Dep/Jemalloc: Update to Jemalloc 5.2.0

(cherry picked from commit 5d986f0be46bf439caf227f3c228ac1c8efddfde)

21 files changed, 3628 insertions, 914 deletions

diff --git a/dep/jemalloc/src/arena.c b/dep/jemalloc/src/arena.c
index 5d55bf1a060..60eac232a93 100644
--- a/dep/jemalloc/src/arena.c
+++ b/dep/jemalloc/src/arena.c
@@ -8,9 +8,10 @@
 #include "jemalloc/internal/extent_mmap.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/rtree.h"
-#include "jemalloc/internal/size_classes.h"
 #include "jemalloc/internal/util.h"
 
+JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+
 /******************************************************************************/
 /* Data. */
 
@@ -40,7 +41,11 @@ const uint64_t h_steps[SMOOTHSTEP_NSTEPS] = {
 #undef STEP
 };
 
-static div_info_t arena_binind_div_info[NBINS];
+static div_info_t arena_binind_div_info[SC_NBINS];
+
+size_t opt_oversize_threshold = OVERSIZE_THRESHOLD_DEFAULT;
+size_t oversize_threshold = OVERSIZE_THRESHOLD_DEFAULT;
+static unsigned huge_arena_ind;
 
 /******************************************************************************/
 /*
@@ -61,7 +66,7 @@ static void arena_bin_lower_slab(tsdn_t *tsdn, arena_t *arena, extent_t *slab,
 /******************************************************************************/
 
 void
-arena_basic_stats_merge(UNUSED tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
+arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
     const char **dss, ssize_t *dirty_decay_ms, ssize_t *muzzy_decay_ms,
     size_t *nactive, size_t *ndirty, size_t *nmuzzy) {
 	*nthreads += arena_nthreads_get(arena, false);
@@ -77,7 +82,8 @@ void
 arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
     const char **dss, ssize_t *dirty_decay_ms, ssize_t *muzzy_decay_ms,
     size_t *nactive, size_t *ndirty, size_t *nmuzzy, arena_stats_t *astats,
-    bin_stats_t *bstats, arena_stats_large_t *lstats) {
+    bin_stats_t *bstats, arena_stats_large_t *lstats,
+    arena_stats_extents_t *estats) {
 	cassert(config_stats);
 
 	arena_basic_stats_merge(tsdn, arena, nthreads, dss, dirty_decay_ms,
@@ -94,6 +100,10 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
 	arena_stats_accum_zu(&astats->retained,
 	    extents_npages_get(&arena->extents_retained) << LG_PAGE);
 
+	atomic_store_zu(&astats->extent_avail,
+	    atomic_load_zu(&arena->extent_avail_cnt, ATOMIC_RELAXED),
+	    ATOMIC_RELAXED);
+
 	arena_stats_accum_u64(&astats->decay_dirty.npurge,
 	    arena_stats_read_u64(tsdn, &arena->stats,
 	    &arena->stats.decay_dirty.npurge));
@@ -122,7 +132,7 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
 	    extents_npages_get(&arena->extents_dirty) +
 	    extents_npages_get(&arena->extents_muzzy)) << LG_PAGE)));
 
-	for (szind_t i = 0; i < NSIZES - NBINS; i++) {
+	for (szind_t i = 0; i < SC_NSIZES - SC_NBINS; i++) {
 		uint64_t nmalloc = arena_stats_read_u64(tsdn, &arena->stats,
 		    &arena->stats.lstats[i].nmalloc);
 		arena_stats_accum_u64(&lstats[i].nmalloc, nmalloc);
@@ -145,7 +155,29 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
 		size_t curlextents = (size_t)(nmalloc - ndalloc);
 		lstats[i].curlextents += curlextents;
 		arena_stats_accum_zu(&astats->allocated_large,
-		    curlextents * sz_index2size(NBINS + i));
+		    curlextents * sz_index2size(SC_NBINS + i));
+	}
+
+	for (pszind_t i = 0; i < SC_NPSIZES; i++) {
+		size_t dirty, muzzy, retained, dirty_bytes, muzzy_bytes,
+		    retained_bytes;
+		dirty = extents_nextents_get(&arena->extents_dirty, i);
+		muzzy = extents_nextents_get(&arena->extents_muzzy, i);
+		retained = extents_nextents_get(&arena->extents_retained, i);
+		dirty_bytes = extents_nbytes_get(&arena->extents_dirty, i);
+		muzzy_bytes = extents_nbytes_get(&arena->extents_muzzy, i);
+		retained_bytes =
+		    extents_nbytes_get(&arena->extents_retained, i);
+
+		atomic_store_zu(&estats[i].ndirty, dirty, ATOMIC_RELAXED);
+		atomic_store_zu(&estats[i].nmuzzy, muzzy, ATOMIC_RELAXED);
+		atomic_store_zu(&estats[i].nretained, retained, ATOMIC_RELAXED);
+		atomic_store_zu(&estats[i].dirty_bytes, dirty_bytes,
+		    ATOMIC_RELAXED);
+		atomic_store_zu(&estats[i].muzzy_bytes, muzzy_bytes,
+		    ATOMIC_RELAXED);
+		atomic_store_zu(&estats[i].retained_bytes, retained_bytes,
+		    ATOMIC_RELAXED);
 	}
 
 	arena_stats_unlock(tsdn, &arena->stats);
@@ -156,7 +188,7 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
 	cache_bin_array_descriptor_t *descriptor;
 	ql_foreach(descriptor, &arena->cache_bin_array_descriptor_ql, link) {
 		szind_t i = 0;
-		for (; i < NBINS; i++) {
+		for (; i < SC_NBINS; i++) {
 			cache_bin_t *tbin = &descriptor->bins_small[i];
 			arena_stats_accum_zu(&astats->tcache_bytes,
 			    tbin->ncached * sz_index2size(i));
@@ -200,8 +232,11 @@ arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
 	nstime_update(&astats->uptime);
 	nstime_subtract(&astats->uptime, &arena->create_time);
 
-	for (szind_t i = 0; i < NBINS; i++) {
-		bin_stats_merge(tsdn, &bstats[i], &arena->bins[i]);
+	for (szind_t i = 0; i < SC_NBINS; i++) {
+		for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+			bin_stats_merge(tsdn, &bstats[i],
+			    &arena->bins[i].bin_shards[j]);
+		}
 	}
 }
 
@@ -236,6 +271,54 @@ arena_slab_reg_alloc(extent_t *slab, const bin_info_t *bin_info) {
 	return ret;
 }
 
+static void
+arena_slab_reg_alloc_batch(extent_t *slab, const bin_info_t *bin_info,
+			   unsigned cnt, void** ptrs) {
+	arena_slab_data_t *slab_data = extent_slab_data_get(slab);
+
+	assert(extent_nfree_get(slab) >= cnt);
+	assert(!bitmap_full(slab_data->bitmap, &bin_info->bitmap_info));
+
+#if (! defined JEMALLOC_INTERNAL_POPCOUNTL) || (defined BITMAP_USE_TREE)
+	for (unsigned i = 0; i < cnt; i++) {
+		size_t regind = bitmap_sfu(slab_data->bitmap,
+					   &bin_info->bitmap_info);
+		*(ptrs + i) = (void *)((uintptr_t)extent_addr_get(slab) +
+		    (uintptr_t)(bin_info->reg_size * regind));
+	}
+#else
+	unsigned group = 0;
+	bitmap_t g = slab_data->bitmap[group];
+	unsigned i = 0;
+	while (i < cnt) {
+		while (g == 0) {
+			g = slab_data->bitmap[++group];
+		}
+		size_t shift = group << LG_BITMAP_GROUP_NBITS;
+		size_t pop = popcount_lu(g);
+		if (pop > (cnt - i)) {
+			pop = cnt - i;
+		}
+
+		/*
+		 * Load from memory locations only once, outside the
+		 * hot loop below.
+		 */
+		uintptr_t base = (uintptr_t)extent_addr_get(slab);
+		uintptr_t regsize = (uintptr_t)bin_info->reg_size;
+		while (pop--) {
+			size_t bit = cfs_lu(&g);
+			size_t regind = shift + bit;
+			*(ptrs + i) = (void *)(base + regsize * regind);
+
+			i++;
+		}
+		slab_data->bitmap[group] = g;
+	}
+#endif
+	extent_nfree_sub(slab, cnt);
+}
+
 #ifndef JEMALLOC_JET
 static
 #endif
@@ -291,11 +374,11 @@ arena_large_malloc_stats_update(tsdn_t *tsdn, arena_t *arena, size_t usize) {
 
 	cassert(config_stats);
 
-	if (usize < LARGE_MINCLASS) {
-		usize = LARGE_MINCLASS;
+	if (usize < SC_LARGE_MINCLASS) {
+		usize = SC_LARGE_MINCLASS;
 	}
 	index = sz_size2index(usize);
-	hindex = (index >= NBINS) ? index - NBINS : 0;
+	hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0;
 
 	arena_stats_add_u64(tsdn, &arena->stats,
 	    &arena->stats.lstats[hindex].nmalloc, 1);
@@ -307,11 +390,11 @@ arena_large_dalloc_stats_update(tsdn_t *tsdn, arena_t *arena, size_t usize) {
 
 	cassert(config_stats);
 
-	if (usize < LARGE_MINCLASS) {
-		usize = LARGE_MINCLASS;
+	if (usize < SC_LARGE_MINCLASS) {
+		usize = SC_LARGE_MINCLASS;
 	}
 	index = sz_size2index(usize);
-	hindex = (index >= NBINS) ? index - NBINS : 0;
+	hindex = (index >= SC_NBINS) ? index - SC_NBINS : 0;
 
 	arena_stats_add_u64(tsdn, &arena->stats,
 	    &arena->stats.lstats[hindex].ndalloc, 1);
@@ -324,6 +407,11 @@ arena_large_ralloc_stats_update(tsdn_t *tsdn, arena_t *arena, size_t oldusize,
 	arena_large_malloc_stats_update(tsdn, arena, usize);
 }
 
+static bool
+arena_may_have_muzzy(arena_t *arena) {
+	return (pages_can_purge_lazy && (arena_muzzy_decay_ms_get(arena) != 0));
+}
+
 extent_t *
 arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena, size_t usize,
     size_t alignment, bool *zero) {
@@ -338,7 +426,7 @@ arena_extent_alloc_large(tsdn_t *tsdn, arena_t *arena, size_t usize,
 	extent_t *extent = extents_alloc(tsdn, arena, &extent_hooks,
 	    &arena->extents_dirty, NULL, usize, sz_large_pad, alignment, false,
 	    szind, zero, &commit);
-	if (extent == NULL) {
+	if (extent == NULL && arena_may_have_muzzy(arena)) {
 		extent = extents_alloc(tsdn, arena, &extent_hooks,
 		    &arena->extents_muzzy, NULL, usize, sz_large_pad, alignment,
 		    false, szind, zero, &commit);
@@ -743,7 +831,7 @@ static size_t
 arena_decay_stashed(tsdn_t *tsdn, arena_t *arena,
     extent_hooks_t **r_extent_hooks, arena_decay_t *decay, extents_t *extents,
     bool all, extent_list_t *decay_extents, bool is_background_thread) {
-	UNUSED size_t nmadvise, nunmapped;
+	size_t nmadvise, nunmapped;
 	size_t npurged;
 
 	if (config_stats) {
@@ -834,7 +922,7 @@ arena_decay_to_limit(tsdn_t *tsdn, arena_t *arena, arena_decay_t *decay,
 	size_t npurge = arena_stash_decayed(tsdn, arena, &extent_hooks, extents,
 	    npages_limit, npages_decay_max, &decay_extents);
 	if (npurge != 0) {
-		UNUSED size_t npurged = arena_decay_stashed(tsdn, arena,
+		size_t npurged = arena_decay_stashed(tsdn, arena,
 		    &extent_hooks, decay, extents, all, &decay_extents,
 		    is_background_thread);
 		assert(npurged == npurge);
@@ -863,7 +951,7 @@ arena_decay_impl(tsdn_t *tsdn, arena_t *arena, arena_decay_t *decay,
 
 	bool epoch_advanced = arena_maybe_decay(tsdn, arena, decay, extents,
 	    is_background_thread);
-	UNUSED size_t npages_new;
+	size_t npages_new;
 	if (epoch_advanced) {
 		/* Backlog is updated on epoch advance. */
 		npages_new = decay->backlog[SMOOTHSTEP_NSTEPS-1];
@@ -954,6 +1042,37 @@ arena_bin_slabs_full_remove(arena_t *arena, bin_t *bin, extent_t *slab) {
 	extent_list_remove(&bin->slabs_full, slab);
 }
 
+static void
+arena_bin_reset(tsd_t *tsd, arena_t *arena, bin_t *bin) {
+	extent_t *slab;
+
+	malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+	if (bin->slabcur != NULL) {
+		slab = bin->slabcur;
+		bin->slabcur = NULL;
+		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
+		arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
+		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+	}
+	while ((slab = extent_heap_remove_first(&bin->slabs_nonfull)) != NULL) {
+		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
+		arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
+		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+	}
+	for (slab = extent_list_first(&bin->slabs_full); slab != NULL;
+	     slab = extent_list_first(&bin->slabs_full)) {
+		arena_bin_slabs_full_remove(arena, bin, slab);
+		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
+		arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
+		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+	}
+	if (config_stats) {
+		bin->stats.curregs = 0;
+		bin->stats.curslabs = 0;
+	}
+	malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
+}
+
 void
 arena_reset(tsd_t *tsd, arena_t *arena) {
 	/*
@@ -983,7 +1102,7 @@ arena_reset(tsd_t *tsd, arena_t *arena) {
 		rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
 		rtree_szind_slab_read(tsd_tsdn(tsd), &extents_rtree, rtree_ctx,
 		    (uintptr_t)ptr, true, &alloc_ctx.szind, &alloc_ctx.slab);
-		assert(alloc_ctx.szind != NSIZES);
+		assert(alloc_ctx.szind != SC_NSIZES);
 
 		if (config_stats || (config_prof && opt_prof)) {
 			usize = sz_index2size(alloc_ctx.szind);
@@ -999,35 +1118,11 @@ arena_reset(tsd_t *tsd, arena_t *arena) {
 	malloc_mutex_unlock(tsd_tsdn(tsd), &arena->large_mtx);
 
 	/* Bins. */
-	for (unsigned i = 0; i < NBINS; i++) {
-		extent_t *slab;
-		bin_t *bin = &arena->bins[i];
-		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
-		if (bin->slabcur != NULL) {
-			slab = bin->slabcur;
-			bin->slabcur = NULL;
-			malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
-			arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
-			malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
-		}
-		while ((slab = extent_heap_remove_first(&bin->slabs_nonfull)) !=
-		    NULL) {
-			malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
-			arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
-			malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
-		}
-		for (slab = extent_list_first(&bin->slabs_full); slab != NULL;
-		    slab = extent_list_first(&bin->slabs_full)) {
-			arena_bin_slabs_full_remove(arena, bin, slab);
-			malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
-			arena_slab_dalloc(tsd_tsdn(tsd), arena, slab);
-			malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+	for (unsigned i = 0; i < SC_NBINS; i++) {
+		for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+			arena_bin_reset(tsd, arena,
+			    &arena->bins[i].bin_shards[j]);
 		}
-		if (config_stats) {
-			bin->stats.curregs = 0;
-			bin->stats.curslabs = 0;
-		}
-		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
 	}
 
 	atomic_store_zu(&arena->nactive, 0, ATOMIC_RELAXED);
@@ -1112,7 +1207,7 @@ arena_slab_alloc_hard(tsdn_t *tsdn, arena_t *arena,
 }
 
 static extent_t *
-arena_slab_alloc(tsdn_t *tsdn, arena_t *arena, szind_t binind,
+arena_slab_alloc(tsdn_t *tsdn, arena_t *arena, szind_t binind, unsigned binshard,
     const bin_info_t *bin_info) {
 	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
 	    WITNESS_RANK_CORE, 0);
@@ -1124,7 +1219,7 @@ arena_slab_alloc(tsdn_t *tsdn, arena_t *arena, szind_t binind,
 	extent_t *slab = extents_alloc(tsdn, arena, &extent_hooks,
 	    &arena->extents_dirty, NULL, bin_info->slab_size, 0, PAGE, true,
 	    binind, &zero, &commit);
-	if (slab == NULL) {
+	if (slab == NULL && arena_may_have_muzzy(arena)) {
 		slab = extents_alloc(tsdn, arena, &extent_hooks,
 		    &arena->extents_muzzy, NULL, bin_info->slab_size, 0, PAGE,
 		    true, binind, &zero, &commit);
@@ -1140,7 +1235,7 @@ arena_slab_alloc(tsdn_t *tsdn, arena_t *arena, szind_t binind,
 
 	/* Initialize slab internals. */
 	arena_slab_data_t *slab_data = extent_slab_data_get(slab);
-	extent_nfree_set(slab, bin_info->nregs);
+	extent_nfree_binshard_set(slab, bin_info->nregs, binshard);
 	bitmap_init(slab_data->bitmap, &bin_info->bitmap_info, false);
 
 	arena_nactive_add(arena, extent_size_get(slab) >> LG_PAGE);
@@ -1150,7 +1245,7 @@ arena_slab_alloc(tsdn_t *tsdn, arena_t *arena, szind_t binind,
 
 static extent_t *
 arena_bin_nonfull_slab_get(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
-    szind_t binind) {
+    szind_t binind, unsigned binshard) {
 	extent_t *slab;
 	const bin_info_t *bin_info;
 
@@ -1166,7 +1261,7 @@ arena_bin_nonfull_slab_get(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
 	/* Allocate a new slab. */
 	malloc_mutex_unlock(tsdn, &bin->lock);
 	/******************************/
-	slab = arena_slab_alloc(tsdn, arena, binind, bin_info);
+	slab = arena_slab_alloc(tsdn, arena, binind, binshard, bin_info);
 	/********************************/
 	malloc_mutex_lock(tsdn, &bin->lock);
 	if (slab != NULL) {
@@ -1193,7 +1288,7 @@ arena_bin_nonfull_slab_get(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
 /* Re-fill bin->slabcur, then call arena_slab_reg_alloc(). */
 static void *
 arena_bin_malloc_hard(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
-    szind_t binind) {
+    szind_t binind, unsigned binshard) {
 	const bin_info_t *bin_info;
 	extent_t *slab;
 
@@ -1202,7 +1297,7 @@ arena_bin_malloc_hard(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
 		arena_bin_slabs_full_insert(arena, bin, bin->slabcur);
 		bin->slabcur = NULL;
 	}
-	slab = arena_bin_nonfull_slab_get(tsdn, arena, bin, binind);
+	slab = arena_bin_nonfull_slab_get(tsdn, arena, bin, binind, binshard);
 	if (bin->slabcur != NULL) {
 		/*
 		 * Another thread updated slabcur while this one ran without the
@@ -1246,46 +1341,75 @@ arena_bin_malloc_hard(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
 	return arena_slab_reg_alloc(slab, bin_info);
 }
 
+/* Choose a bin shard and return the locked bin. */
+bin_t *
+arena_bin_choose_lock(tsdn_t *tsdn, arena_t *arena, szind_t binind,
+    unsigned *binshard) {
+	bin_t *bin;
+	if (tsdn_null(tsdn) || tsd_arena_get(tsdn_tsd(tsdn)) == NULL) {
+		*binshard = 0;
+	} else {
+		*binshard = tsd_binshardsp_get(tsdn_tsd(tsdn))->binshard[binind];
+	}
+	assert(*binshard < bin_infos[binind].n_shards);
+	bin = &arena->bins[binind].bin_shards[*binshard];
+	malloc_mutex_lock(tsdn, &bin->lock);
+
+	return bin;
+}
+
 void
 arena_tcache_fill_small(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
     cache_bin_t *tbin, szind_t binind, uint64_t prof_accumbytes) {
-	unsigned i, nfill;
-	bin_t *bin;
+	unsigned i, nfill, cnt;
 
 	assert(tbin->ncached == 0);
 
 	if (config_prof && arena_prof_accum(tsdn, arena, prof_accumbytes)) {
 		prof_idump(tsdn);
 	}
-	bin = &arena->bins[binind];
-	malloc_mutex_lock(tsdn, &bin->lock);
+
+	unsigned binshard;
+	bin_t *bin = arena_bin_choose_lock(tsdn, arena, binind, &binshard);
+
 	for (i = 0, nfill = (tcache_bin_info[binind].ncached_max >>
-	    tcache->lg_fill_div[binind]); i < nfill; i++) {
+	    tcache->lg_fill_div[binind]); i < nfill; i += cnt) {
 		extent_t *slab;
-		void *ptr;
 		if ((slab = bin->slabcur) != NULL && extent_nfree_get(slab) >
 		    0) {
-			ptr = arena_slab_reg_alloc(slab, &bin_infos[binind]);
+			unsigned tofill = nfill - i;
+			cnt = tofill < extent_nfree_get(slab) ?
+				tofill : extent_nfree_get(slab);
+			arena_slab_reg_alloc_batch(
+			   slab, &bin_infos[binind], cnt,
+			   tbin->avail - nfill + i);
 		} else {
-			ptr = arena_bin_malloc_hard(tsdn, arena, bin, binind);
-		}
-		if (ptr == NULL) {
+			cnt = 1;
+			void *ptr = arena_bin_malloc_hard(tsdn, arena, bin,
+			    binind, binshard);
 			/*
 			 * OOM.  tbin->avail isn't yet filled down to its first
 			 * element, so the successful allocations (if any) must
 			 * be moved just before tbin->avail before bailing out.
 			 */
-			if (i > 0) {
-				memmove(tbin->avail - i, tbin->avail - nfill,
-				    i * sizeof(void *));
+			if (ptr == NULL) {
+				if (i > 0) {
+					memmove(tbin->avail - i,
+						tbin->avail - nfill,
+						i * sizeof(void *));
+				}
+				break;
 			}
-			break;
+			/* Insert such that low regions get used first. */
+			*(tbin->avail - nfill + i) = ptr;
 		}
 		if (config_fill && unlikely(opt_junk_alloc)) {
-			arena_alloc_junk_small(ptr, &bin_infos[binind], true);
+			for (unsigned j = 0; j < cnt; j++) {
+				void* ptr = *(tbin->avail - nfill + i + j);
+				arena_alloc_junk_small(ptr, &bin_infos[binind],
+							true);
+			}
 		}
-		/* Insert such that low regions get used first. */
-		*(tbin->avail - nfill + i) = ptr;
 	}
 	if (config_stats) {
 		bin->stats.nmalloc += i;
@@ -1320,15 +1444,15 @@ arena_malloc_small(tsdn_t *tsdn, arena_t *arena, szind_t binind, bool zero) {
 	size_t usize;
 	extent_t *slab;
 
-	assert(binind < NBINS);
-	bin = &arena->bins[binind];
+	assert(binind < SC_NBINS);
 	usize = sz_index2size(binind);
+	unsigned binshard;
+	bin = arena_bin_choose_lock(tsdn, arena, binind, &binshard);
 
-	malloc_mutex_lock(tsdn, &bin->lock);
 	if ((slab = bin->slabcur) != NULL && extent_nfree_get(slab) > 0) {
 		ret = arena_slab_reg_alloc(slab, &bin_infos[binind]);
 	} else {
-		ret = arena_bin_malloc_hard(tsdn, arena, bin, binind);
+		ret = arena_bin_malloc_hard(tsdn, arena, bin, binind, binshard);
 	}
 
 	if (ret == NULL) {
@@ -1373,13 +1497,13 @@ arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind,
 	assert(!tsdn_null(tsdn) || arena != NULL);
 
 	if (likely(!tsdn_null(tsdn))) {
-		arena = arena_choose(tsdn_tsd(tsdn), arena);
+		arena = arena_choose_maybe_huge(tsdn_tsd(tsdn), arena, size);
 	}
 	if (unlikely(arena == NULL)) {
 		return NULL;
 	}
 
-	if (likely(size <= SMALL_MAXCLASS)) {
+	if (likely(size <= SC_SMALL_MAXCLASS)) {
 		return arena_malloc_small(tsdn, arena, ind, zero);
 	}
 	return large_malloc(tsdn, arena, sz_index2size(ind), zero);
@@ -1390,8 +1514,9 @@ arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
     bool zero, tcache_t *tcache) {
 	void *ret;
 
-	if (usize <= SMALL_MAXCLASS && (alignment < PAGE || (alignment == PAGE
-	    && (usize & PAGE_MASK) == 0))) {
+	if (usize <= SC_SMALL_MAXCLASS
+	    && (alignment < PAGE
+	    || (alignment == PAGE && (usize & PAGE_MASK) == 0))) {
 		/* Small; alignment doesn't require special slab placement. */
 		ret = arena_malloc(tsdn, arena, usize, sz_size2index(usize),
 		    zero, tcache, true);
@@ -1409,8 +1534,8 @@ void
 arena_prof_promote(tsdn_t *tsdn, const void *ptr, size_t usize) {
 	cassert(config_prof);
 	assert(ptr != NULL);
-	assert(isalloc(tsdn, ptr) == LARGE_MINCLASS);
-	assert(usize <= SMALL_MAXCLASS);
+	assert(isalloc(tsdn, ptr) == SC_LARGE_MINCLASS);
+	assert(usize <= SC_SMALL_MAXCLASS);
 
 	rtree_ctx_t rtree_ctx_fallback;
 	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
@@ -1434,15 +1559,15 @@ arena_prof_demote(tsdn_t *tsdn, extent_t *extent, const void *ptr) {
 	cassert(config_prof);
 	assert(ptr != NULL);
 
-	extent_szind_set(extent, NBINS);
+	extent_szind_set(extent, SC_NBINS);
 	rtree_ctx_t rtree_ctx_fallback;
 	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
 	rtree_szind_slab_update(tsdn, &extents_rtree, rtree_ctx, (uintptr_t)ptr,
-	    NBINS, false);
+	    SC_NBINS, false);
 
-	assert(isalloc(tsdn, ptr) == LARGE_MINCLASS);
+	assert(isalloc(tsdn, ptr) == SC_LARGE_MINCLASS);
 
-	return LARGE_MINCLASS;
+	return SC_LARGE_MINCLASS;
 }
 
 void
@@ -1499,7 +1624,7 @@ arena_dalloc_bin_slab(tsdn_t *tsdn, arena_t *arena, extent_t *slab,
 }
 
 static void
-arena_bin_lower_slab(UNUSED tsdn_t *tsdn, arena_t *arena, extent_t *slab,
+arena_bin_lower_slab(tsdn_t *tsdn, arena_t *arena, extent_t *slab,
     bin_t *bin) {
 	assert(extent_nfree_get(slab) > 0);
 
@@ -1526,11 +1651,9 @@ arena_bin_lower_slab(UNUSED tsdn_t *tsdn, arena_t *arena, extent_t *slab,
 }
 
 static void
-arena_dalloc_bin_locked_impl(tsdn_t *tsdn, arena_t *arena, extent_t *slab,
-    void *ptr, bool junked) {
+arena_dalloc_bin_locked_impl(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
+    szind_t binind, extent_t *slab, void *ptr, bool junked) {
 	arena_slab_data_t *slab_data = extent_slab_data_get(slab);
-	szind_t binind = extent_szind_get(slab);
-	bin_t *bin = &arena->bins[binind];
 	const bin_info_t *bin_info = &bin_infos[binind];
 
 	if (!junked && config_fill && unlikely(opt_junk_free)) {
@@ -1554,18 +1677,21 @@ arena_dalloc_bin_locked_impl(tsdn_t *tsdn, arena_t *arena, extent_t *slab,
 }
 
 void
-arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena, extent_t *extent,
-    void *ptr) {
-	arena_dalloc_bin_locked_impl(tsdn, arena, extent, ptr, true);
+arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena, bin_t *bin,
+    szind_t binind, extent_t *extent, void *ptr) {
+	arena_dalloc_bin_locked_impl(tsdn, arena, bin, binind, extent, ptr,
+	    true);
 }
 
 static void
 arena_dalloc_bin(tsdn_t *tsdn, arena_t *arena, extent_t *extent, void *ptr) {
 	szind_t binind = extent_szind_get(extent);
-	bin_t *bin = &arena->bins[binind];
+	unsigned binshard = extent_binshard_get(extent);
+	bin_t *bin = &arena->bins[binind].bin_shards[binshard];
 
 	malloc_mutex_lock(tsdn, &bin->lock);
-	arena_dalloc_bin_locked_impl(tsdn, arena, extent, ptr, false);
+	arena_dalloc_bin_locked_impl(tsdn, arena, bin, binind, extent, ptr,
+	    false);
 	malloc_mutex_unlock(tsdn, &bin->lock);
 }
 
@@ -1580,38 +1706,48 @@ arena_dalloc_small(tsdn_t *tsdn, void *ptr) {
 
 bool
 arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
-    size_t extra, bool zero) {
+    size_t extra, bool zero, size_t *newsize) {
+	bool ret;
 	/* Calls with non-zero extra had to clamp extra. */
-	assert(extra == 0 || size + extra <= LARGE_MAXCLASS);
+	assert(extra == 0 || size + extra <= SC_LARGE_MAXCLASS);
 
-	if (unlikely(size > LARGE_MAXCLASS)) {
-		return true;
+	extent_t *extent = iealloc(tsdn, ptr);
+	if (unlikely(size > SC_LARGE_MAXCLASS)) {
+		ret = true;
+		goto done;
 	}
 
-	extent_t *extent = iealloc(tsdn, ptr);
 	size_t usize_min = sz_s2u(size);
 	size_t usize_max = sz_s2u(size + extra);
-	if (likely(oldsize <= SMALL_MAXCLASS && usize_min <= SMALL_MAXCLASS)) {
+	if (likely(oldsize <= SC_SMALL_MAXCLASS && usize_min
+	    <= SC_SMALL_MAXCLASS)) {
 		/*
 		 * Avoid moving the allocation if the size class can be left the
 		 * same.
 		 */
 		assert(bin_infos[sz_size2index(oldsize)].reg_size ==
 		    oldsize);
-		if ((usize_max > SMALL_MAXCLASS || sz_size2index(usize_max) !=
-		    sz_size2index(oldsize)) && (size > oldsize || usize_max <
-		    oldsize)) {
-			return true;
+		if ((usize_max > SC_SMALL_MAXCLASS
+		    || sz_size2index(usize_max) != sz_size2index(oldsize))
+		    && (size > oldsize || usize_max < oldsize)) {
+			ret = true;
+			goto done;
 		}
 
 		arena_decay_tick(tsdn, extent_arena_get(extent));
-		return false;
-	} else if (oldsize >= LARGE_MINCLASS && usize_max >= LARGE_MINCLASS) {
-		return large_ralloc_no_move(tsdn, extent, usize_min, usize_max,
+		ret = false;
+	} else if (oldsize >= SC_LARGE_MINCLASS
+	    && usize_max >= SC_LARGE_MINCLASS) {
+		ret = large_ralloc_no_move(tsdn, extent, usize_min, usize_max,
 		    zero);
+	} else {
+		ret = true;
 	}
+done:
+	assert(extent == iealloc(tsdn, ptr));
+	*newsize = extent_usize_get(extent);
 
-	return true;
+	return ret;
 }
 
 static void *
@@ -1622,7 +1758,7 @@ arena_ralloc_move_helper(tsdn_t *tsdn, arena_t *arena, size_t usize,
 		    zero, tcache, true);
 	}
 	usize = sz_sa2u(usize, alignment);
-	if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+	if (unlikely(usize == 0 || usize > SC_LARGE_MAXCLASS)) {
 		return NULL;
 	}
 	return ipalloct(tsdn, usize, alignment, zero, tcache, arena);
@@ -1630,22 +1766,30 @@ arena_ralloc_move_helper(tsdn_t *tsdn, arena_t *arena, size_t usize,
 
 void *
 arena_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t oldsize,
-    size_t size, size_t alignment, bool zero, tcache_t *tcache) {
+    size_t size, size_t alignment, bool zero, tcache_t *tcache,
+    hook_ralloc_args_t *hook_args) {
 	size_t usize = sz_s2u(size);
-	if (unlikely(usize == 0 || size > LARGE_MAXCLASS)) {
+	if (unlikely(usize == 0 || size > SC_LARGE_MAXCLASS)) {
 		return NULL;
 	}
 
-	if (likely(usize <= SMALL_MAXCLASS)) {
+	if (likely(usize <= SC_SMALL_MAXCLASS)) {
 		/* Try to avoid moving the allocation. */
-		if (!arena_ralloc_no_move(tsdn, ptr, oldsize, usize, 0, zero)) {
+		UNUSED size_t newsize;
+		if (!arena_ralloc_no_move(tsdn, ptr, oldsize, usize, 0, zero,
+		    &newsize)) {
+			hook_invoke_expand(hook_args->is_realloc
+			    ? hook_expand_realloc : hook_expand_rallocx,
+			    ptr, oldsize, usize, (uintptr_t)ptr,
+			    hook_args->args);
 			return ptr;
 		}
 	}
 
-	if (oldsize >= LARGE_MINCLASS && usize >= LARGE_MINCLASS) {
-		return large_ralloc(tsdn, arena, iealloc(tsdn, ptr), usize,
-		    alignment, zero, tcache);
+	if (oldsize >= SC_LARGE_MINCLASS
+	    && usize >= SC_LARGE_MINCLASS) {
+		return large_ralloc(tsdn, arena, ptr, usize,
+		    alignment, zero, tcache, hook_args);
 	}
 
 	/*
@@ -1658,11 +1802,16 @@ arena_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t oldsize,
 		return NULL;
 	}
 
+	hook_invoke_alloc(hook_args->is_realloc
+	    ? hook_alloc_realloc : hook_alloc_rallocx, ret, (uintptr_t)ret,
+	    hook_args->args);
+	hook_invoke_dalloc(hook_args->is_realloc
+	    ? hook_dalloc_realloc : hook_dalloc_rallocx, ptr, hook_args->args);
+
 	/*
 	 * Junk/zero-filling were already done by
 	 * ipalloc()/arena_malloc().
 	 */
-
 	size_t copysize = (usize < oldsize) ? usize : oldsize;
 	memcpy(ret, ptr, copysize);
 	isdalloct(tsdn, ptr, oldsize, tcache, NULL, true);
@@ -1720,8 +1869,7 @@ arena_retain_grow_limit_get_set(tsd_t *tsd, arena_t *arena, size_t *old_limit,
 	if (new_limit != NULL) {
 		size_t limit = *new_limit;
 		/* Grow no more than the new limit. */
-		if ((new_ind = sz_psz2ind(limit + 1) - 1) >
-		     EXTENT_GROW_MAX_PIND) {
+		if ((new_ind = sz_psz2ind(limit + 1) - 1) >= SC_NPSIZES) {
 			return true;
 		}
 	}
@@ -1773,7 +1921,12 @@ arena_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 		}
 	}
 
-	arena = (arena_t *)base_alloc(tsdn, base, sizeof(arena_t), CACHELINE);
+	unsigned nbins_total = 0;
+	for (i = 0; i < SC_NBINS; i++) {
+		nbins_total += bin_infos[i].n_shards;
+	}
+	size_t arena_size = sizeof(arena_t) + sizeof(bin_t) * nbins_total;
+	arena = (arena_t *)base_alloc(tsdn, base, arena_size, CACHELINE);
 	if (arena == NULL) {
 		goto label_error;
 	}
@@ -1865,7 +2018,7 @@ arena_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 	}
 
 	arena->extent_grow_next = sz_psz2ind(HUGEPAGE);
-	arena->retain_grow_limit = EXTENT_GROW_MAX_PIND;
+	arena->retain_grow_limit = sz_psz2ind(SC_LARGE_MAXCLASS);
 	if (malloc_mutex_init(&arena->extent_grow_mtx, "extent_grow",
 	    WITNESS_RANK_EXTENT_GROW, malloc_mutex_rank_exclusive)) {
 		goto label_error;
@@ -1878,12 +2031,20 @@ arena_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 	}
 
 	/* Initialize bins. */
-	for (i = 0; i < NBINS; i++) {
-		bool err = bin_init(&arena->bins[i]);
-		if (err) {
-			goto label_error;
+	uintptr_t bin_addr = (uintptr_t)arena + sizeof(arena_t);
+	atomic_store_u(&arena->binshard_next, 0, ATOMIC_RELEASE);
+	for (i = 0; i < SC_NBINS; i++) {
+		unsigned nshards = bin_infos[i].n_shards;
+		arena->bins[i].bin_shards = (bin_t *)bin_addr;
+		bin_addr += nshards * sizeof(bin_t);
+		for (unsigned j = 0; j < nshards; j++) {
+			bool err = bin_init(&arena->bins[i].bin_shards[j]);
+			if (err) {
+				goto label_error;
+			}
 		}
 	}
+	assert(bin_addr == (uintptr_t)arena + arena_size);
 
 	arena->base = base;
 	/* Set arena before creating background threads. */
@@ -1900,8 +2061,8 @@ arena_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 		 */
 		assert(!tsdn_null(tsdn));
 		pre_reentrancy(tsdn_tsd(tsdn), arena);
-		if (hooks_arena_new_hook) {
-			hooks_arena_new_hook();
+		if (test_hooks_arena_new_hook) {
+			test_hooks_arena_new_hook();
 		}
 		post_reentrancy(tsdn_tsd(tsdn));
 	}
@@ -1914,20 +2075,75 @@ label_error:
 	return NULL;
 }
 
+arena_t *
+arena_choose_huge(tsd_t *tsd) {
+	/* huge_arena_ind can be 0 during init (will use a0). */
+	if (huge_arena_ind == 0) {
+		assert(!malloc_initialized());
+	}
+
+	arena_t *huge_arena = arena_get(tsd_tsdn(tsd), huge_arena_ind, false);
+	if (huge_arena == NULL) {
+		/* Create the huge arena on demand. */
+		assert(huge_arena_ind != 0);
+		huge_arena = arena_get(tsd_tsdn(tsd), huge_arena_ind, true);
+		if (huge_arena == NULL) {
+			return NULL;
+		}
+		/*
+		 * Purge eagerly for huge allocations, because: 1) number of
+		 * huge allocations is usually small, which means ticker based
+		 * decay is not reliable; and 2) less immediate reuse is
+		 * expected for huge allocations.
+		 */
+		if (arena_dirty_decay_ms_default_get() > 0) {
+			arena_dirty_decay_ms_set(tsd_tsdn(tsd), huge_arena, 0);
+		}
+		if (arena_muzzy_decay_ms_default_get() > 0) {
+			arena_muzzy_decay_ms_set(tsd_tsdn(tsd), huge_arena, 0);
+		}
+	}
+
+	return huge_arena;
+}
+
+bool
+arena_init_huge(void) {
+	bool huge_enabled;
+
+	/* The threshold should be large size class. */
+	if (opt_oversize_threshold > SC_LARGE_MAXCLASS ||
+	    opt_oversize_threshold < SC_LARGE_MINCLASS) {
+		opt_oversize_threshold = 0;
+		oversize_threshold = SC_LARGE_MAXCLASS + PAGE;
+		huge_enabled = false;
+	} else {
+		/* Reserve the index for the huge arena. */
+		huge_arena_ind = narenas_total_get();
+		oversize_threshold = opt_oversize_threshold;
+		huge_enabled = true;
+	}
+
+	return huge_enabled;
+}
+
+bool
+arena_is_huge(unsigned arena_ind) {
+	if (huge_arena_ind == 0) {
+		return false;
+	}
+	return (arena_ind == huge_arena_ind);
+}
+
 void
-arena_boot(void) {
+arena_boot(sc_data_t *sc_data) {
 	arena_dirty_decay_ms_default_set(opt_dirty_decay_ms);
 	arena_muzzy_decay_ms_default_set(opt_muzzy_decay_ms);
-#define REGIND_bin_yes(index, reg_size) 				\
-	div_init(&arena_binind_div_info[(index)], (reg_size));
-#define REGIND_bin_no(index, reg_size)
-#define SC(index, lg_grp, lg_delta, ndelta, psz, bin, pgs,		\
-    lg_delta_lookup)							\
-	REGIND_bin_##bin(index, (1U<<lg_grp) + (ndelta << lg_delta))
-	SIZE_CLASSES
-#undef REGIND_bin_yes
-#undef REGIND_bin_no
-#undef SC
+	for (unsigned i = 0; i < SC_NBINS; i++) {
+		sc_t *sc = &sc_data->sc[i];
+		div_init(&arena_binind_div_info[i],
+		    (1U << sc->lg_base) + (sc->ndelta << sc->lg_delta));
+	}
 }
 
 void
@@ -1972,8 +2188,10 @@ arena_prefork6(tsdn_t *tsdn, arena_t *arena) {
 
 void
 arena_prefork7(tsdn_t *tsdn, arena_t *arena) {
-	for (unsigned i = 0; i < NBINS; i++) {
-		bin_prefork(tsdn, &arena->bins[i]);
+	for (unsigned i = 0; i < SC_NBINS; i++) {
+		for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+			bin_prefork(tsdn, &arena->bins[i].bin_shards[j]);
+		}
 	}
 }
 
@@ -1981,8 +2199,11 @@ void
 arena_postfork_parent(tsdn_t *tsdn, arena_t *arena) {
 	unsigned i;
 
-	for (i = 0; i < NBINS; i++) {
-		bin_postfork_parent(tsdn, &arena->bins[i]);
+	for (i = 0; i < SC_NBINS; i++) {
+		for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+			bin_postfork_parent(tsdn,
+			    &arena->bins[i].bin_shards[j]);
+		}
 	}
 	malloc_mutex_postfork_parent(tsdn, &arena->large_mtx);
 	base_postfork_parent(tsdn, arena->base);
@@ -2025,8 +2246,10 @@ arena_postfork_child(tsdn_t *tsdn, arena_t *arena) {
 		}
 	}
 
-	for (i = 0; i < NBINS; i++) {
-		bin_postfork_child(tsdn, &arena->bins[i]);
+	for (i = 0; i < SC_NBINS; i++) {
+		for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+			bin_postfork_child(tsdn, &arena->bins[i].bin_shards[j]);
+		}
 	}
 	malloc_mutex_postfork_child(tsdn, &arena->large_mtx);
 	base_postfork_child(tsdn, arena->base);
diff --git a/dep/jemalloc/src/background_thread.c b/dep/jemalloc/src/background_thread.c
index 3517a3bb8ed..5ed6c1c959b 100644
--- a/dep/jemalloc/src/background_thread.c
+++ b/dep/jemalloc/src/background_thread.c
@@ -4,6 +4,8 @@
 
 #include "jemalloc/internal/assert.h"
 
+JEMALLOC_DIAGNOSTIC_DISABLE_SPURIOUS
+
 /******************************************************************************/
 /* Data. */
 
@@ -11,7 +13,7 @@
 #define BACKGROUND_THREAD_DEFAULT false
 /* Read-only after initialization. */
 bool opt_background_thread = BACKGROUND_THREAD_DEFAULT;
-size_t opt_max_background_threads = MAX_BACKGROUND_THREAD_LIMIT;
+size_t opt_max_background_threads = MAX_BACKGROUND_THREAD_LIMIT + 1;
 
 /* Used for thread creation, termination and stats. */
 malloc_mutex_t background_thread_lock;
@@ -22,13 +24,9 @@ size_t max_background_threads;
 /* Thread info per-index. */
 background_thread_info_t *background_thread_info;
 
-/* False if no necessary runtime support. */
-bool can_enable_background_thread;
-
 /******************************************************************************/
 
 #ifdef JEMALLOC_PTHREAD_CREATE_WRAPPER
-#include <dlfcn.h>
 
 static int (*pthread_create_fptr)(pthread_t *__restrict, const pthread_attr_t *,
     void *(*)(void *), void *__restrict);
@@ -81,7 +79,7 @@ background_thread_info_init(tsdn_t *tsdn, background_thread_info_t *info) {
 }
 
 static inline bool
-set_current_thread_affinity(UNUSED int cpu) {
+set_current_thread_affinity(int cpu) {
 #if defined(JEMALLOC_HAVE_SCHED_SETAFFINITY)
 	cpu_set_t cpuset;
 	CPU_ZERO(&cpuset);
@@ -510,6 +508,8 @@ background_thread_entry(void *ind_arg) {
 	assert(thread_ind < max_background_threads);
 #ifdef JEMALLOC_HAVE_PTHREAD_SETNAME_NP
 	pthread_setname_np(pthread_self(), "jemalloc_bg_thd");
+#elif defined(__FreeBSD__)
+	pthread_set_name_np(pthread_self(), "jemalloc_bg_thd");
 #endif
 	if (opt_percpu_arena != percpu_arena_disabled) {
 		set_current_thread_affinity((int)thread_ind);
@@ -534,9 +534,8 @@ background_thread_init(tsd_t *tsd, background_thread_info_t *info) {
 	n_background_threads++;
 }
 
-/* Create a new background thread if needed. */
-bool
-background_thread_create(tsd_t *tsd, unsigned arena_ind) {
+static bool
+background_thread_create_locked(tsd_t *tsd, unsigned arena_ind) {
 	assert(have_background_thread);
 	malloc_mutex_assert_owner(tsd_tsdn(tsd), &background_thread_lock);
 
@@ -589,6 +588,19 @@ background_thread_create(tsd_t *tsd, unsigned arena_ind) {
 	return false;
 }
 
+/* Create a new background thread if needed. */
+bool
+background_thread_create(tsd_t *tsd, unsigned arena_ind) {
+	assert(have_background_thread);
+
+	bool ret;
+	malloc_mutex_lock(tsd_tsdn(tsd), &background_thread_lock);
+	ret = background_thread_create_locked(tsd, arena_ind);
+	malloc_mutex_unlock(tsd_tsdn(tsd), &background_thread_lock);
+
+	return ret;
+}
+
 bool
 background_threads_enable(tsd_t *tsd) {
 	assert(n_background_threads == 0);
@@ -622,7 +634,7 @@ background_threads_enable(tsd_t *tsd) {
 		}
 	}
 
-	return background_thread_create(tsd, 0);
+	return background_thread_create_locked(tsd, 0);
 }
 
 bool
@@ -807,21 +819,34 @@ background_thread_stats_read(tsdn_t *tsdn, background_thread_stats_t *stats) {
 #undef BILLION
 #undef BACKGROUND_THREAD_MIN_INTERVAL_NS
 
+#ifdef JEMALLOC_HAVE_DLSYM
+#include <dlfcn.h>
+#endif
+
 static bool
 pthread_create_fptr_init(void) {
 	if (pthread_create_fptr != NULL) {
 		return false;
 	}
+	/*
+	 * Try the next symbol first, because 1) when use lazy_lock we have a
+	 * wrapper for pthread_create; and 2) application may define its own
+	 * wrapper as well (and can call malloc within the wrapper).
+	 */
+#ifdef JEMALLOC_HAVE_DLSYM
 	pthread_create_fptr = dlsym(RTLD_NEXT, "pthread_create");
+#else
+	pthread_create_fptr = NULL;
+#endif
 	if (pthread_create_fptr == NULL) {
-		can_enable_background_thread = false;
-		if (config_lazy_lock || opt_background_thread) {
+		if (config_lazy_lock) {
 			malloc_write("<jemalloc>: Error in dlsym(RTLD_NEXT, "
 			    "\"pthread_create\")\n");
 			abort();
+		} else {
+			/* Fall back to the default symbol. */
+			pthread_create_fptr = pthread_create;
 		}
-	} else {
-		can_enable_background_thread = true;
 	}
 
 	return false;
@@ -866,9 +891,8 @@ background_thread_boot1(tsdn_t *tsdn) {
 	assert(have_background_thread);
 	assert(narenas_total_get() > 0);
 
-	if (opt_max_background_threads == MAX_BACKGROUND_THREAD_LIMIT &&
-	    ncpus < MAX_BACKGROUND_THREAD_LIMIT) {
-		opt_max_background_threads = ncpus;
+	if (opt_max_background_threads > MAX_BACKGROUND_THREAD_LIMIT) {
+		opt_max_background_threads = DEFAULT_NUM_BACKGROUND_THREAD;
 	}
 	max_background_threads = opt_max_background_threads;
 
diff --git a/dep/jemalloc/src/base.c b/dep/jemalloc/src/base.c
index b0324b5d758..f3c61661a20 100644
--- a/dep/jemalloc/src/base.c
+++ b/dep/jemalloc/src/base.c
@@ -262,8 +262,8 @@ base_block_alloc(tsdn_t *tsdn, base_t *base, extent_hooks_t *extent_hooks,
 	 */
 	size_t min_block_size = HUGEPAGE_CEILING(sz_psz2u(header_size + gap_size
 	    + usize));
-	pszind_t pind_next = (*pind_last + 1 < NPSIZES) ? *pind_last + 1 :
-	    *pind_last;
+	pszind_t pind_next = (*pind_last + 1 < sz_psz2ind(SC_LARGE_MAXCLASS)) ?
+	    *pind_last + 1 : *pind_last;
 	size_t next_block_size = HUGEPAGE_CEILING(sz_pind2sz(pind_next));
 	size_t block_size = (min_block_size > next_block_size) ? min_block_size
 	    : next_block_size;
@@ -372,7 +372,7 @@ base_new(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 	base->extent_sn_next = extent_sn_next;
 	base->blocks = block;
 	base->auto_thp_switched = false;
-	for (szind_t i = 0; i < NSIZES; i++) {
+	for (szind_t i = 0; i < SC_NSIZES; i++) {
 		extent_heap_new(&base->avail[i]);
 	}
 	if (config_stats) {
@@ -426,7 +426,7 @@ base_alloc_impl(tsdn_t *tsdn, base_t *base, size_t size, size_t alignment,
 
 	extent_t *extent = NULL;
 	malloc_mutex_lock(tsdn, &base->mtx);
-	for (szind_t i = sz_size2index(asize); i < NSIZES; i++) {
+	for (szind_t i = sz_size2index(asize); i < SC_NSIZES; i++) {
 		extent = extent_heap_remove_first(&base->avail[i]);
 		if (extent != NULL) {
 			/* Use existing space. */
diff --git a/dep/jemalloc/src/bin.c b/dep/jemalloc/src/bin.c
index 0886bc4ea92..bca6b12c352 100644
--- a/dep/jemalloc/src/bin.c
+++ b/dep/jemalloc/src/bin.c
@@ -1,23 +1,68 @@
 #include "jemalloc/internal/jemalloc_preamble.h"
 #include "jemalloc/internal/jemalloc_internal_includes.h"
 
+#include "jemalloc/internal/assert.h"
 #include "jemalloc/internal/bin.h"
+#include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/witness.h"
 
-const bin_info_t bin_infos[NBINS] = {
-#define BIN_INFO_bin_yes(reg_size, slab_size, nregs)			\
-	{reg_size, slab_size, nregs, BITMAP_INFO_INITIALIZER(nregs)},
-#define BIN_INFO_bin_no(reg_size, slab_size, nregs)
-#define SC(index, lg_grp, lg_delta, ndelta, psz, bin, pgs,		\
-    lg_delta_lookup)							\
-	BIN_INFO_bin_##bin((1U<<lg_grp) + (ndelta<<lg_delta),		\
-	    (pgs << LG_PAGE), (pgs << LG_PAGE) / ((1U<<lg_grp) +	\
-	    (ndelta<<lg_delta)))
-	SIZE_CLASSES
-#undef BIN_INFO_bin_yes
-#undef BIN_INFO_bin_no
-#undef SC
-};
+bin_info_t bin_infos[SC_NBINS];
+
+static void
+bin_infos_init(sc_data_t *sc_data, unsigned bin_shard_sizes[SC_NBINS],
+    bin_info_t bin_infos[SC_NBINS]) {
+	for (unsigned i = 0; i < SC_NBINS; i++) {
+		bin_info_t *bin_info = &bin_infos[i];
+		sc_t *sc = &sc_data->sc[i];
+		bin_info->reg_size = ((size_t)1U << sc->lg_base)
+		    + ((size_t)sc->ndelta << sc->lg_delta);
+		bin_info->slab_size = (sc->pgs << LG_PAGE);
+		bin_info->nregs =
+		    (uint32_t)(bin_info->slab_size / bin_info->reg_size);
+		bin_info->n_shards = bin_shard_sizes[i];
+		bitmap_info_t bitmap_info = BITMAP_INFO_INITIALIZER(
+		    bin_info->nregs);
+		bin_info->bitmap_info = bitmap_info;
+	}
+}
+
+bool
+bin_update_shard_size(unsigned bin_shard_sizes[SC_NBINS], size_t start_size,
+    size_t end_size, size_t nshards) {
+	if (nshards > BIN_SHARDS_MAX || nshards == 0) {
+		return true;
+	}
+
+	if (start_size > SC_SMALL_MAXCLASS) {
+		return false;
+	}
+	if (end_size > SC_SMALL_MAXCLASS) {
+		end_size = SC_SMALL_MAXCLASS;
+	}
+
+	/* Compute the index since this may happen before sz init. */
+	szind_t ind1 = sz_size2index_compute(start_size);
+	szind_t ind2 = sz_size2index_compute(end_size);
+	for (unsigned i = ind1; i <= ind2; i++) {
+		bin_shard_sizes[i] = (unsigned)nshards;
+	}
+
+	return false;
+}
+
+void
+bin_shard_sizes_boot(unsigned bin_shard_sizes[SC_NBINS]) {
+	/* Load the default number of shards. */
+	for (unsigned i = 0; i < SC_NBINS; i++) {
+		bin_shard_sizes[i] = N_BIN_SHARDS_DEFAULT;
+	}
+}
+
+void
+bin_boot(sc_data_t *sc_data, unsigned bin_shard_sizes[SC_NBINS]) {
+	assert(sc_data->initialized);
+	bin_infos_init(sc_data, bin_shard_sizes, bin_infos);
+}
 
 bool
 bin_init(bin_t *bin) {
diff --git a/dep/jemalloc/src/ckh.c b/dep/jemalloc/src/ckh.c
index e95e0a3ed59..1bf6df5a115 100644
--- a/dep/jemalloc/src/ckh.c
+++ b/dep/jemalloc/src/ckh.c
@@ -275,7 +275,8 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh) {
 
 		lg_curcells++;
 		usize = sz_sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);
-		if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+		if (unlikely(usize == 0
+		    || usize > SC_LARGE_MAXCLASS)) {
 			ret = true;
 			goto label_return;
 		}
@@ -320,7 +321,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh) {
 	lg_prevbuckets = ckh->lg_curbuckets;
 	lg_curcells = ckh->lg_curbuckets + LG_CKH_BUCKET_CELLS - 1;
 	usize = sz_sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);
-	if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+	if (unlikely(usize == 0 || usize > SC_LARGE_MAXCLASS)) {
 		return;
 	}
 	tab = (ckhc_t *)ipallocztm(tsd_tsdn(tsd), usize, CACHELINE, true, NULL,
@@ -396,7 +397,7 @@ ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
 	ckh->keycomp = keycomp;
 
 	usize = sz_sa2u(sizeof(ckhc_t) << lg_mincells, CACHELINE);
-	if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+	if (unlikely(usize == 0 || usize > SC_LARGE_MAXCLASS)) {
 		ret = true;
 		goto label_return;
 	}
diff --git a/dep/jemalloc/src/ctl.c b/dep/jemalloc/src/ctl.c
index 1e713a3d104..09310a9d191 100644
--- a/dep/jemalloc/src/ctl.c
+++ b/dep/jemalloc/src/ctl.c
@@ -8,7 +8,7 @@
 #include "jemalloc/internal/extent_mmap.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/nstime.h"
-#include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/util.h"
 
 /******************************************************************************/
@@ -85,6 +85,7 @@ CTL_PROTO(opt_retain)
 CTL_PROTO(opt_dss)
 CTL_PROTO(opt_narenas)
 CTL_PROTO(opt_percpu_arena)
+CTL_PROTO(opt_oversize_threshold)
 CTL_PROTO(opt_background_thread)
 CTL_PROTO(opt_max_background_threads)
 CTL_PROTO(opt_dirty_decay_ms)
@@ -126,6 +127,7 @@ INDEX_PROTO(arena_i)
 CTL_PROTO(arenas_bin_i_size)
 CTL_PROTO(arenas_bin_i_nregs)
 CTL_PROTO(arenas_bin_i_slab_size)
+CTL_PROTO(arenas_bin_i_nshards)
 INDEX_PROTO(arenas_bin_i)
 CTL_PROTO(arenas_lextent_i_size)
 INDEX_PROTO(arenas_lextent_i)
@@ -147,6 +149,8 @@ CTL_PROTO(prof_gdump)
 CTL_PROTO(prof_reset)
 CTL_PROTO(prof_interval)
 CTL_PROTO(lg_prof_sample)
+CTL_PROTO(prof_log_start)
+CTL_PROTO(prof_log_stop)
 CTL_PROTO(stats_arenas_i_small_allocated)
 CTL_PROTO(stats_arenas_i_small_nmalloc)
 CTL_PROTO(stats_arenas_i_small_ndalloc)
@@ -170,6 +174,13 @@ CTL_PROTO(stats_arenas_i_lextents_j_ndalloc)
 CTL_PROTO(stats_arenas_i_lextents_j_nrequests)
 CTL_PROTO(stats_arenas_i_lextents_j_curlextents)
 INDEX_PROTO(stats_arenas_i_lextents_j)
+CTL_PROTO(stats_arenas_i_extents_j_ndirty)
+CTL_PROTO(stats_arenas_i_extents_j_nmuzzy)
+CTL_PROTO(stats_arenas_i_extents_j_nretained)
+CTL_PROTO(stats_arenas_i_extents_j_dirty_bytes)
+CTL_PROTO(stats_arenas_i_extents_j_muzzy_bytes)
+CTL_PROTO(stats_arenas_i_extents_j_retained_bytes)
+INDEX_PROTO(stats_arenas_i_extents_j)
 CTL_PROTO(stats_arenas_i_nthreads)
 CTL_PROTO(stats_arenas_i_uptime)
 CTL_PROTO(stats_arenas_i_dss)
@@ -180,6 +191,7 @@ CTL_PROTO(stats_arenas_i_pdirty)
 CTL_PROTO(stats_arenas_i_pmuzzy)
 CTL_PROTO(stats_arenas_i_mapped)
 CTL_PROTO(stats_arenas_i_retained)
+CTL_PROTO(stats_arenas_i_extent_avail)
 CTL_PROTO(stats_arenas_i_dirty_npurge)
 CTL_PROTO(stats_arenas_i_dirty_nmadvise)
 CTL_PROTO(stats_arenas_i_dirty_purged)
@@ -202,6 +214,8 @@ CTL_PROTO(stats_metadata_thp)
 CTL_PROTO(stats_resident)
 CTL_PROTO(stats_mapped)
 CTL_PROTO(stats_retained)
+CTL_PROTO(experimental_hooks_install)
+CTL_PROTO(experimental_hooks_remove)
 
 #define MUTEX_STATS_CTL_PROTO_GEN(n)					\
 CTL_PROTO(stats_##n##_num_ops)						\
@@ -286,6 +300,7 @@ static const ctl_named_node_t opt_node[] = {
 	{NAME("dss"),		CTL(opt_dss)},
 	{NAME("narenas"),	CTL(opt_narenas)},
 	{NAME("percpu_arena"),	CTL(opt_percpu_arena)},
+	{NAME("oversize_threshold"),	CTL(opt_oversize_threshold)},
 	{NAME("background_thread"),	CTL(opt_background_thread)},
 	{NAME("max_background_threads"),	CTL(opt_max_background_threads)},
 	{NAME("dirty_decay_ms"), CTL(opt_dirty_decay_ms)},
@@ -341,7 +356,8 @@ static const ctl_indexed_node_t arena_node[] = {
 static const ctl_named_node_t arenas_bin_i_node[] = {
 	{NAME("size"),		CTL(arenas_bin_i_size)},
 	{NAME("nregs"),		CTL(arenas_bin_i_nregs)},
-	{NAME("slab_size"),	CTL(arenas_bin_i_slab_size)}
+	{NAME("slab_size"),	CTL(arenas_bin_i_slab_size)},
+	{NAME("nshards"),	CTL(arenas_bin_i_nshards)}
 };
 static const ctl_named_node_t super_arenas_bin_i_node[] = {
 	{NAME(""),		CHILD(named, arenas_bin_i)}
@@ -385,9 +401,10 @@ static const ctl_named_node_t	prof_node[] = {
 	{NAME("gdump"),		CTL(prof_gdump)},
 	{NAME("reset"),		CTL(prof_reset)},
 	{NAME("interval"),	CTL(prof_interval)},
-	{NAME("lg_sample"),	CTL(lg_prof_sample)}
+	{NAME("lg_sample"),	CTL(lg_prof_sample)},
+	{NAME("log_start"),	CTL(prof_log_start)},
+	{NAME("log_stop"),	CTL(prof_log_stop)}
 };
-
 static const ctl_named_node_t stats_arenas_i_small_node[] = {
 	{NAME("allocated"),	CTL(stats_arenas_i_small_allocated)},
 	{NAME("nmalloc"),	CTL(stats_arenas_i_small_nmalloc)},
@@ -458,6 +475,23 @@ static const ctl_indexed_node_t stats_arenas_i_lextents_node[] = {
 	{INDEX(stats_arenas_i_lextents_j)}
 };
 
+static const ctl_named_node_t stats_arenas_i_extents_j_node[] = {
+	{NAME("ndirty"),	CTL(stats_arenas_i_extents_j_ndirty)},
+	{NAME("nmuzzy"),	CTL(stats_arenas_i_extents_j_nmuzzy)},
+	{NAME("nretained"),	CTL(stats_arenas_i_extents_j_nretained)},
+	{NAME("dirty_bytes"),	CTL(stats_arenas_i_extents_j_dirty_bytes)},
+	{NAME("muzzy_bytes"),	CTL(stats_arenas_i_extents_j_muzzy_bytes)},
+	{NAME("retained_bytes"), CTL(stats_arenas_i_extents_j_retained_bytes)}
+};
+
+static const ctl_named_node_t super_stats_arenas_i_extents_j_node[] = {
+	{NAME(""),		CHILD(named, stats_arenas_i_extents_j)}
+};
+
+static const ctl_indexed_node_t stats_arenas_i_extents_node[] = {
+	{INDEX(stats_arenas_i_extents_j)}
+};
+
 #define OP(mtx)  MUTEX_PROF_DATA_NODE(arenas_i_mutexes_##mtx)
 MUTEX_PROF_ARENA_MUTEXES
 #undef OP
@@ -479,6 +513,7 @@ static const ctl_named_node_t stats_arenas_i_node[] = {
 	{NAME("pmuzzy"),	CTL(stats_arenas_i_pmuzzy)},
 	{NAME("mapped"),	CTL(stats_arenas_i_mapped)},
 	{NAME("retained"),	CTL(stats_arenas_i_retained)},
+	{NAME("extent_avail"),	CTL(stats_arenas_i_extent_avail)},
 	{NAME("dirty_npurge"),	CTL(stats_arenas_i_dirty_npurge)},
 	{NAME("dirty_nmadvise"), CTL(stats_arenas_i_dirty_nmadvise)},
 	{NAME("dirty_purged"),	CTL(stats_arenas_i_dirty_purged)},
@@ -494,6 +529,7 @@ static const ctl_named_node_t stats_arenas_i_node[] = {
 	{NAME("large"),		CHILD(named, stats_arenas_i_large)},
 	{NAME("bins"),		CHILD(indexed, stats_arenas_i_bins)},
 	{NAME("lextents"),	CHILD(indexed, stats_arenas_i_lextents)},
+	{NAME("extents"),	CHILD(indexed, stats_arenas_i_extents)},
 	{NAME("mutexes"),	CHILD(named, stats_arenas_i_mutexes)}
 };
 static const ctl_named_node_t super_stats_arenas_i_node[] = {
@@ -536,6 +572,15 @@ static const ctl_named_node_t stats_node[] = {
 	{NAME("arenas"),	CHILD(indexed, stats_arenas)}
 };
 
+static const ctl_named_node_t hooks_node[] = {
+	{NAME("install"),	CTL(experimental_hooks_install)},
+	{NAME("remove"),	CTL(experimental_hooks_remove)},
+};
+
+static const ctl_named_node_t experimental_node[] = {
+	{NAME("hooks"),		CHILD(named, hooks)}
+};
+
 static const ctl_named_node_t	root_node[] = {
 	{NAME("version"),	CTL(version)},
 	{NAME("epoch"),		CTL(epoch)},
@@ -548,7 +593,8 @@ static const ctl_named_node_t	root_node[] = {
 	{NAME("arena"),		CHILD(indexed, arena)},
 	{NAME("arenas"),	CHILD(named, arenas)},
 	{NAME("prof"),		CHILD(named, prof)},
-	{NAME("stats"),		CHILD(named, stats)}
+	{NAME("stats"),		CHILD(named, stats)},
+	{NAME("experimental"),	CHILD(named, experimental)}
 };
 static const ctl_named_node_t super_root_node[] = {
 	{NAME(""),		CHILD(named, root)}
@@ -696,10 +742,12 @@ ctl_arena_clear(ctl_arena_t *ctl_arena) {
 		ctl_arena->astats->nmalloc_small = 0;
 		ctl_arena->astats->ndalloc_small = 0;
 		ctl_arena->astats->nrequests_small = 0;
-		memset(ctl_arena->astats->bstats, 0, NBINS *
+		memset(ctl_arena->astats->bstats, 0, SC_NBINS *
 		    sizeof(bin_stats_t));
-		memset(ctl_arena->astats->lstats, 0, (NSIZES - NBINS) *
+		memset(ctl_arena->astats->lstats, 0, (SC_NSIZES - SC_NBINS) *
 		    sizeof(arena_stats_large_t));
+		memset(ctl_arena->astats->estats, 0, SC_NPSIZES *
+		    sizeof(arena_stats_extents_t));
 	}
 }
 
@@ -713,9 +761,9 @@ ctl_arena_stats_amerge(tsdn_t *tsdn, ctl_arena_t *ctl_arena, arena_t *arena) {
 		    &ctl_arena->muzzy_decay_ms, &ctl_arena->pactive,
 		    &ctl_arena->pdirty, &ctl_arena->pmuzzy,
 		    &ctl_arena->astats->astats, ctl_arena->astats->bstats,
-		    ctl_arena->astats->lstats);
+		    ctl_arena->astats->lstats, ctl_arena->astats->estats);
 
-		for (i = 0; i < NBINS; i++) {
+		for (i = 0; i < SC_NBINS; i++) {
 			ctl_arena->astats->allocated_small +=
 			    ctl_arena->astats->bstats[i].curregs *
 			    sz_index2size(i);
@@ -760,6 +808,8 @@ ctl_arena_stats_sdmerge(ctl_arena_t *ctl_sdarena, ctl_arena_t *ctl_arena,
 			    &astats->astats.mapped);
 			accum_atomic_zu(&sdstats->astats.retained,
 			    &astats->astats.retained);
+			accum_atomic_zu(&sdstats->astats.extent_avail,
+			    &astats->astats.extent_avail);
 		}
 
 		ctl_accum_arena_stats_u64(&sdstats->astats.decay_dirty.npurge,
@@ -827,7 +877,8 @@ MUTEX_PROF_ARENA_MUTEXES
 			sdstats->astats.uptime = astats->astats.uptime;
 		}
 
-		for (i = 0; i < NBINS; i++) {
+		/* Merge bin stats. */
+		for (i = 0; i < SC_NBINS; i++) {
 			sdstats->bstats[i].nmalloc += astats->bstats[i].nmalloc;
 			sdstats->bstats[i].ndalloc += astats->bstats[i].ndalloc;
 			sdstats->bstats[i].nrequests +=
@@ -853,7 +904,8 @@ MUTEX_PROF_ARENA_MUTEXES
 			    &astats->bstats[i].mutex_data);
 		}
 
-		for (i = 0; i < NSIZES - NBINS; i++) {
+		/* Merge stats for large allocations. */
+		for (i = 0; i < SC_NSIZES - SC_NBINS; i++) {
 			ctl_accum_arena_stats_u64(&sdstats->lstats[i].nmalloc,
 			    &astats->lstats[i].nmalloc);
 			ctl_accum_arena_stats_u64(&sdstats->lstats[i].ndalloc,
@@ -867,6 +919,22 @@ MUTEX_PROF_ARENA_MUTEXES
 				assert(astats->lstats[i].curlextents == 0);
 			}
 		}
+
+		/* Merge extents stats. */
+		for (i = 0; i < SC_NPSIZES; i++) {
+			accum_atomic_zu(&sdstats->estats[i].ndirty,
+			    &astats->estats[i].ndirty);
+			accum_atomic_zu(&sdstats->estats[i].nmuzzy,
+			    &astats->estats[i].nmuzzy);
+			accum_atomic_zu(&sdstats->estats[i].nretained,
+			    &astats->estats[i].nretained);
+			accum_atomic_zu(&sdstats->estats[i].dirty_bytes,
+			    &astats->estats[i].dirty_bytes);
+			accum_atomic_zu(&sdstats->estats[i].muzzy_bytes,
+			    &astats->estats[i].muzzy_bytes);
+			accum_atomic_zu(&sdstats->estats[i].retained_bytes,
+			    &astats->estats[i].retained_bytes);
+		}
 	}
 }
 
@@ -1378,8 +1446,8 @@ label_return:								\
 
 #define CTL_RO_CGEN(c, n, v, t)						\
 static int								\
-n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,	\
-    size_t *oldlenp, void *newp, size_t newlen) {			\
+n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, \
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {			\
 	int ret;							\
 	t oldval;							\
 									\
@@ -1421,8 +1489,8 @@ label_return:								\
  */
 #define CTL_RO_NL_CGEN(c, n, v, t)					\
 static int								\
-n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,	\
-    size_t *oldlenp, void *newp, size_t newlen) {			\
+n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, \
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {			\
 	int ret;							\
 	t oldval;							\
 									\
@@ -1440,8 +1508,8 @@ label_return:								\
 
 #define CTL_RO_NL_GEN(n, v, t)						\
 static int								\
-n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,	\
-    size_t *oldlenp, void *newp, size_t newlen) {			\
+n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, \
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {			\
 	int ret;							\
 	t oldval;							\
 									\
@@ -1475,8 +1543,8 @@ label_return:								\
 
 #define CTL_RO_CONFIG_GEN(n, t)						\
 static int								\
-n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,	\
-    size_t *oldlenp, void *newp, size_t newlen) {			\
+n##_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, \
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {			\
 	int ret;							\
 	t oldval;							\
 									\
@@ -1494,8 +1562,8 @@ label_return:								\
 CTL_RO_NL_GEN(version, JEMALLOC_VERSION, const char *)
 
 static int
-epoch_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+epoch_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	UNUSED uint64_t newval;
 
@@ -1513,8 +1581,9 @@ label_return:
 }
 
 static int
-background_thread_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+background_thread_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp,
+    void *newp, size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -1544,13 +1613,6 @@ background_thread_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
 
 		background_thread_enabled_set(tsd_tsdn(tsd), newval);
 		if (newval) {
-			if (!can_enable_background_thread) {
-				malloc_printf("<jemalloc>: Error in dlsym("
-			            "RTLD_NEXT, \"pthread_create\"). Cannot "
-				    "enable background_thread\n");
-				ret = EFAULT;
-				goto label_return;
-			}
 			if (background_threads_enable(tsd)) {
 				ret = EFAULT;
 				goto label_return;
@@ -1571,8 +1633,9 @@ label_return:
 }
 
 static int
-max_background_threads_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+max_background_threads_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	size_t oldval;
 
@@ -1605,13 +1668,6 @@ max_background_threads_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
 		}
 
 		if (background_thread_enabled()) {
-			if (!can_enable_background_thread) {
-				malloc_printf("<jemalloc>: Error in dlsym("
-			            "RTLD_NEXT, \"pthread_create\"). Cannot "
-				    "enable background_thread\n");
-				ret = EFAULT;
-				goto label_return;
-			}
 			background_thread_enabled_set(tsd_tsdn(tsd), false);
 			if (background_threads_disable(tsd)) {
 				ret = EFAULT;
@@ -1660,6 +1716,7 @@ CTL_RO_NL_GEN(opt_dss, opt_dss, const char *)
 CTL_RO_NL_GEN(opt_narenas, opt_narenas, unsigned)
 CTL_RO_NL_GEN(opt_percpu_arena, percpu_arena_mode_names[opt_percpu_arena],
     const char *)
+CTL_RO_NL_GEN(opt_oversize_threshold, opt_oversize_threshold, size_t)
 CTL_RO_NL_GEN(opt_background_thread, opt_background_thread, bool)
 CTL_RO_NL_GEN(opt_max_background_threads, opt_max_background_threads, size_t)
 CTL_RO_NL_GEN(opt_dirty_decay_ms, opt_dirty_decay_ms, ssize_t)
@@ -1690,8 +1747,8 @@ CTL_RO_NL_CGEN(config_prof, opt_prof_leak, opt_prof_leak, bool)
 /******************************************************************************/
 
 static int
-thread_arena_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+thread_arena_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	arena_t *oldarena;
 	unsigned newind, oldind;
@@ -1755,8 +1812,9 @@ CTL_TSD_RO_NL_CGEN(config_stats, thread_deallocatedp,
     tsd_thread_deallocatedp_get, uint64_t *)
 
 static int
-thread_tcache_enabled_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+thread_tcache_enabled_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -1776,8 +1834,9 @@ label_return:
 }
 
 static int
-thread_tcache_flush_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+thread_tcache_flush_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 
 	if (!tcache_available(tsd)) {
@@ -1796,8 +1855,9 @@ label_return:
 }
 
 static int
-thread_prof_name_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+thread_prof_name_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 
 	if (!config_prof) {
@@ -1827,8 +1887,9 @@ label_return:
 }
 
 static int
-thread_prof_active_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+thread_prof_active_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -1857,8 +1918,8 @@ label_return:
 /******************************************************************************/
 
 static int
-tcache_create_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+tcache_create_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	unsigned tcache_ind;
 
@@ -1875,8 +1936,8 @@ label_return:
 }
 
 static int
-tcache_flush_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+tcache_flush_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	unsigned tcache_ind;
 
@@ -1895,8 +1956,8 @@ label_return:
 }
 
 static int
-tcache_destroy_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+tcache_destroy_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	unsigned tcache_ind;
 
@@ -2044,9 +2105,8 @@ arena_reset_prepare_background_thread(tsd_t *tsd, unsigned arena_ind) {
 	if (have_background_thread) {
 		malloc_mutex_lock(tsd_tsdn(tsd), &background_thread_lock);
 		if (background_thread_enabled()) {
-			unsigned ind = arena_ind % ncpus;
 			background_thread_info_t *info =
-			    &background_thread_info[ind];
+			    background_thread_info_get(arena_ind);
 			assert(info->state == background_thread_started);
 			malloc_mutex_lock(tsd_tsdn(tsd), &info->mtx);
 			info->state = background_thread_paused;
@@ -2059,9 +2119,8 @@ static void
 arena_reset_finish_background_thread(tsd_t *tsd, unsigned arena_ind) {
 	if (have_background_thread) {
 		if (background_thread_enabled()) {
-			unsigned ind = arena_ind % ncpus;
 			background_thread_info_t *info =
-			    &background_thread_info[ind];
+			    background_thread_info_get(arena_ind);
 			assert(info->state == background_thread_paused);
 			malloc_mutex_lock(tsd_tsdn(tsd), &info->mtx);
 			info->state = background_thread_started;
@@ -2217,6 +2276,17 @@ arena_i_decay_ms_ctl_impl(tsd_t *tsd, const size_t *mib, size_t miblen,
 			ret = EINVAL;
 			goto label_return;
 		}
+		if (arena_is_huge(arena_ind) && *(ssize_t *)newp > 0) {
+			/*
+			 * By default the huge arena purges eagerly.  If it is
+			 * set to non-zero decay time afterwards, background
+			 * thread might be needed.
+			 */
+			if (background_thread_create(tsd, arena_ind)) {
+				ret = EFAULT;
+				goto label_return;
+			}
+		}
 		if (dirty ? arena_dirty_decay_ms_set(tsd_tsdn(tsd), arena,
 		    *(ssize_t *)newp) : arena_muzzy_decay_ms_set(tsd_tsdn(tsd),
 		    arena, *(ssize_t *)newp)) {
@@ -2300,8 +2370,9 @@ label_return:
 }
 
 static int
-arena_i_retain_grow_limit_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+arena_i_retain_grow_limit_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	unsigned arena_ind;
 	arena_t *arena;
@@ -2336,7 +2407,8 @@ label_return:
 }
 
 static const ctl_named_node_t *
-arena_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, size_t i) {
+arena_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen,
+    size_t i) {
 	const ctl_named_node_t *ret;
 
 	malloc_mutex_lock(tsdn, &ctl_mtx);
@@ -2361,8 +2433,8 @@ label_return:
 /******************************************************************************/
 
 static int
-arenas_narenas_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+arenas_narenas_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	unsigned narenas;
 
@@ -2382,8 +2454,9 @@ label_return:
 }
 
 static int
-arenas_decay_ms_ctl_impl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen, bool dirty) {
+arenas_decay_ms_ctl_impl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen, bool dirty) {
 	int ret;
 
 	if (oldp != NULL && oldlenp != NULL) {
@@ -2425,34 +2498,36 @@ arenas_muzzy_decay_ms_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
 CTL_RO_NL_GEN(arenas_quantum, QUANTUM, size_t)
 CTL_RO_NL_GEN(arenas_page, PAGE, size_t)
 CTL_RO_NL_GEN(arenas_tcache_max, tcache_maxclass, size_t)
-CTL_RO_NL_GEN(arenas_nbins, NBINS, unsigned)
+CTL_RO_NL_GEN(arenas_nbins, SC_NBINS, unsigned)
 CTL_RO_NL_GEN(arenas_nhbins, nhbins, unsigned)
 CTL_RO_NL_GEN(arenas_bin_i_size, bin_infos[mib[2]].reg_size, size_t)
 CTL_RO_NL_GEN(arenas_bin_i_nregs, bin_infos[mib[2]].nregs, uint32_t)
 CTL_RO_NL_GEN(arenas_bin_i_slab_size, bin_infos[mib[2]].slab_size, size_t)
+CTL_RO_NL_GEN(arenas_bin_i_nshards, bin_infos[mib[2]].n_shards, uint32_t)
 static const ctl_named_node_t *
-arenas_bin_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, size_t i) {
-	if (i > NBINS) {
+arenas_bin_i_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t i) {
+	if (i > SC_NBINS) {
 		return NULL;
 	}
 	return super_arenas_bin_i_node;
 }
 
-CTL_RO_NL_GEN(arenas_nlextents, NSIZES - NBINS, unsigned)
-CTL_RO_NL_GEN(arenas_lextent_i_size, sz_index2size(NBINS+(szind_t)mib[2]),
+CTL_RO_NL_GEN(arenas_nlextents, SC_NSIZES - SC_NBINS, unsigned)
+CTL_RO_NL_GEN(arenas_lextent_i_size, sz_index2size(SC_NBINS+(szind_t)mib[2]),
     size_t)
 static const ctl_named_node_t *
-arenas_lextent_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen,
-    size_t i) {
-	if (i > NSIZES - NBINS) {
+arenas_lextent_i_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t i) {
+	if (i > SC_NSIZES - SC_NBINS) {
 		return NULL;
 	}
 	return super_arenas_lextent_i_node;
 }
 
 static int
-arenas_create_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+arenas_create_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	extent_hooks_t *extent_hooks;
 	unsigned arena_ind;
@@ -2474,8 +2549,9 @@ label_return:
 }
 
 static int
-arenas_lookup_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+arenas_lookup_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	unsigned arena_ind;
 	void *ptr;
@@ -2506,8 +2582,9 @@ label_return:
 /******************************************************************************/
 
 static int
-prof_thread_active_init_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+prof_thread_active_init_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp, void *newp,
+    size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -2533,8 +2610,8 @@ label_return:
 }
 
 static int
-prof_active_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+prof_active_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -2559,8 +2636,8 @@ label_return:
 }
 
 static int
-prof_dump_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+prof_dump_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	const char *filename = NULL;
 
@@ -2582,8 +2659,8 @@ label_return:
 }
 
 static int
-prof_gdump_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+prof_gdump_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	bool oldval;
 
@@ -2608,8 +2685,8 @@ label_return:
 }
 
 static int
-prof_reset_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
-    size_t *oldlenp, void *newp, size_t newlen) {
+prof_reset_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
 	int ret;
 	size_t lg_sample = lg_prof_sample;
 
@@ -2633,6 +2710,44 @@ label_return:
 CTL_RO_NL_CGEN(config_prof, prof_interval, prof_interval, uint64_t)
 CTL_RO_NL_CGEN(config_prof, lg_prof_sample, lg_prof_sample, size_t)
 
+static int
+prof_log_start_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen) {
+	int ret;
+	
+	const char *filename = NULL;
+
+	if (!config_prof) {
+		return ENOENT;
+	}
+
+	WRITEONLY();
+	WRITE(filename, const char *);
+
+	if (prof_log_start(tsd_tsdn(tsd), filename)) {
+		ret = EFAULT;
+		goto label_return; 
+	}
+
+	ret = 0;
+label_return:
+	return ret;
+}
+
+static int
+prof_log_stop_ctl(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen) {
+	if (!config_prof) {
+		return ENOENT;
+	}
+
+	if (prof_log_stop(tsd_tsdn(tsd))) {
+		return EFAULT;
+	}
+
+	return 0;
+}
+
 /******************************************************************************/
 
 CTL_RO_CGEN(config_stats, stats_allocated, ctl_stats->allocated, size_t)
@@ -2667,6 +2782,10 @@ CTL_RO_CGEN(config_stats, stats_arenas_i_mapped,
 CTL_RO_CGEN(config_stats, stats_arenas_i_retained,
     atomic_load_zu(&arenas_i(mib[2])->astats->astats.retained, ATOMIC_RELAXED),
     size_t)
+CTL_RO_CGEN(config_stats, stats_arenas_i_extent_avail,
+    atomic_load_zu(&arenas_i(mib[2])->astats->astats.extent_avail,
+        ATOMIC_RELAXED),
+    size_t)
 
 CTL_RO_CGEN(config_stats, stats_arenas_i_dirty_npurge,
     ctl_arena_stats_read_u64(
@@ -2765,8 +2884,9 @@ RO_MUTEX_CTL_GEN(arenas_i_bins_j_mutex,
 
 /* Resets all mutex stats, including global, arena and bin mutexes. */
 static int
-stats_mutexes_reset_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
-    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+stats_mutexes_reset_ctl(tsd_t *tsd, const size_t *mib,
+    size_t miblen, void *oldp, size_t *oldlenp,
+    void *newp, size_t newlen) {
 	if (!config_stats) {
 		return ENOENT;
 	}
@@ -2806,9 +2926,11 @@ stats_mutexes_reset_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
 		MUTEX_PROF_RESET(arena->tcache_ql_mtx);
 		MUTEX_PROF_RESET(arena->base->mtx);
 
-		for (szind_t i = 0; i < NBINS; i++) {
-			bin_t *bin = &arena->bins[i];
-			MUTEX_PROF_RESET(bin->lock);
+		for (szind_t i = 0; i < SC_NBINS; i++) {
+			for (unsigned j = 0; j < bin_infos[i].n_shards; j++) {
+				bin_t *bin = &arena->bins[i].bin_shards[j];
+				MUTEX_PROF_RESET(bin->lock);
+			}
 		}
 	}
 #undef MUTEX_PROF_RESET
@@ -2835,9 +2957,9 @@ CTL_RO_CGEN(config_stats, stats_arenas_i_bins_j_curslabs,
     arenas_i(mib[2])->astats->bstats[mib[4]].curslabs, size_t)
 
 static const ctl_named_node_t *
-stats_arenas_i_bins_j_index(tsdn_t *tsdn, const size_t *mib, size_t miblen,
-    size_t j) {
-	if (j > NBINS) {
+stats_arenas_i_bins_j_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t j) {
+	if (j > SC_NBINS) {
 		return NULL;
 	}
 	return super_stats_arenas_i_bins_j_node;
@@ -2856,16 +2978,51 @@ CTL_RO_CGEN(config_stats, stats_arenas_i_lextents_j_curlextents,
     arenas_i(mib[2])->astats->lstats[mib[4]].curlextents, size_t)
 
 static const ctl_named_node_t *
-stats_arenas_i_lextents_j_index(tsdn_t *tsdn, const size_t *mib, size_t miblen,
-    size_t j) {
-	if (j > NSIZES - NBINS) {
+stats_arenas_i_lextents_j_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t j) {
+	if (j > SC_NSIZES - SC_NBINS) {
 		return NULL;
 	}
 	return super_stats_arenas_i_lextents_j_node;
 }
 
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_ndirty,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].ndirty,
+	ATOMIC_RELAXED), size_t);
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_nmuzzy,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].nmuzzy,
+	ATOMIC_RELAXED), size_t);
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_nretained,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].nretained,
+	ATOMIC_RELAXED), size_t);
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_dirty_bytes,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].dirty_bytes,
+	ATOMIC_RELAXED), size_t);
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_muzzy_bytes,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].muzzy_bytes,
+	ATOMIC_RELAXED), size_t);
+CTL_RO_CGEN(config_stats, stats_arenas_i_extents_j_retained_bytes,
+    atomic_load_zu(
+        &arenas_i(mib[2])->astats->estats[mib[4]].retained_bytes,
+	ATOMIC_RELAXED), size_t);
+
 static const ctl_named_node_t *
-stats_arenas_i_index(tsdn_t *tsdn, const size_t *mib, size_t miblen, size_t i) {
+stats_arenas_i_extents_j_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t j) {
+	if (j >= SC_NPSIZES) {
+		return NULL;
+	}
+	return super_stats_arenas_i_extents_j_node;
+}
+
+static const ctl_named_node_t *
+stats_arenas_i_index(tsdn_t *tsdn, const size_t *mib,
+    size_t miblen, size_t i) {
 	const ctl_named_node_t *ret;
 	size_t a;
 
@@ -2881,3 +3038,48 @@ label_return:
 	malloc_mutex_unlock(tsdn, &ctl_mtx);
 	return ret;
 }
+
+static int
+experimental_hooks_install_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+	int ret;
+	if (oldp == NULL || oldlenp == NULL|| newp == NULL) {
+		ret = EINVAL;
+		goto label_return;
+	}
+	/*
+	 * Note: this is a *private* struct.  This is an experimental interface;
+	 * forcing the user to know the jemalloc internals well enough to
+	 * extract the ABI hopefully ensures nobody gets too comfortable with
+	 * this API, which can change at a moment's notice.
+	 */
+	hooks_t hooks;
+	WRITE(hooks, hooks_t);
+	void *handle = hook_install(tsd_tsdn(tsd), &hooks);
+	if (handle == NULL) {
+		ret = EAGAIN;
+		goto label_return;
+	}
+	READ(handle, void *);
+
+	ret = 0;
+label_return:
+	return ret;
+}
+
+static int
+experimental_hooks_remove_ctl(tsd_t *tsd, const size_t *mib, size_t miblen,
+    void *oldp, size_t *oldlenp, void *newp, size_t newlen) {
+	int ret;
+	WRITEONLY();
+	void *handle = NULL;
+	WRITE(handle, void *);
+	if (handle == NULL) {
+		ret = EINVAL;
+		goto label_return;
+	}
+	hook_remove(tsd_tsdn(tsd), handle);
+	ret = 0;
+label_return:
+	return ret;
+}
diff --git a/dep/jemalloc/src/extent.c b/dep/jemalloc/src/extent.c
index 09d6d771817..62086c7d7b5 100644
--- a/dep/jemalloc/src/extent.c
+++ b/dep/jemalloc/src/extent.c
@@ -20,7 +20,7 @@ mutex_pool_t	extent_mutex_pool;
 size_t opt_lg_extent_max_active_fit = LG_EXTENT_MAX_ACTIVE_FIT_DEFAULT;
 
 static const bitmap_info_t extents_bitmap_info =
-    BITMAP_INFO_INITIALIZER(NPSIZES+1);
+    BITMAP_INFO_INITIALIZER(SC_NPSIZES+1);
 
 static void *extent_alloc_default(extent_hooks_t *extent_hooks, void *new_addr,
     size_t size, size_t alignment, bool *zero, bool *commit,
@@ -119,9 +119,13 @@ static void extent_record(tsdn_t *tsdn, arena_t *arena,
 
 /******************************************************************************/
 
-ph_gen(UNUSED, extent_avail_, extent_tree_t, extent_t, ph_link,
+#define ATTR_NONE /* does nothing */
+
+ph_gen(ATTR_NONE, extent_avail_, extent_tree_t, extent_t, ph_link,
     extent_esnead_comp)
 
+#undef ATTR_NONE
+
 typedef enum {
 	lock_result_success,
 	lock_result_failure,
@@ -130,13 +134,16 @@ typedef enum {
 
 static lock_result_t
 extent_rtree_leaf_elm_try_lock(tsdn_t *tsdn, rtree_leaf_elm_t *elm,
-    extent_t **result) {
+    extent_t **result, bool inactive_only) {
 	extent_t *extent1 = rtree_leaf_elm_extent_read(tsdn, &extents_rtree,
 	    elm, true);
 
-	if (extent1 == NULL) {
+	/* Slab implies active extents and should be skipped. */
+	if (extent1 == NULL || (inactive_only && rtree_leaf_elm_slab_read(tsdn,
+	    &extents_rtree, elm, true))) {
 		return lock_result_no_extent;
 	}
+
 	/*
 	 * It's possible that the extent changed out from under us, and with it
 	 * the leaf->extent mapping.  We have to recheck while holding the lock.
@@ -159,7 +166,8 @@ extent_rtree_leaf_elm_try_lock(tsdn_t *tsdn, rtree_leaf_elm_t *elm,
  * address, and NULL otherwise.
  */
 static extent_t *
-extent_lock_from_addr(tsdn_t *tsdn, rtree_ctx_t *rtree_ctx, void *addr) {
+extent_lock_from_addr(tsdn_t *tsdn, rtree_ctx_t *rtree_ctx, void *addr,
+    bool inactive_only) {
 	extent_t *ret = NULL;
 	rtree_leaf_elm_t *elm = rtree_leaf_elm_lookup(tsdn, &extents_rtree,
 	    rtree_ctx, (uintptr_t)addr, false, false);
@@ -168,7 +176,8 @@ extent_lock_from_addr(tsdn_t *tsdn, rtree_ctx_t *rtree_ctx, void *addr) {
 	}
 	lock_result_t lock_result;
 	do {
-		lock_result = extent_rtree_leaf_elm_try_lock(tsdn, elm, &ret);
+		lock_result = extent_rtree_leaf_elm_try_lock(tsdn, elm, &ret,
+		    inactive_only);
 	} while (lock_result == lock_result_failure);
 	return ret;
 }
@@ -182,6 +191,7 @@ extent_alloc(tsdn_t *tsdn, arena_t *arena) {
 		return base_alloc_extent(tsdn, arena->base);
 	}
 	extent_avail_remove(&arena->extent_avail, extent);
+	atomic_fetch_sub_zu(&arena->extent_avail_cnt, 1, ATOMIC_RELAXED);
 	malloc_mutex_unlock(tsdn, &arena->extent_avail_mtx);
 	return extent;
 }
@@ -190,6 +200,7 @@ void
 extent_dalloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent) {
 	malloc_mutex_lock(tsdn, &arena->extent_avail_mtx);
 	extent_avail_insert(&arena->extent_avail, extent);
+	atomic_fetch_add_zu(&arena->extent_avail_cnt, 1, ATOMIC_RELAXED);
 	malloc_mutex_unlock(tsdn, &arena->extent_avail_mtx);
 }
 
@@ -255,7 +266,7 @@ extent_size_quantize_ceil(size_t size) {
 	size_t ret;
 
 	assert(size > 0);
-	assert(size - sz_large_pad <= LARGE_MAXCLASS);
+	assert(size - sz_large_pad <= SC_LARGE_MAXCLASS);
 	assert((size & PAGE_MASK) == 0);
 
 	ret = extent_size_quantize_floor(size);
@@ -284,7 +295,7 @@ extents_init(tsdn_t *tsdn, extents_t *extents, extent_state_t state,
 	    malloc_mutex_rank_exclusive)) {
 		return true;
 	}
-	for (unsigned i = 0; i < NPSIZES+1; i++) {
+	for (unsigned i = 0; i < SC_NPSIZES + 1; i++) {
 		extent_heap_new(&extents->heaps[i]);
 	}
 	bitmap_init(extents->bitmap, &extents_bitmap_info, true);
@@ -305,6 +316,32 @@ extents_npages_get(extents_t *extents) {
 	return atomic_load_zu(&extents->npages, ATOMIC_RELAXED);
 }
 
+size_t
+extents_nextents_get(extents_t *extents, pszind_t pind) {
+	return atomic_load_zu(&extents->nextents[pind], ATOMIC_RELAXED);
+}
+
+size_t
+extents_nbytes_get(extents_t *extents, pszind_t pind) {
+	return atomic_load_zu(&extents->nbytes[pind], ATOMIC_RELAXED);
+}
+
+static void
+extents_stats_add(extents_t *extent, pszind_t pind, size_t sz) {
+	size_t cur = atomic_load_zu(&extent->nextents[pind], ATOMIC_RELAXED);
+	atomic_store_zu(&extent->nextents[pind], cur + 1, ATOMIC_RELAXED);
+	cur = atomic_load_zu(&extent->nbytes[pind], ATOMIC_RELAXED);
+	atomic_store_zu(&extent->nbytes[pind], cur + sz, ATOMIC_RELAXED);
+}
+
+static void
+extents_stats_sub(extents_t *extent, pszind_t pind, size_t sz) {
+	size_t cur = atomic_load_zu(&extent->nextents[pind], ATOMIC_RELAXED);
+	atomic_store_zu(&extent->nextents[pind], cur - 1, ATOMIC_RELAXED);
+	cur = atomic_load_zu(&extent->nbytes[pind], ATOMIC_RELAXED);
+	atomic_store_zu(&extent->nbytes[pind], cur - sz, ATOMIC_RELAXED);
+}
+
 static void
 extents_insert_locked(tsdn_t *tsdn, extents_t *extents, extent_t *extent) {
 	malloc_mutex_assert_owner(tsdn, &extents->mtx);
@@ -318,6 +355,11 @@ extents_insert_locked(tsdn_t *tsdn, extents_t *extents, extent_t *extent) {
 		    (size_t)pind);
 	}
 	extent_heap_insert(&extents->heaps[pind], extent);
+
+	if (config_stats) {
+		extents_stats_add(extents, pind, size);
+	}
+
 	extent_list_append(&extents->lru, extent);
 	size_t npages = size >> LG_PAGE;
 	/*
@@ -340,6 +382,11 @@ extents_remove_locked(tsdn_t *tsdn, extents_t *extents, extent_t *extent) {
 	size_t psz = extent_size_quantize_floor(size);
 	pszind_t pind = sz_psz2ind(psz);
 	extent_heap_remove(&extents->heaps[pind], extent);
+
+	if (config_stats) {
+		extents_stats_sub(extents, pind, size);
+	}
+
 	if (extent_heap_empty(&extents->heaps[pind])) {
 		bitmap_set(extents->bitmap, &extents_bitmap_info,
 		    (size_t)pind);
@@ -371,7 +418,7 @@ extents_fit_alignment(extents_t *extents, size_t min_size, size_t max_size,
 	    &extents_bitmap_info, (size_t)pind); i < pind_max; i =
 	    (pszind_t)bitmap_ffu(extents->bitmap, &extents_bitmap_info,
 	    (size_t)i+1)) {
-		assert(i < NPSIZES);
+		assert(i < SC_NPSIZES);
 		assert(!extent_heap_empty(&extents->heaps[i]));
 		extent_t *extent = extent_heap_first(&extents->heaps[i]);
 		uintptr_t base = (uintptr_t)extent_base_get(extent);
@@ -401,7 +448,7 @@ extents_best_fit_locked(tsdn_t *tsdn, arena_t *arena, extents_t *extents,
 	pszind_t pind = sz_psz2ind(extent_size_quantize_ceil(size));
 	pszind_t i = (pszind_t)bitmap_ffu(extents->bitmap, &extents_bitmap_info,
 	    (size_t)pind);
-	if (i < NPSIZES+1) {
+	if (i < SC_NPSIZES + 1) {
 		/*
 		 * In order to reduce fragmentation, avoid reusing and splitting
 		 * large extents for much smaller sizes.
@@ -429,8 +476,9 @@ extents_first_fit_locked(tsdn_t *tsdn, arena_t *arena, extents_t *extents,
 
 	pszind_t pind = sz_psz2ind(extent_size_quantize_ceil(size));
 	for (pszind_t i = (pszind_t)bitmap_ffu(extents->bitmap,
-	    &extents_bitmap_info, (size_t)pind); i < NPSIZES+1; i =
-	    (pszind_t)bitmap_ffu(extents->bitmap, &extents_bitmap_info,
+	    &extents_bitmap_info, (size_t)pind);
+	    i < SC_NPSIZES + 1;
+	    i = (pszind_t)bitmap_ffu(extents->bitmap, &extents_bitmap_info,
 	    (size_t)i+1)) {
 		assert(!extent_heap_empty(&extents->heaps[i]));
 		extent_t *extent = extent_heap_first(&extents->heaps[i]);
@@ -438,10 +486,10 @@ extents_first_fit_locked(tsdn_t *tsdn, arena_t *arena, extents_t *extents,
 		if (ret == NULL || extent_snad_comp(extent, ret) < 0) {
 			ret = extent;
 		}
-		if (i == NPSIZES) {
+		if (i == SC_NPSIZES) {
 			break;
 		}
-		assert(i < NPSIZES);
+		assert(i < SC_NPSIZES);
 	}
 
 	return ret;
@@ -748,6 +796,7 @@ extent_register_impl(tsdn_t *tsdn, extent_t *extent, bool gdump_add) {
 
 	if (extent_rtree_leaf_elms_lookup(tsdn, rtree_ctx, extent, false, true,
 	    &elm_a, &elm_b)) {
+		extent_unlock(tsdn, extent);
 		return true;
 	}
 
@@ -817,7 +866,7 @@ extent_deregister_impl(tsdn_t *tsdn, extent_t *extent, bool gdump) {
 
 	extent_lock(tsdn, extent);
 
-	extent_rtree_write_acquired(tsdn, elm_a, elm_b, NULL, NSIZES, false);
+	extent_rtree_write_acquired(tsdn, elm_a, elm_b, NULL, SC_NSIZES, false);
 	if (extent_slab_get(extent)) {
 		extent_interior_deregister(tsdn, rtree_ctx, extent);
 		extent_slab_set(extent, false);
@@ -874,7 +923,8 @@ extent_recycle_extract(tsdn_t *tsdn, arena_t *arena,
 	extent_hooks_assure_initialized(arena, r_extent_hooks);
 	extent_t *extent;
 	if (new_addr != NULL) {
-		extent = extent_lock_from_addr(tsdn, rtree_ctx, new_addr);
+		extent = extent_lock_from_addr(tsdn, rtree_ctx, new_addr,
+		    false);
 		if (extent != NULL) {
 			/*
 			 * We might null-out extent to report an error, but we
@@ -958,7 +1008,7 @@ extent_split_interior(tsdn_t *tsdn, arena_t *arena,
 	if (leadsize != 0) {
 		*lead = *extent;
 		*extent = extent_split_impl(tsdn, arena, r_extent_hooks,
-		    *lead, leadsize, NSIZES, false, esize + trailsize, szind,
+		    *lead, leadsize, SC_NSIZES, false, esize + trailsize, szind,
 		    slab, growing_retained);
 		if (*extent == NULL) {
 			*to_leak = *lead;
@@ -970,7 +1020,7 @@ extent_split_interior(tsdn_t *tsdn, arena_t *arena,
 	/* Split the trail. */
 	if (trailsize != 0) {
 		*trail = extent_split_impl(tsdn, arena, r_extent_hooks, *extent,
-		    esize, szind, slab, trailsize, NSIZES, false,
+		    esize, szind, slab, trailsize, SC_NSIZES, false,
 		    growing_retained);
 		if (*trail == NULL) {
 			*to_leak = *extent;
@@ -987,7 +1037,7 @@ extent_split_interior(tsdn_t *tsdn, arena_t *arena,
 		 * splitting occurred.
 		 */
 		extent_szind_set(*extent, szind);
-		if (szind != NSIZES) {
+		if (szind != SC_NSIZES) {
 			rtree_szind_slab_update(tsdn, &extents_rtree, rtree_ctx,
 			    (uintptr_t)extent_addr_get(*extent), szind, slab);
 			if (slab && extent_size_get(*extent) > PAGE) {
@@ -1045,14 +1095,25 @@ extent_recycle_split(tsdn_t *tsdn, arena_t *arena,
 			extent_deregister_no_gdump_sub(tsdn, to_leak);
 			extents_leak(tsdn, arena, r_extent_hooks, extents,
 			    to_leak, growing_retained);
-			assert(extent_lock_from_addr(tsdn, rtree_ctx, leak)
-			    == NULL);
+			assert(extent_lock_from_addr(tsdn, rtree_ctx, leak,
+			    false) == NULL);
 		}
 		return NULL;
 	}
 	unreachable();
 }
 
+static bool
+extent_need_manual_zero(arena_t *arena) {
+	/*
+	 * Need to manually zero the extent on repopulating if either; 1) non
+	 * default extent hooks installed (in which case the purge semantics may
+	 * change); or 2) transparent huge pages enabled.
+	 */
+	return (!arena_has_default_hooks(arena) ||
+		(opt_thp == thp_mode_always));
+}
+
 /*
  * Tries to satisfy the given allocation request by reusing one of the extents
  * in the given extents_t.
@@ -1092,7 +1153,9 @@ extent_recycle(tsdn_t *tsdn, arena_t *arena, extent_hooks_t **r_extent_hooks,
 			    extent, growing_retained);
 			return NULL;
 		}
-		extent_zeroed_set(extent, true);
+		if (!extent_need_manual_zero(arena)) {
+			extent_zeroed_set(extent, true);
+		}
 	}
 
 	if (extent_committed_get(extent)) {
@@ -1113,14 +1176,16 @@ extent_recycle(tsdn_t *tsdn, arena_t *arena, extent_hooks_t **r_extent_hooks,
 
 	if (*zero) {
 		void *addr = extent_base_get(extent);
-		size_t size = extent_size_get(extent);
 		if (!extent_zeroed_get(extent)) {
-			if (pages_purge_forced(addr, size)) {
+			size_t size = extent_size_get(extent);
+			if (extent_need_manual_zero(arena) ||
+			    pages_purge_forced(addr, size)) {
 				memset(addr, 0, size);
 			}
 		} else if (config_debug) {
 			size_t *p = (size_t *)(uintptr_t)addr;
-			for (size_t i = 0; i < size / sizeof(size_t); i++) {
+			/* Check the first page only. */
+			for (size_t i = 0; i < PAGE / sizeof(size_t); i++) {
 				assert(p[i] == 0);
 			}
 		}
@@ -1244,11 +1309,11 @@ extent_grow_retained(tsdn_t *tsdn, arena_t *arena,
 	size_t alloc_size = sz_pind2sz(arena->extent_grow_next + egn_skip);
 	while (alloc_size < alloc_size_min) {
 		egn_skip++;
-		if (arena->extent_grow_next + egn_skip == NPSIZES) {
+		if (arena->extent_grow_next + egn_skip >=
+		    sz_psz2ind(SC_LARGE_MAXCLASS)) {
 			/* Outside legal range. */
 			goto label_err;
 		}
-		assert(arena->extent_grow_next + egn_skip < NPSIZES);
 		alloc_size = sz_pind2sz(arena->extent_grow_next + egn_skip);
 	}
 
@@ -1271,7 +1336,7 @@ extent_grow_retained(tsdn_t *tsdn, arena_t *arena,
 		extent_hook_post_reentrancy(tsdn);
 	}
 
-	extent_init(extent, arena, ptr, alloc_size, false, NSIZES,
+	extent_init(extent, arena, ptr, alloc_size, false, SC_NSIZES,
 	    arena_extent_sn_next(arena), extent_state_active, zeroed,
 	    committed, true);
 	if (ptr == NULL) {
@@ -1341,7 +1406,9 @@ extent_grow_retained(tsdn_t *tsdn, arena_t *arena,
 			    &arena->extents_retained, extent, true);
 			goto label_err;
 		}
-		extent_zeroed_set(extent, true);
+		if (!extent_need_manual_zero(arena)) {
+			extent_zeroed_set(extent, true);
+		}
 	}
 
 	/*
@@ -1375,7 +1442,8 @@ extent_grow_retained(tsdn_t *tsdn, arena_t *arena,
 	if (*zero && !extent_zeroed_get(extent)) {
 		void *addr = extent_base_get(extent);
 		size_t size = extent_size_get(extent);
-		if (pages_purge_forced(addr, size)) {
+		if (extent_need_manual_zero(arena) ||
+		    pages_purge_forced(addr, size)) {
 			memset(addr, 0, size);
 		}
 	}
@@ -1524,9 +1592,15 @@ extent_coalesce(tsdn_t *tsdn, arena_t *arena, extent_hooks_t **r_extent_hooks,
 }
 
 static extent_t *
-extent_try_coalesce(tsdn_t *tsdn, arena_t *arena,
+extent_try_coalesce_impl(tsdn_t *tsdn, arena_t *arena,
     extent_hooks_t **r_extent_hooks, rtree_ctx_t *rtree_ctx, extents_t *extents,
-    extent_t *extent, bool *coalesced, bool growing_retained) {
+    extent_t *extent, bool *coalesced, bool growing_retained,
+    bool inactive_only) {
+	/*
+	 * We avoid checking / locking inactive neighbors for large size
+	 * classes, since they are eagerly coalesced on deallocation which can
+	 * cause lock contention.
+	 */
 	/*
 	 * Continue attempting to coalesce until failure, to protect against
 	 * races with other threads that are thwarted by this one.
@@ -1537,7 +1611,7 @@ extent_try_coalesce(tsdn_t *tsdn, arena_t *arena,
 
 		/* Try to coalesce forward. */
 		extent_t *next = extent_lock_from_addr(tsdn, rtree_ctx,
-		    extent_past_get(extent));
+		    extent_past_get(extent), inactive_only);
 		if (next != NULL) {
 			/*
 			 * extents->mtx only protects against races for
@@ -1563,7 +1637,7 @@ extent_try_coalesce(tsdn_t *tsdn, arena_t *arena,
 
 		/* Try to coalesce backward. */
 		extent_t *prev = extent_lock_from_addr(tsdn, rtree_ctx,
-		    extent_before_get(extent));
+		    extent_before_get(extent), inactive_only);
 		if (prev != NULL) {
 			bool can_coalesce = extent_can_coalesce(arena, extents,
 			    extent, prev);
@@ -1589,6 +1663,22 @@ extent_try_coalesce(tsdn_t *tsdn, arena_t *arena,
 	return extent;
 }
 
+static extent_t *
+extent_try_coalesce(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, rtree_ctx_t *rtree_ctx, extents_t *extents,
+    extent_t *extent, bool *coalesced, bool growing_retained) {
+	return extent_try_coalesce_impl(tsdn, arena, r_extent_hooks, rtree_ctx,
+	    extents, extent, coalesced, growing_retained, false);
+}
+
+static extent_t *
+extent_try_coalesce_large(tsdn_t *tsdn, arena_t *arena,
+    extent_hooks_t **r_extent_hooks, rtree_ctx_t *rtree_ctx, extents_t *extents,
+    extent_t *extent, bool *coalesced, bool growing_retained) {
+	return extent_try_coalesce_impl(tsdn, arena, r_extent_hooks, rtree_ctx,
+	    extents, extent, coalesced, growing_retained, true);
+}
+
 /*
  * Does the metadata management portions of putting an unused extent into the
  * given extents_t (coalesces, deregisters slab interiors, the heap operations).
@@ -1606,7 +1696,7 @@ extent_record(tsdn_t *tsdn, arena_t *arena, extent_hooks_t **r_extent_hooks,
 	malloc_mutex_lock(tsdn, &extents->mtx);
 	extent_hooks_assure_initialized(arena, r_extent_hooks);
 
-	extent_szind_set(extent, NSIZES);
+	extent_szind_set(extent, SC_NSIZES);
 	if (extent_slab_get(extent)) {
 		extent_interior_deregister(tsdn, rtree_ctx, extent);
 		extent_slab_set(extent, false);
@@ -1618,18 +1708,22 @@ extent_record(tsdn_t *tsdn, arena_t *arena, extent_hooks_t **r_extent_hooks,
 	if (!extents->delay_coalesce) {
 		extent = extent_try_coalesce(tsdn, arena, r_extent_hooks,
 		    rtree_ctx, extents, extent, NULL, growing_retained);
-	} else if (extent_size_get(extent) >= LARGE_MINCLASS) {
+	} else if (extent_size_get(extent) >= SC_LARGE_MINCLASS) {
+		assert(extents == &arena->extents_dirty);
 		/* Always coalesce large extents eagerly. */
 		bool coalesced;
-		size_t prev_size;
 		do {
-			prev_size = extent_size_get(extent);
 			assert(extent_state_get(extent) == extent_state_active);
-			extent = extent_try_coalesce(tsdn, arena,
+			extent = extent_try_coalesce_large(tsdn, arena,
 			    r_extent_hooks, rtree_ctx, extents, extent,
 			    &coalesced, growing_retained);
-		} while (coalesced &&
-		    extent_size_get(extent) >= prev_size + LARGE_MINCLASS);
+		} while (coalesced);
+		if (extent_size_get(extent) >= oversize_threshold) {
+			/* Shortcut to purge the oversize extent eagerly. */
+			malloc_mutex_unlock(tsdn, &extents->mtx);
+			arena_decay_extent(tsdn, arena, r_extent_hooks, extent);
+			return;
+		}
 	}
 	extent_deactivate_locked(tsdn, arena, extents, extent);
 
@@ -1652,6 +1746,12 @@ extent_dalloc_gap(tsdn_t *tsdn, arena_t *arena, extent_t *extent) {
 }
 
 static bool
+extent_may_dalloc(void) {
+	/* With retain enabled, the default dalloc always fails. */
+	return !opt_retain;
+}
+
+static bool
 extent_dalloc_default_impl(void *addr, size_t size) {
 	if (!have_dss || !extent_in_dss(addr)) {
 		return extent_dalloc_mmap(addr, size);
@@ -1706,16 +1806,20 @@ extent_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena,
 	witness_assert_depth_to_rank(tsdn_witness_tsdp_get(tsdn),
 	    WITNESS_RANK_CORE, 0);
 
-	/*
-	 * Deregister first to avoid a race with other allocating threads, and
-	 * reregister if deallocation fails.
-	 */
-	extent_deregister(tsdn, extent);
-	if (!extent_dalloc_wrapper_try(tsdn, arena, r_extent_hooks, extent)) {
-		return;
+	/* Avoid calling the default extent_dalloc unless have to. */
+	if (*r_extent_hooks != &extent_hooks_default || extent_may_dalloc()) {
+		/*
+		 * Deregister first to avoid a race with other allocating
+		 * threads, and reregister if deallocation fails.
+		 */
+		extent_deregister(tsdn, extent);
+		if (!extent_dalloc_wrapper_try(tsdn, arena, r_extent_hooks,
+		    extent)) {
+			return;
+		}
+		extent_reregister(tsdn, extent);
 	}
 
-	extent_reregister(tsdn, extent);
 	if (*r_extent_hooks != &extent_hooks_default) {
 		extent_hook_pre_reentrancy(tsdn, arena);
 	}
@@ -2128,22 +2232,23 @@ extent_merge_impl(tsdn_t *tsdn, arena_t *arena,
 
 	if (a_elm_b != NULL) {
 		rtree_leaf_elm_write(tsdn, &extents_rtree, a_elm_b, NULL,
-		    NSIZES, false);
+		    SC_NSIZES, false);
 	}
 	if (b_elm_b != NULL) {
 		rtree_leaf_elm_write(tsdn, &extents_rtree, b_elm_a, NULL,
-		    NSIZES, false);
+		    SC_NSIZES, false);
 	} else {
 		b_elm_b = b_elm_a;
 	}
 
 	extent_size_set(a, extent_size_get(a) + extent_size_get(b));
-	extent_szind_set(a, NSIZES);
+	extent_szind_set(a, SC_NSIZES);
 	extent_sn_set(a, (extent_sn_get(a) < extent_sn_get(b)) ?
 	    extent_sn_get(a) : extent_sn_get(b));
 	extent_zeroed_set(a, extent_zeroed_get(a) && extent_zeroed_get(b));
 
-	extent_rtree_write_acquired(tsdn, a_elm_a, b_elm_b, a, NSIZES, false);
+	extent_rtree_write_acquired(tsdn, a_elm_a, b_elm_b, a, SC_NSIZES,
+	    false);
 
 	extent_unlock2(tsdn, a, b);
 
diff --git a/dep/jemalloc/src/extent_dss.c b/dep/jemalloc/src/extent_dss.c
index 2b1ea9cafa0..6c56cf6568c 100644
--- a/dep/jemalloc/src/extent_dss.c
+++ b/dep/jemalloc/src/extent_dss.c
@@ -154,7 +154,7 @@ extent_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr, size_t size,
 			    (uintptr_t)gap_addr_page;
 			if (gap_size_page != 0) {
 				extent_init(gap, arena, gap_addr_page,
-				    gap_size_page, false, NSIZES,
+				    gap_size_page, false, SC_NSIZES,
 				    arena_extent_sn_next(arena),
 				    extent_state_active, false, true, true);
 			}
@@ -198,7 +198,7 @@ extent_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr, size_t size,
 					extent_t extent;
 
 					extent_init(&extent, arena, ret, size,
-					    size, false, NSIZES,
+					    size, false, SC_NSIZES,
 					    extent_state_active, false, true,
 					    true);
 					if (extent_purge_forced_wrapper(tsdn,
diff --git a/dep/jemalloc/src/hook.c b/dep/jemalloc/src/hook.c
new file mode 100644
index 00000000000..9ac703cf9f5
--- /dev/null
+++ b/dep/jemalloc/src/hook.c
@@ -0,0 +1,195 @@
+#include "jemalloc/internal/jemalloc_preamble.h"
+
+#include "jemalloc/internal/hook.h"
+
+#include "jemalloc/internal/atomic.h"
+#include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/seq.h"
+
+typedef struct hooks_internal_s hooks_internal_t;
+struct hooks_internal_s {
+	hooks_t hooks;
+	bool in_use;
+};
+
+seq_define(hooks_internal_t, hooks)
+
+static atomic_u_t nhooks = ATOMIC_INIT(0);
+static seq_hooks_t hooks[HOOK_MAX];
+static malloc_mutex_t hooks_mu;
+
+bool
+hook_boot() {
+	return malloc_mutex_init(&hooks_mu, "hooks", WITNESS_RANK_HOOK,
+	    malloc_mutex_rank_exclusive);
+}
+
+static void *
+hook_install_locked(hooks_t *to_install) {
+	hooks_internal_t hooks_internal;
+	for (int i = 0; i < HOOK_MAX; i++) {
+		bool success = seq_try_load_hooks(&hooks_internal, &hooks[i]);
+		/* We hold mu; no concurrent access. */
+		assert(success);
+		if (!hooks_internal.in_use) {
+			hooks_internal.hooks = *to_install;
+			hooks_internal.in_use = true;
+			seq_store_hooks(&hooks[i], &hooks_internal);
+			atomic_store_u(&nhooks,
+			    atomic_load_u(&nhooks, ATOMIC_RELAXED) + 1,
+			    ATOMIC_RELAXED);
+			return &hooks[i];
+		}
+	}
+	return NULL;
+}
+
+void *
+hook_install(tsdn_t *tsdn, hooks_t *to_install) {
+	malloc_mutex_lock(tsdn, &hooks_mu);
+	void *ret = hook_install_locked(to_install);
+	if (ret != NULL) {
+		tsd_global_slow_inc(tsdn);
+	}
+	malloc_mutex_unlock(tsdn, &hooks_mu);
+	return ret;
+}
+
+static void
+hook_remove_locked(seq_hooks_t *to_remove) {
+	hooks_internal_t hooks_internal;
+	bool success = seq_try_load_hooks(&hooks_internal, to_remove);
+	/* We hold mu; no concurrent access. */
+	assert(success);
+	/* Should only remove hooks that were added. */
+	assert(hooks_internal.in_use);
+	hooks_internal.in_use = false;
+	seq_store_hooks(to_remove, &hooks_internal);
+	atomic_store_u(&nhooks, atomic_load_u(&nhooks, ATOMIC_RELAXED) - 1,
+	    ATOMIC_RELAXED);
+}
+
+void
+hook_remove(tsdn_t *tsdn, void *opaque) {
+	if (config_debug) {
+		char *hooks_begin = (char *)&hooks[0];
+		char *hooks_end = (char *)&hooks[HOOK_MAX];
+		char *hook = (char *)opaque;
+		assert(hooks_begin <= hook && hook < hooks_end
+		    && (hook - hooks_begin) % sizeof(seq_hooks_t) == 0);
+	}
+	malloc_mutex_lock(tsdn, &hooks_mu);
+	hook_remove_locked((seq_hooks_t *)opaque);
+	tsd_global_slow_dec(tsdn);
+	malloc_mutex_unlock(tsdn, &hooks_mu);
+}
+
+#define FOR_EACH_HOOK_BEGIN(hooks_internal_ptr)				\
+for (int for_each_hook_counter = 0;					\
+    for_each_hook_counter < HOOK_MAX;					\
+    for_each_hook_counter++) {						\
+	bool for_each_hook_success = seq_try_load_hooks(		\
+	    (hooks_internal_ptr), &hooks[for_each_hook_counter]);	\
+	if (!for_each_hook_success) {					\
+		continue;						\
+	}								\
+	if (!(hooks_internal_ptr)->in_use) {				\
+		continue;						\
+	}
+#define FOR_EACH_HOOK_END						\
+}
+
+static bool *
+hook_reentrantp() {
+	/*
+	 * We prevent user reentrancy within hooks.  This is basically just a
+	 * thread-local bool that triggers an early-exit.
+	 *
+	 * We don't fold in_hook into reentrancy.  There are two reasons for
+	 * this:
+	 * - Right now, we turn on reentrancy during things like extent hook
+	 *   execution.  Allocating during extent hooks is not officially
+	 *   supported, but we don't want to break it for the time being.  These
+	 *   sorts of allocations should probably still be hooked, though.
+	 * - If a hook allocates, we may want it to be relatively fast (after
+	 *   all, it executes on every allocator operation).  Turning on
+	 *   reentrancy is a fairly heavyweight mode (disabling tcache,
+	 *   redirecting to arena 0, etc.).  It's possible we may one day want
+	 *   to turn on reentrant mode here, if it proves too difficult to keep
+	 *   this working.  But that's fairly easy for us to see; OTOH, people
+	 *   not using hooks because they're too slow is easy for us to miss.
+	 *
+	 * The tricky part is
+	 * that this code might get invoked even if we don't have access to tsd.
+	 * This function mimics getting a pointer to thread-local data, except
+	 * that it might secretly return a pointer to some global data if we
+	 * know that the caller will take the early-exit path.
+	 * If we return a bool that indicates that we are reentrant, then the
+	 * caller will go down the early exit path, leaving the global
+	 * untouched.
+	 */
+	static bool in_hook_global = true;
+	tsdn_t *tsdn = tsdn_fetch();
+	tcache_t *tcache = tsdn_tcachep_get(tsdn);
+	if (tcache != NULL) {
+		return &tcache->in_hook;
+	}
+	return &in_hook_global;
+}
+
+#define HOOK_PROLOGUE							\
+	if (likely(atomic_load_u(&nhooks, ATOMIC_RELAXED) == 0)) {	\
+		return;							\
+	}								\
+	bool *in_hook = hook_reentrantp();				\
+	if (*in_hook) {							\
+		return;							\
+	}								\
+	*in_hook = true;
+
+#define HOOK_EPILOGUE							\
+	*in_hook = false;
+
+void
+hook_invoke_alloc(hook_alloc_t type, void *result, uintptr_t result_raw,
+    uintptr_t args_raw[3]) {
+	HOOK_PROLOGUE
+
+	hooks_internal_t hook;
+	FOR_EACH_HOOK_BEGIN(&hook)
+		hook_alloc h = hook.hooks.alloc_hook;
+		if (h != NULL) {
+			h(hook.hooks.extra, type, result, result_raw, args_raw);
+		}
+	FOR_EACH_HOOK_END
+
+	HOOK_EPILOGUE
+}
+
+void
+hook_invoke_dalloc(hook_dalloc_t type, void *address, uintptr_t args_raw[3]) {
+	HOOK_PROLOGUE
+	hooks_internal_t hook;
+	FOR_EACH_HOOK_BEGIN(&hook)
+		hook_dalloc h = hook.hooks.dalloc_hook;
+		if (h != NULL) {
+			h(hook.hooks.extra, type, address, args_raw);
+		}
+	FOR_EACH_HOOK_END
+	HOOK_EPILOGUE
+}
+
+void
+hook_invoke_expand(hook_expand_t type, void *address, size_t old_usize,
+    size_t new_usize, uintptr_t result_raw, uintptr_t args_raw[4]) {
+	HOOK_PROLOGUE
+	hooks_internal_t hook;
+	FOR_EACH_HOOK_BEGIN(&hook)
+		hook_expand h = hook.hooks.expand_hook;
+		if (h != NULL) {
+			h(hook.hooks.extra, type, address, old_usize, new_usize,
+			    result_raw, args_raw);
+		}
+	FOR_EACH_HOOK_END
+	HOOK_EPILOGUE
+}
diff --git a/dep/jemalloc/src/jemalloc.c b/dep/jemalloc/src/jemalloc.c
index f93c16fa3cf..c8afa9c4d55 100644
--- a/dep/jemalloc/src/jemalloc.c
+++ b/dep/jemalloc/src/jemalloc.c
@@ -7,12 +7,13 @@
 #include "jemalloc/internal/ctl.h"
 #include "jemalloc/internal/extent_dss.h"
 #include "jemalloc/internal/extent_mmap.h"
+#include "jemalloc/internal/hook.h"
 #include "jemalloc/internal/jemalloc_internal_types.h"
 #include "jemalloc/internal/log.h"
 #include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/rtree.h"
-#include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/sc.h"
 #include "jemalloc/internal/spin.h"
 #include "jemalloc/internal/sz.h"
 #include "jemalloc/internal/ticker.h"
@@ -85,8 +86,10 @@ malloc_mutex_t arenas_lock;
 JEMALLOC_ALIGNED(CACHELINE)
 atomic_p_t		arenas[MALLOCX_ARENA_LIMIT];
 static atomic_u_t	narenas_total; /* Use narenas_total_*(). */
-static arena_t		*a0; /* arenas[0]; read-only after initialization. */
-unsigned		narenas_auto; /* Read-only after initialization. */
+/* Below three are read-only after initialization. */
+static arena_t		*a0; /* arenas[0]. */
+unsigned		narenas_auto;
+unsigned		manual_arena_base;
 
 typedef enum {
 	malloc_init_uninitialized	= 3,
@@ -326,7 +329,7 @@ arena_init_locked(tsdn_t *tsdn, unsigned ind, extent_hooks_t *extent_hooks) {
 	 */
 	arena = arena_get(tsdn, ind, false);
 	if (arena != NULL) {
-		assert(ind < narenas_auto);
+		assert(arena_is_auto(arena));
 		return arena;
 	}
 
@@ -341,12 +344,12 @@ arena_new_create_background_thread(tsdn_t *tsdn, unsigned ind) {
 	if (ind == 0) {
 		return;
 	}
-	if (have_background_thread) {
-		bool err;
-		malloc_mutex_lock(tsdn, &background_thread_lock);
-		err = background_thread_create(tsdn_tsd(tsdn), ind);
-		malloc_mutex_unlock(tsdn, &background_thread_lock);
-		if (err) {
+	/*
+	 * Avoid creating a new background thread just for the huge arena, which
+	 * purges eagerly by default.
+	 */
+	if (have_background_thread && !arena_is_huge(ind)) {
+		if (background_thread_create(tsdn_tsd(tsdn), ind)) {
 			malloc_printf("<jemalloc>: error in background thread "
 				      "creation for arena %u. Abort.\n", ind);
 			abort();
@@ -376,6 +379,14 @@ arena_bind(tsd_t *tsd, unsigned ind, bool internal) {
 		tsd_iarena_set(tsd, arena);
 	} else {
 		tsd_arena_set(tsd, arena);
+		unsigned shard = atomic_fetch_add_u(&arena->binshard_next, 1,
+		    ATOMIC_RELAXED);
+		tsd_binshards_t *bins = tsd_binshardsp_get(tsd);
+		for (unsigned i = 0; i < SC_NBINS; i++) {
+			assert(bin_infos[i].n_shards > 0 &&
+			    bin_infos[i].n_shards <= BIN_SHARDS_MAX);
+			bins->binshard[i] = shard % bin_infos[i].n_shards;
+		}
 	}
 }
 
@@ -761,6 +772,50 @@ init_opt_stats_print_opts(const char *v, size_t vlen) {
 	assert(opts_len == strlen(opt_stats_print_opts));
 }
 
+/* Reads the next size pair in a multi-sized option. */
+static bool
+malloc_conf_multi_sizes_next(const char **slab_size_segment_cur,
+    size_t *vlen_left, size_t *slab_start, size_t *slab_end, size_t *new_size) {
+	const char *cur = *slab_size_segment_cur;
+	char *end;
+	uintmax_t um;
+
+	set_errno(0);
+
+	/* First number, then '-' */
+	um = malloc_strtoumax(cur, &end, 0);
+	if (get_errno() != 0 || *end != '-') {
+		return true;
+	}
+	*slab_start = (size_t)um;
+	cur = end + 1;
+
+	/* Second number, then ':' */
+	um = malloc_strtoumax(cur, &end, 0);
+	if (get_errno() != 0 || *end != ':') {
+		return true;
+	}
+	*slab_end = (size_t)um;
+	cur = end + 1;
+
+	/* Last number */
+	um = malloc_strtoumax(cur, &end, 0);
+	if (get_errno() != 0) {
+		return true;
+	}
+	*new_size = (size_t)um;
+
+	/* Consume the separator if there is one. */
+	if (*end == '|') {
+		end++;
+	}
+
+	*vlen_left -= end - *slab_size_segment_cur;
+	*slab_size_segment_cur = end;
+
+	return false;
+}
+
 static bool
 malloc_conf_next(char const **opts_p, char const **k_p, size_t *klen_p,
     char const **v_p, size_t *vlen_p) {
@@ -850,6 +905,11 @@ malloc_conf_error(const char *msg, const char *k, size_t klen, const char *v,
 	malloc_printf("<jemalloc>: %s: %.*s:%.*s\n", msg, (int)klen, k,
 	    (int)vlen, v);
 	/* If abort_conf is set, error out after processing all options. */
+	const char *experimental = "experimental_";
+	if (strncmp(k, experimental, strlen(experimental)) == 0) {
+		/* However, tolerate experimental features. */
+		return;
+	}
 	had_conf_error = true;
 }
 
@@ -869,7 +929,7 @@ malloc_slow_flag_init(void) {
 }
 
 static void
-malloc_conf_init(void) {
+malloc_conf_init(sc_data_t *sc_data, unsigned bin_shard_sizes[SC_NBINS]) {
 	unsigned i;
 	char buf[PATH_MAX + 1];
 	const char *opts, *k, *v;
@@ -910,7 +970,12 @@ malloc_conf_init(void) {
 			 * Try to use the contents of the "/etc/malloc.conf"
 			 * symbolic link's name.
 			 */
+#ifndef JEMALLOC_READLINKAT
 			linklen = readlink(linkname, buf, sizeof(buf) - 1);
+#else
+			linklen = readlinkat(AT_FDCWD, linkname, buf,
+			    sizeof(buf) - 1);
+#endif
 			if (linklen == -1) {
 				/* No configuration specified. */
 				linklen = 0;
@@ -967,6 +1032,14 @@ malloc_conf_init(void) {
 				}					\
 				continue;				\
 			}
+      /*
+       * One of the CONF_MIN macros below expands, in one of the use points,
+       * to "unsigned integer < 0", which is always false, triggering the
+       * GCC -Wtype-limits warning, which we disable here and re-enable below.
+       */
+      JEMALLOC_DIAGNOSTIC_PUSH
+      JEMALLOC_DIAGNOSTIC_IGNORE_TYPE_LIMITS
+
 #define CONF_MIN_no(um, min)	false
 #define CONF_MIN_yes(um, min)	((um) < (min))
 #define CONF_MAX_no(um, max)	false
@@ -1094,6 +1167,28 @@ malloc_conf_init(void) {
 			}
 			CONF_HANDLE_UNSIGNED(opt_narenas, "narenas", 1,
 			    UINT_MAX, yes, no, false)
+			if (CONF_MATCH("bin_shards")) {
+				const char *bin_shards_segment_cur = v;
+				size_t vlen_left = vlen;
+				do {
+					size_t size_start;
+					size_t size_end;
+					size_t nshards;
+					bool err = malloc_conf_multi_sizes_next(
+					    &bin_shards_segment_cur, &vlen_left,
+					    &size_start, &size_end, &nshards);
+					if (err || bin_update_shard_size(
+					    bin_shard_sizes, size_start,
+					    size_end, nshards)) {
+						malloc_conf_error(
+						    "Invalid settings for "
+						    "bin_shards", k, klen, v,
+						    vlen);
+						break;
+					}
+				} while (vlen_left > 0);
+				continue;
+			}
 			CONF_HANDLE_SSIZE_T(opt_dirty_decay_ms,
 			    "dirty_decay_ms", -1, NSTIME_SEC_MAX * KQU(1000) <
 			    QU(SSIZE_MAX) ? NSTIME_SEC_MAX * KQU(1000) :
@@ -1141,11 +1236,24 @@ malloc_conf_init(void) {
 				CONF_HANDLE_BOOL(opt_xmalloc, "xmalloc")
 			}
 			CONF_HANDLE_BOOL(opt_tcache, "tcache")
-			CONF_HANDLE_SIZE_T(opt_lg_extent_max_active_fit,
-			    "lg_extent_max_active_fit", 0,
-			    (sizeof(size_t) << 3), yes, yes, false)
 			CONF_HANDLE_SSIZE_T(opt_lg_tcache_max, "lg_tcache_max",
 			    -1, (sizeof(size_t) << 3) - 1)
+
+			/*
+			 * The runtime option of oversize_threshold remains
+			 * undocumented.  It may be tweaked in the next major
+			 * release (6.0).  The default value 8M is rather
+			 * conservative / safe.  Tuning it further down may
+			 * improve fragmentation a bit more, but may also cause
+			 * contention on the huge arena.
+			 */
+			CONF_HANDLE_SIZE_T(opt_oversize_threshold,
+			    "oversize_threshold", 0, SC_LARGE_MAXCLASS, no, yes,
+			    false)
+			CONF_HANDLE_SIZE_T(opt_lg_extent_max_active_fit,
+			    "lg_extent_max_active_fit", 0,
+			    (sizeof(size_t) << 3), no, yes, false)
+
 			if (strncmp("percpu_arena", k, klen) == 0) {
 				bool match = false;
 				for (int i = percpu_arena_mode_names_base; i <
@@ -1174,6 +1282,31 @@ malloc_conf_init(void) {
 					   "max_background_threads", 1,
 					   opt_max_background_threads, yes, yes,
 					   true);
+			if (CONF_MATCH("slab_sizes")) {
+				bool err;
+				const char *slab_size_segment_cur = v;
+				size_t vlen_left = vlen;
+				do {
+					size_t slab_start;
+					size_t slab_end;
+					size_t pgs;
+					err = malloc_conf_multi_sizes_next(
+					    &slab_size_segment_cur,
+					    &vlen_left, &slab_start, &slab_end,
+					    &pgs);
+					if (!err) {
+						sc_data_update_slab_size(
+						    sc_data, slab_start,
+						    slab_end, (int)pgs);
+					} else {
+						malloc_conf_error(
+						    "Invalid settings for "
+						    "slab_sizes", k, klen, v,
+						    vlen);
+					}
+				} while (!err && vlen_left > 0);
+				continue;
+			}
 			if (config_prof) {
 				CONF_HANDLE_BOOL(opt_prof, "prof")
 				CONF_HANDLE_CHAR_P(opt_prof_prefix,
@@ -1191,6 +1324,7 @@ malloc_conf_init(void) {
 				CONF_HANDLE_BOOL(opt_prof_gdump, "prof_gdump")
 				CONF_HANDLE_BOOL(opt_prof_final, "prof_final")
 				CONF_HANDLE_BOOL(opt_prof_leak, "prof_leak")
+				CONF_HANDLE_BOOL(opt_prof_log, "prof_log")
 			}
 			if (config_log) {
 				if (CONF_MATCH("log")) {
@@ -1237,6 +1371,8 @@ malloc_conf_init(void) {
 #undef CONF_HANDLE_SIZE_T
 #undef CONF_HANDLE_SSIZE_T
 #undef CONF_HANDLE_CHAR_P
+    /* Re-enable diagnostic "-Wtype-limits" */
+    JEMALLOC_DIAGNOSTIC_POP
 		}
 		if (opt_abort_conf && had_conf_error) {
 			malloc_abort_invalid_conf();
@@ -1275,10 +1411,33 @@ static bool
 malloc_init_hard_a0_locked() {
 	malloc_initializer = INITIALIZER;
 
+	JEMALLOC_DIAGNOSTIC_PUSH
+	JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+	sc_data_t sc_data = {0};
+	JEMALLOC_DIAGNOSTIC_POP
+
+	/*
+	 * Ordering here is somewhat tricky; we need sc_boot() first, since that
+	 * determines what the size classes will be, and then
+	 * malloc_conf_init(), since any slab size tweaking will need to be done
+	 * before sz_boot and bin_boot, which assume that the values they read
+	 * out of sc_data_global are final.
+	 */
+	sc_boot(&sc_data);
+	unsigned bin_shard_sizes[SC_NBINS];
+	bin_shard_sizes_boot(bin_shard_sizes);
+	/*
+	 * prof_boot0 only initializes opt_prof_prefix.  We need to do it before
+	 * we parse malloc_conf options, in case malloc_conf parsing overwrites
+	 * it.
+	 */
 	if (config_prof) {
 		prof_boot0();
 	}
-	malloc_conf_init();
+	malloc_conf_init(&sc_data, bin_shard_sizes);
+	sz_boot(&sc_data);
+	bin_boot(&sc_data, bin_shard_sizes);
+
 	if (opt_stats_print) {
 		/* Print statistics at exit. */
 		if (atexit(stats_print_atexit) != 0) {
@@ -1303,7 +1462,7 @@ malloc_init_hard_a0_locked() {
 	if (config_prof) {
 		prof_boot1();
 	}
-	arena_boot();
+	arena_boot(&sc_data);
 	if (tcache_boot(TSDN_NULL)) {
 		return true;
 	}
@@ -1311,11 +1470,13 @@ malloc_init_hard_a0_locked() {
 	    malloc_mutex_rank_exclusive)) {
 		return true;
 	}
+	hook_boot();
 	/*
 	 * Create enough scaffolding to allow recursive allocation in
 	 * malloc_ncpus().
 	 */
 	narenas_auto = 1;
+	manual_arena_base = narenas_auto + 1;
 	memset(arenas, 0, sizeof(arena_t *) * narenas_auto);
 	/*
 	 * Initialize one arena here.  The rest are lazily created in
@@ -1463,6 +1624,10 @@ malloc_init_narenas(void) {
 		    narenas_auto);
 	}
 	narenas_total_set(narenas_auto);
+	if (arena_init_huge()) {
+		narenas_total_inc();
+	}
+	manual_arena_base = narenas_total_get();
 
 	return false;
 }
@@ -1560,11 +1725,7 @@ malloc_init_hard(void) {
 		 * sets isthreaded) needs to be called without holding any lock.
 		 */
 		background_thread_ctl_init(tsd_tsdn(tsd));
-
-		malloc_mutex_lock(tsd_tsdn(tsd), &background_thread_lock);
-		bool err = background_thread_create(tsd, 0);
-		malloc_mutex_unlock(tsd_tsdn(tsd), &background_thread_lock);
-		if (err) {
+		if (background_thread_create(tsd, 0)) {
 			return true;
 		}
 	}
@@ -1587,8 +1748,7 @@ typedef struct static_opts_s static_opts_t;
 struct static_opts_s {
 	/* Whether or not allocation size may overflow. */
 	bool may_overflow;
-	/* Whether or not allocations of size 0 should be treated as size 1. */
-	bool bump_empty_alloc;
+
 	/*
 	 * Whether to assert that allocations are not of size 0 (after any
 	 * bumping).
@@ -1621,12 +1781,15 @@ struct static_opts_s {
 	 * initialization) options.
 	 */
 	bool slow;
+	/*
+	 * Return size.
+	 */
+	bool usize;
 };
 
 JEMALLOC_ALWAYS_INLINE void
 static_opts_init(static_opts_t *static_opts) {
 	static_opts->may_overflow = false;
-	static_opts->bump_empty_alloc = false;
 	static_opts->assert_nonempty_alloc = false;
 	static_opts->null_out_result_on_error = false;
 	static_opts->set_errno_on_error = false;
@@ -1634,6 +1797,7 @@ static_opts_init(static_opts_t *static_opts) {
 	static_opts->oom_string = "";
 	static_opts->invalid_alignment_string = "";
 	static_opts->slow = false;
+	static_opts->usize = false;
 }
 
 /*
@@ -1648,6 +1812,7 @@ static_opts_init(static_opts_t *static_opts) {
 typedef struct dynamic_opts_s dynamic_opts_t;
 struct dynamic_opts_s {
 	void **result;
+	size_t usize;
 	size_t num_items;
 	size_t item_size;
 	size_t alignment;
@@ -1659,6 +1824,7 @@ struct dynamic_opts_s {
 JEMALLOC_ALWAYS_INLINE void
 dynamic_opts_init(dynamic_opts_t *dynamic_opts) {
 	dynamic_opts->result = NULL;
+	dynamic_opts->usize = 0;
 	dynamic_opts->num_items = 0;
 	dynamic_opts->item_size = 0;
 	dynamic_opts->alignment = 0;
@@ -1722,12 +1888,13 @@ imalloc_sample(static_opts_t *sopts, dynamic_opts_t *dopts, tsd_t *tsd,
 	szind_t ind_large;
 	size_t bumped_usize = usize;
 
-	if (usize <= SMALL_MAXCLASS) {
-		assert(((dopts->alignment == 0) ? sz_s2u(LARGE_MINCLASS) :
-		    sz_sa2u(LARGE_MINCLASS, dopts->alignment))
-		    == LARGE_MINCLASS);
-		ind_large = sz_size2index(LARGE_MINCLASS);
-		bumped_usize = sz_s2u(LARGE_MINCLASS);
+	if (usize <= SC_SMALL_MAXCLASS) {
+		assert(((dopts->alignment == 0) ?
+		    sz_s2u(SC_LARGE_MINCLASS) :
+		    sz_sa2u(SC_LARGE_MINCLASS, dopts->alignment))
+			== SC_LARGE_MINCLASS);
+		ind_large = sz_size2index(SC_LARGE_MINCLASS);
+		bumped_usize = sz_s2u(SC_LARGE_MINCLASS);
 		ret = imalloc_no_sample(sopts, dopts, tsd, bumped_usize,
 		    bumped_usize, ind_large);
 		if (unlikely(ret == NULL)) {
@@ -1811,12 +1978,6 @@ imalloc_body(static_opts_t *sopts, dynamic_opts_t *dopts, tsd_t *tsd) {
 	}
 
 	/* Validate the user input. */
-	if (sopts->bump_empty_alloc) {
-		if (unlikely(size == 0)) {
-			size = 1;
-		}
-	}
-
 	if (sopts->assert_nonempty_alloc) {
 		assert (size != 0);
 	}
@@ -1830,16 +1991,20 @@ imalloc_body(static_opts_t *sopts, dynamic_opts_t *dopts, tsd_t *tsd) {
 
 	if (dopts->alignment == 0) {
 		ind = sz_size2index(size);
-		if (unlikely(ind >= NSIZES)) {
+		if (unlikely(ind >= SC_NSIZES)) {
 			goto label_oom;
 		}
-		if (config_stats || (config_prof && opt_prof)) {
+		if (config_stats || (config_prof && opt_prof) || sopts->usize) {
 			usize = sz_index2size(ind);
-			assert(usize > 0 && usize <= LARGE_MAXCLASS);
+			dopts->usize = usize;
+			assert(usize > 0 && usize
+			    <= SC_LARGE_MAXCLASS);
 		}
 	} else {
 		usize = sz_sa2u(size, dopts->alignment);
-		if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+		dopts->usize = usize;
+		if (unlikely(usize == 0
+		    || usize > SC_LARGE_MAXCLASS)) {
 			goto label_oom;
 		}
 	}
@@ -1875,7 +2040,8 @@ imalloc_body(static_opts_t *sopts, dynamic_opts_t *dopts, tsd_t *tsd) {
 
 		alloc_ctx_t alloc_ctx;
 		if (likely((uintptr_t)tctx == (uintptr_t)1U)) {
-			alloc_ctx.slab = (usize <= SMALL_MAXCLASS);
+			alloc_ctx.slab = (usize
+			    <= SC_SMALL_MAXCLASS);
 			allocation = imalloc_no_sample(
 			    sopts, dopts, tsd, usize, usize, ind);
 		} else if ((uintptr_t)tctx > (uintptr_t)1U) {
@@ -1980,9 +2146,8 @@ label_invalid_alignment:
 	return EINVAL;
 }
 
-/* Returns the errno-style error code of the allocation. */
-JEMALLOC_ALWAYS_INLINE int
-imalloc(static_opts_t *sopts, dynamic_opts_t *dopts) {
+JEMALLOC_ALWAYS_INLINE bool
+imalloc_init_check(static_opts_t *sopts, dynamic_opts_t *dopts) {
 	if (unlikely(!malloc_initialized()) && unlikely(malloc_init())) {
 		if (config_xmalloc && unlikely(opt_xmalloc)) {
 			malloc_write(sopts->oom_string);
@@ -1992,6 +2157,16 @@ imalloc(static_opts_t *sopts, dynamic_opts_t *dopts) {
 		set_errno(ENOMEM);
 		*dopts->result = NULL;
 
+		return false;
+	}
+
+	return true;
+}
+
+/* Returns the errno-style error code of the allocation. */
+JEMALLOC_ALWAYS_INLINE int
+imalloc(static_opts_t *sopts, dynamic_opts_t *dopts) {
+	if (tsd_get_allocates() && !imalloc_init_check(sopts, dopts)) {
 		return ENOMEM;
 	}
 
@@ -2004,19 +2179,18 @@ imalloc(static_opts_t *sopts, dynamic_opts_t *dopts) {
 		sopts->slow = false;
 		return imalloc_body(sopts, dopts, tsd);
 	} else {
+		if (!tsd_get_allocates() && !imalloc_init_check(sopts, dopts)) {
+			return ENOMEM;
+		}
+
 		sopts->slow = true;
 		return imalloc_body(sopts, dopts, tsd);
 	}
 }
-/******************************************************************************/
-/*
- * Begin malloc(3)-compatible functions.
- */
 
-JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
-void JEMALLOC_NOTHROW *
-JEMALLOC_ATTR(malloc) JEMALLOC_ALLOC_SIZE(1)
-je_malloc(size_t size) {
+JEMALLOC_NOINLINE
+void *
+malloc_default(size_t size) {
 	void *ret;
 	static_opts_t sopts;
 	dynamic_opts_t dopts;
@@ -2026,7 +2200,6 @@ je_malloc(size_t size) {
 	static_opts_init(&sopts);
 	dynamic_opts_init(&dopts);
 
-	sopts.bump_empty_alloc = true;
 	sopts.null_out_result_on_error = true;
 	sopts.set_errno_on_error = true;
 	sopts.oom_string = "<jemalloc>: Error in malloc(): out of memory\n";
@@ -2036,12 +2209,107 @@ je_malloc(size_t size) {
 	dopts.item_size = size;
 
 	imalloc(&sopts, &dopts);
+	/*
+	 * Note that this branch gets optimized away -- it immediately follows
+	 * the check on tsd_fast that sets sopts.slow.
+	 */
+	if (sopts.slow) {
+		uintptr_t args[3] = {size};
+		hook_invoke_alloc(hook_alloc_malloc, ret, (uintptr_t)ret, args);
+	}
 
 	LOG("core.malloc.exit", "result: %p", ret);
 
 	return ret;
 }
 
+/******************************************************************************/
+/*
+ * Begin malloc(3)-compatible functions.
+ */
+
+/*
+ * malloc() fastpath.
+ *
+ * Fastpath assumes size <= SC_LOOKUP_MAXCLASS, and that we hit
+ * tcache.  If either of these is false, we tail-call to the slowpath,
+ * malloc_default().  Tail-calling is used to avoid any caller-saved
+ * registers.
+ *
+ * fastpath supports ticker and profiling, both of which will also
+ * tail-call to the slowpath if they fire.
+ */
+JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
+void JEMALLOC_NOTHROW *
+JEMALLOC_ATTR(malloc) JEMALLOC_ALLOC_SIZE(1)
+je_malloc(size_t size) {
+	LOG("core.malloc.entry", "size: %zu", size);
+
+	if (tsd_get_allocates() && unlikely(!malloc_initialized())) {
+		return malloc_default(size);
+	}
+
+	tsd_t *tsd = tsd_get(false);
+	if (unlikely(!tsd || !tsd_fast(tsd) || (size > SC_LOOKUP_MAXCLASS))) {
+		return malloc_default(size);
+	}
+
+	tcache_t *tcache = tsd_tcachep_get(tsd);
+
+	if (unlikely(ticker_trytick(&tcache->gc_ticker))) {
+		return malloc_default(size);
+	}
+
+	szind_t ind = sz_size2index_lookup(size);
+	size_t usize;
+	if (config_stats || config_prof) {
+		usize = sz_index2size(ind);
+	}
+	/* Fast path relies on size being a bin. I.e. SC_LOOKUP_MAXCLASS < SC_SMALL_MAXCLASS */
+	assert(ind < SC_NBINS);
+	assert(size <= SC_SMALL_MAXCLASS);
+
+	if (config_prof) {
+		int64_t bytes_until_sample = tsd_bytes_until_sample_get(tsd);
+		bytes_until_sample -= usize;
+		tsd_bytes_until_sample_set(tsd, bytes_until_sample);
+
+		if (unlikely(bytes_until_sample < 0)) {
+			/*
+			 * Avoid a prof_active check on the fastpath.
+			 * If prof_active is false, set bytes_until_sample to
+			 * a large value.  If prof_active is set to true,
+			 * bytes_until_sample will be reset.
+			 */
+			if (!prof_active) {
+				tsd_bytes_until_sample_set(tsd, SSIZE_MAX);
+			}
+			return malloc_default(size);
+		}
+	}
+
+	cache_bin_t *bin = tcache_small_bin_get(tcache, ind);
+	bool tcache_success;
+	void* ret = cache_bin_alloc_easy(bin, &tcache_success);
+
+	if (tcache_success) {
+		if (config_stats) {
+			*tsd_thread_allocatedp_get(tsd) += usize;
+			bin->tstats.nrequests++;
+		}
+		if (config_prof) {
+			tcache->prof_accumbytes += usize;
+		}
+
+		LOG("core.malloc.exit", "result: %p", ret);
+
+		/* Fastpath success */
+		return ret;
+	}
+
+	return malloc_default(size);
+}
+
 JEMALLOC_EXPORT int JEMALLOC_NOTHROW
 JEMALLOC_ATTR(nonnull(1))
 je_posix_memalign(void **memptr, size_t alignment, size_t size) {
@@ -2055,7 +2323,6 @@ je_posix_memalign(void **memptr, size_t alignment, size_t size) {
 	static_opts_init(&sopts);
 	dynamic_opts_init(&dopts);
 
-	sopts.bump_empty_alloc = true;
 	sopts.min_alignment = sizeof(void *);
 	sopts.oom_string =
 	    "<jemalloc>: Error allocating aligned memory: out of memory\n";
@@ -2068,6 +2335,12 @@ je_posix_memalign(void **memptr, size_t alignment, size_t size) {
 	dopts.alignment = alignment;
 
 	ret = imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {(uintptr_t)memptr, (uintptr_t)alignment,
+			(uintptr_t)size};
+		hook_invoke_alloc(hook_alloc_posix_memalign, *memptr,
+		    (uintptr_t)ret, args);
+	}
 
 	LOG("core.posix_memalign.exit", "result: %d, alloc ptr: %p", ret,
 	    *memptr);
@@ -2090,7 +2363,6 @@ je_aligned_alloc(size_t alignment, size_t size) {
 	static_opts_init(&sopts);
 	dynamic_opts_init(&dopts);
 
-	sopts.bump_empty_alloc = true;
 	sopts.null_out_result_on_error = true;
 	sopts.set_errno_on_error = true;
 	sopts.min_alignment = 1;
@@ -2105,6 +2377,11 @@ je_aligned_alloc(size_t alignment, size_t size) {
 	dopts.alignment = alignment;
 
 	imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {(uintptr_t)alignment, (uintptr_t)size};
+		hook_invoke_alloc(hook_alloc_aligned_alloc, ret,
+		    (uintptr_t)ret, args);
+	}
 
 	LOG("core.aligned_alloc.exit", "result: %p", ret);
 
@@ -2125,7 +2402,6 @@ je_calloc(size_t num, size_t size) {
 	dynamic_opts_init(&dopts);
 
 	sopts.may_overflow = true;
-	sopts.bump_empty_alloc = true;
 	sopts.null_out_result_on_error = true;
 	sopts.set_errno_on_error = true;
 	sopts.oom_string = "<jemalloc>: Error in calloc(): out of memory\n";
@@ -2136,6 +2412,10 @@ je_calloc(size_t num, size_t size) {
 	dopts.zero = true;
 
 	imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {(uintptr_t)num, (uintptr_t)size};
+		hook_invoke_alloc(hook_alloc_calloc, ret, (uintptr_t)ret, args);
+	}
 
 	LOG("core.calloc.exit", "result: %p", ret);
 
@@ -2144,20 +2424,22 @@ je_calloc(size_t num, size_t size) {
 
 static void *
 irealloc_prof_sample(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize,
-    prof_tctx_t *tctx) {
+    prof_tctx_t *tctx, hook_ralloc_args_t *hook_args) {
 	void *p;
 
 	if (tctx == NULL) {
 		return NULL;
 	}
-	if (usize <= SMALL_MAXCLASS) {
-		p = iralloc(tsd, old_ptr, old_usize, LARGE_MINCLASS, 0, false);
+	if (usize <= SC_SMALL_MAXCLASS) {
+		p = iralloc(tsd, old_ptr, old_usize,
+		    SC_LARGE_MINCLASS, 0, false, hook_args);
 		if (p == NULL) {
 			return NULL;
 		}
 		arena_prof_promote(tsd_tsdn(tsd), p, usize);
 	} else {
-		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false);
+		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false,
+		    hook_args);
 	}
 
 	return p;
@@ -2165,7 +2447,7 @@ irealloc_prof_sample(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize,
 
 JEMALLOC_ALWAYS_INLINE void *
 irealloc_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize,
-   alloc_ctx_t *alloc_ctx) {
+   alloc_ctx_t *alloc_ctx, hook_ralloc_args_t *hook_args) {
 	void *p;
 	bool prof_active;
 	prof_tctx_t *old_tctx, *tctx;
@@ -2174,9 +2456,11 @@ irealloc_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize,
 	old_tctx = prof_tctx_get(tsd_tsdn(tsd), old_ptr, alloc_ctx);
 	tctx = prof_alloc_prep(tsd, usize, prof_active, true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
-		p = irealloc_prof_sample(tsd, old_ptr, old_usize, usize, tctx);
+		p = irealloc_prof_sample(tsd, old_ptr, old_usize, usize, tctx,
+		    hook_args);
 	} else {
-		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false);
+		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false,
+		    hook_args);
 	}
 	if (unlikely(p == NULL)) {
 		prof_alloc_rollback(tsd, tctx, true);
@@ -2205,7 +2489,7 @@ ifree(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path) {
 	rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
 	rtree_szind_slab_read(tsd_tsdn(tsd), &extents_rtree, rtree_ctx,
 	    (uintptr_t)ptr, true, &alloc_ctx.szind, &alloc_ctx.slab);
-	assert(alloc_ctx.szind != NSIZES);
+	assert(alloc_ctx.szind != SC_NSIZES);
 
 	size_t usize;
 	if (config_prof && opt_prof) {
@@ -2286,11 +2570,12 @@ isfree(tsd_t *tsd, void *ptr, size_t usize, tcache_t *tcache, bool slow_path) {
 JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
 void JEMALLOC_NOTHROW *
 JEMALLOC_ALLOC_SIZE(2)
-je_realloc(void *ptr, size_t size) {
+je_realloc(void *ptr, size_t arg_size) {
 	void *ret;
 	tsdn_t *tsdn JEMALLOC_CC_SILENCE_INIT(NULL);
 	size_t usize JEMALLOC_CC_SILENCE_INIT(0);
 	size_t old_usize = 0;
+	size_t size = arg_size;
 
 	LOG("core.realloc.entry", "ptr: %p, size: %zu\n", ptr, size);
 
@@ -2305,6 +2590,10 @@ je_realloc(void *ptr, size_t size) {
 			} else {
 				tcache = NULL;
 			}
+
+			uintptr_t args[3] = {(uintptr_t)ptr, size};
+			hook_invoke_dalloc(hook_dalloc_realloc, ptr, args);
+
 			ifree(tsd, ptr, tcache, true);
 
 			LOG("core.realloc.exit", "result: %p", NULL);
@@ -2319,29 +2608,58 @@ je_realloc(void *ptr, size_t size) {
 
 		check_entry_exit_locking(tsd_tsdn(tsd));
 
+
+		hook_ralloc_args_t hook_args = {true, {(uintptr_t)ptr,
+			(uintptr_t)arg_size, 0, 0}};
+
 		alloc_ctx_t alloc_ctx;
 		rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
 		rtree_szind_slab_read(tsd_tsdn(tsd), &extents_rtree, rtree_ctx,
 		    (uintptr_t)ptr, true, &alloc_ctx.szind, &alloc_ctx.slab);
-		assert(alloc_ctx.szind != NSIZES);
+		assert(alloc_ctx.szind != SC_NSIZES);
 		old_usize = sz_index2size(alloc_ctx.szind);
 		assert(old_usize == isalloc(tsd_tsdn(tsd), ptr));
 		if (config_prof && opt_prof) {
 			usize = sz_s2u(size);
-			ret = unlikely(usize == 0 || usize > LARGE_MAXCLASS) ?
-			    NULL : irealloc_prof(tsd, ptr, old_usize, usize,
-			    &alloc_ctx);
+			if (unlikely(usize == 0
+			    || usize > SC_LARGE_MAXCLASS)) {
+				ret = NULL;
+			} else {
+				ret = irealloc_prof(tsd, ptr, old_usize, usize,
+				    &alloc_ctx, &hook_args);
+			}
 		} else {
 			if (config_stats) {
 				usize = sz_s2u(size);
 			}
-			ret = iralloc(tsd, ptr, old_usize, size, 0, false);
+			ret = iralloc(tsd, ptr, old_usize, size, 0, false,
+			    &hook_args);
 		}
 		tsdn = tsd_tsdn(tsd);
 	} else {
 		/* realloc(NULL, size) is equivalent to malloc(size). */
-		void *ret = je_malloc(size);
-		LOG("core.realloc.exit", "result: %p", ret);
+		static_opts_t sopts;
+		dynamic_opts_t dopts;
+
+		static_opts_init(&sopts);
+		dynamic_opts_init(&dopts);
+
+		sopts.null_out_result_on_error = true;
+		sopts.set_errno_on_error = true;
+		sopts.oom_string =
+		    "<jemalloc>: Error in realloc(): out of memory\n";
+
+		dopts.result = &ret;
+		dopts.num_items = 1;
+		dopts.item_size = size;
+
+		imalloc(&sopts, &dopts);
+		if (sopts.slow) {
+			uintptr_t args[3] = {(uintptr_t)ptr, arg_size};
+			hook_invoke_alloc(hook_alloc_realloc, ret,
+			    (uintptr_t)ret, args);
+		}
+
 		return ret;
 	}
 
@@ -2368,10 +2686,9 @@ je_realloc(void *ptr, size_t size) {
 	return ret;
 }
 
-JEMALLOC_EXPORT void JEMALLOC_NOTHROW
-je_free(void *ptr) {
-	LOG("core.free.entry", "ptr: %p", ptr);
-
+JEMALLOC_NOINLINE
+void
+free_default(void *ptr) {
 	UTRACE(ptr, 0, 0);
 	if (likely(ptr != NULL)) {
 		/*
@@ -2397,10 +2714,79 @@ je_free(void *ptr) {
 			} else {
 				tcache = NULL;
 			}
+			uintptr_t args_raw[3] = {(uintptr_t)ptr};
+			hook_invoke_dalloc(hook_dalloc_free, ptr, args_raw);
 			ifree(tsd, ptr, tcache, true);
 		}
 		check_entry_exit_locking(tsd_tsdn(tsd));
 	}
+}
+
+JEMALLOC_ALWAYS_INLINE
+bool free_fastpath(void *ptr, size_t size, bool size_hint) {
+	tsd_t *tsd = tsd_get(false);
+	if (unlikely(!tsd || !tsd_fast(tsd))) {
+		return false;
+	}
+
+	tcache_t *tcache = tsd_tcachep_get(tsd);
+
+	alloc_ctx_t alloc_ctx;
+	/* 
+	 * If !config_cache_oblivious, we can check PAGE alignment to
+	 * detect sampled objects.  Otherwise addresses are
+	 * randomized, and we have to look it up in the rtree anyway.
+	 * See also isfree().
+	 */
+	if (!size_hint || config_cache_oblivious) {
+		rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
+		bool res = rtree_szind_slab_read_fast(tsd_tsdn(tsd), &extents_rtree,
+						      rtree_ctx, (uintptr_t)ptr,
+						      &alloc_ctx.szind, &alloc_ctx.slab);
+		assert(alloc_ctx.szind != SC_NSIZES);
+
+		/* Note: profiled objects will have alloc_ctx.slab set */
+		if (!res || !alloc_ctx.slab) {
+			return false;
+		}
+	} else {
+		/*
+		 * Check for both sizes that are too large, and for sampled objects.
+		 * Sampled objects are always page-aligned.  The sampled object check
+		 * will also check for null ptr.
+		 */
+		if (size > SC_LOOKUP_MAXCLASS || (((uintptr_t)ptr & PAGE_MASK) == 0)) {
+			return false;
+		}
+		alloc_ctx.szind = sz_size2index_lookup(size);
+	}
+
+	if (unlikely(ticker_trytick(&tcache->gc_ticker))) {
+		return false;
+	}
+
+	cache_bin_t *bin = tcache_small_bin_get(tcache, alloc_ctx.szind);
+	cache_bin_info_t *bin_info = &tcache_bin_info[alloc_ctx.szind];
+	if (!cache_bin_dalloc_easy(bin, bin_info, ptr)) {
+		return false;
+	}
+
+	if (config_stats) {
+		size_t usize = sz_index2size(alloc_ctx.szind);
+		*tsd_thread_deallocatedp_get(tsd) += usize;
+	}
+
+	return true;
+}
+
+JEMALLOC_EXPORT void JEMALLOC_NOTHROW
+je_free(void *ptr) {
+	LOG("core.free.entry", "ptr: %p", ptr);
+
+	if (!free_fastpath(ptr, 0, false)) {
+		free_default(ptr);
+	}
+
 	LOG("core.free.exit", "");
 }
 
@@ -2427,7 +2813,6 @@ je_memalign(size_t alignment, size_t size) {
 	static_opts_init(&sopts);
 	dynamic_opts_init(&dopts);
 
-	sopts.bump_empty_alloc = true;
 	sopts.min_alignment = 1;
 	sopts.oom_string =
 	    "<jemalloc>: Error allocating aligned memory: out of memory\n";
@@ -2441,6 +2826,11 @@ je_memalign(size_t alignment, size_t size) {
 	dopts.alignment = alignment;
 
 	imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {alignment, size};
+		hook_invoke_alloc(hook_alloc_memalign, ret, (uintptr_t)ret,
+		    args);
+	}
 
 	LOG("core.memalign.exit", "result: %p", ret);
 	return ret;
@@ -2462,7 +2852,6 @@ je_valloc(size_t size) {
 	static_opts_init(&sopts);
 	dynamic_opts_init(&dopts);
 
-	sopts.bump_empty_alloc = true;
 	sopts.null_out_result_on_error = true;
 	sopts.min_alignment = PAGE;
 	sopts.oom_string =
@@ -2476,6 +2865,10 @@ je_valloc(size_t size) {
 	dopts.alignment = PAGE;
 
 	imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {size};
+		hook_invoke_alloc(hook_alloc_valloc, ret, (uintptr_t)ret, args);
+	}
 
 	LOG("core.valloc.exit", "result: %p\n", ret);
 	return ret;
@@ -2543,6 +2936,82 @@ int __posix_memalign(void** r, size_t a, size_t s) PREALIAS(je_posix_memalign);
  * Begin non-standard functions.
  */
 
+#ifdef JEMALLOC_EXPERIMENTAL_SMALLOCX_API
+
+#define JEMALLOC_SMALLOCX_CONCAT_HELPER(x, y) x ## y
+#define JEMALLOC_SMALLOCX_CONCAT_HELPER2(x, y)  \
+  JEMALLOC_SMALLOCX_CONCAT_HELPER(x, y)
+
+typedef struct {
+	void *ptr;
+	size_t size;
+} smallocx_return_t;
+
+JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
+smallocx_return_t JEMALLOC_NOTHROW
+/*
+ * The attribute JEMALLOC_ATTR(malloc) cannot be used due to:
+ *  - https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86488
+ */
+JEMALLOC_SMALLOCX_CONCAT_HELPER2(je_smallocx_, JEMALLOC_VERSION_GID_IDENT)
+  (size_t size, int flags) {
+	/*
+	 * Note: the attribute JEMALLOC_ALLOC_SIZE(1) cannot be
+	 * used here because it makes writing beyond the `size`
+	 * of the `ptr` undefined behavior, but the objective
+	 * of this function is to allow writing beyond `size`
+	 * up to `smallocx_return_t::size`.
+	 */
+	smallocx_return_t ret;
+	static_opts_t sopts;
+	dynamic_opts_t dopts;
+
+	LOG("core.smallocx.entry", "size: %zu, flags: %d", size, flags);
+
+	static_opts_init(&sopts);
+	dynamic_opts_init(&dopts);
+
+	sopts.assert_nonempty_alloc = true;
+	sopts.null_out_result_on_error = true;
+	sopts.oom_string = "<jemalloc>: Error in mallocx(): out of memory\n";
+	sopts.usize = true;
+
+	dopts.result = &ret.ptr;
+	dopts.num_items = 1;
+	dopts.item_size = size;
+	if (unlikely(flags != 0)) {
+		if ((flags & MALLOCX_LG_ALIGN_MASK) != 0) {
+			dopts.alignment = MALLOCX_ALIGN_GET_SPECIFIED(flags);
+		}
+
+		dopts.zero = MALLOCX_ZERO_GET(flags);
+
+		if ((flags & MALLOCX_TCACHE_MASK) != 0) {
+			if ((flags & MALLOCX_TCACHE_MASK)
+			    == MALLOCX_TCACHE_NONE) {
+				dopts.tcache_ind = TCACHE_IND_NONE;
+			} else {
+				dopts.tcache_ind = MALLOCX_TCACHE_GET(flags);
+			}
+		} else {
+			dopts.tcache_ind = TCACHE_IND_AUTOMATIC;
+		}
+
+		if ((flags & MALLOCX_ARENA_MASK) != 0)
+			dopts.arena_ind = MALLOCX_ARENA_GET(flags);
+	}
+
+	imalloc(&sopts, &dopts);
+	assert(dopts.usize == je_nallocx(size, flags));
+	ret.size = dopts.usize;
+
+	LOG("core.smallocx.exit", "result: %p, size: %zu", ret.ptr, ret.size);
+	return ret;
+}
+#undef JEMALLOC_SMALLOCX_CONCAT_HELPER
+#undef JEMALLOC_SMALLOCX_CONCAT_HELPER2
+#endif
+
 JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
 void JEMALLOC_NOTHROW *
 JEMALLOC_ATTR(malloc) JEMALLOC_ALLOC_SIZE(1)
@@ -2586,6 +3055,11 @@ je_mallocx(size_t size, int flags) {
 	}
 
 	imalloc(&sopts, &dopts);
+	if (sopts.slow) {
+		uintptr_t args[3] = {size, flags};
+		hook_invoke_alloc(hook_alloc_mallocx, ret, (uintptr_t)ret,
+		    args);
+	}
 
 	LOG("core.mallocx.exit", "result: %p", ret);
 	return ret;
@@ -2594,22 +3068,23 @@ je_mallocx(size_t size, int flags) {
 static void *
 irallocx_prof_sample(tsdn_t *tsdn, void *old_ptr, size_t old_usize,
     size_t usize, size_t alignment, bool zero, tcache_t *tcache, arena_t *arena,
-    prof_tctx_t *tctx) {
+    prof_tctx_t *tctx, hook_ralloc_args_t *hook_args) {
 	void *p;
 
 	if (tctx == NULL) {
 		return NULL;
 	}
-	if (usize <= SMALL_MAXCLASS) {
-		p = iralloct(tsdn, old_ptr, old_usize, LARGE_MINCLASS,
-		    alignment, zero, tcache, arena);
+	if (usize <= SC_SMALL_MAXCLASS) {
+		p = iralloct(tsdn, old_ptr, old_usize,
+		    SC_LARGE_MINCLASS, alignment, zero, tcache,
+		    arena, hook_args);
 		if (p == NULL) {
 			return NULL;
 		}
 		arena_prof_promote(tsdn, p, usize);
 	} else {
 		p = iralloct(tsdn, old_ptr, old_usize, usize, alignment, zero,
-		    tcache, arena);
+		    tcache, arena, hook_args);
 	}
 
 	return p;
@@ -2618,7 +3093,7 @@ irallocx_prof_sample(tsdn_t *tsdn, void *old_ptr, size_t old_usize,
 JEMALLOC_ALWAYS_INLINE void *
 irallocx_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t size,
     size_t alignment, size_t *usize, bool zero, tcache_t *tcache,
-    arena_t *arena, alloc_ctx_t *alloc_ctx) {
+    arena_t *arena, alloc_ctx_t *alloc_ctx, hook_ralloc_args_t *hook_args) {
 	void *p;
 	bool prof_active;
 	prof_tctx_t *old_tctx, *tctx;
@@ -2628,10 +3103,10 @@ irallocx_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t size,
 	tctx = prof_alloc_prep(tsd, *usize, prof_active, false);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
 		p = irallocx_prof_sample(tsd_tsdn(tsd), old_ptr, old_usize,
-		    *usize, alignment, zero, tcache, arena, tctx);
+		    *usize, alignment, zero, tcache, arena, tctx, hook_args);
 	} else {
 		p = iralloct(tsd_tsdn(tsd), old_ptr, old_usize, size, alignment,
-		    zero, tcache, arena);
+		    zero, tcache, arena, hook_args);
 	}
 	if (unlikely(p == NULL)) {
 		prof_alloc_rollback(tsd, tctx, false);
@@ -2702,23 +3177,27 @@ je_rallocx(void *ptr, size_t size, int flags) {
 	rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
 	rtree_szind_slab_read(tsd_tsdn(tsd), &extents_rtree, rtree_ctx,
 	    (uintptr_t)ptr, true, &alloc_ctx.szind, &alloc_ctx.slab);
-	assert(alloc_ctx.szind != NSIZES);
+	assert(alloc_ctx.szind != SC_NSIZES);
 	old_usize = sz_index2size(alloc_ctx.szind);
 	assert(old_usize == isalloc(tsd_tsdn(tsd), ptr));
+
+	hook_ralloc_args_t hook_args = {false, {(uintptr_t)ptr, size, flags,
+		0}};
 	if (config_prof && opt_prof) {
 		usize = (alignment == 0) ?
 		    sz_s2u(size) : sz_sa2u(size, alignment);
-		if (unlikely(usize == 0 || usize > LARGE_MAXCLASS)) {
+		if (unlikely(usize == 0
+		    || usize > SC_LARGE_MAXCLASS)) {
 			goto label_oom;
 		}
 		p = irallocx_prof(tsd, ptr, old_usize, size, alignment, &usize,
-		    zero, tcache, arena, &alloc_ctx);
+		    zero, tcache, arena, &alloc_ctx, &hook_args);
 		if (unlikely(p == NULL)) {
 			goto label_oom;
 		}
 	} else {
 		p = iralloct(tsd_tsdn(tsd), ptr, old_usize, size, alignment,
-		    zero, tcache, arena);
+		    zero, tcache, arena, &hook_args);
 		if (unlikely(p == NULL)) {
 			goto label_oom;
 		}
@@ -2752,14 +3231,14 @@ label_oom:
 JEMALLOC_ALWAYS_INLINE size_t
 ixallocx_helper(tsdn_t *tsdn, void *ptr, size_t old_usize, size_t size,
     size_t extra, size_t alignment, bool zero) {
-	size_t usize;
+	size_t newsize;
 
-	if (ixalloc(tsdn, ptr, old_usize, size, extra, alignment, zero)) {
+	if (ixalloc(tsdn, ptr, old_usize, size, extra, alignment, zero,
+	    &newsize)) {
 		return old_usize;
 	}
-	usize = isalloc(tsdn, ptr);
 
-	return usize;
+	return newsize;
 }
 
 static size_t
@@ -2793,17 +3272,19 @@ ixallocx_prof(tsd_t *tsd, void *ptr, size_t old_usize, size_t size,
 	 */
 	if (alignment == 0) {
 		usize_max = sz_s2u(size+extra);
-		assert(usize_max > 0 && usize_max <= LARGE_MAXCLASS);
+		assert(usize_max > 0
+		    && usize_max <= SC_LARGE_MAXCLASS);
 	} else {
 		usize_max = sz_sa2u(size+extra, alignment);
-		if (unlikely(usize_max == 0 || usize_max > LARGE_MAXCLASS)) {
+		if (unlikely(usize_max == 0
+		    || usize_max > SC_LARGE_MAXCLASS)) {
 			/*
 			 * usize_max is out of range, and chances are that
 			 * allocation will fail, but use the maximum possible
 			 * value and carry on with prof_alloc_prep(), just in
 			 * case allocation succeeds.
 			 */
-			usize_max = LARGE_MAXCLASS;
+			usize_max = SC_LARGE_MAXCLASS;
 		}
 	}
 	tctx = prof_alloc_prep(tsd, usize_max, prof_active, false);
@@ -2846,24 +3327,24 @@ je_xallocx(void *ptr, size_t size, size_t extra, int flags) {
 	rtree_ctx_t *rtree_ctx = tsd_rtree_ctx(tsd);
 	rtree_szind_slab_read(tsd_tsdn(tsd), &extents_rtree, rtree_ctx,
 	    (uintptr_t)ptr, true, &alloc_ctx.szind, &alloc_ctx.slab);
-	assert(alloc_ctx.szind != NSIZES);
+	assert(alloc_ctx.szind != SC_NSIZES);
 	old_usize = sz_index2size(alloc_ctx.szind);
 	assert(old_usize == isalloc(tsd_tsdn(tsd), ptr));
 	/*
 	 * The API explicitly absolves itself of protecting against (size +
 	 * extra) numerical overflow, but we may need to clamp extra to avoid
-	 * exceeding LARGE_MAXCLASS.
+	 * exceeding SC_LARGE_MAXCLASS.
 	 *
 	 * Ordinarily, size limit checking is handled deeper down, but here we
 	 * have to check as part of (size + extra) clamping, since we need the
 	 * clamped value in the above helper functions.
 	 */
-	if (unlikely(size > LARGE_MAXCLASS)) {
+	if (unlikely(size > SC_LARGE_MAXCLASS)) {
 		usize = old_usize;
 		goto label_not_resized;
 	}
-	if (unlikely(LARGE_MAXCLASS - size < extra)) {
-		extra = LARGE_MAXCLASS - size;
+	if (unlikely(SC_LARGE_MAXCLASS - size < extra)) {
+		extra = SC_LARGE_MAXCLASS - size;
 	}
 
 	if (config_prof && opt_prof) {
@@ -2882,6 +3363,12 @@ je_xallocx(void *ptr, size_t size, size_t extra, int flags) {
 		*tsd_thread_deallocatedp_get(tsd) += old_usize;
 	}
 label_not_resized:
+	if (unlikely(!tsd_fast(tsd))) {
+		uintptr_t args[4] = {(uintptr_t)ptr, size, extra, flags};
+		hook_invoke_expand(hook_expand_xallocx, ptr, old_usize,
+		    usize, (uintptr_t)usize, args);
+	}
+
 	UTRACE(ptr, size, ptr);
 	check_entry_exit_locking(tsd_tsdn(tsd));
 
@@ -2891,7 +3378,7 @@ label_not_resized:
 
 JEMALLOC_EXPORT size_t JEMALLOC_NOTHROW
 JEMALLOC_ATTR(pure)
-je_sallocx(const void *ptr, UNUSED int flags) {
+je_sallocx(const void *ptr, int flags) {
 	size_t usize;
 	tsdn_t *tsdn;
 
@@ -2954,6 +3441,8 @@ je_dallocx(void *ptr, int flags) {
 		tsd_assert_fast(tsd);
 		ifree(tsd, ptr, tcache, false);
 	} else {
+		uintptr_t args_raw[3] = {(uintptr_t)ptr, flags};
+		hook_invoke_dalloc(hook_dalloc_dallocx, ptr, args_raw);
 		ifree(tsd, ptr, tcache, true);
 	}
 	check_entry_exit_locking(tsd_tsdn(tsd));
@@ -2975,14 +3464,11 @@ inallocx(tsdn_t *tsdn, size_t size, int flags) {
 	return usize;
 }
 
-JEMALLOC_EXPORT void JEMALLOC_NOTHROW
-je_sdallocx(void *ptr, size_t size, int flags) {
+JEMALLOC_NOINLINE void
+sdallocx_default(void *ptr, size_t size, int flags) {
 	assert(ptr != NULL);
 	assert(malloc_initialized() || IS_INITIALIZER);
 
-	LOG("core.sdallocx.entry", "ptr: %p, size: %zu, flags: %d", ptr,
-	    size, flags);
-
 	tsd_t *tsd = tsd_fetch();
 	bool fast = tsd_fast(tsd);
 	size_t usize = inallocx(tsd_tsdn(tsd), size, flags);
@@ -3016,10 +3502,23 @@ je_sdallocx(void *ptr, size_t size, int flags) {
 		tsd_assert_fast(tsd);
 		isfree(tsd, ptr, usize, tcache, false);
 	} else {
+		uintptr_t args_raw[3] = {(uintptr_t)ptr, size, flags};
+		hook_invoke_dalloc(hook_dalloc_sdallocx, ptr, args_raw);
 		isfree(tsd, ptr, usize, tcache, true);
 	}
 	check_entry_exit_locking(tsd_tsdn(tsd));
 
+}
+
+JEMALLOC_EXPORT void JEMALLOC_NOTHROW
+je_sdallocx(void *ptr, size_t size, int flags) {
+	LOG("core.sdallocx.entry", "ptr: %p, size: %zu, flags: %d", ptr,
+		size, flags);
+
+	if (flags !=0 || !free_fastpath(ptr, size, true)) {
+		sdallocx_default(ptr, size, flags);
+	}
+
 	LOG("core.sdallocx.exit", "");
 }
 
@@ -3040,7 +3539,7 @@ je_nallocx(size_t size, int flags) {
 	check_entry_exit_locking(tsdn);
 
 	usize = inallocx(tsdn, size, flags);
-	if (unlikely(usize > LARGE_MAXCLASS)) {
+	if (unlikely(usize > SC_LARGE_MAXCLASS)) {
 		LOG("core.nallocx.exit", "result: %zu", ZU(0));
 		return 0;
 	}
@@ -3256,6 +3755,7 @@ _malloc_prefork(void)
 		}
 	}
 	prof_prefork1(tsd_tsdn(tsd));
+	tsd_prefork(tsd);
 }
 
 #ifndef JEMALLOC_MUTEX_INIT_CB
@@ -3278,6 +3778,8 @@ _malloc_postfork(void)
 
 	tsd = tsd_fetch();
 
+	tsd_postfork_parent(tsd);
+
 	witness_postfork_parent(tsd_witness_tsdp_get(tsd));
 	/* Release all mutexes, now that fork() has completed. */
 	for (i = 0, narenas = narenas_total_get(); i < narenas; i++) {
@@ -3305,6 +3807,8 @@ jemalloc_postfork_child(void) {
 
 	tsd = tsd_fetch();
 
+	tsd_postfork_child(tsd);
+
 	witness_postfork_child(tsd_witness_tsdp_get(tsd));
 	/* Release all mutexes, now that fork() has completed. */
 	for (i = 0, narenas = narenas_total_get(); i < narenas; i++) {
diff --git a/dep/jemalloc/src/large.c b/dep/jemalloc/src/large.c
index 27a2c679876..8e7a781d330 100644
--- a/dep/jemalloc/src/large.c
+++ b/dep/jemalloc/src/large.c
@@ -28,7 +28,7 @@ large_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
 	assert(!tsdn_null(tsdn) || arena != NULL);
 
 	ausize = sz_sa2u(usize, alignment);
-	if (unlikely(ausize == 0 || ausize > LARGE_MAXCLASS)) {
+	if (unlikely(ausize == 0 || ausize > SC_LARGE_MAXCLASS)) {
 		return NULL;
 	}
 
@@ -42,7 +42,7 @@ large_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
 	 */
 	is_zeroed = zero;
 	if (likely(!tsdn_null(tsdn))) {
-		arena = arena_choose(tsdn_tsd(tsdn), arena);
+		arena = arena_choose_maybe_huge(tsdn_tsd(tsdn), arena, usize);
 	}
 	if (unlikely(arena == NULL) || (extent = arena_extent_alloc_large(tsdn,
 	    arena, usize, alignment, &is_zeroed)) == NULL) {
@@ -109,7 +109,7 @@ large_ralloc_no_move_shrink(tsdn_t *tsdn, extent_t *extent, size_t usize) {
 	if (diff != 0) {
 		extent_t *trail = extent_split_wrapper(tsdn, arena,
 		    &extent_hooks, extent, usize + sz_large_pad,
-		    sz_size2index(usize), false, diff, NSIZES, false);
+		    sz_size2index(usize), false, diff, SC_NSIZES, false);
 		if (trail == NULL) {
 			return true;
 		}
@@ -154,17 +154,17 @@ large_ralloc_no_move_expand(tsdn_t *tsdn, extent_t *extent, size_t usize,
 	bool new_mapping;
 	if ((trail = extents_alloc(tsdn, arena, &extent_hooks,
 	    &arena->extents_dirty, extent_past_get(extent), trailsize, 0,
-	    CACHELINE, false, NSIZES, &is_zeroed_trail, &commit)) != NULL
+	    CACHELINE, false, SC_NSIZES, &is_zeroed_trail, &commit)) != NULL
 	    || (trail = extents_alloc(tsdn, arena, &extent_hooks,
 	    &arena->extents_muzzy, extent_past_get(extent), trailsize, 0,
-	    CACHELINE, false, NSIZES, &is_zeroed_trail, &commit)) != NULL) {
+	    CACHELINE, false, SC_NSIZES, &is_zeroed_trail, &commit)) != NULL) {
 		if (config_stats) {
 			new_mapping = false;
 		}
 	} else {
 		if ((trail = extent_alloc_wrapper(tsdn, arena, &extent_hooks,
 		    extent_past_get(extent), trailsize, 0, CACHELINE, false,
-		    NSIZES, &is_zeroed_trail, &commit)) == NULL) {
+		    SC_NSIZES, &is_zeroed_trail, &commit)) == NULL) {
 			return true;
 		}
 		if (config_stats) {
@@ -221,9 +221,10 @@ large_ralloc_no_move(tsdn_t *tsdn, extent_t *extent, size_t usize_min,
 	size_t oldusize = extent_usize_get(extent);
 
 	/* The following should have been caught by callers. */
-	assert(usize_min > 0 && usize_max <= LARGE_MAXCLASS);
+	assert(usize_min > 0 && usize_max <= SC_LARGE_MAXCLASS);
 	/* Both allocation sizes must be large to avoid a move. */
-	assert(oldusize >= LARGE_MINCLASS && usize_max >= LARGE_MINCLASS);
+	assert(oldusize >= SC_LARGE_MINCLASS
+	    && usize_max >= SC_LARGE_MINCLASS);
 
 	if (usize_max > oldusize) {
 		/* Attempt to expand the allocation in-place. */
@@ -270,17 +271,23 @@ large_ralloc_move_helper(tsdn_t *tsdn, arena_t *arena, size_t usize,
 }
 
 void *
-large_ralloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent, size_t usize,
-    size_t alignment, bool zero, tcache_t *tcache) {
-	size_t oldusize = extent_usize_get(extent);
+large_ralloc(tsdn_t *tsdn, arena_t *arena, void *ptr, size_t usize,
+    size_t alignment, bool zero, tcache_t *tcache,
+    hook_ralloc_args_t *hook_args) {
+	extent_t *extent = iealloc(tsdn, ptr);
 
+	size_t oldusize = extent_usize_get(extent);
 	/* The following should have been caught by callers. */
-	assert(usize > 0 && usize <= LARGE_MAXCLASS);
+	assert(usize > 0 && usize <= SC_LARGE_MAXCLASS);
 	/* Both allocation sizes must be large to avoid a move. */
-	assert(oldusize >= LARGE_MINCLASS && usize >= LARGE_MINCLASS);
+	assert(oldusize >= SC_LARGE_MINCLASS
+	    && usize >= SC_LARGE_MINCLASS);
 
 	/* Try to avoid moving the allocation. */
 	if (!large_ralloc_no_move(tsdn, extent, usize, usize, zero)) {
+		hook_invoke_expand(hook_args->is_realloc
+		    ? hook_expand_realloc : hook_expand_rallocx, ptr, oldusize,
+		    usize, (uintptr_t)ptr, hook_args->args);
 		return extent_addr_get(extent);
 	}
 
@@ -295,6 +302,12 @@ large_ralloc(tsdn_t *tsdn, arena_t *arena, extent_t *extent, size_t usize,
 		return NULL;
 	}
 
+	hook_invoke_alloc(hook_args->is_realloc
+	    ? hook_alloc_realloc : hook_alloc_rallocx, ret, (uintptr_t)ret,
+	    hook_args->args);
+	hook_invoke_dalloc(hook_args->is_realloc
+	    ? hook_dalloc_realloc : hook_dalloc_rallocx, ptr, hook_args->args);
+
 	size_t copysize = (usize < oldusize) ? usize : oldusize;
 	memcpy(ret, extent_addr_get(extent), copysize);
 	isdalloct(tsdn, extent_addr_get(extent), oldusize, tcache, NULL, true);
@@ -318,8 +331,9 @@ large_dalloc_prep_impl(tsdn_t *tsdn, arena_t *arena, extent_t *extent,
 		large_dalloc_maybe_junk(extent_addr_get(extent),
 		    extent_usize_get(extent));
 	} else {
-		malloc_mutex_assert_owner(tsdn, &arena->large_mtx);
+		/* Only hold the large_mtx if necessary. */
 		if (!arena_is_auto(arena)) {
+			malloc_mutex_assert_owner(tsdn, &arena->large_mtx);
 			extent_list_remove(&arena->large, extent);
 		}
 	}
@@ -369,3 +383,13 @@ void
 large_prof_tctx_reset(tsdn_t *tsdn, extent_t *extent) {
 	large_prof_tctx_set(tsdn, extent, (prof_tctx_t *)(uintptr_t)1U);
 }
+
+nstime_t
+large_prof_alloc_time_get(const extent_t *extent) {
+	return extent_prof_alloc_time_get(extent);
+}
+
+void
+large_prof_alloc_time_set(extent_t *extent, nstime_t t) {
+	extent_prof_alloc_time_set(extent, t);
+}
diff --git a/dep/jemalloc/src/mutex.c b/dep/jemalloc/src/mutex.c
index 30222b3e582..3f920f5b1cd 100644
--- a/dep/jemalloc/src/mutex.c
+++ b/dep/jemalloc/src/mutex.c
@@ -46,7 +46,7 @@ JEMALLOC_EXPORT int	_pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex,
 void
 malloc_mutex_lock_slow(malloc_mutex_t *mutex) {
 	mutex_prof_data_t *data = &mutex->prof_data;
-	UNUSED nstime_t before = NSTIME_ZERO_INITIALIZER;
+	nstime_t before = NSTIME_ZERO_INITIALIZER;
 
 	if (ncpus == 1) {
 		goto label_spin_done;
@@ -55,7 +55,8 @@ malloc_mutex_lock_slow(malloc_mutex_t *mutex) {
 	int cnt = 0, max_cnt = MALLOC_MUTEX_MAX_SPIN;
 	do {
 		spin_cpu_spinwait();
-		if (!malloc_mutex_trylock_final(mutex)) {
+		if (!atomic_load_b(&mutex->locked, ATOMIC_RELAXED)
+                    && !malloc_mutex_trylock_final(mutex)) {
 			data->n_spin_acquired++;
 			return;
 		}
@@ -144,9 +145,7 @@ malloc_mutex_init(malloc_mutex_t *mutex, const char *name,
 	}
 #  endif
 #elif (defined(JEMALLOC_OS_UNFAIR_LOCK))
-	mutex->lock = OS_UNFAIR_LOCK_INIT;
-#elif (defined(JEMALLOC_OSSPIN))
-	mutex->lock = 0;
+       mutex->lock = OS_UNFAIR_LOCK_INIT;
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
 	if (postpone_init) {
 		mutex->postponed_next = postponed_mutexes;
diff --git a/dep/jemalloc/src/pages.c b/dep/jemalloc/src/pages.c
index 26002692d60..13de27a0087 100644
--- a/dep/jemalloc/src/pages.c
+++ b/dep/jemalloc/src/pages.c
@@ -180,6 +180,35 @@ pages_map(void *addr, size_t size, size_t alignment, bool *commit) {
 	assert(alignment >= PAGE);
 	assert(ALIGNMENT_ADDR2BASE(addr, alignment) == addr);
 
+#if defined(__FreeBSD__) && defined(MAP_EXCL)
+	/*
+	 * FreeBSD has mechanisms both to mmap at specific address without
+	 * touching existing mappings, and to mmap with specific alignment.
+	 */
+	{
+		if (os_overcommits) {
+			*commit = true;
+		}
+
+		int prot = *commit ? PAGES_PROT_COMMIT : PAGES_PROT_DECOMMIT;
+		int flags = mmap_flags;
+
+		if (addr != NULL) {
+			flags |= MAP_FIXED | MAP_EXCL;
+		} else {
+			unsigned alignment_bits = ffs_zu(alignment);
+			assert(alignment_bits > 1);
+			flags |= MAP_ALIGNED(alignment_bits - 1);
+		}
+
+		void *ret = mmap(addr, size, prot, flags, -1, 0);
+		if (ret == MAP_FAILED) {
+			ret = NULL;
+		}
+
+		return ret;
+	}
+#endif
 	/*
 	 * Ideally, there would be a way to specify alignment to mmap() (like
 	 * NetBSD has), but in the absence of such a feature, we have to work
@@ -261,7 +290,7 @@ pages_decommit(void *addr, size_t size) {
 
 bool
 pages_purge_lazy(void *addr, size_t size) {
-	assert(PAGE_ADDR2BASE(addr) == addr);
+	assert(ALIGNMENT_ADDR2BASE(addr, os_page) == addr);
 	assert(PAGE_CEILING(size) == size);
 
 	if (!pages_can_purge_lazy) {
@@ -391,6 +420,10 @@ os_page_detect(void) {
 	GetSystemInfo(&si);
 	return si.dwPageSize;
 #elif defined(__FreeBSD__)
+	/*
+	 * This returns the value obtained from
+	 * the auxv vector, avoiding a syscall.
+	 */
 	return getpagesize();
 #else
 	long result = sysconf(_SC_PAGESIZE);
@@ -544,6 +577,10 @@ init_thp_state(void) {
 	close(fd);
 #endif
 
+        if (nread < 0) {
+		goto label_error; 
+        }
+
 	if (strncmp(buf, sys_state_madvise, (size_t)nread) == 0) {
 		init_system_thp_mode = thp_mode_default;
 	} else if (strncmp(buf, sys_state_always, (size_t)nread) == 0) {
@@ -588,6 +625,11 @@ pages_boot(void) {
 
 	init_thp_state();
 
+#ifdef __FreeBSD__
+	/*
+	 * FreeBSD doesn't need the check; madvise(2) is known to work.
+	 */
+#else
 	/* Detect lazy purge runtime support. */
 	if (pages_can_purge_lazy) {
 		bool committed = false;
@@ -601,6 +643,7 @@ pages_boot(void) {
 		}
 		os_pages_unmap(madv_free_page, PAGE);
 	}
+#endif
 
 	return false;
 }
diff --git a/dep/jemalloc/src/prof.c b/dep/jemalloc/src/prof.c
index 13df641a030..4d7d65db40c 100644
--- a/dep/jemalloc/src/prof.c
+++ b/dep/jemalloc/src/prof.c
@@ -7,6 +7,7 @@
 #include "jemalloc/internal/hash.h"
 #include "jemalloc/internal/malloc_io.h"
 #include "jemalloc/internal/mutex.h"
+#include "jemalloc/internal/emitter.h"
 
 /******************************************************************************/
 
@@ -23,7 +24,7 @@
  */
 #undef _Unwind_Backtrace
 #include <unwind.h>
-#define _Unwind_Backtrace JEMALLOC_HOOK(_Unwind_Backtrace, hooks_libc_hook)
+#define _Unwind_Backtrace JEMALLOC_HOOK(_Unwind_Backtrace, test_hooks_libc_hook)
 #endif
 
 /******************************************************************************/
@@ -38,6 +39,7 @@ bool		opt_prof_gdump = false;
 bool		opt_prof_final = false;
 bool		opt_prof_leak = false;
 bool		opt_prof_accum = false;
+bool		opt_prof_log = false;
 char		opt_prof_prefix[
     /* Minimize memory bloat for non-prof builds. */
 #ifdef JEMALLOC_PROF
@@ -70,6 +72,100 @@ uint64_t	prof_interval = 0;
 
 size_t		lg_prof_sample;
 
+typedef enum prof_logging_state_e prof_logging_state_t;
+enum prof_logging_state_e {
+	prof_logging_state_stopped,
+	prof_logging_state_started,
+	prof_logging_state_dumping
+};
+
+/*
+ * - stopped: log_start never called, or previous log_stop has completed.
+ * - started: log_start called, log_stop not called yet. Allocations are logged.
+ * - dumping: log_stop called but not finished; samples are not logged anymore.
+ */
+prof_logging_state_t prof_logging_state = prof_logging_state_stopped;
+
+#ifdef JEMALLOC_JET
+static bool prof_log_dummy = false;
+#endif
+
+/* Incremented for every log file that is output. */
+static uint64_t log_seq = 0;
+static char log_filename[
+    /* Minimize memory bloat for non-prof builds. */
+#ifdef JEMALLOC_PROF
+    PATH_MAX +
+#endif
+    1];
+
+/* Timestamp for most recent call to log_start(). */
+static nstime_t log_start_timestamp = NSTIME_ZERO_INITIALIZER;
+
+/* Increment these when adding to the log_bt and log_thr linked lists. */
+static size_t log_bt_index = 0;
+static size_t log_thr_index = 0;
+
+/* Linked list node definitions. These are only used in prof.c. */
+typedef struct prof_bt_node_s prof_bt_node_t;
+
+struct prof_bt_node_s {
+	prof_bt_node_t *next;
+	size_t index;
+	prof_bt_t bt;
+	/* Variable size backtrace vector pointed to by bt. */
+	void *vec[1];
+};
+
+typedef struct prof_thr_node_s prof_thr_node_t;
+
+struct prof_thr_node_s {
+	prof_thr_node_t *next;
+	size_t index;
+	uint64_t thr_uid;
+	/* Variable size based on thr_name_sz. */
+	char name[1];
+}; 
+
+typedef struct prof_alloc_node_s prof_alloc_node_t;
+
+/* This is output when logging sampled allocations. */
+struct prof_alloc_node_s {
+	prof_alloc_node_t *next;
+	/* Indices into an array of thread data. */
+	size_t alloc_thr_ind;
+	size_t free_thr_ind;
+
+	/* Indices into an array of backtraces. */
+	size_t alloc_bt_ind;
+	size_t free_bt_ind;
+
+	uint64_t alloc_time_ns;
+	uint64_t free_time_ns;
+
+	size_t usize;
+};
+
+/*
+ * Created on the first call to prof_log_start and deleted on prof_log_stop.
+ * These are the backtraces and threads that have already been logged by an
+ * allocation.
+ */
+static bool log_tables_initialized = false;
+static ckh_t log_bt_node_set;
+static ckh_t log_thr_node_set;
+
+/* Store linked lists for logged data. */
+static prof_bt_node_t *log_bt_first = NULL;
+static prof_bt_node_t *log_bt_last = NULL;
+static prof_thr_node_t *log_thr_first = NULL;
+static prof_thr_node_t *log_thr_last = NULL;
+static prof_alloc_node_t *log_alloc_first = NULL;
+static prof_alloc_node_t *log_alloc_last = NULL;
+
+/* Protects the prof_logging_state and any log_{...} variable. */
+static malloc_mutex_t log_mtx;
+
 /*
  * Table of mutexes that are shared among gctx's.  These are leaf locks, so
  * there is no problem with using them for more than one gctx at the same time.
@@ -145,6 +241,12 @@ static void	prof_tdata_destroy(tsd_t *tsd, prof_tdata_t *tdata,
     bool even_if_attached);
 static char	*prof_thread_name_alloc(tsdn_t *tsdn, const char *thread_name);
 
+/* Hashtable functions for log_bt_node_set and log_thr_node_set. */
+static void prof_thr_node_hash(const void *key, size_t r_hash[2]);
+static bool prof_thr_node_keycomp(const void *k1, const void *k2);
+static void prof_bt_node_hash(const void *key, size_t r_hash[2]);
+static bool prof_bt_node_keycomp(const void *k1, const void *k2);
+
 /******************************************************************************/
 /* Red-black trees. */
 
@@ -242,6 +344,12 @@ prof_malloc_sample_object(tsdn_t *tsdn, const void *ptr, size_t usize,
     prof_tctx_t *tctx) {
 	prof_tctx_set(tsdn, ptr, usize, NULL, tctx);
 
+	/* Get the current time and set this in the extent_t. We'll read this
+	 * when free() is called. */
+	nstime_t t = NSTIME_ZERO_INITIALIZER;
+	nstime_update(&t);
+	prof_alloc_time_set(tsdn, ptr, NULL, t);
+
 	malloc_mutex_lock(tsdn, tctx->tdata->lock);
 	tctx->cnts.curobjs++;
 	tctx->cnts.curbytes += usize;
@@ -253,14 +361,174 @@ prof_malloc_sample_object(tsdn_t *tsdn, const void *ptr, size_t usize,
 	malloc_mutex_unlock(tsdn, tctx->tdata->lock);
 }
 
+static size_t
+prof_log_bt_index(tsd_t *tsd, prof_bt_t *bt) {
+	assert(prof_logging_state == prof_logging_state_started);
+	malloc_mutex_assert_owner(tsd_tsdn(tsd), &log_mtx);
+
+	prof_bt_node_t dummy_node;
+	dummy_node.bt = *bt;
+	prof_bt_node_t *node;
+
+	/* See if this backtrace is already cached in the table. */
+	if (ckh_search(&log_bt_node_set, (void *)(&dummy_node),
+	    (void **)(&node), NULL)) {
+		size_t sz = offsetof(prof_bt_node_t, vec) +
+			        (bt->len * sizeof(void *));
+		prof_bt_node_t *new_node = (prof_bt_node_t *)
+		    iallocztm(tsd_tsdn(tsd), sz, sz_size2index(sz), false, NULL,
+		    true, arena_get(TSDN_NULL, 0, true), true);
+		if (log_bt_first == NULL) {
+			log_bt_first = new_node;
+			log_bt_last = new_node;
+		} else {
+			log_bt_last->next = new_node;
+			log_bt_last = new_node;
+		}
+
+		new_node->next = NULL;
+		new_node->index = log_bt_index;
+		/* 
+		 * Copy the backtrace: bt is inside a tdata or gctx, which
+		 * might die before prof_log_stop is called.
+		 */
+		new_node->bt.len = bt->len;
+		memcpy(new_node->vec, bt->vec, bt->len * sizeof(void *));
+		new_node->bt.vec = new_node->vec;
+
+		log_bt_index++;
+		ckh_insert(tsd, &log_bt_node_set, (void *)new_node, NULL);
+		return new_node->index;
+	} else {
+		return node->index;
+	}
+} 
+static size_t
+prof_log_thr_index(tsd_t *tsd, uint64_t thr_uid, const char *name) {
+	assert(prof_logging_state == prof_logging_state_started);
+	malloc_mutex_assert_owner(tsd_tsdn(tsd), &log_mtx);
+
+	prof_thr_node_t dummy_node;
+	dummy_node.thr_uid = thr_uid;
+	prof_thr_node_t *node;
+
+	/* See if this thread is already cached in the table. */
+	if (ckh_search(&log_thr_node_set, (void *)(&dummy_node),
+	    (void **)(&node), NULL)) {
+		size_t sz = offsetof(prof_thr_node_t, name) + strlen(name) + 1;
+		prof_thr_node_t *new_node = (prof_thr_node_t *)
+		    iallocztm(tsd_tsdn(tsd), sz, sz_size2index(sz), false, NULL,
+		    true, arena_get(TSDN_NULL, 0, true), true);
+		if (log_thr_first == NULL) {
+			log_thr_first = new_node;
+			log_thr_last = new_node;
+		} else {
+			log_thr_last->next = new_node;
+			log_thr_last = new_node;
+		}
+
+		new_node->next = NULL;
+		new_node->index = log_thr_index;
+		new_node->thr_uid = thr_uid;
+		strcpy(new_node->name, name);
+
+		log_thr_index++;
+		ckh_insert(tsd, &log_thr_node_set, (void *)new_node, NULL);
+		return new_node->index;
+	} else {
+		return node->index;
+	}
+}
+
+static void
+prof_try_log(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx) {
+	malloc_mutex_assert_owner(tsd_tsdn(tsd), tctx->tdata->lock);
+
+	prof_tdata_t *cons_tdata = prof_tdata_get(tsd, false);
+	if (cons_tdata == NULL) {
+		/*
+		 * We decide not to log these allocations. cons_tdata will be
+		 * NULL only when the current thread is in a weird state (e.g.
+		 * it's being destroyed).
+		 */
+		return;
+	}	
+
+	malloc_mutex_lock(tsd_tsdn(tsd), &log_mtx);
+
+	if (prof_logging_state != prof_logging_state_started) {
+		goto label_done;
+	}
+
+	if (!log_tables_initialized) {
+		bool err1 = ckh_new(tsd, &log_bt_node_set, PROF_CKH_MINITEMS,
+				prof_bt_node_hash, prof_bt_node_keycomp);
+		bool err2 = ckh_new(tsd, &log_thr_node_set, PROF_CKH_MINITEMS,
+				prof_thr_node_hash, prof_thr_node_keycomp);
+		if (err1 || err2) {
+			goto label_done;
+		}
+		log_tables_initialized = true;
+	}
+
+	nstime_t alloc_time = prof_alloc_time_get(tsd_tsdn(tsd), ptr,
+			          (alloc_ctx_t *)NULL);
+	nstime_t free_time = NSTIME_ZERO_INITIALIZER;
+	nstime_update(&free_time);
+
+	size_t sz = sizeof(prof_alloc_node_t);
+	prof_alloc_node_t *new_node = (prof_alloc_node_t *)
+	    iallocztm(tsd_tsdn(tsd), sz, sz_size2index(sz), false, NULL, true,
+	    arena_get(TSDN_NULL, 0, true), true);
+
+	const char *prod_thr_name = (tctx->tdata->thread_name == NULL)?
+				        "" : tctx->tdata->thread_name;
+	const char *cons_thr_name = prof_thread_name_get(tsd);
+
+	prof_bt_t bt;
+	/* Initialize the backtrace, using the buffer in tdata to store it. */
+	bt_init(&bt, cons_tdata->vec);
+	prof_backtrace(&bt);
+	prof_bt_t *cons_bt = &bt;
+
+	/* We haven't destroyed tctx yet, so gctx should be good to read. */
+	prof_bt_t *prod_bt = &tctx->gctx->bt;
+
+	new_node->next = NULL;
+	new_node->alloc_thr_ind = prof_log_thr_index(tsd, tctx->tdata->thr_uid,
+				      prod_thr_name);
+	new_node->free_thr_ind = prof_log_thr_index(tsd, cons_tdata->thr_uid,
+				     cons_thr_name);
+	new_node->alloc_bt_ind = prof_log_bt_index(tsd, prod_bt);
+	new_node->free_bt_ind = prof_log_bt_index(tsd, cons_bt);
+	new_node->alloc_time_ns = nstime_ns(&alloc_time);
+	new_node->free_time_ns = nstime_ns(&free_time);
+	new_node->usize = usize;
+
+	if (log_alloc_first == NULL) {
+		log_alloc_first = new_node;
+		log_alloc_last = new_node;
+	} else {
+		log_alloc_last->next = new_node;
+		log_alloc_last = new_node;
+	}
+
+label_done:
+	malloc_mutex_unlock(tsd_tsdn(tsd), &log_mtx);	
+}
+
 void
-prof_free_sampled_object(tsd_t *tsd, size_t usize, prof_tctx_t *tctx) {
+prof_free_sampled_object(tsd_t *tsd, const void *ptr, size_t usize, 
+    prof_tctx_t *tctx) {
 	malloc_mutex_lock(tsd_tsdn(tsd), tctx->tdata->lock);
+
 	assert(tctx->cnts.curobjs > 0);
 	assert(tctx->cnts.curbytes >= usize);
 	tctx->cnts.curobjs--;
 	tctx->cnts.curbytes -= usize;
 
+	prof_try_log(tsd, ptr, usize, tctx);
+
 	if (prof_tctx_should_destroy(tsd_tsdn(tsd), tctx)) {
 		prof_tctx_destroy(tsd, tctx);
 	} else {
@@ -871,15 +1139,12 @@ prof_lookup(tsd_t *tsd, prof_bt_t *bt) {
 void
 prof_sample_threshold_update(prof_tdata_t *tdata) {
 #ifdef JEMALLOC_PROF
-	uint64_t r;
-	double u;
-
 	if (!config_prof) {
 		return;
 	}
 
 	if (lg_prof_sample == 0) {
-		tdata->bytes_until_sample = 0;
+		tsd_bytes_until_sample_set(tsd_fetch(), 0);
 		return;
 	}
 
@@ -901,11 +1166,16 @@ prof_sample_threshold_update(prof_tdata_t *tdata) {
 	 *   pp 500
 	 *   (http://luc.devroye.org/rnbookindex.html)
 	 */
-	r = prng_lg_range_u64(&tdata->prng_state, 53);
-	u = (double)r * (1.0/9007199254740992.0L);
-	tdata->bytes_until_sample = (uint64_t)(log(u) /
+	uint64_t r = prng_lg_range_u64(&tdata->prng_state, 53);
+	double u = (double)r * (1.0/9007199254740992.0L);
+	uint64_t bytes_until_sample = (uint64_t)(log(u) /
 	    log(1.0 - (1.0 / (double)((uint64_t)1U << lg_prof_sample))))
 	    + (uint64_t)1U;
+	if (bytes_until_sample > SSIZE_MAX) {
+		bytes_until_sample = SSIZE_MAX;
+	}
+	tsd_bytes_until_sample_set(tsd_fetch(), bytes_until_sample);
+
 #endif
 }
 
@@ -1887,6 +2157,33 @@ prof_bt_keycomp(const void *k1, const void *k2) {
 	return (memcmp(bt1->vec, bt2->vec, bt1->len * sizeof(void *)) == 0);
 }
 
+static void
+prof_bt_node_hash(const void *key, size_t r_hash[2]) {
+	const prof_bt_node_t *bt_node = (prof_bt_node_t *)key;
+	prof_bt_hash((void *)(&bt_node->bt), r_hash);
+}
+
+static bool
+prof_bt_node_keycomp(const void *k1, const void *k2) {
+	const prof_bt_node_t *bt_node1 = (prof_bt_node_t *)k1;
+	const prof_bt_node_t *bt_node2 = (prof_bt_node_t *)k2;
+	return prof_bt_keycomp((void *)(&bt_node1->bt),
+	    (void *)(&bt_node2->bt));
+}
+
+static void
+prof_thr_node_hash(const void *key, size_t r_hash[2]) {
+	const prof_thr_node_t *thr_node = (prof_thr_node_t *)key;
+	hash(&thr_node->thr_uid, sizeof(uint64_t), 0x94122f35U, r_hash);
+}
+
+static bool
+prof_thr_node_keycomp(const void *k1, const void *k2) {
+	const prof_thr_node_t *thr_node1 = (prof_thr_node_t *)k1;
+	const prof_thr_node_t *thr_node2 = (prof_thr_node_t *)k2;
+	return thr_node1->thr_uid == thr_node2->thr_uid;
+}
+
 static uint64_t
 prof_thr_uid_alloc(tsdn_t *tsdn) {
 	uint64_t thr_uid;
@@ -2119,6 +2416,368 @@ prof_active_set(tsdn_t *tsdn, bool active) {
 	return prof_active_old;
 }
 
+#ifdef JEMALLOC_JET
+size_t
+prof_log_bt_count(void) {
+	size_t cnt = 0;
+	prof_bt_node_t *node = log_bt_first;
+	while (node != NULL) {
+		cnt++;
+		node = node->next;
+	}
+	return cnt;
+}
+
+size_t
+prof_log_alloc_count(void) {
+	size_t cnt = 0;
+	prof_alloc_node_t *node = log_alloc_first;
+	while (node != NULL) {
+		cnt++;
+		node = node->next;
+	}
+	return cnt;
+}
+
+size_t
+prof_log_thr_count(void) {
+	size_t cnt = 0;
+	prof_thr_node_t *node = log_thr_first;
+	while (node != NULL) {
+		cnt++;
+		node = node->next;
+	}
+	return cnt;
+}
+
+bool
+prof_log_is_logging(void) {
+	return prof_logging_state == prof_logging_state_started;
+}
+
+bool
+prof_log_rep_check(void) {
+	if (prof_logging_state == prof_logging_state_stopped
+	    && log_tables_initialized) {
+		return true;
+	}
+
+	if (log_bt_last != NULL && log_bt_last->next != NULL) {
+		return true;
+	}
+	if (log_thr_last != NULL && log_thr_last->next != NULL) {
+		return true;
+	}
+	if (log_alloc_last != NULL && log_alloc_last->next != NULL) {
+		return true;
+	}
+
+	size_t bt_count = prof_log_bt_count();
+	size_t thr_count = prof_log_thr_count();
+	size_t alloc_count = prof_log_alloc_count();
+
+
+	if (prof_logging_state == prof_logging_state_stopped) {
+		if (bt_count != 0 || thr_count != 0 || alloc_count || 0) {
+			return true;
+		}
+	}
+
+	prof_alloc_node_t *node = log_alloc_first;
+	while (node != NULL) {
+		if (node->alloc_bt_ind >= bt_count) {
+			return true;
+		}
+		if (node->free_bt_ind >= bt_count) {
+			return true;
+		}
+		if (node->alloc_thr_ind >= thr_count) {
+			return true;
+		}
+		if (node->free_thr_ind >= thr_count) {
+			return true;
+		}
+		if (node->alloc_time_ns > node->free_time_ns) {
+			return true;
+		}
+		node = node->next;
+	}
+
+	return false;
+}
+
+void
+prof_log_dummy_set(bool new_value) {
+	prof_log_dummy = new_value;
+}
+#endif
+
+bool
+prof_log_start(tsdn_t *tsdn, const char *filename) {
+	if (!opt_prof || !prof_booted) {
+		return true;
+	}
+
+	bool ret = false;
+	size_t buf_size = PATH_MAX + 1;
+
+	malloc_mutex_lock(tsdn, &log_mtx);
+
+	if (prof_logging_state != prof_logging_state_stopped) {
+		ret = true;
+	} else if (filename == NULL) {
+		/* Make default name. */
+		malloc_snprintf(log_filename, buf_size, "%s.%d.%"FMTu64".json",
+		    opt_prof_prefix, prof_getpid(), log_seq);
+		log_seq++;
+		prof_logging_state = prof_logging_state_started;
+	} else if (strlen(filename) >= buf_size) {
+		ret = true;
+	} else {
+		strcpy(log_filename, filename);
+		prof_logging_state = prof_logging_state_started;
+	}
+
+	if (!ret) {
+		nstime_update(&log_start_timestamp);
+	}
+
+	malloc_mutex_unlock(tsdn, &log_mtx);
+
+	return ret;
+}
+
+/* Used as an atexit function to stop logging on exit. */
+static void
+prof_log_stop_final(void) {
+	tsd_t *tsd = tsd_fetch();
+	prof_log_stop(tsd_tsdn(tsd));
+}
+
+struct prof_emitter_cb_arg_s {
+	int fd;
+	ssize_t ret;
+};
+
+static void
+prof_emitter_write_cb(void *opaque, const char *to_write) {
+	struct prof_emitter_cb_arg_s *arg =
+	    (struct prof_emitter_cb_arg_s *)opaque;
+	size_t bytes = strlen(to_write);
+#ifdef JEMALLOC_JET
+	if (prof_log_dummy) {
+		return;
+	}
+#endif
+	arg->ret = write(arg->fd, (void *)to_write, bytes);
+}
+
+/*
+ * prof_log_emit_{...} goes through the appropriate linked list, emitting each
+ * node to the json and deallocating it.
+ */
+static void
+prof_log_emit_threads(tsd_t *tsd, emitter_t *emitter) {
+	emitter_json_array_kv_begin(emitter, "threads");
+	prof_thr_node_t *thr_node = log_thr_first;
+	prof_thr_node_t *thr_old_node;
+	while (thr_node != NULL) {
+		emitter_json_object_begin(emitter);
+
+		emitter_json_kv(emitter, "thr_uid", emitter_type_uint64,
+		    &thr_node->thr_uid);
+
+		char *thr_name = thr_node->name;
+
+		emitter_json_kv(emitter, "thr_name", emitter_type_string,
+		    &thr_name);
+
+		emitter_json_object_end(emitter);
+		thr_old_node = thr_node;
+		thr_node = thr_node->next;
+		idalloc(tsd, thr_old_node);
+	}
+	emitter_json_array_end(emitter);
+}
+
+static void
+prof_log_emit_traces(tsd_t *tsd, emitter_t *emitter) {
+	emitter_json_array_kv_begin(emitter, "stack_traces");
+	prof_bt_node_t *bt_node = log_bt_first;
+	prof_bt_node_t *bt_old_node; 
+	/* 
+	 * Calculate how many hex digits we need: twice number of bytes, two for
+	 * "0x", and then one more for terminating '\0'.
+	 */
+	char buf[2 * sizeof(intptr_t) + 3];
+	size_t buf_sz = sizeof(buf);
+	while (bt_node != NULL) {
+		emitter_json_array_begin(emitter);
+		size_t i;
+		for (i = 0; i < bt_node->bt.len; i++) {
+			malloc_snprintf(buf, buf_sz, "%p", bt_node->bt.vec[i]);
+			char *trace_str = buf;
+			emitter_json_value(emitter, emitter_type_string,
+			    &trace_str);
+		}
+		emitter_json_array_end(emitter);
+
+		bt_old_node = bt_node;
+		bt_node = bt_node->next;
+		idalloc(tsd, bt_old_node);
+	}
+	emitter_json_array_end(emitter);
+}
+
+static void
+prof_log_emit_allocs(tsd_t *tsd, emitter_t *emitter) {
+	emitter_json_array_kv_begin(emitter, "allocations");
+	prof_alloc_node_t *alloc_node = log_alloc_first;
+	prof_alloc_node_t *alloc_old_node;
+	while (alloc_node != NULL) {
+		emitter_json_object_begin(emitter);
+
+		emitter_json_kv(emitter, "alloc_thread", emitter_type_size,
+		    &alloc_node->alloc_thr_ind);
+
+		emitter_json_kv(emitter, "free_thread", emitter_type_size,
+		    &alloc_node->free_thr_ind);
+
+		emitter_json_kv(emitter, "alloc_trace", emitter_type_size,
+		    &alloc_node->alloc_bt_ind);
+
+		emitter_json_kv(emitter, "free_trace", emitter_type_size,
+		    &alloc_node->free_bt_ind);
+
+		emitter_json_kv(emitter, "alloc_timestamp",
+		    emitter_type_uint64, &alloc_node->alloc_time_ns);
+
+		emitter_json_kv(emitter, "free_timestamp", emitter_type_uint64,
+		    &alloc_node->free_time_ns);
+
+		emitter_json_kv(emitter, "usize", emitter_type_uint64,
+		    &alloc_node->usize);
+
+		emitter_json_object_end(emitter);
+
+		alloc_old_node = alloc_node;
+		alloc_node = alloc_node->next;
+		idalloc(tsd, alloc_old_node);
+	}
+	emitter_json_array_end(emitter);
+}
+
+static void
+prof_log_emit_metadata(emitter_t *emitter) {
+	emitter_json_object_kv_begin(emitter, "info");
+
+	nstime_t now = NSTIME_ZERO_INITIALIZER;
+
+	nstime_update(&now);
+	uint64_t ns = nstime_ns(&now) - nstime_ns(&log_start_timestamp);
+	emitter_json_kv(emitter, "duration", emitter_type_uint64, &ns);
+
+	char *vers = JEMALLOC_VERSION;
+	emitter_json_kv(emitter, "version",
+	    emitter_type_string, &vers);
+
+	emitter_json_kv(emitter, "lg_sample_rate",
+	    emitter_type_int, &lg_prof_sample);
+
+	int pid = prof_getpid();
+	emitter_json_kv(emitter, "pid", emitter_type_int, &pid);
+
+	emitter_json_object_end(emitter);
+}
+
+
+bool
+prof_log_stop(tsdn_t *tsdn) {
+	if (!opt_prof || !prof_booted) {
+		return true;
+	}
+
+	tsd_t *tsd = tsdn_tsd(tsdn);
+	malloc_mutex_lock(tsdn, &log_mtx);
+
+	if (prof_logging_state != prof_logging_state_started) {
+		malloc_mutex_unlock(tsdn, &log_mtx);
+		return true;
+	}
+
+	/*
+	 * Set the state to dumping. We'll set it to stopped when we're done.
+	 * Since other threads won't be able to start/stop/log when the state is
+	 * dumping, we don't have to hold the lock during the whole method.
+	 */
+	prof_logging_state = prof_logging_state_dumping;
+	malloc_mutex_unlock(tsdn, &log_mtx);
+
+
+	emitter_t emitter;
+
+	/* Create a file. */
+
+	int fd;
+#ifdef JEMALLOC_JET
+	if (prof_log_dummy) {
+		fd = 0;
+	} else {
+		fd = creat(log_filename, 0644);
+	}
+#else
+	fd = creat(log_filename, 0644);
+#endif
+
+	if (fd == -1) {
+		malloc_printf("<jemalloc>: creat() for log file \"%s\" "
+			      " failed with %d\n", log_filename, errno);
+		if (opt_abort) {
+			abort();
+		}
+		return true;
+	}
+
+	/* Emit to json. */
+	struct prof_emitter_cb_arg_s arg;
+	arg.fd = fd;
+	emitter_init(&emitter, emitter_output_json, &prof_emitter_write_cb,
+	    (void *)(&arg));
+
+	emitter_json_object_begin(&emitter);
+	prof_log_emit_metadata(&emitter);
+	prof_log_emit_threads(tsd, &emitter);
+	prof_log_emit_traces(tsd, &emitter);
+	prof_log_emit_allocs(tsd, &emitter);
+	emitter_json_object_end(&emitter);
+
+	/* Reset global state. */
+	if (log_tables_initialized) {
+		ckh_delete(tsd, &log_bt_node_set);
+		ckh_delete(tsd, &log_thr_node_set);
+	}
+	log_tables_initialized = false;
+	log_bt_index = 0;
+	log_thr_index = 0;
+	log_bt_first = NULL;
+	log_bt_last = NULL;
+	log_thr_first = NULL;
+	log_thr_last = NULL;
+	log_alloc_first = NULL;
+	log_alloc_last = NULL;
+
+	malloc_mutex_lock(tsdn, &log_mtx);
+	prof_logging_state = prof_logging_state_stopped;
+	malloc_mutex_unlock(tsdn, &log_mtx);
+
+#ifdef JEMALLOC_JET
+	if (prof_log_dummy) {
+		return false;
+	}
+#endif
+	return close(fd);
+}
+
 const char *
 prof_thread_name_get(tsd_t *tsd) {
 	prof_tdata_t *tdata;
@@ -2355,6 +3014,35 @@ prof_boot2(tsd_t *tsd) {
 			}
 		}
 
+		if (opt_prof_log) {
+			prof_log_start(tsd_tsdn(tsd), NULL);
+		}
+
+		if (atexit(prof_log_stop_final) != 0) {
+			malloc_write("<jemalloc>: Error in atexit() "
+				     "for logging\n");
+			if (opt_abort) {
+				abort();
+			}
+		}
+
+		if (malloc_mutex_init(&log_mtx, "prof_log",
+		    WITNESS_RANK_PROF_LOG, malloc_mutex_rank_exclusive)) {
+			return true;
+		}
+
+		if (ckh_new(tsd, &log_bt_node_set, PROF_CKH_MINITEMS,
+		    prof_bt_node_hash, prof_bt_node_keycomp)) {
+			return true;
+		}
+
+		if (ckh_new(tsd, &log_thr_node_set, PROF_CKH_MINITEMS,
+		    prof_thr_node_hash, prof_thr_node_keycomp)) {
+			return true;
+		}
+
+		log_tables_initialized = true;
+
 		gctx_locks = (malloc_mutex_t *)base_alloc(tsd_tsdn(tsd),
 		    b0get(), PROF_NCTX_LOCKS * sizeof(malloc_mutex_t),
 		    CACHELINE);
@@ -2382,16 +3070,14 @@ prof_boot2(tsd_t *tsd) {
 				return true;
 			}
 		}
-	}
-
 #ifdef JEMALLOC_PROF_LIBGCC
-	/*
-	 * Cause the backtracing machinery to allocate its internal state
-	 * before enabling profiling.
-	 */
-	_Unwind_Backtrace(prof_unwind_init_callback, NULL);
+		/*
+		 * Cause the backtracing machinery to allocate its internal
+		 * state before enabling profiling.
+		 */
+		_Unwind_Backtrace(prof_unwind_init_callback, NULL);
 #endif
-
+	}
 	prof_booted = true;
 
 	return false;
diff --git a/dep/jemalloc/src/rtree.c b/dep/jemalloc/src/rtree.c
index 53702cf7236..4ae41fe2fec 100644
--- a/dep/jemalloc/src/rtree.c
+++ b/dep/jemalloc/src/rtree.c
@@ -39,7 +39,7 @@ rtree_node_dalloc_impl(tsdn_t *tsdn, rtree_t *rtree, rtree_node_elm_t *node) {
 	/* Nodes are never deleted during normal operation. */
 	not_reached();
 }
-UNUSED rtree_node_dalloc_t *JET_MUTABLE rtree_node_dalloc =
+rtree_node_dalloc_t *JET_MUTABLE rtree_node_dalloc =
     rtree_node_dalloc_impl;
 
 static rtree_leaf_elm_t *
@@ -54,7 +54,7 @@ rtree_leaf_dalloc_impl(tsdn_t *tsdn, rtree_t *rtree, rtree_leaf_elm_t *leaf) {
 	/* Leaves are never deleted during normal operation. */
 	not_reached();
 }
-UNUSED rtree_leaf_dalloc_t *JET_MUTABLE rtree_leaf_dalloc =
+rtree_leaf_dalloc_t *JET_MUTABLE rtree_leaf_dalloc =
     rtree_leaf_dalloc_impl;
 
 #ifdef JEMALLOC_JET
diff --git a/dep/jemalloc/src/sc.c b/dep/jemalloc/src/sc.c
new file mode 100644
index 00000000000..89ddb6ba6a9
--- /dev/null
+++ b/dep/jemalloc/src/sc.c
@@ -0,0 +1,313 @@
+#include "jemalloc/internal/jemalloc_preamble.h"
+
+#include "jemalloc/internal/assert.h"
+#include "jemalloc/internal/bit_util.h"
+#include "jemalloc/internal/bitmap.h"
+#include "jemalloc/internal/pages.h"
+#include "jemalloc/internal/sc.h"
+
+/*
+ * This module computes the size classes used to satisfy allocations.  The logic
+ * here was ported more or less line-by-line from a shell script, and because of
+ * that is not the most idiomatic C.  Eventually we should fix this, but for now
+ * at least the damage is compartmentalized to this file.
+ */
+
+sc_data_t sc_data_global;
+
+static size_t
+reg_size_compute(int lg_base, int lg_delta, int ndelta) {
+	return (ZU(1) << lg_base) + (ZU(ndelta) << lg_delta);
+}
+
+/* Returns the number of pages in the slab. */
+static int
+slab_size(int lg_page, int lg_base, int lg_delta, int ndelta) {
+	size_t page = (ZU(1) << lg_page);
+	size_t reg_size = reg_size_compute(lg_base, lg_delta, ndelta);
+
+	size_t try_slab_size = page;
+	size_t try_nregs = try_slab_size / reg_size;
+	size_t perfect_slab_size = 0;
+	bool perfect = false;
+	/*
+	 * This loop continues until we find the least common multiple of the
+	 * page size and size class size.  Size classes are all of the form
+	 * base + ndelta * delta == (ndelta + base/ndelta) * delta, which is
+	 * (ndelta + ngroup) * delta.  The way we choose slabbing strategies
+	 * means that delta is at most the page size and ndelta < ngroup.  So
+	 * the loop executes for at most 2 * ngroup - 1 iterations, which is
+	 * also the bound on the number of pages in a slab chosen by default.
+	 * With the current default settings, this is at most 7.
+	 */
+	while (!perfect) {
+		perfect_slab_size = try_slab_size;
+		size_t perfect_nregs = try_nregs;
+		try_slab_size += page;
+		try_nregs = try_slab_size / reg_size;
+		if (perfect_slab_size == perfect_nregs * reg_size) {
+			perfect = true;
+		}
+	}
+	return (int)(perfect_slab_size / page);
+}
+
+static void
+size_class(
+    /* Output. */
+    sc_t *sc,
+    /* Configuration decisions. */
+    int lg_max_lookup, int lg_page, int lg_ngroup,
+    /* Inputs specific to the size class. */
+    int index, int lg_base, int lg_delta, int ndelta) {
+	sc->index = index;
+	sc->lg_base = lg_base;
+	sc->lg_delta = lg_delta;
+	sc->ndelta = ndelta;
+	sc->psz = (reg_size_compute(lg_base, lg_delta, ndelta)
+	    % (ZU(1) << lg_page) == 0);
+	size_t size = (ZU(1) << lg_base) + (ZU(ndelta) << lg_delta);
+	if (index == 0) {
+		assert(!sc->psz);
+	}
+	if (size < (ZU(1) << (lg_page + lg_ngroup))) {
+		sc->bin = true;
+		sc->pgs = slab_size(lg_page, lg_base, lg_delta, ndelta);
+	} else {
+		sc->bin = false;
+		sc->pgs = 0;
+	}
+	if (size <= (ZU(1) << lg_max_lookup)) {
+		sc->lg_delta_lookup = lg_delta;
+	} else {
+		sc->lg_delta_lookup = 0;
+	}
+}
+
+static void
+size_classes(
+    /* Output. */
+    sc_data_t *sc_data,
+    /* Determined by the system. */
+    size_t lg_ptr_size, int lg_quantum,
+    /* Configuration decisions. */
+    int lg_tiny_min, int lg_max_lookup, int lg_page, int lg_ngroup) {
+	int ptr_bits = (1 << lg_ptr_size) * 8;
+	int ngroup = (1 << lg_ngroup);
+	int ntiny = 0;
+	int nlbins = 0;
+	int lg_tiny_maxclass = (unsigned)-1;
+	int nbins = 0;
+	int npsizes = 0;
+
+	int index = 0;
+
+	int ndelta = 0;
+	int lg_base = lg_tiny_min;
+	int lg_delta = lg_base;
+
+	/* Outputs that we update as we go. */
+	size_t lookup_maxclass = 0;
+	size_t small_maxclass = 0;
+	int lg_large_minclass = 0;
+	size_t large_maxclass = 0;
+
+	/* Tiny size classes. */
+	while (lg_base < lg_quantum) {
+		sc_t *sc = &sc_data->sc[index];
+		size_class(sc, lg_max_lookup, lg_page, lg_ngroup, index,
+		    lg_base, lg_delta, ndelta);
+		if (sc->lg_delta_lookup != 0) {
+			nlbins = index + 1;
+		}
+		if (sc->psz) {
+			npsizes++;
+		}
+		if (sc->bin) {
+			nbins++;
+		}
+		ntiny++;
+		/* Final written value is correct. */
+		lg_tiny_maxclass = lg_base;
+		index++;
+		lg_delta = lg_base;
+		lg_base++;
+	}
+
+	/* First non-tiny (pseudo) group. */
+	if (ntiny != 0) {
+		sc_t *sc = &sc_data->sc[index];
+		/*
+		 * See the note in sc.h; the first non-tiny size class has an
+		 * unusual encoding.
+		 */
+		lg_base--;
+		ndelta = 1;
+		size_class(sc, lg_max_lookup, lg_page, lg_ngroup, index,
+		    lg_base, lg_delta, ndelta);
+		index++;
+		lg_base++;
+		lg_delta++;
+		if (sc->psz) {
+			npsizes++;
+		}
+		if (sc->bin) {
+			nbins++;
+		}
+	}
+	while (ndelta < ngroup) {
+		sc_t *sc = &sc_data->sc[index];
+		size_class(sc, lg_max_lookup, lg_page, lg_ngroup, index,
+		    lg_base, lg_delta, ndelta);
+		index++;
+		ndelta++;
+		if (sc->psz) {
+			npsizes++;
+		}
+		if (sc->bin) {
+			nbins++;
+		}
+	}
+
+	/* All remaining groups. */
+	lg_base = lg_base + lg_ngroup;
+	while (lg_base < ptr_bits - 1) {
+		ndelta = 1;
+		int ndelta_limit;
+		if (lg_base == ptr_bits - 2) {
+			ndelta_limit = ngroup - 1;
+		} else {
+			ndelta_limit = ngroup;
+		}
+		while (ndelta <= ndelta_limit) {
+			sc_t *sc = &sc_data->sc[index];
+			size_class(sc, lg_max_lookup, lg_page, lg_ngroup, index,
+			    lg_base, lg_delta, ndelta);
+			if (sc->lg_delta_lookup != 0) {
+				nlbins = index + 1;
+				/* Final written value is correct. */
+				lookup_maxclass = (ZU(1) << lg_base)
+				    + (ZU(ndelta) << lg_delta);
+			}
+			if (sc->psz) {
+				npsizes++;
+			}
+			if (sc->bin) {
+				nbins++;
+				/* Final written value is correct. */
+				small_maxclass = (ZU(1) << lg_base)
+				    + (ZU(ndelta) << lg_delta);
+				if (lg_ngroup > 0) {
+					lg_large_minclass = lg_base + 1;
+				} else {
+					lg_large_minclass = lg_base + 2;
+				}
+			}
+			large_maxclass = (ZU(1) << lg_base)
+			    + (ZU(ndelta) << lg_delta);
+			index++;
+			ndelta++;
+		}
+		lg_base++;
+		lg_delta++;
+	}
+	/* Additional outputs. */
+	int nsizes = index;
+	unsigned lg_ceil_nsizes = lg_ceil(nsizes);
+
+	/* Fill in the output data. */
+	sc_data->ntiny = ntiny;
+	sc_data->nlbins = nlbins;
+	sc_data->nbins = nbins;
+	sc_data->nsizes = nsizes;
+	sc_data->lg_ceil_nsizes = lg_ceil_nsizes;
+	sc_data->npsizes = npsizes;
+	sc_data->lg_tiny_maxclass = lg_tiny_maxclass;
+	sc_data->lookup_maxclass = lookup_maxclass;
+	sc_data->small_maxclass = small_maxclass;
+	sc_data->lg_large_minclass = lg_large_minclass;
+	sc_data->large_minclass = (ZU(1) << lg_large_minclass);
+	sc_data->large_maxclass = large_maxclass;
+
+	/*
+	 * We compute these values in two ways:
+	 *   - Incrementally, as above.
+	 *   - In macros, in sc.h.
+	 * The computation is easier when done incrementally, but putting it in
+	 * a constant makes it available to the fast paths without having to
+	 * touch the extra global cacheline.  We assert, however, that the two
+	 * computations are equivalent.
+	 */
+	assert(sc_data->npsizes == SC_NPSIZES);
+	assert(sc_data->lg_tiny_maxclass == SC_LG_TINY_MAXCLASS);
+	assert(sc_data->small_maxclass == SC_SMALL_MAXCLASS);
+	assert(sc_data->large_minclass == SC_LARGE_MINCLASS);
+	assert(sc_data->lg_large_minclass == SC_LG_LARGE_MINCLASS);
+	assert(sc_data->large_maxclass == SC_LARGE_MAXCLASS);
+
+	/* 
+	 * In the allocation fastpath, we want to assume that we can
+	 * unconditionally subtract the requested allocation size from
+	 * a ssize_t, and detect passing through 0 correctly.  This
+	 * results in optimal generated code.  For this to work, the
+	 * maximum allocation size must be less than SSIZE_MAX.
+	 */
+	assert(SC_LARGE_MAXCLASS < SSIZE_MAX);
+}
+
+void
+sc_data_init(sc_data_t *sc_data) {
+	assert(!sc_data->initialized);
+
+	int lg_max_lookup = 12;
+
+	size_classes(sc_data, LG_SIZEOF_PTR, LG_QUANTUM, SC_LG_TINY_MIN,
+	    lg_max_lookup, LG_PAGE, 2);
+
+	sc_data->initialized = true;
+}
+
+static void
+sc_data_update_sc_slab_size(sc_t *sc, size_t reg_size, size_t pgs_guess) {
+	size_t min_pgs = reg_size / PAGE;
+	if (reg_size % PAGE != 0) {
+		min_pgs++;
+	}
+	/*
+	 * BITMAP_MAXBITS is actually determined by putting the smallest
+	 * possible size-class on one page, so this can never be 0.
+	 */
+	size_t max_pgs = BITMAP_MAXBITS * reg_size / PAGE;
+
+	assert(min_pgs <= max_pgs);
+	assert(min_pgs > 0);
+	assert(max_pgs >= 1);
+	if (pgs_guess < min_pgs) {
+		sc->pgs = (int)min_pgs;
+	} else if (pgs_guess > max_pgs) {
+		sc->pgs = (int)max_pgs;
+	} else {
+		sc->pgs = (int)pgs_guess;
+	}
+}
+
+void
+sc_data_update_slab_size(sc_data_t *data, size_t begin, size_t end, int pgs) {
+	assert(data->initialized);
+	for (int i = 0; i < data->nsizes; i++) {
+		sc_t *sc = &data->sc[i];
+		if (!sc->bin) {
+			break;
+		}
+		size_t reg_size = reg_size_compute(sc->lg_base, sc->lg_delta,
+		    sc->ndelta);
+		if (begin <= reg_size && reg_size <= end) {
+			sc_data_update_sc_slab_size(sc, reg_size, pgs);
+		}
+	}
+}
+
+void
+sc_boot(sc_data_t *data) {
+	sc_data_init(data);
+}
diff --git a/dep/jemalloc/src/stats.c b/dep/jemalloc/src/stats.c
index 08b9507cfe9..4c427e0d35b 100644
--- a/dep/jemalloc/src/stats.c
+++ b/dep/jemalloc/src/stats.c
@@ -52,6 +52,20 @@ char opt_stats_print_opts[stats_print_tot_num_options+1] = "";
 
 /******************************************************************************/
 
+static uint64_t
+rate_per_second(uint64_t value, uint64_t uptime_ns) {
+	uint64_t billion = 1000000000;
+	if (uptime_ns == 0 || value == 0) {
+		return 0;
+	}
+	if (uptime_ns < billion) {
+		return value;
+	} else {
+		uint64_t uptime_s = uptime_ns / billion;
+		return value / uptime_s;
+	}
+}
+
 /* Calculate x.yyy and output a string (takes a fixed sized char array). */
 static bool
 get_rate_str(uint64_t dividend, uint64_t divisor, char str[6]) {
@@ -104,24 +118,26 @@ mutex_stats_init_cols(emitter_row_t *row, const char *table_name,
 
 #define WIDTH_uint32_t 12
 #define WIDTH_uint64_t 16
-#define OP(counter, counter_type, human)				\
+#define OP(counter, counter_type, human, derived, base_counter)	\
 	col = &col_##counter_type[k_##counter_type];			\
 	++k_##counter_type;						\
 	emitter_col_init(col, row);					\
 	col->justify = emitter_justify_right;				\
-	col->width = WIDTH_##counter_type;				\
+	col->width = derived ? 8 : WIDTH_##counter_type;		\
 	col->type = emitter_type_title;					\
 	col->str_val = human;
 	MUTEX_PROF_COUNTERS
 #undef OP
 #undef WIDTH_uint32_t
 #undef WIDTH_uint64_t
+	col_uint64_t[mutex_counter_total_wait_time_ps].width = 10;
 }
 
 static void
 mutex_stats_read_global(const char *name, emitter_col_t *col_name,
     emitter_col_t col_uint64_t[mutex_prof_num_uint64_t_counters],
-    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters]) {
+    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters],
+    uint64_t uptime) {
 	char cmd[MUTEX_CTL_STR_MAX_LENGTH];
 
 	col_name->str_val = name;
@@ -129,12 +145,17 @@ mutex_stats_read_global(const char *name, emitter_col_t *col_name,
 	emitter_col_t *dst;
 #define EMITTER_TYPE_uint32_t emitter_type_uint32
 #define EMITTER_TYPE_uint64_t emitter_type_uint64
-#define OP(counter, counter_type, human)				\
+#define OP(counter, counter_type, human, derived, base_counter)	\
 	dst = &col_##counter_type[mutex_counter_##counter];		\
 	dst->type = EMITTER_TYPE_##counter_type;			\
-	gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,		\
-	    "mutexes", name, #counter);					\
-	CTL_GET(cmd, (counter_type *)&dst->bool_val, counter_type);
+	if (!derived) {							\
+		gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,	\
+		    "mutexes", name, #counter);				\
+		CTL_GET(cmd, (counter_type *)&dst->bool_val, counter_type);	\
+	} else { \
+	    emitter_col_t *base = &col_##counter_type[mutex_counter_##base_counter];	\
+	    dst->counter_type##_val = rate_per_second(base->counter_type##_val, uptime); \
+	}
 	MUTEX_PROF_COUNTERS
 #undef OP
 #undef EMITTER_TYPE_uint32_t
@@ -145,7 +166,8 @@ static void
 mutex_stats_read_arena(unsigned arena_ind, mutex_prof_arena_ind_t mutex_ind,
     const char *name, emitter_col_t *col_name,
     emitter_col_t col_uint64_t[mutex_prof_num_uint64_t_counters],
-    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters]) {
+    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters],
+    uint64_t uptime) {
 	char cmd[MUTEX_CTL_STR_MAX_LENGTH];
 
 	col_name->str_val = name;
@@ -153,13 +175,17 @@ mutex_stats_read_arena(unsigned arena_ind, mutex_prof_arena_ind_t mutex_ind,
 	emitter_col_t *dst;
 #define EMITTER_TYPE_uint32_t emitter_type_uint32
 #define EMITTER_TYPE_uint64_t emitter_type_uint64
-#define OP(counter, counter_type, human)				\
+#define OP(counter, counter_type, human, derived, base_counter)	\
 	dst = &col_##counter_type[mutex_counter_##counter];		\
 	dst->type = EMITTER_TYPE_##counter_type;			\
-	gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,		\
-	    "arenas.0.mutexes",	arena_mutex_names[mutex_ind], #counter);\
-	CTL_M2_GET(cmd, arena_ind,					\
-	    (counter_type *)&dst->bool_val, counter_type);
+	if (!derived) {                                   \
+		gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,        \
+		    "arenas.0.mutexes", arena_mutex_names[mutex_ind], #counter);\
+		CTL_M2_GET(cmd, arena_ind, (counter_type *)&dst->bool_val, counter_type); \
+	} else {                      \
+		emitter_col_t *base = &col_##counter_type[mutex_counter_##base_counter];	\
+		dst->counter_type##_val = rate_per_second(base->counter_type##_val, uptime); \
+	}
 	MUTEX_PROF_COUNTERS
 #undef OP
 #undef EMITTER_TYPE_uint32_t
@@ -169,19 +195,25 @@ mutex_stats_read_arena(unsigned arena_ind, mutex_prof_arena_ind_t mutex_ind,
 static void
 mutex_stats_read_arena_bin(unsigned arena_ind, unsigned bin_ind,
     emitter_col_t col_uint64_t[mutex_prof_num_uint64_t_counters],
-    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters]) {
+    emitter_col_t col_uint32_t[mutex_prof_num_uint32_t_counters],
+    uint64_t uptime) {
 	char cmd[MUTEX_CTL_STR_MAX_LENGTH];
 	emitter_col_t *dst;
 
 #define EMITTER_TYPE_uint32_t emitter_type_uint32
 #define EMITTER_TYPE_uint64_t emitter_type_uint64
-#define OP(counter, counter_type, human)				\
+#define OP(counter, counter_type, human, derived, base_counter)	\
 	dst = &col_##counter_type[mutex_counter_##counter];		\
 	dst->type = EMITTER_TYPE_##counter_type;			\
-	gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,		\
-	    "arenas.0.bins.0","mutex", #counter);			\
-	CTL_M2_M4_GET(cmd, arena_ind, bin_ind,				\
-	    (counter_type *)&dst->bool_val, counter_type);
+	if (!derived) {                                   \
+		gen_mutex_ctl_str(cmd, MUTEX_CTL_STR_MAX_LENGTH,        \
+		    "arenas.0.bins.0","mutex", #counter);            \
+		CTL_M2_M4_GET(cmd, arena_ind, bin_ind,                \
+		    (counter_type *)&dst->bool_val, counter_type);  \
+	} else {                      \
+		emitter_col_t *base = &col_##counter_type[mutex_counter_##base_counter]; \
+		dst->counter_type##_val = rate_per_second(base->counter_type##_val, uptime); \
+	}
 	MUTEX_PROF_COUNTERS
 #undef OP
 #undef EMITTER_TYPE_uint32_t
@@ -204,19 +236,38 @@ mutex_stats_emit(emitter_t *emitter, emitter_row_t *row,
 
 #define EMITTER_TYPE_uint32_t emitter_type_uint32
 #define EMITTER_TYPE_uint64_t emitter_type_uint64
-#define OP(counter, type, human)					\
-	col = &col_##type[k_##type];						\
-	++k_##type;							\
-	emitter_json_kv(emitter, #counter, EMITTER_TYPE_##type,		\
-	    (const void *)&col->bool_val);
+#define OP(counter, type, human, derived, base_counter)		\
+	if (!derived) {                    \
+		col = &col_##type[k_##type];                        \
+		++k_##type;                            \
+		emitter_json_kv(emitter, #counter, EMITTER_TYPE_##type,        \
+		    (const void *)&col->bool_val); \
+	}
 	MUTEX_PROF_COUNTERS;
 #undef OP
 #undef EMITTER_TYPE_uint32_t
 #undef EMITTER_TYPE_uint64_t
 }
 
+#define COL(row_name, column_name, left_or_right, col_width, etype)      \
+	emitter_col_t col_##column_name;                                     \
+	emitter_col_init(&col_##column_name, &row_name);                     \
+	col_##column_name.justify = emitter_justify_##left_or_right;         \
+	col_##column_name.width = col_width;                                 \
+	col_##column_name.type = emitter_type_##etype;
+
+#define COL_HDR(row_name, column_name, human, left_or_right, col_width, etype)  \
+	COL(row_name, column_name, left_or_right, col_width, etype)	         \
+	emitter_col_t header_##column_name;                                  \
+	emitter_col_init(&header_##column_name, &header_##row_name);         \
+	header_##column_name.justify = emitter_justify_##left_or_right;      \
+	header_##column_name.width = col_width;                              \
+	header_##column_name.type = emitter_type_title;                      \
+	header_##column_name.str_val = human ? human : #column_name;
+
+
 static void
-stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
+stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i, uint64_t uptime) {
 	size_t page;
 	bool in_gap, in_gap_prev;
 	unsigned nbins, j;
@@ -230,43 +281,36 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 
 	emitter_row_t row;
 	emitter_row_init(&row);
-#define COL(name, left_or_right, col_width, etype)			\
-	emitter_col_t col_##name;					\
-	emitter_col_init(&col_##name, &row);				\
-	col_##name.justify = emitter_justify_##left_or_right;		\
-	col_##name.width = col_width;					\
-	col_##name.type = emitter_type_##etype;				\
-	emitter_col_t header_col_##name;				\
-	emitter_col_init(&header_col_##name, &header_row);		\
-	header_col_##name.justify = emitter_justify_##left_or_right;	\
-	header_col_##name.width = col_width;				\
-	header_col_##name.type = emitter_type_title;			\
-	header_col_##name.str_val = #name;
-
-	COL(size, right, 20, size)
-	COL(ind, right, 4, unsigned)
-	COL(allocated, right, 13, uint64)
-	COL(nmalloc, right, 13, uint64)
-	COL(ndalloc, right, 13, uint64)
-	COL(nrequests, right, 13, uint64)
-	COL(curregs, right, 13, size)
-	COL(curslabs, right, 13, size)
-	COL(regs, right, 5, unsigned)
-	COL(pgs, right, 4, size)
+
+	COL_HDR(row, size, NULL, right, 20, size)
+	COL_HDR(row, ind, NULL, right, 4, unsigned)
+	COL_HDR(row, allocated, NULL, right, 13, uint64)
+	COL_HDR(row, nmalloc, NULL, right, 13, uint64)
+	COL_HDR(row, nmalloc_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, ndalloc, NULL, right, 13, uint64)
+	COL_HDR(row, ndalloc_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, nrequests, NULL, right, 13, uint64)
+	COL_HDR(row, nrequests_ps, "(#/sec)", right, 10, uint64)
+	COL_HDR(row, nshards, NULL, right, 9, unsigned)
+	COL_HDR(row, curregs, NULL, right, 13, size)
+	COL_HDR(row, curslabs, NULL, right, 13, size)
+	COL_HDR(row, regs, NULL, right, 5, unsigned)
+	COL_HDR(row, pgs, NULL, right, 4, size)
 	/* To buffer a right- and left-justified column. */
-	COL(justify_spacer, right, 1, title)
-	COL(util, right, 6, title)
-	COL(nfills, right, 13, uint64)
-	COL(nflushes, right, 13, uint64)
-	COL(nslabs, right, 13, uint64)
-	COL(nreslabs, right, 13, uint64)
-#undef COL
+	COL_HDR(row, justify_spacer, NULL, right, 1, title)
+	COL_HDR(row, util, NULL, right, 6, title)
+	COL_HDR(row, nfills, NULL, right, 13, uint64)
+	COL_HDR(row, nfills_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, nflushes, NULL, right, 13, uint64)
+	COL_HDR(row, nflushes_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, nslabs, NULL, right, 13, uint64)
+	COL_HDR(row, nreslabs, NULL, right, 13, uint64)
+	COL_HDR(row, nreslabs_ps, "(#/sec)", right, 8, uint64)
 
 	/* Don't want to actually print the name. */
-	header_col_justify_spacer.str_val = " ";
+	header_justify_spacer.str_val = " ";
 	col_justify_spacer.str_val = " ";
 
-
 	emitter_col_t col_mutex64[mutex_prof_num_uint64_t_counters];
 	emitter_col_t col_mutex32[mutex_prof_num_uint32_t_counters];
 
@@ -284,16 +328,16 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 	 * We print a "bins:" header as part of the table row; we need to adjust
 	 * the header size column to compensate.
 	 */
-	header_col_size.width -=5;
+	header_size.width -=5;
 	emitter_table_printf(emitter, "bins:");
 	emitter_table_row(emitter, &header_row);
-	emitter_json_arr_begin(emitter, "bins");
+	emitter_json_array_kv_begin(emitter, "bins");
 
 	for (j = 0, in_gap = false; j < nbins; j++) {
 		uint64_t nslabs;
 		size_t reg_size, slab_size, curregs;
 		size_t curslabs;
-		uint32_t nregs;
+		uint32_t nregs, nshards;
 		uint64_t nmalloc, ndalloc, nrequests, nfills, nflushes;
 		uint64_t nreslabs;
 
@@ -310,6 +354,7 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 		CTL_M2_GET("arenas.bin.0.size", j, &reg_size, size_t);
 		CTL_M2_GET("arenas.bin.0.nregs", j, &nregs, uint32_t);
 		CTL_M2_GET("arenas.bin.0.slab_size", j, &slab_size, size_t);
+		CTL_M2_GET("arenas.bin.0.nshards", j, &nshards, uint32_t);
 
 		CTL_M2_M4_GET("stats.arenas.0.bins.0.nmalloc", i, j, &nmalloc,
 		    uint64_t);
@@ -330,10 +375,10 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 
 		if (mutex) {
 			mutex_stats_read_arena_bin(i, j, col_mutex64,
-			    col_mutex32);
+			    col_mutex32, uptime);
 		}
 
-		emitter_json_arr_obj_begin(emitter);
+		emitter_json_object_begin(emitter);
 		emitter_json_kv(emitter, "nmalloc", emitter_type_uint64,
 		    &nmalloc);
 		emitter_json_kv(emitter, "ndalloc", emitter_type_uint64,
@@ -351,12 +396,12 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 		emitter_json_kv(emitter, "curslabs", emitter_type_size,
 		    &curslabs);
 		if (mutex) {
-			emitter_json_dict_begin(emitter, "mutex");
+			emitter_json_object_kv_begin(emitter, "mutex");
 			mutex_stats_emit(emitter, NULL, col_mutex64,
 			    col_mutex32);
-			emitter_json_dict_end(emitter);
+			emitter_json_object_end(emitter);
 		}
-		emitter_json_arr_obj_end(emitter);
+		emitter_json_object_end(emitter);
 
 		size_t availregs = nregs * curslabs;
 		char util[6];
@@ -381,17 +426,24 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 		col_ind.unsigned_val = j;
 		col_allocated.size_val = curregs * reg_size;
 		col_nmalloc.uint64_val = nmalloc;
+		col_nmalloc_ps.uint64_val = rate_per_second(nmalloc, uptime);
 		col_ndalloc.uint64_val = ndalloc;
+		col_ndalloc_ps.uint64_val = rate_per_second(ndalloc, uptime);
 		col_nrequests.uint64_val = nrequests;
+		col_nrequests_ps.uint64_val = rate_per_second(nrequests, uptime);
+		col_nshards.unsigned_val = nshards;
 		col_curregs.size_val = curregs;
 		col_curslabs.size_val = curslabs;
 		col_regs.unsigned_val = nregs;
 		col_pgs.size_val = slab_size / page;
 		col_util.str_val = util;
 		col_nfills.uint64_val = nfills;
+		col_nfills_ps.uint64_val = rate_per_second(nfills, uptime);
 		col_nflushes.uint64_val = nflushes;
+		col_nflushes_ps.uint64_val = rate_per_second(nflushes, uptime);
 		col_nslabs.uint64_val = nslabs;
 		col_nreslabs.uint64_val = nreslabs;
+		col_nreslabs_ps.uint64_val = rate_per_second(nreslabs, uptime);
 
 		/*
 		 * Note that mutex columns were initialized above, if mutex ==
@@ -400,7 +452,7 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 
 		emitter_table_row(emitter, &row);
 	}
-	emitter_json_arr_end(emitter); /* Close "bins". */
+	emitter_json_array_end(emitter); /* Close "bins". */
 
 	if (in_gap) {
 		emitter_table_printf(emitter, "                     ---\n");
@@ -408,7 +460,7 @@ stats_arena_bins_print(emitter_t *emitter, bool mutex, unsigned i) {
 }
 
 static void
-stats_arena_lextents_print(emitter_t *emitter, unsigned i) {
+stats_arena_lextents_print(emitter_t *emitter, unsigned i, uint64_t uptime) {
 	unsigned nbins, nlextents, j;
 	bool in_gap, in_gap_prev;
 
@@ -420,34 +472,22 @@ stats_arena_lextents_print(emitter_t *emitter, unsigned i) {
 	emitter_row_t row;
 	emitter_row_init(&row);
 
-#define COL(name, left_or_right, col_width, etype)			\
-	emitter_col_t header_##name;					\
-	emitter_col_init(&header_##name, &header_row);			\
-	header_##name.justify = emitter_justify_##left_or_right;	\
-	header_##name.width = col_width;				\
-	header_##name.type = emitter_type_title;			\
-	header_##name.str_val = #name;					\
-									\
-	emitter_col_t col_##name;					\
-	emitter_col_init(&col_##name, &row);				\
-	col_##name.justify = emitter_justify_##left_or_right;		\
-	col_##name.width = col_width;					\
-	col_##name.type = emitter_type_##etype;
-
-	COL(size, right, 20, size)
-	COL(ind, right, 4, unsigned)
-	COL(allocated, right, 13, size)
-	COL(nmalloc, right, 13, uint64)
-	COL(ndalloc, right, 13, uint64)
-	COL(nrequests, right, 13, uint64)
-	COL(curlextents, right, 13, size)
-#undef COL
+	COL_HDR(row, size, NULL, right, 20, size)
+	COL_HDR(row, ind, NULL, right, 4, unsigned)
+	COL_HDR(row, allocated, NULL, right, 13, size)
+	COL_HDR(row, nmalloc, NULL, right, 13, uint64)
+	COL_HDR(row, nmalloc_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, ndalloc, NULL, right, 13, uint64)
+	COL_HDR(row, ndalloc_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, nrequests, NULL, right, 13, uint64)
+	COL_HDR(row, nrequests_ps, "(#/sec)", right, 8, uint64)
+	COL_HDR(row, curlextents, NULL, right, 13, size)
 
 	/* As with bins, we label the large extents table. */
 	header_size.width -= 6;
 	emitter_table_printf(emitter, "large:");
 	emitter_table_row(emitter, &header_row);
-	emitter_json_arr_begin(emitter, "lextents");
+	emitter_json_array_kv_begin(emitter, "lextents");
 
 	for (j = 0, in_gap = false; j < nlextents; j++) {
 		uint64_t nmalloc, ndalloc, nrequests;
@@ -471,31 +511,122 @@ stats_arena_lextents_print(emitter_t *emitter, unsigned i) {
 		CTL_M2_M4_GET("stats.arenas.0.lextents.0.curlextents", i, j,
 		    &curlextents, size_t);
 
-		emitter_json_arr_obj_begin(emitter);
+		emitter_json_object_begin(emitter);
 		emitter_json_kv(emitter, "curlextents", emitter_type_size,
 		    &curlextents);
-		emitter_json_arr_obj_end(emitter);
+		emitter_json_object_end(emitter);
 
 		col_size.size_val = lextent_size;
 		col_ind.unsigned_val = nbins + j;
 		col_allocated.size_val = curlextents * lextent_size;
 		col_nmalloc.uint64_val = nmalloc;
+		col_nmalloc_ps.uint64_val = rate_per_second(nmalloc, uptime);
 		col_ndalloc.uint64_val = ndalloc;
+		col_ndalloc_ps.uint64_val = rate_per_second(ndalloc, uptime);
 		col_nrequests.uint64_val = nrequests;
+		col_nrequests_ps.uint64_val = rate_per_second(nrequests, uptime);
 		col_curlextents.size_val = curlextents;
 
 		if (!in_gap) {
 			emitter_table_row(emitter, &row);
 		}
 	}
-	emitter_json_arr_end(emitter); /* Close "lextents". */
+	emitter_json_array_end(emitter); /* Close "lextents". */
+	if (in_gap) {
+		emitter_table_printf(emitter, "                     ---\n");
+	}
+}
+
+static void
+stats_arena_extents_print(emitter_t *emitter, unsigned i) {
+	unsigned j;
+	bool in_gap, in_gap_prev;
+	emitter_row_t header_row;
+	emitter_row_init(&header_row);
+	emitter_row_t row;
+	emitter_row_init(&row);
+
+	COL_HDR(row, size, NULL, right, 20, size)
+	COL_HDR(row, ind, NULL, right, 4, unsigned)
+	COL_HDR(row, ndirty, NULL, right, 13, size)
+	COL_HDR(row, dirty, NULL, right, 13, size)
+	COL_HDR(row, nmuzzy, NULL, right, 13, size)
+	COL_HDR(row, muzzy, NULL, right, 13, size)
+	COL_HDR(row, nretained, NULL, right, 13, size)
+	COL_HDR(row, retained, NULL, right, 13, size)
+	COL_HDR(row, ntotal, NULL, right, 13, size)
+	COL_HDR(row, total, NULL, right, 13, size)
+
+	/* Label this section. */
+	header_size.width -= 8;
+	emitter_table_printf(emitter, "extents:");
+	emitter_table_row(emitter, &header_row);
+	emitter_json_array_kv_begin(emitter, "extents");
+
+	in_gap = false;
+	for (j = 0; j < SC_NPSIZES; j++) {
+		size_t ndirty, nmuzzy, nretained, total, dirty_bytes,
+		    muzzy_bytes, retained_bytes, total_bytes;
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.ndirty", i, j,
+		    &ndirty, size_t);
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.nmuzzy", i, j,
+		    &nmuzzy, size_t);
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.nretained", i, j,
+		    &nretained, size_t);
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.dirty_bytes", i, j,
+		    &dirty_bytes, size_t);
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.muzzy_bytes", i, j,
+		    &muzzy_bytes, size_t);
+		CTL_M2_M4_GET("stats.arenas.0.extents.0.retained_bytes", i, j,
+		    &retained_bytes, size_t);
+		total = ndirty + nmuzzy + nretained;
+		total_bytes = dirty_bytes + muzzy_bytes + retained_bytes;
+
+		in_gap_prev = in_gap;
+		in_gap = (total == 0);
+
+		if (in_gap_prev && !in_gap) {
+			emitter_table_printf(emitter,
+			    "                     ---\n");
+		}
+
+		emitter_json_object_begin(emitter);
+		emitter_json_kv(emitter, "ndirty", emitter_type_size, &ndirty);
+		emitter_json_kv(emitter, "nmuzzy", emitter_type_size, &nmuzzy);
+		emitter_json_kv(emitter, "nretained", emitter_type_size,
+		    &nretained);
+
+		emitter_json_kv(emitter, "dirty_bytes", emitter_type_size,
+		    &dirty_bytes);
+		emitter_json_kv(emitter, "muzzy_bytes", emitter_type_size,
+		    &muzzy_bytes);
+		emitter_json_kv(emitter, "retained_bytes", emitter_type_size,
+		    &retained_bytes);
+		emitter_json_object_end(emitter);
+
+		col_size.size_val = sz_pind2sz(j);
+		col_ind.size_val = j;
+		col_ndirty.size_val = ndirty;
+		col_dirty.size_val = dirty_bytes;
+		col_nmuzzy.size_val = nmuzzy;
+		col_muzzy.size_val = muzzy_bytes;
+		col_nretained.size_val = nretained;
+		col_retained.size_val = retained_bytes;
+		col_ntotal.size_val = total;
+		col_total.size_val = total_bytes;
+
+		if (!in_gap) {
+			emitter_table_row(emitter, &row);
+		}
+	}
+	emitter_json_array_end(emitter); /* Close "extents". */
 	if (in_gap) {
 		emitter_table_printf(emitter, "                     ---\n");
 	}
 }
 
 static void
-stats_arena_mutexes_print(emitter_t *emitter, unsigned arena_ind) {
+stats_arena_mutexes_print(emitter_t *emitter, unsigned arena_ind, uint64_t uptime) {
 	emitter_row_t row;
 	emitter_col_t col_name;
 	emitter_col_t col64[mutex_prof_num_uint64_t_counters];
@@ -504,29 +635,29 @@ stats_arena_mutexes_print(emitter_t *emitter, unsigned arena_ind) {
 	emitter_row_init(&row);
 	mutex_stats_init_cols(&row, "", &col_name, col64, col32);
 
-	emitter_json_dict_begin(emitter, "mutexes");
+	emitter_json_object_kv_begin(emitter, "mutexes");
 	emitter_table_row(emitter, &row);
 
 	for (mutex_prof_arena_ind_t i = 0; i < mutex_prof_num_arena_mutexes;
 	    i++) {
 		const char *name = arena_mutex_names[i];
-		emitter_json_dict_begin(emitter, name);
+		emitter_json_object_kv_begin(emitter, name);
 		mutex_stats_read_arena(arena_ind, i, name, &col_name, col64,
-		    col32);
+		    col32, uptime);
 		mutex_stats_emit(emitter, &row, col64, col32);
-		emitter_json_dict_end(emitter); /* Close the mutex dict. */
+		emitter_json_object_end(emitter); /* Close the mutex dict. */
 	}
-	emitter_json_dict_end(emitter); /* End "mutexes". */
+	emitter_json_object_end(emitter); /* End "mutexes". */
 }
 
 static void
 stats_arena_print(emitter_t *emitter, unsigned i, bool bins, bool large,
-    bool mutex) {
+    bool mutex, bool extents) {
 	unsigned nthreads;
 	const char *dss;
 	ssize_t dirty_decay_ms, muzzy_decay_ms;
 	size_t page, pactive, pdirty, pmuzzy, mapped, retained;
-	size_t base, internal, resident, metadata_thp;
+	size_t base, internal, resident, metadata_thp, extent_avail;
 	uint64_t dirty_npurge, dirty_nmadvise, dirty_purged;
 	uint64_t muzzy_npurge, muzzy_nmadvise, muzzy_purged;
 	size_t small_allocated;
@@ -594,98 +725,74 @@ stats_arena_print(emitter_t *emitter, unsigned i, bool bins, bool large,
 	    &muzzy_purged);
 
 	/* Table-style emission. */
-	emitter_col_t decay_type;
-	emitter_col_init(&decay_type, &decay_row);
-	decay_type.justify = emitter_justify_right;
-	decay_type.width = 9;
-	decay_type.type = emitter_type_title;
-	decay_type.str_val = "decaying:";
-
-	emitter_col_t decay_time;
-	emitter_col_init(&decay_time, &decay_row);
-	decay_time.justify = emitter_justify_right;
-	decay_time.width = 6;
-	decay_time.type = emitter_type_title;
-	decay_time.str_val = "time";
-
-	emitter_col_t decay_npages;
-	emitter_col_init(&decay_npages, &decay_row);
-	decay_npages.justify = emitter_justify_right;
-	decay_npages.width = 13;
-	decay_npages.type = emitter_type_title;
-	decay_npages.str_val = "npages";
-
-	emitter_col_t decay_sweeps;
-	emitter_col_init(&decay_sweeps, &decay_row);
-	decay_sweeps.justify = emitter_justify_right;
-	decay_sweeps.width = 13;
-	decay_sweeps.type = emitter_type_title;
-	decay_sweeps.str_val = "sweeps";
-
-	emitter_col_t decay_madvises;
-	emitter_col_init(&decay_madvises, &decay_row);
-	decay_madvises.justify = emitter_justify_right;
-	decay_madvises.width = 13;
-	decay_madvises.type = emitter_type_title;
-	decay_madvises.str_val = "madvises";
-
-	emitter_col_t decay_purged;
-	emitter_col_init(&decay_purged, &decay_row);
-	decay_purged.justify = emitter_justify_right;
-	decay_purged.width = 13;
-	decay_purged.type = emitter_type_title;
-	decay_purged.str_val = "purged";
+	COL(decay_row, decay_type, right, 9, title);
+	col_decay_type.str_val = "decaying:";
+
+	COL(decay_row, decay_time, right, 6, title);
+	col_decay_time.str_val = "time";
+
+	COL(decay_row, decay_npages, right, 13, title);
+	col_decay_npages.str_val = "npages";
+
+	COL(decay_row, decay_sweeps, right, 13, title);
+	col_decay_sweeps.str_val = "sweeps";
+
+	COL(decay_row, decay_madvises, right, 13, title);
+	col_decay_madvises.str_val = "madvises";
+
+	COL(decay_row, decay_purged, right, 13, title);
+	col_decay_purged.str_val = "purged";
 
 	/* Title row. */
 	emitter_table_row(emitter, &decay_row);
 
 	/* Dirty row. */
-	decay_type.str_val = "dirty:";
+	col_decay_type.str_val = "dirty:";
 
 	if (dirty_decay_ms >= 0) {
-		decay_time.type = emitter_type_ssize;
-		decay_time.ssize_val = dirty_decay_ms;
+		col_decay_time.type = emitter_type_ssize;
+		col_decay_time.ssize_val = dirty_decay_ms;
 	} else {
-		decay_time.type = emitter_type_title;
-		decay_time.str_val = "N/A";
+		col_decay_time.type = emitter_type_title;
+		col_decay_time.str_val = "N/A";
 	}
 
-	decay_npages.type = emitter_type_size;
-	decay_npages.size_val = pdirty;
+	col_decay_npages.type = emitter_type_size;
+	col_decay_npages.size_val = pdirty;
 
-	decay_sweeps.type = emitter_type_uint64;
-	decay_sweeps.uint64_val = dirty_npurge;
+	col_decay_sweeps.type = emitter_type_uint64;
+	col_decay_sweeps.uint64_val = dirty_npurge;
 
-	decay_madvises.type = emitter_type_uint64;
-	decay_madvises.uint64_val = dirty_nmadvise;
+	col_decay_madvises.type = emitter_type_uint64;
+	col_decay_madvises.uint64_val = dirty_nmadvise;
 
-	decay_purged.type = emitter_type_uint64;
-	decay_purged.uint64_val = dirty_purged;
+	col_decay_purged.type = emitter_type_uint64;
+	col_decay_purged.uint64_val = dirty_purged;
 
 	emitter_table_row(emitter, &decay_row);
 
 	/* Muzzy row. */
-	decay_type.str_val = "muzzy:";
+	col_decay_type.str_val = "muzzy:";
 
 	if (muzzy_decay_ms >= 0) {
-		decay_time.type = emitter_type_ssize;
-		decay_time.ssize_val = muzzy_decay_ms;
+		col_decay_time.type = emitter_type_ssize;
+		col_decay_time.ssize_val = muzzy_decay_ms;
 	} else {
-		decay_time.type = emitter_type_title;
-		decay_time.str_val = "N/A";
+		col_decay_time.type = emitter_type_title;
+		col_decay_time.str_val = "N/A";
 	}
 
-	decay_npages.type = emitter_type_size;
-	decay_npages.size_val = pmuzzy;
+	col_decay_npages.type = emitter_type_size;
+	col_decay_npages.size_val = pmuzzy;
 
-	decay_sweeps.type = emitter_type_uint64;
-	decay_sweeps.uint64_val = muzzy_npurge;
+	col_decay_sweeps.type = emitter_type_uint64;
+	col_decay_sweeps.uint64_val = muzzy_npurge;
 
-	decay_madvises.type = emitter_type_uint64;
-	decay_madvises.uint64_val = muzzy_nmadvise;
+	col_decay_madvises.type = emitter_type_uint64;
+	col_decay_madvises.uint64_val = muzzy_nmadvise;
 
-	decay_purged.type = emitter_type_uint64;
-	decay_purged.uint64_val = muzzy_purged;
+	col_decay_purged.type = emitter_type_uint64;
+	col_decay_purged.uint64_val = muzzy_purged;
 
 	emitter_table_row(emitter, &decay_row);
 
@@ -693,81 +800,89 @@ stats_arena_print(emitter_t *emitter, unsigned i, bool bins, bool large,
 	emitter_row_t alloc_count_row;
 	emitter_row_init(&alloc_count_row);
 
-	emitter_col_t alloc_count_title;
-	emitter_col_init(&alloc_count_title, &alloc_count_row);
-	alloc_count_title.justify = emitter_justify_left;
-	alloc_count_title.width = 25;
-	alloc_count_title.type = emitter_type_title;
-	alloc_count_title.str_val = "";
-
-	emitter_col_t alloc_count_allocated;
-	emitter_col_init(&alloc_count_allocated, &alloc_count_row);
-	alloc_count_allocated.justify = emitter_justify_right;
-	alloc_count_allocated.width = 12;
-	alloc_count_allocated.type = emitter_type_title;
-	alloc_count_allocated.str_val = "allocated";
-
-	emitter_col_t alloc_count_nmalloc;
-	emitter_col_init(&alloc_count_nmalloc, &alloc_count_row);
-	alloc_count_nmalloc.justify = emitter_justify_right;
-	alloc_count_nmalloc.width = 12;
-	alloc_count_nmalloc.type = emitter_type_title;
-	alloc_count_nmalloc.str_val = "nmalloc";
-
-	emitter_col_t alloc_count_ndalloc;
-	emitter_col_init(&alloc_count_ndalloc, &alloc_count_row);
-	alloc_count_ndalloc.justify = emitter_justify_right;
-	alloc_count_ndalloc.width = 12;
-	alloc_count_ndalloc.type = emitter_type_title;
-	alloc_count_ndalloc.str_val = "ndalloc";
-
-	emitter_col_t alloc_count_nrequests;
-	emitter_col_init(&alloc_count_nrequests, &alloc_count_row);
-	alloc_count_nrequests.justify = emitter_justify_right;
-	alloc_count_nrequests.width = 12;
-	alloc_count_nrequests.type = emitter_type_title;
-	alloc_count_nrequests.str_val = "nrequests";
+	COL(alloc_count_row, count_title, left, 21, title);
+	col_count_title.str_val = "";
+
+	COL(alloc_count_row, count_allocated, right, 16, title);
+	col_count_allocated.str_val = "allocated";
+
+	COL(alloc_count_row, count_nmalloc, right, 16, title);
+	col_count_nmalloc.str_val = "nmalloc";
+	COL(alloc_count_row, count_nmalloc_ps, right, 8, title);
+	col_count_nmalloc_ps.str_val = "(#/sec)";
+
+	COL(alloc_count_row, count_ndalloc, right, 16, title);
+	col_count_ndalloc.str_val = "ndalloc";
+	COL(alloc_count_row, count_ndalloc_ps, right, 8, title);
+	col_count_ndalloc_ps.str_val = "(#/sec)";
+
+	COL(alloc_count_row, count_nrequests, right, 16, title);
+	col_count_nrequests.str_val = "nrequests";
+	COL(alloc_count_row, count_nrequests_ps, right, 10, title);
+	col_count_nrequests_ps.str_val = "(#/sec)";
 
 	emitter_table_row(emitter, &alloc_count_row);
 
+	col_count_nmalloc_ps.type = emitter_type_uint64;
+	col_count_ndalloc_ps.type = emitter_type_uint64;
+	col_count_nrequests_ps.type = emitter_type_uint64;
+
 #define GET_AND_EMIT_ALLOC_STAT(small_or_large, name, valtype)		\
 	CTL_M2_GET("stats.arenas.0." #small_or_large "." #name, i,	\
 	    &small_or_large##_##name, valtype##_t);			\
 	emitter_json_kv(emitter, #name, emitter_type_##valtype,		\
 	    &small_or_large##_##name);					\
-	alloc_count_##name.type = emitter_type_##valtype;		\
-	alloc_count_##name.valtype##_val = small_or_large##_##name;
+	col_count_##name.type = emitter_type_##valtype;		\
+	col_count_##name.valtype##_val = small_or_large##_##name;
 
-	emitter_json_dict_begin(emitter, "small");
-	alloc_count_title.str_val = "small:";
+	emitter_json_object_kv_begin(emitter, "small");
+	col_count_title.str_val = "small:";
 
 	GET_AND_EMIT_ALLOC_STAT(small, allocated, size)
 	GET_AND_EMIT_ALLOC_STAT(small, nmalloc, uint64)
+	col_count_nmalloc_ps.uint64_val =
+	    rate_per_second(col_count_nmalloc.uint64_val, uptime);
 	GET_AND_EMIT_ALLOC_STAT(small, ndalloc, uint64)
+	col_count_ndalloc_ps.uint64_val =
+	    rate_per_second(col_count_ndalloc.uint64_val, uptime);
 	GET_AND_EMIT_ALLOC_STAT(small, nrequests, uint64)
+	col_count_nrequests_ps.uint64_val =
+	    rate_per_second(col_count_nrequests.uint64_val, uptime);
 
 	emitter_table_row(emitter, &alloc_count_row);
-	emitter_json_dict_end(emitter); /* Close "small". */
+	emitter_json_object_end(emitter); /* Close "small". */
 
-	emitter_json_dict_begin(emitter, "large");
-	alloc_count_title.str_val = "large:";
+	emitter_json_object_kv_begin(emitter, "large");
+	col_count_title.str_val = "large:";
 
 	GET_AND_EMIT_ALLOC_STAT(large, allocated, size)
 	GET_AND_EMIT_ALLOC_STAT(large, nmalloc, uint64)
+	col_count_nmalloc_ps.uint64_val =
+	    rate_per_second(col_count_nmalloc.uint64_val, uptime);
 	GET_AND_EMIT_ALLOC_STAT(large, ndalloc, uint64)
+	col_count_ndalloc_ps.uint64_val =
+	    rate_per_second(col_count_ndalloc.uint64_val, uptime);
 	GET_AND_EMIT_ALLOC_STAT(large, nrequests, uint64)
+	col_count_nrequests_ps.uint64_val =
+	    rate_per_second(col_count_nrequests.uint64_val, uptime);
 
 	emitter_table_row(emitter, &alloc_count_row);
-	emitter_json_dict_end(emitter); /* Close "large". */
+	emitter_json_object_end(emitter); /* Close "large". */
 
 #undef GET_AND_EMIT_ALLOC_STAT
 
 	/* Aggregated small + large stats are emitter only in table mode. */
-	alloc_count_title.str_val = "total:";
-	alloc_count_allocated.size_val = small_allocated + large_allocated;
-	alloc_count_nmalloc.uint64_val = small_nmalloc + large_nmalloc;
-	alloc_count_ndalloc.uint64_val = small_ndalloc + large_ndalloc;
-	alloc_count_nrequests.uint64_val = small_nrequests + large_nrequests;
+	col_count_title.str_val = "total:";
+	col_count_allocated.size_val = small_allocated + large_allocated;
+	col_count_nmalloc.uint64_val = small_nmalloc + large_nmalloc;
+	col_count_ndalloc.uint64_val = small_ndalloc + large_ndalloc;
+	col_count_nrequests.uint64_val = small_nrequests + large_nrequests;
+	col_count_nmalloc_ps.uint64_val =
+	    rate_per_second(col_count_nmalloc.uint64_val, uptime);
+	col_count_ndalloc_ps.uint64_val =
+	    rate_per_second(col_count_ndalloc.uint64_val, uptime);
+	col_count_nrequests_ps.uint64_val =
+	    rate_per_second(col_count_nrequests.uint64_val, uptime);
 	emitter_table_row(emitter, &alloc_count_row);
 
 	emitter_row_t mem_count_row;
@@ -776,14 +891,14 @@ stats_arena_print(emitter_t *emitter, unsigned i, bool bins, bool large,
 	emitter_col_t mem_count_title;
 	emitter_col_init(&mem_count_title, &mem_count_row);
 	mem_count_title.justify = emitter_justify_left;
-	mem_count_title.width = 25;
+	mem_count_title.width = 21;
 	mem_count_title.type = emitter_type_title;
 	mem_count_title.str_val = "";
 
 	emitter_col_t mem_count_val;
 	emitter_col_init(&mem_count_val, &mem_count_row);
 	mem_count_val.justify = emitter_justify_right;
-	mem_count_val.width = 12;
+	mem_count_val.width = 16;
 	mem_count_val.type = emitter_type_title;
 	mem_count_val.str_val = "";
 
@@ -809,16 +924,20 @@ stats_arena_print(emitter_t *emitter, unsigned i, bool bins, bool large,
 	GET_AND_EMIT_MEM_STAT(metadata_thp)
 	GET_AND_EMIT_MEM_STAT(tcache_bytes)
 	GET_AND_EMIT_MEM_STAT(resident)
+	GET_AND_EMIT_MEM_STAT(extent_avail)
 #undef GET_AND_EMIT_MEM_STAT
 
 	if (mutex) {
-		stats_arena_mutexes_print(emitter, i);
+		stats_arena_mutexes_print(emitter, i, uptime);
 	}
 	if (bins) {
-		stats_arena_bins_print(emitter, mutex, i);
+		stats_arena_bins_print(emitter, mutex, i, uptime);
 	}
 	if (large) {
-		stats_arena_lextents_print(emitter, i);
+		stats_arena_lextents_print(emitter, i, uptime);
+	}
+	if (extents) {
+		stats_arena_extents_print(emitter, i);
 	}
 }
 
@@ -891,6 +1010,8 @@ stats_general_print(emitter_t *emitter) {
 #define OPT_WRITE_UNSIGNED(name)					\
 	OPT_WRITE(name, uv, usz, emitter_type_unsigned)
 
+#define OPT_WRITE_SIZE_T(name)						\
+	OPT_WRITE(name, sv, ssz, emitter_type_size)
 #define OPT_WRITE_SSIZE_T(name)						\
 	OPT_WRITE(name, ssv, sssz, emitter_type_ssize)
 #define OPT_WRITE_SSIZE_T_MUTABLE(name, altname)			\
@@ -908,11 +1029,12 @@ stats_general_print(emitter_t *emitter) {
 	OPT_WRITE_CHAR_P("dss")
 	OPT_WRITE_UNSIGNED("narenas")
 	OPT_WRITE_CHAR_P("percpu_arena")
+	OPT_WRITE_SIZE_T("oversize_threshold")
 	OPT_WRITE_CHAR_P("metadata_thp")
 	OPT_WRITE_BOOL_MUTABLE("background_thread", "background_thread")
 	OPT_WRITE_SSIZE_T_MUTABLE("dirty_decay_ms", "arenas.dirty_decay_ms")
 	OPT_WRITE_SSIZE_T_MUTABLE("muzzy_decay_ms", "arenas.muzzy_decay_ms")
-	OPT_WRITE_UNSIGNED("lg_extent_max_active_fit")
+	OPT_WRITE_SIZE_T("lg_extent_max_active_fit")
 	OPT_WRITE_CHAR_P("junk")
 	OPT_WRITE_BOOL("zero")
 	OPT_WRITE_BOOL("utrace")
@@ -977,7 +1099,7 @@ stats_general_print(emitter_t *emitter) {
 	 * The json output sticks arena info into an "arenas" dict; the table
 	 * output puts them at the top-level.
 	 */
-	emitter_json_dict_begin(emitter, "arenas");
+	emitter_json_object_kv_begin(emitter, "arenas");
 
 	CTL_GET("arenas.narenas", &uv, unsigned);
 	emitter_kv(emitter, "narenas", "Arenas", emitter_type_unsigned, &uv);
@@ -1018,9 +1140,9 @@ stats_general_print(emitter_t *emitter) {
 	 * (not just omit the printing).
 	 */
 	if (emitter->output == emitter_output_json) {
-		emitter_json_arr_begin(emitter, "bin");
+		emitter_json_array_kv_begin(emitter, "bin");
 		for (unsigned i = 0; i < nbins; i++) {
-			emitter_json_arr_obj_begin(emitter);
+			emitter_json_object_begin(emitter);
 
 			CTL_M2_GET("arenas.bin.0.size", i, &sv, size_t);
 			emitter_json_kv(emitter, "size", emitter_type_size,
@@ -1034,9 +1156,13 @@ stats_general_print(emitter_t *emitter) {
 			emitter_json_kv(emitter, "slab_size", emitter_type_size,
 			    &sv);
 
-			emitter_json_arr_obj_end(emitter);
+			CTL_M2_GET("arenas.bin.0.nshards", i, &u32v, uint32_t);
+			emitter_json_kv(emitter, "nshards", emitter_type_uint32,
+			    &u32v);
+
+			emitter_json_object_end(emitter);
 		}
-		emitter_json_arr_end(emitter); /* Close "bin". */
+		emitter_json_array_end(emitter); /* Close "bin". */
 	}
 
 	unsigned nlextents;
@@ -1045,25 +1171,25 @@ stats_general_print(emitter_t *emitter) {
 	    emitter_type_unsigned, &nlextents);
 
 	if (emitter->output == emitter_output_json) {
-		emitter_json_arr_begin(emitter, "lextent");
+		emitter_json_array_kv_begin(emitter, "lextent");
 		for (unsigned i = 0; i < nlextents; i++) {
-			emitter_json_arr_obj_begin(emitter);
+			emitter_json_object_begin(emitter);
 
 			CTL_M2_GET("arenas.lextent.0.size", i, &sv, size_t);
 			emitter_json_kv(emitter, "size", emitter_type_size,
 			    &sv);
 
-			emitter_json_arr_obj_end(emitter);
+			emitter_json_object_end(emitter);
 		}
-		emitter_json_arr_end(emitter); /* Close "lextent". */
+		emitter_json_array_end(emitter); /* Close "lextent". */
 	}
 
-	emitter_json_dict_end(emitter); /* Close "arenas" */
+	emitter_json_object_end(emitter); /* Close "arenas" */
 }
 
 static void
 stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
-    bool unmerged, bool bins, bool large, bool mutex) {
+    bool unmerged, bool bins, bool large, bool mutex, bool extents) {
 	/*
 	 * These should be deleted.  We keep them around for a while, to aid in
 	 * the transition to the emitter code.
@@ -1095,7 +1221,7 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 	}
 
 	/* Generic global stats. */
-	emitter_json_dict_begin(emitter, "stats");
+	emitter_json_object_kv_begin(emitter, "stats");
 	emitter_json_kv(emitter, "allocated", emitter_type_size, &allocated);
 	emitter_json_kv(emitter, "active", emitter_type_size, &active);
 	emitter_json_kv(emitter, "metadata", emitter_type_size, &metadata);
@@ -1111,14 +1237,14 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 	    resident, mapped, retained);
 
 	/* Background thread stats. */
-	emitter_json_dict_begin(emitter, "background_thread");
+	emitter_json_object_kv_begin(emitter, "background_thread");
 	emitter_json_kv(emitter, "num_threads", emitter_type_size,
 	    &num_background_threads);
 	emitter_json_kv(emitter, "num_runs", emitter_type_uint64,
 	    &background_thread_num_runs);
 	emitter_json_kv(emitter, "run_interval", emitter_type_uint64,
 	    &background_thread_run_interval);
-	emitter_json_dict_end(emitter); /* Close "background_thread". */
+	emitter_json_object_end(emitter); /* Close "background_thread". */
 
 	emitter_table_printf(emitter, "Background threads: %zu, "
 	    "num_runs: %"FMTu64", run_interval: %"FMTu64" ns\n",
@@ -1130,30 +1256,33 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 		emitter_col_t name;
 		emitter_col_t col64[mutex_prof_num_uint64_t_counters];
 		emitter_col_t col32[mutex_prof_num_uint32_t_counters];
+		uint64_t uptime;
 
 		emitter_row_init(&row);
 		mutex_stats_init_cols(&row, "", &name, col64, col32);
 
 		emitter_table_row(emitter, &row);
-		emitter_json_dict_begin(emitter, "mutexes");
+		emitter_json_object_kv_begin(emitter, "mutexes");
+
+		CTL_M2_GET("stats.arenas.0.uptime", 0, &uptime, uint64_t);
 
 		for (int i = 0; i < mutex_prof_num_global_mutexes; i++) {
 			mutex_stats_read_global(global_mutex_names[i], &name,
-			    col64, col32);
-			emitter_json_dict_begin(emitter, global_mutex_names[i]);
+			    col64, col32, uptime);
+			emitter_json_object_kv_begin(emitter, global_mutex_names[i]);
 			mutex_stats_emit(emitter, &row, col64, col32);
-			emitter_json_dict_end(emitter);
+			emitter_json_object_end(emitter);
 		}
 
-		emitter_json_dict_end(emitter); /* Close "mutexes". */
+		emitter_json_object_end(emitter); /* Close "mutexes". */
 	}
 
-	emitter_json_dict_end(emitter); /* Close "stats". */
+	emitter_json_object_end(emitter); /* Close "stats". */
 
 	if (merged || destroyed || unmerged) {
 		unsigned narenas;
 
-		emitter_json_dict_begin(emitter, "stats.arenas");
+		emitter_json_object_kv_begin(emitter, "stats.arenas");
 
 		CTL_GET("arenas.narenas", &narenas, unsigned);
 		size_t mib[3];
@@ -1182,10 +1311,10 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 		if (merged && (ninitialized > 1 || !unmerged)) {
 			/* Print merged arena stats. */
 			emitter_table_printf(emitter, "Merged arenas stats:\n");
-			emitter_json_dict_begin(emitter, "merged");
+			emitter_json_object_kv_begin(emitter, "merged");
 			stats_arena_print(emitter, MALLCTL_ARENAS_ALL, bins,
-			    large, mutex);
-			emitter_json_dict_end(emitter); /* Close "merged". */
+			    large, mutex, extents);
+			emitter_json_object_end(emitter); /* Close "merged". */
 		}
 
 		/* Destroyed stats. */
@@ -1193,10 +1322,10 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 			/* Print destroyed arena stats. */
 			emitter_table_printf(emitter,
 			    "Destroyed arenas stats:\n");
-			emitter_json_dict_begin(emitter, "destroyed");
+			emitter_json_object_kv_begin(emitter, "destroyed");
 			stats_arena_print(emitter, MALLCTL_ARENAS_DESTROYED,
-			    bins, large, mutex);
-			emitter_json_dict_end(emitter); /* Close "destroyed". */
+			    bins, large, mutex, extents);
+			emitter_json_object_end(emitter); /* Close "destroyed". */
 		}
 
 		/* Unmerged stats. */
@@ -1206,18 +1335,18 @@ stats_print_helper(emitter_t *emitter, bool merged, bool destroyed,
 					char arena_ind_str[20];
 					malloc_snprintf(arena_ind_str,
 					    sizeof(arena_ind_str), "%u", i);
-					emitter_json_dict_begin(emitter,
+					emitter_json_object_kv_begin(emitter,
 					    arena_ind_str);
 					emitter_table_printf(emitter,
 					    "arenas[%s]:\n", arena_ind_str);
 					stats_arena_print(emitter, i, bins,
-					    large, mutex);
+					    large, mutex, extents);
 					/* Close "<arena-ind>". */
-					emitter_json_dict_end(emitter);
+					emitter_json_object_end(emitter);
 				}
 			}
 		}
-		emitter_json_dict_end(emitter); /* Close "stats.arenas". */
+		emitter_json_object_end(emitter); /* Close "stats.arenas". */
 	}
 }
 
@@ -1270,17 +1399,17 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 	    cbopaque);
 	emitter_begin(&emitter);
 	emitter_table_printf(&emitter, "___ Begin jemalloc statistics ___\n");
-	emitter_json_dict_begin(&emitter, "jemalloc");
+	emitter_json_object_kv_begin(&emitter, "jemalloc");
 
 	if (general) {
 		stats_general_print(&emitter);
 	}
 	if (config_stats) {
 		stats_print_helper(&emitter, merged, destroyed, unmerged,
-		    bins, large, mutex);
+		    bins, large, mutex, extents);
 	}
 
-	emitter_json_dict_end(&emitter); /* Closes the "jemalloc" dict. */
+	emitter_json_object_end(&emitter); /* Closes the "jemalloc" dict. */
 	emitter_table_printf(&emitter, "--- End jemalloc statistics ---\n");
 	emitter_end(&emitter);
 }
diff --git a/dep/jemalloc/src/sz.c b/dep/jemalloc/src/sz.c
index 9de77e45fff..8633fb05005 100644
--- a/dep/jemalloc/src/sz.c
+++ b/dep/jemalloc/src/sz.c
@@ -2,106 +2,63 @@
 #include "jemalloc/internal/sz.h"
 
 JEMALLOC_ALIGNED(CACHELINE)
-const size_t sz_pind2sz_tab[NPSIZES+1] = {
-#define PSZ_yes(lg_grp, ndelta, lg_delta)				\
-	(((ZU(1)<<lg_grp) + (ZU(ndelta)<<lg_delta))),
-#define PSZ_no(lg_grp, ndelta, lg_delta)
-#define SC(index, lg_grp, lg_delta, ndelta, psz, bin, pgs, lg_delta_lookup) \
-	PSZ_##psz(lg_grp, ndelta, lg_delta)
-	SIZE_CLASSES
-#undef PSZ_yes
-#undef PSZ_no
-#undef SC
-	(LARGE_MAXCLASS + PAGE)
-};
+size_t sz_pind2sz_tab[SC_NPSIZES+1];
+
+static void
+sz_boot_pind2sz_tab(const sc_data_t *sc_data) {
+	int pind = 0;
+	for (unsigned i = 0; i < SC_NSIZES; i++) {
+		const sc_t *sc = &sc_data->sc[i];
+		if (sc->psz) {
+			sz_pind2sz_tab[pind] = (ZU(1) << sc->lg_base)
+			    + (ZU(sc->ndelta) << sc->lg_delta);
+			pind++;
+		}
+	}
+	for (int i = pind; i <= (int)SC_NPSIZES; i++) {
+		sz_pind2sz_tab[pind] = sc_data->large_maxclass + PAGE;
+	}
+}
 
 JEMALLOC_ALIGNED(CACHELINE)
-const size_t sz_index2size_tab[NSIZES] = {
-#define SC(index, lg_grp, lg_delta, ndelta, psz, bin, pgs, lg_delta_lookup) \
-	((ZU(1)<<lg_grp) + (ZU(ndelta)<<lg_delta)),
-	SIZE_CLASSES
-#undef SC
-};
+size_t sz_index2size_tab[SC_NSIZES];
+
+static void
+sz_boot_index2size_tab(const sc_data_t *sc_data) {
+	for (unsigned i = 0; i < SC_NSIZES; i++) {
+		const sc_t *sc = &sc_data->sc[i];
+		sz_index2size_tab[i] = (ZU(1) << sc->lg_base)
+		    + (ZU(sc->ndelta) << (sc->lg_delta));
+	}
+}
 
+/*
+ * To keep this table small, we divide sizes by the tiny min size, which gives
+ * the smallest interval for which the result can change.
+ */
 JEMALLOC_ALIGNED(CACHELINE)
-const uint8_t sz_size2index_tab[] = {
-#if LG_TINY_MIN == 0
-/* The div module doesn't support division by 1. */
-#error "Unsupported LG_TINY_MIN"
-#define S2B_0(i)	i,
-#elif LG_TINY_MIN == 1
-#warning "Dangerous LG_TINY_MIN"
-#define S2B_1(i)	i,
-#elif LG_TINY_MIN == 2
-#warning "Dangerous LG_TINY_MIN"
-#define S2B_2(i)	i,
-#elif LG_TINY_MIN == 3
-#define S2B_3(i)	i,
-#elif LG_TINY_MIN == 4
-#define S2B_4(i)	i,
-#elif LG_TINY_MIN == 5
-#define S2B_5(i)	i,
-#elif LG_TINY_MIN == 6
-#define S2B_6(i)	i,
-#elif LG_TINY_MIN == 7
-#define S2B_7(i)	i,
-#elif LG_TINY_MIN == 8
-#define S2B_8(i)	i,
-#elif LG_TINY_MIN == 9
-#define S2B_9(i)	i,
-#elif LG_TINY_MIN == 10
-#define S2B_10(i)	i,
-#elif LG_TINY_MIN == 11
-#define S2B_11(i)	i,
-#else
-#error "Unsupported LG_TINY_MIN"
-#endif
-#if LG_TINY_MIN < 1
-#define S2B_1(i)	S2B_0(i) S2B_0(i)
-#endif
-#if LG_TINY_MIN < 2
-#define S2B_2(i)	S2B_1(i) S2B_1(i)
-#endif
-#if LG_TINY_MIN < 3
-#define S2B_3(i)	S2B_2(i) S2B_2(i)
-#endif
-#if LG_TINY_MIN < 4
-#define S2B_4(i)	S2B_3(i) S2B_3(i)
-#endif
-#if LG_TINY_MIN < 5
-#define S2B_5(i)	S2B_4(i) S2B_4(i)
-#endif
-#if LG_TINY_MIN < 6
-#define S2B_6(i)	S2B_5(i) S2B_5(i)
-#endif
-#if LG_TINY_MIN < 7
-#define S2B_7(i)	S2B_6(i) S2B_6(i)
-#endif
-#if LG_TINY_MIN < 8
-#define S2B_8(i)	S2B_7(i) S2B_7(i)
-#endif
-#if LG_TINY_MIN < 9
-#define S2B_9(i)	S2B_8(i) S2B_8(i)
-#endif
-#if LG_TINY_MIN < 10
-#define S2B_10(i)	S2B_9(i) S2B_9(i)
-#endif
-#if LG_TINY_MIN < 11
-#define S2B_11(i)	S2B_10(i) S2B_10(i)
-#endif
-#define S2B_no(i)
-#define SC(index, lg_grp, lg_delta, ndelta, psz, bin, pgs, lg_delta_lookup) \
-	S2B_##lg_delta_lookup(index)
-	SIZE_CLASSES
-#undef S2B_3
-#undef S2B_4
-#undef S2B_5
-#undef S2B_6
-#undef S2B_7
-#undef S2B_8
-#undef S2B_9
-#undef S2B_10
-#undef S2B_11
-#undef S2B_no
-#undef SC
-};
+uint8_t sz_size2index_tab[(SC_LOOKUP_MAXCLASS >> SC_LG_TINY_MIN) + 1];
+
+static void
+sz_boot_size2index_tab(const sc_data_t *sc_data) {
+	size_t dst_max = (SC_LOOKUP_MAXCLASS >> SC_LG_TINY_MIN) + 1;
+	size_t dst_ind = 0;
+	for (unsigned sc_ind = 0; sc_ind < SC_NSIZES && dst_ind < dst_max;
+	    sc_ind++) {
+		const sc_t *sc = &sc_data->sc[sc_ind];
+		size_t sz = (ZU(1) << sc->lg_base)
+		    + (ZU(sc->ndelta) << sc->lg_delta);
+		size_t max_ind = ((sz + (ZU(1) << SC_LG_TINY_MIN) - 1)
+				   >> SC_LG_TINY_MIN);
+		for (; dst_ind <= max_ind && dst_ind < dst_max; dst_ind++) {
+			sz_size2index_tab[dst_ind] = sc_ind;
+		}
+	}
+}
+
+void
+sz_boot(const sc_data_t *sc_data) {
+	sz_boot_pind2sz_tab(sc_data);
+	sz_boot_index2size_tab(sc_data);
+	sz_boot_size2index_tab(sc_data);
+}
diff --git a/dep/jemalloc/src/tcache.c b/dep/jemalloc/src/tcache.c
index a769a6b17bc..e7b970d9045 100644
--- a/dep/jemalloc/src/tcache.c
+++ b/dep/jemalloc/src/tcache.c
@@ -4,7 +4,7 @@
 
 #include "jemalloc/internal/assert.h"
 #include "jemalloc/internal/mutex.h"
-#include "jemalloc/internal/size_classes.h"
+#include "jemalloc/internal/sc.h"
 
 /******************************************************************************/
 /* Data. */
@@ -41,7 +41,7 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache) {
 	szind_t binind = tcache->next_gc_bin;
 
 	cache_bin_t *tbin;
-	if (binind < NBINS) {
+	if (binind < SC_NBINS) {
 		tbin = tcache_small_bin_get(tcache, binind);
 	} else {
 		tbin = tcache_large_bin_get(tcache, binind);
@@ -50,7 +50,7 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache) {
 		/*
 		 * Flush (ceiling) 3/4 of the objects below the low water mark.
 		 */
-		if (binind < NBINS) {
+		if (binind < SC_NBINS) {
 			tcache_bin_flush_small(tsd, tcache, tbin, binind,
 			    tbin->ncached - tbin->low_water + (tbin->low_water
 			    >> 2));
@@ -72,7 +72,7 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache) {
 		 * Increase fill count by 2X for small bins.  Make sure
 		 * lg_fill_div stays greater than 0.
 		 */
-		if (binind < NBINS && tcache->lg_fill_div[binind] > 1) {
+		if (binind < SC_NBINS && tcache->lg_fill_div[binind] > 1) {
 			tcache->lg_fill_div[binind]--;
 		}
 	}
@@ -100,28 +100,68 @@ tcache_alloc_small_hard(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
 	return ret;
 }
 
+/* Enabled with --enable-extra-size-check. */
+#ifdef JEMALLOC_EXTRA_SIZE_CHECK
+static void
+tbin_extents_lookup_size_check(tsdn_t *tsdn, cache_bin_t *tbin, szind_t binind,
+    size_t nflush, extent_t **extents){
+	rtree_ctx_t rtree_ctx_fallback;
+	rtree_ctx_t *rtree_ctx = tsdn_rtree_ctx(tsdn, &rtree_ctx_fallback);
+
+	/*
+	 * Verify that the items in the tcache all have the correct size; this
+	 * is useful for catching sized deallocation bugs, also to fail early
+	 * instead of corrupting metadata.  Since this can be turned on for opt
+	 * builds, avoid the branch in the loop.
+	 */
+	szind_t szind;
+	size_t sz_sum = binind * nflush;
+	for (unsigned i = 0 ; i < nflush; i++) {
+		rtree_extent_szind_read(tsdn, &extents_rtree,
+		    rtree_ctx, (uintptr_t)*(tbin->avail - 1 - i), true,
+		    &extents[i], &szind);
+		sz_sum -= szind;
+	}
+	if (sz_sum != 0) {
+		malloc_printf("<jemalloc>: size mismatch in thread cache "
+		    "detected, likely caused by sized deallocation bugs by "
+		    "application. Abort.\n");
+		abort();
+	}
+}
+#endif
+
 void
 tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, cache_bin_t *tbin,
     szind_t binind, unsigned rem) {
 	bool merged_stats = false;
 
-	assert(binind < NBINS);
+	assert(binind < SC_NBINS);
 	assert((cache_bin_sz_t)rem <= tbin->ncached);
 
 	arena_t *arena = tcache->arena;
 	assert(arena != NULL);
 	unsigned nflush = tbin->ncached - rem;
 	VARIABLE_ARRAY(extent_t *, item_extent, nflush);
+
+#ifndef JEMALLOC_EXTRA_SIZE_CHECK
 	/* Look up extent once per item. */
 	for (unsigned i = 0 ; i < nflush; i++) {
 		item_extent[i] = iealloc(tsd_tsdn(tsd), *(tbin->avail - 1 - i));
 	}
-
+#else
+	tbin_extents_lookup_size_check(tsd_tsdn(tsd), tbin, binind, nflush,
+	    item_extent);
+#endif
 	while (nflush > 0) {
 		/* Lock the arena bin associated with the first object. */
 		extent_t *extent = item_extent[0];
-		arena_t *bin_arena = extent_arena_get(extent);
-		bin_t *bin = &bin_arena->bins[binind];
+		unsigned bin_arena_ind = extent_arena_ind_get(extent);
+		arena_t *bin_arena = arena_get(tsd_tsdn(tsd), bin_arena_ind,
+		    false);
+		unsigned binshard = extent_binshard_get(extent);
+		assert(binshard < bin_infos[binind].n_shards);
+		bin_t *bin = &bin_arena->bins[binind].bin_shards[binshard];
 
 		if (config_prof && bin_arena == arena) {
 			if (arena_prof_accum(tsd_tsdn(tsd), arena,
@@ -132,8 +172,7 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, cache_bin_t *tbin,
 		}
 
 		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
-		if (config_stats && bin_arena == arena) {
-			assert(!merged_stats);
+		if (config_stats && bin_arena == arena && !merged_stats) {
 			merged_stats = true;
 			bin->stats.nflushes++;
 			bin->stats.nrequests += tbin->tstats.nrequests;
@@ -145,9 +184,10 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, cache_bin_t *tbin,
 			extent = item_extent[i];
 			assert(ptr != NULL && extent != NULL);
 
-			if (extent_arena_get(extent) == bin_arena) {
+			if (extent_arena_ind_get(extent) == bin_arena_ind
+			    && extent_binshard_get(extent) == binshard) {
 				arena_dalloc_bin_junked_locked(tsd_tsdn(tsd),
-				    bin_arena, extent, ptr);
+				    bin_arena, bin, binind, extent, ptr);
 			} else {
 				/*
 				 * This object was allocated via a different
@@ -169,8 +209,9 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, cache_bin_t *tbin,
 		 * The flush loop didn't happen to flush to this thread's
 		 * arena, so the stats didn't get merged.  Manually do so now.
 		 */
-		bin_t *bin = &arena->bins[binind];
-		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
+		unsigned binshard;
+		bin_t *bin = arena_bin_choose_lock(tsd_tsdn(tsd), arena, binind,
+		    &binshard);
 		bin->stats.nflushes++;
 		bin->stats.nrequests += tbin->tstats.nrequests;
 		tbin->tstats.nrequests = 0;
@@ -193,50 +234,63 @@ tcache_bin_flush_large(tsd_t *tsd, cache_bin_t *tbin, szind_t binind,
 	assert(binind < nhbins);
 	assert((cache_bin_sz_t)rem <= tbin->ncached);
 
-	arena_t *arena = tcache->arena;
-	assert(arena != NULL);
+	arena_t *tcache_arena = tcache->arena;
+	assert(tcache_arena != NULL);
 	unsigned nflush = tbin->ncached - rem;
 	VARIABLE_ARRAY(extent_t *, item_extent, nflush);
+
+#ifndef JEMALLOC_EXTRA_SIZE_CHECK
 	/* Look up extent once per item. */
 	for (unsigned i = 0 ; i < nflush; i++) {
 		item_extent[i] = iealloc(tsd_tsdn(tsd), *(tbin->avail - 1 - i));
 	}
-
+#else
+	tbin_extents_lookup_size_check(tsd_tsdn(tsd), tbin, binind, nflush,
+	    item_extent);
+#endif
 	while (nflush > 0) {
 		/* Lock the arena associated with the first object. */
 		extent_t *extent = item_extent[0];
-		arena_t *locked_arena = extent_arena_get(extent);
-		UNUSED bool idump;
+		unsigned locked_arena_ind = extent_arena_ind_get(extent);
+		arena_t *locked_arena = arena_get(tsd_tsdn(tsd),
+		    locked_arena_ind, false);
+		bool idump;
 
 		if (config_prof) {
 			idump = false;
 		}
 
-		malloc_mutex_lock(tsd_tsdn(tsd), &locked_arena->large_mtx);
+		bool lock_large = !arena_is_auto(locked_arena);
+		if (lock_large) {
+			malloc_mutex_lock(tsd_tsdn(tsd), &locked_arena->large_mtx);
+		}
 		for (unsigned i = 0; i < nflush; i++) {
 			void *ptr = *(tbin->avail - 1 - i);
 			assert(ptr != NULL);
 			extent = item_extent[i];
-			if (extent_arena_get(extent) == locked_arena) {
+			if (extent_arena_ind_get(extent) == locked_arena_ind) {
 				large_dalloc_prep_junked_locked(tsd_tsdn(tsd),
 				    extent);
 			}
 		}
-		if ((config_prof || config_stats) && locked_arena == arena) {
+		if ((config_prof || config_stats) &&
+		    (locked_arena == tcache_arena)) {
 			if (config_prof) {
-				idump = arena_prof_accum(tsd_tsdn(tsd), arena,
-				    tcache->prof_accumbytes);
+				idump = arena_prof_accum(tsd_tsdn(tsd),
+				    tcache_arena, tcache->prof_accumbytes);
 				tcache->prof_accumbytes = 0;
 			}
 			if (config_stats) {
 				merged_stats = true;
 				arena_stats_large_nrequests_add(tsd_tsdn(tsd),
-				    &arena->stats, binind,
+				    &tcache_arena->stats, binind,
 				    tbin->tstats.nrequests);
 				tbin->tstats.nrequests = 0;
 			}
 		}
-		malloc_mutex_unlock(tsd_tsdn(tsd), &locked_arena->large_mtx);
+		if (lock_large) {
+			malloc_mutex_unlock(tsd_tsdn(tsd), &locked_arena->large_mtx);
+		}
 
 		unsigned ndeferred = 0;
 		for (unsigned i = 0; i < nflush; i++) {
@@ -244,7 +298,7 @@ tcache_bin_flush_large(tsd_t *tsd, cache_bin_t *tbin, szind_t binind,
 			extent = item_extent[i];
 			assert(ptr != NULL && extent != NULL);
 
-			if (extent_arena_get(extent) == locked_arena) {
+			if (extent_arena_ind_get(extent) == locked_arena_ind) {
 				large_dalloc_finish(tsd_tsdn(tsd), extent);
 			} else {
 				/*
@@ -270,8 +324,8 @@ tcache_bin_flush_large(tsd_t *tsd, cache_bin_t *tbin, szind_t binind,
 		 * The flush loop didn't happen to flush to this thread's
 		 * arena, so the stats didn't get merged.  Manually do so now.
 		 */
-		arena_stats_large_nrequests_add(tsd_tsdn(tsd), &arena->stats,
-		    binind, tbin->tstats.nrequests);
+		arena_stats_large_nrequests_add(tsd_tsdn(tsd),
+		    &tcache_arena->stats, binind, tbin->tstats.nrequests);
 		tbin->tstats.nrequests = 0;
 	}
 
@@ -363,10 +417,10 @@ tcache_init(tsd_t *tsd, tcache_t *tcache, void *avail_stack) {
 
 	size_t stack_offset = 0;
 	assert((TCACHE_NSLOTS_SMALL_MAX & 1U) == 0);
-	memset(tcache->bins_small, 0, sizeof(cache_bin_t) * NBINS);
-	memset(tcache->bins_large, 0, sizeof(cache_bin_t) * (nhbins - NBINS));
+	memset(tcache->bins_small, 0, sizeof(cache_bin_t) * SC_NBINS);
+	memset(tcache->bins_large, 0, sizeof(cache_bin_t) * (nhbins - SC_NBINS));
 	unsigned i = 0;
-	for (; i < NBINS; i++) {
+	for (; i < SC_NBINS; i++) {
 		tcache->lg_fill_div[i] = 1;
 		stack_offset += tcache_bin_info[i].ncached_max * sizeof(void *);
 		/*
@@ -458,7 +512,7 @@ static void
 tcache_flush_cache(tsd_t *tsd, tcache_t *tcache) {
 	assert(tcache->arena != NULL);
 
-	for (unsigned i = 0; i < NBINS; i++) {
+	for (unsigned i = 0; i < SC_NBINS; i++) {
 		cache_bin_t *tbin = tcache_small_bin_get(tcache, i);
 		tcache_bin_flush_small(tsd, tcache, tbin, i, 0);
 
@@ -466,7 +520,7 @@ tcache_flush_cache(tsd_t *tsd, tcache_t *tcache) {
 			assert(tbin->tstats.nrequests == 0);
 		}
 	}
-	for (unsigned i = NBINS; i < nhbins; i++) {
+	for (unsigned i = SC_NBINS; i < nhbins; i++) {
 		cache_bin_t *tbin = tcache_large_bin_get(tcache, i);
 		tcache_bin_flush_large(tsd, tbin, i, 0, tcache);
 
@@ -491,6 +545,7 @@ tcache_flush(tsd_t *tsd) {
 static void
 tcache_destroy(tsd_t *tsd, tcache_t *tcache, bool tsd_tcache) {
 	tcache_flush_cache(tsd, tcache);
+	arena_t *arena = tcache->arena;
 	tcache_arena_dissociate(tsd_tsdn(tsd), tcache);
 
 	if (tsd_tcache) {
@@ -503,6 +558,23 @@ tcache_destroy(tsd_t *tsd, tcache_t *tcache, bool tsd_tcache) {
 		/* Release both the tcache struct and avail array. */
 		idalloctm(tsd_tsdn(tsd), tcache, NULL, NULL, true, true);
 	}
+
+	/*
+	 * The deallocation and tcache flush above may not trigger decay since
+	 * we are on the tcache shutdown path (potentially with non-nominal
+	 * tsd).  Manually trigger decay to avoid pathological cases.  Also
+	 * include arena 0 because the tcache array is allocated from it.
+	 */
+	arena_decay(tsd_tsdn(tsd), arena_get(tsd_tsdn(tsd), 0, false),
+	    false, false);
+
+	if (arena_nthreads_get(arena, false) == 0 &&
+	    !background_thread_enabled()) {
+		/* Force purging when no threads assigned to the arena anymore. */
+		arena_decay(tsd_tsdn(tsd), arena, false, true);
+	} else {
+		arena_decay(tsd_tsdn(tsd), arena, false, false);
+	}
 }
 
 /* For auto tcache (embedded in TSD) only. */
@@ -532,10 +604,10 @@ tcache_stats_merge(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena) {
 	cassert(config_stats);
 
 	/* Merge and reset tcache stats. */
-	for (i = 0; i < NBINS; i++) {
-		bin_t *bin = &arena->bins[i];
+	for (i = 0; i < SC_NBINS; i++) {
 		cache_bin_t *tbin = tcache_small_bin_get(tcache, i);
-		malloc_mutex_lock(tsdn, &bin->lock);
+		unsigned binshard;
+		bin_t *bin = arena_bin_choose_lock(tsdn, arena, i, &binshard);
 		bin->stats.nrequests += tbin->tstats.nrequests;
 		malloc_mutex_unlock(tsdn, &bin->lock);
 		tbin->tstats.nrequests = 0;
@@ -614,23 +686,32 @@ label_return:
 }
 
 static tcache_t *
-tcaches_elm_remove(tsd_t *tsd, tcaches_t *elm) {
+tcaches_elm_remove(tsd_t *tsd, tcaches_t *elm, bool allow_reinit) {
 	malloc_mutex_assert_owner(tsd_tsdn(tsd), &tcaches_mtx);
 
 	if (elm->tcache == NULL) {
 		return NULL;
 	}
 	tcache_t *tcache = elm->tcache;
-	elm->tcache = NULL;
+	if (allow_reinit) {
+		elm->tcache = TCACHES_ELM_NEED_REINIT;
+	} else {
+		elm->tcache = NULL;
+	}
+
+	if (tcache == TCACHES_ELM_NEED_REINIT) {
+		return NULL;
+	}
 	return tcache;
 }
 
 void
 tcaches_flush(tsd_t *tsd, unsigned ind) {
 	malloc_mutex_lock(tsd_tsdn(tsd), &tcaches_mtx);
-	tcache_t *tcache = tcaches_elm_remove(tsd, &tcaches[ind]);
+	tcache_t *tcache = tcaches_elm_remove(tsd, &tcaches[ind], true);
 	malloc_mutex_unlock(tsd_tsdn(tsd), &tcaches_mtx);
 	if (tcache != NULL) {
+		/* Destroy the tcache; recreate in tcaches_get() if needed. */
 		tcache_destroy(tsd, tcache, false);
 	}
 }
@@ -639,7 +720,7 @@ void
 tcaches_destroy(tsd_t *tsd, unsigned ind) {
 	malloc_mutex_lock(tsd_tsdn(tsd), &tcaches_mtx);
 	tcaches_t *elm = &tcaches[ind];
-	tcache_t *tcache = tcaches_elm_remove(tsd, elm);
+	tcache_t *tcache = tcaches_elm_remove(tsd, elm, false);
 	elm->next = tcaches_avail;
 	tcaches_avail = elm;
 	malloc_mutex_unlock(tsd_tsdn(tsd), &tcaches_mtx);
@@ -652,8 +733,8 @@ bool
 tcache_boot(tsdn_t *tsdn) {
 	/* If necessary, clamp opt_lg_tcache_max. */
 	if (opt_lg_tcache_max < 0 || (ZU(1) << opt_lg_tcache_max) <
-	    SMALL_MAXCLASS) {
-		tcache_maxclass = SMALL_MAXCLASS;
+	    SC_SMALL_MAXCLASS) {
+		tcache_maxclass = SC_SMALL_MAXCLASS;
 	} else {
 		tcache_maxclass = (ZU(1) << opt_lg_tcache_max);
 	}
@@ -673,7 +754,7 @@ tcache_boot(tsdn_t *tsdn) {
 	}
 	stack_nelms = 0;
 	unsigned i;
-	for (i = 0; i < NBINS; i++) {
+	for (i = 0; i < SC_NBINS; i++) {
 		if ((bin_infos[i].nregs << 1) <= TCACHE_NSLOTS_SMALL_MIN) {
 			tcache_bin_info[i].ncached_max =
 			    TCACHE_NSLOTS_SMALL_MIN;
diff --git a/dep/jemalloc/src/hooks.c b/dep/jemalloc/src/test_hooks.c
index 6266ecd47fe..ace00d9c469 100644
--- a/dep/jemalloc/src/hooks.c
+++ b/dep/jemalloc/src/test_hooks.c
@@ -6,7 +6,7 @@
  * from outside the generated library, so that we can use them in test code.
  */
 JEMALLOC_EXPORT
-void (*hooks_arena_new_hook)() = NULL;
+void (*test_hooks_arena_new_hook)() = NULL;
 
 JEMALLOC_EXPORT
-void (*hooks_libc_hook)() = NULL;
+void (*test_hooks_libc_hook)() = NULL;
diff --git a/dep/jemalloc/src/tsd.c b/dep/jemalloc/src/tsd.c
index c1430682dd5..d5fb4d6f82f 100644
--- a/dep/jemalloc/src/tsd.c
+++ b/dep/jemalloc/src/tsd.c
@@ -12,6 +12,10 @@
 static unsigned ncleanups;
 static malloc_tsd_cleanup_t cleanups[MALLOC_TSD_CLEANUPS_MAX];
 
+/* TSD_INITIALIZER triggers "-Wmissing-field-initializer" */
+JEMALLOC_DIAGNOSTIC_PUSH
+JEMALLOC_DIAGNOSTIC_IGNORE_MISSING_STRUCT_FIELD_INITIALIZERS
+
 #ifdef JEMALLOC_MALLOC_THREAD_CLEANUP
 __thread tsd_t JEMALLOC_TLS_MODEL tsd_tls = TSD_INITIALIZER;
 __thread bool JEMALLOC_TLS_MODEL tsd_initialized = false;
@@ -41,6 +45,7 @@ tsd_init_head_t	tsd_init_head = {
 	ql_head_initializer(blocks),
 	MALLOC_MUTEX_INITIALIZER
 };
+
 tsd_wrapper_t tsd_boot_wrapper = {
 	false,
 	TSD_INITIALIZER
@@ -48,17 +53,164 @@ tsd_wrapper_t tsd_boot_wrapper = {
 bool tsd_booted = false;
 #endif
 
+JEMALLOC_DIAGNOSTIC_POP
+
+/******************************************************************************/
+
+/* A list of all the tsds in the nominal state. */
+typedef ql_head(tsd_t) tsd_list_t;
+static tsd_list_t tsd_nominal_tsds = ql_head_initializer(tsd_nominal_tsds);
+static malloc_mutex_t tsd_nominal_tsds_lock;
+
+/* How many slow-path-enabling features are turned on. */
+static atomic_u32_t tsd_global_slow_count = ATOMIC_INIT(0);
+
+static bool
+tsd_in_nominal_list(tsd_t *tsd) {
+	tsd_t *tsd_list;
+	bool found = false;
+	/*
+	 * We don't know that tsd is nominal; it might not be safe to get data
+	 * out of it here.
+	 */
+	malloc_mutex_lock(TSDN_NULL, &tsd_nominal_tsds_lock);
+	ql_foreach(tsd_list, &tsd_nominal_tsds, TSD_MANGLE(tcache).tsd_link) {
+		if (tsd == tsd_list) {
+			found = true;
+			break;
+		}
+	}
+	malloc_mutex_unlock(TSDN_NULL, &tsd_nominal_tsds_lock);
+	return found;
+}
+
+static void
+tsd_add_nominal(tsd_t *tsd) {
+	assert(!tsd_in_nominal_list(tsd));
+	assert(tsd_state_get(tsd) <= tsd_state_nominal_max);
+	ql_elm_new(tsd, TSD_MANGLE(tcache).tsd_link);
+	malloc_mutex_lock(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+	ql_tail_insert(&tsd_nominal_tsds, tsd, TSD_MANGLE(tcache).tsd_link);
+	malloc_mutex_unlock(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+}
+
+static void
+tsd_remove_nominal(tsd_t *tsd) {
+	assert(tsd_in_nominal_list(tsd));
+	assert(tsd_state_get(tsd) <= tsd_state_nominal_max);
+	malloc_mutex_lock(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+	ql_remove(&tsd_nominal_tsds, tsd, TSD_MANGLE(tcache).tsd_link);
+	malloc_mutex_unlock(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+}
+
+static void
+tsd_force_recompute(tsdn_t *tsdn) {
+	/*
+	 * The stores to tsd->state here need to synchronize with the exchange
+	 * in tsd_slow_update.
+	 */
+	atomic_fence(ATOMIC_RELEASE);
+	malloc_mutex_lock(tsdn, &tsd_nominal_tsds_lock);
+	tsd_t *remote_tsd;
+	ql_foreach(remote_tsd, &tsd_nominal_tsds, TSD_MANGLE(tcache).tsd_link) {
+		assert(tsd_atomic_load(&remote_tsd->state, ATOMIC_RELAXED)
+		    <= tsd_state_nominal_max);
+		tsd_atomic_store(&remote_tsd->state, tsd_state_nominal_recompute,
+		    ATOMIC_RELAXED);
+	}
+	malloc_mutex_unlock(tsdn, &tsd_nominal_tsds_lock);
+}
+
+void
+tsd_global_slow_inc(tsdn_t *tsdn) {
+	atomic_fetch_add_u32(&tsd_global_slow_count, 1, ATOMIC_RELAXED);
+	/*
+	 * We unconditionally force a recompute, even if the global slow count
+	 * was already positive.  If we didn't, then it would be possible for us
+	 * to return to the user, have the user synchronize externally with some
+	 * other thread, and then have that other thread not have picked up the
+	 * update yet (since the original incrementing thread might still be
+	 * making its way through the tsd list).
+	 */
+	tsd_force_recompute(tsdn);
+}
+
+void tsd_global_slow_dec(tsdn_t *tsdn) {
+	atomic_fetch_sub_u32(&tsd_global_slow_count, 1, ATOMIC_RELAXED);
+	/* See the note in ..._inc(). */
+	tsd_force_recompute(tsdn);
+}
+
+static bool
+tsd_local_slow(tsd_t *tsd) {
+	return !tsd_tcache_enabled_get(tsd)
+	    || tsd_reentrancy_level_get(tsd) > 0;
+}
+
+bool
+tsd_global_slow() {
+	return atomic_load_u32(&tsd_global_slow_count, ATOMIC_RELAXED) > 0;
+}
 
 /******************************************************************************/
 
+static uint8_t
+tsd_state_compute(tsd_t *tsd) {
+	if (!tsd_nominal(tsd)) {
+		return tsd_state_get(tsd);
+	}
+	/* We're in *a* nominal state; but which one? */
+	if (malloc_slow || tsd_local_slow(tsd) || tsd_global_slow()) {
+		return tsd_state_nominal_slow;
+	} else {
+		return tsd_state_nominal;
+	}
+}
+
 void
 tsd_slow_update(tsd_t *tsd) {
-	if (tsd_nominal(tsd)) {
-		if (malloc_slow || !tsd_tcache_enabled_get(tsd) ||
-		    tsd_reentrancy_level_get(tsd) > 0) {
-			tsd->state = tsd_state_nominal_slow;
+	uint8_t old_state;
+	do {
+		uint8_t new_state = tsd_state_compute(tsd);
+		old_state = tsd_atomic_exchange(&tsd->state, new_state,
+		    ATOMIC_ACQUIRE);
+	} while (old_state == tsd_state_nominal_recompute);
+}
+
+void
+tsd_state_set(tsd_t *tsd, uint8_t new_state) {
+	/* Only the tsd module can change the state *to* recompute. */
+	assert(new_state != tsd_state_nominal_recompute);
+	uint8_t old_state = tsd_atomic_load(&tsd->state, ATOMIC_RELAXED);
+	if (old_state > tsd_state_nominal_max) {
+		/*
+		 * Not currently in the nominal list, but it might need to be
+		 * inserted there.
+		 */
+		assert(!tsd_in_nominal_list(tsd));
+		tsd_atomic_store(&tsd->state, new_state, ATOMIC_RELAXED);
+		if (new_state <= tsd_state_nominal_max) {
+			tsd_add_nominal(tsd);
+		}
+	} else {
+		/*
+		 * We're currently nominal.  If the new state is non-nominal,
+		 * great; we take ourselves off the list and just enter the new
+		 * state.
+		 */
+		assert(tsd_in_nominal_list(tsd));
+		if (new_state > tsd_state_nominal_max) {
+			tsd_remove_nominal(tsd);
+			tsd_atomic_store(&tsd->state, new_state,
+			    ATOMIC_RELAXED);
 		} else {
-			tsd->state = tsd_state_nominal;
+			/*
+			 * This is the tricky case.  We're transitioning from
+			 * one nominal state to another.  The caller can't know
+			 * about any races that are occuring at the same time,
+			 * so we always have to recompute no matter what.
+			 */
+			tsd_slow_update(tsd);
 		}
 	}
 }
@@ -87,6 +239,7 @@ tsd_data_init(tsd_t *tsd) {
 static void
 assert_tsd_data_cleanup_done(tsd_t *tsd) {
 	assert(!tsd_nominal(tsd));
+	assert(!tsd_in_nominal_list(tsd));
 	assert(*tsd_arenap_get_unsafe(tsd) == NULL);
 	assert(*tsd_iarenap_get_unsafe(tsd) == NULL);
 	assert(*tsd_arenas_tdata_bypassp_get_unsafe(tsd) == true);
@@ -97,8 +250,8 @@ assert_tsd_data_cleanup_done(tsd_t *tsd) {
 
 static bool
 tsd_data_init_nocleanup(tsd_t *tsd) {
-	assert(tsd->state == tsd_state_reincarnated ||
-	    tsd->state == tsd_state_minimal_initialized);
+	assert(tsd_state_get(tsd) == tsd_state_reincarnated ||
+	    tsd_state_get(tsd) == tsd_state_minimal_initialized);
 	/*
 	 * During reincarnation, there is no guarantee that the cleanup function
 	 * will be called (deallocation may happen after all tsd destructors).
@@ -117,27 +270,33 @@ tsd_t *
 tsd_fetch_slow(tsd_t *tsd, bool minimal) {
 	assert(!tsd_fast(tsd));
 
-	if (tsd->state == tsd_state_nominal_slow) {
-		/* On slow path but no work needed. */
-		assert(malloc_slow || !tsd_tcache_enabled_get(tsd) ||
-		    tsd_reentrancy_level_get(tsd) > 0 ||
-		    *tsd_arenas_tdata_bypassp_get(tsd));
-	} else if (tsd->state == tsd_state_uninitialized) {
+	if (tsd_state_get(tsd) == tsd_state_nominal_slow) {
+		/*
+		 * On slow path but no work needed.  Note that we can't
+		 * necessarily *assert* that we're slow, because we might be
+		 * slow because of an asynchronous modification to global state,
+		 * which might be asynchronously modified *back*.
+		 */
+	} else if (tsd_state_get(tsd) == tsd_state_nominal_recompute) {
+		tsd_slow_update(tsd);
+	} else if (tsd_state_get(tsd) == tsd_state_uninitialized) {
 		if (!minimal) {
-			tsd->state = tsd_state_nominal;
-			tsd_slow_update(tsd);
-			/* Trigger cleanup handler registration. */
-			tsd_set(tsd);
-			tsd_data_init(tsd);
+			if (tsd_booted) {
+				tsd_state_set(tsd, tsd_state_nominal);
+				tsd_slow_update(tsd);
+				/* Trigger cleanup handler registration. */
+				tsd_set(tsd);
+				tsd_data_init(tsd);
+			}
 		} else {
-			tsd->state = tsd_state_minimal_initialized;
+			tsd_state_set(tsd, tsd_state_minimal_initialized);
 			tsd_set(tsd);
 			tsd_data_init_nocleanup(tsd);
 		}
-	} else if (tsd->state == tsd_state_minimal_initialized) {
+	} else if (tsd_state_get(tsd) == tsd_state_minimal_initialized) {
 		if (!minimal) {
 			/* Switch to fully initialized. */
-			tsd->state = tsd_state_nominal;
+			tsd_state_set(tsd, tsd_state_nominal);
 			assert(*tsd_reentrancy_levelp_get(tsd) >= 1);
 			(*tsd_reentrancy_levelp_get(tsd))--;
 			tsd_slow_update(tsd);
@@ -145,12 +304,12 @@ tsd_fetch_slow(tsd_t *tsd, bool minimal) {
 		} else {
 			assert_tsd_data_cleanup_done(tsd);
 		}
-	} else if (tsd->state == tsd_state_purgatory) {
-		tsd->state = tsd_state_reincarnated;
+	} else if (tsd_state_get(tsd) == tsd_state_purgatory) {
+		tsd_state_set(tsd, tsd_state_reincarnated);
 		tsd_set(tsd);
 		tsd_data_init_nocleanup(tsd);
 	} else {
-		assert(tsd->state == tsd_state_reincarnated);
+		assert(tsd_state_get(tsd) == tsd_state_reincarnated);
 	}
 
 	return tsd;
@@ -214,7 +373,7 @@ void
 tsd_cleanup(void *arg) {
 	tsd_t *tsd = (tsd_t *)arg;
 
-	switch (tsd->state) {
+	switch (tsd_state_get(tsd)) {
 	case tsd_state_uninitialized:
 		/* Do nothing. */
 		break;
@@ -232,7 +391,7 @@ tsd_cleanup(void *arg) {
 	case tsd_state_nominal:
 	case tsd_state_nominal_slow:
 		tsd_do_data_cleanup(tsd);
-		tsd->state = tsd_state_purgatory;
+		tsd_state_set(tsd, tsd_state_purgatory);
 		tsd_set(tsd);
 		break;
 	case tsd_state_purgatory:
@@ -260,6 +419,10 @@ malloc_tsd_boot0(void) {
 	tsd_t *tsd;
 
 	ncleanups = 0;
+	if (malloc_mutex_init(&tsd_nominal_tsds_lock, "tsd_nominal_tsds_lock",
+	    WITNESS_RANK_OMIT, malloc_mutex_rank_exclusive)) {
+		return NULL;
+	}
 	if (tsd_boot0()) {
 		return NULL;
 	}
@@ -310,7 +473,7 @@ _tls_callback(HINSTANCE hinstDLL, DWORD fdwReason, LPVOID lpvReserved) {
 #    pragma comment(linker, "/INCLUDE:_tls_callback")
 #  else
 #    pragma comment(linker, "/INCLUDE:_tls_used")
-#    pragma comment(linker, "/INCLUDE:tls_callback")
+#    pragma comment(linker, "/INCLUDE:" STRINGIFY(tls_callback) )
 #  endif
 #  pragma section(".CRT$XLY",long,read)
 #endif
@@ -349,3 +512,23 @@ tsd_init_finish(tsd_init_head_t *head, tsd_init_block_t *block) {
 	malloc_mutex_unlock(TSDN_NULL, &head->lock);
 }
 #endif
+
+void
+tsd_prefork(tsd_t *tsd) {
+	malloc_mutex_prefork(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+}
+
+void
+tsd_postfork_parent(tsd_t *tsd) {
+	malloc_mutex_postfork_parent(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+}
+
+void
+tsd_postfork_child(tsd_t *tsd) {
+	malloc_mutex_postfork_child(tsd_tsdn(tsd), &tsd_nominal_tsds_lock);
+	ql_new(&tsd_nominal_tsds);
+
+	if (tsd_state_get(tsd) <= tsd_state_nominal_max) {
+		tsd_add_nominal(tsd);
+	}
+}