Merge branch 'dev'

ChangeLog

@@ -4,6 +4,63 @@ brevity.  Much more detail can be found in the git revision history:
 
     https://github.com/jemalloc/jemalloc
 
+* 4.0.1 (September 15, 2015)
+
+  This is a bugfix release that is somewhat high risk due to the amount of
+  refactoring required to address deep xallocx() problems.  As a side effect of
+  these fixes, xallocx() now tries harder to partially fulfill requests for
+  optional extra space.  Note that a couple of minor heap profiling
+  optimizations are included, but these are better thought of as performance
+  fixes that were integral to disovering most of the other bugs.
+
+  Optimizations:
+  - Avoid a chunk metadata read in arena_prof_tctx_set(), since it is in the
+    fast path when heap profiling is enabled.  Additionally, split a special
+    case out into arena_prof_tctx_reset(), which also avoids chunk metadata
+    reads.
+  - Optimize irallocx_prof() to optimistically update the sampler state.  The
+    prior implementation appears to have been a holdover from when
+    rallocx()/xallocx() functionality was combined as rallocm().
+
+  Bug fixes:
+  - Fix TLS configuration such that it is enabled by default for platforms on
+    which it works correctly.
+  - Fix arenas_cache_cleanup() and arena_get_hard() to handle
+    allocation/deallocation within the application's thread-specific data
+    cleanup functions even after arenas_cache is torn down.
+  - Fix xallocx() bugs related to size+extra exceeding HUGE_MAXCLASS.
+  - Fix chunk purge hook calls for in-place huge shrinking reallocation to
+    specify the old chunk size rather than the new chunk size.  This bug caused
+    no correctness issues for the default chunk purge function, but was
+    visible to custom functions set via the "arena.<i>.chunk_hooks" mallctl.
+  - Fix heap profiling bugs:
+    + Fix heap profiling to distinguish among otherwise identical sample sites
+      with interposed resets (triggered via the "prof.reset" mallctl).  This bug
+      could cause data structure corruption that would most likely result in a
+      segfault.
+    + Fix irealloc_prof() to prof_alloc_rollback() on OOM.
+    + Make one call to prof_active_get_unlocked() per allocation event, and use
+      the result throughout the relevant functions that handle an allocation
+      event.  Also add a missing check in prof_realloc().  These fixes protect
+      allocation events against concurrent prof_active changes.
+    + Fix ixallocx_prof() to pass usize_max and zero to ixallocx_prof_sample()
+      in the correct order.
+    + Fix prof_realloc() to call prof_free_sampled_object() after calling
+      prof_malloc_sample_object().  Prior to this fix, if tctx and old_tctx were
+      the same, the tctx could have been prematurely destroyed.
+  - Fix portability bugs:
+    + Don't bitshift by negative amounts when encoding/decoding run sizes in
+      chunk header maps.  This affected systems with page sizes greater than 8
+      KiB.
+    + Rename index_t to szind_t to avoid an existing type on Solaris.
+    + Add JEMALLOC_CXX_THROW to the memalign() function prototype, in order to
+      match glibc and avoid compilation errors when including both
+      jemalloc/jemalloc.h and malloc.h in C++ code.
+    + Don't assume that /bin/sh is appropriate when running size_classes.sh
+      during configuration.
+    + Consider __sparcv9 a synonym for __sparc64__ when defining LG_QUANTUM.
+    + Link tests to librt if it contains clock_gettime(2).
+
 * 4.0.0 (August 17, 2015)
 
   This version contains many speed and space optimizations, both minor and

Makefile.in

@@ -28,6 +28,7 @@ CFLAGS := @CFLAGS@
 LDFLAGS := @LDFLAGS@
 EXTRA_LDFLAGS := @EXTRA_LDFLAGS@
 LIBS := @LIBS@
+TESTLIBS := @TESTLIBS@
 RPATH_EXTRA := @RPATH_EXTRA@
 SO := @so@
 IMPORTLIB := @importlib@
@@ -265,15 +266,15 @@ $(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 $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(TESTLIBS) $(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 $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(filter -lpthread,$(LIBS))) -lm $(TESTLIBS) $(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 $(EXTRA_LDFLAGS)
+	$(CC) $(LDTARGET) $(filter %.$(O),$^) $(call RPATH,$(objroot)lib) $(objroot)lib/$(LIBJEMALLOC).$(IMPORTLIB) $(LDFLAGS) $(filter-out -lm,$(LIBS)) -lm $(TESTLIBS) $(EXTRA_LDFLAGS)
 
 build_lib_shared: $(DSOS)
 build_lib_static: $(STATIC_LIBS)
@@ -343,9 +344,9 @@ check_unit_dir:
 	@mkdir -p $(objroot)test/unit
 check_integration_dir:
 	@mkdir -p $(objroot)test/integration
-check_stress_dir:
+stress_dir:
 	@mkdir -p $(objroot)test/stress
-check_dir: check_unit_dir check_integration_dir check_stress_dir
+check_dir: check_unit_dir check_integration_dir
 
 check_unit: tests_unit check_unit_dir
 	$(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%)
@@ -355,10 +356,10 @@ ifeq ($(enable_prof), 1)
 endif
 check_integration: tests_integration check_integration_dir
 	$(SHELL) $(objroot)test/test.sh $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
-check_stress: tests_stress check_stress_dir
+stress: tests_stress stress_dir
 	$(SHELL) $(objroot)test/test.sh $(TESTS_STRESS:$(srcroot)%.c=$(objroot)%)
 check: tests check_dir check_integration_prof
-	$(SHELL) $(objroot)test/test.sh $(TESTS:$(srcroot)%.c=$(objroot)%)
+	$(SHELL) $(objroot)test/test.sh $(TESTS_UNIT:$(srcroot)%.c=$(objroot)%) $(TESTS_INTEGRATION:$(srcroot)%.c=$(objroot)%)
 
 ifeq ($(enable_code_coverage), 1)
 coverage_unit: check_unit
@@ -372,7 +373,7 @@ coverage_integration: check_integration
 	$(SHELL) $(srcroot)coverage.sh $(srcroot)test/src integration $(C_TESTLIB_INTEGRATION_OBJS)
 	$(SHELL) $(srcroot)coverage.sh $(srcroot)test/integration integration $(TESTS_INTEGRATION_OBJS)
 
-coverage_stress: check_stress
+coverage_stress: stress
 	$(SHELL) $(srcroot)coverage.sh $(srcroot)src pic $(C_PIC_OBJS)
 	$(SHELL) $(srcroot)coverage.sh $(srcroot)src jet $(C_JET_OBJS)
 	$(SHELL) $(srcroot)coverage.sh $(srcroot)test/src stress $(C_TESTLIB_STRESS_OBJS)

configure.ac

@@ -1190,6 +1190,14 @@ 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="${SAVED_LIBS}"
+
 dnl Check if the GNU-specific secure_getenv function exists.
 AC_CHECK_FUNC([secure_getenv],
               [have_secure_getenv="1"],
@@ -1272,13 +1280,16 @@ fi
 ,
 enable_tls=""
 )
-if test "x${enable_tls}" = "x" -a "x${force_tls}" = "x1" ; then
+if test "x${enable_tls}" = "x" ; then
+  if test "x${force_tls}" = "x1" ; then
     AC_MSG_RESULT([Forcing TLS to avoid allocator/threading bootstrap issues])
     enable_tls="1"
-fi
+  elif test "x${force_tls}" = "x0" ; then
-if test "x${enable_tls}" = "x" -a "x${force_tls}" = "x0" ; then
     AC_MSG_RESULT([Forcing no TLS to avoid allocator/threading bootstrap issues])
     enable_tls="0"
+  else
+    enable_tls="1"
+  fi
 fi
 if test "x${enable_tls}" = "x1" ; then
 AC_MSG_CHECKING([for TLS])
@@ -1298,9 +1309,12 @@ else
 fi
 AC_SUBST([enable_tls])
 if test "x${enable_tls}" = "x1" ; then
+  if test "x${force_tls}" = "x0" ; then
+    AC_MSG_WARN([TLS enabled despite being marked unusable on this platform])
+  fi
   AC_DEFINE_UNQUOTED([JEMALLOC_TLS], [ ])
 elif test "x${force_tls}" = "x1" ; then
-  AC_MSG_ERROR([Failed to configure TLS, which is mandatory for correct function])
+  AC_MSG_WARN([TLS disabled despite being marked critical on this platform])
 fi
 
 dnl ============================================================================
@@ -1615,8 +1629,9 @@ AC_CONFIG_COMMANDS([include/jemalloc/internal/public_unnamespace.h], [
 ])
 AC_CONFIG_COMMANDS([include/jemalloc/internal/size_classes.h], [
   mkdir -p "${objroot}include/jemalloc/internal"
-  "${srcdir}/include/jemalloc/internal/size_classes.sh" "${LG_QUANTA}" ${LG_TINY_MIN} "${LG_PAGE_SIZES}" ${LG_SIZE_CLASS_GROUP} > "${objroot}include/jemalloc/internal/size_classes.h"
+  "${SHELL}" "${srcdir}/include/jemalloc/internal/size_classes.sh" "${LG_QUANTA}" ${LG_TINY_MIN} "${LG_PAGE_SIZES}" ${LG_SIZE_CLASS_GROUP} > "${objroot}include/jemalloc/internal/size_classes.h"
 ], [
+  SHELL="${SHELL}"
   srcdir="${srcdir}"
   objroot="${objroot}"
   LG_QUANTA="${LG_QUANTA}"
@@ -1687,6 +1702,7 @@ 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}])

include/jemalloc/internal/arena.h

@@ -39,7 +39,7 @@ typedef struct arena_s arena_t;
 #ifdef JEMALLOC_ARENA_STRUCTS_A
 struct arena_run_s {
 	/* Index of bin this run is associated with. */
-	index_t		binind;
+	szind_t		binind;
 
 	/* Number of free regions in run. */
 	unsigned	nfree;
@@ -424,7 +424,7 @@ extern arena_bin_info_t	arena_bin_info[NBINS];
 extern size_t		map_bias; /* Number of arena chunk header pages. */
 extern size_t		map_misc_offset;
 extern size_t		arena_maxrun; /* Max run size for arenas. */
-extern size_t		arena_maxclass; /* Max size class for arenas. */
+extern size_t		large_maxclass; /* Max large size class. */
 extern unsigned		nlclasses; /* Number of large size classes. */
 extern unsigned		nhclasses; /* Number of huge size classes. */
 
@@ -448,7 +448,7 @@ bool	arena_lg_dirty_mult_set(arena_t *arena, ssize_t lg_dirty_mult);
 void	arena_maybe_purge(arena_t *arena);
 void	arena_purge_all(arena_t *arena);
 void	arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin,
-    index_t binind, uint64_t prof_accumbytes);
+    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
@@ -488,7 +488,7 @@ extern arena_ralloc_junk_large_t *arena_ralloc_junk_large;
 bool	arena_ralloc_no_move(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 extra, size_t alignment, bool zero, tcache_t *tcache);
+    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);
 ssize_t	arena_lg_dirty_mult_default_get(void);
@@ -519,17 +519,19 @@ 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 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);
-index_t	arena_mapbits_binind_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);
 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,
     size_t size, size_t flags);
 void	arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
@@ -539,21 +541,23 @@ void	arena_mapbits_internal_set(arena_chunk_t *chunk, size_t pageind,
 void	arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind,
     size_t size, size_t flags);
 void	arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
-    index_t binind);
+    szind_t binind);
 void	arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind,
-    size_t runind, index_t binind, size_t flags);
+    size_t runind, szind_t binind, size_t flags);
 void	arena_metadata_allocated_add(arena_t *arena, size_t size);
 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);
-index_t	arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
+szind_t	arena_ptr_small_binind_get(const void *ptr, size_t mapbits);
-index_t	arena_bin_index(arena_t *arena, arena_bin_t *bin);
+szind_t	arena_bin_index(arena_t *arena, arena_bin_t *bin);
 unsigned	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, prof_tctx_t *tctx);
+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,
+    const void *old_ptr, prof_tctx_t *old_tctx);
 void	*arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
     tcache_t *tcache);
 arena_t	*arena_aalloc(const void *ptr);
@@ -652,6 +656,22 @@ arena_mapbits_get(arena_chunk_t *chunk, size_t pageind)
 	return (arena_mapbitsp_read(arena_mapbitsp_get(chunk, pageind)));
 }
 
+JEMALLOC_ALWAYS_INLINE size_t
+arena_mapbits_size_decode(size_t mapbits)
+{
+	size_t size;
+
+#if CHUNK_MAP_SIZE_SHIFT > 0
+	size = (mapbits & CHUNK_MAP_SIZE_MASK) >> CHUNK_MAP_SIZE_SHIFT;
+#elif CHUNK_MAP_SIZE_SHIFT == 0
+	size = mapbits & CHUNK_MAP_SIZE_MASK;
+#else
+	size = (mapbits & CHUNK_MAP_SIZE_MASK) << -CHUNK_MAP_SIZE_SHIFT;
+#endif
+
+	return (size);
+}
+
 JEMALLOC_ALWAYS_INLINE size_t
 arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
 {
@@ -659,7 +679,7 @@ arena_mapbits_unallocated_size_get(arena_chunk_t *chunk, size_t pageind)
 
 	mapbits = arena_mapbits_get(chunk, pageind);
 	assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);
-	return ((mapbits & CHUNK_MAP_SIZE_MASK) >> CHUNK_MAP_SIZE_SHIFT);
+	return (arena_mapbits_size_decode(mapbits));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
@@ -670,7 +690,7 @@ arena_mapbits_large_size_get(arena_chunk_t *chunk, size_t pageind)
 	mapbits = arena_mapbits_get(chunk, pageind);
 	assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) ==
 	    (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED));
-	return ((mapbits & CHUNK_MAP_SIZE_MASK) >> CHUNK_MAP_SIZE_SHIFT);
+	return (arena_mapbits_size_decode(mapbits));
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
@@ -684,11 +704,11 @@ arena_mapbits_small_runind_get(arena_chunk_t *chunk, size_t pageind)
 	return (mapbits >> CHUNK_MAP_RUNIND_SHIFT);
 }
 
-JEMALLOC_ALWAYS_INLINE index_t
+JEMALLOC_ALWAYS_INLINE szind_t
 arena_mapbits_binind_get(arena_chunk_t *chunk, size_t pageind)
 {
 	size_t mapbits;
-	index_t binind;
+	szind_t binind;
 
 	mapbits = arena_mapbits_get(chunk, pageind);
 	binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;
@@ -754,6 +774,23 @@ arena_mapbitsp_write(size_t *mapbitsp, size_t mapbits)
 	*mapbitsp = mapbits;
 }
 
+JEMALLOC_ALWAYS_INLINE size_t
+arena_mapbits_size_encode(size_t size)
+{
+	size_t mapbits;
+
+#if CHUNK_MAP_SIZE_SHIFT > 0
+	mapbits = size << CHUNK_MAP_SIZE_SHIFT;
+#elif CHUNK_MAP_SIZE_SHIFT == 0
+	mapbits = size;
+#else
+	mapbits = size >> -CHUNK_MAP_SIZE_SHIFT;
+#endif
+
+	assert((mapbits & ~CHUNK_MAP_SIZE_MASK) == 0);
+	return (mapbits);
+}
+
 JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
     size_t flags)
@@ -761,11 +798,10 @@ arena_mapbits_unallocated_set(arena_chunk_t *chunk, size_t pageind, size_t size,
 	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
 
 	assert((size & PAGE_MASK) == 0);
-	assert(((size << CHUNK_MAP_SIZE_SHIFT) & ~CHUNK_MAP_SIZE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
 	assert((flags & CHUNK_MAP_DECOMMITTED) == 0 || (flags &
 	    (CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
-	arena_mapbitsp_write(mapbitsp, (size << CHUNK_MAP_SIZE_SHIFT) |
+	arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
 	    CHUNK_MAP_BININD_INVALID | flags);
 }
 
@@ -777,10 +813,9 @@ arena_mapbits_unallocated_size_set(arena_chunk_t *chunk, size_t pageind,
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);
 
 	assert((size & PAGE_MASK) == 0);
-	assert(((size << CHUNK_MAP_SIZE_SHIFT) & ~CHUNK_MAP_SIZE_MASK) == 0);
 	assert((mapbits & (CHUNK_MAP_LARGE|CHUNK_MAP_ALLOCATED)) == 0);
-	arena_mapbitsp_write(mapbitsp, (size << CHUNK_MAP_SIZE_SHIFT) | (mapbits
+	arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
-	    & ~CHUNK_MAP_SIZE_MASK));
+	    (mapbits & ~CHUNK_MAP_SIZE_MASK));
 }
 
 JEMALLOC_ALWAYS_INLINE void
@@ -799,18 +834,17 @@ arena_mapbits_large_set(arena_chunk_t *chunk, size_t pageind, size_t size,
 	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
 
 	assert((size & PAGE_MASK) == 0);
-	assert(((size << CHUNK_MAP_SIZE_SHIFT) & ~CHUNK_MAP_SIZE_MASK) == 0);
 	assert((flags & CHUNK_MAP_FLAGS_MASK) == flags);
 	assert((flags & CHUNK_MAP_DECOMMITTED) == 0 || (flags &
 	    (CHUNK_MAP_DIRTY|CHUNK_MAP_UNZEROED)) == 0);
-	arena_mapbitsp_write(mapbitsp, (size << CHUNK_MAP_SIZE_SHIFT) |
+	arena_mapbitsp_write(mapbitsp, arena_mapbits_size_encode(size) |
 	    CHUNK_MAP_BININD_INVALID | flags | CHUNK_MAP_LARGE |
 	    CHUNK_MAP_ALLOCATED);
 }
 
 JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
-    index_t binind)
+    szind_t binind)
 {
 	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
 	size_t mapbits = arena_mapbitsp_read(mapbitsp);
@@ -824,7 +858,7 @@ arena_mapbits_large_binind_set(arena_chunk_t *chunk, size_t pageind,
 
 JEMALLOC_ALWAYS_INLINE void
 arena_mapbits_small_set(arena_chunk_t *chunk, size_t pageind, size_t runind,
-    index_t binind, size_t flags)
+    szind_t binind, size_t flags)
 {
 	size_t *mapbitsp = arena_mapbitsp_get(chunk, pageind);
 
@@ -901,10 +935,10 @@ arena_prof_accum(arena_t *arena, uint64_t accumbytes)
 	}
 }
 
-JEMALLOC_ALWAYS_INLINE index_t
+JEMALLOC_ALWAYS_INLINE szind_t
 arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 {
-	index_t binind;
+	szind_t binind;
 
 	binind = (mapbits & CHUNK_MAP_BININD_MASK) >> CHUNK_MAP_BININD_SHIFT;
 
@@ -916,7 +950,7 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 		size_t rpages_ind;
 		arena_run_t *run;
 		arena_bin_t *bin;
-		index_t run_binind, actual_binind;
+		szind_t run_binind, actual_binind;
 		arena_bin_info_t *bin_info;
 		arena_chunk_map_misc_t *miscelm;
 		void *rpages;
@@ -950,10 +984,10 @@ arena_ptr_small_binind_get(const void *ptr, size_t mapbits)
 #  endif /* JEMALLOC_ARENA_INLINE_A */
 
 #  ifdef JEMALLOC_ARENA_INLINE_B
-JEMALLOC_INLINE index_t
+JEMALLOC_INLINE szind_t
 arena_bin_index(arena_t *arena, arena_bin_t *bin)
 {
-	index_t binind = bin - arena->bins;
+	szind_t binind = bin - arena->bins;
 	assert(binind < NBINS);
 	return (binind);
 }
@@ -1060,7 +1094,7 @@ arena_prof_tctx_get(const void *ptr)
 }
 
 JEMALLOC_INLINE void
-arena_prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
+arena_prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
 {
 	arena_chunk_t *chunk;
 
@@ -1070,17 +1104,59 @@ arena_prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 	if (likely(chunk != ptr)) {
 		size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
+
 		assert(arena_mapbits_allocated_get(chunk, pageind) != 0);
 
-		if (unlikely(arena_mapbits_large_get(chunk, pageind) != 0)) {
+		if (unlikely(usize > SMALL_MAXCLASS || (uintptr_t)tctx >
-			arena_chunk_map_misc_t *elm = arena_miscelm_get(chunk,
+		    (uintptr_t)1U)) {
-			    pageind);
+			arena_chunk_map_misc_t *elm;
+
+			assert(arena_mapbits_large_get(chunk, pageind) != 0);
+
+			elm = arena_miscelm_get(chunk, pageind);
 			atomic_write_p(&elm->prof_tctx_pun, tctx);
+		} else {
+			/*
+			 * tctx must always be initialized for large runs.
+			 * Assert that the surrounding conditional logic is
+			 * equivalent to checking whether ptr refers to a large
+			 * run.
+			 */
+			assert(arena_mapbits_large_get(chunk, pageind) == 0);
 		}
 	} else
 		huge_prof_tctx_set(ptr, tctx);
 }
 
+JEMALLOC_INLINE void
+arena_prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
+    prof_tctx_t *old_tctx)
+{
+
+	cassert(config_prof);
+	assert(ptr != NULL);
+
+	if (unlikely(usize > SMALL_MAXCLASS || (ptr == old_ptr &&
+	    (uintptr_t)old_tctx > (uintptr_t)1U))) {
+		arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
+		if (likely(chunk != ptr)) {
+			size_t pageind;
+			arena_chunk_map_misc_t *elm;
+
+			pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >>
+			    LG_PAGE;
+			assert(arena_mapbits_allocated_get(chunk, pageind) !=
+			    0);
+			assert(arena_mapbits_large_get(chunk, pageind) != 0);
+
+			elm = arena_miscelm_get(chunk, pageind);
+			atomic_write_p(&elm->prof_tctx_pun,
+			    (prof_tctx_t *)(uintptr_t)1U);
+		} else
+			huge_prof_tctx_reset(ptr);
+	}
+}
+
 JEMALLOC_ALWAYS_INLINE void *
 arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
     tcache_t *tcache)
@@ -1098,7 +1174,7 @@ arena_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
 			    zero));
 		} else
 			return (arena_malloc_small(arena, size, zero));
-	} else if (likely(size <= arena_maxclass)) {
+	} else if (likely(size <= large_maxclass)) {
 		/*
 		 * Initialize tcache after checking size in order to avoid
 		 * infinite recursion during tcache initialization.
@@ -1131,7 +1207,7 @@ arena_salloc(const void *ptr, bool demote)
 	size_t ret;
 	arena_chunk_t *chunk;
 	size_t pageind;
-	index_t binind;
+	szind_t binind;
 
 	assert(ptr != NULL);
 
@@ -1190,7 +1266,7 @@ arena_dalloc(tsd_t *tsd, void *ptr, tcache_t *tcache)
 		if (likely((mapbits & CHUNK_MAP_LARGE) == 0)) {
 			/* Small allocation. */
 			if (likely(tcache != NULL)) {
-				index_t binind = arena_ptr_small_binind_get(ptr,
+				szind_t binind = arena_ptr_small_binind_get(ptr,
 				    mapbits);
 				tcache_dalloc_small(tsd, tcache, ptr, binind);
 			} else {
@@ -1242,7 +1318,7 @@ arena_sdalloc(tsd_t *tsd, void *ptr, size_t size, tcache_t *tcache)
 		if (likely(size <= SMALL_MAXCLASS)) {
 			/* Small allocation. */
 			if (likely(tcache != NULL)) {
-				index_t binind = size2index(size);
+				szind_t binind = size2index(size);
 				tcache_dalloc_small(tsd, tcache, ptr, binind);
 			} else {
 				size_t pageind = ((uintptr_t)ptr -

include/jemalloc/internal/huge.h

@@ -13,11 +13,10 @@ void	*huge_malloc(tsd_t *tsd, arena_t *arena, size_t size, bool zero,
     tcache_t *tcache);
 void	*huge_palloc(tsd_t *tsd, arena_t *arena, size_t size, size_t alignment,
     bool zero, tcache_t *tcache);
-bool	huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size,
+bool	huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
-    size_t extra, bool zero);
+    size_t usize_max, bool zero);
 void	*huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize,
-    size_t size, size_t extra, size_t alignment, bool zero,
+    size_t usize, size_t alignment, bool zero, tcache_t *tcache);
-    tcache_t *tcache);
 #ifdef JEMALLOC_JET
 typedef void (huge_dalloc_junk_t)(void *, size_t);
 extern huge_dalloc_junk_t *huge_dalloc_junk;
@@ -27,6 +26,7 @@ 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);
 
 #endif /* JEMALLOC_H_EXTERNS */
 /******************************************************************************/

include/jemalloc/internal/jemalloc_internal.h.in

@@ -184,7 +184,7 @@ static const bool config_cache_oblivious =
 #include "jemalloc/internal/jemalloc_internal_macros.h"
 
 /* Size class index type. */
-typedef unsigned index_t;
+typedef unsigned szind_t;
 
 /*
  * Flags bits:
@@ -232,7 +232,7 @@ typedef unsigned index_t;
 #  ifdef __alpha__
 #    define LG_QUANTUM		4
 #  endif
-#  ifdef __sparc64__
+#  if (defined(__sparc64__) || defined(__sparcv9))
 #    define LG_QUANTUM		4
 #  endif
 #  if (defined(__amd64__) || defined(__x86_64__) || defined(_M_X64))
@@ -511,12 +511,12 @@ void	jemalloc_postfork_child(void);
 #include "jemalloc/internal/huge.h"
 
 #ifndef JEMALLOC_ENABLE_INLINE
-index_t	size2index_compute(size_t size);
+szind_t	size2index_compute(size_t size);
-index_t	size2index_lookup(size_t size);
+szind_t	size2index_lookup(size_t size);
-index_t	size2index(size_t size);
+szind_t	size2index(size_t size);
-size_t	index2size_compute(index_t index);
+size_t	index2size_compute(szind_t index);
-size_t	index2size_lookup(index_t index);
+size_t	index2size_lookup(szind_t index);
-size_t	index2size(index_t index);
+size_t	index2size(szind_t index);
 size_t	s2u_compute(size_t size);
 size_t	s2u_lookup(size_t size);
 size_t	s2u(size_t size);
@@ -527,7 +527,7 @@ arena_t	*arena_get(tsd_t *tsd, unsigned ind, bool init_if_missing,
 #endif
 
 #if (defined(JEMALLOC_ENABLE_INLINE) || defined(JEMALLOC_C_))
-JEMALLOC_INLINE index_t
+JEMALLOC_INLINE szind_t
 size2index_compute(size_t size)
 {
 
@@ -558,7 +558,7 @@ size2index_compute(size_t size)
 	}
 }
 
-JEMALLOC_ALWAYS_INLINE index_t
+JEMALLOC_ALWAYS_INLINE szind_t
 size2index_lookup(size_t size)
 {
 
@@ -571,7 +571,7 @@ size2index_lookup(size_t size)
 	}
 }
 
-JEMALLOC_ALWAYS_INLINE index_t
+JEMALLOC_ALWAYS_INLINE szind_t
 size2index(size_t size)
 {
 
@@ -582,7 +582,7 @@ size2index(size_t size)
 }
 
 JEMALLOC_INLINE size_t
-index2size_compute(index_t index)
+index2size_compute(szind_t index)
 {
 
 #if (NTBINS > 0)
@@ -609,7 +609,7 @@ index2size_compute(index_t index)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-index2size_lookup(index_t index)
+index2size_lookup(szind_t index)
 {
 	size_t ret = (size_t)index2size_tab[index];
 	assert(ret == index2size_compute(index));
@@ -617,7 +617,7 @@ index2size_lookup(index_t index)
 }
 
 JEMALLOC_ALWAYS_INLINE size_t
-index2size(index_t index)
+index2size(szind_t index)
 {
 
 	assert(index < NSIZES);
@@ -705,7 +705,7 @@ sa2u(size_t size, size_t alignment)
 	}
 
 	/* Try for a large size class. */
-	if (likely(size <= arena_maxclass) && likely(alignment < chunksize)) {
+	if (likely(size <= large_maxclass) && likely(alignment < chunksize)) {
 		/*
 		 * We can't achieve subpage alignment, so round up alignment
 		 * to the minimum that can actually be supported.
@@ -976,7 +976,7 @@ u2rz(size_t usize)
 	size_t ret;
 
 	if (usize <= SMALL_MAXCLASS) {
-		index_t binind = size2index(usize);
+		szind_t binind = size2index(usize);
 		ret = arena_bin_info[binind].redzone_size;
 	} else
 		ret = 0;
@@ -1096,7 +1096,7 @@ iralloct(tsd_t *tsd, void *ptr, size_t oldsize, size_t size, size_t alignment,
 		    zero, tcache, arena));
 	}
 
-	return (arena_ralloc(tsd, arena, ptr, oldsize, size, 0, alignment, zero,
+	return (arena_ralloc(tsd, arena, ptr, oldsize, size, alignment, zero,
 	    tcache));
 }
 

include/jemalloc/internal/private_symbols.txt

@@ -50,13 +50,14 @@ arena_mapbits_large_size_get
 arena_mapbitsp_get
 arena_mapbitsp_read
 arena_mapbitsp_write
+arena_mapbits_size_decode
+arena_mapbits_size_encode
 arena_mapbits_small_runind_get
 arena_mapbits_small_set
 arena_mapbits_unallocated_set
 arena_mapbits_unallocated_size_get
 arena_mapbits_unallocated_size_set
 arena_mapbits_unzeroed_get
-arena_maxclass
 arena_maxrun
 arena_maybe_purge
 arena_metadata_allocated_add
@@ -79,6 +80,7 @@ arena_prof_accum_impl
 arena_prof_accum_locked
 arena_prof_promoted
 arena_prof_tctx_get
+arena_prof_tctx_reset
 arena_prof_tctx_set
 arena_ptr_small_binind_get
 arena_purge_all
@@ -249,6 +251,7 @@ huge_dalloc_junk
 huge_malloc
 huge_palloc
 huge_prof_tctx_get
+huge_prof_tctx_reset
 huge_prof_tctx_set
 huge_ralloc
 huge_ralloc_no_move
@@ -283,6 +286,7 @@ ixalloc
 jemalloc_postfork_child
 jemalloc_postfork_parent
 jemalloc_prefork
+large_maxclass
 lg_floor
 malloc_cprintf
 malloc_mutex_init
@@ -377,6 +381,7 @@ prof_reset
 prof_sample_accum_update
 prof_sample_threshold_update
 prof_tctx_get
+prof_tctx_reset
 prof_tctx_set
 prof_tdata_cleanup
 prof_tdata_get

include/jemalloc/internal/prof.h

@@ -90,10 +90,11 @@ struct prof_tctx_s {
 	prof_tdata_t		*tdata;
 
 	/*
-	 * Copy of tdata->thr_uid, necessary because tdata may be defunct during
+	 * Copy of tdata->thr_{uid,discrim}, necessary because tdata may be
-	 * teardown.
+	 * defunct during teardown.
 	 */
 	uint64_t		thr_uid;
+	uint64_t		thr_discrim;
 
 	/* Profiling counters, protected by tdata->lock. */
 	prof_cnt_t		cnts;
@@ -330,14 +331,18 @@ bool	prof_gdump_get_unlocked(void);
 prof_tdata_t	*prof_tdata_get(tsd_t *tsd, bool create);
 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 update);
+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, prof_tctx_t *tctx);
+void	prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx);
+void	prof_tctx_reset(const void *ptr, size_t usize, const void *old_ptr,
+    prof_tctx_t *tctx);
 void	prof_malloc_sample_object(const void *ptr, size_t usize,
     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 updated, size_t old_usize, prof_tctx_t *old_tctx);
+    prof_tctx_t *tctx, bool prof_active, bool updated, const void *old_ptr,
+    size_t old_usize, prof_tctx_t *old_tctx);
 void	prof_free(tsd_t *tsd, const void *ptr, size_t usize);
 #endif
 
@@ -402,13 +407,24 @@ prof_tctx_get(const void *ptr)
 }
 
 JEMALLOC_ALWAYS_INLINE void
-prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
+prof_tctx_set(const void *ptr, size_t usize, prof_tctx_t *tctx)
 {
 
 	cassert(config_prof);
 	assert(ptr != NULL);
 
-	arena_prof_tctx_set(ptr, tctx);
+	arena_prof_tctx_set(ptr, usize, tctx);
+}
+
+JEMALLOC_ALWAYS_INLINE void
+prof_tctx_reset(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);
 }
 
 JEMALLOC_ALWAYS_INLINE bool
@@ -442,7 +458,7 @@ prof_sample_accum_update(tsd_t *tsd, size_t usize, bool update,
 }
 
 JEMALLOC_ALWAYS_INLINE prof_tctx_t *
-prof_alloc_prep(tsd_t *tsd, size_t usize, bool update)
+prof_alloc_prep(tsd_t *tsd, size_t usize, bool prof_active, bool update)
 {
 	prof_tctx_t *ret;
 	prof_tdata_t *tdata;
@@ -450,8 +466,8 @@ prof_alloc_prep(tsd_t *tsd, size_t usize, bool update)
 
 	assert(usize == s2u(usize));
 
-	if (!prof_active_get_unlocked() || likely(prof_sample_accum_update(tsd,
+	if (!prof_active || likely(prof_sample_accum_update(tsd, usize, update,
-	    usize, update, &tdata)))
+	    &tdata)))
 		ret = (prof_tctx_t *)(uintptr_t)1U;
 	else {
 		bt_init(&bt, tdata->vec);
@@ -473,22 +489,24 @@ prof_malloc(const void *ptr, size_t usize, prof_tctx_t *tctx)
 	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
 		prof_malloc_sample_object(ptr, usize, tctx);
 	else
-		prof_tctx_set(ptr, (prof_tctx_t *)(uintptr_t)1U);
+		prof_tctx_set(ptr, usize, (prof_tctx_t *)(uintptr_t)1U);
 }
 
 JEMALLOC_ALWAYS_INLINE void
 prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
-    bool updated, size_t old_usize, prof_tctx_t *old_tctx)
+    bool prof_active, bool updated, const void *old_ptr, size_t old_usize,
+    prof_tctx_t *old_tctx)
 {
+	bool sampled, old_sampled;
 
 	cassert(config_prof);
 	assert(ptr != NULL || (uintptr_t)tctx <= (uintptr_t)1U);
 
-	if (!updated && ptr != NULL) {
+	if (prof_active && !updated && ptr != NULL) {
 		assert(usize == isalloc(ptr, true));
 		if (prof_sample_accum_update(tsd, usize, true, NULL)) {
 			/*
-			 * Don't sample.  The usize passed to PROF_ALLOC_PREP()
+			 * Don't sample.  The usize passed to prof_alloc_prep()
 			 * was larger than what actually got allocated, so a
 			 * backtrace was captured for this allocation, even
 			 * though its actual usize was insufficient to cross the
@@ -498,12 +516,16 @@ prof_realloc(tsd_t *tsd, const void *ptr, size_t usize, prof_tctx_t *tctx,
 		}
 	}
 
-	if (unlikely((uintptr_t)old_tctx > (uintptr_t)1U))
+	sampled = ((uintptr_t)tctx > (uintptr_t)1U);
-		prof_free_sampled_object(tsd, old_usize, old_tctx);
+	old_sampled = ((uintptr_t)old_tctx > (uintptr_t)1U);
-	if (unlikely((uintptr_t)tctx > (uintptr_t)1U))
+
+	if (unlikely(sampled))
 		prof_malloc_sample_object(ptr, usize, tctx);
 	else
-		prof_tctx_set(ptr, (prof_tctx_t *)(uintptr_t)1U);
+		prof_tctx_reset(ptr, usize, old_ptr, old_tctx);
+
+	if (unlikely(old_sampled))
+		prof_free_sampled_object(tsd, old_usize, old_tctx);
 }
 
 JEMALLOC_ALWAYS_INLINE void

include/jemalloc/internal/size_classes.sh

@@ -167,6 +167,8 @@ size_classes() {
           lg_large_minclass=$((${lg_grp} + 2))
         fi
       fi
+      # Final written value is correct:
+      huge_maxclass="((((size_t)1) << ${lg_grp}) + (((size_t)${ndelta}) << ${lg_delta}))"
       index=$((${index} + 1))
       ndelta=$((${ndelta} + 1))
     done
@@ -185,6 +187,7 @@ size_classes() {
   # - lookup_maxclass
   # - small_maxclass
   # - lg_large_minclass
+  # - huge_maxclass
 }
 
 cat <<EOF
@@ -215,6 +218,7 @@ cat <<EOF
  *   LOOKUP_MAXCLASS: Maximum size class included in lookup table.
  *   SMALL_MAXCLASS: Maximum small size class.
  *   LG_LARGE_MINCLASS: Lg of minimum large size class.
+ *   HUGE_MAXCLASS: Maximum (huge) size class.
  */
 
 #define	LG_SIZE_CLASS_GROUP	${lg_g}
@@ -238,6 +242,7 @@ for lg_z in ${lg_zarr} ; do
         echo "#define	LOOKUP_MAXCLASS		${lookup_maxclass}"
         echo "#define	SMALL_MAXCLASS		${small_maxclass}"
         echo "#define	LG_LARGE_MINCLASS	${lg_large_minclass}"
+        echo "#define	HUGE_MAXCLASS		${huge_maxclass}"
         echo "#endif"
         echo
       done

include/jemalloc/internal/tcache.h

@@ -77,7 +77,7 @@ struct tcache_s {
 	ql_elm(tcache_t) link;		/* Used for aggregating stats. */
 	uint64_t	prof_accumbytes;/* Cleared after arena_prof_accum(). */
 	unsigned	ev_cnt;		/* Event count since incremental GC. */
-	index_t		next_gc_bin;	/* Next bin to GC. */
+	szind_t		next_gc_bin;	/* Next bin to GC. */
 	tcache_bin_t	tbins[1];	/* Dynamically sized. */
 	/*
 	 * The pointer stacks associated with tbins follow as a contiguous
@@ -126,10 +126,10 @@ extern tcaches_t	*tcaches;
 size_t	tcache_salloc(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,
-    tcache_bin_t *tbin, index_t binind);
+    tcache_bin_t *tbin, szind_t binind);
 void	tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
-    index_t binind, unsigned rem);
+    szind_t binind, unsigned rem);
-void	tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, index_t binind,
+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,
@@ -161,7 +161,7 @@ void	*tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
 void	*tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
     size_t size, bool zero);
 void	tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr,
-    index_t binind);
+    szind_t binind);
 void	tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr,
     size_t size);
 tcache_t	*tcaches_get(tsd_t *tsd, unsigned ind);
@@ -267,7 +267,7 @@ tcache_alloc_small(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
     bool zero)
 {
 	void *ret;
-	index_t binind;
+	szind_t binind;
 	size_t usize;
 	tcache_bin_t *tbin;
 
@@ -312,7 +312,7 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
     bool zero)
 {
 	void *ret;
-	index_t binind;
+	szind_t binind;
 	size_t usize;
 	tcache_bin_t *tbin;
 
@@ -360,7 +360,7 @@ tcache_alloc_large(tsd_t *tsd, arena_t *arena, tcache_t *tcache, size_t size,
 }
 
 JEMALLOC_ALWAYS_INLINE void
-tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, index_t binind)
+tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, szind_t binind)
 {
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
@@ -386,7 +386,7 @@ tcache_dalloc_small(tsd_t *tsd, tcache_t *tcache, void *ptr, index_t binind)
 JEMALLOC_ALWAYS_INLINE void
 tcache_dalloc_large(tsd_t *tsd, tcache_t *tcache, void *ptr, size_t size)
 {
-	index_t binind;
+	szind_t binind;
 	tcache_bin_t *tbin;
 	tcache_bin_info_t *tbin_info;
 

include/jemalloc/jemalloc_protos.h.in

@@ -56,7 +56,7 @@ JEMALLOC_EXPORT size_t JEMALLOC_NOTHROW	@je_@malloc_usable_size(
 #ifdef JEMALLOC_OVERRIDE_MEMALIGN
 JEMALLOC_EXPORT JEMALLOC_ALLOCATOR JEMALLOC_RESTRICT_RETURN
     void JEMALLOC_NOTHROW	*@je_@memalign(size_t alignment, size_t size)
-    JEMALLOC_ATTR(malloc);
+    JEMALLOC_CXX_THROW JEMALLOC_ATTR(malloc);
 #endif
 
 #ifdef JEMALLOC_OVERRIDE_VALLOC

src/arena.c

@@ -11,7 +11,7 @@ arena_bin_info_t	arena_bin_info[NBINS];
 size_t		map_bias;
 size_t		map_misc_offset;
 size_t		arena_maxrun; /* Max run size for arenas. */
-size_t		arena_maxclass; /* Max size class for arenas. */
+size_t		large_maxclass; /* Max large size class. */
 static size_t	small_maxrun; /* Max run size used for small size classes. */
 static bool	*small_run_tab; /* Valid small run page multiples. */
 unsigned	nlclasses; /* Number of large size classes. */
@@ -39,7 +39,7 @@ JEMALLOC_INLINE_C arena_chunk_map_misc_t *
 arena_miscelm_key_create(size_t size)
 {
 
-	return ((arena_chunk_map_misc_t *)((size << CHUNK_MAP_SIZE_SHIFT) |
+	return ((arena_chunk_map_misc_t *)(arena_mapbits_size_encode(size) |
 	    CHUNK_MAP_KEY));
 }
 
@@ -58,8 +58,7 @@ arena_miscelm_key_size_get(const arena_chunk_map_misc_t *miscelm)
 
 	assert(arena_miscelm_is_key(miscelm));
 
-	return (((uintptr_t)miscelm & CHUNK_MAP_SIZE_MASK) >>
+	return (arena_mapbits_size_decode((uintptr_t)miscelm));
-	    CHUNK_MAP_SIZE_SHIFT);
 }
 
 JEMALLOC_INLINE_C size_t
@@ -73,7 +72,7 @@ arena_miscelm_size_get(arena_chunk_map_misc_t *miscelm)
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(miscelm);
 	pageind = arena_miscelm_to_pageind(miscelm);
 	mapbits = arena_mapbits_get(chunk, pageind);
-	return ((mapbits & CHUNK_MAP_SIZE_MASK) >> CHUNK_MAP_SIZE_SHIFT);
+	return (arena_mapbits_size_decode(mapbits));
 }
 
 JEMALLOC_INLINE_C int
@@ -315,7 +314,7 @@ arena_run_reg_dalloc(arena_run_t *run, void *ptr)
 	arena_chunk_t *chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);
 	size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
 	size_t mapbits = arena_mapbits_get(chunk, pageind);
-	index_t binind = arena_ptr_small_binind_get(ptr, mapbits);
+	szind_t binind = arena_ptr_small_binind_get(ptr, mapbits);
 	arena_bin_info_t *bin_info = &arena_bin_info[binind];
 	unsigned regind = arena_run_regind(run, bin_info, ptr);
 
@@ -426,7 +425,7 @@ arena_run_split_large_helper(arena_t *arena, arena_run_t *run, size_t size,
 {
 	arena_chunk_t *chunk;
 	arena_chunk_map_misc_t *miscelm;
-	size_t flag_dirty, flag_decommitted, run_ind, need_pages, i;
+	size_t flag_dirty, flag_decommitted, run_ind, need_pages;
 	size_t flag_unzeroed_mask;
 
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(run);
@@ -460,6 +459,7 @@ arena_run_split_large_helper(arena_t *arena, arena_run_t *run, size_t size,
 			 * The run is clean, so some pages may be zeroed (i.e.
 			 * never before touched).
 			 */
+			size_t i;
 			for (i = 0; i < need_pages; i++) {
 				if (arena_mapbits_unzeroed_get(chunk, run_ind+i)
 				    != 0)
@@ -508,7 +508,7 @@ arena_run_init_large(arena_t *arena, arena_run_t *run, size_t size, bool zero)
 
 static bool
 arena_run_split_small(arena_t *arena, arena_run_t *run, size_t size,
-    index_t binind)
+    szind_t binind)
 {
 	arena_chunk_t *chunk;
 	arena_chunk_map_misc_t *miscelm;
@@ -780,7 +780,7 @@ arena_chunk_dalloc(arena_t *arena, arena_chunk_t *chunk)
 static void
 arena_huge_malloc_stats_update(arena_t *arena, size_t usize)
 {
-	index_t index = size2index(usize) - nlclasses - NBINS;
+	szind_t index = size2index(usize) - nlclasses - NBINS;
 
 	cassert(config_stats);
 
@@ -793,7 +793,7 @@ arena_huge_malloc_stats_update(arena_t *arena, size_t usize)
 static void
 arena_huge_malloc_stats_update_undo(arena_t *arena, size_t usize)
 {
-	index_t index = size2index(usize) - nlclasses - NBINS;
+	szind_t index = size2index(usize) - nlclasses - NBINS;
 
 	cassert(config_stats);
 
@@ -806,7 +806,7 @@ arena_huge_malloc_stats_update_undo(arena_t *arena, size_t usize)
 static void
 arena_huge_dalloc_stats_update(arena_t *arena, size_t usize)
 {
-	index_t index = size2index(usize) - nlclasses - NBINS;
+	szind_t index = size2index(usize) - nlclasses - NBINS;
 
 	cassert(config_stats);
 
@@ -819,7 +819,7 @@ arena_huge_dalloc_stats_update(arena_t *arena, size_t usize)
 static void
 arena_huge_dalloc_stats_update_undo(arena_t *arena, size_t usize)
 {
-	index_t index = size2index(usize) - nlclasses - NBINS;
+	szind_t index = size2index(usize) - nlclasses - NBINS;
 
 	cassert(config_stats);
 
@@ -1125,7 +1125,7 @@ arena_run_alloc_large(arena_t *arena, size_t size, bool zero)
 }
 
 static arena_run_t *
-arena_run_alloc_small_helper(arena_t *arena, size_t size, index_t binind)
+arena_run_alloc_small_helper(arena_t *arena, size_t size, szind_t binind)
 {
 	arena_run_t *run = arena_run_first_best_fit(arena, size);
 	if (run != NULL) {
@@ -1136,7 +1136,7 @@ arena_run_alloc_small_helper(arena_t *arena, size_t size, index_t binind)
 }
 
 static arena_run_t *
-arena_run_alloc_small(arena_t *arena, size_t size, index_t binind)
+arena_run_alloc_small(arena_t *arena, size_t size, szind_t binind)
 {
 	arena_chunk_t *chunk;
 	arena_run_t *run;
@@ -1889,7 +1889,7 @@ static arena_run_t *
 arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin)
 {
 	arena_run_t *run;
-	index_t binind;
+	szind_t binind;
 	arena_bin_info_t *bin_info;
 
 	/* Look for a usable run. */
@@ -1939,8 +1939,7 @@ arena_bin_nonfull_run_get(arena_t *arena, arena_bin_t *bin)
 static void *
 arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin)
 {
-	void *ret;
+	szind_t binind;
-	index_t binind;
 	arena_bin_info_t *bin_info;
 	arena_run_t *run;
 
@@ -1953,6 +1952,7 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin)
 		 * Another thread updated runcur while this one ran without the
 		 * bin lock in arena_bin_nonfull_run_get().
 		 */
+		void *ret;
 		assert(bin->runcur->nfree > 0);
 		ret = arena_run_reg_alloc(bin->runcur, bin_info);
 		if (run != NULL) {
@@ -1986,13 +1986,11 @@ arena_bin_malloc_hard(arena_t *arena, arena_bin_t *bin)
 }
 
 void
-arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, index_t binind,
+arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, szind_t binind,
     uint64_t prof_accumbytes)
 {
 	unsigned i, nfill;
 	arena_bin_t *bin;
-	arena_run_t *run;
-	void *ptr;
 
 	assert(tbin->ncached == 0);
 
@@ -2002,6 +2000,8 @@ arena_tcache_fill_small(arena_t *arena, tcache_bin_t *tbin, index_t binind,
 	malloc_mutex_lock(&bin->lock);
 	for (i = 0, nfill = (tcache_bin_info[binind].ncached_max >>
 	    tbin->lg_fill_div); i < nfill; i++) {
+		arena_run_t *run;
+		void *ptr;
 		if ((run = bin->runcur) != NULL && run->nfree > 0)
 			ptr = arena_run_reg_alloc(run, &arena_bin_info[binind]);
 		else
@@ -2076,12 +2076,13 @@ arena_redzone_corruption_t *arena_redzone_corruption =
 static void
 arena_redzones_validate(void *ptr, arena_bin_info_t *bin_info, bool reset)
 {
-	size_t size = bin_info->reg_size;
-	size_t redzone_size = bin_info->redzone_size;
-	size_t i;
 	bool error = false;
 
 	if (opt_junk_alloc) {
+		size_t size = bin_info->reg_size;
+		size_t redzone_size = bin_info->redzone_size;
+		size_t i;
+
 		for (i = 1; i <= redzone_size; i++) {
 			uint8_t *byte = (uint8_t *)((uintptr_t)ptr - i);
 			if (*byte != 0xa5) {
@@ -2131,7 +2132,7 @@ arena_dalloc_junk_small_t *arena_dalloc_junk_small =
 void
 arena_quarantine_junk_small(void *ptr, size_t usize)
 {
-	index_t binind;
+	szind_t binind;
 	arena_bin_info_t *bin_info;
 	cassert(config_fill);
 	assert(opt_junk_free);
@@ -2149,7 +2150,7 @@ arena_malloc_small(arena_t *arena, size_t size, bool zero)
 	void *ret;
 	arena_bin_t *bin;
 	arena_run_t *run;
-	index_t binind;
+	szind_t binind;
 
 	binind = size2index(size);
 	assert(binind < NBINS);
@@ -2233,7 +2234,7 @@ arena_malloc_large(arena_t *arena, size_t size, bool zero)
 	ret = (void *)((uintptr_t)arena_miscelm_to_rpages(miscelm) +
 	    random_offset);
 	if (config_stats) {
-		index_t index = size2index(usize) - NBINS;
+		szind_t index = size2index(usize) - NBINS;
 
 		arena->stats.nmalloc_large++;
 		arena->stats.nrequests_large++;
@@ -2326,7 +2327,7 @@ arena_palloc_large(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
 	ret = arena_miscelm_to_rpages(miscelm);
 
 	if (config_stats) {
-		index_t index = size2index(usize) - NBINS;
+		szind_t index = size2index(usize) - NBINS;
 
 		arena->stats.nmalloc_large++;
 		arena->stats.nrequests_large++;
@@ -2356,7 +2357,7 @@ arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
 	    && (usize & PAGE_MASK) == 0))) {
 		/* Small; alignment doesn't require special run placement. */
 		ret = arena_malloc(tsd, arena, usize, zero, tcache);
-	} else if (usize <= arena_maxclass && alignment <= PAGE) {
+	} else if (usize <= large_maxclass && alignment <= PAGE) {
 		/*
 		 * Large; alignment doesn't require special run placement.
 		 * However, the cached pointer may be at a random offset from
@@ -2367,7 +2368,7 @@ arena_palloc(tsd_t *tsd, arena_t *arena, size_t usize, size_t alignment,
 		if (config_cache_oblivious)
 			ret = (void *)((uintptr_t)ret & ~PAGE_MASK);
 	} else {
-		if (likely(usize <= arena_maxclass)) {
+		if (likely(usize <= large_maxclass)) {
 			ret = arena_palloc_large(tsd, arena, usize, alignment,
 			    zero);
 		} else if (likely(alignment <= chunksize))
@@ -2385,7 +2386,7 @@ arena_prof_promoted(const void *ptr, size_t size)
 {
 	arena_chunk_t *chunk;
 	size_t pageind;
-	index_t binind;
+	szind_t binind;
 
 	cassert(config_prof);
 	assert(ptr != NULL);
@@ -2413,7 +2414,7 @@ arena_dissociate_bin_run(arena_chunk_t *chunk, arena_run_t *run,
 	if (run == bin->runcur)
 		bin->runcur = NULL;
 	else {
-		index_t binind = arena_bin_index(extent_node_arena_get(
+		szind_t binind = arena_bin_index(extent_node_arena_get(
 		    &chunk->node), bin);
 		arena_bin_info_t *bin_info = &arena_bin_info[binind];
 
@@ -2477,7 +2478,7 @@ arena_dalloc_bin_locked_impl(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 	arena_run_t *run;
 	arena_bin_t *bin;
 	arena_bin_info_t *bin_info;
-	index_t binind;
+	szind_t binind;
 
 	pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
 	rpages_ind = pageind - arena_mapbits_small_runind_get(chunk, pageind);
@@ -2574,7 +2575,7 @@ arena_dalloc_large_locked_impl(arena_t *arena, arena_chunk_t *chunk,
 		if (!junked)
 			arena_dalloc_junk_large(ptr, usize);
 		if (config_stats) {
-			index_t index = size2index(usize) - NBINS;
+			szind_t index = size2index(usize) - NBINS;
 
 			arena->stats.ndalloc_large++;
 			arena->stats.allocated_large -= usize;
@@ -2621,8 +2622,8 @@ arena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 	arena_run_trim_tail(arena, chunk, run, oldsize + large_pad, size +
 	    large_pad, true);
 	if (config_stats) {
-		index_t oldindex = size2index(oldsize) - NBINS;
+		szind_t oldindex = size2index(oldsize) - NBINS;
-		index_t index = size2index(size) - NBINS;
+		szind_t index = size2index(size) - NBINS;
 
 		arena->stats.ndalloc_large++;
 		arena->stats.allocated_large -= oldsize;
@@ -2641,42 +2642,42 @@ arena_ralloc_large_shrink(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 
 static bool
 arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr,
-    size_t oldsize, size_t size, size_t extra, bool zero)
+    size_t oldsize, size_t usize_min, size_t usize_max, bool zero)
 {
 	size_t pageind = ((uintptr_t)ptr - (uintptr_t)chunk) >> LG_PAGE;
 	size_t npages = (oldsize + large_pad) >> LG_PAGE;
 	size_t followsize;
-	size_t usize_min = s2u(size);
 
 	assert(oldsize == arena_mapbits_large_size_get(chunk, pageind) -
 	    large_pad);
 
 	/* Try to extend the run. */
-	assert(usize_min > oldsize);
 	malloc_mutex_lock(&arena->lock);
-	if (pageind+npages < chunk_npages &&
+	if (pageind+npages >= chunk_npages || arena_mapbits_allocated_get(chunk,
-	    arena_mapbits_allocated_get(chunk, pageind+npages) == 0 &&
+	    pageind+npages) != 0)
-	    (followsize = arena_mapbits_unallocated_size_get(chunk,
+		goto label_fail;
-	    pageind+npages)) >= usize_min - oldsize) {
+	followsize = arena_mapbits_unallocated_size_get(chunk, pageind+npages);
+	if (oldsize + followsize >= usize_min) {
 		/*
 		 * The next run is available and sufficiently large.  Split the
 		 * following run, then merge the first part with the existing
 		 * allocation.
 		 */
 		arena_run_t *run;
-		size_t flag_dirty, flag_unzeroed_mask, splitsize, usize;
+		size_t usize, splitsize, size, flag_dirty, flag_unzeroed_mask;
 
-		usize = s2u(size + extra);
+		usize = usize_max;
 		while (oldsize + followsize < usize)
 			usize = index2size(size2index(usize)-1);
 		assert(usize >= usize_min);
+		assert(usize >= oldsize);
 		splitsize = usize - oldsize;
+		if (splitsize == 0)
+			goto label_fail;
 
 		run = &arena_miscelm_get(chunk, pageind+npages)->run;
-		if (arena_run_split_large(arena, run, splitsize, zero)) {
+		if (arena_run_split_large(arena, run, splitsize, zero))
-			malloc_mutex_unlock(&arena->lock);
+			goto label_fail;
-			return (true);
-		}
 
 		size = oldsize + splitsize;
 		npages = (size + large_pad) >> LG_PAGE;
@@ -2700,8 +2701,8 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 		    pageind+npages-1)));
 
 		if (config_stats) {
-			index_t oldindex = size2index(oldsize) - NBINS;
+			szind_t oldindex = size2index(oldsize) - NBINS;
-			index_t index = size2index(size) - NBINS;
+			szind_t index = size2index(size) - NBINS;
 
 			arena->stats.ndalloc_large++;
 			arena->stats.allocated_large -= oldsize;
@@ -2718,8 +2719,8 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 		malloc_mutex_unlock(&arena->lock);
 		return (false);
 	}
+label_fail:
 	malloc_mutex_unlock(&arena->lock);
-
 	return (true);
 }
 
@@ -2748,98 +2749,107 @@ arena_ralloc_junk_large_t *arena_ralloc_junk_large =
  * always fail if growing an object, and the following run is already in use.
  */
 static bool
-arena_ralloc_large(void *ptr, size_t oldsize, size_t size, size_t extra,
+arena_ralloc_large(void *ptr, size_t oldsize, size_t usize_min,
-    bool zero)
+    size_t usize_max, bool zero)
 {
-	size_t usize;
-
-	/* Make sure extra can't cause size_t overflow. */
-	if (unlikely(extra >= arena_maxclass))
-		return (true);
-
-	usize = s2u(size + extra);
-	if (usize == oldsize) {
-		/* Same size class. */
-		return (false);
-	} else {
 	arena_chunk_t *chunk;
 	arena_t *arena;
 
+	if (oldsize == usize_max) {
+		/* Current size class is compatible and maximal. */
+		return (false);
+	}
+
 	chunk = (arena_chunk_t *)CHUNK_ADDR2BASE(ptr);
 	arena = extent_node_arena_get(&chunk->node);
 
-		if (usize < oldsize) {
+	if (oldsize < usize_max) {
-			/* Fill before shrinking in order avoid a race. */
+		bool ret = arena_ralloc_large_grow(arena, chunk, ptr, oldsize,
-			arena_ralloc_junk_large(ptr, oldsize, usize);
+		    usize_min, usize_max, zero);
-			arena_ralloc_large_shrink(arena, chunk, ptr, oldsize,
-			    usize);
-			return (false);
-		} else {
-			bool ret = arena_ralloc_large_grow(arena, chunk, ptr,
-			    oldsize, size, extra, zero);
 		if (config_fill && !ret && !zero) {
 			if (unlikely(opt_junk_alloc)) {
-					memset((void *)((uintptr_t)ptr +
+				memset((void *)((uintptr_t)ptr + oldsize), 0xa5,
-					    oldsize), 0xa5, isalloc(ptr,
+				    isalloc(ptr, config_prof) - oldsize);
-					    config_prof) - oldsize);
 			} else if (unlikely(opt_zero)) {
-					memset((void *)((uintptr_t)ptr +
+				memset((void *)((uintptr_t)ptr + oldsize), 0,
-					    oldsize), 0, isalloc(ptr,
+				    isalloc(ptr, config_prof) - oldsize);
-					    config_prof) - oldsize);
 			}
 		}
 		return (ret);
 	}
-	}
+
+	assert(oldsize > usize_max);
+	/* Fill before shrinking in order avoid a race. */
+	arena_ralloc_junk_large(ptr, oldsize, usize_max);
+	arena_ralloc_large_shrink(arena, chunk, ptr, oldsize, usize_max);
+	return (false);
 }
 
 bool
 arena_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra,
     bool zero)
 {
+	size_t usize_min, usize_max;
 
-	if (likely(size <= arena_maxclass)) {
+	usize_min = s2u(size);
+	usize_max = s2u(size + extra);
+	if (likely(oldsize <= large_maxclass && usize_min <= large_maxclass)) {
 		/*
 		 * Avoid moving the allocation if the size class can be left the
 		 * same.
 		 */
-		if (likely(oldsize <= arena_maxclass)) {
 		if (oldsize <= SMALL_MAXCLASS) {
-				assert(
+			assert(arena_bin_info[size2index(oldsize)].reg_size ==
-				    arena_bin_info[size2index(oldsize)].reg_size
+			    oldsize);
-				    == oldsize);
+			if ((usize_max <= SMALL_MAXCLASS &&
-				if ((size + extra <= SMALL_MAXCLASS &&
+			    size2index(usize_max) == size2index(oldsize)) ||
-				    size2index(size + extra) ==
+			    (size <= oldsize && usize_max >= oldsize))
-				    size2index(oldsize)) || (size <= oldsize &&
-				    size + extra >= oldsize))
 				return (false);
 		} else {
-				assert(size <= arena_maxclass);
+			if (usize_max > SMALL_MAXCLASS) {
-				if (size + extra > SMALL_MAXCLASS) {
+				if (!arena_ralloc_large(ptr, oldsize, usize_min,
-					if (!arena_ralloc_large(ptr, oldsize,
+				    usize_max, zero))
-					    size, extra, zero))
 					return (false);
 			}
 		}
-		}
 
 		/* Reallocation would require a move. */
 		return (true);
-	} else
+	} else {
-		return (huge_ralloc_no_move(ptr, oldsize, size, extra, zero));
+		return (huge_ralloc_no_move(ptr, oldsize, usize_min, usize_max,
+		    zero));
+	}
+}
+
+static void *
+arena_ralloc_move_helper(tsd_t *tsd, arena_t *arena, size_t usize,
+    size_t alignment, bool zero, tcache_t *tcache)
+{
+
+	if (alignment == 0)
+		return (arena_malloc(tsd, arena, usize, zero, tcache));
+	usize = sa2u(usize, alignment);
+	if (usize == 0)
+		return (NULL);
+	return (ipalloct(tsd, usize, alignment, zero, tcache, arena));
 }
 
 void *
 arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t size,
-    size_t extra, size_t alignment, bool zero, tcache_t *tcache)
+    size_t alignment, bool zero, tcache_t *tcache)
 {
 	void *ret;
+	size_t usize;
 
-	if (likely(size <= arena_maxclass)) {
+	usize = s2u(size);
+	if (usize == 0)
+		return (NULL);
+
+	if (likely(usize <= large_maxclass)) {
 		size_t copysize;
 
 		/* Try to avoid moving the allocation. */
-		if (!arena_ralloc_no_move(ptr, oldsize, size, extra, zero))
+		if (!arena_ralloc_no_move(ptr, oldsize, usize, 0, zero))
 			return (ptr);
 
 		/*
@@ -2847,53 +2857,23 @@ arena_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t size,
 		 * the object.  In that case, fall back to allocating new space
 		 * and copying.
 		 */
-		if (alignment != 0) {
+		ret = arena_ralloc_move_helper(tsd, arena, usize, alignment,
-			size_t usize = sa2u(size + extra, alignment);
+		    zero, tcache);
-			if (usize == 0)
-				return (NULL);
-			ret = ipalloct(tsd, usize, alignment, zero, tcache,
-			    arena);
-		} else {
-			ret = arena_malloc(tsd, arena, size + extra, zero,
-			    tcache);
-		}
-
-		if (ret == NULL) {
-			if (extra == 0)
-				return (NULL);
-			/* Try again, this time without extra. */
-			if (alignment != 0) {
-				size_t usize = sa2u(size, alignment);
-				if (usize == 0)
-					return (NULL);
-				ret = ipalloct(tsd, usize, alignment, zero,
-				    tcache, arena);
-			} else {
-				ret = arena_malloc(tsd, arena, size, zero,
-				    tcache);
-			}
-
 		if (ret == NULL)
 			return (NULL);
-		}
 
 		/*
 		 * Junk/zero-filling were already done by
 		 * ipalloc()/arena_malloc().
 		 */
 
-		/*
+		copysize = (usize < oldsize) ? usize : oldsize;
-		 * Copy at most size bytes (not size+extra), since the caller
-		 * has no expectation that the extra bytes will be reliably
-		 * preserved.
-		 */
-		copysize = (size < oldsize) ? size : oldsize;
 		JEMALLOC_VALGRIND_MAKE_MEM_UNDEFINED(ret, copysize);
 		memcpy(ret, ptr, copysize);
 		isqalloc(tsd, ptr, oldsize, tcache);
 	} else {
-		ret = huge_ralloc(tsd, arena, ptr, oldsize, size, extra,
+		ret = huge_ralloc(tsd, arena, ptr, oldsize, usize, alignment,
-		    alignment, zero, tcache);
+		    zero, tcache);
 	}
 	return (ret);
 }
@@ -3241,7 +3221,6 @@ small_run_size_init(void)
 bool
 arena_boot(void)
 {
-	size_t header_size;
 	unsigned i;
 
 	arena_lg_dirty_mult_default_set(opt_lg_dirty_mult);
@@ -3260,7 +3239,7 @@ arena_boot(void)
 	 */
 	map_bias = 0;
 	for (i = 0; i < 3; i++) {
-		header_size = offsetof(arena_chunk_t, map_bits) +
+		size_t header_size = offsetof(arena_chunk_t, map_bits) +
 		    ((sizeof(arena_chunk_map_bits_t) +
 		    sizeof(arena_chunk_map_misc_t)) * (chunk_npages-map_bias));
 		map_bias = (header_size + PAGE_MASK) >> LG_PAGE;
@@ -3272,17 +3251,17 @@ arena_boot(void)
 
 	arena_maxrun = chunksize - (map_bias << LG_PAGE);
 	assert(arena_maxrun > 0);
-	arena_maxclass = index2size(size2index(chunksize)-1);
+	large_maxclass = index2size(size2index(chunksize)-1);
-	if (arena_maxclass > arena_maxrun) {
+	if (large_maxclass > arena_maxrun) {
 		/*
 		 * For small chunk sizes it's possible for there to be fewer
 		 * non-header pages available than are necessary to serve the
 		 * size classes just below chunksize.
 		 */
-		arena_maxclass = arena_maxrun;
+		large_maxclass = arena_maxrun;
 	}
-	assert(arena_maxclass > 0);
+	assert(large_maxclass > 0);
-	nlclasses = size2index(arena_maxclass) - size2index(SMALL_MAXCLASS);
+	nlclasses = size2index(large_maxclass) - size2index(SMALL_MAXCLASS);
 	nhclasses = NSIZES - nlclasses - NBINS;
 
 	bin_info_init();

src/chunk_dss.c

@@ -69,8 +69,6 @@ void *
 chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
     bool *zero, bool *commit)
 {
-	void *ret;
-
 	cassert(have_dss);
 	assert(size > 0 && (size & chunksize_mask) == 0);
 	assert(alignment > 0 && (alignment & chunksize_mask) == 0);
@@ -84,9 +82,6 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 
 	malloc_mutex_lock(&dss_mtx);
 	if (dss_prev != (void *)-1) {
-		size_t gap_size, cpad_size;
-		void *cpad, *dss_next;
-		intptr_t incr;
 
 		/*
 		 * The loop is necessary to recover from races with other
@@ -94,6 +89,9 @@ chunk_alloc_dss(arena_t *arena, void *new_addr, size_t size, size_t alignment,
 		 * malloc.
 		 */
 		do {
+			void *ret, *cpad, *dss_next;
+			size_t gap_size, cpad_size;
+			intptr_t incr;
 			/* Avoid an unnecessary system call. */
 			if (new_addr != NULL && dss_max != new_addr)
 				break;

src/chunk_mmap.c

@@ -6,14 +6,16 @@
 static void *
 chunk_alloc_mmap_slow(size_t size, size_t alignment, bool *zero, bool *commit)
 {
-	void *ret, *pages;
+	void *ret;
-	size_t alloc_size, leadsize;
+	size_t alloc_size;
 
 	alloc_size = size + alignment - PAGE;
 	/* Beware size_t wrap-around. */
 	if (alloc_size < size)
 		return (NULL);
 	do {
+		void *pages;
+		size_t leadsize;
 		pages = pages_map(NULL, alloc_size);
 		if (pages == NULL)
 			return (NULL);

src/huge.c

@@ -126,18 +126,19 @@ 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,
+huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
-    size_t size, size_t extra, bool zero)
+    size_t usize_max, bool zero)
 {
-	size_t usize_next;
+	size_t usize, usize_next;
 	extent_node_t *node;
 	arena_t *arena;
 	chunk_hooks_t chunk_hooks = CHUNK_HOOKS_INITIALIZER;
 	bool zeroed;
 
 	/* Increase usize to incorporate extra. */
-	while (usize < s2u(size+extra) && (usize_next = s2u(usize+1)) < oldsize)
+	for (usize = usize_min; usize < usize_max && (usize_next = s2u(usize+1))
-		usize = usize_next;
+	    <= oldsize; usize = usize_next)
+		; /* Do nothing. */
 
 	if (oldsize == usize)
 		return;
@@ -148,11 +149,12 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize,
 	/* Fill if necessary (shrinking). */
 	if (oldsize > usize) {
 		size_t sdiff = oldsize - usize;
-		zeroed = !chunk_purge_wrapper(arena, &chunk_hooks, ptr,
-		    CHUNK_CEILING(usize), usize, sdiff);
 		if (config_fill && unlikely(opt_junk_free)) {
 			memset((void *)((uintptr_t)ptr + usize), 0x5a, sdiff);
 			zeroed = false;
+		} else {
+			zeroed = !chunk_purge_wrapper(arena, &chunk_hooks, ptr,
+			    CHUNK_CEILING(oldsize), usize, sdiff);
 		}
 	} else
 		zeroed = true;
@@ -194,6 +196,8 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 	arena = extent_node_arena_get(node);
 	chunk_hooks = chunk_hooks_get(arena);
 
+	assert(oldsize > usize);
+
 	/* Split excess chunks. */
 	cdiff = CHUNK_CEILING(oldsize) - CHUNK_CEILING(usize);
 	if (cdiff != 0 && chunk_hooks.split(ptr, CHUNK_CEILING(oldsize),
@@ -202,14 +206,15 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 
 	if (oldsize > usize) {
 		size_t sdiff = oldsize - usize;
-		zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
-		    CHUNK_ADDR2BASE((uintptr_t)ptr + usize),
-		    CHUNK_CEILING(usize), CHUNK_ADDR2OFFSET((uintptr_t)ptr +
-		    usize), sdiff);
 		if (config_fill && unlikely(opt_junk_free)) {
 			huge_dalloc_junk((void *)((uintptr_t)ptr + usize),
 			    sdiff);
 			zeroed = false;
+		} else {
+			zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
+			    CHUNK_ADDR2BASE((uintptr_t)ptr + usize),
+			    CHUNK_CEILING(oldsize),
+			    CHUNK_ADDR2OFFSET((uintptr_t)ptr + usize), sdiff);
 		}
 	} else
 		zeroed = true;
@@ -228,18 +233,11 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 }
 
 static bool
-huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t size, bool zero) {
+huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t usize, bool zero) {
-	size_t usize;
 	extent_node_t *node;
 	arena_t *arena;
 	bool is_zeroed_subchunk, is_zeroed_chunk;
 
-	usize = s2u(size);
-	if (usize == 0) {
-		/* size_t overflow. */
-		return (true);
-	}
-
 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
 	malloc_mutex_lock(&arena->huge_mtx);
@@ -280,89 +278,76 @@ huge_ralloc_no_move_expand(void *ptr, size_t oldsize, size_t size, bool zero) {
 }
 
 bool
-huge_ralloc_no_move(void *ptr, size_t oldsize, size_t size, size_t extra,
+huge_ralloc_no_move(void *ptr, size_t oldsize, size_t usize_min,
-    bool zero)
+    size_t usize_max, bool zero)
 {
-	size_t usize;
-
-	/* Both allocations must be huge to avoid a move. */
-	if (oldsize < chunksize)
-		return (true);
 
 	assert(s2u(oldsize) == oldsize);
-	usize = s2u(size);
+
-	if (usize == 0) {
+	/* Both allocations must be huge to avoid a move. */
-		/* size_t overflow. */
+	if (oldsize < chunksize || usize_max < chunksize)
 		return (true);
+
+	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, zero))
+			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))
+			return (false);
 	}
 
 	/*
 	 * Avoid moving the allocation if the existing chunk size accommodates
 	 * the new size.
 	 */
-	if (CHUNK_CEILING(oldsize) >= CHUNK_CEILING(usize)
+	if (CHUNK_CEILING(oldsize) >= CHUNK_CEILING(usize_min)
-	    && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(s2u(size+extra))) {
+	    && CHUNK_CEILING(oldsize) <= CHUNK_CEILING(usize_max)) {
-		huge_ralloc_no_move_similar(ptr, oldsize, usize, size, extra,
+		huge_ralloc_no_move_similar(ptr, oldsize, usize_min, usize_max,
 		    zero);
 		return (false);
 	}
 
 	/* Attempt to shrink the allocation in-place. */
-	if (CHUNK_CEILING(oldsize) >= CHUNK_CEILING(usize))
+	if (CHUNK_CEILING(oldsize) > CHUNK_CEILING(usize_max))
-		return (huge_ralloc_no_move_shrink(ptr, oldsize, usize));
+		return (huge_ralloc_no_move_shrink(ptr, oldsize, usize_max));
-
-	/* Attempt to expand the allocation in-place. */
-	if (huge_ralloc_no_move_expand(ptr, oldsize, size + extra, zero)) {
-		if (extra == 0)
 	return (true);
-
-		/* Try again, this time without extra. */
-		return (huge_ralloc_no_move_expand(ptr, oldsize, size, zero));
 }
-	return (false);
+
+static void *
+huge_ralloc_move_helper(tsd_t *tsd, arena_t *arena, size_t usize,
+    size_t alignment, bool zero, tcache_t *tcache)
+{
+
+	if (alignment <= chunksize)
+		return (huge_malloc(tsd, arena, usize, zero, tcache));
+	return (huge_palloc(tsd, arena, usize, alignment, zero, tcache));
 }
 
 void *
-huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t size,
+huge_ralloc(tsd_t *tsd, arena_t *arena, void *ptr, size_t oldsize, size_t usize,
-    size_t extra, size_t alignment, bool zero, tcache_t *tcache)
+    size_t alignment, bool zero, tcache_t *tcache)
 {
 	void *ret;
 	size_t copysize;
 
 	/* Try to avoid moving the allocation. */
-	if (!huge_ralloc_no_move(ptr, oldsize, size, extra, zero))
+	if (!huge_ralloc_no_move(ptr, oldsize, usize, usize, zero))
 		return (ptr);
 
 	/*
-	 * size and oldsize are different enough that we need to use a
+	 * usize and oldsize are different enough that we need to use a
 	 * different size class.  In that case, fall back to allocating new
 	 * space and copying.
 	 */
-	if (alignment > chunksize) {
+	ret = huge_ralloc_move_helper(tsd, arena, usize, alignment, zero,
-		ret = huge_palloc(tsd, arena, size + extra, alignment, zero,
 	    tcache);
-	} else
-		ret = huge_malloc(tsd, arena, size + extra, zero, tcache);
-
-	if (ret == NULL) {
-		if (extra == 0)
-			return (NULL);
-		/* Try again, this time without extra. */
-		if (alignment > chunksize) {
-			ret = huge_palloc(tsd, arena, size, alignment, zero,
-			    tcache);
-		} else
-			ret = huge_malloc(tsd, arena, size, zero, tcache);
-
 	if (ret == NULL)
 		return (NULL);
-	}
 
-	/*
+	copysize = (usize < oldsize) ? usize : oldsize;
-	 * Copy at most size bytes (not size+extra), since the caller has no
-	 * expectation that the extra bytes will be reliably preserved.
-	 */
-	copysize = (size < oldsize) ? size : oldsize;
 	memcpy(ret, ptr, copysize);
 	isqalloc(tsd, ptr, oldsize, tcache);
 	return (ret);
@@ -439,3 +424,10 @@ huge_prof_tctx_set(const void *ptr, prof_tctx_t *tctx)
 	extent_node_prof_tctx_set(node, tctx);
 	malloc_mutex_unlock(&arena->huge_mtx);
 }
+
+void
+huge_prof_tctx_reset(const void *ptr)
+{
+
+	huge_prof_tctx_set(ptr, (prof_tctx_t *)(uintptr_t)1U);
+}

src/jemalloc.c

@@ -179,13 +179,24 @@ static bool			malloc_initializer = NO_INITIALIZER;
 static malloc_mutex_t	init_lock = SRWLOCK_INIT;
 #else
 static malloc_mutex_t	init_lock;
+static bool init_lock_initialized = false;
 
 JEMALLOC_ATTR(constructor)
 static void WINAPI
 _init_init_lock(void)
 {
 
+	/* If another constructor in the same binary is using mallctl to
+	 * e.g. setup chunk hooks, it may end up running before this one,
+	 * and malloc_init_hard will crash trying to lock the uninitialized
+	 * lock. So we force an initialization of the lock in
+	 * malloc_init_hard as well. We don't try to care about atomicity
+	 * of the accessed to the init_lock_initialized boolean, since it
+	 * really only matters early in the process creation, before any
+	 * separate thread normally starts doing anything. */
+	if (!init_lock_initialized)
 		malloc_mutex_init(&init_lock);
+	init_lock_initialized = true;
 }
 
 #ifdef _MSC_VER
@@ -510,17 +521,17 @@ arena_get_hard(tsd_t *tsd, unsigned ind, bool init_if_missing)
 		assert(ind < narenas_actual || !init_if_missing);
 		narenas_cache = (ind < narenas_actual) ? narenas_actual : ind+1;
 
-		if (!*arenas_cache_bypassp) {
+		if (tsd_nominal(tsd) && !*arenas_cache_bypassp) {
 			*arenas_cache_bypassp = true;
 			arenas_cache = (arena_t **)a0malloc(sizeof(arena_t *) *
 			    narenas_cache);
 			*arenas_cache_bypassp = false;
-		} else
+		}
-			arenas_cache = NULL;
 		if (arenas_cache == NULL) {
 			/*
 			 * This function must always tell the truth, even if
-			 * it's slow, so don't let OOM or recursive allocation
+			 * it's slow, so don't let OOM, thread cleanup (note
+			 * tsd_nominal check), nor recursive allocation
 			 * avoidance (note arenas_cache_bypass check) get in the
 			 * way.
 			 */
@@ -531,6 +542,7 @@ arena_get_hard(tsd_t *tsd, unsigned ind, bool init_if_missing)
 			malloc_mutex_unlock(&arenas_lock);
 			return (arena);
 		}
+		assert(tsd_nominal(tsd) && !*arenas_cache_bypassp);
 		tsd_arenas_cache_set(tsd, arenas_cache);
 		tsd_narenas_cache_set(tsd, narenas_cache);
 	}
@@ -649,9 +661,11 @@ arenas_cache_cleanup(tsd_t *tsd)
 	arena_t **arenas_cache;
 
 	arenas_cache = tsd_arenas_cache_get(tsd);
-	if (arenas_cache != NULL)
+	if (arenas_cache != NULL) {
+		tsd_arenas_cache_set(tsd, NULL);
 		a0dalloc(arenas_cache);
 	}
+}
 
 void
 narenas_cache_cleanup(tsd_t *tsd)
@@ -1297,6 +1311,9 @@ static bool
 malloc_init_hard(void)
 {
 
+#if defined(_WIN32) && _WIN32_WINNT < 0x0600
+	_init_init_lock();
+#endif
 	malloc_mutex_lock(&init_lock);
 	if (!malloc_init_hard_needed()) {
 		malloc_mutex_unlock(&init_lock);
@@ -1361,7 +1378,7 @@ imalloc_prof(tsd_t *tsd, size_t usize)
 	void *p;
 	prof_tctx_t *tctx;
 
-	tctx = prof_alloc_prep(tsd, usize, true);
+	tctx = prof_alloc_prep(tsd, usize, prof_active_get_unlocked(), true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U))
 		p = imalloc_prof_sample(tsd, usize, tctx);
 	else
@@ -1451,7 +1468,7 @@ imemalign_prof(tsd_t *tsd, size_t alignment, size_t usize)
 	void *p;
 	prof_tctx_t *tctx;
 
-	tctx = prof_alloc_prep(tsd, usize, true);
+	tctx = prof_alloc_prep(tsd, usize, prof_active_get_unlocked(), true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U))
 		p = imemalign_prof_sample(tsd, alignment, usize, tctx);
 	else
@@ -1582,7 +1599,7 @@ icalloc_prof(tsd_t *tsd, size_t usize)
 	void *p;
 	prof_tctx_t *tctx;
 
-	tctx = prof_alloc_prep(tsd, usize, true);
+	tctx = prof_alloc_prep(tsd, usize, prof_active_get_unlocked(), true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U))
 		p = icalloc_prof_sample(tsd, usize, tctx);
 	else
@@ -1665,7 +1682,7 @@ label_return:
 }
 
 static void *
-irealloc_prof_sample(tsd_t *tsd, void *oldptr, size_t old_usize, size_t usize,
+irealloc_prof_sample(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize,
     prof_tctx_t *tctx)
 {
 	void *p;
@@ -1673,31 +1690,36 @@ irealloc_prof_sample(tsd_t *tsd, void *oldptr, size_t old_usize, size_t usize,
 	if (tctx == NULL)
 		return (NULL);
 	if (usize <= SMALL_MAXCLASS) {
-		p = iralloc(tsd, oldptr, old_usize, LARGE_MINCLASS, 0, false);
+		p = iralloc(tsd, old_ptr, old_usize, LARGE_MINCLASS, 0, false);
 		if (p == NULL)
 			return (NULL);
 		arena_prof_promoted(p, usize);
 	} else
-		p = iralloc(tsd, oldptr, old_usize, usize, 0, false);
+		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false);
 
 	return (p);
 }
 
 JEMALLOC_ALWAYS_INLINE_C void *
-irealloc_prof(tsd_t *tsd, void *oldptr, size_t old_usize, size_t usize)
+irealloc_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t usize)
 {
 	void *p;
+	bool prof_active;
 	prof_tctx_t *old_tctx, *tctx;
 
-	old_tctx = prof_tctx_get(oldptr);
+	prof_active = prof_active_get_unlocked();
-	tctx = prof_alloc_prep(tsd, usize, true);
+	old_tctx = prof_tctx_get(old_ptr);
+	tctx = prof_alloc_prep(tsd, usize, prof_active, true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U))
-		p = irealloc_prof_sample(tsd, oldptr, old_usize, usize, tctx);
+		p = irealloc_prof_sample(tsd, old_ptr, old_usize, usize, tctx);
 	else
-		p = iralloc(tsd, oldptr, old_usize, usize, 0, false);
+		p = iralloc(tsd, old_ptr, old_usize, usize, 0, false);
-	if (p == NULL)
+	if (unlikely(p == NULL)) {
+		prof_alloc_rollback(tsd, tctx, true);
 		return (NULL);
-	prof_realloc(tsd, p, usize, tctx, true, old_usize, old_tctx);
+	}
+	prof_realloc(tsd, p, usize, tctx, prof_active, true, old_ptr, old_usize,
+	    old_tctx);
 
 	return (p);
 }
@@ -1995,7 +2017,7 @@ imallocx_prof(tsd_t *tsd, size_t size, int flags, size_t *usize)
 	if (unlikely(imallocx_flags_decode(tsd, size, flags, usize, &alignment,
 	    &zero, &tcache, &arena)))
 		return (NULL);
-	tctx = prof_alloc_prep(tsd, *usize, true);
+	tctx = prof_alloc_prep(tsd, *usize, prof_active_get_unlocked(), true);
 	if (likely((uintptr_t)tctx == (uintptr_t)1U)) {
 		p = imallocx_maybe_flags(tsd, size, flags, *usize, alignment,
 		    zero, tcache, arena);
@@ -2076,7 +2098,7 @@ label_oom:
 }
 
 static void *
-irallocx_prof_sample(tsd_t *tsd, void *oldptr, size_t old_usize, size_t size,
+irallocx_prof_sample(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t size,
     size_t alignment, size_t usize, bool zero, tcache_t *tcache, arena_t *arena,
     prof_tctx_t *tctx)
 {
@@ -2085,13 +2107,13 @@ irallocx_prof_sample(tsd_t *tsd, void *oldptr, size_t old_usize, size_t size,
 	if (tctx == NULL)
 		return (NULL);
 	if (usize <= SMALL_MAXCLASS) {
-		p = iralloct(tsd, oldptr, old_usize, LARGE_MINCLASS, alignment,
+		p = iralloct(tsd, old_ptr, old_usize, LARGE_MINCLASS, alignment,
 		    zero, tcache, arena);
 		if (p == NULL)
 			return (NULL);
 		arena_prof_promoted(p, usize);
 	} else {
-		p = iralloct(tsd, oldptr, old_usize, size, alignment, zero,
+		p = iralloct(tsd, old_ptr, old_usize, size, alignment, zero,
 		    tcache, arena);
 	}
 
@@ -2099,28 +2121,30 @@ irallocx_prof_sample(tsd_t *tsd, void *oldptr, size_t old_usize, size_t size,
 }
 
 JEMALLOC_ALWAYS_INLINE_C void *
-irallocx_prof(tsd_t *tsd, void *oldptr, size_t old_usize, size_t size,
+irallocx_prof(tsd_t *tsd, void *old_ptr, size_t old_usize, size_t size,
     size_t alignment, size_t *usize, bool zero, tcache_t *tcache,
     arena_t *arena)
 {
 	void *p;
+	bool prof_active;
 	prof_tctx_t *old_tctx, *tctx;
 
-	old_tctx = prof_tctx_get(oldptr);
+	prof_active = prof_active_get_unlocked();
-	tctx = prof_alloc_prep(tsd, *usize, false);
+	old_tctx = prof_tctx_get(old_ptr);
+	tctx = prof_alloc_prep(tsd, *usize, prof_active, true);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
-		p = irallocx_prof_sample(tsd, oldptr, old_usize, size,
+		p = irallocx_prof_sample(tsd, old_ptr, old_usize, size,
 		    alignment, *usize, zero, tcache, arena, tctx);
 	} else {
-		p = iralloct(tsd, oldptr, old_usize, size, alignment, zero,
+		p = iralloct(tsd, old_ptr, old_usize, size, alignment, zero,
 		    tcache, arena);
 	}
 	if (unlikely(p == NULL)) {
-		prof_alloc_rollback(tsd, tctx, false);
+		prof_alloc_rollback(tsd, tctx, true);
 		return (NULL);
 	}
 
-	if (p == oldptr && alignment != 0) {
+	if (p == old_ptr && alignment != 0) {
 		/*
 		 * The allocation did not move, so it is possible that the size
 		 * class is smaller than would guarantee the requested
@@ -2131,7 +2155,8 @@ irallocx_prof(tsd_t *tsd, void *oldptr, size_t old_usize, size_t size,
 		 */
 		*usize = isalloc(p, config_prof);
 	}
-	prof_realloc(tsd, p, *usize, tctx, false, old_usize, old_tctx);
+	prof_realloc(tsd, p, *usize, tctx, prof_active, true, old_ptr,
+	    old_usize, old_tctx);
 
 	return (p);
 }
@@ -2226,7 +2251,7 @@ ixallocx_helper(void *ptr, size_t old_usize, size_t size, size_t extra,
 
 static size_t
 ixallocx_prof_sample(void *ptr, size_t old_usize, size_t size, size_t extra,
-    size_t alignment, size_t max_usize, bool zero, prof_tctx_t *tctx)
+    size_t alignment, size_t usize_max, bool zero, prof_tctx_t *tctx)
 {
 	size_t usize;
 
@@ -2240,7 +2265,7 @@ ixallocx_prof_sample(void *ptr, size_t old_usize, size_t size, size_t extra,
 		    (SMALL_MAXCLASS+1), alignment, zero))
 			return (old_usize);
 		usize = isalloc(ptr, config_prof);
-		if (max_usize < LARGE_MINCLASS)
+		if (usize_max < LARGE_MINCLASS)
 			arena_prof_promoted(ptr, usize);
 	} else {
 		usize = ixallocx_helper(ptr, old_usize, size, extra, alignment,
@@ -2254,9 +2279,11 @@ JEMALLOC_ALWAYS_INLINE_C size_t
 ixallocx_prof(tsd_t *tsd, void *ptr, size_t old_usize, size_t size,
     size_t extra, size_t alignment, bool zero)
 {
-	size_t max_usize, usize;
+	size_t usize_max, usize;
+	bool prof_active;
 	prof_tctx_t *old_tctx, *tctx;
 
+	prof_active = prof_active_get_unlocked();
 	old_tctx = prof_tctx_get(ptr);
 	/*
 	 * usize isn't knowable before ixalloc() returns when extra is non-zero.
@@ -2264,12 +2291,12 @@ ixallocx_prof(tsd_t *tsd, void *ptr, size_t old_usize, size_t size,
 	 * prof_alloc_prep() to decide whether to capture a backtrace.
 	 * prof_realloc() will use the actual usize to decide whether to sample.
 	 */
-	max_usize = (alignment == 0) ? s2u(size+extra) : sa2u(size+extra,
+	usize_max = (alignment == 0) ? s2u(size+extra) : sa2u(size+extra,
 	    alignment);
-	tctx = prof_alloc_prep(tsd, max_usize, false);
+	tctx = prof_alloc_prep(tsd, usize_max, prof_active, false);
 	if (unlikely((uintptr_t)tctx != (uintptr_t)1U)) {
 		usize = ixallocx_prof_sample(ptr, old_usize, size, extra,
-		    alignment, zero, max_usize, tctx);
+		    alignment, usize_max, zero, tctx);
 	} else {
 		usize = ixallocx_helper(ptr, old_usize, size, extra, alignment,
 		    zero);
@@ -2278,7 +2305,8 @@ ixallocx_prof(tsd_t *tsd, void *ptr, size_t old_usize, size_t size,
 		prof_alloc_rollback(tsd, tctx, false);
 		return (usize);
 	}
-	prof_realloc(tsd, ptr, usize, tctx, false, old_usize, old_tctx);
+	prof_realloc(tsd, ptr, usize, tctx, prof_active, false, ptr, old_usize,
+	    old_tctx);
 
 	return (usize);
 }
@@ -2300,6 +2328,17 @@ je_xallocx(void *ptr, size_t size, size_t extra, int flags)
 	tsd = tsd_fetch();
 
 	old_usize = isalloc(ptr, config_prof);
+
+	/* Clamp extra if necessary to avoid (size + extra) overflow. */
+	if (unlikely(size + extra > HUGE_MAXCLASS)) {
+		/* Check for size overflow. */
+		if (unlikely(size > HUGE_MAXCLASS)) {
+			usize = old_usize;
+			goto label_not_resized;
+		}
+		extra = HUGE_MAXCLASS - size;
+	}
+
 	if (config_valgrind && unlikely(in_valgrind))
 		old_rzsize = u2rz(old_usize);
 

src/prof.c

@@ -138,10 +138,17 @@ prof_tctx_comp(const prof_tctx_t *a, const prof_tctx_t *b)
 	uint64_t a_thr_uid = a->thr_uid;
 	uint64_t b_thr_uid = b->thr_uid;
 	int ret = (a_thr_uid > b_thr_uid) - (a_thr_uid < b_thr_uid);
+	if (ret == 0) {
+		uint64_t a_thr_discrim = a->thr_discrim;
+		uint64_t b_thr_discrim = b->thr_discrim;
+		ret = (a_thr_discrim > b_thr_discrim) - (a_thr_discrim <
+		    b_thr_discrim);
 		if (ret == 0) {
 			uint64_t a_tctx_uid = a->tctx_uid;
 			uint64_t b_tctx_uid = b->tctx_uid;
-		ret = (a_tctx_uid > b_tctx_uid) - (a_tctx_uid < b_tctx_uid);
+			ret = (a_tctx_uid > b_tctx_uid) - (a_tctx_uid <
+			    b_tctx_uid);
+		}
 	}
 	return (ret);
 }
@@ -219,7 +226,7 @@ void
 prof_malloc_sample_object(const void *ptr, size_t usize, prof_tctx_t *tctx)
 {
 
-	prof_tctx_set(ptr, tctx);
+	prof_tctx_set(ptr, usize, tctx);
 
 	malloc_mutex_lock(tctx->tdata->lock);
 	tctx->cnts.curobjs++;
@@ -791,6 +798,7 @@ prof_lookup(tsd_t *tsd, prof_bt_t *bt)
 		}
 		ret.p->tdata = tdata;
 		ret.p->thr_uid = tdata->thr_uid;
+		ret.p->thr_discrim = tdata->thr_discrim;
 		memset(&ret.p->cnts, 0, sizeof(prof_cnt_t));
 		ret.p->gctx = gctx;
 		ret.p->tctx_uid = tdata->tctx_uid_next++;
@@ -1569,7 +1577,6 @@ prof_idump(void)
 {
 	tsd_t *tsd;
 	prof_tdata_t *tdata;
-	char filename[PATH_MAX + 1];
 
 	cassert(config_prof);
 
@@ -1585,6 +1592,7 @@ prof_idump(void)
 	}
 
 	if (opt_prof_prefix[0] != '\0') {
+		char filename[PATH_MAX + 1];
 		malloc_mutex_lock(&prof_dump_seq_mtx);
 		prof_dump_filename(filename, 'i', prof_dump_iseq);
 		prof_dump_iseq++;
@@ -1623,7 +1631,6 @@ prof_gdump(void)
 {
 	tsd_t *tsd;
 	prof_tdata_t *tdata;
-	char filename[DUMP_FILENAME_BUFSIZE];
 
 	cassert(config_prof);
 
@@ -1639,6 +1646,7 @@ prof_gdump(void)
 	}
 
 	if (opt_prof_prefix[0] != '\0') {
+		char filename[DUMP_FILENAME_BUFSIZE];
 		malloc_mutex_lock(&prof_dump_seq_mtx);
 		prof_dump_filename(filename, 'u', prof_dump_useq);
 		prof_dump_useq++;

src/tcache.c

@@ -32,7 +32,7 @@ size_t	tcache_salloc(const void *ptr)
 void
 tcache_event_hard(tsd_t *tsd, tcache_t *tcache)
 {
-	index_t binind = tcache->next_gc_bin;
+	szind_t binind = tcache->next_gc_bin;
 	tcache_bin_t *tbin = &tcache->tbins[binind];
 	tcache_bin_info_t *tbin_info = &tcache_bin_info[binind];
 
@@ -72,7 +72,7 @@ tcache_event_hard(tsd_t *tsd, tcache_t *tcache)
 
 void *
 tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
-    tcache_bin_t *tbin, index_t binind)
+    tcache_bin_t *tbin, szind_t binind)
 {
 	void *ret;
 
@@ -87,7 +87,7 @@ tcache_alloc_small_hard(tsd_t *tsd, arena_t *arena, tcache_t *tcache,
 
 void
 tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
-    index_t binind, unsigned rem)
+    szind_t binind, unsigned rem)
 {
 	arena_t *arena;
 	void *ptr;
@@ -166,7 +166,7 @@ tcache_bin_flush_small(tsd_t *tsd, tcache_t *tcache, tcache_bin_t *tbin,
 }
 
 void
-tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, index_t binind,
+tcache_bin_flush_large(tsd_t *tsd, tcache_bin_t *tbin, szind_t binind,
     unsigned rem, tcache_t *tcache)
 {
 	arena_t *arena;
@@ -496,13 +496,13 @@ tcache_boot(void)
 	unsigned i;
 
 	/*
-	 * If necessary, clamp opt_lg_tcache_max, now that arena_maxclass is
+	 * If necessary, clamp opt_lg_tcache_max, now that large_maxclass is
 	 * known.
 	 */
 	if (opt_lg_tcache_max < 0 || (1U << opt_lg_tcache_max) < SMALL_MAXCLASS)
 		tcache_maxclass = SMALL_MAXCLASS;
-	else if ((1U << opt_lg_tcache_max) > arena_maxclass)
+	else if ((1U << opt_lg_tcache_max) > large_maxclass)
-		tcache_maxclass = arena_maxclass;
+		tcache_maxclass = large_maxclass;
 	else
 		tcache_maxclass = (1U << opt_lg_tcache_max);
 

test/integration/chunk.c

@@ -1,5 +1,9 @@
 #include "test/jemalloc_test.h"
 
+#ifdef JEMALLOC_FILL
+const char *malloc_conf = "junk:false";
+#endif
+
 static chunk_hooks_t orig_hooks;
 static chunk_hooks_t old_hooks;
 

test/integration/rallocx.c

@@ -22,7 +22,7 @@ TEST_BEGIN(test_grow_and_shrink)
 			    szs[j-1], szs[j-1]+1);
 			szs[j] = sallocx(q, 0);
 			assert_zu_ne(szs[j], szs[j-1]+1,
-			    "Expected size to at least: %zu", szs[j-1]+1);
+			    "Expected size to be at least: %zu", szs[j-1]+1);
 			p = q;
 		}
 

test/integration/xallocx.c

@@ -48,6 +48,305 @@ TEST_BEGIN(test_no_move_fail)
 }
 TEST_END
 
+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_size)
+{
+	size_t small0, hugemax;
+	void *p;
+
+	/* Get size classes. */
+	small0 = get_small_size(0);
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(small0, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+
+	/* Test smallest supported size. */
+	assert_zu_eq(xallocx(p, 1, 0, 0), small0,
+	    "Unexpected xallocx() behavior");
+
+	/* Test largest supported size. */
+	assert_zu_le(xallocx(p, hugemax, 0, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	/* Test size overflow. */
+	assert_zu_le(xallocx(p, hugemax+1, 0, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, SIZE_T_MAX, 0, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	dallocx(p, 0);
+}
+TEST_END
+
+TEST_BEGIN(test_size_extra_overflow)
+{
+	size_t small0, hugemax;
+	void *p;
+
+	/* Get size classes. */
+	small0 = get_small_size(0);
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(small0, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+
+	/* Test overflows that can be resolved by clamping extra. */
+	assert_zu_le(xallocx(p, hugemax-1, 2, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, hugemax, 1, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	/* Test overflow such that hugemax-size underflows. */
+	assert_zu_le(xallocx(p, hugemax+1, 2, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, hugemax+2, 3, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, SIZE_T_MAX-2, 2, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, SIZE_T_MAX-1, 1, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	dallocx(p, 0);
+}
+TEST_END
+
+TEST_BEGIN(test_extra_small)
+{
+	size_t small0, small1, hugemax;
+	void *p;
+
+	/* Get size classes. */
+	small0 = get_small_size(0);
+	small1 = get_small_size(1);
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(small0, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+
+	assert_zu_eq(xallocx(p, small1, 0, 0), small0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, small1, 0, 0), small0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, small0, small1 - small0, 0), small0,
+	    "Unexpected xallocx() behavior");
+
+	/* Test size+extra overflow. */
+	assert_zu_eq(xallocx(p, small0, hugemax - small0 + 1, 0), small0,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, small0, SIZE_T_MAX - small0, 0), small0,
+	    "Unexpected xallocx() behavior");
+
+	dallocx(p, 0);
+}
+TEST_END
+
+TEST_BEGIN(test_extra_large)
+{
+	size_t smallmax, large0, large1, large2, huge0, hugemax;
+	void *p;
+
+	/* Get size classes. */
+	smallmax = get_small_size(get_nsmall()-1);
+	large0 = get_large_size(0);
+	large1 = get_large_size(1);
+	large2 = get_large_size(2);
+	huge0 = get_huge_size(0);
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(large2, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+
+	assert_zu_eq(xallocx(p, large2, 0, 0), large2,
+	    "Unexpected xallocx() behavior");
+	/* Test size decrease with zero extra. */
+	assert_zu_eq(xallocx(p, large0, 0, 0), large0,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, smallmax, 0, 0), large0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, large2, 0, 0), large2,
+	    "Unexpected xallocx() behavior");
+	/* Test size decrease with non-zero extra. */
+	assert_zu_eq(xallocx(p, large0, large2 - large0, 0), large2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, large1, large2 - large1, 0), large2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, large0, large1 - large0, 0), large1,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, smallmax, large0 - smallmax, 0), large0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, large0, 0, 0), large0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with zero extra. */
+	assert_zu_eq(xallocx(p, large2, 0, 0), large2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, huge0, 0, 0), large2,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, large0, 0, 0), large0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with non-zero extra. */
+	assert_zu_lt(xallocx(p, large0, huge0 - large0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, large0, 0, 0), large0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with non-zero extra. */
+	assert_zu_eq(xallocx(p, large0, large2 - large0, 0), large2,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, large2, 0, 0), large2,
+	    "Unexpected xallocx() behavior");
+	/* Test size+extra overflow. */
+	assert_zu_lt(xallocx(p, large2, hugemax - large2 + 1, 0), huge0,
+	    "Unexpected xallocx() behavior");
+
+	dallocx(p, 0);
+}
+TEST_END
+
+TEST_BEGIN(test_extra_huge)
+{
+	size_t largemax, huge0, huge1, huge2, hugemax;
+	void *p;
+
+	/* Get size classes. */
+	largemax = get_large_size(get_nlarge()-1);
+	huge0 = get_huge_size(0);
+	huge1 = get_huge_size(1);
+	huge2 = get_huge_size(2);
+	hugemax = get_huge_size(get_nhuge()-1);
+
+	p = mallocx(huge2, 0);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+
+	assert_zu_eq(xallocx(p, huge2, 0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	/* Test size decrease with zero extra. */
+	assert_zu_ge(xallocx(p, huge0, 0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+	assert_zu_ge(xallocx(p, largemax, 0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, huge2, 0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	/* Test size decrease with non-zero extra. */
+	assert_zu_eq(xallocx(p, huge0, huge2 - huge0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, huge1, huge2 - huge1, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_eq(xallocx(p, huge0, huge1 - huge0, 0), huge1,
+	    "Unexpected xallocx() behavior");
+	assert_zu_ge(xallocx(p, largemax, huge0 - largemax, 0), huge0,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_ge(xallocx(p, huge0, 0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with zero extra. */
+	assert_zu_le(xallocx(p, huge2, 0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	assert_zu_le(xallocx(p, hugemax+1, 0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_ge(xallocx(p, huge0, 0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with non-zero extra. */
+	assert_zu_le(xallocx(p, huge0, SIZE_T_MAX - huge0, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_ge(xallocx(p, huge0, 0, 0), huge0,
+	    "Unexpected xallocx() behavior");
+	/* Test size increase with non-zero extra. */
+	assert_zu_le(xallocx(p, huge0, huge2 - huge0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+
+	assert_zu_eq(xallocx(p, huge2, 0, 0), huge2,
+	    "Unexpected xallocx() behavior");
+	/* Test size+extra overflow. */
+	assert_zu_le(xallocx(p, huge2, hugemax - huge2 + 1, 0), hugemax,
+	    "Unexpected xallocx() behavior");
+
+	dallocx(p, 0);
+}
+TEST_END
+
 int
 main(void)
 {
@@ -55,5 +354,10 @@ main(void)
 	return (test(
 	    test_same_size,
 	    test_extra_no_move,
-	    test_no_move_fail));
+	    test_no_move_fail,
+	    test_size,
+	    test_size_extra_overflow,
+	    test_extra_small,
+	    test_extra_large,
+	    test_extra_huge));
 }

test/unit/junk.c

@@ -140,7 +140,7 @@ TEST_BEGIN(test_junk_large)
 {
 
 	test_skip_if(!config_fill);
-	test_junk(SMALL_MAXCLASS+1, arena_maxclass);
+	test_junk(SMALL_MAXCLASS+1, large_maxclass);
 }
 TEST_END
 
@@ -148,7 +148,7 @@ TEST_BEGIN(test_junk_huge)
 {
 
 	test_skip_if(!config_fill);
-	test_junk(arena_maxclass+1, chunksize*2);
+	test_junk(large_maxclass+1, chunksize*2);
 }
 TEST_END
 
@@ -172,8 +172,8 @@ arena_ralloc_junk_large_intercept(void *ptr, size_t old_usize, size_t usize)
 {
 
 	arena_ralloc_junk_large_orig(ptr, old_usize, usize);
-	assert_zu_eq(old_usize, arena_maxclass, "Unexpected old_usize");
+	assert_zu_eq(old_usize, large_maxclass, "Unexpected old_usize");
-	assert_zu_eq(usize, shrink_size(arena_maxclass), "Unexpected usize");
+	assert_zu_eq(usize, shrink_size(large_maxclass), "Unexpected usize");
 	most_recently_trimmed = ptr;
 }
 
@@ -181,13 +181,13 @@ TEST_BEGIN(test_junk_large_ralloc_shrink)
 {
 	void *p1, *p2;
 
-	p1 = mallocx(arena_maxclass, 0);
+	p1 = mallocx(large_maxclass, 0);
 	assert_ptr_not_null(p1, "Unexpected mallocx() failure");
 
 	arena_ralloc_junk_large_orig = arena_ralloc_junk_large;
 	arena_ralloc_junk_large = arena_ralloc_junk_large_intercept;
 
-	p2 = rallocx(p1, shrink_size(arena_maxclass), 0);
+	p2 = rallocx(p1, shrink_size(large_maxclass), 0);
 	assert_ptr_eq(p1, p2, "Unexpected move during shrink");
 
 	arena_ralloc_junk_large = arena_ralloc_junk_large_orig;

test/unit/prof_reset.c

@@ -16,6 +16,35 @@ prof_dump_open_intercept(bool propagate_err, const char *filename)
 	return (fd);
 }
 
+static void
+set_prof_active(bool active)
+{
+
+	assert_d_eq(mallctl("prof.active", NULL, NULL, &active, sizeof(active)),
+	    0, "Unexpected mallctl failure");
+}
+
+static size_t
+get_lg_prof_sample(void)
+{
+	size_t lg_prof_sample;
+	size_t sz = sizeof(size_t);
+
+	assert_d_eq(mallctl("prof.lg_sample", &lg_prof_sample, &sz, NULL, 0), 0,
+	    "Unexpected mallctl failure while reading profiling sample rate");
+	return (lg_prof_sample);
+}
+
+static void
+do_prof_reset(size_t lg_prof_sample)
+{
+	assert_d_eq(mallctl("prof.reset", NULL, NULL,
+	    &lg_prof_sample, sizeof(size_t)), 0,
+	    "Unexpected mallctl failure while resetting profile data");
+	assert_zu_eq(lg_prof_sample, get_lg_prof_sample(),
+	    "Expected profile sample rate change");
+}
+
 TEST_BEGIN(test_prof_reset_basic)
 {
 	size_t lg_prof_sample_orig, lg_prof_sample, lg_prof_sample_next;
@@ -30,9 +59,7 @@ TEST_BEGIN(test_prof_reset_basic)
 	    "Unexpected mallctl failure while reading profiling sample rate");
 	assert_zu_eq(lg_prof_sample_orig, 0,
 	    "Unexpected profiling sample rate");
-	sz = sizeof(size_t);
+	lg_prof_sample = get_lg_prof_sample();
-	assert_d_eq(mallctl("prof.lg_sample", &lg_prof_sample, &sz, NULL, 0), 0,
-	    "Unexpected mallctl failure while reading profiling sample rate");
 	assert_zu_eq(lg_prof_sample_orig, lg_prof_sample,
 	    "Unexpected disagreement between \"opt.lg_prof_sample\" and "
 	    "\"prof.lg_sample\"");
@@ -41,10 +68,7 @@ TEST_BEGIN(test_prof_reset_basic)
 	for (i = 0; i < 2; i++) {
 		assert_d_eq(mallctl("prof.reset", NULL, NULL, NULL, 0), 0,
 		    "Unexpected mallctl failure while resetting profile data");
-		sz = sizeof(size_t);
+		lg_prof_sample = get_lg_prof_sample();
-		assert_d_eq(mallctl("prof.lg_sample", &lg_prof_sample, &sz,
-		    NULL, 0), 0, "Unexpected mallctl failure while reading "
-		    "profiling sample rate");
 		assert_zu_eq(lg_prof_sample_orig, lg_prof_sample,
 		    "Unexpected profile sample rate change");
 	}
@@ -52,22 +76,15 @@ TEST_BEGIN(test_prof_reset_basic)
 	/* Test resets with prof.lg_sample changes. */
 	lg_prof_sample_next = 1;
 	for (i = 0; i < 2; i++) {
-		assert_d_eq(mallctl("prof.reset", NULL, NULL,
+		do_prof_reset(lg_prof_sample_next);
-		    &lg_prof_sample_next, sizeof(size_t)), 0,
+		lg_prof_sample = get_lg_prof_sample();
-		    "Unexpected mallctl failure while resetting profile data");
-		sz = sizeof(size_t);
-		assert_d_eq(mallctl("prof.lg_sample", &lg_prof_sample, &sz,
-		    NULL, 0), 0, "Unexpected mallctl failure while reading "
-		    "profiling sample rate");
 		assert_zu_eq(lg_prof_sample, lg_prof_sample_next,
 		    "Expected profile sample rate change");
 		lg_prof_sample_next = lg_prof_sample_orig;
 	}
 
 	/* Make sure the test code restored prof.lg_sample. */
-	sz = sizeof(size_t);
+	lg_prof_sample = get_lg_prof_sample();
-	assert_d_eq(mallctl("prof.lg_sample", &lg_prof_sample, &sz, NULL, 0), 0,
-	    "Unexpected mallctl failure while reading profiling sample rate");
 	assert_zu_eq(lg_prof_sample_orig, lg_prof_sample,
 	    "Unexpected disagreement between \"opt.lg_prof_sample\" and "
 	    "\"prof.lg_sample\"");
@@ -88,15 +105,12 @@ prof_dump_header_intercept(bool propagate_err, const prof_cnt_t *cnt_all)
 
 TEST_BEGIN(test_prof_reset_cleanup)
 {
-	bool active;
 	void *p;
 	prof_dump_header_t *prof_dump_header_orig;
 
 	test_skip_if(!config_prof);
 
-	active = true;
+	set_prof_active(true);
-	assert_d_eq(mallctl("prof.active", NULL, NULL, &active, sizeof(active)),
-	    0, "Unexpected mallctl failure while activating profiling");
 
 	assert_zu_eq(prof_bt_count(), 0, "Expected 0 backtraces");
 	p = mallocx(1, 0);
@@ -124,9 +138,7 @@ TEST_BEGIN(test_prof_reset_cleanup)
 	dallocx(p, 0);
 	assert_zu_eq(prof_bt_count(), 0, "Expected 0 backtraces");
 
-	active = false;
+	set_prof_active(false);
-	assert_d_eq(mallctl("prof.active", NULL, NULL, &active, sizeof(active)),
-	    0, "Unexpected mallctl failure while deactivating profiling");
 }
 TEST_END
 
@@ -182,7 +194,7 @@ thd_start(void *varg)
 
 TEST_BEGIN(test_prof_reset)
 {
-	bool active;
+	size_t lg_prof_sample_orig;
 	thd_t thds[NTHREADS];
 	unsigned thd_args[NTHREADS];
 	unsigned i;
@@ -195,9 +207,10 @@ TEST_BEGIN(test_prof_reset)
 	    "Unexpected pre-existing tdata structures");
 	tdata_count = prof_tdata_count();
 
-	active = true;
+	lg_prof_sample_orig = get_lg_prof_sample();
-	assert_d_eq(mallctl("prof.active", NULL, NULL, &active, sizeof(active)),
+	do_prof_reset(5);
-	    0, "Unexpected mallctl failure while activating profiling");
+
+	set_prof_active(true);
 
 	for (i = 0; i < NTHREADS; i++) {
 		thd_args[i] = i;
@@ -211,9 +224,9 @@ TEST_BEGIN(test_prof_reset)
 	assert_zu_eq(prof_tdata_count(), tdata_count,
 	    "Unexpected remaining tdata structures");
 
-	active = false;
+	set_prof_active(false);
-	assert_d_eq(mallctl("prof.active", NULL, NULL, &active, sizeof(active)),
+
-	    0, "Unexpected mallctl failure while deactivating profiling");
+	do_prof_reset(lg_prof_sample_orig);
 }
 TEST_END
 #undef NTHREADS
@@ -222,6 +235,58 @@ TEST_END
 #undef RESET_INTERVAL
 #undef DUMP_INTERVAL
 
+/* Test sampling at the same allocation site across resets. */
+#define	NITER 10
+TEST_BEGIN(test_xallocx)
+{
+	size_t lg_prof_sample_orig;
+	unsigned i;
+	void *ptrs[NITER];
+
+	test_skip_if(!config_prof);
+
+	lg_prof_sample_orig = get_lg_prof_sample();
+	set_prof_active(true);
+
+	/* Reset profiling. */
+	do_prof_reset(0);
+
+	for (i = 0; i < NITER; i++) {
+		void *p;
+		size_t sz, nsz;
+
+		/* Reset profiling. */
+		do_prof_reset(0);
+
+		/* Allocate small object (which will be promoted). */
+		p = ptrs[i] = mallocx(1, 0);
+		assert_ptr_not_null(p, "Unexpected mallocx() failure");
+
+		/* Reset profiling. */
+		do_prof_reset(0);
+
+		/* Perform successful xallocx(). */
+		sz = sallocx(p, 0);
+		assert_zu_eq(xallocx(p, sz, 0, 0), sz,
+		    "Unexpected xallocx() failure");
+
+		/* Perform unsuccessful xallocx(). */
+		nsz = nallocx(sz+1, 0);
+		assert_zu_eq(xallocx(p, nsz, 0, 0), sz,
+		    "Unexpected xallocx() success");
+	}
+
+	for (i = 0; i < NITER; i++) {
+		/* dallocx. */
+		dallocx(ptrs[i], 0);
+	}
+
+	set_prof_active(false);
+	do_prof_reset(lg_prof_sample_orig);
+}
+TEST_END
+#undef NITER
+
 int
 main(void)
 {
@@ -232,5 +297,6 @@ main(void)
 	return (test(
 	    test_prof_reset_basic,
 	    test_prof_reset_cleanup,
-	    test_prof_reset));
+	    test_prof_reset,
+	    test_xallocx));
 }

test/unit/size_classes.c

@@ -26,7 +26,7 @@ get_max_size_class(void)
 TEST_BEGIN(test_size_classes)
 {
 	size_t size_class, max_size_class;
-	index_t index, max_index;
+	szind_t index, max_index;
 
 	max_size_class = get_max_size_class();
 	max_index = size2index(max_size_class);

test/unit/stats.c

@@ -42,7 +42,7 @@ TEST_BEGIN(test_stats_huge)
 	size_t sz;
 	int expected = config_stats ? 0 : ENOENT;
 
-	p = mallocx(arena_maxclass+1, 0);
+	p = mallocx(large_maxclass+1, 0);
 	assert_ptr_not_null(p, "Unexpected mallocx() failure");
 
 	assert_d_eq(mallctl("epoch", NULL, NULL, &epoch, sizeof(epoch)), 0,
@@ -88,7 +88,7 @@ TEST_BEGIN(test_stats_arenas_summary)
 
 	little = mallocx(SMALL_MAXCLASS, 0);
 	assert_ptr_not_null(little, "Unexpected mallocx() failure");
-	large = mallocx(arena_maxclass, 0);
+	large = mallocx(large_maxclass, 0);
 	assert_ptr_not_null(large, "Unexpected mallocx() failure");
 	huge = mallocx(chunksize, 0);
 	assert_ptr_not_null(huge, "Unexpected mallocx() failure");
@@ -200,7 +200,7 @@ TEST_BEGIN(test_stats_arenas_large)
 	assert_d_eq(mallctl("thread.arena", NULL, NULL, &arena, sizeof(arena)),
 	    0, "Unexpected mallctl() failure");
 
-	p = mallocx(arena_maxclass, 0);
+	p = mallocx(large_maxclass, 0);
 	assert_ptr_not_null(p, "Unexpected mallocx() failure");
 
 	assert_d_eq(mallctl("epoch", NULL, NULL, &epoch, sizeof(epoch)), 0,

test/unit/tsd.c

@@ -56,9 +56,14 @@ static void *
 thd_start(void *arg)
 {
 	data_t d = (data_t)(uintptr_t)arg;
+	void *p;
+
 	assert_x_eq(*data_tsd_get(), DATA_INIT,
 	    "Initial tsd get should return initialization value");
 
+	p = malloc(1);
+	assert_ptr_not_null(p, "Unexpected malloc() failure");
+
 	data_tsd_set(&d);
 	assert_x_eq(*data_tsd_get(), d,
 	    "After tsd set, tsd get should return value that was set");
@@ -67,6 +72,7 @@ thd_start(void *arg)
 	assert_x_eq(*data_tsd_get(), (data_t)(uintptr_t)arg,
 	    "Resetting local data should have no effect on tsd");
 
+	free(p);
 	return (NULL);
 }
 

test/unit/zero.c

@@ -55,7 +55,7 @@ TEST_BEGIN(test_zero_large)
 {
 
 	test_skip_if(!config_fill);
-	test_zero(SMALL_MAXCLASS+1, arena_maxclass);
+	test_zero(SMALL_MAXCLASS+1, large_maxclass);
 }
 TEST_END
 
@@ -63,7 +63,7 @@ TEST_BEGIN(test_zero_huge)
 {
 
 	test_skip_if(!config_fill);
-	test_zero(arena_maxclass+1, chunksize*2);
+	test_zero(large_maxclass+1, chunksize*2);
 }
 TEST_END