Merge branch 'dev'

ChangeLog

@@ -6,6 +6,19 @@ found in the git revision history:
     http://www.canonware.com/cgi-bin/gitweb.cgi?p=jemalloc.git
     git://canonware.com/jemalloc.git
 
+* 3.4.0 (June 2, 2013)
+
+  This version is essentially a small bugfix release, but the addition of
+  aarch64 support requires that the minor version be incremented.
+
+  Bug fixes:
+  - Fix race-triggered deadlocks in chunk_record().  These deadlocks were
+    typically triggered by multiple threads concurrently deallocating huge
+    objects.
+
+  New features:
+  - Add support for the aarch64 architecture.
+
 * 3.3.1 (March 6, 2013)
 
   This version fixes bugs that are typically encountered only when utilizing

doc/jemalloc.xml.in

@@ -432,7 +432,14 @@ for (i = 0; i < nbins; i++) {
     referenced by the symbolic link named <filename
     class="symlink">/etc/malloc.conf</filename>, and the value of the
     environment variable <envar>MALLOC_CONF</envar>, will be interpreted, in
-    that order, from left to right as options.</para>
+    that order, from left to right as options.  Note that
+    <varname>malloc_conf</varname> may be read before
+    <function>main<parameter/></function> is entered, so the declaration of
+    <varname>malloc_conf</varname> should specify an initializer that contains
+    the final value to be read by jemalloc.  <varname>malloc_conf</varname> is
+    a compile-time setting, whereas <filename
+    class="symlink">/etc/malloc.conf</filename> and <envar>MALLOC_CONF</envar>
+    can be safely set any time prior to program invocation.</para>
 
     <para>An options string is a comma-separated list of option:value pairs.
     There is one key corresponding to each <link

include/jemalloc/internal/jemalloc_internal.h.in

@@ -278,6 +278,9 @@ static const bool config_ivsalloc =
 #  ifdef __arm__
 #    define LG_QUANTUM		3
 #  endif
+#  ifdef __aarch64__
+#    define LG_QUANTUM		4
+#  endif
 #  ifdef __hppa__
 #    define LG_QUANTUM		4
 #  endif

src/chunk.c

@@ -214,7 +214,7 @@ chunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,
     size_t size)
 {
 	bool unzeroed;
-	extent_node_t *xnode, *node, *prev, key;
+	extent_node_t *xnode, *node, *prev, *xprev, key;
 
 	unzeroed = pages_purge(chunk, size);
 	VALGRIND_MAKE_MEM_NOACCESS(chunk, size);
@@ -226,6 +226,8 @@ chunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,
 	 * held.
 	 */
 	xnode = base_node_alloc();
+	/* Use xprev to implement conditional deferred deallocation of prev. */
+	xprev = NULL;
 
 	malloc_mutex_lock(&chunks_mtx);
 	key.addr = (void *)((uintptr_t)chunk + size);
@@ -242,8 +244,6 @@ chunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,
 		node->size += size;
 		node->zeroed = (node->zeroed && (unzeroed == false));
 		extent_tree_szad_insert(chunks_szad, node);
-		if (xnode != NULL)
-			base_node_dealloc(xnode);
 	} else {
 		/* Coalescing forward failed, so insert a new node. */
 		if (xnode == NULL) {
@@ -253,10 +253,10 @@ chunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,
 			 * already been purged, so this is only a virtual
 			 * memory leak.
 			 */
-			malloc_mutex_unlock(&chunks_mtx);
+			goto label_return;
-			return;
 		}
 		node = xnode;
+		xnode = NULL; /* Prevent deallocation below. */
 		node->addr = chunk;
 		node->size = size;
 		node->zeroed = (unzeroed == false);
@@ -282,9 +282,19 @@ chunk_record(extent_tree_t *chunks_szad, extent_tree_t *chunks_ad, void *chunk,
 		node->zeroed = (node->zeroed && prev->zeroed);
 		extent_tree_szad_insert(chunks_szad, node);
 
-		base_node_dealloc(prev);
+		xprev = prev;
 	}
+
+label_return:
 	malloc_mutex_unlock(&chunks_mtx);
+	/*
+	 * Deallocate xnode and/or xprev after unlocking chunks_mtx in order to
+	 * avoid potential deadlock.
+	 */
+	if (xnode != NULL)
+		base_node_dealloc(xnode);
+	if (xprev != NULL)
+		base_node_dealloc(prev);
 }
 
 void