https://github.com/mozilla/gecko-dev
Tip revision: 63e2edf7fb1a1794e0a4119c8333d5b7a2a64547 authored by ffxbld on 06 March 2012, 01:57:57 UTC
Added FENNEC_11_0b6_RELEASE FENNEC_11_0b6_BUILD1 tag(s) for changeset d9c4d55a4edb. DONTBUILD CLOSED TREE a=release
Added FENNEC_11_0b6_RELEASE FENNEC_11_0b6_BUILD1 tag(s) for changeset d9c4d55a4edb. DONTBUILD CLOSED TREE a=release
Tip revision: 63e2edf
jsgc.h
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
*
* ***** BEGIN LICENSE BLOCK *****
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
*
* The contents of this file are subject to the Mozilla Public License Version
* 1.1 (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
* http://www.mozilla.org/MPL/
*
* Software distributed under the License is distributed on an "AS IS" basis,
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
* for the specific language governing rights and limitations under the
* License.
*
* The Original Code is Mozilla Communicator client code, released
* March 31, 1998.
*
* The Initial Developer of the Original Code is
* Netscape Communications Corporation.
* Portions created by the Initial Developer are Copyright (C) 1998
* the Initial Developer. All Rights Reserved.
*
* Contributor(s):
*
* Alternatively, the contents of this file may be used under the terms of
* either of the GNU General Public License Version 2 or later (the "GPL"),
* or the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
* in which case the provisions of the GPL or the LGPL are applicable instead
* of those above. If you wish to allow use of your version of this file only
* under the terms of either the GPL or the LGPL, and not to allow others to
* use your version of this file under the terms of the MPL, indicate your
* decision by deleting the provisions above and replace them with the notice
* and other provisions required by the GPL or the LGPL. If you do not delete
* the provisions above, a recipient may use your version of this file under
* the terms of any one of the MPL, the GPL or the LGPL.
*
* ***** END LICENSE BLOCK ***** */
#ifndef jsgc_h___
#define jsgc_h___
/*
* JS Garbage Collector.
*/
#include <setjmp.h>
#include "mozilla/Util.h"
#include "jsalloc.h"
#include "jstypes.h"
#include "jsprvtd.h"
#include "jspubtd.h"
#include "jsdhash.h"
#include "jsgcchunk.h"
#include "jslock.h"
#include "jsutil.h"
#include "jsversion.h"
#include "jsgcstats.h"
#include "jscell.h"
#include "ds/BitArray.h"
#include "gc/Statistics.h"
#include "js/HashTable.h"
#include "js/Vector.h"
#include "js/TemplateLib.h"
struct JSCompartment;
extern "C" void
js_TraceXML(JSTracer *trc, JSXML* thing);
#if JS_STACK_GROWTH_DIRECTION > 0
# define JS_CHECK_STACK_SIZE(limit, lval) ((jsuword)(lval) < limit)
#else
# define JS_CHECK_STACK_SIZE(limit, lval) ((jsuword)(lval) > limit)
#endif
namespace js {
class GCHelperThread;
struct Shape;
namespace gc {
struct Arena;
/*
* This must be an upper bound, but we do not need the least upper bound, so
* we just exclude non-background objects.
*/
const size_t MAX_BACKGROUND_FINALIZE_KINDS = FINALIZE_LIMIT - FINALIZE_OBJECT_LIMIT / 2;
const size_t ArenaShift = 12;
const size_t ArenaSize = size_t(1) << ArenaShift;
const size_t ArenaMask = ArenaSize - 1;
/*
* This is the maximum number of arenas we allow in the FreeCommitted state
* before we trigger a GC_SHRINK to release free arenas to the OS.
*/
const static uint32_t FreeCommittedArenasThreshold = (32 << 20) / ArenaSize;
/*
* The mark bitmap has one bit per each GC cell. For multi-cell GC things this
* wastes space but allows to avoid expensive devisions by thing's size when
* accessing the bitmap. In addition this allows to use some bits for colored
* marking during the cycle GC.
*/
const size_t ArenaCellCount = size_t(1) << (ArenaShift - Cell::CellShift);
const size_t ArenaBitmapBits = ArenaCellCount;
const size_t ArenaBitmapBytes = ArenaBitmapBits / 8;
const size_t ArenaBitmapWords = ArenaBitmapBits / JS_BITS_PER_WORD;
/*
* A FreeSpan represents a contiguous sequence of free cells in an Arena.
* |first| is the address of the first free cell in the span. |last| is the
* address of the last free cell in the span. This last cell holds a FreeSpan
* data structure for the next span unless this is the last span on the list
* of spans in the arena. For this last span |last| points to the last byte of
* the last thing in the arena and no linkage is stored there, so
* |last| == arenaStart + ArenaSize - 1. If the space at the arena end is
* fully used this last span is empty and |first| == |last + 1|.
*
* Thus |first| < |last| implies that we have either the last span with at least
* one element or that the span is not the last and contains at least 2
* elements. In both cases to allocate a thing from this span we need simply
* to increment |first| by the allocation size.
*
* |first| == |last| implies that we have a one element span that records the
* next span. So to allocate from it we need to update the span list head
* with a copy of the span stored at |last| address so the following
* allocations will use that span.
*
* |first| > |last| implies that we have an empty last span and the arena is
* fully used.
*
* Also only for the last span (|last| & 1)! = 0 as all allocation sizes are
* multiples of Cell::CellSize.
*/
struct FreeSpan {
uintptr_t first;
uintptr_t last;
public:
FreeSpan() {}
FreeSpan(uintptr_t first, uintptr_t last)
: first(first), last(last) {
checkSpan();
}
/*
* To minimize the size of the arena header the first span is encoded
* there as offsets from the arena start.
*/
static size_t encodeOffsets(size_t firstOffset, size_t lastOffset) {
/* Check that we can pack the offsets into uint16. */
JS_STATIC_ASSERT(ArenaShift < 16);
JS_ASSERT(firstOffset <= ArenaSize);
JS_ASSERT(lastOffset < ArenaSize);
JS_ASSERT(firstOffset <= ((lastOffset + 1) & ~size_t(1)));
return firstOffset | (lastOffset << 16);
}
/*
* Encoded offsets for a full arena when its first span is the last one
* and empty.
*/
static const size_t FullArenaOffsets = ArenaSize | ((ArenaSize - 1) << 16);
static FreeSpan decodeOffsets(uintptr_t arenaAddr, size_t offsets) {
JS_ASSERT(!(arenaAddr & ArenaMask));
size_t firstOffset = offsets & 0xFFFF;
size_t lastOffset = offsets >> 16;
JS_ASSERT(firstOffset <= ArenaSize);
JS_ASSERT(lastOffset < ArenaSize);
/*
* We must not use | when calculating first as firstOffset is
* ArenaMask + 1 for the empty span.
*/
return FreeSpan(arenaAddr + firstOffset, arenaAddr | lastOffset);
}
void initAsEmpty(uintptr_t arenaAddr = 0) {
JS_ASSERT(!(arenaAddr & ArenaMask));
first = arenaAddr + ArenaSize;
last = arenaAddr | (ArenaSize - 1);
JS_ASSERT(isEmpty());
}
bool isEmpty() const {
checkSpan();
return first > last;
}
bool hasNext() const {
checkSpan();
return !(last & uintptr_t(1));
}
const FreeSpan *nextSpan() const {
JS_ASSERT(hasNext());
return reinterpret_cast<FreeSpan *>(last);
}
FreeSpan *nextSpanUnchecked(size_t thingSize) const {
#ifdef DEBUG
uintptr_t lastOffset = last & ArenaMask;
JS_ASSERT(!(lastOffset & 1));
JS_ASSERT((ArenaSize - lastOffset) % thingSize == 0);
#endif
return reinterpret_cast<FreeSpan *>(last);
}
uintptr_t arenaAddressUnchecked() const {
return last & ~ArenaMask;
}
uintptr_t arenaAddress() const {
checkSpan();
return arenaAddressUnchecked();
}
ArenaHeader *arenaHeader() const {
return reinterpret_cast<ArenaHeader *>(arenaAddress());
}
bool isSameNonEmptySpan(const FreeSpan *another) const {
JS_ASSERT(!isEmpty());
JS_ASSERT(!another->isEmpty());
return first == another->first && last == another->last;
}
bool isWithinArena(uintptr_t arenaAddr) const {
JS_ASSERT(!(arenaAddr & ArenaMask));
/* Return true for the last empty span as well. */
return arenaAddress() == arenaAddr;
}
size_t encodeAsOffsets() const {
/*
* We must use first - arenaAddress(), not first & ArenaMask as
* first == ArenaMask + 1 for an empty span.
*/
uintptr_t arenaAddr = arenaAddress();
return encodeOffsets(first - arenaAddr, last & ArenaMask);
}
/* See comments before FreeSpan for details. */
JS_ALWAYS_INLINE void *allocate(size_t thingSize) {
JS_ASSERT(thingSize % Cell::CellSize == 0);
checkSpan();
uintptr_t thing = first;
if (thing < last) {
/* Bump-allocate from the current span. */
first = thing + thingSize;
} else if (JS_LIKELY(thing == last)) {
/*
* Move to the next span. We use JS_LIKELY as without PGO
* compilers mis-predict == here as unlikely to succeed.
*/
*this = *reinterpret_cast<FreeSpan *>(thing);
} else {
return NULL;
}
checkSpan();
return reinterpret_cast<void *>(thing);
}
/* A version of allocate when we know that the span is not empty. */
JS_ALWAYS_INLINE void *infallibleAllocate(size_t thingSize) {
JS_ASSERT(thingSize % Cell::CellSize == 0);
checkSpan();
uintptr_t thing = first;
if (thing < last) {
first = thing + thingSize;
} else {
JS_ASSERT(thing == last);
*this = *reinterpret_cast<FreeSpan *>(thing);
}
checkSpan();
return reinterpret_cast<void *>(thing);
}
/*
* Allocate from a newly allocated arena. We do not move the free list
* from the arena. Rather we set the arena up as fully used during the
* initialization so to allocate we simply return the first thing in the
* arena and set the free list to point to the second.
*/
JS_ALWAYS_INLINE void *allocateFromNewArena(uintptr_t arenaAddr, size_t firstThingOffset,
size_t thingSize) {
JS_ASSERT(!(arenaAddr & ArenaMask));
uintptr_t thing = arenaAddr | firstThingOffset;
first = thing + thingSize;
last = arenaAddr | ArenaMask;
checkSpan();
return reinterpret_cast<void *>(thing);
}
void checkSpan() const {
#ifdef DEBUG
/* We do not allow spans at the end of the address space. */
JS_ASSERT(last != uintptr_t(-1));
JS_ASSERT(first);
JS_ASSERT(last);
JS_ASSERT(first - 1 <= last);
uintptr_t arenaAddr = arenaAddressUnchecked();
if (last & 1) {
/* The span is the last. */
JS_ASSERT((last & ArenaMask) == ArenaMask);
if (first - 1 == last) {
/* The span is last and empty. The above start != 0 check
* implies that we are not at the end of the address space.
*/
return;
}
size_t spanLength = last - first + 1;
JS_ASSERT(spanLength % Cell::CellSize == 0);
/* Start and end must belong to the same arena. */
JS_ASSERT((first & ~ArenaMask) == arenaAddr);
return;
}
/* The span is not the last and we have more spans to follow. */
JS_ASSERT(first <= last);
size_t spanLengthWithoutOneThing = last - first;
JS_ASSERT(spanLengthWithoutOneThing % Cell::CellSize == 0);
JS_ASSERT((first & ~ArenaMask) == arenaAddr);
/*
* If there is not enough space before the arena end to allocate one
* more thing, then the span must be marked as the last one to avoid
* storing useless empty span reference.
*/
size_t beforeTail = ArenaSize - (last & ArenaMask);
JS_ASSERT(beforeTail >= sizeof(FreeSpan) + Cell::CellSize);
FreeSpan *next = reinterpret_cast<FreeSpan *>(last);
/*
* The GC things on the list of free spans come from one arena
* and the spans are linked in ascending address order with
* at least one non-free thing between spans.
*/
JS_ASSERT(last < next->first);
JS_ASSERT(arenaAddr == next->arenaAddressUnchecked());
if (next->first > next->last) {
/*
* The next span is the empty span that terminates the list for
* arenas that do not have any free things at the end.
*/
JS_ASSERT(next->first - 1 == next->last);
JS_ASSERT(arenaAddr + ArenaSize == next->first);
}
#endif
}
};
/* Every arena has a header. */
struct ArenaHeader {
friend struct FreeLists;
JSCompartment *compartment;
/*
* ArenaHeader::next has two purposes: when unallocated, it points to the
* next available Arena's header. When allocated, it points to the next
* arena of the same size class and compartment.
*/
ArenaHeader *next;
private:
/*
* The first span of free things in the arena. We encode it as the start
* and end offsets within the arena, not as FreeSpan structure, to
* minimize the header size.
*/
size_t firstFreeSpanOffsets;
/*
* One of AllocKind constants or FINALIZE_LIMIT when the arena does not
* contain any GC things and is on the list of empty arenas in the GC
* chunk. The latter allows to quickly check if the arena is allocated
* during the conservative GC scanning without searching the arena in the
* list.
*/
size_t allocKind : 8;
/*
* When recursive marking uses too much stack the marking is delayed and
* the corresponding arenas are put into a stack using the following field
* as a linkage. To distinguish the bottom of the stack from the arenas
* not present in the stack we use an extra flag to tag arenas on the
* stack.
*
* To minimize the ArenaHeader size we record the next delayed marking
* linkage as arenaAddress() >> ArenaShift and pack it with the allocKind
* field and hasDelayedMarking flag. We use 8 bits for the allocKind, not
* ArenaShift - 1, so the compiler can use byte-level memory instructions
* to access it.
*/
public:
size_t hasDelayedMarking : 1;
size_t nextDelayedMarking : JS_BITS_PER_WORD - 8 - 1;
static void staticAsserts() {
/* We must be able to fit the allockind into uint8_t. */
JS_STATIC_ASSERT(FINALIZE_LIMIT <= 255);
/*
* nextDelayedMarkingpacking assumes that ArenaShift has enough bits
* to cover allocKind and hasDelayedMarking.
*/
JS_STATIC_ASSERT(ArenaShift >= 8 + 1);
}
inline uintptr_t address() const;
inline Chunk *chunk() const;
bool allocated() const {
JS_ASSERT(allocKind <= size_t(FINALIZE_LIMIT));
return allocKind < size_t(FINALIZE_LIMIT);
}
void init(JSCompartment *comp, AllocKind kind) {
JS_ASSERT(!allocated());
JS_ASSERT(!hasDelayedMarking);
compartment = comp;
JS_STATIC_ASSERT(FINALIZE_LIMIT <= 255);
allocKind = size_t(kind);
/* See comments in FreeSpan::allocateFromNewArena. */
firstFreeSpanOffsets = FreeSpan::FullArenaOffsets;
}
void setAsNotAllocated() {
allocKind = size_t(FINALIZE_LIMIT);
hasDelayedMarking = 0;
nextDelayedMarking = 0;
}
uintptr_t arenaAddress() const {
return address();
}
Arena *getArena() {
return reinterpret_cast<Arena *>(arenaAddress());
}
AllocKind getAllocKind() const {
JS_ASSERT(allocated());
return AllocKind(allocKind);
}
inline size_t getThingSize() const;
bool hasFreeThings() const {
return firstFreeSpanOffsets != FreeSpan::FullArenaOffsets;
}
inline bool isEmpty() const;
void setAsFullyUsed() {
firstFreeSpanOffsets = FreeSpan::FullArenaOffsets;
}
FreeSpan getFirstFreeSpan() const {
#ifdef DEBUG
checkSynchronizedWithFreeList();
#endif
return FreeSpan::decodeOffsets(arenaAddress(), firstFreeSpanOffsets);
}
void setFirstFreeSpan(const FreeSpan *span) {
JS_ASSERT(span->isWithinArena(arenaAddress()));
firstFreeSpanOffsets = span->encodeAsOffsets();
}
#ifdef DEBUG
void checkSynchronizedWithFreeList() const;
#endif
inline Arena *getNextDelayedMarking() const;
inline void setNextDelayedMarking(Arena *arena);
};
struct Arena {
/*
* Layout of an arena:
* An arena is 4K in size and 4K-aligned. It starts with the ArenaHeader
* descriptor followed by some pad bytes. The remainder of the arena is
* filled with the array of T things. The pad bytes ensure that the thing
* array ends exactly at the end of the arena.
*
* +-------------+-----+----+----+-----+----+
* | ArenaHeader | pad | T0 | T1 | ... | Tn |
* +-------------+-----+----+----+-----+----+
*
* <----------------------------------------> = ArenaSize bytes
* <-------------------> = first thing offset
*/
ArenaHeader aheader;
uint8_t data[ArenaSize - sizeof(ArenaHeader)];
private:
static JS_FRIEND_DATA(const uint32_t) ThingSizes[];
static JS_FRIEND_DATA(const uint32_t) FirstThingOffsets[];
public:
static void staticAsserts();
static size_t thingSize(AllocKind kind) {
return ThingSizes[kind];
}
static size_t firstThingOffset(AllocKind kind) {
return FirstThingOffsets[kind];
}
static size_t thingsPerArena(size_t thingSize) {
JS_ASSERT(thingSize % Cell::CellSize == 0);
/* We should be able to fit FreeSpan in any GC thing. */
JS_ASSERT(thingSize >= sizeof(FreeSpan));
return (ArenaSize - sizeof(ArenaHeader)) / thingSize;
}
static size_t thingsSpan(size_t thingSize) {
return thingsPerArena(thingSize) * thingSize;
}
static bool isAligned(uintptr_t thing, size_t thingSize) {
/* Things ends at the arena end. */
uintptr_t tailOffset = (ArenaSize - thing) & ArenaMask;
return tailOffset % thingSize == 0;
}
uintptr_t address() const {
return aheader.address();
}
uintptr_t thingsStart(AllocKind thingKind) {
return address() | firstThingOffset(thingKind);
}
uintptr_t thingsEnd() {
return address() + ArenaSize;
}
template <typename T>
bool finalize(JSContext *cx, AllocKind thingKind, size_t thingSize, bool background);
};
/* The chunk header (located at the end of the chunk to preserve arena alignment). */
struct ChunkInfo {
Chunk *next;
Chunk **prevp;
/* Free arenas are linked together with aheader.next. */
ArenaHeader *freeArenasHead;
/*
* Decommitted arenas are tracked by a bitmap in the chunk header. We use
* this offset to start our search iteration close to a decommitted arena
* that we can allocate.
*/
uint32_t lastDecommittedArenaOffset;
/* Number of free arenas, either committed or decommitted. */
uint32_t numArenasFree;
/* Number of free, committed arenas. */
uint32_t numArenasFreeCommitted;
/* Number of GC cycles this chunk has survived. */
uint32_t age;
};
/*
* Calculating ArenasPerChunk:
*
* In order to figure out how many Arenas will fit in a chunk, we need to know
* how much extra space is available after we allocate the header data. This
* is a problem because the header size depends on the number of arenas in the
* chunk. The two dependent fields are bitmap and decommittedArenas.
*
* For the mark bitmap, we know that each arena will use a fixed number of full
* bytes: ArenaBitmapBytes. The full size of the header data is this number
* multiplied by the eventual number of arenas we have in the header. We,
* conceptually, distribute this header data among the individual arenas and do
* not include it in the header. This way we do not have to worry about its
* variable size: it gets attached to the variable number we are computing.
*
* For the decommitted arena bitmap, we only have 1 bit per arena, so this
* technique will not work. Instead, we observe that we do not have enough
* header info to fill 8 full arenas: it is currently 4 on 64bit, less on
* 32bit. Thus, with current numbers, we need 64 bytes for decommittedArenas.
* This will not become 63 bytes unless we double the data required in the
* header. Therefore, we just compute the number of bytes required to track
* every possible arena and do not worry about slop bits, since there are too
* few to usefully allocate.
*
* To actually compute the number of arenas we can allocate in a chunk, we
* divide the amount of available space less the header info (not including
* the mark bitmap which is distributed into the arena size) by the size of
* the arena (with the mark bitmap bytes it uses).
*/
const size_t BytesPerArenaWithHeader = ArenaSize + ArenaBitmapBytes;
const size_t ChunkDecommitBitmapBytes = ChunkSize / ArenaSize / JS_BITS_PER_BYTE;
const size_t ChunkBytesAvailable = ChunkSize - sizeof(ChunkInfo) - ChunkDecommitBitmapBytes;
const size_t ArenasPerChunk = ChunkBytesAvailable / BytesPerArenaWithHeader;
/* A chunk bitmap contains enough mark bits for all the cells in a chunk. */
struct ChunkBitmap {
uintptr_t bitmap[ArenaBitmapWords * ArenasPerChunk];
JS_ALWAYS_INLINE void getMarkWordAndMask(const Cell *cell, uint32_t color,
uintptr_t **wordp, uintptr_t *maskp);
JS_ALWAYS_INLINE bool isMarked(const Cell *cell, uint32_t color) {
uintptr_t *word, mask;
getMarkWordAndMask(cell, color, &word, &mask);
return *word & mask;
}
JS_ALWAYS_INLINE bool markIfUnmarked(const Cell *cell, uint32_t color) {
uintptr_t *word, mask;
getMarkWordAndMask(cell, BLACK, &word, &mask);
if (*word & mask)
return false;
*word |= mask;
if (color != BLACK) {
/*
* We use getMarkWordAndMask to recalculate both mask and word as
* doing just mask << color may overflow the mask.
*/
getMarkWordAndMask(cell, color, &word, &mask);
if (*word & mask)
return false;
*word |= mask;
}
return true;
}
JS_ALWAYS_INLINE void unmark(const Cell *cell, uint32_t color) {
uintptr_t *word, mask;
getMarkWordAndMask(cell, color, &word, &mask);
*word &= ~mask;
}
void clear() {
PodArrayZero(bitmap);
}
#ifdef DEBUG
bool noBitsSet(ArenaHeader *aheader) {
/*
* We assume that the part of the bitmap corresponding to the arena
* has the exact number of words so we do not need to deal with a word
* that covers bits from two arenas.
*/
JS_STATIC_ASSERT(ArenaBitmapBits == ArenaBitmapWords * JS_BITS_PER_WORD);
uintptr_t *word, unused;
getMarkWordAndMask(reinterpret_cast<Cell *>(aheader->address()), BLACK, &word, &unused);
for (size_t i = 0; i != ArenaBitmapWords; i++) {
if (word[i])
return false;
}
return true;
}
#endif
};
JS_STATIC_ASSERT(ArenaBitmapBytes * ArenasPerChunk == sizeof(ChunkBitmap));
typedef BitArray<ArenasPerChunk> PerArenaBitmap;
const size_t ChunkPadSize = ChunkSize
- (sizeof(Arena) * ArenasPerChunk)
- sizeof(ChunkBitmap)
- sizeof(PerArenaBitmap)
- sizeof(ChunkInfo);
JS_STATIC_ASSERT(ChunkPadSize < BytesPerArenaWithHeader);
/*
* Chunks contain arenas and associated data structures (mark bitmap, delayed
* marking state).
*/
struct Chunk {
Arena arenas[ArenasPerChunk];
/* Pad to full size to ensure cache alignment of ChunkInfo. */
uint8_t padding[ChunkPadSize];
ChunkBitmap bitmap;
PerArenaBitmap decommittedArenas;
ChunkInfo info;
static Chunk *fromAddress(uintptr_t addr) {
addr &= ~ChunkMask;
return reinterpret_cast<Chunk *>(addr);
}
static bool withinArenasRange(uintptr_t addr) {
uintptr_t offset = addr & ChunkMask;
return offset < ArenasPerChunk * ArenaSize;
}
static size_t arenaIndex(uintptr_t addr) {
JS_ASSERT(withinArenasRange(addr));
return (addr & ChunkMask) >> ArenaShift;
}
uintptr_t address() const {
uintptr_t addr = reinterpret_cast<uintptr_t>(this);
JS_ASSERT(!(addr & ChunkMask));
return addr;
}
bool unused() const {
return info.numArenasFree == ArenasPerChunk;
}
bool noAvailableArenas() const {
return info.numArenasFree == 0;
}
inline void addToAvailableList(JSCompartment *compartment);
inline void removeFromAvailableList();
ArenaHeader *allocateArena(JSCompartment *comp, AllocKind kind);
void releaseArena(ArenaHeader *aheader);
static Chunk *allocate(JSRuntime *rt);
/* Must be called with the GC lock taken. */
static inline void release(JSRuntime *rt, Chunk *chunk);
private:
inline void init();
/* Search for a decommitted arena to allocate. */
jsuint findDecommittedArenaOffset();
ArenaHeader* fetchNextDecommittedArena();
/* Unlink and return the freeArenasHead. */
inline ArenaHeader* fetchNextFreeArena(JSRuntime *rt);
};
JS_STATIC_ASSERT(sizeof(Chunk) == ChunkSize);
class ChunkPool {
Chunk *emptyChunkListHead;
size_t emptyCount;
public:
ChunkPool()
: emptyChunkListHead(NULL),
emptyCount(0) { }
size_t getEmptyCount() const {
return emptyCount;
}
inline bool wantBackgroundAllocation(JSRuntime *rt) const;
/* Must be called with the GC lock taken. */
inline Chunk *get(JSRuntime *rt);
/* Must be called either during the GC or with the GC lock taken. */
inline void put(JSRuntime *rt, Chunk *chunk);
/* Must be called either during the GC or with the GC lock taken. */
void expire(JSRuntime *rt, bool releaseAll);
/* Must be called either during the GC or with the GC lock taken. */
JS_FRIEND_API(int64_t) countCleanDecommittedArenas(JSRuntime *rt);
};
inline uintptr_t
Cell::address() const
{
uintptr_t addr = uintptr_t(this);
JS_ASSERT(addr % Cell::CellSize == 0);
JS_ASSERT(Chunk::withinArenasRange(addr));
return addr;
}
inline ArenaHeader *
Cell::arenaHeader() const
{
uintptr_t addr = address();
addr &= ~ArenaMask;
return reinterpret_cast<ArenaHeader *>(addr);
}
Chunk *
Cell::chunk() const
{
uintptr_t addr = uintptr_t(this);
JS_ASSERT(addr % Cell::CellSize == 0);
addr &= ~(ChunkSize - 1);
return reinterpret_cast<Chunk *>(addr);
}
AllocKind
Cell::getAllocKind() const
{
return arenaHeader()->getAllocKind();
}
#ifdef DEBUG
inline bool
Cell::isAligned() const
{
return Arena::isAligned(address(), arenaHeader()->getThingSize());
}
#endif
inline uintptr_t
ArenaHeader::address() const
{
uintptr_t addr = reinterpret_cast<uintptr_t>(this);
JS_ASSERT(!(addr & ArenaMask));
JS_ASSERT(Chunk::withinArenasRange(addr));
return addr;
}
inline Chunk *
ArenaHeader::chunk() const
{
return Chunk::fromAddress(address());
}
inline bool
ArenaHeader::isEmpty() const
{
/* Arena is empty if its first span covers the whole arena. */
JS_ASSERT(allocated());
size_t firstThingOffset = Arena::firstThingOffset(getAllocKind());
return firstFreeSpanOffsets == FreeSpan::encodeOffsets(firstThingOffset, ArenaMask);
}
inline size_t
ArenaHeader::getThingSize() const
{
JS_ASSERT(allocated());
return Arena::thingSize(getAllocKind());
}
inline Arena *
ArenaHeader::getNextDelayedMarking() const
{
return reinterpret_cast<Arena *>(nextDelayedMarking << ArenaShift);
}
inline void
ArenaHeader::setNextDelayedMarking(Arena *arena)
{
JS_ASSERT(!hasDelayedMarking);
hasDelayedMarking = 1;
nextDelayedMarking = arena->address() >> ArenaShift;
}
JS_ALWAYS_INLINE void
ChunkBitmap::getMarkWordAndMask(const Cell *cell, uint32_t color,
uintptr_t **wordp, uintptr_t *maskp)
{
JS_ASSERT(cell->chunk() == Chunk::fromAddress(reinterpret_cast<uintptr_t>(this)));
size_t bit = (cell->address() & ChunkMask) / Cell::CellSize + color;
JS_ASSERT(bit < ArenaBitmapBits * ArenasPerChunk);
*maskp = uintptr_t(1) << (bit % JS_BITS_PER_WORD);
*wordp = &bitmap[bit / JS_BITS_PER_WORD];
}
static void
AssertValidColor(const void *thing, uint32_t color)
{
#ifdef DEBUG
ArenaHeader *aheader = reinterpret_cast<const js::gc::Cell *>(thing)->arenaHeader();
JS_ASSERT_IF(color, color < aheader->getThingSize() / Cell::CellSize);
#endif
}
inline bool
Cell::isMarked(uint32_t color) const
{
AssertValidColor(this, color);
return chunk()->bitmap.isMarked(this, color);
}
bool
Cell::markIfUnmarked(uint32_t color) const
{
AssertValidColor(this, color);
return chunk()->bitmap.markIfUnmarked(this, color);
}
void
Cell::unmark(uint32_t color) const
{
JS_ASSERT(color != BLACK);
AssertValidColor(this, color);
chunk()->bitmap.unmark(this, color);
}
JSCompartment *
Cell::compartment() const
{
return arenaHeader()->compartment;
}
/*
* Lower limit after which we limit the heap growth
*/
const size_t GC_ALLOCATION_THRESHOLD = 30 * 1024 * 1024;
/*
* A GC is triggered once the number of newly allocated arenas is
* GC_HEAP_GROWTH_FACTOR times the number of live arenas after the last GC
* starting after the lower limit of GC_ALLOCATION_THRESHOLD.
*/
const float GC_HEAP_GROWTH_FACTOR = 3.0f;
/* Perform a Full GC every 20 seconds if MaybeGC is called */
static const int64_t GC_IDLE_FULL_SPAN = 20 * 1000 * 1000;
static inline JSGCTraceKind
MapAllocToTraceKind(AllocKind thingKind)
{
static const JSGCTraceKind map[FINALIZE_LIMIT] = {
JSTRACE_OBJECT, /* FINALIZE_OBJECT0 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT0_BACKGROUND */
JSTRACE_OBJECT, /* FINALIZE_OBJECT2 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT2_BACKGROUND */
JSTRACE_OBJECT, /* FINALIZE_OBJECT4 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT4_BACKGROUND */
JSTRACE_OBJECT, /* FINALIZE_OBJECT8 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT8_BACKGROUND */
JSTRACE_OBJECT, /* FINALIZE_OBJECT12 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT12_BACKGROUND */
JSTRACE_OBJECT, /* FINALIZE_OBJECT16 */
JSTRACE_OBJECT, /* FINALIZE_OBJECT16_BACKGROUND */
JSTRACE_SCRIPT, /* FINALIZE_SCRIPT */
JSTRACE_SHAPE, /* FINALIZE_SHAPE */
JSTRACE_BASE_SHAPE, /* FINALIZE_BASE_SHAPE */
JSTRACE_TYPE_OBJECT,/* FINALIZE_TYPE_OBJECT */
#if JS_HAS_XML_SUPPORT /* FINALIZE_XML */
JSTRACE_XML,
#endif
JSTRACE_STRING, /* FINALIZE_SHORT_STRING */
JSTRACE_STRING, /* FINALIZE_STRING */
JSTRACE_STRING, /* FINALIZE_EXTERNAL_STRING */
};
return map[thingKind];
}
inline JSGCTraceKind
GetGCThingTraceKind(const void *thing);
struct ArenaLists {
/*
* ArenaList::head points to the start of the list. Normally cursor points
* to the first arena in the list with some free things and all arenas
* before cursor are fully allocated. However, as the arena currently being
* allocated from is considered full while its list of free spans is moved
* into the freeList, during the GC or cell enumeration, when an
* unallocated freeList is moved back to the arena, we can see an arena
* with some free cells before the cursor. The cursor is an indirect
* pointer to allow for efficient list insertion at the cursor point and
* other list manipulations.
*/
struct ArenaList {
ArenaHeader *head;
ArenaHeader **cursor;
ArenaList() {
clear();
}
void clear() {
head = NULL;
cursor = &head;
}
};
private:
/*
* For each arena kind its free list is represented as the first span with
* free things. Initially all the spans are initialized as empty. After we
* find a new arena with available things we move its first free span into
* the list and set the arena as fully allocated. way we do not need to
* update the arena header after the initial allocation. When starting the
* GC we only move the head of the of the list of spans back to the arena
* only for the arena that was not fully allocated.
*/
FreeSpan freeLists[FINALIZE_LIMIT];
ArenaList arenaLists[FINALIZE_LIMIT];
#ifdef JS_THREADSAFE
/*
* The background finalization adds the finalized arenas to the list at
* the *cursor position. backgroundFinalizeState controls the interaction
* between the GC lock and the access to the list from the allocation
* thread.
*
* BFS_DONE indicates that the finalizations is not running or cannot
* affect this arena list. The allocation thread can access the list
* outside the GC lock.
*
* In BFS_RUN and BFS_JUST_FINISHED the allocation thread must take the
* lock. The former indicates that the finalization still runs. The latter
* signals that finalization just added to the list finalized arenas. In
* that case the lock effectively serves as a read barrier to ensure that
* the allocation thread see all the writes done during finalization.
*/
enum BackgroundFinalizeState {
BFS_DONE,
BFS_RUN,
BFS_JUST_FINISHED
};
volatile uintptr_t backgroundFinalizeState[FINALIZE_LIMIT];
#endif
public:
ArenaLists() {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i)
freeLists[i].initAsEmpty();
#ifdef JS_THREADSAFE
for (size_t i = 0; i != FINALIZE_LIMIT; ++i)
backgroundFinalizeState[i] = BFS_DONE;
#endif
}
~ArenaLists() {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i) {
#ifdef JS_THREADSAFE
/*
* We can only call this during the shutdown after the last GC when
* the background finalization is disabled.
*/
JS_ASSERT(backgroundFinalizeState[i] == BFS_DONE);
#endif
ArenaHeader **headp = &arenaLists[i].head;
while (ArenaHeader *aheader = *headp) {
*headp = aheader->next;
aheader->chunk()->releaseArena(aheader);
}
}
}
const FreeSpan *getFreeList(AllocKind thingKind) const {
return &freeLists[thingKind];
}
ArenaHeader *getFirstArena(AllocKind thingKind) const {
return arenaLists[thingKind].head;
}
bool arenaListsAreEmpty() const {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i) {
#ifdef JS_THREADSAFE
/*
* The arena cannot be empty if the background finalization is not yet
* done.
*/
if (backgroundFinalizeState[i] != BFS_DONE)
return false;
#endif
if (arenaLists[i].head)
return false;
}
return true;
}
#ifdef DEBUG
bool checkArenaListAllUnmarked() const {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i) {
# ifdef JS_THREADSAFE
/* The background finalization must have stopped at this point. */
JS_ASSERT(backgroundFinalizeState[i] == BFS_DONE ||
backgroundFinalizeState[i] == BFS_JUST_FINISHED);
# endif
for (ArenaHeader *aheader = arenaLists[i].head; aheader; aheader = aheader->next) {
if (!aheader->chunk()->bitmap.noBitsSet(aheader))
return false;
}
}
return true;
}
#endif
#ifdef JS_THREADSAFE
bool doneBackgroundFinalize(AllocKind kind) const {
return backgroundFinalizeState[kind] == BFS_DONE;
}
#endif
/*
* Return the free list back to the arena so the GC finalization will not
* run the finalizers over unitialized bytes from free things.
*/
void purge() {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i) {
FreeSpan *headSpan = &freeLists[i];
if (!headSpan->isEmpty()) {
ArenaHeader *aheader = headSpan->arenaHeader();
JS_ASSERT(!aheader->hasFreeThings());
aheader->setFirstFreeSpan(headSpan);
headSpan->initAsEmpty();
}
}
}
/*
* Temporarily copy the free list heads to the arenas so the code can see
* the proper value in ArenaHeader::freeList when accessing the latter
* outside the GC.
*/
void copyFreeListsToArenas() {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i)
copyFreeListToArena(AllocKind(i));
}
void copyFreeListToArena(AllocKind thingKind) {
FreeSpan *headSpan = &freeLists[thingKind];
if (!headSpan->isEmpty()) {
ArenaHeader *aheader = headSpan->arenaHeader();
JS_ASSERT(!aheader->hasFreeThings());
aheader->setFirstFreeSpan(headSpan);
}
}
/*
* Clear the free lists in arenas that were temporarily set there using
* copyToArenas.
*/
void clearFreeListsInArenas() {
for (size_t i = 0; i != FINALIZE_LIMIT; ++i)
clearFreeListInArena(AllocKind(i));
}
void clearFreeListInArena(AllocKind kind) {
FreeSpan *headSpan = &freeLists[kind];
if (!headSpan->isEmpty()) {
ArenaHeader *aheader = headSpan->arenaHeader();
JS_ASSERT(aheader->getFirstFreeSpan().isSameNonEmptySpan(headSpan));
aheader->setAsFullyUsed();
}
}
/*
* Check that the free list is either empty or were synchronized with the
* arena using copyToArena().
*/
bool isSynchronizedFreeList(AllocKind kind) {
FreeSpan *headSpan = &freeLists[kind];
if (headSpan->isEmpty())
return true;
ArenaHeader *aheader = headSpan->arenaHeader();
if (aheader->hasFreeThings()) {
/*
* If the arena has a free list, it must be the same as one in
* lists.
*/
JS_ASSERT(aheader->getFirstFreeSpan().isSameNonEmptySpan(headSpan));
return true;
}
return false;
}
JS_ALWAYS_INLINE void *allocateFromFreeList(AllocKind thingKind, size_t thingSize) {
return freeLists[thingKind].allocate(thingSize);
}
static void *refillFreeList(JSContext *cx, AllocKind thingKind);
void checkEmptyFreeLists() {
#ifdef DEBUG
for (size_t i = 0; i < mozilla::ArrayLength(freeLists); ++i)
JS_ASSERT(freeLists[i].isEmpty());
#endif
}
void checkEmptyFreeList(AllocKind kind) {
JS_ASSERT(freeLists[kind].isEmpty());
}
void finalizeObjects(JSContext *cx);
void finalizeStrings(JSContext *cx);
void finalizeShapes(JSContext *cx);
void finalizeScripts(JSContext *cx);
#ifdef JS_THREADSAFE
static void backgroundFinalize(JSContext *cx, ArenaHeader *listHead);
#endif
private:
inline void finalizeNow(JSContext *cx, AllocKind thingKind);
inline void finalizeLater(JSContext *cx, AllocKind thingKind);
inline void *allocateFromArena(JSCompartment *comp, AllocKind thingKind);
};
/*
* Initial allocation size for data structures holding chunks is set to hold
* chunks with total capacity of 16MB to avoid buffer resizes during browser
* startup.
*/
const size_t INITIAL_CHUNK_CAPACITY = 16 * 1024 * 1024 / ChunkSize;
/* The number of GC cycles an empty chunk can survive before been released. */
const size_t MAX_EMPTY_CHUNK_AGE = 4;
} /* namespace gc */
struct GCPtrHasher
{
typedef void *Lookup;
static HashNumber hash(void *key) {
return HashNumber(uintptr_t(key) >> JS_GCTHING_ZEROBITS);
}
static bool match(void *l, void *k) { return l == k; }
};
typedef HashMap<void *, uint32_t, GCPtrHasher, SystemAllocPolicy> GCLocks;
struct RootInfo {
RootInfo() {}
RootInfo(const char *name, JSGCRootType type) : name(name), type(type) {}
const char *name;
JSGCRootType type;
};
typedef js::HashMap<void *,
RootInfo,
js::DefaultHasher<void *>,
js::SystemAllocPolicy> RootedValueMap;
/* If HashNumber grows, need to change WrapperHasher. */
JS_STATIC_ASSERT(sizeof(HashNumber) == 4);
struct WrapperHasher
{
typedef Value Lookup;
static HashNumber hash(Value key) {
uint64_t bits = key.asRawBits();
return uint32_t(bits) ^ uint32_t(bits >> 32);
}
static bool match(const Value &l, const Value &k) { return l == k; }
};
typedef HashMap<Value, Value, WrapperHasher, SystemAllocPolicy> WrapperMap;
class AutoValueVector;
class AutoIdVector;
} /* namespace js */
#ifdef DEBUG
extern bool
CheckAllocation(JSContext *cx);
#endif
extern JS_FRIEND_API(JSGCTraceKind)
js_GetGCThingTraceKind(void *thing);
extern JSBool
js_InitGC(JSRuntime *rt, uint32_t maxbytes);
extern void
js_FinishGC(JSRuntime *rt);
extern JSBool
js_AddRoot(JSContext *cx, js::Value *vp, const char *name);
extern JSBool
js_AddGCThingRoot(JSContext *cx, void **rp, const char *name);
#ifdef DEBUG
extern void
js_DumpNamedRoots(JSRuntime *rt,
void (*dump)(const char *name, void *rp, JSGCRootType type, void *data),
void *data);
#endif
extern uint32_t
js_MapGCRoots(JSRuntime *rt, JSGCRootMapFun map, void *data);
/* Table of pointers with count valid members. */
typedef struct JSPtrTable {
size_t count;
void **array;
} JSPtrTable;
extern JSBool
js_LockGCThingRT(JSRuntime *rt, void *thing);
extern void
js_UnlockGCThingRT(JSRuntime *rt, void *thing);
extern JS_FRIEND_API(bool)
IsAboutToBeFinalized(JSContext *cx, const js::gc::Cell *thing);
extern bool
IsAboutToBeFinalized(JSContext *cx, const js::Value &value);
extern JS_FRIEND_API(bool)
js_GCThingIsMarked(void *thing, uintN color);
extern void
js_TraceStackFrame(JSTracer *trc, js::StackFrame *fp);
namespace js {
extern JS_REQUIRES_STACK void
MarkRuntime(JSTracer *trc);
extern void
TraceRuntime(JSTracer *trc);
extern JS_REQUIRES_STACK JS_FRIEND_API(void)
MarkContext(JSTracer *trc, JSContext *acx);
/* Must be called with GC lock taken. */
extern void
TriggerGC(JSRuntime *rt, js::gcstats::Reason reason);
/* Must be called with GC lock taken. */
extern void
TriggerCompartmentGC(JSCompartment *comp, js::gcstats::Reason reason);
extern void
MaybeGC(JSContext *cx);
} /* namespace js */
/*
* Kinds of js_GC invocation.
*/
typedef enum JSGCInvocationKind {
/* Normal invocation. */
GC_NORMAL = 0,
/*
* Called from js_DestroyContext for last JSContext in a JSRuntime, when
* it is imperative that rt->gcPoke gets cleared early in js_GC.
*/
GC_LAST_CONTEXT = 1,
/* Minimize GC triggers and release empty GC chunks right away. */
GC_SHRINK = 2
} JSGCInvocationKind;
/* Pass NULL for |comp| to get a full GC. */
extern void
js_GC(JSContext *cx, JSCompartment *comp, JSGCInvocationKind gckind, js::gcstats::Reason r);
#ifdef JS_THREADSAFE
/*
* This is a helper for code at can potentially run outside JS request to
* ensure that the GC is not running when the function returns.
*
* This function must be called with the GC lock held.
*/
extern void
js_WaitForGC(JSRuntime *rt);
#else /* !JS_THREADSAFE */
# define js_WaitForGC(rt) ((void) 0)
#endif
namespace js {
#ifdef JS_THREADSAFE
class GCHelperThread {
enum State {
IDLE,
SWEEPING,
ALLOCATING,
CANCEL_ALLOCATION,
SHUTDOWN
};
/*
* During the finalization we do not free immediately. Rather we add the
* corresponding pointers to a buffer which we later release on a
* separated thread.
*
* The buffer is implemented as a vector of 64K arrays of pointers, not as
* a simple vector, to avoid realloc calls during the vector growth and to
* not bloat the binary size of the inlined freeLater method. Any OOM
* during buffer growth results in the pointer being freed immediately.
*/
static const size_t FREE_ARRAY_SIZE = size_t(1) << 16;
static const size_t FREE_ARRAY_LENGTH = FREE_ARRAY_SIZE / sizeof(void *);
JSRuntime *const rt;
PRThread *thread;
PRCondVar *wakeup;
PRCondVar *done;
volatile State state;
JSContext *context;
bool shrinkFlag;
Vector<void **, 16, js::SystemAllocPolicy> freeVector;
void **freeCursor;
void **freeCursorEnd;
Vector<js::gc::ArenaHeader *, 64, js::SystemAllocPolicy> finalizeVector;
bool backgroundAllocation;
friend struct js::gc::ArenaLists;
JS_FRIEND_API(void)
replenishAndFreeLater(void *ptr);
static void freeElementsAndArray(void **array, void **end) {
JS_ASSERT(array <= end);
for (void **p = array; p != end; ++p)
js::Foreground::free_(*p);
js::Foreground::free_(array);
}
static void threadMain(void* arg);
void threadLoop();
/* Must be called with the GC lock taken. */
void doSweep();
public:
GCHelperThread(JSRuntime *rt)
: rt(rt),
thread(NULL),
wakeup(NULL),
done(NULL),
state(IDLE),
freeCursor(NULL),
freeCursorEnd(NULL),
backgroundAllocation(true)
{ }
bool init();
void finish();
/* Must be called with the GC lock taken. */
inline void startBackgroundSweep(bool shouldShrink);
/* Must be called with the GC lock taken. */
void waitBackgroundSweepEnd();
/* Must be called with the GC lock taken. */
void waitBackgroundSweepOrAllocEnd();
/* Must be called with the GC lock taken. */
inline void startBackgroundAllocationIfIdle();
bool canBackgroundAllocate() const {
return backgroundAllocation;
}
void disableBackgroundAllocation() {
backgroundAllocation = false;
}
PRThread *getThread() const {
return thread;
}
/*
* Outside the GC lock may give true answer when in fact the sweeping has
* been done.
*/
bool sweeping() const {
return state == SWEEPING;
}
bool shouldShrink() const {
JS_ASSERT(sweeping());
return shrinkFlag;
}
void freeLater(void *ptr) {
JS_ASSERT(!sweeping());
if (freeCursor != freeCursorEnd)
*freeCursor++ = ptr;
else
replenishAndFreeLater(ptr);
}
/* Must be called with the GC lock taken. */
bool prepareForBackgroundSweep(JSContext *cx);
};
#endif /* JS_THREADSAFE */
struct GCChunkHasher {
typedef gc::Chunk *Lookup;
/*
* Strip zeros for better distribution after multiplying by the golden
* ratio.
*/
static HashNumber hash(gc::Chunk *chunk) {
JS_ASSERT(!(jsuword(chunk) & gc::ChunkMask));
return HashNumber(jsuword(chunk) >> gc::ChunkShift);
}
static bool match(gc::Chunk *k, gc::Chunk *l) {
JS_ASSERT(!(jsuword(k) & gc::ChunkMask));
JS_ASSERT(!(jsuword(l) & gc::ChunkMask));
return k == l;
}
};
typedef HashSet<js::gc::Chunk *, GCChunkHasher, SystemAllocPolicy> GCChunkSet;
struct ConservativeGCThreadData {
/*
* The GC scans conservatively between ThreadData::nativeStackBase and
* nativeStackTop unless the latter is NULL.
*/
jsuword *nativeStackTop;
union {
jmp_buf jmpbuf;
jsuword words[JS_HOWMANY(sizeof(jmp_buf), sizeof(jsuword))];
} registerSnapshot;
/*
* Cycle collector uses this to communicate that the native stack of the
* GC thread should be scanned only if the thread have more than the given
* threshold of requests.
*/
unsigned requestThreshold;
ConservativeGCThreadData()
: nativeStackTop(NULL), requestThreshold(0)
{
}
~ConservativeGCThreadData() {
#ifdef JS_THREADSAFE
/*
* The conservative GC scanner should be disabled when the thread leaves
* the last request.
*/
JS_ASSERT(!hasStackToScan());
#endif
}
JS_NEVER_INLINE void recordStackTop();
#ifdef JS_THREADSAFE
void updateForRequestEnd(unsigned suspendCount) {
if (suspendCount)
recordStackTop();
else
nativeStackTop = NULL;
}
#endif
bool hasStackToScan() const {
return !!nativeStackTop;
}
};
template<class T>
struct MarkStack {
T *stack;
T *tos;
T *limit;
bool push(T item) {
if (tos == limit)
return false;
*tos++ = item;
return true;
}
bool push(T item1, T item2, T item3) {
T *nextTos = tos + 3;
if (nextTos > limit)
return false;
tos[0] = item1;
tos[1] = item2;
tos[2] = item3;
tos = nextTos;
return true;
}
bool isEmpty() const {
return tos == stack;
}
T pop() {
JS_ASSERT(!isEmpty());
return *--tos;
}
template<size_t N>
MarkStack(T (&buffer)[N])
: stack(buffer),
tos(buffer),
limit(buffer + N) { }
};
static const size_t MARK_STACK_LENGTH = 32768;
struct GCMarker : public JSTracer {
/*
* We use a common mark stack to mark GC things of different types and use
* the explicit tags to distinguish them when it cannot be deduced from
* the context of push or pop operation.
*
* Currently we need only 4 tags. However that can be extended to 8 if
* necessary as we tag only GC things.
*/
enum StackTag {
ValueArrayTag,
ObjectTag,
TypeTag,
XmlTag,
LastTag = XmlTag
};
static const uintptr_t StackTagMask = 3;
static void staticAsserts() {
JS_STATIC_ASSERT(StackTagMask >= uintptr_t(LastTag));
JS_STATIC_ASSERT(StackTagMask <= gc::Cell::CellMask);
}
private:
/* The color is only applied to objects, functions and xml. */
uint32_t color;
public:
/* Pointer to the top of the stack of arenas we are delaying marking on. */
js::gc::Arena *unmarkedArenaStackTop;
/* Count of arenas that are currently in the stack. */
DebugOnly<size_t> markLaterArenas;
#ifdef JS_DUMP_CONSERVATIVE_GC_ROOTS
js::gc::ConservativeGCStats conservativeStats;
Vector<void *, 0, SystemAllocPolicy> conservativeRoots;
const char *conservativeDumpFileName;
void dumpConservativeRoots();
#endif
MarkStack<uintptr_t> stack;
public:
explicit GCMarker(JSContext *cx);
~GCMarker();
uint32_t getMarkColor() const {
return color;
}
/*
* The only valid color transition during a GC is from black to gray. It is
* wrong to switch the mark color from gray to black. The reason is that the
* cycle collector depends on the invariant that there are no black to gray
* edges in the GC heap. This invariant lets the CC not trace through black
* objects. If this invariant is violated, the cycle collector may free
* objects that are still reachable.
*/
void setMarkColorGray() {
JS_ASSERT(color == gc::BLACK);
color = gc::GRAY;
}
void delayMarkingChildren(const void *thing);
bool hasDelayedChildren() const {
return !!unmarkedArenaStackTop;
}
void markDelayedChildren();
bool isMarkStackEmpty() {
return stack.isEmpty();
}
void drainMarkStack();
inline void processMarkStackTop();
void pushObject(JSObject *obj) {
pushTaggedPtr(ObjectTag, obj);
}
void pushType(types::TypeObject *type) {
pushTaggedPtr(TypeTag, type);
}
void pushXML(JSXML *xml) {
pushTaggedPtr(XmlTag, xml);
}
void pushTaggedPtr(StackTag tag, void *ptr) {
uintptr_t addr = reinterpret_cast<uintptr_t>(ptr);
JS_ASSERT(!(addr & StackTagMask));
if (!stack.push(addr | uintptr_t(tag)))
delayMarkingChildren(ptr);
}
};
void
MarkStackRangeConservatively(JSTracer *trc, Value *begin, Value *end);
typedef void (*IterateCompartmentCallback)(JSContext *cx, void *data, JSCompartment *compartment);
typedef void (*IterateChunkCallback)(JSContext *cx, void *data, gc::Chunk *chunk);
typedef void (*IterateArenaCallback)(JSContext *cx, void *data, gc::Arena *arena,
JSGCTraceKind traceKind, size_t thingSize);
typedef void (*IterateCellCallback)(JSContext *cx, void *data, void *thing,
JSGCTraceKind traceKind, size_t thingSize);
/*
* This function calls |compartmentCallback| on every compartment,
* |arenaCallback| on every in-use arena, and |cellCallback| on every in-use
* cell in the GC heap.
*/
extern JS_FRIEND_API(void)
IterateCompartmentsArenasCells(JSContext *cx, void *data,
IterateCompartmentCallback compartmentCallback,
IterateArenaCallback arenaCallback,
IterateCellCallback cellCallback);
/*
* Invoke chunkCallback on every in-use chunk.
*/
extern JS_FRIEND_API(void)
IterateChunks(JSContext *cx, void *data, IterateChunkCallback chunkCallback);
/*
* Invoke cellCallback on every in-use object of the specified thing kind for
* the given compartment or for all compartments if it is null.
*/
extern JS_FRIEND_API(void)
IterateCells(JSContext *cx, JSCompartment *compartment, gc::AllocKind thingKind,
void *data, IterateCellCallback cellCallback);
} /* namespace js */
extern void
js_FinalizeStringRT(JSRuntime *rt, JSString *str);
/*
* Macro to test if a traversal is the marking phase of the GC.
*/
#define IS_GC_MARKING_TRACER(trc) ((trc)->callback == NULL)
namespace js {
namespace gc {
JSCompartment *
NewCompartment(JSContext *cx, JSPrincipals *principals);
/* Tries to run a GC no matter what (used for GC zeal). */
void
RunDebugGC(JSContext *cx);
const int ZealPokeThreshold = 1;
const int ZealAllocThreshold = 2;
const int ZealVerifierThreshold = 4;
#ifdef JS_GC_ZEAL
/* Check that write barriers have been used correctly. See jsgc.cpp. */
void
VerifyBarriers(JSContext *cx, bool always = false);
#else
static inline void
VerifyBarriers(JSContext *cx, bool always = false)
{
}
#endif
} /* namespace gc */
static inline JSCompartment *
GetObjectCompartment(JSObject *obj) { return reinterpret_cast<js::gc::Cell *>(obj)->compartment(); }
} /* namespace js */
#endif /* jsgc_h___ */