Rewrite ethread library

Large parts of the ethread library have been rewritten. The
ethread library is an Erlang runtime system internal, portable
thread library used by the runtime system itself.

Most notable improvement is a reader optimized rwlock
implementation which dramatically improve the performance of
read-lock/read-unlock operations on multi processor systems by
avoiding ping-ponging of the rwlock cache lines. The reader
optimized rwlock implementation is used by miscellaneous
rwlocks in the runtime system that are known to be read-locked
frequently, and can be enabled on ETS tables by passing the
`{read_concurrency, true}' option upon table creation. See the
documentation of `ets:new/2' for more information.

The ethread library can now also use the libatomic_ops library
for atomic memory accesses. This makes it possible for the
Erlang runtime system to utilize optimized atomic operations
on more platforms than before. Use the
`--with-libatomic_ops=PATH' configure command line argument
when specifying where the libatomic_ops installation is
located. The libatomic_ops library can be downloaded from:
http://www.hpl.hp.com/research/linux/atomic_ops/

The changed API of the ethread library has also caused
modifications in the Erlang runtime system. Preparations for
the to come "delayed deallocation" feature has also been done
since it depends on the ethread library.

Note: When building for x86, the ethread library will now use
instructions that first appeared on the pentium 4 processor. If
you want the runtime system to be compatible with older
processors (back to 486) you need to pass the
`--enable-ethread-pre-pentium4-compatibility' configure command
line argument when configuring the system.

1 files changed, 64 insertions, 45 deletions

diff --git a/erts/emulator/beam/erl_lock_check.c b/erts/emulator/beam/erl_lock_check.c
index cee470ae37..99cc80e259 100644
--- a/erts/emulator/beam/erl_lock_check.c
+++ b/erts/emulator/beam/erl_lock_check.c
@@ -96,10 +96,10 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"proc_status",				"pid"			},
     {	"proc_tab",				NULL			},
     {   "ports_snapshot",                       NULL                    },
-    {	"db_tab",				"address"		},    
-    {	"db_tab_fix",				"address"		},
     {	"meta_name_tab",	         	"address"		},
     {	"meta_main_tab_slot",			"address"		},
+    {	"db_tab",				"address"		},
+    {	"db_tab_fix",				"address"		},
     {	"meta_main_tab_main",			NULL 			},
     {	"db_hash_slot",				"address"		},
     {	"node_table",				NULL			},
@@ -119,9 +119,9 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"child_status",				NULL			},
 #endif
 #ifdef __WIN32__
-    {	"sys_driver_data_lock",			NULL 			}, 
+    {	"sys_driver_data_lock",			NULL 			},
 #endif
-    {	"drv_ev_state_grow",			NULL,   		}, 
+    {	"drv_ev_state_grow",			NULL,   		},
     {	"drv_ev_state",				"address"		},
     {	"safe_hash",				"address"		},
     {   "pollset_rm_list",                      NULL                    },
@@ -153,6 +153,7 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"instr",				NULL			},
     {	"fix_alloc",				"index"			},
     {	"alcu_allocator",			"index"			},
+    {	"alcu_delayed_free",			"index"			},
     {	"mseg",					NULL			},
 #ifdef HALFWORD_HEAP
     {	"pmmap",				NULL			},
@@ -177,17 +178,19 @@ static erts_lc_lock_order_t erts_lock_order[] = {
     {	"pix_lock",				"address"		},
     {	"run_queues_lists",			NULL			},
     {	"sched_stat",				NULL			},
+    {	"run_queue_sleep_list",			"address"		},
 #endif
     {	"alloc_thr_ix_lock",			NULL			},
 #ifdef ERTS_SMP
-    {	"proc_lck_wtr_alloc",			NULL 			},
+    {	"proc_lck_qs_alloc",			NULL 			},
 #endif
 #ifdef __WIN32__
 #ifdef DEBUG
     {   "save_ops_lock",                        NULL                    },
 #endif
 #endif
-    {	"mtrace_buf",				NULL			}
+    {	"mtrace_buf",				NULL			},
+    {	"erts_alloc_hard_debug",		NULL			}
 };
 
 #define ERTS_LOCK_ORDER_SIZE \
@@ -199,6 +202,8 @@ static erts_lc_lock_order_t erts_lock_order[] = {
        & ERTS_LC_FLG_LT_ALL						\
        & ~(ERTS_LC_FLG_LT_SPINLOCK|ERTS_LC_FLG_LT_RWSPINLOCK)))
 
+static __decl_noreturn void  __noreturn lc_abort(void);
+
 static char *
 lock_type(Uint16 flags)
 {
@@ -222,7 +227,7 @@ rw_op_str(Uint16 flags)
 	return " (r)";
     case ERTS_LC_FLG_LO_WRITE:
 	erts_fprintf(stderr, "\nInternal error\n");
-	abort();
+	lc_abort();
     default:
 	break;
     }
@@ -271,28 +276,18 @@ static erts_lc_free_block_t *free_blocks;
 #define ERTS_LC_FB_CHUNK_SIZE 10
 #endif
 
-#ifdef ETHR_HAVE_NATIVE_LOCKS
 static ethr_spinlock_t free_blocks_lock;
-#define ERTS_LC_LOCK	ethr_spin_lock
-#define ERTS_LC_UNLOCK	ethr_spin_unlock
-#else
-static ethr_mutex free_blocks_lock;
-#define ERTS_LC_LOCK	ethr_mutex_lock
-#define ERTS_LC_UNLOCK	ethr_mutex_unlock
-#endif
 
 static ERTS_INLINE void
 lc_lock(void)
 {
-    if (ERTS_LC_LOCK(&free_blocks_lock) != 0)
-	abort();
+    ethr_spin_lock(&free_blocks_lock);
 }
 
 static ERTS_INLINE void
 lc_unlock(void)
 {
-    if (ERTS_LC_UNLOCK(&free_blocks_lock) != 0)
-	abort();
+    ethr_spin_unlock(&free_blocks_lock);
 }
 
 static ERTS_INLINE void lc_free(void *p)
@@ -313,7 +308,7 @@ static void *lc_core_alloc(void)
 {
     lc_unlock();
     erts_fprintf(stderr, "Lock checker out of memory!\n");
-    abort();
+    lc_abort();
 }
 
 #else
@@ -327,7 +322,7 @@ static void *lc_core_alloc(void)
 					  * ERTS_LC_FB_CHUNK_SIZE);
     if (!fbs) {
 	erts_fprintf(stderr, "Lock checker failed to allocate memory!\n");
-	abort();
+	lc_abort();
     }
     for (i = 1; i < ERTS_LC_FB_CHUNK_SIZE - 1; i++) {
 #ifdef DEBUG
@@ -367,11 +362,11 @@ create_locked_locks(char *thread_name)
 {
     erts_lc_locked_locks_t *l_lcks = malloc(sizeof(erts_lc_locked_locks_t));
     if (!l_lcks)
-	abort();
+	lc_abort();
 
     l_lcks->thread_name = strdup(thread_name ? thread_name : "unknown");
     if (!l_lcks->thread_name)
-	abort();
+	lc_abort();
 
     l_lcks->tid = erts_thr_self();
     l_lcks->required.first = NULL;
@@ -513,7 +508,7 @@ uninitialized_lock(void)
 {
     erts_fprintf(stderr, "Performing operations on uninitialized lock!\n");
     print_curr_locks(get_my_locked_locks());
-    abort();
+    lc_abort();
 }
 
 static void
@@ -523,7 +518,7 @@ lock_twice(char *prefix, erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck,
     erts_fprintf(stderr, "%s%s", prefix, rw_op_str(op_flags));
     print_lock(" ", lck, " lock which is already locked by thread!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -533,7 +528,7 @@ unlock_op_mismatch(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck,
     erts_fprintf(stderr, "Unlocking%s ", rw_op_str(op_flags));
     print_lock("", lck, " lock which mismatch previous lock operation!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -541,7 +536,7 @@ unlock_of_not_locked(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
 {
     print_lock("Unlocking ", lck, " lock which is not locked by thread!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -550,7 +545,7 @@ lock_order_violation(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
     print_lock("Lock order violation occured when locking ", lck, "!\n");
     print_curr_locks(l_lcks);
     print_lock_order();
-    abort();
+    lc_abort();
 }
 
 static void
@@ -561,7 +556,7 @@ type_order_violation(char *op, erts_lc_locked_locks_t *l_lcks,
     print_lock(op, lck, "!\n");
     ASSERT(l_lcks);
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -613,7 +608,7 @@ lock_mismatch(erts_lc_locked_locks_t *l_lcks, int exact,
 	}
     }
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -621,7 +616,7 @@ unlock_of_required_lock(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
 {
     print_lock("Unlocking required ", lck, " lock!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -629,7 +624,7 @@ unrequire_of_not_required_lock(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *l
 {
     print_lock("Unrequire on ", lck, " lock not required!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -637,7 +632,7 @@ require_twice(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
 {
     print_lock("Require on ", lck, " lock already required!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 static void
@@ -645,7 +640,7 @@ required_not_locked(erts_lc_locked_locks_t *l_lcks, erts_lc_lock_t *lck)
 {
     print_lock("Required ", lck, " lock not locked!\n");
     print_curr_locks(l_lcks);
-    abort();
+    lc_abort();
 }
 
 
@@ -658,13 +653,23 @@ thread_exit_handler(void)
 	    erts_fprintf(stderr,
 			 "Thread exiting while having locked locks!\n");
 	    print_curr_locks(l_lcks);
-	    abort();
+	    lc_abort();
 	}
 	destroy_locked_locks(l_lcks);
 	/* erts_tsd_set(locks_key, NULL); */
     }
 }
 
+static __decl_noreturn void
+lc_abort(void)
+{
+#ifdef __WIN32__
+    DebugBreak();
+#else
+    abort();
+#endif
+}
+
 void
 erts_lc_set_thread_name(char *thread_name)
 {
@@ -676,7 +681,7 @@ erts_lc_set_thread_name(char *thread_name)
 	free((void *) l_lcks->thread_name);
 	l_lcks->thread_name = strdup(thread_name ? thread_name : "unknown");
 	if (!l_lcks->thread_name)
-	    abort();
+	    lc_abort();
     }
 }
 
@@ -686,7 +691,7 @@ erts_lc_assert_failed(char *file, int line, char *assertion)
     erts_fprintf(stderr, "%s:%d: Lock check assertion \"%s\" failed!\n",
 		 file, line, assertion);
     print_curr_locks(get_my_locked_locks());
-    abort();
+    lc_abort();
     return 0;
 }
 
@@ -699,7 +704,7 @@ void erts_lc_fail(char *fmt, ...)
     va_end(args);
     erts_fprintf(stderr, "\n");
     print_curr_locks(get_my_locked_locks());
-    abort();
+    lc_abort();
 }
 
 
@@ -719,7 +724,7 @@ erts_lc_get_lock_order_id(char *name)
 		     "(update erl_lock_check.c)\n",
 		     name);
     }
-    abort();
+    lc_abort();
     return (Sint16) -1;
 }
 
@@ -895,6 +900,25 @@ erts_lc_check_exact(erts_lc_lock_t *have, int have_len)
     }
 }
 
+void
+erts_lc_check_no_locked_of_type(Uint16 flags)
+{
+    erts_lc_locked_locks_t *l_lcks = get_my_locked_locks();
+    if (l_lcks) {
+	erts_lc_locked_lock_t *l_lck = l_lcks->locked.first;
+	for (l_lck = l_lcks->locked.first; l_lck; l_lck = l_lck->next) {
+	    if (l_lck->flags & flags) {
+		erts_fprintf(stderr,
+			     "Locked lock of type %s found which isn't "
+			     "allowed here!\n",
+			     lock_type(l_lck->flags));
+		print_curr_locks(l_lcks);
+		lc_abort();
+	    }
+	}
+    }
+}
+
 int
 erts_lc_trylock_force_busy_flg(erts_lc_lock_t *lck, Uint16 op_flags)
 {
@@ -1283,13 +1307,8 @@ erts_lc_init(void)
     free_blocks = NULL;
 #endif /* #ifdef ERTS_LC_STATIC_ALLOC */
 
-#ifdef ETHR_HAVE_NATIVE_LOCKS
     if (ethr_spinlock_init(&free_blocks_lock) != 0)
-	abort();
-#else
-    if (ethr_mutex_init(&free_blocks_lock) != 0)
-	abort();
-#endif
+	lc_abort();
 
     erts_tsd_key_create(&locks_key);
 }