Merge branch 'dev'3.0.0

Conflicts:
	ChangeLog
	include/jemalloc/internal/chunk.h
	src/chunk.c
	src/huge.c
	src/jemalloc.c
	test/rallocm.c

1 files changed, 207 insertions, 198 deletions

diff --git a/include/jemalloc/internal/tcache.h b/include/jemalloc/internal/tcache.h
index da3c68c..38d735c 100644
--- a/include/jemalloc/internal/tcache.h
+++ b/include/jemalloc/internal/tcache.h
@@ -1,4 +1,3 @@
-#ifdef JEMALLOC_TCACHE
 /******************************************************************************/
 #ifdef JEMALLOC_H_TYPES
 
@@ -7,6 +6,16 @@ typedef struct tcache_bin_s tcache_bin_t;
 typedef struct tcache_s tcache_t;
 
 /*
+ * tcache pointers close to NULL are used to encode state information that is
+ * used for two purposes: preventing thread caching on a per thread basis and
+ * cleaning up during thread shutdown.
+ */
+#define	TCACHE_STATE_DISABLED		((tcache_t *)(uintptr_t)1)
+#define	TCACHE_STATE_REINCARNATED	((tcache_t *)(uintptr_t)2)
+#define	TCACHE_STATE_PURGATORY		((tcache_t *)(uintptr_t)3)
+#define	TCACHE_STATE_MAX		TCACHE_STATE_PURGATORY
+
+/*
  * Absolute maximum number of cache slots for each small bin in the thread
  * cache.  This is an additional constraint beyond that imposed as: twice the
  * number of regions per run for this size class.
@@ -22,17 +31,26 @@ typedef struct tcache_s tcache_t;
 #define	LG_TCACHE_MAXCLASS_DEFAULT	15
 
 /*
- * (1U << opt_lg_tcache_gc_sweep) is the approximate number of allocation
- * events between full GC sweeps (-1: disabled).  Integer rounding may cause
- * the actual number to be slightly higher, since GC is performed
- * incrementally.
+ * TCACHE_GC_SWEEP is the approximate number of allocation events between
+ * full GC sweeps.  Integer rounding may cause the actual number to be
+ * slightly higher, since GC is performed incrementally.
  */
-#define	LG_TCACHE_GC_SWEEP_DEFAULT	13
+#define	TCACHE_GC_SWEEP			8192
+
+/* Number of tcache allocation/deallocation events between incremental GCs. */
+#define	TCACHE_GC_INCR							\
+    ((TCACHE_GC_SWEEP / NBINS) + ((TCACHE_GC_SWEEP / NBINS == 0) ? 0 : 1))
 
 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
 #ifdef JEMALLOC_H_STRUCTS
 
+typedef enum {
+	tcache_enabled_false   = 0, /* Enable cast to/from bool. */
+	tcache_enabled_true    = 1,
+	tcache_enabled_default = 2
+} tcache_enabled_t;
+
 /*
  * Read-only information associated with each element of tcache_t's tbins array
  * is stored separately, mainly to reduce memory usage.
@@ -42,9 +60,7 @@ struct tcache_bin_info_s {
 };
 
 struct tcache_bin_s {
-#  ifdef JEMALLOC_STATS
 	tcache_bin_stats_t tstats;
-#  endif
 	int		low_water;	/* Min # cached since last GC. */
 	unsigned	lg_fill_div;	/* Fill (ncached_max >> lg_fill_div). */
 	unsigned	ncached;	/* # of cached objects. */
@@ -52,12 +68,8 @@ struct tcache_bin_s {
 };
 
 struct tcache_s {
-#  ifdef JEMALLOC_STATS
 	ql_elm(tcache_t) link;		/* Used for aggregating stats. */
-#  endif
-#  ifdef JEMALLOC_PROF
 	uint64_t	prof_accumbytes;/* Cleared after arena_prof_accum() */
-#  endif
 	arena_t		*arena;		/* This thread's arena. */
 	unsigned	ev_cnt;		/* Event count since incremental GC. */
 	unsigned	next_gc_bin;	/* Next bin to GC. */
@@ -76,29 +88,11 @@ struct tcache_s {
 
 extern bool	opt_tcache;
 extern ssize_t	opt_lg_tcache_max;
-extern ssize_t	opt_lg_tcache_gc_sweep;
 
 extern tcache_bin_info_t	*tcache_bin_info;
 
-/* Map of thread-specific caches. */
-#ifndef NO_TLS
-extern __thread tcache_t	*tcache_tls
-    JEMALLOC_ATTR(tls_model("initial-exec"));
-#  define TCACHE_GET()	tcache_tls
-#  define TCACHE_SET(v)	do {						\
-	tcache_tls = (tcache_t *)(v);					\
-	pthread_setspecific(tcache_tsd, (void *)(v));			\
-} while (0)
-#else
-#  define TCACHE_GET()	((tcache_t *)pthread_getspecific(tcache_tsd))
-#  define TCACHE_SET(v)	do {						\
-	pthread_setspecific(tcache_tsd, (void *)(v));			\
-} while (0)
-#endif
-extern pthread_key_t		tcache_tsd;
-
 /*
- * Number of tcache bins.  There are nbins small-object bins, plus 0 or more
+ * Number of tcache bins.  There are NBINS small-object bins, plus 0 or more
  * large-object bins.
  */
 extern size_t			nhbins;
@@ -106,68 +100,159 @@ extern size_t			nhbins;
 /* Maximum cached size class. */
 extern size_t			tcache_maxclass;
 
-/* Number of tcache allocation/deallocation events between incremental GCs. */
-extern unsigned			tcache_gc_incr;
-
-void	tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-    , tcache_t *tcache
-#endif
-    );
-void	tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-    , tcache_t *tcache
-#endif
-    );
-tcache_t *tcache_create(arena_t *arena);
+size_t	tcache_salloc(const void *ptr);
+void	tcache_event_hard(tcache_t *tcache);
 void	*tcache_alloc_small_hard(tcache_t *tcache, tcache_bin_t *tbin,
     size_t binind);
+void	tcache_bin_flush_small(tcache_bin_t *tbin, size_t binind, unsigned rem,
+    tcache_t *tcache);
+void	tcache_bin_flush_large(tcache_bin_t *tbin, size_t binind, unsigned rem,
+    tcache_t *tcache);
+void	tcache_arena_associate(tcache_t *tcache, arena_t *arena);
+void	tcache_arena_dissociate(tcache_t *tcache);
+tcache_t *tcache_create(arena_t *arena);
 void	tcache_destroy(tcache_t *tcache);
-#ifdef JEMALLOC_STATS
+void	tcache_thread_cleanup(void *arg);
 void	tcache_stats_merge(tcache_t *tcache, arena_t *arena);
-#endif
-bool	tcache_boot(void);
+bool	tcache_boot0(void);
+bool	tcache_boot1(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES
 
 #ifndef JEMALLOC_ENABLE_INLINE
+malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache, tcache_t *)
+malloc_tsd_protos(JEMALLOC_ATTR(unused), tcache_enabled, tcache_enabled_t)
+
 void	tcache_event(tcache_t *tcache);
-tcache_t *tcache_get(void);
+void	tcache_flush(void);
+bool	tcache_enabled_get(void);
+tcache_t *tcache_get(bool create);
+void	tcache_enabled_set(bool enabled);
 void	*tcache_alloc_easy(tcache_bin_t *tbin);
 void	*tcache_alloc_small(tcache_t *tcache, size_t size, bool zero);
 void	*tcache_alloc_large(tcache_t *tcache, size_t size, bool zero);
-void	tcache_dalloc_small(tcache_t *tcache, void *ptr);
+void	tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind);
 void	tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TCACHE_C_))
+/* Map of thread-specific caches. */
+malloc_tsd_externs(tcache, tcache_t *)
+malloc_tsd_funcs(JEMALLOC_INLINE, tcache, tcache_t *, NULL,
+    tcache_thread_cleanup)
+/* Per thread flag that allows thread caches to be disabled. */
+malloc_tsd_externs(tcache_enabled, tcache_enabled_t)
+malloc_tsd_funcs(JEMALLOC_INLINE, tcache_enabled, tcache_enabled_t,
+    tcache_enabled_default, malloc_tsd_no_cleanup)
+
+JEMALLOC_INLINE void
+tcache_flush(void)
+{
+	tcache_t *tcache;
+
+	cassert(config_tcache);
+
+	tcache = *tcache_tsd_get();
+	if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX)
+		return;
+	tcache_destroy(tcache);
+	tcache = NULL;
+	tcache_tsd_set(&tcache);
+}
+
+JEMALLOC_INLINE bool
+tcache_enabled_get(void)
+{
+	tcache_enabled_t tcache_enabled;
+
+	cassert(config_tcache);
+
+	tcache_enabled = *tcache_enabled_tsd_get();
+	if (tcache_enabled == tcache_enabled_default) {
+		tcache_enabled = (tcache_enabled_t)opt_tcache;
+		tcache_enabled_tsd_set(&tcache_enabled);
+	}
+
+	return ((bool)tcache_enabled);
+}
+
+JEMALLOC_INLINE void
+tcache_enabled_set(bool enabled)
+{
+	tcache_enabled_t tcache_enabled;
+	tcache_t *tcache;
+
+	cassert(config_tcache);
+
+	tcache_enabled = (tcache_enabled_t)enabled;
+	tcache_enabled_tsd_set(&tcache_enabled);
+	tcache = *tcache_tsd_get();
+	if (enabled) {
+		if (tcache == TCACHE_STATE_DISABLED) {
+			tcache = NULL;
+			tcache_tsd_set(&tcache);
+		}
+	} else /* disabled */ {
+		if (tcache > TCACHE_STATE_MAX) {
+			tcache_destroy(tcache);
+			tcache = NULL;
+		}
+		if (tcache == NULL) {
+			tcache = TCACHE_STATE_DISABLED;
+			tcache_tsd_set(&tcache);
+		}
+	}
+}
+
 JEMALLOC_INLINE tcache_t *
-tcache_get(void)
+tcache_get(bool create)
 {
 	tcache_t *tcache;
 
-	if ((isthreaded & opt_tcache) == false)
+	if (config_tcache == false)
+		return (NULL);
+	if (config_lazy_lock && isthreaded == false)
 		return (NULL);
 
-	tcache = TCACHE_GET();
-	if ((uintptr_t)tcache <= (uintptr_t)2) {
+	tcache = *tcache_tsd_get();
+	if ((uintptr_t)tcache <= (uintptr_t)TCACHE_STATE_MAX) {
+		if (tcache == TCACHE_STATE_DISABLED)
+			return (NULL);
 		if (tcache == NULL) {
-			tcache = tcache_create(choose_arena());
-			if (tcache == NULL)
-				return (NULL);
-		} else {
-			if (tcache == (void *)(uintptr_t)1) {
+			if (create == false) {
 				/*
-				 * Make a note that an allocator function was
-				 * called after the tcache_thread_cleanup() was
-				 * called.
+				 * Creating a tcache here would cause
+				 * allocation as a side effect of free().
+				 * Ordinarily that would be okay since
+				 * tcache_create() failure is a soft failure
+				 * that doesn't propagate.  However, if TLS
+				 * data are freed via free() as in glibc,
+				 * subtle corruption could result from setting
+				 * a TLS variable after its backing memory is
+				 * freed.
 				 */
-				TCACHE_SET((uintptr_t)2);
+				return (NULL);
+			}
+			if (tcache_enabled_get() == false) {
+				tcache_enabled_set(false); /* Memoize. */
+				return (NULL);
 			}
+			return (tcache_create(choose_arena(NULL)));
+		}
+		if (tcache == TCACHE_STATE_PURGATORY) {
+			/*
+			 * Make a note that an allocator function was called
+			 * after tcache_thread_cleanup() was called.
+			 */
+			tcache = TCACHE_STATE_REINCARNATED;
+			tcache_tsd_set(&tcache);
 			return (NULL);
 		}
+		if (tcache == TCACHE_STATE_REINCARNATED)
+			return (NULL);
+		not_reached();
 	}
 
 	return (tcache);
@@ -177,60 +262,13 @@ JEMALLOC_INLINE void
 tcache_event(tcache_t *tcache)
 {
 
-	if (tcache_gc_incr == 0)
+	if (TCACHE_GC_INCR == 0)
 		return;
 
 	tcache->ev_cnt++;
-	assert(tcache->ev_cnt <= tcache_gc_incr);
-	if (tcache->ev_cnt == tcache_gc_incr) {
-		size_t binind = tcache->next_gc_bin;
-		tcache_bin_t *tbin = &tcache->tbins[binind];
-		tcache_bin_info_t *tbin_info = &tcache_bin_info[binind];
-
-		if (tbin->low_water > 0) {
-			/*
-			 * Flush (ceiling) 3/4 of the objects below the low
-			 * water mark.
-			 */
-			if (binind < nbins) {
-				tcache_bin_flush_small(tbin, binind,
-				    tbin->ncached - tbin->low_water +
-				    (tbin->low_water >> 2)
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-				    , tcache
-#endif
-				    );
-			} else {
-				tcache_bin_flush_large(tbin, binind,
-				    tbin->ncached - tbin->low_water +
-				    (tbin->low_water >> 2)
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-				    , tcache
-#endif
-				    );
-			}
-			/*
-			 * Reduce fill count by 2X.  Limit lg_fill_div such that
-			 * the fill count is always at least 1.
-			 */
-			if ((tbin_info->ncached_max >> (tbin->lg_fill_div+1))
-			    >= 1)
-				tbin->lg_fill_div++;
-		} else if (tbin->low_water < 0) {
-			/*
-			 * Increase fill count by 2X.  Make sure lg_fill_div
-			 * stays greater than 0.
-			 */
-			if (tbin->lg_fill_div > 1)
-				tbin->lg_fill_div--;
-		}
-		tbin->low_water = tbin->ncached;
-
-		tcache->next_gc_bin++;
-		if (tcache->next_gc_bin == nhbins)
-			tcache->next_gc_bin = 0;
-		tcache->ev_cnt = 0;
-	}
+	assert(tcache->ev_cnt <= TCACHE_GC_INCR);
+	if (tcache->ev_cnt == TCACHE_GC_INCR)
+		tcache_event_hard(tcache);
 }
 
 JEMALLOC_INLINE void *
@@ -257,7 +295,7 @@ tcache_alloc_small(tcache_t *tcache, size_t size, bool zero)
 	tcache_bin_t *tbin;
 
 	binind = SMALL_SIZE2BIN(size);
-	assert(binind < nbins);
+	assert(binind < NBINS);
 	tbin = &tcache->tbins[binind];
 	ret = tcache_alloc_easy(tbin);
 	if (ret == NULL) {
@@ -265,24 +303,29 @@ tcache_alloc_small(tcache_t *tcache, size_t size, bool zero)
 		if (ret == NULL)
 			return (NULL);
 	}
-	assert(arena_salloc(ret) == arena_bin_info[binind].reg_size);
+	assert(tcache_salloc(ret) == arena_bin_info[binind].reg_size);
 
 	if (zero == false) {
-#ifdef JEMALLOC_FILL
-		if (opt_junk)
-			memset(ret, 0xa5, size);
-		else if (opt_zero)
-			memset(ret, 0, size);
-#endif
-	} else
+		if (config_fill) {
+			if (opt_junk) {
+				arena_alloc_junk_small(ret,
+				    &arena_bin_info[binind], false);
+			} else if (opt_zero)
+				memset(ret, 0, size);
+		}
+	} else {
+		if (config_fill && opt_junk) {
+			arena_alloc_junk_small(ret, &arena_bin_info[binind],
+			    true);
+		}
+		VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
 		memset(ret, 0, size);
+	}
 
-#ifdef JEMALLOC_STATS
-	tbin->tstats.nrequests++;
-#endif
-#ifdef JEMALLOC_PROF
-	tcache->prof_accumbytes += arena_bin_info[binind].reg_size;
-#endif
+	if (config_stats)
+		tbin->tstats.nrequests++;
+	if (config_prof)
+		tcache->prof_accumbytes += arena_bin_info[binind].reg_size;
 	tcache_event(tcache);
 	return (ret);
 }
@@ -296,7 +339,7 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero)
 
 	size = PAGE_CEILING(size);
 	assert(size <= tcache_maxclass);
-	binind = nbins + (size >> PAGE_SHIFT) - 1;
+	binind = NBINS + (size >> LG_PAGE) - 1;
 	assert(binind < nhbins);
 	tbin = &tcache->tbins[binind];
 	ret = tcache_alloc_easy(tbin);
@@ -309,28 +352,30 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero)
 		if (ret == NULL)
 			return (NULL);
 	} else {
-#ifdef JEMALLOC_PROF
-		arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ret);
-		size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >>
-		    PAGE_SHIFT);
-		chunk->map[pageind-map_bias].bits &= ~CHUNK_MAP_CLASS_MASK;
-#endif
+		if (config_prof && prof_promote && size == PAGE) {
+			arena_chunk_t *chunk =
+			    (arena_chunk_t *)CHUNK_ADDR2BASE(ret);
+			size_t pageind = (((uintptr_t)ret - (uintptr_t)chunk) >>
+			    LG_PAGE);
+			arena_mapbits_large_binind_set(chunk, pageind,
+			    BININD_INVALID);
+		}
 		if (zero == false) {
-#ifdef JEMALLOC_FILL
-			if (opt_junk)
-				memset(ret, 0xa5, size);
-			else if (opt_zero)
-				memset(ret, 0, size);
-#endif
-		} else
+			if (config_fill) {
+				if (opt_junk)
+					memset(ret, 0xa5, size);
+				else if (opt_zero)
+					memset(ret, 0, size);
+			}
+		} else {
+			VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
 			memset(ret, 0, size);
+		}
 
-#ifdef JEMALLOC_STATS
-		tbin->tstats.nrequests++;
-#endif
-#ifdef JEMALLOC_PROF
-		tcache->prof_accumbytes += size;
-#endif
+		if (config_stats)
+			tbin->tstats.nrequests++;
+		if (config_prof)
+			tcache->prof_accumbytes += size;
 	}
 
 	tcache_event(tcache);
@@ -338,45 +383,21 @@ tcache_alloc_large(tcache_t *tcache, size_t size, bool zero)
 }
 
 JEMALLOC_INLINE void
-tcache_dalloc_small(tcache_t *tcache, void *ptr)
+tcache_dalloc_small(tcache_t *tcache, void *ptr, size_t binind)
 {
-	arena_t *arena;
-	arena_chunk_t *chunk;
-	arena_run_t *run;
-	arena_bin_t *bin;
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
-	size_t pageind, binind;
-	arena_chunk_map_t *mapelm;
-
-	assert(arena_salloc(ptr) <= small_maxclass);
-
-	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
-	arena = chunk->arena;
-	pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT;
-	mapelm = &chunk->map[pageind-map_bias];
-	run = (arena_run_t *)((uintptr_t)chunk + (uintptr_t)((pageind -
-	    (mapelm->bits >> PAGE_SHIFT)) << PAGE_SHIFT));
-	dassert(run->magic == ARENA_RUN_MAGIC);
-	bin = run->bin;
-	binind = ((uintptr_t)bin - (uintptr_t)&arena->bins) /
-	    sizeof(arena_bin_t);
-	assert(binind < nbins);
-
-#ifdef JEMALLOC_FILL
-	if (opt_junk)
-		memset(ptr, 0x5a, arena_bin_info[binind].reg_size);
-#endif
+
+	assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
+
+	if (config_fill && opt_junk)
+		arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
 
 	tbin = &tcache->tbins[binind];
 	tbin_info = &tcache_bin_info[binind];
 	if (tbin->ncached == tbin_info->ncached_max) {
 		tcache_bin_flush_small(tbin, binind, (tbin_info->ncached_max >>
-		    1)
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-		    , tcache
-#endif
-		    );
+		    1), tcache);
 	}
 	assert(tbin->ncached < tbin_info->ncached_max);
 	tbin->avail[tbin->ncached] = ptr;
@@ -388,35 +409,24 @@ tcache_dalloc_small(tcache_t *tcache, void *ptr)
 JEMALLOC_INLINE void
 tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size)
 {
-	arena_t *arena;
-	arena_chunk_t *chunk;
-	size_t pageind, binind;
+	size_t binind;
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
 
 	assert((size & PAGE_MASK) == 0);
-	assert(arena_salloc(ptr) > small_maxclass);
-	assert(arena_salloc(ptr) <= tcache_maxclass);
+	assert(tcache_salloc(ptr) > SMALL_MAXCLASS);
+	assert(tcache_salloc(ptr) <= tcache_maxclass);
 
-	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
-	arena = chunk->arena;
-	pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> PAGE_SHIFT;
-	binind = nbins + (size >> PAGE_SHIFT) - 1;
+	binind = NBINS + (size >> LG_PAGE) - 1;
 
-#ifdef JEMALLOC_FILL
-	if (opt_junk)
+	if (config_fill && opt_junk)
 		memset(ptr, 0x5a, size);
-#endif
 
 	tbin = &tcache->tbins[binind];
 	tbin_info = &tcache_bin_info[binind];
 	if (tbin->ncached == tbin_info->ncached_max) {
 		tcache_bin_flush_large(tbin, binind, (tbin_info->ncached_max >>
-		    1)
-#if (defined(JEMALLOC_STATS) || defined(JEMALLOC_PROF))
-		    , tcache
-#endif
-		    );
+		    1), tcache);
 	}
 	assert(tbin->ncached < tbin_info->ncached_max);
 	tbin->avail[tbin->ncached] = ptr;
@@ -428,4 +438,3 @@ tcache_dalloc_large(tcache_t *tcache, void *ptr, size_t size)
 
 #endif /* JEMALLOC_H_INLINES */
 /******************************************************************************/
-#endif /* JEMALLOC_TCACHE */