Merge branch.

Update ChangeLog for 4.2.1.
Fix rallocx() sampling code to not eagerly commit sampler update.
2026-07-23 05:03:10 +00:00 · 2016-06-08 11:41:24 -07:00 · 2016-06-08 10:20:10 -07:00 · 2016-06-08 10:14:25 -07:00 · 2016-06-08 10:12:38 -07:00 · 2016-06-08 10:10:08 -07:00
83 changed files with 5285 additions and 2582 deletions
--- a/4
+++ b/4
@@ -1,10 +1,10 @@
 Unless otherwise specified, files in the jemalloc source distribution are
 subject to the following license:
 --------------------------------------------------------------------------------
-Copyright (C) 2002-2015 Jason Evans <jasone@canonware.com>.
+Copyright (C) 2002-2016 Jason Evans <jasone@canonware.com>.
 All rights reserved.
 Copyright (C) 2007-2012 Mozilla Foundation.  All rights reserved.
-Copyright (C) 2009-2015 Facebook, Inc.  All rights reserved.
+Copyright (C) 2009-2016 Facebook, Inc.  All rights reserved.

 Redistribution and use in source and binary forms, with or without
 modification, are permitted provided that the following conditions are met:
--- a/59
+++ b/59
@@ -4,6 +4,61 @@ brevity.  Much more detail can be found in the git revision history:

    https://github.com/jemalloc/jemalloc

+* 4.2.1 (June 8, 2016)
+
+  Bug fixes:
+  - Fix bootstrapping issues for configurations that require allocation during
+    tsd initialization (e.g. --disable-tls).  (@cferris1000, @jasone)
+  - Fix gettimeofday() version of nstime_update().  (@ronawho)
+  - Fix Valgrind regressions in calloc() and chunk_alloc_wrapper().  (@ronawho)
+  - Fix potential VM map fragmentation regression.  (@jasone)
+  - Fix opt_zero-triggered in-place huge reallocation zeroing.  (@jasone)
+  - Fix heap profiling context leaks in reallocation edge cases.  (@jasone)
+
+* 4.2.0 (May 12, 2016)
+
+  New features:
+  - Add the arena.<i>.reset mallctl, which makes it possible to discard all of
+    an arena's allocations in a single operation.  (@jasone@)
+  - Add the stats.retained and stats.arenas.<i>.retained statistics.  (@jasone)
+  - Add the --with-version configure option.  (@jasone)
+  - Support --with-lg-page values larger than actual page size.  (@jasone)
+
+  Optimizations:
+  - Use pairing heaps rather than red-black trees for various hot data
+    structures.  (@djwatson, @jasone)
+  - Streamline fast paths of rtree operations.  (@jasone)
+  - Optimize the fast paths of calloc() and [m,d,sd]allocx().  (@jasone)
+  - Decommit unused virtual memory if the OS does not overcommit.  (@jasone)
+  - Specify MAP_NORESERVE on Linux if [heuristic] overcommit is active, in order
+    to avoid unfortunate interactions during fork(2).  (@jasone)
+
+  Bug fixes:
+  - Fix chunk accounting related to triggering gdump profiles.  (@jasone)
+  - Link against librt for clock_gettime(2) if glibc < 2.17.  (@jasone)
+  - Scale leak report summary according to sampling probability.  (@jasone)
+
+* 4.1.1 (May 3, 2016)
+
+  This bugfix release resolves a variety of mostly minor issues, though the
+  bitmap fix is critical for 64-bit Windows.
+
+  Bug fixes:
+  - Fix the linear scan version of bitmap_sfu() to shift by the proper amount
+    even when sizeof(long) is not the same as sizeof(void *), as on 64-bit
+    Windows.  (@jasone)
+  - Fix hashing functions to avoid unaligned memory accesses (and resulting
+    crashes).  This is relevant at least to some ARM-based platforms.
+    (@rkmisra)
+  - Fix fork()-related lock rank ordering reversals.  These reversals were
+    unlikely to cause deadlocks in practice except when heap profiling was
+    enabled and active.  (@jasone)
+  - Fix various chunk leaks in OOM code paths.  (@jasone)
+  - Fix malloc_stats_print() to print opt.narenas correctly.  (@jasone)
+  - Fix MSVC-specific build/test issues.  (@rustyx, @yuslepukhin)
+  - Fix a variety of test failures that were due to test fragility rather than
+    core bugs.  (@jasone)
+
 * 4.1.0 (February 28, 2016)

  This release is primarily about optimizations, but it also incorporates a lot
@@ -59,14 +114,14 @@ brevity.  Much more detail can be found in the git revision history:
  Bug fixes:
  - Fix stats.cactive accounting regression.  (@rustyx, @jasone)
  - Handle unaligned keys in hash().  This caused problems for some ARM systems.
-    (@jasone, Christopher Ferris)
+    (@jasone, @cferris1000)
  - Refactor arenas array.  In addition to fixing a fork-related deadlock, this
    makes arena lookups faster and simpler.  (@jasone)
  - Move retained memory allocation out of the default chunk allocation
    function, to a location that gets executed even if the application installs
    a custom chunk allocation function.  This resolves a virtual memory leak.
    (@buchgr)
-  - Fix a potential tsd cleanup leak.  (Christopher Ferris, @jasone)
+  - Fix a potential tsd cleanup leak.  (@cferris1000, @jasone)
  - Fix run quantization.  In practice this bug had no impact unless
    applications requested memory with alignment exceeding one page.
    (@jasone, @djwatson)
--- a/4
+++ b/4
@@ -35,6 +35,10 @@ any of the following arguments (not a definitive list) to 'configure':
    will cause files to be installed into /usr/local/include, /usr/local/lib,
    and /usr/local/man.

+--with-version=<major>.<minor>.<bugfix>-<nrev>-g<gid>
+    Use the specified version string rather than trying to generate one (if in
+    a git repository) or use existing the VERSION file (if present).
+
 --with-rpath=<colon-separated-rpath>
    Embed one or more library paths, so that libjemalloc can find the libraries
    it is linked to.  This works only on ELF-based systems.
--- a/Makefile.in
+++ b/Makefile.in
@@ -28,7 +28,6 @@ CFLAGS := @CFLAGS@
 LDFLAGS := @LDFLAGS@
 EXTRA_LDFLAGS := @EXTRA_LDFLAGS@
 LIBS := @LIBS@
-TESTLIBS := @TESTLIBS@
 RPATH_EXTRA := @RPATH_EXTRA@
 SO := @so@
 IMPORTLIB := @importlib@
@@ -103,7 +102,8 @@ C_SRCS := $(srcroot)src/jemalloc.c \
 	$(srcroot)src/tcache.c \
 	$(srcroot)src/ticker.c \
 	$(srcroot)src/tsd.c \
-	$(srcroot)src/util.c
+	$(srcroot)src/util.c \
+	$(srcroot)src/witness.c
 ifeq ($(enable_valgrind), 1)
 C_SRCS += $(srcroot)src/valgrind.c
 endif
@@ -134,10 +134,14 @@ C_TESTLIB_SRCS := $(srcroot)test/src/btalloc.c $(srcroot)test/src/btalloc_0.c \
 	$(srcroot)test/src/SFMT.c $(srcroot)test/src/test.c \
 	$(srcroot)test/src/thd.c $(srcroot)test/src/timer.c
 C_UTIL_INTEGRATION_SRCS := $(srcroot)src/nstime.c $(srcroot)src/util.c
-TESTS_UNIT := $(srcroot)test/unit/atomic.c \
+TESTS_UNIT := \
+	$(srcroot)test/unit/a0.c \
+	$(srcroot)test/unit/arena_reset.c \
+	$(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/bitmap.c \
 	$(srcroot)test/unit/ckh.c \
 	$(srcroot)test/unit/decay.c \
+	$(srcroot)test/unit/fork.c \
 	$(srcroot)test/unit/hash.c \
 	$(srcroot)test/unit/junk.c \
 	$(srcroot)test/unit/junk_alloc.c \
@@ -147,6 +151,7 @@ TESTS_UNIT := $(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/math.c \
 	$(srcroot)test/unit/mq.c \
 	$(srcroot)test/unit/mtx.c \
+	$(srcroot)test/unit/ph.c \
 	$(srcroot)test/unit/prng.c \
 	$(srcroot)test/unit/prof_accum.c \
 	$(srcroot)test/unit/prof_active.c \
@@ -168,6 +173,7 @@ TESTS_UNIT := $(srcroot)test/unit/atomic.c \
 	$(srcroot)test/unit/nstime.c \
 	$(srcroot)test/unit/tsd.c \
 	$(srcroot)test/unit/util.c \
+	$(srcroot)test/unit/witness.c \
 	$(srcroot)test/unit/zero.c
 TESTS_INTEGRATION := $(srcroot)test/integration/aligned_alloc.c \
 	$(srcroot)test/integration/allocated.c \
@@ -289,15 +295,15 @@ $(STATIC_LIBS):

 $(objroot)test/unit/%$(EXE): $(objroot)test/unit/%.$(O) $(TESTS_UNIT_LINK_OBJS) $(C_JET_OBJS) $(C_TESTLIB_UNIT_OBJS)
 	@mkdir -p $(@D)
-	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(TESTLIBS) $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(EXTRA_LDFLAGS)

 $(objroot)test/integration/%$(EXE): $(objroot)test/integration/%.$(O) $(C_TESTLIB_INTEGRATION_OBJS) $(C_UTIL_INTEGRATION_OBJS) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB)
 	@mkdir -p $(@D)
-	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(filter -lpthread,$(LIBS))) -lm $(TESTLIBS) $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(filter -lpthread,$(LIBS))) -lm $(EXTRA_LDFLAGS)

 $(objroot)test/stress/%$(EXE): $(objroot)test/stress/%.$(O) $(C_JET_OBJS) $(C_TESTLIB_STRESS_OBJS) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB)
 	@mkdir -p $(@D)
-	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(TESTLIBS) $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(EXTRA_LDFLAGS)

 build_lib_shared: $(DSOS)
 build_lib_static: $(STATIC_LIBS)
--- a/configure.ac
+++ b/configure.ac
@@ -141,6 +141,7 @@ if test "x$CFLAGS" = "x" ; then
    JE_CFLAGS_APPEND([-Wall])
    JE_CFLAGS_APPEND([-Werror=declaration-after-statement])
    JE_CFLAGS_APPEND([-Wshorten-64-to-32])
+    JE_CFLAGS_APPEND([-Wsign-compare])
    JE_CFLAGS_APPEND([-pipe])
    JE_CFLAGS_APPEND([-g3])
  elif test "x$je_cv_msvc" = "xyes" ; then
@@ -304,6 +305,7 @@ case "${host}" in
  *-*-freebsd*)
 	CFLAGS="$CFLAGS"
 	abi="elf"
+	AC_DEFINE([JEMALLOC_SYSCTL_VM_OVERCOMMIT], [ ])
 	AC_DEFINE([JEMALLOC_PURGE_MADVISE_FREE], [ ])
 	force_lazy_lock="1"
 	;;
@@ -328,6 +330,7 @@ case "${host}" in
 	CPPFLAGS="$CPPFLAGS -D_GNU_SOURCE"
 	abi="elf"
 	AC_DEFINE([JEMALLOC_HAS_ALLOCA_H])
+	AC_DEFINE([JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY], [ ])
 	AC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED], [ ])
 	AC_DEFINE([JEMALLOC_THREADED_INIT], [ ])
 	AC_DEFINE([JEMALLOC_USE_CXX_THROW], [ ])
@@ -1172,27 +1175,36 @@ dnl ============================================================================
 dnl jemalloc configuration.
 dnl 

-dnl Set VERSION if source directory is inside a git repository.
-if test "x`test ! \"${srcroot}\" && cd \"${srcroot}\"; git rev-parse --is-inside-work-tree 2>/dev/null`" = "xtrue" ; then
-  dnl Pattern globs aren't powerful enough to match both single- and
-  dnl double-digit version numbers, so iterate over patterns to support up to
-  dnl version 99.99.99 without any accidental matches.
-  rm -f "${objroot}VERSION"
-  for pattern in ['[0-9].[0-9].[0-9]' '[0-9].[0-9].[0-9][0-9]' \
-                 '[0-9].[0-9][0-9].[0-9]' '[0-9].[0-9][0-9].[0-9][0-9]' \
-                 '[0-9][0-9].[0-9].[0-9]' '[0-9][0-9].[0-9].[0-9][0-9]' \
-                 '[0-9][0-9].[0-9][0-9].[0-9]' \
-                 '[0-9][0-9].[0-9][0-9].[0-9][0-9]']; do
-    if test ! -e "${objroot}VERSION" ; then
-      (test ! "${srcroot}" && cd "${srcroot}"; git describe --long --abbrev=40 --match="${pattern}") > "${objroot}VERSION.tmp" 2>/dev/null
-      if test $? -eq 0 ; then
-        mv "${objroot}VERSION.tmp" "${objroot}VERSION"
-        break
-      fi
+AC_ARG_WITH([version],
+  [AS_HELP_STRING([--with-version=<major>.<minor>.<bugfix>-<nrev>-g<gid>],
+   [Version string])],
+  [
+    echo "${with_version}" | grep ['^[0-9]\+\.[0-9]\+\.[0-9]\+-[0-9]\+-g[0-9a-f]\+$'] 2>&1 1>/dev/null
+    if test $? -ne 0 ; then
+      AC_MSG_ERROR([${with_version} does not match <major>.<minor>.<bugfix>-<nrev>-g<gid>])
    fi
-  done
-fi
-rm -f "${objroot}VERSION.tmp"
+    echo "$with_version" > "${objroot}VERSION"
+  ], [
+    dnl Set VERSION if source directory is inside a git repository.
+    if test "x`test ! \"${srcroot}\" && cd \"${srcroot}\"; git rev-parse --is-inside-work-tree 2>/dev/null`" = "xtrue" ; then
+      dnl Pattern globs aren't powerful enough to match both single- and
+      dnl double-digit version numbers, so iterate over patterns to support up
+      dnl to version 99.99.99 without any accidental matches.
+      for pattern in ['[0-9].[0-9].[0-9]' '[0-9].[0-9].[0-9][0-9]' \
+                     '[0-9].[0-9][0-9].[0-9]' '[0-9].[0-9][0-9].[0-9][0-9]' \
+                     '[0-9][0-9].[0-9].[0-9]' '[0-9][0-9].[0-9].[0-9][0-9]' \
+                     '[0-9][0-9].[0-9][0-9].[0-9]' \
+                     '[0-9][0-9].[0-9][0-9].[0-9][0-9]']; do
+        (test ! "${srcroot}" && cd "${srcroot}"; git describe --long --abbrev=40 --match="${pattern}") > "${objroot}VERSION.tmp" 2>/dev/null
+        if test $? -eq 0 ; then
+          mv "${objroot}VERSION.tmp" "${objroot}VERSION"
+          break
+        fi
+      done
+    fi
+    rm -f "${objroot}VERSION.tmp"
+  ])
+
 if test ! -e "${objroot}VERSION" ; then
  if test ! -e "${srcroot}VERSION" ; then
    AC_MSG_RESULT(
@@ -1229,13 +1241,8 @@ fi

 CPPFLAGS="$CPPFLAGS -D_REENTRANT"

-dnl Check whether clock_gettime(2) is in libc or librt.  This function is only
-dnl used in test code, so save the result to TESTLIBS to avoid poluting LIBS.
-SAVED_LIBS="${LIBS}"
-LIBS=
-AC_SEARCH_LIBS([clock_gettime], [rt], [TESTLIBS="${LIBS}"])
-AC_SUBST([TESTLIBS])
-LIBS="${SAVED_LIBS}"
+dnl Check whether clock_gettime(2) is in libc or librt.
+AC_SEARCH_LIBS([clock_gettime], [rt])

 dnl Check if the GNU-specific secure_getenv function exists.
 AC_CHECK_FUNC([secure_getenv],
@@ -1741,7 +1748,6 @@ AC_MSG_RESULT([CPPFLAGS           : ${CPPFLAGS}])
 AC_MSG_RESULT([LDFLAGS            : ${LDFLAGS}])
 AC_MSG_RESULT([EXTRA_LDFLAGS      : ${EXTRA_LDFLAGS}])
 AC_MSG_RESULT([LIBS               : ${LIBS}])
-AC_MSG_RESULT([TESTLIBS           : ${TESTLIBS}])
 AC_MSG_RESULT([RPATH_EXTRA        : ${RPATH_EXTRA}])
 AC_MSG_RESULT([])
 AC_MSG_RESULT([XSLTPROC           : ${XSLTPROC}])
--- a/doc/jemalloc.xml.in
+++ b/doc/jemalloc.xml.in
@@ -540,8 +540,8 @@ for (i = 0; i < nbins; i++) {
    are smaller than four times the page size, large size classes are smaller
    than the chunk size (see the <link
    linkend="opt.lg_chunk"><mallctl>opt.lg_chunk</mallctl></link> option), and
-    huge size classes extend from the chunk size up to one size class less than
-    the full address space size.</para>
+    huge size classes extend from the chunk size up to the largest size class
+    that does not exceed <constant>PTRDIFF_MAX</constant>.</para>

    <para>Allocations are packed tightly together, which can be an issue for
    multi-threaded applications.  If you need to assure that allocations do not
@@ -659,7 +659,7 @@ for (i = 0; i < nbins; i++) {
          <entry>[1280 KiB, 1536 KiB, 1792 KiB]</entry>
        </row>
        <row>
-          <entry morerows="6">Huge</entry>
+          <entry morerows="8">Huge</entry>
          <entry>256 KiB</entry>
          <entry>[2 MiB]</entry>
        </row>
@@ -687,6 +687,14 @@ for (i = 0; i < nbins; i++) {
          <entry>...</entry>
          <entry>...</entry>
        </row>
+        <row>
+          <entry>512 PiB</entry>
+          <entry>[2560 PiB, 3 EiB, 3584 PiB, 4 EiB]</entry>
+        </row>
+        <row>
+          <entry>1 EiB</entry>
+          <entry>[5 EiB, 6 EiB, 7 EiB]</entry>
+        </row>
      </tbody>
      </tgroup>
    </table>
@@ -1016,7 +1024,7 @@ for (i = 0; i < nbins; i++) {
        allocate memory during application initialization and then deadlock
        internally when jemalloc in turn calls
        <function>atexit<parameter/></function>, so this option is not
-        univerally usable (though the application can register its own
+        universally usable (though the application can register its own
        <function>atexit<parameter/></function> function with equivalent
        functionality).  Therefore, this option should only be used with care;
        it is primarily intended as a performance tuning aid during application
@@ -1320,7 +1328,7 @@ malloc_conf = "xmalloc:true";]]></programlisting>
        option.  Note that <function>atexit<parameter/></function> may allocate
        memory during application initialization and then deadlock internally
        when jemalloc in turn calls <function>atexit<parameter/></function>, so
-        this option is not univerally usable (though the application can
+        this option is not universally usable (though the application can
        register its own <function>atexit<parameter/></function> function with
        equivalent functionality).  This option is disabled by
        default.</para></listitem>
@@ -1550,6 +1558,23 @@ malloc_conf = "xmalloc:true";]]></programlisting>
        details.</para></listitem>
      </varlistentry>

+      <varlistentry id="arena.i.reset">
+        <term>
+          <mallctl>arena.&lt;i&gt;.reset</mallctl>
+          (<type>void</type>)
+          <literal>--</literal>
+        </term>
+        <listitem><para>Discard all of the arena's extant allocations.  This
+        interface can only be used with arenas created via <link
+        linkend="arenas.extend"><mallctl>arenas.extend</mallctl></link>.  None
+        of the arena's discarded/cached allocations may accessed afterward.  As
+        part of this requirement, all thread caches which were used to
+        allocate/deallocate in conjunction with the arena must be flushed
+        beforehand.  This interface cannot be used if running inside Valgrind,
+        nor if the <link linkend="opt.quarantine">quarantine</link> size is
+        non-zero.</para></listitem>
+      </varlistentry>
+
      <varlistentry id="arena.i.dss">
        <term>
          <mallctl>arena.&lt;i&gt;.dss</mallctl>
@@ -2062,7 +2087,7 @@ typedef struct {
          [<option>--enable-prof</option>]
        </term>
        <listitem><para>Average number of bytes allocated between
-        inverval-based profile dumps.  See the
+        interval-based profile dumps.  See the
        <link
        linkend="opt.lg_prof_interval"><mallctl>opt.lg_prof_interval</mallctl></link>
        option for additional information.</para></listitem>
@@ -2161,6 +2186,25 @@ typedef struct {
        linkend="stats.resident"><mallctl>stats.resident</mallctl></link>.</para></listitem>
      </varlistentry>

+      <varlistentry id="stats.retained">
+        <term>
+          <mallctl>stats.retained</mallctl>
+          (<type>size_t</type>)
+          <literal>r-</literal>
+          [<option>--enable-stats</option>]
+        </term>
+        <listitem><para>Total number of bytes in virtual memory mappings that
+        were retained rather than being returned to the operating system via
+        e.g. <citerefentry><refentrytitle>munmap</refentrytitle>
+        <manvolnum>2</manvolnum></citerefentry>.  Retained virtual memory is
+        typically untouched, decommitted, or purged, so it has no strongly
+        associated physical memory (see <link
+        linkend="arena.i.chunk_hooks">chunk hooks</link> for details).  Retained
+        memory is excluded from mapped memory statistics, e.g. <link
+        linkend="stats.mapped"><mallctl>stats.mapped</mallctl></link>.
+        </para></listitem>
+      </varlistentry>
+
      <varlistentry id="stats.arenas.i.dss">
        <term>
          <mallctl>stats.arenas.&lt;i&gt;.dss</mallctl>
@@ -2241,6 +2285,18 @@ typedef struct {
        <listitem><para>Number of mapped bytes.</para></listitem>
      </varlistentry>

+      <varlistentry id="stats.arenas.i.retained">
+        <term>
+          <mallctl>stats.arenas.&lt;i&gt;.retained</mallctl>
+          (<type>size_t</type>)
+          <literal>r-</literal>
+          [<option>--enable-stats</option>]
+        </term>
+        <listitem><para>Number of retained bytes.  See <link
+        linkend="stats.retained"><mallctl>stats.retained</mallctl></link> for
+        details.</para></listitem>
+      </varlistentry>
+
      <varlistentry id="stats.arenas.i.metadata.mapped">
        <term>
          <mallctl>stats.arenas.&lt;i&gt;.metadata.mapped</mallctl>
--- a/include/jemalloc/internal/arena.h
+++ b/include/jemalloc/internal/arena.h
@@ -36,6 +36,7 @@ typedef enum {
 #define	DECAY_NTICKS_PER_UPDATE	1000

 typedef struct arena_runs_dirty_link_s arena_runs_dirty_link_t;
+typedef struct arena_avail_links_s arena_avail_links_t;
 typedef struct arena_run_s arena_run_t;
 typedef struct arena_chunk_map_bits_s arena_chunk_map_bits_t;
 typedef struct arena_chunk_map_misc_s arena_chunk_map_misc_t;
@@ -153,13 +154,13 @@ struct arena_runs_dirty_link_s {
 */
 struct arena_chunk_map_misc_s {
 	/*
-	 * Linkage for run trees.  There are two disjoint uses:
+	 * Linkage for run heaps.  There are two disjoint uses:
 	 *
-	 * 1) arena_t's runs_avail tree.
+	 * 1) arena_t's runs_avail heaps.
 	 * 2) arena_run_t conceptually uses this linkage for in-use non-full
 	 *    runs, rather than directly embedding linkage.
 	 */
-	rb_node(arena_chunk_map_misc_t)		rb_link;
+	phn(arena_chunk_map_misc_t)		ph_link;

 	union {
 		/* Linkage for list of dirty runs. */
@@ -175,7 +176,7 @@ struct arena_chunk_map_misc_s {
 		arena_run_t			run;
 	};
 };
-typedef rb_tree(arena_chunk_map_misc_t) arena_run_tree_t;
+typedef ph(arena_chunk_map_misc_t) arena_run_heap_t;
 #endif /* JEMALLOC_ARENA_STRUCTS_A */

 #ifdef JEMALLOC_ARENA_STRUCTS_B
@@ -272,13 +273,13 @@ struct arena_bin_s {
 	arena_run_t		*runcur;

 	/*
-	 * Tree of non-full runs.  This tree is used when looking for an
+	 * Heap of non-full runs.  This heap is used when looking for an
 	 * existing run when runcur is no longer usable.  We choose the
 	 * non-full run that is lowest in memory; this policy tends to keep
 	 * objects packed well, and it can also help reduce the number of
 	 * almost-empty chunks.
 	 */
-	arena_run_tree_t	runs;
+	arena_run_heap_t	runs;

 	/* Bin statistics. */
 	malloc_bin_stats_t	stats;
@@ -289,10 +290,18 @@ struct arena_s {
 	unsigned		ind;

 	/*
-	 * Number of threads currently assigned to this arena.  This field is
-	 * synchronized via atomic operations.
+	 * Number of threads currently assigned to this arena, synchronized via
+	 * atomic operations.  Each thread has two distinct assignments, one for
+	 * application-serving allocation, and the other for internal metadata
+	 * allocation.  Internal metadata must not be allocated from arenas
+	 * created via the arenas.extend mallctl, because the arena.<i>.reset
+	 * mallctl indiscriminately discards all allocations for the affected
+	 * arena.
+	 *
+	 *   0: Application allocation.
+	 *   1: Internal metadata allocation.
 	 */
-	unsigned		nthreads;
+	unsigned		nthreads[2];

 	/*
 	 * There are three classes of arena operations from a locking
@@ -321,6 +330,10 @@ struct arena_s {

 	dss_prec_t		dss_prec;

+
+	/* Extant arena chunks. */
+	ql_head(extent_node_t)	achunks;
+
 	/*
 	 * In order to avoid rapid chunk allocation/deallocation when an arena
 	 * oscillates right on the cusp of needing a new chunk, cache the most
@@ -457,10 +470,10 @@ struct arena_s {
 	arena_bin_t		bins[NBINS];

 	/*
-	 * Quantized address-ordered trees of this arena's available runs.  The
-	 * trees are used for first-best-fit run allocation.
+	 * Quantized address-ordered heaps of this arena's available runs.  The
+	 * heaps are used for first-best-fit run allocation.
 	 */
-	arena_run_tree_t	runs_avail[1]; /* Dynamically sized. */
+	arena_run_heap_t	runs_avail[1]; /* Dynamically sized. */
 };

 /* Used in conjunction with tsd for fast arena-related context lookup. */
@@ -505,25 +518,28 @@ void	arena_chunk_cache_maybe_insert(arena_t *arena, extent_node_t *node,
    bool cache);
 void	arena_chunk_cache_maybe_remove(arena_t *arena, extent_node_t *node,
    bool cache);
-extent_node_t	*arena_node_alloc(arena_t *arena);
-void	arena_node_dalloc(arena_t *arena, extent_node_t *node);
-void	*arena_chunk_alloc_huge(arena_t *arena, size_t usize, size_t alignment,
-    bool *zero);
-void	arena_chunk_dalloc_huge(arena_t *arena, void *chunk, size_t usize);
-void	arena_chunk_ralloc_huge_similar(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize);
-void	arena_chunk_ralloc_huge_shrink(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize);
-bool	arena_chunk_ralloc_huge_expand(arena_t *arena, void *chunk,
-    size_t oldsize, size_t usize, bool *zero);
-ssize_t	arena_lg_dirty_mult_get(arena_t *arena);
-bool	arena_lg_dirty_mult_set(arena_t *arena, ssize_t lg_dirty_mult);
-ssize_t	arena_decay_time_get(arena_t *arena);
-bool	arena_decay_time_set(arena_t *arena, ssize_t decay_time);
-void	arena_maybe_purge(arena_t *arena);
-void	arena_purge(arena_t *arena, bool all);
-void	arena_tcache_fill_small(tsd_t *tsd, arena_t *arena, tcache_bin_t *tbin,
-    szind_t binind, uint64_t prof_accumbytes);
+extent_node_t	*arena_node_alloc(tsdn_t *tsdn, arena_t *arena);
+void	arena_node_dalloc(tsdn_t *tsdn, arena_t *arena, extent_node_t *node);
+void	*arena_chunk_alloc_huge(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool *zero);
+void	arena_chunk_dalloc_huge(tsdn_t *tsdn, arena_t *arena, void *chunk,
+    size_t usize);
+void	arena_chunk_ralloc_huge_similar(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize);
+void	arena_chunk_ralloc_huge_shrink(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize);
+bool	arena_chunk_ralloc_huge_expand(tsdn_t *tsdn, arena_t *arena,
+    void *chunk, size_t oldsize, size_t usize, bool *zero);
+ssize_t	arena_lg_dirty_mult_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_lg_dirty_mult_set(tsdn_t *tsdn, arena_t *arena,
+    ssize_t lg_dirty_mult);
+ssize_t	arena_decay_time_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_decay_time_set(tsdn_t *tsdn, arena_t *arena, ssize_t decay_time);
+void	arena_purge(tsdn_t *tsdn, arena_t *arena, bool all);
+void	arena_maybe_purge(tsdn_t *tsdn, arena_t *arena);
+void	arena_reset(tsd_t *tsd, arena_t *arena);
+void	arena_tcache_fill_small(tsdn_t *tsdn, arena_t *arena,
+    tcache_bin_t *tbin, szind_t binind, uint64_t prof_accumbytes);
 void	arena_alloc_junk_small(void *ptr, arena_bin_info_t *bin_info,
    bool zero);
 #ifdef JEMALLOC_JET
@@ -536,17 +552,18 @@ extern arena_dalloc_junk_small_t *arena_dalloc_junk_small;
 void	arena_dalloc_junk_small(void *ptr, arena_bin_info_t *bin_info);
 #endif
 void	arena_quarantine_junk_small(void *ptr, size_t usize);
-void	*arena_malloc_large(tsd_t *tsd, arena_t *arena, szind_t ind, bool zero);
-void	*arena_malloc_hard(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
-    bool zero, tcache_t *tcache);
-void	*arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize,
+void	*arena_malloc_large(tsdn_t *tsdn, arena_t *arena, szind_t ind,
+    bool zero);
+void	*arena_malloc_hard(tsdn_t *tsdn, arena_t *arena, size_t size,
+    szind_t ind, bool zero);
+void	*arena_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
    size_t alignment, bool zero, tcache_t *tcache);
-void	arena_prof_promoted(const void *ptr, size_t size);
-void	arena_dalloc_bin_junked_locked(arena_t *arena, arena_chunk_t *chunk,
-    void *ptr, arena_chunk_map_bits_t *bitselm);
-void	arena_dalloc_bin(arena_t *arena, arena_chunk_t *chunk, void *ptr,
-    size_t pageind, arena_chunk_map_bits_t *bitselm);
-void	arena_dalloc_small(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
+void	arena_prof_promoted(tsdn_t *tsdn, const void *ptr, size_t size);
+void	arena_dalloc_bin_junked_locked(tsdn_t *tsdn, arena_t *arena,
+    arena_chunk_t *chunk, void *ptr, arena_chunk_map_bits_t *bitselm);
+void	arena_dalloc_bin(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
+    void *ptr, size_t pageind, arena_chunk_map_bits_t *bitselm);
+void	arena_dalloc_small(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
    void *ptr, size_t pageind);
 #ifdef JEMALLOC_JET
 typedef void (arena_dalloc_junk_large_t)(void *, size_t);
@@ -554,67 +571,80 @@ extern arena_dalloc_junk_large_t *arena_dalloc_junk_large;
 #else
 void	arena_dalloc_junk_large(void *ptr, size_t usize);
 #endif
-void	arena_dalloc_large_junked_locked(arena_t *arena, arena_chunk_t *chunk,
-    void *ptr);
-void	arena_dalloc_large(tsd_t *tsd, arena_t *arena, arena_chunk_t *chunk,
+void	arena_dalloc_large_junked_locked(tsdn_t *tsdn, arena_t *arena,
+    arena_chunk_t *chunk, void *ptr);
+void	arena_dalloc_large(tsdn_t *tsdn, arena_t *arena, arena_chunk_t *chunk,
    void *ptr);
 #ifdef JEMALLOC_JET
 typedef void (arena_ralloc_junk_large_t)(void *, size_t, size_t);
 extern arena_ralloc_junk_large_t *arena_ralloc_junk_large;
 #endif
-bool	arena_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
-    size_t extra, bool zero);
+bool	arena_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize,
+    size_t size, size_t extra, bool zero);
 void	*arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
    size_t size, size_t alignment, bool zero, tcache_t *tcache);
-dss_prec_t	arena_dss_prec_get(arena_t *arena);
-bool	arena_dss_prec_set(arena_t *arena, dss_prec_t dss_prec);
+dss_prec_t	arena_dss_prec_get(tsdn_t *tsdn, arena_t *arena);
+bool	arena_dss_prec_set(tsdn_t *tsdn, arena_t *arena, dss_prec_t dss_prec);
 ssize_t	arena_lg_dirty_mult_default_get(void);
 bool	arena_lg_dirty_mult_default_set(ssize_t lg_dirty_mult);
 ssize_t	arena_decay_time_default_get(void);
 bool	arena_decay_time_default_set(ssize_t decay_time);
-void	arena_basic_stats_merge(arena_t *arena, unsigned *nthreads,
+void	arena_basic_stats_merge(tsdn_t *tsdn, arena_t *arena,
+    unsigned *nthreads, const char **dss, ssize_t *lg_dirty_mult,
+    ssize_t *decay_time, size_t *nactive, size_t *ndirty);
+void	arena_stats_merge(tsdn_t *tsdn, arena_t *arena, unsigned *nthreads,
    const char **dss, ssize_t *lg_dirty_mult, ssize_t *decay_time,
-    size_t *nactive, size_t *ndirty);
-void	arena_stats_merge(arena_t *arena, unsigned *nthreads, const char **dss,
-    ssize_t *lg_dirty_mult, ssize_t *decay_time, size_t *nactive,
-    size_t *ndirty, arena_stats_t *astats, malloc_bin_stats_t *bstats,
-    malloc_large_stats_t *lstats, malloc_huge_stats_t *hstats);
-unsigned	arena_nthreads_get(arena_t *arena);
-void	arena_nthreads_inc(arena_t *arena);
-void	arena_nthreads_dec(arena_t *arena);
-arena_t	*arena_new(unsigned ind);
+    size_t *nactive, size_t *ndirty, arena_stats_t *astats,
+    malloc_bin_stats_t *bstats, malloc_large_stats_t *lstats,
+    malloc_huge_stats_t *hstats);
+unsigned	arena_nthreads_get(arena_t *arena, bool internal);
+void	arena_nthreads_inc(arena_t *arena, bool internal);
+void	arena_nthreads_dec(arena_t *arena, bool internal);
+arena_t	*arena_new(tsdn_t *tsdn, unsigned ind);
 bool	arena_boot(void);
-void	arena_prefork(arena_t *arena);
-void	arena_postfork_parent(arena_t *arena);
-void	arena_postfork_child(arena_t *arena);
+void	arena_prefork0(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork1(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork2(tsdn_t *tsdn, arena_t *arena);
+void	arena_prefork3(tsdn_t *tsdn, arena_t *arena);
+void	arena_postfork_parent(tsdn_t *tsdn, arena_t *arena);
+void	arena_postfork_child(tsdn_t *tsdn, arena_t *arena);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES

 #ifndef JEMALLOC_ENABLE_INLINE
-arena_chunk_map_bits_t	*arena_bitselm_get(arena_chunk_t *chunk,
+arena_chunk_map_bits_t	*arena_bitselm_get_mutable(arena_chunk_t *chunk,
    size_t pageind);
-arena_chunk_map_misc_t	*arena_miscelm_get(arena_chunk_t *chunk,
+const arena_chunk_map_bits_t	*arena_bitselm_get_const(
+    const arena_chunk_t *chunk, size_t pageind);
+arena_chunk_map_misc_t	*arena_miscelm_get_mutable(arena_chunk_t *chunk,
    size_t pageind);
+const arena_chunk_map_misc_t	*arena_miscelm_get_const(
+    const arena_chunk_t *chunk, size_t pageind);
 size_t	arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm);
-void	*arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm);
+void	*arena_miscelm_to_rpages(const arena_chunk_map_misc_t *miscelm);
 arena_chunk_map_misc_t	*arena_rd_to_miscelm(arena_runs_dirty_link_t *rd);
 arena_chunk_map_misc_t	*arena_run_to_miscelm(arena_run_t *run);
-size_t	*arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbitsp_read(size_t *mapbitsp);
-size_t	arena_mapbits_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_size_decode(size_t mapbits);
-size_t	arena_mapbits_unallocated_size_get(arena_chunk_t *chunk,
+size_t	*arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind);
+const size_t	*arena_mapbitsp_get_const(const arena_chunk_t *chunk,
    size_t pageind);
-size_t	arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind);
-szind_t	arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind);
-size_t	arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbitsp_read(const size_t *mapbitsp);
+size_t	arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_size_decode(size_t mapbits);
+size_t	arena_mapbits_unallocated_size_get(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbits_large_size_get(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbits_small_runind_get(const arena_chunk_t *chunk,
+    size_t pageind);
+szind_t	arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_decommitted_get(const arena_chunk_t *chunk,
+    size_t pageind);
+size_t	arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind);
+size_t	arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind);
 void	arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits);
 size_t	arena_mapbits_size_encode(size_t size);
 void	arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind,
@@ -634,29 +664,31 @@ void	arena_metadata_allocated_sub(arena_t *arena, size_t size);
 size_t	arena_metadata_allocated_get(arena_t *arena);
 bool	arena_prof_accum_impl(arena_t *arena, uint64_t accumbytes);
 bool	arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes);
-bool	arena_prof_accum(arena_t *arena, uint64_t accumbytes);
+bool	arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes);
 szind_t	arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
 szind_t	arena_bin_index(arena_t *arena, arena_bin_t *bin);
 size_t	arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info,
    const void *ptr);
-prof_tctx_t	*arena_prof_tctx_get(const void *ptr);
-void	arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
-void	arena_prof_tctx_reset(const void *ptr, size_t usize,
+prof_tctx_t	*arena_prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	arena_prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void	arena_prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
    const void *old_ptr, prof_tctx_t *old_tctx);
-void	arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks);
-void	arena_decay_tick(tsd_t *tsd, arena_t *arena);
-void	*arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind,
+void	arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks);
+void	arena_decay_tick(tsdn_t *tsdn, arena_t *arena);
+void	*arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind,
    bool zero, tcache_t *tcache, bool slow_path);
 arena_t	*arena_aalloc(const void *ptr);
-size_t	arena_salloc(const void *ptr, bool demote);
-void	arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
-void	arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
+size_t	arena_salloc(tsdn_t *tsdn, const void *ptr, bool demote);
+void	arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool slow_path);
+void	arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
 #endif

 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_ARENA_C_))
 #  ifdef JEMALLOC_ARENA_INLINE_A
 JEMALLOC_ALWAYS_INLINE arena_chunk_map_bits_t *
-arena_bitselm_get(arena_chunk_t *chunk, size_t pageind)
+arena_bitselm_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {

 	assert(pageind >= map_bias);
@@ -665,8 +697,15 @@ arena_bitselm_get(arena_chunk_t *chunk, size_t pageind)
 	return (&chunk->map_bits[pageind-map_bias]);
 }

+JEMALLOC_ALWAYS_INLINE const arena_chunk_map_bits_t *
+arena_bitselm_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_bitselm_get_mutable((arena_chunk_t *)chunk, pageind));
+}
+
 JEMALLOC_ALWAYS_INLINE arena_chunk_map_misc_t *
-arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
+arena_miscelm_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {

 	assert(pageind >= map_bias);
@@ -676,6 +715,13 @@ arena_miscelm_get(arena_chunk_t *chunk, size_t pageind)
 	    (uintptr_t)map_misc_offset) + pageind-map_bias);
 }

+JEMALLOC_ALWAYS_INLINE const arena_chunk_map_misc_t *
+arena_miscelm_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_miscelm_get_mutable((arena_chunk_t *)chunk, pageind));
+}
+
 JEMALLOC_ALWAYS_INLINE size_t
 arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm)
 {
@@ -690,7 +736,7 @@ arena_miscelm_to_pageind(const arena_chunk_map_misc_t *miscelm)
 }

 JEMALLOC_ALWAYS_INLINE void *
-arena_miscelm_to_rpages(arena_chunk_map_misc_t *miscelm)
+arena_miscelm_to_rpages(const arena_chunk_map_misc_t *miscelm)
 {
 	arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
 	size_t pageind = arena_miscelm_to_pageind(miscelm);
@@ -723,24 +769,31 @@ arena_run_to_miscelm(arena_run_t *run)
 }

 JEMALLOC_ALWAYS_INLINE size_t *
-arena_mapbitsp_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbitsp_get_mutable(arena_chunk_t *chunk, size_t pageind)
 {

-	return (&arena_bitselm_get(chunk, pageind)->bits);
+	return (&arena_bitselm_get_mutable(chunk, pageind)->bits);
+}
+
+JEMALLOC_ALWAYS_INLINE const size_t *
+arena_mapbitsp_get_const(const arena_chunk_t *chunk, size_t pageind)
+{
+
+	return (arena_mapbitsp_get_mutable((arena_chunk_t *)chunk, pageind));
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbitsp_read(size_t *mapbitsp)
+arena_mapbitsp_read(const size_t *mapbitsp)
 {

 	return (*mapbitsp);
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_get(const arena_chunk_t *chunk, size_t pageind)
 {

-	return (arena_mapbitsp_read(arena_mapbitsp_get(chunk, pageind)));
+	return (arena_mapbitsp_read(arena_mapbitsp_get_const(chunk, pageind)));
 }

 JEMALLOC_ALWAYS_INLINE size_t
@@ -760,7 +813,7 @@ arena_mapbits_size_decode(size_t mapbits)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_unallocated_size_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -770,7 +823,7 @@ arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_large_size_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -781,7 +834,7 @@ arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_small_runind_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -792,7 +845,7 @@ arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE szind_t
-arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_binind_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
 	szind_t binind;
@@ -804,7 +857,7 @@ arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_dirty_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -815,7 +868,7 @@ arena_mapbits_dirty_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_unzeroed_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -826,7 +879,7 @@ arena_mapbits_unzeroed_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_decommitted_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -837,7 +890,7 @@ arena_mapbits_decommitted_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_large_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -846,7 +899,7 @@ arena_mapbits_large_get(arena_chunk_t *chunk, size_t pageind)
 }

 JEMALLOC_ALWAYS_INLINE size_t
-arena_mapbits_allocated_get(arena_chunk_t *chunk, size_t pageind)
+arena_mapbits_allocated_get(const arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;

@@ -882,7 +935,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
    size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);

 	assert((size & PAGE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
@@ -896,7 +949,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
    size_t size)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);

 	assert((size & PAGE_MASK) == 0);
@@ -908,7 +961,7 @@ arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
 JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_internal_set(arena_chunk_t *chunk, size_t pageind, size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);

 	assert((flags & CHUNK_MAP_UNZEROED) == flags);
 	arena_mapbitsp_write(mapbitsp, flags);
@@ -918,7 +971,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size,
    size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);

 	assert((size & PAGE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
@@ -933,7 +986,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
    szind_t binind)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);

 	assert(binind <= BININD_INVALID);
@@ -947,7 +1000,7 @@ JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,
    szind_t binind, size_t flags)
 {
-	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
+	size_t *mapbitsp = arena_mapbitsp_get_mutable(chunk, pageind);

 	assert(binind < BININD_INVALID);
 	assert(pageind - runind >= map_bias);
@@ -1004,7 +1057,7 @@ arena_prof_accum_locked(arena_t *arena, uint64_t accumbytes)
 }

 JEMALLOC_INLINE bool
-arena_prof_accum(arena_t *arena, uint64_t accumbytes)
+arena_prof_accum(tsdn_t *tsdn, arena_t *arena, uint64_t accumbytes)
 {

 	cassert(config_prof);
@@ -1015,9 +1068,9 @@ arena_prof_accum(arena_t *arena, uint64_t accumbytes)
 	{
 		bool ret;

-		malloc_mutex_lock(&arena->lock);
+		malloc_mutex_lock(tsdn, &arena->lock);
 		ret = arena_prof_accum_impl(arena, accumbytes);
-		malloc_mutex_unlock(&arena->lock);
+		malloc_mutex_unlock(tsdn, &arena->lock);
 		return (ret);
 	}
 }
@@ -1035,12 +1088,12 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		size_t pageind;
 		size_t actual_mapbits;
 		size_t rpages_ind;
-		arena_run_t *run;
+		const arena_run_t *run;
 		arena_bin_t *bin;
 		szind_t run_binind, actual_binind;
 		arena_bin_info_t *bin_info;
-		arena_chunk_map_misc_t *miscelm;
-		void *rpages;
+		const arena_chunk_map_misc_t *miscelm;
+		const void *rpages;

 		assert(binind != BININD_INVALID);
 		assert(binind < NBINS);
@@ -1053,7 +1106,7 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
 		rpages_ind = pageind - arena_mapbits_small_runind_get(chunk,
 		    pageind);
-		miscelm = arena_miscelm_get(chunk, rpages_ind);
+		miscelm = arena_miscelm_get_const(chunk, rpages_ind);
 		run = &miscelm->run;
 		run_binind = run->binind;
 		bin = &arena->bins[run_binind];
@@ -1153,7 +1206,7 @@ arena_run_regind(arena_run_t *run, arena_bin_info_t *bin_info, const void *ptr)
 }

 JEMALLOC_INLINE prof_tctx_t *
-arena_prof_tctx_get(const void *ptr)
+arena_prof_tctx_get(tsdn_t *tsdn, const void *ptr)
 {
 	prof_tctx_t *ret;
 	arena_chunk_t *chunk;
@@ -1169,18 +1222,19 @@ arena_prof_tctx_get(const void *ptr)
 		if (likely((mapbits & CHUNK_MAP_LARGE) == 0))
 			ret = (prof_tctx_t *)(uintptr_t)1U;
 		else {
-			arena_chunk_map_misc_t *elm = arena_miscelm_get(chunk,
-			    pageind);
+			arena_chunk_map_misc_t *elm =
+			    arena_miscelm_get_mutable(chunk, pageind);
 			ret = atomic_read_p(&elm->prof_tctx_pun);
 		}
 	} else
-		ret = huge_prof_tctx_get(ptr);
+		ret = huge_prof_tctx_get(tsdn, ptr);

 	return (ret);
 }

 JEMALLOC_INLINE void
-arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
+arena_prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx)
 {
 	arena_chunk_t *chunk;

@@ -1199,7 +1253,7 @@ arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)

 			assert(arena_mapbits_large_get(chunk, pageind) != 0);

-			elm = arena_miscelm_get(chunk, pageind);
+			elm = arena_miscelm_get_mutable(chunk, pageind);
 			atomic_write_p(&elm->prof_tctx_pun, tctx);
 		} else {
 			/*
@@ -1211,12 +1265,12 @@ arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
 			assert(arena_mapbits_large_get(chunk, pageind) == 0);
 		}
 	} else
-		huge_prof_tctx_set(ptr, tctx);
+		huge_prof_tctx_set(tsdn, ptr, tctx);
 }

 JEMALLOC_INLINE void
-arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
-    prof_tctx_t *old_tctx)
+arena_prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
+    const void *old_ptr, prof_tctx_t *old_tctx)
 {

 	cassert(config_prof);
@@ -1235,56 +1289,59 @@ arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
 			    0);
 			assert(arena_mapbits_large_get(chunk, pageind) != 0);

-			elm = arena_miscelm_get(chunk, pageind);
+			elm = arena_miscelm_get_mutable(chunk, pageind);
 			atomic_write_p(&elm->prof_tctx_pun,
 			    (prof_tctx_t *)(uintptr_t)1U);
 		} else
-			huge_prof_tctx_reset(ptr);
+			huge_prof_tctx_reset(tsdn, ptr);
 	}
 }

 JEMALLOC_ALWAYS_INLINE void
-arena_decay_ticks(tsd_t *tsd, arena_t *arena, unsigned nticks)
+arena_decay_ticks(tsdn_t *tsdn, arena_t *arena, unsigned nticks)
 {
+	tsd_t *tsd;
 	ticker_t *decay_ticker;

-	if (unlikely(tsd == NULL))
+	if (unlikely(tsdn_null(tsdn)))
 		return;
+	tsd = tsdn_tsd(tsdn);
 	decay_ticker = decay_ticker_get(tsd, arena->ind);
 	if (unlikely(decay_ticker == NULL))
 		return;
 	if (unlikely(ticker_ticks(decay_ticker, nticks)))
-		arena_purge(arena, false);
+		arena_purge(tsdn, arena, false);
 }

 JEMALLOC_ALWAYS_INLINE void
-arena_decay_tick(tsd_t *tsd, arena_t *arena)
+arena_decay_tick(tsdn_t *tsdn, arena_t *arena)
 {

-	arena_decay_ticks(tsd, arena, 1);
+	arena_decay_ticks(tsdn, arena, 1);
 }

 JEMALLOC_ALWAYS_INLINE void *
-arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, szind_t ind, bool zero,
+arena_malloc(tsdn_t *tsdn, arena_t *arena, size_t size, szind_t ind, bool zero,
    tcache_t *tcache, bool slow_path)
 {

+	assert(!tsdn_null(tsdn) || tcache == NULL);
 	assert(size != 0);

 	if (likely(tcache != NULL)) {
 		if (likely(size <= SMALL_MAXCLASS)) {
-			return (tcache_alloc_small(tsd, arena, tcache, size,
-			    ind, zero, slow_path));
+			return (tcache_alloc_small(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path));
 		}
 		if (likely(size <= tcache_maxclass)) {
-			return (tcache_alloc_large(tsd, arena, tcache, size,
-			    ind, zero, slow_path));
+			return (tcache_alloc_large(tsdn_tsd(tsdn), arena,
+			    tcache, size, ind, zero, slow_path));
 		}
 		/* (size > tcache_maxclass) case falls through. */
 		assert(size > tcache_maxclass);
 	}

-	return (arena_malloc_hard(tsd, arena, size, ind, zero, tcache));
+	return (arena_malloc_hard(tsdn, arena, size, ind, zero));
 }

 JEMALLOC_ALWAYS_INLINE arena_t *
@@ -1301,7 +1358,7 @@ arena_aalloc(const void *ptr)

 /* Return the size of the allocation pointed to by ptr. */
 JEMALLOC_ALWAYS_INLINE size_t
-arena_salloc(const void *ptr, bool demote)
+arena_salloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {
 	size_t ret;
 	arena_chunk_t *chunk;
@@ -1344,17 +1401,18 @@ arena_salloc(const void *ptr, bool demote)
 			ret = index2size(binind);
 		}
 	} else
-		ret = huge_salloc(ptr);
+		ret = huge_salloc(tsdn, ptr);

 	return (ret);
 }

 JEMALLOC_ALWAYS_INLINE void
-arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
+arena_dalloc(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool slow_path)
 {
 	arena_chunk_t *chunk;
 	size_t pageind, mapbits;

+	assert(!tsdn_null(tsdn) || tcache == NULL);
 	assert(ptr != NULL);

 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
@@ -1367,11 +1425,12 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 			if (likely(tcache != NULL)) {
 				szind_t binind = arena_ptr_small_binind_get(ptr,
 				    mapbits);
-				tcache_dalloc_small(tsd, tcache, ptr, binind,
-				    slow_path);
+				tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
+				    binind, slow_path);
 			} else {
-				arena_dalloc_small(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr, pageind);
+				arena_dalloc_small(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr, pageind);
 			}
 		} else {
 			size_t size = arena_mapbits_large_size_get(chunk,
@@ -1382,22 +1441,26 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)

 			if (likely(tcache != NULL) && size - large_pad <=
 			    tcache_maxclass) {
-				tcache_dalloc_large(tsd, tcache, ptr, size -
-				    large_pad, slow_path);
+				tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr,
+				    size - large_pad, slow_path);
 			} else {
-				arena_dalloc_large(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr);
+				arena_dalloc_large(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr);
 			}
 		}
 	} else
-		huge_dalloc(tsd, ptr, tcache);
+		huge_dalloc(tsdn, ptr);
 }

 JEMALLOC_ALWAYS_INLINE void
-arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+arena_sdalloc(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path)
 {
 	arena_chunk_t *chunk;

+	assert(!tsdn_null(tsdn) || tcache == NULL);
+
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 	if (likely(chunk != ptr)) {
 		if (config_prof && opt_prof) {
@@ -1414,34 +1477,36 @@ arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
 				    pageind) - large_pad;
 			}
 		}
-		assert(s2u(size) == s2u(arena_salloc(ptr, false)));
+		assert(s2u(size) == s2u(arena_salloc(tsdn, ptr, false)));

 		if (likely(size <= SMALL_MAXCLASS)) {
 			/* Small allocation. */
 			if (likely(tcache != NULL)) {
 				szind_t binind = size2index(size);
-				tcache_dalloc_small(tsd, tcache, ptr, binind,
-				    true);
+				tcache_dalloc_small(tsdn_tsd(tsdn), tcache, ptr,
+				    binind, slow_path);
 			} else {
 				size_t pageind = ((uintptr_t)ptr -
 				    (uintptr_t)chunk) >> LG_PAGE;
-				arena_dalloc_small(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr, pageind);
+				arena_dalloc_small(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr, pageind);
 			}
 		} else {
 			assert(config_cache_oblivious || ((uintptr_t)ptr &
 			    PAGE_MASK) == 0);

 			if (likely(tcache != NULL) && size <= tcache_maxclass) {
-				tcache_dalloc_large(tsd, tcache, ptr, size,
-				    true);
+				tcache_dalloc_large(tsdn_tsd(tsdn), tcache, ptr,
+				    size, slow_path);
 			} else {
-				arena_dalloc_large(tsd, extent_node_arena_get(
-				    &chunk->node), chunk, ptr);
+				arena_dalloc_large(tsdn,
+				    extent_node_arena_get(&chunk->node), chunk,
+				    ptr);
 			}
 		}
 	} else
-		huge_dalloc(tsd, ptr, tcache);
+		huge_dalloc(tsdn, ptr);
 }
 #  endif /* JEMALLOC_ARENA_INLINE_B */
 #endif
--- a/include/jemalloc/internal/base.h
+++ b/include/jemalloc/internal/base.h
@@ -9,12 +9,13 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-void	*base_alloc(size_t size);
-void	base_stats_get(size_t *allocated, size_t *resident, size_t *mapped);
+void	*base_alloc(tsdn_t *tsdn, size_t size);
+void	base_stats_get(tsdn_t *tsdn, size_t *allocated, size_t *resident,
+    size_t *mapped);
 bool	base_boot(void);
-void	base_prefork(void);
-void	base_postfork_parent(void);
-void	base_postfork_child(void);
+void	base_prefork(tsdn_t *tsdn);
+void	base_postfork_parent(tsdn_t *tsdn);
+void	base_postfork_child(tsdn_t *tsdn);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
--- a/include/jemalloc/internal/bitmap.h
+++ b/include/jemalloc/internal/bitmap.h
@@ -17,8 +17,8 @@ typedef unsigned long bitmap_t;

 /*
 * Do some analysis on how big the bitmap is before we use a tree.  For a brute
- * force linear search, if we would have to call ffsl more than 2^3 times, use a
- * tree instead.
+ * force linear search, if we would have to call ffs_lu() more than 2^3 times,
+ * use a tree instead.
 */
 #if LG_BITMAP_MAXBITS - LG_BITMAP_GROUP_NBITS > 3
 #  define USE_TREE
@@ -223,7 +223,7 @@ bitmap_sfu(bitmap_t *bitmap, const bitmap_info_t *binfo)
 		i++;
 		g = bitmap[i];
 	}
-	bit = (bit - 1) + (i << 6);
+	bit = (i << LG_BITMAP_GROUP_NBITS) + (bit - 1);
 #endif
 	bitmap_set(bitmap, binfo, bit);
 	return (bit);
--- a/include/jemalloc/internal/chunk.h
+++ b/include/jemalloc/internal/chunk.h
@@ -48,32 +48,32 @@ extern size_t		chunk_npages;

 extern const chunk_hooks_t	chunk_hooks_default;

-chunk_hooks_t	chunk_hooks_get(arena_t *arena);
-chunk_hooks_t	chunk_hooks_set(arena_t *arena,
+chunk_hooks_t	chunk_hooks_get(tsdn_t *tsdn, arena_t *arena);
+chunk_hooks_t	chunk_hooks_set(tsdn_t *tsdn, arena_t *arena,
    const chunk_hooks_t *chunk_hooks);

-bool	chunk_register(const void *chunk, const extent_node_t *node);
+bool	chunk_register(tsdn_t *tsdn, const void *chunk,
+    const extent_node_t *node);
 void	chunk_deregister(const void *chunk, const extent_node_t *node);
 void	*chunk_alloc_base(size_t size);
-void	*chunk_alloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *new_addr, size_t size, size_t alignment, bool *zero,
-    bool dalloc_node);
-void	*chunk_alloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit);
-void	chunk_dalloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, bool committed);
-void	chunk_dalloc_arena(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, bool zeroed, bool committed);
-void	chunk_dalloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, bool committed);
-bool	chunk_purge_arena(arena_t *arena, void *chunk, size_t offset,
+void	*chunk_alloc_cache(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment,
+    bool *zero, bool dalloc_node);
+void	*chunk_alloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *new_addr, size_t size, size_t alignment,
+    bool *zero, bool *commit);
+void	chunk_dalloc_cache(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, bool committed);
+void	chunk_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, bool zeroed,
+    bool committed);
+bool	chunk_purge_wrapper(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, void *chunk, size_t size, size_t offset,
    size_t length);
-bool	chunk_purge_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    void *chunk, size_t size, size_t offset, size_t length);
 bool	chunk_boot(void);
-void	chunk_prefork(void);
-void	chunk_postfork_parent(void);
-void	chunk_postfork_child(void);
+void	chunk_prefork(tsdn_t *tsdn);
+void	chunk_postfork_parent(tsdn_t *tsdn);
+void	chunk_postfork_child(tsdn_t *tsdn);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
--- a/include/jemalloc/internal/chunk_dss.h
+++ b/include/jemalloc/internal/chunk_dss.h
@@ -21,15 +21,15 @@ extern const char *dss_prec_names[];
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-dss_prec_t	chunk_dss_prec_get(void);
-bool	chunk_dss_prec_set(dss_prec_t dss_prec);
-void	*chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size,
-    size_t alignment, bool *zero, bool *commit);
-bool	chunk_in_dss(void *chunk);
+dss_prec_t	chunk_dss_prec_get(tsdn_t *tsdn);
+bool	chunk_dss_prec_set(tsdn_t *tsdn, dss_prec_t dss_prec);
+void	*chunk_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr,
+    size_t size, size_t alignment, bool *zero, bool *commit);
+bool	chunk_in_dss(tsdn_t *tsdn, void *chunk);
 bool	chunk_dss_boot(void);
-void	chunk_dss_prefork(void);
-void	chunk_dss_postfork_parent(void);
-void	chunk_dss_postfork_child(void);
+void	chunk_dss_prefork(tsdn_t *tsdn);
+void	chunk_dss_postfork_parent(tsdn_t *tsdn);
+void	chunk_dss_postfork_child(tsdn_t *tsdn);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
--- a/include/jemalloc/internal/ckh.h
+++ b/include/jemalloc/internal/ckh.h
@@ -64,13 +64,13 @@ struct ckh_s {
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-bool	ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
+bool	ckh_new(tsdn_t *tsdn, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
    ckh_keycomp_t *keycomp);
-void	ckh_delete(tsd_t *tsd, ckh_t *ckh);
+void	ckh_delete(tsdn_t *tsdn, ckh_t *ckh);
 size_t	ckh_count(ckh_t *ckh);
 bool	ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data);
-bool	ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data);
-bool	ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
+bool	ckh_insert(tsdn_t *tsdn, ckh_t *ckh, const void *key, const void *data);
+bool	ckh_remove(tsdn_t *tsdn, ckh_t *ckh, const void *searchkey, void **key,
    void **data);
 bool	ckh_search(ckh_t *ckh, const void *searchkey, void **key, void **data);
 void	ckh_string_hash(const void *key, size_t r_hash[2]);
--- a/include/jemalloc/internal/ctl.h
+++ b/include/jemalloc/internal/ctl.h
@@ -21,13 +21,14 @@ struct ctl_named_node_s {
 	/* If (nchildren == 0), this is a terminal node. */
 	unsigned		nchildren;
 	const			ctl_node_t *children;
-	int			(*ctl)(const size_t *, size_t, void *, size_t *,
-	    void *, size_t);
+	int			(*ctl)(tsd_t *, const size_t *, size_t, void *,
+	    size_t *, void *, size_t);
 };

 struct ctl_indexed_node_s {
 	struct ctl_node_s	node;
-	const ctl_named_node_t	*(*index)(const size_t *, size_t, size_t);
+	const ctl_named_node_t	*(*index)(tsdn_t *, const size_t *, size_t,
+	    size_t);
 };

 struct ctl_arena_stats_s {
@@ -60,6 +61,7 @@ struct ctl_stats_s {
 	size_t			metadata;
 	size_t			resident;
 	size_t			mapped;
+	size_t			retained;
 	unsigned		narenas;
 	ctl_arena_stats_t	*arenas;	/* (narenas + 1) elements. */
 };
@@ -68,16 +70,17 @@ struct ctl_stats_s {
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-int	ctl_byname(const char *name, void *oldp, size_t *oldlenp, void *newp,
-    size_t newlen);
-int	ctl_nametomib(const char *name, size_t *mibp, size_t *miblenp);
-
-int	ctl_bymib(const size_t *mib, size_t miblen, void *oldp, size_t *oldlenp,
+int	ctl_byname(tsd_t *tsd, const char *name, void *oldp, size_t *oldlenp,
    void *newp, size_t newlen);
+int	ctl_nametomib(tsdn_t *tsdn, const char *name, size_t *mibp,
+    size_t *miblenp);
+
+int	ctl_bymib(tsd_t *tsd, const size_t *mib, size_t miblen, void *oldp,
+    size_t *oldlenp, void *newp, size_t newlen);
 bool	ctl_boot(void);
-void	ctl_prefork(void);
-void	ctl_postfork_parent(void);
-void	ctl_postfork_child(void);
+void	ctl_prefork(tsdn_t *tsdn);
+void	ctl_postfork_parent(tsdn_t *tsdn);
+void	ctl_postfork_child(tsdn_t *tsdn);

 #define	xmallctl(name, oldp, oldlenp, newp, newlen) do {		\
 	if (je_mallctl(name, oldp, oldlenp, newp, newlen)		\
--- a/include/jemalloc/internal/extent.h
+++ b/include/jemalloc/internal/extent.h
@@ -48,7 +48,7 @@ struct extent_node_s {
 		/* Linkage for the size/address-ordered tree. */
 		rb_node(extent_node_t)	szad_link;

-		/* Linkage for arena's huge and node_cache lists. */
+		/* Linkage for arena's achunks, huge, and node_cache lists. */
 		ql_elm(extent_node_t)	ql_link;
 	};

--- a/include/jemalloc/internal/hash.h
+++ b/include/jemalloc/internal/hash.h
@@ -53,7 +53,7 @@ hash_get_block_32(const uint32_t *p, int i)
 	if (unlikely((uintptr_t)p & (sizeof(uint32_t)-1)) != 0) {
 		uint32_t ret;

-		memcpy(&ret, &p[i], sizeof(uint32_t));
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint32_t));
 		return (ret);
 	}

@@ -68,7 +68,7 @@ hash_get_block_64(const uint64_t *p, int i)
 	if (unlikely((uintptr_t)p & (sizeof(uint64_t)-1)) != 0) {
 		uint64_t ret;

-		memcpy(&ret, &p[i], sizeof(uint64_t));
+		memcpy(&ret, (uint8_t *)(p + i), sizeof(uint64_t));
 		return (ret);
 	}

--- a/include/jemalloc/internal/huge.h
+++ b/include/jemalloc/internal/huge.h
@@ -9,24 +9,23 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-void	*huge_malloc(tsd_t *tsd, arena_t *arena, size_t usize, bool zero,
-    tcache_t *tcache);
-void	*huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
-    bool zero, tcache_t *tcache);
-bool	huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize,
+void	*huge_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero);
+void	*huge_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool zero);
+bool	huge_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize,
    size_t usize_min, size_t usize_max, bool zero);
 void	*huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
    size_t usize, size_t alignment, bool zero, tcache_t *tcache);
 #ifdef JEMALLOC_JET
-typedef void (huge_dalloc_junk_t)(void *, size_t);
+typedef void (huge_dalloc_junk_t)(tsdn_t *, void *, size_t);
 extern huge_dalloc_junk_t *huge_dalloc_junk;
 #endif
-void	huge_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache);
+void	huge_dalloc(tsdn_t *tsdn, void *ptr);
 arena_t	*huge_aalloc(const void *ptr);
-size_t	huge_salloc(const void *ptr);
-prof_tctx_t	*huge_prof_tctx_get(const void *ptr);
-void	huge_prof_tctx_set(const void *ptr, prof_tctx_t *tctx);
-void	huge_prof_tctx_reset(const void *ptr);
+size_t	huge_salloc(tsdn_t *tsdn, const void *ptr);
+prof_tctx_t	*huge_prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	huge_prof_tctx_set(tsdn_t *tsdn, const void *ptr, prof_tctx_t *tctx);
+void	huge_prof_tctx_reset(tsdn_t *tsdn, const void *ptr);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
--- a/include/jemalloc/internal/jemalloc_internal.h.in
+++ b/include/jemalloc/internal/jemalloc_internal.h.in
@@ -161,6 +161,7 @@ static const bool config_cache_oblivious =
 #include <malloc/malloc.h>
 #endif

+#include "jemalloc/internal/ph.h"
 #define	RB_COMPACT
 #include "jemalloc/internal/rb.h"
 #include "jemalloc/internal/qr.h"
@@ -257,6 +258,9 @@ typedef unsigned szind_t;
 #  ifdef __powerpc__
 #    define LG_QUANTUM		4
 #  endif
+#  ifdef __riscv__
+#    define LG_QUANTUM		4
+#  endif
 #  ifdef __s390__
 #    define LG_QUANTUM		4
 #  endif
@@ -367,6 +371,7 @@ typedef unsigned szind_t;
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/tsd.h"
 #include "jemalloc/internal/mb.h"
@@ -398,6 +403,7 @@ typedef unsigned szind_t;
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/bitmap.h"
@@ -440,6 +446,9 @@ extern bool	in_valgrind;
 /* Number of CPUs. */
 extern unsigned	ncpus;

+/* Number of arenas used for automatic multiplexing of threads and arenas. */
+extern unsigned	narenas_auto;
+
 /*
 * Arenas that are used to service external requests.  Not all elements of the
 * arenas array are necessarily used; arenas are created lazily as needed.
@@ -463,14 +472,14 @@ void	a0dalloc(void *ptr);
 void	*bootstrap_malloc(size_t size);
 void	*bootstrap_calloc(size_t num, size_t size);
 void	bootstrap_free(void *ptr);
-arena_t	*arenas_extend(unsigned ind);
 unsigned	narenas_total_get(void);
-arena_t	*arena_init(unsigned ind);
+arena_t	*arena_init(tsdn_t *tsdn, unsigned ind);
 arena_tdata_t	*arena_tdata_get_hard(tsd_t *tsd, unsigned ind);
-arena_t	*arena_choose_hard(tsd_t *tsd);
+arena_t	*arena_choose_hard(tsd_t *tsd, bool internal);
 void	arena_migrate(tsd_t *tsd, unsigned oldind, unsigned newind);
 void	thread_allocated_cleanup(tsd_t *tsd);
 void	thread_deallocated_cleanup(tsd_t *tsd);
+void	iarena_cleanup(tsd_t *tsd);
 void	arena_cleanup(tsd_t *tsd);
 void	arenas_tdata_cleanup(tsd_t *tsd);
 void	narenas_tdata_cleanup(tsd_t *tsd);
@@ -490,6 +499,7 @@ void	jemalloc_postfork_child(void);
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
+#include "jemalloc/internal/witness.h"
 #include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/bitmap.h"
@@ -521,8 +531,9 @@ void	jemalloc_postfork_child(void);
 #include "jemalloc/internal/smoothstep.h"
 #include "jemalloc/internal/stats.h"
 #include "jemalloc/internal/ctl.h"
-#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/tsd.h"
+#include "jemalloc/internal/witness.h"
+#include "jemalloc/internal/mutex.h"
 #include "jemalloc/internal/mb.h"
 #include "jemalloc/internal/extent.h"
 #include "jemalloc/internal/base.h"
@@ -542,10 +553,12 @@ size_t	s2u_compute(size_t size);
 size_t	s2u_lookup(size_t size);
 size_t	s2u(size_t size);
 size_t	sa2u(size_t size, size_t alignment);
+arena_t	*arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal);
 arena_t	*arena_choose(tsd_t *tsd, arena_t *arena);
+arena_t	*arena_ichoose(tsdn_t *tsdn, arena_t *arena);
 arena_tdata_t	*arena_tdata_get(tsd_t *tsd, unsigned ind,
    bool refresh_if_missing);
-arena_t	*arena_get(unsigned ind, bool init_if_missing);
+arena_t	*arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing);
 ticker_t	*decay_ticker_get(tsd_t *tsd, unsigned ind);
 #endif

@@ -780,19 +793,38 @@ sa2u(size_t size, size_t alignment)

 /* Choose an arena based on a per-thread value. */
 JEMALLOC_INLINE arena_t *
-arena_choose(tsd_t *tsd, arena_t *arena)
+arena_choose_impl(tsd_t *tsd, arena_t *arena, bool internal)
 {
 	arena_t *ret;

 	if (arena != NULL)
 		return (arena);

-	if (unlikely((ret = tsd_arena_get(tsd)) == NULL))
-		ret = arena_choose_hard(tsd);
+	ret = internal ? tsd_iarena_get(tsd) : tsd_arena_get(tsd);
+	if (unlikely(ret == NULL))
+		ret = arena_choose_hard(tsd, internal);

 	return (ret);
 }

+JEMALLOC_INLINE arena_t *
+arena_choose(tsd_t *tsd, arena_t *arena)
+{
+
+	return (arena_choose_impl(tsd, arena, false));
+}
+
+JEMALLOC_INLINE arena_t *
+arena_ichoose(tsdn_t *tsdn, arena_t *arena)
+{
+
+	assert(!tsdn_null(tsdn) || arena != NULL);
+
+	if (!tsdn_null(tsdn))
+		return (arena_choose_impl(tsdn_tsd(tsdn), NULL, true));
+	return (arena);
+}
+
 JEMALLOC_INLINE arena_tdata_t *
 arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing)
 {
@@ -819,7 +851,7 @@ arena_tdata_get(tsd_t *tsd, unsigned ind, bool refresh_if_missing)
 }

 JEMALLOC_INLINE arena_t *
-arena_get(unsigned ind, bool init_if_missing)
+arena_get(tsdn_t *tsdn, unsigned ind, bool init_if_missing)
 {
 	arena_t *ret;

@@ -829,7 +861,7 @@ arena_get(unsigned ind, bool init_if_missing)
 	if (unlikely(ret == NULL)) {
 		ret = atomic_read_p((void *)&arenas[ind]);
 		if (init_if_missing && unlikely(ret == NULL))
-			ret = arena_init(ind);
+			ret = arena_init(tsdn, ind);
 	}
 	return (ret);
 }
@@ -863,30 +895,27 @@ decay_ticker_get(tsd_t *tsd, unsigned ind)

 #ifndef JEMALLOC_ENABLE_INLINE
 arena_t	*iaalloc(const void *ptr);
-size_t	isalloc(const void *ptr, bool demote);
-void	*iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero,
+size_t	isalloc(tsdn_t *tsdn, const void *ptr, bool demote);
+void	*iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero,
    tcache_t *tcache, bool is_metadata, arena_t *arena, bool slow_path);
-void	*imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
-    arena_t *arena);
-void	*imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path);
-void	*icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache,
-    arena_t *arena);
-void	*icalloc(tsd_t *tsd, size_t size, szind_t ind);
-void	*ipallocztm(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+void	*ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero,
+    bool slow_path);
+void	*ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
    tcache_t *tcache, bool is_metadata, arena_t *arena);
-void	*ipalloct(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+void	*ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
    tcache_t *tcache, arena_t *arena);
 void	*ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero);
-size_t	ivsalloc(const void *ptr, bool demote);
+size_t	ivsalloc(tsdn_t *tsdn, const void *ptr, bool demote);
 size_t	u2rz(size_t usize);
-size_t	p2rz(const void *ptr);
-void	idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+size_t	p2rz(tsdn_t *tsdn, const void *ptr);
+void	idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_metadata,
    bool slow_path);
-void	idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache);
 void	idalloc(tsd_t *tsd, void *ptr);
 void	iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path);
-void	isdalloct(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
-void	isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache);
+void	isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
+void	isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path);
 void	*iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
    size_t extra, size_t alignment, bool zero, tcache_t *tcache,
    arena_t *arena);
@@ -894,7 +923,7 @@ void	*iralloct(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
    size_t alignment, bool zero, tcache_t *tcache, arena_t *arena);
 void	*iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
    size_t alignment, bool zero);
-bool	ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
+bool	ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size,
    size_t extra, size_t alignment, bool zero);
 #endif

@@ -910,102 +939,85 @@ iaalloc(const void *ptr)

 /*
 * Typical usage:
+ *   tsdn_t *tsdn = [...]
 *   void *ptr = [...]
- *   size_t sz = isalloc(ptr, config_prof);
+ *   size_t sz = isalloc(tsdn, ptr, config_prof);
 */
 JEMALLOC_ALWAYS_INLINE size_t
-isalloc(const void *ptr, bool demote)
+isalloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {

 	assert(ptr != NULL);
 	/* Demotion only makes sense if config_prof is true. */
 	assert(config_prof || !demote);

-	return (arena_salloc(ptr, demote));
+	return (arena_salloc(tsdn, ptr, demote));
 }

 JEMALLOC_ALWAYS_INLINE void *
-iallocztm(tsd_t *tsd, size_t size, szind_t ind, bool zero, tcache_t *tcache,
+iallocztm(tsdn_t *tsdn, size_t size, szind_t ind, bool zero, tcache_t *tcache,
    bool is_metadata, arena_t *arena, bool slow_path)
 {
 	void *ret;

 	assert(size != 0);
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || arena == NULL || arena->ind < narenas_auto);

-	ret = arena_malloc(tsd, arena, size, ind, zero, tcache, slow_path);
+	ret = arena_malloc(tsdn, arena, size, ind, zero, tcache, slow_path);
 	if (config_stats && is_metadata && likely(ret != NULL)) {
-		arena_metadata_allocated_add(iaalloc(ret), isalloc(ret,
-		    config_prof));
+		arena_metadata_allocated_add(iaalloc(ret),
+		    isalloc(tsdn, ret, config_prof));
 	}
 	return (ret);
 }

 JEMALLOC_ALWAYS_INLINE void *
-imalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
+ialloc(tsd_t *tsd, size_t size, szind_t ind, bool zero, bool slow_path)
 {

-	return (iallocztm(tsd, size, ind, false, tcache, false, arena, true));
+	return (iallocztm(tsd_tsdn(tsd), size, ind, zero, tcache_get(tsd, true),
+	    false, NULL, slow_path));
 }

 JEMALLOC_ALWAYS_INLINE void *
-imalloc(tsd_t *tsd, size_t size, szind_t ind, bool slow_path)
-{
-
-	return (iallocztm(tsd, size, ind, false, tcache_get(tsd, true), false,
-	    NULL, slow_path));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-icalloct(tsd_t *tsd, size_t size, szind_t ind, tcache_t *tcache, arena_t *arena)
-{
-
-	return (iallocztm(tsd, size, ind, true, tcache, false, arena, true));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-icalloc(tsd_t *tsd, size_t size, szind_t ind)
-{
-
-	return (iallocztm(tsd, size, ind, true, tcache_get(tsd, true), false,
-	    NULL, true));
-}
-
-JEMALLOC_ALWAYS_INLINE void *
-ipallocztm(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+ipallocztm(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
    tcache_t *tcache, bool is_metadata, arena_t *arena)
 {
 	void *ret;

 	assert(usize != 0);
 	assert(usize == sa2u(usize, alignment));
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || arena == NULL || arena->ind < narenas_auto);

-	ret = arena_palloc(tsd, arena, usize, alignment, zero, tcache);
+	ret = arena_palloc(tsdn, arena, usize, alignment, zero, tcache);
 	assert(ALIGNMENT_ADDR2BASE(ret, alignment) == ret);
 	if (config_stats && is_metadata && likely(ret != NULL)) {
-		arena_metadata_allocated_add(iaalloc(ret), isalloc(ret,
+		arena_metadata_allocated_add(iaalloc(ret), isalloc(tsdn, ret,
 		    config_prof));
 	}
 	return (ret);
 }

 JEMALLOC_ALWAYS_INLINE void *
-ipalloct(tsd_t *tsd, size_t usize, size_t alignment, bool zero,
+ipalloct(tsdn_t *tsdn, size_t usize, size_t alignment, bool zero,
    tcache_t *tcache, arena_t *arena)
 {

-	return (ipallocztm(tsd, usize, alignment, zero, tcache, false, arena));
+	return (ipallocztm(tsdn, usize, alignment, zero, tcache, false, arena));
 }

 JEMALLOC_ALWAYS_INLINE void *
 ipalloc(tsd_t *tsd, size_t usize, size_t alignment, bool zero)
 {

-	return (ipallocztm(tsd, usize, alignment, zero, tcache_get(tsd, true),
-	    false, NULL));
+	return (ipallocztm(tsd_tsdn(tsd), usize, alignment, zero,
+	    tcache_get(tsd, true), false, NULL));
 }

 JEMALLOC_ALWAYS_INLINE size_t
-ivsalloc(const void *ptr, bool demote)
+ivsalloc(tsdn_t *tsdn, const void *ptr, bool demote)
 {
 	extent_node_t *node;

@@ -1017,7 +1029,7 @@ ivsalloc(const void *ptr, bool demote)
 	assert(extent_node_addr_get(node) == ptr ||
 	    extent_node_achunk_get(node));

-	return (isalloc(ptr, demote));
+	return (isalloc(tsdn, ptr, demote));
 }

 JEMALLOC_INLINE size_t
@@ -1035,39 +1047,34 @@ u2rz(size_t usize)
 }

 JEMALLOC_INLINE size_t
-p2rz(const void *ptr)
+p2rz(tsdn_t *tsdn, const void *ptr)
 {
-	size_t usize = isalloc(ptr, false);
+	size_t usize = isalloc(tsdn, ptr, false);

 	return (u2rz(usize));
 }

 JEMALLOC_ALWAYS_INLINE void
-idalloctm(tsd_t *tsd, void *ptr, tcache_t *tcache, bool is_metadata,
+idalloctm(tsdn_t *tsdn, void *ptr, tcache_t *tcache, bool is_metadata,
    bool slow_path)
 {

 	assert(ptr != NULL);
+	assert(!is_metadata || tcache == NULL);
+	assert(!is_metadata || iaalloc(ptr)->ind < narenas_auto);
 	if (config_stats && is_metadata) {
-		arena_metadata_allocated_sub(iaalloc(ptr), isalloc(ptr,
+		arena_metadata_allocated_sub(iaalloc(ptr), isalloc(tsdn, ptr,
 		    config_prof));
 	}

-	arena_dalloc(tsd, ptr, tcache, slow_path);
-}
-
-JEMALLOC_ALWAYS_INLINE void
-idalloct(tsd_t *tsd, void *ptr, tcache_t *tcache)
-{
-
-	idalloctm(tsd, ptr, tcache, false, true);
+	arena_dalloc(tsdn, ptr, tcache, slow_path);
 }

 JEMALLOC_ALWAYS_INLINE void
 idalloc(tsd_t *tsd, void *ptr)
 {

-	idalloctm(tsd, ptr, tcache_get(tsd, false), false, true);
+	idalloctm(tsd_tsdn(tsd), ptr, tcache_get(tsd, false), false, true);
 }

 JEMALLOC_ALWAYS_INLINE void
@@ -1077,24 +1084,25 @@ iqalloc(tsd_t *tsd, void *ptr, tcache_t *tcache, bool slow_path)
 	if (slow_path && config_fill && unlikely(opt_quarantine))
 		quarantine(tsd, ptr);
 	else
-		idalloctm(tsd, ptr, tcache, false, slow_path);
+		idalloctm(tsd_tsdn(tsd), ptr, tcache, false, slow_path);
 }

 JEMALLOC_ALWAYS_INLINE void
-isdalloct(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+isdalloct(tsdn_t *tsdn, void *ptr, size_t size, tcache_t *tcache,
+    bool slow_path)
 {

-	arena_sdalloc(tsd, ptr, size, tcache);
+	arena_sdalloc(tsdn, ptr, size, tcache, slow_path);
 }

 JEMALLOC_ALWAYS_INLINE void
-isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
+isqalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache, bool slow_path)
 {

-	if (config_fill && unlikely(opt_quarantine))
+	if (slow_path && config_fill && unlikely(opt_quarantine))
 		quarantine(tsd, ptr);
 	else
-		isdalloct(tsd, ptr, size, tcache);
+		isdalloct(tsd_tsdn(tsd), ptr, size, tcache, slow_path);
 }

 JEMALLOC_ALWAYS_INLINE void *
@@ -1107,7 +1115,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 	usize = sa2u(size + extra, alignment);
 	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 		return (NULL);
-	p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
+	p = ipalloct(tsd_tsdn(tsd), usize, alignment, zero, tcache, arena);
 	if (p == NULL) {
 		if (extra == 0)
 			return (NULL);
@@ -1115,7 +1123,8 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 		usize = sa2u(size, alignment);
 		if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 			return (NULL);
-		p = ipalloct(tsd, usize, alignment, zero, tcache, arena);
+		p = ipalloct(tsd_tsdn(tsd), usize, alignment, zero, tcache,
+		    arena);
 		if (p == NULL)
 			return (NULL);
 	}
@@ -1125,7 +1134,7 @@ iralloct_realign(tsd_t *tsd, void *ptr, size_t oldsize, size_t size,
 	 */
 	copysize = (size < oldsize) ? size : oldsize;
 	memcpy(p, ptr, copysize);
-	isqalloc(tsd, ptr, oldsize, tcache);
+	isqalloc(tsd, ptr, oldsize, tcache, true);
 	return (p);
 }

@@ -1161,7 +1170,7 @@ iralloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
 }

 JEMALLOC_ALWAYS_INLINE bool
-ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t extra,
+ixalloc(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t size, size_t extra,
    size_t alignment, bool zero)
 {

@@ -1174,7 +1183,7 @@ ixalloc(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t extra,
 		return (true);
 	}

-	return (arena_ralloc_no_move(tsd, ptr, oldsize, size, extra, zero));
+	return (arena_ralloc_no_move(tsdn, ptr, oldsize, size, extra, zero));
 }
 #endif

--- a/include/jemalloc/internal/jemalloc_internal_defs.h.in
+++ b/include/jemalloc/internal/jemalloc_internal_defs.h.in
@@ -214,6 +214,15 @@
 #undef JEMALLOC_ZONE
 #undef JEMALLOC_ZONE_VERSION

+/*
+ * Methods for determining whether the OS overcommits.
+ * JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY: Linux's
+ *                                         /proc/sys/vm.overcommit_memory file.
+ * JEMALLOC_SYSCTL_VM_OVERCOMMIT: FreeBSD's vm.overcommit sysctl.
+ */
+#undef JEMALLOC_SYSCTL_VM_OVERCOMMIT
+#undef JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY
+
 /*
 * Methods for purging unused pages differ between operating systems.
 *
--- a/include/jemalloc/internal/mb.h
+++ b/include/jemalloc/internal/mb.h
@@ -42,7 +42,7 @@ mb_write(void)
 	    : /* Inputs. */
 	    : "memory" /* Clobbers. */
 	    );
-#else
+#  else
 	/*
 	 * This is hopefully enough to keep the compiler from reordering
 	 * instructions around this one.
@@ -52,7 +52,7 @@ mb_write(void)
 	    : /* Inputs. */
 	    : "memory" /* Clobbers. */
 	    );
-#endif
+#  endif
 }
 #elif (defined(__amd64__) || defined(__x86_64__))
 JEMALLOC_INLINE void
@@ -104,9 +104,9 @@ mb_write(void)
 {
 	malloc_mutex_t mtx;

-	malloc_mutex_init(&mtx);
-	malloc_mutex_lock(&mtx);
-	malloc_mutex_unlock(&mtx);
+	malloc_mutex_init(&mtx, "mb", WITNESS_RANK_OMIT);
+	malloc_mutex_lock(NULL, &mtx);
+	malloc_mutex_unlock(NULL, &mtx);
 }
 #endif
 #endif
--- a/include/jemalloc/internal/mutex.h
+++ b/include/jemalloc/internal/mutex.h
@@ -6,17 +6,21 @@ typedef struct malloc_mutex_s malloc_mutex_t;
 #ifdef _WIN32
 #  define MALLOC_MUTEX_INITIALIZER
 #elif (defined(JEMALLOC_OSSPIN))
-#  define MALLOC_MUTEX_INITIALIZER {0}
+#  define MALLOC_MUTEX_INITIALIZER {0, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #elif (defined(JEMALLOC_MUTEX_INIT_CB))
-#  define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER, NULL}
+#  define MALLOC_MUTEX_INITIALIZER					\
+    {PTHREAD_MUTEX_INITIALIZER, NULL, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #else
 #  if (defined(JEMALLOC_HAVE_PTHREAD_MUTEX_ADAPTIVE_NP) &&		\
       defined(PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP))
 #    define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_ADAPTIVE_NP
-#    define MALLOC_MUTEX_INITIALIZER {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP}
+#    define MALLOC_MUTEX_INITIALIZER					\
+       {PTHREAD_ADAPTIVE_MUTEX_INITIALIZER_NP,				\
+        WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #  else
 #    define MALLOC_MUTEX_TYPE PTHREAD_MUTEX_DEFAULT
-#    define MALLOC_MUTEX_INITIALIZER {PTHREAD_MUTEX_INITIALIZER}
+#    define MALLOC_MUTEX_INITIALIZER					\
+       {PTHREAD_MUTEX_INITIALIZER, WITNESS_INITIALIZER(WITNESS_RANK_OMIT)}
 #  endif
 #endif

@@ -39,6 +43,7 @@ struct malloc_mutex_s {
 #else
 	pthread_mutex_t		lock;
 #endif
+	witness_t		witness;
 };

 #endif /* JEMALLOC_H_STRUCTS */
@@ -52,27 +57,31 @@ extern bool isthreaded;
 #  define isthreaded true
 #endif

-bool	malloc_mutex_init(malloc_mutex_t *mutex);
-void	malloc_mutex_prefork(malloc_mutex_t *mutex);
-void	malloc_mutex_postfork_parent(malloc_mutex_t *mutex);
-void	malloc_mutex_postfork_child(malloc_mutex_t *mutex);
-bool	mutex_boot(void);
+bool	malloc_mutex_init(malloc_mutex_t *mutex, const char *name,
+    witness_rank_t rank);
+void	malloc_mutex_prefork(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_postfork_parent(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_postfork_child(tsdn_t *tsdn, malloc_mutex_t *mutex);
+bool	malloc_mutex_boot(void);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
 #ifdef JEMALLOC_H_INLINES

 #ifndef JEMALLOC_ENABLE_INLINE
-void	malloc_mutex_lock(malloc_mutex_t *mutex);
-void	malloc_mutex_unlock(malloc_mutex_t *mutex);
+void	malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex);
+void	malloc_mutex_assert_not_owner(tsdn_t *tsdn, malloc_mutex_t *mutex);
 #endif

 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MUTEX_C_))
 JEMALLOC_INLINE void
-malloc_mutex_lock(malloc_mutex_t *mutex)
+malloc_mutex_lock(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {

 	if (isthreaded) {
+		witness_assert_not_owner(tsdn, &mutex->witness);
 #ifdef _WIN32
 #  if _WIN32_WINNT >= 0x0600
 		AcquireSRWLockExclusive(&mutex->lock);
@@ -84,14 +93,16 @@ malloc_mutex_lock(malloc_mutex_t *mutex)
 #else
 		pthread_mutex_lock(&mutex->lock);
 #endif
+		witness_lock(tsdn, &mutex->witness);
 	}
 }

 JEMALLOC_INLINE void
-malloc_mutex_unlock(malloc_mutex_t *mutex)
+malloc_mutex_unlock(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {

 	if (isthreaded) {
+		witness_unlock(tsdn, &mutex->witness);
 #ifdef _WIN32
 #  if _WIN32_WINNT >= 0x0600
 		ReleaseSRWLockExclusive(&mutex->lock);
@@ -105,6 +116,22 @@ malloc_mutex_unlock(malloc_mutex_t *mutex)
 #endif
 	}
 }
+
+JEMALLOC_INLINE void
+malloc_mutex_assert_owner(tsdn_t *tsdn, malloc_mutex_t *mutex)
+{
+
+	if (isthreaded)
+		witness_assert_owner(tsdn, &mutex->witness);
+}
+
+JEMALLOC_INLINE void
+malloc_mutex_assert_not_owner(tsdn_t *tsdn, malloc_mutex_t *mutex)
+{
+
+	if (isthreaded)
+		witness_assert_not_owner(tsdn, &mutex->witness);
+}
 #endif

 #endif /* JEMALLOC_H_INLINES */
--- a/include/jemalloc/internal/nstime.h
+++ b/include/jemalloc/internal/nstime.h
@@ -1,13 +1,13 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_TYPES

-#define JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
+#define	JEMALLOC_CLOCK_GETTIME defined(_POSIX_MONOTONIC_CLOCK) \
    && _POSIX_MONOTONIC_CLOCK >= 0

 typedef struct nstime_s nstime_t;

 /* Maximum supported number of seconds (~584 years). */
-#define	NSTIME_SEC_MAX	18446744072
+#define	NSTIME_SEC_MAX	KQU(18446744072)

 #endif /* JEMALLOC_H_TYPES */
 /******************************************************************************/
--- a/include/jemalloc/internal/pages.h
+++ b/include/jemalloc/internal/pages.h
@@ -9,13 +9,14 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS

-void	*pages_map(void *addr, size_t size);
+void	*pages_map(void *addr, size_t size, bool *commit);
 void	pages_unmap(void *addr, size_t size);
 void	*pages_trim(void *addr, size_t alloc_size, size_t leadsize,
-    size_t size);
+    size_t size, bool *commit);
 bool	pages_commit(void *addr, size_t size);
 bool	pages_decommit(void *addr, size_t size);
 bool	pages_purge(void *addr, size_t size);
+void	pages_boot(void);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
--- a/include/jemalloc/internal/ph.h
+++ b/include/jemalloc/internal/ph.h
@@ -0,0 +1,345 @@
+/*
+ * A Pairing Heap implementation.
+ *
+ * "The Pairing Heap: A New Form of Self-Adjusting Heap"
+ * https://www.cs.cmu.edu/~sleator/papers/pairing-heaps.pdf
+ *
+ * With auxiliary twopass list, described in a follow on paper.
+ *
+ * "Pairing Heaps: Experiments and Analysis"
+ * http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.106.2988&rep=rep1&type=pdf
+ *
+ *******************************************************************************
+ */
+
+#ifndef PH_H_
+#define	PH_H_
+
+/* Node structure. */
+#define	phn(a_type)							\
+struct {								\
+	a_type	*phn_prev;						\
+	a_type	*phn_next;						\
+	a_type	*phn_lchild;						\
+}
+
+/* Root structure. */
+#define	ph(a_type)							\
+struct {								\
+	a_type	*ph_root;						\
+}
+
+/* Internal utility macros. */
+#define	phn_lchild_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_lchild)
+#define	phn_lchild_set(a_type, a_field, a_phn, a_lchild) do {		\
+	a_phn->a_field.phn_lchild = a_lchild;				\
+} while (0)
+
+#define	phn_next_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_next)
+#define	phn_prev_set(a_type, a_field, a_phn, a_prev) do {		\
+	a_phn->a_field.phn_prev = a_prev;				\
+} while (0)
+
+#define	phn_prev_get(a_type, a_field, a_phn)				\
+	(a_phn->a_field.phn_prev)
+#define	phn_next_set(a_type, a_field, a_phn, a_next) do {		\
+	a_phn->a_field.phn_next = a_next;				\
+} while (0)
+
+#define	phn_merge_ordered(a_type, a_field, a_phn0, a_phn1, a_cmp) do {	\
+	a_type *phn0child;						\
+									\
+	assert(a_phn0 != NULL);						\
+	assert(a_phn1 != NULL);						\
+	assert(a_cmp(a_phn0, a_phn1) <= 0);				\
+									\
+	phn_prev_set(a_type, a_field, a_phn1, a_phn0);			\
+	phn0child = phn_lchild_get(a_type, a_field, a_phn0);		\
+	phn_next_set(a_type, a_field, a_phn1, phn0child);		\
+	if (phn0child != NULL)						\
+		phn_prev_set(a_type, a_field, phn0child, a_phn1);	\
+	phn_lchild_set(a_type, a_field, a_phn0, a_phn1);		\
+} while (0)
+
+#define	phn_merge(a_type, a_field, a_phn0, a_phn1, a_cmp, r_phn) do {	\
+	if (a_phn0 == NULL)						\
+		r_phn = a_phn1;						\
+	else if (a_phn1 == NULL)					\
+		r_phn = a_phn0;						\
+	else if (a_cmp(a_phn0, a_phn1) < 0) {				\
+		phn_merge_ordered(a_type, a_field, a_phn0, a_phn1,	\
+		    a_cmp);						\
+		r_phn = a_phn0;						\
+	} else {							\
+		phn_merge_ordered(a_type, a_field, a_phn1, a_phn0,	\
+		    a_cmp);						\
+		r_phn = a_phn1;						\
+	}								\
+} while (0)
+
+#define	ph_merge_siblings(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *head = NULL;						\
+	a_type *tail = NULL;						\
+	a_type *phn0 = a_phn;						\
+	a_type *phn1 = phn_next_get(a_type, a_field, phn0);		\
+									\
+	/*								\
+	 * Multipass merge, wherein the first two elements of a FIFO	\
+	 * are repeatedly merged, and each result is appended to the	\
+	 * singly linked FIFO, until the FIFO contains only a single	\
+	 * element.  We start with a sibling list but no reference to	\
+	 * its tail, so we do a single pass over the sibling list to	\
+	 * populate the FIFO.						\
+	 */								\
+	if (phn1 != NULL) {						\
+		a_type *phnrest = phn_next_get(a_type, a_field, phn1);	\
+		if (phnrest != NULL)					\
+			phn_prev_set(a_type, a_field, phnrest, NULL);	\
+		phn_prev_set(a_type, a_field, phn0, NULL);		\
+		phn_next_set(a_type, a_field, phn0, NULL);		\
+		phn_prev_set(a_type, a_field, phn1, NULL);		\
+		phn_next_set(a_type, a_field, phn1, NULL);		\
+		phn_merge(a_type, a_field, phn0, phn1, a_cmp, phn0);	\
+		head = tail = phn0;					\
+		phn0 = phnrest;						\
+		while (phn0 != NULL) {					\
+			phn1 = phn_next_get(a_type, a_field, phn0);	\
+			if (phn1 != NULL) {				\
+				phnrest = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				if (phnrest != NULL) {			\
+					phn_prev_set(a_type, a_field,	\
+					    phnrest, NULL);		\
+				}					\
+				phn_prev_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				phn_prev_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = phnrest;				\
+			} else {					\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = NULL;				\
+			}						\
+		}							\
+		phn0 = head;						\
+		phn1 = phn_next_get(a_type, a_field, phn0);		\
+		if (phn1 != NULL) {					\
+			while (true) {					\
+				head = phn_next_get(a_type, a_field,	\
+				    phn1);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn0) == NULL);			\
+				phn_next_set(a_type, a_field, phn0,	\
+				    NULL);				\
+				assert(phn_prev_get(a_type, a_field,	\
+				    phn1) == NULL);			\
+				phn_next_set(a_type, a_field, phn1,	\
+				    NULL);				\
+				phn_merge(a_type, a_field, phn0, phn1,	\
+				    a_cmp, phn0);			\
+				if (head == NULL)			\
+					break;				\
+				phn_next_set(a_type, a_field, tail,	\
+				    phn0);				\
+				tail = phn0;				\
+				phn0 = head;				\
+				phn1 = phn_next_get(a_type, a_field,	\
+				    phn0);				\
+			}						\
+		}							\
+	}								\
+	r_phn = phn0;							\
+} while (0)
+
+#define	ph_merge_aux(a_type, a_field, a_ph, a_cmp) do {			\
+	a_type *phn = phn_next_get(a_type, a_field, a_ph->ph_root);	\
+	if (phn != NULL) {						\
+		phn_prev_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_next_set(a_type, a_field, a_ph->ph_root, NULL);	\
+		phn_prev_set(a_type, a_field, phn, NULL);		\
+		ph_merge_siblings(a_type, a_field, phn, a_cmp, phn);	\
+		assert(phn_next_get(a_type, a_field, phn) == NULL);	\
+		phn_merge(a_type, a_field, a_ph->ph_root, phn, a_cmp,	\
+		    a_ph->ph_root);					\
+	}								\
+} while (0)
+
+#define	ph_merge_children(a_type, a_field, a_phn, a_cmp, r_phn) do {	\
+	a_type *lchild = phn_lchild_get(a_type, a_field, a_phn);	\
+	if (lchild == NULL)						\
+		r_phn = NULL;						\
+	else {								\
+		ph_merge_siblings(a_type, a_field, lchild, a_cmp,	\
+		    r_phn);						\
+	}								\
+} while (0)
+
+/*
+ * The ph_proto() macro generates function prototypes that correspond to the
+ * functions generated by an equivalently parameterized call to ph_gen().
+ */
+#define	ph_proto(a_attr, a_prefix, a_ph_type, a_type)			\
+a_attr void	a_prefix##new(a_ph_type *ph);				\
+a_attr bool	a_prefix##empty(a_ph_type *ph);				\
+a_attr a_type	*a_prefix##first(a_ph_type *ph);			\
+a_attr void	a_prefix##insert(a_ph_type *ph, a_type *phn);		\
+a_attr a_type	*a_prefix##remove_first(a_ph_type *ph);			\
+a_attr void	a_prefix##remove(a_ph_type *ph, a_type *phn);
+
+/*
+ * The ph_gen() macro generates a type-specific pairing heap implementation,
+ * based on the above cpp macros.
+ */
+#define	ph_gen(a_attr, a_prefix, a_ph_type, a_type, a_field, a_cmp)	\
+a_attr void								\
+a_prefix##new(a_ph_type *ph)						\
+{									\
+									\
+	memset(ph, 0, sizeof(ph(a_type)));				\
+}									\
+a_attr bool								\
+a_prefix##empty(a_ph_type *ph)						\
+{									\
+									\
+	return (ph->ph_root == NULL);					\
+}									\
+a_attr a_type *								\
+a_prefix##first(a_ph_type *ph)						\
+{									\
+									\
+	if (ph->ph_root == NULL)					\
+		return (NULL);						\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+	return (ph->ph_root);						\
+}									\
+a_attr void								\
+a_prefix##insert(a_ph_type *ph, a_type *phn)				\
+{									\
+									\
+	memset(&phn->a_field, 0, sizeof(phn(a_type)));			\
+									\
+	/*								\
+	 * Treat the root as an aux list during insertion, and lazily	\
+	 * merge during a_prefix##remove_first().  For elements that	\
+	 * are inserted, then removed via a_prefix##remove() before the	\
+	 * aux list is ever processed, this makes insert/remove		\
+	 * constant-time, whereas eager merging would make insert	\
+	 * O(log n).							\
+	 */								\
+	if (ph->ph_root == NULL)					\
+		ph->ph_root = phn;					\
+	else {								\
+		phn_next_set(a_type, a_field, phn, phn_next_get(a_type,	\
+		    a_field, ph->ph_root));				\
+		if (phn_next_get(a_type, a_field, ph->ph_root) !=	\
+		    NULL) {						\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, ph->ph_root),	\
+			    phn);					\
+		}							\
+		phn_prev_set(a_type, a_field, phn, ph->ph_root);	\
+		phn_next_set(a_type, a_field, ph->ph_root, phn);	\
+	}								\
+}									\
+a_attr a_type *								\
+a_prefix##remove_first(a_ph_type *ph)					\
+{									\
+	a_type *ret;							\
+									\
+	if (ph->ph_root == NULL)					\
+		return (NULL);						\
+	ph_merge_aux(a_type, a_field, ph, a_cmp);			\
+									\
+	ret = ph->ph_root;						\
+									\
+	ph_merge_children(a_type, a_field, ph->ph_root, a_cmp,		\
+	    ph->ph_root);						\
+									\
+	return (ret);							\
+}									\
+a_attr void								\
+a_prefix##remove(a_ph_type *ph, a_type *phn)				\
+{									\
+	a_type *replace, *parent;					\
+									\
+	/*								\
+	 * We can delete from aux list without merging it, but we need	\
+	 * to merge if we are dealing with the root node.		\
+	 */								\
+	if (ph->ph_root == phn) {					\
+		ph_merge_aux(a_type, a_field, ph, a_cmp);		\
+		if (ph->ph_root == phn) {				\
+			ph_merge_children(a_type, a_field, ph->ph_root,	\
+			    a_cmp, ph->ph_root);			\
+			return;						\
+		}							\
+	}								\
+									\
+	/* Get parent (if phn is leftmost child) before mutating. */	\
+	if ((parent = phn_prev_get(a_type, a_field, phn)) != NULL) {	\
+		if (phn_lchild_get(a_type, a_field, parent) != phn)	\
+			parent = NULL;					\
+	}								\
+	/* Find a possible replacement node, and link to parent. */	\
+	ph_merge_children(a_type, a_field, phn, a_cmp, replace);	\
+	/* Set next/prev for sibling linked list. */			\
+	if (replace != NULL) {						\
+		if (parent != NULL) {					\
+			phn_prev_set(a_type, a_field, replace, parent);	\
+			phn_lchild_set(a_type, a_field, parent,		\
+			    replace);					\
+		} else {						\
+			phn_prev_set(a_type, a_field, replace,		\
+			    phn_prev_get(a_type, a_field, phn));	\
+			if (phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL) {					\
+				phn_next_set(a_type, a_field,		\
+				    phn_prev_get(a_type, a_field, phn),	\
+				    replace);				\
+			}						\
+		}							\
+		phn_next_set(a_type, a_field, replace,			\
+		    phn_next_get(a_type, a_field, phn));		\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    replace);					\
+		}							\
+	} else {							\
+		if (parent != NULL) {					\
+			a_type *next = phn_next_get(a_type, a_field,	\
+			    phn);					\
+			phn_lchild_set(a_type, a_field, parent, next);	\
+			if (next != NULL) {				\
+				phn_prev_set(a_type, a_field, next,	\
+				    parent);				\
+			}						\
+		} else {						\
+			assert(phn_prev_get(a_type, a_field, phn) !=	\
+			    NULL);					\
+			phn_next_set(a_type, a_field,			\
+			    phn_prev_get(a_type, a_field, phn),		\
+			    phn_next_get(a_type, a_field, phn));	\
+		}							\
+		if (phn_next_get(a_type, a_field, phn) != NULL) {	\
+			phn_prev_set(a_type, a_field,			\
+			    phn_next_get(a_type, a_field, phn),		\
+			    phn_prev_get(a_type, a_field, phn));	\
+		}							\
+	}								\
+}
+
+#endif /* PH_H_ */
--- a/include/jemalloc/internal/private_symbols.txt
+++ b/include/jemalloc/internal/private_symbols.txt
@@ -5,10 +5,12 @@ arena_alloc_junk_small
 arena_basic_stats_merge
 arena_bin_index
 arena_bin_info
-arena_bitselm_get
+arena_bitselm_get_const
+arena_bitselm_get_mutable
 arena_boot
 arena_choose
 arena_choose_hard
+arena_choose_impl
 arena_chunk_alloc_huge
 arena_chunk_cache_maybe_insert
 arena_chunk_cache_maybe_remove
@@ -34,6 +36,7 @@ arena_decay_time_set
 arena_dss_prec_get
 arena_dss_prec_set
 arena_get
+arena_ichoose
 arena_init
 arena_lg_dirty_mult_default_get
 arena_lg_dirty_mult_default_set
@@ -60,7 +63,8 @@ arena_mapbits_unallocated_set
 arena_mapbits_unallocated_size_get
 arena_mapbits_unallocated_size_set
 arena_mapbits_unzeroed_get
-arena_mapbitsp_get
+arena_mapbitsp_get_const
+arena_mapbitsp_get_mutable
 arena_mapbitsp_read
 arena_mapbitsp_write
 arena_maxrun
@@ -69,7 +73,8 @@ arena_metadata_allocated_add
 arena_metadata_allocated_get
 arena_metadata_allocated_sub
 arena_migrate
-arena_miscelm_get
+arena_miscelm_get_const
+arena_miscelm_get_mutable
 arena_miscelm_to_pageind
 arena_miscelm_to_rpages
 arena_new
@@ -81,7 +86,10 @@ arena_nthreads_inc
 arena_palloc
 arena_postfork_child
 arena_postfork_parent
-arena_prefork
+arena_prefork0
+arena_prefork1
+arena_prefork2
+arena_prefork3
 arena_prof_accum
 arena_prof_accum_impl
 arena_prof_accum_locked
@@ -97,6 +105,7 @@ arena_ralloc_junk_large
 arena_ralloc_no_move
 arena_rd_to_miscelm
 arena_redzone_corruption
+arena_reset
 arena_run_regind
 arena_run_to_miscelm
 arena_salloc
@@ -123,6 +132,11 @@ atomic_sub_u
 atomic_sub_uint32
 atomic_sub_uint64
 atomic_sub_z
+atomic_write_p
+atomic_write_u
+atomic_write_uint32
+atomic_write_uint64
+atomic_write_z
 base_alloc
 base_boot
 base_postfork_child
@@ -148,7 +162,6 @@ chunk_alloc_dss
 chunk_alloc_mmap
 chunk_alloc_wrapper
 chunk_boot
-chunk_dalloc_arena
 chunk_dalloc_cache
 chunk_dalloc_mmap
 chunk_dalloc_wrapper
@@ -168,7 +181,6 @@ chunk_npages
 chunk_postfork_child
 chunk_postfork_parent
 chunk_prefork
-chunk_purge_arena
 chunk_purge_wrapper
 chunk_register
 chunks_rtree
@@ -200,6 +212,8 @@ extent_node_addr_get
 extent_node_addr_set
 extent_node_arena_get
 extent_node_arena_set
+extent_node_committed_get
+extent_node_committed_set
 extent_node_dirty_insert
 extent_node_dirty_linkage_init
 extent_node_dirty_remove
@@ -210,6 +224,8 @@ extent_node_size_get
 extent_node_size_set
 extent_node_zeroed_get
 extent_node_zeroed_set
+extent_tree_ad_destroy
+extent_tree_ad_destroy_recurse
 extent_tree_ad_empty
 extent_tree_ad_first
 extent_tree_ad_insert
@@ -227,6 +243,8 @@ extent_tree_ad_reverse_iter
 extent_tree_ad_reverse_iter_recurse
 extent_tree_ad_reverse_iter_start
 extent_tree_ad_search
+extent_tree_szad_destroy
+extent_tree_szad_destroy_recurse
 extent_tree_szad_empty
 extent_tree_szad_first
 extent_tree_szad_insert
@@ -273,14 +291,11 @@ huge_ralloc
 huge_ralloc_no_move
 huge_salloc
 iaalloc
+ialloc
 iallocztm
-icalloc
-icalloct
+iarena_cleanup
 idalloc
-idalloct
 idalloctm
-imalloc
-imalloct
 in_valgrind
 index2size
 index2size_compute
@@ -304,7 +319,11 @@ jemalloc_postfork_parent
 jemalloc_prefork
 large_maxclass
 lg_floor
+lg_prof_sample
 malloc_cprintf
+malloc_mutex_assert_not_owner
+malloc_mutex_assert_owner
+malloc_mutex_boot
 malloc_mutex_init
 malloc_mutex_lock
 malloc_mutex_postfork_child
@@ -326,11 +345,13 @@ malloc_write
 map_bias
 map_misc_offset
 mb_write
-mutex_boot
+narenas_auto
 narenas_tdata_cleanup
 narenas_total_get
 ncpus
 nhbins
+nhclasses
+nlclasses
 nstime_add
 nstime_compare
 nstime_copy
@@ -373,6 +394,7 @@ opt_utrace
 opt_xmalloc
 opt_zero
 p2rz
+pages_boot
 pages_commit
 pages_decommit
 pages_map
@@ -384,6 +406,7 @@ pow2_ceil_u64
 pow2_ceil_zu
 prng_lg_range
 prng_range
+prof_active
 prof_active_get
 prof_active_get_unlocked
 prof_active_set
@@ -393,6 +416,7 @@ prof_backtrace
 prof_boot0
 prof_boot1
 prof_boot2
+prof_bt_count
 prof_dump_header
 prof_dump_open
 prof_free
@@ -410,7 +434,8 @@ prof_malloc_sample_object
 prof_mdump
 prof_postfork_child
 prof_postfork_parent
-prof_prefork
+prof_prefork0
+prof_prefork1
 prof_realloc
 prof_reset
 prof_sample_accum_update
@@ -419,6 +444,7 @@ prof_tctx_get
 prof_tctx_reset
 prof_tctx_set
 prof_tdata_cleanup
+prof_tdata_count
 prof_tdata_get
 prof_tdata_init
 prof_tdata_reinit
@@ -470,8 +496,6 @@ tcache_alloc_easy
 tcache_alloc_large
 tcache_alloc_small
 tcache_alloc_small_hard
-tcache_arena_associate
-tcache_arena_dissociate
 tcache_arena_reassociate
 tcache_bin_flush_large
 tcache_bin_flush_small
@@ -506,38 +530,83 @@ ticker_tick
 ticker_ticks
 tsd_arena_get
 tsd_arena_set
+tsd_arenap_get
+tsd_arenas_tdata_bypass_get
+tsd_arenas_tdata_bypass_set
+tsd_arenas_tdata_bypassp_get
+tsd_arenas_tdata_get
+tsd_arenas_tdata_set
+tsd_arenas_tdatap_get
 tsd_boot
 tsd_boot0
 tsd_boot1
 tsd_booted
+tsd_booted_get
 tsd_cleanup
 tsd_cleanup_wrapper
 tsd_fetch
 tsd_get
-tsd_wrapper_get
-tsd_wrapper_set
+tsd_iarena_get
+tsd_iarena_set
+tsd_iarenap_get
 tsd_initialized
 tsd_init_check_recursion
 tsd_init_finish
 tsd_init_head
+tsd_narenas_tdata_get
+tsd_narenas_tdata_set
+tsd_narenas_tdatap_get
+tsd_wrapper_get
+tsd_wrapper_set
 tsd_nominal
 tsd_prof_tdata_get
 tsd_prof_tdata_set
+tsd_prof_tdatap_get
 tsd_quarantine_get
 tsd_quarantine_set
+tsd_quarantinep_get
 tsd_set
 tsd_tcache_enabled_get
 tsd_tcache_enabled_set
+tsd_tcache_enabledp_get
 tsd_tcache_get
 tsd_tcache_set
+tsd_tcachep_get
 tsd_thread_allocated_get
 tsd_thread_allocated_set
+tsd_thread_allocatedp_get
 tsd_thread_deallocated_get
 tsd_thread_deallocated_set
+tsd_thread_deallocatedp_get
 tsd_tls
 tsd_tsd
+tsd_tsdn
+tsd_witness_fork_get
+tsd_witness_fork_set
+tsd_witness_forkp_get
+tsd_witnesses_get
+tsd_witnesses_set
+tsd_witnessesp_get
+tsdn_fetch
+tsdn_null
+tsdn_tsd
 u2rz
 valgrind_freelike_block
 valgrind_make_mem_defined
 valgrind_make_mem_noaccess
 valgrind_make_mem_undefined
+witness_assert_lockless
+witness_assert_not_owner
+witness_assert_owner
+witness_fork_cleanup
+witness_init
+witness_lock
+witness_lock_error
+witness_lockless_error
+witness_not_owner_error
+witness_owner_error
+witness_postfork_child
+witness_postfork_parent
+witness_prefork
+witness_unlock
+witnesses_cleanup
--- a/include/jemalloc/internal/prof.h
+++ b/include/jemalloc/internal/prof.h
@@ -281,7 +281,7 @@ extern uint64_t	prof_interval;
 extern size_t	lg_prof_sample;

 void	prof_alloc_rollback(tsd_t *tsd, prof_tctx_t *tctx, bool updated);
-void	prof_malloc_sample_object(const void *ptr, size_t usize,
+void	prof_malloc_sample_object(tsdn_t *tsdn, const void *ptr, size_t usize,
    prof_tctx_t *tctx);
 void	prof_free_sampled_object(tsd_t *tsd, size_t usize, prof_tctx_t *tctx);
 void	bt_init(prof_bt_t *bt, void **vec);
@@ -293,32 +293,33 @@ size_t	prof_bt_count(void);
 const prof_cnt_t *prof_cnt_all(void);
 typedef int (prof_dump_open_t)(bool, const char *);
 extern prof_dump_open_t *prof_dump_open;
-typedef bool (prof_dump_header_t)(bool, const prof_cnt_t *);
+typedef bool (prof_dump_header_t)(tsdn_t *, bool, const prof_cnt_t *);
 extern prof_dump_header_t *prof_dump_header;
 #endif
-void	prof_idump(void);
-bool	prof_mdump(const char *filename);
-void	prof_gdump(void);
-prof_tdata_t	*prof_tdata_init(tsd_t *tsd);
+void	prof_idump(tsdn_t *tsdn);
+bool	prof_mdump(tsd_t *tsd, const char *filename);
+void	prof_gdump(tsdn_t *tsdn);
+prof_tdata_t	*prof_tdata_init(tsdn_t *tsdn);
 prof_tdata_t	*prof_tdata_reinit(tsd_t *tsd, prof_tdata_t *tdata);
-void	prof_reset(tsd_t *tsd, size_t lg_sample);
+void	prof_reset(tsdn_t *tsdn, size_t lg_sample);
 void	prof_tdata_cleanup(tsd_t *tsd);
-const char	*prof_thread_name_get(void);
-bool	prof_active_get(void);
-bool	prof_active_set(bool active);
+bool	prof_active_get(tsdn_t *tsdn);
+bool	prof_active_set(tsdn_t *tsdn, bool active);
+const char	*prof_thread_name_get(tsd_t *tsd);
 int	prof_thread_name_set(tsd_t *tsd, const char *thread_name);
-bool	prof_thread_active_get(void);
-bool	prof_thread_active_set(bool active);
-bool	prof_thread_active_init_get(void);
-bool	prof_thread_active_init_set(bool active_init);
-bool	prof_gdump_get(void);
-bool	prof_gdump_set(bool active);
+bool	prof_thread_active_get(tsd_t *tsd);
+bool	prof_thread_active_set(tsd_t *tsd, bool active);
+bool	prof_thread_active_init_get(tsdn_t *tsdn);
+bool	prof_thread_active_init_set(tsdn_t *tsdn, bool active_init);
+bool	prof_gdump_get(tsdn_t *tsdn);
+bool	prof_gdump_set(tsdn_t *tsdn, bool active);
 void	prof_boot0(void);
 void	prof_boot1(void);
-bool	prof_boot2(void);
-void	prof_prefork(void);
-void	prof_postfork_parent(void);
-void	prof_postfork_child(void);
+bool	prof_boot2(tsdn_t *tsdn);
+void	prof_prefork0(tsdn_t *tsdn);
+void	prof_prefork1(tsdn_t *tsdn);
+void	prof_postfork_parent(tsdn_t *tsdn);
+void	prof_postfork_child(tsdn_t *tsdn);
 void	prof_sample_threshold_update(prof_tdata_t *tdata);

 #endif /* JEMALLOC_H_EXTERNS */
@@ -329,17 +330,17 @@ void	prof_sample_threshold_update(prof_tdata_t *tdata);
 bool	prof_active_get_unlocked(void);
 bool	prof_gdump_get_unlocked(void);
 prof_tdata_t	*prof_tdata_get(tsd_t *tsd, bool create);
+prof_tctx_t	*prof_tctx_get(tsdn_t *tsdn, const void *ptr);
+void	prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize,
+    prof_tctx_t *tctx);
+void	prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize,
+    const void *old_ptr, prof_tctx_t *tctx);
 bool	prof_sample_accum_update(tsd_t *tsd, size_t usize, bool commit,
    prof_tdata_t **tdata_out);
 prof_tctx_t	*prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active,
    bool update);
-prof_tctx_t	*prof_tctx_get(const void *ptr);
-void	prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
-void	prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
+void	prof_malloc(tsdn_t *tsdn, const void *ptr, size_t usize,
    prof_tctx_t *tctx);
-void	prof_malloc_sample_object(const void *ptr, size_t usize,
-    prof_tctx_t *tctx);
-void	prof_malloc(const void *ptr, size_t usize, prof_tctx_t *tctx);
 void	prof_realloc(tsd_t *tsd, const void *ptr, size_t usize,
    prof_tctx_t *tctx, bool prof_active, bool updated, const void *old_ptr,
    size_t old_usize, prof_tctx_t *old_tctx);
@@ -383,7 +384,7 @@ prof_tdata_get(tsd_t *tsd, bool create)
 	if (create) {
 		if (unlikely(tdata == NULL)) {
 			if (tsd_nominal(tsd)) {
-				tdata = prof_tdata_init(tsd);
+				tdata = prof_tdata_init(tsd_tsdn(tsd));
 				tsd_prof_tdata_set(tsd, tdata);
 			}
 		} else if (unlikely(tdata->expired)) {
@@ -397,34 +398,34 @@ prof_tdata_get(tsd_t *tsd, bool create)
 }

 JEMALLOC_ALWAYS_INLINE prof_tctx_t *
-prof_tctx_get(const void *ptr)
+prof_tctx_get(tsdn_t *tsdn, const void *ptr)
 {

 	cassert(config_prof);
 	assert(ptr != NULL);

-	return (arena_prof_tctx_get(ptr));
+	return (arena_prof_tctx_get(tsdn, ptr));
 }

 JEMALLOC_ALWAYS_INLINE void
-prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
+prof_tctx_set(tsdn_t *tsdn, const void *ptr, size_t usize, prof_tctx_t *tctx)
 {

 	cassert(config_prof);
 	assert(ptr != NULL);

-	arena_prof_tctx_set(ptr, usize, tctx);
+	arena_prof_tctx_set(tsdn, ptr, usize, tctx);
 }

 JEMALLOC_ALWAYS_INLINE void
-prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
+prof_tctx_reset(tsdn_t *tsdn, const void *ptr, size_t usize, const void *old_ptr,
    prof_tctx_t *old_tctx)
 {

 	cassert(config_prof);
 	assert(ptr != NULL);

-	arena_prof_tctx_reset(ptr, usize, old_ptr, old_tctx);
+	arena_prof_tctx_reset(tsdn, ptr, usize, old_ptr, old_tctx);
 }

 JEMALLOC_ALWAYS_INLINE bool
@@ -479,17 +480,17 @@ prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active, bool update)
 }

 JEMALLOC_ALWAYS_INLINE void
-prof_malloc(const void *ptr, size_t usize, prof_tctx_t *tctx)
+prof_malloc(tsdn_t *tsdn, const void *ptr, size_t usize, prof_tctx_t *tctx)
 {

 	cassert(config_prof);
 	assert(ptr != NULL);
-	assert(usize == isalloc(ptr, true));
+	assert(usize == isalloc(tsdn, ptr, true));

 	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
-		prof_malloc_sample_object(ptr, usize, tctx);
+		prof_malloc_sample_object(tsdn, ptr, usize, tctx);
 	else
-		prof_tctx_set(ptr, usize, (prof_tctx_t *)(uintptr_t)1U);
+		prof_tctx_set(tsdn, ptr, usize, (prof_tctx_t *)(uintptr_t)1U);
 }

 JEMALLOC_ALWAYS_INLINE void
@@ -503,7 +504,7 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 	assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U);

 	if (prof_active && !updated && ptr != NULL) {
-		assert(usize == isalloc(ptr, true));
+		assert(usize == isalloc(tsd_tsdn(tsd), ptr, true));
 		if (prof_sample_accum_update(tsd, usize, true, NULL)) {
 			/*
 			 * Don't sample.  The usize passed to prof_alloc_prep()
@@ -512,6 +513,7 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 			 * though its actual usize was insufficient to cross the
 			 * sample threshold.
 			 */
+			prof_alloc_rollback(tsd, tctx, true);
 			tctx = (prof_tctx_t *)(uintptr_t)1U;
 		}
 	}
@@ -520,9 +522,9 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 	old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U);

 	if (unlikely(sampled))
-		prof_malloc_sample_object(ptr, usize, tctx);
+		prof_malloc_sample_object(tsd_tsdn(tsd), ptr, usize, tctx);
 	else
-		prof_tctx_reset(ptr, usize, old_ptr, old_tctx);
+		prof_tctx_reset(tsd_tsdn(tsd), ptr, usize, old_ptr, old_tctx);

 	if (unlikely(old_sampled))
 		prof_free_sampled_object(tsd, old_usize, old_tctx);
@@ -531,10 +533,10 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 JEMALLOC_ALWAYS_INLINE void
 prof_free(tsd_t *tsd, const void *ptr, size_t usize)
 {
-	prof_tctx_t *tctx = prof_tctx_get(ptr);
+	prof_tctx_t *tctx = prof_tctx_get(tsd_tsdn(tsd), ptr);

 	cassert(config_prof);
-	assert(usize == isalloc(ptr, true));
+	assert(usize == isalloc(tsd_tsdn(tsd), ptr, true));

 	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
 		prof_free_sampled_object(tsd, usize, tctx);
--- a/include/jemalloc/internal/rtree.h
+++ b/include/jemalloc/internal/rtree.h
@@ -15,9 +15,10 @@ typedef struct rtree_s rtree_t;
 * machine address width.
 */
 #define	LG_RTREE_BITS_PER_LEVEL	4
-#define	RTREE_BITS_PER_LEVEL	(ZU(1) << LG_RTREE_BITS_PER_LEVEL)
+#define	RTREE_BITS_PER_LEVEL	(1U << LG_RTREE_BITS_PER_LEVEL)
+/* Maximum rtree height. */
 #define	RTREE_HEIGHT_MAX						\
-    ((ZU(1) << (LG_SIZEOF_PTR+3)) / RTREE_BITS_PER_LEVEL)
+    ((1U << (LG_SIZEOF_PTR+3)) / RTREE_BITS_PER_LEVEL)

 /* Used for two-stage lock-free node initialization. */
 #define	RTREE_NODE_INITIALIZING	((rtree_node_elm_t *)0x1)
@@ -111,22 +112,25 @@ unsigned	rtree_start_level(rtree_t *rtree, uintptr_t key);
 uintptr_t	rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level);

 bool	rtree_node_valid(rtree_node_elm_t *node);
-rtree_node_elm_t	*rtree_child_tryread(rtree_node_elm_t *elm);
+rtree_node_elm_t	*rtree_child_tryread(rtree_node_elm_t *elm,
+    bool dependent);
 rtree_node_elm_t	*rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm,
-    unsigned level);
+    unsigned level, bool dependent);
 extent_node_t	*rtree_val_read(rtree_t *rtree, rtree_node_elm_t *elm,
    bool dependent);
 void	rtree_val_write(rtree_t *rtree, rtree_node_elm_t *elm,
    const extent_node_t *val);
-rtree_node_elm_t	*rtree_subtree_tryread(rtree_t *rtree, unsigned level);
-rtree_node_elm_t	*rtree_subtree_read(rtree_t *rtree, unsigned level);
+rtree_node_elm_t	*rtree_subtree_tryread(rtree_t *rtree, unsigned level,
+    bool dependent);
+rtree_node_elm_t	*rtree_subtree_read(rtree_t *rtree, unsigned level,
+    bool dependent);

 extent_node_t	*rtree_get(rtree_t *rtree, uintptr_t key, bool dependent);
 bool	rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val);
 #endif

 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_RTREE_C_))
-JEMALLOC_INLINE unsigned
+JEMALLOC_ALWAYS_INLINE unsigned
 rtree_start_level(rtree_t *rtree, uintptr_t key)
 {
 	unsigned start_level;
@@ -140,7 +144,7 @@ rtree_start_level(rtree_t *rtree, uintptr_t key)
 	return (start_level);
 }

-JEMALLOC_INLINE uintptr_t
+JEMALLOC_ALWAYS_INLINE uintptr_t
 rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level)
 {

@@ -149,37 +153,40 @@ rtree_subkey(rtree_t *rtree, uintptr_t key, unsigned level)
 	    rtree->levels[level].bits) - 1));
 }

-JEMALLOC_INLINE bool
+JEMALLOC_ALWAYS_INLINE bool
 rtree_node_valid(rtree_node_elm_t *node)
 {

 	return ((uintptr_t)node > (uintptr_t)RTREE_NODE_INITIALIZING);
 }

-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_child_tryread(rtree_node_elm_t *elm)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_child_tryread(rtree_node_elm_t *elm, bool dependent)
 {
 	rtree_node_elm_t *child;

 	/* Double-checked read (first read may be stale. */
 	child = elm->child;
-	if (!rtree_node_valid(child))
+	if (!dependent && !rtree_node_valid(child))
 		child = atomic_read_p(&elm->pun);
+	assert(!dependent || child != NULL);
 	return (child);
 }

-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_child_read(rtree_t *rtree, rtree_node_elm_t *elm, unsigned level,
+    bool dependent)
 {
 	rtree_node_elm_t *child;

-	child = rtree_child_tryread(elm);
-	if (unlikely(!rtree_node_valid(child)))
+	child = rtree_child_tryread(elm, dependent);
+	if (!dependent && unlikely(!rtree_node_valid(child)))
 		child = rtree_child_read_hard(rtree, elm, level);
+	assert(!dependent || child != NULL);
 	return (child);
 }

-JEMALLOC_INLINE extent_node_t *
+JEMALLOC_ALWAYS_INLINE extent_node_t *
 rtree_val_read(rtree_t *rtree, rtree_node_elm_t *elm, bool dependent)
 {

@@ -208,54 +215,119 @@ rtree_val_write(rtree_t *rtree, rtree_node_elm_t *elm, const extent_node_t *val)
 	atomic_write_p(&elm->pun, val);
 }

-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_subtree_tryread(rtree_t *rtree, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_subtree_tryread(rtree_t *rtree, unsigned level, bool dependent)
 {
 	rtree_node_elm_t *subtree;

 	/* Double-checked read (first read may be stale. */
 	subtree = rtree->levels[level].subtree;
-	if (!rtree_node_valid(subtree))
+	if (!dependent && unlikely(!rtree_node_valid(subtree)))
 		subtree = atomic_read_p(&rtree->levels[level].subtree_pun);
+	assert(!dependent || subtree != NULL);
 	return (subtree);
 }

-JEMALLOC_INLINE rtree_node_elm_t *
-rtree_subtree_read(rtree_t *rtree, unsigned level)
+JEMALLOC_ALWAYS_INLINE rtree_node_elm_t *
+rtree_subtree_read(rtree_t *rtree, unsigned level, bool dependent)
 {
 	rtree_node_elm_t *subtree;

-	subtree = rtree_subtree_tryread(rtree, level);
-	if (unlikely(!rtree_node_valid(subtree)))
+	subtree = rtree_subtree_tryread(rtree, level, dependent);
+	if (!dependent && unlikely(!rtree_node_valid(subtree)))
 		subtree = rtree_subtree_read_hard(rtree, level);
+	assert(!dependent || subtree != NULL);
 	return (subtree);
 }

-JEMALLOC_INLINE extent_node_t *
+JEMALLOC_ALWAYS_INLINE extent_node_t *
 rtree_get(rtree_t *rtree, uintptr_t key, bool dependent)
 {
 	uintptr_t subkey;
-	unsigned i, start_level;
-	rtree_node_elm_t *node, *child;
+	unsigned start_level;
+	rtree_node_elm_t *node;

 	start_level = rtree_start_level(rtree, key);

-	for (i = start_level, node = rtree_subtree_tryread(rtree, start_level);
-	    /**/; i++, node = child) {
-		if (!dependent && unlikely(!rtree_node_valid(node)))
-			return (NULL);
-		subkey = rtree_subkey(rtree, key, i);
-		if (i == rtree->height - 1) {
-			/*
-			 * node is a leaf, so it contains values rather than
-			 * child pointers.
-			 */
-			return (rtree_val_read(rtree, &node[subkey],
-			    dependent));
-		}
-		assert(i < rtree->height - 1);
-		child = rtree_child_tryread(&node[subkey]);
+	node = rtree_subtree_tryread(rtree, start_level, dependent);
+#define	RTREE_GET_BIAS	(RTREE_HEIGHT_MAX - rtree->height)
+	switch (start_level + RTREE_GET_BIAS) {
+#define	RTREE_GET_SUBTREE(level)					\
+	case level:							\
+		assert(level < (RTREE_HEIGHT_MAX-1));			\
+		if (!dependent && unlikely(!rtree_node_valid(node)))	\
+			return (NULL);					\
+		subkey = rtree_subkey(rtree, key, level -		\
+		    RTREE_GET_BIAS);					\
+		node = rtree_child_tryread(&node[subkey], dependent);	\
+		/* Fall through. */
+#define	RTREE_GET_LEAF(level)						\
+	case level:							\
+		assert(level == (RTREE_HEIGHT_MAX-1));			\
+		if (!dependent && unlikely(!rtree_node_valid(node)))	\
+			return (NULL);					\
+		subkey = rtree_subkey(rtree, key, level -		\
+		    RTREE_GET_BIAS);					\
+		/*							\
+		 * node is a leaf, so it contains values rather than	\
+		 * child pointers.					\
+		 */							\
+		return (rtree_val_read(rtree, &node[subkey],		\
+		    dependent));
+#if RTREE_HEIGHT_MAX > 1
+	RTREE_GET_SUBTREE(0)
+#endif
+#if RTREE_HEIGHT_MAX > 2
+	RTREE_GET_SUBTREE(1)
+#endif
+#if RTREE_HEIGHT_MAX > 3
+	RTREE_GET_SUBTREE(2)
+#endif
+#if RTREE_HEIGHT_MAX > 4
+	RTREE_GET_SUBTREE(3)
+#endif
+#if RTREE_HEIGHT_MAX > 5
+	RTREE_GET_SUBTREE(4)
+#endif
+#if RTREE_HEIGHT_MAX > 6
+	RTREE_GET_SUBTREE(5)
+#endif
+#if RTREE_HEIGHT_MAX > 7
+	RTREE_GET_SUBTREE(6)
+#endif
+#if RTREE_HEIGHT_MAX > 8
+	RTREE_GET_SUBTREE(7)
+#endif
+#if RTREE_HEIGHT_MAX > 9
+	RTREE_GET_SUBTREE(8)
+#endif
+#if RTREE_HEIGHT_MAX > 10
+	RTREE_GET_SUBTREE(9)
+#endif
+#if RTREE_HEIGHT_MAX > 11
+	RTREE_GET_SUBTREE(10)
+#endif
+#if RTREE_HEIGHT_MAX > 12
+	RTREE_GET_SUBTREE(11)
+#endif
+#if RTREE_HEIGHT_MAX > 13
+	RTREE_GET_SUBTREE(12)
+#endif
+#if RTREE_HEIGHT_MAX > 14
+	RTREE_GET_SUBTREE(13)
+#endif
+#if RTREE_HEIGHT_MAX > 15
+	RTREE_GET_SUBTREE(14)
+#endif
+#if RTREE_HEIGHT_MAX > 16
+#  error Unsupported RTREE_HEIGHT_MAX
+#endif
+	RTREE_GET_LEAF(RTREE_HEIGHT_MAX-1)
+#undef RTREE_GET_SUBTREE
+#undef RTREE_GET_LEAF
+	default: not_reached();
 	}
+#undef RTREE_GET_BIAS
 	not_reached();
 }

@@ -268,7 +340,7 @@ rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val)

 	start_level = rtree_start_level(rtree, key);

-	node = rtree_subtree_read(rtree, start_level);
+	node = rtree_subtree_read(rtree, start_level, false);
 	if (node == NULL)
 		return (true);
 	for (i = start_level; /**/; i++, node = child) {
@@ -282,7 +354,7 @@ rtree_set(rtree_t *rtree, uintptr_t key, const extent_node_t *val)
 			return (false);
 		}
 		assert(i + 1 < rtree->height);
-		child = rtree_child_read(rtree, &node[subkey], i);
+		child = rtree_child_read(rtree, &node[subkey], i, false);
 		if (child == NULL)
 			return (true);
 	}
--- a/include/jemalloc/internal/stats.h
+++ b/include/jemalloc/internal/stats.h
@@ -102,6 +102,14 @@ struct arena_stats_s {
 	/* Number of bytes currently mapped. */
 	size_t		mapped;

+	/*
+	 * Number of bytes currently retained as a side effect of munmap() being
+	 * disabled/bypassed.  Retained bytes are technically mapped (though
+	 * always decommitted or purged), but they are excluded from the mapped
+	 * statistic (above).
+	 */
+	size_t		retained;
+
 	/*
 	 * Total number of purge sweeps, total number of madvise calls made,
 	 * and total pages purged in order to keep dirty unused memory under
--- a/include/jemalloc/internal/tcache.h
+++ b/include/jemalloc/internal/tcache.h
@@ -130,27 +130,25 @@ extern size_t	tcache_maxclass;
 */
 extern tcaches_t	*tcaches;

-size_t	tcache_salloc(const void *ptr);
+size_t	tcache_salloc(tsdn_t *tsdn, const void *ptr);
 void	tcache_event_hard(tsd_t *tsd, tcache_t *tcache);
-void	*tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
+void	*tcache_alloc_small_hard(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
    tcache_bin_t *tbin, szind_t binind, bool *tcache_success);
 void	tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
    szind_t binind, unsigned rem);
 void	tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
    unsigned rem, tcache_t *tcache);
-void	tcache_arena_associate(tcache_t *tcache, arena_t *arena);
-void	tcache_arena_reassociate(tcache_t *tcache, arena_t *oldarena,
-    arena_t *newarena);
-void	tcache_arena_dissociate(tcache_t *tcache, arena_t *arena);
+void	tcache_arena_reassociate(tsdn_t *tsdn, tcache_t *tcache,
+    arena_t *oldarena, arena_t *newarena);
 tcache_t *tcache_get_hard(tsd_t *tsd);
-tcache_t *tcache_create(tsd_t *tsd, arena_t *arena);
+tcache_t *tcache_create(tsdn_t *tsdn, arena_t *arena);
 void	tcache_cleanup(tsd_t *tsd);
 void	tcache_enabled_cleanup(tsd_t *tsd);
-void	tcache_stats_merge(tcache_t *tcache, arena_t *arena);
-bool	tcaches_create(tsd_t *tsd, unsigned *r_ind);
+void	tcache_stats_merge(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena);
+bool	tcaches_create(tsdn_t *tsdn, unsigned *r_ind);
 void	tcaches_flush(tsd_t *tsd, unsigned ind);
 void	tcaches_destroy(tsd_t *tsd, unsigned ind);
-bool	tcache_boot(void);
+bool	tcache_boot(tsdn_t *tsdn);

 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
@@ -297,8 +295,8 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		if (unlikely(arena == NULL))
 			return (NULL);

-		ret = tcache_alloc_small_hard(tsd, arena, tcache, tbin, binind,
-			&tcache_hard_success);
+		ret = tcache_alloc_small_hard(tsd_tsdn(tsd), arena, tcache,
+		    tbin, binind, &tcache_hard_success);
 		if (tcache_hard_success == false)
 			return (NULL);
 	}
@@ -310,7 +308,7 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 	 */
 	if (config_prof || (slow_path && config_fill) || unlikely(zero)) {
 		usize = index2size(binind);
-		assert(tcache_salloc(ret) == usize);
+		assert(tcache_salloc(tsd_tsdn(tsd), ret) == usize);
 	}

 	if (likely(!zero)) {
@@ -358,7 +356,7 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		if (unlikely(arena == NULL))
 			return (NULL);

-		ret = arena_malloc_large(tsd, arena, binind, zero);
+		ret = arena_malloc_large(tsd_tsdn(tsd), arena, binind, zero);
 		if (ret == NULL)
 			return (NULL);
 	} else {
@@ -381,9 +379,10 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 		}
 		if (likely(!zero)) {
 			if (slow_path && config_fill) {
-				if (unlikely(opt_junk_alloc))
-					memset(ret, 0xa5, usize);
-				else if (unlikely(opt_zero))
+				if (unlikely(opt_junk_alloc)) {
+					memset(ret, JEMALLOC_ALLOC_JUNK,
+					    usize);
+				} else if (unlikely(opt_zero))
 					memset(ret, 0, usize);
 			}
 		} else
@@ -406,7 +405,7 @@ tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind,
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;

-	assert(tcache_salloc(ptr) <= SMALL_MAXCLASS);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) <= SMALL_MAXCLASS);

 	if (slow_path && config_fill && unlikely(opt_junk_free))
 		arena_dalloc_junk_small(ptr, &arena_bin_info[binind]);
@@ -433,8 +432,8 @@ tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size,
 	tcache_bin_info_t *tbin_info;

 	assert((size & PAGE_MASK) == 0);
-	assert(tcache_salloc(ptr) > SMALL_MAXCLASS);
-	assert(tcache_salloc(ptr) <= tcache_maxclass);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) > SMALL_MAXCLASS);
+	assert(tcache_salloc(tsd_tsdn(tsd), ptr) <= tcache_maxclass);

 	binind = size2index(size);

@@ -458,8 +457,10 @@ JEMALLOC_ALWAYS_INLINE tcache_t *
 tcaches_get(tsd_t *tsd, unsigned ind)
 {
 	tcaches_t *elm = &tcaches[ind];
-	if (unlikely(elm->tcache == NULL))
-		elm->tcache = tcache_create(tsd, arena_choose(tsd, NULL));
+	if (unlikely(elm->tcache == NULL)) {
+		elm->tcache = tcache_create(tsd_tsdn(tsd), arena_choose(tsd,
+		    NULL));
+	}
 	return (elm->tcache);
 }
 #endif
--- a/include/jemalloc/internal/tsd.h
+++ b/include/jemalloc/internal/tsd.h
@@ -13,6 +13,9 @@ typedef struct tsd_init_head_s tsd_init_head_t;
 #endif

 typedef struct tsd_s tsd_t;
+typedef struct tsdn_s tsdn_t;
+
+#define	TSDN_NULL	((tsdn_t *)0)

 typedef enum {
 	tsd_state_uninitialized,
@@ -44,6 +47,7 @@ typedef enum {
 * The result is a set of generated functions, e.g.:
 *
 *   bool example_tsd_boot(void) {...}
+ *   bool example_tsd_booted_get(void) {...}
 *   example_t *example_tsd_get() {...}
 *   void example_tsd_set(example_t *val) {...}
 *
@@ -98,6 +102,8 @@ a_attr void								\
 a_name##tsd_boot1(void);						\
 a_attr bool								\
 a_name##tsd_boot(void);							\
+a_attr bool								\
+a_name##tsd_booted_get(void);						\
 a_attr a_type *								\
 a_name##tsd_get(void);							\
 a_attr void								\
@@ -201,6 +207,12 @@ a_name##tsd_boot(void)							\
 									\
 	return (a_name##tsd_boot0());					\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -246,6 +258,12 @@ a_name##tsd_boot(void)							\
 									\
 	return (a_name##tsd_boot0());					\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -368,6 +386,12 @@ a_name##tsd_boot(void)							\
 	a_name##tsd_boot1();						\
 	return (false);							\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -490,6 +514,12 @@ a_name##tsd_boot(void)							\
 	a_name##tsd_boot1();						\
 	return (false);							\
 }									\
+a_attr bool								\
+a_name##tsd_booted_get(void)						\
+{									\
+									\
+	return (a_name##tsd_booted);					\
+}									\
 /* Get/set. */								\
 a_attr a_type *								\
 a_name##tsd_get(void)							\
@@ -536,12 +566,15 @@ struct tsd_init_head_s {
    O(thread_allocated,		uint64_t)				\
    O(thread_deallocated,	uint64_t)				\
    O(prof_tdata,		prof_tdata_t *)				\
+    O(iarena,			arena_t *)				\
    O(arena,			arena_t *)				\
    O(arenas_tdata,		arena_tdata_t *)			\
    O(narenas_tdata,		unsigned)				\
    O(arenas_tdata_bypass,	bool)					\
    O(tcache_enabled,		tcache_enabled_t)			\
    O(quarantine,		quarantine_t *)				\
+    O(witnesses,		witness_list_t)				\
+    O(witness_fork,		bool)					\

 #define	TSD_INITIALIZER {						\
    tsd_state_uninitialized,						\
@@ -551,10 +584,13 @@ struct tsd_init_head_s {
    NULL,								\
    NULL,								\
    NULL,								\
+    NULL,								\
    0,									\
    false,								\
    tcache_enabled_default,						\
-    NULL								\
+    NULL,								\
+    ql_head_initializer(witnesses),					\
+    false								\
 }

 struct tsd_s {
@@ -565,6 +601,15 @@ MALLOC_TSD
 #undef O
 };

+/*
+ * Wrapper around tsd_t that makes it possible to avoid implicit conversion
+ * between tsd_t and tsdn_t, where tsdn_t is "nullable" and has to be
+ * explicitly converted to tsd_t, which is non-nullable.
+ */
+struct tsdn_s {
+	tsd_t	tsd;
+};
+
 static const tsd_t tsd_initializer = TSD_INITIALIZER;

 malloc_tsd_types(, tsd_t)
@@ -577,7 +622,7 @@ void	*malloc_tsd_malloc(size_t size);
 void	malloc_tsd_dalloc(void *wrapper);
 void	malloc_tsd_no_cleanup(void *arg);
 void	malloc_tsd_cleanup_register(bool (*f)(void));
-bool	malloc_tsd_boot0(void);
+tsd_t	*malloc_tsd_boot0(void);
 void	malloc_tsd_boot1(void);
 #if (!defined(JEMALLOC_MALLOC_THREAD_CLEANUP) && !defined(JEMALLOC_TLS) && \
    !defined(_WIN32))
@@ -595,6 +640,7 @@ void	tsd_cleanup(void *arg);
 malloc_tsd_protos(JEMALLOC_ATTR(unused), , tsd_t)

 tsd_t	*tsd_fetch(void);
+tsdn_t	*tsd_tsdn(tsd_t *tsd);
 bool	tsd_nominal(tsd_t *tsd);
 #define	O(n, t)								\
 t	*tsd_##n##p_get(tsd_t *tsd);					\
@@ -602,6 +648,9 @@ t	tsd_##n##_get(tsd_t *tsd);					\
 void	tsd_##n##_set(tsd_t *tsd, t n);
 MALLOC_TSD
 #undef O
+tsdn_t	*tsdn_fetch(void);
+bool	tsdn_null(const tsdn_t *tsdn);
+tsd_t	*tsdn_tsd(tsdn_t *tsdn);
 #endif

 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_TSD_C_))
@@ -628,6 +677,13 @@ tsd_fetch(void)
 	return (tsd);
 }

+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsd_tsdn(tsd_t *tsd)
+{
+
+	return ((tsdn_t *)tsd);
+}
+
 JEMALLOC_INLINE bool
 tsd_nominal(tsd_t *tsd)
 {
@@ -659,6 +715,32 @@ tsd_##n##_set(tsd_t *tsd, t n)						\
 }
 MALLOC_TSD
 #undef O
+
+JEMALLOC_ALWAYS_INLINE tsdn_t *
+tsdn_fetch(void)
+{
+
+	if (!tsd_booted_get())
+		return (NULL);
+
+	return (tsd_tsdn(tsd_fetch()));
+}
+
+JEMALLOC_ALWAYS_INLINE bool
+tsdn_null(const tsdn_t *tsdn)
+{
+
+	return (tsdn == NULL);
+}
+
+JEMALLOC_ALWAYS_INLINE tsd_t *
+tsdn_tsd(tsdn_t *tsdn)
+{
+
+	assert(!tsdn_null(tsdn));
+
+	return (&tsdn->tsd);
+}
 #endif

 #endif /* JEMALLOC_H_INLINES */
--- a/include/jemalloc/internal/util.h
+++ b/include/jemalloc/internal/util.h
@@ -40,6 +40,10 @@
 */
 #define	MALLOC_PRINTF_BUFSIZE	4096

+/* Junk fill patterns. */
+#define	JEMALLOC_ALLOC_JUNK	((uint8_t)0xa5)
+#define	JEMALLOC_FREE_JUNK	((uint8_t)0x5a)
+
 /*
 * Wrap a cpp argument that contains commas such that it isn't broken up into
 * multiple arguments.
@@ -73,12 +77,12 @@
      JEMALLOC_CLANG_HAS_BUILTIN(__builtin_unreachable)
 #	define unreachable() __builtin_unreachable()
 #  else
-#	define unreachable()
+#	define unreachable() abort()
 #  endif
 #else
 #	define likely(x)   !!(x)
 #	define unlikely(x) !!(x)
-#	define unreachable()
+#	define unreachable() abort()
 #endif

 #include "jemalloc/internal/assert.h"
@@ -106,9 +110,9 @@ void	malloc_write(const char *s);
 * malloc_vsnprintf() supports a subset of snprintf(3) that avoids floating
 * point math.
 */
-int	malloc_vsnprintf(char *str, size_t size, const char *format,
+size_t	malloc_vsnprintf(char *str, size_t size, const char *format,
    va_list ap);
-int	malloc_snprintf(char *str, size_t size, const char *format, ...)
+size_t	malloc_snprintf(char *str, size_t size, const char *format, ...)
    JEMALLOC_FORMAT_PRINTF(3, 4);
 void	malloc_vcprintf(void (*write_cb)(void *, const char *), void *cbopaque,
    const char *format, va_list ap);
--- a/include/jemalloc/internal/valgrind.h
+++ b/include/jemalloc/internal/valgrind.h
@@ -30,15 +30,17 @@
 * calls must be embedded in macros rather than in functions so that when
 * Valgrind reports errors, there are no extra stack frames in the backtraces.
 */
-#define	JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do {		\
-	if (unlikely(in_valgrind && cond))				\
-		VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(ptr), zero);	\
+#define	JEMALLOC_VALGRIND_MALLOC(cond, tsdn, ptr, usize, zero) do {	\
+	if (unlikely(in_valgrind && cond)) {				\
+		VALGRIND_MALLOCLIKE_BLOCK(ptr, usize, p2rz(tsdn, ptr),	\
+		    zero);						\
+	}								\
 } while (0)
-#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, ptr, usize,		\
+#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, tsdn, ptr, usize,	\
    ptr_maybe_null, old_ptr, old_usize, old_rzsize, old_ptr_maybe_null,	\
    zero) do {								\
 	if (unlikely(in_valgrind)) {					\
-		size_t rzsize = p2rz(ptr);				\
+		size_t rzsize = p2rz(tsdn, ptr);			\
 									\
 		if (!maybe_moved || ptr == old_ptr) {			\
 			VALGRIND_RESIZEINPLACE_BLOCK(ptr, old_usize,	\
@@ -81,8 +83,8 @@
 #define	JEMALLOC_VALGRIND_MAKE_MEM_NOACCESS(ptr, usize) do {} while (0)
 #define	JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ptr, usize) do {} while (0)
 #define	JEMALLOC_VALGRIND_MAKE_MEM_DEFINED(ptr, usize) do {} while (0)
-#define	JEMALLOC_VALGRIND_MALLOC(cond, ptr, usize, zero) do {} while (0)
-#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, ptr, usize,		\
+#define	JEMALLOC_VALGRIND_MALLOC(cond, tsdn, ptr, usize, zero) do {} while (0)
+#define	JEMALLOC_VALGRIND_REALLOC(maybe_moved, tsdn, ptr, usize,	\
    ptr_maybe_null, old_ptr, old_usize, old_rzsize, old_ptr_maybe_null,	\
    zero) do {} while (0)
 #define	JEMALLOC_VALGRIND_FREE(ptr, rzsize) do {} while (0)
--- a/include/jemalloc/internal/witness.h
+++ b/include/jemalloc/internal/witness.h
@@ -0,0 +1,249 @@
+/******************************************************************************/
+#ifdef JEMALLOC_H_TYPES
+
+typedef struct witness_s witness_t;
+typedef unsigned witness_rank_t;
+typedef ql_head(witness_t) witness_list_t;
+typedef int witness_comp_t (const witness_t *, const witness_t *);
+
+/*
+ * Lock ranks.  Witnesses with rank WITNESS_RANK_OMIT are completely ignored by
+ * the witness machinery.
+ */
+#define	WITNESS_RANK_OMIT		0U
+
+#define	WITNESS_RANK_INIT		1U
+#define	WITNESS_RANK_CTL		1U
+#define	WITNESS_RANK_ARENAS		2U
+
+#define	WITNESS_RANK_PROF_DUMP		3U
+#define	WITNESS_RANK_PROF_BT2GCTX	4U
+#define	WITNESS_RANK_PROF_TDATAS	5U
+#define	WITNESS_RANK_PROF_TDATA		6U
+#define	WITNESS_RANK_PROF_GCTX		7U
+
+#define	WITNESS_RANK_ARENA		8U
+#define	WITNESS_RANK_ARENA_CHUNKS	9U
+#define	WITNESS_RANK_ARENA_NODE_CACHE	10
+
+#define	WITNESS_RANK_BASE		11U
+
+#define	WITNESS_RANK_LEAF		0xffffffffU
+#define	WITNESS_RANK_ARENA_BIN		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_ARENA_HUGE		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_DSS		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_ACTIVE	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_DUMP_SEQ	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_GDUMP		WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_NEXT_THR_UID	WITNESS_RANK_LEAF
+#define	WITNESS_RANK_PROF_THREAD_ACTIVE_INIT	WITNESS_RANK_LEAF
+
+#define	WITNESS_INITIALIZER(rank) {"initializer", rank, NULL, {NULL, NULL}}
+
+#endif /* JEMALLOC_H_TYPES */
+/******************************************************************************/
+#ifdef JEMALLOC_H_STRUCTS
+
+struct witness_s {
+	/* Name, used for printing lock order reversal messages. */
+	const char		*name;
+
+	/*
+	 * Witness rank, where 0 is lowest and UINT_MAX is highest.  Witnesses
+	 * must be acquired in order of increasing rank.
+	 */
+	witness_rank_t		rank;
+
+	/*
+	 * If two witnesses are of equal rank and they have the samp comp
+	 * function pointer, it is called as a last attempt to differentiate
+	 * between witnesses of equal rank.
+	 */
+	witness_comp_t		*comp;
+
+	/* Linkage for thread's currently owned locks. */
+	ql_elm(witness_t)	link;
+};
+
+#endif /* JEMALLOC_H_STRUCTS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_EXTERNS
+
+void	witness_init(witness_t *witness, const char *name, witness_rank_t rank,
+    witness_comp_t *comp);
+#ifdef JEMALLOC_JET
+typedef void (witness_lock_error_t)(const witness_list_t *, const witness_t *);
+extern witness_lock_error_t *witness_lock_error;
+#else
+void	witness_lock_error(const witness_list_t *witnesses,
+    const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_owner_error_t)(const witness_t *);
+extern witness_owner_error_t *witness_owner_error;
+#else
+void	witness_owner_error(const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_not_owner_error_t)(const witness_t *);
+extern witness_not_owner_error_t *witness_not_owner_error;
+#else
+void	witness_not_owner_error(const witness_t *witness);
+#endif
+#ifdef JEMALLOC_JET
+typedef void (witness_lockless_error_t)(const witness_list_t *);
+extern witness_lockless_error_t *witness_lockless_error;
+#else
+void	witness_lockless_error(const witness_list_t *witnesses);
+#endif
+
+void	witnesses_cleanup(tsd_t *tsd);
+void	witness_fork_cleanup(tsd_t *tsd);
+void	witness_prefork(tsd_t *tsd);
+void	witness_postfork_parent(tsd_t *tsd);
+void	witness_postfork_child(tsd_t *tsd);
+
+#endif /* JEMALLOC_H_EXTERNS */
+/******************************************************************************/
+#ifdef JEMALLOC_H_INLINES
+
+#ifndef JEMALLOC_ENABLE_INLINE
+void	witness_assert_owner(tsdn_t *tsdn, const witness_t *witness);
+void	witness_assert_not_owner(tsdn_t *tsdn, const witness_t *witness);
+void	witness_assert_lockless(tsdn_t *tsdn);
+void	witness_lock(tsdn_t *tsdn, witness_t *witness);
+void	witness_unlock(tsdn_t *tsdn, witness_t *witness);
+#endif
+
+#if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_MUTEX_C_))
+JEMALLOC_INLINE void
+witness_assert_owner(tsdn_t *tsdn, const witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_foreach(w, witnesses, link) {
+		if (w == witness)
+			return;
+	}
+	witness_owner_error(witness);
+}
+
+JEMALLOC_INLINE void
+witness_assert_not_owner(tsdn_t *tsdn, const witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_foreach(w, witnesses, link) {
+		if (w == witness)
+			witness_not_owner_error(witness);
+	}
+}
+
+JEMALLOC_INLINE void
+witness_assert_lockless(tsdn_t *tsdn)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	w = ql_last(witnesses, link);
+	if (w != NULL)
+		witness_lockless_error(witnesses);
+}
+
+JEMALLOC_INLINE void
+witness_lock(tsdn_t *tsdn, witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+	witness_t *w;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witness_assert_not_owner(tsdn, witness);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	w = ql_last(witnesses, link);
+	if (w == NULL) {
+		/* No other locks; do nothing. */
+	} else if (tsd_witness_fork_get(tsd) && w->rank <= witness->rank) {
+		/* Forking, and relaxed ranking satisfied. */
+	} else if (w->rank > witness->rank) {
+		/* Not forking, rank order reversal. */
+		witness_lock_error(witnesses, witness);
+	} else if (w->rank == witness->rank && (w->comp == NULL || w->comp !=
+	    witness->comp || w->comp(w, witness) > 0)) {
+		/*
+		 * Missing/incompatible comparison function, or comparison
+		 * function indicates rank order reversal.
+		 */
+		witness_lock_error(witnesses, witness);
+	}
+
+	ql_elm_new(witness, link);
+	ql_tail_insert(witnesses, witness, link);
+}
+
+JEMALLOC_INLINE void
+witness_unlock(tsdn_t *tsdn, witness_t *witness)
+{
+	tsd_t *tsd;
+	witness_list_t *witnesses;
+
+	if (!config_debug)
+		return;
+
+	if (tsdn_null(tsdn))
+		return;
+	tsd = tsdn_tsd(tsdn);
+	if (witness->rank == WITNESS_RANK_OMIT)
+		return;
+
+	witness_assert_owner(tsdn, witness);
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_remove(witnesses, witness, link);
+}
+#endif
+
+#endif /* JEMALLOC_H_INLINES */
+/******************************************************************************/
--- a/include/jemalloc/jemalloc_macros.h.in
+++ b/include/jemalloc/jemalloc_macros.h.in
@@ -13,11 +13,11 @@

 #  define MALLOCX_LG_ALIGN(la)	((int)(la))
 #  if LG_SIZEOF_PTR == 2
-#    define MALLOCX_ALIGN(a)	((int)(ffs(a)-1))
+#    define MALLOCX_ALIGN(a)	((int)(ffs((int)(a))-1))
 #  else
 #    define MALLOCX_ALIGN(a)						\
-       ((int)(((a) < (size_t)INT_MAX) ? ffs((int)(a))-1 :		\
-       ffs((int)((a)>>32))+31))
+       ((int)(((size_t)(a) < (size_t)INT_MAX) ? ffs((int)(a))-1 :	\
+       ffs((int)(((size_t)(a))>>32))+31))
 #  endif
 #  define MALLOCX_ZERO	((int)0x40)
 /*
@@ -29,7 +29,7 @@
 /*
 * Bias arena index bits so that 0 encodes "use an automatically chosen arena".
 */
-#  define MALLOCX_ARENA(a)	((int)(((a)+1) << 20))
+#  define MALLOCX_ARENA(a)	((((int)(a))+1) << 20)

 #if defined(__cplusplus) && defined(JEMALLOC_USE_CXX_THROW)
 #  define JEMALLOC_CXX_THROW throw()
--- a/include/msvc_compat/windows_extra.h
+++ b/include/msvc_compat/windows_extra.h
@@ -1,26 +1,6 @@
 #ifndef MSVC_COMPAT_WINDOWS_EXTRA_H
 #define	MSVC_COMPAT_WINDOWS_EXTRA_H

-#ifndef ENOENT
-#  define ENOENT ERROR_PATH_NOT_FOUND
-#endif
-#ifndef EINVAL
-#  define EINVAL ERROR_BAD_ARGUMENTS
-#endif
-#ifndef EAGAIN
-#  define EAGAIN ERROR_OUTOFMEMORY
-#endif
-#ifndef EPERM
-#  define EPERM  ERROR_WRITE_FAULT
-#endif
-#ifndef EFAULT
-#  define EFAULT ERROR_INVALID_ADDRESS
-#endif
-#ifndef ENOMEM
-#  define ENOMEM ERROR_NOT_ENOUGH_MEMORY
-#endif
-#ifndef ERANGE
-#  define ERANGE ERROR_INVALID_DATA
-#endif
+#include <errno.h>

 #endif /* MSVC_COMPAT_WINDOWS_EXTRA_H */
--- a/msvc/projects/vc2015/jemalloc/jemalloc.vcxproj
+++ b/msvc/projects/vc2015/jemalloc/jemalloc.vcxproj
@@ -54,7 +54,9 @@
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\jemalloc_internal_macros.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mb.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mutex.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\nstime.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\pages.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ph.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_namespace.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_unnamespace.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\prng.h" />
@@ -69,6 +71,7 @@
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\size_classes.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\stats.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tcache.h" />
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ticker.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tsd.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\util.h" />
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\valgrind.h" />
@@ -103,11 +106,13 @@
    <ClCompile Include="..\..\..\..\src\mutex.c" />
    <ClCompile Include="..\..\..\..\src\nstime.c" />
    <ClCompile Include="..\..\..\..\src\pages.c" />
+    <ClCompile Include="..\..\..\..\src\prng.c" />
    <ClCompile Include="..\..\..\..\src\prof.c" />
    <ClCompile Include="..\..\..\..\src\quarantine.c" />
    <ClCompile Include="..\..\..\..\src\rtree.c" />
    <ClCompile Include="..\..\..\..\src\stats.c" />
    <ClCompile Include="..\..\..\..\src\tcache.c" />
+    <ClCompile Include="..\..\..\..\src\ticker.c" />
    <ClCompile Include="..\..\..\..\src\tsd.c" />
    <ClCompile Include="..\..\..\..\src\util.c" />
  </ItemGroup>
@@ -227,7 +232,7 @@
  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Debug-static|x64'">
    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
    <IntDir>$(Platform)\$(Configuration)\</IntDir>
-    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+    <TargetName>$(ProjectName)-vc$(PlatformToolsetVersion)-$(Configuration)</TargetName>
  </PropertyGroup>
  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
@@ -236,7 +241,7 @@
  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='Release-static|x64'">
    <OutDir>$(SolutionDir)$(Platform)\$(Configuration)\</OutDir>
    <IntDir>$(Platform)\$(Configuration)\</IntDir>
-    <TargetName>$(ProjectName)-$(PlatformToolset)-$(Configuration)</TargetName>
+    <TargetName>$(ProjectName)-vc$(PlatformToolsetVersion)-$(Configuration)</TargetName>
  </PropertyGroup>
  <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
    <ClCompile>
@@ -246,7 +251,7 @@
      <Optimization>Disabled</Optimization>
      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;JEMALLOC_DEBUG;_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -263,7 +268,7 @@
      <PreprocessorDefinitions>JEMALLOC_DEBUG;_REENTRANT;JEMALLOC_EXPORT=;_DEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -279,7 +284,7 @@
      <Optimization>Disabled</Optimization>
      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;JEMALLOC_DEBUG;_DEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -296,8 +301,9 @@
      <PreprocessorDefinitions>JEMALLOC_DEBUG;_REENTRANT;JEMALLOC_EXPORT=;_DEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
      <RuntimeLibrary>MultiThreadedDebug</RuntimeLibrary>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
-      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
+      <DebugInformationFormat>OldStyle</DebugInformationFormat>
+      <MinimalRebuild>false</MinimalRebuild>
    </ClCompile>
    <Link>
      <SubSystem>Windows</SubSystem>
@@ -314,7 +320,7 @@
      <IntrinsicFunctions>true</IntrinsicFunctions>
      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;NDEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -335,7 +341,7 @@
      <PreprocessorDefinitions>_REENTRANT;JEMALLOC_EXPORT=;NDEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -355,7 +361,7 @@
      <IntrinsicFunctions>true</IntrinsicFunctions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
      <PreprocessorDefinitions>_REENTRANT;_WINDLL;DLLEXPORT;NDEBUG;%(PreprocessorDefinitions)</PreprocessorDefinitions>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
    </ClCompile>
    <Link>
@@ -376,8 +382,8 @@
      <PreprocessorDefinitions>_REENTRANT;JEMALLOC_EXPORT=;NDEBUG;_LIB;%(PreprocessorDefinitions)</PreprocessorDefinitions>
      <AdditionalIncludeDirectories>..\..\..\..\include;..\..\..\..\include\msvc_compat;%(AdditionalIncludeDirectories)</AdditionalIncludeDirectories>
      <RuntimeLibrary>MultiThreaded</RuntimeLibrary>
-      <DisableSpecificWarnings>4090;4146;4244;4267;4334</DisableSpecificWarnings>
-      <ProgramDataBaseFileName>$(OutputPath)$(TargetName).pdb</ProgramDataBaseFileName>
+      <DisableSpecificWarnings>4090;4146;4267;4334</DisableSpecificWarnings>
+      <DebugInformationFormat>OldStyle</DebugInformationFormat>
    </ClCompile>
    <Link>
      <SubSystem>Windows</SubSystem>
--- a/msvc/projects/vc2015/jemalloc/jemalloc.vcxproj.filters
+++ b/msvc/projects/vc2015/jemalloc/jemalloc.vcxproj.filters
@@ -101,9 +101,15 @@
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\mutex.h">
      <Filter>Header Files\internal</Filter>
    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\nstime.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\pages.h">
      <Filter>Header Files\internal</Filter>
    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ph.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\private_namespace.h">
      <Filter>Header Files\internal</Filter>
    </ClInclude>
@@ -146,6 +152,9 @@
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tcache.h">
      <Filter>Header Files\internal</Filter>
    </ClInclude>
+    <ClInclude Include="..\..\..\..\include\jemalloc\internal\ticker.h">
+      <Filter>Header Files\internal</Filter>
+    </ClInclude>
    <ClInclude Include="..\..\..\..\include\jemalloc\internal\tsd.h">
      <Filter>Header Files\internal</Filter>
    </ClInclude>
@@ -214,9 +223,15 @@
    <ClCompile Include="..\..\..\..\src\mutex.c">
      <Filter>Source Files</Filter>
    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\nstime.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
    <ClCompile Include="..\..\..\..\src\pages.c">
      <Filter>Source Files</Filter>
    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\prng.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
    <ClCompile Include="..\..\..\..\src\prof.c">
      <Filter>Source Files</Filter>
    </ClCompile>
@@ -232,14 +247,14 @@
    <ClCompile Include="..\..\..\..\src\tcache.c">
      <Filter>Source Files</Filter>
    </ClCompile>
+    <ClCompile Include="..\..\..\..\src\ticker.c">
+      <Filter>Source Files</Filter>
+    </ClCompile>
    <ClCompile Include="..\..\..\..\src\tsd.c">
      <Filter>Source Files</Filter>
    </ClCompile>
    <ClCompile Include="..\..\..\..\src\util.c">
      <Filter>Source Files</Filter>
    </ClCompile>
-    <ClCompile Include="..\..\..\..\src\nstime.c">
-      <Filter>Source Files</Filter>
-    </ClCompile>
  </ItemGroup>
 </Project>
--- a/msvc/projects/vc2015/test_threads/test_threads.vcxproj
+++ b/msvc/projects/vc2015/test_threads/test_threads.vcxproj
@@ -223,7 +223,7 @@
    <Link>
      <SubSystem>Console</SubSystem>
      <GenerateDebugInformation>true</GenerateDebugInformation>
-      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>jemalloc-vc$(PlatformToolsetVersion)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
    </Link>
  </ItemDefinitionGroup>
@@ -306,7 +306,7 @@
      <EnableCOMDATFolding>true</EnableCOMDATFolding>
      <OptimizeReferences>true</OptimizeReferences>
      <AdditionalLibraryDirectories>$(SolutionDir)$(Platform)\$(Configuration)</AdditionalLibraryDirectories>
-      <AdditionalDependencies>jemalloc-$(PlatformToolset)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
+      <AdditionalDependencies>jemalloc-vc$(PlatformToolsetVersion)-$(Configuration).lib;kernel32.lib;user32.lib;gdi32.lib;winspool.lib;comdlg32.lib;advapi32.lib;shell32.lib;ole32.lib;oleaut32.lib;uuid.lib;odbc32.lib;odbccp32.lib;%(AdditionalDependencies)</AdditionalDependencies>
    </Link>
  </ItemDefinitionGroup>
  <ItemGroup>
--- a/src/arena.c
+++ b/src/arena.c
--- a/src/base.c
+++ b/src/base.c
@@ -13,12 +13,13 @@ static size_t		base_mapped;

 /******************************************************************************/

-/* base_mtx must be held. */
 static extent_node_t *
-base_node_try_alloc(void)
+base_node_try_alloc(tsdn_t *tsdn)
 {
 	extent_node_t *node;

+	malloc_mutex_assert_owner(tsdn, &base_mtx);
+
 	if (base_nodes == NULL)
 		return (NULL);
 	node = base_nodes;
@@ -27,33 +28,34 @@ base_node_try_alloc(void)
 	return (node);
 }

-/* base_mtx must be held. */
 static void
-base_node_dalloc(extent_node_t *node)
+base_node_dalloc(tsdn_t *tsdn, extent_node_t *node)
 {

+	malloc_mutex_assert_owner(tsdn, &base_mtx);
+
 	JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(node, sizeof(extent_node_t));
 	*(extent_node_t **)node = base_nodes;
 	base_nodes = node;
 }

-/* base_mtx must be held. */
 static extent_node_t *
-base_chunk_alloc(size_t minsize)
+base_chunk_alloc(tsdn_t *tsdn, size_t minsize)
 {
 	extent_node_t *node;
 	size_t csize, nsize;
 	void *addr;

+	malloc_mutex_assert_owner(tsdn, &base_mtx);
 	assert(minsize != 0);
-	node = base_node_try_alloc();
+	node = base_node_try_alloc(tsdn);
 	/* Allocate enough space to also carve a node out if necessary. */
 	nsize = (node == NULL) ? CACHELINE_CEILING(sizeof(extent_node_t)) : 0;
 	csize = CHUNK_CEILING(minsize + nsize);
 	addr = chunk_alloc_base(csize);
 	if (addr == NULL) {
 		if (node != NULL)
-			base_node_dalloc(node);
+			base_node_dalloc(tsdn, node);
 		return (NULL);
 	}
 	base_mapped += csize;
@@ -76,7 +78,7 @@ base_chunk_alloc(size_t minsize)
 * physical memory usage.
 */
 void *
-base_alloc(size_t size)
+base_alloc(tsdn_t *tsdn, size_t size)
 {
 	void *ret;
 	size_t csize, usize;
@@ -91,14 +93,14 @@ base_alloc(size_t size)

 	usize = s2u(csize);
 	extent_node_init(&key, NULL, NULL, usize, false, false);
-	malloc_mutex_lock(&base_mtx);
+	malloc_mutex_lock(tsdn, &base_mtx);
 	node = extent_tree_szad_nsearch(&base_avail_szad, &key);
 	if (node != NULL) {
 		/* Use existing space. */
 		extent_tree_szad_remove(&base_avail_szad, node);
 	} else {
 		/* Try to allocate more space. */
-		node = base_chunk_alloc(csize);
+		node = base_chunk_alloc(tsdn, csize);
 	}
 	if (node == NULL) {
 		ret = NULL;
@@ -111,7 +113,7 @@ base_alloc(size_t size)
 		extent_node_size_set(node, extent_node_size_get(node) - csize);
 		extent_tree_szad_insert(&base_avail_szad, node);
 	} else
-		base_node_dalloc(node);
+		base_node_dalloc(tsdn, node);
 	if (config_stats) {
 		base_allocated += csize;
 		/*
@@ -123,28 +125,29 @@ base_alloc(size_t size)
 	}
 	JEMALLOC_VALGRIND_MAKE_MEM_DEFINED(ret, csize);
 label_return:
-	malloc_mutex_unlock(&base_mtx);
+	malloc_mutex_unlock(tsdn, &base_mtx);
 	return (ret);
 }

 void
-base_stats_get(size_t *allocated, size_t *resident, size_t *mapped)
+base_stats_get(tsdn_t *tsdn, size_t *allocated, size_t *resident,
+    size_t *mapped)
 {

-	malloc_mutex_lock(&base_mtx);
+	malloc_mutex_lock(tsdn, &base_mtx);
 	assert(base_allocated <= base_resident);
 	assert(base_resident <= base_mapped);
 	*allocated = base_allocated;
 	*resident = base_resident;
 	*mapped = base_mapped;
-	malloc_mutex_unlock(&base_mtx);
+	malloc_mutex_unlock(tsdn, &base_mtx);
 }

 bool
 base_boot(void)
 {

-	if (malloc_mutex_init(&base_mtx))
+	if (malloc_mutex_init(&base_mtx, "base", WITNESS_RANK_BASE))
 		return (true);
 	extent_tree_szad_new(&base_avail_szad);
 	base_nodes = NULL;
@@ -153,22 +156,22 @@ base_boot(void)
 }

 void
-base_prefork(void)
+base_prefork(tsdn_t *tsdn)
 {

-	malloc_mutex_prefork(&base_mtx);
+	malloc_mutex_prefork(tsdn, &base_mtx);
 }

 void
-base_postfork_parent(void)
+base_postfork_parent(tsdn_t *tsdn)
 {

-	malloc_mutex_postfork_parent(&base_mtx);
+	malloc_mutex_postfork_parent(tsdn, &base_mtx);
 }

 void
-base_postfork_child(void)
+base_postfork_child(tsdn_t *tsdn)
 {

-	malloc_mutex_postfork_child(&base_mtx);
+	malloc_mutex_postfork_child(tsdn, &base_mtx);
 }
--- a/src/bitmap.c
+++ b/src/bitmap.c
@@ -74,15 +74,11 @@ bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)
 void
 bitmap_info_init(bitmap_info_t *binfo, size_t nbits)
 {
-	size_t i;

 	assert(nbits > 0);
 	assert(nbits <= (ZU(1) << LG_BITMAP_MAXBITS));

-	i = nbits >> LG_BITMAP_GROUP_NBITS;
-	if (nbits % BITMAP_GROUP_NBITS != 0)
-		i++;
-	binfo->ngroups = i;
+	binfo->ngroups = BITMAP_BITS2GROUPS(nbits);
 	binfo->nbits = nbits;
 }

@@ -99,9 +95,10 @@ bitmap_init(bitmap_t *bitmap, const bitmap_info_t *binfo)
 	size_t extra;

 	memset(bitmap, 0xffU, bitmap_size(binfo));
-	extra = (binfo->nbits % (binfo->ngroups * BITMAP_GROUP_NBITS));
+	extra = (BITMAP_GROUP_NBITS - (binfo->nbits & BITMAP_GROUP_NBITS_MASK))
+	    & BITMAP_GROUP_NBITS_MASK;
 	if (extra != 0)
-		bitmap[binfo->ngroups - 1] >>= (BITMAP_GROUP_NBITS - extra);
+		bitmap[binfo->ngroups - 1] >>= extra;
 }

 #endif /* USE_TREE */
--- a/src/chunk.c
+++ b/src/chunk.c
@@ -49,9 +49,10 @@ const chunk_hooks_t	chunk_hooks_default = {
 * definition.
 */

-static void	chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, bool cache,
-    void *chunk, size_t size, bool zeroed, bool committed);
+static void	chunk_record(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, extent_tree_t *chunks_szad,
+    extent_tree_t *chunks_ad, bool cache, void *chunk, size_t size, bool zeroed,
+    bool committed);

 /******************************************************************************/

@@ -63,23 +64,23 @@ chunk_hooks_get_locked(arena_t *arena)
 }

 chunk_hooks_t
-chunk_hooks_get(arena_t *arena)
+chunk_hooks_get(tsdn_t *tsdn, arena_t *arena)
 {
 	chunk_hooks_t chunk_hooks;

-	malloc_mutex_lock(&arena->chunks_mtx);
+	malloc_mutex_lock(tsdn, &arena->chunks_mtx);
 	chunk_hooks = chunk_hooks_get_locked(arena);
-	malloc_mutex_unlock(&arena->chunks_mtx);
+	malloc_mutex_unlock(tsdn, &arena->chunks_mtx);

 	return (chunk_hooks);
 }

 chunk_hooks_t
-chunk_hooks_set(arena_t *arena, const chunk_hooks_t *chunk_hooks)
+chunk_hooks_set(tsdn_t *tsdn, arena_t *arena, const chunk_hooks_t *chunk_hooks)
 {
 	chunk_hooks_t old_chunk_hooks;

-	malloc_mutex_lock(&arena->chunks_mtx);
+	malloc_mutex_lock(tsdn, &arena->chunks_mtx);
 	old_chunk_hooks = arena->chunk_hooks;
 	/*
 	 * Copy each field atomically so that it is impossible for readers to
@@ -104,14 +105,14 @@ chunk_hooks_set(arena_t *arena, const chunk_hooks_t *chunk_hooks)
 	ATOMIC_COPY_HOOK(split);
 	ATOMIC_COPY_HOOK(merge);
 #undef ATOMIC_COPY_HOOK
-	malloc_mutex_unlock(&arena->chunks_mtx);
+	malloc_mutex_unlock(tsdn, &arena->chunks_mtx);

 	return (old_chunk_hooks);
 }

 static void
-chunk_hooks_assure_initialized_impl(arena_t *arena, chunk_hooks_t *chunk_hooks,
-    bool locked)
+chunk_hooks_assure_initialized_impl(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks, bool locked)
 {
 	static const chunk_hooks_t uninitialized_hooks =
 	    CHUNK_HOOKS_INITIALIZER;
@@ -119,27 +120,28 @@ chunk_hooks_assure_initialized_impl(arena_t *arena, chunk_hooks_t *chunk_hooks,
 	if (memcmp(chunk_hooks, &uninitialized_hooks, sizeof(chunk_hooks_t)) ==
 	    0) {
 		*chunk_hooks = locked ? chunk_hooks_get_locked(arena) :
-		    chunk_hooks_get(arena);
+		    chunk_hooks_get(tsdn, arena);
 	}
 }

 static void
-chunk_hooks_assure_initialized_locked(arena_t *arena,
+chunk_hooks_assure_initialized_locked(tsdn_t *tsdn, arena_t *arena,
    chunk_hooks_t *chunk_hooks)
 {

-	chunk_hooks_assure_initialized_impl(arena, chunk_hooks, true);
+	chunk_hooks_assure_initialized_impl(tsdn, arena, chunk_hooks, true);
 }

 static void
-chunk_hooks_assure_initialized(arena_t *arena, chunk_hooks_t *chunk_hooks)
+chunk_hooks_assure_initialized(tsdn_t *tsdn, arena_t *arena,
+    chunk_hooks_t *chunk_hooks)
 {

-	chunk_hooks_assure_initialized_impl(arena, chunk_hooks, false);
+	chunk_hooks_assure_initialized_impl(tsdn, arena, chunk_hooks, false);
 }

 bool
-chunk_register(const void *chunk, const extent_node_t *node)
+chunk_register(tsdn_t *tsdn, const void *chunk, const extent_node_t *node)
 {

 	assert(extent_node_addr_get(node) == chunk);
@@ -159,7 +161,7 @@ chunk_register(const void *chunk, const extent_node_t *node)
 			high = atomic_read_z(&highchunks);
 		}
 		if (cur > high && prof_gdump_get_unlocked())
-			prof_gdump();
+			prof_gdump(tsdn);
 	}

 	return (false);
@@ -197,7 +199,7 @@ chunk_first_best_fit(arena_t *arena, extent_tree_t *chunks_szad,
 }

 static void *
-chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
+chunk_recycle(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
    extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, bool cache,
    void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit,
    bool dalloc_node)
@@ -219,8 +221,8 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 	/* Beware size_t wrap-around. */
 	if (alloc_size < size)
 		return (NULL);
-	malloc_mutex_lock(&arena->chunks_mtx);
-	chunk_hooks_assure_initialized_locked(arena, chunk_hooks);
+	malloc_mutex_lock(tsdn, &arena->chunks_mtx);
+	chunk_hooks_assure_initialized_locked(tsdn, arena, chunk_hooks);
 	if (new_addr != NULL) {
 		extent_node_t key;
 		extent_node_init(&key, arena, new_addr, alloc_size, false,
@@ -232,7 +234,7 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 	}
 	if (node == NULL || (new_addr != NULL && extent_node_size_get(node) <
 	    size)) {
-		malloc_mutex_unlock(&arena->chunks_mtx);
+		malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
 		return (NULL);
 	}
 	leadsize = ALIGNMENT_CEILING((uintptr_t)extent_node_addr_get(node),
@@ -251,7 +253,7 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 	if (leadsize != 0 &&
 	    chunk_hooks->split(extent_node_addr_get(node),
 	    extent_node_size_get(node), leadsize, size, false, arena->ind)) {
-		malloc_mutex_unlock(&arena->chunks_mtx);
+		malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
 		return (NULL);
 	}
 	/* Remove node from the tree. */
@@ -271,20 +273,21 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 		if (chunk_hooks->split(ret, size + trailsize, size,
 		    trailsize, false, arena->ind)) {
 			if (dalloc_node && node != NULL)
-				arena_node_dalloc(arena, node);
-			malloc_mutex_unlock(&arena->chunks_mtx);
-			chunk_record(arena, chunk_hooks, chunks_szad, chunks_ad,
-			    cache, ret, size + trailsize, zeroed, committed);
+				arena_node_dalloc(tsdn, arena, node);
+			malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
+			chunk_record(tsdn, arena, chunk_hooks, chunks_szad,
+			    chunks_ad, cache, ret, size + trailsize, zeroed,
+			    committed);
 			return (NULL);
 		}
 		/* Insert the trailing space as a smaller chunk. */
 		if (node == NULL) {
-			node = arena_node_alloc(arena);
+			node = arena_node_alloc(tsdn, arena);
 			if (node == NULL) {
-				malloc_mutex_unlock(&arena->chunks_mtx);
-				chunk_record(arena, chunk_hooks, chunks_szad,
-				    chunks_ad, cache, ret, size + trailsize,
-				    zeroed, committed);
+				malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
+				chunk_record(tsdn, arena, chunk_hooks,
+				    chunks_szad, chunks_ad, cache, ret, size +
+				    trailsize, zeroed, committed);
 				return (NULL);
 			}
 		}
@@ -296,16 +299,16 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 		node = NULL;
 	}
 	if (!committed && chunk_hooks->commit(ret, size, 0, size, arena->ind)) {
-		malloc_mutex_unlock(&arena->chunks_mtx);
-		chunk_record(arena, chunk_hooks, chunks_szad, chunks_ad, cache,
-		    ret, size, zeroed, committed);
+		malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
+		chunk_record(tsdn, arena, chunk_hooks, chunks_szad, chunks_ad,
+		    cache, ret, size, zeroed, committed);
 		return (NULL);
 	}
-	malloc_mutex_unlock(&arena->chunks_mtx);
+	malloc_mutex_unlock(tsdn, &arena->chunks_mtx);

 	assert(dalloc_node || node != NULL);
 	if (dalloc_node && node != NULL)
-		arena_node_dalloc(arena, node);
+		arena_node_dalloc(tsdn, arena, node);
 	if (*zero) {
 		if (!zeroed)
 			memset(ret, 0, size);
@@ -328,8 +331,8 @@ chunk_recycle(arena_t *arena, chunk_hooks_t *chunk_hooks,
 * them if they are returned.
 */
 static void *
-chunk_alloc_core(arena_t *arena, void *new_addr, size_t size, size_t alignment,
-    bool *zero, bool *commit, dss_prec_t dss_prec)
+chunk_alloc_core(tsdn_t *tsdn, arena_t *arena, void *new_addr, size_t size,
+    size_t alignment, bool *zero, bool *commit, dss_prec_t dss_prec)
 {
 	void *ret;

@@ -340,8 +343,8 @@ chunk_alloc_core(arena_t *arena, void *new_addr, size_t size, size_t alignment,

 	/* "primary" dss. */
 	if (have_dss && dss_prec == dss_prec_primary && (ret =
-	    chunk_alloc_dss(arena, new_addr, size, alignment, zero, commit)) !=
-	    NULL)
+	    chunk_alloc_dss(tsdn, arena, new_addr, size, alignment, zero,
+	    commit)) != NULL)
 		return (ret);
 	/* mmap. */
 	if ((ret = chunk_alloc_mmap(new_addr, size, alignment, zero, commit)) !=
@@ -349,8 +352,8 @@ chunk_alloc_core(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 		return (ret);
 	/* "secondary" dss. */
 	if (have_dss && dss_prec == dss_prec_secondary && (ret =
-	    chunk_alloc_dss(arena, new_addr, size, alignment, zero, commit)) !=
-	    NULL)
+	    chunk_alloc_dss(tsdn, arena, new_addr, size, alignment, zero,
+	    commit)) != NULL)
 		return (ret);

 	/* All strategies for allocation failed. */
@@ -380,8 +383,8 @@ chunk_alloc_base(size_t size)
 }

 void *
-chunk_alloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
-    size_t size, size_t alignment, bool *zero, bool dalloc_node)
+chunk_alloc_cache(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *new_addr, size_t size, size_t alignment, bool *zero, bool dalloc_node)
 {
 	void *ret;
 	bool commit;
@@ -392,9 +395,9 @@ chunk_alloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
 	assert((alignment & chunksize_mask) == 0);

 	commit = true;
-	ret = chunk_recycle(arena, chunk_hooks, &arena->chunks_szad_cached,
-	    &arena->chunks_ad_cached, true, new_addr, size, alignment, zero,
-	    &commit, dalloc_node);
+	ret = chunk_recycle(tsdn, arena, chunk_hooks,
+	    &arena->chunks_szad_cached, &arena->chunks_ad_cached, true,
+	    new_addr, size, alignment, zero, &commit, dalloc_node);
 	if (ret == NULL)
 		return (NULL);
 	assert(commit);
@@ -404,11 +407,11 @@ chunk_alloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
 }

 static arena_t *
-chunk_arena_get(unsigned arena_ind)
+chunk_arena_get(tsdn_t *tsdn, unsigned arena_ind)
 {
 	arena_t *arena;

-	arena = arena_get(arena_ind, false);
+	arena = arena_get(tsdn, arena_ind, false);
 	/*
 	 * The arena we're allocating on behalf of must have been initialized
 	 * already.
@@ -418,14 +421,12 @@ chunk_arena_get(unsigned arena_ind)
 }

 static void *
-chunk_alloc_default(void *new_addr, size_t size, size_t alignment, bool *zero,
-    bool *commit, unsigned arena_ind)
+chunk_alloc_default_impl(tsdn_t *tsdn, arena_t *arena, void *new_addr,
+    size_t size, size_t alignment, bool *zero, bool *commit)
 {
 	void *ret;
-	arena_t *arena;

-	arena = chunk_arena_get(arena_ind);
-	ret = chunk_alloc_core(arena, new_addr, size, alignment, zero,
+	ret = chunk_alloc_core(tsdn, arena, new_addr, size, alignment, zero,
 	    commit, arena->dss_prec);
 	if (ret == NULL)
 		return (NULL);
@@ -436,44 +437,73 @@ chunk_alloc_default(void *new_addr, size_t size, size_t alignment, bool *zero,
 }

 static void *
-chunk_alloc_retained(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
-    size_t size, size_t alignment, bool *zero, bool *commit)
+chunk_alloc_default(void *new_addr, size_t size, size_t alignment, bool *zero,
+    bool *commit, unsigned arena_ind)
 {
+	tsdn_t *tsdn;
+	arena_t *arena;
+
+	tsdn = tsdn_fetch();
+	arena = chunk_arena_get(tsdn, arena_ind);
+
+	return (chunk_alloc_default_impl(tsdn, arena, new_addr, size, alignment,
+	    zero, commit));
+}
+
+static void *
+chunk_alloc_retained(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit)
+{
+	void *ret;

 	assert(size != 0);
 	assert((size & chunksize_mask) == 0);
 	assert(alignment != 0);
 	assert((alignment & chunksize_mask) == 0);

-	return (chunk_recycle(arena, chunk_hooks, &arena->chunks_szad_retained,
-	    &arena->chunks_ad_retained, false, new_addr, size, alignment, zero,
-	    commit, true));
+	ret = chunk_recycle(tsdn, arena, chunk_hooks,
+	    &arena->chunks_szad_retained, &arena->chunks_ad_retained, false,
+	    new_addr, size, alignment, zero, commit, true);
+
+	if (config_stats && ret != NULL)
+		arena->stats.retained -= size;
+
+	return (ret);
 }

 void *
-chunk_alloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks, void *new_addr,
-    size_t size, size_t alignment, bool *zero, bool *commit)
+chunk_alloc_wrapper(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *new_addr, size_t size, size_t alignment, bool *zero, bool *commit)
 {
 	void *ret;

-	chunk_hooks_assure_initialized(arena, chunk_hooks);
+	chunk_hooks_assure_initialized(tsdn, arena, chunk_hooks);

-	ret = chunk_alloc_retained(arena, chunk_hooks, new_addr, size,
+	ret = chunk_alloc_retained(tsdn, arena, chunk_hooks, new_addr, size,
 	    alignment, zero, commit);
 	if (ret == NULL) {
-		ret = chunk_hooks->alloc(new_addr, size, alignment, zero,
-		    commit, arena->ind);
+		if (chunk_hooks->alloc == chunk_alloc_default) {
+			/* Call directly to propagate tsdn. */
+			ret = chunk_alloc_default_impl(tsdn, arena, new_addr,
+			    size, alignment, zero, commit);
+		} else {
+			ret = chunk_hooks->alloc(new_addr, size, alignment,
+			    zero, commit, arena->ind);
+		}
+
 		if (ret == NULL)
 			return (NULL);
+
+		if (config_valgrind && chunk_hooks->alloc !=
+		    chunk_alloc_default)
+			JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ret, chunksize);
 	}

-	if (config_valgrind && chunk_hooks->alloc != chunk_alloc_default)
-		JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ret, chunksize);
 	return (ret);
 }

 static void
-chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
+chunk_record(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
    extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, bool cache,
    void *chunk, size_t size, bool zeroed, bool committed)
 {
@@ -485,8 +515,8 @@ chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
 	unzeroed = cache || !zeroed;
 	JEMALLOC_VALGRIND_MAKE_MEM_NOACCESS(chunk, size);

-	malloc_mutex_lock(&arena->chunks_mtx);
-	chunk_hooks_assure_initialized_locked(arena, chunk_hooks);
+	malloc_mutex_lock(tsdn, &arena->chunks_mtx);
+	chunk_hooks_assure_initialized_locked(tsdn, arena, chunk_hooks);
 	extent_node_init(&key, arena, (void *)((uintptr_t)chunk + size), 0,
 	    false, false);
 	node = extent_tree_ad_nsearch(chunks_ad, &key);
@@ -511,7 +541,7 @@ chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
 		arena_chunk_cache_maybe_insert(arena, node, cache);
 	} else {
 		/* Coalescing forward failed, so insert a new node. */
-		node = arena_node_alloc(arena);
+		node = arena_node_alloc(tsdn, arena);
 		if (node == NULL) {
 			/*
 			 * Node allocation failed, which is an exceedingly
@@ -520,8 +550,8 @@ chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
 			 * a virtual memory leak.
 			 */
 			if (cache) {
-				chunk_purge_wrapper(arena, chunk_hooks, chunk,
-				    size, 0, size);
+				chunk_purge_wrapper(tsdn, arena, chunk_hooks,
+				    chunk, size, 0, size);
 			}
 			goto label_return;
 		}
@@ -557,16 +587,16 @@ chunk_record(arena_t *arena, chunk_hooks_t *chunk_hooks,
 		extent_tree_szad_insert(chunks_szad, node);
 		arena_chunk_cache_maybe_insert(arena, node, cache);

-		arena_node_dalloc(arena, prev);
+		arena_node_dalloc(tsdn, arena, prev);
 	}

 label_return:
-	malloc_mutex_unlock(&arena->chunks_mtx);
+	malloc_mutex_unlock(tsdn, &arena->chunks_mtx);
 }

 void
-chunk_dalloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
-    size_t size, bool committed)
+chunk_dalloc_cache(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *chunk, size_t size, bool committed)
 {

 	assert(chunk != NULL);
@@ -574,24 +604,51 @@ chunk_dalloc_cache(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
 	assert(size != 0);
 	assert((size & chunksize_mask) == 0);

-	chunk_record(arena, chunk_hooks, &arena->chunks_szad_cached,
+	chunk_record(tsdn, arena, chunk_hooks, &arena->chunks_szad_cached,
 	    &arena->chunks_ad_cached, true, chunk, size, false, committed);
-	arena_maybe_purge(arena);
+	arena_maybe_purge(tsdn, arena);
+}
+
+static bool
+chunk_dalloc_default_impl(tsdn_t *tsdn, void *chunk, size_t size)
+{
+
+	if (!have_dss || !chunk_in_dss(tsdn, chunk))
+		return (chunk_dalloc_mmap(chunk, size));
+	return (true);
+}
+
+static bool
+chunk_dalloc_default(void *chunk, size_t size, bool committed,
+    unsigned arena_ind)
+{
+	tsdn_t *tsdn;
+
+	tsdn = tsdn_fetch();
+
+	return (chunk_dalloc_default_impl(tsdn, chunk, size));
 }

 void
-chunk_dalloc_arena(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
-    size_t size, bool zeroed, bool committed)
+chunk_dalloc_wrapper(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *chunk, size_t size, bool zeroed, bool committed)
 {
+	bool err;

 	assert(chunk != NULL);
 	assert(CHUNK_ADDR2BASE(chunk) == chunk);
 	assert(size != 0);
 	assert((size & chunksize_mask) == 0);

-	chunk_hooks_assure_initialized(arena, chunk_hooks);
+	chunk_hooks_assure_initialized(tsdn, arena, chunk_hooks);
 	/* Try to deallocate. */
-	if (!chunk_hooks->dalloc(chunk, size, committed, arena->ind))
+	if (chunk_hooks->dalloc == chunk_dalloc_default) {
+		/* Call directly to propagate tsdn. */
+		err = chunk_dalloc_default_impl(tsdn, chunk, size);
+	} else
+		err = chunk_hooks->dalloc(chunk, size, committed, arena->ind);
+
+	if (!err)
 		return;
 	/* Try to decommit; purge if that fails. */
 	if (committed) {
@@ -600,29 +657,11 @@ chunk_dalloc_arena(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
 	}
 	zeroed = !committed || !chunk_hooks->purge(chunk, size, 0, size,
 	    arena->ind);
-	chunk_record(arena, chunk_hooks, &arena->chunks_szad_retained,
+	chunk_record(tsdn, arena, chunk_hooks, &arena->chunks_szad_retained,
 	    &arena->chunks_ad_retained, false, chunk, size, zeroed, committed);
-}

-static bool
-chunk_dalloc_default(void *chunk, size_t size, bool committed,
-    unsigned arena_ind)
-{
-
-	if (!have_dss || !chunk_in_dss(chunk))
-		return (chunk_dalloc_mmap(chunk, size));
-	return (true);
-}
-
-void
-chunk_dalloc_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
-    size_t size, bool committed)
-{
-
-	chunk_hooks_assure_initialized(arena, chunk_hooks);
-	chunk_hooks->dalloc(chunk, size, committed, arena->ind);
-	if (config_valgrind && chunk_hooks->dalloc != chunk_dalloc_default)
-		JEMALLOC_VALGRIND_MAKE_MEM_NOACCESS(chunk, size);
+	if (config_stats)
+		arena->stats.retained += size;
 }

 static bool
@@ -643,8 +682,9 @@ chunk_decommit_default(void *chunk, size_t size, size_t offset, size_t length,
 	    length));
 }

-bool
-chunk_purge_arena(arena_t *arena, void *chunk, size_t offset, size_t length)
+static bool
+chunk_purge_default(void *chunk, size_t size, size_t offset, size_t length,
+    unsigned arena_ind)
 {

 	assert(chunk != NULL);
@@ -657,21 +697,12 @@ chunk_purge_arena(arena_t *arena, void *chunk, size_t offset, size_t length)
 	    length));
 }

-static bool
-chunk_purge_default(void *chunk, size_t size, size_t offset, size_t length,
-    unsigned arena_ind)
-{
-
-	return (chunk_purge_arena(chunk_arena_get(arena_ind), chunk, offset,
-	    length));
-}
-
 bool
-chunk_purge_wrapper(arena_t *arena, chunk_hooks_t *chunk_hooks, void *chunk,
-    size_t size, size_t offset, size_t length)
+chunk_purge_wrapper(tsdn_t *tsdn, arena_t *arena, chunk_hooks_t *chunk_hooks,
+    void *chunk, size_t size, size_t offset, size_t length)
 {

-	chunk_hooks_assure_initialized(arena, chunk_hooks);
+	chunk_hooks_assure_initialized(tsdn, arena, chunk_hooks);
 	return (chunk_hooks->purge(chunk, size, offset, length, arena->ind));
 }

@@ -686,23 +717,34 @@ chunk_split_default(void *chunk, size_t size, size_t size_a, size_t size_b,
 }

 static bool
-chunk_merge_default(void *chunk_a, size_t size_a, void *chunk_b, size_t size_b,
-    bool committed, unsigned arena_ind)
+chunk_merge_default_impl(tsdn_t *tsdn, void *chunk_a, void *chunk_b)
 {

 	if (!maps_coalesce)
 		return (true);
-	if (have_dss && chunk_in_dss(chunk_a) != chunk_in_dss(chunk_b))
+	if (have_dss && chunk_in_dss(tsdn, chunk_a) != chunk_in_dss(tsdn,
+	    chunk_b))
 		return (true);

 	return (false);
 }

+static bool
+chunk_merge_default(void *chunk_a, size_t size_a, void *chunk_b, size_t size_b,
+    bool committed, unsigned arena_ind)
+{
+	tsdn_t *tsdn;
+
+	tsdn = tsdn_fetch();
+
+	return (chunk_merge_default_impl(tsdn, chunk_a, chunk_b));
+}
+
 static rtree_node_elm_t *
 chunks_rtree_node_alloc(size_t nelms)
 {

-	return ((rtree_node_elm_t *)base_alloc(nelms *
+	return ((rtree_node_elm_t *)base_alloc(TSDN_NULL, nelms *
 	    sizeof(rtree_node_elm_t)));
 }

@@ -749,22 +791,22 @@ chunk_boot(void)
 }

 void
-chunk_prefork(void)
+chunk_prefork(tsdn_t *tsdn)
 {

-	chunk_dss_prefork();
+	chunk_dss_prefork(tsdn);
 }

 void
-chunk_postfork_parent(void)
+chunk_postfork_parent(tsdn_t *tsdn)
 {

-	chunk_dss_postfork_parent();
+	chunk_dss_postfork_parent(tsdn);
 }

 void
-chunk_postfork_child(void)
+chunk_postfork_child(tsdn_t *tsdn)
 {

-	chunk_dss_postfork_child();
+	chunk_dss_postfork_child(tsdn);
 }
--- a/src/chunk_dss.c
+++ b/src/chunk_dss.c
@@ -41,33 +41,33 @@ chunk_dss_sbrk(intptr_t increment)
 }

 dss_prec_t
-chunk_dss_prec_get(void)
+chunk_dss_prec_get(tsdn_t *tsdn)
 {
 	dss_prec_t ret;

 	if (!have_dss)
 		return (dss_prec_disabled);
-	malloc_mutex_lock(&dss_mtx);
+	malloc_mutex_lock(tsdn, &dss_mtx);
 	ret = dss_prec_default;
-	malloc_mutex_unlock(&dss_mtx);
+	malloc_mutex_unlock(tsdn, &dss_mtx);
 	return (ret);
 }

 bool
-chunk_dss_prec_set(dss_prec_t dss_prec)
+chunk_dss_prec_set(tsdn_t *tsdn, dss_prec_t dss_prec)
 {

 	if (!have_dss)
 		return (dss_prec != dss_prec_disabled);
-	malloc_mutex_lock(&dss_mtx);
+	malloc_mutex_lock(tsdn, &dss_mtx);
 	dss_prec_default = dss_prec;
-	malloc_mutex_unlock(&dss_mtx);
+	malloc_mutex_unlock(tsdn, &dss_mtx);
 	return (false);
 }

 void *
-chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
-    bool *zero, bool *commit)
+chunk_alloc_dss(tsdn_t *tsdn, arena_t *arena, void *new_addr, size_t size,
+    size_t alignment, bool *zero, bool *commit)
 {
 	cassert(have_dss);
 	assert(size > 0 && (size & chunksize_mask) == 0);
@@ -80,7 +80,7 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 	if ((intptr_t)size < 0)
 		return (NULL);

-	malloc_mutex_lock(&dss_mtx);
+	malloc_mutex_lock(tsdn, &dss_mtx);
 	if (dss_prev != (void *)-1) {

 		/*
@@ -122,7 +122,7 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 			if ((uintptr_t)ret < (uintptr_t)dss_max ||
 			    (uintptr_t)dss_next < (uintptr_t)dss_max) {
 				/* Wrap-around. */
-				malloc_mutex_unlock(&dss_mtx);
+				malloc_mutex_unlock(tsdn, &dss_mtx);
 				return (NULL);
 			}
 			incr = gap_size + cpad_size + size;
@@ -130,13 +130,13 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 			if (dss_prev == dss_max) {
 				/* Success. */
 				dss_max = dss_next;
-				malloc_mutex_unlock(&dss_mtx);
+				malloc_mutex_unlock(tsdn, &dss_mtx);
 				if (cpad_size != 0) {
 					chunk_hooks_t chunk_hooks =
 					    CHUNK_HOOKS_INITIALIZER;
-					chunk_dalloc_wrapper(arena,
+					chunk_dalloc_wrapper(tsdn, arena,
 					    &chunk_hooks, cpad, cpad_size,
-					    true);
+					    false, true);
 				}
 				if (*zero) {
 					JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(
@@ -149,25 +149,25 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 			}
 		} while (dss_prev != (void *)-1);
 	}
-	malloc_mutex_unlock(&dss_mtx);
+	malloc_mutex_unlock(tsdn, &dss_mtx);

 	return (NULL);
 }

 bool
-chunk_in_dss(void *chunk)
+chunk_in_dss(tsdn_t *tsdn, void *chunk)
 {
 	bool ret;

 	cassert(have_dss);

-	malloc_mutex_lock(&dss_mtx);
+	malloc_mutex_lock(tsdn, &dss_mtx);
 	if ((uintptr_t)chunk >= (uintptr_t)dss_base
 	    && (uintptr_t)chunk < (uintptr_t)dss_max)
 		ret = true;
 	else
 		ret = false;
-	malloc_mutex_unlock(&dss_mtx);
+	malloc_mutex_unlock(tsdn, &dss_mtx);

 	return (ret);
 }
@@ -178,7 +178,7 @@ chunk_dss_boot(void)

 	cassert(have_dss);

-	if (malloc_mutex_init(&dss_mtx))
+	if (malloc_mutex_init(&dss_mtx, "dss", WITNESS_RANK_DSS))
 		return (true);
 	dss_base = chunk_dss_sbrk(0);
 	dss_prev = dss_base;
@@ -188,27 +188,27 @@ chunk_dss_boot(void)
 }

 void
-chunk_dss_prefork(void)
+chunk_dss_prefork(tsdn_t *tsdn)
 {

 	if (have_dss)
-		malloc_mutex_prefork(&dss_mtx);
+		malloc_mutex_prefork(tsdn, &dss_mtx);
 }

 void
-chunk_dss_postfork_parent(void)
+chunk_dss_postfork_parent(tsdn_t *tsdn)
 {

 	if (have_dss)
-		malloc_mutex_postfork_parent(&dss_mtx);
+		malloc_mutex_postfork_parent(tsdn, &dss_mtx);
 }

 void
-chunk_dss_postfork_child(void)
+chunk_dss_postfork_child(tsdn_t *tsdn)
 {

 	if (have_dss)
-		malloc_mutex_postfork_child(&dss_mtx);
+		malloc_mutex_postfork_child(tsdn, &dss_mtx);
 }

 /******************************************************************************/
--- a/src/chunk_mmap.c
+++ b/src/chunk_mmap.c
@@ -16,18 +16,16 @@ chunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero, bool *commit)
 	do {
 		void *pages;
 		size_t leadsize;
-		pages = pages_map(NULL, alloc_size);
+		pages = pages_map(NULL, alloc_size, commit);
 		if (pages == NULL)
 			return (NULL);
 		leadsize = ALIGNMENT_CEILING((uintptr_t)pages, alignment) -
 		    (uintptr_t)pages;
-		ret = pages_trim(pages, alloc_size, leadsize, size);
+		ret = pages_trim(pages, alloc_size, leadsize, size, commit);
 	} while (ret == NULL);

 	assert(ret != NULL);
 	*zero = true;
-	if (!*commit)
-		*commit = pages_decommit(ret, size);
 	return (ret);
 }

@@ -54,7 +52,7 @@ chunk_alloc_mmap(void *new_addr, size_t size, size_t alignment, bool *zero,
 	assert(alignment != 0);
 	assert((alignment & chunksize_mask) == 0);

-	ret = pages_map(new_addr, size);
+	ret = pages_map(new_addr, size, commit);
 	if (ret == NULL || ret == new_addr)
 		return (ret);
 	assert(new_addr == NULL);
@@ -66,8 +64,6 @@ chunk_alloc_mmap(void *new_addr, size_t size, size_t alignment, bool *zero,

 	assert(ret != NULL);
 	*zero = true;
-	if (!*commit)
-		*commit = pages_decommit(ret, size);
 	return (ret);
 }

--- a/src/ckh.c
+++ b/src/ckh.c
@@ -40,8 +40,8 @@
 /******************************************************************************/
 /* Function prototypes for non-inline static functions. */

-static bool	ckh_grow(tsd_t *tsd, ckh_t *ckh);
-static void	ckh_shrink(tsd_t *tsd, ckh_t *ckh);
+static bool	ckh_grow(tsdn_t *tsdn, ckh_t *ckh);
+static void	ckh_shrink(tsdn_t *tsdn, ckh_t *ckh);

 /******************************************************************************/

@@ -244,7 +244,7 @@ ckh_rebuild(ckh_t *ckh, ckhc_t *aTab)
 }

 static bool
-ckh_grow(tsd_t *tsd, ckh_t *ckh)
+ckh_grow(tsdn_t *tsdn, ckh_t *ckh)
 {
 	bool ret;
 	ckhc_t *tab, *ttab;
@@ -270,8 +270,8 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh)
 			ret = true;
 			goto label_return;
 		}
-		tab = (ckhc_t *)ipallocztm(tsd, usize, CACHELINE, true, NULL,
-		    true, NULL);
+		tab = (ckhc_t *)ipallocztm(tsdn, usize, CACHELINE, true, NULL,
+		    true, arena_ichoose(tsdn, NULL));
 		if (tab == NULL) {
 			ret = true;
 			goto label_return;
@@ -283,12 +283,12 @@ ckh_grow(tsd_t *tsd, ckh_t *ckh)
 		ckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;

 		if (!ckh_rebuild(ckh, tab)) {
-			idalloctm(tsd, tab, tcache_get(tsd, false), true, true);
+			idalloctm(tsdn, tab, NULL, true, true);
 			break;
 		}

 		/* Rebuilding failed, so back out partially rebuilt table. */
-		idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
+		idalloctm(tsdn, ckh->tab, NULL, true, true);
 		ckh->tab = tab;
 		ckh->lg_curbuckets = lg_prevbuckets;
 	}
@@ -299,7 +299,7 @@ label_return:
 }

 static void
-ckh_shrink(tsd_t *tsd, ckh_t *ckh)
+ckh_shrink(tsdn_t *tsdn, ckh_t *ckh)
 {
 	ckhc_t *tab, *ttab;
 	size_t usize;
@@ -314,8 +314,8 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	usize = sa2u(sizeof(ckhc_t) << lg_curcells, CACHELINE);
 	if (unlikely(usize == 0 || usize > HUGE_MAXCLASS))
 		return;
-	tab = (ckhc_t *)ipallocztm(tsd, usize, CACHELINE, true, NULL, true,
-	    NULL);
+	tab = (ckhc_t *)ipallocztm(tsdn, usize, CACHELINE, true, NULL, true,
+	    arena_ichoose(tsdn, NULL));
 	if (tab == NULL) {
 		/*
 		 * An OOM error isn't worth propagating, since it doesn't
@@ -330,7 +330,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	ckh->lg_curbuckets = lg_curcells - LG_CKH_BUCKET_CELLS;

 	if (!ckh_rebuild(ckh, tab)) {
-		idalloctm(tsd, tab, tcache_get(tsd, false), true, true);
+		idalloctm(tsdn, tab, NULL, true, true);
 #ifdef CKH_COUNT
 		ckh->nshrinks++;
 #endif
@@ -338,7 +338,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 	}

 	/* Rebuilding failed, so back out partially rebuilt table. */
-	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
+	idalloctm(tsdn, ckh->tab, NULL, true, true);
 	ckh->tab = tab;
 	ckh->lg_curbuckets = lg_prevbuckets;
 #ifdef CKH_COUNT
@@ -347,7 +347,7 @@ ckh_shrink(tsd_t *tsd, ckh_t *ckh)
 }

 bool
-ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
+ckh_new(tsdn_t *tsdn, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
    ckh_keycomp_t *keycomp)
 {
 	bool ret;
@@ -391,8 +391,8 @@ ckh_new(tsd_t *tsd, ckh_t *ckh, size_t minitems, ckh_hash_t *hash,
 		ret = true;
 		goto label_return;
 	}
-	ckh->tab = (ckhc_t *)ipallocztm(tsd, usize, CACHELINE, true, NULL, true,
-	    NULL);
+	ckh->tab = (ckhc_t *)ipallocztm(tsdn, usize, CACHELINE, true, NULL,
+	    true, arena_ichoose(tsdn, NULL));
 	if (ckh->tab == NULL) {
 		ret = true;
 		goto label_return;
@@ -404,7 +404,7 @@ label_return:
 }

 void
-ckh_delete(tsd_t *tsd, ckh_t *ckh)
+ckh_delete(tsdn_t *tsdn, ckh_t *ckh)
 {

 	assert(ckh != NULL);
@@ -421,9 +421,9 @@ ckh_delete(tsd_t *tsd, ckh_t *ckh)
 	    (unsigned long long)ckh->nrelocs);
 #endif

-	idalloctm(tsd, ckh->tab, tcache_get(tsd, false), true, true);
+	idalloctm(tsdn, ckh->tab, NULL, true, true);
 	if (config_debug)
-		memset(ckh, 0x5a, sizeof(ckh_t));
+		memset(ckh, JEMALLOC_FREE_JUNK, sizeof(ckh_t));
 }

 size_t
@@ -456,7 +456,7 @@ ckh_iter(ckh_t *ckh, size_t *tabind, void **key, void **data)
 }

 bool
-ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data)
+ckh_insert(tsdn_t *tsdn, ckh_t *ckh, const void *key, const void *data)
 {
 	bool ret;

@@ -468,7 +468,7 @@ ckh_insert(tsd_t *tsd, ckh_t *ckh, const void *key, const void *data)
 #endif

 	while (ckh_try_insert(ckh, &key, &data)) {
-		if (ckh_grow(tsd, ckh)) {
+		if (ckh_grow(tsdn, ckh)) {
 			ret = true;
 			goto label_return;
 		}
@@ -480,7 +480,7 @@ label_return:
 }

 bool
-ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
+ckh_remove(tsdn_t *tsdn, ckh_t *ckh, const void *searchkey, void **key,
    void **data)
 {
 	size_t cell;
@@ -502,7 +502,7 @@ ckh_remove(tsd_t *tsd, ckh_t *ckh, const void *searchkey, void **key,
 		    + LG_CKH_BUCKET_CELLS - 2)) && ckh->lg_curbuckets
 		    > ckh->lg_minbuckets) {
 			/* Ignore error due to OOM. */
-			ckh_shrink(tsd, ckh);
+			ckh_shrink(tsdn, ckh);
 		}

 		return (false);
--- a/src/ctl.c
+++ b/src/ctl.c
--- a/src/huge.c
+++ b/src/huge.c
@@ -15,12 +15,21 @@ huge_node_get(const void *ptr)
 }

 static bool
-huge_node_set(const void *ptr, extent_node_t *node)
+huge_node_set(tsdn_t *tsdn, const void *ptr, extent_node_t *node)
 {

 	assert(extent_node_addr_get(node) == ptr);
 	assert(!extent_node_achunk_get(node));
-	return (chunk_register(ptr, node));
+	return (chunk_register(tsdn, ptr, node));
+}
+
+static void
+huge_node_reset(tsdn_t *tsdn, const void *ptr, extent_node_t *node)
+{
+	bool err;
+
+	err = huge_node_set(tsdn, ptr, node);
+	assert(!err);
 }

 static void
@@ -31,18 +40,17 @@ huge_node_unset(const void *ptr, const extent_node_t *node)
 }

 void *
-huge_malloc(tsd_t *tsd, arena_t *arena, size_t usize, bool zero,
-    tcache_t *tcache)
+huge_malloc(tsdn_t *tsdn, arena_t *arena, size_t usize, bool zero)
 {

 	assert(usize == s2u(usize));

-	return (huge_palloc(tsd, arena, usize, chunksize, zero, tcache));
+	return (huge_palloc(tsdn, arena, usize, chunksize, zero));
 }

 void *
-huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
-    bool zero, tcache_t *tcache)
+huge_palloc(tsdn_t *tsdn, arena_t *arena, size_t usize, size_t alignment,
+    bool zero)
 {
 	void *ret;
 	size_t ausize;
@@ -51,14 +59,16 @@ huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,

 	/* Allocate one or more contiguous chunks for this request. */

+	assert(!tsdn_null(tsdn) || arena != NULL);
+
 	ausize = sa2u(usize, alignment);
 	if (unlikely(ausize == 0 || ausize > HUGE_MAXCLASS))
 		return (NULL);
 	assert(ausize >= chunksize);

 	/* Allocate an extent node with which to track the chunk. */
-	node = ipallocztm(tsd, CACHELINE_CEILING(sizeof(extent_node_t)),
-	    CACHELINE, false, tcache, true, arena);
+	node = ipallocztm(tsdn, CACHELINE_CEILING(sizeof(extent_node_t)),
+	    CACHELINE, false, NULL, true, arena_ichoose(tsdn, arena));
 	if (node == NULL)
 		return (NULL);

@@ -67,34 +77,35 @@ huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
 	 * it is possible to make correct junk/zero fill decisions below.
 	 */
 	is_zeroed = zero;
-	arena = arena_choose(tsd, arena);
-	if (unlikely(arena == NULL) || (ret = arena_chunk_alloc_huge(arena,
-	    usize, alignment, &is_zeroed)) == NULL) {
-		idalloctm(tsd, node, tcache, true, true);
+	if (likely(!tsdn_null(tsdn)))
+		arena = arena_choose(tsdn_tsd(tsdn), arena);
+	if (unlikely(arena == NULL) || (ret = arena_chunk_alloc_huge(tsdn,
+	    arena, usize, alignment, &is_zeroed)) == NULL) {
+		idalloctm(tsdn, node, NULL, true, true);
 		return (NULL);
 	}

 	extent_node_init(node, arena, ret, usize, is_zeroed, true);

-	if (huge_node_set(ret, node)) {
-		arena_chunk_dalloc_huge(arena, ret, usize);
-		idalloctm(tsd, node, tcache, true, true);
+	if (huge_node_set(tsdn, ret, node)) {
+		arena_chunk_dalloc_huge(tsdn, arena, ret, usize);
+		idalloctm(tsdn, node, NULL, true, true);
 		return (NULL);
 	}

 	/* Insert node into huge. */
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	ql_elm_new(node, ql_link);
 	ql_tail_insert(&arena->huge, node, ql_link);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	if (zero || (config_fill && unlikely(opt_zero))) {
 		if (!is_zeroed)
 			memset(ret, 0, usize);
 	} else if (config_fill && unlikely(opt_junk_alloc))
-		memset(ret, 0xa5, usize);
+		memset(ret, JEMALLOC_ALLOC_JUNK, usize);

-	arena_decay_tick(tsd, arena);
+	arena_decay_tick(tsdn, arena);
 	return (ret);
 }

@@ -103,7 +114,7 @@ huge_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
 #define	huge_dalloc_junk JEMALLOC_N(huge_dalloc_junk_impl)
 #endif
 static void
-huge_dalloc_junk(void *ptr, size_t usize)
+huge_dalloc_junk(tsdn_t *tsdn, void *ptr, size_t usize)
 {

 	if (config_fill && have_dss && unlikely(opt_junk_free)) {
@@ -111,8 +122,8 @@ huge_dalloc_junk(void *ptr, size_t usize)
 		 * Only bother junk filling if the chunk isn't about to be
 		 * unmapped.
 		 */
-		if (!config_munmap || (have_dss && chunk_in_dss(ptr)))
-			memset(ptr, 0x5a, usize);
+		if (!config_munmap || (have_dss && chunk_in_dss(tsdn, ptr)))
+			memset(ptr, JEMALLOC_FREE_JUNK, usize);
 	}
 }
 #ifdef JEMALLOC_JET
@@ -122,8 +133,8 @@ huge_dalloc_junk_t *huge_dalloc_junk = JEMALLOC_N(huge_dalloc_junk_impl);
 #endif

 static void
-huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
-    size_t usize_max, bool zero)
+huge_ralloc_no_move_similar(tsdn_t *tsdn, void *ptr, size_t oldsize,
+    size_t usize_min, size_t usize_max, bool zero)
 {
 	size_t usize, usize_next;
 	extent_node_t *node;
@@ -147,24 +158,28 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
 	if (oldsize > usize) {
 		size_t sdiff = oldsize - usize;
 		if (config_fill && unlikely(opt_junk_free)) {
-			memset((void *)((uintptr_t)ptr + usize), 0x5a, sdiff);
+			memset((void *)((uintptr_t)ptr + usize),
+			    JEMALLOC_FREE_JUNK, sdiff);
 			post_zeroed = false;
 		} else {
-			post_zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
-			    ptr, CHUNK_CEILING(oldsize), usize, sdiff);
+			post_zeroed = !chunk_purge_wrapper(tsdn, arena,
+			    &chunk_hooks, ptr, CHUNK_CEILING(oldsize), usize,
+			    sdiff);
 		}
 	} else
 		post_zeroed = pre_zeroed;

-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	/* Update the size of the huge allocation. */
+	huge_node_unset(ptr, node);
 	assert(extent_node_size_get(node) != usize);
 	extent_node_size_set(node, usize);
+	huge_node_reset(tsdn, ptr, node);
 	/* Update zeroed. */
 	extent_node_zeroed_set(node, post_zeroed);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

-	arena_chunk_ralloc_huge_similar(arena, ptr, oldsize, usize);
+	arena_chunk_ralloc_huge_similar(tsdn, arena, ptr, oldsize, usize);

 	/* Fill if necessary (growing). */
 	if (oldsize < usize) {
@@ -174,14 +189,15 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
 				    usize - oldsize);
 			}
 		} else if (config_fill && unlikely(opt_junk_alloc)) {
-			memset((void *)((uintptr_t)ptr + oldsize), 0xa5, usize -
-			    oldsize);
+			memset((void *)((uintptr_t)ptr + oldsize),
+			    JEMALLOC_ALLOC_JUNK, usize - oldsize);
 		}
 	}
 }

 static bool
-huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
+huge_ralloc_no_move_shrink(tsdn_t *tsdn, void *ptr, size_t oldsize,
+    size_t usize)
 {
 	extent_node_t *node;
 	arena_t *arena;
@@ -192,7 +208,7 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
 	pre_zeroed = extent_node_zeroed_get(node);
-	chunk_hooks = chunk_hooks_get(arena);
+	chunk_hooks = chunk_hooks_get(tsdn, arena);

 	assert(oldsize > usize);

@@ -205,57 +221,62 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 	if (oldsize > usize) {
 		size_t sdiff = oldsize - usize;
 		if (config_fill && unlikely(opt_junk_free)) {
-			huge_dalloc_junk((void *)((uintptr_t)ptr + usize),
+			huge_dalloc_junk(tsdn, (void *)((uintptr_t)ptr + usize),
 			    sdiff);
 			post_zeroed = false;
 		} else {
-			post_zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
-			    CHUNK_ADDR2BASE((uintptr_t)ptr + usize),
-			    CHUNK_CEILING(oldsize),
+			post_zeroed = !chunk_purge_wrapper(tsdn, arena,
+			    &chunk_hooks, CHUNK_ADDR2BASE((uintptr_t)ptr +
+			    usize), CHUNK_CEILING(oldsize),
 			    CHUNK_ADDR2OFFSET((uintptr_t)ptr + usize), sdiff);
 		}
 	} else
 		post_zeroed = pre_zeroed;

-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	/* Update the size of the huge allocation. */
+	huge_node_unset(ptr, node);
 	extent_node_size_set(node, usize);
+	huge_node_reset(tsdn, ptr, node);
 	/* Update zeroed. */
 	extent_node_zeroed_set(node, post_zeroed);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	/* Zap the excess chunks. */
-	arena_chunk_ralloc_huge_shrink(arena, ptr, oldsize, usize);
+	arena_chunk_ralloc_huge_shrink(tsdn, arena, ptr, oldsize, usize);

 	return (false);
 }

 static bool
-huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t usize, bool zero) {
+huge_ralloc_no_move_expand(tsdn_t *tsdn, void *ptr, size_t oldsize,
+    size_t usize, bool zero) {
 	extent_node_t *node;
 	arena_t *arena;
 	bool is_zeroed_subchunk, is_zeroed_chunk;

 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	is_zeroed_subchunk = extent_node_zeroed_get(node);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	/*
-	 * Copy zero into is_zeroed_chunk and pass the copy to chunk_alloc(), so
-	 * that it is possible to make correct junk/zero fill decisions below.
+	 * Use is_zeroed_chunk to detect whether the trailing memory is zeroed,
+	 * update extent's zeroed field, and zero as necessary.
 	 */
-	is_zeroed_chunk = zero;
-
-	if (arena_chunk_ralloc_huge_expand(arena, ptr, oldsize, usize,
+	is_zeroed_chunk = false;
+	if (arena_chunk_ralloc_huge_expand(tsdn, arena, ptr, oldsize, usize,
 	     &is_zeroed_chunk))
 		return (true);

-	malloc_mutex_lock(&arena->huge_mtx);
-	/* Update the size of the huge allocation. */
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
+	huge_node_unset(ptr, node);
 	extent_node_size_set(node, usize);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	extent_node_zeroed_set(node, extent_node_zeroed_get(node) &&
+	    is_zeroed_chunk);
+	huge_node_reset(tsdn, ptr, node);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	if (zero || (config_fill && unlikely(opt_zero))) {
 		if (!is_zeroed_subchunk) {
@@ -268,15 +289,15 @@ huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t usize, bool zero) {
 			    CHUNK_CEILING(oldsize));
 		}
 	} else if (config_fill && unlikely(opt_junk_alloc)) {
-		memset((void *)((uintptr_t)ptr + oldsize), 0xa5, usize -
-		    oldsize);
+		memset((void *)((uintptr_t)ptr + oldsize), JEMALLOC_ALLOC_JUNK,
+		    usize - oldsize);
 	}

 	return (false);
 }

 bool
-huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t usize_min,
+huge_ralloc_no_move(tsdn_t *tsdn, void *ptr, size_t oldsize, size_t usize_min,
    size_t usize_max, bool zero)
 {

@@ -290,16 +311,16 @@ huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t usize_min,

 	if (CHUNK_CEILING(usize_max) > CHUNK_CEILING(oldsize)) {
 		/* Attempt to expand the allocation in-place. */
-		if (!huge_ralloc_no_move_expand(ptr, oldsize, usize_max,
+		if (!huge_ralloc_no_move_expand(tsdn, ptr, oldsize, usize_max,
 		    zero)) {
-			arena_decay_tick(tsd, huge_aalloc(ptr));
+			arena_decay_tick(tsdn, huge_aalloc(ptr));
 			return (false);
 		}
 		/* Try again, this time with usize_min. */
 		if (usize_min < usize_max && CHUNK_CEILING(usize_min) >
-		    CHUNK_CEILING(oldsize) && huge_ralloc_no_move_expand(ptr,
-		    oldsize, usize_min, zero)) {
-			arena_decay_tick(tsd, huge_aalloc(ptr));
+		    CHUNK_CEILING(oldsize) && huge_ralloc_no_move_expand(tsdn,
+		    ptr, oldsize, usize_min, zero)) {
+			arena_decay_tick(tsdn, huge_aalloc(ptr));
 			return (false);
 		}
 	}
@@ -310,16 +331,17 @@ huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t usize_min,
 	 */
 	if (CHUNK_CEILING(oldsize) >= CHUNK_CEILING(usize_min)
 	    && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(usize_max)) {
-		huge_ralloc_no_move_similar(ptr, oldsize, usize_min, usize_max,
-		    zero);
-		arena_decay_tick(tsd, huge_aalloc(ptr));
+		huge_ralloc_no_move_similar(tsdn, ptr, oldsize, usize_min,
+		    usize_max, zero);
+		arena_decay_tick(tsdn, huge_aalloc(ptr));
 		return (false);
 	}

 	/* Attempt to shrink the allocation in-place. */
 	if (CHUNK_CEILING(oldsize) > CHUNK_CEILING(usize_max)) {
-		if (!huge_ralloc_no_move_shrink(ptr, oldsize, usize_max)) {
-			arena_decay_tick(tsd, huge_aalloc(ptr));
+		if (!huge_ralloc_no_move_shrink(tsdn, ptr, oldsize,
+		    usize_max)) {
+			arena_decay_tick(tsdn, huge_aalloc(ptr));
 			return (false);
 		}
 	}
@@ -327,18 +349,18 @@ huge_ralloc_no_move(tsd_t *tsd, void *ptr, size_t oldsize, size_t usize_min,
 }

 static void *
-huge_ralloc_move_helper(tsd_t *tsd, arena_t *arena, size_t usize,
-    size_t alignment, bool zero, tcache_t *tcache)
+huge_ralloc_move_helper(tsdn_t *tsdn, arena_t *arena, size_t usize,
+    size_t alignment, bool zero)
 {

 	if (alignment <= chunksize)
-		return (huge_malloc(tsd, arena, usize, zero, tcache));
-	return (huge_palloc(tsd, arena, usize, alignment, zero, tcache));
+		return (huge_malloc(tsdn, arena, usize, zero));
+	return (huge_palloc(tsdn, arena, usize, alignment, zero));
 }

 void *
-huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t usize,
-    size_t alignment, bool zero, tcache_t *tcache)
+huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
+    size_t usize, size_t alignment, bool zero, tcache_t *tcache)
 {
 	void *ret;
 	size_t copysize;
@@ -347,7 +369,8 @@ huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t usize,
 	assert(usize > 0 && usize <= HUGE_MAXCLASS);

 	/* Try to avoid moving the allocation. */
-	if (!huge_ralloc_no_move(tsd, ptr, oldsize, usize, usize, zero))
+	if (!huge_ralloc_no_move(tsd_tsdn(tsd), ptr, oldsize, usize, usize,
+	    zero))
 		return (ptr);

 	/*
@@ -355,19 +378,19 @@ huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t usize,
 	 * different size class.  In that case, fall back to allocating new
 	 * space and copying.
 	 */
-	ret = huge_ralloc_move_helper(tsd, arena, usize, alignment, zero,
-	    tcache);
+	ret = huge_ralloc_move_helper(tsd_tsdn(tsd), arena, usize, alignment,
+	    zero);
 	if (ret == NULL)
 		return (NULL);

 	copysize = (usize < oldsize) ? usize : oldsize;
 	memcpy(ret, ptr, copysize);
-	isqalloc(tsd, ptr, oldsize, tcache);
+	isqalloc(tsd, ptr, oldsize, tcache, true);
 	return (ret);
 }

 void
-huge_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
+huge_dalloc(tsdn_t *tsdn, void *ptr)
 {
 	extent_node_t *node;
 	arena_t *arena;
@@ -375,17 +398,17 @@ huge_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
 	huge_node_unset(ptr, node);
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	ql_remove(&arena->huge, node, ql_link);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

-	huge_dalloc_junk(extent_node_addr_get(node),
+	huge_dalloc_junk(tsdn, extent_node_addr_get(node),
 	    extent_node_size_get(node));
-	arena_chunk_dalloc_huge(extent_node_arena_get(node),
+	arena_chunk_dalloc_huge(tsdn, extent_node_arena_get(node),
 	    extent_node_addr_get(node), extent_node_size_get(node));
-	idalloctm(tsd, node, tcache, true, true);
+	idalloctm(tsdn, node, NULL, true, true);

-	arena_decay_tick(tsd, arena);
+	arena_decay_tick(tsdn, arena);
 }

 arena_t *
@@ -396,7 +419,7 @@ huge_aalloc(const void *ptr)
 }

 size_t
-huge_salloc(const void *ptr)
+huge_salloc(tsdn_t *tsdn, const void *ptr)
 {
 	size_t size;
 	extent_node_t *node;
@@ -404,15 +427,15 @@ huge_salloc(const void *ptr)

 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	size = extent_node_size_get(node);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	return (size);
 }

 prof_tctx_t *
-huge_prof_tctx_get(const void *ptr)
+huge_prof_tctx_get(tsdn_t *tsdn, const void *ptr)
 {
 	prof_tctx_t *tctx;
 	extent_node_t *node;
@@ -420,29 +443,29 @@ huge_prof_tctx_get(const void *ptr)

 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	tctx = extent_node_prof_tctx_get(node);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);

 	return (tctx);
 }

 void
-huge_prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
+huge_prof_tctx_set(tsdn_t *tsdn, const void *ptr, prof_tctx_t *tctx)
 {
 	extent_node_t *node;
 	arena_t *arena;

 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
-	malloc_mutex_lock(&arena->huge_mtx);
+	malloc_mutex_lock(tsdn, &arena->huge_mtx);
 	extent_node_prof_tctx_set(node, tctx);
-	malloc_mutex_unlock(&arena->huge_mtx);
+	malloc_mutex_unlock(tsdn, &arena->huge_mtx);
 }

 void
-huge_prof_tctx_reset(const void *ptr)
+huge_prof_tctx_reset(tsdn_t *tsdn, const void *ptr)
 {

-	huge_prof_tctx_set(ptr, (prof_tctx_t *)(uintptr_t)1U);
+	huge_prof_tctx_set(tsdn, ptr, (prof_tctx_t *)(uintptr_t)1U);
 }
--- a/src/jemalloc.c
+++ b/src/jemalloc.c
--- a/src/mutex.c
+++ b/src/mutex.c
@@ -69,7 +69,7 @@ JEMALLOC_EXPORT int	_pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex,
 #endif

 bool
-malloc_mutex_init(malloc_mutex_t *mutex)
+malloc_mutex_init(malloc_mutex_t *mutex, const char *name, witness_rank_t rank)
 {

 #ifdef _WIN32
@@ -103,31 +103,34 @@ malloc_mutex_init(malloc_mutex_t *mutex)
 	}
 	pthread_mutexattr_destroy(&attr);
 #endif
+	if (config_debug)
+		witness_init(&mutex->witness, name, rank, NULL);
 	return (false);
 }

 void
-malloc_mutex_prefork(malloc_mutex_t *mutex)
+malloc_mutex_prefork(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {

-	malloc_mutex_lock(mutex);
+	malloc_mutex_lock(tsdn, mutex);
 }

 void
-malloc_mutex_postfork_parent(malloc_mutex_t *mutex)
+malloc_mutex_postfork_parent(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {

-	malloc_mutex_unlock(mutex);
+	malloc_mutex_unlock(tsdn, mutex);
 }

 void
-malloc_mutex_postfork_child(malloc_mutex_t *mutex)
+malloc_mutex_postfork_child(tsdn_t *tsdn, malloc_mutex_t *mutex)
 {

 #ifdef JEMALLOC_MUTEX_INIT_CB
-	malloc_mutex_unlock(mutex);
+	malloc_mutex_unlock(tsdn, mutex);
 #else
-	if (malloc_mutex_init(mutex)) {
+	if (malloc_mutex_init(mutex, mutex->witness.name,
+	    mutex->witness.rank)) {
 		malloc_printf("<jemalloc>: Error re-initializing mutex in "
 		    "child\n");
 		if (opt_abort)
@@ -137,7 +140,7 @@ malloc_mutex_postfork_child(malloc_mutex_t *mutex)
 }

 bool
-mutex_boot(void)
+malloc_mutex_boot(void)
 {

 #ifdef JEMALLOC_MUTEX_INIT_CB
--- a/src/nstime.c
+++ b/src/nstime.c
@@ -99,7 +99,7 @@ nstime_divide(const nstime_t *time, const nstime_t *divisor)

 #ifdef JEMALLOC_JET
 #undef nstime_update
-#define	nstime_update JEMALLOC_N(nstime_update_impl)
+#define	nstime_update JEMALLOC_N(n_nstime_update)
 #endif
 bool
 nstime_update(nstime_t *time)
@@ -128,9 +128,11 @@ nstime_update(nstime_t *time)
 		time->ns = ts.tv_sec * BILLION + ts.tv_nsec;
 	}
 #else
-	struct timeval tv;
-	gettimeofday(&tv, NULL);
-	time->ns = tv.tv_sec * BILLION + tv.tv_usec * 1000;
+	{
+		struct timeval tv;
+		gettimeofday(&tv, NULL);
+		time->ns = tv.tv_sec * BILLION + tv.tv_usec * 1000;
+	}
 #endif

 	/* Handle non-monotonic clocks. */
@@ -144,5 +146,5 @@ nstime_update(nstime_t *time)
 #ifdef JEMALLOC_JET
 #undef nstime_update
 #define	nstime_update JEMALLOC_N(nstime_update)
-nstime_update_t *nstime_update = JEMALLOC_N(nstime_update_impl);
+nstime_update_t *nstime_update = JEMALLOC_N(n_nstime_update);
 #endif
--- a/src/pages.c
+++ b/src/pages.c
@@ -1,29 +1,49 @@
 #define	JEMALLOC_PAGES_C_
 #include "jemalloc/internal/jemalloc_internal.h"

+#ifdef JEMALLOC_SYSCTL_VM_OVERCOMMIT
+#include <sys/sysctl.h>
+#endif
+
+/******************************************************************************/
+/* Data. */
+
+#ifndef _WIN32
+#  define PAGES_PROT_COMMIT (PROT_READ | PROT_WRITE)
+#  define PAGES_PROT_DECOMMIT (PROT_NONE)
+static int	mmap_flags;
+#endif
+static bool	os_overcommits;
+
 /******************************************************************************/

 void *
-pages_map(void *addr, size_t size)
+pages_map(void *addr, size_t size, bool *commit)
 {
 	void *ret;

 	assert(size != 0);

+	if (os_overcommits)
+		*commit = true;
+
 #ifdef _WIN32
 	/*
 	 * If VirtualAlloc can't allocate at the given address when one is
 	 * given, it fails and returns NULL.
 	 */
-	ret = VirtualAlloc(addr, size, MEM_COMMIT | MEM_RESERVE,
+	ret = VirtualAlloc(addr, size, MEM_RESERVE | (*commit ? MEM_COMMIT : 0),
 	    PAGE_READWRITE);
 #else
 	/*
 	 * We don't use MAP_FIXED here, because it can cause the *replacement*
 	 * of existing mappings, and we only want to create new mappings.
 	 */
-	ret = mmap(addr, size, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON,
-	    -1, 0);
+	{
+		int prot = *commit ? PAGES_PROT_COMMIT : PAGES_PROT_DECOMMIT;
+
+		ret = mmap(addr, size, prot, mmap_flags, -1, 0);
+	}
 	assert(ret != NULL);

 	if (ret == MAP_FAILED)
@@ -67,7 +87,8 @@ pages_unmap(void *addr, size_t size)
 }

 void *
-pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size)
+pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size,
+    bool *commit)
 {
 	void *ret = (void *)((uintptr_t)addr + leadsize);

@@ -77,7 +98,7 @@ pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size)
 		void *new_addr;

 		pages_unmap(addr, alloc_size);
-		new_addr = pages_map(ret, size);
+		new_addr = pages_map(ret, size, commit);
 		if (new_addr == ret)
 			return (ret);
 		if (new_addr)
@@ -101,17 +122,17 @@ static bool
 pages_commit_impl(void *addr, size_t size, bool commit)
 {

-#ifndef _WIN32
-	/*
-	 * The following decommit/commit implementation is functional, but
-	 * always disabled because it doesn't add value beyong improved
-	 * debugging (at the cost of extra system calls) on systems that
-	 * overcommit.
-	 */
-	if (false) {
-		int prot = commit ? (PROT_READ | PROT_WRITE) : PROT_NONE;
-		void *result = mmap(addr, size, prot, MAP_PRIVATE | MAP_ANON |
-		    MAP_FIXED, -1, 0);
+	if (os_overcommits)
+		return (true);
+
+#ifdef _WIN32
+	return (commit ? (addr != VirtualAlloc(addr, size, MEM_COMMIT,
+	    PAGE_READWRITE)) : (!VirtualFree(addr, size, MEM_DECOMMIT)));
+#else
+	{
+		int prot = commit ? PAGES_PROT_COMMIT : PAGES_PROT_DECOMMIT;
+		void *result = mmap(addr, size, prot, mmap_flags | MAP_FIXED,
+		    -1, 0);
 		if (result == MAP_FAILED)
 			return (true);
 		if (result != addr) {
@@ -125,7 +146,6 @@ pages_commit_impl(void *addr, size_t size, bool commit)
 		return (false);
 	}
 #endif
-	return (true);
 }

 bool
@@ -171,3 +191,63 @@ pages_purge(void *addr, size_t size)
 	return (unzeroed);
 }

+#ifdef JEMALLOC_SYSCTL_VM_OVERCOMMIT
+static bool
+os_overcommits_sysctl(void)
+{
+	int vm_overcommit;
+	size_t sz;
+
+	sz = sizeof(vm_overcommit);
+	if (sysctlbyname("vm.overcommit", &vm_overcommit, &sz, NULL, 0) != 0)
+		return (false); /* Error. */
+
+	return ((vm_overcommit & 0x3) == 0);
+}
+#endif
+
+#ifdef JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY
+static bool
+os_overcommits_proc(void)
+{
+	int fd;
+	char buf[1];
+	ssize_t nread;
+
+	fd = open("/proc/sys/vm/overcommit_memory", O_RDONLY);
+	if (fd == -1)
+		return (false); /* Error. */
+
+	nread = read(fd, &buf, sizeof(buf));
+	if (nread < 1)
+		return (false); /* Error. */
+	/*
+	 * /proc/sys/vm/overcommit_memory meanings:
+	 * 0: Heuristic overcommit.
+	 * 1: Always overcommit.
+	 * 2: Never overcommit.
+	 */
+	return (buf[0] == '0' || buf[0] == '1');
+}
+#endif
+
+void
+pages_boot(void)
+{
+
+#ifndef _WIN32
+	mmap_flags = MAP_PRIVATE | MAP_ANON;
+#endif
+
+#ifdef JEMALLOC_SYSCTL_VM_OVERCOMMIT
+	os_overcommits = os_overcommits_sysctl();
+#elif defined(JEMALLOC_PROC_SYS_VM_OVERCOMMIT_MEMORY)
+	os_overcommits = os_overcommits_proc();
+#  ifdef MAP_NORESERVE
+	if (os_overcommits)
+		mmap_flags |= MAP_NORESERVE;
+#  endif
+#else
+	os_overcommits = false;
+#endif
+}
--- a/src/prof.c
+++ b/src/prof.c
--- a/src/quarantine.c
+++ b/src/quarantine.c
@@ -13,24 +13,22 @@
 /* Function prototypes for non-inline static functions. */

 static quarantine_t	*quarantine_grow(tsd_t *tsd, quarantine_t *quarantine);
-static void	quarantine_drain_one(tsd_t *tsd, quarantine_t *quarantine);
-static void	quarantine_drain(tsd_t *tsd, quarantine_t *quarantine,
+static void	quarantine_drain_one(tsdn_t *tsdn, quarantine_t *quarantine);
+static void	quarantine_drain(tsdn_t *tsdn, quarantine_t *quarantine,
    size_t upper_bound);

 /******************************************************************************/

 static quarantine_t *
-quarantine_init(tsd_t *tsd, size_t lg_maxobjs)
+quarantine_init(tsdn_t *tsdn, size_t lg_maxobjs)
 {
 	quarantine_t *quarantine;
 	size_t size;

-	assert(tsd_nominal(tsd));
-
 	size = offsetof(quarantine_t, objs) + ((ZU(1) << lg_maxobjs) *
 	    sizeof(quarantine_obj_t));
-	quarantine = (quarantine_t *)iallocztm(tsd, size, size2index(size),
-	    false, tcache_get(tsd, true), true, NULL, true);
+	quarantine = (quarantine_t *)iallocztm(tsdn, size, size2index(size),
+	    false, NULL, true, arena_get(TSDN_NULL, 0, true), true);
 	if (quarantine == NULL)
 		return (NULL);
 	quarantine->curbytes = 0;
@@ -49,7 +47,7 @@ quarantine_alloc_hook_work(tsd_t *tsd)
 	if (!tsd_nominal(tsd))
 		return;

-	quarantine = quarantine_init(tsd, LG_MAXOBJS_INIT);
+	quarantine = quarantine_init(tsd_tsdn(tsd), LG_MAXOBJS_INIT);
 	/*
 	 * Check again whether quarantine has been initialized, because
 	 * quarantine_init() may have triggered recursive initialization.
@@ -57,7 +55,7 @@ quarantine_alloc_hook_work(tsd_t *tsd)
 	if (tsd_quarantine_get(tsd) == NULL)
 		tsd_quarantine_set(tsd, quarantine);
 	else
-		idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
+		idalloctm(tsd_tsdn(tsd), quarantine, NULL, true, true);
 }

 static quarantine_t *
@@ -65,9 +63,9 @@ quarantine_grow(tsd_t *tsd, quarantine_t *quarantine)
 {
 	quarantine_t *ret;

-	ret = quarantine_init(tsd, quarantine->lg_maxobjs + 1);
+	ret = quarantine_init(tsd_tsdn(tsd), quarantine->lg_maxobjs + 1);
 	if (ret == NULL) {
-		quarantine_drain_one(tsd, quarantine);
+		quarantine_drain_one(tsd_tsdn(tsd), quarantine);
 		return (quarantine);
 	}

@@ -89,18 +87,18 @@ quarantine_grow(tsd_t *tsd, quarantine_t *quarantine)
 		memcpy(&ret->objs[ncopy_a], quarantine->objs, ncopy_b *
 		    sizeof(quarantine_obj_t));
 	}
-	idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
+	idalloctm(tsd_tsdn(tsd), quarantine, NULL, true, true);

 	tsd_quarantine_set(tsd, ret);
 	return (ret);
 }

 static void
-quarantine_drain_one(tsd_t *tsd, quarantine_t *quarantine)
+quarantine_drain_one(tsdn_t *tsdn, quarantine_t *quarantine)
 {
 	quarantine_obj_t *obj = &quarantine->objs[quarantine->first];
-	assert(obj->usize == isalloc(obj->ptr, config_prof));
-	idalloctm(tsd, obj->ptr, NULL, false, true);
+	assert(obj->usize == isalloc(tsdn, obj->ptr, config_prof));
+	idalloctm(tsdn, obj->ptr, NULL, false, true);
 	quarantine->curbytes -= obj->usize;
 	quarantine->curobjs--;
 	quarantine->first = (quarantine->first + 1) & ((ZU(1) <<
@@ -108,24 +106,24 @@ quarantine_drain_one(tsd_t *tsd, quarantine_t *quarantine)
 }

 static void
-quarantine_drain(tsd_t *tsd, quarantine_t *quarantine, size_t upper_bound)
+quarantine_drain(tsdn_t *tsdn, quarantine_t *quarantine, size_t upper_bound)
 {

 	while (quarantine->curbytes > upper_bound && quarantine->curobjs > 0)
-		quarantine_drain_one(tsd, quarantine);
+		quarantine_drain_one(tsdn, quarantine);
 }

 void
 quarantine(tsd_t *tsd, void *ptr)
 {
 	quarantine_t *quarantine;
-	size_t usize = isalloc(ptr, config_prof);
+	size_t usize = isalloc(tsd_tsdn(tsd), ptr, config_prof);

 	cassert(config_fill);
 	assert(opt_quarantine);

 	if ((quarantine = tsd_quarantine_get(tsd)) == NULL) {
-		idalloctm(tsd, ptr, NULL, false, true);
+		idalloctm(tsd_tsdn(tsd), ptr, NULL, false, true);
 		return;
 	}
 	/*
@@ -135,7 +133,7 @@ quarantine(tsd_t *tsd, void *ptr)
 	if (quarantine->curbytes + usize > opt_quarantine) {
 		size_t upper_bound = (opt_quarantine >= usize) ? opt_quarantine
 		    - usize : 0;
-		quarantine_drain(tsd, quarantine, upper_bound);
+		quarantine_drain(tsd_tsdn(tsd), quarantine, upper_bound);
 	}
 	/* Grow the quarantine ring buffer if it's full. */
 	if (quarantine->curobjs == (ZU(1) << quarantine->lg_maxobjs))
@@ -160,11 +158,11 @@ quarantine(tsd_t *tsd, void *ptr)
 			    && usize <= SMALL_MAXCLASS)
 				arena_quarantine_junk_small(ptr, usize);
 			else
-				memset(ptr, 0x5a, usize);
+				memset(ptr, JEMALLOC_FREE_JUNK, usize);
 		}
 	} else {
 		assert(quarantine->curbytes == 0);
-		idalloctm(tsd, ptr, NULL, false, true);
+		idalloctm(tsd_tsdn(tsd), ptr, NULL, false, true);
 	}
 }

@@ -178,8 +176,8 @@ quarantine_cleanup(tsd_t *tsd)

 	quarantine = tsd_quarantine_get(tsd);
 	if (quarantine != NULL) {
-		quarantine_drain(tsd, quarantine, 0);
-		idalloctm(tsd, quarantine, tcache_get(tsd, false), true, true);
+		quarantine_drain(tsd_tsdn(tsd), quarantine, 0);
+		idalloctm(tsd_tsdn(tsd), quarantine, NULL, true, true);
 		tsd_quarantine_set(tsd, NULL);
 	}
 }
--- a/src/rtree.c
+++ b/src/rtree.c
@@ -15,6 +15,8 @@ rtree_new(rtree_t *rtree, unsigned bits, rtree_node_alloc_t *alloc,
 {
 	unsigned bits_in_leaf, height, i;

+	assert(RTREE_HEIGHT_MAX == ((ZU(1) << (LG_SIZEOF_PTR+3)) /
+	    RTREE_BITS_PER_LEVEL));
 	assert(bits > 0 && bits <= (sizeof(uintptr_t) << 3));

 	bits_in_leaf = (bits % RTREE_BITS_PER_LEVEL) == 0 ? RTREE_BITS_PER_LEVEL
--- a/src/stats.c
+++ b/src/stats.c
@@ -259,7 +259,7 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,
 	unsigned nthreads;
 	const char *dss;
 	ssize_t lg_dirty_mult, decay_time;
-	size_t page, pactive, pdirty, mapped;
+	size_t page, pactive, pdirty, mapped, retained;
 	size_t metadata_mapped, metadata_allocated;
 	uint64_t npurge, nmadvise, purged;
 	size_t small_allocated;
@@ -349,6 +349,9 @@ stats_arena_print(void (*write_cb)(void *, const char *), void *cbopaque,
 	CTL_M2_GET("stats.arenas.0.mapped", i, &mapped, size_t);
 	malloc_cprintf(write_cb, cbopaque,
 	    "mapped:                  %12zu\n", mapped);
+	CTL_M2_GET("stats.arenas.0.retained", i, &retained, size_t);
+	malloc_cprintf(write_cb, cbopaque,
+	    "retained:                %12zu\n", retained);
 	CTL_M2_GET("stats.arenas.0.metadata.mapped", i, &metadata_mapped,
 	    size_t);
 	CTL_M2_GET("stats.arenas.0.metadata.allocated", i, &metadata_allocated,
@@ -468,7 +471,7 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 #define	OPT_WRITE_UNSIGNED(n)						\
 		if (je_mallctl("opt."#n, &uv, &usz, NULL, 0) == 0) {	\
 			malloc_cprintf(write_cb, cbopaque,		\
-			"  opt."#n": %zu\n", sv);			\
+			"  opt."#n": %u\n", uv);			\
 		}
 #define	OPT_WRITE_SIZE_T(n)						\
 		if (je_mallctl("opt."#n, &sv, &ssz, NULL, 0) == 0) {	\
@@ -597,7 +600,7 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,

 	if (config_stats) {
 		size_t *cactive;
-		size_t allocated, active, metadata, resident, mapped;
+		size_t allocated, active, metadata, resident, mapped, retained;

 		CTL_GET("stats.cactive", &cactive, size_t *);
 		CTL_GET("stats.allocated", &allocated, size_t);
@@ -605,10 +608,11 @@ stats_print(void (*write_cb)(void *, const char *), void *cbopaque,
 		CTL_GET("stats.metadata", &metadata, size_t);
 		CTL_GET("stats.resident", &resident, size_t);
 		CTL_GET("stats.mapped", &mapped, size_t);
+		CTL_GET("stats.retained", &retained, size_t);
 		malloc_cprintf(write_cb, cbopaque,
 		    "Allocated: %zu, active: %zu, metadata: %zu,"
-		    " resident: %zu, mapped: %zu\n",
-		    allocated, active, metadata, resident, mapped);
+		    " resident: %zu, mapped: %zu, retained: %zu\n",
+		    allocated, active, metadata, resident, mapped, retained);
 		malloc_cprintf(write_cb, cbopaque,
 		    "Current active ceiling: %zu\n",
 		    atomic_read_z(cactive));
--- a/src/tcache.c
+++ b/src/tcache.c
@@ -23,10 +23,11 @@ static tcaches_t	*tcaches_avail;

 /******************************************************************************/

-size_t	tcache_salloc(const void *ptr)
+size_t
+tcache_salloc(tsdn_t *tsdn, const void *ptr)
 {

-	return (arena_salloc(ptr, false));
+	return (arena_salloc(tsdn, ptr, false));
 }

 void
@@ -70,12 +71,12 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache)
 }

 void *
-tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
+tcache_alloc_small_hard(tsdn_t *tsdn, arena_t *arena, tcache_t *tcache,
    tcache_bin_t *tbin, szind_t binind, bool *tcache_success)
 {
 	void *ret;

-	arena_tcache_fill_small(tsd, arena, tbin, binind, config_prof ?
+	arena_tcache_fill_small(tsdn, arena, tbin, binind, config_prof ?
 	    tcache->prof_accumbytes : 0);
 	if (config_prof)
 		tcache->prof_accumbytes = 0;
@@ -106,12 +107,13 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 		arena_bin_t *bin = &bin_arena->bins[binind];

 		if (config_prof && bin_arena == arena) {
-			if (arena_prof_accum(arena, tcache->prof_accumbytes))
-				prof_idump();
+			if (arena_prof_accum(tsd_tsdn(tsd), arena,
+			    tcache->prof_accumbytes))
+				prof_idump(tsd_tsdn(tsd));
 			tcache->prof_accumbytes = 0;
 		}

-		malloc_mutex_lock(&bin->lock);
+		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
 		if (config_stats && bin_arena == arena) {
 			assert(!merged_stats);
 			merged_stats = true;
@@ -128,9 +130,9 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 				size_t pageind = ((uintptr_t)ptr -
 				    (uintptr_t)chunk) >> LG_PAGE;
 				arena_chunk_map_bits_t *bitselm =
-				    arena_bitselm_get(chunk, pageind);
-				arena_dalloc_bin_junked_locked(bin_arena, chunk,
-				    ptr, bitselm);
+				    arena_bitselm_get_mutable(chunk, pageind);
+				arena_dalloc_bin_junked_locked(tsd_tsdn(tsd),
+				    bin_arena, chunk, ptr, bitselm);
 			} else {
 				/*
 				 * This object was allocated via a different
@@ -142,8 +144,8 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 				ndeferred++;
 			}
 		}
-		malloc_mutex_unlock(&bin->lock);
-		arena_decay_ticks(tsd, bin_arena, nflush - ndeferred);
+		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
+		arena_decay_ticks(tsd_tsdn(tsd), bin_arena, nflush - ndeferred);
 	}
 	if (config_stats && !merged_stats) {
 		/*
@@ -151,11 +153,11 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 		 * arena, so the stats didn't get merged.  Manually do so now.
 		 */
 		arena_bin_t *bin = &arena->bins[binind];
-		malloc_mutex_lock(&bin->lock);
+		malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
 		bin->stats.nflushes++;
 		bin->stats.nrequests += tbin->tstats.nrequests;
 		tbin->tstats.nrequests = 0;
-		malloc_mutex_unlock(&bin->lock);
+		malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
 	}

 	memmove(tbin->avail - rem, tbin->avail - tbin->ncached, rem *
@@ -188,7 +190,7 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,

 		if (config_prof)
 			idump = false;
-		malloc_mutex_lock(&locked_arena->lock);
+		malloc_mutex_lock(tsd_tsdn(tsd), &locked_arena->lock);
 		if ((config_prof || config_stats) && locked_arena == arena) {
 			if (config_prof) {
 				idump = arena_prof_accum_locked(arena,
@@ -211,8 +213,8 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 			chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 			if (extent_node_arena_get(&chunk->node) ==
 			    locked_arena) {
-				arena_dalloc_large_junked_locked(locked_arena,
-				    chunk, ptr);
+				arena_dalloc_large_junked_locked(tsd_tsdn(tsd),
+				    locked_arena, chunk, ptr);
 			} else {
 				/*
 				 * This object was allocated via a different
@@ -224,22 +226,23 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 				ndeferred++;
 			}
 		}
-		malloc_mutex_unlock(&locked_arena->lock);
+		malloc_mutex_unlock(tsd_tsdn(tsd), &locked_arena->lock);
 		if (config_prof && idump)
-			prof_idump();
-		arena_decay_ticks(tsd, locked_arena, nflush - ndeferred);
+			prof_idump(tsd_tsdn(tsd));
+		arena_decay_ticks(tsd_tsdn(tsd), locked_arena, nflush -
+		    ndeferred);
 	}
 	if (config_stats && !merged_stats) {
 		/*
 		 * The flush loop didn't happen to flush to this thread's
 		 * arena, so the stats didn't get merged.  Manually do so now.
 		 */
-		malloc_mutex_lock(&arena->lock);
+		malloc_mutex_lock(tsd_tsdn(tsd), &arena->lock);
 		arena->stats.nrequests_large += tbin->tstats.nrequests;
 		arena->stats.lstats[binind - NBINS].nrequests +=
 		    tbin->tstats.nrequests;
 		tbin->tstats.nrequests = 0;
-		malloc_mutex_unlock(&arena->lock);
+		malloc_mutex_unlock(tsd_tsdn(tsd), &arena->lock);
 	}

 	memmove(tbin->avail - rem, tbin->avail - tbin->ncached, rem *
@@ -249,34 +252,26 @@ tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
 		tbin->low_water = tbin->ncached;
 }

-void
-tcache_arena_associate(tcache_t *tcache, arena_t *arena)
+static void
+tcache_arena_associate(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena)
 {

 	if (config_stats) {
 		/* Link into list of extant tcaches. */
-		malloc_mutex_lock(&arena->lock);
+		malloc_mutex_lock(tsdn, &arena->lock);
 		ql_elm_new(tcache, link);
 		ql_tail_insert(&arena->tcache_ql, tcache, link);
-		malloc_mutex_unlock(&arena->lock);
+		malloc_mutex_unlock(tsdn, &arena->lock);
 	}
 }

-void
-tcache_arena_reassociate(tcache_t *tcache, arena_t *oldarena, arena_t *newarena)
-{
-
-	tcache_arena_dissociate(tcache, oldarena);
-	tcache_arena_associate(tcache, newarena);
-}
-
-void
-tcache_arena_dissociate(tcache_t *tcache, arena_t *arena)
+static void
+tcache_arena_dissociate(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena)
 {

 	if (config_stats) {
 		/* Unlink from list of extant tcaches. */
-		malloc_mutex_lock(&arena->lock);
+		malloc_mutex_lock(tsdn, &arena->lock);
 		if (config_debug) {
 			bool in_ql = false;
 			tcache_t *iter;
@@ -289,11 +284,20 @@ tcache_arena_dissociate(tcache_t *tcache, arena_t *arena)
 			assert(in_ql);
 		}
 		ql_remove(&arena->tcache_ql, tcache, link);
-		tcache_stats_merge(tcache, arena);
-		malloc_mutex_unlock(&arena->lock);
+		tcache_stats_merge(tsdn, tcache, arena);
+		malloc_mutex_unlock(tsdn, &arena->lock);
 	}
 }

+void
+tcache_arena_reassociate(tsdn_t *tsdn, tcache_t *tcache, arena_t *oldarena,
+    arena_t *newarena)
+{
+
+	tcache_arena_dissociate(tsdn, tcache, oldarena);
+	tcache_arena_associate(tsdn, tcache, newarena);
+}
+
 tcache_t *
 tcache_get_hard(tsd_t *tsd)
 {
@@ -307,11 +311,11 @@ tcache_get_hard(tsd_t *tsd)
 	arena = arena_choose(tsd, NULL);
 	if (unlikely(arena == NULL))
 		return (NULL);
-	return (tcache_create(tsd, arena));
+	return (tcache_create(tsd_tsdn(tsd), arena));
 }

 tcache_t *
-tcache_create(tsd_t *tsd, arena_t *arena)
+tcache_create(tsdn_t *tsdn, arena_t *arena)
 {
 	tcache_t *tcache;
 	size_t size, stack_offset;
@@ -325,12 +329,12 @@ tcache_create(tsd_t *tsd, arena_t *arena)
 	/* Avoid false cacheline sharing. */
 	size = sa2u(size, CACHELINE);

-	tcache = ipallocztm(tsd, size, CACHELINE, true, false, true,
-	    arena_get(0, false));
+	tcache = ipallocztm(tsdn, size, CACHELINE, true, NULL, true,
+	    arena_get(TSDN_NULL, 0, true));
 	if (tcache == NULL)
 		return (NULL);

-	tcache_arena_associate(tcache, arena);
+	tcache_arena_associate(tsdn, tcache, arena);

 	ticker_init(&tcache->gc_ticker, TCACHE_GC_INCR);

@@ -357,7 +361,7 @@ tcache_destroy(tsd_t *tsd, tcache_t *tcache)
 	unsigned i;

 	arena = arena_choose(tsd, NULL);
-	tcache_arena_dissociate(tcache, arena);
+	tcache_arena_dissociate(tsd_tsdn(tsd), tcache, arena);

 	for (i = 0; i < NBINS; i++) {
 		tcache_bin_t *tbin = &tcache->tbins[i];
@@ -365,9 +369,9 @@ tcache_destroy(tsd_t *tsd, tcache_t *tcache)

 		if (config_stats && tbin->tstats.nrequests != 0) {
 			arena_bin_t *bin = &arena->bins[i];
-			malloc_mutex_lock(&bin->lock);
+			malloc_mutex_lock(tsd_tsdn(tsd), &bin->lock);
 			bin->stats.nrequests += tbin->tstats.nrequests;
-			malloc_mutex_unlock(&bin->lock);
+			malloc_mutex_unlock(tsd_tsdn(tsd), &bin->lock);
 		}
 	}

@@ -376,19 +380,19 @@ tcache_destroy(tsd_t *tsd, tcache_t *tcache)
 		tcache_bin_flush_large(tsd, tbin, i, 0, tcache);

 		if (config_stats && tbin->tstats.nrequests != 0) {
-			malloc_mutex_lock(&arena->lock);
+			malloc_mutex_lock(tsd_tsdn(tsd), &arena->lock);
 			arena->stats.nrequests_large += tbin->tstats.nrequests;
 			arena->stats.lstats[i - NBINS].nrequests +=
 			    tbin->tstats.nrequests;
-			malloc_mutex_unlock(&arena->lock);
+			malloc_mutex_unlock(tsd_tsdn(tsd), &arena->lock);
 		}
 	}

 	if (config_prof && tcache->prof_accumbytes > 0 &&
-	    arena_prof_accum(arena, tcache->prof_accumbytes))
-		prof_idump();
+	    arena_prof_accum(tsd_tsdn(tsd), arena, tcache->prof_accumbytes))
+		prof_idump(tsd_tsdn(tsd));

-	idalloctm(tsd, tcache, false, true, true);
+	idalloctm(tsd_tsdn(tsd), tcache, NULL, true, true);
 }

 void
@@ -412,21 +416,22 @@ tcache_enabled_cleanup(tsd_t *tsd)
 	/* Do nothing. */
 }

-/* Caller must own arena->lock. */
 void
-tcache_stats_merge(tcache_t *tcache, arena_t *arena)
+tcache_stats_merge(tsdn_t *tsdn, tcache_t *tcache, arena_t *arena)
 {
 	unsigned i;

 	cassert(config_stats);

+	malloc_mutex_assert_owner(tsdn, &arena->lock);
+
 	/* Merge and reset tcache stats. */
 	for (i = 0; i < NBINS; i++) {
 		arena_bin_t *bin = &arena->bins[i];
 		tcache_bin_t *tbin = &tcache->tbins[i];
-		malloc_mutex_lock(&bin->lock);
+		malloc_mutex_lock(tsdn, &bin->lock);
 		bin->stats.nrequests += tbin->tstats.nrequests;
-		malloc_mutex_unlock(&bin->lock);
+		malloc_mutex_unlock(tsdn, &bin->lock);
 		tbin->tstats.nrequests = 0;
 	}

@@ -440,13 +445,14 @@ tcache_stats_merge(tcache_t *tcache, arena_t *arena)
 }

 bool
-tcaches_create(tsd_t *tsd, unsigned *r_ind)
+tcaches_create(tsdn_t *tsdn, unsigned *r_ind)
 {
+	arena_t *arena;
 	tcache_t *tcache;
 	tcaches_t *elm;

 	if (tcaches == NULL) {
-		tcaches = base_alloc(sizeof(tcache_t *) *
+		tcaches = base_alloc(tsdn, sizeof(tcache_t *) *
 		    (MALLOCX_TCACHE_MAX+1));
 		if (tcaches == NULL)
 			return (true);
@@ -454,7 +460,10 @@ tcaches_create(tsd_t *tsd, unsigned *r_ind)

 	if (tcaches_avail == NULL && tcaches_past > MALLOCX_TCACHE_MAX)
 		return (true);
-	tcache = tcache_create(tsd, arena_get(0, false));
+	arena = arena_ichoose(tsdn, NULL);
+	if (unlikely(arena == NULL))
+		return (true);
+	tcache = tcache_create(tsdn, arena);
 	if (tcache == NULL)
 		return (true);

@@ -500,7 +509,7 @@ tcaches_destroy(tsd_t *tsd, unsigned ind)
 }

 bool
-tcache_boot(void)
+tcache_boot(tsdn_t *tsdn)
 {
 	unsigned i;

@@ -518,7 +527,7 @@ tcache_boot(void)
 	nhbins = size2index(tcache_maxclass) + 1;

 	/* Initialize tcache_bin_info. */
-	tcache_bin_info = (tcache_bin_info_t *)base_alloc(nhbins *
+	tcache_bin_info = (tcache_bin_info_t *)base_alloc(tsdn, nhbins *
 	    sizeof(tcache_bin_info_t));
 	if (tcache_bin_info == NULL)
 		return (true);
--- a/src/tsd.c
+++ b/src/tsd.c
@@ -77,7 +77,7 @@ tsd_cleanup(void *arg)
 		/* Do nothing. */
 		break;
 	case tsd_state_nominal:
-#define O(n, t)								\
+#define	O(n, t)								\
 		n##_cleanup(tsd);
 MALLOC_TSD
 #undef O
@@ -106,15 +106,17 @@ MALLOC_TSD
 	}
 }

-bool
+tsd_t *
 malloc_tsd_boot0(void)
 {
+	tsd_t *tsd;

 	ncleanups = 0;
 	if (tsd_boot0())
-		return (true);
-	*tsd_arenas_tdata_bypassp_get(tsd_fetch()) = true;
-	return (false);
+		return (NULL);
+	tsd = tsd_fetch();
+	*tsd_arenas_tdata_bypassp_get(tsd) = true;
+	return (tsd);
 }

 void
@@ -169,10 +171,10 @@ tsd_init_check_recursion(tsd_init_head_t *head, tsd_init_block_t *block)
 	tsd_init_block_t *iter;

 	/* Check whether this thread has already inserted into the list. */
-	malloc_mutex_lock(&head->lock);
+	malloc_mutex_lock(NULL, &head->lock);
 	ql_foreach(iter, &head->blocks, link) {
 		if (iter->thread == self) {
-			malloc_mutex_unlock(&head->lock);
+			malloc_mutex_unlock(NULL, &head->lock);
 			return (iter->data);
 		}
 	}
@@ -180,7 +182,7 @@ tsd_init_check_recursion(tsd_init_head_t *head, tsd_init_block_t *block)
 	ql_elm_new(block, link);
 	block->thread = self;
 	ql_tail_insert(&head->blocks, block, link);
-	malloc_mutex_unlock(&head->lock);
+	malloc_mutex_unlock(NULL, &head->lock);
 	return (NULL);
 }

@@ -188,8 +190,8 @@ void
 tsd_init_finish(tsd_init_head_t *head, tsd_init_block_t *block)
 {

-	malloc_mutex_lock(&head->lock);
+	malloc_mutex_lock(NULL, &head->lock);
 	ql_remove(&head->blocks, block, link);
-	malloc_mutex_unlock(&head->lock);
+	malloc_mutex_unlock(NULL, &head->lock);
 }
 #endif
--- a/src/util.c
+++ b/src/util.c
@@ -14,6 +14,7 @@
 		malloc_write("<jemalloc>: Unreachable code reached\n");	\
 		abort();						\
 	}								\
+	unreachable();							\
 } while (0)

 #define	not_implemented() do {						\
@@ -314,10 +315,9 @@ x2s(uintmax_t x, bool alt_form, bool uppercase, char *s, size_t *slen_p)
 	return (s);
 }

-int
+size_t
 malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 {
-	int ret;
 	size_t i;
 	const char *f;

@@ -408,6 +408,8 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 			int prec = -1;
 			int width = -1;
 			unsigned char len = '?';
+			char *s;
+			size_t slen;

 			f++;
 			/* Flags. */
@@ -498,8 +500,6 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 			}
 			/* Conversion specifier. */
 			switch (*f) {
-				char *s;
-				size_t slen;
 			case '%':
 				/* %% */
 				APPEND_C(*f);
@@ -585,21 +585,19 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 		str[i] = '\0';
 	else
 		str[size - 1] = '\0';
-	assert(i < INT_MAX);
-	ret = (int)i;

 #undef APPEND_C
 #undef APPEND_S
 #undef APPEND_PADDED_S
 #undef GET_ARG_NUMERIC
-	return (ret);
+	return (i);
 }

 JEMALLOC_FORMAT_PRINTF(3, 4)
-int
+size_t
 malloc_snprintf(char *str, size_t size, const char *format, ...)
 {
-	int ret;
+	size_t ret;
 	va_list ap;

 	va_start(ap, format);
--- a/src/witness.c
+++ b/src/witness.c
@@ -0,0 +1,136 @@
+#define	JEMALLOC_WITNESS_C_
+#include "jemalloc/internal/jemalloc_internal.h"
+
+void
+witness_init(witness_t *witness, const char *name, witness_rank_t rank,
+    witness_comp_t *comp)
+{
+
+	witness->name = name;
+	witness->rank = rank;
+	witness->comp = comp;
+}
+
+#ifdef JEMALLOC_JET
+#undef witness_lock_error
+#define	witness_lock_error JEMALLOC_N(n_witness_lock_error)
+#endif
+void
+witness_lock_error(const witness_list_t *witnesses, const witness_t *witness)
+{
+	witness_t *w;
+
+	malloc_printf("<jemalloc>: Lock rank order reversal:");
+	ql_foreach(w, witnesses, link) {
+		malloc_printf(" %s(%u)", w->name, w->rank);
+	}
+	malloc_printf(" %s(%u)\n", witness->name, witness->rank);
+	abort();
+}
+#ifdef JEMALLOC_JET
+#undef witness_lock_error
+#define	witness_lock_error JEMALLOC_N(witness_lock_error)
+witness_lock_error_t *witness_lock_error = JEMALLOC_N(n_witness_lock_error);
+#endif
+
+#ifdef JEMALLOC_JET
+#undef witness_owner_error
+#define	witness_owner_error JEMALLOC_N(n_witness_owner_error)
+#endif
+void
+witness_owner_error(const witness_t *witness)
+{
+
+	malloc_printf("<jemalloc>: Should own %s(%u)\n", witness->name,
+	    witness->rank);
+	abort();
+}
+#ifdef JEMALLOC_JET
+#undef witness_owner_error
+#define	witness_owner_error JEMALLOC_N(witness_owner_error)
+witness_owner_error_t *witness_owner_error = JEMALLOC_N(n_witness_owner_error);
+#endif
+
+#ifdef JEMALLOC_JET
+#undef witness_not_owner_error
+#define	witness_not_owner_error JEMALLOC_N(n_witness_not_owner_error)
+#endif
+void
+witness_not_owner_error(const witness_t *witness)
+{
+
+	malloc_printf("<jemalloc>: Should not own %s(%u)\n", witness->name,
+	    witness->rank);
+	abort();
+}
+#ifdef JEMALLOC_JET
+#undef witness_not_owner_error
+#define	witness_not_owner_error JEMALLOC_N(witness_not_owner_error)
+witness_not_owner_error_t *witness_not_owner_error =
+    JEMALLOC_N(n_witness_not_owner_error);
+#endif
+
+#ifdef JEMALLOC_JET
+#undef witness_lockless_error
+#define	witness_lockless_error JEMALLOC_N(n_witness_lockless_error)
+#endif
+void
+witness_lockless_error(const witness_list_t *witnesses)
+{
+	witness_t *w;
+
+	malloc_printf("<jemalloc>: Should not own any locks:");
+	ql_foreach(w, witnesses, link) {
+		malloc_printf(" %s(%u)", w->name, w->rank);
+	}
+	malloc_printf("\n");
+	abort();
+}
+#ifdef JEMALLOC_JET
+#undef witness_lockless_error
+#define	witness_lockless_error JEMALLOC_N(witness_lockless_error)
+witness_lockless_error_t *witness_lockless_error =
+    JEMALLOC_N(n_witness_lockless_error);
+#endif
+
+void
+witnesses_cleanup(tsd_t *tsd)
+{
+
+	witness_assert_lockless(tsd_tsdn(tsd));
+
+	/* Do nothing. */
+}
+
+void
+witness_fork_cleanup(tsd_t *tsd)
+{
+
+	/* Do nothing. */
+}
+
+void
+witness_prefork(tsd_t *tsd)
+{
+
+	tsd_witness_fork_set(tsd, true);
+}
+
+void
+witness_postfork_parent(tsd_t *tsd)
+{
+
+	tsd_witness_fork_set(tsd, false);
+}
+
+void
+witness_postfork_child(tsd_t *tsd)
+{
+#ifndef JEMALLOC_MUTEX_INIT_CB
+	witness_list_t *witnesses;
+
+	witnesses = tsd_witnessesp_get(tsd);
+	ql_new(witnesses);
+#endif
+	tsd_witness_fork_set(tsd, false);
+}
--- a/src/zone.c
+++ b/src/zone.c
@@ -56,7 +56,7 @@ zone_size(malloc_zone_t *zone, void *ptr)
 	 * not work in practice, we must check all pointers to assure that they
 	 * reside within a mapped chunk before determining size.
 	 */
-	return (ivsalloc(ptr, config_prof));
+	return (ivsalloc(tsdn_fetch(), ptr, config_prof));
 }

 static void *
@@ -87,7 +87,7 @@ static void
 zone_free(malloc_zone_t *zone, void *ptr)
 {

-	if (ivsalloc(ptr, config_prof) != 0) {
+	if (ivsalloc(tsdn_fetch(), ptr, config_prof) != 0) {
 		je_free(ptr);
 		return;
 	}
@@ -99,7 +99,7 @@ static void *
 zone_realloc(malloc_zone_t *zone, void *ptr, size_t size)
 {

-	if (ivsalloc(ptr, config_prof) != 0)
+	if (ivsalloc(tsdn_fetch(), ptr, config_prof) != 0)
 		return (je_realloc(ptr, size));

 	return (realloc(ptr, size));
@@ -123,7 +123,7 @@ zone_free_definite_size(malloc_zone_t *zone, void *ptr, size_t size)
 {
 	size_t alloc_size;

-	alloc_size = ivsalloc(ptr, config_prof);
+	alloc_size = ivsalloc(tsdn_fetch(), ptr, config_prof);
 	if (alloc_size != 0) {
 		assert(alloc_size == size);
 		je_free(ptr);
--- a/test/include/test/jemalloc_test.h.in
+++ b/test/include/test/jemalloc_test.h.in
@@ -19,39 +19,6 @@
 #  include <pthread.h>
 #endif

-/******************************************************************************/
-/*
- * Define always-enabled assertion macros, so that test assertions execute even
- * if assertions are disabled in the library code.  These definitions must
- * exist prior to including "jemalloc/internal/util.h".
- */
-#define	assert(e) do {							\
-	if (!(e)) {							\
-		malloc_printf(						\
-		    "<jemalloc>: %s:%d: Failed assertion: \"%s\"\n",	\
-		    __FILE__, __LINE__, #e);				\
-		abort();						\
-	}								\
-} while (0)
-
-#define	not_reached() do {						\
-	malloc_printf(							\
-	    "<jemalloc>: %s:%d: Unreachable code reached\n",		\
-	    __FILE__, __LINE__);					\
-	abort();							\
-} while (0)
-
-#define	not_implemented() do {						\
-	malloc_printf("<jemalloc>: %s:%d: Not implemented\n",		\
-	    __FILE__, __LINE__);					\
-	abort();							\
-} while (0)
-
-#define	assert_not_implemented(e) do {					\
-	if (!(e))							\
-		not_implemented();					\
-} while (0)
-
 #include "test/jemalloc_test_defs.h"

 #ifdef JEMALLOC_OSSPIN
@@ -86,6 +53,14 @@
 #  include "jemalloc/internal/jemalloc_internal_defs.h"
 #  include "jemalloc/internal/jemalloc_internal_macros.h"

+static const bool config_debug =
+#ifdef JEMALLOC_DEBUG
+    true
+#else
+    false
+#endif
+    ;
+
 #  define JEMALLOC_N(n) @private_namespace@##n
 #  include "jemalloc/internal/private_namespace.h"

@@ -149,3 +124,40 @@
 #include "test/thd.h"
 #define	MEXP 19937
 #include "test/SFMT.h"
+
+/******************************************************************************/
+/*
+ * Define always-enabled assertion macros, so that test assertions execute even
+ * if assertions are disabled in the library code.
+ */
+#undef assert
+#undef not_reached
+#undef not_implemented
+#undef assert_not_implemented
+
+#define	assert(e) do {							\
+	if (!(e)) {							\
+		malloc_printf(						\
+		    "<jemalloc>: %s:%d: Failed assertion: \"%s\"\n",	\
+		    __FILE__, __LINE__, #e);				\
+		abort();						\
+	}								\
+} while (0)
+
+#define	not_reached() do {						\
+	malloc_printf(							\
+	    "<jemalloc>: %s:%d: Unreachable code reached\n",		\
+	    __FILE__, __LINE__);					\
+	abort();							\
+} while (0)
+
+#define	not_implemented() do {						\
+	malloc_printf("<jemalloc>: %s:%d: Not implemented\n",		\
+	    __FILE__, __LINE__);					\
+	abort();							\
+} while (0)
+
+#define	assert_not_implemented(e) do {					\
+	if (!(e))							\
+		not_implemented();					\
+} while (0)
--- a/test/include/test/test.h
+++ b/test/include/test/test.h
@@ -311,6 +311,9 @@ label_test_end:								\
 #define	test(...)							\
 	p_test(__VA_ARGS__, NULL)

+#define	test_no_malloc_init(...)					\
+	p_test_no_malloc_init(__VA_ARGS__, NULL)
+
 #define	test_skip_if(e) do {						\
 	if (e) {							\
 		test_skip("%s:%s:%d: Test skipped: (%s)",		\
@@ -324,6 +327,7 @@ void	test_fail(const char *format, ...) JEMALLOC_FORMAT_PRINTF(1, 2);

 /* For private use by macros. */
 test_status_t	p_test(test_t *t, ...);
+test_status_t	p_test_no_malloc_init(test_t *t, ...);
 void	p_test_init(const char *name);
 void	p_test_fini(void);
 void	p_test_fail(const char *prefix, const char *message);
--- a/test/integration/chunk.c
+++ b/test/integration/chunk.c
@@ -121,6 +121,10 @@ TEST_BEGIN(test_chunk)
 {
 	void *p;
 	size_t old_size, new_size, large0, large1, huge0, huge1, huge2, sz;
+	unsigned arena_ind;
+	int flags;
+	size_t hooks_mib[3], purge_mib[3];
+	size_t hooks_miblen, purge_miblen;
 	chunk_hooks_t new_hooks = {
 		chunk_alloc,
 		chunk_dalloc,
@@ -132,10 +136,19 @@ TEST_BEGIN(test_chunk)
 	};
 	bool xallocx_success_a, xallocx_success_b, xallocx_success_c;

+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.extend", &arena_ind, &sz, NULL, 0), 0,
+	    "Unexpected mallctl() failure");
+	flags = MALLOCX_ARENA(arena_ind) | MALLOCX_TCACHE_NONE;
+
 	/* Install custom chunk hooks. */
+	hooks_miblen = sizeof(hooks_mib)/sizeof(size_t);
+	assert_d_eq(mallctlnametomib("arena.0.chunk_hooks", hooks_mib,
+	    &hooks_miblen), 0, "Unexpected mallctlnametomib() failure");
+	hooks_mib[1] = (size_t)arena_ind;
 	old_size = sizeof(chunk_hooks_t);
 	new_size = sizeof(chunk_hooks_t);
-	assert_d_eq(mallctl("arena.0.chunk_hooks", &old_hooks, &old_size,
+	assert_d_eq(mallctlbymib(hooks_mib, hooks_miblen, &old_hooks, &old_size,
 	    &new_hooks, new_size), 0, "Unexpected chunk_hooks error");
 	orig_hooks = old_hooks;
 	assert_ptr_ne(old_hooks.alloc, chunk_alloc, "Unexpected alloc error");
@@ -165,45 +178,49 @@ TEST_BEGIN(test_chunk)
 	    "Unexpected arenas.hchunk.2.size failure");

 	/* Test dalloc/decommit/purge cascade. */
+	purge_miblen = sizeof(purge_mib)/sizeof(size_t);
+	assert_d_eq(mallctlnametomib("arena.0.purge", purge_mib, &purge_miblen),
+	    0, "Unexpected mallctlnametomib() failure");
+	purge_mib[1] = (size_t)arena_ind;
 	do_dalloc = false;
 	do_decommit = false;
-	p = mallocx(huge0 * 2, 0);
+	p = mallocx(huge0 * 2, flags);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
 	did_dalloc = false;
 	did_decommit = false;
 	did_purge = false;
 	did_split = false;
-	xallocx_success_a = (xallocx(p, huge0, 0, 0) == huge0);
-	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
-	    "Unexpected arena.0.purge error");
+	xallocx_success_a = (xallocx(p, huge0, 0, flags) == huge0);
+	assert_d_eq(mallctlbymib(purge_mib, purge_miblen, NULL, NULL, NULL, 0),
+	    0, "Unexpected arena.%u.purge error", arena_ind);
 	if (xallocx_success_a) {
 		assert_true(did_dalloc, "Expected dalloc");
 		assert_false(did_decommit, "Unexpected decommit");
 		assert_true(did_purge, "Expected purge");
 	}
 	assert_true(did_split, "Expected split");
-	dallocx(p, 0);
+	dallocx(p, flags);
 	do_dalloc = true;

 	/* Test decommit/commit and observe split/merge. */
 	do_dalloc = false;
 	do_decommit = true;
-	p = mallocx(huge0 * 2, 0);
+	p = mallocx(huge0 * 2, flags);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
 	did_decommit = false;
 	did_commit = false;
 	did_split = false;
 	did_merge = false;
-	xallocx_success_b = (xallocx(p, huge0, 0, 0) == huge0);
-	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
-	    "Unexpected arena.0.purge error");
+	xallocx_success_b = (xallocx(p, huge0, 0, flags) == huge0);
+	assert_d_eq(mallctlbymib(purge_mib, purge_miblen, NULL, NULL, NULL, 0),
+	    0, "Unexpected arena.%u.purge error", arena_ind);
 	if (xallocx_success_b)
 		assert_true(did_split, "Expected split");
-	xallocx_success_c = (xallocx(p, huge0 * 2, 0, 0) == huge0 * 2);
+	xallocx_success_c = (xallocx(p, huge0 * 2, 0, flags) == huge0 * 2);
 	assert_b_eq(did_decommit, did_commit, "Expected decommit/commit match");
 	if (xallocx_success_b && xallocx_success_c)
 		assert_true(did_merge, "Expected merge");
-	dallocx(p, 0);
+	dallocx(p, flags);
 	do_dalloc = true;
 	do_decommit = false;

@@ -214,42 +231,42 @@ TEST_BEGIN(test_chunk)
 		 * successful xallocx() from size=huge2 to size=huge1 is
 		 * guaranteed to leave trailing purgeable memory.
 		 */
-		p = mallocx(huge2, 0);
+		p = mallocx(huge2, flags);
 		assert_ptr_not_null(p, "Unexpected mallocx() error");
 		did_purge = false;
-		assert_zu_eq(xallocx(p, huge1, 0, 0), huge1,
+		assert_zu_eq(xallocx(p, huge1, 0, flags), huge1,
 		    "Unexpected xallocx() failure");
 		assert_true(did_purge, "Expected purge");
-		dallocx(p, 0);
+		dallocx(p, flags);
 	}

 	/* Test decommit for large allocations. */
 	do_decommit = true;
-	p = mallocx(large1, 0);
+	p = mallocx(large1, flags);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
-	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
-	    "Unexpected arena.0.purge error");
+	assert_d_eq(mallctlbymib(purge_mib, purge_miblen, NULL, NULL, NULL, 0),
+	    0, "Unexpected arena.%u.purge error", arena_ind);
 	did_decommit = false;
-	assert_zu_eq(xallocx(p, large0, 0, 0), large0,
+	assert_zu_eq(xallocx(p, large0, 0, flags), large0,
 	    "Unexpected xallocx() failure");
-	assert_d_eq(mallctl("arena.0.purge", NULL, NULL, NULL, 0), 0,
-	    "Unexpected arena.0.purge error");
+	assert_d_eq(mallctlbymib(purge_mib, purge_miblen, NULL, NULL, NULL, 0),
+	    0, "Unexpected arena.%u.purge error", arena_ind);
 	did_commit = false;
-	assert_zu_eq(xallocx(p, large1, 0, 0), large1,
+	assert_zu_eq(xallocx(p, large1, 0, flags), large1,
 	    "Unexpected xallocx() failure");
 	assert_b_eq(did_decommit, did_commit, "Expected decommit/commit match");
-	dallocx(p, 0);
+	dallocx(p, flags);
 	do_decommit = false;

 	/* Make sure non-huge allocation succeeds. */
-	p = mallocx(42, 0);
+	p = mallocx(42, flags);
 	assert_ptr_not_null(p, "Unexpected mallocx() error");
-	dallocx(p, 0);
+	dallocx(p, flags);

 	/* Restore chunk hooks. */
-	assert_d_eq(mallctl("arena.0.chunk_hooks", NULL, NULL, &old_hooks,
-	    new_size), 0, "Unexpected chunk_hooks error");
-	assert_d_eq(mallctl("arena.0.chunk_hooks", &old_hooks, &old_size,
+	assert_d_eq(mallctlbymib(hooks_mib, hooks_miblen, NULL, NULL,
+	    &old_hooks, new_size), 0, "Unexpected chunk_hooks error");
+	assert_d_eq(mallctlbymib(hooks_mib, hooks_miblen, &old_hooks, &old_size,
 	    NULL, 0), 0, "Unexpected chunk_hooks error");
 	assert_ptr_eq(old_hooks.alloc, orig_hooks.alloc,
 	    "Unexpected alloc error");
--- a/test/integration/mallocx.c
+++ b/test/integration/mallocx.c
@@ -1,5 +1,9 @@
 #include "test/jemalloc_test.h"

+#ifdef JEMALLOC_FILL
+const char *malloc_conf = "junk:false";
+#endif
+
 static unsigned
 get_nsizes_impl(const char *cmd)
 {
@@ -69,35 +73,41 @@ TEST_END

 TEST_BEGIN(test_oom)
 {
-	size_t hugemax, size, alignment;
-
-	hugemax = get_huge_size(get_nhuge()-1);
+	size_t hugemax;
+	bool oom;
+	void *ptrs[3];
+	unsigned i;

 	/*
-	 * It should be impossible to allocate two objects that each consume
-	 * more than half the virtual address space.
+	 * It should be impossible to allocate three objects that each consume
+	 * nearly half the virtual address space.
 	 */
-	{
-		void *p;
-
-		p = mallocx(hugemax, 0);
-		if (p != NULL) {
-			assert_ptr_null(mallocx(hugemax, 0),
-			    "Expected OOM for mallocx(size=%#zx, 0)", hugemax);
-			dallocx(p, 0);
-		}
+	hugemax = get_huge_size(get_nhuge()-1);
+	oom = false;
+	for (i = 0; i < sizeof(ptrs) / sizeof(void *); i++) {
+		ptrs[i] = mallocx(hugemax, 0);
+		if (ptrs[i] == NULL)
+			oom = true;
+	}
+	assert_true(oom,
+	    "Expected OOM during series of calls to mallocx(size=%zu, 0)",
+	    hugemax);
+	for (i = 0; i < sizeof(ptrs) / sizeof(void *); i++) {
+		if (ptrs[i] != NULL)
+			dallocx(ptrs[i], 0);
 	}

 #if LG_SIZEOF_PTR == 3
-	size      = ZU(0x8000000000000000);
-	alignment = ZU(0x8000000000000000);
+	assert_ptr_null(mallocx(0x8000000000000000ULL,
+	    MALLOCX_ALIGN(0x8000000000000000ULL)),
+	    "Expected OOM for mallocx()");
+	assert_ptr_null(mallocx(0x8000000000000000ULL,
+	    MALLOCX_ALIGN(0x80000000)),
+	    "Expected OOM for mallocx()");
 #else
-	size      = ZU(0x80000000);
-	alignment = ZU(0x80000000);
+	assert_ptr_null(mallocx(0x80000000UL, MALLOCX_ALIGN(0x80000000UL)),
+	    "Expected OOM for mallocx()");
 #endif
-	assert_ptr_null(mallocx(size, MALLOCX_ALIGN(alignment)),
-	    "Expected OOM for mallocx(size=%#zx, MALLOCX_ALIGN(%#zx)", size,
-	    alignment);
 }
 TEST_END

--- a/test/integration/xallocx.c
+++ b/test/integration/xallocx.c
@@ -1,5 +1,9 @@
 #include "test/jemalloc_test.h"

+#ifdef JEMALLOC_FILL
+const char *malloc_conf = "junk:false";
+#endif
+
 /*
 * Use a separate arena for xallocx() extension/contraction tests so that
 * internal allocation e.g. by heap profiling can't interpose allocations where
--- a/test/src/test.c
+++ b/test/src/test.c
@@ -60,32 +60,30 @@ p_test_fini(void)
 	malloc_printf("%s: %s\n", test_name, test_status_string(test_status));
 }

-test_status_t
-p_test(test_t *t, ...)
+static test_status_t
+p_test_impl(bool do_malloc_init, test_t *t, va_list ap)
 {
 	test_status_t ret;
-	va_list ap;

-	/*
-	 * Make sure initialization occurs prior to running tests.  Tests are
-	 * special because they may use internal facilities prior to triggering
-	 * initialization as a side effect of calling into the public API.  This
-	 * is a final safety that works even if jemalloc_constructor() doesn't
-	 * run, as for MSVC builds.
-	 */
-	if (nallocx(1, 0) == 0) {
-		malloc_printf("Initialization error");
-		return (test_status_fail);
+	if (do_malloc_init) {
+		/*
+		 * Make sure initialization occurs prior to running tests.
+		 * Tests are special because they may use internal facilities
+		 * prior to triggering initialization as a side effect of
+		 * calling into the public API.
+		 */
+		if (nallocx(1, 0) == 0) {
+			malloc_printf("Initialization error");
+			return (test_status_fail);
+		}
 	}

 	ret = test_status_pass;
-	va_start(ap, t);
 	for (; t != NULL; t = va_arg(ap, test_t *)) {
 		t();
 		if (test_status > ret)
 			ret = test_status;
 	}
-	va_end(ap);

 	malloc_printf("--- %s: %u/%u, %s: %u/%u, %s: %u/%u ---\n",
 	    test_status_string(test_status_pass),
@@ -98,6 +96,34 @@ p_test(test_t *t, ...)
 	return (ret);
 }

+test_status_t
+p_test(test_t *t, ...)
+{
+	test_status_t ret;
+	va_list ap;
+
+	ret = test_status_pass;
+	va_start(ap, t);
+	ret = p_test_impl(true, t, ap);
+	va_end(ap);
+
+	return (ret);
+}
+
+test_status_t
+p_test_no_malloc_init(test_t *t, ...)
+{
+	test_status_t ret;
+	va_list ap;
+
+	ret = test_status_pass;
+	va_start(ap, t);
+	ret = p_test_impl(false, t, ap);
+	va_end(ap);
+
+	return (ret);
+}
+
 void
 p_test_fail(const char *prefix, const char *message)
 {
--- a/test/src/timer.c
+++ b/test/src/timer.c
@@ -32,9 +32,8 @@ timer_ratio(timedelta_t *a, timedelta_t *b, char *buf, size_t buflen)
 	uint64_t t0 = timer_usec(a);
 	uint64_t t1 = timer_usec(b);
 	uint64_t mult;
-	unsigned i = 0;
-	unsigned j;
-	int n;
+	size_t i = 0;
+	size_t j, n;

 	/* Whole. */
 	n = malloc_snprintf(&buf[i], buflen-i, "%"FMTu64, t0 / t1);
--- a/test/stress/microbench.c
+++ b/test/stress/microbench.c
@@ -1,7 +1,8 @@
 #include "test/jemalloc_test.h"

 JEMALLOC_INLINE_C void
-time_func(timedelta_t *timer, uint64_t nwarmup, uint64_t niter, void (*func)(void))
+time_func(timedelta_t *timer, uint64_t nwarmup, uint64_t niter,
+    void (*func)(void))
 {
 	uint64_t i;

--- a/test/unit/a0.c
+++ b/test/unit/a0.c
@@ -0,0 +1,19 @@
+#include "test/jemalloc_test.h"
+
+TEST_BEGIN(test_a0)
+{
+	void *p;
+
+	p = a0malloc(1);
+	assert_ptr_not_null(p, "Unexpected a0malloc() error");
+	a0dalloc(p);
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test_no_malloc_init(
+	    test_a0));
+}
--- a/test/unit/arena_reset.c
+++ b/test/unit/arena_reset.c
@@ -0,0 +1,159 @@
+#include "test/jemalloc_test.h"
+
+#ifdef JEMALLOC_PROF
+const char *malloc_conf = "prof:true,lg_prof_sample:0";
+#endif
+
+static unsigned
+get_nsizes_impl(const char *cmd)
+{
+	unsigned ret;
+	size_t z;
+
+	z = sizeof(unsigned);
+	assert_d_eq(mallctl(cmd, &ret, &z, NULL, 0), 0,
+	    "Unexpected mallctl(\"%s\", ...) failure", cmd);
+
+	return (ret);
+}
+
+static unsigned
+get_nsmall(void)
+{
+
+	return (get_nsizes_impl("arenas.nbins"));
+}
+
+static unsigned
+get_nlarge(void)
+{
+
+	return (get_nsizes_impl("arenas.nlruns"));
+}
+
+static unsigned
+get_nhuge(void)
+{
+
+	return (get_nsizes_impl("arenas.nhchunks"));
+}
+
+static size_t
+get_size_impl(const char *cmd, size_t ind)
+{
+	size_t ret;
+	size_t z;
+	size_t mib[4];
+	size_t miblen = 4;
+
+	z = sizeof(size_t);
+	assert_d_eq(mallctlnametomib(cmd, mib, &miblen),
+	    0, "Unexpected mallctlnametomib(\"%s\", ...) failure", cmd);
+	mib[2] = ind;
+	z = sizeof(size_t);
+	assert_d_eq(mallctlbymib(mib, miblen, &ret, &z, NULL, 0),
+	    0, "Unexpected mallctlbymib([\"%s\", %zu], ...) failure", cmd, ind);
+
+	return (ret);
+}
+
+static size_t
+get_small_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.bin.0.size", ind));
+}
+
+static size_t
+get_large_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.lrun.0.size", ind));
+}
+
+static size_t
+get_huge_size(size_t ind)
+{
+
+	return (get_size_impl("arenas.hchunk.0.size", ind));
+}
+
+TEST_BEGIN(test_arena_reset)
+{
+#define	NHUGE	4
+	unsigned arena_ind, nsmall, nlarge, nhuge, nptrs, i;
+	size_t sz, miblen;
+	void **ptrs;
+	int flags;
+	size_t mib[3];
+	tsdn_t *tsdn;
+
+	test_skip_if((config_valgrind && unlikely(in_valgrind)) || (config_fill
+	    && unlikely(opt_quarantine)));
+
+	sz = sizeof(unsigned);
+	assert_d_eq(mallctl("arenas.extend", &arena_ind, &sz, NULL, 0), 0,
+	    "Unexpected mallctl() failure");
+
+	flags = MALLOCX_ARENA(arena_ind) | MALLOCX_TCACHE_NONE;
+
+	nsmall = get_nsmall();
+	nlarge = get_nlarge();
+	nhuge = get_nhuge() > NHUGE ? NHUGE : get_nhuge();
+	nptrs = nsmall + nlarge + nhuge;
+	ptrs = (void **)malloc(nptrs * sizeof(void *));
+	assert_ptr_not_null(ptrs, "Unexpected malloc() failure");
+
+	/* Allocate objects with a wide range of sizes. */
+	for (i = 0; i < nsmall; i++) {
+		sz = get_small_size(i);
+		ptrs[i] = mallocx(sz, flags);
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, %#x) failure", sz, flags);
+	}
+	for (i = 0; i < nlarge; i++) {
+		sz = get_large_size(i);
+		ptrs[nsmall + i] = mallocx(sz, flags);
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, %#x) failure", sz, flags);
+	}
+	for (i = 0; i < nhuge; i++) {
+		sz = get_huge_size(i);
+		ptrs[nsmall + nlarge + i] = mallocx(sz, flags);
+		assert_ptr_not_null(ptrs[i],
+		    "Unexpected mallocx(%zu, %#x) failure", sz, flags);
+	}
+
+	tsdn = tsdn_fetch();
+
+	/* Verify allocations. */
+	for (i = 0; i < nptrs; i++) {
+		assert_zu_gt(ivsalloc(tsdn, ptrs[i], false), 0,
+		    "Allocation should have queryable size");
+	}
+
+	/* Reset. */
+	miblen = sizeof(mib)/sizeof(size_t);
+	assert_d_eq(mallctlnametomib("arena.0.reset", mib, &miblen), 0,
+	    "Unexpected mallctlnametomib() failure");
+	mib[1] = (size_t)arena_ind;
+	assert_d_eq(mallctlbymib(mib, miblen, NULL, NULL, NULL, 0), 0,
+	    "Unexpected mallctlbymib() failure");
+
+	/* Verify allocations no longer exist. */
+	for (i = 0; i < nptrs; i++) {
+		assert_zu_eq(ivsalloc(tsdn, ptrs[i], false), 0,
+		    "Allocation should no longer exist");
+	}
+
+	free(ptrs);
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_arena_reset));
+}
--- a/test/unit/bitmap.c
+++ b/test/unit/bitmap.c
@@ -101,7 +101,7 @@ TEST_BEGIN(test_bitmap_sfu)
 		bitmap_info_t binfo;
 		bitmap_info_init(&binfo, i);
 		{
-			ssize_t j;
+			size_t j;
 			bitmap_t *bitmap = (bitmap_t *)malloc(
 			    bitmap_size(&binfo));
 			bitmap_init(bitmap, &binfo);
@@ -119,7 +119,7 @@ TEST_BEGIN(test_bitmap_sfu)
 			 * Iteratively unset bits starting at the end, and
 			 * verify that bitmap_sfu() reaches the unset bits.
 			 */
-			for (j = i - 1; j >= 0; j--) {
+			for (j = i - 1; j < i; j--) { /* (i..0] */
 				bitmap_unset(bitmap, &binfo, j);
 				assert_zd_eq(bitmap_sfu(bitmap, &binfo), j,
 				    "First unset bit should the bit previously "
--- a/test/unit/ckh.c
+++ b/test/unit/ckh.c
@@ -2,24 +2,24 @@

 TEST_BEGIN(test_new_delete)
 {
-	tsd_t *tsd;
+	tsdn_t *tsdn;
 	ckh_t ckh;

-	tsd = tsd_fetch();
+	tsdn = tsdn_fetch();

-	assert_false(ckh_new(tsd, &ckh, 2, ckh_string_hash, ckh_string_keycomp),
-	    "Unexpected ckh_new() error");
-	ckh_delete(tsd, &ckh);
+	assert_false(ckh_new(tsdn, &ckh, 2, ckh_string_hash,
+	    ckh_string_keycomp), "Unexpected ckh_new() error");
+	ckh_delete(tsdn, &ckh);

-	assert_false(ckh_new(tsd, &ckh, 3, ckh_pointer_hash,
+	assert_false(ckh_new(tsdn, &ckh, 3, ckh_pointer_hash,
 	    ckh_pointer_keycomp), "Unexpected ckh_new() error");
-	ckh_delete(tsd, &ckh);
+	ckh_delete(tsdn, &ckh);
 }
 TEST_END

 TEST_BEGIN(test_count_insert_search_remove)
 {
-	tsd_t *tsd;
+	tsdn_t *tsdn;
 	ckh_t ckh;
 	const char *strs[] = {
 	    "a string",
@@ -30,17 +30,17 @@ TEST_BEGIN(test_count_insert_search_remove)
 	const char *missing = "A string not in the hash table.";
 	size_t i;

-	tsd = tsd_fetch();
+	tsdn = tsdn_fetch();

-	assert_false(ckh_new(tsd, &ckh, 2, ckh_string_hash, ckh_string_keycomp),
-	    "Unexpected ckh_new() error");
+	assert_false(ckh_new(tsdn, &ckh, 2, ckh_string_hash,
+	    ckh_string_keycomp), "Unexpected ckh_new() error");
 	assert_zu_eq(ckh_count(&ckh), 0,
 	    "ckh_count() should return %zu, but it returned %zu", ZU(0),
 	    ckh_count(&ckh));

 	/* Insert. */
 	for (i = 0; i < sizeof(strs)/sizeof(const char *); i++) {
-		ckh_insert(tsd, &ckh, strs[i], strs[i]);
+		ckh_insert(tsdn, &ckh, strs[i], strs[i]);
 		assert_zu_eq(ckh_count(&ckh), i+1,
 		    "ckh_count() should return %zu, but it returned %zu", i+1,
 		    ckh_count(&ckh));
@@ -85,7 +85,7 @@ TEST_BEGIN(test_count_insert_search_remove)
 		vp = (i & 2) ? &v.p : NULL;
 		k.p = NULL;
 		v.p = NULL;
-		assert_false(ckh_remove(tsd, &ckh, strs[i], kp, vp),
+		assert_false(ckh_remove(tsdn, &ckh, strs[i], kp, vp),
 		    "Unexpected ckh_remove() error");

 		ks = (i & 1) ? strs[i] : (const char *)NULL;
@@ -101,22 +101,22 @@ TEST_BEGIN(test_count_insert_search_remove)
 		    ckh_count(&ckh));
 	}

-	ckh_delete(tsd, &ckh);
+	ckh_delete(tsdn, &ckh);
 }
 TEST_END

 TEST_BEGIN(test_insert_iter_remove)
 {
 #define	NITEMS ZU(1000)
-	tsd_t *tsd;
+	tsdn_t *tsdn;
 	ckh_t ckh;
 	void **p[NITEMS];
 	void *q, *r;
 	size_t i;

-	tsd = tsd_fetch();
+	tsdn = tsdn_fetch();

-	assert_false(ckh_new(tsd, &ckh, 2, ckh_pointer_hash,
+	assert_false(ckh_new(tsdn, &ckh, 2, ckh_pointer_hash,
 	    ckh_pointer_keycomp), "Unexpected ckh_new() error");

 	for (i = 0; i < NITEMS; i++) {
@@ -128,7 +128,7 @@ TEST_BEGIN(test_insert_iter_remove)
 		size_t j;

 		for (j = i; j < NITEMS; j++) {
-			assert_false(ckh_insert(tsd, &ckh, p[j], p[j]),
+			assert_false(ckh_insert(tsdn, &ckh, p[j], p[j]),
 			    "Unexpected ckh_insert() failure");
 			assert_false(ckh_search(&ckh, p[j], &q, &r),
 			    "Unexpected ckh_search() failure");
@@ -143,13 +143,13 @@ TEST_BEGIN(test_insert_iter_remove)
 		for (j = i + 1; j < NITEMS; j++) {
 			assert_false(ckh_search(&ckh, p[j], NULL, NULL),
 			    "Unexpected ckh_search() failure");
-			assert_false(ckh_remove(tsd, &ckh, p[j], &q, &r),
+			assert_false(ckh_remove(tsdn, &ckh, p[j], &q, &r),
 			    "Unexpected ckh_remove() failure");
 			assert_ptr_eq(p[j], q, "Key pointer mismatch");
 			assert_ptr_eq(p[j], r, "Value pointer mismatch");
 			assert_true(ckh_search(&ckh, p[j], NULL, NULL),
 			    "Unexpected ckh_search() success");
-			assert_true(ckh_remove(tsd, &ckh, p[j], &q, &r),
+			assert_true(ckh_remove(tsdn, &ckh, p[j], &q, &r),
 			    "Unexpected ckh_remove() success");
 		}

@@ -184,13 +184,13 @@ TEST_BEGIN(test_insert_iter_remove)
 	for (i = 0; i < NITEMS; i++) {
 		assert_false(ckh_search(&ckh, p[i], NULL, NULL),
 		    "Unexpected ckh_search() failure");
-		assert_false(ckh_remove(tsd, &ckh, p[i], &q, &r),
+		assert_false(ckh_remove(tsdn, &ckh, p[i], &q, &r),
 		    "Unexpected ckh_remove() failure");
 		assert_ptr_eq(p[i], q, "Key pointer mismatch");
 		assert_ptr_eq(p[i], r, "Value pointer mismatch");
 		assert_true(ckh_search(&ckh, p[i], NULL, NULL),
 		    "Unexpected ckh_search() success");
-		assert_true(ckh_remove(tsd, &ckh, p[i], &q, &r),
+		assert_true(ckh_remove(tsdn, &ckh, p[i], &q, &r),
 		    "Unexpected ckh_remove() success");
 		dallocx(p[i], 0);
 	}
@@ -198,7 +198,7 @@ TEST_BEGIN(test_insert_iter_remove)
 	assert_zu_eq(ckh_count(&ckh), 0,
 	    "ckh_count() should return %zu, but it returned %zu",
 	    ZU(0), ckh_count(&ckh));
-	ckh_delete(tsd, &ckh);
+	ckh_delete(tsdn, &ckh);
 #undef NITEMS
 }
 TEST_END
--- a/test/unit/fork.c
+++ b/test/unit/fork.c
@@ -0,0 +1,64 @@
+#include "test/jemalloc_test.h"
+
+#ifndef _WIN32
+#include <sys/wait.h>
+#endif
+
+TEST_BEGIN(test_fork)
+{
+#ifndef _WIN32
+	void *p;
+	pid_t pid;
+
+	p = malloc(1);
+	assert_ptr_not_null(p, "Unexpected malloc() failure");
+
+	pid = fork();
+
+	free(p);
+
+	p = malloc(64);
+	assert_ptr_not_null(p, "Unexpected malloc() failure");
+	free(p);
+
+	if (pid == -1) {
+		/* Error. */
+		test_fail("Unexpected fork() failure");
+	} else if (pid == 0) {
+		/* Child. */
+		exit(0);
+	} else {
+		int status;
+
+		/* Parent. */
+		while (true) {
+			if (waitpid(pid, &status, 0) == -1)
+				test_fail("Unexpected waitpid() failure");
+			if (WIFSIGNALED(status)) {
+				test_fail("Unexpected child termination due to "
+				    "signal %d", WTERMSIG(status));
+				break;
+			}
+			if (WIFEXITED(status)) {
+				if (WEXITSTATUS(status) != 0) {
+					test_fail(
+					    "Unexpected child exit value %d",
+					    WEXITSTATUS(status));
+				}
+				break;
+			}
+		}
+	}
+#else
+	test_skip("fork(2) is irrelevant to Windows");
+#endif
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_fork));
+}
--- a/test/unit/hash.c
+++ b/test/unit/hash.c
@@ -64,14 +64,15 @@ static void
 hash_variant_verify_key(hash_variant_t variant, uint8_t *key)
 {
 	const int hashbytes = hash_variant_bits(variant) / 8;
-	VARIABLE_ARRAY(uint8_t, hashes, hashbytes * 256);
+	const int hashes_size = hashbytes * 256;
+	VARIABLE_ARRAY(uint8_t, hashes, hashes_size);
 	VARIABLE_ARRAY(uint8_t, final, hashbytes);
 	unsigned i;
 	uint32_t computed, expected;

 	memset(key, 0, KEY_SIZE);
-	memset(hashes, 0, sizeof(hashes));
-	memset(final, 0, sizeof(final));
+	memset(hashes, 0, hashes_size);
+	memset(final, 0, hashbytes);

 	/*
 	 * Hash keys of the form {0}, {0,1}, {0,1,2}, ..., {0,1,...,255} as the
@@ -102,17 +103,17 @@ hash_variant_verify_key(hash_variant_t variant, uint8_t *key)
 	/* Hash the result array. */
 	switch (variant) {
 	case hash_variant_x86_32: {
-		uint32_t out = hash_x86_32(hashes, hashbytes*256, 0);
+		uint32_t out = hash_x86_32(hashes, hashes_size, 0);
 		memcpy(final, &out, sizeof(out));
 		break;
 	} case hash_variant_x86_128: {
 		uint64_t out[2];
-		hash_x86_128(hashes, hashbytes*256, 0, out);
+		hash_x86_128(hashes, hashes_size, 0, out);
 		memcpy(final, out, sizeof(out));
 		break;
 	} case hash_variant_x64_128: {
 		uint64_t out[2];
-		hash_x64_128(hashes, hashbytes*256, 0, out);
+		hash_x64_128(hashes, hashes_size, 0, out);
 		memcpy(final, out, sizeof(out));
 		break;
 	} default: not_reached();
--- a/test/unit/junk.c
+++ b/test/unit/junk.c
@@ -29,7 +29,7 @@ arena_dalloc_junk_small_intercept(void *ptr, arena_bin_info_t *bin_info)

 	arena_dalloc_junk_small_orig(ptr, bin_info);
 	for (i = 0; i < bin_info->reg_size; i++) {
-		assert_c_eq(((char *)ptr)[i], 0x5a,
+		assert_u_eq(((uint8_t *)ptr)[i], JEMALLOC_FREE_JUNK,
 		    "Missing junk fill for byte %zu/%zu of deallocated region",
 		    i, bin_info->reg_size);
 	}
@@ -44,7 +44,7 @@ arena_dalloc_junk_large_intercept(void *ptr, size_t usize)

 	arena_dalloc_junk_large_orig(ptr, usize);
 	for (i = 0; i < usize; i++) {
-		assert_c_eq(((char *)ptr)[i], 0x5a,
+		assert_u_eq(((uint8_t *)ptr)[i], JEMALLOC_FREE_JUNK,
 		    "Missing junk fill for byte %zu/%zu of deallocated region",
 		    i, usize);
 	}
@@ -53,10 +53,10 @@ arena_dalloc_junk_large_intercept(void *ptr, size_t usize)
 }

 static void
-huge_dalloc_junk_intercept(void *ptr, size_t usize)
+huge_dalloc_junk_intercept(tsdn_t *tsdn, void *ptr, size_t usize)
 {

-	huge_dalloc_junk_orig(ptr, usize);
+	huge_dalloc_junk_orig(tsdn, ptr, usize);
 	/*
 	 * The conditions under which junk filling actually occurs are nuanced
 	 * enough that it doesn't make sense to duplicate the decision logic in
@@ -69,7 +69,7 @@ huge_dalloc_junk_intercept(void *ptr, size_t usize)
 static void
 test_junk(size_t sz_min, size_t sz_max)
 {
-	char *s;
+	uint8_t *s;
 	size_t sz_prev, sz, i;

 	if (opt_junk_free) {
@@ -82,23 +82,23 @@ test_junk(size_t sz_min, size_t sz_max)
 	}

 	sz_prev = 0;
-	s = (char *)mallocx(sz_min, 0);
+	s = (uint8_t *)mallocx(sz_min, 0);
 	assert_ptr_not_null((void *)s, "Unexpected mallocx() failure");

 	for (sz = sallocx(s, 0); sz <= sz_max;
 	    sz_prev = sz, sz = sallocx(s, 0)) {
 		if (sz_prev > 0) {
-			assert_c_eq(s[0], 'a',
+			assert_u_eq(s[0], 'a',
 			    "Previously allocated byte %zu/%zu is corrupted",
 			    ZU(0), sz_prev);
-			assert_c_eq(s[sz_prev-1], 'a',
+			assert_u_eq(s[sz_prev-1], 'a',
 			    "Previously allocated byte %zu/%zu is corrupted",
 			    sz_prev-1, sz_prev);
 		}

 		for (i = sz_prev; i < sz; i++) {
 			if (opt_junk_alloc) {
-				assert_c_eq(s[i], 0xa5,
+				assert_u_eq(s[i], JEMALLOC_ALLOC_JUNK,
 				    "Newly allocated byte %zu/%zu isn't "
 				    "junk-filled", i, sz);
 			}
@@ -107,7 +107,7 @@ test_junk(size_t sz_min, size_t sz_max)

 		if (xallocx(s, sz+1, 0, 0) == sz) {
 			watch_junking(s);
-			s = (char *)rallocx(s, sz+1, 0);
+			s = (uint8_t *)rallocx(s, sz+1, 0);
 			assert_ptr_not_null((void *)s,
 			    "Unexpected rallocx() failure");
 			assert_true(!opt_junk_free || saw_junking,
@@ -244,7 +244,6 @@ int
 main(void)
 {

-	assert(!config_fill || opt_junk_alloc || opt_junk_free);
 	return (test(
 	    test_junk_small,
 	    test_junk_large,
--- a/test/unit/junk_alloc.c
+++ b/test/unit/junk_alloc.c
@@ -1,3 +1,3 @@
-#define JEMALLOC_TEST_JUNK_OPT "junk:alloc"
+#define	JEMALLOC_TEST_JUNK_OPT "junk:alloc"
 #include "junk.c"
 #undef JEMALLOC_TEST_JUNK_OPT
--- a/test/unit/junk_free.c
+++ b/test/unit/junk_free.c
@@ -1,3 +1,3 @@
-#define JEMALLOC_TEST_JUNK_OPT "junk:free"
+#define	JEMALLOC_TEST_JUNK_OPT "junk:free"
 #include "junk.c"
 #undef JEMALLOC_TEST_JUNK_OPT
--- a/test/unit/ph.c
+++ b/test/unit/ph.c
@@ -0,0 +1,290 @@
+#include "test/jemalloc_test.h"
+
+typedef struct node_s node_t;
+
+struct node_s {
+#define	NODE_MAGIC 0x9823af7e
+	uint32_t magic;
+	phn(node_t) link;
+	uint64_t key;
+};
+
+static int
+node_cmp(const node_t *a, const node_t *b)
+{
+	int ret;
+
+	ret = (a->key > b->key) - (a->key < b->key);
+	if (ret == 0) {
+		/*
+		 * Duplicates are not allowed in the heap, so force an
+		 * arbitrary ordering for non-identical items with equal keys.
+		 */
+		ret = (((uintptr_t)a) > ((uintptr_t)b))
+		    - (((uintptr_t)a) < ((uintptr_t)b));
+	}
+	return (ret);
+}
+
+static int
+node_cmp_magic(const node_t *a, const node_t *b) {
+
+	assert_u32_eq(a->magic, NODE_MAGIC, "Bad magic");
+	assert_u32_eq(b->magic, NODE_MAGIC, "Bad magic");
+
+	return (node_cmp(a, b));
+}
+
+typedef ph(node_t) heap_t;
+ph_gen(static, heap_, heap_t, node_t, link, node_cmp_magic);
+
+static void
+node_print(const node_t *node, unsigned depth)
+{
+	unsigned i;
+	node_t *leftmost_child, *sibling;
+
+	for (i = 0; i < depth; i++)
+		malloc_printf("\t");
+	malloc_printf("%2"FMTu64"\n", node->key);
+
+	leftmost_child = phn_lchild_get(node_t, link, node);
+	if (leftmost_child == NULL)
+		return;
+	node_print(leftmost_child, depth + 1);
+
+	for (sibling = phn_next_get(node_t, link, leftmost_child); sibling !=
+	    NULL; sibling = phn_next_get(node_t, link, sibling)) {
+		node_print(sibling, depth + 1);
+	}
+}
+
+static void
+heap_print(const heap_t *heap)
+{
+	node_t *auxelm;
+
+	malloc_printf("vvv heap %p vvv\n", heap);
+	if (heap->ph_root == NULL)
+		goto label_return;
+
+	node_print(heap->ph_root, 0);
+
+	for (auxelm = phn_next_get(node_t, link, heap->ph_root); auxelm != NULL;
+	    auxelm = phn_next_get(node_t, link, auxelm)) {
+		assert_ptr_eq(phn_next_get(node_t, link, phn_prev_get(node_t,
+		    link, auxelm)), auxelm,
+		    "auxelm's prev doesn't link to auxelm");
+		node_print(auxelm, 0);
+	}
+
+label_return:
+	malloc_printf("^^^ heap %p ^^^\n", heap);
+}
+
+static unsigned
+node_validate(const node_t *node, const node_t *parent)
+{
+	unsigned nnodes = 1;
+	node_t *leftmost_child, *sibling;
+
+	if (parent != NULL) {
+		assert_d_ge(node_cmp_magic(node, parent), 0,
+		    "Child is less than parent");
+	}
+
+	leftmost_child = phn_lchild_get(node_t, link, node);
+	if (leftmost_child == NULL)
+		return (nnodes);
+	assert_ptr_eq((void *)phn_prev_get(node_t, link, leftmost_child),
+	    (void *)node, "Leftmost child does not link to node");
+	nnodes += node_validate(leftmost_child, node);
+
+	for (sibling = phn_next_get(node_t, link, leftmost_child); sibling !=
+	    NULL; sibling = phn_next_get(node_t, link, sibling)) {
+		assert_ptr_eq(phn_next_get(node_t, link, phn_prev_get(node_t,
+		    link, sibling)), sibling,
+		    "sibling's prev doesn't link to sibling");
+		nnodes += node_validate(sibling, node);
+	}
+	return (nnodes);
+}
+
+static unsigned
+heap_validate(const heap_t *heap)
+{
+	unsigned nnodes = 0;
+	node_t *auxelm;
+
+	if (heap->ph_root == NULL)
+		goto label_return;
+
+	nnodes += node_validate(heap->ph_root, NULL);
+
+	for (auxelm = phn_next_get(node_t, link, heap->ph_root); auxelm != NULL;
+	    auxelm = phn_next_get(node_t, link, auxelm)) {
+		assert_ptr_eq(phn_next_get(node_t, link, phn_prev_get(node_t,
+		    link, auxelm)), auxelm,
+		    "auxelm's prev doesn't link to auxelm");
+		nnodes += node_validate(auxelm, NULL);
+	}
+
+label_return:
+	if (false)
+		heap_print(heap);
+	return (nnodes);
+}
+
+TEST_BEGIN(test_ph_empty)
+{
+	heap_t heap;
+
+	heap_new(&heap);
+	assert_true(heap_empty(&heap), "Heap should be empty");
+	assert_ptr_null(heap_first(&heap), "Unexpected node");
+}
+TEST_END
+
+static void
+node_remove(heap_t *heap, node_t *node)
+{
+
+	heap_remove(heap, node);
+
+	node->magic = 0;
+}
+
+static node_t *
+node_remove_first(heap_t *heap)
+{
+	node_t *node = heap_remove_first(heap);
+	node->magic = 0;
+	return (node);
+}
+
+TEST_BEGIN(test_ph_random)
+{
+#define	NNODES 25
+#define	NBAGS 250
+#define	SEED 42
+	sfmt_t *sfmt;
+	uint64_t bag[NNODES];
+	heap_t heap;
+	node_t nodes[NNODES];
+	unsigned i, j, k;
+
+	sfmt = init_gen_rand(SEED);
+	for (i = 0; i < NBAGS; i++) {
+		switch (i) {
+		case 0:
+			/* Insert in order. */
+			for (j = 0; j < NNODES; j++)
+				bag[j] = j;
+			break;
+		case 1:
+			/* Insert in reverse order. */
+			for (j = 0; j < NNODES; j++)
+				bag[j] = NNODES - j - 1;
+			break;
+		default:
+			for (j = 0; j < NNODES; j++)
+				bag[j] = gen_rand64_range(sfmt, NNODES);
+		}
+
+		for (j = 1; j <= NNODES; j++) {
+			/* Initialize heap and nodes. */
+			heap_new(&heap);
+			assert_u_eq(heap_validate(&heap), 0,
+			    "Incorrect node count");
+			for (k = 0; k < j; k++) {
+				nodes[k].magic = NODE_MAGIC;
+				nodes[k].key = bag[k];
+			}
+
+			/* Insert nodes. */
+			for (k = 0; k < j; k++) {
+				heap_insert(&heap, &nodes[k]);
+				if (i % 13 == 12) {
+					/* Trigger merging. */
+					assert_ptr_not_null(heap_first(&heap),
+					    "Heap should not be empty");
+				}
+				assert_u_eq(heap_validate(&heap), k + 1,
+				    "Incorrect node count");
+			}
+
+			assert_false(heap_empty(&heap),
+			    "Heap should not be empty");
+
+			/* Remove nodes. */
+			switch (i % 4) {
+			case 0:
+				for (k = 0; k < j; k++) {
+					assert_u_eq(heap_validate(&heap), j - k,
+					    "Incorrect node count");
+					node_remove(&heap, &nodes[k]);
+					assert_u_eq(heap_validate(&heap), j - k
+					    - 1, "Incorrect node count");
+				}
+				break;
+			case 1:
+				for (k = j; k > 0; k--) {
+					node_remove(&heap, &nodes[k-1]);
+					assert_u_eq(heap_validate(&heap), k - 1,
+					    "Incorrect node count");
+				}
+				break;
+			case 2: {
+				node_t *prev = NULL;
+				for (k = 0; k < j; k++) {
+					node_t *node = node_remove_first(&heap);
+					assert_u_eq(heap_validate(&heap), j - k
+					    - 1, "Incorrect node count");
+					if (prev != NULL) {
+						assert_d_ge(node_cmp(node,
+						    prev), 0,
+						    "Bad removal order");
+					}
+					prev = node;
+				}
+				break;
+			} case 3: {
+				node_t *prev = NULL;
+				for (k = 0; k < j; k++) {
+					node_t *node = heap_first(&heap);
+					assert_u_eq(heap_validate(&heap), j - k,
+					    "Incorrect node count");
+					if (prev != NULL) {
+						assert_d_ge(node_cmp(node,
+						    prev), 0,
+						    "Bad removal order");
+					}
+					node_remove(&heap, node);
+					assert_u_eq(heap_validate(&heap), j - k
+					    - 1, "Incorrect node count");
+					prev = node;
+				}
+				break;
+			} default:
+				not_reached();
+			}
+
+			assert_ptr_null(heap_first(&heap),
+			    "Heap should be empty");
+			assert_true(heap_empty(&heap), "Heap should be empty");
+		}
+	}
+	fini_gen_rand(sfmt);
+#undef NNODES
+#undef SEED
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_ph_empty,
+	    test_ph_random));
+}
--- a/test/unit/prof_reset.c
+++ b/test/unit/prof_reset.c
@@ -94,7 +94,8 @@ TEST_END
 bool prof_dump_header_intercepted = false;
 prof_cnt_t cnt_all_copy = {0, 0, 0, 0};
 static bool
-prof_dump_header_intercept(bool propagate_err, const prof_cnt_t *cnt_all)
+prof_dump_header_intercept(tsdn_t *tsdn, bool propagate_err,
+    const prof_cnt_t *cnt_all)
 {

 	prof_dump_header_intercepted = true;
--- a/test/unit/stats.c
+++ b/test/unit/stats.c
@@ -220,11 +220,11 @@ TEST_BEGIN(test_stats_arenas_large)
 	if (config_stats) {
 		assert_zu_gt(allocated, 0,
 		    "allocated should be greater than zero");
-		assert_zu_gt(nmalloc, 0,
+		assert_u64_gt(nmalloc, 0,
 		    "nmalloc should be greater than zero");
-		assert_zu_ge(nmalloc, ndalloc,
+		assert_u64_ge(nmalloc, ndalloc,
 		    "nmalloc should be at least as large as ndalloc");
-		assert_zu_gt(nrequests, 0,
+		assert_u64_gt(nrequests, 0,
 		    "nrequests should be greater than zero");
 	}

@@ -262,9 +262,9 @@ TEST_BEGIN(test_stats_arenas_huge)
 	if (config_stats) {
 		assert_zu_gt(allocated, 0,
 		    "allocated should be greater than zero");
-		assert_zu_gt(nmalloc, 0,
+		assert_u64_gt(nmalloc, 0,
 		    "nmalloc should be greater than zero");
-		assert_zu_ge(nmalloc, ndalloc,
+		assert_u64_ge(nmalloc, ndalloc,
 		    "nmalloc should be at least as large as ndalloc");
 	}

--- a/test/unit/tsd.c
+++ b/test/unit/tsd.c
@@ -99,6 +99,11 @@ int
 main(void)
 {

+	/* Core tsd bootstrapping must happen prior to data_tsd_boot(). */
+	if (nallocx(1, 0) == 0) {
+		malloc_printf("Initialization error");
+		return (test_status_fail);
+	}
 	data_tsd_boot();

 	return (test(
--- a/test/unit/util.c
+++ b/test/unit/util.c
@@ -4,27 +4,27 @@
 	unsigned i, pow2;						\
 	t x;								\
 									\
-	assert_zu_eq(pow2_ceil_##suf(0), 0, "Unexpected result");	\
+	assert_##suf##_eq(pow2_ceil_##suf(0), 0, "Unexpected result");	\
 									\
 	for (i = 0; i < sizeof(t) * 8; i++) {				\
-		assert_zu_eq(pow2_ceil_##suf(((t)1) << i), ((t)1) << i,	\
-		    "Unexpected result");				\
+		assert_##suf##_eq(pow2_ceil_##suf(((t)1) << i), ((t)1)	\
+		    << i, "Unexpected result");				\
 	}								\
 									\
 	for (i = 2; i < sizeof(t) * 8; i++) {				\
-		assert_zu_eq(pow2_ceil_##suf((((t)1) << i) - 1),	\
+		assert_##suf##_eq(pow2_ceil_##suf((((t)1) << i) - 1),	\
 		    ((t)1) << i, "Unexpected result");			\
 	}								\
 									\
 	for (i = 0; i < sizeof(t) * 8 - 1; i++) {			\
-		assert_zu_eq(pow2_ceil_##suf((((t)1) << i) + 1),	\
+		assert_##suf##_eq(pow2_ceil_##suf((((t)1) << i) + 1),	\
 		    ((t)1) << (i+1), "Unexpected result");		\
 	}								\
 									\
 	for (pow2 = 1; pow2 < 25; pow2++) {				\
 		for (x = (((t)1) << (pow2-1)) + 1; x <= ((t)1) << pow2;	\
 		    x++) {						\
-			assert_zu_eq(pow2_ceil_##suf(x),		\
+			assert_##suf##_eq(pow2_ceil_##suf(x),		\
 			    ((t)1) << pow2,				\
 			    "Unexpected result, x=%"pri, x);		\
 		}							\
@@ -160,14 +160,14 @@ TEST_BEGIN(test_malloc_snprintf_truncated)
 {
 #define	BUFLEN	15
 	char buf[BUFLEN];
-	int result;
+	size_t result;
 	size_t len;
-#define TEST(expected_str_untruncated, ...) do {			\
+#define	TEST(expected_str_untruncated, ...) do {			\
 	result = malloc_snprintf(buf, len, __VA_ARGS__);		\
 	assert_d_eq(strncmp(buf, expected_str_untruncated, len-1), 0,	\
 	    "Unexpected string inequality (\"%s\" vs \"%s\")",		\
-	    buf, expected_str_untruncated);		\
-	assert_d_eq(result, strlen(expected_str_untruncated),		\
+	    buf, expected_str_untruncated);				\
+	assert_zu_eq(result, strlen(expected_str_untruncated),		\
 	    "Unexpected result");					\
 } while (0)

@@ -193,11 +193,11 @@ TEST_BEGIN(test_malloc_snprintf)
 {
 #define	BUFLEN	128
 	char buf[BUFLEN];
-	int result;
+	size_t result;
 #define	TEST(expected_str, ...) do {					\
 	result = malloc_snprintf(buf, sizeof(buf), __VA_ARGS__);	\
 	assert_str_eq(buf, expected_str, "Unexpected output");		\
-	assert_d_eq(result, strlen(expected_str), "Unexpected result");	\
+	assert_zu_eq(result, strlen(expected_str), "Unexpected result");\
 } while (0)

 	TEST("hello", "hello");
--- a/test/unit/witness.c
+++ b/test/unit/witness.c
@@ -0,0 +1,278 @@
+#include "test/jemalloc_test.h"
+
+static witness_lock_error_t *witness_lock_error_orig;
+static witness_owner_error_t *witness_owner_error_orig;
+static witness_not_owner_error_t *witness_not_owner_error_orig;
+static witness_lockless_error_t *witness_lockless_error_orig;
+
+static bool saw_lock_error;
+static bool saw_owner_error;
+static bool saw_not_owner_error;
+static bool saw_lockless_error;
+
+static void
+witness_lock_error_intercept(const witness_list_t *witnesses,
+    const witness_t *witness)
+{
+
+	saw_lock_error = true;
+}
+
+static void
+witness_owner_error_intercept(const witness_t *witness)
+{
+
+	saw_owner_error = true;
+}
+
+static void
+witness_not_owner_error_intercept(const witness_t *witness)
+{
+
+	saw_not_owner_error = true;
+}
+
+static void
+witness_lockless_error_intercept(const witness_list_t *witnesses)
+{
+
+	saw_lockless_error = true;
+}
+
+static int
+witness_comp(const witness_t *a, const witness_t *b)
+{
+
+	assert_u_eq(a->rank, b->rank, "Witnesses should have equal rank");
+
+	return (strcmp(a->name, b->name));
+}
+
+static int
+witness_comp_reverse(const witness_t *a, const witness_t *b)
+{
+
+	assert_u_eq(a->rank, b->rank, "Witnesses should have equal rank");
+
+	return (-strcmp(a->name, b->name));
+}
+
+TEST_BEGIN(test_witness)
+{
+	witness_t a, b;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, NULL);
+	witness_assert_not_owner(tsdn, &a);
+	witness_lock(tsdn, &a);
+	witness_assert_owner(tsdn, &a);
+
+	witness_init(&b, "b", 2, NULL);
+	witness_assert_not_owner(tsdn, &b);
+	witness_lock(tsdn, &b);
+	witness_assert_owner(tsdn, &b);
+
+	witness_unlock(tsdn, &a);
+	witness_unlock(tsdn, &b);
+
+	witness_assert_lockless(tsdn);
+}
+TEST_END
+
+TEST_BEGIN(test_witness_comp)
+{
+	witness_t a, b, c, d;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, witness_comp);
+	witness_assert_not_owner(tsdn, &a);
+	witness_lock(tsdn, &a);
+	witness_assert_owner(tsdn, &a);
+
+	witness_init(&b, "b", 1, witness_comp);
+	witness_assert_not_owner(tsdn, &b);
+	witness_lock(tsdn, &b);
+	witness_assert_owner(tsdn, &b);
+	witness_unlock(tsdn, &b);
+
+	witness_lock_error_orig = witness_lock_error;
+	witness_lock_error = witness_lock_error_intercept;
+	saw_lock_error = false;
+
+	witness_init(&c, "c", 1, witness_comp_reverse);
+	witness_assert_not_owner(tsdn, &c);
+	assert_false(saw_lock_error, "Unexpected witness lock error");
+	witness_lock(tsdn, &c);
+	assert_true(saw_lock_error, "Expected witness lock error");
+	witness_unlock(tsdn, &c);
+
+	saw_lock_error = false;
+
+	witness_init(&d, "d", 1, NULL);
+	witness_assert_not_owner(tsdn, &d);
+	assert_false(saw_lock_error, "Unexpected witness lock error");
+	witness_lock(tsdn, &d);
+	assert_true(saw_lock_error, "Expected witness lock error");
+	witness_unlock(tsdn, &d);
+
+	witness_unlock(tsdn, &a);
+
+	witness_assert_lockless(tsdn);
+
+	witness_lock_error = witness_lock_error_orig;
+}
+TEST_END
+
+TEST_BEGIN(test_witness_reversal)
+{
+	witness_t a, b;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	witness_lock_error_orig = witness_lock_error;
+	witness_lock_error = witness_lock_error_intercept;
+	saw_lock_error = false;
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, NULL);
+	witness_init(&b, "b", 2, NULL);
+
+	witness_lock(tsdn, &b);
+	assert_false(saw_lock_error, "Unexpected witness lock error");
+	witness_lock(tsdn, &a);
+	assert_true(saw_lock_error, "Expected witness lock error");
+
+	witness_unlock(tsdn, &a);
+	witness_unlock(tsdn, &b);
+
+	witness_assert_lockless(tsdn);
+
+	witness_lock_error = witness_lock_error_orig;
+}
+TEST_END
+
+TEST_BEGIN(test_witness_recursive)
+{
+	witness_t a;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	witness_not_owner_error_orig = witness_not_owner_error;
+	witness_not_owner_error = witness_not_owner_error_intercept;
+	saw_not_owner_error = false;
+
+	witness_lock_error_orig = witness_lock_error;
+	witness_lock_error = witness_lock_error_intercept;
+	saw_lock_error = false;
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, NULL);
+
+	witness_lock(tsdn, &a);
+	assert_false(saw_lock_error, "Unexpected witness lock error");
+	assert_false(saw_not_owner_error, "Unexpected witness not owner error");
+	witness_lock(tsdn, &a);
+	assert_true(saw_lock_error, "Expected witness lock error");
+	assert_true(saw_not_owner_error, "Expected witness not owner error");
+
+	witness_unlock(tsdn, &a);
+
+	witness_assert_lockless(tsdn);
+
+	witness_owner_error = witness_owner_error_orig;
+	witness_lock_error = witness_lock_error_orig;
+
+}
+TEST_END
+
+TEST_BEGIN(test_witness_unlock_not_owned)
+{
+	witness_t a;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	witness_owner_error_orig = witness_owner_error;
+	witness_owner_error = witness_owner_error_intercept;
+	saw_owner_error = false;
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, NULL);
+
+	assert_false(saw_owner_error, "Unexpected owner error");
+	witness_unlock(tsdn, &a);
+	assert_true(saw_owner_error, "Expected owner error");
+
+	witness_assert_lockless(tsdn);
+
+	witness_owner_error = witness_owner_error_orig;
+}
+TEST_END
+
+TEST_BEGIN(test_witness_lockful)
+{
+	witness_t a;
+	tsdn_t *tsdn;
+
+	test_skip_if(!config_debug);
+
+	witness_lockless_error_orig = witness_lockless_error;
+	witness_lockless_error = witness_lockless_error_intercept;
+	saw_lockless_error = false;
+
+	tsdn = tsdn_fetch();
+
+	witness_assert_lockless(tsdn);
+
+	witness_init(&a, "a", 1, NULL);
+
+	assert_false(saw_lockless_error, "Unexpected lockless error");
+	witness_assert_lockless(tsdn);
+
+	witness_lock(tsdn, &a);
+	witness_assert_lockless(tsdn);
+	assert_true(saw_lockless_error, "Expected lockless error");
+
+	witness_unlock(tsdn, &a);
+
+	witness_assert_lockless(tsdn);
+
+	witness_lockless_error = witness_lockless_error_orig;
+}
+TEST_END
+
+int
+main(void)
+{
+
+	return (test(
+	    test_witness,
+	    test_witness_comp,
+	    test_witness_reversal,
+	    test_witness_recursive,
+	    test_witness_unlock_not_owned,
+	    test_witness_lockful));
+}
--- a/test/unit/zero.c
+++ b/test/unit/zero.c
@@ -8,39 +8,41 @@ const char *malloc_conf =
 static void
 test_zero(size_t sz_min, size_t sz_max)
 {
-	char *s;
+	uint8_t *s;
 	size_t sz_prev, sz, i;
+#define	MAGIC	((uint8_t)0x61)

 	sz_prev = 0;
-	s = (char *)mallocx(sz_min, 0);
+	s = (uint8_t *)mallocx(sz_min, 0);
 	assert_ptr_not_null((void *)s, "Unexpected mallocx() failure");

 	for (sz = sallocx(s, 0); sz <= sz_max;
 	    sz_prev = sz, sz = sallocx(s, 0)) {
 		if (sz_prev > 0) {
-			assert_c_eq(s[0], 'a',
+			assert_u_eq(s[0], MAGIC,
 			    "Previously allocated byte %zu/%zu is corrupted",
 			    ZU(0), sz_prev);
-			assert_c_eq(s[sz_prev-1], 'a',
+			assert_u_eq(s[sz_prev-1], MAGIC,
 			    "Previously allocated byte %zu/%zu is corrupted",
 			    sz_prev-1, sz_prev);
 		}

 		for (i = sz_prev; i < sz; i++) {
-			assert_c_eq(s[i], 0x0,
+			assert_u_eq(s[i], 0x0,
 			    "Newly allocated byte %zu/%zu isn't zero-filled",
 			    i, sz);
-			s[i] = 'a';
+			s[i] = MAGIC;
 		}

 		if (xallocx(s, sz+1, 0, 0) == sz) {
-			s = (char *)rallocx(s, sz+1, 0);
+			s = (uint8_t *)rallocx(s, sz+1, 0);
 			assert_ptr_not_null((void *)s,
 			    "Unexpected rallocx() failure");
 		}
 	}

 	dallocx(s, 0);
+#undef MAGIC
 }

 TEST_BEGIN(test_zero_small)