diff options
114 files changed, 1487 insertions, 1258 deletions
diff --git a/OTP_VERSION b/OTP_VERSION index 06d4ac2bfd..a56f451a4e 100644 --- a/OTP_VERSION +++ b/OTP_VERSION @@ -1 +1,2 @@ 21.0-rc0 + diff --git a/bootstrap/bin/no_dot_erlang.boot b/bootstrap/bin/no_dot_erlang.boot Binary files differindex e259dc6ffd..fc355b9c10 100644 --- a/bootstrap/bin/no_dot_erlang.boot +++ b/bootstrap/bin/no_dot_erlang.boot diff --git a/bootstrap/lib/compiler/ebin/beam_listing.beam b/bootstrap/lib/compiler/ebin/beam_listing.beam Binary files differindex 74695349ba..281e9c2d03 100644 --- a/bootstrap/lib/compiler/ebin/beam_listing.beam +++ b/bootstrap/lib/compiler/ebin/beam_listing.beam diff --git a/bootstrap/lib/compiler/ebin/compile.beam b/bootstrap/lib/compiler/ebin/compile.beam Binary files differindex 9aea14162a..5387a0074d 100644 --- a/bootstrap/lib/compiler/ebin/compile.beam +++ b/bootstrap/lib/compiler/ebin/compile.beam diff --git a/bootstrap/lib/compiler/ebin/compiler.app b/bootstrap/lib/compiler/ebin/compiler.app index 24ff29cbf1..01e2c478d4 100644 --- a/bootstrap/lib/compiler/ebin/compiler.app +++ b/bootstrap/lib/compiler/ebin/compiler.app @@ -68,8 +68,7 @@ v3_codegen, v3_core, v3_kernel, - v3_kernel_pp, - v3_life + v3_kernel_pp ]}, {registered, []}, {applications, [kernel, stdlib]}, diff --git a/bootstrap/lib/compiler/ebin/v3_codegen.beam b/bootstrap/lib/compiler/ebin/v3_codegen.beam Binary files differindex e7cbc758f1..09ff7b253d 100644 --- a/bootstrap/lib/compiler/ebin/v3_codegen.beam +++ b/bootstrap/lib/compiler/ebin/v3_codegen.beam diff --git a/bootstrap/lib/compiler/ebin/v3_kernel.beam b/bootstrap/lib/compiler/ebin/v3_kernel.beam Binary files differindex 2286d9a646..56af498763 100644 --- a/bootstrap/lib/compiler/ebin/v3_kernel.beam +++ b/bootstrap/lib/compiler/ebin/v3_kernel.beam diff --git a/bootstrap/lib/compiler/ebin/v3_kernel_pp.beam b/bootstrap/lib/compiler/ebin/v3_kernel_pp.beam Binary files differindex e1b8c58c8f..b67d012f34 100644 --- a/bootstrap/lib/compiler/ebin/v3_kernel_pp.beam +++ b/bootstrap/lib/compiler/ebin/v3_kernel_pp.beam diff --git a/bootstrap/lib/compiler/ebin/v3_life.beam b/bootstrap/lib/compiler/ebin/v3_life.beam Binary files differdeleted file mode 100644 index c20070cf34..0000000000 --- a/bootstrap/lib/compiler/ebin/v3_life.beam +++ /dev/null diff --git a/bootstrap/lib/kernel/ebin/application.beam b/bootstrap/lib/kernel/ebin/application.beam Binary files differindex 05ecd20859..d242e045a1 100644 --- a/bootstrap/lib/kernel/ebin/application.beam +++ b/bootstrap/lib/kernel/ebin/application.beam diff --git a/bootstrap/lib/kernel/ebin/code_server.beam b/bootstrap/lib/kernel/ebin/code_server.beam Binary files differindex 536da5c692..f9ca469b0d 100644 --- a/bootstrap/lib/kernel/ebin/code_server.beam +++ b/bootstrap/lib/kernel/ebin/code_server.beam diff --git a/bootstrap/lib/kernel/ebin/disk_log.beam b/bootstrap/lib/kernel/ebin/disk_log.beam Binary files differindex e8a6acd235..fc71affe55 100644 --- a/bootstrap/lib/kernel/ebin/disk_log.beam +++ b/bootstrap/lib/kernel/ebin/disk_log.beam diff --git a/bootstrap/lib/kernel/ebin/disk_log_1.beam b/bootstrap/lib/kernel/ebin/disk_log_1.beam Binary files differindex 648c658f2e..a42a4c1b86 100644 --- a/bootstrap/lib/kernel/ebin/disk_log_1.beam +++ b/bootstrap/lib/kernel/ebin/disk_log_1.beam diff --git a/bootstrap/lib/kernel/ebin/global_group.beam b/bootstrap/lib/kernel/ebin/global_group.beam Binary files differindex ef0913f98d..1e496fdbbf 100644 --- a/bootstrap/lib/kernel/ebin/global_group.beam +++ b/bootstrap/lib/kernel/ebin/global_group.beam diff --git a/bootstrap/lib/kernel/ebin/group.beam b/bootstrap/lib/kernel/ebin/group.beam Binary files differindex 5c5f3bfa73..5eecaecec8 100644 --- a/bootstrap/lib/kernel/ebin/group.beam +++ b/bootstrap/lib/kernel/ebin/group.beam diff --git a/bootstrap/lib/kernel/ebin/group_history.beam b/bootstrap/lib/kernel/ebin/group_history.beam Binary files differindex 9328355025..80afde896f 100644 --- a/bootstrap/lib/kernel/ebin/group_history.beam +++ b/bootstrap/lib/kernel/ebin/group_history.beam diff --git a/bootstrap/lib/kernel/ebin/inet_db.beam b/bootstrap/lib/kernel/ebin/inet_db.beam Binary files differindex 4ce05435f7..0c77cee005 100644 --- a/bootstrap/lib/kernel/ebin/inet_db.beam +++ b/bootstrap/lib/kernel/ebin/inet_db.beam diff --git a/bootstrap/lib/kernel/ebin/inet_dns.beam b/bootstrap/lib/kernel/ebin/inet_dns.beam Binary files differindex ba611799a2..f36b12952e 100644 --- a/bootstrap/lib/kernel/ebin/inet_dns.beam +++ b/bootstrap/lib/kernel/ebin/inet_dns.beam diff --git a/bootstrap/lib/kernel/ebin/user.beam b/bootstrap/lib/kernel/ebin/user.beam Binary files differindex 947bc0f642..e8b49b2b33 100644 --- a/bootstrap/lib/kernel/ebin/user.beam +++ b/bootstrap/lib/kernel/ebin/user.beam diff --git a/bootstrap/lib/stdlib/ebin/base64.beam b/bootstrap/lib/stdlib/ebin/base64.beam Binary files differindex 00c908062a..4d8141a9c1 100644 --- a/bootstrap/lib/stdlib/ebin/base64.beam +++ b/bootstrap/lib/stdlib/ebin/base64.beam diff --git a/bootstrap/lib/stdlib/ebin/c.beam b/bootstrap/lib/stdlib/ebin/c.beam Binary files differindex 21ce30b8e7..3baf074b66 100644 --- a/bootstrap/lib/stdlib/ebin/c.beam +++ b/bootstrap/lib/stdlib/ebin/c.beam diff --git a/bootstrap/lib/stdlib/ebin/dets.beam b/bootstrap/lib/stdlib/ebin/dets.beam Binary files differindex 9e7801f637..e57da59d9d 100644 --- a/bootstrap/lib/stdlib/ebin/dets.beam +++ b/bootstrap/lib/stdlib/ebin/dets.beam diff --git a/bootstrap/lib/stdlib/ebin/edlin.beam b/bootstrap/lib/stdlib/ebin/edlin.beam Binary files differindex 1cabe095b3..06b7ba8b81 100644 --- a/bootstrap/lib/stdlib/ebin/edlin.beam +++ b/bootstrap/lib/stdlib/ebin/edlin.beam diff --git a/bootstrap/lib/stdlib/ebin/erl_compile.beam b/bootstrap/lib/stdlib/ebin/erl_compile.beam Binary files differindex d39b5a29b0..749c05c54f 100644 --- a/bootstrap/lib/stdlib/ebin/erl_compile.beam +++ b/bootstrap/lib/stdlib/ebin/erl_compile.beam diff --git a/bootstrap/lib/stdlib/ebin/erl_tar.beam b/bootstrap/lib/stdlib/ebin/erl_tar.beam Binary files differindex 6887230609..df0d6cf1b8 100644 --- a/bootstrap/lib/stdlib/ebin/erl_tar.beam +++ b/bootstrap/lib/stdlib/ebin/erl_tar.beam diff --git a/bootstrap/lib/stdlib/ebin/ets.beam b/bootstrap/lib/stdlib/ebin/ets.beam Binary files differindex 61ea75fb2b..95fbd243c3 100644 --- a/bootstrap/lib/stdlib/ebin/ets.beam +++ b/bootstrap/lib/stdlib/ebin/ets.beam diff --git a/bootstrap/lib/stdlib/ebin/filename.beam b/bootstrap/lib/stdlib/ebin/filename.beam Binary files differindex d48f6bc6e1..9ac0d3272c 100644 --- a/bootstrap/lib/stdlib/ebin/filename.beam +++ b/bootstrap/lib/stdlib/ebin/filename.beam diff --git a/bootstrap/lib/stdlib/ebin/io_lib.beam b/bootstrap/lib/stdlib/ebin/io_lib.beam Binary files differindex 82642fb353..3884967cba 100644 --- a/bootstrap/lib/stdlib/ebin/io_lib.beam +++ b/bootstrap/lib/stdlib/ebin/io_lib.beam diff --git a/bootstrap/lib/stdlib/ebin/io_lib_pretty.beam b/bootstrap/lib/stdlib/ebin/io_lib_pretty.beam Binary files differindex 6ed1e3b7cc..183e5b6278 100644 --- a/bootstrap/lib/stdlib/ebin/io_lib_pretty.beam +++ b/bootstrap/lib/stdlib/ebin/io_lib_pretty.beam diff --git a/bootstrap/lib/stdlib/ebin/rand.beam b/bootstrap/lib/stdlib/ebin/rand.beam Binary files differindex 887a09385c..2877b269c9 100644 --- a/bootstrap/lib/stdlib/ebin/rand.beam +++ b/bootstrap/lib/stdlib/ebin/rand.beam diff --git a/bootstrap/lib/stdlib/ebin/string.beam b/bootstrap/lib/stdlib/ebin/string.beam Binary files differindex eb00887566..406b2e19de 100644 --- a/bootstrap/lib/stdlib/ebin/string.beam +++ b/bootstrap/lib/stdlib/ebin/string.beam diff --git a/bootstrap/lib/stdlib/ebin/unicode_util.beam b/bootstrap/lib/stdlib/ebin/unicode_util.beam Binary files differindex 4fcf959fe6..1532f67ef4 100644 --- a/bootstrap/lib/stdlib/ebin/unicode_util.beam +++ b/bootstrap/lib/stdlib/ebin/unicode_util.beam diff --git a/bootstrap/lib/stdlib/ebin/zip.beam b/bootstrap/lib/stdlib/ebin/zip.beam Binary files differindex da5131d05f..1582220c5d 100644 --- a/bootstrap/lib/stdlib/ebin/zip.beam +++ b/bootstrap/lib/stdlib/ebin/zip.beam diff --git a/erts/configure.in b/erts/configure.in index 508e99a415..2f6043ee85 100644 --- a/erts/configure.in +++ b/erts/configure.in @@ -595,6 +595,8 @@ AS_HELP_STRING([--enable-pgo], esac ],enable_pgo=default) +LM_CHECK_ENABLE_CFLAG([-fprofile-use -fprofile-correction -Werror],[PROFILE_CORRECTION]) + USE_PGO=false AC_MSG_CHECKING([whether to do PGO of erts]) if test $enable_pgo = no; then @@ -607,22 +609,16 @@ elif test $CROSS_COMPILING = yes; then fi elif test "X$host" = "Xwin32"; then AC_MSG_RESULT([no, not supported in windows]) -elif test "X$PROFILE_GENERATE" = "Xtrue" -a "X$PROFILE_USE" = "Xtrue"; then +elif test "X$PROFILE_GENERATE" = "Xtrue" -a "X$PROFILE_USE" = "Xtrue" -a "X$PROFILE_CORRECTION" = "Xtrue"; then + ## We need -fprofile-generate and -fprofile-correction support to use PGO with + ## gcc as multiple threads run within the executed object files USE_PGO=true - AC_MSG_RESULT([yes, using -fprofile-generate]) PROFILE_COMPILER=gcc -# check if $CC accepts -fprofile-correction, if so we can use PGO on multi-threaded files. - LM_CHECK_ENABLE_CFLAG([-fprofile-use -fprofile-correction -Werror],[PROFILE_CORRECTION]) - if test "X$PROFILE_CORRECTION" = "Xtrue"; then - PROFILE_CORRECTION="-fprofile-correction" - else - PROFILE_CORRECTION="" - fi - AC_SUBST(PROFILE_CORRECTION) + AC_MSG_RESULT([yes, using -fprofile-generate -fprofile-correction]) elif test "X$PROFILE_INSTR_GENERATE" = "Xtrue" -a "X$PROFILE_INSTR_USE" = "Xtrue"; then USE_PGO=true - AC_MSG_RESULT([yes, using -fprofile-instr-generate]) PROFILE_COMPILER=clang + AC_MSG_RESULT([yes, using -fprofile-instr-generate]) else if $enable_pgo = yes; then AC_MSG_ERROR(cannot use PGO with this compiler) diff --git a/erts/emulator/Makefile.in b/erts/emulator/Makefile.in index 3117b24bcc..fdaf253fe5 100644 --- a/erts/emulator/Makefile.in +++ b/erts/emulator/Makefile.in @@ -74,9 +74,8 @@ endif endif ifeq ($(PROFILE_COMPILER), gcc) -PROFILE_CORRECTION=@PROFILE_CORRECTION@ PROFILE_GENERATE=-fprofile-generate -PROFILE_USE=-fprofile-use $(PROFILE_CORRECTION) +PROFILE_USE=-fprofile-use -fprofile-correction PROFILE_USE_DEPS=$(OBJDIR)/%_pu.gcda endif ifeq ($(PROFILE_COMPILER), clang) @@ -816,14 +815,7 @@ $(ERL_TOP)/lib/%.beam: INIT_OBJS = $(OBJDIR)/erl_main.o $(PRELOAD_OBJ) -# -fprofile-correction is needed in order to use PGO on erl_process -# as multiple threads execute in that file. -ifeq ($(PROFILE_CORRECTION),) -PROFILE_OBJS = $(OBJDIR)/beam_emu.o -RUN_OBJS = $(OBJDIR)/erl_process.o -else PROFILE_OBJS = $(OBJDIR)/beam_emu.o $(OBJDIR)/erl_process.o -endif EMU_OBJS = \ $(OBJDIR)/beam_opcodes.o \ diff --git a/erts/emulator/beam/beam_bp.c b/erts/emulator/beam/beam_bp.c index fe1e15701b..0832b3f374 100644 --- a/erts/emulator/beam/beam_bp.c +++ b/erts/emulator/beam/beam_bp.c @@ -998,7 +998,9 @@ do_call_trace(Process* c_p, ErtsCodeInfo* info, Eterm* reg, fixup_cp_before_trace(c_p, &return_to_trace); + ERTS_UNREQ_PROC_MAIN_LOCK(c_p); flags = erts_call_trace(c_p, info, ms, reg, local, &tracer); + ERTS_REQ_PROC_MAIN_LOCK(c_p); /* restore cp after potential fixup */ c_p->cp = cp_save; diff --git a/erts/emulator/beam/beam_debug.c b/erts/emulator/beam/beam_debug.c index 70078c8c59..509aa2a84f 100644 --- a/erts/emulator/beam/beam_debug.c +++ b/erts/emulator/beam/beam_debug.c @@ -408,7 +408,10 @@ print_op(fmtfn_t to, void *to_arg, int op, int size, BeamInstr* addr) addr++; ap = addr; } else { - BeamInstr instr_word = addr++[0]; +#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. diff --git a/erts/emulator/beam/beam_emu.c b/erts/emulator/beam/beam_emu.c index aa94fbf536..60d0008d8f 100644 --- a/erts/emulator/beam/beam_emu.c +++ b/erts/emulator/beam/beam_emu.c @@ -3199,13 +3199,16 @@ 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; diff --git a/erts/emulator/beam/beam_load.c b/erts/emulator/beam/beam_load.c index bbccd0b79d..7331c331a6 100644 --- a/erts/emulator/beam/beam_load.c +++ b/erts/emulator/beam/beam_load.c @@ -5022,7 +5022,7 @@ freeze_code(LoaderState* stp) */ codev[pos] = (BeamInstr) (codev + value); } else { -#ifdef DEBUG +#if defined(DEBUG) && defined(BEAM_WIDE_MASK) Uint w; #endif Sint32 rel = lp->offset + value; diff --git a/erts/emulator/beam/bif.c b/erts/emulator/beam/bif.c index ad555bb195..4b11884f38 100644 --- a/erts/emulator/beam/bif.c +++ b/erts/emulator/beam/bif.c @@ -4200,7 +4200,7 @@ BIF_RETTYPE list_to_port_1(BIF_ALIST_1) cp += 6; /* 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) diff --git a/erts/emulator/beam/erl_io_queue.c b/erts/emulator/beam/erl_io_queue.c index 190ba6bbb9..40d69ea6b0 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 */ 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_msacc.h b/erts/emulator/beam/erl_msacc.h index 2d4637f800..2588dec903 100644 --- a/erts/emulator/beam/erl_msacc.h +++ b/erts/emulator/beam/erl_msacc.h @@ -270,18 +270,32 @@ void erts_msacc_init_thread(char *type, int id, int liberty); #define ERTS_MSACC_PUSH_STATE_M() \ ERTS_MSACC_DECLARE_CACHE(); \ ERTS_MSACC_PUSH_STATE_CACHED_M() -#define ERTS_MSACC_PUSH_STATE_CACHED_M() \ - __erts_msacc_state = ERTS_MSACC_IS_ENABLED_CACHED() ? \ - erts_msacc_get_state_m__(__erts_msacc_cache) : ERTS_MSACC_STATE_OTHER +#define ERTS_MSACC_PUSH_STATE_CACHED_M() \ + do { \ + if (ERTS_MSACC_IS_ENABLED_CACHED()) { \ + ASSERT(!__erts_msacc_cache->unmanaged); \ + __erts_msacc_state = erts_msacc_get_state_m__(__erts_msacc_cache); \ + } else { \ + __erts_msacc_state = ERTS_MSACC_STATE_OTHER; \ + } \ + } while(0) #define ERTS_MSACC_SET_STATE_M(state) \ ERTS_MSACC_DECLARE_CACHE(); \ ERTS_MSACC_SET_STATE_CACHED_M(state) -#define ERTS_MSACC_SET_STATE_CACHED_M(state) \ - if (ERTS_MSACC_IS_ENABLED_CACHED()) \ - erts_msacc_set_state_m__(__erts_msacc_cache, state, 1) -#define ERTS_MSACC_POP_STATE_M() \ - if (ERTS_MSACC_IS_ENABLED_CACHED()) \ - erts_msacc_set_state_m__(__erts_msacc_cache, __erts_msacc_state, 0) +#define ERTS_MSACC_SET_STATE_CACHED_M(state) \ + do { \ + if (ERTS_MSACC_IS_ENABLED_CACHED()) { \ + ASSERT(!__erts_msacc_cache->unmanaged); \ + erts_msacc_set_state_m__(__erts_msacc_cache, state, 1); \ + } \ + } while(0) +#define ERTS_MSACC_POP_STATE_M() \ + do { \ + if (ERTS_MSACC_IS_ENABLED_CACHED()) { \ + ASSERT(!__erts_msacc_cache->unmanaged); \ + erts_msacc_set_state_m__(__erts_msacc_cache, __erts_msacc_state, 0); \ + } \ + } while(0) #define ERTS_MSACC_PUSH_AND_SET_STATE_M(state) \ ERTS_MSACC_PUSH_STATE_M(); ERTS_MSACC_SET_STATE_CACHED_M(state) diff --git a/erts/emulator/beam/erl_node_tables.h b/erts/emulator/beam/erl_node_tables.h index 3bba673435..ee8277b5ea 100644 --- a/erts/emulator/beam/erl_node_tables.h +++ b/erts/emulator/beam/erl_node_tables.h @@ -271,5 +271,6 @@ erts_de_links_unlock(DistEntry *dep) #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_process.c b/erts/emulator/beam/erl_process.c index 3d12b3bc21..6654468fb6 100644 --- a/erts/emulator/beam/erl_process.c +++ b/erts/emulator/beam/erl_process.c @@ -3276,7 +3276,7 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq) erts_aint32_t aux_work = 0; int thr_prgr_active = 1; erts_aint32_t flgs; - ERTS_MSACC_PUSH_STATE_M(); + ERTS_MSACC_PUSH_STATE(); ERTS_LC_ASSERT(erts_lc_runq_is_locked(rq)); @@ -3385,9 +3385,9 @@ scheduler_wait(int *fcalls, ErtsSchedulerData *esdp, ErtsRunQueue *rq) - 1) + 1; } else timeout = -1; - ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_SLEEP); + ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_SLEEP); res = erts_tse_twait(ssi->event, timeout); - ERTS_MSACC_POP_STATE_M(); + ERTS_MSACC_POP_STATE(); current_time = ERTS_SCHEDULER_IS_DIRTY(esdp) ? 0 : erts_get_monotonic_time(esdp); } while (res == EINTR); @@ -9701,7 +9701,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls) | ERTS_PROC_LOCK_STATUS | ERTS_PROC_LOCK_TRACE)); - ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_OTHER); + ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER); if (state & ERTS_PSFLG_FREE) { if (!is_normal_sched) { @@ -9916,7 +9916,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls) case 0: /* No process at all */ default: ASSERT(qmask == 0); - ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_OTHER); + ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_OTHER); goto check_activities_to_run; } @@ -10034,7 +10034,7 @@ Process *erts_schedule(ErtsSchedulerData *esdp, Process *p, int calls) } - ERTS_MSACC_SET_STATE_CACHED_M(ERTS_MSACC_STATE_EMULATOR); + ERTS_MSACC_SET_STATE_CACHED(ERTS_MSACC_STATE_EMULATOR); if (flags & ERTS_RUNQ_FLG_PROTECTED) diff --git a/erts/emulator/beam/erl_process_dump.c b/erts/emulator/beam/erl_process_dump.c index 12ef4aab8a..0b7f361622 100644 --- a/erts/emulator/beam/erl_process_dump.c +++ b/erts/emulator/beam/erl_process_dump.c @@ -728,8 +728,15 @@ static void mark_literal(Eterm* ptr) ap = bsearch(ptr, lit_areas, num_lit_areas, sizeof(ErtsLiteralArea*), search_areas); - ASSERT(ap); - ap[0]->off_heap = (struct erl_off_heap_header *) 1; + + /* + * If the literal was created by native code, this search will not + * find it and ap will be NULL. + */ + + if (ap) { + ap[0]->off_heap = (struct erl_off_heap_header *) 1; + } } diff --git a/erts/emulator/beam/erl_term.h b/erts/emulator/beam/erl_term.h index 6daf043117..5ec6b6b44b 100644 --- a/erts/emulator/beam/erl_term.h +++ b/erts/emulator/beam/erl_term.h @@ -862,7 +862,7 @@ 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 */ diff --git a/erts/emulator/drivers/common/inet_drv.c b/erts/emulator/drivers/common/inet_drv.c index 554c48059f..95d61fcc5d 100644 --- a/erts/emulator/drivers/common/inet_drv.c +++ b/erts/emulator/drivers/common/inet_drv.c @@ -1136,7 +1136,6 @@ 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 +1160,7 @@ 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 +1195,7 @@ static struct erl_drv_entry sctp_inet_driver_entry = NULL, #else packet_inet_drv_input, - packet_inet_drv_output, + NULL, #endif "sctp_inet", NULL, @@ -1262,7 +1261,6 @@ 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 */ @@ -11300,24 +11298,20 @@ 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 +11826,6 @@ 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/hipe/hipe_amd64_bifs.m4 b/erts/emulator/hipe/hipe_amd64_bifs.m4 index b3c9a460bb..6d998c4b55 100644 --- a/erts/emulator/hipe/hipe_amd64_bifs.m4 +++ b/erts/emulator/hipe/hipe_amd64_bifs.m4 @@ -463,6 +463,43 @@ 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) * noproc_primop_interface_1(nbif_name, cbif_name) * noproc_primop_interface_2(nbif_name, cbif_name) diff --git a/erts/emulator/hipe/hipe_bif0.tab b/erts/emulator/hipe/hipe_bif0.tab index 4038ca7ef8..4f73770d24 100644 --- a/erts/emulator/hipe/hipe_bif0.tab +++ b/erts/emulator/hipe/hipe_bif0.tab @@ -140,3 +140,4 @@ atom bs_validate_unicode_retract atom emulate_fpe atom emasculate_binary atom is_divisible +atom is_unicode
\ No newline at end of file diff --git a/erts/emulator/hipe/hipe_bif_list.m4 b/erts/emulator/hipe/hipe_bif_list.m4 index bebe20a18e..08c1ab9318 100644 --- a/erts/emulator/hipe/hipe_bif_list.m4 +++ b/erts/emulator/hipe/hipe_bif_list.m4 @@ -223,6 +223,7 @@ 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_1(nbif_is_unicode, hipe_is_unicode) /* * Mbox primops with implicit P parameter. @@ -247,7 +248,11 @@ nofail_primop_interface_3(nbif_bs_get_float_2, erts_bs_get_float_2) standard_bif_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. diff --git a/erts/emulator/hipe/hipe_native_bif.c b/erts/emulator/hipe/hipe_native_bif.c index 23f64a6991..4d20fbdb90 100644 --- a/erts/emulator/hipe/hipe_native_bif.c +++ b/erts/emulator/hipe/hipe_native_bif.c @@ -491,6 +491,12 @@ BIF_RETTYPE nbif_impl_hipe_bs_validate_unicode(NBIF_ALIST_1) return NIL; } +Uint hipe_is_unicode(Eterm arg) +{ + return (Uint) validate_unicode(arg); +} + + int hipe_bs_validate_unicode_retract(ErlBinMatchBuffer* mb, Eterm arg) { if (!validate_unicode(arg)) { diff --git a/erts/emulator/hipe/hipe_native_bif.h b/erts/emulator/hipe/hipe_native_bif.h index cbc7ab8dc6..aa961ab6d0 100644 --- a/erts/emulator/hipe/hipe_native_bif.h +++ b/erts/emulator/hipe/hipe_native_bif.h @@ -67,6 +67,7 @@ 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)); @@ -92,6 +93,7 @@ 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); diff --git a/erts/emulator/hipe/hipe_primops.h b/erts/emulator/hipe/hipe_primops.h index 6aac5e6205..49e6bdb026 100644 --- a/erts/emulator/hipe/hipe_primops.h +++ b/erts/emulator/hipe/hipe_primops.h @@ -64,6 +64,7 @@ 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) PRIMOP_LIST(am_is_divisible, &nbif_is_divisible) 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). + + + +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. + + + +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. + + + +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. + + + +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). + + + +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. + + + +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/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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew 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 Binary files differnew file mode 100644 index 0000000000..b835c7694e --- /dev/null +++ b/erts/emulator/internal_doc/figures/gc-watermark.png diff --git a/erts/emulator/sys/common/erl_poll.c b/erts/emulator/sys/common/erl_poll.c index 30a595c17a..7aa53e8f36 100644 --- a/erts/emulator/sys/common/erl_poll.c +++ b/erts/emulator/sys/common/erl_poll.c @@ -2059,7 +2059,8 @@ uint32_t epoll_events(int kp_fd, int fd) int ev_fd; uint32_t events; uint64_t data; - if (fscanf(f,"tfd:%d events:%x data:%lx\n", &ev_fd, &events, &data) != 3) { + 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); @@ -2125,7 +2126,8 @@ ERTS_POLL_EXPORT(erts_poll_get_selected_events)(ErtsPollSet *ps, int fd; uint32_t events; uint64_t data; - if (fscanf(f,"tfd:%d events:%x data:%lx\n", &fd, &events, &data) != 3) { + 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); diff --git a/erts/emulator/test/bif_SUITE.erl b/erts/emulator/test/bif_SUITE.erl index 146c37b118..e1b42e5d85 100644 --- a/erts/emulator/test/bif_SUITE.erl +++ b/erts/emulator/test/bif_SUITE.erl @@ -781,14 +781,20 @@ 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. diff --git a/erts/emulator/test/port_SUITE.erl b/erts/emulator/test/port_SUITE.erl index f29528a22f..e46665a881 100644 --- a/erts/emulator/test/port_SUITE.erl +++ b/erts/emulator/test/port_SUITE.erl @@ -1035,6 +1035,9 @@ 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 diff --git a/erts/start_scripts/Makefile b/erts/start_scripts/Makefile index e730422ed8..5a47ad9616 100644 --- a/erts/start_scripts/Makefile +++ b/erts/start_scripts/Makefile @@ -171,7 +171,7 @@ $(ERL_TOP)/bin/no_dot_erlang.script: $(ERLC) $(SCRIPT_PATH) +no_warn_sasl +otp_build +no_dot_erlang -o $@ $(SS_ROOT)/no_dot_erlang.rel ) ## Special target used from system/build/Makefile for source code release bootstrap. -bootstrap_scripts: $(SS_ROOT)/start_clean.rel +bootstrap_scripts: $(SS_ROOT)/start_clean.rel $(SS_ROOT)/no_dot_erlang.rel $(V_at)$(INSTALL_DIR) $(TESTROOT)/bin $(V_at)$(INSTALL_DIR) $(SS_TMP) $(V_at)( cd $(SS_TMP) && \ diff --git a/lib/compiler/src/Makefile b/lib/compiler/src/Makefile index 9b22e5197b..9e96147787 100644 --- a/lib/compiler/src/Makefile +++ b/lib/compiler/src/Makefile @@ -93,16 +93,14 @@ MODULES = \ v3_codegen \ v3_core \ v3_kernel \ - v3_kernel_pp \ - v3_life + v3_kernel_pp BEAM_H = $(wildcard ../priv/beam_h/*.h) HRL_FILES= \ beam_disasm.hrl \ core_parse.hrl \ - v3_kernel.hrl \ - v3_life.hrl + v3_kernel.hrl YRL_FILE = core_parse.yrl @@ -187,7 +185,7 @@ release_docs_spec: # ---------------------------------------------------- $(EBIN)/beam_disasm.beam: $(EGEN)/beam_opcodes.hrl beam_disasm.hrl -$(EBIN)/beam_listing.beam: v3_life.hrl +$(EBIN)/beam_listing.beam: core_parse.hrl v3_kernel.hrl $(EBIN)/beam_validator.beam: beam_disasm.hrl $(EBIN)/cerl.beam: core_parse.hrl $(EBIN)/compile.beam: core_parse.hrl ../../stdlib/include/erl_compile.hrl @@ -200,8 +198,7 @@ $(EBIN)/sys_core_dsetel.beam: core_parse.hrl $(EBIN)/sys_core_fold.beam: core_parse.hrl $(EBIN)/sys_core_fold_lists.beam: core_parse.hrl $(EBIN)/sys_core_inline.beam: core_parse.hrl -$(EBIN)/v3_codegen.beam: v3_life.hrl +$(EBIN)/v3_codegen.beam: v3_kernel.hrl $(EBIN)/v3_core.beam: core_parse.hrl $(EBIN)/v3_kernel.beam: core_parse.hrl v3_kernel.hrl $(EBIN)/v3_kernel_pp.beam: v3_kernel.hrl -$(EBIN)/v3_life.beam: v3_kernel.hrl v3_life.hrl diff --git a/lib/compiler/src/beam_listing.erl b/lib/compiler/src/beam_listing.erl index 94b47cf568..9422f9a78a 100644 --- a/lib/compiler/src/beam_listing.erl +++ b/lib/compiler/src/beam_listing.erl @@ -23,14 +23,12 @@ -include("core_parse.hrl"). -include("v3_kernel.hrl"). --include("v3_life.hrl"). -import(lists, [foreach/2]). -type code() :: cerl:c_module() | beam_utils:module_code() | #k_mdef{} - | {module(),_,_,_} %v3_life | [_]. %form-based format -spec module(file:io_device(), code()) -> 'ok'. @@ -42,13 +40,9 @@ module(File, #k_mdef{}=Kern) -> %% This is a kernel module. io:put_chars(File, v3_kernel_pp:format(Kern)); %%io:put_chars(File, io_lib:format("~p~n", [Kern])); -module(File, {Mod,Exp,Attr,Kern}) -> - %% This is output from beam_life (v3). - io:fwrite(File, "~w.~n~p.~n~p.~n", [Mod,Exp,Attr]), - foreach(fun (F) -> function(File, F) end, Kern); module(Stream, {Mod,Exp,Attr,Code,NumLabels}) -> - %% This is output from beam_codegen. - io:format(Stream, "{module, ~p}. %% version = ~w\n", + %% This is output from v3_codegen. + io:format(Stream, "{module, ~p}. %% version = ~w\n", [Mod, beam_opcodes:format_number()]), io:format(Stream, "\n{exports, ~p}.\n", [Exp]), io:format(Stream, "\n{attributes, ~p}.\n", [Attr]), @@ -68,60 +62,3 @@ format_asm([{label,L}|Is]) -> format_asm([I|Is]) -> [io_lib:format(" ~p", [I]),".\n"|format_asm(Is)]; format_asm([]) -> []. - -function(File, {function,Name,Arity,Args,Body,Vdb,_Anno}) -> - io:nl(File), - io:format(File, "function ~p/~p.\n", [Name,Arity]), - io:format(File, " ~p.\n", [Args]), - print_vdb(File, Vdb), - put(beam_listing_nl, false), - nl(File), - foreach(fun(F) -> format(File, F, []) end, Body), - nl(File), - erase(beam_listing_nl). - -format(File, #l{ke=Ke,i=I,vdb=Vdb}, Ind) -> - nl(File), - ind_format(File, Ind, "~p ", [I]), - print_vdb(File, Vdb), - nl(File), - format(File, Ke, Ind); -format(File, Tuple, Ind) when is_tuple(Tuple) -> - ind_format(File, Ind, "{", []), - format_list(File, tuple_to_list(Tuple), [$\s|Ind]), - ind_format(File, Ind, "}", []); -format(File, List, Ind) when is_list(List) -> - ind_format(File, Ind, "[", []), - format_list(File, List, [$\s|Ind]), - ind_format(File, Ind, "]", []); -format(File, F, Ind) -> - ind_format(File, Ind, "~p", [F]). - -format_list(File, [F], Ind) -> - format(File, F, Ind); -format_list(File, [F|Fs], Ind) -> - format(File, F, Ind), - ind_format(File, Ind, ",", []), - format_list(File, Fs, Ind); -format_list(_, [], _) -> ok. - - -print_vdb(File, [{Var,F,E}|Vs]) -> - io:format(File, "~p:~p..~p ", [Var,F,E]), - print_vdb(File, Vs); -print_vdb(_, []) -> ok. - -ind_format(File, Ind, Format, Args) -> - case get(beam_listing_nl) of - true -> - put(beam_listing_nl, false), - io:put_chars(File, Ind); - false -> ok - end, - io:format(File, Format, Args). - -nl(File) -> - case put(beam_listing_nl, true) of - true -> ok; - false -> io:nl(File) - end. diff --git a/lib/compiler/src/compile.erl b/lib/compiler/src/compile.erl index 1b359d1e59..bc519264fc 100644 --- a/lib/compiler/src/compile.erl +++ b/lib/compiler/src/compile.erl @@ -733,8 +733,6 @@ kernel_passes() -> ?pass(v3_kernel), {iff,dkern,{listing,"kernel"}}, {iff,'to_kernel',{done,"kernel"}}, - {pass,v3_life}, - {iff,dlife,{listing,"life"}}, {pass,v3_codegen}, {iff,dcg,{listing,"codegen"}} | asm_passes()]. @@ -1947,7 +1945,6 @@ pre_load() -> sys_core_fold, v3_codegen, v3_core, - v3_kernel, - v3_life], + v3_kernel], _ = code:ensure_modules_loaded(L), ok. diff --git a/lib/compiler/src/compiler.app.src b/lib/compiler/src/compiler.app.src index 703cf1d1b8..cf32fd251c 100644 --- a/lib/compiler/src/compiler.app.src +++ b/lib/compiler/src/compiler.app.src @@ -68,8 +68,7 @@ v3_codegen, v3_core, v3_kernel, - v3_kernel_pp, - v3_life + v3_kernel_pp ]}, {registered, []}, {applications, [kernel, stdlib]}, diff --git a/lib/compiler/src/v3_codegen.erl b/lib/compiler/src/v3_codegen.erl index ee5bafbc5c..006a6a82d2 100644 --- a/lib/compiler/src/v3_codegen.erl +++ b/lib/compiler/src/v3_codegen.erl @@ -19,25 +19,6 @@ %% %% Purpose : Code generator for Beam. -%% The following assumptions have been made: -%% -%% 1. Matches, i.e. things with {match,M,Ret} wrappers, only return -%% values; no variables are exported. If the match would have returned -%% extra variables then these have been transformed to multiple return -%% values. -%% -%% 2. All BIF's called in guards are gc-safe so there is no need to -%% put thing on the stack in the guard. While this would in principle -%% work it would be difficult to keep track of the stack depth when -%% trimming. -%% -%% The code generation uses variable lifetime information added by -%% the v3_life module to save variables, allocate registers and -%% move registers to the stack when necessary. -%% -%% We try to use a consistent variable name scheme throughout. The -%% StackReg record is always called Bef,Int<n>,Aft. - -module(v3_codegen). %% The main interface. @@ -45,12 +26,14 @@ -import(lists, [member/2,keymember/3,keysort/2,keydelete/3, append/1,flatmap/2,filter/2,foldl/3,foldr/3,mapfoldl/3, - sort/1,reverse/1,reverse/2]). --import(v3_life, [vdb_find/2]). + sort/1,reverse/1,reverse/2,map/2]). +-import(ordsets, [add_element/2,intersection/2,union/2]). -%%-compile([export_all]). +-include("v3_kernel.hrl"). --include("v3_life.hrl"). +%% These are not defined in v3_kernel.hrl. +get_kanno(Kthing) -> element(2, Kthing). +set_kanno(Kthing, Anno) -> setelement(2, Kthing, Anno). %% Main codegen structure. -record(cg, {lcount=1, %Label counter @@ -61,38 +44,273 @@ functable=#{}, %Map of local functions: {Name,Arity}=>Label in_catch=false, %Inside a catch or not. need_frame, %Need a stack frame. - ultimate_failure %Label for ultimate match failure. - }). + ultimate_failure, %Label for ultimate match failure. + ctx %Match context. + }). %% Stack/register state record. -record(sr, {reg=[], %Register table stk=[], %Stack table res=[]}). %Reserved regs: [{reserved,I,V}] --type life_module() :: {module(),_,_,[_]}. +%% Internal records. +-record(cg_need_heap, {anno=[] :: term(), + h=0 :: integer()}). +-record(cg_block, {anno=[] :: term(), + es=[] :: [term()]}). + +-type vdb_entry() :: {atom(),non_neg_integer(),non_neg_integer()}. + +-record(l, {i=0 :: non_neg_integer(), %Op number + vdb=[] :: [vdb_entry()], %Variable database + a=[] :: [term()]}). %Core annotation --spec module(life_module(), [compile:option()]) -> {'ok',beam_asm:module_code()}. +-spec module(#k_mdef{}, [compile:option()]) -> {'ok',beam_asm:module_code()}. -module({Mod,Exp,Attr,Forms}, _Options) -> - {Fs,St} = functions(Forms, {atom,Mod}), - {ok,{Mod,Exp,Attr,Fs,St#cg.lcount}}. +module(#k_mdef{name=Mod,exports=Es,attributes=Attr,body=Forms}, _Opts) -> + {Asm,St} = functions(Forms, {atom,Mod}), + {ok,{Mod,Es,Attr,Asm,St#cg.lcount}}. functions(Forms, AtomMod) -> mapfoldl(fun (F, St) -> function(F, AtomMod, St) end, #cg{lcount=1}, Forms). -function({function,Name,Arity,Asm0,Vb,Vdb,Anno}, AtomMod, St0) -> +function(#k_fdef{anno=#k{a=Anno},func=Name,arity=Arity, + vars=As,body=Kb}, AtomMod, St0) -> try - {Asm,EntryLabel,St} = cg_fun(Vb, Asm0, Vdb, AtomMod, - {Name,Arity}, Anno, St0), - Func = {function,Name,Arity,EntryLabel,Asm}, - {Func,St} + %% Annotate kernel records with variable usage. + #k_match{} = Kb, %Assertion. + Vdb0 = init_vars(As), + {Body,_,Vdb} = body(Kb, 1, Vdb0), + + %% Generate the BEAM assembly code. + {Asm,EntryLabel,St} = cg_fun(Body, As, Vdb, AtomMod, + {Name,Arity}, Anno, St0), + Func = {function,Name,Arity,EntryLabel,Asm}, + {Func,St} catch - Class:Error -> - Stack = erlang:get_stacktrace(), - io:fwrite("Function: ~w/~w\n", [Name,Arity]), - erlang:raise(Class, Error, Stack) + Class:Error -> + Stack = erlang:get_stacktrace(), + io:fwrite("Function: ~w/~w\n", [Name,Arity]), + erlang:raise(Class, Error, Stack) end. +%% This pass creates beam format annotated with variable lifetime +%% information. Each thing is given an index and for each variable we +%% store the first and last index for its occurrence. The variable +%% database, VDB, attached to each thing is only relevant internally +%% for that thing. +%% +%% For nested things like matches the numbering continues locally and +%% the VDB for that thing refers to the variable usage within that +%% thing. Variables which live through a such a thing are internally +%% given a very large last index. Internally the indexes continue +%% after the index of that thing. This creates no problems as the +%% internal variable info never escapes and externally we only see +%% variable which are alive both before or after. +%% +%% This means that variables never "escape" from a thing and the only +%% way to get values from a thing is to "return" them, with 'break' or +%% 'return'. Externally these values become the return values of the +%% thing. This is no real limitation as most nested things have +%% multiple threads so working out a common best variable usage is +%% difficult. + +%% body(Kbody, I, Vdb) -> {[Expr],MaxI,Vdb}. +%% Handle a body. + +body(#k_seq{arg=Ke,body=Kb}, I, Vdb0) -> + %%ok = io:fwrite("life ~w:~p~n", [?LINE,{Ke,I,Vdb0}]), + A = get_kanno(Ke), + Vdb1 = use_vars(union(A#k.us, A#k.ns), I, Vdb0), + {Es,MaxI,Vdb2} = body(Kb, I+1, Vdb1), + E = expr(Ke, I, Vdb2), + {[E|Es],MaxI,Vdb2}; +body(Ke, I, Vdb0) -> + %%ok = io:fwrite("life ~w:~p~n", [?LINE,{Ke,I,Vdb0}]), + A = get_kanno(Ke), + Vdb1 = use_vars(union(A#k.us, A#k.ns), I, Vdb0), + E = expr(Ke, I, Vdb1), + {[E],I,Vdb1}. + +%% expr(Kexpr, I, Vdb) -> Expr. + +expr(#k_test{anno=A}=Test, I, _Vdb) -> + Test#k_test{anno=#l{i=I,a=A#k.a}}; +expr(#k_call{anno=A}=Call, I, _Vdb) -> + Call#k_call{anno=#l{i=I,a=A#k.a}}; +expr(#k_enter{anno=A}=Enter, I, _Vdb) -> + Enter#k_enter{anno=#l{i=I,a=A#k.a}}; +expr(#k_bif{anno=A}=Bif, I, _Vdb) -> + Bif#k_bif{anno=#l{i=I,a=A#k.a}}; +expr(#k_match{anno=A,body=Kb,ret=Rs}, I, Vdb) -> + %% Work out imported variables which need to be locked. + Mdb = vdb_sub(I, I+1, Vdb), + M = match(Kb, A#k.us, I+1, Mdb), + L = #l{i=I,vdb=use_vars(A#k.us, I+1, Mdb),a=A#k.a}, + #k_match{anno=L,body=M,ret=Rs}; +expr(#k_guard_match{anno=A,body=Kb,ret=Rs}, I, Vdb) -> + %% Work out imported variables which need to be locked. + Mdb = vdb_sub(I, I+1, Vdb), + M = match(Kb, A#k.us, I+1, Mdb), + L = #l{i=I,vdb=use_vars(A#k.us, I+1, Mdb),a=A#k.a}, + #k_guard_match{anno=L,body=M,ret=Rs}; +expr(#k_protected{}=Protected, I, Vdb) -> + protected(Protected, I, Vdb); +expr(#k_try{anno=A,arg=Ka,vars=Vs,body=Kb,evars=Evs,handler=Kh}=Try, I, Vdb) -> + %% Lock variables that are alive before the catch and used afterwards. + %% Don't lock variables that are only used inside the try. + Tdb0 = vdb_sub(I, I+1, Vdb), + %% This is the tricky bit. Lock variables in Arg that are used in + %% the body and handler. Add try tag 'variable'. + Ab = get_kanno(Kb), + Ah = get_kanno(Kh), + Tdb1 = use_vars(union(Ab#k.us, Ah#k.us), I+3, Tdb0), + Tdb2 = vdb_sub(I, I+2, Tdb1), + Vnames = fun (Kvar) -> Kvar#k_var.name end, %Get the variable names + {Aes,_,Adb} = body(Ka, I+2, add_var({catch_tag,I+1}, I+1, locked, Tdb2)), + {Bes,_,Bdb} = body(Kb, I+4, new_vars(sort(map(Vnames, Vs)), I+3, Tdb2)), + {Hes,_,Hdb} = body(Kh, I+4, new_vars(sort(map(Vnames, Evs)), I+3, Tdb2)), + L = #l{i=I,vdb=Tdb1,a=A#k.a}, + Try#k_try{anno=L, + arg=#cg_block{es=Aes,anno=#l{i=I+1,vdb=Adb,a=[]}}, + vars=Vs,body=#cg_block{es=Bes,anno=#l{i=I+3,vdb=Bdb,a=[]}}, + evars=Evs,handler=#cg_block{es=Hes,anno=#l{i=I+3,vdb=Hdb,a=[]}}}; +expr(#k_try_enter{anno=A,arg=Ka,vars=Vs,body=Kb,evars=Evs,handler=Kh}, I, Vdb) -> + %% Lock variables that are alive before the catch and used afterwards. + %% Don't lock variables that are only used inside the try. + Tdb0 = vdb_sub(I, I+1, Vdb), + %% This is the tricky bit. Lock variables in Arg that are used in + %% the body and handler. Add try tag 'variable'. + Ab = get_kanno(Kb), + Ah = get_kanno(Kh), + Tdb1 = use_vars(union(Ab#k.us, Ah#k.us), I+3, Tdb0), + Tdb2 = vdb_sub(I, I+2, Tdb1), + Vnames = fun (Kvar) -> Kvar#k_var.name end, %Get the variable names + {Aes,_,Adb} = body(Ka, I+2, add_var({catch_tag,I+1}, I+1, 1000000, Tdb2)), + {Bes,_,Bdb} = body(Kb, I+4, new_vars(sort(map(Vnames, Vs)), I+3, Tdb2)), + {Hes,_,Hdb} = body(Kh, I+4, new_vars(sort(map(Vnames, Evs)), I+3, Tdb2)), + L = #l{i=I,vdb=Tdb1,a=A#k.a}, + #k_try_enter{anno=L, + arg=#cg_block{es=Aes,anno=#l{i=I+1,vdb=Adb,a=[]}}, + vars=Vs,body=#cg_block{es=Bes,anno=#l{i=I+3,vdb=Bdb,a=[]}}, + evars=Evs,handler=#cg_block{es=Hes,anno=#l{i=I+3,vdb=Hdb,a=[]}}}; +expr(#k_catch{anno=A,body=Kb}=Catch, I, Vdb) -> + %% Lock variables that are alive before the catch and used afterwards. + %% Don't lock variables that are only used inside the catch. + %% Add catch tag 'variable'. + Cdb0 = vdb_sub(I, I+1, Vdb), + {Es,_,Cdb1} = body(Kb, I+1, add_var({catch_tag,I}, I, locked, Cdb0)), + L = #l{i=I,vdb=Cdb1,a=A#k.a}, + Catch#k_catch{anno=L,body=#cg_block{es=Es}}; +expr(#k_receive{anno=A,var=V,body=Kb,action=Ka}=Recv, I, Vdb) -> + %% Work out imported variables which need to be locked. + Rdb = vdb_sub(I, I+1, Vdb), + M = match(Kb, add_element(V#k_var.name, A#k.us), I+1, + new_vars([V#k_var.name], I, Rdb)), + {Tes,_,Adb} = body(Ka, I+1, Rdb), + Le = #l{i=I,vdb=use_vars(A#k.us, I+1, Vdb),a=A#k.a}, + Recv#k_receive{anno=Le,body=M, + action=#cg_block{anno=#l{i=I+1,vdb=Adb,a=[]},es=Tes}}; +expr(#k_receive_accept{anno=A}, I, _Vdb) -> + #k_receive_accept{anno=#l{i=I,a=A#k.a}}; +expr(#k_receive_next{anno=A}, I, _Vdb) -> + #k_receive_next{anno=#l{i=I,a=A#k.a}}; +expr(#k_put{anno=A}=Put, I, _Vdb) -> + Put#k_put{anno=#l{i=I,a=A#k.a}}; +expr(#k_break{anno=A}=Break, I, _Vdb) -> + Break#k_break{anno=#l{i=I,a=A#k.a}}; +expr(#k_guard_break{anno=A}=Break, I, Vdb) -> + Locked = [V || {V,_,_} <- Vdb], + L = #l{i=I,a=A#k.a}, + Break#k_guard_break{anno=L,locked=Locked}; +expr(#k_return{anno=A}=Ret, I, _Vdb) -> + Ret#k_return{anno=#l{i=I,a=A#k.a}}. + +%% protected(Kprotected, I, Vdb) -> Protected. +%% Only used in guards. + +protected(#k_protected{anno=A,arg=Ts}=Prot, I, Vdb) -> + %% Lock variables that are alive before try and used afterwards. + %% Don't lock variables that are only used inside the protected + %% expression. + Pdb0 = vdb_sub(I, I+1, Vdb), + {T,MaxI,Pdb1} = body(Ts, I+1, Pdb0), + Pdb2 = use_vars(A#k.ns, MaxI+1, Pdb1), %Save "return" values + Prot#k_protected{arg=T,anno=#l{i=I,a=A#k.a,vdb=Pdb2}}. + +%% match(Kexpr, [LockVar], I, Vdb) -> Expr. +%% Convert match tree to old format. + +match(#k_alt{anno=A,first=Kf,then=Kt}, Ls, I, Vdb0) -> + Vdb1 = use_vars(union(A#k.us, Ls), I, Vdb0), + F = match(Kf, Ls, I+1, Vdb1), + T = match(Kt, Ls, I+1, Vdb1), + #k_alt{anno=[],first=F,then=T}; +match(#k_select{anno=A,var=V,types=Kts}=Select, Ls0, I, Vdb0) -> + Vanno = get_kanno(V), + Ls1 = case member(no_usage, Vanno) of + false -> add_element(V#k_var.name, Ls0); + true -> Ls0 + end, + Vdb1 = use_vars(union(A#k.us, Ls1), I, Vdb0), + Ts = [type_clause(Tc, Ls1, I+1, Vdb1) || Tc <- Kts], + Select#k_select{anno=[],types=Ts}; +match(#k_guard{anno=A,clauses=Kcs}, Ls, I, Vdb0) -> + Vdb1 = use_vars(union(A#k.us, Ls), I, Vdb0), + Cs = [guard_clause(G, Ls, I+1, Vdb1) || G <- Kcs], + #k_guard{anno=[],clauses=Cs}; +match(Other, Ls, I, Vdb0) -> + Vdb1 = use_vars(Ls, I, Vdb0), + {B,_,Vdb2} = body(Other, I+1, Vdb1), + Le = #l{i=I,vdb=Vdb2,a=[]}, + #cg_block{anno=Le,es=B}. + +type_clause(#k_type_clause{anno=A,type=T,values=Kvs}, Ls, I, Vdb0) -> + %%ok = io:format("life ~w: ~p~n", [?LINE,{T,Kvs}]), + Vdb1 = use_vars(union(A#k.us, Ls), I+1, Vdb0), + Vs = [val_clause(Vc, Ls, I+1, Vdb1) || Vc <- Kvs], + #k_type_clause{anno=[],type=T,values=Vs}. + +val_clause(#k_val_clause{anno=A,val=V,body=Kb}, Ls0, I, Vdb0) -> + New = (get_kanno(V))#k.ns, + Bus = (get_kanno(Kb))#k.us, + %%ok = io:format("Ls0 = ~p, Used=~p\n New=~p, Bus=~p\n", [Ls0,Used,New,Bus]), + Ls1 = union(intersection(New, Bus), Ls0), %Lock for safety + Vdb1 = use_vars(union(A#k.us, Ls1), I+1, new_vars(New, I, Vdb0)), + B = match(Kb, Ls1, I+1, Vdb1), + Le = #l{i=I,vdb=use_vars(Bus, I+1, Vdb1),a=A#k.a}, + #k_val_clause{anno=Le,val=V,body=B}. + +guard_clause(#k_guard_clause{anno=A,guard=Kg,body=Kb}, Ls, I, Vdb0) -> + Vdb1 = use_vars(union(A#k.us, Ls), I+2, Vdb0), + Gdb = vdb_sub(I+1, I+2, Vdb1), + G = protected(Kg, I+1, Gdb), + B = match(Kb, Ls, I+2, Vdb1), + Le = #l{i=I,vdb=use_vars((get_kanno(Kg))#k.us, I+2, Vdb1),a=A#k.a}, + #k_guard_clause{anno=Le,guard=G,body=B}. + + +%% Here follows the code generator pass. +%% +%% The following assumptions have been made: +%% +%% 1. Matches, i.e. things with {match,M,Ret} wrappers, only return +%% values; no variables are exported. If the match would have returned +%% extra variables then these have been transformed to multiple return +%% values. +%% +%% 2. All BIF's called in guards are gc-safe so there is no need to +%% put thing on the stack in the guard. While this would in principle +%% work it would be difficult to keep track of the stack depth when +%% trimming. +%% +%% The code generation uses variable lifetime information added by +%% the previous pass to save variables, allocate registers and +%% move registers to the stack when necessary. +%% +%% We try to use a consistent variable name scheme throughout. The +%% StackReg record is always called Bef,Int<n>,Aft. + %% cg_fun([Lkexpr], [HeadVar], Vdb, State) -> {[Ainstr],State} cg_fun(Les, Hvs, Vdb, AtomMod, NameArity, Anno, St0) -> @@ -114,14 +332,14 @@ cg_fun(Les, Hvs, Vdb, AtomMod, NameArity, Anno, St0) -> %% Note that and 'if_end' instruction does not need any %% live x registers, so it will always be safe to jump to %% it. (We never ever expect the jump to be taken, and in - %% must functions there will never be any references to + %% most functions there will never be any references to %% the label in the first place.) %% {UltimateMatchFail,St3} = new_label(St2), %% Create initial stack/register state, clear unused arguments. - Bef = clear_dead(#sr{reg=foldl(fun ({var,V}, Reg) -> + Bef = clear_dead(#sr{reg=foldl(fun (#k_var{name=V}, Reg) -> put_reg(V, Reg) end, [], Hvs), stk=[]}, 0, Vdb), @@ -136,45 +354,43 @@ cg_fun(Les, Hvs, Vdb, AtomMod, NameArity, Anno, St0) -> %% cg(Lkexpr, Vdb, StackReg, State) -> {[Ainstr],StackReg,State}. %% Generate code for a kexpr. -%% Split function into two steps for clarity, not efficiency. -cg(Le, Vdb, Bef, St) -> - cg(Le#l.ke, Le, Vdb, Bef, St). - -cg({block,Es}, Le, Vdb, Bef, St) -> +cg(#cg_block{anno=Le,es=Es}, Vdb, Bef, St) -> block_cg(Es, Le, Vdb, Bef, St); -cg({match,M,Rs}, Le, Vdb, Bef, St) -> +cg(#k_match{anno=Le,body=M,ret=Rs}, Vdb, Bef, St) -> match_cg(M, Rs, Le, Vdb, Bef, St); -cg({guard_match,M,Rs}, Le, Vdb, Bef, St) -> +cg(#k_guard_match{anno=Le,body=M,ret=Rs}, Vdb, Bef, St) -> guard_match_cg(M, Rs, Le, Vdb, Bef, St); -cg({call,Func,As,Rs}, Le, Vdb, Bef, St) -> +cg(#k_call{anno=Le,op=Func,args=As,ret=Rs}, Vdb, Bef, St) -> call_cg(Func, As, Rs, Le, Vdb, Bef, St); -cg({enter,Func,As}, Le, Vdb, Bef, St) -> +cg(#k_enter{anno=Le,op=Func,args=As}, Vdb, Bef, St) -> enter_cg(Func, As, Le, Vdb, Bef, St); -cg({bif,Bif,As,Rs}, Le, Vdb, Bef, St) -> - bif_cg(Bif, As, Rs, Le, Vdb, Bef, St); -cg({gc_bif,Bif,As,Rs}, Le, Vdb, Bef, St) -> - gc_bif_cg(Bif, As, Rs, Le, Vdb, Bef, St); -cg({internal,Bif,As,Rs}, Le, Vdb, Bef, St) -> - internal_cg(Bif, As, Rs, Le, Vdb, Bef, St); -cg({receive_loop,Te,Rvar,Rm,Tes,Rs}, Le, Vdb, Bef, St) -> +cg(#k_bif{anno=Le}=Bif, Vdb, Bef, St) -> + bif_cg(Bif, Le, Vdb, Bef, St); +cg(#k_receive{anno=Le,timeout=Te,var=Rvar,body=Rm,action=Tes,ret=Rs}, + Vdb, Bef, St) -> recv_loop_cg(Te, Rvar, Rm, Tes, Rs, Le, Vdb, Bef, St); -cg(receive_next, Le, Vdb, Bef, St) -> +cg(#k_receive_next{anno=Le}, Vdb, Bef, St) -> recv_next_cg(Le, Vdb, Bef, St); -cg(receive_accept, _Le, _Vdb, Bef, St) -> {[remove_message],Bef,St}; -cg({'try',Ta,Vs,Tb,Evs,Th,Rs}, Le, Vdb, Bef, St) -> +cg(#k_receive_accept{}, _Vdb, Bef, St) -> + {[remove_message],Bef,St}; +cg(#k_try{anno=Le,arg=Ta,vars=Vs,body=Tb,evars=Evs,handler=Th,ret=Rs}, + Vdb, Bef, St) -> try_cg(Ta, Vs, Tb, Evs, Th, Rs, Le, Vdb, Bef, St); -cg({try_enter,Ta,Vs,Tb,Evs,Th}, Le, Vdb, Bef, St) -> +cg(#k_try_enter{anno=Le,arg=Ta,vars=Vs,body=Tb,evars=Evs,handler=Th}, + Vdb, Bef, St) -> try_enter_cg(Ta, Vs, Tb, Evs, Th, Le, Vdb, Bef, St); -cg({'catch',Cb,R}, Le, Vdb, Bef, St) -> +cg(#k_catch{anno=Le,body=Cb,ret=[R]}, Vdb, Bef, St) -> catch_cg(Cb, R, Le, Vdb, Bef, St); -cg({set,Var,Con}, Le, Vdb, Bef, St) -> - set_cg(Var, Con, Le, Vdb, Bef, St); -cg({return,Rs}, Le, Vdb, Bef, St) -> return_cg(Rs, Le, Vdb, Bef, St); -cg({break,Bs}, Le, Vdb, Bef, St) -> break_cg(Bs, Le, Vdb, Bef, St); -cg({guard_break,Bs,N}, Le, Vdb, Bef, St) -> +cg(#k_put{anno=Le,arg=Con,ret=Var}, Vdb, Bef, St) -> + put_cg(Var, Con, Le, Vdb, Bef, St); +cg(#k_return{anno=Le,args=Rs}, Vdb, Bef, St) -> + return_cg(Rs, Le, Vdb, Bef, St); +cg(#k_break{anno=Le,args=Bs}, Vdb, Bef, St) -> + break_cg(Bs, Le, Vdb, Bef, St); +cg(#k_guard_break{anno=Le,args=Bs,locked=N}, Vdb, Bef, St) -> guard_break_cg(Bs, N, Le, Vdb, Bef, St); -cg({need_heap,H}, _Le, _Vdb, Bef, St) -> +cg(#cg_need_heap{h=H}, _Vdb, Bef, St) -> {[{test_heap,H,max_reg(Bef#sr.reg)}],Bef,St}. %% cg_list([Kexpr], FirstI, Vdb, StackReg, St) -> {[Ainstr],StackReg,St}. @@ -191,11 +407,11 @@ cg_list(Kes, I, Vdb, Bef, St0) -> %% Insert need_heap instructions in Kexpr list. Try to be smart and %% collect them together as much as possible. -need_heap(Kes0, I) -> +need_heap(Kes0, _I) -> {Kes,H} = need_heap_0(reverse(Kes0), 0, []), %% Prepend need_heap if necessary. - need_heap_need(I, H) ++ Kes. + need_heap_need(H) ++ Kes. need_heap_0([Ke|Kes], H0, Acc) -> {Ns,H} = need_heap_1(Ke, H0), @@ -203,27 +419,54 @@ need_heap_0([Ke|Kes], H0, Acc) -> need_heap_0([], H, Acc) -> {Acc,H}. -need_heap_1(#l{ke={set,_,{binary,_}},i=I}, H) -> - {need_heap_need(I, H),0}; -need_heap_1(#l{ke={set,_,{map,_,_,_}},i=I}, H) -> - {need_heap_need(I, H),0}; -need_heap_1(#l{ke={set,_,Val}}, H) -> +need_heap_1(#k_put{arg=#k_binary{}}, H) -> + {need_heap_need(H),0}; +need_heap_1(#k_put{arg=#k_map{}}, H) -> + {need_heap_need(H),0}; +need_heap_1(#k_put{arg=Val}, H) -> %% Just pass through adding to needed heap. {[],H + case Val of - {cons,_} -> 2; - {tuple,Es} -> 1 + length(Es); + #k_cons{} -> 2; + #k_tuple{es=Es} -> 1 + length(Es); _Other -> 0 end}; -need_heap_1(#l{ke={bif,_Bif,_As,_Rs}}, H) -> - {[],H}; -need_heap_1(#l{i=I}, H) -> +need_heap_1(#k_bif{}=Bif, H) -> + case is_gc_bif(Bif) of + false -> + {[],H}; + true -> + {need_heap_need(H),0} + end; +need_heap_1(_Ke, H) -> %% Call or call-like instruction such as set_tuple_element/3. - {need_heap_need(I, H),0}. - -need_heap_need(_I, 0) -> []; -need_heap_need(I, H) -> [#l{ke={need_heap,H},i=I}]. - -%% match_cg(Match, [Ret], Le, Vdb, StackReg, State) -> + {need_heap_need(H),0}. + +need_heap_need(0) -> []; +need_heap_need(H) -> [#cg_need_heap{h=H}]. + +%% is_gc_bif(#k_bif{}) -> true|false. +%% is_gc_bif(Name, Arity) -> true|false. +%% Determines whether the BIF Name/Arity might do a GC. + +is_gc_bif(#k_bif{op=#k_remote{name=#k_atom{val=Name}},args=Args}) -> + is_gc_bif(Name, length(Args)); +is_gc_bif(#k_bif{op=#k_internal{}}) -> + true. + +is_gc_bif(hd, 1) -> false; +is_gc_bif(tl, 1) -> false; +is_gc_bif(self, 0) -> false; +is_gc_bif(node, 0) -> false; +is_gc_bif(node, 1) -> false; +is_gc_bif(element, 2) -> false; +is_gc_bif(get, 1) -> false; +is_gc_bif(tuple_size, 1) -> false; +is_gc_bif(Bif, Arity) -> + not (erl_internal:bool_op(Bif, Arity) orelse + erl_internal:new_type_test(Bif, Arity) orelse + erl_internal:comp_op(Bif, Arity)). + +%% match_cg(Matc, [Ret], Le, Vdb, StackReg, State) -> %% {[Ainstr],StackReg,State}. %% Generate code for a match. First save all variables on the stack %% that are to survive after the match. We leave saved variables in @@ -252,7 +495,7 @@ guard_match_cg(M, Rs, Le, Vdb, Bef, St0) -> clear_dead(Aft#sr{reg=Reg}, I, Vdb), St2#cg{break=St1#cg.break}}. -guard_match_regs([{I,gbreakvar}|Rs], [{var,V}|Vs]) -> +guard_match_regs([{I,gbreakvar}|Rs], [#k_var{name=V}|Vs]) -> [{I,V}|guard_match_regs(Rs, Vs)]; guard_match_regs([R|Rs], Vs) -> [R|guard_match_regs(Rs, Vs)]; @@ -264,17 +507,14 @@ guard_match_regs([], []) -> []. %% down as each level which uses this takes its own internal Vdb not %% the outer one. -match_cg(Le, Fail, Bef, St) -> - match_cg(Le#l.ke, Le, Fail, Bef, St). - -match_cg({alt,F,S}, _Le, Fail, Bef, St0) -> +match_cg(#k_alt{first=F,then=S}, Fail, Bef, St0) -> {Tf,St1} = new_label(St0), {Fis,Faft,St2} = match_cg(F, Tf, Bef, St1), {Sis,Saft,St3} = match_cg(S, Fail, Bef, St2), Aft = sr_merge(Faft, Saft), {Fis ++ [{label,Tf}] ++ Sis,Aft,St3}; -match_cg({select,{var,Vname}=V,Scs0}, #l{a=Anno}, Fail, Bef, St) -> - ReuseForContext = member(reuse_for_context, Anno) andalso +match_cg(#k_select{var=#k_var{anno=Vanno,name=Vname}=V,types=Scs0}, Fail, Bef, St) -> + ReuseForContext = member(reuse_for_context, Vanno) andalso find_reg(Vname, Bef#sr.reg) =/= error, Scs = case ReuseForContext of false -> Scs0; @@ -283,10 +523,10 @@ match_cg({select,{var,Vname}=V,Scs0}, #l{a=Anno}, Fail, Bef, St) -> match_fmf(fun (S, F, Sta) -> select_cg(S, V, F, Fail, Bef, Sta) end, Fail, St, Scs); -match_cg({guard,Gcs}, _Le, Fail, Bef, St) -> +match_cg(#k_guard{clauses=Gcs}, Fail, Bef, St) -> match_fmf(fun (G, F, Sta) -> guard_clause_cg(G, F, Bef, Sta) end, Fail, St, Gcs); -match_cg({block,Es}, Le, _Fail, Bef, St) -> +match_cg(#cg_block{anno=Le,es=Es}, _Fail, Bef, St) -> %% Must clear registers and stack of dead variables. Int = clear_dead(Bef, Le#l.i, Le#l.vdb), block_cg(Es, Le, Int, St). @@ -294,8 +534,8 @@ match_cg({block,Es}, Le, _Fail, Bef, St) -> %% bsm_rename_ctx([Clause], Var) -> [Clause] %% We know from an annotation that the register for a binary can %% be reused for the match context because the two are not truly -%% alive at the same time (even though the conservative life time -%% information calculated by v3_life says so). +%% alive at the same time (even though the life time information +%% says so). %% %% The easiest way to have those variables share the same register is %% to rename the variable with the shortest life-span (the match @@ -306,12 +546,14 @@ match_cg({block,Es}, Le, _Fail, Bef, St) -> %% We must also remove all information about the match context %% variable from all life-time information databases (Vdb). -bsm_rename_ctx([#l{ke={type_clause,binary, - [#l{ke={val_clause,{binary,{var,Old}},Ke0}}=L2]}}=L1|Cs], New) -> +bsm_rename_ctx([#k_type_clause{type=k_binary,values=Vcs}=TC|Cs], New) -> + [#k_val_clause{val=#k_binary{segs=#k_var{name=Old}}=Bin, + body=Ke0}=VC0] = Vcs, Ke = bsm_rename_ctx(Ke0, Old, New, false), - [L1#l{ke={type_clause,binary, - [L2#l{ke={val_clause,{binary,{var,New}},Ke}}]}}|bsm_rename_ctx(Cs, New)]; -bsm_rename_ctx([C|Cs], New) -> + VC = VC0#k_val_clause{val=Bin#k_binary{segs=#k_var{name=New}}, + body=Ke}, + [TC#k_type_clause{values=[VC]}|bsm_rename_ctx(Cs, New)]; +bsm_rename_ctx([C|Cs], New) -> [C|bsm_rename_ctx(Cs, New)]; bsm_rename_ctx([], _) -> []. @@ -321,34 +563,24 @@ bsm_rename_ctx([], _) -> []. %% only complicatate things to recurse into blocks not in a protected %% (the match context variable is not live inside them). -bsm_rename_ctx(#l{ke={select,{var,V},Cs0}}=L, Old, New, InProt) -> +bsm_rename_ctx(#k_select{var=#k_var{name=V},types=Cs0}=Sel, + Old, New, InProt) -> Cs = bsm_rename_ctx_list(Cs0, Old, New, InProt), - L#l{ke={select,{var,bsm_rename_var(V, Old, New)},Cs}}; -bsm_rename_ctx(#l{ke={type_clause,Type,Cs0}}=L, Old, New, InProt) -> + Sel#k_select{var=#k_var{name=bsm_rename_var(V, Old, New)},types=Cs}; +bsm_rename_ctx(#k_type_clause{values=Cs0}=TC, Old, New, InProt) -> Cs = bsm_rename_ctx_list(Cs0, Old, New, InProt), - L#l{ke={type_clause,Type,Cs}}; -bsm_rename_ctx(#l{ke={val_clause,{bin_end,V},Ke0}}=L, Old, New, InProt) -> - Ke = bsm_rename_ctx(Ke0, Old, New, InProt), - L#l{ke={val_clause,{bin_end,bsm_rename_var(V, Old, New)},Ke}}; -bsm_rename_ctx(#l{ke={val_clause,{bin_seg,V,Sz,U,Type,Fl,Vs},Ke0}}=L, - Old, New, InProt) -> - Ke = bsm_rename_ctx(Ke0, Old, New, InProt), - L#l{ke={val_clause,{bin_seg,bsm_rename_var(V, Old, New),Sz,U,Type,Fl,Vs},Ke}}; -bsm_rename_ctx(#l{ke={val_clause,{bin_int,V,Sz,U,Fl,Val,Vs},Ke0}}=L, - Old, New, InProt) -> - Ke = bsm_rename_ctx(Ke0, Old, New, InProt), - L#l{ke={val_clause,{bin_int,bsm_rename_var(V, Old, New),Sz,U,Fl,Val,Vs},Ke}}; -bsm_rename_ctx(#l{ke={val_clause,Val,Ke0}}=L, Old, New, InProt) -> + TC#k_type_clause{values=Cs}; +bsm_rename_ctx(#k_val_clause{body=Ke0}=VC, Old, New, InProt) -> Ke = bsm_rename_ctx(Ke0, Old, New, InProt), - L#l{ke={val_clause,Val,Ke}}; -bsm_rename_ctx(#l{ke={alt,F0,S0}}=L, Old, New, InProt) -> + VC#k_val_clause{body=Ke}; +bsm_rename_ctx(#k_alt{first=F0,then=S0}=Alt, Old, New, InProt) -> F = bsm_rename_ctx(F0, Old, New, InProt), S = bsm_rename_ctx(S0, Old, New, InProt), - L#l{ke={alt,F,S}}; -bsm_rename_ctx(#l{ke={guard,Gcs0}}=L, Old, New, InProt) -> + Alt#k_alt{first=F,then=S}; +bsm_rename_ctx(#k_guard{clauses=Gcs0}=Guard, Old, New, InProt) -> Gcs = bsm_rename_ctx_list(Gcs0, Old, New, InProt), - L#l{ke={guard,Gcs}}; -bsm_rename_ctx(#l{ke={guard_clause,G0,B0}}=L, Old, New, InProt) -> + Guard#k_guard{clauses=Gcs}; +bsm_rename_ctx(#k_guard_clause{guard=G0,body=B0}=GC, Old, New, InProt) -> G = bsm_rename_ctx(G0, Old, New, InProt), B = bsm_rename_ctx(B0, Old, New, InProt), %% A guard clause may cause unsaved variables to be saved on the stack. @@ -356,49 +588,49 @@ bsm_rename_ctx(#l{ke={guard_clause,G0,B0}}=L, Old, New, InProt) -> %% same register), it is neither in the stack nor register descriptor %% lists and we would crash when we didn't find it unless we remove %% it from the database. - bsm_forget_var(L#l{ke={guard_clause,G,B}}, Old); -bsm_rename_ctx(#l{ke={protected,Ts0,Rs}}=L, Old, New, _InProt) -> + bsm_forget_var(GC#k_guard_clause{guard=G,body=B}, Old); +bsm_rename_ctx(#k_protected{arg=Ts0}=Prot, Old, New, _InProt) -> InProt = true, Ts = bsm_rename_ctx_list(Ts0, Old, New, InProt), - bsm_forget_var(L#l{ke={protected,Ts,Rs}}, Old); -bsm_rename_ctx(#l{ke={match,Ms0,Rs}}=L, Old, New, InProt) -> + bsm_forget_var(Prot#k_protected{arg=Ts}, Old); +bsm_rename_ctx(#k_match{body=Ms0}=Match, Old, New, InProt) -> Ms = bsm_rename_ctx(Ms0, Old, New, InProt), - L#l{ke={match,Ms,Rs}}; -bsm_rename_ctx(#l{ke={guard_match,Ms0,Rs}}=L, Old, New, InProt) -> + Match#k_match{body=Ms}; +bsm_rename_ctx(#k_guard_match{body=Ms0}=Match, Old, New, InProt) -> Ms = bsm_rename_ctx(Ms0, Old, New, InProt), - L#l{ke={guard_match,Ms,Rs}}; -bsm_rename_ctx(#l{ke={test,_,_,_}}=L, _, _, _) -> L; -bsm_rename_ctx(#l{ke={bif,_,_,_}}=L, _, _, _) -> L; -bsm_rename_ctx(#l{ke={gc_bif,_,_,_}}=L, _, _, _) -> L; -bsm_rename_ctx(#l{ke={set,_,_}}=L, _, _, _) -> L; -bsm_rename_ctx(#l{ke={call,_,_,_}}=L, _, _, _) -> L; -bsm_rename_ctx(#l{ke={block,_}}=L, Old, _, false) -> + Match#k_guard_match{body=Ms}; +bsm_rename_ctx(#k_test{}=Test, _, _, _) -> Test; +bsm_rename_ctx(#k_bif{}=Bif, _, _, _) -> Bif; +bsm_rename_ctx(#k_put{}=Put, _, _, _) -> Put; +bsm_rename_ctx(#k_call{}=Call, _, _, _) -> Call; +bsm_rename_ctx(#cg_block{}=Block, Old, _, false) -> %% This block is not inside a protected. The match context variable cannot %% possibly be live inside the block. - bsm_forget_var(L, Old); -bsm_rename_ctx(#l{ke={block,Bl0}}=L, Old, New, true) -> + bsm_forget_var(Block, Old); +bsm_rename_ctx(#cg_block{es=Es0}=Block, Old, New, true) -> %% A block in a protected. We must recursively rename the variable %% inside the block. - Bl = bsm_rename_ctx_list(Bl0, Old, New, true), - bsm_forget_var(L#l{ke={block,Bl}}, Old); -bsm_rename_ctx(#l{ke={guard_break,Bs,Locked0}}=L0, Old, _New, _InProt) -> + Es = bsm_rename_ctx_list(Es0, Old, New, true), + bsm_forget_var(Block#cg_block{es=Es}, Old); +bsm_rename_ctx(#k_guard_break{locked=Locked0}=Break, Old, _New, _InProt) -> Locked = Locked0 -- [Old], - L = L0#l{ke={guard_break,Bs,Locked}}, - bsm_forget_var(L, Old). + bsm_forget_var(Break#k_guard_break{locked=Locked}, Old). bsm_rename_ctx_list([C|Cs], Old, New, InProt) -> [bsm_rename_ctx(C, Old, New, InProt)| bsm_rename_ctx_list(Cs, Old, New, InProt)]; bsm_rename_ctx_list([], _, _, _) -> []. - + bsm_rename_var(Old, Old, New) -> New; bsm_rename_var(V, _, _) -> V. %% bsm_forget_var(#l{}, Variable) -> #l{} %% Remove a variable from the variable life-time database. -bsm_forget_var(#l{vdb=Vdb}=L, V) -> - L#l{vdb=keydelete(V, 1, Vdb)}. +bsm_forget_var(Ke, V) -> + #l{vdb=Vdb} = L0 = get_kanno(Ke), + L = L0#l{vdb=keydelete(V, 1, Vdb)}, + set_kanno(Ke, L). %% block_cg([Kexpr], Le, Vdb, StackReg, St) -> {[Ainstr],StackReg,St}. %% block_cg([Kexpr], Le, StackReg, St) -> {[Ainstr],StackReg,St}. @@ -421,7 +653,7 @@ cg_block(Kes0, I, Vdb, Bef, St0) -> case basic_block(Kes0) of {Kes1,LastI,Args,Rest} -> Ke = hd(Kes1), - Fb = Ke#l.i, + #l{i=Fb} = get_kanno(Ke), cg_basic_block(Kes1, Fb, LastI, Args, Vdb, Bef, St0); {Kes1,Rest} -> cg_list(Kes1, I, Vdb, Bef, St0) @@ -431,36 +663,46 @@ cg_block(Kes0, I, Vdb, Bef, St0) -> basic_block(Kes) -> basic_block(Kes, []). -basic_block([Le|Les], Acc) -> - case collect_block(Le#l.ke) of - include -> basic_block(Les, [Le|Acc]); +basic_block([Ke|Kes], Acc) -> + case collect_block(Ke) of + include -> basic_block(Kes, [Ke|Acc]); {block_end,As} -> case Acc of [] -> - %% If the basic block does not contain any set instructions, + %% If the basic block does not contain any #k_put{} instructions, %% it serves no useful purpose to do basic block optimizations. - {[Le],Les}; + {[Ke],Kes}; _ -> - {reverse(Acc, [Le]),Le#l.i,As,Les} + #l{i=I} = get_kanno(Ke), + {reverse(Acc, [Ke]),I,As,Kes} end; - no_block -> {reverse(Acc, [Le]),Les} + no_block -> {reverse(Acc, [Ke]),Kes} end. -%% sets that may garbage collect are not allowed in basic blocks. - -collect_block({set,_,{binary,_}}) -> no_block; -collect_block({set,_,{map,_,_,_}}) -> no_block; -collect_block({set,_,_}) -> include; -collect_block({call,{var,_}=Var,As,_Rs}) -> {block_end,As++[Var]}; -collect_block({call,Func,As,_Rs}) -> {block_end,As++func_vars(Func)}; -collect_block({enter,{var,_}=Var,As})-> {block_end,As++[Var]}; -collect_block({enter,Func,As}) -> {block_end,As++func_vars(Func)}; -collect_block({return,Rs}) -> {block_end,Rs}; -collect_block({break,Bs}) -> {block_end,Bs}; +%% #k_put{} instructions that may garbage collect are not allowed in basic blocks. + +collect_block(#k_put{arg=#k_binary{}}) -> + no_block; +collect_block(#k_put{arg=#k_map{}}) -> + no_block; +collect_block(#k_put{}) -> + include; +collect_block(#k_call{op=#k_var{}=Var,args=As}) -> + {block_end,As++[Var]}; +collect_block(#k_call{op=Func,args=As}) -> + {block_end,As++func_vars(Func)}; +collect_block(#k_enter{op=#k_var{}=Var,args=As}) -> + {block_end,As++[Var]}; +collect_block(#k_enter{op=Func,args=As}) -> + {block_end,As++func_vars(Func)}; +collect_block(#k_return{args=Rs}) -> + {block_end,Rs}; +collect_block(#k_break{args=Bs}) -> + {block_end,Bs}; collect_block(_) -> no_block. -func_vars({remote,M,F}) when element(1, M) =:= var; - element(1, F) =:= var -> +func_vars(#k_remote{mod=M,name=F}) + when is_record(M, k_var); is_record(F, k_var) -> [M,F]; func_vars(_) -> []. @@ -478,18 +720,19 @@ cg_basic_block(Kes, Fb, Lf, As, Vdb, Bef, St0) -> {Int0,X0_v0,St0}, need_heap(Kes, Fb)), {Keis,Aft,St1}. -cg_basic_block(#l{ke={need_heap,_}}=Ke, {Inta,X0v,Sta}, _Lf, Vdb) -> +cg_basic_block(#cg_need_heap{}=Ke, {Inta,X0v,Sta}, _Lf, Vdb) -> {Keis,Intb,Stb} = cg(Ke, Vdb, Inta, Sta), {Keis, {Intb,X0v,Stb}}; cg_basic_block(Ke, {Inta,X0_v1,Sta}, Lf, Vdb) -> - {Sis,Intb} = save_carefully(Inta, Ke#l.i, Lf+1, Vdb), - {X0_v2,Intc} = allocate_x0(X0_v1, Ke#l.i, Intb), + #l{i=I} = get_kanno(Ke), + {Sis,Intb} = save_carefully(Inta, I, Lf+1, Vdb), + {X0_v2,Intc} = allocate_x0(X0_v1, I, Intb), Intd = reserve(Intc), {Keis,Inte,Stb} = cg(Ke, Vdb, Intd, Sta), {Sis ++ Keis, {Inte,X0_v2,Stb}}. make_reservation([], _) -> []; -make_reservation([{var,V}|As], I) -> [{I,V}|make_reservation(As, I+1)]; +make_reservation([#k_var{name=V}|As], I) -> [{I,V}|make_reservation(As, I+1)]; make_reservation([A|As], I) -> [{I,A}|make_reservation(As, I+1)]. reserve(Sr) -> Sr#sr{reg=reserve(Sr#sr.res, Sr#sr.reg, Sr#sr.stk)}. @@ -538,7 +781,7 @@ save_carefully([V|Vs], Bef, Acc) -> end. x0_vars([], _Fb, _Lf, _Vdb) -> []; -x0_vars([{var,V}|_], Fb, _Lf, Vdb) -> +x0_vars([#k_var{name=V}|_], Fb, _Lf, Vdb) -> {V,F,_L} = VFL = vdb_find(V, Vdb), x0_vars1([VFL], Fb, F, Vdb); x0_vars([X0|_], Fb, Lf, Vdb) -> @@ -640,23 +883,27 @@ turn_yreg(Other, _MaxY) -> %% wrong. These are different as in the second case there is no need %% to try the next type, it will always fail. -select_cg(#l{ke={type_clause,cons,[S]}}, {var,V}, Tf, Vf, Bef, St) -> +select_cg(#k_type_clause{type=Type,values=Vs}, Var, Tf, Vf, Bef, St) -> + #k_var{name=V} = Var, + select_cg(Type, Vs, V, Tf, Vf, Bef, St). + +select_cg(k_cons, [S], V, Tf, Vf, Bef, St) -> select_cons(S, V, Tf, Vf, Bef, St); -select_cg(#l{ke={type_clause,nil,[S]}}, {var,V}, Tf, Vf, Bef, St) -> +select_cg(k_nil, [S], V, Tf, Vf, Bef, St) -> select_nil(S, V, Tf, Vf, Bef, St); -select_cg(#l{ke={type_clause,binary,[S]}}, {var,V}, Tf, Vf, Bef, St) -> +select_cg(k_binary, [S], V, Tf, Vf, Bef, St) -> select_binary(S, V, Tf, Vf, Bef, St); -select_cg(#l{ke={type_clause,bin_seg,S}}, {var,V}, Tf, _Vf, Bef, St) -> +select_cg(k_bin_seg, S, V, Tf, _Vf, Bef, St) -> select_bin_segs(S, V, Tf, Bef, St); -select_cg(#l{ke={type_clause,bin_int,S}}, {var,V}, Tf, _Vf, Bef, St) -> +select_cg(k_bin_int, S, V, Tf, _Vf, Bef, St) -> select_bin_segs(S, V, Tf, Bef, St); -select_cg(#l{ke={type_clause,bin_end,[S]}}, {var,V}, Tf, _Vf, Bef, St) -> +select_cg(k_bin_end, [S], V, Tf, _Vf, Bef, St) -> select_bin_end(S, V, Tf, Bef, St); -select_cg(#l{ke={type_clause,map,S}}, {var,V}, Tf, Vf, Bef, St) -> +select_cg(k_map, S, V, Tf, Vf, Bef, St) -> select_map(S, V, Tf, Vf, Bef, St); -select_cg(#l{ke={type_clause,literal,S}}, {var,V}, Tf, Vf, Bef, St) -> +select_cg(k_literal, S, V, Tf, Vf, Bef, St) -> select_literal(S, V, Tf, Vf, Bef, St); -select_cg(#l{ke={type_clause,Type,Scs}}, {var,V}, Tf, Vf, Bef, St0) -> +select_cg(Type, Scs, V, Tf, Vf, Bef, St0) -> {Vis,{Aft,St1}} = mapfoldl(fun (S, {Int,Sta}) -> {Val,Is,Inta,Stb} = select_val(S, V, Vf, Bef, Sta), @@ -666,22 +913,29 @@ select_cg(#l{ke={type_clause,Type,Scs}}, {var,V}, Tf, Vf, Bef, St0) -> {Vls,Sis,St2} = select_labels(OptVls, St1, [], []), {select_val_cg(Type, fetch_var(V, Bef), Vls, Tf, Vf, Sis), Aft, St2}. -select_val_cg(tuple, R, [Arity,{f,Lbl}], Tf, Vf, [{label,Lbl}|Sis]) -> +select_val_cg(k_tuple, R, [Arity,{f,Lbl}], Tf, Vf, [{label,Lbl}|Sis]) -> [{test,is_tuple,{f,Tf},[R]},{test,test_arity,{f,Vf},[R,Arity]}|Sis]; -select_val_cg(tuple, R, Vls, Tf, Vf, Sis) -> +select_val_cg(k_tuple, R, Vls, Tf, Vf, Sis) -> [{test,is_tuple,{f,Tf},[R]},{select_tuple_arity,R,{f,Vf},{list,Vls}}|Sis]; select_val_cg(Type, R, [Val, {f,Lbl}], Fail, Fail, [{label,Lbl}|Sis]) -> - [{test,is_eq_exact,{f,Fail},[R,{Type,Val}]}|Sis]; + [{test,is_eq_exact,{f,Fail},[R,{type(Type),Val}]}|Sis]; select_val_cg(Type, R, [Val, {f,Lbl}], Tf, Vf, [{label,Lbl}|Sis]) -> [{test,select_type_test(Type),{f,Tf},[R]}, - {test,is_eq_exact,{f,Vf},[R,{Type,Val}]}|Sis]; + {test,is_eq_exact,{f,Vf},[R,{type(Type),Val}]}|Sis]; select_val_cg(Type, R, Vls0, Tf, Vf, Sis) -> - Vls1 = [case Value of {f,_Lbl} -> Value; _ -> {Type,Value} end || Value <- Vls0], + Vls1 = [case Value of + {f,_Lbl} -> Value; + _ -> {type(Type),Value} + end || Value <- Vls0], [{test,select_type_test(Type),{f,Tf},[R]}, {select_val,R,{f,Vf},{list,Vls1}}|Sis]. - -select_type_test(integer) -> is_integer; -select_type_test(atom) -> is_atom; -select_type_test(float) -> is_float. + +type(k_atom) -> atom; +type(k_float) -> float; +type(k_int) -> integer. + +select_type_test(k_int) -> is_integer; +select_type_test(k_atom) -> is_atom; +select_type_test(k_float) -> is_float. combine([{Is,Vs1}, {Is,Vs2}|Vis]) -> combine([{Is,Vs1 ++ Vs2}|Vis]); combine([V|Vis]) -> [V|combine(Vis)]; @@ -706,36 +960,40 @@ select_literal(S, V, Tf, Vf, Bef, St) -> end, match_fmf(F, Tf, St, S). -select_cons(#l{ke={val_clause,{cons,Es},B},i=I,vdb=Vdb}, V, Tf, Vf, Bef, St0) -> +select_cons(#k_val_clause{val=#k_cons{hd=Hd,tl=Tl},body=B,anno=#l{i=I,vdb=Vdb}}, + V, Tf, Vf, Bef, St0) -> + Es = [Hd,Tl], {Eis,Int,St1} = select_extract_cons(V, Es, I, Vdb, Bef, St0), {Bis,Aft,St2} = match_cg(B, Vf, Int, St1), {[{test,is_nonempty_list,{f,Tf},[fetch_var(V, Bef)]}] ++ Eis ++ Bis,Aft,St2}. -select_nil(#l{ke={val_clause,nil,B}}, V, Tf, Vf, Bef, St0) -> +select_nil(#k_val_clause{val=#k_nil{},body=B}, V, Tf, Vf, Bef, St0) -> {Bis,Aft,St1} = match_cg(B, Vf, Bef, St0), {[{test,is_nil,{f,Tf},[fetch_var(V, Bef)]}] ++ Bis,Aft,St1}. -select_binary(#l{ke={val_clause,{binary,{var,V}},B},i=I,vdb=Vdb}, - V, Tf, Vf, Bef, St0) -> +select_binary(#k_val_clause{val=#k_binary{segs=#k_var{name=V}},body=B, + anno=#l{i=I,vdb=Vdb}}, V, Tf, Vf, Bef, St0) -> + #cg{ctx=OldCtx} = St0, Int0 = clear_dead(Bef#sr{reg=Bef#sr.reg}, I, Vdb), - {Bis0,Aft,St1} = match_cg(B, Vf, Int0, St0), + {Bis0,Aft,St1} = match_cg(B, Vf, Int0, St0#cg{ctx=V}), CtxReg = fetch_var(V, Int0), Live = max_reg(Bef#sr.reg), Bis1 = [{test,bs_start_match2,{f,Tf},Live,[CtxReg,V],CtxReg}, {bs_save2,CtxReg,{V,V}}|Bis0], Bis = finish_select_binary(Bis1), - {Bis,Aft,St1}; -select_binary(#l{ke={val_clause,{binary,{var,Ivar}},B},i=I,vdb=Vdb}, - V, Tf, Vf, Bef, St0) -> + {Bis,Aft,St1#cg{ctx=OldCtx}}; +select_binary(#k_val_clause{val=#k_binary{segs=#k_var{name=Ivar}},body=B, + anno=#l{i=I,vdb=Vdb}}, V, Tf, Vf, Bef, St0) -> + #cg{ctx=OldCtx} = St0, Regs = put_reg(Ivar, Bef#sr.reg), Int0 = clear_dead(Bef#sr{reg=Regs}, I, Vdb), - {Bis0,Aft,St1} = match_cg(B, Vf, Int0, St0), + {Bis0,Aft,St1} = match_cg(B, Vf, Int0, St0#cg{ctx=Ivar}), CtxReg = fetch_var(Ivar, Int0), Live = max_reg(Bef#sr.reg), Bis1 = [{test,bs_start_match2,{f,Tf},Live,[fetch_var(V, Bef),Ivar],CtxReg}, {bs_save2,CtxReg,{Ivar,Ivar}}|Bis0], Bis = finish_select_binary(Bis1), - {Bis,Aft,St1}. + {Bis,Aft,St1#cg{ctx=OldCtx}}. finish_select_binary([{bs_save2,R,Point}=I,{bs_restore2,R,Point}|Is]) -> [I|finish_select_binary(Is)]; @@ -757,9 +1015,16 @@ select_bin_segs(Scs, Ivar, Tf, Bef, St) -> select_bin_seg(S, Ivar, Fail, Bef, Sta) end, Tf, St, Scs). -select_bin_seg(#l{ke={val_clause,{bin_seg,Ctx,Size,U,T,Fs0,Es},B},i=I,vdb=Vdb,a=A}, - Ivar, Fail, Bef, St0) -> +select_bin_seg(#k_val_clause{val=#k_bin_seg{size=Size,unit=U,type=T, + seg=Seg,flags=Fs0,next=Next}, + body=B, + anno=#l{i=I,vdb=Vdb,a=A}}, Ivar, Fail, Bef, St0) -> + Ctx = St0#cg.ctx, Fs = [{anno,A}|Fs0], + Es = case Next of + [] -> [Seg]; + _ -> [Seg,Next] + end, {Mis,Int,St1} = select_extract_bin(Es, Size, U, T, Fs, Fail, I, Vdb, Bef, Ctx, B, St0), {Bis,Aft,St2} = match_cg(B, Fail, Int, St1), @@ -772,9 +1037,12 @@ select_bin_seg(#l{ke={val_clause,{bin_seg,Ctx,Size,U,T,Fs0,Es},B},i=I,vdb=Vdb,a= [{bs_restore2,CtxReg,{Ctx,Ivar}}|Mis++Bis] end, {Is,Aft,St2}; -select_bin_seg(#l{ke={val_clause,{bin_int,Ctx,Sz,U,Fs,Val,Es},B},i=I,vdb=Vdb}, - Ivar, Fail, Bef, St0) -> - {Mis,Int,St1} = select_extract_int(Es, Val, Sz, U, Fs, Fail, +select_bin_seg(#k_val_clause{val=#k_bin_int{size=Sz,unit=U,flags=Fs, + val=Val,next=Next}, + body=B, + anno=#l{i=I,vdb=Vdb}}, Ivar, Fail, Bef, St0) -> + Ctx = St0#cg.ctx, + {Mis,Int,St1} = select_extract_int(Next, Val, Sz, U, Fs, Fail, I, Vdb, Bef, Ctx, St0), {Bis,Aft,St2} = match_cg(B, Fail, Int, St1), CtxReg = fetch_var(Ctx, Bef), @@ -795,7 +1063,7 @@ select_bin_seg(#l{ke={val_clause,{bin_int,Ctx,Sz,U,Fs,Val,Es},B},i=I,vdb=Vdb}, end, {[{bs_restore2,CtxReg,{Ctx,Ivar}}|Is],Aft,St2}. -select_extract_int([{var,Tl}], Val, {integer,Sz}, U, Fs, Vf, +select_extract_int(#k_var{name=Tl}, Val, #k_int{val=Sz}, U, Fs, Vf, I, Vdb, Bef, Ctx, St) -> Bits = U*Sz, Bin = case member(big, Fs) of @@ -816,7 +1084,7 @@ select_extract_int([{var,Tl}], Val, {integer,Sz}, U, Fs, Vf, end, {Is,clear_dead(Bef, I, Vdb),St}. -select_extract_bin([{var,Hd},{var,Tl}], Size0, Unit, Type, Flags, Vf, +select_extract_bin([#k_var{name=Hd},#k_var{name=Tl}], Size0, Unit, Type, Flags, Vf, I, Vdb, Bef, Ctx, _Body, St) -> SizeReg = get_bin_size_reg(Size0, Bef), {Es,Aft} = @@ -839,11 +1107,11 @@ select_extract_bin([{var,Hd},{var,Tl}], Size0, Unit, Type, Flags, Vf, {bs_save2,CtxReg,{Ctx,Tl}}],Int1} end, {Es,clear_dead(Aft, I, Vdb),St}; -select_extract_bin([{var,Hd}], Size, Unit, binary, Flags, Vf, +select_extract_bin([#k_var{name=Hd}], Size, Unit, binary, Flags, Vf, I, Vdb, Bef, Ctx, Body, St) -> %% Match the last segment of a binary. We KNOW that the size %% must be 'all'. - Size = {atom,all}, %Assertion. + #k_atom{val=all} = Size, %Assertion. {Es,Aft} = case vdb_find(Hd, Vdb) of {_,_,Lhd} when Lhd =< I -> @@ -868,7 +1136,7 @@ select_extract_bin([{var,Hd}], Size, Unit, binary, Flags, Vf, Name = bs_get_binary2, Live = max_reg(Bef#sr.reg), {[{test,Name,{f,Vf},Live, - [CtxReg,Size,Unit,{field_flags,Flags}],Rhd}], + [CtxReg,atomic(Size),Unit,{field_flags,Flags}],Rhd}], Int1}; true -> %% Since the matching context will not be used again, @@ -883,43 +1151,42 @@ select_extract_bin([{var,Hd}], Size, Unit, binary, Flags, Vf, Name = bs_get_binary2, Live = max_reg(Int1#sr.reg), {[{test,Name,{f,Vf},Live, - [CtxReg,Size,Unit,{field_flags,Flags}],CtxReg}], + [CtxReg,atomic(Size),Unit,{field_flags,Flags}],CtxReg}], Int1} end end, {Es,clear_dead(Aft, I, Vdb),St}. %% is_context_unused(Ke) -> true | false -%% Simple heurististic to determine whether the code that follows will -%% use the current matching context again. (The information of liveness -%% calculcated by v3_life is too conservative to be useful for this purpose.) -%% 'true' means that the code that follows will definitely not use the context -%% again (because it is a block, not guard or matching code); 'false' that we -%% are not sure (there could be more matching). - -is_context_unused(#l{ke=Ke}) -> - is_context_unused(Ke); -is_context_unused({alt,_First,Then}) -> - %% {alt,First,Then} can be used for different purposes. If the Then part +%% Simple heurististic to determine whether the code that follows +%% will use the current matching context again. (The liveness +%% information is too conservative to be useful for this purpose.) +%% 'true' means that the code that follows will definitely not use +%% the context again (because it is a block, not guard or matching +%% code); 'false' that we are not sure (there could be more +%% matching). + +is_context_unused(#k_alt{then=Then}) -> + %% #k_alt{} can be used for different purposes. If the Then part %% is a block, it means that matching has finished and is used for a guard %% to choose between the matched clauses. is_context_unused(Then); -is_context_unused({block,_}) -> +is_context_unused(#cg_block{}) -> true; is_context_unused(_) -> false. -select_bin_end(#l{ke={val_clause,{bin_end,Ctx},B}}, - Ivar, Tf, Bef, St0) -> +select_bin_end(#k_val_clause{val=#k_bin_end{},body=B}, Ivar, Tf, Bef, St0) -> + Ctx = St0#cg.ctx, {Bis,Aft,St2} = match_cg(B, Tf, Bef, St0), CtxReg = fetch_var(Ctx, Bef), {[{bs_restore2,CtxReg,{Ctx,Ivar}}, {test,bs_test_tail2,{f,Tf},[CtxReg,0]}|Bis],Aft,St2}. -get_bin_size_reg({var,V}, Bef) -> +get_bin_size_reg(#k_var{name=V}, Bef) -> fetch_var(V, Bef); get_bin_size_reg(Literal, _Bef) -> - Literal. + atomic(Literal). build_bs_instr(Type, Vf, CtxReg, Live, SizeReg, Unit, Flags, Rhd) -> {Format,Name} = case Type of @@ -953,11 +1220,18 @@ build_skip_instr(Type, Vf, CtxReg, Live, SizeReg, Unit, Flags) -> {test,Name,{f,Vf},[CtxReg,Live,{field_flags,Flags}]} end. -select_val(#l{ke={val_clause,{tuple,Es},B},i=I,vdb=Vdb}, V, Vf, Bef, St0) -> +select_val(#k_val_clause{val=#k_tuple{es=Es},body=B,anno=#l{i=I,vdb=Vdb}}, + V, Vf, Bef, St0) -> {Eis,Int,St1} = select_extract_tuple(V, Es, I, Vdb, Bef, St0), {Bis,Aft,St2} = match_cg(B, Vf, Int, St1), {length(Es),Eis ++ Bis,Aft,St2}; -select_val(#l{ke={val_clause,{_,Val},B}}, _V, Vf, Bef, St0) -> +select_val(#k_val_clause{val=Val0,body=B}, _V, Vf, Bef, St0) -> + Val = case Val0 of + #k_atom{val=Lit} -> Lit; + #k_float{val=Lit} -> Lit; + #k_int{val=Lit} -> Lit; + #k_literal{val=Lit} -> Lit + end, {Bis,Aft,St1} = match_cg(B, Vf, Bef, St0), {Val,Bis,Aft,St1}. @@ -966,7 +1240,7 @@ select_val(#l{ke={val_clause,{_,Val},B}}, _V, Vf, Bef, St0) -> %% Extract tuple elements, but only if they do not immediately die. select_extract_tuple(Src, Vs, I, Vdb, Bef, St) -> - F = fun ({var,V}, {Int0,Elem}) -> + F = fun (#k_var{name=V}, {Int0,Elem}) -> case vdb_find(V, Vdb) of {V,_,L} when L =< I -> {[], {Int0,Elem+1}}; _Other -> @@ -983,9 +1257,10 @@ select_extract_tuple(Src, Vs, I, Vdb, Bef, St) -> select_map(Scs, V, Tf, Vf, Bef, St0) -> Reg = fetch_var(V, Bef), {Is,Aft,St1} = - match_fmf(fun(#l{ke={val_clause,{map,exact,_,Es},B},i=I,vdb=Vdb}, Fail, St1) -> - select_map_val(V, Es, B, Fail, I, Vdb, Bef, St1) - end, Vf, St0, Scs), + match_fmf(fun(#k_val_clause{val=#k_map{op=exact,es=Es}, + body=B,anno=#l{i=I,vdb=Vdb}}, Fail, St1) -> + select_map_val(V, Es, B, Fail, I, Vdb, Bef, St1) + end, Vf, St0, Scs), {[{test,is_map,{f,Tf},[Reg]}|Is],Aft,St1}. select_map_val(V, Es, B, Fail, I, Vdb, Bef, St0) -> @@ -1002,29 +1277,32 @@ select_extract_map(Src, Vs, Fail, I, Vdb, Bef, St) -> %% Assume keys are term-sorted Rsrc = fetch_var(Src, Bef), - {{HasKs,GetVs,HasVarKs,GetVarVs},Aft} = lists:foldr(fun - ({map_pair,{var,K},{var,V}},{{HasKsi,GetVsi,HasVarVsi,GetVarVsi},Int0}) -> - case vdb_find(V, Vdb) of - {V,_,L} when L =< I -> - RK = fetch_var(K,Int0), - {{HasKsi,GetVsi,[RK|HasVarVsi],GetVarVsi},Int0}; - _Other -> - Reg1 = put_reg(V, Int0#sr.reg), - Int1 = Int0#sr{reg=Reg1}, - RK = fetch_var(K,Int0), - RV = fetch_reg(V,Reg1), - {{HasKsi,GetVsi,HasVarVsi,[[RK,RV]|GetVarVsi]},Int1} - end; - ({map_pair,Key,{var,V}},{{HasKsi,GetVsi,HasVarVsi,GetVarVsi},Int0}) -> - case vdb_find(V, Vdb) of - {V,_,L} when L =< I -> - {{[Key|HasKsi],GetVsi,HasVarVsi,GetVarVsi},Int0}; - _Other -> - Reg1 = put_reg(V, Int0#sr.reg), - Int1 = Int0#sr{reg=Reg1}, - {{HasKsi,[Key,fetch_reg(V, Reg1)|GetVsi],HasVarVsi,GetVarVsi},Int1} - end - end, {{[],[],[],[]},Bef}, Vs), + {{HasKs,GetVs,HasVarKs,GetVarVs},Aft} = + foldr(fun(#k_map_pair{key=#k_var{name=K},val=#k_var{name=V}}, + {{HasKsi,GetVsi,HasVarVsi,GetVarVsi},Int0}) -> + case vdb_find(V, Vdb) of + {V,_,L} when L =< I -> + RK = fetch_var(K,Int0), + {{HasKsi,GetVsi,[RK|HasVarVsi],GetVarVsi},Int0}; + _Other -> + Reg1 = put_reg(V, Int0#sr.reg), + Int1 = Int0#sr{reg=Reg1}, + RK = fetch_var(K,Int0), + RV = fetch_reg(V,Reg1), + {{HasKsi,GetVsi,HasVarVsi,[[RK,RV]|GetVarVsi]},Int1} + end; + (#k_map_pair{key=Key,val=#k_var{name=V}}, + {{HasKsi,GetVsi,HasVarVsi,GetVarVsi},Int0}) -> + case vdb_find(V, Vdb) of + {V,_,L} when L =< I -> + {{[atomic(Key)|HasKsi],GetVsi,HasVarVsi,GetVarVsi},Int0}; + _Other -> + Reg1 = put_reg(V, Int0#sr.reg), + Int1 = Int0#sr{reg=Reg1}, + {{HasKsi,[atomic(Key),fetch_reg(V, Reg1)|GetVsi], + HasVarVsi,GetVarVsi},Int1} + end + end, {{[],[],[],[]},Bef}, Vs), Code = [{test,has_map_fields,{f,Fail},Rsrc,{list,HasKs}} || HasKs =/= []] ++ [{test,has_map_fields,{f,Fail},Rsrc,{list,[K]}} || K <- HasVarKs] ++ @@ -1033,7 +1311,7 @@ select_extract_map(Src, Vs, Fail, I, Vdb, Bef, St) -> {Code, Aft, St}. -select_extract_cons(Src, [{var,Hd}, {var,Tl}], I, Vdb, Bef, St) -> +select_extract_cons(Src, [#k_var{name=Hd}, #k_var{name=Tl}], I, Vdb, Bef, St) -> {Es,Aft} = case {vdb_find(Hd, Vdb), vdb_find(Tl, Vdb)} of {{_,_,Lhd}, {_,_,Ltl}} when Lhd =< I, Ltl =< I -> %% Both head and tail are dead. No need to generate @@ -1056,7 +1334,7 @@ select_extract_cons(Src, [{var,Hd}, {var,Tl}], I, Vdb, Bef, St) -> {Es,Aft,St}. -guard_clause_cg(#l{ke={guard_clause,G,B},vdb=Vdb}, Fail, Bef, St0) -> +guard_clause_cg(#k_guard_clause{anno=#l{vdb=Vdb},guard=G,body=B}, Fail, Bef, St0) -> {Gis,Int,St1} = guard_cg(G, Fail, Vdb, Bef, St0), {Bis,Aft,St} = match_cg(B, Fail, Int, St1), {Gis ++ Bis,Aft,St}. @@ -1069,11 +1347,13 @@ guard_clause_cg(#l{ke={guard_clause,G,B},vdb=Vdb}, Fail, Bef, St0) -> %% the correct exit point. Primops and tests all go to the next %% instruction on success or jump to a failure label. -guard_cg(#l{ke={protected,Ts,Rs},i=I,vdb=Pdb}, Fail, _Vdb, Bef, St) -> +guard_cg(#k_protected{arg=Ts,ret=Rs,anno=#l{i=I,vdb=Pdb}}, Fail, _Vdb, Bef, St) -> protected_cg(Ts, Rs, Fail, I, Pdb, Bef, St); -guard_cg(#l{ke={block,Ts},i=I,vdb=Bdb}, Fail, _Vdb, Bef, St) -> +guard_cg(#cg_block{es=Ts,anno=#l{i=I,vdb=Bdb}}, Fail, _Vdb, Bef, St) -> guard_cg_list(Ts, Fail, I, Bdb, Bef, St); -guard_cg(#l{ke={test,Test,As,Inverted},i=I,vdb=_Tdb}, Fail, Vdb, Bef, St0) -> +guard_cg(#k_test{anno=#l{i=I},op=Test0,args=As,inverted=Inverted}, + Fail, Vdb, Bef, St0) -> + #k_remote{mod=#k_atom{val=erlang},name=#k_atom{val=Test}} = Test0, case Inverted of false -> test_cg(Test, As, Fail, I, Vdb, Bef, St0); @@ -1107,7 +1387,7 @@ protected_cg(Ts, Rs, _Fail, I, Vdb, Bef, St0) -> St2#cg{bfail=Pfail}), %%ok = io:fwrite("cg ~w: ~p~n", [?LINE,{Rs,I,Vdb,Aft}]), %% Set return values to false. - Mis = [{move,{atom,false},fetch_var(V,Aft)}||{var,V} <- Rs], + Mis = [{move,{atom,false},fetch_var(V,Aft)}||#k_var{name=V} <- Rs], {Tis ++ [{jump,{f,Psucc}}, {label,Pfail}] ++ Mis ++ [{label,Psucc}], Aft,St3#cg{bfail=St0#cg.bfail}}. @@ -1132,7 +1412,7 @@ test_cg(is_map, [A], Fail, I, Vdb, Bef, St) -> Arg = cg_reg_arg_prefer_y(A, Bef), Aft = clear_dead(Bef, I, Vdb), {[{test,is_map,{f,Fail},[Arg]}],Aft,St}; -test_cg(is_boolean, [{atom,Val}], Fail, I, Vdb, Bef, St) -> +test_cg(is_boolean, [#k_atom{val=Val}], Fail, I, Vdb, Bef, St) -> Aft = clear_dead(Bef, I, Vdb), Is = case is_boolean(Val) of true -> []; @@ -1178,7 +1458,7 @@ match_fmf(F, LastFail, St0, [H|T]) -> %% frame size. Finally the actual call is made. Call then needs the %% return values filled in. -call_cg({var,_V} = Var, As, Rs, Le, Vdb, Bef, St0) -> +call_cg(#k_var{}=Var, As, Rs, Le, Vdb, Bef, St0) -> {Sis,Int} = cg_setup_call(As++[Var], Bef, Le#l.i, Vdb), %% Put return values in registers. Reg = load_vars(Rs, clear_regs(Int#sr.reg)), @@ -1187,9 +1467,8 @@ call_cg({var,_V} = Var, As, Rs, Le, Vdb, Bef, St0) -> {Frees,Aft} = free_dead(clear_dead(Int#sr{reg=Reg}, Le#l.i, Vdb)), {Sis ++ Frees ++ [line(Le),{call_fun,Arity}],Aft, need_stack_frame(St0)}; -call_cg({remote,Mod,Name}, As, Rs, Le, Vdb, Bef, St0) - when element(1, Mod) =:= var; - element(1, Name) =:= var -> +call_cg(#k_remote{mod=Mod,name=Name}, As, Rs, Le, Vdb, Bef, St0) + when is_record(Mod, k_var); is_record(Name, k_var) -> {Sis,Int} = cg_setup_call(As++[Mod,Name], Bef, Le#l.i, Vdb), %% Put return values in registers. Reg = load_vars(Rs, clear_regs(Int#sr.reg)), @@ -1207,8 +1486,9 @@ call_cg(Func, As, Rs, Le, Vdb, Bef, St0) -> %% %% move {atom,ok} DestReg %% jump FailureLabel - {remote,{atom,erlang},{atom,error}} = Func, %Assertion. - [{var,DestVar}] = Rs, + #k_remote{mod=#k_atom{val=erlang}, + name=#k_atom{val=error}} = Func, %Assertion. + [#k_var{name=DestVar}] = Rs, Int0 = clear_dead(Bef, Le#l.i, Vdb), Reg = put_reg(DestVar, Int0#sr.reg), Int = Int0#sr{reg=Reg}, @@ -1227,11 +1507,11 @@ call_cg(Func, As, Rs, Le, Vdb, Bef, St0) -> {Sis ++ Frees ++ [line(Le)|Call],Aft,St1} end. -build_call({remote,{atom,erlang},{atom,'!'}}, 2, St0) -> +build_call(#k_remote{mod=#k_atom{val=erlang},name=#k_atom{val='!'}}, 2, St0) -> {[send],need_stack_frame(St0)}; -build_call({remote,{atom,Mod},{atom,Name}}, Arity, St0) -> +build_call(#k_remote{mod=#k_atom{val=Mod},name=#k_atom{val=Name}}, Arity, St0) -> {[{call_ext,Arity,{extfunc,Mod,Name,Arity}}],need_stack_frame(St0)}; -build_call(Name, Arity, St0) when is_atom(Name) -> +build_call(#k_local{name=Name}, Arity, St0) when is_atom(Name) -> {Lbl,St1} = local_func_label(Name, Arity, need_stack_frame(St0)), {[{call,Arity,{f,Lbl}}],St1}. @@ -1247,16 +1527,15 @@ free_dead([Any|Stk], Y, Instr, StkAcc) -> free_dead(Stk, Y+1, Instr, [Any|StkAcc]); free_dead([], _, Instr, StkAcc) -> {Instr,reverse(StkAcc)}. -enter_cg({var,_V} = Var, As, Le, Vdb, Bef, St0) -> +enter_cg(#k_var{} = Var, As, Le, Vdb, Bef, St0) -> {Sis,Int} = cg_setup_call(As++[Var], Bef, Le#l.i, Vdb), %% Build complete code and final stack/register state. Arity = length(As), {Sis ++ [line(Le),{call_fun,Arity},return], clear_dead(Int#sr{reg=clear_regs(Int#sr.reg)}, Le#l.i, Vdb), need_stack_frame(St0)}; -enter_cg({remote,Mod,Name}, As, Le, Vdb, Bef, St0) - when element(1, Mod) =:= var; - element(1, Name) =:= var -> +enter_cg(#k_remote{mod=Mod,name=Name}, As, Le, Vdb, Bef, St0) + when is_record(Mod, k_var); is_record(Name, k_var) -> {Sis,Int} = cg_setup_call(As++[Mod,Name], Bef, Le#l.i, Vdb), %% Build complete code and final stack/register state. Arity = length(As), @@ -1274,19 +1553,19 @@ enter_cg(Func, As, Le, Vdb, Bef, St0) -> clear_dead(Int#sr{reg=clear_regs(Int#sr.reg)}, Le#l.i, Vdb), St1}. -build_enter({remote,{atom,erlang},{atom,'!'}}, 2, St0) -> +build_enter(#k_remote{mod=#k_atom{val=erlang},name=#k_atom{val='!'}}, 2, St0) -> {[send,return],need_stack_frame(St0)}; -build_enter({remote,{atom,Mod},{atom,Name}}, Arity, St0) -> +build_enter(#k_remote{mod=#k_atom{val=Mod},name=#k_atom{val=Name}}, Arity, St0) -> St1 = case trap_bif(Mod, Name, Arity) of true -> need_stack_frame(St0); false -> St0 end, {[{call_ext_only,Arity,{extfunc,Mod,Name,Arity}}],St1}; -build_enter(Name, Arity, St0) when is_atom(Name) -> +build_enter(#k_local{name=Name}, Arity, St0) when is_atom(Name) -> {Lbl,St1} = local_func_label(Name, Arity, St0), {[{call_only,Arity,{f,Lbl}}],St1}. -enter_line({remote,{atom,Mod},{atom,Name}}, Arity, Le) -> +enter_line(#k_remote{mod=#k_atom{val=Mod},name=#k_atom{val=Name}}, Arity, Le) -> case erl_bifs:is_safe(Mod, Name, Arity) of false -> %% Tail-recursive call, possibly to a BIF. @@ -1334,6 +1613,22 @@ trap_bif(erlang, group_leader, 2) -> true; trap_bif(erlang, exit, 2) -> true; trap_bif(_, _, _) -> false. +%% bif_cg(#k_bif{}, Le, Vdb, StackReg, State) -> +%% {[Ainstr],StackReg,State}. +%% Generate code a BIF. + +bif_cg(#k_bif{op=#k_internal{name=Name},args=As,ret=Rs}, Le, Vdb, Bef, St) -> + internal_cg(Name, As, Rs, Le, Vdb, Bef, St); +bif_cg(#k_bif{op=#k_remote{mod=#k_atom{val=erlang},name=#k_atom{val=Name}}, + args=As,ret=Rs}, Le, Vdb, Bef, St) -> + Ar = length(As), + case is_gc_bif(Name, Ar) of + false -> + bif_cg(Name, As, Rs, Le, Vdb, Bef, St); + true -> + gc_bif_cg(Name, As, Rs, Le, Vdb, Bef, St) + end. + %% internal_cg(Bif, [Arg], [Ret], Le, Vdb, StackReg, State) -> %% {[Ainstr],StackReg,State}. @@ -1352,8 +1647,8 @@ internal_cg(dsetelement, [Index0,Tuple0,New0], _Rs, Le, Vdb, Bef, St0) -> clear_dead(Bef, Le#l.i, Vdb), St0}; internal_cg(make_fun, [Func0,Arity0|As], Rs, Le, Vdb, Bef, St0) -> %% This behaves more like a function call. - {atom,Func} = Func0, - {integer,Arity} = Arity0, + #k_atom{val=Func} = Func0, + #k_int{val=Arity} = Arity0, {Sis,Int} = cg_setup_call(As, Bef, Le#l.i, Vdb), Reg = load_vars(Rs, clear_regs(Int#sr.reg)), {FuncLbl,St1} = local_func_label(Func, Arity, St0), @@ -1373,7 +1668,7 @@ internal_cg(raise, As, Rs, Le, Vdb, Bef, St) -> %% bif_cg(Bif, [Arg], [Ret], Le, Vdb, StackReg, State) -> %% {[Ainstr],StackReg,State}. -bif_cg(Bif, As, [{var,V}], Le, Vdb, Bef, St0) -> +bif_cg(Bif, As, [#k_var{name=V}], Le, Vdb, Bef, St0) -> Ars = cg_reg_args(As, Bef), %% If we are inside a catch and in a body (not in guard) and the @@ -1411,7 +1706,7 @@ bif_cg(Bif, As, [{var,V}], Le, Vdb, Bef, St0) -> %% gc_bif_cg(Bif, [Arg], [Ret], Le, Vdb, StackReg, State) -> %% {[Ainstr],StackReg,State}. -gc_bif_cg(Bif, As, [{var,V}], Le, Vdb, Bef, St0) -> +gc_bif_cg(Bif, As, [#k_var{name=V}], Le, Vdb, Bef, St0) -> Ars = cg_reg_args(As, Bef), %% If we are inside a catch and in a body (not in guard) and the @@ -1457,7 +1752,7 @@ recv_loop_cg(Te, Rvar, Rm, Tes, Rs, Le, Vdb, Bef, St0) -> %% cg_recv_mesg( ) -> {[Ainstr],Aft,St}. -cg_recv_mesg({var,R}, Rm, Tl, Bef, St0) -> +cg_recv_mesg(#k_var{name=R}, Rm, Tl, Bef, St0) -> Int0 = Bef#sr{reg=put_reg(R, Bef#sr.reg)}, Ret = fetch_reg(R, Int0#sr.reg), %% Int1 = clear_dead(Int0, I, Rm#l.vdb), @@ -1467,22 +1762,22 @@ cg_recv_mesg({var,R}, Rm, Tl, Bef, St0) -> %% cg_recv_wait(Te, Tes, I, Vdb, Int2, St3) -> {[Ainstr],Aft,St}. -cg_recv_wait({atom,infinity}, Tes, I, Bef, St0) -> +cg_recv_wait(#k_atom{val=infinity}, #cg_block{anno=Le,es=Tes}, I, Bef, St0) -> %% We know that the 'after' body will never be executed. %% But to keep the stack and register information up to date, %% we will generate the code for the 'after' body, and then discard it. - Int1 = clear_dead(Bef, I, Tes#l.vdb), - {_,Int2,St1} = cg_block(Tes#l.ke, Tes#l.i, Tes#l.vdb, - Int1#sr{reg=clear_regs(Int1#sr.reg)}, St0), + Int1 = clear_dead(Bef, I, Le#l.vdb), + {_,Int2,St1} = cg_block(Tes, Le#l.i, Le#l.vdb, + Int1#sr{reg=clear_regs(Int1#sr.reg)}, St0), {[{wait,{f,St1#cg.recv}}],Int2,St1}; -cg_recv_wait({integer,0}, Tes, _I, Bef, St0) -> - {Tis,Int,St1} = cg_block(Tes#l.ke, Tes#l.i, Tes#l.vdb, Bef, St0), +cg_recv_wait(#k_int{val=0}, #cg_block{anno=Le,es=Tes}, _I, Bef, St0) -> + {Tis,Int,St1} = cg_block(Tes, Le#l.i, Le#l.vdb, Bef, St0), {[timeout|Tis],Int,St1}; -cg_recv_wait(Te, Tes, I, Bef, St0) -> +cg_recv_wait(Te, #cg_block{anno=Le,es=Tes}, I, Bef, St0) -> Reg = cg_reg_arg(Te, Bef), %% Must have empty registers here! Bug if anything in registers. - Int0 = clear_dead(Bef, I, Tes#l.vdb), - {Tis,Int,St1} = cg_block(Tes#l.ke, Tes#l.i, Tes#l.vdb, + Int0 = clear_dead(Bef, I, Le#l.vdb), + {Tis,Int,St1} = cg_block(Tes, Le#l.i, Le#l.vdb, Int0#sr{reg=clear_regs(Int0#sr.reg)}, St0), {[{wait_timeout,{f,St1#cg.recv},Reg},timeout] ++ Tis,Int,St1}. @@ -1500,7 +1795,7 @@ try_cg(Ta, Vs, Tb, Evs, Th, Rs, Le, Vdb, Bef, St0) -> {B,St1} = new_label(St0), %Body label {H,St2} = new_label(St1), %Handler label {E,St3} = new_label(St2), %End label - TryTag = Ta#l.i, + #l{i=TryTag} = get_kanno(Ta), Int1 = Bef#sr{stk=put_catch(TryTag, Bef#sr.stk)}, TryReg = fetch_stack({catch_tag,TryTag}, Int1#sr.stk), {Ais,Int2,St4} = cg(Ta, Vdb, Int1, St3#cg{break=B,in_catch=true}), @@ -1520,7 +1815,7 @@ try_cg(Ta, Vs, Tb, Evs, Th, Rs, Le, Vdb, Bef, St0) -> try_enter_cg(Ta, Vs, Tb, Evs, Th, Le, Vdb, Bef, St0) -> {B,St1} = new_label(St0), %Body label {H,St2} = new_label(St1), %Handler label - TryTag = Ta#l.i, + #l{i=TryTag} = get_kanno(Ta), Int1 = Bef#sr{stk=put_catch(TryTag, Bef#sr.stk)}, TryReg = fetch_stack({catch_tag,TryTag}, Int1#sr.stk), {Ais,Int2,St3} = cg(Ta, Vdb, Int1, St2#cg{break=B,in_catch=true}), @@ -1538,7 +1833,7 @@ try_enter_cg(Ta, Vs, Tb, Evs, Th, Le, Vdb, Bef, St0) -> %% catch_cg(CatchBlock, Ret, Le, Vdb, Bef, St) -> {[Ainstr],Aft,St}. -catch_cg(C, {var,R}, Le, Vdb, Bef, St0) -> +catch_cg(#cg_block{es=C}, #k_var{name=R}, Le, Vdb, Bef, St0) -> {B,St1} = new_label(St0), CatchTag = Le#l.i, Int1 = Bef#sr{stk=put_catch(CatchTag, Bef#sr.stk)}, @@ -1552,8 +1847,8 @@ catch_cg(C, {var,R}, Le, Vdb, Bef, St0) -> clear_dead(Aft, Le#l.i, Vdb), St2#cg{break=St1#cg.break,in_catch=St1#cg.in_catch}}. -%% set_cg([Var], Constr, Le, Vdb, Bef, St) -> {[Ainstr],Aft,St}. -%% We have to be careful how a 'set' works. First the structure is +%% put_cg([Var], Constr, Le, Vdb, Bef, St) -> {[Ainstr],Aft,St}. +%% We have to be careful how a 'put' works. First the structure is %% built, then it is filled and finally things can be cleared. The %% annotation must reflect this and make sure that the return %% variable is allocated first. @@ -1561,13 +1856,14 @@ catch_cg(C, {var,R}, Le, Vdb, Bef, St0) -> %% put_list and put_map are atomic instructions, both of %% which can safely resuse one of the source registers as target. -set_cg([{var,R}], {cons,Es}, Le, Vdb, Bef, St) -> - [S1,S2] = cg_reg_args(Es, Bef), +put_cg([#k_var{name=R}], #k_cons{hd=Hd,tl=Tl}, Le, Vdb, Bef, St) -> + [S1,S2] = cg_reg_args([Hd,Tl], Bef), Int0 = clear_dead(Bef, Le#l.i, Vdb), Int1 = Int0#sr{reg=put_reg(R, Int0#sr.reg)}, Ret = fetch_reg(R, Int1#sr.reg), {[{put_list,S1,S2,Ret}], Int1, St}; -set_cg([{var,R}], {binary,Segs}, Le, Vdb, Bef, #cg{bfail=Bfail}=St) -> +put_cg([#k_var{name=R}], #k_binary{segs=Segs}, Le, Vdb, Bef, + #cg{bfail=Bfail}=St) -> %% At run-time, binaries are constructed in three stages: %% 1) First the size of the binary is calculated. %% 2) Then the binary is allocated. @@ -1595,7 +1891,9 @@ set_cg([{var,R}], {binary,Segs}, Le, Vdb, Bef, #cg{bfail=Bfail}=St) -> {Sis++Code,Aft,St}; %% Map: single variable key. -set_cg([{var,R}], {map,Op,Map,[{map_pair,{var,_}=K,V}]}, Le, Vdb, Bef, St0) -> +put_cg([#k_var{name=R}], #k_map{op=Op,var=Map, + es=[#k_map_pair{key=#k_var{}=K,val=V}]}, + Le, Vdb, Bef, St0) -> {Sis,Int0} = maybe_adjust_stack(Bef, Le#l.i, Le#l.i+1, Vdb, St0), SrcReg = cg_reg_arg_prefer_y(Map, Int0), @@ -1610,22 +1908,23 @@ set_cg([{var,R}], {map,Op,Map,[{map_pair,{var,_}=K,V}]}, Le, Vdb, Bef, St0) -> Aft = Aft0#sr{reg=put_reg(R, Aft0#sr.reg)}, Target = fetch_reg(R, Aft#sr.reg), - {Is,St1} = set_cg_map(Line, Op, SrcReg, Target, Live, List, St0), + {Is,St1} = put_cg_map(Line, Op, SrcReg, Target, Live, List, St0), {Sis++Is,Aft,St1}; %% Map: (possibly) multiple literal keys. -set_cg([{var,R}], {map,Op,Map,Es}, Le, Vdb, Bef, St0) -> +put_cg([#k_var{name=R}], #k_map{op=Op,var=Map,es=Es}, Le, Vdb, Bef, St0) -> %% assert key literals - [] = [Var||{map_pair,{var,_}=Var,_} <- Es], + [] = [Var || #k_map_pair{key=#k_var{}=Var} <- Es], {Sis,Int0} = maybe_adjust_stack(Bef, Le#l.i, Le#l.i+1, Vdb, St0), SrcReg = cg_reg_arg_prefer_y(Map, Int0), Line = line(Le#l.a), %% fetch registers for values to be put into the map - Pairs = [{K,V} || {_,K,V} <- Es], - List = flatmap(fun({K,V}) -> [K,cg_reg_arg(V,Int0)] end, Pairs), + List = flatmap(fun(#k_map_pair{key=K,val=V}) -> + [atomic(K),cg_reg_arg(V, Int0)] + end, Es), Live = max_reg(Bef#sr.reg), @@ -1634,16 +1933,16 @@ set_cg([{var,R}], {map,Op,Map,Es}, Le, Vdb, Bef, St0) -> Aft = Aft0#sr{reg=put_reg(R, Aft0#sr.reg)}, Target = fetch_reg(R, Aft#sr.reg), - {Is,St1} = set_cg_map(Line, Op, SrcReg, Target, Live, List, St0), + {Is,St1} = put_cg_map(Line, Op, SrcReg, Target, Live, List, St0), {Sis++Is,Aft,St1}; %% Everything else. -set_cg([{var,R}], Con, Le, Vdb, Bef, St) -> +put_cg([#k_var{name=R}], Con, Le, Vdb, Bef, St) -> %% Find a place for the return register first. Int = Bef#sr{reg=put_reg(R, Bef#sr.reg)}, Ret = fetch_reg(R, Int#sr.reg), Ais = case Con of - {tuple,Es} -> + #k_tuple{es=Es} -> [{put_tuple,length(Es),Ret}] ++ cg_build_args(Es, Bef); Other -> [{move,cg_reg_arg(Other, Int),Ret}] @@ -1651,7 +1950,7 @@ set_cg([{var,R}], Con, Le, Vdb, Bef, St) -> {Ais,clear_dead(Int, Le#l.i, Vdb),St}. -set_cg_map(Line, Op0, SrcReg, Target, Live, List, St0) -> +put_cg_map(Line, Op0, SrcReg, Target, Live, List, St0) -> Bfail = St0#cg.bfail, Fail = {f,St0#cg.bfail}, Op = case Op0 of @@ -1865,7 +2164,8 @@ cg_bin_opt_1([I|Is]) -> cg_bin_opt_1([]) -> []. -cg_bin_put({bin_seg,[],S0,U,T,Fs,[E0,Next]}, Fail, Bef) -> +cg_bin_put(#k_bin_seg{size=S0,unit=U,type=T,flags=Fs,seg=E0,next=Next}, + Fail, Bef) -> S1 = cg_reg_arg(S0, Bef), E1 = cg_reg_arg(E0, Bef), {Format,Op} = case T of @@ -1882,7 +2182,7 @@ cg_bin_put({bin_seg,[],S0,U,T,Fs,[E0,Next]}, Fail, Bef) -> utf -> [{Op,Fail,{field_flags,Fs},E1}|cg_bin_put(Next, Fail, Bef)] end; -cg_bin_put({bin_end,[]}, _, _) -> []. +cg_bin_put(#k_bin_end{}, _, _) -> []. cg_build_args(As, Bef) -> [{put,cg_reg_arg(A, Bef)} || A <- As]. @@ -1936,11 +2236,11 @@ get_locked_regs([], _) -> []. cg_reg_args(As, Bef) -> [cg_reg_arg(A, Bef) || A <- As]. -cg_reg_arg({var,V}, Bef) -> fetch_var(V, Bef); -cg_reg_arg(Literal, _) -> Literal. +cg_reg_arg(#k_var{name=V}, Bef) -> fetch_var(V, Bef); +cg_reg_arg(Literal, _) -> atomic(Literal). -cg_reg_arg_prefer_y({var,V}, Bef) -> fetch_var_prefer_y(V, Bef); -cg_reg_arg_prefer_y(Literal, _) -> Literal. +cg_reg_arg_prefer_y(#k_var{name=V}, Bef) -> fetch_var_prefer_y(V, Bef); +cg_reg_arg_prefer_y(Literal, _) -> atomic(Literal). %% cg_setup_call([Arg], Bef, Cur, Vdb) -> {[Instr],Aft}. %% Do the complete setup for a call/enter. @@ -1978,9 +2278,9 @@ cg_call_args(As, Bef, I, Vdb) -> load_arg_regs(Regs, As) -> load_arg_regs(Regs, As, 0). -load_arg_regs([_|Rs], [{var,V}|As], I) -> [{I,V}|load_arg_regs(Rs, As, I+1)]; +load_arg_regs([_|Rs], [#k_var{name=V}|As], I) -> [{I,V}|load_arg_regs(Rs, As, I+1)]; load_arg_regs([_|Rs], [A|As], I) -> [{I,A}|load_arg_regs(Rs, As, I+1)]; -load_arg_regs([], [{var,V}|As], I) -> [{I,V}|load_arg_regs([], As, I+1)]; +load_arg_regs([], [#k_var{name=V}|As], I) -> [{I,V}|load_arg_regs([], As, I+1)]; load_arg_regs([], [A|As], I) -> [{I,A}|load_arg_regs([], As, I+1)]; load_arg_regs(Rs, [], _) -> Rs. @@ -2016,12 +2316,13 @@ move_unsaved([], _, Regs, Acc) -> {Acc,Regs}. gen_moves(As, Sr) -> gen_moves(As, Sr, 0, []). -gen_moves([{var,V}|As], Sr, I, Acc) -> +gen_moves([#k_var{name=V}|As], Sr, I, Acc) -> case fetch_var(V, Sr) of {x,I} -> gen_moves(As, Sr, I+1, Acc); Reg -> gen_moves(As, Sr, I+1, [{move,Reg,{x,I}}|Acc]) end; -gen_moves([A|As], Sr, I, Acc) -> +gen_moves([A0|As], Sr, I, Acc) -> + A = atomic(A0), gen_moves(As, Sr, I+1, [{move,A,{x,I}}|Acc]); gen_moves([], _, _, Acc) -> lists:keysort(3, Acc). @@ -2190,7 +2491,7 @@ fetch_var_prefer_y(V, #sr{reg=Reg,stk=Stk}) -> end. load_vars(Vs, Regs) -> - foldl(fun ({var,V}, Rs) -> put_reg(V, Rs) end, Regs, Vs). + foldl(fun (#k_var{name=V}, Rs) -> put_reg(V, Rs) end, Regs, Vs). %% put_reg(Val, Regs) -> Regs. %% find_reg(Val, Regs) -> {ok,r{R}} | error. @@ -2291,6 +2592,16 @@ put_catch(Tag, [Other|Stk], Acc) -> drop_catch(Tag, [{{catch_tag,Tag}}|Stk]) -> [free|Stk]; drop_catch(Tag, [Other|Stk]) -> [Other|drop_catch(Tag, Stk)]. +%% atomic(Klit) -> Lit. +%% atomic_list([Klit]) -> [Lit]. + +atomic(#k_literal{val=V}) -> {literal,V}; +atomic(#k_int{val=I}) -> {integer,I}; +atomic(#k_float{val=F}) -> {float,F}; +atomic(#k_atom{val=A}) -> {atom,A}; +%%atomic(#k_char{val=C}) -> {char,C}; +atomic(#k_nil{}) -> nil. + %% new_label(St) -> {L,St}. new_label(#cg{lcount=Next}=St) -> @@ -2333,3 +2644,86 @@ flatmapfoldl(F, Accu0, [Hd|Tail]) -> {Rs,Accu2} = flatmapfoldl(F, Accu1, Tail), {R++Rs,Accu2}; flatmapfoldl(_, Accu, []) -> {[],Accu}. + +%% Keep track of life time for variables. +%% +%% init_vars([{var,VarName}]) -> Vdb. +%% new_vars([VarName], I, Vdb) -> Vdb. +%% use_vars([VarName], I, Vdb) -> Vdb. +%% add_var(VarName, F, L, Vdb) -> Vdb. +%% +%% The list of variable names for new_vars/3 and use_vars/3 +%% must be sorted. + +init_vars(Vs) -> + vdb_new(Vs). + +new_vars([], _, Vdb) -> Vdb; +new_vars([V], I, Vdb) -> vdb_store_new(V, {V,I,I}, Vdb); +new_vars(Vs, I, Vdb) -> vdb_update_vars(Vs, Vdb, I). + +use_vars([], _, Vdb) -> + Vdb; +use_vars([V], I, Vdb) -> + case vdb_find(V, Vdb) of + {V,F,L} when I > L -> vdb_update(V, {V,F,I}, Vdb); + {V,_,_} -> Vdb; + error -> vdb_store_new(V, {V,I,I}, Vdb) + end; +use_vars(Vs, I, Vdb) -> vdb_update_vars(Vs, Vdb, I). + +add_var(V, F, L, Vdb) -> + vdb_store_new(V, {V,F,L}, Vdb). + +%% vdb + +vdb_new(Vs) -> + ordsets:from_list([{V,0,0} || #k_var{name=V} <- Vs]). + +-type var() :: atom(). + +-spec vdb_find(var(), [vdb_entry()]) -> 'error' | vdb_entry(). + +vdb_find(V, Vdb) -> + case lists:keyfind(V, 1, Vdb) of + false -> error; + Vd -> Vd + end. + +vdb_update(V, Update, [{V,_,_}|Vdb]) -> + [Update|Vdb]; +vdb_update(V, Update, [Vd|Vdb]) -> + [Vd|vdb_update(V, Update, Vdb)]. + +vdb_store_new(V, New, [{V1,_,_}=Vd|Vdb]) when V > V1 -> + [Vd|vdb_store_new(V, New, Vdb)]; +vdb_store_new(V, New, [{V1,_,_}|_]=Vdb) when V < V1 -> + [New|Vdb]; +vdb_store_new(_, New, []) -> [New]. + +vdb_update_vars([V|_]=Vs, [{V1,_,_}=Vd|Vdb], I) when V > V1 -> + [Vd|vdb_update_vars(Vs, Vdb, I)]; +vdb_update_vars([V|Vs], [{V1,_,_}|_]=Vdb, I) when V < V1 -> + %% New variable. + [{V,I,I}|vdb_update_vars(Vs, Vdb, I)]; +vdb_update_vars([V|Vs], [{_,F,L}=Vd|Vdb], I) -> + %% Existing variable. + if + I > L -> [{V,F,I}|vdb_update_vars(Vs, Vdb, I)]; + true -> [Vd|vdb_update_vars(Vs, Vdb, I)] + end; +vdb_update_vars([V|Vs], [], I) -> + %% New variable. + [{V,I,I}|vdb_update_vars(Vs, [], I)]; +vdb_update_vars([], Vdb, _) -> Vdb. + +%% vdb_sub(Min, Max, Vdb) -> Vdb. +%% Extract variables which are used before and after Min. Lock +%% variables alive after Max. + +vdb_sub(Min, Max, Vdb) -> + [ if L >= Max -> {V,F,locked}; + true -> Vd + end || {V,F,L}=Vd <- Vdb, + F < Min, + L >= Min ]. diff --git a/lib/compiler/src/v3_kernel.erl b/lib/compiler/src/v3_kernel.erl index cbc8bb1303..3eea058153 100644 --- a/lib/compiler/src/v3_kernel.erl +++ b/lib/compiler/src/v3_kernel.erl @@ -160,8 +160,7 @@ function({#c_var{name={F,Arity}=FA},Body}, St0) -> {#ifun{anno=Ab,vars=Kvs,body=B0},[],St2} = expr(Body, new_sub(), St1), {B1,_,St3} = ubody(B0, return, St2), %%B1 = B0, St3 = St2, %Null second pass - {#k_fdef{anno=#k{us=[],ns=[],a=Ab}, - func=F,arity=Arity,vars=Kvs,body=B1},St3} + {make_fdef(#k{us=[],ns=[],a=Ab}, F, Arity, Kvs, B1),St3} catch Class:Error -> Stack = erlang:get_stacktrace(), @@ -2262,9 +2261,8 @@ iletrec_funs_gen(Fs, FreeVs, St) -> Arity0 = length(Vs), {Fb1,_,Lst1} = ubody(Fb0, return, Lst0#kern{ff={N,Arity0}}), Arity = Arity0 + length(FreeVs), - Fun = #k_fdef{anno=#k{us=[],ns=[],a=Fa}, - func=N,arity=Arity, - vars=Vs ++ FreeVs,body=Fb1}, + Fun = make_fdef(#k{us=[],ns=[],a=Fa}, N, Arity, + Vs++FreeVs, Fb1), Lst1#kern{funs=[Fun|Lst1#kern.funs]} end, St, Fs). @@ -2408,8 +2406,7 @@ uexpr(#ifun{anno=A,vars=Vs,body=B0}, {break,Rs}, St0) -> %% No id annotation. Must invent a fun name. new_fun_name(St1) end, - Fun = #k_fdef{anno=#k{us=[],ns=[],a=A},func=Fname,arity=Arity, - vars=Vs ++ Fvs,body=B1}, + Fun = make_fdef(#k{us=[],ns=[],a=A}, Fname, Arity, Vs++Fvs, B1), {#k_bif{anno=#k{us=Free,ns=lit_list_vars(Rs),a=A}, op=#k_internal{name=make_fun,arity=length(Free)+2}, args=[#k_atom{val=Fname},#k_int{val=Arity}|Fvs], @@ -2426,6 +2423,16 @@ uexpr(Lit, {break,Rs0}, St0) -> add_local_function(_, #kern{funs=ignore}=St) -> St; add_local_function(F, #kern{funs=Funs}=St) -> St#kern{funs=[F|Funs]}. +%% Make a #k_fdef{}, making sure that the body is always a #k_match{}. +make_fdef(Anno, Name, Arity, Vs, #k_match{}=Body) -> + #k_fdef{anno=Anno,func=Name,arity=Arity,vars=Vs,body=Body}; +make_fdef(Anno, Name, Arity, Vs, Body) -> + Ka = get_kanno(Body), + Match = #k_match{anno=#k{us=Ka#k.us,ns=[],a=Ka#k.a}, + vars=Vs,body=Body,ret=[]}, + #k_fdef{anno=Anno,func=Name,arity=Arity,vars=Vs,body=Match}. + + %% handle_reuse_annos([#k_var{}], State) -> State. %% In general, it is only safe to reuse a variable for a match context %% if the original value of the variable will no longer be needed. diff --git a/lib/compiler/src/v3_kernel.hrl b/lib/compiler/src/v3_kernel.hrl index 7cd30b25a8..87011b7680 100644 --- a/lib/compiler/src/v3_kernel.hrl +++ b/lib/compiler/src/v3_kernel.hrl @@ -79,7 +79,7 @@ -record(k_guard_clause, {anno=[],guard,body}). -record(k_break, {anno=[],args=[]}). --record(k_guard_break, {anno=[],args=[]}). +-record(k_guard_break, {anno=[],args=[],locked=[]}). -record(k_return, {anno=[],args=[]}). %%k_get_anno(Thing) -> element(2, Thing). diff --git a/lib/compiler/src/v3_life.erl b/lib/compiler/src/v3_life.erl deleted file mode 100644 index be3ade47ff..0000000000 --- a/lib/compiler/src/v3_life.erl +++ /dev/null @@ -1,468 +0,0 @@ -%% -%% %CopyrightBegin% -%% -%% Copyright Ericsson AB 1999-2016. 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 : Convert annotated kernel expressions to annotated beam format. - -%% This module creates beam format annotated with variable lifetime -%% information. Each thing is given an index and for each variable we -%% store the first and last index for its occurrence. The variable -%% database, VDB, attached to each thing is only relevant internally -%% for that thing. -%% -%% For nested things like matches the numbering continues locally and -%% the VDB for that thing refers to the variable usage within that -%% thing. Variables which live through a such a thing are internally -%% given a very large last index. Internally the indexes continue -%% after the index of that thing. This creates no problems as the -%% internal variable info never escapes and externally we only see -%% variable which are alive both before or after. -%% -%% This means that variables never "escape" from a thing and the only -%% way to get values from a thing is to "return" them, with 'break' or -%% 'return'. Externally these values become the return values of the -%% thing. This is no real limitation as most nested things have -%% multiple threads so working out a common best variable usage is -%% difficult. - --module(v3_life). - --export([module/2]). - --export([vdb_find/2]). - --import(lists, [member/2,map/2,reverse/1,sort/1]). --import(ordsets, [add_element/2,intersection/2,union/2]). - --include("v3_kernel.hrl"). --include("v3_life.hrl"). - --type fa() :: {atom(),arity()}. - -%% These are not defined in v3_kernel.hrl. -get_kanno(Kthing) -> element(2, Kthing). -%%set_kanno(Kthing, Anno) -> setelement(2, Kthing, Anno). - --spec module(#k_mdef{}, [compile:option()]) -> - {'ok',{module(),[fa()],[_],[_]}}. - -module(#k_mdef{name=M,exports=Es,attributes=As,body=Fs0}, _Opts) -> - Fs1 = functions(Fs0, []), - {ok,{M,Es,As,Fs1}}. - -functions([F|Fs], Acc) -> - functions(Fs, [function(F)|Acc]); -functions([], Acc) -> reverse(Acc). - -%% function(Kfunc) -> Func. - -function(#k_fdef{anno=#k{a=Anno},func=F,arity=Ar,vars=Vs,body=Kb}) -> - try - As = var_list(Vs), - Vdb0 = init_vars(As), - %% Force a top-level match! - B0 = case Kb of - #k_match{} -> Kb; - _ -> - Ka = get_kanno(Kb), - #k_match{anno=#k{us=Ka#k.us,ns=[],a=Ka#k.a}, - vars=Vs,body=Kb,ret=[]} - end, - {B1,_,Vdb1} = body(B0, 1, Vdb0), - {function,F,Ar,As,B1,Vdb1,Anno} - catch - Class:Error -> - Stack = erlang:get_stacktrace(), - io:fwrite("Function: ~w/~w\n", [F,Ar]), - erlang:raise(Class, Error, Stack) - end. - -%% body(Kbody, I, Vdb) -> {[Expr],MaxI,Vdb}. -%% Handle a body. - -body(#k_seq{arg=Ke,body=Kb}, I, Vdb0) -> - %%ok = io:fwrite("life ~w:~p~n", [?LINE,{Ke,I,Vdb0}]), - A = get_kanno(Ke), - Vdb1 = use_vars(union(A#k.us, A#k.ns), I, Vdb0), - {Es,MaxI,Vdb2} = body(Kb, I+1, Vdb1), - E = expr(Ke, I, Vdb2), - {[E|Es],MaxI,Vdb2}; -body(Ke, I, Vdb0) -> - %%ok = io:fwrite("life ~w:~p~n", [?LINE,{Ke,I,Vdb0}]), - A = get_kanno(Ke), - Vdb1 = use_vars(union(A#k.us, A#k.ns), I, Vdb0), - E = expr(Ke, I, Vdb1), - {[E],I,Vdb1}. - -%% protected(Kprotected, I, Vdb) -> Protected. -%% Only used in guards. - -protected(#k_protected{anno=A,arg=Ts,ret=Rs}, I, Vdb) -> - %% Lock variables that are alive before try and used afterwards. - %% Don't lock variables that are only used inside the protected - %% expression. - Pdb0 = vdb_sub(I, I+1, Vdb), - {T,MaxI,Pdb1} = body(Ts, I+1, Pdb0), - Pdb2 = use_vars(A#k.ns, MaxI+1, Pdb1), %Save "return" values - #l{ke={protected,T,var_list(Rs)},i=I,a=A#k.a,vdb=Pdb2}. - -%% expr(Kexpr, I, Vdb) -> Expr. - -expr(#k_test{anno=A,op=Op,args=As,inverted=Inverted}, I, _Vdb) -> - #l{ke={test,test_op(Op),atomic_list(As),Inverted},i=I,a=A#k.a}; -expr(#k_call{anno=A,op=Op,args=As,ret=Rs}, I, _Vdb) -> - #l{ke={call,call_op(Op),atomic_list(As),var_list(Rs)},i=I,a=A#k.a}; -expr(#k_enter{anno=A,op=Op,args=As}, I, _Vdb) -> - #l{ke={enter,call_op(Op),atomic_list(As)},i=I,a=A#k.a}; -expr(#k_bif{anno=A,op=Op,args=As,ret=Rs}, I, _Vdb) -> - Bif = k_bif(A, Op, As, Rs), - #l{ke=Bif,i=I,a=A#k.a}; -expr(#k_match{anno=A,body=Kb,ret=Rs}, I, Vdb) -> - %% Work out imported variables which need to be locked. - Mdb = vdb_sub(I, I+1, Vdb), - M = match(Kb, A#k.us, I+1, [], Mdb), - #l{ke={match,M,var_list(Rs)},i=I,vdb=use_vars(A#k.us, I+1, Mdb),a=A#k.a}; -expr(#k_guard_match{anno=A,body=Kb,ret=Rs}, I, Vdb) -> - %% Work out imported variables which need to be locked. - Mdb = vdb_sub(I, I+1, Vdb), - M = match(Kb, A#k.us, I+1, [], Mdb), - #l{ke={guard_match,M,var_list(Rs)},i=I,vdb=use_vars(A#k.us, I+1, Mdb),a=A#k.a}; -expr(#k_try{}=Try, I, Vdb) -> - body_try(Try, I, Vdb); -expr(#k_protected{}=Protected, I, Vdb) -> - protected(Protected, I, Vdb); -expr(#k_try_enter{anno=A,arg=Ka,vars=Vs,body=Kb,evars=Evs,handler=Kh}, I, Vdb) -> - %% Lock variables that are alive before the catch and used afterwards. - %% Don't lock variables that are only used inside the try. - Tdb0 = vdb_sub(I, I+1, Vdb), - %% This is the tricky bit. Lock variables in Arg that are used in - %% the body and handler. Add try tag 'variable'. - Ab = get_kanno(Kb), - Ah = get_kanno(Kh), - Tdb1 = use_vars(union(Ab#k.us, Ah#k.us), I+3, Tdb0), - Tdb2 = vdb_sub(I, I+2, Tdb1), - Vnames = fun (Kvar) -> Kvar#k_var.name end, %Get the variable names - {Aes,_,Adb} = body(Ka, I+2, add_var({catch_tag,I+1}, I+1, 1000000, Tdb2)), - {Bes,_,Bdb} = body(Kb, I+4, new_vars(sort(map(Vnames, Vs)), I+3, Tdb2)), - {Hes,_,Hdb} = body(Kh, I+4, new_vars(sort(map(Vnames, Evs)), I+3, Tdb2)), - #l{ke={try_enter,#l{ke={block,Aes},i=I+1,vdb=Adb,a=[]}, - var_list(Vs),#l{ke={block,Bes},i=I+3,vdb=Bdb,a=[]}, - var_list(Evs),#l{ke={block,Hes},i=I+3,vdb=Hdb,a=[]}}, - i=I,vdb=Tdb1,a=A#k.a}; -expr(#k_catch{anno=A,body=Kb,ret=[R]}, I, Vdb) -> - %% Lock variables that are alive before the catch and used afterwards. - %% Don't lock variables that are only used inside the catch. - %% Add catch tag 'variable'. - Cdb0 = vdb_sub(I, I+1, Vdb), - {Es,_,Cdb1} = body(Kb, I+1, add_var({catch_tag,I}, I, locked, Cdb0)), - #l{ke={'catch',Es,variable(R)},i=I,vdb=Cdb1,a=A#k.a}; -expr(#k_receive{anno=A,var=V,body=Kb,timeout=T,action=Ka,ret=Rs}, I, Vdb) -> - %% Work out imported variables which need to be locked. - Rdb = vdb_sub(I, I+1, Vdb), - M = match(Kb, add_element(V#k_var.name, A#k.us), I+1, [], - new_vars([V#k_var.name], I, Rdb)), - {Tes,_,Adb} = body(Ka, I+1, Rdb), - #l{ke={receive_loop,atomic(T),variable(V),M, - #l{ke=Tes,i=I+1,vdb=Adb,a=[]},var_list(Rs)}, - i=I,vdb=use_vars(A#k.us, I+1, Vdb),a=A#k.a}; -expr(#k_receive_accept{anno=A}, I, _Vdb) -> - #l{ke=receive_accept,i=I,a=A#k.a}; -expr(#k_receive_next{anno=A}, I, _Vdb) -> - #l{ke=receive_next,i=I,a=A#k.a}; -expr(#k_put{anno=A,arg=Arg,ret=Rs}, I, _Vdb) -> - #l{ke={set,var_list(Rs),literal(Arg, [])},i=I,a=A#k.a}; -expr(#k_break{anno=A,args=As}, I, _Vdb) -> - #l{ke={break,atomic_list(As)},i=I,a=A#k.a}; -expr(#k_guard_break{anno=A,args=As}, I, Vdb) -> - Locked = [V || {V,_,_} <- Vdb], - #l{ke={guard_break,atomic_list(As),Locked},i=I,a=A#k.a}; -expr(#k_return{anno=A,args=As}, I, _Vdb) -> - #l{ke={return,atomic_list(As)},i=I,a=A#k.a}. - -body_try(#k_try{anno=A,arg=Ka,vars=Vs,body=Kb,evars=Evs,handler=Kh,ret=Rs}, - I, Vdb) -> - %% Lock variables that are alive before the catch and used afterwards. - %% Don't lock variables that are only used inside the try. - Tdb0 = vdb_sub(I, I+1, Vdb), - %% This is the tricky bit. Lock variables in Arg that are used in - %% the body and handler. Add try tag 'variable'. - Ab = get_kanno(Kb), - Ah = get_kanno(Kh), - Tdb1 = use_vars(union(Ab#k.us, Ah#k.us), I+3, Tdb0), - Tdb2 = vdb_sub(I, I+2, Tdb1), - Vnames = fun (Kvar) -> Kvar#k_var.name end, %Get the variable names - {Aes,_,Adb} = body(Ka, I+2, add_var({catch_tag,I+1}, I+1, locked, Tdb2)), - {Bes,_,Bdb} = body(Kb, I+4, new_vars(sort(map(Vnames, Vs)), I+3, Tdb2)), - {Hes,_,Hdb} = body(Kh, I+4, new_vars(sort(map(Vnames, Evs)), I+3, Tdb2)), - #l{ke={'try',#l{ke={block,Aes},i=I+1,vdb=Adb,a=[]}, - var_list(Vs),#l{ke={block,Bes},i=I+3,vdb=Bdb,a=[]}, - var_list(Evs),#l{ke={block,Hes},i=I+3,vdb=Hdb,a=[]}, - var_list(Rs)}, - i=I,vdb=Tdb1,a=A#k.a}. - -%% call_op(Op) -> Op. -%% test_op(Op) -> Op. -%% Do any necessary name translations here to munge into beam format. - -call_op(#k_local{name=N}) -> N; -call_op(#k_remote{mod=M,name=N}) -> {remote,atomic(M),atomic(N)}; -call_op(Other) -> variable(Other). - -test_op(#k_remote{mod=#k_atom{val=erlang},name=#k_atom{val=N}}) -> N. - -%% k_bif(Anno, Op, [Arg], [Ret], Vdb) -> Expr. -%% Build bifs. - -k_bif(_A, #k_internal{name=Name}, As, Rs) -> - {internal,Name,atomic_list(As),var_list(Rs)}; -k_bif(_A, #k_remote{mod=#k_atom{val=erlang},name=#k_atom{val=Name}}, As, Rs) -> - Ar = length(As), - case is_gc_bif(Name, Ar) of - false -> - {bif,Name,atomic_list(As),var_list(Rs)}; - true -> - {gc_bif,Name,atomic_list(As),var_list(Rs)} - end. - -%% match(Kexpr, [LockVar], I, Vdb) -> Expr. -%% Convert match tree to old format. - -match(#k_alt{anno=A,first=Kf,then=Kt}, Ls, I, Ctxt, Vdb0) -> - Vdb1 = use_vars(union(A#k.us, Ls), I, Vdb0), - F = match(Kf, Ls, I+1, Ctxt, Vdb1), - T = match(Kt, Ls, I+1, Ctxt, Vdb1), - #l{ke={alt,F,T},i=I,vdb=Vdb1,a=A#k.a}; -match(#k_select{anno=A,var=V,types=Kts}, Ls0, I, Ctxt, Vdb0) -> - Vanno = get_kanno(V), - Ls1 = case member(no_usage, Vanno) of - false -> add_element(V#k_var.name, Ls0); - true -> Ls0 - end, - Anno = case member(reuse_for_context, Vanno) of - true -> [reuse_for_context|A#k.a]; - false -> A#k.a - end, - Vdb1 = use_vars(union(A#k.us, Ls1), I, Vdb0), - Ts = [type_clause(Tc, Ls1, I+1, Ctxt, Vdb1) || Tc <- Kts], - #l{ke={select,literal(V, Ctxt),Ts},i=I,vdb=Vdb1,a=Anno}; -match(#k_guard{anno=A,clauses=Kcs}, Ls, I, Ctxt, Vdb0) -> - Vdb1 = use_vars(union(A#k.us, Ls), I, Vdb0), - Cs = [guard_clause(G, Ls, I+1, Ctxt, Vdb1) || G <- Kcs], - #l{ke={guard,Cs},i=I,vdb=Vdb1,a=A#k.a}; -match(Other, Ls, I, _Ctxt, Vdb0) -> - Vdb1 = use_vars(Ls, I, Vdb0), - {B,_,Vdb2} = body(Other, I+1, Vdb1), - #l{ke={block,B},i=I,vdb=Vdb2,a=[]}. - -type_clause(#k_type_clause{anno=A,type=T,values=Kvs}, Ls, I, Ctxt, Vdb0) -> - %%ok = io:format("life ~w: ~p~n", [?LINE,{T,Kvs}]), - Vdb1 = use_vars(union(A#k.us, Ls), I+1, Vdb0), - Vs = [val_clause(Vc, Ls, I+1, Ctxt, Vdb1) || Vc <- Kvs], - #l{ke={type_clause,type(T),Vs},i=I,vdb=Vdb1,a=A#k.a}. - -val_clause(#k_val_clause{anno=A,val=V,body=Kb}, Ls0, I, Ctxt0, Vdb0) -> - New = (get_kanno(V))#k.ns, - Bus = (get_kanno(Kb))#k.us, - %%ok = io:format("Ls0 = ~p, Used=~p\n New=~p, Bus=~p\n", [Ls0,Used,New,Bus]), - Ls1 = union(intersection(New, Bus), Ls0), %Lock for safety - Vdb1 = use_vars(union(A#k.us, Ls1), I+1, new_vars(New, I, Vdb0)), - Ctxt = case V of - #k_binary{segs=#k_var{name=C0}} -> C0; - _ -> Ctxt0 - end, - B = match(Kb, Ls1, I+1, Ctxt, Vdb1), - #l{ke={val_clause,literal(V, Ctxt),B},i=I,vdb=use_vars(Bus, I+1, Vdb1),a=A#k.a}. - -guard_clause(#k_guard_clause{anno=A,guard=Kg,body=Kb}, Ls, I, Ctxt, Vdb0) -> - Vdb1 = use_vars(union(A#k.us, Ls), I+2, Vdb0), - Gdb = vdb_sub(I+1, I+2, Vdb1), - G = protected(Kg, I+1, Gdb), - B = match(Kb, Ls, I+2, Ctxt, Vdb1), - #l{ke={guard_clause,G,B}, - i=I,vdb=use_vars((get_kanno(Kg))#k.us, I+2, Vdb1), - a=A#k.a}. - -%% type(Ktype) -> Type. - -type(k_literal) -> literal; -type(k_int) -> integer; -%%type(k_char) -> integer; %Hhhmmm??? -type(k_float) -> float; -type(k_atom) -> atom; -type(k_nil) -> nil; -type(k_cons) -> cons; -type(k_tuple) -> tuple; -type(k_binary) -> binary; -type(k_bin_seg) -> bin_seg; -type(k_bin_int) -> bin_int; -type(k_bin_end) -> bin_end; -type(k_map) -> map. - -%% variable(Klit) -> Lit. -%% var_list([Klit]) -> [Lit]. - -variable(#k_var{name=N}) -> {var,N}. - -var_list(Ks) -> [variable(K) || K <- Ks]. - -%% atomic(Klit) -> Lit. -%% atomic_list([Klit]) -> [Lit]. - -atomic(#k_literal{val=V}) -> {literal,V}; -atomic(#k_var{name=N}) -> {var,N}; -atomic(#k_int{val=I}) -> {integer,I}; -atomic(#k_float{val=F}) -> {float,F}; -atomic(#k_atom{val=N}) -> {atom,N}; -%%atomic(#k_char{val=C}) -> {char,C}; -atomic(#k_nil{}) -> nil. - -atomic_list(Ks) -> [atomic(K) || K <- Ks]. - -%% literal(Klit) -> Lit. -%% literal_list([Klit]) -> [Lit]. - -literal(#k_var{name=N}, _) -> {var,N}; -literal(#k_literal{val=I}, _) -> {literal,I}; -literal(#k_int{val=I}, _) -> {integer,I}; -literal(#k_float{val=F}, _) -> {float,F}; -literal(#k_atom{val=N}, _) -> {atom,N}; -%%literal(#k_char{val=C}, _) -> {char,C}; -literal(#k_nil{}, _) -> nil; -literal(#k_cons{hd=H,tl=T}, Ctxt) -> - {cons,[literal(H, Ctxt),literal(T, Ctxt)]}; -literal(#k_binary{segs=V}, Ctxt) -> - {binary,literal(V, Ctxt)}; -literal(#k_bin_seg{size=S,unit=U,type=T,flags=Fs,seg=Seg,next=[]}, Ctxt) -> - %% Only occurs in patterns. - {bin_seg,Ctxt,literal(S, Ctxt),U,T,Fs,[literal(Seg, Ctxt)]}; -literal(#k_bin_seg{size=S,unit=U,type=T,flags=Fs,seg=Seg,next=N}, Ctxt) -> - {bin_seg,Ctxt,literal(S, Ctxt),U,T,Fs, - [literal(Seg, Ctxt),literal(N, Ctxt)]}; -literal(#k_bin_int{size=S,unit=U,flags=Fs,val=Int,next=N}, Ctxt) -> - %% Only occurs in patterns. - {bin_int,Ctxt,literal(S, Ctxt),U,Fs,Int, - [literal(N, Ctxt)]}; -literal(#k_bin_end{}, Ctxt) -> - {bin_end,Ctxt}; -literal(#k_tuple{es=Es}, Ctxt) -> - {tuple,literal_list(Es, Ctxt)}; -literal(#k_map{op=Op,var=Var,es=Es0}, Ctxt) -> - {map,Op,literal(Var, Ctxt),literal_list(Es0, Ctxt)}; -literal(#k_map_pair{key=K,val=V}, Ctxt) -> - {map_pair,literal(K, Ctxt),literal(V, Ctxt)}. - -literal_list(Ks, Ctxt) -> - [literal(K, Ctxt) || K <- Ks]. - - -%% is_gc_bif(Name, Arity) -> true|false -%% Determines whether the BIF Name/Arity might do a GC. - -is_gc_bif(hd, 1) -> false; -is_gc_bif(tl, 1) -> false; -is_gc_bif(self, 0) -> false; -is_gc_bif(node, 0) -> false; -is_gc_bif(node, 1) -> false; -is_gc_bif(element, 2) -> false; -is_gc_bif(get, 1) -> false; -is_gc_bif(tuple_size, 1) -> false; -is_gc_bif(Bif, Arity) -> - not (erl_internal:bool_op(Bif, Arity) orelse - erl_internal:new_type_test(Bif, Arity) orelse - erl_internal:comp_op(Bif, Arity)). - -%% Keep track of life time for variables. -%% -%% init_vars([{var,VarName}]) -> Vdb. -%% new_vars([VarName], I, Vdb) -> Vdb. -%% use_vars([VarName], I, Vdb) -> Vdb. -%% add_var(VarName, F, L, Vdb) -> Vdb. -%% -%% The list of variable names for new_vars/3 and use_vars/3 -%% must be sorted. - -init_vars(Vs) -> - vdb_new(Vs). - -new_vars([], _, Vdb) -> Vdb; -new_vars([V], I, Vdb) -> vdb_store_new(V, {V,I,I}, Vdb); -new_vars(Vs, I, Vdb) -> vdb_update_vars(Vs, Vdb, I). - -use_vars([], _, Vdb) -> - Vdb; -use_vars([V], I, Vdb) -> - case vdb_find(V, Vdb) of - {V,F,L} when I > L -> vdb_update(V, {V,F,I}, Vdb); - {V,_,_} -> Vdb; - error -> vdb_store_new(V, {V,I,I}, Vdb) - end; -use_vars(Vs, I, Vdb) -> vdb_update_vars(Vs, Vdb, I). - -add_var(V, F, L, Vdb) -> - vdb_store_new(V, {V,F,L}, Vdb). - -%% vdb - -vdb_new(Vs) -> - sort([{V,0,0} || {var,V} <- Vs]). - --type var() :: atom(). - --spec vdb_find(var(), [vdb_entry()]) -> 'error' | vdb_entry(). - -vdb_find(V, Vdb) -> - case lists:keyfind(V, 1, Vdb) of - false -> error; - Vd -> Vd - end. - -vdb_update(V, Update, [{V,_,_}|Vdb]) -> - [Update|Vdb]; -vdb_update(V, Update, [Vd|Vdb]) -> - [Vd|vdb_update(V, Update, Vdb)]. - -vdb_store_new(V, New, [{V1,_,_}=Vd|Vdb]) when V > V1 -> - [Vd|vdb_store_new(V, New, Vdb)]; -vdb_store_new(V, New, [{V1,_,_}|_]=Vdb) when V < V1 -> - [New|Vdb]; -vdb_store_new(_, New, []) -> [New]. - -vdb_update_vars([V|_]=Vs, [{V1,_,_}=Vd|Vdb], I) when V > V1 -> - [Vd|vdb_update_vars(Vs, Vdb, I)]; -vdb_update_vars([V|Vs], [{V1,_,_}|_]=Vdb, I) when V < V1 -> - %% New variable. - [{V,I,I}|vdb_update_vars(Vs, Vdb, I)]; -vdb_update_vars([V|Vs], [{_,F,L}=Vd|Vdb], I) -> - %% Existing variable. - if - I > L -> [{V,F,I}|vdb_update_vars(Vs, Vdb, I)]; - true -> [Vd|vdb_update_vars(Vs, Vdb, I)] - end; -vdb_update_vars([V|Vs], [], I) -> - %% New variable. - [{V,I,I}|vdb_update_vars(Vs, [], I)]; -vdb_update_vars([], Vdb, _) -> Vdb. - -%% vdb_sub(Min, Max, Vdb) -> Vdb. -%% Extract variables which are used before and after Min. Lock -%% variables alive after Max. - -vdb_sub(Min, Max, Vdb) -> - [ if L >= Max -> {V,F,locked}; - true -> Vd - end || {V,F,L}=Vd <- Vdb, F < Min, L >= Min ]. diff --git a/lib/compiler/src/v3_life.hrl b/lib/compiler/src/v3_life.hrl deleted file mode 100644 index 5c76312067..0000000000 --- a/lib/compiler/src/v3_life.hrl +++ /dev/null @@ -1,29 +0,0 @@ -%% -%% %CopyrightBegin% -%% -%% Copyright Ericsson AB 1999-2016. 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% -%% -%% This record contains variable life-time annotation for a -%% kernel expression. Added by v3_life, used by v3_codegen. - --type vdb_entry() :: {atom(),non_neg_integer(),non_neg_integer()}. - --record(l, {ke, %Kernel expression - i=0 :: non_neg_integer(), %Op number - vdb=[] :: [vdb_entry()], %Variable database - a=[] :: [term()]}). %Core annotation - diff --git a/lib/compiler/test/compile_SUITE.erl b/lib/compiler/test/compile_SUITE.erl index 25983c6012..e88a132d59 100644 --- a/lib/compiler/test/compile_SUITE.erl +++ b/lib/compiler/test/compile_SUITE.erl @@ -375,7 +375,6 @@ do_file_listings(DataDir, PrivDir, [File|Files]) -> do_listing(Simple, TargetDir, dcbsm, ".core_bsm"), do_listing(Simple, TargetDir, dsetel, ".dsetel"), do_listing(Simple, TargetDir, dkern, ".kernel"), - do_listing(Simple, TargetDir, dlife, ".life"), do_listing(Simple, TargetDir, dcg, ".codegen"), do_listing(Simple, TargetDir, dblk, ".block"), do_listing(Simple, TargetDir, dexcept, ".except"), diff --git a/lib/compiler/test/misc_SUITE.erl b/lib/compiler/test/misc_SUITE.erl index ea4aaf40a9..b12bcbeeab 100644 --- a/lib/compiler/test/misc_SUITE.erl +++ b/lib/compiler/test/misc_SUITE.erl @@ -171,7 +171,7 @@ silly_coverage(Config) when is_list(Config) -> expect_error(fun() -> sys_core_dsetel:module(BadCoreErlang, []) end), expect_error(fun() -> v3_kernel:module(BadCoreErlang, []) end), - %% v3_life + %% v3_codegen BadKernel = {k_mdef,[],?MODULE, [{foo,0}], [], @@ -179,11 +179,7 @@ silly_coverage(Config) when is_list(Config) -> {k,[],[],[]}, f,0,[], seriously_bad_body}]}, - expect_error(fun() -> v3_life:module(BadKernel, []) end), - - %% v3_codegen - CodegenInput = {?MODULE,[{foo,0}],[],[{function,foo,0,[a|b],a,b,[]}]}, - expect_error(fun() -> v3_codegen:module(CodegenInput, []) end), + expect_error(fun() -> v3_codegen:module(BadKernel, []) end), %% beam_a BeamAInput = {?MODULE,[{foo,0}],[], diff --git a/lib/crypto/c_src/crypto.c b/lib/crypto/c_src/crypto.c index 53fe233790..c4e80e3153 100644 --- a/lib/crypto/c_src/crypto.c +++ b/lib/crypto/c_src/crypto.c @@ -277,7 +277,7 @@ static HMAC_CTX *HMAC_CTX_new() static void HMAC_CTX_free(HMAC_CTX *ctx) { HMAC_CTX_cleanup(ctx); - return CRYPTO_free(ctx); + CRYPTO_free(ctx); } #define EVP_MD_CTX_new() EVP_MD_CTX_create() @@ -2846,7 +2846,7 @@ static ERL_NIF_TERM rsa_generate_key_nif(ErlNifEnv* env, int argc, const ERL_NIF static ERL_NIF_TERM dh_generate_parameters_nif(ErlNifEnv* env, int argc, const ERL_NIF_TERM argv[]) {/* (PrimeLen, Generator) */ int prime_len, generator; - DH* dh_params; + DH* dh_params = NULL; int p_len, g_len; unsigned char *p_ptr, *g_ptr; ERL_NIF_TERM ret_p, ret_g; @@ -2857,8 +2857,8 @@ static ERL_NIF_TERM dh_generate_parameters_nif(ErlNifEnv* env, int argc, const E return enif_make_badarg(env); } - dh_params = DH_generate_parameters(prime_len, generator, NULL, NULL); - if (dh_params == NULL) { + + if (DH_generate_parameters_ex(dh_params, prime_len, generator, NULL)) { return atom_error; } DH_get0_pqg(dh_params, &dh_p, &dh_q, &dh_g); diff --git a/lib/hipe/main/hipe.app.src b/lib/hipe/main/hipe.app.src index 5b2280594f..fb750dd418 100644 --- a/lib/hipe/main/hipe.app.src +++ b/lib/hipe/main/hipe.app.src @@ -236,4 +236,4 @@ {applications, [kernel,stdlib]}, {env, []}, {runtime_dependencies, ["syntax_tools-1.6.14","stdlib-3.4","kernel-5.3", - "erts-9.0","compiler-5.0"]}]}. + "erts-9.2","compiler-5.0"]}]}. diff --git a/lib/hipe/rtl/hipe_rtl_binary_construct.erl b/lib/hipe/rtl/hipe_rtl_binary_construct.erl index 52ea5db382..bc215e3abe 100644 --- a/lib/hipe/rtl/hipe_rtl_binary_construct.erl +++ b/lib/hipe/rtl/hipe_rtl_binary_construct.erl @@ -195,8 +195,13 @@ gen_rtl(BsOP, Dst, Args, TrueLblName, FalseLblName, SystemLimitLblName, ConstTab bs_validate_unicode -> [_Arg] = Args, - [hipe_rtl:mk_call([], bs_validate_unicode, Args, - TrueLblName, FalseLblName, not_remote)]; + [IsUnicode] = create_regs(1), + RetLbl = hipe_rtl:mk_new_label(), + [hipe_rtl:mk_call([IsUnicode], is_unicode, Args, + hipe_rtl:label_name(RetLbl), [], not_remote), + RetLbl, + hipe_rtl:mk_branch(IsUnicode, ne, hipe_rtl:mk_imm(0), + TrueLblName, FalseLblName, 0.99)]; bs_final -> Zero = hipe_rtl:mk_imm(0), diff --git a/lib/inets/doc/src/notes.xml b/lib/inets/doc/src/notes.xml index 10ef84d7cf..1ff6aefaa7 100644 --- a/lib/inets/doc/src/notes.xml +++ b/lib/inets/doc/src/notes.xml @@ -33,7 +33,39 @@ <file>notes.xml</file> </header> - <section><title>Inets 6.4.2</title> + <section><title>Inets 6.4.3</title> + + <section><title>Improvements and New Features</title> + <list> + <item> + <p> + Fix broken handling of POST requests</p> + <p> + New chunk mechanism of body data in POST requests added + in 5d01c70ca399edf28e99dc760506329689fab6ba broke + handling of POST body data not using the new mechanism.</p> + <p> + Own Id: OTP-14656</p> + </item> + <item> + <p> + Make sure ints:stop/2 of the service httpd is synchronous</p> + <p> + Own Id: OTP-14696</p> + </item> + <item> + <p> + Honor status code returned by ESI script and modernize + "location" header handling.</p> + <p> + Own Id: OTP-14716</p> + </item> + </list> + </section> + +</section> + +<section><title>Inets 6.4.2</title> <section><title>Fixed Bugs and Malfunctions</title> <list> diff --git a/lib/inets/src/http_client/httpc_response.erl b/lib/inets/src/http_client/httpc_response.erl index b3b11b74ab..91638f5d2e 100644 --- a/lib/inets/src/http_client/httpc_response.erl +++ b/lib/inets/src/http_client/httpc_response.erl @@ -269,7 +269,7 @@ parse_headers(<<?LF,?LF,Body/binary>>, Header, Headers, MaxHeaderSize, Result, Relaxed); parse_headers(<<?CR,?LF,?CR,?LF,Body/binary>>, Header, Headers, - MaxHeaderSize, Result, _) -> + MaxHeaderSize, Result, Relaxed) -> HTTPHeaders = [lists:reverse(Header) | Headers], Length = lists:foldl(fun(H, Acc) -> length(H) + Acc end, 0, HTTPHeaders), @@ -277,8 +277,42 @@ parse_headers(<<?CR,?LF,?CR,?LF,Body/binary>>, Header, Headers, true -> ResponseHeaderRcord = http_response:headers(HTTPHeaders, #http_response_h{}), - {ok, list_to_tuple( - lists:reverse([Body, ResponseHeaderRcord | Result]))}; + + %% RFC7230, Section 3.3.3 + %% If a message is received with both a Transfer-Encoding and a + %% Content-Length header field, the Transfer-Encoding overrides the + %% Content-Length. Such a message might indicate an attempt to + %% perform request smuggling (Section 9.5) or response splitting + %% (Section 9.4) and ought to be handled as an error. A sender MUST + %% remove the received Content-Length field prior to forwarding such + %% a message downstream. + case ResponseHeaderRcord#http_response_h.'transfer-encoding' of + undefined -> + {ok, list_to_tuple( + lists:reverse([Body, ResponseHeaderRcord | Result]))}; + Value -> + TransferEncoding = string:lowercase(Value), + ContentLength = ResponseHeaderRcord#http_response_h.'content-length', + if + %% Respond without error but remove Content-Length field in relaxed mode + (Relaxed =:= true) + andalso (TransferEncoding =:= "chunked") + andalso (ContentLength =/= "-1") -> + ResponseHeaderRcordFixed = + ResponseHeaderRcord#http_response_h{'content-length' = "-1"}, + {ok, list_to_tuple( + lists:reverse([Body, ResponseHeaderRcordFixed | Result]))}; + %% Respond with error in default (not relaxed) mode + (Relaxed =:= false) + andalso (TransferEncoding =:= "chunked") + andalso (ContentLength =/= "-1") -> + throw({error, {headers_conflict, {'content-length', + 'transfer-encoding'}}}); + true -> + {ok, list_to_tuple( + lists:reverse([Body, ResponseHeaderRcord | Result]))} + end + end; false -> throw({error, {header_too_long, MaxHeaderSize, MaxHeaderSize-Length}}) diff --git a/lib/inets/src/http_server/httpd_esi.erl b/lib/inets/src/http_server/httpd_esi.erl index 9406b47802..fd50934d00 100644 --- a/lib/inets/src/http_server/httpd_esi.erl +++ b/lib/inets/src/http_server/httpd_esi.erl @@ -66,7 +66,7 @@ handle_headers("") -> {ok, [], 200}; handle_headers(Headers) -> NewHeaders = string:tokens(Headers, ?CRLF), - handle_headers(NewHeaders, [], 200). + handle_headers(NewHeaders, [], 200, true). %%%======================================================================== %%% Internal functions @@ -80,21 +80,17 @@ parse_headers([?CR, ?LF, ?CR, ?LF | Rest], Acc) -> parse_headers([Char | Rest], Acc) -> parse_headers(Rest, [Char | Acc]). -handle_headers([], NewHeaders, StatusCode) -> +handle_headers([], NewHeaders, StatusCode, _) -> {ok, NewHeaders, StatusCode}; -handle_headers([Header | Headers], NewHeaders, StatusCode) -> +handle_headers([Header | Headers], NewHeaders, StatusCode, NoESIStatus) -> {FieldName, FieldValue} = httpd_response:split_header(Header, []), case FieldName of "location" -> - case http_request:is_absolut_uri(FieldValue) of - true -> - handle_headers(Headers, - [{FieldName, FieldValue} | NewHeaders], - 302); - false -> - {proceed, FieldValue} - end; + handle_headers(Headers, + [{FieldName, FieldValue} | NewHeaders], + 302, NoESIStatus); + "status" -> NewStatusCode = case httpd_util:split(FieldValue," ",2) of @@ -103,8 +99,9 @@ handle_headers([Header | Headers], NewHeaders, StatusCode) -> _ -> 200 end, - handle_headers(Headers, NewHeaders, NewStatusCode); + handle_headers(Headers, NewHeaders, NewStatusCode, false); _ -> handle_headers(Headers, - [{FieldName, FieldValue}| NewHeaders], StatusCode) - end. + [{FieldName, FieldValue}| NewHeaders], StatusCode, + NoESIStatus) + end. diff --git a/lib/inets/src/http_server/httpd_example.erl b/lib/inets/src/http_server/httpd_example.erl index 45b6deba97..adbbf64685 100644 --- a/lib/inets/src/http_server/httpd_example.erl +++ b/lib/inets/src/http_server/httpd_example.erl @@ -20,7 +20,7 @@ %% -module(httpd_example). -export([print/1]). --export([get/2, put/2, post/2, yahoo/2, test1/2, get_bin/2, peer/2]). +-export([get/2, put/2, post/2, yahoo/2, test1/2, get_bin/2, peer/2,new_status_and_location/2]). -export([newformat/3, post_chunked/3]). %% These are used by the inets test-suite @@ -90,6 +90,9 @@ post(Env,Input) -> yahoo(_Env,_Input) -> "Location: http://www.yahoo.com\r\n\r\n". +new_status_and_location(_Env,_Input) -> + "status:201\r\n Location: http://www.yahoo.com\r\n\r\n". + default(Env,Input) -> [header(), top("Default Example"), diff --git a/lib/inets/src/http_server/mod_esi.erl b/lib/inets/src/http_server/mod_esi.erl index 3a589ca5f0..3206d957d9 100644 --- a/lib/inets/src/http_server/mod_esi.erl +++ b/lib/inets/src/http_server/mod_esi.erl @@ -339,26 +339,21 @@ erl_scheme_webpage_whole(Mod, Func, Env, Input, ModData) -> {Headers, Body} = httpd_esi:parse_headers(lists:flatten(Response)), Length = httpd_util:flatlength(Body), - case httpd_esi:handle_headers(Headers) of - {proceed, AbsPath} -> - {proceed, [{real_name, httpd_util:split_path(AbsPath)} - | ModData#mod.data]}; - {ok, NewHeaders, StatusCode} -> - send_headers(ModData, StatusCode, - [{"content-length", - integer_to_list(Length)}| NewHeaders]), - case ModData#mod.method of - "HEAD" -> - {proceed, [{response, {already_sent, 200, 0}} | - ModData#mod.data]}; - _ -> - httpd_response:send_body(ModData, - StatusCode, Body), - {proceed, [{response, {already_sent, 200, - Length}} | - ModData#mod.data]} - end - end + {ok, NewHeaders, StatusCode} = httpd_esi:handle_headers(Headers), + send_headers(ModData, StatusCode, + [{"content-length", + integer_to_list(Length)}| NewHeaders]), + case ModData#mod.method of + "HEAD" -> + {proceed, [{response, {already_sent, 200, 0}} | + ModData#mod.data]}; + _ -> + httpd_response:send_body(ModData, + StatusCode, Body), + {proceed, [{response, {already_sent, 200, + Length}} | + ModData#mod.data]} + end end. %% New API that allows the dynamic wepage to be sent back to the client @@ -398,29 +393,23 @@ deliver_webpage_chunk(#mod{config_db = Db} = ModData, Pid, Timeout) -> {continue, _} = Continue -> Continue; {Headers, Body} -> - case httpd_esi:handle_headers(Headers) of - {proceed, AbsPath} -> - {proceed, [{real_name, httpd_util:split_path(AbsPath)} - | ModData#mod.data]}; - {ok, NewHeaders, StatusCode} -> - IsDisableChunkedSend = - httpd_response:is_disable_chunked_send(Db), - case (ModData#mod.http_version =/= "HTTP/1.1") or - (IsDisableChunkedSend) of - true -> - send_headers(ModData, StatusCode, - [{"connection", "close"} | - NewHeaders]); - false -> - send_headers(ModData, StatusCode, - [{"transfer-encoding", - "chunked"} | NewHeaders]) - end, - handle_body(Pid, ModData, Body, Timeout, length(Body), - IsDisableChunkedSend) - end; - timeout -> - send_headers(ModData, 504, [{"connection", "close"}]), + {ok, NewHeaders, StatusCode} = httpd_esi:handle_headers(Headers), + IsDisableChunkedSend = httpd_response:is_disable_chunked_send(Db), + case (ModData#mod.http_version =/= "HTTP/1.1") or + (IsDisableChunkedSend) of + true -> + send_headers(ModData, StatusCode, + [{"connection", "close"} | + NewHeaders]); + false -> + send_headers(ModData, StatusCode, + [{"transfer-encoding", + "chunked"} | NewHeaders]) + end, + handle_body(Pid, ModData, Body, Timeout, length(Body), + IsDisableChunkedSend); + timeout -> + send_headers(ModData, 504, [{"connection", "close"}]), httpd_socket:close(ModData#mod.socket_type, ModData#mod.socket), {proceed,[{response, {already_sent, 200, 0}} | ModData#mod.data]} end. @@ -560,15 +549,10 @@ eval(#mod{method = Method} = ModData, ESIBody, Modules) {ok, Response} -> {Headers, _} = httpd_esi:parse_headers(lists:flatten(Response)), - case httpd_esi:handle_headers(Headers) of - {ok, _, StatusCode} -> - {proceed,[{response, {StatusCode, Response}} | - ModData#mod.data]}; - {proceed, AbsPath} -> - {proceed, [{real_name, AbsPath} | - ModData#mod.data]} - end - end; + {ok, _, StatusCode} =httpd_esi:handle_headers(Headers), + {proceed,[{response, {StatusCode, Response}} | + ModData#mod.data]} + end; false -> {proceed,[{status, {403, ModData#mod.request_uri, diff --git a/lib/inets/test/http_format_SUITE.erl b/lib/inets/test/http_format_SUITE.erl index 4e10a97f58..9a13ed3d17 100644 --- a/lib/inets/test/http_format_SUITE.erl +++ b/lib/inets/test/http_format_SUITE.erl @@ -535,7 +535,7 @@ esi_parse_headers(Config) when is_list(Config) -> {"location","http://foo.bar.se"}], 302} = httpd_esi:handle_headers(Headers2), - {proceed,"/foo/bar.html"} = + {ok,[{"location","/foo/bar.html"}], 302} = httpd_esi:handle_headers("location:/foo/bar.html\r\n"). %%-------------------------------------------------------------------- diff --git a/lib/inets/test/httpc_SUITE.erl b/lib/inets/test/httpc_SUITE.erl index a39e786c79..cc166d522e 100644 --- a/lib/inets/test/httpc_SUITE.erl +++ b/lib/inets/test/httpc_SUITE.erl @@ -115,6 +115,7 @@ only_simulated() -> invalid_chunk_size, headers_dummy, headers_with_obs_fold, + headers_conflict_chunked_with_length, empty_response_header, remote_socket_close, remote_socket_close_async, @@ -978,7 +979,6 @@ headers_dummy(Config) when is_list(Config) -> {"If-Range", "Sat, 29 Oct 1994 19:43:31 GMT"}, {"If-Match", "*"}, {"Content-Type", "text/plain"}, - {"Content-Encoding", "chunked"}, {"Content-Length", "6"}, {"Content-Language", "en"}, {"Content-Location", "http://www.foobar.se"}, @@ -1004,6 +1004,18 @@ headers_with_obs_fold(Config) when is_list(Config) -> %%------------------------------------------------------------------------- +headers_conflict_chunked_with_length(doc) -> + ["Test the code for handling headers with both Transfer-Encoding" + "and Content-Length which must receive error in default (not relaxed) mode" + "and must receive successful response in relaxed mode"]; +headers_conflict_chunked_with_length(Config) when is_list(Config) -> + Request = {url(group_name(Config), "/headers_conflict_chunked_with_length.html", Config), []}, + {error, {could_not_parse_as_http, _}} = httpc:request(get, Request, [{relaxed, false}], []), + {ok,{{_,200,_},_,_}} = httpc:request(get, Request, [{relaxed, true}], []), + ok. + +%%------------------------------------------------------------------------- + invalid_headers(Config) -> Request = {url(group_name(Config), "/dummy.html", Config), [{"cookie", undefined}]}, {error, _} = httpc:request(get, Request, [], []). @@ -1869,7 +1881,6 @@ handle_uri(_,"/dummy_headers.html",_,_,Socket,_) -> %% user to evaluate. This is not a valid response %% it only tests that the header handling code works. Head = "HTTP/1.1 200 ok\r\n" ++ - "Content-Length:32\r\n" ++ "Pragma:1#no-cache\r\n" ++ "Via:1.0 fred, 1.1 nowhere.com (Apache/1.1)\r\n" ++ "Warning:1#pseudonym foobar\r\n" ++ @@ -1899,6 +1910,15 @@ handle_uri(_,"/obs_folded_headers.html",_,_,_,_) -> " b\r\n\r\n" "Hello"; +handle_uri(_,"/headers_conflict_chunked_with_length.html",_,_,Socket,_) -> + Head = "HTTP/1.1 200 ok\r\n" + "Content-Length:32\r\n" + "Transfer-Encoding:Chunked\r\n\r\n", + send(Socket, Head), + send(Socket, http_chunk:encode("<HTML><BODY>fo")), + send(Socket, http_chunk:encode("obar</BODY></HTML>")), + http_chunk:encode_last(); + handle_uri(_,"/capital_transfer_encoding.html",_,_,Socket,_) -> Head = "HTTP/1.1 200 ok\r\n" ++ "Transfer-Encoding:Chunked\r\n\r\n", diff --git a/lib/inets/test/httpd_mod.erl b/lib/inets/test/httpd_mod.erl index d9118aa1a4..2035b50248 100644 --- a/lib/inets/test/httpd_mod.erl +++ b/lib/inets/test/httpd_mod.erl @@ -779,9 +779,14 @@ esi(Type, Port, Host, Node) -> [{statuscode, 200}, {no_header, "cache-control"}, {version, "HTTP/1.0"}]), + ok = httpd_test_lib:verify_request(Type, Host, Port, Node, + "GET /cgi-bin/erl/httpd_example:new_status_and_location" + " HTTP/1.1\r\n\r\n", + [{statuscode, 201}, + {header, "Location"}, + {version, "HTTP/1.1"}]), ok. - %%-------------------------------------------------------------------- get(Type, Port, Host, Node) -> ok = httpd_test_lib:verify_request(Type, Host, Port, Node, diff --git a/lib/inets/vsn.mk b/lib/inets/vsn.mk index 34b6902747..108d259823 100644 --- a/lib/inets/vsn.mk +++ b/lib/inets/vsn.mk @@ -19,6 +19,6 @@ # %CopyrightEnd% APPLICATION = inets -INETS_VSN = 6.4.2 +INETS_VSN = 6.4.3 PRE_VSN = APP_VSN = "$(APPLICATION)-$(INETS_VSN)$(PRE_VSN)" diff --git a/lib/kernel/test/gen_tcp_misc_SUITE.erl b/lib/kernel/test/gen_tcp_misc_SUITE.erl index 331864b5de..e47023d201 100644 --- a/lib/kernel/test/gen_tcp_misc_SUITE.erl +++ b/lib/kernel/test/gen_tcp_misc_SUITE.erl @@ -1572,52 +1572,56 @@ fill_sendq(Config) when is_list(Config) -> Master = self(), Server = spawn_link(fun () -> - {ok,L} = gen_tcp:listen - (0, [{active,false},binary, - {reuseaddr,true},{packet,0}]), + {ok,L} = gen_tcp:listen(0, [{active,false},binary, + {reuseaddr,true},{packet,0}]), {ok,Port} = inet:port(L), Master ! {self(),client, fill_sendq_client(Port, Master)}, fill_sendq_srv(L, Master) end), io:format("~p Server~n", [Server]), - receive {Server,client,Client} -> - io:format("~p Client~n", [Client]), - receive {Server,reader,Reader} -> - io:format("~p Reader~n", [Reader]), - fill_sendq_loop(Server, Client, Reader) + receive + {Server,client,Client} -> + io:format("~p Client~n", [Client]), + receive + {Server,reader,Reader} -> + io:format("~p Reader~n", [Reader]), + fill_sendq_loop(Server, Client, Reader) end end. fill_sendq_loop(Server, Client, Reader) -> %% Master %% - receive {Server,send} -> + receive + {Server,send} -> fill_sendq_loop(Server, Client, Reader) after 2000 -> %% Send queue full, sender blocked -> close client. io:format("Send timeout, closing Client...~n", []), Client ! {self(),close}, - receive {Server,[{error,closed}]} -> - io:format("Got server closed.~n"), - receive {Reader,[{error,closed}]} -> - io:format - ("Got reader closed.~n"), - ok - after 3000 -> - ct:fail({timeout,{closed,reader}}) - end; - {Reader,[{error,closed}]} -> - io:format("Got reader closed.~n"), - receive {Server,[{error,closed}]} -> - io:format("Got server closed~n"), - ok - after 3000 -> - ct:fail({timeout,{closed,server}}) - end - after 3000 -> - ct:fail({timeout,{closed,[server,reader]}}) - end + receive + {Server,[{error,closed}]} -> + io:format("Got server closed.~n"), + receive + {Reader,[{error,closed}]} -> + io:format("Got reader closed.~n"), + ok + after 3000 -> + ct:fail({timeout,{closed,reader}}) + end; + {Reader,[{error,closed}]} -> + io:format("Got reader closed.~n"), + receive + {Server,[{error,closed}]} -> + io:format("Got server closed~n"), + ok + after 3000 -> + ct:fail({timeout,{closed,server}}) + end + after 3000 -> + ct:fail({timeout,{closed,[server,reader]}}) + end end. fill_sendq_srv(L, Master) -> diff --git a/lib/observer/src/cdv_bin_cb.erl b/lib/observer/src/cdv_bin_cb.erl index 5502869973..a4a542297c 100644 --- a/lib/observer/src/cdv_bin_cb.erl +++ b/lib/observer/src/cdv_bin_cb.erl @@ -71,6 +71,8 @@ hex_binary_fun(Bin) -> plain_html(io_lib:format("~s",[S])) end. +format_hex(<<>>,_) -> + []; format_hex(<<B1:4,B2:4>>,_) -> [integer_to_list(B1,16),integer_to_list(B2,16)]; format_hex(<<B1:4,B2:4,Bin/binary>>,0) -> diff --git a/lib/observer/src/observer_html_lib.erl b/lib/observer/src/observer_html_lib.erl index 68381bcc7b..8202f91030 100644 --- a/lib/observer/src/observer_html_lib.erl +++ b/lib/observer/src/observer_html_lib.erl @@ -355,11 +355,11 @@ href_proc_bin(From, T, Acc, LTB) -> PreviewStr end end; - [PreviewIntStr,SizeStr,Md5] when From =:= obs -> + [PreviewIntStr,PreviewBitSizeStr,SizeStr,Md5] when From =:= obs -> Size = list_to_integer(SizeStr), PreviewInt = list_to_integer(PreviewIntStr), - PrevSize = (trunc(math:log2(PreviewInt)/8)+1)*8, - PreviewStr = preview_string(Size,<<PreviewInt:PrevSize>>), + PreviewBitSize = list_to_integer(PreviewBitSizeStr), + PreviewStr = preview_string(Size,<<PreviewInt:PreviewBitSize>>), if LTB -> href("TARGET=\"expanded\"", ["#OBSBinary?key1="++PreviewIntStr++ diff --git a/lib/observer/src/observer_lib.erl b/lib/observer/src/observer_lib.erl index e5ffe61d25..52e6c3e52a 100644 --- a/lib/observer/src/observer_lib.erl +++ b/lib/observer/src/observer_lib.erl @@ -810,7 +810,7 @@ progress_dialog_destroy({Dialog,_,_}) -> make_obsbin(Bin,Tab) -> Size = byte_size(Bin), - Preview = + {Preview,PreviewBitSize} = try %% The binary might be a unicode string, in which case we %% don't want to split it in the middle of a grapheme @@ -819,14 +819,14 @@ make_obsbin(Bin,Tab) -> PB1 = string:slice(Bin,0,PL1), PS1 = byte_size(PB1) * 8, <<P1:PS1>> = PB1, - P1 + {P1,PS1} catch _:_ -> %% Probably not a string, so just split anywhere PS2 = min(Size, 10) * 8, <<P2:PS2, _/binary>> = Bin, - P2 + {P2,PS2} end, Hash = erlang:phash2(Bin), Key = {Preview, Size, Hash}, ets:insert(Tab, {Key,Bin}), - ['#OBSBin',Preview,Size,Hash]. + ['#OBSBin',Preview,PreviewBitSize,Size,Hash]. diff --git a/lib/snmp/test/snmp_agent_test.erl b/lib/snmp/test/snmp_agent_test.erl index 3c1a6f2afd..2ed2c4580c 100644 --- a/lib/snmp/test/snmp_agent_test.erl +++ b/lib/snmp/test/snmp_agent_test.erl @@ -605,7 +605,12 @@ init_per_group(multiple_reqs_3 = GroupName, Config) -> init_per_group(test_multi_threaded = GroupName, Config) -> init_mt(snmp_test_lib:init_group_top_dir(GroupName, Config)); init_per_group(test_v3 = GroupName, Config) -> - init_v3(snmp_test_lib:init_group_top_dir(GroupName, Config)); + case snmp_test_lib:crypto_start() of + ok -> + init_v3(snmp_test_lib:init_group_top_dir(GroupName, Config)); + _ -> + {skip, "Crypto did not start"} + end; init_per_group(test_v1_v2 = GroupName, Config) -> init_v1_v2(snmp_test_lib:init_group_top_dir(GroupName, Config)); init_per_group(test_v2 = GroupName, Config) -> @@ -631,11 +636,26 @@ init_per_group(mib_storage_varm_dets = GroupName, Config) -> init_varm_mib_storage_dets( snmp_test_lib:init_group_top_dir(GroupName, Config)); init_per_group(mib_storage_size_check_mnesia = GroupName, Config) -> - init_size_check_msm(snmp_test_lib:init_group_top_dir(GroupName, Config)); + case snmp_test_lib:crypto_start() of + ok -> + init_size_check_msm(snmp_test_lib:init_group_top_dir(GroupName, Config)); + _ -> + {skip, "Crypto did not start"} + end; init_per_group(mib_storage_size_check_dets = GroupName, Config) -> - init_size_check_msd(snmp_test_lib:init_group_top_dir(GroupName, Config)); + case snmp_test_lib:crypto_start() of + ok -> + init_size_check_msm(snmp_test_lib:init_group_top_dir(GroupName, Config)); + _ -> + {skip, "Crypto did not start"} + end; init_per_group(mib_storage_size_check_ets = GroupName, Config) -> - init_size_check_mse(snmp_test_lib:init_group_top_dir(GroupName, Config)); + case snmp_test_lib:crypto_start() of + ok -> + init_size_check_msm(snmp_test_lib:init_group_top_dir(GroupName, Config)); + _ -> + {skip, "Crypto did not start"} + end; init_per_group(mib_storage_mnesia = GroupName, Config) -> init_mib_storage_mnesia(snmp_test_lib:init_group_top_dir(GroupName, Config)); diff --git a/lib/snmp/test/snmp_manager_test.erl b/lib/snmp/test/snmp_manager_test.erl index 4bfeb0f8d1..6ced55f0cc 100644 --- a/lib/snmp/test/snmp_manager_test.erl +++ b/lib/snmp/test/snmp_manager_test.erl @@ -156,16 +156,25 @@ init_per_suite(Config0) when is_list(Config0) -> ?DBG("init_per_suite -> entry with" "~n Config0: ~p", [Config0]), - Config1 = snmp_test_lib:init_suite_top_dir(?MODULE, Config0), - Config2 = snmp_test_lib:fix_data_dir(Config1), - - %% Mib-dirs - %% data_dir is trashed by the test-server / common-test - %% so there is no point in fixing it... - MibDir = snmp_test_lib:lookup(data_dir, Config2), - StdMibDir = filename:join([code:priv_dir(snmp), "mibs"]), - - [{mib_dir, MibDir}, {std_mib_dir, StdMibDir} | Config2]. + %% Preferably this test SUITE should be divided into groups + %% so that if crypto does not work only v3 tests that + %% need crypto will be skipped, but as this is only a + %% problem with one legacy test machine, we will procrastinate + %% until we have a more important reason to fix this. + case snmp_test_lib:crypto_start() of + ok -> + Config1 = snmp_test_lib:init_suite_top_dir(?MODULE, Config0), + Config2 = snmp_test_lib:fix_data_dir(Config1), + %% Mib-dirs + %% data_dir is trashed by the test-server / common-test + %% so there is no point in fixing it... + MibDir = snmp_test_lib:lookup(data_dir, Config2), + StdMibDir = filename:join([code:priv_dir(snmp), "mibs"]), + + [{mib_dir, MibDir}, {std_mib_dir, StdMibDir} | Config2]; + _ -> + {skip, "Crypto did not start"} + end. end_per_suite(Config) when is_list(Config) -> diff --git a/lib/ssh/src/ssh.erl b/lib/ssh/src/ssh.erl index 1a5d48baca..032d87bdad 100644 --- a/lib/ssh/src/ssh.erl +++ b/lib/ssh/src/ssh.erl @@ -188,6 +188,7 @@ daemon(Port) -> daemon(Socket, UserOptions) when is_port(Socket) -> try #{} = Options = ssh_options:handle_options(server, UserOptions), + case valid_socket_to_use(Socket, ?GET_OPT(transport,Options)) of ok -> {ok, {IP,Port}} = inet:sockname(Socket), @@ -461,6 +462,9 @@ open_listen_socket(_Host0, Port0, Options0) -> %%%---------------------------------------------------------------- finalize_start(Host, Port, Profile, Options0, F) -> try + %% throws error:Error if no usable hostkey is found + ssh_connection_handler:available_hkey_algorithms(server, Options0), + sshd_sup:start_child(Host, Port, Profile, Options0) of {error, {already_started, _}} -> @@ -470,6 +474,8 @@ finalize_start(Host, Port, Profile, Options0, F) -> Result = {ok,_} -> F(Options0, Result) catch + error:{shutdown,Err} -> + {error,Err}; exit:{noproc, _} -> {error, ssh_not_started} end. diff --git a/lib/ssh/src/ssh_auth.erl b/lib/ssh/src/ssh_auth.erl index ac64a7bf14..894877f8bf 100644 --- a/lib/ssh/src/ssh_auth.erl +++ b/lib/ssh/src/ssh_auth.erl @@ -145,14 +145,17 @@ get_public_key(SigAlg, #ssh{opts = Opts}) -> case KeyCb:user_key(KeyAlg, [{key_cb_private,KeyCbOpts}|UserOpts]) of {ok, PrivKey} -> try + %% Check the key - the KeyCb may be a buggy plugin + true = ssh_transport:valid_key_sha_alg(PrivKey, KeyAlg), Key = ssh_transport:extract_public_key(PrivKey), public_key:ssh_encode(Key, ssh2_pubkey) of PubKeyBlob -> {ok,{PrivKey,PubKeyBlob}} catch _:_ -> - not_ok + not_ok end; + _Error -> not_ok end. diff --git a/lib/ssh/src/ssh_connection_handler.erl b/lib/ssh/src/ssh_connection_handler.erl index 4158a52a27..802bf62570 100644 --- a/lib/ssh/src/ssh_connection_handler.erl +++ b/lib/ssh/src/ssh_connection_handler.erl @@ -46,6 +46,7 @@ %%% Internal application API -export([start_connection/4, + available_hkey_algorithms/2, open_channel/6, request/6, request/7, reply_request/3, @@ -432,13 +433,12 @@ init_ssh_record(Role, Socket, Opts) -> init_ssh_record(Role, Socket, PeerAddr, Opts). init_ssh_record(Role, _Socket, PeerAddr, Opts) -> - KeyCb = ?GET_OPT(key_cb, Opts), AuthMethods = ?GET_OPT(auth_methods, Opts), S0 = #ssh{role = Role, - key_cb = KeyCb, + key_cb = ?GET_OPT(key_cb, Opts), opts = Opts, userauth_supported_methods = AuthMethods, - available_host_keys = supported_host_keys(Role, KeyCb, Opts), + available_host_keys = available_hkey_algorithms(Role, Opts), random_length_padding = ?GET_OPT(max_random_length_padding, Opts) }, @@ -1544,44 +1544,42 @@ peer_role(client) -> server; peer_role(server) -> client. %%-------------------------------------------------------------------- -supported_host_keys(client, _, Options) -> - try - find_sup_hkeys(Options) - of - [] -> +available_hkey_algorithms(Role, Options) -> + KeyCb = ?GET_OPT(key_cb, Options), + case [A || A <- available_hkey_algos(Options), + (Role==client) orelse available_host_key(KeyCb, A, Options) + ] of + + [] when Role==client -> error({shutdown, "No public key algs"}); - Algs -> - [atom_to_list(A) || A<-Algs] - catch - exit:Reason -> - error({shutdown, Reason}) - end; -supported_host_keys(server, KeyCb, Options) -> - [atom_to_list(A) || A <- find_sup_hkeys(Options), - available_host_key(KeyCb, A, Options) - ]. + [] when Role==server -> + error({shutdown, "No host key available"}); -find_sup_hkeys(Options) -> - case proplists:get_value(public_key, - ?GET_OPT(preferred_algorithms,Options) - ) - of - undefined -> - ssh_transport:default_algorithms(public_key); - L -> - NonSupported = L--ssh_transport:supported_algorithms(public_key), - L -- NonSupported + Algs -> + [atom_to_list(A) || A<-Algs] end. +available_hkey_algos(Options) -> + SupAlgos = ssh_transport:supported_algorithms(public_key), + HKeys = proplists:get_value(public_key, + ?GET_OPT(preferred_algorithms,Options) + ), + NonSupported = HKeys -- SupAlgos, + AvailableAndSupported = HKeys -- NonSupported, + AvailableAndSupported. + %% Alg :: atom() available_host_key({KeyCb,KeyCbOpts}, Alg, Opts) -> UserOpts = ?GET_OPT(user_options, Opts), case KeyCb:host_key(Alg, [{key_cb_private,KeyCbOpts}|UserOpts]) of - {ok,_} -> true; - _ -> false + {ok,Key} -> + %% Check the key - the KeyCb may be a buggy plugin + ssh_transport:valid_key_sha_alg(Key, Alg); + _ -> + false end. diff --git a/lib/ssh/src/ssh_transport.erl b/lib/ssh/src/ssh_transport.erl index e92c727559..d8f7a96c15 100644 --- a/lib/ssh/src/ssh_transport.erl +++ b/lib/ssh/src/ssh_transport.erl @@ -795,8 +795,14 @@ get_host_key(SSH, SignAlg) -> #ssh{key_cb = {KeyCb,KeyCbOpts}, opts = Opts} = SSH, UserOpts = ?GET_OPT(user_options, Opts), case KeyCb:host_key(SignAlg, [{key_cb_private,KeyCbOpts}|UserOpts]) of - {ok, PrivHostKey} -> PrivHostKey; - Result -> exit({error, {Result, unsupported_key_type}}) + {ok, PrivHostKey} -> + %% Check the key - the KeyCb may be a buggy plugin + case valid_key_sha_alg(PrivHostKey, SignAlg) of + true -> PrivHostKey; + false -> exit({error, bad_hostkey}) + end; + Result -> + exit({error, {Result, unsupported_key_type}}) end. extract_public_key(#'RSAPrivateKey'{modulus = N, publicExponent = E}) -> @@ -1830,11 +1836,14 @@ valid_key_sha_alg(#'RSAPrivateKey'{}, 'ssh-rsa' ) -> true; valid_key_sha_alg({_, #'Dss-Parms'{}}, 'ssh-dss') -> true; valid_key_sha_alg(#'DSAPrivateKey'{}, 'ssh-dss') -> true; -valid_key_sha_alg({#'ECPoint'{},{namedCurve,OID}}, Alg) -> sha(OID) == sha(Alg); -valid_key_sha_alg(#'ECPrivateKey'{parameters = {namedCurve,OID}}, Alg) -> sha(OID) == sha(Alg); +valid_key_sha_alg({#'ECPoint'{},{namedCurve,OID}}, Alg) -> valid_key_sha_alg_ec(OID, Alg); +valid_key_sha_alg(#'ECPrivateKey'{parameters = {namedCurve,OID}}, Alg) -> valid_key_sha_alg_ec(OID, Alg); valid_key_sha_alg(_, _) -> false. - +valid_key_sha_alg_ec(OID, Alg) -> + Curve = public_key:oid2ssh_curvename(OID), + Alg == list_to_atom("ecdsa-sha2-" ++ binary_to_list(Curve)). + public_algo(#'RSAPublicKey'{}) -> 'ssh-rsa'; % FIXME: Not right with draft-curdle-rsa-sha2 public_algo({_, #'Dss-Parms'{}}) -> 'ssh-dss'; diff --git a/lib/ssh/test/property_test/ssh_eqc_client_info_timing.erl b/lib/ssh/test/property_test/ssh_eqc_client_info_timing.erl index c07140dc43..19e2754eba 100644 --- a/lib/ssh/test/property_test/ssh_eqc_client_info_timing.erl +++ b/lib/ssh/test/property_test/ssh_eqc_client_info_timing.erl @@ -57,9 +57,9 @@ %%% Properties: -prop_seq(_Config) -> +prop_seq(Config) -> {ok,Pid} = ssh_eqc_event_handler:add_report_handler(), - {_, _, Port} = init_daemon(), + {_, _, Port} = init_daemon(Config), numtests(1000, ?FORALL(Delay, choose(0,100),%% Micro seconds try @@ -86,7 +86,8 @@ any_relevant_error_report(Pid) -> end, Reports). %%%================================================================ -init_daemon() -> +init_daemon(Config) -> ok = begin ssh:stop(), ssh:start() end, - ssh_test_lib:daemon([]). + DataDir = proplists:get_value(data_dir, Config), + ssh_test_lib:daemon([{system_dir,DataDir}]). diff --git a/lib/ssh/test/ssh_basic_SUITE.erl b/lib/ssh/test/ssh_basic_SUITE.erl index db2ae241e5..202b0afe57 100644 --- a/lib/ssh/test/ssh_basic_SUITE.erl +++ b/lib/ssh/test/ssh_basic_SUITE.erl @@ -46,6 +46,7 @@ exec_key_differs2/1, exec_key_differs3/1, exec_key_differs_fail/1, + fail_daemon_start/1, idle_time_client/1, idle_time_server/1, inet6_option/1, @@ -105,6 +106,7 @@ all() -> {group, host_user_key_differs}, {group, key_cb}, {group, internal_error}, + {group, rsa_host_key_is_actualy_ecdsa}, daemon_already_started, double_close, daemon_opt_fd, @@ -121,6 +123,7 @@ groups() -> {ecdsa_sha2_nistp256_key, [], basic_tests()}, {ecdsa_sha2_nistp384_key, [], basic_tests()}, {ecdsa_sha2_nistp521_key, [], basic_tests()}, + {rsa_host_key_is_actualy_ecdsa, [], [fail_daemon_start]}, {host_user_key_differs, [], [exec_key_differs1, exec_key_differs2, exec_key_differs3, @@ -180,6 +183,31 @@ init_per_group(rsa_key, Config) -> false -> {skip, unsupported_pub_key} end; +init_per_group(rsa_host_key_is_actualy_ecdsa, Config) -> + case + lists:member('ssh-rsa', + ssh_transport:default_algorithms(public_key)) and + lists:member('ecdsa-sha2-nistp256', + ssh_transport:default_algorithms(public_key)) + of + true -> + DataDir = proplists:get_value(data_dir, Config), + PrivDir = proplists:get_value(priv_dir, Config), + ssh_test_lib:setup_ecdsa("256", DataDir, PrivDir), + %% The following sets up bad rsa keys: + begin + UserDir = PrivDir, + System = filename:join(UserDir, "system"), + file:copy(filename:join(DataDir, "id_rsa"), filename:join(UserDir, "id_rsa")), + file:rename(filename:join(System, "ssh_host_ecdsa_key"), filename:join(System, "ssh_host_rsa_key")), + file:rename(filename:join(System, "ssh_host_ecdsa_key.pub"), filename:join(System, "ssh_host_rsa_key.pub")), + ssh_test_lib:setup_rsa_known_host(DataDir, UserDir), + ssh_test_lib:setup_rsa_auth_keys(DataDir, UserDir) + end, + Config; + false -> + {skip, unsupported_pub_key} + end; init_per_group(ecdsa_sha2_nistp256_key, Config) -> case lists:member('ecdsa-sha2-nistp256', ssh_transport:default_algorithms(public_key)) of @@ -304,7 +332,8 @@ init_per_group(internal_error, Config) -> DataDir = proplists:get_value(data_dir, Config), PrivDir = proplists:get_value(priv_dir, Config), ssh_test_lib:setup_dsa(DataDir, PrivDir), - file:delete(filename:join(PrivDir, "system/ssh_host_dsa_key")), + %% In the test case the key will be deleted after the daemon start: + %% ... file:delete(filename:join(PrivDir, "system/ssh_host_dsa_key")), Config; init_per_group(dir_options, Config) -> PrivDir = proplists:get_value(priv_dir, Config), @@ -868,12 +897,17 @@ key_callback_options(Config) when is_list(Config) -> %%% Test that client does not hang if disconnects due to internal error internal_error(Config) when is_list(Config) -> process_flag(trap_exit, true), - SystemDir = filename:join(proplists:get_value(priv_dir, Config), system), + PrivDir = proplists:get_value(priv_dir, Config), UserDir = proplists:get_value(priv_dir, Config), + SystemDir = filename:join(PrivDir, system), {Pid, Host, Port} = ssh_test_lib:daemon([{system_dir, SystemDir}, {user_dir, UserDir}, {failfun, fun ssh_test_lib:failfun/2}]), + + %% Now provoke an error in the following connect: + file:delete(filename:join(PrivDir, "system/ssh_host_dsa_key")), + {error, Error} = ssh:connect(Host, Port, [{silently_accept_hosts, true}, {user_dir, UserDir}, @@ -902,6 +936,17 @@ send(Config) when is_list(Config) -> %%-------------------------------------------------------------------- +%%% +fail_daemon_start(Config) when is_list(Config) -> + process_flag(trap_exit, true), + SystemDir = filename:join(proplists:get_value(priv_dir, Config), system), + UserDir = proplists:get_value(priv_dir, Config), + + {error,_} = ssh_test_lib:daemon([{system_dir, SystemDir}, + {user_dir, UserDir}, + {failfun, fun ssh_test_lib:failfun/2}]). + +%%-------------------------------------------------------------------- %%% Test ssh:connection_info([peername, sockname]) peername_sockname(Config) when is_list(Config) -> process_flag(trap_exit, true), @@ -1300,14 +1345,11 @@ shell_exit_status(Config) when is_list(Config) -> %%-------------------------------------------------------------------- %% Due to timing the error message may or may not be delivered to %% the "tcp-application" before the socket closed message is recived -check_error("Invalid state") -> - ok; -check_error("Connection closed") -> - ok; -check_error("Selection of key exchange algorithm failed"++_) -> - ok; -check_error(Error) -> - ct:fail(Error). +check_error("Invalid state") -> ok; +check_error("Connection closed") -> ok; +check_error("Selection of key exchange algorithm failed"++_) -> ok; +check_error("No host key available") -> ok; +check_error(Error) -> ct:fail(Error). basic_test(Config) -> ClientOpts = proplists:get_value(client_opts, Config), diff --git a/lib/ssh/test/ssh_protocol_SUITE.erl b/lib/ssh/test/ssh_protocol_SUITE.erl index 74f802cf57..3e3e151781 100644 --- a/lib/ssh/test/ssh_protocol_SUITE.erl +++ b/lib/ssh/test/ssh_protocol_SUITE.erl @@ -630,11 +630,12 @@ client_handles_keyboard_interactive_0_pwds(Config) -> %%%-------------------------------------------------------------------- -client_info_line(_Config) -> +client_info_line(Config) -> %% A client must not send an info-line. If it does, the server should handle %% handle this gracefully {ok,Pid} = ssh_eqc_event_handler:add_report_handler(), - {_, _, Port} = ssh_test_lib:daemon([]), + DataDir = proplists:get_value(data_dir, Config), + {_, _, Port} = ssh_test_lib:daemon([{system_dir,DataDir}]), %% Fake client: {ok,S} = gen_tcp:connect("localhost",Port,[]), diff --git a/lib/stdlib/test/ets_SUITE.erl b/lib/stdlib/test/ets_SUITE.erl index 5a5e282998..f329a07b7a 100644 --- a/lib/stdlib/test/ets_SUITE.erl +++ b/lib/stdlib/test/ets_SUITE.erl @@ -1682,7 +1682,7 @@ do_random_test() -> ets:delete(Set), verify_etsmem(EtsMem). -%% Ttest various variants of update_element. +%% Test various variants of update_element. update_element(Config) when is_list(Config) -> EtsMem = etsmem(), repeat_for_opts(fun update_element_opts/1), diff --git a/lib/stdlib/test/qlc_SUITE.erl b/lib/stdlib/test/qlc_SUITE.erl index 5e9e03e410..949142ec77 100644 --- a/lib/stdlib/test/qlc_SUITE.erl +++ b/lib/stdlib/test/qlc_SUITE.erl @@ -7871,7 +7871,7 @@ run_test(Config, Extra, {cres, Body, Opts, ExpectedCompileReturn}) -> {module, _} = code:load_abs(AbsFile, Mod), Ms0 = erlang:process_info(self(),messages), - Before = {{get(), lists:sort(ets:all()), Ms0}, pps()}, + Before = {{lget(), lists:sort(ets:all()), Ms0}, pps()}, %% Prepare the check that the qlc module does not call qlc_pt. _ = [unload_pt() || {file, Name} <- [code:is_loaded(qlc_pt)], @@ -7903,7 +7903,7 @@ run_test(Config, Extra, Body) -> wait_for_expected(R, {Strict0,PPS0}=Before, SourceFile, Wait) -> Ms = erlang:process_info(self(),messages), - After = {_,PPS1} = {{get(), lists:sort(ets:all()), Ms}, pps()}, + After = {_,PPS1} = {{lget(), lists:sort(ets:all()), Ms}, pps()}, case {R, After} of {ok, Before} -> ok; @@ -7931,6 +7931,18 @@ wait_for_expected(R, {Strict0,PPS0}=Before, SourceFile, Wait) -> expected({ok,Before}, {R,After}, SourceFile) end. +%% The qlc modules uses the process dictionary for storing names of files. +lget() -> + lists:sort([T || {K, _} = T <- get(), is_qlc_key(K)]). + +%% Copied from the qlc module. +-define(LCACHE_FILE(Ref), {Ref, '$_qlc_cache_tmpfiles_'}). +-define(MERGE_JOIN_FILE, '$_qlc_merge_join_tmpfiles_'). + +is_qlc_key(?LCACHE_FILE(_)) -> true; +is_qlc_key(?MERGE_JOIN_FILE) -> true; +is_qlc_key(_) -> false. + unload_pt() -> erlang:garbage_collect(), % get rid of references to qlc_pt... _ = code:purge(qlc_pt), diff --git a/lib/tools/test/xref_SUITE.erl b/lib/tools/test/xref_SUITE.erl index d651cbcfee..018f632948 100644 --- a/lib/tools/test/xref_SUITE.erl +++ b/lib/tools/test/xref_SUITE.erl @@ -1112,27 +1112,16 @@ read_expected(Version) -> {POS7+2,{FF,{erlang,spawn_opt,4}}}, {POS8+1,{FF,{hej,san,1}}}, {POS8+4,{FF,{a,b,1}}}, - {POS8+4,{FF,{erlang,apply,2}}}, - {POS8+5,{FF,{erlang,apply,2}}}, {POS8+6,{FF,{m,f,1}}}, {POS9+1,{FF,{read,bi,0}}}, {POS9+2,{FF,{a,b,1}}}, - {POS9+2,{FF,{erlang,apply,2}}}, - {POS9+3,{FF,{erlang,apply,2}}}, - {POS9+4,{FF,{erlang,apply,2}}}, {POS9+4,{FF,{erlang,not_a_function,1}}}, {POS9+5,{FF,{mod,func,2}}}, {POS9+6,{FF,{erlang,apply,1}}}, - {POS9+7,{FF,{erlang,apply,2}}}, {POS9+7,{FF,{math,add3,1}}}, {POS9+8,{FF,{q,f,1}}}, - {POS10+4,{FF,{erlang,apply,2}}}, {POS10+5,{FF,{mod1,fun1,1}}}, - {POS11+6,{FF,{erlang,apply,2}}}, - {POS12+1,{FF,{erlang,apply,2}}}, - {POS12+4,{FF,{erlang,apply,2}}}, {POS12+5,{FF,{m3,f3,2}}}, - {POS12+7,{FF,{erlang,apply,2}}}, {POS13+1,{FF,{dm,df,1}}}, {POS13+6,{{read,bi,0},{foo,module_info,0}}}, {POS13+7,{{read,bi,0},{read,module_info,0}}}, @@ -1162,15 +1151,26 @@ read_expected(Version) -> {POS3+3, {FF,{erlang,spawn_link,3}}}, {POS3+4, {FF,{erlang,spawn_link,3}}}, {POS6+4, {FF,{erlang,spawn,3}}}, + {POS8+4,{FF,{erlang,apply,2}}}, + {POS8+5,{FF,{erlang,apply,2}}}, {POS8+6,{FF,{erlang,apply,3}}}, {POS8+7,{FF,{erlang,apply,3}}}, {POS9+1,{FF,{erlang,apply,3}}}, + {POS9+2,{FF,{erlang,apply,2}}}, + {POS9+3,{FF,{erlang,apply,2}}}, + {POS9+4,{FF,{erlang,apply,2}}}, {POS9+5,{FF,{erlang,apply,3}}}, + {POS9+7,{FF,{erlang,apply,2}}}, + {POS10+4,{FF,{erlang,apply,2}}}, {POS11+1,{FF,{erlang,apply,3}}}, {POS11+2,{FF,{erlang,apply,3}}}, {POS11+3,{FF,{erlang,apply,3}}}, {POS11+4,{FF,{erlang,apply,3}}}, + {POS11+6,{FF,{erlang,apply,2}}}, + {POS12+1,{FF,{erlang,apply,2}}}, + {POS12+4,{FF,{erlang,apply,2}}}, {POS12+5,{FF,{erlang,apply,3}}}, + {POS12+7,{FF,{erlang,apply,2}}}, {POS12+8,{FF,{erlang,apply,3}}}, {POS13+5, {{read,bi,0},{erlang,length,1}}}, {POS14+3, {{read,bi,0},{erlang,length,1}}}], diff --git a/otp_versions.table b/otp_versions.table index ba3bccfe67..0cb8c9aa0b 100644 --- a/otp_versions.table +++ b/otp_versions.table @@ -1,3 +1,4 @@ +OTP-20.1.4 : inets-6.4.3 # asn1-5.0.3 common_test-1.15.2 compiler-7.1.3 cosEvent-2.2.1 cosEventDomain-1.2.1 cosFileTransfer-1.2.1 cosNotification-1.2.2 cosProperty-1.2.2 cosTime-1.2.2 cosTransactions-1.3.2 crypto-4.1 debugger-4.2.3 dialyzer-3.2.2 diameter-2.1.2 edoc-0.9.1 eldap-1.2.2 erl_docgen-0.7.1 erl_interface-3.10 erts-9.1.3 et-1.6.1 eunit-2.3.4 hipe-3.16.1 ic-4.4.2 jinterface-1.8 kernel-5.4 megaco-3.18.2 mnesia-4.15.1 observer-2.5 odbc-2.12 orber-3.8.3 os_mon-2.4.3 otp_mibs-1.1.1 parsetools-2.1.5 public_key-1.5 reltool-0.7.5 runtime_tools-1.12.2 sasl-3.1 snmp-5.2.8 ssh-4.6.1 ssl-8.2.1 stdlib-3.4.2 syntax_tools-2.1.3 tools-2.11 wx-1.8.2 xmerl-1.3.15 : OTP-20.1.3 : diameter-2.1.2 erts-9.1.3 snmp-5.2.8 # asn1-5.0.3 common_test-1.15.2 compiler-7.1.3 cosEvent-2.2.1 cosEventDomain-1.2.1 cosFileTransfer-1.2.1 cosNotification-1.2.2 cosProperty-1.2.2 cosTime-1.2.2 cosTransactions-1.3.2 crypto-4.1 debugger-4.2.3 dialyzer-3.2.2 edoc-0.9.1 eldap-1.2.2 erl_docgen-0.7.1 erl_interface-3.10 et-1.6.1 eunit-2.3.4 hipe-3.16.1 ic-4.4.2 inets-6.4.2 jinterface-1.8 kernel-5.4 megaco-3.18.2 mnesia-4.15.1 observer-2.5 odbc-2.12 orber-3.8.3 os_mon-2.4.3 otp_mibs-1.1.1 parsetools-2.1.5 public_key-1.5 reltool-0.7.5 runtime_tools-1.12.2 sasl-3.1 ssh-4.6.1 ssl-8.2.1 stdlib-3.4.2 syntax_tools-2.1.3 tools-2.11 wx-1.8.2 xmerl-1.3.15 : OTP-20.1.2 : diameter-2.1.1 erts-9.1.2 # asn1-5.0.3 common_test-1.15.2 compiler-7.1.3 cosEvent-2.2.1 cosEventDomain-1.2.1 cosFileTransfer-1.2.1 cosNotification-1.2.2 cosProperty-1.2.2 cosTime-1.2.2 cosTransactions-1.3.2 crypto-4.1 debugger-4.2.3 dialyzer-3.2.2 edoc-0.9.1 eldap-1.2.2 erl_docgen-0.7.1 erl_interface-3.10 et-1.6.1 eunit-2.3.4 hipe-3.16.1 ic-4.4.2 inets-6.4.2 jinterface-1.8 kernel-5.4 megaco-3.18.2 mnesia-4.15.1 observer-2.5 odbc-2.12 orber-3.8.3 os_mon-2.4.3 otp_mibs-1.1.1 parsetools-2.1.5 public_key-1.5 reltool-0.7.5 runtime_tools-1.12.2 sasl-3.1 snmp-5.2.7 ssh-4.6.1 ssl-8.2.1 stdlib-3.4.2 syntax_tools-2.1.3 tools-2.11 wx-1.8.2 xmerl-1.3.15 : OTP-20.1.1 : compiler-7.1.3 erts-9.1.1 ssh-4.6.1 # asn1-5.0.3 common_test-1.15.2 cosEvent-2.2.1 cosEventDomain-1.2.1 cosFileTransfer-1.2.1 cosNotification-1.2.2 cosProperty-1.2.2 cosTime-1.2.2 cosTransactions-1.3.2 crypto-4.1 debugger-4.2.3 dialyzer-3.2.2 diameter-2.1 edoc-0.9.1 eldap-1.2.2 erl_docgen-0.7.1 erl_interface-3.10 et-1.6.1 eunit-2.3.4 hipe-3.16.1 ic-4.4.2 inets-6.4.2 jinterface-1.8 kernel-5.4 megaco-3.18.2 mnesia-4.15.1 observer-2.5 odbc-2.12 orber-3.8.3 os_mon-2.4.3 otp_mibs-1.1.1 parsetools-2.1.5 public_key-1.5 reltool-0.7.5 runtime_tools-1.12.2 sasl-3.1 snmp-5.2.7 ssl-8.2.1 stdlib-3.4.2 syntax_tools-2.1.3 tools-2.11 wx-1.8.2 xmerl-1.3.15 : diff --git a/system/doc/oam/oam_intro.xml b/system/doc/oam/oam_intro.xml index d3867f03ca..ead8c026b9 100644 --- a/system/doc/oam/oam_intro.xml +++ b/system/doc/oam/oam_intro.xml @@ -211,7 +211,7 @@ snmp:c("MY-MIB", [{il, ["sasl/priv/mibs"]}]).</code> <p>The following MIBs are defined in the OTP system:</p> <list type="bulleted"> - <item><p><c>OTP-REG)</c> (in SASL) contains the top-level + <item><p><c>OTP-REG</c> (in SASL) contains the top-level OTP registration objects, used by all other MIBs.</p></item> <item><p><c>OTP-TC</c> (in SASL) contains the general Textual Conventions, which can be used by any other MIB.</p></item> |
