Fix xallocx(..., MALLOCX_ZERO) bugs.

Zero all trailing bytes of large allocations when
--enable-cache-oblivious configure option is enabled.  This regression
was introduced by 8a03cf039c (Implement
cache index randomization for large allocations.).

Zero trailing bytes of huge allocations when resizing from/to a size
class that is not a multiple of the chunk size.

ChangeLog

@@ -7,6 +7,10 @@ brevity.  Much more detail can be found in the git revision history:
 * 4.0.3 (XXX)
 
   Bug fixes:
+  - Fix xallocx(..., MALLOCX_ZERO) to zero all trailing bytes of large
+    allocations when --enable-cache-oblivious configure option is enabled.
+  - Fix xallocx(..., MALLOCX_ZERO) to zero trailing bytes of huge allocations
+    when resizing from/to a size class that is not a multiple of the chunk size.
   - Fix prof_tctx_dump_iter() to filter out nodes that were created after heap
     profile dumping started.
 

src/arena.c

@@ -2679,6 +2679,16 @@ arena_ralloc_large_grow(arena_t *arena, arena_chunk_t *chunk, void *ptr,
 		if (arena_run_split_large(arena, run, splitsize, zero))
 			goto label_fail;
 
+		if (config_cache_oblivious && zero) {
+			/*
+			 * Zero the trailing bytes of the original allocation's
+			 * last page, since they are in an indeterminate state.
+			 */
+			assert(PAGE_CEILING(oldsize) == oldsize);
+			memset((void *)((uintptr_t)ptr + oldsize), 0,
+			    PAGE_CEILING((uintptr_t)ptr) - (uintptr_t)ptr);
+		}
+
 		size = oldsize + splitsize;
 		npages = (size + large_pad) >> LG_PAGE;
 

src/huge.c

@@ -133,7 +133,7 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
 	extent_node_t *node;
 	arena_t *arena;
 	chunk_hooks_t chunk_hooks = CHUNK_HOOKS_INITIALIZER;
-	bool zeroed;
+	bool pre_zeroed, post_zeroed;
 
 	/* Increase usize to incorporate extra. */
 	for (usize = usize_min; usize < usize_max && (usize_next = s2u(usize+1))
@@ -145,26 +145,27 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
 
 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
+	pre_zeroed = extent_node_zeroed_get(node);
 
 	/* Fill if necessary (shrinking). */
 	if (oldsize > usize) {
 		size_t sdiff = oldsize - usize;
 		if (config_fill && unlikely(opt_junk_free)) {
 			memset((void *)((uintptr_t)ptr + usize), 0x5a, sdiff);
-			zeroed = false;
+			post_zeroed = false;
 		} else {
-			zeroed = !chunk_purge_wrapper(arena, &chunk_hooks, ptr,
-			    CHUNK_CEILING(oldsize), usize, sdiff);
+			post_zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
+			    ptr, CHUNK_CEILING(oldsize), usize, sdiff);
 		}
 	} else
-		zeroed = true;
+		post_zeroed = pre_zeroed;
 
 	malloc_mutex_lock(&arena->huge_mtx);
 	/* Update the size of the huge allocation. */
 	assert(extent_node_size_get(node) != usize);
 	extent_node_size_set(node, usize);
-	/* Clear node's zeroed field if zeroing failed above. */
-	extent_node_zeroed_set(node, extent_node_zeroed_get(node) && zeroed);
+	/* Update zeroed. */
+	extent_node_zeroed_set(node, post_zeroed);
 	malloc_mutex_unlock(&arena->huge_mtx);
 
 	arena_chunk_ralloc_huge_similar(arena, ptr, oldsize, usize);
@@ -172,7 +173,7 @@ huge_ralloc_no_move_similar(void *ptr, size_t oldsize, size_t usize_min,
 	/* Fill if necessary (growing). */
 	if (oldsize < usize) {
 		if (zero || (config_fill && unlikely(opt_zero))) {
-			if (!zeroed) {
+			if (!pre_zeroed) {
 				memset((void *)((uintptr_t)ptr + oldsize), 0,
 				    usize - oldsize);
 			}
@@ -190,10 +191,11 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 	arena_t *arena;
 	chunk_hooks_t chunk_hooks;
 	size_t cdiff;
-	bool zeroed;
+	bool pre_zeroed, post_zeroed;
 
 	node = huge_node_get(ptr);
 	arena = extent_node_arena_get(node);
+	pre_zeroed = extent_node_zeroed_get(node);
 	chunk_hooks = chunk_hooks_get(arena);
 
 	assert(oldsize > usize);
@@ -209,21 +211,21 @@ huge_ralloc_no_move_shrink(void *ptr, size_t oldsize, size_t usize)
 		if (config_fill && unlikely(opt_junk_free)) {
 			huge_dalloc_junk((void *)((uintptr_t)ptr + usize),
 			    sdiff);
-			zeroed = false;
+			post_zeroed = false;
 		} else {
-			zeroed = !chunk_purge_wrapper(arena, &chunk_hooks,
+			post_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;
+		post_zeroed = pre_zeroed;
 
 	malloc_mutex_lock(&arena->huge_mtx);
 	/* Update the size of the huge allocation. */
 	extent_node_size_set(node, usize);
-	/* Clear node's zeroed field if zeroing failed above. */
-	extent_node_zeroed_set(node, extent_node_zeroed_get(node) && zeroed);
+	/* Update zeroed. */
+	extent_node_zeroed_set(node, post_zeroed);
 	malloc_mutex_unlock(&arena->huge_mtx);
 
 	/* Zap the excess chunks. */

test/integration/xallocx.c

@@ -347,6 +347,121 @@ TEST_BEGIN(test_extra_huge)
 }
 TEST_END
 
+static void
+print_filled_extents(const void *p, uint8_t c, size_t len)
+{
+	const uint8_t *pc = (const uint8_t *)p;
+	size_t i, range0;
+	uint8_t c0;
+
+	malloc_printf("  p=%p, c=%#x, len=%zu:", p, c, len);
+	range0 = 0;
+	c0 = pc[0];
+	for (i = 0; i < len; i++) {
+		if (pc[i] != c0) {
+			malloc_printf(" %#x[%zu..%zu)", c0, range0, i);
+			range0 = i;
+			c0 = pc[i];
+		}
+	}
+	malloc_printf(" %#x[%zu..%zu)\n", c0, range0, i);
+}
+
+static bool
+validate_fill(const void *p, uint8_t c, size_t offset, size_t len)
+{
+	const uint8_t *pc = (const uint8_t *)p;
+	bool err;
+	size_t i;
+
+	for (i = offset, err = false; i < offset+len; i++) {
+		if (pc[i] != c)
+			err = true;
+	}
+
+	if (err)
+		print_filled_extents(p, c, offset + len);
+
+	return (err);
+}
+
+static void
+test_zero(size_t szmin, size_t szmax)
+{
+	size_t sz, nsz;
+	void *p;
+#define	FILL_BYTE 0x7aU
+
+	sz = szmax;
+	p = mallocx(sz, MALLOCX_ZERO);
+	assert_ptr_not_null(p, "Unexpected mallocx() error");
+	assert_false(validate_fill(p, 0x00, 0, sz), "Memory not filled: sz=%zu",
+	    sz);
+
+	/*
+	 * Fill with non-zero so that non-debug builds are more likely to detect
+	 * errors.
+	 */
+	memset(p, FILL_BYTE, sz);
+	assert_false(validate_fill(p, FILL_BYTE, 0, sz),
+	    "Memory not filled: sz=%zu", sz);
+
+	/* Shrink in place so that we can expect growing in place to succeed. */
+	sz = szmin;
+	assert_zu_eq(xallocx(p, sz, 0, MALLOCX_ZERO), sz,
+	    "Unexpected xallocx() error");
+	assert_false(validate_fill(p, FILL_BYTE, 0, sz),
+	    "Memory not filled: sz=%zu", sz);
+
+	for (sz = szmin; sz < szmax; sz = nsz) {
+		nsz = nallocx(sz+1, MALLOCX_ZERO);
+		assert_zu_eq(xallocx(p, sz+1, 0, MALLOCX_ZERO), nsz,
+		    "Unexpected xallocx() failure");
+		assert_false(validate_fill(p, FILL_BYTE, 0, sz),
+		    "Memory not filled: sz=%zu", sz);
+		assert_false(validate_fill(p, 0x00, sz, nsz-sz),
+		    "Memory not filled: sz=%zu, nsz-sz=%zu", sz, nsz-sz);
+		memset((void *)((uintptr_t)p + sz), FILL_BYTE, nsz-sz);
+		assert_false(validate_fill(p, FILL_BYTE, 0, nsz),
+		    "Memory not filled: nsz=%zu", nsz);
+	}
+
+	dallocx(p, 0);
+}
+
+TEST_BEGIN(test_zero_large)
+{
+	size_t large0, largemax;
+
+	/* Get size classes. */
+	large0 = get_large_size(0);
+	largemax = get_large_size(get_nlarge()-1);
+
+	test_zero(large0, largemax);
+}
+TEST_END
+
+TEST_BEGIN(test_zero_huge)
+{
+	size_t huge0, huge1;
+	static const bool maps_coalesce =
+#ifdef JEMALLOC_MAPS_COALESCE
+	    true
+#else
+	    false
+#endif
+	    ;
+
+	/* Get size classes. */
+	huge0 = get_huge_size(0);
+	huge1 = get_huge_size(1);
+
+	if (maps_coalesce)
+		test_zero(huge0, huge0 * 4);
+	test_zero(huge1, huge0 * 2);
+}
+TEST_END
+
 int
 main(void)
 {
@@ -359,5 +474,7 @@ main(void)
 	    test_size_extra_overflow,
 	    test_extra_small,
 	    test_extra_large,
-	    test_extra_huge));
+	    test_extra_huge,
+	    test_zero_large,
+	    test_zero_huge));
 }