Merge remote-tracking branch 'aosp/upstream-release' into fix

Bug: 28860984
Change-Id: I9eaf67f53f9872177d068d660d1e051cecdc82a0

27 files changed, 1431 insertions, 481 deletions

diff --git a/include/jemalloc/internal/arena.h b/include/jemalloc/internal/arena.h
index f234ec6..a8819f9 100644
--- a/include/jemalloc/internal/arena.h
+++ b/include/jemalloc/internal/arena.h
@@ -41,6 +41,7 @@ typedef enum {
 #define	DECAY_NTICKS_PER_UPDATE	1000
 
 typedef struct arena_runs_dirty_link_s arena_runs_dirty_link_t;
+typedef struct arena_avail_links_s arena_avail_links_t;
 typedef struct arena_run_s arena_run_t;
 typedef struct arena_chunk_map_bits_s arena_chunk_map_bits_t;
 typedef struct arena_chunk_map_misc_s arena_chunk_map_misc_t;
@@ -158,13 +159,13 @@ struct arena_runs_dirty_link_s {
  */
 struct arena_chunk_map_misc_s {
 	/*
-	 * Linkage for run trees.  There are two disjoint uses:
+	 * Linkage for run heaps.  There are two disjoint uses:
 	 *
-	 * 1) arena_t's runs_avail tree.
+	 * 1) arena_t's runs_avail heaps.
 	 * 2) arena_run_t conceptually uses this linkage for in-use non-full
 	 *    runs, rather than directly embedding linkage.
 	 */
-	rb_node(arena_chunk_map_misc_t)		rb_link;
+	phn(arena_chunk_map_misc_t)		ph_link;
 
 	union {
 		/* Linkage for list of dirty runs. */
@@ -180,7 +181,7 @@ struct arena_chunk_map_misc_s {
 		arena_run_t			run;
 	};
 };
-typedef rb_tree(arena_chunk_map_misc_t) arena_run_tree_t;
+typedef ph(arena_chunk_map_misc_t) arena_run_heap_t;
 #endif /* JEMALLOC_ARENA_STRUCTS_A */
 
 #ifdef JEMALLOC_ARENA_STRUCTS_B
@@ -277,13 +278,13 @@ struct arena_bin_s {
 	arena_run_t		*runcur;
 
 	/*
-	 * Tree of non-full runs.  This tree is used when looking for an
+	 * Heap of non-full runs.  This heap is used when looking for an
 	 * existing run when runcur is no longer usable.  We choose the
 	 * non-full run that is lowest in memory; this policy tends to keep
 	 * objects packed well, and it can also help reduce the number of
 	 * almost-empty chunks.
 	 */
-	arena_run_tree_t	runs;
+	arena_run_heap_t	runs;
 
 	/* Bin statistics. */
 	malloc_bin_stats_t	stats;
@@ -294,10 +295,18 @@ struct arena_s {
 	unsigned		ind;
 
 	/*
-	 * Number of threads currently assigned to this arena.  This field is
-	 * synchronized via atomic operations.
+	 * Number of threads currently assigned to this arena, synchronized via
+	 * atomic operations.  Each thread has two distinct assignments, one for
+	 * application-serving allocation, and the other for internal metadata
+	 * allocation.  Internal metadata must not be allocated from arenas
+	 * created via the arenas.extend mallctl, because the arena.<i>.reset
+	 * mallctl indiscriminately discards all allocations for the affected
+	 * arena.
+	 *
+	 *   0: Application allocation.
+	 *   1: Internal metadata allocation.
 	 */
-	unsigned		nthreads;
+	unsigned		nthreads[2];
 
 	/*
 	 * There are three classes of arena operations from a locking
@@ -326,6 +335,10 @@ struct arena_s {
 
 	dss_prec_t		dss_prec;
 
+
+	/* Extant arena chunks. */
+	ql_head(extent_node_t)	achunks;
+
 	/*
 	 * In order to avoid rapid chunk allocation/deallocation when an arena
 	 * oscillates right on the cusp of needing a new chunk, cache the most
@@ -462,10 +475,10 @@ struct arena_s {
 	arena_bin_t		bins[NBINS];
 
 	/*
-	 * Quantized address-ordered trees of this arena's available runs.  The
-	 * trees are used for first-best-fit run allocation.
+	 * Quantized address-ordered heaps of this arena's available runs.  The
+	 * heaps are used for first-best-fit run allocation.
 	 */
-	arena_run_tree_t	runs_avail[1]; /* Dynamically sized. */
+	arena_run_heap_t	runs_avail[1]; /* Dynamically sized. */
 };
 
 /* Used in conjunction with tsd for fast arena-related context lookup. */
@@ -510,25 +523,28 @@ void	arena_chunk_cache_maybe_insert(arena_t *arena, extent_node_t *node,
     bool cache);
 void	arena_chunk_cache_maybe_remove(arena_t *arena, extent_node_t *node,
     bool cache);
-extent_node_t	*arena_node_alloc(arena_t *arena);
-void	arena_node_dalloc(arena_t *arena, extent_node_t *node);
-void	*arena_chunk_alloc_huge(arena_t *arena, size_t usize, size_t alignment,
-    bool *zero);
-void	arena_chunk_dalloc_huge(arena_t *arena, void *chunk, size_t usize);
-void	arena_chunk_ralloc_huge_similar(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize);
-void	arena_chunk_ralloc_huge_shrink(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize);
-bool	arena_chunk_ralloc_huge_expand(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize, bool *zero);
-ssize_t	arena_lg_dirty_mult_get(arena_t *arena);
-bool	arena_lg_dirty_mult_set(arena_t *arena, ssize_t lg_dirty_mult);
-ssize_t	arena_decay_time_get(arena_t *arena);
-bool	arena_decay_time_set(arena_t *arena, ssize_t decay_time);
-void	arena_maybe_purge(arena_t *arena);
-void	arena_purge(arena_t *arena, bool all);
-void	arena_tcache_fill_small(tsd_t *tsd, arena_t *arena, tcache_bin_t *tbin,
-    szind_t binind, uint64_t prof_accumbytes);
+extent_node_t	*arena_node_alloc(tsdn_t *tsdn, arena_t *arena);
+void	arena_node_dalloc(tsdn_t *tsdn, arena_t *arena, extent_node_t *node);
+void	*arena_chunk_alloc_huge(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool *zero);
+void	arena_chunk_dalloc_huge(tsdn_t *tsdn, arena_t *arena, void *chunk,
+    size_t usize);
+void	arena_chunk_ralloc_huge_similar(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize);
+void	arena_chunk_ralloc_huge_shrink(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize);
+bool	arena_chunk_ralloc_huge_expand(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize, bool *zero);
+ssize_t	arena_lg_dirty_mult_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_lg_dirty_mult_set(tsdn_t *tsdn, arena_t *arena,
+    ssize_t lg_dirty_mult);
+ssize_t	arena_decay_time_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_decay_time_set(tsdn_t *tsdn, arena_t *arena, ssize_t decay_time);
+void	arena_purge(tsdn_t *tsdn, arena_t *arena, bool all);
+void	arena_maybe_purge(tsdn_t *tsdn, arena_t *arena);
+void	arena_reset(tsd_t *tsd, arena_t *arena);
+void	arena_tcache_fill_small(tsdn_t *tsdn, arena_t *arena,
+    tcache_bin_t *tbin, szind_t binind, uint64_t prof_accumbytes);
 void	arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,
     bool zero);
 #ifdef JEMALLOC_JET
@@ -541,17 +557,18 @@ extern arena_dalloc_junk_small_t *arena_dalloc_junk_small;
 void	arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info);
 #endif
 void	arena_quarantine_junk_small(void *ptr, size_t usize);
-void	*arena_malloc_large(tsd_t *tsd, arena_t *arena, szind_t ind, bool zero);
-void	*arena_malloc_hard(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
-    bool zero, tcache_t *tcache);
-void	*arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize,
+void	*arena_malloc_large(tsdn_t *tsdn, arena_t *arena, szind_t ind,
+    bool zero);
+void	*arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size,
+    szind_t ind, bool zero);
+void	*arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
     size_t alignment, bool zero, tcache_t *tcache);
-void	arena_prof_promoted(const void *ptr, size_t size);
-void	arena_dalloc_bin_junked_locked(arena_t *arena, arena_chunk_t *chunk,
-    void *ptr, arena_chunk_map_bits_t *bitselm);
-void	arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,
-    size_t pageind, arena_chunk_map_bits_t *bitselm);
-void	arena_dalloc_small(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
+void	arena_prof_promoted(tsdn_t *tsdn, const void *ptr, size_t size);
+void	arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena,
+    arena_chunk_t *chunk, void *ptr, arena_chunk_map_bits_t *bitselm);
+void	arena_dalloc_bin(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
+    void *ptr, size_t pageind, arena_chunk_map_bits_t *bitselm);
+void	arena_dalloc_small(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
     void *ptr, size_t pageind);
 #ifdef JEMALLOC_JET
 typedef void (arena_dalloc_junk_large_t)(void *, size_t);
@@ -559,67 +576,80 @@ extern arena_dalloc_junk_large_t *arena_dalloc_junk_large;
 #else
 void	arena_dalloc_junk_large(void *ptr, size_t usize);
 #endif
-void	arena_dalloc_large_junked_locked(arena_t *arena, arena_chunk_t *chunk,
-    void *ptr);
-void	arena_dalloc_large(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
+void	arena_dalloc_large_junked_locked(tsdn_t *tsdn, arena_t *arena,
+    arena_chunk_t *chunk, void *ptr);
+void	arena_dalloc_large(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
     void *ptr);
 #ifdef JEMALLOC_JET
 typedef void (arena_ralloc_junk_large_t)(void *, size_t, size_t);
 extern arena_ralloc_junk_large_t *arena_ralloc_junk_large;
 #endif
-bool	arena_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
-    size_t extra, bool zero);
+bool	arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize,
+    size_t size, size_t extra, bool zero);
 void	*arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
     size_t size, size_t alignment, bool zero, tcache_t *tcache);
-dss_prec_t	arena_dss_prec_get(arena_t *arena);
-bool	arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
+dss_prec_t	arena_dss_prec_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_dss_prec_set(tsdn_t *tsdn, arena_t *arena, dss_prec_t dss_prec);
 ssize_t	arena_lg_dirty_mult_default_get(void);
 bool	arena_lg_dirty_mult_default_set(ssize_t lg_dirty_mult);
 ssize_t	arena_decay_time_default_get(void);
 bool	arena_decay_time_default_set(ssize_t decay_time);
-void	arena_basic_stats_merge(arena_t *arena, unsigned *nthreads,
+void	arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena,
+    unsigned *nthreads, const char **dss, ssize_t *lg_dirty_mult,
+    ssize_t *decay_time, size_t *nactive, size_t *ndirty);
+void	arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
     const char **dss, ssize_t *lg_dirty_mult, ssize_t *decay_time,
-    size_t *nactive, size_t *ndirty);
-void	arena_stats_merge(arena_t *arena, unsigned *nthreads, const char **dss,
-    ssize_t *lg_dirty_mult, ssize_t *decay_time, size_t *nactive,
-    size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats,
-    malloc_large_stats_t *lstats, malloc_huge_stats_t *hstats);
-unsigned	arena_nthreads_get(arena_t *arena);
-void	arena_nthreads_inc(arena_t *arena);
-void	arena_nthreads_dec(arena_t *arena);
-arena_t	*arena_new(unsigned ind);
+    size_t *nactive, size_t *ndirty, arena_stats_t *astats,
+    malloc_bin_stats_t *bstats, malloc_large_stats_t *lstats,
+    malloc_huge_stats_t *hstats);
+unsigned	arena_nthreads_get(arena_t *arena, bool internal);
+void	arena_nthreads_inc(arena_t *arena, bool internal);
+void	arena_nthreads_dec(arena_t *arena, bool internal);
+arena_t	*arena_new(tsdn_t *tsdn, unsigned ind);
 bool	arena_boot(void);
-void	arena_prefork(arena_t *arena);
-void	arena_postfork_parent(arena_t *arena);
-void	arena_postfork_child(arena_t *arena);
+void	arena_prefork0(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork1(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork2(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork3(tsdn_t *tsdn, arena_t *arena);
+void	arena_postfork_parent(tsdn_t *tsdn, arena_t *arena);
+void	arena_postfork_child(tsdn_t *tsdn, arena_t *arena);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES
 
 #ifndef JEMALLOC_ENABLE_INLINE
-arena_chunk_map_bits_t	*arena_bitselm_get(arena_chunk_t *chunk,
+arena_chunk_map_bits_t	*arena_bitselm_get_mutable(arena_chunk_t *chunk,
     size_t pageind);
-arena_chunk_map_misc_t	*arena_miscelm_get(arena_chunk_t *chunk,
+const arena_chunk_map_bits_t	*arena_bitselm_get_const(
+    const arena_chunk_t *chunk, size_t pageind);
+arena_chunk_map_misc_t	*arena_miscelm_get_mutable(arena_chunk_t *chunk,
     size_t pageind);
+const arena_chunk_map_misc_t	*arena_miscelm_get_const(
+    const arena_chunk_t *chunk, size_t pageind);
 size_t	arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm);
-void	*arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm);
+void	*arena_miscelm_to_rpages(const arena_chunk_map_misc_t *miscelm);
 arena_chunk_map_misc_t	*arena_rd_to_miscelm(arena_runs_dirty_link_t *rd);
 arena_chunk_map_misc_t	*arena_run_to_miscelm(arena_run_t *run);
-size_t	*arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbitsp_read(size_t *mapbitsp);
-size_t	arena_mapbits_get(arena_chunk_t *chunk, size_t pageind);
+size_t	*arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind);
+const size_t	*arena_mapbitsp_get_const(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbitsp_read(const size_t *mapbitsp);
+size_t	arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind);
 size_t	arena_mapbits_size_decode(size_t mapbits);
-size_t	arena_mapbits_unallocated_size_get(arena_chunk_t *chunk,
+size_t	arena_mapbits_unallocated_size_get(const arena_chunk_t *chunk,
     size_t pageind);
-size_t	arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind);
-szind_t	arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_large_size_get(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbits_small_runind_get(const arena_chunk_t *chunk,
+    size_t pageind);
+szind_t	arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_decommitted_get(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind);
 void	arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits);
 size_t	arena_mapbits_size_encode(size_t size);
 void	arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind,
@@ -639,29 +669,31 @@ void	arena_metadata_allocated_sub(arena_t *arena, size_t size);
 size_t	arena_metadata_allocated_get(arena_t *arena);
 bool	arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes);
 bool	arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes);
-bool	arena_prof_accum(arena_t *arena, uint64_t accumbytes);
+bool	arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes);
 szind_t	arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
 szind_t	arena_bin_index(arena_t *arena, arena_bin_t *bin);
 size_t	arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,
     const void *ptr);
-prof_tctx_t	*arena_prof_tctx_get(const void *ptr);
-void	arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
-void	arena_prof_tctx_reset(const void *ptr, size_t usize,
+prof_tctx_t	*arena_prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	arena_prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void	arena_prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
     const void *old_ptr, prof_tctx_t *old_tctx);
-void	arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks);
-void	arena_decay_tick(tsd_t *tsd, arena_t *arena);
-void	*arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
+void	arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks);
+void	arena_decay_tick(tsdn_t *tsdn, arena_t *arena);
+void	*arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind,
     bool zero, tcache_t *tcache, bool slow_path);
 arena_t	*arena_aalloc(const void *ptr);
-size_t	arena_salloc(const void *ptr, bool demote);
-void	arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
-void	arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
+size_t	arena_salloc(tsdn_t *tsdn, const void *ptr, bool demote);
+void	arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool slow_path);
+void	arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_))
 #  ifdef JEMALLOC_ARENA_INLINE_A
 JEMALLOC_ALWAYS_INLINE arena_chunk_map_bits_t *
-arena_bitselm_get(arena_chunk_t *chunk, size_t pageind)
+arena_bitselm_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {
 
 	assert(pageind >= map_bias);
@@ -670,8 +702,15 @@ arena_bitselm_get(arena_chunk_t *chunk, size_t pageind)
 	return (&chunk->map_bits[pageind-map_bias]);
 }
 
+JEMALLOC_ALWAYS_INLINE const arena_chunk_map_bits_t *
+arena_bitselm_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_bitselm_get_mutable((arena_chunk_t *)chunk, pageind));
+}
+
 JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
-arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
+arena_miscelm_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {
 
 	assert(pageind >= map_bias);
@@ -681,6 +720,13 @@ arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
 	    (uintptr_t)map_misc_offset) + pageind-map_bias);
 }
 
+JEMALLOC_ALWAYS_INLINE const arena_chunk_map_misc_t *
+arena_miscelm_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_miscelm_get_mutable((arena_chunk_t *)chunk, pageind));
+}
+
 JEMALLOC_ALWAYS_INLINE size_t
 arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm)
 {
@@ -695,7 +741,7 @@ arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm)
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm)
+arena_miscelm_to_rpages(const arena_chunk_map_misc_t *miscelm)
 {
 	arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
 	size_t pageind = arena_miscelm_to_pageind(miscelm);
@@ -728,24 +774,31 @@ arena_run_to_miscelm(arena_run_t *run)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t *
-arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {
 
-	return (&arena_bitselm_get(chunk, pageind)->bits);
+	return (&arena_bitselm_get_mutable(chunk, pageind)->bits);
+}
+
+JEMALLOC_ALWAYS_INLINE const size_t *
+arena_mapbitsp_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_mapbitsp_get_mutable((arena_chunk_t *)chunk, pageind));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbitsp_read(size_t *mapbitsp)
+arena_mapbitsp_read(const size_t *mapbitsp)
 {
 
 	return (*mapbitsp);
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind)
 {
 
-	return (arena_mapbitsp_read(arena_mapbitsp_get(chunk, pageind)));
+	return (arena_mapbitsp_read(arena_mapbitsp_get_const(chunk, pageind)));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
@@ -765,7 +818,7 @@ arena_mapbits_size_decode(size_t mapbits)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_unallocated_size_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -775,7 +828,7 @@ arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_large_size_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -786,7 +839,7 @@ arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_small_runind_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -797,7 +850,7 @@ arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE szind_t
-arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 	szind_t binind;
@@ -809,7 +862,7 @@ arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -820,7 +873,7 @@ arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -831,7 +884,7 @@ arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_decommitted_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -842,7 +895,7 @@ arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -851,7 +904,7 @@ arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 
@@ -887,7 +940,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
     size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 
 	assert((size & PAGE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
@@ -901,7 +954,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
     size_t size)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);
 
 	assert((size & PAGE_MASK) == 0);
@@ -913,7 +966,7 @@ arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
 JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_internal_set(arena_chunk_t *chunk, size_t pageind, size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 
 	assert((flags & CHUNK_MAP_UNZEROED) == flags);
 	arena_mapbitsp_write(mapbitsp, flags);
@@ -923,7 +976,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size,
     size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 
 	assert((size & PAGE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
@@ -938,7 +991,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
     szind_t binind)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);
 
 	assert(binind <= BININD_INVALID);
@@ -952,7 +1005,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,
     szind_t binind, size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 
 	assert(binind < BININD_INVALID);
 	assert(pageind - runind >= map_bias);
@@ -1009,7 +1062,7 @@ arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes)
 }
 
 JEMALLOC_INLINE bool
-arena_prof_accum(arena_t *arena, uint64_t accumbytes)
+arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes)
 {
 
 	cassert(config_prof);
@@ -1020,9 +1073,9 @@ arena_prof_accum(arena_t *arena, uint64_t accumbytes)
 	{
 		bool ret;
 
-		malloc_mutex_lock(&arena->lock);
+		malloc_mutex_lock(tsdn, &arena->lock);
 		ret = arena_prof_accum_impl(arena, accumbytes);
-		malloc_mutex_unlock(&arena->lock);
+		malloc_mutex_unlock(tsdn, &arena->lock);
 		return (ret);
 	}
 }
@@ -1040,12 +1093,12 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		size_t pageind;
 		size_t actual_mapbits;
 		size_t rpages_ind;
-		arena_run_t *run;
+		const arena_run_t *run;
 		arena_bin_t *bin;
 		szind_t run_binind, actual_binind;
 		arena_bin_info_t *bin_info;
-		arena_chunk_map_misc_t *miscelm;
-		void *rpages;
+		const arena_chunk_map_misc_t *miscelm;
+		const void *rpages;
 
 		assert(binind != BININD_INVALID);
 		assert(binind < NBINS);
@@ -1058,7 +1111,7 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
 		rpages_ind = pageind - arena_mapbits_small_runind_get(chunk,
 		    pageind);
-		miscelm = arena_miscelm_get(chunk, rpages_ind);
+		miscelm = arena_miscelm_get_const(chunk, rpages_ind);
 		run = &miscelm->run;
 		run_binind = run->binind;
 		bin = &arena->bins[run_binind];
@@ -1158,7 +1211,7 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 }
 
 JEMALLOC_INLINE prof_tctx_t *
-arena_prof_tctx_get(const void *ptr)
+arena_prof_tctx_get(tsdn_t *tsdn, const void *ptr)
 {
 	prof_tctx_t *ret;
 	arena_chunk_t *chunk;
@@ -1174,18 +1227,19 @@ arena_prof_tctx_get(const void *ptr)
 		if (likely((mapbits & CHUNK_MAP_LARGE) == 0))
 			ret = (prof_tctx_t *)(uintptr_t)1U;
 		else {
-			arena_chunk_map_misc_t *elm = arena_miscelm_get(chunk,
-			    pageind);
+			arena_chunk_map_misc_t *elm =
+			    arena_miscelm_get_mutable(chunk, pageind);
 			ret = atomic_read_p(&elm->prof_tctx_pun);
 		}
 	} else
-		ret = huge_prof_tctx_get(ptr);
+		ret = huge_prof_tctx_get(tsdn, ptr);
 
 	return (ret);
 }
 
 JEMALLOC_INLINE void
-arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
+arena_prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx)
 {
 	arena_chunk_t *chunk;
 
@@ -1204,7 +1258,7 @@ arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
 
 			assert(arena_mapbits_large_get(chunk, pageind) != 0);
 
-			elm = arena_miscelm_get(chunk, pageind);
+			elm = arena_miscelm_get_mutable(chunk, pageind);
 			atomic_write_p(&elm->prof_tctx_pun, tctx);
 		} else {
 			/*
@@ -1216,12 +1270,12 @@ arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
 			assert(arena_mapbits_large_get(chunk, pageind) == 0);
 		}
 	} else
-		huge_prof_tctx_set(ptr, tctx);
+		huge_prof_tctx_set(tsdn, ptr, tctx);
 }
 
 JEMALLOC_INLINE void
-arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
-    prof_tctx_t *old_tctx)
+arena_prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
+    const void *old_ptr, prof_tctx_t *old_tctx)
 {
 
 	cassert(config_prof);
@@ -1240,56 +1294,59 @@ arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
 			    0);
 			assert(arena_mapbits_large_get(chunk, pageind) != 0);
 
-			elm = arena_miscelm_get(chunk, pageind);
+			elm = arena_miscelm_get_mutable(chunk, pageind);
 			atomic_write_p(&elm->prof_tctx_pun,
 			    (prof_tctx_t *)(uintptr_t)1U);
 		} else
-			huge_prof_tctx_reset(ptr);
+			huge_prof_tctx_reset(tsdn, ptr);
 	}
 }
 
 JEMALLOC_ALWAYS_INLINE void
-arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks)
+arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks)
 {
+	tsd_t *tsd;
 	ticker_t *decay_ticker;
 
-	if (unlikely(tsd == NULL))
+	if (unlikely(tsdn_null(tsdn)))
 		return;
+	tsd = tsdn_tsd(tsdn);
 	decay_ticker = decay_ticker_get(tsd, arena->ind);
 	if (unlikely(decay_ticker == NULL))
 		return;
 	if (unlikely(ticker_ticks(decay_ticker, nticks)))
-		arena_purge(arena, false);
+		arena_purge(tsdn, arena, false);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-arena_decay_tick(tsd_t *tsd, arena_t *arena)
+arena_decay_tick(tsdn_t *tsdn, arena_t *arena)
 {
 
-	arena_decay_ticks(tsd, arena, 1);
+	arena_decay_ticks(tsdn, arena, 1);
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind, bool zero,
+arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind, bool zero,
     tcache_t *tcache, bool slow_path)
 {
 
+	assert(!tsdn_null(tsdn) || tcache == NULL);
 	assert(size != 0);
 
 	if (likely(tcache != NULL)) {
 		if (likely(size <= SMALL_MAXCLASS)) {
-			return (tcache_alloc_small(tsd, arena, tcache, size,
-			    ind, zero, slow_path));
+			return (tcache_alloc_small(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path));
 		}
 		if (likely(size <= tcache_maxclass)) {
-			return (tcache_alloc_large(tsd, arena, tcache, size,
-			    ind, zero, slow_path));
+			return (tcache_alloc_large(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path));
 		}
 		/* (size > tcache_maxclass) case falls through. */
 		assert(size > tcache_maxclass);
 	}
 
-	return (arena_malloc_hard(tsd, arena, size, ind, zero, tcache));
+	return (arena_malloc_hard(tsdn, arena, size, ind, zero));
 }
 
 JEMALLOC_ALWAYS_INLINE arena_t *
@@ -1306,7 +1363,7 @@ arena_aalloc(const void *ptr)
 
 /* Return the size of the allocation pointed to by ptr. */
 JEMALLOC_ALWAYS_INLINE size_t
-arena_salloc(const void *ptr, bool demote)
+arena_salloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {
 	size_t ret;
 	arena_chunk_t *chunk;
@@ -1349,17 +1406,18 @@ arena_salloc(const void *ptr, bool demote)
 			ret = index2size(binind);
 		}
 	} else
-		ret = huge_salloc(ptr);
+		ret = huge_salloc(tsdn, ptr);
 
 	return (ret);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
+arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool slow_path)
 {
 	arena_chunk_t *chunk;
 	size_t pageind, mapbits;
 
+	assert(!tsdn_null(tsdn) || tcache == NULL);
 	assert(ptr != NULL);
 
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
@@ -1384,11 +1442,12 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 			if (likely(tcache != NULL)) {
 				szind_t binind = arena_ptr_small_binind_get(ptr,
 				    mapbits);
-				tcache_dalloc_small(tsd, tcache, ptr, binind,
-				    slow_path);
+				tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
+				    binind, slow_path);
 			} else {
-				arena_dalloc_small(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr, pageind);
+				arena_dalloc_small(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr, pageind);
 			}
 		} else {
 			size_t size = arena_mapbits_large_size_get(chunk,
@@ -1399,22 +1458,26 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 
 			if (likely(tcache != NULL) && size - large_pad <=
 			    tcache_maxclass) {
-				tcache_dalloc_large(tsd, tcache, ptr, size -
-				    large_pad, slow_path);
+				tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr,
+				    size - large_pad, slow_path);
 			} else {
-				arena_dalloc_large(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr);
+				arena_dalloc_large(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr);
 			}
 		}
 	} else
-		huge_dalloc(tsd, ptr, tcache);
+		huge_dalloc(tsdn, ptr);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path)
 {
 	arena_chunk_t *chunk;
 
+	assert(!tsdn_null(tsdn) || tcache == NULL);
+
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 	if (likely(chunk != ptr)) {
 		if (config_prof && opt_prof) {
@@ -1431,34 +1494,36 @@ arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
 				    pageind) - large_pad;
 			}
 		}
-		assert(s2u(size) == s2u(arena_salloc(ptr, false)));
+		assert(s2u(size) == s2u(arena_salloc(tsdn, ptr, false)));
 
 		if (likely(size <= SMALL_MAXCLASS)) {
 			/* Small allocation. */
 			if (likely(tcache != NULL)) {
 				szind_t binind = size2index(size);
-				tcache_dalloc_small(tsd, tcache, ptr, binind,
-				    true);
+				tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
+				    binind, slow_path);
 			} else {
 				size_t pageind = ((uintptr_t)ptr -
 				    (uintptr_t)chunk) >> LG_PAGE;
-				arena_dalloc_small(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr, pageind);
+				arena_dalloc_small(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr, pageind);
 			}
 		} else {
 			assert(config_cache_oblivious || ((uintptr_t)ptr &
 			    PAGE_MASK) == 0);
 
 			if (likely(tcache != NULL) && size <= tcache_maxclass) {
-				tcache_dalloc_large(tsd, tcache, ptr, size,
-				    true);
+				tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr,
+				    size, slow_path);
 			} else {
-				arena_dalloc_large(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr);
+				arena_dalloc_large(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr);
 			}
 		}
 	} else
-		huge_dalloc(tsd, ptr, tcache);
+		huge_dalloc(tsdn, ptr);
 }
 #  endif /* JEMALLOC_ARENA_INLINE_B */
 #endif
diff --git a/include/jemalloc/internal/base.h b/include/jemalloc/internal/base.h
index 39e46ee..d6b81e1 100644
--- a/include/jemalloc/internal/base.h
+++ b/include/jemalloc/internal/base.h
@@ -9,12 +9,13 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	*base_alloc(size_t size);
-void	base_stats_get(size_t *allocated, size_t *resident, size_t *mapped);
+void	*base_alloc(tsdn_t *tsdn, size_t size);
+void	base_stats_get(tsdn_t *tsdn, size_t *allocated, size_t *resident,
+    size_t *mapped);
 bool	base_boot(void);
-void	base_prefork(void);
-void	base_postfork_parent(void);
-void	base_postfork_child(void);
+void	base_prefork(tsdn_t *tsdn);
+void	base_postfork_parent(tsdn_t *tsdn);
+void	base_postfork_child(tsdn_t *tsdn);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/bitmap.h b/include/jemalloc/internal/bitmap.h
index 2594e3a..36f38b5 100644
--- a/include/jemalloc/internal/bitmap.h
+++ b/include/jemalloc/internal/bitmap.h
@@ -17,8 +17,8 @@ typedef unsigned long bitmap_t;
 
 /*
  * Do some analysis on how big the bitmap is before we use a tree.  For a brute
- * force linear search, if we would have to call ffsl more than 2^3 times, use a
- * tree instead.
+ * force linear search, if we would have to call ffs_lu() more than 2^3 times,
+ * use a tree instead.
  */
 #if LG_BITMAP_MAXBITS - LG_BITMAP_GROUP_NBITS > 3
 #  define USE_TREE
@@ -223,7 +223,7 @@ bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo)
 		i++;
 		g = bitmap[i];
 	}
-	bit = (bit - 1) + (i << 6);
+	bit = (i << LG_BITMAP_GROUP_NBITS) + (bit - 1);
 #endif
 	bitmap_set(bitmap, binfo, bit);
 	return (bit);
diff --git a/include/jemalloc/internal/chunk.h b/include/jemalloc/internal/chunk.h
index 20be4ba..d6ecdab 100644
--- a/include/jemalloc/internal/chunk.h
+++ b/include/jemalloc/internal/chunk.h
@@ -52,28 +52,32 @@ extern size_t		chunk_npages;
 
 extern const chunk_hooks_t	chunk_hooks_default;
 
-chunk_hooks_t	chunk_hooks_get(arena_t *arena);
-chunk_hooks_t	chunk_hooks_set(arena_t *arena,
+chunk_hooks_t	chunk_hooks_get(tsdn_t *tsdn, arena_t *arena);
+chunk_hooks_t	chunk_hooks_set(tsdn_t *tsdn, arena_t *arena,
     const chunk_hooks_t *chunk_hooks);
 
-bool	chunk_register(const void *chunk, const extent_node_t *node);
+bool	chunk_register(tsdn_t *tsdn, const void *chunk,
+    const extent_node_t *node);
 void	chunk_deregister(const void *chunk, const extent_node_t *node);
 void	*chunk_alloc_base(size_t size);
-void	*chunk_alloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *new_addr, size_t size, size_t alignment, bool *zero,
-    bool dalloc_node);
-void	*chunk_alloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit);
-void	chunk_dalloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, bool committed);
-void	chunk_dalloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, bool zeroed, bool committed);
-bool	chunk_purge_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, size_t offset, size_t length);
+void	*chunk_alloc_cache(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment,
+    bool *zero, bool dalloc_node);
+void	*chunk_alloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment,
+    bool *zero, bool *commit);
+void	chunk_dalloc_cache(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, bool committed);
+void	chunk_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, bool zeroed,
+    bool committed);
+bool	chunk_purge_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, size_t offset,
+    size_t length);
 bool	chunk_boot(void);
-void	chunk_prefork(void);
-void	chunk_postfork_parent(void);
-void	chunk_postfork_child(void);
+void	chunk_prefork(tsdn_t *tsdn);
+void	chunk_postfork_parent(tsdn_t *tsdn);
+void	chunk_postfork_child(tsdn_t *tsdn);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/chunk_dss.h b/include/jemalloc/internal/chunk_dss.h
index 388f46b..724fa57 100644
--- a/include/jemalloc/internal/chunk_dss.h
+++ b/include/jemalloc/internal/chunk_dss.h
@@ -21,15 +21,15 @@ extern const char *dss_prec_names[];
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-dss_prec_t	chunk_dss_prec_get(void);
-bool	chunk_dss_prec_set(dss_prec_t dss_prec);
-void	*chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size,
-    size_t alignment, bool *zero, bool *commit);
-bool	chunk_in_dss(void *chunk);
+dss_prec_t	chunk_dss_prec_get(tsdn_t *tsdn);
+bool	chunk_dss_prec_set(tsdn_t *tsdn, dss_prec_t dss_prec);
+void	*chunk_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr,
+    size_t size, size_t alignment, bool *zero, bool *commit);
+bool	chunk_in_dss(tsdn_t *tsdn, void *chunk);
 bool	chunk_dss_boot(void);
-void	chunk_dss_prefork(void);
-void	chunk_dss_postfork_parent(void);
-void	chunk_dss_postfork_child(void);
+void	chunk_dss_prefork(tsdn_t *tsdn);
+void	chunk_dss_postfork_parent(tsdn_t *tsdn);
+void	chunk_dss_postfork_child(tsdn_t *tsdn);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/ckh.h b/include/jemalloc/internal/ckh.h
index f75ad90..46e151c 100644
--- a/include/jemalloc/internal/ckh.h
+++ b/include/jemalloc/internal/ckh.h
@@ -64,13 +64,13 @@ struct ckh_s {
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-bool	ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
+bool	ckh_new(tsdn_t *tsdn, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
     ckh_keycomp_t *keycomp);
-void	ckh_delete(tsd_t *tsd, ckh_t *ckh);
+void	ckh_delete(tsdn_t *tsdn, ckh_t *ckh);
 size_t	ckh_count(ckh_t *ckh);
 bool	ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);
-bool	ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data);
-bool	ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
+bool	ckh_insert(tsdn_t *tsdn, ckh_t *ckh, const void *key, const void *data);
+bool	ckh_remove(tsdn_t *tsdn, ckh_t *ckh, const void *searchkey, void **key,
     void **data);
 bool	ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data);
 void	ckh_string_hash(const void *key, size_t r_hash[2]);
diff --git a/include/jemalloc/internal/ctl.h b/include/jemalloc/internal/ctl.h
index 9c5e932..af0f6d7 100644
--- a/include/jemalloc/internal/ctl.h
+++ b/include/jemalloc/internal/ctl.h
@@ -21,13 +21,14 @@ struct ctl_named_node_s {
 	/* If (nchildren == 0), this is a terminal node. */
 	unsigned		nchildren;
 	const			ctl_node_t *children;
-	int			(*ctl)(const size_t *, size_t, void *, size_t *,
-	    void *, size_t);
+	int			(*ctl)(tsd_t *, const size_t *, size_t, void *,
+	    size_t *, void *, size_t);
 };
 
 struct ctl_indexed_node_s {
 	struct ctl_node_s	node;
-	const ctl_named_node_t	*(*index)(const size_t *, size_t, size_t);
+	const ctl_named_node_t	*(*index)(tsdn_t *, const size_t *, size_t,
+	    size_t);
 };
 
 struct ctl_arena_stats_s {
@@ -60,6 +61,7 @@ struct ctl_stats_s {
 	size_t			metadata;
 	size_t			resident;
 	size_t			mapped;
+	size_t			retained;
 	unsigned		narenas;
 	ctl_arena_stats_t	*arenas;	/* (narenas + 1) elements. */
 };
@@ -68,16 +70,17 @@ struct ctl_stats_s {
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-int	ctl_byname(const char *name, void *oldp, size_t *oldlenp, void *newp,
-    size_t newlen);
-int	ctl_nametomib(const char *name, size_t *mibp, size_t *miblenp);
-
-int	ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
+int	ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
     void *newp, size_t newlen);
+int	ctl_nametomib(tsdn_t *tsdn, const char *name, size_t *mibp,
+    size_t *miblenp);
+
+int	ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen);
 bool	ctl_boot(void);
-void	ctl_prefork(void);
-void	ctl_postfork_parent(void);
-void	ctl_postfork_child(void);
+void	ctl_prefork(tsdn_t *tsdn);
+void	ctl_postfork_parent(tsdn_t *tsdn);
+void	ctl_postfork_child(tsdn_t *tsdn);
 
 #define	xmallctl(name, oldp, oldlenp, newp, newlen) do {		\
 	if (je_mallctl(name, oldp, oldlenp, newp, newlen)		\
diff --git a/include/jemalloc/internal/extent.h b/include/jemalloc/internal/extent.h
index 386d50e..49d76a5 100644
--- a/include/jemalloc/internal/extent.h
+++ b/include/jemalloc/internal/extent.h
@@ -48,7 +48,7 @@ struct extent_node_s {
 		/* Linkage for the size/address-ordered tree. */
 		rb_node(extent_node_t)	szad_link;
 
-		/* Linkage for arena's huge and node_cache lists. */
+		/* Linkage for arena's achunks, huge, and node_cache lists. */
 		ql_elm(extent_node_t)	ql_link;
 	};
 
diff --git a/include/jemalloc/internal/hash.h b/include/jemalloc/internal/hash.h
index f2d907c..1ff2d9a 100644
--- a/include/jemalloc/internal/hash.h
+++ b/include/jemalloc/internal/hash.h
@@ -53,9 +53,7 @@ hash_get_block_32(const uint32_t *p, int i)
 	if (unlikely((uintptr_t)p & (sizeof(uint32_t)-1)) != 0) {
 		uint32_t ret;
 
-		/* ANDROID change */
-		memcpy(&ret, (uint8_t*)(p+i), sizeof(uint32_t));
-		/* End ANDROID change */
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint32_t));
 		return (ret);
 	}
 
@@ -70,9 +68,7 @@ hash_get_block_64(const uint64_t *p, int i)
 	if (unlikely((uintptr_t)p & (sizeof(uint64_t)-1)) != 0) {
 		uint64_t ret;
 
-		/* ANDROID change */
-		memcpy(&ret, (uint8_t*)(p+i), sizeof(uint64_t));
-		/* End ANDROID change */
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint64_t));
 		return (ret);
 	}
 
diff --git a/include/jemalloc/internal/huge.h b/include/jemalloc/internal/huge.h
index cb6f69e..b5fa9e6 100644
--- a/include/jemalloc/internal/huge.h
+++ b/include/jemalloc/internal/huge.h
@@ -9,24 +9,23 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	*huge_malloc(tsd_t *tsd, arena_t *arena, size_t usize, bool zero,
-    tcache_t *tcache);
-void	*huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
-    bool zero, tcache_t *tcache);
-bool	huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize,
+void	*huge_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero);
+void	*huge_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool zero);
+bool	huge_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize,
     size_t usize_min, size_t usize_max, bool zero);
 void	*huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
     size_t usize, size_t alignment, bool zero, tcache_t *tcache);
 #ifdef JEMALLOC_JET
-typedef void (huge_dalloc_junk_t)(void *, size_t);
+typedef void (huge_dalloc_junk_t)(tsdn_t *, void *, size_t);
 extern huge_dalloc_junk_t *huge_dalloc_junk;
 #endif
-void	huge_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache);
+void	huge_dalloc(tsdn_t *tsdn, void *ptr);
 arena_t	*huge_aalloc(const void *ptr);
-size_t	huge_salloc(const void *ptr);
-prof_tctx_t	*huge_prof_tctx_get(const void *ptr);
-void	huge_prof_tctx_set(const void *ptr, prof_tctx_t *tctx);
-void	huge_prof_tctx_reset(const void *ptr);
+size_t	huge_salloc(tsdn_t *tsdn, const void *ptr);
+prof_tctx_t	*huge_prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	huge_prof_tctx_set(tsdn_t *tsdn, const void *ptr, prof_tctx_t *tctx);
+void	huge_prof_tctx_reset(tsdn_t *tsdn, const void *ptr);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/jemalloc_internal.h.in b/include/jemalloc/internal/jemalloc_internal.h.in
index 3f54391..8f82edd 100644
--- a/include/jemalloc/internal/jemalloc_internal.h.in
+++ b/include/jemalloc/internal/jemalloc_internal.h.in
@@ -161,6 +161,7 @@ static const bool config_cache_oblivious =
 #include <malloc/malloc.h>
 #endif
 
+#include "jemalloc/internal/ph.h"
 #define	RB_COMPACT
 #include "jemalloc/internal/rb.h"
 #include "jemalloc/internal/qr.h"
@@ -257,6 +258,9 @@ typedef unsigned szind_t;
 #  ifdef __powerpc__
 #    define LG_QUANTUM		4
 #  endif
+#  ifdef __riscv__
+#    define LG_QUANTUM		4
+#  endif
 #  ifdef __s390__
 #    define LG_QUANTUM		4
 #  endif
@@ -367,6 +371,7 @@ typedef unsigned szind_t;
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/tsd.h"
 #include "jemalloc/internal/mb.h"
@@ -398,6 +403,7 @@ typedef unsigned szind_t;
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/bitmap.h"
@@ -440,6 +446,9 @@ extern bool	in_valgrind;
 /* Number of CPUs. */
 extern unsigned	ncpus;
 
+/* Number of arenas used for automatic multiplexing of threads and arenas. */
+extern unsigned	narenas_auto;
+
 /*
  * Arenas that are used to service external requests.  Not all elements of the
  * arenas array are necessarily used; arenas are created lazily as needed.
@@ -463,14 +472,14 @@ void	a0dalloc(void *ptr);
 void	*bootstrap_malloc(size_t size);
 void	*bootstrap_calloc(size_t num, size_t size);
 void	bootstrap_free(void *ptr);
-arena_t	*arenas_extend(unsigned ind);
 unsigned	narenas_total_get(void);
-arena_t	*arena_init(unsigned ind);
+arena_t	*arena_init(tsdn_t *tsdn, unsigned ind);
 arena_tdata_t	*arena_tdata_get_hard(tsd_t *tsd, unsigned ind);
-arena_t	*arena_choose_hard(tsd_t *tsd);
+arena_t	*arena_choose_hard(tsd_t *tsd, bool internal);
 void	arena_migrate(tsd_t *tsd, unsigned oldind, unsigned newind);
 void	thread_allocated_cleanup(tsd_t *tsd);
 void	thread_deallocated_cleanup(tsd_t *tsd);
+void	iarena_cleanup(tsd_t *tsd);
 void	arena_cleanup(tsd_t *tsd);
 void	arenas_tdata_cleanup(tsd_t *tsd);
 void	narenas_tdata_cleanup(tsd_t *tsd);
@@ -490,6 +499,7 @@ void	jemalloc_postfork_child(void);
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/bitmap.h"
@@ -521,8 +531,9 @@ void	jemalloc_postfork_child(void);
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
-#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/tsd.h"
+#include "jemalloc/internal/witness.h"
+#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/extent.h"
 #include "jemalloc/internal/base.h"
@@ -542,10 +553,12 @@ size_t	s2u_compute(size_t size);
 size_t	s2u_lookup(size_t size);
 size_t	s2u(size_t size);
 size_t	sa2u(size_t size, size_t alignment);
+arena_t	*arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal);
 arena_t	*arena_choose(tsd_t *tsd, arena_t *arena);
+arena_t	*arena_ichoose(tsdn_t *tsdn, arena_t *arena);
 arena_tdata_t	*arena_tdata_get(tsd_t *tsd, unsigned ind,
     bool refresh_if_missing);
-arena_t	*arena_get(unsigned ind, bool init_if_missing);
+arena_t	*arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing);
 ticker_t	*decay_ticker_get(tsd_t *tsd, unsigned ind);
 #endif
 
@@ -780,19 +793,38 @@ sa2u(size_t size, size_t alignment)
 
 /* Choose an arena based on a per-thread value. */
 JEMALLOC_INLINE arena_t *
-arena_choose(tsd_t *tsd, arena_t *arena)
+arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal)
 {
 	arena_t *ret;
 
 	if (arena != NULL)
 		return (arena);
 
-	if (unlikely((ret = tsd_arena_get(tsd)) == NULL))
-		ret = arena_choose_hard(tsd);
+	ret = internal ? tsd_iarena_get(tsd) : tsd_arena_get(tsd);
+	if (unlikely(ret == NULL))
+		ret = arena_choose_hard(tsd, internal);
 
 	return (ret);
 }
 
+JEMALLOC_INLINE arena_t *
+arena_choose(tsd_t *tsd, arena_t *arena)
+{
+
+	return (arena_choose_impl(tsd, arena, false));
+}
+
+JEMALLOC_INLINE arena_t *
+arena_ichoose(tsdn_t *tsdn, arena_t *arena)
+{
+
+	assert(!tsdn_null(tsdn) || arena != NULL);
+
+	if (!tsdn_null(tsdn))
+		return (arena_choose_impl(tsdn_tsd(tsdn), NULL, true));
+	return (arena);
+}
+
 JEMALLOC_INLINE arena_tdata_t *
 arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing)
 {
@@ -819,7 +851,7 @@ arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing)
 }
 
 JEMALLOC_INLINE arena_t *
-arena_get(unsigned ind, bool init_if_missing)
+arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing)
 {
 	arena_t *ret;
 
@@ -829,7 +861,7 @@ arena_get(unsigned ind, bool init_if_missing)
 	if (unlikely(ret == NULL)) {
 		ret = atomic_read_p((void *)&arenas[ind]);
 		if (init_if_missing && unlikely(ret == NULL))
-			ret = arena_init(ind);
+			ret = arena_init(tsdn, ind);
 	}
 	return (ret);
 }
@@ -863,30 +895,27 @@ decay_ticker_get(tsd_t *tsd, unsigned ind)
 
 #ifndef JEMALLOC_ENABLE_INLINE
 arena_t	*iaalloc(const void *ptr);
-size_t	isalloc(const void *ptr, bool demote);
-void	*iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero,
+size_t	isalloc(tsdn_t *tsdn, const void *ptr, bool demote);
+void	*iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero,
     tcache_t *tcache, bool is_metadata, arena_t *arena, bool slow_path);
-void	*imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
-    arena_t *arena);
-void	*imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path);
-void	*icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
-    arena_t *arena);
-void	*icalloc(tsd_t *tsd, size_t size, szind_t ind);
-void	*ipallocztm(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+void	*ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero,
+    bool slow_path);
+void	*ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
     tcache_t *tcache, bool is_metadata, arena_t *arena);
-void	*ipalloct(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+void	*ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
     tcache_t *tcache, arena_t *arena);
 void	*ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero);
-size_t	ivsalloc(const void *ptr, bool demote);
+size_t	ivsalloc(tsdn_t *tsdn, const void *ptr, bool demote);
 size_t	u2rz(size_t usize);
-size_t	p2rz(const void *ptr);
-void	idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+size_t	p2rz(tsdn_t *tsdn, const void *ptr);
+void	idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_metadata,
     bool slow_path);
-void	idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache);
 void	idalloc(tsd_t *tsd, void *ptr);
 void	iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
-void	isdalloct(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
-void	isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
+void	isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
+void	isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
 void	*iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t extra, size_t alignment, bool zero, tcache_t *tcache,
     arena_t *arena);
@@ -894,7 +923,7 @@ void	*iralloct(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t alignment, bool zero, tcache_t *tcache, arena_t *arena);
 void	*iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
     size_t alignment, bool zero);
-bool	ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
+bool	ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
     size_t extra, size_t alignment, bool zero);
 #endif
 
@@ -910,102 +939,85 @@ iaalloc(const void *ptr)
 
 /*
  * Typical usage:
+ *   tsdn_t *tsdn = [...]
  *   void *ptr = [...]
- *   size_t sz = isalloc(ptr, config_prof);
+ *   size_t sz = isalloc(tsdn, ptr, config_prof);
  */
 JEMALLOC_ALWAYS_INLINE size_t
-isalloc(const void *ptr, bool demote)
+isalloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {
 
 	assert(ptr != NULL);
 	/* Demotion only makes sense if config_prof is true. */
 	assert(config_prof || !demote);
 
-	return (arena_salloc(ptr, demote));
+	return (arena_salloc(tsdn, ptr, demote));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero, tcache_t *tcache,
+iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
     bool is_metadata, arena_t *arena, bool slow_path)
 {
 	void *ret;
 
 	assert(size != 0);
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || arena == NULL || arena->ind < narenas_auto);
 
-	ret = arena_malloc(tsd, arena, size, ind, zero, tcache, slow_path);
+	ret = arena_malloc(tsdn, arena, size, ind, zero, tcache, slow_path);
 	if (config_stats && is_metadata && likely(ret != NULL)) {
-		arena_metadata_allocated_add(iaalloc(ret), isalloc(ret,
-		    config_prof));
+		arena_metadata_allocated_add(iaalloc(ret),
+		    isalloc(tsdn, ret, config_prof));
 	}
 	return (ret);
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
-{
-
-	return (iallocztm(tsd, size, ind, false, tcache, false, arena, true));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path)
+ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path)
 {
 
-	return (iallocztm(tsd, size, ind, false, tcache_get(tsd, true), false,
-	    NULL, slow_path));
+	return (iallocztm(tsd_tsdn(tsd), size, ind, zero, tcache_get(tsd, true),
+	    false, NULL, slow_path));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
-{
-
-	return (iallocztm(tsd, size, ind, true, tcache, false, arena, true));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-icalloc(tsd_t *tsd, size_t size, szind_t ind)
-{
-
-	return (iallocztm(tsd, size, ind, true, tcache_get(tsd, true), false,
-	    NULL, true));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-ipallocztm(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
     tcache_t *tcache, bool is_metadata, arena_t *arena)
 {
 	void *ret;
 
 	assert(usize != 0);
 	assert(usize == sa2u(usize, alignment));
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || arena == NULL || arena->ind < narenas_auto);
 
-	ret = arena_palloc(tsd, arena, usize, alignment, zero, tcache);
+	ret = arena_palloc(tsdn, arena, usize, alignment, zero, tcache);
 	assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);
 	if (config_stats && is_metadata && likely(ret != NULL)) {
-		arena_metadata_allocated_add(iaalloc(ret), isalloc(ret,
+		arena_metadata_allocated_add(iaalloc(ret), isalloc(tsdn, ret,
 		    config_prof));
 	}
 	return (ret);
 }
 
 JEMALLOC_ALWAYS_INLINE void *
-ipalloct(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
     tcache_t *tcache, arena_t *arena)
 {
 
-	return (ipallocztm(tsd, usize, alignment, zero, tcache, false, arena));
+	return (ipallocztm(tsdn, usize, alignment, zero, tcache, false, arena));
 }
 
 JEMALLOC_ALWAYS_INLINE void *
 ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero)
 {
 
-	return (ipallocztm(tsd, usize, alignment, zero, tcache_get(tsd, true),
-	    false, NULL));
+	return (ipallocztm(tsd_tsdn(tsd), usize, alignment, zero,
+	    tcache_get(tsd, true), false, NULL));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-ivsalloc(const void *ptr, bool demote)
+ivsalloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {
 	extent_node_t *node;
 
@@ -1017,7 +1029,7 @@ ivsalloc(const void *ptr, bool demote)
 	assert(extent_node_addr_get(node) == ptr ||
 	    extent_node_achunk_get(node));
 
-	return (isalloc(ptr, demote));
+	return (isalloc(tsdn, ptr, demote));
 }
 
 JEMALLOC_INLINE size_t
@@ -1035,39 +1047,34 @@ u2rz(size_t usize)
 }
 
 JEMALLOC_INLINE size_t
-p2rz(const void *ptr)
+p2rz(tsdn_t *tsdn, const void *ptr)
 {
-	size_t usize = isalloc(ptr, false);
+	size_t usize = isalloc(tsdn, ptr, false);
 
 	return (u2rz(usize));
 }
 
 JEMALLOC_ALWAYS_INLINE void
-idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_metadata,
     bool slow_path)
 {
 
 	assert(ptr != NULL);
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || iaalloc(ptr)->ind < narenas_auto);
 	if (config_stats && is_metadata) {
-		arena_metadata_allocated_sub(iaalloc(ptr), isalloc(ptr,
+		arena_metadata_allocated_sub(iaalloc(ptr), isalloc(tsdn, ptr,
 		    config_prof));
 	}
 
-	arena_dalloc(tsd, ptr, tcache, slow_path);
-}
-
-JEMALLOC_ALWAYS_INLINE void
-idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache)
-{
-
-	idalloctm(tsd, ptr, tcache, false, true);
+	arena_dalloc(tsdn, ptr, tcache, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void
 idalloc(tsd_t *tsd, void *ptr)
 {
 
-	idalloctm(tsd, ptr, tcache_get(tsd, false), false, true);
+	idalloctm(tsd_tsdn(tsd), ptr, tcache_get(tsd, false), false, true);
 }
 
 JEMALLOC_ALWAYS_INLINE void
@@ -1077,24 +1084,25 @@ iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 	if (slow_path && config_fill && unlikely(opt_quarantine))
 		quarantine(tsd, ptr);
 	else
-		idalloctm(tsd, ptr, tcache, false, slow_path);
+		idalloctm(tsd_tsdn(tsd), ptr, tcache, false, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-isdalloct(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path)
 {
 
-	arena_sdalloc(tsd, ptr, size, tcache);
+	arena_sdalloc(tsdn, ptr, size, tcache, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache, bool slow_path)
 {
 
-	if (config_fill && unlikely(opt_quarantine))
+	if (slow_path && config_fill && unlikely(opt_quarantine))
 		quarantine(tsd, ptr);
 	else
-		isdalloct(tsd, ptr, size, tcache);
+		isdalloct(tsd_tsdn(tsd), ptr, size, tcache, slow_path);
 }
 
 JEMALLOC_ALWAYS_INLINE void *
@@ -1107,7 +1115,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 	usize = sa2u(size + extra, alignment);
 	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 		return (NULL);
-	p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
+	p = ipalloct(tsd_tsdn(tsd), usize, alignment, zero, tcache, arena);
 	if (p == NULL) {
 		if (extra == 0)
 			return (NULL);
@@ -1115,7 +1123,8 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 		usize = sa2u(size, alignment);
 		if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 			return (NULL);
-		p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
+		p = ipalloct(tsd_tsdn(tsd), usize, alignment, zero, tcache,
+		    arena);
 		if (p == NULL)
 			return (NULL);
 	}
@@ -1125,7 +1134,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 	 */
 	copysize = (size < oldsize) ? size : oldsize;
 	memcpy(p, ptr, copysize);
-	isqalloc(tsd, ptr, oldsize, tcache);
+	isqalloc(tsd, ptr, oldsize, tcache, true);
 	return (p);
 }
 
@@ -1161,7 +1170,7 @@ iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
 }
 
 JEMALLOC_ALWAYS_INLINE bool
-ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t extra,
+ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
     size_t alignment, bool zero)
 {
 
@@ -1174,7 +1183,7 @@ ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t extra,
 		return (true);
 	}
 
-	return (arena_ralloc_no_move(tsd, ptr, oldsize, size, extra, zero));
+	return (arena_ralloc_no_move(tsdn, ptr, oldsize, size, extra, zero));
 }
 #endif
 
diff --git a/include/jemalloc/internal/jemalloc_internal_defs.h.in b/include/jemalloc/internal/jemalloc_internal_defs.h.in
index 2c75371..7de0cf7 100644
--- a/include/jemalloc/internal/jemalloc_internal_defs.h.in
+++ b/include/jemalloc/internal/jemalloc_internal_defs.h.in
@@ -215,6 +215,15 @@
 #undef JEMALLOC_ZONE_VERSION
 
 /*
+ * Methods for determining whether the OS overcommits.
+ * JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY: Linux's
+ *                                         /proc/sys/vm.overcommit_memory file.
+ * JEMALLOC_SYSCTL_VM_OVERCOMMIT: FreeBSD's vm.overcommit sysctl.
+ */
+#undef JEMALLOC_SYSCTL_VM_OVERCOMMIT
+#undef JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY
+
+/*
  * Methods for purging unused pages differ between operating systems.
  *
  *   madvise(..., MADV_DONTNEED) : On Linux, this immediately discards pages,
diff --git a/include/jemalloc/internal/mb.h b/include/jemalloc/internal/mb.h
index 3cfa787..437c86f 100644
--- a/include/jemalloc/internal/mb.h
+++ b/include/jemalloc/internal/mb.h
@@ -42,7 +42,7 @@ mb_write(void)
 	    : /* Inputs. */
 	    : "memory" /* Clobbers. */
 	    );
-#else
+#  else
 	/*
 	 * This is hopefully enough to keep the compiler from reordering
 	 * instructions around this one.
@@ -52,7 +52,7 @@ mb_write(void)
 	    : /* Inputs. */
 	    : "memory" /* Clobbers. */
 	    );
-#endif
+#  endif
 }
 #elif (defined(__amd64__) || defined(__x86_64__))
 JEMALLOC_INLINE void
@@ -104,9 +104,9 @@ mb_write(void)
 {
 	malloc_mutex_t mtx;
 
-	malloc_mutex_init(&mtx);
-	malloc_mutex_lock(&mtx);
-	malloc_mutex_unlock(&mtx);
+	malloc_mutex_init(&mtx, "mb", WITNESS_RANK_OMIT);
+	malloc_mutex_lock(NULL, &mtx);
+	malloc_mutex_unlock(NULL, &mtx);
 }
 #endif
 #endif
diff --git a/include/jemalloc/internal/mutex.h b/include/jemalloc/internal/mutex.h
index f051f29..5221799 100644
--- a/include/jemalloc/internal/mutex.h
+++ b/include/jemalloc/internal/mutex.h
@@ -6,17 +6,21 @@ typedef struct malloc_mutex_s malloc_mutex_t;
 #ifdef _WIN32
 #  define MALLOC_MUTEX_INITIALIZER
 #elif (defined(JEMALLOC_OSSPIN))
-#  define MALLOC_MUTEX_INITIALIZER {0}
+#  define MALLOC_MUTEX_INITIALIZER {0, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
-#  define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL}
+#  define MALLOC_MUTEX_INITIALIZER					\
+    {PTHREAD_MUTEX_INITIALIZER, NULL, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #else
 #  if (defined(JEMALLOC_HAVE_PTHREAD_MUTEX_ADAPTIVE_NP) &&		\
        defined(PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP))
 #    define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_ADAPTIVE_NP
-#    define MALLOC_MUTEX_INITIALIZER {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP}
+#    define MALLOC_MUTEX_INITIALIZER					\
+       {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP,				\
+        WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #  else
 #    define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT
-#    define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER}
+#    define MALLOC_MUTEX_INITIALIZER					\
+       {PTHREAD_MUTEX_INITIALIZER, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #  endif
 #endif
 
@@ -39,6 +43,7 @@ struct malloc_mutex_s {
 #else
 	pthread_mutex_t		lock;
 #endif
+	witness_t		witness;
 };
 
 #endif /* JEMALLOC_H_STRUCTS */
@@ -52,27 +57,31 @@ extern bool isthreaded;
 #  define isthreaded true
 #endif
 
-bool	malloc_mutex_init(malloc_mutex_t *mutex);
-void	malloc_mutex_prefork(malloc_mutex_t *mutex);
-void	malloc_mutex_postfork_parent(malloc_mutex_t *mutex);
-void	malloc_mutex_postfork_child(malloc_mutex_t *mutex);
-bool	mutex_boot(void);
+bool	malloc_mutex_init(malloc_mutex_t *mutex, const char *name,
+    witness_rank_t rank);
+void	malloc_mutex_prefork(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_postfork_parent(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_postfork_child(tsdn_t *tsdn, malloc_mutex_t *mutex);
+bool	malloc_mutex_boot(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES
 
 #ifndef JEMALLOC_ENABLE_INLINE
-void	malloc_mutex_lock(malloc_mutex_t *mutex);
-void	malloc_mutex_unlock(malloc_mutex_t *mutex);
+void	malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_assert_not_owner(tsdn_t *tsdn, malloc_mutex_t *mutex);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MUTEX_C_))
 JEMALLOC_INLINE void
-malloc_mutex_lock(malloc_mutex_t *mutex)
+malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {
 
 	if (isthreaded) {
+		witness_assert_not_owner(tsdn, &mutex->witness);
 #ifdef _WIN32
 #  if _WIN32_WINNT >= 0x0600
 		AcquireSRWLockExclusive(&mutex->lock);
@@ -84,14 +93,16 @@ malloc_mutex_lock(malloc_mutex_t *mutex)
 #else
 		pthread_mutex_lock(&mutex->lock);
 #endif
+		witness_lock(tsdn, &mutex->witness);
 	}
 }
 
 JEMALLOC_INLINE void
-malloc_mutex_unlock(malloc_mutex_t *mutex)
+malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {
 
 	if (isthreaded) {
+		witness_unlock(tsdn, &mutex->witness);
 #ifdef _WIN32
 #  if _WIN32_WINNT >= 0x0600
 		ReleaseSRWLockExclusive(&mutex->lock);
@@ -105,6 +116,22 @@ malloc_mutex_unlock(malloc_mutex_t *mutex)
 #endif
 	}
 }
+
+JEMALLOC_INLINE void
+malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex)
+{
+
+	if (isthreaded)
+		witness_assert_owner(tsdn, &mutex->witness);
+}
+
+JEMALLOC_INLINE void
+malloc_mutex_assert_not_owner(tsdn_t *tsdn, malloc_mutex_t *mutex)
+{
+
+	if (isthreaded)
+		witness_assert_not_owner(tsdn, &mutex->witness);
+}
 #endif
 
 #endif /* JEMALLOC_H_INLINES */
diff --git a/include/jemalloc/internal/nstime.h b/include/jemalloc/internal/nstime.h
index bd04f04..dc293b7 100644
--- a/include/jemalloc/internal/nstime.h
+++ b/include/jemalloc/internal/nstime.h
@@ -1,13 +1,13 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_TYPES
 
-#define JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
+#define	JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
     && _POSIX_MONOTONIC_CLOCK >= 0
 
 typedef struct nstime_s nstime_t;
 
 /* Maximum supported number of seconds (~584 years). */
-#define	NSTIME_SEC_MAX	18446744072
+#define	NSTIME_SEC_MAX	KQU(18446744072)
 
 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/pages.h b/include/jemalloc/internal/pages.h
index da7eb96..e21effd 100644
--- a/include/jemalloc/internal/pages.h
+++ b/include/jemalloc/internal/pages.h
@@ -9,13 +9,14 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	*pages_map(void *addr, size_t size);
+void	*pages_map(void *addr, size_t size, bool *commit);
 void	pages_unmap(void *addr, size_t size);
 void	*pages_trim(void *addr, size_t alloc_size, size_t leadsize,
-    size_t size);
+    size_t size, bool *commit);
 bool	pages_commit(void *addr, size_t size);
 bool	pages_decommit(void *addr, size_t size);
 bool	pages_purge(void *addr, size_t size);
+void	pages_boot(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
diff --git a/include/jemalloc/internal/ph.h b/include/jemalloc/internal/ph.h
new file mode 100644
index 0000000..4f91c33
--- /dev/null
+++ b/include/jemalloc/internal/ph.h
@@ -0,0 +1,345 @@
+/*
+ * A Pairing Heap implementation.
+ *
+ * "The Pairing Heap: A New Form of Self-Adjusting Heap"
+ * https://www.cs.cmu.edu/~sleator/papers/pairing-heaps.pdf
+ *
+ * With auxiliary twopass list, described in a follow on paper.
+ *
+ * "Pairing Heaps: Experiments and Analysis"
+ * http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.106.2988&rep=rep1&type=pdf
+ *
+ *******************************************************************************
+ */
+
+#ifndef PH_H_
+#define	PH_H_
+
+/* Node structure. */
+#define	phn(a_type)							\
+struct {								\
+	a_type	*phn_prev;						\
+	a_type	*phn_next;						\
+	a_type	*phn_lchild;						\
+}
+
+/* Root structure. */
+#define	ph(a_type)							\
+struct {								\
+	a_type	*ph_root;						\
+}
+
+/* Internal utility macros. */
+#define	phn_lchild_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_lchild)
+#define	phn_lchild_set(a_type, a_field, a_phn, a_lchild) do {		\
+	a_phn->a_field.phn_lchild = a_lchild;				\
+} while (0)
+
+#define	phn_next_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_next)
+#define	phn_prev_set(a_type, a_field, a_phn, a_prev) do {		\
+	a_phn->a_field.phn_prev = a_prev;				\
+} while (0)
+
+#define	phn_prev_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_prev)
+#define	phn_next_set(a_type, a_field, a_phn, a_next) do {		\
+	a_phn->a_field.phn_next = a_next;				\
+} while (0)
+
+#define	phn_merge_ordered(a_type, a_field, a_phn0, a_phn1, a_cmp) do {	\
+	a_type *phn0child;						\
+									\
+	assert(a_phn0 != NULL);						\
+	assert(a_phn1 != NULL);						\
+	assert(a_cmp(a_phn0, a_phn1) <= 0);				\
+									\
+	phn_prev_set(a_type, a_field, a_phn1, a_phn0);			\
+	phn0child = phn_lchild_get(a_type, a_field, a_phn0);		\
+	phn_next_set(a_type, a_field, a_phn1, phn0child);		\
+	if (phn0child != NULL)						\
+		phn_prev_set(a_type, a_field, phn0child, a_phn1);	\
+	phn_lchild_set(a_type, a_field, a_phn0, a_phn1);		\
+} while (0)
+
+#define	phn_merge(a_type, a_field, a_phn0, a_phn1, a_cmp, r_phn) do {	\
+	if (a_phn0 == NULL)						\
+		r_phn = a_phn1;						\
+	else if (a_phn1 == NULL)					\
+		r_phn = a_phn0;						\
+	else if (a_cmp(a_phn0, a_phn1) < 0) {				\
+		phn_merge_ordered(a_type, a_field, a_phn0, a_phn1,	\
+		    a_cmp);						\
+		r_phn = a_phn0;						\
+	} else {							\
+		phn_merge_ordered(a_type, a_field, a_phn1, a_phn0,	\
+		    a_cmp);						\
+		r_phn = a_phn1;						\
+	}								\
+} while (0)
+
+#define	ph_merge_siblings(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *head = NULL;						\
+	a_type *tail = NULL;						\
+	a_type *phn0 = a_phn;						\
+	a_type *phn1 = phn_next_get(a_type, a_field, phn0);		\
+									\
+	/*								\
+	 * Multipass merge, wherein the first two elements of a FIFO	\
+	 * are repeatedly merged, and each result is appended to the	\
+	 * singly linked FIFO, until the FIFO contains only a single	\
+	 * element.  We start with a sibling list but no reference to	\
+	 * its tail, so we do a single pass over the sibling list to	\
+	 * populate the FIFO.						\
+	 */								\
+	if (phn1 != NULL) {						\
+		a_type *phnrest = phn_next_get(a_type, a_field, phn1);	\
+		if (phnrest != NULL)					\
+			phn_prev_set(a_type, a_field, phnrest, NULL);	\
+		phn_prev_set(a_type, a_field, phn0, NULL);		\
+		phn_next_set(a_type, a_field, phn0, NULL);		\
+		phn_prev_set(a_type, a_field, phn1, NULL);		\
+		phn_next_set(a_type, a_field, phn1, NULL);		\
+		phn_merge(a_type, a_field, phn0, phn1, a_cmp, phn0);	\
+		head = tail = phn0;					\
+		phn0 = phnrest;						\
+		while (phn0 != NULL) {					\
+			phn1 = phn_next_get(a_type, a_field, phn0);	\
+			if (phn1 != NULL) {				\
+				phnrest = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				if (phnrest != NULL) {			\
+					phn_prev_set(a_type, a_field,	\
+					    phnrest, NULL);		\
+				}					\
+				phn_prev_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_prev_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = phnrest;				\
+			} else {					\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = NULL;				\
+			}						\
+		}							\
+		phn0 = head;						\
+		phn1 = phn_next_get(a_type, a_field, phn0);		\
+		if (phn1 != NULL) {					\
+			while (true) {					\
+				head = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn0) == NULL);			\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn1) == NULL);			\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				if (head == NULL)			\
+					break;				\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = head;				\
+				phn1 = phn_next_get(a_type, a_field,	\
+				    phn0);				\
+			}						\
+		}							\
+	}								\
+	r_phn = phn0;							\
+} while (0)
+
+#define	ph_merge_aux(a_type, a_field, a_ph, a_cmp) do {			\
+	a_type *phn = phn_next_get(a_type, a_field, a_ph->ph_root);	\
+	if (phn != NULL) {						\
+		phn_prev_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_next_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_prev_set(a_type, a_field, phn, NULL);		\
+		ph_merge_siblings(a_type, a_field, phn, a_cmp, phn);	\
+		assert(phn_next_get(a_type, a_field, phn) == NULL);	\
+		phn_merge(a_type, a_field, a_ph->ph_root, phn, a_cmp,	\
+		    a_ph->ph_root);					\
+	}								\
+} while (0)
+
+#define	ph_merge_children(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *lchild = phn_lchild_get(a_type, a_field, a_phn);	\
+	if (lchild == NULL)						\
+		r_phn = NULL;						\
+	else {								\
+		ph_merge_siblings(a_type, a_field, lchild, a_cmp,	\
+		    r_phn);						\
+	}								\
+} while (0)
+
+/*
+ * The ph_proto() macro generates function prototypes that correspond to the
+ * functions generated by an equivalently parameterized call to ph_gen().
+ */
+#define	ph_proto(a_attr, a_prefix, a_ph_type, a_type)			\
+a_attr void	a_prefix##new(a_ph_type *ph);				\
+a_attr bool	a_prefix##empty(a_ph_type *ph);				\
+a_attr a_type	*a_prefix##first(a_ph_type *ph);			\
+a_attr void	a_prefix##insert(a_ph_type *ph, a_type *phn);		\
+a_attr a_type	*a_prefix##remove_first(a_ph_type *ph);			\
+a_attr void	a_prefix##remove(a_ph_type *ph, a_type *phn);
+
+/*
+ * The ph_gen() macro generates a type-specific pairing heap implementation,
+ * based on the above cpp macros.
+ */
+#define	ph_gen(a_attr, a_prefix, a_ph_type, a_type, a_field, a_cmp)	\
+a_attr void								\
+a_prefix##new(a_ph_type *ph)						\
+{									\
+									\
+	memset(ph, 0, sizeof(ph(a_type)));				\
+}									\
+a_attr bool								\
+a_prefix##empty(a_ph_type *ph)						\
+{									\
+									\
+	return (ph->ph_root == NULL);					\
+}									\
+a_attr a_type *								\
+a_prefix##first(a_ph_type *ph)						\
+{									\
+									\
+	if (ph->ph_root == NULL)					\
+		return (NULL);						\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+	return (ph->ph_root);						\
+}									\
+a_attr void								\
+a_prefix##insert(a_ph_type *ph, a_type *phn)				\
+{									\
+									\
+	memset(&phn->a_field, 0, sizeof(phn(a_type)));			\
+									\
+	/*								\
+	 * Treat the root as an aux list during insertion, and lazily	\
+	 * merge during a_prefix##remove_first().  For elements that	\
+	 * are inserted, then removed via a_prefix##remove() before the	\
+	 * aux list is ever processed, this makes insert/remove		\
+	 * constant-time, whereas eager merging would make insert	\
+	 * O(log n).							\
+	 */								\
+	if (ph->ph_root == NULL)					\
+		ph->ph_root = phn;					\
+	else {								\
+		phn_next_set(a_type, a_field, phn, phn_next_get(a_type,	\
+		    a_field, ph->ph_root));				\
+		if (phn_next_get(a_type, a_field, ph->ph_root) !=	\
+		    NULL) {						\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, ph->ph_root),	\
+			    phn);					\
+		}							\
+		phn_prev_set(a_type, a_field, phn, ph->ph_root);	\
+		phn_next_set(a_type, a_field, ph->ph_root, phn);	\
+	}								\
+}									\
+a_attr a_type *								\
+a_prefix##remove_first(a_ph_type *ph)					\
+{									\
+	a_type *ret;							\
+									\
+	if (ph->ph_root == NULL)					\
+		return (NULL);						\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+									\
+	ret = ph->ph_root;						\
+									\
+	ph_merge_children(a_type, a_field, ph->ph_root, a_cmp,		\
+	    ph->ph_root);						\
+									\
+	return (ret);							\
+}									\
+a_attr void								\
+a_prefix##remove(a_ph_type *ph, a_type *phn)				\
+{									\
+	a_type *replace, *parent;					\
+									\
+	/*								\
+	 * We can delete from aux list without merging it, but we need	\
+	 * to merge if we are dealing with the root node.		\
+	 */								\
+	if (ph->ph_root == phn) {					\
+		ph_merge_aux(a_type, a_field, ph, a_cmp);		\
+		if (ph->ph_root == phn) {				\
+			ph_merge_children(a_type, a_field, ph->ph_root,	\
+			    a_cmp, ph->ph_root);			\
+			return;						\
+		}							\
+	}								\
+									\
+	/* Get parent (if phn is leftmost child) before mutating. */	\
+	if ((parent = phn_prev_get(a_type, a_field, phn)) != NULL) {	\
+		if (phn_lchild_get(a_type, a_field, parent) != phn)	\
+			parent = NULL;					\
+	}								\
+	/* Find a possible replacement node, and link to parent. */	\
+	ph_merge_children(a_type, a_field, phn, a_cmp, replace);	\
+	/* Set next/prev for sibling linked list. */			\
+	if (replace != NULL) {						\
+		if (parent != NULL) {					\
+			phn_prev_set(a_type, a_field, replace, parent);	\
+			phn_lchild_set(a_type, a_field, parent,		\
+			    replace);					\
+		} else {						\
+			phn_prev_set(a_type, a_field, replace,		\
+			    phn_prev_get(a_type, a_field, phn));	\
+			if (phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL) {					\
+				phn_next_set(a_type, a_field,		\
+				    phn_prev_get(a_type, a_field, phn),	\
+				    replace);				\
+			}						\
+		}							\
+		phn_next_set(a_type, a_field, replace,			\
+		    phn_next_get(a_type, a_field, phn));		\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    replace);					\
+		}							\
+	} else {							\
+		if (parent != NULL) {					\
+			a_type *next = phn_next_get(a_type, a_field,	\
+			    phn);					\
+			phn_lchild_set(a_type, a_field, parent, next);	\
+			if (next != NULL) {				\
+				phn_prev_set(a_type, a_field, next,	\
+				    parent);				\
+			}						\
+		} else {						\
+			assert(phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL);					\
+			phn_next_set(a_type, a_field,			\
+			    phn_prev_get(a_type, a_field, phn),		\
+			    phn_next_get(a_type, a_field, phn));	\
+		}							\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    phn_prev_get(a_type, a_field, phn));	\
+		}							\
+	}								\
+}
+
+#endif /* PH_H_ */
diff --git a/include/jemalloc/internal/private_symbols.txt b/include/jemalloc/internal/private_symbols.txt
index 5fcc669..f2b6a55 100644
--- a/include/jemalloc/internal/private_symbols.txt
+++ b/include/jemalloc/internal/private_symbols.txt
@@ -5,10 +5,12 @@ arena_alloc_junk_small
 arena_basic_stats_merge
 arena_bin_index
 arena_bin_info
-arena_bitselm_get
+arena_bitselm_get_const
+arena_bitselm_get_mutable
 arena_boot
 arena_choose
 arena_choose_hard
+arena_choose_impl
 arena_chunk_alloc_huge
 arena_chunk_cache_maybe_insert
 arena_chunk_cache_maybe_remove
@@ -34,6 +36,7 @@ arena_decay_time_set
 arena_dss_prec_get
 arena_dss_prec_set
 arena_get
+arena_ichoose
 arena_init
 arena_lg_dirty_mult_default_get
 arena_lg_dirty_mult_default_set
@@ -60,7 +63,8 @@ arena_mapbits_unallocated_set
 arena_mapbits_unallocated_size_get
 arena_mapbits_unallocated_size_set
 arena_mapbits_unzeroed_get
-arena_mapbitsp_get
+arena_mapbitsp_get_const
+arena_mapbitsp_get_mutable
 arena_mapbitsp_read
 arena_mapbitsp_write
 arena_maxrun
@@ -69,7 +73,8 @@ arena_metadata_allocated_add
 arena_metadata_allocated_get
 arena_metadata_allocated_sub
 arena_migrate
-arena_miscelm_get
+arena_miscelm_get_const
+arena_miscelm_get_mutable
 arena_miscelm_to_pageind
 arena_miscelm_to_rpages
 arena_new
@@ -81,7 +86,10 @@ arena_nthreads_inc
 arena_palloc
 arena_postfork_child
 arena_postfork_parent
-arena_prefork
+arena_prefork0
+arena_prefork1
+arena_prefork2
+arena_prefork3
 arena_prof_accum
 arena_prof_accum_impl
 arena_prof_accum_locked
@@ -97,6 +105,7 @@ arena_ralloc_junk_large
 arena_ralloc_no_move
 arena_rd_to_miscelm
 arena_redzone_corruption
+arena_reset
 arena_run_regind
 arena_run_to_miscelm
 arena_salloc
@@ -123,6 +132,11 @@ atomic_sub_u
 atomic_sub_uint32
 atomic_sub_uint64
 atomic_sub_z
+atomic_write_p
+atomic_write_u
+atomic_write_uint32
+atomic_write_uint64
+atomic_write_z
 base_alloc
 base_boot
 base_postfork_child
@@ -198,6 +212,8 @@ extent_node_addr_get
 extent_node_addr_set
 extent_node_arena_get
 extent_node_arena_set
+extent_node_committed_get
+extent_node_committed_set
 extent_node_dirty_insert
 extent_node_dirty_linkage_init
 extent_node_dirty_remove
@@ -208,6 +224,8 @@ extent_node_size_get
 extent_node_size_set
 extent_node_zeroed_get
 extent_node_zeroed_set
+extent_tree_ad_destroy
+extent_tree_ad_destroy_recurse
 extent_tree_ad_empty
 extent_tree_ad_first
 extent_tree_ad_insert
@@ -225,6 +243,8 @@ extent_tree_ad_reverse_iter
 extent_tree_ad_reverse_iter_recurse
 extent_tree_ad_reverse_iter_start
 extent_tree_ad_search
+extent_tree_szad_destroy
+extent_tree_szad_destroy_recurse
 extent_tree_szad_empty
 extent_tree_szad_first
 extent_tree_szad_insert
@@ -271,14 +291,11 @@ huge_ralloc
 huge_ralloc_no_move
 huge_salloc
 iaalloc
+ialloc
 iallocztm
-icalloc
-icalloct
+iarena_cleanup
 idalloc
-idalloct
 idalloctm
-imalloc
-imalloct
 in_valgrind
 index2size
 index2size_compute
@@ -302,7 +319,11 @@ jemalloc_postfork_parent
 jemalloc_prefork
 large_maxclass
 lg_floor
+lg_prof_sample
 malloc_cprintf
+malloc_mutex_assert_not_owner
+malloc_mutex_assert_owner
+malloc_mutex_boot
 malloc_mutex_init
 malloc_mutex_lock
 malloc_mutex_postfork_child
@@ -324,11 +345,13 @@ malloc_write
 map_bias
 map_misc_offset
 mb_write
-mutex_boot
+narenas_auto
 narenas_tdata_cleanup
 narenas_total_get
 ncpus
 nhbins
+nhclasses
+nlclasses
 nstime_add
 nstime_compare
 nstime_copy
@@ -371,6 +394,7 @@ opt_utrace
 opt_xmalloc
 opt_zero
 p2rz
+pages_boot
 pages_commit
 pages_decommit
 pages_map
@@ -382,6 +406,7 @@ pow2_ceil_u64
 pow2_ceil_zu
 prng_lg_range
 prng_range
+prof_active
 prof_active_get
 prof_active_get_unlocked
 prof_active_set
@@ -391,6 +416,7 @@ prof_backtrace
 prof_boot0
 prof_boot1
 prof_boot2
+prof_bt_count
 prof_dump_header
 prof_dump_open
 prof_free
@@ -408,7 +434,8 @@ prof_malloc_sample_object
 prof_mdump
 prof_postfork_child
 prof_postfork_parent
-prof_prefork
+prof_prefork0
+prof_prefork1
 prof_realloc
 prof_reset
 prof_sample_accum_update
@@ -417,6 +444,7 @@ prof_tctx_get
 prof_tctx_reset
 prof_tctx_set
 prof_tdata_cleanup
+prof_tdata_count
 prof_tdata_get
 prof_tdata_init
 prof_tdata_reinit
@@ -468,8 +496,6 @@ tcache_alloc_easy
 tcache_alloc_large
 tcache_alloc_small
 tcache_alloc_small_hard
-tcache_arena_associate
-tcache_arena_dissociate
 tcache_arena_reassociate
 tcache_bin_flush_large
 tcache_bin_flush_small
@@ -504,38 +530,83 @@ ticker_tick
 ticker_ticks
 tsd_arena_get
 tsd_arena_set
+tsd_arenap_get
+tsd_arenas_tdata_bypass_get
+tsd_arenas_tdata_bypass_set
+tsd_arenas_tdata_bypassp_get
+tsd_arenas_tdata_get
+tsd_arenas_tdata_set
+tsd_arenas_tdatap_get
 tsd_boot
 tsd_boot0
 tsd_boot1
 tsd_booted
+tsd_booted_get
 tsd_cleanup
 tsd_cleanup_wrapper
 tsd_fetch
 tsd_get
-tsd_wrapper_get
-tsd_wrapper_set
+tsd_iarena_get
+tsd_iarena_set
+tsd_iarenap_get
 tsd_initialized
 tsd_init_check_recursion
 tsd_init_finish
 tsd_init_head
+tsd_narenas_tdata_get
+tsd_narenas_tdata_set
+tsd_narenas_tdatap_get
+tsd_wrapper_get
+tsd_wrapper_set
 tsd_nominal
 tsd_prof_tdata_get
 tsd_prof_tdata_set
+tsd_prof_tdatap_get
 tsd_quarantine_get
 tsd_quarantine_set
+tsd_quarantinep_get
 tsd_set
 tsd_tcache_enabled_get
 tsd_tcache_enabled_set
+tsd_tcache_enabledp_get
 tsd_tcache_get
 tsd_tcache_set
+tsd_tcachep_get
 tsd_thread_allocated_get
 tsd_thread_allocated_set
+tsd_thread_allocatedp_get
 tsd_thread_deallocated_get
 tsd_thread_deallocated_set
+tsd_thread_deallocatedp_get
 tsd_tls
 tsd_tsd
+tsd_tsdn
+tsd_witness_fork_get
+tsd_witness_fork_set
+tsd_witness_forkp_get
+tsd_witnesses_get
+tsd_witnesses_set
+tsd_witnessesp_get
+tsdn_fetch
+tsdn_null
+tsdn_tsd
 u2rz
 valgrind_freelike_block
 valgrind_make_mem_defined
 valgrind_make_mem_noaccess
 valgrind_make_mem_undefined
+witness_assert_lockless
+witness_assert_not_owner
+witness_assert_owner
+witness_fork_cleanup
+witness_init
+witness_lock
+witness_lock_error
+witness_lockless_error
+witness_not_owner_error
+witness_owner_error
+witness_postfork_child
+witness_postfork_parent
+witness_prefork
+witness_unlock
+witnesses_cleanup
diff --git a/include/jemalloc/internal/prof.h b/include/jemalloc/internal/prof.h
index a25502a..21dff5f 100644
--- a/include/jemalloc/internal/prof.h
+++ b/include/jemalloc/internal/prof.h
@@ -281,7 +281,7 @@ extern uint64_t	prof_interval;
 extern size_t	lg_prof_sample;
 
 void	prof_alloc_rollback(tsd_t *tsd, prof_tctx_t *tctx, bool updated);
-void	prof_malloc_sample_object(const void *ptr, size_t usize,
+void	prof_malloc_sample_object(tsdn_t *tsdn, const void *ptr, size_t usize,
     prof_tctx_t *tctx);
 void	prof_free_sampled_object(tsd_t *tsd, size_t usize, prof_tctx_t *tctx);
 void	bt_init(prof_bt_t *bt, void **vec);
@@ -293,32 +293,33 @@ size_t	prof_bt_count(void);
 const prof_cnt_t *prof_cnt_all(void);
 typedef int (prof_dump_open_t)(bool, const char *);
 extern prof_dump_open_t *prof_dump_open;
-typedef bool (prof_dump_header_t)(bool, const prof_cnt_t *);
+typedef bool (prof_dump_header_t)(tsdn_t *, bool, const prof_cnt_t *);
 extern prof_dump_header_t *prof_dump_header;
 #endif
-void	prof_idump(void);
-bool	prof_mdump(const char *filename);
-void	prof_gdump(void);
-prof_tdata_t	*prof_tdata_init(tsd_t *tsd);
+void	prof_idump(tsdn_t *tsdn);
+bool	prof_mdump(tsd_t *tsd, const char *filename);
+void	prof_gdump(tsdn_t *tsdn);
+prof_tdata_t	*prof_tdata_init(tsdn_t *tsdn);
 prof_tdata_t	*prof_tdata_reinit(tsd_t *tsd, prof_tdata_t *tdata);
-void	prof_reset(tsd_t *tsd, size_t lg_sample);
+void	prof_reset(tsdn_t *tsdn, size_t lg_sample);
 void	prof_tdata_cleanup(tsd_t *tsd);
-const char	*prof_thread_name_get(void);
-bool	prof_active_get(void);
-bool	prof_active_set(bool active);
+bool	prof_active_get(tsdn_t *tsdn);
+bool	prof_active_set(tsdn_t *tsdn, bool active);
+const char	*prof_thread_name_get(tsd_t *tsd);
 int	prof_thread_name_set(tsd_t *tsd, const char *thread_name);
-bool	prof_thread_active_get(void);
-bool	prof_thread_active_set(bool active);
-bool	prof_thread_active_init_get(void);
-bool	prof_thread_active_init_set(bool active_init);
-bool	prof_gdump_get(void);
-bool	prof_gdump_set(bool active);
+bool	prof_thread_active_get(tsd_t *tsd);
+bool	prof_thread_active_set(tsd_t *tsd, bool active);
+bool	prof_thread_active_init_get(tsdn_t *tsdn);
+bool	prof_thread_active_init_set(tsdn_t *tsdn, bool active_init);
+bool	prof_gdump_get(tsdn_t *tsdn);
+bool	prof_gdump_set(tsdn_t *tsdn, bool active);
 void	prof_boot0(void);
 void	prof_boot1(void);
-bool	prof_boot2(void);
-void	prof_prefork(void);
-void	prof_postfork_parent(void);
-void	prof_postfork_child(void);
+bool	prof_boot2(tsdn_t *tsdn);
+void	prof_prefork0(tsdn_t *tsdn);
+void	prof_prefork1(tsdn_t *tsdn);
+void	prof_postfork_parent(tsdn_t *tsdn);
+void	prof_postfork_child(tsdn_t *tsdn);
 void	prof_sample_threshold_update(prof_tdata_t *tdata);
 
 #endif /* JEMALLOC_H_EXTERNS */
@@ -329,17 +330,17 @@ void	prof_sample_threshold_update(prof_tdata_t *tdata);
 bool	prof_active_get_unlocked(void);
 bool	prof_gdump_get_unlocked(void);
 prof_tdata_t	*prof_tdata_get(tsd_t *tsd, bool create);
+prof_tctx_t	*prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void	prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
+    const void *old_ptr, prof_tctx_t *tctx);
 bool	prof_sample_accum_update(tsd_t *tsd, size_t usize, bool commit,
     prof_tdata_t **tdata_out);
 prof_tctx_t	*prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active,
     bool update);
-prof_tctx_t	*prof_tctx_get(const void *ptr);
-void	prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
-void	prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
-    prof_tctx_t *tctx);
-void	prof_malloc_sample_object(const void *ptr, size_t usize,
+void	prof_malloc(tsdn_t *tsdn, const void *ptr, size_t usize,
     prof_tctx_t *tctx);
-void	prof_malloc(const void *ptr, size_t usize, prof_tctx_t *tctx);
 void	prof_realloc(tsd_t *tsd, const void *ptr, size_t usize,
     prof_tctx_t *tctx, bool prof_active, bool updated, const void *old_ptr,
     size_t old_usize, prof_tctx_t *old_tctx);
@@ -383,7 +384,7 @@ prof_tdata_get(tsd_t *tsd, bool create)
 	if (create) {
 		if (unlikely(tdata == NULL)) {
 			if (tsd_nominal(tsd)) {
-				tdata = prof_tdata_init(tsd);
+				tdata = prof_tdata_init(tsd_tsdn(tsd));
 				tsd_prof_tdata_set(tsd, tdata);
 			}
 		} else if (unlikely(tdata->expired)) {
@@ -397,34 +398,34 @@ prof_tdata_get(tsd_t *tsd, bool create)
 }
 
 JEMALLOC_ALWAYS_INLINE prof_tctx_t *
-prof_tctx_get(const void *ptr)
+prof_tctx_get(tsdn_t *tsdn, const void *ptr)
 {
 
 	cassert(config_prof);
 	assert(ptr != NULL);
 
-	return (arena_prof_tctx_get(ptr));
+	return (arena_prof_tctx_get(tsdn, ptr));
 }
 
 JEMALLOC_ALWAYS_INLINE void
-prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
+prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize, prof_tctx_t *tctx)
 {
 
 	cassert(config_prof);
 	assert(ptr != NULL);
 
-	arena_prof_tctx_set(ptr, usize, tctx);
+	arena_prof_tctx_set(tsdn, ptr, usize, tctx);
 }
 
 JEMALLOC_ALWAYS_INLINE void
-prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
+prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize, const void *old_ptr,
     prof_tctx_t *old_tctx)
 {
 
 	cassert(config_prof);
 	assert(ptr != NULL);
 
-	arena_prof_tctx_reset(ptr, usize, old_ptr, old_tctx);
+	arena_prof_tctx_reset(tsdn, ptr, usize, old_ptr, old_tctx);
 }
 
 JEMALLOC_ALWAYS_INLINE bool
@@ -479,17 +480,17 @@ prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active, bool update)
 }
 
 JEMALLOC_ALWAYS_INLINE void
-prof_malloc(const void *ptr, size_t usize, prof_tctx_t *tctx)
+prof_malloc(tsdn_t *tsdn, const void *ptr, size_t usize, prof_tctx_t *tctx)
 {
 
 	cassert(config_prof);
 	assert(ptr != NULL);
-	assert(usize == isalloc(ptr, true));
+	assert(usize == isalloc(tsdn, ptr, true));
 
 	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
-		prof_malloc_sample_object(ptr, usize, tctx);
+		prof_malloc_sample_object(tsdn, ptr, usize, tctx);
 	else
-		prof_tctx_set(ptr, usize, (prof_tctx_t *)(uintptr_t)1U);
+		prof_tctx_set(tsdn, ptr, usize, (prof_tctx_t *)(uintptr_t)1U);
 }
 
 JEMALLOC_ALWAYS_INLINE void
@@ -503,7 +504,7 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 	assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U);
 
 	if (prof_active && !updated && ptr != NULL) {
-		assert(usize == isalloc(ptr, true));
+		assert(usize == isalloc(tsd_tsdn(tsd), ptr, true));
 		if (prof_sample_accum_update(tsd, usize, true, NULL)) {
 			/*
 			 * Don't sample.  The usize passed to prof_alloc_prep()
@@ -512,6 +513,7 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 			 * though its actual usize was insufficient to cross the
 			 * sample threshold.
 			 */
+			prof_alloc_rollback(tsd, tctx, true);
 			tctx = (prof_tctx_t *)(uintptr_t)1U;
 		}
 	}
@@ -520,9 +522,9 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 	old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U);
 
 	if (unlikely(sampled))
-		prof_malloc_sample_object(ptr, usize, tctx);
+		prof_malloc_sample_object(tsd_tsdn(tsd), ptr, usize, tctx);
 	else
-		prof_tctx_reset(ptr, usize, old_ptr, old_tctx);
+		prof_tctx_reset(tsd_tsdn(tsd), ptr, usize, old_ptr, old_tctx);
 
 	if (unlikely(old_sampled))
 		prof_free_sampled_object(tsd, old_usize, old_tctx);
@@ -531,10 +533,10 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 JEMALLOC_ALWAYS_INLINE void
 prof_free(tsd_t *tsd, const void *ptr, size_t usize)
 {
-	prof_tctx_t *tctx = prof_tctx_get(ptr);
+	prof_tctx_t *tctx = prof_tctx_get(tsd_tsdn(tsd), ptr);
 
 	cassert(config_prof);
-	assert(usize == isalloc(ptr, true));
+	assert(usize == isalloc(tsd_tsdn(tsd), ptr, true));
 
 	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
 		prof_free_sampled_object(tsd, usize, tctx);
diff --git a/include/jemalloc/internal/rtree.h b/include/jemalloc/internal/rtree.h
index 28ae9d1..8d0c584 100644
--- a/include/jemalloc/internal/rtree.h
+++ b/include/jemalloc/internal/rtree.h
@@ -15,9 +15,10 @@ typedef struct rtree_s rtree_t;
  * machine address width.
  */
 #define	LG_RTREE_BITS_PER_LEVEL	4
-#define	RTREE_BITS_PER_LEVEL	(ZU(1) << LG_RTREE_BITS_PER_LEVEL)
+#define	RTREE_BITS_PER_LEVEL	(1U << LG_RTREE_BITS_PER_LEVEL)
+/* Maximum rtree height. */
 #define	RTREE_HEIGHT_MAX						\
-    ((ZU(1) << (LG_SIZEOF_PTR+3)) / RTREE_BITS_PER_LEVEL)
+    ((1U << (LG_SIZEOF_PTR+3)) / RTREE_BITS_PER_LEVEL)
 
 /* Used for two-stage lock-free node initialization. */
 #define	RTREE_NODE_INITIALIZING	((rtree_node_elm_t *)0x1)
@@ -111,22 +112,25 @@ unsigned	rtree_start_level(rtree_t *rtree, uintptr_t key);
 uintptr_t	rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level);
 
 bool	rtree_node_valid(rtree_node_elm_t *node);
-rtree_node_elm_t	*rtree_child_tryread(rtree_node_elm_t *elm);
+rtree_node_elm_t	*rtree_child_tryread(rtree_node_elm_t *elm,
+    bool dependent);
 rtree_node_elm_t	*rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm,
-    unsigned level);
+    unsigned level, bool dependent);
 extent_node_t	*rtree_val_read(rtree_t *rtree, rtree_node_elm_t *elm,
     bool dependent);
 void	rtree_val_write(rtree_t *rtree, rtree_node_elm_t *elm,
     const extent_node_t *val);
-rtree_node_elm_t	*rtree_subtree_tryread(rtree_t *rtree, unsigned level);
-rtree_node_elm_t	*rtree_subtree_read(rtree_t *rtree, unsigned level);
+rtree_node_elm_t	*rtree_subtree_tryread(rtree_t *rtree, unsigned level,
+    bool dependent);
+rtree_node_elm_t	*rtree_subtree_read(rtree_t *rtree, unsigned level,
+    bool dependent);
 
 extent_node_t	*rtree_get(rtree_t *rtree, uintptr_t key, bool dependent);
 bool	rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_RTREE_C_))
-JEMALLOC_INLINE unsigned
+JEMALLOC_ALWAYS_INLINE unsigned
 rtree_start_level(rtree_t *rtree, uintptr_t key)
 {
 	unsigned start_level;
@@ -140,7 +144,7 @@ rtree_start_level(rtree_t *rtree, uintptr_t key)
 	return (start_level);
 }
 
-JEMALLOC_INLINE uintptr_t
+JEMALLOC_ALWAYS_INLINE uintptr_t
 rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level)
 {
 
@@ -149,37 +153,40 @@ rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level)
 	    rtree->levels[level].bits) - 1));
 }
 
-JEMALLOC_INLINE bool
+JEMALLOC_ALWAYS_INLINE bool
 rtree_node_valid(rtree_node_elm_t *node)
 {
 
 	return ((uintptr_t)node > (uintptr_t)RTREE_NODE_INITIALIZING);
 }
 
-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_child_tryread(rtree_node_elm_t *elm)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_child_tryread(rtree_node_elm_t *elm, bool dependent)
 {
 	rtree_node_elm_t *child;
 
 	/* Double-checked read (first read may be stale. */
 	child = elm->child;
-	if (!rtree_node_valid(child))
+	if (!dependent && !rtree_node_valid(child))
 		child = atomic_read_p(&elm->pun);
+	assert(!dependent || child != NULL);
 	return (child);
 }
 
-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm, unsigned level,
+    bool dependent)
 {
 	rtree_node_elm_t *child;
 
-	child = rtree_child_tryread(elm);
-	if (unlikely(!rtree_node_valid(child)))
+	child = rtree_child_tryread(elm, dependent);
+	if (!dependent && unlikely(!rtree_node_valid(child)))
 		child = rtree_child_read_hard(rtree, elm, level);
+	assert(!dependent || child != NULL);
 	return (child);
 }
 
-JEMALLOC_INLINE extent_node_t *
+JEMALLOC_ALWAYS_INLINE extent_node_t *
 rtree_val_read(rtree_t *rtree, rtree_node_elm_t *elm, bool dependent)
 {
 
@@ -208,54 +215,119 @@ rtree_val_write(rtree_t *rtree, rtree_node_elm_t *elm, const extent_node_t *val)
 	atomic_write_p(&elm->pun, val);
 }
 
-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_subtree_tryread(rtree_t *rtree, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_subtree_tryread(rtree_t *rtree, unsigned level, bool dependent)
 {
 	rtree_node_elm_t *subtree;
 
 	/* Double-checked read (first read may be stale. */
 	subtree = rtree->levels[level].subtree;
-	if (!rtree_node_valid(subtree))
+	if (!dependent && unlikely(!rtree_node_valid(subtree)))
 		subtree = atomic_read_p(&rtree->levels[level].subtree_pun);
+	assert(!dependent || subtree != NULL);
 	return (subtree);
 }
 
-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_subtree_read(rtree_t *rtree, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_subtree_read(rtree_t *rtree, unsigned level, bool dependent)
 {
 	rtree_node_elm_t *subtree;
 
-	subtree = rtree_subtree_tryread(rtree, level);
-	if (unlikely(!rtree_node_valid(subtree)))
+	subtree = rtree_subtree_tryread(rtree, level, dependent);
+	if (!dependent && unlikely(!rtree_node_valid(subtree)))
 		subtree = rtree_subtree_read_hard(rtree, level);
+	assert(!dependent || subtree != NULL);
 	return (subtree);
 }
 
-JEMALLOC_INLINE extent_node_t *
+JEMALLOC_ALWAYS_INLINE extent_node_t *
 rtree_get(rtree_t *rtree, uintptr_t key, bool dependent)
 {
 	uintptr_t subkey;
-	unsigned i, start_level;
-	rtree_node_elm_t *node, *child;
+	unsigned start_level;
+	rtree_node_elm_t *node;
 
 	start_level = rtree_start_level(rtree, key);
 
-	for (i = start_level, node = rtree_subtree_tryread(rtree, start_level);
-	    /**/; i++, node = child) {
-		if (!dependent && unlikely(!rtree_node_valid(node)))
-			return (NULL);
-		subkey = rtree_subkey(rtree, key, i);
-		if (i == rtree->height - 1) {
-			/*
-			 * node is a leaf, so it contains values rather than
-			 * child pointers.
-			 */
-			return (rtree_val_read(rtree, &node[subkey],
-			    dependent));
-		}
-		assert(i < rtree->height - 1);
-		child = rtree_child_tryread(&node[subkey]);
+	node = rtree_subtree_tryread(rtree, start_level, dependent);
+#define	RTREE_GET_BIAS	(RTREE_HEIGHT_MAX - rtree->height)
+	switch (start_level + RTREE_GET_BIAS) {
+#define	RTREE_GET_SUBTREE(level)					\
+	case level:							\
+		assert(level < (RTREE_HEIGHT_MAX-1));			\
+		if (!dependent && unlikely(!rtree_node_valid(node)))	\
+			return (NULL);					\
+		subkey = rtree_subkey(rtree, key, level -		\
+		    RTREE_GET_BIAS);					\
+		node = rtree_child_tryread(&node[subkey], dependent);	\
+		/* Fall through. */
+#define	RTREE_GET_LEAF(level)						\
+	case level:							\
+		assert(level == (RTREE_HEIGHT_MAX-1));			\
+		if (!dependent && unlikely(!rtree_node_valid(node)))	\
+			return (NULL);					\
+		subkey = rtree_subkey(rtree, key, level -		\
+		    RTREE_GET_BIAS);					\
+		/*							\
+		 * node is a leaf, so it contains values rather than	\
+		 * child pointers.					\
+		 */							\
+		return (rtree_val_read(rtree, &node[subkey],		\
+		    dependent));
+#if RTREE_HEIGHT_MAX > 1
+	RTREE_GET_SUBTREE(0)
+#endif
+#if RTREE_HEIGHT_MAX > 2
+	RTREE_GET_SUBTREE(1)
+#endif
+#if RTREE_HEIGHT_MAX > 3
+	RTREE_GET_SUBTREE(2)
+#endif
+#if RTREE_HEIGHT_MAX > 4
+	RTREE_GET_SUBTREE(3)
+#endif
+#if RTREE_HEIGHT_MAX > 5
+	RTREE_GET_SUBTREE(4)
+#endif
+#if RTREE_HEIGHT_MAX > 6
+	RTREE_GET_SUBTREE(5)
+#endif
+#if RTREE_HEIGHT_MAX > 7
+	RTREE_GET_SUBTREE(6)
+#endif
+#if RTREE_HEIGHT_MAX > 8
+	RTREE_GET_SUBTREE(7)
+#endif
+#if RTREE_HEIGHT_MAX > 9
+	RTREE_GET_SUBTREE(8)
+#endif
+#if RTREE_HEIGHT_MAX > 10
+	RTREE_GET_SUBTREE(9)
+#endif
+#if RTREE_HEIGHT_MAX > 11
+	RTREE_GET_SUBTREE(10)
+#endif
+#if RTREE_HEIGHT_MAX > 12
+	RTREE_GET_SUBTREE(11)
+#endif
+#if RTREE_HEIGHT_MAX > 13
+	RTREE_GET_SUBTREE(12)
+#endif
+#if RTREE_HEIGHT_MAX > 14
+	RTREE_GET_SUBTREE(13)
+#endif
+#if RTREE_HEIGHT_MAX > 15
+	RTREE_GET_SUBTREE(14)
+#endif
+#if RTREE_HEIGHT_MAX > 16
+#  error Unsupported RTREE_HEIGHT_MAX
+#endif
+	RTREE_GET_LEAF(RTREE_HEIGHT_MAX-1)
+#undef RTREE_GET_SUBTREE
+#undef RTREE_GET_LEAF
+	default: not_reached();
 	}
+#undef RTREE_GET_BIAS
 	not_reached();
 }
 
@@ -268,7 +340,7 @@ rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val)
 
 	start_level = rtree_start_level(rtree, key);
 
-	node = rtree_subtree_read(rtree, start_level);
+	node = rtree_subtree_read(rtree, start_level, false);
 	if (node == NULL)
 		return (true);
 	for (i = start_level; /**/; i++, node = child) {
@@ -282,7 +354,7 @@ rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val)
 			return (false);
 		}
 		assert(i + 1 < rtree->height);
-		child = rtree_child_read(rtree, &node[subkey], i);
+		child = rtree_child_read(rtree, &node[subkey], i, false);
 		if (child == NULL)
 			return (true);
 	}
diff --git a/include/jemalloc/internal/stats.h b/include/jemalloc/internal/stats.h
index 705903a..b621817 100644
--- a/include/jemalloc/internal/stats.h
+++ b/include/jemalloc/internal/stats.h
@@ -103,6 +103,14 @@ struct arena_stats_s {
 	size_t		mapped;
 
 	/*
+	 * Number of bytes currently retained as a side effect of munmap() being
+	 * disabled/bypassed.  Retained bytes are technically mapped (though
+	 * always decommitted or purged), but they are excluded from the mapped
+	 * statistic (above).
+	 */
+	size_t		retained;
+
+	/*
 	 * Total number of purge sweeps, total number of madvise calls made,
 	 * and total pages purged in order to keep dirty unused memory under
 	 * control.
diff --git a/include/jemalloc/internal/tcache.h b/include/jemalloc/internal/tcache.h
index cee91e3..165dd1b 100644
--- a/include/jemalloc/internal/tcache.h
+++ b/include/jemalloc/internal/tcache.h
@@ -142,27 +142,25 @@ extern size_t	tcache_maxclass;
  */
 extern tcaches_t	*tcaches;
 
-size_t	tcache_salloc(const void *ptr);
+size_t	tcache_salloc(tsdn_t *tsdn, const void *ptr);
 void	tcache_event_hard(tsd_t *tsd, tcache_t *tcache);
-void	*tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
+void	*tcache_alloc_small_hard(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
     tcache_bin_t *tbin, szind_t binind, bool *tcache_success);
 void	tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
     szind_t binind, unsigned rem);
 void	tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
     unsigned rem, tcache_t *tcache);
-void	tcache_arena_associate(tcache_t *tcache, arena_t *arena);
-void	tcache_arena_reassociate(tcache_t *tcache, arena_t *oldarena,
-    arena_t *newarena);
-void	tcache_arena_dissociate(tcache_t *tcache, arena_t *arena);
+void	tcache_arena_reassociate(tsdn_t *tsdn, tcache_t *tcache,
+    arena_t *oldarena, arena_t *newarena);
 tcache_t *tcache_get_hard(tsd_t *tsd);
-tcache_t *tcache_create(tsd_t *tsd, arena_t *arena);
+tcache_t *tcache_create(tsdn_t *tsdn, arena_t *arena);
 void	tcache_cleanup(tsd_t *tsd);
 void	tcache_enabled_cleanup(tsd_t *tsd);
-void	tcache_stats_merge(tcache_t *tcache, arena_t *arena);
-bool	tcaches_create(tsd_t *tsd, unsigned *r_ind);
+void	tcache_stats_merge(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena);
+bool	tcaches_create(tsdn_t *tsdn, unsigned *r_ind);
 void	tcaches_flush(tsd_t *tsd, unsigned ind);
 void	tcaches_destroy(tsd_t *tsd, unsigned ind);
-bool	tcache_boot(void);
+bool	tcache_boot(tsdn_t *tsdn);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
@@ -309,8 +307,8 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		if (unlikely(arena == NULL))
 			return (NULL);
 
-		ret = tcache_alloc_small_hard(tsd, arena, tcache, tbin, binind,
-			&tcache_hard_success);
+		ret = tcache_alloc_small_hard(tsd_tsdn(tsd), arena, tcache,
+		    tbin, binind, &tcache_hard_success);
 		if (tcache_hard_success == false)
 			return (NULL);
 	}
@@ -322,7 +320,7 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 	 */
 	if (config_prof || (slow_path && config_fill) || unlikely(zero)) {
 		usize = index2size(binind);
-		assert(tcache_salloc(ret) == usize);
+		assert(tcache_salloc(tsd_tsdn(tsd), ret) == usize);
 	}
 
 	if (likely(!zero)) {
@@ -370,7 +368,7 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		if (unlikely(arena == NULL))
 			return (NULL);
 
-		ret = arena_malloc_large(tsd, arena, binind, zero);
+		ret = arena_malloc_large(tsd_tsdn(tsd), arena, binind, zero);
 		if (ret == NULL)
 			return (NULL);
 	} else {
@@ -393,9 +391,10 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		}
 		if (likely(!zero)) {
 			if (slow_path && config_fill) {
-				if (unlikely(opt_junk_alloc))
-					memset(ret, 0xa5, usize);
-				else if (unlikely(opt_zero))
+				if (unlikely(opt_junk_alloc)) {
+					memset(ret, JEMALLOC_ALLOC_JUNK,
+					    usize);
+				} else if (unlikely(opt_zero))
 					memset(ret, 0, usize);
 			}
 		} else
@@ -418,7 +417,7 @@ tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
 
-	assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) <= SMALL_MAXCLASS);
 
 	if (slow_path && config_fill && unlikely(opt_junk_free))
 		arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
@@ -445,8 +444,8 @@ tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size,
 	tcache_bin_info_t *tbin_info;
 
 	assert((size & PAGE_MASK) == 0);
-	assert(tcache_salloc(ptr) > SMALL_MAXCLASS);
-	assert(tcache_salloc(ptr) <= tcache_maxclass);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) > SMALL_MAXCLASS);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) <= tcache_maxclass);
 
 	binind = size2index(size);
 
@@ -470,8 +469,10 @@ JEMALLOC_ALWAYS_INLINE tcache_t *
 tcaches_get(tsd_t *tsd, unsigned ind)
 {
 	tcaches_t *elm = &tcaches[ind];
-	if (unlikely(elm->tcache == NULL))
-		elm->tcache = tcache_create(tsd, arena_choose(tsd, NULL));
+	if (unlikely(elm->tcache == NULL)) {
+		elm->tcache = tcache_create(tsd_tsdn(tsd), arena_choose(tsd,
+		    NULL));
+	}
 	return (elm->tcache);
 }
 #endif
diff --git a/include/jemalloc/internal/tsd.h b/include/jemalloc/internal/tsd.h
index 16cc2f1..bf11341 100644
--- a/include/jemalloc/internal/tsd.h
+++ b/include/jemalloc/internal/tsd.h
@@ -13,6 +13,9 @@ typedef struct tsd_init_head_s tsd_init_head_t;
 #endif
 
 typedef struct tsd_s tsd_t;
+typedef struct tsdn_s tsdn_t;
+
+#define	TSDN_NULL	((tsdn_t *)0)
 
 typedef enum {
 	tsd_state_uninitialized,
@@ -44,6 +47,7 @@ typedef enum {
  * The result is a set of generated functions, e.g.:
  *
  *   bool example_tsd_boot(void) {...}
+ *   bool example_tsd_booted_get(void) {...}
  *   example_t *example_tsd_get() {...}
  *   void example_tsd_set(example_t *val) {...}
  *
@@ -98,6 +102,8 @@ a_attr void								\
 a_name##tsd_boot1(void);						\
 a_attr bool								\
 a_name##tsd_boot(void);							\
+a_attr bool								\
+a_name##tsd_booted_get(void);						\
 a_attr a_type *								\
 a_name##tsd_get(void);							\
 a_attr void								\
@@ -201,6 +207,12 @@ a_name##tsd_boot(void)							\
 									\
 	return (a_name##tsd_boot0());					\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -246,6 +258,12 @@ a_name##tsd_boot(void)							\
 									\
 	return (a_name##tsd_boot0());					\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -368,6 +386,12 @@ a_name##tsd_boot(void)							\
 	a_name##tsd_boot1();						\
 	return (false);							\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -490,6 +514,12 @@ a_name##tsd_boot(void)							\
 	a_name##tsd_boot1();						\
 	return (false);							\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -536,12 +566,15 @@ struct tsd_init_head_s {
     O(thread_allocated,		uint64_t)				\
     O(thread_deallocated,	uint64_t)				\
     O(prof_tdata,		prof_tdata_t *)				\
+    O(iarena,			arena_t *)				\
     O(arena,			arena_t *)				\
     O(arenas_tdata,		arena_tdata_t *)			\
     O(narenas_tdata,		unsigned)				\
     O(arenas_tdata_bypass,	bool)					\
     O(tcache_enabled,		tcache_enabled_t)			\
     O(quarantine,		quarantine_t *)				\
+    O(witnesses,		witness_list_t)				\
+    O(witness_fork,		bool)					\
 
 #define	TSD_INITIALIZER {						\
     tsd_state_uninitialized,						\
@@ -551,10 +584,13 @@ struct tsd_init_head_s {
     NULL,								\
     NULL,								\
     NULL,								\
+    NULL,								\
     0,									\
     false,								\
     tcache_enabled_default,						\
-    NULL								\
+    NULL,								\
+    ql_head_initializer(witnesses),					\
+    false								\
 }
 
 struct tsd_s {
@@ -565,6 +601,15 @@ MALLOC_TSD
 #undef O
 };
 
+/*
+ * Wrapper around tsd_t that makes it possible to avoid implicit conversion
+ * between tsd_t and tsdn_t, where tsdn_t is "nullable" and has to be
+ * explicitly converted to tsd_t, which is non-nullable.
+ */
+struct tsdn_s {
+	tsd_t	tsd;
+};
+
 static const tsd_t tsd_initializer = TSD_INITIALIZER;
 
 malloc_tsd_types(, tsd_t)
@@ -577,7 +622,7 @@ void	*malloc_tsd_malloc(size_t size);
 void	malloc_tsd_dalloc(void *wrapper);
 void	malloc_tsd_no_cleanup(void *arg);
 void	malloc_tsd_cleanup_register(bool (*f)(void));
-bool	malloc_tsd_boot0(void);
+tsd_t	*malloc_tsd_boot0(void);
 void	malloc_tsd_boot1(void);
 #if (!defined(JEMALLOC_MALLOC_THREAD_CLEANUP) && !defined(JEMALLOC_TLS) && \
     !defined(_WIN32))
@@ -595,6 +640,7 @@ void	tsd_cleanup(void *arg);
 malloc_tsd_protos(JEMALLOC_ATTR(unused), , tsd_t)
 
 tsd_t	*tsd_fetch(void);
+tsdn_t	*tsd_tsdn(tsd_t *tsd);
 bool	tsd_nominal(tsd_t *tsd);
 #define	O(n, t)								\
 t	*tsd_##n##p_get(tsd_t *tsd);					\
@@ -602,6 +648,9 @@ t	tsd_##n##_get(tsd_t *tsd);					\
 void	tsd_##n##_set(tsd_t *tsd, t n);
 MALLOC_TSD
 #undef O
+tsdn_t	*tsdn_fetch(void);
+bool	tsdn_null(const tsdn_t *tsdn);
+tsd_t	*tsdn_tsd(tsdn_t *tsdn);
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TSD_C_))
@@ -628,6 +677,13 @@ tsd_fetch(void)
 	return (tsd);
 }
 
+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsd_tsdn(tsd_t *tsd)
+{
+
+	return ((tsdn_t *)tsd);
+}
+
 JEMALLOC_INLINE bool
 tsd_nominal(tsd_t *tsd)
 {
@@ -659,6 +715,32 @@ tsd_##n##_set(tsd_t *tsd, t n)						\
 }
 MALLOC_TSD
 #undef O
+
+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsdn_fetch(void)
+{
+
+	if (!tsd_booted_get())
+		return (NULL);
+
+	return (tsd_tsdn(tsd_fetch()));
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsdn_null(const tsdn_t *tsdn)
+{
+
+	return (tsdn == NULL);
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsdn_tsd(tsdn_t *tsdn)
+{
+
+	assert(!tsdn_null(tsdn));
+
+	return (&tsdn->tsd);
+}
 #endif
 
 #endif /* JEMALLOC_H_INLINES */
diff --git a/include/jemalloc/internal/util.h b/include/jemalloc/internal/util.h
index b8885bf..a0c2203 100644
--- a/include/jemalloc/internal/util.h
+++ b/include/jemalloc/internal/util.h
@@ -40,6 +40,10 @@
  */
 #define	MALLOC_PRINTF_BUFSIZE	4096
 
+/* Junk fill patterns. */
+#define	JEMALLOC_ALLOC_JUNK	((uint8_t)0xa5)
+#define	JEMALLOC_FREE_JUNK	((uint8_t)0x5a)
+
 /*
  * Wrap a cpp argument that contains commas such that it isn't broken up into
  * multiple arguments.
@@ -73,12 +77,12 @@
       JEMALLOC_CLANG_HAS_BUILTIN(__builtin_unreachable)
 #	define unreachable() __builtin_unreachable()
 #  else
-#	define unreachable()
+#	define unreachable() abort()
 #  endif
 #else
 #	define likely(x)   !!(x)
 #	define unlikely(x) !!(x)
-#	define unreachable()
+#	define unreachable() abort()
 #endif
 
 #include "jemalloc/internal/assert.h"
@@ -106,9 +110,9 @@ void	malloc_write(const char *s);
  * malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating
  * point math.
  */
-int	malloc_vsnprintf(char *str, size_t size, const char *format,
+size_t	malloc_vsnprintf(char *str, size_t size, const char *format,
     va_list ap);
-int	malloc_snprintf(char *str, size_t size, const char *format, ...)
+size_t	malloc_snprintf(char *str, size_t size, const char *format, ...)
     JEMALLOC_FORMAT_PRINTF(3, 4);
 void	malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque,
     const char *format, va_list ap);
diff --git a/include/jemalloc/internal/valgrind.h b/include/jemalloc/internal/valgrind.h
index a3380df..1a86808 100644
--- a/include/jemalloc/internal/valgrind.h
+++ b/include/jemalloc/internal/valgrind.h
@@ -30,15 +30,17 @@
  * calls must be embedded in macros rather than in functions so that when
  * Valgrind reports errors, there are no extra stack frames in the backtraces.
  */
-#define	JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do {		\
-	if (unlikely(in_valgrind && cond))				\
-		VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(ptr), zero);	\
+#define	JEMALLOC_VALGRIND_MALLOC(cond, tsdn, ptr, usize, zero) do {	\
+	if (unlikely(in_valgrind && cond)) {				\
+		VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(tsdn, ptr),	\
+		    zero);						\
+	}								\
 } while (0)
-#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, ptr, usize,		\
+#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, tsdn, ptr, usize,	\
     ptr_maybe_null, old_ptr, old_usize, old_rzsize, old_ptr_maybe_null,	\
     zero) do {								\
 	if (unlikely(in_valgrind)) {					\
-		size_t rzsize = p2rz(ptr);				\
+		size_t rzsize = p2rz(tsdn, ptr);			\
 									\
 		if (!maybe_moved || ptr == old_ptr) {			\
 			VALGRIND_RESIZEINPLACE_BLOCK(ptr, old_usize,	\
@@ -81,8 +83,8 @@
 #define	JEMALLOC_VALGRIND_MAKE_MEM_NOACCESS(ptr, usize) do {} while (0)
 #define	JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ptr, usize) do {} while (0)
 #define	JEMALLOC_VALGRIND_MAKE_MEM_DEFINED(ptr, usize) do {} while (0)
-#define	JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do {} while (0)
-#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, ptr, usize,		\
+#define	JEMALLOC_VALGRIND_MALLOC(cond, tsdn, ptr, usize, zero) do {} while (0)
+#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, tsdn, ptr, usize,	\
     ptr_maybe_null, old_ptr, old_usize, old_rzsize, old_ptr_maybe_null,	\
     zero) do {} while (0)
 #define	JEMALLOC_VALGRIND_FREE(ptr, rzsize) do {} while (0)
diff --git a/include/jemalloc/internal/witness.h b/include/jemalloc/internal/witness.h
new file mode 100644
index 0000000..d78dca2
--- /dev/null
+++ b/include/jemalloc/internal/witness.h
@@ -0,0 +1,249 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct witness_s witness_t;
+typedef unsigned witness_rank_t;
+typedef ql_head(witness_t) witness_list_t;
+typedef int witness_comp_t (const witness_t *, const witness_t *);
+
+/*
+ * Lock ranks.  Witnesses with rank WITNESS_RANK_OMIT are completely ignored by
+ * the witness machinery.
+ */
+#define	WITNESS_RANK_OMIT		0U
+
+#define	WITNESS_RANK_INIT		1U
+#define	WITNESS_RANK_CTL		1U
+#define	WITNESS_RANK_ARENAS		2U
+
+#define	WITNESS_RANK_PROF_DUMP		3U
+#define	WITNESS_RANK_PROF_BT2GCTX	4U
+#define	WITNESS_RANK_PROF_TDATAS	5U
+#define	WITNESS_RANK_PROF_TDATA		6U
+#define	WITNESS_RANK_PROF_GCTX		7U
+
+#define	WITNESS_RANK_ARENA		8U
+#define	WITNESS_RANK_ARENA_CHUNKS	9U
+#define	WITNESS_RANK_ARENA_NODE_CACHE	10
+
+#define	WITNESS_RANK_BASE		11U
+
+#define	WITNESS_RANK_LEAF		0xffffffffU
+#define	WITNESS_RANK_ARENA_BIN		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_ARENA_HUGE		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_DSS		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_ACTIVE	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_DUMP_SEQ	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_GDUMP		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_NEXT_THR_UID	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_THREAD_ACTIVE_INIT	WITNESS_RANK_LEAF
+
+#define	WITNESS_INITIALIZER(rank) {"initializer", rank, NULL, {NULL, NULL}}
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct witness_s {
+	/* Name, used for printing lock order reversal messages. */
+	const char		*name;
+
+	/*
+	 * Witness rank, where 0 is lowest and UINT_MAX is highest.  Witnesses
+	 * must be acquired in order of increasing rank.
+	 */
+	witness_rank_t		rank;
+
+	/*
+	 * If two witnesses are of equal rank and they have the samp comp
+	 * function pointer, it is called as a last attempt to differentiate
+	 * between witnesses of equal rank.
+	 */
+	witness_comp_t		*comp;
+
+	/* Linkage for thread's currently owned locks. */
+	ql_elm(witness_t)	link;
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+void	witness_init(witness_t *witness, const char *name, witness_rank_t rank,
+    witness_comp_t *comp);
+#ifdef JEMALLOC_JET
+typedef void (witness_lock_error_t)(const witness_list_t *, const witness_t *);
+extern witness_lock_error_t *witness_lock_error;
+#else
+void	witness_lock_error(const witness_list_t *witnesses,
+    const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_owner_error_t)(const witness_t *);
+extern witness_owner_error_t *witness_owner_error;
+#else
+void	witness_owner_error(const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_not_owner_error_t)(const witness_t *);
+extern witness_not_owner_error_t *witness_not_owner_error;
+#else
+void	witness_not_owner_error(const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_lockless_error_t)(const witness_list_t *);
+extern witness_lockless_error_t *witness_lockless_error;
+#else
+void	witness_lockless_error(const witness_list_t *witnesses);
+#endif
+
+void	witnesses_cleanup(tsd_t *tsd);
+void	witness_fork_cleanup(tsd_t *tsd);
+void	witness_prefork(tsd_t *tsd);
+void	witness_postfork_parent(tsd_t *tsd);
+void	witness_postfork_child(tsd_t *tsd);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+void	witness_assert_owner(tsdn_t *tsdn, const witness_t *witness);
+void	witness_assert_not_owner(tsdn_t *tsdn, const witness_t *witness);
+void	witness_assert_lockless(tsdn_t *tsdn);
+void	witness_lock(tsdn_t *tsdn, witness_t *witness);
+void	witness_unlock(tsdn_t *tsdn, witness_t *witness);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MUTEX_C_))
+JEMALLOC_INLINE void
+witness_assert_owner(tsdn_t *tsdn, const witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_foreach(w, witnesses, link) {
+		if (w == witness)
+			return;
+	}
+	witness_owner_error(witness);
+}
+
+JEMALLOC_INLINE void
+witness_assert_not_owner(tsdn_t *tsdn, const witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_foreach(w, witnesses, link) {
+		if (w == witness)
+			witness_not_owner_error(witness);
+	}
+}
+
+JEMALLOC_INLINE void
+witness_assert_lockless(tsdn_t *tsdn)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	w = ql_last(witnesses, link);
+	if (w != NULL)
+		witness_lockless_error(witnesses);
+}
+
+JEMALLOC_INLINE void
+witness_lock(tsdn_t *tsdn, witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witness_assert_not_owner(tsdn, witness);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	w = ql_last(witnesses, link);
+	if (w == NULL) {
+		/* No other locks; do nothing. */
+	} else if (tsd_witness_fork_get(tsd) && w->rank <= witness->rank) {
+		/* Forking, and relaxed ranking satisfied. */
+	} else if (w->rank > witness->rank) {
+		/* Not forking, rank order reversal. */
+		witness_lock_error(witnesses, witness);
+	} else if (w->rank == witness->rank && (w->comp == NULL || w->comp !=
+	    witness->comp || w->comp(w, witness) > 0)) {
+		/*
+		 * Missing/incompatible comparison function, or comparison
+		 * function indicates rank order reversal.
+		 */
+		witness_lock_error(witnesses, witness);
+	}
+
+	ql_elm_new(witness, link);
+	ql_tail_insert(witnesses, witness, link);
+}
+
+JEMALLOC_INLINE void
+witness_unlock(tsdn_t *tsdn, witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witness_assert_owner(tsdn, witness);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_remove(witnesses, witness, link);
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
diff --git a/include/jemalloc/jemalloc_macros.h.in b/include/jemalloc/jemalloc_macros.h.in
index 9f356f9..129240e 100644
--- a/include/jemalloc/jemalloc_macros.h.in
+++ b/include/jemalloc/jemalloc_macros.h.in
@@ -13,11 +13,11 @@
 
 #  define MALLOCX_LG_ALIGN(la)	((int)(la))
 #  if LG_SIZEOF_PTR == 2
-#    define MALLOCX_ALIGN(a)	((int)(ffs(a)-1))
+#    define MALLOCX_ALIGN(a)	((int)(ffs((int)(a))-1))
 #  else
 #    define MALLOCX_ALIGN(a)						\
-       ((int)(((a) < (size_t)INT_MAX) ? ffs((int)(a))-1 :		\
-       ffs((int)((a)>>32))+31))
+       ((int)(((size_t)(a) < (size_t)INT_MAX) ? ffs((int)(a))-1 :	\
+       ffs((int)(((size_t)(a))>>32))+31))
 #  endif
 #  define MALLOCX_ZERO	((int)0x40)
 /*
@@ -29,7 +29,7 @@
 /*
  * Bias arena index bits so that 0 encodes "use an automatically chosen arena".
  */
-#  define MALLOCX_ARENA(a)	((int)(((a)+1) << 20))
+#  define MALLOCX_ARENA(a)	((((int)(a))+1) << 20)
 
 #if defined(__cplusplus) && defined(JEMALLOC_USE_CXX_THROW)
 #  define JEMALLOC_CXX_THROW throw()