1 /*
2 * Copyright (c) 2001, 2016, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26 #include "classfile/metadataOnStackMark.hpp"
27 #include "classfile/symbolTable.hpp"
28 #include "code/codeCache.hpp"
29 #include "gc/g1/concurrentMarkThread.inline.hpp"
30 #include "gc/g1/g1CollectedHeap.inline.hpp"
31 #include "gc/g1/g1CollectorPolicy.hpp"
32 #include "gc/g1/g1CollectorState.hpp"
33 #include "gc/g1/g1ConcurrentMark.inline.hpp"
34 #include "gc/g1/g1HeapVerifier.hpp"
35 #include "gc/g1/g1OopClosures.inline.hpp"
36 #include "gc/g1/g1StringDedup.hpp"
37 #include "gc/g1/heapRegion.inline.hpp"
38 #include "gc/g1/heapRegionRemSet.hpp"
39 #include "gc/g1/heapRegionSet.inline.hpp"
40 #include "gc/g1/suspendibleThreadSet.hpp"
41 #include "gc/shared/gcId.hpp"
42 #include "gc/shared/gcTimer.hpp"
43 #include "gc/shared/gcTrace.hpp"
44 #include "gc/shared/gcTraceTime.inline.hpp"
45 #include "gc/shared/genOopClosures.inline.hpp"
46 #include "gc/shared/referencePolicy.hpp"
47 #include "gc/shared/strongRootsScope.hpp"
48 #include "gc/shared/taskqueue.inline.hpp"
49 #include "gc/shared/vmGCOperations.hpp"
50 #include "logging/log.hpp"
51 #include "memory/allocation.hpp"
52 #include "memory/resourceArea.hpp"
53 #include "oops/oop.inline.hpp"
54 #include "runtime/atomic.inline.hpp"
55 #include "runtime/handles.inline.hpp"
56 #include "runtime/java.hpp"
57 #include "runtime/prefetch.inline.hpp"
58 #include "services/memTracker.hpp"
59
60 // Concurrent marking bit map wrapper
61
62 G1CMBitMapRO::G1CMBitMapRO(int shifter) :
63 _bm(),
64 _shifter(shifter) {
65 _bmStartWord = 0;
66 _bmWordSize = 0;
67 }
68
69 HeapWord* G1CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr,
70 const HeapWord* limit) const {
71 // First we must round addr *up* to a possible object boundary.
72 addr = (HeapWord*)align_size_up((intptr_t)addr,
73 HeapWordSize << _shifter);
74 size_t addrOffset = heapWordToOffset(addr);
75 assert(limit != NULL, "limit must not be NULL");
76 size_t limitOffset = heapWordToOffset(limit);
77 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
78 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
79 assert(nextAddr >= addr, "get_next_one postcondition");
80 assert(nextAddr == limit || isMarked(nextAddr),
81 "get_next_one postcondition");
82 return nextAddr;
83 }
84
85 #ifndef PRODUCT
86 bool G1CMBitMapRO::covers(MemRegion heap_rs) const {
87 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
88 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
89 "size inconsistency");
90 return _bmStartWord == (HeapWord*)(heap_rs.start()) &&
91 _bmWordSize == heap_rs.word_size();
92 }
93 #endif
94
95 void G1CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const {
96 _bm.print_on_error(st, prefix);
97 }
98
99 size_t G1CMBitMap::compute_size(size_t heap_size) {
100 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
101 }
102
103 size_t G1CMBitMap::mark_distance() {
104 return MinObjAlignmentInBytes * BitsPerByte;
105 }
106
107 void G1CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) {
108 _bmStartWord = heap.start();
109 _bmWordSize = heap.word_size();
110
111 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start());
112 _bm.set_size(_bmWordSize >> _shifter);
113
114 storage->set_mapping_changed_listener(&_listener);
115 }
116
117 void G1CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) {
118 if (zero_filled) {
119 return;
120 }
121 // We need to clear the bitmap on commit, removing any existing information.
122 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords);
123 _bm->clearRange(mr);
124 }
125
126 // Closure used for clearing the given mark bitmap.
127 class ClearBitmapHRClosure : public HeapRegionClosure {
128 private:
129 G1ConcurrentMark* _cm;
130 G1CMBitMap* _bitmap;
131 bool _may_yield; // The closure may yield during iteration. If yielded, abort the iteration.
132 public:
133 ClearBitmapHRClosure(G1ConcurrentMark* cm, G1CMBitMap* bitmap, bool may_yield) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap), _may_yield(may_yield) {
134 assert(!may_yield || cm != NULL, "CM must be non-NULL if this closure is expected to yield.");
135 }
136
137 virtual bool doHeapRegion(HeapRegion* r) {
138 size_t const chunk_size_in_words = M / HeapWordSize;
139
140 HeapWord* cur = r->bottom();
141 HeapWord* const end = r->end();
142
143 while (cur < end) {
144 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
145 _bitmap->clearRange(mr);
146
147 cur += chunk_size_in_words;
148
149 // Abort iteration if after yielding the marking has been aborted.
150 if (_may_yield && _cm->do_yield_check() && _cm->has_aborted()) {
151 return true;
152 }
153 // Repeat the asserts from before the start of the closure. We will do them
154 // as asserts here to minimize their overhead on the product. However, we
155 // will have them as guarantees at the beginning / end of the bitmap
156 // clearing to get some checking in the product.
157 assert(!_may_yield || _cm->cmThread()->during_cycle(), "invariant");
158 assert(!_may_yield || !G1CollectedHeap::heap()->collector_state()->mark_in_progress(), "invariant");
159 }
160
161 return false;
162 }
163 };
164
165 class ParClearNextMarkBitmapTask : public AbstractGangTask {
166 ClearBitmapHRClosure* _cl;
167 HeapRegionClaimer _hrclaimer;
168 bool _suspendible; // If the task is suspendible, workers must join the STS.
169
170 public:
171 ParClearNextMarkBitmapTask(ClearBitmapHRClosure *cl, uint n_workers, bool suspendible) :
172 _cl(cl), _suspendible(suspendible), AbstractGangTask("Parallel Clear Bitmap Task"), _hrclaimer(n_workers) {}
173
174 void work(uint worker_id) {
175 SuspendibleThreadSetJoiner sts_join(_suspendible);
176 G1CollectedHeap::heap()->heap_region_par_iterate(_cl, worker_id, &_hrclaimer, true);
177 }
178 };
179
180 void G1CMBitMap::clearAll() {
181 G1CollectedHeap* g1h = G1CollectedHeap::heap();
182 ClearBitmapHRClosure cl(NULL, this, false /* may_yield */);
183 uint n_workers = g1h->workers()->active_workers();
184 ParClearNextMarkBitmapTask task(&cl, n_workers, false);
185 g1h->workers()->run_task(&task);
186 guarantee(cl.complete(), "Must have completed iteration.");
187 return;
188 }
189
190 void G1CMBitMap::clearRange(MemRegion mr) {
191 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
192 assert(!mr.is_empty(), "unexpected empty region");
193 // convert address range into offset range
194 _bm.at_put_range(heapWordToOffset(mr.start()),
195 heapWordToOffset(mr.end()), false);
196 }
197
198 G1CMMarkStack::G1CMMarkStack(G1ConcurrentMark* cm) :
199 _base(NULL), _cm(cm)
200 {}
201
202 bool G1CMMarkStack::allocate(size_t capacity) {
203 // allocate a stack of the requisite depth
204 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop)));
205 if (!rs.is_reserved()) {
206 log_warning(gc)("ConcurrentMark MarkStack allocation failure");
207 return false;
208 }
209 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC);
210 if (!_virtual_space.initialize(rs, rs.size())) {
211 log_warning(gc)("ConcurrentMark MarkStack backing store failure");
212 // Release the virtual memory reserved for the marking stack
213 rs.release();
214 return false;
215 }
216 assert(_virtual_space.committed_size() == rs.size(),
217 "Didn't reserve backing store for all of G1ConcurrentMark stack?");
218 _base = (oop*) _virtual_space.low();
219 setEmpty();
220 _capacity = (jint) capacity;
221 _saved_index = -1;
222 _should_expand = false;
223 return true;
224 }
225
226 void G1CMMarkStack::expand() {
227 // Called, during remark, if we've overflown the marking stack during marking.
228 assert(isEmpty(), "stack should been emptied while handling overflow");
229 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted");
230 // Clear expansion flag
231 _should_expand = false;
232 if (_capacity == (jint) MarkStackSizeMax) {
233 log_trace(gc)("(benign) Can't expand marking stack capacity, at max size limit");
234 return;
235 }
236 // Double capacity if possible
237 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax);
238 // Do not give up existing stack until we have managed to
239 // get the double capacity that we desired.
240 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity *
241 sizeof(oop)));
242 if (rs.is_reserved()) {
243 // Release the backing store associated with old stack
244 _virtual_space.release();
245 // Reinitialize virtual space for new stack
246 if (!_virtual_space.initialize(rs, rs.size())) {
247 fatal("Not enough swap for expanded marking stack capacity");
248 }
249 _base = (oop*)(_virtual_space.low());
250 _index = 0;
251 _capacity = new_capacity;
252 } else {
253 // Failed to double capacity, continue;
254 log_trace(gc)("(benign) Failed to expand marking stack capacity from " SIZE_FORMAT "K to " SIZE_FORMAT "K",
255 _capacity / K, new_capacity / K);
256 }
257 }
258
259 void G1CMMarkStack::set_should_expand() {
260 // If we're resetting the marking state because of an
261 // marking stack overflow, record that we should, if
262 // possible, expand the stack.
263 _should_expand = _cm->has_overflown();
264 }
265
266 G1CMMarkStack::~G1CMMarkStack() {
267 if (_base != NULL) {
268 _base = NULL;
269 _virtual_space.release();
270 }
271 }
272
273 void G1CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
274 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
275 jint start = _index;
276 jint next_index = start + n;
277 if (next_index > _capacity) {
278 _overflow = true;
279 return;
280 }
281 // Otherwise.
282 _index = next_index;
283 for (int i = 0; i < n; i++) {
284 int ind = start + i;
285 assert(ind < _capacity, "By overflow test above.");
286 _base[ind] = ptr_arr[i];
287 }
288 }
289
290 bool G1CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
291 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
292 jint index = _index;
293 if (index == 0) {
294 *n = 0;
295 return false;
296 } else {
297 int k = MIN2(max, index);
298 jint new_ind = index - k;
299 for (int j = 0; j < k; j++) {
300 ptr_arr[j] = _base[new_ind + j];
301 }
302 _index = new_ind;
303 *n = k;
304 return true;
305 }
306 }
307
308 void G1CMMarkStack::note_start_of_gc() {
309 assert(_saved_index == -1,
310 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
311 _saved_index = _index;
312 }
313
314 void G1CMMarkStack::note_end_of_gc() {
315 // This is intentionally a guarantee, instead of an assert. If we
316 // accidentally add something to the mark stack during GC, it
317 // will be a correctness issue so it's better if we crash. we'll
318 // only check this once per GC anyway, so it won't be a performance
319 // issue in any way.
320 guarantee(_saved_index == _index,
321 "saved index: %d index: %d", _saved_index, _index);
322 _saved_index = -1;
323 }
324
325 G1CMRootRegions::G1CMRootRegions() :
326 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
327 _should_abort(false), _next_survivor(NULL) { }
328
329 void G1CMRootRegions::init(G1CollectedHeap* g1h, G1ConcurrentMark* cm) {
330 _young_list = g1h->young_list();
331 _cm = cm;
332 }
333
334 void G1CMRootRegions::prepare_for_scan() {
335 assert(!scan_in_progress(), "pre-condition");
336
337 // Currently, only survivors can be root regions.
338 assert(_next_survivor == NULL, "pre-condition");
339 _next_survivor = _young_list->first_survivor_region();
340 _scan_in_progress = (_next_survivor != NULL);
341 _should_abort = false;
342 }
343
344 HeapRegion* G1CMRootRegions::claim_next() {
345 if (_should_abort) {
346 // If someone has set the should_abort flag, we return NULL to
347 // force the caller to bail out of their loop.
348 return NULL;
349 }
350
351 // Currently, only survivors can be root regions.
352 HeapRegion* res = _next_survivor;
353 if (res != NULL) {
354 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
355 // Read it again in case it changed while we were waiting for the lock.
356 res = _next_survivor;
357 if (res != NULL) {
358 if (res == _young_list->last_survivor_region()) {
359 // We just claimed the last survivor so store NULL to indicate
360 // that we're done.
361 _next_survivor = NULL;
362 } else {
363 _next_survivor = res->get_next_young_region();
364 }
365 } else {
366 // Someone else claimed the last survivor while we were trying
367 // to take the lock so nothing else to do.
368 }
369 }
370 assert(res == NULL || res->is_survivor(), "post-condition");
371
372 return res;
373 }
374
375 void G1CMRootRegions::notify_scan_done() {
376 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
377 _scan_in_progress = false;
378 RootRegionScan_lock->notify_all();
379 }
380
381 void G1CMRootRegions::cancel_scan() {
382 notify_scan_done();
383 }
384
385 void G1CMRootRegions::scan_finished() {
386 assert(scan_in_progress(), "pre-condition");
387
388 // Currently, only survivors can be root regions.
389 if (!_should_abort) {
390 assert(_next_survivor == NULL, "we should have claimed all survivors");
391 }
392 _next_survivor = NULL;
393
394 notify_scan_done();
395 }
396
397 bool G1CMRootRegions::wait_until_scan_finished() {
398 if (!scan_in_progress()) return false;
399
400 {
401 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
402 while (scan_in_progress()) {
403 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
404 }
405 }
406 return true;
407 }
408
409 uint G1ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
410 return MAX2((n_par_threads + 2) / 4, 1U);
411 }
412
413 G1ConcurrentMark::G1ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) :
414 _g1h(g1h),
415 _markBitMap1(),
416 _markBitMap2(),
417 _parallel_marking_threads(0),
418 _max_parallel_marking_threads(0),
419 _sleep_factor(0.0),
420 _marking_task_overhead(1.0),
421 _cleanup_list("Cleanup List"),
422 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/),
423 _card_bm((g1h->reserved_region().byte_size() + CardTableModRefBS::card_size - 1) >>
424 CardTableModRefBS::card_shift,
425 false /* in_resource_area*/),
426
427 _prevMarkBitMap(&_markBitMap1),
428 _nextMarkBitMap(&_markBitMap2),
429
430 _markStack(this),
431 // _finger set in set_non_marking_state
432
433 _max_worker_id(ParallelGCThreads),
434 // _active_tasks set in set_non_marking_state
435 // _tasks set inside the constructor
436 _task_queues(new G1CMTaskQueueSet((int) _max_worker_id)),
437 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)),
438
439 _has_overflown(false),
440 _concurrent(false),
441 _has_aborted(false),
442 _restart_for_overflow(false),
443 _concurrent_marking_in_progress(false),
444 _concurrent_phase_status(ConcPhaseNotStarted),
445
446 // _verbose_level set below
447
448 _init_times(),
449 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
450 _cleanup_times(),
451 _total_counting_time(0.0),
452 _total_rs_scrub_time(0.0),
453
454 _parallel_workers(NULL),
455
456 _count_card_bitmaps(NULL),
457 _count_marked_bytes(NULL),
458 _completed_initialization(false) {
459
460 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage);
461 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage);
462
463 // Create & start a ConcurrentMark thread.
464 _cmThread = new ConcurrentMarkThread(this);
465 assert(cmThread() != NULL, "CM Thread should have been created");
466 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
467 if (_cmThread->osthread() == NULL) {
468 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
469 }
470
471 assert(CGC_lock != NULL, "Where's the CGC_lock?");
472 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency");
473 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency");
474
475 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
476 satb_qs.set_buffer_size(G1SATBBufferSize);
477
478 _root_regions.init(_g1h, this);
479
480 if (ConcGCThreads > ParallelGCThreads) {
481 log_warning(gc)("Can't have more ConcGCThreads (%u) than ParallelGCThreads (%u).",
482 ConcGCThreads, ParallelGCThreads);
483 return;
484 }
485 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
486 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
487 // if both are set
488 _sleep_factor = 0.0;
489 _marking_task_overhead = 1.0;
490 } else if (G1MarkingOverheadPercent > 0) {
491 // We will calculate the number of parallel marking threads based
492 // on a target overhead with respect to the soft real-time goal
493 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
494 double overall_cm_overhead =
495 (double) MaxGCPauseMillis * marking_overhead /
496 (double) GCPauseIntervalMillis;
497 double cpu_ratio = 1.0 / (double) os::processor_count();
498 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
499 double marking_task_overhead =
500 overall_cm_overhead / marking_thread_num *
501 (double) os::processor_count();
502 double sleep_factor =
503 (1.0 - marking_task_overhead) / marking_task_overhead;
504
505 FLAG_SET_ERGO(uint, ConcGCThreads, (uint) marking_thread_num);
506 _sleep_factor = sleep_factor;
507 _marking_task_overhead = marking_task_overhead;
508 } else {
509 // Calculate the number of parallel marking threads by scaling
510 // the number of parallel GC threads.
511 uint marking_thread_num = scale_parallel_threads(ParallelGCThreads);
512 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num);
513 _sleep_factor = 0.0;
514 _marking_task_overhead = 1.0;
515 }
516
517 assert(ConcGCThreads > 0, "Should have been set");
518 _parallel_marking_threads = ConcGCThreads;
519 _max_parallel_marking_threads = _parallel_marking_threads;
520
521 _parallel_workers = new WorkGang("G1 Marker",
522 _max_parallel_marking_threads, false, true);
523 if (_parallel_workers == NULL) {
524 vm_exit_during_initialization("Failed necessary allocation.");
525 } else {
526 _parallel_workers->initialize_workers();
527 }
528
529 if (FLAG_IS_DEFAULT(MarkStackSize)) {
530 size_t mark_stack_size =
531 MIN2(MarkStackSizeMax,
532 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE)));
533 // Verify that the calculated value for MarkStackSize is in range.
534 // It would be nice to use the private utility routine from Arguments.
535 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
536 log_warning(gc)("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
537 "must be between 1 and " SIZE_FORMAT,
538 mark_stack_size, MarkStackSizeMax);
539 return;
540 }
541 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
542 } else {
543 // Verify MarkStackSize is in range.
544 if (FLAG_IS_CMDLINE(MarkStackSize)) {
545 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
546 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
547 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
548 "must be between 1 and " SIZE_FORMAT,
549 MarkStackSize, MarkStackSizeMax);
550 return;
551 }
552 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
553 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
554 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
555 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
556 MarkStackSize, MarkStackSizeMax);
557 return;
558 }
559 }
560 }
561 }
562
563 if (!_markStack.allocate(MarkStackSize)) {
564 log_warning(gc)("Failed to allocate CM marking stack");
565 return;
566 }
567
568 _tasks = NEW_C_HEAP_ARRAY(G1CMTask*, _max_worker_id, mtGC);
569 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC);
570
571 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC);
572 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC);
573
574 BitMap::idx_t card_bm_size = _card_bm.size();
575
576 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
577 _active_tasks = _max_worker_id;
578
579 uint max_regions = _g1h->max_regions();
580 for (uint i = 0; i < _max_worker_id; ++i) {
581 G1CMTaskQueue* task_queue = new G1CMTaskQueue();
582 task_queue->initialize();
583 _task_queues->register_queue(i, task_queue);
584
585 _count_card_bitmaps[i] = BitMap(card_bm_size, false);
586 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC);
587
588 _tasks[i] = new G1CMTask(i, this,
589 _count_marked_bytes[i],
590 &_count_card_bitmaps[i],
591 task_queue, _task_queues);
592
593 _accum_task_vtime[i] = 0.0;
594 }
595
596 // Calculate the card number for the bottom of the heap. Used
597 // in biasing indexes into the accounting card bitmaps.
598 _heap_bottom_card_num =
599 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
600 CardTableModRefBS::card_shift);
601
602 // Clear all the liveness counting data
603 clear_all_count_data();
604
605 // so that the call below can read a sensible value
606 _heap_start = g1h->reserved_region().start();
607 set_non_marking_state();
608 _completed_initialization = true;
609 }
610
611 void G1ConcurrentMark::reset() {
612 // Starting values for these two. This should be called in a STW
613 // phase.
614 MemRegion reserved = _g1h->g1_reserved();
615 _heap_start = reserved.start();
616 _heap_end = reserved.end();
617
618 // Separated the asserts so that we know which one fires.
619 assert(_heap_start != NULL, "heap bounds should look ok");
620 assert(_heap_end != NULL, "heap bounds should look ok");
621 assert(_heap_start < _heap_end, "heap bounds should look ok");
622
623 // Reset all the marking data structures and any necessary flags
624 reset_marking_state();
625
626 // We do reset all of them, since different phases will use
627 // different number of active threads. So, it's easiest to have all
628 // of them ready.
629 for (uint i = 0; i < _max_worker_id; ++i) {
630 _tasks[i]->reset(_nextMarkBitMap);
631 }
632
633 // we need this to make sure that the flag is on during the evac
634 // pause with initial mark piggy-backed
635 set_concurrent_marking_in_progress();
636 }
637
638
639 void G1ConcurrentMark::reset_marking_state(bool clear_overflow) {
640 _markStack.set_should_expand();
641 _markStack.setEmpty(); // Also clears the _markStack overflow flag
642 if (clear_overflow) {
643 clear_has_overflown();
644 } else {
645 assert(has_overflown(), "pre-condition");
646 }
647 _finger = _heap_start;
648
649 for (uint i = 0; i < _max_worker_id; ++i) {
650 G1CMTaskQueue* queue = _task_queues->queue(i);
651 queue->set_empty();
652 }
653 }
654
655 void G1ConcurrentMark::set_concurrency(uint active_tasks) {
656 assert(active_tasks <= _max_worker_id, "we should not have more");
657
658 _active_tasks = active_tasks;
659 // Need to update the three data structures below according to the
660 // number of active threads for this phase.
661 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
662 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
663 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
664 }
665
666 void G1ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
667 set_concurrency(active_tasks);
668
669 _concurrent = concurrent;
670 // We propagate this to all tasks, not just the active ones.
671 for (uint i = 0; i < _max_worker_id; ++i)
672 _tasks[i]->set_concurrent(concurrent);
673
674 if (concurrent) {
675 set_concurrent_marking_in_progress();
676 } else {
677 // We currently assume that the concurrent flag has been set to
678 // false before we start remark. At this point we should also be
679 // in a STW phase.
680 assert(!concurrent_marking_in_progress(), "invariant");
681 assert(out_of_regions(),
682 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT,
683 p2i(_finger), p2i(_heap_end));
684 }
685 }
686
687 void G1ConcurrentMark::set_non_marking_state() {
688 // We set the global marking state to some default values when we're
689 // not doing marking.
690 reset_marking_state();
691 _active_tasks = 0;
692 clear_concurrent_marking_in_progress();
693 }
694
695 G1ConcurrentMark::~G1ConcurrentMark() {
696 // The G1ConcurrentMark instance is never freed.
697 ShouldNotReachHere();
698 }
699
700 void G1ConcurrentMark::clearNextBitmap() {
701 G1CollectedHeap* g1h = G1CollectedHeap::heap();
702
703 // Make sure that the concurrent mark thread looks to still be in
704 // the current cycle.
705 guarantee(cmThread()->during_cycle(), "invariant");
706
707 // We are finishing up the current cycle by clearing the next
708 // marking bitmap and getting it ready for the next cycle. During
709 // this time no other cycle can start. So, let's make sure that this
710 // is the case.
711 guarantee(!g1h->collector_state()->mark_in_progress(), "invariant");
712
713 ClearBitmapHRClosure cl(this, _nextMarkBitMap, true /* may_yield */);
714 ParClearNextMarkBitmapTask task(&cl, parallel_marking_threads(), true);
715 _parallel_workers->run_task(&task);
716
717 // Clear the liveness counting data. If the marking has been aborted, the abort()
718 // call already did that.
719 if (cl.complete()) {
720 clear_all_count_data();
721 }
722
723 // Repeat the asserts from above.
724 guarantee(cmThread()->during_cycle(), "invariant");
725 guarantee(!g1h->collector_state()->mark_in_progress(), "invariant");
726 }
727
728 class CheckBitmapClearHRClosure : public HeapRegionClosure {
729 G1CMBitMap* _bitmap;
730 bool _error;
731 public:
732 CheckBitmapClearHRClosure(G1CMBitMap* bitmap) : _bitmap(bitmap) {
733 }
734
735 virtual bool doHeapRegion(HeapRegion* r) {
736 // This closure can be called concurrently to the mutator, so we must make sure
737 // that the result of the getNextMarkedWordAddress() call is compared to the
738 // value passed to it as limit to detect any found bits.
739 // end never changes in G1.
740 HeapWord* end = r->end();
741 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end;
742 }
743 };
744
745 bool G1ConcurrentMark::nextMarkBitmapIsClear() {
746 CheckBitmapClearHRClosure cl(_nextMarkBitMap);
747 _g1h->heap_region_iterate(&cl);
748 return cl.complete();
749 }
750
751 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
752 public:
753 bool doHeapRegion(HeapRegion* r) {
754 r->note_start_of_marking();
755 return false;
756 }
757 };
758
759 void G1ConcurrentMark::checkpointRootsInitialPre() {
760 G1CollectedHeap* g1h = G1CollectedHeap::heap();
761 G1CollectorPolicy* g1p = g1h->g1_policy();
762
763 _has_aborted = false;
764
765 // Initialize marking structures. This has to be done in a STW phase.
766 reset();
767
768 // For each region note start of marking.
769 NoteStartOfMarkHRClosure startcl;
770 g1h->heap_region_iterate(&startcl);
771 }
772
773
774 void G1ConcurrentMark::checkpointRootsInitialPost() {
775 G1CollectedHeap* g1h = G1CollectedHeap::heap();
776
777 // Start Concurrent Marking weak-reference discovery.
778 ReferenceProcessor* rp = g1h->ref_processor_cm();
779 // enable ("weak") refs discovery
780 rp->enable_discovery();
781 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
782
783 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
784 // This is the start of the marking cycle, we're expected all
785 // threads to have SATB queues with active set to false.
786 satb_mq_set.set_active_all_threads(true, /* new active value */
787 false /* expected_active */);
788
789 _root_regions.prepare_for_scan();
790
791 // update_g1_committed() will be called at the end of an evac pause
792 // when marking is on. So, it's also called at the end of the
793 // initial-mark pause to update the heap end, if the heap expands
794 // during it. No need to call it here.
795 }
796
797 /*
798 * Notice that in the next two methods, we actually leave the STS
799 * during the barrier sync and join it immediately afterwards. If we
800 * do not do this, the following deadlock can occur: one thread could
801 * be in the barrier sync code, waiting for the other thread to also
802 * sync up, whereas another one could be trying to yield, while also
803 * waiting for the other threads to sync up too.
804 *
805 * Note, however, that this code is also used during remark and in
806 * this case we should not attempt to leave / enter the STS, otherwise
807 * we'll either hit an assert (debug / fastdebug) or deadlock
808 * (product). So we should only leave / enter the STS if we are
809 * operating concurrently.
810 *
811 * Because the thread that does the sync barrier has left the STS, it
812 * is possible to be suspended for a Full GC or an evacuation pause
813 * could occur. This is actually safe, since the entering the sync
814 * barrier is one of the last things do_marking_step() does, and it
815 * doesn't manipulate any data structures afterwards.
816 */
817
818 void G1ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
819 bool barrier_aborted;
820 {
821 SuspendibleThreadSetLeaver sts_leave(concurrent());
822 barrier_aborted = !_first_overflow_barrier_sync.enter();
823 }
824
825 // at this point everyone should have synced up and not be doing any
826 // more work
827
828 if (barrier_aborted) {
829 // If the barrier aborted we ignore the overflow condition and
830 // just abort the whole marking phase as quickly as possible.
831 return;
832 }
833
834 // If we're executing the concurrent phase of marking, reset the marking
835 // state; otherwise the marking state is reset after reference processing,
836 // during the remark pause.
837 // If we reset here as a result of an overflow during the remark we will
838 // see assertion failures from any subsequent set_concurrency_and_phase()
839 // calls.
840 if (concurrent()) {
841 // let the task associated with with worker 0 do this
842 if (worker_id == 0) {
843 // task 0 is responsible for clearing the global data structures
844 // We should be here because of an overflow. During STW we should
845 // not clear the overflow flag since we rely on it being true when
846 // we exit this method to abort the pause and restart concurrent
847 // marking.
848 reset_marking_state(true /* clear_overflow */);
849
850 log_info(gc)("Concurrent Mark reset for overflow");
851 }
852 }
853
854 // after this, each task should reset its own data structures then
855 // then go into the second barrier
856 }
857
858 void G1ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
859 SuspendibleThreadSetLeaver sts_leave(concurrent());
860 _second_overflow_barrier_sync.enter();
861
862 // at this point everything should be re-initialized and ready to go
863 }
864
865 class G1CMConcurrentMarkingTask: public AbstractGangTask {
866 private:
867 G1ConcurrentMark* _cm;
868 ConcurrentMarkThread* _cmt;
869
870 public:
871 void work(uint worker_id) {
872 assert(Thread::current()->is_ConcurrentGC_thread(),
873 "this should only be done by a conc GC thread");
874 ResourceMark rm;
875
876 double start_vtime = os::elapsedVTime();
877
878 {
879 SuspendibleThreadSetJoiner sts_join;
880
881 assert(worker_id < _cm->active_tasks(), "invariant");
882 G1CMTask* the_task = _cm->task(worker_id);
883 the_task->record_start_time();
884 if (!_cm->has_aborted()) {
885 do {
886 double start_vtime_sec = os::elapsedVTime();
887 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
888
889 the_task->do_marking_step(mark_step_duration_ms,
890 true /* do_termination */,
891 false /* is_serial*/);
892
893 double end_vtime_sec = os::elapsedVTime();
894 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
895 _cm->clear_has_overflown();
896
897 _cm->do_yield_check(worker_id);
898
899 jlong sleep_time_ms;
900 if (!_cm->has_aborted() && the_task->has_aborted()) {
901 sleep_time_ms =
902 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
903 {
904 SuspendibleThreadSetLeaver sts_leave;
905 os::sleep(Thread::current(), sleep_time_ms, false);
906 }
907 }
908 } while (!_cm->has_aborted() && the_task->has_aborted());
909 }
910 the_task->record_end_time();
911 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
912 }
913
914 double end_vtime = os::elapsedVTime();
915 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
916 }
917
918 G1CMConcurrentMarkingTask(G1ConcurrentMark* cm,
919 ConcurrentMarkThread* cmt) :
920 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
921
922 ~G1CMConcurrentMarkingTask() { }
923 };
924
925 // Calculates the number of active workers for a concurrent
926 // phase.
927 uint G1ConcurrentMark::calc_parallel_marking_threads() {
928 uint n_conc_workers = 0;
929 if (!UseDynamicNumberOfGCThreads ||
930 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
931 !ForceDynamicNumberOfGCThreads)) {
932 n_conc_workers = max_parallel_marking_threads();
933 } else {
934 n_conc_workers =
935 AdaptiveSizePolicy::calc_default_active_workers(
936 max_parallel_marking_threads(),
937 1, /* Minimum workers */
938 parallel_marking_threads(),
939 Threads::number_of_non_daemon_threads());
940 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
941 // that scaling has already gone into "_max_parallel_marking_threads".
942 }
943 assert(n_conc_workers > 0, "Always need at least 1");
944 return n_conc_workers;
945 }
946
947 void G1ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
948 // Currently, only survivors can be root regions.
949 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
950 G1RootRegionScanClosure cl(_g1h, this, worker_id);
951
952 const uintx interval = PrefetchScanIntervalInBytes;
953 HeapWord* curr = hr->bottom();
954 const HeapWord* end = hr->top();
955 while (curr < end) {
956 Prefetch::read(curr, interval);
957 oop obj = oop(curr);
958 int size = obj->oop_iterate_size(&cl);
959 assert(size == obj->size(), "sanity");
960 curr += size;
961 }
962 }
963
964 class G1CMRootRegionScanTask : public AbstractGangTask {
965 private:
966 G1ConcurrentMark* _cm;
967
968 public:
969 G1CMRootRegionScanTask(G1ConcurrentMark* cm) :
970 AbstractGangTask("Root Region Scan"), _cm(cm) { }
971
972 void work(uint worker_id) {
973 assert(Thread::current()->is_ConcurrentGC_thread(),
974 "this should only be done by a conc GC thread");
975
976 G1CMRootRegions* root_regions = _cm->root_regions();
977 HeapRegion* hr = root_regions->claim_next();
978 while (hr != NULL) {
979 _cm->scanRootRegion(hr, worker_id);
980 hr = root_regions->claim_next();
981 }
982 }
983 };
984
985 void G1ConcurrentMark::scanRootRegions() {
986 // scan_in_progress() will have been set to true only if there was
987 // at least one root region to scan. So, if it's false, we
988 // should not attempt to do any further work.
989 if (root_regions()->scan_in_progress()) {
990 assert(!has_aborted(), "Aborting before root region scanning is finished not supported.");
991 GCTraceConcTime(Info, gc) tt("Concurrent Root Region Scan");
992
993 _parallel_marking_threads = calc_parallel_marking_threads();
994 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
995 "Maximum number of marking threads exceeded");
996 uint active_workers = MAX2(1U, parallel_marking_threads());
997
998 G1CMRootRegionScanTask task(this);
999 _parallel_workers->set_active_workers(active_workers);
1000 _parallel_workers->run_task(&task);
1001
1002 // It's possible that has_aborted() is true here without actually
1003 // aborting the survivor scan earlier. This is OK as it's
1004 // mainly used for sanity checking.
1005 root_regions()->scan_finished();
1006 }
1007 }
1008
1009 void G1ConcurrentMark::register_concurrent_phase_start(const char* title) {
1010 uint old_val = 0;
1011 do {
1012 old_val = Atomic::cmpxchg(ConcPhaseStarted, &_concurrent_phase_status, ConcPhaseNotStarted);
1013 } while (old_val != ConcPhaseNotStarted);
1014 _g1h->gc_timer_cm()->register_gc_concurrent_start(title);
1015 }
1016
1017 void G1ConcurrentMark::register_concurrent_phase_end_common(bool end_timer) {
1018 if (_concurrent_phase_status == ConcPhaseNotStarted) {
1019 return;
1020 }
1021
1022 uint old_val = Atomic::cmpxchg(ConcPhaseStopping, &_concurrent_phase_status, ConcPhaseStarted);
1023 if (old_val == ConcPhaseStarted) {
1024 _g1h->gc_timer_cm()->register_gc_concurrent_end();
1025 // If 'end_timer' is true, we came here to end timer which needs concurrent phase ended.
1026 // We need to end it before changing the status to 'ConcPhaseNotStarted' to prevent
1027 // starting a new concurrent phase by 'ConcurrentMarkThread'.
1028 if (end_timer) {
1029 _g1h->gc_timer_cm()->register_gc_end();
1030 }
1031 old_val = Atomic::cmpxchg(ConcPhaseNotStarted, &_concurrent_phase_status, ConcPhaseStopping);
1032 assert(old_val == ConcPhaseStopping, "Should not have changed since we entered this scope.");
1033 } else {
1034 do {
1035 // Let other thread finish changing '_concurrent_phase_status' to 'ConcPhaseNotStarted'.
1036 os::naked_short_sleep(1);
1037 } while (_concurrent_phase_status != ConcPhaseNotStarted);
1038 }
1039 }
1040
1041 void G1ConcurrentMark::register_concurrent_phase_end() {
1042 register_concurrent_phase_end_common(false);
1043 }
1044
1045 void G1ConcurrentMark::register_concurrent_gc_end_and_stop_timer() {
1046 register_concurrent_phase_end_common(true);
1047 }
1048
1049 void G1ConcurrentMark::markFromRoots() {
1050 // we might be tempted to assert that:
1051 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1052 // "inconsistent argument?");
1053 // However that wouldn't be right, because it's possible that
1054 // a safepoint is indeed in progress as a younger generation
1055 // stop-the-world GC happens even as we mark in this generation.
1056
1057 _restart_for_overflow = false;
1058
1059 // _g1h has _n_par_threads
1060 _parallel_marking_threads = calc_parallel_marking_threads();
1061 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1062 "Maximum number of marking threads exceeded");
1063
1064 uint active_workers = MAX2(1U, parallel_marking_threads());
1065 assert(active_workers > 0, "Should have been set");
1066
1067 // Parallel task terminator is set in "set_concurrency_and_phase()"
1068 set_concurrency_and_phase(active_workers, true /* concurrent */);
1069
1070 G1CMConcurrentMarkingTask markingTask(this, cmThread());
1071 _parallel_workers->set_active_workers(active_workers);
1072 _parallel_workers->run_task(&markingTask);
1073 print_stats();
1074 }
1075
1076 void G1ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1077 // world is stopped at this checkpoint
1078 assert(SafepointSynchronize::is_at_safepoint(),
1079 "world should be stopped");
1080
1081 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1082
1083 // If a full collection has happened, we shouldn't do this.
1084 if (has_aborted()) {
1085 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1086 return;
1087 }
1088
1089 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1090
1091 if (VerifyDuringGC) {
1092 HandleMark hm; // handle scope
1093 g1h->prepare_for_verify();
1094 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)");
1095 }
1096 g1h->verifier()->check_bitmaps("Remark Start");
1097
1098 G1CollectorPolicy* g1p = g1h->g1_policy();
1099 g1p->record_concurrent_mark_remark_start();
1100
1101 double start = os::elapsedTime();
1102
1103 checkpointRootsFinalWork();
1104
1105 double mark_work_end = os::elapsedTime();
1106
1107 weakRefsWork(clear_all_soft_refs);
1108
1109 if (has_overflown()) {
1110 // Oops. We overflowed. Restart concurrent marking.
1111 _restart_for_overflow = true;
1112 log_develop_trace(gc)("Remark led to restart for overflow.");
1113
1114 // Verify the heap w.r.t. the previous marking bitmap.
1115 if (VerifyDuringGC) {
1116 HandleMark hm; // handle scope
1117 g1h->prepare_for_verify();
1118 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (overflow)");
1119 }
1120
1121 // Clear the marking state because we will be restarting
1122 // marking due to overflowing the global mark stack.
1123 reset_marking_state();
1124 } else {
1125 {
1126 GCTraceTime(Debug, gc, phases) trace("Aggregate Data", g1h->gc_timer_cm());
1127
1128 // Aggregate the per-task counting data that we have accumulated
1129 // while marking.
1130 aggregate_count_data();
1131 }
1132
1133 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1134 // We're done with marking.
1135 // This is the end of the marking cycle, we're expected all
1136 // threads to have SATB queues with active set to true.
1137 satb_mq_set.set_active_all_threads(false, /* new active value */
1138 true /* expected_active */);
1139
1140 if (VerifyDuringGC) {
1141 HandleMark hm; // handle scope
1142 g1h->prepare_for_verify();
1143 Universe::verify(VerifyOption_G1UseNextMarking, "During GC (after)");
1144 }
1145 g1h->verifier()->check_bitmaps("Remark End");
1146 assert(!restart_for_overflow(), "sanity");
1147 // Completely reset the marking state since marking completed
1148 set_non_marking_state();
1149 }
1150
1151 // Expand the marking stack, if we have to and if we can.
1152 if (_markStack.should_expand()) {
1153 _markStack.expand();
1154 }
1155
1156 // Statistics
1157 double now = os::elapsedTime();
1158 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1159 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1160 _remark_times.add((now - start) * 1000.0);
1161
1162 g1p->record_concurrent_mark_remark_end();
1163
1164 G1CMIsAliveClosure is_alive(g1h);
1165 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive);
1166 }
1167
1168 // Base class of the closures that finalize and verify the
1169 // liveness counting data.
1170 class G1CMCountDataClosureBase: public HeapRegionClosure {
1171 protected:
1172 G1CollectedHeap* _g1h;
1173 G1ConcurrentMark* _cm;
1174 CardTableModRefBS* _ct_bs;
1175
1176 BitMap* _region_bm;
1177 BitMap* _card_bm;
1178
1179 // Takes a region that's not empty (i.e., it has at least one
1180 // live object in it and sets its corresponding bit on the region
1181 // bitmap to 1.
1182 void set_bit_for_region(HeapRegion* hr) {
1183 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1184 _region_bm->par_at_put(index, true);
1185 }
1186
1187 public:
1188 G1CMCountDataClosureBase(G1CollectedHeap* g1h,
1189 BitMap* region_bm, BitMap* card_bm):
1190 _g1h(g1h), _cm(g1h->concurrent_mark()),
1191 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
1192 _region_bm(region_bm), _card_bm(card_bm) { }
1193 };
1194
1195 // Closure that calculates the # live objects per region. Used
1196 // for verification purposes during the cleanup pause.
1197 class CalcLiveObjectsClosure: public G1CMCountDataClosureBase {
1198 G1CMBitMapRO* _bm;
1199 size_t _region_marked_bytes;
1200
1201 public:
1202 CalcLiveObjectsClosure(G1CMBitMapRO *bm, G1CollectedHeap* g1h,
1203 BitMap* region_bm, BitMap* card_bm) :
1204 G1CMCountDataClosureBase(g1h, region_bm, card_bm),
1205 _bm(bm), _region_marked_bytes(0) { }
1206
1207 bool doHeapRegion(HeapRegion* hr) {
1208 HeapWord* ntams = hr->next_top_at_mark_start();
1209 HeapWord* start = hr->bottom();
1210
1211 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1212 "Preconditions not met - "
1213 "start: " PTR_FORMAT ", ntams: " PTR_FORMAT ", end: " PTR_FORMAT,
1214 p2i(start), p2i(ntams), p2i(hr->end()));
1215
1216 // Find the first marked object at or after "start".
1217 start = _bm->getNextMarkedWordAddress(start, ntams);
1218
1219 size_t marked_bytes = 0;
1220
1221 while (start < ntams) {
1222 oop obj = oop(start);
1223 int obj_sz = obj->size();
1224 HeapWord* obj_end = start + obj_sz;
1225
1226 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
1227 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end);
1228
1229 // Note: if we're looking at the last region in heap - obj_end
1230 // could be actually just beyond the end of the heap; end_idx
1231 // will then correspond to a (non-existent) card that is also
1232 // just beyond the heap.
1233 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) {
1234 // end of object is not card aligned - increment to cover
1235 // all the cards spanned by the object
1236 end_idx += 1;
1237 }
1238
1239 // Set the bits in the card BM for the cards spanned by this object.
1240 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1241
1242 // Add the size of this object to the number of marked bytes.
1243 marked_bytes += (size_t)obj_sz * HeapWordSize;
1244
1245 // This will happen if we are handling a humongous object that spans
1246 // several heap regions.
1247 if (obj_end > hr->end()) {
1248 break;
1249 }
1250 // Find the next marked object after this one.
1251 start = _bm->getNextMarkedWordAddress(obj_end, ntams);
1252 }
1253
1254 // Mark the allocated-since-marking portion...
1255 HeapWord* top = hr->top();
1256 if (ntams < top) {
1257 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1258 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1259
1260 // Note: if we're looking at the last region in heap - top
1261 // could be actually just beyond the end of the heap; end_idx
1262 // will then correspond to a (non-existent) card that is also
1263 // just beyond the heap.
1264 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1265 // end of object is not card aligned - increment to cover
1266 // all the cards spanned by the object
1267 end_idx += 1;
1268 }
1269 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1270
1271 // This definitely means the region has live objects.
1272 set_bit_for_region(hr);
1273 }
1274
1275 // Update the live region bitmap.
1276 if (marked_bytes > 0) {
1277 set_bit_for_region(hr);
1278 }
1279
1280 // Set the marked bytes for the current region so that
1281 // it can be queried by a calling verification routine
1282 _region_marked_bytes = marked_bytes;
1283
1284 return false;
1285 }
1286
1287 size_t region_marked_bytes() const { return _region_marked_bytes; }
1288 };
1289
1290 // Heap region closure used for verifying the counting data
1291 // that was accumulated concurrently and aggregated during
1292 // the remark pause. This closure is applied to the heap
1293 // regions during the STW cleanup pause.
1294
1295 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure {
1296 G1CollectedHeap* _g1h;
1297 G1ConcurrentMark* _cm;
1298 CalcLiveObjectsClosure _calc_cl;
1299 BitMap* _region_bm; // Region BM to be verified
1300 BitMap* _card_bm; // Card BM to be verified
1301
1302 BitMap* _exp_region_bm; // Expected Region BM values
1303 BitMap* _exp_card_bm; // Expected card BM values
1304
1305 int _failures;
1306
1307 public:
1308 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h,
1309 BitMap* region_bm,
1310 BitMap* card_bm,
1311 BitMap* exp_region_bm,
1312 BitMap* exp_card_bm) :
1313 _g1h(g1h), _cm(g1h->concurrent_mark()),
1314 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm),
1315 _region_bm(region_bm), _card_bm(card_bm),
1316 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm),
1317 _failures(0) { }
1318
1319 int failures() const { return _failures; }
1320
1321 bool doHeapRegion(HeapRegion* hr) {
1322 int failures = 0;
1323
1324 // Call the CalcLiveObjectsClosure to walk the marking bitmap for
1325 // this region and set the corresponding bits in the expected region
1326 // and card bitmaps.
1327 bool res = _calc_cl.doHeapRegion(hr);
1328 assert(res == false, "should be continuing");
1329
1330 // Verify the marked bytes for this region.
1331 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1332 size_t act_marked_bytes = hr->next_marked_bytes();
1333
1334 if (exp_marked_bytes > act_marked_bytes) {
1335 if (hr->is_starts_humongous()) {
1336 // For start_humongous regions, the size of the whole object will be
1337 // in exp_marked_bytes.
1338 HeapRegion* region = hr;
1339 int num_regions;
1340 for (num_regions = 0; region != NULL; num_regions++) {
1341 region = _g1h->next_region_in_humongous(region);
1342 }
1343 if ((num_regions-1) * HeapRegion::GrainBytes >= exp_marked_bytes) {
1344 failures += 1;
1345 } else if (num_regions * HeapRegion::GrainBytes < exp_marked_bytes) {
1346 failures += 1;
1347 }
1348 } else {
1349 // We're not OK if expected marked bytes > actual marked bytes. It means
1350 // we have missed accounting some objects during the actual marking.
1351 failures += 1;
1352 }
1353 }
1354
1355 // Verify the bit, for this region, in the actual and expected
1356 // (which was just calculated) region bit maps.
1357 // We're not OK if the bit in the calculated expected region
1358 // bitmap is set and the bit in the actual region bitmap is not.
1359 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1360
1361 bool expected = _exp_region_bm->at(index);
1362 bool actual = _region_bm->at(index);
1363 if (expected && !actual) {
1364 failures += 1;
1365 }
1366
1367 // Verify that the card bit maps for the cards spanned by the current
1368 // region match. We have an error if we have a set bit in the expected
1369 // bit map and the corresponding bit in the actual bitmap is not set.
1370
1371 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom());
1372 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top());
1373
1374 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1375 expected = _exp_card_bm->at(i);
1376 actual = _card_bm->at(i);
1377
1378 if (expected && !actual) {
1379 failures += 1;
1380 }
1381 }
1382
1383 _failures += failures;
1384
1385 // We could stop iteration over the heap when we
1386 // find the first violating region by returning true.
1387 return false;
1388 }
1389 };
1390
1391 class G1ParVerifyFinalCountTask: public AbstractGangTask {
1392 protected:
1393 G1CollectedHeap* _g1h;
1394 G1ConcurrentMark* _cm;
1395 BitMap* _actual_region_bm;
1396 BitMap* _actual_card_bm;
1397
1398 uint _n_workers;
1399
1400 BitMap* _expected_region_bm;
1401 BitMap* _expected_card_bm;
1402
1403 int _failures;
1404
1405 HeapRegionClaimer _hrclaimer;
1406
1407 public:
1408 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h,
1409 BitMap* region_bm, BitMap* card_bm,
1410 BitMap* expected_region_bm, BitMap* expected_card_bm)
1411 : AbstractGangTask("G1 verify final counting"),
1412 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1413 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1414 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm),
1415 _failures(0),
1416 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1417 assert(VerifyDuringGC, "don't call this otherwise");
1418 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity");
1419 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity");
1420 }
1421
1422 void work(uint worker_id) {
1423 assert(worker_id < _n_workers, "invariant");
1424
1425 VerifyLiveObjectDataHRClosure verify_cl(_g1h,
1426 _actual_region_bm, _actual_card_bm,
1427 _expected_region_bm,
1428 _expected_card_bm);
1429
1430 _g1h->heap_region_par_iterate(&verify_cl, worker_id, &_hrclaimer);
1431
1432 Atomic::add(verify_cl.failures(), &_failures);
1433 }
1434
1435 int failures() const { return _failures; }
1436 };
1437
1438 // Closure that finalizes the liveness counting data.
1439 // Used during the cleanup pause.
1440 // Sets the bits corresponding to the interval [NTAMS, top]
1441 // (which contains the implicitly live objects) in the
1442 // card liveness bitmap. Also sets the bit for each region,
1443 // containing live data, in the region liveness bitmap.
1444
1445 class FinalCountDataUpdateClosure: public G1CMCountDataClosureBase {
1446 public:
1447 FinalCountDataUpdateClosure(G1CollectedHeap* g1h,
1448 BitMap* region_bm,
1449 BitMap* card_bm) :
1450 G1CMCountDataClosureBase(g1h, region_bm, card_bm) { }
1451
1452 bool doHeapRegion(HeapRegion* hr) {
1453 HeapWord* ntams = hr->next_top_at_mark_start();
1454 HeapWord* top = hr->top();
1455
1456 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1457
1458 // Mark the allocated-since-marking portion...
1459 if (ntams < top) {
1460 // This definitely means the region has live objects.
1461 set_bit_for_region(hr);
1462
1463 // Now set the bits in the card bitmap for [ntams, top)
1464 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1465 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1466
1467 // Note: if we're looking at the last region in heap - top
1468 // could be actually just beyond the end of the heap; end_idx
1469 // will then correspond to a (non-existent) card that is also
1470 // just beyond the heap.
1471 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1472 // end of object is not card aligned - increment to cover
1473 // all the cards spanned by the object
1474 end_idx += 1;
1475 }
1476
1477 assert(end_idx <= _card_bm->size(),
1478 "oob: end_idx= " SIZE_FORMAT ", bitmap size= " SIZE_FORMAT,
1479 end_idx, _card_bm->size());
1480 assert(start_idx < _card_bm->size(),
1481 "oob: start_idx= " SIZE_FORMAT ", bitmap size= " SIZE_FORMAT,
1482 start_idx, _card_bm->size());
1483
1484 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1485 }
1486
1487 // Set the bit for the region if it contains live data
1488 if (hr->next_marked_bytes() > 0) {
1489 set_bit_for_region(hr);
1490 }
1491
1492 return false;
1493 }
1494 };
1495
1496 class G1ParFinalCountTask: public AbstractGangTask {
1497 protected:
1498 G1CollectedHeap* _g1h;
1499 G1ConcurrentMark* _cm;
1500 BitMap* _actual_region_bm;
1501 BitMap* _actual_card_bm;
1502
1503 uint _n_workers;
1504 HeapRegionClaimer _hrclaimer;
1505
1506 public:
1507 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm)
1508 : AbstractGangTask("G1 final counting"),
1509 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1510 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1511 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1512 }
1513
1514 void work(uint worker_id) {
1515 assert(worker_id < _n_workers, "invariant");
1516
1517 FinalCountDataUpdateClosure final_update_cl(_g1h,
1518 _actual_region_bm,
1519 _actual_card_bm);
1520
1521 _g1h->heap_region_par_iterate(&final_update_cl, worker_id, &_hrclaimer);
1522 }
1523 };
1524
1525 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1526 G1CollectedHeap* _g1;
1527 size_t _freed_bytes;
1528 FreeRegionList* _local_cleanup_list;
1529 uint _old_regions_removed;
1530 uint _humongous_regions_removed;
1531 HRRSCleanupTask* _hrrs_cleanup_task;
1532
1533 public:
1534 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1535 FreeRegionList* local_cleanup_list,
1536 HRRSCleanupTask* hrrs_cleanup_task) :
1537 _g1(g1),
1538 _freed_bytes(0),
1539 _local_cleanup_list(local_cleanup_list),
1540 _old_regions_removed(0),
1541 _humongous_regions_removed(0),
1542 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1543
1544 size_t freed_bytes() { return _freed_bytes; }
1545 const uint old_regions_removed() { return _old_regions_removed; }
1546 const uint humongous_regions_removed() { return _humongous_regions_removed; }
1547
1548 bool doHeapRegion(HeapRegion *hr) {
1549 if (hr->is_archive()) {
1550 return false;
1551 }
1552 // We use a claim value of zero here because all regions
1553 // were claimed with value 1 in the FinalCount task.
1554 _g1->reset_gc_time_stamps(hr);
1555 hr->note_end_of_marking();
1556
1557 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) {
1558 _freed_bytes += hr->used();
1559 hr->set_containing_set(NULL);
1560 if (hr->is_humongous()) {
1561 _humongous_regions_removed++;
1562 _g1->free_humongous_region(hr, _local_cleanup_list, true);
1563 } else {
1564 _old_regions_removed++;
1565 _g1->free_region(hr, _local_cleanup_list, true);
1566 }
1567 } else {
1568 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1569 }
1570
1571 return false;
1572 }
1573 };
1574
1575 class G1ParNoteEndTask: public AbstractGangTask {
1576 friend class G1NoteEndOfConcMarkClosure;
1577
1578 protected:
1579 G1CollectedHeap* _g1h;
1580 FreeRegionList* _cleanup_list;
1581 HeapRegionClaimer _hrclaimer;
1582
1583 public:
1584 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1585 AbstractGangTask("G1 note end"), _g1h(g1h), _cleanup_list(cleanup_list), _hrclaimer(n_workers) {
1586 }
1587
1588 void work(uint worker_id) {
1589 FreeRegionList local_cleanup_list("Local Cleanup List");
1590 HRRSCleanupTask hrrs_cleanup_task;
1591 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
1592 &hrrs_cleanup_task);
1593 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer);
1594 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1595
1596 // Now update the lists
1597 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed());
1598 {
1599 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1600 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes());
1601
1602 // If we iterate over the global cleanup list at the end of
1603 // cleanup to do this printing we will not guarantee to only
1604 // generate output for the newly-reclaimed regions (the list
1605 // might not be empty at the beginning of cleanup; we might
1606 // still be working on its previous contents). So we do the
1607 // printing here, before we append the new regions to the global
1608 // cleanup list.
1609
1610 G1HRPrinter* hr_printer = _g1h->hr_printer();
1611 if (hr_printer->is_active()) {
1612 FreeRegionListIterator iter(&local_cleanup_list);
1613 while (iter.more_available()) {
1614 HeapRegion* hr = iter.get_next();
1615 hr_printer->cleanup(hr);
1616 }
1617 }
1618
1619 _cleanup_list->add_ordered(&local_cleanup_list);
1620 assert(local_cleanup_list.is_empty(), "post-condition");
1621
1622 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1623 }
1624 }
1625 };
1626
1627 void G1ConcurrentMark::cleanup() {
1628 // world is stopped at this checkpoint
1629 assert(SafepointSynchronize::is_at_safepoint(),
1630 "world should be stopped");
1631 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1632
1633 // If a full collection has happened, we shouldn't do this.
1634 if (has_aborted()) {
1635 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused
1636 return;
1637 }
1638
1639 g1h->verifier()->verify_region_sets_optional();
1640
1641 if (VerifyDuringGC) {
1642 HandleMark hm; // handle scope
1643 g1h->prepare_for_verify();
1644 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)");
1645 }
1646 g1h->verifier()->check_bitmaps("Cleanup Start");
1647
1648 G1CollectorPolicy* g1p = g1h->g1_policy();
1649 g1p->record_concurrent_mark_cleanup_start();
1650
1651 double start = os::elapsedTime();
1652
1653 HeapRegionRemSet::reset_for_cleanup_tasks();
1654
1655 // Do counting once more with the world stopped for good measure.
1656 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
1657
1658 g1h->workers()->run_task(&g1_par_count_task);
1659
1660 if (VerifyDuringGC) {
1661 // Verify that the counting data accumulated during marking matches
1662 // that calculated by walking the marking bitmap.
1663
1664 // Bitmaps to hold expected values
1665 BitMap expected_region_bm(_region_bm.size(), true);
1666 BitMap expected_card_bm(_card_bm.size(), true);
1667
1668 G1ParVerifyFinalCountTask g1_par_verify_task(g1h,
1669 &_region_bm,
1670 &_card_bm,
1671 &expected_region_bm,
1672 &expected_card_bm);
1673
1674 g1h->workers()->run_task(&g1_par_verify_task);
1675
1676 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures");
1677 }
1678
1679 size_t start_used_bytes = g1h->used();
1680 g1h->collector_state()->set_mark_in_progress(false);
1681
1682 double count_end = os::elapsedTime();
1683 double this_final_counting_time = (count_end - start);
1684 _total_counting_time += this_final_counting_time;
1685
1686 if (log_is_enabled(Trace, gc, liveness)) {
1687 G1PrintRegionLivenessInfoClosure cl("Post-Marking");
1688 _g1h->heap_region_iterate(&cl);
1689 }
1690
1691 // Install newly created mark bitMap as "prev".
1692 swapMarkBitMaps();
1693
1694 g1h->reset_gc_time_stamp();
1695
1696 uint n_workers = _g1h->workers()->active_workers();
1697
1698 // Note end of marking in all heap regions.
1699 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers);
1700 g1h->workers()->run_task(&g1_par_note_end_task);
1701 g1h->check_gc_time_stamps();
1702
1703 if (!cleanup_list_is_empty()) {
1704 // The cleanup list is not empty, so we'll have to process it
1705 // concurrently. Notify anyone else that might be wanting free
1706 // regions that there will be more free regions coming soon.
1707 g1h->set_free_regions_coming();
1708 }
1709
1710 // call below, since it affects the metric by which we sort the heap
1711 // regions.
1712 if (G1ScrubRemSets) {
1713 double rs_scrub_start = os::elapsedTime();
1714 g1h->scrub_rem_set(&_region_bm, &_card_bm);
1715 _total_rs_scrub_time += (os::elapsedTime() - rs_scrub_start);
1716 }
1717
1718 // this will also free any regions totally full of garbage objects,
1719 // and sort the regions.
1720 g1h->g1_policy()->record_concurrent_mark_cleanup_end();
1721
1722 // Statistics.
1723 double end = os::elapsedTime();
1724 _cleanup_times.add((end - start) * 1000.0);
1725
1726 // Clean up will have freed any regions completely full of garbage.
1727 // Update the soft reference policy with the new heap occupancy.
1728 Universe::update_heap_info_at_gc();
1729
1730 if (VerifyDuringGC) {
1731 HandleMark hm; // handle scope
1732 g1h->prepare_for_verify();
1733 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (after)");
1734 }
1735
1736 g1h->verifier()->check_bitmaps("Cleanup End");
1737
1738 g1h->verifier()->verify_region_sets_optional();
1739
1740 // We need to make this be a "collection" so any collection pause that
1741 // races with it goes around and waits for completeCleanup to finish.
1742 g1h->increment_total_collections();
1743
1744 // Clean out dead classes and update Metaspace sizes.
1745 if (ClassUnloadingWithConcurrentMark) {
1746 ClassLoaderDataGraph::purge();
1747 }
1748 MetaspaceGC::compute_new_size();
1749
1750 // We reclaimed old regions so we should calculate the sizes to make
1751 // sure we update the old gen/space data.
1752 g1h->g1mm()->update_sizes();
1753 g1h->allocation_context_stats().update_after_mark();
1754
1755 g1h->trace_heap_after_concurrent_cycle();
1756 }
1757
1758 void G1ConcurrentMark::completeCleanup() {
1759 if (has_aborted()) return;
1760
1761 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1762
1763 _cleanup_list.verify_optional();
1764 FreeRegionList tmp_free_list("Tmp Free List");
1765
1766 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : "
1767 "cleanup list has %u entries",
1768 _cleanup_list.length());
1769
1770 // No one else should be accessing the _cleanup_list at this point,
1771 // so it is not necessary to take any locks
1772 while (!_cleanup_list.is_empty()) {
1773 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */);
1774 assert(hr != NULL, "Got NULL from a non-empty list");
1775 hr->par_clear();
1776 tmp_free_list.add_ordered(hr);
1777
1778 // Instead of adding one region at a time to the secondary_free_list,
1779 // we accumulate them in the local list and move them a few at a
1780 // time. This also cuts down on the number of notify_all() calls
1781 // we do during this process. We'll also append the local list when
1782 // _cleanup_list is empty (which means we just removed the last
1783 // region from the _cleanup_list).
1784 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
1785 _cleanup_list.is_empty()) {
1786 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : "
1787 "appending %u entries to the secondary_free_list, "
1788 "cleanup list still has %u entries",
1789 tmp_free_list.length(),
1790 _cleanup_list.length());
1791
1792 {
1793 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
1794 g1h->secondary_free_list_add(&tmp_free_list);
1795 SecondaryFreeList_lock->notify_all();
1796 }
1797 #ifndef PRODUCT
1798 if (G1StressConcRegionFreeing) {
1799 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
1800 os::sleep(Thread::current(), (jlong) 1, false);
1801 }
1802 }
1803 #endif
1804 }
1805 }
1806 assert(tmp_free_list.is_empty(), "post-condition");
1807 }
1808
1809 // Supporting Object and Oop closures for reference discovery
1810 // and processing in during marking
1811
1812 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1813 HeapWord* addr = (HeapWord*)obj;
1814 return addr != NULL &&
1815 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
1816 }
1817
1818 // 'Keep Alive' oop closure used by both serial parallel reference processing.
1819 // Uses the G1CMTask associated with a worker thread (for serial reference
1820 // processing the G1CMTask for worker 0 is used) to preserve (mark) and
1821 // trace referent objects.
1822 //
1823 // Using the G1CMTask and embedded local queues avoids having the worker
1824 // threads operating on the global mark stack. This reduces the risk
1825 // of overflowing the stack - which we would rather avoid at this late
1826 // state. Also using the tasks' local queues removes the potential
1827 // of the workers interfering with each other that could occur if
1828 // operating on the global stack.
1829
1830 class G1CMKeepAliveAndDrainClosure: public OopClosure {
1831 G1ConcurrentMark* _cm;
1832 G1CMTask* _task;
1833 int _ref_counter_limit;
1834 int _ref_counter;
1835 bool _is_serial;
1836 public:
1837 G1CMKeepAliveAndDrainClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1838 _cm(cm), _task(task), _is_serial(is_serial),
1839 _ref_counter_limit(G1RefProcDrainInterval) {
1840 assert(_ref_counter_limit > 0, "sanity");
1841 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1842 _ref_counter = _ref_counter_limit;
1843 }
1844
1845 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1846 virtual void do_oop( oop* p) { do_oop_work(p); }
1847
1848 template <class T> void do_oop_work(T* p) {
1849 if (!_cm->has_overflown()) {
1850 oop obj = oopDesc::load_decode_heap_oop(p);
1851 _task->deal_with_reference(obj);
1852 _ref_counter--;
1853
1854 if (_ref_counter == 0) {
1855 // We have dealt with _ref_counter_limit references, pushing them
1856 // and objects reachable from them on to the local stack (and
1857 // possibly the global stack). Call G1CMTask::do_marking_step() to
1858 // process these entries.
1859 //
1860 // We call G1CMTask::do_marking_step() in a loop, which we'll exit if
1861 // there's nothing more to do (i.e. we're done with the entries that
1862 // were pushed as a result of the G1CMTask::deal_with_reference() calls
1863 // above) or we overflow.
1864 //
1865 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1866 // flag while there may still be some work to do. (See the comment at
1867 // the beginning of G1CMTask::do_marking_step() for those conditions -
1868 // one of which is reaching the specified time target.) It is only
1869 // when G1CMTask::do_marking_step() returns without setting the
1870 // has_aborted() flag that the marking step has completed.
1871 do {
1872 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1873 _task->do_marking_step(mark_step_duration_ms,
1874 false /* do_termination */,
1875 _is_serial);
1876 } while (_task->has_aborted() && !_cm->has_overflown());
1877 _ref_counter = _ref_counter_limit;
1878 }
1879 }
1880 }
1881 };
1882
1883 // 'Drain' oop closure used by both serial and parallel reference processing.
1884 // Uses the G1CMTask associated with a given worker thread (for serial
1885 // reference processing the G1CMtask for worker 0 is used). Calls the
1886 // do_marking_step routine, with an unbelievably large timeout value,
1887 // to drain the marking data structures of the remaining entries
1888 // added by the 'keep alive' oop closure above.
1889
1890 class G1CMDrainMarkingStackClosure: public VoidClosure {
1891 G1ConcurrentMark* _cm;
1892 G1CMTask* _task;
1893 bool _is_serial;
1894 public:
1895 G1CMDrainMarkingStackClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) :
1896 _cm(cm), _task(task), _is_serial(is_serial) {
1897 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
1898 }
1899
1900 void do_void() {
1901 do {
1902 // We call G1CMTask::do_marking_step() to completely drain the local
1903 // and global marking stacks of entries pushed by the 'keep alive'
1904 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
1905 //
1906 // G1CMTask::do_marking_step() is called in a loop, which we'll exit
1907 // if there's nothing more to do (i.e. we've completely drained the
1908 // entries that were pushed as a a result of applying the 'keep alive'
1909 // closure to the entries on the discovered ref lists) or we overflow
1910 // the global marking stack.
1911 //
1912 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted()
1913 // flag while there may still be some work to do. (See the comment at
1914 // the beginning of G1CMTask::do_marking_step() for those conditions -
1915 // one of which is reaching the specified time target.) It is only
1916 // when G1CMTask::do_marking_step() returns without setting the
1917 // has_aborted() flag that the marking step has completed.
1918
1919 _task->do_marking_step(1000000000.0 /* something very large */,
1920 true /* do_termination */,
1921 _is_serial);
1922 } while (_task->has_aborted() && !_cm->has_overflown());
1923 }
1924 };
1925
1926 // Implementation of AbstractRefProcTaskExecutor for parallel
1927 // reference processing at the end of G1 concurrent marking
1928
1929 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
1930 private:
1931 G1CollectedHeap* _g1h;
1932 G1ConcurrentMark* _cm;
1933 WorkGang* _workers;
1934 uint _active_workers;
1935
1936 public:
1937 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
1938 G1ConcurrentMark* cm,
1939 WorkGang* workers,
1940 uint n_workers) :
1941 _g1h(g1h), _cm(cm),
1942 _workers(workers), _active_workers(n_workers) { }
1943
1944 // Executes the given task using concurrent marking worker threads.
1945 virtual void execute(ProcessTask& task);
1946 virtual void execute(EnqueueTask& task);
1947 };
1948
1949 class G1CMRefProcTaskProxy: public AbstractGangTask {
1950 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
1951 ProcessTask& _proc_task;
1952 G1CollectedHeap* _g1h;
1953 G1ConcurrentMark* _cm;
1954
1955 public:
1956 G1CMRefProcTaskProxy(ProcessTask& proc_task,
1957 G1CollectedHeap* g1h,
1958 G1ConcurrentMark* cm) :
1959 AbstractGangTask("Process reference objects in parallel"),
1960 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
1961 ReferenceProcessor* rp = _g1h->ref_processor_cm();
1962 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
1963 }
1964
1965 virtual void work(uint worker_id) {
1966 ResourceMark rm;
1967 HandleMark hm;
1968 G1CMTask* task = _cm->task(worker_id);
1969 G1CMIsAliveClosure g1_is_alive(_g1h);
1970 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
1971 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
1972
1973 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
1974 }
1975 };
1976
1977 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
1978 assert(_workers != NULL, "Need parallel worker threads.");
1979 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
1980
1981 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
1982
1983 // We need to reset the concurrency level before each
1984 // proxy task execution, so that the termination protocol
1985 // and overflow handling in G1CMTask::do_marking_step() knows
1986 // how many workers to wait for.
1987 _cm->set_concurrency(_active_workers);
1988 _workers->run_task(&proc_task_proxy);
1989 }
1990
1991 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
1992 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
1993 EnqueueTask& _enq_task;
1994
1995 public:
1996 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
1997 AbstractGangTask("Enqueue reference objects in parallel"),
1998 _enq_task(enq_task) { }
1999
2000 virtual void work(uint worker_id) {
2001 _enq_task.work(worker_id);
2002 }
2003 };
2004
2005 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2006 assert(_workers != NULL, "Need parallel worker threads.");
2007 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2008
2009 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2010
2011 // Not strictly necessary but...
2012 //
2013 // We need to reset the concurrency level before each
2014 // proxy task execution, so that the termination protocol
2015 // and overflow handling in G1CMTask::do_marking_step() knows
2016 // how many workers to wait for.
2017 _cm->set_concurrency(_active_workers);
2018 _workers->run_task(&enq_task_proxy);
2019 }
2020
2021 void G1ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) {
2022 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes);
2023 }
2024
2025 void G1ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2026 if (has_overflown()) {
2027 // Skip processing the discovered references if we have
2028 // overflown the global marking stack. Reference objects
2029 // only get discovered once so it is OK to not
2030 // de-populate the discovered reference lists. We could have,
2031 // but the only benefit would be that, when marking restarts,
2032 // less reference objects are discovered.
2033 return;
2034 }
2035
2036 ResourceMark rm;
2037 HandleMark hm;
2038
2039 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2040
2041 // Is alive closure.
2042 G1CMIsAliveClosure g1_is_alive(g1h);
2043
2044 // Inner scope to exclude the cleaning of the string and symbol
2045 // tables from the displayed time.
2046 {
2047 GCTraceTime(Debug, gc, phases) trace("Reference Processing", g1h->gc_timer_cm());
2048
2049 ReferenceProcessor* rp = g1h->ref_processor_cm();
2050
2051 // See the comment in G1CollectedHeap::ref_processing_init()
2052 // about how reference processing currently works in G1.
2053
2054 // Set the soft reference policy
2055 rp->setup_policy(clear_all_soft_refs);
2056 assert(_markStack.isEmpty(), "mark stack should be empty");
2057
2058 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2059 // in serial reference processing. Note these closures are also
2060 // used for serially processing (by the the current thread) the
2061 // JNI references during parallel reference processing.
2062 //
2063 // These closures do not need to synchronize with the worker
2064 // threads involved in parallel reference processing as these
2065 // instances are executed serially by the current thread (e.g.
2066 // reference processing is not multi-threaded and is thus
2067 // performed by the current thread instead of a gang worker).
2068 //
2069 // The gang tasks involved in parallel reference processing create
2070 // their own instances of these closures, which do their own
2071 // synchronization among themselves.
2072 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2073 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2074
2075 // We need at least one active thread. If reference processing
2076 // is not multi-threaded we use the current (VMThread) thread,
2077 // otherwise we use the work gang from the G1CollectedHeap and
2078 // we utilize all the worker threads we can.
2079 bool processing_is_mt = rp->processing_is_mt();
2080 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2081 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U);
2082
2083 // Parallel processing task executor.
2084 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2085 g1h->workers(), active_workers);
2086 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2087
2088 // Set the concurrency level. The phase was already set prior to
2089 // executing the remark task.
2090 set_concurrency(active_workers);
2091
2092 // Set the degree of MT processing here. If the discovery was done MT,
2093 // the number of threads involved during discovery could differ from
2094 // the number of active workers. This is OK as long as the discovered
2095 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2096 rp->set_active_mt_degree(active_workers);
2097
2098 // Process the weak references.
2099 const ReferenceProcessorStats& stats =
2100 rp->process_discovered_references(&g1_is_alive,
2101 &g1_keep_alive,
2102 &g1_drain_mark_stack,
2103 executor,
2104 g1h->gc_timer_cm());
2105 g1h->gc_tracer_cm()->report_gc_reference_stats(stats);
2106
2107 // The do_oop work routines of the keep_alive and drain_marking_stack
2108 // oop closures will set the has_overflown flag if we overflow the
2109 // global marking stack.
2110
2111 assert(_markStack.overflow() || _markStack.isEmpty(),
2112 "mark stack should be empty (unless it overflowed)");
2113
2114 if (_markStack.overflow()) {
2115 // This should have been done already when we tried to push an
2116 // entry on to the global mark stack. But let's do it again.
2117 set_has_overflown();
2118 }
2119
2120 assert(rp->num_q() == active_workers, "why not");
2121
2122 rp->enqueue_discovered_references(executor);
2123
2124 rp->verify_no_references_recorded();
2125 assert(!rp->discovery_enabled(), "Post condition");
2126 }
2127
2128 if (has_overflown()) {
2129 // We can not trust g1_is_alive if the marking stack overflowed
2130 return;
2131 }
2132
2133 assert(_markStack.isEmpty(), "Marking should have completed");
2134
2135 // Unload Klasses, String, Symbols, Code Cache, etc.
2136 if (ClassUnloadingWithConcurrentMark) {
2137 bool purged_classes;
2138
2139 {
2140 GCTraceTime(Debug, gc, phases) trace("System Dictionary Unloading", g1h->gc_timer_cm());
2141 purged_classes = SystemDictionary::do_unloading(&g1_is_alive, false /* Defer klass cleaning */);
2142 }
2143
2144 {
2145 GCTraceTime(Debug, gc, phases) trace("Parallel Unloading", g1h->gc_timer_cm());
2146 weakRefsWorkParallelPart(&g1_is_alive, purged_classes);
2147 }
2148 }
2149
2150 if (G1StringDedup::is_enabled()) {
2151 GCTraceTime(Debug, gc, phases) trace("String Deduplication Unlink", g1h->gc_timer_cm());
2152 G1StringDedup::unlink(&g1_is_alive);
2153 }
2154 }
2155
2156 void G1ConcurrentMark::swapMarkBitMaps() {
2157 G1CMBitMapRO* temp = _prevMarkBitMap;
2158 _prevMarkBitMap = (G1CMBitMapRO*)_nextMarkBitMap;
2159 _nextMarkBitMap = (G1CMBitMap*) temp;
2160 }
2161
2162 // Closure for marking entries in SATB buffers.
2163 class G1CMSATBBufferClosure : public SATBBufferClosure {
2164 private:
2165 G1CMTask* _task;
2166 G1CollectedHeap* _g1h;
2167
2168 // This is very similar to G1CMTask::deal_with_reference, but with
2169 // more relaxed requirements for the argument, so this must be more
2170 // circumspect about treating the argument as an object.
2171 void do_entry(void* entry) const {
2172 _task->increment_refs_reached();
2173 HeapRegion* hr = _g1h->heap_region_containing(entry);
2174 if (entry < hr->next_top_at_mark_start()) {
2175 // Until we get here, we don't know whether entry refers to a valid
2176 // object; it could instead have been a stale reference.
2177 oop obj = static_cast<oop>(entry);
2178 assert(obj->is_oop(true /* ignore mark word */),
2179 "Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj));
2180 _task->make_reference_grey(obj, hr);
2181 }
2182 }
2183
2184 public:
2185 G1CMSATBBufferClosure(G1CMTask* task, G1CollectedHeap* g1h)
2186 : _task(task), _g1h(g1h) { }
2187
2188 virtual void do_buffer(void** buffer, size_t size) {
2189 for (size_t i = 0; i < size; ++i) {
2190 do_entry(buffer[i]);
2191 }
2192 }
2193 };
2194
2195 class G1RemarkThreadsClosure : public ThreadClosure {
2196 G1CMSATBBufferClosure _cm_satb_cl;
2197 G1CMOopClosure _cm_cl;
2198 MarkingCodeBlobClosure _code_cl;
2199 int _thread_parity;
2200
2201 public:
2202 G1RemarkThreadsClosure(G1CollectedHeap* g1h, G1CMTask* task) :
2203 _cm_satb_cl(task, g1h),
2204 _cm_cl(g1h, g1h->concurrent_mark(), task),
2205 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
2206 _thread_parity(Threads::thread_claim_parity()) {}
2207
2208 void do_thread(Thread* thread) {
2209 if (thread->is_Java_thread()) {
2210 if (thread->claim_oops_do(true, _thread_parity)) {
2211 JavaThread* jt = (JavaThread*)thread;
2212
2213 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
2214 // however the liveness of oops reachable from nmethods have very complex lifecycles:
2215 // * Alive if on the stack of an executing method
2216 // * Weakly reachable otherwise
2217 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
2218 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
2219 jt->nmethods_do(&_code_cl);
2220
2221 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
2222 }
2223 } else if (thread->is_VM_thread()) {
2224 if (thread->claim_oops_do(true, _thread_parity)) {
2225 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
2226 }
2227 }
2228 }
2229 };
2230
2231 class G1CMRemarkTask: public AbstractGangTask {
2232 private:
2233 G1ConcurrentMark* _cm;
2234 public:
2235 void work(uint worker_id) {
2236 // Since all available tasks are actually started, we should
2237 // only proceed if we're supposed to be active.
2238 if (worker_id < _cm->active_tasks()) {
2239 G1CMTask* task = _cm->task(worker_id);
2240 task->record_start_time();
2241 {
2242 ResourceMark rm;
2243 HandleMark hm;
2244
2245 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
2246 Threads::threads_do(&threads_f);
2247 }
2248
2249 do {
2250 task->do_marking_step(1000000000.0 /* something very large */,
2251 true /* do_termination */,
2252 false /* is_serial */);
2253 } while (task->has_aborted() && !_cm->has_overflown());
2254 // If we overflow, then we do not want to restart. We instead
2255 // want to abort remark and do concurrent marking again.
2256 task->record_end_time();
2257 }
2258 }
2259
2260 G1CMRemarkTask(G1ConcurrentMark* cm, uint active_workers) :
2261 AbstractGangTask("Par Remark"), _cm(cm) {
2262 _cm->terminator()->reset_for_reuse(active_workers);
2263 }
2264 };
2265
2266 void G1ConcurrentMark::checkpointRootsFinalWork() {
2267 ResourceMark rm;
2268 HandleMark hm;
2269 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2270
2271 GCTraceTime(Debug, gc, phases) trace("Finalize Marking", g1h->gc_timer_cm());
2272
2273 g1h->ensure_parsability(false);
2274
2275 // this is remark, so we'll use up all active threads
2276 uint active_workers = g1h->workers()->active_workers();
2277 set_concurrency_and_phase(active_workers, false /* concurrent */);
2278 // Leave _parallel_marking_threads at it's
2279 // value originally calculated in the G1ConcurrentMark
2280 // constructor and pass values of the active workers
2281 // through the gang in the task.
2282
2283 {
2284 StrongRootsScope srs(active_workers);
2285
2286 G1CMRemarkTask remarkTask(this, active_workers);
2287 // We will start all available threads, even if we decide that the
2288 // active_workers will be fewer. The extra ones will just bail out
2289 // immediately.
2290 g1h->workers()->run_task(&remarkTask);
2291 }
2292
2293 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2294 guarantee(has_overflown() ||
2295 satb_mq_set.completed_buffers_num() == 0,
2296 "Invariant: has_overflown = %s, num buffers = " SIZE_FORMAT,
2297 BOOL_TO_STR(has_overflown()),
2298 satb_mq_set.completed_buffers_num());
2299
2300 print_stats();
2301 }
2302
2303 void G1ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2304 // Note we are overriding the read-only view of the prev map here, via
2305 // the cast.
2306 ((G1CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2307 }
2308
2309 HeapRegion*
2310 G1ConcurrentMark::claim_region(uint worker_id) {
2311 // "checkpoint" the finger
2312 HeapWord* finger = _finger;
2313
2314 // _heap_end will not change underneath our feet; it only changes at
2315 // yield points.
2316 while (finger < _heap_end) {
2317 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2318
2319 HeapRegion* curr_region = _g1h->heap_region_containing(finger);
2320
2321 // Above heap_region_containing may return NULL as we always scan claim
2322 // until the end of the heap. In this case, just jump to the next region.
2323 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
2324
2325 // Is the gap between reading the finger and doing the CAS too long?
2326 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2327 if (res == finger && curr_region != NULL) {
2328 // we succeeded
2329 HeapWord* bottom = curr_region->bottom();
2330 HeapWord* limit = curr_region->next_top_at_mark_start();
2331
2332 // notice that _finger == end cannot be guaranteed here since,
2333 // someone else might have moved the finger even further
2334 assert(_finger >= end, "the finger should have moved forward");
2335
2336 if (limit > bottom) {
2337 return curr_region;
2338 } else {
2339 assert(limit == bottom,
2340 "the region limit should be at bottom");
2341 // we return NULL and the caller should try calling
2342 // claim_region() again.
2343 return NULL;
2344 }
2345 } else {
2346 assert(_finger > finger, "the finger should have moved forward");
2347 // read it again
2348 finger = _finger;
2349 }
2350 }
2351
2352 return NULL;
2353 }
2354
2355 #ifndef PRODUCT
2356 class VerifyNoCSetOops VALUE_OBJ_CLASS_SPEC {
2357 private:
2358 G1CollectedHeap* _g1h;
2359 const char* _phase;
2360 int _info;
2361
2362 public:
2363 VerifyNoCSetOops(const char* phase, int info = -1) :
2364 _g1h(G1CollectedHeap::heap()),
2365 _phase(phase),
2366 _info(info)
2367 { }
2368
2369 void operator()(oop obj) const {
2370 guarantee(obj->is_oop(),
2371 "Non-oop " PTR_FORMAT ", phase: %s, info: %d",
2372 p2i(obj), _phase, _info);
2373 guarantee(!_g1h->obj_in_cs(obj),
2374 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d",
2375 p2i(obj), _phase, _info);
2376 }
2377 };
2378
2379 void G1ConcurrentMark::verify_no_cset_oops() {
2380 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2381 if (!G1CollectedHeap::heap()->collector_state()->mark_in_progress()) {
2382 return;
2383 }
2384
2385 // Verify entries on the global mark stack
2386 _markStack.iterate(VerifyNoCSetOops("Stack"));
2387
2388 // Verify entries on the task queues
2389 for (uint i = 0; i < _max_worker_id; ++i) {
2390 G1CMTaskQueue* queue = _task_queues->queue(i);
2391 queue->iterate(VerifyNoCSetOops("Queue", i));
2392 }
2393
2394 // Verify the global finger
2395 HeapWord* global_finger = finger();
2396 if (global_finger != NULL && global_finger < _heap_end) {
2397 // Since we always iterate over all regions, we might get a NULL HeapRegion
2398 // here.
2399 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger);
2400 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
2401 "global finger: " PTR_FORMAT " region: " HR_FORMAT,
2402 p2i(global_finger), HR_FORMAT_PARAMS(global_hr));
2403 }
2404
2405 // Verify the task fingers
2406 assert(parallel_marking_threads() <= _max_worker_id, "sanity");
2407 for (uint i = 0; i < parallel_marking_threads(); ++i) {
2408 G1CMTask* task = _tasks[i];
2409 HeapWord* task_finger = task->finger();
2410 if (task_finger != NULL && task_finger < _heap_end) {
2411 // See above note on the global finger verification.
2412 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger);
2413 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
2414 !task_hr->in_collection_set(),
2415 "task finger: " PTR_FORMAT " region: " HR_FORMAT,
2416 p2i(task_finger), HR_FORMAT_PARAMS(task_hr));
2417 }
2418 }
2419 }
2420 #endif // PRODUCT
2421
2422 // Aggregate the counting data that was constructed concurrently
2423 // with marking.
2424 class AggregateCountDataHRClosure: public HeapRegionClosure {
2425 G1CollectedHeap* _g1h;
2426 G1ConcurrentMark* _cm;
2427 CardTableModRefBS* _ct_bs;
2428 BitMap* _cm_card_bm;
2429 uint _max_worker_id;
2430
2431 public:
2432 AggregateCountDataHRClosure(G1CollectedHeap* g1h,
2433 BitMap* cm_card_bm,
2434 uint max_worker_id) :
2435 _g1h(g1h), _cm(g1h->concurrent_mark()),
2436 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
2437 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { }
2438
2439 bool doHeapRegion(HeapRegion* hr) {
2440 HeapWord* start = hr->bottom();
2441 HeapWord* limit = hr->next_top_at_mark_start();
2442 HeapWord* end = hr->end();
2443
2444 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(),
2445 "Preconditions not met - "
2446 "start: " PTR_FORMAT ", limit: " PTR_FORMAT ", "
2447 "top: " PTR_FORMAT ", end: " PTR_FORMAT,
2448 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end()));
2449
2450 assert(hr->next_marked_bytes() == 0, "Precondition");
2451
2452 if (start == limit) {
2453 // NTAMS of this region has not been set so nothing to do.
2454 return false;
2455 }
2456
2457 // 'start' should be in the heap.
2458 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity");
2459 // 'end' *may* be just beyond the end of the heap (if hr is the last region)
2460 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity");
2461
2462 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
2463 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit);
2464 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end);
2465
2466 // If ntams is not card aligned then we bump card bitmap index
2467 // for limit so that we get the all the cards spanned by
2468 // the object ending at ntams.
2469 // Note: if this is the last region in the heap then ntams
2470 // could be actually just beyond the end of the the heap;
2471 // limit_idx will then correspond to a (non-existent) card
2472 // that is also outside the heap.
2473 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) {
2474 limit_idx += 1;
2475 }
2476
2477 assert(limit_idx <= end_idx, "or else use atomics");
2478
2479 // Aggregate the "stripe" in the count data associated with hr.
2480 uint hrm_index = hr->hrm_index();
2481 size_t marked_bytes = 0;
2482
2483 for (uint i = 0; i < _max_worker_id; i += 1) {
2484 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i);
2485 BitMap* task_card_bm = _cm->count_card_bitmap_for(i);
2486
2487 // Fetch the marked_bytes in this region for task i and
2488 // add it to the running total for this region.
2489 marked_bytes += marked_bytes_array[hrm_index];
2490
2491 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx)
2492 // into the global card bitmap.
2493 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx);
2494
2495 while (scan_idx < limit_idx) {
2496 assert(task_card_bm->at(scan_idx) == true, "should be");
2497 _cm_card_bm->set_bit(scan_idx);
2498 assert(_cm_card_bm->at(scan_idx) == true, "should be");
2499
2500 // BitMap::get_next_one_offset() can handle the case when
2501 // its left_offset parameter is greater than its right_offset
2502 // parameter. It does, however, have an early exit if
2503 // left_offset == right_offset. So let's limit the value
2504 // passed in for left offset here.
2505 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx);
2506 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx);
2507 }
2508 }
2509
2510 // Update the marked bytes for this region.
2511 hr->add_to_marked_bytes(marked_bytes);
2512
2513 // Next heap region
2514 return false;
2515 }
2516 };
2517
2518 class G1AggregateCountDataTask: public AbstractGangTask {
2519 protected:
2520 G1CollectedHeap* _g1h;
2521 G1ConcurrentMark* _cm;
2522 BitMap* _cm_card_bm;
2523 uint _max_worker_id;
2524 uint _active_workers;
2525 HeapRegionClaimer _hrclaimer;
2526
2527 public:
2528 G1AggregateCountDataTask(G1CollectedHeap* g1h,
2529 G1ConcurrentMark* cm,
2530 BitMap* cm_card_bm,
2531 uint max_worker_id,
2532 uint n_workers) :
2533 AbstractGangTask("Count Aggregation"),
2534 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm),
2535 _max_worker_id(max_worker_id),
2536 _active_workers(n_workers),
2537 _hrclaimer(_active_workers) {
2538 }
2539
2540 void work(uint worker_id) {
2541 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id);
2542
2543 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer);
2544 }
2545 };
2546
2547
2548 void G1ConcurrentMark::aggregate_count_data() {
2549 uint n_workers = _g1h->workers()->active_workers();
2550
2551 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm,
2552 _max_worker_id, n_workers);
2553
2554 _g1h->workers()->run_task(&g1_par_agg_task);
2555 }
2556
2557 // Clear the per-worker arrays used to store the per-region counting data
2558 void G1ConcurrentMark::clear_all_count_data() {
2559 // Clear the global card bitmap - it will be filled during
2560 // liveness count aggregation (during remark) and the
2561 // final counting task.
2562 _card_bm.clear();
2563
2564 // Clear the global region bitmap - it will be filled as part
2565 // of the final counting task.
2566 _region_bm.clear();
2567
2568 uint max_regions = _g1h->max_regions();
2569 assert(_max_worker_id > 0, "uninitialized");
2570
2571 for (uint i = 0; i < _max_worker_id; i += 1) {
2572 BitMap* task_card_bm = count_card_bitmap_for(i);
2573 size_t* marked_bytes_array = count_marked_bytes_array_for(i);
2574
2575 assert(task_card_bm->size() == _card_bm.size(), "size mismatch");
2576 assert(marked_bytes_array != NULL, "uninitialized");
2577
2578 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t));
2579 task_card_bm->clear();
2580 }
2581 }
2582
2583 void G1ConcurrentMark::print_stats() {
2584 if (!log_is_enabled(Debug, gc, stats)) {
2585 return;
2586 }
2587 log_debug(gc, stats)("---------------------------------------------------------------------");
2588 for (size_t i = 0; i < _active_tasks; ++i) {
2589 _tasks[i]->print_stats();
2590 log_debug(gc, stats)("---------------------------------------------------------------------");
2591 }
2592 }
2593
2594 // abandon current marking iteration due to a Full GC
2595 void G1ConcurrentMark::abort() {
2596 if (!cmThread()->during_cycle() || _has_aborted) {
2597 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything.
2598 return;
2599 }
2600
2601 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
2602 // concurrent bitmap clearing.
2603 _nextMarkBitMap->clearAll();
2604
2605 // Note we cannot clear the previous marking bitmap here
2606 // since VerifyDuringGC verifies the objects marked during
2607 // a full GC against the previous bitmap.
2608
2609 // Clear the liveness counting data
2610 clear_all_count_data();
2611 // Empty mark stack
2612 reset_marking_state();
2613 for (uint i = 0; i < _max_worker_id; ++i) {
2614 _tasks[i]->clear_region_fields();
2615 }
2616 _first_overflow_barrier_sync.abort();
2617 _second_overflow_barrier_sync.abort();
2618 _has_aborted = true;
2619
2620 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2621 satb_mq_set.abandon_partial_marking();
2622 // This can be called either during or outside marking, we'll read
2623 // the expected_active value from the SATB queue set.
2624 satb_mq_set.set_active_all_threads(
2625 false, /* new active value */
2626 satb_mq_set.is_active() /* expected_active */);
2627
2628 _g1h->trace_heap_after_concurrent_cycle();
2629
2630 _g1h->register_concurrent_cycle_end();
2631 }
2632
2633 static void print_ms_time_info(const char* prefix, const char* name,
2634 NumberSeq& ns) {
2635 log_trace(gc, marking)("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
2636 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
2637 if (ns.num() > 0) {
2638 log_trace(gc, marking)("%s [std. dev = %8.2f ms, max = %8.2f ms]",
2639 prefix, ns.sd(), ns.maximum());
2640 }
2641 }
2642
2643 void G1ConcurrentMark::print_summary_info() {
2644 LogHandle(gc, marking) log;
2645 if (!log.is_trace()) {
2646 return;
2647 }
2648
2649 log.trace(" Concurrent marking:");
2650 print_ms_time_info(" ", "init marks", _init_times);
2651 print_ms_time_info(" ", "remarks", _remark_times);
2652 {
2653 print_ms_time_info(" ", "final marks", _remark_mark_times);
2654 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
2655
2656 }
2657 print_ms_time_info(" ", "cleanups", _cleanup_times);
2658 log.trace(" Final counting total time = %8.2f s (avg = %8.2f ms).",
2659 _total_counting_time, (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
2660 if (G1ScrubRemSets) {
2661 log.trace(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
2662 _total_rs_scrub_time, (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / (double)_cleanup_times.num() : 0.0));
2663 }
2664 log.trace(" Total stop_world time = %8.2f s.",
2665 (_init_times.sum() + _remark_times.sum() + _cleanup_times.sum())/1000.0);
2666 log.trace(" Total concurrent time = %8.2f s (%8.2f s marking).",
2667 cmThread()->vtime_accum(), cmThread()->vtime_mark_accum());
2668 }
2669
2670 void G1ConcurrentMark::print_worker_threads_on(outputStream* st) const {
2671 _parallel_workers->print_worker_threads_on(st);
2672 }
2673
2674 void G1ConcurrentMark::print_on_error(outputStream* st) const {
2675 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
2676 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap));
2677 _prevMarkBitMap->print_on_error(st, " Prev Bits: ");
2678 _nextMarkBitMap->print_on_error(st, " Next Bits: ");
2679 }
2680
2681 // We take a break if someone is trying to stop the world.
2682 bool G1ConcurrentMark::do_yield_check(uint worker_id) {
2683 if (SuspendibleThreadSet::should_yield()) {
2684 SuspendibleThreadSet::yield();
2685 return true;
2686 } else {
2687 return false;
2688 }
2689 }
2690
2691 // Closure for iteration over bitmaps
2692 class G1CMBitMapClosure : public BitMapClosure {
2693 private:
2694 // the bitmap that is being iterated over
2695 G1CMBitMap* _nextMarkBitMap;
2696 G1ConcurrentMark* _cm;
2697 G1CMTask* _task;
2698
2699 public:
2700 G1CMBitMapClosure(G1CMTask *task, G1ConcurrentMark* cm, G1CMBitMap* nextMarkBitMap) :
2701 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
2702
2703 bool do_bit(size_t offset) {
2704 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
2705 assert(_nextMarkBitMap->isMarked(addr), "invariant");
2706 assert( addr < _cm->finger(), "invariant");
2707 assert(addr >= _task->finger(), "invariant");
2708
2709 // We move that task's local finger along.
2710 _task->move_finger_to(addr);
2711
2712 _task->scan_object(oop(addr));
2713 // we only partially drain the local queue and global stack
2714 _task->drain_local_queue(true);
2715 _task->drain_global_stack(true);
2716
2717 // if the has_aborted flag has been raised, we need to bail out of
2718 // the iteration
2719 return !_task->has_aborted();
2720 }
2721 };
2722
2723 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) {
2724 ReferenceProcessor* result = g1h->ref_processor_cm();
2725 assert(result != NULL, "CM reference processor should not be NULL");
2726 return result;
2727 }
2728
2729 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
2730 G1ConcurrentMark* cm,
2731 G1CMTask* task)
2732 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)),
2733 _g1h(g1h), _cm(cm), _task(task)
2734 { }
2735
2736 void G1CMTask::setup_for_region(HeapRegion* hr) {
2737 assert(hr != NULL,
2738 "claim_region() should have filtered out NULL regions");
2739 _curr_region = hr;
2740 _finger = hr->bottom();
2741 update_region_limit();
2742 }
2743
2744 void G1CMTask::update_region_limit() {
2745 HeapRegion* hr = _curr_region;
2746 HeapWord* bottom = hr->bottom();
2747 HeapWord* limit = hr->next_top_at_mark_start();
2748
2749 if (limit == bottom) {
2750 // The region was collected underneath our feet.
2751 // We set the finger to bottom to ensure that the bitmap
2752 // iteration that will follow this will not do anything.
2753 // (this is not a condition that holds when we set the region up,
2754 // as the region is not supposed to be empty in the first place)
2755 _finger = bottom;
2756 } else if (limit >= _region_limit) {
2757 assert(limit >= _finger, "peace of mind");
2758 } else {
2759 assert(limit < _region_limit, "only way to get here");
2760 // This can happen under some pretty unusual circumstances. An
2761 // evacuation pause empties the region underneath our feet (NTAMS
2762 // at bottom). We then do some allocation in the region (NTAMS
2763 // stays at bottom), followed by the region being used as a GC
2764 // alloc region (NTAMS will move to top() and the objects
2765 // originally below it will be grayed). All objects now marked in
2766 // the region are explicitly grayed, if below the global finger,
2767 // and we do not need in fact to scan anything else. So, we simply
2768 // set _finger to be limit to ensure that the bitmap iteration
2769 // doesn't do anything.
2770 _finger = limit;
2771 }
2772
2773 _region_limit = limit;
2774 }
2775
2776 void G1CMTask::giveup_current_region() {
2777 assert(_curr_region != NULL, "invariant");
2778 clear_region_fields();
2779 }
2780
2781 void G1CMTask::clear_region_fields() {
2782 // Values for these three fields that indicate that we're not
2783 // holding on to a region.
2784 _curr_region = NULL;
2785 _finger = NULL;
2786 _region_limit = NULL;
2787 }
2788
2789 void G1CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
2790 if (cm_oop_closure == NULL) {
2791 assert(_cm_oop_closure != NULL, "invariant");
2792 } else {
2793 assert(_cm_oop_closure == NULL, "invariant");
2794 }
2795 _cm_oop_closure = cm_oop_closure;
2796 }
2797
2798 void G1CMTask::reset(G1CMBitMap* nextMarkBitMap) {
2799 guarantee(nextMarkBitMap != NULL, "invariant");
2800 _nextMarkBitMap = nextMarkBitMap;
2801 clear_region_fields();
2802
2803 _calls = 0;
2804 _elapsed_time_ms = 0.0;
2805 _termination_time_ms = 0.0;
2806 _termination_start_time_ms = 0.0;
2807 }
2808
2809 bool G1CMTask::should_exit_termination() {
2810 regular_clock_call();
2811 // This is called when we are in the termination protocol. We should
2812 // quit if, for some reason, this task wants to abort or the global
2813 // stack is not empty (this means that we can get work from it).
2814 return !_cm->mark_stack_empty() || has_aborted();
2815 }
2816
2817 void G1CMTask::reached_limit() {
2818 assert(_words_scanned >= _words_scanned_limit ||
2819 _refs_reached >= _refs_reached_limit ,
2820 "shouldn't have been called otherwise");
2821 regular_clock_call();
2822 }
2823
2824 void G1CMTask::regular_clock_call() {
2825 if (has_aborted()) return;
2826
2827 // First, we need to recalculate the words scanned and refs reached
2828 // limits for the next clock call.
2829 recalculate_limits();
2830
2831 // During the regular clock call we do the following
2832
2833 // (1) If an overflow has been flagged, then we abort.
2834 if (_cm->has_overflown()) {
2835 set_has_aborted();
2836 return;
2837 }
2838
2839 // If we are not concurrent (i.e. we're doing remark) we don't need
2840 // to check anything else. The other steps are only needed during
2841 // the concurrent marking phase.
2842 if (!concurrent()) return;
2843
2844 // (2) If marking has been aborted for Full GC, then we also abort.
2845 if (_cm->has_aborted()) {
2846 set_has_aborted();
2847 return;
2848 }
2849
2850 double curr_time_ms = os::elapsedVTime() * 1000.0;
2851
2852 // (4) We check whether we should yield. If we have to, then we abort.
2853 if (SuspendibleThreadSet::should_yield()) {
2854 // We should yield. To do this we abort the task. The caller is
2855 // responsible for yielding.
2856 set_has_aborted();
2857 return;
2858 }
2859
2860 // (5) We check whether we've reached our time quota. If we have,
2861 // then we abort.
2862 double elapsed_time_ms = curr_time_ms - _start_time_ms;
2863 if (elapsed_time_ms > _time_target_ms) {
2864 set_has_aborted();
2865 _has_timed_out = true;
2866 return;
2867 }
2868
2869 // (6) Finally, we check whether there are enough completed STAB
2870 // buffers available for processing. If there are, we abort.
2871 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2872 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
2873 // we do need to process SATB buffers, we'll abort and restart
2874 // the marking task to do so
2875 set_has_aborted();
2876 return;
2877 }
2878 }
2879
2880 void G1CMTask::recalculate_limits() {
2881 _real_words_scanned_limit = _words_scanned + words_scanned_period;
2882 _words_scanned_limit = _real_words_scanned_limit;
2883
2884 _real_refs_reached_limit = _refs_reached + refs_reached_period;
2885 _refs_reached_limit = _real_refs_reached_limit;
2886 }
2887
2888 void G1CMTask::decrease_limits() {
2889 // This is called when we believe that we're going to do an infrequent
2890 // operation which will increase the per byte scanned cost (i.e. move
2891 // entries to/from the global stack). It basically tries to decrease the
2892 // scanning limit so that the clock is called earlier.
2893
2894 _words_scanned_limit = _real_words_scanned_limit -
2895 3 * words_scanned_period / 4;
2896 _refs_reached_limit = _real_refs_reached_limit -
2897 3 * refs_reached_period / 4;
2898 }
2899
2900 void G1CMTask::move_entries_to_global_stack() {
2901 // local array where we'll store the entries that will be popped
2902 // from the local queue
2903 oop buffer[global_stack_transfer_size];
2904
2905 int n = 0;
2906 oop obj;
2907 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
2908 buffer[n] = obj;
2909 ++n;
2910 }
2911
2912 if (n > 0) {
2913 // we popped at least one entry from the local queue
2914
2915 if (!_cm->mark_stack_push(buffer, n)) {
2916 set_has_aborted();
2917 }
2918 }
2919
2920 // this operation was quite expensive, so decrease the limits
2921 decrease_limits();
2922 }
2923
2924 void G1CMTask::get_entries_from_global_stack() {
2925 // local array where we'll store the entries that will be popped
2926 // from the global stack.
2927 oop buffer[global_stack_transfer_size];
2928 int n;
2929 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
2930 assert(n <= global_stack_transfer_size,
2931 "we should not pop more than the given limit");
2932 if (n > 0) {
2933 // yes, we did actually pop at least one entry
2934 for (int i = 0; i < n; ++i) {
2935 bool success = _task_queue->push(buffer[i]);
2936 // We only call this when the local queue is empty or under a
2937 // given target limit. So, we do not expect this push to fail.
2938 assert(success, "invariant");
2939 }
2940 }
2941
2942 // this operation was quite expensive, so decrease the limits
2943 decrease_limits();
2944 }
2945
2946 void G1CMTask::drain_local_queue(bool partially) {
2947 if (has_aborted()) return;
2948
2949 // Decide what the target size is, depending whether we're going to
2950 // drain it partially (so that other tasks can steal if they run out
2951 // of things to do) or totally (at the very end).
2952 size_t target_size;
2953 if (partially) {
2954 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
2955 } else {
2956 target_size = 0;
2957 }
2958
2959 if (_task_queue->size() > target_size) {
2960 oop obj;
2961 bool ret = _task_queue->pop_local(obj);
2962 while (ret) {
2963 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
2964 assert(!_g1h->is_on_master_free_list(
2965 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
2966
2967 scan_object(obj);
2968
2969 if (_task_queue->size() <= target_size || has_aborted()) {
2970 ret = false;
2971 } else {
2972 ret = _task_queue->pop_local(obj);
2973 }
2974 }
2975 }
2976 }
2977
2978 void G1CMTask::drain_global_stack(bool partially) {
2979 if (has_aborted()) return;
2980
2981 // We have a policy to drain the local queue before we attempt to
2982 // drain the global stack.
2983 assert(partially || _task_queue->size() == 0, "invariant");
2984
2985 // Decide what the target size is, depending whether we're going to
2986 // drain it partially (so that other tasks can steal if they run out
2987 // of things to do) or totally (at the very end). Notice that,
2988 // because we move entries from the global stack in chunks or
2989 // because another task might be doing the same, we might in fact
2990 // drop below the target. But, this is not a problem.
2991 size_t target_size;
2992 if (partially) {
2993 target_size = _cm->partial_mark_stack_size_target();
2994 } else {
2995 target_size = 0;
2996 }
2997
2998 if (_cm->mark_stack_size() > target_size) {
2999 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3000 get_entries_from_global_stack();
3001 drain_local_queue(partially);
3002 }
3003 }
3004 }
3005
3006 // SATB Queue has several assumptions on whether to call the par or
3007 // non-par versions of the methods. this is why some of the code is
3008 // replicated. We should really get rid of the single-threaded version
3009 // of the code to simplify things.
3010 void G1CMTask::drain_satb_buffers() {
3011 if (has_aborted()) return;
3012
3013 // We set this so that the regular clock knows that we're in the
3014 // middle of draining buffers and doesn't set the abort flag when it
3015 // notices that SATB buffers are available for draining. It'd be
3016 // very counter productive if it did that. :-)
3017 _draining_satb_buffers = true;
3018
3019 G1CMSATBBufferClosure satb_cl(this, _g1h);
3020 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3021
3022 // This keeps claiming and applying the closure to completed buffers
3023 // until we run out of buffers or we need to abort.
3024 while (!has_aborted() &&
3025 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
3026 regular_clock_call();
3027 }
3028
3029 _draining_satb_buffers = false;
3030
3031 assert(has_aborted() ||
3032 concurrent() ||
3033 satb_mq_set.completed_buffers_num() == 0, "invariant");
3034
3035 // again, this was a potentially expensive operation, decrease the
3036 // limits to get the regular clock call early
3037 decrease_limits();
3038 }
3039
3040 void G1CMTask::print_stats() {
3041 log_debug(gc, stats)("Marking Stats, task = %u, calls = %d",
3042 _worker_id, _calls);
3043 log_debug(gc, stats)(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
3044 _elapsed_time_ms, _termination_time_ms);
3045 log_debug(gc, stats)(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3046 _step_times_ms.num(), _step_times_ms.avg(),
3047 _step_times_ms.sd());
3048 log_debug(gc, stats)(" max = %1.2lfms, total = %1.2lfms",
3049 _step_times_ms.maximum(), _step_times_ms.sum());
3050 }
3051
3052 bool G1ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, oop& obj) {
3053 return _task_queues->steal(worker_id, hash_seed, obj);
3054 }
3055
3056 /*****************************************************************************
3057
3058 The do_marking_step(time_target_ms, ...) method is the building
3059 block of the parallel marking framework. It can be called in parallel
3060 with other invocations of do_marking_step() on different tasks
3061 (but only one per task, obviously) and concurrently with the
3062 mutator threads, or during remark, hence it eliminates the need
3063 for two versions of the code. When called during remark, it will
3064 pick up from where the task left off during the concurrent marking
3065 phase. Interestingly, tasks are also claimable during evacuation
3066 pauses too, since do_marking_step() ensures that it aborts before
3067 it needs to yield.
3068
3069 The data structures that it uses to do marking work are the
3070 following:
3071
3072 (1) Marking Bitmap. If there are gray objects that appear only
3073 on the bitmap (this happens either when dealing with an overflow
3074 or when the initial marking phase has simply marked the roots
3075 and didn't push them on the stack), then tasks claim heap
3076 regions whose bitmap they then scan to find gray objects. A
3077 global finger indicates where the end of the last claimed region
3078 is. A local finger indicates how far into the region a task has
3079 scanned. The two fingers are used to determine how to gray an
3080 object (i.e. whether simply marking it is OK, as it will be
3081 visited by a task in the future, or whether it needs to be also
3082 pushed on a stack).
3083
3084 (2) Local Queue. The local queue of the task which is accessed
3085 reasonably efficiently by the task. Other tasks can steal from
3086 it when they run out of work. Throughout the marking phase, a
3087 task attempts to keep its local queue short but not totally
3088 empty, so that entries are available for stealing by other
3089 tasks. Only when there is no more work, a task will totally
3090 drain its local queue.
3091
3092 (3) Global Mark Stack. This handles local queue overflow. During
3093 marking only sets of entries are moved between it and the local
3094 queues, as access to it requires a mutex and more fine-grain
3095 interaction with it which might cause contention. If it
3096 overflows, then the marking phase should restart and iterate
3097 over the bitmap to identify gray objects. Throughout the marking
3098 phase, tasks attempt to keep the global mark stack at a small
3099 length but not totally empty, so that entries are available for
3100 popping by other tasks. Only when there is no more work, tasks
3101 will totally drain the global mark stack.
3102
3103 (4) SATB Buffer Queue. This is where completed SATB buffers are
3104 made available. Buffers are regularly removed from this queue
3105 and scanned for roots, so that the queue doesn't get too
3106 long. During remark, all completed buffers are processed, as
3107 well as the filled in parts of any uncompleted buffers.
3108
3109 The do_marking_step() method tries to abort when the time target
3110 has been reached. There are a few other cases when the
3111 do_marking_step() method also aborts:
3112
3113 (1) When the marking phase has been aborted (after a Full GC).
3114
3115 (2) When a global overflow (on the global stack) has been
3116 triggered. Before the task aborts, it will actually sync up with
3117 the other tasks to ensure that all the marking data structures
3118 (local queues, stacks, fingers etc.) are re-initialized so that
3119 when do_marking_step() completes, the marking phase can
3120 immediately restart.
3121
3122 (3) When enough completed SATB buffers are available. The
3123 do_marking_step() method only tries to drain SATB buffers right
3124 at the beginning. So, if enough buffers are available, the
3125 marking step aborts and the SATB buffers are processed at
3126 the beginning of the next invocation.
3127
3128 (4) To yield. when we have to yield then we abort and yield
3129 right at the end of do_marking_step(). This saves us from a lot
3130 of hassle as, by yielding we might allow a Full GC. If this
3131 happens then objects will be compacted underneath our feet, the
3132 heap might shrink, etc. We save checking for this by just
3133 aborting and doing the yield right at the end.
3134
3135 From the above it follows that the do_marking_step() method should
3136 be called in a loop (or, otherwise, regularly) until it completes.
3137
3138 If a marking step completes without its has_aborted() flag being
3139 true, it means it has completed the current marking phase (and
3140 also all other marking tasks have done so and have all synced up).
3141
3142 A method called regular_clock_call() is invoked "regularly" (in
3143 sub ms intervals) throughout marking. It is this clock method that
3144 checks all the abort conditions which were mentioned above and
3145 decides when the task should abort. A work-based scheme is used to
3146 trigger this clock method: when the number of object words the
3147 marking phase has scanned or the number of references the marking
3148 phase has visited reach a given limit. Additional invocations to
3149 the method clock have been planted in a few other strategic places
3150 too. The initial reason for the clock method was to avoid calling
3151 vtime too regularly, as it is quite expensive. So, once it was in
3152 place, it was natural to piggy-back all the other conditions on it
3153 too and not constantly check them throughout the code.
3154
3155 If do_termination is true then do_marking_step will enter its
3156 termination protocol.
3157
3158 The value of is_serial must be true when do_marking_step is being
3159 called serially (i.e. by the VMThread) and do_marking_step should
3160 skip any synchronization in the termination and overflow code.
3161 Examples include the serial remark code and the serial reference
3162 processing closures.
3163
3164 The value of is_serial must be false when do_marking_step is
3165 being called by any of the worker threads in a work gang.
3166 Examples include the concurrent marking code (CMMarkingTask),
3167 the MT remark code, and the MT reference processing closures.
3168
3169 *****************************************************************************/
3170
3171 void G1CMTask::do_marking_step(double time_target_ms,
3172 bool do_termination,
3173 bool is_serial) {
3174 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
3175 assert(concurrent() == _cm->concurrent(), "they should be the same");
3176
3177 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
3178 assert(_task_queues != NULL, "invariant");
3179 assert(_task_queue != NULL, "invariant");
3180 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant");
3181
3182 assert(!_claimed,
3183 "only one thread should claim this task at any one time");
3184
3185 // OK, this doesn't safeguard again all possible scenarios, as it is
3186 // possible for two threads to set the _claimed flag at the same
3187 // time. But it is only for debugging purposes anyway and it will
3188 // catch most problems.
3189 _claimed = true;
3190
3191 _start_time_ms = os::elapsedVTime() * 1000.0;
3192
3193 // If do_stealing is true then do_marking_step will attempt to
3194 // steal work from the other G1CMTasks. It only makes sense to
3195 // enable stealing when the termination protocol is enabled
3196 // and do_marking_step() is not being called serially.
3197 bool do_stealing = do_termination && !is_serial;
3198
3199 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms);
3200 _time_target_ms = time_target_ms - diff_prediction_ms;
3201
3202 // set up the variables that are used in the work-based scheme to
3203 // call the regular clock method
3204 _words_scanned = 0;
3205 _refs_reached = 0;
3206 recalculate_limits();
3207
3208 // clear all flags
3209 clear_has_aborted();
3210 _has_timed_out = false;
3211 _draining_satb_buffers = false;
3212
3213 ++_calls;
3214
3215 // Set up the bitmap and oop closures. Anything that uses them is
3216 // eventually called from this method, so it is OK to allocate these
3217 // statically.
3218 G1CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
3219 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
3220 set_cm_oop_closure(&cm_oop_closure);
3221
3222 if (_cm->has_overflown()) {
3223 // This can happen if the mark stack overflows during a GC pause
3224 // and this task, after a yield point, restarts. We have to abort
3225 // as we need to get into the overflow protocol which happens
3226 // right at the end of this task.
3227 set_has_aborted();
3228 }
3229
3230 // First drain any available SATB buffers. After this, we will not
3231 // look at SATB buffers before the next invocation of this method.
3232 // If enough completed SATB buffers are queued up, the regular clock
3233 // will abort this task so that it restarts.
3234 drain_satb_buffers();
3235 // ...then partially drain the local queue and the global stack
3236 drain_local_queue(true);
3237 drain_global_stack(true);
3238
3239 do {
3240 if (!has_aborted() && _curr_region != NULL) {
3241 // This means that we're already holding on to a region.
3242 assert(_finger != NULL, "if region is not NULL, then the finger "
3243 "should not be NULL either");
3244
3245 // We might have restarted this task after an evacuation pause
3246 // which might have evacuated the region we're holding on to
3247 // underneath our feet. Let's read its limit again to make sure
3248 // that we do not iterate over a region of the heap that
3249 // contains garbage (update_region_limit() will also move
3250 // _finger to the start of the region if it is found empty).
3251 update_region_limit();
3252 // We will start from _finger not from the start of the region,
3253 // as we might be restarting this task after aborting half-way
3254 // through scanning this region. In this case, _finger points to
3255 // the address where we last found a marked object. If this is a
3256 // fresh region, _finger points to start().
3257 MemRegion mr = MemRegion(_finger, _region_limit);
3258
3259 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
3260 "humongous regions should go around loop once only");
3261
3262 // Some special cases:
3263 // If the memory region is empty, we can just give up the region.
3264 // If the current region is humongous then we only need to check
3265 // the bitmap for the bit associated with the start of the object,
3266 // scan the object if it's live, and give up the region.
3267 // Otherwise, let's iterate over the bitmap of the part of the region
3268 // that is left.
3269 // If the iteration is successful, give up the region.
3270 if (mr.is_empty()) {
3271 giveup_current_region();
3272 regular_clock_call();
3273 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
3274 if (_nextMarkBitMap->isMarked(mr.start())) {
3275 // The object is marked - apply the closure
3276 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start());
3277 bitmap_closure.do_bit(offset);
3278 }
3279 // Even if this task aborted while scanning the humongous object
3280 // we can (and should) give up the current region.
3281 giveup_current_region();
3282 regular_clock_call();
3283 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) {
3284 giveup_current_region();
3285 regular_clock_call();
3286 } else {
3287 assert(has_aborted(), "currently the only way to do so");
3288 // The only way to abort the bitmap iteration is to return
3289 // false from the do_bit() method. However, inside the
3290 // do_bit() method we move the _finger to point to the
3291 // object currently being looked at. So, if we bail out, we
3292 // have definitely set _finger to something non-null.
3293 assert(_finger != NULL, "invariant");
3294
3295 // Region iteration was actually aborted. So now _finger
3296 // points to the address of the object we last scanned. If we
3297 // leave it there, when we restart this task, we will rescan
3298 // the object. It is easy to avoid this. We move the finger by
3299 // enough to point to the next possible object header (the
3300 // bitmap knows by how much we need to move it as it knows its
3301 // granularity).
3302 assert(_finger < _region_limit, "invariant");
3303 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger);
3304 // Check if bitmap iteration was aborted while scanning the last object
3305 if (new_finger >= _region_limit) {
3306 giveup_current_region();
3307 } else {
3308 move_finger_to(new_finger);
3309 }
3310 }
3311 }
3312 // At this point we have either completed iterating over the
3313 // region we were holding on to, or we have aborted.
3314
3315 // We then partially drain the local queue and the global stack.
3316 // (Do we really need this?)
3317 drain_local_queue(true);
3318 drain_global_stack(true);
3319
3320 // Read the note on the claim_region() method on why it might
3321 // return NULL with potentially more regions available for
3322 // claiming and why we have to check out_of_regions() to determine
3323 // whether we're done or not.
3324 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
3325 // We are going to try to claim a new region. We should have
3326 // given up on the previous one.
3327 // Separated the asserts so that we know which one fires.
3328 assert(_curr_region == NULL, "invariant");
3329 assert(_finger == NULL, "invariant");
3330 assert(_region_limit == NULL, "invariant");
3331 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
3332 if (claimed_region != NULL) {
3333 // Yes, we managed to claim one
3334 setup_for_region(claimed_region);
3335 assert(_curr_region == claimed_region, "invariant");
3336 }
3337 // It is important to call the regular clock here. It might take
3338 // a while to claim a region if, for example, we hit a large
3339 // block of empty regions. So we need to call the regular clock
3340 // method once round the loop to make sure it's called
3341 // frequently enough.
3342 regular_clock_call();
3343 }
3344
3345 if (!has_aborted() && _curr_region == NULL) {
3346 assert(_cm->out_of_regions(),
3347 "at this point we should be out of regions");
3348 }
3349 } while ( _curr_region != NULL && !has_aborted());
3350
3351 if (!has_aborted()) {
3352 // We cannot check whether the global stack is empty, since other
3353 // tasks might be pushing objects to it concurrently.
3354 assert(_cm->out_of_regions(),
3355 "at this point we should be out of regions");
3356 // Try to reduce the number of available SATB buffers so that
3357 // remark has less work to do.
3358 drain_satb_buffers();
3359 }
3360
3361 // Since we've done everything else, we can now totally drain the
3362 // local queue and global stack.
3363 drain_local_queue(false);
3364 drain_global_stack(false);
3365
3366 // Attempt at work stealing from other task's queues.
3367 if (do_stealing && !has_aborted()) {
3368 // We have not aborted. This means that we have finished all that
3369 // we could. Let's try to do some stealing...
3370
3371 // We cannot check whether the global stack is empty, since other
3372 // tasks might be pushing objects to it concurrently.
3373 assert(_cm->out_of_regions() && _task_queue->size() == 0,
3374 "only way to reach here");
3375 while (!has_aborted()) {
3376 oop obj;
3377 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) {
3378 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
3379 "any stolen object should be marked");
3380 scan_object(obj);
3381
3382 // And since we're towards the end, let's totally drain the
3383 // local queue and global stack.
3384 drain_local_queue(false);
3385 drain_global_stack(false);
3386 } else {
3387 break;
3388 }
3389 }
3390 }
3391
3392 // We still haven't aborted. Now, let's try to get into the
3393 // termination protocol.
3394 if (do_termination && !has_aborted()) {
3395 // We cannot check whether the global stack is empty, since other
3396 // tasks might be concurrently pushing objects on it.
3397 // Separated the asserts so that we know which one fires.
3398 assert(_cm->out_of_regions(), "only way to reach here");
3399 assert(_task_queue->size() == 0, "only way to reach here");
3400 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
3401
3402 // The G1CMTask class also extends the TerminatorTerminator class,
3403 // hence its should_exit_termination() method will also decide
3404 // whether to exit the termination protocol or not.
3405 bool finished = (is_serial ||
3406 _cm->terminator()->offer_termination(this));
3407 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
3408 _termination_time_ms +=
3409 termination_end_time_ms - _termination_start_time_ms;
3410
3411 if (finished) {
3412 // We're all done.
3413
3414 if (_worker_id == 0) {
3415 // let's allow task 0 to do this
3416 if (concurrent()) {
3417 assert(_cm->concurrent_marking_in_progress(), "invariant");
3418 // we need to set this to false before the next
3419 // safepoint. This way we ensure that the marking phase
3420 // doesn't observe any more heap expansions.
3421 _cm->clear_concurrent_marking_in_progress();
3422 }
3423 }
3424
3425 // We can now guarantee that the global stack is empty, since
3426 // all other tasks have finished. We separated the guarantees so
3427 // that, if a condition is false, we can immediately find out
3428 // which one.
3429 guarantee(_cm->out_of_regions(), "only way to reach here");
3430 guarantee(_cm->mark_stack_empty(), "only way to reach here");
3431 guarantee(_task_queue->size() == 0, "only way to reach here");
3432 guarantee(!_cm->has_overflown(), "only way to reach here");
3433 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
3434 } else {
3435 // Apparently there's more work to do. Let's abort this task. It
3436 // will restart it and we can hopefully find more things to do.
3437 set_has_aborted();
3438 }
3439 }
3440
3441 // Mainly for debugging purposes to make sure that a pointer to the
3442 // closure which was statically allocated in this frame doesn't
3443 // escape it by accident.
3444 set_cm_oop_closure(NULL);
3445 double end_time_ms = os::elapsedVTime() * 1000.0;
3446 double elapsed_time_ms = end_time_ms - _start_time_ms;
3447 // Update the step history.
3448 _step_times_ms.add(elapsed_time_ms);
3449
3450 if (has_aborted()) {
3451 // The task was aborted for some reason.
3452 if (_has_timed_out) {
3453 double diff_ms = elapsed_time_ms - _time_target_ms;
3454 // Keep statistics of how well we did with respect to hitting
3455 // our target only if we actually timed out (if we aborted for
3456 // other reasons, then the results might get skewed).
3457 _marking_step_diffs_ms.add(diff_ms);
3458 }
3459
3460 if (_cm->has_overflown()) {
3461 // This is the interesting one. We aborted because a global
3462 // overflow was raised. This means we have to restart the
3463 // marking phase and start iterating over regions. However, in
3464 // order to do this we have to make sure that all tasks stop
3465 // what they are doing and re-initialize in a safe manner. We
3466 // will achieve this with the use of two barrier sync points.
3467
3468 if (!is_serial) {
3469 // We only need to enter the sync barrier if being called
3470 // from a parallel context
3471 _cm->enter_first_sync_barrier(_worker_id);
3472
3473 // When we exit this sync barrier we know that all tasks have
3474 // stopped doing marking work. So, it's now safe to
3475 // re-initialize our data structures. At the end of this method,
3476 // task 0 will clear the global data structures.
3477 }
3478
3479 // We clear the local state of this task...
3480 clear_region_fields();
3481
3482 if (!is_serial) {
3483 // ...and enter the second barrier.
3484 _cm->enter_second_sync_barrier(_worker_id);
3485 }
3486 // At this point, if we're during the concurrent phase of
3487 // marking, everything has been re-initialized and we're
3488 // ready to restart.
3489 }
3490 }
3491
3492 _claimed = false;
3493 }
3494
3495 G1CMTask::G1CMTask(uint worker_id,
3496 G1ConcurrentMark* cm,
3497 size_t* marked_bytes,
3498 BitMap* card_bm,
3499 G1CMTaskQueue* task_queue,
3500 G1CMTaskQueueSet* task_queues)
3501 : _g1h(G1CollectedHeap::heap()),
3502 _worker_id(worker_id), _cm(cm),
3503 _claimed(false),
3504 _nextMarkBitMap(NULL), _hash_seed(17),
3505 _task_queue(task_queue),
3506 _task_queues(task_queues),
3507 _cm_oop_closure(NULL),
3508 _marked_bytes_array(marked_bytes),
3509 _card_bm(card_bm) {
3510 guarantee(task_queue != NULL, "invariant");
3511 guarantee(task_queues != NULL, "invariant");
3512
3513 _marking_step_diffs_ms.add(0.5);
3514 }
3515
3516 // These are formatting macros that are used below to ensure
3517 // consistent formatting. The *_H_* versions are used to format the
3518 // header for a particular value and they should be kept consistent
3519 // with the corresponding macro. Also note that most of the macros add
3520 // the necessary white space (as a prefix) which makes them a bit
3521 // easier to compose.
3522
3523 // All the output lines are prefixed with this string to be able to
3524 // identify them easily in a large log file.
3525 #define G1PPRL_LINE_PREFIX "###"
3526
3527 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT
3528 #ifdef _LP64
3529 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
3530 #else // _LP64
3531 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
3532 #endif // _LP64
3533
3534 // For per-region info
3535 #define G1PPRL_TYPE_FORMAT " %-4s"
3536 #define G1PPRL_TYPE_H_FORMAT " %4s"
3537 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9)
3538 #define G1PPRL_BYTE_H_FORMAT " %9s"
3539 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
3540 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
3541
3542 // For summary info
3543 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT
3544 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT
3545 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB"
3546 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%"
3547
3548 G1PrintRegionLivenessInfoClosure::
3549 G1PrintRegionLivenessInfoClosure(const char* phase_name)
3550 : _total_used_bytes(0), _total_capacity_bytes(0),
3551 _total_prev_live_bytes(0), _total_next_live_bytes(0),
3552 _hum_used_bytes(0), _hum_capacity_bytes(0),
3553 _hum_prev_live_bytes(0), _hum_next_live_bytes(0),
3554 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) {
3555 G1CollectedHeap* g1h = G1CollectedHeap::heap();
3556 MemRegion g1_reserved = g1h->g1_reserved();
3557 double now = os::elapsedTime();
3558
3559 // Print the header of the output.
3560 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
3561 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" HEAP"
3562 G1PPRL_SUM_ADDR_FORMAT("reserved")
3563 G1PPRL_SUM_BYTE_FORMAT("region-size"),
3564 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
3565 HeapRegion::GrainBytes);
3566 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
3567 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3568 G1PPRL_TYPE_H_FORMAT
3569 G1PPRL_ADDR_BASE_H_FORMAT
3570 G1PPRL_BYTE_H_FORMAT
3571 G1PPRL_BYTE_H_FORMAT
3572 G1PPRL_BYTE_H_FORMAT
3573 G1PPRL_DOUBLE_H_FORMAT
3574 G1PPRL_BYTE_H_FORMAT
3575 G1PPRL_BYTE_H_FORMAT,
3576 "type", "address-range",
3577 "used", "prev-live", "next-live", "gc-eff",
3578 "remset", "code-roots");
3579 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3580 G1PPRL_TYPE_H_FORMAT
3581 G1PPRL_ADDR_BASE_H_FORMAT
3582 G1PPRL_BYTE_H_FORMAT
3583 G1PPRL_BYTE_H_FORMAT
3584 G1PPRL_BYTE_H_FORMAT
3585 G1PPRL_DOUBLE_H_FORMAT
3586 G1PPRL_BYTE_H_FORMAT
3587 G1PPRL_BYTE_H_FORMAT,
3588 "", "",
3589 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
3590 "(bytes)", "(bytes)");
3591 }
3592
3593 // It takes as a parameter a reference to one of the _hum_* fields, it
3594 // deduces the corresponding value for a region in a humongous region
3595 // series (either the region size, or what's left if the _hum_* field
3596 // is < the region size), and updates the _hum_* field accordingly.
3597 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
3598 size_t bytes = 0;
3599 // The > 0 check is to deal with the prev and next live bytes which
3600 // could be 0.
3601 if (*hum_bytes > 0) {
3602 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
3603 *hum_bytes -= bytes;
3604 }
3605 return bytes;
3606 }
3607
3608 // It deduces the values for a region in a humongous region series
3609 // from the _hum_* fields and updates those accordingly. It assumes
3610 // that that _hum_* fields have already been set up from the "starts
3611 // humongous" region and we visit the regions in address order.
3612 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
3613 size_t* capacity_bytes,
3614 size_t* prev_live_bytes,
3615 size_t* next_live_bytes) {
3616 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
3617 *used_bytes = get_hum_bytes(&_hum_used_bytes);
3618 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
3619 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
3620 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
3621 }
3622
3623 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
3624 const char* type = r->get_type_str();
3625 HeapWord* bottom = r->bottom();
3626 HeapWord* end = r->end();
3627 size_t capacity_bytes = r->capacity();
3628 size_t used_bytes = r->used();
3629 size_t prev_live_bytes = r->live_bytes();
3630 size_t next_live_bytes = r->next_live_bytes();
3631 double gc_eff = r->gc_efficiency();
3632 size_t remset_bytes = r->rem_set()->mem_size();
3633 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
3634
3635 if (r->is_starts_humongous()) {
3636 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
3637 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
3638 "they should have been zeroed after the last time we used them");
3639 // Set up the _hum_* fields.
3640 _hum_capacity_bytes = capacity_bytes;
3641 _hum_used_bytes = used_bytes;
3642 _hum_prev_live_bytes = prev_live_bytes;
3643 _hum_next_live_bytes = next_live_bytes;
3644 get_hum_bytes(&used_bytes, &capacity_bytes,
3645 &prev_live_bytes, &next_live_bytes);
3646 end = bottom + HeapRegion::GrainWords;
3647 } else if (r->is_continues_humongous()) {
3648 get_hum_bytes(&used_bytes, &capacity_bytes,
3649 &prev_live_bytes, &next_live_bytes);
3650 assert(end == bottom + HeapRegion::GrainWords, "invariant");
3651 }
3652
3653 _total_used_bytes += used_bytes;
3654 _total_capacity_bytes += capacity_bytes;
3655 _total_prev_live_bytes += prev_live_bytes;
3656 _total_next_live_bytes += next_live_bytes;
3657 _total_remset_bytes += remset_bytes;
3658 _total_strong_code_roots_bytes += strong_code_roots_bytes;
3659
3660 // Print a line for this particular region.
3661 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3662 G1PPRL_TYPE_FORMAT
3663 G1PPRL_ADDR_BASE_FORMAT
3664 G1PPRL_BYTE_FORMAT
3665 G1PPRL_BYTE_FORMAT
3666 G1PPRL_BYTE_FORMAT
3667 G1PPRL_DOUBLE_FORMAT
3668 G1PPRL_BYTE_FORMAT
3669 G1PPRL_BYTE_FORMAT,
3670 type, p2i(bottom), p2i(end),
3671 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
3672 remset_bytes, strong_code_roots_bytes);
3673
3674 return false;
3675 }
3676
3677 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
3678 // add static memory usages to remembered set sizes
3679 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
3680 // Print the footer of the output.
3681 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX);
3682 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX
3683 " SUMMARY"
3684 G1PPRL_SUM_MB_FORMAT("capacity")
3685 G1PPRL_SUM_MB_PERC_FORMAT("used")
3686 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
3687 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
3688 G1PPRL_SUM_MB_FORMAT("remset")
3689 G1PPRL_SUM_MB_FORMAT("code-roots"),
3690 bytes_to_mb(_total_capacity_bytes),
3691 bytes_to_mb(_total_used_bytes),
3692 perc(_total_used_bytes, _total_capacity_bytes),
3693 bytes_to_mb(_total_prev_live_bytes),
3694 perc(_total_prev_live_bytes, _total_capacity_bytes),
3695 bytes_to_mb(_total_next_live_bytes),
3696 perc(_total_next_live_bytes, _total_capacity_bytes),
3697 bytes_to_mb(_total_remset_bytes),
3698 bytes_to_mb(_total_strong_code_roots_bytes));
3699 }