erts_driver_t *prev;
+    Eterm name_atom;
     char *name;
     struct {
 	int major;
@@ -209,9 +209,7 @@ struct erts_driver_t_ {
     } version;
     int flags;
     DE_Handle *handle;
-#ifdef ERTS_SMP
-    erts_smp_mtx_t *lock;
-#endif
+    erts_mtx_t *lock;
     ErlDrvEntry *entry;
     ErlDrvData (*start)(ErlDrvPort port, char *command, SysDriverOpts* opts);
     void (*stop)(ErlDrvData drv_data);
@@ -226,8 +224,6 @@ struct erts_driver_t_ {
 			 char *buf, ErlDrvSizeT len,
 			 char **rbuf, ErlDrvSizeT rlen, /* Might be NULL */
 			 unsigned int *flags);
-    void (*event)(ErlDrvData drv_data, ErlDrvEvent event,
-		  ErlDrvEventData event_data);
     void (*ready_input)(ErlDrvData drv_data, ErlDrvEvent event); 
     void (*ready_output)(ErlDrvData drv_data, ErlDrvEvent event);  
     void (*timeout)(ErlDrvData drv_data);
@@ -238,7 +234,7 @@ struct erts_driver_t_ {
 };
 
 extern erts_driver_t *driver_list;
-extern erts_smp_rwmtx_t erts_driver_list_lock;
+extern erts_rwmtx_t erts_driver_list_lock;
 
 extern void erts_ddll_init(void);
 extern void erts_ddll_lock_driver(DE_Handle *dh, char *name);
@@ -299,7 +295,7 @@ extern Eterm node_cookie;
 extern Uint display_items;	/* no of items to display in traces etc */
 
 extern int erts_backtrace_depth;
-extern erts_smp_atomic32_t erts_max_gen_gcs;
+extern erts_atomic32_t erts_max_gen_gcs;
 
 extern int bif_reductions;      /* reductions + fcalls (when doing call_bif) */
 extern int stackdump_on_exit;
@@ -374,7 +370,7 @@ do {\
 	UWord _wsz = ESTACK_COUNT(s);\
 	(dst)->start = erts_alloc((s).alloc_type,\
 				  DEF_ESTACK_SIZE * sizeof(Eterm));\
-	memcpy((dst)->start, (s).start,_wsz*sizeof(Eterm));\
+	sys_memcpy((dst)->start, (s).start,_wsz*sizeof(Eterm));\
 	(dst)->sp = (dst)->start + _wsz;\
 	(dst)->end = (dst)->start + DEF_ESTACK_SIZE;\
         (dst)->edefault = NULL;\
@@ -542,7 +538,7 @@ do {\
 	UWord _wsz = WSTACK_COUNT(s);\
 	(dst)->wstart = erts_alloc(s.alloc_type,\
 				  DEF_WSTACK_SIZE * sizeof(UWord));\
-	memcpy((dst)->wstart, s.wstart,_wsz*sizeof(UWord));\
+	sys_memcpy((dst)->wstart, s.wstart,_wsz*sizeof(UWord));\
 	(dst)->wsp = (dst)->wstart + _wsz;\
 	(dst)->wend = (dst)->wstart + DEF_WSTACK_SIZE;\
         (dst)->wdefault = NULL;\
@@ -866,6 +862,8 @@ void erts_emasculate_writable_binary(ProcBin* pb);
 Eterm erts_new_heap_binary(Process *p, byte *buf, int len, byte** datap);
 Eterm erts_new_mso_binary(Process*, byte*, Uint);
 Eterm new_binary(Process*, byte*, Uint);
+Eterm erts_heap_factory_new_binary(ErtsHeapFactory *hfact, byte *buf,
+                                   Uint len, Uint reserve_size);
 Eterm erts_realloc_binary(Eterm bin, size_t size);
 Eterm erts_build_proc_bin(ErlOffHeap*, Eterm*, Binary*);
 
@@ -892,6 +890,7 @@ void erts_init_trap_export(Export* ep, Eterm m, Eterm f, Uint a,
 			   Eterm (*bif)(Process*, Eterm*, BeamInstr*));
 void erts_init_bif(void);
 Eterm erl_send(Process *p, Eterm to, Eterm msg);
+int erts_set_group_leader(Process *proc, Eterm new_gl);
 
 /* erl_bif_op.c */
 
@@ -910,13 +909,10 @@ typedef struct ErtsLiteralArea_ {
 #define ERTS_LITERAL_AREA_ALLOC_SIZE(N) \
     (sizeof(ErtsLiteralArea) + sizeof(Eterm)*((N) - 1))
 
-extern erts_smp_atomic_t erts_copy_literal_area__;
+extern erts_atomic_t erts_copy_literal_area__;
 #define ERTS_COPY_LITERAL_AREA()					\
-    ((ErtsLiteralArea *) erts_smp_atomic_read_nob(&erts_copy_literal_area__))
+    ((ErtsLiteralArea *) erts_atomic_read_nob(&erts_copy_literal_area__))
 extern Process *erts_literal_area_collector;
-#ifdef ERTS_DIRTY_SCHEDULERS
-extern Process *erts_dirty_process_code_checker;
-#endif
 
 extern Process *erts_code_purger;
 
@@ -956,8 +952,8 @@ void erts_update_ranges(BeamInstr* code, Uint size);
 void erts_remove_from_ranges(BeamInstr* code);
 UWord erts_ranges_sz(void);
 void erts_lookup_function_info(FunctionInfo* fi, BeamInstr* pc, int full_info);
-ErtsLiteralArea** erts_dump_lit_areas;
-Uint erts_dump_num_lit_areas;
+extern ErtsLiteralArea** erts_dump_lit_areas;
+extern Uint erts_dump_num_lit_areas;
 
 /* break.c */
 void init_break_handler(void);
@@ -1074,13 +1070,13 @@ Eterm copy_struct_x(Eterm, Uint, Eterm**, ErlOffHeap*, Uint*, erts_literal_area_
 #define copy_struct_litopt(Obj,Sz,HPP,OH,LitArea) \
     copy_struct_x(Obj,Sz,HPP,OH,NULL,LitArea)
 
-Eterm copy_shallow(Eterm*, Uint, Eterm**, ErlOffHeap*);
+Eterm copy_shallow(Eterm* ERTS_RESTRICT, Uint, Eterm**, ErlOffHeap*);
 
 void erts_move_multi_frags(Eterm** hpp, ErlOffHeap*, ErlHeapFragment* first,
 			   Eterm* refs, unsigned nrefs, int literals);
 
 /* Utilities */
-extern void erts_delete_nodes_monitors(Process *, ErtsProcLocks);
+void erts_monitor_nodes_delete(ErtsMonitor *);
 extern Eterm erts_monitor_nodes(Process *, Eterm, Eterm);
 extern Eterm erts_processes_monitoring_nodes(Process *);
 extern int erts_do_net_exits(DistEntry*, Eterm);
@@ -1113,7 +1109,6 @@ void erts_save_stacktrace(Process* p, struct StackTrace* s, int depth);
 typedef struct {
     Eterm delay_time;
     int context_reds;
-    int input_reds;
 } ErtsModifiedTimings;
 
 extern Export *erts_delay_trap;
@@ -1131,18 +1126,11 @@ extern ErtsModifiedTimings erts_modified_timings[];
 extern int erts_no_line_info;
 extern Eterm erts_error_logger_warnings;
 extern int erts_initialized;
-#if defined(USE_THREADS) && !defined(ERTS_SMP)
-extern erts_tid_t erts_main_thread;
-#endif
 extern int erts_compat_rel;
-extern int erts_use_sender_punish;
 void erl_start(int, char**);
 void erts_usage(void);
 Eterm erts_preloaded(Process* p);
 
-#ifndef ERTS_SMP
-extern void *erts_scheduler_stack_limit;
-#endif
 
 /* erl_md5.c */
 
@@ -1167,7 +1155,7 @@ typedef struct {
 #define ERTS_SPAWN_DRIVER 1
 #define ERTS_SPAWN_EXECUTABLE 2
 #define ERTS_SPAWN_ANY (ERTS_SPAWN_DRIVER | ERTS_SPAWN_EXECUTABLE)
-int erts_add_driver_entry(ErlDrvEntry *drv, DE_Handle *handle, int driver_list_locked);
+int erts_add_driver_entry(ErlDrvEntry *drv, DE_Handle *handle, int driver_list_locked, int taint);
 void erts_destroy_driver(erts_driver_t *drv);
 int erts_save_suspend_process_on_port(Port*, Process*);
 Port *erts_open_driver(erts_driver_t*, Eterm, char*, SysDriverOpts*, int *, int *);
@@ -1186,14 +1174,23 @@ void erts_emergency_close_ports(void);
 void erts_ref_to_driver_monitor(Eterm ref, ErlDrvMonitor *mon);
 Eterm erts_driver_monitor_to_ref(Eterm* hp, const ErlDrvMonitor *mon);
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_COUNT)
+#if defined(ERTS_ENABLE_LOCK_COUNT)
 void erts_lcnt_update_driver_locks(int enable);
 void erts_lcnt_update_port_locks(int enable);
 #endif
 
 /* driver_tab.c */
+typedef struct {
+    ErlDrvEntry* de;
+    int taint;
+} ErtsStaticDriver;
 typedef void *(*ErtsStaticNifInitFPtr)(void);
-ErtsStaticNifInitFPtr erts_static_nif_get_nif_init(const char *name, int len);
+typedef struct ErtsStaticNifEntry_ {
+    const char *nif_name;
+    ErtsStaticNifInitFPtr nif_init;
+    int taint;
+} ErtsStaticNifEntry;
+ErtsStaticNifEntry* erts_static_nif_get_nif_init(const char *name, int len);
 int erts_is_static_nif(void *handle);
 void erts_init_static_drivers(void);
 
@@ -1280,7 +1277,7 @@ char* erts_convert_filename_to_wchar(byte* bytes, Uint size,
                                      char *statbuf, size_t statbuf_size,
                                      ErtsAlcType_t alloc_type, Sint* used,
                                      Uint extra_wchars);
-Eterm erts_convert_native_to_filename(Process *p, byte *bytes);
+Eterm erts_convert_native_to_filename(Process *p, size_t size, byte *bytes);
 Eterm erts_utf8_to_list(Process *p, Uint num, byte *bytes, Uint sz, Uint left,
 			Uint *num_built, Uint *num_eaten, Eterm tail);
 int erts_utf8_to_latin1(byte* dest, const byte* source, int slen);
@@ -1398,7 +1395,7 @@ Uint erts_current_reductions(Process* current, Process *p);
 
 int erts_print_system_version(fmtfn_t to, void *arg, Process *c_p);
 
-int erts_hibernate(Process* c_p, Eterm module, Eterm function, Eterm args, Eterm* reg);
+int erts_hibernate(Process* c_p, Eterm* reg);
 
 ERTS_GLB_FORCE_INLINE int erts_is_literal(Eterm tptr, Eterm *ptr);
 
diff --git a/erts/emulator/beam/hash.c b/erts/emulator/beam/hash.c
index 8548e30e8b..6a31489473 100644
--- a/erts/emulator/beam/hash.c
+++ b/erts/emulator/beam/hash.c
@@ -152,7 +152,7 @@ Hash* hash_init(int type, Hash* h, char* name, int size, HashFunctions fun)
 
     h->bucket = (HashBucket**) fun.meta_alloc(h->meta_alloc_type, sz);
 
-    sys_memzero(h->bucket, sz);
+    memzero(h->bucket, sz);
     h->is_allocated = 0;
     h->name = name;
     h->fun = fun;
@@ -224,7 +224,7 @@ static void rehash(Hash* h, int grow)
     sz = h->size*sizeof(HashBucket*);
 
     new_bucket = (HashBucket **) h->fun.meta_alloc(h->meta_alloc_type, sz);
-    sys_memzero(new_bucket, sz);
+    memzero(new_bucket, sz);
 
     for (i = 0; i < old_size; i++) {
 	HashBucket* b = h->bucket[i];
diff --git a/erts/emulator/beam/index.c b/erts/emulator/beam/index.c
index 7bf1a032c1..be1771b037 100644
--- a/erts/emulator/beam/index.c
+++ b/erts/emulator/beam/index.c
@@ -98,7 +98,7 @@ index_put_entry(IndexTable* t, void* tmpl)
      * Do a write barrier here to allow readers to do lock free iteration.
      * erts_index_num_entries() does matching read barrier.
      */
-    ERTS_SMP_WRITE_MEMORY_BARRIER;
+    ERTS_THR_WRITE_MEMORY_BARRIER;
     t->entries++;
 
     return p;
diff --git a/erts/emulator/beam/index.h b/erts/emulator/beam/index.h
index 6c07571df6..30bc6a1121 100644
--- a/erts/emulator/beam/index.h
+++ b/erts/emulator/beam/index.h
@@ -88,7 +88,7 @@ ERTS_GLB_INLINE int erts_index_num_entries(IndexTable* t)
      * on tables where entries are never erased.
      * index_put_entry() does matching write barrier.
      */
-    ERTS_SMP_READ_MEMORY_BARRIER;
+    ERTS_THR_READ_MEMORY_BARRIER;
     return ret;
 }
 
diff --git a/erts/emulator/beam/instrs.tab b/erts/emulator/beam/instrs.tab
new file mode 100644
index 0000000000..6a531fcc09
--- /dev/null
+++ b/erts/emulator/beam/instrs.tab
@@ -0,0 +1,973 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+// Stack manipulation instructions
+
+allocate(NeedStack, Live) {
+    $AH($NeedStack, 0, $Live);
+}
+
+allocate_heap(NeedStack, NeedHeap, Live) {
+    $AH($NeedStack, $NeedHeap, $Live);
+}
+
+allocate_init(NeedStack, Live, Y) {
+    $AH($NeedStack, 0, $Live);
+    make_blank($Y);
+}
+
+allocate_zero(NeedStack, Live) {
+    Eterm* ptr;
+    int i = $NeedStack;
+    $AH(i, 0, $Live);
+    for (ptr = E + i; ptr > E; ptr--) {
+	 make_blank(*ptr);
+    }
+}
+
+allocate_heap_zero(NeedStack, NeedHeap, Live) {
+    Eterm* ptr;
+    int i = $NeedStack;
+    $AH(i, $NeedHeap, $Live);
+    for (ptr = E + i; ptr > E; ptr--) {
+        make_blank(*ptr);
+    }
+}
+
+// This instruction is probably never used (because it is combined with a
+// a return). However, a future compiler might for some reason emit a
+// deallocate not followed by a return, and that should work.
+
+deallocate(Deallocate) {
+    //| -no_prefetch
+    SET_CP(c_p, (BeamInstr *) cp_val(*E));
+    E = ADD_BYTE_OFFSET(E, $Deallocate);
+}
+
+deallocate_return(Deallocate) {
+    //| -no_next
+    int words_to_pop = $Deallocate;
+    SET_I((BeamInstr *) cp_val(*E));
+    E = ADD_BYTE_OFFSET(E, words_to_pop);
+    CHECK_TERM(x(0));
+    DispatchReturn;
+}
+
+move_deallocate_return(Src, Deallocate) {
+    x(0) = $Src;
+    $deallocate_return($Deallocate);
+}
+
+// Call instructions
+
+DISPATCH_REL(CallDest) {
+    //| -no_next
+    $SET_I_REL($CallDest);
+    DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
+    Dispatch();
+}
+
+DISPATCH_ABS(CallDest) {
+    //| -no_next
+    SET_I((BeamInstr *) $CallDest);
+    DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
+    Dispatch();
+}
+
+i_call(CallDest) {
+    SET_CP(c_p, $NEXT_INSTRUCTION);
+    $DISPATCH_REL($CallDest);
+}
+
+move_call(Src, CallDest) {
+    x(0) = $Src;
+    SET_CP(c_p, $NEXT_INSTRUCTION);
+    $DISPATCH_REL($CallDest);
+}
+
+i_call_last(CallDest, Deallocate) {
+    $deallocate($Deallocate);
+    $DISPATCH_REL($CallDest);
+}
+
+move_call_last(Src, CallDest, Deallocate) {
+    x(0) = $Src;
+    $i_call_last($CallDest, $Deallocate);
+}
+
+i_call_only(CallDest) {
+    $DISPATCH_REL($CallDest);
+}
+
+move_call_only(Src, CallDest) {
+    x(0) = $Src;
+    $i_call_only($CallDest);
+}
+
+DISPATCHX(Dest) {
+    //| -no_next
+    DTRACE_GLOBAL_CALL_FROM_EXPORT(c_p, $Dest);
+    // Dispatchx assumes the Export* is in Arg(0)
+    I = (&$Dest) - 1;
+    Dispatchx();
+}
+
+i_call_ext(Dest) {
+    SET_CP(c_p, $NEXT_INSTRUCTION);
+    $DISPATCHX($Dest);
+}
+
+i_move_call_ext(Src, Dest) {
+    x(0) = $Src;
+    $i_call_ext($Dest);
+}
+
+i_call_ext_only(Dest) {
+    $DISPATCHX($Dest);
+}
+
+i_move_call_ext_only(Dest, Src) {
+    x(0) = $Src;
+    $i_call_ext_only($Dest);
+}
+
+i_call_ext_last(Dest, Deallocate) {
+    $deallocate($Deallocate);
+    $DISPATCHX($Dest);
+}
+
+i_move_call_ext_last(Dest, StackOffset, Src) {
+    x(0) = $Src;
+    $i_call_ext_last($Dest, $StackOffset);
+}
+
+APPLY(I, Deallocate, Next) {
+    //| -no_next
+    HEAVY_SWAPOUT;
+    $Next = apply(c_p, reg, $I, $Deallocate);
+    HEAVY_SWAPIN;
+}
+
+HANDLE_APPLY_ERROR() {
+    I = handle_error(c_p, I, reg, &bif_export[BIF_apply_3]->info.mfa);
+    goto post_error_handling;
+}
+
+i_apply() {
+    BeamInstr *next;
+    $APPLY(NULL, 0, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        SET_CP(c_p, $NEXT_INSTRUCTION);
+        $DISPATCH_ABS(next);
+    }
+    $HANDLE_APPLY_ERROR();
+}
+
+i_apply_last(Deallocate) {
+    BeamInstr *next;
+    $APPLY(I, $Deallocate, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $deallocate($Deallocate);
+        $DISPATCH_ABS(next);
+    }
+    $HANDLE_APPLY_ERROR();
+}
+
+i_apply_only() {
+    BeamInstr *next;
+    $APPLY(I, 0, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $DISPATCH_ABS(next);
+    }
+    $HANDLE_APPLY_ERROR();
+}
+
+FIXED_APPLY(Arity, I, Deallocate, Next) {
+    //| -no_next
+    HEAVY_SWAPOUT;
+    $Next = fixed_apply(c_p, reg, $Arity, $I, $Deallocate);
+    HEAVY_SWAPIN;
+}
+
+apply(Arity) {
+    BeamInstr *next;
+    $FIXED_APPLY($Arity, NULL, 0, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        SET_CP(c_p, $NEXT_INSTRUCTION);
+        $DISPATCH_ABS(next);
+    }
+    $HANDLE_APPLY_ERROR();
+}
+
+apply_last(Arity, Deallocate) {
+    BeamInstr *next;
+    $FIXED_APPLY($Arity, I, $Deallocate, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $deallocate($Deallocate);
+        $DISPATCH_ABS(next);
+    }
+    $HANDLE_APPLY_ERROR();
+}
+
+APPLY_FUN(Next) {
+    HEAVY_SWAPOUT;
+    $Next = apply_fun(c_p, r(0), x(1), reg);
+    HEAVY_SWAPIN;
+}
+
+HANDLE_APPLY_FUN_ERROR() {
+     goto find_func_info;
+}
+
+DISPATCH_FUN(I) {
+    SET_I($I);
+    Dispatchfun();
+}
+
+i_apply_fun() {
+    BeamInstr *next;
+    $APPLY_FUN(next);
+    if (ERTS_LIKELY(next != NULL)) {
+        SET_CP(c_p, $NEXT_INSTRUCTION);
+        $DISPATCH_FUN(next);
+    }
+    $HANDLE_APPLY_FUN_ERROR();
+}
+
+i_apply_fun_last(Deallocate) {
+    BeamInstr *next;
+    $APPLY_FUN(next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $deallocate($Deallocate);
+        $DISPATCH_FUN(next);
+    }
+    $HANDLE_APPLY_FUN_ERROR();
+}
+
+i_apply_fun_only() {
+    BeamInstr *next;
+    $APPLY_FUN(next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $DISPATCH_FUN(next);
+    }
+    $HANDLE_APPLY_FUN_ERROR();
+}
+
+CALL_FUN(Fun, Next) {
+    //| -no_next
+    HEAVY_SWAPOUT;
+    $Next = call_fun(c_p, $Fun, reg, THE_NON_VALUE);
+    HEAVY_SWAPIN;
+}
+
+i_call_fun(Fun) {
+    BeamInstr *next;
+    $CALL_FUN($Fun, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        SET_CP(c_p, $NEXT_INSTRUCTION);
+        $DISPATCH_FUN(next);
+    }
+    $HANDLE_APPLY_FUN_ERROR();
+}
+
+i_call_fun_last(Fun, Deallocate) {
+    BeamInstr *next;
+    $CALL_FUN($Fun, next);
+    if (ERTS_LIKELY(next != NULL)) {
+        $deallocate($Deallocate);
+        $DISPATCH_FUN(next);
+    }
+    $HANDLE_APPLY_FUN_ERROR();
+}
+
+return() {
+    SET_I(c_p->cp);
+    DTRACE_RETURN_FROM_PC(c_p);
+
+    /*
+     * We must clear the CP to make sure that a stale value do not
+     * create a false module dependcy preventing code upgrading.
+     * It also means that we can use the CP in stack backtraces.
+     */
+    c_p->cp = 0;
+    CHECK_TERM(r(0));
+    HEAP_SPACE_VERIFIED(0);
+    DispatchReturn;
+}
+
+get_list(Src, Hd, Tl) {
+    Eterm* tmp_ptr = list_val($Src);
+    Eterm hd, tl;
+    hd = CAR(tmp_ptr);
+    tl = CDR(tmp_ptr);
+    $Hd = hd;
+    $Tl = tl;
+}
+
+get_hd(Src, Hd) {
+    Eterm* tmp_ptr = list_val($Src);
+    $Hd = CAR(tmp_ptr);
+}
+
+get_tl(Src, Tl) {
+    Eterm* tmp_ptr = list_val($Src);
+    $Tl = CDR(tmp_ptr);
+}
+
+i_get(Src, Dst) {
+    $Dst = erts_pd_hash_get(c_p, $Src);
+}
+
+i_get_hash(Src, Hash, Dst) {
+    $Dst = erts_pd_hash_get_with_hx(c_p, $Hash, $Src);
+}
+
+i_get_tuple_element(Src, Element, Dst) {
+    Eterm* src = ADD_BYTE_OFFSET(tuple_val($Src), $Element);
+    $Dst = *src;
+}
+
+i_get_tuple_element2(Src, Element, Dst) {
+    Eterm* src;
+    Eterm* dst;
+    Eterm E1, E2;
+    src = ADD_BYTE_OFFSET(tuple_val($Src), $Element);
+    dst = &($Dst);
+    E1 = src[0];
+    E2 = src[1];
+    dst[0] = E1;
+    dst[1] = E2;
+}
+
+i_get_tuple_element2y(Src, Element, D1, D2) {
+    Eterm* src;
+    Eterm E1, E2;
+    src = ADD_BYTE_OFFSET(tuple_val($Src), $Element);
+    E1 = src[0];
+    E2 = src[1];
+    $D1 = E1;
+    $D2 = E2;
+}
+
+i_get_tuple_element3(Src, Element, Dst) {
+    Eterm* src;
+    Eterm* dst;
+    Eterm E1, E2, E3;
+    src = ADD_BYTE_OFFSET(tuple_val($Src), $Element);
+    dst = &($Dst);
+    E1 = src[0];
+    E2 = src[1];
+    E3 = src[2];
+    dst[0] = E1;
+    dst[1] = E2;
+    dst[2] = E3;
+}
+
+i_element := element_group.fetch.execute;
+
+
+element_group.head() {
+    Eterm element_tuple;
+}
+
+element_group.fetch(Src) {
+    element_tuple = $Src;
+}
+
+element_group.execute(Fail, Index, Dst) {
+    Eterm element_index = $Index;
+    if (ERTS_LIKELY(is_small(element_index) && is_tuple(element_tuple))) {
+        Eterm* tp = tuple_val(element_tuple);
+
+        if ((signed_val(element_index) >= 1) &&
+            (signed_val(element_index) <= arityval(*tp))) {
+            $Dst = tp[signed_val(element_index)];
+            $NEXT0();
+        }
+    }
+    c_p->freason = BADARG;
+    $BIF_ERROR_ARITY_2($Fail, BIF_element_2, element_index, element_tuple);
+}
+
+i_fast_element := fast_element_group.fetch.execute;
+
+fast_element_group.head() {
+    Eterm fast_element_tuple;
+}
+
+fast_element_group.fetch(Src) {
+    fast_element_tuple = $Src;
+}
+
+fast_element_group.execute(Fail, Index, Dst) {
+    if (ERTS_LIKELY(is_tuple(fast_element_tuple))) {
+        Eterm* tp = tuple_val(fast_element_tuple);
+        Eterm pos = $Index;	/* Untagged integer >= 1 */
+        if (pos <= arityval(*tp)) {
+            $Dst = tp[pos];
+            $NEXT0();
+	}
+    }
+    c_p->freason = BADARG;
+    $BIF_ERROR_ARITY_2($Fail, BIF_element_2, make_small($Index), fast_element_tuple);
+}
+
+init(Y) {
+    make_blank($Y);
+}
+
+init2(Y1, Y2) {
+    make_blank($Y1);
+    make_blank($Y2);
+}
+
+init3(Y1, Y2, Y3) {
+    make_blank($Y1);
+    make_blank($Y2);
+    make_blank($Y3);
+}
+
+i_make_fun(FunP, NumFree) {
+    HEAVY_SWAPOUT;
+    x(0) = new_fun(c_p, reg, (ErlFunEntry *) $FunP, $NumFree);
+    HEAVY_SWAPIN;
+}
+
+i_trim(Words) {
+    Uint cp = E[0];
+    E += $Words;
+    E[0] = cp;
+}
+
+move(Src, Dst) {
+    $Dst = $Src;
+}
+
+move3(S1, D1, S2, D2, S3, D3) {
+    $D1 = $S1;
+    $D2 = $S2;
+    $D3 = $S3;
+}
+
+move_dup(Src, D1, D2) {
+    $D1 = $D2 = $Src;
+}
+
+move2_par(S1, D1, S2, D2) {
+    Eterm V1, V2;
+    V1 = $S1;
+    V2 = $S2;
+    $D1 = V1;
+    $D2 = V2;
+}
+
+move_shift(Src, SD, D) {
+    Eterm V;
+    V = $Src;
+    $D = $SD;
+    $SD = V;
+}
+
+move_window3(S1, S2, S3, D) {
+    Eterm xt0, xt1, xt2;
+    Eterm* y = &$D;
+    xt0  = $S1;
+    xt1  = $S2;
+    xt2  = $S3;
+    y[0] = xt0;
+    y[1] = xt1;
+    y[2] = xt2;
+}
+
+move_window4(S1, S2, S3, S4, D) {
+    Eterm xt0, xt1, xt2, xt3;
+    Eterm* y = &$D;
+    xt0  = $S1;
+    xt1  = $S2;
+    xt2  = $S3;
+    xt3  = $S4;
+    y[0] = xt0;
+    y[1] = xt1;
+    y[2] = xt2;
+    y[3] = xt3;
+}
+
+move_window5(S1, S2, S3, S4, S5, D) {
+    Eterm xt0, xt1, xt2, xt3, xt4;
+    Eterm *y = &$D;
+    xt0  = $S1;
+    xt1  = $S2;
+    xt2  = $S3;
+    xt3  = $S4;
+    xt4  = $S5;
+    y[0] = xt0;
+    y[1] = xt1;
+    y[2] = xt2;
+    y[3] = xt3;
+    y[4] = xt4;
+}
+
+move_return(Src) {
+    //| -no_next
+    x(0) = $Src;
+    SET_I(c_p->cp);
+    c_p->cp = 0;
+    DispatchReturn;
+}
+
+move_x1(Src) {
+    x(1) = $Src;
+}
+
+move_x2(Src) {
+    x(2) = $Src;
+}
+
+node(Dst) {
+    $Dst = erts_this_node->sysname;
+}
+
+put_list(Hd, Tl, Dst) {
+    HTOP[0] = $Hd;
+    HTOP[1] = $Tl;
+    $Dst = make_list(HTOP);
+    HTOP += 2;
+}
+
+update_list(Hd, Dst) {
+    HTOP[0] = $Hd;
+    HTOP[1] = $Dst;
+    $Dst = make_list(HTOP);
+    HTOP += 2;
+}
+
+i_put_tuple := i_put_tuple.make.fill;
+
+i_put_tuple.make(Dst) {
+    $Dst = make_tuple(HTOP);
+}
+
+i_put_tuple.fill(Arity) {
+    Eterm* hp = HTOP;
+    Eterm arity = $Arity;
+
+    //| -no_next
+    *hp++ = make_arityval(arity);
+    I = $NEXT_INSTRUCTION;
+    do {
+        Eterm term = *I++;
+        switch (loader_tag(term)) {
+        case LOADER_X_REG:
+            *hp++ = x(loader_x_reg_index(term));
+            break;
+        case LOADER_Y_REG:
+            *hp++ = y(loader_y_reg_index(term));
+            break;
+        default:
+            *hp++ = term;
+            break;
+        }
+    } while (--arity != 0);
+    HTOP = hp;
+    ASSERT(VALID_INSTR(* (Eterm *)I));
+    Goto(*I);
+}
+
+self(Dst) {
+    $Dst = c_p->common.id;
+}
+
+set_tuple_element(Element, Tuple, Offset) {
+    Eterm* p;
+
+    ASSERT(is_tuple($Tuple));
+    p = (Eterm *) ((unsigned char *) tuple_val($Tuple) + $Offset);
+    *p = $Element;
+}
+
+swap(R1, R2) {
+    Eterm V = $R1;
+    $R1 = $R2;
+    $R2 = V;
+}
+
+swap_temp(R1, R2, Tmp) {
+    Eterm V = $R1;
+    $R1 = $R2;
+    $R2 = $Tmp = V;
+}
+
+test_heap(Nh, Live) {
+    $GC_TEST(0, $Nh, $Live);
+}
+
+test_heap_1_put_list(Nh, Reg) {
+    $test_heap($Nh, 1);
+    $put_list($Reg, x(0), x(0));
+}
+
+is_integer_allocate(Fail, Src, NeedStack, Live) {
+    //| -no_prefetch
+    $is_integer($Fail, $Src);
+    $AH($NeedStack, 0, $Live);
+}
+
+is_nonempty_list(Fail, Src) {
+    //| -no_prefetch
+    if (is_not_list($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_nonempty_list_test_heap(Fail, Need, Live) {
+    //| -no_prefetch
+    $is_nonempty_list($Fail, x(0));
+    $test_heap($Need, $Live);
+}
+
+is_nonempty_list_allocate(Fail, Src, Need, Live) {
+    //| -no_prefetch
+    $is_nonempty_list($Fail, $Src);
+    $AH($Need, 0, $Live);
+}
+
+is_nonempty_list_get_list(Fail, Src, Hd, Tl) {
+    //| -no_prefetch
+    $is_nonempty_list($Fail, $Src);
+    $get_list($Src, $Hd, $Tl);
+}
+
+jump(Fail) {
+    $JUMP($Fail);
+}
+
+move_jump(Fail, Src) {
+    x(0) = $Src;
+    $jump($Fail);
+}
+
+//
+// Test instructions.
+//
+
+is_atom(Fail, Src) {
+    if (is_not_atom($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_boolean(Fail, Src) {
+    if (($Src) != am_true && ($Src) != am_false) {
+        $FAIL($Fail);
+    }
+}
+
+is_binary(Fail, Src) {
+    if (is_not_binary($Src) || binary_bitsize($Src) != 0) {
+        $FAIL($Fail);
+    }
+}
+
+is_bitstring(Fail, Src) {
+  if (is_not_binary($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_float(Fail, Src) {
+    if (is_not_float($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_function(Fail, Src) {
+    if ( !(is_any_fun($Src)) ) {
+        $FAIL($Fail);
+    }
+}
+
+is_function2(Fail, Fun, Arity) {
+    if (erl_is_function(c_p, $Fun, $Arity) != am_true ) {
+        $FAIL($Fail);
+    }
+}
+
+is_integer(Fail, Src) {
+    if (is_not_integer($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_list(Fail, Src) {
+    if (is_not_list($Src) && is_not_nil($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_map(Fail, Src) {
+    if (is_not_map($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_nil(Fail, Src) {
+    if (is_not_nil($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_number(Fail, Src) {
+    if (is_not_integer($Src) && is_not_float($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_pid(Fail, Src) {
+    if (is_not_pid($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_port(Fail, Src) {
+    if (is_not_port($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_reference(Fail, Src) {
+    if (is_not_ref($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_tagged_tuple(Fail, Src, Arityval, Tag) {
+    Eterm term = $Src;
+    if (!(BEAM_IS_TUPLE(term) &&
+          (tuple_val(term))[0] == $Arityval &&
+          (tuple_val(term))[1] == $Tag)) {
+        $FAIL($Fail);
+    }
+}
+
+is_tuple(Fail, Src) {
+    if (is_not_tuple($Src)) {
+        $FAIL($Fail);
+    }
+}
+
+is_tuple_of_arity(Fail, Src, Arityval) {
+    Eterm term = $Src;
+    if (!(BEAM_IS_TUPLE(term) && *tuple_val(term) == $Arityval)) {
+        $FAIL($Fail);
+    }
+}
+
+test_arity(Fail, Pointer, Arity) {
+    if (*tuple_val($Pointer) != $Arity) {
+        $FAIL($Fail);
+    }
+}
+
+i_is_eq_exact_immed(Fail, X, Y) {
+    if ($X != $Y) {
+        $FAIL($Fail);
+    }
+}
+
+i_is_ne_exact_immed(Fail, X, Y) {
+    if ($X == $Y) {
+        $FAIL($Fail);
+    }
+}
+
+is_eq_exact(Fail, X, Y) {
+    if (!EQ($X, $Y)) {
+        $FAIL($Fail);
+    }
+}
+
+i_is_eq_exact_literal(Fail, Src, Literal) {
+    Eterm src = $Src;
+    if (is_immed(src) || !eq(src, $Literal)) {
+        $FAIL($Fail);
+    }
+}
+
+is_ne_exact(Fail, X, Y) {
+    if (EQ($X, $Y)) {
+        $FAIL($Fail);
+    }
+}
+
+i_is_ne_exact_literal(Fail, Src, Literal) {
+    Eterm src = $Src;
+    if (!is_immed(src) && eq(src, $Literal)) {
+        $FAIL($Fail);
+    }
+}
+
+is_eq(Fail, X, Y) {
+    CMP_EQ_ACTION($X, $Y, $FAIL($Fail));
+}
+
+is_ne(Fail, X, Y) {
+    CMP_NE_ACTION($X, $Y, $FAIL($Fail));
+}
+
+is_lt(Fail, X, Y) {
+    CMP_LT_ACTION($X, $Y, $FAIL($Fail));
+}
+
+is_ge(Fail, X, Y) {
+    CMP_GE_ACTION($X, $Y, $FAIL($Fail));
+}
+
+badarg(Fail) {
+    $BADARG($Fail);
+    //| -no_next;
+}
+
+badmatch(Src) {
+    c_p->fvalue = $Src;
+    c_p->freason = BADMATCH;
+    goto find_func_info;
+    //| -no_next;
+}
+
+case_end(Src) {
+    c_p->fvalue = $Src;
+    c_p->freason = EXC_CASE_CLAUSE;
+    goto find_func_info;
+    //| -no_next;
+}
+
+if_end() {
+    c_p->freason = EXC_IF_CLAUSE;
+    goto find_func_info;
+    //| -no_next;
+}
+
+system_limit(Fail) {
+    $SYSTEM_LIMIT($Fail);
+    //| -no_next;
+}
+
+catch(Y, Fail) {
+    c_p->catches++;
+    $Y = $Fail;
+}
+
+catch_end(Y) {
+    $try_end($Y);
+    if (is_non_value(r(0))) {
+        c_p->fvalue = NIL;
+        if (x(1) == am_throw) {
+            r(0) = x(2);
+        } else {
+            if (x(1) == am_error) {
+                SWAPOUT;
+                x(2) = add_stacktrace(c_p, x(2), x(3));
+                SWAPIN;
+            }
+            /* only x(2) is included in the rootset here */
+            if (E - HTOP < 3) {
+                SWAPOUT;
+                PROCESS_MAIN_CHK_LOCKS(c_p);
+                FCALLS -= erts_garbage_collect_nobump(c_p, 3, reg+2, 1, FCALLS);
+                ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+                PROCESS_MAIN_CHK_LOCKS(c_p);
+                SWAPIN;
+            }
+            r(0) = TUPLE2(HTOP, am_EXIT, x(2));
+            HTOP += 3;
+        }
+    }
+    CHECK_TERM(r(0));
+}
+
+try_end(Y) {
+    c_p->catches--;
+    make_blank($Y);
+}
+
+try_case(Y) {
+    $try_end($Y);
+    ASSERT(is_non_value(r(0)));
+    c_p->fvalue = NIL;
+    r(0) = x(1);
+    x(1) = x(2);
+    x(2) = x(3);
+}
+
+try_case_end(Src) {
+    c_p->fvalue = $Src;
+    c_p->freason = EXC_TRY_CLAUSE;
+    goto find_func_info;
+    //| -no_next;
+}
+
+i_raise() {
+    Eterm raise_trace = x(2);
+    Eterm raise_value = x(1);
+    struct StackTrace *s;
+
+    c_p->fvalue = raise_value;
+    c_p->ftrace = raise_trace;
+    s = get_trace_from_exc(raise_trace);
+    if (s == NULL) {
+        c_p->freason = EXC_ERROR;
+    } else {
+        c_p->freason = PRIMARY_EXCEPTION(s->freason);
+    }
+    goto find_func_info;
+    //| -no_next
+}
+
+build_stacktrace() {
+    SWAPOUT;
+    x(0) = build_stacktrace(c_p, x(0));
+    SWAPIN;
+}
+
+raw_raise() {
+    Eterm class = x(0);
+    Eterm value = x(1);
+    Eterm stacktrace = x(2);
+
+    if (class == am_error) {
+        c_p->freason = EXC_ERROR & ~EXF_SAVETRACE;
+        c_p->fvalue = value;
+        c_p->ftrace = stacktrace;
+        goto find_func_info;
+    } else if (class == am_exit) {
+        c_p->freason = EXC_EXIT & ~EXF_SAVETRACE;
+        c_p->fvalue = value;
+        c_p->ftrace = stacktrace;
+        goto find_func_info;
+    } else if (class == am_throw) {
+        c_p->freason = EXC_THROWN & ~EXF_SAVETRACE;
+        c_p->fvalue = value;
+        c_p->ftrace = stacktrace;
+        goto find_func_info;
+    } else {
+        x(0) = am_badarg;
+    }
+}
diff --git a/erts/emulator/beam/io.c b/erts/emulator/beam/io.c
index 13b8125a99..2446b3c074 100644
--- a/erts/emulator/beam/io.c
+++ b/erts/emulator/beam/io.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
+ * Copyright Ericsson AB 1996-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -53,6 +53,7 @@
 #include "erl_hl_timer.h"
 #include "erl_time.h"
 #include "erl_io_queue.h"
+#include "erl_proc_sig_queue.h"
 
 extern ErlDrvEntry fd_driver_entry;
 extern ErlDrvEntry vanilla_driver_entry;
@@ -60,13 +61,13 @@ extern ErlDrvEntry spawn_driver_entry;
 #ifndef __WIN32__
 extern ErlDrvEntry forker_driver_entry;
 #endif
-extern ErlDrvEntry *driver_tab[]; /* table of static drivers, only used during initialization */
+extern ErtsStaticDriver driver_tab[]; /* table of static drivers, only used during initialization */
 
 erts_driver_t *driver_list; /* List of all drivers, static and dynamic. */
-erts_smp_rwmtx_t erts_driver_list_lock; /* Mutex for driver list */
-static erts_smp_tsd_key_t driver_list_lock_status_key; /*stop recursive locks when calling 
+erts_rwmtx_t erts_driver_list_lock; /* Mutex for driver list */
+static erts_tsd_key_t driver_list_lock_status_key; /*stop recursive locks when calling
 							 driver init */
-static erts_smp_tsd_key_t driver_list_last_error_key;  /* Save last DDLL error on a 
+static erts_tsd_key_t driver_list_last_error_key;  /* Save last DDLL error on a
 							  per thread basis (for BC interfaces) */
 
 ErtsPTab erts_port erts_align_attribute(ERTS_CACHE_LINE_SIZE); /* The port table */
@@ -94,17 +95,11 @@ static int init_driver(erts_driver_t *, ErlDrvEntry *, DE_Handle *);
 static void terminate_port(Port *p);
 static void pdl_init(void);
 static int driver_failure_term(ErlDrvPort ix, Eterm term, int eof);
-#ifdef ERTS_SMP
 static void driver_monitor_lock_pdl(Port *p);
 static void driver_monitor_unlock_pdl(Port *p);
 #define DRV_MONITOR_LOOKUP_PORT_LOCK_PDL(Port) erts_thr_drvport2port((Port), 1)
 #define DRV_MONITOR_LOCK_PDL(Port) driver_monitor_lock_pdl(Port)
 #define DRV_MONITOR_UNLOCK_PDL(Port) driver_monitor_unlock_pdl(Port)
-#else
-#define DRV_MONITOR_LOOKUP_PORT_LOCK_PDL(Port) erts_thr_drvport2port((Port), 0)
-#define DRV_MONITOR_LOCK_PDL(Port) /* nothing */
-#define DRV_MONITOR_UNLOCK_PDL(Port) /* nothing */
-#endif
 
 #define ERL_SMALL_IO_BIN_LIMIT (4*ERL_ONHEAP_BIN_LIMIT)
 #define SMALL_WRITE_VEC  16
@@ -122,7 +117,7 @@ static ERTS_INLINE int
 is_port_ioq_empty(Port *pp)
 {
     int res;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
     if (!pp->port_data_lock)
 	res = (erts_ioq_size(&pp->ioq) == 0);
     else {
@@ -144,7 +139,7 @@ Uint
 erts_port_ioq_size(Port *pp)
 {
     ErlDrvSizeT res;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
     if (!pp->port_data_lock)
 	res = erts_ioq_size(&pp->ioq);
     else {
@@ -212,14 +207,13 @@ dtrace_drvport_str(ErlDrvPort drvport, char *port_buf)
 static ERTS_INLINE void
 kill_port(Port *pp)
 {
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(pp));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(pp));
     ERTS_TRACER_CLEAR(&ERTS_TRACER(pp));
     erts_ptab_delete_element(&erts_port, &pp->common); /* Time of death */
     erts_port_task_free_port(pp);
     /* In non-smp case the port structure may have been deallocated now */
 }
 
-#ifdef ERTS_SMP
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
 int
@@ -227,12 +221,11 @@ erts_lc_is_port_locked(Port *prt)
 {
     if (!prt)
 	return 0;
-    ERTS_SMP_LC_ASSERT(prt->lock);
-    return erts_smp_lc_mtx_is_locked(prt->lock);
+    ERTS_LC_ASSERT(prt->lock);
+    return erts_lc_mtx_is_locked(prt->lock);
 }
 #endif
 
-#endif /* #ifdef ERTS_SMP */
 
 static void initq(Port* prt);
 
@@ -256,25 +249,21 @@ static ERTS_INLINE void port_init_instr(Port *prt
      * Stuff that need to be initialized with the port id
      * in the instrumented case, but not in the normal case.
      */
-#ifdef ERTS_SMP
     ASSERT(prt->drv_ptr && prt->lock);
     if (!prt->drv_ptr->lock) {
         erts_mtx_init_locked(prt->lock, "port_lock", id, ERTS_LOCK_FLAGS_CATEGORY_IO);
     }
-#endif
     erts_port_task_init_sched(&prt->sched, id);
 }
 
 #if !ERTS_PORT_INIT_INSTR_NEED_ID
 static ERTS_INLINE void port_init_instr_abort(Port *prt)
 {
-#ifdef ERTS_SMP
     ASSERT(prt->drv_ptr && prt->lock);
     if (!prt->drv_ptr->lock) {
 	erts_mtx_unlock(prt->lock);
 	erts_mtx_destroy(prt->lock);
     }
-#endif
     erts_port_task_fini_sched(&prt->sched);
 }
 #endif
@@ -310,15 +299,12 @@ static Port *create_port(char *name,
     erts_aint32_t state = ERTS_PORT_SFLG_CONNECTED;
     erts_aint32_t x_pts_flgs = 0;
 
-#ifdef ERTS_SMP
-    ErtsRunQueue *runq;
     if (!driver_lock) {
 	/* Align size for mutex following port struct */
 	port_size = size = ERTS_ALC_DATA_ALIGN_SIZE(sizeof(Port));
 	size += sizeof(erts_mtx_t);
     }
     else
-#endif
 	port_size = size = ERTS_ALC_DATA_ALIGN_SIZE(sizeof(Port));
 
 #ifdef DEBUG
@@ -352,7 +338,6 @@ static Port *create_port(char *name,
 	p += busy_port_queue_size;
     }
 
-#ifdef ERTS_SMP
     if (driver_lock) {
 	prt->lock = driver_lock;
 	erts_mtx_lock(driver_lock);
@@ -362,17 +347,18 @@ static Port *create_port(char *name,
 	p += sizeof(erts_mtx_t);
 	state |= ERTS_PORT_SFLG_PORT_SPECIFIC_LOCK;
     }
-    if (erts_get_scheduler_data())
-        runq = erts_get_runq_current(NULL);
-    else
-        runq = ERTS_RUNQ_IX(0);
-    erts_smp_atomic_set_nob(&prt->run_queue, (erts_aint_t) runq);
+
+    {
+        ErtsRunQueue *runq;
+        ErtsSchedulerData *esdp = erts_get_scheduler_data();
+        if (esdp)
+            runq = erts_get_runq_current(esdp);
+        else
+            runq = ERTS_RUNQ_IX(0);
+        erts_init_runq_port(prt, runq);
+    }
 
     prt->xports = NULL;
-#else
-    erts_atomic32_init_nob(&prt->refc, 1);
-    prt->cleanup = 0;
-#endif
     
     erts_port_task_pre_init_sched(&prt->sched, busy_port_queue);
 
@@ -381,6 +367,7 @@ static Port *create_port(char *name,
     prt->drv_ptr = driver;
     ERTS_P_LINKS(prt) = NULL;
     ERTS_P_MONITORS(prt) = NULL;
+    ERTS_P_LT_MONITORS(prt) = NULL;
     prt->linebuf = NULL;
     prt->suspended = NULL;
     erts_init_port_data(prt);
@@ -393,7 +380,7 @@ static Port *create_port(char *name,
     prt->common.u.alive.reg = NULL;
     ERTS_PTMR_INIT(prt);
     erts_port_task_handle_init(&prt->timeout_task);
-    erts_smp_atomic_init_nob(&prt->psd, (erts_aint_t) NULL);
+    erts_atomic_init_nob(&prt->psd, (erts_aint_t) NULL);
     prt->async_open_port = NULL;
     prt->drv_data = (SWord) 0;
     prt->os_pid = -1;
@@ -420,10 +407,8 @@ static Port *create_port(char *name,
 #if !ERTS_PORT_INIT_INSTR_NEED_ID
 	port_init_instr_abort(prt);
 #endif
-#ifdef ERTS_SMP
 	if (driver_lock)
 	    erts_mtx_unlock(driver_lock);
-#endif
 	if (enop)
 	    *enop = 0;
 	erts_free(ERTS_ALC_T_PORT, prt);
@@ -436,7 +421,7 @@ static Port *create_port(char *name,
 
     initq(prt);
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     if (erts_port_schedule_all_ops)
 	x_pts_flgs |= ERTS_PTS_FLG_FORCE_SCHED;
@@ -445,29 +430,17 @@ static Port *create_port(char *name,
 	x_pts_flgs |= ERTS_PTS_FLG_PARALLELISM;
 
     if (x_pts_flgs)
-	erts_smp_atomic32_read_bor_nob(&prt->sched.flags, x_pts_flgs);
+	erts_atomic32_read_bor_nob(&prt->sched.flags, x_pts_flgs);
 
     erts_atomic32_set_relb(&prt->state, state);
     return prt;
 }
 
-#ifndef ERTS_SMP
-void
-erts_port_cleanup(Port *prt)
-{
-    if (prt->drv_ptr && prt->drv_ptr->handle)
-	erts_ddll_dereference_driver(prt->drv_ptr->handle);
-    prt->drv_ptr = NULL;
-    erts_port_dec_refc(prt);
-}
-#endif
 
 void
 erts_port_free(Port *prt)
 {
-#if defined(ERTS_SMP) || defined(DEBUG) || defined(ERTS_ENABLE_LOCK_CHECK)
     erts_aint32_t state = erts_atomic32_read_nob(&prt->state);
-#endif
     ERTS_LC_ASSERT(state & (ERTS_PORT_SFLG_INITIALIZING
 			    | ERTS_PORT_SFLG_FREE));
     ASSERT(state & ERTS_PORT_SFLG_PORT_DEBUG);
@@ -481,7 +454,6 @@ erts_port_free(Port *prt)
         prt->async_open_port = NULL;
     }
 
-#ifdef ERTS_SMP
     ASSERT(prt->lock);
     if (state & ERTS_PORT_SFLG_PORT_SPECIFIC_LOCK)
 	erts_mtx_destroy(prt->lock);
@@ -498,7 +470,6 @@ erts_port_free(Port *prt)
      */
     if (prt->drv_ptr->handle)
 	erts_ddll_dereference_driver(prt->drv_ptr->handle);
-#endif
     erts_free(ERTS_ALC_T_PORT, prt);
 }
 
@@ -533,7 +504,7 @@ erts_save_suspend_process_on_port(Port *prt, Process *process)
     int saved;
     erts_aint32_t flags;
     erts_port_task_sched_lock(&prt->sched);
-    flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    flags = erts_atomic32_read_nob(&prt->sched.flags);
     saved = (flags & ERTS_PTS_FLGS_BUSY) && !(flags & ERTS_PTS_FLG_EXIT);
     if (saved)
 	erts_proclist_store_last(&prt->suspended, erts_proclist_create(process));
@@ -577,16 +548,16 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
     erts_mtx_t *driver_lock = NULL;
     int cprt_flgs = 0;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
-    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+    erts_rwmtx_rlock(&erts_driver_list_lock);
     if (!driver) {
 	for (driver = driver_list; driver; driver = driver->next) {
 	    if (sys_strcmp(driver->name, name) == 0)
 		break;
 	}
 	if (!driver) { 
-	    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	    erts_rwmtx_runlock(&erts_driver_list_lock);
 	    ERTS_OPEN_DRIVER_RET(NULL, -3, BADARG);
 	}
     }
@@ -619,7 +590,7 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
 	     */
 	    
 	    for (d = driver_list; d; d = d->next) {
-		if (strcmp(d->name, name) == 0 && 
+		if (sys_strcmp(d->name, name) == 0 && 
 		    erts_ddll_driver_ok(d->handle)) {
 		    driver = d;
 		    break;
@@ -631,19 +602,17 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
     }
 
     if (driver == NULL || (driver != &spawn_driver && opts->exit_status)) {
-	erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	erts_rwmtx_runlock(&erts_driver_list_lock);
 	ERTS_OPEN_DRIVER_RET(NULL, -3, BADARG);
     }
 
-#ifdef ERTS_SMP
     driver_lock = driver->lock;
-#endif
 
     if (driver->handle != NULL) {
 	erts_ddll_increment_port_count(driver->handle);
 	erts_ddll_reference_driver(driver->handle);
     }
-    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+    erts_rwmtx_runlock(&erts_driver_list_lock);
 
     /*
      * We'll set up the port before calling the start function,
@@ -656,9 +625,9 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
     port = create_port(name, driver, driver_lock, cprt_flgs, pid, &port_errno);
     if (!port) {
 	if (driver->handle) {
-	    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+	    erts_rwmtx_rlock(&erts_driver_list_lock);
 	    erts_ddll_decrement_port_count(driver->handle);
-	    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	    erts_rwmtx_runlock(&erts_driver_list_lock);
 	    erts_ddll_dereference_driver(driver->handle);
 	}
 	if (port_errno)
@@ -671,7 +640,7 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
 	trace_port_open(port,
 			pid,
 			erts_atom_put((byte *) port->name,
-				      strlen(port->name),
+				      sys_strlen(port->name),
 				      ERTS_ATOM_ENC_LATIN1,
 				      1));
     }
@@ -726,11 +695,9 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
 	if (IS_TRACED_FL(port, F_TRACE_SCHED_PORTS)) {
 	    trace_sched_ports_where(port, am_out, am_open);
 	}
-#ifdef ERTS_SMP
 	if (port->xports)
 	    erts_port_handle_xports(port);
 	ASSERT(!port->xports);
-#endif
     }
 
     if (error_type) {
@@ -745,9 +712,9 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
 	    port->linebuf = NULL;
 	}
 	if (driver->handle != NULL) {
-	    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+	    erts_rwmtx_rlock(&erts_driver_list_lock);
 	    erts_ddll_decrement_port_count(driver->handle);
-	    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	    erts_rwmtx_runlock(&erts_driver_list_lock);
 	}
 	kill_port(port);
 	erts_port_release(port);
@@ -759,7 +726,6 @@ erts_open_driver(erts_driver_t* driver,	/* Pointer to driver. */
 #undef ERTS_OPEN_DRIVER_RET
 }
 
-#ifdef ERTS_SMP
 
 struct ErtsXPortsList_ {
     ErtsXPortsList *next;
@@ -768,7 +734,6 @@ struct ErtsXPortsList_ {
 
 ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(xports_list, ErtsXPortsList, 50, ERTS_ALC_T_XPORTS_LIST)
 
-#endif
 
 /*
  * Driver function to create new instances of a driver
@@ -785,10 +750,10 @@ driver_create_port(ErlDrvPort creator_port_ix, /* Creating port */
     Port *creator_port;
     Port* port;
     erts_driver_t *driver;
-    Process *rp;
     erts_mtx_t *driver_lock = NULL;
+    ErtsLinkData *ldp;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     /* Need to be called from a scheduler thread */
     if (!erts_get_scheduler_id())
@@ -798,17 +763,13 @@ driver_create_port(ErlDrvPort creator_port_ix, /* Creating port */
     if (creator_port == ERTS_INVALID_ERL_DRV_PORT)
 	return ERTS_INVALID_ERL_DRV_PORT;
 
-    rp = erts_proc_lookup(pid);
-    if (!rp)
-	return ERTS_INVALID_ERL_DRV_PORT;
-
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(creator_port));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(creator_port));
 
     driver = creator_port->drv_ptr;
-    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+    erts_rwmtx_rlock(&erts_driver_list_lock);
     if (!erts_ddll_driver_ok(driver->handle)) {
-	erts_smp_rwmtx_runlock(&erts_driver_list_lock);
-	return ERTS_INVALID_ERL_DRV_PORT;
+	erts_rwmtx_runlock(&erts_driver_list_lock);
+        return ERTS_INVALID_ERL_DRV_PORT;
     }
 
     if (driver->handle != NULL) {
@@ -816,60 +777,57 @@ driver_create_port(ErlDrvPort creator_port_ix, /* Creating port */
 	erts_ddll_reference_referenced_driver(driver->handle);
     }
 
-#ifdef ERTS_SMP
     driver_lock = driver->lock;
-#endif
 
-    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+    erts_rwmtx_runlock(&erts_driver_list_lock);
 
     /* Inherit parallelism flag from parent */
     if (ERTS_PTS_FLG_PARALLELISM &
-	erts_smp_atomic32_read_nob(&creator_port->sched.flags))
+	erts_atomic32_read_nob(&creator_port->sched.flags))
 	cprt_flgs |= ERTS_CREATE_PORT_FLAG_PARALLELISM;
     port = create_port(name, driver, driver_lock, cprt_flgs, pid, NULL);
     if (!port) {
 	if (driver->handle) {
-	    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+	    erts_rwmtx_rlock(&erts_driver_list_lock);
 	    erts_ddll_decrement_port_count(driver->handle);
-	    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	    erts_rwmtx_runlock(&erts_driver_list_lock);
 	    erts_ddll_dereference_driver(driver->handle);
 	}
 	return ERTS_INVALID_ERL_DRV_PORT;
     }
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(port));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(port));
+
+    ldp = erts_link_create(ERTS_LNK_TYPE_PORT,
+                           port->common.id, pid);
+    ASSERT(ldp->a.other.item == pid);
+    ASSERT(ldp->b.other.item == port->common.id);
+    erts_link_tree_insert(&ERTS_P_LINKS(port), &ldp->a);
 
-    erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-    if (ERTS_PROC_IS_EXITING(rp)) {
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
+    if (!erts_proc_sig_send_link(NULL, pid, &ldp->b)) {
+        erts_link_tree_delete(&ERTS_P_LINKS(port), &ldp->a);
+        erts_link_release_both(ldp);
 	if (driver->handle) {
-	    erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+	    erts_rwmtx_rlock(&erts_driver_list_lock);
 	    erts_ddll_decrement_port_count(driver->handle);
-	    erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	    erts_rwmtx_runlock(&erts_driver_list_lock);
 	}
 	kill_port(port);
 	erts_port_release(port);
 	return ERTS_INVALID_ERL_DRV_PORT;
     }
 
-    erts_add_link(&ERTS_P_LINKS(port), LINK_PID, pid);
-    erts_add_link(&ERTS_P_LINKS(rp), LINK_PID, port->common.id);
-    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-
-#ifdef ERTS_SMP
     if (!driver_lock) {
 	ErtsXPortsList *xplp = xports_list_alloc();
 	xplp->port = port;
 	xplp->next = creator_port->xports;
 	creator_port->xports = xplp;
     }
-#endif
 
     port->drv_data = (UWord) drv_data;
 
     return ERTS_Port2ErlDrvPort(port);
 }
 
-#ifdef ERTS_SMP
 int erts_port_handle_xports(Port *prt)
 {
     int reds = 0;
@@ -898,7 +856,6 @@ int erts_port_handle_xports(Port *prt)
     prt->xports = NULL;
     return reds;
 }
-#endif
 
 typedef enum {
     ERTS_TRY_IMM_DRV_CALL_OK,
@@ -945,12 +902,12 @@ try_imm_drv_call(ErtsTryImmDrvCallState *sp)
 	invalid_sched_flags |= ERTS_PTS_FLG_PARALLELISM;
 
     if (sp->pre_chk_sched_flags) {
-	sp->sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+	sp->sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
 	if (sp->sched_flags & invalid_sched_flags)
 	    return ERTS_TRY_IMM_DRV_CALL_INVALID_SCHED_FLAGS;
     }
 
-    if (erts_smp_port_trylock(prt) == EBUSY)
+    if (erts_port_trylock(prt) == EBUSY)
 	return ERTS_TRY_IMM_DRV_CALL_BUSY_LOCK;
 
     invalid_state = sp->state;
@@ -964,7 +921,7 @@ try_imm_drv_call(ErtsTryImmDrvCallState *sp)
     if (prof_runnable_ports)
 	erts_port_task_sched_lock(&prt->sched);
 
-    act = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    act = erts_atomic32_read_nob(&prt->sched.flags);
 
     do {
 	erts_aint32_t new;
@@ -976,7 +933,7 @@ try_imm_drv_call(ErtsTryImmDrvCallState *sp)
 	}
 	exp = act;
 	new = act | ERTS_PTS_FLG_EXEC_IMM;
-	act = erts_smp_atomic32_cmpxchg_mb(&prt->sched.flags, new, exp);
+	act = erts_atomic32_cmpxchg_mb(&prt->sched.flags, new, exp);
     } while (act != exp);
     
     sp->sched_flags = act;
@@ -998,14 +955,17 @@ try_imm_drv_call(ErtsTryImmDrvCallState *sp)
 	    profile_runnable_proc(c_p, am_inactive);
 
 	reds_left_in = ERTS_BIF_REDS_LEFT(c_p);
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+
+        ASSERT((c_p->scheduler_data)->current_port == NULL);
+        (c_p->scheduler_data)->current_port = prt;
     }
 
     ASSERT(0 <= reds_left_in && reds_left_in <= CONTEXT_REDS);
     sp->reds_left_in = reds_left_in;
     prt->reds = CONTEXT_REDS - reds_left_in;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prof_runnable_ports | IS_TRACED_FL(prt, F_TRACE_SCHED_PORTS)) {
 	if (prof_runnable_ports && !(act & (ERTS_PTS_FLG_IN_RUNQ|ERTS_PTS_FLG_EXEC)))
@@ -1043,9 +1003,9 @@ finalize_imm_drv_call(ErtsTryImmDrvCallState *sp)
     if (prof_runnable_ports)
 	erts_port_task_sched_lock(&prt->sched);
 
-    act = erts_smp_atomic32_read_band_mb(&prt->sched.flags,
+    act = erts_atomic32_read_band_mb(&prt->sched.flags,
 					 ~ERTS_PTS_FLG_EXEC_IMM);
-    ERTS_SMP_LC_ASSERT(act & ERTS_PTS_FLG_EXEC_IMM);
+    ERTS_LC_ASSERT(act & ERTS_PTS_FLG_EXEC_IMM);
 
     if (prof_runnable_ports | IS_TRACED_FL(prt, F_TRACE_SCHED_PORTS)) {
 	if (IS_TRACED_FL(prt, F_TRACE_SCHED_PORTS))
@@ -1060,7 +1020,10 @@ finalize_imm_drv_call(ErtsTryImmDrvCallState *sp)
     erts_port_release(prt);
 
     if (c_p) {
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+        ASSERT((c_p->scheduler_data)->current_port == prt);
+        (c_p->scheduler_data)->current_port = NULL;
+
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 
 	if (reds != (CONTEXT_REDS - sp->reds_left_in)) {
 	    int bump_reds = reds - (CONTEXT_REDS - sp->reds_left_in);
@@ -1171,7 +1134,7 @@ port_sched_op_reply(Eterm to, Uint32 *ref_num, Eterm msg, Port* prt)
 				  prt);
 
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
     }
 }
 
@@ -1189,7 +1152,7 @@ erts_schedule_proc2port_signal(Process *c_p,
     int sched_res;
     if (!refp) {
 	if (c_p)
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
     }
     else {
 	ASSERT(c_p);
@@ -1210,21 +1173,10 @@ erts_schedule_proc2port_signal(Process *c_p,
 	 *       otherwise, next receive will *not* work
 	 *       as expected!
 	 */
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-
-	if (ERTS_PROC_PENDING_EXIT(c_p)) {
-	    /* need to exit caller instead */
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	    KILL_CATCHES(c_p);
-	    c_p->freason = EXC_EXIT;
-	    return ERTS_PORT_OP_CALLER_EXIT;
-	}
 
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	c_p->msg.save = c_p->msg.last;
-
-        erts_smp_proc_unlock(c_p, (ERTS_PROC_LOCKS_MSG_RECEIVE
-                                   | ERTS_PROC_LOCK_MAIN));
+        ERTS_RECV_MARK_SAVE(c_p);
+        ERTS_RECV_MARK_SET(c_p);
+        erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
     }
 
 
@@ -1239,7 +1191,7 @@ erts_schedule_proc2port_signal(Process *c_p,
 					task_flags);
 
     if (c_p)
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 
     if (sched_res != 0) {
 	if (refp) {
@@ -1250,9 +1202,9 @@ erts_schedule_proc2port_signal(Process *c_p,
 	     * containing the reference created above...
 	     */
 	    ASSERT(c_p);
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+	    erts_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
 	    JOIN_MESSAGE(c_p);
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+	    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
 	    *refp = NIL;
 	}
 	return ERTS_PORT_OP_DROPPED;
@@ -1282,29 +1234,12 @@ erts_schedule_port2port_signal(Eterm port_num, ErtsProc2PortSigData *sigdp,
 
 static ERTS_INLINE void
 send_badsig(Port *prt) {
-    ErtsProcLocks rp_locks = ERTS_PROC_LOCKS_XSIG_SEND;
-    Process* rp;
     Eterm connected = ERTS_PORT_GET_CONNECTED(prt);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     ERTS_LC_ASSERT(erts_get_scheduler_id());
-
     ASSERT(is_internal_pid(connected));
-
-    rp = erts_proc_lookup_raw(connected);
-    if (rp) {
-	erts_smp_proc_lock(rp, rp_locks);
-	if (!ERTS_PROC_IS_EXITING(rp))
-	    (void) erts_send_exit_signal(NULL,
-					 prt->common.id,
-					 rp,
-					 &rp_locks, 
-					 am_badsig,
-					 NIL,
-					 NULL,
-					 0);
-	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
-    } /* exit sent */
+    erts_proc_sig_send_exit(NULL, prt->common.id, connected,
+                            am_badsig, NIL, 0);
 } /* send_badsig */
 
 static void
@@ -1426,7 +1361,7 @@ call_driver_outputv(int bang_op,
 	ErlDrvSizeT size = evp->size;
 	ERTS_MSACC_PUSH_AND_SET_STATE_M(ERTS_MSACC_STATE_PORT);
 
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt)	
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt)
 			   || ERTS_IS_CRASH_DUMPING);
 
 
@@ -1483,7 +1418,7 @@ port_sig_outputv(Port *prt, erts_aint32_t state, int op, ErtsProc2PortSigData *s
     case ERTS_PROC2PORT_SIG_EXEC:
 	/* Execution of a scheduled outputv() call */
 
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
 	if (state & ERTS_PORT_SFLGS_INVALID_LOOKUP)
 	    reply = am_badarg;
@@ -1539,7 +1474,7 @@ call_driver_output(int bang_op,
     else {
 	ErtsSchedulerData *esdp = erts_get_scheduler_data();
         ERTS_MSACC_PUSH_AND_SET_STATE_M(ERTS_MSACC_STATE_PORT);
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt)	
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt)
 			   || ERTS_IS_CRASH_DUMPING);
 
 #ifdef USE_VM_PROBES
@@ -1590,7 +1525,7 @@ port_sig_output(Port *prt, erts_aint32_t state, int op, ErtsProc2PortSigData *si
     case ERTS_PROC2PORT_SIG_EXEC:
 	/* Execution of a scheduled output() call */
 
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
 	if (state & ERTS_PORT_SFLGS_INVALID_LOOKUP)
 	    reply = am_badarg;
@@ -1805,7 +1740,7 @@ erts_port_output(Process *c_p,
      * Assumes caller have checked that port is valid...
      */
 
-    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
     if (sched_flags & (busy_flgs|ERTS_PTS_FLG_EXIT))
 	return ((sched_flags & ERTS_PTS_FLG_EXIT)
 		? ERTS_PORT_OP_DROPPED
@@ -2188,7 +2123,7 @@ erts_port_output(Process *c_p,
     }
 
     if (!(flags & ERTS_PORT_SIG_FLG_FORCE)) {
-	sched_flags = erts_smp_atomic32_read_acqb(&prt->sched.flags);
+	sched_flags = erts_atomic32_read_acqb(&prt->sched.flags);
 	if (!(sched_flags & ERTS_PTS_FLG_BUSY_PORT)) {
 	    if (async_nosuspend)
 		erts_port_task_tmp_handle_detach(ns_pthp);
@@ -2237,11 +2172,11 @@ call_deliver_port_exit(int bang_op,
     }
 
     if (broken_link) {
-	ErtsLink *lnk = erts_remove_link(&ERTS_P_LINKS(prt), from);
-	if (lnk)
-	    erts_destroy_link(lnk);
-	else
+	ErtsLink *lnk = erts_link_tree_lookup(ERTS_P_LINKS(prt), from);
+	if (!lnk) 
 	    return ERTS_PORT_OP_DROPPED;
+        erts_link_tree_delete(&ERTS_P_LINKS(prt), lnk);
+        erts_link_release(lnk);
     }
 
     if (IS_TRACED_FL(prt, F_TRACE_RECEIVE))
@@ -2410,27 +2345,45 @@ set_port_connected(int bang_op,
 #endif
     }
     else { /* Port BIF operation */
-	Process *rp = erts_proc_lookup_raw(connect);
-	if (!rp)
-	    return ERTS_PORT_OP_DROPPED;
-	erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-	if (ERTS_PROC_IS_EXITING(rp)) {
-	    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	    return ERTS_PORT_OP_DROPPED;
-	}
+        int created;
+        ErtsLink *lnk;
+
+        if (is_not_internal_pid(connect))
+            return ERTS_PORT_OP_DROPPED;
+
+        lnk = erts_link_tree_lookup_create(&ERTS_P_LINKS(prt), &created,
+                                           ERTS_LNK_TYPE_PORT, prt->common.id,
+                                           connect);
+        if (created) {
+            ErtsLinkData *ldp;
+            ErtsLink *olnk = erts_link_to_other(lnk, &ldp);
+            ASSERT(olnk->other.item == prt->common.id);
+            if (!erts_proc_sig_send_link(NULL, connect, olnk)) {
+                erts_link_tree_delete(&ERTS_P_LINKS(prt), lnk);
+                erts_link_release_both(ldp);
+                return ERTS_PORT_OP_DROPPED;
+            }
+            if (IS_TRACED_FL(prt, F_TRACE_PORTS))
+                trace_port(prt, am_getting_linked, connect);
+        }
 
-	erts_add_link(&ERTS_P_LINKS(rp), LINK_PID, prt->common.id);
-	erts_add_link(&ERTS_P_LINKS(prt), LINK_PID, connect);
+#ifdef USE_VM_PROBES
+	if (DTRACE_ENABLED(port_connect)) {
+            Eterm old_connected = ERTS_PORT_GET_CONNECTED(prt);
+            DTRACE_CHARBUF(process_str, DTRACE_TERM_BUF_SIZE);
+            DTRACE_CHARBUF(port_str, DTRACE_TERM_BUF_SIZE);
+            DTRACE_CHARBUF(newprocess_str, DTRACE_TERM_BUF_SIZE);
 
-        if (IS_TRACED_FL(rp, F_TRACE_PROCS))
-            trace_proc(NULL, 0, rp, am_getting_linked, prt->common.id);
+            dtrace_pid_str(old_connected, process_str);
+            erts_snprintf(port_str, sizeof(DTRACE_CHARBUF_NAME(port_str)),
+                          "%T", prt->common.id);
+            dtrace_pid_str(connect, newprocess_str);
+            DTRACE4(port_connect, process_str, port_str, prt->name, newprocess_str);
+	}
+#endif
 
 	ERTS_PORT_SET_CONNECTED(prt, connect);
 
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-
-        if (IS_TRACED_FL(prt, F_TRACE_PORTS))
-            trace_port(prt, am_getting_linked, connect);
         if (IS_TRACED_FL(prt, F_TRACE_RECEIVE))
             trace_port_receive(prt, from, am_connect, connect);
         if (IS_TRACED_FL(prt, F_TRACE_SEND)) {
@@ -2438,18 +2391,6 @@ set_port_connected(int bang_op,
             trace_port_send(prt, from, TUPLE2(hp, prt->common.id, am_connected), 1);
         }
 
-#ifdef USE_VM_PROBES
-	if (DTRACE_ENABLED(port_connect)) {
-	    DTRACE_CHARBUF(process_str, DTRACE_TERM_BUF_SIZE);
-	    DTRACE_CHARBUF(port_str, DTRACE_TERM_BUF_SIZE);
-	    DTRACE_CHARBUF(newprocess_str, DTRACE_TERM_BUF_SIZE);
-
-	    dtrace_pid_str(connect, process_str);
-	    erts_snprintf(port_str, sizeof(DTRACE_CHARBUF_NAME(port_str)), "%T", prt->common.id);
-	    dtrace_proc_str(rp, newprocess_str);
-	    DTRACE4(port_connect, process_str, port_str, prt->name, newprocess_str);
-	}
-#endif
     }
 
     return ERTS_PORT_OP_DONE;
@@ -2537,13 +2478,24 @@ erts_port_connect(Process *c_p,
 }
 
 static void
-port_unlink(Port *prt, Eterm from)
+port_unlink(Port *prt, ErtsLink *lnk)
 {
-    ErtsLink *lnk = erts_remove_link(&ERTS_P_LINKS(prt), from);
-    if (lnk) {
+    ErtsLinkData *ldp;
+    ErtsLink *dlnk, *llnk;
+
+    llnk = erts_link_to_other(lnk, &ldp);
+    dlnk = erts_link_tree_key_delete(&ERTS_P_LINKS(prt), llnk);
+    if (!dlnk)
+        erts_link_release(lnk);
+    else {
         if (IS_TRACED_FL(prt, F_TRACE_PORTS))
-            trace_port(prt, am_getting_unlinked, from);
-	erts_destroy_link(lnk);
+            trace_port(prt, am_getting_unlinked, llnk->other.item);
+        if (dlnk == llnk)
+            erts_link_release_both(ldp);
+        else {
+            erts_link_release(lnk);
+            erts_link_release(dlnk);
+        }
     }
 }
 
@@ -2551,14 +2503,14 @@ static int
 port_sig_unlink(Port *prt, erts_aint32_t state, int op, ErtsProc2PortSigData *sigdp)
 {
     if (op == ERTS_PROC2PORT_SIG_EXEC)
-	port_unlink(prt, sigdp->u.unlink.from);
+	port_unlink(prt, sigdp->u.unlink.lnk);
     if (sigdp->flags & ERTS_P2P_SIG_DATA_FLG_REPLY)
 	port_sched_op_reply(sigdp->caller, sigdp->ref, am_true, prt);
     return ERTS_PORT_REDS_UNLINK;
 }
 
 ErtsPortOpResult
-erts_port_unlink(Process *c_p, Port *prt, Eterm from, Eterm *refp)
+erts_port_unlink(Process *c_p, Port *prt, ErtsLink *lnk, Eterm *refp)
 {
     ErtsProc2PortSigData *sigdp;
     ErtsTryImmDrvCallState try_call_state
@@ -2571,7 +2523,7 @@ erts_port_unlink(Process *c_p, Port *prt, Eterm from, Eterm *refp)
 
     switch (try_imm_drv_call(&try_call_state)) {
     case ERTS_TRY_IMM_DRV_CALL_OK:
-	port_unlink(prt, from);
+	port_unlink(prt, lnk);
 	finalize_imm_drv_call(&try_call_state);
 	BUMP_REDS(c_p, ERTS_PORT_REDS_UNLINK);
 	return ERTS_PORT_OP_DONE;
@@ -2584,11 +2536,12 @@ erts_port_unlink(Process *c_p, Port *prt, Eterm from, Eterm *refp)
 
     sigdp = erts_port_task_alloc_p2p_sig_data();
     sigdp->flags = ERTS_P2P_SIG_TYPE_UNLINK;
-    sigdp->u.unlink.from = from;
-    
+    sigdp->u.unlink.port_id = prt->common.id;
+    sigdp->u.unlink.lnk = lnk;
+
     return erts_schedule_proc2port_signal(c_p,
 					  prt,
-					  c_p ? c_p->common.id : from,
+					  c_p->common.id,
 					  refp,
 					  sigdp,
 					  0,
@@ -2597,45 +2550,24 @@ erts_port_unlink(Process *c_p, Port *prt, Eterm from, Eterm *refp)
 }
 
 static void
-port_link_failure(Eterm port_id, Eterm linker)
+port_link_failure(Eterm port_id, ErtsLink *lnk)
 {
-    Process *rp;
-    ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCKS_XSIG_SEND;
-    ASSERT(is_internal_pid(linker));
-    rp = erts_pid2proc(NULL, 0, linker, rp_locks);
-    if (rp) {
-	ErtsLink *rlnk = erts_remove_link(&ERTS_P_LINKS(rp), port_id);
-	if (rlnk) {
-	    int xres = erts_send_exit_signal(NULL,
-					     port_id,
-					     rp,
-					     &rp_locks, 
-					     am_noproc,
-					     NIL,
-					     NULL,
-					     0);
-	    if (xres >= 0) {
-		/* We didn't exit the process and it is traced */
-		if (IS_TRACED_FL(rp, F_TRACE_PROCS))
-		    trace_proc(NULL, 0, rp, am_getting_unlinked, port_id);
-	    }
-	    if (rp_locks)
-		erts_smp_proc_unlock(rp, rp_locks);
-	}
-    }
+    erts_proc_sig_send_link_exit(NULL, port_id, lnk, am_noproc, NIL);
 }
 
 static void
-port_link(Port *prt, erts_aint32_t state, Eterm to)
+port_link(Port *prt, erts_aint32_t state, ErtsLink *lnk)
 {
-    if (IS_TRACED_FL(prt, F_TRACE_PORTS))
-        trace_port(prt, am_getting_linked, to);
-    if (!(state & ERTS_PORT_SFLGS_INVALID_LOOKUP)) {
-	erts_add_link(&ERTS_P_LINKS(prt), LINK_PID, to);
-    } else {
-	port_link_failure(prt->common.id, to);
-        if (IS_TRACED_FL(prt, F_TRACE_PORTS))
-            trace_port(prt, am_unlink, to);
+    if (state & ERTS_PORT_SFLGS_INVALID_LOOKUP)
+	port_link_failure(prt->common.id, lnk);
+    else {
+        ErtsLink *rlnk;
+        rlnk = erts_link_tree_insert_addr_replace(&ERTS_P_LINKS(prt),
+                                                  lnk);
+        if (rlnk)
+            erts_link_release(rlnk);
+        else if (IS_TRACED_FL(prt, F_TRACE_PORTS))
+            trace_port(prt, am_getting_linked, lnk->other.item);
     }
 }
 
@@ -2643,17 +2575,16 @@ static int
 port_sig_link(Port *prt, erts_aint32_t state, int op, ErtsProc2PortSigData *sigdp)
 {
     if (op == ERTS_PROC2PORT_SIG_EXEC)
-	port_link(prt, state, sigdp->u.link.to);
-    else {
-	port_link_failure(sigdp->u.link.port, sigdp->u.link.to);
-    }
+	port_link(prt, state, sigdp->u.link.lnk);
+    else
+	port_link_failure(sigdp->u.link.port_id, sigdp->u.link.lnk);
     if (sigdp->flags & ERTS_P2P_SIG_DATA_FLG_REPLY)
 	port_sched_op_reply(sigdp->caller, sigdp->ref, am_true, prt);
     return ERTS_PORT_REDS_LINK;
 }
 
 ErtsPortOpResult
-erts_port_link(Process *c_p, Port *prt, Eterm to, Eterm *refp)
+erts_port_link(Process *c_p, Port *prt, ErtsLink *lnk, Eterm *refp)
 {
     ErtsProc2PortSigData *sigdp;
     ErtsTryImmDrvCallState try_call_state
@@ -2666,7 +2597,7 @@ erts_port_link(Process *c_p, Port *prt, Eterm to, Eterm *refp)
 
     switch (try_imm_drv_call(&try_call_state)) {
     case ERTS_TRY_IMM_DRV_CALL_OK:
-	port_link(prt, try_call_state.state, to);
+	port_link(prt, try_call_state.state, lnk);
 	finalize_imm_drv_call(&try_call_state);
 	BUMP_REDS(c_p, ERTS_PORT_REDS_LINK);
 	return ERTS_PORT_OP_DONE;
@@ -2679,12 +2610,12 @@ erts_port_link(Process *c_p, Port *prt, Eterm to, Eterm *refp)
 
     sigdp = erts_port_task_alloc_p2p_sig_data();
     sigdp->flags = ERTS_P2P_SIG_TYPE_LINK;
-    sigdp->u.link.port = prt->common.id;
-    sigdp->u.link.to = to;
+    sigdp->u.link.port_id = prt->common.id;
+    sigdp->u.link.lnk = lnk;
     
     return erts_schedule_proc2port_signal(c_p,
 					  prt,
-					  c_p ? c_p->common.id : to,
+					  c_p->common.id,
 					  refp,
 					  sigdp,
 					  0,
@@ -2693,51 +2624,23 @@ erts_port_link(Process *c_p, Port *prt, Eterm to, Eterm *refp)
 }
 
 static void
-port_monitor_failure(Eterm port_id, Eterm origin, Eterm ref_DOWN)
+port_monitor_failure(Eterm port_id, ErtsMonitor *mon)
 {
-    Process       *origin_p;
-    ErtsProcLocks p_locks = ERTS_PROC_LOCK_LINK;
-    ASSERT(is_internal_pid(origin));
-
-    origin_p = erts_pid2proc(NULL, 0, origin, p_locks);
-    if (! origin_p) { return; }
-
-    /* Send the DOWN message immediately. Ref is made on the fly because
-     * caller has never seen it yet. */
-    erts_queue_monitor_message(origin_p, &p_locks, ref_DOWN,
-                               am_port, port_id, am_noproc);
-    erts_smp_proc_unlock(origin_p, p_locks);
+    erts_proc_sig_send_monitor_down(mon, am_noproc);
 }
 
 /* Origin wants to monitor port Prt. State contains possible error, which has
  * happened just before. Name is either NIL or an atom, if user monitors
  * a port by name. Ref is premade reference that will be returned to user */
 static void
-port_monitor(Port *prt, erts_aint32_t state, Eterm origin,
-             Eterm name, Eterm ref)
+port_monitor(Port *prt, erts_aint32_t state, ErtsMonitor *mon)
 {
-    Eterm name_or_nil = is_atom(name) ? name : NIL;
-
-    ASSERT(is_pid(origin));
-    ASSERT(is_atom(name) || is_port(name) || name == NIL);
-    ASSERT(is_internal_ordinary_ref(ref));
-
-    if (!(state & ERTS_PORT_SFLGS_INVALID_LOOKUP)) {
-        ErtsProcLocks p_locks = ERTS_PROC_LOCK_LINK;
-
-        Process *origin_p = erts_pid2proc(NULL, 0, origin, p_locks);
-        if (! origin_p) {
-            goto failure;
-        }
-        erts_add_monitor(&ERTS_P_MONITORS(origin_p), MON_ORIGIN, ref,
-                         prt->common.id, name_or_nil);
-        erts_add_monitor(&ERTS_P_MONITORS(prt), MON_TARGET, ref,
-                         origin, name_or_nil);
-
-        erts_smp_proc_unlock(origin_p, p_locks);
-    } else {
-failure:
-        port_monitor_failure(prt->common.id, origin, ref);
+    ASSERT(prt);
+    if (state & ERTS_PORT_SFLGS_INVALID_LOOKUP)
+        port_monitor_failure(prt->common.id, mon);
+    else {
+        ASSERT(erts_monitor_is_target(mon));
+        erts_monitor_list_insert(&ERTS_P_LT_MONITORS(prt), mon);
     }
 }
 
@@ -2745,24 +2648,11 @@ static int
 port_sig_monitor(Port *prt, erts_aint32_t state, int op,
                  ErtsProc2PortSigData *sigdp)
 {
-    Eterm hp[ERTS_REF_THING_SIZE];
-    Eterm ref = make_internal_ref(&hp);
-    write_ref_thing(hp, sigdp->ref[0], sigdp->ref[1], sigdp->ref[2]);
-
-    if (op == ERTS_PROC2PORT_SIG_EXEC) {
-        /* erts_add_monitor call inside port_monitor will copy ref from hp */
-        port_monitor(prt, state,
-                     sigdp->u.monitor.origin,
-                     sigdp->u.monitor.name,
-                     ref);
-    } else {
-        port_monitor_failure(sigdp->u.monitor.name,
-                             sigdp->u.monitor.origin,
-                             ref);
-    }
-    if (sigdp->flags & ERTS_P2P_SIG_DATA_FLG_REPLY) {
-        port_sched_op_reply(sigdp->caller, sigdp->ref, am_true, prt);
-    }
+    if (op == ERTS_PROC2PORT_SIG_EXEC)
+        port_monitor(prt, state, sigdp->u.monitor.mon);
+    else
+        port_monitor_failure(sigdp->u.monitor.port_id,
+                             sigdp->u.monitor.mon);
     return ERTS_PORT_REDS_MONITOR;
 }
 
@@ -2770,95 +2660,63 @@ port_sig_monitor(Port *prt, erts_aint32_t state, int op,
  * a reference (ref may be rewritten to be used to serve additionally as a
  * signal id). Name is atom if user monitors port by name or NIL */
 ErtsPortOpResult
-erts_port_monitor(Process *origin, Port *port, Eterm name, Eterm *refp)
+erts_port_monitor(Process *c_p, Port *port, ErtsMonitor *mon)
 {
     ErtsProc2PortSigData *sigdp;
     ErtsTryImmDrvCallState try_call_state
         = ERTS_INIT_TRY_IMM_DRV_CALL_STATE(
-                origin, port, ERTS_PORT_SFLGS_INVALID_LOOKUP,
+                c_p,
+                port,
+                ERTS_PORT_SFLGS_INVALID_LOOKUP,
                 0,
-                0, /* trap_ref is always set so !trap_ref always is false */
+                !0,
                 am_monitor);
 
-    ASSERT(origin);
+    ASSERT(c_p);
     ASSERT(port);
-    ASSERT(is_atom(name) || is_port(name));
-    ASSERT(refp);
+    ASSERT(mon);
 
     switch (try_imm_drv_call(&try_call_state)) {
     case ERTS_TRY_IMM_DRV_CALL_OK:
-        port_monitor(port, try_call_state.state, origin->common.id, name, *refp);
+        port_monitor(port, try_call_state.state, mon);
         finalize_imm_drv_call(&try_call_state);
-        BUMP_REDS(origin, ERTS_PORT_REDS_MONITOR);
+        BUMP_REDS(c_p, ERTS_PORT_REDS_MONITOR);
         return ERTS_PORT_OP_DONE;
     case ERTS_TRY_IMM_DRV_CALL_INVALID_PORT:
-        return ERTS_PORT_OP_BADARG;
+        return ERTS_PORT_OP_DROPPED;
     default:
         break; /* Schedule call instead... */
     }
 
     sigdp = erts_port_task_alloc_p2p_sig_data();
     sigdp->flags = ERTS_P2P_SIG_TYPE_MONITOR;
-    sigdp->u.monitor.origin = origin->common.id;
-    sigdp->u.monitor.name   = name; /* either named monitor, or port id */
+    sigdp->u.monitor.port_id = port->common.id;
+    sigdp->u.monitor.mon = mon;
 
     /* Ref contents will be initialized here */
-    return erts_schedule_proc2port_signal(origin, port, origin->common.id,
-                                          refp, sigdp, 0, NULL,
+    return erts_schedule_proc2port_signal(c_p,
+                                          port,
+                                          c_p->common.id,
+                                          NULL,
+                                          sigdp,
+                                          0,
+                                          NULL,
                                           port_sig_monitor);
 }
 
-static void
-port_demonitor_failure(Eterm port_id, Eterm origin, Eterm ref)
-{
-    Process       *origin_p;
-    ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK;
-    ErtsMonitor   *mon1;
-    ASSERT(is_internal_pid(origin));
-
-    origin_p = erts_pid2proc(NULL, 0, origin, rp_locks);
-    if (! origin_p) { return; }
-
-    /* do not send any DOWN messages, drop monitors on process */
-    mon1 = erts_remove_monitor(&ERTS_P_MONITORS(origin_p), ref);
-    if (mon1 != NULL) {
-        erts_destroy_monitor(mon1);
-    }
-
-    erts_smp_proc_unlock(origin_p, rp_locks);
-}
-
 /* Origin wants to demonitor port Prt. State contains possible error, which has
  * happened just before. Ref is reference to monitor */
 static void
-port_demonitor(Port *port, erts_aint32_t state, Eterm origin, Eterm ref)
+port_demonitor(Port *port, erts_aint32_t state, ErtsMonitor *mon)
 {
-    ASSERT(port);
-    ASSERT(is_pid(origin));
-    ASSERT(is_internal_ref(ref));
-
-    if (!(state & ERTS_PORT_SFLGS_INVALID_LOOKUP)) {
-        ErtsProcLocks p_locks = ERTS_PROC_LOCK_LINK;
-        Process *origin_p = erts_pid2proc(NULL, 0, origin, p_locks);
-        if (origin_p) {
-            ErtsMonitor *mon1 = erts_remove_monitor(&ERTS_P_MONITORS(origin_p),
-                                                    ref);
-            if (mon1 != NULL) {
-                erts_destroy_monitor(mon1);
-            }
-        }
-        if (1) {
-            ErtsMonitor *mon2 = erts_remove_monitor(&ERTS_P_MONITORS(port),
-                                                    ref);
-            if (mon2 != NULL) {
-                erts_destroy_monitor(mon2);
-            }
-        }
-        if (origin_p) { /* when origin is dying, it won't be found */
-            erts_smp_proc_unlock(origin_p, p_locks);
-        }
-    } else {
-        port_demonitor_failure(port->common.id, origin, ref);
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
+    ASSERT(port && mon);
+    ASSERT(erts_monitor_is_origin(mon));
+    if (!erts_monitor_is_in_table(&mdp->target))
+        erts_monitor_release(mon);
+    else {
+        erts_monitor_list_delete(&ERTS_P_LT_MONITORS(port), &mdp->target);
+        erts_monitor_release_both(mdp);
     }
 }
 
@@ -2866,73 +2724,47 @@ static int
 port_sig_demonitor(Port *prt, erts_aint32_t state, int op,
                    ErtsProc2PortSigData *sigdp)
 {
-    Eterm hp[ERTS_REF_THING_SIZE];
-    Eterm ref = make_internal_ref(&hp);
-    write_ref_thing(hp, sigdp->u.demonitor.ref[0],
-                    sigdp->u.demonitor.ref[1],
-                    sigdp->u.demonitor.ref[2]);
-    if (op == ERTS_PROC2PORT_SIG_EXEC) {
-        port_demonitor(prt, state, sigdp->u.demonitor.origin, ref);
-    } else {
-        port_demonitor_failure(sigdp->u.demonitor.name,
-                               sigdp->u.demonitor.origin,
-                               ref);
-    }
-    if (sigdp->flags & ERTS_P2P_SIG_DATA_FLG_REPLY) {
-        port_sched_op_reply(sigdp->caller, sigdp->ref, am_true, prt);
-    }
+    if (op == ERTS_PROC2PORT_SIG_EXEC)
+        port_demonitor(prt, state, sigdp->u.demonitor.mon);
+    else
+        erts_monitor_release(sigdp->u.demonitor.mon);
     return ERTS_PORT_REDS_DEMONITOR;
 }
 
-/* Removes monitor between origin and target, identified by ref.
- * Mode defines normal or relaxed demonitor rules (process is at death) */
-ErtsPortOpResult erts_port_demonitor(Process *origin, ErtsDemonitorMode mode,
-                                     Port *target, Eterm ref,
-                                     Eterm *trap_ref)
+ErtsPortOpResult
+erts_port_demonitor(Process *c_p, Port *prt, ErtsMonitor *mon)
 {
-    Process *c_p = mode == ERTS_PORT_DEMONITOR_NORMAL ? origin : NULL;
     ErtsProc2PortSigData *sigdp;
     ErtsTryImmDrvCallState try_call_state
         = ERTS_INIT_TRY_IMM_DRV_CALL_STATE(
                 c_p,
-                target, ERTS_PORT_SFLGS_INVALID_LOOKUP,
+                prt, ERTS_PORT_SFLGS_INVALID_LOOKUP,
                 0,
-                !trap_ref,
+                !0,
                 am_demonitor);
 
-    ASSERT(origin);
-    ASSERT(target);
-    ASSERT(is_internal_ref(ref));
+    ASSERT(c_p && prt && mon);
 
     switch (try_imm_drv_call(&try_call_state)) {
     case ERTS_TRY_IMM_DRV_CALL_OK:
-        port_demonitor(target, try_call_state.state, origin->common.id, ref);
+        port_demonitor(prt, try_call_state.state, mon);
         finalize_imm_drv_call(&try_call_state);
-        if (mode == ERTS_PORT_DEMONITOR_NORMAL) {
-            BUMP_REDS(origin, ERTS_PORT_REDS_DEMONITOR);
-        }
+        BUMP_REDS(c_p, ERTS_PORT_REDS_DEMONITOR);
         return ERTS_PORT_OP_DONE;
     case ERTS_TRY_IMM_DRV_CALL_INVALID_PORT:
-        return ERTS_PORT_OP_BADARG;
+        return ERTS_PORT_OP_DROPPED;
     default:
         break; /* Schedule call instead... */
     }
 
     sigdp = erts_port_task_alloc_p2p_sig_data();
     sigdp->flags = ERTS_P2P_SIG_TYPE_DEMONITOR;
-    sigdp->u.demonitor.origin = origin->common.id;
-    sigdp->u.demonitor.name = target->common.id;
-    {
-	Uint32 *nums = internal_ref_numbers(ref);
-        /* Start from 1 skip ref arity */
-        sys_memcpy(sigdp->u.demonitor.ref,
-		   nums,
-                   sizeof(sigdp->u.demonitor.ref));
-    }
+    sigdp->u.demonitor.port_id = prt->common.id;
+    sigdp->u.demonitor.mon = mon;
 
     /* Ref contents will be initialized here */
-    return erts_schedule_proc2port_signal(c_p, target, origin->common.id,
-                                          trap_ref, sigdp, 0, NULL,
+    return erts_schedule_proc2port_signal(c_p, prt, c_p->common.id,
+                                          NULL, sigdp, 0, NULL,
                                           port_sig_demonitor);
 }
 
@@ -2941,11 +2773,13 @@ init_ack_send_reply(Port *port, Eterm resp)
 {
 
     if (!is_internal_port(resp)) {
-        Process *rp = erts_proc_lookup_raw(port->async_open_port->to);
-        erts_smp_proc_lock(rp, ERTS_PROC_LOCK_LINK);
-        erts_remove_link(&ERTS_P_LINKS(port), port->async_open_port->to);
-        erts_remove_link(&ERTS_P_LINKS(rp), port->common.id);
-        erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
+        Eterm proc = port->async_open_port->to;
+        ErtsLink *lnk = erts_link_tree_lookup(ERTS_P_LINKS(port),
+                                              proc);
+        if (lnk) {
+            erts_link_tree_delete(&ERTS_P_LINKS(port), lnk);
+            erts_proc_sig_send_unlink(NULL, lnk);
+        }
     }
     port_sched_op_reply(port->async_open_port->to,
                         port->async_open_port->ref,
@@ -2972,7 +2806,7 @@ erl_drv_init_ack(ErlDrvPort ix, ErlDrvData res) {
             break;
         case -2: {
             char *str = erl_errno_id(errno);
-            resp = erts_atom_put((byte *) str, strlen(str),
+            resp = erts_atom_put((byte *) str, sys_strlen(str),
                                  ERTS_ATOM_ENC_LATIN1, 1);
             break;
         }
@@ -3007,11 +2841,11 @@ void erts_init_io(int port_tab_size,
 		  int port_tab_size_ignore_files,
 		  int legacy_port_tab)
 {
-    ErlDrvEntry** dp;
+    ErtsStaticDriver* dp;
     UWord common_element_size;
-    erts_smp_rwmtx_opt_t drv_list_rwmtx_opts = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    drv_list_rwmtx_opts.type = ERTS_SMP_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
-    drv_list_rwmtx_opts.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t drv_list_rwmtx_opts = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    drv_list_rwmtx_opts.type = ERTS_RWMTX_TYPE_EXTREMELY_FREQUENT_READ;
+    drv_list_rwmtx_opts.lived = ERTS_RWMTX_LONG_LIVED;
 
     erts_atomic64_init_nob(&bytes_in, 0);
     erts_atomic64_init_nob(&bytes_out, 0);
@@ -3019,11 +2853,9 @@ void erts_init_io(int port_tab_size,
     common_element_size = ERTS_ALC_DATA_ALIGN_SIZE(sizeof(Port));
     common_element_size += ERTS_ALC_DATA_ALIGN_SIZE(sizeof(ErtsPortTaskBusyPortQ));
     common_element_size += 10; /* name */
-#ifdef ERTS_SMP
     common_element_size += sizeof(erts_mtx_t);
 
     init_xports_list_alloc();
-#endif
 
     pdl_init();
 
@@ -3038,12 +2870,12 @@ void erts_init_io(int port_tab_size,
     else if (port_tab_size < ERTS_MIN_PORTS)
 	port_tab_size = ERTS_MIN_PORTS;
 
-    erts_smp_rwmtx_init_opt(&erts_driver_list_lock, &drv_list_rwmtx_opts, "driver_list", NIL,
+    erts_rwmtx_init_opt(&erts_driver_list_lock, &drv_list_rwmtx_opts, "driver_list", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
     driver_list = NULL;
-    erts_smp_tsd_key_create(&driver_list_lock_status_key,
+    erts_tsd_key_create(&driver_list_lock_status_key,
 			    "erts_driver_list_lock_status_key");
-    erts_smp_tsd_key_create(&driver_list_last_error_key,
+    erts_tsd_key_create(&driver_list_last_error_key,
 			    "erts_driver_list_last_error_key");
 
     erts_ptab_init_table(&erts_port,
@@ -3058,8 +2890,8 @@ void erts_init_io(int port_tab_size,
 
     sys_init_io();
 
-    erts_smp_tsd_set(driver_list_lock_status_key, (void *) 1);
-    erts_smp_rwmtx_rwlock(&erts_driver_list_lock);
+    erts_tsd_set(driver_list_lock_status_key, (void *) 1);
+    erts_rwmtx_rwlock(&erts_driver_list_lock);
 
     init_driver(&fd_driver, &fd_driver_entry, NULL);
     init_driver(&vanilla_driver, &vanilla_driver_entry, NULL);
@@ -3068,23 +2900,20 @@ void erts_init_io(int port_tab_size,
     init_driver(&forker_driver, &forker_driver_entry, NULL);
 #endif
     erts_init_static_drivers();
-    for (dp = driver_tab; *dp != NULL; dp++)
-	erts_add_driver_entry(*dp, NULL, 1);
+    for (dp = driver_tab; dp->de != NULL; dp++)
+	erts_add_driver_entry(dp->de, NULL, 1, dp->taint);
 
-    erts_smp_tsd_set(driver_list_lock_status_key, NULL);
-    erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+    erts_tsd_set(driver_list_lock_status_key, NULL);
+    erts_rwmtx_rwunlock(&erts_driver_list_lock);
 }
 
-#if defined(ERTS_ENABLE_LOCK_COUNT) && defined(ERTS_SMP)
+#if defined(ERTS_ENABLE_LOCK_COUNT)
 static void lcnt_enable_driver_lock_count(erts_driver_t *dp, int enable)
 {
     if (dp->lock) {
         if (enable) {
-            Eterm name_as_atom = erts_atom_put((byte*)dp->name, sys_strlen(dp->name),
-                                                ERTS_ATOM_ENC_LATIN1, 1);
-
-            erts_lcnt_install_new_lock_info(&dp->lock->lcnt, "driver_lock", name_as_atom,
-                ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_IO);
+            erts_lcnt_install_new_lock_info(&dp->lock->lcnt, "driver_lock",
+                dp->name_atom, ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_IO);
         } else {
             erts_lcnt_uninstall(&dp->lock->lcnt);
         }
@@ -3164,7 +2993,8 @@ void erts_lcnt_update_port_locks(int enable) {
     }
 }
 
-#endif /* defined(ERTS_ENABLE_LOCK_COUNT) && defined(ERTS_SMP) */
+#endif /* defined(ERTS_ENABLE_LOCK_COUNT) */
+
 /*
  * Buffering of data when using line oriented I/O on ports
  */
@@ -3343,12 +3173,10 @@ deliver_result(Port *prt, Eterm sender, Eterm pid, Eterm res)
     ErtsProcLocks rp_locks = 0;
     int scheduler = erts_get_scheduler_id() != 0;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     ASSERT(!prt || prt->common.id == sender);
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
-    ASSERT(!prt || erts_lc_is_port_locked(prt));
-#endif
+    ERTS_LC_ASSERT(!prt || erts_lc_is_port_locked(prt));
 
     ASSERT(is_internal_port(sender) && is_internal_pid(pid));
 
@@ -3376,7 +3204,7 @@ deliver_result(Port *prt, Eterm sender, Eterm pid, Eterm res)
 	erts_queue_message(rp, rp_locks, mp, tuple, sender);
 
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
 	if (!scheduler)
 	    erts_proc_dec_refc(rp);
 
@@ -3407,8 +3235,8 @@ static void deliver_read_message(Port* prt, erts_aint32_t state, Eterm to,
     int scheduler = erts_get_scheduler_id() != 0;
     int trace_send = IS_TRACED_FL(prt, F_TRACE_SEND);
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));    
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_CHK_NO_PROC_LOCKS;
 
     need = 3 + 3 + 2*hlen;
 
@@ -3434,24 +3262,11 @@ static void deliver_read_message(Port* prt, erts_aint32_t state, Eterm to,
     if ((state & ERTS_PORT_SFLG_BINARY_IO) == 0) {
 	listp = buf_to_intlist(&hp, buf, len, listp);
     } else if (buf != NULL) {
-	ProcBin* pb;
-	Binary* bptr;
-
-	bptr = erts_bin_nrml_alloc(len);
+	Binary* bptr = erts_bin_nrml_alloc(len);
 	sys_memcpy(bptr->orig_bytes, buf, len);
 
-	pb = (ProcBin *) hp;
-	pb->thing_word = HEADER_PROC_BIN;
-	pb->size = len;
-	pb->next = ohp->first;
-	ohp->first = (struct erl_off_heap_header*)pb;
-	pb->val = bptr;
-	pb->bytes = (byte*) bptr->orig_bytes;
-	pb->flags = 0;
+        listp = erts_build_proc_bin(ohp, hp, bptr);
 	hp += PROC_BIN_SIZE;
-
-	OH_OVERHEAD(ohp, pb->size / sizeof(Eterm));
-	listp = make_binary(pb);
     }
 
     /* Prepend the header */
@@ -3475,7 +3290,7 @@ static void deliver_read_message(Port* prt, erts_aint32_t state, Eterm to,
     ERL_MESSAGE_TOKEN(mp) = am_undefined;
     erts_queue_message(rp, rp_locks, mp, tuple, prt->common.id);
     if (rp_locks)
-	erts_smp_proc_unlock(rp, rp_locks);
+	erts_proc_unlock(rp, rp_locks);
     if (!scheduler)
 	erts_proc_dec_refc(rp);
 }
@@ -3510,7 +3325,7 @@ static void flush_linebuf_messages(Port *prt, erts_aint32_t state)
     LineBufContext lc;
     int ret;
 
-    ERTS_SMP_LC_ASSERT(!prt || erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(!prt || erts_lc_is_port_locked(prt));
 
     if (!prt)
 	return;
@@ -3554,8 +3369,8 @@ deliver_vec_message(Port* prt,			/* Port */
     erts_aint32_t state;
     int trace_send = IS_TRACED_FL(prt, F_TRACE_SEND);
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_CHK_NO_PROC_LOCKS;
 
     /*
      * Check arguments for validity.
@@ -3646,7 +3461,7 @@ deliver_vec_message(Port* prt,			/* Port */
 
     ERL_MESSAGE_TOKEN(mp) = am_undefined;
     erts_queue_message(rp, rp_locks, mp, tuple, prt->common.id);
-    erts_smp_proc_unlock(rp, rp_locks);
+    erts_proc_unlock(rp, rp_locks);
     if (!scheduler)
 	erts_proc_dec_refc(rp);
 }
@@ -3681,8 +3496,8 @@ static void flush_port(Port *p)
 {
     int fpe_was_unmasked;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(p));
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(p));
 
     if (p->drv_ptr->flush != NULL) {
         ERTS_MSACC_PUSH_STATE_M();
@@ -3714,11 +3529,9 @@ static void flush_port(Port *p)
         if (IS_TRACED_FL(p, F_TRACE_SCHED_PORTS)) {
 	    trace_sched_ports_where(p, am_out, am_flush);
 	}
-#ifdef ERTS_SMP
 	if (p->xports)
 	    erts_port_handle_xports(p);
 	ASSERT(!p->xports);
-#endif
     }
     if ((erts_atomic32_read_nob(&p->state) & ERTS_PORT_SFLGS_DEAD) == 0
 	&& is_port_ioq_empty(p)) {
@@ -3736,11 +3549,12 @@ terminate_port(Port *prt)
     erts_aint32_t state;
     ErtsPrtSD *psd;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     ASSERT(!ERTS_P_LINKS(prt));
     ASSERT(!ERTS_P_MONITORS(prt));
+    ASSERT(!ERTS_P_LT_MONITORS(prt));
 
     /* state may be altered by kill_port() below */
     state = erts_atomic32_read_band_nob(&prt->state,
@@ -3779,11 +3593,9 @@ terminate_port(Port *prt)
 	(*drv->stop)((ErlDrvData)prt->drv_data);
 	erts_unblock_fpe(fpe_was_unmasked);
 	ERTS_MSACC_POP_STATE_M();
-#ifdef ERTS_SMP
 	if (prt->xports)
 	    erts_port_handle_xports(prt);
 	ASSERT(!prt->xports);
-#endif
     }
 
     if (is_internal_port(send_closed_port_id)
@@ -3791,9 +3603,9 @@ terminate_port(Port *prt)
         trace_port_send(prt, connected_id, am_closed, 1);
 
     if(drv->handle != NULL) {
-	erts_smp_rwmtx_rlock(&erts_driver_list_lock);
+	erts_rwmtx_rlock(&erts_driver_list_lock);
 	erts_ddll_decrement_port_count(drv->handle); 
-	erts_smp_rwmtx_runlock(&erts_driver_list_lock);
+	erts_rwmtx_runlock(&erts_driver_list_lock);
     }
     stopq(prt);        /* clear queue memory */
     if(prt->linebuf != NULL){
@@ -3803,7 +3615,7 @@ terminate_port(Port *prt)
 
     erts_cleanup_port_data(prt);
 
-    psd = (ErtsPrtSD *) erts_smp_atomic_read_nob(&prt->psd);
+    psd = (ErtsPrtSD *) erts_atomic_read_nob(&prt->psd);
     if (psd)
 	erts_free(ERTS_ALC_T_PRTSD, psd);
 
@@ -3816,7 +3628,7 @@ terminate_port(Port *prt)
      * port has been removed from the port table (in kill_port()).
      */
     if ((state & ERTS_PORT_SFLG_HALT)
-	&& (erts_smp_atomic32_dec_read_nob(&erts_halt_progress) == 0)) {
+	&& (erts_atomic32_dec_read_nob(&erts_halt_progress) == 0)) {
 	erts_port_release(prt); /* We will exit and never return */
 	erts_flush_async_exit(erts_halt_code, "");
     }
@@ -3830,145 +3642,25 @@ erts_terminate_port(Port *pp)
     terminate_port(pp);
 }
 
-static void port_fire_one_monitor(ErtsMonitor *mon, void *ctx0);
-static void sweep_one_monitor(ErtsMonitor *mon, void *vpsc)
-{
-    switch (mon->type) {
-    case MON_ORIGIN: {
-        ErtsMonitor *rmon;
-        Process *rp;
-
-        ASSERT(is_internal_pid(mon->u.pid));
-        rp = erts_pid2proc(NULL, 0, mon->u.pid, ERTS_PROC_LOCK_LINK);
-        if (!rp) {
-            goto done;
-        }
-        rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), mon->ref);
-        erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-        if (rmon == NULL) {
-            goto done;
-        }
-        erts_destroy_monitor(rmon);
-    } break;
-    case MON_TARGET: {
-        port_fire_one_monitor(mon, vpsc); /* forward call */
-    } break;
-    }
- done:
-    erts_destroy_monitor(mon);
-}
-
-
-
 typedef struct {
-    Port *port;
+    Eterm port_id;
     Eterm reason;
-} SweepContext;
+} ErtsPortExitContext;
 
-static void sweep_one_link(ErtsLink *lnk, void *vpsc)
+static void link_port_exit(ErtsLink *lnk, void *vpectxt)
 {
-    SweepContext *psc = vpsc;
-    DistEntry *dep;
-    Process *rp;
-    Eterm port_id = psc->port->common.id;
-
-    ASSERT(lnk->type == LINK_PID);
-
-    if (IS_TRACED_FL(psc->port, F_TRACE_PORTS))
-        trace_port(psc->port, am_unlink, lnk->pid);
-
-    if (is_external_pid(lnk->pid)) {
-	dep = external_pid_dist_entry(lnk->pid);
-	if(dep != erts_this_dist_entry) {
-	    ErtsDistLinkData dld;
-	    ErtsDSigData dsd;
-	    int code;
-	    code = erts_dsig_prepare(&dsd, dep, NULL, ERTS_DSP_NO_LOCK, 0);
-	    switch (code) {
-	    case ERTS_DSIG_PREP_NOT_ALIVE:
-	    case ERTS_DSIG_PREP_NOT_CONNECTED:
-		break;
-	    case ERTS_DSIG_PREP_CONNECTED:
-		erts_remove_dist_link(&dld, port_id, lnk->pid, dep);
-		erts_destroy_dist_link(&dld);
-		code = erts_dsig_send_exit(&dsd, port_id, lnk->pid,
-					   psc->reason);
-		ASSERT(code == ERTS_DSIG_SEND_OK);
-		break;
-	    default:
-		ASSERT(! "Invalid dsig prepare result");
-		break;
-	    }
-	}
-    } else {
-	ErtsProcLocks rp_locks = ERTS_PROC_LOCK_LINK|ERTS_PROC_LOCKS_XSIG_SEND;
-	ASSERT(is_internal_pid(lnk->pid));
-	rp = erts_pid2proc(NULL, 0, lnk->pid, rp_locks);
-	if (rp) {
-	    ErtsLink *rlnk = erts_remove_link(&ERTS_P_LINKS(rp), port_id);
-
-	    if (rlnk) {
-		int xres = erts_send_exit_signal(NULL,
-						 port_id,
-						 rp,
-						 &rp_locks, 
-						 psc->reason,
-						 NIL,
-						 NULL,
-						 0);
-		if (xres >= 0) {
-                    if (rp_locks & ERTS_PROC_LOCKS_XSIG_SEND) {
-                        erts_smp_proc_unlock(rp, ERTS_PROC_LOCKS_XSIG_SEND);
-                        rp_locks &= ~ERTS_PROC_LOCKS_XSIG_SEND;
-                    }
-		    /* We didn't exit the process and it is traced */
-		    if (IS_TRACED_FL(rp, F_TRACE_PROCS))
-			trace_proc(NULL, 0, rp, am_getting_unlinked, port_id);
-		}
-		erts_destroy_link(rlnk);
-	    }
-
-	    erts_smp_proc_unlock(rp, rp_locks);
-	}
-    }
-    erts_destroy_link(lnk);
+    ErtsPortExitContext *pectxt = vpectxt;
+    erts_proc_sig_send_link_exit(NULL, pectxt->port_id,
+                                 lnk, pectxt->reason, NIL);
 }
 
-static void
-port_fire_one_monitor(ErtsMonitor *mon, void *ctx0)
+static void monitor_port_exit(ErtsMonitor *mon, void *vpectxt)
 {
-    Process      *origin;
-    ErtsProcLocks origin_locks;
-
-    if (mon->type != MON_TARGET || ! is_pid(mon->u.pid)) {
-        return;
-    }
-    /*
-     * Proceed here if someone monitors us, we (port) are the target and
-     * origin is some process
-     */
-    origin_locks = ERTS_PROC_LOCKS_MSG_SEND | ERTS_PROC_LOCK_LINK;
-
-    origin = erts_pid2proc(NULL, 0, mon->u.pid, origin_locks);
-    if (origin) {
-        DeclareTmpHeapNoproc(lhp,3);
-        SweepContext    *ctx = (SweepContext *)ctx0;
-        ErtsMonitor     *rmon;
-        Eterm           watched = (is_atom(mon->name)
-                       ? TUPLE2(lhp, mon->name, erts_this_dist_entry->sysname)
-                       : ctx->port->common.id);
-
-        erts_queue_monitor_message(origin, &origin_locks, mon->ref, am_port,
-                                   watched, ctx->reason);
-        UnUseTmpHeapNoproc(3);
-
-        rmon = erts_remove_monitor(&ERTS_P_MONITORS(origin), mon->ref);
-        erts_smp_proc_unlock(origin, origin_locks);
-
-        if (rmon) {
-            erts_destroy_monitor(rmon);
-        }
-    }
+    ErtsPortExitContext *pectxt = vpectxt;
+    if (erts_monitor_is_target(mon))
+        erts_proc_sig_send_monitor_down(mon, pectxt->reason);
+    else
+        erts_proc_sig_send_demonitor(mon);
 }
 
 /* 'from' is sending 'this_port' an exit signal, (this_port must be internal).
@@ -3985,12 +3677,15 @@ int
 erts_deliver_port_exit(Port *prt, Eterm from, Eterm reason, int send_closed,
                        int drop_normal)
 {
-   ErtsLink *lnk;
+   ErtsLink *links;
+   ErtsMonitor *monitors;
+   ErtsMonitor *lt_monitors;
    Eterm modified_reason;
    erts_aint32_t state, set_state_flags;
+   ErtsPortExitContext pectxt;
 
-   ERTS_SMP_CHK_NO_PROC_LOCKS;
-   ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+   ERTS_CHK_NO_PROC_LOCKS;
+   ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
    modified_reason = (reason == am_kill) ? am_killed : reason;
 
@@ -4035,23 +3730,36 @@ erts_deliver_port_exit(Port *prt, Eterm from, Eterm reason, int send_closed,
 
    set_busy_port(ERTS_Port2ErlDrvPort(prt), 0);
 
+   links = ERTS_P_LINKS(prt);
+   ERTS_P_LINKS(prt) = NULL;
+   monitors = ERTS_P_MONITORS(prt);
+   ERTS_P_MONITORS(prt) = NULL;
+   lt_monitors = ERTS_P_LT_MONITORS(prt);
+   ERTS_P_LT_MONITORS(prt) = NULL;
+
    if (prt->common.u.alive.reg != NULL)
        (void) erts_unregister_name(NULL, 0, prt, prt->common.u.alive.reg->name);
 
-   {
-       SweepContext sc = {prt, modified_reason};
-       lnk = ERTS_P_LINKS(prt);
-       ERTS_P_LINKS(prt) = NULL;
-       erts_sweep_links(lnk, &sweep_one_link, &sc);
+   pectxt.port_id = prt->common.id;
+   pectxt.reason = modified_reason;
+
+   if (links)
+       erts_monitor_tree_foreach_delete(&links,
+                                        link_port_exit,
+                                        (void *) &pectxt);
+
+   if (monitors || lt_monitors) {
+       DRV_MONITOR_LOCK_PDL(prt);
+       if (monitors)
+           erts_monitor_tree_foreach_delete(&monitors,
+                                            monitor_port_exit,
+                                            (void *) &pectxt);
+       if (lt_monitors)
+           erts_monitor_list_foreach_delete(&lt_monitors,
+                                            monitor_port_exit,
+                                            (void *) &pectxt);
+       DRV_MONITOR_UNLOCK_PDL(prt);
    }
-   DRV_MONITOR_LOCK_PDL(prt);
-   {
-       SweepContext ctx = {prt, modified_reason};
-       ErtsMonitor *moni = ERTS_P_MONITORS(prt);
-       ERTS_P_MONITORS(prt) = NULL;
-       erts_sweep_monitors(moni, &sweep_one_monitor, &ctx);
-   } 
-   DRV_MONITOR_UNLOCK_PDL(prt);
 
    if ((state & ERTS_PORT_SFLG_DISTRIBUTION) && prt->dist_entry) {
        erts_do_net_exits(prt->dist_entry, modified_reason);
@@ -4257,17 +3965,9 @@ write_port_control_result(int control_flags,
 	else {
 	    dbin = (ErlDrvBinary *) resp_bufp;
 	    if (dbin->orig_size > ERL_ONHEAP_BIN_LIMIT) {
-		ProcBin* pb = (ProcBin *) *hpp;
+                res = erts_build_proc_bin(ohp, *hpp, ErlDrvBinary2Binary(dbin));
 		*hpp += PROC_BIN_SIZE;
-		pb->thing_word = HEADER_PROC_BIN;
-		pb->size = dbin->orig_size;
-		pb->next = ohp->first;
-		ohp->first = (struct erl_off_heap_header *) pb;
-		pb->val = ErlDrvBinary2Binary(dbin);
-		pb->bytes = (byte*) dbin->orig_bytes;
-		pb->flags = 0;
-		OH_OVERHEAD(ohp, dbin->orig_size / sizeof(Eterm));
-		return make_binary(pb);
+                return res;
 	    }
 	    resp_bufp = dbin->orig_bytes;
 	    resp_size = dbin->orig_size;
@@ -4351,7 +4051,7 @@ port_sig_control(Port *prt,
 				      prt);
 
 	    if (rp_locks)
-		erts_smp_proc_unlock(rp, rp_locks);
+		erts_proc_unlock(rp, rp_locks);
 	    goto done;
 	}
     }
@@ -4404,7 +4104,7 @@ erts_port_control(Process* c_p,
     int copy;
     ErtsProc2PortSigData *sigdp;
 
-    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
     if (sched_flags & ERTS_PTS_FLG_EXIT)
 	return ERTS_PORT_OP_BADARG;
 
@@ -4716,11 +4416,11 @@ port_sig_call(Port *prt,
 					      prt);
 
 		    if (rp_locks)
-			erts_smp_proc_unlock(rp, rp_locks);
+			erts_proc_unlock(rp, rp_locks);
 		    goto done;
 		}
 		if (rp_locks)
-		    erts_smp_proc_unlock(rp, rp_locks);
+		    erts_proc_unlock(rp, rp_locks);
 	    }
 	}
     }
@@ -4754,7 +4454,7 @@ erts_port_call(Process* c_p,
     erts_aint32_t sched_flags;
     ErtsProc2PortSigData *sigdp;
 
-    sched_flags = erts_smp_atomic32_read_nob(&prt->sched.flags);
+    sched_flags = erts_atomic32_read_nob(&prt->sched.flags);
     if (sched_flags & ERTS_PTS_FLG_EXIT) {
 	return ERTS_PORT_OP_BADARG;
     }
@@ -4972,7 +4672,7 @@ port_sig_info(Port *prt,
 				      prt);
 	}
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
     }
     return ERTS_PORT_REDS_INFO;
 }
@@ -5041,7 +4741,7 @@ typedef struct {
     Uint sched_id;
     Eterm pid;
     Uint32 refn[ERTS_REF_NUMBERS];
-    erts_smp_atomic32_t refc;
+    erts_atomic32_t refc;
 } ErtsIOBytesReq;
 
 static void
@@ -5091,10 +4791,10 @@ reply_io_bytes(void *vreq)
 	if (req->sched_id == sched_id)
 	    rp_locks &= ~ERTS_PROC_LOCK_MAIN;
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
     }
 
-    if (erts_smp_atomic32_dec_read_nob(&req->refc) == 0)
+    if (erts_atomic32_dec_read_nob(&req->refc) == 0)
 	erts_free(ERTS_ALC_T_IOB_REQ, req);
 }
 
@@ -5117,16 +4817,14 @@ erts_request_io_bytes(Process *c_p)
     req->refn[0] = refn[0];
     req->refn[1] = refn[1];
     req->refn[2] = refn[2];
-    erts_smp_atomic32_init_nob(&req->refc,
+    erts_atomic32_init_nob(&req->refc,
 			       (erts_aint32_t) erts_no_schedulers);
 
-#ifdef ERTS_SMP
     if (erts_no_schedulers > 1)
 	erts_schedule_multi_misc_aux_work(1,
 					  erts_no_schedulers,
 					  reply_io_bytes,
 					  (void *) req);
-#endif
 
     reply_io_bytes((void *) req);
 
@@ -5141,14 +4839,18 @@ typedef struct {
 
 static void prt_one_monitor(ErtsMonitor *mon, void *vprtd)
 {
+    ErtsMonitorData *mdp = erts_monitor_to_data(mon);
     prt_one_lnk_data *prtd = (prt_one_lnk_data *) vprtd;
-    erts_print(prtd->to, prtd->arg, "(%T,%T)", mon->u.pid, mon->ref);
+    if (mon->type == ERTS_MON_TYPE_RESOURCE && erts_monitor_is_target(mon))
+        erts_print(prtd->to, prtd->arg, "(%p,%T)", mon->other.ptr, mdp->ref);
+    else
+        erts_print(prtd->to, prtd->arg, "(%T,%T)", mon->other.item, mdp->ref);
 }
 
 static void prt_one_lnk(ErtsLink *lnk, void *vprtd)
 {
     prt_one_lnk_data *prtd = (prt_one_lnk_data *) vprtd;
-    erts_print(prtd->to, prtd->arg, "%T", lnk->pid);
+    erts_print(prtd->to, prtd->arg, "%T", lnk->other.item);
 }
 
 static void dump_port_state(fmtfn_t to, void *arg, erts_aint32_t state)
@@ -5197,7 +4899,7 @@ static void dump_port_state(fmtfn_t to, void *arg, erts_aint32_t state)
 
 static void dump_port_task_flags(fmtfn_t to, void *arg, Port* p)
 {
-    erts_aint32_t flags = erts_smp_atomic32_read_nob(&p->sched.flags);
+    erts_aint32_t flags = erts_atomic32_read_nob(&p->sched.flags);
     erts_aint32_t unknown = 0;
     char delim = ' ';
 
@@ -5260,15 +4962,18 @@ print_port_info(Port *p, fmtfn_t to, void *arg)
 	prtd.to = to;
 	prtd.arg = arg;
 	erts_print(to, arg, "Links: ");
-	erts_doforall_links(ERTS_P_LINKS(p), &prt_one_lnk, &prtd);
+	erts_link_tree_foreach(ERTS_P_LINKS(p), prt_one_lnk, (void *) &prtd);
 	erts_print(to, arg, "\n");
     }
-    if (ERTS_P_MONITORS(p)) {
+    if (ERTS_P_MONITORS(p) || ERTS_P_LT_MONITORS(p)) {
 	prt_one_lnk_data prtd;
 	prtd.to = to;
 	prtd.arg = arg;
 	erts_print(to, arg, "Monitors: ");
-	erts_doforall_monitors(ERTS_P_MONITORS(p), &prt_one_monitor, &prtd);
+	erts_monitor_tree_foreach(ERTS_P_MONITORS(p), prt_one_monitor,
+                                  (void *) &prtd);
+	erts_monitor_list_foreach(ERTS_P_LT_MONITORS(p), prt_one_monitor,
+                                  (void *) &prtd);
 	erts_print(to, arg, "\n");
     }
     if (p->suspended) {
@@ -5312,14 +5017,14 @@ set_busy_port(ErlDrvPort dprt, int on)
     DTRACE_CHARBUF(port_str, 16);
 #endif
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     prt = erts_drvport2port(dprt);
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return;
 
     if (on) {
-	flags = erts_smp_atomic32_read_bor_acqb(&prt->sched.flags,
+	flags = erts_atomic32_read_bor_acqb(&prt->sched.flags,
 						ERTS_PTS_FLG_BUSY_PORT);
 	if (flags & ERTS_PTS_FLG_BUSY_PORT)
 	    return; /* Already busy */
@@ -5335,7 +5040,7 @@ set_busy_port(ErlDrvPort dprt, int on)
         }
 #endif
     } else {
-	flags = erts_smp_atomic32_read_band_acqb(&prt->sched.flags,
+	flags = erts_atomic32_read_band_acqb(&prt->sched.flags,
 						 ~ERTS_PTS_FLG_BUSY_PORT);
 	if (!(flags & ERTS_PTS_FLG_BUSY_PORT))
 	    return; /* Already non-busy */
@@ -5429,7 +5134,7 @@ int get_port_flags(ErlDrvPort ix)
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return 0;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     flags = 0;
     if (state & ERTS_PORT_SFLG_BINARY_IO)
@@ -5445,8 +5150,8 @@ void erts_raw_port_command(Port* p, byte* buf, Uint len)
     int fpe_was_unmasked;
     ERTS_MSACC_PUSH_STATE_M();
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(p));
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(p));
 
     if (len > (Uint) INT_MAX)
 	erts_exit(ERTS_ABORT_EXIT,
@@ -5475,10 +5180,10 @@ int async_ready(Port *p, void* data)
 {
     int need_free = 1;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (p) {
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(p));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(p));
 	if (p->drv_ptr->ready_async != NULL) {
 	    ERTS_MSACC_PUSH_AND_SET_STATE_M(ERTS_MSACC_STATE_PORT);
 #ifdef USE_VM_PROBES
@@ -5547,24 +5252,17 @@ erts_stale_drv_select(Eterm port,
     switch (mode) {
     case ERL_DRV_READ | ERL_DRV_WRITE:
 	type = "Input/Output";
-	goto deselect;
     case ERL_DRV_WRITE:
 	type = "Output";
-	goto deselect;
     case ERL_DRV_READ:
 	type = "Input";
-    deselect:
-	if (deselect) {
-	    driver_select(drv_port, hndl,
-			  mode | ERL_DRV_USE_NO_CALLBACK,
-			  0);
-	}
-	break;
     default:
-	type = "Event";
-	if (deselect)
-	    driver_event(drv_port, hndl, NULL);
-	break;
+        type = "";
+    }
+    if (deselect) {
+        driver_select(drv_port, hndl,
+                      mode | ERL_DRV_USE_NO_CALLBACK,
+                      0);
     }
 
     dsbufp = erts_create_logger_dsbuf();
@@ -5663,8 +5361,8 @@ void driver_report_exit(ErlDrvPort ix, int status)
    if (prt == ERTS_INVALID_ERL_DRV_PORT)
        return;
 
-   ERTS_SMP_CHK_NO_PROC_LOCKS;
-   ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+   ERTS_CHK_NO_PROC_LOCKS;
+   ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
    pid = ERTS_PORT_GET_CONNECTED(prt);
    ASSERT(is_internal_pid(pid));
@@ -5687,7 +5385,7 @@ void driver_report_exit(ErlDrvPort ix, int status)
    ERL_MESSAGE_TOKEN(mp) = am_undefined;
    erts_queue_message(rp, rp_locks, mp, tuple, prt->common.id);
 
-   erts_smp_proc_unlock(rp, rp_locks);
+   erts_proc_unlock(rp, rp_locks);
    if (!scheduler)
        erts_proc_dec_refc(rp);
 }
@@ -6141,21 +5839,12 @@ driver_deliver_term(Port *prt, Eterm to, ErlDrvTermData* data, int len)
 		mess = make_binary(hbp);
 	    }
 	    else {
-		ProcBin* pbp;
+		Eterm* hp;
 		Binary* bp = erts_bin_nrml_alloc(size);
 		ASSERT(bufp);
 		sys_memcpy((void *) bp->orig_bytes, (void *) bufp, size);
-		pbp = (ProcBin *) erts_produce_heap(&factory,
-						    PROC_BIN_SIZE, HEAP_EXTRA);
-		pbp->thing_word = HEADER_PROC_BIN;
-		pbp->size = size;
-		pbp->next = factory.off_heap->first;
-		factory.off_heap->first = (struct erl_off_heap_header*)pbp;
-		pbp->val = bp;
-		pbp->bytes = (byte*) bp->orig_bytes;
-		pbp->flags = 0;
-		OH_OVERHEAD(factory.off_heap, pbp->size / sizeof(Eterm));
-		mess = make_binary(pbp);
+                hp = erts_produce_heap(&factory, PROC_BIN_SIZE, HEAP_EXTRA);
+                mess = erts_build_proc_bin(factory.off_heap, hp, bp);
 	    }
 	    ptr += 2;
 	    break;
@@ -6327,7 +6016,7 @@ driver_deliver_term(Port *prt, Eterm to, ErlDrvTermData* data, int len)
     }
     if (rp) {
 	if (rp_locks)
-	    erts_smp_proc_unlock(rp, rp_locks);
+	    erts_proc_unlock(rp, rp_locks);
 	if (!scheduler)
 	    erts_proc_dec_refc(rp);
     }
@@ -6342,9 +6031,7 @@ static ERTS_INLINE int
 deliver_term_check_port(ErlDrvTermData port_id, Eterm *connected_p,
                         Port **trace_prt)
 {
-#ifdef ERTS_SMP
     ErtsThrPrgrDelayHandle dhndl = erts_thr_progress_unmanaged_delay();
-#endif
     erts_aint32_t state;
     int res = 1;
     Port *prt = erts_port_lookup_raw((Eterm) port_id);
@@ -6362,24 +6049,20 @@ deliver_term_check_port(ErlDrvTermData port_id, Eterm *connected_p,
 	goto done;
     }
     if (connected_p) {
-#ifdef ERTS_SMP
 	if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED)
 	    ETHR_MEMBAR(ETHR_LoadLoad);
-#endif
 	*connected_p = ERTS_PORT_GET_CONNECTED(prt);
     }
 
 done:
 
-#ifdef ERTS_SMP
     if (dhndl != ERTS_THR_PRGR_DHANDLE_MANAGED) {
-	ERTS_SMP_LC_ASSERT(!prt || !erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(!prt || !erts_lc_is_port_locked(prt));
 	erts_thr_progress_unmanaged_continue(dhndl);
 	ETHR_MEMBAR(ETHR_LoadLoad|ETHR_LoadStore);
     } else
-#endif
     if (res == 1) {
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
         *trace_prt = prt;
     }
     return res;
@@ -6407,13 +6090,13 @@ driver_output_term(ErlDrvPort drvport, ErlDrvTermData* data, int len)
     erts_aint32_t state;
     Port* prt;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     /* NOTE! It *not* safe to access 'drvport' from unmanaged threads. */
     prt = erts_drvport2port_state(drvport, &state);
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1; /* invalid (dead) */
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_CHK_NO_PROC_LOCKS;
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     if (state & ERTS_PORT_SFLG_CLOSING)
 	return 0;
 
@@ -6450,16 +6133,14 @@ driver_send_term(ErlDrvPort drvport,
      * internal data representation for ErlDrvPort.
      */
     Port* prt = NULL;
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
-#ifdef ERTS_SMP
+    ERTS_CHK_NO_PROC_LOCKS;
     if (erts_thr_progress_is_managed_thread())
-#endif
     {
 	erts_aint32_t state;
 	prt = erts_drvport2port_state(drvport, &state);
 	if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	    return -1; /* invalid (dead) */
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 	if (state & ERTS_PORT_SFLG_CLOSING)
 	    return 0;
     }
@@ -6479,11 +6160,11 @@ int driver_output_binary(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     Port* prt = erts_drvport2port_state(ix, &state);
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     if (state & ERTS_PORT_SFLG_CLOSING)
 	return 0;
 
@@ -6493,6 +6174,7 @@ int driver_output_binary(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     else
 	erts_atomic64_add_nob(&bytes_in, (erts_aint64_t) (hlen + len));
     if (state & ERTS_PORT_SFLG_DISTRIBUTION) {
+        erts_atomic64_inc_nob(&prt->dist_entry->in);
 	return erts_net_message(prt,
 				prt->dist_entry,
 				(byte*) hbuf, hlen,
@@ -6518,12 +6200,12 @@ int driver_output2(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     Port* prt = erts_drvport2port_state(ix, &state);
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     if (state & ERTS_PORT_SFLG_CLOSING)
 	return 0;
     
@@ -6533,6 +6215,7 @@ int driver_output2(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     else
 	erts_atomic64_add_nob(&bytes_in, (erts_aint64_t) (hlen + len));
     if (state & ERTS_PORT_SFLG_DISTRIBUTION) {
+        erts_atomic64_inc_nob(&prt->dist_entry->in);
 	if (len == 0)
 	    return erts_net_message(prt,
 				    prt->dist_entry,
@@ -6557,7 +6240,7 @@ int driver_output2(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
 
 int driver_output(ErlDrvPort ix, char* buf, ErlDrvSizeT len)
 {
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     return driver_output2(ix, NULL, 0, buf, len);
 }
 
@@ -6573,7 +6256,7 @@ int driver_outputv(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     erts_aint32_t state;
     ErtsSchedulerData *esdp = erts_get_scheduler_data();
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     ASSERT(vec->size >= skip);
     if (vec->size <= skip)
@@ -6584,7 +6267,7 @@ int driver_outputv(ErlDrvPort ix, char* hbuf, ErlDrvSizeT hlen,
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     if (state & ERTS_PORT_SFLG_CLOSING)
 	return 0;
@@ -6798,7 +6481,6 @@ static ERTS_INLINE void pdl_destroy(ErlDrvPDL pdl)
     erts_free(ERTS_ALC_T_PORT_DATA_LOCK, pdl);
 }
 
-#ifdef ERTS_SMP
 
 static void driver_monitor_lock_pdl(Port *p) {
     if (p->port_data_lock) {
@@ -6807,7 +6489,7 @@ static void driver_monitor_lock_pdl(Port *p) {
     /* Now we either have the port lock or the port_data_lock */
     ERTS_LC_ASSERT(!p->port_data_lock
 		   || erts_lc_mtx_is_locked(&(p->port_data_lock->mtx)));
-    ERTS_SMP_LC_ASSERT(p->port_data_lock
+    ERTS_LC_ASSERT(p->port_data_lock
 		       || erts_lc_is_port_locked(p));
 }
 
@@ -6815,14 +6497,13 @@ static void driver_monitor_unlock_pdl(Port *p) {
     /* We should either have the port lock or the port_data_lock */
     ERTS_LC_ASSERT(!p->port_data_lock
 		   || erts_lc_mtx_is_locked(&(p->port_data_lock->mtx)));
-    ERTS_SMP_LC_ASSERT(p->port_data_lock
+    ERTS_LC_ASSERT(p->port_data_lock
 		       || erts_lc_is_port_locked(p));
     if (p->port_data_lock) {
 	driver_pdl_unlock(p->port_data_lock);
     }
 }
 
-#endif
 
 /*
  * exported driver_pdl_* functions ...
@@ -7025,7 +6706,7 @@ int driver_set_timer(ErlDrvPort ix, unsigned long t)
 {
     Port* prt = erts_drvport2port(ix);
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
@@ -7042,7 +6723,7 @@ int driver_cancel_timer(ErlDrvPort ix)
     Port* prt = erts_drvport2port(ix);
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     erts_cancel_port_timer(prt);
     return 0;
 }
@@ -7053,11 +6734,11 @@ driver_read_timer(ErlDrvPort ix, unsigned long* t)
     Port* prt = erts_drvport2port(ix);
     Sint64 left;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     left = erts_read_port_timer(prt);
     if (left < 0)
@@ -7072,7 +6753,7 @@ int
 driver_get_now(ErlDrvNowData *now_data)
 {
     Uint mega,secs,micro;
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (now_data == NULL) {
 	return -1;
@@ -7129,24 +6810,23 @@ static int do_driver_monitor_process(Port *prt,
 				     ErlDrvMonitor *monitor)
 {
     Eterm buf[ERTS_REF_THING_SIZE];
-    Process *rp;
     Eterm ref;
+    ErtsMonitorData *mdp;
 
-    if (prt->drv_ptr->process_exit == NULL) {
+    if (!prt->drv_ptr->process_exit)
 	return -1;
-    }
-    rp = erts_pid2proc_opt(NULL, 0,
-			   (Eterm) process, ERTS_PROC_LOCK_LINK,
-			   ERTS_P2P_FLG_ALLOW_OTHER_X);
-    if (!rp) {
-	return 1;
-    }
 
     ref = erts_make_ref_in_buffer(buf);
-    erts_add_monitor(&ERTS_P_MONITORS(prt), MON_ORIGIN, ref, rp->common.id, NIL);
-    erts_add_monitor(&ERTS_P_MONITORS(rp), MON_TARGET, ref, prt->common.id, NIL);
+    mdp = erts_monitor_create(ERTS_MON_TYPE_PORT, ref,
+                              prt->common.id, process, NIL);
+
+    if (!erts_proc_sig_send_monitor(&mdp->target, process)) {
+        erts_monitor_release_both(mdp);
+        return 1;
+    }
+
+    erts_monitor_tree_insert(&ERTS_P_MONITORS(prt), &mdp->origin);
 
-    erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
     erts_ref_to_driver_monitor(ref,monitor);
     return 0;
 }
@@ -7160,7 +6840,7 @@ int driver_monitor_process(ErlDrvPort drvport,
 {
     Port *prt;
     int ret;
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     ErtsSchedulerData *sched = erts_get_scheduler_data();
 #endif
 
@@ -7170,7 +6850,7 @@ int driver_monitor_process(ErlDrvPort drvport,
 
     /* Now (in SMP) we should have either the port lock (if we have a scheduler) or the port data lock
        (if we're a driver thread) */
-    ERTS_SMP_LC_ASSERT((sched != NULL || prt->port_data_lock));
+    ERTS_LC_ASSERT((sched != NULL || prt->port_data_lock));
     ret = do_driver_monitor_process(prt,process,monitor);
     DRV_MONITOR_UNLOCK_PDL(prt);
     return ret;
@@ -7179,37 +6859,17 @@ int driver_monitor_process(ErlDrvPort drvport,
 static int do_driver_demonitor_process(Port *prt, const ErlDrvMonitor *monitor)
 {
     Eterm heap[ERTS_REF_THING_SIZE];
-    Process *rp;
     Eterm ref;
     ErtsMonitor *mon;
-    Eterm to;
 
     ref = erts_driver_monitor_to_ref(heap, monitor);
 
-    mon = erts_lookup_monitor(ERTS_P_MONITORS(prt), ref);
-    if (mon == NULL) {
-	return 1;
-    }
-    ASSERT(mon->type == MON_ORIGIN);
-    to = mon->u.pid;
-    ASSERT(is_internal_pid(to));
-    rp = erts_pid2proc_opt(NULL,
-			   0,
-			   to,
-			   ERTS_PROC_LOCK_LINK,
-			   ERTS_P2P_FLG_ALLOW_OTHER_X);
-    mon = erts_remove_monitor(&ERTS_P_MONITORS(prt), ref);
-    if (mon) {
-	erts_destroy_monitor(mon);
-    }
-    if (rp) {
-	ErtsMonitor *rmon;
-	rmon = erts_remove_monitor(&ERTS_P_MONITORS(rp), ref);
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-	if (rmon != NULL) {
-	    erts_destroy_monitor(rmon);
-	}
-    } 
+    mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(prt), ref);
+    if (!mon || !erts_monitor_is_origin(mon))
+        return 1;
+
+    erts_monitor_tree_delete(&ERTS_P_MONITORS(prt), mon);
+    erts_proc_sig_send_demonitor(mon);
     return 0;
 }
 
@@ -7218,7 +6878,7 @@ int driver_demonitor_process(ErlDrvPort drvport,
 {
     Port *prt;
     int ret;
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     ErtsSchedulerData *sched = erts_get_scheduler_data();
 #endif
 
@@ -7228,7 +6888,7 @@ int driver_demonitor_process(ErlDrvPort drvport,
 
     /* Now we should have either the port lock (if we have a scheduler) or the port data lock
        (if we're a driver thread) */
-    ERTS_SMP_LC_ASSERT((sched != NULL || prt->port_data_lock));
+    ERTS_LC_ASSERT((sched != NULL || prt->port_data_lock));
     ret = do_driver_demonitor_process(prt,monitor);
     DRV_MONITOR_UNLOCK_PDL(prt);
     return ret;
@@ -7238,19 +6898,16 @@ static ErlDrvTermData do_driver_get_monitored_process(Port *prt,const ErlDrvMoni
 {
     Eterm ref;
     ErtsMonitor *mon;
-    Eterm to;
     Eterm heap[ERTS_REF_THING_SIZE];
 
     ref = erts_driver_monitor_to_ref(heap, monitor);
 
-    mon = erts_lookup_monitor(ERTS_P_MONITORS(prt), ref);
-    if (mon == NULL) {
+    mon = erts_monitor_tree_lookup(ERTS_P_MONITORS(prt), ref);
+    if (!mon || !erts_monitor_is_origin(mon))
 	return driver_term_nil;
-    }
-    ASSERT(mon->type == MON_ORIGIN);
-    to = mon->u.pid;
-    ASSERT(is_internal_pid(to));
-    return (ErlDrvTermData) to;
+
+    ASSERT(is_internal_pid(mon->other.item));
+    return (ErlDrvTermData) mon->other.item;
 }
 
 
@@ -7259,7 +6916,7 @@ ErlDrvTermData driver_get_monitored_process(ErlDrvPort drvport,
 {
     Port *prt;
     ErlDrvTermData ret;
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     ErtsSchedulerData *sched = erts_get_scheduler_data();
 #endif
 
@@ -7269,7 +6926,7 @@ ErlDrvTermData driver_get_monitored_process(ErlDrvPort drvport,
 
     /* Now we should have either the port lock (if we have a scheduler) or the port data lock
        (if we're a driver thread) */
-    ERTS_SMP_LC_ASSERT((sched != NULL || prt->port_data_lock));
+    ERTS_LC_ASSERT((sched != NULL || prt->port_data_lock));
     ret = do_driver_get_monitored_process(prt,monitor);
     DRV_MONITOR_UNLOCK_PDL(prt);
     return ret;
@@ -7282,24 +6939,27 @@ int driver_compare_monitors(const ErlDrvMonitor *monitor1,
                       ERTS_REF_THING_SIZE*sizeof(Eterm));
 }
 
-void erts_fire_port_monitor(Port *prt, Eterm ref)
+void erts_fire_port_monitor(Port *prt, ErtsMonitor *tmon)
 {
-    ErtsMonitor *rmon;
+    ErtsMonitorData *mdp;
     void (*callback)(ErlDrvData drv_data, ErlDrvMonitor *monitor);
     ErlDrvMonitor drv_monitor;
     int fpe_was_unmasked;
     ERTS_MSACC_PUSH_STATE_M();
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
-    ASSERT(prt->drv_ptr != NULL);    
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ASSERT(prt->drv_ptr != NULL);
+    ASSERT(erts_monitor_is_target(tmon));
+    mdp = erts_monitor_to_data(tmon);
     DRV_MONITOR_LOCK_PDL(prt);
-    if (erts_lookup_monitor(ERTS_P_MONITORS(prt), ref) == NULL) {
+    if (!erts_monitor_is_in_table(&mdp->origin)) {
 	DRV_MONITOR_UNLOCK_PDL(prt);
+        erts_monitor_release(tmon);
 	return;
     }
     callback = prt->drv_ptr->process_exit;
     ASSERT(callback != NULL);
-    erts_ref_to_driver_monitor(ref,&drv_monitor);
+    erts_ref_to_driver_monitor(mdp->ref,&drv_monitor);
     ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_PORT);
     DRV_MONITOR_UNLOCK_PDL(prt);
 #ifdef USE_VM_PROBES
@@ -7323,11 +6983,9 @@ void erts_fire_port_monitor(Port *prt, Eterm ref)
     DRV_MONITOR_LOCK_PDL(prt);
     ERTS_MSACC_POP_STATE_M();
     /* remove monitor *after* callback */
-    rmon = erts_remove_monitor(&ERTS_P_MONITORS(prt), ref);
+    erts_monitor_tree_delete(&ERTS_P_MONITORS(prt), &mdp->origin);
     DRV_MONITOR_UNLOCK_PDL(prt);
-    if (rmon) {
-	erts_destroy_monitor(rmon);
-    }
+    erts_monitor_release_both(mdp);
 }
 
 
@@ -7337,11 +6995,11 @@ driver_failure_term(ErlDrvPort ix, Eterm term, int eof)
     erts_aint32_t state;
     Port* prt = erts_drvport2port_state(ix, &state);
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
     if (prt->async_open_port)
         init_ack_send_reply(prt, prt->common.id);
@@ -7372,34 +7030,19 @@ driver_failure_term(ErlDrvPort ix, Eterm term, int eof)
 int driver_exit(ErlDrvPort ix, int err)
 {
     Port* prt = erts_drvport2port(ix);
-    Process* rp;
-    ErtsLink *lnk, *rlnk = NULL;
+    ErtsLink *lnk;
     Eterm connected;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
   
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
         return -1;
 
     connected = ERTS_PORT_GET_CONNECTED(prt);
-    rp = erts_pid2proc(NULL, 0, connected, ERTS_PROC_LOCK_LINK);
-    if (rp) {
-	rlnk = erts_remove_link(&ERTS_P_LINKS(rp),prt->common.id);
-    }
-
-    lnk = erts_remove_link(&ERTS_P_LINKS(prt), connected);
-
-#ifdef ERTS_SMP
-    if (rp)
-	erts_smp_proc_unlock(rp, ERTS_PROC_LOCK_LINK);
-#endif
-
-    if (rlnk != NULL) {
-	erts_destroy_link(rlnk);
-    }
-
-    if (lnk != NULL) {
-	erts_destroy_link(lnk);
+    lnk = erts_link_tree_lookup(ERTS_P_LINKS(prt), connected);
+    if (lnk) {
+        erts_link_tree_delete(&ERTS_P_LINKS(prt), lnk);
+        erts_proc_sig_send_unlink(NULL, lnk);
     }
 
     if (err == 0)
@@ -7422,7 +7065,7 @@ int driver_failure_atom(ErlDrvPort ix, char* string)
 {
     return driver_failure_term(ix,
 			       erts_atom_put((byte *) string,
-					     strlen(string),
+					     sys_strlen(string),
 					     ERTS_ATOM_ENC_LATIN1,
 					     1),
 			       0);
@@ -7446,7 +7089,7 @@ ErlDrvTermData driver_mk_atom(char* string)
 			     sys_strlen(string),
 			     ERTS_ATOM_ENC_LATIN1,
 			     1);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     return (ErlDrvTermData) am;
 }
 
@@ -7455,27 +7098,27 @@ ErlDrvTermData driver_mk_port(ErlDrvPort ix)
     Port* prt = erts_drvport2port(ix);
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return (ErlDrvTermData) NIL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     return (ErlDrvTermData) prt->common.id;
 }
 
 ErlDrvTermData driver_connected(ErlDrvPort ix)
 {
     Port* prt = erts_drvport2port(ix);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return NIL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     return ERTS_PORT_GET_CONNECTED(prt);
 }
 
 ErlDrvTermData driver_caller(ErlDrvPort ix)
 {
     Port* prt = erts_drvport2port(ix);
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return NIL;
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     return prt->caller;
 }
 
@@ -7484,20 +7127,20 @@ int driver_lock_driver(ErlDrvPort ix)
     Port* prt = erts_drvport2port(ix);
     DE_Handle* dh;
 
-    ERTS_SMP_CHK_NO_PROC_LOCKS;
+    ERTS_CHK_NO_PROC_LOCKS;
 
     if (prt == ERTS_INVALID_ERL_DRV_PORT)
 	return -1;
 
-    erts_smp_rwmtx_rwlock(&erts_driver_list_lock);
+    erts_rwmtx_rwlock(&erts_driver_list_lock);
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
     if ((dh = (DE_Handle*)prt->drv_ptr->handle ) == NULL) {
-	erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+	erts_rwmtx_rwunlock(&erts_driver_list_lock);
 	return -1;
     }
     erts_ddll_lock_driver(dh, prt->drv_ptr->name);
-    erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+    erts_rwmtx_rwunlock(&erts_driver_list_lock);
     return 0;
 }
 
@@ -7505,9 +7148,9 @@ int driver_lock_driver(ErlDrvPort ix)
 static int maybe_lock_driver_list(void) 
 {
     void *rec_lock;
-    rec_lock = erts_smp_tsd_get(driver_list_lock_status_key);
+    rec_lock = erts_tsd_get(driver_list_lock_status_key);
     if (rec_lock == 0) {
-	erts_smp_rwmtx_rwlock(&erts_driver_list_lock);
+	erts_rwmtx_rwlock(&erts_driver_list_lock);
 	return 1;
     }
     return 0;
@@ -7515,7 +7158,7 @@ static int maybe_lock_driver_list(void)
 static void maybe_unlock_driver_list(int doit)
 {
     if (doit) {
-	erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+	erts_rwmtx_rwunlock(&erts_driver_list_lock);
     }
 }
 /* 
@@ -7538,7 +7181,7 @@ void *driver_dl_open(char * path)
 {
     void *ptr;
     int res;
-    int *last_error_p = erts_smp_tsd_get(driver_list_last_error_key);
+    int *last_error_p = erts_tsd_get(driver_list_last_error_key);
     int locked = maybe_lock_driver_list();
     if ((res = erts_sys_ddll_open(path, &ptr, NULL)) == 0) {
 	maybe_unlock_driver_list(locked);
@@ -7546,7 +7189,7 @@ void *driver_dl_open(char * path)
     } else {
 	if (!last_error_p) {
 	    last_error_p = erts_alloc(ERTS_ALC_T_DDLL_ERRCODES, sizeof(int));
-	    erts_smp_tsd_set(driver_list_last_error_key,last_error_p);
+	    erts_tsd_set(driver_list_last_error_key,last_error_p);
 	}
 	*last_error_p = res;
 	maybe_unlock_driver_list(locked);
@@ -7558,7 +7201,7 @@ void *driver_dl_sym(void * handle, char *func_name)
 {
     void *ptr;
     int res;
-    int *last_error_p = erts_smp_tsd_get(driver_list_lock_status_key);
+    int *last_error_p = erts_tsd_get(driver_list_lock_status_key);
     int locked = maybe_lock_driver_list();
     if ((res = erts_sys_ddll_sym(handle, func_name, &ptr)) == 0) {
 	maybe_unlock_driver_list(locked);
@@ -7566,7 +7209,7 @@ void *driver_dl_sym(void * handle, char *func_name)
     } else {
 	if (!last_error_p) {
 	    last_error_p = erts_alloc(ERTS_ALC_T_DDLL_ERRCODES, sizeof(int));
-	    erts_smp_tsd_set(driver_list_lock_status_key,last_error_p);
+	    erts_tsd_set(driver_list_lock_status_key,last_error_p);
 	}
 	*last_error_p = res;
 	maybe_unlock_driver_list(locked);
@@ -7586,7 +7229,7 @@ int driver_dl_close(void *handle)
 char *driver_dl_error(void) 
 {
     char *res;
-    int *last_error_p = erts_smp_tsd_get(driver_list_lock_status_key);
+    int *last_error_p = erts_tsd_get(driver_list_lock_status_key);
     int locked = maybe_lock_driver_list();
     res = erts_ddll_error((last_error_p != NULL) ? (*last_error_p) : ERL_DE_ERROR_UNSPECIFIED); 
     maybe_unlock_driver_list(locked);
@@ -7624,20 +7267,8 @@ driver_system_info(ErlDrvSysInfo *sip, size_t si_size)
 	sip->driver_minor_version = ERL_DRV_EXTENDED_MINOR_VERSION;
 	sip->erts_version = ERLANG_VERSION;
 	sip->otp_release = ERLANG_OTP_RELEASE;
-	sip->thread_support = 
-#ifdef USE_THREADS
-	    1
-#else
-	    0
-#endif
-	    ;
-	sip->smp_support = 
-#ifdef ERTS_SMP
-	    1
-#else
-	    0
-#endif
-	    ;
+	sip->thread_support = 1;
+	sip->smp_support = 1;
 
     }
 
@@ -7663,11 +7294,7 @@ driver_system_info(ErlDrvSysInfo *sip, size_t si_size)
      */
     if (si_size >= ERL_DRV_SYS_INFO_SIZE(dirty_scheduler_support)) {
 	sip->dirty_scheduler_support =
-#ifdef ERTS_DIRTY_SCHEDULERS
 	    1
-#else
-	    0
-#endif
 	    ;
     }
 
@@ -7694,14 +7321,6 @@ no_output_callback(ErlDrvData drv_data, char *buf, ErlDrvSizeT len)
 }
 
 static void
-no_event_callback(ErlDrvData drv_data, ErlDrvEvent event, ErlDrvEventData event_data)
-{
-    Port *prt = get_current_port();
-    report_missing_drv_callback(prt, "Event", "event()");
-    driver_event(ERTS_Port2ErlDrvPort(prt), event, NULL);
-}
-
-static void
 no_ready_input_callback(ErlDrvData drv_data, ErlDrvEvent event)
 {
     Port *prt = get_current_port();
@@ -7735,31 +7354,31 @@ no_stop_select_callback(ErlDrvEvent event, void* private)
 }
 
 #define IS_DRIVER_VERSION_GE(DE,MAJOR,MINOR) \
-    ((DE)->major_version >= (MAJOR) && (DE)->minor_version >= (MINOR))
+    ((DE)->major_version > (MAJOR) ||        \
+     ((DE)->major_version == (MAJOR) && (DE)->minor_version >= (MINOR)))
 
 static int
 init_driver(erts_driver_t *drv, ErlDrvEntry *de, DE_Handle *handle)
 {
+    drv->name_atom = erts_atom_put((byte*)de->driver_name,
+                                   sys_strlen(de->driver_name),
+                                   ERTS_ATOM_ENC_LATIN1, 1);
     drv->name = de->driver_name;
+
     ASSERT(de->extended_marker == ERL_DRV_EXTENDED_MARKER);
     ASSERT(de->major_version >= 2);
     drv->version.major = de->major_version;
     drv->version.minor = de->minor_version;
     drv->flags = de->driver_flags;
     drv->handle = handle;
-#ifdef ERTS_SMP
     if (drv->flags & ERL_DRV_FLAG_USE_PORT_LOCKING) {
         drv->lock = NULL;
     } else {
-        Eterm driver_id = erts_atom_put((byte *) drv->name,
-                                        sys_strlen(drv->name),
-                                        ERTS_ATOM_ENC_LATIN1, 1);
-
         drv->lock = erts_alloc(ERTS_ALC_T_DRIVER_LOCK, sizeof(erts_mtx_t));
 
-        erts_mtx_init(drv->lock, "driver_lock", driver_id, ERTS_LOCK_FLAGS_CATEGORY_IO);
+        erts_mtx_init(drv->lock, "driver_lock", drv->name_atom,
+                      ERTS_LOCK_FLAGS_CATEGORY_IO);
     }
-#endif
     drv->entry = de;
 
     drv->start = de->start;
@@ -7770,7 +7389,6 @@ init_driver(erts_driver_t *drv, ErlDrvEntry *de, DE_Handle *handle)
     drv->outputv = de->outputv;
     drv->control = de->control;
     drv->call = de->call;
-    drv->event = de->event ? de->event : no_event_callback;
     drv->ready_input = de->ready_input ? de->ready_input : no_ready_input_callback;
     drv->ready_output = de->ready_output ? de->ready_output : no_ready_output_callback;
     drv->timeout = de->timeout ? de->timeout : no_timeout_callback;
@@ -7802,12 +7420,10 @@ init_driver(erts_driver_t *drv, ErlDrvEntry *de, DE_Handle *handle)
 void
 erts_destroy_driver(erts_driver_t *drv)
 {
-#ifdef ERTS_SMP
     if (drv->lock) {
-	erts_smp_mtx_destroy(drv->lock);
+	erts_mtx_destroy(drv->lock);
 	erts_free(ERTS_ALC_T_DRIVER_LOCK, drv->lock);
     }
-#endif    
     erts_free(ERTS_ALC_T_DRIVER, drv);
 }
 
@@ -7818,21 +7434,22 @@ erts_destroy_driver(erts_driver_t *drv)
 
 void add_driver_entry(ErlDrvEntry *drv){
     void *rec_lock;
-    rec_lock = erts_smp_tsd_get(driver_list_lock_status_key);
+    rec_lock = erts_tsd_get(driver_list_lock_status_key);
     /* 
      * Ignore result of erts_add_driver_entry, the init is not
      * allowed to fail when drivers are added by drivers.
      */
-    erts_add_driver_entry(drv, NULL, rec_lock != NULL); 
+    erts_add_driver_entry(drv, NULL, rec_lock != NULL, 0);
 }
 
-int erts_add_driver_entry(ErlDrvEntry *de, DE_Handle *handle, int driver_list_locked)
+int erts_add_driver_entry(ErlDrvEntry *de, DE_Handle *handle,
+                          int driver_list_locked, int taint)
 {
     erts_driver_t *dp = erts_alloc(ERTS_ALC_T_DRIVER, sizeof(erts_driver_t));
-    int res;
+    int err = 0;
 
     if (!driver_list_locked) {
-	erts_smp_rwmtx_rwlock(&erts_driver_list_lock);
+	erts_rwmtx_rwlock(&erts_driver_list_lock);
     }
 
     dp->next = driver_list;
@@ -7843,12 +7460,18 @@ int erts_add_driver_entry(ErlDrvEntry *de, DE_Handle *handle, int driver_list_lo
     driver_list = dp;
 
     if (!driver_list_locked) {
-	erts_smp_tsd_set(driver_list_lock_status_key, (void *) 1);
+	erts_tsd_set(driver_list_lock_status_key, (void *) 1);
     }
 
-    res = init_driver(dp, de, handle);
+    if (!err) {
+        err = init_driver(dp, de, handle);
 
-    if (res != 0) {
+        if (taint) {
+            erts_add_taint(dp->name_atom);
+        }
+    }
+
+    if (err) {
 	/* 
 	 * Remove it all again...
 	 */
@@ -7860,10 +7483,10 @@ int erts_add_driver_entry(ErlDrvEntry *de, DE_Handle *handle, int driver_list_lo
     }
 	
     if (!driver_list_locked) {
-	erts_smp_tsd_set(driver_list_lock_status_key, NULL);
-	erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+	erts_tsd_set(driver_list_lock_status_key, NULL);
+	erts_rwmtx_rwunlock(&erts_driver_list_lock);
     }
-    return res;
+    return err;
 }
 
 /* Not allowed for dynamic drivers */
@@ -7872,9 +7495,9 @@ int remove_driver_entry(ErlDrvEntry *drv)
     erts_driver_t *dp;
     void *rec_lock;
     
-    rec_lock = erts_smp_tsd_get(driver_list_lock_status_key);
+    rec_lock = erts_tsd_get(driver_list_lock_status_key);
     if (rec_lock == NULL) {
-	erts_smp_rwmtx_rwlock(&erts_driver_list_lock);
+	erts_rwmtx_rwlock(&erts_driver_list_lock);
     }
     dp = driver_list;
     while (dp && dp->entry != drv)
@@ -7882,7 +7505,7 @@ int remove_driver_entry(ErlDrvEntry *drv)
     if (dp) {
 	if (dp->handle) {
 	    if (rec_lock == NULL) {
-		erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+		erts_rwmtx_rwunlock(&erts_driver_list_lock);
 	    }
 	    return -1;
 	}
@@ -7896,12 +7519,12 @@ int remove_driver_entry(ErlDrvEntry *drv)
 	}
 	erts_destroy_driver(dp);
 	if (rec_lock == NULL) {
-	    erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+	    erts_rwmtx_rwunlock(&erts_driver_list_lock);
 	}
 	return 1;
     }
     if (rec_lock == NULL) {
-	erts_smp_rwmtx_rwunlock(&erts_driver_list_lock);
+	erts_rwmtx_rwunlock(&erts_driver_list_lock);
     }
     return 0;
 }
@@ -7917,13 +7540,27 @@ int null_func(void)
 int
 erl_drv_putenv(const char *key, char *value)
 {
-    return erts_sys_putenv_raw((char*)key, value);
+    switch (erts_sys_explicit_8bit_putenv((char*)key, value)) {
+    case -1: /* Insufficient buffer space */
+        return 1;
+    case 1: /* Success */
+        return 0;
+    default: /* Not found */
+        return -1;
+    }
 }
 
 int
 erl_drv_getenv(const char *key, char *value, size_t *value_size)
 {
-    return erts_sys_getenv_raw((char*)key, value, value_size);
+    switch (erts_sys_explicit_8bit_getenv((char*)key, value, value_size)) {
+    case -1: /* Insufficient buffer space */
+        return 1;
+    case 1: /* Success */
+        return 0;
+    default: /* Not found */
+        return -1;
+    }
 }
 
 /* get heart_port
diff --git a/erts/emulator/beam/lttng-wrapper.h b/erts/emulator/beam/lttng-wrapper.h
index 0bc75c1552..e7f2971bf7 100644
--- a/erts/emulator/beam/lttng-wrapper.h
+++ b/erts/emulator/beam/lttng-wrapper.h
@@ -61,13 +61,13 @@
 #define lttng_proc_to_mfa_str(p, Name)                              \
     do {                                                            \
         if (ERTS_PROC_IS_EXITING((p))) {                            \
-            strcpy(Name, "<exiting>");                              \
+            sys_strcpy(Name, "<exiting>");                              \
         } else {                                                    \
             BeamInstr *_fptr = find_function_from_pc((p)->i);       \
             if (_fptr) {                                            \
                 lttng_mfa_to_str(_fptr[0],_fptr[1],_fptr[2], Name); \
             } else {                                                \
-                strcpy(Name, "<unknown>");                          \
+                sys_strcpy(Name, "<unknown>");                          \
             }                                                       \
         }                                                           \
     } while(0)
diff --git a/erts/emulator/beam/macros.tab b/erts/emulator/beam/macros.tab
new file mode 100644
index 0000000000..494fe8961e
--- /dev/null
+++ b/erts/emulator/beam/macros.tab
@@ -0,0 +1,173 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+//
+// Define a regular expression that will match instructions that
+// perform GC.  That will allow beam_makeops to check for instructions
+// that don't use $REFRESH_GEN_DEST() when they should.
+//
+
+GC_REGEXP=erts_garbage_collect|erts_gc|GcBifFunction;
+
+// $Offset is relative to the start of the instruction (not to the
+// location of the failure label reference). Since combined
+// instructions may increment the instruction pointer (e.g. in
+// 'increment') for some of the instructions in the group, we actually
+// use a virtual start position common to all instructions in the
+// group. To calculate the correct virtual position, we will need to
+// add $IP_ADJUSTMENT to the offset. ($IP_ADJUSTMENT will usually be
+// zero, except in a few bit syntax instructions.)
+
+SET_I_REL(Offset) {
+    ASSERT(VALID_INSTR(*(I + ($Offset) + $IP_ADJUSTMENT)));
+    I += $Offset + $IP_ADJUSTMENT;
+}
+
+SET_CP_I_ABS(Target) {
+    c_p->i = $Target;
+    ASSERT(VALID_INSTR(*c_p->i));
+}
+
+SET_REL_I(Dst, Offset) {
+    $Dst = I + ($Offset);
+    ASSERT(VALID_INSTR(*$Dst));
+}
+
+FAIL(Fail) {
+    //| -no_prefetch
+    $SET_I_REL($Fail);
+    Goto(*I);
+}
+
+JUMP(Fail) {
+    //| -no_next
+    $SET_I_REL($Fail);
+    Goto(*I);
+}
+
+GC_TEST(Ns, Nh, Live) {
+    Uint need = $Nh + $Ns;
+    if (ERTS_UNLIKELY(E - HTOP < need)) {
+       SWAPOUT;
+       PROCESS_MAIN_CHK_LOCKS(c_p);
+       FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, $Live, FCALLS);
+       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+       PROCESS_MAIN_CHK_LOCKS(c_p);
+       SWAPIN;
+    }
+    HEAP_SPACE_VERIFIED($Nh);
+}
+
+GC_TEST_PRESERVE(NeedHeap, Live, PreserveTerm) {
+    Uint need = $NeedHeap;
+    if (ERTS_UNLIKELY(E - HTOP < need)) {
+       SWAPOUT;
+       reg[$Live] = $PreserveTerm;
+       PROCESS_MAIN_CHK_LOCKS(c_p);
+       FCALLS -= erts_garbage_collect_nobump(c_p, need, reg, $Live+1, FCALLS);
+       ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+       PROCESS_MAIN_CHK_LOCKS(c_p);
+       $PreserveTerm = reg[$Live];
+       SWAPIN;
+    }
+    HEAP_SPACE_VERIFIED($NeedHeap);
+}
+
+
+// Make sure that there are NeedStack + NeedHeap + 1 words available
+// on the combined heap/stack segment, then allocates NeedHeap + 1
+// words on the stack and saves CP.
+AH(NeedStack, NeedHeap, Live) {
+    unsigned needed = $NeedStack + 1;
+    $GC_TEST(needed, $NeedHeap, $Live);
+    E -= needed;
+    *E = make_cp(c_p->cp);
+    c_p->cp = 0;
+}
+
+NEXT0() {
+    //| -no_next
+    SET_I((BeamInstr *) $NEXT_INSTRUCTION);
+    Goto(*I);
+}
+
+NEXT(Addr) {
+    //| -no_next
+    SET_I((BeamInstr *) $Addr);
+    Goto(*I);
+}
+
+FAIL_BODY() {
+    //| -no_prefetch
+    goto find_func_info;
+}
+
+FAIL_HEAD_OR_BODY(Fail) {
+    //| -no_prefetch
+
+    /*
+     * In a correctly working program, we expect failures in
+     * guards to be more likely than failures in bodies.
+     */
+
+    if (ERTS_LIKELY($Fail)) {
+        $FAIL($Fail);
+    }
+    goto find_func_info;
+}
+
+BADARG(Fail) {
+    c_p->freason = BADARG;
+    $FAIL_HEAD_OR_BODY($Fail);
+}
+
+BADARITH0() {
+    c_p->freason = BADARITH;
+    goto find_func_info;
+}
+
+SYSTEM_LIMIT(Fail) {
+    c_p->freason = SYSTEM_LIMIT;
+    $FAIL_HEAD_OR_BODY($Fail);
+}
+
+BIF_ERROR_ARITY_1(Fail, BIF, Op1) {
+    //| -no_prefetch
+    if (ERTS_LIKELY($Fail)) {
+        $FAIL($Fail);
+    }
+    reg[0] = $Op1;
+    SWAPOUT;
+    I = handle_error(c_p, I, reg, &bif_export[$BIF]->info.mfa);
+    goto post_error_handling;
+}
+
+BIF_ERROR_ARITY_2(Fail, BIF, Op1, Op2) {
+    //| -no_prefetch
+    if (ERTS_LIKELY($Fail)) {
+        $FAIL($Fail);
+    }
+    reg[0] = $Op1;
+    reg[1] = $Op2;
+    SWAPOUT;
+    I = handle_error(c_p, I, reg, &bif_export[$BIF]->info.mfa);
+    goto post_error_handling;
+}
diff --git a/erts/emulator/beam/map_instrs.tab b/erts/emulator/beam/map_instrs.tab
new file mode 100644
index 0000000000..c594a87298
--- /dev/null
+++ b/erts/emulator/beam/map_instrs.tab
@@ -0,0 +1,159 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+ensure_map(Map) {
+    if (is_not_map($Map)) {
+        c_p->freason = BADMAP;
+        c_p->fvalue = $Map;
+        $FAIL_BODY();
+    }
+}
+
+new_map(Dst, Live, N) {
+    Eterm res;
+
+    HEAVY_SWAPOUT;
+    res = erts_gc_new_map(c_p, reg, $Live, $N, $NEXT_INSTRUCTION);
+    HEAVY_SWAPIN;
+    $REFRESH_GEN_DEST();
+    $Dst = res;
+    $NEXT($NEXT_INSTRUCTION+$N);
+}
+
+i_new_small_map_lit(Dst, Live, Keys) {
+    Eterm res;
+    Uint n;
+    Eterm keys = $Keys;
+
+    HEAVY_SWAPOUT;
+    res = erts_gc_new_small_map_lit(c_p, reg, keys, $Live, $NEXT_INSTRUCTION);
+    HEAVY_SWAPIN;
+    $REFRESH_GEN_DEST();
+    $Dst = res;
+    n = arityval(*tuple_val(keys));
+    $NEXT($NEXT_INSTRUCTION+n);
+}
+
+i_get_map_element(Fail, Src, Key, Dst) {
+    Eterm res = get_map_element($Src, $Key);
+    if (is_non_value(res)) {
+        $FAIL($Fail);
+    }
+    $Dst = res;
+}
+
+i_get_map_element_hash(Fail, Src, Key, Hx, Dst) {
+     Eterm res = get_map_element_hash($Src, $Key, $Hx);
+     if (is_non_value(res)) {
+        $FAIL($Fail);
+     }
+     $Dst = res;
+}
+
+i_get_map_elements(Fail, Src, N) {
+    Eterm map;
+    BeamInstr *fs;
+    Uint sz, n;
+
+    map = $Src;
+
+    /* This instruction assumes Arg1 is a map,
+     * i.e. that it follows a test is_map if needed.
+     */
+
+    n  = (Uint)$N / 3;
+    fs = $NEXT_INSTRUCTION;
+
+    if (is_flatmap(map)) {
+	flatmap_t *mp;
+	Eterm *ks;
+	Eterm *vs;
+
+	mp = (flatmap_t *)flatmap_val(map);
+	sz = flatmap_get_size(mp);
+
+	if (sz == 0) {
+            $FAIL($Fail);
+	}
+
+	ks = flatmap_get_keys(mp);
+	vs = flatmap_get_values(mp);
+
+	while(sz) {
+	    if (EQ((Eterm) fs[0], *ks)) {
+		PUT_TERM_REG(*vs, fs[1]);
+		n--;
+		fs += 3;
+		/* no more values to fetch, we are done */
+		if (n == 0) {
+                    $NEXT(fs);
+		}
+	    }
+	    ks++, sz--, vs++;
+	}
+        $FAIL($Fail);
+    } else {
+	const Eterm *v;
+	Uint32 hx;
+	ASSERT(is_hashmap(map));
+	while(n--) {
+	    hx = fs[2];
+	    ASSERT(hx == hashmap_make_hash((Eterm)fs[0]));
+	    if ((v = erts_hashmap_get(hx, (Eterm)fs[0], map)) == NULL) {
+		$FAIL($Fail);
+	    }
+	    PUT_TERM_REG(*v, fs[1]);
+	    fs += 3;
+	}
+        $NEXT(fs);
+    }
+}
+
+update_map_assoc(Src, Dst, Live, N) {
+    Eterm res;
+    Uint live = $Live;
+
+    reg[live] = $Src;
+    HEAVY_SWAPOUT;
+    res = erts_gc_update_map_assoc(c_p, reg, live, $N, $NEXT_INSTRUCTION);
+    HEAVY_SWAPIN;
+    ASSERT(is_value(res));
+    $REFRESH_GEN_DEST();
+    $Dst = res;
+    $NEXT($NEXT_INSTRUCTION+$N);
+}
+
+update_map_exact(Fail, Src, Dst, Live, N) {
+    Eterm res;
+    Uint live = $Live;
+
+    reg[live] = $Src;
+    HEAVY_SWAPOUT;
+    res = erts_gc_update_map_exact(c_p, reg, live, $N, $NEXT_INSTRUCTION);
+    HEAVY_SWAPIN;
+    if (is_value(res)) {
+        $REFRESH_GEN_DEST();
+        $Dst = res;
+        $NEXT($NEXT_INSTRUCTION+$N);
+    } else {
+        $FAIL_HEAD_OR_BODY($Fail);
+    }
+}
diff --git a/erts/emulator/beam/module.c b/erts/emulator/beam/module.c
index a386407ec2..1712dc803c 100644
--- a/erts/emulator/beam/module.c
+++ b/erts/emulator/beam/module.c
@@ -39,9 +39,9 @@
 
 static IndexTable module_tables[ERTS_NUM_CODE_IX];
 
-erts_smp_rwmtx_t the_old_code_rwlocks[ERTS_NUM_CODE_IX];
+erts_rwmtx_t the_old_code_rwlocks[ERTS_NUM_CODE_IX];
 
-static erts_smp_atomic_t tot_module_bytes;
+static erts_atomic_t tot_module_bytes;
 
 /* SMP note: Active module table lookup and current module instance can be
  *           read without any locks. Old module instances are protected by
@@ -49,8 +49,6 @@ static erts_smp_atomic_t tot_module_bytes;
  *           Staging table is protected by the "code_ix lock". 
  */
 
-#include "erl_smp.h"
-
 void module_info(fmtfn_t to, void *to_arg)
 {
     index_info(to, to_arg, &module_tables[erts_active_code_ix()]);
@@ -84,7 +82,7 @@ void erts_module_instance_init(struct erl_module_instance* modi)
 static Module* module_alloc(Module* tmpl)
 {
     Module* obj = (Module*) erts_alloc(ERTS_ALC_T_MODULE, sizeof(Module));
-    erts_smp_atomic_add_nob(&tot_module_bytes, sizeof(Module));
+    erts_atomic_add_nob(&tot_module_bytes, sizeof(Module));
 
     obj->module = tmpl->module;
     obj->slot.index = -1;
@@ -98,7 +96,7 @@ static Module* module_alloc(Module* tmpl)
 static void module_free(Module* mod)
 {
     erts_free(ERTS_ALC_T_MODULE, mod);
-    erts_smp_atomic_add_nob(&tot_module_bytes, -sizeof(Module));
+    erts_atomic_add_nob(&tot_module_bytes, -sizeof(Module));
 }
 
 void init_module_table(void)
@@ -120,10 +118,10 @@ void init_module_table(void)
     }
 
     for (i=0; i<ERTS_NUM_CODE_IX; i++) {
-        erts_smp_rwmtx_init(&the_old_code_rwlocks[i], "old_code", make_small(i),
+        erts_rwmtx_init(&the_old_code_rwlocks[i], "old_code", make_small(i),
             ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
     }
-    erts_smp_atomic_init_nob(&tot_module_bytes, 0);
+    erts_atomic_init_nob(&tot_module_bytes, 0);
 }
 
 
@@ -159,14 +157,14 @@ static Module* put_module(Eterm mod, IndexTable* mod_tab)
     oldsz = index_table_sz(mod_tab);
     res = (Module*) index_put_entry(mod_tab, (void*) &e);
     newsz = index_table_sz(mod_tab);
-    erts_smp_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
+    erts_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
     return res;
 }
 
 Module*
 erts_put_module(Eterm mod)
 {
-    ERTS_SMP_LC_ASSERT(erts_initialized == 0
+    ERTS_LC_ASSERT(erts_initialized == 0
 		       || erts_has_code_write_permission());
 
     return put_module(mod, &module_tables[erts_staging_code_ix()]);
@@ -184,7 +182,7 @@ int module_code_size(ErtsCodeIndex code_ix)
 
 int module_table_sz(void)
 {
-    return erts_smp_atomic_read_nob(&tot_module_bytes);
+    return erts_atomic_read_nob(&tot_module_bytes);
 }
 
 #ifdef DEBUG
@@ -233,7 +231,7 @@ void module_start_staging(void)
         copy_module(dst_mod, src_mod);
     }
     newsz = index_table_sz(dst);
-    erts_smp_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
+    erts_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
 
     entries_at_start_staging = dst->entries;
     IF_DEBUG(dbg_load_code_ix = erts_staging_code_ix());
@@ -251,7 +249,7 @@ void module_end_staging(int commit)
 	oldsz = index_table_sz(tab);
 	index_erase_latest_from(tab, entries_at_start_staging);
 	newsz = index_table_sz(tab);
-	erts_smp_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
+	erts_atomic_add_nob(&tot_module_bytes, (newsz - oldsz));
     }
 
     IF_DEBUG(dbg_load_code_ix = -1);
diff --git a/erts/emulator/beam/module.h b/erts/emulator/beam/module.h
index 9d258d5dbf..a3f1ce1705 100644
--- a/erts/emulator/beam/module.h
+++ b/erts/emulator/beam/module.h
@@ -45,7 +45,7 @@ typedef struct erl_module {
     int seen;			/* Used by finish_loading() */
 
     struct erl_module_instance curr;
-    struct erl_module_instance old; /* protected by "old_code" rwlock */
+    struct erl_module_instance old; /* active protected by "old_code" rwlock */
     struct erl_module_instance* on_load;
 } Module; 
 
@@ -72,29 +72,29 @@ int erts_is_old_code_rlocked(ErtsCodeIndex);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
-extern erts_smp_rwmtx_t the_old_code_rwlocks[ERTS_NUM_CODE_IX];
+extern erts_rwmtx_t the_old_code_rwlocks[ERTS_NUM_CODE_IX];
 
 ERTS_GLB_INLINE void erts_rwlock_old_code(ErtsCodeIndex code_ix)
 {
-    erts_smp_rwmtx_rwlock(&the_old_code_rwlocks[code_ix]);
+    erts_rwmtx_rwlock(&the_old_code_rwlocks[code_ix]);
 }
 ERTS_GLB_INLINE void erts_rwunlock_old_code(ErtsCodeIndex code_ix)
 {
-    erts_smp_rwmtx_rwunlock(&the_old_code_rwlocks[code_ix]);
+    erts_rwmtx_rwunlock(&the_old_code_rwlocks[code_ix]);
 }
 ERTS_GLB_INLINE void erts_rlock_old_code(ErtsCodeIndex code_ix)
 {
-    erts_smp_rwmtx_rlock(&the_old_code_rwlocks[code_ix]);
+    erts_rwmtx_rlock(&the_old_code_rwlocks[code_ix]);
 }
 ERTS_GLB_INLINE void erts_runlock_old_code(ErtsCodeIndex code_ix)
 {
-    erts_smp_rwmtx_runlock(&the_old_code_rwlocks[code_ix]);
+    erts_rwmtx_runlock(&the_old_code_rwlocks[code_ix]);
 }
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
 ERTS_GLB_INLINE int erts_is_old_code_rlocked(ErtsCodeIndex code_ix)
 {
-    return erts_smp_lc_rwmtx_is_rlocked(&the_old_code_rwlocks[code_ix]);
+    return erts_lc_rwmtx_is_rlocked(&the_old_code_rwlocks[code_ix]);
 }
 #endif
 
diff --git a/erts/emulator/beam/msg_instrs.tab b/erts/emulator/beam/msg_instrs.tab
new file mode 100644
index 0000000000..9bf3aefaca
--- /dev/null
+++ b/erts/emulator/beam/msg_instrs.tab
@@ -0,0 +1,403 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017-2018. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+// /*
+//  * Skeleton for receive statement:
+//  *
+//  *             recv_mark L1                     Optional
+//  *             call make_ref/monitor            Optional
+//  *             ...
+//  *             recv_set L1                      Optional
+//  *      L1:          <-------------------+
+//  *                   <-----------+       |
+//  *                               |       |
+//  *             loop_rec L2 ------+---+   |
+//  *             ...               |   |   |
+//  *             remove_message    |   |   |
+//  *             jump L3           |   |   |
+//  *		...                 |   |   |
+//  *		loop_rec_end L1   --+   |   |
+//  *      L2:          <---------------+   |
+//  *	   	wait L1  -------------------+      or wait_timeout
+//  *		timeout
+//  *
+//  *	 L3:    Code after receive...
+//  *
+//  */
+
+i_recv_mark() {
+    /*
+     * Save the current end of message queue
+     */
+    ERTS_RECV_MARK_SAVE(c_p);
+}
+
+i_recv_set() {
+    /*
+     * If previously saved recv mark, set peek position to it
+     */
+    ERTS_RECV_MARK_SET(c_p);
+    SET_I($NEXT_INSTRUCTION);
+    goto loop_rec_top__;
+    //| -no_next
+}
+
+i_loop_rec(Dest) {
+    //| -no_prefetch
+
+    /*
+     * Pick up the next message and place it in x(0).
+     * If no message, jump to a wait or wait_timeout instruction.
+     */
+
+    ErtsMessage* msgp;
+
+    /* Entry point from recv_set */
+ loop_rec_top__:
+
+    /*
+     * We need to disable GC while matching messages
+     * in the queue. This since messages with data outside
+     * the heap will be corrupted by a GC.
+     */
+    ASSERT(!(c_p->flags & F_DELAY_GC));
+    c_p->flags |= F_DELAY_GC;
+
+    /* Entry point from loop_rec_end (and locally) */
+ loop_rec__:
+
+    if (FCALLS <= 0 && FCALLS <= neg_o_reds) {
+        $SET_CP_I_ABS(I);
+        c_p->flags &= ~F_DELAY_GC;
+        SWAPOUT;
+        c_p->arity = 0;
+        c_p->current = NULL;
+        goto do_schedule;
+    }
+
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+
+    msgp = PEEK_MESSAGE(c_p);
+
+    if (ERTS_UNLIKELY(msgp == NULL)) {
+        int get_out;
+        SWAPOUT;
+        $SET_CP_I_ABS(I);
+        c_p->arity = 0;
+        c_p->current = NULL;
+        FCALLS -= erts_proc_sig_receive_helper(c_p, FCALLS, neg_o_reds,
+                                               &msgp, &get_out);
+        SWAPIN;
+        if (ERTS_UNLIKELY(msgp == NULL)) {
+            if (get_out) {
+                if (get_out < 0) {
+                    ASSERT(FCALLS <= 0 && FCALLS <= neg_o_reds);
+                    goto loop_rec__; /* yield */
+                }
+                else {
+                    ASSERT(ERTS_PROC_IS_EXITING(c_p));
+                    goto do_schedule; /* exit */
+                }
+            }
+
+            /*
+             * If there are no more messages in queue
+             * (and we are not yielding or exiting)
+             * erts_proc_sig_receive_helper()
+             * returns with message queue lock locked...
+             */
+            c_p->flags &= ~F_DELAY_GC;
+            $SET_I_REL($Dest);
+            Goto(*I);		/* Jump to a wait or wait_timeout instruction */
+        }
+    }
+
+    ASSERT(msgp == PEEK_MESSAGE(c_p));
+    ASSERT(msgp && ERTS_SIG_IS_MSG(msgp));
+
+    if (ERTS_UNLIKELY(ERTS_SIG_IS_EXTERNAL_MSG(msgp))) {
+        FCALLS -= 10; /* FIXME: bump appropriate amount... */
+        SWAPOUT; /* erts_decode_dist_message() may write to heap... */
+        if (!erts_decode_dist_message(c_p, ERTS_PROC_LOCK_MAIN, msgp, 0)) {
+            /*
+             * A corrupt distribution message that we weren't able to decode;
+             * remove it...
+             */
+            /* No swapin should be needed */
+            ASSERT(HTOP == c_p->htop && E == c_p->stop);
+            /* TODO: Add DTrace probe for this bad message situation? */
+            UNLINK_MESSAGE(c_p, msgp);
+            msgp->next = NULL;
+            erts_cleanup_messages(msgp);
+            goto loop_rec__;
+        }
+        SWAPIN;
+    }
+
+    ASSERT(msgp == PEEK_MESSAGE(c_p));
+    ASSERT(ERTS_SIG_IS_INTERNAL_MSG(msgp));
+
+    r(0) = ERL_MESSAGE_TERM(msgp);
+}
+
+remove_message() {
+    //| -no_prefetch
+
+    /*
+     * Remove a (matched) message from the message queue.
+     */
+
+    ErtsMessage* msgp;
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+
+    ERTS_CHK_MBUF_SZ(c_p);
+
+    msgp = PEEK_MESSAGE(c_p);
+
+    if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)) {
+        save_calls(c_p, &exp_receive);
+    }
+    if (ERL_MESSAGE_TOKEN(msgp) == NIL) {
+#ifdef USE_VM_PROBES
+        if (DT_UTAG(c_p) != NIL) {
+            if (DT_UTAG_FLAGS(c_p) & DT_UTAG_PERMANENT) {
+                SEQ_TRACE_TOKEN(c_p) = am_have_dt_utag;
+            } else {
+                DT_UTAG(c_p) = NIL;
+                SEQ_TRACE_TOKEN(c_p) = NIL;
+            }
+        } else {
+#endif
+            SEQ_TRACE_TOKEN(c_p) = NIL;
+#ifdef USE_VM_PROBES
+        }
+        DT_UTAG_FLAGS(c_p) &= ~DT_UTAG_SPREADING;
+#endif
+    } else if (ERL_MESSAGE_TOKEN(msgp) != am_undefined) {
+        Eterm msg;
+        SEQ_TRACE_TOKEN(c_p) = ERL_MESSAGE_TOKEN(msgp);
+#ifdef USE_VM_PROBES
+        if (ERL_MESSAGE_TOKEN(msgp) == am_have_dt_utag) {
+            if (DT_UTAG(c_p) == NIL) {
+                DT_UTAG(c_p) = ERL_MESSAGE_DT_UTAG(msgp);
+            }
+            DT_UTAG_FLAGS(c_p) |= DT_UTAG_SPREADING;
+        } else {
+#endif
+            ASSERT(is_tuple(SEQ_TRACE_TOKEN(c_p)));
+            ASSERT(SEQ_TRACE_TOKEN_ARITY(c_p) == 5);
+            ASSERT(is_small(SEQ_TRACE_TOKEN_SERIAL(c_p)));
+            ASSERT(is_small(SEQ_TRACE_TOKEN_LASTCNT(c_p)));
+            ASSERT(is_small(SEQ_TRACE_TOKEN_FLAGS(c_p)));
+            ASSERT(is_pid(SEQ_TRACE_TOKEN_SENDER(c_p)));
+            c_p->seq_trace_lastcnt = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
+            if (c_p->seq_trace_clock < unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p))) {
+                c_p->seq_trace_clock = unsigned_val(SEQ_TRACE_TOKEN_SERIAL(c_p));
+            }
+            msg = ERL_MESSAGE_TERM(msgp);
+            seq_trace_output(SEQ_TRACE_TOKEN(c_p), msg, SEQ_TRACE_RECEIVE, 
+                             c_p->common.id, c_p);
+#ifdef USE_VM_PROBES
+        }
+#endif
+    }
+#ifdef USE_VM_PROBES
+    if (DTRACE_ENABLED(message_receive)) {
+        Eterm token2 = NIL;
+        DTRACE_CHARBUF(receiver_name, DTRACE_TERM_BUF_SIZE);
+        Sint tok_label = 0;
+        Sint tok_lastcnt = 0;
+        Sint tok_serial = 0;
+        Sint len = erts_proc_sig_privqs_len(c_p);
+
+        dtrace_proc_str(c_p, receiver_name);
+        token2 = SEQ_TRACE_TOKEN(c_p);
+        if (have_seqtrace(token2)) {
+            tok_label = SEQ_TRACE_T_DTRACE_LABEL(token2);
+            tok_lastcnt = signed_val(SEQ_TRACE_T_LASTCNT(token2));
+            tok_serial = signed_val(SEQ_TRACE_T_SERIAL(token2));
+        }
+        DTRACE6(message_receive,
+                receiver_name, size_object(ERL_MESSAGE_TERM(msgp)),
+                len, /* This is NOT message queue len, but its something... */
+                tok_label, tok_lastcnt, tok_serial);
+    }
+#endif
+    UNLINK_MESSAGE(c_p, msgp);
+    JOIN_MESSAGE(c_p);
+    CANCEL_TIMER(c_p);
+
+    erts_save_message_in_proc(c_p, msgp);
+    c_p->flags &= ~F_DELAY_GC;
+
+    if (ERTS_IS_GC_DESIRED_INTERNAL(c_p, HTOP, E)) {
+        /*
+         * We want to GC soon but we leave a few
+         * reductions giving the message some time
+         * to turn into garbage.
+         */
+        ERTS_VBUMP_LEAVE_REDS_INTERNAL(c_p, 5, FCALLS);
+    }
+
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    ERTS_CHK_MBUF_SZ(c_p);
+
+    ERTS_VERIFY_UNUSED_TEMP_ALLOC(c_p);
+    PROCESS_MAIN_CHK_LOCKS(c_p);
+}
+
+loop_rec_end(Dest) {
+    //| -no_next
+    /*
+     * Advance the save pointer to the next message (the current
+     * message didn't match), then jump to the loop_rec instruction.
+     */
+
+    ASSERT(c_p->flags & F_DELAY_GC);
+
+    $SET_I_REL($Dest);
+    SAVE_MESSAGE(c_p);
+    FCALLS--;
+    goto loop_rec__;
+}
+
+timeout_locked() {
+    /*
+     * A timeout has occurred.  Reset the save pointer so that the next
+     * receive statement will examine the first message first.
+     */
+
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+    $timeout();
+}
+
+timeout() {
+    if (IS_TRACED_FL(c_p, F_TRACE_RECEIVE)) {
+        trace_receive(c_p, am_clock_service, am_timeout, NULL);
+    }
+    if (ERTS_PROC_GET_SAVED_CALLS_BUF(c_p)) {
+        save_calls(c_p, &exp_timeout);
+    }
+    c_p->flags &= ~F_TIMO;
+    JOIN_MESSAGE(c_p);
+}
+
+TIMEOUT_VALUE() {
+    c_p->freason = EXC_TIMEOUT_VALUE;
+    goto find_func_info;
+    //| -no_next
+}
+
+i_wait_error_locked() {
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+    $TIMEOUT_VALUE();
+}
+
+i_wait_error() {
+    $TIMEOUT_VALUE();
+}
+
+wait_timeout_unlocked_int := wait.lock.int.execute;
+wait_timeout_locked_int := wait.int.execute;
+
+wait_timeout_unlocked := wait.lock.src.execute;
+wait_timeout_locked := wait.src.execute;
+
+wait_unlocked := wait.lock.execute;
+wait_locked := wait.unlocked.execute;
+
+wait.lock() {
+    erts_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+}
+
+wait.unlocked() {
+}
+
+wait.int(Int) {
+    /*
+     * If we have already set the timer, we must NOT set it again.  Therefore,
+     * we must test the F_INSLPQUEUE flag as well as the F_TIMO flag.
+     */
+    if ((c_p->flags & (F_INSLPQUEUE | F_TIMO)) == 0) {
+        BeamInstr** pi = (BeamInstr **) c_p->def_arg_reg;
+        *pi = $NEXT_INSTRUCTION;
+        erts_set_proc_timer_uword(c_p, $Int);
+    }
+}
+
+wait.src(Src) {
+    /*
+     * If we have already set the timer, we must NOT set it again.  Therefore,
+     * we must test the F_INSLPQUEUE flag as well as the F_TIMO flag.
+     */
+    if ((c_p->flags & (F_INSLPQUEUE | F_TIMO)) == 0) {
+        Eterm timeout_value = $Src;
+        if (timeout_value == make_small(0)) {
+            erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+            $NEXT0();
+        } else if (timeout_value == am_infinity) {
+            c_p->flags |= F_TIMO;
+        } else {
+            int tres = erts_set_proc_timer_term(c_p, timeout_value);
+            if (tres == 0) {
+                /*
+                 * The timer routiner will set c_p->i to the value in
+                 * c_p->def_arg_reg[0].  Note that it is safe to use this
+                 * location because there are no living x registers in
+                 * a receive statement.
+                 */
+                BeamInstr** pi = (BeamInstr**) c_p->def_arg_reg;
+                *pi = $NEXT_INSTRUCTION;
+            } else { /* Wrong time */
+                erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+                c_p->freason = EXC_TIMEOUT_VALUE;
+                goto find_func_info;
+            }
+        }
+    }
+}
+
+//
+// Prepare to wait indefinitely for a new message to arrive
+// (or the time set above if falling through from above).
+// When a new message arrives, control will be transferred
+// the loop_rec instruction (at label L1).  In case of
+// of timeout, control will be transferred to the timeout
+// instruction following the wait_timeout instruction.
+//
+
+wait.execute(JumpTarget) {
+    $SET_REL_I(c_p->i, $JumpTarget); /* L1 */
+    SWAPOUT;
+    c_p->arity = 0;
+
+    if (!ERTS_PTMR_IS_TIMED_OUT(c_p)) {
+        erts_atomic32_read_band_relb(&c_p->state,
+                                         ~ERTS_PSFLG_ACTIVE);
+    }
+    ASSERT(!ERTS_PROC_IS_EXITING(c_p));
+    erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+    c_p->current = NULL;
+    goto do_schedule;
+    //| -no_next
+}
diff --git a/erts/emulator/beam/ops.tab b/erts/emulator/beam/ops.tab
index 44613c7d85..8b2d9098a8 100644
--- a/erts/emulator/beam/ops.tab
+++ b/erts/emulator/beam/ops.tab
@@ -59,6 +59,7 @@ put_tuple u==0 d => too_old_compiler
 # All the other instructions.
 #
 
+%cold
 label L
 i_func_info I a a I
 int_code_end
@@ -68,6 +69,8 @@ i_debug_breakpoint
 i_return_time_trace
 i_return_to_trace
 i_yield
+trace_jump W
+%hot
 
 return
 
@@ -96,24 +99,21 @@ line Loc | func_info M F A => func_info M F A | line Loc
 
 line I
 
+allocate t t?
+allocate_heap t I t?
 
-%macro: allocate Allocate -pack
-%macro: allocate_zero AllocateZero -pack
-%macro: allocate_heap AllocateHeap -pack
-%macro: allocate_heap_zero AllocateHeapZero -pack
-%macro: test_heap TestHeap -pack
+%cold
+deallocate Q
+%hot
 
-allocate t t
-allocate_heap t I t
-deallocate I
 init y
-allocate_zero t t
-allocate_heap_zero t I t
+allocate_zero t t?
+allocate_heap_zero t I t?
 
 trim N Remaining => i_trim N
-i_trim I
+i_trim t
 
-test_heap I t
+test_heap I t?
 
 allocate_heap S u==0 R => allocate S R
 allocate_heap_zero S u==0 R => allocate_zero S R
@@ -122,8 +122,6 @@ init2 y y
 init3 y y y
 init Y1 | init Y2 | init Y3 => init3 Y1 Y2 Y3
 init Y1 | init Y2 => init2 Y1 Y2
-%macro: init2 Init2 -pack
-%macro: init3 Init3 -pack
 
 # Selecting values
 
@@ -160,28 +158,20 @@ is_tuple Fail=f S | select_tuple_arity S=d Fail=f Size=u Rest=* => \
 select_tuple_arity S=d Fail=f Size=u Rest=* => \
   gen_select_tuple_arity(S, Fail, Size, Rest)
 
-i_select_val_bins x f I
-i_select_val_bins y f I
+i_select_val_bins xy f? I
 
-i_select_val_lins x f I
-i_select_val_lins y f I
+i_select_val_lins xy f? I
 
-i_select_val2 x f c c f f
-i_select_val2 y f c c f f
+i_select_val2 xy f? c c
 
-i_select_tuple_arity x f I
-i_select_tuple_arity y f I
+i_select_tuple_arity xy f? I
 
-i_select_tuple_arity2 x f A A f f
-i_select_tuple_arity2 y f A A f f
+i_select_tuple_arity2 xy f? A A
 
-i_jump_on_val_zero x f I
-i_jump_on_val_zero y f I
+i_jump_on_val_zero xy f? I
 
-i_jump_on_val x f I I
-i_jump_on_val y f I I
+i_jump_on_val xy f? I W
 
-%macro: get_list GetList -pack
 get_list xy xy xy
 
 # The following get_list instructions using x(0) are frequently used.
@@ -192,50 +182,55 @@ get_list r x y
 get_list r y r
 get_list r x r
 
+get_hd xy xy
+get_tl xy xy
+
 # Old-style catch.
 catch y f
 catch_end y
 
 # Try/catch.
 try Y F => catch Y F
-try_case Y => try_end Y
+
+try_case y
 try_end y
 
+%cold
 try_case_end s
+%hot
 
 # Destructive set tuple element
 
-set_tuple_element s d P
+set_tuple_element s S P
 
 # Get tuple element
 
-%macro: i_get_tuple_element GetTupleElement -pack
 i_get_tuple_element xy P x
 
 %cold
 i_get_tuple_element xy P y
 %hot
 
-%macro: i_get_tuple_element2 GetTupleElement2 -pack
 i_get_tuple_element2 x P x
-
-%macro: i_get_tuple_element2y GetTupleElement2Y -pack
 i_get_tuple_element2y x P y y
 
-%macro: i_get_tuple_element3 GetTupleElement3 -pack
 i_get_tuple_element3 x P x
 
-%macro: is_number IsNumber -fail_action
 %cold
-is_number f x
-is_number f y
+is_number f? xy
 %hot
+
 is_number Fail=f i =>
 is_number Fail=f na => jump Fail
 is_number Fail Literal=q => move Literal x | is_number Fail x
 
 jump f
 
+#
+# Expection rasing instructions. Infrequently executed.
+#
+
+%cold
 case_end NotInX=cy => move NotInX x | case_end x
 badmatch NotInX=cy => move NotInX x | badmatch x
 
@@ -257,9 +252,14 @@ i_raise
 badarg j
 system_limit j
 
+%hot
+
+#
+# Move instructions.
+#
+
 move C=cxy x==0 | jump Lbl => move_jump Lbl C
 
-%macro: move_jump MoveJump -nonext
 move_jump f ncxy
 
 # Movement to and from the stack is common
@@ -283,10 +283,6 @@ move_window X1=x X2=x X3=x X4=x Y1=y Y4=y | move X5=x Y5=y | succ(Y4,Y5) => \
 move_window X1=x X2=x X3=x Y1=y Y3=y => move_window3 X1 X2 X3 Y1
 move_window X1=x X2=x X3=x X4=x Y1=y Y4=y => move_window4 X1 X2 X3 X4 Y1
 
-%macro: move_window3 MoveWindow3 -pack
-%macro: move_window4 MoveWindow4 -pack
-%macro: move_window5 MoveWindow5 -pack
-
 move_window3 x x x y
 move_window4 x x x x y
 move_window5 x x x x x y
@@ -311,10 +307,8 @@ swap_temp R1 R2 Tmp | line Loc | call_ext_only Live Addr | \
 swap_temp R1 R2 Tmp | line Loc | call_ext_last Live Addr D | \
   is_killed(Tmp, Live) => swap R1 R2 | line Loc | call_ext_last Live Addr D
 
-%macro: swap_temp SwapTemp -pack
 swap_temp x xy x
 
-%macro: swap Swap -pack
 swap x xy
 
 move Src=x D1=x | move Src=x D2=x => move_dup Src D1 D2
@@ -358,17 +352,13 @@ move C=aiq X=x==2 => move_x2 C
 move_x1 c
 move_x2 c
 
-%macro: move_shift MoveShift -pack
 move_shift x x x
 move_shift y x x
 move_shift x y x
 move_shift x x y
 
-%macro: move_dup MoveDup -pack
 move_dup xy x xy
 
-%macro: move2_par Move2Par -pack
-
 move2_par x y x y
 move2_par y x y x
 move2_par x x x x
@@ -380,7 +370,6 @@ move2_par y x x y
 move2_par x x y x
 move2_par y x x x
 
-%macro: move3 Move3 -pack
 move3 x y x y x y
 move3 y x y x y x
 move3 x x x x x x
@@ -390,7 +379,6 @@ move3 x x x x x x
 move S=n D=y => init D
 move S=c D=y => move S x | move x D
 
-%macro:move Move -pack
 move x x
 move x y
 move y x
@@ -410,13 +398,15 @@ move r y
 
 loop_rec Fail x==0 | smp_mark_target_label(Fail) => i_loop_rec Fail
 
-label L | wait_timeout Fail Src | smp_already_locked(L) => label L | i_wait_timeout_locked Fail Src
-wait_timeout Fail Src => i_wait_timeout Fail Src
-i_wait_timeout Fail Src=aiq => gen_literal_timeout(Fail, Src)
-i_wait_timeout_locked Fail Src=aiq => gen_literal_timeout_locked(Fail, Src)
+label L | wait_timeout Fail Src | smp_already_locked(L) => \
+    label L | wait_timeout_locked Src Fail
+wait_timeout Fail Src => wait_timeout_unlocked Src Fail
+
+wait_timeout_unlocked Src=aiq Fail => gen_literal_timeout(Fail, Src)
+wait_timeout_locked Src=aiq Fail => gen_literal_timeout_locked(Fail, Src)
 
 label L | wait Fail | smp_already_locked(L) => label L | wait_locked Fail
-wait Fail | smp() => wait_unlocked Fail
+wait Fail => wait_unlocked Fail
 
 label L | timeout | smp_already_locked(L) => label L | timeout_locked
 
@@ -425,15 +415,19 @@ timeout
 timeout_locked
 i_loop_rec f
 loop_rec_end f
-wait f
 wait_locked f
 wait_unlocked f
-i_wait_timeout f I
-i_wait_timeout f s
-i_wait_timeout_locked f I
-i_wait_timeout_locked f s
+
+# Note that a timeout value must fit in 32 bits.
+wait_timeout_unlocked_int I f
+wait_timeout_unlocked s f
+wait_timeout_locked_int I f
+wait_timeout_locked s f
+
+%cold
 i_wait_error
 i_wait_error_locked
+%hot
 
 send
 
@@ -441,56 +435,52 @@ send
 # Optimized comparisons with one immediate/literal operand.
 #
 
-is_eq_exact Lbl R=xy C=ian => i_is_eq_exact_immed Lbl R C
+is_eq_exact Lbl S S =>
+is_eq_exact Lbl C1=c C2=c => move C1 x | is_eq_exact Lbl x C2
+is_eq_exact Lbl C=c R=xy => is_eq_exact Lbl R C
+
+is_eq_exact Lbl R=xy n => is_nil Lbl R
+is_eq_exact Lbl R=xy C=ia => i_is_eq_exact_immed Lbl R C
 is_eq_exact Lbl R=xy C=q => i_is_eq_exact_literal Lbl R C
 
+is_ne_exact Lbl S S => jump Lbl
+is_ne_exact Lbl C1=c C2=c => move C1 x | is_ne_exact Lbl x C2
+is_ne_exact Lbl C=c R=xy => is_ne_exact Lbl R C
+
 is_ne_exact Lbl R=xy C=ian => i_is_ne_exact_immed Lbl R C
 is_ne_exact Lbl R=xy C=q => i_is_ne_exact_literal Lbl R C
 
-%macro: i_is_eq_exact_immed EqualImmed -fail_action
-i_is_eq_exact_immed f r c
-i_is_eq_exact_immed f x c
-i_is_eq_exact_immed f y c
+i_is_eq_exact_immed f? rxy c
 
-i_is_eq_exact_literal f x c
-i_is_eq_exact_literal f y c
+i_is_eq_exact_literal f? xy c
 
-%macro: i_is_ne_exact_immed NotEqualImmed -fail_action
-i_is_ne_exact_immed f x c
-i_is_ne_exact_immed f y c
+i_is_ne_exact_immed f? xy c
 
-i_is_ne_exact_literal f x c
-i_is_ne_exact_literal f y c
+i_is_ne_exact_literal f? xy c
 
 is_eq_exact Lbl Y=y X=x => is_eq_exact Lbl X Y
-%macro: is_eq_exact EqualExact -fail_action -pack
-is_eq_exact f x xy
-is_eq_exact f s s
-
-%macro: is_lt IsLessThan -fail_action
-is_lt f x x
-is_lt f x c
-is_lt f c x
+is_eq_exact f? x xy
+is_eq_exact f? y y
+
+is_ne_exact f? S S
+
+is_lt f? x x
+is_lt f? x c
+is_lt f? c x
 %cold
-is_lt f s s
+is_lt f? s s
 %hot
 
-%macro: is_ge IsGreaterEqual -fail_action
-is_ge f x x
-is_ge f x c
-is_ge f c x
+is_ge f? x x
+is_ge f? x c
+is_ge f? c x
 %cold
-is_ge f s s
+is_ge f? s s
 %hot
 
-%macro: is_ne_exact NotEqualExact -fail_action
-is_ne_exact f s s
-
-%macro: is_eq Equal -fail_action
-is_eq f s s
+is_eq f? s s
 
-%macro: is_ne NotEqual -fail_action
-is_ne f s s
+is_ne f? s s
 
 #
 # Putting things.
@@ -507,9 +497,7 @@ i_put_tuple Dst Arity Puts=* | put S => \
 
 i_put_tuple/2
 
-%macro:i_put_tuple PutTuple -pack -goto:do_put_tuple
-i_put_tuple x I
-i_put_tuple y I
+i_put_tuple xy I
 
 #
 # The instruction "put_list Const [] Dst" were generated in rare
@@ -518,7 +506,9 @@ i_put_tuple y I
 #
 put_list Const=c n Dst => move Const x | put_list x n Dst
 
-%macro:put_list PutList -pack
+put_list Src Dst=x Dst => update_list Src Dst
+
+update_list xyc x
 
 put_list x n x
 put_list y n x
@@ -528,8 +518,6 @@ put_list y x x
 put_list y y x
 put_list x y x
 
-put_list y x x
-
 # put_list SrcReg Constant Dst
 
 put_list x c x
@@ -544,8 +532,6 @@ put_list c y x
 
 # The following put_list instructions using x(0) are frequently used.
 
-put_list y r r
-put_list x r r
 put_list r n r
 put_list r n x
 put_list r x x
@@ -556,10 +542,12 @@ put_list x x r
 put_list s s d
 %hot
 
+
 #
 # Some more only used by the emulator
 #
 
+%cold
 normal_exit
 continue_exit
 apply_bif
@@ -567,6 +555,7 @@ call_nif
 call_error_handler
 error_action_code
 return_trace
+%hot
 
 #
 # Instruction transformations & folded instructions.
@@ -577,27 +566,18 @@ return_trace
 
 move S x==0 | return => move_return S
 
-%macro: move_return MoveReturn -nonext
-move_return x
-move_return c
-move_return n
+move_return xcn
 
 move S x==0 | deallocate D | return => move_deallocate_return S D
 
-%macro: move_deallocate_return MoveDeallocateReturn -pack -nonext
-move_deallocate_return x Q
-move_deallocate_return y Q
-move_deallocate_return c Q
-move_deallocate_return n Q
+move_deallocate_return xycn Q
 
 deallocate D | return => deallocate_return D
 
-%macro: deallocate_return DeallocateReturn -nonext
 deallocate_return Q
 
 test_heap Need u==1 | put_list Y=y x==0 x==0 => test_heap_1_put_list Need Y
 
-%macro: test_heap_1_put_list TestHeapPutList -pack
 test_heap_1_put_list I y
 
 #
@@ -608,9 +588,7 @@ is_tagged_tuple Fail Literal=q Arity Atom => \
     move Literal x | is_tagged_tuple Fail x Arity Atom
 is_tagged_tuple Fail=f c Arity Atom  => jump Fail
 
-%macro:is_tagged_tuple IsTaggedTuple -fail_action
-
-is_tagged_tuple f rxy A a
+is_tagged_tuple f? rxy A a
 
 # Test tuple & arity (head)
 
@@ -618,18 +596,14 @@ is_tuple Fail Literal=q => move Literal x | is_tuple Fail x
 is_tuple Fail=f c => jump Fail
 is_tuple Fail=f S=xy | test_arity Fail=f S=xy Arity => is_tuple_of_arity Fail S Arity
 
-%macro:is_tuple_of_arity IsTupleOfArity -fail_action
-
-is_tuple_of_arity f rxy A
+is_tuple_of_arity f? rxy A
 
-%macro: is_tuple IsTuple -fail_action
-is_tuple f rxy
+is_tuple f? rxy
 
 test_arity Fail Literal=q Arity => move Literal x | test_arity Fail x Arity
 test_arity Fail=f c Arity => jump Fail
 
-%macro: test_arity IsArity -fail_action
-test_arity f xy A
+test_arity f? xy A
 
 get_tuple_element Reg=x P1 D1=x | get_tuple_element Reg=x P2 D2=x | \
    get_tuple_element Reg=x P3 D3=x | \
@@ -650,52 +624,40 @@ is_integer Fail Literal=q => move Literal x | is_integer Fail x
 
 is_integer Fail=f S=x | allocate Need Regs => is_integer_allocate Fail S Need Regs
 
-%macro: is_integer_allocate IsIntegerAllocate -fail_action
-is_integer_allocate f x I I
+is_integer_allocate f? x t t
 
-%macro: is_integer IsInteger -fail_action
-is_integer f xy
+is_integer f? xy
 
 is_list Fail=f n =>
 is_list Fail Literal=q => move Literal x | is_list Fail x
 is_list Fail=f c => jump Fail
-%macro: is_list IsList -fail_action
-is_list f x
+is_list f? x
 %cold
-is_list f y
+is_list f? y
 %hot
 
 is_nonempty_list Fail=f S=x | allocate Need Rs => is_nonempty_list_allocate Fail S Need Rs
 
-%macro:is_nonempty_list_allocate IsNonemptyListAllocate -fail_action -pack
-is_nonempty_list_allocate f rx I t
-
-is_nonempty_list F=f x==0 | test_heap I1 I2 => is_non_empty_list_test_heap F I1 I2
-
-%macro: is_non_empty_list_test_heap IsNonemptyListTestHeap -fail_action -pack
-is_non_empty_list_test_heap f I t
+is_nonempty_list F=f x==0 | test_heap I1 I2 => is_nonempty_list_test_heap F I1 I2
 
 is_nonempty_list Fail=f S=x | get_list S D1=x D2=x => \
   is_nonempty_list_get_list Fail S D1 D2
 
-%macro: is_nonempty_list_get_list IsNonemptyListGetList -fail_action -pack
-is_nonempty_list_get_list f rx x x
-
-%macro: is_nonempty_list IsNonemptyList -fail_action
-is_nonempty_list f xy
+is_nonempty_list_allocate f? rx t t
+is_nonempty_list_test_heap f? I t
+is_nonempty_list_get_list f? rx x x
+is_nonempty_list f? xy
 
-%macro: is_atom IsAtom -fail_action
-is_atom f x
+is_atom f? x
 %cold
-is_atom f y
+is_atom f? y
 %hot
 is_atom Fail=f a =>
 is_atom Fail=f niq => jump Fail
 
-%macro: is_float IsFloat -fail_action
-is_float f x
+is_float f? x
 %cold
-is_float f y
+is_float f? y
 %hot
 is_float Fail=f nai => jump Fail
 is_float Fail Literal=q => move Literal x | is_float Fail x
@@ -703,15 +665,13 @@ is_float Fail Literal=q => move Literal x | is_float Fail x
 is_nil Fail=f n =>
 is_nil Fail=f qia => jump Fail
 
-%macro: is_nil IsNil -fail_action
-is_nil f xy
+is_nil f? xy
 
 is_binary Fail Literal=q => move Literal x | is_binary Fail x
 is_binary Fail=f c => jump Fail
-%macro: is_binary IsBinary -fail_action
-is_binary f x
+is_binary f? x
 %cold
-is_binary f y
+is_binary f? y
 %hot
 
 # XXX Deprecated.
@@ -719,31 +679,27 @@ is_bitstr Fail Term => is_bitstring Fail Term
 
 is_bitstring Fail Literal=q => move Literal x | is_bitstring Fail x
 is_bitstring Fail=f c => jump Fail
-%macro: is_bitstring IsBitstring -fail_action
-is_bitstring f x
+is_bitstring f? x
 %cold
-is_bitstring f y
+is_bitstring f? y
 %hot
 
 is_reference Fail=f cq => jump Fail
-%macro: is_reference IsRef -fail_action
-is_reference f x
+is_reference f? x
 %cold
-is_reference f y
+is_reference f? y
 %hot
 
 is_pid Fail=f cq => jump Fail
-%macro: is_pid IsPid -fail_action
-is_pid f x
+is_pid f? x
 %cold
-is_pid f y
+is_pid f? y
 %hot
 
 is_port Fail=f cq => jump Fail
-%macro: is_port IsPort -fail_action
-is_port f x
+is_port f? x
 %cold
-is_port f y
+is_port f? y
 %hot
 
 is_boolean Fail=f a==am_true =>
@@ -751,22 +707,20 @@ is_boolean Fail=f a==am_false =>
 is_boolean Fail=f ac => jump Fail
 
 %cold
-%macro: is_boolean IsBoolean -fail_action
-is_boolean f xy
+is_boolean f? xy
 %hot
 
-is_function2 Fail=f acq Arity => jump Fail
+is_function2 Fail=f Literal=q Arity | literal_is_export(Literal) =>
+is_function2 Fail=f c Arity => jump Fail
 is_function2 Fail=f Fun a => jump Fail
 
-is_function2 f s s
-%macro: is_function2 IsFunction2 -fail_action
+is_function2 f? S s
 
 # Allocating & initializing.
 allocate Need Regs | init Y => allocate_init Need Regs Y
 init Y1 | init Y2 => init2 Y1 Y2
 
-%macro: allocate_init AllocateInit -pack
-allocate_init t I y
+allocate_init t t? y
 
 #################################################################
 # External function and bif calls.
@@ -1013,16 +967,18 @@ call_ext_last Ar Func D => i_call_ext_last Func D
 call_ext_only Ar Func   => i_call_ext_only Func
 
 i_apply
-i_apply_last P
+i_apply_last Q
 i_apply_only
 
 i_apply_fun
-i_apply_fun_last P
+i_apply_fun_last Q
 i_apply_fun_only
 
+%cold
 i_hibernate
 
 i_perf_counter
+%hot
 
 call_bif e
 
@@ -1045,73 +1001,57 @@ bif2 Fail Bif S1 S2 Dst => i_bif2 Fail Bif S1 S2 Dst
 i_get_hash c I d
 i_get s d
 
-%macro: self Self
 self xy
 
-%macro: node Node
 node x
 %cold
 node y
 %hot
 
-i_fast_element j x I d
-i_fast_element j y I d
+# Note: 'I' is sufficient because this instruction will only be used
+# if the arity fits in 24 bits.
+i_fast_element xy j? I d
 
-i_element j xy s d
+i_element xy j? s d
 
-bif1 f b s d
+bif1 f? b s d
 bif1_body b s d
-i_bif2 f b s s d
+i_bif2 f? b s s d
 i_bif2_body b s s d
 
 #
 # Internal calls.
 #
 
-move S=c x==0 | call Ar P=f => i_move_call S P
-move S=s x==0 | call Ar P=f => move_call S P
-
-i_move_call c f
+move S=cxy x==0 | call Ar P=f => move_call S P
 
-%macro:move_call MoveCall -arg_f -size -nonext
 move_call/2
+move_call cxy f
 
-move_call xy f
-
-move S=c x==0 | call_last Ar P=f D => i_move_call_last P D S
 move S x==0 | call_last Ar P=f D => move_call_last S P D
 
-i_move_call_last f P c
-
-%macro:move_call_last MoveCallLast -arg_f -nonext -pack
-
 move_call_last/3
-move_call_last xy f Q
+move_call_last cxy f Q
 
-move S=c x==0 | call_only Ar P=f => i_move_call_only P S
-move S=x x==0 | call_only Ar P=f => move_call_only S P
+move S=cx x==0 | call_only Ar P=f => move_call_only S P
 
-i_move_call_only f c
-
-%macro:move_call_only MoveCallOnly -arg_f -nonext
 move_call_only/2
-
-move_call_only x f
+move_call_only cx f
 
 call Ar Func        => i_call Func
 call_last Ar Func D => i_call_last Func D
 call_only Ar Func   => i_call_only Func
 
 i_call f
-i_call_last f P
+i_call_last f Q
 i_call_only f
 
 i_call_ext e
-i_call_ext_last e P
+i_call_ext_last e Q
 i_call_ext_only e
 
 i_move_call_ext c e
-i_move_call_ext_last e P c
+i_move_call_ext_last e Q c
 i_move_call_ext_only e c
 
 # Fun calls.
@@ -1119,18 +1059,16 @@ i_move_call_ext_only e c
 call_fun Arity | deallocate D | return => i_call_fun_last Arity D
 call_fun Arity => i_call_fun Arity
 
-i_call_fun I
-i_call_fun_last I P
+i_call_fun t
+i_call_fun_last t Q
 
 make_fun2 OldIndex=u => gen_make_fun2(OldIndex)
 
-%macro: i_make_fun MakeFun -pack
 %cold
-i_make_fun I t
+i_make_fun W t
 %hot
 
-%macro: is_function IsFunction -fail_action
-is_function f xy
+is_function f? xy
 is_function Fail=f c => jump Fail
 
 func_info M F A => i_func_info u M F A
@@ -1139,45 +1077,44 @@ func_info M F A => i_func_info u M F A
 # New bit syntax matching (R11B).
 # ================================================================
 
-%cold
+%warm
 bs_start_match2 Fail=f ica X Y D => jump Fail
 bs_start_match2 Fail Bin X Y D => i_bs_start_match2 Bin Fail X Y D
-i_bs_start_match2 xy f I I d
+i_bs_start_match2 xy f t t x
 
 bs_save2 Reg Index => gen_bs_save(Reg, Index)
-i_bs_save2 x I
+i_bs_save2 x t
 
 bs_restore2 Reg Index => gen_bs_restore(Reg, Index)
-i_bs_restore2 x I
+i_bs_restore2 x t
 
 # Matching integers
 bs_match_string Fail Ms Bits Val => i_bs_match_string Ms Fail Bits Val
 
-i_bs_match_string x f I I
+i_bs_match_string x f W W
 
 # Fetching integers from binaries.
 bs_get_integer2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_integer2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-i_bs_get_integer_small_imm x I f I d
-i_bs_get_integer_imm x I I f I d
-i_bs_get_integer f I I s s d
-i_bs_get_integer_8 x f d
-i_bs_get_integer_16 x f d
-i_bs_get_integer_32 x f I d
+i_bs_get_integer_small_imm x W f? t x
+i_bs_get_integer_imm x W t f? t x
+i_bs_get_integer f? t t x s x
+i_bs_get_integer_8 x f? x
+i_bs_get_integer_16 x f? x
+
+%if ARCH_64
+i_bs_get_integer_32 x f? x
+%endif
 
 # Fetching binaries from binaries.
 bs_get_binary2 Fail=f Ms=x Live=u Sz=sq Unit=u Flags=u Dst=d => \
 			gen_get_binary2(Fail, Ms, Live, Sz, Unit, Flags, Dst)
 
-%macro: i_bs_get_binary_imm2 BsGetBinaryImm_2 -fail_action
-%macro: i_bs_get_binary2 BsGetBinary_2 -fail_action
-%macro: i_bs_get_binary_all2 BsGetBinaryAll_2 -fail_action
-
-i_bs_get_binary_imm2 f x I I I x
-i_bs_get_binary2 f x I s I x
-i_bs_get_binary_all2 f x I I x
-i_bs_get_binary_all_reuse x f I
+i_bs_get_binary_imm2 f? x t W t x
+i_bs_get_binary2 f x t? s t x
+i_bs_get_binary_all2 f? x t t x
+i_bs_get_binary_all_reuse x f? t
 
 # Fetching float from binaries.
 bs_get_float2 Fail=f Ms=x Live=u Sz=s Unit=u Flags=u Dst=d => \
@@ -1185,35 +1122,32 @@ bs_get_float2 Fail=f Ms=x Live=u Sz=s Unit=u Flags=u Dst=d => \
 
 bs_get_float2 Fail=f Ms=x Live=u Sz=q Unit=u Flags=u Dst=d => jump Fail
 
-%macro: i_bs_get_float2 BsGetFloat2 -fail_action
-i_bs_get_float2 f x I s I x
+i_bs_get_float2 f? x t s t x
 
 # Miscellanous
 
 bs_skip_bits2 Fail=f Ms=x Sz=sq Unit=u Flags=u => \
 			gen_skip_bits2(Fail, Ms, Sz, Unit, Flags)
 
-%macro: i_bs_skip_bits_imm2 BsSkipBitsImm2 -fail_action
-i_bs_skip_bits_imm2 f x I
-
-%macro: i_bs_skip_bits2 BsSkipBits2 -fail_action
-i_bs_skip_bits2 f x xy I
-
-%macro: i_bs_skip_bits_all2 BsSkipBitsAll2 -fail_action
-i_bs_skip_bits_all2 f x I 
+i_bs_skip_bits_imm2 f? x W
+i_bs_skip_bits2 f? x xy t
+i_bs_skip_bits_all2 f? x t
 
 bs_test_tail2 Fail=f Ms=x Bits=u==0 => bs_test_zero_tail2 Fail Ms
 bs_test_tail2 Fail=f Ms=x Bits=u => bs_test_tail_imm2 Fail Ms Bits
-bs_test_zero_tail2 f x
-bs_test_tail_imm2 f x I
+bs_test_zero_tail2 f? x
+bs_test_tail_imm2 f? x W
 
 bs_test_unit F Ms Unit=u==8 => bs_test_unit8 F Ms
-bs_test_unit f x I
-bs_test_unit8 f x
+bs_test_unit f? x t
+bs_test_unit8 f? x
 
 # An y register operand for bs_context_to_binary is rare,
 # but can happen because of inlining.
 
+bs_context_to_binary Y=y | line L | badmatch Y => \
+    move Y x | bs_context_to_binary x | line L | badmatch x
+
 bs_context_to_binary Y=y => move Y x | bs_context_to_binary x
 
 bs_context_to_binary x
@@ -1222,14 +1156,14 @@ bs_context_to_binary x
 # Utf8/utf16/utf32 support. (R12B-5)
 #
 bs_get_utf8 Fail=f Ms=x u u Dst=d => i_bs_get_utf8 Ms Fail Dst
-i_bs_get_utf8 x f d
+i_bs_get_utf8 x f? x
 
 bs_skip_utf8 Fail=f Ms=x u u => i_bs_get_utf8 Ms Fail x
 
 bs_get_utf16 Fail=f Ms=x u Flags=u Dst=d => i_bs_get_utf16 Ms Fail Flags Dst
 bs_skip_utf16 Fail=f Ms=x u Flags=u => i_bs_get_utf16 Ms Fail Flags x
 
-i_bs_get_utf16 x f I d
+i_bs_get_utf16 x f? t x
 
 bs_get_utf32 Fail=f Ms=x Live=u Flags=u Dst=d => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags Dst | \
@@ -1238,22 +1172,18 @@ bs_skip_utf32 Fail=f Ms=x Live=u Flags=u => \
 	bs_get_integer2 Fail Ms Live i=32 u=1 Flags x | \
 	i_bs_validate_unicode_retract Fail x Ms
 
-i_bs_validate_unicode_retract j s s
+i_bs_validate_unicode_retract j s S
 %hot
 
 #
 # Constructing binaries
 #
-%cold
+%warm
 
 bs_init2 Fail Sz Words Regs Flags Dst | binary_too_big(Sz) => system_limit Fail
 
-bs_init2 Fail Sz=u Words=u==0 Regs Flags Dst | should_gen_heap_bin(Sz) => \
-	i_bs_init_heap_bin Sz Regs Dst
 bs_init2 Fail Sz=u Words=u==0 Regs Flags Dst => i_bs_init Sz Regs Dst
 
-bs_init2 Fail Sz=u Words Regs Flags Dst | should_gen_heap_bin(Sz) => \
-	i_bs_init_heap_bin_heap Sz Words Regs Dst
 bs_init2 Fail Sz=u Words Regs Flags Dst => \
    i_bs_init_heap Sz Words Regs Dst
 
@@ -1262,15 +1192,13 @@ bs_init2 Fail Sz Words=u==0 Regs Flags Dst => \
 bs_init2 Fail Sz Words Regs Flags Dst => \
   i_bs_init_fail_heap Sz Words Fail Regs Dst
 
-i_bs_init_fail xy j I d
+i_bs_init_fail xy j? t? x
 
-i_bs_init_fail_heap s I j I d
+i_bs_init_fail_heap s I j? t? x
 
-i_bs_init I I d
-i_bs_init_heap_bin I I d
+i_bs_init W t? x
 
-i_bs_init_heap I I I d
-i_bs_init_heap_bin_heap I I I d
+i_bs_init_heap W I t? x
 
 
 bs_init_bits Fail Sz=o Words Regs Flags Dst => system_limit Fail
@@ -1283,16 +1211,16 @@ bs_init_bits Fail Sz Words=u==0 Regs Flags Dst => \
 bs_init_bits Fail Sz Words Regs Flags Dst => \
   i_bs_init_bits_fail_heap Sz Words Fail Regs Dst
 
-i_bs_init_bits_fail xy j I d
+i_bs_init_bits_fail xy j? t? x
 
-i_bs_init_bits_fail_heap s I j I d
+i_bs_init_bits_fail_heap s I j? t? x
 
-i_bs_init_bits I I d
-i_bs_init_bits_heap I I I d
+i_bs_init_bits W t? x
+i_bs_init_bits_heap W I t? x
 
 bs_add Fail S1=i==0 S2 Unit=u==1 D => move S2 D
 
-bs_add j s s I d
+bs_add j? s s t? x
 
 bs_append Fail Size Extra Live Unit Bin Flags Dst => \
   move Bin x | i_bs_append Fail Extra Live Unit Size Dst
@@ -1302,8 +1230,8 @@ bs_private_append Fail Size Unit Bin Flags Dst => \
 
 bs_init_writable
 
-i_bs_append j I I I s d
-i_bs_private_append j I s s d
+i_bs_append j? I t? t s x
+i_bs_private_append j? t s S x
 
 #
 # Storing integers into binaries.
@@ -1312,11 +1240,8 @@ i_bs_private_append j I s s d
 bs_put_integer Fail=j Sz=sq Unit=u Flags=u Src=s => \
 			gen_put_integer(Fail, Sz, Unit, Flags, Src)
 
-%macro: i_new_bs_put_integer NewBsPutInteger
-%macro: i_new_bs_put_integer_imm NewBsPutIntegerImm
-
-i_new_bs_put_integer j s I s
-i_new_bs_put_integer_imm j I I s
+i_new_bs_put_integer j? s t s
+i_new_bs_put_integer_imm j? W t s
 
 #
 # Utf8/utf16/utf32 support. (R12B-5)
@@ -1324,22 +1249,22 @@ i_new_bs_put_integer_imm j I I s
 
 bs_utf8_size j Src=s Dst=d => i_bs_utf8_size Src Dst
 
-i_bs_utf8_size s d
+i_bs_utf8_size s x
 
 bs_utf16_size j Src=s Dst=d => i_bs_utf16_size Src Dst
 
-i_bs_utf16_size s d
+i_bs_utf16_size s x
 
 bs_put_utf8 Fail u Src=s => i_bs_put_utf8 Fail Src
 
-i_bs_put_utf8 j s
+i_bs_put_utf8 j? s
 
-bs_put_utf16 j I s
+bs_put_utf16 j? t s
 
 bs_put_utf32 Fail=j Flags=u Src=s => \
    i_bs_validate_unicode Fail Src | bs_put_integer Fail i=32 u=1 Flags Src
 
-i_bs_validate_unicode j s
+i_bs_validate_unicode j? s
 
 #
 # Storing floats into binaries.
@@ -1349,11 +1274,8 @@ bs_put_float Fail Sz=q Unit Flags Val => badarg Fail
 bs_put_float Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_float(Fail, Sz, Unit, Flags, Src)
 
-%macro: i_new_bs_put_float NewBsPutFloat
-%macro: i_new_bs_put_float_imm NewBsPutFloatImm
-
-i_new_bs_put_float j s I s
-i_new_bs_put_float_imm j I I s
+i_new_bs_put_float j? s t s
+i_new_bs_put_float_imm j? W t s
 
 #
 # Storing binaries into binaries.
@@ -1362,14 +1284,9 @@ i_new_bs_put_float_imm j I I s
 bs_put_binary Fail=j Sz=s Unit=u Flags=u Src=s => \
 			gen_put_binary(Fail, Sz, Unit, Flags, Src)
 
-%macro: i_new_bs_put_binary NewBsPutBinary
-i_new_bs_put_binary j s I s
-
-%macro: i_new_bs_put_binary_imm NewBsPutBinaryImm
-i_new_bs_put_binary_imm j I s
-
-%macro: i_new_bs_put_binary_all NewBsPutBinaryAll
-i_new_bs_put_binary_all j s I
+i_new_bs_put_binary j? s t s
+i_new_bs_put_binary_imm j? W s
+i_new_bs_put_binary_all j? s t
 
 #
 # Warning: The i_bs_put_string and i_new_bs_put_string instructions
@@ -1377,9 +1294,7 @@ i_new_bs_put_binary_all j s I
 # Don't change the instruction format unless you change the loader too.
 #
 
-bs_put_string I I
-
-%hot
+bs_put_string W W
 
 #
 # New floating point instructions (R8).
@@ -1393,11 +1308,13 @@ fnegate p FR1 FR2 => i_fnegate FR1 FR2
 
 fconv Arg=iqan Dst=l => move Arg x | fconv x Dst
 
-fmove q l
-fmove d l
-fmove l d
+fmove Arg=l Dst=d => fstore Arg Dst
+fmove Arg=dq Dst=l => fload Arg Dst
+
+fstore l d
+fload Sq l
 
-fconv d l
+fconv S l
 
 i_fadd l l l
 i_fsub l l l
@@ -1407,51 +1324,88 @@ i_fnegate l l
 
 fclearerror | no_fpe_signals() =>
 fcheckerror p | no_fpe_signals() =>
+
+%unless NO_FPE_SIGNALS
 fcheckerror p => i_fcheckerror
 
 i_fcheckerror
 fclearerror
+%endif
+
+%hot
 
 #
 # New apply instructions in R10B.
 #
 
-apply I
-apply_last I P
+apply t
+apply_last t Q
+
+#
+# Handle compatibility with OTP 17 here.
+#
+
+i_put_map_assoc/4
+
+# We KNOW that in OTP 20 (actually OTP 18 and higher), a put_map_assoc instruction
+# is always preceded by an is_map test. That means that put_map_assoc can never
+# fail and does not need any failure label.
+
+put_map_assoc Fail Map Dst Live Size Rest=* | compiled_with_otp_20_or_higher() => \
+	      i_put_map_assoc Map Dst Live Size Rest
+
+# Translate the put_map_assoc instruction if the module was compiled by a compiler
+# before 20. This is only necessary if the OTP 17 compiler was used, but we
+# have no safe and relatively easy way to know whether OTP 18/19 was used.
+
+put_map_assoc Fail=p Map Dst Live Size Rest=* => \
+	      ensure_map Map | i_put_map_assoc Map Dst Live Size Rest
+put_map_assoc Fail=f Map Dst Live Size Rest=* => \
+	      is_map Fail Map | i_put_map_assoc Map Dst Live Size Rest
+
+ensure_map Lit=q | literal_is_map(Lit) =>
+ensure_map Src=cqy => move Src x | ensure_map x
+
+%cold
+ensure_map x
+%hot
 
 #
-# Map instructions in R17.
+# Map instructions. First introduced in R17.
 #
 
-sorted_put_map_assoc/5
-put_map_assoc F Map Dst Live Size Rest=* | map_key_sort(Size, Rest) => \
-  sorted_put_map_assoc F Map Dst Live Size Rest
+sorted_put_map_assoc/4
+i_put_map_assoc Map Dst Live Size Rest=* | map_key_sort(Size, Rest) => \
+  sorted_put_map_assoc Map Dst Live Size Rest
 
 sorted_put_map_exact/5
 put_map_exact F Map Dst Live Size Rest=* | map_key_sort(Size, Rest) => \
   sorted_put_map_exact F Map Dst Live Size Rest
 
-sorted_put_map_assoc j Map Dst Live Size Rest=* | is_empty_map(Map) => \
+sorted_put_map_assoc Map Dst Live Size Rest=* | is_empty_map(Map) => \
    new_map Dst Live Size Rest
-sorted_put_map_assoc F Src=s Dst Live Size Rest=* => \
-	       update_map_assoc F Src Dst Live Size Rest
-sorted_put_map_assoc F Src Dst Live Size Rest=* => \
-	       move Src x | update_map_assoc F x Dst Live Size Rest
+sorted_put_map_assoc Src=s Dst Live Size Rest=* => \
+	       update_map_assoc Src Dst Live Size Rest
+sorted_put_map_assoc Src Dst Live Size Rest=* => \
+	       move Src x | update_map_assoc x Dst Live Size Rest
 
 sorted_put_map_exact F Src=s Dst Live Size Rest=* => \
 	      update_map_exact F Src Dst Live Size Rest
 sorted_put_map_exact F Src Dst Live Size Rest=* => \
 	      move Src x | update_map_exact F x Dst Live Size Rest
 
-new_map d I I
-update_map_assoc j s d I I
-update_map_exact j s d I I
+new_map Dst Live Size Rest=* | is_small_map_literal_keys(Size, Rest) => \
+   gen_new_small_map_lit(Dst, Live, Size, Rest)
+
+new_map d t I
+i_new_small_map_lit d t q
+update_map_assoc s d t I
+update_map_exact j? s d t I
 
 is_map Fail Lit=q | literal_is_map(Lit) =>
 is_map Fail cq => jump Fail
 
-%macro: is_map IsMap -fail_action
-is_map f xy
+is_map f? xy
 
 ## Transform has_map_fields #{ K1 := _, K2 := _ } to has_map_elements
 
@@ -1460,21 +1414,23 @@ has_map_fields Fail Src Size Rest=* => \
 
 ## Transform get_map_elements(s) #{ K1 := V1, K2 := V2 }
 
-get_map_elements Fail Src=xy Size=u==2 Rest=* => \
+get_map_elements Fail Src Size=u==2 Rest=* => \
     gen_get_map_element(Fail, Src, Size, Rest)
 get_map_elements Fail Src Size Rest=* | map_key_sort(Size, Rest) => \
    gen_get_map_elements(Fail, Src, Size, Rest)
 
-i_get_map_elements f s I
+i_get_map_elements f? s I
 
 i_get_map_element Fail Src=xy Key=y Dst => \
     move Key x | i_get_map_element Fail Src x Dst
 
-%macro: i_get_map_element_hash GetMapElementHash -fail_action
-i_get_map_element_hash f xy c I xy
+i_get_map_element_hash Fail Src=c Key Hash Dst => \
+    move Src x | i_get_map_element_hash Fail x Key Hash Dst
+i_get_map_element_hash f? xy c I xy
 
-%macro: i_get_map_element GetMapElement -fail_action
-i_get_map_element f xy x xy
+i_get_map_element Fail Src=c Key Dst => \
+    move Src x | i_get_map_element Fail x Key Dst
+i_get_map_element f? xy x xy
 
 #
 # Convert the plus operations to a generic plus instruction.
@@ -1509,9 +1465,9 @@ gen_minus p Live Reg=d Int=i Dst | negation_is_small(Int) => \
 # GCing arithmetic instructions.
 #
 
-gen_plus Fail Live S1 S2 Dst => i_plus Fail Live S1 S2 Dst
+gen_plus Fail Live S1 S2 Dst => i_plus S1 S2 Fail Live Dst
 
-gen_minus Fail Live S1 S2 Dst => i_minus Fail Live S1 S2 Dst
+gen_minus Fail Live S1 S2 Dst => i_minus S1 S2 Fail Live Dst
 
 gc_bif2 Fail Live u$bif:erlang:stimes/2 S1 S2 Dst => \
   i_times Fail Live S1 S2 Dst
@@ -1522,15 +1478,15 @@ gc_bif2 Fail Live u$bif:erlang:intdiv/2 S1 S2 Dst => \
   i_int_div Fail Live S1 S2 Dst
 
 gc_bif2 Fail Live u$bif:erlang:rem/2 S1 S2 Dst => \
-  i_rem Fail Live S1 S2 Dst
+  i_rem S1 S2 Fail Live Dst
 
 gc_bif2 Fail Live u$bif:erlang:bsl/2 S1 S2 Dst => \
-  i_bsl Fail Live S1 S2 Dst
+  i_bsl S1 S2 Fail Live Dst
 gc_bif2 Fail Live u$bif:erlang:bsr/2 S1 S2 Dst => \
-  i_bsr Fail Live S1 S2 Dst
+  i_bsr S1 S2 Fail Live Dst
 
 gc_bif2 Fail Live u$bif:erlang:band/2 S1 S2 Dst => \
-  i_band Fail Live S1 S2 Dst
+  i_band S1 S2 Fail Live Dst
 
 gc_bif2 Fail Live u$bif:erlang:bor/2 S1 S2 Dst => \
   i_bor Fail Live S1 S2 Dst
@@ -1540,32 +1496,34 @@ gc_bif2 Fail Live u$bif:erlang:bxor/2 S1 S2 Dst => \
 
 gc_bif1 Fail I u$bif:erlang:bnot/1 Src Dst=d => i_int_bnot Fail Src I Dst
 
-i_increment rxy I I d
+i_increment rxy W t d
+
+i_plus x xy j? t d
+i_plus s s  j? t d
 
-i_plus j I x xy d
-i_plus j I s s d
+i_minus x x j? t d
+i_minus s s j? t d
 
-i_minus j I x x d
-i_minus j I s s d
+i_times j? t s s d
 
-i_times j I s s d
+i_m_div j? t s s d
+i_int_div j? t s s d
 
-i_m_div j I s s d
-i_int_div j I s s d
+i_rem x x j? t d
+i_rem s s j? t d
 
-i_rem j I x x d
-i_rem j I s s d
+i_bsl s s j? t d
+i_bsr s s j? t d
 
-i_bsl j I s s d
-i_bsr j I s s d
+i_band x c j? t d
+i_band s s j? t d
 
-i_band j I x c d
-i_band j I s s d
+i_bor j? I s s d
+i_bxor j? I s s d
 
-i_bor j I s s d
-i_bxor j I s s d
+i_int_bnot Fail Src=c Live Dst => move Src x | i_int_bnot Fail x Live Dst
 
-i_int_bnot j s I d
+i_int_bnot j? S t d
 
 #
 # Old guard BIFs that creates heap fragments are no longer allowed.
@@ -1589,9 +1547,9 @@ gc_bif2 Fail I Bif S1 S2 Dst => \
 gc_bif3 Fail I Bif S1 S2 S3 Dst => \
 	gen_guard_bif3(Fail, I, Bif, S1, S2, S3, Dst)
 
-i_gc_bif1 j I s I d
+i_gc_bif1 j? W s t? d
 
-i_gc_bif2 j I I s s d
+i_gc_bif2 j? W t? s s d
 
 ii_gc_bif3/7
 
@@ -1600,7 +1558,7 @@ ii_gc_bif3/7
 ii_gc_bif3 Fail Bif Live S1 S2 S3 Dst => \
   move S1 x | i_gc_bif3 Fail Bif Live S2 S3 Dst
 
-i_gc_bif3 j I I s s d
+i_gc_bif3 j? W t? s s d
 
 #
 # The following instruction is specially handled in beam_load.c
@@ -1618,8 +1576,20 @@ on_load
 #
 # R14A.
 #
-recv_mark f
+# Modified in OTP 21 because it turns out that we don't need the
+# label after all.
+#
+
+recv_mark f => i_recv_mark
+i_recv_mark
 
 recv_set Fail | label Lbl | loop_rec Lf Reg => \
    i_recv_set | label Lbl | loop_rec Lf Reg
 i_recv_set
+
+#
+# OTP 21.
+#
+
+build_stacktrace
+raw_raise
diff --git a/erts/emulator/beam/packet_parser.c b/erts/emulator/beam/packet_parser.c
index f14910bc72..de1d481105 100644
--- a/erts/emulator/beam/packet_parser.c
+++ b/erts/emulator/beam/packet_parser.c
@@ -200,7 +200,7 @@ static int http_init(void)
     for (i = 0; i < HTTP_HDR_HASH_SIZE; i++)
         http_hdr_hash[i] = NULL;
     for (i = 0; http_hdr_strings[i] != NULL; i++) {
-        ASSERT(strlen(http_hdr_strings[i]) <= HTTP_MAX_NAME_LEN);
+        ASSERT(sys_strlen(http_hdr_strings[i]) <= HTTP_MAX_NAME_LEN);
         http_hdr_table[i].index = i;
         http_hash_insert(http_hdr_strings[i], 
                          &http_hdr_table[i], 
@@ -516,7 +516,7 @@ static http_atom_t* http_hash_lookup(const char* name, int len,
 
     while (ap != NULL) {
         if ((ap->h == h) && (ap->len == len) && 
-            (strncmp(ap->name, name, len) == 0))
+            (sys_strncmp(ap->name, name, len) == 0))
             return ap;
         ap = ap->next;
     }
@@ -656,7 +656,7 @@ int packet_parse_http(const char* buf, int len, int* statep,
     if (*statep == 0) {
         /* start-line = Request-Line | Status-Line */
 
-        if (n >= 5 && (strncmp(buf, "HTTP/", 5) == 0)) {
+        if (n >= 5 && (sys_strncmp(buf, "HTTP/", 5) == 0)) {
             int major  = 0;
             int minor  = 0;
             int status = 0;
@@ -750,7 +750,7 @@ int packet_parse_http(const char* buf, int len, int* statep,
             }
             if (n < 8)
                 return -1;
-            if (strncmp(ptr, "HTTP/", 5) != 0)
+            if (sys_strncmp(ptr, "HTTP/", 5) != 0)
                 return -1;
             ptr += 5;
             n   -= 5;
diff --git a/erts/emulator/beam/register.c b/erts/emulator/beam/register.c
index bf3267cff1..92a0854ad3 100644
--- a/erts/emulator/beam/register.c
+++ b/erts/emulator/beam/register.c
@@ -38,16 +38,15 @@ static Hash process_reg;
 
 #define REG_HASH(term) ((HashValue) atom_val(term))
 
-static erts_smp_rwmtx_t regtab_rwmtx;
+static erts_rwmtx_t regtab_rwmtx;
 
-#define reg_try_read_lock()		erts_smp_rwmtx_tryrlock(&regtab_rwmtx)
-#define reg_try_write_lock()		erts_smp_rwmtx_tryrwlock(&regtab_rwmtx)
-#define reg_read_lock()			erts_smp_rwmtx_rlock(&regtab_rwmtx)
-#define reg_write_lock()		erts_smp_rwmtx_rwlock(&regtab_rwmtx)
-#define reg_read_unlock()		erts_smp_rwmtx_runlock(&regtab_rwmtx)
-#define reg_write_unlock()		erts_smp_rwmtx_rwunlock(&regtab_rwmtx)
+#define reg_try_read_lock()		erts_rwmtx_tryrlock(&regtab_rwmtx)
+#define reg_try_write_lock()		erts_rwmtx_tryrwlock(&regtab_rwmtx)
+#define reg_read_lock()			erts_rwmtx_rlock(&regtab_rwmtx)
+#define reg_write_lock()		erts_rwmtx_rwlock(&regtab_rwmtx)
+#define reg_read_unlock()		erts_rwmtx_runlock(&regtab_rwmtx)
+#define reg_write_unlock()		erts_rwmtx_rwunlock(&regtab_rwmtx)
 
-#ifdef ERTS_SMP
 static ERTS_INLINE void
 reg_safe_read_lock(Process *c_p, ErtsProcLocks *c_p_locks)
 {
@@ -64,7 +63,7 @@ reg_safe_read_lock(Process *c_p, ErtsProcLocks *c_p_locks)
 	}
 
 	/* Release process locks in order to avoid deadlock */
-	erts_smp_proc_unlock(c_p, *c_p_locks);
+	erts_proc_unlock(c_p, *c_p_locks);
 	*c_p_locks = 0;
     }
 
@@ -87,14 +86,13 @@ reg_safe_write_lock(Process *c_p, ErtsProcLocks *c_p_locks)
 	}
 
 	/* Release process locks in order to avoid deadlock */
-	erts_smp_proc_unlock(c_p, *c_p_locks);
+	erts_proc_unlock(c_p, *c_p_locks);
 	*c_p_locks = 0;
     }
 
     reg_write_lock();
 }
 
-#endif
 
 static ERTS_INLINE int
 is_proc_alive(Process *p)
@@ -141,11 +139,11 @@ static void reg_free(RegProc *obj)
 void init_register_table(void)
 {
     HashFunctions f;
-    erts_smp_rwmtx_opt_t rwmtx_opt = ERTS_SMP_RWMTX_OPT_DEFAULT_INITER;
-    rwmtx_opt.type = ERTS_SMP_RWMTX_TYPE_FREQUENT_READ;
-    rwmtx_opt.lived = ERTS_SMP_RWMTX_LONG_LIVED;
+    erts_rwmtx_opt_t rwmtx_opt = ERTS_RWMTX_OPT_DEFAULT_INITER;
+    rwmtx_opt.type = ERTS_RWMTX_TYPE_FREQUENT_READ;
+    rwmtx_opt.lived = ERTS_RWMTX_LONG_LIVED;
 
-    erts_smp_rwmtx_init_opt(&regtab_rwmtx, &rwmtx_opt, "reg_tab", NIL,
+    erts_rwmtx_init_opt(&regtab_rwmtx, &rwmtx_opt, "reg_tab", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
     f.hash = (H_FUN) reg_hash;
@@ -175,7 +173,7 @@ int erts_register_name(Process *c_p, Eterm name, Eterm id)
     Process *proc = NULL;
     Port *port = NULL;
     RegProc r, *rp;
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
 
     if (is_not_atom(name) || name == am_undefined)
 	return res;
@@ -185,7 +183,7 @@ int erts_register_name(Process *c_p, Eterm name, Eterm id)
     else {
 	if (is_not_internal_pid(id) && is_not_internal_port(id))
 	    return res;
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
+	erts_proc_unlock(c_p, ERTS_PROC_LOCK_MAIN);
 	if (is_internal_port(id)) {
 	    port = erts_id2port(id);
 	    if (!port)
@@ -193,15 +191,13 @@ int erts_register_name(Process *c_p, Eterm name, Eterm id)
 	}
     }
 
-#ifdef ERTS_SMP
     {
 	ErtsProcLocks proc_locks = proc ? ERTS_PROC_LOCK_MAIN : 0;
 	reg_safe_write_lock(proc, &proc_locks);
 
 	if (proc && !proc_locks)
-	    erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
     }
-#endif
 
     if (is_internal_pid(id)) {
 	if (!proc)
@@ -215,7 +211,7 @@ int erts_register_name(Process *c_p, Eterm name, Eterm id)
     }
     else {
 	ASSERT(!INVALID_PORT(port, id));
-	ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(port));
+	ERTS_LC_ASSERT(erts_lc_is_port_locked(port));
 	r.pt = port;
 	if (r.pt->common.u.alive.reg)
 	    goto done;
@@ -250,8 +246,8 @@ int erts_register_name(Process *c_p, Eterm name, Eterm id)
 	erts_port_release(port);
     if (c_p != proc) {
 	if (proc)
-	    erts_smp_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
+	    erts_proc_unlock(proc, ERTS_PROC_LOCK_MAIN);
+	erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
     }
     return res;
 }
@@ -272,17 +268,15 @@ erts_whereis_name_to_id(Process *c_p, Eterm name)
     HashValue hval;
     int ix;
     HashBucket* b;
-#ifdef ERTS_SMP
     ErtsProcLocks c_p_locks = 0;
     if (c_p) {
         c_p_locks = ERTS_PROC_LOCK_MAIN;
-        ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
+        ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(c_p);
     }
     reg_safe_read_lock(c_p, &c_p_locks);
 
     if (c_p && !c_p_locks)
-        erts_smp_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
-#endif
+        erts_proc_lock(c_p, ERTS_PROC_LOCK_MAIN);
 
     hval = REG_HASH(name);
     ix = hval % process_reg.size;
@@ -331,7 +325,6 @@ erts_whereis_name(Process *c_p,
     HashValue hval;
     int ix;
     HashBucket* b;
-#ifdef ERTS_SMP
     ErtsProcLocks current_c_p_locks;
     Port *pending_port = NULL;
 
@@ -348,7 +341,6 @@ erts_whereis_name(Process *c_p,
      * - read reg lock
      * - current_c_p_locks (either c_p_locks or 0) on c_p
      */
-#endif
 
     hval = REG_HASH(name);
     ix = hval % process_reg.size;
@@ -370,7 +362,6 @@ erts_whereis_name(Process *c_p,
 	if (!rp)
 	    *proc = NULL;
 	else {
-#ifdef ERTS_SMP
 	    if (!rp->p)
 		*proc = NULL;
 	    else {
@@ -387,17 +378,10 @@ erts_whereis_name(Process *c_p,
 		    *proc = rp->p;
 		else {
 		    if (need_locks)
-			erts_smp_proc_unlock(rp->p, need_locks);
+			erts_proc_unlock(rp->p, need_locks);
 		    *proc = NULL;
 		}
 	    }
-#else
-	    if (rp->p
-		&& ((flags & ERTS_P2P_FLG_ALLOW_OTHER_X) || is_proc_alive(rp->p)))
-		*proc = rp->p;
-	    else
-		*proc = NULL;
-#endif
 	    if (*proc && (flags & ERTS_P2P_FLG_INC_REFC))
 		erts_proc_inc_refc(*proc);
 	}
@@ -407,7 +391,6 @@ erts_whereis_name(Process *c_p,
 	if (!rp || !rp->pt)
 	    *port = NULL;
 	else {
-#ifdef ERTS_SMP
             if (lock_port) {
                 if (pending_port == rp->pt)
                     pending_port = NULL;
@@ -419,11 +402,11 @@ erts_whereis_name(Process *c_p,
                         pending_port = NULL;
                     }
 		    
-                    if (erts_smp_port_trylock(rp->pt) == EBUSY) {
+                    if (erts_port_trylock(rp->pt) == EBUSY) {
                         Eterm id = rp->pt->common.id; /* id read only... */
                         /* Unlock all locks, acquire port lock, and restart... */
                         if (current_c_p_locks) {
-                            erts_smp_proc_unlock(c_p, current_c_p_locks);
+                            erts_proc_unlock(c_p, current_c_p_locks);
                             current_c_p_locks = 0;
                         }
                         reg_read_unlock();
@@ -431,19 +414,16 @@ erts_whereis_name(Process *c_p,
                         goto restart;
                     }
                 }
-                ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(rp->pt));
+                ERTS_LC_ASSERT(erts_lc_is_port_locked(rp->pt));
             }
-#endif
 	    *port = rp->pt;
 	}
     }
 
-#ifdef ERTS_SMP
     if (c_p && !current_c_p_locks)
-	erts_smp_proc_lock(c_p, c_p_locks);
+	erts_proc_lock(c_p, c_p_locks);
     if (pending_port)
 	erts_port_release(pending_port);
-#endif
 
     reg_read_unlock();
 }
@@ -476,7 +456,6 @@ int erts_unregister_name(Process *c_p,
     RegProc r, *rp;
     Port *port = c_prt;
     ErtsProcLocks current_c_p_locks = 0;
-#ifdef ERTS_SMP
 
     /*
      * SMP note: If 'c_prt != NULL' and 'c_prt->reg->name == name',
@@ -492,18 +471,15 @@ int erts_unregister_name(Process *c_p,
  restart:
 
     reg_safe_write_lock(c_p, &current_c_p_locks);
-#endif
 
     r.name = name;
     if (is_non_value(name)) {
 	/* Unregister current process name */
 	ASSERT(c_p);
-#ifdef ERTS_SMP
 	if (current_c_p_locks != c_p_locks) {
-	    erts_smp_proc_lock(c_p, c_p_locks);
+	    erts_proc_lock(c_p, c_p_locks);
 	    current_c_p_locks = c_p_locks;
 	}
-#endif
 	if (c_p->common.u.alive.reg) {
 	    r.name = c_p->common.u.alive.reg->name;
 	} else {
@@ -516,36 +492,34 @@ int erts_unregister_name(Process *c_p,
     if ((rp = (RegProc*) hash_get(&process_reg, (void*) &r)) != NULL) {
 	if (rp->pt) {
 	    if (port != rp->pt) {
-#ifdef ERTS_SMP
 		if (port) {
 		    ASSERT(port != c_prt);
 		    erts_port_release(port);
 		    port = NULL;
 		}
 
-		if (erts_smp_port_trylock(rp->pt) == EBUSY) {
+		if (erts_port_trylock(rp->pt) == EBUSY) {
 		    Eterm id = rp->pt->common.id; /* id read only... */
 		    /* Unlock all locks, acquire port lock, and restart... */
 		    if (current_c_p_locks) {
-			erts_smp_proc_unlock(c_p, current_c_p_locks);
+			erts_proc_unlock(c_p, current_c_p_locks);
 			current_c_p_locks = 0;
 		    }
 		    reg_write_unlock();
 		    port = erts_id2port(id);
 		    goto restart;
 		}
-#endif
 		port = rp->pt;
 	    }
 
 	    ASSERT(rp->pt == port);
-	    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(port));
+	    ERTS_LC_ASSERT(erts_lc_is_port_locked(port));
 
 	    rp->pt->common.u.alive.reg = NULL;
 
 	    if (IS_TRACED_FL(port, F_TRACE_PORTS)) {
                 if (current_c_p_locks) {
-                    erts_smp_proc_unlock(c_p, current_c_p_locks);
+                    erts_proc_unlock(c_p, current_c_p_locks);
                     current_c_p_locks = 0;
                 }
 		trace_port(port, am_unregister, r.name);
@@ -553,7 +527,6 @@ int erts_unregister_name(Process *c_p,
 
 	} else if (rp->p) {
 
-#ifdef ERTS_SMP
 	    erts_proc_safelock(c_p,
 			       current_c_p_locks,
 			       c_p_locks,
@@ -561,17 +534,14 @@ int erts_unregister_name(Process *c_p,
 			       (c_p == rp->p) ?  current_c_p_locks : 0,
 			       ERTS_PROC_LOCK_MAIN);
 	    current_c_p_locks = c_p_locks;
-#endif
 	    rp->p->common.u.alive.reg = NULL;
 	    if (IS_TRACED_FL(rp->p, F_TRACE_PROCS)) {
                 trace_proc(rp->p, (c_p == rp->p) ? c_p_locks : ERTS_PROC_LOCK_MAIN,
                            rp->p, am_unregister, r.name);
 	    }
-#ifdef ERTS_SMP
 	    if (rp->p != c_p) {
-		erts_smp_proc_unlock(rp->p, ERTS_PROC_LOCK_MAIN);
+		erts_proc_unlock(rp->p, ERTS_PROC_LOCK_MAIN);
 	    }
-#endif
 	}
 	hash_erase(&process_reg, (void*) &r);
 	res = 1;
@@ -585,14 +555,12 @@ int erts_unregister_name(Process *c_p,
 	    erts_port_release(port);
 	}
 	if (c_prt) {
-	    erts_smp_port_lock(c_prt);
+	    erts_port_lock(c_prt);
 	}
     }
-#ifdef ERTS_SMP
     if (c_p && !current_c_p_locks) {
-	erts_smp_proc_lock(c_p, c_p_locks);
+	erts_proc_lock(c_p, c_p_locks);
     }
-#endif
     return res;
 }
 
@@ -633,14 +601,12 @@ BIF_RETTYPE registered_0(BIF_ALIST_0)
     Uint need;
     Eterm* hp;
     HashBucket **bucket;
-#ifdef ERTS_SMP
     ErtsProcLocks proc_locks = ERTS_PROC_LOCK_MAIN;
 
-    ERTS_SMP_CHK_HAVE_ONLY_MAIN_PROC_LOCK(BIF_P);
+    ERTS_CHK_HAVE_ONLY_MAIN_PROC_LOCK(BIF_P);
     reg_safe_read_lock(BIF_P, &proc_locks);
     if (!proc_locks)
-	erts_smp_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
-#endif
+	erts_proc_lock(BIF_P, ERTS_PROC_LOCK_MAIN);
 
     bucket = process_reg.bucket;
 
diff --git a/erts/emulator/beam/safe_hash.c b/erts/emulator/beam/safe_hash.c
index 527c9efeca..73306030ae 100644
--- a/erts/emulator/beam/safe_hash.c
+++ b/erts/emulator/beam/safe_hash.c
@@ -62,7 +62,7 @@ static ERTS_INLINE int align_up_pow2(int val)
 */
 static void rehash(SafeHash* h, int grow_limit)
 {
-    if (erts_smp_atomic_xchg_acqb(&h->is_rehashing, 1) != 0) {        
+    if (erts_atomic_xchg_acqb(&h->is_rehashing, 1) != 0) {
 	return; /* already in progress */
     }
     if (h->grow_limit == grow_limit) {
@@ -77,7 +77,7 @@ static void rehash(SafeHash* h, int grow_limit)
 	sys_memzero(new_tab, bytes);
 
 	for (i=0; i<SAFE_HASH_LOCK_CNT; i++) { /* stop all traffic */
-	    erts_smp_mtx_lock(&h->lock_vec[i].mtx);
+	    erts_mtx_lock(&h->lock_vec[i].mtx);
 	}
 
 	h->tab = new_tab;
@@ -95,12 +95,12 @@ static void rehash(SafeHash* h, int grow_limit)
 	}
 
 	for (i=0; i<SAFE_HASH_LOCK_CNT; i++) {
-	    erts_smp_mtx_unlock(&h->lock_vec[i].mtx);
+	    erts_mtx_unlock(&h->lock_vec[i].mtx);
 	}
 	erts_free(h->type, (void *) old_tab);
     }
     /*else already done */
-    erts_smp_atomic_set_relb(&h->is_rehashing, 0);
+    erts_atomic_set_relb(&h->is_rehashing, 0);
 }
 
 
@@ -115,7 +115,7 @@ void safe_hash_get_info(SafeHashInfo *hi, SafeHash *h)
     int objects = 0;
 
     for (lock_ix=0; lock_ix<SAFE_HASH_LOCK_CNT; lock_ix++) {
-	erts_smp_mtx_lock(&h->lock_vec[lock_ix].mtx); 
+	erts_mtx_lock(&h->lock_vec[lock_ix].mtx);
 	size = h->size_mask + 1;
 	for (i = lock_ix; i < size; i += SAFE_HASH_LOCK_CNT) {
 	    int depth = 0;
@@ -128,7 +128,7 @@ void safe_hash_get_info(SafeHashInfo *hi, SafeHash *h)
 	    if (depth > max_depth)
 		max_depth = depth;
 	}
-	erts_smp_mtx_unlock(&h->lock_vec[lock_ix].mtx);	
+	erts_mtx_unlock(&h->lock_vec[lock_ix].mtx);
     }
 
     hi->name  = h->name;
@@ -145,9 +145,9 @@ int safe_hash_table_sz(SafeHash *h)
   int i, size;
   for(i=0; h->name[i]; i++);
   i++;
-  erts_smp_mtx_lock(&h->lock_vec[0].mtx); /* any lock will do to read size */
+  erts_mtx_lock(&h->lock_vec[0].mtx); /* any lock will do to read size */
   size = h->size_mask + 1;
-  erts_smp_mtx_unlock(&h->lock_vec[0].mtx);
+  erts_mtx_unlock(&h->lock_vec[0].mtx);
   return sizeof(SafeHash) + size*sizeof(SafeHashBucket*) + i;
 }
 
@@ -168,10 +168,10 @@ SafeHash* safe_hash_init(ErtsAlcType_t type, SafeHash* h, char* name, erts_lock_
     h->name = name;
     h->fun = fun;
     set_size(h,size);
-    erts_smp_atomic_init_nob(&h->is_rehashing, 0);
-    erts_smp_atomic_init_nob(&h->nitems, 0);
+    erts_atomic_init_nob(&h->is_rehashing, 0);
+    erts_atomic_init_nob(&h->nitems, 0);
     for (i=0; i<SAFE_HASH_LOCK_CNT; i++) {
-        erts_smp_mtx_init(&h->lock_vec[i].mtx, "safe_hash", NIL,
+        erts_mtx_init(&h->lock_vec[i].mtx, "safe_hash", NIL,
             flags);
     }
     return h;
@@ -185,8 +185,8 @@ void* safe_hash_get(SafeHash* h, void* tmpl)
 {
     SafeHashValue hval = h->fun.hash(tmpl);
     SafeHashBucket* b;
-    erts_smp_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
-    erts_smp_mtx_lock(lock);
+    erts_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
+    erts_mtx_lock(lock);
     b = h->tab[hval & h->size_mask];
 	
     while(b != NULL) {
@@ -194,7 +194,7 @@ void* safe_hash_get(SafeHash* h, void* tmpl)
 	    break;
 	b = b->next;
     }
-    erts_smp_mtx_unlock(lock);
+    erts_mtx_unlock(lock);
     return (void*) b;
 }
 
@@ -207,13 +207,13 @@ void* safe_hash_put(SafeHash* h, void* tmpl)
     SafeHashValue hval = h->fun.hash(tmpl);
     SafeHashBucket* b;
     SafeHashBucket** head;
-    erts_smp_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
-    erts_smp_mtx_lock(lock);
+    erts_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
+    erts_mtx_lock(lock);
     head = &h->tab[hval & h->size_mask];
     b = *head;
     while(b != NULL) {
 	if ((b->hvalue == hval) && (h->fun.cmp(tmpl, (void*)b) == 0)) {
-	    erts_smp_mtx_unlock(lock);
+	    erts_mtx_unlock(lock);
 	    return b;
 	}
 	b = b->next;
@@ -224,8 +224,8 @@ void* safe_hash_put(SafeHash* h, void* tmpl)
     b->next = *head;
     *head = b;
     grow_limit = h->grow_limit;
-    erts_smp_mtx_unlock(lock);
-    if (erts_smp_atomic_inc_read_nob(&h->nitems) > grow_limit) {
+    erts_mtx_unlock(lock);
+    if (erts_atomic_inc_read_nob(&h->nitems) > grow_limit) {
 	rehash(h, grow_limit);
     }
     return (void*) b;
@@ -240,36 +240,37 @@ void* safe_hash_erase(SafeHash* h, void* tmpl)
     SafeHashValue hval = h->fun.hash(tmpl);
     SafeHashBucket* b;
     SafeHashBucket** prevp;
-    erts_smp_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
-    erts_smp_mtx_lock(lock);
+    erts_mtx_t* lock = &h->lock_vec[hval % SAFE_HASH_LOCK_CNT].mtx;
+    erts_mtx_lock(lock);
     prevp = &h->tab[hval & h->size_mask];
     b = *prevp;
     while(b != NULL) {
 	if ((b->hvalue == hval) && (h->fun.cmp(tmpl, (void*)b) == 0)) {
 	    *prevp = b->next;
-	    erts_smp_mtx_unlock(lock);
-	    erts_smp_atomic_dec_nob(&h->nitems);
+	    erts_mtx_unlock(lock);
+	    erts_atomic_dec_nob(&h->nitems);
 	    h->fun.free((void*)b);
 	    return tmpl;
 	}
 	prevp = &b->next;
 	b = b->next;
     }
-    erts_smp_mtx_unlock(lock);
+    erts_mtx_unlock(lock);
     return NULL;
 }
 
 /*
-** Call 'func(obj,func_arg2)' for all objects in table. NOT SAFE!!!
+** Call 'func(obj,func_arg2,func_arg3)' for all objects in table. NOT SAFE!!!
 */
-void safe_hash_for_each(SafeHash* h, void (*func)(void *, void *), void *func_arg2)
+void safe_hash_for_each(SafeHash* h, void (*func)(void *, void *, void *),
+                        void *func_arg2, void *func_arg3)
 {
     int i;
 
     for (i = 0; i <= h->size_mask; i++) {
 	SafeHashBucket* b = h->tab[i];
 	while (b != NULL) {
-	    (*func)((void *) b, func_arg2);
+	    (*func)((void *) b, func_arg2, func_arg3);
 	    b = b->next;
 	}
     }
@@ -280,7 +281,7 @@ void erts_lcnt_enable_hash_lock_count(SafeHash *h, erts_lock_flags_t flags, int
     int i;
 
     for(i = 0; i < SAFE_HASH_LOCK_CNT; i++) {
-        erts_smp_mtx_t *lock = &h->lock_vec[i].mtx;
+        erts_mtx_t *lock = &h->lock_vec[i].mtx;
 
         if(enable) {
             erts_lcnt_install_new_lock_info(&lock->lcnt, "safe_hash", NIL,
diff --git a/erts/emulator/beam/safe_hash.h b/erts/emulator/beam/safe_hash.h
index dde48a6de8..af97b4cb4d 100644
--- a/erts/emulator/beam/safe_hash.h
+++ b/erts/emulator/beam/safe_hash.h
@@ -73,11 +73,11 @@ typedef struct
     int size_mask;	      /* (RW) Number of slots - 1 */
     SafeHashBucket** tab;     /* (RW) Vector of bucket pointers (objects) */
     int grow_limit;           /* (RW) Threshold for growing table */
-    erts_smp_atomic_t nitems;       /* (A) Number of items in table */
-    erts_smp_atomic_t is_rehashing; /* (A) Table rehashing in progress */
+    erts_atomic_t nitems;       /* (A) Number of items in table */
+    erts_atomic_t is_rehashing; /* (A) Table rehashing in progress */
 
     union {
-	erts_smp_mtx_t mtx;
+	erts_mtx_t mtx;
 	byte __cache_line__[64];
     }lock_vec[SAFE_HASH_LOCK_CNT];
 
@@ -95,7 +95,7 @@ void* safe_hash_get(SafeHash*, void*);
 void* safe_hash_put(SafeHash*, void*);
 void* safe_hash_erase(SafeHash*, void*);
 
-void  safe_hash_for_each(SafeHash*, void (*func)(void *, void *), void *);
+void  safe_hash_for_each(SafeHash*, void (*func)(void *, void *, void *), void *, void *);
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
 void erts_lcnt_enable_hash_lock_count(SafeHash*, erts_lock_flags_t, int);
diff --git a/erts/emulator/beam/select_instrs.tab b/erts/emulator/beam/select_instrs.tab
new file mode 100644
index 0000000000..2951949d38
--- /dev/null
+++ b/erts/emulator/beam/select_instrs.tab
@@ -0,0 +1,190 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+i_select_val_bins := select_val_bins.fetch.select;
+
+select_val_bins.head() {
+    Eterm select_val;
+}
+
+select_val_bins.fetch(Src) {
+    select_val = $Src;
+}
+
+select_val_bins.select(Fail, NumElements) {
+    struct Singleton {
+        BeamInstr val;
+    };
+    struct Singleton* low;
+    struct Singleton* high;
+    struct Singleton* mid;
+    int bdiff; /* int not long because the arrays aren't that large */
+
+    low = (struct Singleton *) ($NEXT_INSTRUCTION);
+    high = low + $NumElements;
+
+    /* The pointer subtraction (high-low) below must produce
+     * a signed result, because high could be < low. That
+     * requires the compiler to insert quite a bit of code.
+     *
+     * However, high will be > low so the result will be
+     * positive. We can use that knowledge to optimise the
+     * entire sequence, from the initial comparison to the
+     * computation of mid.
+     *
+     * -- Mikael Pettersson, Acumem AB
+     *
+     * Original loop control code:
+     *
+     * while (low < high) {
+     *    mid = low + (high-low) / 2;
+     *
+     */
+    while ((bdiff = (int)((char*)high - (char*)low)) > 0) {
+        unsigned int boffset = ((unsigned int)bdiff >> 1) & ~(sizeof(struct Singleton)-1);
+
+        mid = (struct Singleton*)((char*)low + boffset);
+        if (select_val < mid->val) {
+            high = mid;
+        } else if (select_val > mid->val) {
+            low = mid + 1;
+        } else {
+            Sint32* jump_tab = (Sint32 *) ($NEXT_INSTRUCTION + $NumElements);
+            Sint32 offset = jump_tab[mid - (struct Singleton *)($NEXT_INSTRUCTION)];
+            $JUMP(offset);
+        }
+    }
+    $JUMP($Fail);
+}
+
+i_select_tuple_arity2 := select_val2.src.get_arity.execute;
+i_select_val2 := select_val2.src.execute;
+
+select_val2.head() {
+    Eterm select_val2;
+}
+
+select_val2.src(Src) {
+    select_val2 = $Src;
+}
+
+select_val2.get_arity() {
+    if (ERTS_LIKELY(is_tuple(select_val2))) {
+        select_val2 = *tuple_val(select_val2);
+    } else {
+        select_val2 = NIL;
+    }
+}
+
+select_val2.execute(Fail, T1, T2) {
+    Sint32* jump_tab = (Sint32 *) ($NEXT_INSTRUCTION);
+
+    if (select_val2 == $T1) {
+        $JUMP(jump_tab[0]);
+    } else if (select_val2 == $T2) {
+        $JUMP(jump_tab[1]);
+    } else {
+        $FAIL($Fail);
+    }
+}
+
+i_select_tuple_arity := select_val_lin.fetch.get_arity.execute;
+i_select_val_lins := select_val_lin.fetch.execute;
+
+select_val_lin.head() {
+    Eterm select_val;
+}
+
+select_val_lin.fetch(Src) {
+    select_val = $Src;
+}
+
+select_val_lin.get_arity() {
+    if (ERTS_LIKELY(is_tuple(select_val))) {
+        select_val = *tuple_val(select_val);
+    } else {
+        select_val = NIL;
+    }
+}
+
+select_val_lin.execute(Fail, N) {
+    BeamInstr* vs = $NEXT_INSTRUCTION;
+    int ix = 0;
+
+    for (;;) {
+        if (vs[ix+0] >= select_val) {
+            ix += 0;
+            break;
+        }
+        if (vs[ix+1] >= select_val) {
+            ix += 1;
+            break;
+        }
+        ix += 2;
+    }
+
+    if (vs[ix] == select_val) {
+        Sint32* jump_tab = (Sint32 *) ($NEXT_INSTRUCTION + $N);
+        Eterm offset = jump_tab[ix];
+        $JUMP(offset);
+    } else {
+        $JUMP($Fail);
+    }
+}
+
+JUMP_ON_VAL(Fail, Index, N, Base) {
+    if (is_small($Index)) {
+        $Index = (Uint) (signed_val($Index) - $Base);
+        if ($Index < $N) {
+            Sint32* jump_tab = (Sint32 *) ($NEXT_INSTRUCTION);
+            $JUMP(jump_tab[$Index]);
+        }
+    }
+    $FAIL($Fail);
+}
+
+i_jump_on_val_zero := jump_on_val_zero.fetch.execute;
+
+jump_on_val_zero.head() {
+    Eterm index;
+}
+
+jump_on_val_zero.fetch(Src) {
+    index = $Src;
+}
+
+jump_on_val_zero.execute(Fail, N) {
+    $JUMP_ON_VAL($Fail, index, $N, 0);
+}
+
+i_jump_on_val := jump_on_val.fetch.execute;
+
+jump_on_val.head() {
+    Eterm index;
+}
+
+jump_on_val.fetch(Src) {
+    index = $Src;
+}
+
+jump_on_val.execute(Fail, N, Base) {
+    $JUMP_ON_VAL($Fail, index, $N, $Base);
+}
diff --git a/erts/emulator/beam/sys.h b/erts/emulator/beam/sys.h
index 704d567337..be6ab57eeb 100644
--- a/erts/emulator/beam/sys.h
+++ b/erts/emulator/beam/sys.h
@@ -21,7 +21,7 @@
 #ifndef __SYS_H__
 #define __SYS_H__
 
-#if !defined(__GNUC__)
+#if !defined(__GNUC__) || defined(__e2k__)
 #  define ERTS_AT_LEAST_GCC_VSN__(MAJ, MIN, PL) 0
 #elif !defined(__GNUC_MINOR__)
 #  define ERTS_AT_LEAST_GCC_VSN__(MAJ, MIN, PL) \
@@ -34,9 +34,6 @@
   (((__GNUC__ << 24) | (__GNUC_MINOR__ << 12) | __GNUC_PATCHLEVEL__) >= (((MAJ) << 24) | ((MIN) << 12) | (PL)))
 #endif
 
-#if defined(ERTS_DIRTY_SCHEDULERS) && !defined(ERTS_SMP)
-# error "Dirty schedulers not supported without smp support"
-#endif
 
 #ifdef ERTS_INLINE
 #  ifndef ERTS_CAN_INLINE
@@ -221,12 +218,6 @@ __decl_noreturn void __noreturn erl_assert_error(const char* expr, const char *f
 #  define ASSERT(e) ((void) 1)
 #endif
 
-#ifdef ERTS_SMP
-#  define ERTS_SMP_ASSERT(e) ASSERT(e)
-#else
-#  define ERTS_SMP_ASSERT(e) ((void)1)
-#endif
-
 /* ERTS_UNDEF can be used to silence false warnings about
  * "variable may be used uninitialized" while keeping the variable
  * marked as undefined by valgrind.
@@ -327,9 +318,9 @@ __decl_noreturn void __noreturn erl_assert_error(const char* expr, const char *f
 #endif
 
 #if SIZEOF_VOID_P == SIZEOF_LONG
-typedef unsigned long Eterm;
-typedef unsigned long Uint;
-typedef long          Sint;
+typedef unsigned long Eterm erts_align_attribute(sizeof(long));
+typedef unsigned long Uint  erts_align_attribute(sizeof(long));
+typedef long          Sint  erts_align_attribute(sizeof(long));
 #define SWORD_CONSTANT(Const) Const##L
 #define UWORD_CONSTANT(Const) Const##UL
 #define ERTS_UWORD_MAX ULONG_MAX
@@ -337,9 +328,9 @@ typedef long          Sint;
 #define ERTS_SIZEOF_ETERM SIZEOF_LONG
 #define ErtsStrToSint strtol
 #elif SIZEOF_VOID_P == SIZEOF_INT
-typedef unsigned int Eterm;
-typedef unsigned int Uint;
-typedef int          Sint;
+typedef unsigned int Eterm erts_align_attribute(sizeof(int));
+typedef unsigned int Uint  erts_align_attribute(sizeof(int));
+typedef int          Sint  erts_align_attribute(sizeof(int));
 #define SWORD_CONSTANT(Const) Const
 #define UWORD_CONSTANT(Const) Const##U
 #define ERTS_UWORD_MAX UINT_MAX
@@ -347,9 +338,9 @@ typedef int          Sint;
 #define ERTS_SIZEOF_ETERM SIZEOF_INT
 #define ErtsStrToSint strtol
 #elif SIZEOF_VOID_P == SIZEOF_LONG_LONG
-typedef unsigned long long Eterm;
-typedef unsigned long long Uint;
-typedef long long          Sint;
+typedef unsigned long long Eterm erts_align_attribute(sizeof(long long));
+typedef unsigned long long Uint  erts_align_attribute(sizeof(long long));
+typedef long long          Sint  erts_align_attribute(sizeof(long long));
 #define SWORD_CONSTANT(Const) Const##LL
 #define UWORD_CONSTANT(Const) Const##ULL
 #define ERTS_UWORD_MAX ULLONG_MAX
@@ -375,29 +366,11 @@ typedef UWord BeamInstr;
 #    define HAVE_INT64 1
 typedef unsigned long Uint64;
 typedef long          Sint64;
-#    ifdef ULONG_MAX
-#      define ERTS_UINT64_MAX ULONG_MAX
-#    endif
-#    ifdef LONG_MAX
-#      define ERTS_SINT64_MAX LONG_MAX
-#    endif
-#    ifdef LONG_MIN
-#      define ERTS_SINT64_MIN LONG_MIN
-#    endif
 #    define ErtsStrToSint64 strtol
 #  elif SIZEOF_LONG_LONG == 8
 #    define HAVE_INT64 1
 typedef unsigned long long Uint64;
 typedef long long          Sint64;
-#    ifdef ULLONG_MAX
-#      define ERTS_UINT64_MAX ULLONG_MAX
-#    endif
-#    ifdef LLONG_MAX
-#      define ERTS_SINT64_MAX LLONG_MAX
-#    endif
-#    ifdef LLONG_MIN
-#      define ERTS_SINT64_MIN LLONG_MIN
-#    endif
 #    define ErtsStrToSint64 strtoll
 #  else
 #    error "No 64-bit integer type found"
@@ -411,7 +384,7 @@ typedef long long          Sint64;
 #  define ERTS_SINT64_MAX ((Sint64) ((((Uint64) 1) << 63)-1))
 #endif
 #ifndef ERTS_SINT64_MIN
-#  define ERTS_SINT64_MIN (-1*(((Sint64) 1) << 63))
+#  define ERTS_SINT64_MIN ((Sint64) ((((Uint64) 1) << 63)))
 #endif
 
 #if SIZEOF_LONG == 4
@@ -424,6 +397,16 @@ typedef int          Sint32;
 #error Found no appropriate type to use for 'Uint32' and 'Sint32'
 #endif
 
+#ifndef ERTS_UINT32_MAX
+#  define ERTS_UINT32_MAX (~((Uint32) 0))
+#endif
+#ifndef ERTS_SINT32_MAX
+#  define ERTS_SINT32_MAX ((Sint32) ((((Uint32) 1) << 31)-1))
+#endif
+#ifndef ERTS_SINT32_MIN
+#  define ERTS_SINT32_MIN ((Sint32) ((((Uint32) 1) << 31)))
+#endif
+
 #if SIZEOF_INT == 2
 typedef unsigned int Uint16;
 typedef int          Sint16;
@@ -434,6 +417,16 @@ typedef short          Sint16;
 #error Found no appropriate type to use for 'Uint16' and 'Sint16'
 #endif
 
+#ifndef ERTS_UINT16_MAX
+#  define ERTS_UINT16_MAX (~((Uint16) 0))
+#endif
+#ifndef ERTS_SINT16_MAX
+#  define ERTS_SINT16_MAX ((Sint16) ((((Uint16) 1) << 15)-1))
+#endif
+#ifndef ERTS_SINT16_MIN
+#  define ERTS_SINT16_MIN ((Sint16) ((((Uint16) 1) << 15)))
+#endif
+
 #if CHAR_BIT == 8
 typedef unsigned char byte;
 #else
@@ -470,41 +463,25 @@ typedef union {
 
 #include "erl_lock_check.h"
 
-/* needed by erl_smp.h */
+/* needed by erl_threads.h */
 int erts_send_warning_to_logger_str_nogl(char *);
 
-#include "erl_smp.h"
+#include "erl_threads.h"
 
 #ifdef ERTS_WANT_BREAK_HANDLING
-#  ifdef ERTS_SMP
-extern erts_smp_atomic32_t erts_break_requested;
+extern erts_atomic32_t erts_break_requested;
 #    define ERTS_BREAK_REQUESTED \
-  ((int) erts_smp_atomic32_read_nob(&erts_break_requested))
-#  else
-extern volatile int erts_break_requested;
-#    define ERTS_BREAK_REQUESTED erts_break_requested
-#  endif
+  ((int) erts_atomic32_read_nob(&erts_break_requested))
 void erts_do_break_handling(void);
 #endif
 
-#if !defined(ERTS_SMP) && !defined(__WIN32__)
-extern volatile Uint erts_signal_state;
-#define ERTS_SIGNAL_STATE erts_signal_state
-void erts_handle_signal_state(void);
-#endif
 
-#ifdef ERTS_SMP
-extern erts_smp_atomic32_t erts_writing_erl_crash_dump;
+extern erts_atomic32_t erts_writing_erl_crash_dump;
 extern erts_tsd_key_t erts_is_crash_dumping_key;
 #define ERTS_SOMEONE_IS_CRASH_DUMPING \
-  ((int) erts_smp_atomic32_read_mb(&erts_writing_erl_crash_dump))
+  ((int) erts_atomic32_read_mb(&erts_writing_erl_crash_dump))
 #define ERTS_IS_CRASH_DUMPING \
   ((int) (SWord) erts_tsd_get(erts_is_crash_dumping_key))
-#else
-extern volatile int erts_writing_erl_crash_dump;
-#define ERTS_SOMEONE_IS_CRASH_DUMPING erts_writing_erl_crash_dump
-#define ERTS_IS_CRASH_DUMPING erts_writing_erl_crash_dump
-#endif
 
 /* Deal with memcpy() vs bcopy() etc. We want to use the mem*() functions,
    but be able to fall back on bcopy() etc on systems that don't have
@@ -600,8 +577,6 @@ __decl_noreturn void __noreturn erts_exit(int n, char*, ...);
     erts_exit(ERTS_ABORT_EXIT, "%s:%d:%s(): Internal error: %s\n", \
 	     __FILE__, __LINE__, __func__, What)
 
-Eterm erts_check_io_info(void *p);
-
 UWord erts_sys_get_page_size(void);
 
 /* Size of misc memory allocated from system dependent code */
@@ -643,7 +618,7 @@ int erts_send_info_to_logger_nogl(erts_dsprintf_buf_t *);
 int erts_send_warning_to_logger_nogl(erts_dsprintf_buf_t *);
 int erts_send_error_to_logger_nogl(erts_dsprintf_buf_t *);
 int erts_send_info_to_logger_str_nogl(char *);
-/* needed by erl_smp.h (declared above)
+/* needed by erl_threads.h (declared above)
    int erts_send_warning_to_logger_str_nogl(char *); */
 int erts_send_error_to_logger_str_nogl(char *);
 
@@ -663,6 +638,8 @@ typedef struct preload {
  */
 typedef Eterm ErtsTracer;
 
+#include "erl_osenv.h"
+
 /*
  * This structure contains options to all built in drivers.
  * None of the drivers use all of the fields.
@@ -678,8 +655,7 @@ typedef struct _SysDriverOpts {
     int hide_window;		/* Hide this windows (Windows). */
     int exit_status;		/* Report exit status of subprocess. */
     int overlapped_io;          /* Only has effect on windows NT et al */
-    char *envir;		/* Environment of the port process, */
-				/* in Windows format. */
+    erts_osenv_t envir;		/* Environment of the port process */
     char **argv;                /* Argument vector in Unix'ish format. */
     char *wd;			/* Working directory. */
     unsigned spawn_type;        /* Bitfield of ERTS_SPAWN_DRIVER | 
@@ -764,11 +740,7 @@ extern char *erts_sys_ddll_error(int code);
 /*
  * System interfaces for startup.
  */
-void erts_sys_schedule_interrupt(int set);
-#ifdef ERTS_SMP
-void erts_sys_schedule_interrupt_timed(int, ErtsMonotonicTime);
 void erts_sys_main_thread(void);
-#endif
 
 extern int erts_sys_prepare_crash_dump(int secs);
 extern void erts_sys_pre_init(void);
@@ -813,15 +785,12 @@ void set_break_quit(void (*)(void), void (*)(void));
 
 void os_flavor(char*, unsigned);
 void os_version(int*, int*, int*);
-void init_getenv_state(GETENV_STATE *);
-char * getenv_string(GETENV_STATE *);
-void fini_getenv_state(GETENV_STATE *);
 
 #define HAVE_ERTS_CHECK_IO_DEBUG
 typedef struct {
     int no_used_fds;
     int no_driver_select_structs;
-    int no_driver_event_structs;
+    int no_enif_select_structs;
 } ErtsCheckIoDebugInfo;
 int erts_check_io_debug(ErtsCheckIoDebugInfo *ip);
 
@@ -836,32 +805,36 @@ int sys_double_to_chars_ext(double, char*, size_t, size_t);
 int sys_double_to_chars_fast(double, char*, int, int, int);
 void sys_get_pid(char *, size_t);
 
-/* erts_sys_putenv() returns, 0 on success and a value != 0 on failure. */
-int erts_sys_putenv(char *key, char *value);
-/* Simple variant used from drivers, raw eightbit interface */
-int erts_sys_putenv_raw(char *key, char *value);
-/* erts_sys_getenv() returns 0 on success (length of value string in
-   *size), a value > 0 if value buffer is too small (*size is set to needed
-   size), and a value < 0 on failure. */
-int erts_sys_getenv(char *key, char *value, size_t *size);
-/* Simple variant used from drivers, raw eightbit interface */
-int erts_sys_getenv_raw(char *key, char *value, size_t *size);
-/* erts_sys_getenv__() is only allowed to be used in early init phase */
-int erts_sys_getenv__(char *key, char *value, size_t *size);
-/* erst_sys_unsetenv() returns 0 on success and a value != 0 on failure. */
-int erts_sys_unsetenv(char *key);
+/* erl_drv_get/putenv have been implicitly 8-bit for so long that we can't
+ * change them without breaking things on Windows. Their return values are
+ * identical to erts_osenv_get/putenv */
+int erts_sys_explicit_8bit_getenv(char *key, char *value, size_t *size);
+int erts_sys_explicit_8bit_putenv(char *key, char *value);
+
+/* This is identical to erts_sys_explicit_8bit_getenv but falls down to the
+ * host OS implementation instead of erts_osenv. */
+int erts_sys_explicit_host_getenv(char *key, char *value, size_t *size);
+
+const erts_osenv_t *erts_sys_rlock_global_osenv(void);
+void erts_sys_runlock_global_osenv(void);
+
+erts_osenv_t *erts_sys_rwlock_global_osenv(void);
+void erts_sys_rwunlock_global_osenv(void);
 
 /* Easier to use, but not as efficient, environment functions */
 char *erts_read_env(char *key);
 void erts_free_read_env(void *value);
 
-#if defined(ERTS_SMP)
-#if defined(ERTS_THR_HAVE_SIG_FUNCS) && !defined(ETHR_UNUSABLE_SIGUSRX)
+#if defined(ERTS_THR_HAVE_SIG_FUNCS) &&                         \
+    (!defined(ETHR_UNUSABLE_SIGUSRX) || defined(SIGRTMIN))
 extern void sys_thr_resume(erts_tid_t tid);
 extern void sys_thr_suspend(erts_tid_t tid);
-#define ERTS_SYS_SUSPEND_SIGNAL SIGUSR2
-#endif
-#endif
+#ifdef SIGRTMIN
+#define ERTS_SYS_SUSPEND_SIGNAL (SIGRTMIN+1)
+#else
+#define ERTS_SYS_SUSPEND_SIGNAL (SIGUSR2)
+#endif /* SIGRTMIN */
+#endif /* HAVE_SIG_FUNCS */
 
 /* utils.c */
 
@@ -1023,157 +996,83 @@ erts_refc_read(erts_refc_t *refcp, erts_aint_t min_val)
     return val;
 }
 
-#endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
+#endif  /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
 
-typedef erts_smp_atomic_t erts_smp_refc_t;
-
-ERTS_GLB_INLINE void erts_smp_refc_init(erts_smp_refc_t *refcp, erts_aint_t val);
-ERTS_GLB_INLINE void erts_smp_refc_inc(erts_smp_refc_t *refcp, erts_aint_t min_val);
-ERTS_GLB_INLINE erts_aint_t erts_smp_refc_inc_unless(erts_smp_refc_t *refcp,
-                                                     erts_aint_t unless_val,
-                                                     erts_aint_t min_val);
-ERTS_GLB_INLINE erts_aint_t erts_smp_refc_inctest(erts_smp_refc_t *refcp,
-                                                  erts_aint_t min_val);
-ERTS_GLB_INLINE void erts_smp_refc_dec(erts_smp_refc_t *refcp, erts_aint_t min_val);
-ERTS_GLB_INLINE erts_aint_t erts_smp_refc_dectest(erts_smp_refc_t *refcp,
-                                                  erts_aint_t min_val);
-ERTS_GLB_INLINE void erts_smp_refc_add(erts_smp_refc_t *refcp, erts_aint_t diff,
-                                       erts_aint_t min_val);
-ERTS_GLB_INLINE erts_aint_t erts_smp_refc_read(erts_smp_refc_t *refcp,
-                                               erts_aint_t min_val);
+/* Thin wrappers around memcpy and friends, which should always be used in
+ * place of plain memcpy, memset, etc.
+ *
+ * Passing NULL to any of these functions is undefined behavior even though it
+ * may seemingly work when the length (if any) is zero; a compiler can take
+ * this as a hint that the passed operand may *never* be NULL and then optimize
+ * based on that information.
+ */
+ERTS_GLB_INLINE void *sys_memcpy(void *dest, const void *src, size_t n);
+ERTS_GLB_INLINE void *sys_memmove(void *dest, const void *src, size_t n);
+ERTS_GLB_INLINE int sys_memcmp(const void *s1, const void *s2, size_t n);
+ERTS_GLB_INLINE void *sys_memset(void *s, int c, size_t n);
+ERTS_GLB_INLINE void *sys_memzero(void *s, size_t n);
+ERTS_GLB_INLINE int sys_strcmp(const char *s1, const char *s2);
+ERTS_GLB_INLINE int sys_strncmp(const char *s1, const char *s2, size_t n);
+ERTS_GLB_INLINE char *sys_strcpy(char *dest, const char *src);
+ERTS_GLB_INLINE char *sys_strncpy(char *dest, const char *src, size_t n);
+ERTS_GLB_INLINE size_t sys_strlen(const char *s);
 
 #if ERTS_GLB_INLINE_INCL_FUNC_DEF
 
-ERTS_GLB_INLINE void
-erts_smp_refc_init(erts_smp_refc_t *refcp, erts_aint_t val)
+ERTS_GLB_INLINE void *sys_memcpy(void *dest, const void *src, size_t n)
 {
-    erts_smp_atomic_init_nob((erts_smp_atomic_t *) refcp, val);
+    ASSERT(dest != NULL && src != NULL);
+    return memcpy(dest,src,n);
 }
-
-ERTS_GLB_INLINE void
-erts_smp_refc_inc(erts_smp_refc_t *refcp, erts_aint_t min_val)
+ERTS_GLB_INLINE void *sys_memmove(void *dest, const void *src, size_t n)
 {
-#ifdef ERTS_REFC_DEBUG
-    erts_aint_t val = erts_smp_atomic_inc_read_nob((erts_smp_atomic_t *) refcp);
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_inc(): Bad refc found (refc=%ld < %ld)!\n",
-		 val, min_val);
-#else
-    erts_smp_atomic_inc_nob((erts_smp_atomic_t *) refcp);
-#endif
+    ASSERT(dest != NULL && src != NULL);
+    return memmove(dest,src,n);
 }
-
-ERTS_GLB_INLINE erts_aint_t
-erts_smp_refc_inc_unless(erts_smp_refc_t *refcp,
-                         erts_aint_t unless_val,
-                         erts_aint_t min_val)
+ERTS_GLB_INLINE int sys_memcmp(const void *s1, const void *s2, size_t n)
 {
-    erts_aint_t val = erts_smp_atomic_read_nob((erts_smp_atomic_t *) refcp);
-    while (1) {
-        erts_aint_t exp, new;
-#ifdef ERTS_REFC_DEBUG
-        if (val < 0)
-            erts_exit(ERTS_ABORT_EXIT,
-                      "erts_smp_refc_inc_unless(): Bad refc found (refc=%ld < %ld)!\n",
-                      val, min_val);
-#endif
-        if (val == unless_val)
-            return val;
-        new = val + 1;
-        exp = val;
-        val = erts_smp_atomic_cmpxchg_nob((erts_smp_atomic_t *) refcp, new, exp);
-        if (val == exp)
-            return new;
-    }
+    ASSERT(s1 != NULL && s2 != NULL);
+    return memcmp(s1,s2,n);
 }
-
-
-ERTS_GLB_INLINE erts_aint_t
-erts_smp_refc_inctest(erts_smp_refc_t *refcp, erts_aint_t min_val)
+ERTS_GLB_INLINE void *sys_memset(void *s, int c, size_t n)
 {
-    erts_aint_t val = erts_smp_atomic_inc_read_nob((erts_smp_atomic_t *) refcp);
-#ifdef ERTS_REFC_DEBUG
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_inctest(): Bad refc found (refc=%ld < %ld)!\n",
-		 val, min_val);
-#endif
-    return val;
+    ASSERT(s != NULL);
+    return memset(s,c,n);
 }
-
-ERTS_GLB_INLINE void
-erts_smp_refc_dec(erts_smp_refc_t *refcp, erts_aint_t min_val)
+ERTS_GLB_INLINE void *sys_memzero(void *s, size_t n)
 {
-#ifdef ERTS_REFC_DEBUG
-    erts_aint_t val = erts_smp_atomic_dec_read_nob((erts_smp_atomic_t *) refcp);
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_dec(): Bad refc found (refc=%ld < %ld)!\n",
-		 val, min_val);
-#else
-    erts_smp_atomic_dec_nob((erts_smp_atomic_t *) refcp);
-#endif
+    ASSERT(s != NULL);
+    return memset(s,'\0',n);
 }
-
-ERTS_GLB_INLINE erts_aint_t
-erts_smp_refc_dectest(erts_smp_refc_t *refcp, erts_aint_t min_val)
+ERTS_GLB_INLINE int sys_strcmp(const char *s1, const char *s2)
 {
-    erts_aint_t val = erts_smp_atomic_dec_read_nob((erts_smp_atomic_t *) refcp);
-#ifdef ERTS_REFC_DEBUG
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_dectest(): Bad refc found (refc=%ld < %ld)!\n",
-		 val, min_val);
-#endif
-    return val;
+    ASSERT(s1 != NULL && s2 != NULL);
+    return strcmp(s1,s2);
 }
-
-ERTS_GLB_INLINE void
-erts_smp_refc_add(erts_smp_refc_t *refcp, erts_aint_t diff, erts_aint_t min_val)
+ERTS_GLB_INLINE int sys_strncmp(const char *s1, const char *s2, size_t n)
 {
-#ifdef ERTS_REFC_DEBUG
-    erts_aint_t val = erts_smp_atomic_add_read_nob((erts_smp_atomic_t *) refcp, diff);
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_add(%ld): Bad refc found (refc=%ld < %ld)!\n",
-		 diff, val, min_val);
-#else
-    erts_smp_atomic_add_nob((erts_smp_atomic_t *) refcp, diff);
-#endif
+    ASSERT(s1 != NULL && s2 != NULL);
+    return strncmp(s1,s2,n);
 }
+ERTS_GLB_INLINE char *sys_strcpy(char *dest, const char *src)
+{
+    ASSERT(dest != NULL && src != NULL);
+    return strcpy(dest,src);
 
-ERTS_GLB_INLINE erts_aint_t
-erts_smp_refc_read(erts_smp_refc_t *refcp, erts_aint_t min_val)
+}
+ERTS_GLB_INLINE char *sys_strncpy(char *dest, const char *src, size_t n)
 {
-    erts_aint_t val = erts_smp_atomic_read_nob((erts_smp_atomic_t *) refcp);
-#ifdef ERTS_REFC_DEBUG
-    if (val < min_val)
-	erts_exit(ERTS_ABORT_EXIT,
-		 "erts_smp_refc_read(): Bad refc found (refc=%ld < %ld)!\n",
-		 val, min_val);
-#endif
-    return val;
+    ASSERT(dest != NULL && src != NULL);
+    return strncpy(dest,src,n);
+}
+ERTS_GLB_INLINE size_t sys_strlen(const char *s)
+{
+    ASSERT(s != NULL);
+    return strlen(s);
 }
-
 #endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
 
-
-#ifdef ERTS_ENABLE_KERNEL_POLL
-extern int erts_use_kernel_poll;
-#endif
-
-#define sys_memcpy(s1,s2,n)  memcpy(s1,s2,n)
-#define sys_memmove(s1,s2,n) memmove(s1,s2,n)
-#define sys_memcmp(s1,s2,n)  memcmp(s1,s2,n)
-#define sys_memset(s,c,n)    memset(s,c,n)
-#define sys_memzero(s, n)    memset(s,'\0',n)
-#define sys_strcmp(s1,s2)    strcmp(s1,s2)
-#define sys_strncmp(s1,s2,n) strncmp(s1,s2,n)
-#define sys_strcpy(s1,s2)    strcpy(s1,s2)
-#define sys_strncpy(s1,s2,n) strncpy(s1,s2,n)
-#define sys_strlen(s)        strlen(s)
-
 /* define function symbols (needed in sys_drv_api) */
 #define sys_fp_alloc     sys_alloc
 #define sys_fp_realloc   sys_realloc
@@ -1341,4 +1240,52 @@ int erts_get_printable_characters(void);
 
 void erts_init_sys_common_misc(void);
 
+ERTS_GLB_INLINE Sint erts_raw_env_7bit_ascii_char_need(int encoding);
+ERTS_GLB_INLINE byte *erts_raw_env_7bit_ascii_char_put(byte c, byte *p,
+                                                       int encoding);
+ERTS_GLB_INLINE int  erts_raw_env_char_is_7bit_ascii_char(byte c, byte *p,
+                                                          int encoding);
+ERTS_GLB_INLINE byte *erts_raw_env_next_char(byte *p, int encoding);
+
+#if ERTS_GLB_INLINE_INCL_FUNC_DEF
+
+ERTS_GLB_INLINE Sint
+erts_raw_env_7bit_ascii_char_need(int encoding)
+{
+    return (encoding == ERL_FILENAME_WIN_WCHAR) ? 2 : 1;
+}
+
+ERTS_GLB_INLINE byte *
+erts_raw_env_7bit_ascii_char_put(byte c,
+                                 byte *p,
+                                 int encoding)
+{
+    *(p++) = c;
+    if (encoding == ERL_FILENAME_WIN_WCHAR)
+        *(p++) = 0;
+    return p;
+}
+
+ERTS_GLB_INLINE int
+erts_raw_env_char_is_7bit_ascii_char(byte c,
+                                     byte *p,
+                                     int encoding)
+{
+    if (encoding == ERL_FILENAME_WIN_WCHAR)
+        return (p[0] == c) & (p[1] == 0);
+    else
+        return p[0] == c;
+}
+
+ERTS_GLB_INLINE byte *
+erts_raw_env_next_char(byte *p, int encoding)
+{
+    if (encoding == ERL_FILENAME_WIN_WCHAR)
+        return p + 2;
+    else
+        return p + 1;
+}
+
+#endif /* #if ERTS_GLB_INLINE_INCL_FUNC_DEF */
+
 #endif
diff --git a/erts/emulator/beam/trace_instrs.tab b/erts/emulator/beam/trace_instrs.tab
new file mode 100644
index 0000000000..3eee81c053
--- /dev/null
+++ b/erts/emulator/beam/trace_instrs.tab
@@ -0,0 +1,168 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+return_trace() {
+    ErtsCodeMFA* mfa = (ErtsCodeMFA *)(E[0]);
+
+    SWAPOUT;		/* Needed for shared heap */
+    ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+    erts_trace_return(c_p, mfa, r(0), ERTS_TRACER_FROM_ETERM(E+1)/* tracer */);
+    ERTS_REQ_PROC_MAIN_LOCK(c_p);
+    SWAPIN;
+    c_p->cp = NULL;
+    SET_I((BeamInstr *) cp_val(E[2]));
+    E += 3;
+    Goto(*I);
+    //| -no_next
+}
+
+i_generic_breakpoint() {
+    BeamInstr real_I;
+    HEAVY_SWAPOUT;
+    real_I = erts_generic_breakpoint(c_p, erts_code_to_codeinfo(I), reg);
+    HEAVY_SWAPIN;
+    ASSERT(VALID_INSTR(real_I));
+    Goto(real_I);
+    //| -no_next
+}
+
+i_return_time_trace() {
+    BeamInstr *pc = (BeamInstr *) (UWord) E[0];
+    SWAPOUT;
+    erts_trace_time_return(c_p, erts_code_to_codeinfo(pc));
+    SWAPIN;
+    c_p->cp = NULL;
+    SET_I((BeamInstr *) cp_val(E[1]));
+    E += 2;
+    Goto(*I);
+    //| -no_next
+}
+
+i_return_to_trace() {
+    if (IS_TRACED_FL(c_p, F_TRACE_RETURN_TO)) {
+        Uint *cpp = (Uint*) E;
+        for(;;) {
+            ASSERT(is_CP(*cpp));
+            if (IsOpCode(*cp_val(*cpp), return_trace)) {
+                do
+                    ++cpp;
+                while (is_not_CP(*cpp));
+                cpp += 2;
+            } else if (IsOpCode(*cp_val(*cpp), i_return_to_trace)) {
+                do
+                    ++cpp;
+                while (is_not_CP(*cpp));
+            } else {
+                break;
+            }
+        }
+        SWAPOUT;		/* Needed for shared heap */
+        ERTS_UNREQ_PROC_MAIN_LOCK(c_p);
+        erts_trace_return_to(c_p, cp_val(*cpp));
+        ERTS_REQ_PROC_MAIN_LOCK(c_p);
+        SWAPIN;
+    }
+    c_p->cp = NULL;
+    SET_I((BeamInstr *) cp_val(E[0]));
+    E += 1;
+    Goto(*I);
+    //| -no_next
+}
+
+i_yield() {
+    /* This is safe as long as REDS_IN(c_p) is never stored
+     * in c_p->arg_reg[0]. It is currently stored in c_p->def_arg_reg[5].
+     */
+    c_p->arg_reg[0] = am_true;
+    c_p->arity = 1; /* One living register (the 'true' return value) */
+    SWAPOUT;
+    $SET_CP_I_ABS($NEXT_INSTRUCTION);
+    c_p->current = NULL;
+    goto do_schedule;
+    //| -no_next
+}
+
+i_hibernate() {
+    HEAVY_SWAPOUT;
+    if (erts_hibernate(c_p, reg)) {
+        FCALLS = c_p->fcalls;
+        c_p->flags &= ~F_HIBERNATE_SCHED;
+        goto do_schedule;
+    } else {
+        HEAVY_SWAPIN;
+        I = handle_error(c_p, I, reg, &bif_export[BIF_hibernate_3]->info.mfa);
+        goto post_error_handling;
+    }
+    //| -no_next
+}
+
+// This is optimised as an instruction because
+// it has to be very very fast.
+
+i_perf_counter() {
+    ErtsSysPerfCounter ts;
+
+    ts = erts_sys_perf_counter();
+    if (IS_SSMALL(ts)) {
+        r(0) = make_small((Sint)ts);
+    } else {
+        $GC_TEST(0, ERTS_SINT64_HEAP_SIZE(ts), 0);
+        r(0) = make_big(HTOP);
+#if defined(ARCH_32)
+        if (ts >= (((Uint64) 1) << 32)) {
+            *HTOP = make_pos_bignum_header(2);
+            BIG_DIGIT(HTOP, 0) = (Uint) (ts & ((Uint) 0xffffffff));
+            BIG_DIGIT(HTOP, 1) = (Uint) ((ts >> 32) & ((Uint) 0xffffffff));
+            HTOP += 3;
+        }
+        else
+#endif
+            {
+                *HTOP = make_pos_bignum_header(1);
+                BIG_DIGIT(HTOP, 0) = (Uint) ts;
+                HTOP += 2;
+            }
+    }
+}
+
+i_debug_breakpoint() {
+    HEAVY_SWAPOUT;
+    I = call_error_handler(c_p, erts_code_to_codemfa(I), reg, am_breakpoint);
+    HEAVY_SWAPIN;
+    if (I) {
+        Goto(*I);
+    }
+    goto handle_error;
+    //| -no_next
+}
+
+
+
+//
+// Special jump instruction used for tracing. Takes an absolute
+// failure address.
+//
+
+trace_jump(Fail) {
+    //| -no_next
+    SET_I((BeamInstr *) $Fail);
+    Goto(*I);
+}
diff --git a/erts/emulator/beam/utils.c b/erts/emulator/beam/utils.c
index d7116bd2c3..87709edc15 100644
--- a/erts/emulator/beam/utils.c
+++ b/erts/emulator/beam/utils.c
@@ -43,7 +43,6 @@
 #include "erl_printf.h"
 #include "erl_threads.h"
 #include "erl_lock_count.h"
-#include "erl_smp.h"
 #include "erl_time.h"
 #include "erl_thr_progress.h"
 #include "erl_thr_queue.h"
@@ -60,6 +59,7 @@
 #endif
 #define ERTS_WANT_NFUNC_SCHED_INTERNALS__
 #include "erl_nfunc_sched.h"
+#include "erl_proc_sig_queue.h"
 
 #undef M_TRIM_THRESHOLD
 #undef M_TOP_PAD
@@ -140,7 +140,7 @@ Eterm*
 erts_set_hole_marker(Eterm* ptr, Uint sz)
 {
     Eterm* p = ptr;
-    int i;
+    Uint i;
 
     for (i = 0; i < sz; i++) {
 	*p++ = ERTS_HOLE_MARKER;
@@ -1924,184 +1924,164 @@ make_internal_hash(Eterm term, Uint32 salt)
 #undef HCONST
 #undef MIX
 
+/* error_logger !
+   {log, Level, format, [args], #{ gl, pid, time, error_logger => #{tag, emulator => true} }}
+*/
 static Eterm
-do_allocate_logger_message(Eterm gleader, Eterm **hp, ErlOffHeap **ohp,
-			   ErlHeapFragment **bp, Process **p, Uint sz)
+do_allocate_logger_message(Eterm gleader, ErtsMonotonicTime *ts, Eterm *pid,
+                           Eterm **hp, ErlOffHeap **ohp,
+			   ErlHeapFragment **bp, Uint sz)
 {
     Uint gl_sz;
     gl_sz = IS_CONST(gleader) ? 0 : size_object(gleader);
-    sz = sz + gl_sz;
+    sz = sz + gl_sz + 6 /*outer 5-tuple*/
+        + MAP2_SZ /* error_logger map */;
+
+    *pid = erts_get_current_pid();
 
-#ifndef ERTS_SMP
-#ifdef USE_THREADS
-    if (!erts_get_scheduler_data()) /* Must be scheduler thread */
-	*p = NULL;
+    if (is_nil(gleader) && is_non_value(*pid)) {
+        sz += MAP2_SZ /* metadata map no gl, no pid */;
+    } else if (is_nil(gleader) || is_non_value(*pid))
+        sz += MAP3_SZ /* metadata map no gl or no pid*/;
     else
-#endif
-    {
-	*p = erts_whereis_process(NULL, 0, am_error_logger, 0, 0);
-	if (*p) {
-	    erts_aint32_t state = erts_smp_atomic32_read_acqb(&(*p)->state);
-	    if (state & (ERTS_PSFLG_RUNNING|ERTS_PSFLG_RUNNING_SYS))
-		*p = NULL;
-	}
-    }
+        sz += MAP4_SZ /* metadata map w gl w pid*/;
 
-    if (!*p) {
-	return NIL;
-    }
+    *ts = ERTS_MONOTONIC_TO_USEC(erts_get_monotonic_time(NULL)) + ERTS_MONOTONIC_OFFSET_USEC;
+    erts_bld_sint64(NULL, &sz, *ts);
 
-    /* So we have an error logger, lets build the message */
-#endif
     *bp = new_message_buffer(sz);
     *ohp = &(*bp)->off_heap;
     *hp = (*bp)->mem;
 
-    return (is_nil(gleader)
-	  ? am_noproc
-	  : (IS_CONST(gleader)
-	     ? gleader
-	     : copy_struct(gleader,gl_sz,hp,*ohp)));
+    return copy_struct(gleader,gl_sz,hp,*ohp);
 }
 
-static void do_send_logger_message(Eterm *hp, ErlOffHeap *ohp, ErlHeapFragment *bp,
-				   Process *p, Eterm message)
+static void do_send_logger_message(Eterm gl, Eterm tag, Eterm format, Eterm args,
+                                   ErtsMonotonicTime ts, Eterm pid,
+                                   Eterm *hp, ErlHeapFragment *bp)
 {
-#ifdef HARDDEBUG
-    erts_fprintf(stderr, "%T\n", message);
-#endif
-#ifdef ERTS_SMP
-    {
-	Eterm from = erts_get_current_pid();
-	if (is_not_internal_pid(from))
-	    from = NIL;
-	erts_queue_error_logger_message(from, message, bp);
+    Eterm message, md, el_tag = tag;
+    Eterm time = erts_bld_sint64(&hp, NULL, ts);
+
+    /* This mapping is needed for the backwards compatible error_logger */
+    switch (tag) {
+    case am_info: el_tag = am_info_msg; break;
+    case am_warning: el_tag = am_warning_msg; break;
+    default:
+        ASSERT(am_error);
+        break;
     }
-#else
-    {
-	ErtsMessage *mp = erts_alloc_message(0, NULL);
-	mp->data.heap_frag = bp;
-	erts_queue_message(p, 0, mp, message, am_system);
+
+    md = MAP2(hp, am_emulator, am_true, ERTS_MAKE_AM("tag"), el_tag);
+    hp += MAP2_SZ;
+
+    if (is_nil(gl) && is_non_value(pid)) {
+        /* no gl and no pid, probably from a port */
+        md = MAP2(hp,
+                  am_error_logger, md,
+                  am_time, time);
+        hp += MAP2_SZ;
+        pid = NIL;
+    } else if (is_nil(gl)) {
+        /* no gl */
+        md = MAP3(hp,
+                  am_error_logger, md,
+                  am_pid, pid,
+                  am_time, time);
+        hp += MAP3_SZ;
+    } else if (is_non_value(pid)) {
+        /* no gl */
+        md = MAP3(hp,
+                  am_error_logger, md,
+                  ERTS_MAKE_AM("gl"), gl,
+                  am_time, time);
+        hp += MAP3_SZ;
+        pid = NIL;
+    } else {
+        md = MAP4(hp,
+                  am_error_logger, md,
+                  ERTS_MAKE_AM("gl"), gl,
+                  am_pid, pid,
+                  am_time, time);
+        hp += MAP4_SZ;
     }
-#endif
+
+    message = TUPLE5(hp, am_log, tag, format, args, md);
+    erts_queue_error_logger_message(pid, message, bp);
 }
 
-/* error_logger !
-   {notify,{info_msg,gleader,{emulator,format,[args]}}} |
-   {notify,{error,gleader,{emulator,format,[args]}}} |
-   {notify,{warning_msg,gleader,{emulator,format,[args}]}} */
-static int do_send_to_logger(Eterm tag, Eterm gleader, char *buf, int len)
+static int do_send_to_logger(Eterm tag, Eterm gl, char *buf, size_t len)
 {
     Uint sz;
-    Eterm gl;
-    Eterm list,args,format,tuple1,tuple2,tuple3;
+    Eterm list, args, format, pid;
+    ErtsMonotonicTime ts;
 
     Eterm *hp = NULL;
     ErlOffHeap *ohp = NULL;
     ErlHeapFragment *bp = NULL;
-    Process *p = NULL;
-
-    ASSERT(is_atom(tag));
-
-    if (len <= 0) {
-	return -1;
-    }
 
     sz = len * 2 /* message list */ + 2 /* cons surrounding message list */
-	+ 3 /*outer 2-tuple*/ + 4 /* middle 3-tuple */ + 4 /*inner 3-tuple */
 	+ 8 /* "~s~n" */;
 
     /* gleader size is accounted and allocated next */
-    gl = do_allocate_logger_message(gleader, &hp, &ohp, &bp, &p, sz);
-
-    if(is_nil(gl)) {
-       /* buf *always* points to a null terminated string */
-       erts_fprintf(stderr, "(no error logger present) %T: \"%s\"\n",
-                    tag, buf);
-       return 0;
-    }
+    gl = do_allocate_logger_message(gl, &ts, &pid, &hp, &ohp, &bp, sz);
 
     list = buf_to_intlist(&hp, buf, len, NIL);
     args = CONS(hp,list,NIL);
     hp += 2;
     format = buf_to_intlist(&hp, "~s~n", 4, NIL);
-    tuple1 = TUPLE3(hp, am_emulator, format, args);
-    hp += 4;
-    tuple2 = TUPLE3(hp, tag, gl, tuple1);
-    hp += 4;
-    tuple3 = TUPLE2(hp, am_notify, tuple2);
 
-    do_send_logger_message(hp, ohp, bp, p, tuple3);
+    do_send_logger_message(gl, tag, format, args, ts, pid, hp, bp);
     return 0;
 }
 
-static int do_send_term_to_logger(Eterm tag, Eterm gleader,
-				  char *buf, int len, Eterm args)
+static int do_send_term_to_logger(Eterm tag, Eterm gl,
+				  char *buf, size_t len, Eterm args)
 {
     Uint sz;
-    Eterm gl;
     Uint args_sz;
-    Eterm format,tuple1,tuple2,tuple3;
+    Eterm format, pid;
+    ErtsMonotonicTime ts;
 
     Eterm *hp = NULL;
     ErlOffHeap *ohp = NULL;
     ErlHeapFragment *bp = NULL;
-    Process *p = NULL;
 
-    ASSERT(is_atom(tag));
+    ASSERT(len > 0);
 
     args_sz = size_object(args);
-    sz = len * 2 /* format */ + args_sz
-	+ 3 /*outer 2-tuple*/ + 4 /* middle 3-tuple */ + 4 /*inner 3-tuple */;
+    sz = len * 2 /* format */ + args_sz;
 
     /* gleader size is accounted and allocated next */
-    gl = do_allocate_logger_message(gleader, &hp, &ohp, &bp, &p, sz);
-
-    if(is_nil(gl)) {
-       /* buf *always* points to a null terminated string */
-       erts_fprintf(stderr, "(no error logger present) %T: \"%s\" %T\n",
-                    tag, buf, args);
-       return 0;
-    }
+    gl = do_allocate_logger_message(gl, &ts, &pid, &hp, &ohp, &bp, sz);
 
     format = buf_to_intlist(&hp, buf, len, NIL);
     args = copy_struct(args, args_sz, &hp, ohp);
-    tuple1 = TUPLE3(hp, am_emulator, format, args);
-    hp += 4;
-    tuple2 = TUPLE3(hp, tag, gl, tuple1);
-    hp += 4;
-    tuple3 = TUPLE2(hp, am_notify, tuple2);
 
-    do_send_logger_message(hp, ohp, bp, p, tuple3);
+    do_send_logger_message(gl, tag, format, args, ts, pid, hp, bp);
     return 0;
 }
 
 static ERTS_INLINE int
-send_info_to_logger(Eterm gleader, char *buf, int len)
+send_info_to_logger(Eterm gleader, char *buf, size_t len)
 {
-    return do_send_to_logger(am_info_msg, gleader, buf, len);
+    return do_send_to_logger(am_info, gleader, buf, len);
 }
 
 static ERTS_INLINE int
-send_warning_to_logger(Eterm gleader, char *buf, int len)
+send_warning_to_logger(Eterm gleader, char *buf, size_t len)
 {
-    Eterm tag;
-    switch (erts_error_logger_warnings) {
-    case am_info:	tag = am_info_msg;	break;
-    case am_warning:	tag = am_warning_msg;	break;
-    default:		tag = am_error;		break;
-    }
-    return do_send_to_logger(tag, gleader, buf, len);
+    return do_send_to_logger(erts_error_logger_warnings, gleader, buf, len);
 }
 
 static ERTS_INLINE int
-send_error_to_logger(Eterm gleader, char *buf, int len)
+send_error_to_logger(Eterm gleader, char *buf, size_t len)
 {
     return do_send_to_logger(am_error, gleader, buf, len);
 }
 
 static ERTS_INLINE int
-send_error_term_to_logger(Eterm gleader, char *buf, int len, Eterm args)
+send_error_term_to_logger(Eterm gleader, char *buf, size_t len, Eterm args)
 {
     return do_send_term_to_logger(am_error, gleader, buf, len, args);
 }
@@ -3182,6 +3162,9 @@ tailrecur_ne:
 		    int cmp;
 		    byte* a_ptr;
 		    byte* b_ptr;
+		    if (eq_only && a_size != b_size) {
+		        RETURN_NEQ(a_size - b_size);
+		    }
 		    ERTS_GET_BINARY_BYTES(a, a_ptr, a_bitoffs, a_bitsize);
 		    ERTS_GET_BINARY_BYTES(b, b_ptr, b_bitoffs, b_bitsize);
 		    if ((a_bitsize | b_bitsize | a_bitoffs | b_bitoffs) == 0) {
@@ -3557,7 +3540,7 @@ store_external_or_ref_(Uint **hpp, ErlOffHeap* oh, Eterm ns)
     if (is_external_header(*from_hp)) {
 	ExternalThing *etp = (ExternalThing *) from_hp;
 	ASSERT(is_external(ns));
-	erts_smp_refc_inc(&etp->node->refc, 2);
+	erts_refc_inc(&etp->node->refc, 2);
     }
     else if (is_ordinary_ref_thing(from_hp))
 	return make_internal_ref(to_hp);
@@ -4390,15 +4373,20 @@ erts_read_env(char *key)
     char *value = erts_alloc(ERTS_ALC_T_TMP, value_len);
     int res;
     while (1) {
-	res = erts_sys_getenv_raw(key, value, &value_len);
-	if (res <= 0)
-	    break;
-	value = erts_realloc(ERTS_ALC_T_TMP, value, value_len);
+        res = erts_sys_explicit_8bit_getenv(key, value, &value_len);
+
+        if (res >= 0) {
+            break;
+        }
+
+        value = erts_realloc(ERTS_ALC_T_TMP, value, value_len);
     }
-    if (res != 0) {
-	erts_free(ERTS_ALC_T_TMP, value);
-	return NULL;
+
+    if (res != 1) {
+        erts_free(ERTS_ALC_T_TMP, value);
+        return NULL;
     }
+
     return value;
 }
 
@@ -4712,22 +4700,6 @@ void
 erts_interval_init(erts_interval_t *icp)
 {
     erts_atomic64_init_nob(&icp->counter.atomic, 0);
-#ifdef DEBUG
-    icp->smp_api = 0;
-#endif
-}
-
-void
-erts_smp_interval_init(erts_interval_t *icp)
-{
-#ifdef ERTS_SMP
-    erts_interval_init(icp);
-#else
-    icp->counter.not_atomic = 0;
-#endif
-#ifdef DEBUG
-    icp->smp_api = 1;
-#endif
 }
 
 static ERTS_INLINE Uint64
@@ -4767,79 +4739,25 @@ ensure_later_interval_acqb(erts_interval_t *icp, Uint64 ic)
 Uint64
 erts_step_interval_nob(erts_interval_t *icp)
 {
-    ASSERT(!icp->smp_api);
     return step_interval_nob(icp);
 }
 
 Uint64
 erts_step_interval_relb(erts_interval_t *icp)
 {
-    ASSERT(!icp->smp_api);
-    return step_interval_relb(icp);
-}
-
-Uint64
-erts_smp_step_interval_nob(erts_interval_t *icp)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
-    return step_interval_nob(icp);
-#else
-    return ++icp->counter.not_atomic;
-#endif
-}
-
-Uint64
-erts_smp_step_interval_relb(erts_interval_t *icp)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
     return step_interval_relb(icp);
-#else
-    return ++icp->counter.not_atomic;
-#endif
 }
 
 Uint64
 erts_ensure_later_interval_nob(erts_interval_t *icp, Uint64 ic)
 {
-    ASSERT(!icp->smp_api);
     return ensure_later_interval_nob(icp, ic);
 }
 
 Uint64
 erts_ensure_later_interval_acqb(erts_interval_t *icp, Uint64 ic)
 {
-    ASSERT(!icp->smp_api);
-    return ensure_later_interval_acqb(icp, ic);
-}
-
-Uint64
-erts_smp_ensure_later_interval_nob(erts_interval_t *icp, Uint64 ic)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
-    return ensure_later_interval_nob(icp, ic);
-#else
-    if (icp->counter.not_atomic > ic)
-	return icp->counter.not_atomic;
-    else
-	return ++icp->counter.not_atomic;
-#endif
-}
-
-Uint64
-erts_smp_ensure_later_interval_acqb(erts_interval_t *icp, Uint64 ic)
-{
-    ASSERT(icp->smp_api);
-#ifdef ERTS_SMP
     return ensure_later_interval_acqb(icp, ic);
-#else
-    if (icp->counter.not_atomic > ic)
-	return icp->counter.not_atomic;
-    else
-	return ++icp->counter.not_atomic;
-#endif
 }
 
 /*
diff --git a/erts/emulator/drivers/common/efile_drv.c b/erts/emulator/drivers/common/efile_drv.c
deleted file mode 100644
index 3a7b3bb50c..0000000000
--- a/erts/emulator/drivers/common/efile_drv.c
+++ /dev/null
@@ -1,4249 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 1996-2017. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-/*
- * Purpose: Provides file and directory operations.
- *
- * This file is generic, and does the work of decoding the commands
- * and encoding the responses.  System-specific functions are found in
- * the unix_efile.c and win_efile.c files.
- */
-
-/* Operations */
-
-#define FILE_OPEN		 1 /* Essential for startup */
-#define FILE_READ		 2
-#define FILE_LSEEK		 3
-#define FILE_WRITE		 4
-#define FILE_FSTAT		 5 /* Essential for startup */
-#define FILE_PWD                 6 /* Essential for startup */
-#define FILE_READDIR             7 /* Essential for startup */
-#define FILE_CHDIR               8
-#define FILE_FSYNC               9
-#define FILE_MKDIR              10
-#define FILE_DELETE             11
-#define FILE_RENAME             12
-#define FILE_RMDIR              13
-#define FILE_TRUNCATE           14
-#define FILE_READ_FILE          15 /* Essential for startup */
-#define FILE_WRITE_INFO		16
-#define FILE_LSTAT            	19
-#define FILE_READLINK        	20
-#define FILE_LINK             	21
-#define FILE_SYMLINK          	22
-#define FILE_CLOSE		23
-#define FILE_PWRITEV		24
-#define FILE_PREADV		25
-#define FILE_SETOPT		26
-#define FILE_IPREAD             27
-#define FILE_ALTNAME            28
-#define FILE_READ_LINE          29
-#define FILE_FDATASYNC          30
-#define FILE_FADVISE            31
-#define FILE_SENDFILE           32
-#define FILE_FALLOCATE          33
-#define FILE_CLOSE_ON_PORT_EXIT 34
-/* Return codes */
-
-#define FILE_RESP_OK         0
-#define FILE_RESP_ERROR      1
-#define FILE_RESP_DATA       2
-#define FILE_RESP_NUMBER     3
-#define FILE_RESP_INFO       4
-#define FILE_RESP_NUMERR     5
-#define FILE_RESP_LDATA      6
-#define FILE_RESP_N2DATA     7
-#define FILE_RESP_EOF        8
-#define FILE_RESP_FNAME      9
-#define FILE_RESP_ALL_DATA  10
-#define FILE_RESP_LFNAME    11
-
-/* Options */
-
-#define FILE_OPT_DELAYED_WRITE 0
-#define FILE_OPT_READ_AHEAD    1
-
-/* IPREAD variants */
-
-#define IPREAD_S32BU_P32BU 0
-
-/* Limits */
-
-#define FILE_SEGMENT_READ  (256*1024)
-#define FILE_SEGMENT_WRITE (256*1024)
-
-/* Internal */
-
-/* Set to 1 to test having read_ahead implicitly for read_line */ 
-#define ALWAYS_READ_LINE_AHEAD 0
-
-
-/* Must not be possible to get from malloc()! */
-#define FILE_FD_INVALID ((Sint)(-1))
-
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-
-#include <ctype.h>
-#include <sys/types.h>
-#include <stdlib.h>
-
-/* Need (NON)BLOCKING macros for sendfile */
-#ifndef WANT_NONBLOCKING
-#define WANT_NONBLOCKING
-#endif
-
-#include "sys.h"
-
-#include "erl_driver.h"
-#include "erl_efile.h"
-#include "erl_threads.h"
-#include "gzio.h"
-#include "dtrace-wrapper.h" 
-
-
-static ErlDrvSysInfo sys_info;
-
-/* For explanation of this var, see comment for same var in erl_async.c */
-static unsigned gcc_optimizer_hack = 0;
-
-#ifdef  USE_VM_PROBES
-
-#define DTRACE_EFILE_BUFSIZ 128
-
-#define DTRACE_INVOKE_SETUP(op) \
-    do { DTRACE3(efile_drv_int_entry, d->sched_i1, d->sched_i2, op); } while (0)
-#define DTRACE_INVOKE_SETUP_BY_NAME(op) \
-    struct t_data *d = (struct t_data *) data ; \
-    DTRACE_INVOKE_SETUP(op)
-#define DTRACE_INVOKE_RETURN(op) \
-    do { DTRACE3(efile_drv_int_return, d->sched_i1, d->sched_i2, \
-                 op); } while (0) ; gcc_optimizer_hack++ ;
-
-/* Assign human-friendlier id numbers to scheduler & I/O worker threads */
-int             dt_driver_idnum = 0;
-int             dt_driver_io_worker_base = 5000;
-erts_mtx_t      dt_driver_mutex;
-pthread_key_t   dt_driver_key;
-
-typedef struct {
-    int         thread_num;
-    Uint64      tag;
-} dt_private;
-
-dt_private *get_dt_private(int);
-#else  /* USE_VM_PROBES */
-#define DTRACE_INVOKE_SETUP(op)            do {} while (0)
-#define DTRACE_INVOKE_SETUP_BY_NAME(op)    do {} while (0)
-#define DTRACE_INVOKE_RETURN(op)           do {} while (0)
-#endif  /* USE_VM_PROBES */
-
-/* #define TRACE 1 */
-#ifdef TRACE
-#    define TRACE_C(c) do { putchar(c); fflush(stdout); } while (0)
-#    define TRACE_S(s) do { fputs((s), stdout); fflush(stdout); } while (0)
-#    define TRACE_F(args) do { printf args ;fflush(stdout); } while (0)
-#else
-#    define TRACE_C(c) ((void)(0))
-#    define TRACE_S(s) ((void)(0))
-#    define TRACE_F(args) ((void)(0))
-#endif
-
-
-#ifdef USE_THREADS
-#define THRDS_AVAILABLE (sys_info.async_threads > 0)
-#ifdef HARDDEBUG /* HARDDEBUG in io.c is expected too */
-#define TRACE_DRIVER fprintf(stderr, "Efile: ")
-#else
-#define TRACE_DRIVER
-#endif
-#define MUTEX_INIT(m, p) do { IF_THRDS { TRACE_DRIVER; (m = driver_pdl_create(p)); } } while (0)
-#define MUTEX_LOCK(m)    do { IF_THRDS { TRACE_DRIVER; driver_pdl_lock(m);   } } while (0)
-#define MUTEX_UNLOCK(m)  do { IF_THRDS { TRACE_DRIVER; driver_pdl_unlock(m); } } while (0)
-#else
-#define THRDS_AVAILABLE (0)
-#define MUTEX_INIT(m, p)
-#define MUTEX_LOCK(m)
-#define MUTEX_UNLOCK(m)
-#endif
-#define IF_THRDS if (THRDS_AVAILABLE)
-
-
-#define SENDFILE_FLGS_USE_THREADS (1 << 0)
-/**
- * On DARWIN sendfile can deadlock with close if called in
- * different threads. So until Apple fixes so that sendfile
- * is not buggy we disable usage of the async pool for
- * DARWIN. The testcase t_sendfile_crashduring reproduces
- * this error when using +A 10 and enabling SENDFILE_FLGS_USE_THREADS.
- */
-#if defined(__APPLE__) && defined(__MACH__)
-#define USE_THRDS_FOR_SENDFILE(DATA) 0
-#else
-#define USE_THRDS_FOR_SENDFILE(DATA) (DATA->flags & SENDFILE_FLGS_USE_THREADS)
-#endif /* defined(__APPLE__) && defined(__MACH__) */
-
-
-
-#if 0
-/* Experimental, for forcing all file operations to use the same thread. */
-   static unsigned file_fixed_key = 1;
-#  define KEY(desc) (&file_fixed_key)
-#else
-#  define KEY(desc) (&(desc)->key)
-#endif
-
-#ifndef MAX
-#  define MAX(x, y) (((x) > (y)) ? (x) : (y))
-#endif
-
-#ifdef FILENAMES_16BIT
-#ifdef USE_VM_PROBES
-#error 16bit characters in filenames and dtrace in combination is not supported.
-#endif
-#  define FILENAME_BYTELEN(Str) filename_len_16bit(Str)
-#  define FILENAME_COPY(To,From) filename_cpy_16bit((To),(From)) 
-#  define FILENAME_CHARSIZE 2
-
-   static int filename_len_16bit(char *str) 
-   {
-       char *p = str;
-       while(*p != '\0' || p[1] != '\0') {
-	   p += 2;
-       }
-       return (p - str);
-   }
-
-   static void filename_cpy_16bit(char *to, char *from) 
-   {
-       while(*from != '\0' || from[1] != '\0') {
-	   *to++ = *from++;
-	   *to++ = *from++;
-       }
-       *to++ = *from++;
-       *to++ = *from++;
-   }
-
-#else
-#  define FILENAME_BYTELEN(Str) strlen(Str)
-#  define FILENAME_COPY(To,From) strcpy(To,From) 
-#  define FILENAME_CHARSIZE 1
-#endif
-
-#if     (MAXPATHLEN+1)*FILENAME_CHARSIZE+1 > BUFSIZ
-#  define    RESBUFSIZE  ((MAXPATHLEN+1)*FILENAME_CHARSIZE+1)
-#else
-#  define    RESBUFSIZE  BUFSIZ
-#endif
-
-#define READDIR_CHUNKS (5)
-
-
-
-#if ALWAYS_READ_LINE_AHEAD
-#define DEFAULT_LINEBUF_SIZE 2048
-#else
-#define DEFAULT_LINEBUF_SIZE 512 /* Small, it's usually discarded anyway */ 
-#endif
-
-typedef unsigned char uchar;
-
-static ErlDrvData file_start(ErlDrvPort port, char* command);
-static int file_init(void);
-static void file_stop(ErlDrvData);
-static void file_output(ErlDrvData, char* buf, ErlDrvSizeT len);
-static ErlDrvSSizeT file_control(ErlDrvData, unsigned int command,
-				 char* buf, ErlDrvSizeT len,
-				 char **rbuf, ErlDrvSizeT rlen);
-static void file_timeout(ErlDrvData);
-static void file_outputv(ErlDrvData, ErlIOVec*);
-static void file_async_ready(ErlDrvData, ErlDrvThreadData);
-static void file_flush(ErlDrvData);
-
-#ifdef HAVE_SENDFILE
-static void file_ready_output(ErlDrvData data, ErlDrvEvent event);
-static void file_stop_select(ErlDrvEvent event, void* _);
-#endif /* HAVE_SENDFILE */
-
-
-enum e_timer {timer_idle, timer_again, timer_write};
-#ifdef HAVE_SENDFILE
-enum e_sendfile {sending, not_sending};
-#define SENDFILE_USE_THREADS (1 << 0)
-#endif /* HAVE_SENDFILE */
-
-struct t_data;
-
-typedef struct {
-    SWord           fd;
-    ErlDrvPort      port;
-    unsigned int    key;      /* Async queue key */
-    unsigned        flags;    /* Original flags from FILE_OPEN. */
-    void          (*invoke)(void *);
-    struct t_data  *d;
-    void          (*free)(void *);
-    struct t_data  *cq_head;  /* Queue of incoming commands */
-    struct t_data  *cq_tail;  /* -""- */
-    enum e_timer    timer_state;
-#ifdef HAVE_SENDFILE
-    enum e_sendfile sendfile_state;
-#endif /* HAVE_SENDFILE */
-    size_t          read_bufsize;
-    ErlDrvBinary   *read_binp;
-    size_t          read_offset;
-    size_t          read_size;
-    size_t          write_bufsize;
-    unsigned long   write_delay;
-    int             write_error;
-    Efile_error     write_errInfo;
-    ErlDrvPDL       q_mtx;    /* Mutex for the driver queue, known by the emulator. Also used for
-				 mutual exclusion when accessing field(s) below. */
-    size_t          write_buffered;
-#ifdef USE_VM_PROBES
-    int             idnum;      /* Unique ID # for this driver thread/desc */
-    char            port_str[DTRACE_TERM_BUF_SIZE];
-#endif
-} file_descriptor;
-
-
-static int reply_error(file_descriptor*, Efile_error* errInfo);
-
-struct erl_drv_entry efile_driver_entry = {
-    file_init,
-    file_start,
-    file_stop,
-    file_output,
-    NULL,
-#ifdef HAVE_SENDFILE
-    file_ready_output,
-#else
-    NULL,
-#endif /* HAVE_SENDFILE */
-    "efile",
-    NULL,
-    NULL,
-    file_control,
-    file_timeout,
-    file_outputv,
-    file_async_ready,
-    file_flush,
-    NULL,
-    NULL,
-    ERL_DRV_EXTENDED_MARKER,
-    ERL_DRV_EXTENDED_MAJOR_VERSION,
-    ERL_DRV_EXTENDED_MINOR_VERSION,
-    ERL_DRV_FLAG_USE_PORT_LOCKING,
-    NULL,
-    NULL,
-#ifdef HAVE_SENDFILE
-    file_stop_select
-#else
-    NULL
-#endif /* HAVE_SENDFILE */
-};
-
-
-
-static int thread_short_circuit;
-
-#define DRIVER_ASYNC(level, desc, f_invoke, data, f_free) \
-if (thread_short_circuit >= (level)) { \
-    (*(f_invoke))(data); \
-    file_async_ready((ErlDrvData)(desc), (data)); \
-} else { \
-    driver_async((desc)->port, KEY(desc), (f_invoke), (data), (f_free)); \
-}
-
-
-
-struct t_pbuf_spec {
-    Sint64 offset;
-    size_t size;
-};
-
-struct t_pwritev {
-    ErlDrvPort         port;
-    ErlDrvPDL          q_mtx;
-    size_t             size;
-    unsigned           cnt;
-    unsigned           n;
-    struct t_pbuf_spec specs[1];
-};
-
-struct t_preadv {
-    ErlIOVec eiov;
-    unsigned n;
-    unsigned cnt;
-    size_t   size;
-    Sint64   offsets[1];
-};
-
-#define READDIR_BUFSIZE (8*1024)*READDIR_CHUNKS
-#if READDIR_BUFSIZE < (1 + (2 + MAXPATHLEN)*FILENAME_CHARSIZE*READDIR_CHUNKS)
-#  undef READDIR_BUFSIZE
-#  define READDIR_BUFSIZE (1 + (2 + MAXPATHLEN)*FILENAME_CHARSIZE*READDIR_CHUNKS)
-#endif
-
-struct t_readdir_buf {
-     struct t_readdir_buf *next;
-     size_t n;
-     char buf[READDIR_BUFSIZE];
-};
-
-struct t_data
-{
-    struct t_data *next;
-    int            command;
-    int            level;
-    void         (*invoke)(void *);
-    void         (*free)(void *);
-    void           *data_to_free; /* used by FILE_CLOSE_ON_PORT_EXIT only */
-    int            again;
-    int            reply;
-#ifdef  USE_VM_PROBES
-    int               sched_i1;
-    Uint64            sched_i2;
-    char              sched_utag[DTRACE_EFILE_BUFSIZ+1];
-#endif
-    int            result_ok;
-    Efile_error    errInfo;
-    int            flags;
-    SWord          fd;
-    int            is_fd_unused;
-    /**/
-    Efile_info        info;
-    EFILE_DIR_HANDLE  dir_handle; /* Handle to open directory. */
-    ErlDrvBinary     *bin;
-    int               drive;
-    size_t            n;
-    /*off_t             offset;*/
-    /*size_t            bytesRead; Bytes read from the file. */
-    /**/
-    union {
-	struct {
-	    Sint64 offset;
-	    int    origin;
-	    Sint64 location;
-	} lseek;
-	struct {
-	    ErlDrvPort    port;
-	    ErlDrvPDL     q_mtx;
-	    size_t        size;
-	    size_t        reply_size;
-	} writev;
-	struct t_pwritev pwritev;
-	struct t_preadv  preadv;
-	struct {
-	    ErlDrvBinary *binp;
-	    size_t        bin_offset;
-	    size_t        bin_size;
-	    size_t        size;
-	} read;
-	struct {
-	    ErlDrvBinary *binp; /* in - out */
-	    size_t        read_offset; /* in - out */
-	    size_t        read_size; /* in - out */
-	    size_t        nl_pos; /* out */
-	    short         nl_skip; /* out, 0 or 1 */
-#if !ALWAYS_READ_LINE_AHEAD
-	    short         read_ahead; /* in, bool */
-#endif
-	} read_line;
-	struct {
-	    ErlDrvBinary *binp;
-	    int           size;
-	    int           offset;
-	} read_file;
-	struct {
-	    struct t_readdir_buf *first_buf;
-	    struct t_readdir_buf *last_buf;
-	} read_dir;
-	struct {
-	    Sint64 offset;
-	    Sint64 length;
-	    int advise;
-	} fadvise;
-#ifdef HAVE_SENDFILE
-	struct {
-	    ErlDrvPort port;
-	    ErlDrvPDL q_mtx;
-	    int out_fd;
-	    off_t offset;
-	    Uint64 nbytes;
-	    Uint64 written;
-	} sendfile;
-#endif /* HAVE_SENDFILE */
-	struct {
-	    Sint64 offset;
-	    Sint64 length;
-	} fallocate;
-    } c;
-    char b[1];
-};
-
-#define EF_ALLOC(S)		driver_alloc((S))
-#define EF_REALLOC(P, S)	driver_realloc((P), (S))
-#define EF_SAFE_ALLOC(S)	ef_safe_alloc((S))
-#define EF_SAFE_REALLOC(P, S)	ef_safe_realloc((P), (S))
-#define EF_FREE(P)		do { if((P)) driver_free((P)); } while(0)
-
-static void *ef_safe_alloc(Uint s)
-{
-    void *p = EF_ALLOC(s);
-    if (!p) erts_exit(ERTS_ERROR_EXIT, "efile drv: Can't allocate %lu bytes of memory\n", (unsigned long)s);
-    return p;
-}
-
-/*********************************************************************
- * ErlIOVec manipulation functions.
- */
-
-/* char EV_CHAR_P(ErlIOVec *ev, int p, int q) */
-#define EV_CHAR_P(ev, p, q)			\
-    (((char *)(ev)->iov[q].iov_base) + (p))
-
-/* int EV_GET_CHAR(ErlIOVec *ev, char *p, int *pp, int *qp) */
-#define EV_GET_CHAR(ev, p, pp, qp) efile_ev_get_char(ev, p ,pp, qp)
-static int
-efile_ev_get_char(ErlIOVec *ev, char *p, size_t *pp, size_t *qp) {
-    if (*pp + 1 <= ev->iov[*qp].iov_len) {
-	*p = *EV_CHAR_P(ev, *pp, *qp);
-	if (*pp + 1 < ev->iov[*qp].iov_len)
-	    *pp += 1;
-	else {
-	    *qp += 1;
-	    *pp = 0;
-	}
-	return !0;
-    }
-    return 0;
-}
-
-/* Uint32 EV_UINT32(ErlIOVec *ev, int p, int q)*/
-#define EV_UINT32(ev, p, q)						\
-    ((Uint32) ((unsigned char *)(ev)->iov[q].iov_base)[p])
-
-/* int EV_GET_UINT32(ErlIOVec *ev, Uint32 *p, int *pp, int *qp) */
-#define EV_GET_UINT32(ev, p, pp, qp) efile_ev_get_uint32(ev, p, pp, qp)
-static int
-efile_ev_get_uint32(ErlIOVec *ev, Uint32 *p, size_t *pp, size_t *qp) {
-    if (*pp + 4 <= ev->iov[*qp].iov_len) {
-	*p = (EV_UINT32(ev, *pp,   *qp) << 24)
-	    | (EV_UINT32(ev, *pp + 1, *qp) << 16)
-	    | (EV_UINT32(ev, *pp + 2, *qp) << 8)
-	    | (EV_UINT32(ev, *pp + 3, *qp));
-	if (*pp + 4 < ev->iov[*qp].iov_len)
-	    *pp += 4;
-	else {
-	    *qp += 1;
-	    *pp = 0;
-	}
-	return !0;
-    }
-    return 0;
-}
-
-/* Uint64 EV_UINT64(ErlIOVec *ev, int p, int q)*/
-#define EV_UINT64(ev, p, q)						\
-    ((Uint64) ((unsigned char *)(ev)->iov[q].iov_base)[p])
-
-/* int EV_GET_UINT64(ErlIOVec *ev, Uint64 *p, int *pp, int *qp) */
-#define EV_GET_UINT64(ev, p, pp, qp) efile_ev_get_uint64(ev, p, pp, qp)
-static int
-efile_ev_get_uint64(ErlIOVec *ev, Uint64 *p, size_t *pp, size_t *qp) {
-    if (*pp + 8 <= ev->iov[*qp].iov_len) {
-	*p = (EV_UINT64(ev, *pp, *qp) << 56)
-	    | (EV_UINT64(ev, *pp + 1, *qp) << 48)
-	    | (EV_UINT64(ev, *pp + 2, *qp) << 40)
-	    | (EV_UINT64(ev, *pp + 3, *qp) << 32)
-	    | (EV_UINT64(ev, *pp + 4, *qp) << 24)
-	    | (EV_UINT64(ev, *pp + 5, *qp) << 16)
-	    | (EV_UINT64(ev, *pp + 6, *qp) << 8)
-	    | (EV_UINT64(ev, *pp + 7, *qp));
-	if (*pp + 8 < ev->iov[*qp].iov_len)
-	    *pp += 8;
-	else {
-	    *qp += 1;
-	    *pp = 0;
-	}
-	return !0;
-    }
-    return 0;
-}
-
-/* int EV_GET_SINT64(ErlIOVec *ev, Uint64 *p, int *pp, int *qp) */
-#define EV_GET_SINT64(ev, p, pp, qp) efile_ev_get_sint64(ev, p, pp, qp)
-static int
-efile_ev_get_sint64(ErlIOVec *ev, Sint64 *p, size_t *pp, size_t *qp) {
-    Uint64 *tmp = (Uint64*)p;
-    return EV_GET_UINT64(ev, tmp, pp, qp);
-}
-
-#if 0
-
-static void ev_clear(ErlIOVec *ev) {
-    ASSERT(ev);
-    ev->size = 0;
-    ev->vsize = 0;
-    ev->iov = NULL;
-    ev->binv = NULL;
-}
-
-/* Assumes that ->iov and ->binv were allocated with sys_alloc().
- */
-static void ev_free(ErlIOVec *ev) {
-    if (! ev) {
-	return;
-    }
-    if (ev->vsize > 0) {
-	int i;
-	ASSERT(ev->iov);
-	ASSERT(ev->binv);
-	for (i = 0; i < ev->vsize; i++) {
-	    if (ev->binv[i]) {
-		driver_free_binary(ev->binv[i]);
-	    }
-	}
-	EF_FREE(ev->iov);
-	EF_FREE(ev->binv);
-    }
-}
-
-/* Copy the contents from source to dest.
- * Data in binaries is not copied, just the pointers; 
- * and refc is incremented.
- */
-static ErlIOVec *ev_copy(ErlIOVec *dest, ErlIOVec *source) {
-    int *ip;
-    ASSERT(dest);
-    ASSERT(source);
-    if (source->vsize == 0) {
-	/* Empty source */
-	ev_clear(dest);
-	return dest;
-    }
-    /* Allocate ->iov and ->binv */
-    dest->iov = EF_ALLOC(sizeof(*dest->iov) * source->vsize);
-    if (! dest->iov) {
-	return NULL;
-    }
-    dest->binv = EF_ALLOC(sizeof(*dest->binv) * source->vsize);
-    if (! dest->binv) {
-	EF_FREE(dest->iov);
-	return NULL;
-    }
-    dest->size = source->size;
-    /* Copy one vector element at the time. 
-     * Use *ip as an alias for dest->vsize to improve readabiliy.
-     * Keep dest consistent in every iteration by using 
-     * dest->vsize==*ip as loop variable.
-     */
-    for (ip = &dest->vsize, *ip = 0;  *ip < source->vsize;  (*ip)++) {
-	if (source->iov[*ip].iov_len == 0) {
-	    /* Empty vector element */
-	    dest->iov[*ip].iov_len = 0;
-	    dest->iov[*ip].iov_base = NULL;
-	    dest->binv[*ip] = NULL;
-	} else {
-	    /* Non empty vector element */
-	    if (source->binv[*ip]) {
-		/* Contents in binary - copy pointers and increment refc */
-		dest->iov[*ip] = source->iov[*ip];
-		dest->binv[*ip] = source->binv[*ip];
-		driver_binary_inc_refc(source->binv[*ip]);
-	    } else {
-		/* Contents not in binary - allocate new binary and copy data */
-		if (! (dest->binv[*ip] = 
-		       driver_alloc_binary(source->iov[*ip].iov_len))) {
-		    goto failed;
-		}
-		sys_memcpy(dest->binv[*ip]->orig_bytes,
-			   source->iov[*ip].iov_base,
-			   source->iov[*ip].iov_len);
-		dest->iov[*ip].iov_base = dest->binv[*ip]->orig_bytes;
-		dest->iov[*ip].iov_len = source->iov[*ip].iov_len;
-	    }
-	}
-    }
-    return dest;
- failed:
-    ev_free(dest);
-    return NULL;
-}
-
-#endif
-
-
-
-/*********************************************************************
- * Command queue functions
- */
-
-static void cq_enq(file_descriptor *desc, struct t_data *d) {
-    ASSERT(d);
-    if (desc->cq_head) {
-	ASSERT(desc->cq_tail);
-	ASSERT(!desc->cq_tail->next);
-	desc->cq_tail = desc->cq_tail->next = d;
-    } else {
-	ASSERT(desc->cq_tail == NULL);
-	desc->cq_head = desc->cq_tail = d;
-    }
-    d->next = NULL;
-}
-
-static struct t_data *cq_deq(file_descriptor *desc) {
-    struct t_data *d = desc->cq_head;
-    ASSERT(d || (!d && !desc->cq_tail));
-    if (d) {
-	ASSERT(!d->next || (d->next && desc->cq_tail != d));
-	if ((desc->cq_head = d->next) == NULL) {
-	    ASSERT(desc->cq_tail == d);
-	    desc->cq_tail = NULL;
-	}
-    }	
-    return d;
-}
-
-
-/*********************************************************************
- * Driver entry point -> init
- */
-static int 
-file_init(void)
-{
-    char buf[21]; /* enough to hold any 64-bit integer */
-    size_t bufsz = sizeof(buf);
-    thread_short_circuit = (erl_drv_getenv("ERL_EFILE_THREAD_SHORT_CIRCUIT",
-					   buf,
-					   &bufsz) == 0
-			    ? atoi(buf)
-			    : 0);
-    driver_system_info(&sys_info, sizeof(ErlDrvSysInfo));
-
-    /* run initiation of efile_driver if needed */
-    efile_init();
-
-#ifdef  USE_VM_PROBES
-    erts_mtx_init(&dt_driver_mutex, "efile_drv dtrace mutex", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
-    pthread_key_create(&dt_driver_key, NULL);
-#endif  /* USE_VM_PROBES */
-
-    return 0;
-}
-
-
-/*********************************************************************
- * Driver entry point -> start
- */
-static ErlDrvData 
-file_start(ErlDrvPort port, char* command) 
-
-{
-    file_descriptor* desc;
-
-    if ((desc = (file_descriptor*) EF_ALLOC(sizeof(file_descriptor)))
-	== NULL) {
-	errno = ENOMEM;
-	return ERL_DRV_ERROR_ERRNO;
-    }
-    desc->fd = FILE_FD_INVALID;
-    desc->port = port;
-    desc->key = driver_async_port_key(port);
-    desc->flags = 0;
-    desc->invoke = NULL;
-    desc->d = NULL;
-    desc->free = NULL;
-    desc->cq_head = NULL;
-    desc->cq_tail = NULL;
-    desc->timer_state = timer_idle;
-#ifdef HAVE_SENDFILE
-    desc->sendfile_state = not_sending;
-#endif
-    desc->read_bufsize = 0;
-    desc->read_binp = NULL;
-    desc->read_offset = 0;
-    desc->read_size = 0;
-    desc->write_delay = 0L;
-    desc->write_bufsize = 0;
-    desc->write_error = 0;
-    MUTEX_INIT(desc->q_mtx, port); /* Refc is one, referenced by emulator now */
-    desc->write_buffered = 0;
-#ifdef  USE_VM_PROBES
-    dtrace_drvport_str(port, desc->port_str);
-    get_dt_private(0);           /* throw away return value */
-#endif  /* USE_VM_PROBES */
-    return (ErlDrvData) desc;
-}
-
-static void do_close(int flags, SWord fd) {
-    if (flags & EFILE_COMPRESSED) {
-	erts_gzclose((ErtsGzFile)(fd));
-    } else {
-	efile_closefile((int) fd);
-    }
-}
-
-static void invoke_close(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    DTRACE_INVOKE_SETUP(FILE_CLOSE);
-    d->again = 0;
-    do_close(d->flags, d->fd);
-    DTRACE_INVOKE_RETURN(FILE_CLOSE);
-}
-
-static void free_data(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-
-    switch (d->command) {
-    case FILE_OPEN:
-        if (d->is_fd_unused && d->fd != FILE_FD_INVALID) {
-            /* This is OK to do in scheduler thread because there can be no async op
-               ongoing for this fd here, as we exited during async open.
-               Ideally, this close should happen in an async thread too, but that would
-               require a substantial rewrite, as we are here because of a dead port and
-               cannot schedule async jobs for that port any more... */
-            do_close(d->flags, d->fd);
-        }
-        break;
-    case FILE_CLOSE_ON_PORT_EXIT:
-        EF_FREE(d->data_to_free);
-        break;
-    }
-
-    EF_FREE(data);
-}
-
-
-/*
- * Sends back an error reply to Erlang.
- */
-
-static void reply_posix_error(file_descriptor *desc, int posix_errno) {
-    char response[256];		/* Response buffer. */
-    char* s;
-    char* t;
-    
-    /*
-     * Contents of buffer sent back:
-     *
-     * +-----------------------------------------+
-     * | FILE_RESP_ERROR | Posix error id string |
-     * +-----------------------------------------+
-     */
-
-    TRACE_C('E');
-
-    response[0] = FILE_RESP_ERROR;
-    for (s = erl_errno_id(posix_errno), t = response+1; *s; s++, t++)
-	*t = tolower(*s);
-    driver_output2(desc->port, response, t-response, NULL, 0);
-}
-
-static void reply_Uint_posix_error(file_descriptor *desc, Uint num, 
-				   int posix_errno) {
-    char response[256];		/* Response buffer. */
-    char* s;
-    char* t;
-    
-    /*
-     * Contents of buffer sent back:
-     *
-     * +----------------------------------------------------------------------+
-     * | FILE_RESP_NUMERR | 64-bit number (big-endian) | Posix error id string |
-     * +----------------------------------------------------------------------+
-     */
-
-    TRACE_C('N');
-
-    response[0] = FILE_RESP_NUMERR;
-#if SIZEOF_VOID_P == 4
-    put_int32(0, response+1);
-#else
-    put_int32(num>>32, response+1);
-#endif
-    put_int32((Uint32)num, response+1+4);
-    for (s = erl_errno_id(posix_errno), t = response+1+4+4; *s; s++, t++)
-	*t = tolower(*s);
-    driver_output2(desc->port, response, t-response, NULL, 0);
-}
-
-#ifdef HAVE_SENDFILE
-static void reply_string_error(file_descriptor *desc, char* str) {
-    char response[256];		/* Response buffer. */
-    char* s;
-    char* t;
-
-    response[0] = FILE_RESP_ERROR;
-    for (s = str, t = response+1; *s; s++, t++)
-	*t = tolower(*s);
-    driver_output2(desc->port, response, t-response, NULL, 0);
-}
-#endif
-
-static int reply_error(file_descriptor *desc, 
-		       Efile_error *errInfo) /* The error codes. */
-{
-    reply_posix_error(desc, errInfo->posix_errno);
-    return 0;
-}
-
-static int reply_Uint_error(file_descriptor *desc, Uint num, 
-			    Efile_error *errInfo) /* The error codes. */
-{
-    reply_Uint_posix_error(desc, num, errInfo->posix_errno);
-    return 0;
-}
-
-static int reply_ok(file_descriptor *desc) {
-    char c = FILE_RESP_OK;
-
-    driver_output2(desc->port, &c, 1, NULL, 0);
-    return 0;
-}
-
-static int reply(file_descriptor *desc, int ok, Efile_error *errInfo) {
-    if (!ok) {
-	reply_error(desc, errInfo);
-    } else {
-	TRACE_C('K');
-	reply_ok(desc);
-    }
-    return 0;
-}
-
-static int reply_Uint(file_descriptor *desc, Uint result) {
-    char tmp[1+4+4];
-
-    /*
-     * Contents of buffer sent back:
-     *
-     * +-----------------------------------------------+
-     * | FILE_RESP_NUMBER | 64-bit number (big-endian) |
-     * +-----------------------------------------------+
-     */
-
-    TRACE_C('R');
-
-    tmp[0] = FILE_RESP_NUMBER;
-#if SIZEOF_VOID_P == 4
-    put_int32(0, tmp+1);
-#else
-    put_int32(result>>32, tmp+1);
-#endif
-    put_int32((Uint32)result, tmp+1+4);
-    driver_output2(desc->port, tmp, sizeof(tmp), NULL, 0);
-    return 0;
-}
-
-static int reply_Sint64(file_descriptor *desc, Sint64 result) {
-    char tmp[1+4+4];
-
-    /*
-     * Contents of buffer sent back:
-     *
-     * +-----------------------------------------------+
-     * | FILE_RESP_NUMBER | 64-bit number (big-endian) |
-     * +-----------------------------------------------+
-     */
-
-    TRACE_C('R');
-
-    tmp[0] = FILE_RESP_NUMBER;
-    put_int64(result, tmp+1);
-    driver_output2(desc->port, tmp, sizeof(tmp), NULL, 0);
-    return 0;
-}
-
-#if 0
-static void reply_again(file_descriptor *desc) {
-    char tmp[1];
-    tmp[0] = FILE_RESP_AGAIN;
-    driver_output2(desc->port, tmp, sizeof(tmp), NULL, 0);
-}
-#endif
-
-static void reply_ev(file_descriptor *desc, char response, ErlIOVec *ev) {
-    char tmp[1];
-    /* Data arriving at the Erlang process:
-     * [Response, Binary0, Binary1, .... | BinaryN-1]
-     */
-    tmp[0] = response;
-    driver_outputv(desc->port, tmp, sizeof(tmp), ev, 0);
-}
-
-static void reply_data(file_descriptor *desc, 
-		       ErlDrvBinary *binp, size_t offset, size_t len) {
-    char header[1+4+4];
-    /* Data arriving at the Erlang process:
-     * [?FILE_RESP_DATA, 64-bit length (big-endian) | Data]
-     */
-    header[0] = FILE_RESP_DATA;
-#if SIZEOF_SIZE_T == 4
-    put_int32(0, header+1);
-#else
-    put_int32(len>>32, header+1);
-#endif
-    put_int32((Uint32)len, header+1+4);
-    driver_output_binary(desc->port, header, sizeof(header),
-			 binp, offset, len);
-}
-
-static void reply_buf(file_descriptor *desc, char *buf, size_t len) {
-    char header[1+4+4];
-    /* Data arriving at the Erlang process:
-     * [?FILE_RESP_DATA, 64-bit length (big-endian) | Data]
-     */
-    header[0] = FILE_RESP_DATA;
-#if SIZEOF_SIZE_T == 4
-    put_int32(0, header+1);
-#else
-    put_int32(len>>32, header+1);
-#endif
-    put_int32((Uint32)len, header+1+4);
-    driver_output2(desc->port, header, sizeof(header), buf, len);
-}
-
-static int reply_eof(file_descriptor *desc) {
-    char c = FILE_RESP_EOF;
-
-    driver_output2(desc->port, &c, 1, NULL, 0);
-    return 0;
-}
- 
-static void invoke_name(void *data, int (*f)(Efile_error *, char *))
-{
-    struct t_data *d = (struct t_data *) data;
-    char *name = (char *) d->b;
-
-    d->again = 0;
-    d->result_ok = (*f)(&d->errInfo, name);
-}
-
-static void invoke_mkdir(void *data)
-{
-    DTRACE_INVOKE_SETUP_BY_NAME(FILE_MKDIR);
-    invoke_name(data, efile_mkdir);
-    DTRACE_INVOKE_RETURN(FILE_MKDIR);
-}
-
-static void invoke_rmdir(void *data)
-{
-    DTRACE_INVOKE_SETUP_BY_NAME(FILE_RMDIR);
-    invoke_name(data, efile_rmdir);
-    DTRACE_INVOKE_RETURN(FILE_RMDIR);
-}
-
-static void invoke_delete_file(void *data)
-{
-    DTRACE_INVOKE_SETUP_BY_NAME(FILE_DELETE);
-    invoke_name(data, efile_delete_file);
-    DTRACE_INVOKE_RETURN(FILE_DELETE);
-}
-
-static void invoke_chdir(void *data)
-{
-    DTRACE_INVOKE_SETUP_BY_NAME(FILE_CHDIR);
-    invoke_name(data, efile_chdir);
-    DTRACE_INVOKE_RETURN(FILE_CHDIR);
-}
-
-static void invoke_fdatasync(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int fd = (int) d->fd;
-    DTRACE_INVOKE_SETUP(FILE_FDATASYNC);
-
-    d->again = 0;
-    d->result_ok = efile_fdatasync(&d->errInfo, fd);
-    DTRACE_INVOKE_RETURN(FILE_FDATASYNC);
-}
-
-static void invoke_fsync(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int fd = (int) d->fd;
-    DTRACE_INVOKE_SETUP(FILE_FSYNC);
-
-    d->again = 0;
-    d->result_ok = efile_fsync(&d->errInfo, fd);
-    DTRACE_INVOKE_RETURN(FILE_FSYNC);
-}
-
-static void invoke_truncate(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int fd = (int) d->fd;
-    DTRACE_INVOKE_SETUP(FILE_TRUNCATE);
-
-    d->again = 0;
-    d->result_ok = efile_truncate_file(&d->errInfo, &fd, d->flags);
-    DTRACE_INVOKE_RETURN(FILE_TRUNCATE);
-}
-
-static void invoke_read(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int status, segment;
-    size_t size, read_size;
-    DTRACE_INVOKE_SETUP(FILE_READ);
-
-    segment = d->again && d->c.read.bin_size >= 2*FILE_SEGMENT_READ;
-    if (segment) {
-	size = FILE_SEGMENT_READ;
-    } else {
-	size = d->c.read.bin_size;
-    }
-    read_size = size;
-    if (d->flags & EFILE_COMPRESSED) {
-	read_size = erts_gzread((ErtsGzFile)d->fd,
-				d->c.read.binp->orig_bytes + d->c.read.bin_offset,
-				size);
-	status = (read_size != (size_t) -1);
-	if (!status) {
-	    d->errInfo.posix_errno = EIO;
-	}
-    } else {
-	status = efile_read(&d->errInfo, d->flags, (int) d->fd,
-			    d->c.read.binp->orig_bytes + d->c.read.bin_offset,
-			    size,
-			    &read_size);
-    }
-    if ( (d->result_ok = status)) {
-	ASSERT(read_size <= size);
-	d->c.read.bin_offset += read_size;
-	if (read_size < size || !segment) {
-	    d->c.read.bin_size = 0;
-	    d->again = 0;
-	} else {
-	    d->c.read.bin_size -= read_size;
-	}
-    } else {
-	d->again = 0;
-    }
-    DTRACE_INVOKE_RETURN(FILE_READ);
-}
-
-static void free_read(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-
-    driver_free_binary(d->c.read.binp);
-    EF_FREE(d);
-}
-
-static void invoke_read_line(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int status;
-    size_t read_size = 0;
-    int local_loop = (d->again == 0);
-    DTRACE_INVOKE_SETUP(FILE_READ_LINE);
-
-    do {
-	size_t size = (d->c.read_line.binp)->orig_size - 
-	    d->c.read_line.read_offset - d->c.read_line.read_size;
-	if (size == 0) {
-	    /* Need more place */
-	    ErlDrvSizeT need = (d->c.read_line.read_size >= DEFAULT_LINEBUF_SIZE) ?
-		d->c.read_line.read_size + DEFAULT_LINEBUF_SIZE : DEFAULT_LINEBUF_SIZE;
-	    ErlDrvBinary   *newbin;
-#if !ALWAYS_READ_LINE_AHEAD
-	    /* Use read_ahead size if need does not exceed it */
-	    if (need < (d->c.read_line.binp)->orig_size && 
-		d->c.read_line.read_ahead)
-	      need = (d->c.read_line.binp)->orig_size;
-#endif
-	    newbin = driver_alloc_binary(need);
-	    if (newbin == NULL) {
-		d->result_ok = 0;
-		d->errInfo.posix_errno = ENOMEM;
-		d->again = 0;
-		break;
-	    }
-	    memcpy(newbin->orig_bytes, (d->c.read_line.binp)->orig_bytes + d->c.read_line.read_offset,  
-		   d->c.read_line.read_size);
-	    driver_free_binary(d->c.read_line.binp);
-	    d->c.read_line.binp = newbin;
-	    d->c.read_line.read_offset = 0;
-	    size = need - d->c.read_line.read_size;
-	}
-	if (d->flags & EFILE_COMPRESSED) {
-	    read_size = erts_gzread((ErtsGzFile)d->fd,
-				    d->c.read_line.binp->orig_bytes + 
-				    d->c.read_line.read_offset + d->c.read_line.read_size,
-				    size);
-	    status = (read_size != (size_t) -1);
-	    if (!status) {
-		d->errInfo.posix_errno = EIO;
-	    }
-	} else {
-	    status = efile_read(&d->errInfo, d->flags, (int) d->fd,
-				d->c.read_line.binp->orig_bytes + 
-				d->c.read_line.read_offset + d->c.read_line.read_size,
-				size,
-				&read_size);
-	}
-	if ( (d->result_ok = status)) {
-	    void *nl_ptr = memchr((d->c.read_line.binp)->orig_bytes + 
-				  d->c.read_line.read_offset + d->c.read_line.read_size,'\n',read_size);
-	    ASSERT(read_size <= size);
-	    d->c.read_line.read_size += read_size;
-	    if (nl_ptr != NULL) {
-		/* If found, we're done */
-		d->c.read_line.nl_pos = ((char *) nl_ptr) - 
-		    ((char *) ((d->c.read_line.binp)->orig_bytes)) + 1;
-		if (d->c.read_line.nl_pos > 1 &&
-		    *(((char *) nl_ptr) - 1) == '\r') {
-		    --d->c.read_line.nl_pos;
-		    *(((char *) nl_ptr) - 1) = '\n';
-		    d->c.read_line.nl_skip = 1;
-		} else {
-		    d->c.read_line.nl_skip = 0;
-		}
-		d->again = 0;
-#if !ALWAYS_READ_LINE_AHEAD
-		if (!(d->c.read_line.read_ahead)) {
-		    /* Ouch! Undo buffering... */
-		    size_t too_much = d->c.read_line.read_size - d->c.read_line.nl_skip - 
-			(d->c.read_line.nl_pos - d->c.read_line.read_offset);
-		    d->c.read_line.read_size -= too_much;
-		    ASSERT(d->c.read_line.read_size >= 0);
-		    if (d->flags & EFILE_COMPRESSED) {
-			Sint64 location = erts_gzseek((ErtsGzFile)d->fd,
-						      -((Sint64) too_much), EFILE_SEEK_CUR);
-			if (location == -1) {
-			    d->result_ok = 0;
-			    d->errInfo.posix_errno = errno;
-			}
-		    } else {
-			Sint64 location;
-			d->result_ok = efile_seek(&d->errInfo, (int) d->fd, 
-						-((Sint64) too_much), EFILE_SEEK_CUR,
-						&location);
-		    }
-		}
-#endif
-		break;
-	    } else if (read_size == 0) {
-		d->c.read_line.nl_pos = 
-		    d->c.read_line.read_offset + d->c.read_line.read_size;
-		d->c.read_line.nl_skip = 0;
-		d->again = 0;
-		break;
-	    }
-	} else {
-	    d->again = 0;
-	    break;
-	}
-    } while (local_loop);
-    DTRACE_INVOKE_RETURN(FILE_READ_LINE);
-}
-
-static void free_read_line(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-
-    driver_free_binary(d->c.read_line.binp);
-    EF_FREE(d);
-}
-
-static void invoke_read_file(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    size_t read_size;
-    int chop;
-    DTRACE_INVOKE_SETUP(FILE_READ_FILE);
-    
-    if (! d->c.read_file.binp) { /* First invocation only */
-	int fd;
-	Sint64 size;
-	
-	if (! (d->result_ok = 
-	       efile_openfile(&d->errInfo, d->b, 
-			      EFILE_MODE_READ, &fd, &size))) {
-	    goto done;
-	}
-	d->fd = fd;
-	d->c.read_file.size = (int) size;
-	if (size < 0 || size != d->c.read_file.size ||
-	    ! (d->c.read_file.binp = 
-	       driver_alloc_binary(d->c.read_file.size))) {
-	    d->result_ok = 0;
-	    d->errInfo.posix_errno = ENOMEM;
-	    goto close;
-	}
-	d->c.read_file.offset = 0;
-    }
-    /* Invariant: d->c.read_file.size >= d->c.read_file.offset */
-    
-    read_size = (size_t) (d->c.read_file.size - d->c.read_file.offset);
-    if (! read_size) goto close;
-    chop = d->again && read_size >= FILE_SEGMENT_READ*2;
-    if (chop) read_size = FILE_SEGMENT_READ;
-    d->result_ok = 
-	efile_read(&d->errInfo, 
-		   EFILE_MODE_READ, 
-		   (int) d->fd, 
-		   d->c.read_file.binp->orig_bytes + d->c.read_file.offset,
-		   read_size, 
-		   &read_size);
-    if (d->result_ok) {
-	d->c.read_file.offset += read_size;
-	if (chop) goto chop_done; /* again */
-    }
- close:
-    efile_closefile((int) d->fd);
- done:
-    d->again = 0;
- chop_done:
-    DTRACE_INVOKE_RETURN(FILE_READ_FILE);
-}
-
-static void free_read_file(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-
-    if (d->c.read_file.binp) driver_free_binary(d->c.read_file.binp);
-    EF_FREE(d);
-}
-
-
-
-static void invoke_preadv(void *data)
-{
-    struct t_data   *d = (struct t_data *) data;
-    struct t_preadv *c = &d->c.preadv;
-    ErlIOVec        *ev = &c->eiov;
-    size_t           bytes_read_so_far = 0;
-    unsigned char   *p = (unsigned char *)ev->iov[0].iov_base + 4+4+8*c->cnt;
-    DTRACE_INVOKE_SETUP(FILE_PREADV);
-
-    while (c->cnt < c->n) {
-	size_t read_size = ev->iov[1 + c->cnt].iov_len - c->size;
-	size_t bytes_read = 0;
-	int chop = d->again 
-	    && bytes_read_so_far + read_size >= 2*FILE_SEGMENT_READ;
-	if (chop) {
-	    ASSERT(bytes_read_so_far < FILE_SEGMENT_READ);
-	    read_size = FILE_SEGMENT_READ + FILE_SEGMENT_READ/2
-		- bytes_read_so_far;
-	}
-	if ( (d->result_ok 
-	      = efile_pread(&d->errInfo, 
-			    (int) d->fd,
-			    c->offsets[c->cnt] + c->size,
-			    ((char *)ev->iov[1 + c->cnt].iov_base) + c->size,
-			    read_size,
-			    &bytes_read))) {
-	    bytes_read_so_far += bytes_read;
-	    if (chop && bytes_read == read_size) {
-		c->size += bytes_read;
-		goto done;
-	    }
-	    ASSERT(bytes_read <= read_size);
-	    ev->iov[1 + c->cnt].iov_len = bytes_read + c->size;
-	    ev->size += bytes_read + c->size;
-	    put_int64(bytes_read + c->size, p); p += 8;
-	    c->size = 0;
-	    c->cnt++;
-	    if (d->again 
-		&& bytes_read_so_far >= FILE_SEGMENT_READ
-		&& c->cnt < c->n) {
-		goto done;
-	    }
-	} else {
-	    /* In case of a read error, ev->size will not be correct,
-	     * which does not matter since no read data is returned
-	     * to Erlang.
-	     */
-	    break;
-	}
-    }					
-    d->again = 0;
- done:
-    DTRACE_INVOKE_RETURN(FILE_PREADV);
-}
-
-static void free_preadv(void *data) {
-    struct t_data *d = data;
-    int            i;
-    ErlIOVec      *ev = &d->c.preadv.eiov;
-    
-    for(i = 0; i < ev->vsize; i++) {
-	driver_free_binary(ev->binv[i]);
-    }
-    EF_FREE(d);
-}
-
-static void invoke_ipread(void *data)
-{
-    struct t_data   *d = data;
-    struct t_preadv *c = &d->c.preadv;
-    ErlIOVec        *ev = &c->eiov;
-    size_t bytes_read = 0;
-    char buf[2*sizeof(Uint32)];
-    Uint32 offset, size;
-    DTRACE_INVOKE_SETUP(FILE_IPREAD);
-    
-    /* Read indirection header */
-    if (! efile_pread(&d->errInfo, (int) d->fd, c->offsets[0], 
-		      buf, sizeof(buf), &bytes_read)) {
-	goto error;
-    }
-    if (bytes_read != sizeof(buf)) goto done; /* eof */
-    size = get_int32(buf);
-    offset = get_int32(buf+4);
-    if (size > c->size) goto done; /* eof */
-    c->n = 1;
-    c->cnt = 0;
-    c->size = 0;
-    c->offsets[0] = offset;
-    if (! (ev->binv[0] = driver_alloc_binary(3*8))) {
-	d->errInfo.posix_errno = ENOMEM;
-	goto error;
-    }
-    ev->vsize = 1;
-    ev->iov[0].iov_len = 3*8;
-    ev->iov[0].iov_base = ev->binv[0]->orig_bytes;
-    ev->size = ev->iov[0].iov_len;
-    put_int64(offset, ev->iov[0].iov_base);
-    put_int64(size, ((char *)ev->iov[0].iov_base) + 2*8);
-    if (size == 0) {
-	put_int64(size, ((char *)ev->iov[0].iov_base) + 8);
-	goto done;
-    }
-    if (! (ev->binv[1] = driver_alloc_binary(size))) {
-	d->errInfo.posix_errno = ENOMEM;
-	goto error;
-    }
-    ev->vsize = 2;
-    ev->iov[1].iov_len = size;
-    ev->iov[1].iov_base = ev->binv[1]->orig_bytes;
-    /* Read data block */
-    d->invoke = invoke_preadv;
-    invoke_preadv(data);
-    DTRACE_INVOKE_RETURN(FILE_IPREAD);
-    return;
- error:
-    d->result_ok = 0;
-    d->again = 0;
-    DTRACE_INVOKE_RETURN(FILE_IPREAD);
-    return;
- done:
-    d->result_ok = !0;
-    d->again = 0;
-    DTRACE_INVOKE_RETURN(FILE_IPREAD);
-}
-
-/* invoke_writev and invoke_pwritev are the only thread functions that
- * access non-thread data i.e the port queue and a mutex in the port
- * structure that is used to lock the port queue.
- *
- * The port will normally not be terminated until the port queue is
- * empty, but if the port is killed, i.e., exit(Port, kill) is called,
- * it will terminate regardless of the port queue state. When the
- * port is invalid driver_peekq() returns NULL and set the size to -1,
- * and driver_sizeq() returns -1.
- */
-
-static void invoke_writev(void *data) {
-    struct t_data *d = (struct t_data *) data;
-    SysIOVec      *iov0;
-    SysIOVec      *iov;
-    int            iovlen;
-    int            iovcnt;
-    size_t         size;
-    size_t         p;
-    int            segment;
-    DTRACE_INVOKE_SETUP(FILE_WRITE);
-
-    segment = d->again && d->c.writev.size >= 2*FILE_SEGMENT_WRITE;
-    if (segment) {
-	size = FILE_SEGMENT_WRITE;
-    } else {
-	size = d->c.writev.size;
-    }
-
-    /* Copy the io vector to avoid locking the port que while writing,
-     * also, both we and efile_writev might/will change the SysIOVec
-     * when segmenting or due to partial write and we do not want to
-     * tamper with the actual queue that we get from driver_peekq
-     */
-    MUTEX_LOCK(d->c.writev.q_mtx); /* Lock before accessing the port queue */
-    iov0 = driver_peekq(d->c.writev.port, &iovlen);
-
-    /* Calculate iovcnt */
-    for (p = 0, iovcnt = 0;
-	 p < size && iovcnt < iovlen;
-	 p += iov0[iovcnt++].iov_len)
-	;
-    iov = EF_SAFE_ALLOC(sizeof(SysIOVec)*iovcnt);
-    memcpy(iov,iov0,iovcnt*sizeof(SysIOVec));
-    MUTEX_UNLOCK(d->c.writev.q_mtx);
-    /* Let go of lock until we deque from original vector */
-
-    if (iovlen > 0) {
-	ASSERT(iov[iovcnt-1].iov_len > p - size);
-	iov[iovcnt-1].iov_len -= p - size;
-	if (d->flags & EFILE_COMPRESSED) {
-	    int i, status = 1;
-	    for (i = 0; i < iovcnt; i++) {
-		if (iov[i].iov_base && iov[i].iov_len > 0) {
-		    /* Just in case, I do not know what gzwrite does
-		     * with errno.
-		     */
-		    errno = EINVAL; 
-		    status = erts_gzwrite((ErtsGzFile)d->fd,
-					  iov[i].iov_base,
-					  iov[i].iov_len) == iov[i].iov_len;
-		    if (! status) {
-			d->errInfo.posix_errno =
-			    d->errInfo.os_errno = errno; /* XXX Correct? */
-			break;
-		    }
-		}
-	    }
-	    d->result_ok = status;
-	} else {
-	    d->result_ok = efile_writev(&d->errInfo, 
-					d->flags, (int) d->fd,
-					iov, iovcnt);
-	}
-    } else if (iovlen == 0) {
-	d->result_ok = 1;
-    }
-    else { /* Port has terminated */
-	d->result_ok = 0;
-	d->errInfo.posix_errno = d->errInfo.os_errno = EINVAL;
-    }
-    EF_FREE(iov);
-
-    if (! d->result_ok) {
-	d->again = 0;
-	MUTEX_LOCK(d->c.writev.q_mtx);
-	driver_deq(d->c.writev.port, d->c.writev.size);
-	MUTEX_UNLOCK(d->c.writev.q_mtx);
-    } else {
-	if (! segment) {
-	    d->again = 0;
-	}
-	d->c.writev.size -= size;
-	TRACE_F(("w%lu", (unsigned long)size));
-	MUTEX_LOCK(d->c.writev.q_mtx);
-	driver_deq(d->c.writev.port, size);
-	MUTEX_UNLOCK(d->c.writev.q_mtx);
-    }
-
-
-    DTRACE_INVOKE_RETURN(FILE_WRITE);
-}
-
-static void invoke_pwd(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    DTRACE_INVOKE_SETUP(FILE_PWD);
-
-    d->again = 0;
-    d->result_ok = efile_getdcwd(&d->errInfo,d->drive, d->b+1,
-				 RESBUFSIZE-1);
-    DTRACE_INVOKE_RETURN(FILE_PWD);
-}
-
-static void invoke_readlink(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char resbuf[RESBUFSIZE];	/* Result buffer. */
-    DTRACE_INVOKE_SETUP(FILE_READLINK);
-
-    d->again = 0;
-    d->result_ok = efile_readlink(&d->errInfo, d->b, resbuf+1,
-				  RESBUFSIZE-1);
-    if (d->result_ok != 0)
-	FILENAME_COPY((char *) d->b + 1, resbuf+1);
-    DTRACE_INVOKE_RETURN(FILE_READLINK);
-}
-
-static void invoke_altname(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char resbuf[RESBUFSIZE];	/* Result buffer. */
-    DTRACE_INVOKE_SETUP(FILE_ALTNAME);
-
-    d->again = 0;
-    d->result_ok = efile_altname(&d->errInfo, d->b, resbuf+1,
-				  RESBUFSIZE-1);
-    if (d->result_ok != 0)
-	FILENAME_COPY((char *) d->b + 1, resbuf+1);
-    DTRACE_INVOKE_RETURN(FILE_ALTNAME);
-}
-
-static void invoke_pwritev(void *data) {
-    struct t_data* const d = (struct t_data *) data;
-    struct t_pwritev * const c = &d->c.pwritev;
-    SysIOVec         *iov0;
-    SysIOVec         *iov;
-    int               iovlen;
-    int               iovcnt;
-    size_t            p;
-    int               segment;
-    size_t            size, write_size, written;
-    DTRACE_INVOKE_SETUP(FILE_PWRITEV);
-
-    segment = d->again && c->size >= 2*FILE_SEGMENT_WRITE;
-    if (segment) {
-	size = FILE_SEGMENT_WRITE;
-    } else {
-	size = c->size;
-    }
-    d->result_ok = !0;
-    p = 0;
-    /* Lock the queue just for a while, we don't want it locked during write */
-    MUTEX_LOCK(c->q_mtx);
-    iov0 = driver_peekq(c->port, &iovlen);
-    iov = EF_SAFE_ALLOC(sizeof(SysIOVec)*iovlen);
-    memcpy(iov,iov0,sizeof(SysIOVec)*iovlen);
-    MUTEX_UNLOCK(c->q_mtx);
-
-    if (iovlen < 0)
-	goto error; /* Port terminated */
-    for (iovcnt = 0, written = 0;
-	 c->cnt < c->n && iovcnt < iovlen && written < size;
-	 c->cnt++) {
-	int chop;
-	write_size = c->specs[c->cnt].size;
-	if (iov[iovcnt].iov_len - p < write_size) {
-	    goto error;
-	}
-	chop = segment && written + write_size >= 2*FILE_SEGMENT_WRITE;
-	if (chop) {
-	    ASSERT(written < FILE_SEGMENT_WRITE);
-	    write_size = FILE_SEGMENT_WRITE + FILE_SEGMENT_WRITE/2 
-		- written;
-	}
-	d->result_ok = efile_pwrite(&d->errInfo, (int) d->fd,
-				    (char *)(iov[iovcnt].iov_base) + p,
-				    write_size,
-				    c->specs[c->cnt].offset);
-	if (! d->result_ok) {
-	    d->again = 0;
-	    goto deq_error;
-	}
-	written += write_size; 
-	c->size -= write_size;
-	if (chop) { 
-	    c->specs[c->cnt].offset += write_size;
-	    c->specs[c->cnt].size -= write_size;
-	    /* Schedule out (d->again != 0) */
-	    break;
-	}
-	/* Move forward in buffer */
-	p += write_size;
-	ASSERT(iov[iovcnt].iov_len >= p);
-	if (iov[iovcnt].iov_len == p) {
-	    /* Move to next iov[], we trust that it is not a 
-	     * zero length vector, and thereby depend on that
-	     * such are not queued.
-	     */
-	    iovcnt++; p = 0;
-	}
-    }
-    if (! segment) {
-	if (c->cnt != c->n) {
-	    /* Mismatch between number of 
-	     * pos/size specs vs number of queued buffers .
-	     */
-	error:
-	    d->errInfo.posix_errno = EINVAL;
-	    d->result_ok = 0;
-	    d->again = 0;
-	deq_error:
-	    MUTEX_LOCK(c->q_mtx);
-	    driver_deq(c->port, c->size);
-	    MUTEX_UNLOCK(c->q_mtx);
-
-	    goto done;
-	} else {
-	    ASSERT(written == size);
-	    d->again = 0;
-	}
-    } else {
-      ASSERT(written >= FILE_SEGMENT_WRITE);
-    }
-      
-    MUTEX_LOCK(c->q_mtx);
-    driver_deq(c->port, written);
-    MUTEX_UNLOCK(c->q_mtx);
- done:
-    EF_FREE(iov); /* Free our copy of the vector, nothing to restore */
-    
-    DTRACE_INVOKE_RETURN(FILE_PWRITEV);
-}
-
-static void invoke_flstat(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-
-    DTRACE3(efile_drv_int_entry, d->sched_i1, d->sched_i2,
-            d->command == FILE_LSTAT ? FILE_LSTAT : FILE_FSTAT);
-    d->again = 0;
-    d->result_ok = efile_fileinfo(&d->errInfo, &d->info,
-				  d->b, d->command == FILE_LSTAT);
-    DTRACE3(efile_drv_int_entry, d->sched_i1, d->sched_i2,
-            d->command == FILE_LSTAT ? FILE_LSTAT : FILE_FSTAT);
-    gcc_optimizer_hack++;
-}
-
-static void invoke_link(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char *name = d->b;
-    char *new_name;
-    DTRACE_INVOKE_SETUP(FILE_LINK);
-
-    d->again = 0;
-    new_name = name+FILENAME_BYTELEN(name)+FILENAME_CHARSIZE;
-    d->result_ok = efile_link(&d->errInfo, name, new_name);
-    DTRACE_INVOKE_RETURN(FILE_LINK);
-}
-
-static void invoke_symlink(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char *name = d->b;
-    char *new_name;
-    DTRACE_INVOKE_SETUP(FILE_SYMLINK);
-
-    d->again = 0;
-    new_name = name+FILENAME_BYTELEN(name)+FILENAME_CHARSIZE;
-    d->result_ok = efile_symlink(&d->errInfo, name, new_name);
-    DTRACE_INVOKE_RETURN(FILE_SYMLINK);
-}
-
-static void invoke_rename(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char *name = d->b;
-    char *new_name;
-    DTRACE_INVOKE_SETUP(FILE_RENAME);
-
-    d->again = 0;
-    new_name = name+FILENAME_BYTELEN(name)+FILENAME_CHARSIZE;
-    d->result_ok = efile_rename(&d->errInfo, name, new_name);
-    DTRACE_INVOKE_RETURN(FILE_RENAME);
-}
-
-static void invoke_write_info(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    DTRACE_INVOKE_SETUP(FILE_WRITE_INFO);
-
-    d->again = 0;
-    d->result_ok = efile_write_info(&d->errInfo, &d->info, d->b);
-    DTRACE_INVOKE_RETURN(FILE_WRITE_INFO);
-}
-
-static void invoke_lseek(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int status;
-    DTRACE_INVOKE_SETUP(FILE_LSEEK);
-
-    d->again = 0;
-    if (d->flags & EFILE_COMPRESSED) {
-	int offset = (int) d->c.lseek.offset;
-	
-	if (offset != d->c.lseek.offset) {
-	    d->errInfo.posix_errno = EINVAL;
-	    status = 0;
-	} else {
-	    d->c.lseek.location = erts_gzseek((ErtsGzFile)d->fd,
-					      offset, d->c.lseek.origin);
-	    if (d->c.lseek.location == -1) {
-		d->errInfo.posix_errno = errno;
-		status = 0;
-	    } else {
-		status = 1;
-	    }
-	}
-    } else {
-	status = efile_seek(&d->errInfo, (int) d->fd, 
-			    d->c.lseek.offset, d->c.lseek.origin,
-			    &d->c.lseek.location);
-    }
-    d->result_ok = status;
-    DTRACE_INVOKE_RETURN(FILE_LSEEK);
-}
-
-static void invoke_readdir(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    char *p = NULL;
-    size_t file_bs;
-    size_t n = 0, total = 0;
-    struct t_readdir_buf *b = NULL;
-    int res = 0;
-    DTRACE_INVOKE_SETUP(FILE_READDIR);
-
-    d->again = 0;
-    d->errInfo.posix_errno = 0;
-
-    do {
-	total   = READDIR_BUFSIZE;
-	n       = 1;
-	b       = EF_SAFE_ALLOC(sizeof(struct t_readdir_buf));
-	b->next = NULL;
-	
-	if (d->c.read_dir.last_buf) {
-	    d->c.read_dir.last_buf->next = b;
-	} else {
-	    d->c.read_dir.first_buf = b;
-	}
-	d->c.read_dir.last_buf = b;
-
-	p       = &b->buf[0];
-	p[0]    = FILE_RESP_LFNAME;
-	file_bs = READDIR_BUFSIZE - n;
-
-	do {
-	    res = efile_readdir(&d->errInfo, d->b, &d->dir_handle, p + n + 2, &file_bs);
-
-	    if (res) {
-		put_int16((Uint16)file_bs, p + n);
-		n += 2 + file_bs;
-		file_bs = READDIR_BUFSIZE - n;
-	    }
-	} while( res && ((total - n - 2) >= MAXPATHLEN*FILENAME_CHARSIZE));
-
-	b->n = n;
-    } while(res);
-
-    d->result_ok = (d->errInfo.posix_errno == 0);
-    DTRACE_INVOKE_RETURN(FILE_READDIR);
-}
-
-static void invoke_open(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int status = 1;		/* Status of open call. */
-    DTRACE_INVOKE_SETUP(FILE_OPEN);
-
-    d->again = 0;
-    if ((d->flags & EFILE_COMPRESSED) == 0) {
-	int fd;
-	status = efile_openfile(&d->errInfo, d->b, d->flags, &fd, NULL);
-	d->fd = fd;
-    } else {
-	char* mode = NULL;
-
-	if (((d->flags & (EFILE_MODE_READ_WRITE)) == EFILE_MODE_READ_WRITE) ||
-	    (d->flags & EFILE_MODE_APPEND)) {
-	    status = 0;
-	    d->errInfo.posix_errno = EINVAL;
-	} else {
-	    status = efile_may_openfile(&d->errInfo, d->b);
-	    if (status || (d->errInfo.posix_errno != EISDIR)) {
-		mode = (d->flags & EFILE_MODE_READ) ? "rb" : "wb";
-		d->fd = (SWord) erts_gzopen(d->b, mode);
-		if ((ErtsGzFile)d->fd) {
-		    status = 1;
-		} else {
-		    if (errno == 0) {
-			errno = ENOMEM;
-		    }
-		    d->errInfo.posix_errno = errno;
-		    status = 0;
-		}
-	    }
-	}
-    }
-
-    d->result_ok = status;
-    if (!status) {
-        d->fd = FILE_FD_INVALID;
-    }
-    DTRACE_INVOKE_RETURN(FILE_OPEN);
-}
-
-static void invoke_fadvise(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int fd = (int) d->fd;
-    off_t offset = (off_t) d->c.fadvise.offset;
-    off_t length = (off_t) d->c.fadvise.length;
-    int advise = (int) d->c.fadvise.advise;
-    DTRACE_INVOKE_SETUP(FILE_FADVISE);
-
-    d->again = 0;
-    d->result_ok = efile_fadvise(&d->errInfo, fd, offset, length, advise);
-    DTRACE_INVOKE_RETURN(FILE_FADVISE);
-}
-
-#ifdef HAVE_SENDFILE
-static void invoke_sendfile(void *data)
-{
-    struct t_data *d = (struct t_data *)data;
-    int fd = d->fd;
-    int out_fd = (int)d->c.sendfile.out_fd;
-    Uint64 nbytes = d->c.sendfile.nbytes;
-    int result = 0;
-    d->again = 0;
-
-    result = efile_sendfile(&d->errInfo, fd, out_fd, &d->c.sendfile.offset, &nbytes, NULL);
-
-    d->c.sendfile.written += nbytes;
-
-    if (result == 1 || (result == 0 && USE_THRDS_FOR_SENDFILE(d))) {
-      d->result_ok = 0;
-    } else if (result == 0 && (d->errInfo.posix_errno == EAGAIN
-				 || d->errInfo.posix_errno == EINTR)) {
-      if ((d->c.sendfile.nbytes - nbytes) != 0) {
-	d->result_ok = 1;
-	if (d->c.sendfile.nbytes != 0)
-	  d->c.sendfile.nbytes -= nbytes;
-      } else if (nbytes == 0 && d->c.sendfile.nbytes == 0) {
-	d->result_ok = 1;
-      } else
-	d->result_ok = 0;
-    } else {
-	d->result_ok = -1;
-    }
-}
-
-static void free_sendfile(void *data) {
-    struct t_data *d = (struct t_data *)data;
-    if (USE_THRDS_FOR_SENDFILE(d)) {
-	SET_NONBLOCKING(d->c.sendfile.out_fd);
-    } else {
-	MUTEX_LOCK(d->c.sendfile.q_mtx);
-	driver_deq(d->c.sendfile.port,1);
-	MUTEX_UNLOCK(d->c.sendfile.q_mtx);
-	driver_select(d->c.sendfile.port, (ErlDrvEvent)(long)d->c.sendfile.out_fd,
-                      ERL_DRV_USE_NO_CALLBACK|ERL_DRV_WRITE, 0);
-    }
-    EF_FREE(data);
-}
-
-static void file_ready_output(ErlDrvData data, ErlDrvEvent event)
-{
-    file_descriptor* fd = (file_descriptor*) data;
-
-    switch (fd->d->command) {
-    case FILE_SENDFILE:
-	driver_select(fd->d->c.sendfile.port, event,
-		      (int)ERL_DRV_WRITE,(int) 0);
-	invoke_sendfile((void *)fd->d);
-	file_async_ready(data, (ErlDrvThreadData)fd->d);
-	break;
-    default:
-	break;
-    }
-}
-
-static void file_stop_select(ErlDrvEvent event, void* _)
-{
-
-}
-
-static int flush_sendfile(file_descriptor *desc,void *_) {
-    if (desc->sendfile_state == sending) {
-	desc->d->result_ok = -1;
-	desc->d->errInfo.posix_errno = ECONNABORTED;
-	file_async_ready((ErlDrvData)desc,(ErlDrvThreadData)desc->d);
-    }
-    return 1;
-}
-#endif /* HAVE_SENDFILE */
-
-
-static void invoke_fallocate(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    int fd = (int) d->fd;
-    Sint64 offset = d->c.fallocate.offset;
-    Sint64 length = d->c.fallocate.length;
-
-    d->again = 0;
-    d->result_ok = efile_fallocate(&d->errInfo, fd, offset, length);
-}
-
-static void free_readdir(void *data)
-{
-    struct t_data *d = (struct t_data *) data;
-    struct t_readdir_buf *b1 = d->c.read_dir.first_buf;
-
-    while (b1) {
-	struct t_readdir_buf *b2 = b1;
-	b1 = b1->next;
-	EF_FREE(b2);
-    }
-    EF_FREE(d);
-}
-
-
-
-static void try_free_read_bin(file_descriptor *desc) {
-    if ((desc->read_size == 0)
-	&& (desc->read_offset >= desc->read_binp->orig_size)) {
-	ASSERT(desc->read_offset == desc->read_binp->orig_size);
-	driver_free_binary(desc->read_binp);
-	desc->read_binp = NULL;
-	desc->read_offset = 0;
-	desc->read_size = 0;
-    }
-}
-
-
-
-static int try_again(file_descriptor *desc, struct t_data *d) {
-    if (! d->again)
-	return 0;
-    if (desc->timer_state != timer_idle) {
-	driver_cancel_timer(desc->port);
-    }
-    desc->timer_state = timer_again;
-    desc->invoke = d->invoke;
-    desc->d = d;
-    desc->free = d->free;
-    driver_set_timer(desc->port, 0L);
-    return !0;
-}
-
-
-
-static void cq_execute(file_descriptor *desc) {
-    struct t_data *d;
-    register void *void_ptr; /* Soft cast variable */
-    if (desc->timer_state == timer_again)
-	return;
-#ifdef HAVE_SENDFILE
-    if (desc->sendfile_state == sending)
-	return;
-#endif
-    if (! (d = cq_deq(desc)))
-	return;
-    TRACE_F(("x%i", (int) d->command));
-    d->again = sys_info.async_threads == 0;
-    DRIVER_ASYNC(d->level, desc, d->invoke, void_ptr=d, d->free);
-}
-
-static struct t_data *async_write(file_descriptor *desc, int *errp,
-		       int reply, Uint32 reply_size
-#ifdef USE_VM_PROBES
-		       ,Sint64 *dt_i1, Sint64 *dt_i2, Sint64 *dt_i3
-#endif
-) {
-    struct t_data *d;
-    if (! (d = EF_ALLOC(sizeof(struct t_data) - 1))) {
-	if (errp) *errp = ENOMEM;
-	return NULL;
-    }
-    TRACE_F(("w%lu", (unsigned long)desc->write_buffered));
-    d->command = FILE_WRITE;
-    d->fd = desc->fd;
-    d->flags = desc->flags;
-    d->c.writev.port = desc->port;
-    d->c.writev.q_mtx = desc->q_mtx;
-    d->c.writev.size = desc->write_buffered;
-#ifdef USE_VM_PROBES
-    if (dt_i1 != NULL) {
-        *dt_i1 = d->fd;
-        *dt_i2 = d->flags;
-        *dt_i3 = d->c.writev.size;
-    }
-#endif
-    d->reply = reply;
-    d->c.writev.reply_size = reply_size;
-    d->invoke = invoke_writev;
-    d->free = free_data;
-    d->level = 1;
-    cq_enq(desc, d);
-    desc->write_buffered = 0;
-    return d;
-}
-
-static int flush_write(file_descriptor *desc, int *errp
-#ifdef USE_VM_PROBES
-                       , dt_private *dt_priv, char *dt_utag
-#endif
-) {
-    int    result = 0;
-#ifdef USE_VM_PROBES
-    Sint64 dt_i1 = 0, dt_i2 = 0, dt_i3 = 0;
-#endif
-    struct t_data *d = NULL;
-
-    MUTEX_LOCK(desc->q_mtx);
-    if (desc->write_buffered > 0) {
-	if ((d = async_write(desc, errp, 0, 0
-#ifdef USE_VM_PROBES
-			     ,&dt_i1, &dt_i2, &dt_i3
-#endif
-			     )) == NULL) {
-            result = -1;
-        }
-    }
-    MUTEX_UNLOCK(desc->q_mtx);
-#ifdef USE_VM_PROBES
-    if (d != NULL) {
-        d->sched_i1 = dt_priv->thread_num;
-        d->sched_i2 = dt_priv->tag;
-        d->sched_utag[0] = '\0';
-        if (dt_utag != NULL) {
-            if (dt_utag[0] == '\0') {
-                dt_utag = NULL;
-            } else {
-                strncpy(d->sched_utag, dt_utag, sizeof(d->sched_utag) - 1);
-                d->sched_utag[sizeof(d->sched_utag) - 1] = '\0';
-            }
-        }
-        DTRACE11(efile_drv_entry, dt_priv->thread_num, dt_priv->tag++,
-                 dt_utag, FILE_WRITE,
-                 NULL, NULL, dt_i1, dt_i2, dt_i3, 0, desc->port_str);
-    }
-#endif /* USE_VM_PROBES */
-    return result;
-}
-
-static int check_write_error(file_descriptor *desc, int *errp) {
-    if (desc->write_error) {
-	if (errp) *errp = desc->write_errInfo.posix_errno;
-	desc->write_error = 0;
-	return -1;
-    }
-    return 0;
-}
-
-static int flush_write_check_error(file_descriptor *desc, int *errp
-#ifdef USE_VM_PROBES
-                                   , dt_private *dt_priv, char *dt_utag
-#endif
-				   ) {
-    int r;
-    if ( (r = flush_write(desc, errp
-#ifdef USE_VM_PROBES
-			  , dt_priv, dt_utag
-#endif
-			  )) != 0) {
-	check_write_error(desc, NULL);
-	return r;
-    } else {
-	return check_write_error(desc, errp);
-    }
-}
-
-static struct t_data *async_lseek(file_descriptor *desc, int *errp, int reply,
-				  Sint64 offset, int origin
-#ifdef USE_VM_PROBES
-				  , Sint64 *dt_i1, Sint64 *dt_i2, Sint64 *dt_i3
-#endif
-				  ) {
-    struct t_data *d;
-    if (! (d = EF_ALLOC(sizeof(struct t_data)))) {
-	*errp = ENOMEM;
-	return NULL;
-    }
-    d->flags = desc->flags;
-    d->fd = desc->fd;
-    d->command = FILE_LSEEK;
-    d->reply = reply;
-    d->c.lseek.offset = offset;
-    d->c.lseek.origin = origin;
-#ifdef USE_VM_PROBES
-    if (dt_i1 != NULL) {
-        *dt_i1 = d->fd;
-        *dt_i2 = d->c.lseek.offset;
-        *dt_i3 = d->c.lseek.origin;
-    }
-#endif
-    d->invoke = invoke_lseek;
-    d->free = free_data;
-    d->level = 1;
-    cq_enq(desc, d);
-    return d;
-}
-
-static void flush_read(file_descriptor *desc) {
-    desc->read_offset = 0;
-    desc->read_size = 0;
-    if (desc->read_binp) {
-	driver_free_binary(desc->read_binp);
-	desc->read_binp = NULL;
-    }
-}
-
-static int lseek_flush_read(file_descriptor *desc, int *errp
-#ifdef USE_VM_PROBES
-			    ,dt_private *dt_priv, char *dt_utag
-#endif
-			    ) {
-    int r = 0;
-    size_t read_size = desc->read_size;
-#ifdef USE_VM_PROBES
-    Sint64 dt_i1 = 0, dt_i2 = 0, dt_i3 = 0;
-#endif
-    struct t_data *d;
-
-    flush_read(desc);
-    if (read_size != 0) {
-	if ((d = async_lseek(desc, errp, 0,
-                             -((ssize_t)read_size), EFILE_SEEK_CUR
-#ifdef USE_VM_PROBES
-			     , &dt_i1, &dt_i2, &dt_i3
-#endif
-			     )) == NULL) {
-            r = -1;
-        } else {
-#ifdef USE_VM_PROBES
-            d->sched_i1 = dt_priv->thread_num;
-            d->sched_i2 = dt_priv->tag;
-            d->sched_utag[0] = '\0';
-            if (dt_utag != NULL) {
-                if (dt_utag[0] == '\0') {
-                    dt_utag = NULL;
-                } else {
-                    strncpy(d->sched_utag, dt_utag, sizeof(d->sched_utag) - 1);
-                    d->sched_utag[sizeof(d->sched_utag) - 1] = '\0';
-                }
-            }
-            DTRACE11(efile_drv_entry, dt_priv->thread_num, dt_priv->tag++,
-                     dt_utag, FILE_LSEEK,
-                     NULL, NULL, dt_i1, dt_i2, dt_i3, 0, desc->port_str);
-#endif /* USE_VM_PROBES */
-        }
-    }
-    return r;
-}
-
-
-/*********************************************************************
- * Driver entry point -> stop
- * The close has to be scheduled on async thread, so that currently active
- * async operation does not suddenly have the ground disappearing under their feet...
- */
-static void 
-file_stop(ErlDrvData e)
-{
-    file_descriptor* desc = (file_descriptor*)e;
-
-    TRACE_C('p');
-
-    IF_THRDS {
-	flush_read(desc);
-	if (desc->fd != FILE_FD_INVALID) {
-	    struct t_data *d = EF_SAFE_ALLOC(sizeof(struct t_data));
-	    d->command = FILE_CLOSE_ON_PORT_EXIT;
-	    d->reply = !0;
-	    d->fd = desc->fd;
-	    d->flags = desc->flags;
-	    d->invoke = invoke_close;
-	    d->free = free_data;
-	    d->level = 2;
-	    d->data_to_free = (void *) desc;
-	    cq_enq(desc, d);
-	    desc->fd = FILE_FD_INVALID;
-	    desc->flags = 0;
-	    cq_execute(desc);
-	} else {
-	    EF_FREE(desc);
-	}
-    } else {
-	if (desc->fd != FILE_FD_INVALID) {
-	    do_close(desc->flags, desc->fd);
-	    desc->fd = FILE_FD_INVALID;
-	    desc->flags = 0;
-	}
-	if (desc->read_binp) {
-	    driver_free_binary(desc->read_binp);
-	}
-	EF_FREE(desc);
-    }
-}
-
-/*********************************************************************
- * Driver entry point -> ready_async
- */
-static void 
-file_async_ready(ErlDrvData e, ErlDrvThreadData data)
-{
-    file_descriptor *desc = (file_descriptor*)e;
-    struct t_data *d = (struct t_data *) data;
-    char header[5];		/* result code + count */
-    char resbuf[RESBUFSIZE];	/* Result buffer. */
-#ifdef  USE_VM_PROBES
-    int sched_i1 = d->sched_i1, sched_i2 = d->sched_i2, command = d->command,
-        result_ok = d->result_ok,
-        posix_errno = d->result_ok ? 0 : d->errInfo.posix_errno;
-    DTRACE_CHARBUF(sched_utag, DTRACE_EFILE_BUFSIZ+1);
-
-    sched_utag[0] = '\0';
-    if (DTRACE_ENABLED(efile_drv_return)) {
-        strncpy(sched_utag, d->sched_utag, DTRACE_EFILE_BUFSIZ);
-        sched_utag[DTRACE_EFILE_BUFSIZ] = '\0';
-    }
-#endif  /* USE_VM_PROBES */
-
-    TRACE_C('r');
-
-    if (try_again(desc, d)) {
-        /* DTRACE TODO: what kind of probe makes sense here? */
-	return;
-    }
-
-    switch (d->command)
-    {
-    case FILE_READ:
-	if (!d->result_ok) {
-	    reply_error(desc, &d->errInfo);
-	} else {
-	    size_t available_bytes = 
-		d->c.read.bin_offset + d->c.read.bin_size - desc->read_offset;
-	    if (available_bytes < d->c.read.size) {
-		d->c.read.size = available_bytes;
-	    }
-	    TRACE_C('D');
-	    reply_data(desc, d->c.read.binp, 
-		       desc->read_offset, d->c.read.size);
-	    desc->read_offset += d->c.read.size;
-	    desc->read_size = 
-		d->c.read.bin_offset + d->c.read.bin_size - desc->read_offset;
-	    try_free_read_bin(desc);
-	}
-	free_read(data);
-	break;
-      case FILE_READ_LINE:
-	  /* The read_line structure differs from the read structure.
-	     The data->read_offset and d->c.read_line.read_offset are copies, as are
-	     data->read_size and d->c.read_line.read_size
-             The read_line function does not kniow in advance how large the binary has to be,
-	     why new allocation (but not reallocation of the old binary, for obvious reasons) 
-	     may happen in the worker thread. */
-	if (!d->result_ok) {
-	    reply_error(desc, &d->errInfo);
-	} else {
-	    size_t len = d->c.read_line.nl_pos - d->c.read_line.read_offset;
-	    TRACE_C('L');
-	    reply_data(desc, d->c.read_line.binp, 
-		       d->c.read_line.read_offset, len);
-	    desc->read_offset = d->c.read_line.read_offset + d->c.read_line.nl_skip + len;
-	    desc->read_size = 
-		d->c.read_line.read_size - d->c.read_line.nl_skip - len;
-	    if (desc->read_binp != d->c.read_line.binp) { /* New binary allocated */
-		driver_free_binary(desc->read_binp);
-		desc->read_binp =  d->c.read_line.binp;
-		driver_binary_inc_refc(desc->read_binp);
-	    }
-#if !ALWAYS_READ_LINE_AHEAD
-	    ASSERT(desc->read_bufsize > 0 || desc->read_size == 0);
-	    if (desc->read_bufsize == 0) {
-		desc->read_offset = desc->read_binp->orig_size; /* triggers cleanup */
-	    }
-#endif
-	    try_free_read_bin(desc);
-	}
-	free_read_line(data);
-	break;
-      case FILE_READ_FILE:
-	if (!d->result_ok)
-	    reply_error(desc, &d->errInfo);
-	else {
-	    header[0] = FILE_RESP_ALL_DATA;
-	    TRACE_C('R');
-	    driver_output_binary(desc->port, header, 1,
-				 d->c.read_file.binp,
-				 0, d->c.read_file.offset);
-	}
-	free_read_file(data);
-	break;
-      case FILE_WRITE:
-	  if (d->reply) {
-	      if (! d->result_ok) {
-		  reply_error(desc, &d->errInfo);
-	      } else {
-		  reply_Uint(desc, d->c.writev.reply_size);
-	      }
-	  } else {
-	      if (! d->result_ok) {
-		  desc->write_error = !0;
-		  desc->write_errInfo = d->errInfo;
-	      }
-	  }
-	  free_data(data);
-	  break;
-      case FILE_LSEEK:
-	  if (d->reply) {
-	      if (d->result_ok)
-		  reply_Sint64(desc, d->c.lseek.location);
-	      else
-		  reply_error(desc, &d->errInfo);
-	  }
-	  free_data(data);
-	  break;
-      case FILE_MKDIR:
-      case FILE_RMDIR:
-      case FILE_CHDIR:
-      case FILE_DELETE:
-      case FILE_FDATASYNC:
-      case FILE_FSYNC:
-      case FILE_TRUNCATE:
-      case FILE_LINK:
-      case FILE_SYMLINK:
-      case FILE_RENAME:
-      case FILE_WRITE_INFO:
-      case FILE_FADVISE:
-      case FILE_FALLOCATE:
-	reply(desc, d->result_ok, &d->errInfo);
-	free_data(data);
-	break;
-      case FILE_ALTNAME:
-      case FILE_PWD:
-      case FILE_READLINK:
-        {
-	    int length;
-	    char *resbuf = d->b;
-
-	    if (!d->result_ok)
-		reply_error(desc, &d->errInfo);
-	    else {
-		resbuf[0] = FILE_RESP_FNAME;
-		length = 1+FILENAME_BYTELEN((char*) resbuf+1);
-		TRACE_C('R');
-		driver_output2(desc->port, resbuf, 1, resbuf+1, length-1);
-	    }
-	    free_data(data);
-	    break;
-	}
-      case FILE_OPEN:
-	if (!d->result_ok) {
-	    reply_error(desc, &d->errInfo);
-	} else {
-	    ASSERT(d->is_fd_unused);
-	    desc->fd = d->fd;
-	    desc->flags = d->flags;
-	    d->is_fd_unused = 0;
-	    reply_Uint(desc, d->fd);
-	}
-	free_data(data);
-	break;
-      case FILE_FSTAT:
-      case FILE_LSTAT:
-        {
-	    if (d->result_ok) {
-		resbuf[0] = FILE_RESP_INFO;
-
-		put_int32(d->info.size_high,         &resbuf[1 + ( 0 * 4)]);
-		put_int32(d->info.size_low,          &resbuf[1 + ( 1 * 4)]);
-		put_int32(d->info.type,              &resbuf[1 + ( 2 * 4)]);
-
-		/* Note 64 bit indexing in resbuf here */
-		put_int64(d->info.accessTime,        &resbuf[1 + ( 3 * 4)]);
-		put_int64(d->info.modifyTime,        &resbuf[1 + ( 5 * 4)]);
-		put_int64(d->info.cTime,             &resbuf[1 + ( 7 * 4)]);
-
-		put_int32(d->info.mode,              &resbuf[1 + ( 9 * 4)]);
-		put_int32(d->info.links,             &resbuf[1 + (10 * 4)]);
-		put_int32(d->info.major_device,      &resbuf[1 + (11 * 4)]);
-		put_int32(d->info.minor_device,      &resbuf[1 + (12 * 4)]);
-		put_int32(d->info.inode,             &resbuf[1 + (13 * 4)]);
-		put_int32(d->info.uid,               &resbuf[1 + (14 * 4)]);
-		put_int32(d->info.gid,               &resbuf[1 + (15 * 4)]);
-		put_int32(d->info.access,            &resbuf[1 + (16 * 4)]);
-
-#define RESULT_SIZE (1 + (17 * 4))
-		TRACE_C('R');
-		driver_output2(desc->port, resbuf, RESULT_SIZE, NULL, 0);
-#undef RESULT_SIZE
-	    } else
-		reply_error(desc, &d->errInfo);
-	}
-	free_data(data);
-	break;
-      case FILE_READDIR:
-	if (!d->result_ok) {
-	    reply_error(desc, &d->errInfo);
-	} else {
-	    struct t_readdir_buf *b1 = d->c.read_dir.first_buf;
-	    char   op = FILE_RESP_LFNAME;
-
-	    TRACE_C('R');
-	    ASSERT(b1);
-
-	    while (b1) {
-		struct t_readdir_buf *b2 = b1;
-		char *p = &b1->buf[0];
-		driver_output2(desc->port, p, 1, p + 1, b1->n - 1);
-		b1 = b1->next;
-		EF_FREE(b2);
-	    }
-	    driver_output2(desc->port, &op, 1, NULL, 0);
-    
-	    d->c.read_dir.first_buf = NULL;
-	    d->c.read_dir.last_buf = NULL;
-	}
-	free_readdir(data);
-	break;
-      case FILE_CLOSE:
-	  if (d->reply) {
-	      TRACE_C('K');
-	      reply_ok(desc);
-#ifdef USE_VM_PROBES
-              result_ok = 1;
-#endif
-	  }
-	  free_data(data);
-	  break;
-      case FILE_PWRITEV:
-	  if (!d->result_ok) {
-	      reply_Uint_error(desc, d->c.pwritev.cnt, &d->errInfo);
-	  } else {
-	      reply_Uint(desc, d->c.pwritev.n);
-	  }
-	  free_data(data);
-	  break;
-      case FILE_PREADV:
-	  if (!d->result_ok) {
-	      reply_error(desc, &d->errInfo);
-	  } else {
-	      reply_ev(desc, FILE_RESP_LDATA, &d->c.preadv.eiov);
-	  }
-	  free_preadv(data);
-	  break;
-      case FILE_IPREAD:
-	  if (!d->result_ok) {
-	      reply_error(desc, &d->errInfo);
-	  } else if (!d->c.preadv.eiov.vsize) {
-	      reply_eof(desc);
-	  } else {
-	      reply_ev(desc, FILE_RESP_N2DATA, &d->c.preadv.eiov);
-	  }
-	  free_preadv(data);
-	  break;
-#ifdef HAVE_SENDFILE
-      case FILE_SENDFILE:
-	  if (d->result_ok == -1) {
-	      if (d->errInfo.posix_errno == ECONNRESET ||
-		  d->errInfo.posix_errno == ENOTCONN ||
-		  d->errInfo.posix_errno == EPIPE)
-		  reply_string_error(desc,"closed");
-	      else
-		  reply_error(desc, &d->errInfo);
-	      desc->sendfile_state = not_sending;
-	      free_sendfile(data);
-	  } else if (d->result_ok == 0) {
-	      reply_Sint64(desc, d->c.sendfile.written);
-	      desc->sendfile_state = not_sending;
-	      free_sendfile(data);
-	  } else if (d->result_ok == 1) { /* If we are using select to send the rest of the data */
-	      desc->sendfile_state = sending;
-	      desc->d = d;
-	      driver_select(desc->port, (ErlDrvEvent)(long)d->c.sendfile.out_fd,
-			    ERL_DRV_USE|ERL_DRV_WRITE, 1);
-	  }
-	  break;
-#endif
-      case FILE_CLOSE_ON_PORT_EXIT:
-	  /* See file_stop. However this is never invoked after the port is killed. */
-	  free_data(data);
-	  desc = NULL;
-	  /* This is it for this port, so just send dtrace and return, avoid doing anything to the freed data */
-	  DTRACE6(efile_drv_return, sched_i1, sched_i2, sched_utag,
-		  command, result_ok, posix_errno);
-	  return;
-      default:
-	abort();
-    }
-    DTRACE6(efile_drv_return, sched_i1, sched_i2, sched_utag,
-            command, result_ok, posix_errno);
-    if (desc->write_buffered != 0 && desc->timer_state == timer_idle ) {
-	desc->timer_state = timer_write;
-	driver_set_timer(desc->port, desc->write_delay);
-    }
-    cq_execute(desc);
-
-}
-
-
-/*********************************************************************
- * Driver entry point -> output
- */
-static void 
-file_output(ErlDrvData e, char* buf, ErlDrvSizeT count)
-{
-    file_descriptor* desc = (file_descriptor*)e;
-    Efile_error errInfo;	/* The error codes for the last operation. */
-    Sint fd;			/* The file descriptor for this port, if any,
-				 * -1 if none.
-				 */
-    char* name;			/* Points to the filename in buf. */
-    int command;
-    struct t_data *d = NULL;
-#ifdef  USE_VM_PROBES
-    char *dt_utag = NULL;
-    char *dt_s1 = NULL, *dt_s2 = NULL;
-    Sint64 dt_i1 = 0;
-    Sint64 dt_i2 = 0;
-    Sint64 dt_i3 = 0;
-    Sint64 dt_i4 = 0;
-    dt_private *dt_priv = get_dt_private(0);
-#endif  /* USE_VM_PROBES */
-
-    TRACE_C('o');
-
-    fd  = desc->fd;
-    name = buf+1;
-    command = *(uchar*)buf++;
-
-    switch(command) {
-
-    case FILE_MKDIR:
-    {
-	d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE);
-	
-	FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	dt_s1 = d->b;
-	dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	d->command = command;
-	d->invoke = invoke_mkdir;
-	d->free = free_data;
-	d->level = 2;
-	goto done;
-    }
-    case FILE_RMDIR:
-    {
-	d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE);
-	
-	FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	dt_s1 = d->b;
-	dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	d->command = command;
-	d->invoke = invoke_rmdir;
-	d->free = free_data;
-	d->level = 2;
-	goto done;
-    }
-    case FILE_DELETE:
-    {
-	d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE);
-	
-	FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	dt_s1 = d->b;
-	dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	d->command = command;
-	d->invoke = invoke_delete_file;
-	d->free = free_data;
-	d->level = 2;
-	goto done;
-    }
-    case FILE_RENAME:
-	{
-	    char* new_name;
-	    int namelen = FILENAME_BYTELEN(name)+FILENAME_CHARSIZE;
-	    new_name = name+namelen;
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1
-			      + namelen
-			      + FILENAME_BYTELEN(new_name) + FILENAME_CHARSIZE);
-	
-	    FILENAME_COPY(d->b, name);
-	    FILENAME_COPY(d->b + namelen, new_name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_s2 = d->b + namelen;
-	    dt_utag = buf + namelen + FILENAME_BYTELEN(new_name) + FILENAME_CHARSIZE;
-#endif
-	    d->flags = desc->flags;
-	    d->fd = fd;
-	    d->command = command;
-	    d->invoke = invoke_rename;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-    case FILE_CHDIR:
-    {
-	d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE);
-	
-	FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	dt_s1 = d->b;
-	dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	d->command = command;
-	d->invoke = invoke_chdir;
-	d->free = free_data;
-	d->level = 2;
-	goto done;
-    }
-    case FILE_PWD:
-        {
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + RESBUFSIZE + 1);
-	
-	    d->drive = *(uchar*)buf;
-#ifdef USE_VM_PROBES
-	    dt_utag = buf + 1;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_pwd;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_READDIR: 
-#ifdef USE_THREADS
-	if (sys_info.async_threads > 0)
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + 
-			      FILENAME_CHARSIZE);
-	
-	    FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	    d->dir_handle = NULL;
-	    d->command = command;
-	    d->invoke = invoke_readdir;
-	    d->free = free_readdir;
-	    d->level = 2;
-	    d->c.read_dir.first_buf = NULL;
-	    d->c.read_dir.last_buf = NULL;
-	    goto done;
-	}
-	else   
-#endif
-	{
-	    size_t resbufsize;
-	    size_t n = 0, total = 0;
-	    int res = 0;
-	    char resbuf[READDIR_BUFSIZE];
-
-	    EFILE_DIR_HANDLE dir_handle; /* Handle to open directory. */
-
-	    total               = READDIR_BUFSIZE;
-	    errInfo.posix_errno = 0;
-	    dir_handle          = NULL;
-	    resbuf[0]           = FILE_RESP_LFNAME;
-
-#ifdef USE_VM_PROBES
-	    dt_s1 = name;
-	    dt_utag = name + FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-#endif
-	    /* Fill the buffer with multiple directory listings before sending it to the
-	     * receiving process. READDIR_CHUNKS is minimum number of files sent to the
-	     * receiver.
-	     * Format for each driver_output2:
-	     * ------------------------------------
-	     * | Type   | Len     | Filename  | ...
-	     * | 1 byte | 2 bytes | Len bytes | ...
-	     * ------------------------------------
-	     */
-
-	    do {
-		n = 1;
-		resbufsize = READDIR_BUFSIZE - n;
-
-		do {
-		    res = efile_readdir(&errInfo, name, &dir_handle, resbuf + n + 2, &resbufsize);
-
-		    if (res) {
-			put_int16((Uint16)resbufsize, resbuf + n);
-			n += 2 + resbufsize;
-			resbufsize = READDIR_BUFSIZE - n;
-		    }
-		} while( res && ((total - n - 2) >= MAXPATHLEN*FILENAME_CHARSIZE));
-
-		if (n > 1) {
-		    driver_output2(desc->port, resbuf, 1, resbuf + 1, n - 1);
-		}
-	    } while(res);
-
-	    if (errInfo.posix_errno != 0) {
-		reply_error(desc, &errInfo);
-		return;
-	    }
-#ifdef USE_VM_PROBES
-	    if (dt_utag != NULL && dt_utag[0] == '\0') {
-                dt_utag = NULL;
-            } 
-
-	    DTRACE11(efile_drv_entry, dt_priv->thread_num, dt_priv->tag,
-		     dt_utag, command, name, dt_s2,
-		     dt_i1, dt_i2, dt_i3, dt_i4, desc->port_str);
-	    DTRACE6(efile_drv_return, dt_priv->thread_num, dt_priv->tag++, 
-		    dt_utag, command, 1, 0);
-#endif
-	    TRACE_C('R');
-	    driver_output2(desc->port, resbuf, 1, NULL, 0);
-	    return;
-	}
-    case FILE_OPEN:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(buf+4) + 
-			      FILENAME_CHARSIZE);
-	
-	    d->flags = get_int32((uchar*)buf);
-	    name = buf+4;
-	    FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	    dt_i1 = d->flags;
-	    dt_s1 = d->b;
-	    dt_utag = name + FILENAME_BYTELEN(d->b) + FILENAME_CHARSIZE;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_open;
-	    d->free = free_data;
-	    d->level = 2;
-	    d->is_fd_unused = 1;
-	    goto done;
-	}
-
-    case FILE_FDATASYNC:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data));
-	    
-	    d->fd = fd;
-#ifdef USE_VM_PROBES
-	    dt_utag = name;
-	    dt_i1 = fd;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_fdatasync;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_FSYNC:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data));
-	    
-	    d->fd = fd;
-#ifdef USE_VM_PROBES
-	    dt_utag = name;
-	    dt_i1 = fd;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_fsync;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-
-    case FILE_FSTAT: 
-    case FILE_LSTAT:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + FILENAME_BYTELEN(name) + 
-			      FILENAME_CHARSIZE);
-	    
-	    FILENAME_COPY(d->b, name);
-	    d->fd = fd;
-#ifdef USE_VM_PROBES
-	    dt_utag = name + FILENAME_BYTELEN(d->b) + FILENAME_CHARSIZE;
-	    if (command == FILE_LSTAT) {
-		dt_s1 = d->b;
-	    } else {
-		dt_i1 = fd;
-	    }
-#endif
-	    d->command = command;
-	    d->invoke = invoke_flstat;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-	
-    case FILE_TRUNCATE:
-        {
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data));
-	    
-	    d->flags = desc->flags;
-	    d->fd = fd;
-#ifdef USE_VM_PROBES
-	    dt_utag = name;
-	    dt_i1 = fd;
-	    dt_i2 = d->flags;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_truncate;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_WRITE_INFO:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1
-			      + FILENAME_BYTELEN(buf + 9*4) + FILENAME_CHARSIZE);
-	    
-	    d->info.mode       = get_int32(buf + 0 * 4);
-	    d->info.uid        = get_int32(buf + 1 * 4);
-	    d->info.gid        = get_int32(buf + 2 * 4);
-	    d->info.accessTime = get_int64(buf + 3 * 4);
-	    d->info.modifyTime = get_int64(buf + 5 * 4);
-	    d->info.cTime      = get_int64(buf + 7 * 4);
-
-	    FILENAME_COPY(d->b, buf + 9*4);
-#ifdef USE_VM_PROBES
-	    dt_i1              = d->info.mode;
-	    dt_i2              = d->info.uid;
-	    dt_i3              = d->info.gid;
-	    dt_s1 = d->b;
-	    dt_utag = buf + 9 * 4 + FILENAME_BYTELEN(d->b) + FILENAME_CHARSIZE;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_write_info;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_READLINK:
-	{
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + 
-			      MAX(RESBUFSIZE, (FILENAME_BYTELEN(name) +  
-					       FILENAME_CHARSIZE))  + 1);
-	    FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_utag = name + FILENAME_BYTELEN(d->b) + FILENAME_CHARSIZE;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_readlink;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_ALTNAME:
-	{
-	     d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1 + 
-			       MAX(RESBUFSIZE, (FILENAME_BYTELEN(name) +  
-						FILENAME_CHARSIZE))  + 1);
-	    FILENAME_COPY(d->b, name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_utag = name + FILENAME_BYTELEN(d->b) + FILENAME_CHARSIZE;
-#endif
-	    d->command = command;
-	    d->invoke = invoke_altname;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-
-    case FILE_LINK:
-	{
-	    char* new_name;
-	    int namelen = FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-
-	    new_name = name+namelen;
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1
-			      + namelen
-			      + FILENAME_BYTELEN(new_name) + FILENAME_CHARSIZE);
-	
-	    FILENAME_COPY(d->b, name);
-	    FILENAME_COPY(d->b + namelen, new_name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_s2 = d->b + namelen;
-	    dt_utag = buf + namelen + FILENAME_BYTELEN(dt_s2) + FILENAME_CHARSIZE;
-#endif
-	    d->flags = desc->flags;
-	    d->fd = fd;
-	    d->command = command;
-	    d->invoke = invoke_link;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_SYMLINK:
-	{
-	    char* new_name;
-	    int namelen = FILENAME_BYTELEN(name) + FILENAME_CHARSIZE;
-
-	    new_name = name+namelen;
-	    d = EF_SAFE_ALLOC(sizeof(struct t_data) - 1
-			      + namelen
-			      + FILENAME_BYTELEN(new_name) + FILENAME_CHARSIZE);
-	
-	    FILENAME_COPY(d->b, name);
-	    FILENAME_COPY(d->b + namelen, new_name);
-#ifdef USE_VM_PROBES
-	    dt_s1 = d->b;
-	    dt_s2 = d->b + namelen;
-	    dt_utag = buf + namelen + FILENAME_BYTELEN(dt_s2) + FILENAME_CHARSIZE;
-#endif
-	    d->flags = desc->flags;
-	    d->fd = fd;
-	    d->command = command;
-	    d->invoke = invoke_symlink;
-	    d->free = free_data;
-	    d->level = 2;
-	    goto done;
-	}
-
-    case FILE_FADVISE:
-    {
-        d = EF_SAFE_ALLOC(sizeof(struct t_data));
-
-        d->fd = fd;
-        d->command = command;
-        d->invoke = invoke_fadvise;
-        d->free = free_data;
-        d->level = 2;
-        d->c.fadvise.offset = get_int64((uchar*) buf);
-        d->c.fadvise.length = get_int64(((uchar*) buf) + sizeof(Sint64));
-        d->c.fadvise.advise = get_int32(((uchar*) buf) + 2 * sizeof(Sint64));
-#ifdef USE_VM_PROBES
-        dt_i1 = d->fd;
-        dt_i2 = d->c.fadvise.offset;
-        dt_i3 = d->c.fadvise.length;
-        dt_i4 = d->c.fadvise.advise;
-        dt_utag = buf + 3 * sizeof(Sint64);
-#endif
-        goto done;
-    }
-
-    case FILE_FALLOCATE:
-    {
-        d = EF_SAFE_ALLOC(sizeof(struct t_data));
-
-        d->fd = fd;
-        d->command = command;
-        d->invoke = invoke_fallocate;
-        d->free = free_data;
-        d->level = 2;
-        d->c.fallocate.offset = get_int64((uchar*) buf);
-        d->c.fallocate.length = get_int64(((uchar*) buf) + sizeof(Sint64));
-        goto done;
-    }
-
-    }
-
-    /*
-     * Ignore anything else -- let the caller hang.
-     */
-     
-    return;
-
- done:
-    if (d) {
-#ifdef USE_VM_PROBES
-	d->sched_i1 = dt_priv->thread_num;
-	d->sched_i2 = dt_priv->tag;
-	d->sched_utag[0] = '\0';
-	if (dt_utag != NULL) {
-	    if (dt_utag[0] == '\0') {
-		dt_utag = NULL;
-	    } else {
-		strncpy(d->sched_utag, dt_utag, sizeof(d->sched_utag) - 1);
-		d->sched_utag[sizeof(d->sched_utag) - 1] = '\0';
-	    }
-	}
-	DTRACE11(efile_drv_entry, dt_priv->thread_num, dt_priv->tag++,
-		 dt_utag, command, dt_s1, dt_s2,
-		 dt_i1, dt_i2, dt_i3, dt_i4, desc->port_str);
-#endif
-	cq_enq(desc, d);
-    }
-}
-
-/*********************************************************************
- * Driver entry point -> flush
- */
-static void 
-file_flush(ErlDrvData e) {
-    file_descriptor *desc = (file_descriptor *)e;
-#ifdef DEBUG
-    int r;
-#endif
-#ifdef  USE_VM_PROBES
-    dt_private *dt_priv = get_dt_private(dt_driver_io_worker_base);
-#endif
-
-    TRACE_C('f');
-
-#ifdef HAVE_SENDFILE
-    flush_sendfile(desc, NULL);
-#endif
-
-#ifdef DEBUG
-    r = 
-#endif
-         flush_write(desc, NULL
-#ifdef USE_VM_PROBES
-		     , dt_priv, (desc->d == NULL) ? NULL : desc->d->sched_utag
-#endif
-		     );
-    /* Only possible reason for bad return value is ENOMEM, and 
-     * there is nobody to tell...
-     */
-#ifdef DEBUG
-    ASSERT(r == 0); 
-#endif
-    cq_execute(desc);
-}
-
-
-
-/*********************************************************************
- * Driver entry point -> control
- * Only debug functionality...
- */
-static ErlDrvSSizeT
-file_control(ErlDrvData e, unsigned int command, 
-	     char* buf, ErlDrvSizeT len, char **rbuf, ErlDrvSizeT rlen) {
-    file_descriptor *desc = (file_descriptor *)e;
-    switch (command) {
-    case 'K' :
-	if (rlen < 4) {
-	    *rbuf = EF_ALLOC(4);
-	}
-	(*rbuf)[0] = ((desc->key) >> 24) & 0xFF;
-	(*rbuf)[1] = ((desc->key) >> 16) & 0xFF;
-	(*rbuf)[2] = ((desc->key) >> 8) & 0xFF;
-	(*rbuf)[3] = (desc->key) & 0xFF;
-	return 4;
-    default:
-	return 0;
-    }
-}
-
-/*********************************************************************
- * Driver entry point -> timeout
- */
-static void 
-file_timeout(ErlDrvData e) {
-    file_descriptor *desc = (file_descriptor *)e;
-    enum e_timer timer_state = desc->timer_state;
-#ifdef  USE_VM_PROBES
-    dt_private *dt_priv = get_dt_private(dt_driver_io_worker_base);
-#endif
-
-    TRACE_C('t');
-
-    desc->timer_state = timer_idle;
-    switch (timer_state) {
-    case timer_idle:
-	ASSERT(0);
-	break;
-    case timer_again:
-	ASSERT(desc->invoke);
-	ASSERT(desc->free);
-	driver_async(desc->port, KEY(desc), desc->invoke, desc->d, desc->free);
-	break;
-    case timer_write: {
-#ifdef DEBUG
-	int r = 
-#endif
-	         flush_write(desc, NULL
-#ifdef USE_VM_PROBES
-			     , dt_priv, (desc->d == NULL) ? NULL : desc->d->sched_utag
-#endif
-			     );
-	/* Only possible reason for bad return value is ENOMEM, and 
-	 * there is nobody to tell...
-	 */
-	ASSERT(r == 0); 
-	cq_execute(desc);
-    } break;
-    } /* case */
-}
-
-
-
-/*********************************************************************
- * Driver entry point -> outputv
- */
-static void 
-file_outputv(ErlDrvData e, ErlIOVec *ev) {
-    file_descriptor* desc = (file_descriptor*)e;
-    char command;
-    size_t p, q;
-    int err;
-    struct t_data *d = NULL;
-#ifdef USE_VM_PROBES
-    Sint64 dt_i1 = 0, dt_i2 = 0, dt_i3 = 0;
-    Sint64 dt_i4 = 0;
-    char *dt_utag = NULL;
-    char *dt_s1 = NULL;
-    dt_private *dt_priv = get_dt_private(dt_driver_io_worker_base);
-#endif
-
-    TRACE_C('v');
-
-    p = 0; q = 1;
-    if (! EV_GET_CHAR(ev, &command, &p, &q)) {
-	/* Empty command */
-	reply_posix_error(desc, EINVAL);
-	goto done;
-    }
-    /* 'command' contains the decoded command number,
-     * 'p' and 'q' point out the next byte in the command:
-     * ((char *)ev->iov[q].iov_base) + p;
-     */
-    
-    TRACE_F(("%i", (int) command));
-
-    switch (command) {
-
-    case FILE_CLOSE: {
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	flush_read(desc);
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (desc->fd != FILE_FD_INVALID) {
-	    if (! (d = EF_ALLOC(sizeof(struct t_data)))) {
-		reply_posix_error(desc, ENOMEM);
-	    } else {
-		d->command = command;
-		d->reply = !0;
-		d->fd = desc->fd;
-		d->flags = desc->flags;
-#ifdef USE_VM_PROBES
-		dt_i1 = d->fd;
-		dt_i2 = d->flags;
-#endif
-		d->invoke = invoke_close;
-		d->free = free_data;
-		d->level = 2;
-		cq_enq(desc, d);
-		desc->fd = FILE_FD_INVALID;
-		desc->flags = 0;
-	    }
-	} else {
-	    reply_posix_error(desc, EBADF);
-	}
-    } goto done;
-
-    case FILE_READ: {
-	Uint32 sizeH, sizeL;
-	size_t size, alloc_size;
-
-	if (!EV_GET_UINT32(ev, &sizeH, &p, &q)
-	    || !EV_GET_UINT32(ev, &sizeL, &p, &q)) {
-	    /* Wrong buffer length to contain the read count */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-#if ALWAYS_READ_LINE_AHEAD
-	if (desc->read_bufsize == 0 && desc->read_binp != NULL && desc->read_size > 0) {
-	    /* We have allocated a buffer for line mode but should not really have a 
-	       read-ahead buffer... */
-	    if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-				 , dt_priv, dt_utag
-#endif
-				 ) < 0) {
-		reply_posix_error(desc, err);
-		goto done;
-	    }
-	}
-#endif
-#if SIZEOF_SIZE_T == 4
-	if (sizeH != 0) {
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	size = sizeL;
-#else
-	size = ((size_t)sizeH << 32) | sizeL;
-#endif
-	if ((desc->fd == FILE_FD_INVALID)
-	    || (! (desc->flags & EFILE_MODE_READ)) ) {
-	    reply_posix_error(desc, EBADF);
-	    goto done;
-	}
-	if (size == 0) {
-	    reply_buf(desc, &command, 0);
-	    goto done;
-	}
-	if (desc->read_size >= size) {
-	    /* We already have all data */
-	    TRACE_C('D');
-	    reply_data(desc, desc->read_binp, desc->read_offset, size);
-	    desc->read_offset += size;
-	    desc->read_size -= size;
-	    try_free_read_bin(desc);
-	    goto done;
-	}
-	/* We may have some of the data 
-	 */
-	/* Justification for the following strange formula:
-	 * If the read request is for such a large block as more than 
-	 * half the buffer size it may lead to a lot of unnecessary copying, 
-	 * since the tail of the old buffer is copied to the head of the
-	 * new, and if the tail is almost half the buffer it is a lot
-	 * to copy. Therefore allocate the exact amount needed in 
-	 * this case, giving no lingering tail. */
-	alloc_size = 
-	    size > (desc->read_bufsize>>1) ? 
-	    size : desc->read_bufsize;
-	if (! desc->read_binp) {
-	    /* Need to allocate a new binary for the result */
-	    if (! (desc->read_binp = driver_alloc_binary(alloc_size))) {
-		reply_posix_error(desc, ENOMEM);
-		goto done;
-	    }
-	} else {
-	    /* We already have a buffer */
-	    if (desc->read_binp->orig_size - desc->read_offset < size) {
-		/* Need to allocate a new binary for the result */
-		ErlDrvBinary *binp;
-		if (! (binp = driver_alloc_binary(alloc_size))) {
-		    reply_posix_error(desc, ENOMEM);
-		    goto done;
-		}
-		/* Move data we already have to the new binary */
-		sys_memcpy(binp->orig_bytes, 
-			   desc->read_binp->orig_bytes + desc->read_offset,
-			   desc->read_size);
-		driver_free_binary(desc->read_binp);
-		desc->read_offset = 0;
-		desc->read_binp = binp;
-	    }
-	} 
-	if (! (d = EF_ALLOC(sizeof(struct t_data)))) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	d->command = command;
-	d->reply = !0;
-	d->fd = desc->fd;
-	d->flags = desc->flags;
-	d->c.read.binp = desc->read_binp;
-	d->c.read.bin_offset = desc->read_offset + desc->read_size;
-	d->c.read.bin_size = desc->read_binp->orig_size - d->c.read.bin_offset;
-	d->c.read.size = size;
-#ifdef USE_VM_PROBES
-	dt_i1 = d->fd;
-	dt_i2 = d->flags;
-	dt_i3 = d->c.read.size;
-#endif
-	driver_binary_inc_refc(d->c.read.binp);
-	d->invoke = invoke_read;
-	d->free = free_read;
-	d->level = 1;
-	cq_enq(desc, d);
-    } goto done; /* case FILE_READ: */
-
-    case FILE_READ_LINE: {
-	/*
-	 * Icky little creature... We do mostly as ordinary file read, but with a few differences.
-	 * 1) We have to scan for proper newline sequence if there is a buffer already, we cannot know 
-	 *    in advance if the buffer contains a whole line without scanning.
-	 * 2) We do not know how large the buffer needs to be in advance. We give a default buffer,
-	 *    but the worker may need to allocate a new one. Freeing the old and rereferencing a newly 
-	 *    allocated binary + dealing with offsets and lengts are done in file_async ready
-	 *    for this OP.
-	 */
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (ev->size != 1
-#ifdef USE_VM_PROBES
-	    + FILENAME_BYTELEN(dt_utag) + FILENAME_CHARSIZE
-#endif
-	    ) {
-	    /* Wrong command length */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	if ((desc->fd == FILE_FD_INVALID)
-	    || (! (desc->flags & EFILE_MODE_READ)) ) {
-	    reply_posix_error(desc, EBADF);
-	    goto done;
-	}
-	if (desc->read_size > 0) {
-	    /* look for '\n' in what we'we already got */
-	    void *nl_ptr = memchr(desc->read_binp->orig_bytes + desc->read_offset,'\n',desc->read_size);
-	    if (nl_ptr != NULL) {
-		/* If found, we're done */
-		int skip = 0;
-		size_t size = ((char *) nl_ptr) - 
-		    ((char *) (desc->read_binp->orig_bytes + desc->read_offset)) + 1;
-		if (size > 1 &&
-		    *(((char *) nl_ptr) - 1) == '\r') {
-		    *(((char *) nl_ptr) - 1) = '\n';		    
-		    skip = 1;
-		    --size;
-		} 
-		reply_data(desc, desc->read_binp, desc->read_offset, size);
-		desc->read_offset += (size + skip);
-		desc->read_size -= (size + skip);
-		try_free_read_bin(desc);
-		goto done;
-	    }
-	}
-	/* Now, it's up to the thread to work out the need for more buffers and such, it's
-	   no use doing it in this thread as we do not have the information required anyway. 
-	   Even a NULL buffer could be handled by the thread, but code is simplified by us 
-	   allocating it */
-	if (! desc->read_binp) {
-	    int alloc_size = (desc->read_bufsize > DEFAULT_LINEBUF_SIZE) ? desc->read_bufsize : 
-		DEFAULT_LINEBUF_SIZE;
-	    /* Allocate a new binary for the result */
-	    if (! (desc->read_binp = driver_alloc_binary(alloc_size))) {
-		reply_posix_error(desc, ENOMEM);
-		goto done;
-	    }
-	}	
-	if (! (d = EF_ALLOC(sizeof(struct t_data)))) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-
-	d->command = command;
-	d->reply = !0;
-	d->fd = desc->fd;
-	d->flags = desc->flags;
-	d->c.read_line.binp = desc->read_binp;
-	d->c.read_line.read_offset = desc->read_offset;
-	d->c.read_line.read_size = desc->read_size;
-#ifdef USE_VM_PROBES
-	dt_i1 = d->fd;
-	dt_i2 = d->flags;
-	dt_i3 = d->c.read_line.read_offset;
-#endif
-#if !ALWAYS_READ_LINE_AHEAD
-	d->c.read_line.read_ahead = (desc->read_bufsize > 0);
-#ifdef USE_VM_PROBES
-	dt_i4 = d->c.read_line.read_ahead;
-#endif
-#endif 
-	driver_binary_inc_refc(d->c.read.binp);
-	d->invoke = invoke_read_line;
-	d->free = free_read_line;
-	d->level = 1;
-	cq_enq(desc, d);
-    } goto done;
-    case FILE_WRITE: { /* Dtrace: The dtrace user tag is not last in message, 
-			  but follows the message tag directly. 
-			  This is handled specially in prim_file.erl */
-	ErlDrvSizeT skip = 1;
-	ErlDrvSizeT size = ev->size - skip;
-
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-	skip += FILENAME_BYTELEN(dt_utag) + FILENAME_CHARSIZE;
-	size = ev->size - skip;
-#endif
-	if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			     , dt_priv, dt_utag
-#endif
-			     ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (! (desc->flags & EFILE_MODE_WRITE)) {
-	    reply_posix_error(desc, EBADF);
-	    goto done;
-	}
-	if (size == 0) {
-	    reply_Uint(desc, size);
-	    goto done;
-	}
-	MUTEX_LOCK(desc->q_mtx);
-	if (driver_enqv(desc->port, ev, skip)) {
-	    MUTEX_UNLOCK(desc->q_mtx);
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	desc->write_buffered += size;
-	if (desc->write_buffered < desc->write_bufsize) {
-	    MUTEX_UNLOCK(desc->q_mtx);
-	    reply_Uint(desc, size);
-	    if (desc->timer_state == timer_idle) {
-		desc->timer_state = timer_write;
-		driver_set_timer(desc->port, desc->write_delay);
-	    }
-	} else {
-	    if ((d = async_write(desc, &err, !0, size
-#ifdef USE_VM_PROBES
-				 , &dt_i1, &dt_i2, &dt_i3
-#endif
-				 )) == NULL) {
-		MUTEX_UNLOCK(desc->q_mtx);
-		reply_posix_error(desc, err);
-		goto done;
-	    } else {
-		MUTEX_UNLOCK(desc->q_mtx);
-	    }
-	}
-    } goto done; /* case FILE_WRITE */
-
-    case FILE_PWRITEV: { /* Dtrace: The dtrace user tag is not last in message, 
-			   but follows the message tag directly. 
-			   This is handled specially in prim_file.erl */
-	Uint32 i, j, n; 
-	size_t total;
-#ifdef USE_VM_PROBES
-	char dt_tmp;
-	int dt_utag_bytes = 1;
-
-	dt_utag = EV_CHAR_P(ev, p, q);
-	/* This will work for UTF-8, but not for UTF-16 - extra reminder here */
-#ifdef FILENAMES_16BIT 
-#error 16bit characters in filenames and dtrace in combination is not supported.
-#endif
-	while (EV_GET_CHAR(ev, &dt_tmp, &p, &q) && dt_tmp != '\0') {
-	    dt_utag_bytes++;
-	}
-#endif
-	if (ev->size < 1+4
-#ifdef USE_VM_PROBES
-	    + dt_utag_bytes
-#endif
-	    || !EV_GET_UINT32(ev, &n, &p, &q)) {
-	    /* Buffer too short to contain even the number of pos/size specs */
-	    reply_Uint_posix_error(desc, 0, EINVAL);
-	    goto done;
-	}
-	if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			     , dt_priv, dt_utag
-#endif
-			     ) < 0) {
-	    reply_Uint_posix_error(desc, 0, err);
-	    goto done;
-	}
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_Uint_posix_error(desc, 0, err);
-	    goto done;
-	}
-	if (n == 0) {
-	    /* Trivial case - nothing to write */
-	    if (ev->size != 1+4) {
-		reply_posix_error(desc, err);
-	    } else {
-		reply_Uint(desc, 0);
-	    }
-	    goto done;
-	}
-	if (ev->size < 1+4+8*(2*n)
-#ifdef USE_VM_PROBES
-	    + dt_utag_bytes
-#endif
-	    ) {
-	    /* Buffer too short to contain even the pos/size specs */
-	    reply_Uint_posix_error(desc, 0, EINVAL);
-	    goto done;
-	}
-	d = EF_ALLOC(sizeof(struct t_data) 
-		     + (n * sizeof(struct t_pbuf_spec)));
-	if (! d) {
-	    reply_Uint_posix_error(desc, 0, ENOMEM);
-	    goto done;
-	}
-	d->command = command;
-	d->reply = !0;
-	d->fd = desc->fd;
-	d->flags = desc->flags;
-#ifdef USE_VM_PROBES
-	dt_i1 = d->fd;
-	dt_i2 = d->flags;
-#endif
-	d->c.pwritev.port = desc->port;
-	d->c.pwritev.q_mtx = desc->q_mtx;
-	d->c.pwritev.n = n;
-	d->c.pwritev.cnt = 0;
-	total = 0;
-	j = 0;
-	/* Create pos/size specs in the thread data structure
-	 * for all non-zero size binaries. Calculate total size.
-	 */
-	for(i = 0; i < n; i++) {
-	    Uint32 sizeH, sizeL;
-	    size_t size;
-	    if (   !EV_GET_SINT64(ev, &d->c.pwritev.specs[i].offset, &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeH, &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeL, &p, &q)) {
-		/* Misalignment in buffer */
-		reply_Uint_posix_error(desc, 0, EINVAL);
-		EF_FREE(d);
-		goto done;
-	    }
-#if SIZEOF_SIZE_T == 4
-	    if (sizeH != 0) {
-		reply_Uint_posix_error(desc, 0, EINVAL);
-		EF_FREE(d);
-		goto done;
-	    }
-	    size = sizeL;
-#else
-	    size = ((size_t)sizeH<<32) | sizeL;
-#endif
-	    if (size > 0) {
-		total += size;
-		d->c.pwritev.specs[j].size = size;
-		j++;
-	    }
-	}
-	d->c.pwritev.size = total;
-#ifdef USE_VM_PROBES
-	dt_i3 = d->c.pwritev.size;
-#endif
-	if (j == 0) {
-	    /* Trivial case - nothing to write */
-	    EF_FREE(d);
-	    reply_Uint(desc, 0);
-	} else {
-	    ErlDrvSizeT skip = 1 + 4 + 8 * (2*n) 
-#ifdef USE_VM_PROBES
-		+ dt_utag_bytes
-#endif
-		;
-	    if (skip + total != ev->size) {
-		/* Actual amount of data does not match 
-		 * total of all pos/size specs
-		 */
-		EF_FREE(d);
-		reply_Uint_posix_error(desc, 0, EINVAL);
-	    } else {
-		/* Enqueue the data */
-		MUTEX_LOCK(desc->q_mtx);
-		driver_enqv(desc->port, ev, skip);
-		MUTEX_UNLOCK(desc->q_mtx);
-		/* Execute the command */
-		d->invoke = invoke_pwritev;
-		d->free = free_data;
-		d->level = 1;
-		cq_enq(desc, d);
-	    }
-	}
-    } goto done; /* case FILE_PWRITEV: */
-
-    case FILE_PREADV: { /* Dtrace: The dtrace user tag is not last in message, 
-			   but follows the message tag directly. 
-			   This is handled specially in prim_file.erl */
-	register void * void_ptr;
-	Uint32 i, n;
-	ErlIOVec *res_ev;
-#ifdef USE_VM_PROBES
-	char dt_tmp;
-	int dt_utag_bytes = 1;
-	/* This will work for UTF-8, but not for UTF-16 - extra reminder here */
-#ifdef FILENAMES_16BIT 
-#error 16bit characters in filenames and dtrace in combination is not supported.
-#endif
-	dt_utag = EV_CHAR_P(ev, p, q);
-	while (EV_GET_CHAR(ev, &dt_tmp, &p, &q) && dt_tmp != '\0') {
-	    dt_utag_bytes++;
-	}
-#endif
-	if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			     , dt_priv, dt_utag
-#endif
-			     ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (ev->size < 1+8
-#ifdef USE_VM_PROBES
-	    + dt_utag_bytes
-#endif
-	    || !EV_GET_UINT32(ev, &n, &p, &q)
-	    || !EV_GET_UINT32(ev, &n, &p, &q)) {
-	    /* Buffer too short to contain even the number of pos/size specs */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	if (ev->size < 1+8+8*(2*n)
-#ifdef USE_VM_PROBES
-	    + dt_utag_bytes
-#endif
-	    ) {
-	    /* Buffer wrong length to contain the pos/size specs */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	/* Create the thread data structure with the contained ErlIOVec 
-	 * and corresponding binaries for the response 
-	 */
-	d = EF_ALLOC(sizeof(*d)
-		     + (n * sizeof(*d->c.preadv.offsets))
-		     + ((1+n) * (sizeof(*res_ev->iov)
-				 + sizeof(*res_ev->binv))));
-	if (! d) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	d->command = command;
-	d->reply = !0;
-	d->fd = desc->fd;
-	d->flags = desc->flags;
-#ifdef USE_VM_PROBES
-	dt_i1 = d->fd;
-	dt_i2 = d->flags;
-#endif
-	d->c.preadv.n = n;
-	d->c.preadv.cnt = 0;
-	d->c.preadv.size = 0;
-	res_ev = &d->c.preadv.eiov;
-	/* XXX possible alignment problems here for weird machines */
-	res_ev->vsize = 1+d->c.preadv.n;
-	res_ev->iov = void_ptr = &d->c.preadv.offsets[d->c.preadv.n];
-	res_ev->binv = void_ptr = &res_ev->iov[res_ev->vsize];
-	/* Read in the pos/size specs and allocate binaries for the results */
-	for (i = 1; i < 1+n; i++) {
-	    Uint32 sizeH, sizeL;
-	    size_t size;
-	    if (   !EV_GET_SINT64(ev, &d->c.preadv.offsets[i-1], &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeH, &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeL, &p, &q)) {
-		reply_posix_error(desc, EINVAL);
-		break;
-	    }
-#if SIZEOF_SIZE_T == 4
-	    if (sizeH != 0) {
-		reply_posix_error(desc, EINVAL);
-		break;
-	    }
-	    size = sizeL;
-#else
-	    size = ((size_t)sizeH<<32) | sizeL;
-#endif
-#ifdef USE_VM_PROBES
-	    dt_i3 += size;
-#endif
-	    if (! (res_ev->binv[i] = driver_alloc_binary(size))) {
-		reply_posix_error(desc, ENOMEM);
-		break;
-	    } else {
-		res_ev->iov[i].iov_len  = size;
-		res_ev->iov[i].iov_base = res_ev->binv[i]->orig_bytes;
-	    }
-	}
-	if (i < 1+n) {
-	    for (i--; i > 0; i--) {
-		driver_free_binary(res_ev->binv[i]);
-	    }
-	    EF_FREE(d);
-	    goto done;
-	}
-	/* Allocate the header binary (index 0) */
-	res_ev->binv[0] = driver_alloc_binary(4+4+8*n);
-	if (! res_ev->binv[0]) {
-	    reply_posix_error(desc, ENOMEM);
-	    for (i = 1; i < 1+n; i++) {
-		driver_free_binary(res_ev->binv[i]);
-	    }
-	    EF_FREE(d);
-	    goto done;
-	}
-	res_ev->iov[0].iov_len = 4+4+8*n;
-	res_ev->iov[0].iov_base = res_ev->binv[0]->orig_bytes;
-	/* Fill in the number of buffers in the header */
-	put_int32(0, res_ev->iov[0].iov_base);
-	put_int32(n, (char *)(res_ev->iov[0].iov_base) + 4);
-	/**/
-	res_ev->size = res_ev->iov[0].iov_len;
-	if (n == 0) {
-	    /* Trivial case - nothing to read */
-	    reply_ev(desc, FILE_RESP_LDATA, res_ev);
-	    free_preadv(d);
-	    goto done;
-	} else {
-	    d->invoke = invoke_preadv;
-	    d->free = free_preadv;
-	    d->level = 1;
-	    cq_enq(desc, d);
-	}
-    } goto done; /* case FILE_PREADV: */
-
-    case FILE_LSEEK: {
-	Sint64 offset;		/* Offset for seek */
-	Uint32 origin;		/* Origin of seek. */
-
-	if (ev->size < 1+8+4
-	    || !EV_GET_SINT64(ev, &offset, &p, &q)
-	    || !EV_GET_UINT32(ev, &origin, &p, &q)) {
-	    /* Wrong length of buffer to contain offset and origin */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			     , dt_priv, dt_utag
-#endif
-			     ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if ((d = async_lseek(desc, &err, !0, offset, origin
-#ifdef USE_VM_PROBES
-			     , &dt_i1, &dt_i2, &dt_i3
-#endif
-			     )) == NULL) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-    } goto done;
-
-    case FILE_READ_FILE: {
-	char *filename;
-	if (ev->size < 1+1) {
-	    /* Buffer contains empty name */
-	    reply_posix_error(desc, ENOENT);
-	    goto done;
-	}
-#ifndef USE_VM_PROBES
-	/* In the dtrace case, the iov has an extra element, the dtrace utag - we will need 
-	   another test to see that
-	   the filename is in a single buffer: */
-	if (ev->size-1 != ev->iov[q].iov_len-p) {
-	    /* Name not in one single buffer */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-#else
-	if (((byte *)ev->iov[q].iov_base)[ev->iov[q].iov_len-1] != '\0') {
-	    /* Name not in one single buffer */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}	
-#endif
-	filename = EV_CHAR_P(ev, p, q);
-	d = EF_ALLOC(sizeof(struct t_data) -1 + FILENAME_BYTELEN(filename) + FILENAME_CHARSIZE);
-	if (! d) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	d->command = command;
-	d->reply = !0;
-	/* Copy name */
-	FILENAME_COPY(d->b, filename);
-#ifdef USE_VM_PROBES
-	{
-	    char dt_tmp;
-
-	    /* This will work for UTF-8, but not for UTF-16 - extra reminder here */
-#ifdef FILENAMES_16BIT 
-#error 16bit characters in filenames and dtrace in combination is not supported.
-#endif
-	    while (EV_GET_CHAR(ev, &dt_tmp, &p, &q) && dt_tmp != '\0') 
-		;
-	    dt_s1 = d->b;
-	    dt_utag = EV_CHAR_P(ev, p, q);
-	}
-#endif
-	d->c.read_file.binp = NULL;
-	d->invoke = invoke_read_file;
-	d->free = free_read_file;
-	d->level = 2;
-	cq_enq(desc, d);
-    } goto done;
-
-    case FILE_IPREAD: {
-	/* This operation cheets by using invoke_preadv() and free_preadv()
-	 * plus its own invoke_ipread. Therefore the result format is 
-	 * a bit awkward - the header binary contains one extra 64 bit
-	 * field that invoke_preadv() fortunately ignores,
-	 * and the first 64 bit field does not contain the number of 
-	 * data binaries which invoke_preadv() also ignores.
-	 */
-	register void * void_ptr;
-	char mode;
-	Sint64 hdr_offset;
-	Uint32 max_size;
-	ErlIOVec *res_ev;
-	int vsize;
-	if (! EV_GET_CHAR(ev, &mode, &p, &q)) {
-	    /* Empty command */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	if (mode != IPREAD_S32BU_P32BU) {
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	if (ev->size < 1+1+8+4
-	    || !EV_GET_SINT64(ev, &hdr_offset, &p, &q)
-	    || !EV_GET_UINT32(ev, &max_size, &p, &q)) {
-	    /* Buffer too short to contain 
-	     * the header offset and max size spec */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-#ifdef USE_VM_PROBES
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			     , dt_priv, dt_utag
-#endif
-			     ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				    , dt_priv, dt_utag
-#endif
-				    ) < 0) {
-	    reply_posix_error(desc, err);
-	    goto done;
-	}
-	/* Create the thread data structure with the contained ErlIOVec 
-	 * and corresponding binaries for the response 
-	 */
-	vsize = 2;
-	d = EF_ALLOC(sizeof(*d) + 
-		     vsize*(sizeof(*res_ev->iov) + sizeof(*res_ev->binv)));
-	if (! d) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	d->command = command;
-	d->reply = !0;
-	d->fd = desc->fd;
-	d->flags = desc->flags;
-	d->c.preadv.offsets[0] = hdr_offset;
-	d->c.preadv.size = max_size;
-#ifdef USE_VM_PROBES
-	dt_i1 = d->fd;
-	dt_i2 = d->flags;
-	dt_i3 = d->c.preadv.offsets[0];
-	dt_i4 = d->c.preadv.size;
-#endif
-	res_ev = &d->c.preadv.eiov;
-	/* XXX possible alignment problems here for weird machines */
-	res_ev->iov = void_ptr = d + 1;
-	res_ev->binv = void_ptr = res_ev->iov + vsize;
-	res_ev->size = 0;
-	res_ev->vsize = 0;
-	d->invoke = invoke_ipread;
-	d->free = free_preadv;
-	d->level = 1;
-	cq_enq(desc, d);
-    } goto done; /* case FILE_IPREAD: */
-
-    case FILE_SETOPT: {
-	char opt;
-
-	if (ev->size < 1+1
-	    || !EV_GET_CHAR(ev, &opt, &p, &q)) {
-	    /* Buffer too short to contain even the option type */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-#ifdef USE_VM_PROBES
-	dt_i1 = opt;
-	dt_utag = EV_CHAR_P(ev, p, q);
-#endif
-	switch (opt) {
-	case FILE_OPT_DELAYED_WRITE: {
-	    Uint32 sizeH, sizeL, delayH, delayL;
-	    if (ev->size != 1+1+4*sizeof(Uint32)
-#ifdef USE_VM_PROBES
-		+ FILENAME_BYTELEN(dt_utag) + FILENAME_CHARSIZE
-#endif
-		|| !EV_GET_UINT32(ev, &sizeH, &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeL, &p, &q)
-		|| !EV_GET_UINT32(ev, &delayH, &p, &q)
-		|| !EV_GET_UINT32(ev, &delayL, &p, &q)) {
-		/* Buffer has wrong length to contain the option values */
-		reply_posix_error(desc, EINVAL);
-		goto done;
-	    }
-#if SIZEOF_SIZE_T == 4
-	    if (sizeH != 0) {
-		reply_posix_error(desc, EINVAL);
-		goto done;
-	    }
-	    desc->write_bufsize = sizeL;
-#else
-	    desc->write_bufsize = ((size_t)sizeH << 32) | sizeL;
-#endif
-#if SIZEOF_LONG == 4
-	    if (delayH != 0) {
-		reply_posix_error(desc, EINVAL);
-		goto done;
-	    }
-	    desc->write_delay = delayL;
-#else
-	    desc->write_delay = ((unsigned long)delayH << 32) | delayL;
-#endif
-#ifdef USE_VM_PROBES
-	    dt_i2 = desc->write_delay;
-#endif
-	    TRACE_C('K');
-	    reply_ok(desc);
-	} goto done;
-	case FILE_OPT_READ_AHEAD: {
-	    Uint32 sizeH, sizeL;
-	    if (ev->size != 1+1+2*sizeof(Uint32)
-#ifdef USE_VM_PROBES
-		+ FILENAME_BYTELEN(dt_utag)+FILENAME_CHARSIZE
-#endif
-		|| !EV_GET_UINT32(ev, &sizeH, &p, &q)
-		|| !EV_GET_UINT32(ev, &sizeL, &p, &q)) {
-		/* Buffer has wrong length to contain the option values */
-		reply_posix_error(desc, EINVAL);
-		goto done;
-	    }
-#if SIZEOF_SIZE_T == 4
-	    if (sizeH != 0) {
-		reply_posix_error(desc, EINVAL);
-		goto done;
-	    }
-	    desc->read_bufsize = sizeL;
-#else
-	    desc->read_bufsize = ((size_t)sizeH << 32) | sizeL;
-#endif
-#ifdef USE_VM_PROBES
-	    dt_i2 = desc->read_bufsize;
-#endif
-	    TRACE_C('K');
-	    reply_ok(desc);
-	} goto done;
-	default:
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	} /* case FILE_OPT_DELAYED_WRITE: */
-    } ASSERT(0); goto done; /* case FILE_SETOPT: */
-
-    case FILE_SENDFILE: {
-
-#ifdef HAVE_SENDFILE
-        struct t_data *d;
-	Uint32 out_fd, offsetH, offsetL, hd_len, tl_len;
-	Uint64 nbytes;
-	char flags;
-
-	if (ev->size < 1 + 7 * sizeof(Uint32) + sizeof(char)
-		|| !EV_GET_UINT32(ev, &out_fd, &p, &q)
-		|| !EV_GET_CHAR(ev, &flags, &p, &q)
-		|| !EV_GET_UINT32(ev, &offsetH, &p, &q)
-		|| !EV_GET_UINT32(ev, &offsetL, &p, &q)
-		|| !EV_GET_UINT64(ev, &nbytes, &p, &q)
-		|| !EV_GET_UINT32(ev, &hd_len, &p, &q)
-		|| !EV_GET_UINT32(ev, &tl_len, &p, &q)) {
-	    /* Buffer has wrong length to contain all the needed values */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-
-	if (hd_len != 0 || tl_len != 0) {
-	    /* We do not allow header, trailers */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-
-	
-	if (flags & SENDFILE_FLGS_USE_THREADS && !THRDS_AVAILABLE) {
-	    /* We do not allow use_threads flag on a system where
-	       no threads are available. */
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-
-	d = EF_SAFE_ALLOC(sizeof(struct t_data));
-	d->fd = desc->fd;
-	d->command = command;
-	d->invoke = invoke_sendfile;
-	d->free = free_sendfile;
-	d->flags = flags;
-	d->level = 2;
-
-	d->c.sendfile.out_fd = (int) out_fd;
-	d->c.sendfile.written = 0;
-	d->c.sendfile.port = desc->port;
-	d->c.sendfile.q_mtx = desc->q_mtx;
-
-    #if SIZEOF_OFF_T == 4
-	if (offsetH != 0) {
-	    reply_posix_error(desc, EINVAL);
-	    goto done;
-	}
-	d->c.sendfile.offset = (off_t) offsetL;
-    #else
-	d->c.sendfile.offset = ((off_t) offsetH << 32) | offsetL;
-    #endif
-
-	d->c.sendfile.nbytes = nbytes;
-
-	if (USE_THRDS_FOR_SENDFILE(d)) {
-	    SET_BLOCKING(d->c.sendfile.out_fd);
-	} else {
-	    /**
-	     * Write a place holder to queue in order to force file_flush
-	     * to be called before the driver is closed.
-	     */
-	    char tmp[1] = "";
-	    MUTEX_LOCK(d->c.sendfile.q_mtx);
-	    if (driver_enq(d->c.sendfile.port, tmp, 1)) {
-	        MUTEX_UNLOCK(d->c.sendfile.q_mtx);
-		reply_posix_error(desc, ENOMEM);
-		goto done;
-	    }
-	    MUTEX_UNLOCK(d->c.sendfile.q_mtx);
-	}
-
-	cq_enq(desc, d);
-#else
-	reply_posix_error(desc, ENOTSUP);
-#endif
-	goto done;
-        } /* case FILE_SENDFILE: */
-
-    } /* switch(command) */
-
-    if (lseek_flush_read(desc, &err
-#ifdef USE_VM_PROBES
-			 , dt_priv, dt_utag
-#endif
-			 ) < 0) {
-	reply_posix_error(desc, err);
-	goto done;
-    }
-    if (flush_write_check_error(desc, &err
-#ifdef USE_VM_PROBES
-				, dt_priv, dt_utag
-#endif
-				) < 0) {
-	reply_posix_error(desc, err);
-	goto done;
-    } else {
-	/* Flatten buffer and send it to file_output(desc, buf, len) */
-	int len = ev->size;
-	char *buf = EF_ALLOC(len);
-	if (! buf) {
-	    reply_posix_error(desc, ENOMEM);
-	    goto done;
-	}
-	driver_vec_to_buf(ev, buf, len);
-	file_output((ErlDrvData) desc, buf, len);
-	EF_FREE(buf);
-	goto done;
-    }
-
- done:
-    if (d != NULL) {
-#ifdef USE_VM_PROBES
-	/*
-	 * If d == NULL, then either:
-	 *    1). There was an error of some sort, or
-	 *    2). The command given to us is actually implemented
-	 *	  by file_output() instead.
-	 *
-	 * Case #1 is probably a TODO item, perhaps?
-	 * Case #2 we definitely don't want to activate a probe.
-	 */
-	d->sched_i1 = dt_priv->thread_num;
-	d->sched_i2 = dt_priv->tag;
-	d->sched_utag[0] = '\0';
-	if (dt_utag != NULL) {
-	    if (dt_utag[0] == '\0') {
-                dt_utag = NULL;
-            } else {
-		strncpy(d->sched_utag, dt_utag, sizeof(d->sched_utag) - 1);
-		d->sched_utag[sizeof(d->sched_utag) - 1] = '\0';
-	    }
-	}
-	DTRACE11(efile_drv_entry, dt_priv->thread_num, dt_priv->tag++,
-		 dt_utag, command, dt_s1, NULL, dt_i1, dt_i2, dt_i3, dt_i4,
-		 desc->port_str);
-#endif
-    }
-    cq_execute(desc);
-}
-
-#ifdef  USE_VM_PROBES
-dt_private *
-get_dt_private(int base)
-{
-    dt_private *dt_priv = (dt_private *) pthread_getspecific(dt_driver_key);
-
-    if (dt_priv == NULL) {
-	dt_priv = EF_SAFE_ALLOC(sizeof(dt_private));
-	erts_mtx_lock(&dt_driver_mutex);
-	dt_priv->thread_num = (base + dt_driver_idnum++);
-	erts_mtx_unlock(&dt_driver_mutex);
-	dt_priv->tag = 0;
-	pthread_setspecific(dt_driver_key, dt_priv);
-    }
-    return dt_priv;
-}
-#endif  /* USE_VM_PROBES */
diff --git a/erts/emulator/drivers/common/erl_efile.h b/erts/emulator/drivers/common/erl_efile.h
deleted file mode 100644
index b7f063b4f2..0000000000
--- a/erts/emulator/drivers/common/erl_efile.h
+++ /dev/null
@@ -1,176 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 1997-2016. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-/*
- * Defines the interfaces between the generic efile driver and its
- * operating-system dependent helpers.
- */
-
-#include "sys.h"
-#include "erl_driver.h"
-
-/*
- * Open modes for efile_openfile().
- */
-#define EFILE_MODE_READ 1
-#define EFILE_MODE_WRITE 2	/* Implies truncating file when used alone. */
-#define EFILE_MODE_READ_WRITE 	3
-#define EFILE_MODE_APPEND	4
-#define EFILE_COMPRESSED 	8
-#define EFILE_MODE_EXCL        16
-#define EFILE_NO_TRUNCATE      32 /* Special for reopening on VxWorks */
-#define EFILE_MODE_SYNC        64
-
-/*
- * Seek modes for efile_seek().
- */
-#define	EFILE_SEEK_SET	0
-#define	EFILE_SEEK_CUR	1
-#define	EFILE_SEEK_END	2
-
-/*
- * File types returned by efile_fileinfo().
- */
-#define FT_DEVICE 	1
-#define FT_DIRECTORY 	2
-#define FT_REGULAR	3
-#define FT_SYMLINK	4
-#define FT_OTHER	5
-
-/*
- * Access attributes returned by efile_fileinfo() (the bits can be ORed
- * together).
- */
-#define FA_NONE    	0
-#define FA_WRITE 	1
-#define FA_READ		2
-
-/* Some OS'es (i.e. Windows) has filenames in wide charaqcters. That requires special handling */
-/* Note that we do *not* honor alignment in the communication to the OS specific driver, */
-/* which is not a problem on x86, but might be on other platforms. The OS specific efile */
-/* implementation is expected to align if needed */
-#ifdef __WIN32__
-#define FILENAMES_16BIT 1
-#endif
-
-/* We use sendfilev if it exist on solaris */
-#if !defined(HAVE_SENDFILE) && defined(HAVE_SENDFILEV)
-#define HAVE_SENDFILE
-#endif
-
-/*
- * An handle to an open directory.  To be cast to the correct type
- * in the system-dependent directory functions.
- */
-
-typedef struct _Efile_Dir_Handle* EFILE_DIR_HANDLE;
-
-/*
- * Error information from the last call.
- */  
-typedef struct _Efile_error {
-    int posix_errno;		/* Posix error number, as in <errno.h>. */
-    int os_errno;		/* Os-dependent error number (not used). */
-} Efile_error;
-
-/*
- * Describes what is returned by file:file_info/1.
- */
-
-typedef struct _Efile_info {
-    Uint32 size_low;		/* Size of file, lower 32 bits.. */
-    Uint32 size_high;		/* Size of file, higher 32 bits. */
-    Uint32 type;		/* Type of file -- one of FT_*. */
-    Uint32 access;		/* Access to file -- one of FA_*. */
-    Uint32 mode;		/* Access permissions -- bit field. */
-    Uint32 links;		/* Number of links to file. */
-    Uint32 major_device;	/* Major device or file system. */
-    Uint32 minor_device;	/* Minor device (for devices). */
-    Uint32 inode;		/* Inode number. */
-    Uint32 uid;			/* User id of owner. */
-    Uint32 gid;			/* Group id of owner. */
-    Sint64 accessTime;		/* Last time the file was accessed. */
-    Sint64 modifyTime;		/* Last time the file was modified. */
-    Sint64 cTime;		/* Creation time (Windows) or last
-				 * inode change (Unix).
-				 */
-} Efile_info;
-
-
-#ifdef HAVE_SENDFILE
-/*
- * Describes the structure of headers/trailers for sendfile
- */
-struct t_sendfile_hdtl {
-    SysIOVec *headers;
-    int hdr_cnt;
-    SysIOVec *trailers;
-    int trl_cnt;
-};
-#endif /* HAVE_SENDFILE */
-
-/*
- * Functions.
- */
-int efile_init(void);
-int efile_mkdir(Efile_error* errInfo, char* name);
-int efile_rmdir(Efile_error* errInfo, char* name);
-int efile_delete_file(Efile_error* errInfo, char* name);
-int efile_rename(Efile_error* errInfo, char* src, char* dst);
-int efile_chdir(Efile_error* errInfo, char* name);
-int efile_getdcwd(Efile_error* errInfo, int drive,
-		  char* buffer, size_t size);
-int efile_readdir(Efile_error* errInfo, char* name, 
-		  EFILE_DIR_HANDLE* dir_handle,
-		  char* buffer, size_t *size);
-int efile_openfile(Efile_error* errInfo, char* name, int flags,
-		   int* pfd, Sint64* pSize);
-void efile_closefile(int fd);
-int efile_fdatasync(Efile_error* errInfo, int fd);
-int efile_fsync(Efile_error* errInfo, int fd);
-int efile_fileinfo(Efile_error* errInfo, Efile_info* pInfo,
-		   char *name, int info_for_link);
-int efile_write_info(Efile_error* errInfo, Efile_info* pInfo, char *name);
-int efile_write(Efile_error* errInfo, int flags, int fd, 
-		char* buf, size_t count);
-int efile_writev(Efile_error* errInfo, int flags, int fd, 
-		 SysIOVec* iov, int iovcnt);
-int efile_read(Efile_error* errInfo, int flags, int fd, 
-	       char* buf, size_t count, size_t* pBytesRead);
-int efile_seek(Efile_error* errInfo, int fd, 
-	       Sint64 offset, int origin, Sint64* new_location);
-int efile_truncate_file(Efile_error* errInfo, int *fd, int flags);
-int efile_pwrite(Efile_error* errInfo, int fd, 
-		 char* buf, size_t count, Sint64 offset);
-int efile_pread(Efile_error* errInfo, int fd, 
-		Sint64 offset, char* buf, size_t count, size_t* pBytesRead);
-int efile_readlink(Efile_error* errInfo, char *name, 
-		   char* buffer, size_t size);
-int efile_altname(Efile_error* errInfo, char *name, 
-		  char* buffer, size_t size);
-int efile_link(Efile_error* errInfo, char* old, char* new);
-int efile_symlink(Efile_error* errInfo, char* old, char* new);
-int efile_may_openfile(Efile_error* errInfo, char *name);
-int efile_fadvise(Efile_error* errInfo, int fd, Sint64 offset, Sint64 length,
-		  int advise);
-#ifdef HAVE_SENDFILE
-int efile_sendfile(Efile_error* errInfo, int in_fd, int out_fd,
-		      off_t *offset, Uint64 *nbytes, struct t_sendfile_hdtl *hdtl);
-#endif /* HAVE_SENDFILE */
-int efile_fallocate(Efile_error* errInfo, int fd, Sint64 offset, Sint64 length);
diff --git a/erts/emulator/drivers/common/gzio.c b/erts/emulator/drivers/common/gzio.c
index f60c781894..86c3b07cea 100644
--- a/erts/emulator/drivers/common/gzio.c
+++ b/erts/emulator/drivers/common/gzio.c
@@ -19,726 +19,16 @@
 #include <unistd.h>
 #endif
 #include <ctype.h>
+
 #include "erl_driver.h"
-#include "erl_efile.h"
 #include "sys.h"
 
-#ifdef __WIN32__
-#ifndef HAVE_CONFLICTING_FREAD_DECLARATION
-#define HAVE_CONFLICTING_FREAD_DECLARATION
-#endif
-#define FILENAMES_16BIT 1
-#endif
-
-#ifdef STDC
-#  define zstrerror(errnum) strerror(errnum)
-#else
-#  define zstrerror(errnum) ""
-#endif
-
 #include "gzio_zutil.h"
 #include "erl_zlib.h"
 #include "gzio.h"
 
-/********struct internal_state {int dummy;}; / * for buggy compilers */
-
-#define Z_BUFSIZE 4096
-
-#define ALLOC(size) driver_alloc(size)
-#define TRYFREE(p) {if (p) driver_free(p);}
-
 static int gz_magic[2] = {0x1f, 0x8b}; /* gzip magic header */
 
-/* gzip flag byte */
-#define ASCII_FLAG   0x01 /* bit 0 set: file probably ascii text */
-#define HEAD_CRC     0x02 /* bit 1 set: header CRC present */
-#define EXTRA_FIELD  0x04 /* bit 2 set: extra field present */
-#define ORIG_NAME    0x08 /* bit 3 set: original file name present */
-#define COMMENT      0x10 /* bit 4 set: file comment present */
-#define RESERVED     0xE0 /* bits 5..7: reserved */
-
-typedef struct gz_stream {
-    z_stream stream;
-    int      z_err;   /* error code for last stream operation */
-    int      z_eof;   /* set if end of input file */
-#ifdef UNIX
-    int      file;    /* .gz file descriptor */
-#else
-    FILE     *file;   /* .gz file */
-#endif
-    Byte     *inbuf;  /* input buffer */
-    Byte     *outbuf; /* output buffer */
-    uLong    crc;     /* crc32 of uncompressed data */
-    char     *msg;    /* error message */
-    char     *path;   /* path name for debugging only */
-    int      transparent; /* 1 if input file is not a .gz file */
-    char     mode;    /* 'w' or 'r' */
-    int      position; /* Position (for seek) */
-    int (*destroy)(struct gz_stream*); /* Function to destroy
-					    *  this structure. */
-} gz_stream;
-
-local ErtsGzFile gz_open      (const char *path, const char *mode);
-local int    get_byte     (gz_stream *s);
-local void   check_header (gz_stream *s);
-local int    destroy      (gz_stream *s);
-local uLong  getLong      (gz_stream *s);
-
-#ifdef UNIX
-/*
- * In Solaris 8 and earlier, fopen() and its friends cannot handle 
- * file descriptors larger than 255. Therefore, we use read()/write()
- * on all Unix systems.
- */
-# define ERTS_GZWRITE(File, Buf, Count) write((File), (Buf), (Count))
-# define ERTS_GZREAD(File, Buf, Count) read((File), (Buf), (Count))
-#else
-/*
- * On all other operating systems, using fopen(), fread()/fwrite(), since
- * there is not guaranteed to exist any read()/write() (not part of
- * ANSI/ISO-C).
- */
-# define ERTS_GZWRITE(File, Buf, Count) fwrite((Buf), 1, (Count), (File))
-# define ERTS_GZREAD(File, Buf, Count) fread((Buf), 1, (Count), (File))
-#endif
-
-/*
- * Ripped from efile_drv.c
- */
-
-#ifdef FILENAMES_16BIT
-#  define FILENAME_BYTELEN(Str) filename_len_16bit(Str)
-#  define FILENAME_COPY(To,From) filename_cpy_16bit((To),(From)) 
-#  define FILENAME_CHARSIZE 2
-
-   static int filename_len_16bit(const char *str) 
-   {
-       const char *p = str;
-       while(*p != '\0' || p[1] != '\0') {
-	   p += 2;
-       }
-       return (p - str);
-   }
-
-   static void filename_cpy_16bit(char *to, const char *from) 
-   {
-       while(*from != '\0' || from[1] != '\0') {
-	   *to++ = *from++;
-	   *to++ = *from++;
-       }
-       *to++ = *from++;
-       *to++ = *from++;
-   }
-
-#else
-#  define FILENAME_BYTELEN(Str) strlen(Str)
-#  define FILENAME_COPY(To,From) strcpy(To,From) 
-#  define FILENAME_CHARSIZE 1
-#endif
-
-/* ===========================================================================
-     Opens a gzip (.gz) file for reading or writing. The mode parameter
-   is as in fopen ("rb" or "wb"). The file is given either by file descriptor
-   or path name (if fd == -1).
-     gz_open return NULL if the file could not be opened or if there was
-   insufficient memory to allocate the (de)compression state; errno
-   can be checked to distinguish the two cases (if errno is zero, the
-   zlib error is Z_MEM_ERROR).
-*/
-local ErtsGzFile gz_open (path, mode)
-    const char *path;
-    const char *mode;
-{
-    int err;
-    int level = Z_DEFAULT_COMPRESSION; /* compression level */
-    char *p = (char*)mode;
-    gz_stream *s;
-    char fmode[80]; /* copy of mode, without the compression level */
-    char *m = fmode;
-
-    if (!path || !mode) return Z_NULL;
-
-    s = (gz_stream *)ALLOC(sizeof(gz_stream));
-    if (!s) return Z_NULL;
-
-    erl_zlib_alloc_init(&s->stream);
-    s->stream.next_in = s->inbuf = Z_NULL;
-    s->stream.next_out = s->outbuf = Z_NULL;
-    s->stream.avail_in = s->stream.avail_out = 0;
-#ifdef UNIX
-    s->file = -1;
-#else
-    s->file = NULL;
-#endif
-    s->z_err = Z_OK;
-    s->z_eof = 0;
-    s->crc = crc32(0L, Z_NULL, 0);
-    s->msg = NULL;
-    s->transparent = 0;
-    s->position = 0;
-    s->destroy = destroy;
-
-    s->path = (char*)ALLOC(FILENAME_BYTELEN(path)+FILENAME_CHARSIZE);
-    if (s->path == NULL) {
-        return s->destroy(s), (ErtsGzFile)Z_NULL;
-    }
-    FILENAME_COPY(s->path, path); /* do this early for debugging */
-
-    s->mode = '\0';
-    do {
-        if (*p == 'r')
-	    s->mode = 'r';
-	if (*p == 'w' || *p == 'a')
-	    s->mode = 'w';
-	if (isdigit((int)*p)) {
-	    level = *p - '0';
-	} else {
-	    *m++ = *p;		/* Copy the mode */
-	}
-    } while (*p++ && m < fmode + sizeof(fmode) - 1);
-    *m = '\0';
-    if (s->mode == '\0')
-	return s->destroy(s), (ErtsGzFile)Z_NULL;
-    
-    if (s->mode == 'w') {
-        err = deflateInit2(&(s->stream), level,
-                           Z_DEFLATED, MAX_WBITS+16, DEF_MEM_LEVEL, 0);
-        /* windowBits is passed < 0 to suppress zlib header */
-
-        s->stream.next_out = s->outbuf = (Byte*)ALLOC(Z_BUFSIZE);
-
-        if (err != Z_OK || s->outbuf == Z_NULL) {
-            return s->destroy(s), (ErtsGzFile)Z_NULL;
-        }
-    } else {
-	/*
-	 * It is tempting to use the built-in support in zlib
-	 * for handling GZIP headers, but unfortunately it
-	 * cannot handle multiple GZIP headers (which occur when
-	 * several GZIP files have been concatenated).
-	 */
-
-	err = inflateInit2(&(s->stream), -MAX_WBITS);
-        s->stream.next_in  = s->inbuf = (Byte*)ALLOC(Z_BUFSIZE);
-
-        if (err != Z_OK || s->inbuf == Z_NULL) {
-            return s->destroy(s), (ErtsGzFile)Z_NULL;
-        }
-    }
-    s->stream.avail_out = Z_BUFSIZE;
-
-    errno = 0;
-#if defined(FILENAMES_16BIT)
-    {
-	FILE* efile_wfopen(const WCHAR* name, const WCHAR* mode);
-	WCHAR wfmode[80];
-	int i = 0;
-	int j;
-	for(j = 0; fmode[j] != '\0'; ++j) {
-	    wfmode[i++] = (WCHAR) fmode[j];
-	}
-	wfmode[i++] = L'\0';
-	s->file = efile_wfopen((WCHAR *)path, wfmode);
-	if (s->file == NULL) {
-	    return s->destroy(s), (ErtsGzFile)Z_NULL;
-	}
-    }
-#elif defined(UNIX)
-    if (s->mode == 'r') {
-	s->file = open(path, O_RDONLY);
-    } else {
-	s->file = open(path, O_WRONLY | O_CREAT | O_TRUNC, 0666);
-    }
-    if (s->file == -1) {
-        return s->destroy(s), (ErtsGzFile)Z_NULL;
-    }
-#else
-    s->file = fopen(path, fmode);
-    if (s->file == NULL) {
-        return s->destroy(s), (ErtsGzFile)Z_NULL;
-    }
-#endif
-    if (s->mode == 'r') {
-	check_header(s); /* skip the .gz header */
-    }
-    return (ErtsGzFile)s;
-}
-
-/* ===========================================================================
-   Rewind a gzfile back to the beginning.
-*/
-
-local int gz_rewind (gz_stream *s)
-{
-    TRYFREE(s->msg);
-
-#ifdef UNIX
-    lseek(s->file, 0L, SEEK_SET);
-#else
-    fseek(s->file, 0L, SEEK_SET);
-#endif
-    inflateReset(&(s->stream));
-    s->stream.next_in = Z_NULL;
-    s->stream.next_out = Z_NULL;
-    s->stream.avail_in = s->stream.avail_out = 0;
-    s->z_err = Z_OK;
-    s->z_eof = 0;
-    s->crc = crc32(0L, Z_NULL, 0);
-    s->msg = NULL;
-    s->position = 0;
-    s->stream.next_in = s->inbuf;
-
-    s->stream.avail_out = Z_BUFSIZE;
-
-    check_header(s);		/* skip the .gz header */
-    return 1;
-}
-
-/* ===========================================================================
-     Opens a gzip (.gz) file for reading or writing.
-*/
-ErtsGzFile erts_gzopen (path, mode)
-    const char *path;
-    const char *mode;
-{
-    return gz_open (path, mode);
-}
-
-
-/* ===========================================================================
-     Read a byte from a gz_stream; update next_in and avail_in. Return EOF
-   for end of file.
-   IN assertion: the stream s has been successfully opened for reading.
-*/
-local int get_byte(s)
-    gz_stream *s;
-{
-    if (s->z_eof) return EOF;
-    if (s->stream.avail_in == 0) {
-#ifdef UNIX
-	ssize_t res;
-	errno = 0;
-	res = ERTS_GZREAD(s->file, s->inbuf, Z_BUFSIZE);
-	if (res == 0) {
-	    s->stream.avail_in = 0;
-	    s->z_eof = 1;
-	    return EOF;
-	} else if (res < 0) {
-	    s->stream.avail_in = 0;
-	    s->z_eof = 1;
-	    s->z_err = Z_ERRNO;
-	    return EOF;
-	} else {
-	    s->stream.avail_in = (uInt) res;
-	}
-#else
-	errno = 0;
-	s->stream.avail_in = ERTS_GZREAD(s->file, s->inbuf, Z_BUFSIZE);
-	if (s->stream.avail_in == 0) {
-	    s->z_eof = 1;
-	    if (s->file && ferror(s->file))
-		s->z_err = Z_ERRNO;
-	    return EOF;
-	}
-#endif
-	s->stream.next_in = s->inbuf;
-    }
-    s->stream.avail_in--;
-    return *(s->stream.next_in)++;
-}
-
-/* ===========================================================================
-      Check the gzip header of a gz_stream opened for reading. Set the stream
-    mode to transparent if the gzip magic header is not present; set s->err
-    to Z_DATA_ERROR if the magic header is present but the rest of the header
-    is incorrect.
-    IN assertion: the stream s has already been created sucessfully;
-       s->stream.avail_in is zero for the first time, but may be non-zero
-       for concatenated .gz files.
-*/
-local void check_header(s)
-    gz_stream *s;
-{
-    int method; /* method byte */
-    int flags;  /* flags byte */
-    uInt len;
-    int c;
-
-    /* Check the gzip magic header */
-    for (len = 0; len < 2; len++) {
-	c = get_byte(s);
-	if (c != gz_magic[len]) {
-	    if (len != 0) s->stream.avail_in++, s->stream.next_in--;
-	    if (c != EOF) {
-		s->stream.avail_in++, s->stream.next_in--;
-		s->transparent = 1;
-	    }
-	    s->z_err = s->stream.avail_in != 0 ? Z_OK : Z_STREAM_END;
-	    return;
-	}
-    }
-    method = get_byte(s);
-    flags = get_byte(s);
-    if (method != Z_DEFLATED || (flags & RESERVED) != 0) {
-	s->z_err = Z_DATA_ERROR;
-	return;
-    }
-
-    /* Discard time, xflags and OS code: */
-    for (len = 0; len < 6; len++) (void)get_byte(s);
-
-    if ((flags & EXTRA_FIELD) != 0) { /* skip the extra field */
-	len  =  (uInt)get_byte(s);
-	len += ((uInt)get_byte(s))<<8;
-	/* len is garbage if EOF but the loop below will quit anyway */
-	while (len-- != 0 && get_byte(s) != EOF) ;
-    }
-    if ((flags & ORIG_NAME) != 0) { /* skip the original file name */
-	while ((c = get_byte(s)) != 0 && c != EOF) ;
-    }
-    if ((flags & COMMENT) != 0) {   /* skip the .gz file comment */
-	while ((c = get_byte(s)) != 0 && c != EOF) ;
-    }
-    if ((flags & HEAD_CRC) != 0) {  /* skip the header crc */
-	for (len = 0; len < 2; len++) (void)get_byte(s);
-    }
-    s->z_err = s->z_eof ? Z_DATA_ERROR : Z_OK;
-}
-
- /* ===========================================================================
- * Cleanup then free the given gz_stream. Return a zlib error code.
-   Try freeing in the reverse order of allocations.
- */
-local int destroy (s)
-    gz_stream *s;
-{
-    int err = Z_OK;
-
-    if (!s) return Z_STREAM_ERROR;
-
-    TRYFREE(s->msg);
-
-    if (s->stream.state != NULL) {
-       if (s->mode == 'w') {
-           err = deflateEnd(&(s->stream));
-       } else if (s->mode == 'r') {
-           err = inflateEnd(&(s->stream));
-       }
-    }
-#ifdef UNIX
-    if (s->file != -1 && close(s->file)) {
-        err = Z_ERRNO;
-    }
-#else
-    if (s->file != NULL && fclose(s->file)) {
-        err = Z_ERRNO;
-    }
-#endif
-    if (s->z_err < 0) err = s->z_err;
-
-    TRYFREE(s->inbuf);
-    TRYFREE(s->outbuf);
-    TRYFREE(s->path);
-    TRYFREE(s);
-    return err;
-}
-
-/* ===========================================================================
-     Reads the given number of uncompressed bytes from the compressed file.
-   gzread returns the number of bytes actually read (0 for end of file).
-*/
-int
-erts_gzread(ErtsGzFile file, voidp buf, unsigned len)
-{
-    gz_stream *s = (gz_stream*)file;
-    Bytef *start = buf; /* starting point for crc computation */
-    Byte  *next_out; /* == stream.next_out but not forced far (for MSDOS) */
-
-    if (s == NULL || s->mode != 'r') return Z_STREAM_ERROR;
-
-    if (s->z_err == Z_DATA_ERROR || s->z_err == Z_ERRNO) return -1;
-    if (s->z_err == Z_STREAM_END) return 0;  /* EOF */
-
-    s->stream.next_out = next_out = buf;
-    s->stream.avail_out = len;
-
-    while (s->stream.avail_out != 0) {
-
-	if (s->transparent) {
-	    /* Copy first the lookahead bytes: */
-	    uInt n = s->stream.avail_in;
-	    if (n > s->stream.avail_out) n = s->stream.avail_out;
-	    if (n > 0) {
-		zmemcpy(s->stream.next_out, s->stream.next_in, n);
-		next_out += n;
-		s->stream.next_out = next_out;
-		s->stream.next_in   += n;
-		s->stream.avail_out -= n;
-		s->stream.avail_in  -= n;
-	    }
-	    if (s->stream.avail_out > 0) {
-		s->stream.avail_out -= ERTS_GZREAD(s->file, next_out,
-						   s->stream.avail_out);
-	    }
-	    len -= s->stream.avail_out;
-	    s->stream.total_in  += (uLong)len;
-	    s->stream.total_out += (uLong)len;
-            if (len == 0) s->z_eof = 1;
-	    s->position += (int)len;
-	    return (int)len;
-	}
-        if (s->stream.avail_in == 0 && !s->z_eof) {
-#ifdef UNIX
-	    ssize_t res;
-	    errno = 0;
-	    res = ERTS_GZREAD(s->file, s->inbuf, Z_BUFSIZE);
-	    if (res == 0) {
-		s->stream.avail_in = 0;
-		s->z_eof = 1;
-		return EOF;
-	    } else if (res < 0) {
-		s->stream.avail_in = 0;
-		s->z_eof = 1;
-		s->z_err = Z_ERRNO;
-		return EOF;
-	    } else {
-		s->stream.avail_in = (uInt) res;
-	    }
-#else
-	    errno = 0;
-            s->stream.avail_in = ERTS_GZREAD(s->file, s->inbuf, Z_BUFSIZE);
-            if (s->stream.avail_in == 0) {
-                s->z_eof = 1;
-		if (s->file && ferror(s->file)) {
-		    s->z_err = Z_ERRNO;
-		    break;
-		}
-            }
-#endif
-            s->stream.next_in = s->inbuf;
-        }
-        s->z_err = inflate(&(s->stream), Z_NO_FLUSH);
-
-	if (s->z_err == Z_STREAM_END) {
-	    /* Check CRC and original size */
-	    s->crc = crc32(s->crc, start, (uInt)(s->stream.next_out - start));
-	    start = s->stream.next_out;
-
-	    if (getLong(s) != s->crc) {
-		s->z_err = Z_DATA_ERROR;
-	    } else {
-	        (void)getLong(s);
-                /* The uncompressed length returned by above getlong() may
-                 * be different from s->stream.total_out) in case of
-		 * concatenated .gz files. Check for such files:
-		 */
-		check_header(s);
-		if (s->z_err == Z_OK) {
-		    uLong total_in = s->stream.total_in;
-		    uLong total_out = s->stream.total_out;
-
-		    inflateReset(&(s->stream));
-		    s->stream.total_in = total_in;
-		    s->stream.total_out = total_out;
-		    s->crc = crc32(0L, Z_NULL, 0);
-		}
-	    }
-	}
-	if (s->z_err != Z_OK || s->z_eof) break;
-    }
-    s->crc = crc32(s->crc, start, (uInt)(s->stream.next_out - start));
-
-    s->position += (int)(len - s->stream.avail_out);
-
-    return (int)(len - s->stream.avail_out);
-}
-
-/* ===========================================================================
-     Writes the given number of uncompressed bytes into the compressed file.
-   gzwrite returns the number of bytes actually written (0 in case of error).
-*/
-int
-erts_gzwrite(ErtsGzFile file, voidp buf, unsigned len)
-{
-    gz_stream *s = (gz_stream*)file;
-
-    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
-
-    s->stream.next_in = buf;
-    s->stream.avail_in = len;
-
-    while (s->stream.avail_in != 0) {
-
-        if (s->stream.avail_out == 0) {
-
-            s->stream.next_out = s->outbuf;
-            if (ERTS_GZWRITE(s->file, s->outbuf, Z_BUFSIZE) != Z_BUFSIZE) {
-                s->z_err = Z_ERRNO;
-                break;
-            }
-            s->stream.avail_out = Z_BUFSIZE;
-        }
-        s->z_err = deflate(&(s->stream), Z_NO_FLUSH);
-        if (s->z_err != Z_OK) break;
-    }
-    s->position += (int)(len - s->stream.avail_in);
-    return (int)(len - s->stream.avail_in);
-}
-
-/*
- * For use by Erlang file driver.
- *
- * XXX Limitations:
- *  - SEEK_END is not allowed (length of file is not known).
- *  - When writing, only forward seek is supported.
- */
-
-int
-erts_gzseek(ErtsGzFile file, int offset, int whence)
-{
-    int pos;
-    gz_stream* s = (gz_stream *) file;
-
-    switch (whence) {
-    case EFILE_SEEK_SET: whence = SEEK_SET; break;
-    case EFILE_SEEK_CUR: whence = SEEK_CUR; break;
-    case EFILE_SEEK_END: whence = SEEK_END; break;
-    default:
-	errno = EINVAL;
-	return -1;
-    }
-
-    if (s == NULL) {
-	errno = EINVAL;
-	return -1;
-    }
-    if (s->z_err == Z_DATA_ERROR || s->z_err == Z_ERRNO) {
-	errno = EIO;
-	return -1;
-    }
-
-    switch (whence) {
-    case SEEK_SET: pos = offset; break;
-    case SEEK_CUR: pos = s->position+offset; break;
-    case SEEK_END: 
-    default:
-      errno = EINVAL; return -1;
-    }
-
-    if (pos == s->position) {
-	return pos;
-    }
-
-    if (pos < s->position) {
-	if (s->mode == 'w') {
-	    errno = EINVAL;
-	    return -1;
-	}
-	gz_rewind(s);
-    }
-
-    while (s->position < pos) {
-	char buf[512];
-	int n;
-	int save_pos = s->position;
-
-	n = pos - s->position;
-	if (n > sizeof(buf))
-	    n = sizeof(buf);
-
-	if (s->mode == 'r') {
-	    erts_gzread(file, buf, n);
-	} else {
-	    memset(buf, '\0', n);
-	    erts_gzwrite(file, buf, n);
-	}
-	if (save_pos == s->position) break;
-    }
-
-    return s->position;
-}
-
-/* ===========================================================================
-     Flushes all pending output into the compressed file. The parameter
-   flush is as in the deflate() function.
-     gzflush should be called only when strictly necessary because it can
-   degrade compression.
-*/
-int
-erts_gzflush(ErtsGzFile file, int flush)
-{
-    uInt len;
-    int done = 0;
-    gz_stream *s = (gz_stream*)file;
-
-    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
-
-    s->stream.avail_in = 0; /* should be zero already anyway */
-
-    for (;;) {
-        len = Z_BUFSIZE - s->stream.avail_out;
-
-        if (len != 0) {
-            if ((uInt)ERTS_GZWRITE(s->file, s->outbuf, len) != len) {
-                s->z_err = Z_ERRNO;
-                return Z_ERRNO;
-            }
-            s->stream.next_out = s->outbuf;
-            s->stream.avail_out = Z_BUFSIZE;
-        }
-        if (done) break;
-        s->z_err = deflate(&(s->stream), flush);
-
-        /* deflate has finished flushing only when it hasn't used up
-         * all the available space in the output buffer: 
-         */
-        done = (s->stream.avail_out != 0 || s->z_err == Z_STREAM_END);
- 
-        if (s->z_err != Z_OK && s->z_err != Z_STREAM_END) break;
-    }
-#ifndef UNIX
-    fflush(s->file);
-#endif
-    return  s->z_err == Z_STREAM_END ? Z_OK : s->z_err;
-}
-
-/* ===========================================================================
-   Reads a long in LSB order from the given gz_stream. Sets 
-*/
-local uLong getLong (s)
-    gz_stream *s;
-{
-    uLong x = (uLong)get_byte(s);
-    int c;
-
-    x += ((uLong)get_byte(s))<<8;
-    x += ((uLong)get_byte(s))<<16;
-    c = get_byte(s);
-    if (c == EOF) s->z_err = Z_DATA_ERROR;
-    x += ((uLong)c)<<24;
-    return x;
-}
-
-/* ===========================================================================
-     Flushes all pending output if necessary, closes the compressed file
-   and deallocates all the (de)compression state.
-*/
-int
-erts_gzclose(ErtsGzFile file)
-{
-    int err;
-    gz_stream *s = (gz_stream*)file;
-
-    if (s == NULL) return Z_STREAM_ERROR;
-
-    if (s->mode == 'w') {
-        err = erts_gzflush (file, Z_FINISH);
-        if (err != Z_OK) return s->destroy(s);
-    }
-    return s->destroy(s);
-}
-
-
 /* ===========================================================================
    Uncompresses the buffer given and returns a pointer to a binary.
    If the buffer was not compressed with gzip, the buffer contents
diff --git a/erts/emulator/drivers/common/gzio.h b/erts/emulator/drivers/common/gzio.h
index ee0ebe7bd8..e331b5208b 100644
--- a/erts/emulator/drivers/common/gzio.h
+++ b/erts/emulator/drivers/common/gzio.h
@@ -20,13 +20,5 @@
 
 #include "zlib.h"
 
-typedef struct erts_gzFile* ErtsGzFile;
-
-ErtsGzFile erts_gzopen (const char *path, const char *mode);
-int erts_gzread(ErtsGzFile file, voidp buf, unsigned len);
-int erts_gzwrite(ErtsGzFile file, voidp buf, unsigned len);
-int erts_gzseek(ErtsGzFile, int, int);
-int erts_gzflush(ErtsGzFile file, int flush);
-int erts_gzclose(ErtsGzFile file);
 ErlDrvBinary* erts_gzinflate_buffer(char*, uLong);
 ErlDrvBinary* erts_gzdeflate_buffer(char*, uLong);
diff --git a/erts/emulator/drivers/common/inet_drv.c b/erts/emulator/drivers/common/inet_drv.c
index 554c48059f..1a68f65b52 100644
--- a/erts/emulator/drivers/common/inet_drv.c
+++ b/erts/emulator/drivers/common/inet_drv.c
@@ -63,6 +63,20 @@
 #include <sys/un.h>
 #endif
 
+#ifdef HAVE_SENDFILE
+#if defined(__linux__) || (defined(__sun) && defined(__SVR4))
+    #include <sys/sendfile.h>
+#elif defined(__FreeBSD__) || defined(__DragonFly__)
+    /* Need to define __BSD_VISIBLE in order to expose prototype of sendfile */
+    #define __BSD_VISIBLE 1
+    #include <sys/socket.h>
+#endif
+#endif
+
+#if defined(__APPLE__) && defined(__MACH__) && !defined(__DARWIN__)
+    #define __DARWIN__ 1
+#endif
+
 /* All platforms fail on malloc errors. */
 #define FATAL_MALLOC
 
@@ -701,6 +715,7 @@ static size_t my_strnlen(const char *s, size_t maxlen)
 #define TCP_REQ_RECV           42
 #define TCP_REQ_UNRECV         43
 #define TCP_REQ_SHUTDOWN       44
+#define TCP_REQ_SENDFILE       45
 /* UDP and SCTP requests */
 #define PACKET_REQ_RECV        60 /* Common for UDP and SCTP         */
 /* #define SCTP_REQ_LISTEN       61 MERGED Different from TCP; not for UDP */
@@ -723,6 +738,7 @@ static size_t my_strnlen(const char *s, size_t maxlen)
 #define TCP_ADDF_DELAYED_ECONNRESET 128 /* An ECONNRESET error occurred on send or shutdown */
 #define TCP_ADDF_SHUTDOWN_WR_DONE 256 /* A shutdown(sock, SHUT_WR) or SHUT_RDWR was made */
 #define TCP_ADDF_LINGER_ZERO 	  512 /* Discard driver queue on port close */
+#define TCP_ADDF_SENDFILE         1024 /* Send from an fd instead of the driver queue */
 
 /* *_REQ_* replies */
 #define INET_REP_ERROR       0
@@ -1136,7 +1152,6 @@ static int        packet_inet_init(void);
 static void       packet_inet_stop(ErlDrvData);
 static void       packet_inet_command(ErlDrvData, char*, ErlDrvSizeT);
 static void       packet_inet_drv_input(ErlDrvData data, ErlDrvEvent event);
-static void	  packet_inet_drv_output(ErlDrvData data, ErlDrvEvent event);
 static ErlDrvData udp_inet_start(ErlDrvPort, char* command);
 #ifdef HAVE_SCTP
 static ErlDrvData sctp_inet_start(ErlDrvPort, char* command);
@@ -1161,7 +1176,7 @@ static struct erl_drv_entry udp_inet_driver_entry =
     NULL, 
 #else
     packet_inet_drv_input,
-    packet_inet_drv_output,
+    NULL,
 #endif
     "udp_inet",
     NULL,
@@ -1196,7 +1211,7 @@ static struct erl_drv_entry sctp_inet_driver_entry =
     NULL, 
 #else
     packet_inet_drv_input,
-    packet_inet_drv_output,
+    NULL,
 #endif
     "sctp_inet",
     NULL,
@@ -1236,6 +1251,21 @@ typedef struct {
     inet_async_multi_op *multi_first;/* NULL == no multi-accept-queue, op is in ordinary queue */
     inet_async_multi_op *multi_last;
     MultiTimerData *mtd;        /* Timer structures for multiple accept */
+#ifdef HAVE_SENDFILE
+    struct {
+        ErlDrvSizeT ioq_skip;   /* The number of bytes in the queue at the time
+                                 * sendfile was issued, which must be sent
+                                 * before issuing the sendfile call itself. */
+        int dup_file_fd;        /* The file handle to send from; this is
+                                 * duplicated when sendfile is issued to
+                                 * reduce (but not eliminate) the impact of a
+                                 * nasty race, so we have to remember to close
+                                 * it. */
+        Uint64 bytes_sent;
+        Uint64 offset;
+        Uint64 length;
+    } sendfile;
+#endif
 } tcp_descriptor;
 
 /* send function */
@@ -1246,6 +1276,8 @@ static int tcp_deliver(tcp_descriptor* desc, int len);
 
 static int tcp_shutdown_error(tcp_descriptor* desc, int err);
 
+static int tcp_inet_sendfile(tcp_descriptor* desc);
+
 static int tcp_inet_output(tcp_descriptor* desc, HANDLE event);
 static int tcp_inet_input(tcp_descriptor* desc, HANDLE event);
 
@@ -1262,7 +1294,6 @@ typedef struct {
 
 
 static int packet_inet_input(udp_descriptor* udesc, HANDLE event);
-static int packet_inet_output(udp_descriptor* udesc, HANDLE event);
 #endif
 
 /* convert descriptor pointer to inet_descriptor pointer */
@@ -1331,6 +1362,9 @@ static ErlDrvTermData am_ipv6_v6only;
 static ErlDrvTermData am_netns;
 static ErlDrvTermData am_bind_to_device;
 #endif
+#ifdef HAVE_SENDFILE
+static ErlDrvTermData am_sendfile;
+#endif
 
 static char str_eafnosupport[] = "eafnosupport";
 static char str_einval[] = "einval";
@@ -3877,6 +3911,10 @@ static int inet_init()
     INIT_ATOM(https);
     INIT_ATOM(scheme);
 
+#ifdef HAVE_SENDFILE
+    INIT_ATOM(sendfile);
+#endif
+
     /* add TCP, UDP and SCTP drivers */
     add_driver_entry(&tcp_inet_driver_entry);
 #ifdef HAVE_UDP
@@ -9272,6 +9310,13 @@ static void tcp_inet_stop(ErlDrvData e)
  * will be freed through tcp_inet_stop later on. */
 static void tcp_desc_close(tcp_descriptor* desc)
 {
+#ifdef HAVE_SENDFILE
+    if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
+        desc->tcp_add_flags &= ~TCP_ADDF_SENDFILE;
+        close(desc->sendfile.dup_file_fd);
+    }
+#endif
+
     tcp_clear_input(desc);
     tcp_clear_output(desc);
 
@@ -9610,6 +9655,60 @@ static ErlDrvSSizeT tcp_inet_ctl(ErlDrvData e, unsigned int cmd,
 	    return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
 	}
     }
+
+    case TCP_REQ_SENDFILE: {
+#ifdef HAVE_SENDFILE
+        const ErlDrvSizeT required_len =
+            sizeof(desc->sendfile.dup_file_fd) +
+            sizeof(Uint64) * 2;
+
+        int raw_file_fd;
+
+        DEBUGF(("tcp_inet_ctl(%ld): SENDFILE\r\n", (long)desc->inet.port));
+
+        if (len != required_len) {
+            return ctl_error(EINVAL, rbuf, rsize);
+        } else if (!IS_CONNECTED(INETP(desc))) {
+            return ctl_error(ENOTCONN, rbuf, rsize);
+        }
+
+        sys_memcpy(&raw_file_fd, buf, sizeof(raw_file_fd));
+        buf += sizeof(raw_file_fd);
+
+        desc->sendfile.dup_file_fd = dup(raw_file_fd);
+
+        if(desc->sendfile.dup_file_fd == -1) {
+            return ctl_error(errno, rbuf, rsize);
+        }
+
+        desc->sendfile.offset = get_int64(buf);
+        buf += sizeof(Uint64);
+
+        desc->sendfile.length = get_int64(buf);
+        buf += sizeof(Uint64);
+
+        ASSERT(desc->sendfile.offset >= 0);
+        ASSERT(desc->sendfile.length >= 0);
+
+        desc->sendfile.ioq_skip = driver_sizeq(desc->inet.port);
+        desc->sendfile.bytes_sent = 0;
+
+        desc->inet.caller = driver_caller(desc->inet.port);
+        desc->tcp_add_flags |= TCP_ADDF_SENDFILE;
+
+        /* See if we can finish sending without selecting & rescheduling. */
+        tcp_inet_sendfile(desc);
+
+        if(desc->sendfile.length > 0) {
+            sock_select(INETP(desc), FD_WRITE, 1);
+        }
+
+        return ctl_reply(INET_REP_OK, NULL, 0, rbuf, rsize);
+#else
+        return ctl_error(ENOTSUP, rbuf, rsize);
+#endif
+    }
+
     default:
 	DEBUGF(("tcp_inet_ctl(%ld): %u\r\n", (long)desc->inet.port, cmd)); 
 	return inet_ctl(INETP(desc), cmd, buf, len, rbuf, rsize);
@@ -9749,12 +9848,27 @@ static void tcp_inet_commandv(ErlDrvData e, ErlIOVec* ev)
 static void tcp_inet_flush(ErlDrvData e)
 {
     tcp_descriptor* desc = (tcp_descriptor*)e;
-    if (!(desc->inet.event_mask & FD_WRITE)) {
-	/* Discard send queue to avoid hanging port (OTP-7615) */
-	tcp_clear_output(desc);
+    int discard_output;
+
+    /* Discard send queue to avoid hanging port (OTP-7615) */
+    discard_output = !(desc->inet.event_mask & FD_WRITE);
+
+    discard_output |= desc->tcp_add_flags & TCP_ADDF_LINGER_ZERO;
+
+#ifdef HAVE_SENDFILE
+    /* The old file driver aborted when it was stopped during sendfile, so
+     * we'll clear the flag and discard all output. */
+    if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
+        desc->tcp_add_flags &= ~TCP_ADDF_SENDFILE;
+        close(desc->sendfile.dup_file_fd);
+
+        discard_output = 1;
+    }
+#endif
+
+    if (discard_output) {
+        tcp_clear_output(desc);
     }
-    if (desc->tcp_add_flags & TCP_ADDF_LINGER_ZERO)
-	tcp_clear_output(desc);
 }
 
 static void tcp_inet_process_exit(ErlDrvData e, ErlDrvMonitor *monitorp) 
@@ -9816,6 +9930,12 @@ static int tcp_recv_closed(tcp_descriptor* desc)
 	set_busy_port(desc->inet.port, 0);
 	inet_reply_error_am(INETP(desc), am_closed);
 	DEBUGF(("tcp_recv_closed(%ld): busy reply 'closed'\r\n", port));
+    } else {
+        /* No blocking send op to reply to right now.
+         * If next op is a send, make sure it returns {error,closed}
+         * rather than {error,enotconn}.
+         */
+        desc->tcp_add_flags |= TCP_ADDF_DELAYED_CLOSE_SEND;
     }
     if (!desc->inet.active) {
 	/* We must cancel any timer here ! */
@@ -10649,7 +10769,9 @@ static int tcp_sendv(tcp_descriptor* desc, ErlIOVec* ev)
 	ev->size += h_len;
     }
 
-    if ((sz = driver_sizeq(ix)) > 0) {
+    sz = driver_sizeq(ix);
+
+    if ((desc->tcp_add_flags & TCP_ADDF_SENDFILE) || sz > 0) {
 	driver_enqv(ix, ev, 0);
 	if (sz+ev->size >= desc->high) {
 	    DEBUGF(("tcp_sendv(%ld): s=%d, sender forced busy\r\n",
@@ -10743,8 +10865,9 @@ static int tcp_send(tcp_descriptor* desc, char* ptr, ErlDrvSizeT len)
 
     inet_output_count(INETP(desc), len+h_len);
 
+    sz = driver_sizeq(ix);
 
-    if ((sz = driver_sizeq(ix)) > 0) {
+    if ((desc->tcp_add_flags & TCP_ADDF_SENDFILE) || sz > 0) {
 	if (h_len > 0)
 	    driver_enq(ix, buf, h_len);
 	driver_enq(ix, ptr, len);
@@ -10834,6 +10957,246 @@ static void tcp_inet_drv_input(ErlDrvData data, ErlDrvEvent event)
     (void)tcp_inet_input((tcp_descriptor*)data, (HANDLE)event);
 }
 
+#ifdef HAVE_SENDFILE
+static int tcp_sendfile_completed(tcp_descriptor* desc) {
+    ErlDrvTermData spec[LOAD_PORT_CNT + LOAD_TUPLE_CNT * 2 +
+        LOAD_ATOM_CNT * 2 + LOAD_UINT_CNT * 2];
+    Uint32 sent_low, sent_high;
+    int i;
+
+    desc->tcp_add_flags &= ~TCP_ADDF_SENDFILE;
+    close(desc->sendfile.dup_file_fd);
+
+    /* While we flushed the output queue prior to sending the file, we've
+     * deferred clearing busy status until now as there's no point in doing so
+     * while we still have a file to send.
+     *
+     * The watermark is checked since more data may have been added while we
+     * were sending the file. */
+
+    if (driver_sizeq(desc->inet.port) <= desc->low) {
+        if (IS_BUSY(INETP(desc))) {
+            desc->inet.caller = desc->inet.busy_caller;
+            desc->inet.state &= ~INET_F_BUSY;
+
+            set_busy_port(desc->inet.port, 0);
+
+            /* if we have a timer then cancel and send ok to client */
+            if (desc->busy_on_send) {
+                driver_cancel_timer(desc->inet.port);
+                desc->busy_on_send = 0;
+            }
+
+            inet_reply_ok(INETP(desc));
+        }
+    }
+
+    if (driver_sizeq(desc->inet.port) == 0) {
+        sock_select(INETP(desc), FD_WRITE, 0);
+        send_empty_out_q_msgs(INETP(desc));
+
+        if (desc->tcp_add_flags & TCP_ADDF_PENDING_SHUTDOWN) {
+            tcp_shutdown_async(desc);
+        }
+    }
+
+    sent_low = ((Uint64)desc->sendfile.bytes_sent >> 0) & 0xFFFFFFFF;
+    sent_high = ((Uint64)desc->sendfile.bytes_sent >> 32) & 0xFFFFFFFF;
+
+    i = LOAD_ATOM(spec, 0, am_sendfile);
+    i = LOAD_PORT(spec, i, desc->inet.dport);
+    i = LOAD_ATOM(spec, i, am_ok);
+    i = LOAD_UINT(spec, i, sent_low);
+    i = LOAD_UINT(spec, i, sent_high);
+    i = LOAD_TUPLE(spec, i, 3);
+    i = LOAD_TUPLE(spec, i, 3);
+
+    ASSERT(i == sizeof(spec)/sizeof(*spec));
+
+    return erl_drv_output_term(desc->inet.dport, spec, i);
+}
+
+static int tcp_sendfile_aborted(tcp_descriptor* desc, int socket_error) {
+    ErlDrvTermData spec[LOAD_PORT_CNT + LOAD_TUPLE_CNT * 2 + LOAD_ATOM_CNT * 3];
+    int i;
+
+    /* We don't clean up sendfile state here, as that's done in tcp_desc_close
+     * following normal error handling. All we do here is report the failure. */
+
+    i = LOAD_ATOM(spec, 0, am_sendfile);
+    i = LOAD_PORT(spec, i, desc->inet.dport);
+    i = LOAD_ATOM(spec, i, am_error);
+
+    switch (socket_error) {
+    case ECONNRESET:
+    case ENOTCONN:
+    case EPIPE:
+        i = LOAD_ATOM(spec, i, am_closed);
+        break;
+    default:
+        i = LOAD_ATOM(spec, i, error_atom(socket_error));
+    }
+
+    i = LOAD_TUPLE(spec, i, 2);
+    i = LOAD_TUPLE(spec, i, 3);
+
+    ASSERT(i == sizeof(spec)/sizeof(*spec));
+
+    return erl_drv_output_term(desc->inet.dport, spec, i);
+}
+
+static int tcp_inet_sendfile(tcp_descriptor* desc) {
+    ErlDrvPort ix = desc->inet.port;
+    int result = 0;
+    ssize_t n;
+
+    DEBUGF(("tcp_inet_sendfile(%ld) {s=%d\r\n", (long)ix, desc->inet.s));
+
+    /* If there was any data in the queue by the time sendfile was issued,
+     * we'll need to skip it first. Note that we don't clear busy status until
+     * we're finished sending the file. */
+    while (desc->sendfile.ioq_skip > 0) {
+        ssize_t bytes_to_send;
+        SysIOVec* iov;
+        int vsize;
+
+        ASSERT(driver_sizeq(ix) >= desc->sendfile.ioq_skip);
+
+        if ((iov = driver_peekq(ix, &vsize)) == NULL) {
+            ERTS_INTERNAL_ERROR("ioq empty when sendfile.ioq_skip > 0");
+        }
+
+        bytes_to_send = MIN(desc->sendfile.ioq_skip, iov[0].iov_len);
+        n = sock_send(desc->inet.s, iov[0].iov_base, bytes_to_send, 0);
+
+        if (!IS_SOCKET_ERROR(n)) {
+            desc->sendfile.ioq_skip -= n;
+            driver_deq(ix, n);
+        } else if (sock_errno() == ERRNO_BLOCK) {
+#ifdef __WIN32__
+            desc->inet.send_would_block = 1;
+#endif
+            goto done;
+        } else if (sock_errno() != EINTR) {
+            goto socket_error;
+        }
+    }
+
+    while (desc->sendfile.length > 0) {
+        /* For some reason the maximum ssize_t cannot be used as the max size.
+         * 1GB seems to work on all platforms */
+        const Sint64 SENDFILE_CHUNK_SIZE = ((1UL << 30) - 1);
+
+        ssize_t bytes_to_send = MIN(SENDFILE_CHUNK_SIZE, desc->sendfile.length);
+        off_t offset = desc->sendfile.offset;
+
+#if defined(__linux__)
+        n = sendfile(desc->inet.s, desc->sendfile.dup_file_fd, &offset,
+            bytes_to_send);
+#elif defined(__FreeBSD__) || defined(__DragonFly__) || defined(__DARWIN__)
+        {
+            off_t bytes_sent;
+            int error;
+
+    #if defined(__DARWIN__)
+            bytes_sent = bytes_to_send;
+
+            error = sendfile(desc->sendfile.dup_file_fd, desc->inet.s, offset,
+                &bytes_sent, NULL, 0);
+            n = bytes_sent;
+    #else
+            error = sendfile(desc->sendfile.dup_file_fd, desc->inet.s, offset,
+                bytes_to_send, NULL, &bytes_sent, 0);
+            n = bytes_sent;
+    #endif
+
+            if(error < 0) {
+                /* EAGAIN/EINTR report partial success by setting bytes_sent,
+                 * so we have to skip error handling if nonzero, and skip EOF
+                 * handling if zero, as it's possible that we didn't manage to
+                 * send anything at all before being interrupted by a
+                 * signal. */
+                if((errno != EAGAIN && errno != EINTR) || bytes_sent == 0) {
+                    n = -1;
+                }
+            }
+        }
+#elif defined(__sun) && defined(__SVR4) && defined(HAVE_SENDFILEV)
+        {
+            sendfilevec_t sfvec[1];
+            size_t bytes_sent;
+            ssize_t error;
+
+            sfvec[0].sfv_fd = desc->sendfile.dup_file_fd;
+            sfvec[0].sfv_len = bytes_to_send;
+            sfvec[0].sfv_off = offset;
+            sfvec[0].sfv_flag = 0;
+
+            error = sendfilev(desc->inet.s, sfvec, 1, &bytes_sent);
+            n = bytes_sent;
+
+            if(error < 0) {
+                if(errno == EINVAL) {
+                    /* On some solaris versions (I've seen it on SunOS 5.10),
+                     * using a sfv_len larger than the filesize will result in
+                     * a (-1 && errno == EINVAL). We translate this to a
+                     * successful send of the data.*/
+                } else {
+                    /* EAGAIN/EINTR behavior is identical to *BSD. */
+                    if((errno != EAGAIN && errno != EINTR) || bytes_sent == 0) {
+                        n = -1;
+                    }
+                }
+            }
+        }
+#else
+        #error "Unsupported sendfile syscall; update configure test."
+#endif
+
+        if (n > 0) {
+            desc->sendfile.bytes_sent += n;
+            desc->sendfile.offset += n;
+            desc->sendfile.length -= n;
+        } else if (n == 0) {
+            /* EOF. */
+            desc->sendfile.length = 0;
+            break;
+        } else if (IS_SOCKET_ERROR(n) && sock_errno() != EINTR) {
+            if (sock_errno() != ERRNO_BLOCK) {
+                goto socket_error;
+            }
+
+#ifdef __WIN32__
+            desc->inet.send_would_block = 1;
+#endif
+            break;
+        }
+    }
+
+    if (desc->sendfile.length == 0) {
+        tcp_sendfile_completed(desc);
+    }
+
+    goto done;
+
+socket_error: {
+        int socket_errno = sock_errno();
+
+        DEBUGF(("tcp_inet_sendfile(%ld): send errno = %d (errno %d)\r\n",
+            (long)desc->inet.port, socket_errno, errno));
+
+        result = tcp_send_error(desc, socket_errno);
+        tcp_sendfile_aborted(desc, socket_errno);
+
+        goto done;
+    }
+
+done:
+    DEBUGF(("tcp_inet_sendfile(%ld) }\r\n", (long)desc->inet.port));
+    return result;
+}
+#endif /* HAVE_SENDFILE */
+
 /* socket ready for ouput:
 ** 1. INET_STATE_CONNECTING => non block connect ?
 ** 2. INET_STATE_CONNECTED  => write output
@@ -10894,7 +11257,14 @@ static int tcp_inet_output(tcp_descriptor* desc, HANDLE event)
 	async_ok(INETP(desc));
     }
     else if (IS_CONNECTED(INETP(desc))) {
-	for (;;) {
+
+#ifdef HAVE_SENDFILE
+        if(desc->tcp_add_flags & TCP_ADDF_SENDFILE) {
+            return tcp_inet_sendfile(desc);
+        }
+#endif
+
+        for (;;) {
 	    int vsize;
 	    ssize_t n;
 	    SysIOVec* iov;
@@ -11300,24 +11670,20 @@ static ErlDrvSSizeT packet_inet_ctl(ErlDrvData e, unsigned int cmd, char* buf,
 		 (desc->sfamily, &remote, &buf, &len)) != NULL)
 	        return ctl_xerror(xerror, rbuf, rsize);
 	
-	    sock_select(desc, FD_CONNECT, 1);
 	    code = sock_connect(desc->s, &remote.sa, len);
 
 	    if (IS_SOCKET_ERROR(code) && (sock_errno() == EINPROGRESS)) {
 		/* XXX: Unix only -- WinSock would have a different cond! */
-		desc->state = INET_STATE_CONNECTING;
 		if (timeout != INET_INFINITY)
 		    driver_set_timer(desc->port, timeout);
 		enq_async(desc, tbuf, INET_REQ_CONNECT);
+		async_ok(desc);
 	    }
 	    else if (code == 0) { /* OK we are connected */
-		sock_select(desc, FD_CONNECT, 0);
-		desc->state = INET_STATE_CONNECTED;
 		enq_async(desc, tbuf, INET_REQ_CONNECT);
 		async_ok(desc);
 	    }
 	    else {
-		sock_select(desc, FD_CONNECT, 0);
 		return ctl_error(sock_errno(), rbuf, rsize);
 	    }
 	    return ctl_reply(INET_REP_OK, tbuf, 2, rbuf, rsize);
@@ -11832,77 +12198,6 @@ static int packet_inet_input(udp_descriptor* udesc, HANDLE event)
     return count;
 }
 
-static void packet_inet_drv_output(ErlDrvData e, ErlDrvEvent event)
-{
-    (void)  packet_inet_output((udp_descriptor*)e, (HANDLE)event);
-}
-
-/* UDP/SCTP socket ready for output:
-**	This is a Back-End for Non-Block SCTP Connect (INET_STATE_CONNECTING)
-*/
-static int packet_inet_output(udp_descriptor* udesc, HANDLE event)
-{
-    inet_descriptor* desc = INETP(udesc);
-    int ret = 0;
-    ErlDrvPort ix = desc->port;
-
-    DEBUGF(("packet_inet_output(%ld) {s=%d\r\n", 
-	    (long)desc->port, desc->s));
-
-    if (desc->state == INET_STATE_CONNECTING) {
-	sock_select(desc, FD_CONNECT, 0);
-
-	driver_cancel_timer(ix);  /* posssibly cancel a timer */
-#ifndef __WIN32__
-	/*
-	 * XXX This is strange.  This *should* work on Windows NT too,
-	 * but doesn't.  An bug in Winsock 2.0 for Windows NT?
-	 *
-	 * See "Unix Netwok Programming", W.R.Stevens, p 412 for a
-	 * discussion about Unix portability and non blocking connect.
-	 */
-
-#ifndef SO_ERROR
-	{
-	    int sz = sizeof(desc->remote);
-	    int code = sock_peer(desc->s,
-				 (struct sockaddr*) &desc->remote, &sz);
-
-	    if (IS_SOCKET_ERROR(code)) {
-		desc->state = INET_STATE_OPEN;  /* restore state */
-		ret =  async_error(desc, sock_errno());
-		goto done;
-	    }
-	}
-#else
-	{
-	    int error = 0;	/* Has to be initiated, we check it */
-	    unsigned int sz = sizeof(error);   /* even if we get -1 */
-	    int code = sock_getopt(desc->s, SOL_SOCKET, SO_ERROR, 
-				   (void *)&error, &sz);
-
-	    if ((code < 0) || error) {
-		desc->state = INET_STATE_OPEN;  /* restore state */
-		ret = async_error(desc, error);
-		goto done;
-	    }
-	}
-#endif /* SO_ERROR */
-#endif /* !__WIN32__ */
-
-	desc->state = INET_STATE_CONNECTED;
-	async_ok(desc);
-    }
-    else {
-	sock_select(desc,FD_CONNECT,0);
-
-	DEBUGF(("packet_inet_output(%ld): bad state: %04x\r\n", 
-		(long)desc->port, desc->state));
-    }
- done:
-    DEBUGF(("packet_inet_output(%ld) }\r\n", (long)desc->port));
-    return ret;
-}
 #endif
 
 /*---------------------------------------------------------------------------*/
diff --git a/erts/emulator/drivers/unix/ttsl_drv.c b/erts/emulator/drivers/unix/ttsl_drv.c
index 2a508b02eb..7355df6059 100644
--- a/erts/emulator/drivers/unix/ttsl_drv.c
+++ b/erts/emulator/drivers/unix/ttsl_drv.c
@@ -892,8 +892,8 @@ static void ttysl_from_tty(ErlDrvData ttysl_data, ErlDrvEvent fd)
 		tpos = 0;
 	    }
 	}
-    } else {
-        DEBUGLOG(("ttysl_from_tty: driver failure in read(%d,..) = %d\n", (int)(SWord)fd, i)); 
+    } else if (errno != EAGAIN && errno != EWOULDBLOCK) {
+        DEBUGLOG(("ttysl_from_tty: driver failure in read(%d,..) = %d (errno = %d)\n", (int)(SWord)fd, i, errno));
 	driver_failure(ttysl_port, -1);
     }
 }
diff --git a/erts/emulator/drivers/unix/unix_efile.c b/erts/emulator/drivers/unix/unix_efile.c
deleted file mode 100644
index 33e4d75ef7..0000000000
--- a/erts/emulator/drivers/unix/unix_efile.c
+++ /dev/null
@@ -1,1102 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 1997-2017. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-/*
- * Purpose: Provides file and directory operations for Unix.
- */
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-#if defined(HAVE_POSIX_FALLOCATE) && !defined(__sun) && !defined(__sun__)
-#define _XOPEN_SOURCE 600
-#endif
-#if !defined(_GNU_SOURCE) && defined(HAVE_LINUX_FALLOC_H)
-#define _GNU_SOURCE
-#endif
-#include "sys.h"
-#include "erl_driver.h"
-#include "erl_efile.h"
-#include <utime.h>
-#ifdef HAVE_UNISTD_H
-#include <unistd.h>
-#endif
-#ifdef HAVE_SYS_UIO_H
-#include <sys/types.h>
-#include <sys/uio.h>
-#if defined(HAVE_SENDFILE) && (defined(__FreeBSD__) || defined(__DragonFly__))
-/* Need to define __BSD_VISIBLE in order to expose prototype of sendfile */
-#define __BSD_VISIBLE 1
-#include <sys/socket.h>
-#endif
-#endif
-#if defined(HAVE_SENDFILE) && (defined(__linux__) || (defined(__sun) && defined(__SVR4)))
-#include <sys/sendfile.h>
-#endif
-
-#if defined(__APPLE__) && defined(__MACH__) && !defined(__DARWIN__)
-#define __DARWIN__ 1
-#endif
-
-#if defined(__DARWIN__) || defined(HAVE_LINUX_FALLOC_H) || defined(HAVE_POSIX_FALLOCATE)
-#include <fcntl.h>
-#endif
-
-#ifdef HAVE_LINUX_FALLOC_H
-#include <linux/falloc.h>
-#endif
-
-#ifdef SUNOS4
-#  define getcwd(buf, size) getwd(buf)
-#endif
-
-/* Find a definition of MAXIOV, that is used in the code later. */
-#if defined IOV_MAX
-#define MAXIOV IOV_MAX
-#elif defined UIO_MAXIOV
-#define MAXIOV UIO_MAXIOV
-#else
-#define MAXIOV 16
-#endif
-
-
-/*
- * Macros for testing file types.
- */
-
-#define ISDIR(st) (S_ISDIR((st).st_mode))
-#define ISREG(st) (S_ISREG((st).st_mode))
-#define ISDEV(st) (S_ISCHR((st).st_mode) || S_ISBLK((st).st_mode))
-#define ISLNK(st) (S_ISLNK((st).st_mode))
-#ifdef NO_UMASK
-#define FILE_MODE 0644
-#define DIR_MODE  0755
-#else
-#define FILE_MODE 0666
-#define DIR_MODE  0777
-#endif
-
-#define IS_DOT_OR_DOTDOT(s) \
-    (s[0] == '.' && (s[1] == '\0' || (s[1] == '.' && s[2] == '\0')))
-
-static int check_error(int result, Efile_error* errInfo);
-
-static int
-check_error(int result, Efile_error *errInfo)
-{
-    if (result < 0) {
-	errInfo->posix_errno = errInfo->os_errno = errno;
-	return 0;
-    }
-    return 1;
-}
-
-int
-efile_init() {
-   return 1;
-}
-
-int
-efile_mkdir(Efile_error* errInfo,	/* Where to return error codes. */
-	    char* name)			/* Name of directory to create. */
-{
-#ifdef NO_MKDIR_MODE
-    return check_error(mkdir(name), errInfo);
-#else
-    return check_error(mkdir(name, DIR_MODE), errInfo);
-#endif
-}
-
-int
-efile_rmdir(Efile_error* errInfo,	/* Where to return error codes. */
-	    char* name)			/* Name of directory to delete. */
-{
-    if (rmdir(name) == 0) {
-	return 1;
-    }
-    if (errno == ENOTEMPTY) {
-	errno = EEXIST;
-    }
-    if (errno == EEXIST) {
-	int saved_errno = errno;
-	struct stat file_stat;
-	struct stat cwd_stat;
-
-	/*
-	 *  The error code might be wrong if this is the current directory.
-	 */
-
-	if (stat(name, &file_stat) == 0 && stat(".", &cwd_stat) == 0 &&
-	    file_stat.st_ino == cwd_stat.st_ino &&
-	    file_stat.st_dev == cwd_stat.st_dev) {
-	    saved_errno = EINVAL;
-	}
-	errno = saved_errno;
-    }
-    return check_error(-1, errInfo);
-}
-
-int
-efile_delete_file(Efile_error* errInfo,	/* Where to return error codes. */
-		  char* name)		/* Name of file to delete. */
-{
-    if (unlink(name) == 0) {
-	return 1;
-    }
-    if (errno == EISDIR) {	/* Linux sets the wrong error code. */
-	errno = EPERM;
-    }
-    return check_error(-1, errInfo);
-}
-
-/*
- *---------------------------------------------------------------------------
- *
- *      Changes the name of an existing file or directory, from src to dst.
- *	If src and dst refer to the same file or directory, does nothing
- *	and returns success.  Otherwise if dst already exists, it will be
- *	deleted and replaced by src subject to the following conditions:
- *	    If src is a directory, dst may be an empty directory.
- *	    If src is a file, dst may be a file.
- *	In any other situation where dst already exists, the rename will
- *	fail.  
- *
- * Results:
- *	If the directory was successfully created, returns 1.
- *	Otherwise the return value is 0 and errno is set to
- *	indicate the error.  Some possible values for errno are:
- *
- *	EACCES:     src or dst parent directory can't be read and/or written.
- *	EEXIST:	    dst is a non-empty directory.
- *	EINVAL:	    src is a root directory or dst is a subdirectory of src.
- *	EISDIR:	    dst is a directory, but src is not.
- *	ENOENT:	    src doesn't exist, or src or dst is "".
- *	ENOTDIR:    src is a directory, but dst is not.  
- *	EXDEV:	    src and dst are on different filesystems.
- *	
- * Side effects:
- *	The implementation of rename may allow cross-filesystem renames,
- *	but the caller should be prepared to emulate it with copy and
- *	delete if errno is EXDEV.
- *
- *---------------------------------------------------------------------------
- */
-
-int
-efile_rename(Efile_error* errInfo,	/* Where to return error codes. */
-	     char* src,		        /* Original name. */
-	     char* dst)			/* New name. */
-{
-    if (rename(src, dst) == 0) {
-	return 1;
-    }
-    if (errno == ENOTEMPTY) {
-	errno = EEXIST;
-    }
-#if defined (sparc)
-    /*
-     * SunOS 4.1.4 reports overwriting a non-empty directory with a
-     * directory as EINVAL instead of EEXIST (first rule out the correct
-     * EINVAL result code for moving a directory into itself).  Must be
-     * conditionally compiled because realpath() is only defined on SunOS.
-     */
-
-    if (errno == EINVAL) {
-	char srcPath[MAXPATHLEN], dstPath[MAXPATHLEN];
-	DIR *dirPtr;
-	struct dirent *dirEntPtr;
-
-#ifdef PURIFY
-	memset(srcPath, '\0', sizeof(srcPath));
-	memset(dstPath, '\0', sizeof(dstPath));
-#endif
-
-	if ((realpath(src, srcPath) != NULL)
-		&& (realpath(dst, dstPath) != NULL)
-		&& (strncmp(srcPath, dstPath, strlen(srcPath)) != 0)) {
-	    dirPtr = opendir(dst);
-	    if (dirPtr != NULL) {
-		while ((dirEntPtr = readdir(dirPtr)) != NULL) {
-		    if ((strcmp(dirEntPtr->d_name, ".") != 0) &&
-			    (strcmp(dirEntPtr->d_name, "..") != 0)) {
-			errno = EEXIST;
-			closedir(dirPtr);
-			return check_error(-1, errInfo);
-		    }
-		}
-		closedir(dirPtr);
-	    }
-	}
-	errno = EINVAL;
-    }
-#endif	/* sparc */
-
-    if (strcmp(src, "/") == 0) {
-	/*
-	 * Alpha reports renaming / as EBUSY and Linux reports it as EACCES,
-	 * instead of EINVAL.
-	 */
-	 
-	errno = EINVAL;
-    }
-
-    /*
-     * DEC Alpha OSF1 V3.0 returns EACCES when attempting to move a
-     * file across filesystems and the parent directory of that file is
-     * not writable.  Most other systems return EXDEV.  Does nothing to
-     * correct this behavior.
-     */
-
-    return check_error(-1, errInfo);
-}
-
-int
-efile_chdir(Efile_error* errInfo,   /* Where to return error codes. */
-	    char* name)		    /* Name of directory to make current. */
-{
-    return check_error(chdir(name), errInfo);
-}
-
-
-int
-efile_getdcwd(Efile_error* errInfo,	/* Where to return error codes. */
-	      int drive,		/* 0 - current, 1 - A, 2 - B etc. */
-	      char* buffer,		/* Where to return the current 
-					   directory. */
-	      size_t size)		/* Size of buffer. */
-{
-    if (drive == 0) {
-	if (getcwd(buffer, size) == NULL)
-	    return check_error(-1, errInfo);
-	
-#ifdef SIMSPARCSOLARIS
-	/* We get "host:" prepended to the dirname - remove!. */
-	{
-	  int i = 0;
-	  int j = 0;
-	  while ((buffer[i] != ':') && (buffer[i] != '\0')) i++;
-	  if (buffer[i] == ':') {
-	    i++;
-	    while ((buffer[j++] = buffer[i++]) != '\0');
-	  }
-	}
-#endif	      
-	return 1;
-    }
-
-    /*
-     * Drives other than 0 is not supported on Unix.
-     */
-
-    errno = ENOTSUP;
-    return check_error(-1, errInfo);
-}
-
-int
-efile_readdir(Efile_error* errInfo,	/* Where to return error codes. */
-	      char* name,		/* Name of directory to open. */
-	      EFILE_DIR_HANDLE* p_dir_handle,	/* Pointer to directory 
-						   handle of
-						   open directory.*/
-	      char* buffer,		/* Pointer to buffer for 
-					   one filename. */
-	      size_t *size)		/* in-out Size of buffer, length
-					   of name. */
-{
-    DIR *dp;			/* Pointer to directory structure. */
-    struct dirent* dirp;	/* Pointer to directory entry. */
-
-    /*
-     * If this is the first call, we must open the directory.
-     */
-
-    if (*p_dir_handle == NULL) {
-	dp = opendir(name);
-	if (dp == NULL)
-	    return check_error(-1, errInfo);
-	*p_dir_handle = (EFILE_DIR_HANDLE) dp;
-    }
-
-    /*
-     * Retrieve the name of the next file using the directory handle.
-     */
-
-    dp = *((DIR **)((void *)p_dir_handle));
-    for (;;) {
-	dirp = readdir(dp);
-	if (dirp == NULL) {
-	    closedir(dp);
-	    return 0;
-	}
-	if (IS_DOT_OR_DOTDOT(dirp->d_name))
-	    continue;
-	buffer[0] = '\0';
-	strncat(buffer, dirp->d_name, (*size)-1);
-	*size = strlen(dirp->d_name);
-	return 1;
-    }
-}
-
-int
-efile_openfile(Efile_error* errInfo,	/* Where to return error codes. */
-	       char* name,		/* Name of directory to open. */
-	       int flags,		/* Flags to user for opening. */
-	       int* pfd,		/* Where to store the file 
-					   descriptor. */
-	       Sint64 *pSize)		/* Where to store the size of the 
-					   file. */
-{
-    struct stat statbuf;
-    int fd;
-    int mode;			/* Open mode. */
-
-    switch (flags & (EFILE_MODE_READ|EFILE_MODE_WRITE)) {
-    case EFILE_MODE_READ:
-	mode = O_RDONLY;
-	break;
-    case EFILE_MODE_WRITE:
-	if (flags & EFILE_NO_TRUNCATE)
-	    mode = O_WRONLY | O_CREAT;
-	else
-	    mode = O_WRONLY | O_CREAT | O_TRUNC;
-	break;
-    case EFILE_MODE_READ_WRITE:
-	mode = O_RDWR | O_CREAT;
-	break;
-    default:
-	errno = EINVAL;
-	return check_error(-1, errInfo);
-    }
-
-    if (flags & EFILE_MODE_APPEND) {
-	mode &= ~O_TRUNC;
-	mode |= O_APPEND;
-    }
-    if (flags & EFILE_MODE_EXCL) {
-	mode |= O_EXCL;
-    }
-    if (flags & EFILE_MODE_SYNC) {
-#ifdef O_SYNC
-	mode |= O_SYNC;
-#else
-	errno = ENOTSUP;
-	return check_error(-1, errInfo);
-#endif
-    }
-
-#ifdef HAVE_FSTAT
-    while (((fd = open(name, mode, FILE_MODE)) < 0) && (errno == EINTR));
-    if (!check_error(fd, errInfo)) return 0;
-#endif
-
-    if (
-#ifdef HAVE_FSTAT
-        fstat(fd, &statbuf) < 0
-#else
-        stat(name, &statbuf) < 0
-#endif
-        ) {
-        /* statbuf is undefined: if the caller depends on it,
-           i.e. invoke_read_file(), fail the call immediately */
-        if (pSize && flags == EFILE_MODE_READ) {
-            check_error(-1, errInfo);
-#ifdef HAVE_FSTAT
-            efile_closefile(fd);
-#endif
-            return 0;
-        }
-    }
-    else if (! ISREG(statbuf)) {
-        struct stat nullstatbuf;
-	/*
-	 * For UNIX only, here is some ugly code to allow
-	 * /dev/null to be opened as a file.
-	 */
-        if ( (stat("/dev/null", &nullstatbuf) < 0)
-             || (statbuf.st_ino != nullstatbuf.st_ino)
-             || (statbuf.st_dev != nullstatbuf.st_dev) ) {
-#ifdef HAVE_FSTAT
-            efile_closefile(fd);
-#endif
-	    errno = EISDIR;
-	    return check_error(-1, errInfo);
-        }
-    }
-
-#ifndef HAVE_FSTAT
-    while (((fd = open(name, mode, FILE_MODE)) < 0) && (errno == EINTR));
-    if (!check_error(fd, errInfo)) return 0;
-#endif
-
-    *pfd = fd;
-    if (pSize) *pSize = statbuf.st_size;
-    return 1;
-}
-
-int 
-efile_may_openfile(Efile_error* errInfo, char *name) {
-    struct stat statbuf;	/* Information about the file */
-    int result;
-    
-    result = stat(name, &statbuf);
-    if (!check_error(result, errInfo))
-	return 0;
-    if (!ISREG(statbuf)) {
-	errno = EISDIR;
-	return check_error(-1, errInfo);
-    }
-    return 1;
-}
-
-void
-efile_closefile(int fd)
-{
-    close(fd);
-}
-
-int
-efile_fdatasync(Efile_error *errInfo, /* Where to return error codes. */
-	    int fd)               /* File descriptor for file to sync data. */
-{
-#if defined(HAVE_FDATASYNC) && !defined(__DARWIN__)
-    return check_error(fdatasync(fd), errInfo);
-#else
-    return efile_fsync(errInfo, fd);
-#endif
-}
-
-int
-efile_fsync(Efile_error *errInfo, /* Where to return error codes. */
-	    int fd)               /* File descriptor for file to sync. */
-{
-#ifdef NO_FSYNC
-  undefined fsync /* XXX: Really? */
-#else
-#if defined(__DARWIN__) && defined(F_FULLFSYNC)
-    return check_error(fcntl(fd, F_FULLFSYNC), errInfo);
-#else
-    return check_error(fsync(fd), errInfo);
-#endif /* __DARWIN__ */
-#endif /* NO_FSYNC */
-}
-
-int
-efile_fileinfo(Efile_error* errInfo, Efile_info* pInfo, 
-	       char* name, int info_for_link)
-{
-    struct stat statbuf;	/* Information about the file */
-    int result;
-
-    if (info_for_link) {
-	result = lstat(name, &statbuf);
-    } else {
-	result = stat(name, &statbuf);
-    }	
-    if (!check_error(result, errInfo)) {
-	return 0;
-    }
-
-#if SIZEOF_OFF_T == 4
-    pInfo->size_high = 0;
-#else
-    pInfo->size_high = (Uint32)(statbuf.st_size >> 32);
-#endif
-    pInfo->size_low = (Uint32)statbuf.st_size;
-
-#ifdef NO_ACCESS
-    /* Just look at read/write access for owner. */
-
-    pInfo->access = ((statbuf.st_mode >> 6) & 07) >> 1;
-
-#else
-    pInfo->access = FA_NONE;
-    if (access(name, R_OK) == 0)
-	pInfo->access |= FA_READ;
-    if (access(name, W_OK) == 0)
-	pInfo->access |= FA_WRITE;
-
-#endif	
-
-    if (ISDEV(statbuf))
-	pInfo->type = FT_DEVICE;
-    else if (ISDIR(statbuf))
-	pInfo->type = FT_DIRECTORY;
-    else if (ISREG(statbuf))
-	pInfo->type = FT_REGULAR;
-    else if (ISLNK(statbuf))
-	pInfo->type = FT_SYMLINK;
-    else
-	pInfo->type = FT_OTHER;
-
-    pInfo->accessTime   = (Sint64)statbuf.st_atime;
-    pInfo->modifyTime   = (Sint64)statbuf.st_mtime;
-    pInfo->cTime        = (Sint64)statbuf.st_ctime;
-
-    pInfo->mode         = statbuf.st_mode;
-    pInfo->links        = statbuf.st_nlink;
-    pInfo->major_device = statbuf.st_dev;
-    pInfo->minor_device = statbuf.st_rdev;
-    pInfo->inode        = statbuf.st_ino;
-    pInfo->uid          = statbuf.st_uid;
-    pInfo->gid          = statbuf.st_gid;
-
-    return 1;
-}
-
-int
-efile_write_info(Efile_error *errInfo, Efile_info *pInfo, char *name)
-{
-    struct utimbuf tval;
-
-    /*
-     * On some systems chown will always fail for a non-root user unless
-     * POSIX_CHOWN_RESTRICTED is not set.  Others will succeed as long as 
-     * you don't try to chown a file to someone besides youself.
-     */
-
-    if (chown(name, pInfo->uid, pInfo->gid) && errno != EPERM) {
-	return check_error(-1, errInfo);
-    }
-
-    if (pInfo->mode != -1) {
-	mode_t newMode = pInfo->mode & (S_ISUID | S_ISGID |
-					S_IRWXU | S_IRWXG | S_IRWXO);
-	if (chmod(name, newMode)) {
-	    newMode &= ~(S_ISUID | S_ISGID);
-	    if (chmod(name, newMode)) {
-		return check_error(-1, errInfo);
-	    }
-	}
-    }
-
-    tval.actime  = (time_t)pInfo->accessTime;
-    tval.modtime = (time_t)pInfo->modifyTime;
-
-    return check_error(utime(name, &tval), errInfo);
-}
-
-
-int
-efile_write(Efile_error* errInfo,	/* Where to return error codes. */
-	    int flags,			/* Flags given when file was 
-					   opened. */
-	    int fd,			/* File descriptor to write to. */
-	    char* buf,			/* Buffer to write. */
-	    size_t count)		/* Number of bytes to write. */
-{
-    ssize_t written;			/* Bytes written in last operation. */
-
-    while (count > 0) {
-	if ((written = write(fd, buf, count)) < 0) {
-	    if (errno != EINTR)
-		return check_error(-1, errInfo);
-	    else
-		written = 0;
-	}
-	ASSERT(written <= count);
-	buf += written;
-	count -= written;
-    }
-    return 1;
-}
-
-int
-efile_writev(Efile_error* errInfo,   /* Where to return error codes */
-	     int flags,              /* Flags given when file was
-				      * opened */
-	     int fd,                 /* File descriptor to write to */
-	     SysIOVec* iov,          /* Vector of buffer structs.
-				      * The structs may be changed i.e.
-				      * due to incomplete writes */
-	     int iovcnt)             /* Number of structs in vector */
-{
-    int cnt = 0;                     /* Buffers so far written */
-
-    ASSERT(iovcnt >= 0);
-    
-    while (cnt < iovcnt) {
-	if ((! iov[cnt].iov_base) || (iov[cnt].iov_len <= 0)) {
-	    /* Empty buffer - skip */
-	    cnt++;
-	} else { /* Non-empty buffer */
-	    ssize_t w;                   /* Bytes written in this call */
-#ifdef HAVE_WRITEV
-	    int b = iovcnt - cnt;        /* Buffers to write */
-	    /* Use as many buffers as MAXIOV allows */
-	    if (b > MAXIOV)
-		b = MAXIOV;
-	    if (b > 1) {
-		do {
-		    w = writev(fd, &iov[cnt], b);
-		} while (w < 0 && errno == EINTR);
-		if (w < 0 && errno == EINVAL) {
-		    goto single_write;
-		}
-	    } else
-	    single_write:
-		/* Degenerated io vector - use regular write */
-#endif
-		{
-		    do {
-			size_t iov_len = iov[cnt].iov_len;
-			size_t limit = 1024*1024*1024; /* 1GB */
-			if (iov_len > limit) {
-			    iov_len = limit;
-			}
-			w = write(fd, iov[cnt].iov_base, iov_len);
-		    } while (w < 0 && errno == EINTR);
-		    ASSERT(w <= iov[cnt].iov_len ||
-			   (w == -1 && errno != EINTR));
-		}
-	    if (w < 0) return check_error(-1, errInfo);
-	    /* Move forward to next buffer to write */
-	    for (; cnt < iovcnt && w > 0; cnt++) {
-		if (iov[cnt].iov_base && iov[cnt].iov_len > 0) {
-		    if (w < iov[cnt].iov_len) {
-			/* Adjust the buffer for next write */
-			iov[cnt].iov_len -= w;
-			iov[cnt].iov_base = ((char *)iov[cnt].iov_base) + w;
-			w = 0;
-			break;
-		    } else {
-			w -= iov[cnt].iov_len;
-		    }
-		}
-	    }
-	    ASSERT(w == 0);
-	} /* else Non-empty buffer */
-    } /* while (cnt< iovcnt) */
-    return 1;
-}
-
-int
-efile_read(Efile_error* errInfo,     /* Where to return error codes. */
-	   int flags,		     /* Flags given when file was opened. */
-	   int fd,		     /* File descriptor to read from. */
-	   char* buf,		     /* Buffer to read into. */
-	   size_t count,	     /* Number of bytes to read. */
-	   size_t *pBytesRead)	     /* Where to return number of 
-					bytes read. */
-{
-    ssize_t n;
-
-    for (;;) {
-	if ((n = read(fd, buf, count)) >= 0)
-	    break;
-	else if (errno != EINTR)
-	    return check_error(-1, errInfo);
-    }
-    *pBytesRead = (size_t) n;
-    return 1;
-}
-
-
-/* pread() and pwrite()                                                   */
-/* Some unix systems, notably Solaris has these syscalls                  */
-/* It is especially nice for i.e. the dets module to have support         */
-/* for this, even if the underlying OS dosn't support it, it is           */
-/* reasonably easy to work around by first calling seek, and then         */
-/* calling read().                                                        */
-/* This later strategy however changes the file pointer, which pread()    */
-/* does not do. We choose to ignore this and say that the location        */
-/* of the file pointer is undefined after a call to any of the p functions*/
-
-
-int
-efile_pread(Efile_error* errInfo,     /* Where to return error codes. */
-	    int fd,		      /* File descriptor to read from. */
-	    Sint64 offset,            /* Offset in bytes from BOF. */
-	    char* buf,		      /* Buffer to read into. */
-	    size_t count,	      /* Number of bytes to read. */
-	    size_t *pBytesRead)	      /* Where to return 
-					 number of bytes read. */
-{
-#if defined(HAVE_PREAD) && defined(HAVE_PWRITE)
-    ssize_t n;
-    off_t off = (off_t) offset;
-    if (off != offset) {
-	errno = EINVAL;
-	return check_error(-1, errInfo);
-    }
-    for (;;) {
-	if ((n = pread(fd, buf, count, offset)) >= 0)
-	    break;
-	else if (errno != EINTR)
-	    return check_error(-1, errInfo);
-    }
-    *pBytesRead = (size_t) n;
-    return 1;
-#else
-    {
-	int res = efile_seek(errInfo, fd, offset, EFILE_SEEK_SET, NULL);
-	if (res) {
-	    return efile_read(errInfo, 0, fd, buf, count, pBytesRead);
-	} else {
-	    return res;
-	}
-    }
-#endif
-}
-
-
-
-int
-efile_pwrite(Efile_error* errInfo,  /* Where to return error codes. */
-	     int fd,		    /* File descriptor to write to. */
-	     char* buf,		    /* Buffer to write. */
-	     size_t count,	    /* Number of bytes to write. */
-	     Sint64 offset)	    /* where to write it */
-{ 
-#if defined(HAVE_PREAD) && defined(HAVE_PWRITE)
-    ssize_t written;		    /* Bytes written in last operation. */
-    off_t off = (off_t) offset;
-    if (off != offset) {
-	errno = EINVAL;
-	return check_error(-1, errInfo);
-    }
-    
-    while (count > 0) {
-	if ((written = pwrite(fd, buf, count, offset)) < 0) {
-	    if (errno != EINTR)
-		return check_error(-1, errInfo);
-	    else
-		written = 0;
-	}
-	ASSERT(written <= count);
-	buf += written;
-	count -= written;
-	offset += written;
-    }
-    return 1;
-#else  /* For unix systems that don't support pread() and pwrite() */    
-    {
-	int res = efile_seek(errInfo, fd, offset, EFILE_SEEK_SET, NULL);
-	
-	if (res) {
-	    return efile_write(errInfo, 0, fd, buf, count);
-	} else {
-	    return res;
-	}
-    }
-#endif
-}
-
-
-int
-efile_seek(Efile_error* errInfo,      /* Where to return error codes. */
-	   int fd,                    /* File descriptor to do the seek on. */
-	   Sint64 offset,             /* Offset in bytes from the given 
-					 origin. */ 
-	   int origin,                /* Origin of seek (SEEK_SET, SEEK_CUR,
-				         SEEK_END). */ 
-	   Sint64 *new_location)      /* Resulting new location in file. */
-{
-    off_t off, result;
-
-    switch (origin) {
-    case EFILE_SEEK_SET: origin = SEEK_SET; break;
-    case EFILE_SEEK_CUR: origin = SEEK_CUR; break;
-    case EFILE_SEEK_END: origin = SEEK_END; break;
-    default:
-	errno = EINVAL;
-	return check_error(-1, errInfo);
-    }
-    off = (off_t) offset;
-    if (off != offset) {
-	errno = EINVAL;
-	return check_error(-1, errInfo);
-    }
-    
-    errno = 0;
-    result = lseek(fd, off, origin);
-
-    /*
-     * Note that the man page for lseek (on SunOs 5) says:
-     * 
-     * "if fildes is a remote file  descriptor  and  offset  is
-     * negative,  lseek()  returns  the  file pointer even if it is
-     * negative."
-     */
-
-    if (result < 0 && errno == 0)
-	errno = EINVAL;
-    if (result < 0)
-	return check_error(-1, errInfo);
-    if (new_location) {
-	*new_location = result;
-    }
-    return 1;
-}
-
-
-int
-efile_truncate_file(Efile_error* errInfo, int *fd, int flags)
-{
-#ifndef NO_FTRUNCATE
-    off_t offset;
-
-    return check_error((offset = lseek(*fd, 0, 1)) >= 0 &&
-		       ftruncate(*fd, offset) == 0 ? 1 : -1,
-		       errInfo);
-#else
-    return 1;
-#endif
-}
-
-int
-efile_readlink(Efile_error* errInfo, char* name, char* buffer, size_t size)
-{
-    int len;
-    ASSERT(size > 0);
-    len = readlink(name, buffer, size-1);
-    if (len == -1) {
-	return check_error(-1, errInfo);
-    }
-    buffer[len] = '\0';
-    return 1;
-}
-
-int
-efile_altname(Efile_error* errInfo, char* name, char* buffer, size_t size)
-{
-    errno = ENOTSUP;
-    return check_error(-1, errInfo);
-}
-
-int
-efile_link(Efile_error* errInfo, char* old, char* new)
-{
-    return check_error(link(old, new), errInfo);
-}
-
-int
-efile_symlink(Efile_error* errInfo, char* old, char* new)
-{
-    return check_error(symlink(old, new), errInfo);
-}
-
-int
-efile_fadvise(Efile_error* errInfo, int fd, Sint64 offset,
-		  Sint64 length, int advise)
-{
-#ifdef HAVE_POSIX_FADVISE
-    return check_error(posix_fadvise(fd, offset, length, advise), errInfo);
-#else
-    return check_error(0, errInfo);
-#endif
-}
-
-#ifdef HAVE_SENDFILE
-/* For some reason the maximum size_t cannot be used as the max size
-   3GB seems to work on all platforms */
-#define SENDFILE_CHUNK_SIZE ((1UL << 30) -1)
-
-/*
- * sendfile: The implementation of the sendfile system call varies
- * a lot on different *nix platforms so to make the api similar in all
- * we have to emulate some things in linux and play with variables on
- * bsd/darwin.
- *
- * All of the calls will split a command which tries to send more than
- * SENDFILE_CHUNK_SIZE of data at once.
- *
- * On platforms where *nbytes of 0 does not mean the entire file, this is
- * simulated.
- *
- * It could be possible to implement header/trailer in sendfile. Though
- * you would have to emulate it in linux and on BSD/Darwin some complex
- * calculations have to be made when using a non blocking socket to figure
- * out how much of the header/file/trailer was sent in each command.
- *
- * The semantics of the API is this:
- * Return value: 1 if all data was sent and the function does not need to
- * be called again. 0 if an error occures OR if there is more data which
- * has to be sent (EAGAIN or EINTR will be set appropriately)
- *
- * The amount of data written in a call is returned through nbytes.
- *
- */
-
-int
-efile_sendfile(Efile_error* errInfo, int in_fd, int out_fd,
-	       off_t *offset, Uint64 *nbytes, struct t_sendfile_hdtl* hdtl)
-{
-    Uint64 written = 0;
-#if defined(__linux__)
-    ssize_t retval;
-    do {
-      /* check if *nbytes is 0 or greater than chunk size */
-      if (*nbytes == 0 || *nbytes > SENDFILE_CHUNK_SIZE)
-	retval = sendfile(out_fd, in_fd, offset, SENDFILE_CHUNK_SIZE);
-      else
-	retval = sendfile(out_fd, in_fd, offset, *nbytes);
-      if (retval > 0) {
-	written += retval;
-	*nbytes -= retval;
-      }
-    } while (retval == SENDFILE_CHUNK_SIZE);
-    if (written != 0) {
-      /* -1 is not returned by the linux API so we have to simulate it */
-      retval = -1;
-      errno = EAGAIN;
-    }
-#elif defined(__sun) && defined(__SVR4) && defined(HAVE_SENDFILEV)
-    ssize_t retval;
-    size_t len;
-    sendfilevec_t fdrec;
-    fdrec.sfv_fd = in_fd;
-    fdrec.sfv_flag = 0;
-    do {
-      fdrec.sfv_off = *offset;
-      len = 0;
-      /* check if *nbytes is 0 or greater than chunk size */
-      if (*nbytes == 0 || *nbytes > SENDFILE_CHUNK_SIZE)
-	fdrec.sfv_len = SENDFILE_CHUNK_SIZE;
-      else
-	fdrec.sfv_len = *nbytes;
-
-      retval = sendfilev(out_fd, &fdrec, 1, &len);
-
-      if (retval == -1 && errno == EINVAL) {
-          /* On some solaris versions (I've seen it on SunOS 5.10),
-             using a sfv_len larger then a filesize will result in
-             a -1 && errno == EINVAL return. We translate this so
-             a successful send of the data.*/
-          retval = len;
-      }
-
-      if (retval != -1 || errno == EAGAIN || errno == EINTR) {
-        *offset += len;
-	*nbytes -= len;
-	written += len;
-      }
-    } while (len == SENDFILE_CHUNK_SIZE);
-#elif defined(__DARWIN__)
-    int retval;
-    off_t len;
-    do {
-      /* check if *nbytes is 0 or greater than chunk size */
-      if(*nbytes > SENDFILE_CHUNK_SIZE)
-	len = SENDFILE_CHUNK_SIZE;
-      else
-	len = *nbytes;
-      retval = sendfile(in_fd, out_fd, *offset, &len, NULL, 0);
-      if (retval != -1 || errno == EAGAIN || errno == EINTR) {
-        *offset += len;
-	*nbytes -= len;
-	written += len;
-      }
-    } while (len == SENDFILE_CHUNK_SIZE);
-#elif defined(__FreeBSD__) || defined(__DragonFly__)
-    off_t len;
-    int retval;
-    do {
-      if (*nbytes > SENDFILE_CHUNK_SIZE)
-	retval = sendfile(in_fd, out_fd, *offset, SENDFILE_CHUNK_SIZE,
-			  NULL, &len, 0);
-      else
-	retval = sendfile(in_fd, out_fd, *offset, *nbytes, NULL, &len, 0);
-      if (retval != -1 || errno == EAGAIN || errno == EINTR) {
-	*offset += len;
-	*nbytes -= len;
-	written += len;
-      }
-    } while(len == SENDFILE_CHUNK_SIZE);
-#endif
-    *nbytes = written;
-    return check_error(retval, errInfo);
-}
-#endif /* HAVE_SENDFILE */
-
-#ifdef HAVE_POSIX_FALLOCATE
-static int
-call_posix_fallocate(int fd, Sint64 offset, Sint64 length)
-{
-    int ret;
-
-    /*
-     * On Linux and Solaris for example, posix_fallocate() returns
-     * a positive error number on error and it does not set errno.
-     * On FreeBSD however (9.0 at least), it returns -1 on error
-     * and it sets errno.
-     */
-    do {
-        ret = posix_fallocate(fd, (off_t) offset, (off_t) length);
-        if (ret > 0) {
-            errno = ret;
-            ret = -1;
-        }
-    } while (ret != 0 && errno == EINTR);
-
-    return ret;
-}
-#endif /* HAVE_POSIX_FALLOCATE */
-
-int
-efile_fallocate(Efile_error* errInfo, int fd, Sint64 offset, Sint64 length)
-{
-#if defined HAVE_FALLOCATE
-    /* Linux specific, more efficient than posix_fallocate. */
-    int ret;
-
-    do {
-        ret = fallocate(fd, FALLOC_FL_KEEP_SIZE, (off_t) offset, (off_t) length);
-    } while (ret != 0 && errno == EINTR);
-
-#if defined HAVE_POSIX_FALLOCATE
-    /* Fallback to posix_fallocate if available. */
-    if (ret != 0) {
-        ret = call_posix_fallocate(fd, offset, length);
-    }
-#endif
-
-    return check_error(ret, errInfo);
-#elif defined F_PREALLOCATE
-    /* Mac OS X specific, equivalent to posix_fallocate. */
-    int ret;
-    fstore_t fs;
-
-    memset(&fs, 0, sizeof(fs));
-    fs.fst_flags = F_ALLOCATECONTIG;
-    fs.fst_posmode = F_VOLPOSMODE;
-    fs.fst_offset = (off_t) offset;
-    fs.fst_length = (off_t) length;
-
-    ret = fcntl(fd, F_PREALLOCATE, &fs);
-
-    if (-1 == ret) {
-        fs.fst_flags = F_ALLOCATEALL;
-        ret = fcntl(fd, F_PREALLOCATE, &fs);
-
-#if defined HAVE_POSIX_FALLOCATE
-        /* Fallback to posix_fallocate if available. */
-        if (-1 == ret) {
-            ret = call_posix_fallocate(fd, offset, length);
-        }
-#endif
-    }
-
-    return check_error(ret, errInfo);
-#elif defined HAVE_POSIX_FALLOCATE
-    /* Other Unixes, use posix_fallocate if available. */
-    return check_error(call_posix_fallocate(fd, offset, length), errInfo);
-#else
-    errno = ENOTSUP;
-    return check_error(-1, errInfo);
-#endif
-}
diff --git a/erts/emulator/drivers/win32/win_efile.c b/erts/emulator/drivers/win32/win_efile.c
deleted file mode 100644
index 2d366b5833..0000000000
--- a/erts/emulator/drivers/win32/win_efile.c
+++ /dev/null
@@ -1,2058 +0,0 @@
-/*
- * %CopyrightBegin%
- *
- * Copyright Ericsson AB 1997-2016. All Rights Reserved.
- *
- * Licensed under the Apache License, Version 2.0 (the "License");
- * you may not use this file except in compliance with the License.
- * You may obtain a copy of the License at
- *
- *     http://www.apache.org/licenses/LICENSE-2.0
- *
- * Unless required by applicable law or agreed to in writing, software
- * distributed under the License is distributed on an "AS IS" BASIS,
- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- * See the License for the specific language governing permissions and
- * limitations under the License.
- *
- * %CopyrightEnd%
- */
-/*
- * Purpose: Provides file and directory operations for Windows.
- */
-
-#include <windows.h>
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-#include "sys.h"
-#include <ctype.h>
-#include <wchar.h>
-#include "erl_efile.h"
-
-#define DBG_TRACE_MASK 0
-/* 1 = file name ops
- * 2 = file descr ops
- * 4 = errors
- * 8 = path name conversion
- */
-#if !DBG_TRACE_MASK
-#  define DBG_TRACE(M,S)
-#  define DBG_TRACE1(M,FMT,A)
-#  define DBG_TRACE2(M,FMT,A,B)
-#else
-#  define DBG_TRACE(M,S) do { if ((M)&DBG_TRACE_MASK) fwprintf(stderr, L"DBG_TRACE %d: %s\r\n", __LINE__, (WCHAR*)(S)); }while(0)
-#  define DBG_TRACE1(M,FMT,A) do { if ((M)&DBG_TRACE_MASK) fwprintf(stderr, L"DBG_TRACE %d: " L##FMT L"\r\n", __LINE__, (A)); }while(0)
-#  define DBG_TRACE2(M,FMT,A,B) do { if ((M)&DBG_TRACE_MASK) fwprintf(stderr, L"DBG_TRACE %d: " L##FMT L"\r\n", __LINE__, (A), (B)); }while(0)
-#endif
-
-/*
- * Microsoft-specific function to map a WIN32 error code to a Posix errno.
- */
-
-#define ISSLASH(a)  ((a) == L'\\' || (a) == L'/')
-#define ISDIR(st) (((st).st_mode&S_IFMT) == S_IFDIR)
-#define ISREG(st) (((st).st_mode&S_IFMT) == S_IFREG)
-
-#define IS_DOT_OR_DOTDOT(s) \
-    ((s)[0] == L'.' && ((s)[1] == L'\0' || ((s)[1] == L'.' && (s)[2] == L'\0')))
-
-#define FILE_SHARE_FLAGS (FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE)
-
-#ifndef INVALID_FILE_ATTRIBUTES
-#define INVALID_FILE_ATTRIBUTES ((DWORD) 0xFFFFFFFF)
-#endif
-
-#define TICKS_PER_SECOND (10000000ULL)
-#define EPOCH_DIFFERENCE (11644473600LL)
-
-#define FILETIME_TO_EPOCH(epoch, ft) \
-    do { \
-	ULARGE_INTEGER ull; \
-	ull.LowPart  = (ft).dwLowDateTime; \
-	ull.HighPart = (ft).dwHighDateTime; \
-	(epoch) = ((ull.QuadPart / TICKS_PER_SECOND) - EPOCH_DIFFERENCE); \
-    } while(0)
-
-#define EPOCH_TO_FILETIME(ft, epoch) \
-    do { \
-	ULARGE_INTEGER ull; \
-	ull.QuadPart = (((epoch) + EPOCH_DIFFERENCE) * TICKS_PER_SECOND); \
-	(ft).dwLowDateTime = ull.LowPart; \
-	(ft).dwHighDateTime = ull.HighPart; \
-    } while(0)
-
-
-static int check_error(int result, Efile_error* errInfo);
-static int set_error(Efile_error* errInfo);
-static int set_os_errno(Efile_error* errInfo, DWORD os_errno);
-static int is_root_unc_name(const WCHAR *path);
-static int extract_root(WCHAR *name);
-static unsigned short dos_to_posix_mode(int attr, const WCHAR *name);
-
-
-struct wpath_tmp_buffer {
-    struct wpath_tmp_buffer* next;
-    WCHAR buffer[1];
-};
-
-typedef struct {
-    Efile_error* errInfo;
-    struct wpath_tmp_buffer* buf_list;
-}Efile_call_state;
-
-static void call_state_init(Efile_call_state* state, Efile_error* errInfo)
-{
-    state->errInfo = errInfo;
-    state->buf_list = NULL;
-}
-static WCHAR* wpath_tmp_alloc(Efile_call_state* state, size_t len)
-{
-    size_t sz = offsetof(struct wpath_tmp_buffer, buffer)
-	+ (len+1)*sizeof(WCHAR);
-    struct wpath_tmp_buffer* p = driver_alloc(sz);
-    p->next = state->buf_list;
-    state->buf_list = p;
-    return p->buffer;
-}
-static void call_state_free(Efile_call_state* state)
-{
-    while(state->buf_list) {
-	struct wpath_tmp_buffer* next = state->buf_list->next;
-	driver_free(state->buf_list);
-	state->buf_list = next;
-    }
-}
-static WCHAR* get_cwd_wpath_tmp(Efile_call_state* state)
-{
-    WCHAR dummy;
-    DWORD size = GetCurrentDirectoryW(0, &dummy);
-    WCHAR* ret = NULL;
-
-    if (size) {
-	ret = wpath_tmp_alloc(state, size);
-	if (!GetCurrentDirectoryW(size, ret)) {
-	    ret = NULL;
-	}
-    }
-    return ret;
-}
-static WCHAR* get_full_wpath_tmp(Efile_call_state* state,
-                                 const WCHAR* file,
-				 WCHAR** file_part,
-				 DWORD extra)
-{
-    WCHAR dummy;
-    DWORD size = GetFullPathNameW(file, 0, &dummy, NULL);
-    WCHAR* ret = NULL;
-
-    if (size) {
-	int ok;
-	ret = wpath_tmp_alloc(state, size + extra);
-	if (file_part) {
-	    ok = (GetFullPathNameW(file, size, ret, file_part) != 0);
-	}
-	else {
-	    ok = (_wfullpath(ret, file, size) != NULL);
-	}
-	if (!ok) {
-	    ret = NULL;
-	}
-    }
-    return ret;
-}
-
-static void ensure_wpath_max(Efile_call_state* state, WCHAR** pathp, size_t max);
-static int do_rmdir(Efile_call_state*, char* name);
-static int do_rename(Efile_call_state*, char* src, char* dst);
-static int do_readdir(Efile_call_state*, char* name, EFILE_DIR_HANDLE*, char* buffer, size_t *size);
-static int do_fileinfo(Efile_call_state*, Efile_info*, char* orig_name, int info_for_link);
-static char* do_readlink(Efile_call_state*, char* name, char* buffer, size_t size);
-static int do_altname(Efile_call_state*, char* orig_name, char* buffer, size_t size);
-
-
-static int errno_map(DWORD last_error) {
-
-    switch (last_error) {
-    case ERROR_SUCCESS:
-	return 0;
-    case ERROR_INVALID_FUNCTION:
-    case ERROR_INVALID_DATA:
-    case ERROR_INVALID_PARAMETER:
-    case ERROR_INVALID_TARGET_HANDLE:
-    case ERROR_INVALID_CATEGORY:
-    case ERROR_NEGATIVE_SEEK:
-	return EINVAL;
-    case ERROR_DIR_NOT_EMPTY:
-	return EEXIST;
-    case ERROR_BAD_FORMAT:
-	return ENOEXEC;
-    case ERROR_PATH_NOT_FOUND:
-    case ERROR_FILE_NOT_FOUND:
-    case ERROR_NO_MORE_FILES:
-	return ENOENT;
-    case ERROR_TOO_MANY_OPEN_FILES:
-	return EMFILE;
-    case ERROR_ACCESS_DENIED:
-    case ERROR_INVALID_ACCESS:
-    case ERROR_CURRENT_DIRECTORY:
-    case ERROR_SHARING_VIOLATION:
-    case ERROR_LOCK_VIOLATION:
-    case ERROR_INVALID_PASSWORD:
-    case ERROR_DRIVE_LOCKED:
-	return EACCES;
-    case ERROR_INVALID_HANDLE:
-	return EBADF;
-    case ERROR_NOT_ENOUGH_MEMORY:
-    case ERROR_OUTOFMEMORY:
-    case ERROR_OUT_OF_STRUCTURES:
-	return ENOMEM;
-    case ERROR_INVALID_DRIVE:
-    case ERROR_BAD_UNIT:
-    case ERROR_NOT_READY:
-    case ERROR_REM_NOT_LIST:
-    case ERROR_DUP_NAME:
-    case ERROR_BAD_NETPATH:
-    case ERROR_NETWORK_BUSY:
-    case ERROR_DEV_NOT_EXIST:
-    case ERROR_BAD_NET_NAME:
-	return ENXIO;
-    case ERROR_NOT_SAME_DEVICE:
-	return EXDEV;
-    case ERROR_WRITE_PROTECT:
-	return EROFS;
-    case ERROR_BAD_LENGTH:
-    case ERROR_BUFFER_OVERFLOW:
-	return E2BIG;
-    case ERROR_SEEK:
-    case ERROR_SECTOR_NOT_FOUND:
-	return ESPIPE;
-    case ERROR_NOT_DOS_DISK:
-	return ENODEV;
-    case ERROR_GEN_FAILURE:
-	return ENODEV;
-    case ERROR_SHARING_BUFFER_EXCEEDED:
-    case ERROR_NO_MORE_SEARCH_HANDLES:
-	return EMFILE;
-    case ERROR_HANDLE_EOF:
-    case ERROR_BROKEN_PIPE:
-	return EPIPE;
-    case ERROR_HANDLE_DISK_FULL:
-    case ERROR_DISK_FULL:
-	return ENOSPC;
-    case ERROR_NOT_SUPPORTED:
-	return ENOTSUP;
-    case ERROR_FILE_EXISTS:
-    case ERROR_ALREADY_EXISTS:
-    case ERROR_CANNOT_MAKE:
-	return EEXIST;
-    case ERROR_ALREADY_ASSIGNED:
-	return EBUSY;
-    case ERROR_NO_PROC_SLOTS:
-	return EAGAIN;
-    case ERROR_CANT_RESOLVE_FILENAME:
-	return EMLINK;
-    case ERROR_PRIVILEGE_NOT_HELD:
-	return EPERM;
-    case ERROR_ARENA_TRASHED:
-    case ERROR_INVALID_BLOCK:
-    case ERROR_BAD_ENVIRONMENT:
-    case ERROR_BAD_COMMAND:
-    case ERROR_CRC:
-    case ERROR_OUT_OF_PAPER:
-    case ERROR_READ_FAULT:
-    case ERROR_WRITE_FAULT:
-    case ERROR_WRONG_DISK:
-    case ERROR_NET_WRITE_FAULT:
-	return EIO;
-    default: /* not to do with files I expect. */
-	return EIO;
-    }	
-}
-
-static int
-check_error(int result, Efile_error* errInfo)
-{
-    if (result < 0) {
-	errInfo->posix_errno = errno;
-	errInfo->os_errno = GetLastError();
-	DBG_TRACE2(4, "ERROR os_error=%d errno=%d @@@@@@@@@@@@@@@@@@@@@@@@@@@@",
-	        errInfo->os_errno, errInfo->posix_errno);
-	return 0;
-    }
-    return 1;
-}
-
-static void
-save_last_error(Efile_error* errInfo)
-{
-    errInfo->posix_errno = errno;
-    errInfo->os_errno = GetLastError();
-    DBG_TRACE2(4, "ERROR os_error=%d errno=%d $$$$$$$$$$$$$$$$$$$$$$$$$$$$$",
-	errInfo->os_errno, errInfo->posix_errno);
-}
-
-
-/*
- * Fills the provided error information structure with information
- * with the error code given by GetLastError() and its corresponding
- * Posix error number.
- *
- * Returns 0.
- */
-
-static int
-set_error(Efile_error* errInfo)
-{
-    set_os_errno(errInfo, GetLastError());
-    return 0;
-}
-
-
-static int
-set_os_errno(Efile_error* errInfo, DWORD os_errno)
-{
-    errInfo->os_errno = os_errno;
-    errInfo->posix_errno = errno_map(os_errno);
-    DBG_TRACE2(4, "ERROR os_error=%d errno=%d ############################",
-	    errInfo->os_errno, errInfo->posix_errno);
-    return 0;
-}
-
-int
-efile_init() {
-   return 1;
-}
-
-/*
- * A writev with Unix semantics, but with Windows arguments 
- */
-static int 
-win_writev(Efile_error* errInfo,
-	   HANDLE fd,                  /* handle to file */
-	   FILE_SEGMENT_ELEMENT iov[], /* array of buffer pointers */
-	   DWORD *size)                /* number of bytes to write */
-{
-    OVERLAPPED ov;
-    ov.Offset = 0L;
-    ov.OffsetHigh = 0L;
-    ov.hEvent = CreateEvent(NULL, TRUE, FALSE, NULL);
-    if (ov.hEvent == NULL)
-	return set_error(errInfo);
-    if (! write_file_gather(fd, iov, *size, NULL, &ov))
-	return set_error(errInfo);
-    if (WaitForSingleObject(ov.hEvent, INFINITE) != WAIT_OBJECT_0)
-	return set_error(errInfo);
-    if (! GetOverlappedResult(fd, &ov, size, FALSE))
-	return set_error(errInfo);
-    return 1;
-}
-
-
-/* Check '*pathp' and convert it if needed to something that windows will accept.
- * Typically use UNC path with \\?\ prefix if absolute path is longer than 260.
- */
-static void ensure_wpath(Efile_call_state* state, WCHAR** pathp)
-{
-    ensure_wpath_max(state, pathp, MAX_PATH);
-}
-
-static void ensure_wpath_max(Efile_call_state* state, WCHAR** pathp, size_t max)
-{
-    WCHAR* path = *pathp;
-    WCHAR* p;
-    size_t len = wcslen(path);
-    int unc_fixup = 0;
-
-    if (path[0] == 0) {
-	DBG_TRACE(8, L"Let empty path pass through");
-	return;
-    }
-
-    DBG_TRACE1(8,"IN: %s", path);
-
-    if (path[1] == L':' && ISSLASH(path[2])) { /* absolute path */
-	if (len >= max) {
-	    WCHAR *src, *dst;
-
-	    *pathp = wpath_tmp_alloc(state, 4+len+1);
-	    dst = *pathp;
-	    wcscpy(dst, L"\\\\?\\");
-	    for (src=path,dst+=4; *src; src++) {
-		if (*src == L'/') {
-		    if (dst[-1] != L'\\') {
-			*dst++ =  L'\\';
-		    }
-		    /*else ignore redundant slashes */
-		}
-		else
-		    *dst++ = *src;
-	    }
-	    *dst = 0;
-	    unc_fixup = 1;
-	}
-    }
-    else if (!(ISSLASH(path[0]) && ISSLASH(path[1]))) { /* relative path */
-	DWORD cwdLen = GetCurrentDirectoryW(0, NULL);
-	DWORD absLen = cwdLen + 1 + len;
-	if (absLen >= max) {
-	    WCHAR *fullPath = wpath_tmp_alloc(state, 4+4+absLen);
-	    DWORD fullLen;
-
-	    fullLen = GetFullPathNameW(path, 4 + absLen, fullPath+4, NULL);
-	    if (fullLen >= 4+absLen) {
-		*pathp = path;
-		DBG_TRACE2(8,"ensure_wpath FAILED absLen=%u %s", (int)absLen, path);
-		return;
-	    }
-	    /* GetFullPathNameW can return paths longer than MAX_PATH without the \\?\ prefix.
-	     * At least seen on Windows 7. Go figure...
-	     */
-	    if (fullLen >= max && wcsncmp(fullPath+4, L"\\\\?\\", 4) != 0) {
-		wcsncpy(fullPath, L"\\\\?\\", 4);
-		*pathp = fullPath;
-	    }
-	    else {
-		*pathp = fullPath + 4;
-	    }
-	}
-    }
-
-    if (unc_fixup) {
-	WCHAR* endp;
-
-	p = *pathp;
-	len = wcslen(p);
-	endp = p + len;
-	if (len > 4) {
-	    p += 4;
-	    while (*p) {
-		if (p[0] == L'\\' && p[1] == L'.') {
-		    if (p[2] == L'\\' || !p[2]) { /* single dot */
-			wmemmove(p, p+2, (&endp[1] - &p[2]));
-			endp -= 2;
-		    }
-		    else if (p[2] == L'.' && (p[3] == L'\\' || !p[3])) { /* double dot */
-			WCHAR* r;
-			for (r=p-1; *r == L'\\'; --r)
-			    /*skip redundant slashes*/;
-			for (; *r != L'\\'; --r)
-			    /*find start of prev directory*/;
-			if (r < *pathp + 6)
-			    break;
-			wmemmove(r, p+3, (&endp[1] - &p[3]));
-			p = r;
-		    }
-		    else p += 3;
-		}
-		else ++p;
-	    }
-	}
-    }
-    DBG_TRACE1(8,"OUT: %s", *pathp);
-}
-
-int
-efile_mkdir(Efile_error* errInfo,	/* Where to return error codes. */
-	    char* name)			/* Name of directory to create. */
-{
-    Efile_call_state state;
-    WCHAR* wname = (WCHAR*)name;
-    int ret;
-
-    DBG_TRACE(1, name);
-    call_state_init(&state, errInfo);
-    ensure_wpath_max(&state, &wname, 248); /* Yes, 248 limit for normal paths */
-
-    ret = (int) CreateDirectoryW(wname, NULL);
-    if (!ret)
-	set_error(errInfo);
-
-    call_state_free(&state);
-    return ret;
-}
-
-int
-efile_rmdir(Efile_error* errInfo,	/* Where to return error codes. */
-	    char* name)			/* Name of directory to delete. */
-{
-    Efile_call_state state;
-    int ret;
-
-    DBG_TRACE(1, name);
-    call_state_init(&state, errInfo);
-    ret = do_rmdir(&state, name);
-    call_state_free(&state);
-    return ret;
-}
-
-static int do_rmdir(Efile_call_state* state, char* name)
-{
-    OSVERSIONINFO os;
-    DWORD attr;
-    WCHAR *wname = (WCHAR *) name;
-    WCHAR *buffer = NULL;
-
-    ensure_wpath(state, &wname);
-
-    if (RemoveDirectoryW(wname) != FALSE) {
-	return 1;
-    }
-    errno = errno_map(GetLastError());
-    if (errno == EACCES) {
-	attr = GetFileAttributesW(wname);
-	if (attr != (DWORD) -1) {
-	    if ((attr & FILE_ATTRIBUTE_DIRECTORY) == 0) {
-		/* 
-		 * Windows 95 reports calling RemoveDirectory on a file as an 
-		 * EACCES, not an ENOTDIR.
-		 */
-		
-		errno = ENOTDIR;
-		goto end;
-	    }
-
-	    /* 
-	     * Windows 95 reports removing a non-empty directory as
-	     * an EACCES, not an EEXIST.  If the directory is not empty,
-	     * change errno so caller knows what's going on.
-	     */
-
-	    os.dwOSVersionInfoSize = sizeof(os);
-	    GetVersionEx(&os);
-	    if (os.dwPlatformId == VER_PLATFORM_WIN32_WINDOWS) {
-		HANDLE handle;
-		WIN32_FIND_DATAW data;
-		int len = wcslen(wname);
-
-		buffer = wpath_tmp_alloc(state, len + 4);
-		wcscpy(buffer, wname);
-		if (buffer[0] && buffer[len-1] != L'\\' && buffer[len-1] != L'/') {
-		    wcscat(buffer, L"\\");
-		}
-		wcscat(buffer, L"*.*");
-		handle = FindFirstFileW(buffer, &data);
-		if (handle != INVALID_HANDLE_VALUE) {
-		    while (1) {
-			if ((wcscmp(data.cFileName, L".") != 0)
-				&& (wcscmp(data.cFileName, L"..") != 0)) {
-			    /*
-			     * Found something in this directory.
-			     */
-
-			    errno = EEXIST;
-			    break;
-			}
-			if (FindNextFileW(handle, &data) == FALSE) {
-			    break;
-			}
-		    }
-		    FindClose(handle);
-		}
-	    }
-	}
-    }
-
-    if (errno == ENOTEMPTY) {
-	/* 
-	 * Posix allows both EEXIST or ENOTEMPTY, but we'll always
-	 * return EEXIST to allow easy matching in Erlang code.
-	 */
-
-	errno = EEXIST;
-    }
-
- end:
-    save_last_error(state->errInfo);
-    return 0;
-}
-
-int
-efile_delete_file(Efile_error* errInfo,		/* Where to return error codes. */
-		  char* name)			/* Name of file to delete. */
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(1, name);
-    call_state_init(&state, errInfo);
-    ret = do_delete_file(&state, name);
-    call_state_free(&state);
-    return ret;
-}
-
-static int do_delete_file(Efile_call_state* state, char* name)
-{
-    DWORD attr;
-    WCHAR *wname = (WCHAR *) name;
-
-    ensure_wpath(state, &wname);
-
-    if (DeleteFileW(wname) != FALSE) {
-	return 1;
-    }
-
-    errno = errno_map(GetLastError());
-    if (errno == EACCES) {
-        attr = GetFileAttributesW(wname);
-	if (attr != (DWORD) -1) {
-	    if (attr & FILE_ATTRIBUTE_DIRECTORY) {
-		/*
-		 * Windows NT reports removing a directory as EACCES instead
-		 * of EPERM.
-		 */
-
-		errno = EPERM;
-	    }
-	}
-    } else if (errno == ENOENT) {
-        attr = GetFileAttributesW(wname);
-	if (attr != (DWORD) -1) {
-	    if (attr & FILE_ATTRIBUTE_DIRECTORY) {
-	    	/*
-		 * Windows 95 reports removing a directory as ENOENT instead 
-		 * of EPERM.
-		 */
-
-		errno = EPERM;
-	    }
-	}
-    } else if (errno == EINVAL) {
-	/*
-	 * Windows NT reports removing a char device as EINVAL instead of
-	 * EACCES.
-	 */
-	
-	errno = EACCES;
-    }
-
-    return check_error(-1, state->errInfo);
-}
-
-/*
- *---------------------------------------------------------------------------
- *
- *      Changes the name of an existing file or directory, from src to dst.
- *	If src and dst refer to the same file or directory, does nothing
- *	and returns success.  Otherwise if dst already exists, it will be
- *	deleted and replaced by src subject to the following conditions:
- *	    If src is a directory, dst may be an empty directory.
- *	    If src is a file, dst may be a file.
- *	In any other situation where dst already exists, the rename will
- *	fail.  
- *
- *	Some possible error codes:
- *
- *	EACCES:     src or dst parent directory can't be read and/or written.
- *	EEXIST:	    dst is a non-empty directory.
- *	EINVAL:	    src is a root directory or dst is a subdirectory of src.
- *	EISDIR:	    dst is a directory, but src is not.
- *	ENOENT:	    src doesn't exist, or src or dst is "".
- *	ENOTDIR:    src is a directory, but dst is not.  
- *	EXDEV:	    src and dst are on different filesystems.
- *	
- * Side effects:
- *	The implementation of rename may allow cross-filesystem renames,
- *	but the caller should be prepared to emulate it with copy and
- *	delete if errno is EXDEV.
- *
- *---------------------------------------------------------------------------
- */
-
-int
-efile_rename(Efile_error* errInfo, char* src, char* dst)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(1, src);
-    call_state_init(&state, errInfo);
-    ret = do_rename(&state, src, dst);
-    call_state_free(&state);
-    return ret;
-}
-
-static int
-do_rename(Efile_call_state* state,
-	  char* src,			/* Original name. */
-	  char* dst)			/* New name. */
-{
-    DWORD srcAttr, dstAttr;
-    WCHAR *wsrc = (WCHAR *) src;
-    WCHAR *wdst = (WCHAR *) dst;
-
-    ensure_wpath(state, &wsrc);
-    ensure_wpath(state, &wdst);
-
-    if (MoveFileW(wsrc, wdst) != FALSE) {
-	return 1;
-    }
-
-    errno = errno_map(GetLastError());
-    srcAttr = GetFileAttributesW(wsrc);
-    dstAttr = GetFileAttributesW(wdst);
-    if (srcAttr == (DWORD) -1) {
-	srcAttr = 0;
-    }
-    if (dstAttr == (DWORD) -1) {
-	dstAttr = 0;
-    }
-
-    if (errno == EBADF) {
-	errno = EACCES;
-	return check_error(-1, state->errInfo);
-    }
-    if (errno == EACCES) {
-	decode:
-	if (srcAttr & FILE_ATTRIBUTE_DIRECTORY) {
-	    WCHAR *srcPath, *dstPath;
-	    WCHAR *srcRest, *dstRest;
-	    int size;
-
-	    srcPath = get_full_wpath_tmp(state, wsrc, &srcRest, 0);
-	    if (!srcPath) {
-		save_last_error(state->errInfo);
-		return 0;
-	    }
-
-	    dstPath = get_full_wpath_tmp(state, wdst, &dstRest, 0);
-	    if (!dstPath) {
-		save_last_error(state->errInfo);
-		return 0;
-	    }
-
-	    if (srcRest == NULL) {
-		srcRest = srcPath + wcslen(srcPath);
-	    }
-	    if (_wcsnicmp(srcPath, dstPath, srcRest - srcPath) == 0) {
-		/*
-		 * Trying to move a directory into itself.
-		 */
-
-		errno = EINVAL;
-	    }
-	    if (extract_root(srcPath)) {
-		/*
-		 * Attempt to move a root directory.  Never allowed.
-		 */
-		errno = EINVAL;
-	    }
-
-	    (void) extract_root(dstPath);
-	    if (dstPath[0] == L'\0') {
-		/*
-		 * The filename was invalid.  (Don't know why,
-		 * but play it safe.)
-		 */
-		errno = EINVAL;
-	    }
-	    if (_wcsicmp(srcPath, dstPath) != 0) {
-		/*
-		 * If src is a directory and dst filesystem != src
-		 * filesystem, errno should be EXDEV.  It is very
-		 * important to get this behavior, so that the caller
-		 * can respond to a cross filesystem rename by
-		 * simulating it with copy and delete.  The MoveFile
-		 * system call already handles the case of moving a
-		 * *file* between filesystems.
-		 */
-
-		errno = EXDEV;
-	    }
-	}
-
-	/*
-	 * Other types of access failure is that dst is a read-only
-	 * filesystem, that an open file referred to src or dest, or that
-	 * src or dest specified the current working directory on the
-	 * current filesystem.  EACCES is returned for those cases.
-	 */
-
-    } else if (errno == EEXIST) {
-	/*
-	 * Reports EEXIST any time the target already exists.  If it makes
-	 * sense, remove the old file and try renaming again.
-	 */
-
-	if (srcAttr & FILE_ATTRIBUTE_DIRECTORY) {
-	    if (dstAttr & FILE_ATTRIBUTE_DIRECTORY) {
-		/*
-		 * Overwrite empty dst directory with src directory.  The
-		 * following call will remove an empty directory.  If it
-		 * fails, it's because it wasn't empty.
-		 */
-
-		if (RemoveDirectoryW(wdst)) {
-		    /*
-		     * Now that that empty directory is gone, we can try
-		     * renaming again.  If that fails, we'll put this empty
-		     * directory back, for completeness.
-		     */
-
-		    if (MoveFileW(wsrc, wdst) != FALSE) {
-			return 1;
-		    }
-
-		    /*
-		     * Some new error has occurred.  Don't know what it
-		     * could be, but report this one.
-		     */
-
-		    errno = errno_map(GetLastError());
-		    CreateDirectoryW(wdst, NULL);
-		    SetFileAttributesW(wdst, dstAttr);
-		    if (errno == EACCES) {
-			/*
-			 * Decode the EACCES to a more meaningful error.
-			 */
-
-			goto decode;
-		    }
-		}
-	    } else {	/* (dstAttr & FILE_ATTRIBUTE_DIRECTORY) == 0 */
-		errno = ENOTDIR;
-	    }
-	} else {    /* (srcAttr & FILE_ATTRIBUTE_DIRECTORY) == 0 */
-	    if (dstAttr & FILE_ATTRIBUTE_DIRECTORY) {
-		errno = EISDIR;
-	    } else {
-		/*
-		 * Overwrite existing file by:
-		 * 
-		 * 1. Rename existing file to temp name.
-		 * 2. Rename old file to new name.
-		 * 3. If success, delete temp file.  If failure,
-		 *    put temp file back to old name.
-		 */
-
-		WCHAR *tempName;
-		int result;
-		WCHAR *rest;
-		
-		tempName = get_full_wpath_tmp(state, wdst, &rest, 14);
-		if (!tempName || !rest) {
-		    save_last_error(state->errInfo);
-		    return 0;
-		}
-
-		*rest = L'\0';
-		result = -1;
-		if (GetTempFileNameW(tempName, L"erlr", 0, tempName) != 0) {
-		    /*
-		     * Strictly speaking, need the following DeleteFile and
-		     * MoveFile to be joined as an atomic operation so no
-		     * other app comes along in the meantime and creates the
-		     * same temp file.
-		     */
-		     
-		    DeleteFileW(tempName);
-		    if (MoveFileW(wdst, tempName) != FALSE) {
-			if (MoveFileW(wsrc, wdst) != FALSE) {
-			    SetFileAttributesW(tempName, FILE_ATTRIBUTE_NORMAL);
-			    DeleteFileW(tempName);
-			    return 1;
-			} else {
-			    DeleteFileW(wdst);
-			    MoveFileW(tempName, wdst);
-			}
-		    } 
-
-		    /*
-		     * Can't backup dst file or move src file.  Return that
-		     * error.  Could happen if an open file refers to dst.
-		     */
-
-		    errno = errno_map(GetLastError());
-		    if (errno == EACCES) {
-			/*
-			 * Decode the EACCES to a more meaningful error.
-			 */
-			goto decode;
-		    }
-		}
-		return result;
-	    }
-	}
-    }
-    return check_error(-1, state->errInfo);
-}
-
-int
-efile_chdir(Efile_error* errInfo,	/* Where to return error codes. */
-	    char* name)			/* Name of directory to make current. */
-{

-    /* We don't even try to handle long paths here

-     * as current working directory is always limited to MAX_PATH

-     * even if we use UNC paths and SetCurrentDirectoryW()

-     */

-    int success = check_error(_wchdir((WCHAR *) name), errInfo);

-    if (!success && errInfo->posix_errno == EINVAL)

-	/* POSIXification of errno */

-	errInfo->posix_errno = ENOENT;
-    return success;
-}
-
-int
-efile_getdcwd(Efile_error* errInfo,		/* Where to return error codes. */
-	      int drive,			/* 0 - current, 1 - A, 2 - B etc. */
-	      char* buffer,			/* Where to return the current directory. */
-	      size_t size)			/* Size of buffer. */
-{
-    WCHAR *wbuffer = (WCHAR *) buffer;
-    size_t wbuffer_size = size / 2; 
-    DBG_TRACE(1, L"#getdcwd#");
-    if (_wgetdcwd(drive, wbuffer, wbuffer_size) == NULL) {
-	return check_error(-1, errInfo);
-    }
-    DBG_TRACE1(8, "getdcwd OS=%s", wbuffer);
-    if (wcsncmp(wbuffer, L"\\\\?\\", 4) == 0) {
-	wmemmove(wbuffer, wbuffer+4, wcslen(wbuffer+4)+1);
-    }
-    for ( ; *wbuffer; wbuffer++) 
-	if (*wbuffer == L'\\')
-	    *wbuffer = L'/';
-    DBG_TRACE1(8, "getdcwd ERLANG=%s", (WCHAR*)buffer);
-    return 1;
-}
-
-int
-efile_readdir(Efile_error* errInfo, char* name, EFILE_DIR_HANDLE* dir_handle,
-	      char* buffer, size_t *size)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(dir_handle?2:1, name);
-    call_state_init(&state, errInfo);
-    ret = do_readdir(&state, name, dir_handle, buffer, size);
-    call_state_free(&state);
-    return ret;
-}
-
-static int do_readdir(Efile_call_state* state,
-                      char* name,                   /* Name of directory to list */
-		      EFILE_DIR_HANDLE* dir_handle, /* Handle of opened directory or NULL */
-		      char* buffer,                 /* Buffer to put one filename in */
-		      size_t *size)                 /* in-out size of buffer/size of filename excluding zero
-					               termination in bytes*/
-{
-    HANDLE dir;			/* Handle to directory. */
-    WIN32_FIND_DATAW findData;	/* Data found by FindFirstFile() or FindNext(). */
-    /* Alignment is not honored, this works on x86 because of alignment fixup by processor.
-       Not perfect, but faster than alinging by hand (really) */
-    WCHAR *wbuffer = (WCHAR *) buffer;
-
-    /*
-     * First time we must setup everything.
-     */
-
-    if (*dir_handle == NULL) {
-	WCHAR *wname = (WCHAR *) name;
-	WCHAR* wildcard;
-	int length;
-	WCHAR* s;
-
-	ensure_wpath_max(state, &wname, MAX_PATH-2);
-	length = wcslen(wname);
-
-	wildcard = wpath_tmp_alloc(state, length+3);
-
-	wcscpy(wildcard, wname);
-	s = wildcard+length-1;
-	if (*s != L'/' && *s != L'\\')
-	    *++s = L'\\';
-	*++s = L'*';
-	*++s = L'\0';
-	DEBUGF(("Reading %ws\n", wildcard));
-	dir = FindFirstFileW(wildcard, &findData);
-	if (dir == INVALID_HANDLE_VALUE) {
-	    set_error(state->errInfo);
-	    return 0;
-	}
-	*dir_handle = (EFILE_DIR_HANDLE) dir;
-
-	if (!IS_DOT_OR_DOTDOT(findData.cFileName)) {
-	    wcscpy(wbuffer, findData.cFileName);
-	    *size = wcslen(wbuffer)*2;
-	    return 1;
-	}
-    }
-
-    /*
-     * Retrieve the name of the next file using the directory handle.
-     */
-
-    dir = (HANDLE) *dir_handle;
-
-    for (;;) {
-	if (FindNextFileW(dir, &findData)) {
-	    if (IS_DOT_OR_DOTDOT(findData.cFileName))
-		continue;
-	    wcscpy(wbuffer, findData.cFileName);
-	    *size = wcslen(wbuffer)*2;
-	    return 1;
-	}
-
-	if (GetLastError() == ERROR_NO_MORE_FILES) {
-	    state->errInfo->posix_errno = state->errInfo->os_errno = 0;
-	}
-	else {
-	    set_error(state->errInfo);
-	}
-	FindClose(dir);
-	return 0;
-    }
-}
-
-int
-efile_openfile(Efile_error* errInfo, char* name, int flags, int* pfd, Sint64* pSize)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE1(1, "openfile(%s)", name);
-    call_state_init(&state, errInfo);
-    ret = do_openfile(&state, name, flags, pfd, pSize);
-    call_state_free(&state);
-    return ret;
-}
-
-static
-int do_openfile(Efile_call_state* state,        /* Where to return error codes. */
-	        char* name,			/* Name of directory to open. */
-	        int flags,			/* Flags to use for opening. */
-	        int* pfd,			/* Where to store the file descriptor. */
-	        Sint64* pSize)			/* Where to store the size of the file. */
-{
-    Efile_error* errInfo = state->errInfo;
-    BY_HANDLE_FILE_INFORMATION fileInfo; /* File information from a handle. */
-    HANDLE fd;			/* Handle to open file. */
-    DWORD access;		/* Access mode: GENERIC_READ, GENERIC_WRITE. */
-    DWORD crFlags;
-    DWORD flagsAndAttrs = FILE_ATTRIBUTE_NORMAL;
-    WCHAR *wname = (WCHAR *) name;
-
-    switch (flags & (EFILE_MODE_READ|EFILE_MODE_WRITE)) {
-    case EFILE_MODE_READ:
-	access = GENERIC_READ;
-	crFlags = OPEN_EXISTING;
-	break;
-    case EFILE_MODE_WRITE:
-	access = GENERIC_WRITE;
-	crFlags = CREATE_ALWAYS;
-	break;
-    case EFILE_MODE_READ_WRITE:
-	access = GENERIC_READ|GENERIC_WRITE;
-	crFlags = OPEN_ALWAYS;
-	break;
-    default:
-	errno = EINVAL;
-	check_error(-1, errInfo);
-	return 0;
-    }
-
-    if (flags & EFILE_MODE_SYNC) {
-        flagsAndAttrs = FILE_FLAG_WRITE_THROUGH;
-    }
-
-    if (flags & EFILE_MODE_APPEND) {
-	crFlags = OPEN_ALWAYS;
-    }
-    if (flags & EFILE_MODE_EXCL) {
-	crFlags = CREATE_NEW;
-    }
-    ensure_wpath(state, &wname);
-    fd = CreateFileW(wname, access,
-		    FILE_SHARE_FLAGS,
-		    NULL, crFlags, flagsAndAttrs, NULL);
-
-    /*
-     * Check for errors.
-     */
-
-    if (fd == INVALID_HANDLE_VALUE) {
-	DWORD attr;
-
-	set_error(errInfo);
-
-	/*
-	 * If the error is EACESS, the reason could be that we tried to
-	 * open a directory.  In that case, we'll change the error code
-	 * to EISDIR.
-	 */
-	if (errInfo->posix_errno && 
-	    (attr = GetFileAttributesW(wname)) != INVALID_FILE_ATTRIBUTES && 
-	    (attr & FILE_ATTRIBUTE_DIRECTORY)) {
-	    errInfo->posix_errno = EISDIR;
-	}
-	return 0;
-    }
-
-    /*
-     * Get and return the length of the open file.
-     */
-       
-    if (!GetFileInformationByHandle(fd, &fileInfo))
-	return set_error(errInfo);
-    *pfd = (int) fd;
-    if (pSize) {
-	*pSize = (Sint64)
-	    (((Uint64)fileInfo.nFileSizeHigh << 32) | 
-	     (Uint64)fileInfo.nFileSizeLow);
-    }
-    return 1;
-}
-
-int 
-efile_may_openfile(Efile_error* errInfo, char *name)
-{
-    Efile_call_state state;
-    WCHAR *wname = (WCHAR *) name;
-    DWORD attr;
-    int ret;
-
-    DBG_TRACE(1, name);
-    call_state_init(&state, errInfo);
-    ensure_wpath(&state, &wname);
-    if ((attr = GetFileAttributesW(wname)) == INVALID_FILE_ATTRIBUTES) {
-	errno = ENOENT;
-	ret = check_error(-1, errInfo);
-    }
-    else if (attr & FILE_ATTRIBUTE_DIRECTORY) {
-	errno = EISDIR;
-	ret = check_error(-1, errInfo);
-    }
-    else ret = 1;
-
-    call_state_free(&state);
-    return ret;
-}
-
-void
-efile_closefile(fd)
-int fd;				/* File descriptor for file to close. */
-{
-    DBG_TRACE(2, L"");
-    CloseHandle((HANDLE) fd);
-}
-
-FILE* efile_wfopen(const WCHAR* name, const WCHAR* mode)
-{
-    Efile_call_state state;
-    Efile_error dummy;
-    FILE* f;
-    call_state_init(&state, &dummy);
-    ensure_wpath(&state, (WCHAR**)&name);
-    f = _wfopen(name, mode);
-    call_state_free(&state);
-    return f;
-}
-
-int
-efile_fdatasync(errInfo, fd)
-Efile_error* errInfo;		/* Where to return error codes. */
-int fd;				/* File descriptor for file to sync. */
-{
-    DBG_TRACE(2, L"");
-    /* Not available in Windows, just call regular fsync */
-    return efile_fsync(errInfo, fd);
-}
-
-int
-efile_fsync(errInfo, fd)
-Efile_error* errInfo;		/* Where to return error codes. */
-int fd;				/* File descriptor for file to sync. */
-{
-    DBG_TRACE(2, L"");
-    if (!FlushFileBuffers((HANDLE) fd)) {
-	return check_error(-1, errInfo);
-    }
-    return 1;
-}
-
-int
-efile_fileinfo(Efile_error* errInfo, Efile_info* pInfo,
-	       char* orig_name, int info_for_link)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(1, L"");
-    call_state_init(&state, errInfo);
-    ret = do_fileinfo(&state, pInfo, orig_name, info_for_link);
-    call_state_free(&state);
-    return ret;
-}
-
-static int
-do_fileinfo(Efile_call_state* state, Efile_info* pInfo,
-	    char* orig_name, int info_for_link)
-{
-    Efile_error* errInfo = state->errInfo;
-    HANDLE findhandle;		/* Handle returned by FindFirstFile(). */
-    WIN32_FIND_DATAW findbuf;	/* Data return by FindFirstFile(). */
-    WCHAR* name = NULL;
-    WCHAR* win_path;
-    int name_len;
-    int drive;			/* Drive for filename (1 = A:, 2 = B: etc). */
-    WCHAR *worig_name = (WCHAR *) orig_name;
-
-    ensure_wpath(state, &worig_name);
-    /* Don't allow wildcards to be interpreted by system */
-
-
-    /*
-     * Move the name to a buffer and make sure to remove a trailing
-     * slash, because it causes FindFirstFile() to fail on Win95.
-     */
-
-    name_len = wcslen(worig_name);
-
-    name = wpath_tmp_alloc(state, name_len+1);
-    wcscpy(name, worig_name);
-    if (name_len > 2 && ISSLASH(name[name_len-1]) &&
-	name[name_len-2] != L':') {
-	name[name_len-1] = L'\0';
-    }
-
-    win_path = name;
-    if (wcsncmp(name, L"\\\\?\\", 4) == 0) {
-	win_path += 4;
-    }
-
-    if (wcspbrk(win_path, L"?*")) {
-    enoent:
-	errInfo->posix_errno = ENOENT;
-	errInfo->os_errno = ERROR_FILE_NOT_FOUND;
-        return 0;
-    }
-
-    /* Try to get disk from name.  If none, get current disk.  */
-
-    if (win_path[1] != L':') {
-	WCHAR* cwd_path = get_cwd_wpath_tmp(state);
-        drive = 0;
-	if (cwd_path[1] == L':') {
-	    drive = towlower(cwd_path[0]) - L'a' + 1;
-	}
-    } else if (*win_path && win_path[2] == L'\0') {
-	/*
-	 * X: and nothing more is an error.
-	 */
-	errInfo->posix_errno = ENOENT;
-	errInfo->os_errno = ERROR_FILE_NOT_FOUND;
-	return 0;
-    } else {
-        drive = towlower(*win_path) - L'a' + 1;
-    }
-
-    findhandle = FindFirstFileW(name, &findbuf);
-    if (findhandle == INVALID_HANDLE_VALUE) {
-	WCHAR* path = NULL;
-
-        if (!(wcspbrk(name, L"./\\") &&
-	      (path = get_full_wpath_tmp(state, name, NULL, 0)) &&
-	      /* root dir. ('C:\') or UNC root dir. ('\\server\share\') */
-	      ((wcslen(path) == 3) || is_root_unc_name(path)) &&
-	      (GetDriveTypeW(path) > 1)   ) ) {
-
-	    errInfo->posix_errno = ENOENT;
-	    errInfo->os_errno = ERROR_FILE_NOT_FOUND;
-	    return 0;
-	}
-
-        /*
-         * Root directories (such as C:\ or \\server\share\ are fabricated.
-         */
-	
-        findbuf.dwFileAttributes = FILE_ATTRIBUTE_DIRECTORY;
-        findbuf.nFileSizeHigh = 0;
-        findbuf.nFileSizeLow = 0;
-        findbuf.cFileName[0] = L'\0';
-
-	pInfo->links = 1;
-	pInfo->cTime = pInfo->accessTime = pInfo->modifyTime = 0;
-    } else {
-	SYSTEMTIME SystemTime;
-        FILETIME LocalFTime;
-
-	/*first check if we are a symlink */
-	if (!info_for_link && (findbuf.dwFileAttributes &
-			       FILE_ATTRIBUTE_REPARSE_POINT)){
-	    /*
-	     * given that we know this is a symlink,
-	     we should be able to find its target */
-	    WCHAR* target_name = (WCHAR*) do_readlink(state, (char *) name, NULL, 0);
-	    if (target_name) {
-		FindClose(findhandle);
-		return do_fileinfo(state, pInfo,
-				   (char *) target_name, info_for_link);
-	    }
-	}
-
-	/* number of links: */
-	{
-	    HANDLE handle;	/* Handle returned by CreateFile() */
-	    BY_HANDLE_FILE_INFORMATION fileInfo; /* from  CreateFile() */
-
-            /* We initialise nNumberOfLinks as GetFileInformationByHandle
-               does not always initialise this field */
-            fileInfo.nNumberOfLinks = 1;
-	    if (handle = CreateFileW(name, GENERIC_READ, FILE_SHARE_FLAGS, NULL,
-				    OPEN_EXISTING, 0, NULL)) {
-		GetFileInformationByHandle(handle, &fileInfo);
-		pInfo->links = fileInfo.nNumberOfLinks;
-		CloseHandle(handle);
-	    } else {
-		pInfo->links = 1;
-	    }	
-	}
-
-        FILETIME_TO_EPOCH(pInfo->modifyTime, findbuf.ftLastWriteTime);
-
-        if (findbuf.ftLastAccessTime.dwLowDateTime == 0 &&
-	    findbuf.ftLastAccessTime.dwHighDateTime == 0) {
-	    pInfo->accessTime = pInfo->modifyTime;
-	} else {
-	    FILETIME_TO_EPOCH(pInfo->accessTime, findbuf.ftLastAccessTime);
-	}
-
-        if (findbuf.ftCreationTime.dwLowDateTime == 0 &&
-	    findbuf.ftCreationTime.dwHighDateTime == 0) {
-	    pInfo->cTime = pInfo->modifyTime;
-	} else {
-	    FILETIME_TO_EPOCH(pInfo->cTime ,findbuf.ftCreationTime);
-	}
-        FindClose(findhandle);
-    }
-
-    pInfo->size_low = findbuf.nFileSizeLow;
-    pInfo->size_high = findbuf.nFileSizeHigh;
-	
-    if (info_for_link && (findbuf.dwFileAttributes &
-			  FILE_ATTRIBUTE_REPARSE_POINT))
-	pInfo->type = FT_SYMLINK;
-    else if (findbuf.dwFileAttributes & FILE_ATTRIBUTE_DIRECTORY)
-	pInfo->type = FT_DIRECTORY;
-    else
-	pInfo->type = FT_REGULAR;
-
-    if (findbuf.dwFileAttributes & FILE_ATTRIBUTE_READONLY)
-	pInfo->access = FA_READ;
-    else
-	pInfo->access = FA_READ|FA_WRITE;
-
-    pInfo->mode = dos_to_posix_mode(findbuf.dwFileAttributes, name);
-    pInfo->major_device = drive;
-    pInfo->minor_device = 0;
-    pInfo->inode = 0;
-    pInfo->uid = 0;
-    pInfo->gid = 0;
-    
-    return 1;
-}
-
-int
-efile_write_info(Efile_error* errInfo,
-		 Efile_info* pInfo,
-		 char* name)
-{
-    Efile_call_state state;
-    int ret;
-    call_state_init(&state, errInfo);
-    ret = do_write_info(&state, pInfo, name);
-    call_state_free(&state);
-    return ret;
-}
-
-static int
-do_write_info(Efile_call_state* state,
-              Efile_info* pInfo,
-	      char* name)
-{
-    Efile_error* errInfo = state->errInfo;
-    SYSTEMTIME timebuf;
-    FILETIME ModifyFileTime;
-    FILETIME AccessFileTime;
-    FILETIME CreationFileTime;
-    HANDLE fd;
-    DWORD attr;
-    DWORD tempAttr;
-    WCHAR *wname = (WCHAR *) name;
-
-    DBG_TRACE(1, name);
-
-    ensure_wpath(state, &wname);
-
-    /*
-     * Get the attributes for the file.
-     */
-
-    tempAttr = attr = GetFileAttributesW(wname);
-    if (attr == 0xffffffff) {
-	return set_error(errInfo);
-    }
-    if (pInfo->mode != -1) {
-	if (pInfo->mode & _S_IWRITE) {
-	    /* clear read only bit */
-	    attr &= ~FILE_ATTRIBUTE_READONLY;
-	} else {
-	    /* set read only bit */
-	    attr |= FILE_ATTRIBUTE_READONLY;
-	}
-    }
-
-    /*
-     * Construct all file times.
-     */
-
-    EPOCH_TO_FILETIME(ModifyFileTime,   pInfo->modifyTime);
-    EPOCH_TO_FILETIME(AccessFileTime,   pInfo->accessTime);
-    EPOCH_TO_FILETIME(CreationFileTime, pInfo->cTime);
-
-    /*
-     * If necessary, set the file times.
-     */
-
-    /*
-     * If the has read only access, we must temporarily turn on
-     * write access (this is necessary for native filesystems,
-     * but not for NFS filesystems).
-     */
-
-    if (tempAttr & FILE_ATTRIBUTE_READONLY) {
-	tempAttr &= ~FILE_ATTRIBUTE_READONLY;
-	if (!SetFileAttributesW(wname, tempAttr)) {
-	    return set_error(errInfo);
-	}
-    }
-
-    fd = CreateFileW(wname, GENERIC_READ|GENERIC_WRITE,
-	    FILE_SHARE_FLAGS,
-	    NULL, OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL);
-    if (fd != INVALID_HANDLE_VALUE) {
-	BOOL result = SetFileTime(fd, &CreationFileTime, &AccessFileTime, &ModifyFileTime);
-	if (!result) {
-	    return set_error(errInfo);
-	}
-	CloseHandle(fd);
-    }
-
-    /*
-     * If the file doesn't have the correct attributes, set them now.
-     * (It could have been done before setting the file times, above).
-     */
-
-    if (tempAttr != attr) {
-	if (!SetFileAttributesW(wname, attr)) {
-	    return set_error(errInfo);
-	}
-    }
-    return 1;
-}
-
-
-int
-efile_pwrite(errInfo, fd, buf, count, offset)
-Efile_error* errInfo;		/* Where to return error codes. */
-int fd;				/* File descriptor to write to. */
-char* buf;			/* Buffer to write. */
-size_t count;			/* Number of bytes to write. */
-Sint64 offset;			/* where to write it */
-{
-    int res;
-    DBG_TRACE(2, L"");
-    res = efile_seek(errInfo, fd, offset, EFILE_SEEK_SET, NULL);
-    if (res) {
-	return efile_write(errInfo, EFILE_MODE_WRITE, fd, buf, count);
-    } else {
-	return res;
-    }
-}
-
-/* position and read/write as a single atomic op */
-int
-efile_pread(errInfo, fd, offset, buf, count, pBytesRead)
-Efile_error* errInfo;		/* Where to return error codes. */
-int fd;				/* File descriptor to read from. */
-Sint64 offset;			/* Offset in bytes from BOF. */
-char* buf;			/* Buffer to read into. */
-size_t count;			/* Number of bytes to read. */
-size_t* pBytesRead;		/* Where to return number of bytes read. */
-{
-    int res;
-    DBG_TRACE(2, L"");
-    res = efile_seek(errInfo, fd, offset, EFILE_SEEK_SET, NULL);
-    if (res) {
-	return efile_read(errInfo, EFILE_MODE_READ, fd, buf, count, pBytesRead);
-    } else {
-	return res;
-    }
-}
-
-
-
-int
-efile_write(errInfo, flags, fd, buf, count)
-Efile_error* errInfo;		/* Where to return error codes. */
-int flags;			/* Flags given when file was opened. */
-int fd;				/* File descriptor to write to. */
-char* buf;			/* Buffer to write. */
-size_t count;			/* Number of bytes to write. */
-{
-    DWORD written;		/* Bytes written in last operation. */
-    OVERLAPPED overlapped;
-    OVERLAPPED* pOverlapped = NULL;
-
-    DBG_TRACE(2, L"");
-    if (flags & EFILE_MODE_APPEND) {
-	memset(&overlapped, 0, sizeof(overlapped));
-	overlapped.Offset = 0xffffffff;
-	overlapped.OffsetHigh = 0xffffffff;
-	pOverlapped = &overlapped;
-    }
-    while (count > 0) {
-	if (!WriteFile((HANDLE) fd, buf, count, &written, pOverlapped))
-	    return set_error(errInfo);
-	buf += written;
-	count -= written;
-    }
-    return 1;
-}
-
-int
-efile_writev(Efile_error* errInfo,   /* Where to return error codes */
-	     int flags,              /* Flags given when file was
-				      * opened */
-	     int fd,                 /* File descriptor to write to */
-	     SysIOVec* iov,          /* Vector of buffer structs.
-				      * The structs are unchanged 
-				      * after the call */
-	     int iovcnt)             /* Number of structs in vector */
-{
-    int cnt;                         /* Buffers so far written */
-    OVERLAPPED overlapped;
-    OVERLAPPED* pOverlapped = NULL;
-
-    DBG_TRACE(2, L"");
-    ASSERT(iovcnt >= 0);
-    
-    if (flags & EFILE_MODE_APPEND) {
-	memset(&overlapped, 0, sizeof(overlapped));
-	overlapped.Offset = 0xffffffff;
-	overlapped.OffsetHigh = 0xffffffff;
-	pOverlapped = &overlapped;
-    }
-    for (cnt = 0; cnt < iovcnt; cnt++) {
-	if (iov[cnt].iov_base && iov[cnt].iov_len > 0) {
-	    /* Non-empty buffer */
-	    int p;                   /* Position in buffer */
-	    int w = iov[cnt].iov_len;/* Bytes written in this call */
-	    for (p = 0; p < iov[cnt].iov_len; p += w) {
-		if (!WriteFile((HANDLE) fd, 
-			       iov[cnt].iov_base + p, 
-			       iov[cnt].iov_len - p, 
-			       &w, 
-			       pOverlapped))
-		    return set_error(errInfo);
-	    }
-	}
-    }	
-    return 1;
-}
-
-int
-efile_read(errInfo, flags, fd, buf, count, pBytesRead)
-Efile_error* errInfo;		/* Where to return error codes. */
-int flags;			/* Flags given when file was opened. */
-int fd;				/* File descriptor to read from. */
-char* buf;			/* Buffer to read into. */
-size_t count;			/* Number of bytes to read. */
-size_t* pBytesRead;		/* Where to return number of bytes read. */
-{
-    DWORD nbytes = 0;
-
-    DBG_TRACE(2, L"");
-    if (!ReadFile((HANDLE) fd, buf, count, &nbytes, NULL))
-	return set_error(errInfo);
-
-    *pBytesRead = nbytes;
-    return 1;
-}
-
-int
-efile_seek(errInfo, fd, offset, origin, new_location)
-Efile_error* errInfo;		/* Where to return error codes. */
-int fd;				/* File descriptor to do the seek on. */
-Sint64 offset;			/* Offset in bytes from the given origin. */
-int origin;			/* Origin of seek (SEEK_SET, SEEK_CUR,
-				 * SEEK_END).
-				 */
-Sint64* new_location;		/* Resulting new location in file. */
-{
-    LARGE_INTEGER off, new_loc;
-    
-    DBG_TRACE(2, L"");
-    switch (origin) {
-    case EFILE_SEEK_SET: origin = FILE_BEGIN; break;
-    case EFILE_SEEK_CUR: origin = FILE_CURRENT; break;
-    case EFILE_SEEK_END: origin = FILE_END; break;
-    default:
-	errno = EINVAL;
-	check_error(-1, errInfo);
-	break;
-    }
-    
-    off.QuadPart = offset;
-    if (! SetFilePointerEx((HANDLE) fd, off,
-	new_location ? &new_loc : NULL, origin)) {
-	return set_error(errInfo);
-    }
-    if (new_location) {
-	*new_location = new_loc.QuadPart;
-	DEBUGF(("efile_seek(offset=%ld, origin=%d) -> %ld\n", 
-		(long) offset, origin, (long) *new_location));
-    } else {
-	DEBUGF(("efile_seek(offset=%ld, origin=%d)\n", (long) offset, origin));
-    }
-    return 1;
-}
-
-int
-efile_truncate_file(errInfo, fd, flags)
-Efile_error* errInfo;		/* Where to return error codes. */
-int *fd;				/* File descriptor for file to truncate. */
-int flags;
-{
-    DBG_TRACE(2, L"");
-    if (!SetEndOfFile((HANDLE) (*fd)))
-	return set_error(errInfo);
-    return 1;
-}
-
-
-/*
- * is_root_unc_name - returns TRUE if the argument is a UNC name specifying
- *      a root share.  That is, if it is of the form \\server\share\.
- *      This routine will also return true if the argument is of the
- *      form \\server\share (no trailing slash) but Win32 currently
- *      does not like that form.
- *
- *      Forward slashes ('/') may be used instead of backslashes ('\').
- */
-
-static int
-is_root_unc_name(const WCHAR *path)
-{
-    /*
-     * If a root UNC name, path will start with 2 (but not 3) slashes
-     */
-
-    if ((wcslen(path) >= 5) /* minimum string is "//x/y" */
-	&& ISSLASH(path[0]) && ISSLASH(path[1]))
-    {
-        const WCHAR *p = path + 2;
-
-        /*
-         * find the slash between the server name and share name
-         */
-        while ( * ++ p )
-            if ( ISSLASH(*p) )
-                break ;
-
-        if ( *p && p[1] )
-        {
-            /*
-             * is there a further slash?
-             */
-            while ( * ++ p )
-                if ( ISSLASH(*p) )
-                    break ;
-
-            /*
-             * just final slash (or no final slash)
-             */
-            if ( !*p || !p[1])
-                return 1;
-        }
-    }
-
-    return 0 ;
-}
-
-/*
- * Extracts the root part of an absolute filename (by modifying the string
- * pointed to by the name argument).  The name can start
- * with either a driver letter (for example, C:\), or a UNC name
- * (for example, \\guinness\bjorn).
- *
- * If the name is invalid, the buffer will be modified to point to
- * an empty string.
- *
- * Returns: 1 if the name consists of just the root part, 0 if
- * 	    the name was longer.
- */
-
-static int
-extract_root(WCHAR* name)
-{
-    int len = wcslen(name);
-
-    if (iswalpha(name[0]) && name[1] == L':' && ISSLASH(name[2])) {
-	WCHAR c = name[3];
-	name[3] = L'\0';
-	return c == L'\0';
-    } else if (len < 5 || !ISSLASH(name[0]) || !ISSLASH(name[1])) {
-	goto error;
-    } else {			/* Try to find the end of the UNC name. */
-	WCHAR* p;
-	WCHAR c;
-
-        /*
-         * Find the slash between the server name and share name.
-         */
-
-	for (p = name + 2; *p; p++)
-            if (ISSLASH(*p))
-                break;
-	if (*p == L'\0')
-	    goto error;
-
-	/*
-	 * Find the slash after the share name.
-	 */
-
-	for (p++; *p; p++)
-            if (ISSLASH(*p))
-                break;
-	c = *p;
-	*p = L'\0';
-	return c == L'\0' || p[1] == L'\0';
-    }
-
- error:
-    *name = L'\0';
-    return 1;
-}
-
-static unsigned short
-dos_to_posix_mode(int attr, const WCHAR *name)
-{
-    register unsigned short uxmode;
-    unsigned dosmode;
-    register const WCHAR *p;
-
-    dosmode = attr & 0xff;
-    if ((p = name)[1] == L':')
-	p += 2;
-
-    /* check to see if this is a directory - note we must make a special
-     * check for the root, which DOS thinks is not a directory
-     */
-
-    uxmode = (unsigned short)
-	(((ISSLASH(*p) && !p[1]) || (dosmode & FILE_ATTRIBUTE_DIRECTORY) ||
-	       *p == L'\0') ? _S_IFDIR|_S_IEXEC : _S_IFREG);
-
-    /* If attribute byte does not have read-only bit, it is read-write */
-
-    uxmode |= (dosmode & FILE_ATTRIBUTE_READONLY) ?
-	_S_IREAD : (_S_IREAD|_S_IWRITE);
-
-    /* see if file appears to be executable - check extension of name */
-
-    if (p = wcsrchr(name, L'.')) {
-        if (!_wcsicmp(p, L".exe") ||
-	    !_wcsicmp(p, L".cmd") ||
-	    !_wcsicmp(p, L".bat") ||
-	    !_wcsicmp(p, L".com"))
-            uxmode |= _S_IEXEC;
-    }
-
-    /* propagate user read/write/execute bits to group/other fields */
-
-    uxmode |= (uxmode & 0700) >> 3;
-    uxmode |= (uxmode & 0700) >> 6;
-
-    return uxmode;
-}
-
-
-int
-efile_readlink(Efile_error* errInfo, char* name, char* buffer, size_t size)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(1, name);
-    call_state_init(&state, errInfo);
-    ret = !!do_readlink(&state, name, buffer, size);
-    call_state_free(&state);
-    return ret;
-}
-
-/* If buffer==0, return buffer allocated by wpath_tmp_allocate
-*/
-static char*
-do_readlink(Efile_call_state* state, char* name, char* buffer, size_t size)
-{
-    /*
-     * load dll and see if we have CreateSymbolicLink at runtime:
-     * (Vista only)
-     */
-    HINSTANCE hModule = NULL;
-    WCHAR *wname = (WCHAR *) name;
-    WCHAR *wbuffer = (WCHAR *) buffer;
-    DWORD wsize = size / sizeof(WCHAR);
-    char* ret = NULL;
-
-    if ((hModule = LoadLibrary("kernel32.dll")) != NULL) {
-	typedef DWORD (WINAPI * GETFINALPATHNAMEBYHANDLEPTR)(
-							     HANDLE hFile,
-							     LPCWSTR lpFilePath,
-							     DWORD cchFilePath,
-							     DWORD dwFlags);
-
-	GETFINALPATHNAMEBYHANDLEPTR pGetFinalPathNameByHandle =
-	    (GETFINALPATHNAMEBYHANDLEPTR)GetProcAddress(hModule, "GetFinalPathNameByHandleW");
-
-	if (pGetFinalPathNameByHandle != NULL) {
-	    DWORD fileAttributes;
-	    ensure_wpath(state, &wname);
-	    /* first check if file is a symlink; {error, einval} otherwise */
-	    fileAttributes = GetFileAttributesW(wname);
-	    if ((fileAttributes & FILE_ATTRIBUTE_REPARSE_POINT)) {
-		DWORD success = 0;
-		HANDLE h = CreateFileW(wname, GENERIC_READ, FILE_SHARE_FLAGS, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
-		int len;
-		if(h != INVALID_HANDLE_VALUE) {
-		    if (!wbuffer) { /* dynamic allocation */
-			WCHAR dummy;
-			wsize = pGetFinalPathNameByHandle(h, &dummy, 0, 0);
-			if (wsize) {
-			    wbuffer = wpath_tmp_alloc(state, wsize);
-			}
-		    }
-		    if (wbuffer
-			&& (success = pGetFinalPathNameByHandle(h, wbuffer, wsize, 0))
-			&& success < wsize) {
-			WCHAR* wp;
-
-			/* GetFinalPathNameByHandle prepends path with "\\?\": */
-			len = wcslen(wbuffer);
-			wmemmove(wbuffer,wbuffer+4,len-3);
-			if (len - 4 >= 2 && wbuffer[1] == L':' && wbuffer[0] >= L'A' &&
-			    wbuffer[0] <= L'Z') {
-			    wbuffer[0] = wbuffer[0] + L'a' - L'A';
-			}
-			
-			for (wp=wbuffer ; *wp; wp++)
-			    if (*wp == L'\\')
-				*wp = L'/';
-		    }
-		    CloseHandle(h);
-		}
-		if (success) {
-		    ret = (char*) wbuffer;
-		} else {
-		    set_error(state->errInfo);
-		}
-	    } else {
-		errno = EINVAL;
-		save_last_error(state->errInfo);
-	    }
-	    goto done;
-	}
-    }
-    errno = ENOTSUP;
-    save_last_error(state->errInfo);
-
-done:
-    if (hModule)
-	FreeLibrary(hModule);
-    return ret;
-}
-
-
-int
-efile_altname(Efile_error* errInfo, char* orig_name, char* buffer, size_t size)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE(1, orig_name);
-    call_state_init(&state, errInfo);
-    ret = do_altname(&state, orig_name, buffer, size);
-    call_state_free(&state);
-    return ret;
-}
-
-static int
-do_altname(Efile_call_state* state, char* orig_name, char* buffer, size_t size)
-{
-    WIN32_FIND_DATAW wfd;
-    HANDLE fh;
-    WCHAR* name;
-    int name_len;
-    WCHAR* full_path = NULL;
-    WCHAR *worig_name = (WCHAR *) orig_name;
-    WCHAR *wbuffer = (WCHAR *) buffer;
-    int drive;			/* Drive for filename (1 = A:, 2 = B: etc). */
-
-    /* Don't allow wildcards to be interpreted by system */
-
-    if (wcspbrk(worig_name, L"?*")) {
-    enoent:
-	state->errInfo->posix_errno = ENOENT;
-	state->errInfo->os_errno = ERROR_FILE_NOT_FOUND;
-        return 0;
-    }
-
-    /*
-     * Move the name to a buffer and make sure to remove a trailing
-     * slash, because it causes FindFirstFile() to fail on Win95.
-     */
-    ensure_wpath(state, &worig_name);
-    name_len = wcslen(worig_name);
-
-    name = wpath_tmp_alloc(state, name_len + 1);
-    wcscpy(name, worig_name);
-    if (name_len > 2 && ISSLASH(name[name_len-1]) &&
-	name[name_len-2] != L':') {
-	name[name_len-1] = L'\0';
-    }
-    
-    /* Try to get disk from name.  If none, get current disk.  */
-
-    if (name[1] != L':') {
-	WCHAR* cwd_path = get_cwd_wpath_tmp(state);
-        drive = 0;
-        if (cwd_path[1] == L':') {
-	    drive = towlower(cwd_path[0]) - L'a' + 1;
-	}
-    } else if (*name && name[2] == L'\0') {
-	/*
-	 * X: and nothing more is an error.
-	 */
-	goto enoent;
-    } else {
-        drive = towlower(*name) - L'a' + 1;
-    }
-    fh = FindFirstFileW(name,&wfd);
-    if (fh == INVALID_HANDLE_VALUE) {
-	DWORD fff_error = GetLastError();
-        if (!(wcspbrk(name, L"./\\") &&
-	      (full_path = get_full_wpath_tmp(state, name, NULL, 0)) &&
-	      /* root dir. ('C:\') or UNC root dir. ('\\server\share\') */
-	      ((wcslen(full_path) == 3) || is_root_unc_name(full_path)) &&
-	      (GetDriveTypeW(full_path) > 1)   ) ) {
-
-	    set_os_errno(state->errInfo, fff_error);
-	    return 0;
-	}
-        /*
-         * Root directories (such as C:\ or \\server\share\ are fabricated.
-         */
-	wcscpy(wbuffer,name);
-	return 1;
-    }
-	
-    wcscpy(wbuffer,wfd.cAlternateFileName);
-    if (!*wbuffer) {
-	wcscpy(wbuffer,wfd.cFileName);
-    }
-    FindClose(fh);
-    return 1;
-}
-
-
-int
-efile_link(Efile_error* errInfo, char* old, char* new)
-{
-    Efile_call_state state;
-    WCHAR *wold = (WCHAR *) old;
-    WCHAR *wnew = (WCHAR *) new;
-    int ret;
-    DBG_TRACE(1, old);
-    call_state_init(&state, errInfo);
-    ensure_wpath(&state, &wold);
-    ensure_wpath(&state, &wnew);
-    if(!CreateHardLinkW(wnew, wold, NULL)) {
-	ret = set_error(errInfo);
-    }
-    else ret =1;
-    call_state_free(&state);
-    return ret;
-}
-
-int
-efile_symlink(Efile_error* errInfo, char* old, char* new)
-{
-    Efile_call_state state;
-    int ret;
-    DBG_TRACE2(1, "symlink(%s <- %s)", old, new);
-    call_state_init(&state, errInfo);
-    ret = do_symlink(&state, old, new);
-    call_state_free(&state);
-    return ret;
-}
-
-static int
-do_symlink(Efile_call_state* state, char* old, char* new)
-{
-    /*
-     * Load dll and see if we have CreateSymbolicLink at runtime:
-     * (Vista only)
-     */
-    HINSTANCE hModule = NULL;
-    WCHAR *wold = (WCHAR *) old;
-    WCHAR *wnew = (WCHAR *) new;
-
-    DBG_TRACE(1, old);
-    if ((hModule = LoadLibrary("kernel32.dll")) != NULL) {
-	typedef BOOLEAN (WINAPI  * CREATESYMBOLICLINKFUNCPTR) (
-	     LPCWSTR lpSymlinkFileName,
-	     LPCWSTR lpTargetFileName,
-	     DWORD dwFlags);
-
-	CREATESYMBOLICLINKFUNCPTR pCreateSymbolicLink =
-	    (CREATESYMBOLICLINKFUNCPTR) GetProcAddress(hModule,
-						       "CreateSymbolicLinkW");
-	/* A for MBCS, W for UNICODE... char* above implies 'W'! */
-	if (pCreateSymbolicLink != NULL) {
-	  ensure_wpath(state, &wold);
-	  ensure_wpath(state, &wnew);
-	  {
-	    DWORD attr = GetFileAttributesW(wold);
-	    int flag = (attr != INVALID_FILE_ATTRIBUTES &&
-			attr & FILE_ATTRIBUTE_DIRECTORY) ? 1 : 0;
-	    /*  SYMBOLIC_LINK_FLAG_DIRECTORY = 1 */
-	    BOOLEAN success = pCreateSymbolicLink(wnew, wold, flag);
-	    FreeLibrary(hModule);
-
-	    if (success) {
-		return 1;
-	    } else {
-		return set_error(state->errInfo);
-	    }
-	  }
-	} else
-	    FreeLibrary(hModule);
-    }
-    errno = ENOTSUP;
-    return check_error(-1, state->errInfo);
-}
-
-int
-efile_fadvise(Efile_error* errInfo, int fd, Sint64 offset,
-	    Sint64 length, int advise)
-{
-    DBG_TRACE(2, L"");
-    /* posix_fadvise is not available on Windows, do nothing */
-    errno = ERROR_SUCCESS;
-    return check_error(0, errInfo);
-}
-
-int
-efile_fallocate(Efile_error* errInfo, int fd, Sint64 offset, Sint64 length)
-{
-    DBG_TRACE(2, L"");
-    /* No file preallocation method available in Windows. */
-    errno = errno_map(ERROR_NOT_SUPPORTED);
-    SetLastError(ERROR_NOT_SUPPORTED);
-
-    return check_error(-1, errInfo);
-}
diff --git a/erts/emulator/hipe/hipe_amd64.c b/erts/emulator/hipe/hipe_amd64.c
index e3cff4a4ba..f23f341e6d 100644
--- a/erts/emulator/hipe/hipe_amd64.c
+++ b/erts/emulator/hipe/hipe_amd64.c
@@ -28,6 +28,7 @@
 #include "error.h"
 #include "bif.h"
 #include "big.h"	/* term_to_Sint() */
+#include "erl_binary.h"
 
 #include "hipe_arch.h"
 #include "hipe_bif0.h"
@@ -38,6 +39,8 @@
 #undef ERL_FUN_SIZE
 #include "hipe_literals.h"
 
+static void patch_trampoline(void *trampoline, void *destAddress);
+
 const Uint sse2_fnegate_mask[2] = {0x8000000000000000,0};
 
 void hipe_patch_load_fe(Uint64 *address, Uint64 value)
@@ -52,9 +55,9 @@ int hipe_patch_insn(void *address, Uint64 value, Eterm type)
     switch (type) {
       case am_closure:
       case am_constant:
+      case am_c_const:
 	*(Uint64*)address = value;
 	break;
-      case am_c_const:
       case am_atom:
 	/* check that value fits in an unsigned imm32 */
 	/* XXX: are we sure it's not really a signed imm32? */
@@ -71,14 +74,18 @@ int hipe_patch_insn(void *address, Uint64 value, Eterm type)
 
 int hipe_patch_call(void *callAddress, void *destAddress, void *trampoline)
 {
-    Sint rel32;
+    Sint64 destOffset = (Sint64)destAddress - (Sint64)callAddress - 4;
 
-    ASSERT(trampoline == NULL);
+    if ((destOffset < -0x80000000L) || (destOffset >= 0x80000000L)) {
+        destOffset = (Sint64)trampoline - (Sint64)callAddress - 4;
 
-    rel32 = (Sint)destAddress - (Sint)callAddress - 4;
-    if ((Sint)(Sint32)rel32 != rel32)
-	return -1;
-    *(Uint32*)callAddress = (Uint32)rel32;
+        if ((destOffset < -0x80000000L) || (destOffset >= 0x80000000L))
+            return -1;
+
+        patch_trampoline(trampoline, destAddress);
+    }
+
+    *(Uint32*)callAddress = (Uint32)destOffset;
     hipe_flush_icache_word(callAddress);
     return 0;
 }
@@ -96,12 +103,80 @@ static void *alloc_code(unsigned int alloc_bytes)
     return erts_alloc(ERTS_ALC_T_HIPE_EXEC, alloc_bytes);
 }
 
+static int check_callees(Eterm callees)
+{
+    Eterm *tuple;
+    Uint arity;
+    Uint i;
+
+    if (is_not_tuple(callees))
+	return -1;
+    tuple = tuple_val(callees);
+    arity = arityval(tuple[0]);
+    for (i = 1; i <= arity; ++i) {
+	Eterm mfa = tuple[i];
+	if (is_atom(mfa))
+	    continue;
+	if (is_not_tuple(mfa) ||
+	    tuple_val(mfa)[0] != make_arityval(3) ||
+	    is_not_atom(tuple_val(mfa)[1]) ||
+	    is_not_atom(tuple_val(mfa)[2]) ||
+	    is_not_small(tuple_val(mfa)[3]) ||
+	    unsigned_val(tuple_val(mfa)[3]) > 255)
+	    return -1;
+    }
+    return arity;
+}
+
+#define TRAMPOLINE_BYTES 12
+
+static void generate_trampolines(unsigned char *address,
+                                 int nrcallees, Eterm callees,
+                                 unsigned char **trampvec)
+{
+    unsigned char *trampoline = address;
+    int i;
+
+    for(i = 0; i < nrcallees; ++i) {
+        trampoline[0] = 0x48;           /* movabsq $..., %rax; */
+        trampoline[1] = 0xb8;
+        *(void**)(trampoline+2) = NULL; /* callee's address */
+        trampoline[10] = 0xff;          /* jmpq *%rax */
+        trampoline[11] = 0xe0;
+        trampvec[i] = trampoline;
+        trampoline += TRAMPOLINE_BYTES;
+    }
+    hipe_flush_icache_range(address, nrcallees*TRAMPOLINE_BYTES);
+}
+
+static void patch_trampoline(void *trampoline, void *destAddress)
+{
+    unsigned char *tp = (unsigned char*) trampoline;
+
+    ASSERT(tp[0] == 0x48 && tp[1] == 0xb8);
+
+    *(void**)(tp+2) = destAddress; /* callee's address */
+    hipe_flush_icache_word(tp+2);
+}
+
 void *hipe_alloc_code(Uint nrbytes, Eterm callees, Eterm *trampolines, Process *p)
 {
-    if (is_not_nil(callees))
+    int nrcallees;
+    Eterm trampvecbin;
+    unsigned char **trampvec;
+    unsigned char *address;
+
+    nrcallees = check_callees(callees);
+    if (nrcallees < 0)
 	return NULL;
-    *trampolines = NIL;
-    return alloc_code(nrbytes);
+
+    trampvecbin = new_binary(p, NULL, nrcallees*sizeof(unsigned char*));
+    trampvec = (unsigned char **)binary_bytes(trampvecbin);
+
+    address = alloc_code(nrbytes + nrcallees*TRAMPOLINE_BYTES);
+    generate_trampolines(address + nrbytes, nrcallees, callees, trampvec);
+    *trampolines = trampvecbin;
+    return address;
 }
 
 void hipe_free_code(void* code, unsigned int bytes)
@@ -129,10 +204,9 @@ void *hipe_make_native_stub(void *callee_exp, unsigned int beamArity)
      */
     unsigned int codeSize;
     unsigned char *code, *codep;
-    unsigned int callEmuOffset;
 
-    codeSize =	/* 23, 26, 29, or 32 bytes */
-      23 +	/* 23 when all offsets are 8-bit */
+    codeSize =	/* 30, 33, 36, or 39 bytes */
+      30 +	/* 30 when all offsets are 8-bit */
       (P_CALLEE_EXP >= 128 ? 3 : 0) +
       ((P_CALLEE_EXP + 4) >= 128 ? 3 : 0) +
       (P_ARITY >= 128 ? 3 : 0);
@@ -197,14 +271,15 @@ void *hipe_make_native_stub(void *callee_exp, unsigned int beamArity)
     codep[0] = beamArity;
     codep += 1;
 
-    /* jmp callemu; 5 bytes */
-    callEmuOffset = (unsigned char*)nbif_callemu - (code + codeSize);
-    codep[0] = 0xe9;
-    codep[1] =  callEmuOffset        & 0xFF;
-    codep[2] = (callEmuOffset >>  8) & 0xFF;
-    codep[3] = (callEmuOffset >> 16) & 0xFF;
-    codep[4] = (callEmuOffset >> 24) & 0xFF;
-    codep += 5;
+    /* jmp callemu; 12 bytes */
+    codep[0] = 0x48;
+    codep[1] = 0xb8;
+    codep += 2;
+    *(Uint64*)codep = (Uint64)nbif_callemu;
+    codep += 8;
+    codep[0] = 0xff;
+    codep[1] = 0xe0;
+    codep += 2;
 
     ASSERT(codep == code + codeSize);
 
diff --git a/erts/emulator/hipe/hipe_amd64_bifs.m4 b/erts/emulator/hipe/hipe_amd64_bifs.m4
index aff10f1528..cf4c59c9af 100644
--- a/erts/emulator/hipe/hipe_amd64_bifs.m4
+++ b/erts/emulator/hipe/hipe_amd64_bifs.m4
@@ -39,7 +39,7 @@ define(HANDLE_GOT_MBUF,`
 3:	call nbif_$1_gc_after_bif	/* `HANDLE_GOT_MBUF' */
 	jmp 2b')
 
-`#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+`#if defined(ERTS_ENABLE_LOCK_CHECK)
 #  define CALL_BIF(F) \
 		movq CSYM(nbif_impl_##F)@GOTPCREL(%rip), %r11; \
 		movq %r11, P_BIF_CALLEE(P); \
@@ -462,42 +462,6 @@ ASYM($1):
 	TYPE_FUNCTION(ASYM($1))
 #endif')
 
-/*
- * nogc_bif_interface_1(nbif_name, cbif_name)
- *
- * Generate native interface for a bif with implicit P
- * The bif can fail but cannot do GC.
- */
-
-define(nogc_bif_interface_1,
-`
-#ifndef HAVE_$1
-#`define' HAVE_$1
-	TEXT
-	.align	4
-	GLOBAL(ASYM($1))
-ASYM($1):
-	/* set up the parameters */
-	movq	P, %rdi
-	NBIF_ARG(%rsi,1,0)
-
-	/* make the call on the C stack */
-	SWITCH_ERLANG_TO_C
-	pushq	%rsi
-	movq	%rsp, %rsi	/* Eterm* BIF__ARGS */
-	sub	$(8), %rsp	/* stack frame 16-byte alignment */
-	CALL_BIF($2)
-	add	$(1*8 + 8), %rsp
-	SWITCH_C_TO_ERLANG
-
-	/* throw exception if failure, otherwise return */
-	TEST_GOT_EXN
-	jz	nbif_1_simple_exception
-	NBIF_RET(1)
-	SET_SIZE(ASYM($1))
-	TYPE_FUNCTION(ASYM($1))
-#endif')
-
 
 /*
  * noproc_primop_interface_0(nbif_name, cbif_name)
diff --git a/erts/emulator/hipe/hipe_arm_bifs.m4 b/erts/emulator/hipe/hipe_arm_bifs.m4
index a9097dabde..554faa2567 100644
--- a/erts/emulator/hipe/hipe_arm_bifs.m4
+++ b/erts/emulator/hipe/hipe_arm_bifs.m4
@@ -29,7 +29,7 @@ include(`hipe/hipe_arm_asm.m4')
 	.p2align 2
 	.arm
 
-`#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+`#if defined(ERTS_ENABLE_LOCK_CHECK)
 #  define CALL_BIF(F)	ldr r14, =nbif_impl_##F; str r14, [r0, #P_BIF_CALLEE]; bl hipe_debug_bif_wrapper
 #else
 #  define CALL_BIF(F)	bl	nbif_impl_##F
diff --git a/erts/emulator/hipe/hipe_bif0.c b/erts/emulator/hipe/hipe_bif0.c
index d8427a4c36..a8be64e08d 100644
--- a/erts/emulator/hipe/hipe_bif0.c
+++ b/erts/emulator/hipe/hipe_bif0.c
@@ -53,8 +53,6 @@
 #include "hipe_literals.h"
 #endif
 
-#define BeamOpCode(Op)	((Uint)BeamOp(Op))
-
 
 int term_to_Sint32(Eterm term, Sint *sp)
 {
@@ -615,7 +613,7 @@ static ErtsCodeInfo* hipe_find_emu_address(Eterm mod, Eterm name, unsigned int a
     n = code_hdr->num_functions;
     for (i = 0; i < n; ++i) {
 	ErtsCodeInfo *ci = code_hdr->functions[i];
-	ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
+	ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
 	if (ci->mfa.function == name && ci->mfa.arity == arity)
 	    return ci;
     }
@@ -1000,7 +998,7 @@ BIF_RETTYPE hipe_bifs_set_native_address_in_fe_2(BIF_ALIST_2)
 	BIF_ERROR(BIF_P, BADARG);
 
     fe->native_address = native_address;
-    if (erts_smp_refc_dectest(&fe->refc, 0) == 0)
+    if (erts_refc_dectest(&fe->refc, 0) == 0)
 	erts_erase_fun_entry(fe);
     BIF_RET(am_true);
 }
@@ -1048,7 +1046,7 @@ static struct {
      * they create a new stub for the mfa, which forces locking.
      * XXX: Redesign apply et al to avoid those updates.
      */
-    erts_smp_rwmtx_t lock;
+    erts_rwmtx_t lock;
 } hipe_mfa_info_table;
 
 Hash mod2mfa_tab; /* map from module atom to list of hipe_mfa_info */
@@ -1114,7 +1112,7 @@ static struct hipe_mfa_info* mod2mfa_put(struct hipe_mfa_info* mfa)
 struct hipe_ref {
     struct hipe_ref_head head;    /* list of refs to same calleee */
     void *address;
-#if defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
+#if defined(__x86_64__) || defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
     void *trampoline;
 #endif
     unsigned int flags;
@@ -1129,28 +1127,28 @@ struct hipe_ref {
 
 static inline void hipe_mfa_info_table_init_lock(void)
 {
-    erts_smp_rwmtx_init(&hipe_mfa_info_table.lock, "hipe_mfait_lock", NIL,
+    erts_rwmtx_init(&hipe_mfa_info_table.lock, "hipe_mfait_lock", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 }
 
 static inline void hipe_mfa_info_table_rlock(void)
 {
-    erts_smp_rwmtx_rlock(&hipe_mfa_info_table.lock);
+    erts_rwmtx_rlock(&hipe_mfa_info_table.lock);
 }
 
 static inline void hipe_mfa_info_table_runlock(void)
 {
-    erts_smp_rwmtx_runlock(&hipe_mfa_info_table.lock);
+    erts_rwmtx_runlock(&hipe_mfa_info_table.lock);
 }
 
 static inline void hipe_mfa_info_table_rwlock(void)
 {
-    erts_smp_rwmtx_rwlock(&hipe_mfa_info_table.lock);
+    erts_rwmtx_rwlock(&hipe_mfa_info_table.lock);
 }
 
 static inline void hipe_mfa_info_table_rwunlock(void)
 {
-    erts_smp_rwmtx_rwunlock(&hipe_mfa_info_table.lock);
+    erts_rwmtx_rwunlock(&hipe_mfa_info_table.lock);
 }
 
 static ERTS_INLINE
@@ -1545,7 +1543,7 @@ BIF_RETTYPE hipe_bifs_add_ref_2(BIF_ALIST_2)
 
     ref = erts_alloc(ERTS_ALC_T_HIPE_LL, sizeof(struct hipe_ref));
     ref->address = address;
-#if defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
+#if defined(__x86_64__) || defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
     ref->trampoline = trampoline;
 #endif
     ref->flags = flags;
@@ -1636,7 +1634,7 @@ void hipe_purge_refs(struct hipe_ref* first_ref, Eterm caller_module,
 {
     struct hipe_ref* ref = first_ref;
 
-    ERTS_SMP_LC_ASSERT(is_blocking == erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(is_blocking == erts_thr_progress_is_blocking());
 
     while (ref) {
         struct hipe_ref* free_ref = ref;
@@ -1682,9 +1680,9 @@ void hipe_purge_sdescs(struct hipe_sdesc* first_sdesc, Eterm module,
 {
     struct hipe_sdesc* sdesc = first_sdesc;
 
-    ERTS_SMP_LC_ASSERT(is_blocking == erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(is_blocking == erts_thr_progress_is_blocking());
 
-    ERTS_SMP_LC_ASSERT(is_blocking); /*XXX Fix safe sdesc destruction */
+    ERTS_LC_ASSERT(is_blocking); /*XXX Fix safe sdesc destruction */
 
     while (sdesc) {
         struct hipe_sdesc* free_sdesc = sdesc;
@@ -1702,7 +1700,7 @@ void hipe_purge_module(Module* modp, int is_blocking)
 {
     ASSERT(modp);
 
-    ERTS_SMP_LC_ASSERT(is_blocking == erts_smp_thr_progress_is_blocking());
+    ERTS_LC_ASSERT(is_blocking == erts_thr_progress_is_blocking());
 
     DBG_TRACE_MFA(make_atom(modp->module), 0, 0, "hipe_purge_module");
 
@@ -1711,7 +1709,7 @@ void hipe_purge_module(Module* modp, int is_blocking)
          * Remove all hipe_ref's (external calls) from the old module instance
          */
         if (modp->old.hipe_code->first_hipe_ref) {
-            ERTS_SMP_LC_ASSERT(is_blocking);
+            ERTS_LC_ASSERT(is_blocking);
 
             hipe_purge_refs(modp->old.hipe_code->first_hipe_ref,
                             make_atom(modp->module), is_blocking);
@@ -1722,7 +1720,7 @@ void hipe_purge_module(Module* modp, int is_blocking)
          * Remove all hipe_sdesc's for the old module instance
          */
         if (modp->old.hipe_code->first_hipe_sdesc) {
-            ERTS_SMP_LC_ASSERT(is_blocking);
+            ERTS_LC_ASSERT(is_blocking);
 
             hipe_purge_sdescs(modp->old.hipe_code->first_hipe_sdesc,
                              make_atom(modp->module), is_blocking);
@@ -1773,8 +1771,8 @@ void hipe_redirect_to_module(Module* modp)
     struct hipe_mfa_info *p;
     struct hipe_ref_head* refh;
 
-    ERTS_SMP_LC_ASSERT(erts_smp_thr_progress_is_blocking() ||
-                       erts_is_multi_scheduling_blocked());
+    ERTS_LC_ASSERT(erts_thr_progress_is_blocking() ||
+                   erts_is_multi_scheduling_blocked());
 
     for (p = mod2mfa_get(modp); p; p = p->next_in_mod) {
         if (p->new_address) {
@@ -1821,7 +1819,7 @@ void hipe_redirect_to_module(Module* modp)
 	    if (ref->flags & REF_FLAG_IS_LOAD_MFA)
 		res = hipe_patch_insn(ref->address, (Uint)p->remote_address, am_load_mfa);
 	    else {
-#if defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
+#if defined(__x86_64__) || defined(__arm__) || defined(__powerpc__) || defined(__ppc__) || defined(__powerpc64__)
                 void* trampoline = ref->trampoline;
 #else
                 void* trampoline = NULL;
diff --git a/erts/emulator/hipe/hipe_bif0.tab b/erts/emulator/hipe/hipe_bif0.tab
index 4f73770d24..6728e20123 100644
--- a/erts/emulator/hipe/hipe_bif0.tab
+++ b/erts/emulator/hipe/hipe_bif0.tab
@@ -109,6 +109,7 @@ atom suspend_0
 atom gc_1
 atom hipe_apply
 atom rethrow
+atom raw_raise
 atom find_na_or_make_stub
 atom nonclosure_address
 atom atomic_inc
@@ -135,7 +136,6 @@ atom bs_utf16_size
 atom bs_put_utf16be
 atom bs_put_utf16le
 atom bs_get_utf16
-atom bs_validate_unicode
 atom bs_validate_unicode_retract
 atom emulate_fpe
 atom emasculate_binary
diff --git a/erts/emulator/hipe/hipe_bif1.c b/erts/emulator/hipe/hipe_bif1.c
index 3d3df4fd48..73d07f0ce5 100644
--- a/erts/emulator/hipe/hipe_bif1.c
+++ b/erts/emulator/hipe/hipe_bif1.c
@@ -32,11 +32,10 @@
 #include "big.h"
 #include "error.h"
 #include "beam_load.h"
+#include "erl_vm.h"
 #include "hipe_bif0.h"
 #include "hipe_bif1.h"
 
-#define BeamOpCode(Op)	((Uint)BeamOp(Op))
-
 BIF_RETTYPE hipe_bifs_call_count_on_1(BIF_ALIST_1)
 {
     ErtsCodeInfo *ci;
@@ -46,17 +45,17 @@ BIF_RETTYPE hipe_bifs_call_count_on_1(BIF_ALIST_1)
     ci = hipe_bifs_find_pc_from_mfa(BIF_ARG_1);
     if (!ci)
 	BIF_ERROR(BIF_P, BADARG);
-    ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
     pc = erts_codeinfo_to_code(ci);
-    if (pc[0] == BeamOpCode(op_hipe_trap_call))
+    if (BeamIsOpCode(pc[0], op_hipe_trap_call))
 	BIF_ERROR(BIF_P, BADARG);
-    if (pc[0] == BeamOpCode(op_hipe_call_count))
+    if (BeamIsOpCode(pc[0], op_hipe_call_count))
 	BIF_RET(NIL);
     hcc = erts_alloc(ERTS_ALC_T_HIPE_SL, sizeof(*hcc));
     hcc->count = 0;
     hcc->opcode = pc[0];
     ci->u.hcc = hcc;
-    pc[0] = BeamOpCode(op_hipe_call_count);
+    pc[0] = BeamOpCodeAddr(op_hipe_call_count);
     BIF_RET(am_true);
 }
 
@@ -70,9 +69,9 @@ BIF_RETTYPE hipe_bifs_call_count_off_1(BIF_ALIST_1)
     ci = hipe_bifs_find_pc_from_mfa(BIF_ARG_1);
     if (!ci)
 	BIF_ERROR(BIF_P, BADARG);
-    ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
     pc = erts_codeinfo_to_code(ci);
-    if (pc[0] != BeamOpCode(op_hipe_call_count))
+    if (! BeamIsOpCode(pc[0], op_hipe_call_count))
 	BIF_RET(am_false);
     hcc = ci->u.hcc;
     count = hcc->count;
@@ -91,9 +90,9 @@ BIF_RETTYPE hipe_bifs_call_count_get_1(BIF_ALIST_1)
     ci = hipe_bifs_find_pc_from_mfa(BIF_ARG_1);
     if (!ci)
 	BIF_ERROR(BIF_P, BADARG);
-    ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
     pc = erts_codeinfo_to_code(ci);
-    if (pc[0] != BeamOpCode(op_hipe_call_count))
+    if (! BeamIsOpCode(pc[0], op_hipe_call_count))
 	BIF_RET(am_false);
     hcc = ci->u.hcc;
     BIF_RET(make_small(hcc->count));
@@ -109,9 +108,9 @@ BIF_RETTYPE hipe_bifs_call_count_clear_1(BIF_ALIST_1)
     ci = hipe_bifs_find_pc_from_mfa(BIF_ARG_1);
     if (!ci)
 	BIF_ERROR(BIF_P, BADARG);
-    ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
+    ASSERT(BeamIsOpCode(ci->op, op_i_func_info_IaaI));
     pc = erts_codeinfo_to_code(ci);
-    if (pc[0] != BeamOpCode(op_hipe_call_count))
+    if (! BeamIsOpCode(pc[0], op_hipe_call_count))
 	BIF_RET(am_false);
     hcc = ci->u.hcc;
     count = hcc->count;
diff --git a/erts/emulator/hipe/hipe_bif2.c b/erts/emulator/hipe/hipe_bif2.c
index e04d3d32d1..df377b2153 100644
--- a/erts/emulator/hipe/hipe_bif2.c
+++ b/erts/emulator/hipe/hipe_bif2.c
@@ -45,7 +45,7 @@ static void proc_unlock(Process* c_p, Process* rp)
 	locks &= ~ERTS_PROC_LOCK_MAIN;
     }
     if (rp && locks) {
-	erts_smp_proc_unlock(rp, locks);
+	erts_proc_unlock(rp, locks);
     }
 }
 
@@ -153,14 +153,14 @@ BIF_RETTYPE hipe_bifs_modeswitch_debug_off_0(BIF_ALIST_0)
     BIF_RET(am_true);
 }
 
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 
 BIF_RETTYPE hipe_debug_bif_wrapper(NBIF_ALIST_1);
 
-#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P) \
+#    define ERTS_REQ_PROC_MAIN_LOCK(P) \
        if ((P)) erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN,\
 					  __FILE__, __LINE__)
-#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P) \
+#    define ERTS_UNREQ_PROC_MAIN_LOCK(P) \
         if ((P)) erts_proc_lc_unrequire_lock((P), ERTS_PROC_LOCK_MAIN)
 
 BIF_RETTYPE hipe_debug_bif_wrapper(NBIF_ALIST_1)
@@ -168,13 +168,13 @@ BIF_RETTYPE hipe_debug_bif_wrapper(NBIF_ALIST_1)
     typedef BIF_RETTYPE nBif(NBIF_ALIST_1);
     nBif* fp = (nBif*) (BIF_P->hipe.bif_callee);
     BIF_RETTYPE res;
-    ERTS_SMP_UNREQ_PROC_MAIN_LOCK(BIF_P);
+    ERTS_UNREQ_PROC_MAIN_LOCK(BIF_P);
     res = (*fp)(NBIF_CALL_ARGS);
-    ERTS_SMP_REQ_PROC_MAIN_LOCK(BIF_P);
+    ERTS_REQ_PROC_MAIN_LOCK(BIF_P);
     return res;
 }
 
-#endif /* ERTS_ENABLE_LOCK_CHECK && ERTS_SMP */
+#endif /* ERTS_ENABLE_LOCK_CHECK*/
 
 
 BIF_RETTYPE hipe_bifs_debug_native_called_2(BIF_ALIST_2)
@@ -190,3 +190,8 @@ BIF_RETTYPE hipe_bifs_llvm_fix_pinned_regs_0(BIF_ALIST_0)
 {
     BIF_RET(am_ok);
 }
+
+BIF_RETTYPE hipe_bifs_build_stacktrace_1(BIF_ALIST_1)
+{
+    BIF_RET(build_stacktrace(BIF_P, BIF_ARG_1));
+}
diff --git a/erts/emulator/hipe/hipe_bif2.tab b/erts/emulator/hipe/hipe_bif2.tab
index bbcb577be0..c4da44606a 100644
--- a/erts/emulator/hipe/hipe_bif2.tab
+++ b/erts/emulator/hipe/hipe_bif2.tab
@@ -32,3 +32,4 @@ bif hipe_bifs:modeswitch_debug_on/0
 bif hipe_bifs:modeswitch_debug_off/0
 bif hipe_bifs:debug_native_called/2
 bif hipe_bifs:llvm_fix_pinned_regs/0
+bif hipe_bifs:build_stacktrace/1
diff --git a/erts/emulator/hipe/hipe_bif_list.m4 b/erts/emulator/hipe/hipe_bif_list.m4
index b2fccdadef..33b3cc1ee5 100644
--- a/erts/emulator/hipe/hipe_bif_list.m4
+++ b/erts/emulator/hipe/hipe_bif_list.m4
@@ -186,6 +186,7 @@ gc_bif_interface_1(nbif_erase_1, erase_1)
 gc_bif_interface_1(nbif_erts_internal_garbage_collect_1, erts_internal_garbage_collect_1)
 gc_nofail_primop_interface_1(nbif_gc_1, hipe_gc)
 gc_bif_interface_2(nbif_put_2, put_2)
+gc_bif_interface_2(nbif_hipe_bifs_build_stacktrace, hipe_bifs_build_stacktrace_1)
 
 /*
  * Debug BIFs that need read access to the full state.
@@ -219,10 +220,11 @@ standard_bif_interface_1(nbif_bnot_1, bnot_1)
 standard_bif_interface_1(nbif_set_timeout, hipe_set_timeout)
 standard_bif_interface_1(nbif_conv_big_to_float, hipe_conv_big_to_float)
 standard_bif_interface_2(nbif_rethrow, hipe_rethrow)
+standard_bif_interface_3(nbif_raw_raise, hipe_raw_raise)
 standard_bif_interface_3(nbif_find_na_or_make_stub, hipe_find_na_or_make_stub)
 standard_bif_interface_2(nbif_nonclosure_address, hipe_nonclosure_address)
 nocons_nofail_primop_interface_0(nbif_fclearerror_error, hipe_fclearerror_error)
-standard_bif_interface_2(nbif_is_divisible, hipe_is_divisible)
+noproc_primop_interface_2(nbif_is_divisible, hipe_is_divisible)
 noproc_primop_interface_1(nbif_is_unicode, hipe_is_unicode)
 
 /*
@@ -248,11 +250,6 @@ nofail_primop_interface_3(nbif_bs_get_float_2, erts_bs_get_float_2)
 nocons_nofail_primop_interface_3(nbif_bs_put_utf8, hipe_bs_put_utf8)
 standard_bif_interface_3(nbif_bs_put_utf16be, hipe_bs_put_utf16be)
 standard_bif_interface_3(nbif_bs_put_utf16le, hipe_bs_put_utf16le)
-ifdef(`nogc_bif_interface_1',`
-nogc_bif_interface_1(nbif_bs_validate_unicode, hipe_bs_validate_unicode)
-',`
-standard_bif_interface_1(nbif_bs_validate_unicode, hipe_bs_validate_unicode)
-')
 
 /*
  * Bit-syntax primops without any P parameter.
@@ -267,18 +264,12 @@ noproc_primop_interface_2(nbif_bs_get_utf16, erts_bs_get_utf16)
 noproc_primop_interface_2(nbif_bs_validate_unicode_retract, hipe_bs_validate_unicode_retract)
 
 /*
- * Bit-syntax primops. The ERTS_SMP runtime system requires P,
+ * Bit-syntax primops. The runtime system requires P,
  * hence the use of nocons_nofail_primop_interface_N().
- * When ERTS_SMP is disabled, noproc_primop_interface_N()
- * should be used instead.
  */
 nocons_nofail_primop_interface_5(nbif_bs_put_small_float, hipe_bs_put_small_float)
 noproc_primop_interface_5(nbif_bs_put_bits, hipe_bs_put_bits)
-ifelse(ERTS_SMP,1,`
 nocons_nofail_primop_interface_5(nbif_bs_put_big_integer, hipe_bs_put_big_integer)
-',`
-noproc_primop_interface_5(nbif_bs_put_big_integer, hipe_bs_put_big_integer)
-')dnl
 
 nofail_primop_interface_0(nbif_check_get_msg, hipe_check_get_msg)
 
@@ -288,13 +279,8 @@ nocons_nofail_primop_interface_0(nbif_emulate_fpe, hipe_emulate_fpe)
 
 noproc_primop_interface_1(nbif_emasculate_binary, hipe_emasculate_binary)
 
-/*
- * SMP-specific stuff
- */
-ifelse(ERTS_SMP,1,`
 nocons_nofail_primop_interface_0(nbif_clear_timeout, hipe_clear_timeout)
 noproc_primop_interface_1(nbif_atomic_inc, hipe_atomic_inc)
-',)dnl
 
 /*
  * BIFs that disable GC while trapping are called via a wrapper
diff --git a/erts/emulator/hipe/hipe_gc.c b/erts/emulator/hipe/hipe_gc.c
index 1a4a4c7952..aaedba1afd 100644
--- a/erts/emulator/hipe/hipe_gc.c
+++ b/erts/emulator/hipe/hipe_gc.c
@@ -91,7 +91,7 @@ Eterm *fullsweep_nstack(Process *p, Eterm *n_htop)
 			ASSERT(is_boxed(val));
 			*nsp_i = val;
 		    } else if (!erts_is_literal(gval, ptr)) {
-			move_boxed(&ptr, val, &n_htop, nsp_i);
+			move_boxed(ptr, val, &n_htop, nsp_i);
 		    }
 		} else if (is_list(gval)) {
 		    Eterm *ptr = list_val(gval);
@@ -100,7 +100,7 @@ Eterm *fullsweep_nstack(Process *p, Eterm *n_htop)
 			*nsp_i = ptr[1];
 		    } else if (!erts_is_literal(gval, ptr)) {
 			ASSERT(erts_dbg_within_proc(ptr, p, NULL));
-			move_cons(&ptr, val, &n_htop, nsp_i);
+			move_cons(ptr, val, &n_htop, nsp_i);
 		    }
 		}
 	    }
@@ -206,10 +206,10 @@ void gensweep_nstack(Process *p, Eterm **ptr_old_htop, Eterm **ptr_n_htop)
 			ASSERT(is_boxed(val));
 			*nsp_i = val;
 		    } else if (ErtsInArea(ptr, mature, mature_size)) {
-			move_boxed(&ptr, val, &old_htop, nsp_i);
+			move_boxed(ptr, val, &old_htop, nsp_i);
 		    } else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
 			ASSERT(erts_dbg_within_proc(ptr, p, NULL));
-			move_boxed(&ptr, val, &n_htop, nsp_i);
+			move_boxed(ptr, val, &n_htop, nsp_i);
 		    }
 		} else if (is_list(gval)) {
 		    Eterm *ptr = list_val(gval);
@@ -217,10 +217,10 @@ void gensweep_nstack(Process *p, Eterm **ptr_old_htop, Eterm **ptr_n_htop)
 		    if (IS_MOVED_CONS(val)) {
 			*nsp_i = ptr[1];
 		    } else if (ErtsInArea(ptr, mature, mature_size)) {
-			move_cons(&ptr, val, &old_htop, nsp_i);
+			move_cons(ptr, val, &old_htop, nsp_i);
 		    } else if (ErtsInYoungGen(gval, ptr, oh, oh_size)) {
 			ASSERT(erts_dbg_within_proc(ptr, p, NULL));
-			move_cons(&ptr, val, &n_htop, nsp_i);
+			move_cons(ptr, val, &n_htop, nsp_i);
 		    }
 		}
 	    }
@@ -278,7 +278,7 @@ Eterm *sweep_literals_nstack(Process *p, Eterm *old_htop, char *area,
 			ASSERT(is_boxed(val));
 			*nsp_i = val;
 		    } else if (ErtsInArea(ptr, area, area_size)) {
-			move_boxed(&ptr, val, &old_htop, nsp_i);
+			move_boxed(ptr, val, &old_htop, nsp_i);
 		    }
 		} else if (is_list(gval)) {
 		    Eterm *ptr = list_val(gval);
@@ -286,7 +286,7 @@ Eterm *sweep_literals_nstack(Process *p, Eterm *old_htop, char *area,
 		    if (IS_MOVED_CONS(val)) {
 			*nsp_i = ptr[1];
 		    } else if (ErtsInArea(ptr, area, area_size)) {
-			move_cons(&ptr, val, &old_htop, nsp_i);
+			move_cons(ptr, val, &old_htop, nsp_i);
 		    }
 		}
 	    }
diff --git a/erts/emulator/hipe/hipe_instrs.tab b/erts/emulator/hipe/hipe_instrs.tab
new file mode 100644
index 0000000000..a01baebddf
--- /dev/null
+++ b/erts/emulator/hipe/hipe_instrs.tab
@@ -0,0 +1,141 @@
+// -*- c -*-
+//
+// %CopyrightBegin%
+//
+// Copyright Ericsson AB 2017. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+//
+// %CopyrightEnd%
+//
+
+
+HIPE_MODE_SWITCH(Cmd) {
+    SWAPOUT;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    c_p->fcalls = FCALLS;
+    c_p->def_arg_reg[4] = -neg_o_reds;
+    c_p = hipe_mode_switch(c_p, $Cmd, reg);
+}
+
+hipe_trap_call := hipe_trap.call.post;
+hipe_trap_call_closure := hipe_trap.call_closure.post;
+hipe_trap_return := hipe_trap.return.post;
+hipe_trap_throw := hipe_trap.throw.post;
+hipe_trap_resume := hipe_trap.resume.post;
+
+hipe_trap.call() {
+    /*
+     * I[-5]: &&lb_i_func_info_IaaI
+     * I[-4]: Native code callee (inserted by HiPE)
+     * I[-3]: Module (tagged atom)
+     * I[-2]: Function (tagged atom)
+     * I[-1]: Arity (untagged integer)
+     * I[ 0]: &&lb_hipe_trap_call
+     * ... remainder of original BEAM code
+     */
+    ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
+    ASSERT(IsOpCode(ci->op, i_func_info_IaaI));
+    c_p->hipe.u.ncallee = ci->u.ncallee;
+    ++hipe_trap_count;
+    $HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL | (ci->mfa.arity << 8));
+}
+
+hipe_trap.call_closure() {
+    ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
+    ASSERT(IsOpCode(ci->op, i_func_info_IaaI));
+    c_p->hipe.u.ncallee = ci->u.ncallee;
+    ++hipe_trap_count;
+    $HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_CALL_CLOSURE | (ci->mfa.arity << 8));
+}
+
+hipe_trap.return() {
+    $HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RETURN);
+}
+
+hipe_trap.throw() {
+    $HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_THROW);
+}
+
+hipe_trap.resume() {
+    $HIPE_MODE_SWITCH(HIPE_MODE_SWITCH_CMD_RESUME);
+}
+
+hipe_trap.post() {
+#ifdef DEBUG
+    pid = c_p->common.id; /* may have switched process... */
+#endif
+    reg = erts_proc_sched_data(c_p)->x_reg_array;
+    freg = erts_proc_sched_data(c_p)->f_reg_array;
+    ERL_BITS_RELOAD_STATEP(c_p);
+    /* XXX: this abuse of def_arg_reg[] is horrid! */
+    neg_o_reds = -c_p->def_arg_reg[4];
+    FCALLS = c_p->fcalls;
+    SWAPIN;
+    ERTS_DBG_CHK_REDS(c_p, FCALLS);
+    switch( c_p->def_arg_reg[3] ) {
+    case HIPE_MODE_SWITCH_RES_RETURN:
+        ASSERT(is_value(reg[0]));
+        SET_I(c_p->cp);
+        c_p->cp = 0;
+        Goto(*I);
+    case HIPE_MODE_SWITCH_RES_CALL_EXPORTED:
+        c_p->i = c_p->hipe.u.callee_exp->addressv[erts_active_code_ix()];
+        /*fall through*/
+    case HIPE_MODE_SWITCH_RES_CALL_BEAM:
+        SET_I(c_p->i);
+        Dispatch();
+    case HIPE_MODE_SWITCH_RES_CALL_CLOSURE:
+        /* This can be used to call any function value, but currently
+           it's only used to call closures referring to unloaded
+           modules. */
+        {
+            BeamInstr *next;
+
+            next = call_fun(c_p, c_p->arity - 1, reg, THE_NON_VALUE);
+            HEAVY_SWAPIN;
+            if (next != NULL) {
+                SET_I(next);
+                Dispatchfun();
+            }
+            goto find_func_info;
+        }
+    case HIPE_MODE_SWITCH_RES_THROW:
+        c_p->cp = NULL;
+        I = handle_error(c_p, I, reg, NULL);
+        goto post_error_handling;
+    default:
+        erts_exit(ERTS_ERROR_EXIT, "hipe_mode_switch: result %u\n", c_p->def_arg_reg[3]);
+    }
+    //| -no_next;
+}
+
+hipe_call_count() {
+    /*
+     * I[-5]: &&lb_i_func_info_IaaI
+     * I[-4]: pointer to struct hipe_call_count (inserted by HiPE)
+     * I[-3]: Module (tagged atom)
+     * I[-2]: Function (tagged atom)
+     * I[-1]: Arity (untagged integer)
+     * I[ 0]: &&lb_hipe_call_count
+     * ... remainder of original BEAM code
+     */
+    ErtsCodeInfo *ci = erts_code_to_codeinfo(I);
+    struct hipe_call_count *hcc = ci->u.hcc;
+    ASSERT(IsOpCode(ci->op, i_func_info_IaaI));
+    ASSERT(hcc != NULL);
+    ASSERT(VALID_INSTR(hcc->opcode));
+    ++(hcc->count);
+    Goto(hcc->opcode);
+    //| -no_next;
+}
diff --git a/erts/emulator/hipe/hipe_mkliterals.c b/erts/emulator/hipe/hipe_mkliterals.c
index 1ebe4e1188..74e793577c 100644
--- a/erts/emulator/hipe/hipe_mkliterals.c
+++ b/erts/emulator/hipe/hipe_mkliterals.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2001-2016. All Rights Reserved.
+ * Copyright Ericsson AB 2001-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -441,9 +441,7 @@ static const struct rts_param rts_params[] = {
     { 11, "ERL_FUN_SIZE", 1, ERL_FUN_SIZE },
 
     { 12, "P_SCHED_DATA",
-#ifdef ERTS_SMP
       1, offsetof(struct process, scheduler_data)
-#endif
     },
     { 14, "P_FP_EXCEPTION",
 #if !defined(NO_FPE_SIGNALS) || defined(HIPE)
@@ -453,22 +451,26 @@ static const struct rts_param rts_params[] = {
     /* This flag is always defined, but its value is configuration-dependent. */
     { 15, "ERTS_IS_SMP",
       1,
-#if defined(ERTS_SMP)
+      1
+    },
+    /* This flag is always defined, but its value is configuration-dependent. */
+    { 16, "ERTS_NO_FPE_SIGNALS",
+      1,
+#if defined(NO_FPE_SIGNALS)
       1
 #else
       0
 #endif
     },
     /* This flag is always defined, but its value is configuration-dependent. */
-    { 16, "ERTS_NO_FPE_SIGNALS",
+    { 17, "ERTS_USE_LITERAL_TAG",
       1,
-#if defined(NO_FPE_SIGNALS)
+#if defined(TAG_LITERAL_PTR)
       1
 #else
       0
 #endif
     },
-    /* This parameter is always defined, but its value depends on ERTS_SMP. */
     { 19, "MSG_MESSAGE",
       1, offsetof(struct erl_mesg, m[0])
     },
@@ -513,12 +515,12 @@ static const struct rts_param rts_params[] = {
 #endif
     },
     { 48, "P_BIF_CALLEE",
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
 	1, offsetof(struct process, hipe.bif_callee)
 #endif
     },
-    { 49, "P_MSG_FIRST", 1, offsetof(struct process, msg.first) },
-    { 50, "P_MSG_SAVE", 1, offsetof(struct process, msg.save) },
+    { 49, "P_MSG_FIRST", 1, offsetof(struct process, sig_qs.first) },
+    { 50, "P_MSG_SAVE", 1, offsetof(struct process, sig_qs.save) },
     { 51, "P_CALLEE_EXP", 1, offsetof(struct process, hipe.u.callee_exp) },
 
     { 52, "THE_NON_VALUE", 1, (int)THE_NON_VALUE },
@@ -528,6 +530,9 @@ static const struct rts_param rts_params[] = {
       1, offsetof(struct process, hipe.gc_is_unsafe)
 #endif
     },
+
+    { 54, "P_MSG_LAST", 1, offsetof(struct process, sig_qs.last) },
+    { 55, "P_MSG_SAVED_LAST", 1, offsetof(struct process, sig_qs.saved_last) },
 };
 
 #define NR_PARAMS	ARRAY_SIZE(rts_params)
diff --git a/erts/emulator/hipe/hipe_mode_switch.c b/erts/emulator/hipe/hipe_mode_switch.c
index ba7ae1e6a8..0a65e317ed 100644
--- a/erts/emulator/hipe/hipe_mode_switch.c
+++ b/erts/emulator/hipe/hipe_mode_switch.c
@@ -36,15 +36,15 @@
 #include "hipe_stack.h"
 #include "hipe_bif0.h"	/* hipe_mfa_info_table_init() */
 
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
-#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P) \
+#if defined(ERTS_ENABLE_LOCK_CHECK)
+#    define ERTS_REQ_PROC_MAIN_LOCK(P) \
         if ((P)) erts_proc_lc_require_lock((P), ERTS_PROC_LOCK_MAIN,	\
 					   __FILE__, __LINE__)
-#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P) \
+#    define ERTS_UNREQ_PROC_MAIN_LOCK(P) \
         if ((P)) erts_proc_lc_unrequire_lock((P), ERTS_PROC_LOCK_MAIN)
 #else
-#    define ERTS_SMP_REQ_PROC_MAIN_LOCK(P)
-#    define ERTS_SMP_UNREQ_PROC_MAIN_LOCK(P)
+#    define ERTS_REQ_PROC_MAIN_LOCK(P)
+#    define ERTS_UNREQ_PROC_MAIN_LOCK(P)
 #endif
 
 
@@ -155,8 +155,6 @@ void hipe_check_pcb(Process *p, const char *file, unsigned line)
 #include "hipe_arm_glue.h"
 #endif
 
-#define BeamOpCode(Op)		((Uint)BeamOp(Op))
-
 Uint hipe_beam_pc_return[1];	/* needed in hipe_debug.c */
 Uint hipe_beam_pc_throw[1];	/* needed in hipe_debug.c */
 Uint hipe_beam_pc_resume[1];	/* needed by hipe_set_timeout() */
@@ -166,9 +164,9 @@ void hipe_mode_switch_init(void)
 {
     hipe_arch_glue_init();
 
-    hipe_beam_pc_return[0] = BeamOpCode(op_hipe_trap_return);
-    hipe_beam_pc_throw[0] = BeamOpCode(op_hipe_trap_throw);
-    hipe_beam_pc_resume[0] = BeamOpCode(op_hipe_trap_resume);
+    hipe_beam_pc_return[0] = BeamOpCodeAddr(op_hipe_trap_return);
+    hipe_beam_pc_throw[0] = BeamOpCodeAddr(op_hipe_trap_throw);
+    hipe_beam_pc_resume[0] = BeamOpCodeAddr(op_hipe_trap_resume);
 
     hipe_beam_catch_throw =
 	make_catch(beam_catches_cons(hipe_beam_pc_throw, BEAM_CATCHES_NIL));
@@ -182,8 +180,8 @@ void hipe_set_call_trap(ErtsCodeInfo* ci, void *nfun, int is_closure)
     HIPE_ASSERT(ci->op == BeamOpCode(op_i_func_info_IaaI));
     bfun[0] =
 	is_closure
-	? BeamOpCode(op_hipe_trap_call_closure)
-	: BeamOpCode(op_hipe_trap_call);
+	? BeamOpCodeAddr(op_hipe_trap_call_closure)
+	: BeamOpCodeAddr(op_hipe_trap_call);
     ci->u.ncallee = (void (*)(void)) nfun;
 }
 
@@ -394,7 +392,7 @@ Process *hipe_mode_switch(Process *p, unsigned cmd, Eterm reg[])
 	      goto do_schedule;
 	  }
 
-	  if (!(erts_smp_atomic32_read_acqb(&p->state) & ERTS_PSFLG_ACTIVE)) {
+	  if (!(erts_atomic32_read_acqb(&p->state) & ERTS_PSFLG_ACTIVE)) {
 	      for (i = 0; i < p->arity; ++i)
 		  p->arg_reg[i] = reg[i]; 	      
 	      goto do_schedule;
@@ -490,19 +488,24 @@ Process *hipe_mode_switch(Process *p, unsigned cmd, Eterm reg[])
       case HIPE_MODE_SWITCH_RES_WAIT:
       case HIPE_MODE_SWITCH_RES_WAIT_TIMEOUT: {
 	  /* same semantics, different debug trace messages */
-#ifdef ERTS_SMP
 	  /* XXX: BEAM has different entries for the locked and unlocked
 	     cases. HiPE doesn't, so we must check dynamically. */
-	  if (p->hipe_smp.have_receive_locks)
-	      p->hipe_smp.have_receive_locks = 0;
+	  if (p->flags & F_HIPE_RECV_LOCKED)
+	      p->flags &= ~F_HIPE_RECV_LOCKED;
 	  else
-	      erts_smp_proc_lock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-#endif
+	      erts_proc_lock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
 	  p->i = hipe_beam_pc_resume;
 	  p->arity = 0;
-	  erts_smp_atomic32_read_band_relb(&p->state,
-					   ~ERTS_PSFLG_ACTIVE);
-	  erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+          if (erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_EXITING)
+              ASSERT(erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_ACTIVE);
+          else if (!(p->flags & F_HIPE_RECV_YIELD))
+              erts_atomic32_read_band_relb(&p->state, ~ERTS_PSFLG_ACTIVE);
+          else {
+              /* Yielded from receive */
+              ERTS_VBUMP_ALL_REDS(p);
+              p->flags &= ~F_HIPE_RECV_YIELD;
+          }
+	  erts_proc_unlock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
       do_schedule:
 	  {
 	      struct saved_calls *scb;
@@ -513,21 +516,19 @@ Process *hipe_mode_switch(Process *p, unsigned cmd, Eterm reg[])
 
               /* The process may have died while it was executing,
                  if so we return out from native code to the interpreter */
-              if (erts_smp_atomic32_read_nob(&p->state) & ERTS_PSFLG_EXITING)
+              if (erts_atomic32_read_nob(&p->state) & ERTS_PSFLG_EXITING)
                   p->i = beam_exit;
 #ifdef DEBUG
 	      ASSERT(p->debug_reds_in == reds_in);
 #endif
 	      p->flags &= ~F_HIPE_MODE;
 
-	      ERTS_SMP_UNREQ_PROC_MAIN_LOCK(p);
+	      ERTS_UNREQ_PROC_MAIN_LOCK(p);
 	      p = erts_schedule(NULL, p, reds_in - p->fcalls);
-	      ERTS_SMP_REQ_PROC_MAIN_LOCK(p);
+	      ERTS_REQ_PROC_MAIN_LOCK(p);
 	      ASSERT(!(p->flags & F_HIPE_MODE));
-#ifdef ERTS_SMP
-	      p->hipe_smp.have_receive_locks = 0;
+	      p->flags &= ~F_HIPE_RECV_LOCKED;
 	      reg = p->scheduler_data->x_reg_array;
-#endif
 	  }
 	  {
 	      Eterm *argp;
@@ -651,10 +652,10 @@ void hipe_inc_nstack(Process *p)
 	p->hipe.nsp = new_nstack + (p->hipe.nsp - old_nstack);
 	p->hipe.nstack = new_nstack;
 	if (p->hipe.nstgraylim)
-	    p->hipe.nstgraylim = 
+	    p->hipe.nstgraylim =
 		new_nstack + (p->hipe.nstgraylim - old_nstack);
 	if (p->hipe.nstblacklim)
-	    p->hipe.nstblacklim = 
+	    p->hipe.nstblacklim =
 		new_nstack + (p->hipe.nstblacklim - old_nstack);
     }
 }
@@ -668,7 +669,8 @@ void hipe_inc_nstack(Process *p)
     Eterm *new_nstack = erts_alloc(ERTS_ALC_T_HIPE_STK, new_size*sizeof(Eterm));
     unsigned used_size = p->hipe.nstend - p->hipe.nsp;
 
-    sys_memcpy(new_nstack+new_size-used_size, p->hipe.nsp, used_size*sizeof(Eterm));
+    if (used_size)
+        sys_memcpy(new_nstack+new_size-used_size, p->hipe.nsp, used_size*sizeof(Eterm));
     if (p->hipe.nstgraylim)
 	p->hipe.nstgraylim = new_nstack + new_size - (p->hipe.nstend - p->hipe.nstgraylim);
     if (p->hipe.nstblacklim)
diff --git a/erts/emulator/hipe/hipe_native_bif.c b/erts/emulator/hipe/hipe_native_bif.c
index 6ab7a9e1de..211ce0492a 100644
--- a/erts/emulator/hipe/hipe_native_bif.c
+++ b/erts/emulator/hipe/hipe_native_bif.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2001-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2001-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -35,6 +35,7 @@
 #include "hipe_native_bif.h"
 #include "hipe_arch.h"
 #include "hipe_stack.h"
+#include "erl_proc_sig_queue.h"
 
 /*
  * These are wrappers for BIFs that may trigger a native
@@ -143,12 +144,10 @@ BIF_RETTYPE nbif_impl_hipe_set_timeout(NBIF_ALIST_1)
     else {
 	int tres = erts_set_proc_timer_term(p, timeout_value);
 	if (tres != 0) { /* Wrong time */
-#ifdef ERTS_SMP
-	    if (p->hipe_smp.have_receive_locks) {
-		p->hipe_smp.have_receive_locks = 0;
-		erts_smp_proc_unlock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+	    if (p->flags & F_HIPE_RECV_LOCKED) {
+                p->flags &= ~F_HIPE_RECV_LOCKED;
+		erts_proc_unlock(p, ERTS_PROC_LOCKS_MSG_RECEIVE);
 	    }
-#endif
 	    BIF_ERROR(p, EXC_TIMEOUT_VALUE);
 	}
     }
@@ -256,6 +255,8 @@ void hipe_handle_exception(Process *c_p)
     /* Synthesized to avoid having to generate code for it. */
     c_p->def_arg_reg[0] = exception_tag[GET_EXC_CLASS(c_p->freason)];
 
+    ERTS_RECV_MARK_CLEAR(c_p);  /* No longer safe to use this position */
+
     hipe_find_handler(c_p);
 }
 
@@ -314,6 +315,32 @@ BIF_RETTYPE nbif_impl_hipe_rethrow(NBIF_ALIST_2)
     }
 }
 
+/* Called via standard_bif_interface_3 */
+BIF_RETTYPE nbif_impl_hipe_raw_raise(NBIF_ALIST_3)
+{
+    Process *c_p = BIF_P;
+    Eterm class = BIF_ARG_1;
+    Eterm value = BIF_ARG_2;
+    Eterm stacktrace = BIF_ARG_3;
+    Eterm reason;
+
+    if (class == am_error) {
+	c_p->fvalue = value;
+	reason = EXC_ERROR;
+    } else if (class == am_exit) {
+	c_p->fvalue = value;
+	reason = EXC_EXIT;
+    } else if (class == am_throw) {
+	c_p->fvalue = value;
+	reason = EXC_THROWN;
+    } else {
+        return am_badarg;
+    }
+    reason &= ~EXF_SAVETRACE;
+    c_p->ftrace = stacktrace;
+    BIF_ERROR(c_p, reason);
+}
+
 /*
  * Support for compiled binary syntax operations.
  */
@@ -335,9 +362,7 @@ Binary *hipe_bs_reallocate(Binary* oldbptr, int newsize)
 }
 
 int hipe_bs_put_big_integer(
-#ifdef ERTS_SMP
     Process *p,
-#endif
     Eterm arg, Uint num_bits, byte* base, unsigned offset, unsigned flags)
 {
     byte *save_bin_buf;
@@ -483,15 +508,6 @@ static int validate_unicode(Eterm arg)
     return 1;
 }
 
-BIF_RETTYPE nbif_impl_hipe_bs_validate_unicode(NBIF_ALIST_1)
-{
-    Process *p = BIF_P;
-    Eterm arg = BIF_ARG_1;
-    if (!validate_unicode(arg))
-	BIF_ERROR(p, BADARG);
-    return NIL;
-}
-
 Uint hipe_is_unicode(Eterm arg)
 {
     return (Uint) validate_unicode(arg);
@@ -507,16 +523,12 @@ int hipe_bs_validate_unicode_retract(ErlBinMatchBuffer* mb, Eterm arg)
     return 1;
 }
 
-/* Called via standard_bif_interface_2 */
-BIF_RETTYPE nbif_impl_hipe_is_divisible(NBIF_ALIST_2)
+Uint hipe_is_divisible(Uint dividend, Uint divisor)
 {
-    /* Arguments are Eterm-sized unsigned integers */
-    Uint dividend = BIF_ARG_1;
-    Uint divisor = BIF_ARG_2;
     if (dividend % divisor) {
-	BIF_ERROR(BIF_P, BADARG);
+        return 0;
     } else {
-	return NIL;
+        return 1;
     }
 }
 
@@ -533,42 +545,55 @@ Eterm hipe_check_get_msg(Process *c_p)
     msgp = PEEK_MESSAGE(c_p);
 
     if (!msgp) {
-#ifdef ERTS_SMP
-	erts_smp_proc_lock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	/* Make sure messages wont pass exit signals... */
-	if (ERTS_PROC_PENDING_EXIT(c_p)) {
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	    return THE_NON_VALUE; /* Will be rescheduled for exit */
-	}
-	ERTS_SMP_MSGQ_MV_INQ2PRIVQ(c_p);
-	msgp = PEEK_MESSAGE(c_p);
-	if (msgp)
-	    erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
-	else {
-	    /* XXX: BEAM doesn't need this */
-	    c_p->hipe_smp.have_receive_locks = 1;
-#endif
-	    c_p->flags &= ~F_DELAY_GC;
-	    return THE_NON_VALUE;
-#ifdef ERTS_SMP
-	}
-#endif
+        int get_out;
+        c_p->i = NULL;
+        c_p->arity = 0;
+        c_p->current = NULL;
+        (void) erts_proc_sig_receive_helper(c_p, CONTEXT_REDS/4, 0,
+                                            &msgp, &get_out);
+        /* FIXME: Need to bump reductions... */
+        if (!msgp) {
+            if (get_out) {
+                if (get_out < 0)
+                    c_p->flags |= F_HIPE_RECV_YIELD; /* yield... */
+                /* else: go exit... */
+                return THE_NON_VALUE;
+            }
+
+            /*
+             * If there are no more messages in queue
+             * (and we are not yielding or exiting)
+             * erts_proc_sig_receive_helper()
+             * returns with message queue lock locked...
+             */
+
+            /* XXX: BEAM doesn't need this */
+            c_p->flags |= F_HIPE_RECV_LOCKED;
+            c_p->flags &= ~F_DELAY_GC;
+            return THE_NON_VALUE;
+        }
     }
 
-    if (is_non_value(ERL_MESSAGE_TERM(msgp))
-	&& !erts_decode_dist_message(c_p, ERTS_PROC_LOCK_MAIN, msgp, 0)) {
-	/*
-	 * A corrupt distribution message that we weren't able to decode;
-	 * remove it...
-	 */
-	ASSERT(!msgp->data.attached);
-	UNLINK_MESSAGE(c_p, msgp);
-	msgp->next = NULL;
-	erts_cleanup_messages(msgp);
-	goto next_message;
+    ASSERT(msgp == PEEK_MESSAGE(c_p));
+    ASSERT(msgp && ERTS_SIG_IS_MSG(msgp));
+
+    if (ERTS_SIG_IS_EXTERNAL_MSG(msgp)) {
+        /* FIXME: bump appropriate amount... */
+        if (!erts_decode_dist_message(c_p, ERTS_PROC_LOCK_MAIN, msgp, 0)) {
+            /*
+             * A corrupt distribution message that we weren't able to decode;
+             * remove it...
+             */
+            /* TODO: Add DTrace probe for this bad message situation? */
+            UNLINK_MESSAGE(c_p, msgp);
+            msgp->next = NULL;
+            erts_cleanup_messages(msgp);
+            goto next_message;
+        }
     }
 
-    ASSERT(is_value(ERL_MESSAGE_TERM(msgp)));
+    ASSERT(msgp == PEEK_MESSAGE(c_p));
+    ASSERT(ERTS_SIG_IS_INTERNAL_MSG(msgp));
 
     return ERL_MESSAGE_TERM(msgp);
 }
@@ -576,7 +601,6 @@ Eterm hipe_check_get_msg(Process *c_p)
 /*
  * SMP-specific stuff
  */
-#ifdef ERTS_SMP
 
 /*
  * This is like the timeout BEAM instruction.
@@ -587,14 +611,12 @@ void hipe_clear_timeout(Process *c_p)
      * A timeout has occurred.  Reset the save pointer so that the next
      * receive statement will examine the first message first.
      */
-#ifdef ERTS_SMP
     /* XXX: BEAM has different entries for the locked and unlocked
        cases. HiPE doesn't, so we must check dynamically. */
-    if (c_p->hipe_smp.have_receive_locks) {
-	c_p->hipe_smp.have_receive_locks = 0;
-	erts_smp_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
+    if (c_p->flags & F_HIPE_RECV_LOCKED) {
+	c_p->flags &= ~F_HIPE_RECV_LOCKED;
+	erts_proc_unlock(c_p, ERTS_PROC_LOCKS_MSG_RECEIVE);
     }
-#endif
     if (IS_TRACED_FL(c_p, F_TRACE_RECEIVE)) {
 	trace_receive(c_p, am_clock_service, am_timeout, NULL);
     }
@@ -604,7 +626,6 @@ void hipe_clear_timeout(Process *c_p)
 
 void hipe_atomic_inc(int *counter)
 {
-    erts_smp_atomic_inc_nob((erts_smp_atomic_t*)counter);
+    erts_atomic_inc_nob((erts_atomic_t*)counter);
 }
 
-#endif
diff --git a/erts/emulator/hipe/hipe_native_bif.h b/erts/emulator/hipe/hipe_native_bif.h
index 6321e66e7a..ba42b126be 100644
--- a/erts/emulator/hipe/hipe_native_bif.h
+++ b/erts/emulator/hipe/hipe_native_bif.h
@@ -36,6 +36,7 @@ AEXTERN(int,nbif_suspend_msg,(void));
 AEXTERN(int,nbif_suspend_msg_timeout,(void));
 
 AEXTERN(Eterm,nbif_rethrow,(Process*, Eterm, Eterm));
+AEXTERN(Eterm,nbif_raw_raise,(Process*, Eterm, Eterm, Eterm));
 AEXTERN(Eterm,nbif_set_timeout,(Process*, Eterm));
 
 AEXTERN(Eterm,nbif_gc_1,(void));
@@ -66,10 +67,9 @@ AEXTERN(Eterm,nbif_bs_utf16_size,(Eterm));
 AEXTERN(Eterm,nbif_bs_put_utf16be,(Process*,Eterm,byte*,unsigned int));
 AEXTERN(Eterm,nbif_bs_put_utf16le,(Process*,Eterm,byte*,unsigned int));
 AEXTERN(Eterm,nbif_bs_get_utf16,(void));
-AEXTERN(Eterm,nbif_bs_validate_unicode,(Process*,Eterm));
 AEXTERN(Uint,nbif_is_unicode,(Eterm));
 AEXTERN(Eterm,nbif_bs_validate_unicode_retract,(void));
-AEXTERN(void,nbif_is_divisible,(Process*,Uint,Uint));
+AEXTERN(Uint,nbif_is_divisible,(Uint,Uint));
 
 AEXTERN(void,nbif_select_msg,(Process*));
 AEXTERN(Eterm,nbif_cmp_2,(void));
@@ -83,6 +83,7 @@ void hipe_gc(Process*, Eterm);
 BIF_RETTYPE nbif_impl_hipe_set_timeout(NBIF_ALIST_1);
 void hipe_handle_exception(Process*);
 BIF_RETTYPE nbif_impl_hipe_rethrow(NBIF_ALIST_2);
+BIF_RETTYPE nbif_impl_hipe_raw_raise(NBIF_ALIST_3);
 char *hipe_bs_allocate(int);
 Binary *hipe_bs_reallocate(Binary*, int);
 int hipe_bs_put_small_float(Process*, Eterm, Uint, byte*, unsigned, unsigned);
@@ -92,11 +93,10 @@ Eterm hipe_bs_put_utf8(Process*, Eterm arg, byte* base, Uint offset);
 Eterm hipe_bs_utf16_size(Eterm);
 BIF_RETTYPE nbif_impl_hipe_bs_put_utf16be(NBIF_ALIST_3);
 BIF_RETTYPE nbif_impl_hipe_bs_put_utf16le(NBIF_ALIST_3);
-BIF_RETTYPE nbif_impl_hipe_bs_validate_unicode(NBIF_ALIST_1);
 Uint hipe_is_unicode(Eterm);
 struct erl_bin_match_buffer;
 int hipe_bs_validate_unicode_retract(struct erl_bin_match_buffer*, Eterm);
-BIF_RETTYPE nbif_impl_hipe_is_divisible(NBIF_ALIST_2);
+Uint hipe_is_divisible(Uint, Uint);
 
 #ifdef NO_FPE_SIGNALS
 AEXTERN(void,nbif_emulate_fpe,(Process*));
@@ -106,14 +106,13 @@ void hipe_emulate_fpe(Process*);
 AEXTERN(void,nbif_emasculate_binary,(Eterm));
 void hipe_emasculate_binary(Eterm);
 
+AEXTERN(BIF_RETTYPE,nbif_hipe_bifs_build_stacktrace,(Process*,Eterm));
+BIF_RETTYPE hipe_bifs_build_stacktrace_1(BIF_ALIST_1);
+
 /*
  * Stuff that is different in SMP and non-SMP.
  */
-#ifdef ERTS_SMP
 int hipe_bs_put_big_integer(Process*, Eterm, Uint, byte*, unsigned, unsigned);
-#else
-int hipe_bs_put_big_integer(Eterm, Uint, byte*, unsigned, unsigned);
-#endif
 
 AEXTERN(Eterm,nbif_check_get_msg,(Process*));
 Eterm hipe_check_get_msg(Process*);
@@ -124,12 +123,10 @@ BIF_RETTYPE hipe_bifs_debug_native_called_2(BIF_ALIST_2);
 /*
  * SMP-specific stuff
  */
-#ifdef ERTS_SMP
 AEXTERN(void,nbif_atomic_inc,(void));
 AEXTERN(void,nbif_clear_timeout,(Process*));
 void hipe_atomic_inc(int*);
 void hipe_clear_timeout(Process*);
-#endif
 
 #define BIF_LIST(M,F,A,B,C,I)	AEXTERN(Eterm,nbif_##C,(void));
 #include "erl_bif_list.h"
diff --git a/erts/emulator/hipe/hipe_ops.tab b/erts/emulator/hipe/hipe_ops.tab
index 96e4c0da91..19a3820a6a 100644
--- a/erts/emulator/hipe/hipe_ops.tab
+++ b/erts/emulator/hipe/hipe_ops.tab
@@ -23,4 +23,7 @@ hipe_trap_call_closure
 hipe_trap_return
 hipe_trap_throw
 hipe_trap_resume
+
+%cold
 hipe_call_count
+%hot
diff --git a/erts/emulator/hipe/hipe_ppc_bifs.m4 b/erts/emulator/hipe/hipe_ppc_bifs.m4
index 79a8bef77d..283fbbb200 100644
--- a/erts/emulator/hipe/hipe_ppc_bifs.m4
+++ b/erts/emulator/hipe/hipe_ppc_bifs.m4
@@ -25,7 +25,7 @@ include(`hipe/hipe_ppc_asm.m4')
 #`include' "config.h"
 #`include' "hipe_literals.h"
 
-`#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+`#if defined(ERTS_ENABLE_LOCK_CHECK)
 #  define CALL_BIF(F)	STORE_IA(CSYM(nbif_impl_##F), P_BIF_CALLEE(P), r29); bl CSYM(hipe_debug_bif_wrapper) 
 #else
 #  define CALL_BIF(F)	bl	CSYM(nbif_impl_##F)
diff --git a/erts/emulator/hipe/hipe_primops.h b/erts/emulator/hipe/hipe_primops.h
index 77f0dfe7e5..c5f10672f3 100644
--- a/erts/emulator/hipe/hipe_primops.h
+++ b/erts/emulator/hipe/hipe_primops.h
@@ -41,13 +41,12 @@ PRIMOP_LIST(am_bnot, &nbif_bnot_1)
 
 PRIMOP_LIST(am_gc_1, &nbif_gc_1)
 PRIMOP_LIST(am_check_get_msg, &nbif_check_get_msg)
-#ifdef ERTS_SMP
 PRIMOP_LIST(am_atomic_inc, &nbif_atomic_inc)
 PRIMOP_LIST(am_clear_timeout, &nbif_clear_timeout)
-#endif
 PRIMOP_LIST(am_select_msg, &nbif_select_msg)
 PRIMOP_LIST(am_set_timeout, &nbif_set_timeout)
 PRIMOP_LIST(am_rethrow, &nbif_rethrow)
+PRIMOP_LIST(am_raw_raise, &nbif_raw_raise)
 
 
 PRIMOP_LIST(am_bs_get_integer_2, &nbif_bs_get_integer_2)
@@ -65,7 +64,6 @@ PRIMOP_LIST(am_bs_utf16_size, &nbif_bs_utf16_size)
 PRIMOP_LIST(am_bs_put_utf16be, &nbif_bs_put_utf16be)
 PRIMOP_LIST(am_bs_put_utf16le, &nbif_bs_put_utf16le)
 PRIMOP_LIST(am_bs_get_utf16, &nbif_bs_get_utf16)
-PRIMOP_LIST(am_bs_validate_unicode, &nbif_bs_validate_unicode)
 PRIMOP_LIST(am_is_unicode, &nbif_is_unicode)
 PRIMOP_LIST(am_bs_validate_unicode_retract, &nbif_bs_validate_unicode_retract)
 
@@ -86,6 +84,7 @@ PRIMOP_LIST(am_emulate_fpe, &nbif_emulate_fpe)
 #endif
 PRIMOP_LIST(am_emasculate_binary, &nbif_emasculate_binary)
 PRIMOP_LIST(am_debug_native_called, &nbif_hipe_bifs_debug_native_called)
+PRIMOP_LIST(am_build_stacktrace, &nbif_hipe_bifs_build_stacktrace)
 
 #if defined(__sparc__)
 #include "hipe_sparc_primops.h"
diff --git a/erts/emulator/hipe/hipe_process.h b/erts/emulator/hipe/hipe_process.h
index cc92bf653c..18354ba0a6 100644
--- a/erts/emulator/hipe/hipe_process.h
+++ b/erts/emulator/hipe/hipe_process.h
@@ -49,7 +49,7 @@ struct hipe_process_state {
 #ifdef NO_FPE_SIGNALS
     double float_result;        /* to be checked for inf/NaN by hipe_emulate_fpe */ 
 #endif
-#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     void (*bif_callee)(void);   /* When calling BIF's via debug wrapper */
 #endif
 #ifdef DEBUG
@@ -82,15 +82,4 @@ static __inline__ void hipe_delete_process(struct hipe_process_state *p)
 	erts_free(ERTS_ALC_T_HIPE_STK, (void*)p->nstack);
 }
 
-#ifdef ERTS_SMP
-struct hipe_process_state_smp {
-    int have_receive_locks;
-};
-
-static __inline__ void hipe_init_process_smp(struct hipe_process_state_smp *p)
-{
-    p->have_receive_locks = 0;
-}
-#endif
-
 #endif /* HIPE_PROCESS_H */
diff --git a/erts/emulator/hipe/hipe_signal.h b/erts/emulator/hipe/hipe_signal.h
index 5d8621135b..524def11a4 100644
--- a/erts/emulator/hipe/hipe_signal.h
+++ b/erts/emulator/hipe/hipe_signal.h
@@ -27,13 +27,9 @@
 
 #if defined(__i386__) || defined(__x86_64__)
 extern void hipe_signal_init(void);
-#else
-static __inline__ void hipe_signal_init(void) { }
-#endif
-
-#if defined(ERTS_SMP) && (defined(__i386__) || defined(__x86_64__))
 extern void hipe_thread_signal_init(void);
 #else
+static __inline__ void hipe_signal_init(void) { }
 static __inline__ void hipe_thread_signal_init(void) { }
 #endif
 
diff --git a/erts/emulator/hipe/hipe_sparc_bifs.m4 b/erts/emulator/hipe/hipe_sparc_bifs.m4
index 14330c2f1c..1b49fa57fd 100644
--- a/erts/emulator/hipe/hipe_sparc_bifs.m4
+++ b/erts/emulator/hipe/hipe_sparc_bifs.m4
@@ -28,7 +28,7 @@ include(`hipe/hipe_sparc_asm.m4')
 	.section ".text"
 	.align	4
 
-`#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+`#if defined(ERTS_ENABLE_LOCK_CHECK)
 #  define CALL_BIF(F)	set nbif_impl_##F, %o7; st %o7, [%o0+P_BIF_CALLEE]; call hipe_debug_bif_wrapper 
 #else
 #  define CALL_BIF(F)	call	nbif_impl_##F
diff --git a/erts/emulator/hipe/hipe_x86_bifs.m4 b/erts/emulator/hipe/hipe_x86_bifs.m4
index aecf67dc1b..9cb343d067 100644
--- a/erts/emulator/hipe/hipe_x86_bifs.m4
+++ b/erts/emulator/hipe/hipe_x86_bifs.m4
@@ -31,7 +31,7 @@ include(`hipe/hipe_x86_asm.m4')
 #define TEST_GOT_EXN	cmpl	$THE_NON_VALUE,%eax
 #endif'
 
-`#if defined(ERTS_ENABLE_LOCK_CHECK) && defined(ERTS_SMP)
+`#if defined(ERTS_ENABLE_LOCK_CHECK)
 #  define CALL_BIF(F)	movl $CSYM(nbif_impl_##F), P_BIF_CALLEE(P); call CSYM(hipe_debug_bif_wrapper) 
 #else
 #  define CALL_BIF(F)	call	CSYM(nbif_impl_##F)
diff --git a/erts/emulator/hipe/hipe_x86_signal.c b/erts/emulator/hipe/hipe_x86_signal.c
index be68d7d463..d3b6933155 100644
--- a/erts/emulator/hipe/hipe_x86_signal.c
+++ b/erts/emulator/hipe/hipe_x86_signal.c
@@ -45,10 +45,8 @@
 #include <signal.h>
 #include <stdio.h>
 #include <stdlib.h>
-#ifdef ERTS_SMP
 #include "sys.h"
 #include "erl_alloc.h"
-#endif
 #include "hipe_signal.h"
 
 #if defined(__GLIBC__) && __GLIBC__ == 2 && (__GLIBC_MINOR__ >= 3)
@@ -259,7 +257,6 @@ static void hipe_sigaltstack(void *ss_sp)
     }
 }
 
-#ifdef ERTS_SMP
 /*
  * Set up alternate signal stack for an Erlang process scheduler thread.
  */
@@ -269,7 +266,6 @@ void hipe_thread_signal_init(void)
        We use it to suppress false leak report from valgrind */
     hipe_sigaltstack(erts_alloc_permanent_cache_aligned(ERTS_ALC_T_HIPE_LL, SIGSTKSZ));
 }
-#endif
 
 /*
  * Set up alternate signal stack for the main thread,
@@ -277,10 +273,6 @@ void hipe_thread_signal_init(void)
  */
 static void hipe_sigaltstack_init(void)
 {
-#if !defined(ERTS_SMP)
-    static unsigned long my_sigstack[SIGSTKSZ/sizeof(long)];
-    hipe_sigaltstack(my_sigstack);
-#endif
 }
 
 /*
diff --git a/erts/emulator/internal_doc/GarbageCollection.md b/erts/emulator/internal_doc/GarbageCollection.md
new file mode 100644
index 0000000000..1d9e3f4160
--- /dev/null
+++ b/erts/emulator/internal_doc/GarbageCollection.md
@@ -0,0 +1,187 @@
+# Erlang Garbage Collector
+
+Erlang manages dynamic memory with a [tracing garbage collector](https://en.wikipedia.org/wiki/Tracing_garbage_collection). More precisely a per process generational semi-space copying collector using [Cheney's](#cheney) copy collection algorithm together with a global large object space.
+
+## Overview
+
+Each Erlang process has its own stack and heap which are allocated in the same memory block and grow towards each other. When the stack and the heap [meet](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/beam_emu.c#L387), the garbage collector is triggered and memory is reclaimed. If not enough memory was reclaimed, the heap will grow.
+
+### Creating Data
+
+Terms are created on the heap by evaluating expressions. There are two major types of terms: [immediate terms](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_term.h#L88-L97) which require no heap space (small integers, atoms, pids, port ids etc) and cons or [boxed terms](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_term.h#L106-L120) (tuple, big num, binaries etc) that do require heap space. Immediate terms do not need any heap space because they are embedded into the containing structure.
+
+Let's look at an example that returns a tuple with the newly created data.
+
+```erlang
+data(Foo) ->
+   Cons = [42|Foo],
+   Literal = {text, "hello world!"},
+   {tag, Cons, Literal}.
+```
+
+In this example we first create a new cons cell with an integer and a tuple with some text. Then a tuple of size three wrapping the other values with an atom tag is created and returned.
+
+On the heap tuples require a word size for each of its elements as well as for the header. Cons cells always require two words. Adding these things together, we get seven words for the tuples and 26 words for the cons cells. The string `"hello world!"` is a list of cons cells and thus requires 24 words. The atom `tag` and the integer `42` do not require any additional heap memory since it is an *immediate*. Adding all the terms together, the heap space required in this example should be 33 words.
+
+Compiling this code to beam assembly (`erlc -S`) shows exactly what is happening.
+
+```erlang
+    ...
+    {test_heap,6,1}.
+    {put_list,{integer,42},{x,0},{x,1}}.
+    {put_tuple,3,{x,0}}.
+    {put,{atom,tag}}.
+    {put,{x,1}}.
+    {put,{literal,{text,"hello world!"}}}.
+    return.
+```
+
+Looking at the assembler code we can see three things; The heap requirement in this function turns out to be only six words, as seen by the `{test_heap,6,1}` instruction. All the allocations are combined to a single instruction. The bulk of the data `{text, "hello world!"}` is a *literal*. Literals, sometimes referred to as constants, are not allocated in the function since they are a part of the module and allocated at load time.
+
+If there is not enough space available on the heap to satisfy the `test_heap` instructions request for memory, then a garbage collection is initiated. It may happen immediately in the `test_heap` instruction, or it can be delayed until a later time depending on what state the process is in. If the garbage collection is delayed, any memory needed will be allocated in heap fragments. Heap fragments are extra memory blocks that are a part of the young heap, but are not allocated in the contigious area where terms normally reside. See [The young heap](#the-young-heap) for more details.
+
+### The collector
+
+Erlang has a copying semi-space garbage collector. This means that when doing a garbage collection, the terms are copied from one distinct area, called the *from space*, to a new clean area, called the *to space*. The collector starts by [scanning the root-set](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1980) (stack, registers, etc).
+
+![Garbage collection: initial values](figures/gc-start.png)
+
+It follows all the pointers from the root-set to the heap and copies each term word by word to the *to space*.
+
+After the header word has been copied a [*move marker*](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.h#L45-L46) is destructively placed in it pointing to the term in the *to space*. Any other term that points to the already moved term will [see this move marker](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1125) and copy the referring pointer instead. For example, if the have the following Erlang code:
+
+```erlang
+foo(Arg) ->
+    T = {test, Arg},
+    {wrapper, T, T, T}.
+```
+
+Only one copy of T exists on the heap and during the garbage collection only the first time T is encountered will it be copied.
+
+![Garbage collection: root set scan](figures/gc-rootset-scan.png)
+
+After [all terms](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1089) referenced by the root-set have been copied, the collector scans the *to space* and copies all terms that these terms reference. When scanning, the collector steps through each term on the *to space* and any term still referencing the *from space* is copied over to the *to space*. Some terms contain non-term data (the payload of a on heap binary for instance). When encountered by the collector, these values are simply skipped.
+
+![Garbage collection: heap scan](figures/gc-heap-scan1.png)
+
+Every term object we can reach is copied to the *to space* and stored on top off the *scan stop* line, and then the scan stop is moved to the end of the last object.
+
+![Garbage collection: heap scan](figures/gc-heap-stop.png)
+
+When *scan stop* marker [catches up](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1103) to the *scan start* marker, the garbage collection is done. At this point we can [deallocate](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1206) the entire *from space* and therefore reclaim the entire young heap.
+
+## Generational Garbage Collection
+
+In addition to the collection algorithm described above, the Erlang garbage collector also provides generational garbage collection. An additional heap, called the old heap, is used where the long lived data is stored. The original heap is called the young heap, or sometimes the allocation heap.
+
+With this in mind we can look at the Erlang's garbage collection again. During the copy stage anything that should be copied to the young *to space* is instead copied to the old *to space* *if* it is [below the *high-watermark*](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L1127).
+
+![Garbage collection: heap scan](figures/gc-watermark.png)
+
+The [*high-watermark*](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_process.h#L1021) is placed where the previous garbage collection (described in [Overview](#overview)) ended and we have introduced a new area called the old heap. When doing the normal garbage collection pass, any term that is located below the high-watermark is copied to the old *to space* instead of the young.
+
+![Garbage collection: heap scan](figures/gc-watermark-2.png)
+
+In the next garbage collection, any pointers to the old heap will be ignored and not scanned. This way the garbage collector does not have to scan the long-lived terms.
+
+Generational garbage collection aims to increase performance at the expense of memory. This is achieved because only the young, smaller, heap is considered in most garbage collections.
+
+The generational [hypothesis](#ungar) predicts that most terms tend to die young, and for an immutable language such as Erlang, young terms die even faster than in other languages. So for most usage patterns the data in the new heap will die very soon after it is allocated. This is good because it limits the amount of data copied to the old heap and also because the garbage collection algorithm used is proportional to the amount of live data on the heap.
+
+One critical issue to note here is that any term on the young heap can reference terms on the old heap but *no* term on the old heap may refer to a term on the young heap. This is due to the nature of the copy algorithm. Anything referenced by an old heap term is not included in the reference tree, root-set and its followers, and hence is not copied. If it was, the data would be lost, fire and brimstone would rise to cover the earth. Fortunately, this comes naturally for Erlang because the terms are immutable and thus there can be no pointers modified on the old heap to point to the young heap.
+
+To reclaim data from the old heap, both young and old heaps are included during the collection and copied to a common *to space*. Both the *from space* of the young and old heap are then deallocated and the procedure will start over from the beginning. This type of garbage collection is called a full sweep and is triggered when the size of the area under the high-watermark is larger than the size of the free area of the old heap. It can also be triggered by doing a manual call to [erlang:garbage_collect()](http://erlang.org/doc/man/erlang.html#garbage_collect-0), or by running into the young garbage collection limit set by [spawn_opt(fun(),[{fullsweep_after, N}])](http://erlang.org/doc/man/erlang.html#spawn_opt-4) where N is the number of young garbage collections to do before forcing a garbage collection of both young and old heap.
+
+## The young heap
+
+The young heap, or the allocation heap, consists of the stack and heap as described in the Overview. However, it also includes any heap fragments that are attached to the heap. All of the heap fragments are considered to be above the high-watermark and part of the young generation. Heap fragments contain terms that either did not fit on the heap, or were created by another process and then attached to the heap. For instance if the bif binary_to_term created a term which does not fit on the current heap without doing a garbage collection, it will create a heap-fragment for the term and then schedule a garbage collection for later. Also if a message is sent to the process, the payload may be placed in a heap-fragment and that fragment is added to young heap when the message is matched in a receive clause.
+
+This procedure differs from how it worked prior to Erlang/OTP 19.0. Before 19.0, only a contiguous memory block where the young heap and stack resided was considered to be part of the young heap. Heap fragments and messages were immediately copied into the young heap before they could be inspected by the Erlang program. The behaviour introduced in 19.0 is superior in many ways - most significantly it reduces the number of necessary copy operations and the root set for garbage collection.
+
+## Sizing the heap
+
+As mentioned in the Overview the size of the heap [grows](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L247) to accommodate more data. Heaps grow in two stages, first a [variation of the Fibonacci sequence](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L199-L208) is used starting at 233 words. Then at about 1 mega words the heap only [grows in 20% increments](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L215-L227).
+
+There are two occasions when the young heap grows:
+
+* if the total size of the heap + message and heap fragments exceeds the current heap size.
+* if after a fullsweep, the total amount of live objects is greater than 75%.
+
+There are two occasions when the young heap is shrunk:
+
+* if after a young collection, the total amount of live objects is less than 25% of the heap and the young heap is "big"
+* if after a fullsweep, the total amount of live objects is less than 25% of the heap.
+
+The old heap is always one step ahead in the heap growth stages than the young heap.
+
+## Literals
+
+When garbage collecting a heap (young or old) all literals are left in place and not copied. To figure out if a term should be copied or not when doing a garbage collection the following pseudo code is used:
+
+```c
+if (erts_is_literal(ptr) || (on_old_heap(ptr) && !fullsweep)) {
+  /* literal or non fullsweep - do not copy */
+} else {
+  copy(ptr);
+}
+```
+
+The [`erts_is_literal`](https://github.com/erlang/otp/blob/OTP-19.0/erts/emulator/beam/global.h#L1452-L1465) check works differently on different architectures and operating systems.
+
+On 64 bit systems that allow mapping of unreserved virtual memory areas (most operating systems except Windows), an area of size 1 GB (by default) is mapped and then all literals are placed within that area. Then all that has to be done to determine if something is a literal or not is [two quick pointer checks](https://github.com/erlang/otp/blob/OTP-19.0/erts/emulator/beam/erl_alloc.h#L322-L324). This system relies on the fact that a memory page that has not been touched yet does not take any actual space. So even if 1 GB of virtual memory is mapped, only the memory which is actually needed for literals is allocated in ram. The size of the literal area is configurable through the +MIscs erts_alloc option.
+
+On 32 bit systems, there is not enough virtual memory space to allocate 1 GB for just literals, so instead small 256 KB sized literal regions are created on demand and a card mark bit-array of the entire 32 bit memory space is then used to determine if a term is a literal or not. Since the total memory space is only 32 bits, the card mark bit-array is only 256 words large. On a 64 bit system the same bit-array would have to be 1 tera words large, so this technique is only viable on 32 bit systems. Doing [lookups in the array](https://github.com/erlang/otp/blob/OTP-19.0/erts/emulator/beam/erl_alloc.h#L316-L319) is a little more expensive then just doing the pointer checks that can be done in 64 bit systems, but not extremely so.
+
+On 64 bit windows, on which erts_alloc cannot do unreserved virtual memory mappings, a [special tag](https://github.com/erlang/otp/blob/OTP-19.0/erts/emulator/beam/erl_term.h#L59) within the Erlang term object is used to determine if something [is a literal or not](https://github.com/erlang/otp/blob/OTP-19.0/erts/emulator/beam/erl_term.h#L248-L252). This is very cheap, however, the tag is only available on 64 bit machines, and it is possible to do a great deal of other nice optimizations with this tag in the future (like for instance a more compact list implementation) so it is not used on operating systems where it is not needed.
+
+This behaviour is different from how it worked prior to Erlang/OTP 19.0. Before 19.0 the literal check was done by checking if the pointer pointed to the young or old heap block. If it did not, then it was considered a literal. This lead to considerable overhead and strange memory usage scenarios, so it was removed in 19.0.
+
+## Binary heap
+
+The binary heap works as a large object space for binary terms that are greater than 64 bytes (from now on called off-heap binaries). The binary heap is [reference counted](https://en.wikipedia.org/wiki/Reference_counting) and a pointer to the off-heap binary is stored on the process heap. To keep track of when to decrement the reference counter of the off-heap binary, a linked list (the MSO - mark and sweep object list) containing funs and externals as well as off-heap binaries is woven through the heap. After a garbage collection is done, the [MSO list is swept](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L2299) and any off-heap binary that does not have a [move marker](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L2325) written into the header words has its reference [decremented and is potentially freed](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L2344-L2367).
+
+All items in the MSO list are ordered by the time they were added to the process heap, so when doing a minor garbage collection, the MSO sweeper only has to sweep until it [encounters an off-heap binary that is on the old heap](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_gc.c#L2369).
+
+### Virtual Binary heap
+
+Each process has a virtual binary heap associated with it that has the size of all the current off-heap binaries that the process has references to. The virtual binary heap also has a limit and grows and shrinks depending on how off-heap binaries are used by the process. The same growth and shrink mechanisms are used for the binary heap and for the term heap, so first a Fibonacci like series and then 20% growth.
+
+The virtual binary heap exists in order to [trigger](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/beam_emu.c#L364) garbage collections earlier when potentially there is a very large amount of off-heap binary data that could be reclaimed. This approach does not catch all problems with binary memory not being released soon enough, but it does catch a lot of them.
+
+## Messages
+
+Messages can become a part of the process heap at different times. This depends on how the process is configured.
+We can configure the behaviour of each process using `process_flag(message_queue_data, off_heap | on_heap)` or we can set a default for all processes at start using the option `+hmqd`.
+
+What do these different configurations do and when should we use them?
+Let's start by going through what happens when one Erlang process sends a message to another.
+The sending process needs to do a couple of things:
+
+1. calculate [how large](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_message.c#L1031) the message to be sent is
+2. [allocate enough space](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_message.c#L1033) to fit the entire message
+3. [copy](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_message.c#L1040) the message payload
+4. allocate a message container with some meta data
+5. [insert](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_message.c#L502) the message container in the receiver process' [message queue](https://github.com/erlang/otp/blob/OTP-18.0/erts/emulator/beam/erl_process.h#L1042)
+
+The process flag `message_queue_data`, of the receiver process, controls the message allocating strategy of the sender process in step 2 and also how the message data is treated by the garbage collector.
+
+The procedure above is different from how it worked prior to 19.0. Before 19.0 there was no configuration option, the behaviour was always very similar to how the `on_heap` option is in 19.0.
+
+### Message allocating strategies
+
+If set to `on_heap`, the sending process will first attempt to allocate the space for the message directly on the young heap block of the receiving process.
+This is not always possible as it requires taking the *main lock* of the receiving process. The main lock is also held when the process is executing. The possibility for a lock conflict is thus likely in an intensely collaborating system.
+If the sending process cannot acquire the main lock, a heap fragment is instead created for the message and the message payload is copied onto that.
+With the `off_heap` option the sender process always creates heap fragments for messages sent to that process.
+
+There are a bunch of different tradeoffs that come into play when trying to figure out which of the strategies you want to use.
+
+Using `off_heap` may seem like a nice way to get a more scalable system as you get very little contention on the main locks, however, allocating a heap fragment is more expensive than allocating on the heap of the receiving process. So if it is very unlikely that contention will occur, it is more efficient to try to allocate the message directly on the receiving process' heap.
+
+Using `on_heap` will force all messages to be part of on the young heap which will increase the amount of data that the garbage collector has to move. So if a garbage collection is triggered while processing a large amount of messages, they will be copied to the young heap. This in turn will lead to that the messages will quickly be promoted to the old heap and thus increase its size. This may be good or bad depending on exactly what the process does. A large old heap means that the young heap will also be larger, which in turn means that less garbage collections will be triggered while processing the message queue. This will temporarly increase the throughput of the process at the cost of more memory usage. However, if after all the messages have been consumed the process enters a state where a lot less messages are being received. Then it may be a long time before the next fullsweep garbage collection happens and the messages that are on the old heap will be there until that happens. So while `on_heap` is potentially faster than the other modes, it uses more memory for a longer time. This mode is the legacy mode which is almost how the message queue was handled before Erlang/OTP 19.0.
+
+Which one of these strategies is best depends a lot on what the process is doing and how it interacts with other processes. So, as always, profile the application and see how it behaves with the different options.
+
+   <a name="cheney">[1]</a>: C. J. Cheney. A nonrecursive list compacting algorithm. Commun. ACM, 13(11):677–678, Nov. 1970.
+
+   <a name="ungar">[2]</a>: D. Ungar. Generation scavenging: A non-disruptive high performance storage reclamation algorithm. SIGSOFT Softw. Eng. Notes, 9(3):157–167, Apr. 1984.
diff --git a/erts/emulator/internal_doc/beam_makeops.md b/erts/emulator/internal_doc/beam_makeops.md
new file mode 100644
index 0000000000..1da8d2ab05
--- /dev/null
+++ b/erts/emulator/internal_doc/beam_makeops.md
@@ -0,0 +1,1846 @@
+The beam\_makeops script
+=======================
+
+This document describes the **beam\_makeops** script.
+
+Introduction
+------------
+
+The **beam\_makeops** Perl script is used at build-time by both the
+compiler and runtime system.  Given a number of input files (all with
+the extension `.tab`), it will generate source files used by the
+Erlang compiler and by the runtime system to load and execute BEAM
+instructions.
+
+Essentially those `.tab` files define:
+
+* External generic BEAM instructions.  They are the instructions that
+are known to both the compiler and the runtime system.  Generic
+instructions are stable between releases.  New generic instructions
+with high numbers than previous instructions can be added in major
+releases.  The OTP 20 release has 159 external generic instructions.
+
+* Internal generic instructions.  They are known only to the runtime
+system and can be changed at any time without compatibility issues.
+They are created by transformation rules (described next).
+
+* Rules for transforming one or more generic instructions to other
+generic instructions.  The transformation rules allow combining,
+splitting, and removal of instructions, as well as shuffling operands.
+Because of the transformation rules, the runtime can have many
+internal generic instructions that are only known to runtime system.
+
+* Specific BEAM instructions.  The specific instructions are the
+instructions that are actually executed by the runtime system.  They
+can be changed at any time without causing compatibility issues.
+The loader translates generic instructions to specific instructions.
+In general, for each generic instruction, there exists a family of
+specific instructions.  The OTP 20 release has 389 specific
+instructions.
+
+* The implementation of specific instructions.
+
+Generic instructions have typed operands. Here are a few examples of
+operands for `move/2`:
+
+    {move,{atom,id},{x,5}}.
+    {move,{x,3},{x,0}}.
+    {move,{x,2},{y,1}}.
+
+When those instructions are loaded, the loader rewrites them
+to specific instructions:
+
+    move_cx id 5
+    move_xx 3 0
+    move_xy 2 1
+
+Corresponding to each generic instruction, there is a family of
+specific instructions.  The types that an instance of a specific
+instruction can handle are encoded in the instruction names.  For
+example, `move_xy` takes an X register number as the first operand and
+a Y register number as the second operand.  `move_cx` takes a tagged
+Erlang term as the first operand and an X register number as the
+second operand.
+
+An example: the move instruction
+--------------------------------
+
+Using the `move` instruction as an example, we will give a quick
+tour to show the main features of **beam\_makeops**.
+
+In the `compiler` application, in the file `genop.tab`, there is the
+following line:
+
+    64: move/2
+
+This is a definition of an external generic BEAM instruction. Most
+importantly it specifices that the opcode is 64.  It also defines that
+it has two operands.  The BEAM assembler will use the opcode when
+creating `.beam` files.  The compiler does not really need the arity,
+but it will use it as an internal sanity check when assembling the
+BEAM code.
+
+Let's have a look at `ops.tab` in `erts/emulator/beam`, where the
+specific `move` instructions are defined.  Here are a few of them:
+
+    move x x
+    move x y
+    move c x
+
+Each specific instructions is defined by following the name of the
+instruction with the types for each operand.  An operand type is a
+single letter.  For example, `x` means an X register, `y`
+means a Y register, and `c` is a "constant" (a tagged term such as
+an integer, an atom, or a literal).
+
+Now let's look at the implementation of the `move` instruction.  There
+are multiple files containing implementations of instructions in the
+`erts/emulator/beam` directory.  The `move` instruction is defined in
+`instrs.tab`.  It looks like this:
+
+    move(Src, Dst) {
+        $Dst = $Src;
+    }
+
+The implementation for an instruction largely follows the C syntax,
+except that the variables in the function head don't have any types.
+The `$` before an identifier denotes a macro expansion.  Thus,
+`$Src` will expand to the code to pick up the source operand for
+the instruction and `$Dst` to the code for the destination register.
+
+We will look at the code for each specific instruction in turn.  To
+make the code easier to understand, let's first look at the memory
+layout for the instruction `{move,{atom,id},{x,5}}`:
+
+         +--------------------+--------------------+
+    I -> |                 40 |       &&lb_move_cx |
+         +--------------------+--------------------+
+         |                        Tagged atom 'id' |
+         +--------------------+--------------------+
+
+This example and all other examples in the document assumes a 64-bit
+archictecture, and furthermore that pointers to C code fit in 32 bits.
+
+`I` in the BEAM virtual machine is the instruction pointer.  When BEAM
+executes an instruction, `I` points to the first word of the
+instruction.
+
+`&&lb_move_cx` is the address to C code that implements `move_cx`.  It
+is stored in the lower 32 bits of the word.  In the upper 32 bits is
+the byte offset to the X register; the register number 5 has been
+multiplied by the word size size 8.
+
+In the next word the tagged atom `id` is stored.
+
+With that background, we can look at the generated code for `move_cx`
+in `beam_hot.h`:
+
+    OpCase(move_cx):
+    {
+      BeamInstr next_pf = BeamCodeAddr(I[2]);
+      xb(BeamExtraData(I[0])) = I[1];
+      I += 2;
+      ASSERT(VALID_INSTR(next_pf));
+      GotoPF(next_pf);
+    }
+
+We will go through each line in turn.
+
+* `OpCase(move_cx):` defines a label for the instruction.  The
+`OpCase()` macro is defined in `beam_emu.c`.  It will expand this line
+to `lb_move_cx:`.
+
+* `BeamInstr next_pf = BeamCodeAddr(I[2]);` fetches the pointer to
+code for the next instruction to be executed.  The `BeamCodeAddr()`
+macro extracts the pointer from the lower 32 bits of the instruction
+word.
+
+* `xb(BeamExtraData(I[0])) = I[1];` is the expansion of `$Dst = $Src`.
+`BeamExtraData()` is a macro that will extract the upper 32 bits from
+the instruction word.  In this example, it will return 40 which is the
+byte offset for X register 5.  The `xb()` macro will cast a byte
+pointer to an `Eterm` pointer and dereference it.  The `I[1]` on
+the right side of the `=` fetches an Erlang term (the atom `id` in
+this case).
+
+* `I += 2` advances the instruction pointer to the next
+instruction.
+
+* In a debug-compiled emulator, `ASSERT(VALID_INSTR(next_pf));` makes
+sure that `next_pf` is a valid instruction (that is, that it points
+within the `process_main()` function in `beam_emu.c`).
+
+* `GotoPF(next_pf);` transfers control to the next instruction.
+
+Now let's look at the implementation of `move_xx`:
+
+    OpCase(move_xx):
+    {
+      Eterm tmp_packed1 = BeamExtraData(I[0]);
+      BeamInstr next_pf = BeamCodeAddr(I[1]);
+      xb((tmp_packed1>>BEAM_TIGHT_SHIFT)) = xb(tmp_packed1&BEAM_TIGHT_MASK);
+      I += 1;
+      ASSERT(VALID_INSTR(next_pf));
+      GotoPF(next_pf);
+    }
+
+We will go through the lines that are new or have changed compared to
+`move_cx`.
+
+* `Eterm tmp_packed1 = BeamExtraData(I[0]);` picks up both X register
+numbers packed into the upper 32 bits of the instruction word.
+
+* `BeamInstr next_pf = BeamCodeAddr(I[1]);` pre-fetches the address of
+the next instruction. Note that because both X registers operands fits
+into the instruction word, the next instruction is in the very next
+word.
+
+* `xb((tmp_packed1>>BEAM_TIGHT_SHIFT)) = xb(tmp_packed1&BEAM_TIGHT_MASK);`
+copies the source to the destination.  (For a 64-bit architecture,
+`BEAM_TIGHT_SHIFT` is 16 and `BEAM_TIGHT_MASK` is `0xFFFF`.)
+
+* `I += 1;` advances the instruction pointer to the next instruction.
+
+`move_xy` is almost identical to `move_xx`.  The only difference is
+the use of the `yb()` macro instead of `xb()` to reference the
+destination register:
+
+    OpCase(move_xy):
+    {
+      Eterm tmp_packed1 = BeamExtraData(I[0]);
+      BeamInstr next_pf = BeamCodeAddr(I[1]);
+      yb((tmp_packed1>>BEAM_TIGHT_SHIFT)) = xb(tmp_packed1&BEAM_TIGHT_MASK);
+      I += 1;
+      ASSERT(VALID_INSTR(next_pf));
+      GotoPF(next_pf);
+    }
+
+### Transformation rules ###
+
+Next let's look at how we can do some optimizations using transformation
+rules.  For simple instructions such as `move/2`, the instruction dispatch
+overhead can be substantial.  A simple optimization is to combine common
+instructions sequences to a single instruction.  One such common sequence
+is multiple `move` instructions moving X registers to Y registers.
+
+Using the following rule we can combine two `move` instructions
+to a `move2` instruction:
+
+    move X1=x Y1=y | move X2=x Y2=y => move2 X1 Y1 X2 Y2
+
+The left side of the arrow (`=>`) is a pattern.  If the pattern
+matches, the matching instructions will be replaced by the
+instructions on the right side.  Variables in a pattern must start
+with an uppercase letter just as in Erlang.  A pattern variable may be
+followed `=` and one or more type letters to constrain the match to
+one of those types.  The variables that are bound on the left side can
+be used on the right side.
+
+We will also need to define a specific instruction and an implementation:
+
+    # In ops.tab
+    move2 x y x y
+
+    // In instrs.tab
+    move2(S1, D1, S2, D2) {
+        Eterm V1, V2;
+        V1 = $S1;
+        V2 = $S2;
+        $D1 = V1;
+        $D2 = V2;
+    }
+
+When the loader has found a match and replaced the matched instructions,
+it will match the new instructions against the transformation rules.
+Because of that, we can define the rule for a `move3/6` instruction
+as follows:
+
+    move2 X1=x Y1=y X2=x Y2=y | move X3=x Y3=y => \
+          move3 X1 Y1 X2 Y2 X3 Y3
+
+(A `\` before a newline can be used to break a long line for readability.)
+
+It would also be possible to define it like this:
+
+    move X1=x Y1=y | move X2=x Y2=y | move X3=x Y3=y => \
+         move3 X1 Y1 X2 Y2 X3 Y3
+
+but in that case it must be defined before the rule for `move2/4`
+because the first matching rule will be applied.
+
+One must be careful not to create infinite loops.  For example, if we
+for some reason would want to reverse the operand order for the `move`
+instruction, we must not do like this:
+
+    move Src Dst => move Dst Src
+
+The loader would swap the operands forever.  To avoid the loop, we must
+rename the instruction.  For example:
+
+    move Src Dst => assign Dst Src
+
+This concludes the quick tour of the features of **beam\_makeops**.
+
+Short overview of instruction loading
+-------------------------------------
+
+To give some background to the rest of this document, here follows a
+quick overview of how instructions are loaded.
+
+* The loader reads and decodes one instruction at a time from the BEAM
+code and creates a generic instruction.  Many transformation rules
+must look at multiple instructions, so the loader will
+keep multiple generic instructions in a linked list.
+
+* The loader tries to apply transformation rules against the
+generic instructions in the linked list.  If a rule matches, the
+matched instructions will be removed and replaced with new
+generic instructions constructed from the right side of the
+transformation.
+
+* If a transformation rule matched, the loader applies the
+transformation rules again.
+
+* If no transformation rule match, the loader will begin rewriting
+the first of generic instructions to a specific instruction.
+
+* First the loader will search for a specific operation where the
+types for all operands match the type for the generic instruction.
+The first matching instruction will be selected.  **beam\_makeops**
+has ordered the specific instructions so that instructions with more
+specific operands comes before instructions with less specific
+operands.  For example, `move_nx` is more specific than `move_cx`.  If
+the first operand is `[]` (NIL), `move_nx` will be selected.
+
+* Given the opcode for the selected specific instruction, the loader
+looks up the pointer to the C code for the instruction and stores
+in the code area for the module being loaded.
+
+* The loader translates each operand to a machine word and stores it
+in the code area.  The operand type for the selected specific
+instruction guides the translation.  For example, if the type is `e`,
+the value of the operand is an index into an arry of external
+functions and will be translated to a pointer to the export entry for
+the function to call.  If the type is `x`, the number of the X
+register will be multiplied by the word size to produce a byte offset.
+
+* The loader runs the packing engine to pack multiple operands into a
+single word.  The packing engine is controlled by a small program,
+which is a string where each character is an instruction.  For
+example, the code to pack the operands for `move_xy` is `"22#"` (on a
+64-bit machine).  That program will pack the byte offsets for both
+registers into the same word as the pointer to C code.
+
+Running beam_makeops
+--------------------
+
+**beam\_makeops** is found in `$ERL_TOP/erts/emulator/utils`.  Options
+start with a hyphen (`-`).  The options are followed by the name of
+the input files.  By convention, all input files have the extension
+`.tab`, but is not enforced by **beam\_makeops**.
+
+### The -outdir option ###
+
+The option `-outdir Directory` specifies the output directory for
+the generated files.  Default is the current working directory.
+
+### Running beam_makeops for the compiler ###
+
+Give the option `-compiler` to produce output files for the compiler.
+The following files will be written to the output directory:
+
+* `beam_opcodes.erl` - Used primarily by `beam_asm` and `beam_diasm`.
+
+* `beam_opcode.hrl` - Used by `beam_asm`.  It contains tag definitions
+used for encoding instruction operands.
+
+The input file should only contain the definition of BEAM_FORMAT_NUMBER
+and external generic instructions.  (Everything else would be ignored.)
+
+### Running beam_makeops for the emulator ###
+
+Give the option `-emulator` to produce output files for the emulator.
+The following output files will be generated in the output directory.
+
+* `beam_hot.h`, `beam_warm.h`, `beam_cold.`h - Implementation of
+instructions.  Included inside the `process_main()` function in
+`beam_emu.c`.
+
+* `beam_opcodes.c` - Defines static data used by the loader
+(`beam_load.c`).  Data about generic instructions, specific
+instructions (including how to pack their operands), and
+transformation rules are all part of this file.
+
+* `beam_opcodes.h` - Miscellanous preprocessor definitions, mainly
+used by `beam_load.c` but also by `beam_{hot,warm,cold}.h`.
+
+* `beam_pred_funcs.h` - Included by `beam_load.c`.  Contains defines
+needed to call guard constraints in transformation rules.
+
+* `beam_tr_funcs.h` - Included by `beam_load.c`.  Contains defines
+needed to call a C function to the right of a transformation rule.
+
+The following options can be given:
+
+* `wordsize 32|64` - Defines the word size.  Default is 32.
+
+* `code-model Model` - The code model as given to `-mcmodel` option
+for GCC.  Default is `unknown`.  If the code model is `small` (and
+the word size is 64 bits), **beam\_makeops** will pack operands
+into the upper 32 bits of the instruction word.
+
+* `DSymbol=0|1` - Defines the value for a symbol.  The symbol can be
+used in `%if` and `%unless` directives.
+
+Syntax of .tab files
+--------------------
+
+### Comments ###
+
+Any line starting with `#` is a comment and is ignored.
+
+A line with `//` is also a comment.  It is recommended to only
+use this style of comments in files that define implementations of
+instructions.
+
+A long line can be broken into shorter lines by a placing a`\` before
+the newline.
+
+### Variable definitions ###
+
+A variable definition binds a variable to a Perl variable.  It is only
+meaningful to add a new definition if **beam\_makeops** is updated
+at the same time to use the variable.  A variable definition looks this:
+
+*name*=*value*[;]
+
+where *name* is the name of a Perl variable in **beam\_makeops**,
+and *value* is the value to be given to the variable.  The line
+can optionally end with a `;` (to avoid messing up the
+C indentation mode in Emacs).
+
+Here follows a description of the variables that are defined.
+
+#### BEAM\_FORMAT\_NUMBER ####
+
+`genop.tab` has the following definition:
+
+    BEAM_FORMAT_NUMBER=0
+
+It defines the version of the instruction set (which will be
+included in the code header in the BEAM code).  Theoretically,
+the version could be bumped, and all instructions changed.
+In practice, we would have two support two instruction sets
+in the runtime system for at least two releases, so it will
+probably never happen in practice.
+
+#### GC\_REGEXP ####
+
+In `macros.tab`, there is a definition of `GC_REGEXP`.
+It will be described in [a later section](#the-gc_regexp-definition).
+
+### Directives ###
+
+There are directives to classify specific instructions depending
+on how frequently used they are:
+
+* `%hot` - Implementation will be placed in `beam_hot.h`. Frequently
+executed instructions.
+
+* `%warm` - Implementation will be placed in `beam_warm.h`.  Binary
+syntax instructions.
+
+* `%cold` - Implementation will be placed in `beam_cold.h`. Trace
+instructions and infrequently used instructions.
+
+Default is `%hot`.  The directives will be applied to declarations
+of the specific instruction that follow.  Here is an example:
+
+    %cold
+    is_number f? xy
+    %hot
+
+#### Conditional compilation directives ####
+
+The `%if` directive includes a range of lines if a condition is
+true.  For example:
+
+    %if ARCH_64
+    i_bs_get_integer_32 x f? x
+    %endif
+
+The specific instruction `i_bs_get_integer_32` will only be defined
+on a 64-bit machine.
+
+The condition can be inverted by using `%unless` instead of `%if`:
+
+    %unless NO_FPE_SIGNALS
+    fcheckerror p => i_fcheckerror
+    i_fcheckerror
+    fclearerror
+    %endif
+
+It is also possible to add an `%else` clause:
+
+    %if ARCH_64
+    BS_SAFE_MUL(A, B, Fail, Dst) {
+        Uint64 res = ($A) * ($B);
+        if (res / $B != $A) {
+            $Fail;
+        }
+        $Dst = res;
+    }
+    %else
+    BS_SAFE_MUL(A, B, Fail, Dst) {
+        Uint64 res = (Uint64)($A) * (Uint64)($B);
+        if ((res >> (8*sizeof(Uint))) != 0) {
+            $Fail;
+        }
+        $Dst = res;
+    }
+    %endif
+
+#### Symbols that are defined in directives ####
+
+The following symbols are always defined.
+
+* `ARCH_64` - is 1 for a 64-bit machine, and 0 otherwise.
+* `ARCH_32` - is 1 for 32-bit machine, and 1 otherwise.
+
+The `Makefile` for building the emulator currently defines the
+following symbols by using the `-D` option on the command line for
+**beam\_makeops**.
+
+* `NO_FPE_SIGNALS` - 1 if FPE signals are not enable in runtime system,
+0 otherwise.
+* `USE_VM_PROBES` - 1 if the runtime system is compiled to use VM probes (support for dtrace or systemtap), 0 otherwise.
+
+### Defining external generic instructions ###
+
+External generic BEAM instructions are known to both the compiler and
+the runtime system.  They remain stable between releases.  A new major
+release may add more external generic instructions, but must not change
+the semantics for a previously defined instruction.
+
+The syntax for an external generic instruction is as follows:
+
+*opcode*: [-]*name*/*arity*
+
+*opcode* is an integer greater than or equal to 1.
+
+*name* is an identifier starting with a lowercase letter.  *arity* is
+an integer denoting the number of operands.
+
+*name* can optionally be preceded by `-` to indicate that it has been
+obsoleted.  The compiler is not allowed to generate BEAM files that
+use obsolete instructions and the loader will refuse to load BEAM
+files that use obsolete instructions.
+
+It only makes sense to define external generic instructions in the
+file `genop.tab` in `lib/compiler/src`, because the compiler must
+know about them in order to use them.
+
+New instructions must be added at the end of the file, with higher
+numbers than the previous instructions.
+
+### Defining internal generic instructions ###
+
+Internal generic instructions are known only to the runtime
+system and can be changed at any time without compatibility issues.
+
+There are two ways to define internal generic instructions:
+
+* Implicitly when a specific instruction is defined.  This is by far
+the most common way.  Whenever a specific instruction is created,
+**beam\_makeops** automatically creates an internal generic instruction
+if it does not previously exist.
+
+* Explicitly.  This is necessary only when a generic instruction does
+not have any corresponding specific instruction.
+
+The syntax for an internal generic instruction is as follows:
+
+*name*/*arity*
+
+*name* is an identifier starting with a lowercase letter.  *arity* is
+an integer denoting the number of operands.
+
+### About generic instructions in general ###
+
+Each generic instruction has an opcode.  The opcode is an integer,
+greater than or equal to 1. For an external generic instruction, it
+must be explicitly given `genop.tab`, while internal generic
+instructions are automatically numbered by **beam\_makeops**.
+
+The identity of a generic instruction is its name combined with its
+arity.  That means that it is allowed to define two distinct generic
+instructions having the same name but with different arities.  For
+example:
+
+    move_window/5
+    move_window/6
+
+Each operand of a generic instruction is tagged with its type.  A generic
+instruction can have one of the following types:
+
+* `x` - X register.
+
+* `y` - Y register.
+
+* `l` - Floating point register number.
+
+* `i` - Tagged literal integer.
+
+* `a` - Tagged literal atom.
+
+* `n` - NIL (`[]`, the empty list).
+
+* `q` - Literal that don't fit in a word, that is an object stored on
+the heap such as a list or tuple.  Any heap object type is supported,
+even types that don't have real literals such as external references.
+
+* `f` - Non-zero failure label.
+
+* `p` - Zero failure label.
+
+* `u` - Untagged integer that fits in a machine word.  It is used for many
+different purposes, such as the number of live registers in `test_heap/2`,
+as a reference to the export for `call_ext/2`, and as the flags operand for
+binary syntax instructions.  When the generic instruction is translated to a
+specific instruction, the type for the operand in the specific operation will
+tell the loader how to treat the operand.
+
+* `o` - Overflow.  If the value for an `u` operand does not fit in a machine
+word, the type of the operand will be changed to `o` (with no associated
+value).  Currently only used internally in the loader in the guard constraint
+function `binary_too_big()`.
+
+* `v` - Arity value.  Only used internally in the loader.
+
+
+### Defining specific instructions ###
+
+The specific instructions are known only to the runtime system and
+are the instructions that are actually executed.  They can be changed
+at any time without causing compatibility issues.
+
+A specific instruction can have at most 6 operands.
+
+A specific instruction is defined by first giving its name followed by
+the types for each operand.  For example:
+
+     move x y
+
+Internally, for example in the generated code and in the output from
+the BEAM disassembler, the instruction `move x y` will be called `move_xy`.
+
+The name for a specific instruction is an identifier starting with a
+lowercase letter.  A type is an lowercase or uppercase letter.
+
+All specific instructions with a given name must have the same number
+of operands. That is, the following is **not** allowed:
+
+     move x x
+     move x y x y
+
+Here follows the type letters that more or less directly corresponds
+to the types for generic instructions.
+
+* `x` - X register.  Will be loaded as a byte offset to the X register
+relative to the base of X register array.  (Can be packed with other
+operands.)
+
+* `y` - Y register.  Will be loaded as a byte offset to the Y register
+relative to the stack frame. (Can be packed with other operands.)
+
+* `r` - X register 0.  An implicit operand that will not be stored in
+the loaded code.
+
+* `l` - Floating point register number.  (Can be packed with other
+operands.)
+
+* `i` - Tagged literal integer (a SMALL that will fit in one word).
+
+* `a` - Tagged atom.
+
+* `n` - NIL or the empty list.  (Will not be stored in the loaded code.)
+
+* `q` - Tagged CONS or BOXED pointer.  That is, a term such as a list
+or tuple.  Any heap object type is supported, even types that don't
+have real literals such as external references.
+
+* `f` - Failure label (non-zero).  The target for a branch
+or call instruction.
+
+* `p` - The 0 failure label, meaning that an exception should be raised
+if the instruction fails.  (Will not be stored in the loaded code.)
+
+* `c` - Any literal term; that is, immediate literals such as SMALL,
+and CONS or BOXED pointers to literals.  (Can be used where the
+operand in the generic instruction has one of the types `i`, `a`, `n`,
+or `q`.)
+
+The types that follow do a type test of the operand at runtime; thus,
+they are generally more expensive in terms of runtime than the types
+described earlier.  However, those operand types are needed to avoid a
+combinatorial explosion in the number of specific instructions and
+overall code size of `process_main()`.
+
+* `s` - Tagged source: X register, Y register, or a literal term.  The
+tag will be tested at runtime to retrieve the value from an X
+register, a Y register, or simply use the value as a tagged Erlang
+term.  (Implementation note: An X register is tagged as a pid, and a Y
+register as a port.  Therefore the literal term must not contain a
+port or pid.)
+
+* `S` - Tagged source register (X or Y).  The tag will be tested at
+runtime to retrieve the value from an X register or a Y register.  Slighly
+cheaper than `s`.
+
+* `d` - Tagged destination register (X or Y).  The tag will be tested
+at runtime to set up a pointer to the destination register.  If the
+instrution performs a garbarge collection, it must use the
+`$REFRESH_GEN_DEST()` macro to refresh the pointer before storing to
+it (there are more details about that in a later section).
+
+* `j` - A failure label (combination of `f` and `p`).  If the branch target 0,
+an exception will be raised if instruction fails, otherwise control will be
+transfered to the target address.
+
+The types that follows are all applied to an operand that has the `u`
+type.
+
+* `t` - An untagged integer that will fit in 12 bits (0-4096).  It can be
+packed with other operands in a word.  Most often used as the number
+of live registers in instructions such as `test_heap`.
+
+* `I` - An untagged integer that will fit in 32 bits.  It can be
+packed with other operands in a word on a 64-bit system.
+
+* `W` - Untagged integer or pointer.  Not possible to pack with other
+operands.
+
+* `e` - Pointer to an export entry.  Use by call instructions that call
+other modules, such as `call_ext`.
+
+* `L` - A label.  Only used by the `label/1` instruction.
+
+* `b` - Pointer to BIF.  Used by instructions that BIFs, such as
+`call_bif`.
+
+* `A` - A tagged arityvalue.  Used in instructions that test the arity
+of a tuple.
+
+* `P` - A byte offset into a tuple.
+
+* `Q` - A byte offset into the stack.  Used for updating the frame
+pointer register.  Can be packed with other operands.
+
+When the loader translates a generic instruction a specific
+instruction, it will choose the most specific instruction that will
+fit the types.  Consider the following two instructions:
+
+    move c x
+    move n x
+
+The `c` operand can encode any literal value, including NIL.  The
+`n` operand only works for NIL.  If we have the generic instruction
+`{move,nil,{x,1}}`, the loader will translate it to `move_nx 1`
+because `move n x` is more specific.  `move_nx` could be slightly
+faster or smaller (depending on the architecture), because the `[]`
+is not stored explicitly as an operand.
+
+#### Syntactic sugar for specific instructions ####
+
+It is possible to specify more than one type letter for each operand.
+Here is an example:
+
+    move cxy xy
+
+This is syntactic sugar for:
+
+    move c x
+    move c y
+    move x x
+    move x y
+    move y x
+    move y y
+
+Note the difference between `move c xy` and `move c d`.  Note that `move c xy`
+is equivalent to the following two definitions:
+
+    move c x
+    move c y
+
+On the other hand, `move c d` is a single instruction.  At runtime,
+the `d` operand will be tested to see whether it refers to an X
+register or a Y register, and a pointer to the register will be set
+up.
+
+#### The '?' type modifier ####
+
+The character `?` can be added to the end of an operand to indicate
+that the operand will not be used every time the instruction is executed.
+For example:
+
+    allocate_heap t I t?
+    is_eq_exact f? x xy
+
+In `allocate_heap`, the last operand is the number of live registers.
+It will only be used if there is not enough heap space and a garbage
+collection must be performed.
+
+In `is_eq_exact`, the failure address (the first operand) will only be
+used if the two register operands are not equal.
+
+Knowing that an operand is not always used can improve how packing
+is done for some instructions.
+
+For the `allocate_heap` instruction, without the `?` the packing would
+be done like this:
+
+         +--------------------+--------------------+
+    I -> |       Stack needed | &&lb_allocate_heap +
+         +--------------------+--------------------+
+         |        Heap needed | Live registers     +
+         +--------------------+--------------------+
+
+"Stack needed" and "Heap needed" are always used, but they are in
+different words.  Thus, at runtime the `allocate_heap` instruction
+must read both words from memory even though it will not always use
+"Live registers".
+
+With the `?`, the operands will be packed like this:
+
+         +--------------------+--------------------+
+    I -> |     Live registers | &&lb_allocate_heap +
+         +--------------------+--------------------+
+         |        Heap needed |       Stack needed +
+         +--------------------+--------------------+
+
+Now "Stack needed" and "Heap needed" are in the same word.
+
+### Defining transformation rules ###
+
+Transformation rules are used to rewrite generic instructions to other
+generic instructions.  The transformations rules are applied
+repeatedly until no rule match.  At that point, the first instruction
+in the resulting instruction sequence will be converted to a specific
+instruction and added to the code for the module being loaded.  Then
+the transformation rules for the remaining instructions are run in the
+same way.
+
+A rule is recognized by its right-pointer arrow: `=>`.  To the left of
+the arrow is one or more instruction patterns, separated by `|`.  To
+the right of the arrow is zero or more instructions, separated by `|`.
+If the instructions from the BEAM code matches the instruction
+patterns on the left side, they will be replaced with instructions on
+the right side (or removed if there are no instructions on the right).
+
+#### Defining instruction patterns ####
+
+We will start looking at the patterns on the left side of the arrow.
+
+A pattern for an instruction consists of its name, followed by a pattern
+for each of its operands.  The operand patterns are separated by spaces.
+
+The simplest possible pattern is a variable.  Just like in Erlang,
+a variable must begin with an uppercase letter.  If the same variable is
+used in multiple operands, the pattern will only match if the operands
+are equal.  For example:
+
+    move Same Same =>
+
+This pattern will match if the operands for `move` are the same.  If
+the pattern match, the instruction will be removed.  (That used to be an
+actual rule a long time ago when the compiler would occasionally produce
+instructions such as `{move,{x,2},{x,2}}`.)
+
+Variables that have been bound on the left side can be used on the
+right side.  For example, this rule will rewrite all `move` instructions
+to `assign` instructions with the operands swapped:
+
+    move Src Dst => assign Dst Src
+
+If we only want to match operands of a certain type, we can
+use a type constraint.  A type constraint consists of one or more
+lowercase letters, each specifying a type.  For example:
+
+    is_integer Fail an => jump Fail
+
+The second operand pattern, `an`, will match if the second operand is
+either an atom or NIL (the empty list).  In case of a match, the
+`is_integer/2` instruction will be replaced with a `jump/1`
+instruction.
+
+An operand pattern can bind a variable and constrain the type at the
+same time by following the variable with a `=` and the constraint.
+For example:
+
+    is_eq_exact Fail=f R=xy C=q => i_is_eq_exact_literal Fail R C
+
+Here the `is_eq_exact` instruction is replaced with a specialized instruction
+that only compares literals, but only if the first operand is a register and
+the second operand is a literal.
+
+#### Further constraining patterns ####
+
+In addition to specifying a type letter, the actual value for the type can
+be specified.  For example:
+
+    move C=c x==1 => move_x1 C
+
+Here the second operand of `move` is constrained to be X register 1.
+
+When specifying an atom constraint, the atom is written as it would be
+in the C source code.  That is, it needs an `am_` prefix, and it must
+be listed in `atom.names`.  For example:
+
+    is_boolean Fail=f a==am_true =>
+    is_boolean Fail=f a==am_false =>
+
+There are several constraints available for testing whether a call is to a BIF
+or a function.
+
+The constraint `u$is_bif` will test whether the given operand refers to a BIF.
+For example:
+
+    call_ext u Bif=u$is_bif => call_bif Bif
+    call_ext u Func         => i_call_ext Func
+
+The `call_ext` instruction can be used to call functions written in
+Erlang as well as BIFs (or more properly called SNIFs).  The
+`u$is_bif` constraint will match if the operand refers to a BIF (that
+is, if it is listed in the file `bif.tab`).  Note that `u$is_bif`
+should only be applied to operands that are known to contain an index
+to the import table chunk in the BEAM file (such operands have the
+type `b` or `e` in the corresponding specific instruction).  If
+applied to other `u` operands, it will at best return a nonsense
+result.
+
+The `u$is_not_bif` constraint matches if the operand does not refer to
+a BIF (not listed in `bif.tab`).  For example:
+
+    move S X0=x==0 | line Loc | call_ext_last Ar Func=u$is_not_bif D => \
+         move S X0 | call_ext_last Ar Func D
+
+The `u$bif:Module:Name/Arity` constraint tests whether the given
+operand refers to a specific BIF.  Note that `Module:Name/Arity`
+**must** be an existing BIF defined in `bif.tab`, or there will
+be a compilation error.  It is useful when a call to a specific BIF
+should be replaced with an instruction as in this example:
+
+    gc_bif2 Fail Live u$bif:erlang:splus/2 S1 S2 Dst => \
+         gen_plus Fail Live S1 S2 Dst
+
+Here the call to the GC BIF `'+'/2` will be replaced with the instruction
+`gen_plus/5`.  Note that the same name as used in the C source code must be
+used for the BIF, which in this case is `splus`.  It is defined like this
+in `bit.tab`:
+
+    ubif erlang:'+'/2 splus_2
+
+The `u$func:Module:Name/Arity` will test whether the given operand is a
+a specific function.  Here is an example:
+
+    bif1 Fail u$func:erlang:is_constant/1 Src Dst => too_old_compiler
+
+`is_constant/1` used to be a BIF a long time ago.  The transformation
+replaces the call with the `too_old_compiler` instruction which will produce
+a nicer error message than the default error would be for a missing guard BIF.
+
+#### Type constraints allowed in patterns ####
+
+Here are all type letters that are allowed on the left side of a transformation
+rule.
+
+* `u` - An untagged integer that fits in a machine word.
+
+* `x` - X register.
+
+* `y` - Y register.
+
+* `l` - Floating point register number.
+
+* `i` - Tagged literal integer.
+
+* `a` - Tagged literal atom.
+
+* `n` - NIL (`[]`, the empty list).
+
+* `q` - Literals that don't fit in a word, such as list or tuples.
+
+* `f` - Non-zero failure label.
+
+* `p` - The zero failure label.
+
+* `j` - Any label.  Equivalent to `fp`.
+
+* `c` - Any literal term.  Equivalent to `ainq`.
+
+* `s` - X register, Y register, or any literal term.  Equivalent to `xyc`.
+
+* `d` - X or Y register.  Equivalent to `xy`.  (In a pattern `d` will
+match both source and destination registers.  As an operand in a specific
+instruction, it must only be used for a destination register.)
+
+* `o` - Overflow.  An untagged integer that does not fit in a machine word.
+
+#### Guard constraints ####
+
+If the constraints described so far is not enough, additional
+constraints can be written in C in `beam_load.c` and be called as a
+guard function on the left side of the transformation.  If the guard
+function returns a non-zero value, the matching of the rule will
+continue, otherwise the match will fail.  For example:
+
+    ensure_map Lit=q | literal_is_map(Lit) =>
+
+The guard test `literal_is_map/1` tests whether the given literal is a map.
+If the literal is a map, the instruction is unnecessary and can be removed.
+
+It is outside the scope for this document to describe in detail how such
+guard functions are written, but for the curious here is the implementation
+of `literal_is_map()`:
+
+    static int
+    literal_is_map(LoaderState* stp, GenOpArg Lit)
+    {
+        Eterm term;
+
+        ASSERT(Lit.type == TAG_q);
+        term = stp->literals[Lit.val].term;
+        return is_map(term);
+    }
+
+#### Handling instruction with variable number of operands ####
+
+Some instructions, such as `select_val/3`, essentially has a variable
+number of operands.  Such instructions have a `{list,[...]}` operand
+as their last operand in the BEAM assembly code.  For example:
+
+    {select_val,{x,0},
+                {f,1},
+                {list,[{atom,b},{f,4},{atom,a},{f,5}]}}.
+
+The loader will convert a `{list,[...]}` operand to an `u` operand whose
+value is the number of elements in the list, followed by each element in
+the list.  The instruction above would be translated to the following
+generic instruction:
+
+    {select_val,{x,0},{f,1},{u,4},{atom,b},{f,4},{atom,a},{f,5}}
+
+To match a variable number of arguments we need to use the special
+operand type `*` like this:
+
+    select_val Src=aiq Fail=f Size=u List=* => \
+        i_const_select_val Src Fail Size List
+
+This transformation renames a `select_val/3` instruction
+with a constant source operand to `i_const_select_val/3`.
+
+#### Constructing new instructions on the right side ####
+
+The most common operand on the right side is a variable that was bound while
+matching the left side.  For example:
+
+    trim N Remaining => i_trim N
+
+An operand can also be a type letter to construct an operand of that type.
+Each type has a default value.  For example, the type `x` has the default
+value 1023, which is the highest X register.  That makes `x` on the right
+side a convenient shortcut for a temporary X register.  For example:
+
+    is_number Fail Literal=q => move Literal x | is_number Fail x
+
+If the second operand for `is_number/2` is a literal, it will be moved to
+X register 1023.  Then `is_number/2` will test whether the value stored in
+X register 1023 is a number.
+
+This kind of transformation is useful when it is rare that an operand can
+be anything else but a register.  In the case of `is_number/2`, the second
+operand is always a register unless the compiler optimizations have been
+disabled.
+
+If the default value is not suitable, the type letter can be followed
+by `=` and a value.  Most types take an integer value.  The value for
+an atom is written the same way as in the C source code.  For example,
+the atom `false` is written as `am_false`.  The atom must be listed in
+`atom.names`.
+
+Here is an example showing how values can be specified:
+
+    bs_put_utf32 Fail=j Flags=u Src=s => \
+        i_bs_validate_unicode Fail Src | \
+        bs_put_integer Fail i=32 u=1 Flags Src
+
+#### Type letters on the right side ####
+
+Here follows all types that are allowed to be used in operands for
+instructions being constructed on the right side of a transformation
+rule.
+
+* `u` - Construct an untagged integer.  The default value is 0.
+
+* `x` - X register.  The default value is 1023.  That makes `x` convenient to
+use as a temporary X register.
+
+* `y` - Y register.  The default value is 0.
+
+* `l` - Foating point register number.  The default value is 0.
+
+* `i` - Tagged literal integer.  The default value is 0.
+
+* `a` - Tagged atom.  The default value is the empty atom (`am_Empty`).
+
+* `n` - NIL (`[]`, the empty list).
+
+#### Function call on the right side ####
+
+Transformations that are not possible to describe with the rule
+language as described here can be written as a C function in
+`beam_load.c` and called from the right side of a transformation.  The
+left side of the transformation will perform the match and bind
+operands to variables.  The variables can then be passed to a
+generator function on the right side.  For example:
+
+    bif2 Fail=j u$bif:erlang:element/2 Index=s Tuple=xy Dst=d => \
+        gen_element(Jump, Index, Tuple, Dst)
+
+This transformation rule matches a call to the BIF `element/2`.
+The operands will be captured and the function `gen_element()` will
+be called.
+
+`gen_element()` will produce one of two instructions depending
+on `Index`.  If `Index` is an integer in the range from 1 up to
+the maximum tuple size, the instruction `i_fast_element/2` will
+be produced, otherwise the instruction `i_element/4` will be
+produced.  The corresponding specific instructions are:
+
+    i_fast_element xy j? I d
+    i_element xy j? s d
+
+The `i_fast_element/2` instruction is faster because the tuple is
+already an untagged integer.  It also knows that the index is at least
+1, so it does not have to test for that.  The `i_element/4`
+instruction will have to fetch the index from a register, test that it
+is an integer, and untag the integer.
+
+It is outside the scope of this document to describe in detail how
+generator functions are written, but for the curious, here is the
+implementation of `gen_element()`:
+
+    static GenOp*
+    gen_element(LoaderState* stp, GenOpArg Fail,
+       GenOpArg Index, GenOpArg Tuple, GenOpArg Dst)
+    {
+        GenOp* op;
+
+        NEW_GENOP(stp, op);
+        op->arity = 4;
+        op->next = NULL;
+
+        if (Index.type == TAG_i && Index.val > 0 &&
+           Index.val <= ERTS_MAX_TUPLE_SIZE &&
+           (Tuple.type == TAG_x || Tuple.type == TAG_y)) {
+            op->op = genop_i_fast_element_4;
+            op->a[0] = Tuple;
+            op->a[1] = Fail;
+            op->a[2].type = TAG_u;
+            op->a[2].val = Index.val;
+            op->a[3] = Dst;
+        } else {
+            op->op = genop_i_element_4;
+            op->a[0] = Tuple;
+            op->a[1] = Fail;
+            op->a[2] = Index;
+            op->a[3] = Dst;
+        }
+
+        return op;
+    }
+}
+
+### Defining the implementation ###
+
+The actual implementation of instructions are also defined in `.tab`
+files processed by **beam\_makeops**.  For practical reasons,
+instruction definitions are stored in several files, at the time of
+writing in the following files:
+
+    bif_instrs.tab
+    arith_instrs.tab
+    bs_instrs.tab
+    float_instrs.tab
+    instrs.tab
+    map_instrs.tab
+    msg_instrs.tab
+    select_instrs.tab
+    trace_instrs.tab
+
+There is also a file that only contains macro definitions:
+
+    macros.tab
+
+The syntax of each file is similar to C code.  In fact, most of
+the contents *is* C code, interspersed with macro invocations.
+
+To allow Emacs to auto-indent the code, each file starts with the
+following line:
+
+    // -*- c -*-
+
+To avoid messing up the indentation, all comments are written
+as C++ style comments (`//`) instead of `#`.  Note that a comment
+must start at the beginning of a line.
+
+The meat of an instruction definition file are macro definitions.
+We have seen this macro definition before:
+
+    move(Src, Dst) {
+        $Dst = $Src;
+    }
+
+A macro definitions must start at the beginning of the line (no spaces
+allowed), the opening curly bracket must be on the same line, and the
+finishing curly bracket must be at the beginning of a line.  It is
+recommended that the macro body is properly indented.
+
+As a convention, the macro arguments in the head all start with an
+uppercase letter.  In the body, the macro arguments can be expanded
+by preceding them with `$`.
+
+A macro definition whose name and arity matches a family of
+specific instructions is assumed to be the implementation of that
+instruction.
+
+A macro can also be invoked from within another macro.  For example,
+`move_deallocate_return/2` avoids repeating code by invoking
+`$deallocate_return()` as a macro:
+
+    move_deallocate_return(Src, Deallocate) {
+        x(0) = $Src;
+        $deallocate_return($Deallocate);
+    }
+
+Here is the definition of `deallocate_return/1`:
+
+    deallocate_return(Deallocate) {
+        //| -no_next
+        int words_to_pop = $Deallocate;
+        SET_I((BeamInstr *) cp_val(*E));
+        E = ADD_BYTE_OFFSET(E, words_to_pop);
+        CHECK_TERM(x(0));
+        DispatchReturn;
+    }
+
+The expanded code for `move_deallocate_return` will look this:
+
+    OpCase(move_deallocate_return_cQ):
+    {
+      x(0) = I[1];
+      do {
+        int words_to_pop = Qb(BeamExtraData(I[0]));
+        SET_I((BeamInstr *) cp_val(*E));
+        E = ADD_BYTE_OFFSET(E, words_to_pop);
+        CHECK_TERM(x(0));
+        DispatchReturn;
+      } while (0);
+    }
+
+When expanding macros, **beam\_makeops** wraps the expansion in a
+`do`/`while` wrapper unless **beam\_makeops** can clearly see that no
+wrapper is needed.  In this case, the wrapper is needed.
+
+Note that arguments for macros cannot be complex expressions, because
+the arguments are split on `,`.  For example, the following would
+not work because **beam\_makeops** would split the expression into
+two arguments:
+
+    $deallocate_return(get_deallocation(y, $Deallocate));
+
+#### Code generation directives ####
+
+Within macro definitions, `//` comments are in general not treated
+specially.  They will be copied to the file with the generated code
+along with the rest of code in the body.
+
+However, there is an exception. Within a macro definition, a line that
+starts with whitespace followed by `//|` is treated specially.  The
+rest of the line is assumed to contain directives to control code
+generation.
+
+Currently, two code generation directives are recognized:
+
+* `-no_prefetch`
+* `-no_next`
+
+##### The -no_prefetch directive #####
+
+To see what `-no_prefetch` does, let's first look at the default code
+generation.  Here is the code generated for `move_cx`:
+
+    OpCase(move_cx):
+    {
+      BeamInstr next_pf = BeamCodeAddr(I[2]);
+      xb(BeamExtraData(I[0])) = I[1];
+      I += 2;
+      ASSERT(VALID_INSTR(next_pf));
+      GotoPF(next_pf);
+    }
+
+Note that the very first thing done is to fetch the address to the
+next instruction.  The reason is that it usually improves performance.
+
+Just as a demonstration, we can add a `-no_prefetch` directive to
+the `move/2` instruction:
+
+    move(Src, Dst) {
+        //| -no_prefetch
+        $Dst = $Src;
+    }
+
+We can see that the prefetch is no longer done:
+
+    OpCase(move_cx):
+    {
+      xb(BeamExtraData(I[0])) = I[1];
+      I += 2;
+      ASSERT(VALID_INSTR(*I));
+      Goto(*I);
+    }
+
+When would we want to turn off the prefetch in practice?
+
+In instructions that will not always execute the next instruction.
+For example:
+
+    is_atom(Fail, Src) {
+        if (is_not_atom($Src)) {
+            $FAIL($Fail);
+        }
+    }
+
+    // From macros.tab
+    FAIL(Fail) {
+        //| -no_prefetch
+        $SET_I_REL($Fail);
+        Goto(*I);
+    }
+
+`is_atom/2` may either execute the next instruction (if the second
+operand is an atom) or branch to the failure label.
+
+The generated code looks like this:
+
+    OpCase(is_atom_fx):
+    {
+      if (is_not_atom(xb(I[1]))) {
+        ASSERT(VALID_INSTR(*(I + (fb(BeamExtraData(I[0]))) + 0)));
+        I += fb(BeamExtraData(I[0])) + 0;;
+        Goto(*I);;
+      }
+      I += 2;
+      ASSERT(VALID_INSTR(*I));
+      Goto(*I);
+    }
+
+##### The -no_next directive #####
+
+Next we will look at when the `-no_next` directive can be used.  Here
+is the `jump/1` instruction:
+
+    jump(Fail) {
+        $JUMP($Fail);
+    }
+
+    // From macros.tab
+    JUMP(Fail) {
+        //| -no_next
+        $SET_I_REL($Fail);
+        Goto(*I);
+    }
+
+The generated code looks like this:
+
+    OpCase(jump_f):
+    {
+      ASSERT(VALID_INSTR(*(I + (fb(BeamExtraData(I[0]))) + 0)));
+      I += fb(BeamExtraData(I[0])) + 0;;
+      Goto(*I);;
+    }
+
+If we remove the `-no_next` directive, the code would look like this:
+
+    OpCase(jump_f):
+    {
+      BeamInstr next_pf = BeamCodeAddr(I[1]);
+      ASSERT(VALID_INSTR(*(I + (fb(BeamExtraData(I[0]))) + 0)));
+      I += fb(BeamExtraData(I[0])) + 0;;
+      Goto(*I);;
+      I += 1;
+      ASSERT(VALID_INSTR(next_pf));
+      GotoPF(next_pf);
+    }
+
+In the end, the C compiler will probably optimize this code to the
+same native code as the first version, but the first version is certainly
+much easier to read for human readers.
+
+#### Macros in the macros.tab file ####
+
+The file `macros.tab` contains many useful macros.  When implementing
+new instructions it is good practice to look through `macros.tab` to
+see if any of existing macros can be used rather than re-inventing
+the wheel.
+
+We will describe a few of the most useful macros here.
+
+##### The GC_REGEXP definition #####
+
+The following line defines a regular expression that will recognize
+a call to a function that does a garbage collection:
+
+     GC_REGEXP=erts_garbage_collect|erts_gc|GcBifFunction;
+
+The purpose is that **beam\_makeops** can verify that an instruction
+that does a garbage collection and has an `d` operand uses the
+`$REFRESH_GEN_DEST()` macro.
+
+If you need to define a new function that does garbage collection,
+you should give it the prefix `erts_gc_`.  If that is not possible
+you should update the regular expression so that it will match your
+new function.
+
+##### FAIL(Fail) #####
+
+Branch to `$Fail`.  Will suppress prefetch (`-no_prefetch`).  Typical use:
+
+    is_nonempty_list(Fail, Src) {
+        if (is_not_list($Src)) {
+            $FAIL($Fail);
+        }
+    }
+
+##### JUMP(Fail) #####
+
+Branch to `$Fail`.  Suppresses generation of dispatch of the next
+instruction (`-no_next`).  Typical use:
+
+    jump(Fail) {
+        $JUMP($Fail);
+    }
+
+##### GC_TEST(NeedStack, NeedHeap, Live) #####
+
+`$GC_TEST(NeedStack, NeedHeap, Live)` tests that given amount of
+stack space and heap space is available.  If not it will do a
+garbage collection.  Typical use:
+
+    test_heap(Nh, Live) {
+        $GC_TEST(0, $Nh, $Live);
+    }
+
+##### AH(NeedStack, NeedHeap, Live) #####
+
+`AH(NeedStack, NeedHeap, Live)` allocates a stack frame and
+optionally additional heap space.
+
+#### Pre-defined macros and variables ####
+
+**beam\_makeops** defines several built-in macros and pre-bound variables.
+
+##### The NEXT_INSTRUCTION pre-bound variable #####
+
+The NEXT_INSTRUCTION is a pre-bound variable that is available in
+all instructions.  It expands to the address of the next instruction.
+
+Here is an example:
+
+    i_call(CallDest) {
+        SET_CP(c_p, $NEXT_INSTRUCTION);
+        $DISPATCH_REL($CallDest);
+    }
+
+When calling a function, the return address is first stored in `c_p->cp`
+(using the `SET_CP()` macro defined in `beam_emu.c`), and then control is
+transferred to the callee.  Here is the generated code:
+
+    OpCase(i_call_f):
+    {
+      SET_CP(c_p, I+1);
+      ASSERT(VALID_INSTR(*(I + (fb(BeamExtraData(I[0]))) + 0)));
+      I += fb(BeamExtraData(I[0])) + 0;;
+      DTRACE_LOCAL_CALL(c_p, erts_code_to_codemfa(I));
+      Dispatch();;
+    }
+
+We can see that that `$NEXT_INSTRUCTION` has been expanded to `I+1`.
+That makes sense since the size of the `i_call_f/1` instruction is
+one word.
+
+##### The IP_ADJUSTMENT pre-bound variable #####
+
+`$IP_ADJUSTMENT` is usually 0.  In a few combined instructions
+(described below) it can be non-zero.  It is used like this
+in `macros.tab`:
+
+    SET_I_REL(Offset) {
+        ASSERT(VALID_INSTR(*(I + ($Offset) + $IP_ADJUSTMENT)));
+        I += $Offset + $IP_ADJUSTMENT;
+    }
+
+Avoid using `IP_ADJUSTMENT` directly.  Use `SET_I_REL()` or
+one of the macros that invoke such as `FAIL()` or `JUMP()`
+defined in `macros.tab`.
+
+#### Pre-defined macro functions ####
+
+##### The IF() macro #####
+
+`$IF(Expr, IfTrue, IfFalse)` evaluates `Expr`, which must be a valid
+Perl expression (which for simple numeric expressions have the same
+syntax as C).  If `Expr` evaluates to 0, the entire `IF()` expression will be
+replaced with `IfFalse`, otherwise it will be replaced with `IfTrue`.
+
+See the description of `OPERAND_POSITION()` for an example.
+
+##### The OPERAND\_POSITION() macro #####
+
+`$OPERAND_POSITION(Expr)` returns the position for `Expr`, if
+`Expr` is an operand that is not packed.  The first operand is
+at position 1.
+
+Returns 0 otherwise.
+
+This macro could be used like this in order to share code:
+
+    FAIL(Fail) {
+        //| -no_prefetch
+        $IF($OPERAND_POSITION($Fail) == 1 && $IP_ADJUSTMENT == 0,
+            goto common_jump,
+            $DO_JUMP($Fail));
+    }
+
+    DO_JUMP(Fail) {
+        $SET_I_REL($Fail);
+        Goto(*I));
+    }
+
+    // In beam_emu.c:
+    common_jump:
+       I += I[1];
+       Goto(*I));
+
+
+#### The $REFRESH\_GEN\_DEST() macro ####
+
+When a specific instruction has a `d` operand, early during execution
+of the instruction, a pointer will be initialized to point to the X or
+Y register in question.
+
+If there is a garbage collection before the result is stored,
+the stack will move and if the `d` operand refered to a Y
+register, the pointer will no longer be valid.  (Y registers are
+stored on the stack.)
+
+In those circumstances, `$REFRESH_GEN_DEST()` must be invoked
+to set up the pointer again.  **beam\_makeops** will notice
+if there is a call to a function that does a garbage collection and
+`$REFRESH_GEN_DEST()` is not called.
+
+Here is a complete example.  The `new_map` instruction is defined
+like this:
+
+    new_map d t I
+
+It is implemented like this:
+
+    new_map(Dst, Live, N) {
+        Eterm res;
+
+        HEAVY_SWAPOUT;
+        res = erts_gc_new_map(c_p, reg, $Live, $N, $NEXT_INSTRUCTION);
+        HEAVY_SWAPIN;
+        $REFRESH_GEN_DEST();
+        $Dst = res;
+        $NEXT($NEXT_INSTRUCTION+$N);
+    }
+
+If we have forgotten the `$REFRESH_GEN_DEST()` there would be a message
+similar to this:
+
+    pointer to destination register is invalid after GC -- use $REFRESH_GEN_DEST()
+    ... from the body of new_map at beam/map_instrs.tab(30)
+
+#### Combined instructions ####
+
+**Problem**: For frequently executed instructions we want to use
+"fast" operands types such as `x` and `y`, as opposed to `s` or `S`.
+To avoid an explosion in code size, we want to share most of the
+implementation between the instructions.  Here are the specific
+instructions for `i_increment/5`:
+
+    i_increment r W t d
+    i_increment x W t d
+    i_increment y W t d
+
+The `i_increment` instruction is implemented like this:
+
+    i_increment(Source, IncrementVal, Live, Dst) {
+        Eterm increment_reg_source = $Source;
+        Eterm increment_val = $IncrementVal;
+        Uint live;
+        Eterm result;
+
+        if (ERTS_LIKELY(is_small(increment_reg_val))) {
+            Sint i = signed_val(increment_reg_val) + increment_val;
+            if (ERTS_LIKELY(IS_SSMALL(i))) {
+                $Dst = make_small(i);
+                $NEXT0();
+            }
+        }
+        live = $Live;
+        HEAVY_SWAPOUT;
+        reg[live] = increment_reg_val;
+        reg[live+1] = make_small(increment_val);
+        result = erts_gc_mixed_plus(c_p, reg, live);
+        HEAVY_SWAPIN;
+        ERTS_HOLE_CHECK(c_p);
+        if (ERTS_LIKELY(is_value(result))) {
+            $REFRESH_GEN_DEST();
+            $Dst = result;
+            $NEXT0();
+        }
+        ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
+        goto find_func_info;
+    }
+
+There will be three almost identical copies of the code.  Given the
+size of the code, that could be too high cost to pay.
+
+To avoid the three copies of the code, we could use only one specific
+instruction:
+
+    i_increment S W t d
+
+(The same implementation as above will work.)
+
+That reduces the code size, but is slower because `S` means that
+there will be extra code to test whether the operand refers to an X
+register or a Y register.
+
+**Solution**: We can use "combined instructions".  Combined
+instructions are combined from instruction fragments.  The
+bulk of the code can be shared.
+
+Here we will show how `i_increment` can be implemented as a combined
+instruction.  We will show each individual fragment first, and then
+show how to connect them together.  First we will need a variable that
+we can store the value fetched from the register in:
+
+    increment.head() {
+        Eterm increment_reg_val;
+    }
+
+The name `increment` is the name of the group that the fragment
+belongs to.  Note that it does not need to have the same
+name as the instruction.  The group name is followed by `.` and
+the name of the fragment.  The name `head` is pre-defined.
+The code in it will be placed at the beginning of a block, so
+that all fragments in the group can access it.
+
+Next we define the fragment that will pick up the value from the
+register from the first operand:
+
+    increment.fetch(Src) {
+        increment_reg_val = $Src;
+    }
+
+We call this fragment `fetch`.  This fragment will be duplicated three
+times, one for each value of the first operand (`r`, `x`, and `y`).
+
+Next we define the main part of the code that do the actual incrementing.
+
+    increment.execute(IncrementVal, Live, Dst) {
+        Eterm increment_val = $IncrementVal;
+        Uint live;
+        Eterm result;
+
+        if (ERTS_LIKELY(is_small(increment_reg_val))) {
+            Sint i = signed_val(increment_reg_val) + increment_val;
+            if (ERTS_LIKELY(IS_SSMALL(i))) {
+                $Dst = make_small(i);
+                $NEXT0();
+            }
+        }
+        live = $Live;
+        HEAVY_SWAPOUT;
+        reg[live] = increment_reg_val;
+        reg[live+1] = make_small(increment_val);
+        result = erts_gc_mixed_plus(c_p, reg, live);
+        HEAVY_SWAPIN;
+        ERTS_HOLE_CHECK(c_p);
+        if (ERTS_LIKELY(is_value(result))) {
+            $REFRESH_GEN_DEST();
+            $Dst = result;
+            $NEXT0();
+        }
+        ASSERT(c_p->freason != BADMATCH || is_value(c_p->fvalue));
+        goto find_func_info;
+    }
+
+We call this fragment `execute`.  It will handle the three remaining
+operands (`W t d`).  There will only be one copy of this fragment.
+
+Now that we have defined the fragments, we need to inform
+**beam\_makeops** how they should be connected:
+
+    i_increment := increment.fetch.execute;
+
+To the left of the `:=` is the name of the specific instruction that
+should be implemented by the fragments, in this case `i_increment`.
+To the right of `:=` is the name of the group with the fragments,
+followed by a `.`.  Then the name of the fragments in the group are
+listed in the order they should be executed.  Note that the `head`
+fragment is not listed.
+
+The line ends in `;` (to avoid messing up the indentation in Emacs).
+
+(Note that in practice the `:=` line is usually placed before the
+fragments.)
+
+The generated code looks like this:
+
+    {
+      Eterm increment_reg_val;
+      OpCase(i_increment_rWtd):
+      {
+        increment_reg_val = r(0);
+      }
+      goto increment__execute;
+
+      OpCase(i_increment_xWtd):
+      {
+        increment_reg_val = xb(BeamExtraData(I[0]));
+      }
+      goto increment__execute;
+
+      OpCase(i_increment_yWtd):
+      {
+        increment_reg_val = yb(BeamExtraData(I[0]));
+      }
+      goto increment__execute;
+
+      increment__execute:
+      {
+        // Here follows the code from increment.execute()
+        .
+        .
+        .
+    }
+
+##### Some notes about combined instructions #####
+
+The operands that are different must be at
+the beginning of the instruction.  All operands in the last
+fragment must have the same operands in all variants of
+the specific instruction.
+
+As an example, the following specific instructions cannot be
+implemented as a combined instruction:
+
+    i_times j? t x x d
+    i_times j? t x y d
+    i_times j? t s s d
+
+We would have to change the order of the operands so that the
+two operands that are different are placed first:
+
+    i_times x x j? t d
+    i_times x y j? t d
+    i_times s s j? t d
+
+We can then define:
+
+    i_times := times.fetch.execute;
+
+    times.head {
+        Eterm op1, op2;
+    }
+
+    times.fetch(Src1, Src2) {
+        op1 = $Src1;
+        op2 = $Src2;
+    }
+
+    times.execute(Fail, Live, Dst) {
+        // Multiply op1 and op2.
+        .
+        .
+        .
+    }
+
+Several instructions can share a group.  As an example, the following
+instructions have different names, but in the end they all create a
+binary.  The last two operands are common for all of them:
+
+    i_bs_init_fail       xy j? t? x
+    i_bs_init_fail_heap s I j? t? x
+    i_bs_init                W t? x
+    i_bs_init_heap         W I t? x
+
+The instructions are defined like this (formatted with extra
+spaces for clarity):
+
+    i_bs_init_fail_heap := bs_init . fail_heap . verify . execute;
+    i_bs_init_fail      := bs_init . fail      . verify . execute;
+    i_bs_init           := bs_init .           .  plain . execute;
+    i_bs_init_heap      := bs_init .               heap . execute;
+
+Note that the first two instruction have three fragments, while the
+other two only have two fragments.  Here are the fragments:
+
+    bs_init_bits.head() {
+        Eterm num_bits_term;
+        Uint num_bits;
+        Uint alloc;
+    }
+
+    bs_init_bits.plain(NumBits) {
+        num_bits = $NumBits;
+        alloc = 0;
+    }
+
+    bs_init_bits.heap(NumBits, Alloc) {
+        num_bits = $NumBits;
+        alloc = $Alloc;
+    }
+
+    bs_init_bits.fail(NumBitsTerm) {
+        num_bits_term = $NumBitsTerm;
+        alloc = 0;
+    }
+
+    bs_init_bits.fail_heap(NumBitsTerm, Alloc) {
+        num_bits_term = $NumBitsTerm;
+        alloc = $Alloc;
+    }
+
+    bs_init_bits.verify(Fail) {
+        // Verify the num_bits_term, fail using $FAIL
+        // if there is a problem.
+	.
+	.
+	.
+    }
+
+    bs_init_bits.execute(Live, Dst) {
+       // Long complicated code to a create a binary.
+       .
+       .
+       .
+    }
+
+The full definitions of those instructions can be found in `bs_instrs.tab`.
+The generated code can be found in `beam_warm.h`.
diff --git a/erts/emulator/internal_doc/figures/.gitignore b/erts/emulator/internal_doc/figures/.gitignore
new file mode 100644
index 0000000000..c2813ac866
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/.gitignore
@@ -0,0 +1 @@
+*.eps
\ No newline at end of file
diff --git a/erts/emulator/internal_doc/figures/Makefile b/erts/emulator/internal_doc/figures/Makefile
new file mode 100644
index 0000000000..111cb393fb
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/Makefile
@@ -0,0 +1,20 @@
+# In order to update the figures you have to have both dia
+# and imagemagick installed.
+
+DIAGRAMS=$(wildcard *.dia)
+EPS_DIAGRAMS=$(patsubst %.dia,%.eps,$(DIAGRAMS))
+PNG_DIAGRAMS=$(patsubst %.dia,%.png,$(DIAGRAMS))
+
+diagrams: $(EPS_DIAGRAMS)
+
+png: $(PNG_DIAGRAMS)
+
+update_png: png
+	git add $(PNG_DIAGRAMS)
+	git commit -m "Update internal docs figures"
+
+%.eps: %.dia
+	dia --export=$@ $<
+
+%.png: %.eps
+	convert $< -resize 65% $@
diff --git a/erts/emulator/internal_doc/figures/gc-heap-scan1.dia b/erts/emulator/internal_doc/figures/gc-heap-scan1.dia
new file mode 100644
index 0000000000..b8a32dc092
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-heap-scan1.dia
diff --git a/erts/emulator/internal_doc/figures/gc-heap-scan1.png b/erts/emulator/internal_doc/figures/gc-heap-scan1.png
new file mode 100644
index 0000000000..9724fc8698
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-heap-scan1.png
diff --git a/erts/emulator/internal_doc/figures/gc-heap-stop.dia b/erts/emulator/internal_doc/figures/gc-heap-stop.dia
new file mode 100644
index 0000000000..af219958c6
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-heap-stop.dia
diff --git a/erts/emulator/internal_doc/figures/gc-heap-stop.png b/erts/emulator/internal_doc/figures/gc-heap-stop.png
new file mode 100644
index 0000000000..ec79790d36
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-heap-stop.png
diff --git a/erts/emulator/internal_doc/figures/gc-rootset-scan.dia b/erts/emulator/internal_doc/figures/gc-rootset-scan.dia
new file mode 100644
index 0000000000..d6147740e5
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-rootset-scan.dia
diff --git a/erts/emulator/internal_doc/figures/gc-rootset-scan.png b/erts/emulator/internal_doc/figures/gc-rootset-scan.png
new file mode 100644
index 0000000000..06509f83c3
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-rootset-scan.png
diff --git a/erts/emulator/internal_doc/figures/gc-start.dia b/erts/emulator/internal_doc/figures/gc-start.dia
new file mode 100644
index 0000000000..41832d29b0
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-start.dia
diff --git a/erts/emulator/internal_doc/figures/gc-start.png b/erts/emulator/internal_doc/figures/gc-start.png
new file mode 100644
index 0000000000..6c29e9adb0
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-start.png
diff --git a/erts/emulator/internal_doc/figures/gc-watermark-2.dia b/erts/emulator/internal_doc/figures/gc-watermark-2.dia
new file mode 100644
index 0000000000..2ea3be3fb2
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-watermark-2.dia
diff --git a/erts/emulator/internal_doc/figures/gc-watermark-2.png b/erts/emulator/internal_doc/figures/gc-watermark-2.png
new file mode 100644
index 0000000000..bca110282d
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-watermark-2.png
diff --git a/erts/emulator/internal_doc/figures/gc-watermark.dia b/erts/emulator/internal_doc/figures/gc-watermark.dia
new file mode 100644
index 0000000000..a5de7565dc
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-watermark.dia
diff --git a/erts/emulator/internal_doc/figures/gc-watermark.png b/erts/emulator/internal_doc/figures/gc-watermark.png
new file mode 100644
index 0000000000..b835c7694e
--- /dev/null
+++ b/erts/emulator/internal_doc/figures/gc-watermark.png
diff --git a/erts/emulator/nifs/common/prim_buffer_nif.c b/erts/emulator/nifs/common/prim_buffer_nif.c
new file mode 100644
index 0000000000..a8ef5fc355
--- /dev/null
+++ b/erts/emulator/nifs/common/prim_buffer_nif.c
@@ -0,0 +1,512 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#define STATIC_ERLANG_NIF 1
+
+#include "erl_nif.h"
+#include "config.h"
+#include "sys.h"
+
+#ifdef VALGRIND
+#  include <valgrind/memcheck.h>
+#endif
+
+#define ACCUMULATOR_SIZE (2 << 10)
+
+#define FIND_NIF_RESCHEDULE_SIZE (1 << 20)
+
+/* NIF interface declarations */
+static int load(ErlNifEnv *env, void** priv_data, ERL_NIF_TERM load_info);
+static int upgrade(ErlNifEnv *env, void** priv_data, void** old_priv_data, ERL_NIF_TERM load_info);
+static void unload(ErlNifEnv *env, void* priv_data);
+
+static ErlNifResourceType *rtype_buffer;
+
+static ERL_NIF_TERM am_ok;
+static ERL_NIF_TERM am_error;
+
+static ERL_NIF_TERM am_lock_order_violation;
+
+static ERL_NIF_TERM am_acquired;
+static ERL_NIF_TERM am_busy;
+
+static ERL_NIF_TERM am_continue;
+
+static ERL_NIF_TERM am_out_of_memory;
+static ERL_NIF_TERM am_not_found;
+
+typedef struct {
+#ifdef DEBUG
+    erts_atomic32_t concurrent_users;
+#endif
+
+    ErlNifBinary accumulator;
+    size_t accumulated_bytes;
+    int accumulator_present;
+
+    ErlNifIOQueue *queue;
+
+    erts_atomic32_t external_lock;
+} buffer_data_t;
+
+static ERL_NIF_TERM new_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM peek_head_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM skip_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM size_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM write_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM copying_read_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM find_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM trylock_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM unlock_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ErlNifFunc nif_funcs[] = {
+    {"new", 0, new_nif},
+    {"size", 1, size_nif},
+    {"peek_head", 1, peek_head_nif},
+    {"copying_read", 2, copying_read_nif},
+    {"write", 2, write_nif},
+    {"skip", 2, skip_nif},
+    {"find_byte_index", 2, find_nif},
+    {"try_lock", 1, trylock_nif},
+    {"unlock", 1, unlock_nif},
+};
+
+ERL_NIF_INIT(prim_buffer, nif_funcs, load, NULL, upgrade, unload)
+
+static void gc_buffer(ErlNifEnv *env, void* data);
+
+static int load(ErlNifEnv *env, void** priv_data, ERL_NIF_TERM load_info)
+{
+    am_ok = enif_make_atom(env, "ok");
+    am_error = enif_make_atom(env, "error");
+
+    am_lock_order_violation = enif_make_atom(env, "lock_order_violation");
+    am_acquired = enif_make_atom(env, "acquired");
+    am_busy = enif_make_atom(env, "busy");
+
+    am_continue = enif_make_atom(env, "continue");
+
+    am_out_of_memory = enif_make_atom(env, "out_of_memory");
+    am_not_found = enif_make_atom(env, "not_found");
+
+    rtype_buffer = enif_open_resource_type(env, NULL, "gc_buffer", gc_buffer,
+        ERL_NIF_RT_CREATE, NULL);
+
+    *priv_data = NULL;
+
+    return 0;
+}
+
+static void unload(ErlNifEnv *env, void* priv_data)
+{
+
+}
+
+static int upgrade(ErlNifEnv *env, void** priv_data, void** old_priv_data, ERL_NIF_TERM load_info)
+{
+    if(*old_priv_data != NULL) {
+        return -1; /* Don't know how to do that */
+    }
+
+    if(*priv_data != NULL) {
+        return -1; /* Don't know how to do that */
+    }
+
+    if(load(env, priv_data, load_info)) {
+        return -1;
+    }
+
+    return 0;
+}
+
+static void gc_buffer(ErlNifEnv *env, void* data) {
+    buffer_data_t *buffer = (buffer_data_t*)data;
+
+    if(buffer->accumulator_present) {
+        enif_release_binary(&buffer->accumulator);
+    }
+
+    enif_ioq_destroy(buffer->queue);
+}
+
+static int get_buffer_data(ErlNifEnv *env, ERL_NIF_TERM opaque, buffer_data_t **buffer) {
+    return enif_get_resource(env, opaque, rtype_buffer, (void **)buffer);
+}
+
+/* Copies a number of bytes from the head of the iovec, skipping "vec_skip"
+ * vector elements followed by "byte_skip" bytes on the target vector. */
+static void copy_from_iovec(SysIOVec *iovec, int vec_len, int vec_skip,
+        size_t byte_skip, size_t size, char *data) {
+
+    size_t bytes_copied, skip_offset;
+    int vec_index;
+
+    skip_offset = byte_skip;
+    vec_index = vec_skip;
+    bytes_copied = 0;
+
+    while(bytes_copied < size) {
+        size_t block_size, copy_size;
+        char *block_start;
+
+        ASSERT(vec_index < vec_len);
+
+        block_start = (char*)iovec[vec_index].iov_base;
+        block_size = iovec[vec_index].iov_len;
+
+        copy_size = MIN(size - bytes_copied, block_size - skip_offset);
+        sys_memcpy(&data[bytes_copied], &block_start[skip_offset], copy_size);
+
+        bytes_copied += copy_size;
+        skip_offset = 0;
+
+        vec_index++;
+    }
+}
+
+/* Convenience function for copy_from_iovec over queues. */
+static void copy_from_queue(ErlNifIOQueue *queue, int queue_skip,
+    size_t byte_skip, size_t size, char *data) {
+
+    SysIOVec *queued_data;
+    int queue_length;
+
+    queued_data = enif_ioq_peek(queue, &queue_length);
+    ASSERT(queue_skip < queue_length);
+
+    copy_from_iovec(queued_data, queue_length, queue_skip, byte_skip, size, data);
+}
+
+static int enqueue_write_accumulator(buffer_data_t *buffer) {
+    ASSERT(!buffer->accumulator_present ^ (buffer->accumulated_bytes > 0));
+
+    if(buffer->accumulator_present && buffer->accumulated_bytes > 0) {
+        if(!enif_realloc_binary(&buffer->accumulator, buffer->accumulated_bytes)) {
+            return 0;
+        } else if(!enif_ioq_enq_binary(buffer->queue, &buffer->accumulator, 0)) {
+            return 0;
+        }
+
+        /* The queue owns the accumulator now. */
+        buffer->accumulator_present = 0;
+        buffer->accumulated_bytes = 0;
+    }
+
+    return 1;
+}
+
+static int combine_small_writes(buffer_data_t *buffer, ErlNifIOVec *iovec) {
+    ASSERT(!buffer->accumulator_present ^ (buffer->accumulated_bytes > 0));
+
+    if(buffer->accumulated_bytes + iovec->size >= ACCUMULATOR_SIZE) {
+        if(iovec->size >= (ACCUMULATOR_SIZE / 2)) {
+            return 0;
+        }
+
+        if(!enqueue_write_accumulator(buffer)) {
+            return 0;
+        }
+    }
+
+    if(!buffer->accumulator_present) {
+        if(!enif_alloc_binary(ACCUMULATOR_SIZE, &buffer->accumulator)) {
+            return 0;
+        }
+
+        buffer->accumulator_present = 1;
+    }
+
+    copy_from_iovec(iovec->iov, iovec->iovcnt, 0, 0, iovec->size,
+            (char*)&buffer->accumulator.data[buffer->accumulated_bytes]);
+    buffer->accumulated_bytes += iovec->size;
+
+    return 1;
+}
+
+/* *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** */
+
+static ERL_NIF_TERM new_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+    ERL_NIF_TERM result;
+
+    buffer = (buffer_data_t*)enif_alloc_resource(rtype_buffer, sizeof(buffer_data_t));
+    buffer->queue = enif_ioq_create(ERL_NIF_IOQ_NORMAL);
+
+    if(buffer->queue != NULL) {
+#ifdef DEBUG
+        erts_atomic32_init_nob(&buffer->concurrent_users, 0);
+#endif
+        erts_atomic32_init_nob(&buffer->external_lock, 0);
+
+        buffer->accumulator_present = 0;
+        buffer->accumulated_bytes = 0;
+
+        result = enif_make_resource(env, buffer);
+    } else {
+        result = enif_raise_exception(env, am_out_of_memory);
+    }
+
+    enif_release_resource(buffer);
+
+    return result;
+}
+
+static ERL_NIF_TERM size_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    size_t total_size;
+
+    if(argc != 1 || !get_buffer_data(env, argv[0], &buffer)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    total_size = enif_ioq_size(buffer->queue);
+
+    if(buffer->accumulator_present) {
+        total_size += buffer->accumulated_bytes;
+    } else {
+        ASSERT(buffer->accumulated_bytes == 0);
+    }
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    return enif_make_uint64(env, total_size);
+}
+
+static ERL_NIF_TERM copying_read_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    ERL_NIF_TERM result;
+    unsigned char *data;
+    Uint64 block_size;
+
+    if(argc != 2 || !get_buffer_data(env, argv[0], &buffer)
+                 || !enif_get_uint64(env, argv[1], &block_size)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    if(!enqueue_write_accumulator(buffer)) {
+        return enif_raise_exception(env, am_out_of_memory);
+    }
+
+    if(enif_ioq_size(buffer->queue) < block_size) {
+        return enif_make_badarg(env);
+    }
+
+    data = enif_make_new_binary(env, block_size, &result);
+
+    if(block_size > 0) {
+        copy_from_queue(buffer->queue, 0, 0, block_size, (char*)data);
+        enif_ioq_deq(buffer->queue, block_size, NULL);
+    }
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    return result;
+}
+
+static ERL_NIF_TERM write_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    ErlNifIOVec vec, *iovec = &vec;
+    ERL_NIF_TERM tail;
+
+    if(argc != 2 || !get_buffer_data(env, argv[0], &buffer)
+                 || !enif_inspect_iovec(env, 64, argv[1], &tail, &iovec)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    if(!combine_small_writes(buffer, iovec)) {
+        if(!enqueue_write_accumulator(buffer) || !enif_ioq_enqv(buffer->queue, iovec, 0)) {
+            return enif_raise_exception(env, am_out_of_memory);
+        }
+    }
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    if(!enif_is_empty_list(env, tail)) {
+        const ERL_NIF_TERM new_argv[2] = {argv[0], tail};
+
+        return enif_schedule_nif(env, "write", 0, &write_nif, argc, new_argv);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM peek_head_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    ERL_NIF_TERM result;
+
+    if(argc != 1 || !get_buffer_data(env, argv[0], &buffer)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    if(!enqueue_write_accumulator(buffer)) {
+        return enif_raise_exception(env, am_out_of_memory);
+    }
+
+    if(!enif_ioq_peek_head(env, buffer->queue, NULL, &result)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    return result;
+}
+
+static ERL_NIF_TERM skip_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    Uint64 block_size;
+
+    if(argc != 2 || !get_buffer_data(env, argv[0], &buffer)
+                 || !enif_get_uint64(env, argv[1], &block_size)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    if(!enqueue_write_accumulator(buffer)) {
+        return enif_raise_exception(env, am_out_of_memory);
+    } else if(enif_ioq_size(buffer->queue) < block_size) {
+        return enif_make_badarg(env);
+    }
+
+    enif_ioq_deq(buffer->queue, block_size, NULL);
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM find_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    int queue_length, queue_index;
+    SysIOVec *queued_data;
+    size_t queue_size;
+
+    size_t search_offset;
+    int needle;
+
+    if(argc != 2 || !get_buffer_data(env, argv[0], &buffer)
+                 || !enif_get_int(env, argv[1], &needle)) {
+        return enif_make_badarg(env);
+    }
+
+    ASSERT(erts_atomic32_inc_read_acqb(&buffer->concurrent_users) == 1);
+
+    if(!enqueue_write_accumulator(buffer)) {
+        return enif_raise_exception(env, am_out_of_memory);
+    } else if(needle < 0 || needle > 255) {
+        return enif_make_badarg(env);
+    }
+
+    queued_data = enif_ioq_peek(buffer->queue, &queue_length);
+    queue_size = enif_ioq_size(buffer->queue);
+    queue_index = 0;
+
+    search_offset = 0;
+
+    if(queue_size > (FIND_NIF_RESCHEDULE_SIZE / 100)) {
+        if(enif_thread_type() == ERL_NIF_THR_NORMAL_SCHEDULER) {
+            int timeslice_percent;
+
+            if(queue_size >= FIND_NIF_RESCHEDULE_SIZE) {
+                ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+                return enif_schedule_nif(env, "find",
+                    ERL_NIF_DIRTY_JOB_CPU_BOUND, &find_nif, argc, argv);
+            }
+
+            timeslice_percent = (queue_size * 100) / FIND_NIF_RESCHEDULE_SIZE;
+            enif_consume_timeslice(env, timeslice_percent);
+        }
+    }
+
+    while(queue_index < queue_length) {
+        char *needle_address;
+        char *block_start;
+        size_t block_size;
+
+        block_start = queued_data[queue_index].iov_base;
+        block_size = queued_data[queue_index].iov_len;
+
+        needle_address = memchr(block_start, needle, block_size);
+
+        if(needle_address != NULL) {
+            size_t result = search_offset + (needle_address - block_start);
+
+            ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+            return enif_make_tuple2(env, am_ok, enif_make_uint64(env, result));
+        }
+
+        search_offset += block_size;
+        queue_index++;
+    }
+
+    ASSERT(erts_atomic32_dec_read_relb(&buffer->concurrent_users) == 0);
+
+    return am_not_found;
+}
+
+/* */
+
+static ERL_NIF_TERM trylock_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    if(argc != 1 || !get_buffer_data(env, argv[0], &buffer)) {
+        return enif_make_badarg(env);
+    }
+
+    if(erts_atomic32_cmpxchg_acqb(&buffer->external_lock, 1, 0) == 0) {
+        return am_acquired;
+    }
+
+    return am_busy;
+}
+
+static ERL_NIF_TERM unlock_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    buffer_data_t *buffer;
+
+    if(argc != 1 || !get_buffer_data(env, argv[0], &buffer)) {
+        return enif_make_badarg(env);
+    }
+
+    if(erts_atomic32_cmpxchg_relb(&buffer->external_lock, 0, 1) == 0) {
+        return enif_raise_exception(env, am_lock_order_violation);
+    }
+
+    return am_ok;
+}
diff --git a/erts/emulator/nifs/common/prim_file_nif.c b/erts/emulator/nifs/common/prim_file_nif.c
new file mode 100644
index 0000000000..bbd9becb47
--- /dev/null
+++ b/erts/emulator/nifs/common/prim_file_nif.c
@@ -0,0 +1,1237 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#define STATIC_ERLANG_NIF 1
+
+#include "erl_nif.h"
+#include "config.h"
+#include "sys.h"
+
+#ifdef VALGRIND
+#  include <valgrind/memcheck.h>
+#endif
+
+#include "erl_driver.h"
+#include "prim_file_nif.h"
+
+/* NIF interface declarations */
+static int load(ErlNifEnv *env, void** priv_data, ERL_NIF_TERM load_info);
+static int upgrade(ErlNifEnv *env, void** priv_data, void** old_priv_data, ERL_NIF_TERM load_info);
+static void unload(ErlNifEnv *env, void* priv_data);
+
+static ErlNifResourceType *efile_resource_type;
+
+static ERL_NIF_TERM am_ok;
+static ERL_NIF_TERM am_error;
+static ERL_NIF_TERM am_continue;
+
+static ERL_NIF_TERM am_file_info;
+
+/* File modes */
+static ERL_NIF_TERM am_read;
+static ERL_NIF_TERM am_write;
+static ERL_NIF_TERM am_exclusive;
+static ERL_NIF_TERM am_append;
+static ERL_NIF_TERM am_sync;
+static ERL_NIF_TERM am_skip_type_check;
+
+/* enum efile_access_t; read and write are defined above.*/
+static ERL_NIF_TERM am_read_write;
+static ERL_NIF_TERM am_none;
+
+/* enum efile_advise_t */
+static ERL_NIF_TERM am_normal;
+static ERL_NIF_TERM am_random;
+static ERL_NIF_TERM am_sequential;
+static ERL_NIF_TERM am_will_need;
+static ERL_NIF_TERM am_dont_need;
+static ERL_NIF_TERM am_no_reuse;
+
+/* enum efile_filetype_t */
+static ERL_NIF_TERM am_device;
+static ERL_NIF_TERM am_directory;
+static ERL_NIF_TERM am_regular;
+static ERL_NIF_TERM am_symlink;
+static ERL_NIF_TERM am_other;
+
+/* enum efile_seek_t, 'eof' marker. */
+static ERL_NIF_TERM am_bof;
+static ERL_NIF_TERM am_cur;
+static ERL_NIF_TERM am_eof;
+
+static ERL_NIF_TERM read_info_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM set_permissions_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM set_owner_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM set_time_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM read_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM list_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM make_hard_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM make_soft_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM rename_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM make_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM del_file_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM del_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM get_device_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM get_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM set_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM read_file_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM get_handle_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM altname_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+static ERL_NIF_TERM open_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]);
+
+/* All file handle operations are passed through a wrapper that handles state
+ * transitions, marking it as busy during the course of the operation, and
+ * closing on completion if the owner died in the middle of an operation.
+ *
+ * This is pretty ugly but required as there's no way to tell when it's safe to
+ * asynchronously close a file; the event could have fired just before landing
+ * in a system call which will fail with EBADF at best or alias a newly opened
+ * fd at worst.
+ *
+ * The old driver got away with enqueueing the close operation on the same
+ * async queue as all of its other operations, but since dirty schedulers use a
+ * single global queue there's no natural way to schedule an asynchronous close
+ * "behind" other operations.
+ *
+ * The states may transition as follows:
+ *
+ * IDLE ->
+ *      BUSY (file_handle_wrapper) |
+ *      CLOSED (owner_death_callback)
+ *
+ * BUSY ->
+ *      IDLE (file_handle_wrapper)
+ *      CLOSED (close_nif_impl)
+ *      CLOSE_PENDING (owner_death_callback)
+ *
+ * CLOSE_PENDING ->
+ *      CLOSED (file_handle_wrapper)
+ */
+
+typedef ERL_NIF_TERM (*file_op_impl_t)(efile_data_t *d, ErlNifEnv *env,
+    int argc, const ERL_NIF_TERM argv[]);
+static ERL_NIF_TERM file_handle_wrapper(file_op_impl_t operation, ErlNifEnv *env,
+    int argc, const ERL_NIF_TERM argv[]);
+
+#define WRAP_FILE_HANDLE_EXPORT(name) \
+    static ERL_NIF_TERM name ## _impl (efile_data_t *d, ErlNifEnv *env, \
+        int argc, const ERL_NIF_TERM argv[]);\
+    static ERL_NIF_TERM name(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) { \
+        return file_handle_wrapper( name ## _impl , env, argc, argv); \
+    }
+
+WRAP_FILE_HANDLE_EXPORT(close_nif)
+WRAP_FILE_HANDLE_EXPORT(read_nif)
+WRAP_FILE_HANDLE_EXPORT(write_nif)
+WRAP_FILE_HANDLE_EXPORT(pread_nif)
+WRAP_FILE_HANDLE_EXPORT(pwrite_nif)
+WRAP_FILE_HANDLE_EXPORT(seek_nif)
+WRAP_FILE_HANDLE_EXPORT(sync_nif)
+WRAP_FILE_HANDLE_EXPORT(truncate_nif)
+WRAP_FILE_HANDLE_EXPORT(allocate_nif)
+WRAP_FILE_HANDLE_EXPORT(advise_nif)
+WRAP_FILE_HANDLE_EXPORT(get_handle_nif)
+WRAP_FILE_HANDLE_EXPORT(ipread_s32bu_p32bu_nif)
+
+static ErlNifFunc nif_funcs[] = {
+    /* File handle ops */
+    {"open_nif", 2, open_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"close_nif", 1, close_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"read_nif", 2, read_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"write_nif", 2, write_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"pread_nif", 3, pread_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"pwrite_nif", 3, pwrite_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"seek_nif", 3, seek_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"sync_nif", 2, sync_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"truncate_nif", 1, truncate_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"allocate_nif", 3, allocate_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"advise_nif", 4, advise_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+
+    /* Filesystem ops */
+    {"make_hard_link_nif", 2, make_hard_link_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"make_soft_link_nif", 2, make_soft_link_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"rename_nif", 2, rename_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"read_info_nif", 2, read_info_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"set_permissions_nif", 2, set_permissions_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"set_owner_nif", 3, set_owner_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"set_time_nif", 4, set_time_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"read_link_nif", 1, read_link_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"list_dir_nif", 1, list_dir_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"make_dir_nif", 1, make_dir_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"del_file_nif", 1, del_file_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"del_dir_nif", 1, del_dir_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"get_device_cwd_nif", 1, get_device_cwd_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"set_cwd_nif", 1, set_cwd_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"get_cwd_nif", 0, get_cwd_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+
+    /* These operations are equivalent to chained calls of other operations,
+     * but have been moved down to avoid excessive rescheduling. */
+    {"ipread_s32bu_p32bu_nif", 3, ipread_s32bu_p32bu_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+    {"read_file_nif", 1, read_file_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+
+    /* Internal ops. */
+    {"get_handle_nif", 1, get_handle_nif},
+    {"altname_nif", 1, altname_nif, ERL_NIF_DIRTY_JOB_IO_BOUND},
+};
+
+ERL_NIF_INIT(prim_file, nif_funcs, load, NULL, upgrade, unload)
+
+static void owner_death_callback(ErlNifEnv* env, void* obj, ErlNifPid* pid, ErlNifMonitor* mon);
+static void gc_callback(ErlNifEnv *env, void* data);
+
+static int load(ErlNifEnv *env, void** priv_data, ERL_NIF_TERM load_info)
+{
+    ErlNifResourceTypeInit callbacks;
+
+    am_ok = enif_make_atom(env, "ok");
+    am_error = enif_make_atom(env, "error");
+    am_continue = enif_make_atom(env, "continue");
+
+    am_read = enif_make_atom(env, "read");
+    am_write = enif_make_atom(env, "write");
+    am_exclusive = enif_make_atom(env, "exclusive");
+    am_append = enif_make_atom(env, "append");
+    am_sync = enif_make_atom(env, "sync");
+    am_skip_type_check = enif_make_atom(env, "skip_type_check");
+
+    am_read_write = enif_make_atom(env, "read_write");
+    am_none = enif_make_atom(env, "none");
+
+    am_normal = enif_make_atom(env, "normal");
+    am_random = enif_make_atom(env, "random");
+    am_sequential = enif_make_atom(env, "sequential");
+    am_will_need = enif_make_atom(env, "will_need");
+    am_dont_need = enif_make_atom(env, "dont_need");
+    am_no_reuse = enif_make_atom(env, "no_reuse");
+
+    am_device = enif_make_atom(env, "device");
+    am_directory = enif_make_atom(env, "directory");
+    am_regular = enif_make_atom(env, "regular");
+    am_symlink = enif_make_atom(env, "symlink");
+    am_other = enif_make_atom(env, "other");
+
+    am_file_info = enif_make_atom(env, "file_info");
+
+    am_bof = enif_make_atom(env, "bof");
+    am_cur = enif_make_atom(env, "cur");
+    am_eof = enif_make_atom(env, "eof");
+
+    callbacks.down = owner_death_callback;
+    callbacks.dtor = gc_callback;
+    callbacks.stop = NULL;
+
+    efile_resource_type = enif_open_resource_type_x(env, "efile", &callbacks,
+        ERL_NIF_RT_CREATE, NULL);
+
+    *priv_data = NULL;
+
+    return 0;
+}
+
+static void unload(ErlNifEnv *env, void* priv_data)
+{
+
+}
+
+static int upgrade(ErlNifEnv *env, void** priv_data, void** old_priv_data, ERL_NIF_TERM load_info)
+{
+    if(*old_priv_data != NULL) {
+        return -1; /* Don't know how to do that */
+    }
+    if(*priv_data != NULL) {
+        return -1; /* Don't know how to do that */
+    }
+    if(load(env, priv_data, load_info)) {
+        return -1;
+    }
+    return 0;
+}
+
+static ERL_NIF_TERM posix_error_to_tuple(ErlNifEnv *env, posix_errno_t posix_errno) {
+    ERL_NIF_TERM error = enif_make_atom(env, erl_errno_id(posix_errno));
+    return enif_make_tuple2(env, am_error, error);
+}
+
+static int get_file_data(ErlNifEnv *env, ERL_NIF_TERM opaque, efile_data_t **d) {
+    return enif_get_resource(env, opaque, efile_resource_type, (void **)d);
+}
+
+static ERL_NIF_TERM file_handle_wrapper(file_op_impl_t operation, ErlNifEnv *env,
+        int argc, const ERL_NIF_TERM argv[]) {
+
+    efile_data_t *d;
+
+    enum efile_state_t previous_state;
+    ERL_NIF_TERM result;
+
+    if(argc < 1 || !get_file_data(env, argv[0], &d)) {
+        return enif_make_badarg(env);
+    }
+
+    previous_state = erts_atomic32_cmpxchg_acqb(&d->state,
+        EFILE_STATE_BUSY, EFILE_STATE_IDLE);
+
+    if(previous_state == EFILE_STATE_IDLE) {
+        result = operation(d, env, argc - 1, &argv[1]);
+
+        previous_state = erts_atomic32_cmpxchg_relb(&d->state,
+            EFILE_STATE_IDLE, EFILE_STATE_BUSY);
+
+        ASSERT(previous_state != EFILE_STATE_IDLE);
+
+        if(previous_state == EFILE_STATE_CLOSE_PENDING) {
+            /* This is the only point where a change from CLOSE_PENDING is
+             * possible, and we're running synchronously, so we can't race with
+             * anything else here. */
+            erts_atomic32_set_acqb(&d->state, EFILE_STATE_CLOSED);
+            efile_close(d);
+        }
+    } else {
+        /* CLOSE_PENDING should be impossible at this point since it requires
+         * a transition from BUSY; the only valid state here is CLOSED. */
+        ASSERT(previous_state == EFILE_STATE_CLOSED);
+
+        result = posix_error_to_tuple(env, EINVAL);
+    }
+
+    return result;
+}
+
+static void owner_death_callback(ErlNifEnv* env, void* obj, ErlNifPid* pid, ErlNifMonitor* mon) {
+    efile_data_t *d = (efile_data_t*)obj;
+
+    (void)env;
+    (void)pid;
+    (void)mon;
+
+    for(;;) {
+        enum efile_state_t previous_state;
+
+        previous_state = erts_atomic32_cmpxchg_acqb(&d->state,
+            EFILE_STATE_CLOSED, EFILE_STATE_IDLE);
+
+        switch(previous_state) {
+        case EFILE_STATE_IDLE:
+            efile_close(d);
+            return;
+        case EFILE_STATE_CLOSE_PENDING:
+        case EFILE_STATE_CLOSED:
+            /* We're either already closed or managed to mark ourselves for
+             * closure in the previous iteration. */
+            return;
+        case EFILE_STATE_BUSY:
+            /* Schedule ourselves to be closed once the current operation
+             * finishes, retrying the [IDLE -> CLOSED] transition in case we
+             * narrowly passed the [BUSY -> IDLE] one. */
+            erts_atomic32_cmpxchg_nob(&d->state,
+                EFILE_STATE_CLOSE_PENDING, EFILE_STATE_BUSY);
+            break;
+        }
+    }
+}
+
+static void gc_callback(ErlNifEnv *env, void* data) {
+    efile_data_t *d = (efile_data_t*)data;
+
+    enum efile_state_t previous_state;
+
+    (void)env;
+
+    previous_state = erts_atomic32_cmpxchg_acqb(&d->state,
+        EFILE_STATE_CLOSED, EFILE_STATE_IDLE);
+
+    ASSERT(previous_state != EFILE_STATE_CLOSE_PENDING &&
+        previous_state != EFILE_STATE_BUSY);
+
+    if(previous_state == EFILE_STATE_IDLE) {
+        efile_close(d);
+    }
+}
+
+static ERL_NIF_TERM efile_filetype_to_atom(enum efile_filetype_t type) {
+    switch(type) {
+        case EFILE_FILETYPE_DEVICE: return am_device;
+        case EFILE_FILETYPE_DIRECTORY: return am_directory;
+        case EFILE_FILETYPE_REGULAR: return am_regular;
+        case EFILE_FILETYPE_SYMLINK: return am_symlink;
+        case EFILE_FILETYPE_OTHER: return am_other;
+    }
+
+    return am_other;
+}
+
+static ERL_NIF_TERM efile_access_to_atom(enum efile_access_t type) {
+    if(type & EFILE_ACCESS_READ && !(type & EFILE_ACCESS_WRITE)) {
+        return am_read;
+    } else if(type & EFILE_ACCESS_WRITE && !(type & EFILE_ACCESS_READ)) {
+        return am_write;
+    } else if(type & EFILE_ACCESS_READ_WRITE) {
+        return am_read_write;
+    }
+
+    return am_none;
+}
+
+static enum efile_modes_t efile_translate_modelist(ErlNifEnv *env, ERL_NIF_TERM list) {
+    enum efile_modes_t modes;
+    ERL_NIF_TERM head, tail;
+
+    modes = 0;
+
+    while(enif_get_list_cell(env, list, &head, &tail)) {
+        if(enif_is_identical(head, am_read)) {
+            modes |= EFILE_MODE_READ;
+        } else if(enif_is_identical(head, am_write)) {
+            modes |= EFILE_MODE_WRITE;
+        } else if(enif_is_identical(head, am_exclusive)) {
+            modes |= EFILE_MODE_EXCLUSIVE;
+        } else if(enif_is_identical(head, am_append)) {
+            modes |= EFILE_MODE_APPEND;
+        } else if(enif_is_identical(head, am_sync)) {
+            modes |= EFILE_MODE_SYNC;
+        } else if(enif_is_identical(head, am_skip_type_check)) {
+            modes |= EFILE_MODE_SKIP_TYPE_CHECK;
+        } else {
+            /* Modes like 'raw', 'ram', 'delayed_writes' etc are handled
+             * further up the chain. */
+        }
+
+        list = tail;
+    }
+
+    if(modes & (EFILE_MODE_APPEND | EFILE_MODE_EXCLUSIVE)) {
+        /* 'append' and 'exclusive' are documented as "open for writing." */
+        modes |= EFILE_MODE_WRITE;
+    } else if(!(modes & EFILE_MODE_READ_WRITE)) {
+        /* Defaulting to read if !(W|R) is undocumented, but specifically
+         * tested against in file_SUITE. */
+        modes |= EFILE_MODE_READ;
+    }
+
+    return modes;
+}
+
+static ERL_NIF_TERM open_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+    efile_data_t *d;
+
+    ErlNifPid controlling_process;
+    enum efile_modes_t modes;
+    ERL_NIF_TERM result;
+    efile_path_t path;
+
+    if(argc != 2 || !enif_is_list(env, argv[1])) {
+        return enif_make_badarg(env);
+    }
+
+    modes = efile_translate_modelist(env, argv[1]);
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_open(&path, modes, efile_resource_type, &d))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    result = enif_make_resource(env, d);
+    enif_release_resource(d);
+
+    enif_self(env, &controlling_process);
+
+    if(enif_monitor_process(env, d, &controlling_process, &d->monitor)) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
+
+static ERL_NIF_TERM close_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    enum efile_state_t previous_state;
+
+    if(argc != 0) {
+        return enif_make_badarg(env);
+    }
+
+    previous_state = erts_atomic32_cmpxchg_acqb(&d->state,
+        EFILE_STATE_CLOSED, EFILE_STATE_BUSY);
+
+    ASSERT(previous_state == EFILE_STATE_CLOSE_PENDING ||
+        previous_state == EFILE_STATE_BUSY);
+
+    if(previous_state == EFILE_STATE_BUSY) {
+        enif_demonitor_process(env, d, &d->monitor);
+
+        if(!efile_close(d)) {
+            return posix_error_to_tuple(env, d->posix_errno);
+        }
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM read_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    Sint64 bytes_read, block_size;
+    SysIOVec read_vec[1];
+    ErlNifBinary result;
+
+    if(argc != 1 || !enif_is_number(env, argv[0])) {
+        return enif_make_badarg(env);
+    }
+
+    if(!enif_get_int64(env, argv[0], &block_size) || block_size < 0) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    if(!enif_alloc_binary(block_size, &result)) {
+        return posix_error_to_tuple(env, ENOMEM);
+    }
+
+    read_vec[0].iov_base = result.data;
+    read_vec[0].iov_len = result.size;
+
+    bytes_read = efile_readv(d, read_vec, 1);
+    ASSERT(bytes_read <= block_size);
+
+    if(bytes_read < 0) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    } else if(bytes_read == 0) {
+        enif_release_binary(&result);
+        return am_eof;
+    }
+
+    if(bytes_read < block_size && !enif_realloc_binary(&result, bytes_read)) {
+        ERTS_INTERNAL_ERROR("Failed to shrink read result.");
+    }
+
+    return enif_make_tuple2(env, am_ok, enif_make_binary(env, &result));
+}
+
+static ERL_NIF_TERM write_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    ErlNifIOVec vec, *input = &vec;
+    Sint64 bytes_written;
+    ERL_NIF_TERM tail;
+
+    if(argc != 1 || !enif_inspect_iovec(env, 64, argv[0], &tail, &input)) {
+        return enif_make_badarg(env);
+    }
+
+    bytes_written = efile_writev(d, input->iov, input->iovcnt);
+
+    if(bytes_written < 0) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    if(!enif_is_empty_list(env, tail)) {
+        ASSERT(bytes_written > 0);
+        return enif_make_tuple2(env, am_continue, tail);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM pread_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    Sint64 bytes_read, block_size, offset;
+    SysIOVec read_vec[1];
+    ErlNifBinary result;
+
+    if(argc != 2 || !enif_is_number(env, argv[0])
+                 || !enif_is_number(env, argv[1])) {
+        return enif_make_badarg(env);
+    }
+
+    if(!enif_get_int64(env, argv[0], &offset) ||
+       !enif_get_int64(env, argv[1], &block_size) ||
+       (offset < 0 || block_size < 0)) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    if(!enif_alloc_binary(block_size, &result)) {
+        return posix_error_to_tuple(env, ENOMEM);
+    }
+
+    read_vec[0].iov_base = result.data;
+    read_vec[0].iov_len = result.size;
+
+    bytes_read = efile_preadv(d, offset, read_vec, 1);
+
+    if(bytes_read < 0) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    } else if(bytes_read == 0) {
+        enif_release_binary(&result);
+        return am_eof;
+    }
+
+    if(bytes_read < block_size && !enif_realloc_binary(&result, bytes_read)) {
+        ERTS_INTERNAL_ERROR("Failed to shrink pread result.");
+    }
+
+    return enif_make_tuple2(env, am_ok, enif_make_binary(env, &result));
+}
+
+static ERL_NIF_TERM pwrite_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    ErlNifIOVec vec, *input = &vec;
+    Sint64 bytes_written, offset;
+    ERL_NIF_TERM tail;
+
+    if(argc != 2 || !enif_is_number(env, argv[0])
+                 || !enif_inspect_iovec(env, 64, argv[1], &tail, &input)) {
+        return enif_make_badarg(env);
+    }
+
+    if(!enif_get_int64(env, argv[0], &offset) || offset < 0) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    bytes_written = efile_pwritev(d, offset, input->iov, input->iovcnt);
+
+    if(bytes_written < 0) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    if(!enif_is_empty_list(env, tail)) {
+        ASSERT(bytes_written > 0);
+        return enif_make_tuple3(env, am_continue,
+            enif_make_int64(env, bytes_written), tail);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM seek_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    Sint64 new_position, offset;
+    enum efile_seek_t seek;
+
+    if(argc != 2 || !enif_get_int64(env, argv[1], &offset)) {
+        return enif_make_badarg(env);
+    }
+
+    if(enif_is_identical(argv[0], am_bof)) {
+        seek = EFILE_SEEK_BOF;
+    } else if(enif_is_identical(argv[0], am_cur)) {
+        seek = EFILE_SEEK_CUR;
+    } else if(enif_is_identical(argv[0], am_eof)) {
+        seek = EFILE_SEEK_EOF;
+    } else {
+        return enif_make_badarg(env);
+    }
+
+    if(!efile_seek(d, seek, offset, &new_position)) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, enif_make_uint64(env, new_position));
+}
+
+static ERL_NIF_TERM sync_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    int data_only;
+
+    if(argc != 1 || !enif_get_int(env, argv[0], &data_only)) {
+        return enif_make_badarg(env);
+    }
+
+    if(!efile_sync(d, data_only)) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM truncate_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    if(argc != 0) {
+        return enif_make_badarg(env);
+    }
+
+    if(!efile_truncate(d)) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM allocate_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    Sint64 offset, length;
+
+    if(argc != 2 || !enif_is_number(env, argv[0])
+                 || !enif_is_number(env, argv[1])) {
+        return enif_make_badarg(env);
+    }
+
+    if(!enif_get_int64(env, argv[0], &offset) ||
+       !enif_get_int64(env, argv[1], &length) ||
+       (offset < 0 || length < 0)) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    if(!efile_allocate(d, offset, length)) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM advise_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    enum efile_advise_t advise;
+    Sint64 offset, length;
+
+    if(argc != 3 || !enif_is_number(env, argv[0])
+                 || !enif_is_number(env, argv[1])) {
+        return enif_make_badarg(env);
+    }
+
+    if(!enif_get_int64(env, argv[0], &offset) ||
+       !enif_get_int64(env, argv[1], &length) ||
+       (offset < 0 || length < 0)) {
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    if(enif_is_identical(argv[2], am_normal)) {
+        advise = EFILE_ADVISE_NORMAL;
+    } else if(enif_is_identical(argv[2], am_random)) {
+        advise = EFILE_ADVISE_RANDOM;
+    } else if(enif_is_identical(argv[2], am_sequential)) {
+        advise = EFILE_ADVISE_SEQUENTIAL;
+    } else if(enif_is_identical(argv[2], am_will_need)) {
+        advise = EFILE_ADVISE_WILL_NEED;
+    } else if(enif_is_identical(argv[2], am_dont_need)) {
+        advise = EFILE_ADVISE_DONT_NEED;
+    } else if(enif_is_identical(argv[2], am_no_reuse)) {
+        advise = EFILE_ADVISE_NO_REUSE;
+    } else {
+        /* The tests check for EINVAL instead of badarg. Sigh. */
+        return posix_error_to_tuple(env, EINVAL);
+    }
+
+    if(!efile_advise(d, offset, length, advise)) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    }
+
+    return am_ok;
+}
+
+/* This undocumented function reads a pointer and then reads the data block
+ * described by said pointer. It was reverse-engineered from the old
+ * implementation so while all tests pass it may not be entirely correct. Our
+ * current understanding is as follows:
+ *
+ * Pointer layout:
+ *
+ *     <<Size:1/integer-unit:32, Offset:1/integer-unit:32>>
+ *
+ * Where Offset is the -absolute- address to the data block.
+ *
+ * *) If we fail to read the pointer block in its entirety, we return eof.
+ * *) If the provided max_payload_size is larger than Size, we return eof.
+ * *) If we fail to read any data whatsoever at Offset, we return
+ *    {ok, {Size, Offset, eof}}
+ * *) Otherwise, we return {ok, {Size, Offset, Data}}. Note that the size
+ *    of Data may be smaller than Size if we encounter EOF before we could
+ *    read the entire block.
+ *
+ * On errors we'll return {error, posix()} regardless of whether they
+ * happened before or after reading the pointer block. */
+static ERL_NIF_TERM ipread_s32bu_p32bu_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    Sint64 payload_offset, payload_size;
+
+    SysIOVec read_vec[1];
+    Sint64 bytes_read;
+
+    ErlNifBinary payload;
+
+    if(argc != 2 || !enif_is_number(env, argv[0])
+                 || !enif_is_number(env, argv[1])) {
+        return enif_make_badarg(env);
+    }
+
+    {
+        Sint64 max_payload_size, pointer_offset;
+        unsigned char pointer_block[8];
+
+        if(!enif_get_int64(env, argv[0], &pointer_offset) ||
+           !enif_get_int64(env, argv[1], &max_payload_size) ||
+           (pointer_offset < 0 || max_payload_size >= 1u << 31)) {
+            return posix_error_to_tuple(env, EINVAL);
+        }
+
+        read_vec[0].iov_base = pointer_block;
+        read_vec[0].iov_len = sizeof(pointer_block);
+
+        bytes_read = efile_preadv(d, pointer_offset, read_vec, 1);
+
+        if(bytes_read < 0) {
+            return posix_error_to_tuple(env, d->posix_errno);
+        } else if(bytes_read < sizeof(pointer_block)) {
+            return am_eof;
+        }
+
+        payload_size = (Uint32)get_int32(&pointer_block[0]);
+        payload_offset = (Uint32)get_int32(&pointer_block[4]);
+
+        if(payload_size > max_payload_size) {
+            return am_eof;
+        }
+    }
+
+    if(!enif_alloc_binary(payload_size, &payload)) {
+        return posix_error_to_tuple(env, ENOMEM);
+    }
+
+    read_vec[0].iov_base = payload.data;
+    read_vec[0].iov_len = payload.size;
+
+    bytes_read = efile_preadv(d, payload_offset, read_vec, 1);
+
+    if(bytes_read < 0) {
+        return posix_error_to_tuple(env, d->posix_errno);
+    } else if(bytes_read == 0) {
+        enif_release_binary(&payload);
+
+        return enif_make_tuple2(env, am_ok,
+                enif_make_tuple3(env,
+                    enif_make_uint(env, payload_size),
+                    enif_make_uint(env, payload_offset),
+                    am_eof));
+    }
+
+    if(bytes_read < payload.size && !enif_realloc_binary(&payload, bytes_read)) {
+        ERTS_INTERNAL_ERROR("Failed to shrink ipread payload.");
+    }
+
+    return enif_make_tuple2(env, am_ok,
+        enif_make_tuple3(env,
+            enif_make_uint(env, payload_size),
+            enif_make_uint(env, payload_offset),
+            enif_make_binary(env, &payload)));
+}
+
+static ERL_NIF_TERM get_handle_nif_impl(efile_data_t *d, ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    if(argc != 0) {
+        return enif_make_badarg(env);
+    }
+
+    return efile_get_handle(env, d);
+}
+
+static ERL_NIF_TERM read_info_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_fileinfo_t info = {0};
+    efile_path_t path;
+    int follow_links;
+
+    if(argc != 2 || !enif_get_int(env, argv[1], &follow_links)) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_read_info(&path, follow_links, &info))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    /* #file_info as declared in file.hrl */
+    return enif_make_tuple(env, 14,
+        am_file_info,
+        enif_make_uint64(env, info.size),
+        efile_filetype_to_atom(info.type),
+        efile_access_to_atom(info.access),
+        enif_make_int64(env, MAX(EFILE_MIN_FILETIME, info.a_time)),
+        enif_make_int64(env, MAX(EFILE_MIN_FILETIME, info.m_time)),
+        enif_make_int64(env, MAX(EFILE_MIN_FILETIME, info.c_time)),
+        enif_make_uint(env, info.mode),
+        enif_make_uint(env, info.links),
+        enif_make_uint(env, info.major_device),
+        enif_make_uint(env, info.minor_device),
+        enif_make_uint(env, info.inode),
+        enif_make_uint(env, info.uid),
+        enif_make_uint(env, info.gid)
+    );
+}
+
+static ERL_NIF_TERM set_permissions_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+    Uint32 permissions;
+
+    if(argc != 2 || !enif_get_uint(env, argv[1], &permissions)) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_set_permissions(&path, permissions))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM set_owner_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+    Sint32 uid, gid;
+
+    if(argc != 3 || !enif_get_int(env, argv[1], &uid)
+                 || !enif_get_int(env, argv[2], &gid)) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_set_owner(&path, uid, gid))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM set_time_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    Sint64 accessed, modified, created;
+    efile_path_t path;
+
+    if(argc != 4 || !enif_get_int64(env, argv[1], &accessed)
+                 || !enif_get_int64(env, argv[2], &modified)
+                 || !enif_get_int64(env, argv[3], &created)) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_set_time(&path, accessed, modified, created))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM read_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+    ERL_NIF_TERM result;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_read_link(env, &path, &result))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
+
+static ERL_NIF_TERM list_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+    ERL_NIF_TERM result;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_list_dir(env, &path, &result))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
+
+static ERL_NIF_TERM rename_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t existing_path, new_path;
+
+    if(argc != 2) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &existing_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_marshal_path(env, argv[1], &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_rename(&existing_path, &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM make_hard_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t existing_path, new_path;
+
+    if(argc != 2) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &existing_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_marshal_path(env, argv[1], &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_make_hard_link(&existing_path, &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM make_soft_link_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t existing_path, new_path;
+
+    if(argc != 2) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &existing_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_marshal_path(env, argv[1], &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_make_soft_link(&existing_path, &new_path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM make_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_make_dir(&path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM del_file_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_del_file(&path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM del_dir_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_del_dir(&path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+static ERL_NIF_TERM get_device_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    ERL_NIF_TERM result;
+    int device_index;
+
+    if(argc != 1 || !enif_get_int(env, argv[0], &device_index)) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_get_device_cwd(env, device_index, &result))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
+
+static ERL_NIF_TERM get_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+    ERL_NIF_TERM result;
+
+    if(argc != 0) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_get_cwd(env, &result))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
+
+static ERL_NIF_TERM set_cwd_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_set_cwd(&path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return am_ok;
+}
+
+/** @brief Reads an entire file into \c result, stopping after \c size bytes or
+ * EOF. It will read until EOF if size is 0. */
+static posix_errno_t read_file(efile_data_t *d, size_t size, ErlNifBinary *result) {
+    size_t initial_buffer_size;
+    ssize_t bytes_read;
+
+    if(size == 0) {
+        initial_buffer_size = 16 << 10;
+    } else {
+        initial_buffer_size = size;
+    }
+
+    if(!enif_alloc_binary(initial_buffer_size, result)) {
+        return ENOMEM;
+    }
+
+    bytes_read = 0;
+
+    for(;;) {
+        ssize_t block_bytes_read;
+        SysIOVec read_vec[1];
+
+        read_vec[0].iov_base = result->data + bytes_read;
+        read_vec[0].iov_len = result->size - bytes_read;
+
+        block_bytes_read = efile_readv(d, read_vec, 1);
+
+        if(block_bytes_read < 0) {
+            enif_release_binary(result);
+            return d->posix_errno;
+        }
+
+        bytes_read += block_bytes_read;
+
+        if(block_bytes_read < (result->size - bytes_read)) {
+            /* EOF */
+            break;
+        } else if(bytes_read == size) {
+            break;
+        }
+
+        if(!enif_realloc_binary(result, bytes_read * 2)) {
+            enif_release_binary(result);
+            return ENOMEM;
+        }
+    }
+
+    /* The file may have shrunk since we queried its size, so we have to do
+     * this even when the size is known. */
+    if(bytes_read < result->size && !enif_realloc_binary(result, bytes_read)) {
+        ERTS_INTERNAL_ERROR("Failed to shrink read_file result.");
+    }
+
+    return 0;
+}
+
+static ERL_NIF_TERM read_file_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_fileinfo_t info = {0};
+    efile_path_t path;
+    efile_data_t *d;
+
+    ErlNifBinary result;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_read_info(&path, 1, &info))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_open(&path, EFILE_MODE_READ, efile_resource_type, &d))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    posix_errno = read_file(d, info.size, &result);
+    enif_release_resource(d);
+
+    if(posix_errno) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, enif_make_binary(env, &result));
+}
+
+static ERL_NIF_TERM altname_nif(ErlNifEnv *env, int argc, const ERL_NIF_TERM argv[]) {
+    posix_errno_t posix_errno;
+
+    efile_path_t path;
+    ERL_NIF_TERM result;
+
+    if(argc != 1) {
+        return enif_make_badarg(env);
+    }
+
+    if((posix_errno = efile_marshal_path(env, argv[0], &path))) {
+        return posix_error_to_tuple(env, posix_errno);
+    } else if((posix_errno = efile_altname(env, &path, &result))) {
+        return posix_error_to_tuple(env, posix_errno);
+    }
+
+    return enif_make_tuple2(env, am_ok, result);
+}
diff --git a/erts/emulator/nifs/common/prim_file_nif.h b/erts/emulator/nifs/common/prim_file_nif.h
new file mode 100644
index 0000000000..4194cdc7d9
--- /dev/null
+++ b/erts/emulator/nifs/common/prim_file_nif.h
@@ -0,0 +1,240 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+typedef int posix_errno_t;
+
+enum efile_modes_t {
+    EFILE_MODE_READ = (1 << 0),
+    EFILE_MODE_WRITE = (1 << 1), /* Implies truncating file when used alone. */
+    EFILE_MODE_APPEND = (1 << 2),
+    EFILE_MODE_EXCLUSIVE = (1 << 3),
+    EFILE_MODE_SYNC = (1 << 4),
+
+    EFILE_MODE_SKIP_TYPE_CHECK = (1 << 5), /* Special for device files on Unix. */
+    EFILE_MODE_NO_TRUNCATE = (1 << 6), /* Special for reopening on VxWorks. */
+
+    EFILE_MODE_READ_WRITE = EFILE_MODE_READ | EFILE_MODE_WRITE
+};
+
+enum efile_access_t {
+    EFILE_ACCESS_NONE = 0,
+    EFILE_ACCESS_READ = 1,
+    EFILE_ACCESS_WRITE = 2,
+    EFILE_ACCESS_READ_WRITE = EFILE_ACCESS_READ | EFILE_ACCESS_WRITE
+};
+
+enum efile_seek_t {
+    EFILE_SEEK_BOF,
+    EFILE_SEEK_CUR,
+    EFILE_SEEK_EOF
+};
+
+enum efile_filetype_t {
+    EFILE_FILETYPE_DEVICE,
+    EFILE_FILETYPE_DIRECTORY,
+    EFILE_FILETYPE_REGULAR,
+    EFILE_FILETYPE_SYMLINK,
+    EFILE_FILETYPE_OTHER
+};
+
+enum efile_advise_t {
+    EFILE_ADVISE_NORMAL,
+    EFILE_ADVISE_RANDOM,
+    EFILE_ADVISE_SEQUENTIAL,
+    EFILE_ADVISE_WILL_NEED,
+    EFILE_ADVISE_DONT_NEED,
+    EFILE_ADVISE_NO_REUSE
+};
+
+enum efile_state_t {
+    EFILE_STATE_IDLE = 0,
+    EFILE_STATE_BUSY = 1,
+    EFILE_STATE_CLOSE_PENDING = 2,
+    EFILE_STATE_CLOSED = 3
+};
+
+typedef struct {
+    Sint64 size;            /* Size of file */
+    Uint32 type;            /* Type of file -- one of EFILE_FILETYPE_*. */
+    Uint32 access;          /* Access to file -- one of EFILE_ACCESS_*. */
+    Uint32 mode;            /* Access permissions -- bit field. */
+    Uint32 links;           /* Number of links to file. */
+    Uint32 major_device;    /* Major device or file system. */
+    Uint32 minor_device;    /* Minor device (for devices). */
+    Uint32 inode;           /* Inode number. */
+    Uint32 uid;             /* User id of owner. */
+    Uint32 gid;             /* Group id of owner. */
+    Sint64 a_time;          /* Last time the file was accessed. */
+    Sint64 m_time;          /* Last time the file was modified. */
+    Sint64 c_time;          /* Windows: creation time, Unix: last inode
+                             * change. */
+} efile_fileinfo_t;
+
+/* The smallest value that can be converted freely between universal, local,
+ * and POSIX time, as required by read_file_info/2. Corresponds to
+ * {{1902,1,1},{0,0,0}} */
+#define EFILE_MIN_FILETIME -2145916800
+
+/* Initializes an efile_data_t; must be used in efile_open on success. */
+#define EFILE_INIT_RESOURCE(__d, __modes) do { \
+        erts_atomic32_init_acqb(&(__d)->state, EFILE_STATE_IDLE); \
+        (__d)->posix_errno = 0; \
+        (__d)->modes = __modes; \
+    } while(0)
+
+typedef struct {
+    erts_atomic32_t state;
+
+    posix_errno_t posix_errno;
+    enum efile_modes_t modes;
+
+    ErlNifMonitor monitor;
+} efile_data_t;
+
+typedef ErlNifBinary efile_path_t;
+
+/* @brief Translates the given "raw name" into the format expected by the APIs
+ * used by the underlying implementation. The result is transient and does not
+ * need to be released.
+ *
+ * This may change the structure of the path and its results should never be
+ * passed on to the user. Refer to the OS-specific implementation for details.
+ *
+ * @param path The term to translate; it must have been encoded with
+ * prim_file:internal_native2name for compatibility reasons. */
+posix_errno_t efile_marshal_path(ErlNifEnv *env, ERL_NIF_TERM path, efile_path_t *result);
+
+/* @brief Returns the underlying handle as an implementation-defined term.
+ *
+ * This is an internal function intended to support tests and tricky
+ * operations like sendfile(2). */
+ERL_NIF_TERM efile_get_handle(ErlNifEnv *env, efile_data_t *d);
+
+/* @brief Read until EOF or the given iovec has been filled.
+ *
+ * @return -1 on failure, or the number of bytes read on success. The return
+ * value will be 0 if no bytes could be read before EOF or the end of the
+ * iovec. */
+Sint64 efile_readv(efile_data_t *d, SysIOVec *iov, int iovlen);
+
+/* @brief Write the entirety of the given iovec.
+ *
+ * @return -1 on failure, or the number of bytes written on success. "Partial"
+ * failures will be reported with -1 and not the number of bytes we managed to
+ * write to disk before the failure. */
+Sint64 efile_writev(efile_data_t *d, SysIOVec *iov, int iovlen);
+
+/* @brief As \c efile_readv, but starting from a file offset. */
+Sint64 efile_preadv(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen);
+
+/* @brief As \c efile_writev, but starting from a file offset. */
+Sint64 efile_pwritev(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen);
+
+int efile_seek(efile_data_t *d, enum efile_seek_t seek, Sint64 offset, Sint64 *new_position);
+
+int efile_sync(efile_data_t *d, int data_only);
+
+int efile_advise(efile_data_t *d, Sint64 offset, Sint64 length, enum efile_advise_t advise);
+int efile_allocate(efile_data_t *d, Sint64 offset, Sint64 length);
+int efile_truncate(efile_data_t *d);
+
+posix_errno_t efile_open(const efile_path_t *path, enum efile_modes_t modes,
+        ErlNifResourceType *nif_type, efile_data_t **d);
+
+/** @brief Closes a file. The file must have entered the CLOSED state prior to
+ * calling this to prevent double close. */
+int efile_close(efile_data_t *d);
+
+/* **** **** **** **** **** **** **** **** **** **** **** **** **** **** **** */
+
+posix_errno_t efile_read_info(const efile_path_t *path, int follow_link, efile_fileinfo_t *result);
+
+/** @brief Sets the file times to the given values. Refer to efile_fileinfo_t
+ * for a description of each. */
+posix_errno_t efile_set_time(const efile_path_t *path, Sint64 a_time, Sint64 m_time, Sint64 c_time);
+
+/** @brief On Unix, this sets the file permissions according to the docs for
+ * file:write_file_info/2. On Windows it uses the "owner write permission" flag
+ * to toggle whether the file is read-only or not. */
+posix_errno_t efile_set_permissions(const efile_path_t *path, Uint32 permissions);
+
+/** @brief On Unix, this will set the owner/group to the given values. It will
+ * do nothing on other platforms. */
+posix_errno_t efile_set_owner(const efile_path_t *path, Sint32 owner, Sint32 group);
+
+/** @brief Resolves the final path of the given link. */
+posix_errno_t efile_read_link(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result);
+
+/** @brief Lists the contents of the given directory.
+ * @param result [out] A list of all the directory/file names contained in the
+ * given directory. */
+posix_errno_t efile_list_dir(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result);
+
+/** @brief Changes the name of an existing file or directory, from old_path
+ * to new_path.
+ *
+ * If old_path and new_path refer to the same file or directory, it does
+ * nothing and returns success. Otherwise if new_path already exists, it will
+ * be deleted and replaced by src subject to the following conditions:
+ *
+ *     If old_path is a directory, new_path may be an empty directory.
+ *     If old_path is a file, new_path may be a file.
+ *
+ * Neither of these are guaranteed to be atomic. In any other situation where
+ * new_path already exists, the rename will fail.
+ *
+ * Some possible error codes:
+ *
+ * - EACCES:     Either paths or one of their parent directories can't be read
+ *               and/or written.
+ * - EEXIST:     new_path is a non-empty directory.
+ * - EINVAL:     old_path is a root directory or new_path is a subdirectory
+ *               of new_path.
+ * - EISDIR:     new_path is a directory, but old_path is not.
+ * - ENOTDIR:    old_path is a directory, but new_path is not.
+ * - ENOENT:     old_path doesn't exist, or either path is "".
+ * - EXDEV:      The paths are on different filesystems.
+ *
+ * The implementation of rename may allow cross-filesystem renames,
+ * but the caller should be prepared to emulate it with copy and
+ * delete if errno is EXDEV. */
+posix_errno_t efile_rename(const efile_path_t *old_path, const efile_path_t *new_path);
+
+posix_errno_t efile_make_hard_link(const efile_path_t *existing_path, const efile_path_t *new_path);
+posix_errno_t efile_make_soft_link(const efile_path_t *existing_path, const efile_path_t *new_path);
+posix_errno_t efile_make_dir(const efile_path_t *path);
+
+posix_errno_t efile_del_file(const efile_path_t *path);
+posix_errno_t efile_del_dir(const efile_path_t *path);
+
+posix_errno_t efile_get_cwd(ErlNifEnv *env, ERL_NIF_TERM *result);
+posix_errno_t efile_set_cwd(const efile_path_t *path);
+
+/** @brief A Windows-specific function for returning the working directory of a
+ * given device.
+ *
+ * @param device_index The drive index; 1 for A, 2 for B, etc.
+ * @param result [out] The working directory of the given device
+ */
+posix_errno_t efile_get_device_cwd(ErlNifEnv *env, int device_index, ERL_NIF_TERM *result);
+
+/** @brief A Windows-specific function for returning the 8.3-name of a given
+ * file or directory. */
+posix_errno_t efile_altname(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result);
diff --git a/erts/emulator/nifs/unix/unix_prim_file.c b/erts/emulator/nifs/unix/unix_prim_file.c
new file mode 100644
index 0000000000..2b112dda76
--- /dev/null
+++ b/erts/emulator/nifs/unix/unix_prim_file.c
@@ -0,0 +1,944 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#include "erl_nif.h"
+#include "config.h"
+#include "sys.h"
+
+#ifdef VALGRIND
+#  include <valgrind/memcheck.h>
+#endif
+
+#include "prim_file_nif.h"
+
+#if defined(__APPLE__) && defined(__MACH__) && !defined(__DARWIN__)
+#define __DARWIN__ 1
+#endif
+
+#if defined(__DARWIN__) || defined(HAVE_LINUX_FALLOC_H) || defined(HAVE_POSIX_FALLOCATE)
+#include <fcntl.h>
+#endif
+
+#ifdef HAVE_LINUX_FALLOC_H
+#include <linux/falloc.h>
+#endif
+
+#include <utime.h>
+
+/* Macros for testing file types. */
+#ifdef NO_UMASK
+#define FILE_MODE 0644
+#define DIR_MODE  0755
+#else
+#define FILE_MODE 0666
+#define DIR_MODE  0777
+#endif
+
+/* Old platforms might not have IOV_MAX defined. */
+#if !defined(IOV_MAX) && defined(UIO_MAXIOV)
+#define IOV_MAX UIO_MAXIOV
+#elif !defined(IOV_MAX)
+#define IOV_MAX 16
+#endif
+
+typedef struct {
+    efile_data_t common;
+    int fd;
+} efile_unix_t;
+
+static int has_invalid_null_termination(const ErlNifBinary *path) {
+    const char *null_pos, *end_pos;
+
+    null_pos = memchr(path->data, '\0', path->size);
+    end_pos = (const char*)&path->data[path->size] - 1;
+
+    if(null_pos == NULL) {
+        return 1;
+    }
+
+    /* prim_file:internal_name2native sometimes feeds us data that is "doubly"
+     * NUL-terminated, so we'll accept any number of trailing NULs so long as
+     * they aren't interrupted by anything else. */
+    while(null_pos < end_pos && (*null_pos) == '\0') {
+        null_pos++;
+    }
+
+    return null_pos != end_pos;
+}
+
+posix_errno_t efile_marshal_path(ErlNifEnv *env, ERL_NIF_TERM path, efile_path_t *result) {
+    if(!enif_inspect_binary(env, path, result)) {
+        return EINVAL;
+    }
+
+    if(has_invalid_null_termination(result)) {
+        return EINVAL;
+    }
+
+    return 0;
+}
+
+ERL_NIF_TERM efile_get_handle(ErlNifEnv *env, efile_data_t *d) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+    ERL_NIF_TERM result;
+    unsigned char *bits;
+
+    bits = enif_make_new_binary(env, sizeof(u->fd), &result);
+    memcpy(bits, &u->fd, sizeof(u->fd));
+
+    return result;
+}
+
+static int open_file_type_check(const efile_path_t *path, int fd) {
+    struct stat file_info;
+    int error;
+
+#ifndef HAVE_FSTAT
+    error = stat((const char*)path->data, &file_info);
+    (void)fd;
+#else
+    error = fstat(fd, &file_info);
+    (void)path;
+#endif
+
+    if(error < 0) {
+        /* If we failed to stat assume success and let the next call handle the
+         * error. The old driver checked whether the file was to be used
+         * immediately in a read within the call, but the new implementation
+         * never does that. */
+         return 1;
+    }
+
+    /* Allow everything that isn't a directory, and error out on the next call
+     * if it's unsupported. */
+    if(S_ISDIR(file_info.st_mode)) {
+        return 0;
+    }
+
+    return 1;
+}
+
+posix_errno_t efile_open(const efile_path_t *path, enum efile_modes_t modes,
+        ErlNifResourceType *nif_type, efile_data_t **d) {
+
+    int flags, fd;
+
+    flags = 0;
+
+    if(modes & EFILE_MODE_READ && !(modes & EFILE_MODE_WRITE)) {
+        flags |= O_RDONLY;
+    } else if(modes & EFILE_MODE_WRITE && !(modes & EFILE_MODE_READ)) {
+        if(!(modes & EFILE_MODE_NO_TRUNCATE)) {
+            flags |= O_TRUNC;
+        }
+
+        flags |= O_WRONLY | O_CREAT;
+    } else if(modes & EFILE_MODE_READ_WRITE) {
+        flags |= O_RDWR | O_CREAT;
+    } else {
+        return EINVAL;
+    }
+
+    if(modes & EFILE_MODE_APPEND) {
+        flags &= ~O_TRUNC;
+        flags |= O_APPEND;
+    }
+
+    if(modes & EFILE_MODE_EXCLUSIVE) {
+        flags |= O_EXCL;
+    }
+
+    if(modes & EFILE_MODE_SYNC) {
+#ifndef O_SYNC
+        return ENOTSUP;
+#else
+        flags |= O_SYNC;
+#endif
+    }
+
+    do {
+        fd = open((const char*)path->data, flags, FILE_MODE);
+    } while(fd == -1 && errno == EINTR);
+
+    if(fd != -1) {
+        efile_unix_t *u;
+
+        if(!(modes & EFILE_MODE_SKIP_TYPE_CHECK) && !open_file_type_check(path, fd)) {
+            close(fd);
+
+            /* This is blatantly incorrect, but we're documented as returning
+             * this for everything that isn't a file. */
+            return EISDIR;
+        }
+
+        u = (efile_unix_t*)enif_alloc_resource(nif_type, sizeof(efile_unix_t));
+        u->fd = fd;
+
+        EFILE_INIT_RESOURCE(&u->common, modes);
+        (*d) = &u->common;
+
+        return 0;
+    }
+
+    (*d) = NULL;
+    return errno;
+}
+
+int efile_close(efile_data_t *d) {
+    efile_unix_t *u = (efile_unix_t*)d;
+    int fd;
+
+    ASSERT(erts_atomic32_read_nob(&d->state) == EFILE_STATE_CLOSED);
+    ASSERT(u->fd != -1);
+
+    fd = u->fd;
+    u->fd = -1;
+
+    /* close(2) either always closes (*BSD, Linux) or leaves the fd in an
+     * undefined state (POSIX 2008, Solaris), so we must not retry on EINTR. */
+
+    if(close(fd) < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    return 1;
+}
+
+static void shift_iov(SysIOVec **iov, int *iovlen, ssize_t shift) {
+    SysIOVec *head_vec = (*iov);
+
+    ASSERT(shift >= 0);
+
+    while(shift > 0) {
+        ASSERT(head_vec < &(*iov)[*iovlen]);
+
+        if(shift < head_vec->iov_len) {
+            head_vec->iov_base = (char*)head_vec->iov_base + shift;
+            head_vec->iov_len -= shift;
+            break;
+        } else {
+            shift -= head_vec->iov_len;
+            head_vec++;
+        }
+    }
+
+    (*iovlen) -= head_vec - (*iov);
+    (*iov) = head_vec;
+}
+
+Sint64 efile_readv(efile_data_t *d, SysIOVec *iov, int iovlen) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+    Sint64 bytes_read;
+    ssize_t result;
+
+    bytes_read = 0;
+
+    do {
+        int use_fallback = 0;
+
+        if(iovlen < 1) {
+            result = 0;
+            break;
+        }
+
+        /* writev(2) implies readv(2) */
+#ifdef HAVE_WRITEV
+        result = readv(u->fd, iov, MIN(IOV_MAX, iovlen));
+
+        /* Fall back to using read(2) if readv(2) reports that the combined
+         * size of iov is greater than SSIZE_T_MAX. */
+        use_fallback = (result < 0 && errno == EINVAL);
+#else
+        use_fallback = 1;
+#endif
+
+        if(use_fallback) {
+            result = read(u->fd, iov->iov_base, iov->iov_len);
+        }
+
+        if(result > 0) {
+            shift_iov(&iov, &iovlen, result);
+            bytes_read += result;
+        }
+    } while(result > 0 || (result < 0 && errno == EINTR));
+
+    u->common.posix_errno = errno;
+
+    if(result == 0 && bytes_read > 0) {
+        return bytes_read;
+    }
+
+    return result;
+}
+
+Sint64 efile_writev(efile_data_t *d, SysIOVec *iov, int iovlen) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+    Sint64 bytes_written;
+    ssize_t result;
+
+    bytes_written = 0;
+
+    do {
+        int use_fallback = 0;
+
+        if(iovlen < 1) {
+            result = 0;
+            break;
+        }
+
+#ifdef HAVE_WRITEV
+        result = writev(u->fd, iov, MIN(IOV_MAX, iovlen));
+
+        /* Fall back to using write(2) if writev(2) reports that the combined
+         * size of iov is greater than SSIZE_T_MAX. */
+        use_fallback = (result < 0 && errno == EINVAL);
+#else
+        use_fallback = 1;
+#endif
+
+        if(use_fallback) {
+            result = write(u->fd, iov->iov_base, iov->iov_len);
+        }
+
+        if(result > 0) {
+            shift_iov(&iov, &iovlen, result);
+            bytes_written += result;
+        }
+    } while(result > 0 || (result < 0 && errno == EINTR));
+
+    u->common.posix_errno = errno;
+
+    if(result == 0 && bytes_written > 0) {
+        return bytes_written;
+    }
+
+    return result;
+}
+
+Sint64 efile_preadv(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+    Uint64 bytes_read;
+    Sint64 result;
+
+#if !defined(HAVE_PREADV) && !defined(HAVE_PREAD)
+    /* This function is documented as leaving the file position undefined, but
+     * the old driver always reset it so there's probably code in the wild that
+     * relies on this behavior. */
+    off_t original_position = lseek(u->fd, 0, SEEK_CUR);
+
+    if(original_position < 0 || lseek(u->fd, offset, SEEK_SET) < 0) {
+        u->common.posix_errno = errno;
+        return -1;
+    }
+#endif
+
+    bytes_read = 0;
+
+    do {
+        if(iovlen < 1) {
+            result = 0;
+            break;
+        }
+
+#if defined(HAVE_PREADV)
+        result = preadv(u->fd, iov, MIN(IOV_MAX, iovlen), offset);
+#elif defined(HAVE_PREAD)
+        result = pread(u->fd, iov->iov_base, iov->iov_len, offset);
+#else
+        result = read(u->fd, iov->iov_base, iov->iov_len);
+#endif
+
+        if(result > 0) {
+            shift_iov(&iov, &iovlen, result);
+            bytes_read += result;
+            offset += result;
+        }
+    } while(result > 0 || (result < 0 && errno == EINTR));
+
+    u->common.posix_errno = errno;
+
+#if !defined(HAVE_PREADV) && !defined(HAVE_PREAD)
+    if(result >= 0) {
+        if(lseek(u->fd, original_position, SEEK_SET) < 0) {
+            u->common.posix_errno = errno;
+            return -1;
+        }
+    }
+#endif
+
+    if(result == 0 && bytes_read > 0) {
+        return bytes_read;
+    }
+
+    return result;
+}
+
+Sint64 efile_pwritev(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+    Sint64 bytes_written;
+    ssize_t result;
+
+#if !defined(HAVE_PWRITEV) && !defined(HAVE_PWRITE)
+    off_t original_position = lseek(u->fd, 0, SEEK_CUR);
+
+    if(original_position < 0 || lseek(u->fd, offset, SEEK_SET) < 0) {
+        u->common.posix_errno = errno;
+        return -1;
+    }
+#endif
+
+    bytes_written = 0;
+
+    do {
+        if(iovlen < 1) {
+            result = 0;
+            break;
+        }
+
+#if defined(HAVE_PWRITEV)
+        result = pwritev(u->fd, iov, MIN(IOV_MAX, iovlen), offset);
+#elif defined(HAVE_PWRITE)
+        result = pwrite(u->fd, iov->iov_base, iov->iov_len, offset);
+#else
+        result = write(u->fd, iov->iov_base, iov->iov_len);
+#endif
+
+        if(result > 0) {
+            shift_iov(&iov, &iovlen, result);
+            bytes_written += result;
+            offset += result;
+        }
+    } while(result > 0 || (result < 0 && errno == EINTR));
+
+    u->common.posix_errno = errno;
+
+#if !defined(HAVE_PWRITEV) && !defined(HAVE_PWRITE)
+    if(result >= 0) {
+        if(lseek(u->fd, original_position, SEEK_SET) < 0) {
+            u->common.posix_errno = errno;
+            return -1;
+        }
+    }
+#endif
+
+    if(result == 0 && bytes_written > 0) {
+        return bytes_written;
+    }
+
+    return result;
+}
+
+int efile_seek(efile_data_t *d, enum efile_seek_t seek, Sint64 offset, Sint64 *new_position) {
+    efile_unix_t *u = (efile_unix_t*)d;
+    off_t result;
+    int whence;
+
+    switch(seek) {
+        case EFILE_SEEK_BOF: whence = SEEK_SET; break;
+        case EFILE_SEEK_CUR: whence = SEEK_CUR; break;
+        case EFILE_SEEK_EOF: whence = SEEK_END; break;
+        default: ERTS_INTERNAL_ERROR("Invalid seek parameter");
+    }
+
+    result = lseek(u->fd, offset, whence);
+
+    /*
+     * The man page for lseek (on SunOs 5) says:
+     *
+     * "if fildes is a remote file descriptor and offset is negative, lseek()
+     * returns the file pointer even if it is negative."
+     */
+    if(result < 0 && errno == 0) {
+        errno = EINVAL;
+    }
+
+    if(result < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    (*new_position) = result;
+
+    return 1;
+}
+
+int efile_sync(efile_data_t *d, int data_only) {
+    efile_unix_t *u = (efile_unix_t*)d;
+
+#if defined(HAVE_FDATASYNC) && !defined(__DARWIN__)
+    if(data_only) {
+        if(fdatasync(u->fd) < 0) {
+            u->common.posix_errno = errno;
+            return 0;
+        }
+
+        return 1;
+    }
+#endif
+
+#if defined(__DARWIN__) && defined(F_FULLFSYNC)
+    if(fcntl(u->fd, F_FULLFSYNC) < 0) {
+#else
+    if(fsync(u->fd) < 0) {
+#endif
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    return 1;
+}
+
+int efile_advise(efile_data_t *d, Sint64 offset, Sint64 length, enum efile_advise_t advise) {
+#ifdef HAVE_POSIX_FADVISE
+    efile_unix_t *u = (efile_unix_t*)d;
+    int p_advise;
+
+    switch(advise) {
+        case EFILE_ADVISE_NORMAL: p_advise = POSIX_FADV_NORMAL; break;
+        case EFILE_ADVISE_RANDOM: p_advise = POSIX_FADV_RANDOM; break;
+        case EFILE_ADVISE_SEQUENTIAL: p_advise = POSIX_FADV_SEQUENTIAL; break;
+        case EFILE_ADVISE_WILL_NEED: p_advise = POSIX_FADV_WILLNEED; break;
+        case EFILE_ADVISE_DONT_NEED: p_advise = POSIX_FADV_DONTNEED; break;
+        case EFILE_ADVISE_NO_REUSE: p_advise = POSIX_FADV_NOREUSE; break;
+        default:
+            u->common.posix_errno = EINVAL;
+            return 0;
+    }
+
+    if(posix_fadvise(u->fd, offset, length, p_advise) < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    return 1;
+#else
+    /* We'll pretend to support this syscall as it's only a recommendation even
+     * on systems that do support it. */
+    return 1;
+#endif
+}
+
+int efile_allocate(efile_data_t *d, Sint64 offset, Sint64 length) {
+    efile_unix_t *u = (efile_unix_t*)d;
+    int ret = -1;
+
+    /* We prefer OS-specific methods, but fall back to posix_fallocate on
+     * failure. It's unclear whether this has any practical benefit on
+     * modern systems, but the old driver did it. */
+
+#if defined(HAVE_FALLOCATE)
+    /* Linux-specific */
+    do {
+        ret = fallocate(u->fd, FALLOC_FL_KEEP_SIZE, offset, length);
+    } while(ret < 0 && errno == EINTR);
+#elif defined(F_PREALLOCATE)
+    /* Mac-specific */
+    fstore_t fs = {};
+
+    fs.fst_flags = F_ALLOCATECONTIG;
+    fs.fst_posmode = F_VOLPOSMODE;
+    fs.fst_offset = offset;
+    fs.fst_length = length;
+
+    ret = fcntl(u->fd, F_PREALLOCATE, &fs);
+    if(ret < 0) {
+        fs.fst_flags = F_ALLOCATEALL;
+        ret = fcntl(u->fd, F_PREALLOCATE, &fs);
+    }
+#elif !defined(HAVE_POSIX_FALLOCATE)
+    u->common.posix_errno = ENOTSUP;
+    return 0;
+#endif
+
+#ifdef HAVE_POSIX_FALLOCATE
+    if(ret < 0) {
+        do {
+            ret = posix_fallocate(u->fd, offset, length);
+
+            /* On Linux and Solaris for example, posix_fallocate() returns a
+             * positive error number on error and it does not set errno. On
+             * FreeBSD however (9.0 at least), it returns -1 on error and it
+             * sets errno. */
+            if (ret > 0) {
+                errno = ret;
+                ret = -1;
+            }
+        } while(ret < 0 && errno == EINTR);
+    }
+#endif
+
+    if(ret < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    return 1;
+}
+
+int efile_truncate(efile_data_t *d) {
+    efile_unix_t *u = (efile_unix_t*)d;
+    off_t offset;
+
+    offset = lseek(u->fd, 0, SEEK_CUR);
+
+    if(offset < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    if(ftruncate(u->fd, offset) < 0) {
+        u->common.posix_errno = errno;
+        return 0;
+    }
+
+    return 1;
+}
+
+posix_errno_t efile_read_info(const efile_path_t *path, int follow_links, efile_fileinfo_t *result) {
+    struct stat data;
+
+    if(follow_links) {
+        if(stat((const char*)path->data, &data) < 0) {
+            return errno;
+        }
+    } else {
+        if(lstat((const char*)path->data, &data) < 0) {
+            return errno;
+        }
+    }
+
+    if(S_ISCHR(data.st_mode) || S_ISBLK(data.st_mode)) {
+        result->type = EFILE_FILETYPE_DEVICE;
+    } else if(S_ISDIR(data.st_mode)) {
+        result->type = EFILE_FILETYPE_DIRECTORY;
+    } else if(S_ISREG(data.st_mode)) {
+        result->type = EFILE_FILETYPE_REGULAR;
+    } else if(S_ISLNK(data.st_mode)) {
+        result->type = EFILE_FILETYPE_SYMLINK;
+    } else {
+        result->type = EFILE_FILETYPE_OTHER;
+    }
+
+    result->a_time = (Sint64)data.st_atime;
+    result->m_time = (Sint64)data.st_mtime;
+    result->c_time = (Sint64)data.st_ctime;
+    result->size = data.st_size;
+
+    result->major_device = data.st_dev;
+    result->minor_device = data.st_rdev;
+    result->links = data.st_nlink;
+    result->inode = data.st_ino;
+    result->mode = data.st_mode;
+    result->uid = data.st_uid;
+    result->gid = data.st_gid;
+
+#ifndef NO_ACCESS
+    result->access = EFILE_ACCESS_NONE;
+
+    if(access((const char*)path->data, R_OK) == 0) {
+        result->access |= EFILE_ACCESS_READ;
+    }
+    if(access((const char*)path->data, W_OK) == 0) {
+        result->access |= EFILE_ACCESS_WRITE;
+    }
+#else
+    /* Just look at read/write access for owner. */
+    result->access = ((data.st_mode >> 6) & 07) >> 1;
+#endif
+
+    return 0;
+}
+
+posix_errno_t efile_set_permissions(const efile_path_t *path, Uint32 permissions) {
+    const mode_t MUTABLE_MODES = (S_ISUID | S_ISGID | S_IRWXU | S_IRWXG | S_IRWXO);
+    mode_t new_modes = permissions & MUTABLE_MODES;
+
+    if(chmod((const char*)path->data, new_modes) < 0) {
+        new_modes &= ~(S_ISUID | S_ISGID);
+
+        if (chmod((const char*)path->data, new_modes) < 0) {
+            return errno;
+        }
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_set_owner(const efile_path_t *path, Sint32 owner, Sint32 group) {
+    if(chown((const char*)path->data, owner, group) < 0) {
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_set_time(const efile_path_t *path, Sint64 a_time, Sint64 m_time, Sint64 c_time) {
+    struct utimbuf tval;
+
+    tval.actime = (time_t)a_time;
+    tval.modtime = (time_t)m_time;
+
+    (void)c_time;
+
+    if(utime((const char*)path->data, &tval) < 0) {
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_read_link(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    ErlNifBinary result_bin;
+
+    if(!enif_alloc_binary(256, &result_bin)) {
+        return ENOMEM;
+    }
+
+    for(;;) {
+        ssize_t bytes_copied;
+
+        bytes_copied = readlink((const char*)path->data, (char*)result_bin.data,
+            result_bin.size);
+
+        if(bytes_copied <= 0) {
+            posix_errno_t saved_errno = errno;
+            enif_release_binary(&result_bin);
+            return saved_errno;
+        } else if(bytes_copied < result_bin.size) {
+            if(!enif_realloc_binary(&result_bin, bytes_copied)) {
+                enif_release_binary(&result_bin);
+                return ENOMEM;
+            }
+
+            (*result) = enif_make_binary(env, &result_bin);
+
+            return 0;
+        }
+
+        /* The result didn't fit into the buffer, so we'll try again with a
+         * larger one. */
+
+        if(!enif_realloc_binary(&result_bin, result_bin.size * 2)) {
+            enif_release_binary(&result_bin);
+            return ENOMEM;
+        }
+    }
+}
+
+static int is_ignored_name(int name_length, const char *name) {
+    if(name_length == 1 && name[0] == '.') {
+        return 1;
+    } else if(name_length == 2 && memcmp(name, "..", 2) == 0) {
+        return 1;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_list_dir(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    ERL_NIF_TERM list_head;
+
+    struct dirent *dir_entry;
+    DIR *dir_stream;
+
+    dir_stream = opendir((const char*)path->data);
+    if(dir_stream == NULL) {
+        posix_errno_t saved_errno = errno;
+        *result = enif_make_list(env, 0);
+        return saved_errno;
+    }
+
+    list_head = enif_make_list(env, 0);
+    dir_entry = readdir(dir_stream);
+
+    while(dir_entry != NULL) {
+        int name_length = strlen(dir_entry->d_name);
+
+        if(!is_ignored_name(name_length, dir_entry->d_name)) {
+            unsigned char *name_bytes;
+            ERL_NIF_TERM name_term;
+
+            name_bytes = enif_make_new_binary(env, name_length, &name_term);
+            sys_memcpy(name_bytes, dir_entry->d_name, name_length);
+
+            list_head = enif_make_list_cell(env, name_term, list_head);
+        }
+
+        dir_entry = readdir(dir_stream);
+    }
+
+    (*result) = list_head;
+    closedir(dir_stream);
+
+    return 0;
+}
+
+posix_errno_t efile_rename(const efile_path_t *old_path, const efile_path_t *new_path) {
+    if(rename((const char*)old_path->data, (const char*)new_path->data) < 0) {
+        if(errno == ENOTEMPTY) {
+            return EEXIST;
+        }
+
+        if(strcmp((const char*)old_path->data, "/") == 0) {
+            /* Alpha reports renaming / as EBUSY and Linux reports it as EACCES
+             * instead of EINVAL.*/
+             return EINVAL;
+        }
+
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_make_hard_link(const efile_path_t *existing_path, const efile_path_t *new_path) {
+    if(link((const char*)existing_path->data, (const char*)new_path->data) < 0) {
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_make_soft_link(const efile_path_t *existing_path, const efile_path_t *new_path) {
+    if(symlink((const char*)existing_path->data, (const char*)new_path->data) < 0) {
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_make_dir(const efile_path_t *path) {
+#ifdef NO_MKDIR_MODE
+    if(mkdir((const char*)path->data) < 0) {
+#else
+    if(mkdir((const char*)path->data, DIR_MODE) < 0) {
+#endif
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_del_file(const efile_path_t *path) {
+    if(unlink((const char*)path->data) < 0) {
+        /* Linux sets the wrong error code. */
+        if(errno == EISDIR) {
+            return EPERM;
+        }
+
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_del_dir(const efile_path_t *path) {
+    if(rmdir((const char*)path->data) < 0) {
+        posix_errno_t saved_errno = errno;
+
+        if(saved_errno == ENOTEMPTY) {
+            saved_errno = EEXIST;
+        }
+
+        /* The error code might be wrong if we're trying to delete the current
+         * directory. */
+        if(saved_errno == EEXIST) {
+            struct stat path_stat, cwd_stat;
+            int has_stat;
+
+            has_stat = (stat((const char*)path->data, &path_stat) == 0);
+            has_stat &= (stat(".", &cwd_stat) == 0);
+
+            if(has_stat && path_stat.st_ino == cwd_stat.st_ino) {
+                if(path_stat.st_dev == cwd_stat.st_dev) {
+                    return EINVAL;
+                }
+            }
+        }
+
+        return saved_errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_set_cwd(const efile_path_t *path) {
+    if(chdir((const char*)path->data) < 0) {
+        return errno;
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_get_device_cwd(ErlNifEnv *env, int device_index, ERL_NIF_TERM *result) {
+    (void)device_index;
+    (void)result;
+    (void)env;
+
+    return ENOTSUP;
+}
+
+posix_errno_t efile_get_cwd(ErlNifEnv *env, ERL_NIF_TERM *result) {
+    ErlNifBinary result_bin;
+    size_t bytes_copied;
+
+    if(!enif_alloc_binary(256, &result_bin)) {
+        return ENOMEM;
+    }
+
+    while(getcwd((char*)result_bin.data, result_bin.size) == NULL) {
+        posix_errno_t saved_errno = errno;
+
+        if(saved_errno != ERANGE) {
+            enif_release_binary(&result_bin);
+            return saved_errno;
+        } else {
+            if(!enif_realloc_binary(&result_bin, result_bin.size * 2)) {
+                enif_release_binary(&result_bin);
+                return ENOMEM;
+            }
+        }
+    }
+
+    /* getcwd(2) guarantees null-termination. */
+    bytes_copied = strlen((const char*)result_bin.data);
+
+    if(!enif_realloc_binary(&result_bin, bytes_copied)) {
+        enif_release_binary(&result_bin);
+        return ENOMEM;
+    }
+
+    (*result) = enif_make_binary(env, &result_bin);
+
+    return 0;
+}
+
+posix_errno_t efile_altname(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    (void)path;
+    (void)result;
+
+    return ENOTSUP;
+}
diff --git a/erts/emulator/nifs/win32/win_prim_file.c b/erts/emulator/nifs/win32/win_prim_file.c
new file mode 100644
index 0000000000..044bee62cf
--- /dev/null
+++ b/erts/emulator/nifs/win32/win_prim_file.c
@@ -0,0 +1,1474 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#include "erl_nif.h"
+#include "config.h"
+#include "sys.h"
+
+#include "prim_file_nif.h"
+
+#include <winioctl.h>
+#include <windows.h>
+#include <strsafe.h>
+#include <wchar.h>
+
+#define IS_SLASH(a)  ((a) == L'\\' || (a) == L'/')
+
+#define FILE_SHARE_FLAGS (FILE_SHARE_READ | FILE_SHARE_WRITE | FILE_SHARE_DELETE)
+
+#define LP_PREFIX L"\\\\?\\"
+#define LP_PREFIX_SIZE (sizeof(LP_PREFIX) - sizeof(WCHAR))
+#define LP_PREFIX_LENGTH (LP_PREFIX_SIZE / sizeof(WCHAR))
+
+#define PATH_LENGTH(path) (path->size / sizeof(WCHAR) - 1)
+
+#define ASSERT_PATH_FORMAT(path) \
+    do { \
+        ASSERT(PATH_LENGTH(path) >= 4 && \
+            !memcmp(path->data, LP_PREFIX, LP_PREFIX_SIZE)); \
+        ASSERT(PATH_LENGTH(path) == wcslen((WCHAR*)path->data)); \
+    } while(0)
+
+#define TICKS_PER_SECOND (10000000ULL)
+#define EPOCH_DIFFERENCE (11644473600LL)
+
+#define FILETIME_TO_EPOCH(epoch, ft) \
+    do { \
+        ULARGE_INTEGER ull; \
+        ull.LowPart  = (ft).dwLowDateTime; \
+        ull.HighPart = (ft).dwHighDateTime; \
+        (epoch) = ((ull.QuadPart / TICKS_PER_SECOND) - EPOCH_DIFFERENCE); \
+    } while(0)
+
+#define EPOCH_TO_FILETIME(ft, epoch) \
+    do { \
+        ULARGE_INTEGER ull; \
+        ull.QuadPart = (((epoch) + EPOCH_DIFFERENCE) * TICKS_PER_SECOND); \
+        (ft).dwLowDateTime = ull.LowPart; \
+        (ft).dwHighDateTime = ull.HighPart; \
+    } while(0)
+
+typedef struct {
+    efile_data_t common;
+    HANDLE handle;
+} efile_win_t;
+
+static int windows_to_posix_errno(DWORD last_error);
+
+static int has_invalid_null_termination(const ErlNifBinary *path) {
+    const WCHAR *null_pos, *end_pos;
+
+    null_pos = wmemchr((const WCHAR*)path->data, L'\0', path->size);
+    end_pos = (const WCHAR*)&path->data[path->size] - 1;
+
+    if(null_pos == NULL) {
+        return 1;
+    }
+
+    /* prim_file:internal_name2native sometimes feeds us data that is "doubly"
+     * NUL-terminated, so we'll accept any number of trailing NULs so long as
+     * they aren't interrupted by anything else. */
+    while(null_pos < end_pos && (*null_pos) == L'\0') {
+        null_pos++;
+    }
+
+    return null_pos != end_pos;
+}
+
+static posix_errno_t get_full_path(ErlNifEnv *env, WCHAR *input, efile_path_t *result) {
+    DWORD maximum_length, actual_length;
+    int add_long_prefix;
+
+    maximum_length = GetFullPathNameW(input, 0, NULL, NULL);
+    add_long_prefix = 0;
+
+    if(maximum_length == 0) {
+        /* POSIX doesn't have the concept of a "path error" in the same way
+         * Windows does, so we'll return ENOENT since that's what most POSIX
+         * APIs would return if they were fed such garbage. */
+        return ENOENT;
+    }
+
+    maximum_length += LP_PREFIX_LENGTH;
+
+    if(!enif_alloc_binary(maximum_length * sizeof(WCHAR), result)) {
+        return ENOMEM;
+    }
+
+    actual_length = GetFullPathNameW(input, maximum_length, (WCHAR*)result->data, NULL);
+
+    if(actual_length < maximum_length) {
+        int has_long_path_prefix;
+        WCHAR *path_start;
+
+        /* Make sure we have a long-path prefix; GetFullPathNameW only adds one
+         * if the path is relative. */
+        has_long_path_prefix = actual_length >= LP_PREFIX_LENGTH &&
+            !sys_memcmp(result->data, LP_PREFIX, LP_PREFIX_SIZE);
+
+        if(!has_long_path_prefix) {
+            sys_memmove(result->data + LP_PREFIX_SIZE, result->data,
+                (actual_length + 1) * sizeof(WCHAR));
+            sys_memcpy(result->data, LP_PREFIX, LP_PREFIX_SIZE);
+            actual_length += LP_PREFIX_LENGTH;
+        }
+
+        path_start = (WCHAR*)result->data;
+
+        /* We're removing trailing slashes since quite a few APIs refuse to
+         * work with them, and none require them. We only check the last
+         * character since GetFullPathNameW folds slashes together. */
+        if(IS_SLASH(path_start[actual_length - 1])) {
+            if(path_start[actual_length - 2] != L':') {
+                path_start[actual_length - 1] = L'\0';
+                actual_length--;
+            }
+        }
+
+        if(!enif_realloc_binary(result, (actual_length + 1) * sizeof(WCHAR))) {
+            enif_release_binary(result);
+            return ENOMEM;
+        }
+
+        enif_make_binary(env, result);
+        return 0;
+    }
+
+    /* We may end up here if the current directory changes to something longer
+     * between/during GetFullPathName. There's nothing sensible we can do about
+     * this. */
+
+    enif_release_binary(result);
+
+    return EINVAL;
+}
+
+posix_errno_t efile_marshal_path(ErlNifEnv *env, ERL_NIF_TERM path, efile_path_t *result) {
+    ErlNifBinary raw_path;
+
+    if(!enif_inspect_binary(env, path, &raw_path)) {
+        return EINVAL;
+    } else if(raw_path.size % sizeof(WCHAR)) {
+        return EINVAL;
+    }
+
+    if(has_invalid_null_termination(&raw_path)) {
+        return EINVAL;
+    }
+
+    return get_full_path(env, (WCHAR*)raw_path.data, result);
+}
+
+ERL_NIF_TERM efile_get_handle(ErlNifEnv *env, efile_data_t *d) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    ERL_NIF_TERM result;
+    unsigned char *bits;
+
+    bits = enif_make_new_binary(env, sizeof(w->handle), &result);
+    memcpy(bits, &w->handle, sizeof(w->handle));
+
+    return result;
+}
+
+/** @brief Converts a native path to the preferred form in "erlang space,"
+ * without path-prefixes, forward-slashes, or NUL terminators. */
+static int normalize_path_result(ErlNifBinary *path) {
+    WCHAR *path_iterator, *path_start, *path_end;
+    int length;
+
+    path_start = (WCHAR*)path->data;
+    length = wcslen(path_start);
+
+    ASSERT(length < path->size / sizeof(WCHAR));
+
+    /* Get rid of the long-path prefix, if present. */
+    if(length >= LP_PREFIX_LENGTH) {
+        if(!sys_memcmp(path_start, LP_PREFIX, LP_PREFIX_SIZE)) {
+            length -= LP_PREFIX_LENGTH;
+
+            sys_memmove(path_start, &path_start[LP_PREFIX_LENGTH],
+                length * sizeof(WCHAR));
+        }
+    }
+
+    path_end = &path_start[length];
+    path_iterator = path_start;
+
+    /* Convert drive letters to lowercase, if present. */
+    if(length >= 2 && path_start[1] == L':') {
+        WCHAR drive_letter = path_start[0];
+
+        if(drive_letter >= L'A' && drive_letter <= L'Z') {
+            path_start[0] = drive_letter - L'A' + L'a';
+        }
+    }
+
+    while(path_iterator < path_end) {
+        if(*path_iterator == L'\\') {
+            *path_iterator = L'/';
+        }
+
+        path_iterator++;
+    }
+
+    /* Truncate the result to its actual length; we don't want to include the
+     * NUL terminator. */
+    return enif_realloc_binary(path, length * sizeof(WCHAR));
+}
+
+/* @brief Checks whether all the given attributes are set on the object at the
+ * given path. Note that it assumes false on errors. */
+static int has_file_attributes(const efile_path_t *path, DWORD mask) {
+    DWORD attributes = GetFileAttributesW((WCHAR*)path->data);
+
+    if(attributes == INVALID_FILE_ATTRIBUTES) {
+        return 0;
+    }
+
+    return !!((attributes & mask) == mask);
+}
+
+static int is_ignored_name(int name_length, const WCHAR *name) {
+    if(name_length == 1 && name[0] == L'.') {
+        return 1;
+    } else if(name_length == 2 && !sys_memcmp(name, L"..", 2 * sizeof(WCHAR))) {
+        return 1;
+    }
+
+    return 0;
+}
+
+static int get_drive_number(const efile_path_t *path) {
+    const WCHAR *path_start;
+    int length;
+
+    ASSERT_PATH_FORMAT(path);
+
+    path_start = (WCHAR*)path->data + LP_PREFIX_LENGTH;
+    length = PATH_LENGTH(path) - LP_PREFIX_LENGTH;
+
+    if(length >= 2 && path_start[1] == L':') {
+        WCHAR drive_letter = path_start[0];
+
+        if(drive_letter >= L'A' && drive_letter <= L'Z') {
+            return drive_letter - L'A' + 1;
+        } else if(drive_letter >= L'a' && drive_letter <= L'z') {
+            return drive_letter - L'a' + 1;
+        }
+    }
+
+    return -1;
+}
+
+/* @brief Checks whether two *paths* are on the same mount point; they don't
+ * have to refer to existing or accessible files/directories. */
+static int has_same_mount_point(const efile_path_t *path_a, const efile_path_t *path_b) {
+    WCHAR *mount_a, *mount_b;
+    int result = 0;
+
+    mount_a = enif_alloc(path_a->size);
+    mount_b = enif_alloc(path_b->size);
+
+    if(mount_a != NULL && mount_b != NULL) {
+        int length_a, length_b;
+
+        length_a = PATH_LENGTH(path_a);
+        length_b = PATH_LENGTH(path_b);
+
+        if(GetVolumePathNameW((WCHAR*)path_a->data, mount_a, length_a)) {
+            ASSERT(wcslen(mount_a) <= length_a);
+
+            if(GetVolumePathNameW((WCHAR*)path_b->data, mount_b, length_b)) {
+                ASSERT(wcslen(mount_b) <= length_b);
+
+                result = !_wcsicmp(mount_a, mount_b);
+            }
+        }
+    }
+
+    if(mount_b != NULL) {
+        enif_free(mount_b);
+    }
+
+    if(mount_a != NULL) {
+        enif_free(mount_a);
+    }
+
+    return result;
+}
+
+/* Mirrors the PathIsRootW function of the shell API, but doesn't choke on
+ * paths longer than MAX_PATH. */
+static int is_path_root(const efile_path_t *path) {
+    const WCHAR *path_start, *path_end;
+    int length;
+
+    ASSERT_PATH_FORMAT(path);
+
+    path_start = (WCHAR*)path->data + LP_PREFIX_LENGTH;
+    length = PATH_LENGTH(path) - LP_PREFIX_LENGTH;
+
+    path_end = &path_start[length];
+
+    if(length == 1) {
+        /* A single \ refers to the root of the current working directory. */
+        return IS_SLASH(path_start[0]);
+    } else if(length == 3 && iswalpha(path_start[0]) && path_start[1] == L':') {
+        /* Drive letter. */
+        return IS_SLASH(path_start[2]);
+    } else if(length >= 4) {
+        /* Check whether we're a UNC root, eg. \\server, \\server\share */
+        const WCHAR *path_iterator;
+
+        if(!IS_SLASH(path_start[0]) || !IS_SLASH(path_start[1])) {
+            return 0;
+        }
+
+        path_iterator = path_start + 2;
+
+        /* Slide to the slash between the server and share names, if present. */
+        while(path_iterator < path_end && !IS_SLASH(*path_iterator)) {
+            path_iterator++;
+        }
+
+        /* Slide past the end of the string, stopping at the first slash we
+         * encounter. */
+        do {
+            path_iterator++;
+        }  while(path_iterator < path_end && !IS_SLASH(*path_iterator));
+
+        /* If we're past the end of the string and it didnt't end with a slash,
+         * then we're a root path. */
+        return path_iterator >= path_end && !IS_SLASH(path_start[length - 1]);
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_open(const efile_path_t *path, enum efile_modes_t modes,
+        ErlNifResourceType *nif_type, efile_data_t **d) {
+
+    DWORD attributes, access_flags, open_mode;
+    HANDLE handle;
+
+    ASSERT_PATH_FORMAT(path);
+
+    access_flags = 0;
+    open_mode = 0;
+
+    if(modes & EFILE_MODE_READ && !(modes & EFILE_MODE_WRITE)) {
+        access_flags = GENERIC_READ;
+        open_mode = OPEN_EXISTING;
+    } else if(modes & EFILE_MODE_WRITE && !(modes & EFILE_MODE_READ)) {
+        access_flags = GENERIC_WRITE;
+        open_mode = CREATE_ALWAYS;
+    } else if(modes & EFILE_MODE_READ_WRITE) {
+        access_flags = GENERIC_READ | GENERIC_WRITE;
+        open_mode = OPEN_ALWAYS;
+    } else {
+        return EINVAL;
+    }
+
+    if(modes & EFILE_MODE_APPEND) {
+        access_flags |= FILE_APPEND_DATA;
+        open_mode = OPEN_ALWAYS;
+    }
+
+    if(modes & EFILE_MODE_EXCLUSIVE) {
+        open_mode = CREATE_NEW;
+    }
+
+    if(modes & EFILE_MODE_SYNC) {
+        attributes = FILE_FLAG_WRITE_THROUGH;
+    } else {
+        attributes = FILE_ATTRIBUTE_NORMAL;
+    }
+
+    handle = CreateFileW((WCHAR*)path->data, access_flags,
+        FILE_SHARE_FLAGS, NULL, open_mode, attributes, NULL);
+
+    if(handle != INVALID_HANDLE_VALUE) {
+        efile_win_t *w;
+
+        w = (efile_win_t*)enif_alloc_resource(nif_type, sizeof(efile_win_t));
+        w->handle = handle;
+
+        EFILE_INIT_RESOURCE(&w->common, modes);
+        (*d) = &w->common;
+
+        return 0;
+    } else {
+        DWORD last_error = GetLastError();
+
+        /* Rewrite all failures on directories to EISDIR to match the old
+         * driver. */
+        if(has_file_attributes(path, FILE_ATTRIBUTE_DIRECTORY)) {
+            return EISDIR;
+        }
+
+        return windows_to_posix_errno(last_error);
+    }
+}
+
+int efile_close(efile_data_t *d) {
+    efile_win_t *w = (efile_win_t*)d;
+    HANDLE handle;
+
+    ASSERT(erts_atomic32_read_nob(&d->state) == EFILE_STATE_CLOSED);
+    ASSERT(w->handle != INVALID_HANDLE_VALUE);
+
+    handle = w->handle;
+    w->handle = INVALID_HANDLE_VALUE;
+
+    if(!CloseHandle(handle)) {
+        w->common.posix_errno = windows_to_posix_errno(GetLastError());
+        return 0;
+    }
+
+    return 1;
+}
+
+static void shift_overlapped(OVERLAPPED *overlapped, DWORD shift) {
+    LARGE_INTEGER offset;
+
+    ASSERT(shift >= 0);
+
+    offset.HighPart = overlapped->OffsetHigh;
+    offset.LowPart = overlapped->Offset;
+
+    /* ~(Uint64)0 is a magic value ("append to end of file") which needs to be
+     * preserved. Other positions resulting in overflow would have errored out
+     * just prior to this point. */
+    if(offset.QuadPart != ERTS_UINT64_MAX) {
+        offset.QuadPart += shift;
+    }
+
+    /* All unused fields must be zeroed for the next call. */
+    sys_memset(overlapped, 0, sizeof(*overlapped));
+    overlapped->OffsetHigh = offset.HighPart;
+    overlapped->Offset = offset.LowPart;
+}
+
+static void shift_iov(SysIOVec **iov, int *iovlen, DWORD shift) {
+    SysIOVec *head_vec = (*iov);
+
+    ASSERT(shift >= 0);
+
+    while(shift > 0) {
+        ASSERT(head_vec < &(*iov)[*iovlen]);
+
+        if(shift < head_vec->iov_len) {
+            head_vec->iov_base = (char*)head_vec->iov_base + shift;
+            head_vec->iov_len -= shift;
+            break;
+        } else {
+            shift -= head_vec->iov_len;
+            head_vec++;
+        }
+    }
+
+    (*iovlen) -= head_vec - (*iov);
+    (*iov) = head_vec;
+}
+
+typedef BOOL (WINAPI *io_op_t)(HANDLE, LPVOID, DWORD, LPDWORD, LPOVERLAPPED);
+
+static Sint64 internal_sync_io(efile_win_t *w, io_op_t operation,
+        SysIOVec *iov, int iovlen, OVERLAPPED *overlapped) {
+
+    Sint64 bytes_processed = 0;
+
+    for(;;) {
+        DWORD block_bytes_processed, last_error;
+        BOOL succeeded;
+
+        if(iovlen < 1) {
+            return bytes_processed;
+        }
+
+        succeeded = operation(w->handle, iov->iov_base, iov->iov_len,
+            &block_bytes_processed, overlapped);
+        last_error = GetLastError();
+
+        if(!succeeded && (last_error != ERROR_HANDLE_EOF)) {
+            w->common.posix_errno = windows_to_posix_errno(last_error);
+            return -1;
+        } else if(block_bytes_processed == 0) {
+            /* EOF */
+            return bytes_processed;
+        }
+
+        if(overlapped != NULL) {
+            shift_overlapped(overlapped, block_bytes_processed);
+        }
+
+        shift_iov(&iov, &iovlen, block_bytes_processed);
+
+        bytes_processed += block_bytes_processed;
+    }
+}
+
+Sint64 efile_readv(efile_data_t *d, SysIOVec *iov, int iovlen) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    return internal_sync_io(w, ReadFile, iov, iovlen, NULL);
+}
+
+Sint64 efile_writev(efile_data_t *d, SysIOVec *iov, int iovlen) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    OVERLAPPED __overlapped, *overlapped;
+    Uint64 bytes_written;
+
+    if(w->common.modes & EFILE_MODE_APPEND) {
+        overlapped = &__overlapped;
+
+        sys_memset(overlapped, 0, sizeof(*overlapped));
+        overlapped->OffsetHigh = 0xFFFFFFFF;
+        overlapped->Offset = 0xFFFFFFFF;
+    } else {
+        overlapped = NULL;
+    }
+
+    return internal_sync_io(w, WriteFile, iov, iovlen, overlapped);
+}
+
+Sint64 efile_preadv(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    OVERLAPPED overlapped;
+
+    sys_memset(&overlapped, 0, sizeof(overlapped));
+    overlapped.OffsetHigh = (offset >> 32) & 0xFFFFFFFF;
+    overlapped.Offset = offset & 0xFFFFFFFF;
+
+    return internal_sync_io(w, ReadFile, iov, iovlen, &overlapped);
+}
+
+Sint64 efile_pwritev(efile_data_t *d, Sint64 offset, SysIOVec *iov, int iovlen) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    OVERLAPPED overlapped;
+
+    sys_memset(&overlapped, 0, sizeof(overlapped));
+    overlapped.OffsetHigh = (offset >> 32) & 0xFFFFFFFF;
+    overlapped.Offset = offset & 0xFFFFFFFF;
+
+    return internal_sync_io(w, WriteFile, iov, iovlen, &overlapped);
+}
+
+int efile_seek(efile_data_t *d, enum efile_seek_t seek, Sint64 offset, Sint64 *new_position) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    LARGE_INTEGER large_offset, large_new_position;
+    DWORD whence;
+
+    switch(seek) {
+        case EFILE_SEEK_BOF: whence = FILE_BEGIN; break;
+        case EFILE_SEEK_CUR: whence = FILE_CURRENT; break;
+        case EFILE_SEEK_EOF: whence = FILE_END; break;
+        default: ERTS_INTERNAL_ERROR("Invalid seek parameter");
+    }
+
+    large_offset.QuadPart = offset;
+
+    if(!SetFilePointerEx(w->handle, large_offset, &large_new_position, whence)) {
+        w->common.posix_errno = windows_to_posix_errno(GetLastError());
+        return 0;
+    }
+
+    (*new_position) = large_new_position.QuadPart;
+
+    return 1;
+}
+
+int efile_sync(efile_data_t *d, int data_only) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    /* Windows doesn't support data-only syncing. */
+    (void)data_only;
+
+    if(!FlushFileBuffers(w->handle)) {
+        w->common.posix_errno = windows_to_posix_errno(GetLastError());
+        return 0;
+    }
+
+    return 1;
+}
+
+int efile_advise(efile_data_t *d, Sint64 offset, Sint64 length, enum efile_advise_t advise) {
+    /* Windows doesn't support this, but we'll pretend it does since the call
+     * is only a recommendation even on systems that do support it. */
+
+    (void)d;
+    (void)offset;
+    (void)length;
+    (void)advise;
+
+    return 1;
+}
+
+int efile_allocate(efile_data_t *d, Sint64 offset, Sint64 length) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    (void)d;
+    (void)offset;
+    (void)length;
+
+    w->common.posix_errno = ENOTSUP;
+
+    return 0;
+}
+
+int efile_truncate(efile_data_t *d) {
+    efile_win_t *w = (efile_win_t*)d;
+
+    if(!SetEndOfFile(w->handle)) {
+        w->common.posix_errno = windows_to_posix_errno(GetLastError());
+        return 0;
+    }
+
+    return 1;
+}
+
+static int is_executable_file(const efile_path_t *path) {
+    /* We're using the file extension in order to be quirks-compliant with the
+     * old driver, which never bothered to check the actual permissions. We
+     * could easily do so now (cf. GetNamedSecurityInfo) but the execute
+     * permission is only relevant for files that are started with the default
+     * loader, and batch files run just fine with read permission alone. */
+
+    int length = PATH_LENGTH(path);
+
+    if(length >= 4) {
+        const WCHAR *last_four = &((WCHAR*)path->data)[length - 4];
+
+        if (!_wcsicmp(last_four, L".exe") ||
+            !_wcsicmp(last_four, L".cmd") ||
+            !_wcsicmp(last_four, L".bat") ||
+            !_wcsicmp(last_four, L".com")) {
+            return 1;
+        }
+    }
+
+    return 0;
+}
+
+/* Returns whether the path refers to a link-like object, e.g. a junction
+ * point, symbolic link, or mounted folder. */
+static int is_name_surrogate(const efile_path_t *path) {
+    HANDLE handle;
+    int result;
+
+    handle = CreateFileW((const WCHAR*)path->data, GENERIC_READ,
+                         FILE_SHARE_FLAGS, NULL, OPEN_EXISTING,
+                         FILE_FLAG_OPEN_REPARSE_POINT |
+                         FILE_FLAG_BACKUP_SEMANTICS,
+                         NULL);
+    result = 0;
+
+    if(handle != INVALID_HANDLE_VALUE) {
+        REPARSE_GUID_DATA_BUFFER reparse_buffer;
+        LPDWORD unused_length;
+        BOOL success;
+
+        success = DeviceIoControl(handle,
+                                  FSCTL_GET_REPARSE_POINT, NULL, 0,
+                                  &reparse_buffer, sizeof(reparse_buffer),
+                                  &unused_length, NULL);
+
+        /* ERROR_MORE_DATA is tolerated since we're guaranteed to have filled
+         * the field we want. */
+        if(success || GetLastError() == ERROR_MORE_DATA) {
+            result = IsReparseTagNameSurrogate(reparse_buffer.ReparseTag);
+        }
+
+        CloseHandle(handle);
+     }
+
+     return result;
+}
+
+posix_errno_t efile_read_info(const efile_path_t *path, int follow_links, efile_fileinfo_t *result) {
+    BY_HANDLE_FILE_INFORMATION native_file_info;
+    DWORD attributes;
+    int is_link;
+
+    sys_memset(&native_file_info, 0, sizeof(native_file_info));
+    is_link = 0;
+
+    attributes = GetFileAttributesW((WCHAR*)path->data);
+
+    if(attributes == INVALID_FILE_ATTRIBUTES) {
+        DWORD last_error = GetLastError();
+
+        /* Querying a network share root fails with ERROR_BAD_NETPATH, so we'll
+         * fake it as a directory just like local roots. */
+        if(!is_path_root(path) || last_error != ERROR_BAD_NETPATH) {
+            return windows_to_posix_errno(last_error);
+        }
+
+        attributes = FILE_ATTRIBUTE_DIRECTORY;
+    } else if(is_path_root(path)) {
+        /* Local (or mounted) roots can be queried with GetFileAttributesW but
+         * lack support for GetFileInformationByHandle, so we'll skip that
+         * part. */
+    } else {
+        HANDLE handle;
+
+        if(attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
+            is_link = is_name_surrogate(path);
+        }
+
+        if(follow_links && is_link) {
+            posix_errno_t posix_errno;
+            efile_path_t resolved_path;
+
+            posix_errno = internal_read_link(path, &resolved_path);
+
+            if(posix_errno != 0) {
+                return posix_errno;
+            }
+
+            return efile_read_info(&resolved_path, 0, result);
+        }
+
+        handle = CreateFileW((const WCHAR*)path->data, GENERIC_READ,
+            FILE_SHARE_FLAGS, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS,
+            NULL);
+
+        /* The old driver never cared whether this succeeded. */
+        if(handle != INVALID_HANDLE_VALUE) {
+            GetFileInformationByHandle(handle, &native_file_info);
+            CloseHandle(handle);
+        }
+
+        FILETIME_TO_EPOCH(result->m_time, native_file_info.ftLastWriteTime);
+        FILETIME_TO_EPOCH(result->a_time, native_file_info.ftLastAccessTime);
+        FILETIME_TO_EPOCH(result->c_time, native_file_info.ftCreationTime);
+
+        if(result->m_time == -EPOCH_DIFFERENCE) {
+            /* Default to 1970 just like the old driver. */
+            result->m_time = 0;
+        }
+
+        if(result->a_time == -EPOCH_DIFFERENCE) {
+            result->a_time = result->m_time;
+        }
+
+        if(result->c_time == -EPOCH_DIFFERENCE) {
+            result->c_time = result->m_time;
+        }
+    }
+
+    if(is_link) {
+        result->type = EFILE_FILETYPE_SYMLINK;
+        /* This should be _S_IFLNK, but the old driver always set
+         * non-directories to _S_IFREG. */
+        result->mode |= _S_IFREG;
+    } else if(attributes & FILE_ATTRIBUTE_DIRECTORY) {
+        result->type = EFILE_FILETYPE_DIRECTORY;
+        result->mode |= _S_IFDIR | _S_IEXEC;
+    } else {
+        if(is_executable_file(path)) {
+            result->mode |= _S_IEXEC;
+        }
+
+        result->type = EFILE_FILETYPE_REGULAR;
+        result->mode |= _S_IFREG;
+    }
+
+    if(!(attributes & FILE_ATTRIBUTE_READONLY)) {
+        result->access = EFILE_ACCESS_READ | EFILE_ACCESS_WRITE;
+        result->mode |= _S_IREAD | _S_IWRITE;
+    } else {
+        result->access = EFILE_ACCESS_READ;
+        result->mode |= _S_IREAD;
+    }
+
+    /* Propagate user mode-bits to group/other fields */
+    result->mode |= (result->mode & 0700) >> 3;
+    result->mode |= (result->mode & 0700) >> 6;
+
+    result->size =
+        ((Uint64)native_file_info.nFileSizeHigh << 32ull) |
+        (Uint64)native_file_info.nFileSizeLow;
+
+    result->links = MAX(1, native_file_info.nNumberOfLinks);
+
+    result->major_device = get_drive_number(path);
+    result->minor_device = 0;
+    result->inode = 0;
+    result->uid = 0;
+    result->gid = 0;
+
+    return 0;
+}
+
+posix_errno_t efile_set_permissions(const efile_path_t *path, Uint32 permissions) {
+    DWORD attributes = GetFileAttributesW((WCHAR*)path->data);
+
+    if(attributes == INVALID_FILE_ATTRIBUTES) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    if(permissions & _S_IWRITE) {
+        attributes &= ~FILE_ATTRIBUTE_READONLY;
+    } else {
+        attributes |= FILE_ATTRIBUTE_READONLY;
+    }
+
+    if(SetFileAttributesW((WCHAR*)path->data, attributes)) {
+        return 0;
+    }
+
+    return windows_to_posix_errno(GetLastError());
+}
+
+posix_errno_t efile_set_owner(const efile_path_t *path, Sint32 owner, Sint32 group) {
+    (void)path;
+    (void)owner;
+    (void)group;
+
+    return 0;
+}
+
+posix_errno_t efile_set_time(const efile_path_t *path, Sint64 a_time, Sint64 m_time, Sint64 c_time) {
+    FILETIME accessed, modified, created;
+    DWORD last_error, attributes;
+    HANDLE handle;
+
+    attributes = GetFileAttributesW((WCHAR*)path->data);
+
+    if(attributes == INVALID_FILE_ATTRIBUTES) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    /* If the file is read-only, we have to make it temporarily writable while
+     * setting new metadata. */
+    if(attributes & FILE_ATTRIBUTE_READONLY) {
+        DWORD without_readonly = attributes & ~FILE_ATTRIBUTE_READONLY;
+
+        if(!SetFileAttributesW((WCHAR*)path->data, without_readonly)) {
+            return windows_to_posix_errno(GetLastError());
+        }
+    }
+
+    EPOCH_TO_FILETIME(modified, m_time);
+    EPOCH_TO_FILETIME(accessed, a_time);
+    EPOCH_TO_FILETIME(created, c_time);
+
+    handle = CreateFileW((WCHAR*)path->data, GENERIC_READ | GENERIC_WRITE,
+        FILE_SHARE_FLAGS, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
+    last_error = GetLastError();
+
+    if(handle != INVALID_HANDLE_VALUE) {
+        if(SetFileTime(handle, &created, &accessed, &modified)) {
+            last_error = ERROR_SUCCESS;
+        } else {
+            last_error = GetLastError();
+        }
+
+        CloseHandle(handle);
+    }
+
+    if(attributes & FILE_ATTRIBUTE_READONLY) {
+        SetFileAttributesW((WCHAR*)path->data, attributes);
+    }
+
+    return windows_to_posix_errno(last_error);
+}
+
+static posix_errno_t internal_read_link(const efile_path_t *path, efile_path_t *result) {
+    DWORD required_length, actual_length;
+    HANDLE link_handle;
+    DWORD last_error;
+
+    link_handle = CreateFileW((WCHAR*)path->data, GENERIC_READ,
+        FILE_SHARE_FLAGS, NULL, OPEN_EXISTING, FILE_FLAG_BACKUP_SEMANTICS, NULL);
+    last_error = GetLastError();
+
+    if(link_handle == INVALID_HANDLE_VALUE) {
+        return windows_to_posix_errno(last_error);
+    }
+
+    required_length = GetFinalPathNameByHandleW(link_handle, NULL, 0, 0);
+    last_error = GetLastError();
+
+    if(required_length <= 0) {
+        CloseHandle(link_handle);
+        return windows_to_posix_errno(last_error);
+    }
+
+    /* Unlike many other path functions (eg. GetFullPathNameW), this one
+     * includes the NUL terminator in its required length. */
+    if(!enif_alloc_binary(required_length * sizeof(WCHAR), result)) {
+        CloseHandle(link_handle);
+        return ENOMEM;
+    }
+
+    actual_length = GetFinalPathNameByHandleW(link_handle,
+        (WCHAR*)result->data, required_length, 0);
+    last_error = GetLastError();
+
+    CloseHandle(link_handle);
+
+    if(actual_length == 0 || actual_length >= required_length) {
+        enif_release_binary(result);
+        return windows_to_posix_errno(last_error);
+    }
+
+    /* GetFinalPathNameByHandle always prepends with "\\?\" and NUL-terminates,
+     * so we never have to touch-up the resulting path. */
+
+    ASSERT_PATH_FORMAT(result);
+
+    return 0;
+}
+
+posix_errno_t efile_read_link(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    posix_errno_t posix_errno;
+    ErlNifBinary result_bin;
+    DWORD attributes;
+
+    ASSERT_PATH_FORMAT(path);
+
+    attributes = GetFileAttributesW((WCHAR*)path->data);
+
+    if(attributes == INVALID_FILE_ATTRIBUTES) {
+        return windows_to_posix_errno(GetLastError());
+    } else if(!(attributes & FILE_ATTRIBUTE_REPARSE_POINT)) {
+        return EINVAL;
+    }
+
+    if(!is_name_surrogate(path)) {
+        return EINVAL;
+    }
+
+    posix_errno = internal_read_link(path, &result_bin);
+
+    if(posix_errno == 0) {
+        if(!normalize_path_result(&result_bin)) {
+            enif_release_binary(&result_bin);
+            return ENOMEM;
+        }
+
+        (*result) = enif_make_binary(env, &result_bin);
+    }
+
+    return posix_errno;
+}
+
+posix_errno_t efile_list_dir(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    ERL_NIF_TERM list_head;
+    WIN32_FIND_DATAW data;
+    HANDLE search_handle;
+    WCHAR *search_path;
+    DWORD last_error;
+
+    ASSERT_PATH_FORMAT(path);
+
+    search_path = enif_alloc(path->size + 2 * sizeof(WCHAR));
+
+    if(search_path == NULL) {
+        return ENOMEM;
+    }
+
+    sys_memcpy(search_path, path->data, path->size);
+    search_path[PATH_LENGTH(path) + 0] = L'\\';
+    search_path[PATH_LENGTH(path) + 1] = L'*';
+    search_path[PATH_LENGTH(path) + 2] = L'\0';
+
+    search_handle = FindFirstFileW(search_path, &data);
+    last_error = GetLastError();
+
+    enif_free(search_path);
+
+    if(search_handle == INVALID_HANDLE_VALUE) {
+        return windows_to_posix_errno(last_error);
+    }
+
+    list_head = enif_make_list(env, 0);
+
+    do {
+        int name_length = wcslen(data.cFileName);
+
+        if(!is_ignored_name(name_length, data.cFileName)) {
+            unsigned char *name_bytes;
+            ERL_NIF_TERM name_term;
+            size_t name_size;
+
+            name_size = name_length * sizeof(WCHAR);
+
+            name_bytes = enif_make_new_binary(env, name_size, &name_term);
+            sys_memcpy(name_bytes, data.cFileName, name_size);
+
+            list_head = enif_make_list_cell(env, name_term, list_head);
+        }
+    } while(FindNextFileW(search_handle, &data));
+
+    FindClose(search_handle);
+    (*result) = list_head;
+
+    return 0;
+}
+
+posix_errno_t efile_rename(const efile_path_t *old_path, const efile_path_t *new_path) {
+    BOOL old_is_directory, new_is_directory;
+    DWORD move_flags, last_error;
+
+    ASSERT_PATH_FORMAT(old_path);
+    ASSERT_PATH_FORMAT(new_path);
+
+    move_flags = MOVEFILE_COPY_ALLOWED | MOVEFILE_WRITE_THROUGH;
+
+    if(MoveFileExW((WCHAR*)old_path->data, (WCHAR*)new_path->data, move_flags)) {
+        return 0;
+    }
+
+    last_error = GetLastError();
+
+    old_is_directory = has_file_attributes(old_path, FILE_ATTRIBUTE_DIRECTORY);
+    new_is_directory = has_file_attributes(new_path, FILE_ATTRIBUTE_DIRECTORY);
+
+    switch(last_error) {
+    case ERROR_SHARING_VIOLATION:
+    case ERROR_ACCESS_DENIED:
+        if(old_is_directory) {
+            BOOL moved_into_itself;
+
+            moved_into_itself = (old_path->size <= new_path->size) &&
+                !_wcsnicmp((WCHAR*)old_path->data, (WCHAR*)new_path->data,
+                    PATH_LENGTH(old_path));
+
+            if(moved_into_itself) {
+                return EINVAL;
+            } else if(is_path_root(old_path)) {
+                return EINVAL;
+            }
+
+            /* Renaming a directory across volumes needs to be rewritten as
+             * EXDEV so that the caller can respond by simulating it with
+             * copy/delete operations.
+             *
+             * Files are handled through MOVEFILE_COPY_ALLOWED. */
+            if(!has_same_mount_point(old_path, new_path)) {
+                return EXDEV;
+            }
+        }
+        break;
+    case ERROR_PATH_NOT_FOUND:
+    case ERROR_FILE_NOT_FOUND:
+        return ENOENT;
+    case ERROR_ALREADY_EXISTS:
+    case ERROR_FILE_EXISTS:
+        if(old_is_directory && !new_is_directory) {
+            return ENOTDIR;
+        } else if(!old_is_directory && new_is_directory) {
+            return EISDIR;
+        } else if(old_is_directory && new_is_directory) {
+            /* This will fail if the destination isn't empty. */
+            if(RemoveDirectoryW((WCHAR*)new_path->data)) {
+                return efile_rename(old_path, new_path);
+            }
+
+            return EEXIST;
+        } else if(!old_is_directory && !new_is_directory) {
+            /* This is pretty iffy; the public documentation says that the
+             * operation may EACCES on some systems when either file is open,
+             * which gives us room to use MOVEFILE_REPLACE_EXISTING and be done
+             * with it, but the old implementation simulated Unix semantics and
+             * there's a lot of code that relies on that.
+             *
+             * The simulation renames the destination to a scratch name to get
+             * around the fact that it's impossible to open (and by extension
+             * rename) a file that's been deleted while open. It has a few
+             * drawbacks though;
+             *
+             * 1) It's not atomic as there's a small window where there's no
+             *    file at all on the destination path.
+             * 2) It will confuse applications that subscribe to folder
+             *    changes.
+             * 3) It will fail if we lack general permission to write in the
+             *    same folder. */
+
+            WCHAR *swap_path = enif_alloc(new_path->size + sizeof(WCHAR) * 64);
+
+            if(swap_path == NULL) {
+                return ENOMEM;
+            } else {
+                static LONGLONG unique_counter = 0;
+                WCHAR *swap_path_end;
+
+                /* We swap in the same folder as the destination to be
+                 * reasonably sure that it's on the same volume. Note that
+                 * we're avoiding GetTempFileNameW as it will fail on long
+                 * paths. */
+
+                sys_memcpy(swap_path, (WCHAR*)new_path->data, new_path->size);
+                swap_path_end = swap_path + PATH_LENGTH(new_path);
+
+                while(!IS_SLASH(*swap_path_end)) {
+                    ASSERT(swap_path_end > swap_path);
+                    swap_path_end--;
+                }
+
+                StringCchPrintfW(&swap_path_end[1], 64, L"erl-%lx-%llx.tmp",
+                    GetCurrentProcessId(), unique_counter);
+                InterlockedIncrement64(&unique_counter);
+            }
+
+            if(MoveFileExW((WCHAR*)new_path->data, swap_path, MOVEFILE_REPLACE_EXISTING)) {
+                if(MoveFileExW((WCHAR*)old_path->data, (WCHAR*)new_path->data, move_flags)) {
+                    last_error = ERROR_SUCCESS;
+                    DeleteFileW(swap_path);
+                } else {
+                    last_error = GetLastError();
+                    MoveFileW(swap_path, (WCHAR*)new_path->data);
+                }
+            } else {
+                last_error = GetLastError();
+                DeleteFileW(swap_path);
+            }
+
+            enif_free(swap_path);
+
+            return windows_to_posix_errno(last_error);
+        }
+
+        return EEXIST;
+    }
+
+    return windows_to_posix_errno(last_error);
+}
+
+posix_errno_t efile_make_hard_link(const efile_path_t *existing_path, const efile_path_t *new_path) {
+    ASSERT_PATH_FORMAT(existing_path);
+    ASSERT_PATH_FORMAT(new_path);
+
+    if(!CreateHardLinkW((WCHAR*)new_path->data, (WCHAR*)existing_path->data, NULL)) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_make_soft_link(const efile_path_t *existing_path, const efile_path_t *new_path) {
+    DWORD link_flags;
+
+    ASSERT_PATH_FORMAT(existing_path);
+    ASSERT_PATH_FORMAT(new_path);
+
+    if(has_file_attributes(existing_path, FILE_ATTRIBUTE_DIRECTORY)) {
+        link_flags = SYMBOLIC_LINK_FLAG_DIRECTORY;
+    } else {
+        link_flags = 0;
+    }
+
+    if(!CreateSymbolicLinkW((WCHAR*)new_path->data, (WCHAR*)existing_path->data, link_flags)) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_make_dir(const efile_path_t *path) {
+    ASSERT_PATH_FORMAT(path);
+
+    if(!CreateDirectoryW((WCHAR*)path->data, NULL)) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_del_file(const efile_path_t *path) {
+    ASSERT_PATH_FORMAT(path);
+
+    if(!DeleteFileW((WCHAR*)path->data)) {
+        DWORD last_error = GetLastError();
+
+        switch(last_error) {
+        case ERROR_INVALID_NAME:
+            /* Attempted to delete a device or similar. */
+            return EACCES;
+        case ERROR_ACCESS_DENIED:
+            /* Windows NT reports removing a directory as EACCES instead of
+             * EPERM. */
+            if(has_file_attributes(path, FILE_ATTRIBUTE_DIRECTORY)) {
+                return EPERM;
+            }
+            break;
+        }
+
+        return windows_to_posix_errno(last_error);
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_del_dir(const efile_path_t *path) {
+    ASSERT_PATH_FORMAT(path);
+
+    if(!RemoveDirectoryW((WCHAR*)path->data)) {
+        DWORD last_error = GetLastError();
+
+        if(last_error == ERROR_DIRECTORY) {
+            return ENOTDIR;
+        }
+
+        return windows_to_posix_errno(last_error);
+    }
+
+    return 0;
+}
+
+posix_errno_t efile_set_cwd(const efile_path_t *path) {
+    const WCHAR *path_start;
+
+    ASSERT_PATH_FORMAT(path);
+
+    /* We have to use _wchdir since that's the only function that updates the
+     * per-drive working directory, but it naively assumes that all paths
+     * starting with \\ are UNC paths, so we have to skip the \\?\-prefix. */
+    path_start = (WCHAR*)path->data + LP_PREFIX_LENGTH;
+
+    if(_wchdir(path_start)) {
+        return windows_to_posix_errno(GetLastError());
+    }
+
+    return 0;
+}
+
+static int is_valid_drive(int device_index) {
+    WCHAR drive_path[4] = {L'?', L':', L'\\', L'\0'};
+
+    if(device_index == 0) {
+        /* Default drive; always valid. */
+        return 1;
+    } else if(device_index > (L'Z' - L'A' + 1)) {
+        return 0;
+    }
+
+    drive_path[0] = device_index + L'A' - 1;
+
+    switch(GetDriveTypeW(drive_path)) {
+    case DRIVE_NO_ROOT_DIR:
+    case DRIVE_UNKNOWN:
+        return 0;
+    }
+
+    return 1;
+}
+
+posix_errno_t efile_get_device_cwd(ErlNifEnv *env, int device_index, ERL_NIF_TERM *result) {
+    ErlNifBinary result_bin;
+
+    /* _wgetdcwd might crash the entire emulator on debug builds since the CRT
+     * invalid parameter handler asserts if passed a non-existent drive (Or
+     * simply one that has been unmounted), so we check it ourselves to avoid
+     * that. */
+    if(!is_valid_drive(device_index)) {
+        return EACCES;
+    }
+
+    if(!enif_alloc_binary(MAX_PATH * sizeof(WCHAR), &result_bin)) {
+        return ENOMEM;
+    }
+
+    if(_wgetdcwd(device_index, (WCHAR*)result_bin.data, MAX_PATH) == NULL) {
+        enif_release_binary(&result_bin);
+        return EACCES;
+    }
+
+    if(!normalize_path_result(&result_bin)) {
+        enif_release_binary(&result_bin);
+        return ENOMEM;
+    }
+
+    (*result) = enif_make_binary(env, &result_bin);
+
+    return 0;
+}
+
+posix_errno_t efile_get_cwd(ErlNifEnv *env, ERL_NIF_TERM *result) {
+    return efile_get_device_cwd(env, 0, result);
+}
+
+posix_errno_t efile_altname(ErlNifEnv *env, const efile_path_t *path, ERL_NIF_TERM *result) {
+    ErlNifBinary result_bin;
+
+    ASSERT_PATH_FORMAT(path);
+
+    if(is_path_root(path)) {
+        /* Root paths can't be queried so we'll just return them as they are. */
+        if(!enif_alloc_binary(path->size, &result_bin)) {
+            return ENOMEM;
+        }
+
+        sys_memcpy(result_bin.data, path->data, path->size);
+    } else {
+        WIN32_FIND_DATAW data;
+        HANDLE handle;
+
+        WCHAR *name_buffer;
+        int name_length;
+
+        /* Reject path wildcards. */
+        if(wcspbrk(&((const WCHAR*)path->data)[4], L"?*")) {
+            return ENOENT;
+        }
+
+        handle = FindFirstFileW((const WCHAR*)path->data, &data);
+
+        if(handle == INVALID_HANDLE_VALUE) {
+            return windows_to_posix_errno(GetLastError());
+        }
+
+        FindClose(handle);
+
+        name_length = wcslen(data.cAlternateFileName);
+
+        if(name_length > 0) {
+            name_buffer = data.cAlternateFileName;
+        } else {
+            name_length = wcslen(data.cFileName);
+            name_buffer = data.cFileName;
+        }
+
+        /* Include NUL-terminator; it will be removed after normalization. */
+        name_length += 1;
+
+        if(!enif_alloc_binary(name_length * sizeof(WCHAR), &result_bin)) {
+            return ENOMEM;
+        }
+
+        sys_memcpy(result_bin.data, name_buffer, name_length * sizeof(WCHAR));
+    }
+
+    if(!normalize_path_result(&result_bin)) {
+        enif_release_binary(&result_bin);
+        return ENOMEM;
+    }
+
+    (*result) = enif_make_binary(env, &result_bin);
+
+    return 0;
+}
+
+static int windows_to_posix_errno(DWORD last_error) {
+    switch(last_error) {
+    case ERROR_SUCCESS:
+        return 0;
+    case ERROR_INVALID_FUNCTION:
+    case ERROR_INVALID_DATA:
+    case ERROR_INVALID_PARAMETER:
+    case ERROR_INVALID_TARGET_HANDLE:
+    case ERROR_INVALID_CATEGORY:
+    case ERROR_NEGATIVE_SEEK:
+        return EINVAL;
+    case ERROR_DIR_NOT_EMPTY:
+        return EEXIST;
+    case ERROR_BAD_FORMAT:
+        return ENOEXEC;
+    case ERROR_PATH_NOT_FOUND:
+    case ERROR_FILE_NOT_FOUND:
+    case ERROR_NO_MORE_FILES:
+    case ERROR_INVALID_NAME:
+        return ENOENT;
+    case ERROR_TOO_MANY_OPEN_FILES:
+        return EMFILE;
+    case ERROR_ACCESS_DENIED:
+    case ERROR_INVALID_ACCESS:
+    case ERROR_CURRENT_DIRECTORY:
+    case ERROR_SHARING_VIOLATION:
+    case ERROR_LOCK_VIOLATION:
+    case ERROR_INVALID_PASSWORD:
+    case ERROR_DRIVE_LOCKED:
+        return EACCES;
+    case ERROR_INVALID_HANDLE:
+        return EBADF;
+    case ERROR_NOT_ENOUGH_MEMORY:
+    case ERROR_OUTOFMEMORY:
+    case ERROR_OUT_OF_STRUCTURES:
+        return ENOMEM;
+    case ERROR_INVALID_DRIVE:
+    case ERROR_BAD_UNIT:
+    case ERROR_NOT_READY:
+    case ERROR_REM_NOT_LIST:
+    case ERROR_DUP_NAME:
+    case ERROR_BAD_NETPATH:
+    case ERROR_NETWORK_BUSY:
+    case ERROR_DEV_NOT_EXIST:
+    case ERROR_BAD_NET_NAME:
+        return ENXIO;
+    case ERROR_NOT_SAME_DEVICE:
+        return EXDEV;
+    case ERROR_WRITE_PROTECT:
+        return EROFS;
+    case ERROR_BAD_LENGTH:
+    case ERROR_BUFFER_OVERFLOW:
+        return E2BIG;
+    case ERROR_SEEK:
+    case ERROR_SECTOR_NOT_FOUND:
+        return ESPIPE;
+    case ERROR_NOT_DOS_DISK:
+        return ENODEV;
+    case ERROR_GEN_FAILURE:
+        return ENODEV;
+    case ERROR_SHARING_BUFFER_EXCEEDED:
+    case ERROR_NO_MORE_SEARCH_HANDLES:
+        return EMFILE;
+    case ERROR_HANDLE_EOF:
+    case ERROR_BROKEN_PIPE:
+        return EPIPE;
+    case ERROR_HANDLE_DISK_FULL:
+    case ERROR_DISK_FULL:
+        return ENOSPC;
+    case ERROR_NOT_SUPPORTED:
+        return ENOTSUP;
+    case ERROR_FILE_EXISTS:
+    case ERROR_ALREADY_EXISTS:
+    case ERROR_CANNOT_MAKE:
+        return EEXIST;
+    case ERROR_ALREADY_ASSIGNED:
+        return EBUSY;
+    case ERROR_NO_PROC_SLOTS:
+        return EAGAIN;
+    case ERROR_CANT_RESOLVE_FILENAME:
+        return EMLINK;
+    case ERROR_PRIVILEGE_NOT_HELD:
+        return EPERM;
+    case ERROR_ARENA_TRASHED:
+    case ERROR_INVALID_BLOCK:
+    case ERROR_BAD_ENVIRONMENT:
+    case ERROR_BAD_COMMAND:
+    case ERROR_CRC:
+    case ERROR_OUT_OF_PAPER:
+    case ERROR_READ_FAULT:
+    case ERROR_WRITE_FAULT:
+    case ERROR_WRONG_DISK:
+    case ERROR_NET_WRITE_FAULT:
+        return EIO;
+    default: /* not to do with files I expect. */
+        return EIO;
+    }
+}
diff --git a/erts/emulator/pcre/README.pcre_update.md b/erts/emulator/pcre/README.pcre_update.md
index 8caf575d31..599e3d0d12 100644
--- a/erts/emulator/pcre/README.pcre_update.md
+++ b/erts/emulator/pcre/README.pcre_update.md
@@ -2,7 +2,7 @@
 
 ## The basic changes to the PCRE library
 
-To work with the Erlang VM, PCRE has been changed in two important ways:
+To work with the Erlang VM, PCRE has been changed in three important ways:
 
 1. The main execution machine in pcre\_exec has been modified so that
 matching can be interrupted and restarted. This functionality utilizes
diff --git a/erts/emulator/sys/common/erl_check_io.c b/erts/emulator/sys/common/erl_check_io.c
index 799f67fc45..3e77dce1cd 100644
--- a/erts/emulator/sys/common/erl_check_io.c
+++ b/erts/emulator/sys/common/erl_check_io.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  * 
- * Copyright Ericsson AB 2006-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2006-2018. All Rights Reserved.
  * 
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -29,7 +29,6 @@
 #endif
 
 #define ERL_CHECK_IO_C__
-#define ERTS_WANT_BREAK_HANDLING
 #ifndef WANT_NONBLOCKING
 #  define WANT_NONBLOCKING
 #endif
@@ -44,76 +43,79 @@
 #define ERTS_WANT_TIMER_WHEEL_API
 #include "erl_time.h"
 
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-#  include "safe_hash.h"
-#  define DRV_EV_STATE_HTAB_SIZE 1024
+#if 0
+#define DEBUG_PRINT(FMT, ...) erts_printf(FMT "\r\n", ##__VA_ARGS__)
+#define DEBUG_PRINT_FD(FMT, STATE, ...)                                 \
+    DEBUG_PRINT("%d: " FMT " (ev=%s, ac=%s, flg=%d)",                   \
+                (STATE) ? (STATE)->fd : (ErtsSysFdType)-1, ##__VA_ARGS__, \
+                ev2str((STATE) ? (STATE)->events : ERTS_POLL_EV_NONE),  \
+                ev2str((STATE) ? (STATE)->active_events : ERTS_POLL_EV_NONE), \
+                (STATE) ? (STATE)->flags : ERTS_EV_FLAG_CLEAR)
+#define DEBUG_PRINT_MODE
+#else
+#define DEBUG_PRINT(...)
 #endif
 
-typedef char EventStateType;
-#define ERTS_EV_TYPE_NONE     ((EventStateType) 0)
-#define ERTS_EV_TYPE_DRV_SEL  ((EventStateType) 1) /* driver_select */
-#define ERTS_EV_TYPE_DRV_EV   ((EventStateType) 2) /* driver_event */
-#define ERTS_EV_TYPE_STOP_USE ((EventStateType) 3) /* pending stop_select */
-#define ERTS_EV_TYPE_NIF      ((EventStateType) 4) /* enif_select */
-#define ERTS_EV_TYPE_STOP_NIF ((EventStateType) 5) /* pending nif stop */
-
-typedef char EventStateFlags;
-#define ERTS_EV_FLAG_USED         ((EventStateFlags) 1)   /* ERL_DRV_USE has been turned on */
-#define ERTS_EV_FLAG_DEFER_IN_EV  ((EventStateFlags) 2)
-#define ERTS_EV_FLAG_DEFER_OUT_EV ((EventStateFlags) 4)
-
-#ifdef DEBUG
-#  define ERTS_ACTIVE_FD_INC 2
-#else
-#  define ERTS_ACTIVE_FD_INC 128
+#ifndef DEBUG_PRINT_FD
+#define DEBUG_PRINT_FD(...)
 #endif
 
-#define ERTS_CHECK_IO_POLL_RES_LEN 512
+#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
+#  include "safe_hash.h"
+#  define DRV_EV_STATE_HTAB_SIZE 1024
+#endif
 
-#if defined(ERTS_KERNEL_POLL_VERSION)
-#  define ERTS_CIO_EXPORT(FUNC) FUNC ## _kp
-#elif defined(ERTS_NO_KERNEL_POLL_VERSION)
-#  define ERTS_CIO_EXPORT(FUNC) FUNC ## _nkp
+typedef enum {
+    ERTS_EV_TYPE_NONE     = 0,
+    ERTS_EV_TYPE_DRV_SEL  = 1, /* driver_select */
+    ERTS_EV_TYPE_STOP_USE = 2, /* pending stop_select */
+    ERTS_EV_TYPE_NIF      = 3, /* enif_select */
+    ERTS_EV_TYPE_STOP_NIF = 4  /* pending nif stop */
+} EventStateType;
+
+typedef enum {
+    ERTS_EV_FLAG_CLEAR         = 0,
+    ERTS_EV_FLAG_USED          = 1,   /* ERL_DRV_USE has been turned on */
+#ifdef ERTS_ENABLE_KERNEL_POLL
+    ERTS_EV_FLAG_FALLBACK      = 2,   /* Set when kernel poll rejected fd
+                                         and it was put in the nkp version */
 #else
-#  define ERTS_CIO_EXPORT(FUNC) FUNC
+    ERTS_EV_FLAG_FALLBACK      = ERTS_EV_FLAG_CLEAR,
 #endif
 
-#define ERTS_CIO_HAVE_DRV_EVENT \
-  (ERTS_POLL_USE_POLL && !ERTS_POLL_USE_KERNEL_POLL)
+    /* Combinations */
+    ERTS_EV_FLAG_USED_FALLBACK = ERTS_EV_FLAG_USED | ERTS_EV_FLAG_FALLBACK
+} EventStateFlags;
 
-#define ERTS_CIO_POLL_CTL	ERTS_POLL_EXPORT(erts_poll_control)
-#define ERTS_CIO_POLL_CTLV	ERTS_POLL_EXPORT(erts_poll_controlv)
-#define ERTS_CIO_POLL_WAIT	ERTS_POLL_EXPORT(erts_poll_wait)
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#define ERTS_CIO_POLL_AS_INTR 	ERTS_POLL_EXPORT(erts_poll_async_sig_interrupt)
-#endif
-#define ERTS_CIO_POLL_INTR 	ERTS_POLL_EXPORT(erts_poll_interrupt)
-#define ERTS_CIO_POLL_INTR_TMD	ERTS_POLL_EXPORT(erts_poll_interrupt_timed)
-#define ERTS_CIO_NEW_POLLSET 	ERTS_POLL_EXPORT(erts_poll_create_pollset)
-#define ERTS_CIO_FREE_POLLSET	ERTS_POLL_EXPORT(erts_poll_destroy_pollset)
-#define ERTS_CIO_POLL_MAX_FDS	ERTS_POLL_EXPORT(erts_poll_max_fds)
-#define ERTS_CIO_POLL_INIT	ERTS_POLL_EXPORT(erts_poll_init)
-#define ERTS_CIO_POLL_INFO	ERTS_POLL_EXPORT(erts_poll_info)
+#define flag2str(flags)                                                 \
+    ((flags) == ERTS_EV_FLAG_CLEAR ? "CLEAR" :                          \
+     ((flags) == ERTS_EV_FLAG_USED ? "USED" :                           \
+      ((flags) == ERTS_EV_FLAG_FALLBACK ? "FLBK" :                      \
+       ((flags) == ERTS_EV_FLAG_USED_FALLBACK ? "USED|FLBK" : "ERROR"))))
 
-#define GET_FD(fd) fd
+/* How many events that can be handled at once by one erts_poll_wait call */
+#define ERTS_CHECK_IO_POLL_RES_LEN 512
 
-static struct pollset_info
+/* Each I/O Poll Thread has one ErtsPollThread each. The ps field
+   can point to either a private ErtsPollSet or a shared one.
+   At the moment only kqueue and epoll pollsets can be
+   shared across threads.
+*/
+typedef struct erts_poll_thread
 {
-    ErtsPollSet ps;
-    erts_smp_atomic_t in_poll_wait;        /* set while doing poll */
-    struct {
-	int six; /* start index */
-	int eix; /* end index */
-	erts_smp_atomic32_t no;
-	int size;
-	ErtsSysFdType *array;
-    } active_fd;
-#ifdef ERTS_SMP
-    struct removed_fd* removed_list;       /* list of deselected fd's*/
-    erts_smp_spinlock_t removed_list_lock;
-#endif
-}pollset;
-#define NUM_OF_POLLSETS 1
+    ErtsPollSet *ps;
+    ErtsPollResFd *pollres;
+    int pollres_len;
+} ErtsPollThread;
+
+/* pollsetv contains pointers to the ErtsPollSets that are in use.
+ * Which pollset to use is determined by hashing the fd.
+ */
+static ErtsPollSet **pollsetv;
+#if ERTS_POLL_USE_FALLBACK
+static ErtsPollSet *flbk_pollset;
+#endif
+static ErtsPollThread *psiv;
 
 typedef struct {
 #ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
@@ -122,98 +124,151 @@ typedef struct {
     ErtsSysFdType fd;
     struct {
 	ErtsDrvSelectDataState *select;   /* ERTS_EV_TYPE_DRV_SEL */
-#if ERTS_CIO_HAVE_DRV_EVENT
-	ErtsDrvEventDataState *event;     /* ERTS_EV_TYPE_DRV_EV */
-#endif
         ErtsNifSelectDataState *nif;      /* ERTS_EV_TYPE_NIF */
         union {
             erts_driver_t*  drv_ptr;    /* ERTS_EV_TYPE_STOP_USE */
             ErtsResource* resource;   /* ERTS_EV_TYPE_STOP_NIF */
-        }stop;
+        } stop;
     } driver;
-    ErtsPollEvents events;
-    unsigned short remove_cnt; /* number of removed_fd's referring to this fd */
+    ErtsPollEvents events; /* The events that have been selected upon */
+    ErtsPollEvents active_events; /* The events currently active in the pollset */
     EventStateType type;
     EventStateFlags flags;
 } ErtsDrvEventState;
 
-#ifdef ERTS_SMP
-struct removed_fd {
-    struct removed_fd *next;
+struct drv_ev_state_shared {
+
+    union {
+        erts_mtx_t lck;
+        byte _cache_line_alignment[ERTS_ALC_CACHE_LINE_ALIGN_SIZE(sizeof(erts_mtx_t))];
+    } locks[ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT];
+
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    ErtsSysFdType fd;
+    int max_fds;
+    erts_atomic_t len;
+    ErtsDrvEventState *v;
+    erts_mtx_t grow_lock; /* prevent lock-hogging of racing growers */
 #else
-    ErtsDrvEventState* state;
-    #ifdef DEBUG
-    ErtsSysFdType fd;
-    #endif
+    SafeHash tab;
+    int num_prealloc;
+    ErtsDrvEventState *prealloc_first;
+    erts_spinlock_t prealloc_lock;
 #endif
-
 };
-#endif
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-static int max_fds = -1;
-#endif
-#define DRV_EV_STATE_LOCK_CNT 16
-static union {
-    erts_smp_mtx_t lck;
-    byte _cache_line_alignment[64];
-}drv_ev_state_locks[DRV_EV_STATE_LOCK_CNT];
+int ERTS_WRITE_UNLIKELY(erts_no_pollsets) = 1;
+int ERTS_WRITE_UNLIKELY(erts_no_poll_threads) = 1;
+struct drv_ev_state_shared drv_ev_state;
 
-#ifdef ERTS_SMP
-static ERTS_INLINE erts_smp_mtx_t* fd_mtx(ErtsSysFdType fd)
-{
+static ERTS_INLINE int fd_hash(ErtsSysFdType fd) {
     int hash = (int)fd;
 # ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
     hash ^= (hash >> 9);
 # endif
-    return &drv_ev_state_locks[hash % DRV_EV_STATE_LOCK_CNT].lck;
+    return hash;
+}
+
+static ERTS_INLINE erts_mtx_t* fd_mtx(ErtsSysFdType fd)
+{
+    return &drv_ev_state.locks[fd_hash(fd) % ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT].lck;
 }
-#else
-#  define fd_mtx(fd) NULL
-#endif
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
 
-static erts_smp_atomic_t drv_ev_state_len;
-static ErtsDrvEventState *drv_ev_state;
-static erts_smp_mtx_t drv_ev_state_grow_lock; /* prevent lock-hogging of racing growers */
+static ERTS_INLINE ErtsDrvEventState *get_drv_ev_state(ErtsSysFdType fd)
+{
+    return &drv_ev_state.v[(int) fd];
+}
+
+#define new_drv_ev_state(State, fd) (State)
+#define erase_drv_ev_state(State)
 
-#else
-static SafeHash drv_ev_state_tab;
-static int num_state_prealloc;
-static ErtsDrvEventState *state_prealloc_first;
-erts_smp_spinlock_t state_prealloc_lock;
+static ERTS_INLINE int grow_drv_ev_state(ErtsSysFdType fd) {
+    int i;
+    int old_len;
+    int new_len;
+
+    if ((unsigned)fd >= (unsigned)erts_atomic_read_nob(&drv_ev_state.len)) {
 
-static ERTS_INLINE ErtsDrvEventState *hash_get_drv_ev_state(ErtsSysFdType fd)
+	if (fd < 0 || fd >= drv_ev_state.max_fds)
+            return 0;
+
+        erts_mtx_lock(&drv_ev_state.grow_lock);
+        old_len = erts_atomic_read_nob(&drv_ev_state.len);
+        if (fd >= old_len) {
+            new_len = erts_poll_new_table_len(old_len, fd + 1);
+            if (new_len > drv_ev_state.max_fds)
+                new_len = drv_ev_state.max_fds;
+
+            for (i=0; i<ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT; i++) { /* lock all fd's */
+                erts_mtx_lock(&drv_ev_state.locks[i].lck);
+            }
+            drv_ev_state.v = (drv_ev_state.v
+                              ? erts_realloc(ERTS_ALC_T_DRV_EV_STATE,
+                                             drv_ev_state.v,
+                                             sizeof(ErtsDrvEventState)*new_len)
+                              : erts_alloc(ERTS_ALC_T_DRV_EV_STATE,
+                                           sizeof(ErtsDrvEventState)*new_len));
+            ERTS_CT_ASSERT(ERTS_EV_TYPE_NONE == 0);
+            sys_memzero(drv_ev_state.v+old_len,
+                        sizeof(ErtsDrvEventState) * (new_len - old_len));
+            for (i = old_len; i < new_len; i++) {
+                drv_ev_state.v[i].fd = (ErtsSysFdType) i;
+            }
+            erts_atomic_set_nob(&drv_ev_state.len, new_len);
+            for (i=0; i<ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT; i++) {
+                erts_mtx_unlock(&drv_ev_state.locks[i].lck);
+            }
+        }
+        /*else already grown by racing thread */
+
+        erts_mtx_unlock(&drv_ev_state.grow_lock);
+    }
+    return 1;
+}
+
+static int drv_ev_state_len(void)
+{
+    return erts_atomic_read_nob(&drv_ev_state.len);
+}
+
+#else /* !ERTS_SYS_CONTINOUS_FD_NUMBERS */
+
+static ERTS_INLINE ErtsDrvEventState *get_drv_ev_state(ErtsSysFdType fd)
 {
     ErtsDrvEventState tmpl;
     tmpl.fd = fd;
-    return  (ErtsDrvEventState *) safe_hash_get(&drv_ev_state_tab, (void *) &tmpl);
+    return  (ErtsDrvEventState *) safe_hash_get(&drv_ev_state.tab, (void *) &tmpl);
 }
 
-static ERTS_INLINE ErtsDrvEventState* hash_new_drv_ev_state(ErtsSysFdType fd)
+static ERTS_INLINE ErtsDrvEventState* new_drv_ev_state(ErtsDrvEventState *state,
+                                                       ErtsSysFdType fd)
 {
     ErtsDrvEventState tmpl;
+
+    if (state)
+        return state;
+
     tmpl.fd = fd;
     tmpl.driver.select = NULL;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    tmpl.driver.event = NULL;
-#endif
     tmpl.driver.nif = NULL;
     tmpl.driver.stop.drv_ptr = NULL;
     tmpl.events = 0;
-    tmpl.remove_cnt = 0;
+    tmpl.active_events = 0;
     tmpl.type = ERTS_EV_TYPE_NONE;
     tmpl.flags = 0;
-    return  (ErtsDrvEventState *) safe_hash_put(&drv_ev_state_tab, (void *) &tmpl);
+
+    return  (ErtsDrvEventState *) safe_hash_put(&drv_ev_state.tab, (void *) &tmpl);
+}
+
+static ERTS_INLINE void erase_drv_ev_state(ErtsDrvEventState *state)
+{
+    safe_hash_erase(&drv_ev_state.tab, (void *) state);
 }
 
-static ERTS_INLINE void hash_erase_drv_ev_state(ErtsDrvEventState *state)
+static int drv_ev_state_len(void)
 {
-    ASSERT(state->remove_cnt == 0);
-    safe_hash_erase(&drv_ev_state_tab, (void *) state);
+    return erts_atomic_read_nob(&drv_ev_state.tab.nitems);
 }
 
 #endif /* !ERTS_SYS_CONTINOUS_FD_NUMBERS */
@@ -233,52 +288,47 @@ static void print_nif_select_op(erts_dsprintf_buf_t*, ErtsSysFdType,
 static void drv_select_large_fd_error(ErlDrvPort, ErtsSysFdType, int, int);
 static void nif_select_large_fd_error(ErtsSysFdType, int, ErtsResource*,Eterm ref);
 #endif
-#if ERTS_CIO_HAVE_DRV_EVENT
-static void drv_event_steal(ErlDrvPort ix, ErtsDrvEventState *state,
-			ErlDrvEventData event_data);
-static void print_drv_event_op(erts_dsprintf_buf_t *dsbufp,
-                               ErlDrvPort, ErtsSysFdType, ErlDrvEventData);
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-static void event_large_fd_error(ErlDrvPort, ErtsSysFdType, ErlDrvEventData);
-#endif
-#endif
 static void
 steal_pending_stop_use(erts_dsprintf_buf_t*, ErlDrvPort, ErtsDrvEventState*,
                        int mode, int on);
 static void
 steal_pending_stop_nif(erts_dsprintf_buf_t *dsbufp, ErtsResource*,
                        ErtsDrvEventState *state, int mode, int on);
-
-#ifdef ERTS_SMP
-ERTS_SCHED_PREF_QUICK_ALLOC_IMPL(removed_fd, struct removed_fd, 64, ERTS_ALC_T_FD_LIST)
+static ERTS_INLINE void
+check_fd_cleanup(ErtsDrvEventState *state,
+		 ErtsDrvSelectDataState **free_select,
+                 ErtsNifSelectDataState **free_nif);
+static ERTS_INLINE void iready(Eterm id, ErtsDrvEventState *state);
+static ERTS_INLINE void oready(Eterm id, ErtsDrvEventState *state);
+#ifdef DEBUG_PRINT_MODE
+static char *drvmode2str(int mode);
+static char *nifmode2str(enum ErlNifSelectFlags mode);
 #endif
 
 static ERTS_INLINE void
-init_iotask(ErtsIoTask *io_task)
+init_iotask(ErtsIoTask *io_task, ErtsSysFdType fd)
 {
     erts_port_task_handle_init(&io_task->task);
-    erts_smp_atomic_init_nob(&io_task->executed_time, ~((erts_aint_t) 0));
+    io_task->fd = fd;
 }
 
 static ERTS_INLINE int
-is_iotask_active(ErtsIoTask *io_task, erts_aint_t current_cio_time)
-{    
+is_iotask_active(ErtsIoTask *io_task)
+{
     if (erts_port_task_is_scheduled(&io_task->task))
 	return 1;
-    if (erts_smp_atomic_read_nob(&io_task->executed_time) == current_cio_time)
-	return 1;
     return 0;
 }
 
 static ERTS_INLINE ErtsDrvSelectDataState *
-alloc_drv_select_data(void)
+alloc_drv_select_data(ErtsSysFdType fd)
 {
     ErtsDrvSelectDataState *dsp = erts_alloc(ERTS_ALC_T_DRV_SEL_D_STATE,
 					     sizeof(ErtsDrvSelectDataState));
     dsp->inport = NIL;
     dsp->outport = NIL;
-    init_iotask(&dsp->iniotask);
-    init_iotask(&dsp->outiotask);
+    init_iotask(&dsp->iniotask, fd);
+    init_iotask(&dsp->outiotask, fd);
     return dsp;
 }
 
@@ -289,8 +339,6 @@ alloc_nif_select_data(void)
 					     sizeof(ErtsNifSelectDataState));
     dsp->in.pid = NIL;
     dsp->out.pid = NIL;
-    dsp->in.ddeselect_cnt = 0;
-    dsp->out.ddeselect_cnt = 0;
     return dsp;
 }
 
@@ -308,209 +356,136 @@ free_nif_select_data(ErtsNifSelectDataState *dsp)
     erts_free(ERTS_ALC_T_NIF_SEL_D_STATE, dsp);
 }
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-
-static ERTS_INLINE ErtsDrvEventDataState *
-alloc_drv_event_data(void)
+static ERTS_INLINE int
+get_pollset_id(ErtsSysFdType fd)
 {
-    ErtsDrvEventDataState *dep = erts_alloc(ERTS_ALC_T_DRV_EV_D_STATE,
-					    sizeof(ErtsDrvEventDataState));
-    dep->port = NIL;
-    dep->data = NULL;
-    dep->removed_events = 0;
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    dep->deferred_events = 0;
-#endif
-    init_iotask(&dep->iotask);
-    return dep;
+    return fd_hash(fd) % erts_no_pollsets;
 }
 
-static ERTS_INLINE void
-free_drv_event_data(ErtsDrvEventDataState *dep)
+static ERTS_INLINE ErtsPollSet *
+get_pollset(ErtsSysFdType fd)
 {
-    ASSERT(!erts_port_task_is_scheduled(&dep->iotask.task));
-    erts_free(ERTS_ALC_T_DRV_EV_D_STATE, dep);   
+    return pollsetv[get_pollset_id(fd)];
 }
 
-#endif /* ERTS_CIO_HAVE_DRV_EVENT */
-
-static ERTS_INLINE void
-remember_removed(ErtsDrvEventState *state, struct pollset_info* psi)
+#if ERTS_POLL_USE_FALLBACK
+static ERTS_INLINE ErtsPollSet *
+get_fallback(void)
 {
-#ifdef ERTS_SMP
-    struct removed_fd *fdlp;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(fd_mtx(state->fd)));
-    if (erts_smp_atomic_read_nob(&psi->in_poll_wait)) {
-	state->remove_cnt++;
-	ASSERT(state->remove_cnt > 0);
-	fdlp = removed_fd_alloc();
-    #if defined(ERTS_SYS_CONTINOUS_FD_NUMBERS) || defined(DEBUG)
-    	fdlp->fd = state->fd;
-    #endif
-    #ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	fdlp->state = state;
-    #endif
-	erts_smp_spin_lock(&psi->removed_list_lock);
-	fdlp->next = psi->removed_list;
-	psi->removed_list = fdlp;
-	erts_smp_spin_unlock(&psi->removed_list_lock);
-    }
-#endif
+    return flbk_pollset;
 }
+#endif
 
-
-static ERTS_INLINE int
-is_removed(ErtsDrvEventState *state)
+/*
+ * Place a fd within a pollset. This will automatically use
+ * the fallback ps if needed.
+ */
+static ERTS_INLINE ErtsPollEvents
+erts_io_control_wakeup(ErtsDrvEventState *state, ErtsPollOp op,
+                       ErtsPollEvents pe, int *wake_poller)
 {
-#ifdef ERTS_SMP
-    /* Note that there is a possible race here, where an fd is removed
-       (increasing remove_cnt) and then added again just before erts_poll_wait
-       is called by erts_check_io. Any polled event on the re-added fd will then
-       be falsely ignored. But that does not matter, as the event will trigger
-       again next time erl_check_io is called. */
-    return state->remove_cnt > 0;
-#else
-    return 0;
+    ErtsSysFdType fd = state->fd;
+    ErtsPollEvents res = 0;
+    EventStateFlags flags = state->flags;
+
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(fd_mtx(state->fd)));
+
+    if (!(flags & ERTS_EV_FLAG_FALLBACK)) {
+        res = erts_poll_control(get_pollset(fd), fd, op, pe, wake_poller);
+
+#if ERTS_POLL_USE_FALLBACK
+        if (op == ERTS_POLL_OP_ADD && res == ERTS_POLL_EV_NVAL) {
+            /* When an add fails with NVAL, the poll/kevent operation could not
+               put that fd in the pollset, so we instead put it into a fallback pollset */
+            state->flags |= ERTS_EV_FLAG_FALLBACK;
+            res = erts_poll_control_flbk(get_fallback(), fd, op, pe, wake_poller);
+        }
+    } else {
+        ASSERT(op != ERTS_POLL_OP_ADD);
+        res = erts_poll_control_flbk(get_fallback(), fd, op, pe, wake_poller);
 #endif
+    }
+
+    return res;
 }
 
-static void
-forget_removed(struct pollset_info* psi)
+static ERTS_INLINE ErtsPollEvents
+erts_io_control(ErtsDrvEventState *state, ErtsPollOp op, ErtsPollEvents pe)
 {
-#ifdef ERTS_SMP
-    struct removed_fd* fdlp;
-    struct removed_fd* tofree;
+    int wake_poller = 0;
+    return erts_io_control_wakeup(state, op, pe, &wake_poller);
+}
 
-    /* Fast track: if (atomic_ptr(removed_list)==NULL) return; */
+/* ToDo: Was inline in erl_check_io.h but now need struct erts_poll_thread */
+void
+erts_io_notify_port_task_executed(ErtsPortTaskType type,
+                                  ErtsPortTaskHandle *pthp,
+                                  void (*reset_handle)(ErtsPortTaskHandle *))
+{
+    ErtsIoTask *itp = ErtsContainerStruct(pthp, ErtsIoTask, task);
+    ErtsSysFdType fd = itp->fd;
+    erts_mtx_t *mtx = fd_mtx(fd);
+    int active_events;
+    ErtsDrvEventState *state;
+    ErtsDrvSelectDataState *free_select = NULL;
+    ErtsNifSelectDataState *free_nif = NULL;
 
-    erts_smp_spin_lock(&psi->removed_list_lock);
-    fdlp = psi->removed_list;
-    psi->removed_list = NULL;
-    erts_smp_spin_unlock(&psi->removed_list_lock);
+    erts_mtx_lock(mtx);
+    state = get_drv_ev_state(fd);
 
-    while (fdlp) {
-        ErtsResource* resource = NULL;
-	erts_driver_t* drv_ptr = NULL;
-	erts_smp_mtx_t* mtx;
-	ErtsSysFdType fd;
-	ErtsDrvEventState *state;
+    active_events = state->active_events;
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	fd = fdlp->fd;
-	mtx = fd_mtx(fd);
-	erts_smp_mtx_lock(mtx);
-	state = &drv_ev_state[(int) fd];
-#else
-	state = fdlp->state;
-	fd = state->fd;
-	ASSERT(fd == fdlp->fd);
-	mtx = fd_mtx(fd);
-	erts_smp_mtx_lock(mtx);
-#endif
-	ASSERT(state->remove_cnt > 0);
-	if (--state->remove_cnt == 0) {
-	    switch (state->type) {
-	    case ERTS_EV_TYPE_STOP_NIF:
-		/* Now we can call stop */
-		resource = state->driver.stop.resource;
-                state->driver.stop.resource = NULL;
-                ASSERT(resource);
-		state->type = ERTS_EV_TYPE_NONE;
-		state->flags &= ~ERTS_EV_FLAG_USED;
-                goto case_ERTS_EV_TYPE_NONE;
-
-	    case ERTS_EV_TYPE_STOP_USE:
-		/* Now we can call stop_select */
-		drv_ptr = state->driver.stop.drv_ptr;
-		ASSERT(drv_ptr);
-		state->type = ERTS_EV_TYPE_NONE;
-		state->flags &= ~ERTS_EV_FLAG_USED;
-		state->driver.stop.drv_ptr = NULL;
-		/* Fall through */
-            case ERTS_EV_TYPE_NONE:
-            case_ERTS_EV_TYPE_NONE:
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-		hash_erase_drv_ev_state(state);
-#endif
-		break;
-	    case ERTS_EV_TYPE_DRV_SEL:
-	    case ERTS_EV_TYPE_DRV_EV:
-		break;
-	    default:
-		ASSERT(0);
-	    }
-	}
-	erts_smp_mtx_unlock(mtx);
-	if (drv_ptr) {
-	    int was_unmasked = erts_block_fpe();
-	    DTRACE1(driver_stop_select, drv_ptr->name);
-	    LTTNG1(driver_stop_select, drv_ptr->name);
-	    (*drv_ptr->stop_select) ((ErlDrvEvent) fd, NULL);
-	    erts_unblock_fpe(was_unmasked);
-	    if (drv_ptr->handle) {
-		erts_ddll_dereference_driver(drv_ptr->handle);
-	    }
-	}
-        if (resource) {
-            erts_resource_stop(resource, (ErlNifEvent)fd, 0);
-            enif_release_resource(resource->data);
-        }
+    switch (type) {
+    case ERTS_PORT_TASK_INPUT:
 
-	tofree = fdlp;
-	fdlp = fdlp->next;
-	removed_fd_free(tofree);
+        DEBUG_PRINT_FD("executed ready_input", state);
+
+        ASSERT(!(state->active_events & ERTS_POLL_EV_IN));
+        if (state->events & ERTS_POLL_EV_IN)
+            active_events |= ERTS_POLL_EV_IN;
+        break;
+    case ERTS_PORT_TASK_OUTPUT:
+
+        DEBUG_PRINT_FD("executed ready_output", state);
+
+        ASSERT(!(state->active_events & ERTS_POLL_EV_OUT));
+        if (state->events & ERTS_POLL_EV_OUT)
+            active_events |= ERTS_POLL_EV_OUT;
+        break;
+    default:
+        erts_exit(ERTS_ABORT_EXIT, "Invalid IO port task type");
+        break;
     }
-#endif /* ERTS_SMP */
-}
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-static void
-grow_drv_ev_state(int min_ix)
-{
-    int i;
-    int old_len;
-    int new_len;
+    reset_handle(pthp);
 
-    erts_smp_mtx_lock(&drv_ev_state_grow_lock);
-    old_len = erts_smp_atomic_read_nob(&drv_ev_state_len);
-    if (min_ix >= old_len) {
-        new_len = erts_poll_new_table_len(old_len, min_ix + 1);
-        if (new_len > max_fds)
-            new_len = max_fds;
+    if (active_events) {
+        /* This is not needed if active_events has not changed */
+        if (state->active_events != active_events) {
+            ErtsPollEvents new_events;
+            state->active_events = active_events;
+            new_events = erts_io_control(state, ERTS_POLL_OP_MOD, active_events);
 
-	for (i=0; i<DRV_EV_STATE_LOCK_CNT; i++) { /* lock all fd's */
-	    erts_smp_mtx_lock(&drv_ev_state_locks[i].lck);
-	}
-	drv_ev_state = (drv_ev_state
-			? erts_realloc(ERTS_ALC_T_DRV_EV_STATE,
-				       drv_ev_state,
-				       sizeof(ErtsDrvEventState)*new_len)
-			: erts_alloc(ERTS_ALC_T_DRV_EV_STATE,
-				     sizeof(ErtsDrvEventState)*new_len));
-	for (i = old_len; i < new_len; i++) {
-	    drv_ev_state[i].fd = (ErtsSysFdType) i;
-	    drv_ev_state[i].driver.select = NULL;
-#if ERTS_CIO_HAVE_DRV_EVENT
-	    drv_ev_state[i].driver.event = NULL;
-#endif
-	    drv_ev_state[i].driver.stop.drv_ptr = NULL;
-            drv_ev_state[i].driver.nif = NULL;
-	    drv_ev_state[i].events = 0;
-	    drv_ev_state[i].remove_cnt = 0;
-	    drv_ev_state[i].type = ERTS_EV_TYPE_NONE;
-	    drv_ev_state[i].flags = 0;
-	}
-	erts_smp_atomic_set_nob(&drv_ev_state_len, new_len);
-	for (i=0; i<DRV_EV_STATE_LOCK_CNT; i++) {
-	    erts_smp_mtx_unlock(&drv_ev_state_locks[i].lck);
-	}
+            /* We were unable to re-insert the fd into the pollset, signal the callback. */
+            if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
+                if (active_events & ERTS_POLL_EV_IN)
+                    iready(state->driver.select->inport, state);
+                if (active_events & ERTS_POLL_EV_OUT)
+                    oready(state->driver.select->outport, state);
+                state->active_events = 0;
+            }
+        }
+    } else {
+        check_fd_cleanup(state, &free_select, &free_nif);
     }
-    /*else already grown by racing thread */
 
-    erts_smp_mtx_unlock(&drv_ev_state_grow_lock);
-}
-#endif /* ERTS_SYS_CONTINOUS_FD_NUMBERS */
+    erts_mtx_unlock(mtx);
 
+    if (free_select)
+        free_drv_select_data(free_select);
+    if (free_nif)
+        free_nif_select_data(free_nif);
+}
 
 static ERTS_INLINE void
 abort_task(Eterm id, ErtsPortTaskHandle *pthp, EventStateType type)
@@ -527,13 +502,6 @@ abort_tasks(ErtsDrvEventState *state, int mode)
     switch (mode) {
     case 0: check_type:
 	switch (state->type) {
-#if ERTS_CIO_HAVE_DRV_EVENT
-	case ERTS_EV_TYPE_DRV_EV:
-	    abort_task(state->driver.event->port,
-		       &state->driver.event->iotask.task,
-		       ERTS_EV_TYPE_DRV_EV);
-	    return;
-#endif
         case ERTS_EV_TYPE_NIF:
 	case ERTS_EV_TYPE_NONE:
 	    return;
@@ -563,16 +531,14 @@ abort_tasks(ErtsDrvEventState *state, int mode)
 static void
 deselect(ErtsDrvEventState *state, int mode)
 {
-    int do_wake = 0;
     ErtsPollEvents rm_events;
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(fd_mtx(state->fd)));
-    ASSERT(state->events);
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(fd_mtx(state->fd)));
 
     abort_tasks(state, mode);
 
-    if (!mode)
+    if (!mode) {
 	rm_events = state->events;
-    else {
+    } else {
 	rm_events = 0;
 	ASSERT(state->type == ERTS_EV_TYPE_DRV_SEL);
 	if (mode & ERL_DRV_READ) {
@@ -585,67 +551,63 @@ deselect(ErtsDrvEventState *state, int mode)
 	}
     }
 
-    state->events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd, rm_events, 0, &do_wake);
+    state->events &= ~rm_events;
+    state->active_events &= ~rm_events;
 
     if (!(state->events)) {
+        erts_io_control(state, ERTS_POLL_OP_DEL, 0);
 	switch (state->type) {
         case ERTS_EV_TYPE_NIF:
             state->driver.nif->in.pid = NIL;
             state->driver.nif->out.pid = NIL;
-            state->driver.nif->in.ddeselect_cnt = 0;
-            state->driver.nif->out.ddeselect_cnt = 0;
-            enif_release_resource(state->driver.stop.resource);
+            enif_release_resource(state->driver.stop.resource->data);
             state->driver.stop.resource = NULL;
             break;
 	case ERTS_EV_TYPE_DRV_SEL:
 	    state->driver.select->inport = NIL;
 	    state->driver.select->outport = NIL;
 	    break;
-#if ERTS_CIO_HAVE_DRV_EVENT
-	case ERTS_EV_TYPE_DRV_EV:
-	    state->driver.event->port = NIL;
-	    state->driver.event->data = NULL;
-	    state->driver.event->removed_events = (ErtsPollEvents) 0;
-	    break;
-#endif
 	case ERTS_EV_TYPE_NONE:
 	    break;
 	default:
 	    ASSERT(0);
 	    break;
 	}
-	    
 	state->type = ERTS_EV_TYPE_NONE;
-	state->flags &= ~ERTS_EV_FLAG_USED;
-	remember_removed(state, &pollset);
+	state->flags = 0;
+    } else {
+        ErtsPollEvents new_events =
+            erts_io_control(state, ERTS_POLL_OP_MOD, state->active_events);
+
+        /* We were unable to re-insert the fd into the pollset, signal the callback. */
+        if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
+            if (state->active_events & ERTS_POLL_EV_IN)
+                iready(state->driver.select->inport, state);
+            if (state->active_events & ERTS_POLL_EV_OUT)
+                oready(state->driver.select->outport, state);
+            state->active_events = 0;
+        }
     }
 }
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-#  define IS_FD_UNKNOWN(state) ((state)->type == ERTS_EV_TYPE_NONE && (state)->remove_cnt == 0)
+#  define IS_FD_UNKNOWN(state) ((state)->type == ERTS_EV_TYPE_NONE)
 #else
 #  define IS_FD_UNKNOWN(state) ((state) == NULL)
 #endif
 
 static ERTS_INLINE void
 check_fd_cleanup(ErtsDrvEventState *state,
-#if ERTS_CIO_HAVE_DRV_EVENT
-		 ErtsDrvEventDataState **free_event,
-#endif
 		 ErtsDrvSelectDataState **free_select,
                  ErtsNifSelectDataState **free_nif)
 {
-    erts_aint_t current_cio_time;
-
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_mtx_is_locked(fd_mtx(state->fd)));
-
-    current_cio_time = erts_smp_atomic_read_acqb(&erts_check_io_time);
+    ERTS_LC_ASSERT(erts_lc_mtx_is_locked(fd_mtx(state->fd)));
     *free_select = NULL;
     if (state->driver.select
 	&& (state->type != ERTS_EV_TYPE_DRV_SEL)
-	&& !is_iotask_active(&state->driver.select->iniotask, current_cio_time)
-	&& !is_iotask_active(&state->driver.select->outiotask, current_cio_time)) {
-	
+	&& !is_iotask_active(&state->driver.select->iniotask)
+	&& !is_iotask_active(&state->driver.select->outiotask)) {
+
 	*free_select = state->driver.select;
 	state->driver.select = NULL;
     }
@@ -656,382 +618,85 @@ check_fd_cleanup(ErtsDrvEventState *state,
         state->driver.nif = NULL;
     }
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-    *free_event = NULL;
-    if (state->driver.event
-	&& (state->type != ERTS_EV_TYPE_DRV_EV)
-	&& !is_iotask_active(&state->driver.event->iotask,  current_cio_time)) {
-	
-	*free_event = state->driver.event;
-	state->driver.event = NULL;
-    }
-#endif
-
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
     if (((state->type != ERTS_EV_TYPE_NONE)
-	 | state->remove_cnt
-#if ERTS_CIO_HAVE_DRV_EVENT
-	 | (state->driver.event != NULL)
-#endif
+         | (state->driver.nif != NULL)
 	 | (state->driver.select != NULL)) == 0) {
 
-	hash_erase_drv_ev_state(state);
-
+	erase_drv_ev_state(state);
     }
-#endif
 }
 
-static ERTS_INLINE int
-check_cleanup_active_fd(ErtsSysFdType fd,
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-			ErtsPollControlEntry *pce,
-			int *pce_ix,
-#endif
-			erts_aint_t current_cio_time,
-                        int may_sleep)
-{
-    ErtsDrvEventState *state;
-    int active = 0;
-    erts_smp_mtx_t *mtx = fd_mtx(fd);
-    void *free_select = NULL;
-    void *free_nif = NULL;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    void *free_event = NULL;
-#endif
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    ErtsPollEvents evon = 0, evoff = 0;
-#endif
-
-    erts_smp_mtx_lock(mtx);
-
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    state = &drv_ev_state[(int) fd];
+#ifdef __WIN32__
+# define MUST_DEFER(MAY_SLEEP) 1
 #else
-    state = hash_get_drv_ev_state(fd); /* may be NULL! */
-    if (state)
-#endif
-    {
-	if (state->driver.select) {
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-	    if (is_iotask_active(&state->driver.select->iniotask, current_cio_time)) {
-		active = 1;
-		if ((state->events & ERTS_POLL_EV_IN)
-		    && !(state->flags & ERTS_EV_FLAG_DEFER_IN_EV)) {
-		    evoff |= ERTS_POLL_EV_IN;
-		    state->flags |= ERTS_EV_FLAG_DEFER_IN_EV;
-		}
-	    }
-	    else if (state->flags & ERTS_EV_FLAG_DEFER_IN_EV) {
-		if (state->events & ERTS_POLL_EV_IN)
-		    evon |= ERTS_POLL_EV_IN;
-		state->flags &= ~ERTS_EV_FLAG_DEFER_IN_EV;
-	    }
-	    if (is_iotask_active(&state->driver.select->outiotask, current_cio_time)) {
-		active = 1;
-		if ((state->events & ERTS_POLL_EV_OUT)
-		    && !(state->flags & ERTS_EV_FLAG_DEFER_OUT_EV)) {
-		    evoff |= ERTS_POLL_EV_OUT;
-		    state->flags |= ERTS_EV_FLAG_DEFER_OUT_EV;
-		}
-	    }
-	    else if (state->flags & ERTS_EV_FLAG_DEFER_OUT_EV) {
-		if (state->events & ERTS_POLL_EV_OUT)
-		    evon |= ERTS_POLL_EV_OUT;
-		state->flags &= ~ERTS_EV_FLAG_DEFER_OUT_EV;
-	    }
-	    if (active)
-		(void) 0;
-	    else
-#else
-	    if (is_iotask_active(&state->driver.select->iniotask, current_cio_time)
-		|| is_iotask_active(&state->driver.select->outiotask, current_cio_time))
-		active = 1;
-	    else
-#endif
-		if (state->type != ERTS_EV_TYPE_DRV_SEL) {
-		free_select = state->driver.select;
-		state->driver.select = NULL;
-	    }
-	}
-
-        if (state->driver.nif) {
-            ErtsPollEvents rm_events = 0;
-            if (state->driver.nif->in.ddeselect_cnt) {
-                ASSERT(state->type == ERTS_EV_TYPE_NIF);
-                ASSERT(state->events & ERTS_POLL_EV_IN);
-                ASSERT(is_nil(state->driver.nif->in.pid));
-                if (may_sleep || state->driver.nif->in.ddeselect_cnt == 1) {
-                    rm_events = ERTS_POLL_EV_IN;
-                    state->driver.nif->in.ddeselect_cnt = 0;
-                }
-            }
-            if (state->driver.nif->out.ddeselect_cnt) {
-                ASSERT(state->type == ERTS_EV_TYPE_NIF);
-                ASSERT(state->events & ERTS_POLL_EV_OUT);
-                ASSERT(is_nil(state->driver.nif->out.pid));
-                if (may_sleep || state->driver.nif->out.ddeselect_cnt == 1) {
-                    rm_events |= ERTS_POLL_EV_OUT;
-                    state->driver.nif->out.ddeselect_cnt = 0;
-                }
-            }
-            if (rm_events) {
-                int do_wake = 0;
-                state->events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd,
-                                                  rm_events, 0, &do_wake);
-            }
-            if (state->events)
-                active = 1;
-            else if (state->type != ERTS_EV_TYPE_NIF) {
-                free_nif = state->driver.nif;
-                state->driver.nif = NULL;
-            }
-        }
-
-#if ERTS_CIO_HAVE_DRV_EVENT
-	if (state->driver.event) {
-	    if (is_iotask_active(&state->driver.event->iotask, current_cio_time)) {
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-		ErtsPollEvents evs = state->events & ~state->driver.event->deferred_events;
-		if (evs) {
-		    evoff |= evs;
-		    state->driver.event->deferred_events |= evs;
-		}
-#endif
-		active = 1;
-	    }
-	    else if (state->type != ERTS_EV_TYPE_DRV_EV) {
-		free_event = state->driver.event;
-		state->driver.event = NULL;
-	    }
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-	    else {
-		ErtsPollEvents evs = state->events & state->driver.event->deferred_events;
-		if (evs) {
-		    evon |= evs;
-		    state->driver.event->deferred_events = 0;
-		}
-	    }
-#endif
-
-	}
-#endif
-
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	if (((state->type != ERTS_EV_TYPE_NONE) | state->remove_cnt | active) == 0)
-	    hash_erase_drv_ev_state(state);
-#endif
-
-    }
-
-    erts_smp_mtx_unlock(mtx);
-
-    if (free_select)
-	free_drv_select_data(free_select);
-    if (free_nif)
-        free_nif_select_data(free_nif);
-#if ERTS_CIO_HAVE_DRV_EVENT
-    if (free_event)
-	free_drv_event_data(free_event);
-#endif
-
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    if (evoff) {
-	ErtsPollControlEntry *pcep = &pce[(*pce_ix)++];
-	pcep->fd = fd;
-	pcep->events = evoff;
-	pcep->on = 0;
-    }
-    if (evon) {
-	ErtsPollControlEntry *pcep = &pce[(*pce_ix)++];
-	pcep->fd = fd;
-	pcep->events = evon;
-	pcep->on = 1;
-    }
-#endif
-
-    return active;
-}
-
-static void
-check_cleanup_active_fds(erts_aint_t current_cio_time, int may_sleep)
-{
-    int six = pollset.active_fd.six;
-    int eix = pollset.active_fd.eix;
-    erts_aint32_t no = erts_smp_atomic32_read_dirty(&pollset.active_fd.no);
-    int size = pollset.active_fd.size;
-    int ix = six;
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    /* every fd might add two entries */
-    Uint pce_sz = 2*sizeof(ErtsPollControlEntry)*no;
-    ErtsPollControlEntry *pctrl_entries = (pce_sz
-					   ? erts_alloc(ERTS_ALC_T_TMP, pce_sz)
-					   : NULL);
-    int pctrl_ix = 0;
-#endif
-
-    while (ix != eix) {
-	ErtsSysFdType fd = pollset.active_fd.array[ix];
-	int nix = ix + 1;
-	if (nix >= size)
-	    nix = 0;
-	ASSERT(fd != ERTS_SYS_FD_INVALID);
-	if (!check_cleanup_active_fd(fd,
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-				     pctrl_entries,
-				     &pctrl_ix,
-#endif
-				     current_cio_time,
-                                     may_sleep)) {
-	    no--;
-	    if (ix == six) {
-#ifdef DEBUG
-		pollset.active_fd.array[ix] = ERTS_SYS_FD_INVALID;
-#endif
-		six = nix;
-	    }
-	    else {
-		pollset.active_fd.array[ix] = pollset.active_fd.array[six];
-#ifdef DEBUG
-		pollset.active_fd.array[six] = ERTS_SYS_FD_INVALID;
-#endif
-		six++;
-		if (six >= size)
-		    six = 0;
-	    }
-	}
-	ix = nix;
-    }
-
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    ASSERT(pctrl_ix <= pce_sz/sizeof(ErtsPollControlEntry));
-    if (pctrl_ix)
-	ERTS_CIO_POLL_CTLV(pollset.ps, pctrl_entries, pctrl_ix);
-    if (pctrl_entries)
-	erts_free(ERTS_ALC_T_TMP, pctrl_entries);
-#endif
-
-    pollset.active_fd.six = six;
-    pollset.active_fd.eix = eix;
-    erts_smp_atomic32_set_relb(&pollset.active_fd.no, no);
-}
-
-static void grow_active_fds(void)
-{
-    ASSERT(pollset.active_fd.six == pollset.active_fd.eix);
-    pollset.active_fd.six = 0;
-    pollset.active_fd.eix = pollset.active_fd.size;
-    pollset.active_fd.size += ERTS_ACTIVE_FD_INC;
-    pollset.active_fd.array = erts_realloc(ERTS_ALC_T_ACTIVE_FD_ARR,
-                                           pollset.active_fd.array,
-                                           pollset.active_fd.size*sizeof(ErtsSysFdType));
-#ifdef DEBUG
-    {
-        int i;
-        for (i = pollset.active_fd.eix + 1; i < pollset.active_fd.size; i++)
-            pollset.active_fd.array[i] = ERTS_SYS_FD_INVALID;
-    }
+# define MUST_DEFER(MAY_SLEEP) (MAY_SLEEP)
 #endif
-}
-
-static ERTS_INLINE void
-add_active_fd(ErtsSysFdType fd)
-{
-    int eix = pollset.active_fd.eix;
-    int size = pollset.active_fd.size;
-    
-    pollset.active_fd.array[eix] = fd;
-
-    erts_smp_atomic32_set_relb(&pollset.active_fd.no,
-			       (erts_smp_atomic32_read_dirty(&pollset.active_fd.no)
-				+ 1));
-
-    eix++;
-    if (eix >= size)
-	eix = 0;
-    pollset.active_fd.eix = eix;
-
-    if (pollset.active_fd.six == eix) {
-        grow_active_fds();
-    }
-}
 
 int
-ERTS_CIO_EXPORT(driver_select)(ErlDrvPort ix,
-			       ErlDrvEvent e,
-			       int mode,
-			       int on)
+driver_select(ErlDrvPort ix, ErlDrvEvent e, int mode, int on)
 {
     void (*stop_select_fn)(ErlDrvEvent, void*) = NULL;
     Port *prt = erts_drvport2port(ix);
     Eterm id = erts_drvport2id(ix);
     ErtsSysFdType fd = (ErtsSysFdType) e;
     ErtsPollEvents ctl_events = (ErtsPollEvents) 0;
-    ErtsPollEvents new_events, old_events;
+    ErtsPollEvents old_events;
+    ErtsPollEvents new_events;
+    ErtsPollOp ctl_op = ERTS_POLL_OP_MOD;
     ErtsDrvEventState *state;
-    int wake_poller;
+    int wake_poller = 0;
     int ret;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    ErtsDrvEventDataState *free_event = NULL;
-#endif
     ErtsDrvSelectDataState *free_select = NULL;
     ErtsNifSelectDataState *free_nif = NULL;
 #ifdef USE_VM_PROBES
     DTRACE_CHARBUF(name, 64);
 #endif
+    ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_CHECK_IO);
 
-    if (prt == ERTS_INVALID_ERL_DRV_PORT)
+    if (prt == ERTS_INVALID_ERL_DRV_PORT) {
+        ERTS_MSACC_POP_STATE();
 	return -1;
+    }
 
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
+    ERTS_LC_ASSERT(erts_lc_is_port_locked(prt));
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if ((unsigned)fd >= (unsigned)erts_smp_atomic_read_nob(&drv_ev_state_len)) {
-	if (fd < 0) {
-	    return -1;    
-	}
-	if (fd >= max_fds) {
-	    drv_select_large_fd_error(ix, fd, mode, on);
-	    return -1;
-	}
-	grow_drv_ev_state(fd);
+    if (!grow_drv_ev_state(fd)) {
+        if (fd > 0) drv_select_large_fd_error(ix, fd, mode, on);
+        ERTS_MSACC_POP_STATE();
+        return -1;
     }
 #endif
 
-    erts_smp_mtx_lock(fd_mtx(fd));
+    erts_mtx_lock(fd_mtx(fd));
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    state = &drv_ev_state[(int) fd];
-#else
-    state = hash_get_drv_ev_state(fd); /* may be NULL! */
-#endif
+    state = get_drv_ev_state(fd); /* may be NULL! */
 
-    if (!on && (mode&ERL_DRV_USE_NO_CALLBACK) == ERL_DRV_USE) {
-	if (IS_FD_UNKNOWN(state)) {
-	    /* fast track to stop_select callback */
-	    stop_select_fn = prt->drv_ptr->stop_select;
-#ifdef USE_VM_PROBES
-	    strncpy(name, prt->drv_ptr->name,
-		    sizeof(DTRACE_CHARBUF_NAME(name))-1);
-	    name[sizeof(name)-1] = '\0';
-#endif
-	    ret = 0;
-	    goto done_unknown;
-	}
-	mode |= (ERL_DRV_READ | ERL_DRV_WRITE);
-	wake_poller = 1; /* to eject fd from pollset (if needed) */
-    }
-    else wake_poller = 0;
+    DEBUG_PRINT_FD("driver_select(%T, %p, %s, %d)",
+                   state, id, fd, drvmode2str(mode), on);
 
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if (state == NULL) {
-	state = hash_new_drv_ev_state(fd);
+    if (!on) {
+        if (IS_FD_UNKNOWN(state)) {
+            if ((mode&ERL_DRV_USE_NO_CALLBACK) == ERL_DRV_USE) {
+                /* fast track to stop_select callback */
+                stop_select_fn = prt->drv_ptr->stop_select;
+        #ifdef USE_VM_PROBES
+                strncpy(name, prt->drv_ptr->name,
+                        sizeof(DTRACE_CHARBUF_NAME(name))-1);
+                name[sizeof(name)-1] = '\0';
+        #endif
+            }
+            ret = 0;
+            goto done_unknown;
+        }
+        else if ((mode&ERL_DRV_USE_NO_CALLBACK) == ERL_DRV_USE) {
+            mode |= (ERL_DRV_READ | ERL_DRV_WRITE);
+        }
     }
-#endif
+
+    state = new_drv_ev_state(state, fd);
 
     switch (state->type) {
-#if ERTS_CIO_HAVE_DRV_EVENT
-    case ERTS_EV_TYPE_DRV_EV:
-#endif
     case ERTS_EV_TYPE_NIF:
         drv_select_steal(ix, state, mode, on);
         break;
@@ -1053,7 +718,9 @@ ERTS_CIO_EXPORT(driver_select)(ErlDrvPort ix,
         ASSERT(state->type == ERTS_EV_TYPE_NONE);
         break;
 
-    }}
+    }
+    default: break;
+    }
 
     if (mode & ERL_DRV_READ) {
 	if (state->type == ERTS_EV_TYPE_DRV_SEL) {
@@ -1061,7 +728,7 @@ ERTS_CIO_EXPORT(driver_select)(ErlDrvPort ix,
 	    if (owner != id && is_not_nil(owner))
 		drv_select_steal(ix, state, mode, on);
 	}
-	ctl_events |= ERTS_POLL_EV_IN;
+        ctl_events = ERTS_POLL_EV_IN;
     }
     if (mode & ERL_DRV_WRITE) {
 	if (state->type == ERTS_EV_TYPE_DRV_SEL) {
@@ -1070,100 +737,105 @@ ERTS_CIO_EXPORT(driver_select)(ErlDrvPort ix,
 		drv_select_steal(ix, state, mode, on);
 	}
 	ctl_events |= ERTS_POLL_EV_OUT;
-    }	
+    }
+
 
     ASSERT((state->type == ERTS_EV_TYPE_DRV_SEL) ||
 	   (state->type == ERTS_EV_TYPE_NONE && !state->events));
 
-    if (!on && !(state->flags & ERTS_EV_FLAG_USED) 
-	&& state->events && !(state->events & ~ctl_events)) {	
-	/* Old driver removing all events. At least wake poller.
-	   It will not make close() 100% safe but it will prevent
-	   actions delayed by poll timeout. */
-	wake_poller = 1;
-    }
+    old_events = state->events;
 
-    new_events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd, ctl_events, on, &wake_poller);
+    if (on) {
+        ctl_events &= ~old_events;
+        state->events |= ctl_events;
+        if (ctl_events & ERTS_POLL_EV_IN && (!state->driver.select || !is_iotask_active(&state->driver.select->iniotask)))
+            state->active_events |= ERTS_POLL_EV_IN;
+        if (ctl_events & ERTS_POLL_EV_OUT && (!state->driver.select || !is_iotask_active(&state->driver.select->outiotask)))
+            state->active_events |= ERTS_POLL_EV_OUT;
+        if (old_events == 0 && !(state->flags & ERTS_EV_FLAG_USED)) {
+            ctl_op = ERTS_POLL_OP_ADD;
+        }
+    }
+    else {
+        ctl_events &= old_events;
+        state->events &= ~ctl_events;
+        state->active_events &= ~ctl_events;
 
-    if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
-	if (state->type == ERTS_EV_TYPE_DRV_SEL && !state->events) {
-	    state->type = ERTS_EV_TYPE_NONE;
-	    state->flags &= ~ERTS_EV_FLAG_USED;
-	    state->driver.select->inport = NIL;
-	    state->driver.select->outport = NIL;
-	}
-	ret = -1;
-	goto done;
+        if (!state->events) {
+            if (!(state->flags & ERTS_EV_FLAG_USED) || mode & ERL_DRV_USE)
+                ctl_op = ERTS_POLL_OP_DEL;
+        }
     }
 
-    old_events = state->events;
+    if (ctl_events || ctl_op == ERTS_POLL_OP_DEL) {
 
-    ASSERT(on
-	   ? (new_events == (state->events | ctl_events))
-	   : (new_events == (state->events & ~ctl_events)));
+        new_events = erts_io_control_wakeup(state, ctl_op,
+                                            state->active_events,
+                                            &wake_poller);
 
-    ASSERT(state->type == ERTS_EV_TYPE_DRV_SEL
-	   || state->type == ERTS_EV_TYPE_NONE);
+        ASSERT(state->type == ERTS_EV_TYPE_DRV_SEL || state->type == ERTS_EV_TYPE_NONE);
+    }
 
-    state->events = new_events;
-    if (ctl_events) {
-	if (on) {
+    if (on) {
+        if (ctl_events) {
 	    if (!state->driver.select)
-		state->driver.select = alloc_drv_select_data();
+		state->driver.select = alloc_drv_select_data(state->fd);
 	    if (state->type == ERTS_EV_TYPE_NONE)
 		state->type = ERTS_EV_TYPE_DRV_SEL;
 	    ASSERT(state->type == ERTS_EV_TYPE_DRV_SEL);
-	    if (ctl_events & ERTS_POLL_EV_IN)
+	    if (ctl_events & ERTS_POLL_EV_IN) {
 		state->driver.select->inport = id;
-	    if (ctl_events & ERTS_POLL_EV_OUT)
+                if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL))
+                    iready(id, state);
+            }
+	    if (ctl_events & ERTS_POLL_EV_OUT) {
 		state->driver.select->outport = id;
-	    if (mode & ERL_DRV_USE) {
+                if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL))
+                    oready(id, state);
+            }
+	    if (mode & ERL_DRV_USE)
 		state->flags |= ERTS_EV_FLAG_USED;
-	    }
-	}
-	else { /* off */
-	    if (state->type == ERTS_EV_TYPE_DRV_SEL) {
-		if (ctl_events & ERTS_POLL_EV_IN) {
-		    abort_tasks(state, ERL_DRV_READ);
-		    state->driver.select->inport = NIL;
-		}
-		if (ctl_events & ERTS_POLL_EV_OUT) {
-		    abort_tasks(state, ERL_DRV_WRITE);
-		    state->driver.select->outport = NIL;
-		}
-		if (new_events == 0) {
-		    if (old_events != 0) {
-			remember_removed(state, &pollset);
-		    }		    
-		    if ((mode & ERL_DRV_USE) || !(state->flags & ERTS_EV_FLAG_USED)) {
-			state->type = ERTS_EV_TYPE_NONE;
-			state->flags &= ~ERTS_EV_FLAG_USED;
-		    }
-		    /*else keep it, as fd will probably be selected upon again */
-		}
-	    }
-	    if ((mode & ERL_DRV_USE_NO_CALLBACK) == ERL_DRV_USE) {
-		erts_driver_t* drv_ptr = prt->drv_ptr;
-		ASSERT(new_events==0);
-		if (state->remove_cnt == 0 || !wake_poller) {
-		    /* Safe to close fd now as it is not in pollset
-		       or there was no need to eject fd (kernel poll) */
-		    stop_select_fn = drv_ptr->stop_select;
+        }
+    }
+    else { /* off */
+        if (state->type == ERTS_EV_TYPE_DRV_SEL) {
+            if (ctl_events & ERTS_POLL_EV_IN) {
+                abort_tasks(state, ERL_DRV_READ);
+                state->driver.select->inport = NIL;
+            }
+            if (ctl_events & ERTS_POLL_EV_OUT) {
+                abort_tasks(state, ERL_DRV_WRITE);
+                state->driver.select->outport = NIL;
+            }
+            if (state->events == 0) {
+                if ((mode & ERL_DRV_USE) || !(state->flags & ERTS_EV_FLAG_USED)) {
+                    state->type = ERTS_EV_TYPE_NONE;
+                    state->flags = 0;
+                }
+                /*else keep it, as fd will probably be selected upon again */
+            }
+        }
+        if ((mode & ERL_DRV_USE_NO_CALLBACK) == ERL_DRV_USE) {
+            erts_driver_t* drv_ptr = prt->drv_ptr;
+            ASSERT(state->events==0);
+            if (!wake_poller) {
+                /* Safe to close fd now as it is not in pollset
+                   or there was no need to eject fd (kernel poll) */
+                stop_select_fn = drv_ptr->stop_select;
 #ifdef USE_VM_PROBES
-		    strncpy(name, prt->drv_ptr->name, sizeof(name)-1);
-		    name[sizeof(name)-1] = '\0';
+                strncpy(name, prt->drv_ptr->name, sizeof(name)-1);
+                name[sizeof(name)-1] = '\0';
 #endif
-		}
-		else {
-		    /* Not safe to close fd, postpone stop_select callback. */
-		    state->type = ERTS_EV_TYPE_STOP_USE;
-		    state->driver.stop.drv_ptr = drv_ptr;
-		    if (drv_ptr->handle) {
-			erts_ddll_reference_referenced_driver(drv_ptr->handle);
-		    }
-		}
-	    }
-	}
+            }
+            else {
+                /* Not safe to close fd, postpone stop_select callback. */
+                state->type = ERTS_EV_TYPE_STOP_USE;
+                state->driver.stop.drv_ptr = drv_ptr;
+                if (drv_ptr->handle) {
+                    erts_ddll_reference_referenced_driver(drv_ptr->handle);
+                }
+            }
+        }
     }
    
     ret = 0;
@@ -1171,14 +843,11 @@ ERTS_CIO_EXPORT(driver_select)(ErlDrvPort ix,
 done:
 
     check_fd_cleanup(state,
-#if ERTS_CIO_HAVE_DRV_EVENT
-		     &free_event,
-#endif
 		     &free_select,
                      &free_nif);
 
 done_unknown:
-    erts_smp_mtx_unlock(fd_mtx(fd));
+    erts_mtx_unlock(fd_mtx(fd));
     if (stop_select_fn) {
 	int was_unmasked = erts_block_fpe();
 	DTRACE1(driver_stop_select, name);
@@ -1191,61 +860,47 @@ done_unknown:
     if (free_nif)
         free_nif_select_data(free_nif);
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-    if (free_event)
-	free_drv_event_data(free_event);
-#endif
+    ERTS_MSACC_POP_STATE();
+
     return ret;
 }
 
 int
-ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
-                             ErlNifEvent e,
-                             enum ErlNifSelectFlags mode,
-                             void* obj,
-                             const ErlNifPid* pid,
-                             Eterm ref)
+enif_select(ErlNifEnv* env,
+            ErlNifEvent e,
+            enum ErlNifSelectFlags mode,
+            void* obj,
+            const ErlNifPid* pid,
+            Eterm ref)
 {
     int on;
     ErtsResource* resource = DATA_TO_RESOURCE(obj);
     ErtsSysFdType fd = (ErtsSysFdType) e;
     ErtsPollEvents ctl_events = (ErtsPollEvents) 0;
-    ErtsPollEvents new_events, old_events;
+    ErtsPollEvents old_events;
+    ErtsPollOp ctl_op = ERTS_POLL_OP_MOD;
     ErtsDrvEventState *state;
-    int wake_poller;
-    int ret;
+    int ret, wake_poller = 0;
     enum { NO_STOP=0, CALL_STOP, CALL_STOP_AND_RELEASE } call_stop = NO_STOP;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    ErtsDrvEventDataState *free_event = NULL;
-#endif
     ErtsDrvSelectDataState *free_select = NULL;
     ErtsNifSelectDataState *free_nif = NULL;
-#ifdef USE_VM_PROBES
-    DTRACE_CHARBUF(name, 64);
-#endif
 
-    ASSERT(!(resource->monitors && resource->monitors->is_dying));
+    ASSERT(!resource->monitors);
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if ((unsigned)fd >= (unsigned)erts_smp_atomic_read_nob(&drv_ev_state_len)) {
-	if (fd < 0) {
-	    return INT_MIN | ERL_NIF_SELECT_INVALID_EVENT;
-	}
-	if (fd >= max_fds) {
-	    nif_select_large_fd_error(fd, mode, resource, ref);
-            return INT_MIN | ERL_NIF_SELECT_INVALID_EVENT;
-	}
-	grow_drv_ev_state(fd);
+    if (!grow_drv_ev_state(fd)) {
+        if (fd > 0) nif_select_large_fd_error(fd, mode, resource, ref);
+        return INT_MIN | ERL_NIF_SELECT_INVALID_EVENT;
     }
 #endif
 
-    erts_smp_mtx_lock(fd_mtx(fd));
+    erts_mtx_lock(fd_mtx(fd));
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    state = &drv_ev_state[(int) fd];
-#else
-    state = hash_get_drv_ev_state(fd); /* may be NULL! */
-#endif
+    state = get_drv_ev_state(fd); /* may be NULL! */
+
+    DEBUG_PRINT_FD("enif_select(%T, %d, %s, %p, %T, %T)",
+                   state, env->proc->common.id, fd, nifmode2str(mode), resource,
+                   pid ? pid->pid : THE_NON_VALUE, ref);
 
     if (mode & ERL_NIF_SELECT_STOP) {
         ASSERT(resource->type->stop);
@@ -1257,13 +912,12 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         }
         on = 0;
         mode = ERL_DRV_READ | ERL_DRV_WRITE | ERL_DRV_USE;
-        wake_poller = 1; /* to eject fd from pollset (if needed) */
         ctl_events = ERTS_POLL_EV_IN | ERTS_POLL_EV_OUT;
+        ctl_op = ERTS_POLL_OP_DEL;
     }
     else {
         on = 1;
         ASSERT(mode);
-        wake_poller = 0;
         if (mode & ERL_DRV_READ) {
             ctl_events |= ERTS_POLL_EV_IN;
         }
@@ -1272,11 +926,7 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         }
     }
 
-#ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if (state == NULL) {
-	state = hash_new_drv_ev_state(fd);
-    }
-#endif
+    state = new_drv_ev_state(state,fd);
 
     switch (state->type) {
     case ERTS_EV_TYPE_NIF:
@@ -1287,9 +937,6 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         if (state->driver.stop.resource != resource)
             nif_select_steal(state, ERL_DRV_READ | ERL_DRV_WRITE, resource, ref);
         break;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    case ERTS_EV_TYPE_DRV_EV:
-#endif
     case ERTS_EV_TYPE_DRV_SEL:
         nif_select_steal(state, mode, resource, ref);
         break;
@@ -1310,37 +957,52 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         }
         ASSERT(state->type == ERTS_EV_TYPE_NONE);
         break;
-    }}
+    }
+    default: break;
+    }
 
     ASSERT((state->type == ERTS_EV_TYPE_NIF) ||
 	   (state->type == ERTS_EV_TYPE_NONE && !state->events));
 
-    new_events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd, ctl_events, on, &wake_poller);
+    old_events = state->events;
 
-    if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
-	if (state->type == ERTS_EV_TYPE_NIF && !state->events) {
-	    state->type = ERTS_EV_TYPE_NONE;
-	    state->flags &= ~ERTS_EV_FLAG_USED;
-	    state->driver.nif->in.pid = NIL;
-	    state->driver.nif->out.pid = NIL;
-            state->driver.nif->in.ddeselect_cnt = 0;
-            state->driver.nif->out.ddeselect_cnt = 0;
-            state->driver.stop.resource = NULL;
-	}
-	ret = INT_MIN | ERL_NIF_SELECT_FAILED;
-	goto done;
+    if (on) {
+        ctl_events &= ~old_events;
+        state->events |= ctl_events;
+        state->active_events |= ctl_events;
+        if (state->type == ERTS_EV_TYPE_NONE)
+            ctl_op = ERTS_POLL_OP_ADD;
+    }
+    else {
+        ctl_events &= old_events;
+        state->events &= ~ctl_events;
+        state->active_events &= ~ctl_events;
     }
 
-    old_events = state->events;
+    if (ctl_events || ctl_op == ERTS_POLL_OP_DEL) {
+        ErtsPollEvents new_events;
+
+        new_events = erts_io_control_wakeup(state, ctl_op,
+                                            state->active_events,
+                                            &wake_poller);
 
-    ASSERT(on
-	   ? (new_events == (state->events | ctl_events))
-	   : (new_events == (state->events & ~ctl_events)));
+        if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
+            if (state->type == ERTS_EV_TYPE_NIF && !old_events) {
+                state->type = ERTS_EV_TYPE_NONE;
+                state->flags = 0;
+                state->driver.nif->in.pid = NIL;
+                state->driver.nif->out.pid = NIL;
+                state->driver.stop.resource = NULL;
+            }
+            ret = INT_MIN | ERL_NIF_SELECT_FAILED;
+            goto done;
+        }
+        ASSERT(new_events == state->events);
+    }
 
     ASSERT(state->type == ERTS_EV_TYPE_NIF
 	   || state->type == ERTS_EV_TYPE_NONE);
 
-    state->events = new_events;
     if (on) {
         const Eterm recipient = pid ? pid->pid : env->proc->common.id;
         Uint32* refn;
@@ -1353,7 +1015,7 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         }
         ASSERT(state->type == ERTS_EV_TYPE_NIF);
         ASSERT(state->driver.stop.resource == resource);
-        if (ctl_events & ERTS_POLL_EV_IN) {
+        if (mode & ERL_DRV_READ) {
             state->driver.nif->in.pid = recipient;
             if (is_immed(ref)) {
                 state->driver.nif->in.immed = ref;
@@ -1361,13 +1023,11 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
                 ASSERT(is_internal_ref(ref));
                 refn = internal_ref_numbers(ref);
                 state->driver.nif->in.immed = THE_NON_VALUE;
-                state->driver.nif->in.refn[0] = refn[0];
-                state->driver.nif->in.refn[1] = refn[1];
-                state->driver.nif->in.refn[2] = refn[2];
+                sys_memcpy(state->driver.nif->in.refn, refn,
+                           sizeof(state->driver.nif->in.refn));
             }
-            state->driver.nif->in.ddeselect_cnt = 0;
         }
-        if (ctl_events & ERTS_POLL_EV_OUT) {
+        if (mode & ERL_DRV_WRITE) {
             state->driver.nif->out.pid = recipient;
             if (is_immed(ref)) {
                 state->driver.nif->out.immed = ref;
@@ -1375,11 +1035,9 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
                 ASSERT(is_internal_ref(ref));
                 refn = internal_ref_numbers(ref);
                 state->driver.nif->out.immed = THE_NON_VALUE;
-                state->driver.nif->out.refn[0] = refn[0];
-                state->driver.nif->out.refn[1] = refn[1];
-                state->driver.nif->out.refn[2] = refn[2];
+                sys_memcpy(state->driver.nif->out.refn, refn,
+                           sizeof(state->driver.nif->out.refn));
             }
-            state->driver.nif->out.ddeselect_cnt = 0;
         }
         ret = 0;
     }
@@ -1387,14 +1045,9 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
         if (state->type == ERTS_EV_TYPE_NIF) {
             state->driver.nif->in.pid = NIL;
             state->driver.nif->out.pid = NIL;
-            state->driver.nif->in.ddeselect_cnt = 0;
-            state->driver.nif->out.ddeselect_cnt = 0;
-            if (old_events != 0) {
-                remember_removed(state, &pollset);
-            }
         }
-        ASSERT(new_events==0);
-        if (state->remove_cnt == 0 || !wake_poller) {
+        ASSERT(state->events==0);
+        if (!wake_poller) {
             /*
              * Safe to close fd now as it is not in pollset
              * or there was no need to eject fd (kernel poll)
@@ -1426,14 +1079,11 @@ ERTS_CIO_EXPORT(enif_select)(ErlNifEnv* env,
 done:
 
     check_fd_cleanup(state,
-#if ERTS_CIO_HAVE_DRV_EVENT
-		     &free_event,
-#endif
 		     &free_select,
                      &free_nif);
 
 done_unknown:
-    erts_smp_mtx_unlock(fd_mtx(fd));
+    erts_mtx_unlock(fd_mtx(fd));
     if (call_stop) {
         erts_resource_stop(resource, (ErlNifEvent)fd, 1);
         if (call_stop == CALL_STOP_AND_RELEASE) {
@@ -1445,162 +1095,9 @@ done_unknown:
     if (free_nif)
         free_nif_select_data(free_nif);
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-    if (free_event)
-	free_drv_event_data(free_event);
-#endif
     return ret;
 }
 
-
-int
-ERTS_CIO_EXPORT(driver_event)(ErlDrvPort ix,
-			      ErlDrvEvent e,
-			      ErlDrvEventData event_data)
-{
-#if !ERTS_CIO_HAVE_DRV_EVENT
-    return -1;
-#else
-    ErtsSysFdType fd = (ErtsSysFdType) e;
-    ErtsPollEvents events;
-    ErtsPollEvents add_events;
-    ErtsPollEvents remove_events;
-    Eterm id = erts_drvport2id(ix);
-    ErtsDrvEventState *state;
-    int do_wake = 0;
-    int ret;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    ErtsDrvEventDataState *free_event;
-#endif
-    ErtsDrvSelectDataState *free_select;
-    ErtsNifSelectDataState *free_nif;
-    Port *prt = erts_drvport2port(ix);
-
-    if (prt == ERTS_INVALID_ERL_DRV_PORT)
-	return -1;
-
-    ERTS_SMP_LC_ASSERT(erts_lc_is_port_locked(prt));
-
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if ((unsigned)fd >= (unsigned)erts_smp_atomic_read_nob(&drv_ev_state_len)) {
-	if (fd < 0)
-	    return -1;    
-	if (fd >= max_fds) {
-	    event_large_fd_error(ix, fd, event_data);
-	    return -1;
-	}
-	grow_drv_ev_state(fd);
-    }
-#endif
-
-    erts_smp_mtx_lock(fd_mtx(fd));
-
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    state = &drv_ev_state[(int) fd];
-#else
-    /* Could use hash_new directly, but want to keep the normal case fast */
-    state = hash_get_drv_ev_state(fd);
-    if (state == NULL) {
-	state = hash_new_drv_ev_state(fd);
-    }
-#endif
-
-    switch (state->type) {
-    case ERTS_EV_TYPE_DRV_EV:
-	if (state->driver.event->port == id) break;
-	/*fall through*/
-    case ERTS_EV_TYPE_DRV_SEL:
-	drv_event_steal(ix, state, event_data);
-	break;
-    case ERTS_EV_TYPE_STOP_USE: {
-	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-	print_drv_event_op(dsbufp, ix, fd, event_data);
-	steal_pending_stop_use(dsbufp, ix, state, 0, 1);
-	break;
-      }
-    }
-
-    ASSERT(state->type == ERTS_EV_TYPE_DRV_EV
-	   || state->type == ERTS_EV_TYPE_NONE);
-
-    events = state->events;
-
-    if (!event_data) {
-	remove_events = events;
-	add_events = 0;
-    }
-    else {
-	remove_events = ~event_data->events & events;
-	add_events = ~events & event_data->events;
-    }
-
-    if (add_events) {
-	events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd, add_events, 1, &do_wake);
-	if (events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
-	    ret = -1;
-	    goto done;
-	}
-    }
-    if (remove_events) {
-	events = ERTS_CIO_POLL_CTL(pollset.ps, state->fd, remove_events, 0, &do_wake);
-	if (events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
-	    ret = -1;
-	    goto done;
-	}
-    }
-    if (event_data && event_data->events != 0) {
-	if (state->type == ERTS_EV_TYPE_DRV_EV) {
-	    state->driver.event->removed_events &= ~add_events;
-	    state->driver.event->removed_events |= remove_events;
-	}
-	else {
-	    if (!state->driver.event)
-		state->driver.event = alloc_drv_event_data();
-	    state->driver.event->port = id;
-	    state->driver.event->removed_events = (ErtsPollEvents) 0;
-	    state->type = ERTS_EV_TYPE_DRV_EV;
-	}
-	state->driver.event->data = event_data;
-    }
-    else {
-	if (state->type == ERTS_EV_TYPE_DRV_EV) {
-	    abort_tasks(state, 0);
-	    state->driver.event->port = NIL;
-	    state->driver.event->data = NULL;
-	    state->driver.event->removed_events = (ErtsPollEvents) 0;
-	}
-	state->type = ERTS_EV_TYPE_NONE;
-	remember_removed(state, &pollset);
-    }
-    state->events = events;
-    ASSERT(event_data ? events == event_data->events : events == 0); 
-
-    ret = 0;
-
-done:
-
-    check_fd_cleanup(state,
-#if ERTS_CIO_HAVE_DRV_EVENT
-		     &free_event,
-#endif
-		     &free_select,
-                     &free_nif);
-
-    erts_smp_mtx_unlock(fd_mtx(fd));
-
-    if (free_select)
-	free_drv_select_data(free_select);
-    if (free_nif)
-        free_nif_select_data(free_nif);
-#if ERTS_CIO_HAVE_DRV_EVENT
-    if (free_event)
-	free_drv_event_data(free_event);
-#endif
-
-    return ret;
-#endif
-}
-
 static ERTS_INLINE int
 chk_stale(Eterm id, ErtsDrvEventState *state, int mode)
 {
@@ -1632,11 +1129,6 @@ need2steal(ErtsDrvEventState *state, int mode)
         do_steal = 1;
         break;
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-    case ERTS_EV_TYPE_DRV_EV:
-	do_steal |= chk_stale(state->driver.event->port, state, 0);
-	break;
-#endif
     case ERTS_EV_TYPE_STOP_USE:
     case ERTS_EV_TYPE_STOP_NIF:
         ASSERT(0);
@@ -1670,7 +1162,7 @@ print_driver_name(erts_dsprintf_buf_t *dsbufp, Eterm id)
 static void
 steal(erts_dsprintf_buf_t *dsbufp, ErtsDrvEventState *state, int mode)
 {
-    erts_dsprintf(dsbufp, "stealing control of fd=%d from ", (int) GET_FD(state->fd));
+    erts_dsprintf(dsbufp, "stealing control of fd=%d from ", (int) state->fd);
     switch (state->type) {
     case ERTS_EV_TYPE_DRV_SEL: {
 	int deselect_mode = 0;
@@ -1694,7 +1186,7 @@ steal(erts_dsprintf_buf_t *dsbufp, ErtsDrvEventState *state, int mode)
 	if (deselect_mode)
 	    deselect(state, deselect_mode);
 	else {
-	    erts_dsprintf(dsbufp, "no one", (int) GET_FD(state->fd));
+	    erts_dsprintf(dsbufp, "no one", (int) state->fd);
 	    ASSERT(0);
 	}
 	erts_dsprintf(dsbufp, "\n");
@@ -1719,30 +1211,13 @@ steal(erts_dsprintf_buf_t *dsbufp, ErtsDrvEventState *state, int mode)
         erts_dsprintf(dsbufp, "\n");
         break;
     }
-#if ERTS_CIO_HAVE_DRV_EVENT
-    case ERTS_EV_TYPE_DRV_EV: {
-	Eterm eid = state->driver.event->port;
-	if (is_nil(eid)) {
-	    erts_dsprintf(dsbufp, "no one", (int) state->fd);
-	    ASSERT(0);
-	}
-	else {
-	    erts_dsprintf(dsbufp, "event driver ");
-	    print_driver_name(dsbufp, eid);
-	    erts_dsprintf(dsbufp, "%T ", eid);
-	}
-	erts_dsprintf(dsbufp, "\n");
-	deselect(state, 0);
-	break;
-    }
-#endif
     case ERTS_EV_TYPE_STOP_USE:
     case ERTS_EV_TYPE_STOP_NIF: {
 	ASSERT(0);
 	break;
     }
     default:
-	erts_dsprintf(dsbufp, "no one\n", (int) GET_FD(state->fd));
+	erts_dsprintf(dsbufp, "no one\n", (int) state->fd);
 	ASSERT(0);
     }
 }
@@ -1756,7 +1231,7 @@ print_drv_select_op(erts_dsprintf_buf_t *dsbufp,
 		  "driver_select(%p, %d,%s%s%s%s, %d) "
 		  "by ",
 		  ix,
-		  (int) GET_FD(fd),
+		  (int) fd,
 		  mode & ERL_DRV_READ ? " ERL_DRV_READ" : "",
 		  mode & ERL_DRV_WRITE ? " ERL_DRV_WRITE" : "",
 		  mode & ERL_DRV_USE ? " ERL_DRV_USE" : "",
@@ -1773,7 +1248,7 @@ print_nif_select_op(erts_dsprintf_buf_t *dsbufp,
 {
     erts_dsprintf(dsbufp,
 		  "enif_select(_, %d,%s%s%s, %T:%T, %T) ",
-		  (int) GET_FD(fd),
+		  (int) fd,
 		  mode & ERL_NIF_SELECT_READ ? " READ" : "",
 		  mode & ERL_NIF_SELECT_WRITE ? " WRITE" : "",
 		  mode & ERL_NIF_SELECT_STOP ? " STOP" : "",
@@ -1812,7 +1287,7 @@ large_fd_error_common(erts_dsprintf_buf_t *dsbufp, ErtsSysFdType fd)
 {
     erts_dsprintf(dsbufp,
 		  "fd=%d is larger than the largest allowed fd=%d\n",
-		  (int) fd, max_fds - 1);
+		  (int) fd, drv_ev_state.max_fds - 1);
 }
 
 static void
@@ -1868,7 +1343,7 @@ steal_pending_stop_use(erts_dsprintf_buf_t *dsbufp, ErlDrvPort ix,
             erts_ddll_dereference_driver(state->driver.stop.drv_ptr->handle);
         }
         state->type = ERTS_EV_TYPE_NONE;
-        state->flags &= ~ERTS_EV_FLAG_USED;
+        state->flags = 0;
         state->driver.stop.drv_ptr = NULL;
     }
     else {
@@ -1905,9 +1380,9 @@ steal_pending_stop_nif(erts_dsprintf_buf_t *dsbufp, ErtsResource* resource,
         erts_dsprintf(dsbufp, "called before stop was called for NIF resource %T:%T\n",
                       rt->module, rt->name);
 
-        enif_release_resource(state->driver.stop.resource);
+        enif_release_resource(state->driver.stop.resource->data);
         state->type = ERTS_EV_TYPE_NONE;
-        state->flags &= ~ERTS_EV_FLAG_USED;
+        state->flags = 0;
         state->driver.stop.resource = NULL;
     }
     else {
@@ -1917,59 +1392,9 @@ steal_pending_stop_nif(erts_dsprintf_buf_t *dsbufp, ErtsResource* resource,
 
 }
 
-
-#if ERTS_CIO_HAVE_DRV_EVENT
-
-static void
-print_drv_event_op(erts_dsprintf_buf_t *dsbufp,
-                   ErlDrvPort ix, ErtsSysFdType fd, ErlDrvEventData event_data)
-{
-    Port *pp = erts_drvport2port(ix);
-    erts_dsprintf(dsbufp, "driver_event(%p, %d, ", ix, (int) fd);
-    if (!event_data)
-	erts_dsprintf(dsbufp, "NULL");
-    else
-	erts_dsprintf(dsbufp, "{0x%x, 0x%x}",
-		      (unsigned int) event_data->events,
-		      (unsigned int) event_data->revents);
-    erts_dsprintf(dsbufp, ") by ");
-    if (pp != ERTS_INVALID_ERL_DRV_PORT)
-	print_driver_name(dsbufp, pp->common.id);
-    erts_dsprintf(dsbufp, "driver %T ", pp != ERTS_INVALID_ERL_DRV_PORT ? pp->common.id : NIL);
-}
-
-static void
-drv_event_steal(ErlDrvPort ix, ErtsDrvEventState *state, ErlDrvEventData event_data)
-{
-    if (need2steal(state, ERL_DRV_READ|ERL_DRV_WRITE)) {
-	erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-	print_drv_event_op(dsbufp, ix, state->fd, event_data);
-	steal(dsbufp, state, ERL_DRV_READ|ERL_DRV_WRITE);
-	erts_send_error_to_logger_nogl(dsbufp);
-    }
-    else if (state->type == ERTS_EV_TYPE_DRV_SEL) {
-	ASSERT(state->flags & ERTS_EV_FLAG_USED);
-	deselect(state, 0);
-    }
-}
-
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-static void
-event_large_fd_error(ErlDrvPort ix, ErtsSysFdType fd, ErlDrvEventData event_data)
-{
-    erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
-    print_drv_event_op(dsbufp, ix, fd, event_data);
-    erts_dsprintf(dsbufp, "failed: ");
-    large_fd_error_common(dsbufp, fd);
-    erts_send_error_to_logger_nogl(dsbufp);
-}
-#endif
-#endif
-
 static ERTS_INLINE int
 io_task_schedule_allowed(ErtsDrvEventState *state,
-			 ErtsPortTaskType type,
-			 erts_aint_t current_cio_time)
+			 ErtsPortTaskType type)
 {
     ErtsIoTask *io_task;
 
@@ -1977,71 +1402,53 @@ io_task_schedule_allowed(ErtsDrvEventState *state,
     case ERTS_PORT_TASK_INPUT:
 	if (!state->driver.select)
 	    return 0;
-#if ERTS_CIO_HAVE_DRV_EVENT
-	if (state->driver.event)
-	    return 0;
-#endif
 	io_task = &state->driver.select->iniotask;
 	break;
     case ERTS_PORT_TASK_OUTPUT:
 	if (!state->driver.select)
 	    return 0;
-#if ERTS_CIO_HAVE_DRV_EVENT
-	if (state->driver.event)
-	    return 0;
-#endif
 	io_task = &state->driver.select->outiotask;
 	break;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    case ERTS_PORT_TASK_EVENT:
-	if (!state->driver.event)
-	    return 0;
-	if (state->driver.select)
-	    return 0;
-	io_task = &state->driver.event->iotask;
-	break;
-#endif
     default:
 	ERTS_INTERNAL_ERROR("Invalid I/O-task type");
 	return 0;
     }
 
-    return !is_iotask_active(io_task, current_cio_time);
+    return !is_iotask_active(io_task);
 }
 
 static ERTS_INLINE void
-iready(Eterm id, ErtsDrvEventState *state, erts_aint_t current_cio_time)
+iready(Eterm id, ErtsDrvEventState *state)
 {
     if (io_task_schedule_allowed(state,
-				 ERTS_PORT_TASK_INPUT,
-				 current_cio_time)) {
+				 ERTS_PORT_TASK_INPUT)) {
 	ErtsIoTask *iotask = &state->driver.select->iniotask;
-	erts_smp_atomic_set_nob(&iotask->executed_time, current_cio_time);
 	if (erts_port_task_schedule(id,
 				    &iotask->task,
 				    ERTS_PORT_TASK_INPUT,
 				    (ErlDrvEvent) state->fd) != 0) {
 	    stale_drv_select(id, state, ERL_DRV_READ);
-	}
-	add_active_fd(state->fd);
+	} else {
+            DEBUG_PRINT_FD("schedule ready_input(%T, %d)",
+                           state, id, state->fd);
+        }
     }
 }
 
 static ERTS_INLINE void
-oready(Eterm id, ErtsDrvEventState *state, erts_aint_t current_cio_time)
+oready(Eterm id, ErtsDrvEventState *state)
 {
     if (io_task_schedule_allowed(state,
-				 ERTS_PORT_TASK_OUTPUT,
-				 current_cio_time)) {
+				 ERTS_PORT_TASK_OUTPUT)) {
 	ErtsIoTask *iotask = &state->driver.select->outiotask;
-	erts_smp_atomic_set_nob(&iotask->executed_time, current_cio_time);
 	if (erts_port_task_schedule(id,
 				    &iotask->task,
 				    ERTS_PORT_TASK_OUTPUT,
 				    (ErlDrvEvent) state->fd) != 0) {
 	    stale_drv_select(id, state, ERL_DRV_WRITE);
-	}
-	add_active_fd(state->fd);
+	} else {
+            DEBUG_PRINT_FD("schedule ready_output(%T, %d)", state, id, state->fd);
+        }
     }
 }
 
@@ -2090,141 +1497,62 @@ send_event_tuple(struct erts_nif_select_event* e, ErtsResource* resource,
     erts_queue_message(rp, rp_locks, mp, tuple, am_system);
 
     if (rp_locks)
-        erts_smp_proc_unlock(rp, rp_locks);
+        erts_proc_unlock(rp, rp_locks);
 }
 
-
-#if ERTS_CIO_HAVE_DRV_EVENT
-static ERTS_INLINE void
-eready(Eterm id, ErtsDrvEventState *state, ErlDrvEventData event_data,
-       erts_aint_t current_cio_time)
-{
-    if (io_task_schedule_allowed(state,
-				 ERTS_PORT_TASK_EVENT,
-				 current_cio_time)) {
-	ErtsIoTask *iotask = &state->driver.event->iotask;
-	erts_smp_atomic_set_nob(&iotask->executed_time, current_cio_time);
-	if (erts_port_task_schedule(id,
-				    &iotask->task,
-				    ERTS_PORT_TASK_EVENT,
-				    (ErlDrvEvent) state->fd,
-				    event_data) != 0) {
-	    stale_drv_select(id, state, 0);
-	}
-	add_active_fd(state->fd);
-    }
-}
-#endif
-
 static void bad_fd_in_pollset(ErtsDrvEventState *, Eterm inport, Eterm outport);
 
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
 void
-ERTS_CIO_EXPORT(erts_check_io_async_sig_interrupt)(void)
+erts_check_io_interrupt(ErtsPollThread *psi, int set)
 {
-    ERTS_CIO_POLL_AS_INTR(pollset.ps);
-}
+    if (psi) {
+#if ERTS_POLL_USE_FALLBACK
+        if (psi->ps == get_fallback()) {
+            erts_poll_interrupt_flbk(psi->ps, set);
+            return;
+        }
 #endif
-
-void
-ERTS_CIO_EXPORT(erts_check_io_interrupt)(int set)
-{
-    ERTS_CIO_POLL_INTR(pollset.ps, set);
+        erts_poll_interrupt(psi->ps, set);
+    }
 }
 
-void
-ERTS_CIO_EXPORT(erts_check_io_interrupt_timed)(int set,
-					       ErtsMonotonicTime timeout_time)
-{
-    ERTS_CIO_POLL_INTR_TMD(pollset.ps, set, timeout_time);
+ErtsPollThread *
+erts_create_pollset_thread(int id) {
+    return psiv+id;
 }
 
-#if !ERTS_CIO_DEFER_ACTIVE_EVENTS
-/*
- * Number of ignored events, for a lingering fd added by enif_select(),
- * until we deselect fd-event from pollset.
- */
-# define ERTS_NIF_DELAYED_DESELECT 20
-#else
-/* Disable delayed deselect as pollset cannot handle active events */
-# define ERTS_NIF_DELAYED_DESELECT 1
-#endif
-
 void
-ERTS_CIO_EXPORT(erts_check_io)(int do_wait)
+erts_check_io(ErtsPollThread *psi)
 {
-    ErtsPollResFd *pollres;
     int pollres_len;
-    ErtsMonotonicTime timeout_time;
     int poll_ret, i;
-    erts_aint_t current_cio_time;
-    ErtsSchedulerData *esdp = erts_get_scheduler_data();
-
-    ASSERT(esdp);
+    ERTS_MSACC_PUSH_AND_SET_STATE(ERTS_MSACC_STATE_CHECK_IO);
 
  restart:
 
-#ifdef ERTS_BREAK_REQUESTED
-    if (ERTS_BREAK_REQUESTED)
-	erts_do_break_handling();
-#endif
-
-#ifdef ERTS_SIGNAL_STATE /* ifndef ERTS_SMP */
-    if (ERTS_SIGNAL_STATE) {
-        erts_handle_signal_state();
-    }
-#endif
-
-    /* Figure out timeout value */
-    timeout_time = (do_wait
-		    ? erts_check_next_timeout_time(esdp)
-		    : ERTS_POLL_NO_TIMEOUT /* poll only */);
-
-    /*
-     * No need for an atomic inc op when incrementing
-     * erts_check_io_time, since only one thread can
-     * check io at a time.
-     */
-    current_cio_time = erts_smp_atomic_read_dirty(&erts_check_io_time);
-    current_cio_time++;
-    erts_smp_atomic_set_relb(&erts_check_io_time, current_cio_time);
-
-    check_cleanup_active_fds(current_cio_time,
-                             timeout_time != ERTS_POLL_NO_TIMEOUT);
-
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_check_exact(NULL, 0); /* No locks should be locked */
 #endif
 
-    pollres_len = erts_smp_atomic32_read_dirty(&pollset.active_fd.no) + ERTS_CHECK_IO_POLL_RES_LEN;
+    pollres_len = psi->pollres_len;
 
-    pollres = erts_alloc(ERTS_ALC_T_TMP, sizeof(ErtsPollResFd)*pollres_len);
+#if ERTS_POLL_USE_FALLBACK
+    if (psi->ps == get_fallback()) {
 
-    erts_smp_atomic_set_nob(&pollset.in_poll_wait, 1);
+        poll_ret = erts_poll_wait_flbk(psi->ps, psi->pollres, &pollres_len);
 
-    poll_ret = ERTS_CIO_POLL_WAIT(pollset.ps, pollres, &pollres_len, timeout_time);
+    } else
+#endif
+    {
+        poll_ret = erts_poll_wait(psi->ps, psi->pollres, &pollres_len);
+    }
 
 #ifdef ERTS_ENABLE_LOCK_CHECK
     erts_lc_check_exact(NULL, 0); /* No locks should be locked */
 #endif
 
-#ifdef ERTS_BREAK_REQUESTED
-    if (ERTS_BREAK_REQUESTED)
-	erts_do_break_handling();
-#endif
-
-
-#ifdef ERTS_SIGNAL_STATE /* ifndef ERTS_SMP */
-    if (ERTS_SIGNAL_STATE) {
-        erts_handle_signal_state();
-    }
-#endif
-
-
     if (poll_ret != 0) {
-	erts_smp_atomic_set_nob(&pollset.in_poll_wait, 0);
-	forget_removed(&pollset);
-	erts_free(ERTS_ALC_T_TMP, pollres);
+
 	if (poll_ret == EAGAIN) {
 	    goto restart;
 	}
@@ -2240,65 +1568,86 @@ ERTS_CIO_EXPORT(erts_check_io)(int do_wait)
 			  erl_errno_id(poll_ret), poll_ret);
 	    erts_send_error_to_logger_nogl(dsbufp);
 	}
+        ERTS_MSACC_POP_STATE();
 	return;
     }
 
     for (i = 0; i < pollres_len; i++) {
 
-	ErtsSysFdType fd = (ErtsSysFdType) pollres[i].fd;
+        erts_driver_t* drv_ptr = NULL;
+        ErtsResource* resource = NULL;
+        ErtsDrvSelectDataState *free_select = NULL;
+        ErtsNifSelectDataState *free_nif = NULL;
+	ErtsSysFdType fd = (ErtsSysFdType) ERTS_POLL_RES_GET_FD(&psi->pollres[i]);
 	ErtsDrvEventState *state;
+        ErtsPollEvents revents;
 
-	erts_smp_mtx_lock(fd_mtx(fd));
+	erts_mtx_lock(fd_mtx(fd));
+
+	state = get_drv_ev_state(fd);
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	state = &drv_ev_state[ (int) fd];
-#else
-	state = hash_get_drv_ev_state(fd);
 	if (!state) {
-	    goto next_pollres;
+            erts_mtx_unlock(fd_mtx(fd));
+            continue;
 	}
-#endif
 
-	/* Skip this fd if it was removed from pollset */
-	if (is_removed(state)) {
-	    goto next_pollres;
-	}
+        revents = ERTS_POLL_RES_GET_EVTS(&psi->pollres[i]);
+
+        DEBUG_PRINT_FD("triggered %s", state, ev2str(revents));
+
+        if (revents & ERTS_POLL_EV_ERR) {
+            /*
+             * Handle error events by triggering all in/out events
+             * that has been selected on.
+             * We *do not* want to call a callback that corresponds
+             * to an event not selected.
+             */
+            revents = state->active_events;
+            state->active_events = 0;
+        } else {
+
+            /* Disregard any events that are not active at the moment,
+               for instance this could happen if the driver/nif does
+               select/deselect in rapid succession. */
+            revents &= state->active_events | ERTS_POLL_EV_NVAL;
+            state->active_events &= ~revents;
+
+            /* Reactivate the poll op if there are still active events */
+            if (state->active_events) {
+                ErtsPollEvents new_events;
+                DEBUG_PRINT_FD("re-enable %s", state, ev2str(state->active_events));
+
+                new_events = erts_io_control(state, ERTS_POLL_OP_MOD, state->active_events);
+
+                /* Unable to re-enable the fd, signal all callbacks */
+                if (new_events & (ERTS_POLL_EV_ERR|ERTS_POLL_EV_NVAL)) {
+                    revents |= state->active_events;
+                    state->active_events = 0;
+                }
+            }
+        }
 
 	switch (state->type) {
 	case ERTS_EV_TYPE_DRV_SEL: { /* Requested via driver_select()... */
-	    ErtsPollEvents revents = pollres[i].events;
-
-            if (revents & ERTS_POLL_EV_ERR) {
-                /*
-                 * Handle error events by triggering all in/out events
-                 * that the driver has selected.
-                 * We *do not* want to call a callback that corresponds
-		 * to an event not selected.
-                 */
-                revents = state->events;
-            }
-            else {
-                revents &= (state->events | ERTS_POLL_EV_NVAL);
-            }
 
             if (revents & (ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT)) {
 		if (revents & ERTS_POLL_EV_OUT) {
-		    oready(state->driver.select->outport, state, current_cio_time);
+		    oready(state->driver.select->outport, state);
 		}
 		/* Someone might have deselected input since revents
 		   was read (true also on the non-smp emulator since
 		   oready() may have been called); therefore, update
 		   revents... */
-		revents &= state->events;
+                revents &= state->events;
 		if (revents & ERTS_POLL_EV_IN) {
-		    iready(state->driver.select->inport, state, current_cio_time);
+		    iready(state->driver.select->inport, state);
 		}
 	    }
 	    else if (revents & ERTS_POLL_EV_NVAL) {
 		bad_fd_in_pollset(state,
                                   state->driver.select->inport,
                                   state->driver.select->outport);
-		add_active_fd(state->fd);
+                check_fd_cleanup(state, &free_select, &free_nif);
 	    }
 	    break;
 	}
@@ -2307,112 +1656,116 @@ ERTS_CIO_EXPORT(erts_check_io)(int do_wait)
             struct erts_nif_select_event in = {NIL};
             struct erts_nif_select_event out = {NIL};
             ErtsResource* resource = NULL;
-            ErtsPollEvents revents = pollres[i].events;
-
-            if (revents & ERTS_POLL_EV_ERR) {
-                /*
-                 * Handle error events by triggering all in/out events
-                 * that the NIF has selected.
-                 * We *do not* want to send a message that corresponds
-		 * to an event not selected.
-                 */
-                revents = state->events;
-            }
-            else {
-                revents &= (state->events | ERTS_POLL_EV_NVAL);
-            }
 
             if (revents & (ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT)) {
                 if (revents & ERTS_POLL_EV_OUT) {
                     if (is_not_nil(state->driver.nif->out.pid)) {
                         out = state->driver.nif->out;
                         resource = state->driver.stop.resource;
-                        state->driver.nif->out.ddeselect_cnt = ERTS_NIF_DELAYED_DESELECT;
                         state->driver.nif->out.pid = NIL;
-                        add_active_fd(state->fd);
-                    }
-                    else {
-                        ASSERT(state->driver.nif->out.ddeselect_cnt >= 2);
-                        state->driver.nif->out.ddeselect_cnt--;
                     }
                 }
                 if (revents & ERTS_POLL_EV_IN) {
                     if (is_not_nil(state->driver.nif->in.pid)) {
                         in = state->driver.nif->in;
                         resource = state->driver.stop.resource;
-                        state->driver.nif->in.ddeselect_cnt = ERTS_NIF_DELAYED_DESELECT;
                         state->driver.nif->in.pid = NIL;
-                        add_active_fd(state->fd);
-                    }
-                    else {
-                        ASSERT(state->driver.nif->in.ddeselect_cnt >= 2);
-                        state->driver.nif->in.ddeselect_cnt--;
                     }
                 }
+                state->events &= ~revents;
             }
             else if (revents & ERTS_POLL_EV_NVAL) {
                 bad_fd_in_pollset(state, NIL, NIL);
-                add_active_fd(state->fd);
+                check_fd_cleanup(state, &free_select, &free_nif);
             }
 
-#ifdef ERTS_SMP
-            erts_smp_mtx_unlock(fd_mtx(fd));
-#endif
+            erts_mtx_unlock(fd_mtx(fd));
+
             if (is_not_nil(in.pid)) {
                 send_event_tuple(&in, resource, am_ready_input);
             }
             if (is_not_nil(out.pid)) {
                 send_event_tuple(&out, resource, am_ready_output);
             }
-            goto next_pollres_unlocked;
+            continue;
         }
 
-#if ERTS_CIO_HAVE_DRV_EVENT
-	case ERTS_EV_TYPE_DRV_EV: { /* Requested via driver_event()... */
-	    ErlDrvEventData event_data;
-	    ErtsPollEvents revents;
-	    ASSERT(state->driver.event);
-	    ASSERT(state->driver.event->data);
-	    event_data = state->driver.event->data;
-	    revents = pollres[i].events;
-	    revents &= ~state->driver.event->removed_events;
-
-	    if (revents) {
-		event_data->events = state->events;
-		event_data->revents = revents;
-		eready(state->driver.event->port, state, event_data, current_cio_time);
-	    }
-	    break;
-	}
-#endif
+        case ERTS_EV_TYPE_STOP_NIF: {
+            resource = state->driver.stop.resource;
+            state->type = ERTS_EV_TYPE_NONE;
+            goto case_ERTS_EV_TYPE_NONE;
+        }
 
+        case ERTS_EV_TYPE_STOP_USE: {
+#if ERTS_POLL_USE_FALLBACK
+            ASSERT(psi->ps == get_fallback());
+#endif
+            drv_ptr = state->driver.stop.drv_ptr;
+            state->type = ERTS_EV_TYPE_NONE;
+            /* fallthrough */
 	case ERTS_EV_TYPE_NONE: /* Deselected ... */
+        case_ERTS_EV_TYPE_NONE:
+            ASSERT(!state->events && !state->active_events && !state->flags);
+            check_fd_cleanup(state, &free_select, &free_nif);
 	    break;
+        }
 
 	default: { /* Error */
 	    erts_dsprintf_buf_t *dsbufp;
 	    dsbufp = erts_create_logger_dsbuf();
 	    erts_dsprintf(dsbufp,
 			  "Invalid event request type for fd in erts_poll()! "
-			  "fd=%d, event request type=%sd\n", (int) state->fd,
+			  "fd=%d, event request type=%d\n", (int) state->fd,
 			  (int) state->type);
 	    ASSERT(0);
 	    deselect(state, 0);
-	    add_active_fd(state->fd);
 	    break;
 	}
 	}
 
-	next_pollres:;
-#ifdef ERTS_SMP
-	erts_smp_mtx_unlock(fd_mtx(fd));
-#endif
-        next_pollres_unlocked:;
-    }
+	erts_mtx_unlock(fd_mtx(fd));
 
-    erts_smp_atomic_set_nob(&pollset.in_poll_wait, 0);
-    erts_free(ERTS_ALC_T_TMP, pollres);
-    forget_removed(&pollset);
+        if (drv_ptr) {
+            int was_unmasked = erts_block_fpe();
+            DTRACE1(driver_stop_select, drv_ptr->name);
+            LTTNG1(driver_stop_select, drv_ptr->name);
+            (*drv_ptr->stop_select)((ErlDrvEvent) fd, NULL);
+            erts_unblock_fpe(was_unmasked);
+            if (drv_ptr->handle) {
+		erts_ddll_dereference_driver(drv_ptr->handle);
+	    }
+        }
+        if (resource) {
+            erts_resource_stop(resource, (ErlNifEvent)fd, 0);
+            enif_release_resource(resource->data);
+        }
+        if (free_select)
+            free_drv_select_data(free_select);
+        if (free_nif)
+            free_nif_select_data(free_nif);
+    }
+
+    /* The entire pollres array was filled with events,
+     * grow it for the next call. We do this for two reasons:
+     * 1. Pulling out more events in on go will increase throughput
+     * 2. If the polling implementation is not fair, this will make
+     *    sure that we get all fds that we can. i.e. if 12 fds are
+     *    constantly active, but we only have a pollres_len of 10,
+     *    two of the fds may never be triggered depending on what the
+     *    kernel decides to do.
+     **/
+    if (pollres_len == psi->pollres_len) {
+        int ev_state_len = drv_ev_state_len();
+        erts_free(ERTS_ALC_T_POLLSET, psi->pollres);
+        psi->pollres_len *= 2;
+        /* Never grow it larger than the current drv_ev_state.len size */
+        if (psi->pollres_len > ev_state_len)
+            psi->pollres_len = ev_state_len;
+        psi->pollres = erts_alloc(ERTS_ALC_T_POLLSET,
+                                  sizeof(ErtsPollResFd) * psi->pollres_len);
+    }
+
+    ERTS_MSACC_POP_STATE();
 }
 
 static void
@@ -2506,16 +1859,16 @@ static int drv_ev_state_cmp(void *des1, void *des2)
 static void *drv_ev_state_alloc(void *des_tmpl)
 {
     ErtsDrvEventState *evstate;
-    erts_smp_spin_lock(&state_prealloc_lock);
-    if (state_prealloc_first == NULL) {
-	erts_smp_spin_unlock(&state_prealloc_lock);
+    erts_spin_lock(&drv_ev_state.prealloc_lock);
+    if (drv_ev_state.prealloc_first == NULL) {
+	erts_spin_unlock(&drv_ev_state.prealloc_lock);
 	evstate = (ErtsDrvEventState *) 
 	    erts_alloc(ERTS_ALC_T_DRV_EV_STATE, sizeof(ErtsDrvEventState));
     } else {
-	evstate = state_prealloc_first;
-	state_prealloc_first = (ErtsDrvEventState *) evstate->hb.next;
-	--num_state_prealloc;
-	erts_smp_spin_unlock(&state_prealloc_lock);
+	evstate = drv_ev_state.prealloc_first;
+	drv_ev_state.prealloc_first = (ErtsDrvEventState *) evstate->hb.next;
+	--drv_ev_state.num_prealloc;
+	erts_spin_unlock(&drv_ev_state.prealloc_lock);
     }
     /* XXX: Already valid data if prealloced, could ignore template! */
     *evstate = *((ErtsDrvEventState *) des_tmpl);
@@ -2525,60 +1878,200 @@ static void *drv_ev_state_alloc(void *des_tmpl)
 
 static void drv_ev_state_free(void *des)
 {
-    erts_smp_spin_lock(&state_prealloc_lock);
-    ((ErtsDrvEventState *) des)->hb.next = &state_prealloc_first->hb;
-    state_prealloc_first = (ErtsDrvEventState *) des;
-    ++num_state_prealloc;
-    erts_smp_spin_unlock(&state_prealloc_lock);
+    erts_spin_lock(&drv_ev_state.prealloc_lock);
+    ((ErtsDrvEventState *) des)->hb.next = &drv_ev_state.prealloc_first->hb;
+    drv_ev_state.prealloc_first = (ErtsDrvEventState *) des;
+    ++drv_ev_state.num_prealloc;
+    erts_spin_unlock(&drv_ev_state.prealloc_lock);
 }
 #endif
 
+#define ERTS_MAX_NO_OF_POLL_THREADS ERTS_MAX_NO_OF_SCHEDULERS
+
+static char *
+get_arg(char* rest, char** argv, int* ip)
+{
+    int i = *ip;
+    if (*rest == '\0') {
+	if (argv[i+1] == NULL) {
+	    erts_fprintf(stderr, "too few arguments\n");
+	    erts_usage();
+	}
+        argv[i++] = NULL;
+        rest = argv[i];
+    }
+    argv[i] = NULL;
+    *ip = i;
+    return rest;
+}
+
+static void
+parse_args(int *argc, char **argv, int concurrent_waiters)
+{
+    int i = 0, j;
+    int no_pollsets = 0, no_poll_threads = 0,
+        no_pollsets_percentage = 0,
+        no_poll_threads_percentage = 0;
+    ASSERT(argc && argv);
+    while (i < *argc) {
+	if(argv[i][0] == '-') {
+	    switch (argv[i][1]) {
+            case 'I': {
+                if (strncmp(argv[i]+2, "Ot", 2) == 0) {
+                    char *arg = get_arg(argv[i]+4, argv, &i);
+                    if (sscanf(arg, "%d", &no_poll_threads) != 1 ||
+                        no_poll_threads < 1 ||
+                        ERTS_MAX_NO_OF_POLL_THREADS < no_poll_threads) {
+                        erts_fprintf(stderr,"bad I/O poll threads number: %s\n", arg);
+                        erts_usage();
+                    }
+                } else if (strncmp(argv[i]+2, "Op", 3) == 0) {
+                    char *arg = get_arg(argv[i]+4, argv, &i);
+                    if (sscanf(arg, "%d", &no_pollsets) != 1 ||
+                        no_pollsets < 1) {
+                        erts_fprintf(stderr,"bad I/O pollset number: %s\n", arg);
+                        erts_usage();
+                    }
+                } else if (strncmp(argv[i]+2, "OPt", 4) == 0) {
+                    char *arg = get_arg(argv[i]+5, argv, &i);
+                    if (sscanf(arg, "%d", &no_poll_threads_percentage) != 1 ||
+                        no_poll_threads_percentage < 0 ||
+                        no_poll_threads_percentage > 100) {
+                        erts_fprintf(stderr,"bad I/O poll thread percentage number: %s\n", arg);
+                        erts_usage();
+                    }
+                } else if (strncmp(argv[i]+2, "OPp", 4) == 0) {
+                    char *arg = get_arg(argv[i]+5, argv, &i);
+                    if (sscanf(arg, "%d", &no_pollsets_percentage) != 1 ||
+                        no_pollsets_percentage < 0 ||
+                        no_pollsets_percentage > 100) {
+                        erts_fprintf(stderr,"bad I/O pollset percentage number: %s\n", arg);
+                        erts_usage();
+                    }
+                } else {
+                    break;
+                }
+                break;
+            }
+            case 'K':
+                (void)get_arg(argv[i]+2, argv, &i);
+                break;
+            case '-':
+                goto args_parsed;
+            default:
+                break;
+            }
+        }
+        i++;
+    }
+
+args_parsed:
+
+    if (!concurrent_waiters) {
+        no_pollsets = no_poll_threads;
+        no_pollsets_percentage = 100;
+    }
+
+    if (no_poll_threads == 0) {
+        if (no_poll_threads_percentage == 0)
+            no_poll_threads = 1; /* This is the default */
+        else {
+            no_poll_threads = erts_no_schedulers * no_poll_threads_percentage / 100;
+            if (no_poll_threads < 1)
+                no_poll_threads = 1;
+        }
+    }
+
+    if (no_pollsets == 0) {
+        if (no_pollsets_percentage == 0)
+            no_pollsets = 1; /* This is the default */
+        else {
+            no_pollsets = no_poll_threads * no_pollsets_percentage / 100;
+            if (no_pollsets < 1)
+                no_pollsets = 1;
+        }
+    }
+
+    if (no_poll_threads < no_pollsets) {
+        erts_fprintf(stderr,
+                     "number of IO poll threads has to be greater or equal to "
+                     "the number of \nIO pollsets. Current values are set to: \n"
+                     "  -IOt %d -IOp %d\n",
+                     no_poll_threads, no_pollsets);
+        erts_usage();
+    }
+
+    /* Handled arguments have been marked with NULL. Slide arguments
+       not handled towards the beginning of argv. */
+    for (i = 0, j = 0; i < *argc; i++) {
+	if (argv[i])
+	    argv[j++] = argv[i];
+    }
+    *argc = j;
+
+    erts_no_pollsets = no_pollsets;
+    erts_no_poll_threads = no_poll_threads;
+}
+
 void
-ERTS_CIO_EXPORT(erts_init_check_io)(void)
+erts_init_check_io(int *argc, char **argv)
 {
+    int j, concurrent_waiters, no_poll_threads;
     ERTS_CT_ASSERT((INT_MIN & (ERL_NIF_SELECT_STOP_CALLED |
                                ERL_NIF_SELECT_STOP_SCHEDULED |
                                ERL_NIF_SELECT_INVALID_EVENT |
                                ERL_NIF_SELECT_FAILED)) == 0);
-    erts_smp_atomic_init_nob(&erts_check_io_time, 0);
-    erts_smp_atomic_init_nob(&pollset.in_poll_wait, 0);
-
-    ERTS_CIO_POLL_INIT();
-    pollset.ps = ERTS_CIO_NEW_POLLSET();
-
-    pollset.active_fd.six = 0;
-    pollset.active_fd.eix = 0;
-    erts_smp_atomic32_init_nob(&pollset.active_fd.no, 0);
-    pollset.active_fd.size = ERTS_ACTIVE_FD_INC;
-    pollset.active_fd.array = erts_alloc(ERTS_ALC_T_ACTIVE_FD_ARR,
-					 sizeof(ErtsSysFdType)*ERTS_ACTIVE_FD_INC);
-#ifdef DEBUG
-    {
-	int i;
-	for (i = 0; i < ERTS_ACTIVE_FD_INC; i++)
-	    pollset.active_fd.array[i] = ERTS_SYS_FD_INVALID;
-    }
+
+
+    erts_poll_init(&concurrent_waiters);
+#if ERTS_POLL_USE_FALLBACK
+    erts_poll_init_flbk(NULL);
 #endif
 
+    parse_args(argc, argv, concurrent_waiters);
 
-#ifdef ERTS_SMP
-    init_removed_fd_alloc();
-    pollset.removed_list = NULL;
-    erts_smp_spinlock_init(&pollset.removed_list_lock, "pollset_rm_list", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
-    {
-        int i;
-        for (i=0; i<DRV_EV_STATE_LOCK_CNT; i++) {
-            erts_smp_mtx_init(&drv_ev_state_locks[i].lck, "drv_ev_state", make_small(i),
-                ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
-        }
-    }
+    /* Create the actual pollsets */
+    pollsetv = erts_alloc(ERTS_ALC_T_POLLSET,sizeof(ErtsPollSet *) * erts_no_pollsets);
+
+    for (j=0; j < erts_no_pollsets; j++)
+        pollsetv[j] = erts_poll_create_pollset(j);
+
+#if ERTS_POLL_USE_FALLBACK
+    flbk_pollset = erts_poll_create_pollset_flbk(-1);
 #endif
+
+    no_poll_threads = erts_no_poll_threads;
+#if ERTS_POLL_USE_FALLBACK
+    no_poll_threads++;
+#endif
+
+    psiv = erts_alloc(ERTS_ALC_T_POLLSET, sizeof(ErtsPollThread) * no_poll_threads);
+
+#if ERTS_POLL_USE_FALLBACK
+    psiv[0].pollres_len = ERTS_CHECK_IO_POLL_RES_LEN;
+    psiv[0].pollres = erts_alloc(ERTS_ALC_T_POLLSET,
+        sizeof(ErtsPollResFd) * ERTS_CHECK_IO_POLL_RES_LEN);
+    psiv[0].ps = get_fallback();
+    psiv++;
+#endif
+
+    for (j = 0; j < erts_no_poll_threads; j++) {
+        psiv[j].pollres_len = ERTS_CHECK_IO_POLL_RES_LEN;
+        psiv[j].pollres = erts_alloc(ERTS_ALC_T_POLLSET,
+                                      sizeof(ErtsPollResFd) * ERTS_CHECK_IO_POLL_RES_LEN);
+        psiv[j].ps = pollsetv[j % erts_no_pollsets];
+    }
+
+    for (j=0; j < ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT; j++) {
+        erts_mtx_init(&drv_ev_state.locks[j].lck, "drv_ev_state", make_small(j),
+                          ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
+    }
+
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    max_fds = ERTS_CIO_POLL_MAX_FDS();
-    erts_smp_atomic_init_nob(&drv_ev_state_len, 0);
-    drv_ev_state = NULL;
-    erts_smp_mtx_init(&drv_ev_state_grow_lock, "drv_ev_state_grow", NIL,
+    drv_ev_state.max_fds = erts_poll_max_fds();
+    erts_atomic_init_nob(&drv_ev_state.len, 0);
+    drv_ev_state.v = NULL;
+    erts_mtx_init(&drv_ev_state.grow_lock, "drv_ev_state_grow", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
 #else
     {
@@ -2587,153 +2080,182 @@ ERTS_CIO_EXPORT(erts_init_check_io)(void)
 	hf.cmp = &drv_ev_state_cmp;
 	hf.alloc = &drv_ev_state_alloc;
 	hf.free = &drv_ev_state_free;
-	num_state_prealloc = 0;
-	state_prealloc_first = NULL;
-	erts_smp_spinlock_init(&state_prealloc_lock,"state_prealloc", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
-
-	safe_hash_init(ERTS_ALC_T_DRV_EV_STATE, &drv_ev_state_tab, "drv_ev_state_tab",
+	drv_ev_state.num_prealloc = 0;
+	drv_ev_state.prealloc_first = NULL;
+	erts_spinlock_init(&drv_ev_state.prealloc_lock, "state_prealloc", NIL,
+                               ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
+	safe_hash_init(ERTS_ALC_T_DRV_EV_STATE, &drv_ev_state.tab, "drv_ev_state_tab",
             ERTS_LOCK_FLAGS_CATEGORY_IO, DRV_EV_STATE_HTAB_SIZE, hf);
     }
 #endif
 }
 
 int
-ERTS_CIO_EXPORT(erts_check_io_max_files)(void)
+erts_check_io_max_files(void)
 {
 #ifdef  ERTS_SYS_CONTINOUS_FD_NUMBERS
-    return max_fds;
+    return drv_ev_state.max_fds;
 #else
-    return ERTS_POLL_EXPORT(erts_poll_max_fds)();
+    return erts_poll_max_fds();
 #endif
 }
 
 Uint
-ERTS_CIO_EXPORT(erts_check_io_size)(void)
+erts_check_io_size(void)
 {
-    Uint res;
+    Uint res = 0;
     ErtsPollInfo pi;
-    ERTS_CIO_POLL_INFO(pollset.ps, &pi);
-    res = pi.memory_size;
+    int i;
+
+#if ERTS_POLL_USE_FALLBACK
+    erts_poll_info(get_fallback(), &pi);
+    res += pi.memory_size;
+#endif
+
+    for (i = 0; i < erts_no_pollsets; i++) {
+        erts_poll_info(pollsetv[i], &pi);
+        res += pi.memory_size;
+    }
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    res += sizeof(ErtsDrvEventState) * erts_smp_atomic_read_nob(&drv_ev_state_len);
+    res += sizeof(ErtsDrvEventState) * erts_atomic_read_nob(&drv_ev_state.len);
 #else
-    res += safe_hash_table_sz(&drv_ev_state_tab);
+    res += safe_hash_table_sz(&drv_ev_state.tab);
     {
 	SafeHashInfo hi;
-	safe_hash_get_info(&hi, &drv_ev_state_tab);
+	safe_hash_get_info(&hi, &drv_ev_state.tab);
 	res += hi.objs * sizeof(ErtsDrvEventState);
     }
-    erts_smp_spin_lock(&state_prealloc_lock);
-    res += num_state_prealloc * sizeof(ErtsDrvEventState);
-    erts_smp_spin_unlock(&state_prealloc_lock);
+    erts_spin_lock(&drv_ev_state.prealloc_lock);
+    res += drv_ev_state.num_prealloc * sizeof(ErtsDrvEventState);
+    erts_spin_unlock(&drv_ev_state.prealloc_lock);
 #endif
     return res;
 }
 
 Eterm
-ERTS_CIO_EXPORT(erts_check_io_info)(void *proc)
+erts_check_io_info(void *proc)
 {
     Process *p = (Process *) proc;
-    Eterm tags[16], values[16], res;
-    Uint sz, *szp, *hp, **hpp, memory_size;
-    Sint i;
-    ErtsPollInfo pi;
-    erts_aint_t cio_time = erts_smp_atomic_read_acqb(&erts_check_io_time);
-    int active_fds = (int) erts_smp_atomic32_read_acqb(&pollset.active_fd.no);
+    Eterm tags[16], values[16], res, list = NIL;
+    Uint sz, *szp, *hp, **hpp;
+    ErtsPollInfo *piv;
+    Sint i, j = 0, len;
+    int no_pollsets = erts_no_pollsets + ERTS_POLL_USE_FALLBACK;
+    ERTS_CT_ASSERT(ERTS_POLL_USE_FALLBACK == 0 || ERTS_POLL_USE_FALLBACK == 1);
 
-    while (1) {
-	erts_aint_t post_cio_time;
-	int post_active_fds;
+    piv = erts_alloc(ERTS_ALC_T_TMP, sizeof(ErtsPollInfo) * no_pollsets);
 
-	ERTS_CIO_POLL_INFO(pollset.ps, &pi);
+#if ERTS_POLL_USE_FALLBACK
+    erts_poll_info_flbk(get_fallback(), &piv[0]);
+    piv[0].poll_threads = 1;
+    piv[0].active_fds = 0;
+    piv++;
+#endif
 
-	post_cio_time = erts_smp_atomic_read_mb(&erts_check_io_time);
-	post_active_fds = (int) erts_smp_atomic32_read_acqb(&pollset.active_fd.no);
-	if (cio_time == post_cio_time && active_fds == post_active_fds)
-	    break;
-	cio_time = post_cio_time;
-	active_fds = post_active_fds;
+    for (j = 0; j < erts_no_pollsets; j++) {
+        erts_poll_info(pollsetv[j], &piv[j]);
+        piv[j].active_fds = 0;
+        piv[j].poll_threads = erts_no_poll_threads / erts_no_pollsets;
+        if (erts_no_poll_threads % erts_no_pollsets > j)
+            piv[j].poll_threads++;
     }
 
-    memory_size = pi.memory_size;
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    memory_size += sizeof(ErtsDrvEventState) * erts_smp_atomic_read_nob(&drv_ev_state_len);
+    i = 0;
+    erts_mtx_lock(&drv_ev_state.grow_lock);
+    len = erts_atomic_read_nob(&drv_ev_state.len);
+    for (i = 0; i < ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT; i++) {
+        erts_mtx_lock(&drv_ev_state.locks[i].lck);
+        for (j = i; j < len; j+=ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT) {
+            ErtsDrvEventState *state = get_drv_ev_state(j);
+            int pollsetid = get_pollset_id(j);
+            ASSERT(fd_mtx(j) == &drv_ev_state.locks[i].lck);
+            if (state->flags & ERTS_EV_FLAG_FALLBACK)
+                pollsetid = -1;
+            if (state->driver.select
+                && (state->type == ERTS_EV_TYPE_DRV_SEL)
+                && (is_iotask_active(&state->driver.select->iniotask)
+                    || is_iotask_active(&state->driver.select->outiotask)))
+                piv[pollsetid].active_fds++;
+        }
+        erts_mtx_unlock(&drv_ev_state.locks[i].lck);
+    }
+    erts_mtx_unlock(&drv_ev_state.grow_lock);
+
+    piv[0].memory_size += sizeof(ErtsDrvEventState) * erts_atomic_read_nob(&drv_ev_state.len);
 #else
-    memory_size += safe_hash_table_sz(&drv_ev_state_tab);
+    piv[0].memory_size += safe_hash_table_sz(&drv_ev_state.tab);
     {
-	SafeHashInfo hi;
-	safe_hash_get_info(&hi, &drv_ev_state_tab);
-	memory_size += hi.objs * sizeof(ErtsDrvEventState);
+        SafeHashInfo hi;
+        safe_hash_get_info(&hi, &drv_ev_state.tab);
+        piv[0].memory_size += hi.objs * sizeof(ErtsDrvEventState);
     }
-    erts_smp_spin_lock(&state_prealloc_lock);
-    memory_size += num_state_prealloc * sizeof(ErtsDrvEventState);
-    erts_smp_spin_unlock(&state_prealloc_lock);
+    erts_spin_lock(&drv_ev_state.prealloc_lock);
+    piv[0].memory_size += drv_ev_state.num_prealloc * sizeof(ErtsDrvEventState);
+    erts_spin_unlock(&drv_ev_state.prealloc_lock);
 #endif
 
     hpp = NULL;
     szp = &sz;
     sz = 0;
 
- bld_it:
-    i = 0;
+    piv -= ERTS_POLL_USE_FALLBACK;
 
-    tags[i] = erts_bld_atom(hpp, szp, "name");
-    values[i++] = erts_bld_atom(hpp, szp, "erts_poll");
+ bld_it:
 
-    tags[i] = erts_bld_atom(hpp, szp, "primary");
-    values[i++] = erts_bld_atom(hpp, szp, pi.primary);
+    for (j = no_pollsets-1; j >= 0; j--) {
+        i = 0;
 
-    tags[i] = erts_bld_atom(hpp, szp, "fallback");
-    values[i++] = erts_bld_atom(hpp, szp, pi.fallback ? pi.fallback : "false");
+        tags[i] = erts_bld_atom(hpp, szp, "name");
+        values[i++] = erts_bld_atom(hpp, szp, "erts_poll");
 
-    tags[i] = erts_bld_atom(hpp, szp, "kernel_poll");
-    values[i++] = erts_bld_atom(hpp, szp,
-				pi.kernel_poll ? pi.kernel_poll : "false");
+        tags[i] = erts_bld_atom(hpp, szp, "primary");
+        values[i++] = erts_bld_atom(hpp, szp, piv[j].primary);
 
-    tags[i] = erts_bld_atom(hpp, szp, "memory_size");
-    values[i++] = erts_bld_uint(hpp, szp, memory_size);
+        tags[i] = erts_bld_atom(hpp, szp, "kernel_poll");
+        values[i++] = erts_bld_atom(hpp, szp,
+                                    piv[j].kernel_poll ? piv[j].kernel_poll : "false");
 
-    tags[i] = erts_bld_atom(hpp, szp, "total_poll_set_size");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.poll_set_size);
+        tags[i] = erts_bld_atom(hpp, szp, "memory_size");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].memory_size);
 
-    if (pi.fallback) {
-	tags[i] = erts_bld_atom(hpp, szp, "fallback_poll_set_size");
-	values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.fallback_poll_set_size);
-    }
+        tags[i] = erts_bld_atom(hpp, szp, "total_poll_set_size");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].poll_set_size);
 
-    tags[i] = erts_bld_atom(hpp, szp, "lazy_updates");
-    values[i++] = pi.lazy_updates ? am_true : am_false;
+        tags[i] = erts_bld_atom(hpp, szp, "lazy_updates");
+        values[i++] = piv[j].lazy_updates ? am_true : am_false;
 
-    if (pi.lazy_updates) {
-	tags[i] = erts_bld_atom(hpp, szp, "pending_updates");
-	values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.pending_updates);
-    }
+        tags[i] = erts_bld_atom(hpp, szp, "pending_updates");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].pending_updates);
 
-    tags[i] = erts_bld_atom(hpp, szp, "batch_updates");
-    values[i++] = pi.batch_updates ? am_true : am_false;
+        tags[i] = erts_bld_atom(hpp, szp, "batch_updates");
+        values[i++] = piv[j].batch_updates ? am_true : am_false;
 
-    tags[i] = erts_bld_atom(hpp, szp, "concurrent_updates");
-    values[i++] = pi.concurrent_updates ? am_true : am_false;
+        tags[i] = erts_bld_atom(hpp, szp, "concurrent_updates");
+        values[i++] = piv[j].concurrent_updates ? am_true : am_false;
 
-    tags[i] = erts_bld_atom(hpp, szp, "max_fds");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.max_fds);
+        tags[i] = erts_bld_atom(hpp, szp, "fallback");
+        values[i++] = piv[j].is_fallback ? am_true : am_false;
 
-    tags[i] = erts_bld_atom(hpp, szp, "active_fds");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) active_fds);
+        tags[i] = erts_bld_atom(hpp, szp, "max_fds");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].max_fds);
 
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-    tags[i] = erts_bld_atom(hpp, szp, "no_avoided_wakeups");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.no_avoided_wakeups);
+        tags[i] = erts_bld_atom(hpp, szp, "active_fds");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].active_fds);
 
-    tags[i] = erts_bld_atom(hpp, szp, "no_avoided_interrupts");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.no_avoided_interrupts);
+        tags[i] = erts_bld_atom(hpp, szp, "poll_threads");
+        values[i++] = erts_bld_uint(hpp, szp, piv[j].poll_threads);
 
-    tags[i] = erts_bld_atom(hpp, szp, "no_interrupt_timed");
-    values[i++] = erts_bld_uint(hpp, szp, (Uint) pi.no_interrupt_timed);
-#endif
+        res = erts_bld_2tup_list(hpp, szp, i, tags, values);
 
-    res = erts_bld_2tup_list(hpp, szp, i, tags, values);
+        if (!hpp) {
+            *szp += 2;
+        }
+        else {
+            list = CONS(*hpp, res, list);
+            *hpp += 2;
+        }
+    }
 
     if (!hpp) {
 	hp = HAlloc(p, sz);
@@ -2742,386 +2264,442 @@ ERTS_CIO_EXPORT(erts_check_io_info)(void *proc)
 	goto bld_it;
     }
 
-    return res;
+    erts_free(ERTS_ALC_T_TMP, piv);
+
+    return list;
 }
 
 static ERTS_INLINE ErtsPollEvents
-print_events(ErtsPollEvents ev)
+print_events(erts_dsprintf_buf_t *dsbufp, ErtsPollEvents ev)
 {
     int first = 1;
+    if(ev == ERTS_POLL_EV_NONE) {
+        erts_dsprintf(dsbufp, "N/A");
+        return 0;
+    }
     if(ev & ERTS_POLL_EV_IN) {
 	ev &= ~ERTS_POLL_EV_IN;
-	erts_printf("%s%s", first ? "" : "|", "IN");
+	erts_dsprintf(dsbufp, "%s%s", first ? "" : "|", "IN");
 	first = 0;
     }
     if(ev & ERTS_POLL_EV_OUT) {
 	ev &= ~ERTS_POLL_EV_OUT;
-	erts_printf("%s%s", first ? "" : "|", "OUT");
+	erts_dsprintf(dsbufp, "%s%s", first ? "" : "|", "OUT");
 	first = 0;
     }
     /* The following should not appear... */
     if(ev & ERTS_POLL_EV_NVAL) {
-	erts_printf("%s%s", first ? "" : "|", "NVAL");
+	erts_dsprintf(dsbufp, "%s%s", first ? "" : "|", "NVAL");
 	first = 0;
     }
     if(ev & ERTS_POLL_EV_ERR) {
-	erts_printf("%s%s", first ? "" : "|", "ERR");
+	erts_dsprintf(dsbufp, "%s%s", first ? "" : "|", "ERR");
 	first = 0;
     }
     if (ev)
-	erts_printf("%s0x%b32x", first ? "" : "|", (Uint32) ev);
+	erts_dsprintf(dsbufp, "%s0x%b32x", first ? "" : "|", (Uint32) ev);
     return ev;
 }
 
+static ERTS_INLINE void
+print_flags(erts_dsprintf_buf_t *dsbufp, EventStateFlags f)
+{
+    const char* delim = "";
+    if(f & ERTS_EV_FLAG_USED) {
+	erts_dsprintf(dsbufp, "%s","USED");
+	delim = "|";
+    }
+    if(f & ERTS_EV_FLAG_FALLBACK) {
+	erts_dsprintf(dsbufp, "%s%s", delim, "FLBK");
+        delim = "|";
+    }
+}
+
+#ifdef DEBUG_PRINT_MODE
+
+static ERTS_INLINE char *
+drvmode2str(int mode) {
+    switch (mode) {
+    case ERL_DRV_READ|ERL_DRV_USE: return "READ|USE";
+    case ERL_DRV_WRITE|ERL_DRV_USE: return "WRITE|USE";
+    case ERL_DRV_READ|ERL_DRV_WRITE|ERL_DRV_USE: return "READ|WRITE|USE";
+    case ERL_DRV_USE: return "USE";
+    case ERL_DRV_READ: return "READ";
+    case ERL_DRV_WRITE: return "WRITE";
+    case ERL_DRV_READ|ERL_DRV_WRITE: return "READ|WRITE";
+    default: return "UNKNOWN";
+    }
+}
+
+static ERTS_INLINE char *
+nifmode2str(enum ErlNifSelectFlags mode) {
+    switch (mode) {
+    case ERL_NIF_SELECT_READ: return "READ";
+    case ERL_NIF_SELECT_WRITE: return "WRITE";
+    case ERL_NIF_SELECT_STOP: return "STOP";
+    default: return "UNKNOWN";
+    }
+}
+
+#endif
+
 typedef struct {
     int used_fds;
     int num_errors;
     int no_driver_select_structs;
-    int no_driver_event_structs;
+    int no_enif_select_structs;
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
     int internal_fds;
     ErtsPollEvents *epep;
 #endif
 } IterDebugCounters;
 
-static void doit_erts_check_io_debug(void *vstate, void *vcounters)
+static int erts_debug_print_checkio_state(erts_dsprintf_buf_t *dsbufp,
+                                          ErtsDrvEventState *state,
+                                          ErtsPollEvents ep_events,
+                                          int internal)
 {
-    ErtsDrvEventState *state = (ErtsDrvEventState *) vstate;
-    IterDebugCounters *counters = (IterDebugCounters *) vcounters;
-    ErtsPollEvents cio_events = state->events;
-    ErtsSysFdType fd = state->fd;
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    int internal = 0;
-    ErtsPollEvents ep_events = counters->epep[(int) fd];
-#endif
-    int err = 0;
-
 #if defined(HAVE_FSTAT) && !defined(NO_FSTAT_ON_SYS_FD_TYPE)
     struct stat stat_buf;
 #endif
-
-    if (state->driver.select)
-	counters->no_driver_select_structs++;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    if (state->driver.event)
-	counters->no_driver_event_structs++;
-#endif
-
+    ErtsSysFdType fd = state->fd;
+    ErtsPollEvents cio_events = state->events;
+    int err = 0;
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    if (state->events || ep_events) {
-	if (ep_events & ERTS_POLL_EV_NVAL) {
-	    ep_events &= ~ERTS_POLL_EV_NVAL;
-	    internal = 1;
-	    counters->internal_fds++;
-	}
-	else
-	    counters->used_fds++;
-#else
-    if (state->events) {
-	counters->used_fds++;
+    ErtsPollEvents aio_events = state->active_events;
 #endif
-	
-	erts_printf("fd=%d ", (int) fd);
-	
+    erts_dsprintf(dsbufp, "pollset=%d fd=%d ",
+                state->flags & ERTS_EV_FLAG_FALLBACK ? -1 : get_pollset_id(fd), (int) fd);
+
 #if defined(HAVE_FSTAT) && !defined(NO_FSTAT_ON_SYS_FD_TYPE)
-	if (fstat((int) fd, &stat_buf) < 0)
-	    erts_printf("type=unknown ");
-	else {
-	    erts_printf("type=");
+    if (fstat((int) fd, &stat_buf) < 0)
+        erts_dsprintf(dsbufp, "type=unknown ");
+    else {
+        erts_dsprintf(dsbufp, "type=");
 #ifdef S_ISSOCK
-	    if (S_ISSOCK(stat_buf.st_mode))
-		erts_printf("sock ");
-	    else
+        if (S_ISSOCK(stat_buf.st_mode))
+            erts_dsprintf(dsbufp, "sock ");
+        else
 #endif
 #ifdef S_ISFIFO
 	    if (S_ISFIFO(stat_buf.st_mode))
-		erts_printf("fifo ");
+		erts_dsprintf(dsbufp, "fifo ");
 	    else
 #endif
 #ifdef S_ISCHR
-	    if (S_ISCHR(stat_buf.st_mode))
-		erts_printf("chr ");
-	    else
+                if (S_ISCHR(stat_buf.st_mode))
+                    erts_dsprintf(dsbufp, "chr ");
+                else
 #endif
 #ifdef S_ISDIR
-	    if (S_ISDIR(stat_buf.st_mode))
-		erts_printf("dir ");
-	    else
+                    if (S_ISDIR(stat_buf.st_mode))
+                        erts_dsprintf(dsbufp, "dir ");
+                    else
 #endif
 #ifdef S_ISBLK
-	    if (S_ISBLK(stat_buf.st_mode))
-		erts_printf("blk ");
-	    else
+                        if (S_ISBLK(stat_buf.st_mode))
+                            erts_dsprintf(dsbufp, "blk ");
+                        else
 #endif
 #ifdef S_ISREG
-	    if (S_ISREG(stat_buf.st_mode))
-		erts_printf("reg ");
-	    else
+                            if (S_ISREG(stat_buf.st_mode))
+                                erts_dsprintf(dsbufp, "reg ");
+                            else
 #endif
 #ifdef S_ISLNK
-	    if (S_ISLNK(stat_buf.st_mode))
-		erts_printf("lnk ");
-	    else
+                                if (S_ISLNK(stat_buf.st_mode))
+                                    erts_dsprintf(dsbufp, "lnk ");
+                                else
 #endif
 #ifdef S_ISDOOR
-	    if (S_ISDOOR(stat_buf.st_mode))
-		erts_printf("door ");
-	    else
+                                    if (S_ISDOOR(stat_buf.st_mode))
+                                        erts_dsprintf(dsbufp, "door ");
+                                    else
 #endif
 #ifdef S_ISWHT
-	    if (S_ISWHT(stat_buf.st_mode))
-		erts_printf("wht ");
-	    else
+                                        if (S_ISWHT(stat_buf.st_mode))
+                                            erts_dsprintf(dsbufp, "wht ");
+                                        else
 #endif
 #ifdef S_ISXATTR
-	    if (S_ISXATTR(stat_buf.st_mode))
-		erts_printf("xattr ");
-	    else
+                                            if (S_ISXATTR(stat_buf.st_mode))
+                                                erts_dsprintf(dsbufp, "xattr ");
+                                            else
 #endif
-		erts_printf("unknown ");
-	}
+                                                erts_dsprintf(dsbufp, "unknown ");
+    }
 #else
-	erts_printf("type=unknown ");
+    erts_dsprintf(dsbufp, "type=unknown ");
 #endif
 
-	if (state->type == ERTS_EV_TYPE_DRV_SEL) {
-	    erts_printf("driver_select ");
-	    
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	    if (internal) {
-		erts_printf("internal ");
-		err = 1;		    
-	    }
-	    
-	    if (cio_events == ep_events) {
-		erts_printf("ev=");
-		if (print_events(cio_events) != 0)
-		    err = 1;
-	    }
-	    else {
-		err = 1;
-		erts_printf("cio_ev=");
-		print_events(cio_events);
-		erts_printf(" ep_ev=");
-		print_events(ep_events);
-	    }
-#else
-	    if (print_events(cio_events) != 0)
-		err = 1;
-#endif
-	    erts_printf(" ");
-	    if (cio_events & ERTS_POLL_EV_IN) {
-		Eterm id = state->driver.select->inport;
-		if (is_nil(id)) {
-		    erts_printf("inport=none inname=none indrv=none ");
-		    err = 1;
-		}
-		else {
-		    ErtsPortNames *pnp = erts_get_port_names(id, ERTS_INVALID_ERL_DRV_PORT);
-		    erts_printf(" inport=%T inname=%s indrv=%s ",
-				id,
-				pnp->name ? pnp->name : "unknown",
-				(pnp->driver_name
-				 ? pnp->driver_name
-				 : "unknown"));
-		    erts_free_port_names(pnp);
-		}
-	    }
-	    if (cio_events & ERTS_POLL_EV_OUT) {
-		Eterm id = state->driver.select->outport;
-		if (is_nil(id)) {
-		    erts_printf("outport=none outname=none outdrv=none ");
-		    err = 1;
-		}
-		else {
-		    ErtsPortNames *pnp = erts_get_port_names(id, ERTS_INVALID_ERL_DRV_PORT);
-		    erts_printf(" outport=%T outname=%s outdrv=%s ",
-				id,
-				pnp->name ? pnp->name : "unknown",
-				(pnp->driver_name
-				 ? pnp->driver_name
-				 : "unknown"));
-		    erts_free_port_names(pnp);
-		}
-	    }
-	}
-        else if (state->type == ERTS_EV_TYPE_NIF) {
-            ErtsResource* r;
-            erts_printf("enif_select ");
+    if (state->type == ERTS_EV_TYPE_DRV_SEL) {
+        erts_dsprintf(dsbufp, "driver_select ");
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-            if (internal) {
-                erts_printf("internal ");
-                err = 1;
-            }
-
+        if (internal) {
+            erts_dsprintf(dsbufp, "internal ");
+            err = 1;
+        }
+        if (aio_events == cio_events) {
             if (cio_events == ep_events) {
-                erts_printf("ev=");
-                if (print_events(cio_events) != 0)
+                erts_dsprintf(dsbufp, "ev=");
+                if (print_events(dsbufp, cio_events) != 0)
                     err = 1;
             }
             else {
+                ErtsPollEvents ev = cio_events;
+                if (ev != ep_events && ep_events != ERTS_POLL_EV_NONE)
+                    err = 1;
+                erts_dsprintf(dsbufp, "cio_ev=");
+                print_events(dsbufp, cio_events);
+                erts_dsprintf(dsbufp, " ep_ev=");
+                print_events(dsbufp, ep_events);
+            }
+        } else {
+            erts_dsprintf(dsbufp, "cio_ev=");
+            print_events(dsbufp, cio_events);
+            erts_dsprintf(dsbufp, " aio_ev=");
+            print_events(dsbufp, aio_events);
+            if ((aio_events != ep_events && ep_events != ERTS_POLL_EV_NONE) ||
+                (aio_events != 0 && ep_events == ERTS_POLL_EV_NONE)) {
+                erts_dsprintf(dsbufp, " ep_ev=");
+                print_events(dsbufp, ep_events);
                 err = 1;
-                erts_printf("cio_ev=");
-                print_events(cio_events);
-                erts_printf(" ep_ev=");
-                print_events(ep_events);
             }
+        }
 #else
-            if (print_events(cio_events) != 0)
+        if (print_events(dsbufp, cio_events) != 0)
+            err = 1;
+#endif
+        erts_dsprintf(dsbufp, " ");
+        if (cio_events & ERTS_POLL_EV_IN) {
+            Eterm id = state->driver.select->inport;
+            if (is_nil(id)) {
+                erts_dsprintf(dsbufp, "inport=none inname=none indrv=none ");
                 err = 1;
-#endif
-            erts_printf(" inpid=%T dd_cnt=%b32d", state->driver.nif->in.pid,
-                        state->driver.nif->in.ddeselect_cnt);
-            erts_printf(" outpid=%T dd_cnt=%b32d", state->driver.nif->out.pid,
-                        state->driver.nif->out.ddeselect_cnt);
-            r = state->driver.stop.resource;
-            erts_printf(" resource=%p(%T:%T)", r, r->type->module, r->type->name);
+            }
+            else {
+                ErtsPortNames *pnp = erts_get_port_names(id, ERTS_INVALID_ERL_DRV_PORT);
+                erts_dsprintf(dsbufp, " inport=%T inname=%s indrv=%s ",
+                            id,
+                            pnp->name ? pnp->name : "unknown",
+                            (pnp->driver_name
+                             ? pnp->driver_name
+                             : "unknown"));
+                erts_free_port_names(pnp);
+            }
         }
-#if ERTS_CIO_HAVE_DRV_EVENT
-	else if (state->type == ERTS_EV_TYPE_DRV_EV) {
-	    Eterm id;
-	    erts_printf("driver_event ");
+        if (cio_events & ERTS_POLL_EV_OUT) {
+            Eterm id = state->driver.select->outport;
+            if (is_nil(id)) {
+                erts_dsprintf(dsbufp, "outport=none outname=none outdrv=none ");
+                err = 1;
+            }
+            else {
+                ErtsPortNames *pnp = erts_get_port_names(id, ERTS_INVALID_ERL_DRV_PORT);
+                erts_dsprintf(dsbufp, " outport=%T outname=%s outdrv=%s ",
+                            id,
+                            pnp->name ? pnp->name : "unknown",
+                            (pnp->driver_name
+                             ? pnp->driver_name
+                             : "unknown"));
+                erts_free_port_names(pnp);
+            }
+        }
+    }
+    else if (state->type == ERTS_EV_TYPE_NIF) {
+        ErtsResource* r;
+        erts_dsprintf(dsbufp, "enif_select ");
+
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	    if (internal) {
-		erts_printf("internal ");
-		err = 1;		    
-	    }
-	    if (cio_events == ep_events) {
-		erts_printf("ev=0x%b32x", (Uint32) cio_events);
-	    }
-	    else {
-		err = 1;
-		erts_printf("cio_ev=0x%b32x", (Uint32) cio_events);
-		erts_printf(" ep_ev=0x%b32x", (Uint32) ep_events);
-	    }
+        if (internal) {
+            erts_dsprintf(dsbufp, "internal ");
+            err = 1;
+        }
+
+        if (cio_events == ep_events) {
+            erts_dsprintf(dsbufp, "ev=");
+            if (print_events(dsbufp, cio_events) != 0)
+                err = 1;
+        }
+        else {
+            err = 1;
+            erts_dsprintf(dsbufp, "cio_ev=");
+            print_events(dsbufp, cio_events);
+            erts_dsprintf(dsbufp, " ep_ev=");
+            print_events(dsbufp, ep_events);
+        }
 #else
-	    erts_printf("ev=0x%b32x", (Uint32) cio_events);
+        if (print_events(dsbufp, cio_events) != 0)
+            err = 1;
 #endif
-	    id = state->driver.event->port;
-	    if (is_nil(id)) {
-		erts_printf(" port=none name=none drv=none ");
-		err = 1;
-	    }
-	    else {
-		ErtsPortNames *pnp = erts_get_port_names(id, ERTS_INVALID_ERL_DRV_PORT);
-		erts_printf(" port=%T name=%s drv=%s ",
-			    id,
-			    pnp->name ? pnp->name : "unknown",
-			    (pnp->driver_name
-			     ? pnp->driver_name
-			     : "unknown"));
-		erts_free_port_names(pnp);
-	    }
-	}
+        erts_dsprintf(dsbufp, " inpid=%T", state->driver.nif->in.pid);
+        erts_dsprintf(dsbufp, " outpid=%T", state->driver.nif->out.pid);
+        r = state->driver.stop.resource;
+        erts_dsprintf(dsbufp, " resource=%p(%T:%T)", r, r->type->module, r->type->name);
+    }
+#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
+    else if (internal) {
+        erts_dsprintf(dsbufp, "internal ");
+        if (cio_events) {
+            err = 1;
+            erts_dsprintf(dsbufp, "cio_ev=");
+            print_events(dsbufp, cio_events);
+        }
+        if (ep_events) {
+            erts_dsprintf(dsbufp, "ep_ev=");
+            print_events(dsbufp, ep_events);
+        }
+    }
 #endif
+    else {
+        err = 1;
+        erts_dsprintf(dsbufp, "control_type=%d ", (int)state->type);
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	else if (internal) {
-	    erts_printf("internal ");
-	    if (cio_events) {
-		err = 1;
-		erts_printf("cio_ev=");
-		print_events(cio_events);
-	    }
-	    if (ep_events) {
-		erts_printf("ep_ev=");
-		print_events(ep_events);
-	    }
-	}
+        if (cio_events == ep_events) {
+            erts_dsprintf(dsbufp, "ev=");
+            print_events(dsbufp, cio_events);
+        }
+        else {
+            erts_dsprintf(dsbufp, "cio_ev="); print_events(dsbufp, cio_events);
+            erts_dsprintf(dsbufp, " ep_ev="); print_events(dsbufp, ep_events);
+        }
+#else
+        erts_dsprintf(dsbufp, "ev=0x%b32x", (Uint32) cio_events);
 #endif
-	else {
-	    err = 1;
-	    erts_printf("control_type=%d ", (int)state->type);
+    }
+
+    erts_dsprintf(dsbufp, " flags="); print_flags(dsbufp, state->flags);
+    if (err) {
+        erts_dsprintf(dsbufp, " ERROR");
+    }
+    erts_dsprintf(dsbufp, "\r\n");
+    return err;
+}
+
+static void doit_erts_check_io_debug(void *vstate, void *vcounters,
+                                     erts_dsprintf_buf_t *dsbufp)
+{
+    ErtsDrvEventState *state = (ErtsDrvEventState *) vstate;
+    IterDebugCounters *counters = (IterDebugCounters *) vcounters;
+    int internal = 0;
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-	    if (cio_events == ep_events) {
-                erts_printf("ev=");
-                print_events(cio_events);
-	    }
-	    else {
-		erts_printf("cio_ev="); print_events(cio_events);
-		erts_printf(" ep_ev="); print_events(ep_events);
-	    }
+    ErtsSysFdType fd = state->fd;
+    ErtsPollEvents ep_events = counters->epep[(int) fd];
 #else
-	    erts_printf("ev=0x%b32x", (Uint32) cio_events);
+    ErtsPollEvents ep_events = ERTS_POLL_EV_NONE;
 #endif
+
+    if (state->driver.select) {
+	counters->no_driver_select_structs++;
+        ASSERT(state->events || (ep_events != 0 && ep_events != ERTS_POLL_EV_NONE));
+    }
+    if (state->driver.nif) {
+        counters->no_enif_select_structs++;
+        ASSERT(state->events || (ep_events != 0 && ep_events != ERTS_POLL_EV_NONE));
+    }
+
+#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
+    if (state->events || (ep_events != 0 && ep_events != ERTS_POLL_EV_NONE)) {
+	if (ep_events & ERTS_POLL_EV_NVAL) {
+	    ep_events &= ~ERTS_POLL_EV_NVAL;
+	    internal = 1;
+	    counters->internal_fds++;
 	}
-	
-	if (err) {
+	else
+	    counters->used_fds++;
+#else
+    if (state->events) {
+	counters->used_fds++;
+#endif
+	if (erts_debug_print_checkio_state(dsbufp, state, ep_events, internal)) {
 	    counters->num_errors++;
-	    erts_printf(" ERROR");
 	}
-	erts_printf("\n");
     }
 }
-    
+
+/* ciodpi can be NULL when called from etp-commands */
 int
-ERTS_CIO_EXPORT(erts_check_io_debug)(ErtsCheckIoDebugInfo *ciodip)
+erts_check_io_debug(ErtsCheckIoDebugInfo *ciodip)
 {
+    erts_dsprintf_buf_t *dsbufp = erts_create_logger_dsbuf();
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    int fd, len;
+    int fd, len, i;
 #endif
-    IterDebugCounters counters;
+    IterDebugCounters counters = {0};
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
     ErtsDrvEventState null_des;
 
     null_des.driver.select = NULL;
-#if ERTS_CIO_HAVE_DRV_EVENT
-    null_des.driver.event = NULL;
-#endif
+    null_des.driver.nif = NULL;
     null_des.driver.stop.drv_ptr = NULL;
     null_des.events = 0;
-    null_des.remove_cnt = 0;
     null_des.type = ERTS_EV_TYPE_NONE;
+    null_des.flags = 0;
+
+    counters.epep = erts_alloc(ERTS_ALC_T_TMP,
+                               sizeof(ErtsPollEvents)*drv_ev_state.max_fds);
 #endif
 
-    erts_printf("--- fds in pollset --------------------------------------\n");
 
-#if defined(ERTS_SMP) && defined(ERTS_ENABLE_LOCK_CHECK)
+#if defined(ERTS_ENABLE_LOCK_CHECK)
     erts_lc_check_exact(NULL, 0); /* No locks should be locked */
 #endif
 
-    erts_smp_thr_progress_block(); /* stop the world to avoid messy locking */
+    if (ciodip)
+        erts_thr_progress_block(); /* stop the world to avoid messy locking */
 
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    counters.epep = erts_alloc(ERTS_ALC_T_TMP, sizeof(ErtsPollEvents)*max_fds);
-    ERTS_POLL_EXPORT(erts_poll_get_selected_events)(pollset.ps, counters.epep, max_fds);
-    counters.internal_fds = 0;
-#endif
-    counters.used_fds = 0;
-    counters.num_errors = 0;
-    counters.no_driver_select_structs = 0;
-    counters.no_driver_event_structs = 0;
+    len = erts_atomic_read_nob(&drv_ev_state.len);
 
-#ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    len = erts_smp_atomic_read_nob(&drv_ev_state_len);
+#if ERTS_POLL_USE_FALLBACK
+    erts_dsprintf(dsbufp, "--- fds in flbk pollset ---------------------------------\n");
+    erts_poll_get_selected_events_flbk(get_fallback(), counters.epep,
+                                       drv_ev_state.max_fds);
     for (fd = 0; fd < len; fd++) {
-	doit_erts_check_io_debug((void *) &drv_ev_state[fd], (void *) &counters);
+        if (drv_ev_state.v[fd].flags & ERTS_EV_FLAG_FALLBACK)
+            doit_erts_check_io_debug(&drv_ev_state.v[fd], &counters, dsbufp);
+    }
+#endif
+    erts_dsprintf(dsbufp, "--- fds in pollset --------------------------------------\n");
+
+    for (i = 0; i < erts_no_pollsets; i++) {
+        erts_poll_get_selected_events(pollsetv[i],
+                                      counters.epep,
+                                      drv_ev_state.max_fds);
+        for (fd = 0; fd < len; fd++) {
+            if (!(drv_ev_state.v[fd].flags & ERTS_EV_FLAG_FALLBACK)
+                && get_pollset_id(fd) == i)
+                doit_erts_check_io_debug(&drv_ev_state.v[fd], &counters, dsbufp);
+        }
     }
-    for ( ; fd < max_fds; fd++) {
-	null_des.fd = fd;
-	doit_erts_check_io_debug((void *) &null_des, (void *) &counters);
+    for (fd = len ; fd < drv_ev_state.max_fds; fd++) {
+        null_des.fd = fd;
+        doit_erts_check_io_debug(&null_des, &counters, dsbufp);
     }
 #else
-    safe_hash_for_each(&drv_ev_state_tab, &doit_erts_check_io_debug, (void *) &counters);
+    safe_hash_for_each(&drv_ev_state.tab, &doit_erts_check_io_debug,
+                       &counters, dsbufp);
 #endif
 
-    erts_smp_thr_progress_unblock();
+    if (ciodip)
+        erts_thr_progress_unblock();
 
-    ciodip->no_used_fds = counters.used_fds;
-    ciodip->no_driver_select_structs = counters.no_driver_select_structs;
-    ciodip->no_driver_event_structs = counters.no_driver_event_structs;
+    if (ciodip) {
+        ciodip->no_used_fds = counters.used_fds;
+        ciodip->no_driver_select_structs = counters.no_driver_select_structs;
+        ciodip->no_enif_select_structs = counters.no_enif_select_structs;
+    }
 
-    erts_printf("\n");
-    erts_printf("used fds=%d\n", counters.used_fds);
-    erts_printf("Number of driver_select() structures=%d\n", counters.no_driver_select_structs);
-#if ERTS_CIO_HAVE_DRV_EVENT
-    erts_printf("Number of driver_event() structures=%d\n", counters.no_driver_event_structs);
-#endif
+    erts_dsprintf(dsbufp, "\n");
+    erts_dsprintf(dsbufp, "used fds=%d\n", counters.used_fds);
+    erts_dsprintf(dsbufp, "Number of driver_select() structures=%d\n", counters.no_driver_select_structs);
+    erts_dsprintf(dsbufp, "Number of enif_select() structures=%d\n", counters.no_enif_select_structs);
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    erts_printf("internal fds=%d\n", counters.internal_fds);
+    erts_dsprintf(dsbufp, "internal fds=%d\n", counters.internal_fds);
 #endif
-    erts_printf("---------------------------------------------------------\n");
-    fflush(stdout);
+    erts_dsprintf(dsbufp, "---------------------------------------------------------\n");
+    erts_send_error_to_logger_nogl(dsbufp);
 #ifdef ERTS_SYS_CONTINOUS_FD_NUMBERS
     erts_free(ERTS_ALC_T_TMP, (void *) counters.epep);
 #endif
@@ -3130,11 +2708,19 @@ ERTS_CIO_EXPORT(erts_check_io_debug)(ErtsCheckIoDebugInfo *ciodip)
 }
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
-void ERTS_CIO_EXPORT(erts_lcnt_update_cio_locks)(int enable) {
+void erts_lcnt_update_cio_locks(int enable) {
+    int i;
 #ifndef ERTS_SYS_CONTINOUS_FD_NUMBERS
-    erts_lcnt_enable_hash_lock_count(&drv_ev_state_tab, ERTS_LOCK_FLAGS_CATEGORY_IO, enable);
+    erts_lcnt_enable_hash_lock_count(&drv_ev_state.tab, ERTS_LOCK_FLAGS_CATEGORY_IO, enable);
 #else
     (void)enable;
 #endif
+
+#if ERTS_POLL_USE_FALLBACK
+    erts_lcnt_enable_pollset_lock_count_flbk(get_fallback(), enable);
+#endif
+
+    for (i = 0; i < erts_no_pollsets; i++)
+        erts_lcnt_enable_pollset_lock_count(pollsetv[i], enable);
 }
 #endif /* ERTS_ENABLE_LOCK_COUNT */
diff --git a/erts/emulator/sys/common/erl_check_io.h b/erts/emulator/sys/common/erl_check_io.h
index 2d3bb98afa..443ef1264c 100644
--- a/erts/emulator/sys/common/erl_check_io.h
+++ b/erts/emulator/sys/common/erl_check_io.h
@@ -18,10 +18,11 @@
  * %CopyrightEnd%
  */
 
-/*
- * Description:	Check I/O
+/**
+ * @description Check I/O, a cross platform IO polling framework for ERTS
  *
- * Author: 	Rickard Green
+ * @author Rickard Green
+ * @author Lukas Larsson
  */
 
 #ifndef ERL_CHECK_IO_H__
@@ -30,79 +31,79 @@
 #include "sys.h"
 #include "erl_sys_driver.h"
 
-#ifdef ERTS_ENABLE_KERNEL_POLL
-
-int driver_select_kp(ErlDrvPort, ErlDrvEvent, int, int);
-int driver_select_nkp(ErlDrvPort, ErlDrvEvent, int, int);
-int enif_select_kp(ErlNifEnv*, ErlNifEvent, enum ErlNifSelectFlags, void*, const ErlNifPid*, Eterm);
-int enif_select_nkp(ErlNifEnv*, ErlNifEvent, enum ErlNifSelectFlags, void*, const ErlNifPid*, Eterm);
-int driver_event_kp(ErlDrvPort, ErlDrvEvent, ErlDrvEventData);
-int driver_event_nkp(ErlDrvPort, ErlDrvEvent, ErlDrvEventData);
-Uint erts_check_io_size_kp(void);
-Uint erts_check_io_size_nkp(void);
-Eterm erts_check_io_info_kp(void *);
-Eterm erts_check_io_info_nkp(void *);
-int erts_check_io_max_files_kp(void);
-int erts_check_io_max_files_nkp(void);
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-void erts_check_io_async_sig_interrupt_kp(void);
-void erts_check_io_async_sig_interrupt_nkp(void);
-#endif
-void erts_check_io_interrupt_kp(int);
-void erts_check_io_interrupt_nkp(int);
-void erts_check_io_interrupt_timed_kp(int, ErtsMonotonicTime);
-void erts_check_io_interrupt_timed_nkp(int, ErtsMonotonicTime);
-void erts_check_io_kp(int);
-void erts_check_io_nkp(int);
-void erts_init_check_io_kp(void);
-void erts_init_check_io_nkp(void);
-int erts_check_io_debug_kp(ErtsCheckIoDebugInfo *);
-int erts_check_io_debug_nkp(ErtsCheckIoDebugInfo *);
-
-#ifdef ERTS_ENABLE_LOCK_COUNT
-void erts_lcnt_update_cio_locks_kp(int enable);
-void erts_lcnt_update_cio_locks_nkp(int enable);
-#endif
-
-#else /* !ERTS_ENABLE_KERNEL_POLL */
+/** @brief a structure that is used by each polling thread */
+struct erts_poll_thread;
 
+/**
+ * Get the memory size of the check io framework
+ */
 Uint erts_check_io_size(void);
-Eterm erts_check_io_info(void *);
+/**
+ * Returns an Eterm with information about all the pollsets active at the
+ * moment.
+ *
+ * @param proc the Process* to allocate the result on. It is passed as
+ *  void * because of header include problems.
+ */
+Eterm erts_check_io_info(void *proc);
+/**
+ * Should be called when a port IO task has been executed in order to re-enable
+ * or clear the information about the fd.
+ *
+ * @param type The type of event that has been completed.
+ * @param handle The port task handle of the event.
+ * @param reset A function pointer to be called when the port task handle
+ *  should be reset.
+ */
+void erts_io_notify_port_task_executed(ErtsPortTaskType type,
+                                       ErtsPortTaskHandle *handle,
+                                       void (*reset)(ErtsPortTaskHandle *));
+/**
+ * Returns the maximum number of fds that the check io framework can handle.
+ */
 int erts_check_io_max_files(void);
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-void erts_check_io_async_sig_interrupt(void);
-#endif
-void erts_check_io_interrupt(int);
-void erts_check_io_interrupt_timed(int, ErtsMonotonicTime);
-void erts_check_io(int);
-void erts_init_check_io(void);
-
+/**
+ * Called by any thread that should check for new IO events. This function will
+ * not return unless erts_check_io_interrupt(pt, 1) is called by another thread.
+ *
+ * @param pt the poll thread structure to use.
+ */
+void erts_check_io(struct erts_poll_thread *pt);
+/**
+ * Initialize the check io framework. This function will parse the arguments
+ * and delete any entries that it is interested in.
+ *
+ * @param argc the number of arguments
+ * @param argv an array with the arguments
+ */
+void erts_init_check_io(int *argc, char **argv);
+/**
+ * Interrupt the poll thread so that it can execute other code.
+ *
+ * Should be called with set = 0 by the waiting thread before calling
+ * erts_check_io.
+ *
+ * @param pt the poll thread to wake
+ * @param set whether to set or clear the interrupt flag
+ */
+void erts_check_io_interrupt(struct erts_poll_thread *pt, int set);
+/**
+ * Create a new poll thread structure that is associated with the number no.
+ * It is the callers responsibility that no is unique.
+ */
+struct erts_poll_thread* erts_create_pollset_thread(int no);
 #ifdef ERTS_ENABLE_LOCK_COUNT
+/**
+ * Toggle lock counting on all check io locks
+ */
 void erts_lcnt_update_cio_locks(int enable);
 #endif
 
-#endif
-
-extern erts_smp_atomic_t erts_check_io_time;
-
 typedef struct {
     ErtsPortTaskHandle task;
-    erts_smp_atomic_t executed_time;
+    ErtsSysFdType fd;
 } ErtsIoTask;
 
-ERTS_GLB_INLINE void erts_io_notify_port_task_executed(ErtsPortTaskHandle *pthp);
-
-#if ERTS_GLB_INLINE_INCL_FUNC_DEF
-
-ERTS_GLB_INLINE void
-erts_io_notify_port_task_executed(ErtsPortTaskHandle *pthp)
-{
-    ErtsIoTask *itp = (ErtsIoTask *) (((char *) pthp) - offsetof(ErtsIoTask, task));
-    erts_aint_t ci_time = erts_smp_atomic_read_acqb(&erts_check_io_time);
-    erts_smp_atomic_set_relb(&itp->executed_time, ci_time);
-}
-
-#endif
 
 #endif /*  ERL_CHECK_IO_H__ */
 
@@ -110,6 +111,16 @@ erts_io_notify_port_task_executed(ErtsPortTaskHandle *pthp)
 #define ERL_CHECK_IO_INTERNAL__
 #endif
 
+#define ERTS_CHECK_IO_DRV_EV_STATE_LOCK_CNT 128
+
+/* Controls how many pollsets to allocate. Fd's are hashed into
+   each pollset based on the FD. When doing non-concurrent updates
+   there will be one pollset per thread.
+*/
+extern int erts_no_pollsets;
+extern int erts_no_poll_threads;
+
+
 #ifndef ERL_CHECK_IO_INTERNAL__
 #define ERL_CHECK_IO_INTERNAL__
 #include "erl_poll.h"
@@ -122,24 +133,8 @@ erts_io_notify_port_task_executed(ErtsPortTaskHandle *pthp)
  */
 #  define ERTS_CIO_DEFER_ACTIVE_EVENTS 1
 #else
-#  define ERTS_CIO_DEFER_ACTIVE_EVENTS 0
-#endif
-
-/*
- * ErtsDrvEventDataState is used by driver_event() which is almost never
- * used. We allocate ErtsDrvEventDataState separate since we dont wan't
- * the size of ErtsDrvEventState to increase due to driver_event()
- * information.
- */
-typedef struct {
-    Eterm port;
-    ErlDrvEventData data;
-    ErtsPollEvents removed_events;
-#if ERTS_CIO_DEFER_ACTIVE_EVENTS
-    ErtsPollEvents deferred_events;
+#  define ERTS_CIO_DEFER_ACTIVE_EVENTS 1
 #endif
-    ErtsIoTask iotask;
-} ErtsDrvEventDataState;
 
 typedef struct {
     Eterm inport;
diff --git a/erts/emulator/sys/common/erl_mmap.c b/erts/emulator/sys/common/erl_mmap.c
index 214ed01c82..145503bea7 100644
--- a/erts/emulator/sys/common/erl_mmap.c
+++ b/erts/emulator/sys/common/erl_mmap.c
@@ -24,7 +24,6 @@
 #define ERTS_WANT_MEM_MAPPERS
 #include "sys.h"
 #include "erl_process.h"
-#include "erl_smp.h"
 #include "atom.h"
 #include "erl_mmap.h"
 #include <stddef.h>
@@ -62,11 +61,11 @@
     (((UWord) (PTR)) - ((UWord) mm->sa.bot)			\
      < ((UWord) mm->sua.top) - ((UWord) mm->sa.bot))
 #define ERTS_MMAP_IN_SUPERALIGNED_AREA(PTR)				\
-    (ERTS_SMP_LC_ASSERT(erts_lc_mtx_is_locked(&mm->mtx)),	\
+    (ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&mm->mtx)),	\
      (((UWord) (PTR)) - ((UWord) mm->sa.bot)			\
       < ((UWord) mm->sa.top) - ((UWord) mm->sa.bot)))
 #define ERTS_MMAP_IN_SUPERUNALIGNED_AREA(PTR)				\
-    (ERTS_SMP_LC_ASSERT(erts_lc_mtx_is_locked(&mm->mtx)),	\
+    (ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&mm->mtx)),	\
      (((UWord) (PTR)) - ((UWord) mm->sua.bot)			\
       < ((UWord) mm->sua.top) - ((UWord) mm->sua.bot)))
 
@@ -199,10 +198,10 @@ static ErtsMMapOp mmap_ops[ERTS_MMAP_OP_RINGBUF_SZ];
 
 #define ERTS_MMAP_OP_LCK(RES, IN_SZ, OUT_SZ)			\
     do {							\
-	erts_smp_mtx_lock(&mm->mtx);			\
+	erts_mtx_lock(&mm->mtx);			\
 	ERTS_MMAP_OP_START((IN_SZ));				\
 	ERTS_MMAP_OP_END((RES), (OUT_SZ));			\
-	erts_smp_mtx_unlock(&mm->mtx);			\
+	erts_mtx_unlock(&mm->mtx);			\
     } while (0)
 
 #define ERTS_MUNMAP_OP(PTR, SZ)					\
@@ -221,9 +220,9 @@ static ErtsMMapOp mmap_ops[ERTS_MMAP_OP_RINGBUF_SZ];
 
 #define ERTS_MUNMAP_OP_LCK(PTR, SZ)				\
     do {							\
-	erts_smp_mtx_lock(&mm->mtx);			\
+	erts_mtx_lock(&mm->mtx);			\
 	ERTS_MUNMAP_OP((PTR), (SZ));				\
-	erts_smp_mtx_unlock(&mm->mtx);			\
+	erts_mtx_unlock(&mm->mtx);			\
     } while (0)
 
 #define ERTS_MREMAP_OP_START(OLD_PTR, OLD_SZ, IN_SZ)		\
@@ -249,10 +248,10 @@ static ErtsMMapOp mmap_ops[ERTS_MMAP_OP_RINGBUF_SZ];
 
 #define ERTS_MREMAP_OP_LCK(RES, OLD_PTR, OLD_SZ, IN_SZ, OUT_SZ)	\
     do {							\
-	erts_smp_mtx_lock(&mm->mtx);			\
+	erts_mtx_lock(&mm->mtx);			\
 	ERTS_MREMAP_OP_START((OLD_PTR), (OLD_SZ), (IN_SZ));	\
 	ERTS_MREMAP_OP_END((RES), (OUT_SZ));			\
-	erts_smp_mtx_unlock(&mm->mtx);			\
+	erts_mtx_unlock(&mm->mtx);			\
     } while (0)
 
 #define ERTS_MMAP_OP_ABORT()					\
@@ -297,11 +296,10 @@ typedef struct {
 }ErtsFreeSegMap;
 
 struct ErtsMemMapper_ {
-    int (*reserve_physical)(char *, UWord, int exec);
+    int (*reserve_physical)(char *, UWord);
     void (*unreserve_physical)(char *, UWord);
     int supercarrier;
     int no_os_mmap;
-    int executable;   /* is client a native code allocator? */
     /*
      * Super unaligned area is located above super aligned
      * area. That is, `sa.bot` is beginning of the super
@@ -321,7 +319,7 @@ struct ErtsMemMapper_ {
 #if HAVE_MMAP && (!defined(MAP_ANON) && !defined(MAP_ANONYMOUS))
     int mmap_fd;
 #endif
-    erts_smp_mtx_t mtx;
+    erts_mtx_t mtx;
     struct {
 	char *free_list;
 	char *unused_start;
@@ -359,10 +357,6 @@ char* erts_literals_start;
 UWord erts_literals_size;
 #endif
 
-#ifdef ERTS_HAVE_EXEC_MMAPPER
-ErtsMemMapper erts_exec_mmapper;
-#endif
-
 
 #define ERTS_MMAP_SIZE_SC_SA_INC(SZ) 						\
     do {									\
@@ -1241,7 +1235,6 @@ Eterm build_free_seg_list(Process* p, ErtsFreeSegMap* map)
 
 #if HAVE_MMAP
 #  define ERTS_MMAP_PROT		(PROT_READ|PROT_WRITE)
-#  define ERTS_MMAP_PROT_EXEC		(PROT_READ|PROT_WRITE|PROT_EXEC)
 #  if defined(MAP_ANONYMOUS)
 #    define ERTS_MMAP_FLAGS		(MAP_ANON|MAP_PRIVATE)
 #    define ERTS_MMAP_FD		(-1)
@@ -1255,26 +1248,24 @@ Eterm build_free_seg_list(Process* p, ErtsFreeSegMap* map)
 #endif
 
 static ERTS_INLINE void *
-os_mmap(void *hint_ptr, UWord size, int try_superalign, int executable)
+os_mmap(void *hint_ptr, UWord size, int try_superalign)
 {
 #if HAVE_MMAP
-    const int prot = executable ? ERTS_MMAP_PROT_EXEC : ERTS_MMAP_PROT;
     void *res;
 #ifdef MAP_ALIGN
     if (try_superalign)
-	res = mmap((void *) ERTS_SUPERALIGNED_SIZE, size, prot,
+	res = mmap((void *) ERTS_SUPERALIGNED_SIZE, size, ERTS_MMAP_PROT,
 		   ERTS_MMAP_FLAGS|MAP_ALIGN, ERTS_MMAP_FD, 0);
     else
 #endif
-	res = mmap((void *) hint_ptr, size, prot,
+	res = mmap((void *) hint_ptr, size, ERTS_MMAP_PROT,
 		   ERTS_MMAP_FLAGS, ERTS_MMAP_FD, 0);
     if (res == MAP_FAILED)
 	return NULL;
     return res;
 #elif HAVE_VIRTUALALLOC
-    const DWORD prot = executable ? PAGE_EXECUTE_READWRITE : PAGE_READWRITE;
     return (void *) VirtualAlloc(NULL, (SIZE_T) size,
-				 MEM_COMMIT|MEM_RESERVE, prot);
+				 MEM_COMMIT|MEM_RESERVE, PAGE_READWRITE);
 #else
 #  error "missing mmap() or similar"
 #endif
@@ -1331,7 +1322,6 @@ os_mremap(void *ptr, UWord old_size, UWord new_size, int try_superalign)
 #if HAVE_MMAP
 
 #define ERTS_MMAP_RESERVE_PROT		(ERTS_MMAP_PROT)
-#define ERTS_MMAP_RESERVE_PROT_EXEC	(ERTS_MMAP_PROT_EXEC)
 #define ERTS_MMAP_RESERVE_FLAGS		(ERTS_MMAP_FLAGS|MAP_FIXED)
 #define ERTS_MMAP_UNRESERVE_PROT	(PROT_NONE)
 #if defined(__FreeBSD__)
@@ -1347,10 +1337,9 @@ os_mremap(void *ptr, UWord old_size, UWord new_size, int try_superalign)
 #endif /* __FreeBSD__ */
 
 static int
-os_reserve_physical(char *ptr, UWord size, int exec)
+os_reserve_physical(char *ptr, UWord size)
 {
-    const int prot = exec ? ERTS_MMAP_RESERVE_PROT_EXEC : ERTS_MMAP_RESERVE_PROT;
-    void *res = mmap((void *) ptr, (size_t) size, prot,
+    void *res = mmap((void *) ptr, (size_t) size, ERTS_MMAP_RESERVE_PROT,
 		     ERTS_MMAP_RESERVE_FLAGS, ERTS_MMAP_FD, 0);
     if (res == (void *) MAP_FAILED)
 	return 0;
@@ -1367,37 +1356,11 @@ os_unreserve_physical(char *ptr, UWord size)
 }
 
 static void *
-os_mmap_virtual(char *ptr, UWord size, int exec)
+os_mmap_virtual(char *ptr, UWord size)
 {
     int flags = ERTS_MMAP_VIRTUAL_FLAGS;
     void* res;
 
-#ifdef ERTS_ALC_A_EXEC
-    if (exec) {
-        ASSERT(!ptr);
-        /* OTP-19.0: Nice hack below cut-and-pasted from hipe_amd64.c */
-
-#  ifdef MAP_32BIT
-        /* If we got MAP_32BIT (Linux), then use that to ask for low memory */
-        flags |= MAP_32BIT;
-#  else
-        /* FreeBSD doesn't have MAP_32BIT, and it doesn't respect
-           a plain map_hint (returns high mappings even though the
-           hint refers to a free area), so we have to use both map_hint
-           and MAP_FIXED to get addresses below the 2GB boundary.
-           This is even worse than the Linux/ppc64 case.
-           Similarly, Solaris 10 doesn't have MAP_32BIT,
-           and it doesn't respect a plain map_hint. */
-        ptr = (char*)(512*1024*1024); /* 0.5GB */
-
-#    if defined(__FreeBSD__) || defined(__sun__)
-        flags |= MAP_FIXED;
-#    endif
-#  endif /* !MAP_32BIT */
-    }
-#else /* !ERTS_ALC_A_EXEC */
-    ASSERT(!exec);
-#endif
     res = mmap((void *) ptr, (size_t) size, ERTS_MMAP_VIRTUAL_PROT,
                flags, ERTS_MMAP_FD, 0);
     if (res == (void *) MAP_FAILED)
@@ -1412,7 +1375,7 @@ os_mmap_virtual(char *ptr, UWord size, int exec)
 
 #endif /* ERTS_HAVE_OS_MMAP */
 
-static int reserve_noop(char *ptr, UWord size, int exec)
+static int reserve_noop(char *ptr, UWord size)
 {
 #ifdef ERTS_MMAP_DEBUG_FILL_AREAS
     Uint32 *uip, *end = (Uint32 *) (ptr + size);
@@ -1455,7 +1418,7 @@ alloc_desc_insert_free_seg(ErtsMemMapper* mm,
 
 #if ERTS_HAVE_OS_MMAP
     if (!mm->no_os_mmap) {
-        ptr = os_mmap(mm->desc.new_area_hint, ERTS_PAGEALIGNED_SIZE, 0, 0);
+        ptr = os_mmap(mm->desc.new_area_hint, ERTS_PAGEALIGNED_SIZE, 0);
         if (ptr) {
             mm->desc.new_area_hint = ptr+ERTS_PAGEALIGNED_SIZE;
             ERTS_MMAP_SIZE_OS_INC(ERTS_PAGEALIGNED_SIZE);
@@ -1474,7 +1437,7 @@ alloc_desc_insert_free_seg(ErtsMemMapper* mm,
     da_map = &mm->sua.map;
     desc = lookup_free_seg(da_map, ERTS_PAGEALIGNED_SIZE);
     if (desc) {
-	if (mm->reserve_physical(desc->start, ERTS_PAGEALIGNED_SIZE, 0))
+	if (mm->reserve_physical(desc->start, ERTS_PAGEALIGNED_SIZE))
 	    ERTS_MMAP_SIZE_SC_SUA_INC(ERTS_PAGEALIGNED_SIZE);
 	else
 	    desc = NULL;
@@ -1484,7 +1447,7 @@ alloc_desc_insert_free_seg(ErtsMemMapper* mm,
 	da_map = &mm->sa.map;
 	desc = lookup_free_seg(da_map, ERTS_PAGEALIGNED_SIZE);
 	if (desc) {
-	    if (mm->reserve_physical(desc->start, ERTS_PAGEALIGNED_SIZE, 0))
+	    if (mm->reserve_physical(desc->start, ERTS_PAGEALIGNED_SIZE))
 		ERTS_MMAP_SIZE_SC_SA_INC(ERTS_PAGEALIGNED_SIZE);
 	    else
 		desc = NULL;
@@ -1536,7 +1499,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 	ErtsFreeSegDesc *desc;
 	Uint32 superaligned = (ERTS_MMAPFLG_SUPERALIGNED & flags);
 
-	erts_smp_mtx_lock(&mm->mtx);
+	erts_mtx_lock(&mm->mtx);
 
 	ERTS_MMAP_OP_START(*sizep);
 
@@ -1545,7 +1508,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 	    if (desc) {
 		seg = desc->start;
 		end = seg+asize;
-		if (!mm->reserve_physical(seg, asize, mm->executable))
+		if (!mm->reserve_physical(seg, asize))
 		    goto supercarrier_reserve_failure;
 		if (desc->end == end) {
 		    delete_free_seg(&mm->sua.map, desc);
@@ -1560,8 +1523,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 	    }
 
 	    if (asize <= mm->sua.bot - mm->sa.top) {
-		if (!mm->reserve_physical(mm->sua.bot - asize, asize,
-                                          mm->executable))
+		if (!mm->reserve_physical(mm->sua.bot - asize, asize))
 		    goto supercarrier_reserve_failure;
 		mm->sua.bot -= asize;
 		seg = mm->sua.bot;
@@ -1577,8 +1539,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 	    char *start = seg = desc->start;
 	    seg = (char *) ERTS_SUPERALIGNED_CEILING(seg);
 	    end = seg+asize;
-	    if (!mm->reserve_physical(start, (UWord) (end - start),
-                                      mm->executable))
+	    if (!mm->reserve_physical(start, (UWord) (end - start)))
 		goto supercarrier_reserve_failure;
 	    ERTS_MMAP_SIZE_SC_SA_INC(asize);
 	    if (desc->end == end) {
@@ -1610,8 +1571,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 
 	    if (asize + (seg - start) <= mm->sua.bot - start) {
 		end = seg + asize;
-		if (!mm->reserve_physical(start, (UWord) (end - start),
-                                          mm->executable))
+		if (!mm->reserve_physical(start, (UWord) (end - start)))
 		    goto supercarrier_reserve_failure;
 		mm->sa.top = end;
 		ERTS_MMAP_SIZE_SC_SA_INC(asize);
@@ -1633,8 +1593,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 
 		seg = (char *) ERTS_SUPERALIGNED_CEILING(org_start);
 		end = seg + asize;
-		if (!mm->reserve_physical(seg, (UWord) (org_end - seg),
-                                          mm->executable))
+		if (!mm->reserve_physical(seg, (UWord) (org_end - seg)))
 		    goto supercarrier_reserve_failure;
 		ERTS_MMAP_SIZE_SC_SUA_INC(asize);
 		if (org_start != seg) {
@@ -1660,20 +1619,20 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 	}
 
 	ERTS_MMAP_OP_ABORT();
-	erts_smp_mtx_unlock(&mm->mtx);
+	erts_mtx_unlock(&mm->mtx);
     }
 
 #if ERTS_HAVE_OS_MMAP
     /* Map using OS primitives */
     if (!(ERTS_MMAPFLG_SUPERCARRIER_ONLY & flags) && !mm->no_os_mmap) {
 	if (!(ERTS_MMAPFLG_SUPERALIGNED & flags)) {
-	    seg = os_mmap(NULL, asize, 0, mm->executable);
+	    seg = os_mmap(NULL, asize, 0);
 	    if (!seg)
 		goto failure;
 	}
 	else {
 	    asize = ERTS_SUPERALIGNED_CEILING(*sizep);
-	    seg = os_mmap(NULL, asize, 1, mm->executable);
+	    seg = os_mmap(NULL, asize, 1);
 	    if (!seg)
 		goto failure;
 
@@ -1683,8 +1642,7 @@ erts_mmap(ErtsMemMapper* mm, Uint32 flags, UWord *sizep)
 
 		os_munmap(seg, asize);
 
-		ptr = os_mmap(NULL, asize + ERTS_SUPERALIGNED_SIZE, 1,
-                              mm->executable);
+		ptr = os_mmap(NULL, asize + ERTS_SUPERALIGNED_SIZE, 1);
 		if (!ptr)
 		    goto failure;
 
@@ -1724,13 +1682,13 @@ supercarrier_success:
 #endif
 
     ERTS_MMAP_OP_END(seg, asize);
-    erts_smp_mtx_unlock(&mm->mtx);
+    erts_mtx_unlock(&mm->mtx);
 
     *sizep = asize;
     return (void *) seg;
 
 supercarrier_reserve_failure:
-    erts_smp_mtx_unlock(&mm->mtx);
+    erts_mtx_unlock(&mm->mtx);
     *sizep = 0;
     return NULL;
 }
@@ -1760,7 +1718,7 @@ erts_munmap(ErtsMemMapper* mm, Uint32 flags, void *ptr, UWord size)
 	start = (char *) ptr;
 	end = start + size;
 
-	erts_smp_mtx_lock(&mm->mtx);
+	erts_mtx_lock(&mm->mtx);
 
 	ERTS_MUNMAP_OP(ptr, size);
 
@@ -1829,7 +1787,7 @@ erts_munmap(ErtsMemMapper* mm, Uint32 flags, void *ptr, UWord size)
 	    if (unres_sz)
 		mm->unreserve_physical(((char *) ptr) + ad_sz, unres_sz);
 
-            erts_smp_mtx_unlock(&mm->mtx);
+            erts_mtx_unlock(&mm->mtx);
 	}
     }
 }
@@ -1948,12 +1906,12 @@ erts_mremap(ErtsMemMapper* mm,
 		 ? ERTS_SUPERALIGNED_CEILING(*sizep)
 		 : ERTS_PAGEALIGNED_CEILING(*sizep));
 
-	erts_smp_mtx_lock(&mm->mtx);
+	erts_mtx_lock(&mm->mtx);
 
 	if (ERTS_MMAP_IN_SUPERALIGNED_AREA(ptr)
 	    ? (!superaligned && lookup_free_seg(&mm->sua.map, asize))
 	    : (superaligned && lookup_free_seg(&mm->sa.map, asize))) {
-	    erts_smp_mtx_unlock(&mm->mtx);
+	    erts_mtx_unlock(&mm->mtx);
 	    /*
 	     * Segment currently in wrong area (due to a previous memory
 	     * shortage), move it to the right area.
@@ -2019,8 +1977,7 @@ erts_mremap(ErtsMemMapper* mm,
 
 	    if (next && new_end <= next->end) {
 		if (!mm->reserve_physical(((char *) ptr) + old_size,
-                                          asize - old_size,
-                                          mm->executable))
+                                          asize - old_size))
 		    goto supercarrier_reserve_failure;
 		if (new_end < next->end)
 		    resize_free_seg(&mm->sua.map, next, new_end, next->end);
@@ -2038,8 +1995,7 @@ erts_mremap(ErtsMemMapper* mm,
 	    if (end == mm->sa.top) {
 		if (new_end <= mm->sua.bot) {
 		    if (!mm->reserve_physical(((char *) ptr) + old_size,
-                                              asize - old_size,
-                                              mm->executable))
+                                              asize - old_size))
 			goto supercarrier_reserve_failure;
 		    mm->sa.top = new_end;
 		    new_ptr = ptr;
@@ -2051,8 +2007,7 @@ erts_mremap(ErtsMemMapper* mm,
 		adjacent_free_seg(&mm->sa.map, start, end, &prev, &next);
 		if (next && new_end <= next->end) {
 		    if (!mm->reserve_physical(((char *) ptr) + old_size,
-                                              asize - old_size,
-                                              mm->executable))
+                                              asize - old_size))
 			goto supercarrier_reserve_failure;
 		    if (new_end < next->end)
 			resize_free_seg(&mm->sa.map, next, new_end, next->end);
@@ -2068,7 +2023,7 @@ erts_mremap(ErtsMemMapper* mm,
 	}
 
 	ERTS_MMAP_OP_ABORT();
-	erts_smp_mtx_unlock(&mm->mtx);
+	erts_mtx_unlock(&mm->mtx);
 
 	/* Failed to resize... */
     }
@@ -2090,14 +2045,14 @@ supercarrier_resize_success:
 #endif
 
     ERTS_MREMAP_OP_END(new_ptr, asize);
-    erts_smp_mtx_unlock(&mm->mtx);
+    erts_mtx_unlock(&mm->mtx);
 
     *sizep = asize;
     return new_ptr;
 
 supercarrier_reserve_failure:
     ERTS_MREMAP_OP_END(NULL, old_size);
-    erts_smp_mtx_unlock(&mm->mtx);
+    erts_mtx_unlock(&mm->mtx);
     *sizep = old_size;
     return NULL;
     
@@ -2172,7 +2127,7 @@ static void hard_dbg_mseg_init(void);
 #endif
 
 void
-erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
+erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init)
 {
     static int is_first_call = 1;
     int virtual_map = 0;
@@ -2204,7 +2159,6 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
     mm->supercarrier = 0;
     mm->reserve_physical = reserve_noop;
     mm->unreserve_physical = unreserve_noop;
-    mm->executable = executable;
 
 #if HAVE_MMAP && !defined(MAP_ANON)
     mm->mmap_fd = open("/dev/zero", O_RDWR);
@@ -2212,7 +2166,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	erts_exit(1, "erts_mmap: Failed to open /dev/zero\n");
 #endif
 
-    erts_smp_mtx_init(&mm->mtx, "erts_mmap", NIL,
+    erts_mtx_init(&mm->mtx, "erts_mmap", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
     if (is_first_call) {
         erts_mtx_init(&am.init_mutex, "mmap_init_atoms", NIL,
@@ -2226,7 +2180,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	ptr = (char *) ERTS_PAGEALIGNED_CEILING(init->virtual_range.start);
 	end = (char *) ERTS_PAGEALIGNED_FLOOR(init->virtual_range.end);
 	sz = end - ptr;
-	start = os_mmap_virtual(ptr, sz, executable);
+	start = os_mmap_virtual(ptr, sz);
 	if (!start || start > ptr || start >= end)
 	    erts_exit(1,
 		     "erts_mmap: Failed to create virtual range for super carrier\n");
@@ -2251,7 +2205,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	sz = ERTS_PAGEALIGNED_CEILING(init->scs);
 #ifdef ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION
 	if (!init->scrpm) {
-	    start = os_mmap_virtual(NULL, sz, executable);
+	    start = os_mmap_virtual(NULL, sz);
 	    mm->reserve_physical = os_reserve_physical;
 	    mm->unreserve_physical = os_unreserve_physical;
 	    virtual_map = 1;
@@ -2263,7 +2217,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	     * The whole supercarrier will by physically
 	     * reserved all the time.
 	     */
-	    start = os_mmap(NULL, sz, 1, executable);
+	    start = os_mmap(NULL, sz, 1);
 	}
 	if (!start)
 	    erts_exit(1,
@@ -2337,7 +2291,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	    mm->sua.top -= ERTS_PAGEALIGNED_SIZE;
 	    mm->size.supercarrier.used.total += ERTS_PAGEALIGNED_SIZE;
 #ifdef ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION
-	    if (!virtual_map || os_reserve_physical(mm->sua.top, ERTS_PAGEALIGNED_SIZE, 0))
+	    if (!virtual_map || os_reserve_physical(mm->sua.top, ERTS_PAGEALIGNED_SIZE))
 #endif
 		add_free_desc_area(mm, mm->sua.top, end);
             mm->desc.reserved += (end - mm->sua.top) / sizeof(ErtsFreeSegDesc);
@@ -2350,7 +2304,7 @@ erts_mmap_init(ErtsMemMapper* mm, ErtsMMapInit *init, int executable)
 	 * will be used for free segment descritors.
 	 */
 #ifdef ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION
-	if (virtual_map && !os_reserve_physical(start, mm->sa.bot - start, 0))
+	if (virtual_map && !os_reserve_physical(start, mm->sa.bot - start))
 	    erts_exit(1, "erts_mmap: Failed to reserve physical memory for descriptors\n");
 #endif
 	mm->desc.unused_start = start;
@@ -2407,7 +2361,7 @@ Eterm erts_mmap_info(ErtsMemMapper* mm,
     Eterm res = THE_NON_VALUE;
 
     if (!hpp) {
-        erts_smp_mtx_lock(&mm->mtx);
+        erts_mtx_lock(&mm->mtx);
         emis->sizes[0] = mm->size.supercarrier.total;
         emis->sizes[1] = mm->sa.top  - mm->sa.bot;
         emis->sizes[2] = mm->sua.top - mm->sua.bot;
@@ -2423,7 +2377,7 @@ Eterm erts_mmap_info(ErtsMemMapper* mm,
         emis->segs[5] = mm->sua.map.nseg;
 
         emis->os_used = mm->size.os.used;
-        erts_smp_mtx_unlock(&mm->mtx);
+        erts_mtx_unlock(&mm->mtx);
     }
 
     list[lix] = erts_mmap_info_options(mm, "option ", print_to_p, print_to_arg,
@@ -2543,14 +2497,14 @@ Eterm erts_mmap_debug_info(Process* p)
         Eterm *hp, *hp_end;
         Uint may_need;
 
-        erts_smp_mtx_lock(&mm->mtx);
+        erts_mtx_lock(&mm->mtx);
         values[0] = (UWord)mm->sa.bot;
         values[1] = (UWord)mm->sa.top;
         values[2] = (UWord)mm->sua.bot;
         values[3] = (UWord)mm->sua.top;
         sa_list = build_free_seg_list(p, &mm->sa.map);
         sua_list = build_free_seg_list(p, &mm->sua.map);
-        erts_smp_mtx_unlock(&mm->mtx);
+        erts_mtx_unlock(&mm->mtx);
 
         may_need = 4*(2+3+2) + 2*(2+3);
         hp = HAlloc(p, may_need);
@@ -2560,8 +2514,8 @@ Eterm erts_mmap_debug_info(Process* p)
                                             sizeof(values)/sizeof(*values),
                                             tags, values);
 
-        sa_list = TUPLE2(hp, am_atom_put("sa_free_segs",12), sa_list); hp+=3;
-        sua_list = TUPLE2(hp, am_atom_put("sua_free_segs",13), sua_list); hp+=3;
+        sa_list = TUPLE2(hp, ERTS_MAKE_AM("sa_free_segs"), sa_list); hp+=3;
+        sua_list = TUPLE2(hp, ERTS_MAKE_AM("sua_free_segs"), sua_list); hp+=3;
         list = CONS(hp, sua_list, list); hp+=2;
         list = CONS(hp, sa_list, list); hp+=2;
 
diff --git a/erts/emulator/sys/common/erl_mmap.h b/erts/emulator/sys/common/erl_mmap.h
index 2a07d93c8c..c1f9668b18 100644
--- a/erts/emulator/sys/common/erl_mmap.h
+++ b/erts/emulator/sys/common/erl_mmap.h
@@ -93,11 +93,6 @@ typedef struct {
 #define ERTS_MMAP_INIT_LITERAL_INITER \
     {{NULL, NULL}, {NULL, NULL}, ERTS_LITERAL_VIRTUAL_AREA_SIZE, 1, (1 << 10), 0}
 
-#define ERTS_HIPE_EXEC_VIRTUAL_AREA_SIZE (UWORD_CONSTANT(512)*1024*1024)
-
-#define ERTS_MMAP_INIT_HIPE_EXEC_INITER \
-    {{NULL, NULL}, {NULL, NULL}, ERTS_HIPE_EXEC_VIRTUAL_AREA_SIZE, 1, (1 << 10), 0}
-
 
 #define ERTS_SUPERALIGNED_SIZE \
     (1 << ERTS_MMAP_SUPERALIGNED_BITS)
@@ -140,7 +135,7 @@ void *erts_mmap(ErtsMemMapper*, Uint32 flags, UWord *sizep);
 void erts_munmap(ErtsMemMapper*, Uint32 flags, void *ptr, UWord size);
 void *erts_mremap(ErtsMemMapper*, Uint32 flags, void *ptr, UWord old_size, UWord *sizep);
 int erts_mmap_in_supercarrier(ErtsMemMapper*, void *ptr);
-void erts_mmap_init(ErtsMemMapper*, ErtsMMapInit*, int executable);
+void erts_mmap_init(ErtsMemMapper*, ErtsMMapInit*);
 struct erts_mmap_info_struct
 {
     UWord sizes[6];
@@ -165,15 +160,6 @@ extern ErtsMemMapper erts_dflt_mmapper;
 extern ErtsMemMapper erts_literal_mmapper;
 #  endif
 
-#  if defined(ERTS_ALC_A_EXEC) && defined(__x86_64__)
-   /*
-    * On x86_64, exec_alloc employs its own super carrier 'erts_exec_mmaper'
-    * to ensure low memory for HiPE AMD64 small code model.
-    */
-#   define ERTS_HAVE_EXEC_MMAPPER
-extern ErtsMemMapper erts_exec_mmapper;
-#  endif
-
 # endif /* ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION */
 #endif /* ERTS_WANT_MEM_MAPPERS */
 
diff --git a/erts/emulator/sys/common/erl_mseg.c b/erts/emulator/sys/common/erl_mseg.c
index d69a79dc2a..ced3d61525 100644
--- a/erts/emulator/sys/common/erl_mseg.c
+++ b/erts/emulator/sys/common/erl_mseg.c
@@ -188,7 +188,6 @@ typedef union {
 static int no_mseg_allocators;
 static ErtsAlgndMsegAllctr_t *aligned_mseg_allctr;
 
-#ifdef ERTS_SMP
 
 #define ERTS_MSEG_ALLCTR_IX(IX) \
   (&aligned_mseg_allctr[(IX)].mseg_alloc)
@@ -199,18 +198,6 @@ static ErtsAlgndMsegAllctr_t *aligned_mseg_allctr;
 #define ERTS_MSEG_ALLCTR_OPT(OPT) \
   ((OPT)->sched_spec ? ERTS_MSEG_ALLCTR_SS() : ERTS_MSEG_ALLCTR_IX(0))
 
-#else
-
-#define ERTS_MSEG_ALLCTR_IX(IX) \
-  (&aligned_mseg_allctr[0].mseg_alloc)
-
-#define ERTS_MSEG_ALLCTR_SS() \
-  (&aligned_mseg_allctr[0].mseg_alloc)
-
-#define ERTS_MSEG_ALLCTR_OPT(OPT) \
-  (&aligned_mseg_allctr[0].mseg_alloc)
-
-#endif
 
 #define ERTS_MSEG_LOCK(MA)		\
 do {					\
@@ -352,11 +339,11 @@ mseg_recreate(ErtsMsegAllctr_t *ma, Uint flags, void *old_seg, UWord old_size, U
 do {									\
     if ((MA)->is_thread_safe)						\
 	ERTS_LC_ASSERT(erts_lc_mtx_is_locked(&(MA)->mtx)		\
-		       || erts_smp_thr_progress_is_blocking()		\
+		       || erts_thr_progress_is_blocking()		\
 		       || ERTS_IS_CRASH_DUMPING);			\
     else								\
 	ERTS_LC_ASSERT((MA)->ix == (int) erts_get_scheduler_id()	\
-		       || erts_smp_thr_progress_is_blocking()		\
+		       || erts_thr_progress_is_blocking()		\
 		       || ERTS_IS_CRASH_DUMPING);			\
 } while (0)
 #else
@@ -1404,11 +1391,7 @@ erts_mseg_init(ErtsMsegInit_t *init)
     int i;
     UWord x;
 
-#ifdef ERTS_SMP
     no_mseg_allocators = init->nos + 1;
-#else
-    no_mseg_allocators = 1;
-#endif
 
     x = (UWord) malloc(sizeof(ErtsAlgndMsegAllctr_t)
 		       *no_mseg_allocators
@@ -1423,15 +1406,9 @@ erts_mseg_init(ErtsMsegInit_t *init)
     erts_mtx_init(&init_atoms_mutex, "mseg_init_atoms", NIL,
         ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 
-#ifdef ERTS_HAVE_EXEC_MMAPPER
-    /* Initialize erts_exec_mapper *FIRST*, to increase probability
-     * of getting low memory for HiPE AMD64's small code model.
-     */
-    erts_mmap_init(&erts_exec_mmapper, &init->exec_mmap, 1);
-#endif
-    erts_mmap_init(&erts_dflt_mmapper, &init->dflt_mmap, 0);
+    erts_mmap_init(&erts_dflt_mmapper, &init->dflt_mmap);
 #if defined(ARCH_64) && defined(ERTS_HAVE_OS_PHYSICAL_MEMORY_RESERVATION)
-    erts_mmap_init(&erts_literal_mmapper, &init->literal_mmap, 0);
+    erts_mmap_init(&erts_literal_mmapper, &init->literal_mmap);
 #endif
 
     if (!IS_2POW(GET_PAGE_SIZE))
diff --git a/erts/emulator/sys/common/erl_mseg.h b/erts/emulator/sys/common/erl_mseg.h
index bba0dec499..af275c18be 100644
--- a/erts/emulator/sys/common/erl_mseg.h
+++ b/erts/emulator/sys/common/erl_mseg.h
@@ -61,7 +61,6 @@ typedef struct {
     Uint nos;
     ErtsMMapInit dflt_mmap;
     ErtsMMapInit literal_mmap;
-    ErtsMMapInit exec_mmap;
 } ErtsMsegInit_t;
 
 #define ERTS_MSEG_INIT_DEFAULT_INITIALIZER				\
@@ -72,7 +71,6 @@ typedef struct {
     1000,		/* cci:   Cache check interval		*/	\
     ERTS_MMAP_INIT_DEFAULT_INITER,					\
     ERTS_MMAP_INIT_LITERAL_INITER,                                      \
-    ERTS_MMAP_INIT_HIPE_EXEC_INITER                                     \
 }
 
 typedef struct {
diff --git a/erts/emulator/sys/common/erl_os_monotonic_time_extender.c b/erts/emulator/sys/common/erl_os_monotonic_time_extender.c
index d53190fdd5..341845cc2a 100644
--- a/erts/emulator/sys/common/erl_os_monotonic_time_extender.c
+++ b/erts/emulator/sys/common/erl_os_monotonic_time_extender.c
@@ -23,7 +23,6 @@
 #endif
 #include "erl_os_monotonic_time_extender.h"
 
-#ifdef USE_THREADS
 
 static void *os_monotonic_time_extender(void *vstatep)
 {
@@ -49,30 +48,22 @@ static void *os_monotonic_time_extender(void *vstatep)
 }
 
 static erts_tid_t os_monotonic_extender_tid;
-#endif
 
 void
 erts_init_os_monotonic_time_extender(ErtsOsMonotonicTimeExtendState *statep,
 				     Uint32 (*raw_os_monotonic_time)(void),
 				     int check_seconds)
 {
-#ifdef USE_THREADS
     statep->raw_os_monotonic_time = raw_os_monotonic_time;
     erts_atomic32_init_nob(&statep->extend[0], (erts_aint32_t) 0);
     erts_atomic32_init_nob(&statep->extend[1], (erts_aint32_t) 0);
     statep->check_interval = check_seconds;
 
-#else
-    statep->extend[0] = (Uint32) 0;
-    statep->extend[1] = (Uint32) 0;
-    statep->last_msb = (ErtsMonotonicTime) 0;
-#endif
 }
 
 void
 erts_late_init_os_monotonic_time_extender(ErtsOsMonotonicTimeExtendState *statep)
 {
-#ifdef USE_THREADS
     erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER;
     thr_opts.detached = 1;
     thr_opts.suggested_stack_size = 4;
@@ -85,5 +76,4 @@ erts_late_init_os_monotonic_time_extender(ErtsOsMonotonicTimeExtendState *statep
 		    os_monotonic_time_extender,
 		    (void*) statep,
 		    &thr_opts);
-#endif
 }
diff --git a/erts/emulator/sys/common/erl_os_monotonic_time_extender.h b/erts/emulator/sys/common/erl_os_monotonic_time_extender.h
index 8089c9aed9..53c32579d5 100644
--- a/erts/emulator/sys/common/erl_os_monotonic_time_extender.h
+++ b/erts/emulator/sys/common/erl_os_monotonic_time_extender.h
@@ -25,36 +25,16 @@
 #include "erl_threads.h"
 
 typedef struct {
-#ifdef USE_THREADS
     Uint32 (*raw_os_monotonic_time)(void);
     erts_atomic32_t extend[2];
     int check_interval;
-#else
-    Uint32 extend[2];
-    ErtsMonotonicTime last_msb;
-#endif
 } ErtsOsMonotonicTimeExtendState;
 
-#ifdef USE_THREADS
-#  define ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(S, RT) ((void) 1)
 #  define ERTS_EXTEND_OS_MONOTONIC_TIME(S, RT)				\
     ((((ErtsMonotonicTime)						\
        erts_atomic32_read_nob(&((S)->extend[((int) ((RT) >> 31)) & 1]))) \
       << 32)								\
      + (RT))
-#else
-#  define ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(S, RT)			\
-    do {								\
-	Uint32 msb__ = (RT) & (((Uint32) 1) << 31);			\
-	if (msb__ != (S)->last_msb) {					\
-	    int ix__ = ((int) ((S)->last_msb >> 31)) & 1;		\
-	    (S)->extend[ix__]++;					\
-	    (S)->last_msb = msb;					\
-	}								\
-    } while (0)
-#  define ERTS_EXTEND_OS_MONOTONIC_TIME(S, RT)				\
-    ((((ErtsMonotonicTime) (S)->extend[((int) ((RT) >> 31)) & 1]) << 32) + (RT))
-#endif
 
 void
 erts_init_os_monotonic_time_extender(ErtsOsMonotonicTimeExtendState *statep,
diff --git a/erts/emulator/sys/common/erl_osenv.c b/erts/emulator/sys/common/erl_osenv.c
new file mode 100644
index 0000000000..9f54d1dff0
--- /dev/null
+++ b/erts/emulator/sys/common/erl_osenv.c
@@ -0,0 +1,396 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#include "erl_osenv.h"
+
+#include "global.h"
+#include "erl_alloc.h"
+#include "erl_process.h"
+
+#define STACKBUF_SIZE (512)
+
+typedef struct __env_rbtnode_t {
+    struct __env_rbtnode_t *parent;
+    struct __env_rbtnode_t *left;
+    struct __env_rbtnode_t *right;
+
+    int is_red;
+
+    erts_osenv_data_t key;
+    erts_osenv_data_t value;
+} env_rbtnode_t;
+
+#define ERTS_RBT_PREFIX env
+#define ERTS_RBT_T env_rbtnode_t
+#define ERTS_RBT_KEY_T erts_osenv_data_t
+#define ERTS_RBT_FLAGS_T int
+#define ERTS_RBT_INIT_EMPTY_TNODE(T) \
+    do { \
+        (T)->parent = NULL; \
+        (T)->left = NULL; \
+        (T)->right = NULL; \
+        (T)->is_red = 0; \
+    } while(0)
+#define ERTS_RBT_IS_RED(T) ((T)->is_red)
+#define ERTS_RBT_SET_RED(T) ((T)->is_red = 1)
+#define ERTS_RBT_IS_BLACK(T) (!ERTS_RBT_IS_RED(T))
+#define ERTS_RBT_SET_BLACK(T) ((T)->is_red = 0)
+#define ERTS_RBT_GET_FLAGS(T) ((T)->is_red)
+#define ERTS_RBT_SET_FLAGS(T, F) ((T)->is_red = F)
+#define ERTS_RBT_GET_PARENT(T) ((T)->parent)
+#define ERTS_RBT_SET_PARENT(T, P) ((T)->parent = P)
+#define ERTS_RBT_GET_RIGHT(T) ((T)->right)
+#define ERTS_RBT_SET_RIGHT(T, R) ((T)->right = (R))
+#define ERTS_RBT_GET_LEFT(T) ((T)->left)
+#define ERTS_RBT_SET_LEFT(T, L) ((T)->left = (L))
+#define ERTS_RBT_GET_KEY(T) ((T)->key)
+#define ERTS_RBT_IS_LT(KX, KY) (compare_env_keys(KX, KY) < 0)
+#define ERTS_RBT_IS_EQ(KX, KY) (compare_env_keys(KX, KY) == 0)
+#define ERTS_RBT_WANT_FOREACH_DESTROY
+#define ERTS_RBT_WANT_FOREACH
+#define ERTS_RBT_WANT_REPLACE
+#define ERTS_RBT_WANT_DELETE
+#define ERTS_RBT_WANT_INSERT
+#define ERTS_RBT_WANT_LOOKUP
+
+static int compare_env_keys(const erts_osenv_data_t a, const erts_osenv_data_t b);
+
+#include "erl_rbtree.h"
+
+static int compare_env_keys(const erts_osenv_data_t a, const erts_osenv_data_t b) {
+    int relation = sys_memcmp(a.data, b.data, MIN(a.length, b.length));
+
+    if(relation != 0) {
+        return relation;
+    }
+
+    if(a.length < b.length) {
+        return -1;
+    } else if(a.length == b.length) {
+        return 0;
+    } else {
+        return 1;
+    }
+}
+
+static void *convert_value_to_native(Eterm term, char *stackbuf,
+        int stackbuf_size, Sint *length) {
+    int encoding;
+    void *result;
+
+    if(is_atom(term)) {
+        return NULL;
+    }
+
+    encoding = erts_get_native_filename_encoding();
+    *length = erts_native_filename_need(term, encoding);
+
+    if(*length < 0) {
+        return NULL;
+    } else if(*length >= stackbuf_size) {
+        result = erts_alloc(ERTS_ALC_T_TMP, *length);
+    } else {
+        result = stackbuf;
+    }
+
+    erts_native_filename_put(term, encoding, (byte*)result);
+
+    return result;
+}
+
+static void *convert_key_to_native(Eterm term, char *stackbuf,
+        int stackbuf_size, Sint *length) {
+    byte *name_iterator, *name_end;
+    void *result;
+    int encoding;
+
+    result = convert_value_to_native(term, stackbuf, stackbuf_size, length);
+
+    if(result == NULL || length == 0) {
+        return NULL;
+    }
+
+    encoding = erts_get_native_filename_encoding();
+
+    name_iterator = (byte*)result;
+    name_end = &name_iterator[*length];
+
+#ifdef __WIN32__
+    /* Windows stores per-drive working directories as variables starting with
+     * '=', so we skip the first character to tolerate that. */
+    name_iterator = erts_raw_env_next_char(name_iterator, encoding);
+#endif
+
+    while(name_iterator < name_end) {
+        if(erts_raw_env_char_is_7bit_ascii_char('=', name_iterator, encoding)) {
+            if(result != stackbuf) {
+                erts_free(ERTS_ALC_T_TMP, result);
+            }
+
+            return NULL;
+        }
+
+        name_iterator = erts_raw_env_next_char(name_iterator, encoding);
+    }
+
+    return result;
+}
+
+void erts_osenv_init(erts_osenv_t *env) {
+    env->variable_count = 0;
+    env->content_size = 0;
+    env->tree = NULL;
+}
+
+static void destroy_foreach(env_rbtnode_t *node, void *_state) {
+    erts_free(ERTS_ALC_T_ENVIRONMENT, node);
+    (void)_state;
+}
+
+void erts_osenv_clear(erts_osenv_t *env) {
+    env_rbt_foreach_destroy(&env->tree, &destroy_foreach, NULL);
+    erts_osenv_init(env);
+}
+
+struct __env_merge {
+    int overwrite_existing;
+    erts_osenv_t *env;
+};
+
+static void merge_foreach(env_rbtnode_t *node, void *_state) {
+    struct __env_merge *state = (struct __env_merge*)(_state);
+    env_rbtnode_t *existing_node;
+
+    existing_node = env_rbt_lookup(state->env->tree, node->key);
+
+    if(existing_node == NULL || state->overwrite_existing) {
+        erts_osenv_put_native(state->env, &node->key, &node->value);
+    }
+}
+
+void erts_osenv_merge(erts_osenv_t *env, const erts_osenv_t *with, int overwrite) {
+    struct __env_merge merge_state;
+
+    merge_state.overwrite_existing = overwrite;
+    merge_state.env = env;
+
+    env_rbt_foreach(with->tree, merge_foreach, &merge_state);
+}
+
+struct __env_foreach_term {
+    erts_osenv_foreach_term_cb_t user_callback;
+    struct process *process;
+    void *user_state;
+};
+
+static void foreach_term_wrapper(env_rbtnode_t *node, void *_state) {
+    struct __env_foreach_term *state = (struct __env_foreach_term*)_state;
+    Eterm key, value;
+
+    key = erts_convert_native_to_filename(state->process,
+        node->key.length, (byte*)node->key.data);
+    value = erts_convert_native_to_filename(state->process,
+        node->value.length, (byte*)node->value.data);
+
+    state->user_callback(state->process, state->user_state, key, value);
+}
+
+void erts_osenv_foreach_term(const erts_osenv_t *env, struct process *process,
+        void *state, erts_osenv_foreach_term_cb_t callback) {
+    struct __env_foreach_term wrapper_state;
+
+    wrapper_state.user_callback = callback;
+    wrapper_state.user_state = state;
+    wrapper_state.process = process;
+
+    env_rbt_foreach(env->tree, foreach_term_wrapper, &wrapper_state);
+}
+
+int erts_osenv_get_term(const erts_osenv_t *env, Process *process,
+        Eterm key_term, Eterm *out_term) {
+    char key_stackbuf[STACKBUF_SIZE];
+    erts_osenv_data_t key;
+    int result;
+
+    key.data = convert_key_to_native(key_term, key_stackbuf,
+        STACKBUF_SIZE, &key.length);
+    result = -1;
+
+    if(key.data != NULL) {
+        env_rbtnode_t *node;
+
+        node = env_rbt_lookup(env->tree, key);
+        result = 0;
+
+        if(node != NULL) {
+            (*out_term) = erts_convert_native_to_filename(process,
+                node->value.length, (byte*)node->value.data);
+            result = 1;
+        }
+
+        if(key.data != key_stackbuf) {
+            erts_free(ERTS_ALC_T_TMP, key.data);
+        }
+    }
+
+    return result;
+}
+
+int erts_osenv_put_term(erts_osenv_t *env, Eterm key_term, Eterm value_term) {
+    char key_stackbuf[STACKBUF_SIZE], value_stackbuf[STACKBUF_SIZE];
+    erts_osenv_data_t key, value;
+    int result;
+
+    key.data = convert_key_to_native(key_term, key_stackbuf,
+        STACKBUF_SIZE, &key.length);
+    value.data = convert_value_to_native(value_term, value_stackbuf,
+        STACKBUF_SIZE, &value.length);
+    result = -1;
+
+    if(value.data != NULL && key.data != NULL) {
+        result = erts_osenv_put_native(env, &key, &value);
+    }
+
+    if(value.data != NULL && value.data != value_stackbuf) {
+        erts_free(ERTS_ALC_T_TMP, value.data);
+    }
+
+    if(key.data != NULL && key.data != key_stackbuf) {
+        erts_free(ERTS_ALC_T_TMP, key.data);
+    }
+
+    return result;
+}
+
+int erts_osenv_unset_term(erts_osenv_t *env, Eterm key_term) {
+    char key_stackbuf[STACKBUF_SIZE];
+    erts_osenv_data_t key;
+    int result;
+
+    key.data = convert_key_to_native(key_term, key_stackbuf,
+        STACKBUF_SIZE, &key.length);
+    result = -1;
+
+    if(key.data != NULL) {
+        result = erts_osenv_unset_native(env, &key);
+
+        if(key.data != key_stackbuf) {
+            erts_free(ERTS_ALC_T_TMP, key.data);
+        }
+    }
+
+    return result;
+}
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+struct __env_foreach_native {
+    erts_osenv_foreach_native_cb_t user_callback;
+    void *user_state;
+};
+
+static void foreach_native_wrapper(env_rbtnode_t *node, void *_state) {
+    struct __env_foreach_native *state = (struct __env_foreach_native*)_state;
+
+    state->user_callback(state->user_state, &node->key, &node->value);
+}
+
+void erts_osenv_foreach_native(const erts_osenv_t *env, void *state,
+        erts_osenv_foreach_native_cb_t callback) {
+    struct __env_foreach_native wrapper_state;
+
+    wrapper_state.user_callback = callback;
+    wrapper_state.user_state = state;
+
+    env_rbt_foreach(env->tree, foreach_native_wrapper, &wrapper_state);
+}
+
+int erts_osenv_get_native(const erts_osenv_t *env,
+        const erts_osenv_data_t *key,
+        erts_osenv_data_t *value) {
+    env_rbtnode_t *node = env_rbt_lookup(env->tree, *key);
+
+    if(node != NULL) {
+        if(value != NULL) {
+            if(node->value.length > value->length) {
+                return -1;
+            }
+
+            sys_memcpy(value->data, node->value.data, node->value.length);
+            value->length = node->value.length;
+        }
+
+        return 1;
+    }
+
+    return 0;
+}
+
+int erts_osenv_put_native(erts_osenv_t *env, const erts_osenv_data_t *key,
+        const erts_osenv_data_t *value) {
+    env_rbtnode_t *old_node, *new_node;
+
+    new_node = erts_alloc(ERTS_ALC_T_ENVIRONMENT, sizeof(env_rbtnode_t) +
+        key->length + value->length);
+
+    new_node->key.data = (char*)(&new_node[1]);
+    new_node->key.length = key->length;
+    new_node->value.data = &((char*)new_node->key.data)[key->length];
+    new_node->value.length = value->length;
+
+    sys_memcpy(new_node->key.data, key->data, key->length);
+    sys_memcpy(new_node->value.data, value->data, value->length);
+
+    old_node = env_rbt_lookup(env->tree, *key);
+
+    if(old_node != NULL) {
+        env->content_size -= old_node->value.length;
+        env->content_size -= old_node->key.length;
+        env_rbt_replace(&env->tree, old_node, new_node);
+    } else {
+        env_rbt_insert(&env->tree, new_node);
+        env->variable_count++;
+    }
+
+    env->content_size += new_node->value.length;
+    env->content_size += new_node->key.length;
+
+    if(old_node != NULL) {
+        erts_free(ERTS_ALC_T_ENVIRONMENT, old_node);
+    }
+
+    return 1;
+}
+
+int erts_osenv_unset_native(erts_osenv_t *env, const erts_osenv_data_t *key) {
+    env_rbtnode_t *old_node = env_rbt_lookup(env->tree, *key);
+
+    if(old_node != NULL) {
+        env->content_size -= old_node->value.length;
+        env->content_size -= old_node->key.length;
+        env->variable_count -= 1;
+
+        env_rbt_delete(&env->tree, old_node);
+        erts_free(ERTS_ALC_T_ENVIRONMENT, old_node);
+        return 1;
+    }
+
+    return 0;
+}
diff --git a/erts/emulator/sys/common/erl_osenv.h b/erts/emulator/sys/common/erl_osenv.h
new file mode 100644
index 0000000000..4777f2148a
--- /dev/null
+++ b/erts/emulator/sys/common/erl_osenv.h
@@ -0,0 +1,121 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+/* This is a replacement for getenv(3) and friends, operating on instances so
+ * we can keep a common implementation for both the global and local (per-port)
+ * environments.
+ *
+ * The instances are not thread-safe on their own but unlike getenv(3) we're
+ * guaranteed to be the only user, so placing locks around all our accesses
+ * will suffice.
+ *
+ * Use erts_sys_rwlock_global_osenv to access the global environment. */
+
+#ifndef __ERL_OSENV_H__
+#define __ERL_OSENV_H__
+
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+typedef struct __erts_osenv_data_t erts_osenv_data_t;
+
+typedef struct __erts_osenv_t {
+    struct __env_rbtnode_t *tree;
+    int variable_count;
+    int content_size;
+} erts_osenv_t;
+
+#include "sys.h"
+
+struct __erts_osenv_data_t {
+    Sint length;
+    void *data;
+};
+
+void erts_osenv_init(erts_osenv_t *env);
+void erts_osenv_clear(erts_osenv_t *env);
+
+/* @brief Merges \c with into \c env
+ *
+ * @param overwrite Whether to overwrite existing entries or keep them as they
+ * are. */
+void erts_osenv_merge(erts_osenv_t *env, const erts_osenv_t *with, int overwrite);
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+/* @brief Copies env[key] into \c value
+ *
+ * @return 1 on success, 0 if the key couldn't be found, and -1 if the input
+ * was invalid. */
+int erts_osenv_get_term(const erts_osenv_t *env, struct process *process,
+    Eterm key, Eterm *value);
+
+/* @brief Copies \c value into \c env[key]
+ *
+ * @return 1 on success, -1 if the input was invalid. */
+int erts_osenv_put_term(erts_osenv_t *env, Eterm key, Eterm value);
+
+/* @brief Removes \c env[key]
+ *
+ * @return 1 on success, 0 if the key couldn't be found, and -1 if the input
+ * was invalid. */
+int erts_osenv_unset_term(erts_osenv_t *env, Eterm key);
+
+/* @brief Copies env[key] into \c value
+ *
+ * @param value [in,out] The buffer to copy the value into, may be NULL if you
+ * only wish to query presence.
+ *
+ * @return 1 on success, 0 if the key couldn't be found, and -1 if if the value
+ * didn't fit into the buffer. */
+int erts_osenv_get_native(const erts_osenv_t *env, const erts_osenv_data_t *key,
+    erts_osenv_data_t *value);
+
+/* @brief Copies \c value into \c env[key]
+ *
+ * @return 1 on success, -1 on failure. */
+int erts_osenv_put_native(erts_osenv_t *env, const erts_osenv_data_t *key,
+    const erts_osenv_data_t *value);
+
+/* @brief Removes \c key from the env.
+ *
+ * @return 1 on success, 0 if the key couldn't be found. */
+int erts_osenv_unset_native(erts_osenv_t *env, const erts_osenv_data_t *key);
+
+/* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * */
+
+typedef void (*erts_osenv_foreach_term_cb_t)(struct process *process,
+    void *state, Eterm key, Eterm value);
+
+typedef void (*erts_osenv_foreach_native_cb_t)(void *state,
+    const erts_osenv_data_t *key,
+    const erts_osenv_data_t *value);
+
+/* @brief Walks through all environment variables, calling \c callback for each
+ * one. It's unsafe to modify \c env within the callback. */
+void erts_osenv_foreach_term(const erts_osenv_t *env, struct process *process,
+    void *state, erts_osenv_foreach_term_cb_t callback);
+
+/* @copydoc erts_osenv_foreach_term */
+void erts_osenv_foreach_native(const erts_osenv_t *env, void *state,
+    erts_osenv_foreach_native_cb_t callback);
+
+#endif
diff --git a/erts/emulator/sys/common/erl_poll.c b/erts/emulator/sys/common/erl_poll.c
index 52a8b6a53f..ced8a4a2a7 100644
--- a/erts/emulator/sys/common/erl_poll.c
+++ b/erts/emulator/sys/common/erl_poll.c
@@ -18,9 +18,8 @@
  * %CopyrightEnd%
  */
 
-/*
- * Description:	Poll interface suitable for ERTS with or without
- *              SMP support.
+/**
+ * @description Poll interface suitable for ERTS
  *
  *		The interface is currently implemented using:
  *		- select
@@ -29,12 +28,36 @@
  *              - epoll with poll or select as fallback
  *              - kqueue with poll or select as fallback
  *
- *		Some time in the future it will also be
- *		implemented using Solaris ports.
  *
+ * @author Rickard Green
+ * @author Lukas Larsson
+ *
+ * There are two major different implementations off IO polling in this
+ * file. The concurrent and non-concurrent implementations.
+ * When available epoll/kqueue are used to implement the concurrent
+ * versions. poll, select and dev/poll use non-concurrent updates.
+ *
+ * Concurrent version:
+ *   In the concurrent version erts_poll_control directly modifies
+ *   the kernel pollset without waking the thread that is waiting
+ *   on events. Also the ErtsPollResFd type is directly mapped to
+ *   the native event type, so no extra copying is needed. Note that
+ *   as no locking at all is done, fds can be triggered that have been
+ *   removed from the pollset. The check_io layer has to deal with this.
+ *
+ * Non-concurrent version:
+ *   In the non-concurrent version, the pollset has an internal representation
+ *   of the pollset that is updated by erts_poll_control. When an fd is updated,
+ *   its number is placed in the update request queue and then the waiting thread
+ *   is woken in order to see the change. The internal data in the pollset is
+ *   protected by a mutex that has to be taken by both the modifying and waiting
+ *   thread at different times.
  *
+ *   The non-concurrent pollset cannot have fd's closed in it while a thread is
+ *   waiting on that fd. In order to fix this, when an ERTS_POLL_OP_DEL command
+ *   is issued, the fd is marked as closing and the waiting thread is woken. The
+ *   fd is then returned in the waiting threads results as ERTS_POLL_EV_NONE.
  *
- * Author: 	Rickard Green
  */
 
 #ifdef HAVE_CONFIG_H
@@ -62,6 +85,8 @@
 #  ifdef SYS_SELECT_H
 #    include <sys/select.h>
 #  endif
+#elif defined(_DARWIN_UNLIMITED_SELECT)
+#  undef _DARWIN_UNLIMITED_SELECT
 #endif
 #ifdef NO_SYSCONF
 #  if ERTS_POLL_USE_SELECT
@@ -83,23 +108,35 @@
 #error "Missing implementation of erts_poll()"
 #endif
 
-#if defined(ERTS_KERNEL_POLL_VERSION) && !ERTS_POLL_USE_KERNEL_POLL
-#error "Missing kernel poll implementation of erts_poll()"
-#endif
+#if 0
+#define ERTS_POLL_DEBUG_PRINT 1
 
-#if defined(ERTS_NO_KERNEL_POLL_VERSION) && ERTS_POLL_USE_KERNEL_POLL
-#error "Kernel poll used when it shouldn't be used"
-#endif
+#define DEBUG_PRINT(FMT, PS, ...)                                       \
+    do {                                                                \
+        int myerrno = errno;                                            \
+        erts_printf("%d: " FMT "\r\n", (PS)->id, ##__VA_ARGS__);        \
+        errno = myerrno;                                                \
+    } while(0)
 
-#if 0
-#define ERTS_POLL_DEBUG_PRINT
+/* Define to print info about modifications done to each fd */
+#define DEBUG_PRINT_FD(FMT, PS, FD, ...) DEBUG_PRINT("%d: " FMT, PS, FD, ##__VA_ARGS__)
+/* Define to print entry and exit from erts_poll_wait (can be very spammy) */
+//#define DEBUG_PRINT_WAIT(FMT, PS, ...) DEBUG_PRINT(FMT, PS, ##__VA_ARGS__)
+
+#else
+#define ERTS_POLL_DEBUG_PRINT 0
+#define DEBUG_PRINT(...)
 #endif
 
-#if defined(DEBUG) && 0
-#define HARD_DEBUG
+#ifndef DEBUG_PRINT_FD
+#define DEBUG_PRINT_FD(...)
+#endif
+#ifndef DEBUG_PRINT_WAIT
+#define DEBUG_PRINT_WAIT(...)
 #endif
 
-#ifdef _DARWIN_UNLIMITED_SELECT
+
+#if defined(_DARWIN_UNLIMITED_SELECT) && ERTS_POLL_USE_SELECT
 typedef struct {
     size_t sz;
     fd_set* ptr;
@@ -145,74 +182,43 @@ int ERTS_SELECT(int nfds, ERTS_fd_set *readfds, ERTS_fd_set *writefds,
 #  define ERTS_SELECT	select
 #endif
 
-#define ERTS_POLL_USE_BATCH_UPDATE_POLLSET (ERTS_POLL_USE_DEVPOLL \
-					    || ERTS_POLL_USE_KQUEUE)
-#define ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE \
-   (defined(ERTS_SMP) || ERTS_POLL_USE_KERNEL_POLL || ERTS_POLL_USE_POLL)
+#define ERTS_POLL_IS_FALLBACK (ERTS_POLL_USE_POLL || ERTS_POLL_USE_SELECT) && ERTS_ENABLE_KERNEL_POLL
 
-#define ERTS_POLL_USE_CONCURRENT_UPDATE \
-   (defined(ERTS_SMP) && ERTS_POLL_USE_EPOLL)
+#define ERTS_POLL_USE_CONCURRENT_UPDATE (ERTS_POLL_USE_EPOLL || ERTS_POLL_USE_KQUEUE)
 
-#define ERTS_POLL_COALESCE_KP_RES (ERTS_POLL_USE_KQUEUE || ERTS_POLL_USE_EPOLL)
+#define ERTS_POLL_USE_WAKEUP_PIPE (!ERTS_POLL_USE_CONCURRENT_UPDATE)
 
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#  define ERTS_POLL_ASYNC_INTERRUPT_SUPPORT 1
-#else
-#  define ERTS_POLL_ASYNC_INTERRUPT_SUPPORT 0
-#endif
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 
-#define ERTS_POLL_USE_WAKEUP_PIPE \
-   (ERTS_POLL_ASYNC_INTERRUPT_SUPPORT || defined(USE_THREADS))
-
-#ifdef ERTS_SMP
+#define ERTS_POLLSET_SET_HAVE_UPDATE_REQUESTS(PS) \
+  erts_atomic32_set_nob(&(PS)->have_update_requests, (erts_aint32_t) 1)
+#define ERTS_POLLSET_UNSET_HAVE_UPDATE_REQUESTS(PS) \
+  erts_atomic32_set_nob(&(PS)->have_update_requests, (erts_aint32_t) 0)
+#define ERTS_POLLSET_HAVE_UPDATE_REQUESTS(PS) \
+  ((int) erts_atomic32_read_nob(&(PS)->have_update_requests))
 
 #define ERTS_POLLSET_LOCK(PS) \
-  erts_smp_mtx_lock(&(PS)->mtx)
+  erts_mtx_lock(&(PS)->mtx)
 #define ERTS_POLLSET_UNLOCK(PS) \
-  erts_smp_mtx_unlock(&(PS)->mtx)
-
-#define ERTS_POLLSET_SET_POLLED_CHK(PS) \
-  ((int) erts_atomic32_xchg_nob(&(PS)->polled, (erts_aint32_t) 1))
-#define ERTS_POLLSET_UNSET_POLLED(PS) \
-  erts_atomic32_set_nob(&(PS)->polled, (erts_aint32_t) 0)
-#define ERTS_POLLSET_IS_POLLED(PS) \
-  ((int) erts_atomic32_read_nob(&(PS)->polled))
+  erts_mtx_unlock(&(PS)->mtx)
 
 #else
 
-#define ERTS_POLLSET_LOCK(PS)
-#define ERTS_POLLSET_UNLOCK(PS)
-#define ERTS_POLLSET_SET_POLLED_CHK(PS) 0
-#define ERTS_POLLSET_UNSET_POLLED(PS)
-#define ERTS_POLLSET_IS_POLLED(PS) 0
-
-#endif
-
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
-#define ERTS_POLLSET_SET_HAVE_UPDATE_REQUESTS(PS) \
-  erts_smp_atomic32_set_nob(&(PS)->have_update_requests, (erts_aint32_t) 1)
-#define ERTS_POLLSET_UNSET_HAVE_UPDATE_REQUESTS(PS) \
-  erts_smp_atomic32_set_nob(&(PS)->have_update_requests, (erts_aint32_t) 0)
-#define ERTS_POLLSET_HAVE_UPDATE_REQUESTS(PS) \
-  ((int) erts_smp_atomic32_read_nob(&(PS)->have_update_requests))
-#else
 #define ERTS_POLLSET_SET_HAVE_UPDATE_REQUESTS(PS)
 #define ERTS_POLLSET_UNSET_HAVE_UPDATE_REQUESTS(PS)
 #define ERTS_POLLSET_HAVE_UPDATE_REQUESTS(PS) 0
-#endif
 
-#if ERTS_POLL_USE_FALLBACK
-#  if ERTS_POLL_USE_POLL
-#    define ERTS_POLL_NEED_FALLBACK(PS) ((PS)->no_poll_fds > 1)
-#  elif ERTS_POLL_USE_SELECT
-#    define ERTS_POLL_NEED_FALLBACK(PS) ((PS)->no_select_fds > 1)
-#  endif
+#define ERTS_POLLSET_LOCK(PS)
+#define ERTS_POLLSET_UNLOCK(PS)
+
 #endif
+
 /*
  * --- Data types ------------------------------------------------------------
  */
 
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+
 #define ERTS_POLLSET_UPDATE_REQ_BLOCK_SIZE 128
 
 typedef struct ErtsPollSetUpdateRequestsBlock_ ErtsPollSetUpdateRequestsBlock;
@@ -222,206 +228,146 @@ struct ErtsPollSetUpdateRequestsBlock_ {
     int fds[ERTS_POLLSET_UPDATE_REQ_BLOCK_SIZE];
 };
 
-#endif
-
-
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
 #  define ERTS_POLL_FD_FLG_INURQ	(((unsigned short) 1) << 0)
-#endif
-#if ERTS_POLL_USE_FALLBACK
-#  define ERTS_POLL_FD_FLG_INFLBCK	(((unsigned short) 1) << 1)
-#  define ERTS_POLL_FD_FLG_USEFLBCK	(((unsigned short) 1) << 2)
-#endif
-#if ERTS_POLL_USE_KERNEL_POLL || defined(ERTS_SMP)
-#  define ERTS_POLL_FD_FLG_RST		(((unsigned short) 1) << 3)
-#endif
+#  define ERTS_POLL_FD_FLG_RST		(((unsigned short) 1) << 1)
+
 typedef struct {
 #if ERTS_POLL_USE_POLL
     int pix;
 #endif
+
     ErtsPollEvents used_events;
     ErtsPollEvents events;
-#if ERTS_POLL_COALESCE_KP_RES
-    unsigned short res_ev_ix;
-#endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE || ERTS_POLL_USE_FALLBACK
     unsigned short flags;
-#endif
 
 } ErtsFdStatus;
 
-
-#if ERTS_POLL_COALESCE_KP_RES
-/* res_ev_ix max value */
-#define ERTS_POLL_MAX_RES ((1 << sizeof(unsigned short)*8) - 1)
-#endif
-
-#if ERTS_POLL_USE_KQUEUE
-
-#define ERTS_POLL_KQ_OP_HANDLED			1
-#define ERTS_POLL_KQ_OP_DEL_R			2
-#define ERTS_POLL_KQ_OP_DEL_W			3
-#define ERTS_POLL_KQ_OP_ADD_R			4
-#define ERTS_POLL_KQ_OP_ADD_W			5
-#define ERTS_POLL_KQ_OP_ADD2_R			6
-#define ERTS_POLL_KQ_OP_ADD2_W			7
-
 #endif
 
-struct ErtsPollSet_ {
-    ErtsPollSet next;
+/*
+ * This struct is not really exported, but it's nice to
+ * get unique names in debugger for kp/nkp
+ */
+struct ERTS_POLL_EXPORT(erts_pollset) {
+    int id;
     int internal_fd_limit;
-    ErtsFdStatus *fds_status;
-    erts_smp_atomic_t no_of_user_fds;
-    int fds_status_len;
+    erts_atomic_t no_of_user_fds;
+
 #if ERTS_POLL_USE_KERNEL_POLL
     int kp_fd;
-    int res_events_len;
-#if ERTS_POLL_USE_EPOLL
-    struct epoll_event *res_events;
-#elif ERTS_POLL_USE_KQUEUE
-    struct kevent *res_events;
-#elif ERTS_POLL_USE_DEVPOLL
-    struct pollfd *res_events;
-#endif
 #endif /* ERTS_POLL_USE_KERNEL_POLL */
+
 #if ERTS_POLL_USE_POLL
     int next_poll_fds_ix;
     int no_poll_fds;
     int poll_fds_len;
-    struct pollfd*poll_fds;
+    struct pollfd *poll_fds;
 #elif ERTS_POLL_USE_SELECT
     int next_sel_fd;
     int max_fd;
-#if ERTS_POLL_USE_FALLBACK
-    int no_select_fds;
-#endif
     ERTS_fd_set input_fds;
     ERTS_fd_set res_input_fds;
     ERTS_fd_set output_fds;
     ERTS_fd_set res_output_fds;
+#elif ERTS_POLL_USE_DEVPOLL
+    struct pollfd *poll_fds;
+    int poll_fds_ix;
 #endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+    ErtsFdStatus *fds_status;
+    int fds_status_len;
     ErtsPollSetUpdateRequestsBlock update_requests;
     ErtsPollSetUpdateRequestsBlock *curr_upd_req_block;
-    erts_smp_atomic32_t have_update_requests;
-#endif
-#ifdef ERTS_SMP
-    erts_atomic32_t polled;
-    erts_smp_mtx_t mtx;
+    erts_atomic32_t have_update_requests;
+    erts_mtx_t mtx;
+    erts_atomic32_t wakeup_state;
 #endif
+
 #if ERTS_POLL_USE_WAKEUP_PIPE
     int wake_fds[2];
 #endif
-#if ERTS_POLL_USE_TIMERFD
-    int timer_fd;
-#endif
-#if ERTS_POLL_USE_FALLBACK
-    int fallback_used;
-#endif
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
-    erts_atomic32_t wakeup_state;
-#endif
-    erts_atomic64_t timeout_time;
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-    erts_smp_atomic_t no_avoided_wakeups;
-    erts_smp_atomic_t no_avoided_interrupts;
-    erts_smp_atomic_t no_interrupt_timed;
-#endif
 };
 
 void erts_silence_warn_unused_result(long unused);
 static void fatal_error(char *format, ...);
-static void fatal_error_async_signal_safe(char *error_str);
 
 static int max_fds = -1;
-static ErtsPollSet pollsets;
-static erts_smp_mtx_t pollsets_lock;
 
 #if ERTS_POLL_USE_POLL
 
+#if !ERTS_POLL_IS_FALLBACK
+static ERTS_INLINE short ev2pollev(ErtsPollEvents ev)
+{
+    return ERTS_POLL_EV_E2N(ev);
+}
+
+static ERTS_INLINE ErtsPollEvents pollev2ev(short ev)
+{
+    return ERTS_POLL_EV_N2E(ev);
+}
+
+#else /* ERTS_POLL_IS_FALLBACK */
+
 static ERTS_INLINE short
 ev2pollev(ErtsPollEvents ev)
 {
-#if !ERTS_POLL_USE_FALLBACK || ERTS_POLL_USE_KQUEUE
-    return ERTS_POLL_EV_E2N(ev);
-#else /* Note, we only map events we are interested in */
     short res_ev = (short) 0;
     if (ev & ERTS_POLL_EV_IN)
-	res_ev |= ERTS_POLL_EV_NKP_IN;
+       res_ev |= ERTS_POLL_EV_NKP_IN;
     if (ev & ERTS_POLL_EV_OUT)
-	res_ev |= ERTS_POLL_EV_NKP_OUT;
+       res_ev |= ERTS_POLL_EV_NKP_OUT;
     return res_ev;
-#endif
 }
 
 static ERTS_INLINE ErtsPollEvents
 pollev2ev(short ev)
 {
-#if !ERTS_POLL_USE_FALLBACK || ERTS_POLL_USE_KQUEUE
-    return ERTS_POLL_EV_N2E(ev);
-#else /* Note, we only map events we are interested in */
     ErtsPollEvents res_ev = (ErtsPollEvents) 0;
     if (ev & ERTS_POLL_EV_NKP_IN)
-	res_ev |= ERTS_POLL_EV_IN;
+       res_ev |= ERTS_POLL_EV_IN;
     if (ev & ERTS_POLL_EV_NKP_OUT)
-	res_ev |= ERTS_POLL_EV_OUT;
+       res_ev |= ERTS_POLL_EV_OUT;
     if (ev & ERTS_POLL_EV_NKP_ERR)
-	res_ev |= ERTS_POLL_EV_ERR;
+       res_ev |= ERTS_POLL_EV_ERR;
     if (ev & ERTS_POLL_EV_NKP_NVAL)
-	res_ev |= ERTS_POLL_EV_NVAL;
-    return res_ev;
-#endif
+       res_ev |= ERTS_POLL_EV_NVAL;
+   return res_ev;
 }
 
-#endif
+#endif /* !ERTS_POLL_IS_FALLBACK */
+
+#endif /* ERTS_POLL_USE_POLL */
+
 
 #ifdef HARD_DEBUG
 static void check_poll_result(ErtsPollResFd pr[], int len);
-#if ERTS_POLL_USE_DEVPOLL
-static void check_poll_status(ErtsPollSet ps);
-#endif /* ERTS_POLL_USE_DEVPOLL */
 #endif /* HARD_DEBUG */
-#ifdef ERTS_POLL_DEBUG_PRINT
+#if ERTS_POLL_USE_DEVPOLL && defined(DEBUG)
+static void check_poll_status(ErtsPollSet *ps);
+#endif /* ERTS_POLL_USE_DEVPOLL && DEBUG */
 static void print_misc_debug_info(void);
+#if ERTS_POLL_USE_EPOLL
+uint32_t epoll_events(int kp_fd, int fd);
 #endif
 
-static ERTS_INLINE void
-init_timeout_time(ErtsPollSet ps)
-{
-    erts_atomic64_init_nob(&ps->timeout_time,
-			   (erts_aint64_t) ERTS_MONOTONIC_TIME_MAX);
-}
-
-static ERTS_INLINE void
-set_timeout_time(ErtsPollSet ps, ErtsMonotonicTime time)
-{
-    erts_atomic64_set_relb(&ps->timeout_time,
-			   (erts_aint64_t) time);
-}
-
-static ERTS_INLINE ErtsMonotonicTime
-get_timeout_time(ErtsPollSet ps)
-{
-    return (ErtsMonotonicTime) erts_atomic64_read_acqb(&ps->timeout_time);
-}
 
 #define ERTS_POLL_NOT_WOKEN	0
 #define ERTS_POLL_WOKEN		-1
 #define ERTS_POLL_WOKEN_INTR	1
 
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 static ERTS_INLINE void
-reset_wakeup_state(ErtsPollSet ps)
+reset_wakeup_state(ErtsPollSet *ps)
 {
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
     erts_atomic32_set_mb(&ps->wakeup_state, ERTS_POLL_NOT_WOKEN);
-#endif
 }
+#endif
 
 static ERTS_INLINE int
-is_woken(ErtsPollSet ps)
+is_woken(ErtsPollSet *ps)
 {
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     return erts_atomic32_read_acqb(&ps->wakeup_state) != ERTS_POLL_NOT_WOKEN;
 #else
     return 0;
@@ -429,9 +375,9 @@ is_woken(ErtsPollSet ps)
 }
 
 static ERTS_INLINE int
-is_interrupted_reset(ErtsPollSet ps)
+is_interrupted_reset(ErtsPollSet *ps)
 {
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     return (erts_atomic32_xchg_acqb(&ps->wakeup_state, ERTS_POLL_NOT_WOKEN)
 	    == ERTS_POLL_WOKEN_INTR);
 #else
@@ -440,9 +386,9 @@ is_interrupted_reset(ErtsPollSet ps)
 }
 
 static ERTS_INLINE void
-woke_up(ErtsPollSet ps)
+woke_up(ErtsPollSet *ps)
 {
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     erts_aint32_t wakeup_state = erts_atomic32_read_acqb(&ps->wakeup_state);
     if (wakeup_state == ERTS_POLL_NOT_WOKEN)
 	(void) erts_atomic32_cmpxchg_nob(&ps->wakeup_state,
@@ -459,28 +405,23 @@ woke_up(ErtsPollSet ps)
 #if ERTS_POLL_USE_WAKEUP_PIPE
 
 static ERTS_INLINE void
-wake_poller(ErtsPollSet ps, int interrupted, int async_signal_safe)
+wake_poller(ErtsPollSet *ps, int interrupted)
 {
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     int wake;
-    if (async_signal_safe)
-	wake = 1;
-    else {
-	erts_aint32_t wakeup_state;
-	if (!interrupted)
-	    wakeup_state = erts_atomic32_cmpxchg_relb(&ps->wakeup_state,
-						      ERTS_POLL_WOKEN,
-						      ERTS_POLL_NOT_WOKEN);
-	else
-            wakeup_state = erts_atomic32_xchg_relb(&ps->wakeup_state,
-                                                   ERTS_POLL_WOKEN_INTR);
-	wake = wakeup_state == ERTS_POLL_NOT_WOKEN;
-    }
-    /*
-     * NOTE: This function might be called from signal handlers in the
-     *       non-smp case; therefore, it has to be async-signal safe in
-     *       the non-smp case.
-     */
-    if (wake) {
+    erts_aint32_t wakeup_state;
+    if (!interrupted)
+        wakeup_state = erts_atomic32_cmpxchg_relb(&ps->wakeup_state,
+                                                  ERTS_POLL_WOKEN,
+                                                  ERTS_POLL_NOT_WOKEN);
+    else
+        wakeup_state = erts_atomic32_xchg_relb(&ps->wakeup_state,
+                                               ERTS_POLL_WOKEN_INTR);
+    wake = wakeup_state == ERTS_POLL_NOT_WOKEN;
+
+    if (wake)
+#endif
+    {
 	ssize_t res;
 	if (ps->wake_fds[1] < 0)
 	    return; /* Not initialized yet */
@@ -489,36 +430,27 @@ wake_poller(ErtsPollSet ps, int interrupted, int async_signal_safe)
 	    res = write(ps->wake_fds[1], "!", 1);
 	} while (res < 0 && errno == EINTR);
 	if (res <= 0 && errno != ERRNO_BLOCK) {
-	    if (async_signal_safe)
-		fatal_error_async_signal_safe(__FILE__
-					      ":XXX:wake_poller(): "
-					      "Failed to write on wakeup pipe\n");
-	    else
-		fatal_error("%s:%d:wake_poller(): "
-			    "Failed to write to wakeup pipe fd=%d: "
-			    "%s (%d)\n",
-			    __FILE__, __LINE__,
-			    ps->wake_fds[1],
-			    erl_errno_id(errno), errno);
+            fatal_error("%s:%d:wake_poller(): "
+                        "Failed to write to wakeup pipe fd=%d: "
+                        "%s (%d)\n",
+                        __FILE__, __LINE__,
+                        ps->wake_fds[1],
+                        erl_errno_id(errno), errno);
 	}
     }
 }
 
 static ERTS_INLINE void
-cleanup_wakeup_pipe(ErtsPollSet ps)
+cleanup_wakeup_pipe(ErtsPollSet *ps)
 {
-#if ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
     int intr = 0;
-#endif
     int fd = ps->wake_fds[0];
     int res;
     do {
 	char buf[32];
 	res = read(fd, buf, sizeof(buf));
-#if ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
 	if (res > 0)
 	    intr = 1;
-#endif
     } while (res > 0 || (res < 0 && errno == EINTR));
     if (res < 0 && errno != ERRNO_BLOCK) {
 	fatal_error("%s:%d:cleanup_wakeup_pipe(): "
@@ -528,14 +460,14 @@ cleanup_wakeup_pipe(ErtsPollSet ps)
 		    fd,
 		    erl_errno_id(errno), errno);
     }
-#if ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     if (intr)
 	erts_atomic32_set_nob(&ps->wakeup_state, ERTS_POLL_WOKEN_INTR);
 #endif
 }
 
 static void
-create_wakeup_pipe(ErtsPollSet ps)
+create_wakeup_pipe(ErtsPollSet *ps)
 {
     int do_wake = 0;
     int wake_fds[2];
@@ -552,20 +484,13 @@ create_wakeup_pipe(ErtsPollSet ps)
     SET_NONBLOCKING(wake_fds[0]);
     SET_NONBLOCKING(wake_fds[1]);
 
-#ifdef ERTS_POLL_DEBUG_PRINT
-    erts_printf("wakeup fds = {%d, %d}\n", wake_fds[0], wake_fds[1]);
-#endif
+    DEBUG_PRINT("wakeup fds = {%d, %d}", ps, wake_fds[0], wake_fds[1]);
 
     ERTS_POLL_EXPORT(erts_poll_control)(ps,
 					wake_fds[0],
+                                        ERTS_POLL_OP_ADD,
 					ERTS_POLL_EV_IN,
-					1, &do_wake);
-#if ERTS_POLL_USE_FALLBACK
-    /* We depend on the wakeup pipe being handled by kernel poll */
-    if (ps->fds_status[wake_fds[0]].flags & ERTS_POLL_FD_FLG_INFLBCK)
-	fatal_error("%s:%d:create_wakeup_pipe(): Internal error\n",
-		    __FILE__, __LINE__);
-#endif
+                                        &do_wake);
     if (ps->internal_fd_limit <= wake_fds[1])
 	ps->internal_fd_limit = wake_fds[1] + 1;
     if (ps->internal_fd_limit <= wake_fds[0])
@@ -574,84 +499,16 @@ create_wakeup_pipe(ErtsPollSet ps)
     ps->wake_fds[1] = wake_fds[1];
 }
 
-#endif /* ERTS_POLL_USE_WAKEUP_PIPE */
-
-/*
- * --- timer fd -----------------------------------------------------------
- */
-
-#if ERTS_POLL_USE_TIMERFD
-
-/* We use the timerfd when using epoll_wait to get high accuracy
-   timeouts, i.e. we want to sleep with < ms accuracy. */
-
-static void
-create_timerfd(ErtsPollSet ps)
-{
-    int do_wake = 0;
-    int timer_fd;
-    timer_fd = timerfd_create(CLOCK_MONOTONIC,0);
-    ERTS_POLL_EXPORT(erts_poll_control)(ps,
-					timer_fd,
-					ERTS_POLL_EV_IN,
-					1, &do_wake);
-#if ERTS_POLL_USE_FALLBACK
-    /* We depend on the wakeup pipe being handled by kernel poll */
-    if (ps->fds_status[timer_fd].flags & ERTS_POLL_FD_FLG_INFLBCK)
-	fatal_error("%s:%d:create_wakeup_pipe(): Internal error\n",
-		    __FILE__, __LINE__);
-#endif
-    if (ps->internal_fd_limit <= timer_fd)
-	ps->internal_fd_limit = timer_fd + 1;
-    ps->timer_fd = timer_fd;
-}
-
-static ERTS_INLINE void
-timerfd_set(ErtsPollSet ps, struct itimerspec *its)
-{
-#ifdef DEBUG
-    struct itimerspec old_its;
-    int res;
-    res = timerfd_settime(ps->timer_fd, 0, its, &old_its);
-    ASSERT(res == 0);
-    ASSERT(old_its.it_interval.tv_sec == 0 &&
-           old_its.it_interval.tv_nsec == 0 &&
-           old_its.it_value.tv_sec == 0 &&
-           old_its.it_value.tv_nsec == 0);
-
-#else
-    timerfd_settime(ps->timer_fd, 0, its, NULL);
 #endif
-}
-
-static ERTS_INLINE int
-timerfd_clear(ErtsPollSet ps, int res, int max_res) {
-
-    struct itimerspec its;
-    /* we always have to clear the timer */
-    its.it_interval.tv_sec = 0;
-    its.it_interval.tv_nsec = 0;
-    its.it_value.tv_sec = 0;
-    its.it_value.tv_nsec = 0;
-    timerfd_settime(ps->timer_fd, 0, &its, NULL);
-
-    /* only timeout fd triggered */
-    if (res == 1 && ps->res_events[0].data.fd == ps->timer_fd)
-        return 0;
-
-    return res;
-}
-
-#endif /* ERTS_POLL_USE_TIMERFD */
-
 
 /*
  * --- Poll set update requests ----------------------------------------------
  */
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 
 static ERTS_INLINE void
-enqueue_update_request(ErtsPollSet ps, int fd)
+enqueue_update_request(ErtsPollSet *ps, int fd)
 {
     ErtsPollSetUpdateRequestsBlock *urqbp;
 
@@ -666,13 +523,11 @@ enqueue_update_request(ErtsPollSet ps, int fd)
     urqbp = ps->curr_upd_req_block;
 
     if (urqbp->len == ERTS_POLLSET_UPDATE_REQ_BLOCK_SIZE) {
-	ASSERT(!urqbp->next);
 	urqbp = erts_alloc(ERTS_ALC_T_POLLSET_UPDREQ,
 			   sizeof(ErtsPollSetUpdateRequestsBlock));
-	ps->curr_upd_req_block->next = urqbp;
-	ps->curr_upd_req_block = urqbp;
-	urqbp->next = NULL;
+	urqbp->next = ps->curr_upd_req_block;
 	urqbp->len = 0;
+	ps->curr_upd_req_block = urqbp;
     }
 
     ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_INURQ;
@@ -680,29 +535,29 @@ enqueue_update_request(ErtsPollSet ps, int fd)
 }
 
 static ERTS_INLINE void
-free_update_requests_block(ErtsPollSet ps,
+free_update_requests_block(ErtsPollSet *ps,
 			   ErtsPollSetUpdateRequestsBlock *urqbp)
 {
     if (urqbp != &ps->update_requests)
 	erts_free(ERTS_ALC_T_POLLSET_UPDREQ, (void *) urqbp);
     else {
-	urqbp->next = NULL;
 	urqbp->len = 0;
     }
 }
 
-#endif /* ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE */
+#endif /* !ERTS_POLL_USE_CONCURRENT_UPDATE */
 
 /*
  * --- Growing poll set structures -------------------------------------------
  */
 
-#ifndef ERTS_KERNEL_POLL_VERSION   /* only one shared implementation */
+#if !ERTS_NO_KERNEL_POLL_VERSION || !ERTS_ENABLE_KERNEL_POLL
+/* only one shared implementation */
 
 #define ERTS_FD_TABLE_MIN_LENGTH	1024
 #define ERTS_FD_TABLE_EXP_THRESHOLD	(2048*1024)
 
-int erts_poll_new_table_len (int old_len, int need_len)
+int erts_poll_new_table_len(int old_len, int need_len)
 {
     int new_len;
 
@@ -712,7 +567,7 @@ int erts_poll_new_table_len (int old_len, int need_len)
     }
     else {
         new_len = old_len;
-        do {            
+        do {
             if (new_len < ERTS_FD_TABLE_EXP_THRESHOLD)
                 new_len *= 2;
             else
@@ -725,30 +580,9 @@ int erts_poll_new_table_len (int old_len, int need_len)
 }
 #endif
 
-#if ERTS_POLL_USE_KERNEL_POLL
-static void
-grow_res_events(ErtsPollSet ps, int new_len)
-{
-    size_t new_size = sizeof(
-#if ERTS_POLL_USE_EPOLL
-	struct epoll_event
-#elif ERTS_POLL_USE_DEVPOLL
-	struct pollfd
-#elif ERTS_POLL_USE_KQUEUE
-	struct kevent
-#endif
-	) * erts_poll_new_table_len(ps->res_events_len, new_len);
-    /* We do not need to save previously stored data */
-    if (ps->res_events)
-	erts_free(ERTS_ALC_T_POLL_RES_EVS, ps->res_events);
-    ps->res_events = erts_alloc(ERTS_ALC_T_POLL_RES_EVS, new_size);
-    ps->res_events_len = new_len;
-}
-#endif /* ERTS_POLL_USE_KERNEL_POLL */
-
 #if ERTS_POLL_USE_POLL
 static void
-grow_poll_fds(ErtsPollSet ps, int min_ix)
+grow_poll_fds(ErtsPollSet *ps, int min_ix)
 {
     int i;
     int new_len = erts_poll_new_table_len(ps->poll_fds_len, min_ix + 1);
@@ -796,8 +630,9 @@ ensure_select_fds(int fd, ERTS_fd_set* in, ERTS_fd_set* out)
 #  define ensure_select_fds(fd, in, out) do {} while(0)
 #endif /* _DARWIN_UNLIMITED_SELECT */
 
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 static void
-grow_fds_status(ErtsPollSet ps, int min_fd)
+grow_fds_status(ErtsPollSet *ps, int min_fd)
 {
     int i;
     int new_len = erts_poll_new_table_len(ps->fds_status_len, min_fd + 1);
@@ -816,430 +651,26 @@ grow_fds_status(ErtsPollSet ps, int min_fd)
 #endif
 	ps->fds_status[i].used_events = (ErtsPollEvents) 0;
 	ps->fds_status[i].events = (ErtsPollEvents) 0;
-#if ERTS_POLL_COALESCE_KP_RES
-	ps->fds_status[i].res_ev_ix = (unsigned short) ERTS_POLL_MAX_RES;
-#endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE || ERTS_POLL_USE_FALLBACK
 	ps->fds_status[i].flags = (unsigned short) 0;
-#endif
     }
     ps->fds_status_len = new_len;
 }
+#endif
 
 /*
  * --- Selecting fd to poll on -----------------------------------------------
  */
 
-#if ERTS_POLL_USE_FALLBACK
-static int update_fallback_pollset(ErtsPollSet ps, int fd);
-#endif
-
-static ERTS_INLINE int
-need_update(ErtsPollSet ps, int fd)
-{
-#if ERTS_POLL_USE_KERNEL_POLL
-    int reset;
-#endif
-
-    ASSERT(fd < ps->fds_status_len);
-
-#if ERTS_POLL_USE_KERNEL_POLL
-    reset = (int) (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST);
-    if (reset && !ps->fds_status[fd].used_events) {
-	ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_RST;
-	reset = 0;
-    }
-#elif defined(ERTS_SMP)
-    ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_RST;
-#endif
-
-    if (ps->fds_status[fd].used_events != ps->fds_status[fd].events)
-	return 1;
-
-#if ERTS_POLL_USE_KERNEL_POLL
-    return reset;
-#else
-    return 0;
-#endif
-}
-
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-
-#if ERTS_POLL_USE_KQUEUE
-#define ERTS_POLL_MIN_BATCH_BUF_SIZE 128
-#else
-#define ERTS_POLL_MIN_BATCH_BUF_SIZE 64
-#endif
-
-typedef struct {
-    int len;
-    int size;
-#if ERTS_POLL_USE_DEVPOLL
-    struct pollfd *buf;
-#elif ERTS_POLL_USE_KQUEUE
-    struct kevent *buf;
-    struct kevent *ebuf;
-#endif
-} ErtsPollBatchBuf;
-
-
-static ERTS_INLINE void
-setup_batch_buf(ErtsPollSet ps, ErtsPollBatchBuf *bbp)
-{
-    bbp->len = 0;
-#if ERTS_POLL_USE_DEVPOLL
-    bbp->size = ps->res_events_len;
-    bbp->buf = ps->res_events;
-#elif ERTS_POLL_USE_KQUEUE
-    bbp->size = ps->res_events_len/2;
-    bbp->buf = ps->res_events;
-    bbp->ebuf = bbp->buf + bbp->size;
-#endif
-}
-
-
-#if ERTS_POLL_USE_DEVPOLL
-
-static void
-write_batch_buf(ErtsPollSet ps, ErtsPollBatchBuf *bbp)
-{
-    ssize_t wres;
-    char *buf = (char *) bbp->buf;
-    size_t buf_size = sizeof(struct pollfd)*bbp->len;
-    
-    while (1) {
-	wres = write(ps->kp_fd, (void *) buf, buf_size);
-	if (wres < 0) {
-	    if (errno == EINTR)
-		continue;
-	    fatal_error("%s:%d:write_batch_buf(): "
-			"Failed to write to /dev/poll: "
-			"%s (%d)\n",
-			__FILE__, __LINE__,
-			erl_errno_id(errno), errno);
-	}
-	buf_size -= wres;
-	if (buf_size <= 0)
-	    break;
-	buf += wres;
-    }
-
-    if (buf_size < 0) {
-	fatal_error("%s:%d:write_devpoll_buf(): Internal error\n",
-		    __FILE__, __LINE__);
-    }
-    bbp->len = 0;
-}
-
-#elif ERTS_POLL_USE_KQUEUE
-
-static void
-write_batch_buf(ErtsPollSet ps, ErtsPollBatchBuf *bbp)
-{
-    int res;
-    int len = bbp->len;
-    struct kevent *buf = bbp->buf;
-    struct timespec ts = {0, 0};
-
-    do {
-	res = kevent(ps->kp_fd, buf, len, NULL, 0, &ts);
-    } while (res < 0 && errno == EINTR);
-    if (res < 0) {
-	int i;
-	struct kevent *ebuf = bbp->ebuf;
-	do {
-	    res = kevent(ps->kp_fd, buf, len, ebuf, len, &ts);
-	} while (res < 0 && errno == EINTR);
-	if (res < 0) {
-	    fatal_error("%s:%d: kevent() failed: %s (%d)\n",
-			__FILE__, __LINE__, erl_errno_id(errno), errno);
-	}
-	for (i = 0; i < res; i++) {
-	    if (ebuf[i].flags & EV_ERROR) {
-		short filter;
-		int fd = (int) ebuf[i].ident;
-
-		switch ((int) (long) ebuf[i].udata) {
-
-		    /*
-		     * Since we use a lazy update approach EV_DELETE will
-		     * frequently fail. This since kqueue automatically
-		     * removes a file descriptor that is closed from the
-		     * poll set.
-		     */
-		case ERTS_POLL_KQ_OP_DEL_R:
-		case ERTS_POLL_KQ_OP_DEL_W:
-		case ERTS_POLL_KQ_OP_HANDLED:
-		    break;
-
-		    /*
-		     * According to the kqueue man page EVFILT_READ support
-		     * does not imply EVFILT_WRITE support; therefore,
-		     * if an EV_ADD fail, we may have to remove other
-		     * events on this fd in the kqueue pollset before
-		     * adding fd to the fallback pollset.
-		     */
-		case ERTS_POLL_KQ_OP_ADD_W:
-		    if (ps->fds_status[fd].used_events & ERTS_POLL_EV_IN) {
-			filter = EVFILT_READ;
-			goto rm_add_fb;
-		    }
-		    goto add_fb;
-		case ERTS_POLL_KQ_OP_ADD_R:
-		    if (ps->fds_status[fd].used_events & ERTS_POLL_EV_OUT) {
-			filter = EVFILT_WRITE;
-			goto rm_add_fb;
-		    }
-		    goto add_fb;
-		case ERTS_POLL_KQ_OP_ADD2_W:
-		case ERTS_POLL_KQ_OP_ADD2_R: {
-		    int j;
-		    for (j = i+1; j < res; j++) {
-			if (fd == (int) ebuf[j].ident) {
-			    ebuf[j].udata = (void *) ERTS_POLL_KQ_OP_HANDLED;
-			    if (!(ebuf[j].flags & EV_ERROR)) {
-				switch ((int) (long) ebuf[j].udata) {
-				case ERTS_POLL_KQ_OP_ADD2_W:
-				    filter = EVFILT_WRITE;
-				    goto rm_add_fb;
-				case ERTS_POLL_KQ_OP_ADD2_R:
-				    filter = EVFILT_READ;
-				    goto rm_add_fb;
-				default:
-				    fatal_error("%s:%d:write_batch_buf(): "
-						"Internal error",
-						__FILE__, __LINE__);
-				    break;
-				}
-			    }
-			    goto add_fb;
-			}
-		    }
-		    /* The other add succeded... */
-		    filter = ((((int) (long) ebuf[i].udata)
-			       == ERTS_POLL_KQ_OP_ADD2_W)
-			      ? EVFILT_READ
-			      : EVFILT_WRITE);
-		rm_add_fb:
-		    { 
-			struct kevent kev;
-			struct timespec ts = {0, 0};
-			EV_SET(&kev, fd, filter, EV_DELETE, 0, 0, 0);
-			(void) kevent(ps->kp_fd, &kev, 1, NULL, 0, &ts);
-		    }
-
-		add_fb:
-		    ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_USEFLBCK;
-		    ASSERT(ps->fds_status[fd].used_events);
-		    ps->fds_status[fd].used_events = 0;
-		    erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-		    update_fallback_pollset(ps, fd);
-		    ASSERT(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK);
-		    break;
-		}
-		default:
-		    fatal_error("%s:%d:write_batch_buf(): Internal error",
-				__FILE__, __LINE__);
-		    break;
-		}
-	    }
-	}
-    }
-    bbp->len = 0;
-}
-
-#endif /* ERTS_POLL_USE_KQUEUE */
-
-static ERTS_INLINE void
-batch_update_pollset(ErtsPollSet ps, int fd, ErtsPollBatchBuf *bbp)
-{
-    int buf_len;
-#if ERTS_POLL_USE_DEVPOLL
-    short events;
-    struct pollfd *buf;
-#elif ERTS_POLL_USE_KQUEUE
-    struct kevent *buf;
-#endif
-
-#ifdef ERTS_POLL_DEBUG_PRINT
-    erts_printf("Doing lazy update on fd=%d\n", fd);
-#endif
-
-    if (!need_update(ps, fd))
-	return;
-
-    /* Make sure we have room for at least maximum no of entries
-       per fd */
-    if (bbp->size - bbp->len < 2)
-	write_batch_buf(ps, bbp);
-
-    buf_len = bbp->len;
-    buf = bbp->buf;
-
-    ASSERT(fd < ps->fds_status_len);
-
-#if ERTS_POLL_USE_DEVPOLL
-    events = ERTS_POLL_EV_E2N(ps->fds_status[fd].events);
-    if (!events) {
-	buf[buf_len].events = POLLREMOVE;
-	erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-    }
-    else if (!ps->fds_status[fd].used_events) {
-	buf[buf_len].events = events;
-	erts_smp_atomic_inc_nob(&ps->no_of_user_fds);
-    }
-    else {
-	if ((ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST)
-	    || (ps->fds_status[fd].used_events & ~events)) {
-	    /* Reset or removed events... */
-	    buf[buf_len].fd = fd;
-	    buf[buf_len].events = POLLREMOVE;
-	    buf[buf_len++].revents = 0;
-	}
-	buf[buf_len].events = events;
-    }
-    buf[buf_len].fd = fd;
-    buf[buf_len++].revents = 0;
-
-#elif ERTS_POLL_USE_KQUEUE
-
-    if (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK) {
-	if (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_USEFLBCK)
-	    update_fallback_pollset(ps, fd);
-	else { /* Remove from fallback and try kqueue */
-	    ErtsPollEvents events = ps->fds_status[fd].events;
-	    ps->fds_status[fd].events = (ErtsPollEvents) 0;
-	    update_fallback_pollset(ps, fd);
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-	    if (events) {
-		ps->fds_status[fd].events = events;
-		goto try_kqueue;
-	    }
-	}
-    }
-    else {
-	ErtsPollEvents events, used_events;
-	int mod_w, mod_r;
-    try_kqueue:
-	events = ERTS_POLL_EV_E2N(ps->fds_status[fd].events);
-	used_events = ERTS_POLL_EV_E2N(ps->fds_status[fd].used_events);
-	if (!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST)) {
-	    if (!used_events &&
-		(events & ERTS_POLL_EV_IN) && (events & ERTS_POLL_EV_OUT))
-		goto do_add_rw;
-	    mod_r = ((events & ERTS_POLL_EV_IN)
-		     != (used_events & ERTS_POLL_EV_IN));
-	    mod_w = ((events & ERTS_POLL_EV_OUT)
-		     != (used_events & ERTS_POLL_EV_OUT));
-	    goto do_mod;
-	}
-	else { /* Reset */
-	    if ((events & ERTS_POLL_EV_IN) && (events & ERTS_POLL_EV_OUT)) {
-	    do_add_rw:
-		EV_SET(&buf[buf_len], fd, EVFILT_READ, EV_ADD,
-		       0, 0, (void *) ERTS_POLL_KQ_OP_ADD2_R);
-		buf_len++;
-		EV_SET(&buf[buf_len], fd, EVFILT_WRITE, EV_ADD,
-		       0, 0, (void *) ERTS_POLL_KQ_OP_ADD2_W);
-		buf_len++;
-
-	    }
-	    else {
-		mod_r = 1;
-		mod_w = 1;
-	    do_mod:
-		if (mod_r) {
-		    if (events & ERTS_POLL_EV_IN) {
-			EV_SET(&buf[buf_len], fd, EVFILT_READ, EV_ADD,
-			       0, 0, (void *) ERTS_POLL_KQ_OP_ADD_R);
-			buf_len++;
-		    }
-		    else if (used_events & ERTS_POLL_EV_IN) {
-			EV_SET(&buf[buf_len], fd, EVFILT_READ, EV_DELETE,
-			       0, 0, (void *) ERTS_POLL_KQ_OP_DEL_R);
-			buf_len++;
-		    }
-		}
-		if (mod_w) {
-		    if (events & ERTS_POLL_EV_OUT) {
-			EV_SET(&buf[buf_len], fd, EVFILT_WRITE, EV_ADD,
-			       0, 0, (void *) ERTS_POLL_KQ_OP_ADD_W);
-			buf_len++;
-		    }
-		    else if (used_events & ERTS_POLL_EV_OUT) {
-			EV_SET(&buf[buf_len], fd, EVFILT_WRITE, EV_DELETE,
-			       0, 0, (void *) ERTS_POLL_KQ_OP_DEL_W);
-			buf_len++;
-		    }
-		}
-	    }
-	}
-	if (used_events) {
-	    if (!events) {
-		erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-	    }
-	}
-	else {
-	    if (events)
-		erts_smp_atomic_inc_nob(&ps->no_of_user_fds);
-	}
-	ASSERT((events & ~(ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT)) == 0);
-	ASSERT((used_events & ~(ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT)) == 0);
-    }	    
-
-#endif
-
-    ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_RST;
-    ps->fds_status[fd].used_events = ps->fds_status[fd].events;
-
-    bbp->len = buf_len;
-}
-
-#else /* !ERTS_POLL_USE_BATCH_UPDATE_POLLSET */
-
 #if ERTS_POLL_USE_EPOLL
 static int
-#if ERTS_POLL_USE_CONCURRENT_UPDATE
-conc_update_pollset(ErtsPollSet ps, int fd, int *update_fallback)
-#else
-update_pollset(ErtsPollSet ps, int fd)
-#endif
+update_pollset(ErtsPollSet *ps, int fd, ErtsPollOp op, ErtsPollEvents events)
 {
     int res;
-    int op;
+    int epoll_op = EPOLL_CTL_MOD;
     struct epoll_event epe_templ;
     struct epoll_event epe;
 
-    ASSERT(fd < ps->fds_status_len);
-
-    if (!need_update(ps, fd))
-	return 0;
-
-#ifdef ERTS_POLL_DEBUG_PRINT
-    erts_printf("Doing update on fd=%d\n", fd);
-#endif
-    if (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK) {
-#if ERTS_POLL_USE_CONCURRENT_UPDATE
-	if (!*update_fallback) {
-	    *update_fallback = 1;
-	    return 0;
-	}
-#endif
-	if (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_USEFLBCK) {
-	    return update_fallback_pollset(ps, fd);
-	}
-	else { /* Remove from fallback and try epoll */
-	    ErtsPollEvents events = ps->fds_status[fd].events;
-	    ps->fds_status[fd].events = (ErtsPollEvents) 0;
-	    res = update_fallback_pollset(ps, fd);
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-	    if (!events)
-		return res;
-	    ps->fds_status[fd].events = events;
-	}
-    }
-
-    epe_templ.events = ERTS_POLL_EV_E2N(ps->fds_status[fd].events);
+    epe_templ.events = ERTS_POLL_EV_E2N(events) | EPOLLONESHOT;
     epe_templ.data.fd = fd;
 
 #ifdef VALGRIND
@@ -1247,30 +678,24 @@ update_pollset(ErtsPollSet ps, int fd)
     memset((void *) &epe.data, 0, sizeof(epoll_data_t));
 #endif
 
-    if (epe_templ.events && ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST) {
-	do {
-	    /* We init 'epe' every time since epoll_ctl() may modify it
-	       (not declared const and not documented as const). */
-	    epe.events = epe_templ.events;
-	    epe.data.fd = epe_templ.data.fd;
-	    res = epoll_ctl(ps->kp_fd, EPOLL_CTL_DEL, fd, &epe);
-	} while (res != 0 && errno == EINTR);
-	erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-	ps->fds_status[fd].used_events = 0;
-    }
-
-    if (!epe_templ.events) {
+    switch (op) {
+    case ERTS_POLL_OP_DEL:
 	/* A note on EPOLL_CTL_DEL: linux kernel versions before 2.6.9
 	   need a non-NULL event pointer even though it is ignored... */
-	op = EPOLL_CTL_DEL;
-	erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-    }
-    else if (!ps->fds_status[fd].used_events) {
-	op = EPOLL_CTL_ADD;
-	erts_smp_atomic_inc_nob(&ps->no_of_user_fds);
-    }
-    else {
-	op = EPOLL_CTL_MOD;
+        epoll_op = EPOLL_CTL_DEL;
+        epe_templ.events = 0;
+	erts_atomic_dec_nob(&ps->no_of_user_fds);
+        break;
+    case ERTS_POLL_OP_ADD:
+        epoll_op = EPOLL_CTL_ADD;
+	erts_atomic_inc_nob(&ps->no_of_user_fds);
+        break;
+    case ERTS_POLL_OP_MOD:
+        epoll_op = EPOLL_CTL_MOD;
+        break;
+    default:
+        ASSERT(0);
+        break;
     }
 
     do {
@@ -1278,33 +703,32 @@ update_pollset(ErtsPollSet ps, int fd)
 	   (not declared const and not documented as const). */
 	epe.events = epe_templ.events;
 	epe.data.fd = epe_templ.data.fd;
-	res = epoll_ctl(ps->kp_fd, op, fd, &epe);
+	res = epoll_ctl(ps->kp_fd, epoll_op, fd, &epe);
     } while (res != 0 && errno == EINTR);
 
-#if defined(ERTS_POLL_DEBUG_PRINT) && 1
+#if ERTS_POLL_DEBUG_PRINT
     {
 	int saved_errno = errno;
-	erts_printf("%s = epoll_ctl(%d, %s, %d, {Ox%x, %d})\n",
-		     res == 0 ? "0" : erl_errno_id(errno),
-		     ps->kp_fd,
-		     (op == EPOLL_CTL_ADD
-		      ? "EPOLL_CTL_ADD"
-		      : (op == EPOLL_CTL_MOD
-			 ? "EPOLL_CTL_MOD"
-			 : (op == EPOLL_CTL_DEL
-			    ? "EPOLL_CTL_DEL"
-			    : "UNKNOWN"))),
-		     fd,
-		     epe_templ.events,
-		     fd);
+	DEBUG_PRINT_FD("%s = epoll_ctl(%d, %s, %d, {0x%x, %d})",
+                       ps, fd,
+                       res == 0 ? "0" : erl_errno_id(errno),
+                       ps->kp_fd,
+                       (epoll_op == EPOLL_CTL_ADD
+                        ? "EPOLL_CTL_ADD"
+                        : (epoll_op == EPOLL_CTL_MOD
+                           ? "EPOLL_CTL_MOD"
+                           : (epoll_op == EPOLL_CTL_DEL
+                              ? "EPOLL_CTL_DEL"
+                              : "UNKNOWN"))),
+                       fd,
+                       epe_templ.events,
+                       fd);
 	errno = saved_errno;
     }
 #endif
-    if (res == 0)
-	ps->fds_status[fd].used_events = ps->fds_status[fd].events;
-    else {
+    if (res != 0) {
 	switch (op) {
-	case EPOLL_CTL_MOD:
+	case ERTS_POLL_OP_MOD:
 	    epe.events = 0;
 	    do {
 		/* We init 'epe' every time since epoll_ctl() may modify it
@@ -1313,29 +737,18 @@ update_pollset(ErtsPollSet ps, int fd)
 		epe.data.fd = fd;
 		res = epoll_ctl(ps->kp_fd, EPOLL_CTL_DEL, fd, &epe);
 	    } while (res != 0 && errno == EINTR);
-	    ps->fds_status[fd].used_events = 0;
 	/* Fall through ... */
-	case EPOLL_CTL_ADD: {
-	    ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_USEFLBCK;
-	    erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-#if ERTS_POLL_USE_CONCURRENT_UPDATE
-	    if (!*update_fallback) {
-		*update_fallback = 1;
-		return 0;
-	    }
-#endif
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-	    res = update_fallback_pollset(ps, fd);
-	    ASSERT(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK);
+	case ERTS_POLL_OP_ADD: {
+	    erts_atomic_dec_nob(&ps->no_of_user_fds);
+            res = ERTS_POLL_EV_NVAL;
 	    break;
 	}
-	case EPOLL_CTL_DEL: {
+	case ERTS_POLL_OP_DEL: {
 	    /*
 	     * Since we use a lazy update approach EPOLL_CTL_DEL will
 	     * frequently fail. This since epoll automatically removes
 	     * a filedescriptor that is closed from the poll set.
 	     */
-	    ps->fds_status[fd].used_events = 0;
 	    res = 0;
 	    break;
 	}
@@ -1344,68 +757,282 @@ update_pollset(ErtsPollSet ps, int fd)
 			__FILE__, __LINE__);
 	    break;
 	}
+    } else {
+        res = events;
     }
-    ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_RST;
     return res;
 }
 
-#if ERTS_POLL_USE_CONCURRENT_UPDATE
+#endif /* ERTS_POLL_USE_EPOLL */
+
+#if ERTS_POLL_USE_KQUEUE
+
+/* Some versions of the EV_SET macro used kevp multiple times,
+   so we define out own version that make sure that it is safe
+   to do kevp++ in the argument list. */
+#define ERTS_EV_SET(kevp, a, b, c, f) do {              \
+        struct kevent *kevp_ = kevp;                    \
+        EV_SET(kevp_, a, b, c, 0, 0, f);                \
+    } while(0)
+
 static int
-update_pollset(ErtsPollSet ps, int fd)
+update_pollset(ErtsPollSet *ps, int fd, ErtsPollOp op, ErtsPollEvents events)
 {
-    int update_fallback = 1;
-    return conc_update_pollset(ps, fd, &update_fallback);
-}
-#endif
-
-#endif /* ERTS_POLL_USE_EPOLL */
+    int res = 0, len = 0;
+    struct kevent evts[2];
+    struct timespec ts = {0, 0};
 
-#endif /* ERTS_POLL_USE_BATCH_UPDATE_POLLSET */
+#if defined(EV_DISPATCH) && !defined(__OpenBSD__)
+    /* If we have EV_DISPATCH we use it, unless we are on OpenBSD as the
+       behavior of EV_EOF seems to be edge triggered there and we need it
+       to be level triggered.
+
+       The kevent descriptions for both read and write are added on OP_ADD
+       and removed on OP_DEL. And then after than only EV_ENABLE|EV_DISPATCH
+       are used.
+
+       It could be possible to not modify the pollset when disabling and/or
+       deleting events, but that may cause the poll threads to be awoken
+       a lot more than they should so we take the cost here instead of
+       in the poll thread.
+
+       Note: We need to have EV_DISPATCH both when the event is enabled and
+       disabled, as otherwise the event may be triggered twice on each re-arm.
+       Not sure if this is intended or not (can't find anything about it in the
+       man page), but it seems to be the way it works...
+    */
+
+    if (op == ERTS_POLL_OP_DEL) {
+        erts_atomic_dec_nob(&ps->no_of_user_fds);
+        /* We could probably skip this delete, do we want to? */
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_READ, EV_DELETE, (void *) 0);
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_WRITE, EV_DELETE, (void *) 0);
+    } else if (op == ERTS_POLL_OP_ADD) {
+        uint32_t flags;
+        erts_atomic_inc_nob(&ps->no_of_user_fds);
+
+        flags = EV_ADD|EV_DISPATCH;
+        flags |= ((events & ERTS_POLL_EV_IN) ? 0 : EV_DISABLE);
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_READ, flags, (void *) ERTS_POLL_EV_IN);
+
+        flags = EV_ADD|EV_DISPATCH;
+        flags |= ((events & ERTS_POLL_EV_OUT) ? 0 : EV_DISABLE);
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_WRITE, flags, (void *) ERTS_POLL_EV_OUT);
+    } else {
+        uint32_t flags;
+        ASSERT(op == ERTS_POLL_OP_MOD);
 
-#if ERTS_POLL_USE_POLL || ERTS_POLL_USE_SELECT || ERTS_POLL_USE_FALLBACK
+        flags = EV_DISPATCH;
+        flags |= (events & ERTS_POLL_EV_IN) ? EV_ENABLE : EV_DISABLE;
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_READ, flags, (void *) ERTS_POLL_EV_IN);
 
-#if ERTS_POLL_USE_FALLBACK
-static int update_fallback_pollset(ErtsPollSet ps, int fd)
+        flags = EV_DISPATCH;
+        flags |= (events & ERTS_POLL_EV_OUT) ? EV_ENABLE : EV_DISABLE;
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_WRITE, flags, (void *) ERTS_POLL_EV_OUT);
+    }
 #else
-static int update_pollset(ErtsPollSet ps, int fd)
+    uint32_t flags = EV_ADD|EV_ONESHOT;
+
+    if (op == ERTS_POLL_OP_DEL) {
+        erts_atomic_dec_nob(&ps->no_of_user_fds);
+        /* We don't do anything when a delete is issued. The fds will be removed
+           when they are triggered, or when they are closed. */
+        events = 0;
+    } else if (op == ERTS_POLL_OP_ADD) {
+        erts_atomic_inc_nob(&ps->no_of_user_fds);
+    }
+
+    if (events & ERTS_POLL_EV_IN) {
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_READ, flags, (void *) ERTS_POLL_EV_IN);
+    }
+    if (events & ERTS_POLL_EV_OUT) {
+        ERTS_EV_SET(&evts[len++], fd, EVFILT_WRITE, flags, (void *) ERTS_POLL_EV_OUT);
+    }
+
 #endif
-{
-#ifdef ERTS_POLL_DEBUG_PRINT
-#if ERTS_POLL_USE_FALLBACK
-    erts_printf("Doing fallback update on fd=%d\n", fd);
+    if (len)
+        do {
+            res = kevent(ps->kp_fd, evts, len, NULL, 0, &ts);
+        } while (res < 0 && errno == EINTR);
+#if ERTS_POLL_DEBUG_PRINT
+    {
+        int saved_errno = errno, i;
+        char keventb[255], *keventbp = keventb;
+        if (res < 0)
+            keventbp += sprintf(keventbp,"%s = ",erl_errno_id(saved_errno));
+        else
+            keventbp += sprintf(keventbp,"%d = ",res);
+        keventbp += sprintf(keventbp, "kevent(%d, {",ps->kp_fd);
+        for (i = 0; i < len; i++) {
+            const char *flags = "UNKNOWN";
+            if (evts[i].flags == EV_DELETE) flags = "EV_DELETE";
+            if (evts[i].flags == (EV_ADD|EV_ONESHOT)) flags = "EV_ADD|EV_ONESHOT";
+#ifdef EV_DISPATCH
+            if (evts[i].flags == (EV_ADD|EV_DISPATCH)) flags = "EV_ADD|EV_DISPATCH";
+            if (evts[i].flags == (EV_ADD|EV_DISABLE)) flags = "EV_ADD|EV_DISABLE";
+            if (evts[i].flags == (EV_ENABLE|EV_DISPATCH)) flags = "EV_ENABLE|EV_DISPATCH";
+            if (evts[i].flags == EV_DISABLE) flags = "EV_DISABLE";
+            if (evts[i].flags == (EV_DISABLE|EV_DISPATCH)) flags = "EV_DISABLE|EV_DISABLE";
+#endif
+
+            keventbp += sprintf(keventbp, "%s{%lu, %s, %s}",i > 0 ? ", " : "",
+                                evts[i].ident,
+                                (evts[i].filter == EVFILT_READ
+                                 ? "EVFILT_READ"
+                                 : (evts[i].filter == EVFILT_WRITE
+                                    ? "EVFILT_WRITE"
+                                    : "UNKNOWN")), flags);
+        }
+        keventbp += sprintf(keventbp, "}, %d)", len);
+        DEBUG_PRINT_FD("%s", ps, fd, keventb);
+        errno = saved_errno;
+    }
+#endif
+    if (res < 0) {
+        if (op != ERTS_POLL_OP_DEL) {
+#ifdef EV_RECEIPT
+            struct kevent receipt_evts[2];
+            len = 0;
+            ERTS_EV_SET(&evts[len++], fd, EVFILT_WRITE, EV_DELETE|EV_RECEIPT, (void *) 0);
+            ERTS_EV_SET(&evts[len++], fd, EVFILT_READ, EV_DELETE|EV_RECEIPT, (void *) 0);
+            do {
+                res = kevent(ps->kp_fd, evts, len, receipt_evts, 2, &ts);
+            } while (res < 0 && errno == EINTR);
 #else
-    erts_printf("Doing update on fd=%d\n", fd);
+            ERTS_EV_SET(&evts[0], fd, EVFILT_WRITE, EV_DELETE, (void *) 0);
+            do {
+                res = kevent(ps->kp_fd, evts, 1, NULL, 0, &ts);
+            } while (res < 0 && errno == EINTR);
+            ERTS_EV_SET(&evts[0], fd, EVFILT_READ, EV_DELETE, (void *) 0);
+            do {
+                res = kevent(ps->kp_fd, evts, 1, NULL, 0, &ts);
+            } while (res < 0 && errno == EINTR);
+#endif
+            if (op == ERTS_POLL_OP_ADD)
+                erts_atomic_dec_nob(&ps->no_of_user_fds);
+            events = ERTS_POLL_EV_NVAL;
+        } else
+            events = 0;
+    }
+    return events;
+}
+
+#endif /* ERTS_POLL_USE_KQUEUE */
+
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+
+static ERTS_INLINE void
+init_batch_update(ErtsPollSet *ps, int len)
+{
+#if ERTS_POLL_USE_DEVPOLL
+    ASSERT(ps->poll_fds == NULL);
+    ps->poll_fds = erts_alloc(ERTS_ALC_T_TMP, sizeof(struct pollfd) * len);
+    ps->poll_fds_ix = 0;
 #endif
+}
+
+static ERTS_INLINE void
+write_batch_update(ErtsPollSet *ps)
+{
+#if ERTS_POLL_USE_DEVPOLL
+    ssize_t wres;
+    char *buf = (char *) ps->poll_fds;
+    size_t buf_size = sizeof(struct pollfd)*ps->poll_fds_ix;
+
+    while (1) {
+	wres = write(ps->kp_fd, (void *) buf, buf_size);
+	if (wres < 0) {
+	    if (errno == EINTR)
+		continue;
+	    fatal_error("%s:%d:write_batch_buf(): "
+			"Failed to write to /dev/poll: "
+			"%s (%d)\n",
+			__FILE__, __LINE__,
+			erl_errno_id(errno), errno);
+	}
+#if ERTS_POLL_DEBUG_PRINT
+        {
+            int saved_errno = errno, i;
+            char devpollb[2048], *devpollbp = devpollb;
+            devpollbp += sprintf(devpollbp, "%d = devpoll(%d, {", wres, ps->kp_fd);
+            for (i = 0; i < wres / sizeof(struct pollfd); i++) {
+                if (devpollbp == devpollb)
+                    devpollbp += sprintf(devpollbp, "%d = devpoll(%d, {", wres, ps->kp_fd);
+                devpollbp += sprintf(devpollbp, "%s{fd = %d, events = %s}",
+                                     i > 0 ? ", " : "",
+                                     ps->poll_fds[i].fd,
+                                     ev2str(ps->poll_fds[i].events));
+                if (devpollbp - devpollb > 512) {
+                    devpollbp += sprintf(devpollbp, "}, %d)", ps->poll_fds_ix);
+                    DEBUG_PRINT("%s", ps, devpollb);
+                    devpollbp = devpollb;
+                }
+            }
+            devpollbp += sprintf(devpollbp, "}, %d)", ps->poll_fds_ix);
+            DEBUG_PRINT("%s", ps, devpollb);
+            errno = saved_errno;
+        }
 #endif
 
-    ASSERT(fd < ps->fds_status_len);
-#if ERTS_POLL_USE_FALLBACK
-    ASSERT(ps->fds_status[fd].used_events
-	   ? (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK)
-	   : (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_USEFLBCK));
+	buf_size -= wres;
+	if (buf_size <= 0)
+	    break;
+	buf += wres;
+    }
+
+    if (buf_size < 0) {
+	fatal_error("%s:%d:write_devpoll_buf(): Internal error\n",
+		    __FILE__, __LINE__);
+    }
+    erts_free(ERTS_ALC_T_TMP, ps->poll_fds);
+    ps->poll_fds = NULL;
 #endif
+}
 
-    if (!need_update(ps, fd))
-	return 0;
+static ERTS_INLINE int
+need_update(ErtsPollSet *ps, int fd, int *resetp)
+{
+    int reset;
+    ASSERT(fd < ps->fds_status_len);
 
-#if ERTS_POLL_USE_FALLBACK
+    reset = (int) (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST);
     ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_RST;
-#endif
+
+    *resetp = reset;
+
+    if (reset || ps->fds_status[fd].used_events != ps->fds_status[fd].events)
+	return 1;
+
+    return 0;
+}
+
+static int update_pollset(ErtsPollSet *ps, ErtsPollResFd pr[], int fd)
+{
+    int res = 0, reset = 0;
+    ErtsPollEvents events = ps->fds_status[fd].events;
+    ASSERT(fd < ps->fds_status_len);
+
+    if (!need_update(ps, fd, &reset))
+	return res;
 
 #if ERTS_POLL_USE_POLL	/* --- poll -------------------------------- */
-    if (!ps->fds_status[fd].events) {
+    if (!events) {
 	int pix = ps->fds_status[fd].pix;
 	int last_pix;
+
+        if (reset) {
+            /* When a fd has been reset, we tell the caller of erts_poll_wait
+               this by setting the fd as ERTS_POLL_EV_NONE */
+            ERTS_POLL_RES_SET_FD(&pr[res], fd);
+            ERTS_POLL_RES_SET_EVTS(&pr[res], ERTS_POLL_EV_NONE);
+            DEBUG_PRINT_FD("trig %s (poll)", ps, fd, ev2str(ERTS_POLL_EV_NONE));
+            res++;
+        }
+
 	if (pix < 0) {
-#if ERTS_POLL_USE_FALLBACK
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-#endif
-	    return -1;
+	    return res;
 	}
-#if ERTS_POLL_USE_FALLBACK
-	ASSERT(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK);
-#endif
-	erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
+	erts_atomic_dec_nob(&ps->no_of_user_fds);
 	last_pix = --ps->no_poll_fds;
 	if (pix != last_pix) {
 	/* Move last pix to this pix */
@@ -1421,127 +1048,151 @@ static int update_pollset(ErtsPollSet ps, int fd)
 	/* Clear this fd status */
 	ps->fds_status[fd].pix = -1;
 	ps->fds_status[fd].used_events = (ErtsPollEvents) 0;
-#if ERTS_POLL_USE_FALLBACK
-	ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_INFLBCK;
-#endif
+
     }
     else {
 	int pix = ps->fds_status[fd].pix;
 	if (pix < 0) {
-#if ERTS_POLL_USE_FALLBACK
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK)
-		    || fd == ps->kp_fd);
-#endif
-	    erts_smp_atomic_inc_nob(&ps->no_of_user_fds);
+	    erts_atomic_inc_nob(&ps->no_of_user_fds);
 	    ps->fds_status[fd].pix = pix = ps->no_poll_fds++;
 	    if (pix >= ps->poll_fds_len)
 		grow_poll_fds(ps, pix);
 	    ps->poll_fds[pix].fd = fd;
 	    ps->fds_status[fd].pix = pix;
-#if ERTS_POLL_USE_FALLBACK
-	    ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_INFLBCK;
-#endif
 	}
 
-#if ERTS_POLL_USE_FALLBACK
-	ASSERT(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK);
-#endif
-
 	/* Events to be used in next poll */
-	ps->poll_fds[pix].events = ev2pollev(ps->fds_status[fd].events);
+	ps->poll_fds[pix].events = ev2pollev(events);
 	if (ps->poll_fds[pix].revents) {
 	    /* Remove result events that we should not poll for anymore */
 	    ps->poll_fds[pix].revents
 		&= ev2pollev(~(~ps->fds_status[fd].used_events
-			       & ps->fds_status[fd].events));
+			       & events));
 	}
 	/* Save events to be used in next poll */
-	ps->fds_status[fd].used_events = ps->fds_status[fd].events;
+	ps->fds_status[fd].used_events = events;
     }
-    return 0;
+    return res;
 #elif ERTS_POLL_USE_SELECT	/* --- select ------------------------------ */
-    {
-	ErtsPollEvents events = ps->fds_status[fd].events;
+    if (!events) {
+
+        if (reset) {
+            /* When a fd has been reset, we tell the caller of erts_poll_wait
+               this by setting the fd as ERTS_POLL_EV_NONE */
+            ERTS_POLL_RES_SET_FD(&pr[res], fd);
+            ERTS_POLL_RES_SET_EVTS(&pr[res], ERTS_POLL_EV_NONE);
+            DEBUG_PRINT_FD("trig %s (select)", ps, fd, ev2str(ERTS_POLL_EV_NONE));
+            res++;
+        }
+
+        ERTS_FD_CLR(fd, &ps->input_fds);
+        ERTS_FD_CLR(fd, &ps->output_fds);
+
+        if (ps->fds_status[fd].used_events) {
+            erts_atomic_dec_nob(&ps->no_of_user_fds);
+            ps->fds_status[fd].used_events = (ErtsPollEvents) 0;
+        }
+
+        if (fd == ps->max_fd) {
+            int max = ps->max_fd;
+            for (max = ps->max_fd; max >= 0; max--)
+                if (ps->fds_status[max].used_events)
+                    break;
+            ps->max_fd = max;
+        }
+
+    } else {
+
 	ensure_select_fds(fd, &ps->input_fds, &ps->output_fds);
-	if ((ERTS_POLL_EV_IN & events)
-	    != (ERTS_POLL_EV_IN & ps->fds_status[fd].used_events)) {
-	    if (ERTS_POLL_EV_IN & events) {
-		ERTS_FD_SET(fd, &ps->input_fds);
-	    }
-	    else {
-		ERTS_FD_CLR(fd, &ps->input_fds);
-	    }
-	}
-	if ((ERTS_POLL_EV_OUT & events)
-	    != (ERTS_POLL_EV_OUT & ps->fds_status[fd].used_events)) {
-	    if (ERTS_POLL_EV_OUT & events) {
-		ERTS_FD_SET(fd, &ps->output_fds);
-	    }
-	    else {
-		ERTS_FD_CLR(fd, &ps->output_fds);
-	    }
-	}
 
-	if (!ps->fds_status[fd].used_events) {
-	    ASSERT(events);
-	    erts_smp_atomic_inc_nob(&ps->no_of_user_fds);
-#if ERTS_POLL_USE_FALLBACK
-	    ps->no_select_fds++;
-	    ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_INFLBCK;
-#endif
-	}
-	else if (!events) {
-	    ASSERT(ps->fds_status[fd].used_events);
-	    erts_smp_atomic_dec_nob(&ps->no_of_user_fds);
-	    ps->fds_status[fd].events = events;
-#if ERTS_POLL_USE_FALLBACK
-	    ps->no_select_fds--;
-	    ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_INFLBCK;
-#endif
-	}
+        if (!ps->fds_status[fd].used_events)
+            erts_atomic_inc_nob(&ps->no_of_user_fds);
+
+        if (events & ERTS_POLL_EV_IN)
+            ERTS_FD_SET(fd, &ps->input_fds);
+        else
+            ERTS_FD_CLR(fd, &ps->input_fds);
+
+        if (events & ERTS_POLL_EV_OUT)
+            ERTS_FD_SET(fd, &ps->output_fds);
+        else
+            ERTS_FD_CLR(fd, &ps->output_fds);
 
 	ps->fds_status[fd].used_events = events;
 
-	if (events && fd > ps->max_fd)
-	    ps->max_fd = fd;
-	else if (!events && fd == ps->max_fd) {
-	    int max = ps->max_fd;
-	    for (max = ps->max_fd; max >= 0; max--)
-		if (ps->fds_status[max].used_events)
-		    break;
-	    ps->max_fd = max;
-	}
+        if (fd > ps->max_fd)
+            ps->max_fd = fd;
     }
-    return 0;
-#endif
-}
 
-#endif /* ERTS_POLL_USE_POLL || ERTS_POLL_USE_SELECT || ERTS_POLL_USE_FALLBACK */
+    return res;
+#elif ERTS_POLL_USE_DEVPOLL
 
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+    if (!events) {
 
-static void
-handle_update_requests(ErtsPollSet ps)
-{
-    ErtsPollSetUpdateRequestsBlock *urqbp = &ps->update_requests;
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-    ErtsPollBatchBuf bb;
-    setup_batch_buf(ps, &bb);
+        if (reset) {
+            /* When a fd has been reset, we tell the caller of erts_poll_wait
+               this by setting the fd as ERTS_POLL_EV_NONE */
+            ERTS_POLL_RES_SET_FD(&pr[res], fd);
+            ERTS_POLL_RES_SET_EVTS(&pr[res], ERTS_POLL_EV_NONE);
+            DEBUG_PRINT_FD("trig %s (devpoll)", ps, fd, ev2str(ERTS_POLL_EV_NONE));
+            res++;
+        }
+
+        ps->poll_fds[ps->poll_fds_ix].fd = fd;
+        ps->poll_fds[ps->poll_fds_ix].revents = 0;
+        ps->poll_fds[ps->poll_fds_ix++].events = POLLREMOVE;
+
+        if (ps->fds_status[fd].used_events) {
+            erts_atomic_dec_nob(&ps->no_of_user_fds);
+            ps->fds_status[fd].used_events = 0;
+        }
+
+    } else {
+        if (!ps->fds_status[fd].used_events) {
+            erts_atomic_inc_nob(&ps->no_of_user_fds);
+        }
+        ps->poll_fds[ps->poll_fds_ix].fd = fd;
+        ps->poll_fds[ps->poll_fds_ix].revents = 0;
+        ps->poll_fds[ps->poll_fds_ix++].events = ERTS_POLL_EV_E2N(events);
+        ps->fds_status[fd].used_events = ps->fds_status[fd].events;
+    }
+
+    return res;
 #endif
+}
+
+static int
+handle_update_requests(ErtsPollSet *ps, ErtsPollResFd pr[], int no_fds)
+{
+    int res = 0;
+    ErtsPollSetUpdateRequestsBlock *urqbp = ps->curr_upd_req_block;
 
     while (urqbp) {
 	ErtsPollSetUpdateRequestsBlock *free_urqbp = urqbp;
 	int i;
 	int len = urqbp->len;
+
+        init_batch_update(ps, len);
+
 	for (i = 0; i < len; i++) {
 	    int fd = urqbp->fds[i];
 	    ASSERT(fd < ps->fds_status_len);
-	    ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_INURQ;
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-	    batch_update_pollset(ps, fd, &bb);
-#else
-	    update_pollset(ps, fd);
-#endif
+            ASSERT(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INURQ);
+
+            /* We have run out of PollResFd slots to put results in,
+               so we yield here and return later for more. */
+            if (res == no_fds && pr != NULL) {
+                memmove(urqbp->fds, urqbp->fds+i, sizeof(int) * (len - i));
+                urqbp->len -= i;
+                ps->curr_upd_req_block = urqbp;
+                write_batch_update(ps);
+                return res;
+            }
+
+            if (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INURQ) {
+                ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_INURQ;
+                res += update_pollset(ps, pr + res, fd);
+            }
 	}
 
 	free_urqbp = urqbp;
@@ -1549,12 +1200,9 @@ handle_update_requests(ErtsPollSet ps)
 
 	free_update_requests_block(ps, free_urqbp);
 
-    }
+        write_batch_update(ps);
 
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-    if (bb.len)
-	write_batch_buf(ps, &bb);
-#endif
+    }
 
     ps->curr_upd_req_block = &ps->update_requests;
 
@@ -1563,17 +1211,19 @@ handle_update_requests(ErtsPollSet ps)
 #endif
 
     ERTS_POLLSET_UNSET_HAVE_UPDATE_REQUESTS(ps);
+    return res;
 }
 
-#endif /* ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE */
+#endif /* !ERTS_POLL_USE_CONCURRENT_UPDATE */
 
 static ERTS_INLINE ErtsPollEvents
-poll_control(ErtsPollSet ps, int fd, ErtsPollEvents events, int on, int *do_wake)
+poll_control(ErtsPollSet *ps, int fd, ErtsPollOp op,
+             ErtsPollEvents events, int *do_wake)
 {
     ErtsPollEvents new_events;
 
     if (fd < ps->internal_fd_limit || fd >= max_fds) {
-	if (fd < 0) {
+	if (fd < 0 || fd >= max_fds) {
 	    new_events = ERTS_POLL_EV_ERR;
 	    goto done;
 	}
@@ -1589,124 +1239,49 @@ poll_control(ErtsPollSet ps, int fd, ErtsPollEvents events, int on, int *do_wake
 	    goto done;
 	}
 #endif
-#if ERTS_POLL_USE_TIMERFD
-	if (fd == ps->timer_fd) {
-	    new_events = ERTS_POLL_EV_NVAL;
-	    goto done;
-	}
-#endif
     }
 
+#if ERTS_POLL_USE_CONCURRENT_UPDATE
+
+    new_events = update_pollset(ps, fd, op, events);
+
+#else /* !ERTS_POLL_USE_CONCURRENT_UPDATE */
     if (fd >= ps->fds_status_len)
 	grow_fds_status(ps, fd);
 
     ASSERT(fd < ps->fds_status_len);
 
-    new_events = ps->fds_status[fd].events;
-
-    if (events == 0) {
-	*do_wake = 0;
-	goto done;
-    }
-
-    if (on)
-	new_events |= events;
-    else
-	new_events &= ~events;
-
-    if (new_events == (ErtsPollEvents) 0) {
-#if ERTS_POLL_USE_KERNEL_POLL || defined(ERTS_SMP)
-	ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_RST;
-#endif
-#if ERTS_POLL_USE_FALLBACK
-	ps->fds_status[fd].flags &= ~ERTS_POLL_FD_FLG_USEFLBCK;
-#endif
-    }
-
-    ps->fds_status[fd].events = new_events;
-
-    if (new_events == ps->fds_status[fd].used_events
-#if ERTS_POLL_USE_KERNEL_POLL || defined(ERTS_SMP)
-	&& !(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST)
-#endif
-	) {
-	*do_wake = 0;
-	goto done;
-    }
-
-#if !ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
-    if (update_pollset(ps, fd) != 0)
-	new_events = ERTS_POLL_EV_ERR;
-#else /* ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE */
-
-#if ERTS_POLL_USE_CONCURRENT_UPDATE
-    if (ERTS_POLLSET_IS_POLLED(ps)) {
-	int update_fallback = 0;
-	conc_update_pollset(ps, fd, &update_fallback);
-	if (!update_fallback) {
-	    *do_wake = 0; /* no need to wake kernel poller */
-	    goto done;
-	}
+    if (op == ERTS_POLL_OP_DEL) {
+        ps->fds_status[fd].flags |= ERTS_POLL_FD_FLG_RST;
+        ps->fds_status[fd].events = 0;
+        *do_wake = 1;
+    } else if (op == ERTS_POLL_OP_ADD) {
+        ASSERT(ps->fds_status[fd].events == 0);
+        ps->fds_status[fd].events = events;
+        *do_wake = 1;
+    } else {
+        ASSERT(op == ERTS_POLL_OP_MOD);
+        ps->fds_status[fd].events = events;
+        *do_wake = 1;
     }
-#endif
+    new_events = ps->fds_status[fd].events;
 
     enqueue_update_request(ps, fd);
-	
-#ifdef ERTS_SMP
-    /*
-     * If new events have been added, we need to wake up the
-     * polling thread, but if events have been removed we don't.
-     */
-    if ((new_events && (ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_RST))
-	|| (~ps->fds_status[fd].used_events & new_events))
-	*do_wake = 1;
-#endif /* ERTS_SMP */
-
-#endif /* ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE */
+
+#endif /* !ERTS_POLL_USE_CONCURRENT_UPDATE */
 
  done:
-#ifdef ERTS_POLL_DEBUG_PRINT
-     erts_printf("0x%x = poll_control(ps, %d, 0x%x, %s) do_wake=%d\n",
-		 (int) new_events, fd, (int) events, (on ? "on" : "off"), *do_wake);
-#endif
+    DEBUG_PRINT_FD("%s = %s(%p, %d, %s, %s) do_wake=%d",
+                   ps, fd, ev2str(new_events), __FUNCTION__, ps,
+                   fd, op2str(op), ev2str(events), *do_wake);
     return new_events;
 }
 
-void
-ERTS_POLL_EXPORT(erts_poll_controlv)(ErtsPollSet ps,
-				     ErtsPollControlEntry pcev[],
-				     int len)
-{
-    int i;
-    int do_wake;
-    int final_do_wake = 0;
-
-    ERTS_POLLSET_LOCK(ps);
-
-    for (i = 0; i < len; i++) {
-	do_wake = 0;
-	pcev[i].events = poll_control(ps,
-				      pcev[i].fd,
-				      pcev[i].events,
-				      pcev[i].on,
-				      &do_wake);
-	final_do_wake |= do_wake;
-    }
-
-    ERTS_POLLSET_UNLOCK(ps);
-
-#ifdef ERTS_SMP
-    if (final_do_wake)
-	wake_poller(ps, 0, 0);
-#endif /* ERTS_SMP */
-
-}
-
 ErtsPollEvents
-ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet ps,
+ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet *ps,
 				    ErtsSysFdType fd,
+                                    ErtsPollOp op,
 				    ErtsPollEvents events,
-				    int on,
 				    int* do_wake) /* In: Wake up polling thread */
 				                  /* Out: Poller is woken */
 {
@@ -1714,15 +1289,15 @@ ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet ps,
 
     ERTS_POLLSET_LOCK(ps);
 
-    res = poll_control(ps, fd, events, on, do_wake);
+    res = poll_control(ps, fd, op, events, do_wake);
 
     ERTS_POLLSET_UNLOCK(ps);
 
-#ifdef ERTS_SMP
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     if (*do_wake) {
-	wake_poller(ps, 0, 0);
+	wake_poller(ps, 0);
     }
-#endif /* ERTS_SMP */
+#endif
 
     return res;
 }
@@ -1734,188 +1309,64 @@ ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet ps,
 #if ERTS_POLL_USE_KERNEL_POLL
 
 static ERTS_INLINE int
-save_kp_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res, int chk_fds_res)
+ERTS_POLL_EXPORT(save_result)(ErtsPollSet *ps, ErtsPollResFd pr[], int max_res, int chk_fds_res, int ebadf)
 {
-    int res = 0;
-    int i;
-    int n = chk_fds_res < max_res ? chk_fds_res : max_res;
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE || ERTS_POLL_DEBUG_PRINT || ERTS_POLL_USE_WAKEUP_PIPE
+    int n = chk_fds_res < max_res ? chk_fds_res : max_res, i;
+    int res = n;
 #if ERTS_POLL_USE_WAKEUP_PIPE
     int wake_fd = ps->wake_fds[0];
 #endif
-#if ERTS_POLL_USE_TIMERFD
-    int timer_fd = ps->timer_fd;
-#endif
 
     for (i = 0; i < n; i++) {
-
-#if ERTS_POLL_USE_EPOLL		/* --- epoll ------------------------------- */
-
-	if (ps->res_events[i].events) {
-	    int fd = ps->res_events[i].data.fd;
-	    int ix;
-	    ErtsPollEvents revents;
-#if ERTS_POLL_USE_WAKEUP_PIPE
-	    if (fd == wake_fd) {
-		cleanup_wakeup_pipe(ps);
-		continue;
-	    }
-#endif
-#if ERTS_POLL_USE_TIMERFD
-            if (fd == timer_fd) {
-                continue;
-            }
+        int fd = ERTS_POLL_RES_GET_FD(&pr[i]);
+#ifdef DEBUG_PRINT_MODE
+        ErtsPollEvents evts = ERTS_POLL_RES_GET_EVTS(pr+i);
 #endif
-	    ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-	    /* epoll_wait() can repeat the same fd in result array... */
-	    ix = (int) ps->fds_status[fd].res_ev_ix;
-	    ASSERT(ix >= 0);
-	    if (ix >= res || pr[ix].fd != fd) {
-		ix = res;
-		pr[ix].fd = fd;
-		pr[ix].events = (ErtsPollEvents) 0;
-	    }
-
-	    revents = ERTS_POLL_EV_N2E(ps->res_events[i].events);
-	    pr[ix].events |= revents;
-	    if (revents) {
-		if (res == ix) {
-		    ps->fds_status[fd].res_ev_ix = (unsigned short) ix;
-		    res++;
-		}
-	    }
-	}
-
-#elif ERTS_POLL_USE_KQUEUE	/* --- kqueue ------------------------------ */
 
-	struct kevent *ev;
-	int fd;
-	int ix;
-
-	ev = &ps->res_events[i];
-	fd = (int) ev->ident;
-	ASSERT(fd < ps->fds_status_len);
-	ASSERT(!(ps->fds_status[fd].flags & ERTS_POLL_FD_FLG_INFLBCK));
-	ix = (int) ps->fds_status[fd].res_ev_ix;
-
-	ASSERT(ix >= 0);
-	if (ix >= res || pr[ix].fd != fd) {
-	    ix = res;
-	    pr[ix].fd = (int) ev->ident;
-	    pr[ix].events = (ErtsPollEvents) 0;
-	}
-
-	if (ev->filter == EVFILT_READ) {
-#if ERTS_POLL_USE_WAKEUP_PIPE
-	    if (fd == wake_fd) {
-		cleanup_wakeup_pipe(ps);
-		continue;
-	    }
+        DEBUG_PRINT_FD("trig %s (%s)", ps, fd,
+                       ev2str(evts),
+#if ERTS_POLL_USE_KQUEUE
+                       "kqueue"
+#elif ERTS_POLL_USE_EPOLL
+                       "epoll"
+#else
+                       "/dev/poll"
 #endif
-	    pr[ix].events |= ERTS_POLL_EV_IN;
-	}
-	else if (ev->filter == EVFILT_WRITE)
-	    pr[ix].events |= ERTS_POLL_EV_OUT;
-	if (ev->flags & (EV_ERROR|EV_EOF)) {
-	    if ((ev->flags & EV_ERROR) && (((int) ev->data) == EBADF))
-		pr[ix].events |= ERTS_POLL_EV_NVAL;
-	    else
-		pr[ix].events |= ERTS_POLL_EV_ERR;
-	}
-	if (pr[ix].events) {
-	    if (res == ix) {
-		ps->fds_status[fd].res_ev_ix = (unsigned short) ix;
-		res++;
-	    }
-	}
+            );
 
-#elif ERTS_POLL_USE_DEVPOLL	/* --- devpoll ----------------------------- */
-
-	if (ps->res_events[i].revents) {
-	    int fd = ps->res_events[i].fd;
-	    ErtsPollEvents revents;
 #if ERTS_POLL_USE_WAKEUP_PIPE
-	    if (fd == wake_fd) {
-		cleanup_wakeup_pipe(ps);
-		continue;
-	    }
-#endif
-#if ERTS_POLL_USE_TIMERFD
-	    if (fd == timer_fd) {
-		continue;
-	    }
+        if (fd == wake_fd) {
+            cleanup_wakeup_pipe(ps);
+            ERTS_POLL_RES_SET_EVTS(&pr[i], ERTS_POLL_EV_NONE);
+            if (n == 1)
+                return 0;
+        }
 #endif
-	    revents = ERTS_POLL_EV_N2E(ps->res_events[i].events);
-	    pr[res].fd = fd;
-	    pr[res].events = revents;
-	    res++;
-	}
-
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+        else {
+            /* Reset the events to emulate ONESHOT semantics */
+            ps->fds_status[fd].events = 0;
+            enqueue_update_request(ps, fd);
+        }
 #endif
-
     }
 
     return res;
-}
-
-#endif /* ERTS_POLL_USE_KERNEL_POLL */
-
-#if ERTS_POLL_USE_FALLBACK
-
-static int
-get_kp_results(ErtsPollSet ps, ErtsPollResFd pr[], int max_res)
-{
-    int res;
-#if ERTS_POLL_USE_KQUEUE
-    struct timespec ts = {0, 0};
-#endif
-
-    if (max_res > ps->res_events_len)
-	grow_res_events(ps, max_res);
-
-    do {
-#if ERTS_POLL_USE_EPOLL
-	res = epoll_wait(ps->kp_fd, ps->res_events, max_res, 0);
-#elif ERTS_POLL_USE_KQUEUE
-	res = kevent(ps->kp_fd, NULL, 0, ps->res_events, max_res, &ts);
-#endif
-    } while (res < 0 && errno == EINTR);
-
-    if (res < 0) {
-	fatal_error("%s:%d: %s() failed: %s (%d)\n",
-		    __FILE__, __LINE__,
-#if ERTS_POLL_USE_EPOLL
-		    "epoll_wait",
-#elif ERTS_POLL_USE_KQUEUE
-		    "kevent",
+#else
+    ASSERT(chk_fds_res <= max_res);
+    return chk_fds_res;
 #endif
-		    erl_errno_id(errno), errno);
-    }
-
-    return save_kp_result(ps, pr, max_res, res);
 }
 
-#endif /* ERTS_POLL_USE_FALLBACK */
-
-
+#else /* !ERTS_POLL_USE_KERNEL_POLL */
 
 static ERTS_INLINE int
-save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
-		 int chk_fds_res, int ebadf)
+ERTS_POLL_EXPORT(save_result)(ErtsPollSet *ps, ErtsPollResFd pr[], int max_res, int chk_fds_res, int ebadf)
 {
-#if ERTS_POLL_USE_DEVPOLL
-    return save_kp_result(ps, pr, max_res, chk_fds_res);
-#elif ERTS_POLL_USE_FALLBACK
-    if (!ps->fallback_used)
-	return save_kp_result(ps, pr, max_res, chk_fds_res);
-    else
-#endif /* ERTS_POLL_USE_FALLBACK */
-    {
-
 #if ERTS_POLL_USE_POLL	/* --- poll -------------------------------- */
 	int res = 0;
-#if ERTS_POLL_USE_WAKEUP_PIPE && !ERTS_POLL_USE_FALLBACK 
 	int wake_fd = ps->wake_fds[0];
-#endif
 	int i, first_ix, end_ix;
 
 	/*
@@ -1932,23 +1383,30 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 		if (ps->poll_fds[i].revents != (short) 0) {
 		    int fd = ps->poll_fds[i].fd;
 		    ErtsPollEvents revents;
-#if ERTS_POLL_USE_FALLBACK
-		    if (fd == ps->kp_fd) {
-			res += get_kp_results(ps, &pr[res], max_res-res);
-			i++;
-			continue;
-		    }
-#elif ERTS_POLL_USE_WAKEUP_PIPE
 		    if (fd == wake_fd) {
 			cleanup_wakeup_pipe(ps);
 			i++;
 			continue;
 		    }
-#endif
 		    revents = pollev2ev(ps->poll_fds[i].revents);
-		    pr[res].fd = fd;
-		    pr[res].events = revents;
+		    ERTS_POLL_RES_SET_FD(&pr[res], fd);
+		    ERTS_POLL_RES_SET_EVTS(&pr[res], revents);
+
+                    /* If an fd returns as error, we may want to check the
+                       update_requests queue to see if it has been reset
+                       before delivering the result?!?! This should allow
+                       the user to do driver_dselect + close without waiting
+                       for stop_select... */
+
+                    DEBUG_PRINT_FD("trig %s (poll)", ps, ERTS_POLL_RES_GET_FD(&pr[res]),
+                                   ev2str(ERTS_POLL_RES_GET_EVTS(&pr[res])));
+
 		    res++;
+
+                    /* Clear the events for this fd in order to mimic
+                       how epoll ONESHOT works */
+                    ps->fds_status[fd].events = 0;
+                    enqueue_update_request(ps, fd);
 		}
 		i++;
 	    }
@@ -1964,9 +1422,7 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 
 #elif ERTS_POLL_USE_SELECT	/* --- select ------------------------------ */
 	int res = 0;
-#if ERTS_POLL_USE_WAKEUP_PIPE && !ERTS_POLL_USE_FALLBACK 
 	int wake_fd = ps->wake_fds[0];
-#endif
 	int fd, first_fd, end_fd;
 
 	/*
@@ -1979,29 +1435,23 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 	if (!ebadf) {
 	    while (1) {
 		while (fd < end_fd && res < max_res) {
-
-		    pr[res].events = (ErtsPollEvents) 0;
+                    ErtsPollEvents events = 0;
 		    if (ERTS_FD_ISSET(fd, &ps->res_input_fds)) {
-#if ERTS_POLL_USE_FALLBACK
-			if (fd == ps->kp_fd) {
-			    res += get_kp_results(ps, &pr[res], max_res-res);
-			    fd++;
-			    continue;
-			}
-#elif ERTS_POLL_USE_WAKEUP_PIPE
 			if (fd == wake_fd) {
 			    cleanup_wakeup_pipe(ps);
 			    fd++;
 			    continue;
 			}
-#endif
-			pr[res].events |= ERTS_POLL_EV_IN;
+			events |= ERTS_POLL_EV_IN;
 		    }
 		    if (ERTS_FD_ISSET(fd, &ps->res_output_fds))
-			pr[res].events |= ERTS_POLL_EV_OUT;
-		    if (pr[res].events) {
-			pr[res].fd = fd;
+			events |= ERTS_POLL_EV_OUT;
+		    if (events) {
+                        ERTS_POLL_RES_SET_FD(&pr[res], fd);
+                        ERTS_POLL_RES_SET_EVTS(&pr[res], events);
 			res++;
+                        ps->fds_status[fd].events = 0;
+                        enqueue_update_request(ps, fd);
 		    }
 		    fd++;
 		}
@@ -2034,7 +1484,7 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 			if (ps->fds_status[fd].events & ERTS_POLL_EV_OUT) {
 			    oset = &ps->res_output_fds;
 			    ERTS_FD_ZERO(oset);
-			    ERTS_FD_SET(fd, oset);			
+			    ERTS_FD_SET(fd, oset);
 			}
 			do {
 			    /* Initiate 'tv' each time;
@@ -2043,49 +1493,31 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 			    sres = ERTS_SELECT(ps->max_fd+1, iset, oset, NULL, &tv);
 			} while (sres < 0 && errno == EINTR);
 			if (sres < 0) {
-#if ERTS_POLL_USE_FALLBACK
-			    if (fd == ps->kp_fd) {
-				res += get_kp_results(ps,
-						      &pr[res],
-						      max_res-res);
-				fd++;
-				continue;
-			    }
-#elif ERTS_POLL_USE_WAKEUP_PIPE
 			    if (fd == wake_fd) {
 				cleanup_wakeup_pipe(ps);
 				fd++;
 				continue;
 			    }
-#endif
-			    pr[res].fd = fd;
-			    pr[res].events = ERTS_POLL_EV_NVAL;
+                            ERTS_POLL_RES_SET_FD(&pr[res], fd);
+                            ERTS_POLL_RES_SET_EVTS(&pr[res], ERTS_POLL_EV_NVAL);
 			    res++;
 			}
 			else if (sres > 0) {
-			    pr[res].fd = fd;
+                            ErtsPollEvents events = 0;
+                            ERTS_POLL_RES_SET_FD(&pr[res], fd);
 			    if (iset && ERTS_FD_ISSET(fd, iset)) {
-#if ERTS_POLL_USE_FALLBACK
-				if (fd == ps->kp_fd) {
-				    res += get_kp_results(ps,
-							  &pr[res],
-							  max_res-res);
-				    fd++;
-				    continue;
-				}
-#elif ERTS_POLL_USE_WAKEUP_PIPE
 				if (fd == wake_fd) {
 				    cleanup_wakeup_pipe(ps);
 				    fd++;
 				    continue;
 				}
-#endif
-				pr[res].events |= ERTS_POLL_EV_IN;
+				events |= ERTS_POLL_EV_IN;
 			    }
 			    if (oset && ERTS_FD_ISSET(fd, oset)) {
-				pr[res].events |= ERTS_POLL_EV_OUT;
+				events |= ERTS_POLL_EV_OUT;
 			    }
-			    ASSERT(pr[res].events);
+			    ASSERT(events);
+                            ERTS_POLL_RES_SET_EVTS(&pr[res], events);
 			    res++;
 			}
 		    }
@@ -2100,377 +1532,145 @@ save_poll_result(ErtsPollSet ps, ErtsPollResFd pr[], int max_res,
 	}
 	ps->next_sel_fd = fd;
 	return res;
-#endif
-    }
-}
-
-static ERTS_INLINE ErtsMonotonicTime
-get_timeout(ErtsPollSet ps,
-	    int resolution,
-	    ErtsMonotonicTime timeout_time)
-{
-    ErtsMonotonicTime timeout, save_timeout_time;
-
-    if (timeout_time == ERTS_POLL_NO_TIMEOUT) {
-	save_timeout_time = ERTS_MONOTONIC_TIME_MIN;
-	timeout = 0;
-    }
-    else {
-	ErtsMonotonicTime diff_time, current_time;
-	current_time = erts_get_monotonic_time(NULL);
-	diff_time = timeout_time - current_time;
-	if (diff_time <= 0) {
-	    save_timeout_time = ERTS_MONOTONIC_TIME_MIN;
-	    timeout = 0;
-	}
-	else {
-	    save_timeout_time = current_time;
-	    switch (resolution) {
-	    case 1000:
-		/* Round up to nearest even milli second */
-		timeout = ERTS_MONOTONIC_TO_MSEC(diff_time - 1) + 1;
-		if (timeout > (ErtsMonotonicTime) INT_MAX)
-		    timeout = (ErtsMonotonicTime) INT_MAX;
-		save_timeout_time += ERTS_MSEC_TO_MONOTONIC(timeout);
-		timeout -= ERTS_PREMATURE_TIMEOUT(timeout, 1000);
-		break;
-	    case 1000000:
-		/* Round up to nearest even micro second */
-		timeout = ERTS_MONOTONIC_TO_USEC(diff_time - 1) + 1;
-		save_timeout_time += ERTS_USEC_TO_MONOTONIC(timeout);
-		timeout -= ERTS_PREMATURE_TIMEOUT(timeout, 1000*1000);
-		break;
-	    case 1000000000:
-		/* Round up to nearest even nano second */
-		timeout = ERTS_MONOTONIC_TO_NSEC(diff_time - 1) + 1;
-		save_timeout_time += ERTS_NSEC_TO_MONOTONIC(timeout);
-		timeout -= ERTS_PREMATURE_TIMEOUT(timeout, 1000*1000*1000);
-		break;
-	    default:
-		ERTS_INTERNAL_ERROR("Invalid resolution");
-		timeout = 0;
-		save_timeout_time = 0;
-		break;
-	    }
-	}
-    }
-    set_timeout_time(ps, save_timeout_time);
-    return timeout;
-}
-
-#if ERTS_POLL_USE_SELECT
-
-static ERTS_INLINE int
-get_timeout_timeval(ErtsPollSet ps,
-		    SysTimeval *tvp,
-		    ErtsMonotonicTime timeout_time)
-{
-    ErtsMonotonicTime timeout = get_timeout(ps,
-					    1000*1000,
-					    timeout_time);
-
-    if (!timeout) {
-	tvp->tv_sec = 0;
-	tvp->tv_usec = 0;
-
-	return 0;
-    }
-    else {
-	ErtsMonotonicTime sec = timeout/(1000*1000);
-	tvp->tv_sec = sec;
-	tvp->tv_usec = timeout - sec*(1000*1000);
-
-	ASSERT(tvp->tv_sec >= 0);
-	ASSERT(tvp->tv_usec >= 0);
-	ASSERT(tvp->tv_usec < 1000*1000);
-
-	return !0;
-    }
-
-}
-
-#endif
-
-#if ERTS_POLL_USE_KQUEUE || (ERTS_POLL_USE_POLL && defined(HAVE_PPOLL)) || ERTS_POLL_USE_TIMERFD
-
-static ERTS_INLINE int
-get_timeout_timespec(ErtsPollSet ps,
-		     struct timespec *tsp,
-		     ErtsMonotonicTime timeout_time)
-{
-    ErtsMonotonicTime timeout = get_timeout(ps,
-					    1000*1000*1000,
-					    timeout_time);
-
-    if (!timeout) {
-	tsp->tv_sec = 0;
-	tsp->tv_nsec = 0;
-	return 0;
-    }
-    else {
-	ErtsMonotonicTime sec = timeout/(1000*1000*1000);
-	tsp->tv_sec = sec;
-	tsp->tv_nsec = timeout - sec*(1000*1000*1000);
-
-	ASSERT(tsp->tv_sec >= 0);
-	ASSERT(tsp->tv_nsec >= 0);
-	ASSERT(tsp->tv_nsec < 1000*1000*1000);
-
-	return !0;
-    }
+#endif /* ERTS_POLL_USE_SELECT */
 }
 
-#endif
-
-#if ERTS_POLL_USE_TIMERFD
+#endif /* !ERTS_POLL_USE_KERNEL_POLL */
 
 static ERTS_INLINE int
-get_timeout_itimerspec(ErtsPollSet ps,
-                       struct itimerspec *itsp,
-                       ErtsMonotonicTime timeout_time)
-{
-
-    itsp->it_interval.tv_sec = 0;
-    itsp->it_interval.tv_nsec = 0;
-
-    return get_timeout_timespec(ps, &itsp->it_value, timeout_time);
-}
-
-#endif
-
-static ERTS_INLINE int
-check_fd_events(ErtsPollSet ps, ErtsMonotonicTime timeout_time, int max_res)
+check_fd_events(ErtsPollSet *ps, ErtsPollResFd pr[], int do_wait, int max_res)
 {
     int res;
-    ERTS_MSACC_PUSH_STATE_M();
-    if (erts_smp_atomic_read_nob(&ps->no_of_user_fds) == 0
-	&& timeout_time == ERTS_POLL_NO_TIMEOUT) {
-	/* Nothing to poll and zero timeout; done... */
-	return 0;
-    }
-    else {
-	int timeout;
-#if ERTS_POLL_USE_FALLBACK
-	if (!(ps->fallback_used = ERTS_POLL_NEED_FALLBACK(ps))) {
-
-#if ERTS_POLL_USE_EPOLL		/* --- epoll ------------------------------- */
-	    if (max_res > ps->res_events_len)
-		grow_res_events(ps, max_res);
-#if ERTS_POLL_USE_TIMERFD
-            {
-                struct itimerspec its;
-                timeout = get_timeout_itimerspec(ps, &its, timeout_time);
-                if (timeout) {
-#ifdef ERTS_SMP
-                    erts_thr_progress_prepare_wait(NULL);
-#endif
-                    ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-                    timerfd_set(ps, &its);
-                    res = epoll_wait(ps->kp_fd, ps->res_events, max_res, -1);
-                    res = timerfd_clear(ps, res, max_res);
-                } else {
-                    res = epoll_wait(ps->kp_fd, ps->res_events, max_res, 0);
-                }
-            }
-#else /* !ERTS_POLL_USE_TIMERFD */
-	    timeout = (int) get_timeout(ps, 1000, timeout_time);
-            if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-                ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-            }
-	    res = epoll_wait(ps->kp_fd, ps->res_events, max_res, timeout);
-#endif /* !ERTS_POLL_USE_TIMERFD */
-#elif ERTS_POLL_USE_KQUEUE	/* --- kqueue ------------------------------ */
-	    struct timespec ts;
-	    if (max_res > ps->res_events_len)
-		grow_res_events(ps, max_res);
-	    timeout = get_timeout_timespec(ps, &ts, timeout_time);
-            if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-            }
-	    res = kevent(ps->kp_fd, NULL, 0, ps->res_events, max_res, &ts);
-#endif				/* ----------------------------------------- */
-	}
-	else /* use fallback (i.e. poll() or select()) */
-#endif /* ERTS_POLL_USE_FALLBACK */
-	{
+    int timeout = do_wait ? -1 : 0;
+    DEBUG_PRINT_WAIT("Entering check_fd_events(), do_wait=%d", ps, do_wait);
+    {
+#if ERTS_POLL_USE_EPOLL                /* --- epoll ------------------------------- */
+        res = epoll_wait(ps->kp_fd, pr, max_res, timeout);
+
+#elif ERTS_POLL_USE_KQUEUE     /* --- kqueue ------------------------------ */
+        struct timespec ts = {0, 0};
+        struct timespec *tsp = timeout ? NULL : &ts;
+        res = kevent(ps->kp_fd, NULL, 0, pr, max_res, tsp);
+#elif ERTS_POLL_USE_DEVPOLL	/* --- devpoll ----------------------------- */
+        /*
+         * The ioctl() will fail with EINVAL on Solaris 10 if dp_nfds
+         * is set too high. dp_nfds should not be set greater than
+         * the maximum number of file descriptors in the poll set.
+         */
+        struct dvpoll poll_res;
+        int nfds = (int) erts_atomic_read_nob(&ps->no_of_user_fds) + 1 /* wakeup pipe */;
+        poll_res.dp_nfds = nfds < max_res ? nfds : max_res;
+        poll_res.dp_fds = pr;
+        poll_res.dp_timeout = timeout;
+        res = ioctl(ps->kp_fd, DP_POLL, &poll_res);
 
-#if ERTS_POLL_USE_DEVPOLL	/* --- devpoll ----------------------------- */
-	    /*
-	     * The ioctl() will fail with EINVAL on Solaris 10 if dp_nfds
-	     * is set too high. dp_nfds should not be set greater than
-	     * the maximum number of file descriptors in the poll set.
-	     */
-	    struct dvpoll poll_res;
-	    int nfds = (int) erts_smp_atomic_read_nob(&ps->no_of_user_fds);
-#if ERTS_POLL_USE_WAKEUP_PIPE
-	    nfds++; /* Wakeup pipe */
-#endif
-	    timeout = (int) get_timeout(ps, 1000, timeout_time);
-	    poll_res.dp_nfds = nfds < max_res ? nfds : max_res;
-	    if (poll_res.dp_nfds > ps->res_events_len)
-		grow_res_events(ps, poll_res.dp_nfds);
-	    poll_res.dp_fds = ps->res_events;
-	    if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-                ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-            }
-	    poll_res.dp_timeout = timeout;
-	    res = ioctl(ps->kp_fd, DP_POLL, &poll_res);
-#elif ERTS_POLL_USE_POLL && defined(HAVE_PPOLL)	/* --- ppoll ---------------- */
-            struct timespec ts;
-	    timeout = get_timeout_timespec(ps, &ts, timeout_time);
-            if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-            }
-            res = ppoll(ps->poll_fds, ps->no_poll_fds, &ts, NULL);
 #elif ERTS_POLL_USE_POLL        /* --- poll --------------------------------- */
-	    timeout = (int) get_timeout(ps, 1000, timeout_time);
 
-	    if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-            }
-	    res = poll(ps->poll_fds, ps->no_poll_fds, timeout);
+        res = poll(ps->poll_fds, ps->no_poll_fds, timeout);
+
 #elif ERTS_POLL_USE_SELECT	/* --- select ------------------------------ */
-	    SysTimeval to;
-	    timeout = get_timeout_timeval(ps, &to, timeout_time);
+        SysTimeval tv = {0, 0};
+        SysTimeval *tvp = timeout ? NULL : &tv;
 
-	    ERTS_FD_COPY(&ps->input_fds, &ps->res_input_fds);
-	    ERTS_FD_COPY(&ps->output_fds, &ps->res_output_fds);
+        ERTS_FD_COPY(&ps->input_fds, &ps->res_input_fds);
+        ERTS_FD_COPY(&ps->output_fds, &ps->res_output_fds);
 
-	    if (timeout) {
-#ifdef ERTS_SMP
-		erts_thr_progress_prepare_wait(NULL);
-#endif
-		ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
-	    }
-	    res = ERTS_SELECT(ps->max_fd + 1,
-			      &ps->res_input_fds,
-			      &ps->res_output_fds,
-			      NULL,
-			      &to);
-#ifdef ERTS_SMP
-	    if (timeout) {
-		erts_thr_progress_finalize_wait(NULL);
-		ERTS_MSACC_POP_STATE_M();
-	    }
-	    if (res < 0
-		&& errno == EBADF
-		&& ERTS_POLLSET_HAVE_UPDATE_REQUESTS(ps)) {
-		/*
-		 * This may have happened because another thread deselected
-		 * a fd in our poll set and then closed it, i.e. the driver
-		 * behaved correctly. We wan't to avoid looking for a bad
-		 * fd, that may even not exist anymore. Therefore, handle
-		 * update requests and try again.
-		 *
-		 * We don't know how much of the timeout is left; therfore,
-		 * we use a zero timeout. If no error occur and no events
-		 * have triggered, we fake an EAGAIN error and let the caller
-		 * restart us.
-		 */
-		to.tv_sec = 0;
-		to.tv_usec = 0;
-		ERTS_POLLSET_LOCK(ps);
-		handle_update_requests(ps);
-		ERTS_POLLSET_UNLOCK(ps);
-		res = ERTS_SELECT(ps->max_fd + 1,
-				  &ps->res_input_fds,
-				  &ps->res_output_fds,
-				  NULL,
-				  &to);
-		if (res == 0) {
-		    errno = EAGAIN;
-		    res = -1;
-		}
-	    }
-#endif /* ERTS_SMP */
-	    return res;
+        res = ERTS_SELECT(ps->max_fd + 1,
+                          &ps->res_input_fds,
+                          &ps->res_output_fds,
+                          NULL,
+                          tvp);
 #endif				/* ----------------------------------------- */
-	}
-	if (timeout) {
-#ifdef ERTS_SMP
-	    erts_thr_progress_finalize_wait(NULL);
-#endif
-	    ERTS_MSACC_POP_STATE_M();
-	}
-	return res;
     }
+    DEBUG_PRINT_WAIT("Leaving check_fd_events(), res=%d", ps, res);
+    return res;
 }
 
 int
-ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet ps,
+ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
 				 ErtsPollResFd pr[],
-				 int *len,
-				 ErtsMonotonicTime timeout_time)
+				 int *len)
 {
-    ErtsMonotonicTime to;
-    int res, no_fds;
+    int res, no_fds, used_fds = 0;
     int ebadf = 0;
-#ifdef ERTS_SMP
+    int do_wait;
     int ps_locked = 0;
-#endif
+    ERTS_MSACC_DECLARE_CACHE();
 
     no_fds = *len;
-#ifdef ERTS_POLL_MAX_RES
-    if (no_fds >= ERTS_POLL_MAX_RES)
-	no_fds = ERTS_POLL_MAX_RES;
-#endif
-
     *len = 0;
+    ASSERT(no_fds > 0);
 
-#ifdef ERTS_POLL_DEBUG_PRINT
-    erts_printf("Entering erts_poll_wait(), timeout_time=%bps\n",
-		timeout_time);
-#endif
-
-    if (ERTS_POLLSET_SET_POLLED_CHK(ps)) {
-	res = EINVAL; /* Another thread is in erts_poll_wait()
-			 on this pollset... */
-	goto done;
-    }
-
-    to = (is_woken(ps)
-	  ? ERTS_POLL_NO_TIMEOUT /* Use zero timeout */
-	  : timeout_time);
-
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     if (ERTS_POLLSET_HAVE_UPDATE_REQUESTS(ps)) {
 	ERTS_POLLSET_LOCK(ps);
-	handle_update_requests(ps);
+	used_fds = handle_update_requests(ps, pr, no_fds);
 	ERTS_POLLSET_UNLOCK(ps);
+
+        if (used_fds == no_fds) {
+            *len = used_fds;
+            return 0;
+        }
     }
 #endif
 
+    do_wait = !is_woken(ps) && used_fds == 0;
+
+    DEBUG_PRINT_WAIT("Entering %s(), do_wait=%d", ps, __FUNCTION__, do_wait);
+
+    if (do_wait) {
+        erts_thr_progress_prepare_wait(NULL);
+        ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
+    }
+
     while (1) {
-	res = check_fd_events(ps, to, no_fds);
+	res = check_fd_events(ps, pr + used_fds, do_wait, no_fds - used_fds);
+
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+        if (res < 0
+            && errno == EBADF
+            && ERTS_POLLSET_HAVE_UPDATE_REQUESTS(ps)) {
+            /*
+             * This may have happened because another thread deselected
+             * a fd in our poll set and then closed it, i.e. the driver
+             * behaved correctly. We wan't to avoid looking for a bad
+             * fd, that may even not exist anymore. Therefore, handle
+             * update requests and try again. This behaviour should only
+             * happen when using SELECT as the polling mechanism.
+             */
+            ERTS_POLLSET_LOCK(ps);
+            used_fds += handle_update_requests(ps, pr + used_fds, no_fds - used_fds);
+            if (used_fds == no_fds) {
+                *len = used_fds;
+                ERTS_POLLSET_UNLOCK(ps);
+                return 0;
+            }
+            res = check_fd_events(ps, pr + used_fds, 0, no_fds - used_fds);
+            /* Keep the lock over the non-blocking poll in order to not
+               get any nasty races happening. */
+            ERTS_POLLSET_UNLOCK(ps);
+            if (res == 0) {
+                errno = EAGAIN;
+                res = -1;
+            }
+        }
+#endif
+
 	if (res != 0)
 	    break;
-	if (to == ERTS_POLL_NO_TIMEOUT)
-	    break;
-	if (erts_get_monotonic_time(NULL) >= timeout_time)
-	    break;
+        if (!do_wait)
+            break;
+    }
+
+    if (do_wait) {
+        erts_thr_progress_finalize_wait(NULL);
+        ERTS_MSACC_UPDATE_CACHE();
+        ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_CHECK_IO);
     }
 
     woke_up(ps);
 
-    if (res == 0) {
-	res = ETIMEDOUT;
-    }
-    else if (res < 0) {
+    if (res < 0) {
 #if ERTS_POLL_USE_SELECT
 	if (errno == EBADF) {
 	    ebadf = 1;
@@ -2484,33 +1684,24 @@ ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet ps,
     save_results:
 #endif
 
-#ifdef ERTS_SMP
 	ps_locked = 1;
 	ERTS_POLLSET_LOCK(ps);
-#endif
 
-	no_fds = save_poll_result(ps, pr, no_fds, res, ebadf);
+	used_fds += ERTS_POLL_EXPORT(save_result)(ps, pr + used_fds, no_fds - used_fds, res, ebadf);
 
 #ifdef HARD_DEBUG
-	check_poll_result(pr, no_fds);
+	check_poll_result(pr, used_fds);
 #endif
 
-	res = (no_fds == 0 ? (is_interrupted_reset(ps) ? EINTR : EAGAIN) : 0);
-	*len = no_fds;
+	res = (used_fds == 0 ? (is_interrupted_reset(ps) ? EINTR : EAGAIN) : 0);
+	*len = used_fds;
     }
 
-#ifdef ERTS_SMP
     if (ps_locked)
 	ERTS_POLLSET_UNLOCK(ps);
-    ERTS_POLLSET_UNSET_POLLED(ps);
-#endif
 
- done:
-    set_timeout_time(ps, ERTS_MONOTONIC_TIME_MAX);
-#ifdef ERTS_POLL_DEBUG_PRINT
-    erts_printf("Leaving %s = erts_poll_wait()\n",
-		 res == 0 ? "0" : erl_errno_id(res));
-#endif
+    DEBUG_PRINT_WAIT("Leaving %s = %s(len = %d)", ps,
+                     res == 0 ? "0" : erl_errno_id(res), __FUNCTION__, *len);
 
     return res;
 }
@@ -2520,54 +1711,13 @@ ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet ps,
  */
 
 void
-ERTS_POLL_EXPORT(erts_poll_interrupt)(ErtsPollSet ps, int set)
+ERTS_POLL_EXPORT(erts_poll_interrupt)(ErtsPollSet *ps, int set)
 {
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     if (!set)
 	reset_wakeup_state(ps);
     else
-	wake_poller(ps, 1, 0);
-#endif
-}
-
-#if ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
-void
-ERTS_POLL_EXPORT(erts_poll_async_sig_interrupt)(ErtsPollSet ps)
-{
-    /*
-     * NOTE: This function is called from signal handlers, it,
-     *       therefore, it has to be async-signal safe.
-     */
-    wake_poller(ps, 1, 1);
-}
-#endif
-
-/*
- * erts_poll_interrupt_timed():
- *   If 'set' != 0, interrupt thread blocked in erts_poll_wait() if it
- * is not guaranteed that it will timeout before 'msec' milli seconds.
- */
-void
-ERTS_POLL_EXPORT(erts_poll_interrupt_timed)(ErtsPollSet ps,
-					    int set,
-					    ErtsMonotonicTime timeout_time)
-{
-#if ERTS_POLL_ASYNC_INTERRUPT_SUPPORT || defined(ERTS_SMP)
-    if (!set)
-	reset_wakeup_state(ps);
-    else {
-	ErtsMonotonicTime max_wait_time = get_timeout_time(ps);
-	if (max_wait_time > timeout_time)
-	    wake_poller(ps, 1, 0);
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-	else {
-	    if (ERTS_POLLSET_IS_POLLED(ps))
-		erts_smp_atomic_inc_nob(&ps->no_avoided_wakeups);
-	    erts_smp_atomic_inc_nob(&ps->no_avoided_interrupts);
-	}
-	erts_smp_atomic_inc_nob(&ps->no_interrupt_timed);
-#endif
-    }
+	wake_poller(ps, 1);
 #endif
 }
 
@@ -2581,14 +1731,19 @@ ERTS_POLL_EXPORT(erts_poll_max_fds)(void)
  */
 
 void
-ERTS_POLL_EXPORT(erts_poll_init)(void)
+ERTS_POLL_EXPORT(erts_poll_init)(int *concurrent_updates)
 {
-    erts_smp_mtx_init(&pollsets_lock, "pollsets_lock", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
-    pollsets = NULL;
 
     errno = 0;
 
+    if (concurrent_updates) {
+#if ERTS_POLL_USE_CONCURRENT_UPDATE
+        *concurrent_updates = 1;
+#else
+        *concurrent_updates = 0;
+#endif
+    }
+
 #if !defined(NO_SYSCONF)
     max_fds = sysconf(_SC_OPEN_MAX);
 #elif ERTS_POLL_USE_SELECT
@@ -2607,37 +1762,28 @@ ERTS_POLL_EXPORT(erts_poll_init)(void)
 	fatal_error("erts_poll_init(): Failed to get max number of files: %s\n",
 		    erl_errno_id(errno));
 
-#ifdef ERTS_POLL_DEBUG_PRINT
     print_misc_debug_info();
-#endif
 }
 
-ErtsPollSet
-ERTS_POLL_EXPORT(erts_poll_create_pollset)(void)
+ErtsPollSet *
+ERTS_POLL_EXPORT(erts_poll_create_pollset)(int id)
 {
 #if ERTS_POLL_USE_KERNEL_POLL
     int kp_fd;
 #endif
-    ErtsPollSet ps = erts_alloc(ERTS_ALC_T_POLLSET,
-				sizeof(struct ErtsPollSet_));
+    ErtsPollSet *ps = erts_alloc(ERTS_ALC_T_POLLSET,
+				sizeof(struct ERTS_POLL_EXPORT(erts_pollset)));
+    ps->id = id;
     ps->internal_fd_limit = 0;
-    ps->fds_status = NULL;
-    ps->fds_status_len = 0;
-    erts_smp_atomic_init_nob(&ps->no_of_user_fds, 0);
+    erts_atomic_init_nob(&ps->no_of_user_fds, 0);
 #if ERTS_POLL_USE_KERNEL_POLL
     ps->kp_fd = -1;
 #if ERTS_POLL_USE_EPOLL
     kp_fd = epoll_create(256);
-    ps->res_events_len = 0;
-    ps->res_events = NULL;
 #elif ERTS_POLL_USE_DEVPOLL
     kp_fd = open("/dev/poll", O_RDWR);
-    ps->res_events_len = 0;
-    ps->res_events = NULL;
 #elif ERTS_POLL_USE_KQUEUE
     kp_fd = kqueue();
-    ps->res_events_len = 0;
-    ps->res_events = NULL;
 #endif
     if (kp_fd < 0)
 	fatal_error("erts_poll_create_pollset(): Failed to "
@@ -2651,10 +1797,6 @@ ERTS_POLL_EXPORT(erts_poll_create_pollset)(void)
 		    ": %s (%d)\n",
 		    erl_errno_id(errno), errno);
 #endif /* ERTS_POLL_USE_KERNEL_POLL */
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-    /* res_events is also used as write buffer */
-    grow_res_events(ps, ERTS_POLL_MIN_BATCH_BUF_SIZE);
-#endif
 #if ERTS_POLL_USE_POLL
     ps->next_poll_fds_ix = 0;
     ps->no_poll_fds = 0;
@@ -2663,9 +1805,6 @@ ERTS_POLL_EXPORT(erts_poll_create_pollset)(void)
 #elif ERTS_POLL_USE_SELECT
     ps->next_sel_fd = 0;
     ps->max_fd = -1;
-#if ERTS_POLL_USE_FALLBACK
-    ps->no_select_fds = 0;
-#endif
 #ifdef _DARWIN_UNLIMITED_SELECT
     ps->input_fds.sz = 0;
     ps->input_fds.ptr = NULL;
@@ -2682,133 +1821,66 @@ ERTS_POLL_EXPORT(erts_poll_create_pollset)(void)
     ERTS_FD_ZERO(&ps->res_output_fds);
 #endif
 #endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+    ps->fds_status = NULL;
+    ps->fds_status_len = 0;
     ps->update_requests.next = NULL;
     ps->update_requests.len = 0;
     ps->curr_upd_req_block = &ps->update_requests;
-    erts_smp_atomic32_init_nob(&ps->have_update_requests, 0);
-#endif
-#ifdef ERTS_SMP
-    erts_atomic32_init_nob(&ps->polled, 0);
-    erts_smp_mtx_init(&ps->mtx, "pollset", NIL, ERTS_LOCK_FLAGS_CATEGORY_IO);
-#endif
-#if defined(USE_THREADS) || ERTS_POLL_ASYNC_INTERRUPT_SUPPORT
-    erts_atomic32_init_nob(&ps->wakeup_state, (erts_aint32_t) 0);
-#endif
-#if ERTS_POLL_USE_WAKEUP_PIPE
-    create_wakeup_pipe(ps);
-#endif
-#if ERTS_POLL_USE_TIMERFD
-    create_timerfd(ps);
-#endif
-#if ERTS_POLL_USE_FALLBACK
-    if (kp_fd >= ps->fds_status_len)
-	grow_fds_status(ps, kp_fd);
-    /* Force kernel poll fd into fallback (poll/select) set */
-    ps->fds_status[kp_fd].flags
-	|= ERTS_POLL_FD_FLG_INFLBCK|ERTS_POLL_FD_FLG_USEFLBCK;
-    {
-	int do_wake = 0;
-	ERTS_POLL_EXPORT(erts_poll_control)(ps, kp_fd, ERTS_POLL_EV_IN, 1,
-					    &do_wake);
-    }    
+    erts_atomic32_init_nob(&ps->have_update_requests, 0);
+    erts_mtx_init(&ps->mtx, "pollset", NIL, ERTS_LOCK_FLAGS_CATEGORY_IO);
 #endif
 #if ERTS_POLL_USE_KERNEL_POLL
     if (ps->internal_fd_limit <= kp_fd)
 	ps->internal_fd_limit = kp_fd + 1;
     ps->kp_fd = kp_fd;
 #endif
-    init_timeout_time(ps);
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-    erts_smp_atomic_init_nob(&ps->no_avoided_wakeups, 0);
-    erts_smp_atomic_init_nob(&ps->no_avoided_interrupts, 0);
-    erts_smp_atomic_init_nob(&ps->no_interrupt_timed, 0);
-#endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
-    handle_update_requests(ps);
-#endif
-#if ERTS_POLL_USE_FALLBACK
-    ps->fallback_used = 0;
-#endif
-    erts_smp_atomic_set_nob(&ps->no_of_user_fds, 0); /* Don't count wakeup pipe and fallback fd */
-
-    erts_smp_mtx_lock(&pollsets_lock);
-    ps->next = pollsets;
-    pollsets = ps;
-    erts_smp_mtx_unlock(&pollsets_lock);
-
-    return ps;
-}
-
-void
-ERTS_POLL_EXPORT(erts_poll_destroy_pollset)(ErtsPollSet ps)
-{
-
-    if (ps->fds_status)
-	erts_free(ERTS_ALC_T_FD_STATUS, (void *) ps->fds_status);
-
-#if ERTS_POLL_USE_EPOLL
-    if (ps->kp_fd >= 0)
-	close(ps->kp_fd);
-    if (ps->res_events)
-	erts_free(ERTS_ALC_T_POLL_RES_EVS, (void *) ps->res_events);
-#elif ERTS_POLL_USE_DEVPOLL
-    if (ps->kp_fd >= 0)
-	close(ps->kp_fd);
-    if (ps->res_events)
-	erts_free(ERTS_ALC_T_POLL_RES_EVS, (void *) ps->res_events);
-#elif ERTS_POLL_USE_POLL
-    if (ps->poll_fds)
-	erts_free(ERTS_ALC_T_POLL_FDS, (void *) ps->poll_fds);
-#elif ERTS_POLL_USE_SELECT
-#ifdef _DARWIN_UNLIMITED_SELECT
-    if (ps->input_fds.ptr)
-	erts_free(ERTS_ALC_T_SELECT_FDS, (void *) ps->input_fds.ptr);
-    if (ps->res_input_fds.ptr)
-	erts_free(ERTS_ALC_T_SELECT_FDS, (void *) ps->res_input_fds.ptr);
-    if (ps->output_fds.ptr)
-	erts_free(ERTS_ALC_T_SELECT_FDS, (void *) ps->output_fds.ptr);
-    if (ps->res_output_fds.ptr)
-	erts_free(ERTS_ALC_T_SELECT_FDS, (void *) ps->res_output_fds.ptr);
-#endif
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
+    erts_atomic32_init_nob(&ps->wakeup_state, (erts_aint32_t) 0);
+    create_wakeup_pipe(ps);
+    handle_update_requests(ps, NULL, 0);
+    cleanup_wakeup_pipe(ps);
 #endif
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if ERTS_POLL_USE_KERNEL_POLL && (defined(__DARWIN__) || defined(__APPLE__) && defined(__MACH__))
     {
-	ErtsPollSetUpdateRequestsBlock *urqbp = ps->update_requests.next;
-	while (urqbp) {
-	    ErtsPollSetUpdateRequestsBlock *free_urqbp = urqbp;
-	    urqbp = urqbp->next;
-	    free_update_requests_block(ps, free_urqbp);
-	}
-    }
-#endif
-#ifdef ERTS_SMP
-    erts_smp_mtx_destroy(&ps->mtx);
-#endif
-#if ERTS_POLL_USE_WAKEUP_PIPE
-    if (ps->wake_fds[0] >= 0)
-	close(ps->wake_fds[0]);
-    if (ps->wake_fds[1] >= 0)
-	close(ps->wake_fds[1]);
-#endif
-#if ERTS_POLL_USE_TIMERFD
-    if (ps->timer_fd >= 0)
-        close(ps->timer_fd);
-#endif
+        /*
+         * Using kqueue on OS X is a mess of brokenness...
+         *
+         * On OS X version older than 15.6 (i.e. OS X El Capitan released in July 2015),
+         * a thread waiting in kevent is not woken if an event is inserted into the kqueue
+         * by another thread and the event becomes ready. However if a new call to kevent
+         * is done by the waiting thread, the new event is found.
+         *
+         * So on effected OS X versions we could trigger the wakeup pipe so that
+         * the waiters will be woken and re-issue the kevent. However...
+         *
+         * On OS X version older then 16 (i.e. OS X Sierra released in September 2016),
+         * running the emulator driver_SUITE smp_select testcase consistently causes a
+         * kernel panic. I don't know why or what events that trigger it. But it seems
+         * like updates of the pollset while another thread is sleeping in it Creates
+         * some kind of race that triggers the kernel panic.
+         *
+         * So to deal with this, the erts configure check what OS X version is run
+         * and only enabled kernel poll on OS X 16 or newer. In addition, if someone
+         * attempts to compile Erlang on OS X 16 and then run it on OS X 15, we do the
+         * run-time check below to disallow this.
+         */
+        int major, minor, build;
+        os_version(&major,&minor,&build);
+        if (major < 16) {
+            erts_fprintf(stderr,"BROKEN KQUEUE!\n"
+                         "Erlang has been compiled with kernel-poll support,\n"
+                         "but this OS X version is known to have kernel bugs\n"
+                         "when using kernel-poll. You have two options:\n"
+                         " 1) update to a newer OS X version (OS X Sierra or newer)\n"
+                         " 2) recompile erlang without kernel-poll support\n");
+            erts_exit(1, "");
+        }
+    }
+#endif
+    erts_atomic_set_nob(&ps->no_of_user_fds, 0); /* Don't count wakeup pipe and fallback fd */
 
-    erts_smp_mtx_lock(&pollsets_lock);
-    if (ps == pollsets)
-	pollsets = pollsets->next;
-    else {
-	ErtsPollSet prev_ps;
-	for (prev_ps = pollsets; ps != prev_ps->next; prev_ps = prev_ps->next)
-            ;
-	ASSERT(ps == prev_ps->next);
-	prev_ps->next = ps->next;
-    }
-    erts_smp_mtx_unlock(&pollsets_lock);
-
-    erts_free(ERTS_ALC_T_POLLSET, (void *) ps);
+    return ps;
 }
 
 /*
@@ -2816,24 +1888,18 @@ ERTS_POLL_EXPORT(erts_poll_destroy_pollset)(ErtsPollSet ps)
  */
 
 void
-ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
+ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet *ps, ErtsPollInfo *pip)
 {
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     int pending_updates;
 #endif
     Uint size = 0;
 
     ERTS_POLLSET_LOCK(ps);
 
-    size += sizeof(struct ErtsPollSet_);
+    size += sizeof(struct ERTS_POLL_EXPORT(erts_pollset));
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     size += ps->fds_status_len*sizeof(ErtsFdStatus);
-
-#if ERTS_POLL_USE_EPOLL
-    size += ps->res_events_len*sizeof(struct epoll_event);
-#elif ERTS_POLL_USE_DEVPOLL
-    size += ps->res_events_len*sizeof(struct pollfd);
-#elif ERTS_POLL_USE_KQUEUE
-    size += ps->res_events_len*sizeof(struct kevent);
 #endif
 
 #if ERTS_POLL_USE_POLL
@@ -2845,7 +1911,7 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
 #endif
 #endif
 
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     {
 	ErtsPollSetUpdateRequestsBlock *urqbp = ps->update_requests.next;
 	pending_updates = ps->update_requests.len;
@@ -2857,7 +1923,7 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
     }
 #endif
 
-    pip->primary = 
+    pip->primary =
 #if ERTS_POLL_USE_KQUEUE
 	"kqueue"
 #elif ERTS_POLL_USE_EPOLL
@@ -2871,17 +1937,7 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
 #endif
 	;
 
-    pip->fallback = 
-#if !ERTS_POLL_USE_FALLBACK
-	NULL
-#elif ERTS_POLL_USE_POLL
-	"poll"
-#elif ERTS_POLL_USE_SELECT
-	"select"
-#endif
-	;
-
-    pip->kernel_poll = 
+    pip->kernel_poll =
 #if !ERTS_POLL_USE_KERNEL_POLL
 	NULL
 #elif ERTS_POLL_USE_KQUEUE
@@ -2895,34 +1951,13 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
 
     pip->memory_size = size;
 
-    pip->poll_set_size = (int) erts_smp_atomic_read_nob(&ps->no_of_user_fds);
-#if ERTS_POLL_USE_WAKEUP_PIPE
+    pip->poll_set_size = (int) erts_atomic_read_nob(&ps->no_of_user_fds);
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     pip->poll_set_size++; /* Wakeup pipe */
 #endif
-#if ERTS_POLL_USE_TIMERFD
-    pip->poll_set_size++; /* timerfd */
-#endif
-
-    pip->fallback_poll_set_size =
-#if !ERTS_POLL_USE_FALLBACK
-	0
-#elif ERTS_POLL_USE_POLL
-	ps->no_poll_fds
-#elif ERTS_POLL_USE_SELECT
-	ps->no_select_fds
-#endif
-	;
-
-#if ERTS_POLL_USE_FALLBACK
-    /* If only kp_fd is in fallback poll set we don't use fallback... */
-    if (pip->fallback_poll_set_size == 1)
-	pip->fallback_poll_set_size = 0;
-    else
-	pip->poll_set_size++; /* kp_fd */
-#endif
 
     pip->lazy_updates =
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 	1
 #else
 	0
@@ -2930,21 +1965,13 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
 	;
 
     pip->pending_updates = 
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 	pending_updates
 #else
 	0
 #endif
 	;
 
-    pip->batch_updates = 
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
-	1
-#else
-	0
-#endif
-	;
-
     pip->concurrent_updates =
 #if ERTS_POLL_USE_CONCURRENT_UPDATE
 	1
@@ -2953,13 +1980,23 @@ ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet ps, ErtsPollInfo *pip)
 #endif
 	;
 
-    pip->max_fds = max_fds;
+    pip->is_fallback =
+#if ERTS_POLL_IS_FALLBACK
+        1
+#else
+        0
+#endif
+        ;
 
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-    pip->no_avoided_wakeups = erts_smp_atomic_read_nob(&ps->no_avoided_wakeups);
-    pip->no_avoided_interrupts = erts_smp_atomic_read_nob(&ps->no_avoided_interrupts);
-    pip->no_interrupt_timed = erts_smp_atomic_read_nob(&ps->no_interrupt_timed);
+    pip->batch_updates =
+#if ERTS_POLL_USE_DEVPOLL
+        1
+#else
+        0
 #endif
+        ;
+
+    pip->max_fds = max_fds;
 
     ERTS_POLLSET_UNLOCK(ps);
 
@@ -2995,35 +2032,61 @@ fatal_error(char *format, ...)
     abort();
 }
 
-static void
-fatal_error_async_signal_safe(char *error_str)
+/*
+ * --- Debug -----------------------------------------------------------------
+ */
+
+#if ERTS_POLL_USE_EPOLL
+uint32_t epoll_events(int kp_fd, int fd)
 {
-    if (ERTS_SOMEONE_IS_CRASH_DUMPING || ERTS_GOT_SIGUSR1) {
-	/* See comment above in fatal_error() */
-	return;
+    /* For epoll we read the information about what is selected upon from the proc fs.*/
+    char fname[30];
+    FILE *f;
+    unsigned int pos, flags, mnt_id;
+    int line = 0;
+    sprintf(fname,"/proc/%d/fdinfo/%d",getpid(), kp_fd);
+    f = fopen(fname,"r");
+    if (!f) {
+        fprintf(stderr,"failed to open file %s, errno = %d\n", fname, errno);
+        ASSERT(0);
+        return 0;
     }
-    if (error_str) {
-	int len = 0;
-	while (error_str[len])
-	    len++;
-	if (len) {
-	    /* async signal safe */
-	    erts_silence_warn_unused_result(write(2, error_str, len));
-	}
+    if (fscanf(f,"pos:\t%x\nflags:\t%x", &pos, &flags) != 2) {
+        fprintf(stderr,"failed to parse file %s, errno = %d\n", fname, errno);
+        ASSERT(0);
+        return 0;
     }
-    abort();
+    if (fscanf(f,"\nmnt_id:\t%x\n", &mnt_id));
+    line += 3;
+    while (!feof(f)) {
+        /* tfd:       10 events: 40000019 data:       180000000a */
+        int ev_fd;
+        uint32_t events;
+        uint64_t data;
+        if (fscanf(f,"tfd:%d events:%x data:%llx\n", &ev_fd, &events,
+                   (unsigned long long*)&data) != 3) {
+            fprintf(stderr,"failed to parse file %s on line %d, errno = %d\n", fname,
+                    line,
+                    errno);
+            return 0;
+        }
+        if (fd == ev_fd) {
+            fclose(f);
+            return events;
+        }
+    }
+    fclose(f);
+    return 0;
 }
- 
-/*
- * --- Debug -----------------------------------------------------------------
- */
+#endif
 
 void
-ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet ps,
+ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet *ps,
 						ErtsPollEvents ev[],
 						int len)
 {
     int fd;
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     ERTS_POLLSET_LOCK(ps);
     for (fd = 0; fd < len; fd++) {
 	if (fd >= ps->fds_status_len)
@@ -3031,12 +2094,7 @@ ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet ps,
 	else {
 	    ev[fd] = ps->fds_status[fd].events;
             if (
-#if ERTS_POLL_USE_WAKEUP_PIPE
                 fd == ps->wake_fds[0] || fd == ps->wake_fds[1] ||
-#endif
-#if ERTS_POLL_USE_TIMERFD
-                fd == ps->timer_fd ||
-#endif
 #if ERTS_POLL_USE_KERNEL_POLL
                 fd == ps->kp_fd ||
 #endif
@@ -3045,7 +2103,51 @@ ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet ps,
 	}
     }
     ERTS_POLLSET_UNLOCK(ps);
+#elif ERTS_POLL_USE_EPOLL
 
+    /* For epoll we read the information about what is selected upon from the proc fs.*/
+    char fname[30];
+    FILE *f;
+    unsigned int pos, flags, mnt_id;
+    int line = 0;
+    sprintf(fname,"/proc/%d/fdinfo/%d",getpid(), ps->kp_fd);
+    for (fd = 0; fd < len; fd++)
+        ev[fd] = ERTS_POLL_EV_NONE;
+    f = fopen(fname,"r");
+    if (!f) {
+        fprintf(stderr,"failed to open file %s, errno = %d\n", fname, errno);
+        return;
+    }
+    if (fscanf(f,"pos:\t%x\nflags:\t%x", &pos, &flags) != 2) {
+        fprintf(stderr,"failed to parse file %s, errno = %d\n", fname, errno);
+        ASSERT(0);
+        return;
+    }
+    if (fscanf(f,"\nmnt_id:\t%x\n", &mnt_id));
+    line += 3;
+    while (!feof(f)) {
+        /* tfd:       10 events: 40000019 data:       180000000a */
+        int fd;
+        uint32_t events;
+        uint64_t data;
+        if (fscanf(f,"tfd:%d events:%x data:%llx\n", &fd, &events,
+                   (unsigned long long*)&data) != 3) {
+            fprintf(stderr,"failed to parse file %s on line %d, errno = %d\n",
+                    fname, line, errno);
+            ASSERT(0);
+            return;
+        }
+        data &= 0xFFFFFFFF;
+        ASSERT(fd == data);
+        /* Events are the events that are being monitored, which of course include
+           error and hup events, but we are only interested in IN/OUT events */
+        ev[fd] = (ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT) & ERTS_POLL_EV_N2E(events);
+        line++;
+    }
+#else
+    for (fd = 0; fd < len; fd++)
+        ev[fd] = ERTS_POLL_EV_NONE;
+#endif
 }
 
 #ifdef HARD_DEBUG
@@ -3065,10 +2167,10 @@ check_poll_result(ErtsPollResFd pr[], int len)
 }
 
 
-#if ERTS_POLL_USE_DEVPOLL
+#if ERTS_POLL_USE_DEVPOLL && defined(DEBUG)
 
 static void
-check_poll_status(ErtsPollSet ps)
+check_poll_status(ErtsPollSet *ps)
 {
     int i;
     for (i = 0; i < ps->fds_status_len; i++) {
@@ -3100,34 +2202,24 @@ check_poll_status(ErtsPollSet ps)
 #endif /* ERTS_POLL_USE_DEVPOLL */
 #endif /* HARD_DEBUG */
 
-#ifdef ERTS_POLL_DEBUG_PRINT
 static void
 print_misc_debug_info(void)
 {
-    erts_printf("erts_poll using: %s lazy_updates:%s batch_updates:%s\n",
+#if ERTS_POLL_DEBUG_PRINT
+    erts_printf("erts_poll using: %s lazy_updates:%s\n",
 #if ERTS_POLL_USE_KQUEUE
 		"kqueue"
 #elif ERTS_POLL_USE_EPOLL
 		"epoll"
 #elif ERTS_POLL_USE_DEVPOLL
 		"/dev/poll"
-#endif
-#if ERTS_POLL_USE_FALLBACK
-		"-"
-#endif
-#if ERTS_POLL_USE_POLL
+#elif ERTS_POLL_USE_POLL
 		"poll"
 #elif ERTS_POLL_USE_SELECT
 		"select"
 #endif
 		,
-#if ERTS_POLL_USE_UPDATE_REQUESTS_QUEUE
-		"true"
-#else
-		"false"
-#endif
-		,
-#if ERTS_POLL_USE_BATCH_UPDATE_POLLSET
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
 		"true"
 #else
 		"false"
@@ -3146,29 +2238,20 @@ print_misc_debug_info(void)
 #ifdef FD_SETSIZE
     erts_printf("FD_SETSIZE=%d\n", FD_SETSIZE);
 #endif
-}
-    
 #endif
+}
 
 #ifdef ERTS_ENABLE_LOCK_COUNT
-static void erts_lcnt_enable_pollset_lock_count(ErtsPollSet pollset, int enable) {
+void ERTS_POLL_EXPORT(erts_lcnt_enable_pollset_lock_count)(ErtsPollSet *pollset, int enable)
+{
+#if !ERTS_POLL_USE_CONCURRENT_UPDATE
     if(enable) {
         erts_lcnt_install_new_lock_info(&pollset->mtx.lcnt, "pollset_rm", NIL,
             ERTS_LOCK_TYPE_MUTEX | ERTS_LOCK_FLAGS_CATEGORY_IO);
     } else {
         erts_lcnt_uninstall(&pollset->mtx.lcnt);
     }
-}
-
-void ERTS_POLL_EXPORT(erts_lcnt_update_pollset_locks)(int enable) {
-    ErtsPollSet iterator;
-
-    erts_smp_mtx_lock(&pollsets_lock);
-
-    for(iterator = pollsets; iterator != NULL; iterator = iterator->next) {
-        erts_lcnt_enable_pollset_lock_count(iterator, enable);
-    }
-
-    erts_smp_mtx_unlock(&pollsets_lock);
+#endif
+    return;
 }
 #endif
diff --git a/erts/emulator/sys/common/erl_poll.h b/erts/emulator/sys/common/erl_poll.h
index b3b4d79984..e9a667cac1 100644
--- a/erts/emulator/sys/common/erl_poll.h
+++ b/erts/emulator/sys/common/erl_poll.h
@@ -18,11 +18,31 @@
  * %CopyrightEnd%
  */
 
-/*
- * Description:	Poll interface suitable for ERTS with or without
- *              SMP support.
+/**
+ * @description: Poll interface suitable for ERTS with SMP support.
+ *
+ * @author: 	Rickard Green
+ * @author: 	Lukas Larsson
+ *
+ * This header file exports macros and functions that are used to
+ * react to I/O polling events from file descriptors or wait-able
+ * objects. The API exported is the following:
  *
- * Author: 	Rickard Green
+ * defines:
+ *   ERTS_POLL_EV_NONE - No events have been set. This is not the same as 0.
+ *   ERTS_POLL_EV_IN   - Represent an IN event
+ *   ERTS_POLL_EV_OUT  - Represent an OUT event
+ *   ERTS_POLL_EV_ERR  - Represent an error event
+ *   ERTS_POLL_EV_NVAL - Represent an invalid event
+ *
+ * macro functions:
+ *   ErtsSysFdType ERTS_POLL_RES_GET_FD(ErtsPollResFd *evt);
+ *   void ERTS_POLL_RES_SET_FD(ErtsPollResFd *evt, ErtsSysFdType fd);
+ *   ErtsPollEvents ERTS_POLL_RES_GET_EVTS(ErtsPollResFd *evt)
+ *   void ERTS_POLL_RES_SET_EVTS(ErtsPollResFd *evt, ErtsPollEvents fd);
+ *
+ * functions:
+ *   See erl_poll_api.h
  */
 
 #ifndef ERL_POLL_H__
@@ -32,33 +52,27 @@
 
 #define ERTS_POLL_NO_TIMEOUT ERTS_MONOTONIC_TIME_MIN
 
-#if 0
-#define ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-#endif
-
 #ifdef ERTS_ENABLE_KERNEL_POLL
-#  if defined(ERTS_KERNEL_POLL_VERSION)
-#    define ERTS_POLL_EXPORT(FUNC) FUNC ## _kp
+#  undef ERTS_ENABLE_KERNEL_POLL
+#  define ERTS_ENABLE_KERNEL_POLL 1
+#  if defined(ERTS_NO_KERNEL_POLL_VERSION)
+#    define ERTS_POLL_EXPORT(FUNC) FUNC ## _flbk
+#    undef ERTS_NO_KERNEL_POLL_VERSION
+#    define ERTS_NO_KERNEL_POLL_VERSION 1
+#    define ERTS_KERNEL_POLL_VERSION 0
 #  else
-#    define ERTS_POLL_EXPORT(FUNC) FUNC ## _nkp
-#    undef ERTS_POLL_DISABLE_KERNEL_POLL
-#    define ERTS_POLL_DISABLE_KERNEL_POLL
+#    undef ERTS_KERNEL_POLL_VERSION
+#    define ERTS_KERNEL_POLL_VERSION 1
+#    define ERTS_NO_KERNEL_POLL_VERSION 0
+#    define ERTS_POLL_EXPORT(FUNC) FUNC
 #  endif
 #else
 #    define ERTS_POLL_EXPORT(FUNC) FUNC
-#    undef ERTS_POLL_DISABLE_KERNEL_POLL
-#    define ERTS_POLL_DISABLE_KERNEL_POLL
-#endif
-
-#ifdef ERTS_POLL_DISABLE_KERNEL_POLL
-#  undef HAVE_SYS_EPOLL_H
-#  undef HAVE_SYS_EVENT_H
-#  undef HAVE_SYS_DEVPOLL_H
+#    define ERTS_ENABLE_KERNEL_POLL 0
+#    define ERTS_NO_KERNEL_POLL_VERSION 1
+#    define ERTS_KERNEL_POLL_VERSION 0
 #endif
 
-#undef ERTS_POLL_USE_KERNEL_POLL
-#define ERTS_POLL_USE_KERNEL_POLL 0
-
 #undef ERTS_POLL_USE_KQUEUE
 #define ERTS_POLL_USE_KQUEUE 0
 #undef ERTS_POLL_USE_EPOLL
@@ -70,68 +84,96 @@
 #undef ERTS_POLL_USE_SELECT
 #define ERTS_POLL_USE_SELECT 0
 
-#if defined(HAVE_SYS_EVENT_H)
-#  undef ERTS_POLL_USE_KQUEUE
-#  define ERTS_POLL_USE_KQUEUE 1
-#  undef ERTS_POLL_USE_KERNEL_POLL
-#  define ERTS_POLL_USE_KERNEL_POLL 1
-#elif defined(HAVE_SYS_EPOLL_H)
-#  undef ERTS_POLL_USE_EPOLL
-#  define ERTS_POLL_USE_EPOLL 1
-#  undef ERTS_POLL_USE_KERNEL_POLL
-#  define ERTS_POLL_USE_KERNEL_POLL 1
-#elif defined(HAVE_SYS_DEVPOLL_H)
-#  undef ERTS_POLL_USE_DEVPOLL
-#  define ERTS_POLL_USE_DEVPOLL 1
-#  undef ERTS_POLL_USE_KERNEL_POLL
-#  define ERTS_POLL_USE_KERNEL_POLL 1
+/* Defines which structure that erts_poll_wait should use to wait with
+   and how events should be represented */
+#define ERTS_POLL_USE_EPOLL_EVS 0
+#define ERTS_POLL_USE_KQUEUE_EVS 0
+#define ERTS_POLL_USE_DEVPOLL_EVS 0
+#define ERTS_POLL_USE_POLL_EVS 0
+#define ERTS_POLL_USE_SELECT_EVS 0
+
+#define ERTS_POLL_USE_KERNEL_POLL ERTS_KERNEL_POLL_VERSION
+
+#if ERTS_ENABLE_KERNEL_POLL
+#  if defined(HAVE_SYS_EVENT_H)
+#    undef ERTS_POLL_USE_KQUEUE_EVS
+#    define ERTS_POLL_USE_KQUEUE_EVS 1
+#    undef ERTS_POLL_USE_KQUEUE
+#    define ERTS_POLL_USE_KQUEUE ERTS_KERNEL_POLL_VERSION
+#  elif defined(HAVE_SYS_EPOLL_H)
+#    undef ERTS_POLL_USE_EPOLL_EVS
+#    define ERTS_POLL_USE_EPOLL_EVS 1
+#    undef ERTS_POLL_USE_EPOLL
+#    define ERTS_POLL_USE_EPOLL ERTS_KERNEL_POLL_VERSION
+#  elif defined(HAVE_SYS_DEVPOLL_H)
+#    undef ERTS_POLL_USE_DEVPOLL_EVS
+#    define ERTS_POLL_USE_DEVPOLL_EVS 1
+#    undef ERTS_POLL_USE_DEVPOLL
+#    define ERTS_POLL_USE_DEVPOLL ERTS_KERNEL_POLL_VERSION
+#  else
+#    error "Missing kernel poll implementation of erts_poll()"
+#  endif
 #endif
 
-#define ERTS_POLL_USE_FALLBACK (ERTS_POLL_USE_KQUEUE || ERTS_POLL_USE_EPOLL)
-
-#if !ERTS_POLL_USE_KERNEL_POLL || ERTS_POLL_USE_FALLBACK
+#if ERTS_NO_KERNEL_POLL_VERSION
 #  if defined(ERTS_USE_POLL)
+#    undef ERTS_POLL_USE_POLL_EVS
+#    define ERTS_POLL_USE_POLL_EVS 1
 #    undef ERTS_POLL_USE_POLL
 #    define ERTS_POLL_USE_POLL 1
 #  elif !defined(__WIN32__)
+#    undef ERTS_POLL_USE_SELECT_EVS
+#    define ERTS_POLL_USE_SELECT_EVS 1
 #    undef ERTS_POLL_USE_SELECT
 #    define ERTS_POLL_USE_SELECT 1
 #  endif
 #endif
 
-#define ERTS_POLL_USE_TIMERFD 0
+#define ERTS_POLL_USE_FALLBACK (ERTS_POLL_USE_KQUEUE || ERTS_POLL_USE_EPOLL)
 
 typedef Uint32 ErtsPollEvents;
-#undef ERTS_POLL_EV_E2N
+
+typedef enum {
+    ERTS_POLL_OP_ADD = 0,          /* Add the FD to the pollset */
+    ERTS_POLL_OP_MOD = 1,          /* Modify the FD in the pollset */
+    ERTS_POLL_OP_DEL = 2           /* Delete the FD from the pollset */
+} ErtsPollOp;
+
+#define op2str(op) (op == ERTS_POLL_OP_ADD ? "add" :            \
+                    (op == ERTS_POLL_OP_MOD ? "mod" : "del"))
 
 #if defined(__WIN32__)	/* --- win32  --------------------------------------- */
 
-#define ERTS_POLL_EV_IN   1
-#define ERTS_POLL_EV_OUT  2
-#define ERTS_POLL_EV_ERR  4
-#define ERTS_POLL_EV_NVAL 8
+#define ERTS_POLL_EV_IN    1
+#define ERTS_POLL_EV_OUT   2
+#define ERTS_POLL_EV_ERR   4
+#define ERTS_POLL_EV_NVAL  8
 
-#elif ERTS_POLL_USE_EPOLL	/* --- epoll ------------------------------- */
+#define ERTS_POLL_EV_E2N(EV) 		(EV)
+#define ERTS_POLL_EV_N2E(EV) 		(EV)
 
-#include <sys/epoll.h>
+#elif ERTS_POLL_USE_EPOLL_EVS	/* --- epoll ------------------------------- */
 
-#ifdef HAVE_SYS_TIMERFD_H
-#include <sys/timerfd.h>
-#undef ERTS_POLL_USE_TIMERFD
-#define ERTS_POLL_USE_TIMERFD 1
-#endif
+#include <sys/epoll.h>
 
 #define ERTS_POLL_EV_E2N(EV) \
   ((uint32_t) (EV))
 #define ERTS_POLL_EV_N2E(EV) \
-  ((ErtsPollEvents) (EV))
+  ((ErtsPollEvents) (EV) & ~EPOLLONESHOT)
 
 #define ERTS_POLL_EV_IN			ERTS_POLL_EV_N2E(EPOLLIN)
 #define ERTS_POLL_EV_OUT		ERTS_POLL_EV_N2E(EPOLLOUT)
 #define ERTS_POLL_EV_NVAL		ERTS_POLL_EV_N2E(EPOLLET)
 #define ERTS_POLL_EV_ERR		ERTS_POLL_EV_N2E(EPOLLERR|EPOLLHUP)
 
-#elif ERTS_POLL_USE_DEVPOLL	/* --- devpoll ----------------------------- */
+typedef struct epoll_event ErtsPollResFd;
+
+#define ERTS_POLL_RES_GET_FD(evt) ((ErtsSysFdType)((evt)->data.fd))
+#define ERTS_POLL_RES_SET_FD(evt, ident) (evt)->data.fd = ident
+#define ERTS_POLL_RES_GET_EVTS(evt) ERTS_POLL_EV_N2E((evt)->events)
+#define ERTS_POLL_RES_SET_EVTS(evt, evts) (evt)->events = ERTS_POLL_EV_E2N(evts)
+
+#elif ERTS_POLL_USE_DEVPOLL_EVS	/* --- devpoll ----------------------------- */
 
 #include <sys/devpoll.h>
 
@@ -145,12 +187,37 @@ typedef Uint32 ErtsPollEvents;
 #define ERTS_POLL_EV_NVAL		ERTS_POLL_EV_N2E(POLLNVAL)
 #define ERTS_POLL_EV_ERR		ERTS_POLL_EV_N2E(POLLERR|POLLHUP)
 
-#elif ERTS_POLL_USE_KQUEUE	/* --- kqueue ------------------------------ */
+typedef struct pollfd ErtsPollResFd;
+
+#define ERTS_POLL_RES_GET_FD(evt) ((ErtsSysFdType)((evt)->fd))
+#define ERTS_POLL_RES_SET_FD(evt, ident) (evt)->fd = ident
+#define ERTS_POLL_RES_GET_EVTS(evt) ERTS_POLL_EV_N2E((evt)->revents)
+#define ERTS_POLL_RES_SET_EVTS(evt, evts) (evt)->revents = ERTS_POLL_EV_E2N(evts)
+
+#elif ERTS_POLL_USE_KQUEUE_EVS	/* --- kqueue ------------------------------ */
 /* Kqueue use fallback defines (poll() or select()) */
+
+#include <sys/event.h>
+
+#ifdef ERTS_USE_POLL
+#  undef ERTS_POLL_USE_POLL_EVS
+#  define ERTS_POLL_USE_POLL_EVS   1
+#elif !defined(__WIN32__)
+#  undef ERTS_POLL_USE_SELECT_EVS
+#  define ERTS_POLL_USE_SELECT_EVS 1
 #endif
 
-#if ERTS_POLL_USE_POLL	/* --- poll -------------------------------- */
+typedef struct kevent ErtsPollResFd;
+
+#define ERTS_POLL_RES_GET_FD(evt) ((ErtsSysFdType)((evt)->ident))
+#define ERTS_POLL_RES_SET_FD(evt, fd) (evt)->ident = fd
+#define ERTS_POLL_RES_GET_EVTS(evt) ERTS_POLL_EV_N2E((ErtsPollEvents)(evt)->udata)
+#define ERTS_POLL_RES_SET_EVTS(evt, evts) (evt)->udata = (void*)(UWord)(ERTS_POLL_EV_E2N(evts))
+
+#endif
 
+#if ERTS_POLL_USE_POLL_EVS
+				/* --- poll -------------------------------- */
 #include <poll.h>
 
 #define ERTS_POLL_EV_NKP_E2N(EV) \
@@ -169,7 +236,7 @@ typedef Uint32 ErtsPollEvents;
 #define ERTS_POLL_EV_NKP_NVAL		ERTS_POLL_EV_N2E(POLLNVAL)
 #define ERTS_POLL_EV_NKP_ERR		ERTS_POLL_EV_N2E(POLLERR|POLLHUP)
 
-#elif ERTS_POLL_USE_SELECT	/* --- select ------------------------------ */
+#elif ERTS_POLL_USE_SELECT_EVS	/* --- select ------------------------------ */
 
 #define ERTS_POLL_EV_NKP_E2N(EV) (EV)
 #define ERTS_POLL_EV_NKP_N2E(EV) (EV)
@@ -195,73 +262,65 @@ typedef Uint32 ErtsPollEvents;
 
 #endif
 
-typedef struct ErtsPollSet_ *ErtsPollSet;
+#if !ERTS_ENABLE_KERNEL_POLL
 
-typedef struct {
-    ErtsSysFdType fd;
-    ErtsPollEvents events;
-    int on;
-} ErtsPollControlEntry;
-
-typedef struct {
+typedef struct _ErtsPollResFd {
     ErtsSysFdType fd;
     ErtsPollEvents events;
 } ErtsPollResFd;
 
+#define ERTS_POLL_RES_GET_FD(evt) (evt)->fd
+#define ERTS_POLL_RES_SET_FD(evt, ident) (evt)->fd = (ident)
+#define ERTS_POLL_RES_GET_EVTS(evt) ERTS_POLL_EV_N2E((evt)->events)
+#define ERTS_POLL_RES_SET_EVTS(evt, evts) (evt)->events = ERTS_POLL_EV_E2N(evts)
+
+#endif
+
+#define ERTS_POLL_EV_NONE (UINT_MAX & ~(ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT|ERTS_POLL_EV_NVAL|ERTS_POLL_EV_ERR))
+
+#define ev2str(ev)                                                     \
+    (((ev) == 0 || (ev) == ERTS_POLL_EV_NONE) ? "NONE" :               \
+     ((ev) == ERTS_POLL_EV_IN ? "IN" :                                 \
+      ((ev) == ERTS_POLL_EV_OUT ? "OUT" :                              \
+       ((ev) == (ERTS_POLL_EV_IN|ERTS_POLL_EV_OUT) ? "IN|OUT" :        \
+        ((ev) & ERTS_POLL_EV_ERR ? "ERR" :                             \
+         ((ev) & ERTS_POLL_EV_NVAL ? "NVAL" : "OTHER"))))))
+
+
+typedef struct ERTS_POLL_EXPORT(erts_pollset) ErtsPollSet;
+
 typedef struct {
     char *primary;
-    char *fallback;
     char *kernel_poll;
     Uint memory_size;
-    int poll_set_size;
-    int fallback_poll_set_size;
+    Uint poll_set_size;
     int lazy_updates;
-    int pending_updates;
+    Uint pending_updates;
     int batch_updates;
     int concurrent_updates;
-    int max_fds;
-#ifdef ERTS_POLL_COUNT_AVOIDED_WAKEUPS
-    long no_avoided_wakeups;
-    long no_avoided_interrupts;
-    long no_interrupt_timed;
-#endif
+    int is_fallback;
+    Uint max_fds;
+    Uint active_fds;
+    Uint poll_threads;
 } ErtsPollInfo;
 
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-void		ERTS_POLL_EXPORT(erts_poll_async_sig_interrupt)(ErtsPollSet);
+#if defined(ERTS_POLL_USE_FALLBACK) && ERTS_KERNEL_POLL_VERSION
+#  undef ERTS_POLL_EXPORT
+#  define ERTS_POLL_EXPORT(FUNC) FUNC ## _flbk
+#  include "erl_poll_api.h"
+#  undef ERTS_POLL_EXPORT
+#  define ERTS_POLL_EXPORT(FUNC) FUNC
+#elif !defined(ERTS_POLL_USE_FALLBACK)
+#  define ERTS_POLL_USE_FALLBACK 0
 #endif
-void		ERTS_POLL_EXPORT(erts_poll_interrupt)(ErtsPollSet,
-						      int);
-void		ERTS_POLL_EXPORT(erts_poll_interrupt_timed)(ErtsPollSet,
-							    int,
-							    ErtsMonotonicTime);
-ErtsPollEvents	ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet,
-						    ErtsSysFdType,
-						    ErtsPollEvents,
-						    int on,
-						    int* wake_poller
-						    );
-void		ERTS_POLL_EXPORT(erts_poll_controlv)(ErtsPollSet,
-						     ErtsPollControlEntry [],
-						     int on);
-int		ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet,
-						 ErtsPollResFd [],
-						 int *,
-						 ErtsMonotonicTime);
-int		ERTS_POLL_EXPORT(erts_poll_max_fds)(void);
-void		ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet,
-						 ErtsPollInfo *);
-ErtsPollSet	ERTS_POLL_EXPORT(erts_poll_create_pollset)(void);
-void		ERTS_POLL_EXPORT(erts_poll_destroy_pollset)(ErtsPollSet);
-void		ERTS_POLL_EXPORT(erts_poll_init)(void);
-void		ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet,
-								ErtsPollEvents [],
-								int);
 
-int erts_poll_new_table_len(int old_len, int need_len);
+#include "erl_poll_api.h"
 
-#ifdef ERTS_ENABLE_LOCK_COUNT
-void ERTS_POLL_EXPORT(erts_lcnt_update_pollset_locks)(int enable);
-#endif
+/**
+ * Get the next size of the array that holds the file descriptors.
+ * This function is used in order for the check io array and the
+ * pollset array to be of the same size.
+ */
+int erts_poll_new_table_len(int old_len, int need_len);
 
 #endif /* #ifndef ERL_POLL_H__ */
diff --git a/erts/emulator/sys/common/erl_poll_api.h b/erts/emulator/sys/common/erl_poll_api.h
new file mode 100644
index 0000000000..04beb37d1c
--- /dev/null
+++ b/erts/emulator/sys/common/erl_poll_api.h
@@ -0,0 +1,122 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2006-2016. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+/**
+ * @description: Poll interface functions
+ * @author Lukas Larsson
+ *
+ * The functions in the header are used to interact with the poll
+ * implementation. Iff the kernel-poll implementation needs a fallback
+ * pollset, then all functions are exported twice. Once with a _flbk
+ * suffix and once without any suffix. If no fallback is needed, then
+ * only the non-suffix version is exported.
+ */
+
+/**
+ * Initialize the poll implementation. Has to be called before any other function.
+ * @param[out] concurrent_waiters if not NULL, set to 1 if more then one thread
+ * is allowed to wait in the pollsets at the same time.
+ */
+void ERTS_POLL_EXPORT(erts_poll_init)(int *concurrent_waiters);
+/**
+ * @brief Create a new pollset.
+ * @param id The unique debug id of this pollset.
+ */
+ErtsPollSet *ERTS_POLL_EXPORT(erts_poll_create_pollset)(int id);
+
+/**
+ * Modify the contents of a pollset. This function can be called while one
+ * (or possibly more) thread is waiting in the pollset.
+ *
+ * @param ps the pollset to modify
+ * @param fd the file descriptor to modify
+ * @param op the type of operation to do. Normal usage is ADD,MOD...MOD,DEL.
+ * @param evts the events that we are changing interest to. Ignored if op is DEL.
+ * @param[in] wake_poller if set to 1 any thread waiting in the pollset will be woken.
+ *   This parameter is ignored if the pollset supports concurrent waiters.
+ * @param[out] wake_poller set to 1 if the waiting thread was woken.
+ * @return The events set, or ERTS_POLL_EV_NVAL if it was not possible to add the
+ * fd to the pollset.
+ */
+ErtsPollEvents ERTS_POLL_EXPORT(erts_poll_control)(ErtsPollSet *ps,
+                                                   ErtsSysFdType fd,
+                                                   ErtsPollOp op,
+                                                   ErtsPollEvents evts,
+                                                   int *wake_poller);
+
+/**
+ * Wait for events to be ready in the pollset. If the erts_poll_init call
+ * set concurrent_waiters to 1, then multiple threads are allowed to call
+ * this function at the same time.
+ *
+ * When an event has been triggered on a fd, that event is disabled. To
+ * re-enable it the implementation has to call erts_poll_control again.
+ *
+ * @param ps the pollset to wait for events in
+ * @param res an array of fd results that the ready fds are put in.
+ * @param[in] length the length of the res array
+ * @param[out] length the number of ready events returned in res
+ * @return 0 on success, else the ERRNO of the error that happened.
+ */
+int ERTS_POLL_EXPORT(erts_poll_wait)(ErtsPollSet *ps,
+                                     ErtsPollResFd res[],
+                                     int *length);
+/**
+ * Interrupt the thread waiting in the pollset. This function should be called
+ * with set = 0 before any thread calls erts_poll_wait in order to clear any
+ * interrupts that have happened while the thread was awake.
+ *
+ * This function has no effect on pollsets that support concurrent waiters.
+ *
+ * @param ps the pollset to wake
+ * @param set if 1, interrupt the pollset, if 0 clear the interrupt flag.
+ */
+void ERTS_POLL_EXPORT(erts_poll_interrupt)(ErtsPollSet *ps, int set);
+
+/* Debug functions */
+
+/**
+ * Get the maximum number of fds supported by the pollset
+ */
+int ERTS_POLL_EXPORT(erts_poll_max_fds)(void);
+/**
+ * Get information about the given pollset
+ */
+void ERTS_POLL_EXPORT(erts_poll_info)(ErtsPollSet *ps,
+                                      ErtsPollInfo *info);
+/**
+ * Get information about which events are currently selected.
+ *
+ * The unix fd is used to index into the array, so naturally this function does
+ * not work on windows. If the pollset cannot figure out what the selected
+ * events for a given fd is, it is set to ERTS_POLL_EV_NONE.
+ *
+ * @param ps the pollset to get events from
+ * @param evts an array of which events are selected on.
+ */
+void ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet *ps,
+                                                     ErtsPollEvents evts[],
+                                                     int length);
+
+#ifdef ERTS_ENABLE_LOCK_COUNT
+/**
+ * Enable lock counting of any locks within the pollset.
+ */
+void ERTS_POLL_EXPORT(erts_lcnt_enable_pollset_lock_count)(ErtsPollSet *, int enable);
+#endif
diff --git a/erts/emulator/sys/common/erl_sys_common_misc.c b/erts/emulator/sys/common/erl_sys_common_misc.c
index ba03a11405..41a6fcb7e1 100644
--- a/erts/emulator/sys/common/erl_sys_common_misc.c
+++ b/erts/emulator/sys/common/erl_sys_common_misc.c
@@ -45,14 +45,6 @@
 #endif
 #endif
 
-/*
- * erts_check_io_time is used by the erl_check_io implementation. The
- * global erts_check_io_time variable is declared here since there
- * (often) exist two versions of erl_check_io (kernel-poll and
- * non-kernel-poll), and we dont want two versions of this variable.
- */
-erts_smp_atomic_t erts_check_io_time;
-
 /* Written once and only once */
 
 static int filename_encoding = ERL_FILENAME_UNKNOWN;
@@ -150,7 +142,16 @@ sys_double_to_chars(double fp, char *buffer, size_t buffer_size)
     return sys_double_to_chars_ext(fp, buffer, buffer_size, SYS_DEFAULT_FLOAT_DECIMALS);
 }
 
-/* Convert float to string using fixed point notation.
+
+#if SIZEOF_LONG == 8
+# define round_int64 lround
+#elif SIZEOF_LONG_LONG == 8
+# define round_int64 llround
+#else
+# error "No 64-bit integer type?"
+#endif
+
+/* Convert float to string
  *   decimals must be >= 0
  *   if compact != 0, the trailing 0's will be truncated
  */
@@ -162,80 +163,35 @@ sys_double_to_chars_fast(double f, char *buffer, int buffer_size, int decimals,
     #define FRAC_SIZE            52
     #define EXP_SIZE             11
     #define EXP_MASK             (((Uint64)1 << EXP_SIZE) - 1)
-    #define MAX_DECIMALS         (sizeof(cs_sys_double_pow10) \
-				   / sizeof(cs_sys_double_pow10[0]))
+    #define MAX_DECIMALS         (sizeof(pow10v) / sizeof(pow10v[0]))
     #define FRAC_MASK            (((Uint64)1 << FRAC_SIZE) - 1)
     #define FRAC_MASK2           (((Uint64)1 << (FRAC_SIZE + 1)) - 1)
     #define MAX_FLOAT            ((Uint64)1 << (FRAC_SIZE+1))
 
-    static const double cs_sys_double_pow10[] = {
-        SYS_DOUBLE_RND_CONST / 1e0,
-        SYS_DOUBLE_RND_CONST / 1e1,
-        SYS_DOUBLE_RND_CONST / 1e2,
-        SYS_DOUBLE_RND_CONST / 1e3,
-        SYS_DOUBLE_RND_CONST / 1e4,
-        SYS_DOUBLE_RND_CONST / 1e5,
-        SYS_DOUBLE_RND_CONST / 1e6,
-        SYS_DOUBLE_RND_CONST / 1e7,
-        SYS_DOUBLE_RND_CONST / 1e8,
-        SYS_DOUBLE_RND_CONST / 1e9,
-        SYS_DOUBLE_RND_CONST / 1e10,
-        SYS_DOUBLE_RND_CONST / 1e11,
-        SYS_DOUBLE_RND_CONST / 1e12,
-        SYS_DOUBLE_RND_CONST / 1e13,
-        SYS_DOUBLE_RND_CONST / 1e14,
-        SYS_DOUBLE_RND_CONST / 1e15,
-        SYS_DOUBLE_RND_CONST / 1e16,
-        SYS_DOUBLE_RND_CONST / 1e17,
-        SYS_DOUBLE_RND_CONST / 1e18
+    static const double pow10v[] = {
+        1e0,  1e1,  1e2,  1e3,  1e4,  1e5,  1e6,  1e7,  1e8, 1e9,
+        1e10, 1e11, 1e12, 1e13, 1e14, 1e15, 1e16, 1e17, 1e18
     };
 
-    Uint64 mantissa, int_part, frac_part;
-    int exp;
-    int fbits;
-    int max;
+    double af;
+    Uint64 int_part, frac_part;
     int neg;
-    double fr;
-    union { Uint64 L; double F; } x;
     char *p = buffer;
 
     if (decimals < 0)
         return -1;
 
-    if (f >= 0) {
-        neg = 0;
-        fr  = decimals < MAX_DECIMALS ? (f + cs_sys_double_pow10[decimals]) : f;
-        x.F = fr;
-    } else {
+    if (f < 0) {
         neg = 1;
-        fr  = decimals < MAX_DECIMALS ? (f - cs_sys_double_pow10[decimals]) : f;
-        x.F = -fr;
+        af = -f;
     }
-
-    exp      = (x.L >> FRAC_SIZE) & EXP_MASK;
-    mantissa = x.L & FRAC_MASK;
-
-    if (exp == EXP_MASK) {
-        if (mantissa == 0) {
-            if (neg)
-                *p++ = '-';
-            *p++ = 'i';
-            *p++ = 'n';
-            *p++ = 'f';
-        } else {
-            *p++ = 'n';
-            *p++ = 'a';
-            *p++ = 'n';
-        }
-        *p = '\0';
-        return p - buffer;
+    else {
+        neg = 0;
+        af = f;
     }
 
-    exp      -= EXP_MASK >> 1;
-    mantissa |= ((Uint64)1 << FRAC_SIZE);
-
     /* Don't bother with optimizing too large numbers or too large precision */
-    if (x.F > MAX_FLOAT || decimals >= MAX_DECIMALS) {
+    if (af > MAX_FLOAT || decimals >= MAX_DECIMALS) {
         int len = erts_snprintf(buffer, buffer_size, "%.*f", decimals, f);
         char* p = buffer + len;
         if (len >= buffer_size)
@@ -245,77 +201,64 @@ sys_double_to_chars_fast(double f, char *buffer, int buffer_size, int decimals,
             p = find_first_trailing_zero(p);
         *p = '\0';
         return p - buffer;
-    } else if (exp >= FRAC_SIZE) {
-        int_part  = mantissa << (exp - FRAC_SIZE);
-        frac_part = 0;
-        fbits = FRAC_SIZE;  /* not important as frac_part==0 */
-    } else if (exp >= 0) {
-        fbits = FRAC_SIZE - exp;
-        int_part  = mantissa >> fbits;
-        frac_part = mantissa & (((Uint64)1 << fbits) -1);
-    } else /* if (exp < 0) */ {
-        int_part = 0;
-        frac_part = mantissa;
-        fbits = FRAC_SIZE - exp;
     }
 
-    if (!int_part) {
-        if (neg)
-            *p++ = '-';
-        *p++ = '0';
-    } else {
-        int ret, i, n;
-        while (int_part != 0) {
-            *p++ = (char)((int_part % 10) + '0');
-            int_part /= 10;
-        }
-        if (neg)
-            *p++ = '-';
-        /* Reverse string */
-        ret = p - buffer;
-        for (i = 0, n = ret/2; i < n; i++) {
-            int  j = ret - i - 1;
-            char c = buffer[i];
-            buffer[i] = buffer[j];
-            buffer[j] = c;
-        }
-    }
-
-    if (decimals > 0) {
-        int i;
-        *p++ = '.';
-
-        max = buffer_size - (p - buffer) - 1 /* leave room for trailing '\0' */;
+    if (decimals) {
+        double int_f = floor(af);
+        double frac_f = round((af - int_f) * pow10v[decimals]);
 
-        if (decimals > max)
-            return -1;  /* the number is not large enough to fit in the buffer */
+        int_part = (Uint64)int_f;
+        frac_part = (Uint64)frac_f;
 
-        max = decimals;
+        if (frac_f >= pow10v[decimals]) {
+            /* rounding overflow carry into int_part */
+            int_part++;
+            frac_part = 0;
+        }
 
-        for (i = 0; i < max; i++) {
-            if (frac_part > (ERTS_UINT64_MAX/5)) {
-                frac_part >>= 3;
-                fbits -= 3;
+        do {
+            Uint64 n;
+            if (!frac_part) {
+                do {
+                    *p++ = '0';
+                } while (--decimals);
+                break;
             }
+            n = frac_part / 10;
+            *p++ = (char)((frac_part - n*10) + '0');
+            frac_part = n;
+        } while (--decimals);
 
-            /* Multiply by 10 (5*2) to extract decimal digit as integer part */
-            frac_part *= 5;
-            fbits--;
+        *p++ = '.';
+    }
+    else
+        int_part = (Uint64)round_int64(af);
 
-            if (fbits >= 64) {
-                *p++ = '0';
-            }
-            else {
-                *p++ = (char)((frac_part >> fbits) + '0');
-                frac_part &= ((Uint64)1 << fbits) - 1;
-            }
+    if (!int_part) {
+        *p++ = '0';
+    } else {
+        do {
+            Uint64 n = int_part / 10;
+            *p++ = (char)((int_part - n*10) + '0');
+            int_part = n;
+        } while (int_part);
+    }
+    if (neg)
+        *p++ = '-';
+
+    {/* Reverse string */
+        int i = 0;
+        int j = p - buffer - 1;
+        for ( ; i < j; i++, j--) {
+            char tmp = buffer[i];
+            buffer[i] = buffer[j];
+            buffer[j] = tmp;
         }
-
-        /* Delete trailing zeroes */
-        if (compact)
-            p = find_first_trailing_zero(p);
     }
 
+    /* Delete trailing zeroes */
+    if (compact)
+        p = find_first_trailing_zero(p);
     *p = '\0';
     return p - buffer;
 }
diff --git a/erts/emulator/sys/unix/erl_child_setup.c b/erts/emulator/sys/unix/erl_child_setup.c
index 69fc6c2879..10601529a4 100644
--- a/erts/emulator/sys/unix/erl_child_setup.c
+++ b/erts/emulator/sys/unix/erl_child_setup.c
@@ -56,6 +56,8 @@
 #include <stdio.h>
 #include <stdarg.h>
 #include <sys/wait.h>
+#include <sys/types.h>
+#include <sys/socket.h>
 
 #define WANT_NONBLOCKING
 
@@ -437,6 +439,21 @@ main(int argc, char *argv[])
         exit(1);
     }
 
+    /* Ignore SIGTERM.
+       Some container environments send SIGTERM to all processes
+       when terminating. We don't want erl_child_setup to terminate
+       in these cases as that will prevent beam from properly
+       cleaning up.
+    */
+    sa.sa_handler = SIG_IGN;
+    sigemptyset(&sa.sa_mask);
+    sa.sa_flags = 0;
+
+    if (sigaction(SIGTERM, &sa, 0) == -1) {
+        perror(NULL);
+        exit(1);
+    }
+
     forker_hash_init();
 
     SET_CLOEXEC(uds_fd);
diff --git a/erts/emulator/sys/unix/erl_unix_sys.h b/erts/emulator/sys/unix/erl_unix_sys.h
index b83837a7d2..5bfe5a8e2d 100644
--- a/erts/emulator/sys/unix/erl_unix_sys.h
+++ b/erts/emulator/sys/unix/erl_unix_sys.h
@@ -132,12 +132,8 @@
 /* File descriptors are numbers anc consecutively allocated on Unix */
 #define  ERTS_SYS_CONTINOUS_FD_NUMBERS
 
-#ifndef ERTS_SMP
-#  undef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#  define ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#endif
 
-typedef void *GETENV_STATE;
+void erts_sys_env_init(void);
 
 /*
 ** For the erl_timer_sup module.
@@ -268,7 +264,7 @@ erts_os_monotonic_time(void)
 ERTS_GLB_INLINE void
 erts_os_times(ErtsMonotonicTime *mtimep, ErtsSystemTime *stimep)
 {
-    return (*erts_sys_time_data__.r.o.os_times)(mtimep, stimep);
+    (*erts_sys_time_data__.r.o.os_times)(mtimep, stimep);
 }
 
 #endif /* ERTS_OS_TIMES_INLINE_FUNC_PTR_CALL__ */
@@ -296,6 +292,8 @@ erts_sys_perf_counter()
 
 /*
  * Functions for measuring CPU time
+ *
+ * Note that gethrvtime is time per process and clock_gettime is per thread.
  */
 
 #if (defined(HAVE_GETHRVTIME) || defined(HAVE_CLOCK_GETTIME_CPU_TIME))
@@ -304,15 +302,15 @@ typedef struct timespec SysTimespec;
 
 #if defined(HAVE_GETHRVTIME)
 #define sys_gethrvtime() gethrvtime()
-#define sys_get_proc_cputime(t,tp) (t) = sys_gethrvtime(), \
-                                   (tp).tv_sec = (time_t)((t)/1000000000LL), \
-                                   (tp).tv_nsec = (long)((t)%1000000000LL)
+#define sys_get_cputime(t,tp) (t) = sys_gethrvtime(), \
+        (tp).tv_sec = (time_t)((t)/1000000000LL),     \
+        (tp).tv_nsec = (long)((t)%1000000000LL)
 int sys_start_hrvtime(void);
 int sys_stop_hrvtime(void);
 
 #elif defined(HAVE_CLOCK_GETTIME_CPU_TIME)
 #define sys_clock_gettime(cid,tp) clock_gettime((cid),&(tp))
-#define sys_get_proc_cputime(t,tp) sys_clock_gettime(CLOCK_PROCESS_CPUTIME_ID,(tp))
+#define sys_get_cputime(t,tp) sys_clock_gettime(CLOCK_THREAD_CPUTIME_ID,(tp))
 
 #endif
 #endif
@@ -358,9 +356,7 @@ extern void erts_sys_unix_later_init(void);
 #ifdef NO_FPE_SIGNALS
 
 #define erts_get_current_fp_exception() NULL
-#ifdef ERTS_SMP
 #define erts_thread_init_fp_exception() do{}while(0)
-#endif
 #  define __ERTS_FP_CHECK_INIT(fpexnp) do {} while (0)
 #  define __ERTS_FP_ERROR(fpexnp, f, Action) if (!isfinite(f)) { Action; } else {}
 #  define __ERTS_FP_ERROR_THOROUGH(fpexnp, f, Action) __ERTS_FP_ERROR(fpexnp, f, Action)
@@ -373,9 +369,7 @@ extern void erts_sys_unix_later_init(void);
 #else /* !NO_FPE_SIGNALS */
 
 extern volatile unsigned long *erts_get_current_fp_exception(void);
-#ifdef ERTS_SMP
 extern void erts_thread_init_fp_exception(void);
-#endif
 #  if (defined(__i386__) || defined(__x86_64__)) && defined(__GNUC__)
 #    define erts_fwait(fpexnp,f) \
 	__asm__ __volatile__("fwait" : "=m"(*(fpexnp)) : "m"(f))
@@ -442,10 +436,8 @@ void erts_sys_unblock_fpe(int);
 
 
 /* Threads */
-#ifdef USE_THREADS
 extern int init_async(int);
 extern int exit_async(void);
-#endif
 
 #define ERTS_EXIT_AFTER_DUMP _exit
 
diff --git a/erts/emulator/sys/unix/sys.c b/erts/emulator/sys/unix/sys.c
index c1fa660cf9..189ca083d7 100644
--- a/erts/emulator/sys/unix/sys.c
+++ b/erts/emulator/sys/unix/sys.c
@@ -58,15 +58,10 @@
 #define __DARWIN__ 1
 #endif
 
-#ifdef USE_THREADS
 #include "erl_threads.h"
-#endif
 
 #include "erl_mseg.h"
 
-extern char **environ;
-erts_smp_rwmtx_t environ_rwmtx;
-
 #define MAX_VSIZE 16		/* Max number of entries allowed in an I/O
 				 * vector sock_sendv().
 				 */
@@ -79,7 +74,7 @@ erts_smp_rwmtx_t environ_rwmtx;
 
 #include "erl_check_io.h"
 #include "erl_cpu_topology.h"
-
+#include "erl_osenv.h"
 extern int  driver_interrupt(int, int);
 extern void do_break(void);
 
@@ -94,19 +89,12 @@ extern void erts_sys_init_float(void);
 static int debug_log = 0;
 #endif
 
-#ifdef ERTS_SMP
-static erts_smp_atomic32_t have_prepared_crash_dump;
+static erts_atomic32_t have_prepared_crash_dump;
 #define ERTS_PREPARED_CRASH_DUMP \
-  ((int) erts_smp_atomic32_xchg_nob(&have_prepared_crash_dump, 1))
-#else
-static volatile int have_prepared_crash_dump;
-#define ERTS_PREPARED_CRASH_DUMP \
-  (have_prepared_crash_dump++)
-#endif
+  ((int) erts_atomic32_xchg_nob(&have_prepared_crash_dump, 1))
 
-erts_smp_atomic_t sys_misc_mem_sz;
+erts_atomic_t sys_misc_mem_sz;
 
-#if defined(ERTS_SMP)
 static void smp_sig_notify(int signum);
 static int sig_notify_fds[2] = {-1, -1};
 
@@ -114,7 +102,6 @@ static int sig_notify_fds[2] = {-1, -1};
 static int sig_suspend_fds[2] = {-1, -1};
 #endif
 
-#endif
 
 jmp_buf erts_sys_sigsegv_jmp;
 
@@ -128,38 +115,12 @@ static int max_files = -1;
 /* 
  * a few variables used by the break handler 
  */
-#ifdef ERTS_SMP
-erts_smp_atomic32_t erts_break_requested;
+erts_atomic32_t erts_break_requested;
 #define ERTS_SET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
 #define ERTS_UNSET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
-#else
-volatile int erts_break_requested = 0;
-#define ERTS_SET_BREAK_REQUESTED (erts_break_requested = 1)
-#define ERTS_UNSET_BREAK_REQUESTED (erts_break_requested = 0)
-#endif
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
 
-#ifndef ERTS_SMP
-static Eterm signalstate_sigterm[] = {
-    am_sigint,   /* 0 */
-    am_sighup,   /* 1 */
-    am_sigquit,  /* 2 */
-    am_sigabrt,  /* 3 */
-    am_sigalrm,  /* 4 */
-    am_sigterm,  /* 5 */
-    am_sigusr1,  /* 6 */
-    am_sigusr2,  /* 7 */
-    am_sigchld,  /* 8 */
-    am_sigstop,  /* 9 */
-    am_sigtstp   /* 10 */
-};
-
-volatile Uint erts_signal_state = 0;
-#define ERTS_SET_SIGNAL_STATE(S) (erts_signal_state |= signum_to_signalstate(S))
-#define ERTS_CLEAR_SIGNAL_STATE (erts_signal_state = 0)
-static ERTS_INLINE Uint signum_to_signalstate(int signum);
-#endif
 
 /* set early so the break handler has access to initial mode */
 static struct termios initial_tty_mode;
@@ -167,137 +128,6 @@ static int replace_intr = 0;
 /* assume yes initially, ttsl_init will clear it */
 int using_oldshell = 1;
 
-#ifdef ERTS_ENABLE_KERNEL_POLL
-
-int erts_use_kernel_poll = 0;
-
-struct {
-    int (*select)(ErlDrvPort, ErlDrvEvent, int, int);
-    int (*enif_select)(ErlNifEnv*, ErlNifEvent, enum ErlNifSelectFlags, void*, const ErlNifPid*, Eterm);
-    int (*event)(ErlDrvPort, ErlDrvEvent, ErlDrvEventData);
-    void (*check_io_as_interrupt)(void);
-    void (*check_io_interrupt)(int);
-    void (*check_io_interrupt_tmd)(int, ErtsMonotonicTime);
-    void (*check_io)(int);
-    Uint (*size)(void);
-    Eterm (*info)(void *);
-    int (*check_io_debug)(ErtsCheckIoDebugInfo *);
-} io_func = {0};
-
-
-int
-driver_select(ErlDrvPort port, ErlDrvEvent event, int mode, int on)
-{
-    return (*io_func.select)(port, event, mode, on);
-}
-
-int
-driver_event(ErlDrvPort port, ErlDrvEvent event, ErlDrvEventData event_data)
-{
-    return (*io_func.event)(port, event, event_data);
-}
-
-int enif_select(ErlNifEnv* env, ErlNifEvent event,
-                enum ErlNifSelectFlags flags, void* obj, const ErlNifPid* pid, Eterm ref)
-{
-    return (*io_func.enif_select)(env, event, flags, obj, pid, ref);
-}
-
-
-Eterm erts_check_io_info(void *p)
-{
-    return (*io_func.info)(p);
-}
-
-int
-erts_check_io_debug(ErtsCheckIoDebugInfo *ip)
-{
-    return (*io_func.check_io_debug)(ip);
-}
-
-
-static void
-init_check_io(void)
-{
-    if (erts_use_kernel_poll) {
-	io_func.select			= driver_select_kp;
-        io_func.enif_select		= enif_select_kp;
-	io_func.event			= driver_event_kp;
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-	io_func.check_io_as_interrupt	= erts_check_io_async_sig_interrupt_kp;
-#endif
-	io_func.check_io_interrupt	= erts_check_io_interrupt_kp;
-	io_func.check_io_interrupt_tmd	= erts_check_io_interrupt_timed_kp;
-	io_func.check_io		= erts_check_io_kp;
-	io_func.size			= erts_check_io_size_kp;
-	io_func.info			= erts_check_io_info_kp;
-	io_func.check_io_debug		= erts_check_io_debug_kp;
-	erts_init_check_io_kp();
-	max_files = erts_check_io_max_files_kp();
-    }
-    else {
-	io_func.select			= driver_select_nkp;
-        io_func.enif_select		= enif_select_nkp;
-	io_func.event			= driver_event_nkp;
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-	io_func.check_io_as_interrupt	= erts_check_io_async_sig_interrupt_nkp;
-#endif
-	io_func.check_io_interrupt	= erts_check_io_interrupt_nkp;
-	io_func.check_io_interrupt_tmd	= erts_check_io_interrupt_timed_nkp;
-	io_func.check_io		= erts_check_io_nkp;
-	io_func.size			= erts_check_io_size_nkp;
-	io_func.info			= erts_check_io_info_nkp;
-	io_func.check_io_debug		= erts_check_io_debug_nkp;
-	erts_init_check_io_nkp();
-	max_files = erts_check_io_max_files_nkp();
-    }
-}
-
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#define ERTS_CHK_IO_AS_INTR()	(*io_func.check_io_as_interrupt)()
-#else
-#define ERTS_CHK_IO_AS_INTR()	(*io_func.check_io_interrupt)(1)
-#endif
-#define ERTS_CHK_IO_INTR	(*io_func.check_io_interrupt)
-#define ERTS_CHK_IO_INTR_TMD	(*io_func.check_io_interrupt_tmd)
-#define ERTS_CHK_IO		(*io_func.check_io)
-#define ERTS_CHK_IO_SZ		(*io_func.size)
-
-#else /* !ERTS_ENABLE_KERNEL_POLL */
-
-static void
-init_check_io(void)
-{
-    erts_init_check_io();
-    max_files = erts_check_io_max_files();
-}
-
-#ifdef ERTS_POLL_NEED_ASYNC_INTERRUPT_SUPPORT
-#define ERTS_CHK_IO_AS_INTR()	erts_check_io_async_sig_interrupt()
-#else
-#define ERTS_CHK_IO_AS_INTR()	erts_check_io_interrupt(1)
-#endif
-#define ERTS_CHK_IO_INTR	erts_check_io_interrupt
-#define ERTS_CHK_IO_INTR_TMD	erts_check_io_interrupt_timed
-#define ERTS_CHK_IO		erts_check_io
-#define ERTS_CHK_IO_SZ		erts_check_io_size
-
-#endif
-
-void
-erts_sys_schedule_interrupt(int set)
-{
-    ERTS_CHK_IO_INTR(set);
-}
-
-#ifdef ERTS_SMP
-void
-erts_sys_schedule_interrupt_timed(int set, ErtsMonotonicTime timeout_time)
-{
-    ERTS_CHK_IO_INTR_TMD(set, timeout_time);
-}
-#endif
-
 UWord
 erts_sys_get_page_size(void)
 {
@@ -313,8 +143,8 @@ erts_sys_get_page_size(void)
 Uint
 erts_sys_misc_mem_sz(void)
 {
-    Uint res = ERTS_CHK_IO_SZ();
-    res += erts_smp_atomic_read_mb(&sys_misc_mem_sz);
+    Uint res = erts_check_io_size();
+    res += erts_atomic_read_mb(&sys_misc_mem_sz);
     return res;
 }
 
@@ -339,7 +169,6 @@ MALLOC_USE_HASH(1);
 #endif
 #endif
 
-#ifdef USE_THREADS
 
 #ifdef ERTS_THR_HAVE_SIG_FUNCS
 
@@ -418,19 +247,15 @@ thr_create_prepare_child(void *vtcdp)
     erts_sched_bind_atthrcreate_child(tcdp->sched_bind_data);
 }
 
-#endif /* #ifdef USE_THREADS */
 
 void
 erts_sys_pre_init(void)
 {
-#ifdef USE_THREADS
     erts_thr_init_data_t eid = ERTS_THR_INIT_DATA_DEF_INITER;
-#endif
 
     erts_printf_add_cr_to_stdout = 1;
     erts_printf_add_cr_to_stderr = 1;
 
-#ifdef USE_THREADS
 
     eid.thread_create_child_func = thr_create_prepare_child;
     /* Before creation in parent */
@@ -452,23 +277,15 @@ erts_sys_pre_init(void)
     erts_lc_init();
 #endif
 
-#endif /* USE_THREADS */
 
     erts_init_sys_time_sup();
 
-#ifdef USE_THREADS
 
-#ifdef ERTS_SMP
-    erts_smp_atomic32_init_nob(&erts_break_requested, 0);
-    erts_smp_atomic32_init_nob(&have_prepared_crash_dump, 0);
-#else
-    erts_break_requested = 0;
-    have_prepared_crash_dump = 0;
-#endif
+    erts_atomic32_init_nob(&erts_break_requested, 0);
+    erts_atomic32_init_nob(&have_prepared_crash_dump, 0);
 
-#endif /* USE_THREADS */
 
-    erts_smp_atomic_init_nob(&sys_misc_mem_sz, 0);
+    erts_atomic_init_nob(&sys_misc_mem_sz, 0);
 
     {
       /*
@@ -531,10 +348,8 @@ SIGFUNC sys_signal(int sig, SIGFUNC func)
     return(oact.sa_handler);
 }
 
-#ifdef USE_THREADS
 #undef  sigprocmask
 #define sigprocmask erts_thr_sigmask
-#endif
 
 void sys_sigblock(int sig)
 {
@@ -636,10 +451,10 @@ prepare_crash_dump(int secs)
     close(crashdump_companion_cube_fd);
 
     envsz = sizeof(env);
-    i = erts_sys_getenv__("ERL_CRASH_DUMP_NICE", env, &envsz);
+    i = erts_sys_explicit_8bit_getenv("ERL_CRASH_DUMP_NICE", env, &envsz);
     if (i >= 0) {
 	int nice_val;
-	nice_val = i != 0 ? 0 : atoi(env);
+	nice_val = i != 1 ? 0 : atoi(env);
 	if (nice_val > 39) {
 	    nice_val = 39;
 	}
@@ -672,7 +487,7 @@ static void signal_notify_requested(Eterm type) {
         erts_queue_message(p, locks, msgp, msg, am_system);
 
         if (locks)
-            erts_smp_proc_unlock(p, locks);
+            erts_proc_unlock(p, locks);
         erts_proc_dec_refc(p);
     }
 }
@@ -685,23 +500,17 @@ break_requested(void)
    * just set a flag - checked for and handled by
    * scheduler threads erts_check_io() (not signal handler).
    */
-#ifdef DEBUG			
-  fprintf(stderr,"break!\n");
-#endif
   if (ERTS_BREAK_REQUESTED)
       erts_exit(ERTS_INTR_EXIT, "");
 
   ERTS_SET_BREAK_REQUESTED;
-  ERTS_CHK_IO_AS_INTR(); /* Make sure we don't sleep in poll */
+  /* Wake aux thread to get handle break */
+  erts_aux_thread_poke();
 }
 
 static RETSIGTYPE request_break(int signum)
 {
-#ifdef ERTS_SMP
     smp_sig_notify(signum);
-#else
-    break_requested();
-#endif
 }
 
 #ifdef ETHR_UNUSABLE_SIGUSRX
@@ -789,8 +598,6 @@ signalterm_to_signum(Eterm signal)
     }
 }
 
-#ifdef ERTS_SMP
-
 static ERTS_INLINE Eterm
 signum_to_signalterm(int signum)
 {
@@ -812,37 +619,9 @@ signum_to_signalterm(int signum)
     }
 }
 
-#endif
-
-#ifndef ERTS_SMP
-static ERTS_INLINE Uint
-signum_to_signalstate(int signum)
-{
-    switch (signum) {
-    case SIGINT:  return (1 <<  0);
-    case SIGHUP:  return (1 <<  1);
-    case SIGQUIT: return (1 <<  2);
-    case SIGABRT: return (1 <<  3);
-    case SIGALRM: return (1 <<  4);
-    case SIGTERM: return (1 <<  5);
-    case SIGUSR1: return (1 <<  6);
-    case SIGUSR2: return (1 <<  7);
-    case SIGCHLD: return (1 <<  8);
-    case SIGSTOP: return (1 <<  9);
-    case SIGTSTP: return (1 << 10);
-    default:      return 0;
-    }
-}
-#endif
-
 static RETSIGTYPE generic_signal_handler(int signum)
 {
-#ifdef ERTS_SMP
     smp_sig_notify(signum);
-#else
-    ERTS_SET_SIGNAL_STATE(signum);
-    ERTS_CHK_IO_AS_INTR(); /* Make sure we don't sleep in poll */
-#endif
 }
 
 int erts_set_signal(Eterm signal, Eterm type) {
@@ -914,7 +693,7 @@ erts_sys_unix_later_init(void)
 
 int sys_max_files(void)
 {
-   return(max_files);
+   return max_files;
 }
 
 /************************** OS info *******************************/
@@ -967,34 +746,6 @@ void os_version(int *pMajor, int *pMinor, int *pBuild) {
     *pBuild = get_number(&release); /* Pointer to build number. */
 }
 
-void init_getenv_state(GETENV_STATE *state)
-{
-   erts_smp_rwmtx_rlock(&environ_rwmtx);
-   *state = NULL;
-}
-
-char *getenv_string(GETENV_STATE *state0)
-{
-   char **state = (char **) *state0;
-   char *cp;
-
-   ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&environ_rwmtx));
-
-   if (state == NULL)
-      state = environ;
-
-   cp = *state++;
-   *state0 = (GETENV_STATE) state;
-
-   return cp;
-}
-
-void fini_getenv_state(GETENV_STATE *state)
-{
-   *state = NULL;
-   erts_smp_rwmtx_runlock(&environ_rwmtx);
-}
-
 void erts_do_break_handling(void)
 {
     struct termios temp_mode;
@@ -1005,7 +756,7 @@ void erts_do_break_handling(void)
      * therefore, make sure that all threads but this one are blocked before
      * proceeding!
      */
-    erts_smp_thr_progress_block();
+    erts_thr_progress_block();
 
     /* during break we revert to initial settings */
     /* this is done differently for oldshell */
@@ -1033,25 +784,9 @@ void erts_do_break_handling(void)
       tcsetattr(0,TCSANOW,&temp_mode);
     }
 
-    erts_smp_thr_progress_unblock();
+    erts_thr_progress_unblock();
 }
 
-#ifdef ERTS_SIGNAL_STATE
-void erts_handle_signal_state(void) {
-    Uint signal_state = ERTS_SIGNAL_STATE;
-    Uint i = 0;
-
-    ERTS_CLEAR_SIGNAL_STATE;
-
-    while (signal_state) {
-        if (signal_state & 0x1) {
-            signal_notify_requested(signalstate_sigterm[i]);
-        }
-        i++;
-        signal_state = signal_state >> 1;
-    }
-}
-#endif
 
 /* Fills in the systems representation of the jam/beam process identifier.
 ** The Pid is put in STRING representation in the supplied buffer,
@@ -1064,90 +799,6 @@ void sys_get_pid(char *buffer, size_t buffer_size){
     erts_snprintf(buffer, buffer_size, "%lu",(unsigned long) p);
 }
 
-int
-erts_sys_putenv_raw(char *key, char *value) {
-    return erts_sys_putenv(key, value);
-}
-int
-erts_sys_putenv(char *key, char *value)
-{
-    int res;
-    char *env;
-    Uint need = strlen(key) + strlen(value) + 2;
-
-#ifdef HAVE_COPYING_PUTENV
-    env = erts_alloc(ERTS_ALC_T_TMP, need);
-#else
-    env = erts_alloc(ERTS_ALC_T_PUTENV_STR, need);
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, need);
-#endif
-    strcpy(env,key);
-    strcat(env,"=");
-    strcat(env,value);
-    erts_smp_rwmtx_rwlock(&environ_rwmtx);
-    res = putenv(env);
-    erts_smp_rwmtx_rwunlock(&environ_rwmtx);
-#ifdef HAVE_COPYING_PUTENV
-    erts_free(ERTS_ALC_T_TMP, env);
-#endif
-    return res;
-}
-
-int
-erts_sys_getenv__(char *key, char *value, size_t *size)
-{
-    int res;
-    char *orig_value = getenv(key);
-    if (!orig_value)
-	res = -1;
-    else {
-	size_t len = sys_strlen(orig_value);
-	if (len >= *size) {
-	    *size = len + 1;
-	    res = 1;
-	}
-	else {
-	    *size = len;
-	    sys_memcpy((void *) value, (void *) orig_value, len+1);
-	    res = 0;
-	}
-    }
-    return res;
-}
-
-int
-erts_sys_getenv_raw(char *key, char *value, size_t *size) {
-    return erts_sys_getenv(key, value, size);
-}
-
-/*
- * erts_sys_getenv
- * returns:
- *  -1, if environment key is not set with a value
- *   0, if environment key is set and value fits into buffer size
- *   1, if environment key is set but does not fit into buffer size
- *      size is set with the needed buffer size value
- */
-
-int
-erts_sys_getenv(char *key, char *value, size_t *size)
-{
-    int res;
-    erts_smp_rwmtx_rlock(&environ_rwmtx);
-    res = erts_sys_getenv__(key, value, size);
-    erts_smp_rwmtx_runlock(&environ_rwmtx);
-    return res;
-}
-
-int
-erts_sys_unsetenv(char *key)
-{
-    int res;
-    erts_smp_rwmtx_rwlock(&environ_rwmtx);
-    res = unsetenv(key);
-    erts_smp_rwmtx_rwunlock(&environ_rwmtx);
-    return res;
-}
 
 void sys_init_io(void) { }
 void erts_sys_alloc_init(void) { }
@@ -1286,16 +937,6 @@ erl_assert_error(const char* expr, const char* func, const char* file, int line)
     fprintf(stderr, "%s:%d:%s() Assertion failed: %s\n",
             file, line, func, expr);
     fflush(stderr);
-#if !defined(ERTS_SMP) && 0
-    /* Writing a crashdump from a failed assertion when smp support
-     * is enabled almost a guaranteed deadlocking, don't even bother.
-     *
-     * It could maybe be useful (but I'm not convinced) to write the
-     * crashdump if smp support is disabled...
-     */
-    if (erts_initialized)
-	erl_crash_dump(file, line, "Assertion failed: %s\n", expr);
-#endif
     abort();
 }
 
@@ -1317,22 +958,7 @@ erl_debug(char* fmt, ...)
 
 #endif /* DEBUG */
 
-/*
- * Called from schedule() when it runs out of runnable processes,
- * or when Erlang code has performed INPUT_REDUCTIONS reduction
- * steps. runnable == 0 iff there are no runnable Erlang processes.
- */
-void
-erl_sys_schedule(int runnable)
-{
-    ERTS_CHK_IO(!runnable);
-    ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
-}
-
-
-#ifdef ERTS_SMP
-
-static erts_smp_tid_t sig_dispatcher_tid;
+static erts_tid_t sig_dispatcher_tid;
 
 static void
 smp_sig_notify(int signum)
@@ -1406,7 +1032,7 @@ signal_dispatcher_thread_func(void *unused)
                 }
                 signal_notify_requested(signal);
         }
-        ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
+        ERTS_LC_ASSERT(!erts_thr_progress_is_blocking());
     }
     return NULL;
 }
@@ -1414,7 +1040,7 @@ signal_dispatcher_thread_func(void *unused)
 static void
 init_smp_sig_notify(void)
 {
-    erts_smp_thr_opts_t thr_opts = ERTS_SMP_THR_OPTS_DEFAULT_INITER;
+    erts_thr_opts_t thr_opts = ERTS_THR_OPTS_DEFAULT_INITER;
     thr_opts.detached = 1;
     thr_opts.name = "sys_sig_dispatcher";
 
@@ -1426,7 +1052,7 @@ init_smp_sig_notify(void)
     }
 
     /* Start signal handler thread */
-    erts_smp_thr_create(&sig_dispatcher_tid,
+    erts_thr_create(&sig_dispatcher_tid,
 			signal_dispatcher_thread_func,
 			NULL,
 			&thr_opts);
@@ -1519,103 +1145,15 @@ erts_sys_main_thread(void)
     }
 }
 
-#endif /* ERTS_SMP */
-
-#ifdef ERTS_ENABLE_KERNEL_POLL /* get_value() is currently only used when
-				  kernel-poll is enabled */
-
-/* Get arg marks argument as handled by
-   putting NULL in argv */
-static char *
-get_value(char* rest, char** argv, int* ip)
-{
-    char *param = argv[*ip]+1;
-    argv[*ip] = NULL;
-    if (*rest == '\0') {
-	char *next = argv[*ip + 1];
-	if (next[0] == '-'
-	    && next[1] == '-'
-	    &&  next[2] == '\0') {
-	    erts_fprintf(stderr, "bad \"%s\" value: \n", param);
-	    erts_usage();
-	}
-	(*ip)++;
-	argv[*ip] = NULL;
-	return next;
-    }
-    return rest;
-}
-
-#endif /* ERTS_ENABLE_KERNEL_POLL */
-
 void
 erl_sys_args(int* argc, char** argv)
 {
-    int i, j;
-
-    erts_smp_rwmtx_init(&environ_rwmtx, "environ", NIL,
-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
-
-    i = 1;
-
     ASSERT(argc && argv);
 
-    while (i < *argc) {
-	if(argv[i][0] == '-') {
-	    switch (argv[i][1]) {
-#ifdef ERTS_ENABLE_KERNEL_POLL
-	    case 'K': {
-		char *arg = get_value(argv[i] + 2, argv, &i);
-		if (strcmp("true", arg) == 0) {
-		    erts_use_kernel_poll = 1;
-		}
-		else if (strcmp("false", arg) == 0) {
-		    erts_use_kernel_poll = 0;
-		}
-		else {
-		    erts_fprintf(stderr, "bad \"K\" value: %s\n", arg);
-		    erts_usage();
-		}
-		break;
-	    }
-#endif
-	    case '-':
-		goto done_parsing;
-	    default:
-		break;
-	    }
-	}
-	i++;
-    }
-
- done_parsing:
-
-#ifdef ERTS_ENABLE_KERNEL_POLL
-    if (erts_use_kernel_poll) {
-	char no_kp[10];
-	size_t no_kp_sz = sizeof(no_kp);
-	int res = erts_sys_getenv_raw("ERL_NO_KERNEL_POLL", no_kp, &no_kp_sz);
-	if (res > 0
-	    || (res == 0
-		&& sys_strcmp("false", no_kp) != 0
-		&& sys_strcmp("FALSE", no_kp) != 0)) {
-	    erts_use_kernel_poll = 0;
-	}
-    }
-#endif
-
-    init_check_io();
+    max_files = erts_check_io_max_files();
 
-#ifdef ERTS_SMP
     init_smp_sig_notify();
     init_smp_sig_suspend();
-#endif
 
-    /* Handled arguments have been marked with NULL. Slide arguments
-       not handled towards the beginning of argv. */
-    for (i = 0, j = 0; i < *argc; i++) {
-	if (argv[i])
-	    argv[j++] = argv[i];
-    }
-    *argc = j;
+    erts_sys_env_init();
 }
diff --git a/erts/emulator/sys/unix/sys_drivers.c b/erts/emulator/sys/unix/sys_drivers.c
index 834706d86f..117855acf0 100644
--- a/erts/emulator/sys/unix/sys_drivers.c
+++ b/erts/emulator/sys/unix/sys_drivers.c
@@ -50,17 +50,15 @@
 #include <sys/ioctl.h>
 #endif
 
+#include <sys/types.h>
+#include <sys/socket.h>
+
 #define WANT_NONBLOCKING    /* must define this to pull in defs from sys.h */
 #include "sys.h"
 
-#ifdef USE_THREADS
 #include "erl_threads.h"
-#endif
 
-extern char **environ;
-extern erts_smp_rwmtx_t environ_rwmtx;
-
-extern erts_smp_atomic_t sys_misc_mem_sz;
+extern erts_atomic_t sys_misc_mem_sz;
 
 static Eterm forker_port;
 
@@ -86,12 +84,6 @@ static Eterm forker_port;
 #define MAXIOV 16
 #endif
 
-#ifdef USE_THREADS
-#  define FDBLOCK 1
-#else
-#  define FDBLOCK 0
-#endif
-
 /* Used by the fd driver iff the fd could not be set to non-blocking */
 typedef struct ErtsSysBlocking_ {
     ErlDrvPDL pdl;
@@ -178,9 +170,7 @@ void
 erl_sys_late_init(void)
 {
     SysDriverOpts opts;
-#ifdef ERTS_SMP
     Port *port;
-#endif
 
     sys_signal(SIGPIPE, SIG_IGN); /* Ignore - we'll handle the write failure */
 
@@ -190,20 +180,16 @@ erl_sys_late_init(void)
     opts.read_write = 0;
     opts.hide_window = 0;
     opts.wd = NULL;
-    opts.envir = NULL;
+    erts_osenv_init(&opts.envir);
     opts.exit_status = 0;
     opts.overlapped_io = 0;
     opts.spawn_type = ERTS_SPAWN_ANY;
     opts.argv = NULL;
     opts.parallelism = erts_port_parallelism;
 
-#ifdef ERTS_SMP
     port =
-#endif
         erts_open_driver(&forker_driver, make_internal_pid(0), "forker", &opts, NULL, NULL);
-#ifdef ERTS_SMP
     erts_mtx_unlock(port->lock);
-#endif
     erts_sys_unix_later_init(); /* Need to be called after forker has been started */
 }
 
@@ -220,10 +206,8 @@ static ErlDrvData vanilla_start(ErlDrvPort, char*, SysDriverOpts*);
 
 /* II.III FD prototypes */
 static ErlDrvData fd_start(ErlDrvPort, char*, SysDriverOpts*);
-#if FDBLOCK
 static void fd_async(void *);
 static void fd_ready_async(ErlDrvData drv_data, ErlDrvThreadData thread_data);
-#endif
 static ErlDrvSSizeT fd_control(ErlDrvData, unsigned int, char *, ErlDrvSizeT,
 			       char **, ErlDrvSizeT);
 static void fd_stop(ErlDrvData);
@@ -287,11 +271,7 @@ struct erl_drv_entry fd_driver_entry = {
     fd_control,
     NULL,
     outputv,
-#if FDBLOCK
     fd_ready_async, /* ready_async */
-#else
-    NULL,
-#endif
     fd_flush, /* flush */
     NULL, /* call */
     NULL, /* event */
@@ -363,7 +343,7 @@ static int set_blocking_data(ErtsSysDriverData *dd) {
 
     dd->blocking = erts_alloc(ERTS_ALC_T_SYS_BLOCKING, sizeof(ErtsSysBlocking));
 
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, sizeof(ErtsSysBlocking));
+    erts_atomic_add_nob(&sys_misc_mem_sz, sizeof(ErtsSysBlocking));
 
     dd->blocking->pdl = driver_pdl_create(dd->port_num);
     dd->blocking->res = 0;
@@ -406,7 +386,7 @@ create_driver_data(ErlDrvPort port_num,
         size += sizeof(ErtsSysFdData);
 
     data = erts_alloc(ERTS_ALC_T_DRV_TAB,size);
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, size);
+    erts_atomic_add_nob(&sys_misc_mem_sz, size);
 
     driver_data = (ErtsSysDriverData*)data;
     data += sizeof(*driver_data);
@@ -441,7 +421,7 @@ create_driver_data(ErlDrvPort port_num,
             data += sizeof(*driver_data->ofd);
             init_fd_data(driver_data->ofd, ofd);
         }
-        if (is_blocking && FDBLOCK)
+        if (is_blocking)
             if (!set_blocking_data(driver_data)) {
                 erts_free(ERTS_ALC_T_DRV_TAB, driver_data);
                 return NULL;
@@ -463,85 +443,55 @@ static void close_pipes(int ifd[2], int ofd[2])
     close(ofd[1]);
 }
 
-static char **build_unix_environment(char *block)
+struct __add_spawn_env_state {
+    struct iovec *iov;
+    int *iov_index;
+
+    Sint32 *payload_size;
+    char *env_block;
+};
+
+static void add_spawn_env_block_foreach(void *_state,
+                                        const erts_osenv_data_t *key,
+                                        const erts_osenv_data_t *value)
 {
-    int i;
-    int j;
-    int len;
-    char *cp;
-    char **cpp;
-    char** old_env;
-    
-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&environ_rwmtx));
-
-    cp = block;
-    len = 0;
-    while (*cp != '\0') {
-	cp += strlen(cp) + 1;
-	len++;
-    }
-    old_env = environ;
-    while (*old_env++ != NULL) {
-	len++;
-    }
-    
-    cpp = (char **) erts_alloc_fnf(ERTS_ALC_T_ENVIRONMENT,
-				   sizeof(char *) * (len+1));
-    if (cpp == NULL) {
-	return NULL;
-    }
+    struct __add_spawn_env_state *state;
+    struct iovec *iov;
 
-    cp = block;
-    len = 0;
-    while (*cp != '\0') {
-	cpp[len] = cp;
-	cp += strlen(cp) + 1;
-	len++;
-    }
-    
-    i = len;
-    for (old_env = environ; *old_env; old_env++) {
-	char* old = *old_env;
-
-	for (j = 0; j < len; j++) {
-	    char *s, *t;
-
-	    /* check if cpp[j] equals old
-	       before the = sign,
-	       i.e.
-	       "TMPDIR=/tmp/" */
-	    s = cpp[j];
-	    t = old;
-	    while (*s == *t && *s != '=') {
-		s++, t++;
-	    }
-	    if (*s == '=' && *t == '=') {
-		break;
-	    }
-	}
+    state = (struct __add_spawn_env_state*)(_state);
+    iov = &state->iov[*state->iov_index];
 
-	if (j == len) {		/* New version not found */
-	    cpp[len++] = old;
-	}
-    }
+    iov->iov_base = state->env_block;
 
-    for (j = 0; j < i; ) {
-        size_t last = strlen(cpp[j])-1;
-	if (cpp[j][last] == '=' && strchr(cpp[j], '=') == cpp[j]+last) {
-	    cpp[j] = cpp[--len];
-	    if (len < i) {
-		i--;
-	    } else {
-		j++;
-	    }
-	}
-	else {
-	    j++;
-	}
-    }
+    sys_memcpy(state->env_block, key->data, key->length);
+    state->env_block += key->length;
+    *state->env_block++ = '=';
+    sys_memcpy(state->env_block, value->data, value->length);
+    state->env_block += value->length;
+    *state->env_block++ = '\0';
+
+    iov->iov_len = state->env_block - (char*)iov->iov_base;
+
+    (*state->payload_size) += iov->iov_len;
+    (*state->iov_index)++;
+}
+
+static void *add_spawn_env_block(const erts_osenv_t *env, struct iovec *iov,
+                                  int *iov_index, Sint32 *payload_size) {
+    struct __add_spawn_env_state add_state;
+    char *env_block;
+
+    env_block = erts_alloc(ERTS_ALC_T_TMP, env->content_size +
+        env->variable_count * sizeof("=\0"));
+
+    add_state.iov = iov;
+    add_state.iov_index = iov_index;
+    add_state.env_block = env_block;
+    add_state.payload_size = payload_size;
+
+    erts_osenv_foreach_native(env, &add_state, add_spawn_env_block_foreach);
 
-    cpp[len] = NULL;
-    return cpp;
+    return env_block;
 }
 
 static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
@@ -551,7 +501,6 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
 #define CMD_LINE_PREFIX_STR_SZ (sizeof(CMD_LINE_PREFIX_STR) - 1)
 
     int len;
-    char **new_environ;
     ErtsSysDriverData *dd;
     char *cmd_line;
     char wd_buff[MAXPATHLEN+1];
@@ -618,19 +567,7 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
 	memcpy((void *) (cmd_line + CMD_LINE_PREFIX_STR_SZ), (void *) name, len);
 	cmd_line[CMD_LINE_PREFIX_STR_SZ + len] = '\0';
 	len = CMD_LINE_PREFIX_STR_SZ + len + 1;
-    }
-
-    erts_smp_rwmtx_rlock(&environ_rwmtx);
-
-    if (opts->envir == NULL) {
-	new_environ = environ;
-    } else if ((new_environ = build_unix_environment(opts->envir)) == NULL) {
-	erts_smp_rwmtx_runlock(&environ_rwmtx);
-        close_pipes(ifd, ofd);
-	erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
-	errno = ENOMEM;
-	return ERL_DRV_ERROR_ERRNO;
-    }
+}
 
     if ((cwd = getcwd(wd_buff, MAXPATHLEN+1)) == NULL) {
         /* on some OSs this call opens a fd in the
@@ -639,9 +576,6 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
         int err = errno;
         close_pipes(ifd, ofd);
         erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
-        if (new_environ != environ)
-            erts_free(ERTS_ALC_T_ENVIRONMENT, (void *) new_environ);
-        erts_smp_rwmtx_runlock(&environ_rwmtx);
         errno = err;
         return ERL_DRV_ERROR_ERRNO;
     }
@@ -649,6 +583,7 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
     wd = opts->wd;
 
     {
+        void *environment_block;
         struct iovec *io_vector;
         int iov_len = 5;
         char nullbuff[] = "\0";
@@ -661,10 +596,8 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
 
         if (wd) iov_len++;
 
-        /* count number of elements in environment */
-        while(new_environ[env_len] != NULL)
-            env_len++;
-        iov_len += 1 + env_len; /* num envs including size int */
+        /* num envs including size int */
+        iov_len += 1 + opts->envir.variable_count;
 
         /* count number of element in argument list */
         if (opts->spawn_type == ERTS_SPAWN_EXECUTABLE) {
@@ -681,10 +614,7 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
 
         if (!io_vector) {
             close_pipes(ifd, ofd);
-            erts_smp_rwmtx_runlock(&environ_rwmtx);
             erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
-            if (new_environ != environ)
-                erts_free(ERTS_ALC_T_ENVIRONMENT, (void *) new_environ);
             errno = ENOMEM;
             return ERL_DRV_ERROR_ERRNO;
         }
@@ -719,16 +649,13 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
         io_vector[i++].iov_len = 1;
         buffsz += io_vector[i-1].iov_len;
 
+        env_len = htonl(opts->envir.variable_count);
         io_vector[i].iov_base = (void*)&env_len;
-        env_len = htonl(env_len);
         io_vector[i++].iov_len = sizeof(env_len);
         buffsz += io_vector[i-1].iov_len;
 
-        for (j = 0; new_environ[j] != NULL; j++) {
-            io_vector[i].iov_base = new_environ[j];
-            io_vector[i++].iov_len = strlen(new_environ[j]) + 1;
-            buffsz += io_vector[i-1].iov_len;
-        }
+        environment_block = add_spawn_env_block(&opts->envir, io_vector, &i,
+            &buffsz);
 
         /* only append arguments if this was a spawn_executable */
         if (opts->spawn_type == ERTS_SPAWN_EXECUTABLE) {
@@ -764,9 +691,6 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
                 int err = errno;
                 close_pipes(ifd, ofd);
                 erts_free(ERTS_ALC_T_TMP, io_vector);
-                if (new_environ != environ)
-                    erts_free(ERTS_ALC_T_ENVIRONMENT, (void *) new_environ);
-                erts_smp_rwmtx_runlock(&environ_rwmtx);
                 erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
                 errno = err;
                 return ERL_DRV_ERROR_ERRNO;
@@ -787,16 +711,12 @@ static ErlDrvData spawn_start(ErlDrvPort port_num, char* name,
             driver_select(port_num, ofd[1], ERL_DRV_WRITE|ERL_DRV_USE, 1);
         }
 
+        erts_free(ERTS_ALC_T_TMP, environment_block);
         erts_free(ERTS_ALC_T_TMP, io_vector);
     }
 
     erts_free(ERTS_ALC_T_TMP, (void *) cmd_line);
 
-    if (new_environ != environ)
-	erts_free(ERTS_ALC_T_ENVIRONMENT, (void *) new_environ);
-
-    erts_smp_rwmtx_runlock(&environ_rwmtx);
-
     dd = create_driver_data(port_num, ifd[0], ofd[1], opts->packet_bytes,
                              DO_WRITE | DO_READ, opts->exit_status,
                              0, 0);
@@ -1068,8 +988,8 @@ static void clear_fd_data(ErtsSysFdData *fdd)
 {
     if (fdd->sz > 0) {
 	erts_free(ERTS_ALC_T_FD_ENTRY_BUF, (void *) fdd->buf);
-	ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= fdd->sz);
-	erts_smp_atomic_add_nob(&sys_misc_mem_sz, -1*fdd->sz);
+	ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= fdd->sz);
+	erts_atomic_add_nob(&sys_misc_mem_sz, -1*fdd->sz);
     }
     fdd->buf = NULL;
     fdd->sz = 0;
@@ -1092,13 +1012,11 @@ static void fd_stop(ErlDrvData ev)  /* Does not close the fds */
     ErlDrvPort prt = dd->port_num;
     int sz = sizeof(ErtsSysDriverData);
 
-#if FDBLOCK
     if (dd->blocking) {
         erts_free(ERTS_ALC_T_SYS_BLOCKING, dd->blocking);
         dd->blocking = NULL;
         sz += sizeof(ErtsSysBlocking);
     }
-#endif
 
     if (dd->ifd) {
         sz += sizeof(ErtsSysFdData);
@@ -1110,7 +1028,7 @@ static void fd_stop(ErlDrvData ev)  /* Does not close the fds */
     }
 
      erts_free(ERTS_ALC_T_DRV_TAB, dd);
-     erts_smp_atomic_add_nob(&sys_misc_mem_sz, -sz);
+     erts_atomic_add_nob(&sys_misc_mem_sz, -sz);
 }
 
 static void fd_flush(ErlDrvData ev)
@@ -1191,19 +1109,19 @@ static void outputv(ErlDrvData e, ErlIOVec* ev)
     ev->iov[0].iov_len = pb;
     ev->size += pb;
 
-    if (dd->blocking && FDBLOCK)
+    if (dd->blocking)
         driver_pdl_lock(dd->blocking->pdl);
 
     if ((sz = driver_sizeq(ix)) > 0) {
 	driver_enqv(ix, ev, 0);
 
-        if (dd->blocking && FDBLOCK)
+        if (dd->blocking)
             driver_pdl_unlock(dd->blocking->pdl);
 
 	if (sz + ev->size >= (1 << 13))
 	    set_busy_port(ix, 1);
     }
-    else if (!dd->blocking || !FDBLOCK) {
+    else if (!dd->blocking) {
         /* We try to write directly if the fd in non-blocking */
 	int vsize = ev->vsize > MAX_VSIZE ? MAX_VSIZE : ev->vsize;
 
@@ -1220,7 +1138,6 @@ static void outputv(ErlDrvData e, ErlIOVec* ev)
 	driver_enqv(ix, ev, n);  /* n is the skip value */
 	driver_select(ix, ofd, ERL_DRV_WRITE|ERL_DRV_USE, 1);
     }
-#if FDBLOCK
     else {
         if (ev->size != 0) {
             driver_enqv(ix, ev, 0);
@@ -1231,7 +1148,6 @@ static void outputv(ErlDrvData e, ErlIOVec* ev)
             driver_pdl_unlock(dd->blocking->pdl);
         }
     }
-#endif
     /* return 0;*/
 }
 
@@ -1303,7 +1219,7 @@ static int port_inp_failure(ErtsSysDriverData *dd, int res)
         clear_fd_data(dd->ifd);
     }
 
-    if (dd->blocking && FDBLOCK) {
+    if (dd->blocking) {
         driver_pdl_lock(dd->blocking->pdl);
         if (driver_sizeq(dd->port_num) > 0) {
             driver_pdl_unlock(dd->blocking->pdl);
@@ -1408,7 +1324,7 @@ static void ready_input(ErlDrvData e, ErlDrvEvent ready_fd)
 
             if (dd->ifd->fd < 0) {
                 driver_select(port_num, abs(dd->ifd->fd), ERL_DRV_READ|ERL_DRV_USE, 0);
-                erts_smp_atomic_add_nob(&sys_misc_mem_sz, -sizeof(ErtsSysFdData));
+                erts_atomic_add_nob(&sys_misc_mem_sz, -sizeof(ErtsSysFdData));
                 dd->ifd = NULL;
             }
 
@@ -1514,7 +1430,7 @@ static void ready_input(ErlDrvData e, ErlDrvEvent ready_fd)
 			port_inp_failure(dd, -1);
 		    }
 		    else {
-			erts_smp_atomic_add_nob(&sys_misc_mem_sz, h);
+			erts_atomic_add_nob(&sys_misc_mem_sz, h);
 			sys_memcpy(buf, cpos, bytes_left);
 			dd->ifd->buf = buf;
 			dd->ifd->sz = h;
@@ -1549,7 +1465,7 @@ static void ready_output(ErlDrvData e, ErlDrvEvent ready_fd)
                should close the output fd as soon as the command has
                been sent. */
             driver_select(ix, ready_fd, ERL_DRV_WRITE|ERL_DRV_USE, 0);
-            erts_smp_atomic_add_nob(&sys_misc_mem_sz, -sizeof(ErtsSysFdData));
+            erts_atomic_add_nob(&sys_misc_mem_sz, -sizeof(ErtsSysFdData));
             dd->ofd = NULL;
         }
         if (dd->terminating)
@@ -1579,7 +1495,6 @@ static void stop_select(ErlDrvEvent fd, void* _)
     close((int)fd);
 }
 
-#if FDBLOCK
 
 static void
 fd_async(void *async_data)
@@ -1658,7 +1573,6 @@ void fd_ready_async(ErlDrvData drv_data,
     return; /* 0; */
 }
 
-#endif
 
 /* Forker driver */
 
@@ -1678,15 +1592,13 @@ static ErlDrvData forker_start(ErlDrvPort port_num, char* name,
 
     forker_port = erts_drvport2id(port_num);
 
-    res = erts_sys_getenv_raw("BINDIR", bindir, &bindirsz);
-    if (res != 0) {
-        if (res < 0)
-            erts_exit(1,
-                     "Environment variable BINDIR is not set\n");
-        if (res > 0)
-            erts_exit(1,
-                     "Value of environment variable BINDIR is too large\n");
+    res = erts_sys_explicit_8bit_getenv("BINDIR", bindir, &bindirsz);
+    if (res == 0) {
+        erts_exit(1, "Environment variable BINDIR is not set\n");
+    } else if(res < 0) {
+        erts_exit(1, "Value of environment variable BINDIR is too large\n");
     }
+
     if (bindir[0] != DIR_SEPARATOR_CHAR)
         erts_exit(1,
                  "Environment variable BINDIR does not contain an"
@@ -1749,8 +1661,6 @@ static ErlDrvData forker_start(ErlDrvPort port_num, char* name,
 
     SET_NONBLOCKING(forker_fd);
 
-    driver_select(port_num, forker_fd, ERL_DRV_READ|ERL_DRV_USE, 1);
-
     return (ErlDrvData)port_num;
 }
 
@@ -1847,10 +1757,19 @@ static void forker_ready_output(ErlDrvData e, ErlDrvEvent fd)
 static ErlDrvSSizeT forker_control(ErlDrvData e, unsigned int cmd, char *buf,
                                    ErlDrvSizeT len, char **rbuf, ErlDrvSizeT rlen)
 {
+    static int first_call = 1;
     ErtsSysForkerProto *proto = (ErtsSysForkerProto *)buf;
     ErlDrvPort port_num = (ErlDrvPort)e;
     int res;
 
+    if (first_call) {
+        /*
+         * Do driver_select here when schedulers and their pollsets have started.
+         */
+        driver_select(port_num, forker_fd, ERL_DRV_READ|ERL_DRV_USE, 1);
+        first_call = 0;
+    }
+
     driver_enq(port_num, buf, len);
     if (driver_sizeq(port_num) > sizeof(*proto)) {
         return 0;
diff --git a/erts/emulator/sys/unix/sys_env.c b/erts/emulator/sys/unix/sys_env.c
new file mode 100644
index 0000000000..4d8301f985
--- /dev/null
+++ b/erts/emulator/sys/unix/sys_env.c
@@ -0,0 +1,133 @@
+
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#include "sys.h"
+#include "erl_osenv.h"
+#include "erl_alloc.h"
+
+#include "erl_thr_progress.h"
+
+static erts_osenv_t sysenv_global_env;
+static erts_rwmtx_t sysenv_rwmtx;
+
+extern char **environ;
+
+static void import_initial_env(void);
+
+void erts_sys_env_init() {
+    erts_rwmtx_init(&sysenv_rwmtx, "environ", NIL,
+        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
+
+    erts_osenv_init(&sysenv_global_env);
+    import_initial_env();
+}
+
+const erts_osenv_t *erts_sys_rlock_global_osenv() {
+    erts_rwmtx_rlock(&sysenv_rwmtx);
+    return &sysenv_global_env;
+}
+
+erts_osenv_t *erts_sys_rwlock_global_osenv() {
+    erts_rwmtx_rwlock(&sysenv_rwmtx);
+    return &sysenv_global_env;
+}
+
+void erts_sys_rwunlock_global_osenv() {
+    erts_rwmtx_rwunlock(&sysenv_rwmtx);
+}
+
+void erts_sys_runlock_global_osenv() {
+    erts_rwmtx_runlock(&sysenv_rwmtx);
+}
+
+int erts_sys_explicit_8bit_putenv(char *key, char *value) {
+    erts_osenv_data_t env_key, env_value;
+    int result;
+
+    env_key.length = sys_strlen(key);
+    env_key.data = key;
+
+    env_value.length = sys_strlen(value);
+    env_value.data = value;
+
+    {
+        erts_osenv_t *env = erts_sys_rwlock_global_osenv();
+        result = erts_osenv_put_native(env, &env_key, &env_value);
+        erts_sys_rwunlock_global_osenv();
+    }
+
+    return result;
+}
+
+int erts_sys_explicit_8bit_getenv(char *key, char *value, size_t *size) {
+    erts_osenv_data_t env_key, env_value;
+    int result;
+
+    env_key.length = sys_strlen(key);
+    env_key.data = key;
+
+    /* Reserve space for NUL termination. */
+    env_value.length = *size - 1;
+    env_value.data = value;
+
+    {
+        const erts_osenv_t *env = erts_sys_rlock_global_osenv();
+        result = erts_osenv_get_native(env, &env_key, &env_value);
+        erts_sys_runlock_global_osenv();
+    }
+
+    if(result == 1) {
+        value[env_value.length] = '\0';
+    }
+
+    *size = env_value.length;
+
+    return result;
+}
+
+int erts_sys_explicit_host_getenv(char *key, char *value, size_t *size) {
+    char *orig_value;
+    size_t length;
+    
+    orig_value = getenv(key);
+
+    if(orig_value == NULL) {
+        return 0;
+    }
+
+    length = sys_strlen(orig_value);
+
+    if (length >= *size) {
+        *size = length + 1;
+        return -1;
+    }
+
+    sys_memcpy((void*)value, (void*)orig_value, length + 1);
+    *size = length;
+
+    return 1;
+}
+
+static void import_initial_env(void) {
+    char **environ_iterator, *environ_variable;
+
+    environ_iterator = environ;
+
+    while ((environ_variable = *(environ_iterator++)) != NULL) {
+        char *separator_index = strchr(environ_variable, '=');
+
+        if (separator_index != NULL) {
+            erts_osenv_data_t env_key, env_value;
+
+            env_key.length = separator_index - environ_variable;
+            env_key.data = environ_variable;
+    
+            env_value.length = sys_strlen(separator_index) - 1;
+            env_value.data = separator_index + 1;
+
+            erts_osenv_put_native(&sysenv_global_env, &env_key, &env_value);
+        }
+    }
+}
diff --git a/erts/emulator/sys/unix/sys_float.c b/erts/emulator/sys/unix/sys_float.c
index 6435da086f..a82c15bd32 100644
--- a/erts/emulator/sys/unix/sys_float.c
+++ b/erts/emulator/sys/unix/sys_float.c
@@ -39,7 +39,6 @@ erts_sys_init_float(void)
 
 #else  /* !NO_FPE_SIGNALS */
 
-#ifdef ERTS_SMP
 static erts_tsd_key_t fpe_key;
 
 /* once-only initialisation early in the main thread (via erts_sys_init_float()) */
@@ -61,11 +60,6 @@ static ERTS_INLINE volatile unsigned long *erts_thread_get_fp_exception(void)
 {
     return (volatile unsigned long*)erts_tsd_get(fpe_key);
 }
-#else /* !SMP */
-#define erts_init_fp_exception()	/*empty*/
-static volatile unsigned long fp_exception;
-#define erts_thread_get_fp_exception()	(&fp_exception)
-#endif /* SMP */
 
 volatile unsigned long *erts_get_current_fp_exception(void)
 {
@@ -659,11 +653,9 @@ void erts_sys_init_float(void)
 
 void erts_thread_init_float(void)
 {
-#ifdef ERTS_SMP
     /* This allows Erlang schedulers to leave Erlang-process context
        and still have working FP exceptions. XXX: is this needed? */
     erts_thread_init_fp_exception();
-#endif
 
 #ifndef NO_FPE_SIGNALS
     /* NOTE:
diff --git a/erts/emulator/sys/unix/sys_time.c b/erts/emulator/sys/unix/sys_time.c
index 102ef7bebf..ef05380d17 100644
--- a/erts/emulator/sys/unix/sys_time.c
+++ b/erts/emulator/sys/unix/sys_time.c
@@ -160,7 +160,7 @@ struct sys_time_internal_state_read_mostly__ {
 
 #ifdef ERTS_SYS_TIME_INTERNAL_STATE_WRITE_FREQ__
 struct sys_time_internal_state_write_freq__ {
-    erts_smp_mtx_t mtx;
+    erts_mtx_t mtx;
 #if defined(__linux__) && defined(OS_MONOTONIC_TIME_USING_CLOCK_GETTIME)
     ErtsMonotonicTime last_delivered;
 #endif
@@ -304,7 +304,7 @@ sys_init_time(ErtsSysInitTimeResult *init_resp)
 	    erts_sys_time_data__.r.o.os_times =
 		clock_gettime_times_verified;
 #endif
-            erts_smp_mtx_init(&internal_state.w.f.mtx, "os_monotonic_time", NIL,
+            erts_mtx_init(&internal_state.w.f.mtx, "os_monotonic_time", NIL,
                 ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_IO);
 	    internal_state.w.f.last_delivered
 		= clock_gettime_monotonic();
@@ -525,12 +525,12 @@ static ErtsMonotonicTime clock_gettime_monotonic_verified(void)
     mtime = (ErtsMonotonicTime) posix_clock_gettime(MONOTONIC_CLOCK_ID,
 						    MONOTONIC_CLOCK_ID_STR);
 
-    erts_smp_mtx_lock(&internal_state.w.f.mtx);
+    erts_mtx_lock(&internal_state.w.f.mtx);
     if (mtime < internal_state.w.f.last_delivered)
 	mtime = internal_state.w.f.last_delivered;
     else
 	internal_state.w.f.last_delivered = mtime;
-    erts_smp_mtx_unlock(&internal_state.w.f.mtx);
+    erts_mtx_unlock(&internal_state.w.f.mtx);
 
     return mtime;
 }
@@ -547,12 +547,12 @@ static void clock_gettime_times_verified(ErtsMonotonicTime *mtimep,
 			      WALL_CLOCK_ID_STR,
 			      stimep);
 
-    erts_smp_mtx_lock(&internal_state.w.f.mtx);
+    erts_mtx_lock(&internal_state.w.f.mtx);
     if (*mtimep < internal_state.w.f.last_delivered)
 	*mtimep = internal_state.w.f.last_delivered;
     else
 	internal_state.w.f.last_delivered = *mtimep;
-    erts_smp_mtx_unlock(&internal_state.w.f.mtx);
+    erts_mtx_unlock(&internal_state.w.f.mtx);
 }
 
 #endif /* defined(OS_SYSTEM_TIME_USING_CLOCK_GETTIME) */
@@ -878,8 +878,6 @@ ErtsMonotonicTime
 erts_os_monotonic_time(void)
 {
     Uint32 ticks = get_tick_count();
-    ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
-				      ticks);
     return ERTS_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
 					 ticks) << internal_state.r.o.times_shift;
 }
diff --git a/erts/emulator/sys/unix/sys_uds.c b/erts/emulator/sys/unix/sys_uds.c
index dd0a3b03ff..278c6b6ba1 100644
--- a/erts/emulator/sys/unix/sys_uds.c
+++ b/erts/emulator/sys/unix/sys_uds.c
@@ -18,6 +18,42 @@
  * %CopyrightEnd%
  */
 
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#if defined(__sun__) && !defined(_XOPEN_SOURCE)
+#define _XOPEN_SOURCE 500
+#endif
+
+#include <limits.h>
+
+#include <sys/types.h>
+#include <sys/socket.h>
+
+#ifdef HAVE_SYS_SOCKETIO_H
+#   include <sys/socketio.h>
+#endif
+#ifdef HAVE_SYS_SOCKIO_H
+#   include <sys/sockio.h>
+#endif
+
+#ifdef HAVE_NET_ERRNO_H
+#include <net/errno.h>
+#endif
+
+#ifdef HAVE_DIRENT_H
+#  include <dirent.h>
+#endif
+
+#ifdef HAVE_UNISTD_H
+#   include <unistd.h>
+#endif
+
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+
 #include "sys_uds.h"
 
 int
diff --git a/erts/emulator/sys/unix/sys_uds.h b/erts/emulator/sys/unix/sys_uds.h
index a598102d5c..26c91d6a00 100644
--- a/erts/emulator/sys/unix/sys_uds.h
+++ b/erts/emulator/sys/unix/sys_uds.h
@@ -21,18 +21,6 @@
 #ifndef _ERL_UNIX_UDS_H
 #define _ERL_UNIX_UDS_H
 
-#ifdef HAVE_CONFIG_H
-#  include "config.h"
-#endif
-
-#if defined(__sun__) && !defined(_XOPEN_SOURCE)
-#define _XOPEN_SOURCE 500
-#endif
-
-#include <limits.h>
-
-#include <sys/types.h>
-#include <sys/socket.h>
 #include <sys/uio.h>
 
 #if defined IOV_MAX
@@ -43,8 +31,6 @@
 #define MAXIOV 16
 #endif
 
-#include "sys.h"
-
 int sys_uds_readv(int fd, struct iovec *iov, size_t iov_len,
                   int *fds, int fd_count, int flags);
 int sys_uds_read(int fd, char *buff, size_t len,
diff --git a/erts/emulator/sys/win32/erl_poll.c b/erts/emulator/sys/win32/erl_poll.c
index 8743f83a50..fd4c745c3b 100644
--- a/erts/emulator/sys/win32/erl_poll.c
+++ b/erts/emulator/sys/win32/erl_poll.c
@@ -34,6 +34,7 @@
  */
 
 /*#define HARDDEBUG */
+/*#define HARDTRACE */
 #ifdef HARDDEBUG
 #ifdef HARDTRACE
 #define HARDTRACEF(X) my_debug_printf##X
@@ -50,7 +51,7 @@ static void my_debug_printf(char *fmt, ...)
     va_start(args, fmt);
     erts_vsnprintf(buffer,1024,fmt,args);
     va_end(args);
-    erts_fprintf(stderr,"%s\r\n",buffer);
+    erts_printf("%s\r\n",buffer);
 }
 #else
 #define HARDTRACEF(X)
@@ -274,53 +275,35 @@ typedef struct _Waiter {
 /*
  * The structure for a pollset. There can currently be only one...
  */
-struct ErtsPollSet_ {
+struct erts_pollset {
     Waiter** waiter;
     int allocated_waiters;  /* Size ow waiter array */ 
     int num_waiters;	    /* Number of waiter threads. */
-    int restore_events;        /* Tells us to restore waiters events 
-				  next time around */
     HANDLE event_io_ready;     /* To be used when waiting for io */
     /* These are used to wait for workers to enter standby */
     volatile int standby_wait_counter; /* Number of threads to wait for */
     CRITICAL_SECTION standby_crit;     /* CS to guard the counter */
-    HANDLE standby_wait_event;         /* Event signalled when counte == 0 */
+    HANDLE standby_wait_event;         /* Event signalled when counter == 0 */
     erts_atomic32_t wakeup_state;
-#ifdef ERTS_SMP
-    erts_smp_mtx_t mtx;
-#endif
-    erts_atomic64_t timeout_time;
+    erts_mtx_t mtx;
 };
 
-#ifdef ERTS_SMP
 
 #define ERTS_POLLSET_LOCK(PS) \
-  erts_smp_mtx_lock(&(PS)->mtx)
+  erts_mtx_lock(&(PS)->mtx)
 #define ERTS_POLLSET_UNLOCK(PS) \
-  erts_smp_mtx_unlock(&(PS)->mtx)
-
-#else
+  erts_mtx_unlock(&(PS)->mtx)
 
-#define ERTS_POLLSET_LOCK(PS)
-#define ERTS_POLLSET_UNLOCK(PS)
-
-#endif
 
 /*
  * Communication with sys_interrupt
  */
 
-#ifdef ERTS_SMP
-extern erts_smp_atomic32_t erts_break_requested;
+extern erts_atomic32_t erts_break_requested;
 #define ERTS_SET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
 #define ERTS_UNSET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
-#else
-extern volatile int erts_break_requested;
-#define ERTS_SET_BREAK_REQUESTED (erts_break_requested = 1)
-#define ERTS_UNSET_BREAK_REQUESTED (erts_break_requested = 0)
-#endif
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
 
 static erts_mtx_t break_waiter_lock;
 static HANDLE break_happened_event;
@@ -367,39 +350,19 @@ do { \
     wait_standby(PS); \
  } while(0)
 
-static ERTS_INLINE void
-init_timeout_time(ErtsPollSet ps)
-{
-    erts_atomic64_init_nob(&ps->timeout_time,
-			   (erts_aint64_t) ERTS_MONOTONIC_TIME_MAX);
-}
-
-static ERTS_INLINE void
-set_timeout_time(ErtsPollSet ps, ErtsMonotonicTime time)
-{
-    erts_atomic64_set_relb(&ps->timeout_time,
-			   (erts_aint64_t) time);
-}
-
-static ERTS_INLINE ErtsMonotonicTime
-get_timeout_time(ErtsPollSet ps)
-{
-    return (ErtsMonotonicTime) erts_atomic64_read_acqb(&ps->timeout_time);
-}
-
 #define ERTS_POLL_NOT_WOKEN		((erts_aint32_t) 0)
 #define ERTS_POLL_WOKEN_IO_READY	((erts_aint32_t) 1)
 #define ERTS_POLL_WOKEN_INTR		((erts_aint32_t) 2)
 #define ERTS_POLL_WOKEN_TIMEDOUT	((erts_aint32_t) 3)
 
 static ERTS_INLINE int
-is_io_ready(ErtsPollSet ps)
+is_io_ready(ErtsPollSet *ps)
 {
     return erts_atomic32_read_nob(&ps->wakeup_state) == ERTS_POLL_WOKEN_IO_READY;
 }
 
 static ERTS_INLINE void
-woke_up(ErtsPollSet ps)
+woke_up(ErtsPollSet *ps)
 {
     if (erts_atomic32_read_nob(&ps->wakeup_state) == ERTS_POLL_NOT_WOKEN)
 	erts_atomic32_cmpxchg_nob(&ps->wakeup_state,
@@ -422,7 +385,7 @@ woke_up(ErtsPollSet ps)
 }
 
 static ERTS_INLINE int
-wakeup_cause(ErtsPollSet ps)
+wakeup_cause(ErtsPollSet *ps)
 {
     int res;
     erts_aint32_t wakeup_state = erts_atomic32_read_acqb(&ps->wakeup_state);
@@ -445,46 +408,8 @@ wakeup_cause(ErtsPollSet ps)
     return res;
 }
 
-static ERTS_INLINE DWORD
-poll_wait_timeout(ErtsPollSet ps, ErtsMonotonicTime timeout_time)
-{
-    ErtsMonotonicTime current_time, diff_time, timeout;
-
-    if (timeout_time == ERTS_POLL_NO_TIMEOUT) {
-    no_timeout:
-	set_timeout_time(ps, ERTS_MONOTONIC_TIME_MIN);
-	woke_up(ps);
-	return (DWORD) 0;
-    }
-
-    current_time = erts_get_monotonic_time(NULL);
-    diff_time = timeout_time - current_time;
-    if (diff_time <= 0)
-	goto no_timeout;
-
-    /* Round up to nearest milli second */
-    timeout = (ERTS_MONOTONIC_TO_MSEC(diff_time - 1) + 1);
-    if (timeout > INT_MAX)
-	timeout = INT_MAX; /* Also prevents DWORD overflow */
-
-    set_timeout_time(ps, current_time + ERTS_MSEC_TO_MONOTONIC(timeout));
-
-    ResetEvent(ps->event_io_ready);
-    /*
-     * Since we don't know the internals of ResetEvent() we issue
-     * a memory barrier as a safety precaution ensuring that
-     * the load of wakeup_state wont be reordered with stores made
-     * by ResetEvent().
-     */
-    ERTS_THR_MEMORY_BARRIER;
-    if (erts_atomic32_read_nob(&ps->wakeup_state) != ERTS_POLL_NOT_WOKEN)
-	return (DWORD) 0;
-
-    return (DWORD) timeout;
-}
-
 static ERTS_INLINE void
-wake_poller(ErtsPollSet ps, int io_ready)
+wake_poller(ErtsPollSet *ps, int io_ready)
 {
     erts_aint32_t wakeup_state;
     if (io_ready) {
@@ -519,13 +444,13 @@ wake_poller(ErtsPollSet ps, int io_ready)
 }
 
 static ERTS_INLINE void
-reset_io_ready(ErtsPollSet ps)
+reset_io_ready(ErtsPollSet *ps)
 {
     erts_atomic32_set_nob(&ps->wakeup_state, ERTS_POLL_NOT_WOKEN);
 }
 
 static ERTS_INLINE void
-restore_io_ready(ErtsPollSet ps)
+restore_io_ready(ErtsPollSet *ps)
 {
     erts_atomic32_set_nob(&ps->wakeup_state, ERTS_POLL_WOKEN_IO_READY);
 }
@@ -535,13 +460,13 @@ restore_io_ready(ErtsPollSet ps)
  * notifying a poller thread about I/O ready.
  */
 static ERTS_INLINE void
-notify_io_ready(ErtsPollSet ps)
+notify_io_ready(ErtsPollSet *ps)
 {
     wake_poller(ps, 1);
 }
 
 static ERTS_INLINE void
-reset_interrupt(ErtsPollSet ps)
+reset_interrupt(ErtsPollSet *ps)
 {
     /* We need to keep io-ready if set */
     erts_aint32_t wakeup_state = erts_atomic32_read_nob(&ps->wakeup_state);
@@ -558,12 +483,12 @@ reset_interrupt(ErtsPollSet ps)
 }
 
 static ERTS_INLINE void
-set_interrupt(ErtsPollSet ps)
+set_interrupt(ErtsPollSet *ps)
 {
     wake_poller(ps, 0);
 }
 
-static void setup_standby_wait(ErtsPollSet ps, int num_threads)
+static void setup_standby_wait(ErtsPollSet *ps, int num_threads)
 {
     EnterCriticalSection(&(ps->standby_crit));
     ps->standby_wait_counter = num_threads;
@@ -571,7 +496,7 @@ static void setup_standby_wait(ErtsPollSet ps, int num_threads)
     LeaveCriticalSection(&(ps->standby_crit));
 }
 
-static void signal_standby(ErtsPollSet ps) 
+static void signal_standby(ErtsPollSet *ps)
 {
     EnterCriticalSection(&(ps->standby_crit));
     --(ps->standby_wait_counter);
@@ -585,7 +510,7 @@ static void signal_standby(ErtsPollSet ps)
     LeaveCriticalSection(&(ps->standby_crit));
 }
 
-static void wait_standby(ErtsPollSet ps)
+static void wait_standby(ErtsPollSet *ps)
 {
     WaitForSingleObject(ps->standby_wait_event,INFINITE);
 }
@@ -653,7 +578,7 @@ static void consistency_check(Waiter* w)
 
 #endif
 
-static void new_waiter(ErtsPollSet ps)
+static void new_waiter(ErtsPollSet *ps)
 {
     register Waiter* w;
     DWORD tid;			/* Id for thread. */
@@ -747,7 +672,7 @@ static void *break_waiter(void *param)
 static void *threaded_waiter(void *param)
 {
     register Waiter* w = (Waiter *) param;
-    ErtsPollSet ps = (ErtsPollSet) w->xdata;
+    ErtsPollSet *ps = (ErtsPollSet*) w->xdata;
 #ifdef HARD_POLL_DEBUG2
     HANDLE oold_fired[64];
     int num_oold_fired;
@@ -850,9 +775,9 @@ event_happened:
 	    ASSERT(i >= WAIT_OBJECT_0+1);
 	    i -= WAIT_OBJECT_0;
 	    ASSERT(i >= 1);
-	    w->active_events--;
 	    HARDDEBUGF(("i = %d, a,h,t = %d,%d,%d",i,
 			w->active_events, w->highwater, w->total_events));
+	    w->active_events--;
 #ifdef HARD_POLL_DEBUG2
 	    fired[num_fired++] = w->events[i];
 #endif
@@ -882,7 +807,7 @@ event_happened:
  * The actual adding and removing from pollset utilities 
  */
 
-static int set_driver_select(ErtsPollSet ps, HANDLE event, ErtsPollEvents mode)
+static int set_driver_select(ErtsPollSet *ps, HANDLE event, ErtsPollEvents mode)
 {
     int i;
     int best_waiter = -1;	/* The waiter with lowest number of events. */
@@ -972,13 +897,13 @@ static int set_driver_select(ErtsPollSet ps, HANDLE event, ErtsPollEvents mode)
 #endif
     erts_mtx_unlock(&w->mtx);
     START_WAITER(ps,w);
-    HARDDEBUGF(("add select %d %d %d %d",best_waiter,
+    HARDDEBUGF(("%d: add select %d %d %d %d", event, best_waiter,
 		w->active_events,w->highwater,w->total_events));
     return mode;
 }
 
 
-static int cancel_driver_select(ErtsPollSet ps, HANDLE event)
+static int cancel_driver_select(ErtsPollSet *ps, HANDLE event)
 {
     int i;
 
@@ -1033,7 +958,7 @@ static int cancel_driver_select(ErtsPollSet ps, HANDLE event)
  * Interface functions
  */
 
-void  erts_poll_interrupt(ErtsPollSet ps, int set /* bool */)
+void  erts_poll_interrupt(ErtsPollSet *ps, int set /* bool */)
 {
     HARDTRACEF(("In erts_poll_interrupt(%d)",set));
     if (!set)
@@ -1043,35 +968,23 @@ void  erts_poll_interrupt(ErtsPollSet ps, int set /* bool */)
     HARDTRACEF(("Out erts_poll_interrupt(%d)",set));
 }
 
-void erts_poll_interrupt_timed(ErtsPollSet ps,
-			       int set /* bool */,
-			       ErtsMonotonicTime timeout_time)
-{
-    HARDTRACEF(("In erts_poll_interrupt_timed(%d,%ld)",set,timeout_time));
-    if (!set)
-	reset_interrupt(ps);
-    else if (get_timeout_time(ps) > timeout_time)
-	set_interrupt(ps);
-    HARDTRACEF(("Out erts_poll_interrupt_timed"));
-}
-
 
 /*
  * Windows is special, there is actually only one event type, and
  * the only difference between ERTS_POLL_EV_IN and ERTS_POLL_EV_OUT
  * is which driver callback will eventually be called.
  */
-static ErtsPollEvents do_poll_control(ErtsPollSet ps,
-				      ErtsSysFdType fd,
-				      ErtsPollEvents pe,
-				      int on /* bool */)
+static ErtsPollEvents do_poll_control(ErtsPollSet *ps,
+                                      ErtsSysFdType fd,
+                                      ErtsPollOp op,
+                                      ErtsPollEvents pe)
 {
     HANDLE event = (HANDLE) fd;
     ErtsPollEvents mode;
     ErtsPollEvents result;
     ASSERT(event != INVALID_HANDLE_VALUE);
 
-    if (on) {
+    if (op != ERTS_POLL_OP_DEL) {
 	if (pe & ERTS_POLL_EV_IN || !(pe & ERTS_POLL_EV_OUT )) {
 	    mode = ERTS_POLL_EV_IN;
 	} else {
@@ -1084,51 +997,30 @@ static ErtsPollEvents do_poll_control(ErtsPollSet ps,
     return result;
 }
 
-ErtsPollEvents erts_poll_control(ErtsPollSet ps,
+ErtsPollEvents erts_poll_control(ErtsPollSet *ps,
 				 ErtsSysFdType fd,
+                                 ErtsPollOp op,
 				 ErtsPollEvents pe,
-				 int on,
 				 int* do_wake) /* In: Wake up polling thread */
 				               /* Out: Poller is woken */
 {
     ErtsPollEvents result;
-    HARDTRACEF(("In erts_poll_control(0x%08X, %u, %d)",(unsigned long) fd, (unsigned) pe, on));
+    HARDTRACEF(("In erts_poll_control(0x%08X, %s, %s)",
+                (unsigned long) fd, op2str(op), ev2str(pe)));
     ERTS_POLLSET_LOCK(ps);
-    result=do_poll_control(ps,fd,pe,on);
+    result=do_poll_control(ps, fd, op, pe);
     ERTS_POLLSET_UNLOCK(ps);
     *do_wake = 0; /* Never any need to wake polling threads on windows */
     HARDTRACEF(("Out erts_poll_control -> %u",(unsigned) result));
     return result;
 }
 
-void erts_poll_controlv(ErtsPollSet ps,
-			ErtsPollControlEntry pcev[],
-			int len)
-{
-    int i;
-    int hshur = 0;
-    int do_wake = 0;
-
-    HARDTRACEF(("In erts_poll_controlv(%d)",len));
-    ERTS_POLLSET_LOCK(ps);
-
-    for (i = 0; i < len; i++) {
-	pcev[i].events = do_poll_control(ps,
-					 pcev[i].fd,
-					 pcev[i].events,
-					 pcev[i].on);
-    }
-    ERTS_POLLSET_UNLOCK(ps);
-    HARDTRACEF(("Out erts_poll_controlv"));
-}
-
-int erts_poll_wait(ErtsPollSet ps,
+int erts_poll_wait(ErtsPollSet *ps,
 		   ErtsPollResFd pr[],
-		   int *len,
-		   ErtsMonotonicTime timeout_time)
+		   int *len)
 {
     int no_fds;
-    DWORD timeout;
+    DWORD timeout = INFINITE;
     EventData* ev;
     int res = 0;
     int num = 0;
@@ -1139,42 +1031,6 @@ int erts_poll_wait(ErtsPollSet ps,
     HARDTRACEF(("In erts_poll_wait"));
     ERTS_POLLSET_LOCK(ps);
 
-    if (!is_io_ready(ps) && ps->restore_events) {
-	HARDDEBUGF(("Restore events: %d",ps->num_waiters));
-	ps->restore_events = 0;
-	for (i = 0; i < ps->num_waiters; ++i) {
-	   Waiter* w = ps->waiter[i];
-	   erts_mtx_lock(&w->mtx);
-	   HARDDEBUGF(("Maybe reset %d %d %d %d",i,
-		       w->active_events,w->highwater,w->total_events));
-	   if (w->active_events < w->total_events) {
-	       erts_mtx_unlock(&w->mtx);
-	       STOP_WAITER(ps,w);
-	       HARDDEBUGF(("Need reset %d %d %d %d",i,
-			   w->active_events,w->highwater,w->total_events));
-	       erts_mtx_lock(&w->mtx);	       
-	       /* Need reset, just check that it doesn't have got more to tell */
-	       if (w->highwater != w->active_events) {
-		   HARDDEBUGF(("Oups!"));
-		   /* Oups, got signalled before we took the lock, can't reset */
-		   if(!is_io_ready(ps)) {
-		       erts_exit(ERTS_ERROR_EXIT,"Internal error: "
-				"Inconsistent io structures in erl_poll.\n");
-		   }
-		   START_WAITER(ps,w);
-		   erts_mtx_unlock(&w->mtx);
-		   ps->restore_events = 1; 
-		   continue;
-	       }
-	       w->active_events = w->highwater =  w->total_events;
-	       START_WAITER(ps,w);
-	       erts_mtx_unlock(&w->mtx);
-	   } else {
-	       erts_mtx_unlock(&w->mtx);
-	   }
-	}
-    }
-
     no_fds = *len;
 
 #ifdef ERTS_POLL_MAX_RES
@@ -1182,26 +1038,29 @@ int erts_poll_wait(ErtsPollSet ps,
 	no_fds = ERTS_POLL_MAX_RES;
 #endif
 
-    timeout = poll_wait_timeout(ps, timeout_time);
-
-    /*HARDDEBUGF(("timeout = %ld",(long) timeout));*/
+    ResetEvent(ps->event_io_ready);
+    /*
+     * Since we don't know the internals of ResetEvent() we issue
+     * a memory barrier as a safety precaution ensuring that
+     * the load of wakeup_state wont be reordered with stores made
+     * by ResetEvent().
+     */
+    ERTS_THR_MEMORY_BARRIER;
+    if (erts_atomic32_read_nob(&ps->wakeup_state) != ERTS_POLL_NOT_WOKEN)
+       timeout = (DWORD) 0;
 
-    if (timeout > 0 && !erts_atomic32_read_nob(&break_waiter_state)) {
+    if (!erts_atomic32_read_nob(&break_waiter_state)) {
 	HANDLE harr[2] = {ps->event_io_ready, break_happened_event};
 	int num_h = 2;
-        ERTS_MSACC_PUSH_STATE_M();
+        ERTS_MSACC_PUSH_STATE();
 
 	HARDDEBUGF(("Start waiting %d [%d]",num_h, (int) timeout));
 	ERTS_POLLSET_UNLOCK(ps);
-#ifdef ERTS_SMP
 	erts_thr_progress_prepare_wait(NULL);
-#endif
-        ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP);
+        ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP);
 	WaitForMultipleObjects(num_h, harr, FALSE, timeout);
-#ifdef ERTS_SMP
 	erts_thr_progress_finalize_wait(NULL);
-#endif
-        ERTS_MSACC_POP_STATE_M();
+        ERTS_MSACC_POP_STATE();
 	ERTS_POLLSET_LOCK(ps);
 	HARDDEBUGF(("Stop waiting %d [%d]",num_h, (int) timeout));
 	woke_up(ps);
@@ -1234,7 +1093,7 @@ int erts_poll_wait(ErtsPollSet ps,
 
     reset_io_ready(ps);
 
-    n = ps->num_waiters;	
+    n = ps->num_waiters;
 
     for (i = 0; i < n; i++) {
 	Waiter* w = ps->waiter[i];
@@ -1260,11 +1119,10 @@ int erts_poll_wait(ErtsPollSet ps,
 		HARDDEBUGF(("To many FD's to report!"));
 		goto done;
 	    }
-	    HARDDEBUGF(("SET! Restore events"));
-	    ps->restore_events = 1;
 	    HARDDEBUGF(("Report %d,%d",i,j));
-	    pr[num].fd = (ErtsSysFdType) w->events[j];
-	    pr[num].events = w->evdata[j]->mode;
+	    ERTS_POLL_RES_SET_FD(&pr[num], w->events[j]);
+	    ERTS_POLL_RES_SET_EVTS(&pr[num], w->evdata[j]->mode);
+            remove_event_from_set(w, j);
 #ifdef HARD_POLL_DEBUG
 	    poll_debug_reported(w->events[j],w->highwater | (j << 16));
 	    poll_debug_reported(w->events[j],first | (last << 16));
@@ -1272,13 +1130,14 @@ int erts_poll_wait(ErtsPollSet ps,
 	    ++num;
 	}
 
+        w->total_events = w->highwater = w->active_events;
+
 #ifdef DEBUG
 	consistency_check(w);
 #endif
 	erts_mtx_unlock(&w->mtx);
     }
  done:
-    set_timeout_time(ps, ERTS_MONOTONIC_TIME_MAX);
     *len = num;
     ERTS_POLLSET_UNLOCK(ps);
     HARDTRACEF(("Out erts_poll_wait"));
@@ -1293,7 +1152,7 @@ int erts_poll_max_fds(void)
     return res;
 }
 
-void erts_poll_info(ErtsPollSet ps,
+void erts_poll_info(ErtsPollSet *ps,
 		    ErtsPollInfo *pip)
 {
     Uint size = 0;
@@ -1303,7 +1162,7 @@ void erts_poll_info(ErtsPollSet ps,
     HARDTRACEF(("In erts_poll_info"));
     ERTS_POLLSET_LOCK(ps);
 
-    size += sizeof(struct ErtsPollSet_);
+    size += sizeof(struct erts_pollset);
     size += sizeof(Waiter *) * ps->allocated_waiters;
     for (i = 0; i < ps->num_waiters; ++i) {
 	Waiter *w = ps->waiter[i];
@@ -1318,16 +1177,12 @@ void erts_poll_info(ErtsPollSet ps,
 
     pip->primary = "WaitForMultipleObjects"; 
 
-    pip->fallback = NULL;
-
     pip->kernel_poll = NULL;
 
     pip->memory_size = size;
 
     pip->poll_set_size = num_events;
 
-    pip->fallback_poll_set_size = 0;
-
     pip->lazy_updates = 0;
 
     pip->pending_updates = 0;
@@ -1335,6 +1190,8 @@ void erts_poll_info(ErtsPollSet ps,
     pip->batch_updates = 0;
 
     pip->concurrent_updates = 0;
+
+    pip->is_fallback = 0;
     ERTS_POLLSET_UNLOCK(ps);
 
     pip->max_fds = erts_poll_max_fds();
@@ -1342,10 +1199,10 @@ void erts_poll_info(ErtsPollSet ps,
 
 }
 
-ErtsPollSet erts_poll_create_pollset(void)
+ErtsPollSet *erts_poll_create_pollset(int no)
 {
-    ErtsPollSet ps = SEL_ALLOC(ERTS_ALC_T_POLLSET,
-			       sizeof(struct ErtsPollSet_));
+    ErtsPollSet *ps = SEL_ALLOC(ERTS_ALC_T_POLLSET,
+			       sizeof(struct erts_pollset));
     HARDTRACEF(("In erts_poll_create_pollset"));
 
     ps->num_waiters = 0;
@@ -1356,19 +1213,15 @@ ErtsPollSet erts_poll_create_pollset(void)
     ps->standby_wait_counter = 0;
     ps->event_io_ready = CreateManualEvent(FALSE);
     ps->standby_wait_event = CreateManualEvent(FALSE); 
-    ps->restore_events = 0;
 
     erts_atomic32_init_nob(&ps->wakeup_state, ERTS_POLL_NOT_WOKEN);
-#ifdef ERTS_SMP
-    erts_smp_mtx_init(&ps->mtx, "pollset", NIL, ERTS_LOCK_FLAGS_CATEGORY_IO);
-#endif
-    init_timeout_time(ps);
+    erts_mtx_init(&ps->mtx, "pollset", NIL, ERTS_LOCK_FLAGS_CATEGORY_IO);
 
     HARDTRACEF(("Out erts_poll_create_pollset"));
     return ps;
 }
 
-void erts_poll_destroy_pollset(ErtsPollSet ps)
+void erts_poll_destroy_pollset(ErtsPollSet *ps)
 {
     int i;
     HARDTRACEF(("In erts_poll_destroy_pollset"));
@@ -1391,9 +1244,7 @@ void erts_poll_destroy_pollset(ErtsPollSet ps)
     CloseHandle(ps->event_io_ready);
     CloseHandle(ps->standby_wait_event);
     ERTS_POLLSET_UNLOCK(ps);
-#ifdef ERTS_SMP
-    erts_smp_mtx_destroy(&ps->mtx);
-#endif
+    erts_mtx_destroy(&ps->mtx);
     SEL_FREE(ERTS_ALC_T_POLLSET, (void *) ps);
     HARDTRACEF(("Out erts_poll_destroy_pollset"));
 }
@@ -1401,14 +1252,16 @@ void erts_poll_destroy_pollset(ErtsPollSet ps)
 /*
  * Actually mostly initializes the friend module sys_interrupt...
  */
-void  erts_poll_init(void)
+void  erts_poll_init(int *concurrent_updates)
 {
-    erts_tid_t thread;
 
 #ifdef HARD_POLL_DEBUG
     poll_debug_init();
 #endif
 
+    if (concurrent_updates)
+        *concurrent_updates = 0;
+
     HARDTRACEF(("In erts_poll_init"));
     erts_sys_break_event = CreateManualEvent(FALSE);
 
@@ -1417,21 +1270,26 @@ void  erts_poll_init(void)
     break_happened_event = CreateManualEvent(FALSE);
     erts_atomic32_init_nob(&break_waiter_state, 0); 
 
+    HARDTRACEF(("Out erts_poll_init"));
+}
+
+void erts_poll_late_init(void)
+{
+    erts_tid_t thread;
     erts_thr_create(&thread, &break_waiter, NULL, NULL);
     ERTS_UNSET_BREAK_REQUESTED;
-    HARDTRACEF(("Out erts_poll_init"));
 }
 
 /*
  * Non windows friendly interface, not used when fd's are not continous
  */
-void  erts_poll_get_selected_events(ErtsPollSet ps,
+void  erts_poll_get_selected_events(ErtsPollSet *ps,
 				    ErtsPollEvents ev[],
 				    int len)
 {
     int i;
     HARDTRACEF(("In erts_poll_get_selected_events"));
     for (i = 0; i < len; ++i)
-	ev[i] = 0;
+	ev[i] = ERTS_POLL_EV_NONE;
     HARDTRACEF(("Out erts_poll_get_selected_events"));
 }
diff --git a/erts/emulator/sys/win32/erl_win32_sys_ddll.c b/erts/emulator/sys/win32/erl_win32_sys_ddll.c
index 274133a346..fc2179328f 100644
--- a/erts/emulator/sys/win32/erl_win32_sys_ddll.c
+++ b/erts/emulator/sys/win32/erl_win32_sys_ddll.c
@@ -46,9 +46,17 @@ static TWinDynDriverCallbacks wddc;
 static TWinDynNifCallbacks nif_callbacks;
 
 void erl_sys_ddll_init(void) {
+    WCHAR cwd_buffer[MAX_PATH];
+
     tls_index = TlsAlloc();
     ERL_INIT_CALLBACK_STRUCTURE(wddc);
 
+    /* LOAD_WITH_ALTERED_SEARCH_PATH removes the startup directory from the
+     * search path, so we add it separately to be backwards compatible. */
+    if (GetCurrentDirectoryW(sizeof(cwd_buffer), cwd_buffer)) {
+        SetDllDirectoryW(cwd_buffer);
+    }
+
 #define ERL_NIF_API_FUNC_DECL(RET,NAME,ARGS) nif_callbacks.NAME = NAME
 #include "erl_nif_api_funcs.h"
 #undef ERL_NIF_API_FUNC_DECL
@@ -81,7 +89,10 @@ int erts_sys_ddll_open(const char *full_name, void **handle, ErtsSysDdllError* e
 						   ERTS_ALC_T_TMP, &used, EXT_LEN);
     wcscpy(&wcp[used/2 - 1], FILE_EXT_WCHAR);
 
-    if ((hinstance = LoadLibraryW(wcp)) == NULL) {
+    /* LOAD_WITH_ALTERED_SEARCH_PATH adds the specified DLL's directory to the
+     * dependency search path. This also removes the directory we started in,
+     * but we've explicitly added that in in erl_sys_ddll_init. */
+    if ((hinstance = LoadLibraryExW(wcp, NULL, LOAD_WITH_ALTERED_SEARCH_PATH)) == NULL) {
 	code = ERL_DE_DYNAMIC_ERROR_OFFSET - GetLastError();
 	if (err != NULL) {
 	    err->str = erts_sys_ddll_error(code);
diff --git a/erts/emulator/sys/win32/erl_win_dyn_driver.h b/erts/emulator/sys/win32/erl_win_dyn_driver.h
index 6f28d513c2..0d1a6d4c87 100644
--- a/erts/emulator/sys/win32/erl_win_dyn_driver.h
+++ b/erts/emulator/sys/win32/erl_win_dyn_driver.h
@@ -40,7 +40,6 @@ WDD_TYPEDEF(int, driver_exit, (ErlDrvPort, int));
 WDD_TYPEDEF(int, driver_failure_eof, (ErlDrvPort));
 WDD_TYPEDEF(void, erl_drv_busy_msgq_limits, (ErlDrvPort, ErlDrvSizeT *, ErlDrvSizeT *));
 WDD_TYPEDEF(int, driver_select, (ErlDrvPort, ErlDrvEvent, int, int));
-WDD_TYPEDEF(int, driver_event, (ErlDrvPort, ErlDrvEvent,ErlDrvEventData));
 WDD_TYPEDEF(int, driver_output, (ErlDrvPort, char *, ErlDrvSizeT));
 WDD_TYPEDEF(int, driver_output2, (ErlDrvPort, char *, ErlDrvSizeT ,char *, ErlDrvSizeT));
 WDD_TYPEDEF(int, driver_output_binary, (ErlDrvPort, char *, ErlDrvSizeT, ErlDrvBinary*, ErlDrvSizeT, ErlDrvSizeT));
@@ -162,7 +161,7 @@ typedef struct {
     WDD_FTYPE(driver_failure_eof) *driver_failure_eof;
     WDD_FTYPE(erl_drv_busy_msgq_limits) *erl_drv_busy_msgq_limits;
     WDD_FTYPE(driver_select) *driver_select;
-    WDD_FTYPE(driver_event) *driver_event;
+    void *REMOVED_driver_event;
     WDD_FTYPE(driver_output) *driver_output;
     WDD_FTYPE(driver_output2) *driver_output2;
     WDD_FTYPE(driver_output_binary) *driver_output_binary;
@@ -276,7 +275,6 @@ extern TWinDynDriverCallbacks WinDynDriverCallbacks;
 #define driver_failure_eof (WinDynDriverCallbacks.driver_failure_eof)
 #define erl_drv_busy_msgq_limits (WinDynDriverCallbacks.erl_drv_busy_msgq_limits)
 #define driver_select (WinDynDriverCallbacks.driver_select)
-#define driver_event (WinDynDriverCallbacks.driver_event)
 #define driver_output (WinDynDriverCallbacks.driver_output)
 #define driver_output2 (WinDynDriverCallbacks.driver_output2)
 #define driver_output_binary (WinDynDriverCallbacks.driver_output_binary)
@@ -414,7 +412,7 @@ do {				                        \
 ((W).driver_failure_eof) = driver_failure_eof;		\
 ((W).erl_drv_busy_msgq_limits) = erl_drv_busy_msgq_limits;\
 ((W).driver_select) = driver_select;			\
-((W).driver_event) = driver_event;			\
+((W).REMOVED_driver_event) = NULL;			\
 ((W).driver_output) = driver_output;			\
 ((W).driver_output2) = driver_output2;			\
 ((W).driver_output_binary) = driver_output_binary;	\
diff --git a/erts/emulator/sys/win32/erl_win_sys.h b/erts/emulator/sys/win32/erl_win_sys.h
index 78005aada9..1f53452d17 100644
--- a/erts/emulator/sys/win32/erl_win_sys.h
+++ b/erts/emulator/sys/win32/erl_win_sys.h
@@ -311,10 +311,8 @@ typedef long ssize_t;
 #endif
 
 /* Threads */
-#ifdef USE_THREADS
 int init_async(int);
 int exit_async(void);
-#endif
 
 #define ERTS_HAVE_TRY_CATCH 1
 
diff --git a/erts/emulator/sys/win32/sys.c b/erts/emulator/sys/win32/sys.c
index 15c59109b1..a1c630d68a 100644
--- a/erts/emulator/sys/win32/sys.c
+++ b/erts/emulator/sys/win32/sys.c
@@ -77,14 +77,13 @@ static int create_pipe(LPHANDLE, LPHANDLE, BOOL, BOOL);
 static int application_type(const wchar_t* originalName, wchar_t fullPath[MAX_PATH],
 			   BOOL search_in_path, BOOL handle_quotes,
 			   int *error_return);
+static void *build_env_block(const erts_osenv_t *env);
 
 HANDLE erts_service_event;
 
-#ifdef ERTS_SMP
-static erts_smp_tsd_key_t win32_errstr_key;
-#endif
+static erts_tsd_key_t win32_errstr_key;
 
-static erts_smp_atomic_t pipe_creation_counter;
+static erts_atomic_t pipe_creation_counter;
 
 /* Results from application_type(_w) is one of */
 #define APPL_NONE 0
@@ -94,10 +93,8 @@ static erts_smp_atomic_t pipe_creation_counter;
 
 static int driver_write(long, HANDLE, byte*, int);
 static int create_file_thread(struct async_io* aio, int mode);
-#ifdef ERTS_SMP
 static void close_active_handle(DriverData *, HANDLE handle);
 static DWORD WINAPI threaded_handle_closer(LPVOID param);
-#endif
 static DWORD WINAPI threaded_reader(LPVOID param);
 static DWORD WINAPI threaded_writer(LPVOID param);
 static DWORD WINAPI threaded_exiter(LPVOID param);
@@ -136,7 +133,7 @@ static OSVERSIONINFO int_os_version;	/* Version information for Win32. */
     Disabled the use of CancelIoEx as its been seen to cause problem with some
     drivers. Not sure what to blame; faulty drivers or some form of invalid use.
 */
-#if defined(ERTS_SMP) && defined(USE_CANCELIOEX)
+#if defined(USE_CANCELIOEX)
 static BOOL (WINAPI *fpCancelIoEx)(HANDLE,LPOVERLAPPED);
 #endif
 
@@ -145,7 +142,7 @@ static BOOL (WINAPI *fpCancelIoEx)(HANDLE,LPOVERLAPPED);
    - call erl_start() to parse arguments and do other init
 */
 
-static erts_smp_atomic_t sys_misc_mem_sz;
+static erts_atomic_t sys_misc_mem_sz;
 
 HMODULE beam_module = NULL;
 
@@ -196,7 +193,7 @@ Uint
 erts_sys_misc_mem_sz(void)
 {
     Uint res = (Uint) erts_check_io_size();
-    res += (Uint) erts_smp_atomic_read_mb(&sys_misc_mem_sz);
+    res += (Uint) erts_atomic_read_mb(&sys_misc_mem_sz);
     return res;
 }
 
@@ -450,9 +447,7 @@ typedef struct async_io {
 				 * the console for Windows NT).
 				 */
   HANDLE fd;			/* Handle for file or pipe. */
-#ifdef ERTS_SMP
   int async_io_active;          /* if true, a close of the file will signal the event in ov */
-#endif
   OVERLAPPED ov;		/* Control structure for overlapped reading.
 				 * When overlapped reading is simulated with
 				 * a thread, the fields are used as follows:
@@ -665,7 +660,7 @@ new_driver_data(ErlDrvPort port_num, int packet_bytes, int wait_objs_required, i
     dp->inbuf = DRV_BUF_ALLOC(dp->inBufSize);
     if (dp->inbuf == NULL)
 	goto buf_alloc_error;
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, dp->inBufSize);
+    erts_atomic_add_nob(&sys_misc_mem_sz, dp->inBufSize);
     dp->outBufSize = 0;
     dp->outbuf = NULL;
     dp->port_num = port_num;
@@ -691,7 +686,6 @@ buf_alloc_error:
 static void
 release_driver_data(DriverData* dp)
 {
-#ifdef ERTS_SMP
 #ifdef USE_CANCELIOEX
     if (fpCancelIoEx != NULL) {
 	if (dp->in.thread == (HANDLE) -1 && dp->in.fd != INVALID_HANDLE_VALUE) {
@@ -734,18 +728,10 @@ release_driver_data(DriverData* dp)
 	    DEBUGF(("...done\n"));
 	}
     }
-#else
-	if (dp->in.thread == (HANDLE) -1 && dp->in.fd != INVALID_HANDLE_VALUE) {
-	     CancelIo(dp->in.fd); 
-	}
-	if (dp->out.thread == (HANDLE) -1 && dp->out.fd != INVALID_HANDLE_VALUE) {
-	    CancelIo(dp->out.fd); 
-	}
-#endif
 
     if (dp->inbuf != NULL) {
-	ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= dp->inBufSize);
-	erts_smp_atomic_add_nob(&sys_misc_mem_sz, -1*dp->inBufSize);
+	ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= dp->inBufSize);
+	erts_atomic_add_nob(&sys_misc_mem_sz, -1*dp->inBufSize);
 	DRV_BUF_FREE(dp->inbuf);
 	dp->inBufSize = 0;
 	dp->inbuf = NULL;
@@ -753,8 +739,8 @@ release_driver_data(DriverData* dp)
     ASSERT(dp->inBufSize == 0);
 
     if (dp->outbuf != NULL) {
-	ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
-	erts_smp_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
+	ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
+	erts_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
 	DRV_BUF_FREE(dp->outbuf);
 	dp->outBufSize = 0;
 	dp->outbuf = NULL;
@@ -777,7 +763,6 @@ release_driver_data(DriverData* dp)
     unrefer_driver_data(dp);
 }
 
-#ifdef ERTS_SMP
 
 struct handles_to_be_closed {
     HANDLE handles[MAXIMUM_WAIT_OBJECTS];
@@ -870,7 +855,6 @@ threaded_handle_closer(LPVOID param)
     DEBUGF(("threaded_handle_closer %p terminating\r\n", htbc));
     return 0;
 }
-#endif /* ERTS_SMP */
 
 /*
  * Stores input and output file descriptors in the DriverData structure,
@@ -946,9 +930,7 @@ init_async_io(DriverData *dp, AsyncIo* aio, int use_threads)
     aio->flushReplyEvent = NULL;
     aio->pendingError = 0;
     aio->bytesTransferred = 0;
-#ifdef ERTS_SMP
     aio->async_io_active = 0;
-#endif
     aio->ov.hEvent = CreateManualEvent(FALSE);
     if (aio->ov.hEvent == NULL)
 	return -1;
@@ -1029,9 +1011,7 @@ async_read_file(AsyncIo* aio, LPVOID buf, DWORD numToRead)
 	ResetEvent(aio->ov.hEvent);
 	SetEvent(aio->ioAllowed);
     } else {
-#ifdef ERTS_SMP
 	aio->async_io_active = 1; /* Will get 0 when the event actually happened */
-#endif
 	if (ReadFile(aio->fd, buf, numToRead,
 		     &aio->bytesTransferred, &aio->ov)) {
 	    DEBUGF(("async_read_file: ReadFile() suceeded: %d bytes\n",
@@ -1079,16 +1059,12 @@ async_write_file(AsyncIo* aio,		/* Pointer to async control block. */
 	ResetEvent(aio->ov.hEvent);
 	SetEvent(aio->ioAllowed);
     } else {
-#ifdef ERTS_SMP
 	aio->async_io_active = 1; /* Will get 0 when the event actually happened */
-#endif
 	if (WriteFile(aio->fd, buf, numToWrite,
 		      &aio->bytesTransferred, &aio->ov)) {
 	    DEBUGF(("async_write_file: WriteFile() suceeded: %d bytes\n",
 		    aio->bytesTransferred));
-#ifdef ERTS_SMP
 	    aio->async_io_active = 0; /* The event will not be signalled */
-#endif
 	    ResetEvent(aio->ov.hEvent);
 	    return TRUE;
 	} else {
@@ -1190,7 +1166,7 @@ static int
 spawn_init(void)
 {
     int i;
-#if defined(ERTS_SMP) && defined(USE_CANCELIOEX)
+#if defined(USE_CANCELIOEX)
     HMODULE module = GetModuleHandle("kernel32");
     fpCancelIoEx = (BOOL (WINAPI *)(HANDLE,LPOVERLAPPED))
 	((module != NULL) ? GetProcAddress(module,"CancelIoEx") : NULL);
@@ -1215,7 +1191,6 @@ spawn_start(ErlDrvPort port_num, char* utf8_name, SysDriverOpts* opts)
     int ok;
     int neededSelects = 0;
     SECURITY_ATTRIBUTES sa = {sizeof(SECURITY_ATTRIBUTES), NULL, TRUE};
-    char* envir = opts->envir;
     int errno_return = -1;
     wchar_t *name;
     int len;
@@ -1290,29 +1265,33 @@ spawn_start(ErlDrvPort port_num, char* utf8_name, SysDriverOpts* opts)
 	name[i] = L'\0';
     }
     DEBUGF(("Spawning \"%S\"\n", name));
-    envir = win_build_environment(envir); /* Always a unicode environment */ 
-    ok = create_child_process(name,
-			      hChildStdin,
-			      hChildStdout,
-			      hChildStderr,
-			      &dp->port_pid,
-			      &pid,
-			      opts->hide_window,
-			      (LPVOID) envir,
-			      (wchar_t *) opts->wd,
-			      opts->spawn_type,
-			      (wchar_t **) opts->argv,
-			      &errno_return);
-    CloseHandle(hChildStdin);
-    CloseHandle(hChildStdout);
-    if (close_child_stderr && hChildStderr != INVALID_HANDLE_VALUE &&
-	hChildStderr != 0) {
-	CloseHandle(hChildStderr);
-    }
-    erts_free(ERTS_ALC_T_TMP, name);
-
-    if (envir != NULL) {
-	erts_free(ERTS_ALC_T_ENVIRONMENT, envir);
+
+    {
+        void *environment_block = build_env_block(&opts->envir);
+
+        ok = create_child_process(name,
+                                  hChildStdin,
+                                  hChildStdout,
+                                  hChildStderr,
+                                  &dp->port_pid,
+                                  &pid,
+                                  opts->hide_window,
+                                  environment_block,
+                                  (wchar_t *) opts->wd,
+                                  opts->spawn_type,
+                                  (wchar_t **) opts->argv,
+                                  &errno_return);
+
+        CloseHandle(hChildStdin);
+        CloseHandle(hChildStdout);
+
+        if (close_child_stderr && hChildStderr != INVALID_HANDLE_VALUE &&
+            hChildStderr != 0) {
+            CloseHandle(hChildStderr);
+        }
+
+        erts_free(ERTS_ALC_T_TMP, environment_block);
+        erts_free(ERTS_ALC_T_TMP, name);
     }
 
     if (!ok) {
@@ -1363,6 +1342,41 @@ spawn_start(ErlDrvPort port_num, char* utf8_name, SysDriverOpts* opts)
     return retval;
 }
 
+struct __build_env_state {
+    WCHAR *next_variable;
+};
+
+static void build_env_foreach(void *_state, const erts_osenv_data_t *key,
+                              const erts_osenv_data_t *value)
+{
+    struct __build_env_state *state = (struct __build_env_state*)(_state);
+
+    sys_memcpy(state->next_variable, key->data, key->length);
+    state->next_variable += (int)key->length / sizeof(WCHAR);
+    *state->next_variable++ = L'=';
+
+    sys_memcpy(state->next_variable, value->data, value->length);
+    state->next_variable += (int)value->length / sizeof(WCHAR);
+    *state->next_variable++ = L'\0';
+}
+
+/* Builds an environment block suitable for CreateProcessW. */
+static void *build_env_block(const erts_osenv_t *env) {
+    struct __build_env_state build_state;
+    WCHAR *env_block;
+
+    env_block = erts_alloc(ERTS_ALC_T_TMP, env->content_size +
+        (env->variable_count * sizeof(L"=\0") + sizeof(L'\0')));
+
+    build_state.next_variable = env_block;
+
+    erts_osenv_foreach_native(env, &build_state, build_env_foreach);
+
+    (*build_state.next_variable) = L'\0';
+
+    return env_block;
+}
+
 static int
 create_file_thread(AsyncIo* aio, int mode)
 {
@@ -1762,7 +1776,7 @@ static int create_pipe(HANDLE *phRead, HANDLE *phWrite, BOOL inheritRead, BOOL o
      * Otherwise, create named pipes.
      */
 
-    calls = (UWord) erts_smp_atomic_inc_read_nob(&pipe_creation_counter);
+    calls = (UWord) erts_atomic_inc_read_nob(&pipe_creation_counter);
     erts_snprintf(pipe_name, sizeof(pipe_name),
 		  "\\\\.\\pipe\\erlang44_%d_%bpu", getpid(), calls);
 
@@ -2447,7 +2461,7 @@ output(ErlDrvData drv_data, char* buf, ErlDrvSizeT len)
     }
 
     dp->outBufSize = pb+len;
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, dp->outBufSize);
+    erts_atomic_add_nob(&sys_misc_mem_sz, dp->outBufSize);
 
     /*
      * Store header bytes (if any).
@@ -2476,8 +2490,8 @@ output(ErlDrvData drv_data, char* buf, ErlDrvSizeT len)
     } else {
 	dp->out.ov.Offset += pb+len; /* For vanilla driver. */
 	/* XXX OffsetHigh should be changed too. */
-	ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
-	erts_smp_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
+	ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
+	erts_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
 	DRV_BUF_FREE(dp->outbuf);
 	dp->outBufSize = 0;
 	dp->outbuf = NULL;
@@ -2511,11 +2525,9 @@ ready_input(ErlDrvData drv_data, ErlDrvEvent ready_event)
     int pb;
 
     pb = dp->packet_bytes;
-#ifdef ERTS_SMP
     if(dp->in.thread == (HANDLE) -1) {
 	dp->in.async_io_active = 0;
     }
-#endif
     DEBUGF(("ready_input: dp %p, event 0x%x\n", dp, ready_event));
 
     /*
@@ -2590,8 +2602,8 @@ ready_input(ErlDrvData drv_data, ErlDrvEvent ready_event)
 			    error = ERROR_NOT_ENOUGH_MEMORY;
 			    break; /* Break out of loop into error handler. */
 			}
-			ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= dp->inBufSize);
-			erts_smp_atomic_add_nob(&sys_misc_mem_sz,
+			ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= dp->inBufSize);
+			erts_atomic_add_nob(&sys_misc_mem_sz,
 						dp->totalNeeded - dp->inBufSize);
 			dp->inBufSize = dp->totalNeeded;
 			dp->inbuf = new_buf;
@@ -2680,11 +2692,9 @@ ready_output(ErlDrvData drv_data, ErlDrvEvent ready_event)
     DriverData *dp = (DriverData *) drv_data;
     int error;
 
-#ifdef ERTS_SMP
     if(dp->out.thread == (HANDLE) -1) {
 	dp->out.async_io_active = 0;
     }
-#endif
     DEBUGF(("ready_output(%p, 0x%x)\n", drv_data, ready_event));
     set_busy_port(dp->port_num, 0);
     if (!(dp->outbuf)) {
@@ -2692,8 +2702,8 @@ ready_output(ErlDrvData drv_data, ErlDrvEvent ready_event)
 	   write... */
 	return;
     }
-    ASSERT(erts_smp_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
+    ASSERT(erts_atomic_read_nob(&sys_misc_mem_sz) >= dp->outBufSize);
+    erts_atomic_add_nob(&sys_misc_mem_sz, -1*dp->outBufSize);
     DRV_BUF_FREE(dp->outbuf);
     dp->outBufSize = 0;
     dp->outbuf = NULL;
@@ -2743,7 +2753,6 @@ sys_init_io(void)
     max_files = 2*erts_ptab_max(&erts_port);
 }
 
-#ifdef ERTS_SMP
 void
 erts_sys_main_thread(void)
 {
@@ -2756,7 +2765,6 @@ erts_sys_main_thread(void)
 	WaitForSingleObject(dummy, INFINITE);
     }
 }
-#endif
 
 void erts_sys_alloc_init(void)
 {
@@ -2843,7 +2851,7 @@ Preload* sys_preloaded(void)
 			   (num_preloaded+1)*sizeof(Preload));
     res_name = erts_alloc(ERTS_ALC_T_PRELOADED,
 			  (num_preloaded+1)*sizeof(unsigned));
-    erts_smp_atomic_add_nob(&sys_misc_mem_sz,
+    erts_atomic_add_nob(&sys_misc_mem_sz,
 			    (num_preloaded+1)*sizeof(Preload)
 			    + (num_preloaded+1)*sizeof(unsigned));
     for (i = 0; i < num_preloaded; i++) {
@@ -2856,7 +2864,7 @@ Preload* sys_preloaded(void)
 	n = GETWORD(data);
 	data += 2;
 	preloaded[i].name = erts_alloc(ERTS_ALC_T_PRELOADED, n+1);
-	erts_smp_atomic_add_nob(&sys_misc_mem_sz, n+1);
+	erts_atomic_add_nob(&sys_misc_mem_sz, n+1);
 	sys_memcpy(preloaded[i].name, data, n);
 	preloaded[i].name[n] = '\0';
 	data += n;
@@ -2938,11 +2946,7 @@ sys_get_key(int fd)
 
 char* win32_errorstr(int error)
 {
-#ifdef SMP
-  LPTSTR lpBufPtr = erts_smp_tsd_get(win32_errstr_key);
-#else
-  static LPTSTR lpBufPtr = NULL;
-#endif
+  LPTSTR lpBufPtr = erts_tsd_get(win32_errstr_key);
   if (lpBufPtr) {
       LocalFree(lpBufPtr);
   }
@@ -2956,9 +2960,7 @@ char* win32_errorstr(int error)
 		0,
 		NULL);
   SetLastError(error);
-#ifdef ERTS_SMP
-  erts_smp_tsd_set(win32_errstr_key,lpBufPtr);
-#endif
+  erts_tsd_set(win32_errstr_key,lpBufPtr);
   return lpBufPtr;
 }
 
@@ -3131,7 +3133,6 @@ check_supported_os_version(void)
 #endif
 }
 
-#ifdef USE_THREADS
 
 typedef struct {
     int sched_bind_data;
@@ -3176,19 +3177,15 @@ thr_create_prepare_child(void *vtcdp)
     erts_sched_bind_atthrcreate_child(tcdp->sched_bind_data);
 }
 
-#endif /* USE_THREADS */
 
 void
 erts_sys_pre_init(void)
 {
-#ifdef USE_THREADS
     erts_thr_init_data_t eid = ERTS_THR_INIT_DATA_DEF_INITER;
-#endif
     int_os_version.dwOSVersionInfoSize = sizeof(OSVERSIONINFO);
     GetVersionEx(&int_os_version);
     check_supported_os_version();
 
-#ifdef USE_THREADS
     eid.thread_create_child_func = thr_create_prepare_child;
     /* Before creation in parent */
     eid.thread_create_prepare_func = thr_create_prepare;
@@ -3209,11 +3206,10 @@ erts_sys_pre_init(void)
     erts_lc_init();
 #endif
 
-#endif /* USE_THREADS */
 
     erts_init_sys_time_sup();
 
-    erts_smp_atomic_init_nob(&sys_misc_mem_sz, 0);
+    erts_atomic_init_nob(&sys_misc_mem_sz, 0);
 }
 
 void noinherit_std_handle(DWORD type)
@@ -3233,11 +3229,9 @@ void erl_sys_init(void)
     noinherit_std_handle(STD_INPUT_HANDLE);
     noinherit_std_handle(STD_ERROR_HANDLE);
 
-#ifdef ERTS_SMP
-    erts_smp_tsd_key_create(&win32_errstr_key,"win32_errstr_key");
+    erts_tsd_key_create(&win32_errstr_key,"win32_errstr_key");
     InitializeCriticalSection(&htbc_lock);
-#endif
-    erts_smp_atomic_init_nob(&pipe_creation_counter,0);
+    erts_atomic_init_nob(&pipe_creation_counter,0);
     /*
      * Test if we have named pipes or not.
      */
@@ -3280,42 +3274,16 @@ void erl_sys_init(void)
 	SetStdHandle(STD_ERROR_HANDLE, GetStdHandle(STD_OUTPUT_HANDLE));
     }
     erts_sys_init_float();
-    erts_init_check_io();
 
     /* Suppress windows error message popups */
     SetErrorMode(SetErrorMode(0) |
 		 SEM_FAILCRITICALERRORS | SEM_NOOPENFILEERRORBOX); 
 }
+void erts_poll_late_init(void);
 
 void
 erl_sys_late_init(void)
 {
     /* do nothing */
+    erts_poll_late_init();
 }
-
-void
-erts_sys_schedule_interrupt(int set)
-{
-    erts_check_io_interrupt(set);
-}
-
-#ifdef ERTS_SMP
-void
-erts_sys_schedule_interrupt_timed(int set, ErtsMonotonicTime timeout_time)
-{
-    erts_check_io_interrupt_timed(set, timeout_time);
-}
-#endif
-
-/*
- * Called from schedule() when it runs out of runnable processes,
- * or when Erlang code has performed INPUT_REDUCTIONS reduction
- * steps. runnable == 0 iff there are no runnable Erlang processes.
- */
-void
-erl_sys_schedule(int runnable)
-{
-    erts_check_io(!runnable);
-    ERTS_SMP_LC_ASSERT(!erts_thr_progress_is_blocking());
-}
-
diff --git a/erts/emulator/sys/win32/sys_env.c b/erts/emulator/sys/win32/sys_env.c
index 8fcee1cbb6..c78161b344 100644
--- a/erts/emulator/sys/win32/sys_env.c
+++ b/erts/emulator/sys/win32/sys_env.c
@@ -1,319 +1,212 @@
-/*

- * %CopyrightBegin%

- *

- * Copyright Ericsson AB 2002-2016. All Rights Reserved.

- *

- * Licensed under the Apache License, Version 2.0 (the "License");

- * you may not use this file except in compliance with the License.

- * You may obtain a copy of the License at

- *

- *     http://www.apache.org/licenses/LICENSE-2.0

- *

- * Unless required by applicable law or agreed to in writing, software

- * distributed under the License is distributed on an "AS IS" BASIS,

- * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.

- * See the License for the specific language governing permissions and

- * limitations under the License.

- *

- * %CopyrightEnd%

- */

-

-#ifdef HAVE_CONFIG_H

-#  include "config.h"

-#endif

-

-#include "sys.h"

-#include "erl_sys_driver.h"

-#include "erl_alloc.h"

-

-static WCHAR *merge_environment(WCHAR *current, WCHAR *add);

-static WCHAR *arg_to_env(WCHAR **arg);

-static WCHAR **env_to_arg(WCHAR *env);

-static WCHAR **find_arg(WCHAR **arg, WCHAR *str);

-static int compare(const void *a, const void *b);

-

-static erts_smp_rwmtx_t environ_rwmtx;

-

-void

-erts_sys_env_init(void)

-{

-    erts_smp_rwmtx_init(&environ_rwmtx, "environ", NIL,

-        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);

-}

-

-int

-erts_sys_putenv_raw(char *key, char *value)

-{

-    int res;

-    erts_smp_rwmtx_rwlock(&environ_rwmtx);

-    res = (SetEnvironmentVariable((LPCTSTR) key,

-				  (LPCTSTR) value) ? 0 : 1);

-    erts_smp_rwmtx_rwunlock(&environ_rwmtx);

-    return res;

-}

-

-int

-erts_sys_putenv(char *key, char *value)

-{

-    int res;

-    WCHAR *wkey = (WCHAR *) key;

-    WCHAR *wvalue = (WCHAR *) value;

-    erts_smp_rwmtx_rwlock(&environ_rwmtx);

-    res = (SetEnvironmentVariableW(wkey,

-				   wvalue) ? 0 : 1);

-    erts_smp_rwmtx_rwunlock(&environ_rwmtx);

-    return res;

-}

-

-int

-erts_sys_getenv(char *key, char *value, size_t *size)

-{

-    size_t req_size = 0;

-    int res = 0;

-    DWORD new_size;

-    WCHAR *wkey = (WCHAR *) key;

-    WCHAR *wvalue = (WCHAR *) value;

-    DWORD wsize = *size / (sizeof(WCHAR) / sizeof(char));

-

-    SetLastError(0);

-    erts_smp_rwmtx_rlock(&environ_rwmtx);

-    new_size = GetEnvironmentVariableW(wkey,

-				       wvalue,

-				      (DWORD) wsize);

-    res = !new_size && GetLastError() == ERROR_ENVVAR_NOT_FOUND ? -1 : 0;

-    erts_smp_rwmtx_runlock(&environ_rwmtx);

-    if (res < 0)

-	return res;

-    res = new_size > wsize ? 1 : 0;

-    *size = new_size * (sizeof(WCHAR) / sizeof(char));

-    return res;

-}

-int

-erts_sys_getenv__(char *key, char *value, size_t *size)

-{

-    size_t req_size = 0;

-    int res = 0;

-    DWORD new_size;

-

-    SetLastError(0);

-    new_size = GetEnvironmentVariable((LPCTSTR) key,

-				      (LPTSTR) value,

-				      (DWORD) *size);

-    res = !new_size && GetLastError() == ERROR_ENVVAR_NOT_FOUND ? -1 : 0;

-    if (res < 0)

-	return res;

-    res = new_size > *size ? 1 : 0;

-    *size = new_size;

-    return res;

-}

-

-int

-erts_sys_getenv_raw(char *key, char *value, size_t *size)

-{

-    int res;

-    erts_smp_rwmtx_rlock(&environ_rwmtx);

-    res = erts_sys_getenv__(key, value, size);

-    erts_smp_rwmtx_runlock(&environ_rwmtx);

-    return res;

-}

-

-void init_getenv_state(GETENV_STATE *state)

-{

-    erts_smp_rwmtx_rlock(&environ_rwmtx);

-    state->environment_strings = GetEnvironmentStringsW();

-    state->next_string = state->environment_strings;

-}

-

-char *getenv_string(GETENV_STATE *state)

-{

-    ERTS_SMP_LC_ASSERT(erts_smp_lc_rwmtx_is_rlocked(&environ_rwmtx));

-    if (state->next_string[0] == L'\0') {

-	return NULL;

-    } else {

-	WCHAR *res = state->next_string;

-	state->next_string += wcslen(res) + 1;

-	return (char *) res;

-    }

-}

-

-void fini_getenv_state(GETENV_STATE *state)

-{

-    FreeEnvironmentStringsW(state->environment_strings);

-    state->environment_strings = state->next_string = NULL;

-    erts_smp_rwmtx_runlock(&environ_rwmtx);

-}

-

-int erts_sys_unsetenv(char *key)

-{

-    int res = 0;

-    WCHAR *wkey = (WCHAR *) key;

-

-    SetLastError(0);

-    erts_smp_rwmtx_rlock(&environ_rwmtx);

-    GetEnvironmentVariableW(wkey,

-                            NULL,

-                            0);

-    if (GetLastError() != ERROR_ENVVAR_NOT_FOUND) {

-        res = (SetEnvironmentVariableW(wkey,

-                                       NULL) ? 0 : 1);

-    }

-    erts_smp_rwmtx_runlock(&environ_rwmtx);

-    return res;

-}

-

-char*

-win_build_environment(char* new_env)

-{

-    if (new_env == NULL) {

-	return NULL;

-    } else {

-	WCHAR *tmp, *merged, *tmp_new;

-

-	tmp_new = (WCHAR *) new_env;

-	

-	erts_smp_rwmtx_rlock(&environ_rwmtx);

-	tmp = GetEnvironmentStringsW();

-	merged = merge_environment(tmp, tmp_new);

-

-	FreeEnvironmentStringsW(tmp);

-	erts_smp_rwmtx_runlock(&environ_rwmtx);

-	return (char *) merged;

-    }

-}

-

-static WCHAR *

-merge_environment(WCHAR *old, WCHAR *add)

-{

-    WCHAR **a_arg = env_to_arg(add);

-    WCHAR **c_arg = env_to_arg(old);

-    WCHAR *ret;

-    int i, j;

-    

-    for(i = 0; c_arg[i] != NULL; ++i)

-	;

-

-    for(j = 0; a_arg[j] != NULL; ++j)

-	;

-

-    c_arg = erts_realloc(ERTS_ALC_T_TMP,

-			 c_arg, (i+j+1) * sizeof(WCHAR *));

-

-    for(j = 0; a_arg[j] != NULL; ++j){

-	WCHAR **tmp;

-	WCHAR *current = a_arg[j];

-	WCHAR *eq_p = wcschr(current,L'=');

-	int unset = (eq_p!=NULL && eq_p[1]==L'\0');

-

-	if ((tmp = find_arg(c_arg, current)) != NULL) {

-	    if (!unset) {

-		*tmp = current;

-	    } else {

-		*tmp = c_arg[--i];

-		c_arg[i] = NULL;

-	    }

-	} else if (!unset) {

-	    c_arg[i++] = current;

-	    c_arg[i] = NULL;

-	}

-    }

-    ret = arg_to_env(c_arg);

-    erts_free(ERTS_ALC_T_TMP, c_arg);

-    erts_free(ERTS_ALC_T_TMP, a_arg);

-    return ret;

-}

-

-static WCHAR**

-find_arg(WCHAR **arg, WCHAR *str)

-{

-    WCHAR *tmp;

-    int len;

-

-    if ((tmp = wcschr(str, L'=')) != NULL) {

-	tmp++;

-	len = tmp - str;

-	while (*arg != NULL){

-	    if (_wcsnicmp(*arg, str, len) == 0){

-		return arg;

-	    }

-	    ++arg;

-	}

-    }

-    return NULL;

-}

-

-static int

-compare(const void *a, const void *b)

-{

-    WCHAR *s1 = *((WCHAR **) a);

-    WCHAR *s2 = *((WCHAR **) b);

-    WCHAR *e1 = wcschr(s1,L'=');

-    WCHAR *e2 = wcschr(s2,L'=');

-    int ret;

-    int len;

-  

-    if(!e1)

-	e1 = s1 + wcslen(s1);

-    if(!e2)

-	e2 = s2 + wcslen(s2);

-  

-    if((e1 - s1) > (e2 - s2))

-	len = (e2 - s2);

-    else

-	len = (e1 - s1);

-  

-    ret = _wcsnicmp(s1,s2,len);

-    if (ret == 0)

-	return ((e1 - s1) - (e2 - s2));

-    else

-	return ret;

-}

-

-static WCHAR**

-env_to_arg(WCHAR *env)

-{

-    WCHAR **ret;

-    WCHAR *tmp;

-    int i;

-    int num_strings = 0;

-

-    for(tmp = env; *tmp != '\0'; tmp += wcslen(tmp)+1) {

-	++num_strings;

-    }

-    ret = erts_alloc(ERTS_ALC_T_TMP, sizeof(WCHAR *) * (num_strings + 1));

-    i = 0;

-    for(tmp = env; *tmp != '\0'; tmp += wcslen(tmp)+1){

-	ret[i++] = tmp;

-    }

-    ret[i] = NULL;

-    return ret;

-}

-

-static WCHAR *

-arg_to_env(WCHAR **arg)

-{

-    WCHAR *block;

-    WCHAR *ptr;

-    int i;

-    int totlen = 1;		/* extra '\0' */

-

-    for(i = 0; arg[i] != NULL; ++i) {

-	totlen += wcslen(arg[i])+1;

-    }

-

-    /* sort the environment vector */

-    qsort(arg, i, sizeof(WCHAR *), &compare);

-

-    if (totlen == 1){

-	block = erts_alloc(ERTS_ALC_T_ENVIRONMENT, 2 * sizeof(WCHAR));

-	block[0] = block[1] = '\0'; 

-    } else {

-	block = erts_alloc(ERTS_ALC_T_ENVIRONMENT, totlen * sizeof(WCHAR));

-	ptr = block;

-	for(i=0; arg[i] != NULL; ++i){

-	    wcscpy(ptr, arg[i]);

-	    ptr += wcslen(ptr)+1;

-	}

-	*ptr = '\0';

-    }

-    return block;

-}

+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2002-2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#include "sys.h"
+#include "erl_sys_driver.h"
+#include "erl_alloc.h"
+
+static erts_osenv_t sysenv_global_env;
+static erts_rwmtx_t sysenv_rwmtx;
+
+static void import_initial_env(void);
+
+void erts_sys_env_init() {
+    erts_rwmtx_init(&sysenv_rwmtx, "environ", NIL,
+        ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
+
+    erts_osenv_init(&sysenv_global_env);
+    import_initial_env();
+}
+
+const erts_osenv_t *erts_sys_rlock_global_osenv() {
+    erts_rwmtx_rlock(&sysenv_rwmtx);
+    return &sysenv_global_env;
+}
+
+erts_osenv_t *erts_sys_rwlock_global_osenv() {
+    erts_rwmtx_rwlock(&sysenv_rwmtx);
+    return &sysenv_global_env;
+}
+
+void erts_sys_runlock_global_osenv() {
+    erts_rwmtx_runlock(&sysenv_rwmtx);
+}
+
+void erts_sys_rwunlock_global_osenv() {
+    erts_rwmtx_rwunlock(&sysenv_rwmtx);
+}
+
+int erts_sys_explicit_host_getenv(char *key, char *value, size_t *size) {
+    size_t new_size = GetEnvironmentVariableA(key, value, (DWORD)*size);
+
+    if(new_size == 0 && GetLastError() == ERROR_ENVVAR_NOT_FOUND) {
+        return 0;
+    } else if(new_size > *size) {
+        return -1;
+    }
+
+    *size = new_size;
+    return 1;
+}
+
+int erts_sys_explicit_8bit_putenv(char *key, char *value) {
+    WCHAR *wide_key, *wide_value;
+    int key_length, value_length;
+    int result;
+
+    /* Note that we do *NOT* honor the filename encoding flags (+fnu/+fnl)
+     * here; the previous implementation used SetEnvironmentVariableA and
+     * things may break if we step away from that. */
+
+    key_length = MultiByteToWideChar(CP_ACP, 0, key, -1, NULL, 0);
+    value_length = MultiByteToWideChar(CP_ACP, 0, value, -1, NULL, 0);
+
+    /* Report "not found" if either string isn't convertible. */
+    if(key_length == 0 || value_length == 0) {
+        return 0;
+    }
+
+    wide_key = erts_alloc(ERTS_ALC_T_TMP, key_length * sizeof(WCHAR));
+    wide_value = erts_alloc(ERTS_ALC_T_TMP, value_length * sizeof(WCHAR));
+
+    MultiByteToWideChar(CP_ACP, 0, key, -1, wide_key, key_length);
+    MultiByteToWideChar(CP_ACP, 0, value, -1, wide_value, value_length);
+
+    {
+        erts_osenv_data_t env_key, env_value;
+        erts_osenv_t *env;
+
+        env = erts_sys_rwlock_global_osenv();
+
+        /* -1 to exclude the NUL terminator. */
+        env_key.length = (key_length - 1) * sizeof(WCHAR);
+        env_key.data = wide_key;
+
+        env_value.length = (value_length - 1) * sizeof(WCHAR);
+        env_value.data = wide_value;
+
+        result = erts_osenv_put_native(env, &env_key, &env_value);
+        erts_sys_rwunlock_global_osenv();
+    }
+
+    erts_free(ERTS_ALC_T_TMP, wide_key);
+    erts_free(ERTS_ALC_T_TMP, wide_value);
+
+    return result;
+}
+
+int erts_sys_explicit_8bit_getenv(char *key, char *value, size_t *size) {
+    erts_osenv_data_t env_key, env_value;
+    int key_length, value_length, result;
+    WCHAR *wide_key, *wide_value;
+
+    key_length = MultiByteToWideChar(CP_ACP, 0, key, -1, NULL, 0);
+
+    /* Report "not found" if the string isn't convertible. */
+    if(key_length == 0) {
+        return 0;
+    }
+
+    wide_key = erts_alloc(ERTS_ALC_T_TMP, key_length * sizeof(WCHAR));
+    MultiByteToWideChar(CP_ACP, 0, key, -1, wide_key, key_length);
+
+    /* We assume that the worst possible size is twice the output buffer width,
+     * as we could theoretically be on a code page that requires surrogates. */
+    value_length = (*size) * 2;
+    wide_value = erts_alloc(ERTS_ALC_T_TMP, value_length * sizeof(WCHAR));
+
+    {
+        const erts_osenv_t *env = erts_sys_rlock_global_osenv();
+
+        /* -1 to exclude the NUL terminator. */
+        env_key.length = (key_length - 1) * sizeof(WCHAR);
+        env_key.data = wide_key;
+
+        env_value.length = value_length * sizeof(WCHAR);
+        env_value.data = wide_value;
+
+        result = erts_osenv_get_native(env, &env_key, &env_value);
+        erts_sys_runlock_global_osenv();
+    }
+
+    if(result == 1 && env_value.length > 0) {
+        /* This function doesn't NUL-terminate if the provided size is >= 0,
+         * so we pass (*size - 1) to reserve space for it and then do it
+         * manually. */
+        *size = WideCharToMultiByte(CP_ACP, 0, env_value.data,
+            env_value.length / sizeof(WCHAR), value, *size - 1, NULL, NULL);
+
+        if(*size == 0) {
+            if(GetLastError() == ERROR_INSUFFICIENT_BUFFER) {
+                result = -1;
+            } else {
+                result = 0;
+            }
+        }
+    } else {
+        *size = 0;
+    }
+
+    if(*size > 0) {
+        value[*size] = '\0';
+    }
+
+    erts_free(ERTS_ALC_T_TMP, wide_key);
+    erts_free(ERTS_ALC_T_TMP, wide_value);
+
+    return result;
+}
+
+static void import_initial_env(void) {
+    WCHAR *environment_block, *current_variable;
+
+    environment_block = GetEnvironmentStringsW();
+    current_variable = environment_block;
+
+    while(wcslen(current_variable) > 0) {
+        WCHAR *separator_index = wcschr(current_variable, L'=');
+
+        /* We tolerate environment variables starting with '=' as the per-drive
+         * working directories are stored this way. */
+        if(separator_index == current_variable) {
+            separator_index = wcschr(separator_index + 1, L'=');
+        }
+
+        if(separator_index != NULL && separator_index != current_variable) {
+            erts_osenv_data_t env_key, env_value;
+
+            env_key.length = (separator_index - current_variable) * sizeof(WCHAR);
+            env_key.data = current_variable;
+
+            env_value.length = (wcslen(separator_index) - 1) * sizeof(WCHAR);
+            env_value.data = separator_index + 1;
+
+            erts_osenv_put_native(&sysenv_global_env, &env_key, &env_value);
+        }
+
+        current_variable += wcslen(current_variable) + 1;
+    }
+
+    FreeEnvironmentStringsW(environment_block);
+}
diff --git a/erts/emulator/sys/win32/sys_interrupt.c b/erts/emulator/sys/win32/sys_interrupt.c
index df838960eb..02aa50500f 100644
--- a/erts/emulator/sys/win32/sys_interrupt.c
+++ b/erts/emulator/sys/win32/sys_interrupt.c
@@ -35,17 +35,11 @@
 #  define WIN_SYS_INLINE __forceinline
 #endif
 
-#ifdef ERTS_SMP
-erts_smp_atomic32_t erts_break_requested;
+erts_atomic32_t erts_break_requested;
 #define ERTS_SET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 1)
 #define ERTS_UNSET_BREAK_REQUESTED \
-  erts_smp_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
-#else
-volatile int erts_break_requested = 0;
-#define ERTS_SET_BREAK_REQUESTED (erts_break_requested = 1)
-#define ERTS_UNSET_BREAK_REQUESTED (erts_break_requested = 0)
-#endif
+  erts_atomic32_set_nob(&erts_break_requested, (erts_aint32_t) 0)
 
 extern int nohup;
 HANDLE erts_sys_break_event = NULL;
@@ -57,14 +51,14 @@ void erts_do_break_handling(void)
      * therefore, make sure that all threads but this one are blocked before
      * proceeding!
      */
-    erts_smp_thr_progress_block();
+    erts_thr_progress_block();
     /* call the break handling function, reset the flag */
     do_break();
 
     ResetEvent(erts_sys_break_event);
     ERTS_UNSET_BREAK_REQUESTED;
 
-    erts_smp_thr_progress_unblock();
+    erts_thr_progress_unblock();
 }
 
 
diff --git a/erts/emulator/sys/win32/sys_time.c b/erts/emulator/sys/win32/sys_time.c
index 88131aaa6a..25c2ad385c 100644
--- a/erts/emulator/sys/win32/sys_time.c
+++ b/erts/emulator/sys/win32/sys_time.c
@@ -95,7 +95,7 @@ struct sys_time_internal_state_read_mostly__ {
 };
 
 struct sys_time_internal_state_write_freq__ {
-    erts_smp_mtx_t mtime_mtx;
+    erts_mtx_t mtime_mtx;
     ULONGLONG wrap;
     ULONGLONG last_tick_count;
 };
@@ -187,8 +187,6 @@ os_monotonic_time_gtc32(void)
 {
     ErtsMonotonicTime mtime;
     Uint32 ticks = (Uint32) GetTickCount();
-    ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
-				      ticks);
     mtime = ERTS_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
 					  ticks);
     mtime <<= ERTS_GET_TICK_COUNT_TIME_UNIT_SHIFT;
@@ -205,8 +203,6 @@ os_times_gtc32(ErtsMonotonicTime *mtimep, ErtsSystemTime *stimep)
     ticks = (Uint32) GetTickCount();
     GetSystemTime(&st);
 
-    ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
-				      ticks);
     mtime = ERTS_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
 					  ticks);
     mtime <<= ERTS_GET_TICK_COUNT_TIME_UNIT_SHIFT;
@@ -265,8 +261,6 @@ sys_hrtime_gtc32(void)
 {
     ErtsSysHrTime time;
     Uint32 ticks = (Uint32) GetTickCount();
-    ERTS_CHK_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
-				      tick_count);
     time = (ErtsSysHrTime) ERTS_EXTEND_OS_MONOTONIC_TIME(&internal_state.wr.m.os_mtime_xtnd,
 							 ticks);
     time *= (ErtsSysHrTime) (1000 * 1000);
@@ -300,7 +294,7 @@ sys_init_time(ErtsSysInitTimeResult *init_resp)
     module = GetModuleHandle(kernel_dll_name);
     if (!module) {
     get_tick_count:
-        erts_smp_mtx_init(&internal_state.w.f.mtime_mtx, "os_monotonic_time", NIL,
+        erts_mtx_init(&internal_state.w.f.mtime_mtx, "os_monotonic_time", NIL,
             ERTS_LOCK_FLAGS_PROPERTY_STATIC | ERTS_LOCK_FLAGS_CATEGORY_GENERIC);
 	internal_state.w.f.wrap = 0;
 	internal_state.w.f.last_tick_count = 0;
diff --git a/erts/emulator/test/Makefile b/erts/emulator/test/Makefile
index b17170c8b8..bf00de2204 100644
--- a/erts/emulator/test/Makefile
+++ b/erts/emulator/test/Makefile
@@ -57,6 +57,7 @@ MODULES= \
 	dirty_nif_SUITE \
 	distribution_SUITE \
 	driver_SUITE \
+	dump_SUITE \
 	efile_SUITE \
 	erts_debug_SUITE \
 	estone_SUITE \
diff --git a/erts/emulator/test/alloc_SUITE.erl b/erts/emulator/test/alloc_SUITE.erl
index 3d29776530..88ff2a7a92 100644
--- a/erts/emulator/test/alloc_SUITE.erl
+++ b/erts/emulator/test/alloc_SUITE.erl
@@ -67,16 +67,11 @@ mseg_clear_cache(Cfg) -> drv_case(Cfg).
 cpool(Cfg) -> drv_case(Cfg).
 
 migration(Cfg) ->
-    case erlang:system_info(smp_support) of
-	true ->
-            %% Enable test_alloc.
-            %% Disable driver_alloc to avoid recursive alloc_util calls
-            %% through enif_mutex_create() in my_creating_mbc().
-	    drv_case(Cfg, concurrent, "+MZe true +MRe false"),
-	    drv_case(Cfg, concurrent, "+MZe true +MRe false +MZas ageffcbf");
-	false ->
-	    {skipped, "No smp"}
-    end.
+    %% Enable test_alloc.
+    %% Disable driver_alloc to avoid recursive alloc_util calls
+    %% through enif_mutex_create() in my_creating_mbc().
+    drv_case(Cfg, concurrent, "+MZe true +MRe false"),
+    drv_case(Cfg, concurrent, "+MZe true +MRe false +MZas ageffcbf").
 
 erts_mmap(Config) when is_list(Config) ->
     case {os:type(), mmsc_flags()} of
@@ -100,10 +95,11 @@ mmsc_flags() ->
 mmsc_flags(Env) ->
     case os:getenv(Env) of
 	false -> false;
-	V -> case string:str(V, "+MMsc") of
-		 0 -> false;
-		 P -> Env ++ "=" ++ string:substr(V, P)
-	     end
+	V ->
+            case string:find(V, "+MMsc") of
+                nomatch -> false;
+                SubStr -> Env ++ "=" ++ SubStr
+            end
     end.
 
 erts_mmap_do(Config, SCO, SCRPM, SCRFSD) ->
diff --git a/erts/emulator/test/beam_SUITE.erl b/erts/emulator/test/beam_SUITE.erl
index 6a54fa87e0..bdf8f6c34e 100644
--- a/erts/emulator/test/beam_SUITE.erl
+++ b/erts/emulator/test/beam_SUITE.erl
@@ -113,20 +113,41 @@ packed_registers(Config) when is_list(Config) ->
 		   VarName = list_to_atom("M"++integer_to_list(V)),
 		   merl:var(VarName)
 	       end || V <- Seq],
+    MoreNewVars = [begin
+                       VarName = list_to_atom("MM"++integer_to_list(V)),
+                       merl:var(VarName)
+                   end || V <- Seq],
+    TupleEls = [?Q("id(_@Value@)") || {_,Value} <- S0],
     S = [?Q("_@Var = id(_@Value@)") || {Var,Value} <- S0],
     Code = ?Q(["-module('@Mod@').\n"
 	       "-export([f/0]).\n"
 	       "f() ->\n"
+               "Tuple = id({_@TupleEls}),\n"
+               "{_@MoreNewVars} = Tuple,\n"
 	       "_@S,\n"
 	       "_ = id(0),\n"
 	       "L = [_@Vars],\n"
 	       "_ = id(1),\n"
 	       "[_@NewVars] = L,\n"		%Test get_list/3.
 	       "_ = id(2),\n"
-	       "id([_@Vars,_@NewVars]).\n"
+	       "id([_@Vars,_@NewVars,_@MoreNewVars]).\n"
 	       "id(I) -> I.\n"]),
     merl:compile_and_load(Code),
-    CombinedSeq = Seq ++ Seq,
+
+    %% Optionally print the generated code.
+    PrintCode = false,                          %Change to true to print code.
+
+    case PrintCode of
+        false ->
+            ok;
+        true ->
+            merl:print(Code),
+            erts_debug:df(Mod),
+            {ok,Dis} = file:read_file(atom_to_list(Mod)++".dis"),
+            io:put_chars(Dis)
+    end,
+
+    CombinedSeq = Seq ++ Seq ++ Seq,
     CombinedSeq = Mod:f(),
 
     %% Clean up.
diff --git a/erts/emulator/test/beam_literals_SUITE.erl b/erts/emulator/test/beam_literals_SUITE.erl
index 09761263e2..b447ca0210 100644
--- a/erts/emulator/test/beam_literals_SUITE.erl
+++ b/erts/emulator/test/beam_literals_SUITE.erl
@@ -248,35 +248,58 @@ literal_type_tests(Config) when is_list(Config) ->
     ok.
 
 make_test([{is_function=T,L}|Ts]) ->
-    [test(T, L),test(T, 0, L)|make_test(Ts)];
+    [guard_test(T, L),guard_test(T, 0, L),body_test(T, L),body_test(T, 0, L)|make_test(Ts)];
 make_test([{T,L}|Ts]) ->
-    [test(T, L)|make_test(Ts)];
+    [guard_test(T, L),body_test(T, L)|make_test(Ts)];
 make_test([]) -> [].
 
-test(T, L) ->
-    S = lists:flatten(io_lib:format("begin io:format(\"~~p~n\", [{~p,~p}]), if ~w(~w) -> true; true -> false end end. ", [T, L, T, L])),
-    {ok,Toks,_Line} = erl_scan:string(S),
-    {ok,E} = erl_parse:parse_exprs(Toks),
-    {value,Val,_Bs} = erl_eval:exprs(E, []),
+guard_test(_, L) when is_function(L) ->
+    %% Skip guard tests with exports - they are not literals
+    {atom,erl_anno:new(0),true};
+guard_test(T, L) ->
+    S = io_lib:format("begin io:format(\"~~p~n\", [{~p,~p}]), if ~w(~w) -> true; true -> false end end. ", [T, L, T, L]),
+    {Val,Expr} = eval_string(S),
+    Anno = erl_anno:new(0),
+    {match,Anno,{atom,Anno,Val},Expr}.
+
+guard_test(_, _, L) when is_function(L) ->
+    %% Skip guard tests with exports - they are not literals
+    {atom,erl_anno:new(0),true};
+guard_test(T, A, L) ->
+    S = io_lib:format("begin io:format(\"~~p~n\", [{~p,~p,~p}]), if ~w(~w, ~w) -> true; true -> false end end. ", [T,L,A,T,L,A]),
+    {Val,Expr} = eval_string(S),
+    Anno = erl_anno:new(0),
+    {match,Anno,{atom,Anno,Val},Expr}.
+
+body_test(T, L) ->
+    S = io_lib:format("begin io:format(\"~~p~n\", [{~p,~p}]), ~w(~w) end. ", [T,L,T,L]),
+    {Val,Expr} = eval_string(S),
     Anno = erl_anno:new(0),
-    {match,Anno,{atom,Anno,Val},hd(E)}.
+    {match,Anno,{atom,Anno,Val},Expr}.
 
-test(T, A, L) ->
-    S = lists:flatten(io_lib:format("begin io:format(\"~~p~n\", [{~p,~p,~p}]), if ~w(~w, ~w) -> true; true -> false end end. ",
-				    [T,L,A,T,L,A])),
-    {ok,Toks,_Line} = erl_scan:string(S),
+body_test(T, A, L) ->
+    S = io_lib:format("begin io:format(\"~~p~n\", [{~p,~p,~p}]), ~w(~w,~w) end. ", [T,L,A,T,L,A]),
+    {Val,Expr} = eval_string(S),
+    Anno = erl_anno:new(0),
+    {match,Anno,{atom,Anno,Val},Expr}.
+
+eval_string(S) ->
+    {ok,Toks,_Line} = erl_scan:string(lists:flatten(S)),
     {ok,E} = erl_parse:parse_exprs(Toks),
     {value,Val,_Bs} = erl_eval:exprs(E, []),
-    Anno = erl_anno:new(0),
-    {match,Anno,{atom,Anno,Val},hd(E)}.
-    
+    {Val,hd(E)}.
+
 literals() ->
     [42,
      3.14,
      -3,
      32982724987789283473473838474,
      [],
-     xxxx].
+     "abc",
+     <<"abc">>,
+     {},
+     xxxx,
+     fun erlang:erase/0].
 
 type_tests() ->
     [is_boolean,
diff --git a/erts/emulator/test/bif_SUITE.erl b/erts/emulator/test/bif_SUITE.erl
index 04b7f2de15..32bfcd5520 100644
--- a/erts/emulator/test/bif_SUITE.erl
+++ b/erts/emulator/test/bif_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 2005-2016. All Rights Reserved.
+%% Copyright Ericsson AB 2005-2018. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -33,7 +33,10 @@
 	 atom_to_binary/1,min_max/1, erlang_halt/1,
          erl_crash_dump_bytes/1,
 	 is_builtin/1, error_stacktrace/1,
-	 error_stacktrace_during_call_trace/1]).
+	 error_stacktrace_during_call_trace/1,
+         group_leader_prio/1, group_leader_prio_dirty/1,
+         is_process_alive/1,
+         process_info_blast/1]).
 
 suite() ->
     [{ct_hooks,[ts_install_cth]},
@@ -46,7 +49,9 @@ all() ->
      display, display_string, list_to_utf8_atom,
      atom_to_binary, binary_to_atom, binary_to_existing_atom,
      erl_crash_dump_bytes, min_max, erlang_halt, is_builtin,
-     error_stacktrace, error_stacktrace_during_call_trace].
+     error_stacktrace, error_stacktrace_during_call_trace,
+     group_leader_prio, group_leader_prio_dirty,
+     is_process_alive, process_info_blast].
 
 %% Uses erlang:display to test that erts_printf does not do deep recursion
 display(Config) when is_list(Config) ->
@@ -731,10 +736,12 @@ erlang_halt(Config) when is_list(Config) ->
                                  [broken_halt, "Validate correct crash dump"]),
     {ok,_} = wait_until_stable_size(CrashDump,-1),
     {ok, Bin} = file:read_file(CrashDump),
-    case {string:str(binary_to_list(Bin),"\n=end\n"),
-          string:str(binary_to_list(Bin),"\r\n=end\r\n")} of
-        {0,0} -> ct:fail("Could not find end marker in crash dump");
-        _ -> ok
+    case {string:find(Bin, <<"\n=end\n">>),
+          string:find(Bin, <<"\r\n=end\r\n">>)} of
+        {nomatch,nomatch} ->
+            ct:fail("Could not find end marker in crash dump");
+        {_,_} ->
+            ok
     end.
 
 wait_until_stable_size(_File,-10) ->
@@ -779,14 +786,20 @@ is_builtin(_Config) ->
 		       {F,A} <- M:module_info(exports)],
     Exp = ordsets:from_list(Exp0),
 
-    %% erlang:apply/3 is considered to be built-in, but is not
-    %% implemented as other BIFs.
+    %% Built-ins implemented as special instructions.
+    Instructions = [{erlang,apply,2},{erlang,apply,3},{erlang,yield,0}],
 
-    Builtins0 = [{erlang,apply,3}|erlang:system_info(snifs)],
+    Builtins0 = Instructions ++ erlang:system_info(snifs),
     Builtins = ordsets:from_list(Builtins0),
-    NotBuiltin = ordsets:subtract(Exp, Builtins),
-    _ = [true = erlang:is_builtin(M, F, A) || {M,F,A} <- Builtins],
-    _ = [false = erlang:is_builtin(M, F, A) || {M,F,A} <- NotBuiltin],
+
+    Fakes = [{M,F,42} || {M,F,_} <- Instructions],
+    All = ordsets:from_list(Fakes ++ Exp),
+    NotBuiltin = ordsets:subtract(All, Builtins),
+
+    _ = [{true,_} = {erlang:is_builtin(M, F, A),MFA} ||
+            {M,F,A}=MFA <- Builtins],
+    _ = [{false,_} = {erlang:is_builtin(M, F, A),MFA} ||
+            {M,F,A}=MFA <- NotBuiltin],
 
     ok.
 
@@ -817,7 +830,6 @@ error_stacktrace_during_call_trace(Config) when is_list(Config) ->
 	end
     end,
     ok.
-    
 
 error_stacktrace_test() ->
     Types = [apply_const_last, apply_const, apply_last,
@@ -955,9 +967,210 @@ do_error_1(call) ->
     erlang:error(id(oops)).
 
 
+group_leader_prio(Config) when is_list(Config) ->
+    group_leader_prio_test(false).
+
+group_leader_prio_dirty(Config) when is_list(Config) ->
+    group_leader_prio_test(true).
+
+group_leader_prio_test(Dirty) ->
+    %%
+    %% Unfortunately back in the days node local group_leader/2 was not
+    %% implemented as sending an asynchronous signal to the process to change
+    %% group leader for. Instead it has always been synchronously changed, and
+    %% nothing in the documentation have hinted otherwise... Therefore I do not
+    %% dare the change this.
+    %%
+    %% In order to prevent priority inversion, the priority of the receiver of
+    %% the group leader signal is elevated while handling incoming signals if
+    %% the sender has a higher priority than the receiver. This test tests that
+    %% the priority elevation actually works...
+    %%
+    Tester = self(),
+    Init = erlang:whereis(init),
+    GL = erlang:group_leader(),
+    process_flag(priority, max),
+    {TestProcFun, NTestProcs}
+        = case Dirty of
+              false ->
+                  %% These processes will handle all incoming signals
+                  %% by them selves...
+                  {fun () ->
+                           Tester ! {alive, self()},
+                           receive after infinity -> ok end
+                   end,
+                   100};
+              true ->
+                  %% These processes wont handle incoming signals by
+                  %% them selves since they are stuck on dirty schedulers
+                  %% when we try to change group leader. A dirty process
+                  %% signal handler process (system process) will be notified
+                  %% of the need to handle incoming signals for these processes,
+                  %% and will instead handle the signal for these processes...
+                  {fun () ->
+                           %% The following sends the message '{alive, self()}'
+                           %% to Tester once on a dirty io scheduler, then wait
+                           %% there until the process terminates...
+                           erts_debug:dirty_io(alive_waitexiting, Tester)
+                   end,
+                   erlang:system_info(dirty_io_schedulers)}
+          end,
+    TPs = lists:map(fun (_) ->
+                            spawn_opt(TestProcFun,
+                                      [link, {priority, normal}])
+                    end, lists:seq(1, NTestProcs)),
+    lists:foreach(fun (TP) -> receive {alive, TP} -> ok end end, TPs),
+    TLs = lists:map(fun (_) ->
+                            spawn_opt(fun () -> tok_loop() end,
+                                      [link, {priority, high}])
+                    end,
+                    lists:seq(1, 2*erlang:system_info(schedulers))),
+    %% Wait to ensure distribution of high prio processes over schedulers...
+    receive after 1000 -> ok end,
+    %%
+    %% Test that we can get group-leader signals through to normal prio
+    %% processes from a max prio process even though all schedulers are filled
+    %% with executing high prio processes.
+    %%
+    lists:foreach(fun (_) ->
+                          lists:foreach(fun (TP) ->
+                                                erlang:yield(),
+                                                %% whitebox -- Enqueue some signals on it
+                                                %% preventing us from hogging its main lock
+                                                %% and set group-leader directly....
+                                                erlang:demonitor(erlang:monitor(process, TP)),
+                                                true = erlang:group_leader(Init, TP),
+                                                {group_leader, Init} = process_info(TP, group_leader),
+                                                erlang:demonitor(erlang:monitor(process, TP)),
+                                                true = erlang:group_leader(GL, TP),
+                                                {group_leader, GL} = process_info(TP, group_leader)
+                                        end,
+                                        TPs)
+                  end,
+                  lists:seq(1,100)),
+    %%
+    %% Also test when it is exiting...
+    %%
+    lists:foreach(fun (TP) ->
+                          erlang:yield(),
+                          M = erlang:monitor(process, TP),
+                          unlink(TP),
+                          exit(TP, bang),
+                          badarg = try
+                                       true = erlang:group_leader(Init, TP)
+                                   catch
+                                       error : What -> What
+                                   end,
+                          receive
+                              {'DOWN', M, process, TP, Reason} ->
+                                  bang = Reason
+                          end
+                  end,
+                  TPs),
+    lists:foreach(fun (TL) ->
+                          M = erlang:monitor(process, TL),
+                          unlink(TL),
+                          exit(TL, bang),
+                          receive
+                              {'DOWN', M, process, TL, Reason} ->
+                                  bang = Reason
+                          end
+                  end,
+                  TLs),
+    ok.
+
+is_process_alive(Config) when is_list(Config) ->
+    process_flag(priority, max),
+    Ps = lists:map(fun (_) ->
+                           spawn_opt(fun () -> tok_loop() end,
+                                     [{priority, high}, link])
+                   end,
+                   lists:seq(1, 2*erlang:system_info(schedulers))),
+    receive after 1000 -> ok end, %% Wait for load to spread
+    lists:foreach(fun (P) ->
+                          %% Ensure that signal order is preserved
+                          %% and that we are not starved due to
+                          %% priority inversion
+                          true = erlang:is_process_alive(P),
+                          unlink(P),
+                          true = erlang:is_process_alive(P),
+                          exit(P, kill),
+                          false = erlang:is_process_alive(P)
+                  end,
+                  Ps),
+    ok.
 
+process_info_blast(Config) when is_list(Config) ->
+    Tester = self(),
+    NoAttackers = 1000,
+    NoAL = lists:seq(1, NoAttackers),
+    Consume = make_ref(),
+    Victim = spawn_link(fun () ->
+                                receive
+                                    Consume ->
+                                        ok
+                                end,
+                                consume_msgs()
+                        end),
+    AFun = fun () ->
+                   Victim ! hej,
+                   Res = process_info(Victim, message_queue_len),
+                   Tester ! {self(), Res}
+           end,
+    Attackers0 = lists:map(fun (_) ->
+                                   spawn_link(AFun)
+                           end,
+                           NoAL),
+    lists:foreach(fun (A) ->
+                          receive
+                              {A, Res} ->
+                                  case Res of
+                                      {message_queue_len, Len} when Len > 0, Len =< NoAttackers ->
+                                          Len;
+                                      Error ->
+                                          exit({unexpected, Error})
+                                  end
+                          end
+                  end,
+                  Attackers0),
+    Attackers1 = lists:map(fun (_) ->
+                                   spawn_link(AFun)
+                           end,
+                           NoAL),
+    Victim ! Consume,
+    lists:foreach(fun (A) ->
+                          receive
+                              {A, Res} ->
+                                  case Res of
+                                      {message_queue_len, Len} when Len >= 0, Len =< 2*NoAttackers+1 ->
+                                          ok;
+                                      undefined ->
+                                          ok;
+                                      Error ->
+                                          exit({unexpected, Error})
+                                  end
+                          end
+                  end,
+                  Attackers1),
+    KillFun = fun (P) ->
+                      unlink(P),
+                      exit(P, kill),
+                      false = erlang:is_process_alive(P)
+              end,
+    lists:foreach(fun (A) -> KillFun(A) end, Attackers0),
+    lists:foreach(fun (A) -> KillFun(A) end, Attackers1),
+    KillFun(Victim),
+    ok.
 
-%% Helpers
+consume_msgs() ->
+    receive
+        _ ->
+            consume_msgs()
+    after 0 ->
+              ok
+    end.
+                              
+%% helpers
     
 id(I) -> I.
 
@@ -997,3 +1210,11 @@ hostname([$@ | Hostname]) ->
     list_to_atom(Hostname);
 hostname([_C | Cs]) ->
     hostname(Cs).
+
+tok_loop() ->
+    tok_loop(hej).
+
+tok_loop(hej) ->
+    tok_loop(hopp);
+tok_loop(hopp) ->
+    tok_loop(hej).
diff --git a/erts/emulator/test/big_SUITE.erl b/erts/emulator/test/big_SUITE.erl
index c308760211..5939d024ae 100644
--- a/erts/emulator/test/big_SUITE.erl
+++ b/erts/emulator/test/big_SUITE.erl
@@ -339,6 +339,13 @@ system_limit(Config) when is_list(Config) ->
     {'EXIT',{system_limit,_}} = (catch apply(erlang, id('bsl'), [Maxbig,2])),
     {'EXIT',{system_limit,_}} = (catch id(1) bsl (1 bsl 45)),
     {'EXIT',{system_limit,_}} = (catch id(1) bsl (1 bsl 69)),
+
+    %% There should be no system_limit exception when shifting a zero.
+    0 = id(0) bsl (1 bsl 128),
+    0 = id(0) bsr -(1 bsl 128),
+    Erlang = id(erlang),
+    0 = Erlang:'bsl'(id(0), 1 bsl 128),
+    0 = Erlang:'bsr'(id(0), -(1 bsl 128)),
     ok.
 
 maxbig() ->
diff --git a/erts/emulator/test/binary_SUITE.erl b/erts/emulator/test/binary_SUITE.erl
index 374f91e487..a3c3daac15 100644
--- a/erts/emulator/test/binary_SUITE.erl
+++ b/erts/emulator/test/binary_SUITE.erl
@@ -48,6 +48,7 @@
 	 bad_list_to_binary/1, bad_binary_to_list/1,
 	 t_split_binary/1, bad_split/1,
 	 terms/1, terms_float/1, float_middle_endian/1,
+         b2t_used_big/1,
 	 external_size/1, t_iolist_size/1,
 	 t_hash/1,
 	 bad_size/1,
@@ -57,7 +58,9 @@
 	 otp_5484/1,otp_5933/1,
 	 ordering/1,unaligned_order/1,gc_test/1,
 	 bit_sized_binary_sizes/1,
-	 otp_6817/1,deep/1,obsolete_funs/1,robustness/1,otp_8117/1,
+	 otp_6817/1,deep/1,
+         term2bin_tuple_fallbacks/1,
+         robustness/1,otp_8117/1,
 	 otp_8180/1, trapping/1, large/1,
 	 error_after_yield/1, cmp_old_impl/1]).
 
@@ -72,12 +75,14 @@ all() ->
      t_split_binary, bad_split,
      bad_list_to_binary, bad_binary_to_list, terms,
      terms_float, float_middle_endian, external_size, t_iolist_size,
+     b2t_used_big,
      bad_binary_to_term_2, safe_binary_to_term2,
      bad_binary_to_term, bad_terms, t_hash, bad_size,
      bad_term_to_binary, more_bad_terms, otp_5484, otp_5933,
      ordering, unaligned_order, gc_test,
      bit_sized_binary_sizes, otp_6817, otp_8117, deep,
-     obsolete_funs, robustness, otp_8180, trapping, large,
+     term2bin_tuple_fallbacks,
+     robustness, otp_8180, trapping, large,
      error_after_yield, cmp_old_impl].
 
 groups() -> 
@@ -426,40 +431,77 @@ bad_term_to_binary(Config) when is_list(Config) ->
 
 terms(Config) when is_list(Config) ->
     TestFun = fun(Term) ->
-		      try
-			  S = io_lib:format("~p", [Term]),
-			  io:put_chars(S)
-		      catch
-			  error:badarg ->
-			      io:put_chars("bit sized binary")
-		      end,
+                      S = io_lib:format("~p", [Term]),
+                      io:put_chars(S),
 		      Bin = term_to_binary(Term),
 		      case erlang:external_size(Bin) of
 			  Sz when is_integer(Sz), size(Bin) =< Sz ->
 			      ok
 		      end,
-              Bin1 = term_to_binary(Term, [{minor_version, 1}]),
-              case erlang:external_size(Bin1, [{minor_version, 1}]) of
-              Sz1 when is_integer(Sz1), size(Bin1) =< Sz1 ->
-                  ok
-              end,
+                      Bin1 = term_to_binary(Term, [{minor_version, 1}]),
+                      case erlang:external_size(Bin1, [{minor_version, 1}]) of
+                          Sz1 when is_integer(Sz1), size(Bin1) =< Sz1 ->
+                              ok
+                      end,
 		      Term = binary_to_term_stress(Bin),
 		      Term = binary_to_term_stress(Bin, [safe]),
-		      Unaligned = make_unaligned_sub_binary(Bin),
-		      Term = binary_to_term_stress(Unaligned),
-		      Term = binary_to_term_stress(Unaligned, []),
-		      Term = binary_to_term_stress(Bin, [safe]),
+                      Bin_sz = byte_size(Bin),
+		      {Term,Bin_sz} = binary_to_term_stress(Bin, [used]),
+
+                      BinE = <<Bin/binary, 1, 2, 3>>,
+		      {Term,Bin_sz} = binary_to_term_stress(BinE, [used]),
+
+		      BinU = make_unaligned_sub_binary(Bin),
+		      Term = binary_to_term_stress(BinU),
+		      Term = binary_to_term_stress(BinU, []),
+		      Term = binary_to_term_stress(BinU, [safe]),
+		      {Term,Bin_sz} = binary_to_term_stress(BinU, [used]),
+
+                      BinUE = make_unaligned_sub_binary(BinE),
+		      {Term,Bin_sz} = binary_to_term_stress(BinUE, [used]),
+
 		      BinC = erlang:term_to_binary(Term, [compressed]),
+                      BinC_sz = byte_size(BinC),
+		      true = BinC_sz =< size(Bin),
 		      Term = binary_to_term_stress(BinC),
-		      true = size(BinC) =< size(Bin),
+		      {Term, BinC_sz} = binary_to_term_stress(BinC, [used]),
+
 		      Bin = term_to_binary(Term, [{compressed,0}]),
 		      terms_compression_levels(Term, size(Bin), 1),
-		      UnalignedC = make_unaligned_sub_binary(BinC),
-		      Term = binary_to_term_stress(UnalignedC)
+
+		      BinUC = make_unaligned_sub_binary(BinC),
+		      Term = binary_to_term_stress(BinUC),
+                      {Term,BinC_sz} = binary_to_term_stress(BinUC, [used]),
+
+                      BinCE = <<BinC/binary, 1, 2, 3>>,
+		      {Term,BinC_sz} = binary_to_term_stress(BinCE, [used]),
+
+		      BinUCE = make_unaligned_sub_binary(BinCE),
+		      Term = binary_to_term_stress(BinUCE),
+                      {Term,BinC_sz} = binary_to_term_stress(BinUCE, [used])
 	      end,
     test_terms(TestFun),
     ok.
 
+%% Test binary_to_term(_, [used]) returning a big Used integer.
+b2t_used_big(_Config) ->
+    case erlang:system_info(wordsize) of
+        8 ->
+            {skipped, "This is not a 32-bit machine"};
+        4 ->
+            %% Use a long utf8 atom for large external format but compact on heap.
+            BigAtom = binary_to_atom(<< <<16#F0908D88:32>> || _ <- lists:seq(1,255) >>,
+                                     utf8),
+            Atoms = (1 bsl 17) + (1 bsl 9),
+            BigAtomList = lists:duplicate(Atoms, BigAtom),
+            BigBin = term_to_binary(BigAtomList),
+            {BigAtomList, Used} = binary_to_term(BigBin, [used]),
+            2 = erts_debug:size(Used),
+            Used = byte_size(BigBin),
+            Used = 1 + 1 + 4 + Atoms*(1+2+4*255) + 1,
+            ok
+    end.
+
 terms_compression_levels(Term, UncompressedSz, Level) when Level < 10 ->
     BinC = erlang:term_to_binary(Term, [{compressed,Level}]),
     Term = binary_to_term_stress(BinC),
@@ -1161,7 +1203,7 @@ very_big_num(0, Result) ->
     Result.
 
 make_port() ->
-    open_port({spawn, efile}, [eof]).
+    hd(erlang:ports()).
 
 make_pid() ->
     spawn_link(?MODULE, sleeper, []).
@@ -1262,40 +1304,28 @@ deep_roundtrip(T) ->
     B = term_to_binary(T),
     T = binary_to_term(B).
 
-obsolete_funs(Config) when is_list(Config) ->
+term2bin_tuple_fallbacks(Config) when is_list(Config) ->
     erts_debug:set_internal_state(available_internal_state, true),
 
-    X = id({1,2,3}),
-    Y = id([a,b,c,d]),
-    Z = id({x,y,z}),
-    obsolete_fun(fun() -> ok end),
-    obsolete_fun(fun() -> X end),
-    obsolete_fun(fun(A) -> {A,X} end),
-    obsolete_fun(fun() -> {X,Y} end),
-    obsolete_fun(fun() -> {X,Y,Z} end),
-
-    obsolete_fun(fun ?MODULE:all/1),
+    term2bin_tf(fun ?MODULE:all/1),
+    term2bin_tf(<<1:1>>),
+    term2bin_tf(<<90,80:7>>),
 
     erts_debug:set_internal_state(available_internal_state, false),
     ok.
 
-obsolete_fun(Fun) ->
-    Tuple = case erlang:fun_info(Fun, type) of
-		{type,external} ->
-		    {module,M} = erlang:fun_info(Fun, module),
-		    {name,F} = erlang:fun_info(Fun, name),
-		    {M,F};
-		{type,local} ->
-		    {module,M} = erlang:fun_info(Fun, module),
-		    {index,I} = erlang:fun_info(Fun, index),
-		    {uniq,U} = erlang:fun_info(Fun, uniq),
-		    {env,E} = erlang:fun_info(Fun, env),
-		    {'fun',M,I,U,list_to_tuple(E)}
-	    end,
-    Tuple = no_fun_roundtrip(Fun).
-
-no_fun_roundtrip(Term) ->
-    binary_to_term_stress(erts_debug:get_internal_state({term_to_binary_no_funs,Term})).
+term2bin_tf(Term) ->
+    Tuple = case Term of
+                Fun when is_function(Fun) ->
+                    {type, external} = erlang:fun_info(Fun, type),
+                    {module,M} = erlang:fun_info(Fun, module),
+                    {name,F} = erlang:fun_info(Fun, name),
+                    {M,F};
+                BS when bit_size(BS) rem 8 =/= 0 ->
+                    Bits = bit_size(BS) rem 8,
+                    {<<BS/bitstring, 0:(8-Bits)>>, Bits}
+            end,
+    Tuple = binary_to_term_stress(erts_debug:get_internal_state({term_to_binary_tuple_fallbacks,Term})).
 
 %% Test non-standard encodings never generated by term_to_binary/1
 %% but recognized by binary_to_term/1.
@@ -1443,13 +1473,13 @@ error_after_yield(Type, M, F, AN, AFun, TrapFunc) ->
 					   apply(M, F, A),
 					   exit({unexpected_success, {M, F, A}})
 				       catch
-					   error:Type ->
+					   error:Type:Stk ->
 					       erlang:trace(self(),false,[running,{tracer,Tracer}]),
 					       %% We threw the exception from the native
 					       %% function we trapped to, but we want
 					       %% the BIF that originally was called
 					       %% to appear in the stack trace.
-					       [{M, F, A, _} | _] = erlang:get_stacktrace()
+					       [{M, F, A, _} | _] = Stk
 				       end
 			       end),
     receive
diff --git a/erts/emulator/test/call_trace_SUITE.erl b/erts/emulator/test/call_trace_SUITE.erl
index 1251d644ae..d19f7f81ad 100644
--- a/erts/emulator/test/call_trace_SUITE.erl
+++ b/erts/emulator/test/call_trace_SUITE.erl
@@ -1116,8 +1116,8 @@ get_deep_4_loc(Arg) ->
         deep_4(Arg),
         ct:fail(should_not_return_to_here)
     catch
-        _:_ ->
-            [{?MODULE,deep_4,1,Loc0}|_] = erlang:get_stacktrace(),
+        _:_:Stk ->
+            [{?MODULE,deep_4,1,Loc0}|_] = Stk,
             Loc0
     end.
 
diff --git a/erts/emulator/test/code_SUITE.erl b/erts/emulator/test/code_SUITE.erl
index 661a2ee6c9..9c6dc3ff83 100644
--- a/erts/emulator/test/code_SUITE.erl
+++ b/erts/emulator/test/code_SUITE.erl
@@ -957,7 +957,7 @@ erl_544(Config) when is_list(Config) ->
                 StackFun = fun(_, _, _) -> false end,
                 FormatFun = fun (Term, _) -> io_lib:format("~tp", [Term]) end,
                 Formated =
-                    lib:format_stacktrace(1, Stack, StackFun, FormatFun),
+                    erl_error:format_stacktrace(1, Stack, StackFun, FormatFun),
                 true = is_list(Formated),
                 ok
             after
diff --git a/erts/emulator/test/ddll_SUITE.erl b/erts/emulator/test/ddll_SUITE.erl
index 031b05790d..4998fc08be 100644
--- a/erts/emulator/test/ddll_SUITE.erl
+++ b/erts/emulator/test/ddll_SUITE.erl
@@ -775,7 +775,7 @@ errors(Config) when is_list(Config) ->
     {error, bad_driver_name} = erl_ddll:load_driver(Path, wrongname_drv),
 
     %% We assume that there is a statically linked driver named "ddll":
-    {error, linked_in_driver} = erl_ddll:unload_driver(efile),
+    {error, linked_in_driver} = erl_ddll:unload_driver(ram_file_drv),
     {error, not_loaded} = erl_ddll:unload_driver("__pucko_driver__"),
 
     case os:type() of
diff --git a/erts/emulator/test/decode_packet_SUITE.erl b/erts/emulator/test/decode_packet_SUITE.erl
index 54ee4d5567..0ccdbd7ee8 100644
--- a/erts/emulator/test/decode_packet_SUITE.erl
+++ b/erts/emulator/test/decode_packet_SUITE.erl
@@ -239,7 +239,7 @@ packet_size(Config) when is_list(Config) ->
     %% Test OTP-9389, long HTTP header lines.
     Opts = [{packet_size, 128}],
     Pkt = list_to_binary(["GET / HTTP/1.1\r\nHost: localhost\r\nLink: /",
-                          string:chars($Y, 64), "\r\n\r\n"]),
+                          lists:duplicate(64, $Y), "\r\n\r\n"]),
     <<Pkt1:50/binary, Pkt2/binary>> = Pkt,
     {ok, {http_request,'GET',{abs_path,"/"},{1,1}}, Rest1} =
     erlang:decode_packet(http, Pkt1, Opts),
@@ -250,7 +250,7 @@ packet_size(Config) when is_list(Config) ->
     erlang:decode_packet(httph, list_to_binary([Rest2, Pkt2]), Opts),
 
     Pkt3 = list_to_binary(["GET / HTTP/1.1\r\nHost: localhost\r\nLink: /",
-                           string:chars($Y, 129), "\r\n\r\n"]),
+                           lists:duplicate(129, $Y), "\r\n\r\n"]),
     {ok, {http_request,'GET',{abs_path,"/"},{1,1}}, Rest3} =
     erlang:decode_packet(http, Pkt3, Opts),
     {ok, {http_header,_,'Host',_,"localhost"}, Rest4} =
@@ -509,9 +509,9 @@ decode_line(Bin,MaxLen) ->
     end.
 
 find_in_binary(Byte, Bin) ->
-    case string:chr(binary_to_list(Bin),Byte) of
-        0 -> notfound;	    
-        P -> P
+    case string:find(Bin, [Byte]) of
+        nomatch -> notfound;
+        Suffix -> byte_size(Bin) - byte_size(Suffix) + 1
     end.    
 
 ssl(Config) when is_list(Config) ->
@@ -562,7 +562,7 @@ decode_pkt(Type,Bin,Opts) ->
 otp_9389(Config) when is_list(Config) ->
     Opts = [{packet_size, 16384}, {line_length, 3000}],
     Pkt = list_to_binary(["GET / HTTP/1.1\r\nHost: localhost\r\nLink: /",
-                          string:chars($X, 8192),
+                          lists:duplicate(8192, $X),
                           "\r\nContent-Length: 0\r\n\r\n"]),
     <<Pkt1:5000/binary, Pkt2/binary>> = Pkt,
     {ok, {http_request,'GET',{abs_path,"/"},{1,1}}, Rest1} =
diff --git a/erts/emulator/test/dgawd_handler.erl b/erts/emulator/test/dgawd_handler.erl
index 52cdd26427..29b9d6ac7b 100644
--- a/erts/emulator/test/dgawd_handler.erl
+++ b/erts/emulator/test/dgawd_handler.erl
@@ -42,10 +42,10 @@
 %%====================================================================
 
 install() ->
-    gen_event:add_handler(error_logger, ?MODULE, []).
+    error_logger:add_report_handler(?MODULE, []).
 
 restore() ->
-    gen_event:delete_handler(error_logger, ?MODULE, []).
+    error_logger:delete_report_handler(?MODULE).
 
 got_dgawd_report() ->
     gen_event:call(error_logger, ?MODULE, got_dgawd_report, 10*60*1000).
diff --git a/erts/emulator/test/dirty_bif_SUITE.erl b/erts/emulator/test/dirty_bif_SUITE.erl
index 981ec4d48d..46eb0cba58 100644
--- a/erts/emulator/test/dirty_bif_SUITE.erl
+++ b/erts/emulator/test/dirty_bif_SUITE.erl
@@ -108,90 +108,80 @@ dirty_bif_exception(Config) when is_list(Config) ->
 			      erts_debug:dirty_cpu(error, Error),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty_cpu,[error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk1 ->
+				  [{erts_debug,dirty_cpu,[error, Error],_}|_] = Stk1,
 				  ok
 			  end,
 			  try
 			      apply(erts_debug,dirty_cpu,[error, Error]),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty_cpu,[error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk2 ->
+				  [{erts_debug,dirty_cpu,[error, Error],_}|_] = Stk2,
 				  ok
 			  end,
 			  try
 			      erts_debug:dirty_io(error, Error),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty_io,[error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk3 ->
+				  [{erts_debug,dirty_io,[error, Error],_}|_] = Stk3,
 				  ok
 			  end,
 			  try
 			      apply(erts_debug,dirty_io,[error, Error]),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty_io,[error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk4 ->
+				  [{erts_debug,dirty_io,[error, Error],_}|_] = Stk4,
 				  ok
 			  end,
 			  try
 			      erts_debug:dirty(normal, error, Error),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[normal, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk5 ->
+				  [{erts_debug,dirty,[normal, error, Error],_}|_] = Stk5,
 				  ok
 			  end,
 			  try
 			      apply(erts_debug,dirty,[normal, error, Error]),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[normal, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk6 ->
+				  [{erts_debug,dirty,[normal, error, Error],_}|_] = Stk6,
 				  ok
 			  end,
 			  try
 			      erts_debug:dirty(dirty_cpu, error, Error),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[dirty_cpu, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk7 ->
+				  [{erts_debug,dirty,[dirty_cpu, error, Error],_}|_] = Stk7,
 				  ok
 			  end,
 			  try
 			      apply(erts_debug,dirty,[dirty_cpu, error, Error]),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[dirty_cpu, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk8 ->
+				  [{erts_debug,dirty,[dirty_cpu, error, Error],_}|_] = Stk8,
 				  ok
 			  end,
 			  try
 			      erts_debug:dirty(dirty_io, error, Error),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[dirty_io, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk9 ->
+				  [{erts_debug,dirty,[dirty_io, error, Error],_}|_] = Stk9,
 				  ok
 			  end,
 			  try
 			      apply(erts_debug,dirty,[dirty_io, error, Error]),
 			      ct:fail(expected_exception)
 			  catch
-			      error:ErrorType ->
-				  [{erts_debug,dirty,[dirty_io, error, Error],_}|_]
-				      = erlang:get_stacktrace(),
+			      error:ErrorType:Stk10 ->
+				  [{erts_debug,dirty,[dirty_io, error, Error],_}|_] = Stk10,
 				  ok
 			  end
 		  end,
@@ -204,25 +194,22 @@ dirty_bif_multischedule_exception(Config) when is_list(Config) ->
     try
 	erts_debug:dirty_cpu(reschedule,1001)
     catch
-	error:badarg ->
-	    [{erts_debug,dirty_cpu,[reschedule, 1001],_}|_]
-		= erlang:get_stacktrace(),
+	error:badarg:Stk1 ->
+	    [{erts_debug,dirty_cpu,[reschedule, 1001],_}|_] = Stk1,
 	    ok
     end,
     try
 	erts_debug:dirty_io(reschedule,1001)
     catch
-	error:badarg ->
-	    [{erts_debug,dirty_io,[reschedule, 1001],_}|_]
-		= erlang:get_stacktrace(),
+	error:badarg:Stk2 ->
+	    [{erts_debug,dirty_io,[reschedule, 1001],_}|_] = Stk2,
 	    ok
     end,
     try
 	erts_debug:dirty(normal,reschedule,1001)
     catch
-	error:badarg ->
-	    [{erts_debug,dirty,[normal,reschedule,1001],_}|_]
-		= erlang:get_stacktrace(),
+	error:badarg:Stk3 ->
+	    [{erts_debug,dirty,[normal,reschedule,1001],_}|_] = Stk3,
 	    ok
     end.
 
@@ -230,6 +217,11 @@ dirty_scheduler_exit(Config) when is_list(Config) ->
     {ok, Node} = start_node(Config, "+SDio 1"),
     [ok] = mcall(Node,
                  [fun() ->
+                          %% Perform a dry run to ensure that all required code
+                          %% is loaded. Otherwise the test will fail since code
+                          %% loading is done through dirty IO and it won't make
+                          %% any progress during this test.
+                          _DryRun = test_dirty_scheduler_exit(),
 			  Start = erlang:monotonic_time(millisecond),
                           ok = test_dirty_scheduler_exit(),
 			  End = erlang:monotonic_time(millisecond),
@@ -246,23 +238,22 @@ test_dse(0,Pids) ->
     timer:sleep(100),
     kill_dse(Pids,[]);
 test_dse(N,Pids) ->
-    Pid = spawn_link(fun () -> erts_debug:dirty_io(wait, 5000) end),
+    Pid = spawn_link(fun () -> erts_debug:dirty_io(wait, 1000) end),
     test_dse(N-1,[Pid|Pids]).
 
 kill_dse([],Killed) ->
-    wait_dse(Killed);
+    wait_dse(Killed, ok);
 kill_dse([Pid|Pids],AlreadyKilled) ->
     exit(Pid,kill),
     kill_dse(Pids,[Pid|AlreadyKilled]).
 
-wait_dse([]) ->
-    ok;
-wait_dse([Pid|Pids]) ->
+wait_dse([], Result) ->
+    Result;
+wait_dse([Pid|Pids], Result) ->
     receive
-        {'EXIT',Pid,Reason} ->
-	    killed = Reason
-    end,
-    wait_dse(Pids).
+        {'EXIT', Pid, killed} -> wait_dse(Pids, Result);
+        {'EXIT', Pid, _Other} -> wait_dse(Pids, failed)
+    end.
 
 dirty_call_while_terminated(Config) when is_list(Config) ->
     Me = self(),
diff --git a/erts/emulator/test/dirty_nif_SUITE.erl b/erts/emulator/test/dirty_nif_SUITE.erl
index 13806fd5c4..93d0ac392c 100644
--- a/erts/emulator/test/dirty_nif_SUITE.erl
+++ b/erts/emulator/test/dirty_nif_SUITE.erl
@@ -109,9 +109,8 @@ dirty_nif_exception(Config) when is_list(Config) ->
 	call_dirty_nif_exception(1),
 	ct:fail(expected_badarg)
     catch
-	error:badarg ->
-	    [{?MODULE,call_dirty_nif_exception,[1],_}|_] =
-		erlang:get_stacktrace(),
+	error:badarg:Stk1 ->
+	    [{?MODULE,call_dirty_nif_exception,[1],_}|_] = Stk1,
 	    ok
     end,
     try
@@ -121,9 +120,8 @@ dirty_nif_exception(Config) when is_list(Config) ->
 	call_dirty_nif_exception(0),
 	ct:fail(expected_badarg)
     catch
-	error:badarg ->
-	    [{?MODULE,call_dirty_nif_exception,[0],_}|_] =
-		erlang:get_stacktrace(),
+	error:badarg:Stk2 ->
+	    [{?MODULE,call_dirty_nif_exception,[0],_}|_] = Stk2,
 	    ok
     end,
     %% this checks that a dirty NIF can raise various terms as
@@ -138,8 +136,8 @@ nif_raise_exceptions(NifFunc) ->
                             erlang:apply(?MODULE,NifFunc,[Term]),
                             ct:fail({expected,Term})
                         catch
-                            error:Term ->
-                                [{?MODULE,NifFunc,[Term],_}|_] = erlang:get_stacktrace(),
+                            error:Term:Stk ->
+                                [{?MODULE,NifFunc,[Term],_}|_] = Stk,
                                 ok
                         end
                 end, ok, ExcTerms).
@@ -151,6 +149,11 @@ dirty_scheduler_exit(Config) when is_list(Config) ->
     [ok] = mcall(Node,
                  [fun() ->
                           ok = erlang:load_nif(NifLib, []),
+                          %% Perform a dry run to ensure that all required code
+                          %% is loaded. Otherwise the test will fail since code
+                          %% loading is done through dirty IO and it won't make
+                          %% any progress during this test.
+                          _DryRun = test_dirty_scheduler_exit(),
 			  Start = erlang:monotonic_time(millisecond),
                           ok = test_dirty_scheduler_exit(),
 			  End = erlang:monotonic_time(millisecond),
@@ -171,19 +174,18 @@ test_dse(N,Pids) ->
     test_dse(N-1,[Pid|Pids]).
 
 kill_dse([],Killed) ->
-    wait_dse(Killed);
+    wait_dse(Killed, ok);
 kill_dse([Pid|Pids],AlreadyKilled) ->
     exit(Pid,kill),
     kill_dse(Pids,[Pid|AlreadyKilled]).
 
-wait_dse([]) ->
-    ok;
-wait_dse([Pid|Pids]) ->
+wait_dse([], Result) ->
+    Result;
+wait_dse([Pid|Pids], Result) ->
     receive
-        {'EXIT',Pid,Reason} ->
-	    killed = Reason
-    end,
-    wait_dse(Pids).
+        {'EXIT', Pid, killed} -> wait_dse(Pids, Result);
+        {'EXIT', Pid, _Other} -> wait_dse(Pids, failed)
+    end.
 
 dirty_call_while_terminated(Config) when is_list(Config) ->
     Me = self(),
@@ -287,9 +289,9 @@ access_dirty_heap(Dirty, RGL, N, R) ->
 %% dirty NIF where the main lock is needed for that access do not get
 %% blocked. Each test passes its pid to dirty_sleeper, which sends a
 %% 'ready' message when it's running on a dirty scheduler and just before
-%% it starts a 6 second sleep. When it receives the message, it verifies
+%% it starts a 2 second sleep. When it receives the message, it verifies
 %% that access to the dirty process is as it expects.  After the dirty
-%% process finishes its 6 second sleep but before it returns from the dirty
+%% process finishes its 2 second sleep but before it returns from the dirty
 %% scheduler, it sends a 'done' message. If the tester already received
 %% that message, the test fails because it means attempting to access the
 %% dirty process waited for that process to return to a regular scheduler,
@@ -353,7 +355,7 @@ dirty_process_trace(Config) when is_list(Config) ->
 			      error(missing_trace_return_message)
 		      end
 	      after
-		  6500 ->
+		  2500 ->
 		      error(missing_done_message)
 	      end,
 	      ok
@@ -380,7 +382,7 @@ code_purge(Config) when is_list(Config) ->
     Start = erlang:monotonic_time(),
     {Pid1, Mon1} = spawn_monitor(fun () ->
 				       dirty_code_test:func(fun () ->
-								    %% Sleep for 6 seconds
+								    %% Sleep for 2 seconds
 								    %% in dirty nif...
 								    dirty_sleeper()
 							    end)
@@ -388,7 +390,7 @@ code_purge(Config) when is_list(Config) ->
     {module, dirty_code_test} = erlang:load_module(dirty_code_test, Bin),
     {Pid2, Mon2} = spawn_monitor(fun () ->
 				       dirty_code_test:func(fun () ->
-								    %% Sleep for 6 seconds
+								    %% Sleep for 2 seconds
 								    %% in dirty nif...
 								    dirty_sleeper()
 							    end)
@@ -490,7 +492,7 @@ test_dirty_process_access(Start, Test, Finish) ->
 			 ok
 		 end
 	 after
-	     3000 ->
+	     1000 ->
 		 error(timeout)
 	 end,
     ok = Finish(NifPid).
diff --git a/erts/emulator/test/dirty_nif_SUITE_data/dirty_nif_SUITE.c b/erts/emulator/test/dirty_nif_SUITE_data/dirty_nif_SUITE.c
index 0321b9898f..a94a2c0b02 100644
--- a/erts/emulator/test/dirty_nif_SUITE_data/dirty_nif_SUITE.c
+++ b/erts/emulator/test/dirty_nif_SUITE_data/dirty_nif_SUITE.c
@@ -1,7 +1,7 @@
 /*
  * %CopyrightBegin%
  *
- * Copyright Ericsson AB 2009-2017. All Rights Reserved.
+ * Copyright Ericsson AB 2009-2018. All Rights Reserved.
  *
  * Licensed under the Apache License, Version 2.0 (the "License");
  * you may not use this file except in compliance with the License.
@@ -112,15 +112,12 @@ static ERL_NIF_TERM send_from_dirty_nif(ErlNifEnv* env, int argc, const ERL_NIF_
 {
     ERL_NIF_TERM result;
     ErlNifPid pid;
-    ErlNifEnv* menv;
     int res;
 
     if (!enif_get_local_pid(env, argv[0], &pid))
 	return enif_make_badarg(env);
     result = enif_make_tuple2(env, enif_make_atom(env, "ok"), enif_make_pid(env, &pid));
-    menv = enif_alloc_env();
-    res = enif_send(env, &pid, menv, result);
-    enif_free_env(menv);
+    res = enif_send(env, &pid, NULL, result);
     if (!res)
 	return enif_make_badarg(env);
     else
@@ -131,15 +128,12 @@ static ERL_NIF_TERM send_wait_from_dirty_nif(ErlNifEnv* env, int argc, const ERL
 {
     ERL_NIF_TERM result;
     ErlNifPid pid;
-    ErlNifEnv* menv;
     int res;
 
     if (!enif_get_local_pid(env, argv[0], &pid))
 	return enif_make_badarg(env);
     result = enif_make_tuple2(env, enif_make_atom(env, "ok"), enif_make_pid(env, &pid));
-    menv = enif_alloc_env();
-    res = enif_send(env, &pid, menv, result);
-    enif_free_env(menv);
+    res = enif_send(env, &pid, NULL, result);
 
 #ifdef __WIN32__
     Sleep(2000);
@@ -211,22 +205,17 @@ dirty_sleeper(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
     /* If we get a pid argument, it indicates a process involved in the
        test wants a message from us. Prior to the sleep we send a 'ready'
        message, and then after the sleep, send a 'done' message. */
-    if (argc == 1 && enif_get_local_pid(env, argv[0], &pid)) {
-        msg_env = enif_alloc_env();
-        enif_send(env, &pid, msg_env, enif_make_atom(msg_env, "ready"));
-    }
+    if (argc == 1 && enif_get_local_pid(env, argv[0], &pid))
+        enif_send(env, &pid, NULL, enif_make_atom(env, "ready"));
 
 #ifdef __WIN32__
-    Sleep(6000);
+    Sleep(2000);
 #else
-    sleep(6);
+    sleep(2);
 #endif
 
-    if (argc == 1) {
-        assert(msg_env != NULL);
-        enif_send(env, &pid, msg_env, enif_make_atom(msg_env, "done"));
-        enif_free_env(msg_env);
-    }
+    if (argc == 1)
+        enif_send(env, &pid, NULL, enif_make_atom(env, "done"));
 
     return enif_make_atom(env, "ok");
 }
@@ -247,8 +236,8 @@ static ERL_NIF_TERM dirty_call_while_terminated_nif(ErlNifEnv* env, int argc, co
 
     self_term = enif_make_pid(env, &self);
 
-    result = enif_make_tuple2(env, enif_make_atom(env, "dirty_alive"), self_term);
     menv = enif_alloc_env();
+    result = enif_make_tuple2(menv, enif_make_atom(menv, "dirty_alive"), self_term);
     res = enif_send(env, &to, menv, result);
     enif_free_env(menv);
     if (!res)
@@ -259,9 +248,7 @@ static ERL_NIF_TERM dirty_call_while_terminated_nif(ErlNifEnv* env, int argc, co
 	;
 
     result = enif_make_tuple2(env, enif_make_atom(env, "dirty_dead"), self_term);
-    menv = enif_alloc_env();
-    res = enif_send(env, &to, menv, result);
-    enif_free_env(menv);
+    res = enif_send(env, &to, NULL, result);
 
 #ifdef __WIN32__
     Sleep(1000);
diff --git a/erts/emulator/test/distribution_SUITE.erl b/erts/emulator/test/distribution_SUITE.erl
index e731b68f2f..45dd922ff0 100644
--- a/erts/emulator/test/distribution_SUITE.erl
+++ b/erts/emulator/test/distribution_SUITE.erl
@@ -35,13 +35,18 @@
 
 -include_lib("common_test/include/ct.hrl").
 
+%-define(Line, erlang:display({line,?LINE}),).
+-define(Line,).
+
 -export([all/0, suite/0, groups/0,
          ping/1, bulk_send_small/1,
+         group_leader/1,
+         optimistic_dflags/1,
          bulk_send_big/1, bulk_send_bigbig/1,
          local_send_small/1, local_send_big/1,
          local_send_legal/1, link_to_busy/1, exit_to_busy/1,
          lost_exit/1, link_to_dead/1, link_to_dead_new_node/1,
-         applied_monitor_node/1, ref_port_roundtrip/1, nil_roundtrip/1,
+         ref_port_roundtrip/1, nil_roundtrip/1,
          trap_bif_1/1, trap_bif_2/1, trap_bif_3/1,
          stop_dist/1,
          dist_auto_connect_never/1, dist_auto_connect_once/1,
@@ -61,12 +66,14 @@
 
 %% Internal exports.
 -export([sender/3, receiver2/2, dummy_waiter/0, dead_process/0,
+         group_leader_1/1,
+         optimistic_dflags_echo/0, optimistic_dflags_sender/1,
          roundtrip/1, bounce/1, do_dist_auto_connect/1, inet_rpc_server/1,
          dist_parallel_sender/3, dist_parallel_receiver/0,
          dist_evil_parallel_receiver/0]).
 
 %% epmd_module exports
--export([start_link/0, register_node/2, register_node/3, port_please/2]).
+-export([start_link/0, register_node/2, register_node/3, port_please/2, address_please/3]).
 
 suite() ->
     [{ct_hooks,[ts_install_cth]},
@@ -74,8 +81,10 @@ suite() ->
 
 all() ->
     [ping, {group, bulk_send}, {group, local_send},
+     group_leader,
+     optimistic_dflags,
      link_to_busy, exit_to_busy, lost_exit, link_to_dead,
-     link_to_dead_new_node, applied_monitor_node,
+     link_to_dead_new_node,
      ref_port_roundtrip, nil_roundtrip, stop_dist,
      {group, trap_bif}, {group, dist_auto_connect},
      dist_parallel_send, atom_roundtrip, unicode_atom_roundtrip,
@@ -124,6 +133,96 @@ ping(Config) when is_list(Config) ->
 
     ok.
 
+%% Test erlang:group_leader(_, ExternalPid), i.e. DOP_GROUP_LEADER
+group_leader(Config) when is_list(Config) ->
+    ?Line Sock = start_relay_node(group_leader_1, []),
+    ?Line Sock2 = start_relay_node(group_leader_2, []),
+    try
+        ?Line Node2 = inet_rpc_nodename(Sock2),
+        ?Line {ok, ok} = do_inet_rpc(Sock, ?MODULE, group_leader_1, [Node2])
+    after
+        ?Line stop_relay_node(Sock),
+        ?Line stop_relay_node(Sock2)
+    end,
+    ok.
+
+group_leader_1(Node2) ->
+    ?Line ExtPid = spawn(Node2, fun F() ->
+                                        receive {From, group_leader} ->
+                                                From ! {self(), group_leader, group_leader()}
+                                        end,
+                                        F()
+                                end),
+    ?Line GL1 = self(),
+    ?Line group_leader(GL1, ExtPid),
+    ?Line ExtPid ! {self(), group_leader},
+    ?Line {ExtPid, group_leader, GL1} = receive_one(),
+
+    %% Kill connection and repeat test when group_leader/2 triggers auto-connect
+    ?Line net_kernel:monitor_nodes(true),
+    ?Line net_kernel:disconnect(Node2),
+    ?Line {nodedown, Node2} = receive_one(),
+    ?Line GL2 = spawn(fun() -> dummy end),
+    ?Line group_leader(GL2, ExtPid),
+    ?Line {nodeup, Node2} = receive_one(),
+    ?Line ExtPid ! {self(), group_leader},
+    ?Line {ExtPid, group_leader, GL2} = receive_one(),
+    ok.
+
+%% Test optimistic distribution flags toward pending connections (DFLAG_DIST_HOPEFULLY)
+optimistic_dflags(Config) when is_list(Config) ->
+    ?Line Sender = start_relay_node(optimistic_dflags_sender, []),
+    ?Line Echo = start_relay_node(optimistic_dflags_echo, []),
+    try
+        ?Line {ok, ok} = do_inet_rpc(Echo, ?MODULE, optimistic_dflags_echo, []),
+
+        ?Line EchoNode = inet_rpc_nodename(Echo),
+        ?Line {ok, ok} = do_inet_rpc(Sender, ?MODULE, optimistic_dflags_sender, [EchoNode])
+    after
+        ?Line stop_relay_node(Sender),
+        ?Line stop_relay_node(Echo)
+    end,
+    ok.
+
+optimistic_dflags_echo() ->
+    P = spawn(fun F() ->
+                      receive {From, Term} ->
+                              From ! {self(), Term}
+                      end,
+                      F()
+              end),
+    register(optimistic_dflags_echo, P),
+    optimistic_dflags_echo ! {self(), hello},
+    {P, hello} = receive_one(),
+    ok.
+
+optimistic_dflags_sender(EchoNode) ->
+    ?Line net_kernel:monitor_nodes(true),
+
+    optimistic_dflags_do(EchoNode, <<1:1>>),
+    optimistic_dflags_do(EchoNode, fun lists:map/2),
+    ok.
+
+optimistic_dflags_do(EchoNode, Term) ->
+    ?Line {optimistic_dflags_echo, EchoNode} ! {self(), Term},
+    ?Line {nodeup, EchoNode} = receive_one(),
+    ?Line {EchoPid, Term} = receive_one(),
+    %% repeat with pid destination
+    ?Line net_kernel:disconnect(EchoNode),
+    ?Line {nodedown, EchoNode} = receive_one(),
+    ?Line EchoPid ! {self(), Term},
+    ?Line {nodeup, EchoNode} = receive_one(),
+    ?Line {EchoPid, Term} = receive_one(),
+
+    ?Line net_kernel:disconnect(EchoNode),
+    ?Line {nodedown, EchoNode} = receive_one(),
+    ok.
+
+
+receive_one() ->
+    receive M -> M after 1000 -> timeout end.
+
+
 bulk_send_small(Config) when is_list(Config) ->
     bulk_send(64, 32).
 
@@ -420,18 +519,20 @@ make_busy(Node, Time) when is_integer(Time) ->
     Own = 500,
     freeze_node(Node, Time+Own),
     Data = make_busy_data(),
+    DCtrl = dctrl(Node),
     %% first make port busy
     Pid = spawn_link(fun () ->
                              forever(fun () ->
-                                             dport_reg_send(Node,
-                                                            '__noone__',
-                                                            Data)
+                                             dctrl_dop_reg_send(Node,
+                                                                '__noone__',
+                                                                Data)
                                      end)
                      end),
     receive after Own -> ok end,
     until(fun () ->
-                  case process_info(Pid, status) of
-                      {status, suspended} -> true;
+                  case {DCtrl, process_info(Pid, status)} of
+                      {DPrt, {status, suspended}} when is_port(DPrt) -> true;
+                      {DPid, {status, waiting}} when is_pid(DPid) -> true;
                       _ -> false
                   end
           end),
@@ -637,31 +738,11 @@ link_to_dead_new_node(Config) when is_list(Config) ->
     end,
     ok.
 
-%% Test that monitor_node/2 works when applied.
-applied_monitor_node(Config) when is_list(Config) ->
-    NonExisting = list_to_atom("__non_existing__@" ++ hostname()),
-
-    %% Tail-recursive call to apply (since the node is non-existing,
-    %% there will be a trap).
-
-    true = tail_apply(erlang, monitor_node, [NonExisting, true]),
-    [{nodedown, NonExisting}] = test_server:messages_get(),
-
-    %% Ordinary call (with trap).
-
-    true = apply(erlang, monitor_node, [NonExisting, true]),
-    [{nodedown, NonExisting}] = test_server:messages_get(),
-
-    ok.
-
-tail_apply(M, F, A) ->
-    apply(M, F, A).
-
 %% Test that sending a port or reference to another node and back again
 %% doesn't correct them in any way.
 ref_port_roundtrip(Config) when is_list(Config) ->
     process_flag(trap_exit, true),
-    Port = open_port({spawn, efile}, []),
+    Port = make_port(),
     Ref = make_ref(),
     {ok, Node} = start_node(ref_port_roundtrip),
     net_adm:ping(Node),
@@ -682,6 +763,9 @@ ref_port_roundtrip(Config) when is_list(Config) ->
     end,
     ok.
 
+make_port() ->
+    hd(erlang:ports()).
+
 roundtrip(Term) ->
     exit(Term).
 
@@ -713,7 +797,7 @@ show_term(Term) ->
 
 %% Tests behaviour after net_kernel:stop (OTP-2586).
 stop_dist(Config) when is_list(Config) ->
-    Str = os:cmd(atom_to_list(lib:progname())
+    Str = os:cmd(ct:get_progname()
                  ++ " -noshell -pa "
                  ++ proplists:get_value(data_dir, Config)
                  ++ " -s run"),
@@ -784,8 +868,8 @@ dist_auto_connect_once(Config) when is_list(Config) ->
     {ok, pong} = do_inet_rpc(Sock2,net_adm,ping,[NN]),
     {ok,[NN2]} = do_inet_rpc(Sock,erlang,nodes,[]),
     {ok,[NN]} = do_inet_rpc(Sock2,erlang,nodes,[]),
-    [_,HostPartPeer] = string:tokens(atom_to_list(NN),"@"),
-    [_,MyHostPart] = string:tokens(atom_to_list(node()),"@"),
+    [_,HostPartPeer] = string:lexemes(atom_to_list(NN),"@"),
+    [_,MyHostPart] = string:lexemes(atom_to_list(node()),"@"),
     % Give net_kernel a chance to change the state of the node to up to.
     receive after 1000 -> ok end,
     case HostPartPeer of
@@ -890,9 +974,9 @@ dist_auto_connect_start(Name, Value) when is_list(Name), is_atom(Value) ->
     ModuleDir = filename:dirname(code:which(?MODULE)),
     ValueStr = atom_to_list(Value),
     Cookie = atom_to_list(erlang:get_cookie()),
-    Cmd = lists:concat(
+    Cmd = lists:append(
             [%"xterm -e ",
-             atom_to_list(lib:progname()),
+             ct:get_progname(),
              %	     " -noinput ",
              " -detached ",
              long_or_short(), " ", Name,
@@ -1156,8 +1240,6 @@ contended_atom_cache_entry_test(Config, Type) ->
     spawn_link(
       SNode,
       fun () ->
-              erts_debug:set_internal_state(available_internal_state,
-                                            true),
               Master = self(),
               CIX = get_cix(),
               TestAtoms = case Type of
@@ -1242,7 +1324,7 @@ get_cix(CIX) when is_integer(CIX), CIX < 0 ->
 get_cix(CIX) when is_integer(CIX) ->
     get_cix(CIX,
             unwanted_cixs(),
-            erts_debug:get_internal_state(max_atom_out_cache_index)).
+            get_internal_state(max_atom_out_cache_index)).
 
 get_cix(CIX, Unwanted, MaxCIX) when CIX > MaxCIX ->
     get_cix(0, Unwanted, MaxCIX);
@@ -1254,8 +1336,8 @@ get_cix(CIX, Unwanted, MaxCIX) ->
 
 unwanted_cixs() ->
     lists:map(fun (Node) ->
-                      erts_debug:get_internal_state({atom_out_cache_index,
-                                                     Node})
+                      get_internal_state({atom_out_cache_index,
+                                          Node})
               end,
               nodes()).
 
@@ -1264,7 +1346,7 @@ get_conflicting_atoms(_CIX, 0) ->
     [];
 get_conflicting_atoms(CIX, N) ->
     Atom = list_to_atom("atom" ++ integer_to_list(erlang:unique_integer([positive]))),
-    case erts_debug:get_internal_state({atom_out_cache_index, Atom}) of
+    case get_internal_state({atom_out_cache_index, Atom}) of
         CIX ->
             [Atom|get_conflicting_atoms(CIX, N-1)];
         _ ->
@@ -1275,7 +1357,7 @@ get_conflicting_unicode_atoms(_CIX, 0) ->
     [];
 get_conflicting_unicode_atoms(CIX, N) ->
     Atom = string_to_atom([16#1f608] ++ "atom" ++ integer_to_list(erlang:unique_integer([positive]))),
-    case erts_debug:get_internal_state({atom_out_cache_index, Atom}) of
+    case get_internal_state({atom_out_cache_index, Atom}) of
         CIX ->
             [Atom|get_conflicting_unicode_atoms(CIX, N-1)];
         _ ->
@@ -1738,37 +1820,38 @@ bad_dist_ext_check_msgs([M|Ms]) ->
             bad_dist_ext_check_msgs(Ms)
     end.
 
+ensure_dctrl(Node) ->
+    case dctrl(Node) of
+        undefined ->
+            pong = net_adm:ping(Node),
+            dctrl(Node);
+        DCtrl ->
+            DCtrl
+    end.
 
-dport_reg_send(Node, Name, Msg) ->
-    DPrt = case dport(Node) of
-               undefined ->
-                   pong = net_adm:ping(Node),
-                   dport(Node);
-               Prt ->
-                   Prt
-           end,
-    port_command(DPrt, [dmsg_hdr(),
-                        dmsg_ext({?DOP_REG_SEND,
-                                  self(),
-                                  ?COOKIE,
-                                  Name}),
-                        dmsg_ext(Msg)]).
-
-
-dport_send(To, Msg) ->
+dctrl_send(DPrt, Data) when is_port(DPrt) ->
+    port_command(DPrt, Data);
+dctrl_send(DPid, Data) when is_pid(DPid) ->
+    Ref = make_ref(),
+    DPid ! {send, self(), Ref, Data},
+    receive {Ref, Res} -> Res end.
+
+dctrl_dop_reg_send(Node, Name, Msg) ->
+    dctrl_send(ensure_dctrl(Node),
+               [dmsg_hdr(),
+                dmsg_ext({?DOP_REG_SEND,
+                          self(),
+                          ?COOKIE,
+                          Name}),
+                dmsg_ext(Msg)]).
+
+dctrl_dop_send(To, Msg) ->
     Node = node(To),
-    DPrt = case dport(Node) of
-               undefined ->
-                   pong = net_adm:ping(Node),
-                   dport(Node);
-               Prt ->
-                   Prt
-           end,
-    port_command(DPrt, [dmsg_hdr(),
-                        dmsg_ext({?DOP_SEND,
-                                  ?COOKIE,
-                                  To}),
-                        dmsg_ext(Msg)]).
+    dctrl_send(ensure_dctrl(Node),
+               [dmsg_hdr(),
+                dmsg_ext({?DOP_SEND, ?COOKIE, To}),
+                dmsg_ext(Msg)]).
+
 send_bad_structure(Offender,Victim,Bad,WhereToPutSelf) ->
     send_bad_structure(Offender,Victim,Bad,WhereToPutSelf,[]).
 send_bad_structure(Offender,Victim,Bad,WhereToPutSelf,PayLoad) ->
@@ -1779,7 +1862,7 @@ send_bad_structure(Offender,Victim,Bad,WhereToPutSelf,PayLoad) ->
                   Node = node(Victim),
                   pong = net_adm:ping(Node),
                   erlang:monitor_node(Node, true),
-                  DPrt = dport(Node),
+                  DCtrl = dctrl(Node),
                   Bad1 = case WhereToPutSelf of
                              0 ->
                                  Bad;
@@ -1797,7 +1880,7 @@ send_bad_structure(Offender,Victim,Bad,WhereToPutSelf,PayLoad) ->
                   after 10 -> ok
                   end,
 
-                  port_command(DPrt, DData),
+                  dctrl_send(DCtrl, DData),
 
                   receive {nodedown, Node} -> ok
                   after 5000 -> exit("missing nodedown")
@@ -1832,11 +1915,11 @@ send_bad_msgs(BadNode, To, Repeat, BadTerm) when is_atom(BadNode),
                fun () ->
                        Node = node(To),
                        pong = net_adm:ping(Node),
-                       DPrt = dport(Node),
+                       DCtrl = dctrl(Node),
                        DData = [dmsg_hdr(),
                                 dmsg_ext({?DOP_SEND, ?COOKIE, To}),
                                 BadTerm],
-                       repeat(fun () -> port_command(DPrt, DData) end, Repeat),
+		       repeat(fun () -> dctrl_send(DCtrl, DData) end, Repeat),
                        Parent ! Done
                end),
     receive Done -> ok end.
@@ -1858,11 +1941,12 @@ send_bad_ctl(BadNode, ToNode) when is_atom(BadNode), is_atom(ToNode) ->
                                        replace}),
                        CtlBeginSize = size(Ctl) - size(Replace),
                        <<CtlBegin:CtlBeginSize/binary, Replace/binary>> = Ctl,
-                       port_command(dport(ToNode),
-                                    [dmsg_fake_hdr2(),
-                                     CtlBegin,
-                                     dmsg_bad_atom_cache_ref(),
-                                     dmsg_ext({a, message})]),
+                       DCtrl = dctrl(ToNode),
+                       Data = [dmsg_fake_hdr2(),
+                               CtlBegin,
+                               dmsg_bad_atom_cache_ref(),
+                               dmsg_ext({a, message})],
+                       dctrl_send(DCtrl, Data),
                        Parent ! Done
                end),
     receive Done -> ok end.
@@ -1875,17 +1959,32 @@ send_bad_dhdr(BadNode, ToNode) when is_atom(BadNode), is_atom(ToNode) ->
     spawn_link(BadNode,
                fun () ->
                        pong = net_adm:ping(ToNode),
-                       port_command(dport(ToNode), dmsg_bad_hdr()),
+                       dctrl_send(dctrl(ToNode), dmsg_bad_hdr()),
                        Parent ! Done
                end),
     receive Done -> ok end.
 
-dport(Node) when is_atom(Node) ->
-    case catch erts_debug:get_internal_state(available_internal_state) of
-        true -> true;
-        _ -> erts_debug:set_internal_state(available_internal_state, true)
-    end,
-    erts_debug:get_internal_state({dist_port, Node}).
+dctrl(Node) when is_atom(Node) ->
+    get_internal_state({dist_ctrl, Node}).
+
+get_internal_state(Op) ->
+    try erts_debug:get_internal_state(Op) of
+        R -> R
+    catch
+        error:undef ->
+            erts_debug:set_internal_state(available_internal_state, true),
+            erts_debug:get_internal_state(Op)
+    end.
+
+set_internal_state(Op, Val) ->
+    try erts_debug:set_internal_state(Op, Val) of
+        R -> R
+    catch
+        error:undef ->
+            erts_debug:set_internal_state(available_internal_state, true),
+            erts_debug:set_internal_state(Op, Val)
+    end.
+
 
 dmsg_hdr() ->
     [131, % Version Magic
@@ -1987,6 +2086,11 @@ port_please(_Name, _Ip) ->
 	    {port, Port, Version}
     end.
 
+address_please(_Name, _Address, _AddressFamily) ->
+    %% Use localhost.
+    IP = {127,0,0,1},
+    {ok, IP}.
+
 %%% Utilities
 
 timestamp() ->
@@ -2028,11 +2132,9 @@ freeze_node(Node, MS) ->
     Freezer = self(),
     spawn_link(Node,
                fun () ->
-                       erts_debug:set_internal_state(available_internal_state,
-                                                     true),
-                       dport_send(Freezer, DoingIt),
+                       dctrl_dop_send(Freezer, DoingIt),
                        receive after Own -> ok end,
-                       erts_debug:set_internal_state(block, MS+Own)
+                       set_internal_state(block, MS+Own)
                end),
     receive DoingIt -> ok end,
     receive after Own -> ok end.
@@ -2082,7 +2184,7 @@ start_relay_node(Node, Args) ->
                                       [{args, Args ++
                                         " -setcookie "++Cookie++" -pa "++Pa++" "++
                                         RunArg}]),
-    [N,H] = string:tokens(atom_to_list(NN),"@"),
+    [N,H] = string:lexemes(atom_to_list(NN),"@"),
     {ok, Sock} = gen_tcp:accept(LSock),
     pang = net_adm:ping(NN),
     {N,H,Sock}.
@@ -2236,8 +2338,7 @@ forever(Fun) ->
     forever(Fun).
 
 abort(Why) ->
-    erts_debug:set_internal_state(available_internal_state, true),
-    erts_debug:set_internal_state(abort, Why).
+    set_internal_state(abort, Why).
 
 
 start_busy_dist_port_tracer() ->
diff --git a/erts/emulator/test/driver_SUITE.erl b/erts/emulator/test/driver_SUITE.erl
index c31ceb4d4b..40c7cc11e1 100644
--- a/erts/emulator/test/driver_SUITE.erl
+++ b/erts/emulator/test/driver_SUITE.erl
@@ -43,9 +43,9 @@
          outputv_errors/1,
          driver_unloaded/1,
          io_ready_exit/1,
+         use_fallback_pollset/0,
          use_fallback_pollset/1,
          bad_fd_in_pollset/1,
-         driver_event/1,
          fd_change/1,
          steal_control/1,
          otp_6602/1,
@@ -58,11 +58,9 @@
          ioq_exit_ready_output/1,
          ioq_exit_timeout/1,
          ioq_exit_ready_async/1,
-         ioq_exit_event/1,
          ioq_exit_ready_input_async/1,
          ioq_exit_ready_output_async/1,
          ioq_exit_timeout_async/1,
-         ioq_exit_event_async/1,
          zero_extended_marker_garb_drv/1,
          invalid_extended_marker_drv/1,
          larger_major_vsn_drv/1,
@@ -82,10 +80,13 @@
          async_blast/1,
          thr_msg_blast/1,
          consume_timeslice/1,
+         env/1,
          z_test/1]).
 
 -export([bin_prefix/2]).
 
+-export([get_check_io_total/1]).   % for z_SUITE.erl
+
 -include_lib("common_test/include/ct.hrl").
 
 
@@ -119,17 +120,26 @@
 -define(heap_binary_size, 64).
 
 init_per_testcase(Case, Config) when is_atom(Case), is_list(Config) ->
-    case catch erts_debug:get_internal_state(available_internal_state) of
-        true -> ok;
-        _ -> erts_debug:set_internal_state(available_internal_state, true)
-    end,
+    CIOD = rpc(Config,
+               fun() ->
+                       case catch erts_debug:get_internal_state(available_internal_state) of
+                           true -> ok;
+                           _ -> erts_debug:set_internal_state(available_internal_state, true)
+                       end,
+                       erts_debug:get_internal_state(check_io_debug)
+               end),
     erlang:display({init_per_testcase, Case}),
-    0 = element(1, erts_debug:get_internal_state(check_io_debug)),
+    0 = element(1, CIOD),
     [{testcase, Case}|Config].
 
-end_per_testcase(Case, _Config) ->
+end_per_testcase(Case, Config) ->
     erlang:display({end_per_testcase, Case}),
-    0 = element(1, erts_debug:get_internal_state(check_io_debug)),
+    CIOD = rpc(Config,
+               fun() ->
+                       get_stable_check_io_info(),
+                       erts_debug:get_internal_state(check_io_debug)
+               end),
+    0 = element(1, CIOD),
     ok.
 
 suite() ->
@@ -137,10 +147,13 @@ suite() ->
      {timetrap, {minutes, 1}}].
 
 all() -> %% Keep a_test first and z_test last...
-    [a_test, outputv_errors, outputv_echo, queue_echo, {group, timer},
-     driver_unloaded, io_ready_exit, use_fallback_pollset,
-     bad_fd_in_pollset, driver_event, fd_change,
-     steal_control, otp_6602, driver_system_info_base_ver,
+    [a_test, outputv_errors, outputv_echo, queue_echo,
+     {group, timer},
+     driver_unloaded, io_ready_exit, otp_6602,
+     {group, polling},
+     {group, poll_thread},
+     {group, poll_set},
+     driver_system_info_base_ver,
      driver_system_info_prev_ver,
      driver_system_info_current_ver, driver_monitor,
      {group, ioq_exit}, zero_extended_marker_garb_drv,
@@ -148,24 +161,31 @@ all() -> %% Keep a_test first and z_test last...
      larger_minor_vsn_drv, smaller_major_vsn_drv,
      smaller_minor_vsn_drv, peek_non_existing_queue,
      otp_6879, caller, many_events, missing_callbacks,
-     smp_select, driver_select_use,
      thread_mseg_alloc_cache_clean,
      otp_9302,
      thr_free_drv,
      async_blast,
      thr_msg_blast,
      consume_timeslice,
+     env,
      z_test].
 
 groups() -> 
     [{timer, [],
       [timer_measure, timer_cancel, timer_delay,
        timer_change]},
+     {poll_thread, [], [{group, polling}]},
+     {poll_set, [], [{group, polling}]},
+     {polling, [],
+      [a_test, use_fallback_pollset,
+       bad_fd_in_pollset, fd_change,
+       steal_control, smp_select,
+       driver_select_use, z_test]},
      {ioq_exit, [],
       [ioq_exit_ready_input, ioq_exit_ready_output,
-       ioq_exit_timeout, ioq_exit_ready_async, ioq_exit_event,
+       ioq_exit_timeout, ioq_exit_ready_async,
        ioq_exit_ready_input_async, ioq_exit_ready_output_async,
-       ioq_exit_timeout_async, ioq_exit_event_async]}].
+       ioq_exit_timeout_async]}].
 
 init_per_suite(Config) ->
     Config.
@@ -173,10 +193,28 @@ init_per_suite(Config) ->
 end_per_suite(_Config) ->
     catch erts_debug:set_internal_state(available_internal_state, false).
 
+init_per_group(poll_thread, Config) ->
+    [{node_args, "+IOt 2"} | Config];
+init_per_group(poll_set, Config) ->
+    [{node_args, "+IOt 2 +IOp 2"} | Config];
+init_per_group(polling, Config) ->
+    case proplists:get_value(node_args, Config) of
+        undefined ->
+            Config;
+        Args ->
+            {ok, Node} = start_node(polling, Args),
+            [{node, Node} | Config]
+    end;
 init_per_group(_GroupName, Config) ->
     Config.
 
 end_per_group(_GroupName, Config) ->
+    case proplists:get_value(node, Config) of
+        undefined ->
+            ok;
+        Node ->
+            stop_node(Node)
+    end,
     Config.
 
 %% Test sending bad types to port with an outputv-capable driver.
@@ -778,21 +816,23 @@ io_ready_exit(Config) when is_list(Config) ->
 -define(CHKIO_STOP, 0).
 -define(CHKIO_USE_FALLBACK_POLLSET, 1).
 -define(CHKIO_BAD_FD_IN_POLLSET, 2).
--define(CHKIO_DRIVER_EVENT, 3).
 -define(CHKIO_FD_CHANGE, 4).
 -define(CHKIO_STEAL, 5).
 -define(CHKIO_STEAL_AUX, 6).
 -define(CHKIO_SMP_SELECT, 7).
 -define(CHKIO_DRV_USE, 8).
 
+use_fallback_pollset() ->
+    [{timetrap, {minutes, 2}}].
+
 use_fallback_pollset(Config) when is_list(Config) ->
+    rpc(Config, fun() -> use_fallback_pollset_t(Config) end).
+
+use_fallback_pollset_t(Config) when is_list(Config) ->
     FlbkFun = fun () ->
-                      ChkIoDuring = erlang:system_info(check_io),
-                      case lists:keysearch(fallback_poll_set_size,
-                                           1,
-                                           ChkIoDuring) of
-                          {value,
-                           {fallback_poll_set_size, N}} when N > 0 ->
+                      {Flbk, _} = get_fallback(erlang:system_info(check_io)),
+                      case lists:keysearch(total_poll_set_size, 1, Flbk) of
+                          {value, {total_poll_set_size, N}} when N > 0 ->
                               ok;
                           Error ->
                               ct:fail({failed_to_use_fallback, Error})
@@ -814,6 +854,7 @@ use_fallback_pollset(Config) when is_list(Config) ->
           Skip ->
               {fun () -> ok end, Skip, ok}
       end,
+    io:format("Node = ~p~n",[node()]),
     case chkio_test_fini(chkio_test(Handel,
                                     ?CHKIO_USE_FALLBACK_POLLSET,
                                     fun () ->
@@ -825,27 +866,31 @@ use_fallback_pollset(Config) when is_list(Config) ->
     end.
 
 bad_fd_in_pollset(Config) when is_list(Config) ->
-    chkio_test_fini(chkio_test(chkio_test_init(Config),
-                               ?CHKIO_BAD_FD_IN_POLLSET,
-                               fun () -> sleep(1000) end)).
-
-driver_event(Config) when is_list(Config) ->
-    chkio_test_fini(chkio_test(chkio_test_init(Config),
-                               ?CHKIO_DRIVER_EVENT,
-                               fun () -> sleep(1000) end)).
+    rpc(Config,
+        fun() ->
+                chkio_test_fini(chkio_test(chkio_test_init(Config),
+                                           ?CHKIO_BAD_FD_IN_POLLSET,
+                                           fun () -> sleep(1000) end))
+        end).
 
 fd_change(Config) when is_list(Config) ->
-    chkio_test_fini(chkio_test(chkio_test_init(Config),
-                               ?CHKIO_FD_CHANGE,
-                               fun () -> sleep(1000) end)).
+    rpc(Config,
+        fun() ->
+                chkio_test_fini(chkio_test(chkio_test_init(Config),
+                                           ?CHKIO_FD_CHANGE,
+                                           fun () -> sleep(1000) end))
+        end).
 
 steal_control(Config) when is_list(Config) ->
-    chkio_test_fini(case chkio_test_init(Config) of
-                        {erts_poll_info, _} = Hndl ->
-                            steal_control_test(Hndl);
-                        Skip ->
-                            Skip
-                    end).
+    rpc(Config,
+        fun() ->
+                chkio_test_fini(case chkio_test_init(Config) of
+                                    {erts_poll_info, _} = Hndl ->
+                                        steal_control_test(Hndl);
+                                    Skip ->
+                                        Skip
+                                end)
+        end).
 
 steal_control_test(Hndl = {erts_poll_info, Before}) ->
     Port = open_chkio_port(),
@@ -887,7 +932,7 @@ chkio_test_init(Config) when is_list(Config) ->
     ChkIo = get_stable_check_io_info(),
     case catch lists:keysearch(name, 1, ChkIo) of
         {value, {name, erts_poll}} ->
-            io:format("Before test: ~p~n", [ChkIo]),
+            ct:log("Before test: ~p~n", [ChkIo]),
             Path = proplists:get_value(data_dir, Config),
             erl_ddll:start(),
             ok = load_driver(Path, 'chkio_drv'),
@@ -948,8 +993,9 @@ chkio_test({erts_poll_info, Before},
         "ok" ->
             chk_chkio_port(Port),
             Fun(),
-            During = erlang:system_info(check_io),
+            During = get_check_io_total(erlang:system_info(check_io)),
             erlang:display(During),
+
             0 = element(1, erts_debug:get_internal_state(check_io_debug)),
             io:format("During test: ~p~n", [During]),
             chk_chkio_port(Port),
@@ -1001,22 +1047,83 @@ verify_chkio_state(Before, After) ->
     ok.
 
 get_stable_check_io_info() ->
-    ChkIo = erlang:system_info(check_io),
-    PendUpdNo = case lists:keysearch(pending_updates, 1, ChkIo) of
-                    {value, {pending_updates, PendNo}} ->
-                        PendNo;
-                    false ->
-                        0
-                end,
-    {value, {active_fds, ActFds}} = lists:keysearch(active_fds, 1, ChkIo),
+    get_stable_check_io_info(10).
+get_stable_check_io_info(0) ->
+    get_check_io_total(erlang:system_info(check_io));
+get_stable_check_io_info(N) ->
+    ChkIo = get_check_io_total(erlang:system_info(check_io)),
+    PendUpdNo = proplists:get_value(pending_updates, ChkIo, 0),
+    ActFds = proplists:get_value(active_fds, ChkIo),
     case {PendUpdNo, ActFds} of
         {0, 0} ->
             ChkIo;
         _ ->
-            receive after 10 -> ok end,
-            get_stable_check_io_info()
+            receive after 100 -> ok end,
+            get_stable_check_io_info(N-1)
+    end.
+
+%% Merge return from erlang:system_info(check_io)
+%% as if it was one big pollset.
+get_check_io_total(ChkIo) ->
+    ct:log("ChkIo = ~p~n",[ChkIo]),
+    {Fallback, Rest} = get_fallback(ChkIo),
+    add_fallback_infos(Fallback,
+      lists:foldl(fun(Pollset, Acc) ->
+                          lists:zipwith(fun(A, B) ->
+                                                add_pollset_infos(A,B)
+                                        end,
+                                        Pollset, Acc)
+                  end,
+                  hd(Rest), tl(Rest))).
+
+add_pollset_infos({Tag, A}=TA , {Tag, B}=TB) ->
+    case tag_type(Tag) of
+        sum ->
+            {Tag, A + B};
+        const ->
+            case A of
+                B -> TA;
+                _ ->
+                    ct:fail("Unexpected diff in pollsets ~p != ~p",
+                            [TA,TB])
+            end
+    end.
+
+get_fallback([MaybeFallback | ChkIo] = AllChkIo) ->
+    case proplists:get_value(fallback, MaybeFallback) of
+        true ->
+            {MaybeFallback, ChkIo};
+        false ->
+            {undefined, AllChkIo}
     end.
 
+add_fallback_infos(undefined, Acc) ->
+    Acc;
+add_fallback_infos(Flbk, Acc) ->
+    lists:zipwith(fun({Tag, A}=TA, {Tag, B}=TB) ->
+                          case tag_type(Tag) of
+                              sum -> {Tag, A + B};
+                              const when Tag =:= fallback -> TA;
+                              const -> TB
+                          end
+                  end,
+                  Flbk, Acc).
+
+tag_type(name) -> const;
+tag_type(primary) -> const;
+tag_type(fallback) -> const;
+tag_type(kernel_poll) -> const;
+tag_type(memory_size) -> sum;
+tag_type(total_poll_set_size) -> sum;
+tag_type(lazy_updates) -> const;
+tag_type(pending_updates) -> sum;
+tag_type(batch_updates) -> const;
+tag_type(concurrent_updates) -> const;
+tag_type(max_fds) -> const;
+tag_type(active_fds) -> sum;
+tag_type(poll_threads) -> sum.
+
+
 %% Missed port lock when stealing control of fd from a
 %% driver that didn't use the same lock. The lock checker
 %% used to trigger on this and dump core.
@@ -1087,9 +1194,9 @@ check_driver_system_info_result(Result) ->
     io:format("All names: ~p~n", [?EXPECTED_SYSTEM_INFO_NAMES]),
     io:format("Result: ~p~n", [Result]),
     {[], Ns, DDVSN} = chk_sis(lists:map(fun (Str) ->
-                                                string:tokens(Str, "=")
+                                                string:lexemes(Str, "=")
                                         end,
-                                        string:tokens(Result, " ")),
+                                        string:lexemes(Result, " ")),
                               ?EXPECTED_SYSTEM_INFO_NAMES),
     case {DDVSN,
           drv_vsn_str2tup(erlang:system_info(driver_version))} of
@@ -1144,8 +1251,6 @@ check_si_res(["thread", "false"]) ->
     false = erlang:system_info(threads);
 check_si_res(["smp", "true"]) ->
     true = erlang:system_info(smp_support);
-check_si_res(["smp", "false"]) ->
-    false = erlang:system_info(smp_support);
 
 %% Data added in second version of driver_system_info() (driver version 1.1)
 check_si_res(["async_thrs", Value]) ->
@@ -1338,11 +1443,9 @@ driver_monitor(Config) when is_list(Config) ->
 -define(IOQ_EXIT_READY_OUTPUT, 2).
 -define(IOQ_EXIT_TIMEOUT, 3).
 -define(IOQ_EXIT_READY_ASYNC, 4).
--define(IOQ_EXIT_EVENT, 5).
 -define(IOQ_EXIT_READY_INPUT_ASYNC, 6).
 -define(IOQ_EXIT_READY_OUTPUT_ASYNC, 7).
 -define(IOQ_EXIT_TIMEOUT_ASYNC, 8).
--define(IOQ_EXIT_EVENT_ASYNC, 9).
 
 ioq_exit_test(Config, TestNo) ->
     Drv = ioq_exit_drv,
@@ -1395,9 +1498,6 @@ ioq_exit_timeout(Config) when is_list(Config) ->
 ioq_exit_ready_async(Config) when is_list(Config) ->
     ioq_exit_test(Config, ?IOQ_EXIT_READY_ASYNC).
 
-ioq_exit_event(Config) when is_list(Config) ->
-    ioq_exit_test(Config, ?IOQ_EXIT_EVENT).
-
 ioq_exit_ready_input_async(Config) when is_list(Config) ->
     ioq_exit_test(Config, ?IOQ_EXIT_READY_INPUT_ASYNC).
 
@@ -1407,9 +1507,6 @@ ioq_exit_ready_output_async(Config) when is_list(Config) ->
 ioq_exit_timeout_async(Config) when is_list(Config) ->
     ioq_exit_test(Config, ?IOQ_EXIT_TIMEOUT_ASYNC).
 
-ioq_exit_event_async(Config) when is_list(Config) ->
-    ioq_exit_test(Config, ?IOQ_EXIT_EVENT_ASYNC).
-
 
 vsn_mismatch_test(Config, LoadResult) ->
     Path = proplists:get_value(data_dir, Config),
@@ -1643,7 +1740,7 @@ missing_callbacks(Config) when is_list(Config) ->
 smp_select(Config) when is_list(Config) -> 
     case os:type() of
         {win32,_} -> {skipped, "Test not implemented for this OS"};
-        _ -> smp_select0(Config)
+        _ -> rpc(Config, fun() -> smp_select0(Config) end)
     end.
 
 smp_select0(Config) ->
@@ -1702,7 +1799,7 @@ smp_select_wait(Pids, TimeoutMsg) ->
 driver_select_use(Config) when is_list(Config) -> 
     case os:type() of
         {win32,_} -> {skipped, "Test not implemented for this OS"};
-        _ -> driver_select_use0(Config)
+        _ -> rpc(Config, fun() -> driver_select_use0(Config) end)
     end.
 
 driver_select_use0(Config) ->
@@ -1947,44 +2044,39 @@ thr_msg_blast_receiver_proc(Port, Max, Parent, Done) ->
     end.
 
 thr_msg_blast(Config) when is_list(Config) ->
-    case erlang:system_info(smp_support) of
-        false ->
-            {skipped, "Non-SMP emulator; nothing to test..."};
-        true ->
-            Path = proplists:get_value(data_dir, Config),
-            erl_ddll:start(),
-            ok = load_driver(Path, thr_msg_blast_drv),
-            MemBefore = driver_alloc_size(),
-            Start = os:timestamp(),
-            Port = open_port({spawn, thr_msg_blast_drv}, []),
-            true = is_port(Port),
-            Done = make_ref(),
-            Me = self(),
-            spawn(fun () ->
-                          thr_msg_blast_receiver_proc(Port, 1, Me, Done)
-                  end),
-            receive
-                Done -> ok
-            end,
-            ok = thr_msg_blast_receiver(Port, 0, 32*10000),
-            port_close(Port),
-            End = os:timestamp(),
-            receive
-                Garbage ->
-                    ct:fail({received_garbage, Port, Garbage})
-            after 2000 ->
-                      ok
-            end,
-            MemAfter = driver_alloc_size(),
-            io:format("MemBefore=~p, MemAfter=~p~n",
-                      [MemBefore, MemAfter]),
-            ThrMsgBlastTime = timer:now_diff(End,Start)/1000000,
-            io:format("ThrMsgBlastTime=~p~n", [ThrMsgBlastTime]),
-            MemBefore = MemAfter,
-            Res = {thr_msg_blast_time, ThrMsgBlastTime},
-            erlang:display(Res),
-            Res
-    end.
+    Path = proplists:get_value(data_dir, Config),
+    erl_ddll:start(),
+    ok = load_driver(Path, thr_msg_blast_drv),
+    MemBefore = driver_alloc_size(),
+    Start = os:timestamp(),
+    Port = open_port({spawn, thr_msg_blast_drv}, []),
+    true = is_port(Port),
+    Done = make_ref(),
+    Me = self(),
+    spawn(fun () ->
+                  thr_msg_blast_receiver_proc(Port, 1, Me, Done)
+          end),
+    receive
+        Done -> ok
+    end,
+    ok = thr_msg_blast_receiver(Port, 0, 32*10000),
+    port_close(Port),
+    End = os:timestamp(),
+    receive
+        Garbage ->
+            ct:fail({received_garbage, Port, Garbage})
+    after 2000 ->
+            ok
+    end,
+    MemAfter = driver_alloc_size(),
+    io:format("MemBefore=~p, MemAfter=~p~n",
+              [MemBefore, MemAfter]),
+    ThrMsgBlastTime = timer:now_diff(End,Start)/1000000,
+    io:format("ThrMsgBlastTime=~p~n", [ThrMsgBlastTime]),
+    MemBefore = MemAfter,
+    Res = {thr_msg_blast_time, ThrMsgBlastTime},
+    erlang:display(Res),
+    Res.
 
 -define(IN_RANGE(LoW_, VaLuE_, HiGh_),
         case in_range(LoW_, VaLuE_, HiGh_) of
@@ -2273,11 +2365,56 @@ count_proc_sched(Ps, PNs) ->
               PNs
     end.
 
+%%
+%% Tests whether erl_drv_putenv reflects in os:getenv and vice versa.
+%%
+env(Config) when is_list(Config) ->
+    ok = load_driver(proplists:get_value(data_dir, Config), env_drv),
+    Port = open_port({spawn_driver, env_drv}, []),
+    true = is_port(Port),
+
+    Keys = ["env_drv_a_key", "env_drv_b_key", "env_drv_c_key"],
+    Values = ["a_value", "b_value", "c_value"],
+
+    [env_put_test(Port, Key, Value) || Key <- Keys, Value <- Values],
+    [env_get_test(Port, Key, Value) || Key <- Keys, Value <- Values],
+    [env_oversize_test(Port, Key) || Key <- Keys],
+    [env_notfound_test(Port, Key) || Key <- Keys],
+
+    true = port_close(Port),
+    erl_ddll:unload_driver(env_drv),
+    ok.
+
+env_control(Port, Command, Key, Value) ->
+    KeyBin = list_to_binary(Key),
+    ValueBin = list_to_binary(Value),
+    Header = <<(byte_size(KeyBin)), (byte_size(ValueBin))>>,
+    Payload = <<KeyBin/binary, ValueBin/binary>>,
+    port_control(Port, Command, <<Header/binary, Payload/binary>>).
+
+env_put_test(Port, Key, Value) ->
+    os:unsetenv(Key),
+    [0] = env_control(Port, 0, Key, Value),
+    Value = os:getenv(Key).
+
+env_get_test(Port, Key, ExpectedValue) ->
+    true = os:putenv(Key, ExpectedValue),
+    [0] = env_control(Port, 1, Key, ExpectedValue).
+
+env_oversize_test(Port, Key) ->
+    os:putenv(Key, [$A || _ <- lists:seq(1, 1024)]),
+    [127] = env_control(Port, 1, Key, "").
+
+env_notfound_test(Port, Key) ->
+    true = os:unsetenv(Key),
+    [255] = env_control(Port, 1, Key, "").
+
+
 a_test(Config) when is_list(Config) ->
-    check_io_debug().
+    rpc(Config, fun check_io_debug/0).
 
 z_test(Config) when is_list(Config) ->
-    check_io_debug().
+    rpc(Config, fun check_io_debug/0).
 
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 %% 		Utilities
@@ -2285,8 +2422,8 @@ z_test(Config) when is_list(Config) ->
 
 check_io_debug() ->
     get_stable_check_io_info(),
-    {NoErrorFds, NoUsedFds, NoDrvSelStructs, NoDrvEvStructs} = CheckIoDebug
-    = erts_debug:get_internal_state(check_io_debug),
+    {NoErrorFds, NoUsedFds, NoDrvSelStructs, NoEnifSelStructs}
+        = CheckIoDebug = erts_debug:get_internal_state(check_io_debug),
     HasGetHost = has_gethost(),
     ct:log("check_io_debug: ~p~n"
            "HasGetHost: ~p",[CheckIoDebug, HasGetHost]),
@@ -2299,7 +2436,7 @@ check_io_debug() ->
             %% one extra used fd that is not selected on
             ok
     end,
-    0 = NoDrvEvStructs,
+    0 = NoEnifSelStructs,
     ok.
 
 has_gethost() ->
@@ -2351,7 +2488,7 @@ wait_until(Fun) ->
     end.
 
 drv_vsn_str2tup(Str) ->
-    [Major, Minor] = string:tokens(Str, "."),
+    [Major, Minor] = string:lexemes(Str, "."),
     {list_to_integer(Major), list_to_integer(Minor)}.
 
 %% Build port data from a template.
@@ -2452,8 +2589,18 @@ stop_driver(Port, Name) ->
     ok = erl_ddll:stop().
 
 load_driver(Dir, Driver) ->
+    Before = erlang:system_info(taints),
     case erl_ddll:load_driver(Dir, Driver) of
-        ok -> ok;
+        ok ->
+            After = erlang:system_info(taints),
+            case lists:member(Driver, Before) of
+                true ->
+                    After = Before;
+                false ->
+                    true = lists:member(Driver, After),
+                    Before = lists:delete(Driver, After)
+            end,
+            ok;
         {error, Error} = Res ->
             io:format("~s\n", [erl_ddll:format_error(Error)]),
             Res
@@ -2469,15 +2616,19 @@ sleep(Ms) when is_integer(Ms), Ms >= 0 ->
 
 
 start_node(Config) when is_list(Config) ->
+    start_node(proplists:get_value(testcase, Config));
+start_node(Name) ->
+    start_node(Name, "").
+start_node(NodeName, Args) ->
     Pa = filename:dirname(code:which(?MODULE)),
     Name = list_to_atom(atom_to_list(?MODULE)
                         ++ "-"
-                        ++ atom_to_list(proplists:get_value(testcase, Config))
+                        ++ atom_to_list(NodeName)
                         ++ "-"
                         ++ integer_to_list(erlang:system_time(second))
                         ++ "-"
                         ++ integer_to_list(erlang:unique_integer([positive]))),
-    test_server:start_node(Name, slave, [{args, "-pa "++Pa}]).
+    test_server:start_node(Name, slave, [{args, Args ++ " -pa "++Pa}]).
 
 stop_node(Node) ->
     test_server:stop_node(Node).
@@ -2491,14 +2642,6 @@ wait_deallocations() ->
     end.
 
 driver_alloc_size() ->
-    case erlang:system_info(smp_support) of
-        true ->
-            ok;
-        false ->
-            %% driver_alloc also used by elements in lock-free queues,
-            %% give these some time to be deallocated...
-            receive after 100 -> ok end
-    end,
     wait_deallocations(),
     case erlang:system_info({allocator_sizes, driver_alloc}) of
         false ->
@@ -2518,3 +2661,35 @@ driver_alloc_size() ->
                       Sz0+Sz
               end, 0, CS)
     end.
+
+rpc(Config, Fun) ->
+    case proplists:get_value(node, Config) of
+        undefined ->
+            Fun();
+        Node ->
+            Self = self(),
+            Ref = make_ref(),
+            Pid = spawn(Node,
+                        fun() ->
+                                Result
+                                    = try Fun() of
+                                          Res -> Res
+                                      catch E:R:Stk ->
+                                              {'EXIT',E,R,Stk}
+                                      end,
+                                Self ! {Ref, Result}
+                        end),
+            MRef = monitor(process, Pid),
+            receive
+                {'DOWN', MRef, _Type, _Object, Info} ->
+                    erlang:error({died, Pid, Info});
+                {Ref, {'EXIT',E,R,ST}} ->
+                    erlang:demonitor(MRef, [flush]),
+                    erlang:raise(E,R,ST);
+                {Ref, Ret} ->
+                    erlang:demonitor(MRef, [flush]),
+                    Ret;
+                Other ->
+                    ct:fail(Other)
+            end
+    end.
diff --git a/erts/emulator/test/driver_SUITE_data/Makefile.src b/erts/emulator/test/driver_SUITE_data/Makefile.src
index 1fedd72200..bcabaa689d 100644
--- a/erts/emulator/test/driver_SUITE_data/Makefile.src
+++ b/erts/emulator/test/driver_SUITE_data/Makefile.src
@@ -16,7 +16,8 @@ MISC_DRVS =		outputv_drv@dll@ \
 			thr_free_drv@dll@ \
 			async_blast_drv@dll@ \
 			thr_msg_blast_drv@dll@ \
-			consume_timeslice_drv@dll@
+			consume_timeslice_drv@dll@ \
+			env_drv@dll@
 
 SYS_INFO_DRVS = 	sys_info_base_drv@dll@ \
 			sys_info_prev_drv@dll@ \
diff --git a/erts/emulator/test/driver_SUITE_data/chkio_drv.c b/erts/emulator/test/driver_SUITE_data/chkio_drv.c
index e55b9e10ba..ee8f28e8b1 100644
--- a/erts/emulator/test/driver_SUITE_data/chkio_drv.c
+++ b/erts/emulator/test/driver_SUITE_data/chkio_drv.c
@@ -42,7 +42,6 @@
 #define CHKIO_STOP				0
 #define CHKIO_USE_FALLBACK_POLLSET		1
 #define CHKIO_BAD_FD_IN_POLLSET			2
-#define CHKIO_DRIVER_EVENT			3
 #define CHKIO_FD_CHANGE				4
 #define CHKIO_STEAL				5
 #define CHKIO_STEAL_AUX				6
@@ -67,15 +66,6 @@ typedef struct {
 } ChkioFallbackData;
 
 typedef struct {
-    int in_fd;
-    struct erl_drv_event_data in_data;
-    int in_ok;
-    int out_fd;
-    struct erl_drv_event_data out_data;
-    int out_ok;
-} ChkioDriverEvent;
-
-typedef struct {
     int fds[2];
     int same_fd;
 } ChkioFdChange;
@@ -86,14 +76,10 @@ typedef struct {
 
 typedef struct {
     int driver_select_fds[2];
-    int driver_event_fds[2];
-    struct erl_drv_event_data event_data[2];
 } ChkioSteal;
 
 typedef struct {
     int driver_select_fds[2];
-    int driver_event_fds[2];
-    struct erl_drv_event_data event_data[2];
 } ChkioStealAux;
 
 
@@ -141,7 +127,6 @@ static ErlDrvData chkio_drv_start(ErlDrvPort, char *);
 static void chkio_drv_stop(ErlDrvData);
 static void chkio_drv_ready_input(ErlDrvData, ErlDrvEvent);
 static void chkio_drv_ready_output(ErlDrvData, ErlDrvEvent);
-static void chkio_drv_ready_event(ErlDrvData, ErlDrvEvent, ErlDrvEventData);
 static ErlDrvSSizeT chkio_drv_control(ErlDrvData, unsigned int,
 				      char *, ErlDrvSizeT, char **, ErlDrvSizeT);
 static void chkio_drv_timeout(ErlDrvData);
@@ -164,7 +149,7 @@ static ErlDrvEntry chkio_drv_entry = {
     NULL, /* ready_async */
     NULL, /* flush */
     NULL, /* call */
-    chkio_drv_ready_event,
+    NULL, /* unused_event_callback */
 
     ERL_DRV_EXTENDED_MARKER,
     ERL_DRV_EXTENDED_MAJOR_VERSION,
@@ -243,25 +228,6 @@ stop_use_fallback_pollset(ChkioDrvData *cddp)
 }
 
 static void
-stop_driver_event(ChkioDrvData *cddp)
-{
-    if (cddp->test_data) {
-	ChkioDriverEvent *cdep = cddp->test_data;
-	cddp->test_data = NULL;
-
-	if (cdep->in_fd >= 0) {
-	    driver_event(cddp->port, (ErlDrvEvent) (ErlDrvSInt) cdep->in_fd, NULL);
-	    close(cdep->in_fd);
-	}
-	if (cdep->out_fd >= 0) {
-	    driver_event(cddp->port, (ErlDrvEvent) (ErlDrvSInt) cdep->out_fd, NULL);
-	    close(cdep->out_fd);
-	}
-	driver_free(cdep);	    
-    }
-}
-
-static void
 stop_fd_change(ChkioDrvData *cddp)
 {
     if (cddp->test_data) {
@@ -305,14 +271,6 @@ stop_steal(ChkioDrvData *cddp)
 			  (ErlDrvEvent) (ErlDrvSInt) csp->driver_select_fds[1],
 			  DO_WRITE,
 			  0);
-	if (csp->driver_event_fds[0] >= 0)
-	    driver_event(cddp->port,
-			 (ErlDrvEvent) (ErlDrvSInt) csp->driver_event_fds[0],
-			 NULL);
-	if (csp->driver_event_fds[1] >= 0)
-	    driver_event(cddp->port,
-			 (ErlDrvEvent) (ErlDrvSInt) csp->driver_event_fds[1],
-			 NULL);
 	driver_free(csp);
     }
 }
@@ -327,10 +285,6 @@ stop_steal_aux(ChkioDrvData *cddp)
 	    close(csap->driver_select_fds[0]);
 	if (csap->driver_select_fds[1] >= 0)
 	    close(csap->driver_select_fds[1]);
-	if (csap->driver_event_fds[0] >= 0)
-	    close(csap->driver_event_fds[0]);
-	if (csap->driver_event_fds[1] >= 0)
-	    close(csap->driver_event_fds[1]);
 	driver_free(csap);
     }
 }
@@ -354,10 +308,13 @@ static void free_smp_select(ChkioSmpSelect* pip, ErlDrvPort port)
 	abort();
     }
     case Selected:
-	driver_select(port, (ErlDrvEvent)(ErlDrvSInt)pip->read_fd, DO_READ, 0);
-	/*fall through*/ 
     case Opened:
-	close(pip->read_fd);
+        TRACEF(("%T: Close pipe [%d->%d]\n", driver_mk_port(port), pip->write_fd,
+                pip->read_fd));
+        if (pip->wasSelected)
+            driver_select(port, (ErlDrvEvent)(ErlDrvSInt)pip->read_fd, DO_READ|ERL_DRV_USE, 0);
+        else
+            close(pip->read_fd);
 	close(pip->write_fd);
 	pip->state = Closed;
 	break;
@@ -445,9 +402,6 @@ chkio_drv_stop(ErlDrvData drv_data) {
     case CHKIO_BAD_FD_IN_POLLSET:
 	stop_bad_fd_in_pollset(cddp);
 	break;
-    case CHKIO_DRIVER_EVENT:
-	stop_driver_event(cddp);
-	break;
     case CHKIO_FD_CHANGE:
 	stop_fd_change(cddp);
 	break;
@@ -557,6 +511,9 @@ chkio_drv_ready_input(ErlDrvData drv_data, ErlDrvEvent event)
 	    driver_failure_atom(cddp->port, "input_fd_not_found");
 	break;
     }
+    case CHKIO_FD_CHANGE:
+        /* This may be triggered when an fd is closed while being selected on. */
+        break;
     case CHKIO_STEAL:
 	break;
     case CHKIO_STEAL_AUX:
@@ -621,55 +578,6 @@ chkio_drv_ready_input(ErlDrvData drv_data, ErlDrvEvent event)
 }
 
 static void
-chkio_drv_ready_event(ErlDrvData drv_data,
-		      ErlDrvEvent event,
-		      ErlDrvEventData event_data)
-{
-#ifdef UNIX
-    ChkioDrvData *cddp = (ChkioDrvData *) drv_data;
-    switch (cddp->test) {
-    case CHKIO_DRIVER_EVENT: {
-#ifdef HAVE_POLL_H
-	ChkioDriverEvent *cdep = cddp->test_data;
-	int fd = (int) (ErlDrvSInt) event;
-	if (fd == cdep->in_fd) {
-	    if (event_data->events == POLLIN
-		&& event_data->revents == POLLIN) {
-		cdep->in_ok++;
-	    }
-	    else {
-		driver_failure_atom(cddp->port, "invalid_input_fd_events");
-	    }
-	    break;
-	}
-	if (fd == cdep->out_fd) {
-	    if (event_data->events == POLLOUT
-		&& event_data->revents == POLLOUT) {
-		cdep->out_ok++;
-	    }
-	    else {
-		driver_failure_atom(cddp->port, "invalid_output_fd_events");
-	    }
-	    break;
-	}
-#endif
-    }
-    case CHKIO_STEAL:
-#ifdef HAVE_POLL_H
-	break;
-#endif
-    case CHKIO_STEAL_AUX:
-#ifdef HAVE_POLL_H
-	break;
-#endif
-    default:
-	driver_failure_atom(cddp->port, "unexpected_ready_event");
-	break;
-    }
-#endif /* UNIX */
-}
-
-static void
 chkio_drv_timeout(ErlDrvData drv_data)
 {
 #ifdef UNIX
@@ -779,25 +687,6 @@ chkio_drv_control(ErlDrvData drv_data,
 	    res_len = -1;
 	    stop_bad_fd_in_pollset(cddp);
 	    break;
-	case CHKIO_DRIVER_EVENT: {
-	    ChkioDriverEvent *cdep = cddp->test_data;
-	    if (!cdep->in_ok || !cdep->out_ok) {
-		if (!cdep->in_ok)
-		    driver_failure_atom(cddp->port, "got_no_input_events");
-		if (!cdep->out_ok)
-		    driver_failure_atom(cddp->port, "got_no_output_events");
-	    }
-	    else {
-		char *c = driver_alloc(sizeof(char)*2*30);
-		if (!c)
-		    driver_failure_posix(cddp->port, ENOMEM);
-		*rbuf = c;
-		res_len = sprintf(c, "in=%d\nout=%d\n",
-				  cdep->in_ok, cdep->out_ok);
-	    }
-	    stop_driver_event(cddp);
-	    break;
-	}
 	case CHKIO_FD_CHANGE: {
 	    ChkioFdChange *cfcp = cddp->test_data;
 	    if (!cfcp->same_fd)
@@ -937,69 +826,6 @@ chkio_drv_control(ErlDrvData drv_data,
 	res_len = -1;
 	break;
     }
-    case CHKIO_DRIVER_EVENT: {
-#ifndef HAVE_POLL_H
-	res_str = "skip: Need the poll.h header for this test, but it doesn't exist";
-	res_len = -1;
-#else /* HAVE_POLL_H */
-	int in_fd = open("/dev/zero", O_RDONLY);
-	int out_fd = open("/dev/null", O_WRONLY);
-
-	if (in_fd < 0 || out_fd < 0) {
-	    if (in_fd >= 0)
-		close(in_fd);
-	    if (out_fd >= 0)
-		close(out_fd);
-	    driver_failure_posix(cddp->port, errno);
-	}
-	else {
-	    ChkioDriverEvent *cdep = driver_alloc(sizeof(ChkioDriverEvent));
-	    if (!cdep)
-		driver_failure_posix(cddp->port, ENOMEM);
-	    else {
-		int res;
-		cddp->test_data = cdep;
-
-		cdep->in_fd = in_fd;
-		cdep->in_data.events = POLLIN;
-		cdep->in_data.revents = 0;
-		cdep->in_ok = 0;
-
-		res = driver_event(cddp->port,
-				   (ErlDrvEvent) (ErlDrvSInt) in_fd,
-				   &cdep->in_data);
-		if (res < 0) {
-		    res_str = "skip: driver_event() not supported";
-		    res_len = -1;
-		    close(in_fd);
-		    close(out_fd);
-		    cdep->in_fd = -1;
-		    cdep->out_fd = -1;
-		}
-		else {
-		    res_str = "ok";
-		    res_len = -1;
-
-		    cdep->out_fd = out_fd;
-		    cdep->out_data.events = POLLOUT;
-		    cdep->out_data.revents = 0;
-		    cdep->out_ok = 0;
-
-		    res = driver_event(cddp->port,
-				       (ErlDrvEvent) (ErlDrvSInt) out_fd,
-				       &cdep->out_data);
-		    if (res < 0) {
-			close(out_fd);
-			cdep->out_fd = -1;
-			driver_failure_atom(cddp->port, "driver_event_failed");
-		    }
-		}
-
-	    }
-	}
-#endif /* HAVE_POLL_H */
-	break;
-    }
     case CHKIO_FD_CHANGE: {
 	ChkioFdChange *cfcp = driver_alloc(sizeof(ChkioFdChange));
 	if (!cfcp)
@@ -1028,58 +854,19 @@ chkio_drv_control(ErlDrvData drv_data,
 	    res_len = -1;
 	}
 	else {
-	    int driver_event_fds[2];
 	    int driver_select_fds[2];
 	    cddp->test_data = csp; 
 	    memcpy(c, buf, len);
 	    c[len] = '\0';
 	    if (sscanf(c,
-		       "fds:%d:%d:%d:%d",
+		       "fds:%d:%d",
 		       &driver_select_fds[0],
-		       &driver_select_fds[1],
-		       &driver_event_fds[0],
-		       &driver_event_fds[1]) != 4)
-		driver_failure_atom(cddp->port, "bad_input");
+		       &driver_select_fds[1]) != 2)
+                driver_failure_atom(cddp->port, "bad_input");
 	    else {
 		int res = 0;
-		if (driver_event_fds[0] < 0) { /* Have no working driver_event() ... */
-		    csp->driver_select_fds[0] = driver_select_fds[0]; /* In */
-		    csp->driver_select_fds[1] = driver_select_fds[1]; /* Out */
-		    csp->driver_event_fds[0] = -1;
-		    csp->driver_event_fds[1] = -1;
-		}
-		else { /* Have working driver_event() ... */
-#ifndef HAVE_POLL_H
-		    driver_failure_atom(cddp->port, "unexpected_result");
-		    res = -1;
-#else
-		    csp->driver_select_fds[0] = driver_select_fds[0]; /* In */
-		    csp->driver_event_fds[1] = driver_select_fds[1]; /* Out */
-		    csp->driver_event_fds[0] = driver_event_fds[0]; /* In */
-		    csp->driver_select_fds[1] = driver_event_fds[1]; /* Out */
-
-		    /* Steal with driver_event() */
-
-		    csp->event_data[0].events = POLLIN;
-		    csp->event_data[0].revents = 0;
-		    res = driver_event(cddp->port,
-				       (ErlDrvEvent) (ErlDrvSInt) csp->driver_event_fds[0],
-				       &csp->event_data[0]);
-		    if (res < 0)
-			driver_failure_atom(cddp->port,
-					    "driver_event_failed_to_steal");
-		    if (res >= 0) {
-			csp->event_data[1].events = POLLOUT;
-			csp->event_data[1].revents = 0;
-			res = driver_event(cddp->port,
-					   (ErlDrvEvent) (ErlDrvSInt) csp->driver_event_fds[1],
-					   &csp->event_data[1]);
-			if (res < 0)
-			    driver_failure_atom(cddp->port,
-						"driver_event_failed_to_steal");
-		    }
-#endif
-		}
+                csp->driver_select_fds[0] = driver_select_fds[0]; /* In */
+                csp->driver_select_fds[1] = driver_select_fds[1]; /* Out */
 
 		/* Steal with driver_select() */
 		if (res >= 0) {
@@ -1109,37 +896,17 @@ chkio_drv_control(ErlDrvData drv_data,
 	break;
     }
     case CHKIO_STEAL_AUX: {
-	int read_fds[2];
-	int write_fds[2];
+	int read_fd;
+	int write_fd;
 
-	read_fds[0] = open("/dev/zero", O_RDONLY);
-	write_fds[0] = open("/dev/null", O_WRONLY);
-
-#ifdef HAVE_POLL_H
-	read_fds[1] = open("/dev/zero", O_RDONLY);
-	write_fds[1] = open("/dev/null", O_WRONLY);
-#else
-	read_fds[1] = -1;
-	write_fds[1] = -1;
-#endif
+	read_fd = open("/dev/zero", O_RDONLY);
+	write_fd = open("/dev/null", O_WRONLY);
 	
-	if (read_fds[0] < 0
-	    || write_fds[0] < 0
-#ifdef HAVE_POLL_H
-	    || read_fds[1] < 0 
-	    || write_fds[1] < 0
-#endif
-	    ) {
-	    if (read_fds[0] < 0)
-		close(read_fds[0]);
-	    if (write_fds[0] < 0)
-		close(write_fds[0]);
-#ifdef HAVE_POLL_H
-	    if (read_fds[1] < 0)
-		close(read_fds[1]);
-	    if (write_fds[1] < 0)
-		close(write_fds[1]);
-#endif
+	if (read_fd < 0 || write_fd < 0) {
+	    if (read_fd < 0)
+		close(read_fd);
+	    if (write_fd < 0)
+		close(write_fd);
 	    driver_failure_posix(cddp->port, errno);
 	}
 	else {
@@ -1153,11 +920,8 @@ chkio_drv_control(ErlDrvData drv_data,
 		int res;
 		cddp->test_data = csap;
 
-		csap->driver_select_fds[0] = read_fds[0];
-		csap->driver_select_fds[1] = write_fds[0];
-
-		csap->driver_event_fds[0] = read_fds[1];
-		csap->driver_event_fds[1] = write_fds[1];
+		csap->driver_select_fds[0] = read_fd;
+		csap->driver_select_fds[1] = write_fd;
 
 		res = driver_select(cddp->port,
 				    (ErlDrvEvent) (ErlDrvSInt) csap->driver_select_fds[0],
@@ -1173,32 +937,6 @@ chkio_drv_control(ErlDrvData drv_data,
 		    if (res < 0)
 			driver_failure_atom(cddp->port, "driver_select_failed");
 		}
-#ifdef HAVE_POLL_H
-		if (res >= 0) {
-		    csap->event_data[0].events = POLLIN;
-		    csap->event_data[0].revents = 0;
-		    res = driver_event(cddp->port,
-				       (ErlDrvEvent) (ErlDrvSInt) csap->driver_event_fds[0],
-				       &csap->event_data[0]);
-		    if (res < 0) {
-			close(csap->driver_event_fds[0]);
-			csap->driver_event_fds[0] = -1;
-			close(csap->driver_event_fds[1]);
-			csap->driver_event_fds[1] = -1;
-			res = 0;
-		    }
-		    else {
-			csap->event_data[1].events = POLLOUT;
-			csap->event_data[1].revents = 0;
-			res = driver_event(cddp->port,
-					   (ErlDrvEvent) (ErlDrvSInt) csap->driver_event_fds[1],
-					   &csap->event_data[1]);
-			if (res < 0)
-			    driver_failure_atom(cddp->port,
-						"driver_event_failed");
-		    }
-		}
-#endif
 		if (res < 0) {
 		    res_str = "error";
 		    res_len = -1;
@@ -1213,11 +951,9 @@ chkio_drv_control(ErlDrvData drv_data,
 		    else {
 			*rbuf = c;
 			res_len = sprintf(c,
-					  "fds:%d:%d:%d:%d",
+					  "fds:%d:%d",
 					  csap->driver_select_fds[0],
-					  csap->driver_select_fds[1],
-					  csap->driver_event_fds[0],
-					  csap->driver_event_fds[1]);
+					  csap->driver_select_fds[1]);
 		    }
 		}
 	    }
@@ -1256,7 +992,7 @@ chkio_drv_control(ErlDrvData drv_data,
 		}
 		TRACEF(("%T: Created pipe [%d->%d]\n", cddp->id, fds[1], fds[0]));
 		pip->read_fd = fds[0];
-		pip->write_fd = fds[1];	       
+		pip->write_fd = fds[1];
 		pip->state = Opened;
 		pip->wasSelected = 0;
 		pip->next_write = pip->next_read = rand_r(&pip->rand_state) % 1024;
@@ -1266,7 +1002,8 @@ chkio_drv_control(ErlDrvData drv_data,
 	    }/*fall through*/
 	    case Opened: {
 		if (op & 1) {
-		    TRACEF(("%T: Write %d to opened pipe [%d->%d]\n", cddp->id, pip->next_write, pip->write_fd, pip->read_fd));
+		    TRACEF(("%T: Write %d to opened pipe [%d->%d]\n", cddp->id,
+                            pip->next_write, pip->write_fd, pip->read_fd));
 		    if (write(pip->write_fd, &pip->next_write, sizeof(int)) != sizeof(int)) {
 			fprintf(stderr, "Failed to write to pipe fd=%d, errno=%d\n", pip->write_fd, errno);
 			abort();
@@ -1275,8 +1012,12 @@ chkio_drv_control(ErlDrvData drv_data,
 		}
 		op >>= 1;
 		if (pip->wasSelected && (op & 1)) {
-		    TRACEF(("%T: Close pipe [%d->%d]\n", cddp->id, pip->write_fd, pip->read_fd));
-		    if (close(pip->read_fd) || close(pip->write_fd)) {
+		    TRACEF(("%T: Close pipe [%d->%d]\n", cddp->id, pip->write_fd,
+                            pip->read_fd));
+                    drv_use_singleton.fd_stop_select = -2; /* disable stop_select asserts */
+		    if (driver_select(cddp->port, (ErlDrvEvent)(ErlDrvSInt)pip->read_fd,
+                                      DO_READ|ERL_DRV_USE, 0)
+                        || close(pip->write_fd)) {
 			fprintf(stderr, "Failed to close pipe, errno=%d\n", errno);
 			abort();
 		    }
@@ -1284,8 +1025,10 @@ chkio_drv_control(ErlDrvData drv_data,
 		    break;
 		}
 		else {
-		    TRACEF(("%T: Select on pipe [%d->%d]\n", cddp->id, pip->write_fd, pip->read_fd));
-		    if (driver_select(cddp->port, (ErlDrvEvent)(ErlDrvSInt)pip->read_fd, DO_READ, 1)) {
+		    TRACEF(("%T: Select on pipe [%d->%d]\n", cddp->id,
+                            pip->write_fd, pip->read_fd));
+		    if (driver_select(cddp->port, (ErlDrvEvent)(ErlDrvSInt)pip->read_fd,
+                                      DO_READ|ERL_DRV_USE, 1)) {
 			fprintf(stderr, "driver_select failed for fd=%d\n", pip->read_fd);
 			abort();
 		    }
@@ -1293,13 +1036,13 @@ chkio_drv_control(ErlDrvData drv_data,
 		    pip->wasSelected = 1;
 		    op >>= 1;
 		    if (pip->next_write != pip->next_read) { /* pipe not empty */
-			if (op & 1) {  
+			if (op & 1) {
 			    pip->state = Waiting; /* Wait for reader */
 			    break;
 			}
 			op >>= 1;
 		    }
-		}  
+		}
 	    }/*fall through*/
 	    case Selected:
 		if (op & 1) {
@@ -1328,7 +1071,7 @@ chkio_drv_control(ErlDrvData drv_data,
 			fprintf(stderr, "Failed to write to pipe fd=%d, errno=%d\n", pip->write_fd, errno);
 			abort();
 		    }
-		    pip->next_write++;		    
+		    pip->next_write++;
 		}
 		break;
             case Waiting:
@@ -1586,7 +1329,12 @@ static void chkio_drv_stop_select(ErlDrvEvent e, void* null)
     if (!(drv_use_singleton.fd_stop_select < 0)) {
 	assert_print("fd_stop_select<0", __LINE__); abort(); 
     }
-    drv_use_singleton.fd_stop_select = (int)(long)e;
+    /* fd_stop_select counting is disabled if this is set to -2 */
+    if (drv_use_singleton.fd_stop_select == -2) {
+        TRACEF(("closing %d\n", (int)(long)e));
+        close((int)(long)e);
+    } else
+        drv_use_singleton.fd_stop_select = (int)(long)e;
     /* Can't call chkio_drv_use directly here. That could even be recursive.
      * Next timeout will detect it instead.
      */
diff --git a/erts/emulator/test/driver_SUITE_data/env_drv.c b/erts/emulator/test/driver_SUITE_data/env_drv.c
new file mode 100644
index 0000000000..0e910eeb84
--- /dev/null
+++ b/erts/emulator/test/driver_SUITE_data/env_drv.c
@@ -0,0 +1,108 @@
+/*
+ * %CopyrightBegin%
+ *
+ * Copyright Ericsson AB 2017. All Rights Reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ *
+ * %CopyrightEnd%
+ */
+
+/* Tests whether erl_drv_putenv/erl_drv_getenv work correctly and reflect
+ * changes to os:putenv/getenv. */
+
+#include <string.h>
+#include <stdio.h>
+
+#include "erl_driver.h"
+
+static ErlDrvSSizeT env_drv_ctl(ErlDrvData drv_data, unsigned int cmd,
+        char* buf, ErlDrvSizeT len, char** rbuf, ErlDrvSizeT rsize);
+
+static ErlDrvEntry env_drv_entry = { 
+    NULL /* init */,
+    NULL /* start */,
+    NULL /* stop */,
+    NULL /* output */,
+    NULL /* ready_input */,
+    NULL /* ready_output */,
+    "env_drv",
+    NULL /* finish */,
+    NULL /* handle */,
+    env_drv_ctl,
+    NULL /* timeout */,
+    NULL /* outputv*/,
+    NULL /* ready_async */,
+    NULL /* flush */,
+    NULL /* call*/,
+    NULL /* event */,
+    ERL_DRV_EXTENDED_MARKER,
+    ERL_DRV_EXTENDED_MAJOR_VERSION,
+    ERL_DRV_EXTENDED_MINOR_VERSION,
+    ERL_DRV_FLAG_USE_PORT_LOCKING,
+    NULL /* handle2 */,
+    NULL /* handle_monitor */
+};
+
+DRIVER_INIT(env_drv) {
+    return &env_drv_entry;
+}
+
+static int test_putenv(ErlDrvData drv_data, char *buf, ErlDrvSizeT len) {
+    char key[256], value[256];
+    int key_len, value_len;
+
+    key_len = buf[0];
+    value_len = buf[1];
+
+    sprintf(key, "%.*s", key_len, &buf[2]);
+    sprintf(value, "%.*s", value_len, &buf[2 + key_len]);
+
+    return erl_drv_putenv(key, value);
+}
+
+static int test_getenv(ErlDrvData drv_data, char *buf, ErlDrvSizeT len) {
+    char expected_value[256], stored_value[256], key[256];
+    int expected_value_len, key_len;
+    size_t stored_value_len;
+    int res;
+
+    key_len = buf[0];
+    sprintf(key, "%.*s", key_len, &buf[2]);
+
+    expected_value_len = buf[1];
+    sprintf(expected_value, "%.*s", expected_value_len, &buf[2 + key_len]);
+
+    stored_value_len = sizeof(stored_value);
+    res = erl_drv_getenv(key, stored_value, &stored_value_len);
+
+    if(res == 0) {
+        return strcmp(stored_value, expected_value) != 0;
+    } else if(res == 1) {
+        return 127;
+    }
+
+    return 255;
+}
+
+static ErlDrvSSizeT env_drv_ctl(ErlDrvData drv_data, unsigned int cmd,
+        char* buf, ErlDrvSizeT len, char** rbuf, ErlDrvSizeT rsize) {
+
+    if(cmd == 0) {
+        (**rbuf) = (char)test_putenv(drv_data, buf, len);
+    } else {
+        (**rbuf) = (char)test_getenv(drv_data, buf, len);
+    }
+
+    return 1;
+}
diff --git a/erts/emulator/test/driver_SUITE_data/ioq_exit_drv.c b/erts/emulator/test/driver_SUITE_data/ioq_exit_drv.c
index d87c2bec93..fa58e9d5ec 100644
--- a/erts/emulator/test/driver_SUITE_data/ioq_exit_drv.c
+++ b/erts/emulator/test/driver_SUITE_data/ioq_exit_drv.c
@@ -25,8 +25,7 @@
  *		- ready_input(),
  *		- ready_output(),
  *		- timeout(),
- *		- driver_async() -> read_async(), and
- *		- event()
+ *		- driver_async() -> read_async()
  */
 
 #ifndef UNIX
@@ -65,11 +64,9 @@ typedef enum {
     IOQ_EXIT_READY_OUTPUT = 2,
     IOQ_EXIT_TIMEOUT = 3,
     IOQ_EXIT_READY_ASYNC = 4,
-    IOQ_EXIT_EVENT = 5,
     IOQ_EXIT_READY_INPUT_ASYNC = 6,
     IOQ_EXIT_READY_OUTPUT_ASYNC = 7,
     IOQ_EXIT_TIMEOUT_ASYNC = 8,
-    IOQ_EXIT_EVENT_ASYNC = 9
 } IOQExitTest;
 
 typedef struct {
@@ -80,9 +77,6 @@ typedef struct {
     int outstanding_async_task;
     long async_task;
     ErlDrvPDL pdl;
-#ifdef HAVE_POLL_H
-    struct erl_drv_event_data event_data;
-#endif
 } IOQExitDrvData;
 
 #define EV2FD(EV) ((int) ((long) (EV)))
@@ -97,8 +91,6 @@ static ErlDrvSSizeT control(ErlDrvData, unsigned int,
 static void timeout(ErlDrvData drv_data);
 static void ready_async(ErlDrvData drv_data, ErlDrvThreadData thread_data);
 static void flush(ErlDrvData drv_data);
-static void event(ErlDrvData drv_data, ErlDrvEvent event,
-		  ErlDrvEventData event_data);
 static void async_invoke(void*);
 static void do_driver_async(IOQExitDrvData *);
 
@@ -118,7 +110,7 @@ static ErlDrvEntry ioq_exit_drv_entry = {
     ready_async,
     flush,
     NULL /* call */,
-    event,
+    NULL /* unused_event_callback*/,
     ERL_DRV_EXTENDED_MARKER,
     ERL_DRV_EXTENDED_MAJOR_VERSION,
     ERL_DRV_EXTENDED_MINOR_VERSION,
@@ -149,10 +141,6 @@ start(ErlDrvPort port, char *command)
     ddp->outstanding_async_task = 0;
     ddp->async_task = -1;
     ddp->pdl = driver_pdl_create(port);
-#ifdef HAVE_POLL_H
-    ddp->event_data.events = (short) 0;
-    ddp->event_data.revents = (short) 0;
-#endif
 
     return (ErlDrvData) ddp;
 }
@@ -192,27 +180,6 @@ static ErlDrvSSizeT control(ErlDrvData drv_data,
 #else
 	goto done;
 #endif
-    case IOQ_EXIT_EVENT:
-    case IOQ_EXIT_EVENT_ASYNC:
-#ifdef UNIX
-#ifdef HAVE_POLL_H
-	ddp->ofd = open("/dev/null", O_WRONLY);
-	if (ddp->ofd < 0) {
-	    driver_failure_posix(ddp->port, errno);
-	    return 0;
-	}
-	else if (driver_event(ddp->port, FD2EV(ddp->ofd), NULL) != 0) {
-	    res_str = "skip: driver_event() not supported";
-	    goto done;
-	}
-#else
-	res_str = "skip: No poll.h found which is needed for this test";
-	goto done;
-#endif
-	break;
-#else /* UNIX */
-	goto done;
-#endif
     case IOQ_EXIT_TIMEOUT:
     case IOQ_EXIT_TIMEOUT_ASYNC:
 	break;
@@ -266,13 +233,6 @@ static void stop(ErlDrvData drv_data)
 		close(ddp->ofd);
 	    }
 	    break;
-	case IOQ_EXIT_EVENT:
-	case IOQ_EXIT_EVENT_ASYNC:
-	    if (ddp->ofd >= 0) {
-		driver_event(ddp->port, FD2EV(ddp->ofd), NULL);
-		close(ddp->ofd);
-	    }
-	    break;
 #endif
 	case IOQ_EXIT_TIMEOUT:
 	case IOQ_EXIT_TIMEOUT_ASYNC:
@@ -302,13 +262,6 @@ static void flush(ErlDrvData drv_data)
     case IOQ_EXIT_READY_OUTPUT_ASYNC:
 	driver_select(ddp->port, FD2EV(ddp->ofd), DO_WRITE, 1);
 	break;
-    case IOQ_EXIT_EVENT:
-    case IOQ_EXIT_EVENT_ASYNC:
-#ifdef HAVE_POLL_H
-	ddp->event_data.events |= POLLOUT;
-	driver_event(ddp->port, FD2EV(ddp->ofd), &ddp->event_data);
-#endif
-	break;
 #endif
     case IOQ_EXIT_TIMEOUT:
     case IOQ_EXIT_TIMEOUT_ASYNC:
@@ -395,30 +348,6 @@ static void ready_async(ErlDrvData drv_data, ErlDrvThreadData thread_data)
     }
 }
 
-static void event(ErlDrvData drv_data,
-		  ErlDrvEvent event,
-		  ErlDrvEventData event_data)
-{
-    IOQExitDrvData *ddp = (IOQExitDrvData *) drv_data;
-
-    PRINTF(("event(%p, %d, %p) called\r\n", drv_data, EV2FD(event), event_data));
-
-#if defined(UNIX) && defined(HAVE_POLL_H)
-    if (ddp->ofd == EV2FD(event)) {
-	driver_event(ddp->port, FD2EV(ddp->ofd), NULL);
-	close(ddp->ofd);
-	ddp->ofd = -1;
-	if (ddp->test == IOQ_EXIT_EVENT_ASYNC)
-	    do_driver_async(ddp);
-	else {
-	    driver_pdl_lock(ddp->pdl);
-	    driver_deq(ddp->port, 1);
-	    driver_pdl_unlock(ddp->pdl);
-	}
-    }
-#endif
-}
-
 static void async_invoke(void *arg)
 {
     PRINTF(("async_invoke(%p) called\r\n", arg));
diff --git a/erts/emulator/test/driver_SUITE_data/missing_callback_drv.c b/erts/emulator/test/driver_SUITE_data/missing_callback_drv.c
index e7480d2e00..14838f0377 100644
--- a/erts/emulator/test/driver_SUITE_data/missing_callback_drv.c
+++ b/erts/emulator/test/driver_SUITE_data/missing_callback_drv.c
@@ -41,10 +41,6 @@
 typedef struct {
     int ofd;
     int ifd;
-    int efd;
-#ifdef HAVE_POLL_H
-    struct erl_drv_event_data edata;
-#endif
 } mcd_data_t;
 
 static ErlDrvData start(ErlDrvPort port, char *command);
@@ -90,7 +86,6 @@ start(ErlDrvPort port, char *command)
 
     mcd->ofd = -1;
     mcd->ifd = -1;
-    mcd->efd = -1;
 
 #ifdef UNIX
 
@@ -105,15 +100,6 @@ start(ErlDrvPort port, char *command)
 	goto error;
     if (driver_select(port, (ErlDrvEvent) (long) mcd->ifd, DO_READ, 1) != 0)
 	goto error;
-
-#ifdef HAVE_POLL_H
-    mcd->efd = open("/dev/null", O_WRONLY);
-    if (mcd->efd < 0)
-	goto error;
-    mcd->edata.events = POLLOUT;
-    mcd->edata.revents = 0;
-    driver_event(port, (ErlDrvEvent) (long) mcd->efd, &mcd->edata);
-#endif
 #endif
 
     driver_set_timer(port, 0);
@@ -135,10 +121,6 @@ stop(ErlDrvData data)
 	    close(mcd->ofd);
 	if (mcd->ifd >= 0)
 	    close(mcd->ifd);
-#ifdef HAVE_POLL_H
-	if (mcd->efd >= 0)
-	    close(mcd->efd);
-#endif
 #endif
 	driver_free(mcd);
     }
diff --git a/erts/emulator/test/dump_SUITE.erl b/erts/emulator/test/dump_SUITE.erl
new file mode 100644
index 0000000000..8d18d46d92
--- /dev/null
+++ b/erts/emulator/test/dump_SUITE.erl
@@ -0,0 +1,125 @@
+%%
+%% %CopyrightBegin%
+%%
+%% Copyright Ericsson AB 2005-2017. All Rights Reserved.
+%%
+%% Licensed under the Apache License, Version 2.0 (the "License");
+%% you may not use this file except in compliance with the License.
+%% You may obtain a copy of the License at
+%%
+%%     http://www.apache.org/licenses/LICENSE-2.0
+%%
+%% Unless required by applicable law or agreed to in writing, software
+%% distributed under the License is distributed on an "AS IS" BASIS,
+%% WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+%% See the License for the specific language governing permissions and
+%% limitations under the License.
+%%
+%% %CopyrightEnd%
+%%
+
+-module(dump_SUITE).
+
+-include_lib("common_test/include/ct.hrl").
+
+-export([all/0, suite/0, init_per_testcase/2, end_per_testcase/2]).
+
+-export([signal_abort/1]).
+
+-export([load/0]).
+
+-include_lib("kernel/include/file.hrl").
+
+suite() ->
+    [{ct_hooks,[ts_install_cth]},
+     {timetrap, {minutes, 2}}].
+
+all() ->
+    [signal_abort].
+
+init_per_testcase(signal_abort, Config) ->
+    SO = erlang:system_info(schedulers_online),
+    erts_debug:set_internal_state(available_internal_state, true),
+    Dump = erts_debug:get_internal_state(scheduler_dump),
+    erts_debug:set_internal_state(available_internal_state, false),
+    if SO < 3 ->
+            {skip, "not enough schedulers"};
+       not Dump ->
+            {skip, "the platform does not support scheduler dump"};
+       Dump ->
+            Config
+    end.
+
+end_per_testcase(_, Config) ->
+    Config.
+
+%%%
+%%% The test cases -------------------------------------------------------------
+%%%
+
+%% Test that a snapshot is taken of other schedulers using a signal
+%% when a crash dump is generated.
+signal_abort(Config) ->
+
+    Dump = filename:join(proplists:get_value(priv_dir, Config),"signal_abort.dump"),
+
+    {ok, Node} = start_node(Config),
+
+    _P1 = spawn(Node, ?MODULE, load, []),
+    _P2 = spawn(Node, ?MODULE, load, []),
+    _P3 = spawn(Node, ?MODULE, load, []),
+    _P4 = spawn(Node, ?MODULE, load, []),
+    _P5 = spawn(Node, ?MODULE, load, []),
+    _P6 = spawn(Node, ?MODULE, load, []),
+
+    timer:sleep(500),
+
+    true = rpc:call(Node, os, putenv, ["ERL_CRASH_DUMP",Dump]),
+    rpc:call(Node, erlang, halt, ["dump"]),
+
+    {ok, Bin} = get_dump_when_done(Dump),
+
+    ct:log("~s",[Bin]),
+
+    {match, Matches} = re:run(Bin,"Current Process: <",[global]),
+
+    ct:log("Found ~p",[Matches]),
+
+    true = length(Matches) > 1,
+
+    file:delete(Dump),
+
+    ok.
+
+get_dump_when_done(Dump) ->
+    case file:read_file_info(Dump) of
+        {ok, #file_info{ size = Sz }} ->
+            get_dump_when_done(Dump, Sz);
+        {error, enoent} ->
+            timer:sleep(1000),
+            get_dump_when_done(Dump)
+    end.
+
+get_dump_when_done(Dump, Sz) ->
+    timer:sleep(1000),
+    case file:read_file_info(Dump) of
+        {ok, #file_info{ size = Sz }} ->
+            file:read_file(Dump);
+        {ok, #file_info{ size = NewSz }} ->
+            get_dump_when_done(Dump, NewSz)
+    end.
+
+load() ->
+    lists:seq(1,10000),
+    load().
+
+start_node(Config) when is_list(Config) ->
+    Pa = filename:dirname(code:which(?MODULE)),
+    Name = list_to_atom(atom_to_list(?MODULE)
+                        ++ "-"
+                        ++ atom_to_list(proplists:get_value(testcase, Config))
+                        ++ "-"
+                        ++ integer_to_list(erlang:system_time(second))
+                        ++ "-"
+                        ++ integer_to_list(erlang:unique_integer([positive]))),
+    test_server:start_node(Name, slave, [{args, "-pa "++Pa}]).
diff --git a/erts/emulator/test/efile_SUITE.erl b/erts/emulator/test/efile_SUITE.erl
index f0e1bcf04b..821381bf0d 100644
--- a/erts/emulator/test/efile_SUITE.erl
+++ b/erts/emulator/test/efile_SUITE.erl
@@ -19,93 +19,17 @@
 
 -module(efile_SUITE).
 -export([all/0, suite/0]).
--export([iter_max_files/1, async_dist/1]).
+-export([iter_max_files/1, proc_zero_sized_files/1]).
 
--export([do_iter_max_files/2, do_async_dist/1]).
+-export([do_iter_max_files/2]).
 
 -include_lib("common_test/include/ct.hrl").
+-include_lib("stdlib/include/assert.hrl").
 
 suite() -> [{ct_hooks,[ts_install_cth]}].
 
 all() -> 
-    [iter_max_files, async_dist].
-
-do_async_dist(Dir) ->
-    X = 100,
-    AT = erlang:system_info(thread_pool_size),
-    Keys = file_keys(Dir,AT*X,[],[]),
-    Tab = ets:new(x,[ordered_set]),
-    [ ets:insert(Tab,{N,0}) || N <- lists:seq(0,AT-1) ],
-    [ ets:update_counter(Tab,(N rem AT),1) || N <- Keys ],
-    Res = [ V || {_,V} <- ets:tab2list(Tab) ],
-    ets:delete(Tab),
-    {Res, sdev(Res)/X}.
-
-sdev(List) ->
-    Len = length(List),
-    Mean = lists:sum(List)/Len,
-    math:sqrt(lists:sum([ (X - Mean) * (X - Mean) || X <- List ]) / Len).
-
-file_keys(_,0,FdList,FnList) ->
-    [ file:close(FD) || FD <- FdList ],
-    [ file:delete(FN) || FN <- FnList ],
-    [];
-file_keys(Dir,Num,FdList,FnList) ->
-    Name = "dummy"++integer_to_list(Num),
-    FN = filename:join([Dir,Name]),
-    case file:open(FN,[write,raw]) of
-        {ok,FD} ->
-            {file_descriptor,prim_file,{Port,_}} = FD,
-            <<X:32/integer-big>> = 
-            iolist_to_binary(erlang:port_control(Port,$K,[])),
-            [X | file_keys(Dir,Num-1,[FD|FdList],[FN|FnList])];
-        {error,_} ->
-            % Try freeing up FD's if there are any
-            case FdList of 
-                [] ->
-                    exit({cannot_open_file,FN});
-                _ ->
-                    [ file:close(FD) || FD <- FdList ],
-                    [ file:delete(F) || F <- FnList ],
-                    file_keys(Dir,Num,[],[])
-            end
-    end.
-
-%% Check that the distribution of files over async threads is fair
-async_dist(Config) when is_list(Config) ->
-    DataDir = proplists:get_value(data_dir,Config),
-    Dir = filename:dirname(code:which(?MODULE)),
-    AsyncSizes = [7,10,100,255,256,64,63,65],
-    Max = 0.5,
-
-    lists:foreach(fun(Size) ->
-                          {ok,Node} = 
-                          test_server:start_node
-                          (test_iter_max_files,slave,
-                           [{args,
-                             "+A "++integer_to_list(Size)++
-                             " -pa " ++ Dir}]),
-                          {Distr,SD} = rpc:call(Node,?MODULE,do_async_dist,
-                                                [DataDir]),
-                          test_server:stop_node(Node),
-                          if
-                              SD > Max ->
-                                  io:format("Bad async queue distribution for "
-                                            "~p async threads:~n"
-                                            "    Standard deviation is ~p~n"
-                                            "    Key distribution:~n    ~lp~n",
-                                            [Size,SD,Distr]),
-                                  exit({bad_async_dist,Size,SD,Distr});
-                              true ->
-                                  io:format("OK async queue distribution for "
-                                            "~p async threads:~n"
-                                            "    Standard deviation is ~p~n"
-                                            "    Key distribution:~n    ~lp~n",
-                                            [Size,SD,Distr]),
-                                  ok
-                          end
-                  end, AsyncSizes),
-    ok.
+    [iter_max_files, proc_zero_sized_files].
 
 %%
 %% Open as many files as possible. Do this several times and check 
@@ -113,17 +37,23 @@ async_dist(Config) when is_list(Config) ->
 %%
 
 iter_max_files(Config) when is_list(Config) ->
+    case os:type() of
+        {win32, _} -> {skip, "Windows lacks a hard limit on file handles"};
+        _ -> iter_max_files_1(Config)
+    end.
+
+iter_max_files_1(Config) ->
     DataDir = proplists:get_value(data_dir,Config),
     TestFile = filename:join(DataDir, "existing_file"),
     N = 10,
-    %% Run on a different node in order to set the max ports
+    %% Run on a different node in order to make the test more stable.
     Dir = filename:dirname(code:which(?MODULE)),
     {ok,Node} = test_server:start_node(test_iter_max_files,slave,
-                                       [{args,"+Q 1524 -pa " ++ Dir}]),
+                                       [{args,"-pa " ++ Dir}]),
     L = rpc:call(Node,?MODULE,do_iter_max_files,[N, TestFile]),
     test_server:stop_node(Node),
     io:format("Number of files opened in each test:~n~w\n", [L]),
-    all_equal(L),
+    verify_max_files(L),
     Head = hd(L),
     if  Head >= 2 -> ok;
         true -> ct:fail(too_few_files)
@@ -135,12 +65,15 @@ do_iter_max_files(N, Name) when N > 0 ->
 do_iter_max_files(_, _) ->
     [].
 
-all_equal([E, E| T]) ->
-    all_equal([E| T]);
-all_equal([_]) ->
-    ok;
-all_equal([]) ->
-    ok.
+%% The attempts shouldn't vary too much; we used to require that they were all
+%% exactly equal, but after we reimplemented the file driver as a NIF we
+%% noticed that the only reason it was stable on Darwin was because the port
+%% limit was hit before ulimit.
+verify_max_files(Attempts) ->
+    N = length(Attempts),
+    Mean = lists:sum(Attempts) / N,
+    Variance = lists:sum([(X - Mean) * (X - Mean) || X <- Attempts]) / N,
+    true = math:sqrt(Variance) =< 1 + (Mean / 1000).
 
 max_files(Name) ->
     Fds = open_files(Name),
@@ -162,3 +95,44 @@ open_files(Name) ->
             %		  io:format("Error reason: ~p", [_Reason]),
             []
     end.
+
+%% @doc If /proc filesystem exists (no way to know if it is real proc or just
+%% a /proc directory), let's read some zero sized files 500 times each, while
+%% ensuring that response isn't empty << >>
+proc_zero_sized_files(Config) when is_list(Config) ->
+    {Type, Flavor} = os:type(),
+    %% Some files which exist on Linux but might be missing on other systems
+    Inputs = ["/proc/cpuinfo",
+              "/proc/meminfo",
+              "/proc/partitions",
+              "/proc/swaps",
+              "/proc/version",
+              "/proc/uptime",
+              %% curproc is present on freebsd
+              "/proc/curproc/cmdline"],
+    case filelib:is_dir("/proc") of
+        false -> {skip, "/proc not found"}; % skip the test if no /proc
+        _ when Type =:= unix andalso Flavor =:= sunos ->
+            %% SunOS has a /proc, but no zero sized special files
+            {skip, "sunos does not have any zero sized special files"};
+        true ->
+            %% Take away files which do not exist in proc
+            Inputs1 = lists:filter(fun filelib:is_file/1, Inputs),
+
+            %% Fail if none of mentioned files exist in /proc, did we just get
+            %% a normal /proc directory without any special files?
+            ?assertNotEqual([], Inputs1),
+
+            %% For 6 inputs and 500 attempts each this do run anywhere
+            %% between 500 and 3000 function calls.
+            lists:foreach(
+                fun(Filename) -> do_proc_zero_sized(Filename, 500) end,
+                Inputs1)
+    end.
+
+%% @doc Test one file N times to also trigger possible leaking fds and memory
+do_proc_zero_sized(_Filename, 0) -> ok;
+do_proc_zero_sized(Filename, N) ->
+    Data = file:read_file(Filename),
+    ?assertNotEqual(<<>>, Data),
+    do_proc_zero_sized(Filename, N-1).
diff --git a/erts/emulator/test/emulator_smoke.spec b/erts/emulator/test/emulator_smoke.spec
index b2d0de8835..fc98ba6823 100644
--- a/erts/emulator/test/emulator_smoke.spec
+++ b/erts/emulator/test/emulator_smoke.spec
@@ -7,3 +7,4 @@
     [consistency],"Not reliable in October and March"}.
 {cases,'Dir',crypto_SUITE,[t_md5]}.
 {cases,'Dir',float_SUITE,[fpe,cmp_integer]}.
+{cases,'Dir',erts_debug_SUITE,[df]}.
diff --git a/erts/emulator/test/erl_link_SUITE.erl b/erts/emulator/test/erl_link_SUITE.erl
index 5622cce980..ed444f2599 100644
--- a/erts/emulator/test/erl_link_SUITE.erl
+++ b/erts/emulator/test/erl_link_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 2001-2017. All Rights Reserved.
+%% Copyright Ericsson AB 2001-2018. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -53,26 +53,17 @@
 -export([test_proc/0]).
 
 
--define(LINK_UNDEF, 0).
--define(LINK_PID,   1).
--define(LINK_NODE,  3).
-
-
-% These are to be kept in sync with erl_monitors.h 
--define(MON_ORIGIN, 1).
--define(MON_TARGET, 2).
-
-
--record(erl_link, {type = ?LINK_UNDEF,
+-record(erl_link, {type,           % process | port | dist_process
                    pid = [],
-                   targets = []}).
+                   id}).
 
 % This is to be kept in sync with erl_bif_info.c (make_monitor_list)
 
--record(erl_monitor, {type,        % MON_ORIGIN or MON_TARGET
-                      ref,
+-record(erl_monitor, {type,        % process | port | time_offset | dist_process | resource | node | nodes | suspend
+                      dir,         % origin | target
+                      ref,         % Reference | []
                       pid,         % Process or nodename
-                      name = []}). % registered name or []
+                      extra = []}). % registered name, integer or, []
 
 
 suite() ->
@@ -106,7 +97,7 @@ end_per_suite(_Config) ->
 links(Config) when is_list(Config) ->
     common_link_test(node(), node()),
     true = link(self()),
-    [] = find_erl_link(self(), ?LINK_PID, self()),
+    [] = find_erl_link(self(), process, self()),
     true = unlink(self()),
     ok.
 
@@ -191,6 +182,7 @@ monitor_nodes(Config) when is_list(Config) ->
     monitor_node(A, false),
     monitor_node(B, true),
     monitor_node(C, true),
+    receive after 1000 -> ok end,
     monitor_node(C, false),
     monitor_node(C, true),
     monitor_node(B, true),
@@ -200,13 +192,16 @@ monitor_nodes(Config) when is_list(Config) ->
     monitor_node(A, true),
     check_monitor_node(self(), A, 1),
     check_monitor_node(self(), B, 3),
+    ok = receive {nodedown, C} -> ok after 1000 -> timeout end,
+    %%OTP-21: monitor_node(_,false) does not trigger auto-connect anymore
+    %%        and therefore no nodedown if it fails.
+    %%ok = receive {nodedown, C} -> ok after 1000 -> timeout end,
+    ok = receive {nodedown, C} -> ok after 1000 -> timeout end,
+    ok = receive {nodedown, D} -> ok after 1000 -> timeout end,
+    ok = receive {nodedown, D} -> ok after 1000 -> timeout end,
     check_monitor_node(self(), C, 0),
     check_monitor_node(self(), D, 0),
-    receive {nodedown, C} -> ok end,
-    receive {nodedown, C} -> ok end,
-    receive {nodedown, C} -> ok end,
-    receive {nodedown, D} -> ok end,
-    receive {nodedown, D} -> ok end,
+
     stop_node(A),
     receive {nodedown, A} -> ok end,
     check_monitor_node(self(), A, 0),
@@ -301,7 +296,8 @@ run_common_process_monitors(TP1, TP2) ->
     wait_until(fun () -> is_proc_dead(TP2) end),
     ok = tp_call(TP1, fun () ->
                               receive
-                                  {'DOWN',R2,process,TP2O,bye} ->
+                                  {'DOWN',R2,process,TP2O,Reason1} ->
+                                      bye = Reason1,
                                       ok
                               end
                       end),
@@ -310,7 +306,8 @@ run_common_process_monitors(TP1, TP2) ->
     R3 = tp_call(TP1, fun () -> erlang:monitor(process, TP2) end),
     ok = tp_call(TP1, fun () ->
                               receive
-                                  {'DOWN',R3,process,TP2O,noproc} ->
+                                  {'DOWN',R3,process,TP2O,Reason2} ->
+                                      noproc = Reason2,
                                       ok
                               end
                       end),
@@ -533,7 +530,7 @@ freeze_node(Node, MS) ->
                fun () ->
                        erts_debug:set_internal_state(available_internal_state,
                                                      true),
-                       dport_send(Freezer, DoingIt),
+                       dctrl_dop_send(Freezer, DoingIt),
                        receive after Own -> ok end,
                        erts_debug:set_internal_state(block, MS+Own)
                end),
@@ -544,20 +541,22 @@ make_busy(Node, Time) when is_integer(Time) ->
     Own = 500,
     freeze_node(Node, Time+Own), 
     Data = busy_data(),
+    DCtrl = dctrl(Node),
     %% first make port busy
     Pid = spawn_link(fun () ->
                              forever(fun () ->
-                                             dport_reg_send(Node,
-                                                            '__noone__',
-                                                            Data)
+                                             dctrl_dop_reg_send(Node,
+                                                                '__noone__',
+                                                                Data)
                                      end)
                      end),
     receive after Own -> ok end,
     wait_until(fun () ->
-                       case process_info(Pid, status) of
-                           {status, suspended} -> true;
-                           _ -> false
-                       end
+                  case {DCtrl, process_info(Pid, status)} of
+                      {DPrt, {status, suspended}} when is_port(DPrt) -> true;
+                      {DPid, {status, waiting}} when is_pid(DPid) -> true;
+                      _ -> false
+                  end
                end),
     %% then dist entry
     make_busy(Node, [nosuspend], Data),
@@ -693,22 +692,10 @@ test_proc() ->
     end,
     test_proc().
 
-expand_link_list([#erl_link{type = ?LINK_NODE, targets = N} = Rec | T]) ->
-    lists:duplicate(N,Rec#erl_link{targets = []}) ++ expand_link_list(T);
-expand_link_list([#erl_link{targets = [#erl_link{pid = Pid}]} = Rec | T]) ->
-    [Rec#erl_link{targets = [Pid]} | expand_link_list(T)];
-expand_link_list([#erl_link{targets = [#erl_link{pid = Pid}|TT]} = Rec | T]) ->
-    [ Rec#erl_link{targets = [Pid]} | expand_link_list( 
-                                        [Rec#erl_link{targets = TT} | T])]; 
-expand_link_list([#erl_link{targets = []} = Rec | T]) ->
-    [Rec | expand_link_list(T)];
-expand_link_list([]) ->
-    [].
-
 get_local_link_list(Obj) ->
     case catch erts_debug:get_internal_state({link_list, Obj}) of
         LL when is_list(LL) ->
-            expand_link_list(LL);
+            LL;
         _ ->
             []
     end.
@@ -717,7 +704,7 @@ get_remote_link_list(Node, Obj) ->
     case catch rpc:call(Node, erts_debug, get_internal_state,
                         [{link_list, Obj}]) of
         LL when is_list(LL) ->
-            expand_link_list(LL);
+            LL;
         _ ->
             []
     end.
@@ -771,82 +758,106 @@ get_monitor_list(undefined) ->
 
 
 find_erl_monitor(Pid, Ref) when is_reference(Ref) ->
+    MonitorList = get_monitor_list(Pid),
+    io:format("~p MonitorList: ~p~n", [Pid, MonitorList]),
     lists:foldl(fun (#erl_monitor{ref = R} = EL, Acc) when R == Ref ->
                         [EL|Acc];
                     (_, Acc) ->
                         Acc
                 end,
                 [],
-                get_monitor_list(Pid)).
-
-% find_erl_link(Obj, Ref) when is_reference(Ref) -> 
-%     lists:foldl(fun (#erl_link{ref = R} = EL, Acc) when R == Ref ->
-% 			      [EL|Acc];
-% 			  (_, Acc) ->
-% 			      Acc
-% 		      end,
-% 		      [],
-% 		      get_link_list(Obj)).
-
-find_erl_link(Obj, Type, [Item, Data]) when is_pid(Item);
-                                            is_port(Item);
-                                            is_atom(Item) -> 
-    lists:foldl(fun (#erl_link{type = T, pid = I, targets = D} = EL,
+                MonitorList);
+find_erl_monitor(Pid, Item) ->
+    MonitorList = get_monitor_list(Pid),
+    io:format("~p MonitorList: ~p~n", [Pid, MonitorList]),
+    lists:foldl(fun (#erl_monitor{pid = I} = EL, Acc) when I == Item ->
+                        [EL|Acc];
+                    (_, Acc) ->
+                        Acc
+                end,
+                [],
+                MonitorList).
+
+
+find_erl_link(Obj, Type, Item) when is_pid(Item); is_port(Item) -> 
+    LinkList = get_link_list(Obj),
+    io:format("~p LinkList: ~p~n", [Obj, LinkList]),
+    lists:foldl(fun (#erl_link{type = T, pid = I} = EL,
                      Acc) when T == Type, I == Item ->
-                        case Data of
-                            D ->
-                                [EL|Acc];
-                            [] ->
-                                [EL|Acc];
-                            _ ->
-                                Acc
-                        end;
+                        [EL|Acc];
                     (_, Acc) ->
                         Acc
                 end,
                 [],
-                get_link_list(Obj));
-find_erl_link(Obj, Type, Item) when is_pid(Item); is_port(Item); is_atom(Item) ->
-    find_erl_link(Obj, Type, [Item, []]). 
+                LinkList);
+find_erl_link(Obj, Type, Id) when is_integer(Id) -> 
+    %% Find by Id
+    LinkList = get_link_list(Obj),
+    io:format("~p LinkList: ~p~n", [Obj, LinkList]),
+    lists:foldl(fun (#erl_link{type = T, id = I} = EL,
+                     Acc) when T == Type, I == Id ->
+                        [EL|Acc];
+                    (_, Acc) ->
+                        Acc
+                end,
+                [],
+                LinkList).
 
 
+get_link_type(A, B) when is_port(A);
+                         is_port(B) ->
+    port;
+get_link_type(A, B) when is_pid(A),
+                         is_pid(B) ->
+    case node(A) == node(B) of
+        true ->
+            process;
+        false ->
+            dist_process
+    end.
 
+check_link(A, B) when node(A) == node(B) ->
+    LinkType = get_link_type(A, B),
+    [#erl_link{type = LinkType,
+               pid = B,
+               id = Id}] = find_erl_link(A, LinkType, B),
+    [#erl_link{type = LinkType,
+               pid = A,
+               id = Id}] = find_erl_link(B, LinkType, A),
+    [] = find_erl_link({node(A), node(B)},
+                       LinkType,
+                       A),
+    [] = find_erl_link({node(B), node(A)},
+                       LinkType,
+                       B),
+    ok;
 check_link(A, B) ->
-    [#erl_link{type = ?LINK_PID,
+    [#erl_link{type = dist_process,
                pid = B,
-               targets = []}] = find_erl_link(A, ?LINK_PID, B),
-    [#erl_link{type = ?LINK_PID,
+               id = IdA}] = find_erl_link(A, dist_process, B),
+    [#erl_link{type = dist_process,
                pid = A,
-               targets = []}] = find_erl_link(B, ?LINK_PID, A),
-    case node(A) == node(B) of
-        false ->
-            [#erl_link{type = ?LINK_PID,
-                       pid = A,
-                       targets = [B]}] = find_erl_link({node(A),
-                                                        node(B)},
-                                                       ?LINK_PID,
-                                                       [A, [B]]),
-            [#erl_link{type = ?LINK_PID,
-                       pid = B,
-                       targets = [A]}] = find_erl_link({node(B),
-                                                        node(A)},
-                                                       ?LINK_PID,
-                                                       [B, [A]]);
-        true ->
-            [] = find_erl_link({node(A), node(B)},
-                               ?LINK_PID,
-                               [A, [B]]),
-            [] = find_erl_link({node(B), node(A)},
-                               ?LINK_PID,
-                               [B, [A]])
-    end,
+               id = IdA}] = find_erl_link({node(A),
+                                           node(B)},
+                                          dist_process,
+                                          IdA),
+    [#erl_link{type = dist_process,
+               pid = A,
+               id = IdB}] = find_erl_link(B, dist_process, A),
+    [#erl_link{type = dist_process,
+               pid = B,
+               id = IdB}] = find_erl_link({node(B),
+                                           node(A)},
+                                          dist_process,
+                                          IdB),
     ok.
 
 check_unlink(A, B) ->
-    [] = find_erl_link(A, ?LINK_PID, B),
-    [] = find_erl_link(B, ?LINK_PID, A),
-    [] = find_erl_link({node(A), node(B)}, ?LINK_PID, [A, [B]]),
-    [] = find_erl_link({node(B), node(A)}, ?LINK_PID, [B, [A]]),
+    LinkType = get_link_type(A, B),
+    [] = find_erl_link(A, LinkType, B),
+    [] = find_erl_link(B, LinkType, A),
+    [] = find_erl_link({node(A), node(B)}, dist_process, A),
+    [] = find_erl_link({node(B), node(A)}, dist_process, B),
     ok.
 
 check_process_monitor(From, {Name, Node}, Ref) when is_pid(From),
@@ -859,22 +870,26 @@ check_process_monitor(From, {Name, Node}, Ref) when is_pid(From),
                                                     is_atom(Node),
                                                     is_reference(Ref) ->
     MonitoredPid = rpc:call(Node, erlang, whereis, [Name]),
-    [#erl_monitor{type = ?MON_ORIGIN,
+    [#erl_monitor{type = dist_process,
+                  dir = origin,
                   ref = Ref,
                   pid = Node,
-                  name = Name}] = find_erl_monitor(From, Ref),
-    [#erl_monitor{type = ?MON_TARGET,
+                  extra = Name}] = find_erl_monitor(From, Ref),
+    [#erl_monitor{type = dist_process,
+                  dir = target,
                   ref = Ref,
                   pid = From,
-                  name = Name}] = 	find_erl_monitor({node(From), Node}, Ref),
-    [#erl_monitor{type = ?MON_ORIGIN,
+                  extra = Name}] = find_erl_monitor({node(From), Node}, Ref),
+    [#erl_monitor{type = dist_process,
+                  dir = origin,
                   ref = Ref,
                   pid = MonitoredPid,
-                  name = Name}] = find_erl_monitor({Node, node(From)}, Ref),
-    [#erl_monitor{type = ?MON_TARGET,
+                  extra = Name}] = find_erl_monitor({Node, node(From)}, Ref),
+    [#erl_monitor{type = dist_process,
+                  dir = target,
                   ref = Ref,
                   pid = From,
-                  name = Name}] = find_erl_monitor(MonitoredPid, Ref),
+                  extra = Name}] = find_erl_monitor(MonitoredPid, Ref),
     ok;
 check_process_monitor(From, Name, Ref) when is_pid(From),
                                             is_atom(Name),
@@ -882,27 +897,36 @@ check_process_monitor(From, Name, Ref) when is_pid(From),
                                             is_reference(Ref) ->
     MonitoredPid = rpc:call(node(From), erlang, whereis, [Name]),
 
-    [#erl_monitor{type = ?MON_ORIGIN,
+    [#erl_monitor{type = process,
+                  dir = origin,
                   ref = Ref,
                   pid = MonitoredPid,
-                  name = Name}] = find_erl_monitor(From, Ref),
+                  extra = Name}] = find_erl_monitor(From, Ref),
 
 
-    [#erl_monitor{type = ?MON_TARGET,
+    [#erl_monitor{type = process,
+                  dir = target,
                   ref = Ref,
                   pid = From,
-                  name = Name}] = find_erl_monitor(MonitoredPid,Ref),
+                  extra = Name}] = find_erl_monitor(MonitoredPid,Ref),
     ok;
 check_process_monitor(From, To, Ref) when is_pid(From),
                                           is_pid(To),
                                           is_reference(Ref) ->
-    OriMon = [#erl_monitor{type = ?MON_ORIGIN,
+    MonType = case node(From) == node(To) of
+                  true -> process;
+                  false -> dist_process
+              end,
+
+    OriMon = [#erl_monitor{type = MonType,
+                           dir = origin,
                            ref = Ref,
                            pid = To}],
 
     OriMon = find_erl_monitor(From, Ref), 
 
-    TargMon = [#erl_monitor{type = ?MON_TARGET,
+    TargMon = [#erl_monitor{type = MonType,
+                            dir = target,
                             ref = Ref,
                             pid = From}],
     TargMon = find_erl_monitor(To, Ref),
@@ -910,7 +934,11 @@ check_process_monitor(From, To, Ref) when is_pid(From),
 
     case node(From) == node(To) of
         false ->
-            TargMon = find_erl_monitor({node(From), node(To)}, Ref),
+            DistTargMon = [#erl_monitor{type = dist_process,
+                                        dir = target,
+                                        ref = Ref,
+                                        pid = From}],
+            DistTargMon = find_erl_monitor({node(From), node(To)}, Ref),
             OriMon = find_erl_monitor({node(To), node(From)}, Ref);
         true ->
             [] = find_erl_monitor({node(From), node(From)}, Ref)
@@ -981,19 +1009,36 @@ check_process_demonitor(From, To, Ref) when is_pid(From),
     ok.
 
 no_of_monitor_node(From, Node) when is_pid(From), is_atom(Node) ->
-    length(find_erl_link(From, ?LINK_NODE, Node)).
+    case find_erl_monitor(From, Node) of
+        [] -> 0;
+        [#erl_monitor{type = node,
+                      dir = origin,
+                      pid = Node,
+                      extra = N}] -> N
+    end.
 
+check_monitor_node(From, Node, 0) when is_pid(From),
+                                        is_atom(Node) ->
+    [] = find_erl_monitor(From, Node),
+    [] = find_erl_monitor({node(From), Node}, From);
 check_monitor_node(From, Node, No) when is_pid(From),
                                         is_atom(Node),
                                         is_integer(No),
-                                        No >= 0 ->
-    LL = lists:duplicate(No, #erl_link{type = ?LINK_NODE, pid = Node}),
-    DLL = lists:duplicate(No, #erl_link{type = ?LINK_NODE, pid = From}),
-    LL = find_erl_link(From, ?LINK_NODE, Node),
-    DLL = find_erl_link({node(From), Node}, ?LINK_NODE, From),
-    ok.
-
-
+                                        No > 0 ->
+    [#erl_monitor{type = node,
+                  dir = origin,
+                  pid = Node,
+                  extra = No}] = find_erl_monitor(From, Node),
+    [#erl_monitor{type = node,
+                  dir = target,
+                  pid = From}] = find_erl_monitor({node(From), Node}, From).
+
+connection_id(Node) ->
+    try
+        erts_debug:get_internal_state({connection_id, Node})
+    catch
+        _:_ -> -1
+    end.
 
 hostname() ->
     from($@, atom_to_list(node())).
@@ -1048,42 +1093,45 @@ stop_node(Node) ->
 -define(DOP_DEMONITOR_P,	20).
 -define(DOP_MONITOR_P_EXIT,	21).
 
-dport_send(To, Msg) ->
-    Node = node(To),
-    DPrt = case dport(Node) of
-               undefined ->
-                   pong = net_adm:ping(Node),
-                   dport(Node);
-               Prt ->
-                   Prt
-           end,
-    port_command(DPrt, [dmsg_hdr(),
-                        dmsg_ext({?DOP_SEND,
-                                  ?COOKIE,
-                                  To}),
-                        dmsg_ext(Msg)]).
-
-dport_reg_send(Node, Name, Msg) ->
-    DPrt = case dport(Node) of
-               undefined ->
-                   pong = net_adm:ping(Node),
-                   dport(Node);
-               Prt ->
-                   Prt
-           end,
-    port_command(DPrt, [dmsg_hdr(),
-                        dmsg_ext({?DOP_REG_SEND,
-                                  self(),
-                                  ?COOKIE,
-                                  Name}),
-                        dmsg_ext(Msg)]).
-
-dport(Node) when is_atom(Node) ->
+ensure_dctrl(Node) ->
+    case dctrl(Node) of
+        undefined ->
+            pong = net_adm:ping(Node),
+            dctrl(Node);
+        DCtrl ->
+            DCtrl
+    end.
+
+dctrl_send(DPrt, Data) when is_port(DPrt) ->
+    port_command(DPrt, Data);
+dctrl_send(DPid, Data) when is_pid(DPid) ->
+    Ref = make_ref(),
+    DPid ! {send, self(), Ref, Data},
+    receive {Ref, Res} -> Res end.
+
+dctrl_dop_send(To, Msg) ->
+    dctrl_send(ensure_dctrl(node(To)),
+               [dmsg_hdr(),
+                dmsg_ext({?DOP_SEND,
+                          ?COOKIE,
+                          To}),
+                dmsg_ext(Msg)]).
+
+dctrl_dop_reg_send(Node, Name, Msg) ->
+    dctrl_send(ensure_dctrl(Node),
+               [dmsg_hdr(),
+                dmsg_ext({?DOP_REG_SEND,
+                          self(),
+                          ?COOKIE,
+                          Name}),
+                dmsg_ext(Msg)]).
+
+dctrl(Node) when is_atom(Node) ->
     case catch erts_debug:get_internal_state(available_internal_state) of
         true -> true;
         _ -> erts_debug:set_internal_state(available_internal_state, true)
     end,
-    erts_debug:get_internal_state({dist_port, Node}).
+    erts_debug:get_internal_state({dist_ctrl, Node}).
 
 dmsg_hdr() ->
     [131, % Version Magic
diff --git a/erts/emulator/test/estone_SUITE.erl b/erts/emulator/test/estone_SUITE.erl
index 8b336b366d..c9c1867049 100644
--- a/erts/emulator/test/estone_SUITE.erl
+++ b/erts/emulator/test/estone_SUITE.erl
@@ -20,7 +20,7 @@
 -module(estone_SUITE).
 %% Test functions
 -export([all/0, suite/0, groups/0,
-	 estone/1, estone_bench/1]).
+	 estone/1, estone_bench/1, pgo/0]).
 
 %% Internal exports for EStone tests
 -export([lists/1,
@@ -44,9 +44,9 @@
 	 links/1,lproc/1,
 	 run_micro/3,p1/1,ppp/3,macro/2,micros/0]).
 
-
--include_lib("common_test/include/ct.hrl").
+-ifndef(PGO).
 -include_lib("common_test/include/ct_event.hrl").
+-endif.
 
 %% EStone defines
 -define(TOTAL, (3000 * 1000 * 100)).   %% 300 secs
@@ -85,13 +85,28 @@ estone(Config) when is_list(Config) ->
 estone_bench(Config) ->
     DataDir = proplists:get_value(data_dir,Config),
     L = ?MODULE:macro(?MODULE:micros(),DataDir),
-    [ct_event:notify(
-       #event{name = benchmark_data, 
-	      data = [{name,proplists:get_value(title,Mark)},
-		      {value,proplists:get_value(estones,Mark)}]})
-     || Mark <- L],
+    {Total, Stones} = sum_micros(L, 0, 0),
+    notify([[{title,"ESTONES"}, {estones, Stones}] | L]),
     L.
 
+-ifndef(PGO).
+notify(Marks) ->
+    [ct_event:notify(
+       #event{name = benchmark_data,
+	      data = [{name,proplists:get_value(title, Mark)},
+		      {value,proplists:get_value(estones, Mark)}]})
+     || Mark <- Marks].
+-else.
+notify(_) ->
+    ok.
+-endif.
+
+%% The benchmarks to run in order to guide PGO (profile guided optimisation)
+pgo() ->
+    %% We run all benchmarks except the port_io as we don't want to
+    %% have to build a custom port.
+    Micros = ?MODULE:micros() -- [micro(port_io)],
+    ?MODULE:macro(Micros,[]).
 
 %%
 %% Calculate CPU speed
@@ -364,7 +379,7 @@ monotonic_time() ->
     try erlang:monotonic_time() catch error:undef -> erlang:now() end.
 
 subtr(Before, After) when is_integer(Before), is_integer(After) ->
-    erlang:convert_time_unit(After-Before, native, microsecond);
+    erlang:convert_time_unit(After-Before, native, 1000000);
 subtr({_,_,_}=Before, {_,_,_}=After) ->
     timer:now_diff(After, Before).
 
diff --git a/erts/emulator/test/exception_SUITE.erl b/erts/emulator/test/exception_SUITE.erl
index aaca522da6..60d14ce841 100644
--- a/erts/emulator/test/exception_SUITE.erl
+++ b/erts/emulator/test/exception_SUITE.erl
@@ -21,9 +21,10 @@
 -module(exception_SUITE).
 
 -export([all/0, suite/0,
-         badmatch/1, pending_errors/1, nil_arith/1,
+         badmatch/1, pending_errors/1, nil_arith/1, top_of_stacktrace/1,
          stacktrace/1, nested_stacktrace/1, raise/1, gunilla/1, per/1,
-         exception_with_heap_frag/1, line_numbers/1]).
+         exception_with_heap_frag/1, backtrace_depth/1,
+         line_numbers/1]).
 
 -export([bad_guy/2]).
 -export([crash/1]).
@@ -31,14 +32,18 @@
 -include_lib("common_test/include/ct.hrl").
 -import(lists, [foreach/2]).
 
+%% The range analysis of the HiPE compiler results in a system limit error
+%% during compilation instead of at runtime, so do not perform this analysis.
+-compile([{hipe, [no_icode_range]}]).
+
 suite() ->
     [{ct_hooks,[ts_install_cth]},
      {timetrap, {minutes, 1}}].
 
 all() -> 
-    [badmatch, pending_errors, nil_arith, stacktrace,
-     nested_stacktrace, raise, gunilla, per,
-     exception_with_heap_frag, line_numbers].
+    [badmatch, pending_errors, nil_arith, top_of_stacktrace,
+     stacktrace, nested_stacktrace, raise, gunilla, per,
+     exception_with_heap_frag, backtrace_depth, line_numbers].
 
 -define(try_match(E),
         catch ?MODULE:bar(),
@@ -241,7 +246,54 @@ ba_bnot(A) ->
     io:format("bnot ~p", [A]),
     {'EXIT', {badarith, _}} = (catch bnot A).
 
+%% Test that BIFs are added to the top of the stacktrace.
+
+top_of_stacktrace(Conf) when is_list(Conf) ->
+    %% Arithmetic operators
+    {'EXIT', {badarith, [{erlang, '+', [1, ok], _} | _]}} = (catch my_add(1, ok)),
+    {'EXIT', {badarith, [{erlang, '-', [1, ok], _} | _]}} = (catch my_minus(1, ok)),
+    {'EXIT', {badarith, [{erlang, '*', [1, ok], _} | _]}} = (catch my_times(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'div', [1, ok], _} | _]}} = (catch my_div(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'div', [1, 0], _} | _]}} = (catch my_div(1, 0)),
+    {'EXIT', {badarith, [{erlang, 'rem', [1, ok], _} | _]}} = (catch my_rem(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'rem', [1, 0], _} | _]}} = (catch my_rem(1, 0)),
+
+    %% Bit operators
+    {'EXIT', {badarith, [{erlang, 'band', [1, ok], _} | _]}} = (catch my_band(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'bor', [1, ok], _} | _]}} = (catch my_bor(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'bsl', [1, ok], _} | _]}} = (catch my_bsl(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'bsr', [1, ok], _} | _]}} = (catch my_bsr(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'bxor', [1, ok], _} | _]}} = (catch my_bxor(1, ok)),
+    {'EXIT', {badarith, [{erlang, 'bnot', [ok], _} | _]}} = (catch my_bnot(ok)),
+
+    %% Tuples
+    {'EXIT', {badarg, [{erlang, element, [1, ok], _} | _]}} = (catch my_element(1, ok)),
+    {'EXIT', {badarg, [{erlang, element, [ok, {}], _} | _]}} = (catch my_element(ok, {})),
+    {'EXIT', {badarg, [{erlang, element, [1, {}], _} | _]}} = (catch my_element(1, {})),
+    {'EXIT', {badarg, [{erlang, element, [1, {}], _} | _]}} = (catch element(1, erlang:make_tuple(0, ok))),
+
+    %% System limits
+    Maxbig = maxbig(),
+    MinusMaxbig = -Maxbig,
+    {'EXIT', {system_limit, [{erlang, '+', [Maxbig, 1], _} | _]}} = (catch my_add(Maxbig, 1)),
+    {'EXIT', {system_limit, [{erlang, '+', [Maxbig, 1], _} | _]}} = (catch my_add(maxbig_gc(), 1)),
+    {'EXIT', {system_limit, [{erlang, '-', [MinusMaxbig, 1], _} | _]}} = (catch my_minus(-Maxbig, 1)),
+    {'EXIT', {system_limit, [{erlang, '-', [MinusMaxbig, 1], _} | _]}} = (catch my_minus(-maxbig_gc(), 1)),
+    {'EXIT', {system_limit, [{erlang, '*', [Maxbig, 2], _} | _]}} = (catch my_times(Maxbig, 2)),
+    {'EXIT', {system_limit, [{erlang, '*', [Maxbig, 2], _} | _]}} = (catch my_times(maxbig_gc(), 2)),
+    {'EXIT', {system_limit, [{erlang, 'bnot', [Maxbig], _} | _]}} = (catch my_bnot(Maxbig)),
+    {'EXIT', {system_limit, [{erlang, 'bnot', [Maxbig], _} | _]}} = (catch my_bnot(maxbig_gc())),
+    ok.
+
+maxbig() ->
+    %% We assume that the maximum arity is (1 bsl 19) - 1.
+    Ws = erlang:system_info(wordsize),
+    (((1 bsl ((16777184 * (Ws div 4))-1)) - 1) bsl 1) + 1.
 
+maxbig_gc() ->
+    Maxbig = maxbig(),
+    erlang:garbage_collect(),
+    Maxbig.
 
 stacktrace(Conf) when is_list(Conf) ->
     Tag = make_ref(),
@@ -253,9 +305,9 @@ stacktrace(Conf) when is_list(Conf) ->
     St1 = erase(stacktrace1),
     St1 = erase(stacktrace2),
     St1 = erlang:get_stacktrace(),
-    {caught2,{error,badarith},[{?MODULE,my_add,2,_}|_]=St2} =
+    {caught2,{error,badarith},[{erlang,'+',[0,a],_},{?MODULE,my_add,2,_}|_]=St2} =
     stacktrace_1({'div',{1,0}}, error, {'add',{0,a}}),
-    [{?MODULE,my_div,2,_}|_] = erase(stacktrace1),
+    [{erlang,'div',[1,0],_},{?MODULE,my_div,2,_}|_] = erase(stacktrace1),
     St2 = erase(stacktrace2),
     St2 = erlang:get_stacktrace(),
     {caught2,{error,{try_clause,V}},[{?MODULE,stacktrace_1,3,_}|_]=St3} =
@@ -308,13 +360,13 @@ nested_stacktrace(Conf) when is_list(Conf) ->
     nested_stacktrace_1({{value,{V,x1}},void,{V,x1}},
                         {void,void,void}),
     {caught1,
-     [{?MODULE,my_add,2,_}|_],
+     [{erlang,'+',[V,x1],_},{?MODULE,my_add,2,_}|_],
      value2,
-     [{?MODULE,my_add,2,_}|_]} =
+     [{erlang,'+',[V,x1],_},{?MODULE,my_add,2,_}|_]} =
     nested_stacktrace_1({{'add',{V,x1}},error,badarith},
                         {{value,{V,x2}},void,{V,x2}}),
     {caught1,
-     [{?MODULE,my_add,2,_}|_],
+     [{erlang,'+',[V,x1],_},{?MODULE,my_add,2,_}|_],
      {caught2,[{erlang,abs,[V],_}|_]},
      [{erlang,abs,[V],_}|_]} =
     nested_stacktrace_1({{'add',{V,x1}},error,badarith},
@@ -344,18 +396,18 @@ raise(Conf) when is_list(Conf) ->
     try 
         try foo({'div',{1,0}}) 
         catch
-            error:badarith ->
+            error:badarith:A0 ->
                 put(raise, A0 = erlang:get_stacktrace()),
                 erlang:raise(error, badarith, A0)
         end
     catch
-        error:badarith ->
+        error:badarith:A1 ->
             A1 = erlang:get_stacktrace(),
             A1 = get(raise)
     end,
     A = erlang:get_stacktrace(),
     A = get(raise),
-    [{?MODULE,my_div,2,_}|_] = A,
+    [{erlang,'div',[1, 0], _},{?MODULE,my_div,2,_}|_] = A,
     %%
     N = 8, % Must be even
     N = erlang:system_flag(backtrace_depth, N),
@@ -404,11 +456,20 @@ foo({raise,{Class,Reason,Stacktrace}}) ->
     erlang:raise(Class, Reason, Stacktrace).
 %%foo(function_clause) -> % must not be defined!
 
-my_div(A, B) ->
-    A div B.
+my_add(A, B) -> A + B.
+my_minus(A, B) -> A - B.
+my_times(A, B) -> A * B.
+my_div(A, B) -> A div B.
+my_rem(A, B) -> A rem B.
+
+my_band(A, B) -> A band B.
+my_bor(A, B) -> A bor B.
+my_bsl(A, B) -> A bsl B.
+my_bsr(A, B) -> A bsr B.
+my_bxor(A, B) -> A bxor B.
+my_bnot(A) -> bnot A.
 
-my_add(A, B) ->
-    A + B.
+my_element(A, B) -> element(A, B).
 
 my_abs(X) -> abs(X).
 
@@ -512,6 +573,57 @@ do_exception_with_heap_frag(Bin, [Sz|Sizes]) ->
     do_exception_with_heap_frag(Bin, Sizes);
 do_exception_with_heap_frag(_, []) -> ok.
 
+backtrace_depth(Config) when is_list(Config) ->
+    _ = [do_backtrace_depth(D) || D <- lists:seq(0, 8)],
+    ok.
+
+do_backtrace_depth(D) ->
+    Old = erlang:system_flag(backtrace_depth, D),
+    try
+        Expected = max(1, D),
+        do_backtrace_depth_1(Expected)
+    after
+        _ = erlang:system_flag(backtrace_depth, Old)
+    end.
+
+do_backtrace_depth_1(D) ->
+    Exit = fun() ->
+                   error(reason)
+           end,
+    HandCrafted = fun() ->
+                          {'EXIT',{_,Stk0}} = (catch error(get_stacktrace)),
+                          %% Fool the compiler to force a hand-crafted
+                          %% stacktrace.
+                          Stk = [hd(Stk0)|tl(Stk0)],
+                          erlang:raise(error, reason, Stk)
+                  end,
+    PassedOn = fun() ->
+                       try error(get_stacktrace)
+                       catch error:_:Stk ->
+                               %% Just pass on the given stacktrace.
+                               erlang:raise(error, reason, Stk)
+                       end
+               end,
+    do_backtrace_depth_2(D, Exit),
+    do_backtrace_depth_2(D, HandCrafted),
+    do_backtrace_depth_2(D, PassedOn),
+    ok.
+
+do_backtrace_depth_2(D, Exc) ->
+    try
+        Exc()
+    catch
+        error:reason:Stk ->
+            if
+                length(Stk) =/= D ->
+                    io:format("Expected depth: ~p\n", [D]),
+                    io:format("~p\n", [Stk]),
+                    error(bad_depth);
+                true ->
+                    ok
+            end
+    end.
+
 line_numbers(Config) when is_list(Config) ->
     {'EXIT',{{case_clause,bad_tag},
              [{?MODULE,line1,2,
@@ -606,6 +718,15 @@ line_numbers(Config) when is_list(Config) ->
               {?MODULE,line_numbers,1,_}|_]}} =
     (catch applied_bif_2()),
 
+    {'EXIT',{badarith,
+             [{?MODULE,increment1,1,[{file,"increment.erl"},{line,45}]},
+              {?MODULE,line_numbers,1,_}|_]}} =
+        (catch increment1(x)),
+    {'EXIT',{badarith,
+             [{?MODULE,increment2,1,[{file,"increment.erl"},{line,48}]},
+              {?MODULE,line_numbers,1,_}|_]}} =
+        (catch increment2(x)),
+
     ok.
 
 id(I) -> I.
@@ -706,3 +827,15 @@ applied_bif_2() ->				%Line 8
     R = process_info(self(), current_location),	%Line 9
     fail = R,					%Line 10
     ok.						%Line 11
+
+%% The increment instruction used to decrement the instruction
+%% pointer, which would cause the line number in a stack trace to
+%% be the previous line number.
+
+-file("increment.erl", 42).
+increment1(Arg) ->                              %Line 43
+    Res = id(Arg),                              %Line 44
+    Res + 1.                                    %Line 45
+increment2(Arg) ->                              %Line 46
+    _ = id(Arg),                                %Line 47
+    Arg + 1.                                    %Line 48
diff --git a/erts/emulator/test/guard_SUITE.erl b/erts/emulator/test/guard_SUITE.erl
index 1a93a9f5c2..f2c1595392 100644
--- a/erts/emulator/test/guard_SUITE.erl
+++ b/erts/emulator/test/guard_SUITE.erl
@@ -500,7 +500,7 @@ all_types() ->
      {atom, xxxx},
      {ref, make_ref()},
      {pid, self()},
-     {port, open_port({spawn, efile}, [])},
+     {port, make_port()},
      {function, fun(_) -> "" end},
      {function, fun erlang:abs/1},
      {binary, list_to_binary([])},
@@ -551,4 +551,7 @@ type_test(bitstring, X) when is_bitstring(X) ->
 type_test(function, X) when is_function(X) ->
     function.
 
+make_port() ->
+    hd(erlang:ports()).
+
 id(I) -> I.
diff --git a/erts/emulator/test/lcnt_SUITE.erl b/erts/emulator/test/lcnt_SUITE.erl
index 4e52c2813c..dfffd662e2 100644
--- a/erts/emulator/test/lcnt_SUITE.erl
+++ b/erts/emulator/test/lcnt_SUITE.erl
@@ -87,8 +87,9 @@ wait_for_empty_lock_list() ->
     wait_for_empty_lock_list(10).
 wait_for_empty_lock_list(Tries) when Tries > 0 ->
     try_flush_cleanup_ops(),
-    case erts_debug:lcnt_collect() of
-        [{duration, _}, {locks, []}] ->
+    [{duration, _}, {locks, Locks}] = erts_debug:lcnt_collect(),
+    case remove_untoggleable_locks(Locks) of
+        [] ->
             ok;
         _ ->
             timer:sleep(50),
@@ -124,7 +125,7 @@ toggle_lock_counting(Config) when is_list(Config) ->
 get_lock_info_for(Categories) when is_list(Categories) ->
     ok = erts_debug:lcnt_control(mask, Categories),
     [{duration, _}, {locks, Locks}] = erts_debug:lcnt_collect(),
-    Locks;
+    remove_untoggleable_locks(Locks);
 
 get_lock_info_for(Category) when is_atom(Category) ->
     get_lock_info_for([Category]).
@@ -178,3 +179,13 @@ registered_db_tables(Config) when is_list(Config) ->
             (_Lock) -> false
         end, DbLocks),
     ok.
+
+%% Not all locks can be toggled on or off due to technical limitations, so we
+%% need to filter them out when checking whether we successfully disabled lock
+%% counting.
+remove_untoggleable_locks([]) ->
+    [];
+remove_untoggleable_locks([{resource_monitors, _, _, _} | T]) ->
+    remove_untoggleable_locks(T);
+remove_untoggleable_locks([H | T]) ->
+    [H | remove_untoggleable_locks(T)].
diff --git a/erts/emulator/test/lttng_SUITE.erl b/erts/emulator/test/lttng_SUITE.erl
index a012fa1da2..19c3844c40 100644
--- a/erts/emulator/test/lttng_SUITE.erl
+++ b/erts/emulator/test/lttng_SUITE.erl
@@ -81,7 +81,6 @@ end_per_testcase(Case, _Config) ->
 
 %% Not tested yet
 %%   org_erlang_otp:driver_process_exit
-%%   org_erlang_otp:driver_event
 
 %% tracepoints
 %%
@@ -100,7 +99,6 @@ end_per_testcase(Case, _Config) ->
 %%   org_erlang_otp:driver_flush
 %%   org_erlang_otp:driver_stop_select
 %%   org_erlang_otp:driver_timeout
-%%   org_erlang_otp:driver_event
 %%   org_erlang_otp:driver_ready_output
 %%   org_erlang_otp:driver_ready_input
 %%   org_erlang_otp:driver_output
@@ -431,7 +429,6 @@ txt() ->
       "%%   org_erlang_otp:driver_flush\n"
       "%%   org_erlang_otp:driver_stop_select\n"
       "%%   org_erlang_otp:driver_timeout\n"
-      "%%   org_erlang_otp:driver_event\n"
       "%%   org_erlang_otp:driver_ready_output\n"
       "%%   org_erlang_otp:driver_ready_input\n"
       "%%   org_erlang_otp:driver_output\n"
diff --git a/erts/emulator/test/map_SUITE.erl b/erts/emulator/test/map_SUITE.erl
index 02f3c89318..f93c637650 100644
--- a/erts/emulator/test/map_SUITE.erl
+++ b/erts/emulator/test/map_SUITE.erl
@@ -36,7 +36,9 @@
          t_map_equal/1,
          t_map_compare/1,
          t_map_size/1,
+         t_map_get/1,
          t_is_map/1,
+         t_is_map_key/1,
 
          %% Specific Map BIFs
          t_bif_map_get/1,
@@ -52,7 +54,7 @@
          t_bif_map_values/1,
          t_bif_map_to_list/1,
          t_bif_map_from_list/1,
-         t_bif_erts_internal_maps_to_list/1,
+         t_bif_map_next/1,
 
          %% erlang
          t_erlang_hash/1,
@@ -119,12 +121,12 @@ all() -> [t_build_and_match_literals, t_build_and_match_literals_large,
           t_bif_map_update,
           t_bif_map_values,
           t_bif_map_to_list, t_bif_map_from_list,
-          t_bif_erts_internal_maps_to_list,
+          t_bif_map_next,
 
           %% erlang
           t_erlang_hash, t_map_encode_decode,
           t_gc_rare_map_overflow,
-          t_map_size, t_is_map,
+          t_map_size, t_map_get, t_is_map,
 
           %% non specific BIF related
           t_bif_build_and_check,
@@ -680,6 +682,88 @@ t_map_size(Config) when is_list(Config) ->
               end),
     ok.
 
+t_map_get(Config) when is_list(Config) ->
+    %% small map
+    1    = map_get(a, id(#{a=>1})),
+    2    = map_get(b, id(#{a=>1, b=>2})),
+    "hi" = map_get("hello", id(#{a=>1, "hello"=>"hi"})),
+    "tuple hi" = map_get({1,1.0}, id(#{a=>a, {1,1.0}=>"tuple hi"})),
+
+    M0    = id(#{ k1=>"v1", <<"k2">> => <<"v3">> }),
+    "v4" = map_get(<<"k2">>, M0#{<<"k2">> => "v4"}),
+
+    %% large map
+    M1   = maps:from_list([{I,I}||I<-lists:seq(1,100)] ++
+			  [{a,1},{b,2},{"hello","hi"},{{1,1.0},"tuple hi"},
+			   {k1,"v1"},{<<"k2">>,"v3"}]),
+    1    = map_get(a, M1),
+    2    = map_get(b, M1),
+    "hi" = map_get("hello", M1),
+    "tuple hi" = map_get({1,1.0}, M1),
+    "v3" = map_get(<<"k2">>, M1),
+
+    %% error cases
+    do_badmap(fun(T) ->
+		      {'EXIT',{{badmap,T},[{erlang,map_get,_,_}|_]}} =
+			  (catch map_get(a, T))
+	      end),
+
+    {'EXIT',{{badkey,{1,1}},[{erlang,map_get,_,_}|_]}} =
+	(catch map_get({1,1}, id(#{{1,1.0}=>"tuple"}))),
+    {'EXIT',{{badkey,a},[{erlang,map_get,_,_}|_]}} = (catch map_get(a, id(#{}))),
+    {'EXIT',{{badkey,a},[{erlang,map_get,_,_}|_]}} =
+	(catch map_get(a, id(#{b=>1, c=>2}))),
+
+    %% in guards
+    M2 = id(#{a=>1}),
+    true = if map_get(a, M2) =:= 1 -> true; true -> false end,
+    false = if map_get(x, M2) =:= 1 -> true; true -> false end,
+    do_badmap(fun
+        (T) when map_get(x, T) =:= 1 -> ok;
+        (T) -> false = is_map(T)
+    end),
+    ok.
+
+t_is_map_key(Config) when is_list(Config) ->
+    %% small map
+    true = is_map_key(a, id(#{a=>1})),
+    true = is_map_key(b, id(#{a=>1, b=>2})),
+    true = is_map_key("hello", id(#{a=>1, "hello"=>"hi"})),
+    true = is_map_key({1,1.0}, id(#{a=>a, {1,1.0}=>"tuple hi"})),
+
+    M0   = id(#{ k1=>"v1", <<"k2">> => <<"v3">> }),
+    true = is_map_key(<<"k2">>, M0#{<<"k2">> => "v4"}),
+
+    %% large map
+    M1   = maps:from_list([{I,I}||I<-lists:seq(1,100)] ++
+			  [{a,1},{b,2},{"hello","hi"},{{1,1.0},"tuple hi"},
+			   {k1,"v1"},{<<"k2">>,"v3"}]),
+    true = is_map_key(a, M1),
+    true = is_map_key(b, M1),
+    true = is_map_key("hello", M1),
+    true = is_map_key({1,1.0}, M1),
+    true = is_map_key(<<"k2">>, M1),
+
+    %% error cases
+    do_badmap(fun(T) ->
+		      {'EXIT',{{badmap,T},[{erlang,is_map_key,_,_}|_]}} =
+			  (catch is_map_key(a, T))
+	      end),
+
+    false = is_map_key({1,1}, id(#{{1,1.0}=>"tuple"})),
+    false = is_map_key(a, id(#{})),
+    false = is_map_key(a, id(#{b=>1, c=>2})),
+
+    %% in guards
+    M2 = id(#{a=>1}),
+    true = if is_map_key(a, M2) -> true; true -> false end,
+    false = if is_map_key(x, M2) -> true; true -> false end,
+    do_badmap(fun
+        (T) when is_map_key(T, x) =:= 1 -> ok;
+        (T) -> false = is_map(T)
+    end),
+    ok.
+
 build_and_check_size([K|Ks],N,M0) ->
     N = map_size(M0),
     M1 = M0#{ K => K },
@@ -2364,41 +2448,71 @@ t_bif_map_from_list(Config) when is_list(Config) ->
     {'EXIT', {badarg,_}} = (catch maps:from_list(id(42))),
     ok.
 
-t_bif_erts_internal_maps_to_list(Config) when is_list(Config) ->
-    %% small maps
-    [] = erts_internal:maps_to_list(#{},-1),
-    [] = erts_internal:maps_to_list(#{},-2),
-    [] = erts_internal:maps_to_list(#{},10),
-    [{a,1},{b,2}] = lists:sort(erts_internal:maps_to_list(#{a=>1,b=>2}, 2)),
-    [{a,1},{b,2}] = lists:sort(erts_internal:maps_to_list(#{a=>1,b=>2}, -1)),
-    [{_,_}] = erts_internal:maps_to_list(#{a=>1,b=>2}, 1),
-    [{a,1},{b,2},{c,3}] = lists:sort(erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},-2)),
-    [{a,1},{b,2},{c,3}] = lists:sort(erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},3)),
-    [{a,1},{b,2},{c,3}] = lists:sort(erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},5)),
-    [{_,_},{_,_}] = erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},2),
-    [{_,_}] = erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},1),
-    [] = erts_internal:maps_to_list(#{c=>3,a=>1,b=>2},0),
-
-    %% big maps
-    M = maps:from_list([{I,ok}||I <- lists:seq(1,500)]),
-    [] = erts_internal:maps_to_list(M,0),
-    [{_,_}] = erts_internal:maps_to_list(M,1),
-    [{_,_},{_,_}] = erts_internal:maps_to_list(M,2),
-    Ls1 = erts_internal:maps_to_list(M,10),
-    10 = length(Ls1),
-    Ls2 = erts_internal:maps_to_list(M,20),
-    20 = length(Ls2),
-    Ls3 = erts_internal:maps_to_list(M,120),
-    120 = length(Ls3),
-    Ls4 = erts_internal:maps_to_list(M,-1),
-    500 = length(Ls4),
+t_bif_map_next(Config) when is_list(Config) ->
 
-    %% error cases
-    {'EXIT', {{badmap,[{a,b},b]},_}} = (catch erts_internal:maps_to_list(id([{a,b},b]),id(1))),
-    {'EXIT', {badarg,_}} = (catch erts_internal:maps_to_list(id(#{}),id(a))),
-    {'EXIT', {badarg,_}} = (catch erts_internal:maps_to_list(id(#{1=>2}),id(<<>>))),
+    erts_debug:set_internal_state(available_internal_state, true),
+
+    try
+
+        none = maps:next(maps:iterator(id(#{}))),
+
+        verify_iterator(#{}),
+        verify_iterator(#{a => 1, b => 2, c => 3}),
+
+        %% Use fatmap in order to test iterating in very deep maps
+        FM = fatmap(43),
+        verify_iterator(FM),
+
+        {'EXIT', {{badmap,[{a,b},b]},_}} = (catch maps:iterator(id([{a,b},b]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id(a))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([a|FM]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([1|#{}]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([-1|#{}]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([-1|FM]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([16#FFFFFFFFFFFFFFFF|FM]))),
+        {'EXIT', {badarg,_}} = (catch maps:next(id([-16#FFFFFFFFFFFFFFFF|FM]))),
+
+        %% This us a whitebox test that the error code works correctly.
+        %% It uses a path for a tree of depth 4 and tries to do next on
+        %% each of those paths.
+        (fun F(0) -> ok;
+             F(N) ->
+                 try maps:next([N|FM]) of
+                     none ->
+                         F(N-1);
+                     {_K,_V,_I} ->
+                         F(N-1)
+                 catch error:badarg ->
+                         F(N-1)
+                 end
+         end)(16#FFFF),
+
+        ok
+    after
+            erts_debug:set_internal_state(available_internal_state, false)
+    end.
+
+verify_iterator(Map) ->
+    KVs = t_fold(fun(K, V, A) -> [{K, V} | A] end, [], Map),
+
+    %% Verify that KVs created by iterating Map is of
+    %% correct size and contains all elements
+    true = length(KVs) == maps:size(Map),
+    [maps:get(K, Map) || {K, _} <- KVs],
     ok.
 
+
+t_fold(Fun, Init, Map) ->
+    t_fold_1(Fun, Init, maps:iterator(Map)).
+
+t_fold_1(Fun, Acc, Iter) ->
+    case maps:next(Iter) of
+        {K, V, NextIter} ->
+            t_fold_1(Fun, Fun(K,V,Acc), NextIter);
+        none ->
+            Acc
+    end.
+
 t_bif_build_and_check(Config) when is_list(Config) ->
     ok = check_build_and_remove(750,[fun(K) -> [K,K] end,
 				     fun(K) -> [float(K),K] end,
diff --git a/erts/emulator/test/map_SUITE_data/badmap_17.beam b/erts/emulator/test/map_SUITE_data/badmap_17.beam
index 277fc34b94..6f79bb8c2c 100644
--- a/erts/emulator/test/map_SUITE_data/badmap_17.beam
+++ b/erts/emulator/test/map_SUITE_data/badmap_17.beam
diff --git a/erts/emulator/test/map_SUITE_data/badmap_17.erl b/erts/emulator/test/map_SUITE_data/badmap_17.erl
index 0ec65e0e33..887fc2e5e3 100644
--- a/erts/emulator/test/map_SUITE_data/badmap_17.erl
+++ b/erts/emulator/test/map_SUITE_data/badmap_17.erl
@@ -1,7 +1,7 @@
 -module(badmap_17).
 -export([update/1]).
 
-%% Compile this source file with OTP 17.
+%% Compile this source file with OTP 17.0.
 
 update(Map) ->
     try
@@ -17,10 +17,42 @@ update(Map) ->
     catch
 	error:{badmap,Map} ->
 	    ok
-    end.
+    end,
+    try
+	update_3(Map),
+	error(update_did_not_fail)
+    catch
+	error:{badmap,Map} ->
+	    ok
+    end,
+    ok = update_4(Map),
+    ok = update_5(Map),
+    ok.
 
 update_1(M) ->
     M#{a=>42}.
 
 update_2(M) ->
     M#{a:=42}.
+
+update_3(M) ->
+    id(M),
+    M#{a=>42}.
+
+update_4(M) when M#{a=>b} =:= M ->
+    did_not_fail;
+update_4(_) ->
+    ok.
+
+update_5(M) ->
+    id(M),
+    case id(true) of
+        true when M#{a=>b} =:= M ->
+            did_not_fail;
+        true ->
+            ok
+    end.
+
+id(I) ->
+    I.
+
diff --git a/erts/emulator/test/match_spec_SUITE.erl b/erts/emulator/test/match_spec_SUITE.erl
index 08a7b4560c..4415d8d1b9 100644
--- a/erts/emulator/test/match_spec_SUITE.erl
+++ b/erts/emulator/test/match_spec_SUITE.erl
@@ -885,6 +885,26 @@ maps(Config) when is_list(Config) ->
         erlang:match_spec_test(#{<<"b">> =>"camembert","c"=>"cabécou", "wat"=>"hi", b=><<"other">>},
                                [{#{<<"b">> => '$1',"wat" => '$2'},[],[#{a=>'$1',b=>'$2'}]}],
                                table),
+
+    {ok,1,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[],[{map_size,'$1'}]}],table),
+    {ok,'EXIT',[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[],[{map_size,'$1'}]}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[{map_size,'$1'}],['$_']}], table),
+    {ok,true,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[{'=:=',{map_size,'$1'},1}],[true]}], table),
+
+    {ok,1,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[],[{map_get,a,'$1'}]}], table),
+    {ok,'EXIT',[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[],[{map_get,b,'$1'}]}], table),
+    {ok,'EXIT',[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[],[{map_get,b,'$1'}]}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[{map_get,b,'$1'}],['$_']}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[{map_get,b,'$1'}],['$_']}], table),
+    {ok,true,[],[]} = erlang:match_spec_test(#{a => true}, [{'$1',[{map_get,a,'$1'}],[true]}], table),
+
+    {ok,true,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[],[{is_map_key,a,'$1'}]}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[],[{is_map_key,b,'$1'}]}], table),
+    {ok,'EXIT',[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[],[{is_map_key,a,'$1'}]}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(#{a => 1}, [{'$1',[{is_map_key,b,'$1'}],['$_']}], table),
+    {ok,false,[],[]} = erlang:match_spec_test(not_a_map, [{'$1',[{is_map_key,b,'$1'}],['$_']}], table),
+    {ok,true,[],[]} = erlang:match_spec_test(#{a => true}, [{'$1',[{is_map_key,a,'$1'}],[true]}], table),
+
     %% large maps
 
     Ls0 = [{I,<<I:32>>}||I <- lists:seq(1,415)],
diff --git a/erts/emulator/test/module_info_SUITE.erl b/erts/emulator/test/module_info_SUITE.erl
index ba9b564fdc..46a3bba732 100644
--- a/erts/emulator/test/module_info_SUITE.erl
+++ b/erts/emulator/test/module_info_SUITE.erl
@@ -23,7 +23,7 @@
 -include_lib("common_test/include/ct.hrl").
 
 -export([all/0, suite/0,
-	 exports/1,functions/1,deleted/1,native/1,info/1]).
+	 exports/1,functions/1,deleted/1,native/1,info/1,nifs/1]).
 
 %%-compile(native).
 
@@ -38,7 +38,7 @@ all() ->
     modules().
 
 modules() ->
-    [exports, functions, deleted, native, info].
+    [exports, functions, deleted, native, info, nifs].
 
 %% Should return all functions exported from this module. (local)
 all_exported() ->
@@ -62,12 +62,24 @@ exports(Config) when is_list(Config) ->
     All = lists:sort(?MODULE:module_info(exports)),
     ok.
 
-%% Test that the list of exported functions from this module is correct.
+%% Test that the list of local and exported functions from this module is
+%% correct.
 functions(Config) when is_list(Config) ->
     All = all_functions(),
     All = lists:sort(?MODULE:module_info(functions)),
     ok.
 
+nifs(Config) when is_list(Config) ->
+    [] = ?MODULE:module_info(nifs),
+
+    %% erl_tracer is guaranteed to be present and contain these NIFs
+    TraceNIFs = erl_tracer:module_info(nifs),
+    true = lists:member({enabled, 3}, TraceNIFs),
+    true = lists:member({trace, 5}, TraceNIFs),
+    2 = length(TraceNIFs),
+
+    ok.
+
 %% Test that deleted modules cause badarg
 deleted(Config) when is_list(Config) ->
     Data = proplists:get_value(data_dir, Config),
diff --git a/erts/emulator/test/monitor_SUITE.erl b/erts/emulator/test/monitor_SUITE.erl
index 9d772480d9..c7250a9d26 100644
--- a/erts/emulator/test/monitor_SUITE.erl
+++ b/erts/emulator/test/monitor_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 1999-2017. All Rights Reserved.
+%% Copyright Ericsson AB 1999-2018. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -86,7 +86,7 @@ case_2(Config) when is_list(Config) ->
     R = erlang:monitor(process, B),
     B ! R,
     receive
-        {'EXIT', _} -> ok;
+        true -> ok;
         Other ->
             ct:fail({rec, Other})
     end,
@@ -98,7 +98,7 @@ case_2a(Config) when is_list(Config) ->
     {B,R} = spawn_monitor(?MODULE, y2, [self()]),
     B ! R,
     receive
-        {'EXIT', _} -> ok;
+        true -> ok;
         Other ->
             ct:fail({rec, Other})
     end,
@@ -182,7 +182,7 @@ demon_e_1(Config) when is_list(Config) ->
            end ),
     receive 
         {P2, ref, R2} -> 
-            demon_error(R2, badarg),
+            true = erlang:demonitor(R2),
             P2 ! {self(), stop};
         Other2 ->
             ct:fail({rec, Other2})
@@ -729,8 +729,8 @@ named_down(Config) when is_list(Config) ->
                            end),
     ?assertEqual(true, register(Name, NamedProc)),
     unlink(NamedProc),
-    exit(NamedProc, bang),
     Mon = erlang:monitor(process, Name),
+    exit(NamedProc, bang),
     receive {'DOWN',Mon, _, _, bang} -> ok
     after 3000 -> ?assert(false) end,
     ?assertEqual(true, register(Name, self())),
diff --git a/erts/emulator/test/nif_SUITE.erl b/erts/emulator/test/nif_SUITE.erl
index ac2b136a38..100fa006e7 100644
--- a/erts/emulator/test/nif_SUITE.erl
+++ b/erts/emulator/test/nif_SUITE.erl
@@ -25,13 +25,14 @@
 %%-define(CHECK(Exp,Got), Exp = Got).
 
 -include_lib("common_test/include/ct.hrl").
+-include_lib("stdlib/include/assert.hrl").
 
 -export([all/0, suite/0, groups/0,
          init_per_group/2, end_per_group/2,
 	 init_per_testcase/2, end_per_testcase/2,
          basic/1, reload_error/1, upgrade/1, heap_frag/1,
          t_on_load/1,
-         select/1,
+         select/1, select_steal/1,
          monitor_process_a/1,
          monitor_process_b/1,
          monitor_process_c/1,
@@ -42,9 +43,9 @@
 	 types/1, many_args/1, binaries/1, get_string/1, get_atom/1,
 	 maps/1,
 	 api_macros/1,
-	 from_array/1, iolist_as_binary/1, resource/1, resource_binary/1, 
+	 from_array/1, iolist_as_binary/1, resource/1, resource_binary/1,
 	 resource_takeover/1,
-	 threading/1, send/1, send2/1, send3/1, send_threaded/1, neg/1, 
+	 threading/1, send/1, send2/1, send3/1, send_threaded/1, neg/1,
 	 is_checks/1,
 	 get_length/1, make_atom/1, make_string/1, reverse_list_test/1,
 	 otp_9828/1,
@@ -79,7 +80,7 @@ all() ->
     [{group, G} || G <- api_groups()]
         ++
     [reload_error, heap_frag, types, many_args,
-     select,
+     select, select_steal,
      {group, monitor},
      monitor_frenzy,
      hipe,
@@ -144,7 +145,8 @@ init_per_testcase(nif_whereis_threaded, Config) ->
         true -> Config;
         false -> {skip, "No thread support"}
     end;
-init_per_testcase(select, Config) ->
+init_per_testcase(Select, Config) when Select =:= select;
+                                       Select =:= select_steal ->
     case os:type() of
         {win32,_} ->
             {skip, "Test not yet implemented for windows"};
@@ -152,6 +154,9 @@ init_per_testcase(select, Config) ->
             Config
     end;
 init_per_testcase(_Case, Config) ->
+    %% Clear any resource dtor data before test starts in case another tc
+    %% left it in a bad state
+    catch last_resource_dtor_call(),
     Config.
 
 end_per_testcase(t_on_load, _Config) ->
@@ -590,7 +595,71 @@ select_3(_Config) ->
     {_,_,2} = last_resource_dtor_call(),
     ok.
 
-check_stop_ret(?ERL_NIF_SELECT_STOP_CALLED) -> ok;    
+%% @doc The stealing child process for the select_steal test. Duplicates given
+%% W/RFds and runs select on them to steal
+select_steal_child_process(Parent, RFd) ->
+    %% Duplicate the resource with the same FD
+    {R2Fd, _R2Ptr} = dupe_resource_nif(RFd),
+    Ref2 = make_ref(),
+
+    %% Try to select from the child pid (steal from parent)
+    ?assertEqual(0, select_nif(R2Fd, ?ERL_NIF_SELECT_READ, R2Fd, null, Ref2)),
+    ?assertEqual([], flush(0)),
+    ?assertEqual(eagain, read_nif(R2Fd, 1)),
+
+    %% Check that now events arrive to this temporary process
+    Parent ! {self(), stage1}, % signal parent to send the <<"stolen1">>
+
+    %% Receive <<"stolen1">> via enif_select
+    ?assertEqual(0, select_nif(R2Fd, ?ERL_NIF_SELECT_READ, R2Fd, null, Ref2)),
+    ?assertMatch([{select, R2Fd, Ref2, ready_input}], flush()),
+    ?assertEqual(<<"stolen1">>, read_nif(R2Fd, 7)),
+
+    clear_select_nif(R2Fd),
+
+    % do not do this here - stop_selecting(R2Fd, R2Rsrc, Ref2),
+    Parent ! {self(), done}.
+
+%% @doc Similar to select/1 test, make a double ended pipe. Then try to steal
+%% the socket, see what happens.
+select_steal(Config) when is_list(Config) ->
+    ensure_lib_loaded(Config),
+
+    Ref = make_ref(),
+    {{RFd, RPtr}, {WFd, WPtr}} = pipe_nif(),
+
+    %% Bind the socket to current pid in enif_select
+    ?assertEqual(0, select_nif(RFd, ?ERL_NIF_SELECT_READ, RFd, null, Ref)),
+    ?assertEqual([], flush(0)),
+
+    %% Spawn a process and do some stealing
+    Parent = self(),
+    Pid = spawn_link(fun() -> select_steal_child_process(Parent, RFd) end),
+
+    %% Signal from the child to send the first message
+    {Pid, stage1} = receive_any(),
+    ?assertEqual(ok, write_nif(WFd, <<"stolen1">>)),
+
+    ?assertMatch([{Pid, done}], flush(1)),  % synchronize with the child
+
+    %% Try to select from the parent pid (steal back)
+    ?assertEqual(0, select_nif(RFd, ?ERL_NIF_SELECT_READ, RFd, Pid, Ref)),
+
+    %% Ensure that no data is hanging and close.
+    %% Rfd is stolen at this point.
+    check_stop_ret(select_nif(WFd, ?ERL_NIF_SELECT_STOP, WFd, null, Ref)),
+    ?assertMatch([{fd_resource_stop, WPtr, _}], flush()),
+    {1, {WPtr, 1}} = last_fd_stop_call(),
+
+    check_stop_ret(select_nif(RFd, ?ERL_NIF_SELECT_STOP, RFd, null, Ref)),
+    ?assertMatch([{fd_resource_stop, RPtr, _}], flush()),
+    {1, {RPtr, _DirectCall}} = last_fd_stop_call(),
+
+    ?assert(is_closed_nif(WFd)),
+
+    ok.
+
+check_stop_ret(?ERL_NIF_SELECT_STOP_CALLED) -> ok;
 check_stop_ret(?ERL_NIF_SELECT_STOP_SCHEDULED) -> ok.
 
 write_full(W, C) ->
@@ -1101,6 +1170,12 @@ maps(Config) when is_list(Config) ->
     {1, M2} = make_map_remove_nif(M2, "key3"),
     {0, undefined} = make_map_remove_nif(self(), key),
 
+    M1 = maps_from_list_nif(maps:to_list(M1)),
+    M2 = maps_from_list_nif(maps:to_list(M2)),
+    M3 = maps_from_list_nif(maps:to_list(M3)),
+
+    has_duplicate_keys = maps_from_list_nif([{1,1},{1,1}]),
+
     verify_tmpmem(TmpMem),
     ok.
  
@@ -1723,14 +1798,9 @@ send2(Config) when is_list(Config) ->
 
 %% Send msg from user thread
 send_threaded(Config) when is_list(Config) ->
-    case erlang:system_info(smp_support) of
-	true ->
-	    send2_do1(fun(ME,To) -> send_blob_thread_dbg(ME,To,join) end),
-	    send2_do1(fun(ME,To) -> send_blob_thread_and_join(ME,To) end),
-	    ok;
-	false ->
-	    {skipped,"No threaded send on non-SMP"}
-    end.
+    send2_do1(fun(ME,To) -> send_blob_thread_dbg(ME,To,join) end),
+    send2_do1(fun(ME,To) -> send_blob_thread_and_join(ME,To) end),
+    ok.
 
 
 send2_do1(SendBlobF) ->
@@ -2193,9 +2263,8 @@ nif_schedule(Config) when is_list(Config) ->
     {B,A} = call_nif_schedule(A, B),
     ok = try call_nif_schedule(1, 2)
 	 catch
-	     error:badarg ->
-		 [{?MODULE,call_nif_schedule,[1,2],_}|_] =
-		     erlang:get_stacktrace(),
+	     error:badarg:Stk ->
+		 [{?MODULE,call_nif_schedule,[1,2],_}|_] = Stk,
 		 ok
 	 end,
     ok.
@@ -2365,8 +2434,8 @@ nif_raise_exceptions(NifFunc) ->
                             erlang:apply(?MODULE,NifFunc,[Term]),
                             ct:fail({expected,Term})
                         catch
-                            error:Term ->
-                                [{?MODULE,NifFunc,[Term],_}|_] = erlang:get_stacktrace(),
+                            error:Term:Stk ->
+                                [{?MODULE,NifFunc,[Term],_}|_] = Stk,
                                 ok
                         end
                 end, ok, ExcTerms).
@@ -2594,16 +2663,23 @@ nif_term_to_binary(Config) ->
 
 nif_binary_to_term(Config) ->
     ensure_lib_loaded(Config),
-    T = {#{ok => nok}, <<0:8096>>, lists:seq(1,100)},
+    BigMap = maps:from_list([{I,-I} || I <- lists:seq(1,100)]),
+    [nif_binary_to_term_do(T)
+     || T <- [{#{ok => nok}, <<0:8096>>, lists:seq(1,100)},
+              atom, 42, self(), BigMap]],
+    ok.
+
+nif_binary_to_term_do(T) ->
+    Dummy = [true|false],
     Bin = term_to_binary(T),
     Len = byte_size(Bin),
-    {Len,T} = binary_to_term_nif(Bin, undefined, 0),
+    {Len,T,Dummy} = binary_to_term_nif(Bin, undefined, 0),
     Len = binary_to_term_nif(Bin, self(), 0),
-    T = receive M -> M after 1000 -> timeout end,
+    {T,Dummy} = receive M -> M after 1000 -> timeout end,
 
-    {Len, T} = binary_to_term_nif(Bin, undefined, ?ERL_NIF_BIN2TERM_SAFE),
+    {Len,T,Dummy} = binary_to_term_nif(Bin, undefined, ?ERL_NIF_BIN2TERM_SAFE),
     false = binary_to_term_nif(<<131,100,0,14,"undefined_atom">>,
-			   undefined, ?ERL_NIF_BIN2TERM_SAFE),
+                               undefined, ?ERL_NIF_BIN2TERM_SAFE),
     false = binary_to_term_nif(Bin, undefined, 1),
     ok.
 
@@ -2973,14 +3049,17 @@ nif_ioq(Config) ->
          {enqbraw,a},
          {enqbraw,a, 5},
          {peek,   a},
+         {peek_head,  a},
          {deq,    a, 42},
 
          %% Test enqv
          {enqv,   a, 2, 100},
+         {peek_head,  a},
          {deq,    a, all},
 
          %% This skips all elements but one in the iolist
          {enqv,   a, 5, iolist_size(nif_ioq_payload(5)) - 1},
+         {peek_head,  a},
          {peek,   a},
 
          %% Ensure that enqueued refc binaries are intact after a roundtrip.
@@ -3026,9 +3105,13 @@ nif_ioq(Config) ->
 
     Q = ioq_nif(create),
 
+    false = ioq_nif(peek_head, Q),
+
     {'EXIT', {badarg, _}} = (catch ioq_nif(deq, Q, 1)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, 1, 1234)),
 
+    false = ioq_nif(peek_head, Q),
+
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, [atom_in_list], 0)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, [make_ref()], 0)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, [256], 0)),
@@ -3037,6 +3120,8 @@ nif_ioq(Config) ->
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, [1 bsl 64], 0)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(enqv, Q, [{tuple}], 0)),
 
+    false = ioq_nif(peek_head, Q),
+
     {'EXIT', {badarg, _}} = (catch ioq_nif(inspect,  [atom_in_list], use_stack)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(inspect,  [make_ref()], no_stack)),
     {'EXIT', {badarg, _}} = (catch ioq_nif(inspect,  [256], use_stack)),
@@ -3097,6 +3182,20 @@ nif_ioq_run([{peek, Name} = H|T], State) ->
 
     true = iolist_to_binary(B) == iolist_to_binary(Data),
     nif_ioq_run(T, State);
+nif_ioq_run([{peek_head, Name} = H|T], State) ->
+    #{ q := IOQ, b := B } = maps:get(Name, State),
+    RefData = iolist_to_binary(B),
+
+    ct:log("~p", [H]),
+
+    {true, QueueHead} = ioq_nif(peek_head, IOQ),
+    true = byte_size(QueueHead) > 0,
+
+    {RefHead, _Tail} = split_binary(RefData, byte_size(QueueHead)),
+
+    true = QueueHead =:= RefHead,
+
+    nif_ioq_run(T, State);
 nif_ioq_run([{deq, Name, all}|T], State) ->
     #{ q := IOQ, b := B } = maps:get(Name, State),
     Size = ioq_nif(size, IOQ),
@@ -3224,10 +3323,12 @@ binary_to_term_nif(_, _, _) -> ?nif_stub.
 port_command_nif(_, _) -> ?nif_stub.
 format_term_nif(_,_) -> ?nif_stub.
 select_nif(_,_,_,_,_) -> ?nif_stub.
+dupe_resource_nif(_) -> ?nif_stub.
 pipe_nif() -> ?nif_stub.
 write_nif(_,_) -> ?nif_stub.
 read_nif(_,_) -> ?nif_stub.
 is_closed_nif(_) -> ?nif_stub.
+clear_select_nif(_) -> ?nif_stub.
 last_fd_stop_call() -> ?nif_stub.
 alloc_monitor_resource_nif() -> ?nif_stub.
 monitor_process_nif(_,_,_,_) -> ?nif_stub.
diff --git a/erts/emulator/test/nif_SUITE_data/nif_SUITE.c b/erts/emulator/test/nif_SUITE_data/nif_SUITE.c
index b47d013bd2..155bda6df0 100644
--- a/erts/emulator/test/nif_SUITE_data/nif_SUITE.c
+++ b/erts/emulator/test/nif_SUITE_data/nif_SUITE.c
@@ -2140,24 +2140,45 @@ static ERL_NIF_TERM make_map_remove_nif(ErlNifEnv* env, int argc, const ERL_NIF_
 /* maps */
 static ERL_NIF_TERM maps_from_list_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 {
-    ERL_NIF_TERM cell = argv[0];
-    ERL_NIF_TERM map  = enif_make_new_map(env);
-    ERL_NIF_TERM tuple;
-    const ERL_NIF_TERM *pair;
-    int arity = -1;
+    ERL_NIF_TERM *keys, *values;
+    ERL_NIF_TERM result, cell;
+    unsigned count;
+
+    if (argc != 1 || !enif_get_list_length(env, argv[0], &count)) {
+        return enif_make_badarg(env);
+    }
 
-    if (argc != 1 && !enif_is_list(env, cell)) return enif_make_badarg(env);
+    keys = enif_alloc(sizeof(ERL_NIF_TERM) * count * 2);
+    values = keys + count;
 
-    /* assume sorted keys */
+    cell = argv[0];
+    count = 0;
 
-    while (!enif_is_empty_list(env,cell)) {
-	if (!enif_get_list_cell(env, cell, &tuple, &cell)) return enif_make_badarg(env);
-	if (enif_get_tuple(env,tuple,&arity,&pair)) {
-	    enif_make_map_put(env, map, pair[0], pair[1], &map);
-	}
+    while (!enif_is_empty_list(env, cell)) {
+        const ERL_NIF_TERM *pair;
+        ERL_NIF_TERM tuple;
+        int arity;
+
+        if (!enif_get_list_cell(env, cell, &tuple, &cell)
+            || !enif_get_tuple(env, tuple, &arity, &pair)
+            || arity != 2) {
+            enif_free(keys);
+            return enif_make_badarg(env);
+        }
+
+        keys[count] = pair[0];
+        values[count] = pair[1];
+
+        count++;
+    }
+
+    if (!enif_make_map_from_arrays(env, keys, values, count, &result)) {
+        result = enif_make_atom(env, "has_duplicate_keys");
     }
 
-    return map;
+    enif_free(keys);
+
+    return result;
 }
 
 static ERL_NIF_TERM sorted_list_from_maps_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]) {
@@ -2384,7 +2405,7 @@ static ERL_NIF_TERM term_to_binary(ErlNifEnv* env, int argc, const ERL_NIF_TERM
 static ERL_NIF_TERM binary_to_term(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 {
     ErlNifBinary bin;
-    ERL_NIF_TERM term, ret_term;
+    ERL_NIF_TERM term, dummy, ret_term;
     ErlNifPid pid;
     ErlNifEnv *msg_env = env;
     unsigned int opts;
@@ -2397,6 +2418,9 @@ static ERL_NIF_TERM binary_to_term(ErlNifEnv* env, int argc, const ERL_NIF_TERM
 	|| !enif_get_uint(env, argv[2], &opts))
         return enif_make_badarg(env);
 
+    /* build dummy heap term first to provoke OTP-15080 */
+    dummy = enif_make_list_cell(msg_env, atom_true, atom_false);
+
     ret = enif_binary_to_term(msg_env, bin.data, bin.size, &term,
 			      (ErlNifBinaryToTerm)opts);
     if (!ret)
@@ -2404,11 +2428,12 @@ static ERL_NIF_TERM binary_to_term(ErlNifEnv* env, int argc, const ERL_NIF_TERM
 
     ret_term = enif_make_uint64(env, ret);
     if (msg_env != env) {
-        enif_send(env, &pid, msg_env, term);
+        enif_send(env, &pid, msg_env,
+                  enif_make_tuple2(msg_env, term, dummy));
         enif_free_env(msg_env);
         return ret_term;
     } else {
-        return enif_make_tuple2(env, ret_term, term);
+        return enif_make_tuple3(env, ret_term, term, dummy);
     }
 }
 
@@ -2448,6 +2473,13 @@ static int get_fd(ErlNifEnv* env, ERL_NIF_TERM term, struct fd_resource** rsrc)
     return 1;
 }
 
+/* Returns: badarg
+ *    Or an enif_select result, which is a combination of bits:
+ *    ERL_NIF_SELECT_STOP_CALLED = 1
+ *    ERL_NIF_SELECT_STOP_SCHEDULED = 2
+ *    ERL_NIF_SELECT_INVALID_EVENT = 4
+ *    ERL_NIF_SELECT_FAILED = 8
+ */
 static ERL_NIF_TERM select_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 {
     struct fd_resource* fdr;
@@ -2479,6 +2511,9 @@ static ERL_NIF_TERM select_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv
 }
 
 #ifndef __WIN32__
+/*
+ * Create a read-write pipe with two fds (to read and to write)
+ */
 static ERL_NIF_TERM pipe_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 {
     struct fd_resource* read_rsrc;
@@ -2514,6 +2549,30 @@ static ERL_NIF_TERM pipe_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]
                enif_make_tuple2(env, write_fd, make_pointer(env, write_rsrc)));
 }
 
+/*
+ * Create (dupe) of a resource with the same fd, to test stealing
+ */
+static ERL_NIF_TERM dupe_resource_nif(ErlNifEnv* env, int argc,
+                                      const ERL_NIF_TERM argv[]) {
+    struct fd_resource* orig_rsrc;
+
+    if (!get_fd(env, argv[0], &orig_rsrc)) {
+        return enif_make_badarg(env);
+    } else {
+        struct fd_resource* new_rsrc;
+        ERL_NIF_TERM new_fd;
+
+        new_rsrc = enif_alloc_resource(fd_resource_type,
+                                       sizeof(struct fd_resource));
+        new_rsrc->fd = orig_rsrc->fd;
+        new_rsrc->was_selected = 0;
+        new_fd = enif_make_resource(env, new_rsrc);
+        enif_release_resource(new_rsrc);
+
+        return enif_make_tuple2(env, new_fd, make_pointer(env, new_rsrc));
+    }
+}
+
 static ERL_NIF_TERM write_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 {
     struct fd_resource* fdr;
@@ -2589,6 +2648,20 @@ static ERL_NIF_TERM is_closed_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM a
 
     return fdr->fd < 0 ? atom_true : atom_false;
 }
+
+static ERL_NIF_TERM clear_select_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
+{
+    struct fd_resource* fdr = NULL;
+
+    if (!get_fd(env, argv[0], &fdr))
+        return enif_make_badarg(env);
+
+    fdr->fd = -1;
+    fdr->was_selected = 0;
+
+    return atom_ok;
+}
+
 #endif /* !__WIN32__ */
 
 
@@ -2790,7 +2863,7 @@ unsigned rand_bits(struct frenzy_rand_bits* rnd, unsigned int nbits)
 
 struct frenzy_monitor {
     ErlNifMutex* lock;
-    enum { 
+    volatile enum {
         MON_FREE, MON_FREE_DOWN, MON_FREE_DEMONITOR,
         MON_TRYING, MON_ACTIVE, MON_PENDING
     } state;
@@ -3152,13 +3225,24 @@ static void frenzy_resource_down(ErlNifEnv* env, void* obj, ErlNifPid* pid,
     DBG_TRACE3("DOWN pid=%T, r=%p rix=%u\n", pid->pid, r, r->rix);
 
     for (mix = 0; mix < FRENZY_MONITORS_MAX; mix++) {
-        if (r->monv[mix].pid.pid == pid->pid && r->monv[mix].state >= MON_TRYING) {
+        int state1 = r->monv[mix].state;
+        /* First do dirty access of pid and state without the lock */
+        if (r->monv[mix].pid.pid == pid->pid && state1 >= MON_TRYING) {
+            int state2;
             enif_mutex_lock(r->monv[mix].lock);
-            if (enif_compare_monitors(mon, &r->monv[mix].mon) == 0) {
-                assert(r->monv[mix].state >= MON_ACTIVE);
-                r->monv[mix].state = MON_FREE_DOWN;
-                enif_mutex_unlock(r->monv[mix].lock);
-                return;
+            state2 = r->monv[mix].state;
+            if (state2 >= MON_ACTIVE) {
+                if (enif_compare_monitors(mon, &r->monv[mix].mon) == 0) {
+                    r->monv[mix].state = MON_FREE_DOWN;
+                    enif_mutex_unlock(r->monv[mix].lock);
+                    return;
+                }
+            }
+            else {
+                assert(state2 != MON_TRYING);
+                assert(state1 == MON_TRYING ||  /* racing monitor failed */
+                       state2 == MON_FREE_DEMONITOR || /* racing demonitor */
+                       state2 == MON_FREE_DOWN);       /* racing down */
             }
             enif_mutex_unlock(r->monv[mix].lock);
         }
@@ -3355,6 +3439,15 @@ static ERL_NIF_TERM ioq(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[])
 		return enif_make_badarg(env);
 	    else
 		return enif_make_atom(env, "true");
+        } else if (enif_is_identical(argv[0], enif_make_atom(env, "peek_head"))) {
+            ERL_NIF_TERM head_term;
+
+            if(enif_ioq_peek_head(env, ioq->q, NULL, &head_term)) {
+                return enif_make_tuple2(env,
+                    enif_make_atom(env, "true"), head_term);
+            }
+
+            return enif_make_atom(env, "false");
         } else if (enif_is_identical(argv[0], enif_make_atom(env, "peek"))) {
             int iovlen, num, i, off = 0;
             SysIOVec *iov = enif_ioq_peek(ioq->q, &iovlen);
@@ -3476,8 +3569,10 @@ static ErlNifFunc nif_funcs[] =
 #ifndef __WIN32__
     {"pipe_nif", 0, pipe_nif},
     {"write_nif", 2, write_nif},
+    {"dupe_resource_nif", 1, dupe_resource_nif},
     {"read_nif", 2, read_nif},
     {"is_closed_nif", 1, is_closed_nif},
+    {"clear_select_nif", 1, clear_select_nif},
 #endif
     {"last_fd_stop_call", 0, last_fd_stop_call},
     {"alloc_monitor_resource_nif", 0, alloc_monitor_resource_nif},
diff --git a/erts/emulator/test/node_container_SUITE.erl b/erts/emulator/test/node_container_SUITE.erl
index 8e9e3cb05a..7df001fec5 100644
--- a/erts/emulator/test/node_container_SUITE.erl
+++ b/erts/emulator/test/node_container_SUITE.erl
@@ -405,6 +405,7 @@ node_table_gc(Config) when is_list(Config) ->
     PreKnown = nodes(known),
     io:format("PreKnown = ~p~n", [PreKnown]),
     make_node_garbage(0, 200000, 1000, []),
+    receive after 1000 -> ok end, %% Wait for thread progress...
     PostKnown = nodes(known),
     PostAreas = erlang:system_info(allocated_areas),
     io:format("PostKnown = ~p~n", [PostKnown]),
@@ -965,6 +966,8 @@ check_refc(ThisNodeName,ThisCreation,Table,EntryList) when is_list(EntryList) ->
                         {case Referrer of
                              {system,delayed_delete_timer} ->
                                  true;
+                             {system,thread_progress_delete_timer} ->
+                                 true;
                              _ ->
                                  DDT
                          end,
diff --git a/erts/emulator/test/num_bif_SUITE.erl b/erts/emulator/test/num_bif_SUITE.erl
index d950179882..290bb61fc8 100644
--- a/erts/emulator/test/num_bif_SUITE.erl
+++ b/erts/emulator/test/num_bif_SUITE.erl
@@ -146,7 +146,7 @@ t_float_to_string(Config) when is_list(Config) ->
 				     123456789012345678.0, [{decimals, 237}])),
     {'EXIT', {badarg, _}} = (catch float_to_binary(
 				     123456789012345678.0, [{decimals, 237}])),
-    test_fts("1." ++ string:copies("0", 249) ++ "e+00",
+    test_fts("1." ++ lists:duplicate(249, $0) ++ "e+00",
 	     1.0,  [{scientific, 249}, compact]),
 
     X1 = float_to_list(1.0),
@@ -213,6 +213,20 @@ fts_rand_float_decimals(N) ->
     [begin
          F0 = rand_float_reasonable(),
          L0 = float_to_list(F0, [{decimals, D}]),
+         case conform_with_io_lib_format_os(F0,D) of
+             false -> ok;
+             true ->
+                 IOL = lists:flatten(io_lib:format("~.*f", [D, F0])),
+                 true = case L0 =:= IOL of
+                            true -> true;
+                            false ->
+                                io:format("F0 = ~w ~w\n",  [F0, <<F0/float>>]),
+                                io:format("decimals = ~w\n",  [D]),
+                                io:format("float_to_list = ~s\n",  [L0]),
+                                io:format("io_lib:format = ~s\n",  [IOL]),
+                                false
+                        end
+         end,
          L1 = case D of
                   0 -> L0 ++ ".0";
                   _ -> L0
@@ -234,6 +248,26 @@ fts_rand_float_decimals(N) ->
 
     fts_rand_float_decimals(N-1).
 
+conform_with_io_lib_format_os(F, D) ->
+    case os:type() of
+        {win32,_} ->
+            %% io_lib:format("~.*f") buggy on windows? OTP-15010
+            false;
+        _ ->
+            conform_with_io_lib_format(F, D)
+    end.
+
+conform_with_io_lib_format(_, 0) ->
+    %% io_lib:format("~.*f") does not support zero decimals
+    false;
+conform_with_io_lib_format(_, D) when D > 10 ->
+    %% Seems float_to_list gets it slightly wrong sometimes for many decimals
+    false;
+conform_with_io_lib_format(F, D) ->
+    %% io_lib:format prints '0' for input bits beyond mantissa precision
+    %% float_to_list treats those unknown input bits as if they were zeros.
+    math:log2(abs(F) * math:pow(10,D)) < 54.
+
 max_diff_decimals(F, D) ->
     IntBits = floor(math:log2(abs(F))) + 1,
     FracBits = (52 - IntBits),
diff --git a/erts/emulator/test/port_SUITE.erl b/erts/emulator/test/port_SUITE.erl
index 730a17d7e8..eb9b94a316 100644
--- a/erts/emulator/test/port_SUITE.erl
+++ b/erts/emulator/test/port_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 1997-2017. All Rights Reserved.
+%% Copyright Ericsson AB 1997-2018. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -109,7 +109,6 @@
     mon_port_pid_demonitor/1,
     mon_port_remote_on_remote/1,
     mon_port_driver_die/1,
-    mon_port_driver_die_demonitor/1,
     mul_basic/1,
     mul_slow_writes/1,
     name1/1,
@@ -180,8 +179,7 @@ all() ->
      mon_port_bad_named,
      mon_port_pid_demonitor,
      mon_port_name_demonitor,
-     mon_port_driver_die,
-     mon_port_driver_die_demonitor
+     mon_port_driver_die
     ].
 
 groups() ->
@@ -967,7 +965,7 @@ env_slave(File, Env) ->
 
 env_slave(File, Env, Body) ->
     file:write_file(File, term_to_binary(Body)),
-    Program = atom_to_list(lib:progname()),
+    Program = ct:get_progname(),
     Dir = filename:dirname(code:which(?MODULE)),
     Cmd = Program ++ " -pz " ++ Dir ++
     " -noinput -run " ++ ?MODULE_STRING ++ " env_slave_main " ++
@@ -1035,6 +1033,9 @@ huge_env(Config) when is_list(Config) ->
     try erlang:open_port({spawn,Cmd},[exit_status, {env, Env}]) of
         P ->
             receive
+                {P, {exit_status,N}} = M when N > 127->
+                    %% If exit status is > 127 something went very wrong
+                    ct:fail("Open port failed got ~p",[M]);
                 {P, {exit_status,N}} = M ->
                     %% We test that the exit status is an integer, this means
                     %% that the child program has started. If we get an atom
@@ -1128,7 +1129,7 @@ try_bad_args(Args) ->
 cd(Config)  when is_list(Config) ->
     ct:timetrap({minutes, 1}),
 
-    Program = atom_to_list(lib:progname()),
+    Program = ct:get_progname(),
     DataDir = proplists:get_value(data_dir, Config),
     TestDir = filename:join(DataDir, "dir"),
     Cmd = Program ++ " -pz " ++ DataDir ++
@@ -1190,7 +1191,7 @@ cd(Config)  when is_list(Config) ->
 %% be relative the new cwd and not the original
 cd_relative(Config) ->
 
-    Program = atom_to_list(lib:progname()),
+    Program = ct:get_progname(),
     DataDir = proplists:get_value(data_dir, Config),
     TestDir = filename:join(DataDir, "dir"),
 
@@ -1213,7 +1214,7 @@ cd_relative(Config) ->
 
 relative_cd() ->
 
-    Program = atom_to_list(lib:progname()),
+    Program = ct:get_progname(),
     ok = file:set_cwd(".."),
     {ok, Cwd} = file:get_cwd(),
 
@@ -1662,13 +1663,7 @@ spawn_executable(Config) when is_list(Config) ->
     [ExactFile2,"hello world","dlrow olleh"] =
     run_echo_args_2(unicode:characters_to_binary("\""++ExactFile2++"\" "++"\"hello world\" \"dlrow olleh\"")),
 
-    ExeExt =
-    case string:to_lower(lists:last(string:tokens(ExactFile2,"."))) of
-        "exe" ->
-            ".exe";
-        _ ->
-            ""
-    end,
+    ExeExt = filename:extension(ExactFile2),
     Executable2 = "spoky name"++ExeExt,
     file:copy(ExactFile1,filename:join([SpaceDir,Executable2])),
     ExactFile3 = filename:nativename(filename:join([SpaceDir,Executable2])),
@@ -1836,7 +1831,7 @@ collect_data(Port) ->
     end.
 
 parse_echo_args_output(Data) ->
-    [lists:last(string:tokens(S,"|")) || S <- string:tokens(Data,"\r\n")].
+    [lists:last(string:lexemes(S,"|")) || S <- string:lexemes(Data,["\r\n",$\n])].
 
 %% Test that the emulator does not mix up ports when the port table wraps
 mix_up_ports(Config) when is_list(Config) ->
@@ -1962,7 +1957,7 @@ max_ports() ->
     erlang:system_info(port_limit).
 
 port_ix(Port) when is_port(Port) ->
-    ["#Port",_,PortIxStr] = string:tokens(erlang:port_to_list(Port),
+    ["#Port",_,PortIxStr] = string:lexemes(erlang:port_to_list(Port),
                                           "<.>"),
     list_to_integer(PortIxStr).
 
@@ -2786,7 +2781,7 @@ mon_port_driver_die(Config) ->
     end,
     ok.
 
-
+-ifdef(DISABLED_TESTCASE).
 %% 1. Spawn a port which will sleep 3 seconds
 %% 2. Monitor port
 %% 3. Port driver and dies horribly (via C driver_failure call). This should
@@ -2822,6 +2817,7 @@ mon_port_driver_die_demonitor(Config) ->
     after 5000 -> ?assert(false)
     end,
     ok.
+-endif.
 
 %% @doc Makes a controllable port for testing. Underlying mechanism of this
 %% port is not important, only important is our ability to close/kill it or
diff --git a/erts/emulator/test/port_trace_SUITE.erl b/erts/emulator/test/port_trace_SUITE.erl
index c78dc754a9..eba8f194e0 100644
--- a/erts/emulator/test/port_trace_SUITE.erl
+++ b/erts/emulator/test/port_trace_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 1999-2017. All Rights Reserved.
+%% Copyright Ericsson AB 1999-2018. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -78,13 +78,6 @@ end_per_group(_GroupName, Config) ->
     Config.
 
 
-init_per_testcase(driver_remote_send_term, Config) ->
-    case erlang:system_info(smp_support) of
-        false ->
-            {skip,"Only supported on smp systems"};
-        true ->
-            init_per_testcase(driver_remote_send_term_smp, Config)
-    end;
 init_per_testcase(Func, Config) when is_atom(Func), is_list(Config) ->
     erlang:trace(all, false, [all]),
     os:unsetenv("OUTPUTV"),
@@ -209,8 +202,7 @@ ports(_Config) ->
     erlang:port_close(Prt),
 
     [{trace,Prt,closed,normal},
-     {trace,Prt,unregister,port_trace_SUITE},
-     {trace,Prt,unlink,S}] = flush(),
+     {trace,Prt,unregister,port_trace_SUITE}] = flush(),
 
     ok.
 
@@ -482,8 +474,7 @@ failure_test(Failure, Reason) ->
         process_flag(trap_exit, false)
     end,
     [{trace, Prt, 'receive', {S, {command, Failure}}},
-     {trace, Prt, closed, Reason},
-     {trace, Prt, unlink, S}] = flush(),
+     {trace, Prt, closed, Reason}] = flush(),
 
     ok.
 
@@ -606,13 +597,11 @@ close(Prt, Flags) ->
 
     if Recv, Ports ->
             [{trace, Prt, 'receive', {S, close}},
-             {trace, Prt, closed, normal},
-             {trace, Prt, unlink, S}] = flush();
+             {trace, Prt, closed, normal}] = flush();
        Recv ->
             [{trace, Prt, 'receive', {S, close}}] = flush();
        Ports ->
-            [{trace, Prt, closed, normal},
-             {trace, Prt, unlink, S}] = flush();
+            [{trace, Prt, closed, normal}] = flush();
        true ->
             [] = flush()
     end.
diff --git a/erts/emulator/test/process_SUITE.erl b/erts/emulator/test/process_SUITE.erl
index a8bcfac84d..585c5a1871 100644
--- a/erts/emulator/test/process_SUITE.erl
+++ b/erts/emulator/test/process_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 1997-2017. All Rights Reserved.
+%% Copyright Ericsson AB 1997-2018. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -42,8 +42,9 @@
 	 process_info_lock_reschedule2/1,
 	 process_info_lock_reschedule3/1,
          process_info_garbage_collection/1,
+         process_info_smoke_all/1,
+         process_info_status_handled_signal/1,
 	 bump_reductions/1, low_prio/1, binary_owner/1, yield/1, yield2/1,
-	 process_status_exiting/1,
 	 otp_4725/1, bad_register/1, garbage_collect/1, otp_6237/1,
 	 process_info_messages/1, process_flag_badarg/1, process_flag_heap_size/1,
 	 spawn_opt_heap_size/1, spawn_opt_max_heap_size/1,
@@ -80,7 +81,8 @@ all() ->
      process_info_lock_reschedule2,
      process_info_lock_reschedule3,
      process_info_garbage_collection,
-     process_status_exiting,
+     process_info_smoke_all,
+     process_info_status_handled_signal,
      bump_reductions, low_prio, yield, yield2, otp_4725,
      bad_register, garbage_collect, process_info_messages,
      process_flag_badarg, process_flag_heap_size,
@@ -512,14 +514,20 @@ pio_current_location(N, Pid, Pi, Looper) ->
     case Where of
 	{erlang,process_info,2,[]} ->
 	    pio_current_location(N-1, Pid, Pi+1, Looper);
+	{erts_internal,await_result,1, Loc} when is_list(Loc) ->
+	    pio_current_location(N-1, Pid, Pi+1, Looper);
 	{?MODULE,process_info_looper,1,Loc} when is_list(Loc) ->
-	    pio_current_location(N-1, Pid, Pi, Looper+1)
+	    pio_current_location(N-1, Pid, Pi, Looper+1);
+	_ ->
+	    exit({unexpected_location, Where})
     end.
 
 pio_current_stacktrace() ->
     L = [begin
-	     {current_stacktrace,Stk} = process_info(P, current_stacktrace),
-	     {P,Stk}
+	     case process_info(P, current_stacktrace) of
+                 {current_stacktrace, Stk} -> {P,Stk};
+                 undefined -> {P, []}
+             end
 	 end || P <- processes()],
     [erlang:garbage_collect(P) || {P,_} <- L],
     erlang:garbage_collect(),
@@ -841,28 +849,6 @@ process_info_lock_reschedule3(Config) when is_list(Config) ->
 		  ct:fail(BadStatus)
 	  end.
 
-process_status_exiting(Config) when is_list(Config) ->
-    %% Make sure that erts_debug:get_internal_state({process_status,P})
-    %% returns exiting if it is in status P_EXITING.
-    erts_debug:set_internal_state(available_internal_state,true),
-    Prio = process_flag(priority, max),
-    P = spawn_opt(fun () -> receive after infinity -> ok end end,
-			[{priority, normal}]),
-    erlang:yield(),
-    %% The tok_loop processes are here to make it hard for the exiting
-    %% process to be scheduled in for exit...
-    TokLoops = lists:map(fun (_) ->
-		spawn_opt(fun tok_loop/0,
-		    [link,{priority, high}])
-	end, lists:seq(1, erlang:system_info(schedulers_online))),
-    exit(P, boom),
-    wait_until(fun() ->
-		exiting =:= erts_debug:get_internal_state({process_status,P})
-	end),
-    lists:foreach(fun (Tok) -> unlink(Tok), exit(Tok,bang) end, TokLoops),
-    process_flag(priority, Prio),
-    ok.
-
 otp_4725(Config) when is_list(Config) ->
     Tester = self(),
     Ref1 = make_ref(),
@@ -997,10 +983,110 @@ process_info_garbage_collection(_Config) ->
 gv(Key,List) ->
     proplists:get_value(Key,List).
 
+process_info_smoke_all_tester() ->
+    register(process_info_smoke_all_tester, self()),
+    put(ets_ref, ets:new(blupp, [])),
+    put(binary, [list_to_binary(lists:duplicate(1000, 1)),
+                 list_to_binary(lists:duplicate(1000, 2))]),
+    process_info_smoke_all_tester_loop().
+
+process_info_smoke_all_tester_loop() ->
+    receive
+        {other_process, Pid} ->
+            case get(procs) of
+                undefined -> put(procs, [Pid]);
+                Procs -> put(procs, [Pid|Procs])
+            end,
+            erlang:monitor(process, Pid),
+            link(Pid),
+            process_info_smoke_all_tester_loop()
+    end.
+
+process_info_smoke_all(Config) when is_list(Config) ->
+    AllPIOptions = [registered_name,
+                    current_function,
+                    initial_call,
+                    messages,
+                    message_queue_len,
+                    links,
+                    monitors,
+                    monitored_by,
+                    dictionary,
+                    trap_exit,
+                    error_handler,
+                    heap_size,
+                    stack_size,
+                    memory,
+                    garbage_collection,
+                    group_leader,
+                    reductions,
+                    priority,
+                    trace,
+                    binary,
+                    sequential_trace_token,
+                    catchlevel,
+                    backtrace,
+                    last_calls,
+                    total_heap_size,
+                    suspending,
+                    min_heap_size,
+                    min_bin_vheap_size,
+                    max_heap_size,
+                    current_location,
+                    current_stacktrace,
+                    message_queue_data,
+                    garbage_collection_info,
+                    magic_ref,
+                    fullsweep_after],
+
+    {ok, Node} = start_node(Config, ""),
+    RP = spawn_link(Node, fun process_info_smoke_all_tester/0),
+    LP = spawn_link(fun process_info_smoke_all_tester/0),
+    RP ! {other_process, LP},
+    LP ! {other_process, RP},
+    LP ! {other_process, self()},
+    LP ! ets:new(blapp, []),
+    LP ! ets:new(blipp, []),
+    LP ! list_to_binary(lists:duplicate(1000, 3)),
+    receive after 1000 -> ok end,
+    _MLP = erlang:monitor(process, LP),
+    true = is_process_alive(LP),
+    PI = process_info(LP, AllPIOptions),
+    io:format("~p~n", [PI]),
+    garbage_collect(),
+    unlink(RP),
+    unlink(LP),
+    exit(RP, kill),
+    exit(LP, kill),
+    false = is_process_alive(LP),
+    stop_node(Node),
+    ok.
+
+process_info_status_handled_signal(Config) when is_list(Config) ->
+    P = spawn_link(fun () ->
+                           receive after infinity -> ok end
+                   end),
+    wait_until(fun () ->
+                       process_info(P, status) == {status, waiting}
+               end),
+    %%
+    %% The 'messages' option will force a process-info-request
+    %% signal to be scheduled on the process. Ensure that status
+    %% 'waiting' is reported even though it is actually running
+    %% when handling the request. We want it to report the status
+    %% it would have had if it had not been handling the
+    %% process-info-request...
+    %%
+    [{status, waiting}, {messages, []}] = process_info(P, [status, messages]),
+    unlink(P),
+    exit(P, kill),
+    false = erlang:is_process_alive(P),
+    ok.
+
 %% Tests erlang:bump_reductions/1.
 bump_reductions(Config) when is_list(Config) ->
     erlang:garbage_collect(),
-    receive after 1 -> ok end,		% Clear reductions.
+    erlang:yield(),		% Clear reductions.
     {reductions,R1} = process_info(self(), reductions),
     true = erlang:bump_reductions(100),
     {reductions,R2} = process_info(self(), reductions),
@@ -2131,36 +2217,44 @@ handle_event(Event, Pid) ->
 
 processes_term_proc_list(Config) when is_list(Config) ->
     Tester = self(),
-    as_expected = processes_term_proc_list_test(false),
-    {ok, Node} = start_node(Config, "+Mis true"),
-    RT = spawn_link(Node, fun () ->
-		receive after 1000 -> ok end,
-		processes_term_proc_list_test(false),
-		Tester ! {it_worked, self()}
-	end),
-    receive {it_worked, RT} -> ok end,
-    stop_node(Node),
+
+    Run = fun(Args) ->
+              {ok, Node} = start_node(Config, Args),
+              RT = spawn_link(Node, fun () ->
+                              receive after 1000 -> ok end,
+                              as_expected = processes_term_proc_list_test(false),
+                              Tester ! {it_worked, self()}
+                      end),
+              receive {it_worked, RT} -> ok end,
+              stop_node(Node)
+          end,
+
+    %% We have to run this test case with +S1 since instrument:allocations()
+    %% will report a free()'d block as present until it's actually deallocated
+    %% by its employer.
+    Run("+MSe true +MSatags false +S1"),
+    Run("+MSe true +MSatags true +S1"),
+
     ok.
-    
+
 -define(CHK_TERM_PROC_LIST(MC, XB),
 	chk_term_proc_list(?LINE, MC, XB)).
 
 chk_term_proc_list(Line, MustChk, ExpectBlks) ->
-    case {MustChk, instrument:memory_status(types)} of
-	{false, false} ->
+    Allocs = instrument:allocations(#{ allocator_types => [sl_alloc] }),
+    case {MustChk, Allocs} of
+	{false, {error, not_enabled}} ->
 	    not_enabled;
-	{_, MS} ->
-	    {value,
-	     {ptab_list_deleted_el,
-	      DL}} = lists:keysearch(ptab_list_deleted_el, 1, MS),
-	    case lists:keysearch(blocks, 1, DL) of
-		{value, {blocks, ExpectBlks, _, _}} ->
-		    ok;
-		{value, {blocks, Blks, _, _}} ->
-		    exit({line, Line,
-			  mismatch, expected, ExpectBlks, actual, Blks});
-		Unexpected ->
-		    exit(Unexpected)
+	{_, {ok, {_Shift, _Unscanned, ByOrigin}}} ->
+            ByType = maps:get(system, ByOrigin, #{}),
+            Hist = maps:get(ptab_list_deleted_el, ByType, {}),
+	    case lists:sum(tuple_to_list(Hist)) of
+		ExpectBlks ->
+                    ok;
+		Blks ->
+                    exit({line, Line, mismatch,
+                          expected, ExpectBlks,
+                          actual, Blks})
 	    end
     end,
     ok.
diff --git a/erts/emulator/test/receive_SUITE.erl b/erts/emulator/test/receive_SUITE.erl
index a12019ec83..1fe11428b4 100644
--- a/erts/emulator/test/receive_SUITE.erl
+++ b/erts/emulator/test/receive_SUITE.erl
@@ -25,14 +25,16 @@
 -include_lib("common_test/include/ct.hrl").
 
 -export([all/0, suite/0,
-	 call_with_huge_message_queue/1,receive_in_between/1]).
+	 call_with_huge_message_queue/1,receive_in_between/1,
+         receive_opt_exception/1,receive_opt_recursion/1]).
 
 suite() ->
     [{ct_hooks,[ts_install_cth]},
      {timetrap, {minutes, 3}}].
 
-all() -> 
-    [call_with_huge_message_queue, receive_in_between].
+all() ->
+    [call_with_huge_message_queue, receive_in_between,
+     receive_opt_exception, receive_opt_recursion].
 
 call_with_huge_message_queue(Config) when is_list(Config) ->
     Pid = spawn_link(fun echo_loop/0),
@@ -113,6 +115,60 @@ receive_one() ->
 	dummy -> ok
     end.
 
+receive_opt_exception(_Config) ->
+    Recurse = fun() ->
+                      %% Overwrite with the same mark,
+                      %% and never consume it.
+                      ThrowFun = fun() -> throw(aborted) end,
+                      aborted = (catch do_receive_opt_exception(ThrowFun)),
+                      ok
+              end,
+    do_receive_opt_exception(Recurse),
+
+    %% Eat the second message.
+    receive
+        Ref when is_reference(Ref) -> ok
+    end.
+
+do_receive_opt_exception(Disturber) ->
+    %% Create a receive mark.
+    Ref = make_ref(),
+    self() ! Ref,
+    Disturber(),
+    receive
+        Ref ->
+            ok
+    after 0 ->
+            error(the_expected_message_was_not_there)
+    end.
+
+receive_opt_recursion(_Config) ->
+    Recurse = fun() ->
+                      %% Overwrite with the same mark,
+                      %% and never consume it.
+                      NoOp = fun() -> ok end,
+                      BlackHole = spawn(NoOp),
+                      expected = do_receive_opt_recursion(BlackHole, NoOp, true),
+                      ok
+              end,
+    do_receive_opt_recursion(self(), Recurse, false),
+    ok.
+
+do_receive_opt_recursion(Recipient, Disturber, IsInner) ->
+    Ref = make_ref(),
+    Recipient ! Ref,
+    Disturber(),
+    receive
+        Ref -> ok
+    after 0 ->
+            case IsInner of
+                true ->
+                    expected;
+                false ->
+                    error(the_expected_message_was_not_there)
+            end
+    end.
+
 %%%
 %%% Common helpers.
 %%%
diff --git a/erts/emulator/test/scheduler_SUITE.erl b/erts/emulator/test/scheduler_SUITE.erl
index af33de237c..7eebbe8b19 100644
--- a/erts/emulator/test/scheduler_SUITE.erl
+++ b/erts/emulator/test/scheduler_SUITE.erl
@@ -57,6 +57,7 @@
 	 scheduler_suspend_basic/1,
 	 scheduler_suspend/1,
 	 dirty_scheduler_threads/1,
+         poll_threads/1,
 	 reader_groups/1]).
 
 suite() ->
@@ -72,6 +73,7 @@ all() ->
      {group, scheduler_bind}, scheduler_threads,
      scheduler_suspend_basic, scheduler_suspend,
      dirty_scheduler_threads,
+     poll_threads,
      reader_groups].
 
 groups() -> 
@@ -892,11 +894,9 @@ adjust_schedulers_online() ->
 
 read_affinity(Data) ->
     Exp = "pid " ++ os:getpid() ++ "'s current affinity mask",
-    case string:tokens(Data, ":") of
+    case string:lexemes(Data, ":") of
 	[Exp, DirtyAffinityStr] ->
-	    AffinityStr = string:strip(string:strip(DirtyAffinityStr,
-						    both, $ ),
-				       both, $\n),
+	    AffinityStr = string:trim(DirtyAffinityStr),
 	    case catch erlang:list_to_integer(AffinityStr, 16) of
 		Affinity when is_integer(Affinity) ->
 		    Affinity;
@@ -1083,7 +1083,6 @@ sbt_test(Config, CpuTCmd, ClBt, Bt, LP) ->
     ok.
 
 scheduler_threads(Config) when is_list(Config) ->
-    SmpSupport = erlang:system_info(smp_support),
     {Sched, SchedOnln, _} = get_sstate(Config, ""),
     %% Configure half the number of both the scheduler threads and
     %% the scheduler threads online.
@@ -1095,10 +1094,7 @@ scheduler_threads(Config) when is_list(Config) ->
     %% setting using +SP to 50% scheduler threads and 25% scheduler
     %% threads online. The result should be 2x scheduler threads and
     %% 1x scheduler threads online.
-    TwiceSched = case SmpSupport of
-                     false -> 1;
-                     true -> Sched*2
-                 end,
+    TwiceSched = Sched*2,
     FourSched = integer_to_list(Sched*4),
     FourSchedOnln = integer_to_list(SchedOnln*4),
     CombinedCmd1 = "+S "++FourSched++":"++FourSchedOnln++" +SP50:25",
@@ -1121,8 +1117,8 @@ scheduler_threads(Config) when is_list(Config) ->
 	    ResetCmd = "+S "++FourSched++":"++FourSchedOnln++" +S 0:0",
 	    {LProc, LProcAvail, _} = get_sstate(Config, ResetCmd),
 	    %% Test negative +S settings, but only for SMP-enabled emulators
-	    case {SmpSupport, LProc > 1, LProcAvail > 1} of
-		{true, true, true} ->
+	    case {LProc > 1, LProcAvail > 1} of
+		{true, true} ->
 		    SchedMinus1 = LProc-1,
 		    SchedOnlnMinus1 = LProcAvail-1,
 		    {SchedMinus1, SchedOnlnMinus1, _} = get_sstate(Config, "+S -1"),
@@ -1157,9 +1153,6 @@ dirty_scheduler_threads_test(Config) ->
     ok.
 
 dirty_schedulers_online_test() ->
-    dirty_schedulers_online_test(erlang:system_info(smp_support)).
-dirty_schedulers_online_test(false) -> ok;
-dirty_schedulers_online_test(true) ->
     dirty_schedulers_online_smp_test(erlang:system_info(schedulers_online)).
 dirty_schedulers_online_smp_test(SchedOnln) when SchedOnln < 4 -> ok;
 dirty_schedulers_online_smp_test(SchedOnln) ->
@@ -1453,6 +1446,79 @@ sst5_loop(N) ->
     erlang:system_flag(multi_scheduling, unblock_normal),
     sst5_loop(N-1).
 
+poll_threads(Config) when is_list(Config) ->
+    {Conc, PollType, KP} = get_ioconfig(Config),
+    {Sched, SchedOnln, _} = get_sstate(Config, ""),
+
+    [1, 1] = get_ionum(Config,"+IOt 2 +IOp 2"),
+    [1, 1, 1, 1, 1] = get_ionum(Config,"+IOt 5 +IOp 5"),
+
+    [1, 1] = get_ionum(Config, "+S 2 +IOPt 100 +IOPp 100"),
+
+    if
+        Conc ->
+            [5] = get_ionum(Config,"+IOt 5 +IOp 1"),
+            [3, 2] = get_ionum(Config,"+IOt 5 +IOp 2"),
+            [2, 2, 2, 2, 2] = get_ionum(Config,"+IOt 10 +IOPp 50"),
+
+            [2] = get_ionum(Config, "+S 2 +IOPt 100"),
+            [4] = get_ionum(Config, "+S 4 +IOPt 100"),
+            [4] = get_ionum(Config, "+S 4:2 +IOPt 100"),
+            [4, 4] = get_ionum(Config, "+S 8 +IOPt 100 +IOPp 25"),
+
+            fail = get_ionum(Config, "+IOt 1 +IOp 2"),
+
+            ok;
+        not Conc ->
+            [1, 1, 1, 1, 1] = get_ionum(Config,"+IOt 5 +IOp 1"),
+            [1, 1, 1, 1, 1] = get_ionum(Config,"+IOt 5 +IOp 2"),
+            [1, 1, 1, 1, 1, 1, 1, 1, 1, 1] = get_ionum(Config,"+IOt 10 +IOPp 50"),
+
+            [1, 1] = get_ionum(Config, "+S 2 +IOPt 100"),
+            [1, 1, 1, 1] = get_ionum(Config, "+S 4 +IOPt 100"),
+            [1, 1, 1, 1] = get_ionum(Config, "+S 4:2 +IOPt 100"),
+            [1, 1, 1, 1, 1, 1, 1, 1] = get_ionum(Config, "+S 8 +IOPt 100 +IOPp 25"),
+
+            [1] = get_ionum(Config, "+IOt 1 +IOp 2"),
+
+            ok
+    end,
+
+    fail = get_ionum(Config, "+IOt 1 +IOPp 101"),
+    fail = get_ionum(Config, "+IOt 0"),
+    fail = get_ionum(Config, "+IOPt 101"),
+
+    ok.
+
+get_ioconfig(Config) ->
+    [PS | _] = get_iostate(Config, ""),
+    {proplists:get_value(concurrent_updates, PS),
+     proplists:get_value(primary, PS),
+     proplists:get_value(kernel_poll, PS)}.
+
+get_ionum(Config, Cmd) ->
+    case get_iostate(Config, Cmd) of
+        fail -> fail;
+        PSs ->
+            lists:reverse(
+              lists:sort(
+                [proplists:get_value(poll_threads, PS) || PS <- PSs]))
+    end.
+
+get_iostate(Config, Cmd)->
+    case start_node(Config, Cmd) of
+        {ok, Node} ->
+            [IOStates] = mcall(Node,[fun () ->
+                                             erlang:system_info(check_io)
+                                     end]),
+            IO = [IOState || IOState <- IOStates,
+                             proplists:get_value(fallback, IOState) == false],
+            stop_node(Node),
+            IO;
+        {error,timeout} ->
+            fail
+    end.
+
 reader_groups(Config) when is_list(Config) ->
     %% White box testing. These results are correct, but other results
     %% could be too...
@@ -1777,18 +1843,24 @@ mcall(Node, Funs) ->
     Parent = self(),
     Refs = lists:map(fun (Fun) ->
                              Ref = make_ref(),
-                             spawn_link(Node,
-                                        fun () ->
-                                                Res = Fun(),
-                                                unlink(Parent),
-                                                Parent ! {Ref, Res}
-                                        end),
-                             Ref
+                             Pid = spawn(Node,
+                                         fun () ->
+                                                 Res = Fun(),
+                                                 unlink(Parent),
+                                                 Parent ! {Ref, Res}
+                                         end),
+                             MRef = erlang:monitor(process, Pid),
+                             {Ref, MRef}
                      end, Funs),
-    lists:map(fun (Ref) ->
+    lists:map(fun ({Ref, MRef}) ->
                       receive
                           {Ref, Res} ->
-                              Res
+                              receive
+                                  {'DOWN',MRef,_,_,_} ->
+                                      Res
+                              end;
+                          {'DOWN',MRef,_,_,Reason} ->
+                              Reason
                       end
               end, Refs).
 
diff --git a/erts/emulator/test/sensitive_SUITE.erl b/erts/emulator/test/sensitive_SUITE.erl
index c3e303bbd1..9b23a30e88 100644
--- a/erts/emulator/test/sensitive_SUITE.erl
+++ b/erts/emulator/test/sensitive_SUITE.erl
@@ -413,7 +413,7 @@ my_process_info(Pid, Tag) ->
 
 t_process_display(Config) when is_list(Config) ->
     Dir = filename:dirname(code:which(?MODULE)),
-    Cmd = atom_to_list(lib:progname()) ++ " -noinput -pa " ++ Dir ++
+    Cmd = ct:get_progname() ++ " -noinput -pa " ++ Dir ++
     " -run " ++ ?MODULE_STRING ++ " remote_process_display",
     io:put_chars(Cmd),
     P = open_port({spawn,Cmd}, [in,stderr_to_stdout,eof]),
diff --git a/erts/emulator/test/signal_SUITE.erl b/erts/emulator/test/signal_SUITE.erl
index f1d11d1814..fab2f45f28 100644
--- a/erts/emulator/test/signal_SUITE.erl
+++ b/erts/emulator/test/signal_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %% 
-%% Copyright Ericsson AB 2006-2017. All Rights Reserved.
+%% Copyright Ericsson AB 2006-2018. All Rights Reserved.
 %% 
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -34,23 +34,9 @@
 -export([init_per_testcase/2, end_per_testcase/2]).
 
 % Test cases
--export([xm_sig_order/1,
-	 pending_exit_unlink_process/1,
-	 pending_exit_unlink_dist_process/1,
-	 pending_exit_unlink_port/1,
-	 pending_exit_trap_exit/1,
-	 pending_exit_receive/1,
-	 pending_exit_exit/1,
-	 pending_exit_gc/1,
-	 pending_exit_is_process_alive/1,
-	 pending_exit_process_display/1,
-	 pending_exit_process_info_1/1,
-	 pending_exit_process_info_2/1,
-	 pending_exit_group_leader/1,
-	 exit_before_pending_exit/1]).
+-export([xm_sig_order/1]).
 
 init_per_testcase(Func, Config) when is_atom(Func), is_list(Config) ->
-    available_internal_state(true),
     [{testcase, Func}|Config].
 
 end_per_testcase(_Func, _Config) ->
@@ -60,24 +46,14 @@ init_per_suite(Config) ->
     Config.
 
 end_per_suite(_Config) ->
-    available_internal_state(true),
-    catch erts_debug:set_internal_state(not_running_optimization, true),
-    available_internal_state(false).
+    ok.
 
 suite() ->
     [{ct_hooks,[ts_install_cth]},
      {timetrap, {minutes, 2}}].
 
 all() -> 
-    [xm_sig_order, pending_exit_unlink_process,
-     pending_exit_unlink_dist_process,
-     pending_exit_unlink_port, pending_exit_trap_exit,
-     pending_exit_receive, pending_exit_trap_exit,
-     pending_exit_gc, pending_exit_is_process_alive,
-     pending_exit_process_display,
-     pending_exit_process_info_1,
-     pending_exit_process_info_2, pending_exit_group_leader,
-     exit_before_pending_exit].
+    [xm_sig_order].
 
 
 %% Test that exit signals and messages are received in correct order
@@ -113,367 +89,9 @@ xm_sig_order_proc() ->
     end,
     xm_sig_order_proc().
 
-pending_exit_unlink_process(Config) when is_list(Config) ->
-    pending_exit_test(self(), unlink).
-
-pending_exit_unlink_dist_process(Config) when is_list(Config) ->
-    {ok, Node} = start_node(Config),
-    From = spawn(Node, fun () -> receive after infinity -> ok end end),
-    Res = pending_exit_test(From, unlink),
-    stop_node(Node),
-    Res.
-
-pending_exit_unlink_port(Config) when is_list(Config) ->
-    pending_exit_test(hd(erlang:ports()), unlink).
-
-pending_exit_trap_exit(Config) when is_list(Config) ->
-    pending_exit_test(self(), trap_exit).
-
-pending_exit_receive(Config) when is_list(Config) ->
-    pending_exit_test(self(), 'receive').
-
-pending_exit_exit(Config) when is_list(Config) ->
-    pending_exit_test(self(), exit).
-
-pending_exit_gc(Config) when is_list(Config) ->
-    pending_exit_test(self(), gc).
-
-pending_exit_test(From, Type) ->
-    case catch erlang:system_info(smp_support) of
-        true ->
-            OTE = process_flag(trap_exit, true),
-            Ref = make_ref(),
-            Master = self(),
-            ExitBySignal = case Type of
-                               gc ->
-                                   lists:duplicate(10000,
-                                                   exit_by_signal);
-                               _ ->
-                                   exit_by_signal
-                           end,
-            Pid = spawn_link(
-                    fun () ->
-                            receive go -> ok end,
-                            false = have_pending_exit(),
-                            exit = fake_exit(From,
-                                             self(),
-                                             ExitBySignal),
-                            true = have_pending_exit(),
-                            Master ! {self(), Ref, Type},
-                            case Type of
-                                gc ->
-                                    force_gc(),
-                                    erlang:yield();
-                                unlink ->
-                                    unlink(From);
-                                trap_exit ->
-                                    process_flag(trap_exit, true);
-                                'receive' ->
-                                    receive _ -> ok
-                                    after 0 -> ok
-                                    end;
-                                exit ->
-                                    ok
-                            end,
-                            exit(exit_by_myself)
-                    end),
-            Mon = erlang:monitor(process, Pid),
-            Pid ! go,
-            Reason = receive
-                         {'DOWN', Mon, process, Pid, R} ->
-                             receive
-                                 {Pid, Ref, Type} ->
-                                     ok
-                             after 0 ->
-                                       ct:fail(premature_exit)
-                             end,
-                             case Type of
-                                 exit ->
-                                     exit_by_myself = R;
-                                 _ ->
-                                     ExitBySignal = R
-                             end
-                     end,
-            receive
-                {'EXIT', Pid, R2} ->
-                    Reason = R2
-            end,
-            process_flag(trap_exit, OTE),
-            ok,
-            {comment, "Test only valid with current SMP emulator."};
-        _ ->
-            {skipped, "SMP support not enabled. Test only valid with current SMP emulator."}
-    end.
-
-
-
-exit_before_pending_exit(Config) when is_list(Config) ->
-    %% This is a testcase testcase very specific to the smp
-    %% implementation as it is of the time of writing.
-    %%
-    %% The testcase tries to check that a process can
-    %% exit by itself even though it has a pending exit.
-    OTE = process_flag(trap_exit, true),
-    Master = self(),
-    Tester = spawn_link(
-               fun () ->
-                       Opts = case {erlang:system_info(run_queues),
-                                    erlang:system_info(schedulers_online)} of
-                                  {RQ, SO} when RQ =:= 1; SO =:= 1 -> [];
-                                  _ ->
-                                      process_flag(scheduler, 1),
-                                      [{scheduler, 2}]
-                              end,
-                       P = self(),
-                       Exiter = spawn_opt(fun () ->
-                                                  receive
-                                                      {exit_me, P, R} ->
-                                                          exit(P, R)
-                                                  end
-                                          end, Opts),
-                       erlang:yield(),
-                       Exiter ! {exit_me, self(), exited_by_exiter},
-                       %% We want to get a pending exit
-                       %% before we exit ourselves. We
-                       %% don't want to be scheduled out
-                       %% since we will then see the
-                       %% pending exit.
-                       %%
-                       %% Do something that takes
-                       %% relatively long time but
-                       %% consumes few reductions...
-                       repeat(fun() -> erlang:system_info(procs) end,10),
-                       %% ... then exit.
-                       Master ! {self(),
-                                 pending_exit,
-                                 have_pending_exit()},
-                       exit(exited_by_myself)
-               end),
-    PendingExit = receive {Tester, pending_exit, PE} -> PE end,
-    receive
-	      {'EXIT', Tester, exited_by_myself} ->
-		  process_flag(trap_exit, OTE),
-		  ok;
-	      Msg ->
-		  ct:fail({unexpected_message, Msg})
-	  end,
-    NoScheds = integer_to_list(erlang:system_info(schedulers_online)),
-    {comment,
-     "Was "
-     ++ case PendingExit of
-	    true -> "";
-	    false ->"*not*"
-	end ++ " able to trigger a pending exit. "
-     ++ "Running on " ++ NoScheds ++ " scheduler(s). "
-     ++ "This test is only interesting with at least two schedulers."}.
-
--define(PE_INFO_REPEAT, 100).
-
-pending_exit_is_process_alive(Config) when is_list(Config) ->
-    S = exit_op_test_init(),
-    TestFun = fun (P) -> false = is_process_alive(P) end,
-    repeated_exit_op_test(TestFun, ?PE_INFO_REPEAT),
-    verify_pending_exit_success(S),
-    comment().
-
-pending_exit_process_info_1(Config) when is_list(Config) ->
-    S = exit_op_test_init(),
-    TestFun = fun (P) ->
-			    undefined = process_info(P)
-		    end,
-    repeated_exit_op_test(TestFun, ?PE_INFO_REPEAT),
-    verify_pending_exit_success(S),
-    comment().
-
-pending_exit_process_info_2(Config) when is_list(Config) ->
-    S0 = exit_op_test_init(),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, messages)
-				end, ?PE_INFO_REPEAT),
-    S1 = verify_pending_exit_success(S0),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, status)
-				end, ?PE_INFO_REPEAT),
-    S2 = verify_pending_exit_success(S1),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, links)
-				end, ?PE_INFO_REPEAT),
-    S3 = verify_pending_exit_success(S2),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [messages])
-				end, ?PE_INFO_REPEAT),
-    S4 = verify_pending_exit_success(S3),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [status])
-				end, ?PE_INFO_REPEAT),
-    S5 = verify_pending_exit_success(S4),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [links])
-				end, ?PE_INFO_REPEAT),
-    S6 = verify_pending_exit_success(S5),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [status,
-								     links])
-				end, ?PE_INFO_REPEAT),
-    S7 = verify_pending_exit_success(S6),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [messages,
-								     status])
-				end, ?PE_INFO_REPEAT),
-    S8 = verify_pending_exit_success(S7),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [messages,
-								     links])
-				end, ?PE_INFO_REPEAT),
-    S9 = verify_pending_exit_success(S8),
-    repeated_exit_op_test(
-	    fun (P) ->
-		    undefined = process_info(P, [message_queue_len,
-						 status])
-	    end, ?PE_INFO_REPEAT),
-    S10 = verify_pending_exit_success(S9),
-    repeated_exit_op_test(fun (P) ->
-					undefined = process_info(P, [messages,
-								     links,
-								     status])
-				end, ?PE_INFO_REPEAT),
-    verify_pending_exit_success(S10),
-    comment().
-
-pending_exit_process_display(Config) when is_list(Config) ->
-    S = exit_op_test_init(),
-    TestFun = fun (P) ->
-			    badarg = try
-					 erlang:process_display(P, backtrace)
-				     catch
-					 error:badarg -> badarg
-				     end
-		    end,
-    repeated_exit_op_test(TestFun, ?PE_INFO_REPEAT),
-    verify_pending_exit_success(S),
-    comment().
-
-pending_exit_group_leader(Config) when is_list(Config) ->
-    S = exit_op_test_init(),
-    TestFun = fun (P) ->
-			    badarg = try
-					 group_leader(self(), P)
-				     catch
-					 error:badarg -> badarg
-				     end
-		    end,
-    repeated_exit_op_test(TestFun, ?PE_INFO_REPEAT),
-    verify_pending_exit_success(S),
-    comment().
-
 %%
 %% -- Internal utils --------------------------------------------------------
 %%
-exit_op_test_init() ->
-    put(no_pending_exit_success, 0),
-    put(no_pending_exit_tries, 0),
-    {case {erlang:system_info(run_queues),
-	   erlang:system_info(schedulers_online)} of
-	 {RQ, SO} when RQ =:= 1; SO =:= 1 -> false;
-	 _ -> true
-     end, 0, 0}.
-
-verify_pending_exit_success({false, _, _} = S) ->
-    S;
-verify_pending_exit_success({true, S, T}) ->
-    NewS = get(no_pending_exit_success),
-    NewT = get(no_pending_exit_tries),
-    case NewT =:= T of
-	true -> ok;
-	_ -> case NewS > S of
-		 true -> ok;
-		 _ -> exit(no_pending_exits)
-	     end
-    end,
-    {true, NewS, NewT}.
-
-comment() ->
-    {comment,
-     "Pending exit trigger ratio "
-     ++ integer_to_list(get(no_pending_exit_success))
-     ++ "/"
-     ++ integer_to_list(get(no_pending_exit_tries))
-     ++ "."
-     ++ case get(not_running_opt_test) of
-	    true -> " No 'not running optimization' to disable.";
-	    _ -> ""
-	end}.
-
-repeated_exit_op_test(TestFun, N) ->
-    WorkFun0 = fun () ->
-		       lists:sort(lists:reverse(lists:seq(1, 1000)))
-	       end,
-    repeat(fun () -> exit_op_test(TestFun, WorkFun0) end, N),
-    try erts_debug:set_internal_state(not_running_optimization, false) of
-	Bool when Bool == true; Bool == false -> 
-	    WorkFun1 = fun () ->
-			       erts_debug:set_internal_state(sleep, 0),
-			       lists:sort(lists:reverse(lists:seq(1, 1000)))
-			    end,
-	    repeat(fun () ->
-			   exit_op_test(TestFun, WorkFun1)
-		   end, N)
-    catch
-	error:notsup -> put(not_running_opt_test, true)
-    after
-	catch erts_debug:set_internal_state(not_running_optimization, true)
-    end.
-
-exit_op_test(TestFun, WorkFun) ->
-    Opts = case {erlang:system_info(run_queues),
-		 erlang:system_info(schedulers_online)} of
-	       {RQ, SO} when RQ =:= 1; SO =:= 1 -> [];
-	       _ ->
-		   process_flag(scheduler, 1),
-		   [{scheduler, 2}]
-	   end,
-    Master = self(),
-    Going = make_ref(),
-    P = spawn_opt(fun () ->
-			  loop(10, WorkFun),
-			  Master ! Going,
-			  loop(infinity, WorkFun)
-		  end, Opts),
-    receive Going -> ok end,
-    loop(10, WorkFun),
-    erlang:yield(),
-    exit(P, bang),
-    PE0 = have_pending_exit(P),
-    TestFun(P),
-    PE = case PE0 of
-	     true -> true;
-	     _ -> false
-	 end,
-    case {PE, get(no_pending_exit_success), get(no_pending_exit_tries)} of
-	{true, undefined, undefined} ->
-	    put(no_pending_exit_success, 1),
-	    put(no_pending_exit_tries, 1);
-	{false, undefined, undefined} ->
-	    put(no_pending_exit_success, 0),
-	    put(no_pending_exit_tries, 1);
-	{true, S, T} ->
-	    put(no_pending_exit_success, S+1),
-	    put(no_pending_exit_tries, T+1);
-	{false, _S, T} ->
-	    put(no_pending_exit_tries, T+1)
-    end,
-    ok.
-
-loop(infinity, WorkFun) ->
-    do_loop(infinity, WorkFun);
-loop(0, _WorkFun) ->
-    ok;
-loop(N, WorkFun) when is_integer(N) ->
-    do_loop(N-1, WorkFun).
-
-do_loop(N, WorkFun) ->
-    WorkFun(),
-    loop(N, WorkFun).
 
 repeat(_Fun, N) when is_integer(N), N =< 0 ->
     ok;
@@ -491,30 +109,3 @@ start_node(Config) ->
 
 stop_node(Node) ->
     test_server:stop_node(Node).
-
-have_pending_exit() ->
-    have_pending_exit(self()).
-
-have_pending_exit(Pid) ->
-    erts_debug:get_internal_state({have_pending_exit, Pid}).
-
-force_gc() ->
-    erts_debug:set_internal_state(force_gc, self()).
-
-fake_exit(From, To, Reason) ->
-    erts_debug:set_internal_state(send_fake_exit_signal, {From, To, Reason}).
-
-available_internal_state(Bool) when Bool == true; Bool == false ->
-    case {Bool,
-          (catch erts_debug:get_internal_state(available_internal_state))} of
-        {true, true} ->
-            true;
-        {false, true} ->
-            erts_debug:set_internal_state(available_internal_state, false),
-            true;
-        {true, _} ->
-            erts_debug:set_internal_state(available_internal_state, true),
-            false;
-        {false, _} ->
-            false
-    end.
diff --git a/erts/emulator/test/smoke_test_SUITE.erl b/erts/emulator/test/smoke_test_SUITE.erl
index 41bb07b84c..b3d34103f1 100644
--- a/erts/emulator/test/smoke_test_SUITE.erl
+++ b/erts/emulator/test/smoke_test_SUITE.erl
@@ -70,6 +70,20 @@ boot_combo(Config) when is_list(Config) ->
 	chk_boot(Config, "+Ktrue", NOOP),
 	chk_boot(Config, "+A42", A42),
 	chk_boot(Config, "+Ktrue +A42", A42),
+
+        WBTArgs = ["very_short", "short", "medium", "long", "very_long"],
+        WTArgs = ["very_low", "low", "medium", "high", "very_high"],
+        [chk_boot(Config,
+                 " +sbwt " ++ WBT ++
+                 " +sbwtdcpu " ++ WBT ++
+                 " +sbwtdio " ++ WBT ++
+                 " +swt " ++ WT ++
+                 " +swtdcpu " ++ WT ++
+                 " +swtdio " ++ WT, NOOP) || WBT <- WBTArgs,  WT <- WTArgs],
+
+        WSArgs = ["legacy", "default"],
+        [chk_boot(Config, " +sws " ++ WS, NOOP) || WS <- WSArgs],
+
 	%% A lot more combos could be implemented...
 	ok
     after
@@ -88,11 +102,9 @@ native_atomics(Config) when is_list(Config) ->
     {value,{NA32Key, NA32, _}} = lists:keysearch(NA32Key, 1, EthreadInfo),
     {value,{NA64Key, NA64, _}} = lists:keysearch(NA64Key, 1, EthreadInfo),
     {value,{DWNAKey, DWNA, _}} = lists:keysearch(DWNAKey, 1, EthreadInfo),
-    case {erlang:system_info(build_type), erlang:system_info(smp_support), NA32, NA64, DWNA} of
-	{opt, true, "no", "no", _} ->
+    case {erlang:system_info(build_type), NA32, NA64, DWNA} of
+	{opt, "no", "no", _} ->
 	    ct:fail(optimized_smp_runtime_without_native_atomics);
-	{_, false, "no", "no", _} ->
-	    {comment, "No native atomics"};
 	_ ->
 	    {comment,
 	     NA32 ++ " 32-bit, "
diff --git a/erts/emulator/test/statistics_SUITE.erl b/erts/emulator/test/statistics_SUITE.erl
index 40cc940a94..3f2897242e 100644
--- a/erts/emulator/test/statistics_SUITE.erl
+++ b/erts/emulator/test/statistics_SUITE.erl
@@ -310,6 +310,13 @@ scheduler_wall_time_all(Config) when is_list(Config) ->
     scheduler_wall_time_test(scheduler_wall_time_all).
 
 scheduler_wall_time_test(Type) ->
+    case string:find(erlang:system_info(system_version),
+                     "dirty-schedulers-TEST") == nomatch of
+        true -> run_scheduler_wall_time_test(Type);
+        false -> {skip, "Cannot be run with dirty-schedulers-TEST build"}
+    end.
+
+run_scheduler_wall_time_test(Type) ->
     %% Should return undefined if system_flag is not turned on yet
     undefined = statistics(Type),
     %% Turn on statistics
@@ -638,9 +645,7 @@ msacc(Config) ->
                         (aux, 0) ->
                              %% aux will be zero if we do not have smp support
                              %% or no async threads
-                             case erlang:system_info(smp_support) orelse
-                                  erlang:system_info(thread_pool_size) > 0
-                             of
+                             case erlang:system_info(thread_pool_size) > 0 of
                                  false ->
                                      ok;
                                  true ->
@@ -676,6 +681,16 @@ msacc_test(TmpFile) ->
     ets:insert(Tid, {1, hello}),
     ets:delete(Tid),
 
+    %% Check some IO
+    {ok, L} = gen_tcp:listen(0, [{active, true},{reuseaddr,true}]),
+    {ok, Port} = inet:port(L),
+    Pid = spawn(fun() ->
+                        {ok, S} = gen_tcp:accept(L),
+                        (fun F() -> receive M -> F() end end)()
+                end),
+    {ok, C} = gen_tcp:connect("localhost", Port, []),
+    [begin gen_tcp:send(C,"hello"),timer:sleep(1) end || _ <- lists:seq(1,100)],
+
     %% Collect some garbage
     [erlang:garbage_collect() || _ <- lists:seq(1,100)],
 
diff --git a/erts/emulator/test/system_info_SUITE.erl b/erts/emulator/test/system_info_SUITE.erl
index 56522039da..7309908337 100644
--- a/erts/emulator/test/system_info_SUITE.erl
+++ b/erts/emulator/test/system_info_SUITE.erl
@@ -37,6 +37,7 @@
 
 -export([process_count/1, system_version/1, misc_smoke_tests/1,
          heap_size/1, wordsize/1, memory/1, ets_limit/1, atom_limit/1,
+         ets_count/1,
          atom_count/1]).
 
 suite() ->
@@ -45,6 +46,7 @@ suite() ->
 
 all() -> 
     [process_count, system_version, misc_smoke_tests,
+     ets_count,
      heap_size, wordsize, memory, ets_limit, atom_limit, atom_count].
 
 %%%
@@ -308,9 +310,9 @@ memory_test(_Config) ->
 
     mem_workers_call(MWs,
 		     fun () ->
-			     list_to_atom("an ugly atom "++integer_to_list(erlang:system_info(scheduler_id))),
-			     list_to_atom("another ugly atom "++integer_to_list(erlang:system_info(scheduler_id))),
-			     list_to_atom("yet another ugly atom "++integer_to_list(erlang:system_info(scheduler_id)))
+			     _ = list_to_atom("an ugly atom "++integer_to_list(erlang:system_info(scheduler_id))),
+			     _ = list_to_atom("another ugly atom "++integer_to_list(erlang:system_info(scheduler_id))),
+			     _ = list_to_atom("yet another ugly atom "++integer_to_list(erlang:system_info(scheduler_id)))
 		     end, []),
     cmp_memory(MWs, "new atoms"),
 
@@ -478,6 +480,21 @@ get_node_name(Config) ->
 		 ++ "-"
 		 ++ integer_to_list(erlang:unique_integer([positive]))).
 
+ets_count(Config) when is_list(Config) ->
+    [ets_count_do([Type | Named])
+     || Type <- [set, bag, duplicate_bag, ordered_set],
+        Named <- [[named_table], []]
+    ],
+    ok.
+
+ets_count_do(Opts) ->
+    Before = erlang:system_info(ets_count),
+    T = ets:new(?MODULE, Opts),
+    After = erlang:system_info(ets_count),
+    After = Before + 1,
+    ets:delete(T),
+    Before = erlang:system_info(ets_count).
+
 
 %% Verify system_info(ets_limit) reflects max ETS table settings.
 ets_limit(Config0) when is_list(Config0) ->
diff --git a/erts/emulator/test/system_profile_SUITE.erl b/erts/emulator/test/system_profile_SUITE.erl
index 9b678fcff9..ae27bfe9df 100644
--- a/erts/emulator/test/system_profile_SUITE.erl
+++ b/erts/emulator/test/system_profile_SUITE.erl
@@ -95,18 +95,20 @@ do_runnable_procs({TsType, TsTypeFlag}) ->
     % FIXME: Set #laps and #nodes in config file
     Nodes = 10,
     Laps = 10,
-    Master = ring(Nodes),
+    All = ring(Nodes, [link,monitor]),
+    [Master | _] = All,
     undefined = erlang:system_profile(Pid, [runnable_procs]++TsTypeFlag),
     % loop a message
     ok = ring_message(Master, message, Laps),
+    ok = kill_ring(Master),
+    [receive {'DOWN', _, process, P, _} -> ok end || P <- All],
     Events = get_profiler_events(),
-    kill_em_all = kill_ring(Master),
     erlang:system_profile(undefined, []),
     put(master, Master),
     put(laps, Laps),
     true = has_runnable_event(TsType, Events),
     Pids = sort_events_by_pid(Events),
-    ok = check_events(TsType, Pids),
+    ok = check_events(TsType, Pids, (Laps+1)*2+2, (Laps+1)*2),
     erase(),
     exit(Pid,kill),
     ok.
@@ -139,16 +141,15 @@ do_runnable_ports({TsType, TsTypeFlag}, Config) ->
     erlang:system_profile(undefined, []),
     true = has_runnable_event(TsType, Events),
     Pids = sort_events_by_pid(Events),
-    ok = check_events(TsType, Pids),
+    ok = check_events(TsType, Pids, Laps*2+2, Laps*2),
     erase(),
     exit(Pid,kill),
     ok.
 
 %% Tests system_profiling with scheduler.
 scheduler(Config) when is_list(Config) ->
-    case {erlang:system_info(smp_support), erlang:system_info(schedulers_online)} of
-	{false,_} -> {skipped, "No need for scheduler test when smp support is disabled."};
-	{_,    1} -> {skipped, "No need for scheduler test when only one scheduler online."};
+    case erlang:system_info(schedulers_online) of
+	1 -> {skipped, "No need for scheduler test when only one scheduler online."};
 	_ ->
 	    Nodes = 10,
 	    lists:foreach(fun (TsType) ->
@@ -172,12 +173,12 @@ dont_profile_profiler(Config) when is_list(Config) ->
 
     Nodes = 10,
     Laps = 10,
-    Master = ring(Nodes),
+    [Master|_] = ring(Nodes, [link]),
     undefined = erlang:system_profile(Pid, [runnable_procs]),
     % loop a message
     ok = ring_message(Master, message, Laps),
     erlang:system_profile(undefined, []),
-    kill_em_all = kill_ring(Master),
+    ok = kill_ring(Master),
     Events = get_profiler_events(),
     false  = has_profiler_pid_event(Events, Pid),
 
@@ -249,27 +250,28 @@ check_block_system({TsType, TsTypeFlag}, Nodes) ->
 
 %%% Check events
 
-check_events(_TsType, []) -> ok;
-check_events(TsType, [Pid | Pids]) ->
+check_events(_TsType, [], _, _) -> ok;
+check_events(TsType, [Pid | Pids], ExpMaster, ExpMember) ->
     Master = get(master),
-    Laps = get(laps),
     CheckPids = get(pids),
     {Events, N} = get_pid_events(Pid),
     ok = check_event_flow(Events),
     ok = check_event_ts(TsType, Events),
     IsMember = lists:member(Pid, CheckPids),
-    case Pid of
-    	Master ->
-	    io:format("Expected ~p and got ~p profile events from ~p: ok~n", [Laps*2+2, N, Pid]),
-	    N = Laps*2 + 2,
-    	    check_events(TsType, Pids);
-	Pid when IsMember == true ->
-	    io:format("Expected ~p and got ~p profile events from ~p: ok~n", [Laps*2, N, Pid]),
-	    N = Laps*2,
-    	    check_events(TsType, Pids);
-	Pid ->
-	    check_events(TsType, Pids)
-    end.
+    {Title,Exp} = case Pid of
+                      Master -> {master,ExpMaster};
+                      Pid when IsMember == true -> {member,ExpMember};
+                      _ -> {other,N}
+                  end,
+    ok = case N of
+             Exp -> ok;
+             _ ->
+                 io:format("Expected ~p and got ~p profile events from ~p ~p:~n~p~n",
+                           [Exp, N, Title, Pid, Events]),
+                 error
+         end,
+    check_events(TsType, Pids, ExpMaster, ExpMember).
+
 
 %% timestamp consistency check for descending timestamps
 
@@ -297,7 +299,13 @@ check_event_ts(TsType, [{Pid, _, _, TS1}=Event | Events], {Pid,_,_,TS0}) ->
 %% consistency check for active vs. inactive activity (runnable)
 
 check_event_flow(Events) ->
-    check_event_flow(Events, undefined).
+    case check_event_flow(Events, undefined) of
+        ok -> ok;
+        Error ->
+            io:format("Events = ~p\n", [Events]),
+            Error
+    end.
+
 check_event_flow([], _) -> ok;
 check_event_flow([Event | PidEvents], undefined) ->
     check_event_flow(PidEvents, Event);
@@ -337,10 +345,11 @@ sort_events_by_pid([Event | Events],Pids) ->
 %% API
 
 % Returns master pid
-ring(N) -> 
-    Pids = build_ring(N, []),
+ring(N, SpawnOpt) -> 
+    Pids = build_ring(N, [], SpawnOpt),
     put(pids, Pids),
-    setup_ring(Pids).
+    setup_ring(Pids),
+    Pids.
 
 ring_message(Master, Message, Laps) ->
     Master ! {message, Master, Laps, Message},
@@ -348,13 +357,19 @@ ring_message(Master, Message, Laps) ->
     	{laps_complete, Master} -> ok
     end.
 
-kill_ring(Master) -> Master ! kill_em_all.
+kill_ring(Master) ->
+    Master ! kill_em_all,
+    ok.
 
 %% Process ring helpers
 
-build_ring(0, Pids) -> Pids;
-build_ring(N, Pids) ->
-    build_ring(N - 1, [spawn_link(?MODULE, ring_loop, [undefined]) | Pids]).
+build_ring(0, Pids, _) -> Pids;
+build_ring(N, Pids, SpawnOpt) ->
+    Pid = case spawn_opt(?MODULE, ring_loop, [undefined], SpawnOpt) of
+              {P,_} -> P;
+              P -> P
+          end,
+    build_ring(N-1, [Pid | Pids], SpawnOpt).
 
 setup_ring([Master | Relayers]) ->
     % Relayers may not include the master pid
@@ -383,15 +398,13 @@ ring_loop(RelayTo) ->
         {message, Master, Lap, Msg}=Message ->
             case {self(), Lap} of
                 {Master, 0} ->
-                    get(supervisor) ! {laps_complete, self()},
-                    ring_loop(RelayTo);
+                    get(supervisor) ! {laps_complete, self()};
                 {Master, Lap} ->
-                    RelayTo ! {message, Master, Lap - 1, Msg},
-                    ring_loop(RelayTo);
+                    RelayTo ! {message, Master, Lap - 1, Msg};
                 _ ->
-                    RelayTo ! Message,
-                    ring_loop(RelayTo)
-            end
+                    RelayTo ! Message
+            end,
+            ring_loop(RelayTo)
     end.
 
 %%%
diff --git a/erts/emulator/test/trace_SUITE.erl b/erts/emulator/test/trace_SUITE.erl
index a81aa64057..138aefb29c 100644
--- a/erts/emulator/test/trace_SUITE.erl
+++ b/erts/emulator/test/trace_SUITE.erl
@@ -1,7 +1,7 @@
 %%
 %% %CopyrightBegin%
 %%
-%% Copyright Ericsson AB 1997-2017. All Rights Reserved.
+%% Copyright Ericsson AB 1997-2018. All Rights Reserved.
 %%
 %% Licensed under the Apache License, Version 2.0 (the "License");
 %% you may not use this file except in compliance with the License.
@@ -29,7 +29,7 @@
          receive_trace/1, link_receive_call_correlation/1, self_send/1,
 	 timeout_trace/1, send_trace/1,
 	 procs_trace/1, dist_procs_trace/1, procs_new_trace/1,
-	 suspend/1, mutual_suspend/1, suspend_exit/1, suspender_exit/1,
+	 suspend/1, suspend_exit/1, suspender_exit/1,
 	 suspend_system_limit/1, suspend_opts/1, suspend_waiting/1,
 	 new_clear/1, existing_clear/1, tracer_die/1,
 	 set_on_spawn/1, set_on_first_spawn/1, cpu_timestamp/1,
@@ -53,7 +53,7 @@ all() ->
     [cpu_timestamp, receive_trace, link_receive_call_correlation,
      self_send, timeout_trace,
      send_trace, procs_trace, dist_procs_trace, suspend,
-     mutual_suspend, suspend_exit, suspender_exit,
+     suspend_exit, suspender_exit,
      suspend_system_limit, suspend_opts, suspend_waiting,
      new_clear, existing_clear, tracer_die, set_on_spawn,
      set_on_first_spawn, set_on_link, set_on_first_link,
@@ -235,7 +235,7 @@ link_receive_call_correlation(Config) when is_list(Config) ->
     1 = erlang:trace(Receiver, true, ['receive', procs, call, timestamp, scheduler_id]),
     1 = erlang:trace_pattern({?MODULE, receive_msg, '_'}, [], [local]),
 
-    Num = 100000,
+    Num = 100,
 
     (fun F(0) -> [];
          F(N) ->
@@ -255,7 +255,7 @@ link_receive_call_correlation(Config) when is_list(Config) ->
 
     Msgs = (fun F() -> receive M -> [M | F()] after 1 -> [] end end)(),
 
-    case check_consistent(Receiver, Num, Num, Num, Msgs) of
+    case check_consistent(Receiver, Num, Num, Num, Msgs, false, undefined) of
         ok ->
             ok;
         {error, Reason} ->
@@ -265,20 +265,63 @@ link_receive_call_correlation(Config) when is_list(Config) ->
 
 -define(schedid, , _).
 
-check_consistent(_Pid, Recv, Call, _LU, [Msg | _]) when Recv > Call ->
+check_consistent(_Pid, Recv, Call, _LU, [Msg | _], _Received, _LinkedN) when Recv > Call ->
     {error, Msg};
-check_consistent(Pid, Recv, Call, LU, [Msg | Msgs]) ->
+check_consistent(Pid, Recv, Call, LU, [Msg | Msgs], false, undefined) ->
 
     case Msg of
         {trace, Pid, 'receive', Recv ?schedid} ->
-            check_consistent(Pid,Recv - 1, Call, LU, Msgs);
+            check_consistent(Pid,Recv - 1, Call, LU, Msgs, true, undefined);
         {trace_ts, Pid, 'receive', Recv ?schedid, _} ->
-            check_consistent(Pid,Recv - 1, Call, LU, Msgs);
+            check_consistent(Pid,Recv - 1, Call, LU, Msgs, true, undefined);
 
         {trace, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid} ->
-            check_consistent(Pid,Recv, Call - 1, LU, Msgs);
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, false, undefined);
         {trace_ts, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid, _} ->
-            check_consistent(Pid,Recv, Call - 1, LU, Msgs);
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, false, undefined);
+
+        {trace, Pid, _, _Self ?schedid} ->
+            check_consistent(Pid, Recv, Call, LU, Msgs, false, undefined);
+        {trace_ts, Pid, _, _Self ?schedid, _} ->
+            check_consistent(Pid, Recv, Call, LU, Msgs, false, undefined);
+
+        Msg ->
+            {error, Msg}
+    end;
+check_consistent(Pid, Recv, Call, LU, [Msg | Msgs], true, undefined) ->
+
+    case Msg of
+        {trace, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid} ->
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, true, undefined);
+        {trace_ts, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid, _} ->
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, true, undefined);
+
+        {trace, Pid, getting_linked, _Self ?schedid} ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, Recv rem 2);
+        {trace_ts, Pid, getting_linked, _Self ?schedid, _} ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, Recv rem 2);
+
+        {trace, Pid, getting_unlinked, _Self ?schedid} ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, (Recv+1) rem 2);
+        {trace_ts, Pid, getting_unlinked, _Self ?schedid, _} ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, (Recv+1) rem 2);
+
+        Msg ->
+            {error, Msg}
+    end;
+check_consistent(Pid, Recv, Call, LU, [Msg | Msgs], true, LinkedN) ->
+    UnlinkedN = (LinkedN + 1) rem 2,
+
+    case Msg of
+        {trace, Pid, 'receive', Recv ?schedid} when Recv == LU ->
+            check_consistent(Pid,Recv - 1, Call, LU, Msgs, true, LinkedN);
+        {trace_ts, Pid, 'receive', Recv ?schedid, _} when Recv == LU ->
+            check_consistent(Pid,Recv - 1, Call, LU, Msgs, true, LinkedN);
+
+        {trace, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid} ->
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, true, LinkedN);
+        {trace_ts, Pid, call, {?MODULE, receive_msg, [Call]} ?schedid, _} ->
+            check_consistent(Pid,Recv, Call - 1, LU, Msgs, true, LinkedN);
 
         %% We check that for each receive we have gotten a
         %% getting_linked or getting_unlinked message. Also
@@ -286,38 +329,38 @@ check_consistent(Pid, Recv, Call, LU, [Msg | Msgs]) ->
         %% message we expect to receive is an even number
         %% and odd number for getting_unlinked.
         {trace, Pid, getting_linked, _Self ?schedid}
-          when Recv rem 2 == 0, Recv == LU ->
-            check_consistent(Pid, Recv, Call, LU - 1, Msgs);
+          when Recv rem 2 == LinkedN ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, LinkedN);
         {trace_ts, Pid, getting_linked, _Self ?schedid, _}
-          when Recv rem 2 == 0, Recv == LU ->
-            check_consistent(Pid, Recv, Call, LU - 1, Msgs);
+          when Recv rem 2 == LinkedN ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, LinkedN);
 
         {trace, Pid, getting_unlinked, _Self ?schedid}
-          when Recv rem 2 == 1, Recv == LU ->
-            check_consistent(Pid, Recv, Call, LU - 1, Msgs);
+          when Recv rem 2 == UnlinkedN ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, LinkedN);
         {trace_ts, Pid, getting_unlinked, _Self ?schedid, _}
-          when Recv rem 2 == 1, Recv == LU ->
-            check_consistent(Pid, Recv, Call, LU - 1, Msgs);
+          when Recv rem 2 == UnlinkedN ->
+            check_consistent(Pid, Recv, Call, LU - 1, Msgs, true, LinkedN);
 
         {trace,Pid,'receive',Ignore ?schedid}
           when Ignore == stop; Ignore == timeout ->
-            check_consistent(Pid, Recv, Call, LU, Msgs);
+            check_consistent(Pid, Recv, Call, LU, Msgs, true, LinkedN);
         {trace_ts,Pid,'receive',Ignore ?schedid,_}
           when Ignore == stop; Ignore == timeout ->
-            check_consistent(Pid, Recv, Call, LU, Msgs);
+            check_consistent(Pid, Recv, Call, LU, Msgs, true, LinkedN);
 
         {trace, Pid, exit, normal ?schedid} ->
-            check_consistent(Pid, Recv, Call, LU, Msgs);
+            check_consistent(Pid, Recv, Call, LU, Msgs, true, LinkedN);
         {trace_ts, Pid, exit, normal  ?schedid, _} ->
-            check_consistent(Pid, Recv, Call, LU, Msgs);
+            check_consistent(Pid, Recv, Call, LU, Msgs, true, LinkedN);
         {'EXIT', Pid, normal} ->
-            check_consistent(Pid, Recv, Call, LU, Msgs);
+            check_consistent(Pid, Recv, Call, LU, Msgs, true, LinkedN);
         Msg ->
             {error, Msg}
     end;
-check_consistent(_, 0, 0, 0, []) ->
+check_consistent(_, 0, 0, 1, [], true, _) ->
     ok;
-check_consistent(_, Recv, Call, LU, []) ->
+check_consistent(_, Recv, Call, LU, [], _, _) ->
     {error,{Recv, Call, LU}}.
 
 receive_msg(M) ->
@@ -1191,55 +1234,6 @@ do_suspend(Pid, N) ->
     erlang:yield(),
     do_suspend(Pid, N-1).
 
-
-
-mutual_suspend(Config) when is_list(Config) ->
-    TimeoutSecs = 5*60,
-    ct:timetrap({seconds, TimeoutSecs}),
-    Parent = self(),
-    Fun = fun () ->
-                  receive
-                      {go, Pid} ->
-                          do_mutual_suspend(Pid, 100000)
-                  end,
-                  Parent ! {done, self()},
-                  receive after infinity -> ok end
-          end,
-    P1 = spawn_link(Fun),
-    P2 = spawn_link(Fun),
-    T1 = erlang:start_timer((TimeoutSecs - 5)*1000, self(), oops),
-    T2 = erlang:start_timer((TimeoutSecs - 5)*1000, self(), oops),
-    P1 ! {go, P2},
-    P2 ! {go, P1},
-    Res1 = receive
-               {done, P1} -> done;
-               {timeout,T1,_} -> timeout
-           end,
-    Res2 = receive
-               {done, P2} -> done;
-               {timeout,T2,_} -> timeout
-           end,
-    P1S = process_info(P1, status),
-    P2S = process_info(P2, status),
-    io:format("P1S=~p P2S=~p", [P1S, P2S]),
-    false = {status, suspended} == P1S,
-    false = {status, suspended} == P2S,
-    unlink(P1), exit(P1, bang),
-    unlink(P2), exit(P2, bang),
-    done = Res1,
-    done = Res2,
-    ok.
-
-do_mutual_suspend(_Pid, 0) ->
-    ok;
-do_mutual_suspend(Pid, N) ->
-    %% Suspend a process and test that it is suspended.
-    true = erlang:suspend_process(Pid),
-    {status, suspended} = process_info(Pid, status),
-    %% Unsuspend the process.
-    true = erlang:resume_process(Pid),
-    do_mutual_suspend(Pid, N-1).		
-
 suspend_exit(Config) when is_list(Config) ->
     ct:timetrap({minutes, 2}),
     rand:seed(exsplus, {4711,17,4711}),
@@ -1470,7 +1464,8 @@ suspend_opts(Config) when is_list(Config) ->
                              dbl_async = AA,
                              synced = S,
                              async_once = AO} = Acc) ->
-                 erlang:suspend_process(Tok, [asynchronous]),
+                 Tag = {make_ref(), self()},
+                 erlang:suspend_process(Tok, [{asynchronous, Tag}]),
                  Res = case {suspend_count(Tok), N rem 4} of
                            {0, 2} ->
                                erlang:suspend_process(Tok,
@@ -1506,7 +1501,11 @@ suspend_opts(Config) when is_list(Config) ->
                            _ ->
                                Acc
                        end,
-                 erlang:resume_process(Tok),
+                 receive
+                     {Tag, Result} ->
+                         suspended = Result,
+                         erlang:resume_process(Tok)
+                 end,
                  erlang:yield(),
                  Res
          end,
diff --git a/erts/emulator/test/tracer_SUITE.erl b/erts/emulator/test/tracer_SUITE.erl
index ab7d047bc3..070462b0f1 100644
--- a/erts/emulator/test/tracer_SUITE.erl
+++ b/erts/emulator/test/tracer_SUITE.erl
@@ -30,7 +30,8 @@
 -export([load/1, unload/1, reload/1, invalid_tracers/1]).
 -export([send/1, recv/1, call/1, call_return/1, spawn/1, exit/1,
          link/1, unlink/1, getting_linked/1, getting_unlinked/1,
-         register/1, unregister/1, in/1, out/1, gc_start/1, gc_end/1]).
+         register/1, unregister/1, in/1, out/1, gc_start/1, gc_end/1,
+         seq_trace/1]).
 
 suite() -> [{ct_hooks,[ts_install_cth]},
             {timetrap, {minutes, 1}}].
@@ -41,7 +42,8 @@ all() ->
 groups() ->
     [{ basic, [], [send, recv, call, call_return, spawn, exit,
                    link, unlink, getting_linked, getting_unlinked,
-                   register, unregister, in, out, gc_start, gc_end]}].
+                   register, unregister, in, out, gc_start, gc_end,
+                   seq_trace]}].
 
 init_per_suite(Config) ->
     erlang:trace_pattern({'_','_','_'}, false, [local]),
@@ -583,6 +585,24 @@ gc_end(_Config) ->
 
     test(gc_major_end, garbage_collection, Tc, Expect, false).
 
+seq_trace(_Config) ->
+
+    seq_trace:set_system_tracer({tracer_test,
+                                 {#{ seq_trace => trace }, self(), []}}),
+    erlang:spawn(fun() ->
+                         seq_trace:set_token(label,17),
+                         seq_trace:set_token(print,true),
+                         seq_trace:print(17,"**** Trace Started ****")
+                 end),
+    receive
+        {seq_trace, _, 17, {print, _, _, _, _}, _} ->
+            ok;
+        M ->
+            ct:fail("~p~n",[M])
+    after 100 ->
+            ct:fail(timeout)
+    end.
+
 test(Event, Tc, Expect) ->
     test(Event, Tc, Expect, false).
 test(Event, Tc, Expect, Removes) ->
@@ -603,7 +623,7 @@ test(Event, TraceFlag, Tc, Expect, _Removes, Dies) ->
 
     Expect(Pid1, State1, Opts),
     receive M11 -> ct:fail({unexpected, M11}) after 0 -> ok end,
-    if not Dies ->
+    if not Dies andalso Event /= in ->
             {flags, [TraceFlag]} = erlang:trace_info(Pid1, flags),
             {tracer, {tracer_test, State1}} = erlang:trace_info(Pid1, tracer),
             erlang:trace(Pid1, false, [TraceFlag]);
@@ -620,7 +640,7 @@ test(Event, TraceFlag, Tc, Expect, _Removes, Dies) ->
     Expect(Pid1T, State1, Opts#{ scheduler_id => number,
                                  timestamp => timestamp}),
     receive M11T -> ct:fail({unexpected, M11T}) after 0 -> ok end,
-    if not Dies ->
+    if not Dies andalso Event /= in ->
             {flags, [scheduler_id, TraceFlag, timestamp]}
                 = erlang:trace_info(Pid1T, flags),
             {tracer, {tracer_test, State1}} = erlang:trace_info(Pid1T, tracer),
@@ -635,7 +655,7 @@ test(Event, TraceFlag, Tc, Expect, _Removes, Dies) ->
     Tc(Pid2),
     ok = trace_delivered(Pid2),
     receive M2 -> ct:fail({unexpected, M2}) after 0 -> ok end,
-    if not Dies ->
+    if not Dies andalso Event /= in ->
             {flags, [TraceFlag]} = erlang:trace_info(Pid2, flags),
             {tracer, {tracer_test, State2}} = erlang:trace_info(Pid2, tracer),
             erlang:trace(Pid2, false, [TraceFlag]);
diff --git a/erts/emulator/test/tuple_SUITE.erl b/erts/emulator/test/tuple_SUITE.erl
index 79b681b4d1..baf41180e0 100644
--- a/erts/emulator/test/tuple_SUITE.erl
+++ b/erts/emulator/test/tuple_SUITE.erl
@@ -134,6 +134,13 @@ t_element(Config) when is_list(Config) ->
     {'EXIT', {badarg, _}} = (catch element(1, id(42))),
     {'EXIT', {badarg, _}} = (catch element(id(1.5), id({a,b}))),
 
+    %% Make sure that the loader does not reject the module when
+    %% huge literal index values are used.
+    {'EXIT', {badarg, _}} = (catch element((1 bsl 24)-1, id({a,b,c}))),
+    {'EXIT', {badarg, _}} = (catch element(1 bsl 24, id({a,b,c}))),
+    {'EXIT', {badarg, _}} = (catch element(1 bsl 32, id({a,b,c}))),
+    {'EXIT', {badarg, _}} = (catch element(1 bsl 64, id({a,b,c}))),
+
     ok.
 
 get_elements([Element|Rest], Tuple, Pos) ->
diff --git a/erts/emulator/test/z_SUITE.erl b/erts/emulator/test/z_SUITE.erl
index feea7432a9..103f9f1550 100644
--- a/erts/emulator/test/z_SUITE.erl
+++ b/erts/emulator/test/z_SUITE.erl
@@ -37,6 +37,7 @@
 -export([schedulers_alive/1, node_container_refc_check/1,
 	 long_timers/1, pollset_size/1,
 	 check_io_debug/1, get_check_io_info/0,
+         lc_graph/1,
          leaked_processes/1]).
 
 suite() ->
@@ -46,6 +47,7 @@ suite() ->
 all() -> 
     [schedulers_alive, node_container_refc_check,
      long_timers, pollset_size, check_io_debug,
+     lc_graph,
      %% Make sure that the leaked_processes/1 is always
      %% run last.
      leaked_processes].
@@ -289,6 +291,12 @@ has_gethost([P|T]) ->
 has_gethost([]) ->
     false.
 
+lc_graph(Config) when is_list(Config) ->
+    %% Create "lc_graph" file in current working dir
+    %% if lock checker is enabled
+    erts_debug:lc_graph(),
+    ok.
+
 leaked_processes(Config) when is_list(Config) ->
     %% Replace the defualt timetrap with a timetrap with
     %% known pid.
@@ -330,7 +338,7 @@ display_check_io(ChkIo) ->
     ok.
 
 get_check_io_info() ->
-    ChkIo = erlang:system_info(check_io),
+    ChkIo = driver_SUITE:get_check_io_total(erlang:system_info(check_io)),
     PendUpdNo = case lists:keysearch(pending_updates, 1, ChkIo) of
 		    {value, {pending_updates, PendNo}} ->
 			PendNo;
diff --git a/erts/emulator/utils/beam_emu_vars b/erts/emulator/utils/beam_emu_vars
new file mode 100755
index 0000000000..c798a4dada
--- /dev/null
+++ b/erts/emulator/utils/beam_emu_vars
@@ -0,0 +1,122 @@
+#!/usr/bin/perl -w
+use strict;
+
+# Analyse beam_emu.s and try to find out the registers
+# used for the important variables in process_main().
+#
+# Works for .s files from clang or gcc. For gcc, the -fverbose-asm
+# option must be used.
+#
+# Example:
+#
+# $ beam-emu-vars -vars 'c_p E HTOP FCALLS I reg freg' beam_emu.s
+# E: %r13
+# FCALLS: %rcx:98 %rax:88 16(%rsp):50 %rdi:6
+# HTOP: %r10:382 64(%rsp):88 72(%rsp):9 24(%rsp):7 %rcx:6 %r15:6 80(%rsp):3 88(%rsp):2
+# I: %rbx
+# c_p: %rbp
+# freg: 48(%rsp):11 %rcx:8 %rdi:5 %rax:4
+# reg: %r12
+#
+# That means that E, I, c_p, reg seems to be assigned to permanent registers.
+# HTOP seems to be assigned %r10, but it is saved to a scratch location
+# before any function calls. FCALLS and freg seems to be saved in a location on
+# the stack and loaded into a register when used.
+#
+# The exit status will be 0 if all variables are assigned to registers (most of
+# the time), and 1 if one or more variables are assigned to a stack location.
+
+my $vars = 'c_p E FCALLS freg HTOP I reg';
+
+while (@ARGV and $ARGV[0] =~ /^-(.*)/) {
+    $_ = $1;
+    shift;
+    ($vars = shift), next if /^vars/;
+    die "$0: Bad option: -$_\n";
+}
+
+my @vars = split(" ", $vars);
+my %vars;
+@vars{@vars} = @vars;
+
+my $inside;
+my %count;
+
+if (@ARGV != 1) {
+    usage();
+}
+
+while (<>) {
+    if (!$inside && /[.]globl\s*_?process_main/) {
+	$inside = 1;
+    } elsif ($inside && /[.]globl/) {
+	last;
+    }
+    if ($inside) {
+	if (/##DEBUG_VALUE:\s*process_main:([A-Za-z]*)\s*<-\s*(.*)/) {
+	    # clang
+	    my($var,$reg) = ($1,$2);
+	    next if $reg =~ /^[-\d]+$/; # Ignore if number.
+	    $count{$var}->{$reg}++ if $vars{$var};
+	    next;
+	}
+
+	# Parse gcc verbose arguments. Comments are marked with
+	# one '#' (clang marks its comments with two '#').
+	my($src,$dst,$comment) = /movq\s+([^#]+), ([^#]+)#(?!#)\s*(.*)/;
+	next unless $comment;
+	$dst =~ s/\s*$//;
+	my($vsrc,$vdst) = split /,/, $comment, 2;
+	$vdst =~ s/^\s//;
+	update_count(\%count, $vsrc, $src);
+	update_count(\%count, $vdst, $dst);
+	if ($vars{$vdst} and $vsrc eq '%sfp') {
+	    $count{$vdst}->{$src}++;
+	}
+    }
+}
+
+my @first;
+
+OUTER:
+for my $var (sort keys %count) {
+    my $total = 0;
+
+    foreach my $reg (keys %{$count{$var}}) {
+	$total += $count{$var}->{$reg}++;
+    }
+
+    foreach my $reg (keys %{$count{$var}}) {
+	if ($count{$var}->{$reg} > 0.9*$total) {
+	    print "$var: $reg\n";
+	    push @first, $var;
+	    next OUTER;
+	}
+    }
+
+    my @r;
+    foreach my $reg (keys %{$count{$var}}) {
+	push @r, $reg;
+    }
+    @r = sort { $count{$var}->{$b} <=> $count{$var}->{$a} } @r;
+    @r = map { "$_:$count{$var}->{$_}" } @r;
+    push @first, $r[0];
+    print "$var: ", join(' ', @r), "\n";
+}
+
+foreach (@first) {
+    exit 1 if /%rsp/;
+}
+exit 0;
+
+sub update_count {
+    my($count_ref,$var,$reg) = @_;
+    return unless $vars{$var};
+    ${${$count_ref}{$var}}{$reg}++;
+}
+
+sub usage {
+    die qq[usage: beam_emu_vars [ -vars "var1 var2..." ] <filename>.s\n\n] .
+	"The exit status is 0 if all variables are assigned to registers,\n" .
+	"and 1 if one or more variables are allocated to a stack location.\n";
+}
diff --git a/erts/emulator/utils/beam_makeops b/erts/emulator/utils/beam_makeops
index 0a30553f71..da994fae3e 100755
--- a/erts/emulator/utils/beam_makeops
+++ b/erts/emulator/utils/beam_makeops
@@ -19,45 +19,40 @@
 # %CopyrightEnd%
 #
 use strict;
-use vars qw($BEAM_FORMAT_NUMBER);
+use vars qw($BEAM_FORMAT_NUMBER $GC_REGEXP);
+use constant COLD => 0;
+use constant WARM => 1;
+use constant HOT => 2;
+
+# Instructions for packing
+use constant PACK_JUMP => 1;
+use constant PACK_IN_INSTR_WORD => 2;
+use constant PACK_OPT_IN_INSTR_WORD => 4;
+
+# Packing commands
+use constant PACK_CMD_TIGHTEST => '1';
+use constant PACK_CMD_TIGHT => '2';
+use constant PACK_CMD_LOOSE => '3';
+use constant PACK_CMD_WIDE => '4';
 
 $BEAM_FORMAT_NUMBER = undef;
+$GC_REGEXP = undef;
 
 my $target = \&emulator_output;
 my $outdir = ".";		# Directory for output files.
 my $verbose = 0;
-my $hot = 1;
+my $hotness = 1;
 my $num_file_opcodes = 0;
 my $wordsize = 32;
-my %defs;			# Defines (from command line).
+my $code_pointers_are_short = 0; # Whether code pointers (to C code) are short.
+my $code_model = 'unknown';
+my %defs;                        # Defines (from command line).
 
 # This is shift counts and mask for the packer.
 my $WHOLE_WORD = '';
-my @pack_instr;
-my @pack_shift;
-my @pack_mask;
 
-$pack_instr[2] = ['6', 'i'];
-$pack_instr[3] = ['0', '0', 'i'];
-$pack_instr[4] = ['6', '6', '6', 'i']; # Only for 64 bit wordsize
-
-$pack_shift[2] = ['0', 'BEAM_LOOSE_SHIFT'];
-$pack_shift[3] = ['0', 'BEAM_TIGHT_SHIFT', '(2*BEAM_TIGHT_SHIFT)'];
-$pack_shift[4] = ['0', 'BEAM_LOOSE_SHIFT', # Only for 64 bit wordsize
-		  '(2*BEAM_LOOSE_SHIFT)',
-		  '(3*BEAM_LOOSE_SHIFT)'];
-
-$pack_mask[2]  = ['BEAM_LOOSE_MASK', $WHOLE_WORD];
-$pack_mask[3]  = ['BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK'];
-$pack_mask[4]  = ['BEAM_LOOSE_MASK', # Only for 64 bit wordsize
-		  'BEAM_LOOSE_MASK',
-		  'BEAM_LOOSE_MASK',
-		  $WHOLE_WORD];
-
-# Mapping from packagable arguments to number of packed arguments per
-# word.  Initialized after the wordsize is known.
-
-my @args_per_word;
+my @basic_pack_options = (0);
+my @extended_pack_options = @basic_pack_options;
 
 # There are two types of instructions: generic and specific.
 # The generic instructions are those generated by the Beam compiler.
@@ -83,6 +78,14 @@ my %num_specific;
 my %gen_to_spec;
 my %specific_op;
 
+# The following hashes are used for error checking.
+my %print_name;
+my %specific_op_arity;
+
+# Information about each specific operator. Key is the print name (e.g. get_list_xxy).
+# Value is a hash.
+my %spec_op_info;
+
 my %gen_arity;
 my @gen_arity;
 
@@ -91,17 +94,22 @@ my @op_to_name;
 
 my @obsolete;
 
-my %macro;
-my %macro_flags;
+# Instructions and micro instructions implemented in C.
+my %c_code;                     # C code block, location, arguments.
+my %c_code_used;                # Used or not.
 
-my %hot_code;
-my %cold_code;
+# Definitions for instructions combined from micro instructions.
+my %combined_instrs;
+
+my @generated_code;             # Generated code.
+my %sort_order;
 
 my @unnumbered_generic;
 my %unnumbered;
 
 my %is_transformed;
 
+
 #
 # Pre-processor.
 #
@@ -128,7 +136,10 @@ my $loader_types = "nprvlqo";
 my $genop_types = $compiler_types . $loader_types;
 
 #
-# Defines the argument types and their loaded size assuming no packing.
+# Define the operand types and their loaded size assuming no packing.
+#
+# Those are the types that can be used in the definition of a specific
+# instruction.
 #
 my %arg_size = ('r' => 0,	# x(0) - x register zero
 		'x' => 1,	# x(N), N > 0 - x register
@@ -138,23 +149,48 @@ my %arg_size = ('r' => 0,	# x(0) - x register zero
 		'n' => 0,	# NIL (implicit)
 		'c' => 1,	# tagged constant (integer, atom, nil)
 		's' => 1,	# tagged source; any of the above
+                'S' => 1,       # tagged source register (x or y)
 		'd' => 1,	# tagged destination register (r, x, y)
 		'f' => 1,	# failure label
 		'j' => 1,	# either 'f' or 'p'
 		'e' => 1,	# pointer to export entry
 		'L' => 0,	# label
-		'I' => 1,	# untagged integer
-		't' => 1,	# untagged integer -- can be packed
+		't' => 1,	# untagged integer (12 bits) -- can be packed
+		'I' => 1,	# untagged integer (32 bits) -- can be packed
+                'W' => 1,       # untagged integer/pointer (one word)
 		'b' => 1,	# pointer to bif
 		'A' => 1,	# arity value
 		'P' => 1,	# byte offset into tuple or stack
 		'Q' => 1,	# like 'P', but packable
-		'h' => 1,	# character
+		'h' => 1,	# character (not used)
 		'l' => 1,	# float reg
 		'q' => 1,	# literal term
 	     );
 
 #
+# Define the types that may be used in a transformation rule.
+#
+# %pattern_type defines the types that may be used in a pattern
+# on the left side.
+#
+# %construction_type defines the types that may be used when
+# constructing a new instruction on the right side (a subset of
+# the pattern types that are possible to construct).
+#
+my $pattern_types = "acdfjilnopqsuxy";
+my %pattern_type;
+@pattern_type{split("", $pattern_types)} = (1) x length($pattern_types);
+
+my %construction_type;
+foreach my $type (keys %pattern_type) {
+    $construction_type{$type} = 1
+        if index($genop_types, $type) >= 0;
+}
+foreach my $makes_no_sense ('f', 'j', 'o', 'p', 'q') {
+    delete $construction_type{$makes_no_sense};
+}
+
+#
 # Generate bits.
 #
 my %type_bit;
@@ -186,16 +222,17 @@ sub define_type_bit {
     define_type_bit('s', $type_bit{'d'} | $type_bit{'i'} |
 		    $type_bit{'a'} | $type_bit{'n'} |
 		    $type_bit{'q'});
+    define_type_bit('S', $type_bit{'d'});
     define_type_bit('j', $type_bit{'f'} | $type_bit{'p'});
 
-    # Aliases (for matching purposes).
-    define_type_bit('I', $type_bit{'u'});
+    # Aliases of 'u'.  Those specify how to load the operand and
+    # what kind of packing can be done.
     define_type_bit('t', $type_bit{'u'});
+    define_type_bit('I', $type_bit{'u'});
+    define_type_bit('W', $type_bit{'u'});
     define_type_bit('A', $type_bit{'u'});
     define_type_bit('L', $type_bit{'u'});
     define_type_bit('b', $type_bit{'u'});
-    define_type_bit('N', $type_bit{'u'});
-    define_type_bit('U', $type_bit{'u'});
     define_type_bit('e', $type_bit{'u'});
     define_type_bit('P', $type_bit{'u'});
     define_type_bit('Q', $type_bit{'u'});
@@ -222,6 +259,12 @@ $match_engine_ops{'TOP_fail'} = 1;
 	sanity("tag '$tag': primitive tags must be named with lowercase letters")
 	    unless $tag =~ /^[a-z]$/;
     }
+
+    foreach my $tag (keys %arg_size) {
+        defined $type_bit{$tag} or
+            sanity("the tag '$tag' has a size in %arg_size, " .
+                   "but has no defined bit pattern");
+    }
 }
 
 #
@@ -235,32 +278,75 @@ while (@ARGV && $ARGV[0] =~ /^-(.*)/) {
     ($target = \&compiler_output), next if /^compiler/;
     ($outdir = shift), next if /^outdir/;
     ($wordsize = shift), next if /^wordsize/;
+    ($code_model = shift), next if /^code-model/;
     ($verbose = 1), next if /^v/;
     ($defs{$1} = $2), next if /^D(\w+)=(\w+)/;
     die "$0: Bad option: -$_\n";
 }
 
+if ($wordsize == 32) {
+    $defs{'ARCH_32'} = 1;
+    $defs{'ARCH_64'} = 0;
+} elsif ($wordsize == 64) {
+    $defs{'ARCH_32'} = 0;
+    $defs{'ARCH_64'} = 1;
+    $code_pointers_are_short = $code_model eq 'small';
+}
+
 #
-# Initialize number of arguments per packed word.
+# Initialize pack options.
 #
 
-$args_per_word[2] = 2;
-$args_per_word[3] = 3;
-$args_per_word[4] = 2;
-$args_per_word[5] = 3;
-$args_per_word[6] = 3;
-
 if ($wordsize == 64) {
-    $pack_mask[3] = ['BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK', $WHOLE_WORD];
-    $args_per_word[4] = 4;
+    @basic_pack_options = (0,PACK_JUMP);
+    @extended_pack_options = @basic_pack_options;
+    if ($code_pointers_are_short) {
+        foreach (@basic_pack_options) {
+            push @extended_pack_options, $_ | PACK_IN_INSTR_WORD;
+        }
+    }
+}
+
+#
+# Add placeholders for built-in macros.
+#
+
+my %predef_macros =
+    (OPERAND_POSITION => ['Expr'],
+     IF => ['Expr','IfTrue','IfFalse'],
+     REFRESH_GEN_DEST => [],
+    );
+foreach my $name (keys %predef_macros) {
+    my @args = @{$predef_macros{$name}};
+    my $body = join(':', map { '$' . $_ } @args);
+    $c_code{$name} = [$body,"built-in macro",@args],
+    $c_code_used{$name} = 1;
 }
 
 #
 # Parse the input files.
 #
 
+my $in_c_code = '';
+my $c_code_block;
+my $c_code_loc;
+my @c_args;
+
 while (<>) {
     my($op_num);
+    if ($in_c_code) {
+        if (/^\}/) {
+            my $name = $in_c_code;
+            my $block = $c_code_block;
+            $in_c_code = '';
+            $block =~ s/^    //mg;
+            chomp $block;
+            $c_code{$name} = [$block,$c_code_loc,@c_args];
+        } else {
+            $c_code_block .= $_;
+        }
+        next;
+    }
     chomp;
     if (s/\\$//) {
 	$_ .= <>;
@@ -268,6 +354,7 @@ while (<>) {
     }
     next if /^\s*$/;
     next if /^\#/;
+    next if m@^//@;
 
     #
     # Handle %if.
@@ -310,36 +397,24 @@ while (<>) {
     #
     if (/^([\w_][\w\d_]+)=(.*)/) {
 	no strict 'refs';
-	my($name) = $1;
-	$$name = $2;
+	my $name = $1;
+        my $value = $2;
+        $value =~ s/;\s*$//;
+	$$name = $value;
 	next;
     }
 
     #
-    # Handle %hot/%cold.
+    # Handle %hot, %warm, and %cold.
     # 
     if (/^\%hot/) {
-	$hot = 1;
+	$hotness = HOT;
 	next;
+    } elsif (/^\%warm/) {
+	$hotness = WARM;
+        next;
     } elsif (/^\%cold/) {
-	$hot = 0;
-	next;
-    }
-    
-    #
-    # Handle macro definitions.
-    #
-    if (/^\%macro:(.*)/) {
-	my($op, $macro, @flags) = split(' ', $1);
-	defined($macro) and $macro =~ /^-/ and
-	    error("A macro must not start with a hyphen");
-	foreach (@flags) {
-	    /^-/ or error("Flags for macros should start with a hyphen");
-	}
-	error("Macro for '$op' is already defined")
-	    if defined $macro{$op};
-	$macro{$op} = $macro;
-	$macro_flags{$op} = join('', @flags);
+	$hotness = COLD;
 	next;
     }
 
@@ -352,6 +427,31 @@ while (<>) {
     }
 
     #
+    # Handle C code blocks.
+    #
+    if (/^(\w[\w.]*)\(([^\)]*)\)\s*{/) {
+        my $name = $1;
+        $in_c_code = $name;
+        $c_code_block = '';
+        @c_args = parse_c_args($2);
+        $c_code_loc = "$ARGV($.)";
+        if (defined $c_code{$name}) {
+            my $where = $c_code{$name}->[1];
+            error("$name: already defined at $where");
+        }
+        next;
+    }
+
+    #
+    # Handle definition of instructions in terms of
+    # micro instructions.
+    #
+    if (/^(\w+)\s*:=\s*([\w.]+)\s*;\s*$/) {
+        $combined_instrs{$1} = ["$ARGV($.)",$2];
+        next;
+    }
+
+    #
     # Parse off the number of the operation.
     #
     $op_num = undef;
@@ -394,15 +494,7 @@ while (<>) {
     # Parse specific instructions (only present in emulator/loader):
     #    Name Arg1 Arg2...
     #
-    my($name, @args) = split;
-    error("too many operands")
-	if @args > $max_spec_operands;
-    syntax_check($name, @args);
-    my $arity = @args;
-    if (defined $gen_opnum{$name,$arity} and $obsolete[$gen_opnum{$name,$arity}]) {
-	error("specific instructions may not be specified for obsolete instructions");
-    }
-    save_specific_ops($name, $arity, $hot, @args);
+    my($name,$arity) = parse_specific_op($_);
     if (defined $op_num) {
 	error("specific instructions must not be numbered");
     } elsif (!defined($gen_arity{$name}) && !defined($unnumbered{$name,$arity})) {
@@ -449,6 +541,18 @@ $num_file_opcodes = @gen_opname;
 &$target();
 
 #
+# Ensure that all C code implementations have been used.
+#
+{
+    my(@unused) = grep(!$c_code_used{$_}, keys %c_code);
+    foreach my $unused (@unused) {
+        my(undef,$where) = @{$c_code{$unused}};
+        warn "$where: $unused is unused\n";
+    }
+    die "\n" if @unused;
+}
+
+#
 # Produce output needed by the emulator/loader.
 #
 
@@ -458,6 +562,36 @@ sub emulator_output {
     my $key;			# Loop variable.
 
     #
+    # Generate code and meta information for all instructions.
+    #
+    foreach $key (keys %specific_op) {
+	foreach (@{$specific_op{$key}}) {
+	    my($name, $hotness, @args) = @$_;
+	    my $print_name = print_name($name, @args);
+
+	    my($size, $code, $pack_spec) = cg_basic(name => $name, args => \@args);
+	    if (defined $code) {
+                $code = "OpCase($print_name):\n$code";
+                push @generated_code, [$hotness,$code,($print_name)];
+	    }
+
+            # Note: Some of the information below will be modified
+            # for combined instructions.
+            my %info = ('size' => $size,
+                        'pack_spec' => $pack_spec,
+                        'adj' => 0,
+                        'args' => \@args);
+            $spec_op_info{$print_name} = \%info;
+	}
+    }
+
+    #
+    # Combine micro instruction into instruction blocks and generate
+    # code for them.
+    #
+    combine_micro_instructions();
+
+    #
     # Information about opcodes (beam_opcodes.c).
     #
     $name = "$outdir/beam_opcodes.c";
@@ -488,7 +622,7 @@ sub emulator_output {
     #
     # Generate code for specific ops.
     #
-    my($spec_opnum) = 0;
+    my $spec_opnum = 0;
     print "const OpEntry opc[] = {\n";
     foreach $key (sort keys %specific_op) {
 	$gen_to_spec{$key} = $spec_opnum;
@@ -503,40 +637,26 @@ sub emulator_output {
 	foreach (@{$specific_op{$key}}) {
 	    my($name, $hot, @args) = @{$_};
 	    my($sign) = join('', @args);
+            $sign =~ s/[?]//g;
 
 	    # The primitive types should sort before other types.
 
-	    my($sort_key) = $sign;
+	    my $sort_key = $sign;
 	    eval "\$sort_key =~ tr/$genop_types/./";
 	    $sort_key .= ":$sign";
-	    $items{$sort_key} = [$name, $hot, $sign, @args];
+            my $print_name = print_name($name, @args);
+	    $items{$sort_key} = $print_name;
 	}
 
 	#
 	# Now call the generator for the sorted result.
 	#
-	foreach (sort keys %items) {
-	    my($name, $hot, $sign, @args) = @{$items{$_}};
+	foreach my $sort_key (sort keys %items) {
+            my $print_name = $items{$sort_key};
+            my $info = $spec_op_info{$print_name};
+            my(@args) = @{$info->{'args'}};
+            @args = map { s/[?]$//; $_ } @args;
 	    my $arity = @args;
-	    my($instr) = "${name}_$sign";
-	    $instr =~ s/_$//;
-
-	    #
-	    # Call a generator to calculate size and generate macros
-	    # for the emulator.
-	    #
-	    my($size, $code, $pack) = basic_generator($name, $hot, @args);
-
-	    #
-	    # Save the generated $code for later.
-	    #
-	    if (defined $code) {
-		if ($hot) {
-		    push(@{$hot_code{$code}}, $instr);
-		} else {
-		    push(@{$cold_code{$code}}, $instr);
-		}
-	    }
 
 	    #
 	    # Calculate the bit mask which should be used to match this
@@ -558,7 +678,6 @@ sub emulator_output {
 	    }
 
 	    printf "/* %3d */  ", $spec_opnum;
-	    my $print_name = $sign ne '' ? "${name}_$sign" : $name;
 	    my $init = "{";
 	    my $sep = "";
 	    foreach (@bits) {
@@ -566,8 +685,12 @@ sub emulator_output {
 		$sep = ",";
 	    }
 	    $init .= "}";
-	    init_item($print_name, $init, $involves_r, $size, $pack, $sign);
-	    $op_to_name[$spec_opnum] = $instr;
+            my $adj = $info->{'adj'};
+            my $size = $info->{'size'};
+            my $pack_spec = $info->{'pack_spec'};
+            my $sign = join '', @args;
+	    init_item($print_name, $init, $involves_r, $size, $adj, $pack_spec, $sign);
+	    $op_to_name[$spec_opnum] = $print_name;
 	    $spec_opnum++;
 	}
     }
@@ -646,12 +769,19 @@ sub emulator_output {
 	print "#if !defined(ARCH_64)\n";
 	print qq[  #error "64-bit architecture assumed, but ARCH_64 not defined"\n];
 	print "#endif\n";
-	print "#define BEAM_WIDE_MASK 0xFFFFUL\n";
-	print "#define BEAM_LOOSE_MASK 0xFFFFUL\n";
-	print "#define BEAM_TIGHT_MASK 0xFFFFUL\n";
+        if ($code_pointers_are_short) {
+            print "#if !defined(CODE_MODEL_SMALL)\n";
+            print qq[  #error "small code model assumed, but CODE_MODEL_SMALL not defined"\n];
+            print "#endif\n";
+        }
+	print "#define BEAM_WIDE_MASK 0xFFFFFFFFull\n";
+	print "#define BEAM_LOOSE_MASK 0xFFFFull\n";
+	print "#define BEAM_TIGHT_MASK 0xFFFFull\n";
+	print "#define BEAM_TIGHTEST_MASK 0x3FFull\n";
 	print "#define BEAM_WIDE_SHIFT 32\n";
 	print "#define BEAM_LOOSE_SHIFT 16\n";
 	print "#define BEAM_TIGHT_SHIFT 16\n";
+	print "#define BEAM_TIGHTEST_SHIFT 10\n";
     }
     print "\n";
 
@@ -750,13 +880,24 @@ sub emulator_output {
     $name = "$outdir/beam_hot.h";
     open(STDOUT, ">$name") || die "Failed to open $name for writing: $!\n";
     comment('C');
-    print_code(\%hot_code);
+    print_code(HOT);
+
+    $name = "$outdir/beam_warm.h";
+    open(STDOUT, ">$name") || die "Failed to open $name for writing: $!\n";
+    comment('C');
+    print_code(WARM);
 
     $name = "$outdir/beam_cold.h";
     open(STDOUT, ">$name") || die "Failed to open $name for writing: $!\n";
     comment('C');
-    print_code(\%cold_code);
+    print_code(COLD);
+}
 
+sub print_name {
+    my($name,@args) = @_;
+    my $sign = join '', @args;
+    $sign =~ s/[?]//g;
+    $sign ne '' ? "${name}_$sign" : $name;
 }
 
 sub init_item {
@@ -784,29 +925,47 @@ sub q {
 }
 
 sub print_code {
-    my($ref) = @_;
-    my(%sorted);
-    my($key, $label);		# Loop variables.
-
-    foreach $key (keys %$ref) {
-	my($sort_key);
-	my($code) = '';
-	foreach $label (@{$ref->{$key}}) {
-	    $code .= "OpCase($label):\n";
-	    $sort_key = $label;
-	}
-	foreach (split("\n", $key)) {
-	    $code .= "    $_\n";
-	}
-	$code .= "\n";
-	$sorted{$sort_key} = $code;
+    my($include_hot) = @_;
+    my %sorted;
+
+    foreach my $ref (@generated_code) {
+        my($hot,$code,@labels) = @$ref;
+        next unless $hot == $include_hot;
+        my($sort_key) = @labels; # Use the first label as sort key.
+        $sorted{$sort_key} = $code;
     }
 
     foreach (sort keys %sorted) {
-	print $sorted{$_};
+	print_indented_code($sorted{$_});
     }
 }
 
+sub print_indented_code {
+    my(@code) = @_;
+
+    foreach my $chunk (@code) {
+        my $indent = 0;
+        foreach (split "\n", $chunk) {
+            s/^\s*//;
+            if (/\}/) {
+                $indent -= 2;
+            }
+            if ($_ eq '') {
+                print "\n";
+            } elsif (/^#/) {
+                print $_, "\n";
+            } else {
+                print ' ' x $indent, $_, "\n";
+            }
+            if (/\{/) {
+                $indent += 2;
+            }
+        }
+        print "\n";
+    }
+}
+
+
 #
 # Produce output needed by the compiler back-end (assembler).
 #
@@ -857,40 +1016,82 @@ sub compiler_output {
 }
 
 #
-# Check an operation for validity.
+# Parse and store a specific operation.
 #
-sub syntax_check {
-    my($name, @args) = @_;
-    my($i);
+sub parse_specific_op {
+    my($name, @args) = split " ", shift;
+    my $arity = @args;
 
+    # Check for various errors.
     error("Bad opcode name '$name'")
 	unless $name =~ /^[a-z][\w\d_]*$/;
-    for ($i = 0; $i < @args; $i++) {
-        foreach my $type (split(//, $args[$i])) {
+    error("too many operands")
+	if @args > $max_spec_operands;
+    for (my $i = 0; $i < $arity; $i++) {
+        my $arg = $args[$i];
+        $arg =~ s/[?]$//;
+        foreach my $type (split(//, $arg)) {
             error("Argument " . ($i+1) . ": invalid type '$type'")
                 unless defined $arg_size{$type};
         }
     }
-}
-
-sub save_specific_ops {
-    my($name,$arity,$hot,@args) = @_;
-    my(@res) = ("");
+    if (defined $gen_opnum{$name,$arity} and $obsolete[$gen_opnum{$name,$arity}]) {
+	error("specific instructions may not be specified for obsolete instructions");
+    }
 
+    # Expand operands with multiple types to multiple instructions.
+    # (For example, "get_list xy xy xy" will be expanded to six instructions.)
+    my @res = ([]);
     foreach my $arg (@args) {
-        my @new_res = ();
+        my @old_res = @res;
+        @res = ();
+        my $marker = ($arg =~ s/[?]$//) ? '?' : '';
         foreach my $type (split(//, $arg)) {
-            foreach my $args (@res) {
-                push @new_res, "$args$type";
+            foreach my $args_ref (@old_res) {
+                my @args = @$args_ref;
+                push @args, "$type$marker";
+                push @res, \@args;
             }
         }
-        @res = @new_res;
     }
+
+    # Store each specific instruction.
     my $key = "$name/$arity";
-    foreach my $args (@res) {
-        @args = split //, $args;
-        push @{$specific_op{$key}}, [$name,$hot,@args];
+    foreach my $args_ref (@res) {
+        @args = @$args_ref;
+        my $arity = @args;
+        my $loc = "$ARGV($.)";
+        if (defined $specific_op_arity{$name}) {
+            my($prev_arity,$loc) = @{$specific_op_arity{$name}};
+            if ($arity != $prev_arity) {
+                error("$name defined with arity $arity, " .
+                      "but previously defined with arity $prev_arity at $loc");
+            }
+        }
+        $specific_op_arity{$name} = [$arity,$loc];
+        my $print_name = print_name($name, @args);
+        if (defined $print_name{$print_name}) {
+            error("$name @args: already defined at " .
+                  $print_name{$print_name});
+        }
+        $print_name{$print_name} = $loc;
+        push @{$specific_op{$key}}, [$name,$hotness,@args];
+    }
+
+    # Done.
+    ($name,$arity);
+}
+
+sub parse_c_args {
+    local($_) = @_;
+    my @res;
+
+    while (s/^(\w[\w\d]*)\s*//) {
+        push @res, $1;
+        s/^,\s*// or last;
     }
+    $_ eq '' or error("garbage in argument list: $_");
+    @res;
 }
 
 sub error {
@@ -934,58 +1135,314 @@ sub comment {
 }
 
 #
-# Basic implementation of instruction in emulator loop
-# (assuming no packing).
+# Combine micro instruction into instruction blocks.
 #
+sub combine_micro_instructions {
+    my %groups;
+
+    # Sanity check, normalize micro instructions.
+    foreach my $instr (keys %combined_instrs) {
+        my $ref = $combined_instrs{$instr};
+        my($def_loc,$def) = @$ref;
+        my($group,@subs) = split /[.]/, $def;
+        my $arity = 0;
+        @subs = map { "$group.$_" } @subs;
+        foreach my $s (@subs) {
+            my $code = $c_code{$s};
+            defined $code or
+                error("$def_loc: no definition of $s");
+            $c_code_used{$s} = 1;
+            my(undef,undef,@c_args) = @{$code};
+            $arity += scalar(@c_args);
+        }
+        push @{$groups{$group}}, [$instr,$arity,@subs];
+    }
 
-sub basic_generator {
-    my($name, $hot, @args) = @_;
-    my($size) = 0;
-    my($macro) = '';
-    my($flags) = '';
-    my(@f);
-    my(@f_types);
-    my($fail_type);
-    my($prefix) = '';
-    my($tmp_arg_num) = 1;
-    my($pack_spec) = '';
-    my($var_decls) = '';
-    my($i);
-    my($no_prefetch) = 0;
+    # Now generate code for each group.
+    foreach my $group (sort keys %groups) {
+        my($hotness,$code,@labels) =
+            combine_instruction_group($group, @{$groups{$group}});
+        push @generated_code, [$hotness,$code,@labels];
+    }
+}
 
-    # The following argument types should be included as macro arguments.
-    my(%incl_arg) = ('c' => 1,
-		     'i' => 1,
-		     'a' => 1,
-		     'A' => 1,
-		     'N' => 1,
-		     'U' => 1,
-		     'I' => 1,
-		     't' => 1,
-		     'P' => 1,
-		     'Q' => 1,
-		     );
+sub combine_instruction_group {
+    my($group,@in_instrs) = @_;
+    my $gcode = '';             # Code for the entire group.
+    my $group_hotness = COLD;
 
-    # Pick up the macro to use and its flags (if any).
+    # Get code for the head of the group (if any).
+    my $head_name = "$group.head";
+    $c_code_used{$head_name} = 1;
+    my $head_code_ref = $c_code{$head_name};
+    if (defined $head_code_ref) {
+        my($head_code,$where,@c_args) = @{$head_code_ref};
+        @c_args and error("$where: no arguments allowed for " .
+                          "head function '$head_name()'");
+        $gcode = $head_code . "\n";
+    }
 
-    $macro = $macro{$name} if defined $macro{$name};
-    $flags = $macro_flags{$name} if defined $macro_flags{$name};
+    # Variables.
+    my %offsets;
+    my @instrs;
+    my %num_references;         # Number of references from other sub instructions.
+    my $group_size = 999;
 
     #
-    # Add any arguments to be included as macro arguments (for instance,
-    # 'p' is usually not an argument, except for calls).
+    # Calculate the number of references from other sub instructions.
+    # This number is useful in several ways:
+    #
+    # * If this number is 0, it is only used as the entry point for a
+    #   function, implying that it does not need a label and that operands
+    #   can be packed into the instruction word.
+    #
+    # * We'll use this number in the sort key, as a tie breaker for sub instructions
+    #   at the same instruction offset.
     #
+    foreach my $ref_instr (@in_instrs) {
+        my(undef,undef,$first_sub,@other_subs) = @$ref_instr;
+        $num_references{$first_sub} += 0; # Make sure it is defined.
+        foreach my $sub (@other_subs) {
+            $num_references{$sub}++;
+        }
+    }
 
-    while ($flags =~ /-arg_(\w)/g) {
-	$incl_arg{$1} = 1;
-    };
+    # Do basic error checking. Associate operands of instructions
+    # with the correct micro instructions. Calculate offsets for micro
+    # instructions.
+    foreach my $ref_instr (@in_instrs) {
+        my($specific,$arity,@subs) = @$ref_instr;
+        my $specific_key = "$specific/$arity";
+        my $specific_op_ref = $specific_op{$specific_key};
+        error("no $specific_key instruction")
+            unless defined $specific_op_ref;
+        foreach my $specific_op (@$specific_op_ref) {
+            my($name, $hotness, @args) = @{$specific_op};
+            $group_hotness = $hotness unless $group_hotness >= $hotness;
+            my $offset = 0;
+            my @rest = @args;
+            my @new_subs;
+            my $print_name = print_name($specific, @args);
+            my $opcase = $print_name;
+            my $last = $subs[$#subs];
+            foreach my $s (@subs) {
+                my $code = $c_code{$s};
+                my(undef,undef,@c_args) = @{$code};
+                my @first;
+                foreach (0..$#c_args) {
+                    push @first, shift @rest;
+                }
+                my $size = cg_combined_size(name => $s,
+                                            first => $num_references{$s} == 0,
+                                            args => \@first);
+                $offsets{$s} = $offset
+                    unless defined $offsets{$s} and $offsets{$s} < $offset;
+                $offset += $size - 1;
+                my $label = micro_label($s);
+                push @new_subs, [$opcase,$label,$s,$size-1,@first];
+                $opcase = '';
+            }
+            $spec_op_info{$print_name}->{'size'} = $offset + 1;
+            $group_size = $offset if $group_size >= $offset;
+            push @instrs, [$specific_key,@new_subs];
+        }
+    }
+
+    # Link the sub instructions for each instructions to each
+    # other.
+    my @all_instrs;
+    foreach my $instr (@instrs) {
+        my($specific_key,@subs) = @{$instr};
+        for (my $i = 0; $i < @subs; $i++) {
+            my($opcase,$label,$s,$size,@args) = @{$subs[$i]};
+            my $next = '';
+            (undef,$next) = @{$subs[$i+1]} if $i < $#subs;
+            my $instr_info = "$opcase:$label:$next:$s:$size:@args";
+            push @all_instrs, [$label,$s,$offsets{$s},$instr_info];
+        }
+    }
+
+    my %order_to_instrs;
+    my %label_to_offset;
+    my %order_to_offset;
+    foreach my $instr (@all_instrs) {
+        my($label,$s,$offset,$instr_info) = @$instr;
+        my $sort_key = sprintf("%02d.%02d", $offset, $num_references{$s});
+        push @{$order_to_instrs{$sort_key}}, $instr_info;
+        $label_to_offset{$label} = $offset;
+        $order_to_offset{$sort_key} = $offset;
+    }
+
+    my(@slots) = sort {$a <=> $b} keys %order_to_instrs;
+
+    # Now generate the code for the entire group.
+    my $offset = 0;
+    my @opcase_labels;
+    my %down;
+    my %up;
+    for(my $i = 0; $i < @slots; $i++) {
+        my $key = $slots[$i];
+
+        # Sort micro-instructions with OpCase before other micro-instructions.
+        my(@instrs) = @{$order_to_instrs{$key}};
+        my $order_func = sub {
+            my $a_key = ($a =~ /^:/) ? "1$a" : "0$a";
+            my $b_key = ($b =~ /^:/) ? "1$b" : "0$b";
+            $a_key cmp $b_key;
+        };
+        @instrs = sort $order_func @instrs;
+
+        my %seen;
+        foreach my $instr (@instrs) {
+            my($opcase,$label,$next,$s,$size,$args) = split ":", $instr;
+            my(@first) = split " ", $args;
+
+            my $seen_key = "$label:$next:" . scalar(@first);
+            next if $opcase eq '' and $seen{$seen_key};
+            $seen{$seen_key} = 1;
+            $seen_key .= $opcase;
+
+            if ($opcase ne '') {
+                $gcode .= "OpCase($opcase):\n";
+                push @opcase_labels, $opcase;
+            }
+            if ($num_references{$s}) {
+                $gcode .= "$label:\n";
+            }
+
+            my $flags = '';
+            my $transfer_to_next = '';
+            my $inc = 0;
+
+            unless ($i == $#slots) {
+                $flags = "-no_next";
+                my $next_offset = $label_to_offset{$next};
+                $inc = ($offset + $size) - $next_offset;
+                $transfer_to_next = "I += $inc;\n" if $inc;
+                $transfer_to_next .= "goto $next;\n\n";
+            }
+
+            my($gen_code,$down,$up) =
+                cg_combined_code(name => $s,
+                                 first => $num_references{$s} == 0,
+                                 extra_comments => $flags,
+                                 offset => $offset,
+                                 comp_size => $group_size-$offset,
+                                 inc => $inc,
+                                 args =>\@first);
+            my $spec_label = "$opcase$label";
+            $down{$spec_label} = $down;
+            $up{$spec_label} = $up;
+            $gcode .= $gen_code . $transfer_to_next;
+        }
+        $offset = $order_to_offset{$slots[$i+1]} if $i < $#slots;
+    }
+
+    foreach my $print_name (@opcase_labels) {
+        my $info = $spec_op_info{$print_name};
+        $info->{'adj'} = $info->{'size'} - $group_size - 1;
+    }
 
     #
-    # Pack arguments if requested.
+    # Assemble pack specifications for all instructions in the group.
     #
+    foreach my $instr (@instrs) {
+        my(undef,@subs) = @{$instr};
+        my $down = '';
+        my $up = '';
+        for (my $i = 0; $i < @subs; $i++) {
+            my($opcase,$label) = @{$subs[$i]};
+            my $spec_label = "$opcase$label";
+            if (defined $down{$spec_label}) {
+                $down = $down{$spec_label} . $down;
+                $up = $up . $up{$spec_label};
+            }
+        }
+        my $print_name = $subs[0]->[0];
+        my $info = $spec_op_info{$print_name};
+        $info->{'pack_spec'} = build_pack_spec("$down:$up");
+    }
+
+    ($group_hotness,"{\n$gcode\n}\n\n",@opcase_labels);
+}
+
+sub micro_label {
+    my $label = shift;
+    $label =~ s/[.]/__/g;
+    $label;
+}
 
-    if ($flags =~ /-pack/ && $hot) {
-        ($prefix, $pack_spec, @args) = do_pack(@args);
+
+#
+# Basic code generation for one instruction.
+#
+
+sub cg_basic {
+    my %params = (@_, pack_options => \@extended_pack_options);
+    my($size,$code,$pack_spec) = code_gen(%params);
+    $pack_spec = build_pack_spec($pack_spec);
+    ($size,$code,$pack_spec);
+}
+
+#
+# Calculate size for a micro instruction.
+#
+
+sub cg_combined_size {
+    my %params = (@_,
+                  pack_options => \@basic_pack_options,
+                  size_only => 1);
+    $params{pack_options} = \@extended_pack_options
+        if $params{first};
+    my($size) = code_gen(%params);
+    $size;
+}
+
+#
+# Generate code for a micro instruction.
+#
+
+sub cg_combined_code {
+    my %params = (@_, pack_options => \@basic_pack_options);
+    $params{pack_options} = \@extended_pack_options
+        if $params{first};
+    my($size,$code,$pack_spec) = code_gen(%params);
+    if ($pack_spec eq '') {
+        ($code,'','');
+    } else {
+        my($down,$up) = split /:/, $pack_spec;
+        ($code,$down,$up);
+    }
+}
+
+sub code_gen {
+    my %params = (extra_comments => '',
+                  offset => 0,
+                  inc => 0,
+                  size_only => 0,
+                  @_);
+    my $name = $params{name};
+    my $extra_comments = $params{extra_comments};
+    my $offset = $params{offset};
+    my $inc = $params{inc};
+    my @args = @{$params{args}};
+
+    my $size = 0;
+    my $flags = '';
+    my @f;
+    my $prefix = '';
+    my $tmp_arg_num = 1;
+    my $pack_spec = '';
+    my $var_decls = '';
+
+    #
+    # Pack arguments for hot code with an implementation.
+    #
+
+    my $c_code_ref = $c_code{$name};
+    if (defined $c_code_ref and $name ne 'catch') {
+        my $pack_options = $params{pack_options};
+        ($var_decls, $pack_spec, @args) = do_pack($name, $offset, $pack_options, @args);
     }
 
     #
@@ -993,259 +1450,696 @@ sub basic_generator {
     # the macro.
     #
 
+    my $need_block = 0;
+    my $arg_offset = $offset;
+    my $has_gen_dest = 0;
+    @args = map { s/[?]$//g; $_ } @args;
     foreach (@args) {
 	my($this_size) = $arg_size{$_};
       SWITCH:
 	{
-	    /^pack:(\d):(.*)/ and do { push(@f, $2);
-				       push(@f_types, 'packed');
-				       $this_size = $1;
-				       last SWITCH;
-				   };
-	    /r/    and do { push(@f, "r(0)"); push(@f_types, $_); last SWITCH };
-	    /[xy]/ and do { push(@f, "$_" . "b(Arg($size))");
-			     push(@f_types, $_);
-			     last SWITCH;
-			};
-	    /n/    and do { push(@f, "NIL"); push(@f_types, $_); last SWITCH };
-	    /s/    and do { my($tmp) = "targ$tmp_arg_num";
-			    $var_decls .= "Eterm $tmp; ";
-			    $tmp_arg_num++;
-			    push(@f, $tmp);
-			    push(@f_types, $_);
-			    $prefix .= "GetR($size, $tmp);\n";
-			    last SWITCH; };
-	    /d/    and do { $var_decls .= "Eterm dst; Eterm* dst_ptr; ";
-			    push(@f, "*dst_ptr");
-			    push(@f_types, $_);
-			    $prefix .= "dst = Arg($size);\n";
-			    $prefix .= "dst_ptr = REG_TARGET_PTR(dst);\n";
-			    last SWITCH;
-			};
-	    defined($incl_arg{$_})
-		and do { push(@f, "Arg($size)");
-			 push(@f_types, $_);
-			 last SWITCH;
-		     };
-
-	    /[fp]/ and do { $fail_type = $_; last SWITCH };
-
-	    /[eLIFEbASjPowlq]/ and do { last SWITCH; };
+	    /^packed:d:(\d):(.*)/ and do {
+                $var_decls .= "Eterm dst = $2;\n" .
+                    "Eterm* dst_ptr = REG_TARGET_PTR(dst);\n";
+                push(@f, "*dst_ptr");
+                $this_size = $1;
+                $has_gen_dest = 1;
+                last SWITCH;
+            };
+	    /^packed:[a-zA-z]:(\d):(.*)/ and do {
+                push(@f, $2);
+                $this_size = $1;
+                last SWITCH;
+            };
+	    /r/ and do {
+                push(@f, "r(0)");
+                last SWITCH;
+            };
+	    /[lxyS]/ and do {
+                push(@f, $_ . "b(" . arg_offset($arg_offset) . ")");
+                last SWITCH;
+            };
+	    /n/ and do {
+                push(@f, "NIL");
+                last SWITCH;
+            };
+	    /s/ and do {
+                my($tmp) = "targ$tmp_arg_num";
+                $var_decls .= "Eterm $tmp;\n";
+                $tmp_arg_num++;
+                push(@f, $tmp);
+                $prefix .= "GetR($arg_offset, $tmp);\n";
+                $need_block = 1;
+                last SWITCH;
+            };
+	    /d/ and do {
+                $var_decls .= "Eterm dst = " . arg_offset($arg_offset) . ";\n" .
+                    "Eterm* dst_ptr = REG_TARGET_PTR(dst);\n";
+                push(@f, "*dst_ptr");
+                $has_gen_dest = 1;
+                last SWITCH;
+            };
+            defined $arg_size{$_} and do {
+                push @f, arg_offset($arg_offset);
+                last SWITCH;
+            };
 
 	    die "$name: The generator can't handle $_, at";
 	}
 	$size += $this_size;
+        $arg_offset += $this_size;
     }
 
     #
-    # Add a fail action macro if requested.
+    # If the implementation is in beam_emu.c or if
+    # the caller only wants the size, we are done.
     #
+    if (not defined $c_code_ref or $params{size_only}) {
+        return ($size+1, undef, '');
+    }
 
-    $flags =~ /-fail_action/ and do {
-	$no_prefetch = 1;
-	if (!defined $fail_type) {
-	    my($i);
-	    for ($i = 0; $i < @f_types; $i++) {
-		local($_) = $f_types[$i];
-		/[rxycians]/ and do { push(@f, "Badmatch($f[$i])"); next };
-	    }
-	} elsif ($fail_type eq 'f') {
-	    push(@f, "ClauseFail()");
-	} else {
-	    my($i);
-	    for ($i = 0; $i < @f_types; $i++) {
-		local($_) = $f_types[$i];
-		/[rxycians]/ and do { push(@f, "Badmatch($f[$i])"); next };
-	    }
-	}
-    };
+    my $group_size = ($params{comp_size} || $size) + $inc;
 
     #
-    # Add a size argument if requested.
+    # Generate main body of the implementation.
     #
+    my($c_code,$where,@c_args) = @{$c_code_ref};
+    my %bindings;
+    $c_code_used{$name} = 1;
 
-    $flags =~ /-size/ and do {
-	push(@f, $size);
-    };
+    if (@f != @c_args) {
+        error("$where: defining '$name' with ", scalar(@c_args),
+              " arguments instead of expected ", scalar(@f), " arguments");
+    }
 
-    # Generate the macro if requested.
-    my($code);
-    if (defined $macro{$name}) {
-	my($macro_code) = "$prefix$macro(" . join(', ', @f) . ");";
-	$var_decls .= "BeamInstr tmp_packed1;"
-	    if $macro_code =~ /tmp_packed1/;
-	$var_decls .= "BeamInstr tmp_packed2;"
-	    if $macro_code =~ /tmp_packed2/;
-	if ($flags =~ /-nonext/) {
-	    $code = join("\n",
-			 "{ $var_decls",
-			 $macro_code,
-			 "}");
-	} elsif ($flags =~ /-goto:(\S*)/) {
-	    my $goto = $1;
-	    $code = join("\n",
-			 "{ $var_decls",
-			 $macro_code,
-			 "I += $size + 1;",
-			 "goto $goto;",
-			 "}");
-	} elsif ($no_prefetch) {
-	    $code = join("\n",
-			 "{ $var_decls",
-			 $macro_code,
-			 "Next($size);",
-			 "}", "");
-	} else {
-	    $code = join("\n",
-			 "{ $var_decls",
-			 "BeamInstr* next;",
-			 "PreFetch($size, next);",
-			 "$macro_code",
-			 "NextPF($size, next);",
-			 "}", "");
-	}
+    for (my $i = 0; $i < @f; $i++) {
+        my $var = $c_args[$i];
+        $bindings{$var} = $f[$i];
+    }
+    $bindings{'NEXT_INSTRUCTION'} = "I+" . ($group_size+$offset+1);
+    $bindings{'IP_ADJUSTMENT'} = $inc;
+    $c_code = eval { expand_all($c_code, \%bindings) };
+    unless (defined $c_code) {
+        warn $@;
+        error("... from the body of $name at $where");
+    }
+    my(@comments) = $c_code =~ m@//[|]\s*(.*)@g;
+    $c_code =~ s@//[|]\s*(.*)\n?@@g;
+    $flags = "@comments $extra_comments";
+
+    #
+    # Generate code for transferring to the next instruction.
+    #
+    my $dispatch_next;
+    my $instr_offset = $group_size + $offset + 1;
+
+    if ($flags =~ /-no_next/) {
+        $dispatch_next = "";
+    } elsif ($flags =~ /-no_prefetch/) {
+        $dispatch_next = "\nI += $instr_offset;\n" .
+            "ASSERT(VALID_INSTR(*I));\n" .
+            "Goto(*I);";
+    } else {
+        $var_decls .= "BeamInstr next_pf = BeamCodeAddr(I[$instr_offset]);\n";
+        $dispatch_next = "\nI += $instr_offset;\n" .
+            "ASSERT(VALID_INSTR(next_pf));\n" .
+            "GotoPF(next_pf);";
+    }
+
+    #
+    # Assemble the complete code for the instruction.
+    #
+    my $body = "$c_code$dispatch_next";
+    if ($need_block) {
+        $body = "$prefix\{\n$body\n}";
+    } else {
+        $body = "$prefix$body";
+    }
+    my $code = join("\n",
+                    "{",
+                    "$var_decls$body",
+                    "}", "");
+
+    # Make sure that $REFRESH_GEN_DEST() is used when a
+    # general destination ('d') may have been clobbered by
+    # a GC.
+    my $gc_error = verify_gc_code($code, $has_gen_dest);
+    if (defined $gc_error) {
+        warn $gc_error;
+        error("... from the body of $name at $where");
+    }
+
+    # Done.
+    ($size+1, $code, $pack_spec);
+}
+
+sub verify_gc_code {
+    my $code = shift;
+    my $has_gen_dest = shift;
+
+    return unless $has_gen_dest;
+
+    if ($code =~ /$GC_REGEXP/o) {
+        my $code_after_gc = substr($code, $+[0]);
+        unless ($code_after_gc =~ /dst_ptr = REG_TARGET_PTR/) {
+            return "pointer to destination register is invalid after GC -- " .
+                "use \$REFRESH_GEN_DEST()\n";
+        }
+    }
+    return undef;
+}
+
+sub arg_offset {
+    my $offset = shift;
+    "I[" . ($offset+1) . "]";
+}
+
+sub expand_all {
+    my($code,$bindings_ref) = @_;
+    my %bindings = %{$bindings_ref};
+
+    # Expand all $Var occurrences.
+    $code =~ s/[\$](\w[\w\d]*)(?!\()/defined $bindings{$1} ? $bindings{$1} : "\$$1"/ge;
+
+    # Find calls to macros, $name(...), and expand them.
+    my $res = "";
+    while ($code =~ /[\$](\w[\w\d]*)\(/) {
+        my $macro_name = $1;
+        my $keep = substr($code, 0, $-[0]);
+        my $after = substr($code, $+[0]);
+
+        my $body;
+        ($body,$code) = expand_macro($macro_name, $after, \%bindings);
+        $res .= "$keep$body";
+    }
+
+    $res . $code;
+}
+
+sub expand_macro {
+    my($name,$rest,$bindings_ref) = @_;
+
+    my $c_code = $c_code{$name};
+    defined $c_code or
+        error("calling undefined macro '$name'...");
+    $c_code_used{$name} = 1;
+    my ($body,$where,@vars) = @{$c_code};
+
+    # Separate the arguments into @args;
+    my @args;
+    my $level = 1;
+    my %inc = ('(' => 1, ')' => -1,
+               '[' => 1, ']' => -1,
+               '{' => 1, '}' => -1);
+    my $arg = undef;
+    while ($rest =~ /([,\(\[\{\}\]\)]|([^,\(\[\{\}\]\)]*))/g) {
+        my $token = $1;
+        my $inc = $inc{$token} || 0;
+        $level += $inc;
+        if ($level == 0) {
+            $rest = substr($rest, pos($rest));
+            push @args, $arg if defined $arg;
+            last;
+        }
+        if ($token eq ',') {
+            if ($level == 1) {
+                push @args, $arg;
+                $arg = "";
+            }
+            next;
+        }
+        $arg .= $token;
+    }
+
+    # Trim leading whitespace from each argument.
+    foreach my $arg (@args) {
+        $arg =~ s/^\s*//;
+    }
+
+    # Make sure that the number of arguments are correct.
+    if (@vars != @args) {
+        error("calling $name with ", scalar(@args),
+              " arguments instead of expected ", scalar(@vars), " arguments...");
+    }
+
+    # Now combine bindings from the parameter names and arguments.
+    my %bindings = %{$bindings_ref};
+    my %new_bindings;
+
+    # Keep the special, pre-defined bindings.
+    foreach my $key (qw(NEXT_INSTRUCTION IP_ADJUSTMENT)) {
+        $new_bindings{$key} = $bindings{$key};
+    }
+
+    for (my $i = 0; $i < @vars; $i++) {
+        my $arg = $args[$i];
+        $arg = eval { expand_all($arg, \%bindings) };
+        unless (defined $arg) {
+            warn $@;
+            die "... from the body of $name at $where\n";
+        }
+        $new_bindings{$vars[$i]} = $arg;
+    }
+
+    $body = eval { expand_all($body, \%new_bindings) };
+    unless (defined $body) {
+        warn $@;
+        die "... from the body of $name at $where\n";
+    }
+
+    # Handle built-in macros.
+    if ($name eq 'OPERAND_POSITION') {
+        if ($body =~ /^I\[(\d+)\]$/) {
+            $body = $1;
+        } else {
+            $body = 0;
+        }
+    } elsif ($name eq 'IF') {
+        my $expr = $new_bindings{Expr};
+        my $bool = eval $expr;
+        if ($@ ne '') {
+            &error("bad expression '$expr' in \$IF()");
+        }
+        my $part = $bool ? 'IfTrue' : 'IfFalse';
+        $body = $new_bindings{$part};
+    } elsif ($name eq 'REFRESH_GEN_DEST') {
+        $body = "dst_ptr = REG_TARGET_PTR(dst)";
     }
 
-    # Return the size and code for the macro (if any).
-    $size++;
-    ($size, $code, $pack_spec);
+
+    # Wrap body if needed and return result.
+    $body = "do {\n$body\n} while (0)"
+        if needs_do_wrapper($body);
+    ($body,$rest);
+}
+
+# Conservative heuristic to determine whether a do { ... } while(0)
+# wrapper is needed.
+sub needs_do_wrapper {
+    local $_ = shift;
+
+    s@^//[|][^\n]*\n@@;
+    s@^\s*@@s;
+    s@^/[*].*[*]/\s*@@s;
+    return 1 if /^(Eterm|Uint|Sint|int|unsigned)/; # Definitely needed.
+    return 0 if /^do/;
+    return 0 if /^SET_I/;
+    return 0 if /^SET_CP/;
+    return 0 if /^ERTS_NO_FPE_CHECK_INIT/;
+    return 0 if /^ASSERT/;
+    return 0 if /^DTRACE/;
+    return 0 if /^[A-Za-z_]*\s*=/;
+    return 0 if /^c_p->/;
+    return 0 if /^[A-Z_]*SWAPOUT/;
+    return 0 if /^if\s*[(]/;
+    return 0 if /^goto\b/;
+    return 0 if /^\d+/;
+    return 1;                   # Not sure, say that it is needed.
 }
 
 sub do_pack {
-    my(@args) = @_;
+    my($name,$offset,$pack_opts_ref,@args) = @_;
+    my @pack_opts = @$pack_opts_ref;
+    my $opt_arg_pos = -1;
+
+    # Look for an optional use operand not as the first argument.
+    if (@args and $args[0] !~ /[?]$/) {
+        for (my $pos = 0; $pos < @args; $pos++) {
+            if ($args[$pos] =~ /[?]$/) {
+                $opt_arg_pos = $pos;
+                last;
+            }
+        }
+    }
+
+    @args = map { s/[?]$//; $_ } @args; # Remove any optional use marker.
+
+    # If there is an optional operand, extend the array of pack options.
+    if ($opt_arg_pos >= 0) {
+        my @new_pack_opts = grep { $_ & PACK_IN_INSTR_WORD } @pack_opts;
+        @new_pack_opts = map {
+            ($_ & ~ PACK_IN_INSTR_WORD) | PACK_OPT_IN_INSTR_WORD;
+        } @new_pack_opts;
+        push @pack_opts, @new_pack_opts;
+    }
+
+    my $ret = ['', ':', @args];
+    my $score = 0;
+
+    foreach my $options (@pack_opts) {
+        my $this_opt_arg_pos = ($options & PACK_OPT_IN_INSTR_WORD) ? $opt_arg_pos : -1;
+        my($this_score,$this_result) =
+            do_pack_one($name, $options, $this_opt_arg_pos, $offset, @args);
+        if ($this_score > $score) {
+            $ret = $this_result;
+            $score = $this_score;
+        }
+    }
+    return @$ret;
+}
+
+sub do_pack_one {
+    my($name,$options,$opt_arg_pos,$offset,@args) = @_;
     my($packable_args) = 0;
-    my @is_packable;		# Packability (boolean) for each argument.
-    my $wide_packing = 0;
-    my(@orig_args) = @args;
+    my @bits_needed;            # Bits needed for each argument.
+    my $pack_in_iw = $options & PACK_IN_INSTR_WORD;
+
+    #
+    # Define the minimum number of bits needed for the packable argument types.
+    #
+    my %bits_needed = ('x' => 10,
+                       'y' => 10,
+                       'Q' => 10,
+                       'l' => 10,
+                       'S' => 16,
+                       'd' => 16,
+                       't' => 16);
+    if ($wordsize == 64) {
+        $bits_needed{'I'} = 32;
+        if ($options & PACK_JUMP) {
+            $bits_needed{'f'} = 32;
+            $bits_needed{'j'} = 32;
+        }
+    }
 
     #
-    # Count the number of packable arguments.  If we encounter any 's' or 'd'
-    # arguments, packing is not possible.
+    # Count the number of packable arguments.
     #
-    my $packable_types = "xytQ";
     foreach my $arg (@args) {
-	if ($arg =~ /^[$packable_types]/) {
+        if (defined $bits_needed{$arg}) {
 	    $packable_args++;
-	    push @is_packable, 1;
-	} elsif ($arg =~ /^I/ and $wordsize == 64 and $packable_args < 2) {
-	    $wide_packing = 1;
-	    push @is_packable, 1;
-	    if (++$packable_args == 2) {
-		# We can only pack two arguments. Turn off packing
-		# for the rest of the arguments.
-		$packable_types = "\xFF";
-	    }
-	} elsif ($arg =~ /^[sd]/) {
-	    return ('', '', @args);
-	} elsif ($arg =~ /^[scq]/ and $packable_args > 0) {
-	    # When packing, this operand will be picked up from the
-	    # code array, put onto the packing stack, and later put
-	    # back into a different location in the code. The problem
-	    # is that if this operand is a literal, the original
-	    # location in the code would have been remembered in a
-	    # literal patch.  For packing to work, we would have to
-	    # adjust the position in the literal patch. For the
-	    # moment, adding additional instructions to the packing
-	    # engine to handle this does not seem worth it, so we will
-	    # just turn off packing.
-	    return ('', '', @args);
+            push @bits_needed, $bits_needed{$arg};
 	} else {
-	    push @is_packable, 0;
+	    push @bits_needed, 0;
 	}
+        if ($arg =~ /^[fj]$/) {
+            # Only pack the first occurrence of 'f' or 'j'.
+            delete $bits_needed{'f'};
+            delete $bits_needed{'j'};
+        }
     }
 
     #
-    # Get out of here if too few or too many arguments.
+    # Return if there is nothing to pack.
     #
-    return ('', '', @args) if $packable_args < 2;
+    if ($packable_args == 0) {
+        return (-1);
+    } elsif ($packable_args == 1 and $options == 0) {
+        return (-1);
+    }
 
-    my($size) = 0;
-    my($pack_prefix) = '';
-    my($down) = '';		# Pack commands (towards instruction
-				# beginning).
-    my($up) = '';		# Pack commands (storing back while
-				# moving forward).
+    #
+    # Determine how many arguments we should pack into each word.
+    #
+    my @args_per_word;
+    my @need_wide_mask;
+    my $bits;
+    my $this_wordsize;
+    my $word = -1;
+
+    my $next_word = sub {
+        $word++;
+        $args_per_word[$word] = 0;
+        $need_wide_mask[$word] = 0;
+        $bits = 0;
+        $this_wordsize = $wordsize;
+    };
 
-    my $args_per_word = $args_per_word[$packable_args];
-    my @shift;
-    my @mask;
-    my @instr;
+    $next_word->();
+    $this_wordsize = 32 if $pack_in_iw;
+    for (my $arg_num = 0; $arg_num < @args; $arg_num++) {
+            my $needed = $bits_needed[$arg_num];
 
-    if ($wide_packing) {
-	@shift = ('0', 'BEAM_WIDE_SHIFT');
-	@mask = ('BEAM_WIDE_MASK', $WHOLE_WORD);
-	@instr = ('w', 'i');
-    } else {
-	@shift = @{$pack_shift[$args_per_word]};
-	@mask = @{$pack_mask[$args_per_word]};
-	@instr = @{$pack_instr[$args_per_word]};
+            next unless $needed;
+            next if $arg_num == $opt_arg_pos;
+
+            if ($bits+$needed > $this_wordsize) { # Does not fit.
+                $next_word->();
+            }
+            if ($args_per_word[$word] == 4) { # Can't handle more than 4 args.
+                $next_word->();
+            }
+            if ($needed == 32 and $args_per_word[$word] > 1) {
+                # Must only pack two arguments in this word, and there
+                # are already at least two arguments here.
+                $next_word->();
+            }
+            $args_per_word[$word]++;
+            $bits += $needed;
+            if ($needed == 32) {
+                $need_wide_mask[$word]++;
+            }
+            if ($need_wide_mask[$word] and $bits > 32) {
+                # Can only pack two things in a word where one
+                # item is 32 bits. Force the next item into
+                # the next word.
+                $bits = $this_wordsize;
+            }
     }
 
     #
+    # Try to balance packing between words.
+    #
+    if (@args_per_word == 1 and $args_per_word[0] == 1 and $pack_in_iw) {
+        # Don't rebalance.
+    } elsif ($args_per_word[$#args_per_word] == 1) {
+        if ($args_per_word[$#args_per_word-1] < 3) {
+            pop @args_per_word;
+        } else {
+            $args_per_word[$#args_per_word-1]--;
+            $args_per_word[$#args_per_word]++;
+        }
+    } elsif (@args_per_word == 2 and
+             $args_per_word[0] == 4 and
+             $args_per_word[1] == 2) {
+        $args_per_word[0] = 3;
+        $args_per_word[1] = 3;
+    } elsif (@args_per_word == 2 and
+             $args_per_word[0] == 3 and
+             $args_per_word[1] == 1) {
+        $args_per_word[0] = 2;
+        $args_per_word[1] = 2;
+    }
+
+    my $size = 0;
+    my $pack_prefix = '';
+    my $down = '';		# Pack commands (towards instruction
+				# beginning).
+    my $up = '';		# Pack commands (storing back while
+				# moving forward).
+    my $arg_num = 0;            # Number of argument.
+
+    # Skip an unpackable argument. Also handle packing of
+    # an single operand into the instruction word.
+    my $skip_unpackable = sub {
+        my($arg) = @_;
+
+        if ($arg_num == $opt_arg_pos) {
+            my $pack = chr(ord('#') + $arg_num);
+            $down = PACK_CMD_WIDE . "$pack$down";
+            my $unpack = "BeamExtraData(I[0])";
+            $args[$arg_num] = "packed:$arg:0:${arg}b($unpack)";
+        } elsif ($arg_size{$arg}) {
+            # Save the argument on the pack engine's stack.
+            my $push = 'g';
+            if ($type_bit{$arg} & $type_bit{'q'}) {
+                # The operand may be a literal.
+                $push = 'q';
+            } elsif ($type_bit{$arg} & $type_bit{'f'}) {
+                # The operand may be a failure label.
+                $push = 'f';
+            }
+            $down = "$push${down}";
+            $up = "${up}p";
+        }
+    };
+
+    #
     # Now generate the packing instructions.  One complication is that
     # the packing engine works from right-to-left, but we must generate
     # the instructions from left-to-right because we must calculate
     # instruction sizes from left-to-right.
+    for (my $word = 0; $word < @args_per_word; $word++) {
+        my $ap = 0;              # Argument number within word.
+        my $packed_var = "tmp_packed" . ($word+1);
+        my $args_per_word = $args_per_word[$word];
+        my $pack_word_size = ($pack_in_iw && $word == 0) ? 32 : $wordsize;
+
+        my($shref,$mref,$iref,$unpack_suffix)  =
+            get_pack_parameters($name, $args_per_word, $pack_word_size,
+                                $need_wide_mask[$word]);
+        my @shift = @$shref;
+        my @mask = @$mref;
+        my @instr = @$iref;
+
+        while ($ap < $args_per_word) {
+            my $reg = $args[$arg_num];
+            my $this_size = $arg_size{$reg};
+
+            if ($bits_needed[$arg_num]) {
+                $this_size = 0;
+
+                if ($ap == 0) {
+                    my $packed_data;
+                    if ($pack_in_iw and $word == 0) {
+                        $packed_data = "BeamExtraData(I[0])";
+                        if ($args_per_word == 1) {
+                            $packed_var = $packed_data;
+                        } else {
+                            $pack_prefix .= "Eterm $packed_var = $packed_data;\n";
+                        }
+                        my $pack = chr(ord('#') + $size);
+                        $down = "$pack$down";
+                    } else {
+                        $packed_data = arg_offset($size + $offset);
+                        $pack_prefix .= "Eterm $packed_var = $packed_data;\n";
+                        $down = "P$down";
+                        $up .= "p";
+                        $this_size = 1;
+                    }
+                }
+
+                $down = "$instr[$ap]$down";
+                my $unpack = make_unpack($packed_var, $shift[$ap], $mask[$ap]);
+                my $macro = "$reg$unpack_suffix";
+                $args[$arg_num] = "packed:$reg:$this_size:$macro($unpack)";
+
+                $ap++;
+            } else {
+                $skip_unpackable->($reg);
+            }
+            $size += $this_size;
+            $arg_num++;
+        }
+    }
+
     #
-    # XXX Packing 3 't's in one word won't work.  Sorry.
-
-    my $did_some_packing = 0;	# Nothing packed yet.
-    my($ap) = 0;		# Argument number within word.
-    my($tmpnum) = 1;		# Number of temporary variable.
-    my($expr) = '';
-    for (my $i = 0; $i < @args; $i++) {
-	my($reg) = $args[$i];
-	my($this_size) = $arg_size{$reg};
-	if ($is_packable[$i]) {
-	    $this_size = 0;
-	    $did_some_packing = 1;
-
-	    if ($ap == 0) {
-		$pack_prefix .= "tmp_packed$tmpnum = Arg($size);\n";
-		$up .= "p";
-		$down = "P$down";
-		$this_size = 1;
-	    }
+    # Skip any unpackable arguments at the end.
+    #
+    while ($arg_num < @args) {
+        my $arg = $args[$arg_num];
+        $skip_unpackable->($arg);
+        $size += $arg_size{$arg};
+        $arg_num++;
+    }
 
-	    $down = "$instr[$ap]$down";
-	    my($unpack) = make_unpack($tmpnum, $shift[$ap], $mask[$ap]);
-	    $args[$i] = "pack:$this_size:$reg" . "b($unpack)";
+    my $pack_spec = "$down:$up";
+    my $score = pack_score($options, @args);
 
-	    if (++$ap == $args_per_word) {
-		$ap = 0;
-		$tmpnum++;
-	    }
-	} elsif ($arg_size{$reg} && $did_some_packing) {
-	    #
-	    # This is an argument that can't be packed.  Normally, we must
-	    # save it on the pack engine's stack, unless:
-	    #
-	    # 1. The argument has zero size (e.g. r(0)).  Such arguments
-	    #    will not be loaded.  They disappear.
-	    # 2. If the argument is on the left of the first packed argument,
-	    #    the packing engine will never access it (because the engine
-	    #    operates from right-to-left).
-	    #
+    return ($score, [$pack_prefix,$pack_spec,@args]);
+}
 
-	    $down = "g${down}";
-	    $up = "${up}p";
-	}
-	$size += $this_size;
+sub get_pack_parameters {
+    my($name,$args_per_word,$pack_word_size,$wide_mask) = @_;
+    my(@shift,@mask,@instr);
+    my $unpack_suffix = 'b';
+
+    if ($wide_mask and $args_per_word > 1) {
+        @shift = ('0', 'BEAM_WIDE_SHIFT');
+        @mask = ('BEAM_WIDE_MASK', $WHOLE_WORD);
+        @instr = (PACK_CMD_WIDE) x 2;
+    } elsif ($args_per_word == 1) {
+        @shift = ('0');
+        @mask = ($WHOLE_WORD);
+        @instr = (PACK_CMD_WIDE);
+    } elsif ($args_per_word == 2) {
+        if ($pack_word_size != $wordsize) {
+            # 64-bit word size, pack 32 bits into instruction word.
+            @shift = ('0', 'BEAM_TIGHT_SHIFT');
+            @mask  = ('BEAM_TIGHT_MASK', $WHOLE_WORD);
+            @instr = (PACK_CMD_TIGHT) x 2;
+        } else {
+            # 32/64 bit word size
+            @shift = ('0', 'BEAM_LOOSE_SHIFT');
+            @mask  = ('BEAM_LOOSE_MASK', $WHOLE_WORD);
+            @instr = (PACK_CMD_LOOSE) x 2;
+        }
+    } elsif ($args_per_word == 3) {
+        if ($pack_word_size != $wordsize) {
+            # 64-bit word size, pack 3 register numbers into instruction word.
+            @shift = ('0', 'BEAM_TIGHTEST_SHIFT', '(2*BEAM_TIGHTEST_SHIFT)');
+            @mask = ('BEAM_TIGHTEST_MASK', 'BEAM_TIGHTEST_MASK', $WHOLE_WORD);
+            @instr = (PACK_CMD_TIGHTEST) x 3;
+            $unpack_suffix = '';
+        } else {
+            # 32/64 bit word size.
+            @shift = ('0', 'BEAM_TIGHT_SHIFT', '(2*BEAM_TIGHT_SHIFT)');
+            if ($wordsize == 32) {
+                @mask = ('BEAM_TIGHT_MASK') x 3;
+            } elsif ($wordsize == 64) {
+                @mask = ('BEAM_TIGHT_MASK', 'BEAM_TIGHT_MASK', $WHOLE_WORD);
+            }
+            @instr = (PACK_CMD_TIGHT) x 3;
+        }
+    } elsif ($args_per_word == 4) {
+        # 64 bit word size only.
+        @shift = ('0',
+                  'BEAM_LOOSE_SHIFT',
+                  '(2*BEAM_LOOSE_SHIFT)',
+		  '(3*BEAM_LOOSE_SHIFT)');
+        @mask = ('BEAM_LOOSE_MASK', 'BEAM_LOOSE_MASK',
+                 'BEAM_LOOSE_MASK', $WHOLE_WORD);
+        @instr = (PACK_CMD_LOOSE) x 4;
     }
 
-    my $pack_spec = $down . $up;
-    return ($pack_prefix, $pack_spec, @args);
+    unless (@shift) {
+        error("$name: internal packing error: args_per_word=$args_per_word, " .
+              "pack_word_size=$pack_word_size");
+    }
+
+    (\@shift,\@mask,\@instr,$unpack_suffix);
+}
+
+sub pack_score {
+    my($options,@args) = @_;
+    my $size = 0;
+
+    # Calculate the number of words.
+    foreach (@args) {
+        if (/^packed:[^:]*:(\d+)/) {
+            $size += $1;
+        } else {
+            $size += $arg_size{$_}
+        }
+    }
+
+    # Less numbers of words give a higher score; for the same number of
+    # words, using PACK_JUMP or PACK_IN_INSTR_WORD gives a lower score.
+    my $score = 1 + 10*($max_spec_operands - $size);
+    if (($options & PACK_OPT_IN_INSTR_WORD) != 0) {
+        $score += 4;
+    } elsif ($options == PACK_IN_INSTR_WORD) {
+        $score += 0;
+    } elsif ($options == PACK_JUMP) {
+        $score += 1;
+    } elsif ($options == (PACK_JUMP|PACK_IN_INSTR_WORD)) {
+        $score += 2;
+    } elsif ($options == 0) {
+        $score += 3;
+    }
+    $score;
 }
 
 sub make_unpack {
-    my($tmpnum, $shift, $mask) = @_;
+    my($packed_var, $shift, $mask) = @_;
 
-    my($e) = "tmp_packed$tmpnum";
+    my $e = $packed_var;
     $e = "($e>>$shift)" if $shift;
     $e .= "&$mask" unless $mask eq $WHOLE_WORD;
     $e;
 }
 
+sub build_pack_spec {
+    my $pack_spec = shift;
+    return '' if $pack_spec eq '';
+    my($down,$up) = split /:/, $pack_spec;
+    while ($down =~ /[gfq]$/ and $up =~ /^p/) {
+        $down = substr($down, 0, -1);
+        $up = substr($up, 1);
+    }
+    "$down$up";
+}
+
 sub quote {
     local($_) = @_;
     return "'$_'" if $_ eq 'try';
@@ -1286,8 +2180,11 @@ sub parse_transformation {
     #
 
     my @to;
-    if ($to =~ /^(\w+)\((.*?)\)/) {
-	my($name, $arglist) = ($1, $2);
+    if ($to =~ /^(\w+)\((.*?)\)(.*)/) {
+	my($name, $arglist, $garbage) = ($1, $2, $3);
+        if ($garbage =~ /\S/) {
+            error("garbage after call to '$name()'");
+        }
 	@to = (compile_transform_function($name, split(/\s*,\s*/, $arglist)));
     } else {
 	@to = split(/\s*\|\s*/, $to);
@@ -1357,12 +2254,19 @@ sub tr_parse_op {
     if (/^([a-z*]+)(.*)/) {
 	$type = $1;
 	$_ = $2;
+        error("$type: only a single type is allowed on right side of transformations")
+            if not $src and length($type) > 1;
 	foreach (split('', $type)) {
-	    error("bad type in $op")
-		unless defined $type_bit{$_} or $type eq '*';
-	    $_ eq 'r' and
-		error("$op: 'r' is not allowed in transformations")
-	}
+            next if $src and $type eq '*';
+            error("$op: not a type")
+                unless defined $type_bit{$_};
+            error("$op: the type '$_' is not allowed in transformations")
+                unless defined $pattern_type{$_};
+            if (not $src) {
+                error("$op: type '$_' is not allowed on the right side of transformations")
+                    unless defined $construction_type{$_};
+            }
+        }
     }
 
     # Get an optional condition. (In source.)
@@ -1395,10 +2299,18 @@ sub tr_parse_op {
     }
 
     # Get an optional value. (In destination.)
-    $type_val = $type eq 'x' ? 1023 : 0;
+    if ($type eq 'x') {
+        $type_val = 1023;
+    } elsif ($type eq 'a') {
+        $type_val = 'am_Empty';
+    } else {
+        $type_val = 0;
+    }
     if (/^=(.*)/) {
-	error("value not allowed in source: $op")
+	error("$op: value not allowed in source")
 	    if $src;
+        error("$op: the type 'n' must not be given a value")
+            if $type eq 'n';
 	$type_val = $1;
 	$_ = '';
     }
@@ -1408,13 +2320,16 @@ sub tr_parse_op {
     error("garbage '$_' after operand: $op")
 	unless /^\s*$/;
 
-    # Test that destination has no conditions.
+    # Check the conditions.
 
-    unless ($src) {
-	error("condition not allowed in destination: $op")
+    if ($src) {
+        error("$op: the type '$type' is not allowed to be compared with a literal value")
+            if $cond and not $construction_type{$type};
+    } else {
+	error("$op: condition not allowed in destination")
 	    if $cond;
-	error("variable name and type cannot be combined in destination: $op")
-	    if $var && $type;
+	error("$op: variable name and type cannot be combined in destination")
+	    if $var and $type;
     }
 
     ($var,$type,$type_val,$cond,$cond_val);
diff --git a/erts/emulator/utils/make_driver_tab b/erts/emulator/utils/make_driver_tab
index ffb5f58ebf..b7bca1dc3a 100755
--- a/erts/emulator/utils/make_driver_tab
+++ b/erts/emulator/utils/make_driver_tab
@@ -30,6 +30,7 @@ use File::Basename;
 my $file = "";
 my $nif = "";
 my @emu_drivers = ();
+my @emu_nifs = ();
 my @static_drivers = ();
 my @static_nifs = ();
 my $mode = 1;
@@ -61,7 +62,7 @@ while (@ARGV) {
     } elsif ($mode == 2) {
         $d = basename $d;
         $d =~ s/_nif(\..*|)$//; # strip nif.* or just nif
-        push(@static_nifs, $d);
+        push(@emu_nifs, $d);
         next;
     }
     $d = basename $d;
@@ -94,37 +95,33 @@ foreach (@static_drivers) {
 }
 
 # The array itself
-print "\nErlDrvEntry *driver_tab[] =\n{\n";
+print "\nErtsStaticDriver driver_tab[] =\n{\n";
 
 foreach (@emu_drivers) {
-    print "    &${_}driver_entry,\n";
+    print "    {&${_}driver_entry, 0},\n";
 }
 
 foreach (@static_drivers) {
-    print "    NULL, /* ${_} */\n";
+    print "    {NULL, 1}, /* ${_} */\n";
 }
-print "    NULL\n};\n";
+print "    {NULL}\n};\n";
 
 print "void erts_init_static_drivers() {\n";
 
 my $index = 0;
 foreach (@static_drivers) {
-    print "    driver_tab[".(scalar @emu_drivers+$index)."] = ${_}_driver_init();\n";
+    print "    driver_tab[".(scalar @emu_drivers+$index)."].de = ${_}_driver_init();\n";
     $index++;
 }
 
 print "}\n";
 
-print <<EOF;
-
-typedef struct ErtsStaticNifEntry_ {
-    const char *nif_name;
-    ErtsStaticNifInitFPtr nif_init;
-} ErtsStaticNifEntry;
-
-EOF
-
 # prototypes
+foreach (@emu_nifs) {
+    my $d = ${_};
+    $d =~ s/\.debug//; # strip .debug
+    print "void *".$d."_nif_init(void);\n";
+}
 foreach (@static_nifs) {
     my $d = ${_};
     $d =~ s/\.debug//; # strip .debug
@@ -134,20 +131,25 @@ foreach (@static_nifs) {
 # The array itself
 print "static ErtsStaticNifEntry static_nif_tab[] =\n{\n";
 
+foreach (@emu_nifs) {
+    my $d = ${_};
+    $d =~ s/\.debug//; # strip .debug
+    print "    {\"${_}\", &".$d."_nif_init, 0},\n";
+}
 foreach (@static_nifs) {
     my $d = ${_};
     $d =~ s/\.debug//; # strip .debug
-    print "{\"${_}\",&".$d."_nif_init},\n";
+    print "    {\"${_}\", &".$d."_nif_init, 1},\n";
 }
 
 print "    {NULL,NULL}\n};\n";
 
 print <<EOF;
-ErtsStaticNifInitFPtr erts_static_nif_get_nif_init(const char *name, int len) {
+ErtsStaticNifEntry* erts_static_nif_get_nif_init(const char *name, int len) {
     ErtsStaticNifEntry* p;
     for (p = static_nif_tab; p->nif_name != NULL; p++)
         if (strncmp(p->nif_name, name, len) == 0 && p->nif_name[len] == 0)
-            return p->nif_init;
+            return p;
     return NULL;
 }
 
diff --git a/erts/emulator/utils/make_tables b/erts/emulator/utils/make_tables
index 47e1528958..094a35ae4b 100755
--- a/erts/emulator/utils/make_tables
+++ b/erts/emulator/utils/make_tables
@@ -59,7 +59,6 @@ my %dirty_bif_tab;
 
 my @bif;
 my @bif_info;
-my $dirty_schedulers = 'no';
 my $dirty_schedulers_test = 'no';
 my $hipe = 'no';
 
@@ -73,10 +72,6 @@ while (@ARGV && $ARGV[0] =~ /^-(\w+)/) {
 	$include = shift;
 	die "No directory for -include argument specified"
 	    unless defined $include;
-    } elsif($opt eq '-ds') {
-	$dirty_schedulers = shift;
-	die "No -ds argument specified"
-	    unless defined $dirty_schedulers;
     } elsif($opt eq '-dst') {
 	$dirty_schedulers_test = shift;
 	die "No -dst argument specified"
@@ -140,21 +135,19 @@ while (<>) {
 	push(@bif_info, [$type, $sched_type, $alias3, $alias]);
     } elsif ($type eq 'dirty-cpu' or $type eq 'dirty-io'
 	     or $type eq 'dirty-cpu-test' or $type eq 'dirty-io-test') {
-	if ($dirty_schedulers eq 'yes') {
-	    my($bif,$other) = (@args);
-	    $bif =~ m@^([a-z_.'0-9]+):(.*)/(\d)$@ or error("invalid BIF");
-	    my($mod,$name,$arity) = ($1,$2,$3);
-	    my $mfa = "$mod:$name/$arity";
-	    if (($type eq 'dirty-cpu')
-		or (($dirty_schedulers_test eq 'yes')
-		    and ($type eq 'dirty-cpu-test'))) {
-		$dirty_bif_tab{$mfa} = 'dirty_cpu';
-	    } elsif (($type eq 'dirty-io')
-		or (($dirty_schedulers_test eq 'yes')
-		    and ($type eq 'dirty-io-test'))) {
-		$dirty_bif_tab{$mfa} = 'dirty_io';
-	    }
-	}
+        my($bif,$other) = (@args);
+        $bif =~ m@^([a-z_.'0-9]+):(.*)/(\d)$@ or error("invalid BIF");
+        my($mod,$name,$arity) = ($1,$2,$3);
+        my $mfa = "$mod:$name/$arity";
+        if (($type eq 'dirty-cpu')
+            or (($dirty_schedulers_test eq 'yes')
+                and ($type eq 'dirty-cpu-test'))) {
+            $dirty_bif_tab{$mfa} = 'dirty_cpu';
+        } elsif (($type eq 'dirty-io')
+                 or (($dirty_schedulers_test eq 'yes')
+                     and ($type eq 'dirty-io-test'))) {
+            $dirty_bif_tab{$mfa} = 'dirty_io';
+        }
     } else {
 	error("invalid line");
     }