blob: 3aeb5190455eff253e3be9d658ec5227ba576649 [file] [log] [blame]
/*
* Copyright (C) 2013 Google Inc. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are
* met:
*
* * Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above
* copyright notice, this list of conditions and the following disclaimer
* in the documentation and/or other materials provided with the
* distribution.
* * Neither the name of Google Inc. nor the names of its
* contributors may be used to endorse or promote products derived from
* this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
* OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
* LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
* THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
* OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "config.h"
#include "wtf/PartitionAlloc.h"
#include <string.h>
#ifndef NDEBUG
#include <stdio.h>
#endif
// Two partition pages are used as guard / metadata page so make sure the super
// page size is bigger.
COMPILE_ASSERT(WTF::kPartitionPageSize * 4 <= WTF::kSuperPageSize, ok_super_page_size);
COMPILE_ASSERT(!(WTF::kSuperPageSize % WTF::kPartitionPageSize), ok_super_page_multiple);
// Four system pages gives us room to hack out a still-guard-paged piece
// of metadata in the middle of a guard partition page.
COMPILE_ASSERT(WTF::kSystemPageSize * 4 <= WTF::kPartitionPageSize, ok_partition_page_size);
COMPILE_ASSERT(!(WTF::kPartitionPageSize % WTF::kSystemPageSize), ok_partition_page_multiple);
COMPILE_ASSERT(sizeof(WTF::PartitionPage) <= WTF::kPageMetadataSize, PartitionPage_not_too_big);
COMPILE_ASSERT(sizeof(WTF::PartitionSuperPageExtentEntry) <= WTF::kPageMetadataSize, PartitionSuperPageExtentEntry_not_too_big);
COMPILE_ASSERT(WTF::kPageMetadataSize * WTF::kNumPartitionPagesPerSuperPage <= WTF::kSystemPageSize, page_metadata_fits_in_hole);
namespace WTF {
static size_t partitionBucketPageSize(size_t size)
{
double bestWasteRatio = 1.0f;
size_t bestPages = 0;
for (size_t i = kNumSystemPagesPerPartitionPage - 1; i <= kNumSystemPagesPerPartitionPage; ++i) {
size_t pageSize = kSystemPageSize * i;
size_t numSlots = pageSize / size;
size_t waste = pageSize - (numSlots * size);
// Leave a page unfaulted is not free; the page will occupy an empty page table entry.
// Make a simple attempt to account for that.
waste += sizeof(void*) * (kNumSystemPagesPerPartitionPage - i);
double wasteRatio = (double) waste / (double) pageSize;
if (wasteRatio < bestWasteRatio) {
bestWasteRatio = wasteRatio;
bestPages = i;
}
}
ASSERT(bestPages > 0);
return bestPages * kSystemPageSize;
}
static ALWAYS_INLINE void partitionPageMarkFree(PartitionPage* page)
{
ASSERT(!partitionPageIsFree(page));
page->numAllocatedSlots = -1;
}
WTF_EXPORT void partitionAllocInit(PartitionRoot* root, size_t numBuckets, size_t maxAllocation)
{
ASSERT(!root->initialized);
root->initialized = true;
root->lock = 0;
root->totalSizeOfSuperPages = 0;
root->numBuckets = numBuckets;
root->maxAllocation = maxAllocation;
size_t i;
for (i = 0; i < root->numBuckets; ++i) {
PartitionBucket* bucket = &root->buckets()[i];
bucket->root = root;
bucket->activePagesHead = &root->seedPage;
bucket->freePagesHead = 0;
bucket->numFullPages = 0;
size_t size = partitionBucketSize(bucket);
bucket->pageSize = partitionBucketPageSize(size);
}
root->nextSuperPage = 0;
root->nextPartitionPage = 0;
root->nextPartitionPageEnd = 0;
root->currentExtent = &root->firstExtent;
root->firstExtent.superPageBase = 0;
root->firstExtent.superPagesEnd = 0;
root->firstExtent.next = 0;
root->seedPage.bucket = &root->seedBucket;
root->seedPage.activePageNext = 0;
root->seedPage.numAllocatedSlots = 0;
root->seedPage.numUnprovisionedSlots = 0;
partitionPageSetFreelistHead(&root->seedPage, 0);
// We mark the seed page as free to make sure it is skipped by our logic to
// find a new active page.
partitionPageMarkFree(&root->seedPage);
root->seedBucket.root = root;
root->seedBucket.activePagesHead = 0;
root->seedBucket.freePagesHead = 0;
root->seedBucket.numFullPages = 0;
}
static bool partitionAllocShutdownBucket(PartitionBucket* bucket)
{
// Failure here indicates a memory leak.
bool noLeaks = !bucket->numFullPages;
PartitionPage* page = bucket->activePagesHead;
if (page == &bucket->root->seedPage)
return noLeaks;
do {
if (page->numAllocatedSlots)
noLeaks = false;
page = page->activePageNext;
} while (page);
return noLeaks;
}
bool partitionAllocShutdown(PartitionRoot* root)
{
bool noLeaks = true;
ASSERT(root->initialized);
root->initialized = false;
size_t i;
for (i = 0; i < root->numBuckets; ++i) {
PartitionBucket* bucket = &root->buckets()[i];
if (!partitionAllocShutdownBucket(bucket))
noLeaks = false;
}
// Now that we've examined all partition pages in all buckets, it's safe
// to free all our super pages. We first collect the super page pointers
// on the stack because some of them are themselves store in super pages.
char* superPages[kMaxPartitionSize / kSuperPageSize];
size_t numSuperPages = 0;
PartitionSuperPageExtentEntry* entry = &root->firstExtent;
while (entry) {
char* superPage = entry->superPageBase;
while (superPage != entry->superPagesEnd) {
superPages[numSuperPages] = superPage;
numSuperPages++;
superPage += kSuperPageSize;
}
entry = entry->next;
}
ASSERT(numSuperPages == root->totalSizeOfSuperPages / kSuperPageSize);
for (size_t i = 0; i < numSuperPages; ++i) {
SuperPageBitmap::unregisterSuperPage(superPages[i]);
freePages(superPages[i], kSuperPageSize);
}
return noLeaks;
}
static ALWAYS_INLINE void* partitionAllocPage(PartitionRoot* root)
{
if (LIKELY(root->nextPartitionPage != 0)) {
// In this case, we can still hand out pages from a previous
// super page allocation.
char* ret = root->nextPartitionPage;
root->nextPartitionPage += kPartitionPageSize;
if (UNLIKELY(root->nextPartitionPage == root->nextPartitionPageEnd)) {
// We ran out, need to get more pages next time.
root->nextPartitionPage = 0;
root->nextPartitionPageEnd = 0;
}
return ret;
}
// Need a new super page.
root->totalSizeOfSuperPages += kSuperPageSize;
RELEASE_ASSERT(root->totalSizeOfSuperPages <= kMaxPartitionSize);
char* requestedAddress = root->nextSuperPage;
char* superPage = reinterpret_cast<char*>(allocPages(requestedAddress, kSuperPageSize, kSuperPageSize));
SuperPageBitmap::registerSuperPage(superPage);
root->nextSuperPage = superPage + kSuperPageSize;
char* ret = superPage + kPartitionPageSize;
root->nextPartitionPage = ret + kPartitionPageSize;
root->nextPartitionPageEnd = root->nextSuperPage - kPartitionPageSize;
// Make the first partition page in the super page a guard page, but leave a // hole in the middle.
// This is where we put page metadata and also a tiny amount of extent
// metadata.
setSystemPagesInaccessible(superPage, kSystemPageSize);
setSystemPagesInaccessible(superPage + (kSystemPageSize * 2), kPartitionPageSize - (kSystemPageSize * 2));
// Also make the last partition page a guard page.
setSystemPagesInaccessible(superPage + (kSuperPageSize - kPartitionPageSize), kPartitionPageSize);
// If we were after a specific address, but didn't get it, assume that
// the system chose a lousy address and re-randomize the next
// allocation.
if (requestedAddress && requestedAddress != superPage)
root->nextSuperPage = 0;
// We allocated a new super page so update super page metadata.
// First check if this is a new extent or not.
PartitionSuperPageExtentEntry* currentExtent = root->currentExtent;
bool isNewExtent = (superPage != requestedAddress);
if (UNLIKELY(isNewExtent)) {
if (currentExtent->superPageBase) {
// We already have a super page, so need to allocate metadata in the linked list.
PartitionSuperPageExtentEntry* newEntry = reinterpret_cast<PartitionSuperPageExtentEntry*>(partitionSuperPageToMetadataArea(currentExtent->superPageBase));
newEntry->next = 0;
currentExtent->next = newEntry;
currentExtent = newEntry;
root->currentExtent = newEntry;
}
currentExtent->superPageBase = superPage;
currentExtent->superPagesEnd = superPage + kSuperPageSize;
} else {
// We allocated next to an existing extent so just nudge the size up a little.
currentExtent->superPagesEnd += kSuperPageSize;
ASSERT(ret >= currentExtent->superPageBase && ret < currentExtent->superPagesEnd);
}
return ret;
}
static ALWAYS_INLINE void partitionUnusePage(PartitionPage* page)
{
void* addr = partitionPageToPointer(page);
decommitSystemPages(addr, kPartitionPageSize);
}
static ALWAYS_INLINE size_t partitionBucketSlots(const PartitionBucket* bucket)
{
return bucket->pageSize / partitionBucketSize(bucket);
}
static ALWAYS_INLINE void partitionPageReset(PartitionPage* page, PartitionBucket* bucket)
{
ASSERT(page != &bucket->root->seedPage);
page->numAllocatedSlots = 0;
page->numUnprovisionedSlots = partitionBucketSlots(bucket);
ASSERT(page->numUnprovisionedSlots > 1);
page->bucket = bucket;
// NULLing the freelist is not strictly necessary but it makes an ASSERT in partitionPageFillFreelist simpler.
partitionPageSetFreelistHead(page, 0);
}
static ALWAYS_INLINE char* partitionPageAllocAndFillFreelist(PartitionPage* page)
{
ASSERT(page != &page->bucket->root->seedPage);
size_t numSlots = page->numUnprovisionedSlots;
ASSERT(numSlots);
PartitionBucket* bucket = page->bucket;
// We should only get here when _every_ slot is either used or unprovisioned.
// (The third state is "on the freelist". If we have a non-empty freelist, we should not get here.)
ASSERT(numSlots + page->numAllocatedSlots == partitionBucketSlots(bucket));
// Similarly, make explicitly sure that the freelist is empty.
ASSERT(!partitionPageFreelistHead(page));
size_t size = partitionBucketSize(bucket);
char* base = reinterpret_cast<char*>(partitionPageToPointer(page));
char* returnObject = base + (size * page->numAllocatedSlots);
char* nextFreeObject = returnObject + size;
char* subPageLimit = reinterpret_cast<char*>((reinterpret_cast<uintptr_t>(returnObject) + kSystemPageSize) & kSystemPageBaseMask);
size_t numNewFreelistEntries = 0;
if (LIKELY(subPageLimit > nextFreeObject))
numNewFreelistEntries = (subPageLimit - nextFreeObject) / size;
// We always return an object slot -- that's the +1 below.
// We do not neccessarily create any new freelist entries, because we cross sub page boundaries frequently for large bucket sizes.
numSlots -= (numNewFreelistEntries + 1);
page->numUnprovisionedSlots = numSlots;
page->numAllocatedSlots++;
if (LIKELY(numNewFreelistEntries)) {
PartitionFreelistEntry* entry = reinterpret_cast<PartitionFreelistEntry*>(nextFreeObject);
partitionPageSetFreelistHead(page, entry);
while (--numNewFreelistEntries) {
nextFreeObject += size;
PartitionFreelistEntry* nextEntry = reinterpret_cast<PartitionFreelistEntry*>(nextFreeObject);
entry->next = partitionFreelistMask(nextEntry);
entry = nextEntry;
}
entry->next = partitionFreelistMask(0);
} else {
partitionPageSetFreelistHead(page, 0);
}
return returnObject;
}
// This helper function scans the active page list for a suitable new active
// page, starting at the page passed in. When it finds a suitable new active
// page (one that has free slots), it is also set as the new active page. If
// there is no suitable new active page, the current active page is set to null.
static ALWAYS_INLINE void partitionSetNewActivePage(PartitionBucket* bucket, PartitionPage* page)
{
ASSERT(page == &bucket->root->seedPage || page->bucket == bucket);
for (; page; page = page->activePageNext) {
// Skip over free pages. We need this check first so that we can trust
// that the page union field really containts a freelist pointer.
if (UNLIKELY(partitionPageIsFree(page)))
continue;
// Page is usable if it has something on the freelist, or unprovisioned
// slots that can be turned into a freelist.
if (LIKELY(partitionPageFreelistHead(page) != 0) || LIKELY(page->numUnprovisionedSlots))
break;
// If we get here, we found a full page. Skip over it too, and also tag
// it as full (via a negative value). We need it tagged so that free'ing
// can tell, and move it back into the active page list.
ASSERT(page->numAllocatedSlots == static_cast<int>(partitionBucketSlots(bucket)));
page->numAllocatedSlots = -page->numAllocatedSlots;
++bucket->numFullPages;
}
bucket->activePagesHead = page;
}
static ALWAYS_INLINE void partitionPageSetFreePageNext(PartitionPage* page, PartitionPage* nextFree)
{
ASSERT(partitionPageIsFree(page));
page->u.freePageNext = nextFree;
}
static ALWAYS_INLINE PartitionPage* partitionPageFreePageNext(PartitionPage* page)
{
ASSERT(partitionPageIsFree(page));
return page->u.freePageNext;
}
void* partitionAllocSlowPath(PartitionBucket* bucket)
{
// The slow path is called when the freelist is empty.
ASSERT(!partitionPageFreelistHead(bucket->activePagesHead));
// First, look for a usable page in the existing active pages list.
PartitionPage* page = bucket->activePagesHead;
partitionSetNewActivePage(bucket, page);
page = bucket->activePagesHead;
if (LIKELY(page != 0)) {
if (LIKELY(partitionPageFreelistHead(page) != 0)) {
PartitionFreelistEntry* ret = partitionPageFreelistHead(page);
partitionPageSetFreelistHead(page, partitionFreelistMask(ret->next));
page->numAllocatedSlots++;
return ret;
}
ASSERT(page->numUnprovisionedSlots);
return partitionPageAllocAndFillFreelist(page);
}
// Second, look in our list of freed but reserved pages.
PartitionPage* newPage = bucket->freePagesHead;
if (LIKELY(newPage != 0)) {
ASSERT(newPage != &bucket->root->seedPage);
bucket->freePagesHead = partitionPageFreePageNext(newPage);
} else {
// Third. If we get here, we need a brand new page.
void* rawNewPage = partitionAllocPage(bucket->root);
newPage = partitionPointerToPageNoAlignmentCheck(rawNewPage);
}
newPage->activePageNext = 0;
partitionPageReset(newPage, bucket);
bucket->activePagesHead = newPage;
return partitionPageAllocAndFillFreelist(newPage);
}
void partitionFreeSlowPath(PartitionPage* page)
{
PartitionBucket* bucket = page->bucket;
ASSERT(page != &bucket->root->seedPage);
ASSERT(bucket->activePagesHead != &bucket->root->seedPage);
if (LIKELY(page->numAllocatedSlots == 0)) {
// Page became fully unused.
// If it's the current page, change it!
if (LIKELY(page == bucket->activePagesHead)) {
PartitionPage* newPage = 0;
if (page->activePageNext) {
partitionSetNewActivePage(bucket, page->activePageNext);
newPage = bucket->activePagesHead;
}
ASSERT(newPage != page);
if (UNLIKELY(!newPage)) {
// We do not free the last active page in a bucket.
// To do so is a serious performance issue as we can get into
// a repeating state change between 0 and 1 objects of a given
// size allocated; and each state change incurs either a system
// call or a page fault, or more.
page->activePageNext = 0;
bucket->activePagesHead = page;
return;
}
bucket->activePagesHead = newPage;
}
partitionUnusePage(page);
// We actually leave the freed page in the active list as well as
// putting it on the free list. We'll evict it from the active list soon
// enough in partitionAllocSlowPath. Pulling this trick enables us to
// use a singly-linked page list for all cases, which is critical in
// keeping the page metadata structure down to 32-bytes in size.
partitionPageMarkFree(page);
partitionPageSetFreePageNext(page, bucket->freePagesHead);
bucket->freePagesHead = page;
} else {
// Ensure that the page is full. That's the only valid case if we
// arrive here.
ASSERT(page->numAllocatedSlots < 0);
// Fully used page became partially used. It must be put back on the
// non-full page list. Also make it the current page to increase the
// chances of it being filled up again. The old current page will be
// the next page.
page->numAllocatedSlots = -page->numAllocatedSlots - 2;
ASSERT(page->numAllocatedSlots == static_cast<int>(partitionBucketSlots(bucket) - 1));
page->activePageNext = bucket->activePagesHead;
bucket->activePagesHead = page;
--bucket->numFullPages;
// Note: there's an opportunity here to look to see if the old active
// page is now freeable.
}
}
void* partitionReallocGeneric(PartitionRoot* root, void* ptr, size_t newSize)
{
RELEASE_ASSERT(newSize <= QuantizedAllocation::kMaxUnquantizedAllocation);
#if defined(MEMORY_TOOL_REPLACES_ALLOCATOR)
return realloc(ptr, newSize);
#else
size_t oldIndex;
if (LIKELY(partitionPointerIsValid(root, ptr))) {
void* realPtr = partitionCookieFreePointerAdjust(ptr);
PartitionBucket* bucket = partitionPointerToPage(realPtr)->bucket;
ASSERT(bucket->root == root);
oldIndex = bucket - root->buckets();
} else {
oldIndex = root->numBuckets;
}
size_t allocSize = QuantizedAllocation::quantizedSize(newSize);
allocSize = partitionCookieSizeAdjustAdd(allocSize);
size_t newIndex = allocSize >> kBucketShift;
if (newIndex > root->numBuckets)
newIndex = root->numBuckets;
if (oldIndex == newIndex) {
// Same bucket. But kNumBuckets indicates the fastMalloc "bucket" so in
// that case we're not done; we have to forward to fastRealloc.
if (oldIndex == root->numBuckets)
return WTF::fastRealloc(ptr, newSize);
return ptr;
}
// This realloc cannot be resized in-place. Sadness.
void* ret = partitionAllocGeneric(root, newSize);
size_t copySize = oldIndex << kBucketShift;
copySize = partitionCookieSizeAdjustSubtract(copySize);
if (newSize < copySize)
copySize = newSize;
memcpy(ret, ptr, copySize);
partitionFreeGeneric(root, ptr);
return ret;
#endif
}
#if CPU(32BIT)
unsigned char SuperPageBitmap::s_bitmap[1 << (32 - kSuperPageShift - 3)];
static int bitmapLock = 0;
void SuperPageBitmap::registerSuperPage(void* ptr)
{
ASSERT(!isPointerInSuperPage(ptr));
uintptr_t raw = reinterpret_cast<uintptr_t>(ptr);
raw >>= kSuperPageShift;
size_t byteIndex = raw >> 3;
size_t bit = raw & 7;
ASSERT(byteIndex < sizeof(s_bitmap));
// The read/modify/write is not guaranteed atomic, so take a lock.
spinLockLock(&bitmapLock);
s_bitmap[byteIndex] |= (1 << bit);
spinLockUnlock(&bitmapLock);
}
void SuperPageBitmap::unregisterSuperPage(void* ptr)
{
ASSERT(isPointerInSuperPage(ptr));
uintptr_t raw = reinterpret_cast<uintptr_t>(ptr);
raw >>= kSuperPageShift;
size_t byteIndex = raw >> 3;
size_t bit = raw & 7;
ASSERT(byteIndex < sizeof(s_bitmap));
// The read/modify/write is not guaranteed atomic, so take a lock.
spinLockLock(&bitmapLock);
s_bitmap[byteIndex] &= ~(1 << bit);
spinLockUnlock(&bitmapLock);
}
#endif
#ifndef NDEBUG
void partitionDumpStats(const PartitionRoot& root)
{
size_t i;
size_t totalLive = 0;
size_t totalResident = 0;
size_t totalFreeable = 0;
for (i = 0; i < root.numBuckets; ++i) {
const PartitionBucket& bucket = root.buckets()[i];
if (bucket.activePagesHead == &bucket.root->seedPage && !bucket.freePagesHead && !bucket.numFullPages) {
// Empty bucket with no freelist or full pages. Skip reporting it.
continue;
}
size_t numFreePages = 0;
PartitionPage* freePages = bucket.freePagesHead;
while (freePages) {
++numFreePages;
freePages = partitionPageFreePageNext(freePages);
}
size_t bucketSlotSize = partitionBucketSize(&bucket);
size_t bucketNumSlots = partitionBucketSlots(&bucket);
size_t bucketUsefulStorage = bucketSlotSize * bucketNumSlots;
size_t bucketWaste = bucket.pageSize - bucketUsefulStorage;
size_t numActiveBytes = bucket.numFullPages * bucketUsefulStorage;
size_t numResidentBytes = bucket.numFullPages * bucket.pageSize;
size_t numFreeableBytes = 0;
size_t numActivePages = 0;
const PartitionPage* page = bucket.activePagesHead;
do {
if (page != &bucket.root->seedPage) {
++numActivePages;
numActiveBytes += (page->numAllocatedSlots * bucketSlotSize);
size_t pageBytesResident = (bucketNumSlots - page->numUnprovisionedSlots) * bucketSlotSize;
// Round up to system page size.
pageBytesResident = (pageBytesResident + kSystemPageOffsetMask) & kSystemPageBaseMask;
numResidentBytes += pageBytesResident;
if (!page->numAllocatedSlots)
numFreeableBytes += pageBytesResident;
}
page = page->activePageNext;
} while (page != bucket.activePagesHead);
totalLive += numActiveBytes;
totalResident += numResidentBytes;
totalFreeable += numFreeableBytes;
printf("bucket size %zu (pageSize %zu waste %zu): %zu alloc/%zu commit/%zu freeable bytes, %zu/%zu/%zu full/active/free pages\n", bucketSlotSize, static_cast<size_t>(bucket.pageSize), bucketWaste, numActiveBytes, numResidentBytes, numFreeableBytes, static_cast<size_t>(bucket.numFullPages), numActivePages, numFreePages);
}
printf("total live: %zu bytes\n", totalLive);
printf("total resident: %zu bytes\n", totalResident);
printf("total freeable: %zu bytes\n", totalFreeable);
fflush(stdout);
}
#endif // !NDEBUG
} // namespace WTF