blob: 779a9be65f7fce0658dea8eca219c35a76e7124f [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 "platform/heap/Heap.h"
#include "platform/ScriptForbiddenScope.h"
#include "platform/Task.h"
#include "platform/TraceEvent.h"
#include "platform/heap/CallbackStack.h"
#include "platform/heap/MarkingVisitorImpl.h"
#include "platform/heap/ThreadState.h"
#include "public/platform/Platform.h"
#include "wtf/AddressSpaceRandomization.h"
#include "wtf/Assertions.h"
#include "wtf/LeakAnnotations.h"
#include "wtf/PassOwnPtr.h"
#if ENABLE(GC_PROFILE_MARKING)
#include "wtf/HashMap.h"
#include "wtf/HashSet.h"
#include "wtf/text/StringBuilder.h"
#include "wtf/text/StringHash.h"
#include <stdio.h>
#include <utility>
#endif
#if ENABLE(GC_PROFILE_HEAP)
#include "platform/TracedValue.h"
#endif
#if OS(POSIX)
#include <sys/mman.h>
#include <unistd.h>
#elif OS(WIN)
#include <windows.h>
#endif
namespace blink {
#if ENABLE(GC_PROFILE_MARKING)
static String classOf(const void* object)
{
if (const GCInfo* gcInfo = Heap::findGCInfo(reinterpret_cast<Address>(const_cast<void*>(object))))
return gcInfo->m_className;
return "unknown";
}
#endif
static bool vTableInitialized(void* objectPointer)
{
return !!(*reinterpret_cast<Address*>(objectPointer));
}
#if OS(WIN)
static bool IsPowerOf2(size_t power)
{
return !((power - 1) & power);
}
#endif
static Address roundToBlinkPageBoundary(void* base)
{
return reinterpret_cast<Address>((reinterpret_cast<uintptr_t>(base) + blinkPageOffsetMask) & blinkPageBaseMask);
}
static size_t roundToOsPageSize(size_t size)
{
return (size + WTF::kSystemPageSize - 1) & ~(WTF::kSystemPageSize - 1);
}
class MemoryRegion {
public:
MemoryRegion(Address base, size_t size)
: m_base(base)
, m_size(size)
{
ASSERT(size > 0);
}
bool contains(Address addr) const
{
return m_base <= addr && addr < (m_base + m_size);
}
bool contains(const MemoryRegion& other) const
{
return contains(other.m_base) && contains(other.m_base + other.m_size - 1);
}
void release()
{
#if OS(POSIX)
int err = munmap(m_base, m_size);
RELEASE_ASSERT(!err);
#else
bool success = VirtualFree(m_base, 0, MEM_RELEASE);
RELEASE_ASSERT(success);
#endif
}
WARN_UNUSED_RETURN bool commit()
{
#if OS(POSIX)
return !mprotect(m_base, m_size, PROT_READ | PROT_WRITE);
#else
void* result = VirtualAlloc(m_base, m_size, MEM_COMMIT, PAGE_READWRITE);
return !!result;
#endif
}
void decommit()
{
#if OS(POSIX)
int err = mprotect(m_base, m_size, PROT_NONE);
RELEASE_ASSERT(!err);
// FIXME: Consider using MADV_FREE on MacOS.
madvise(m_base, m_size, MADV_DONTNEED);
#else
bool success = VirtualFree(m_base, m_size, MEM_DECOMMIT);
RELEASE_ASSERT(success);
#endif
}
Address base() const { return m_base; }
size_t size() const { return m_size; }
private:
Address m_base;
size_t m_size;
};
// A PageMemoryRegion represents a chunk of reserved virtual address
// space containing a number of blink heap pages. On Windows, reserved
// virtual address space can only be given back to the system as a
// whole. The PageMemoryRegion allows us to do that by keeping track
// of the number of pages using it in order to be able to release all
// of the virtual address space when there are no more pages using it.
class PageMemoryRegion : public MemoryRegion {
public:
~PageMemoryRegion()
{
release();
}
void pageDeleted(Address page)
{
markPageUnused(page);
if (!--m_numPages) {
Heap::removePageMemoryRegion(this);
delete this;
}
}
void markPageUsed(Address page)
{
ASSERT(!m_inUse[index(page)]);
m_inUse[index(page)] = true;
}
void markPageUnused(Address page)
{
m_inUse[index(page)] = false;
}
static PageMemoryRegion* allocateLargePage(size_t size)
{
return allocate(size, 1);
}
static PageMemoryRegion* allocateNormalPages()
{
return allocate(blinkPageSize * blinkPagesPerRegion, blinkPagesPerRegion);
}
BaseHeapPage* pageFromAddress(Address address)
{
ASSERT(contains(address));
if (!m_inUse[index(address)])
return nullptr;
if (m_isLargePage)
return pageFromObject(base());
return pageFromObject(address);
}
private:
PageMemoryRegion(Address base, size_t size, unsigned numPages)
: MemoryRegion(base, size)
, m_isLargePage(numPages == 1)
, m_numPages(numPages)
{
for (size_t i = 0; i < blinkPagesPerRegion; ++i)
m_inUse[i] = false;
}
unsigned index(Address address)
{
ASSERT(contains(address));
if (m_isLargePage)
return 0;
size_t offset = blinkPageAddress(address) - base();
ASSERT(offset % blinkPageSize == 0);
return offset / blinkPageSize;
}
static PageMemoryRegion* allocate(size_t size, unsigned numPages)
{
// Compute a random blink page aligned address for the page memory
// region and attempt to get the memory there.
Address randomAddress = reinterpret_cast<Address>(WTF::getRandomPageBase());
Address alignedRandomAddress = roundToBlinkPageBoundary(randomAddress);
#if OS(POSIX)
Address base = static_cast<Address>(mmap(alignedRandomAddress, size, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0));
if (base == roundToBlinkPageBoundary(base))
return new PageMemoryRegion(base, size, numPages);
// We failed to get a blink page aligned chunk of memory.
// Unmap the chunk that we got and fall back to overallocating
// and selecting an aligned sub part of what we allocate.
if (base != MAP_FAILED) {
int error = munmap(base, size);
RELEASE_ASSERT(!error);
}
size_t allocationSize = size + blinkPageSize;
for (int attempt = 0; attempt < 10; ++attempt) {
base = static_cast<Address>(mmap(alignedRandomAddress, allocationSize, PROT_NONE, MAP_ANON | MAP_PRIVATE, -1, 0));
if (base != MAP_FAILED)
break;
randomAddress = reinterpret_cast<Address>(WTF::getRandomPageBase());
alignedRandomAddress = roundToBlinkPageBoundary(randomAddress);
}
RELEASE_ASSERT(base != MAP_FAILED);
Address end = base + allocationSize;
Address alignedBase = roundToBlinkPageBoundary(base);
Address regionEnd = alignedBase + size;
// If the allocated memory was not blink page aligned release
// the memory before the aligned address.
if (alignedBase != base)
MemoryRegion(base, alignedBase - base).release();
// Free the additional memory at the end of the page if any.
if (regionEnd < end)
MemoryRegion(regionEnd, end - regionEnd).release();
return new PageMemoryRegion(alignedBase, size, numPages);
#else
Address base = static_cast<Address>(VirtualAlloc(alignedRandomAddress, size, MEM_RESERVE, PAGE_NOACCESS));
if (base) {
ASSERT(base == alignedRandomAddress);
return new PageMemoryRegion(base, size, numPages);
}
// We failed to get the random aligned address that we asked
// for. Fall back to overallocating. On Windows it is
// impossible to partially release a region of memory
// allocated by VirtualAlloc. To avoid wasting virtual address
// space we attempt to release a large region of memory
// returned as a whole and then allocate an aligned region
// inside this larger region.
size_t allocationSize = size + blinkPageSize;
for (int attempt = 0; attempt < 3; ++attempt) {
base = static_cast<Address>(VirtualAlloc(0, allocationSize, MEM_RESERVE, PAGE_NOACCESS));
RELEASE_ASSERT(base);
VirtualFree(base, 0, MEM_RELEASE);
Address alignedBase = roundToBlinkPageBoundary(base);
base = static_cast<Address>(VirtualAlloc(alignedBase, size, MEM_RESERVE, PAGE_NOACCESS));
if (base) {
ASSERT(base == alignedBase);
return new PageMemoryRegion(alignedBase, size, numPages);
}
}
// We failed to avoid wasting virtual address space after
// several attempts.
base = static_cast<Address>(VirtualAlloc(0, allocationSize, MEM_RESERVE, PAGE_NOACCESS));
RELEASE_ASSERT(base);
// FIXME: If base is by accident blink page size aligned
// here then we can create two pages out of reserved
// space. Do this.
Address alignedBase = roundToBlinkPageBoundary(base);
return new PageMemoryRegion(alignedBase, size, numPages);
#endif
}
bool m_isLargePage;
bool m_inUse[blinkPagesPerRegion];
unsigned m_numPages;
};
// Representation of the memory used for a Blink heap page.
//
// The representation keeps track of two memory regions:
//
// 1. The virtual memory reserved from the system in order to be able
// to free all the virtual memory reserved. Multiple PageMemory
// instances can share the same reserved memory region and
// therefore notify the reserved memory region on destruction so
// that the system memory can be given back when all PageMemory
// instances for that memory are gone.
//
// 2. The writable memory (a sub-region of the reserved virtual
// memory region) that is used for the actual heap page payload.
//
// Guard pages are created before and after the writable memory.
class PageMemory {
public:
~PageMemory()
{
__lsan_unregister_root_region(m_writable.base(), m_writable.size());
m_reserved->pageDeleted(writableStart());
}
WARN_UNUSED_RETURN bool commit()
{
m_reserved->markPageUsed(writableStart());
return m_writable.commit();
}
void decommit()
{
m_reserved->markPageUnused(writableStart());
m_writable.decommit();
}
void markUnused() { m_reserved->markPageUnused(writableStart()); }
PageMemoryRegion* region() { return m_reserved; }
Address writableStart() { return m_writable.base(); }
static PageMemory* setupPageMemoryInRegion(PageMemoryRegion* region, size_t pageOffset, size_t payloadSize)
{
// Setup the payload one OS page into the page memory. The
// first os page is the guard page.
Address payloadAddress = region->base() + pageOffset + WTF::kSystemPageSize;
return new PageMemory(region, MemoryRegion(payloadAddress, payloadSize));
}
// Allocate a virtual address space for one blink page with the
// following layout:
//
// [ guard os page | ... payload ... | guard os page ]
// ^---{ aligned to blink page size }
//
static PageMemory* allocate(size_t payloadSize)
{
ASSERT(payloadSize > 0);
// Virtual memory allocation routines operate in OS page sizes.
// Round up the requested size to nearest os page size.
payloadSize = roundToOsPageSize(payloadSize);
// Overallocate by 2 times OS page size to have space for a
// guard page at the beginning and end of blink heap page.
size_t allocationSize = payloadSize + 2 * WTF::kSystemPageSize;
PageMemoryRegion* pageMemoryRegion = PageMemoryRegion::allocateLargePage(allocationSize);
PageMemory* storage = setupPageMemoryInRegion(pageMemoryRegion, 0, payloadSize);
RELEASE_ASSERT(storage->commit());
return storage;
}
private:
PageMemory(PageMemoryRegion* reserved, const MemoryRegion& writable)
: m_reserved(reserved)
, m_writable(writable)
{
ASSERT(reserved->contains(writable));
// Register the writable area of the memory as part of the LSan root set.
// Only the writable area is mapped and can contain C++ objects. Those
// C++ objects can contain pointers to objects outside of the heap and
// should therefore be part of the LSan root set.
__lsan_register_root_region(m_writable.base(), m_writable.size());
}
PageMemoryRegion* m_reserved;
MemoryRegion m_writable;
};
class GCScope {
public:
explicit GCScope(ThreadState::StackState stackState)
: m_state(ThreadState::current())
, m_safePointScope(stackState)
, m_parkedAllThreads(false)
{
TRACE_EVENT0("blink_gc", "Heap::GCScope");
const char* samplingState = TRACE_EVENT_GET_SAMPLING_STATE();
if (m_state->isMainThread())
TRACE_EVENT_SET_SAMPLING_STATE("blink_gc", "BlinkGCWaiting");
m_state->checkThread();
// FIXME: in an unlikely coincidence that two threads decide
// to collect garbage at the same time, avoid doing two GCs in
// a row.
if (LIKELY(ThreadState::stopThreads())) {
m_parkedAllThreads = true;
}
if (m_state->isMainThread())
TRACE_EVENT_SET_NONCONST_SAMPLING_STATE(samplingState);
}
bool allThreadsParked() { return m_parkedAllThreads; }
~GCScope()
{
// Only cleanup if we parked all threads in which case the GC happened
// and we need to resume the other threads.
if (LIKELY(m_parkedAllThreads)) {
ThreadState::resumeThreads();
}
}
private:
ThreadState* m_state;
ThreadState::SafePointScope m_safePointScope;
bool m_parkedAllThreads; // False if we fail to park all threads
};
#if ENABLE(ASSERT)
NO_SANITIZE_ADDRESS
void HeapObjectHeader::zapMagic()
{
checkHeader();
m_magic = zappedMagic;
}
#endif
void HeapObjectHeader::finalize(const GCInfo* gcInfo, Address object, size_t objectSize)
{
ASSERT(gcInfo);
if (gcInfo->hasFinalizer()) {
gcInfo->m_finalize(object);
}
#if ENABLE(ASSERT) || defined(LEAK_SANITIZER) || defined(ADDRESS_SANITIZER)
// In Debug builds, memory is zapped when it's freed, and the zapped memory
// is zeroed out when the memory is reused. Memory is also zapped when
// using Leak Sanitizer because the heap is used as a root region for LSan
// and therefore pointers in unreachable memory could hide leaks.
for (size_t i = 0; i < objectSize; ++i)
object[i] = finalizedZapValue;
// Zap the primary vTable entry (secondary vTable entries are not zapped).
if (gcInfo->hasVTable()) {
*(reinterpret_cast<uintptr_t*>(object)) = zappedVTable;
}
#endif
// In Release builds, the entire object is zeroed out when it is added to
// the free list. This happens right after sweeping the page and before the
// thread commences execution.
}
NO_SANITIZE_ADDRESS
void GeneralHeapObjectHeader::finalize()
{
HeapObjectHeader::finalize(m_gcInfo, payload(), payloadSize());
}
template<typename Header>
void LargeObject<Header>::sweep()
{
Heap::increaseMarkedObjectSize(size());
heapObjectHeader()->unmark();
}
template<typename Header>
bool LargeObject<Header>::isEmpty()
{
return !heapObjectHeader()->isMarked();
}
template<typename Header>
void LargeObject<Header>::checkAndMarkPointer(Visitor* visitor, Address address)
{
ASSERT(contains(address));
if (!objectContains(address) || heapObjectHeader()->isDead())
return;
#if ENABLE(GC_PROFILE_MARKING)
visitor->setHostInfo(&address, "stack");
#endif
mark(visitor);
}
template<typename Header>
void LargeObject<Header>::markUnmarkedObjectsDead()
{
Header* header = heapObjectHeader();
if (header->isMarked())
header->unmark();
else
header->markDead();
}
#if ENABLE(ASSERT)
static bool isUninitializedMemory(void* objectPointer, size_t objectSize)
{
// Scan through the object's fields and check that they are all zero.
Address* objectFields = reinterpret_cast<Address*>(objectPointer);
for (size_t i = 0; i < objectSize / sizeof(Address); ++i) {
if (objectFields[i] != 0)
return false;
}
return true;
}
#endif
template<>
void LargeObject<GeneralHeapObjectHeader>::mark(Visitor* visitor)
{
GeneralHeapObjectHeader* header = heapObjectHeader();
if (header->hasVTable() && !vTableInitialized(payload())) {
visitor->markHeaderNoTracing(header);
ASSERT(isUninitializedMemory(header->payload(), header->payloadSize()));
} else {
visitor->markHeader(header, header->traceCallback());
}
}
template<>
void LargeObject<HeapObjectHeader>::mark(Visitor* visitor)
{
ASSERT(gcInfo());
if (gcInfo()->hasVTable() && !vTableInitialized(payload())) {
HeapObjectHeader* header = heapObjectHeader();
visitor->markHeaderNoTracing(header);
ASSERT(isUninitializedMemory(header->payload(), header->payloadSize()));
} else {
visitor->mark(heapObjectHeader(), gcInfo()->m_trace);
}
}
template<>
void LargeObject<GeneralHeapObjectHeader>::finalize()
{
heapObjectHeader()->finalize();
}
template<>
void LargeObject<HeapObjectHeader>::finalize()
{
ASSERT(gcInfo());
HeapObjectHeader::finalize(gcInfo(), payload(), payloadSize());
}
template<typename Header>
ThreadHeap<Header>::ThreadHeap(ThreadState* state, int index)
: m_currentAllocationPoint(nullptr)
, m_remainingAllocationSize(0)
, m_lastRemainingAllocationSize(0)
, m_firstPage(nullptr)
, m_firstLargeObject(nullptr)
, m_firstPageAllocatedDuringSweeping(nullptr)
, m_lastPageAllocatedDuringSweeping(nullptr)
, m_firstLargeObjectAllocatedDuringSweeping(nullptr)
, m_lastLargeObjectAllocatedDuringSweeping(nullptr)
, m_threadState(state)
, m_index(index)
, m_promptlyFreedSize(0)
{
clearFreeLists();
}
template<typename Header>
FreeList<Header>::FreeList()
: m_biggestFreeListIndex(0)
{
}
template<typename Header>
ThreadHeap<Header>::~ThreadHeap()
{
ASSERT(!m_firstPage);
ASSERT(!m_firstLargeObject);
}
template<typename Header>
void ThreadHeap<Header>::cleanupPages()
{
clearFreeLists();
// Add the ThreadHeap's pages to the orphanedPagePool.
for (HeapPage<Header>* page = m_firstPage; page; page = page->m_next) {
Heap::decreaseAllocatedSpace(blinkPageSize);
Heap::orphanedPagePool()->addOrphanedPage(m_index, page);
}
m_firstPage = nullptr;
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->m_next) {
Heap::decreaseAllocatedSpace(largeObject->size());
Heap::orphanedPagePool()->addOrphanedPage(m_index, largeObject);
}
m_firstLargeObject = nullptr;
}
template<typename Header>
void ThreadHeap<Header>::updateRemainingAllocationSize()
{
if (m_lastRemainingAllocationSize > remainingAllocationSize()) {
Heap::increaseAllocatedObjectSize(m_lastRemainingAllocationSize - remainingAllocationSize());
m_lastRemainingAllocationSize = remainingAllocationSize();
}
ASSERT(m_lastRemainingAllocationSize == remainingAllocationSize());
}
template<typename Header>
Address ThreadHeap<Header>::outOfLineAllocate(size_t allocationSize, const GCInfo* gcInfo)
{
ASSERT(allocationSize > remainingAllocationSize());
if (allocationSize > blinkPageSize / 2)
return allocateLargeObject(allocationSize, gcInfo);
updateRemainingAllocationSize();
threadState()->scheduleGCOrForceConservativeGCIfNeeded();
ASSERT(allocationSize >= allocationGranularity);
Address result = allocateFromFreeList(allocationSize, gcInfo);
if (result)
return result;
setAllocationPoint(nullptr, 0);
if (coalesce()) {
result = allocateFromFreeList(allocationSize, gcInfo);
if (result)
return result;
}
addPageToHeap(gcInfo);
result = allocateFromFreeList(allocationSize, gcInfo);
RELEASE_ASSERT(result);
return result;
}
static bool shouldUseFirstFitForHeap(int heapIndex)
{
// For an allocation of size N, should a heap perform a first-fit
// matching within the sized bin that N belongs to?
//
// Theory: quickly reusing a previously freed backing store block stands
// a chance of maintaining cached presence of that block (maintains
// "locality".) This is preferable to starting to bump allocate from a
// new and bigger block, which is what allocateFromFreeList() does by
// default. Hence, the backing store heaps are considered for binned
// first-fit matching.
//
// This appears to hold true through performance expermentation; at
// least no signficant performance regressions have been observed.
//
// This theory of improved performance does not hold true for other
// heap types. We are currently seeking an understanding of why;
// larger amounts of small block fragmentation might be one reason
// for it. TBC.
//
switch (heapIndex) {
case VectorBackingHeap:
case InlineVectorBackingHeap:
case HashTableBackingHeap:
return true;
default:
return false;
}
}
template<typename Header>
Address ThreadHeap<Header>::allocateFromFreeList(size_t allocationSize, const GCInfo* gcInfo)
{
// The freelist allocation scheme is currently as follows:
//
// - If the heap is of an appropriate type, try to pick the first
// entry from the sized bin corresponding to |allocationSize|.
// [See shouldUseFirstFitForHeap() comment for motivation on why.]
//
// - If that didn't satisfy the allocation, try reusing a block
// from the largest bin. The underlying reasoning being that
// we want to amortize this slow allocation call by carving
// off as a large a free block as possible in one go; a block
// that will service this block and let following allocations
// be serviced quickly by bump allocation.
//
// - Fail; allocation cannot be serviced by the freelist.
// The allocator will handle that failure by requesting more
// heap pages from the OS and re-initiate the allocation request.
//
int index = FreeList<Header>::bucketIndexForSize(allocationSize) + 1;
if (index <= m_freeList.m_biggestFreeListIndex && shouldUseFirstFitForHeap(m_index)) {
if (FreeListEntry* entry = m_freeList.m_freeLists[index]) {
entry->unlink(&m_freeList.m_freeLists[index]);
if (!m_freeList.m_freeLists[index] && index == m_freeList.m_biggestFreeListIndex) {
// Biggest bucket drained, adjust biggest index downwards.
int maxIndex = m_freeList.m_biggestFreeListIndex - 1;
for (; maxIndex >= 0 && !m_freeList.m_freeLists[maxIndex]; --maxIndex) { }
m_freeList.m_biggestFreeListIndex = maxIndex < 0 ? 0 : maxIndex;
}
// Allocate into the freelist block without disturbing the current allocation area.
ASSERT(entry->size() >= allocationSize);
if (entry->size() > allocationSize)
addToFreeList(entry->address() + allocationSize, entry->size() - allocationSize);
Heap::increaseAllocatedObjectSize(allocationSize);
return allocateAtAddress(entry->address(), allocationSize, gcInfo);
}
// Failed to find a first-fit freelist entry; fall into the standard case of
// chopping off the largest free block and bump allocate from it.
}
size_t bucketSize = 1 << m_freeList.m_biggestFreeListIndex;
index = m_freeList.m_biggestFreeListIndex;
for (; index > 0; --index, bucketSize >>= 1) {
FreeListEntry* entry = m_freeList.m_freeLists[index];
if (allocationSize > bucketSize) {
// Final bucket candidate; check initial entry if it is able
// to service this allocation. Do not perform a linear scan,
// as it is considered too costly.
if (!entry || entry->size() < allocationSize)
break;
}
if (entry) {
entry->unlink(&m_freeList.m_freeLists[index]);
setAllocationPoint(entry->address(), entry->size());
ASSERT(hasCurrentAllocationArea());
ASSERT(remainingAllocationSize() >= allocationSize);
m_freeList.m_biggestFreeListIndex = index;
return allocateSize(allocationSize, gcInfo);
}
}
m_freeList.m_biggestFreeListIndex = index;
return nullptr;
}
#if ENABLE(ASSERT)
template<typename Header>
static bool isLargeObjectAligned(LargeObject<Header>* largeObject, Address address)
{
// Check that a large object is blinkPageSize aligned (modulo the osPageSize
// for the guard page).
return reinterpret_cast<Address>(largeObject) - WTF::kSystemPageSize == roundToBlinkPageStart(reinterpret_cast<Address>(largeObject));
}
template<typename Header>
BaseHeapPage* ThreadHeap<Header>::findPageFromAddress(Address address)
{
for (HeapPage<Header>* page = m_firstPage; page; page = page->next()) {
if (page->contains(address))
return page;
}
for (HeapPage<Header>* page = m_firstPageAllocatedDuringSweeping; page; page = page->next()) {
if (page->contains(address))
return page;
}
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next()) {
ASSERT(isLargeObjectAligned(largeObject, address));
if (largeObject->contains(address))
return largeObject;
}
for (LargeObject<Header>* largeObject = m_firstLargeObjectAllocatedDuringSweeping; largeObject; largeObject = largeObject->next()) {
ASSERT(isLargeObjectAligned(largeObject, address));
if (largeObject->contains(address))
return largeObject;
}
return nullptr;
}
#endif
#if ENABLE(GC_PROFILE_MARKING)
template<typename Header>
const GCInfo* ThreadHeap<Header>::findGCInfoOfLargeObject(Address address)
{
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next()) {
if (largeObject->contains(address))
return largeObject->gcInfo();
}
return nullptr;
}
#endif
#if ENABLE(GC_PROFILE_HEAP)
#define GC_PROFILE_HEAP_PAGE_SNAPSHOT_THRESHOLD 0
template<typename Header>
void ThreadHeap<Header>::snapshot(TracedValue* json, ThreadState::SnapshotInfo* info)
{
ASSERT(isConsistentForSweeping());
size_t previousPageCount = info->pageCount;
json->beginArray("pages");
for (HeapPage<Header>* page = m_firstPage; page; page = page->next(), ++info->pageCount) {
// FIXME: To limit the size of the snapshot we only output "threshold" many page snapshots.
if (info->pageCount < GC_PROFILE_HEAP_PAGE_SNAPSHOT_THRESHOLD) {
json->beginArray();
json->pushInteger(reinterpret_cast<intptr_t>(page));
page->snapshot(json, info);
json->endArray();
} else {
page->snapshot(0, info);
}
}
json->endArray();
json->beginArray("largeObjects");
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next()) {
json->beginDictionary();
largeObject->snapshot(json, info);
json->endDictionary();
}
json->endArray();
json->setInteger("pageCount", info->pageCount - previousPageCount);
}
#endif
template<typename Header>
void FreeList<Header>::addToFreeList(Address address, size_t size)
{
ASSERT(size < blinkPagePayloadSize());
// The free list entries are only pointer aligned (but when we allocate
// from them we are 8 byte aligned due to the header size).
ASSERT(!((reinterpret_cast<uintptr_t>(address) + sizeof(Header)) & allocationMask));
ASSERT(!(size & allocationMask));
ASAN_POISON_MEMORY_REGION(address, size);
FreeListEntry* entry;
if (size < sizeof(*entry)) {
// Create a dummy header with only a size and freelist bit set.
ASSERT(size >= sizeof(HeapObjectHeader));
// Free list encode the size to mark the lost memory as freelist memory.
new (NotNull, address) HeapObjectHeader(HeapObjectHeader::freeListEncodedSize(size));
// This memory gets lost. Sweeping can reclaim it.
return;
}
entry = new (NotNull, address) FreeListEntry(size);
#if defined(ADDRESS_SANITIZER)
// For ASan we don't add the entry to the free lists until the
// asanDeferMemoryReuseCount reaches zero. However we always add entire
// pages to ensure that adding a new page will increase the allocation
// space.
if (HeapPage<Header>::payloadSize() != size && !entry->shouldAddToFreeList())
return;
#endif
int index = bucketIndexForSize(size);
entry->link(&m_freeLists[index]);
if (index > m_biggestFreeListIndex)
m_biggestFreeListIndex = index;
}
template<typename Header>
bool ThreadHeap<Header>::expandObject(Header* header, size_t newSize)
{
// It's possible that Vector requests a smaller expanded size because
// Vector::shrinkCapacity can set a capacity smaller than the actual payload
// size.
if (header->payloadSize() >= newSize)
return true;
size_t allocationSize = allocationSizeFromSize(newSize);
ASSERT(allocationSize > header->size());
size_t expandSize = allocationSize - header->size();
if (header->payloadEnd() == m_currentAllocationPoint && expandSize <= m_remainingAllocationSize) {
m_currentAllocationPoint += expandSize;
m_remainingAllocationSize -= expandSize;
// Unpoison the memory used for the object (payload).
ASAN_UNPOISON_MEMORY_REGION(header->payloadEnd(), expandSize);
FILL_ZERO_IF_NOT_PRODUCTION(header->payloadEnd(), expandSize);
header->setSize(allocationSize);
ASSERT(findPageFromAddress(header->payloadEnd() - 1));
return true;
}
return false;
}
template<typename Header>
void ThreadHeap<Header>::shrinkObject(Header* header, size_t newSize)
{
ASSERT(header->payloadSize() > newSize);
size_t allocationSize = allocationSizeFromSize(newSize);
ASSERT(header->size() > allocationSize);
size_t shrinkSize = header->size() - allocationSize;
if (header->payloadEnd() == m_currentAllocationPoint) {
m_currentAllocationPoint -= shrinkSize;
m_remainingAllocationSize += shrinkSize;
FILL_ZERO_IF_PRODUCTION(m_currentAllocationPoint, shrinkSize);
ASAN_POISON_MEMORY_REGION(m_currentAllocationPoint, shrinkSize);
header->setSize(allocationSize);
} else {
ASSERT(shrinkSize >= sizeof(HeapObjectHeader));
HeapObjectHeader* freedHeader = new (NotNull, header->payloadEnd() - shrinkSize) HeapObjectHeader(shrinkSize);
freedHeader->markPromptlyFreed();
ASSERT(pageFromObject(reinterpret_cast<Address>(header)) == findPageFromAddress(reinterpret_cast<Address>(header)));
m_promptlyFreedSize += shrinkSize;
header->setSize(allocationSize);
}
}
template<typename Header>
void ThreadHeap<Header>::promptlyFreeObject(Header* header)
{
ASSERT(!m_threadState->sweepForbidden());
header->checkHeader();
Address address = reinterpret_cast<Address>(header);
Address payload = header->payload();
size_t size = header->size();
size_t payloadSize = header->payloadSize();
ASSERT(size > 0);
ASSERT(pageFromObject(address) == findPageFromAddress(address));
{
ThreadState::SweepForbiddenScope forbiddenScope(m_threadState);
HeapObjectHeader::finalize(header->gcInfo(), payload, payloadSize);
if (address + size == m_currentAllocationPoint) {
m_currentAllocationPoint = address;
if (m_lastRemainingAllocationSize == m_remainingAllocationSize) {
Heap::decreaseAllocatedObjectSize(size);
m_lastRemainingAllocationSize += size;
}
m_remainingAllocationSize += size;
FILL_ZERO_IF_PRODUCTION(address, size);
ASAN_POISON_MEMORY_REGION(address, size);
return;
}
FILL_ZERO_IF_PRODUCTION(payload, payloadSize);
header->markPromptlyFreed();
}
m_promptlyFreedSize += size;
}
template<typename Header>
bool ThreadHeap<Header>::coalesce()
{
// Don't coalesce heaps if there are not enough promptly freed entries
// to be coalesced.
//
// FIXME: This threshold is determined just to optimize blink_perf
// benchmarks. Coalescing is very sensitive to the threashold and
// we need further investigations on the coalescing scheme.
if (m_promptlyFreedSize < 1024 * 1024)
return false;
if (m_threadState->sweepForbidden())
return false;
ASSERT(!hasCurrentAllocationArea());
TRACE_EVENT0("blink_gc", "ThreadHeap::coalesce");
// Rebuild free lists.
m_freeList.clear();
size_t freedSize = 0;
for (HeapPage<Header>* page = m_firstPage; page; page = page->next()) {
page->clearObjectStartBitMap();
Address startOfGap = page->payload();
for (Address headerAddress = startOfGap; headerAddress < page->end(); ) {
HeapObjectHeader* basicHeader = reinterpret_cast<HeapObjectHeader*>(headerAddress);
size_t size = basicHeader->size();
ASSERT(size > 0);
ASSERT(size < blinkPagePayloadSize());
if (basicHeader->isPromptlyFreed()) {
ASSERT(size >= sizeof(Header));
FILL_ZERO_IF_PRODUCTION(headerAddress, sizeof(Header));
freedSize += size;
headerAddress += size;
continue;
}
if (basicHeader->isFree()) {
// Zero the memory in the free list header to maintain the
// invariant that memory on the free list is zero filled.
// The rest of the memory is already on the free list and is
// therefore already zero filled.
FILL_ZERO_IF_PRODUCTION(headerAddress, size < sizeof(FreeListEntry) ? size : sizeof(FreeListEntry));
headerAddress += size;
continue;
}
if (startOfGap != headerAddress)
addToFreeList(startOfGap, headerAddress - startOfGap);
headerAddress += size;
startOfGap = headerAddress;
}
if (startOfGap != page->end())
addToFreeList(startOfGap, page->end() - startOfGap);
}
Heap::decreaseAllocatedObjectSize(freedSize);
ASSERT(m_promptlyFreedSize == freedSize);
m_promptlyFreedSize = 0;
return true;
}
template<typename Header>
Address ThreadHeap<Header>::allocateLargeObject(size_t size, const GCInfo* gcInfo)
{
// Caller already added space for object header and rounded up to allocation
// alignment
ASSERT(!(size & allocationMask));
size_t allocationSize = sizeof(LargeObject<Header>) + size;
// Ensure that there is enough space for alignment. If the header
// is not a multiple of 8 bytes we will allocate an extra
// headerPadding<Header> bytes to ensure it 8 byte aligned.
allocationSize += headerPadding<Header>();
// If ASan is supported we add allocationGranularity bytes to the allocated
// space and poison that to detect overflows
#if defined(ADDRESS_SANITIZER)
allocationSize += allocationGranularity;
#endif
updateRemainingAllocationSize();
m_threadState->scheduleGCOrForceConservativeGCIfNeeded();
m_threadState->shouldFlushHeapDoesNotContainCache();
PageMemory* pageMemory = PageMemory::allocate(allocationSize);
m_threadState->allocatedRegionsSinceLastGC().append(pageMemory->region());
Address largeObjectAddress = pageMemory->writableStart();
Address headerAddress = largeObjectAddress + sizeof(LargeObject<Header>) + headerPadding<Header>();
memset(headerAddress, 0, size);
Header* header = new (NotNull, headerAddress) Header(size, gcInfo);
Address result = headerAddress + sizeof(*header);
ASSERT(!(reinterpret_cast<uintptr_t>(result) & allocationMask));
LargeObject<Header>* largeObject = new (largeObjectAddress) LargeObject<Header>(pageMemory, gcInfo, threadState());
header->checkHeader();
// Poison the object header and allocationGranularity bytes after the object
ASAN_POISON_MEMORY_REGION(header, sizeof(*header));
ASAN_POISON_MEMORY_REGION(largeObject->address() + largeObject->size(), allocationGranularity);
// Use a separate list for large objects allocated during sweeping to make
// sure that we do not accidentally sweep objects that have been
// allocated during sweeping.
if (m_threadState->sweepForbidden()) {
if (!m_lastLargeObjectAllocatedDuringSweeping)
m_lastLargeObjectAllocatedDuringSweeping = largeObject;
largeObject->link(&m_firstLargeObjectAllocatedDuringSweeping);
} else {
largeObject->link(&m_firstLargeObject);
}
Heap::increaseAllocatedSpace(largeObject->size());
Heap::increaseAllocatedObjectSize(largeObject->size());
return result;
}
template<typename Header>
void ThreadHeap<Header>::freeLargeObject(LargeObject<Header>* object)
{
object->finalize();
Heap::decreaseAllocatedSpace(object->size());
// Unpoison the object header and allocationGranularity bytes after the
// object before freeing.
ASAN_UNPOISON_MEMORY_REGION(object->heapObjectHeader(), sizeof(Header));
ASAN_UNPOISON_MEMORY_REGION(object->address() + object->size(), allocationGranularity);
if (object->terminating()) {
ASSERT(ThreadState::current()->isTerminating());
// The thread is shutting down and this page is being removed as a part
// of the thread local GC. In that case the object could be traced in
// the next global GC if there is a dangling pointer from a live thread
// heap to this dead thread heap. To guard against this, we put the
// page into the orphaned page pool and zap the page memory. This
// ensures that tracing the dangling pointer in the next global GC just
// crashes instead of causing use-after-frees. After the next global
// GC, the orphaned pages are removed.
Heap::orphanedPagePool()->addOrphanedPage(m_index, object);
} else {
ASSERT(!ThreadState::current()->isTerminating());
PageMemory* memory = object->storage();
object->~LargeObject<Header>();
delete memory;
}
}
template<typename DataType>
PagePool<DataType>::PagePool()
{
for (int i = 0; i < NumberOfHeaps; ++i) {
m_pool[i] = nullptr;
}
}
FreePagePool::~FreePagePool()
{
for (int index = 0; index < NumberOfHeaps; ++index) {
while (PoolEntry* entry = m_pool[index]) {
m_pool[index] = entry->next;
PageMemory* memory = entry->data;
ASSERT(memory);
delete memory;
delete entry;
}
}
}
void FreePagePool::addFreePage(int index, PageMemory* memory)
{
// When adding a page to the pool we decommit it to ensure it is unused
// while in the pool. This also allows the physical memory, backing the
// page, to be given back to the OS.
memory->decommit();
MutexLocker locker(m_mutex[index]);
PoolEntry* entry = new PoolEntry(memory, m_pool[index]);
m_pool[index] = entry;
}
PageMemory* FreePagePool::takeFreePage(int index)
{
MutexLocker locker(m_mutex[index]);
while (PoolEntry* entry = m_pool[index]) {
m_pool[index] = entry->next;
PageMemory* memory = entry->data;
ASSERT(memory);
delete entry;
if (memory->commit())
return memory;
// We got some memory, but failed to commit it, try again.
delete memory;
}
return nullptr;
}
BaseHeapPage::BaseHeapPage(PageMemory* storage, const GCInfo* gcInfo, ThreadState* state)
: m_storage(storage)
, m_gcInfo(gcInfo)
, m_threadState(state)
, m_terminating(false)
{
ASSERT(isPageHeaderAddress(reinterpret_cast<Address>(this)));
}
void BaseHeapPage::markOrphaned()
{
m_threadState = nullptr;
m_gcInfo = nullptr;
m_terminating = false;
// Since we zap the page payload for orphaned pages we need to mark it as
// unused so a conservative pointer won't interpret the object headers.
storage()->markUnused();
}
OrphanedPagePool::~OrphanedPagePool()
{
for (int index = 0; index < NumberOfHeaps; ++index) {
while (PoolEntry* entry = m_pool[index]) {
m_pool[index] = entry->next;
BaseHeapPage* page = entry->data;
delete entry;
PageMemory* memory = page->storage();
ASSERT(memory);
page->~BaseHeapPage();
delete memory;
}
}
}
void OrphanedPagePool::addOrphanedPage(int index, BaseHeapPage* page)
{
page->markOrphaned();
PoolEntry* entry = new PoolEntry(page, m_pool[index]);
m_pool[index] = entry;
}
NO_SANITIZE_ADDRESS
void OrphanedPagePool::decommitOrphanedPages()
{
ASSERT(ThreadState::current()->isInGC());
#if ENABLE(ASSERT)
// No locking needed as all threads are at safepoints at this point in time.
for (ThreadState* state : ThreadState::attachedThreads())
ASSERT(state->isAtSafePoint());
#endif
for (int index = 0; index < NumberOfHeaps; ++index) {
PoolEntry* entry = m_pool[index];
PoolEntry** prevNext = &m_pool[index];
while (entry) {
BaseHeapPage* page = entry->data;
// Check if we should reuse the memory or just free it.
// Large object memory is not reused but freed, normal blink heap
// pages are reused.
// NOTE: We call the destructor before freeing or adding to the
// free page pool.
PageMemory* memory = page->storage();
if (page->isLargeObject()) {
page->~BaseHeapPage();
delete memory;
} else {
page->~BaseHeapPage();
clearMemory(memory);
Heap::freePagePool()->addFreePage(index, memory);
}
PoolEntry* deadEntry = entry;
entry = entry->next;
*prevNext = entry;
delete deadEntry;
}
}
}
NO_SANITIZE_ADDRESS
void OrphanedPagePool::clearMemory(PageMemory* memory)
{
#if defined(ADDRESS_SANITIZER)
// Don't use memset when running with ASan since this needs to zap
// poisoned memory as well and the NO_SANITIZE_ADDRESS annotation
// only works for code in this method and not for calls to memset.
Address base = memory->writableStart();
for (Address current = base; current < base + blinkPagePayloadSize(); ++current)
*current = 0;
#else
memset(memory->writableStart(), 0, blinkPagePayloadSize());
#endif
}
#if ENABLE(ASSERT)
bool OrphanedPagePool::contains(void* object)
{
for (int index = 0; index < NumberOfHeaps; ++index) {
for (PoolEntry* entry = m_pool[index]; entry; entry = entry->next) {
BaseHeapPage* page = entry->data;
if (page->contains(reinterpret_cast<Address>(object)))
return true;
}
}
return false;
}
#endif
template<>
void ThreadHeap<GeneralHeapObjectHeader>::addPageToHeap(const GCInfo* gcInfo)
{
// When adding a page to the ThreadHeap using GeneralHeapObjectHeaders the
// GCInfo on the heap should be unused (ie. nullptr).
allocatePage(nullptr);
}
template<>
void ThreadHeap<HeapObjectHeader>::addPageToHeap(const GCInfo* gcInfo)
{
// When adding a page to the ThreadHeap using HeapObjectHeaders store the
// GCInfo on the heap since it is the same for all objects
ASSERT(gcInfo);
allocatePage(gcInfo);
}
template <typename Header>
void ThreadHeap<Header>::freePage(HeapPage<Header>* page)
{
Heap::decreaseAllocatedSpace(blinkPageSize);
if (page->terminating()) {
// The thread is shutting down and this page is being removed as a part
// of the thread local GC. In that case the object could be traced in
// the next global GC if there is a dangling pointer from a live thread
// heap to this dead thread heap. To guard against this, we put the
// page into the orphaned page pool and zap the page memory. This
// ensures that tracing the dangling pointer in the next global GC just
// crashes instead of causing use-after-frees. After the next global
// GC, the orphaned pages are removed.
Heap::orphanedPagePool()->addOrphanedPage(m_index, page);
} else {
PageMemory* memory = page->storage();
page->~HeapPage<Header>();
Heap::freePagePool()->addFreePage(m_index, memory);
}
}
template<typename Header>
void ThreadHeap<Header>::allocatePage(const GCInfo* gcInfo)
{
m_threadState->shouldFlushHeapDoesNotContainCache();
PageMemory* pageMemory = Heap::freePagePool()->takeFreePage(m_index);
// We continue allocating page memory until we succeed in committing one.
while (!pageMemory) {
// Allocate a memory region for blinkPagesPerRegion pages that
// will each have the following layout.
//
// [ guard os page | ... payload ... | guard os page ]
// ^---{ aligned to blink page size }
PageMemoryRegion* region = PageMemoryRegion::allocateNormalPages();
m_threadState->allocatedRegionsSinceLastGC().append(region);
// Setup the PageMemory object for each of the pages in the region.
size_t offset = 0;
for (size_t i = 0; i < blinkPagesPerRegion; ++i) {
PageMemory* memory = PageMemory::setupPageMemoryInRegion(region, offset, blinkPagePayloadSize());
// Take the first possible page ensuring that this thread actually
// gets a page and add the rest to the page pool.
if (!pageMemory) {
if (memory->commit())
pageMemory = memory;
else
delete memory;
} else {
Heap::freePagePool()->addFreePage(m_index, memory);
}
offset += blinkPageSize;
}
}
HeapPage<Header>* page = new (pageMemory->writableStart()) HeapPage<Header>(pageMemory, this, gcInfo);
// Use a separate list for pages allocated during sweeping to make
// sure that we do not accidentally sweep objects that have been
// allocated during sweeping.
if (m_threadState->sweepForbidden()) {
if (!m_lastPageAllocatedDuringSweeping)
m_lastPageAllocatedDuringSweeping = page;
page->link(&m_firstPageAllocatedDuringSweeping);
} else {
page->link(&m_firstPage);
}
Heap::increaseAllocatedSpace(blinkPageSize);
addToFreeList(page->payload(), HeapPage<Header>::payloadSize());
}
#if ENABLE(ASSERT)
template<typename Header>
bool ThreadHeap<Header>::pagesToBeSweptContains(Address address)
{
for (HeapPage<Header>* page = m_firstPage; page; page = page->next()) {
if (page->contains(address))
return true;
}
return false;
}
template<typename Header>
bool ThreadHeap<Header>::pagesAllocatedDuringSweepingContains(Address address)
{
for (HeapPage<Header>* page = m_firstPageAllocatedDuringSweeping; page; page = page->next()) {
if (page->contains(address))
return true;
}
return false;
}
#endif
template<typename Header>
size_t ThreadHeap<Header>::objectPayloadSizeForTesting()
{
ASSERT(isConsistentForSweeping());
ASSERT(!m_firstPageAllocatedDuringSweeping);
ASSERT(!m_firstLargeObjectAllocatedDuringSweeping);
size_t objectPayloadSize = 0;
for (HeapPage<Header>* page = m_firstPage; page; page = page->next())
objectPayloadSize += page->objectPayloadSizeForTesting();
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next())
objectPayloadSize += largeObject->objectPayloadSizeForTesting();
return objectPayloadSize;
}
template<typename Header>
void ThreadHeap<Header>::sweepNormalPages()
{
TRACE_EVENT0("blink_gc", "ThreadHeap::sweepNormalPages");
HeapPage<Header>* page = m_firstPage;
HeapPage<Header>** previousNext = &m_firstPage;
while (page) {
if (page->isEmpty()) {
HeapPage<Header>* next = page->next();
page->unlink(previousNext);
freePage(page);
page = next;
} else {
page->sweep(this);
previousNext = &page->m_next;
page = page->next();
}
}
}
template<typename Header>
void ThreadHeap<Header>::sweepLargePages()
{
TRACE_EVENT0("blink_gc", "ThreadHeap::sweepLargePages");
LargeObject<Header>* largeObject = m_firstLargeObject;
LargeObject<Header>** previousNext = &m_firstLargeObject;
while (largeObject) {
if (largeObject->isEmpty()) {
LargeObject<Header>* next = largeObject->next();
largeObject->unlink(previousNext);
freeLargeObject(largeObject);
largeObject = next;
} else {
largeObject->sweep();
previousNext = &largeObject->m_next;
largeObject = largeObject->next();
}
}
}
// STRICT_ASAN_FINALIZATION_CHECKING turns on poisoning of all objects during
// sweeping to catch cases where dead objects touch each other. This is not
// turned on by default because it also triggers for cases that are safe.
// Examples of such safe cases are context life cycle observers and timers
// embedded in garbage collected objects.
#define STRICT_ASAN_FINALIZATION_CHECKING 0
template<typename Header>
void ThreadHeap<Header>::sweep()
{
ASSERT(isConsistentForSweeping());
#if defined(ADDRESS_SANITIZER) && STRICT_ASAN_FINALIZATION_CHECKING
// When using ASan do a pre-sweep where all unmarked objects are
// poisoned before calling their finalizer methods. This can catch
// the case where the finalizer of an object tries to modify
// another object as part of finalization.
for (HeapPage<Header>* page = m_firstPage; page; page = page->next())
page->poisonUnmarkedObjects();
#endif
sweepNormalPages();
sweepLargePages();
}
template<typename Header>
void ThreadHeap<Header>::postSweepProcessing()
{
// If pages have been allocated during sweeping, link them into
// the list of pages.
if (m_firstPageAllocatedDuringSweeping) {
m_lastPageAllocatedDuringSweeping->m_next = m_firstPage;
m_firstPage = m_firstPageAllocatedDuringSweeping;
m_lastPageAllocatedDuringSweeping = nullptr;
m_firstPageAllocatedDuringSweeping = nullptr;
}
if (m_firstLargeObjectAllocatedDuringSweeping) {
m_lastLargeObjectAllocatedDuringSweeping->m_next = m_firstLargeObject;
m_firstLargeObject = m_firstLargeObjectAllocatedDuringSweeping;
m_lastLargeObjectAllocatedDuringSweeping = nullptr;
m_firstLargeObjectAllocatedDuringSweeping = nullptr;
}
}
#if ENABLE(ASSERT)
template<typename Header>
bool ThreadHeap<Header>::isConsistentForSweeping()
{
// A thread heap is consistent for sweeping if none of the pages to be swept
// contain a freelist block or the current allocation point.
for (size_t i = 0; i < blinkPageSizeLog2; ++i) {
for (FreeListEntry* freeListEntry = m_freeList.m_freeLists[i]; freeListEntry; freeListEntry = freeListEntry->next()) {
if (pagesToBeSweptContains(freeListEntry->address()))
return false;
ASSERT(pagesAllocatedDuringSweepingContains(freeListEntry->address()));
}
}
if (hasCurrentAllocationArea()) {
if (pagesToBeSweptContains(currentAllocationPoint()))
return false;
ASSERT(pagesAllocatedDuringSweepingContains(currentAllocationPoint()));
}
return true;
}
#endif
template<typename Header>
void ThreadHeap<Header>::makeConsistentForSweeping()
{
setAllocationPoint(nullptr, 0);
clearFreeLists();
}
template<typename Header>
void ThreadHeap<Header>::markUnmarkedObjectsDead()
{
ASSERT(isConsistentForSweeping());
for (HeapPage<Header>* page = m_firstPage; page; page = page->next()) {
page->markUnmarkedObjectsDead();
}
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next()) {
largeObject->markUnmarkedObjectsDead();
}
}
template<typename Header>
void ThreadHeap<Header>::clearFreeLists()
{
m_promptlyFreedSize = 0;
m_freeList.clear();
}
template<typename Header>
void FreeList<Header>::clear()
{
m_biggestFreeListIndex = 0;
for (size_t i = 0; i < blinkPageSizeLog2; ++i)
m_freeLists[i] = nullptr;
}
template<typename Header>
int FreeList<Header>::bucketIndexForSize(size_t size)
{
ASSERT(size > 0);
int index = -1;
while (size) {
size >>= 1;
index++;
}
return index;
}
template<typename Header>
HeapPage<Header>::HeapPage(PageMemory* storage, ThreadHeap<Header>* heap, const GCInfo* gcInfo)
: BaseHeapPage(storage, gcInfo, heap->threadState())
, m_next(nullptr)
{
static_assert(!(sizeof(HeapPage<Header>) & allocationMask), "page header incorrectly aligned");
m_objectStartBitMapComputed = false;
ASSERT(isPageHeaderAddress(reinterpret_cast<Address>(this)));
}
template<typename Header>
size_t HeapPage<Header>::objectPayloadSizeForTesting()
{
size_t objectPayloadSize = 0;
Address headerAddress = payload();
ASSERT(headerAddress != end());
do {
Header* header = reinterpret_cast<Header*>(headerAddress);
if (!header->isFree()) {
header->checkHeader();
objectPayloadSize += header->payloadSize();
}
ASSERT(header->size() < blinkPagePayloadSize());
headerAddress += header->size();
ASSERT(headerAddress <= end());
} while (headerAddress < end());
return objectPayloadSize;
}
template<typename Header>
bool HeapPage<Header>::isEmpty()
{
HeapObjectHeader* header = reinterpret_cast<HeapObjectHeader*>(payload());
return header->isFree() && header->size() == payloadSize();
}
template<typename Header>
void HeapPage<Header>::sweep(ThreadHeap<Header>* heap)
{
clearObjectStartBitMap();
Address startOfGap = payload();
for (Address headerAddress = startOfGap; headerAddress < end(); ) {
HeapObjectHeader* basicHeader = reinterpret_cast<HeapObjectHeader*>(headerAddress);
ASSERT(basicHeader->size() > 0);
ASSERT(basicHeader->size() < blinkPagePayloadSize());
if (basicHeader->isFree()) {
size_t size = basicHeader->size();
// Zero the memory in the free list header to maintain the
// invariant that memory on the free list is zero filled.
// The rest of the memory is already on the free list and is
// therefore already zero filled.
FILL_ZERO_IF_PRODUCTION(headerAddress, size < sizeof(FreeListEntry) ? size : sizeof(FreeListEntry));
headerAddress += size;
continue;
}
// At this point we know this is a valid object of type Header
Header* header = static_cast<Header*>(basicHeader);
header->checkHeader();
if (!header->isMarked()) {
// For ASan we unpoison the specific object when calling the
// finalizer and poison it again when done to allow the object's own
// finalizer to operate on the object, but not have other finalizers
// be allowed to access it.
ASAN_UNPOISON_MEMORY_REGION(header->payload(), header->payloadSize());
finalize(header);
size_t size = header->size();
// This memory will be added to the freelist. Maintain the invariant
// that memory on the freelist is zero filled.
FILL_ZERO_IF_PRODUCTION(headerAddress, size);
ASAN_POISON_MEMORY_REGION(header->payload(), header->payloadSize());
headerAddress += size;
continue;
}
if (startOfGap != headerAddress)
heap->addToFreeList(startOfGap, headerAddress - startOfGap);
header->unmark();
headerAddress += header->size();
Heap::increaseMarkedObjectSize(header->size());
startOfGap = headerAddress;
}
if (startOfGap != end())
heap->addToFreeList(startOfGap, end() - startOfGap);
}
template<typename Header>
void HeapPage<Header>::markUnmarkedObjectsDead()
{
for (Address headerAddress = payload(); headerAddress < end();) {
Header* header = reinterpret_cast<Header*>(headerAddress);
ASSERT(header->size() < blinkPagePayloadSize());
// Check if a free list entry first since we cannot call
// isMarked on a free list entry.
if (header->isFree()) {
headerAddress += header->size();
continue;
}
if (header->isMarked())
header->unmark();
else
header->markDead();
headerAddress += header->size();
}
}
template<typename Header>
void HeapPage<Header>::populateObjectStartBitMap()
{
memset(&m_objectStartBitMap, 0, objectStartBitMapSize);
Address start = payload();
for (Address headerAddress = start; headerAddress < end();) {
Header* header = reinterpret_cast<Header*>(headerAddress);
size_t objectOffset = headerAddress - start;
ASSERT(!(objectOffset & allocationMask));
size_t objectStartNumber = objectOffset / allocationGranularity;
size_t mapIndex = objectStartNumber / 8;
ASSERT(mapIndex < objectStartBitMapSize);
m_objectStartBitMap[mapIndex] |= (1 << (objectStartNumber & 7));
headerAddress += header->size();
ASSERT(headerAddress <= end());
}
m_objectStartBitMapComputed = true;
}
template<typename Header>
void HeapPage<Header>::clearObjectStartBitMap()
{
m_objectStartBitMapComputed = false;
}
static int numberOfLeadingZeroes(uint8_t byte)
{
if (!byte)
return 8;
int result = 0;
if (byte <= 0x0F) {
result += 4;
byte = byte << 4;
}
if (byte <= 0x3F) {
result += 2;
byte = byte << 2;
}
if (byte <= 0x7F)
result++;
return result;
}
template<typename Header>
Header* HeapPage<Header>::findHeaderFromAddress(Address address)
{
if (address < payload())
return nullptr;
if (!isObjectStartBitMapComputed())
populateObjectStartBitMap();
size_t objectOffset = address - payload();
size_t objectStartNumber = objectOffset / allocationGranularity;
size_t mapIndex = objectStartNumber / 8;
ASSERT(mapIndex < objectStartBitMapSize);
size_t bit = objectStartNumber & 7;
uint8_t byte = m_objectStartBitMap[mapIndex] & ((1 << (bit + 1)) - 1);
while (!byte) {
ASSERT(mapIndex > 0);
byte = m_objectStartBitMap[--mapIndex];
}
int leadingZeroes = numberOfLeadingZeroes(byte);
objectStartNumber = (mapIndex * 8) + 7 - leadingZeroes;
objectOffset = objectStartNumber * allocationGranularity;
Address objectAddress = objectOffset + payload();
Header* header = reinterpret_cast<Header*>(objectAddress);
if (header->isFree())
return nullptr;
header->checkHeader();
return header;
}
template<typename Header>
void HeapPage<Header>::checkAndMarkPointer(Visitor* visitor, Address address)
{
ASSERT(contains(address));
Header* header = findHeaderFromAddress(address);
if (!header || header->isDead())
return;
#if ENABLE(GC_PROFILE_MARKING)
visitor->setHostInfo(&address, "stack");
#endif
if (hasVTable(header) && !vTableInitialized(header->payload())) {
visitor->markHeaderNoTracing(header);
ASSERT(isUninitializedMemory(header->payload(), header->payloadSize()));
} else {
visitor->markHeader(header, traceCallback(header));
}
}
#if ENABLE(GC_PROFILE_MARKING)
template<typename Header>
const GCInfo* HeapPage<Header>::findGCInfo(Address address)
{
if (address < payload())
return nullptr;
if (gcInfo()) // For non GeneralHeapObjects.
return gcInfo();
Header* header = findHeaderFromAddress(address);
if (!header)
return nullptr;
return header->gcInfo();
}
#endif
#if ENABLE(GC_PROFILE_HEAP)
template<typename Header>
void HeapPage<Header>::snapshot(TracedValue* json, ThreadState::SnapshotInfo* info)
{
Header* header = nullptr;
for (Address addr = payload(); addr < end(); addr += header->size()) {
header = reinterpret_cast<Header*>(addr);
if (json)
json->pushInteger(header->encodedSize());
if (header->isFree()) {
info->freeSize += header->size();
continue;
}
const GCInfo* gcinfo = header->gcInfo() ? header->gcInfo() : gcInfo();
size_t tag = info->getClassTag(gcinfo);
size_t age = header->age();
if (json)
json->pushInteger(tag);
if (header->isMarked()) {
info->liveCount[tag] += 1;
info->liveSize[tag] += header->size();
// Count objects that are live when promoted to the final generation.
if (age == maxHeapObjectAge - 1)
info->generations[tag][maxHeapObjectAge] += 1;
header->incAge();
} else {
info->deadCount[tag] += 1;
info->deadSize[tag] += header->size();
// Count objects that are dead before the final generation.
if (age < maxHeapObjectAge)
info->generations[tag][age] += 1;
}
}
}
#endif
#if defined(ADDRESS_SANITIZER)
template<typename Header>
void HeapPage<Header>::poisonUnmarkedObjects()
{
for (Address headerAddress = payload(); headerAddress < end(); ) {
Header* header = reinterpret_cast<Header*>(headerAddress);
ASSERT(header->size() < blinkPagePayloadSize());
if (!header->isFree() && !header->isMarked())
ASAN_POISON_MEMORY_REGION(header->payload(), header->payloadSize());
headerAddress += header->size();
}
}
#endif
template<>
inline void HeapPage<GeneralHeapObjectHeader>::finalize(GeneralHeapObjectHeader* header)
{
header->finalize();
}
template<>
inline void HeapPage<HeapObjectHeader>::finalize(HeapObjectHeader* header)
{
ASSERT(gcInfo());
HeapObjectHeader::finalize(gcInfo(), header->payload(), header->payloadSize());
}
template<>
inline TraceCallback HeapPage<HeapObjectHeader>::traceCallback(HeapObjectHeader* header)
{
ASSERT(gcInfo());
return gcInfo()->m_trace;
}
template<>
inline TraceCallback HeapPage<GeneralHeapObjectHeader>::traceCallback(GeneralHeapObjectHeader* header)
{
return header->traceCallback();
}
template<>
inline bool HeapPage<HeapObjectHeader>::hasVTable(HeapObjectHeader* header)
{
ASSERT(gcInfo());
return gcInfo()->hasVTable();
}
template<>
inline bool HeapPage<GeneralHeapObjectHeader>::hasVTable(GeneralHeapObjectHeader* header)
{
return header->hasVTable();
}
template<typename Header>
size_t LargeObject<Header>::objectPayloadSizeForTesting()
{
return payloadSize();
}
#if ENABLE(GC_PROFILE_HEAP)
template<typename Header>
void LargeObject<Header>::snapshot(TracedValue* json, ThreadState::SnapshotInfo* info)
{
Header* header = heapObjectHeader();
size_t tag = info->getClassTag(header->gcInfo());
size_t age = header->age();
if (header->isMarked()) {
info->liveCount[tag] += 1;
info->liveSize[tag] += header->size();
// Count objects that are live when promoted to the final generation.
if (age == maxHeapObjectAge - 1)
info->generations[tag][maxHeapObjectAge] += 1;
header->incAge();
} else {
info->deadCount[tag] += 1;
info->deadSize[tag] += header->size();
// Count objects that are dead before the final generation.
if (age < maxHeapObjectAge)
info->generations[tag][age] += 1;
}
if (json) {
json->setInteger("class", tag);
json->setInteger("size", header->size());
json->setInteger("isMarked", header->isMarked());
}
}
#endif
void HeapDoesNotContainCache::flush()
{
if (m_hasEntries) {
for (int i = 0; i < numberOfEntries; ++i)
m_entries[i] = nullptr;
m_hasEntries = false;
}
}
size_t HeapDoesNotContainCache::hash(Address address)
{
size_t value = (reinterpret_cast<size_t>(address) >> blinkPageSizeLog2);
value ^= value >> numberOfEntriesLog2;
value ^= value >> (numberOfEntriesLog2 * 2);
value &= numberOfEntries - 1;
return value & ~1; // Returns only even number.
}
bool HeapDoesNotContainCache::lookup(Address address)
{
ASSERT(ThreadState::current()->isInGC());
size_t index = hash(address);
ASSERT(!(index & 1));
Address cachePage = roundToBlinkPageStart(address);
if (m_entries[index] == cachePage)
return m_entries[index];
if (m_entries[index + 1] == cachePage)
return m_entries[index + 1];
return false;
}
void HeapDoesNotContainCache::addEntry(Address address)
{
ASSERT(ThreadState::current()->isInGC());
m_hasEntries = true;
size_t index = hash(address);
ASSERT(!(index & 1));
Address cachePage = roundToBlinkPageStart(address);
m_entries[index + 1] = m_entries[index];
m_entries[index] = cachePage;
}
void Heap::flushHeapDoesNotContainCache()
{
s_heapDoesNotContainCache->flush();
}
enum MarkingMode {
GlobalMarking,
ThreadLocalMarking,
};
template <MarkingMode Mode>
class MarkingVisitor final : public Visitor, public MarkingVisitorImpl<MarkingVisitor<Mode>> {
public:
using Impl = MarkingVisitorImpl<MarkingVisitor<Mode>>;
friend class MarkingVisitorImpl<MarkingVisitor<Mode>>;
#if ENABLE(GC_PROFILE_MARKING)
using LiveObjectSet = HashSet<uintptr_t>;
using LiveObjectMap = HashMap<String, LiveObjectSet>;
using ObjectGraph = HashMap<uintptr_t, std::pair<uintptr_t, String>>;
#endif
MarkingVisitor()
: Visitor(Mode == GlobalMarking ? Visitor::GlobalMarkingVisitorType : Visitor::GenericVisitorType)
{
}
// We need both HeapObjectHeader and GeneralHeapObjectHeader versions to
// correctly find the payload.
virtual void markHeader(HeapObjectHeader* header, TraceCallback callback) override
{
Impl::visitHeader(header, header->payload(), callback);
}
virtual void markHeader(GeneralHeapObjectHeader* header, TraceCallback callback) override
{
Impl::visitHeader(header, header->payload(), callback);
}
virtual void mark(const void* objectPointer, TraceCallback callback) override
{
Impl::mark(objectPointer, callback);
}
virtual void registerDelayedMarkNoTracing(const void* object) override
{
Impl::registerDelayedMarkNoTracing(object);
}
virtual void registerWeakMembers(const void* closure, const void* objectPointer, WeakPointerCallback callback) override
{
Impl::registerWeakMembers(closure, objectPointer, callback);
}
virtual void registerWeakTable(const void* closure, EphemeronCallback iterationCallback, EphemeronCallback iterationDoneCallback)
{
Impl::registerWeakTable(closure, iterationCallback, iterationDoneCallback);
}
#if ENABLE(ASSERT)
virtual bool weakTableRegistered(const void* closure)
{
return Impl::weakTableRegistered(closure);
}
#endif
virtual bool isMarked(const void* objectPointer) override
{
return Impl::isMarked(objectPointer);
}
virtual bool ensureMarked(const void* objectPointer) override
{
return Impl::ensureMarked(objectPointer);
}
// This macro defines the necessary visitor methods for typed heaps
#define DEFINE_VISITOR_METHODS(Type) \
virtual void mark(const Type* objectPointer, TraceCallback callback) override \
{ \
Impl::mark(objectPointer, callback); \
} \
virtual bool isMarked(const Type* objectPointer) override \
{ \
return Impl::isMarked(objectPointer); \
} \
virtual bool ensureMarked(const Type* objectPointer) override \
{ \
static_assert(!NeedsAdjustAndMark<Type>::value, "ensureMarked can only be used on non adjusted types"); \
return Impl::ensureMarked(objectPointer); \
}
FOR_EACH_TYPED_HEAP(DEFINE_VISITOR_METHODS)
#undef DEFINE_VISITOR_METHODS
#if ENABLE(GC_PROFILE_MARKING)
virtual void recordObjectGraphEdge(const void* objectPointer) override
{
MutexLocker locker(objectGraphMutex());
String className(classOf(objectPointer));
{
LiveObjectMap::AddResult result = currentlyLive().add(className, LiveObjectSet());
result.storedValue->value.add(reinterpret_cast<uintptr_t>(objectPointer));
}
ObjectGraph::AddResult result = objectGraph().add(reinterpret_cast<uintptr_t>(objectPointer), std::make_pair(reinterpret_cast<uintptr_t>(m_hostObject), m_hostName));
ASSERT(result.isNewEntry);
// fprintf(stderr, "%s[%p] -> %s[%p]\n", m_hostName.ascii().data(), m_hostObject, className.ascii().data(), objectPointer);
}
void reportStats()
{
fprintf(stderr, "\n---------- AFTER MARKING -------------------\n");
for (LiveObjectMap::iterator it = currentlyLive().begin(), end = currentlyLive().end(); it != end; ++it) {
fprintf(stderr, "%s %u", it->key.ascii().data(), it->value.size());
if (it->key == "blink::Document")
reportStillAlive(it->value, previouslyLive().get(it->key));
fprintf(stderr, "\n");
}
previouslyLive().swap(currentlyLive());
currentlyLive().clear();
for (uintptr_t object : objectsToFindPath()) {
dumpPathToObjectFromObjectGraph(objectGraph(), object);
}
}
static void reportStillAlive(LiveObjectSet current, LiveObjectSet previous)
{
int count = 0;
fprintf(stderr, " [previously %u]", previous.size());
for (uintptr_t object : current) {
if (previous.find(object) == previous.end())
continue;
count++;
}
if (!count)
return;
fprintf(stderr, " {survived 2GCs %d: ", count);
for (uintptr_t object : current) {
if (previous.find(object) == previous.end())
continue;
fprintf(stderr, "%ld", object);
if (--count)
fprintf(stderr, ", ");
}
ASSERT(!count);
fprintf(stderr, "}");
}
static void dumpPathToObjectFromObjectGraph(const ObjectGraph& graph, uintptr_t target)
{
ObjectGraph::const_iterator it = graph.find(target);
if (it == graph.end())
return;
fprintf(stderr, "Path to %lx of %s\n", target, classOf(reinterpret_cast<const void*>(target)).ascii().data());
while (it != graph.end()) {
fprintf(stderr, "<- %lx of %s\n", it->value.first, it->value.second.utf8().data());
it = graph.find(it->value.first);
}
fprintf(stderr, "\n");
}
static void dumpPathToObjectOnNextGC(void* p)
{
objectsToFindPath().add(reinterpret_cast<uintptr_t>(p));
}
static Mutex& objectGraphMutex()
{
AtomicallyInitializedStatic(Mutex&, mutex = *new Mutex);
return mutex;
}
static LiveObjectMap& previouslyLive()
{
DEFINE_STATIC_LOCAL(LiveObjectMap, map, ());
return map;
}
static LiveObjectMap& currentlyLive()
{
DEFINE_STATIC_LOCAL(LiveObjectMap, map, ());
return map;
}
static ObjectGraph& objectGraph()
{
DEFINE_STATIC_LOCAL(ObjectGraph, graph, ());
return graph;
}
static HashSet<uintptr_t>& objectsToFindPath()
{
DEFINE_STATIC_LOCAL(HashSet<uintptr_t>, set, ());
return set;
}
#endif
protected:
virtual void registerWeakCellWithCallback(void** cell, WeakPointerCallback callback) override
{
Impl::registerWeakCellWithCallback(cell, callback);
}
inline bool shouldMarkObject(const void* objectPointer)
{
if (Mode != ThreadLocalMarking)
return true;
BaseHeapPage* page = pageFromObject(objectPointer);
ASSERT(!page->orphaned());
// When doing a thread local GC, the marker checks if
// the object resides in another thread's heap. If it
// does, the object should not be marked & traced.
return page->terminating();
}
};
void Heap::init()
{
ThreadState::init();
s_markingStack = new CallbackStack();
s_postMarkingCallbackStack = new CallbackStack();
s_weakCallbackStack = new CallbackStack();
s_ephemeronStack = new CallbackStack();
s_heapDoesNotContainCache = new HeapDoesNotContainCache();
s_markingVisitor = new MarkingVisitor<GlobalMarking>();
s_freePagePool = new FreePagePool();
s_orphanedPagePool = new OrphanedPagePool();
s_allocatedObjectSize = 0;
s_allocatedSpace = 0;
s_markedObjectSize = 0;
}
void Heap::shutdown()
{
s_shutdownCalled = true;
ThreadState::shutdownHeapIfNecessary();
}
void Heap::doShutdown()
{
// We don't want to call doShutdown() twice.
if (!s_markingVisitor)
return;
ASSERT(!ThreadState::attachedThreads().size());
delete s_markingVisitor;
s_markingVisitor = nullptr;
delete s_heapDoesNotContainCache;
s_heapDoesNotContainCache = nullptr;
delete s_freePagePool;
s_freePagePool = nullptr;
delete s_orphanedPagePool;
s_orphanedPagePool = nullptr;
delete s_weakCallbackStack;
s_weakCallbackStack = nullptr;
delete s_postMarkingCallbackStack;
s_postMarkingCallbackStack = nullptr;
delete s_markingStack;
s_markingStack = nullptr;
delete s_ephemeronStack;
s_ephemeronStack = nullptr;
delete s_regionTree;
s_regionTree = nullptr;
ThreadState::shutdown();
ASSERT(Heap::allocatedSpace() == 0);
}
#if ENABLE(ASSERT)
BaseHeapPage* Heap::findPageFromAddress(Address address)
{
ASSERT(ThreadState::current()->isInGC());
for (ThreadState* state : ThreadState::attachedThreads()) {
if (BaseHeapPage* page = state->findPageFromAddress(address))
return page;
}
return nullptr;
}
bool Heap::containedInHeapOrOrphanedPage(void* object)
{
return findPageFromAddress(object) || orphanedPagePool()->contains(object);
}
#endif
Address Heap::checkAndMarkPointer(Visitor* visitor, Address address)
{
ASSERT(ThreadState::current()->isInGC());
#if !ENABLE(ASSERT)
if (s_heapDoesNotContainCache->lookup(address))
return nullptr;
#endif
if (BaseHeapPage* page = lookup(address)) {
ASSERT(page->contains(address));
ASSERT(!page->orphaned());
ASSERT(!s_heapDoesNotContainCache->lookup(address));
page->checkAndMarkPointer(visitor, address);
// FIXME: We only need to set the conservative flag if
// checkAndMarkPointer actually marked the pointer.
s_lastGCWasConservative = true;
return address;
}
#if !ENABLE(ASSERT)
s_heapDoesNotContainCache->addEntry(address);
#else
if (!s_heapDoesNotContainCache->lookup(address))
s_heapDoesNotContainCache->addEntry(address);
#endif
return nullptr;
}
#if ENABLE(GC_PROFILE_MARKING)
const GCInfo* Heap::findGCInfo(Address address)
{
return ThreadState::findGCInfoFromAllThreads(address);
}
#endif
#if ENABLE(GC_PROFILE_MARKING)
void Heap::dumpPathToObjectOnNextGC(void* p)
{
static_cast<MarkingVisitor<GlobalMarking>*>(s_markingVisitor)->dumpPathToObjectOnNextGC(p);
}
String Heap::createBacktraceString()
{
int framesToShow = 3;
int stackFrameSize = 16;
ASSERT(stackFrameSize >= framesToShow);
using FramePointer = void*;
FramePointer* stackFrame = static_cast<FramePointer*>(alloca(sizeof(FramePointer) * stackFrameSize));
WTFGetBacktrace(stackFrame, &stackFrameSize);
StringBuilder builder;
builder.append("Persistent");
bool didAppendFirstName = false;
// Skip frames before/including "blink::Persistent".
bool didSeePersistent = false;
for (int i = 0; i < stackFrameSize && framesToShow > 0; ++i) {
FrameToNameScope frameToName(stackFrame[i]);
if (!frameToName.nullableName())
continue;
if (strstr(frameToName.nullableName(), "blink::Persistent")) {
didSeePersistent = true;
continue;
}
if (!didSeePersistent)
continue;
if (!didAppendFirstName) {
didAppendFirstName = true;
builder.append(" ... Backtrace:");
}
builder.append("\n\t");
builder.append(frameToName.nullableName());
--framesToShow;
}
return builder.toString().replace("blink::", "");
}
#endif
void Heap::pushTraceCallback(void* object, TraceCallback callback)
{
ASSERT(Heap::containedInHeapOrOrphanedPage(object));
CallbackStack::Item* slot = s_markingStack->allocateEntry();
*slot = CallbackStack::Item(object, callback);
}
bool Heap::popAndInvokeTraceCallback(CallbackStack* stack, Visitor* visitor)
{
CallbackStack::Item* item = stack->pop();
if (!item)
return false;
#if ENABLE(GC_PROFILE_MARKING)
visitor->setHostInfo(item->object(), classOf(item->object()));
#endif
item->call(visitor);
return true;
}
void Heap::pushPostMarkingCallback(void* object, TraceCallback callback)
{
ASSERT(!Heap::orphanedPagePool()->contains(object));
CallbackStack::Item* slot = s_postMarkingCallbackStack->allocateEntry();
*slot = CallbackStack::Item(object, callback);
}
bool Heap::popAndInvokePostMarkingCallback(Visitor* visitor)
{
if (CallbackStack::Item* item = s_postMarkingCallbackStack->pop()) {
item->call(visitor);
return true;
}
return false;
}
void Heap::pushWeakCellPointerCallback(void** cell, WeakPointerCallback callback)
{
ASSERT(!Heap::orphanedPagePool()->contains(cell));
CallbackStack::Item* slot = s_weakCallbackStack->allocateEntry();
*slot = CallbackStack::Item(cell, callback);
}
void Heap::pushWeakPointerCallback(void* closure, void* object, WeakPointerCallback callback)
{
BaseHeapPage* page = pageFromObject(object);
ASSERT(!page->orphaned());
ThreadState* state = page->threadState();
state->pushWeakPointerCallback(closure, callback);
}
bool Heap::popAndInvokeWeakPointerCallback(Visitor* visitor)
{
// For weak processing we should never reach orphaned pages since orphaned
// pages are not traced and thus objects on those pages are never be
// registered as objects on orphaned pages. We cannot assert this here
// since we might have an off-heap collection. We assert it in
// Heap::pushWeakPointerCallback.
if (CallbackStack::Item* item = s_weakCallbackStack->pop()) {
item->call(visitor);
return true;
}
return false;
}
void Heap::registerWeakTable(void* table, EphemeronCallback iterationCallback, EphemeronCallback iterationDoneCallback)
{
{
// Check that the ephemeron table being pushed onto the stack is not on
// an orphaned page.
ASSERT(!Heap::orphanedPagePool()->contains(table));
CallbackStack::Item* slot = s_ephemeronStack->allocateEntry();
*slot = CallbackStack::Item(table, iterationCallback);
}
// Register a post-marking callback to tell the tables that
// ephemeron iteration is complete.
pushPostMarkingCallback(table, iterationDoneCallback);
}
#if ENABLE(ASSERT)
bool Heap::weakTableRegistered(const void* table)
{
ASSERT(s_ephemeronStack);
return s_ephemeronStack->hasCallbackForObject(table);
}
#endif
void Heap::preGC()
{
ASSERT(!ThreadState::current()->isInGC());
for (ThreadState* state : ThreadState::attachedThreads())
state->preGC();
}
void Heap::postGC()
{
ASSERT(ThreadState::current()->isInGC());
for (ThreadState* state : ThreadState::attachedThreads())
state->postGC();
}
void Heap::collectGarbage(ThreadState::StackState stackState, ThreadState::GCType gcType)
{
ThreadState* state = ThreadState::current();
state->setGCState(ThreadState::StoppingOtherThreads);
GCScope gcScope(stackState);
// Check if we successfully parked the other threads. If not we bail out of
// the GC.
if (!gcScope.allThreadsParked()) {
state->scheduleGC();
return;
}
if (state->isMainThread())
ScriptForbiddenScope::enter();
s_lastGCWasConservative = false;
TRACE_EVENT2("blink_gc", "Heap::collectGarbage",
"precise", stackState == ThreadState::NoHeapPointersOnStack,
"forced", gcType == ThreadState::ForcedGC);
TRACE_EVENT_SCOPED_SAMPLING_STATE("blink_gc", "BlinkGC");
double timeStamp = WTF::currentTimeMS();
#if ENABLE(GC_PROFILE_MARKING)
static_cast<MarkingVisitor<GlobalMarking>*>(s_markingVisitor)->objectGraph().clear();
#endif
// Disallow allocation during garbage collection (but not during the
// finalization that happens when the gcScope is torn down).
ThreadState::NoAllocationScope noAllocationScope(state);
preGC();
Heap::resetMarkedObjectSize();
Heap::resetAllocatedObjectSize();
// 1. Trace persistent roots.
ThreadState::visitPersistentRoots(s_markingVisitor);
// 2. Trace objects reachable from the persistent roots including
// ephemerons.
processMarkingStack(s_markingVisitor);
// 3. Trace objects reachable from the stack. We do this independent of the
// given stackState since other threads might have a different stack state.
ThreadState::visitStackRoots(s_markingVisitor);
// 4. Trace objects reachable from the stack "roots" including ephemerons.
// Only do the processing if we found a pointer to an object on one of the
// thread stacks.
if (lastGCWasConservative())
processMarkingStack(s_markingVisitor);
postMarkingProcessing(s_markingVisitor);
globalWeakProcessing(s_markingVisitor);
// Now we can delete all orphaned pages because there are no dangling
// pointers to the orphaned pages. (If we have such dangling pointers,
// we should have crashed during marking before getting here.)
orphanedPagePool()->decommitOrphanedPages();
postGC();
#if ENABLE(GC_PROFILE_MARKING)
static_cast<MarkingVisitor<GlobalMarking>*>(s_markingVisitor)->reportStats();
#endif
if (Platform::current()) {
Platform::current()->histogramCustomCounts("BlinkGC.CollectGarbage", WTF::currentTimeMS() - timeStamp, 0, 10 * 1000, 50);
Platform::current()->histogramCustomCounts("BlinkGC.TotalObjectSpace", Heap::allocatedObjectSize() / 1024, 0, 4 * 1024 * 1024, 50);
Platform::current()->histogramCustomCounts("BlinkGC.TotalAllocatedSpace", Heap::allocatedSpace() / 1024, 0, 4 * 1024 * 1024, 50);
}
if (state->isMainThread())
ScriptForbiddenScope::exit();
}
void Heap::collectGarbageForTerminatingThread(ThreadState* state)
{
// We explicitly do not enter a safepoint while doing thread specific
// garbage collection since we don't want to allow a global GC at the
// same time as a thread local GC.
{
MarkingVisitor<ThreadLocalMarking> markingVisitor;
ThreadState::NoAllocationScope noAllocationScope(state);
state->preGC();
// 1. Trace the thread local persistent roots. For thread local GCs we
// don't trace the stack (ie. no conservative scanning) since this is
// only called during thread shutdown where there should be no objects
// on the stack.
// We also assume that orphaned pages have no objects reachable from
// persistent handles on other threads or CrossThreadPersistents. The
// only cases where this could happen is if a subsequent conservative
// global GC finds a "pointer" on the stack or due to a programming
// error where an object has a dangling cross-thread pointer to an
// object on this heap.
state->visitPersistents(&markingVisitor);
// 2. Trace objects reachable from the thread's persistent roots
// including ephemerons.
processMarkingStack(&markingVisitor);
postMarkingProcessing(&markingVisitor);
globalWeakProcessing(&markingVisitor);
state->postGC();
}
state->performPendingSweep();
}
void Heap::processMarkingStack(Visitor* markingVisitor)
{
// Ephemeron fixed point loop.
do {
{
// Iteratively mark all objects that are reachable from the objects
// currently pushed onto the marking stack.
TRACE_EVENT0("blink_gc", "Heap::processMarkingStackSingleThreaded");
while (popAndInvokeTraceCallback(s_markingStack, markingVisitor)) { }
}
{
// Mark any strong pointers that have now become reachable in
// ephemeron maps.
TRACE_EVENT0("blink_gc", "Heap::processEphemeronStack");
s_ephemeronStack->invokeEphemeronCallbacks(markingVisitor);
}
// Rerun loop if ephemeron processing queued more objects for tracing.
} while (!s_markingStack->isEmpty());
}
void Heap::postMarkingProcessing(Visitor* markingVisitor)
{
TRACE_EVENT0("blink_gc", "Heap::postMarkingProcessing");
// Call post-marking callbacks including:
// 1. the ephemeronIterationDone callbacks on weak tables to do cleanup
// (specifically to clear the queued bits for weak hash tables), and
// 2. the markNoTracing callbacks on collection backings to mark them
// if they are only reachable from their front objects.
while (popAndInvokePostMarkingCallback(markingVisitor)) { }
s_ephemeronStack->clear();
// Post-marking callbacks should not trace any objects and
// therefore the marking stack should be empty after the
// post-marking callbacks.
ASSERT(s_markingStack->isEmpty());
}
void Heap::globalWeakProcessing(Visitor* markingVisitor)
{
TRACE_EVENT0("blink_gc", "Heap::globalWeakProcessing");
// Call weak callbacks on objects that may now be pointing to dead objects.
while (popAndInvokeWeakPointerCallback(markingVisitor)) { }
// It is not permitted to trace pointers of live objects in the weak
// callback phase, so the marking stack should still be empty here.
ASSERT(s_markingStack->isEmpty());
}
void Heap::collectAllGarbage()
{
// FIXME: Oilpan: we should perform a single GC and everything
// should die. Unfortunately it is not the case for all objects
// because the hierarchy was not completely moved to the heap and
// some heap allocated objects own objects that contain persistents
// pointing to other heap allocated objects.
for (int i = 0; i < 5; ++i)
collectGarbage(ThreadState::NoHeapPointersOnStack);
}
template<typename Header>
void ThreadHeap<Header>::prepareHeapForTermination()
{
for (HeapPage<Header>* page = m_firstPage; page; page = page->next()) {
page->setTerminating();
}
for (LargeObject<Header>* largeObject = m_firstLargeObject; largeObject; largeObject = largeObject->next()) {
largeObject->setTerminating();
}
}
size_t Heap::objectPayloadSizeForTesting()
{
size_t objectPayloadSize = 0;
for (ThreadState* state : ThreadState::attachedThreads()) {
state->setGCState(ThreadState::GCRunning);
state->makeConsistentForSweeping();
objectPayloadSize += state->objectPayloadSizeForTesting();
state->setGCState(ThreadState::SweepScheduled);
state->setGCState(ThreadState::Sweeping);
state->setGCState(ThreadState::NoGCScheduled);
}
return objectPayloadSize;
}
template<typename HeapTraits, typename HeapType, typename HeaderType>
void HeapAllocator::backingFree(void* address)
{
ThreadState* state = ThreadState::current();
if (!address || state->isInGC())
return;
if (state->sweepForbidden())
return;
// Don't promptly free large objects because their page is never reused
// and don't free backings allocated on other threads.
BaseHeapPage* page = pageFromObject(address);
if (page->isLargeObject() || page->threadState() != state)
return;
HeaderType* header = HeaderType::fromPayload(address);
header->checkHeader();
int heapIndex = HeapTraits::index(header->payloadSize());
HeapType* heap = static_cast<HeapType*>(state->heap(heapIndex));
heap->promptlyFreeObject(header);
}
void HeapAllocator::freeVectorBacking(void* address)
{
using HeapTraits = HeapIndexTrait<VectorBackingHeap>;
backingFree<HeapTraits, HeapTraits::HeapType, HeapTraits::HeaderType>(address);
}
void HeapAllocator::freeInlineVectorBacking(void* address)
{
using HeapTraits = HeapIndexTrait<InlineVectorBackingHeap>;
backingFree<HeapTraits, HeapTraits::HeapType, HeapTraits::HeaderType>(address);
}
void HeapAllocator::freeHashTableBacking(void* address)
{
using HeapTraits = HeapIndexTrait<HashTableBackingHeap>;
backingFree<HeapTraits, HeapTraits::HeapType, HeapTraits::HeaderType>(address);
}
template<typename HeapTraits, typename HeapType, typename HeaderType>
bool HeapAllocator::backingExpand(void* address, size_t newSize)
{
ThreadState* state = ThreadState::current();
if (!address || state->isInGC())
return false;
if (state->sweepForbidden())
return false;
ASSERT(state->isAllocationAllowed());
BaseHeapPage* page = pageFromObject(address);
if (page->isLargeObject() || page->threadState() != state)
return false;
HeaderType* header = HeaderType::fromPayload(address);
header->checkHeader();
int heapIndex = HeapTraits::index(header->payloadSize());
HeapType* heap = static_cast<HeapType*>(state->heap(heapIndex));
return heap->expandObject(header, newSize);
}
bool HeapAllocator::expandVectorBacking(void* address, size_t newSize)
{
using HeapTraits = HeapIndexTrait<VectorBackingHeap>;
return backingExpand<HeapTraits, HeapTraits::HeapType, HeapTraits::HeaderType>(address, newSize);
}
bool HeapAllocator::expandInlineVectorBacking(void* address, size_t newSize)
{
using HeapTraits = HeapIndexTrait<InlineVectorBackingHeap>;
return backingExpand<HeapTraits, HeapTraits::HeapType, HeapTraits::HeaderType>(address, newSize);
}
template<typename HeapTraits>
void HeapAllocator::backingShrink(void* address, size_t quantizedCurrentSize, size_t quantizedShrunkSize)
{
// We shrink the object only if the shrinking will make a non-small
// prompt-free block.
// FIXME: Optimize the threshold size.
if (quantizedCurrentSize <= quantizedShrunkSize + sizeof(HeapObjectHeader) + sizeof(void*) * 32)
return;
ThreadState* state = ThreadState::current();
if (!address || state->isInGC())
return;
if (state->sweepForbidden())
return;
ASSERT(state->isAllocationAllowed());
BaseHeapPage* page = pageFromObject(address);
if (page->isLargeObject()) {
// We do nothing for large objects.
// FIXME: This wastes unused memory. If this increases memory
// consumption, we should reallocate a new large object and shrink the
// memory usage.
return;
}
if (page->threadState() != state)
return;
typename HeapTraits::HeaderType* header = HeapTraits::HeaderType::fromPayload(address);
header->checkHeader();
int heapIndex = HeapTraits::index(header->payloadSize());
static_cast<typename HeapTraits::HeapType*>(state->heap(heapIndex))->shrinkObject(header, quantizedShrunkSize);
}
void HeapAllocator::shrinkVectorBackingInternal(void* address, size_t quantizedCurrentSize, size_t quantizedShrunkSize)
{
backingShrink<HeapIndexTrait<VectorBackingHeap>>(address, quantizedCurrentSize, quantizedShrunkSize);
}
void HeapAllocator::shrinkInlineVectorBackingInternal(void* address, size_t quantizedCurrentSize, size_t quantizedShrunkSize)
{
backingShrink<HeapIndexTrait<InlineVectorBackingHeap>>(address, quantizedCurrentSize, quantizedShrunkSize);
}
BaseHeapPage* Heap::lookup(Address address)
{
ASSERT(ThreadState::current()->isInGC());
if (!s_regionTree)
return nullptr;
if (PageMemoryRegion* region = s_regionTree->lookup(address)) {
BaseHeapPage* page = region->pageFromAddress(address);
return page && !page->orphaned() ? page : nullptr;
}
return nullptr;
}
static Mutex& regionTreeMutex()
{
AtomicallyInitializedStatic(Mutex&, mutex = *new Mutex);
return mutex;
}
void Heap::removePageMemoryRegion(PageMemoryRegion* region)
{
// Deletion of large objects (and thus their regions) can happen
// concurrently on sweeper threads. Removal can also happen during thread
// shutdown, but that case is safe. Regardless, we make all removals
// mutually exclusive.
MutexLocker locker(regionTreeMutex());
RegionTree::remove(region, &s_regionTree);
}
void Heap::addPageMemoryRegion(PageMemoryRegion* region)
{
RegionTree::add(new RegionTree(region), &s_regionTree);
}
PageMemoryRegion* Heap::RegionTree::lookup(Address address)
{
RegionTree* current = s_regionTree;
while (current) {
Address base = current->m_region->base();
if (address < base) {
current = current->m_left;
continue;
}
if (address >= base + current->m_region->size()) {
current = current->m_right;
continue;
}
ASSERT(current->m_region->contains(address));
return current->m_region;
}
return nullptr;
}
void Heap::RegionTree::add(RegionTree* newTree, RegionTree** context)
{
ASSERT(newTree);
Address base = newTree->m_region->base();
for (RegionTree* current = *context; current; current = *context) {
ASSERT(!current->m_region->contains(base));
context = (base < current->m_region->base()) ? &current->m_left : &current->m_right;
}
*context = newTree;
}
void Heap::RegionTree::remove(PageMemoryRegion* region, RegionTree** context)
{
ASSERT(region);
ASSERT(context);
Address base = region->base();
RegionTree* current = *context;
for (; current; current = *context) {
if (region == current->m_region)
break;
context = (base < current->m_region->base()) ? &current->m_left : &current->m_right;
}
// Shutdown via detachMainThread might not have populated the region tree.
if (!current)
return;
*context = nullptr;
if (current->m_left) {
add(current->m_left, context);
current->m_left = nullptr;
}
if (current->m_right) {
add(current->m_right, context);
current->m_right = nullptr;
}
delete current;
}
// Force template instantiations for the types that we need.
template class HeapPage<GeneralHeapObjectHeader>;
template class HeapPage<HeapObjectHeader>;
template class ThreadHeap<GeneralHeapObjectHeader>;
template class ThreadHeap<HeapObjectHeader>;
Visitor* Heap::s_markingVisitor;
CallbackStack* Heap::s_markingStack;
CallbackStack* Heap::s_postMarkingCallbackStack;
CallbackStack* Heap::s_weakCallbackStack;
CallbackStack* Heap::s_ephemeronStack;
HeapDoesNotContainCache* Heap::s_heapDoesNotContainCache;
bool Heap::s_shutdownCalled = false;
bool Heap::s_lastGCWasConservative = false;
FreePagePool* Heap::s_freePagePool;
OrphanedPagePool* Heap::s_orphanedPagePool;
Heap::RegionTree* Heap::s_regionTree = nullptr;
size_t Heap::s_allocatedObjectSize = 0;
size_t Heap::s_allocatedSpace = 0;
size_t Heap::s_markedObjectSize = 0;
} // namespace blink