Loading...
--- libmalloc/libmalloc-166.251.2/src/nanov2_malloc.c
+++ libmalloc/libmalloc-657.80.3/src/nanov2_malloc.c
@@ -29,25 +29,51 @@
 #pragma mark Forward Declarations
 
 #if OS_VARIANT_NOTRESOLVED
-static void nanov2_statistics(nanozonev2_t *nanozone, malloc_statistics_t *stats);
+static kern_return_t
+nanov2_statistics_task_printer(task_t task, vm_address_t zone_address,
+							   memory_reader_t reader, print_task_printer_t printer,
+							   malloc_statistics_t *stats);
+static kern_return_t
+nanov2_statistics_task(task_t task, vm_address_t zone_address,
+					   memory_reader_t reader, malloc_statistics_t *stats);
 #endif // OS_VARIANT_NOTRESOLVED
 
 #pragma mark -
 #pragma mark Externals for resolved functions
 
-extern void *nanov2_allocate(nanozonev2_t *nanozone, size_t rounded_size,
-		boolean_t clear);
-extern void nanov2_free_to_block(nanozonev2_t *nanozone, void *ptr,
-		nanov2_size_class_t size_class);
-extern boolean_t nanov2_madvise_block(nanozonev2_t *nanozone,
-		nanov2_block_meta_t *block_metap, nanov2_block_t *blockp,
-		nanov2_size_class_t size_class);
-extern size_t nanov2_pointer_size(nanozonev2_t *nanozone, void *ptr,
+MALLOC_NOEXPORT extern size_t nanov2_pointer_size(nanozonev2_t *nanozone, void *ptr,
 		boolean_t allow_inner);
-extern size_t nanov2_pressure_relief(nanozonev2_t *nanozone, size_t goal);
+MALLOC_NOEXPORT extern size_t nanov2_pressure_relief(nanozonev2_t *nanozone, size_t goal);
 
 #if OS_VARIANT_RESOLVED
-extern boolean_t nanov2_allocate_new_region(nanozonev2_t *nanozone);
+MALLOC_ALWAYS_INLINE MALLOC_INLINE size_t
+nanov2_pointer_size_inline(nanozonev2_t *nanozone, void *ptr, boolean_t allow_inner,
+		nanov2_size_class_t *size_class_out, nanov2_block_meta_t **block_metap_out);
+
+MALLOC_ALWAYS_INLINE MALLOC_INLINE void *
+nanov2_allocate_from_block_inline(nanozonev2_t *nanozone,
+		nanov2_block_meta_t *block_metap, nanov2_size_class_t size_class,
+		nanov2_block_meta_t **madvise_block_metapp_out, bool *corruption);
+
+static void *
+nanov2_allocate_outlined(nanozonev2_t *nanozone,
+		nanov2_block_meta_t **block_metapp, size_t rounded_size,
+		nanov2_size_class_t size_class, int allocation_index,
+		nanov2_block_meta_t *madvise_block_metap, void *corrupt_slot,
+		bool clear, bool typed, malloc_type_id_t type_id);
+
+MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_block_meta_t *
+nanov2_free_to_block_inline(nanozonev2_t *nanozone, void *ptr,
+		nanov2_size_class_t size_class, nanov2_block_meta_t *block_metap);
+
+static boolean_t nanov2_madvise_block_locked(
+		nanozonev2_t *nanozone, nanov2_block_meta_t *block_metap,
+		nanov2_block_t *blockp, nanov2_size_class_t size_class, uint32_t expected_state);
+static void nanov2_madvise_block(nanozonev2_t *nanozone,
+		nanov2_block_meta_t *block_metap, nanov2_size_class_t size_class,
+		uint32_t expected_state);
+
+MALLOC_NOEXPORT extern nanov2_arena_t *nanov2_allocate_new_region(nanozonev2_t *nanozone);
 #endif // OS_VARIANT_RESOLVED
 
 #pragma mark -
@@ -104,9 +130,9 @@
 #if OS_VARIANT_NOTRESOLVED
 
 // Madvise policy. Set from the MallocNanoMadvisePolicy environment variable.
-nanov2_madvise_policy_t nanov2_madvise_policy;
-
-nanov2_policy_config_t nanov2_policy_config = {
+MALLOC_NOEXPORT nanov2_madvise_policy_t nanov2_madvise_policy;
+
+MALLOC_NOEXPORT nanov2_policy_config_t nanov2_policy_config = {
 	.block_scan_policy = NANO_SCAN_CAPACITY_BASED,
 	.block_scan_min_capacity = DEFAULT_SCAN_MIN_CAPACITY,
 	.block_scan_max_capacity = DEFAULT_SCAN_MAX_CAPACITY,
@@ -117,8 +143,8 @@
 
 #else  // OS_VARIANT_NOTRESOLVED
 
-extern nanov2_policy_config_t nanov2_policy_config;
-extern nanov2_madvise_policy_t nanov2_madvise_policy;
+MALLOC_NOEXPORT extern nanov2_policy_config_t nanov2_policy_config;
+MALLOC_NOEXPORT extern nanov2_madvise_policy_t nanov2_madvise_policy;
 
 #endif // OS_VARIANT_NOTRESOLVED
 
@@ -133,7 +159,7 @@
 // up to 64. One unit corresponds to BLOCKS_PER_UNIT blocks in the corresponding
 // size class, so 64 units maps to a total of 64 * 64 = 4096 blocks and each
 // block is 16K, making a total of 64MB, which is the size of an arena.
-int block_units_by_size_class[] = {
+static int block_units_by_size_class[] = {
 	2,	// 16-byte allocations (less 1 for the metadata block)
 	10,	// 32-byte allocations
 	11,	// 48-byte allocations
@@ -163,18 +189,18 @@
 // Offsets to the first and last blocks for each size class within an arena, in
 // the logical address space. These tables are constructed from the values in
 // the block_units_by_size_class table.
-int first_block_offset_by_size_class[NANO_SIZE_CLASSES];
-int last_block_offset_by_size_class[NANO_SIZE_CLASSES];
+MALLOC_NOEXPORT int first_block_offset_by_size_class[NANO_SIZE_CLASSES];
+MALLOC_NOEXPORT int last_block_offset_by_size_class[NANO_SIZE_CLASSES];
 
 // Table mapping the part of a logical address that depends on size class to
 // the size class. Also built from the block_units_by_size_class table.
-int ptr_offset_to_size_class[TOTAL_BLOCK_UNITS];
+MALLOC_NOEXPORT int ptr_offset_to_size_class[TOTAL_BLOCK_UNITS];
 
 // Number of slots in a block, indexed by size class. Note that there is a small
 // amount of wastage in some size classes because the block size is not always
 // exactly divisible by the allocation size. The number of wasted bytes is shown
 // in parentheses in the comments below.
-const int slots_by_size_class[] = {
+MALLOC_NOEXPORT const int slots_by_size_class[] = {
 	NANOV2_BLOCK_SIZE/(1 * NANO_REGIME_QUANTA_SIZE),  	// 16 bytes: 1024	(0)
 	NANOV2_BLOCK_SIZE/(2 * NANO_REGIME_QUANTA_SIZE),	// 32 bytes: 512	(0)
 	NANOV2_BLOCK_SIZE/(3 * NANO_REGIME_QUANTA_SIZE),	// 48 bytes: 341	(16)
@@ -233,13 +259,14 @@
 
 // Given a block metadata pointer, returns whether the block is active (that is,
 // it is being used for allocations, it has allocations that have not been freed,
-// or is waiting to be madvised).
+// or is waiting to be madvised  and is not a guard block).
 static MALLOC_ALWAYS_INLINE MALLOC_INLINE boolean_t
 nanov2_is_block_active(nanov2_block_meta_t block_meta)
 {
 	return block_meta.next_slot != SLOT_NULL
 			&& block_meta.next_slot != SLOT_MADVISING
-			&& block_meta.next_slot != SLOT_MADVISED;
+			&& block_meta.next_slot != SLOT_MADVISED
+			&& block_meta.next_slot != SLOT_GUARD;
 }
 
 #if OS_VARIANT_RESOLVED
@@ -306,7 +333,6 @@
 	return (void *)(((uintptr_t)ptr) & NANOV2_ARENA_ADDRESS_MASK);
 }
 
-#if OS_VARIANT_RESOLVED
 // Given a pointer that is assumed to be in the Nano zone, returns the address
 // of its containing region. Works for both real and logical pointers.
 static MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_region_t *
@@ -314,7 +340,6 @@
 {
 	return (nanov2_region_t *)(((uintptr_t)ptr) & NANOV2_REGION_ADDRESS_MASK);
 }
-#endif // OS_VARIANT_RESOLVED
 
 // Given a pointer that is assumed to be in the Nano zone, returns the real
 // address of its metadata block. Works for both real and logical pointers.
@@ -388,15 +413,40 @@
 	return (nanov2_arena_t *)region;
 }
 
+#if OS_VARIANT_RESOLVED
+// Given an atomically-observed current_region_next_arena pointer, returns
+// whether or not it's a usable arena or a limit arena (indicating exhaustion of
+// the current region).
+static MALLOC_ALWAYS_INLINE MALLOC_INLINE bool
+nanov2_current_region_next_arena_is_limit(
+		nanov2_arena_t *current_region_next_arena)
+{
+	// The first arena of a region is never stored in current_region_next_arena,
+	// so a value at the beginning of a region must be a limit arena.
+	return current_region_next_arena == (nanov2_arena_t *)(
+			nanov2_region_address_for_ptr(current_region_next_arena));
+}
+#endif // OS_VARIANT_RESOLVED
+
+// Given an atomically-observed current_region_next_arena pointer, returns the
+// base of the current region at the time of the observation.
+static MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_region_t *
+nanov2_current_region_base(nanov2_arena_t *current_region_next_arena)
+{
+	return nanov2_region_address_for_ptr(
+			(void *)(((uintptr_t)current_region_next_arena) - 1));
+}
+
 // Given a region pointer, returns a pointer to the arena after the last
 // active arena in the region.
 static MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_arena_t *
-nanov2_limit_arena_for_region(nanozonev2_t *nanozone, nanov2_region_t *region)
+nanov2_limit_arena_for_region(nanozonev2_t __unused *nanozone,
+		nanov2_region_t *region, nanov2_arena_t *current_region_next_arena)
 {
 	// The first arena is colocated with the region itself.
 	nanov2_arena_t *limit_arena;
-	if (region == nanozone->current_region_base) {
-		limit_arena = nanozone->current_region_next_arena;
+	if (region == nanov2_current_region_base(current_region_next_arena)) {
+		limit_arena = current_region_next_arena;
 	} else {
 		limit_arena = nanov2_first_arena_for_region(region + 1);
 	}
@@ -415,16 +465,51 @@
 			nanov2_metablock_meta_index(nanozone)];
 }
 
+#if OS_VARIANT_RESOLVED
 // Given a pointer to a region, returns a pointer to the region that follows it,
-// or NULL if there isn't one.
+// or NULL if there isn't one. We may observe linkage to a new region that
+// hasn't yet actually been installed into current_region_next_arena; ignore the
+// linkage in this case.
 static MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_region_t *
-nanov2_next_region_for_region(nanozonev2_t *nanozone, nanov2_region_t *region)
+nanov2_next_region_for_region(nanozonev2_t *nanozone, nanov2_region_t *region,
+		nanov2_arena_t *current_region_next_arena)
 {
 	nanov2_region_linkage_t *linkage =
 			nanov2_region_linkage_for_region(nanozone, region);
-	int offset = linkage->next_region_offset;
-	return offset ? region + offset : NULL;
-}
+	int offset = os_atomic_load(&linkage->next_region_offset, relaxed);
+	if (!offset) {
+		return NULL;
+	}
+
+	nanov2_region_t *next_region = region + offset;
+	return (nanov2_arena_t *)next_region < current_region_next_arena ?
+			next_region : NULL;
+}
+#endif // OS_VARIANT_RESOLVED
+
+#if OS_VARIANT_NOTRESOLVED
+// Given a pointer to a region, returns a pointer to the region that follows it,
+// or NULL if there isn't one. This variant is used when mapping the nanozone
+// for another process.
+static MALLOC_ALWAYS_INLINE MALLOC_INLINE nanov2_region_t *
+nanov2_next_region_for_region_offset(nanozonev2_t *nanozone,
+		nanov2_region_t *region, off_t region_offset,
+		nanov2_arena_t *current_region_next_arena)
+{
+	nanov2_region_linkage_t *linkage =
+			nanov2_region_linkage_for_region(nanozone, region);
+	nanov2_region_linkage_t *mapped_linkage = (nanov2_region_linkage_t *)(
+			((uintptr_t)linkage + region_offset));
+	int offset = os_atomic_load(&mapped_linkage->next_region_offset, relaxed);
+	if (!offset) {
+		return NULL;
+	}
+
+	nanov2_region_t *next_region = region + offset;
+	return (nanov2_arena_t *)next_region < current_region_next_arena ?
+			next_region : NULL;
+}
+#endif // OS_VARIANT_NOTRESOLVED
 
 // Given the index of a slot in a block of a given size and the base address of
 // the block, returns a pointer to the start of the slot. This works for both
@@ -570,15 +655,15 @@
 #if CONFIG_NANO_USES_HYPER_SHIFT
 	if (os_likely(nano_common_max_magazines_is_ncpu)) {
 		// Default case is max magazines == physical number of CPUs, which
-		// must be > _os_cpu_number() >> hyper_shift, so the modulo
+		// must be > _malloc_cpu_number() >> hyper_shift, so the modulo
 		// operation is not required.
-		return _os_cpu_number() >> hyper_shift;
+		return (_malloc_cpu_number() >> hyper_shift) & MAX_CURRENT_BLOCKS_MASK;
 	}
 #else // CONFIG_NANO_USES_HYPER_SHIFT
 	if (os_likely(nano_common_max_magazines_is_ncpu)) {
 		// Default case is max magazines == logical number of CPUs, which
-		// must be > _os_cpu_number() so the modulo operation is not required.
-		return _os_cpu_number();
+		// must be > _malloc_cpu_number() so the modulo operation is not required.
+		return _malloc_cpu_number() & MAX_CURRENT_BLOCKS_MASK;
 	}
 #endif // CONFIG_NANO_USES_HYPER_SHIFT
 
@@ -588,11 +673,64 @@
 #endif // CONFIG_NANO_USES_HYPER_SHIFT
 
 	if (os_likely(_os_cpu_number_override == -1)) {
-		return (_os_cpu_number() >> shift) % nano_common_max_magazines;
-	}
-	return (_os_cpu_number_override >> shift) % nano_common_max_magazines;
+		return ((_malloc_cpu_number() >> shift) % nano_common_max_magazines) &
+				MAX_CURRENT_BLOCKS_MASK;
+	}
+	return ((_os_cpu_number_override >> shift) % nano_common_max_magazines) &
+			MAX_CURRENT_BLOCKS_MASK;
 }
 #endif // OS_VARIANT_RESOLVED
+
+#pragma mark -
+#pragma mark Guard Blocks
+
+// Converts a given block (specified by absolute block number) in an arena into
+// a guard block. The block will be marked as in-use so that it is not available
+// for allocations and its permissions are set to PROT_READ. Note that
+// PROT_READ is used instead of PROT_NONE because the latter breaks the
+// enumerator, which tries to map the whole region and fails if there are
+// PROT_NONE pages in the range. We can't fix that in the allocator because the
+// code that does the mapping is part of the sampling tools and is simply
+// invoked as a callback from the enumerator.
+static MALLOC_ALWAYS_INLINE MALLOC_INLINE void
+nanov2_create_guard_block(nanozonev2_t *nanozone, nanov2_arena_t *arena,
+		nanov2_block_index_t block_index) {
+	// Mark the block as in-use in the meta data
+	static nanov2_block_meta_t in_use_block = {
+		.in_use = 1,
+		.next_slot = SLOT_GUARD
+	};
+	nanov2_meta_index_t	block_meta_index =
+			nanov2_block_index_to_meta_index(block_index);
+	nanov2_arena_metablock_t *block_metap = nanov2_metablock_address_for_ptr(
+			nanozone, arena);
+	block_metap->arena_block_meta[block_meta_index] = in_use_block;
+	void *block_ptr = &arena->blocks[block_index];
+
+	// Apply PROT_NONE to the block itself.
+	kern_return_t err = mprotect(block_ptr, NANOV2_BLOCK_SIZE, PROT_READ);
+	if (err != KERN_SUCCESS) {
+		malloc_report(ASL_LEVEL_ERR, "Failed to create guard block at %p (%d)\n",
+				block_ptr, err);
+	}
+}
+
+// Creates the guard blocks for an arena, if required. The guard blocks are
+// the first and last physical blocks in the arena that are not the metadata
+// block.
+static MALLOC_ALWAYS_INLINE MALLOC_INLINE void
+nanov2_init_guard_blocks(nanozonev2_t *nanozone, nanov2_arena_t *arena)
+{
+	if (nanozone->debug_flags & MALLOC_ALL_GUARD_PAGE_FLAGS) {
+		// Use the first and last blocks in the arena as guard regions,
+		// avoiding the metadata block.
+		nanov2_meta_index_t meta_index = nanov2_metablock_meta_index(nanozone);
+		nanov2_create_guard_block(nanozone, arena, meta_index == 0 ? 1 : 0);
+		nanov2_create_guard_block(nanozone, arena,
+				meta_index == NANOV2_BLOCKS_PER_ARENA - 1 ?
+					NANOV2_BLOCKS_PER_ARENA - 2 : NANOV2_BLOCKS_PER_ARENA - 1);
+	}
+}
 
 #pragma mark -
 #pragma mark Allocator Initialization
@@ -921,69 +1059,165 @@
 // determine whether the pointer is for a Nano V2 allocation and, if not,
 // delegates to the helper zone. Returns 0 if the pointer is not to memory
 // allocated by Nano V2 or attributable to the helper zone.
-size_t
+MALLOC_NOEXPORT size_t
 nanov2_size(nanozonev2_t *nanozone, const void *ptr)
 {
-	size_t size = nanov2_pointer_size(nanozone, (void *)ptr, FALSE);
+	size_t size = nanov2_pointer_size_inline(nanozone, (void *)ptr, FALSE,
+			NULL, NULL);
 	return size ? size : nanozone->helper_zone->size(nanozone->helper_zone, ptr);
 }
 
-void *
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
+void
+nanov2_bzero(void *ptr, size_t size)
+{
+	// TODO: inline bzero from libplatform
+	bzero(ptr, size);
+}
+
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
+static void *
+_nanov2_allocate(nanozonev2_t *nanozone, size_t rounded_size, bool clear,
+		bool typed, malloc_type_id_t type_id)
+{
+	nanov2_block_meta_t *madvise_block_metap = NULL;
+	nanov2_size_class_t size_class = nanov2_size_class_from_size(rounded_size);
+
+	// Get the index of the pointer to the block from which we are should be
+	// allocating. This currently depends on the physical CPU number.
+	int allocation_index = nanov2_get_allocation_block_index();
+
+	// Get the current allocation block meta data pointer. If this is NULL, we
+	// need to find a new allocation block.
+	nanov2_block_meta_t **block_metapp =
+			&nanozone->current_block[size_class][allocation_index];
+	nanov2_block_meta_t *block_metap = os_atomic_load(block_metapp, relaxed);
+	bool corruption = false;
+	void *ptr = NULL;
+	if (block_metap) {
+		// Fast path: we have a block -- try to allocate from it.
+		ptr = nanov2_allocate_from_block_inline(nanozone, block_metap,
+				size_class, &madvise_block_metap, &corruption);
+		if (ptr && !corruption) {
+			if (clear && malloc_zero_policy != MALLOC_ZERO_ON_FREE) {
+				nanov2_bzero(ptr, rounded_size);
+			} else {
+				// Always clear the double-free guard so that we can recognize that
+				// this block is not on the free list.
+				nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
+				os_atomic_store(&slotp->double_free_guard, 0, relaxed);
+
+				// We know the body of the allocation is already clear, so we just
+				// need to clean up the next_slot word to get to all-zero.  Do so in
+				// all cases, even if a cleared allocation is not requested, to
+				// prevent any leakage through the next_slot bits.
+				os_atomic_store(&slotp->next_slot, 0, relaxed);
+			}
+			return ptr;
+		}
+	}
+
+	return nanov2_allocate_outlined(nanozone, block_metapp, rounded_size,
+			size_class, allocation_index, madvise_block_metap, ptr, clear,
+			typed, type_id);
+}
+
+MALLOC_NOEXPORT void *
 nanov2_malloc(nanozonev2_t *nanozone, size_t size)
 {
 	size_t rounded_size = _nano_common_good_size(size);
 	if (rounded_size <= NANO_MAX_SIZE) {
-		void *ptr = nanov2_allocate(nanozone, rounded_size, FALSE);
-		if (ptr) {
-			if (os_unlikely(size && (nanozone->debug_flags & MALLOC_DO_SCRIBBLE))) {
-				memset(ptr, SCRIBBLE_BYTE, size);
+		return _nanov2_allocate(nanozone, rounded_size, false, false,
+				MALLOC_TYPE_ID_NONE);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
+	return nanozone->helper_zone->malloc(nanozone->helper_zone, size);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_malloc_type(nanozonev2_t *nanozone, size_t size,
+		malloc_type_id_t type_id)
+{
+	size_t rounded_size = _nano_common_good_size(size);
+	if (rounded_size <= NANO_MAX_SIZE) {
+		return _nanov2_allocate(nanozone, rounded_size, false, true, type_id);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_malloc(helper_zone, size, type_id);
+}
+
+MALLOC_NOEXPORT void
+nanov2_free_definite_size(nanozonev2_t *nanozone, void *ptr, size_t size)
+{
+	if (ptr && nanov2_has_valid_signature(ptr)) {
+		nanov2_size_class_t size_class = nanov2_size_class_from_size(size);
+
+		if (malloc_zero_policy == MALLOC_ZERO_ON_FREE) {
+			if (size_class != 0) {
+				nanov2_bzero((char *)ptr + sizeof(nanov2_free_slot_t),
+						size - sizeof(nanov2_free_slot_t));
 			}
-			return ptr;
-		}
-	}
-
-	// If we reach this point, we couldn't allocate, so delegate to the
-	// helper zone.
-	return nanozone->helper_zone->malloc(nanozone->helper_zone, size);
-}
-
-void
-nanov2_free_definite_size(nanozonev2_t *nanozone, void *ptr, size_t size)
-{
-	// Check whether it's a Nano pointer and get the size. We should only get
-	// here if it is and furthermore we already know that "size" is the actual
-	// rounded size, so don't waste time rechecking that. This is just a
-	// sanity check.
-	if (ptr && nanov2_has_valid_signature(ptr)) {
-		if (os_unlikely(nanozone->debug_flags & MALLOC_DO_SCRIBBLE)) {
-			memset(ptr, SCRABBLE_BYTE, size);
-		}
-		nanov2_free_to_block(nanozone, ptr, nanov2_size_class_from_size(size));
+		}
+
+		nanov2_block_meta_t *madvise_block_metap = nanov2_free_to_block_inline(
+				nanozone, ptr, size_class, NULL);
+		if (madvise_block_metap) {
+			nanov2_madvise_block(nanozone, madvise_block_metap, size_class,
+					SLOT_CAN_MADVISE);
+		}
 		return;
 	}
 	return nanozone->helper_zone->free_definite_size(nanozone->helper_zone, ptr,
 			size);
 }
 
-void
-nanov2_free(nanozonev2_t *nanozone, void *ptr)
-{
-	if (ptr && nanov2_has_valid_signature(ptr)) {
+static void
+_nanov2_free(nanozonev2_t *nanozone, void *ptr, bool try)
+{
+	if (ptr) {
 		// Check whether it's a Nano pointer and get the size. If it's not
 		// Nano, pass it to the helper zone.
-		size_t size = nanov2_pointer_size(nanozone, ptr, FALSE);
+		nanov2_size_class_t size_class;
+		nanov2_block_meta_t *block_metap;
+		size_t size = nanov2_pointer_size_inline(nanozone, ptr, FALSE,
+				&size_class, &block_metap);
 		if (size) {
-			if (os_unlikely(nanozone->debug_flags & MALLOC_DO_SCRIBBLE)) {
-				memset(ptr, SCRABBLE_BYTE, size);
+			if (malloc_zero_policy == MALLOC_ZERO_ON_FREE) {
+				if (size > sizeof(nanov2_free_slot_t)) {
+					nanov2_bzero((char *)ptr + sizeof(nanov2_free_slot_t),
+							size - sizeof(nanov2_free_slot_t));
+				}
 			}
-			nanov2_free_to_block(nanozone, ptr, nanov2_size_class_from_size(size));
+
+			nanov2_block_meta_t *madvise_block_metap = nanov2_free_to_block_inline(
+					nanozone, ptr, size_class, block_metap);
+			if (madvise_block_metap) {
+				nanov2_madvise_block(nanozone, madvise_block_metap, size_class,
+						SLOT_CAN_MADVISE);
+			}
 			return;
 		}
 	}
-	return nanozone->helper_zone->free(nanozone->helper_zone, ptr);
-}
-
-void *
+	return try ? nanozone->helper_zone->try_free_default(nanozone->helper_zone, ptr) :
+			nanozone->helper_zone->free(nanozone->helper_zone, ptr);
+}
+
+MALLOC_NOEXPORT void
+nanov2_free(nanozonev2_t *nanozone, void *ptr)
+{
+	_nanov2_free(nanozone, ptr, false);
+}
+
+MALLOC_NOEXPORT void
+nanov2_try_free_default(nanozonev2_t *nanozone, void *ptr)
+{
+	_nanov2_free(nanozone, ptr, true);
+}
+
+MALLOC_NOEXPORT void *
 nanov2_calloc(nanozonev2_t *nanozone, size_t num_items, size_t size)
 {
 	size_t total_bytes;
@@ -992,15 +1226,58 @@
 	}
 	size_t rounded_size = _nano_common_good_size(total_bytes);
 	if (total_bytes <= NANO_MAX_SIZE) {
-		void *ptr = nanov2_allocate(nanozone, rounded_size, TRUE);
-		if (ptr) {
-			return ptr;
-		}
-	}
-
-	// If we reach this point, we couldn't allocate, so delegate to the
-	// helper zone.
+		return _nanov2_allocate(nanozone, rounded_size, true, false,
+				MALLOC_TYPE_ID_NONE);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
 	return nanozone->helper_zone->calloc(nanozone->helper_zone, 1, total_bytes);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_calloc_type(nanozonev2_t *nanozone, size_t num_items, size_t size,
+		malloc_type_id_t type_id)
+{
+	size_t total_bytes;
+	if (calloc_get_size(num_items, size, 0, &total_bytes)) {
+		return NULL;
+	}
+	size_t rounded_size = _nano_common_good_size(total_bytes);
+	if (total_bytes <= NANO_MAX_SIZE) {
+		return _nanov2_allocate(nanozone, rounded_size, true, true, type_id);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_calloc(helper_zone, 1, total_bytes,
+			type_id);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_malloc_zero_on_alloc(nanozonev2_t *nanozone, size_t size)
+{
+	size_t rounded_size = _nano_common_good_size(size);
+	if (rounded_size <= NANO_MAX_SIZE) {
+		return _nanov2_allocate(nanozone, rounded_size, true, false,
+				MALLOC_TYPE_ID_NONE);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
+	return nanozone->helper_zone->malloc(nanozone->helper_zone, size);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_malloc_type_zero_on_alloc(nanozonev2_t *nanozone, size_t size,
+		malloc_type_id_t type_id)
+{
+	size_t rounded_size = _nano_common_good_size(size);
+	if (rounded_size <= NANO_MAX_SIZE) {
+		return _nanov2_allocate(nanozone, rounded_size, true, true, type_id);
+	}
+
+	// Too big for nano, so delegate to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_malloc(helper_zone, size, type_id);
 }
 #endif // OS_VARIANT_RESOLVED
 
@@ -1014,25 +1291,44 @@
 #endif // OS_VARIANT_NOTRESOLVED
 
 #if OS_VARIANT_RESOLVED
-void *
-nanov2_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size)
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
+static void *
+_nanov2_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size, bool typed,
+		malloc_type_id_t type_id)
 {
 	// If we are given a NULL pointer, just allocate memory of the requested
 	// size.
 	if (ptr == NULL) {
-		return nanov2_malloc(nanozone, new_size);
-	}
+		if (typed) {
+			return nanov2_malloc_type(nanozone, new_size, type_id);
+		} else {
+			return nanov2_malloc(nanozone, new_size);
+		}
+	}
+
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
 
 	size_t old_size = nanov2_pointer_size(nanozone, ptr, FALSE);
 	if (!old_size) {
 		// Not a Nano pointer - let the helper deal with it
-		return nanozone->helper_zone->realloc(nanozone->helper_zone, ptr, new_size);
+		if (typed) {
+			return helper_zone->malloc_type_realloc(helper_zone, ptr, new_size,
+					type_id);
+		} else {
+			return helper_zone->realloc(helper_zone, ptr, new_size);
+		}
 	}
 
 	void *new_ptr;
 	if (new_size > NANO_MAX_SIZE) {
 		// Too large for Nano. Try to allocate from the helper zone.
-		new_ptr = nanozone->helper_zone->malloc(nanozone->helper_zone, new_size);
+		if (typed) {
+			new_ptr = helper_zone->malloc_type_malloc(helper_zone, new_size,
+					type_id);
+		} else {
+			new_ptr = helper_zone->malloc(helper_zone, new_size);
+		}
+
 		if (!new_ptr) {
 			// Failed to allocate - leave the existing allocation alone.
 			return NULL;
@@ -1044,17 +1340,24 @@
 		return nanov2_malloc(nanozone, 0);
 	} else {
 		size_t new_good_size = _nano_common_good_size(new_size);
-		if (new_good_size > old_size || new_good_size <= old_size/2) {
+		if (new_good_size > old_size || new_good_size <= old_size / 2) {
 			// Growing or shrinking to less than half size - we need to
 			// reallocate.
-			new_ptr = nanov2_malloc(nanozone, new_good_size);
+			if (typed) {
+				new_ptr = nanov2_malloc_type(nanozone, new_good_size, type_id);
+			} else {
+				new_ptr = nanov2_malloc(nanozone, new_good_size);
+			}
+
 			if (!new_ptr) {
 				// Failed to allocate - leave the existing allocation alone.
 				return NULL;
 			}
 		} else {
 			// Same size or shrinking by less than half size. Keep the same
-			// allocation and clear the area that's being released.
+			// allocation and scribble the area that's being released. Nothing
+			// to do for zero-on-free yet; that will be taken care of when the
+			// shrunk allocation is freed.
 			if (new_size != old_size) {
 				MALLOC_ASSERT(new_size < old_size);
 				if (os_unlikely(nanozone->debug_flags & MALLOC_DO_SCRIBBLE)) {
@@ -1073,6 +1376,19 @@
 
 	return new_ptr;
 }
+
+MALLOC_NOEXPORT void *
+nanov2_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size)
+{
+	return _nanov2_realloc(nanozone, ptr, new_size, false, MALLOC_TYPE_ID_NONE);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_realloc_type(nanozonev2_t *nanozone, void *ptr, size_t new_size,
+		malloc_type_id_t type_id)
+{
+	return _nanov2_realloc(nanozone, ptr, new_size, true, type_id);
+}
 #endif // OS_VARIANT_RESOLVED
 
 #if OS_VARIANT_NOTRESOLVED
@@ -1086,13 +1402,14 @@
 #endif // OS_VARIANT_NOTRESOLVED
 
 #if OS_VARIANT_RESOLVED
-boolean_t
+MALLOC_NOEXPORT boolean_t
 nanov2_claimed_address(nanozonev2_t *nanozone, void *ptr)
 {
-	return nanov2_pointer_size(nanozone, ptr, TRUE) != 0;
-}
-
-unsigned
+	return nanov2_pointer_size(nanozone, ptr, TRUE)
+			|| malloc_zone_claimed_address(nanozone->helper_zone, ptr);
+}
+
+MALLOC_NOEXPORT unsigned
 nanov2_batch_malloc(nanozonev2_t *nanozone, size_t size, void **results,
 		unsigned count)
 {
@@ -1100,7 +1417,8 @@
 	size_t rounded_size = _nano_common_good_size(size);
 	if (rounded_size <= NANO_MAX_SIZE) {
 		while (allocated < count) {
-			void *ptr = nanov2_allocate(nanozone, rounded_size, FALSE);
+			// TODO: nanov2_malloc will redo _nano_common_good_size
+			void *ptr = nanov2_malloc(nanozone, rounded_size);
 			if (!ptr) {
 				break;
 			}
@@ -1119,7 +1437,7 @@
 			nanozone->helper_zone, size, results, count - allocated);
 }
 
-void
+MALLOC_NOEXPORT void
 nanov2_batch_free(nanozonev2_t *nanozone, void **to_be_freed, unsigned count)
 {
 	if (count) {
@@ -1131,19 +1449,36 @@
 		}
 	}
 }
-#endif // OS_VARIANT_RESOLVED
-
-#if OS_VARIANT_NOTRESOLVED
-static void *
+
+MALLOC_NOEXPORT void *
 nanov2_memalign(nanozonev2_t *nanozone, size_t alignment, size_t size)
 {
-	// Always delegate this to the helper zone.
+	// Serve directly if the requested alignment is trivially satisfied by our
+	// baseline alignment (16 bytes)
+	if (alignment <= NANO_REGIME_QUANTA_SIZE) {
+		return nanov2_malloc(nanozone, size);
+	}
+
+	// Otherwise delegate to the helper zone
 	return nanozone->helper_zone->memalign(nanozone->helper_zone, alignment,
 			size);
 }
-#endif // OS_VARIANT_NOTRESOLVED
-
-#if OS_VARIANT_RESOLVED
+
+MALLOC_NOEXPORT void *
+nanov2_memalign_type(nanozonev2_t *nanozone, size_t alignment, size_t size,
+		malloc_type_id_t type_id)
+{
+	// Serve directly if the requested alignment is trivially satisfied by our
+	// baseline alignment (16 bytes)
+	if (alignment <= NANO_REGIME_QUANTA_SIZE) {
+		return nanov2_malloc_type(nanozone, size, type_id);
+	}
+
+	// Otherwise delegate to the helper zone
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_memalign(helper_zone, alignment, size,
+			type_id);
+}
 
 size_t
 nanov2_pressure_relief(nanozonev2_t *nanozone, size_t goal)
@@ -1165,9 +1500,12 @@
 	// until we reach our goal.
 	nanov2_region_t *region = nanozone->first_region_base;
 	nanov2_meta_index_t metablock_meta_index = nanov2_metablock_meta_index(nanozone);
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&nanozone->current_region_next_arena, acquire);
 	while (region) {
 		nanov2_arena_t *arena = nanov2_first_arena_for_region(region);
-		nanov2_arena_t *arena_after_region = nanov2_limit_arena_for_region(nanozone, region);
+		nanov2_arena_t *arena_after_region = nanov2_limit_arena_for_region(
+				nanozone, region, current_region_next_arena);
 		while (arena < arena_after_region) {
 			// Scan all of the blocks in the arena, skipping the metadata block.
 			nanov2_arena_metablock_t *meta_blockp =
@@ -1186,8 +1524,9 @@
 					if (meta.next_slot == SLOT_CAN_MADVISE) {
 						nanov2_block_t *blockp = nanov2_block_address_from_meta_index(
 								nanozone, arena, i);
-						if (nanov2_madvise_block(nanozone, block_metap,
-								blockp, nanov2_size_class_for_ptr(nanozone, blockp))) {
+						if (nanov2_madvise_block_locked(nanozone, block_metap,
+								blockp, nanov2_size_class_for_ptr(nanozone, blockp),
+								SLOT_CAN_MADVISE)) {
 							total += NANOV2_BLOCK_SIZE;
 						}
 					}
@@ -1199,7 +1538,8 @@
 			}
 			arena++;
 		}
-		region = nanov2_next_region_for_region(nanozone, region);
+		region = nanov2_next_region_for_region(nanozone, region,
+				current_region_next_arena);
 	}
 
 done:
@@ -1248,15 +1588,13 @@
 	kern_return_t kr;
 	bitarray_t slots;
 
-	if (!reader) {
-		reader = nano_common_default_reader;
-	}
+	reader = reader_or_in_memory_fallback(reader, task);
 
 	kr = reader(task, zone_address, sizeof(nanozonev2_t), (void **)&nanozone);
 	if (kr) {
 		return kr;
 	}
-	boolean_t self_zone = (nanozonev2_t *)zone_address == nanozone;
+	boolean_t self_zone = mach_task_is_self(task) && (nanozonev2_t *)zone_address == nanozone;
 	memcpy(&zone_copy, nanozone, sizeof(zone_copy));
 	nanozone = &zone_copy;
 	nanov2_meta_index_t metablock_meta_index = nanov2_metablock_meta_index(nanozone);
@@ -1264,6 +1602,8 @@
 	// Process the zone one region at a time. Report each in-use block as a
 	// pointer range and each in-use slot as a pointer.
 	nanov2_region_t *region = nanozone->first_region_base;
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&nanozone->current_region_next_arena, acquire);
 	while (region) {
 		mach_vm_address_t vm_addr = (mach_vm_address_t)NULL;
 		kern_return_t kr = reader(task, (vm_address_t)region, NANOV2_REGION_SIZE, (void **)&vm_addr);
@@ -1275,7 +1615,8 @@
 		// and its mapped address in this process.
 		mach_vm_offset_t ptr_offset = (mach_vm_address_t)region - vm_addr;
 		nanov2_arena_t *arena = nanov2_first_arena_for_region(region);
-		nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(nanozone, region);
+		nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(nanozone, region,
+				current_region_next_arena);
 		vm_range_t ptr_range;
 		while (arena < limit_arena) {
 			// Find the metadata block and process every entry, apart from the
@@ -1352,7 +1693,7 @@
 							bitarray_zap(slots, log_size, next_slot);
 							void *ptr = nanov2_slot_in_block_ptr(blockp, size_class, next_slot);
 							nanov2_free_slot_t *slotp = NANOV2_ZONE_PTR_TO_MAPPED_PTR(nanov2_free_slot_t *, ptr, ptr_offset);
-							next_slot = slotp->next_slot;
+							next_slot = (uint16_t)slotp->next_slot;
 							free_list_count++;
 						}
 						// Add a range for each slot that is not on the freelist,
@@ -1371,7 +1712,7 @@
 							ranges[range_count].size = slot_size;
 							range_count++;
 						}
-						free(slots);
+						_free(slots);
 					}
 					if (range_count) {
 						// Notify the in-use pointers that we found.
@@ -1389,7 +1730,8 @@
 		nanov2_region_linkage_t *mapped_region_linkagep =
 				NANOV2_ZONE_PTR_TO_MAPPED_PTR(nanov2_region_linkage_t *,
 				region_linkagep, ptr_offset);
-		int offset = mapped_region_linkagep->next_region_offset;
+		int offset = os_atomic_load(&mapped_region_linkagep->next_region_offset,
+				relaxed);
 		region = offset ? region + offset : NULL;
 	}
 	return 0;
@@ -1408,72 +1750,94 @@
 static boolean_t
 nanov2_check(nanozonev2_t *nanozone)
 {
-	// Does nothing, just like Nano V1.
+	// Does nothing
 	return 1;
 }
 
 static void
-nanov2_print(nanozonev2_t *nanozone, boolean_t verbose)
-{
+nanov2_print(task_t task, unsigned level, vm_address_t zone_address,
+		memory_reader_t reader, print_task_printer_t printer)
+{
+    // Ensure that we have configured enough of the allocator to be able to
+    // examine its data structures. In tools that do not directly use Nano, we
+    // won't have done this yet. nanov2_configure() runs the initialization
+    // only once.
+    nanov2_configure();
+
+	nanozonev2_t *mapped_nanozone;
+	if (reader(task, (vm_address_t)zone_address, sizeof(nanozonev2_t),
+            (void **)&mapped_nanozone)) {
+        printer("Failed to map nanozonev2_s at %p\n", zone_address);
+        return;
+    }
+
 	// Zone-wide statistics
 	malloc_statistics_t stats;
-	nanov2_statistics_t *nano_stats = &nanozone->statistics;
-	nanov2_statistics(nanozone, &stats);
-	malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-			"Nanozonev2 %p: blocks in use: %llu, size in use: %llu allocated size: %llu, "
-			"allocated regions: %d, region holes: %d\n",
-			nanozone, (uint64_t)stats.blocks_in_use, (uint64_t)stats.size_in_use,
-			(uint64_t)stats.size_allocated, nano_stats->allocated_regions,
-			nano_stats->region_address_clashes);
+	nanov2_statistics_task_printer(task, zone_address, reader, printer, &stats);
+	nanov2_statistics_t *nano_stats = &mapped_nanozone->statistics;
+	printer("Nanozonev2 %p: blocks in use: %llu, size in use: %llu "
+			"allocated size: %llu, allocated regions: %d, region holes: %d\n",
+			zone_address, (uint64_t)stats.blocks_in_use,
+			(uint64_t)stats.size_in_use, (uint64_t)stats.size_allocated,
+			nano_stats->allocated_regions, nano_stats->region_address_clashes);
 
 #if DEBUG_MALLOC
 	// Per-size class statistics
-	malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-			"\nPer size-class statistics:\n");
+	printer("\nPer size-class statistics:\n");
 	for (int i = 0; i < NANO_SIZE_CLASSES; i++) {
 		nanov2_size_class_statistics *cs = &nano_stats->size_class_statistics[i];
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-			"  Class %d: ", i);
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-			"total alloc: %llu, total frees: %llu, madvised blocks: %llu, madvise races: %llu",
-			cs->total_allocations, cs->total_frees, cs->madvised_blocks,
-			cs->madvise_races);
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "\n");
+		printer("  Class %d: ", i);
+		printer("total alloc: %llu, total frees: %llu, madvised blocks: %llu, "
+				"madvise races: %llu",
+				cs->total_allocations, cs->total_frees, cs->madvised_blocks,
+				cs->madvise_races);
+		printer("\n");
 	}
 #endif // DEBUG_MALLOC
 
 	// Per-context block pointers.
-	malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-		"Current Allocation Blocks By Size Class/Context [CPU]\n");
+	printer("Current Allocation Blocks By Size Class/Context [CPU]\n");
 	for (int i = 0; i < NANO_SIZE_CLASSES; i++) {
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-				"  Class %d: ", i);
+		printer("  Class %d: ", i);
 		for (int j = 0; j < MAX_CURRENT_BLOCKS; j++) {
-			if (nanozone->current_block[i][j]) {
-				malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-						"%d: %p; ", j, nanozone->current_block[i][j]);
+			if (mapped_nanozone->current_block[i][j]) {
+				printer("%d: %p; ", j, mapped_nanozone->current_block[i][j]);
 			}
 		}
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "\n");
-	}
-
-	nanov2_meta_index_t metablock_meta_index = nanov2_metablock_meta_index(nanozone);
-	nanov2_region_t *region = nanozone->first_region_base;
+		printer("\n");
+	}
+
+	nanov2_meta_index_t metablock_meta_index =
+			nanov2_metablock_meta_index(mapped_nanozone);
+	nanov2_region_t *region = mapped_nanozone->first_region_base;
+	// Use a single, consistent snapshot of current_region_next_arena throughout
+	// iteration, ignoring any arenas or regions allocated after it.
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&mapped_nanozone->current_region_next_arena, acquire);
 	int region_index = 0;
 	while (region) {
-		malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-				"\nRegion %d: base address %p\n", region_index, region);
+		printer("\nRegion %d: base address %p\n", region_index, region);
+		nanov2_region_t *mapped_region;
+		if (reader(task, (vm_address_t)region, sizeof(nanov2_region_t),
+				(void **)&mapped_region)) {
+			printer("Failed to map nanov2 region at %p\n", region);
+			return;
+		}
+        off_t region_offset = (uintptr_t)mapped_region - (uintptr_t)region;
 
 		nanov2_arena_t *arena = nanov2_first_arena_for_region(region);
-		nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(nanozone, region);
+		nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(
+				mapped_nanozone, region, current_region_next_arena);
 		int arena_index = 0;
 		while (arena < limit_arena) {
 			// Find the metadata block and process every entry, apart from the
 			// one for the metadata block itself.
 			nanov2_arena_metablock_t *arena_meta_blockp =
-					nanov2_metablock_address_for_ptr(nanozone, arena);
-
+					nanov2_metablock_address_for_ptr(mapped_nanozone, arena);
+			nanov2_arena_metablock_t *mapped_arena_meta_blockp =
+				(nanov2_arena_metablock_t *)((uintptr_t)arena_meta_blockp + region_offset);
 			nanov2_block_meta_t *block_metap = &arena_meta_blockp->arena_block_meta[0];
+			nanov2_block_meta_t *mapped_block_metap = &mapped_arena_meta_blockp->arena_block_meta[0];
 
 			int active_blocks = 0;
 			int madvisable_blocks = 0;
@@ -1485,7 +1849,7 @@
 					// Skip the metadata block.
 					continue;
 				}
-				nanov2_block_meta_t meta = block_metap[i];
+				nanov2_block_meta_t meta = mapped_block_metap[i];
 				switch (meta.next_slot) {
 				case SLOT_NULL:
 					unused_blocks++;
@@ -1504,8 +1868,7 @@
 					break;
 				}
 			}
-			malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-					"Arena #%d: base address %p. Blocks - active: %d, "
+			printer("Arena #%d: base address %p. Blocks - active: %d, "
 					"madvisable: %d, madvising: %d, madvised: %d, unused: %d\n",
 					arena_index, arena, active_blocks, madvisable_blocks,
 					madvising_blocks, madvised_blocks, unused_blocks);
@@ -1520,9 +1883,9 @@
 					// Skip the metadata block.
 					continue;
 				}
-				nanov2_block_meta_t meta = block_metap[i];
+				nanov2_block_meta_t meta = mapped_block_metap[i];
 				nanov2_size_class_t size_class =
-						nanov2_size_class_for_meta_index(nanozone, i);
+						nanov2_size_class_for_meta_index(mapped_nanozone, i);
 				switch (meta.next_slot) {
 				case SLOT_FULL:
 				case SLOT_BUMP:
@@ -1537,85 +1900,97 @@
 					break;
 				}
 			}
-			malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-					"Size classes with allocated blocks: ");
+			printer("Size classes with allocated blocks: ");
 			for (int i = 0; i < NANO_SIZE_CLASSES; i++) {
 				if (non_empty_size_classes[i]) {
-					malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-						"%d ", i);
+					printer("%d ", i);
 				}
 			}
-			malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX, "\n");
-
-			for (nanov2_meta_index_t i = 0; i < NANOV2_BLOCKS_PER_ARENA; i++) {
-				if (i == metablock_meta_index) {
-					// Skip the metadata block.
-					continue;
-				}
-				nanov2_block_meta_t meta = block_metap[i];
-				if (!nanov2_is_block_active(meta) && !verbose) {
-					continue;
-				}
-				nanov2_size_class_t size_class =
-						nanov2_size_class_for_meta_index(nanozone, i);
-				char *slot_text;
-				switch (meta.next_slot) {
-				case SLOT_NULL:
-					slot_text = "NOT USED";
-					break;
-				case SLOT_FULL:
-					slot_text = "FULL";
-					break;
-				case SLOT_CAN_MADVISE:
-					slot_text = "CAN MADVISE";
-					break;
-				case SLOT_MADVISING:
-					slot_text = "MADVISING";
-					break;
-				case SLOT_MADVISED:
-					slot_text = "MADVISED";
-					break;
-				default:
-					slot_text = NULL;
-					break;
-				}
-				malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-						"    Block %d: base %p; metadata: %p, size %d (class %d) in-use: %d ",
-						i, nanov2_block_address_from_meta_index(nanozone, arena, i),
-						&block_metap[i], nanov2_size_from_size_class(size_class),
-						size_class, meta.in_use);
-				if (slot_text) {
-					malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-						"%s\n", slot_text);
-				} else {
-					int allocated = slots_by_size_class[size_class] - meta.free_count - 1;
-					if (meta.next_slot == SLOT_BUMP) {
-						malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-								"BUMP (free list empty)");
+			printer("\n");
+
+			if (level >= MALLOC_VERBOSE_PRINT_LEVEL) {
+				for (nanov2_meta_index_t i = 0; i < NANOV2_BLOCKS_PER_ARENA; i++) {
+					if (i == metablock_meta_index) {
+						// Skip the metadata block.
+						continue;
+					}
+					nanov2_block_meta_t meta = mapped_block_metap[i];
+					if (!nanov2_is_block_active(meta)) {
+						continue;
+					}
+					nanov2_size_class_t size_class =
+							nanov2_size_class_for_meta_index(mapped_nanozone, i);
+					char *slot_text;
+					switch (meta.next_slot) {
+					case SLOT_NULL:
+						slot_text = "NOT USED";
+						break;
+					case SLOT_FULL:
+						slot_text = "FULL";
+						break;
+					case SLOT_CAN_MADVISE:
+						slot_text = "CAN MADVISE";
+						break;
+					case SLOT_MADVISING:
+						slot_text = "MADVISING";
+						break;
+					case SLOT_MADVISED:
+						slot_text = "MADVISED";
+						break;
+					default:
+						slot_text = NULL;
+						break;
+					}
+					printer("    Block %d: base %p; metadata: %p, size %d "
+							"(class %d) in-use: %d ",
+							i, nanov2_block_address_from_meta_index(mapped_nanozone, arena, i),
+							&block_metap[i], nanov2_size_from_size_class(size_class),
+							size_class, meta.in_use);
+					if (slot_text) {
+						printer("%s\n", slot_text);
 					} else {
-						malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-							"next_slot (1-based) = %d", meta.next_slot);
+						int allocated = slots_by_size_class[size_class] - meta.free_count - 1;
+						if (meta.next_slot == SLOT_BUMP) {
+							printer("BUMP (free list empty)");
+						} else {
+							printer("next_slot (1-based) = %d", meta.next_slot);
 
 					}
-					malloc_report(MALLOC_REPORT_NOLOG | MALLOC_REPORT_NOPREFIX,
-						", allocated slots: %d, free slots = %d, occupancy: %d%%\n",
-						allocated, meta.free_count + 1,
-						(100 * allocated)/slots_by_size_class[size_class]);
+						printer(", allocated slots: %d, free slots = %d, "
+								"occupancy: %d%%\n",
+								allocated, meta.free_count + 1,
+								(100 * allocated)/slots_by_size_class[size_class]);
+					}
 				}
 			}
 			arena++;
 			arena_index++;
 		}
 
-		region = nanov2_next_region_for_region(nanozone, region);
+		region = nanov2_next_region_for_region_offset(mapped_nanozone, region,
+                region_offset, current_region_next_arena);
 		region_index++;
 	}
+}
+
+static void
+nanov2_print_self(nanozonev2_t *nanozone, boolean_t verbose)
+{
+	nanov2_print(mach_task_self(), verbose ? MALLOC_VERBOSE_PRINT_LEVEL : 0,
+			(vm_address_t)nanozone, _malloc_default_reader, malloc_report_simple);
+}
+
+static void
+nanov2_print_task(task_t task, unsigned level, vm_address_t zone_address,
+		memory_reader_t reader, print_task_printer_t printer)
+{
+	nanov2_print(task, level, zone_address, reader, printer);
 }
 
 static void
 nanov2_log(malloc_zone_t *zone, void *log_address)
 {
-	// Does nothing, just like Nano V1.
+	// Does nothing
 }
 
 static void
@@ -1644,35 +2019,74 @@
 }
 
 static void
-nanov2_statistics(nanozonev2_t *nanozone, malloc_statistics_t *stats)
-{
+nanov2_null_printer(const char __unused *fmt, ...)
+{
+}
+
+static kern_return_t
+nanov2_statistics(task_t task, vm_address_t zone_address,
+		memory_reader_t reader, print_task_printer_t printer,
+		malloc_statistics_t *stats)
+{
+	printer = printer ? printer : nanov2_null_printer;
+	reader = reader_or_in_memory_fallback(reader, task);
+
+	kern_return_t err;
+
+    // Ensure that we have configured enough of the allocator to be able to
+    // examine its data structures. In tools that do not directly use Nano, we
+    // won't have done this yet. nanov2_configure() runs the initialization
+    // only once.
+    nanov2_configure();
+
 	memset(stats, '\0', sizeof(*stats));
 
+	nanozonev2_t *mapped_nanozone;
+	err = reader(task, (vm_address_t)zone_address, sizeof(nanozonev2_t),
+				 (void **)&mapped_nanozone);
+	if (err) {
+        printer("Failed to map nanozonev2_s at %p\n", zone_address);
+        return err;
+    }
+
 	nanov2_region_t *region;
-	nanov2_arena_t * arena;
-	nanov2_meta_index_t metadata_block_index = nanov2_metablock_meta_index(nanozone);
+	nanov2_arena_t *arena;
+	nanov2_meta_index_t metadata_block_index =
+			nanov2_metablock_meta_index(mapped_nanozone);
 
 	// Iterate over each arena in each region. Within each region, add
 	// statistics for each slot in each block, excluding the meta data block.
-	for (region = nanozone->first_region_base; region;
-			region = nanov2_next_region_for_region(nanozone, region)) {
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&mapped_nanozone->current_region_next_arena, acquire);
+	for (region = mapped_nanozone->first_region_base; region;) {
+        nanov2_region_t *mapped_region;
+		err = reader(task, (vm_address_t)region, sizeof(nanov2_region_t), (void **)&mapped_region);
+        if (err) {
+            printer("Failed to map nanov2 region at %p\n", region);
+            return err;
+        }
+        off_t region_offset = (uintptr_t)mapped_region - (uintptr_t)region;
 		for (arena = nanov2_first_arena_for_region(region);
-				arena < nanov2_limit_arena_for_region(nanozone, region);
+				arena < nanov2_limit_arena_for_region(mapped_nanozone, region,
+						current_region_next_arena);
 				arena++) {
 			nanov2_arena_metablock_t *meta_block =
-					nanov2_metablock_address_for_ptr(nanozone, arena);
+					nanov2_metablock_address_for_ptr(mapped_nanozone, arena);
+			nanov2_arena_metablock_t *mapped_meta_block =
+				(nanov2_arena_metablock_t *)((uintptr_t)meta_block + region_offset);
 			for (nanov2_meta_index_t i = 0; i < NANOV2_BLOCKS_PER_ARENA; i++) {
 				if (i == metadata_block_index) {
 					// Skip the metadata block.
 					continue;
 				}
 
-				nanov2_block_meta_t *block_metap = &meta_block->arena_block_meta[i];
+				nanov2_block_meta_t *mapped_block_metap = &mapped_meta_block->arena_block_meta[i];
 				nanov2_size_class_t size_class =
-						nanov2_size_class_for_meta_index(nanozone, i);
+						nanov2_size_class_for_meta_index(mapped_nanozone, i);
 				int slot_size = nanov2_size_from_size_class(size_class);
 
-				nanov2_block_meta_t meta = os_atomic_load(block_metap, relaxed);
+				nanov2_block_meta_t meta =
+						os_atomic_load(mapped_block_metap, relaxed);
 				int slots_in_use = 0;
 				switch (meta.next_slot) {
 				case SLOT_NULL:
@@ -1682,6 +2096,8 @@
 				case SLOT_MADVISING:
 					// FALLTHRU
 				case SLOT_MADVISED:
+					// FALLTHRU
+				case SLOT_GUARD:
 					// These blocks have no active content.
 					break;
 				case SLOT_FULL:
@@ -1703,18 +2119,43 @@
 				}
 			}
 		}
-	}
-}
+        region = nanov2_next_region_for_region_offset(mapped_nanozone,
+                region, region_offset, current_region_next_arena);
+	}
+	return KERN_SUCCESS;
+}
+
+static void
+nanov2_statistics_self(nanozonev2_t *nanozone, malloc_statistics_t *stats)
+{
+	nanov2_statistics(mach_task_self(), (vm_address_t)nanozone,
+			_malloc_default_reader, malloc_report_simple, stats);
+}
+
+static kern_return_t
+nanov2_statistics_task_printer(task_t task, vm_address_t zone_address,
+		memory_reader_t reader, print_task_printer_t printer,
+		malloc_statistics_t *stats)
+{
+	return nanov2_statistics(task, zone_address, reader, printer, stats);
+}
+
+static kern_return_t
+nanov2_statistics_task(task_t task, vm_address_t zone_address, memory_reader_t reader, malloc_statistics_t *stats)
+{
+	return nanov2_statistics(task, zone_address, reader, NULL, stats);
+}
+
 
 static const struct malloc_introspection_t nanov2_introspect = {
 	.enumerator = 	(void *)nanov2_ptr_in_use_enumerator,
 	.good_size =	(void *)nanov2_good_size,
 	.check = 		(void *)nanov2_check,
-	.print =		(void *)nanov2_print,
+	.print =		(void *)nanov2_print_self,
 	.log = 			(void *)nanov2_log,
 	.force_lock = 	(void *)nanov2_force_lock,
 	.force_unlock =	(void *)nanov2_force_unlock,
-	.statistics = 	(void *)nanov2_statistics,
+	.statistics = 	(void *)nanov2_statistics_self,
 	.zone_locked =	(void *)nanov2_locked,
 	.enable_discharge_checking = NULL,
 	.disable_discharge_checking = NULL,
@@ -1724,6 +2165,8 @@
 	.enumerate_unavailable_without_blocks = NULL,
 #endif // __BLOCKS__
 	.reinit_lock = 	(void *)nanov2_reinit_lock,
+	.print_task = 	(void *)nanov2_print_task,
+	.task_statistics = (void*)nanov2_statistics_task,
 };
 
 #endif // OS_VARIANT_NOTRESOLVED
@@ -1737,8 +2180,10 @@
 // the allocation, or 0 if the pointer does not correspond to an active
 // allocation. If allow_inner is true, the pointer need not point to the start
 // of the allocation.
-size_t
-nanov2_pointer_size(nanozonev2_t *nanozone, void *ptr, boolean_t allow_inner)
+MALLOC_ALWAYS_INLINE MALLOC_INLINE size_t
+nanov2_pointer_size_inline(nanozonev2_t *nanozone, void *ptr,
+		boolean_t allow_inner, nanov2_size_class_t *size_class_out,
+		nanov2_block_meta_t **block_metap_out)
 {
 	// First check the address signature.
 	if (!nanov2_has_valid_signature((void *)ptr)) {
@@ -1750,25 +2195,40 @@
 		return 0;
 	}
 
+	// Atomically load the value of current_region_next_arena. No thread is
+	// allowed to allocate from an arena until it observes a greater value of
+	// current_region_next_arena, which must have happened before now if we're
+	// being called in the context of a deallocation, so we can safely use it as
+	// the upper bound for an overall address range check.
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&nanozone->current_region_next_arena, relaxed);
+
 	// Bounds check against the active address space.
 	if (ptr < (void *)nanozone->first_region_base ||
-			ptr > (void *)nanozone->current_region_next_arena) {
+			ptr > (void *)current_region_next_arena) {
 		return 0;
 	}
 
 #if NANOV2_MULTIPLE_REGIONS
 	// Need to check that the region part is valid because there could be holes.
 	// Do this only if we know there is a hole.
-	// NOTE: in M2 convergence, use a hashed structure to make this more
-	// efficient.
-	if (nanozone->statistics.region_address_clashes) {
+	//
+	// If we're looking at a legitimately-allocated nano pointer, a load-acquire
+	// of current_region_next_arena must have already happened when its
+	// containing arena was first allocated from, so any region_address_clashes
+	// increment that preceded the store-release of current_region_next_arena
+	// should be visible.
+	//
+	// TODO: use a hashed structure to make this more efficient.
+	if (os_atomic_load(&nanozone->statistics.region_address_clashes, relaxed)) {
 		nanov2_region_t *ptr_region = nanov2_region_address_for_ptr(ptr);
 		nanov2_region_t *region = nanozone->first_region_base;
 		while (region) {
 			if (ptr_region == region) {
 				break;
 			}
-			region = nanov2_next_region_for_region(nanozone, region);
+			region = nanov2_next_region_for_region(nanozone, region,
+					current_region_next_arena);
 		}
 		if (!region) {
 			// Reached the end of the region list without matching - not a
@@ -1805,27 +2265,42 @@
 		return 0;
 	}
 
+	if (size_class_out) {
+		*size_class_out = size_class;
+	}
+	if (block_metap_out) {
+		*block_metap_out = block_metap;
+	}
 	return size;
+}
+
+size_t
+nanov2_pointer_size(nanozonev2_t *nanozone, void *ptr, boolean_t allow_inner)
+{
+	return nanov2_pointer_size_inline(nanozone, ptr, allow_inner, NULL, NULL);
 }
 
 #pragma mark -
 #pragma mark Madvise Management
 
-// Given a pointer to a block and its metadata, calls madvise() on that block
-// if it is in state SLOT_CAN_MADVISE. Returns true on success, false if the
-// block is not in the correct state or if the state changed during the
-// operation.
+// Given a pointer to a block and its metadata, calls madvise() on that block if
+// it is still in the state we expect, either SLOT_CAN_MADVISE or SLOT_MADVISED
+// (the latter expected when we need to pessimistically re-madvise a block we
+// may have touched while racing to allocate against a transition to
+// SLOT_CAN_MADVISE). Returns true on success, false if the block is not in the
+// correct state or if the state changed during the operation.
 //
 // This function must be called with the zone's madvise_lock held
-boolean_t
-nanov2_madvise_block(nanozonev2_t *nanozone, nanov2_block_meta_t *block_metap,
-		nanov2_block_t *blockp, nanov2_size_class_t size_class)
+static boolean_t
+nanov2_madvise_block_locked(nanozonev2_t *nanozone,
+		nanov2_block_meta_t *block_metap, nanov2_block_t *blockp,
+		nanov2_size_class_t size_class, uint32_t expected_state)
 {
 	_malloc_lock_assert_owner(&nanozone->madvise_lock);
 
 	boolean_t madvised = FALSE;
 	nanov2_block_meta_t old_meta = os_atomic_load(block_metap, relaxed);
-	if (old_meta.next_slot == SLOT_CAN_MADVISE) {
+	if (old_meta.next_slot == expected_state) {
 		// Nobody raced with us. We can safely madvise this block. First change
 		// the state to SLOT_MADVISING so that other threads don't try to
 		// grab the block for new allocations.
@@ -1867,6 +2342,18 @@
 	return madvised;
 }
 
+static void
+nanov2_madvise_block(nanozonev2_t *nanozone, nanov2_block_meta_t *block_metap,
+		nanov2_size_class_t size_class, uint32_t expected_state)
+{
+	nanov2_block_t *blockp = nanov2_block_address_from_meta_ptr(nanozone,
+			block_metap);
+	_malloc_lock_lock(&nanozone->madvise_lock);
+	nanov2_madvise_block_locked(nanozone, block_metap, blockp, size_class,
+			expected_state);
+	_malloc_lock_unlock(&nanozone->madvise_lock);
+}
+
 #endif // OS_VARIANT_RESOLVED
 
 #pragma mark -
@@ -1874,68 +2361,124 @@
 
 #if OS_VARIANT_NOTRESOLVED
 
-#if NANOV2_MULTIPLE_REGIONS
-static nanov2_addr_t nanov2_max_region_base = {
-	.fields.nano_signature = NANOZONE_SIGNATURE,
-	.fields.nano_region = NANOV2_MAX_REGION_NUMBER
-};
-#endif // NANOV2_MULTIPLE_REGIONS
-
+#if CONFIG_NANO_RESERVE_REGIONS
+// Update protection for region to DEFAULT
+static bool
+nanov2_unprotect_region(nanov2_region_t *region)
+{
+	MALLOC_TRACE(TRACE_nanov2_region_protection | DBG_FUNC_START,
+			(uint64_t)region, 0, 0, 0);
+	bool result = nano_common_unprotect_vm_space((mach_vm_address_t)region,
+			NANOV2_REGION_SIZE);
+	MALLOC_TRACE(TRACE_nanov2_region_protection | DBG_FUNC_END,
+			(uint64_t)region, result, 0, 0);
+	return result;
+}
+
+// Reserve VA at [base, base+num_regions*REGION_SIZE].
+// Note: permissions must still be granted on reserved region with `nanov2_unprotect_region`
+static bool
+nanov2_reserve_regions(nanov2_region_t *base, unsigned int num_regions)
+{
+	MALLOC_TRACE(TRACE_nanov2_region_reservation | DBG_FUNC_START,
+			(uint64_t)base, num_regions, 0, 0);
+	bool result = nano_common_reserve_vm_space((mach_vm_address_t)base,
+			(NANOV2_REGION_SIZE * (mach_vm_size_t)num_regions));
+	MALLOC_TRACE(TRACE_nanov2_region_reservation | DBG_FUNC_END,
+			(uint64_t)base, num_regions, result, 0);
+
+	return result;
+}
+#else
 // Attempts to allocate VM space for a region at a given address and returns
 // whether the allocation succeeded.
-static boolean_t
+static bool
 nanov2_allocate_region(nanov2_region_t *region)
 {
 	MALLOC_TRACE(TRACE_nanov2_region_allocation | DBG_FUNC_START,
 			(uint64_t)region, 0, 0, 0);
-	boolean_t result = nano_common_allocate_vm_space((mach_vm_address_t)region,
+	bool result = nano_common_allocate_vm_space((mach_vm_address_t)region,
 			NANOV2_REGION_SIZE);
 	MALLOC_TRACE(TRACE_nanov2_region_allocation | DBG_FUNC_END,
 			(uint64_t)region, result, 0, 0);
 	return result;
 }
+#endif // CONFIG_NANO_RESERVE_REGIONS
 
 // Allocates a new region adjacent to the current one. If the allocation fails,
 // keep sliding up by the size of a region until we either succeed or run out of
-// address space. The caller must own the Nanozone regions lock.
-boolean_t
+// address space. The caller must own the Nanozone regions lock. Returns the
+// first arena of the newly-allocated region if successful, or NULL otherwise.
+MALLOC_NOEXPORT nanov2_arena_t *
 nanov2_allocate_new_region(nanozonev2_t *nanozone)
 {
 #if NANOV2_MULTIPLE_REGIONS
-	boolean_t result = FALSE;
+	bool allocated = false;
+
+	nanov2_addr_t nanov2_max_region_base = {
+		.fields.nano_signature = NANOZONE_SIGNATURE,
+		.fields.nano_region = nano_max_region,
+	};
 
 	_malloc_lock_assert_owner(&nanozone->regions_lock);
-	nanov2_region_t *current_region = nanozone->current_region_base;
-	nanov2_region_t *next_region = (nanov2_region_t *)nanozone->current_region_limit;
+	nanov2_region_t *current_region = nanov2_current_region_base(
+			os_atomic_load(&nanozone->current_region_next_arena, relaxed));
+	nanov2_region_t *next_region = current_region + 1;
+
 	while ((void *)next_region <= nanov2_max_region_base.addr) {
+#if CONFIG_NANO_RESERVE_REGIONS
+		if (!nanov2_unprotect_region(next_region)) {
+			MALLOC_REPORT_FATAL_ERROR(next_region,
+					"Nano: Unable to raise protection on pre-allocated region");
+		}
+		nanozone->statistics.allocated_regions++;
+		allocated = true;
+		break;
+#else // CONFIG_NANO_RESERVE_REGIONS
 		if (nanov2_allocate_region(next_region)) {
-			nanozone->current_region_base = next_region;
-			nanozone->current_region_next_arena = (nanov2_arena_t *)next_region;
-			nanozone->current_region_limit = next_region + 1;
 			nanozone->statistics.allocated_regions++;
-			result = TRUE;
+			allocated = true;
 			break;
 		}
 		next_region++;
-		nanozone->statistics.region_address_clashes++;
-	}
-
-	if (result) {
-		// Link this region to the previous one.
-		nanov2_region_linkage_t *current_region_linkage =
-				nanov2_region_linkage_for_region(nanozone, current_region);
-		nanov2_region_linkage_t *next_region_linkage =
-				nanov2_region_linkage_for_region(nanozone, next_region);
-		uint16_t offset = next_region - current_region;
-		current_region_linkage->next_region_offset = offset;
-		next_region_linkage->next_region_offset = 0;
-	}
-
-	return result;
+
+		// Loaded atomically in nanov2_pointer_size() to determine whether or
+		// not it's necessary to walk the region list, so we need to increment
+		// atomically here. Published by the store-release of
+		// current_region_next_arena.
+		os_atomic_inc(&nanozone->statistics.region_address_clashes, relaxed);
+#endif // CONFIG_NANO_RESERVE_REGIONS
+	}
+
+	if (!allocated) {
+		return NULL;
+	}
+
+	// Link this region to the previous one.
+	nanov2_region_linkage_t *current_region_linkage =
+			nanov2_region_linkage_for_region(nanozone, current_region);
+
+	// The linkage of the next region is in pristine memory, so already zero -
+	// don't touch it.
+
+	// Store-release the linkage update so any dependent loads through it
+	// observe the (implicit zero-)initialization of the next region.
+	uint16_t offset = next_region - current_region;
+	os_atomic_store(&current_region_linkage->next_region_offset, offset,
+			release);
+
+	// Store-release the update to current_region_next_arena to publish the
+	// linkage update. Pairs with load-acquires of current_region_next_arena
+	// followed by walks of the region list.
+	nanov2_arena_t *first_arena = nanov2_first_arena_for_region(next_region);
+	os_atomic_store(&nanozone->current_region_next_arena, first_arena + 1,
+			release);
+
+	return first_arena;
 #else // NANOV2_MULTIPLE_REGIONS
 	// On iOS, only one region is supported, so we fail since the first
 	// region is allocated separately.
-	return FALSE;
+	return NULL;
 #endif // CONFIG_NANOV2_MULTIPLE_REGIONS
 }
 #endif // OS_VARIANT_NOTRESOLVED
@@ -1944,18 +2487,40 @@
 #pragma mark Allocation
 
 #if OS_VARIANT_RESOLVED
+
+MALLOC_NOINLINE MALLOC_NORETURN
+static void
+nanov2_guard_corruption_detected(void *corrupt_slot)
+{
+	uint64_t guard = *(uint64_t *)corrupt_slot;
+	malloc_zone_error(MALLOC_ABORT_ON_CORRUPTION, true,
+			"Heap corruption detected, free list is damaged at %p\n"
+			"*** Incorrect guard value: %llu\n", corrupt_slot,
+			(unsigned long long)guard);
+	__builtin_unreachable();
+}
 
 // Allocates memory from the block that corresponds to a given block meta data
 // pointer. The memory is taken from the free list if possible, or from the
 // unused region of the block if not. If the block is no longer in use or is
 // full, NULL is returned and the caller is expected to find another block to
 // allocate from.
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
 void *
-nanov2_allocate_from_block(nanozonev2_t *nanozone,
-		nanov2_block_meta_t *block_metap, nanov2_size_class_t size_class)
+nanov2_allocate_from_block_inline(nanozonev2_t *nanozone,
+		nanov2_block_meta_t *block_metap, nanov2_size_class_t size_class,
+		nanov2_block_meta_t **madvise_block_metap_out, bool *corruption)
 {
 	nanov2_block_meta_view_t old_meta_view;
-	old_meta_view.meta = os_atomic_load(block_metap, relaxed);
+
+	// Our loads of the block metadata use dependency ordering, which guarantees
+	// that any loads we do from a slot pointer derived from the metadata value
+	// as we do below will observe all of the stores preceding the store-release
+	// of that value we observed.  This allows us to safely rely on the contents
+	// of the slot updated when it was last freed, including the double-free
+	// guard and zeroing done by zero-on-free (which is required for correctness
+	// in the case of calloc).
+	old_meta_view.meta = os_atomic_load(block_metap, dependency);
 
 	// Calculating blockp and ptr is relatively expensive. Do both lazily to
 	// minimize the time in the block starting with "again:" and ending with the
@@ -1996,7 +2561,7 @@
 		slot = old_meta_view.meta.next_slot - 1; // meta.next_slot is 1-based.
 		ptr = nanov2_slot_in_block_ptr(blockp, size_class, slot);
 		nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
-		new_meta.next_slot = slot_full ? SLOT_FULL : slotp->next_slot;
+		new_meta.next_slot = slot_full ? SLOT_FULL : (uint16_t)slotp->next_slot;
 	}
 
 	// Write the updated meta data; try again if we raced with another thread.
@@ -2005,21 +2570,8 @@
 		if (old_meta_view.meta.next_slot == SLOT_CAN_MADVISE ||
 				old_meta_view.meta.next_slot == SLOT_MADVISING ||
 				old_meta_view.meta.next_slot == SLOT_MADVISED) {
-			_malloc_lock_lock(&nanozone->madvise_lock);
-			if (old_meta_view.meta.next_slot == SLOT_MADVISED) {
-				// We raced against another thread madvising this block. We need
-				// to redo the madvise because we may have touched it when
-				// reading the next pointer in the freelist.
-				if (!blockp) {
-					blockp = nanov2_block_address_from_meta_ptr(nanozone, block_metap);
-				}
-				if (mvm_madvise_free(nanozone, nanov2_region_address_for_ptr(blockp),
-						(uintptr_t)blockp, (uintptr_t)(blockp + 1), NULL, FALSE)) {
-					malloc_zone_error(0, false,
-							"Failed to remadvise block at blockp: %p, error: %d\n", blockp, errno);
-				}
-			}
-			_malloc_lock_unlock(&nanozone->madvise_lock);
+			*madvise_block_metap_out = block_metap;
+			return NULL;
 		}
 		goto again;
 	}
@@ -2031,27 +2583,40 @@
 		ptr = nanov2_slot_in_block_ptr(blockp, size_class, slot);
 	}
 
-	nanov2_free_slot_t *slotp =
-			(nanov2_free_slot_t *)os_atomic_force_dependency_on(ptr,
-			(unsigned long)old_meta_view.bits);
 	if (from_free_list) {
 		// We grabbed the item from the free list. Check the free list canary
 		// and crash if it's not valid. We can't do this check before the
 		// cmpxchgv because another thread may race with us, claim the slot and
 		// write to it.
+		nanov2_free_slot_t *slotp = ptr;
 		uintptr_t guard = os_atomic_load(&slotp->double_free_guard, relaxed);
-		if ((guard ^ nanozone->slot_freelist_cookie) != (uintptr_t)ptr) {
-			malloc_zone_error(MALLOC_ABORT_ON_CORRUPTION, false,
-					"Heap corruption detected, free list is damaged at %p\n"
-					"*** Incorrect guard value: %lu\n", ptr, guard);
-			__builtin_unreachable();
-		}
-	}
-	
+		if (os_unlikely((guard ^ nanozone->slot_freelist_cookie) != (uintptr_t)ptr)) {
+			*corruption = true;
+		}
+	}
+
 #if DEBUG_MALLOC
 	nanozone->statistics.size_class_statistics[size_class].total_allocations++;
 #endif // DEBUG_MALLOC
 
+	return ptr;
+}
+
+static void *
+nanov2_allocate_from_block(nanozonev2_t *nanozone,
+		nanov2_block_meta_t *block_metap, nanov2_size_class_t size_class)
+{
+	nanov2_block_meta_t *madvise_block_metap = NULL;
+	bool corruption = false;
+	void *ptr = nanov2_allocate_from_block_inline(nanozone, block_metap,
+			size_class, &madvise_block_metap, &corruption);
+	if (os_unlikely(corruption)) {
+		nanov2_guard_corruption_detected(ptr);
+	}
+	if (madvise_block_metap) {
+		nanov2_madvise_block(nanozone, madvise_block_metap, size_class,
+				SLOT_MADVISED);
+	}
 	return ptr;
 }
 
@@ -2266,7 +2831,7 @@
 //
 // In order to avoid races, this function must be called with the
 // current_block_lock for the calling context [CPU] and size class locked.
-MALLOC_NOINLINE void *
+MALLOC_NOEXPORT MALLOC_NOINLINE void *
 nanov2_find_block_and_allocate(nanozonev2_t *nanozone,
 		nanov2_size_class_t size_class, nanov2_block_meta_t **block_metapp)
 {
@@ -2286,8 +2851,13 @@
 	start_region = nanov2_region_address_for_ptr(arena);
 	nanov2_arena_t *start_arena = arena;
 	nanov2_region_t *region = start_region;
-	nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(nanozone, start_region);
-	nanov2_arena_t *initial_region_next_arena = nanozone->current_region_next_arena;
+	// The load-acquire pairs with store-release in nanov2_allocate_new_region()
+	// to make the most recent region linkage update visible when we load it in
+	// nanov2_next_region_for_region() below.
+	nanov2_arena_t *initial_region_next_arena = os_atomic_load(
+			&nanozone->current_region_next_arena, acquire);
+	nanov2_arena_t *limit_arena = nanov2_limit_arena_for_region(nanozone,
+			start_region, initial_region_next_arena);
 	do {
 		nanov2_block_meta_t *block_metap = nanov2_find_block_in_arena(nanozone,
 				arena, size_class, start_block);
@@ -2323,13 +2893,15 @@
 		start_block = NULL;
 		arena++;
 		if (arena >= limit_arena) {
-			region = nanov2_next_region_for_region(nanozone, region);
+			region = nanov2_next_region_for_region(nanozone, region,
+					initial_region_next_arena);
 			if (!region) {
 				// Reached the last region -- loop back to the first.
 				region = nanozone->first_region_base;
 			}
 			arena = nanov2_first_arena_for_region(region);
-			limit_arena = nanov2_limit_arena_for_region(nanozone, region);
+			limit_arena = nanov2_limit_arena_for_region(nanozone, region,
+					initial_region_next_arena);
 		}
 	} while (arena != start_arena);
 
@@ -2340,24 +2912,42 @@
 	}
 
 	// Allocate a new arena and maybe a new region. To do either of those
-	// things, we need to take the regions_lock. After doing so, check that
-	// the state is unchanged. If it has, just assume that we might have some
-	// new space to allocate into and try again.
+	// things, we need to take the regions_lock. After doing so, check that the
+	// state is unchanged. If it has, just assume that we might have some new
+	// space to allocate into and try again.
+
 	boolean_t failed = FALSE;
-	arena = initial_region_next_arena;
+
 	_malloc_lock_lock(&nanozone->regions_lock);
-	if (nanozone->current_region_next_arena == arena) {
-		if ((void *)arena >= nanozone->current_region_limit) {
+	nanov2_arena_t *current_region_next_arena = os_atomic_load(
+			&nanozone->current_region_next_arena, relaxed);
+	if (current_region_next_arena == initial_region_next_arena) {
+		if (nanov2_current_region_next_arena_is_limit(
+				current_region_next_arena)) {
 			// Reached the end of the region. Allocate a new one, if we can.
-			if (nanov2_allocate_new_region(nanozone)) {
-				arena = nanozone->current_region_next_arena++;
-			} else {
+			arena = nanov2_allocate_new_region(nanozone);
+			if (!arena) {
 				failed = TRUE;
 			}
 		} else {
-			// Assign the new arena, in the same region.
-			nanozone->current_region_next_arena = arena + 1;
-		}
+			// Assign the new arena, in the current region.
+			arena = current_region_next_arena;
+
+			// Bump current_region_next_arena by 1. No need for an atomic add
+			// because we're under the regions_lock.
+			os_atomic_store(&nanozone->current_region_next_arena,
+					current_region_next_arena + 1, relaxed);
+		}
+
+		// Set up the guard blocks for the new arena, if requested
+		if (!failed) {
+			nanov2_init_guard_blocks(nanozone, arena);
+		}
+	} else {
+		// The arena just before current_region_next_arena is always the most
+		// recently allocated arena. Let's retry from that arena, which was
+		// allocated in the time since we started our last try.
+		arena = current_region_next_arena - 1;
 	}
 	_malloc_lock_unlock(&nanozone->regions_lock);
 
@@ -2374,46 +2964,49 @@
 	return NULL;
 }
 
-// Allocates memory of a given size (which must be a multiple of the Nano
-// quantum size) and optionally clears it (for calloc).
+// This function is called when a fast-path allocation from a given (size_class,
+// allocation_index) has been tried and failed, and we need to act on
+// observations from that attempt and/or retry the allocation.  Its rather
+// tortured calling contract is designed to allow the caller to avoid pushing a
+// frame and pass along as much of what it has already computed as possible.
 //
-// Allocation is attempted first from the block last used for the caller's
-// context (which is initially the physical CPU by default). If there is no
-// last block, or the block is full or now out of use, find another one, if
-// possible. See the comments for nanov2_get_allocation_block() for the details.
+// If @corrupt_slot is non-NULL it means we detected corruption of the slot's
+// guard on the fast path, and we need to report that corruption.
+//
+// If @madvise_block_metap is non-NULL it means we raced with another thread
+// madvising the block we tried to allocate from and need to re-madvise it.
 //
 // If the allocation fails, NULL is returned.
-void *
-nanov2_allocate(nanozonev2_t *nanozone, size_t rounded_size, boolean_t clear)
+static void *
+nanov2_allocate_outlined(nanozonev2_t *nanozone, nanov2_block_meta_t **block_metapp,
+		size_t rounded_size, nanov2_size_class_t size_class,
+		int allocation_index, nanov2_block_meta_t *madvise_block_metap,
+		void *corrupt_slot, bool clear, bool typed, malloc_type_id_t type_id)
 {
 	void *ptr = NULL;
-	nanov2_size_class_t size_class = nanov2_size_class_from_size(rounded_size);
-	MALLOC_ASSERT(size_class < NANO_SIZE_CLASSES);
-	MALLOC_ASSERT(rounded_size != 0);
-	nanov2_block_meta_t *block_metap;
-	nanov2_block_meta_t **block_metapp;
-
-	// Get the index of the pointer to the block from which we are should be
-	// allocating. This currently depends on the physical CPU number.
-	int allocation_index = nanov2_get_allocation_block_index() & MAX_CURRENT_BLOCKS_MASK;
-
-	// Get the current allocation block meta data pointer. If this is NULL,
-	// we need to find a new allocation block.
-	block_metapp = &nanozone->current_block[size_class][allocation_index];
-	block_metap = os_atomic_load(block_metapp, relaxed);
-	if (block_metap) {
-		// Fast path: we have a block -- try to allocate from it.
-		ptr = nanov2_allocate_from_block(nanozone, block_metap, size_class);
-		if (ptr) {
-			goto done;
-		}
+
+	if (os_unlikely(corrupt_slot)) {
+		nanov2_guard_corruption_detected(corrupt_slot);
+	}
+
+	// If we need to re-madvise the old block that we might have accidentally
+	// touched out of turn, do so now.
+	if (madvise_block_metap) {
+		nanov2_madvise_block(nanozone, madvise_block_metap, size_class,
+				SLOT_MADVISED);
 	}
 
 	// No current allocation block, or we were unable to allocate. We need to
 	// get a new block. Before doing so, delegate to the helper allocator if
 	// the size class was full and has not released enough memory yet.
 	if (nanozone->delegate_allocations & (1 << size_class)) {
-		ptr = nanozone->helper_zone->malloc(nanozone->helper_zone, rounded_size);
+		malloc_zone_t *helper_zone = nanozone->helper_zone;
+		if (typed) {
+			ptr = helper_zone->malloc_type_malloc(helper_zone, rounded_size,
+					type_id);
+		} else {
+			ptr = helper_zone->malloc(helper_zone, rounded_size);
+		}
 		goto done;
 	}
 
@@ -2423,7 +3016,7 @@
 	_malloc_lock_s *lock = &nanozone->current_block_lock[size_class][allocation_index];
 	_malloc_lock_lock(lock);
 
-	block_metap = os_atomic_load(block_metapp, relaxed);
+	nanov2_block_meta_t *block_metap = os_atomic_load(block_metapp, relaxed);
 	if (block_metap) {
 		ptr = nanov2_allocate_from_block(nanozone, block_metap, size_class);
 		if (ptr) {
@@ -2445,21 +3038,49 @@
 		// We could not find a block to allocate from -- make future
 		// allocations for this size class go to the helper zone until
 		// we have enough free space.
-		_malloc_lock_lock(&nanozone->delegate_allocations_lock);
-		nanozone->delegate_allocations |= 1 << size_class;
-		_malloc_lock_unlock(&nanozone->delegate_allocations_lock);
+		os_atomic_or(&nanozone->delegate_allocations,
+				(uint16_t)(1 << size_class), relaxed);
+
+		malloc_zone_t *helper_zone = nanozone->helper_zone;
+		if (typed) {
+			ptr = helper_zone->malloc_type_malloc(helper_zone, rounded_size,
+					type_id);
+		} else {
+			ptr = helper_zone->malloc(helper_zone, rounded_size);
+		}
 	}
 
 done:
-	if (ptr) {
-		if (clear) {
-			memset(ptr, '\0', rounded_size);
-		} else {
+	if (os_likely(ptr)) {
+		switch (malloc_zero_policy) {
+		case MALLOC_ZERO_ON_FREE: {
 			// Always clear the double-free guard so that we can recognize that
 			// this block is not on the free list.
 			nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
 			os_atomic_store(&slotp->double_free_guard, 0, relaxed);
-		}
+
+			// We know the body of the allocation is already clear, so we just
+			// need to clean up the next_slot word to get to all-zero.  Do so in
+			// all cases, even if a cleared allocation is not requested, to
+			// prevent any leakage through the next_slot bits.
+			os_atomic_store(&slotp->next_slot, 0, relaxed);
+			break;
+		}
+		case MALLOC_ZERO_NONE:
+			if (!clear) {
+				// Always clear the double-free guard so that we can recognize that
+				// this block is not on the free list.
+				nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
+				os_atomic_store(&slotp->double_free_guard, 0, relaxed);
+				break;
+			}
+			MALLOC_FALLTHROUGH;
+		case MALLOC_ZERO_ON_ALLOC:
+			memset(ptr, '\0', rounded_size);
+			break;
+		}
+	} else {
+		malloc_set_errno_fast(MZ_POSIX, ENOMEM);
 	}
 	return ptr;
 }
@@ -2468,20 +3089,32 @@
 #pragma mark Freeing
 
 // Frees an allocation to its owning block and updates the block's state.
-// If the block becomes empty, it is marked as SLOT_CAN_MADVISE and is
-// madvised immediately if the policy is NANO_MADVISE_IMMEDIATE.
-void
-nanov2_free_to_block(nanozonev2_t *nanozone, void *ptr,
-		nanov2_size_class_t size_class)
+//
+// If the block becomes empty, it is marked as SLOT_CAN_MADVISE and we return
+// the block to the caller to madvise if dictated by policy.
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
+nanov2_block_meta_t *
+nanov2_free_to_block_inline(nanozonev2_t *nanozone, void *ptr,
+		nanov2_size_class_t size_class, nanov2_block_meta_t *block_metap)
 {
 	nanov2_block_t *blockp = nanov2_block_address_for_ptr(ptr);
-	nanov2_block_meta_t *block_metap = nanov2_meta_ptr_for_ptr(nanozone, ptr);
+	if (!block_metap) {
+		block_metap = nanov2_meta_ptr_for_ptr(nanozone, ptr);
+	}
 
 	// Release the slot memory onto the block's freelist.
 	nanov2_block_meta_t old_meta = os_atomic_load(block_metap, relaxed);
 	int slot_count = slots_by_size_class[size_class];
 	nanov2_block_meta_t new_meta;
 	boolean_t was_full;
+
+	nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
+	// All of the free slot content (double_free_guard, next_slot word and the
+	// zeroed remainder of the slot) must be visible when the os_atomic_cmpxchgv
+	// completes, so the metadata updates on either path below need a release
+	// barrier.
+	os_atomic_store(&slotp->double_free_guard,
+			nanozone->slot_freelist_cookie ^ (uintptr_t)ptr, relaxed);
 
 again:
 	was_full = old_meta.next_slot == SLOT_FULL;
@@ -2491,12 +3124,14 @@
 	boolean_t freeing_last_active_slot = !was_full &&
 			new_meta.free_count == slots_by_size_class[size_class] - 1;
 	if (freeing_last_active_slot) {
+		os_atomic_store(&slotp->next_slot, SLOT_NULL, relaxed);
+
 		// Releasing the last active slot onto the free list. Mark the block as
 		// ready to be madvised if it's not in use, otherwise reset next_slot
 		// to SLOT_BUMP.
 		new_meta.next_slot = new_meta.in_use ? SLOT_BUMP : SLOT_CAN_MADVISE;
 		// Write the updated meta data; try again if we raced with another thread.
-		if (!os_atomic_cmpxchgv(block_metap, old_meta, new_meta, &old_meta, relaxed)) {
+		if (!os_atomic_cmpxchgv(block_metap, old_meta, new_meta, &old_meta, release)) {
 			goto again;
 		}
 
@@ -2504,20 +3139,14 @@
 		// is to do so immediately.
 		if (new_meta.next_slot == SLOT_CAN_MADVISE &&
 				nanov2_madvise_policy == NANO_MADVISE_IMMEDIATE) {
-			_malloc_lock_lock(&nanozone->madvise_lock);
-			nanov2_madvise_block(nanozone, block_metap, blockp, size_class);
-			_malloc_lock_unlock(&nanozone->madvise_lock);
+			return block_metap;
 		}
 	} else {
 		int slot_index = nanov2_slot_index_in_block(blockp, size_class, ptr);
 		new_meta.next_slot = slot_index + 1;  // meta.next_slot is 1-based
-		nanov2_free_slot_t *slotp = (nanov2_free_slot_t *)ptr;
-		slotp->next_slot = was_full ? SLOT_BUMP : old_meta.next_slot;
-		os_atomic_store(&slotp->double_free_guard,
-				nanozone->slot_freelist_cookie ^ (uintptr_t)ptr, relaxed);
-
-		// The double_free_guard change must be visible when the os_atomic_cmpxchgv
-		// completes.
+		os_atomic_store(&slotp->next_slot,
+				was_full ? SLOT_BUMP : old_meta.next_slot, relaxed);
+
 		// Write the updated meta data; try again if we raced with another thread.
 		if (!os_atomic_cmpxchgv(block_metap, old_meta, new_meta, &old_meta, release)) {
 			goto again;
@@ -2531,14 +3160,14 @@
 	uint16_t class_mask = 1 << size_class;
 	if (!new_meta.in_use && (nanozone->delegate_allocations & class_mask) &&
 			(new_meta.free_count >= 0.75 * slot_count)) {
-		_malloc_lock_lock(&nanozone->delegate_allocations_lock);
-		nanozone->delegate_allocations &= ~class_mask;
-		_malloc_lock_unlock(&nanozone->delegate_allocations_lock);
+		os_atomic_and(&nanozone->delegate_allocations, ~class_mask, relaxed);
 	}
 
 #if DEBUG_MALLOC
 	nanozone->statistics.size_class_statistics[size_class].total_frees++;
 #endif // DEBUG_MALLOC
+
+	return NULL;
 }
 
 #endif // OS_VARIANT_RESOLVED
@@ -2551,11 +3180,9 @@
 malloc_zone_t *
 nanov2_create_zone(malloc_zone_t *helper_zone, unsigned debug_flags)
 {
-	// Note: It is important that nanov2_create_zone resets _malloc_engaged_nano
-	// if it is unable to enable the nanozone (and chooses not to abort). As
-	// several functions rely on _malloc_engaged_nano to determine if they
-	// should manipulate the nanozone, and these should not run if we failed
-	// to create the zone.
+	// Note: It is not necessary that nanov2_create_zone resets _malloc_engaged_nano
+	// if it is unable to enable the nanozone - functions that need to determine
+	// whether the nanozone is preset should test initial_nano_zone.
 	MALLOC_ASSERT(_malloc_engaged_nano == NANO_V2);
 
 	// Get memory for the zone and disable Nano if we fail.
@@ -2567,9 +3194,18 @@
 	}
 
 	// Set up the basic_zone portion of the nanozonev2 structure
-	nanozone->basic_zone.version = 10;
+	nanozone->basic_zone.version = 16;
 	nanozone->basic_zone.size = OS_RESOLVED_VARIANT_ADDR(nanov2_size);
-	nanozone->basic_zone.malloc = OS_RESOLVED_VARIANT_ADDR(nanov2_malloc);
+	if (malloc_zero_policy == MALLOC_ZERO_ON_ALLOC) {
+		nanozone->basic_zone.malloc =
+				OS_RESOLVED_VARIANT_ADDR(nanov2_malloc_zero_on_alloc);
+		nanozone->basic_zone.malloc_type_malloc =
+				OS_RESOLVED_VARIANT_ADDR(nanov2_malloc_type_zero_on_alloc);
+	} else {
+		nanozone->basic_zone.malloc = OS_RESOLVED_VARIANT_ADDR(nanov2_malloc);
+		nanozone->basic_zone.malloc_type_malloc =
+				OS_RESOLVED_VARIANT_ADDR(nanov2_malloc_type);
+	}
 	nanozone->basic_zone.calloc = OS_RESOLVED_VARIANT_ADDR(nanov2_calloc);
 	nanozone->basic_zone.valloc = (void *)nanov2_valloc;
 	nanozone->basic_zone.free = OS_RESOLVED_VARIANT_ADDR(nanov2_free);
@@ -2579,10 +3215,18 @@
 	nanozone->basic_zone.batch_free = OS_RESOLVED_VARIANT_ADDR(nanov2_batch_free);
 	nanozone->basic_zone.introspect =
 			(struct malloc_introspection_t *)&nanov2_introspect;
-	nanozone->basic_zone.memalign = (void *)nanov2_memalign;
+	nanozone->basic_zone.memalign = OS_RESOLVED_VARIANT_ADDR(nanov2_memalign);
 	nanozone->basic_zone.free_definite_size = OS_RESOLVED_VARIANT_ADDR(nanov2_free_definite_size);
 	nanozone->basic_zone.pressure_relief = OS_RESOLVED_VARIANT_ADDR(nanov2_pressure_relief);
 	nanozone->basic_zone.claimed_address = OS_RESOLVED_VARIANT_ADDR(nanov2_claimed_address);
+	nanozone->basic_zone.try_free_default = OS_RESOLVED_VARIANT_ADDR(nanov2_try_free_default);
+
+	// zone_type == MALLOC_ZONE_TYPE_UNKNOWN
+	// No malloc_with_options
+
+	nanozone->basic_zone.malloc_type_calloc = OS_RESOLVED_VARIANT_ADDR(nanov2_calloc_type);
+	nanozone->basic_zone.malloc_type_realloc = OS_RESOLVED_VARIANT_ADDR(nanov2_realloc_type);
+	nanozone->basic_zone.malloc_type_memalign = OS_RESOLVED_VARIANT_ADDR(nanov2_memalign_type);
 
 	// Set these both to zero as required by CFAllocator.
 	nanozone->basic_zone.reserved1 = 0;
@@ -2590,12 +3234,6 @@
 
 	// Prevent overwriting the function pointers in basic_zone.
 	mprotect(nanozone, sizeof(nanozone->basic_zone), PROT_READ);
-
-	// Nano V2 zone does not support MALLOC_ADD_GUARD_PAGES
-	if (debug_flags & MALLOC_ADD_GUARD_PAGES) {
-		malloc_report(ASL_LEVEL_INFO, "nano does not support guard pages\n");
-		debug_flags &= ~MALLOC_ADD_GUARD_PAGES;
-	}
 
 	// Set up the remainder of the nanozonev2 structure
 	nanozone->debug_flags = debug_flags;
@@ -2627,26 +3265,43 @@
 	_malloc_lock_init(&nanozone->madvise_lock);
 
 	// Allocate the initial region. If this does not succeed, we disable Nano.
-	nanov2_addr_t p = {.fields.nano_signature = NANOZONE_SIGNATURE};
-	nanov2_region_t *region = (nanov2_region_t *)p.addr;
-	boolean_t result = nanov2_allocate_region(region);
+	nanov2_region_t *region = (nanov2_region_t *)NANOZONE_BASE_REGION_ADDRESS;
+
+	bool result;
+#if CONFIG_NANO_RESERVE_REGIONS
+	unsigned int num_regions = (nano_max_region + 1);
+	result = nanov2_reserve_regions(region, num_regions);
+	if (result) {
+		result = nanov2_unprotect_region(region);
+		if (!result) {
+			malloc_report(ASL_LEVEL_ERR,
+					"unable to protect initial region\n");
+			nano_common_deallocate_pages((void *)region,
+					num_regions * (size_t)NANOV2_REGION_SIZE, 0);
+		}
+	}
+#else // CONFIG_NANO_RESERVE_REGIONS
+	result = nanov2_allocate_region(region);
+#endif // CONFIG_NANO_RESERVE_REGIONS
 	if (!result) {
-		nano_common_deallocate_pages(nanozone, NANOZONEV2_ZONE_PAGED_SIZE, 0);
+		nano_common_deallocate_pages((void *)nanozone,
+				NANOZONEV2_ZONE_PAGED_SIZE, 0);
 		_malloc_engaged_nano = NANO_NONE;
 		malloc_report(ASL_LEVEL_NOTICE, "nano zone abandoned due to inability "
-				"to preallocate reserved vm space.\n");
+				"to reserve vm space.\n");
 		return NULL;
 	}
 	nanov2_region_linkage_t *region_linkage =
 			nanov2_region_linkage_for_region(nanozone, region);
-	region_linkage->next_region_offset = 0;
+	os_atomic_store(&region_linkage->next_region_offset, 0, relaxed);
 
 	// Install the first region and pre-allocate the first arena.
 	nanozone->first_region_base = region;
-	nanozone->current_region_base = region;
-	nanozone->current_region_next_arena = ((nanov2_arena_t *)region) + 1;
-	nanozone->current_region_limit = region + 1;
+	os_atomic_store(&nanozone->current_region_next_arena,
+			((nanov2_arena_t *)region) + 1, release);
 	nanozone->statistics.allocated_regions = 1;
+	// Set up the guard blocks for the initial arena, if requested
+	nanov2_init_guard_blocks(nanozone, (nanov2_arena_t *)region);
 
 	return (malloc_zone_t *)nanozone;
 }
@@ -2663,11 +3318,20 @@
 // leak, but this is better than possibly crashing.
 
 #if OS_VARIANT_RESOLVED
-void *
+MALLOC_NOEXPORT void *
 nanov2_forked_malloc(nanozonev2_t *nanozone, size_t size)
 {
 	// Just hand to the helper zone.
 	return nanozone->helper_zone->malloc(nanozone->helper_zone, size);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_forked_malloc_type(nanozonev2_t *nanozone, size_t size,
+		malloc_type_id_t type_id)
+{
+	// Just hand to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_malloc(helper_zone, size, type_id);
 }
 #endif // OS_VARIANT_RESOLVED
 
@@ -2681,11 +3345,38 @@
 			size);
 }
 
+static void *
+nanov2_forked_calloc_type(nanozonev2_t *nanozone, size_t num_items, size_t size,
+		malloc_type_id_t type_id)
+{
+	// Just hand to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_calloc(helper_zone, num_items, size, type_id);
+}
+
+static void *
+nanov2_forked_memalign(nanozonev2_t *nanozone, size_t alignment, size_t size)
+{
+	// Just hand to the helper zone.
+	return nanozone->helper_zone->memalign(nanozone->helper_zone, alignment,
+			size);
+}
+
+static void *
+nanov2_forked_memalign_type(nanozonev2_t *nanozone, size_t alignment,
+		size_t size, malloc_type_id_t type_id)
+{
+	// Just hand to the helper zone.
+	malloc_zone_t *helper_zone = nanozone->helper_zone;
+	return helper_zone->malloc_type_memalign(helper_zone, alignment, size,
+			type_id);
+}
+
 #endif // OS_VARIANT_NOTRESOLVED
 
 #if OS_VARIANT_RESOLVED
 
-void
+MALLOC_NOEXPORT void
 nanov2_forked_free(nanozonev2_t *nanozone, void *ptr)
 {
 	if (!ptr) {
@@ -2707,27 +3398,37 @@
 	/* NOTREACHED */
 }
 
-void
+MALLOC_NOEXPORT void
 nanov2_forked_free_definite_size(nanozonev2_t *nanozone, void *ptr, size_t size)
 {
 	nanov2_forked_free(nanozone, ptr);
 }
 
-void *
-nanov2_forked_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size)
+MALLOC_ALWAYS_INLINE MALLOC_INLINE
+static void *
+_nanov2_forked_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size,
+		bool typed, malloc_type_id_t type_id)
 {
 	// could occur through malloc_zone_realloc() path
 	if (!ptr) {
 		// If ptr is a null pointer, realloc() shall be equivalent to malloc()
 		// for the specified size.
-		return nanov2_forked_malloc(nanozone, new_size);
+		if (typed) {
+			return nanov2_forked_malloc_type(nanozone, new_size, type_id);
+		} else {
+			return nanov2_forked_malloc(nanozone, new_size);
+		}
 	}
 
 	size_t old_size = nanov2_pointer_size(nanozone, ptr, FALSE);
 	if (!old_size) {
 		// not-nano pointer, hand down to helper zone
 		malloc_zone_t *zone = (malloc_zone_t *)(nanozone->helper_zone);
-		return zone->realloc(zone, ptr, new_size);
+		if (typed) {
+			return zone->malloc_type_realloc(zone, ptr, new_size, type_id);
+		} else {
+			return zone->realloc(zone, ptr, new_size);
+		}
 	} else {
 		if (!new_size) {
 			// If size is 0 and ptr is not a null pointer, the object pointed to
@@ -2739,8 +3440,15 @@
 			return nanov2_forked_malloc(nanozone, 1);
 		}
 
-		void *new_ptr = nanozone->helper_zone->malloc(nanozone->helper_zone,
+		void *new_ptr;
+		if (typed) {
+			malloc_zone_t *helper_zone = nanozone->helper_zone;
+			new_ptr = helper_zone->malloc_type_malloc(helper_zone, new_size,
+					type_id);
+		} else {
+			new_ptr = nanozone->helper_zone->malloc(nanozone->helper_zone,
 				new_size);
+		}
 		if (new_ptr) {
 			size_t valid_size = MIN(old_size, new_size);
 			memcpy(new_ptr, ptr, valid_size);
@@ -2755,6 +3463,20 @@
 	}
 	/* NOTREACHED */
 }
+
+MALLOC_NOEXPORT void *
+nanov2_forked_realloc(nanozonev2_t *nanozone, void *ptr, size_t new_size)
+{
+	return _nanov2_forked_realloc(nanozone, ptr, new_size, false,
+			MALLOC_TYPE_ID_NONE);
+}
+
+MALLOC_NOEXPORT void *
+nanov2_forked_realloc_type(nanozonev2_t *nanozone, void *ptr, size_t new_size,
+		malloc_type_id_t type_id)
+{
+	return _nanov2_forked_realloc(nanozone, ptr, new_size, true, type_id);
+}
 #endif // OS_VARIANT_RESOLVED
 
 #if OS_VARIANT_NOTRESOLVED
@@ -2771,7 +3493,7 @@
 
 #if OS_VARIANT_RESOLVED
 
-void
+MALLOC_NOEXPORT void
 nanov2_forked_batch_free(nanozonev2_t *nanozone, void **to_be_freed,
 		unsigned count)
 {
@@ -2819,10 +3541,18 @@
 	nanozone->basic_zone.batch_free = OS_RESOLVED_VARIANT_ADDR(nanov2_forked_batch_free);
 	nanozone->basic_zone.introspect =
 			(struct malloc_introspection_t *)&nanov2_introspect;// Unchanged
-	nanozone->basic_zone.memalign = (void *)nanov2_memalign; 	// Unchanged
+	nanozone->basic_zone.memalign = (void *)nanov2_forked_memalign;
 	nanozone->basic_zone.free_definite_size =
 			OS_RESOLVED_VARIANT_ADDR(nanov2_forked_free_definite_size);
 	nanozone->basic_zone.claimed_address = nanov2_forked_claimed_address;
+	nanozone->basic_zone.try_free_default = NULL; // Fall back to old protocol
+	nanozone->basic_zone.malloc_type_malloc =
+			OS_RESOLVED_VARIANT_ADDR(nanov2_forked_malloc_type);
+	nanozone->basic_zone.malloc_type_calloc = (void *)nanov2_forked_calloc_type;
+	nanozone->basic_zone.malloc_type_realloc =
+			OS_RESOLVED_VARIANT_ADDR(nanov2_forked_realloc_type);
+	nanozone->basic_zone.malloc_type_memalign =
+			(void *)nanov2_forked_memalign_type;
 	mprotect(nanozone, sizeof(nanozone->basic_zone), PROT_READ);
 }