1 /*
2 * Copyright (C) 2011 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 * http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17 #include "mark_sweep.h"
18
19 #include <atomic>
20 #include <climits>
21 #include <functional>
22 #include <numeric>
23 #include <vector>
24
25 #include "base/bounded_fifo.h"
26 #include "base/enums.h"
27 #include "base/file_utils.h"
28 #include "base/logging.h" // For VLOG.
29 #include "base/macros.h"
30 #include "base/mutex-inl.h"
31 #include "base/systrace.h"
32 #include "base/time_utils.h"
33 #include "base/timing_logger.h"
34 #include "gc/accounting/card_table-inl.h"
35 #include "gc/accounting/heap_bitmap-inl.h"
36 #include "gc/accounting/mod_union_table.h"
37 #include "gc/accounting/space_bitmap-inl.h"
38 #include "gc/heap.h"
39 #include "gc/reference_processor.h"
40 #include "gc/space/large_object_space.h"
41 #include "gc/space/space-inl.h"
42 #include "mark_sweep-inl.h"
43 #include "mirror/object-inl.h"
44 #include "runtime.h"
45 #include "scoped_thread_state_change-inl.h"
46 #include "thread-current-inl.h"
47 #include "thread_list.h"
48
49 namespace art {
50 namespace gc {
51 namespace collector {
52
53 // Performance options.
54 static constexpr bool kUseRecursiveMark = false;
55 static constexpr bool kUseMarkStackPrefetch = true;
56 static constexpr size_t kSweepArrayChunkFreeSize = 1024;
57 static constexpr bool kPreCleanCards = true;
58
59 // Parallelism options.
60 static constexpr bool kParallelCardScan = true;
61 static constexpr bool kParallelRecursiveMark = true;
62 // Don't attempt to parallelize mark stack processing unless the mark stack is at least n
63 // elements. This is temporary until we reduce the overhead caused by allocating tasks, etc.. Not
64 // having this can add overhead in ProcessReferences since we may end up doing many calls of
65 // ProcessMarkStack with very small mark stacks.
66 static constexpr size_t kMinimumParallelMarkStackSize = 128;
67 static constexpr bool kParallelProcessMarkStack = true;
68
69 // Profiling and information flags.
70 static constexpr bool kProfileLargeObjects = false;
71 static constexpr bool kMeasureOverhead = false;
72 static constexpr bool kCountTasks = false;
73 static constexpr bool kCountMarkedObjects = false;
74
75 // Turn off kCheckLocks when profiling the GC since it slows the GC down by up to 40%.
76 static constexpr bool kCheckLocks = kDebugLocking;
77 static constexpr bool kVerifyRootsMarked = kIsDebugBuild;
78
79 // If true, revoke the rosalloc thread-local buffers at the
80 // checkpoint, as opposed to during the pause.
81 static constexpr bool kRevokeRosAllocThreadLocalBuffersAtCheckpoint = true;
82
BindBitmaps()83 void MarkSweep::BindBitmaps() {
84 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
85 WriterMutexLock mu(Thread::Current(), *Locks::heap_bitmap_lock_);
86 // Mark all of the spaces we never collect as immune.
87 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
88 if (space->GetGcRetentionPolicy() == space::kGcRetentionPolicyNeverCollect) {
89 immune_spaces_.AddSpace(space);
90 }
91 }
92 }
93
MarkSweep(Heap * heap,bool is_concurrent,const std::string & name_prefix)94 MarkSweep::MarkSweep(Heap* heap, bool is_concurrent, const std::string& name_prefix)
95 : GarbageCollector(heap,
96 name_prefix +
97 (is_concurrent ? "concurrent mark sweep": "mark sweep")),
98 current_space_bitmap_(nullptr),
99 mark_bitmap_(nullptr),
100 mark_stack_(nullptr),
101 gc_barrier_(new Barrier(0)),
102 mark_stack_lock_("mark sweep mark stack lock", kMarkSweepMarkStackLock),
103 is_concurrent_(is_concurrent),
104 live_stack_freeze_size_(0) {
105 std::string error_msg;
106 MemMap* mem_map = MemMap::MapAnonymous(
107 "mark sweep sweep array free buffer", nullptr,
108 RoundUp(kSweepArrayChunkFreeSize * sizeof(mirror::Object*), kPageSize),
109 PROT_READ | PROT_WRITE, false, false, &error_msg);
110 CHECK(mem_map != nullptr) << "Couldn't allocate sweep array free buffer: " << error_msg;
111 sweep_array_free_buffer_mem_map_.reset(mem_map);
112 }
113
InitializePhase()114 void MarkSweep::InitializePhase() {
115 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
116 mark_stack_ = heap_->GetMarkStack();
117 DCHECK(mark_stack_ != nullptr);
118 immune_spaces_.Reset();
119 no_reference_class_count_.StoreRelaxed(0);
120 normal_count_.StoreRelaxed(0);
121 class_count_.StoreRelaxed(0);
122 object_array_count_.StoreRelaxed(0);
123 other_count_.StoreRelaxed(0);
124 reference_count_.StoreRelaxed(0);
125 large_object_test_.StoreRelaxed(0);
126 large_object_mark_.StoreRelaxed(0);
127 overhead_time_ .StoreRelaxed(0);
128 work_chunks_created_.StoreRelaxed(0);
129 work_chunks_deleted_.StoreRelaxed(0);
130 mark_null_count_.StoreRelaxed(0);
131 mark_immune_count_.StoreRelaxed(0);
132 mark_fastpath_count_.StoreRelaxed(0);
133 mark_slowpath_count_.StoreRelaxed(0);
134 {
135 // TODO: I don't think we should need heap bitmap lock to Get the mark bitmap.
136 ReaderMutexLock mu(Thread::Current(), *Locks::heap_bitmap_lock_);
137 mark_bitmap_ = heap_->GetMarkBitmap();
138 }
139 if (!GetCurrentIteration()->GetClearSoftReferences()) {
140 // Always clear soft references if a non-sticky collection.
141 GetCurrentIteration()->SetClearSoftReferences(GetGcType() != collector::kGcTypeSticky);
142 }
143 }
144
RunPhases()145 void MarkSweep::RunPhases() {
146 Thread* self = Thread::Current();
147 InitializePhase();
148 Locks::mutator_lock_->AssertNotHeld(self);
149 if (IsConcurrent()) {
150 GetHeap()->PreGcVerification(this);
151 {
152 ReaderMutexLock mu(self, *Locks::mutator_lock_);
153 MarkingPhase();
154 }
155 ScopedPause pause(this);
156 GetHeap()->PrePauseRosAllocVerification(this);
157 PausePhase();
158 RevokeAllThreadLocalBuffers();
159 } else {
160 ScopedPause pause(this);
161 GetHeap()->PreGcVerificationPaused(this);
162 MarkingPhase();
163 GetHeap()->PrePauseRosAllocVerification(this);
164 PausePhase();
165 RevokeAllThreadLocalBuffers();
166 }
167 {
168 // Sweeping always done concurrently, even for non concurrent mark sweep.
169 ReaderMutexLock mu(self, *Locks::mutator_lock_);
170 ReclaimPhase();
171 }
172 GetHeap()->PostGcVerification(this);
173 FinishPhase();
174 }
175
ProcessReferences(Thread * self)176 void MarkSweep::ProcessReferences(Thread* self) {
177 WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
178 GetHeap()->GetReferenceProcessor()->ProcessReferences(
179 true,
180 GetTimings(),
181 GetCurrentIteration()->GetClearSoftReferences(),
182 this);
183 }
184
PausePhase()185 void MarkSweep::PausePhase() {
186 TimingLogger::ScopedTiming t("(Paused)PausePhase", GetTimings());
187 Thread* self = Thread::Current();
188 Locks::mutator_lock_->AssertExclusiveHeld(self);
189 if (IsConcurrent()) {
190 // Handle the dirty objects if we are a concurrent GC.
191 WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
192 // Re-mark root set.
193 ReMarkRoots();
194 // Scan dirty objects, this is only required if we are doing concurrent GC.
195 RecursiveMarkDirtyObjects(true, accounting::CardTable::kCardDirty);
196 }
197 {
198 TimingLogger::ScopedTiming t2("SwapStacks", GetTimings());
199 WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
200 heap_->SwapStacks();
201 live_stack_freeze_size_ = heap_->GetLiveStack()->Size();
202 // Need to revoke all the thread local allocation stacks since we just swapped the allocation
203 // stacks and don't want anybody to allocate into the live stack.
204 RevokeAllThreadLocalAllocationStacks(self);
205 }
206 heap_->PreSweepingGcVerification(this);
207 // Disallow new system weaks to prevent a race which occurs when someone adds a new system
208 // weak before we sweep them. Since this new system weak may not be marked, the GC may
209 // incorrectly sweep it. This also fixes a race where interning may attempt to return a strong
210 // reference to a string that is about to be swept.
211 Runtime::Current()->DisallowNewSystemWeaks();
212 // Enable the reference processing slow path, needs to be done with mutators paused since there
213 // is no lock in the GetReferent fast path.
214 GetHeap()->GetReferenceProcessor()->EnableSlowPath();
215 }
216
PreCleanCards()217 void MarkSweep::PreCleanCards() {
218 // Don't do this for non concurrent GCs since they don't have any dirty cards.
219 if (kPreCleanCards && IsConcurrent()) {
220 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
221 Thread* self = Thread::Current();
222 CHECK(!Locks::mutator_lock_->IsExclusiveHeld(self));
223 // Process dirty cards and add dirty cards to mod union tables, also ages cards.
224 heap_->ProcessCards(GetTimings(), false, true, false);
225 // The checkpoint root marking is required to avoid a race condition which occurs if the
226 // following happens during a reference write:
227 // 1. mutator dirties the card (write barrier)
228 // 2. GC ages the card (the above ProcessCards call)
229 // 3. GC scans the object (the RecursiveMarkDirtyObjects call below)
230 // 4. mutator writes the value (corresponding to the write barrier in 1.)
231 // This causes the GC to age the card but not necessarily mark the reference which the mutator
232 // wrote into the object stored in the card.
233 // Having the checkpoint fixes this issue since it ensures that the card mark and the
234 // reference write are visible to the GC before the card is scanned (this is due to locks being
235 // acquired / released in the checkpoint code).
236 // The other roots are also marked to help reduce the pause.
237 MarkRootsCheckpoint(self, false);
238 MarkNonThreadRoots();
239 MarkConcurrentRoots(
240 static_cast<VisitRootFlags>(kVisitRootFlagClearRootLog | kVisitRootFlagNewRoots));
241 // Process the newly aged cards.
242 RecursiveMarkDirtyObjects(false, accounting::CardTable::kCardDirty - 1);
243 // TODO: Empty allocation stack to reduce the number of objects we need to test / mark as live
244 // in the next GC.
245 }
246 }
247
RevokeAllThreadLocalAllocationStacks(Thread * self)248 void MarkSweep::RevokeAllThreadLocalAllocationStacks(Thread* self) {
249 if (kUseThreadLocalAllocationStack) {
250 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
251 Locks::mutator_lock_->AssertExclusiveHeld(self);
252 heap_->RevokeAllThreadLocalAllocationStacks(self);
253 }
254 }
255
MarkingPhase()256 void MarkSweep::MarkingPhase() {
257 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
258 Thread* self = Thread::Current();
259 BindBitmaps();
260 FindDefaultSpaceBitmap();
261 // Process dirty cards and add dirty cards to mod union tables.
262 // If the GC type is non sticky, then we just clear the cards of the
263 // alloc space instead of aging them.
264 //
265 // Note that it is fine to clear the cards of the alloc space here,
266 // in the case of a concurrent (non-sticky) mark-sweep GC (whose
267 // marking phase _is_ performed concurrently with mutator threads
268 // running and possibly dirtying cards), as the whole alloc space
269 // will be traced in that case, starting *after* this call to
270 // Heap::ProcessCards (see calls to MarkSweep::MarkRoots and
271 // MarkSweep::MarkReachableObjects). References held by objects on
272 // cards that became dirty *after* the actual marking work started
273 // will be marked in the pause (see MarkSweep::PausePhase), in a
274 // *non-concurrent* way to prevent races with mutator threads.
275 //
276 // TODO: Do we need some sort of fence between the call to
277 // Heap::ProcessCard and the calls to MarkSweep::MarkRoot /
278 // MarkSweep::MarkReachableObjects below to make sure write
279 // operations in the card table clearing the alloc space's dirty
280 // cards (during the call to Heap::ProcessCard) are not reordered
281 // *after* marking actually starts?
282 heap_->ProcessCards(GetTimings(),
283 /* use_rem_sets */ false,
284 /* process_alloc_space_cards */ true,
285 /* clear_alloc_space_cards */ GetGcType() != kGcTypeSticky);
286 WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
287 MarkRoots(self);
288 MarkReachableObjects();
289 // Pre-clean dirtied cards to reduce pauses.
290 PreCleanCards();
291 }
292
293 class MarkSweep::ScanObjectVisitor {
294 public:
ScanObjectVisitor(MarkSweep * const mark_sweep)295 explicit ScanObjectVisitor(MarkSweep* const mark_sweep) ALWAYS_INLINE
296 : mark_sweep_(mark_sweep) {}
297
operator ()(ObjPtr<mirror::Object> obj) const298 void operator()(ObjPtr<mirror::Object> obj) const
299 ALWAYS_INLINE
300 REQUIRES(Locks::heap_bitmap_lock_)
301 REQUIRES_SHARED(Locks::mutator_lock_) {
302 if (kCheckLocks) {
303 Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
304 Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
305 }
306 mark_sweep_->ScanObject(obj.Ptr());
307 }
308
309 private:
310 MarkSweep* const mark_sweep_;
311 };
312
UpdateAndMarkModUnion()313 void MarkSweep::UpdateAndMarkModUnion() {
314 for (const auto& space : immune_spaces_.GetSpaces()) {
315 const char* name = space->IsZygoteSpace()
316 ? "UpdateAndMarkZygoteModUnionTable"
317 : "UpdateAndMarkImageModUnionTable";
318 DCHECK(space->IsZygoteSpace() || space->IsImageSpace()) << *space;
319 TimingLogger::ScopedTiming t(name, GetTimings());
320 accounting::ModUnionTable* mod_union_table = heap_->FindModUnionTableFromSpace(space);
321 if (mod_union_table != nullptr) {
322 mod_union_table->UpdateAndMarkReferences(this);
323 } else {
324 // No mod-union table, scan all the live bits. This can only occur for app images.
325 space->GetLiveBitmap()->VisitMarkedRange(reinterpret_cast<uintptr_t>(space->Begin()),
326 reinterpret_cast<uintptr_t>(space->End()),
327 ScanObjectVisitor(this));
328 }
329 }
330 }
331
MarkReachableObjects()332 void MarkSweep::MarkReachableObjects() {
333 UpdateAndMarkModUnion();
334 // Recursively mark all the non-image bits set in the mark bitmap.
335 RecursiveMark();
336 }
337
ReclaimPhase()338 void MarkSweep::ReclaimPhase() {
339 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
340 Thread* const self = Thread::Current();
341 // Process the references concurrently.
342 ProcessReferences(self);
343 SweepSystemWeaks(self);
344 Runtime* const runtime = Runtime::Current();
345 runtime->AllowNewSystemWeaks();
346 // Clean up class loaders after system weaks are swept since that is how we know if class
347 // unloading occurred.
348 runtime->GetClassLinker()->CleanupClassLoaders();
349 {
350 WriterMutexLock mu(self, *Locks::heap_bitmap_lock_);
351 GetHeap()->RecordFreeRevoke();
352 // Reclaim unmarked objects.
353 Sweep(false);
354 // Swap the live and mark bitmaps for each space which we modified space. This is an
355 // optimization that enables us to not clear live bits inside of the sweep. Only swaps unbound
356 // bitmaps.
357 SwapBitmaps();
358 // Unbind the live and mark bitmaps.
359 GetHeap()->UnBindBitmaps();
360 }
361 }
362
FindDefaultSpaceBitmap()363 void MarkSweep::FindDefaultSpaceBitmap() {
364 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
365 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
366 accounting::ContinuousSpaceBitmap* bitmap = space->GetMarkBitmap();
367 // We want to have the main space instead of non moving if possible.
368 if (bitmap != nullptr &&
369 space->GetGcRetentionPolicy() == space::kGcRetentionPolicyAlwaysCollect) {
370 current_space_bitmap_ = bitmap;
371 // If we are not the non moving space exit the loop early since this will be good enough.
372 if (space != heap_->GetNonMovingSpace()) {
373 break;
374 }
375 }
376 }
377 CHECK(current_space_bitmap_ != nullptr) << "Could not find a default mark bitmap\n"
378 << heap_->DumpSpaces();
379 }
380
ExpandMarkStack()381 void MarkSweep::ExpandMarkStack() {
382 ResizeMarkStack(mark_stack_->Capacity() * 2);
383 }
384
ResizeMarkStack(size_t new_size)385 void MarkSweep::ResizeMarkStack(size_t new_size) {
386 // Rare case, no need to have Thread::Current be a parameter.
387 if (UNLIKELY(mark_stack_->Size() < mark_stack_->Capacity())) {
388 // Someone else acquired the lock and expanded the mark stack before us.
389 return;
390 }
391 std::vector<StackReference<mirror::Object>> temp(mark_stack_->Begin(), mark_stack_->End());
392 CHECK_LE(mark_stack_->Size(), new_size);
393 mark_stack_->Resize(new_size);
394 for (auto& obj : temp) {
395 mark_stack_->PushBack(obj.AsMirrorPtr());
396 }
397 }
398
MarkObject(mirror::Object * obj)399 mirror::Object* MarkSweep::MarkObject(mirror::Object* obj) {
400 MarkObject(obj, nullptr, MemberOffset(0));
401 return obj;
402 }
403
MarkObjectNonNullParallel(mirror::Object * obj)404 inline void MarkSweep::MarkObjectNonNullParallel(mirror::Object* obj) {
405 DCHECK(obj != nullptr);
406 if (MarkObjectParallel(obj)) {
407 MutexLock mu(Thread::Current(), mark_stack_lock_);
408 if (UNLIKELY(mark_stack_->Size() >= mark_stack_->Capacity())) {
409 ExpandMarkStack();
410 }
411 // The object must be pushed on to the mark stack.
412 mark_stack_->PushBack(obj);
413 }
414 }
415
IsNullOrMarkedHeapReference(mirror::HeapReference<mirror::Object> * ref,bool do_atomic_update ATTRIBUTE_UNUSED)416 bool MarkSweep::IsNullOrMarkedHeapReference(mirror::HeapReference<mirror::Object>* ref,
417 bool do_atomic_update ATTRIBUTE_UNUSED) {
418 mirror::Object* obj = ref->AsMirrorPtr();
419 if (obj == nullptr) {
420 return true;
421 }
422 return IsMarked(obj);
423 }
424
425 class MarkSweep::MarkObjectSlowPath {
426 public:
MarkObjectSlowPath(MarkSweep * mark_sweep,mirror::Object * holder=nullptr,MemberOffset offset=MemberOffset (0))427 explicit MarkObjectSlowPath(MarkSweep* mark_sweep,
428 mirror::Object* holder = nullptr,
429 MemberOffset offset = MemberOffset(0))
430 : mark_sweep_(mark_sweep),
431 holder_(holder),
432 offset_(offset) {}
433
operator ()(const mirror::Object * obj) const434 void operator()(const mirror::Object* obj) const NO_THREAD_SAFETY_ANALYSIS {
435 if (kProfileLargeObjects) {
436 // TODO: Differentiate between marking and testing somehow.
437 ++mark_sweep_->large_object_test_;
438 ++mark_sweep_->large_object_mark_;
439 }
440 space::LargeObjectSpace* large_object_space = mark_sweep_->GetHeap()->GetLargeObjectsSpace();
441 if (UNLIKELY(obj == nullptr || !IsAligned<kPageSize>(obj) ||
442 (kIsDebugBuild && large_object_space != nullptr &&
443 !large_object_space->Contains(obj)))) {
444 // Lowest priority logging first:
445 PrintFileToLog("/proc/self/maps", LogSeverity::FATAL_WITHOUT_ABORT);
446 MemMap::DumpMaps(LOG_STREAM(FATAL_WITHOUT_ABORT), true);
447 // Buffer the output in the string stream since it is more important than the stack traces
448 // and we want it to have log priority. The stack traces are printed from Runtime::Abort
449 // which is called from LOG(FATAL) but before the abort message.
450 std::ostringstream oss;
451 oss << "Tried to mark " << obj << " not contained by any spaces" << std::endl;
452 if (holder_ != nullptr) {
453 size_t holder_size = holder_->SizeOf();
454 ArtField* field = holder_->FindFieldByOffset(offset_);
455 oss << "Field info: "
456 << " holder=" << holder_
457 << " holder is "
458 << (mark_sweep_->GetHeap()->IsLiveObjectLocked(holder_)
459 ? "alive" : "dead")
460 << " holder_size=" << holder_size
461 << " holder_type=" << holder_->PrettyTypeOf()
462 << " offset=" << offset_.Uint32Value()
463 << " field=" << (field != nullptr ? field->GetName() : "nullptr")
464 << " field_type="
465 << (field != nullptr ? field->GetTypeDescriptor() : "")
466 << " first_ref_field_offset="
467 << (holder_->IsClass()
468 ? holder_->AsClass()->GetFirstReferenceStaticFieldOffset(
469 kRuntimePointerSize)
470 : holder_->GetClass()->GetFirstReferenceInstanceFieldOffset())
471 << " num_of_ref_fields="
472 << (holder_->IsClass()
473 ? holder_->AsClass()->NumReferenceStaticFields()
474 : holder_->GetClass()->NumReferenceInstanceFields())
475 << std::endl;
476 // Print the memory content of the holder.
477 for (size_t i = 0; i < holder_size / sizeof(uint32_t); ++i) {
478 uint32_t* p = reinterpret_cast<uint32_t*>(holder_);
479 oss << &p[i] << ": " << "holder+" << (i * sizeof(uint32_t)) << " = " << std::hex << p[i]
480 << std::endl;
481 }
482 }
483 oss << "Attempting see if it's a bad thread root" << std::endl;
484 mark_sweep_->VerifySuspendedThreadRoots(oss);
485 LOG(FATAL) << oss.str();
486 }
487 }
488
489 private:
490 MarkSweep* const mark_sweep_;
491 mirror::Object* const holder_;
492 MemberOffset offset_;
493 };
494
MarkObjectNonNull(mirror::Object * obj,mirror::Object * holder,MemberOffset offset)495 inline void MarkSweep::MarkObjectNonNull(mirror::Object* obj,
496 mirror::Object* holder,
497 MemberOffset offset) {
498 DCHECK(obj != nullptr);
499 if (kUseBakerReadBarrier) {
500 // Verify all the objects have the correct state installed.
501 obj->AssertReadBarrierState();
502 }
503 if (immune_spaces_.IsInImmuneRegion(obj)) {
504 if (kCountMarkedObjects) {
505 ++mark_immune_count_;
506 }
507 DCHECK(mark_bitmap_->Test(obj));
508 } else if (LIKELY(current_space_bitmap_->HasAddress(obj))) {
509 if (kCountMarkedObjects) {
510 ++mark_fastpath_count_;
511 }
512 if (UNLIKELY(!current_space_bitmap_->Set(obj))) {
513 PushOnMarkStack(obj); // This object was not previously marked.
514 }
515 } else {
516 if (kCountMarkedObjects) {
517 ++mark_slowpath_count_;
518 }
519 MarkObjectSlowPath visitor(this, holder, offset);
520 // TODO: We already know that the object is not in the current_space_bitmap_ but MarkBitmap::Set
521 // will check again.
522 if (!mark_bitmap_->Set(obj, visitor)) {
523 PushOnMarkStack(obj); // Was not already marked, push.
524 }
525 }
526 }
527
PushOnMarkStack(mirror::Object * obj)528 inline void MarkSweep::PushOnMarkStack(mirror::Object* obj) {
529 if (UNLIKELY(mark_stack_->Size() >= mark_stack_->Capacity())) {
530 // Lock is not needed but is here anyways to please annotalysis.
531 MutexLock mu(Thread::Current(), mark_stack_lock_);
532 ExpandMarkStack();
533 }
534 // The object must be pushed on to the mark stack.
535 mark_stack_->PushBack(obj);
536 }
537
MarkObjectParallel(mirror::Object * obj)538 inline bool MarkSweep::MarkObjectParallel(mirror::Object* obj) {
539 DCHECK(obj != nullptr);
540 if (kUseBakerReadBarrier) {
541 // Verify all the objects have the correct state installed.
542 obj->AssertReadBarrierState();
543 }
544 if (immune_spaces_.IsInImmuneRegion(obj)) {
545 DCHECK(IsMarked(obj) != nullptr);
546 return false;
547 }
548 // Try to take advantage of locality of references within a space, failing this find the space
549 // the hard way.
550 accounting::ContinuousSpaceBitmap* object_bitmap = current_space_bitmap_;
551 if (LIKELY(object_bitmap->HasAddress(obj))) {
552 return !object_bitmap->AtomicTestAndSet(obj);
553 }
554 MarkObjectSlowPath visitor(this);
555 return !mark_bitmap_->AtomicTestAndSet(obj, visitor);
556 }
557
MarkHeapReference(mirror::HeapReference<mirror::Object> * ref,bool do_atomic_update ATTRIBUTE_UNUSED)558 void MarkSweep::MarkHeapReference(mirror::HeapReference<mirror::Object>* ref,
559 bool do_atomic_update ATTRIBUTE_UNUSED) {
560 MarkObject(ref->AsMirrorPtr(), nullptr, MemberOffset(0));
561 }
562
563 // Used to mark objects when processing the mark stack. If an object is null, it is not marked.
MarkObject(mirror::Object * obj,mirror::Object * holder,MemberOffset offset)564 inline void MarkSweep::MarkObject(mirror::Object* obj,
565 mirror::Object* holder,
566 MemberOffset offset) {
567 if (obj != nullptr) {
568 MarkObjectNonNull(obj, holder, offset);
569 } else if (kCountMarkedObjects) {
570 ++mark_null_count_;
571 }
572 }
573
574 class MarkSweep::VerifyRootMarkedVisitor : public SingleRootVisitor {
575 public:
VerifyRootMarkedVisitor(MarkSweep * collector)576 explicit VerifyRootMarkedVisitor(MarkSweep* collector) : collector_(collector) { }
577
VisitRoot(mirror::Object * root,const RootInfo & info)578 void VisitRoot(mirror::Object* root, const RootInfo& info) OVERRIDE
579 REQUIRES_SHARED(Locks::mutator_lock_, Locks::heap_bitmap_lock_) {
580 CHECK(collector_->IsMarked(root) != nullptr) << info.ToString();
581 }
582
583 private:
584 MarkSweep* const collector_;
585 };
586
VisitRoots(mirror::Object *** roots,size_t count,const RootInfo & info ATTRIBUTE_UNUSED)587 void MarkSweep::VisitRoots(mirror::Object*** roots,
588 size_t count,
589 const RootInfo& info ATTRIBUTE_UNUSED) {
590 for (size_t i = 0; i < count; ++i) {
591 MarkObjectNonNull(*roots[i]);
592 }
593 }
594
VisitRoots(mirror::CompressedReference<mirror::Object> ** roots,size_t count,const RootInfo & info ATTRIBUTE_UNUSED)595 void MarkSweep::VisitRoots(mirror::CompressedReference<mirror::Object>** roots,
596 size_t count,
597 const RootInfo& info ATTRIBUTE_UNUSED) {
598 for (size_t i = 0; i < count; ++i) {
599 MarkObjectNonNull(roots[i]->AsMirrorPtr());
600 }
601 }
602
603 class MarkSweep::VerifyRootVisitor : public SingleRootVisitor {
604 public:
VerifyRootVisitor(std::ostream & os)605 explicit VerifyRootVisitor(std::ostream& os) : os_(os) {}
606
VisitRoot(mirror::Object * root,const RootInfo & info)607 void VisitRoot(mirror::Object* root, const RootInfo& info) OVERRIDE
608 REQUIRES_SHARED(Locks::mutator_lock_, Locks::heap_bitmap_lock_) {
609 // See if the root is on any space bitmap.
610 auto* heap = Runtime::Current()->GetHeap();
611 if (heap->GetLiveBitmap()->GetContinuousSpaceBitmap(root) == nullptr) {
612 space::LargeObjectSpace* large_object_space = heap->GetLargeObjectsSpace();
613 if (large_object_space != nullptr && !large_object_space->Contains(root)) {
614 os_ << "Found invalid root: " << root << " " << info << std::endl;
615 }
616 }
617 }
618
619 private:
620 std::ostream& os_;
621 };
622
VerifySuspendedThreadRoots(std::ostream & os)623 void MarkSweep::VerifySuspendedThreadRoots(std::ostream& os) {
624 VerifyRootVisitor visitor(os);
625 Runtime::Current()->GetThreadList()->VisitRootsForSuspendedThreads(&visitor);
626 }
627
MarkRoots(Thread * self)628 void MarkSweep::MarkRoots(Thread* self) {
629 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
630 if (Locks::mutator_lock_->IsExclusiveHeld(self)) {
631 // If we exclusively hold the mutator lock, all threads must be suspended.
632 Runtime::Current()->VisitRoots(this);
633 RevokeAllThreadLocalAllocationStacks(self);
634 } else {
635 MarkRootsCheckpoint(self, kRevokeRosAllocThreadLocalBuffersAtCheckpoint);
636 // At this point the live stack should no longer have any mutators which push into it.
637 MarkNonThreadRoots();
638 MarkConcurrentRoots(
639 static_cast<VisitRootFlags>(kVisitRootFlagAllRoots | kVisitRootFlagStartLoggingNewRoots));
640 }
641 }
642
MarkNonThreadRoots()643 void MarkSweep::MarkNonThreadRoots() {
644 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
645 Runtime::Current()->VisitNonThreadRoots(this);
646 }
647
MarkConcurrentRoots(VisitRootFlags flags)648 void MarkSweep::MarkConcurrentRoots(VisitRootFlags flags) {
649 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
650 // Visit all runtime roots and clear dirty flags.
651 Runtime::Current()->VisitConcurrentRoots(this, flags);
652 }
653
654 class MarkSweep::DelayReferenceReferentVisitor {
655 public:
DelayReferenceReferentVisitor(MarkSweep * collector)656 explicit DelayReferenceReferentVisitor(MarkSweep* collector) : collector_(collector) {}
657
operator ()(ObjPtr<mirror::Class> klass,ObjPtr<mirror::Reference> ref) const658 void operator()(ObjPtr<mirror::Class> klass, ObjPtr<mirror::Reference> ref) const
659 REQUIRES(Locks::heap_bitmap_lock_)
660 REQUIRES_SHARED(Locks::mutator_lock_) {
661 collector_->DelayReferenceReferent(klass, ref);
662 }
663
664 private:
665 MarkSweep* const collector_;
666 };
667
668 template <bool kUseFinger = false>
669 class MarkSweep::MarkStackTask : public Task {
670 public:
MarkStackTask(ThreadPool * thread_pool,MarkSweep * mark_sweep,size_t mark_stack_size,StackReference<mirror::Object> * mark_stack)671 MarkStackTask(ThreadPool* thread_pool,
672 MarkSweep* mark_sweep,
673 size_t mark_stack_size,
674 StackReference<mirror::Object>* mark_stack)
675 : mark_sweep_(mark_sweep),
676 thread_pool_(thread_pool),
677 mark_stack_pos_(mark_stack_size) {
678 // We may have to copy part of an existing mark stack when another mark stack overflows.
679 if (mark_stack_size != 0) {
680 DCHECK(mark_stack != nullptr);
681 // TODO: Check performance?
682 std::copy(mark_stack, mark_stack + mark_stack_size, mark_stack_);
683 }
684 if (kCountTasks) {
685 ++mark_sweep_->work_chunks_created_;
686 }
687 }
688
689 static const size_t kMaxSize = 1 * KB;
690
691 protected:
692 class MarkObjectParallelVisitor {
693 public:
MarkObjectParallelVisitor(MarkStackTask<kUseFinger> * chunk_task,MarkSweep * mark_sweep)694 ALWAYS_INLINE MarkObjectParallelVisitor(MarkStackTask<kUseFinger>* chunk_task,
695 MarkSweep* mark_sweep)
696 : chunk_task_(chunk_task), mark_sweep_(mark_sweep) {}
697
operator ()(mirror::Object * obj,MemberOffset offset,bool is_static ATTRIBUTE_UNUSED) const698 ALWAYS_INLINE void operator()(mirror::Object* obj,
699 MemberOffset offset,
700 bool is_static ATTRIBUTE_UNUSED) const
701 REQUIRES_SHARED(Locks::mutator_lock_) {
702 Mark(obj->GetFieldObject<mirror::Object>(offset));
703 }
704
VisitRootIfNonNull(mirror::CompressedReference<mirror::Object> * root) const705 void VisitRootIfNonNull(mirror::CompressedReference<mirror::Object>* root) const
706 REQUIRES_SHARED(Locks::mutator_lock_) {
707 if (!root->IsNull()) {
708 VisitRoot(root);
709 }
710 }
711
VisitRoot(mirror::CompressedReference<mirror::Object> * root) const712 void VisitRoot(mirror::CompressedReference<mirror::Object>* root) const
713 REQUIRES_SHARED(Locks::mutator_lock_) {
714 if (kCheckLocks) {
715 Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
716 Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
717 }
718 Mark(root->AsMirrorPtr());
719 }
720
721 private:
Mark(mirror::Object * ref) const722 ALWAYS_INLINE void Mark(mirror::Object* ref) const REQUIRES_SHARED(Locks::mutator_lock_) {
723 if (ref != nullptr && mark_sweep_->MarkObjectParallel(ref)) {
724 if (kUseFinger) {
725 std::atomic_thread_fence(std::memory_order_seq_cst);
726 if (reinterpret_cast<uintptr_t>(ref) >=
727 static_cast<uintptr_t>(mark_sweep_->atomic_finger_.LoadRelaxed())) {
728 return;
729 }
730 }
731 chunk_task_->MarkStackPush(ref);
732 }
733 }
734
735 MarkStackTask<kUseFinger>* const chunk_task_;
736 MarkSweep* const mark_sweep_;
737 };
738
739 class ScanObjectParallelVisitor {
740 public:
ScanObjectParallelVisitor(MarkStackTask<kUseFinger> * chunk_task)741 ALWAYS_INLINE explicit ScanObjectParallelVisitor(MarkStackTask<kUseFinger>* chunk_task)
742 : chunk_task_(chunk_task) {}
743
744 // No thread safety analysis since multiple threads will use this visitor.
operator ()(mirror::Object * obj) const745 void operator()(mirror::Object* obj) const
746 REQUIRES(Locks::heap_bitmap_lock_)
747 REQUIRES_SHARED(Locks::mutator_lock_) {
748 MarkSweep* const mark_sweep = chunk_task_->mark_sweep_;
749 MarkObjectParallelVisitor mark_visitor(chunk_task_, mark_sweep);
750 DelayReferenceReferentVisitor ref_visitor(mark_sweep);
751 mark_sweep->ScanObjectVisit(obj, mark_visitor, ref_visitor);
752 }
753
754 private:
755 MarkStackTask<kUseFinger>* const chunk_task_;
756 };
757
~MarkStackTask()758 virtual ~MarkStackTask() {
759 // Make sure that we have cleared our mark stack.
760 DCHECK_EQ(mark_stack_pos_, 0U);
761 if (kCountTasks) {
762 ++mark_sweep_->work_chunks_deleted_;
763 }
764 }
765
766 MarkSweep* const mark_sweep_;
767 ThreadPool* const thread_pool_;
768 // Thread local mark stack for this task.
769 StackReference<mirror::Object> mark_stack_[kMaxSize];
770 // Mark stack position.
771 size_t mark_stack_pos_;
772
MarkStackPush(mirror::Object * obj)773 ALWAYS_INLINE void MarkStackPush(mirror::Object* obj)
774 REQUIRES_SHARED(Locks::mutator_lock_) {
775 if (UNLIKELY(mark_stack_pos_ == kMaxSize)) {
776 // Mark stack overflow, give 1/2 the stack to the thread pool as a new work task.
777 mark_stack_pos_ /= 2;
778 auto* task = new MarkStackTask(thread_pool_,
779 mark_sweep_,
780 kMaxSize - mark_stack_pos_,
781 mark_stack_ + mark_stack_pos_);
782 thread_pool_->AddTask(Thread::Current(), task);
783 }
784 DCHECK(obj != nullptr);
785 DCHECK_LT(mark_stack_pos_, kMaxSize);
786 mark_stack_[mark_stack_pos_++].Assign(obj);
787 }
788
Finalize()789 virtual void Finalize() {
790 delete this;
791 }
792
793 // Scans all of the objects
Run(Thread * self ATTRIBUTE_UNUSED)794 virtual void Run(Thread* self ATTRIBUTE_UNUSED)
795 REQUIRES(Locks::heap_bitmap_lock_)
796 REQUIRES_SHARED(Locks::mutator_lock_) {
797 ScanObjectParallelVisitor visitor(this);
798 // TODO: Tune this.
799 static const size_t kFifoSize = 4;
800 BoundedFifoPowerOfTwo<mirror::Object*, kFifoSize> prefetch_fifo;
801 for (;;) {
802 mirror::Object* obj = nullptr;
803 if (kUseMarkStackPrefetch) {
804 while (mark_stack_pos_ != 0 && prefetch_fifo.size() < kFifoSize) {
805 mirror::Object* const mark_stack_obj = mark_stack_[--mark_stack_pos_].AsMirrorPtr();
806 DCHECK(mark_stack_obj != nullptr);
807 __builtin_prefetch(mark_stack_obj);
808 prefetch_fifo.push_back(mark_stack_obj);
809 }
810 if (UNLIKELY(prefetch_fifo.empty())) {
811 break;
812 }
813 obj = prefetch_fifo.front();
814 prefetch_fifo.pop_front();
815 } else {
816 if (UNLIKELY(mark_stack_pos_ == 0)) {
817 break;
818 }
819 obj = mark_stack_[--mark_stack_pos_].AsMirrorPtr();
820 }
821 DCHECK(obj != nullptr);
822 visitor(obj);
823 }
824 }
825 };
826
827 class MarkSweep::CardScanTask : public MarkStackTask<false> {
828 public:
CardScanTask(ThreadPool * thread_pool,MarkSweep * mark_sweep,accounting::ContinuousSpaceBitmap * bitmap,uint8_t * begin,uint8_t * end,uint8_t minimum_age,size_t mark_stack_size,StackReference<mirror::Object> * mark_stack_obj,bool clear_card)829 CardScanTask(ThreadPool* thread_pool,
830 MarkSweep* mark_sweep,
831 accounting::ContinuousSpaceBitmap* bitmap,
832 uint8_t* begin,
833 uint8_t* end,
834 uint8_t minimum_age,
835 size_t mark_stack_size,
836 StackReference<mirror::Object>* mark_stack_obj,
837 bool clear_card)
838 : MarkStackTask<false>(thread_pool, mark_sweep, mark_stack_size, mark_stack_obj),
839 bitmap_(bitmap),
840 begin_(begin),
841 end_(end),
842 minimum_age_(minimum_age),
843 clear_card_(clear_card) {}
844
845 protected:
846 accounting::ContinuousSpaceBitmap* const bitmap_;
847 uint8_t* const begin_;
848 uint8_t* const end_;
849 const uint8_t minimum_age_;
850 const bool clear_card_;
851
Finalize()852 virtual void Finalize() {
853 delete this;
854 }
855
Run(Thread * self)856 virtual void Run(Thread* self) NO_THREAD_SAFETY_ANALYSIS {
857 ScanObjectParallelVisitor visitor(this);
858 accounting::CardTable* card_table = mark_sweep_->GetHeap()->GetCardTable();
859 size_t cards_scanned = clear_card_
860 ? card_table->Scan<true>(bitmap_, begin_, end_, visitor, minimum_age_)
861 : card_table->Scan<false>(bitmap_, begin_, end_, visitor, minimum_age_);
862 VLOG(heap) << "Parallel scanning cards " << reinterpret_cast<void*>(begin_) << " - "
863 << reinterpret_cast<void*>(end_) << " = " << cards_scanned;
864 // Finish by emptying our local mark stack.
865 MarkStackTask::Run(self);
866 }
867 };
868
GetThreadCount(bool paused) const869 size_t MarkSweep::GetThreadCount(bool paused) const {
870 // Use less threads if we are in a background state (non jank perceptible) since we want to leave
871 // more CPU time for the foreground apps.
872 if (heap_->GetThreadPool() == nullptr || !Runtime::Current()->InJankPerceptibleProcessState()) {
873 return 1;
874 }
875 return (paused ? heap_->GetParallelGCThreadCount() : heap_->GetConcGCThreadCount()) + 1;
876 }
877
ScanGrayObjects(bool paused,uint8_t minimum_age)878 void MarkSweep::ScanGrayObjects(bool paused, uint8_t minimum_age) {
879 accounting::CardTable* card_table = GetHeap()->GetCardTable();
880 ThreadPool* thread_pool = GetHeap()->GetThreadPool();
881 size_t thread_count = GetThreadCount(paused);
882 // The parallel version with only one thread is faster for card scanning, TODO: fix.
883 if (kParallelCardScan && thread_count > 1) {
884 Thread* self = Thread::Current();
885 // Can't have a different split for each space since multiple spaces can have their cards being
886 // scanned at the same time.
887 TimingLogger::ScopedTiming t(paused ? "(Paused)ScanGrayObjects" : __FUNCTION__,
888 GetTimings());
889 // Try to take some of the mark stack since we can pass this off to the worker tasks.
890 StackReference<mirror::Object>* mark_stack_begin = mark_stack_->Begin();
891 StackReference<mirror::Object>* mark_stack_end = mark_stack_->End();
892 const size_t mark_stack_size = mark_stack_end - mark_stack_begin;
893 // Estimated number of work tasks we will create.
894 const size_t mark_stack_tasks = GetHeap()->GetContinuousSpaces().size() * thread_count;
895 DCHECK_NE(mark_stack_tasks, 0U);
896 const size_t mark_stack_delta = std::min(CardScanTask::kMaxSize / 2,
897 mark_stack_size / mark_stack_tasks + 1);
898 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
899 if (space->GetMarkBitmap() == nullptr) {
900 continue;
901 }
902 uint8_t* card_begin = space->Begin();
903 uint8_t* card_end = space->End();
904 // Align up the end address. For example, the image space's end
905 // may not be card-size-aligned.
906 card_end = AlignUp(card_end, accounting::CardTable::kCardSize);
907 DCHECK_ALIGNED(card_begin, accounting::CardTable::kCardSize);
908 DCHECK_ALIGNED(card_end, accounting::CardTable::kCardSize);
909 // Calculate how many bytes of heap we will scan,
910 const size_t address_range = card_end - card_begin;
911 // Calculate how much address range each task gets.
912 const size_t card_delta = RoundUp(address_range / thread_count + 1,
913 accounting::CardTable::kCardSize);
914 // If paused and the space is neither zygote nor image space, we could clear the dirty
915 // cards to avoid accumulating them to increase card scanning load in the following GC
916 // cycles. We need to keep dirty cards of image space and zygote space in order to track
917 // references to the other spaces.
918 bool clear_card = paused && !space->IsZygoteSpace() && !space->IsImageSpace();
919 // Create the worker tasks for this space.
920 while (card_begin != card_end) {
921 // Add a range of cards.
922 size_t addr_remaining = card_end - card_begin;
923 size_t card_increment = std::min(card_delta, addr_remaining);
924 // Take from the back of the mark stack.
925 size_t mark_stack_remaining = mark_stack_end - mark_stack_begin;
926 size_t mark_stack_increment = std::min(mark_stack_delta, mark_stack_remaining);
927 mark_stack_end -= mark_stack_increment;
928 mark_stack_->PopBackCount(static_cast<int32_t>(mark_stack_increment));
929 DCHECK_EQ(mark_stack_end, mark_stack_->End());
930 // Add the new task to the thread pool.
931 auto* task = new CardScanTask(thread_pool,
932 this,
933 space->GetMarkBitmap(),
934 card_begin,
935 card_begin + card_increment,
936 minimum_age,
937 mark_stack_increment,
938 mark_stack_end,
939 clear_card);
940 thread_pool->AddTask(self, task);
941 card_begin += card_increment;
942 }
943 }
944
945 // Note: the card scan below may dirty new cards (and scan them)
946 // as a side effect when a Reference object is encountered and
947 // queued during the marking. See b/11465268.
948 thread_pool->SetMaxActiveWorkers(thread_count - 1);
949 thread_pool->StartWorkers(self);
950 thread_pool->Wait(self, true, true);
951 thread_pool->StopWorkers(self);
952 } else {
953 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
954 if (space->GetMarkBitmap() != nullptr) {
955 // Image spaces are handled properly since live == marked for them.
956 const char* name = nullptr;
957 switch (space->GetGcRetentionPolicy()) {
958 case space::kGcRetentionPolicyNeverCollect:
959 name = paused ? "(Paused)ScanGrayImageSpaceObjects" : "ScanGrayImageSpaceObjects";
960 break;
961 case space::kGcRetentionPolicyFullCollect:
962 name = paused ? "(Paused)ScanGrayZygoteSpaceObjects" : "ScanGrayZygoteSpaceObjects";
963 break;
964 case space::kGcRetentionPolicyAlwaysCollect:
965 name = paused ? "(Paused)ScanGrayAllocSpaceObjects" : "ScanGrayAllocSpaceObjects";
966 break;
967 default:
968 LOG(FATAL) << "Unreachable";
969 UNREACHABLE();
970 }
971 TimingLogger::ScopedTiming t(name, GetTimings());
972 ScanObjectVisitor visitor(this);
973 bool clear_card = paused && !space->IsZygoteSpace() && !space->IsImageSpace();
974 if (clear_card) {
975 card_table->Scan<true>(space->GetMarkBitmap(),
976 space->Begin(),
977 space->End(),
978 visitor,
979 minimum_age);
980 } else {
981 card_table->Scan<false>(space->GetMarkBitmap(),
982 space->Begin(),
983 space->End(),
984 visitor,
985 minimum_age);
986 }
987 }
988 }
989 }
990 }
991
992 class MarkSweep::RecursiveMarkTask : public MarkStackTask<false> {
993 public:
RecursiveMarkTask(ThreadPool * thread_pool,MarkSweep * mark_sweep,accounting::ContinuousSpaceBitmap * bitmap,uintptr_t begin,uintptr_t end)994 RecursiveMarkTask(ThreadPool* thread_pool,
995 MarkSweep* mark_sweep,
996 accounting::ContinuousSpaceBitmap* bitmap,
997 uintptr_t begin,
998 uintptr_t end)
999 : MarkStackTask<false>(thread_pool, mark_sweep, 0, nullptr),
1000 bitmap_(bitmap),
1001 begin_(begin),
1002 end_(end) {}
1003
1004 protected:
1005 accounting::ContinuousSpaceBitmap* const bitmap_;
1006 const uintptr_t begin_;
1007 const uintptr_t end_;
1008
Finalize()1009 virtual void Finalize() {
1010 delete this;
1011 }
1012
1013 // Scans all of the objects
Run(Thread * self)1014 virtual void Run(Thread* self) NO_THREAD_SAFETY_ANALYSIS {
1015 ScanObjectParallelVisitor visitor(this);
1016 bitmap_->VisitMarkedRange(begin_, end_, visitor);
1017 // Finish by emptying our local mark stack.
1018 MarkStackTask::Run(self);
1019 }
1020 };
1021
1022 // Populates the mark stack based on the set of marked objects and
1023 // recursively marks until the mark stack is emptied.
RecursiveMark()1024 void MarkSweep::RecursiveMark() {
1025 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1026 // RecursiveMark will build the lists of known instances of the Reference classes. See
1027 // DelayReferenceReferent for details.
1028 if (kUseRecursiveMark) {
1029 const bool partial = GetGcType() == kGcTypePartial;
1030 ScanObjectVisitor scan_visitor(this);
1031 auto* self = Thread::Current();
1032 ThreadPool* thread_pool = heap_->GetThreadPool();
1033 size_t thread_count = GetThreadCount(false);
1034 const bool parallel = kParallelRecursiveMark && thread_count > 1;
1035 mark_stack_->Reset();
1036 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
1037 if ((space->GetGcRetentionPolicy() == space::kGcRetentionPolicyAlwaysCollect) ||
1038 (!partial && space->GetGcRetentionPolicy() == space::kGcRetentionPolicyFullCollect)) {
1039 current_space_bitmap_ = space->GetMarkBitmap();
1040 if (current_space_bitmap_ == nullptr) {
1041 continue;
1042 }
1043 if (parallel) {
1044 // We will use the mark stack the future.
1045 // CHECK(mark_stack_->IsEmpty());
1046 // This function does not handle heap end increasing, so we must use the space end.
1047 uintptr_t begin = reinterpret_cast<uintptr_t>(space->Begin());
1048 uintptr_t end = reinterpret_cast<uintptr_t>(space->End());
1049 atomic_finger_.StoreRelaxed(AtomicInteger::MaxValue());
1050
1051 // Create a few worker tasks.
1052 const size_t n = thread_count * 2;
1053 while (begin != end) {
1054 uintptr_t start = begin;
1055 uintptr_t delta = (end - begin) / n;
1056 delta = RoundUp(delta, KB);
1057 if (delta < 16 * KB) delta = end - begin;
1058 begin += delta;
1059 auto* task = new RecursiveMarkTask(thread_pool,
1060 this,
1061 current_space_bitmap_,
1062 start,
1063 begin);
1064 thread_pool->AddTask(self, task);
1065 }
1066 thread_pool->SetMaxActiveWorkers(thread_count - 1);
1067 thread_pool->StartWorkers(self);
1068 thread_pool->Wait(self, true, true);
1069 thread_pool->StopWorkers(self);
1070 } else {
1071 // This function does not handle heap end increasing, so we must use the space end.
1072 uintptr_t begin = reinterpret_cast<uintptr_t>(space->Begin());
1073 uintptr_t end = reinterpret_cast<uintptr_t>(space->End());
1074 current_space_bitmap_->VisitMarkedRange(begin, end, scan_visitor);
1075 }
1076 }
1077 }
1078 }
1079 ProcessMarkStack(false);
1080 }
1081
RecursiveMarkDirtyObjects(bool paused,uint8_t minimum_age)1082 void MarkSweep::RecursiveMarkDirtyObjects(bool paused, uint8_t minimum_age) {
1083 ScanGrayObjects(paused, minimum_age);
1084 ProcessMarkStack(paused);
1085 }
1086
ReMarkRoots()1087 void MarkSweep::ReMarkRoots() {
1088 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1089 Locks::mutator_lock_->AssertExclusiveHeld(Thread::Current());
1090 Runtime::Current()->VisitRoots(this, static_cast<VisitRootFlags>(
1091 kVisitRootFlagNewRoots | kVisitRootFlagStopLoggingNewRoots | kVisitRootFlagClearRootLog));
1092 if (kVerifyRootsMarked) {
1093 TimingLogger::ScopedTiming t2("(Paused)VerifyRoots", GetTimings());
1094 VerifyRootMarkedVisitor visitor(this);
1095 Runtime::Current()->VisitRoots(&visitor);
1096 }
1097 }
1098
SweepSystemWeaks(Thread * self)1099 void MarkSweep::SweepSystemWeaks(Thread* self) {
1100 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1101 ReaderMutexLock mu(self, *Locks::heap_bitmap_lock_);
1102 Runtime::Current()->SweepSystemWeaks(this);
1103 }
1104
1105 class MarkSweep::VerifySystemWeakVisitor : public IsMarkedVisitor {
1106 public:
VerifySystemWeakVisitor(MarkSweep * mark_sweep)1107 explicit VerifySystemWeakVisitor(MarkSweep* mark_sweep) : mark_sweep_(mark_sweep) {}
1108
IsMarked(mirror::Object * obj)1109 virtual mirror::Object* IsMarked(mirror::Object* obj)
1110 OVERRIDE
1111 REQUIRES_SHARED(Locks::mutator_lock_, Locks::heap_bitmap_lock_) {
1112 mark_sweep_->VerifyIsLive(obj);
1113 return obj;
1114 }
1115
1116 MarkSweep* const mark_sweep_;
1117 };
1118
VerifyIsLive(const mirror::Object * obj)1119 void MarkSweep::VerifyIsLive(const mirror::Object* obj) {
1120 if (!heap_->GetLiveBitmap()->Test(obj)) {
1121 // TODO: Consider live stack? Has this code bitrotted?
1122 CHECK(!heap_->allocation_stack_->Contains(obj))
1123 << "Found dead object " << obj << "\n" << heap_->DumpSpaces();
1124 }
1125 }
1126
VerifySystemWeaks()1127 void MarkSweep::VerifySystemWeaks() {
1128 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1129 // Verify system weaks, uses a special object visitor which returns the input object.
1130 VerifySystemWeakVisitor visitor(this);
1131 Runtime::Current()->SweepSystemWeaks(&visitor);
1132 }
1133
1134 class MarkSweep::CheckpointMarkThreadRoots : public Closure, public RootVisitor {
1135 public:
CheckpointMarkThreadRoots(MarkSweep * mark_sweep,bool revoke_ros_alloc_thread_local_buffers_at_checkpoint)1136 CheckpointMarkThreadRoots(MarkSweep* mark_sweep,
1137 bool revoke_ros_alloc_thread_local_buffers_at_checkpoint)
1138 : mark_sweep_(mark_sweep),
1139 revoke_ros_alloc_thread_local_buffers_at_checkpoint_(
1140 revoke_ros_alloc_thread_local_buffers_at_checkpoint) {
1141 }
1142
VisitRoots(mirror::Object *** roots,size_t count,const RootInfo & info ATTRIBUTE_UNUSED)1143 void VisitRoots(mirror::Object*** roots, size_t count, const RootInfo& info ATTRIBUTE_UNUSED)
1144 OVERRIDE REQUIRES_SHARED(Locks::mutator_lock_)
1145 REQUIRES(Locks::heap_bitmap_lock_) {
1146 for (size_t i = 0; i < count; ++i) {
1147 mark_sweep_->MarkObjectNonNullParallel(*roots[i]);
1148 }
1149 }
1150
VisitRoots(mirror::CompressedReference<mirror::Object> ** roots,size_t count,const RootInfo & info ATTRIBUTE_UNUSED)1151 void VisitRoots(mirror::CompressedReference<mirror::Object>** roots,
1152 size_t count,
1153 const RootInfo& info ATTRIBUTE_UNUSED)
1154 OVERRIDE REQUIRES_SHARED(Locks::mutator_lock_)
1155 REQUIRES(Locks::heap_bitmap_lock_) {
1156 for (size_t i = 0; i < count; ++i) {
1157 mark_sweep_->MarkObjectNonNullParallel(roots[i]->AsMirrorPtr());
1158 }
1159 }
1160
Run(Thread * thread)1161 virtual void Run(Thread* thread) OVERRIDE NO_THREAD_SAFETY_ANALYSIS {
1162 ScopedTrace trace("Marking thread roots");
1163 // Note: self is not necessarily equal to thread since thread may be suspended.
1164 Thread* const self = Thread::Current();
1165 CHECK(thread == self || thread->IsSuspended() || thread->GetState() == kWaitingPerformingGc)
1166 << thread->GetState() << " thread " << thread << " self " << self;
1167 thread->VisitRoots(this, kVisitRootFlagAllRoots);
1168 if (revoke_ros_alloc_thread_local_buffers_at_checkpoint_) {
1169 ScopedTrace trace2("RevokeRosAllocThreadLocalBuffers");
1170 mark_sweep_->GetHeap()->RevokeRosAllocThreadLocalBuffers(thread);
1171 }
1172 // If thread is a running mutator, then act on behalf of the garbage collector.
1173 // See the code in ThreadList::RunCheckpoint.
1174 mark_sweep_->GetBarrier().Pass(self);
1175 }
1176
1177 private:
1178 MarkSweep* const mark_sweep_;
1179 const bool revoke_ros_alloc_thread_local_buffers_at_checkpoint_;
1180 };
1181
MarkRootsCheckpoint(Thread * self,bool revoke_ros_alloc_thread_local_buffers_at_checkpoint)1182 void MarkSweep::MarkRootsCheckpoint(Thread* self,
1183 bool revoke_ros_alloc_thread_local_buffers_at_checkpoint) {
1184 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1185 CheckpointMarkThreadRoots check_point(this, revoke_ros_alloc_thread_local_buffers_at_checkpoint);
1186 ThreadList* thread_list = Runtime::Current()->GetThreadList();
1187 // Request the check point is run on all threads returning a count of the threads that must
1188 // run through the barrier including self.
1189 size_t barrier_count = thread_list->RunCheckpoint(&check_point);
1190 // Release locks then wait for all mutator threads to pass the barrier.
1191 // If there are no threads to wait which implys that all the checkpoint functions are finished,
1192 // then no need to release locks.
1193 if (barrier_count == 0) {
1194 return;
1195 }
1196 Locks::heap_bitmap_lock_->ExclusiveUnlock(self);
1197 Locks::mutator_lock_->SharedUnlock(self);
1198 {
1199 ScopedThreadStateChange tsc(self, kWaitingForCheckPointsToRun);
1200 gc_barrier_->Increment(self, barrier_count);
1201 }
1202 Locks::mutator_lock_->SharedLock(self);
1203 Locks::heap_bitmap_lock_->ExclusiveLock(self);
1204 }
1205
SweepArray(accounting::ObjectStack * allocations,bool swap_bitmaps)1206 void MarkSweep::SweepArray(accounting::ObjectStack* allocations, bool swap_bitmaps) {
1207 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1208 Thread* self = Thread::Current();
1209 mirror::Object** chunk_free_buffer = reinterpret_cast<mirror::Object**>(
1210 sweep_array_free_buffer_mem_map_->BaseBegin());
1211 size_t chunk_free_pos = 0;
1212 ObjectBytePair freed;
1213 ObjectBytePair freed_los;
1214 // How many objects are left in the array, modified after each space is swept.
1215 StackReference<mirror::Object>* objects = allocations->Begin();
1216 size_t count = allocations->Size();
1217 // Change the order to ensure that the non-moving space last swept as an optimization.
1218 std::vector<space::ContinuousSpace*> sweep_spaces;
1219 space::ContinuousSpace* non_moving_space = nullptr;
1220 for (space::ContinuousSpace* space : heap_->GetContinuousSpaces()) {
1221 if (space->IsAllocSpace() &&
1222 !immune_spaces_.ContainsSpace(space) &&
1223 space->GetLiveBitmap() != nullptr) {
1224 if (space == heap_->GetNonMovingSpace()) {
1225 non_moving_space = space;
1226 } else {
1227 sweep_spaces.push_back(space);
1228 }
1229 }
1230 }
1231 // Unlikely to sweep a significant amount of non_movable objects, so we do these after
1232 // the other alloc spaces as an optimization.
1233 if (non_moving_space != nullptr) {
1234 sweep_spaces.push_back(non_moving_space);
1235 }
1236 // Start by sweeping the continuous spaces.
1237 for (space::ContinuousSpace* space : sweep_spaces) {
1238 space::AllocSpace* alloc_space = space->AsAllocSpace();
1239 accounting::ContinuousSpaceBitmap* live_bitmap = space->GetLiveBitmap();
1240 accounting::ContinuousSpaceBitmap* mark_bitmap = space->GetMarkBitmap();
1241 if (swap_bitmaps) {
1242 std::swap(live_bitmap, mark_bitmap);
1243 }
1244 StackReference<mirror::Object>* out = objects;
1245 for (size_t i = 0; i < count; ++i) {
1246 mirror::Object* const obj = objects[i].AsMirrorPtr();
1247 if (kUseThreadLocalAllocationStack && obj == nullptr) {
1248 continue;
1249 }
1250 if (space->HasAddress(obj)) {
1251 // This object is in the space, remove it from the array and add it to the sweep buffer
1252 // if needed.
1253 if (!mark_bitmap->Test(obj)) {
1254 if (chunk_free_pos >= kSweepArrayChunkFreeSize) {
1255 TimingLogger::ScopedTiming t2("FreeList", GetTimings());
1256 freed.objects += chunk_free_pos;
1257 freed.bytes += alloc_space->FreeList(self, chunk_free_pos, chunk_free_buffer);
1258 chunk_free_pos = 0;
1259 }
1260 chunk_free_buffer[chunk_free_pos++] = obj;
1261 }
1262 } else {
1263 (out++)->Assign(obj);
1264 }
1265 }
1266 if (chunk_free_pos > 0) {
1267 TimingLogger::ScopedTiming t2("FreeList", GetTimings());
1268 freed.objects += chunk_free_pos;
1269 freed.bytes += alloc_space->FreeList(self, chunk_free_pos, chunk_free_buffer);
1270 chunk_free_pos = 0;
1271 }
1272 // All of the references which space contained are no longer in the allocation stack, update
1273 // the count.
1274 count = out - objects;
1275 }
1276 // Handle the large object space.
1277 space::LargeObjectSpace* large_object_space = GetHeap()->GetLargeObjectsSpace();
1278 if (large_object_space != nullptr) {
1279 accounting::LargeObjectBitmap* large_live_objects = large_object_space->GetLiveBitmap();
1280 accounting::LargeObjectBitmap* large_mark_objects = large_object_space->GetMarkBitmap();
1281 if (swap_bitmaps) {
1282 std::swap(large_live_objects, large_mark_objects);
1283 }
1284 for (size_t i = 0; i < count; ++i) {
1285 mirror::Object* const obj = objects[i].AsMirrorPtr();
1286 // Handle large objects.
1287 if (kUseThreadLocalAllocationStack && obj == nullptr) {
1288 continue;
1289 }
1290 if (!large_mark_objects->Test(obj)) {
1291 ++freed_los.objects;
1292 freed_los.bytes += large_object_space->Free(self, obj);
1293 }
1294 }
1295 }
1296 {
1297 TimingLogger::ScopedTiming t2("RecordFree", GetTimings());
1298 RecordFree(freed);
1299 RecordFreeLOS(freed_los);
1300 t2.NewTiming("ResetStack");
1301 allocations->Reset();
1302 }
1303 sweep_array_free_buffer_mem_map_->MadviseDontNeedAndZero();
1304 }
1305
Sweep(bool swap_bitmaps)1306 void MarkSweep::Sweep(bool swap_bitmaps) {
1307 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1308 // Ensure that nobody inserted items in the live stack after we swapped the stacks.
1309 CHECK_GE(live_stack_freeze_size_, GetHeap()->GetLiveStack()->Size());
1310 {
1311 TimingLogger::ScopedTiming t2("MarkAllocStackAsLive", GetTimings());
1312 // Mark everything allocated since the last GC as live so that we can sweep concurrently,
1313 // knowing that new allocations won't be marked as live.
1314 accounting::ObjectStack* live_stack = heap_->GetLiveStack();
1315 heap_->MarkAllocStackAsLive(live_stack);
1316 live_stack->Reset();
1317 DCHECK(mark_stack_->IsEmpty());
1318 }
1319 for (const auto& space : GetHeap()->GetContinuousSpaces()) {
1320 if (space->IsContinuousMemMapAllocSpace()) {
1321 space::ContinuousMemMapAllocSpace* alloc_space = space->AsContinuousMemMapAllocSpace();
1322 TimingLogger::ScopedTiming split(
1323 alloc_space->IsZygoteSpace() ? "SweepZygoteSpace" : "SweepMallocSpace",
1324 GetTimings());
1325 RecordFree(alloc_space->Sweep(swap_bitmaps));
1326 }
1327 }
1328 SweepLargeObjects(swap_bitmaps);
1329 }
1330
SweepLargeObjects(bool swap_bitmaps)1331 void MarkSweep::SweepLargeObjects(bool swap_bitmaps) {
1332 space::LargeObjectSpace* los = heap_->GetLargeObjectsSpace();
1333 if (los != nullptr) {
1334 TimingLogger::ScopedTiming split(__FUNCTION__, GetTimings());
1335 RecordFreeLOS(los->Sweep(swap_bitmaps));
1336 }
1337 }
1338
1339 // Process the "referent" field lin a java.lang.ref.Reference. If the referent has not yet been
1340 // marked, put it on the appropriate list in the heap for later processing.
DelayReferenceReferent(ObjPtr<mirror::Class> klass,ObjPtr<mirror::Reference> ref)1341 void MarkSweep::DelayReferenceReferent(ObjPtr<mirror::Class> klass, ObjPtr<mirror::Reference> ref) {
1342 heap_->GetReferenceProcessor()->DelayReferenceReferent(klass, ref, this);
1343 }
1344
1345 class MarkVisitor {
1346 public:
MarkVisitor(MarkSweep * const mark_sweep)1347 ALWAYS_INLINE explicit MarkVisitor(MarkSweep* const mark_sweep) : mark_sweep_(mark_sweep) {}
1348
operator ()(mirror::Object * obj,MemberOffset offset,bool is_static ATTRIBUTE_UNUSED) const1349 ALWAYS_INLINE void operator()(mirror::Object* obj,
1350 MemberOffset offset,
1351 bool is_static ATTRIBUTE_UNUSED) const
1352 REQUIRES(Locks::heap_bitmap_lock_)
1353 REQUIRES_SHARED(Locks::mutator_lock_) {
1354 if (kCheckLocks) {
1355 Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
1356 Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
1357 }
1358 mark_sweep_->MarkObject(obj->GetFieldObject<mirror::Object>(offset), obj, offset);
1359 }
1360
VisitRootIfNonNull(mirror::CompressedReference<mirror::Object> * root) const1361 void VisitRootIfNonNull(mirror::CompressedReference<mirror::Object>* root) const
1362 REQUIRES(Locks::heap_bitmap_lock_)
1363 REQUIRES_SHARED(Locks::mutator_lock_) {
1364 if (!root->IsNull()) {
1365 VisitRoot(root);
1366 }
1367 }
1368
VisitRoot(mirror::CompressedReference<mirror::Object> * root) const1369 void VisitRoot(mirror::CompressedReference<mirror::Object>* root) const
1370 REQUIRES(Locks::heap_bitmap_lock_)
1371 REQUIRES_SHARED(Locks::mutator_lock_) {
1372 if (kCheckLocks) {
1373 Locks::mutator_lock_->AssertSharedHeld(Thread::Current());
1374 Locks::heap_bitmap_lock_->AssertExclusiveHeld(Thread::Current());
1375 }
1376 mark_sweep_->MarkObject(root->AsMirrorPtr());
1377 }
1378
1379 private:
1380 MarkSweep* const mark_sweep_;
1381 };
1382
1383 // Scans an object reference. Determines the type of the reference
1384 // and dispatches to a specialized scanning routine.
ScanObject(mirror::Object * obj)1385 void MarkSweep::ScanObject(mirror::Object* obj) {
1386 MarkVisitor mark_visitor(this);
1387 DelayReferenceReferentVisitor ref_visitor(this);
1388 ScanObjectVisit(obj, mark_visitor, ref_visitor);
1389 }
1390
ProcessMarkStackParallel(size_t thread_count)1391 void MarkSweep::ProcessMarkStackParallel(size_t thread_count) {
1392 Thread* self = Thread::Current();
1393 ThreadPool* thread_pool = GetHeap()->GetThreadPool();
1394 const size_t chunk_size = std::min(mark_stack_->Size() / thread_count + 1,
1395 static_cast<size_t>(MarkStackTask<false>::kMaxSize));
1396 CHECK_GT(chunk_size, 0U);
1397 // Split the current mark stack up into work tasks.
1398 for (auto* it = mark_stack_->Begin(), *end = mark_stack_->End(); it < end; ) {
1399 const size_t delta = std::min(static_cast<size_t>(end - it), chunk_size);
1400 thread_pool->AddTask(self, new MarkStackTask<false>(thread_pool, this, delta, it));
1401 it += delta;
1402 }
1403 thread_pool->SetMaxActiveWorkers(thread_count - 1);
1404 thread_pool->StartWorkers(self);
1405 thread_pool->Wait(self, true, true);
1406 thread_pool->StopWorkers(self);
1407 mark_stack_->Reset();
1408 CHECK_EQ(work_chunks_created_.LoadSequentiallyConsistent(),
1409 work_chunks_deleted_.LoadSequentiallyConsistent())
1410 << " some of the work chunks were leaked";
1411 }
1412
1413 // Scan anything that's on the mark stack.
ProcessMarkStack(bool paused)1414 void MarkSweep::ProcessMarkStack(bool paused) {
1415 TimingLogger::ScopedTiming t(paused ? "(Paused)ProcessMarkStack" : __FUNCTION__, GetTimings());
1416 size_t thread_count = GetThreadCount(paused);
1417 if (kParallelProcessMarkStack && thread_count > 1 &&
1418 mark_stack_->Size() >= kMinimumParallelMarkStackSize) {
1419 ProcessMarkStackParallel(thread_count);
1420 } else {
1421 // TODO: Tune this.
1422 static const size_t kFifoSize = 4;
1423 BoundedFifoPowerOfTwo<mirror::Object*, kFifoSize> prefetch_fifo;
1424 for (;;) {
1425 mirror::Object* obj = nullptr;
1426 if (kUseMarkStackPrefetch) {
1427 while (!mark_stack_->IsEmpty() && prefetch_fifo.size() < kFifoSize) {
1428 mirror::Object* mark_stack_obj = mark_stack_->PopBack();
1429 DCHECK(mark_stack_obj != nullptr);
1430 __builtin_prefetch(mark_stack_obj);
1431 prefetch_fifo.push_back(mark_stack_obj);
1432 }
1433 if (prefetch_fifo.empty()) {
1434 break;
1435 }
1436 obj = prefetch_fifo.front();
1437 prefetch_fifo.pop_front();
1438 } else {
1439 if (mark_stack_->IsEmpty()) {
1440 break;
1441 }
1442 obj = mark_stack_->PopBack();
1443 }
1444 DCHECK(obj != nullptr);
1445 ScanObject(obj);
1446 }
1447 }
1448 }
1449
IsMarked(mirror::Object * object)1450 inline mirror::Object* MarkSweep::IsMarked(mirror::Object* object) {
1451 if (immune_spaces_.IsInImmuneRegion(object)) {
1452 return object;
1453 }
1454 if (current_space_bitmap_->HasAddress(object)) {
1455 return current_space_bitmap_->Test(object) ? object : nullptr;
1456 }
1457 return mark_bitmap_->Test(object) ? object : nullptr;
1458 }
1459
FinishPhase()1460 void MarkSweep::FinishPhase() {
1461 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1462 if (kCountScannedTypes) {
1463 VLOG(gc)
1464 << "MarkSweep scanned"
1465 << " no reference objects=" << no_reference_class_count_.LoadRelaxed()
1466 << " normal objects=" << normal_count_.LoadRelaxed()
1467 << " classes=" << class_count_.LoadRelaxed()
1468 << " object arrays=" << object_array_count_.LoadRelaxed()
1469 << " references=" << reference_count_.LoadRelaxed()
1470 << " other=" << other_count_.LoadRelaxed();
1471 }
1472 if (kCountTasks) {
1473 VLOG(gc) << "Total number of work chunks allocated: " << work_chunks_created_.LoadRelaxed();
1474 }
1475 if (kMeasureOverhead) {
1476 VLOG(gc) << "Overhead time " << PrettyDuration(overhead_time_.LoadRelaxed());
1477 }
1478 if (kProfileLargeObjects) {
1479 VLOG(gc) << "Large objects tested " << large_object_test_.LoadRelaxed()
1480 << " marked " << large_object_mark_.LoadRelaxed();
1481 }
1482 if (kCountMarkedObjects) {
1483 VLOG(gc) << "Marked: null=" << mark_null_count_.LoadRelaxed()
1484 << " immune=" << mark_immune_count_.LoadRelaxed()
1485 << " fastpath=" << mark_fastpath_count_.LoadRelaxed()
1486 << " slowpath=" << mark_slowpath_count_.LoadRelaxed();
1487 }
1488 CHECK(mark_stack_->IsEmpty()); // Ensure that the mark stack is empty.
1489 mark_stack_->Reset();
1490 Thread* const self = Thread::Current();
1491 ReaderMutexLock mu(self, *Locks::mutator_lock_);
1492 WriterMutexLock mu2(self, *Locks::heap_bitmap_lock_);
1493 heap_->ClearMarkedObjects();
1494 }
1495
RevokeAllThreadLocalBuffers()1496 void MarkSweep::RevokeAllThreadLocalBuffers() {
1497 if (kRevokeRosAllocThreadLocalBuffersAtCheckpoint && IsConcurrent()) {
1498 // If concurrent, rosalloc thread-local buffers are revoked at the
1499 // thread checkpoint. Bump pointer space thread-local buffers must
1500 // not be in use.
1501 GetHeap()->AssertAllBumpPointerSpaceThreadLocalBuffersAreRevoked();
1502 } else {
1503 TimingLogger::ScopedTiming t(__FUNCTION__, GetTimings());
1504 GetHeap()->RevokeAllThreadLocalBuffers();
1505 }
1506 }
1507
1508 } // namespace collector
1509 } // namespace gc
1510 } // namespace art
1511