/* * Copyright (c) 1999 Apple Computer, Inc. All rights reserved. * * @APPLE_LICENSE_HEADER_START@ * * The contents of this file constitute Original Code as defined in and * are subject to the Apple Public Source License Version 1.1 (the * "License"). You may not use this file except in compliance with the * License. Please obtain a copy of the License at * http://www.apple.com/publicsource and read it before using this file. * * This Original Code and all software distributed under the License are * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the * License for the specific language governing rights and limitations * under the License. * * @APPLE_LICENSE_HEADER_END@ */ /* Author: Bertrand Serlet, August 1999 */ #import "scalable_malloc.h" #define __POSIX_LIB__ #import #import // for spin lock #import #include /********************* DEFINITIONS ************************/ static unsigned vm_page_shift = 0; // guaranteed to be intialized by zone creation #define DEBUG_MALLOC 0 // set to one to debug malloc itself #define DEBUG_CLIENT 0 // set to one to help debug a nasty memory smasher #if DEBUG_MALLOC #warning DEBUG ENABLED #define INLINE #else #define INLINE inline #endif #define CHECK_REGIONS (1 << 31) #define VM_COPY_THRESHOLD (40 * 1024) // When all memory is touched after a copy, vm_copy() is always a lose // But if the memory is only read, vm_copy() wins over memmove() at 3 or 4 pages (on a G3/300MHz) #define KILL_THRESHOLD (32 * 1024) #define LARGE_THRESHOLD (3 * vm_page_size) // at or above this use "large" #define SHIFT_QUANTUM 4 // Required for AltiVec #define QUANTUM (1 << SHIFT_QUANTUM) // allocation quantum #define MIN_BLOCK 1 // minimum size, in QUANTUM multiples /* The header of a small block in use contains its size (expressed as multiples of QUANTUM, and header included), or 0; If 0 then the block is either free (in which case the size is directly at the block itself), or the last block (indicated by either being beyond range, or having 0 in the block itself) */ #define PTR_HEADER_SIZE (sizeof(msize_t)) #define FOLLOWING_PTR(ptr,msize) (((char *)(ptr)) + ((msize) << SHIFT_QUANTUM)) #define PREVIOUS_MSIZE(ptr) ((msize_t *)(ptr))[-2] #define THIS_FREE 0x8000 // indicates this block is free #define PREV_FREE 0x4000 // indicates previous block is free #define MSIZE_FLAGS_FOR_PTR(ptr) (((msize_t *)(ptr))[-1]) #define REGION_SIZE (1 << (16 - 2 + SHIFT_QUANTUM)) // since we only have 16 bits for msize_t, and 1 bit is taken by THIS_FREE, and 1 by PREV_FREE #define INITIAL_NUM_REGIONS 8 // must always be at least 2 to always have 1 slot empty #define CHECKSUM_MAGIC 0x357B #define PROTECT_SMALL 0 // Should be 0: 1 is too slow for normal use #define LARGE_CACHE_SIZE 1 // define hysterisis of large chunks #define MAX_LARGE_SIZE_TO_CACHE (128*1024) /* blocks larger than this are not cached */ #define MAX_RECORDER_BUFFER 256 #define MAX_GRAIN 64 typedef unsigned short msize_t; // a size in multiples of SHIFT_QUANTUM typedef struct { unsigned checksum; void *previous; void *next; } free_list_t; typedef struct { unsigned address_and_num_pages; // this type represents both an address and a number of pages // the low bits are the number of pages // the high bits are the address // note that the exact number of bits used for depends on the page size // also, this cannot represent pointers larger than 1 << (vm_page_shift * 2) } compact_range_t; typedef vm_address_t region_t; typedef compact_range_t large_entry_t; typedef vm_range_t huge_entry_t; typedef unsigned short grain_t; typedef struct { malloc_zone_t basic_zone; pthread_lock_t lock; unsigned debug_flags; void *log_address; /* Regions for small objects */ unsigned num_regions; region_t *regions; // this array is always created with 1 extra slot to be able to add a region without taking memory right away unsigned last_region_hit; free_list_t *free_list[MAX_GRAIN]; unsigned num_bytes_free_in_last_region; // these bytes are cleared unsigned num_small_objects; unsigned num_bytes_in_small_objects; /* Cache of recently freed large object */ vm_range_t large_to_deallocate[LARGE_CACHE_SIZE]; // Entries that are 0 should be discarded /* large objects: vm_page_shift <= log2(size) < 2 *vm_page_shift */ unsigned num_large_objects_in_use; // does not count the large entries in large_to_deallocate unsigned num_large_entries; unsigned num_bytes_in_large_objects; large_entry_t *large_entries; // large_entries are hashed by location // large_entries that are 0 should be discarded /* huge objects: log2(size) >= 2 *vm_page_shift */ unsigned num_bytes_in_huge_objects; unsigned num_huge_entries; huge_entry_t *huge_entries; } szone_t; static void *szone_malloc(szone_t *szone, size_t size); static void *szone_valloc(szone_t *szone, size_t size); static INLINE void *szone_malloc_should_clear(szone_t *szone, size_t size, boolean_t cleared_requested); static void szone_free(szone_t *szone, void *ptr); static size_t szone_good_size(szone_t *szone, size_t size); static boolean_t szone_check_all(szone_t *szone, const char *function); static void szone_print(szone_t *szone, boolean_t verbose); static INLINE region_t *region_for_ptr_no_lock(szone_t *szone, const void *ptr); static vm_range_t large_free_no_lock(szone_t *szone, large_entry_t *entry); #define LOG(szone,ptr) (szone->log_address && (szone->num_small_objects > 8) && (((unsigned)szone->log_address == -1) || (szone->log_address == (void *)(ptr)))) /********************* ACCESSOR MACROS ************************/ #define SZONE_LOCK(szone) LOCK(szone->lock) #define SZONE_UNLOCK(szone) UNLOCK(szone->lock) #define CHECK(szone,fun) if ((szone)->debug_flags & CHECK_REGIONS) szone_check_all(szone, fun) #define REGION_ADDRESS(region) (region) #define REGION_END(region) (region+REGION_SIZE) #define LARGE_ENTRY_ADDRESS(entry) (((entry).address_and_num_pages >> vm_page_shift) << vm_page_shift) #define LARGE_ENTRY_NUM_PAGES(entry) ((entry).address_and_num_pages & ((1 << vm_page_shift) - 1)) #define LARGE_ENTRY_SIZE(entry) (LARGE_ENTRY_NUM_PAGES(entry) << vm_page_shift) #define LARGE_ENTRY_MATCHES(entry,ptr) (!(((entry).address_and_num_pages - (unsigned)(ptr)) >> vm_page_shift)) #define LARGE_ENTRY_IS_EMPTY(entry) (!((entry).address_and_num_pages)) /********************* VERY LOW LEVEL UTILITIES ************************/ static void szone_error(szone_t *szone, const char *msg, const void *ptr) { if (szone) SZONE_UNLOCK(szone); if (ptr) { malloc_printf("*** malloc[%d]: error for object %p: %s\n", getpid(), ptr, msg); #if DEBUG_MALLOC szone_print(szone, 1); #endif } else { malloc_printf("*** malloc[%d]: error: %s\n", getpid(), msg); } #if DEBUG_CLIENT malloc_printf("*** Sleeping to help debug\n"); sleep(3600); // to help debug #endif } static void protect(szone_t *szone, vm_address_t address, vm_size_t size, unsigned protection, unsigned debug_flags) { kern_return_t err; if (!(debug_flags & SCALABLE_MALLOC_DONT_PROTECT_PRELUDE)) { err = vm_protect(mach_task_self(), address - vm_page_size, vm_page_size, 0, protection); if (err) malloc_printf("*** malloc[%d]: Can't protect(%x) region for prelude guard page at 0x%x\n", getpid(), protection, address - vm_page_size); } if (!(debug_flags & SCALABLE_MALLOC_DONT_PROTECT_POSTLUDE)) { err = vm_protect(mach_task_self(), (vm_address_t)(address + size), vm_page_size, 0, protection); if (err) malloc_printf("*** malloc[%d]: Can't protect(%x) region for postlude guard page at 0x%x\n", getpid(), protection, address + size); } } static vm_address_t allocate_pages(szone_t *szone, size_t size, unsigned debug_flags, int vm_page_label) { kern_return_t err; vm_address_t addr; boolean_t add_guard_pages = debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES; size_t allocation_size = round_page(size); if (!allocation_size) allocation_size = vm_page_size; if (add_guard_pages) allocation_size += 2 * vm_page_size; err = vm_allocate(mach_task_self(), &addr, allocation_size, vm_page_label | 1); if (err) { szone_error(szone, "Can't allocate region", NULL); return NULL; } if (add_guard_pages) { addr += vm_page_size; protect(szone, addr, size, 0, debug_flags); } return addr; } static void deallocate_pages(szone_t *szone, vm_address_t addr, size_t size, unsigned debug_flags) { kern_return_t err; boolean_t add_guard_pages = debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES; if (add_guard_pages) { addr -= vm_page_size; size += 2 * vm_page_size; } err = vm_deallocate(mach_task_self(), addr, size); if (err) { szone_error(szone, "Can't deallocate_pages region", (void *)addr); } } static kern_return_t _szone_default_reader(task_t task, vm_address_t address, vm_size_t size, void **ptr) { *ptr = (void *)address; return 0; } /********************* RANGE UTILITIES ************************/ static const vm_range_t zero_range = {0, 0}; static vm_range_t coalesce_range(vm_range_t *ranges, unsigned count, vm_range_t range) { // Given a sequence of ranges and a range, tries to find an abutting range // If no, returns original range // Else zeroes out coalesced range, and reapplies with coalesced range unsigned index = count; vm_range_t *current = ranges; while (index--) { vm_range_t this = *current++; if (!this.size) continue; if (this.address + this.size == range.address) { range.address = this.address; range.size += this.size; current[-1] = zero_range; return coalesce_range(ranges, count, range); } if (range.address + range.size == this.address) { range.size += this.size; current[-1] = zero_range; return coalesce_range(ranges, count, range); } } return range; } static INLINE vm_range_t *first_zero_range(vm_range_t *ranges, unsigned count) { // Given a sequence of ranges, find the first empty slot // or returns NULL while (count--) { if (!ranges->size) return ranges; ranges++; } return NULL; } static vm_range_t *largest_range(vm_range_t *ranges, unsigned count) { // Given a sequence of ranges, find the largest range // Returns NULL on empty arrays vm_range_t *largest_range; if (!count) return NULL; largest_range = ranges; count--; ranges++; while (count--) { if (ranges->size > largest_range->size) largest_range = ranges; ranges++; } return largest_range; } static vm_range_t *first_range_greater_or_equal(vm_range_t *ranges, unsigned count, vm_size_t size) { // Given a sequence of ranges, find the first range greater than range // Returns NULL when none found while (count--) { if (ranges->size >= size) return ranges; ranges++; } return NULL; } /********************* FREE LIST UTILITIES ************************/ static INLINE grain_t grain_for_msize(szone_t *szone, msize_t msize) { // assumes msize >= MIN_BLOCK #if DEBUG_MALLOC if (msize < MIN_BLOCK) { szone_error(szone, "grain_for_msize: msize too small", NULL); } #endif return (msize < MAX_GRAIN + MIN_BLOCK) ? msize - MIN_BLOCK : MAX_GRAIN - 1; } static INLINE msize_t msize_for_grain(szone_t *szone, grain_t grain) { // 0 if multiple sizes return (grain < MAX_GRAIN - 1) ? grain + MIN_BLOCK : 0; } static INLINE void free_list_checksum(szone_t *szone, free_list_t *ptr) { // We always checksum, as testing whether to do it (based on szone->debug_flags) is as fast as doing it if (ptr->checksum != (((unsigned)ptr->previous) ^ ((unsigned)ptr->next) ^ CHECKSUM_MAGIC)) { szone_error(szone, "Incorrect check sum for freed object - object was probably modified after beeing freed; break at szone_error", ptr); } } static INLINE void free_list_set_checksum(szone_t *szone, free_list_t *ptr) { // We always set checksum, as testing whether to do it (based on szone->debug_flags) is slower than just doing it ptr->checksum = ((unsigned)ptr->previous) ^ ((unsigned)ptr->next) ^ CHECKSUM_MAGIC; } static void free_list_add_ptr(szone_t *szone, void *ptr, msize_t msize) { // Adds an item to the proper free list // Also marks the header of the block properly grain_t grain = grain_for_msize(szone, msize); free_list_t *free_ptr = ptr; free_list_t *free_head = szone->free_list[grain]; msize_t *follower = (msize_t *)FOLLOWING_PTR(ptr, msize); #if DEBUG_MALLOC if (LOG(szone,ptr)) malloc_printf("In free_list_add_ptr(), ptr=%p, msize=%d\n", ptr, msize); if (((unsigned)ptr) & (QUANTUM - 1)) { szone_error(szone, "free_list_add_ptr: Unaligned ptr", ptr); } #endif MSIZE_FLAGS_FOR_PTR(ptr) = msize | THIS_FREE; if (free_head) { free_list_checksum(szone, free_head); #if DEBUG_MALLOC if (free_head->previous) { malloc_printf("ptr=%p grain=%d free_head=%p previous=%p\n", ptr, grain, free_head, free_head->previous); szone_error(szone, "free_list_add_ptr: Internal invariant broken (free_head->previous)", ptr); } if (!(MSIZE_FLAGS_FOR_PTR(free_head) & THIS_FREE)) { malloc_printf("ptr=%p grain=%d free_head=%p\n", ptr, grain, free_head); szone_error(szone, "free_list_add_ptr: Internal invariant broken (free_head is not a free pointer)", ptr); } if ((grain != MAX_GRAIN-1) && (MSIZE_FLAGS_FOR_PTR(free_head) != (THIS_FREE | msize))) { malloc_printf("ptr=%p grain=%d free_head=%p previous_msize=%d\n", ptr, grain, free_head, MSIZE_FLAGS_FOR_PTR(free_head)); szone_error(szone, "free_list_add_ptr: Internal invariant broken (incorrect msize)", ptr); } #endif free_head->previous = free_ptr; free_list_set_checksum(szone, free_head); } free_ptr->previous = NULL; free_ptr->next = free_head; free_list_set_checksum(szone, free_ptr); szone->free_list[grain] = free_ptr; // mark the end of this block PREVIOUS_MSIZE(follower) = msize; MSIZE_FLAGS_FOR_PTR(follower) |= PREV_FREE; } static void free_list_remove_ptr(szone_t *szone, void *ptr, msize_t msize) { // Removes item in the proper free list // msize could be read, but all callers have it so we pass it in grain_t grain = grain_for_msize(szone, msize); free_list_t *free_ptr = ptr; free_list_t *next = free_ptr->next; free_list_t *previous = free_ptr->previous; #if DEBUG_MALLOC if (LOG(szone,ptr)) malloc_printf("In free_list_remove_ptr(), ptr=%p, msize=%d\n", ptr, msize); #endif free_list_checksum(szone, free_ptr); if (!previous) { #if DEBUG_MALLOC if (szone->free_list[grain] != ptr) { malloc_printf("ptr=%p grain=%d msize=%d szone->free_list[grain]=%p\n", ptr, grain, msize, szone->free_list[grain]); szone_error(szone, "free_list_remove_ptr: Internal invariant broken (szone->free_list[grain])", ptr); return; } #endif szone->free_list[grain] = next; } else { previous->next = next; free_list_set_checksum(szone, previous); } if (next) { next->previous = previous; free_list_set_checksum(szone, next); } MSIZE_FLAGS_FOR_PTR(FOLLOWING_PTR(ptr, msize)) &= ~ PREV_FREE; } static boolean_t free_list_check(szone_t *szone, grain_t grain) { unsigned count = 0; free_list_t *ptr = szone->free_list[grain]; free_list_t *previous = NULL; while (ptr) { msize_t msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); count++; if (!(msize_and_free & THIS_FREE)) { malloc_printf("*** malloc[%d]: In-use ptr in free list grain=%d count=%d ptr=%p\n", getpid(), grain, count, ptr); return 0; } if (((unsigned)ptr) & (QUANTUM - 1)) { malloc_printf("*** malloc[%d]: Unaligned ptr in free list grain=%d count=%d ptr=%p\n", getpid(), grain, count, ptr); return 0; } if (!region_for_ptr_no_lock(szone, ptr)) { malloc_printf("*** malloc[%d]: Ptr not in szone grain=%d count=%d ptr=%p\n", getpid(), grain, count, ptr); return 0; } free_list_checksum(szone, ptr); if (ptr->previous != previous) { malloc_printf("*** malloc[%d]: Previous incorrectly set grain=%d count=%d ptr=%p\n", getpid(), grain, count, ptr); return 0; } if ((grain != MAX_GRAIN-1) && (msize_and_free != (msize_for_grain(szone, grain) | THIS_FREE))) { malloc_printf("*** malloc[%d]: Incorrect msize for grain=%d count=%d ptr=%p msize=%d\n", getpid(), grain, count, ptr, msize_and_free); return 0; } previous = ptr; ptr = ptr->next; } return 1; } /********************* SMALL BLOCKS MANAGEMENT ************************/ static INLINE region_t *region_for_ptr_no_lock(szone_t *szone, const void *ptr) { region_t *first_region = szone->regions; region_t *region = first_region + szone->last_region_hit; region_t this = *region; if ((unsigned)ptr - (unsigned)REGION_ADDRESS(this) < (unsigned)REGION_SIZE) { return region; } else { // We iterate in reverse order becase last regions are more likely region = first_region + szone->num_regions; while (region != first_region) { this = *(--region); if ((unsigned)ptr - (unsigned)REGION_ADDRESS(this) < (unsigned)REGION_SIZE) { szone->last_region_hit = region - first_region; return region; } } return NULL; } } static INLINE void small_free_no_lock(szone_t *szone, region_t *region, void *ptr, msize_t msize_and_free) { msize_t msize = msize_and_free & ~ PREV_FREE; size_t original_size = msize << SHIFT_QUANTUM; void *next_block = ((char *)ptr + original_size); msize_t next_msize_and_free; #if DEBUG_MALLOC if (LOG(szone,ptr)) malloc_printf("In small_free_no_lock(), ptr=%p, msize=%d\n", ptr, msize); if (msize < MIN_BLOCK) { malloc_printf("In small_free_no_lock(), ptr=%p, msize=%d\n", ptr, msize); szone_error(szone, "Trying to free small block that is too small", ptr); } #endif if (((vm_address_t)next_block < REGION_END(*region)) && ((next_msize_and_free = MSIZE_FLAGS_FOR_PTR(next_block)) & THIS_FREE)) { // If the next block is free, we coalesce msize_t next_msize = next_msize_and_free & ~THIS_FREE; if (LOG(szone,ptr)) malloc_printf("In small_free_no_lock(), for ptr=%p, msize=%d coalesced next block=%p next_msize=%d\n", ptr, msize, next_block, next_msize); free_list_remove_ptr(szone, next_block, next_msize); msize += next_msize; } // Let's try to coalesce backwards now if (msize_and_free & PREV_FREE) { msize_t previous_msize = PREVIOUS_MSIZE(ptr); void *previous = ptr - (previous_msize << SHIFT_QUANTUM); #if DEBUG_MALLOC if (LOG(szone,previous)) malloc_printf("In small_free_no_lock(), coalesced backwards for %p previous=%p, msize=%d\n", ptr, previous, previous_msize); if (!previous_msize || (previous_msize >= (((vm_address_t)ptr - REGION_ADDRESS(*region)) >> SHIFT_QUANTUM))) { szone_error(szone, "Invariant 1 broken when coalescing backwards", ptr); } if (MSIZE_FLAGS_FOR_PTR(previous) != (previous_msize | THIS_FREE)) { malloc_printf("previous=%p its_msize_and_free=0x%x previous_msize=%d\n", previous, MSIZE_FLAGS_FOR_PTR(previous), previous_msize); szone_error(szone, "Invariant 3 broken when coalescing backwards", ptr); } #endif free_list_remove_ptr(szone, previous, previous_msize); ptr = previous; msize += previous_msize; #if DEBUG_MALLOC if (msize & PREV_FREE) { malloc_printf("In small_free_no_lock(), after coalescing with previous ptr=%p, msize=%d previous_msize=%d\n", ptr, msize, previous_msize); szone_error(szone, "Incorrect coalescing", ptr); } #endif } if (szone->debug_flags & SCALABLE_MALLOC_DO_SCRIBBLE) { if (!msize) { szone_error(szone, "Incorrect size information - block header was damaged", ptr); } else { memset(ptr, 0x55, (msize << SHIFT_QUANTUM) - PTR_HEADER_SIZE); } } free_list_add_ptr(szone, ptr, msize); CHECK(szone, "small_free_no_lock: added to free list"); szone->num_small_objects--; szone->num_bytes_in_small_objects -= original_size; // we use original_size and not msize to avoid double counting the coalesced blocks } static void *small_malloc_from_region_no_lock(szone_t *szone, msize_t msize) { // Allocates from the last region or a freshly allocated region region_t *last_region = szone->regions + szone->num_regions - 1; vm_address_t new_address; void *ptr; msize_t msize_and_free; unsigned region_capacity; ptr = (void *)(REGION_END(*last_region) - szone->num_bytes_free_in_last_region + PTR_HEADER_SIZE); #if DEBUG_MALLOC if (((vm_address_t)ptr) & (QUANTUM - 1)) { szone_error(szone, "Invariant broken while using end of region", ptr); } #endif msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); #if DEBUG_MALLOC if (msize_and_free != PREV_FREE && msize_and_free != 0) { malloc_printf("*** malloc[%d]: msize_and_free = %d\n", getpid(), msize_and_free); szone_error(szone, "Invariant broken when allocating at end of zone", ptr); } #endif // In order to make sure we don't have 2 free pointers following themselves, if the last item is a free item, we combine it and clear it if (msize_and_free == PREV_FREE) { msize_t previous_msize = PREVIOUS_MSIZE(ptr); void *previous = ptr - (previous_msize << SHIFT_QUANTUM); #if DEBUG_MALLOC if (LOG(szone, ptr)) malloc_printf("Combining last with free space at %p\n", ptr); if (!previous_msize || (previous_msize >= (((vm_address_t)ptr - REGION_ADDRESS(*last_region)) >> SHIFT_QUANTUM)) || (MSIZE_FLAGS_FOR_PTR(previous) != (previous_msize | THIS_FREE))) { szone_error(szone, "Invariant broken when coalescing backwards at end of zone", ptr); } #endif free_list_remove_ptr(szone, previous, previous_msize); szone->num_bytes_free_in_last_region += previous_msize << SHIFT_QUANTUM; memset(previous, 0, previous_msize << SHIFT_QUANTUM); MSIZE_FLAGS_FOR_PTR(previous) = 0; ptr = previous; } // first try at the end of the last region CHECK(szone, __PRETTY_FUNCTION__); if (szone->num_bytes_free_in_last_region >= (msize << SHIFT_QUANTUM)) { szone->num_bytes_free_in_last_region -= (msize << SHIFT_QUANTUM); szone->num_small_objects++; szone->num_bytes_in_small_objects += msize << SHIFT_QUANTUM; MSIZE_FLAGS_FOR_PTR(ptr) = msize; return ptr; } // time to create a new region new_address = allocate_pages(szone, REGION_SIZE, 0, VM_MAKE_TAG(VM_MEMORY_MALLOC_SMALL)); if (!new_address) { // out of memory! return NULL; } // let's prepare to free the remnants of last_region if (szone->num_bytes_free_in_last_region >= QUANTUM) { msize_t this_msize = szone->num_bytes_free_in_last_region >> SHIFT_QUANTUM; // malloc_printf("Entering last block %p size=%d\n", ptr, this_msize << SHIFT_QUANTUM); if (this_msize >= MIN_BLOCK) { free_list_add_ptr(szone, ptr, this_msize); } else { // malloc_printf("Leaking last block at %p\n", ptr); } szone->num_bytes_free_in_last_region -= this_msize << SHIFT_QUANTUM; // to avoid coming back here } last_region[1] = new_address; szone->num_regions++; szone->num_bytes_free_in_last_region = REGION_SIZE - QUANTUM + PTR_HEADER_SIZE - (msize << SHIFT_QUANTUM); ptr = (void *)(new_address + QUANTUM); // waste the first bytes region_capacity = (MSIZE_FLAGS_FOR_PTR(szone->regions) * QUANTUM - PTR_HEADER_SIZE) / sizeof(region_t); if (szone->num_regions >= region_capacity) { unsigned new_capacity = region_capacity * 2 + 1; msize_t new_msize = (new_capacity * sizeof(region_t) + PTR_HEADER_SIZE + QUANTUM - 1) / QUANTUM; region_t *new_regions = ptr; // malloc_printf("Now %d regions growing regions %p to %d\n", szone->num_regions, szone->regions, new_capacity); MSIZE_FLAGS_FOR_PTR(new_regions) = new_msize; szone->num_small_objects++; szone->num_bytes_in_small_objects += new_msize << SHIFT_QUANTUM; memcpy(new_regions, szone->regions, szone->num_regions * sizeof(region_t)); // We intentionally leak the previous regions pointer to avoid multi-threading crashes if another thread was reading it (unlocked) while we are changing it // Given that in practise the number of regions is typically a handful, this should not be a big deal szone->regions = new_regions; ptr += (new_msize << SHIFT_QUANTUM); szone->num_bytes_free_in_last_region -= (new_msize << SHIFT_QUANTUM); // malloc_printf("Regions is now %p next ptr is %p\n", szone->regions, ptr); } szone->num_small_objects++; szone->num_bytes_in_small_objects += msize << SHIFT_QUANTUM; MSIZE_FLAGS_FOR_PTR(ptr) = msize; return ptr; } static boolean_t szone_check_region(szone_t *szone, region_t *region) { void *ptr = (void *)REGION_ADDRESS(*region) + QUANTUM; vm_address_t region_end = REGION_END(*region); int is_last_region = region == szone->regions + szone->num_regions - 1; msize_t prev_free = 0; while ((vm_address_t)ptr < region_end) { msize_t msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); if (!(msize_and_free & THIS_FREE)) { msize_t msize = msize_and_free & ~PREV_FREE; if ((msize_and_free & PREV_FREE) != prev_free) { malloc_printf("*** malloc[%d]: invariant broken for %p (prev_free=%d) this msize=%d\n", getpid(), ptr, prev_free, msize_and_free); return 0; } if (!msize) { int extra = (is_last_region) ? szone->num_bytes_free_in_last_region : QUANTUM; if (((unsigned)(ptr + extra)) < region_end) { malloc_printf("*** malloc[%d]: invariant broken at region end: ptr=%p extra=%d index=%d num_regions=%d end=%p\n", getpid(), ptr, extra, region - szone->regions, szone->num_regions, (void *)region_end); return 0; } break; // last encountered } if (msize > (LARGE_THRESHOLD / QUANTUM)) { malloc_printf("*** malloc[%d]: invariant broken for %p this msize=%d - size is too large\n", getpid(), ptr, msize_and_free); return 0; } if ((msize < MIN_BLOCK) && ((unsigned)ptr != region_end - QUANTUM)) { malloc_printf("*** malloc[%d]: invariant broken for %p this msize=%d - size is too small\n", getpid(), ptr, msize_and_free); return 0; } ptr += msize << SHIFT_QUANTUM; prev_free = 0; if (is_last_region && ((vm_address_t)ptr - PTR_HEADER_SIZE > region_end - szone->num_bytes_free_in_last_region)) { malloc_printf("*** malloc[%d]: invariant broken for %p this msize=%d - block extends beyond allocated region\n", getpid(), ptr, msize_and_free); } } else { // free pointer msize_t msize = msize_and_free & ~THIS_FREE; free_list_t *free_head = ptr; msize_t *follower = (void *)FOLLOWING_PTR(ptr, msize); if ((msize_and_free & PREV_FREE) && !prev_free) { malloc_printf("*** malloc[%d]: invariant broken for free block %p this msize=%d: PREV_FREE set while previous block is in use\n", getpid(), ptr, msize); return 0; } if (msize < MIN_BLOCK) { malloc_printf("*** malloc[%d]: invariant broken for free block %p this msize=%d\n", getpid(), ptr, msize); return 0; } if (prev_free) { malloc_printf("*** malloc[%d]: invariant broken for %p (2 free in a row)\n", getpid(), ptr); return 0; } free_list_checksum(szone, free_head); if (free_head->previous && !(MSIZE_FLAGS_FOR_PTR(free_head->previous) & THIS_FREE)) { malloc_printf("*** malloc[%d]: invariant broken for %p (previous %p is not a free pointer)\n", getpid(), ptr, free_head->previous); return 0; } if (free_head->next && !(MSIZE_FLAGS_FOR_PTR(free_head->next) & THIS_FREE)) { malloc_printf("*** malloc[%d]: invariant broken for %p (next is not a free pointer)\n", getpid(), ptr); return 0; } if (PREVIOUS_MSIZE(follower) != msize) { malloc_printf("*** malloc[%d]: invariant broken for free %p followed by %p in region [%x-%x] (end marker incorrect) should be %d; in fact %d\n", getpid(), ptr, follower, REGION_ADDRESS(*region), region_end, msize, PREVIOUS_MSIZE(follower)); return 0; } ptr = follower; prev_free = PREV_FREE; } } return 1; } static kern_return_t small_in_use_enumerator(task_t task, void *context, unsigned type_mask, vm_address_t region_address, unsigned num_regions, memory_reader_t reader, vm_range_recorder_t recorder) { region_t *regions; unsigned index = 0; vm_range_t buffer[MAX_RECORDER_BUFFER]; unsigned count = 0; kern_return_t err; err = reader(task, region_address, sizeof(region_t) * num_regions, (void **)®ions); if (err) return err; while (index < num_regions) { region_t region = regions[index++]; vm_range_t range = {REGION_ADDRESS(region), REGION_SIZE}; vm_address_t start = range.address + QUANTUM; // malloc_printf("Enumerating small ptrs for Region starting at 0x%x\n", start); if (type_mask & MALLOC_PTR_REGION_RANGE_TYPE) recorder(task, context, MALLOC_PTR_REGION_RANGE_TYPE, &range, 1); if (type_mask & MALLOC_PTR_IN_USE_RANGE_TYPE) while (start < range.address + range.size) { void *previous; msize_t msize_and_free; err = reader(task, start - PTR_HEADER_SIZE, QUANTUM, (void **)&previous); if (err) return err; previous += PTR_HEADER_SIZE; msize_and_free = MSIZE_FLAGS_FOR_PTR(previous); if (!(msize_and_free & THIS_FREE)) { // Block in use msize_t msize = msize_and_free & ~PREV_FREE; if (!msize) break; // last encountered buffer[count].address = start; buffer[count].size = (msize << SHIFT_QUANTUM) - PTR_HEADER_SIZE; count++; if (count >= MAX_RECORDER_BUFFER) { recorder(task, context, MALLOC_PTR_IN_USE_RANGE_TYPE, buffer, count); count = 0; } start += msize << SHIFT_QUANTUM; } else { // free pointer msize_t msize = msize_and_free & ~THIS_FREE; start += msize << SHIFT_QUANTUM; } } // malloc_printf("End region - count=%d\n", count); } if (count) recorder(task, context, MALLOC_PTR_IN_USE_RANGE_TYPE, buffer, count); return 0; } static INLINE void *small_malloc_from_free_list(szone_t *szone, msize_t msize, boolean_t *locked) { void *ptr; msize_t this_msize; free_list_t **free_list; free_list_t **limit = szone->free_list + MAX_GRAIN - 1; // first try the small grains free_list = szone->free_list + grain_for_msize(szone, msize); while (free_list < limit) { // try bigger grains ptr = *free_list; if (ptr) { if (!*locked) { *locked = 1; SZONE_LOCK(szone); CHECK(szone, __PRETTY_FUNCTION__); } ptr = *free_list; if (ptr) { // optimistic test worked free_list_t *next; next = ((free_list_t *)ptr)->next; if (next) { next->previous = NULL; free_list_set_checksum(szone, next); } *free_list = next; this_msize = MSIZE_FLAGS_FOR_PTR(ptr) & ~THIS_FREE; MSIZE_FLAGS_FOR_PTR(FOLLOWING_PTR(ptr, this_msize)) &= ~ PREV_FREE; goto add_leftover_and_proceed; } } free_list++; } // We now check the large grains for one that is big enough if (!*locked) { *locked = 1; SZONE_LOCK(szone); CHECK(szone, __PRETTY_FUNCTION__); } ptr = *free_list; while (ptr) { this_msize = MSIZE_FLAGS_FOR_PTR(ptr) & ~THIS_FREE; if (this_msize >= msize) { free_list_remove_ptr(szone, ptr, this_msize); goto add_leftover_and_proceed; } ptr = ((free_list_t *)ptr)->next; } return NULL; add_leftover_and_proceed: if (this_msize >= msize + MIN_BLOCK) { if (LOG(szone,ptr)) malloc_printf("In small_malloc_should_clear(), adding leftover ptr=%p, this_msize=%d\n", ptr, this_msize); free_list_add_ptr(szone, ptr + (msize << SHIFT_QUANTUM), this_msize - msize); this_msize = msize; } szone->num_small_objects++; szone->num_bytes_in_small_objects += this_msize << SHIFT_QUANTUM; #if DEBUG_MALLOC if (LOG(szone,ptr)) malloc_printf("In small_malloc_should_clear(), ptr=%p, this_msize=%d, msize=%d\n", ptr, this_msize, msize); #endif MSIZE_FLAGS_FOR_PTR(ptr) = this_msize; return ptr; } static INLINE void *small_malloc_should_clear(szone_t *szone, msize_t msize, boolean_t cleared_requested) { boolean_t locked = 0; void *ptr; #if DEBUG_MALLOC if (! (msize & 0xffff)) { szone_error(szone, "Invariant broken (!msize) in allocation (region)", NULL); } if (msize < MIN_BLOCK) { szone_error(szone, "Invariant broken (msize too small) in allocation (region)", NULL); } #endif ptr = small_malloc_from_free_list(szone, msize, &locked); if (ptr) { CHECK(szone, __PRETTY_FUNCTION__); SZONE_UNLOCK(szone); if (cleared_requested) memset(ptr, 0, (msize << SHIFT_QUANTUM) - PTR_HEADER_SIZE); return ptr; } else { if (!locked) SZONE_LOCK(szone); CHECK(szone, __PRETTY_FUNCTION__); ptr = small_malloc_from_region_no_lock(szone, msize); // we don't clear because this freshly allocated space is pristine CHECK(szone, __PRETTY_FUNCTION__); SZONE_UNLOCK(szone); } return ptr; } static INLINE void *small_malloc_cleared_no_lock(szone_t *szone, msize_t msize) { // tries to allocate a small, cleared block boolean_t locked = 1; void *ptr; ptr = small_malloc_from_free_list(szone, msize, &locked); if (ptr) { memset(ptr, 0, (msize << SHIFT_QUANTUM) - PTR_HEADER_SIZE); return ptr; } else { ptr = small_malloc_from_region_no_lock(szone, msize); // we don't clear because this freshly allocated space is pristine } return ptr; } /********************* LARGE ENTRY UTILITIES ************************/ #if DEBUG_MALLOC static void large_cache_debug_print(szone_t *szone) { unsigned index = LARGE_CACHE_SIZE; malloc_printf("Cache to be dealloced: "); while (index--) { vm_range_t range = szone->large_to_deallocate[index]; if (range.size) malloc_printf("0x%x(%dKB) ", range.address, range.size/1024); } malloc_printf("\n"); } static void large_debug_print(szone_t *szone) { unsigned num_large_entries = szone->num_large_entries; unsigned index = num_large_entries; while (index--) { large_entry_t *range = szone->large_entries + index; large_entry_t entry = *range; if (!LARGE_ENTRY_IS_EMPTY(entry)) malloc_printf("%d: 0x%x(%dKB); ", index, LARGE_ENTRY_ADDRESS(entry), LARGE_ENTRY_SIZE(entry)/1024); } malloc_printf("\n"); } #endif static large_entry_t *large_entry_for_pointer_no_lock(szone_t *szone, const void *ptr) { // result only valid during a lock unsigned num_large_entries = szone->num_large_entries; unsigned hash_index; unsigned index; if (!num_large_entries) return NULL; hash_index = ((unsigned)ptr >> vm_page_shift) % num_large_entries; index = hash_index; do { large_entry_t *range = szone->large_entries + index; large_entry_t entry = *range; if (LARGE_ENTRY_MATCHES(entry, ptr)) return range; if (LARGE_ENTRY_IS_EMPTY(entry)) return NULL; // end of chain index++; if (index == num_large_entries) index = 0; } while (index != hash_index); return NULL; } static void large_entry_insert_no_lock(szone_t *szone, large_entry_t range) { unsigned num_large_entries = szone->num_large_entries; unsigned hash_index = (range.address_and_num_pages >> vm_page_shift) % num_large_entries; unsigned index = hash_index; // malloc_printf("Before insertion of 0x%x\n", LARGE_ENTRY_ADDRESS(range)); do { large_entry_t *entry = szone->large_entries + index; if (LARGE_ENTRY_IS_EMPTY(*entry)) { *entry = range; return; // end of chain } index++; if (index == num_large_entries) index = 0; } while (index != hash_index); } static INLINE void large_entries_rehash_after_entry_no_lock(szone_t *szone, large_entry_t *entry) { unsigned num_large_entries = szone->num_large_entries; unsigned hash_index = entry - szone->large_entries; unsigned index = hash_index; do { large_entry_t range; index++; if (index == num_large_entries) index = 0; range = szone->large_entries[index]; if (LARGE_ENTRY_IS_EMPTY(range)) return; szone->large_entries[index].address_and_num_pages = 0; large_entry_insert_no_lock(szone, range); // this will reinsert in the proper place } while (index != hash_index); } static INLINE large_entry_t *large_entries_alloc_no_lock(szone_t *szone, unsigned num) { size_t size = num * sizeof(large_entry_t); boolean_t is_vm_allocation = size >= LARGE_THRESHOLD; if (is_vm_allocation) { return (void *)allocate_pages(szone, round_page(size), 0, VM_MAKE_TAG(VM_MEMORY_MALLOC_LARGE)); } else { return small_malloc_cleared_no_lock(szone, (size + PTR_HEADER_SIZE + QUANTUM - 1) >> SHIFT_QUANTUM); } } static void large_entries_free_no_lock(szone_t *szone, large_entry_t *entries, unsigned num) { size_t size = num * sizeof(large_entry_t); boolean_t is_vm_allocation = size >= LARGE_THRESHOLD; if (is_vm_allocation) { deallocate_pages(szone, (vm_address_t)entries, round_page(size), 0); } else { region_t *region = region_for_ptr_no_lock(szone, entries); msize_t msize_and_free = MSIZE_FLAGS_FOR_PTR(entries); if (msize_and_free & THIS_FREE) { szone_error(szone, "Object already freed being freed", entries); return; } small_free_no_lock(szone, region, entries, msize_and_free); } } static void large_entries_grow_no_lock(szone_t *szone) { unsigned old_num_entries = szone->num_large_entries; large_entry_t *old_entries = szone->large_entries; unsigned new_num_entries = (old_num_entries) ? old_num_entries * 2 + 1 : 15; // always an odd number for good hashing large_entry_t *new_entries = large_entries_alloc_no_lock(szone, new_num_entries); unsigned index = old_num_entries; szone->num_large_entries = new_num_entries; szone->large_entries = new_entries; // malloc_printf("_grow_large_entries old_num_entries=%d new_num_entries=%d\n", old_num_entries, new_num_entries); while (index--) { large_entry_t oldRange = old_entries[index]; if (!LARGE_ENTRY_IS_EMPTY(oldRange)) large_entry_insert_no_lock(szone, oldRange); } if (old_entries) large_entries_free_no_lock(szone, old_entries, old_num_entries); } static vm_range_t large_free_no_lock(szone_t *szone, large_entry_t *entry) { // frees the specific entry in the size table // returns a range to truly deallocate, taking into account vm_range_t range; range.address = LARGE_ENTRY_ADDRESS(*entry); range.size = LARGE_ENTRY_SIZE(*entry); szone->num_large_objects_in_use --; szone->num_bytes_in_large_objects -= range.size; if (szone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES) { protect(szone, range.address, range.size, VM_PROT_READ | VM_PROT_WRITE, szone->debug_flags); range.address -= vm_page_size; range.size += 2 * vm_page_size; } // printf("Entry is 0x%x=%d; cache is 0x%x ; found=0x%x\n", entry, entry-szone->large_entries, szone->large_entries, large_entry_for_pointer_no_lock(szone, (void *)range.address)); entry->address_and_num_pages = 0; large_entries_rehash_after_entry_no_lock(szone, entry); #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, (void *)range.address)) { malloc_printf("*** malloc[%d]: Freed entry 0x%x still in use; num_large_entries=%d\n", getpid(), range.address, szone->num_large_entries); large_cache_debug_print(szone); large_debug_print(szone); sleep(3600); } #endif return range; } static vm_range_t large_find_better_range_to_deallocate(szone_t *szone, vm_range_t range) { // enters the specified large entry into the cache of freed entries // returns a range to truly deallocate vm_range_t *range_to_use; vm_range_t vm_range_to_deallocate; // if the specified range in larger than MAX_LARGE_SIZE_TO_CACHE the range is not cached if (range.size > MAX_LARGE_SIZE_TO_CACHE) return range; range = coalesce_range(szone->large_to_deallocate, LARGE_CACHE_SIZE, range); range_to_use = first_zero_range(szone->large_to_deallocate, LARGE_CACHE_SIZE); if (range_to_use) { // we fill an empty slot *range_to_use = range; return zero_range; } // we always try to deallocate the largest chunk range_to_use = largest_range(szone->large_to_deallocate, LARGE_CACHE_SIZE); if (!range_to_use) return range; vm_range_to_deallocate = *range_to_use; *range_to_use = range; return vm_range_to_deallocate; } static kern_return_t large_in_use_enumerator(task_t task, void *context, unsigned type_mask, vm_address_t large_entries_address, unsigned num_entries, memory_reader_t reader, vm_range_recorder_t recorder) { unsigned index = 0; vm_range_t buffer[MAX_RECORDER_BUFFER]; unsigned count = 0; large_entry_t *entries; kern_return_t err; err = reader(task, large_entries_address, sizeof(large_entry_t) * num_entries, (void **)&entries); if (err) return err; index = num_entries; if ((type_mask & MALLOC_ADMIN_REGION_RANGE_TYPE) && (num_entries * sizeof(large_entry_t) >= LARGE_THRESHOLD)) { vm_range_t range; range.address = large_entries_address; range.size = round_page(num_entries * sizeof(large_entry_t)); recorder(task, context, MALLOC_ADMIN_REGION_RANGE_TYPE, &range, 1); } if (type_mask & (MALLOC_PTR_IN_USE_RANGE_TYPE | MALLOC_PTR_REGION_RANGE_TYPE)) while (index--) { large_entry_t entry = entries[index]; if (!LARGE_ENTRY_IS_EMPTY(entry)) { vm_range_t range; range.address = LARGE_ENTRY_ADDRESS(entry); range.size = LARGE_ENTRY_SIZE(entry); buffer[count++] = range; if (count >= MAX_RECORDER_BUFFER) { recorder(task, context, MALLOC_PTR_IN_USE_RANGE_TYPE | MALLOC_PTR_REGION_RANGE_TYPE, buffer, count); count = 0; } } } if (count) recorder(task, context, MALLOC_PTR_IN_USE_RANGE_TYPE | MALLOC_PTR_REGION_RANGE_TYPE, buffer, count); return 0; } /********************* HUGE ENTRY UTILITIES ************************/ static huge_entry_t *huge_entry_for_pointer_no_lock(szone_t *szone, const void *ptr) { unsigned index = szone->num_huge_entries; while (index--) { huge_entry_t *huge = szone->huge_entries + index; if (huge->address == (vm_address_t)ptr) return huge; } return NULL; } static void huge_entry_append(szone_t *szone, huge_entry_t huge) { // We do a little dance with locking because doing allocation (even in the default szone) may cause something to get freed in this szone, with a deadlock huge_entry_t *new_huge_entries = NULL; SZONE_LOCK(szone); while (1) { unsigned num_huge_entries; num_huge_entries = szone->num_huge_entries; SZONE_UNLOCK(szone); // malloc_printf("In huge_entry_append currentEntries=%d\n", num_huge_entries); if (new_huge_entries) szone_free(szone, new_huge_entries); new_huge_entries = szone_malloc(szone, (num_huge_entries + 1) * sizeof(huge_entry_t)); SZONE_LOCK(szone); if (num_huge_entries == szone->num_huge_entries) { // No change - our malloc still applies huge_entry_t *old_huge_entries = szone->huge_entries; if (num_huge_entries) memcpy(new_huge_entries, old_huge_entries, num_huge_entries * sizeof(huge_entry_t)); new_huge_entries[szone->num_huge_entries++] = huge; szone->huge_entries = new_huge_entries; SZONE_UNLOCK(szone); szone_free(szone, old_huge_entries); // malloc_printf("Done huge_entry_append now=%d\n", szone->num_huge_entries); return; } // try again! } } static kern_return_t huge_in_use_enumerator(task_t task, void *context, unsigned type_mask, vm_address_t huge_entries_address, unsigned num_entries, memory_reader_t reader, vm_range_recorder_t recorder) { huge_entry_t *entries; kern_return_t err; err = reader(task, huge_entries_address, sizeof(huge_entry_t) * num_entries, (void **)&entries); if (err) return err; if (num_entries) recorder(task, context, MALLOC_PTR_IN_USE_RANGE_TYPE | MALLOC_PTR_REGION_RANGE_TYPE, entries, num_entries); return 0; } static void *large_and_huge_malloc(szone_t *szone, unsigned num_pages, boolean_t cleared_requested) { vm_address_t addr = 0; boolean_t cleared_needed = 0; // by default blocks will be freshly allocated and therefore no need to clean them if (!num_pages) num_pages = 1; // minimal allocation size for this szone // malloc_printf("In large_and_huge_malloc for %dKB\n", num_pages * vm_page_size / 1024); if (num_pages >= (1 << vm_page_shift)) { huge_entry_t huge; huge.size = num_pages << vm_page_shift; addr = allocate_pages(szone, huge.size, szone->debug_flags, VM_MAKE_TAG(VM_MEMORY_MALLOC_HUGE)); if (!addr) return NULL; huge.address = addr; huge_entry_append(szone, huge); SZONE_LOCK(szone); szone->num_bytes_in_huge_objects += huge.size; } else { vm_size_t size = num_pages << vm_page_shift; large_entry_t entry; boolean_t add_guard_pages = szone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES; vm_size_t guard_pages = (add_guard_pages) ? 2 * vm_page_size : 0; vm_range_t *range_to_use; cleared_needed = cleared_requested; // in the "large" case set by default SZONE_LOCK(szone); // First check in the list large_to_deallocate if we can reuse // malloc_printf("In szone_malloc checking recently deallocated\n"); range_to_use = first_range_greater_or_equal(szone->large_to_deallocate, LARGE_CACHE_SIZE, size + guard_pages); if (range_to_use) { // that one will do! addr = range_to_use->address + ((add_guard_pages) ? vm_page_size : 0); if (add_guard_pages) protect(szone, addr, size, 0, szone->debug_flags); // malloc_printf("In szone_malloc found recently deallocated at 0x%x for %d pages\n", addr, num_pages); if (range_to_use->size == size + guard_pages) { *range_to_use = zero_range; } else { range_to_use->address += size + guard_pages; range_to_use->size -= size + guard_pages; } #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, (void *)addr)) { malloc_printf("Entry about to be reused already in use: 0x%x\n", addr); large_debug_print(szone); sleep(3600); } #endif } if (!addr) { // we need to really allocate_pages a new region SZONE_UNLOCK(szone); addr = allocate_pages(szone, size, szone->debug_flags, VM_MAKE_TAG(VM_MEMORY_MALLOC_LARGE)); cleared_needed = 0; // since we allocated the pages, no need to clean them if (LOG(szone, addr)) malloc_printf("In szone_malloc true large allocation at %p for %dKB\n", (void *)addr, size / 1024); SZONE_LOCK(szone); if (!addr) return NULL; #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, (void *)addr)) { malloc_printf("Freshly allocated is already in use: 0x%x\n", addr); large_debug_print(szone); sleep(3600); } #endif } if ((szone->num_large_objects_in_use + 1) * 4 > szone->num_large_entries) { // density of hash table too high; grow table // we do that under lock to avoid a race // malloc_printf("In szone_malloc growing hash table current=%d\n", szone->num_large_entries); large_entries_grow_no_lock(szone); } // malloc_printf("Inserting large entry (0x%x, %dKB)\n", addr, num_pages * vm_page_size / 1024); entry.address_and_num_pages = addr | num_pages; #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, (void *)addr)) { malloc_printf("Entry about to be added already in use: 0x%x\n", addr); large_debug_print(szone); sleep(3600); } #endif large_entry_insert_no_lock(szone, entry); #if DEBUG_MALLOC if (!large_entry_for_pointer_no_lock(szone, (void *)addr)) { malloc_printf("Can't find entry just added\n"); large_debug_print(szone); sleep(3600); } #endif // malloc_printf("Inserted large entry (0x%x, %d pages)\n", addr, num_pages); szone->num_large_objects_in_use ++; szone->num_bytes_in_large_objects += size; } SZONE_UNLOCK(szone); if (cleared_needed) memset((void *)addr, 0, num_pages << vm_page_shift); return (void *)addr; } /********************* Zone call backs ************************/ static void szone_free(szone_t *szone, void *ptr) { region_t *region; large_entry_t *entry; vm_range_t vm_range_to_deallocate; huge_entry_t *huge; if (LOG(szone, ptr)) malloc_printf("In szone_free with %p\n", ptr); if (!ptr) return; if ((vm_address_t)ptr & (QUANTUM - 1)) { szone_error(szone, "Non-aligned pointer being freed", ptr); return; } // try a small pointer region = region_for_ptr_no_lock(szone, ptr); if (region) { // this is indeed a valid pointer msize_t msize_and_free; SZONE_LOCK(szone); msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); if (msize_and_free & THIS_FREE) { szone_error(szone, "Object already freed being freed", ptr); return; } CHECK(szone, __PRETTY_FUNCTION__); small_free_no_lock(szone, region, ptr, msize_and_free); CHECK(szone, __PRETTY_FUNCTION__); SZONE_UNLOCK(szone); return; } if (((unsigned)ptr) & (vm_page_size - 1)) { szone_error(szone, "Non-page-aligned, non-allocated pointer being freed", ptr); return; } SZONE_LOCK(szone); entry = large_entry_for_pointer_no_lock(szone, ptr); if (entry) { // malloc_printf("Ready for deallocation [0x%x-%dKB]\n", LARGE_ENTRY_ADDRESS(*entry), LARGE_ENTRY_SIZE(*entry)/1024); if (KILL_THRESHOLD && (LARGE_ENTRY_SIZE(*entry) > KILL_THRESHOLD)) { // We indicate to the VM system that these pages contain garbage and therefore don't need to be swapped out vm_msync(mach_task_self(), LARGE_ENTRY_ADDRESS(*entry), LARGE_ENTRY_SIZE(*entry), VM_SYNC_KILLPAGES); } vm_range_to_deallocate = large_free_no_lock(szone, entry); vm_range_to_deallocate = large_find_better_range_to_deallocate(szone, vm_range_to_deallocate); #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, ptr)) { malloc_printf("*** malloc[%d]: Just after freeing 0x%x still in use num_large_entries=%d\n", getpid(), ptr, szone->num_large_entries); large_cache_debug_print(szone); large_debug_print(szone); sleep(3600); } #endif } else if ((huge = huge_entry_for_pointer_no_lock(szone, ptr))) { vm_range_to_deallocate = *huge; *huge = szone->huge_entries[--szone->num_huge_entries]; // last entry fills that spot szone->num_bytes_in_huge_objects -= vm_range_to_deallocate.size; } else { #if DEBUG_MALLOC large_debug_print(szone); #endif szone_error(szone, "Pointer being freed was not allocated", ptr); return; } CHECK(szone, __PRETTY_FUNCTION__); SZONE_UNLOCK(szone); // we release the lock asap // we deallocate_pages, including guard pages if (vm_range_to_deallocate.address) { // malloc_printf("About to deallocate 0x%x size %dKB\n", vm_range_to_deallocate.address, vm_range_to_deallocate.size / 1024); #if DEBUG_MALLOC if (large_entry_for_pointer_no_lock(szone, (void *)vm_range_to_deallocate.address)) { malloc_printf("*** malloc[%d]: Invariant broken: 0x%x still in use num_large_entries=%d\n", getpid(), vm_range_to_deallocate.address, szone->num_large_entries); large_cache_debug_print(szone); large_debug_print(szone); sleep(3600); } #endif deallocate_pages(szone, vm_range_to_deallocate.address, vm_range_to_deallocate.size, 0); } } static INLINE void *szone_malloc_should_clear(szone_t *szone, size_t size, boolean_t cleared_requested) { void *ptr; if (!((szone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES) && PROTECT_SMALL) && (size < LARGE_THRESHOLD)) { // think small size_t msize = (size + PTR_HEADER_SIZE + QUANTUM - 1) >> SHIFT_QUANTUM; if (msize < MIN_BLOCK) msize = MIN_BLOCK; ptr = small_malloc_should_clear(szone, msize, cleared_requested); #if DEBUG_MALLOC if ((MSIZE_FLAGS_FOR_PTR(ptr) & ~ PREV_FREE) < msize) { malloc_printf("ptr=%p this=%d msize=%d\n", ptr, MSIZE_FLAGS_FOR_PTR(ptr), (int)msize); szone_error(szone, "Pointer allocated has improper size (1)", ptr); return NULL; } if ((MSIZE_FLAGS_FOR_PTR(ptr) & ~ PREV_FREE) < MIN_BLOCK) { malloc_printf("ptr=%p this=%d msize=%d\n", ptr, MSIZE_FLAGS_FOR_PTR(ptr), (int)msize); szone_error(szone, "Pointer allocated has improper size (2)", ptr); return NULL; } #endif } else { unsigned num_pages; num_pages = round_page(size) >> vm_page_shift; ptr = large_and_huge_malloc(szone, num_pages, cleared_requested); } if (LOG(szone, ptr)) malloc_printf("szone_malloc returned %p\n", ptr); return ptr; } static void *szone_malloc(szone_t *szone, size_t size) { return szone_malloc_should_clear(szone, size, 0); } static void *szone_calloc(szone_t *szone, size_t num_items, size_t size) { return szone_malloc_should_clear(szone, num_items * size, 1); } static void *szone_valloc(szone_t *szone, size_t size) { void *ptr; unsigned num_pages; num_pages = round_page(size) >> vm_page_shift; ptr = large_and_huge_malloc(szone, num_pages, 1); if (LOG(szone, ptr)) malloc_printf("szone_valloc returned %p\n", ptr); return ptr; } static size_t szone_size(szone_t *szone, const void *ptr) { size_t size = 0; region_t *region; large_entry_t *entry; huge_entry_t *huge; if (!ptr) return 0; if (LOG(szone, ptr)) malloc_printf("In szone_size for %p (szone=%p)\n", ptr, szone); if ((vm_address_t)ptr & (QUANTUM - 1)) return 0; if ((((unsigned)ptr) & (vm_page_size - 1)) && (MSIZE_FLAGS_FOR_PTR(ptr) & THIS_FREE)) { // not page aligned, but definitely not in use return 0; } // Try a small pointer region = region_for_ptr_no_lock(szone, ptr); // malloc_printf("FOUND REGION %p\n", region); if (region) { // this is indeed a valid pointer msize_t msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); return (msize_and_free & THIS_FREE) ? 0 : ((msize_and_free & ~PREV_FREE) << SHIFT_QUANTUM) - PTR_HEADER_SIZE; } if (((unsigned)ptr) & (vm_page_size - 1)) { return 0; } SZONE_LOCK(szone); entry = large_entry_for_pointer_no_lock(szone, ptr); if (entry) { size = LARGE_ENTRY_SIZE(*entry); } else if ((huge = huge_entry_for_pointer_no_lock(szone, ptr))) { size = huge->size; } SZONE_UNLOCK(szone); // malloc_printf("szone_size for large/huge %p returned %d\n", ptr, (unsigned)size); if (LOG(szone, ptr)) malloc_printf("szone_size for %p returned %d\n", ptr, (unsigned)size); return size; } static INLINE int szone_try_realloc_in_place(szone_t *szone, void *ptr, size_t old_size, size_t new_size) { // returns 1 on success void *next_block = (char *)ptr + old_size + PTR_HEADER_SIZE; msize_t next_msize_and_free; msize_t next_msize; region_t region; msize_t coalesced_msize; msize_t leftover_msize; msize_t new_msize_and_free; void *following_ptr; SZONE_LOCK(szone); region = szone->regions[szone->num_regions - 1]; if (((vm_address_t)ptr >= region) && ((vm_address_t)ptr < region + REGION_SIZE) && ((vm_address_t)next_block == REGION_END(region) - szone->num_bytes_free_in_last_region + PTR_HEADER_SIZE)) { // This could be optimized but it is so rare it's not worth it SZONE_UNLOCK(szone); return 0; } // If the next block is free, we coalesce next_msize_and_free = MSIZE_FLAGS_FOR_PTR(next_block); #if DEBUG_MALLOC if ((vm_address_t)next_block & (QUANTUM - 1)) { szone_error(szone, "Internal invariant broken in realloc(next_block)", next_block); } if (next_msize_and_free & PREV_FREE) { malloc_printf("szone_try_realloc_in_place: 0x%x=PREV_FREE|%d\n", next_msize_and_free, next_msize_and_free & ~PREV_FREE); SZONE_UNLOCK(szone); return 0; } #endif next_msize = next_msize_and_free & ~THIS_FREE; if (!(next_msize_and_free & THIS_FREE) || !next_msize || (old_size + (next_msize << SHIFT_QUANTUM) < new_size)) { SZONE_UNLOCK(szone); return 0; } coalesced_msize = (new_size - old_size + QUANTUM - 1) >> SHIFT_QUANTUM; leftover_msize = next_msize - coalesced_msize; new_msize_and_free = MSIZE_FLAGS_FOR_PTR(ptr); // malloc_printf("Realloc in place for %p; current msize=%d next_msize=%d wanted=%d\n", ptr, MSIZE_FLAGS_FOR_PTR(ptr), next_msize, new_size); free_list_remove_ptr(szone, next_block, next_msize); if ((leftover_msize < MIN_BLOCK) || (leftover_msize < coalesced_msize / 4)) { // don't bother splitting it off // malloc_printf("No leftover "); coalesced_msize = next_msize; leftover_msize = 0; } else { void *leftover = next_block + (coalesced_msize << SHIFT_QUANTUM); // malloc_printf("Leftover "); free_list_add_ptr(szone, leftover, leftover_msize); } new_msize_and_free += coalesced_msize; MSIZE_FLAGS_FOR_PTR(ptr) = new_msize_and_free; following_ptr = FOLLOWING_PTR(ptr, new_msize_and_free & ~PREV_FREE); MSIZE_FLAGS_FOR_PTR(following_ptr) &= ~ PREV_FREE; #if DEBUG_MALLOC { msize_t ms = MSIZE_FLAGS_FOR_PTR(following_ptr); msize_t pms = PREVIOUS_MSIZE(FOLLOWING_PTR(following_ptr, ms & ~THIS_FREE)); malloc_printf("Following ptr of coalesced (%p) has msize_and_free=0x%x=%s%d end_of_block_marker=%d\n", following_ptr, ms, (ms & THIS_FREE) ? "THIS_FREE|" : "", ms & ~THIS_FREE, pms); } if (LOG(szone,ptr)) malloc_printf("In szone_realloc(), ptr=%p, msize=%d\n", ptr, MSIZE_FLAGS_FOR_PTR(ptr)); #endif CHECK(szone, __PRETTY_FUNCTION__); szone->num_bytes_in_small_objects += coalesced_msize << SHIFT_QUANTUM; SZONE_UNLOCK(szone); // malloc_printf("Extended ptr %p for realloc old=%d desired=%d new=%d leftover=%d\n", ptr, (unsigned)old_size, (unsigned)new_size, (unsigned)szone_size(szone, ptr), leftover_msize << SHIFT_QUANTUM); return 1; } static void *szone_realloc(szone_t *szone, void *ptr, size_t new_size) { size_t old_size = 0; void *newPtr; if (LOG(szone, ptr)) malloc_printf("In szone_realloc for %p, %d\n", ptr, (unsigned)new_size); if (!ptr) return szone_malloc(szone, new_size); old_size = szone_size(szone, ptr); if (!old_size) { szone_error(szone, "Pointer being reallocated was not allocated", ptr); return NULL; } if (old_size >= new_size) return ptr; if (!((szone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES) && PROTECT_SMALL) && (new_size < LARGE_THRESHOLD)) { // We now try to realloc in place if (szone_try_realloc_in_place(szone, ptr, old_size, new_size)) return ptr; } newPtr = szone_malloc(szone, new_size); if ((old_size > VM_COPY_THRESHOLD) && (old_size < (1 << (vm_page_shift + vm_page_shift)))) { // we know it's a large block, and not a huge block kern_return_t err = 0; err = vm_copy(mach_task_self(), (vm_address_t)ptr, old_size, (vm_address_t)newPtr); if (err) { szone_error(szone, "Can't vm_copy region", ptr); } else { large_entry_t *entry; vm_range_t range; SZONE_LOCK(szone); entry = large_entry_for_pointer_no_lock(szone, ptr); if (!entry) { szone_error(szone, "Can't find entry for large copied block", ptr); } range = large_free_no_lock(szone, entry); SZONE_UNLOCK(szone); // we release the lock asap // we truly deallocate_pages, including guard pages deallocate_pages(szone, range.address, range.size, 0); if (LOG(szone, ptr)) malloc_printf("szone_realloc returned %p for %d\n", newPtr, (unsigned)new_size); return newPtr; } } else { memcpy(newPtr, ptr, old_size); } szone_free(szone, ptr); if (LOG(szone, ptr)) malloc_printf("szone_realloc returned %p for %d\n", newPtr, (unsigned)new_size); return newPtr; } static void szone_destroy(szone_t *szone) { unsigned index; index = szone->num_large_entries; while (index--) { large_entry_t *entry = szone->large_entries + index; if (!LARGE_ENTRY_IS_EMPTY(*entry)) { large_entry_t range; range = *entry; // we deallocate_pages, including guard pages deallocate_pages(szone, LARGE_ENTRY_ADDRESS(range), LARGE_ENTRY_SIZE(range), szone->debug_flags); } } if (szone->num_large_entries * sizeof(large_entry_t) >= LARGE_THRESHOLD) large_entries_free_no_lock(szone, szone->large_entries, szone->num_large_entries); // we do not free in the small chunk case index = LARGE_CACHE_SIZE; while (index--) { vm_range_t range = szone->large_to_deallocate[index]; if (range.size) deallocate_pages(szone, range.address, range.size, 0); } index = szone->num_huge_entries; while (index--) { huge_entry_t *huge = szone->huge_entries + index; deallocate_pages(szone, huge->address, huge->size, szone->debug_flags); } // and now we free regions, with regions[0] as the last one (the final harakiri) index = szone->num_regions; while (index--) { // we skip the first region, that is the zone itself region_t region = szone->regions[index]; deallocate_pages(szone, REGION_ADDRESS(region), REGION_SIZE, 0); } } static size_t szone_good_size(szone_t *szone, size_t size) { if (!((szone->debug_flags & SCALABLE_MALLOC_ADD_GUARD_PAGES) && PROTECT_SMALL) && (size < LARGE_THRESHOLD)) { // think small msize_t msize = (size + PTR_HEADER_SIZE + QUANTUM - 1) >> SHIFT_QUANTUM; if (msize < MIN_BLOCK) msize = MIN_BLOCK; return (msize << SHIFT_QUANTUM) - PTR_HEADER_SIZE; } else { unsigned num_pages; num_pages = round_page(size) >> vm_page_shift; if (!num_pages) num_pages = 1; // minimal allocation size for this return num_pages << vm_page_shift; } } unsigned szone_check_counter = 0; unsigned szone_check_start = 0; unsigned szone_check_modulo = 1; static boolean_t szone_check_all(szone_t *szone, const char *function) { unsigned index = 0; SZONE_LOCK(szone); while (index < szone->num_regions) { region_t *region = szone->regions + index++; if (!szone_check_region(szone, region)) { SZONE_UNLOCK(szone); szone->debug_flags &= ~ CHECK_REGIONS; malloc_printf("*** malloc[%d]: Region %d incorrect szone_check_all(%s) counter=%d\n", getpid(), index-1, function, szone_check_counter); szone_error(szone, "Check: region incorrect", NULL); return 0; } } index = 0; while (index < MAX_GRAIN) { if (! free_list_check(szone, index)) { SZONE_UNLOCK(szone); szone->debug_flags &= ~ CHECK_REGIONS; malloc_printf("*** malloc[%d]: Free list incorrect (grain=%d) szone_check_all(%s) counter=%d\n", getpid(), index, function, szone_check_counter); szone_error(szone, "Check: free list incorrect", NULL); return 0; } index++; } SZONE_UNLOCK(szone); return 1; } static boolean_t szone_check(szone_t *szone) { if (! (++szone_check_counter % 10000)) { malloc_printf("At szone_check counter=%d\n", szone_check_counter); } if (szone_check_counter < szone_check_start) return 1; if (szone_check_counter % szone_check_modulo) return 1; return szone_check_all(szone, ""); } static kern_return_t szone_ptr_in_use_enumerator(task_t task, void *context, unsigned type_mask, vm_address_t zone_address, memory_reader_t reader, vm_range_recorder_t recorder) { szone_t *szone; kern_return_t err; if (!reader) reader = _szone_default_reader; // malloc_printf("Enumerator for zone 0x%x\n", zone_address); err = reader(task, zone_address, sizeof(szone_t), (void **)&szone); if (err) return err; // malloc_printf("Small ptrs enumeration for zone 0x%x\n", zone_address); err = small_in_use_enumerator(task, context, type_mask, (vm_address_t)szone->regions, szone->num_regions, reader, recorder); if (err) return err; // malloc_printf("Large ptrs enumeration for zone 0x%x\n", zone_address); err = large_in_use_enumerator(task, context, type_mask, (vm_address_t)szone->large_entries, szone->num_large_entries, reader, recorder); if (err) return err; // malloc_printf("Huge ptrs enumeration for zone 0x%x\n", zone_address); err = huge_in_use_enumerator(task, context, type_mask, (vm_address_t)szone->huge_entries, szone->num_huge_entries, reader, recorder); return err; } static void szone_print_free_list(szone_t *szone) { grain_t grain = MAX_GRAIN; malloc_printf("Free Sizes: "); while (grain--) { free_list_t *ptr = szone->free_list[grain]; if (ptr) { unsigned count = 0; while (ptr) { count++; // malloc_printf("%p ", ptr); ptr = ptr->next; } malloc_printf("%s%d[%d] ", (grain == MAX_GRAIN-1) ? ">=" : "", (grain+1)*QUANTUM, count); } } malloc_printf("\n"); } static void szone_print(szone_t *szone, boolean_t verbose) { unsigned info[scalable_zone_info_count]; unsigned index; unsigned num = 0; index = LARGE_CACHE_SIZE; while (index--) if (szone->large_to_deallocate[index].size) num++; index = 0; scalable_zone_info((void *)szone, info, scalable_zone_info_count); malloc_printf("Scalable zone %p: inUse=%d(%dKB) small=%d(%dKB) large=%d(%dKB) to_be_deallocated=%d huge=%d(%dKB) guard_page=%d\n", szone, info[0], info[1] / 1024, info[2], info[3] / 1024, info[4], info[5] / 1024, num, info[6], info[7] / 1024, info[8]); malloc_printf("%d regions: \n", szone->num_regions); while (index < szone->num_regions) { region_t *region = szone->regions + index; unsigned counts[512]; unsigned ci = 0; unsigned in_use = 0; vm_address_t start = REGION_ADDRESS(*region) + QUANTUM; memset(counts, 0, 512 * sizeof(unsigned)); while (start < REGION_END(*region)) { msize_t msize_and_free = MSIZE_FLAGS_FOR_PTR(start); if (!(msize_and_free & THIS_FREE)) { msize_t msize = msize_and_free & ~PREV_FREE; if (!msize) break; // last encountered // block in use if (msize < 512) counts[msize]++; start += msize << SHIFT_QUANTUM; in_use++; } else { msize_t msize = msize_and_free & ~THIS_FREE; // free block start += msize << SHIFT_QUANTUM; } } malloc_printf("Region [0x%x-0x%x, %dKB] \tIn_use=%d ", REGION_ADDRESS(*region), REGION_END(*region), (int)REGION_SIZE / 1024, in_use); if (verbose) { malloc_printf("\n\tSizes in use: "); while (ci < 512) { if (counts[ci]) malloc_printf("%d[%d] ", ci << SHIFT_QUANTUM, counts[ci]); ci++; } } malloc_printf("\n"); index++; } if (verbose) szone_print_free_list(szone); malloc_printf("Free in last zone %d\n", szone->num_bytes_free_in_last_region); } static void szone_log(malloc_zone_t *zone, void *log_address) { szone_t *szone = (void *)zone; szone->log_address = log_address; } static void szone_force_lock(szone_t *szone) { // malloc_printf("szone_force_lock\n"); SZONE_LOCK(szone); } static void szone_force_unlock(szone_t *szone) { // malloc_printf("szone_force_unlock\n"); SZONE_UNLOCK(szone); } static struct malloc_introspection_t szone_introspect = {(void *)szone_ptr_in_use_enumerator, (void *)szone_good_size, (void *)szone_check, (void *)szone_print, szone_log, (void *)szone_force_lock, (void *)szone_force_unlock}; malloc_zone_t *create_scalable_zone(size_t initial_size, unsigned debug_flags) { szone_t *szone; vm_address_t addr; size_t msize; size_t msize_used = 0; if (!vm_page_shift) { unsigned page; vm_page_shift = 12; // the minimal for page sizes page = 1 << vm_page_shift; while (page != vm_page_size) { page += page; vm_page_shift++;}; if (MIN_BLOCK * QUANTUM < sizeof(free_list_t) + PTR_HEADER_SIZE) { malloc_printf("*** malloc[%d]: inconsistant parameters\n", getpid()); } } addr = allocate_pages(NULL, REGION_SIZE, 0, VM_MAKE_TAG(VM_MEMORY_MALLOC)); if (!addr) return NULL; szone = (void *)(addr + QUANTUM); msize = (sizeof(szone_t) + PTR_HEADER_SIZE + QUANTUM-1) >> SHIFT_QUANTUM; MSIZE_FLAGS_FOR_PTR(szone) = msize; msize_used += msize; szone->num_small_objects++; szone->basic_zone.size = (void *)szone_size; szone->basic_zone.malloc = (void *)szone_malloc; szone->basic_zone.calloc = (void *)szone_calloc; szone->basic_zone.valloc = (void *)szone_valloc; szone->basic_zone.free = (void *)szone_free; szone->basic_zone.realloc = (void *)szone_realloc; szone->basic_zone.destroy = (void *)szone_destroy; szone->basic_zone.introspect = &szone_introspect; LOCK_INIT(szone->lock); szone->debug_flags = debug_flags; szone->regions = (void *)((char *)szone + (msize << SHIFT_QUANTUM)); // we always reserve room for a few regions msize = (sizeof(region_t) * INITIAL_NUM_REGIONS + PTR_HEADER_SIZE + QUANTUM-1) >> SHIFT_QUANTUM; if (msize < MIN_BLOCK) msize = MIN_BLOCK; MSIZE_FLAGS_FOR_PTR(szone->regions) = msize; msize_used += msize; szone->num_small_objects++; szone->regions[0] = addr; szone->num_regions = 1; szone->num_bytes_free_in_last_region = REGION_SIZE - ((msize_used+1) << SHIFT_QUANTUM) + PTR_HEADER_SIZE; CHECK(szone, __PRETTY_FUNCTION__); return (malloc_zone_t *)szone; } /********* The following is private API for debug and perf tools ************/ void scalable_zone_info(malloc_zone_t *zone, unsigned *info_to_fill, unsigned count) { szone_t *szone = (void *)zone; unsigned info[scalable_zone_info_count]; // We do not lock to facilitate debug info[2] = szone->num_small_objects; info[3] = szone->num_bytes_in_small_objects; info[4] = szone->num_large_objects_in_use; info[5] = szone->num_bytes_in_large_objects; info[6] = szone->num_huge_entries; info[7] = szone->num_bytes_in_huge_objects; info[8] = szone->debug_flags; info[0] = info[2] + info[4] + info[6]; info[1] = info[3] + info[5] + info[7]; memcpy(info_to_fill, info, sizeof(unsigned)*count); }