Bug 799090 - Update jemalloc3 to commit d0ffd8e. r=jlebar

memory/jemalloc/src/ChangeLog

@@ -6,6 +6,21 @@ found in the git revision history:
     http://www.canonware.com/cgi-bin/gitweb.cgi?p=jemalloc.git
     git://canonware.com/jemalloc.git
 
+* 3.x.x (XXX not yet released)
+
+  New features:
+  - Auto-detect whether running inside Valgrind, thus removing the need to
+    manually specify MALLOC_CONF=valgrind:true.
+
+  Incompatible changes:
+  - Disable tcache by default if running inside Valgrind, in order to avoid
+    making unallocated objects appear reachable to Valgrind.
+
+  Bug fixes:
+  - Fix heap profiling crash if sampled object is freed via realloc(p, 0).
+  - Remove const from __*_hook variable declarations, so that glibc can modify
+    them during process forking.
+
 * 3.0.0 (May 11, 2012)
 
   Although this version adds some major new features, the primary focus is on

memory/jemalloc/src/VERSION

@@ -1 +1 @@
-3.0.0-0-gfc9b1dbf69f59d7ecfc4ac68da9847e017e1d046
+1.0.0-357-gd0ffd8ed4f6aa4cf7248028eddfcb35f93247fe4

memory/jemalloc/src/configure

@@ -4448,6 +4448,7 @@ MKLIB='ar crus $@'
 CC_MM=1
 
 default_munmap="1"
+JEMALLOC_USABLE_SIZE_CONST="const"
 case "${host}" in
   *-*-darwin*)
 	CFLAGS="$CFLAGS"
@@ -4477,6 +4478,7 @@ case "${host}" in
 
 	$as_echo "#define JEMALLOC_THREADED_INIT  " >>confdefs.h
 
+	JEMALLOC_USABLE_SIZE_CONST=""
 	default_munmap="0"
 	;;
   *-*-netbsd*)
@@ -4552,6 +4554,10 @@ $as_echo "Unsupported operating system: ${host}" >&6; }
 	abi="elf"
 	;;
 esac
+cat >>confdefs.h <<_ACEOF
+#define JEMALLOC_USABLE_SIZE_CONST $JEMALLOC_USABLE_SIZE_CONST
+_ACEOF
+
 
 
 

memory/jemalloc/src/configure.ac

@@ -237,6 +237,7 @@ dnl Define cpp macros in CPPFLAGS, rather than doing AC_DEFINE(macro), since the
 dnl definitions need to be seen before any headers are included, which is a pain
 dnl to make happen otherwise.
 default_munmap="1"
+JEMALLOC_USABLE_SIZE_CONST="const"
 case "${host}" in
   *-*-darwin*)
 	CFLAGS="$CFLAGS"
@@ -262,6 +263,7 @@ case "${host}" in
 	abi="elf"
 	AC_DEFINE([JEMALLOC_PURGE_MADVISE_DONTNEED], [ ])
 	AC_DEFINE([JEMALLOC_THREADED_INIT], [ ])
+	JEMALLOC_USABLE_SIZE_CONST=""
 	default_munmap="0"
 	;;
   *-*-netbsd*)
@@ -323,6 +325,7 @@ case "${host}" in
 	abi="elf"
 	;;
 esac
+AC_DEFINE_UNQUOTED([JEMALLOC_USABLE_SIZE_CONST], [$JEMALLOC_USABLE_SIZE_CONST])
 AC_SUBST([abi])
 AC_SUBST([RPATH])
 AC_SUBST([LD_PRELOAD_VAR])

memory/jemalloc/src/doc/jemalloc.xml.in

@@ -846,7 +846,9 @@ for (i = 0; i < nbins; i++) {
         <literal>0x5a</literal>.  This is intended for debugging and will
         impact performance negatively.  This option is disabled by default
         unless <option>--enable-debug</option> is specified during
-        configuration, in which case it is enabled by default.</para></listitem>
+        configuration, in which case it is enabled by default unless running
+        inside <ulink
+        url="http://valgrind.org/">Valgrind</ulink>.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.quarantine">
@@ -865,8 +867,9 @@ for (i = 0; i < nbins; i++) {
         enabled.  This feature is of particular use in combination with <ulink
         url="http://valgrind.org/">Valgrind</ulink>, which can detect attempts
         to access quarantined objects.  This is intended for debugging and will
-        impact performance negatively.  The default quarantine size is
+        impact performance negatively.  The default quarantine size is 0 unless
-        0.</para></listitem>
+        running inside Valgrind, in which case the default is 16
+        MiB.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.redzone">
@@ -885,7 +888,7 @@ for (i = 0; i < nbins; i++) {
         which needs redzones in order to do effective buffer overflow/underflow
         detection.  This option is intended for debugging and will impact
         performance negatively.  This option is disabled by
-        default.</para></listitem>
+        default unless running inside Valgrind.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.zero">
@@ -926,15 +929,9 @@ for (i = 0; i < nbins; i++) {
           [<option>--enable-valgrind</option>]
         </term>
         <listitem><para><ulink url="http://valgrind.org/">Valgrind</ulink>
-        support enabled/disabled.  If enabled, several other options are
+        support enabled/disabled.  This option is vestigal because jemalloc
-        automatically modified during options processing to work well with
+        auto-detects whether it is running inside Valgrind.  This option is
-        Valgrind: <link linkend="opt.junk"><mallctl>opt.junk</mallctl></link>
+        disabled by default, unless running inside Valgrind.</para></listitem>
-        and <link linkend="opt.zero"><mallctl>opt.zero</mallctl></link> are set
-        to false, <link
-        linkend="opt.quarantine"><mallctl>opt.quarantine</mallctl></link> is
-        set to 16 MiB, and <link
-        linkend="opt.redzone"><mallctl>opt.redzone</mallctl></link> is set to
-        true.  This option is disabled by default.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.xmalloc">
@@ -972,7 +969,8 @@ malloc_conf = "xmalloc:true";]]></programlisting>
         <link
         linkend="opt.lg_tcache_max"><mallctl>opt.lg_tcache_max</mallctl></link>
         option for related tuning information.  This option is enabled by
-        default.</para></listitem>
+        default unless running inside <ulink
+        url="http://valgrind.org/">Valgrind</ulink>.</para></listitem>
       </varlistentry>
 
       <varlistentry id="opt.lg_tcache_max">
@@ -1865,9 +1863,7 @@ malloc_conf = "xmalloc:true";]]></programlisting>
     it detects, because the performance impact for storing such information
     would be prohibitive.  However, jemalloc does integrate with the most
     excellent <ulink url="http://valgrind.org/">Valgrind</ulink> tool if the
-    <option>--enable-valgrind</option> configuration option is enabled and the
+    <option>--enable-valgrind</option> configuration option is enabled.</para>
-    <link linkend="opt.valgrind"><mallctl>opt.valgrind</mallctl></link> option
-    is enabled.</para>
   </refsect1>
   <refsect1 id="diagnostic_messages">
     <title>DIAGNOSTIC MESSAGES</title>

memory/jemalloc/src/include/jemalloc/internal/chunk.h

@@ -45,6 +45,9 @@ extern size_t		arena_maxclass; /* Max size class for arenas. */
 void	*chunk_alloc(size_t size, size_t alignment, bool base, bool *zero);
 void	chunk_dealloc(void *chunk, size_t size, bool unmap);
 bool	chunk_boot(void);
+void	chunk_prefork(void);
+void	chunk_postfork_parent(void);
+void	chunk_postfork_child(void);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/

memory/jemalloc/src/include/jemalloc/internal/chunk_mmap.h

@@ -9,7 +9,7 @@
 /******************************************************************************/
 #ifdef JEMALLOC_H_EXTERNS
 
-void	pages_purge(void *addr, size_t length);
+bool	pages_purge(void *addr, size_t length);
 
 void	*chunk_alloc_mmap(size_t size, size_t alignment, bool *zero);
 bool	chunk_dealloc_mmap(void *chunk, size_t size);

memory/jemalloc/src/include/jemalloc/internal/ctl.h

@@ -75,6 +75,9 @@ 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,
     void *newp, size_t newlen);
 bool	ctl_boot(void);
+void	ctl_prefork(void);
+void	ctl_postfork_parent(void);
+void	ctl_postfork_child(void);
 
 #define	xmallctl(name, oldp, oldlenp, newp, newlen) do {		\
 	if (je_mallctl(name, oldp, oldlenp, newp, newlen)		\

memory/jemalloc/src/include/jemalloc/internal/extent.h

@@ -23,6 +23,9 @@ struct extent_node_s {
 
 	/* Total region size. */
 	size_t			size;
+
+	/* True if zero-filled; used by chunk recycling code. */
+	bool			zeroed;
 };
 typedef rb_tree(extent_node_t) extent_tree_t;
 

memory/jemalloc/src/include/jemalloc/internal/jemalloc_internal.h.in

@@ -270,6 +270,9 @@ static const bool config_ivsalloc =
 #  ifdef __arm__
 #    define LG_QUANTUM		3
 #  endif
+#  ifdef __hppa__
+#    define LG_QUANTUM		4
+#  endif
 #  ifdef __mips__
 #    define LG_QUANTUM		3
 #  endif
@@ -424,6 +427,7 @@ static const bool config_ivsalloc =
 		VALGRIND_FREELIKE_BLOCK(ptr, rzsize);			\
 } while (0)
 #else
+#define	RUNNING_ON_VALGRIND	((unsigned)0)
 #define	VALGRIND_MALLOCLIKE_BLOCK(addr, sizeB, rzB, is_zeroed)
 #define	VALGRIND_RESIZEINPLACE_BLOCK(addr, oldSizeB, newSizeB, rzB)
 #define	VALGRIND_FREELIKE_BLOCK(addr, rzB)

memory/jemalloc/src/include/jemalloc/internal/private_namespace.h

@@ -59,6 +59,7 @@
 #define	arenas_lock JEMALLOC_N(arenas_lock)
 #define	arenas_lrun_i_index JEMALLOC_N(arenas_lrun_i_index)
 #define	arenas_tls JEMALLOC_N(arenas_tls)
+#define	arenas_tsd JEMALLOC_N(arenas_tsd)
 #define	arenas_tsd_boot JEMALLOC_N(arenas_tsd_boot)
 #define	arenas_tsd_cleanup_wrapper JEMALLOC_N(arenas_tsd_cleanup_wrapper)
 #define	arenas_tsd_get JEMALLOC_N(arenas_tsd_get)
@@ -104,6 +105,9 @@
 #define	chunk_dss_prefork JEMALLOC_N(chunk_dss_prefork)
 #define	chunk_in_dss JEMALLOC_N(chunk_in_dss)
 #define	chunk_npages JEMALLOC_N(chunk_npages)
+#define	chunk_postfork_child JEMALLOC_N(chunk_postfork_child)
+#define	chunk_postfork_parent JEMALLOC_N(chunk_postfork_parent)
+#define	chunk_prefork JEMALLOC_N(chunk_prefork)
 #define	chunks_mtx JEMALLOC_N(chunks_mtx)
 #define	chunks_rtree JEMALLOC_N(chunks_rtree)
 #define	chunksize JEMALLOC_N(chunksize)
@@ -129,6 +133,9 @@
 #define	ctl_bymib JEMALLOC_N(ctl_bymib)
 #define	ctl_byname JEMALLOC_N(ctl_byname)
 #define	ctl_nametomib JEMALLOC_N(ctl_nametomib)
+#define	ctl_postfork_child JEMALLOC_N(ctl_postfork_child)
+#define	ctl_postfork_parent JEMALLOC_N(ctl_postfork_parent)
+#define	ctl_prefork JEMALLOC_N(ctl_prefork)
 #define	extent_tree_ad_first JEMALLOC_N(extent_tree_ad_first)
 #define	extent_tree_ad_insert JEMALLOC_N(extent_tree_ad_insert)
 #define	extent_tree_ad_iter JEMALLOC_N(extent_tree_ad_iter)
@@ -161,6 +168,7 @@
 #define	extent_tree_szad_reverse_iter_recurse JEMALLOC_N(extent_tree_szad_reverse_iter_recurse)
 #define	extent_tree_szad_reverse_iter_start JEMALLOC_N(extent_tree_szad_reverse_iter_start)
 #define	extent_tree_szad_search JEMALLOC_N(extent_tree_szad_search)
+#define	get_errno JEMALLOC_N(get_errno)
 #define	hash JEMALLOC_N(hash)
 #define	huge_allocated JEMALLOC_N(huge_allocated)
 #define	huge_boot JEMALLOC_N(huge_boot)
@@ -254,6 +262,9 @@
 #define	prof_lookup JEMALLOC_N(prof_lookup)
 #define	prof_malloc JEMALLOC_N(prof_malloc)
 #define	prof_mdump JEMALLOC_N(prof_mdump)
+#define	prof_postfork_child JEMALLOC_N(prof_postfork_child)
+#define	prof_postfork_parent JEMALLOC_N(prof_postfork_parent)
+#define	prof_prefork JEMALLOC_N(prof_prefork)
 #define	prof_promote JEMALLOC_N(prof_promote)
 #define	prof_realloc JEMALLOC_N(prof_realloc)
 #define	prof_sample_accum_update JEMALLOC_N(prof_sample_accum_update)
@@ -264,6 +275,7 @@
 #define	prof_tdata_init JEMALLOC_N(prof_tdata_init)
 #define	prof_tdata_initialized JEMALLOC_N(prof_tdata_initialized)
 #define	prof_tdata_tls JEMALLOC_N(prof_tdata_tls)
+#define	prof_tdata_tsd JEMALLOC_N(prof_tdata_tsd)
 #define	prof_tdata_tsd_boot JEMALLOC_N(prof_tdata_tsd_boot)
 #define	prof_tdata_tsd_cleanup_wrapper JEMALLOC_N(prof_tdata_tsd_cleanup_wrapper)
 #define	prof_tdata_tsd_get JEMALLOC_N(prof_tdata_tsd_get)
@@ -278,9 +290,13 @@
 #define	rtree_get JEMALLOC_N(rtree_get)
 #define	rtree_get_locked JEMALLOC_N(rtree_get_locked)
 #define	rtree_new JEMALLOC_N(rtree_new)
+#define	rtree_postfork_child JEMALLOC_N(rtree_postfork_child)
+#define	rtree_postfork_parent JEMALLOC_N(rtree_postfork_parent)
+#define	rtree_prefork JEMALLOC_N(rtree_prefork)
 #define	rtree_set JEMALLOC_N(rtree_set)
 #define	s2u JEMALLOC_N(s2u)
 #define	sa2u JEMALLOC_N(sa2u)
+#define	set_errno JEMALLOC_N(set_errno)
 #define	stats_arenas_i_bins_j_index JEMALLOC_N(stats_arenas_i_bins_j_index)
 #define	stats_arenas_i_index JEMALLOC_N(stats_arenas_i_index)
 #define	stats_arenas_i_lruns_j_index JEMALLOC_N(stats_arenas_i_lruns_j_index)
@@ -311,6 +327,7 @@
 #define	tcache_enabled_initialized JEMALLOC_N(tcache_enabled_initialized)
 #define	tcache_enabled_set JEMALLOC_N(tcache_enabled_set)
 #define	tcache_enabled_tls JEMALLOC_N(tcache_enabled_tls)
+#define	tcache_enabled_tsd JEMALLOC_N(tcache_enabled_tsd)
 #define	tcache_enabled_tsd_boot JEMALLOC_N(tcache_enabled_tsd_boot)
 #define	tcache_enabled_tsd_cleanup_wrapper JEMALLOC_N(tcache_enabled_tsd_cleanup_wrapper)
 #define	tcache_enabled_tsd_get JEMALLOC_N(tcache_enabled_tsd_get)
@@ -325,6 +342,7 @@
 #define	tcache_stats_merge JEMALLOC_N(tcache_stats_merge)
 #define	tcache_thread_cleanup JEMALLOC_N(tcache_thread_cleanup)
 #define	tcache_tls JEMALLOC_N(tcache_tls)
+#define	tcache_tsd JEMALLOC_N(tcache_tsd)
 #define	tcache_tsd_boot JEMALLOC_N(tcache_tsd_boot)
 #define	tcache_tsd_cleanup_wrapper JEMALLOC_N(tcache_tsd_cleanup_wrapper)
 #define	tcache_tsd_get JEMALLOC_N(tcache_tsd_get)
@@ -332,6 +350,7 @@
 #define	thread_allocated_booted JEMALLOC_N(thread_allocated_booted)
 #define	thread_allocated_initialized JEMALLOC_N(thread_allocated_initialized)
 #define	thread_allocated_tls JEMALLOC_N(thread_allocated_tls)
+#define	thread_allocated_tsd JEMALLOC_N(thread_allocated_tsd)
 #define	thread_allocated_tsd_boot JEMALLOC_N(thread_allocated_tsd_boot)
 #define	thread_allocated_tsd_cleanup_wrapper JEMALLOC_N(thread_allocated_tsd_cleanup_wrapper)
 #define	thread_allocated_tsd_get JEMALLOC_N(thread_allocated_tsd_get)

memory/jemalloc/src/include/jemalloc/internal/prof.h

@@ -223,6 +223,9 @@ void	prof_tdata_cleanup(void *arg);
 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);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/
@@ -506,7 +509,7 @@ prof_realloc(const void *ptr, size_t size, prof_thr_cnt_t *cnt,
 	if ((uintptr_t)cnt > (uintptr_t)1U) {
 		prof_ctx_set(ptr, cnt->ctx);
 		cnt->epoch++;
-	} else
+	} else if (ptr != NULL)
 		prof_ctx_set(ptr, (prof_ctx_t *)(uintptr_t)1U);
 	/*********/
 	mb_write();

memory/jemalloc/src/include/jemalloc/internal/rtree.h

@@ -36,6 +36,9 @@ struct rtree_s {
 #ifdef JEMALLOC_H_EXTERNS
 
 rtree_t	*rtree_new(unsigned bits);
+void	rtree_prefork(rtree_t *rtree);
+void	rtree_postfork_parent(rtree_t *rtree);
+void	rtree_postfork_child(rtree_t *rtree);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/

memory/jemalloc/src/include/jemalloc/jemalloc.h.in

@@ -59,7 +59,8 @@ JEMALLOC_EXPORT void *	je_memalign(size_t alignment, size_t size)
 JEMALLOC_EXPORT void *	je_valloc(size_t size) JEMALLOC_ATTR(malloc);
 #endif
 
-JEMALLOC_EXPORT size_t	je_malloc_usable_size(const void *ptr);
+JEMALLOC_EXPORT size_t	je_malloc_usable_size(
+    JEMALLOC_USABLE_SIZE_CONST void *ptr);
 JEMALLOC_EXPORT void	je_malloc_stats_print(void (*write_cb)(void *,
     const char *), void *je_cbopaque, const char *opts);
 JEMALLOC_EXPORT int	je_mallctl(const char *name, void *oldp,

memory/jemalloc/src/include/jemalloc/jemalloc_defs.h.in

@@ -221,6 +221,15 @@
 #undef JEMALLOC_OVERRIDE_MEMALIGN
 #undef JEMALLOC_OVERRIDE_VALLOC
 
+/*
+ * At least Linux omits the "const" in:
+ *
+ *   size_t malloc_usable_size(const void *ptr);
+ *
+ * Match the operating system's prototype.
+ */
+#undef JEMALLOC_USABLE_SIZE_CONST
+
 /*
  * Darwin (OS X) uses zones to work around Mach-O symbol override shortcomings.
  */

memory/jemalloc/src/src/arena.c

@@ -551,24 +551,12 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk)
 {
 	ql_head(arena_chunk_map_t) mapelms;
 	arena_chunk_map_t *mapelm;
-	size_t pageind, flag_unzeroed;
+	size_t pageind;
 	size_t ndirty;
 	size_t nmadvise;
 
 	ql_new(&mapelms);
 
-	flag_unzeroed =
-#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED
-   /*
-    * madvise(..., MADV_DONTNEED) results in zero-filled pages for anonymous
-    * mappings, but not for file-backed mappings.
-    */
-	    0
-#else
-	    CHUNK_MAP_UNZEROED
-#endif
-	    ;
-
 	/*
 	 * If chunk is the spare, temporarily re-allocate it, 1) so that its
 	 * run is reinserted into runs_avail_dirty, and 2) so that it cannot be
@@ -603,26 +591,12 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk)
 			assert(arena_mapbits_dirty_get(chunk, pageind) ==
 			    arena_mapbits_dirty_get(chunk, pageind+npages-1));
 			if (arena_mapbits_dirty_get(chunk, pageind) != 0) {
-				size_t i;
-
 				arena_avail_tree_remove(
 				    &arena->runs_avail_dirty, mapelm);
 
-				arena_mapbits_unzeroed_set(chunk, pageind,
-				    flag_unzeroed);
 				arena_mapbits_large_set(chunk, pageind,
 				    (npages << LG_PAGE), 0);
-				/*
-				 * Update internal elements in the page map, so
-				 * that CHUNK_MAP_UNZEROED is properly set.
-				 */
-				for (i = 1; i < npages - 1; i++) {
-					arena_mapbits_unzeroed_set(chunk,
-					    pageind+i, flag_unzeroed);
-				}
 				if (npages > 1) {
-					arena_mapbits_unzeroed_set(chunk,
-					    pageind+npages-1, flag_unzeroed);
 					arena_mapbits_large_set(chunk,
 					    pageind+npages-1, 0, 0);
 				}
@@ -685,14 +659,30 @@ arena_chunk_purge(arena_t *arena, arena_chunk_t *chunk)
 		    sizeof(arena_chunk_map_t)) + map_bias;
 		size_t npages = arena_mapbits_large_size_get(chunk, pageind) >>
 		    LG_PAGE;
+		bool unzeroed;
+		size_t flag_unzeroed, i;
 
 		assert(pageind + npages <= chunk_npages);
 		assert(ndirty >= npages);
 		if (config_debug)
 			ndirty -= npages;
-
+		unzeroed = pages_purge((void *)((uintptr_t)chunk + (pageind <<
-		pages_purge((void *)((uintptr_t)chunk + (pageind << LG_PAGE)),
+		    LG_PAGE)), (npages << LG_PAGE));
-		    (npages << LG_PAGE));
+		flag_unzeroed = unzeroed ? CHUNK_MAP_UNZEROED : 0;
+		/*
+		 * Set the unzeroed flag for all pages, now that pages_purge()
+		 * has returned whether the pages were zeroed as a side effect
+		 * of purging.  This chunk map modification is safe even though
+		 * the arena mutex isn't currently owned by this thread,
+		 * because the run is marked as allocated, thus protecting it
+		 * from being modified by any other thread.  As long as these
+		 * writes don't perturb the first and last elements'
+		 * CHUNK_MAP_ALLOCATED bits, behavior is well defined.
+		 */
+		for (i = 0; i < npages; i++) {
+			arena_mapbits_unzeroed_set(chunk, pageind+i,
+			    flag_unzeroed);
+		}
 		if (config_stats)
 			nmadvise++;
 	}

memory/jemalloc/src/src/chunk.c

@@ -43,6 +43,7 @@ chunk_recycle(size_t size, size_t alignment, bool base, bool *zero)
 	extent_node_t *node;
 	extent_node_t key;
 	size_t alloc_size, leadsize, trailsize;
+	bool zeroed;
 
 	if (base) {
 		/*
@@ -107,17 +108,18 @@ chunk_recycle(size_t size, size_t alignment, bool base, bool *zero)
 	}
 	malloc_mutex_unlock(&chunks_mtx);
 
-	if (node != NULL)
+	zeroed = false;
+	if (node != NULL) {
+		if (node->zeroed) {
+			zeroed = true;
+			*zero = true;
+		}
 		base_node_dealloc(node);
-#ifdef JEMALLOC_PURGE_MADVISE_DONTNEED
+	}
-	/* Pages are zeroed as a side effect of pages_purge(). */
+	if (zeroed == false && *zero) {
-	*zero = true;
-#else
-	if (*zero) {
 		VALGRIND_MAKE_MEM_UNDEFINED(ret, size);
 		memset(ret, 0, size);
 	}
-#endif
 	return (ret);
 }
 
@@ -191,9 +193,10 @@ label_return:
 static void
 chunk_record(void *chunk, size_t size)
 {
+	bool unzeroed;
 	extent_node_t *xnode, *node, *prev, key;
 
-	pages_purge(chunk, size);
+	unzeroed = pages_purge(chunk, size);
 
 	/*
 	 * Allocate a node before acquiring chunks_mtx even though it might not
@@ -216,6 +219,7 @@ chunk_record(void *chunk, size_t size)
 		extent_tree_szad_remove(&chunks_szad, node);
 		node->addr = chunk;
 		node->size += size;
+		node->zeroed = (node->zeroed && (unzeroed == false));
 		extent_tree_szad_insert(&chunks_szad, node);
 		if (xnode != NULL)
 			base_node_dealloc(xnode);
@@ -234,6 +238,7 @@ chunk_record(void *chunk, size_t size)
 		node = xnode;
 		node->addr = chunk;
 		node->size = size;
+		node->zeroed = (unzeroed == false);
 		extent_tree_ad_insert(&chunks_ad, node);
 		extent_tree_szad_insert(&chunks_szad, node);
 	}
@@ -253,6 +258,7 @@ chunk_record(void *chunk, size_t size)
 		extent_tree_szad_remove(&chunks_szad, node);
 		node->addr = prev->addr;
 		node->size += prev->size;
+		node->zeroed = (node->zeroed && prev->zeroed);
 		extent_tree_szad_insert(&chunks_szad, node);
 
 		base_node_dealloc(prev);
@@ -312,3 +318,33 @@ chunk_boot(void)
 
 	return (false);
 }
+
+void
+chunk_prefork(void)
+{
+
+	malloc_mutex_lock(&chunks_mtx);
+	if (config_ivsalloc)
+		rtree_prefork(chunks_rtree);
+	chunk_dss_prefork();
+}
+
+void
+chunk_postfork_parent(void)
+{
+
+	chunk_dss_postfork_parent();
+	if (config_ivsalloc)
+		rtree_postfork_parent(chunks_rtree);
+	malloc_mutex_postfork_parent(&chunks_mtx);
+}
+
+void
+chunk_postfork_child(void)
+{
+
+	chunk_dss_postfork_child();
+	if (config_ivsalloc)
+		rtree_postfork_child(chunks_rtree);
+	malloc_mutex_postfork_child(&chunks_mtx);
+}

memory/jemalloc/src/src/chunk_mmap.c

@@ -113,22 +113,30 @@ pages_trim(void *addr, size_t alloc_size, size_t leadsize, size_t size)
 #endif
 }
 
-void
+bool
 pages_purge(void *addr, size_t length)
 {
+	bool unzeroed;
 
 #ifdef _WIN32
 	VirtualAlloc(addr, length, MEM_RESET, PAGE_READWRITE);
+	unzeroed = true;
 #else
 #  ifdef JEMALLOC_PURGE_MADVISE_DONTNEED
 #    define JEMALLOC_MADV_PURGE MADV_DONTNEED
+#    define JEMALLOC_MADV_ZEROS true
 #  elif defined(JEMALLOC_PURGE_MADVISE_FREE)
 #    define JEMALLOC_MADV_PURGE MADV_FREE
+#    define JEMALLOC_MADV_ZEROS false
 #  else
 #    error "No method defined for purging unused dirty pages."
 #  endif
-	madvise(addr, length, JEMALLOC_MADV_PURGE);
+	int err = madvise(addr, length, JEMALLOC_MADV_PURGE);
+	unzeroed = (JEMALLOC_MADV_ZEROS == false || err != 0);
+#  undef JEMALLOC_MADV_PURGE
+#  undef JEMALLOC_MADV_ZEROS
 #endif
+	return (unzeroed);
 }
 
 static void *

memory/jemalloc/src/src/ctl.c

@@ -827,6 +827,27 @@ ctl_boot(void)
 	return (false);
 }
 
+void
+ctl_prefork(void)
+{
+
+	malloc_mutex_lock(&ctl_mtx);
+}
+
+void
+ctl_postfork_parent(void)
+{
+
+	malloc_mutex_postfork_parent(&ctl_mtx);
+}
+
+void
+ctl_postfork_child(void)
+{
+
+	malloc_mutex_postfork_child(&ctl_mtx);
+}
+
 /******************************************************************************/
 /* *_ctl() functions. */
 
@@ -1032,8 +1053,8 @@ thread_tcache_enabled_ctl(const size_t *mib, size_t miblen, void *oldp,
 	}
 	READ(oldval, bool);
 
-label_return:
 	ret = 0;
+label_return:
 	return (ret);
 }
 

memory/jemalloc/src/src/jemalloc.c

@@ -254,12 +254,13 @@ malloc_ncpus(void)
 	result = si.dwNumberOfProcessors;
 #else
 	result = sysconf(_SC_NPROCESSORS_ONLN);
+#endif
 	if (result == -1) {
 		/* Error. */
 		ret = 1;
-	}
+	}  else {
-#endif
+    ret = (unsigned)result;
-	ret = (unsigned)result;
+  }
 
 	return (ret);
 }
@@ -377,6 +378,22 @@ malloc_conf_init(void)
 	const char *opts, *k, *v;
 	size_t klen, vlen;
 
+	/*
+	 * Automatically configure valgrind before processing options.  The
+	 * valgrind option remains in jemalloc 3.x for compatibility reasons.
+	 */
+	if (config_valgrind) {
+		opt_valgrind = (RUNNING_ON_VALGRIND != 0) ? true : false;
+		if (config_fill && opt_valgrind) {
+			opt_junk = false;
+			assert(opt_zero == false);
+			opt_quarantine = JEMALLOC_VALGRIND_QUARANTINE_DEFAULT;
+			opt_redzone = true;
+		}
+		if (config_tcache && opt_valgrind)
+			opt_tcache = false;
+	}
+
 	for (i = 0; i < 3; i++) {
 		/* Get runtime configuration. */
 		switch (i) {
@@ -553,20 +570,7 @@ malloc_conf_init(void)
 				CONF_HANDLE_BOOL(opt_utrace, "utrace")
 			}
 			if (config_valgrind) {
-				bool hit;
+				CONF_HANDLE_BOOL(opt_valgrind, "valgrind")
-				CONF_HANDLE_BOOL_HIT(opt_valgrind,
-				    "valgrind", hit)
-				if (config_fill && opt_valgrind && hit) {
-					opt_junk = false;
-					opt_zero = false;
-					if (opt_quarantine == 0) {
-						opt_quarantine =
-						    JEMALLOC_VALGRIND_QUARANTINE_DEFAULT;
-					}
-					opt_redzone = true;
-				}
-				if (hit)
-					continue;
 			}
 			if (config_xmalloc) {
 				CONF_HANDLE_BOOL(opt_xmalloc, "xmalloc")
@@ -1262,11 +1266,10 @@ je_valloc(size_t size)
  * passed an extra argument for the caller return address, which will be
  * ignored.
  */
-JEMALLOC_EXPORT void (* const __free_hook)(void *ptr) = je_free;
+JEMALLOC_EXPORT void (* __free_hook)(void *ptr) = je_free;
-JEMALLOC_EXPORT void *(* const __malloc_hook)(size_t size) = je_malloc;
+JEMALLOC_EXPORT void *(* __malloc_hook)(size_t size) = je_malloc;
-JEMALLOC_EXPORT void *(* const __realloc_hook)(void *ptr, size_t size) =
+JEMALLOC_EXPORT void *(* __realloc_hook)(void *ptr, size_t size) = je_realloc;
-    je_realloc;
+JEMALLOC_EXPORT void *(* __memalign_hook)(size_t alignment, size_t size) =
-JEMALLOC_EXPORT void *(* const __memalign_hook)(size_t alignment, size_t size) =
     je_memalign;
 #endif
 
@@ -1279,7 +1282,7 @@ JEMALLOC_EXPORT void *(* const __memalign_hook)(size_t alignment, size_t size) =
  */
 
 size_t
-je_malloc_usable_size(const void *ptr)
+je_malloc_usable_size(JEMALLOC_USABLE_SIZE_CONST void *ptr)
 {
 	size_t ret;
 
@@ -1611,6 +1614,27 @@ je_nallocm(size_t *rsize, size_t size, int flags)
  * malloc during fork().
  */
 
+/*
+ * If an application creates a thread before doing any allocation in the main
+ * thread, then calls fork(2) in the main thread followed by memory allocation
+ * in the child process, a race can occur that results in deadlock within the
+ * child: the main thread may have forked while the created thread had
+ * partially initialized the allocator.  Ordinarily jemalloc prevents
+ * fork/malloc races via the following functions it registers during
+ * initialization using pthread_atfork(), but of course that does no good if
+ * the allocator isn't fully initialized at fork time.  The following library
+ * constructor is a partial solution to this problem.  It may still possible to
+ * trigger the deadlock described above, but doing so would involve forking via
+ * a library constructor that runs before jemalloc's runs.
+ */
+JEMALLOC_ATTR(constructor)
+static void
+jemalloc_constructor(void)
+{
+
+	malloc_init();
+}
+
 #ifndef JEMALLOC_MUTEX_INIT_CB
 void
 jemalloc_prefork(void)
@@ -1628,14 +1652,16 @@ _malloc_prefork(void)
 	assert(malloc_initialized);
 
 	/* Acquire all mutexes in a safe order. */
+	ctl_prefork();
 	malloc_mutex_prefork(&arenas_lock);
 	for (i = 0; i < narenas; i++) {
 		if (arenas[i] != NULL)
 			arena_prefork(arenas[i]);
 	}
+	prof_prefork();
+	chunk_prefork();
 	base_prefork();
 	huge_prefork();
-	chunk_dss_prefork();
 }
 
 #ifndef JEMALLOC_MUTEX_INIT_CB
@@ -1655,14 +1681,16 @@ _malloc_postfork(void)
 	assert(malloc_initialized);
 
 	/* Release all mutexes, now that fork() has completed. */
-	chunk_dss_postfork_parent();
 	huge_postfork_parent();
 	base_postfork_parent();
+	chunk_postfork_parent();
+	prof_postfork_parent();
 	for (i = 0; i < narenas; i++) {
 		if (arenas[i] != NULL)
 			arena_postfork_parent(arenas[i]);
 	}
 	malloc_mutex_postfork_parent(&arenas_lock);
+	ctl_postfork_parent();
 }
 
 void
@@ -1673,14 +1701,16 @@ jemalloc_postfork_child(void)
 	assert(malloc_initialized);
 
 	/* Release all mutexes, now that fork() has completed. */
-	chunk_dss_postfork_child();
 	huge_postfork_child();
 	base_postfork_child();
+	chunk_postfork_child();
+	prof_postfork_child();
 	for (i = 0; i < narenas; i++) {
 		if (arenas[i] != NULL)
 			arena_postfork_child(arenas[i]);
 	}
 	malloc_mutex_postfork_child(&arenas_lock);
+	ctl_postfork_child();
 }
 
 /******************************************************************************/

memory/jemalloc/src/src/mutex.c

@@ -64,7 +64,7 @@ pthread_create(pthread_t *__restrict thread,
 /******************************************************************************/
 
 #ifdef JEMALLOC_MUTEX_INIT_CB
-int	_pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex,
+JEMALLOC_EXPORT int	_pthread_mutex_init_calloc_cb(pthread_mutex_t *mutex,
     void *(calloc_cb)(size_t, size_t));
 #endif
 

memory/jemalloc/src/src/prof.c

@@ -1270,4 +1270,46 @@ prof_boot2(void)
 	return (false);
 }
 
+void
+prof_prefork(void)
+{
+
+	if (opt_prof) {
+		unsigned i;
+
+		malloc_mutex_lock(&bt2ctx_mtx);
+		malloc_mutex_lock(&prof_dump_seq_mtx);
+		for (i = 0; i < PROF_NCTX_LOCKS; i++)
+			malloc_mutex_lock(&ctx_locks[i]);
+	}
+}
+
+void
+prof_postfork_parent(void)
+{
+
+	if (opt_prof) {
+		unsigned i;
+
+		for (i = 0; i < PROF_NCTX_LOCKS; i++)
+			malloc_mutex_postfork_parent(&ctx_locks[i]);
+		malloc_mutex_postfork_parent(&prof_dump_seq_mtx);
+		malloc_mutex_postfork_parent(&bt2ctx_mtx);
+	}
+}
+
+void
+prof_postfork_child(void)
+{
+
+	if (opt_prof) {
+		unsigned i;
+
+		for (i = 0; i < PROF_NCTX_LOCKS; i++)
+			malloc_mutex_postfork_child(&ctx_locks[i]);
+		malloc_mutex_postfork_child(&prof_dump_seq_mtx);
+		malloc_mutex_postfork_child(&bt2ctx_mtx);
+	}
+}
+
 /******************************************************************************/

memory/jemalloc/src/src/rtree.c

@@ -44,3 +44,24 @@ rtree_new(unsigned bits)
 
 	return (ret);
 }
+
+void
+rtree_prefork(rtree_t *rtree)
+{
+
+	malloc_mutex_prefork(&rtree->mutex);
+}
+
+void
+rtree_postfork_parent(rtree_t *rtree)
+{
+
+	malloc_mutex_postfork_parent(&rtree->mutex);
+}
+
+void
+rtree_postfork_child(rtree_t *rtree)
+{
+
+	malloc_mutex_postfork_child(&rtree->mutex);
+}

memory/jemalloc/src/src/util.c

@@ -377,7 +377,6 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 		case '\0': goto label_out;
 		case '%': {
 			bool alt_form = false;
-			bool zero_pad = false;
 			bool left_justify = false;
 			bool plus_space = false;
 			bool plus_plus = false;
@@ -398,10 +397,6 @@ malloc_vsnprintf(char *str, size_t size, const char *format, va_list ap)
 					assert(alt_form == false);
 					alt_form = true;
 					break;
-				case '0':
-					assert(zero_pad == false);
-					zero_pad = true;
-					break;
 				case '-':
 					assert(left_justify == false);
 					left_justify = true;

memory/jemalloc/upstream.info

@@ -1,2 +1,2 @@
 UPSTREAM_REPO=git://canonware.com/jemalloc.git
-UPSTREAM_COMMIT=3.0.0
+UPSTREAM_COMMIT=d0ffd8ed4f6aa4cf7248028eddfcb35f93247fe4