1 /*
2 * Copyright (c) 2001, 2015, 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/concurrentMark.inline.hpp"
30 #include "gc/g1/concurrentMarkThread.inline.hpp"
31 #include "gc/g1/g1CollectedHeap.inline.hpp"
32 #include "gc/g1/g1CollectorPolicy.hpp"
33 #include "gc/g1/g1ErgoVerbose.hpp"
34 #include "gc/g1/g1Log.hpp"
35 #include "gc/g1/g1OopClosures.inline.hpp"
36 #include "gc/g1/g1RemSet.hpp"
37 #include "gc/g1/g1StringDedup.hpp"
38 #include "gc/g1/heapRegion.inline.hpp"
39 #include "gc/g1/heapRegionManager.inline.hpp"
40 #include "gc/g1/heapRegionRemSet.hpp"
41 #include "gc/g1/heapRegionSet.inline.hpp"
42 #include "gc/g1/suspendibleThreadSet.hpp"
43 #include "gc/shared/gcTimer.hpp"
44 #include "gc/shared/gcTrace.hpp"
45 #include "gc/shared/gcTraceTime.hpp"
46 #include "gc/shared/genOopClosures.inline.hpp"
47 #include "gc/shared/referencePolicy.hpp"
48 #include "gc/shared/strongRootsScope.hpp"
49 #include "gc/shared/taskqueue.inline.hpp"
50 #include "gc/shared/vmGCOperations.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 CMBitMapRO::CMBitMapRO(int shifter) :
63 _bm(),
64 _shifter(shifter) {
65 _bmStartWord = 0;
66 _bmWordSize = 0;
67 }
68
69 HeapWord* CMBitMapRO::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 if (limit == NULL) {
76 limit = _bmStartWord + _bmWordSize;
77 }
78 size_t limitOffset = heapWordToOffset(limit);
79 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
80 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
81 assert(nextAddr >= addr, "get_next_one postcondition");
82 assert(nextAddr == limit || isMarked(nextAddr),
83 "get_next_one postcondition");
84 return nextAddr;
85 }
86
87 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(const HeapWord* addr,
88 const HeapWord* limit) const {
89 size_t addrOffset = heapWordToOffset(addr);
90 if (limit == NULL) {
91 limit = _bmStartWord + _bmWordSize;
92 }
93 size_t limitOffset = heapWordToOffset(limit);
94 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset);
95 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
96 assert(nextAddr >= addr, "get_next_one postcondition");
97 assert(nextAddr == limit || !isMarked(nextAddr),
98 "get_next_one postcondition");
99 return nextAddr;
100 }
101
102 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const {
103 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check");
104 return (int) (diff >> _shifter);
105 }
106
107 #ifndef PRODUCT
108 bool CMBitMapRO::covers(MemRegion heap_rs) const {
109 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
110 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
111 "size inconsistency");
112 return _bmStartWord == (HeapWord*)(heap_rs.start()) &&
113 _bmWordSize == heap_rs.word_size();
114 }
115 #endif
116
117 void CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const {
118 _bm.print_on_error(st, prefix);
119 }
120
121 size_t CMBitMap::compute_size(size_t heap_size) {
122 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance());
123 }
124
125 size_t CMBitMap::mark_distance() {
126 return MinObjAlignmentInBytes * BitsPerByte;
127 }
128
129 void CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) {
130 _bmStartWord = heap.start();
131 _bmWordSize = heap.word_size();
132
133 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start());
134 _bm.set_size(_bmWordSize >> _shifter);
135
136 storage->set_mapping_changed_listener(&_listener);
137 }
138
139 void CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) {
140 if (zero_filled) {
141 return;
142 }
143 // We need to clear the bitmap on commit, removing any existing information.
144 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords);
145 _bm->clearRange(mr);
146 }
147
148 // Closure used for clearing the given mark bitmap.
149 class ClearBitmapHRClosure : public HeapRegionClosure {
150 private:
151 ConcurrentMark* _cm;
152 CMBitMap* _bitmap;
153 bool _may_yield; // The closure may yield during iteration. If yielded, abort the iteration.
154 public:
155 ClearBitmapHRClosure(ConcurrentMark* cm, CMBitMap* bitmap, bool may_yield) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap), _may_yield(may_yield) {
156 assert(!may_yield || cm != NULL, "CM must be non-NULL if this closure is expected to yield.");
157 }
158
159 virtual bool doHeapRegion(HeapRegion* r) {
160 size_t const chunk_size_in_words = M / HeapWordSize;
161
162 HeapWord* cur = r->bottom();
163 HeapWord* const end = r->end();
164
165 while (cur < end) {
166 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end));
167 _bitmap->clearRange(mr);
168
169 cur += chunk_size_in_words;
170
171 // Abort iteration if after yielding the marking has been aborted.
172 if (_may_yield && _cm->do_yield_check() && _cm->has_aborted()) {
173 return true;
174 }
175 // Repeat the asserts from before the start of the closure. We will do them
176 // as asserts here to minimize their overhead on the product. However, we
177 // will have them as guarantees at the beginning / end of the bitmap
178 // clearing to get some checking in the product.
179 assert(!_may_yield || _cm->cmThread()->during_cycle(), "invariant");
180 assert(!_may_yield || !G1CollectedHeap::heap()->mark_in_progress(), "invariant");
181 }
182
183 return false;
184 }
185 };
186
187 class ParClearNextMarkBitmapTask : public AbstractGangTask {
188 ClearBitmapHRClosure* _cl;
189 HeapRegionClaimer _hrclaimer;
190 bool _suspendible; // If the task is suspendible, workers must join the STS.
191
192 public:
193 ParClearNextMarkBitmapTask(ClearBitmapHRClosure *cl, uint n_workers, bool suspendible) :
194 _cl(cl), _suspendible(suspendible), AbstractGangTask("Parallel Clear Bitmap Task"), _hrclaimer(n_workers) {}
195
196 void work(uint worker_id) {
197 SuspendibleThreadSetJoiner sts_join(_suspendible);
198 G1CollectedHeap::heap()->heap_region_par_iterate(_cl, worker_id, &_hrclaimer, true);
199 }
200 };
201
202 void CMBitMap::clearAll() {
203 G1CollectedHeap* g1h = G1CollectedHeap::heap();
204 ClearBitmapHRClosure cl(NULL, this, false /* may_yield */);
205 uint n_workers = g1h->workers()->active_workers();
206 ParClearNextMarkBitmapTask task(&cl, n_workers, false);
207 g1h->workers()->run_task(&task);
208 guarantee(cl.complete(), "Must have completed iteration.");
209 return;
210 }
211
212 void CMBitMap::markRange(MemRegion mr) {
213 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
214 assert(!mr.is_empty(), "unexpected empty region");
215 assert((offsetToHeapWord(heapWordToOffset(mr.end())) ==
216 ((HeapWord *) mr.end())),
217 "markRange memory region end is not card aligned");
218 // convert address range into offset range
219 _bm.at_put_range(heapWordToOffset(mr.start()),
220 heapWordToOffset(mr.end()), true);
221 }
222
223 void CMBitMap::clearRange(MemRegion mr) {
224 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
225 assert(!mr.is_empty(), "unexpected empty region");
226 // convert address range into offset range
227 _bm.at_put_range(heapWordToOffset(mr.start()),
228 heapWordToOffset(mr.end()), false);
229 }
230
231 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr,
232 HeapWord* end_addr) {
233 HeapWord* start = getNextMarkedWordAddress(addr);
234 start = MIN2(start, end_addr);
235 HeapWord* end = getNextUnmarkedWordAddress(start);
236 end = MIN2(end, end_addr);
237 assert(start <= end, "Consistency check");
238 MemRegion mr(start, end);
239 if (!mr.is_empty()) {
240 clearRange(mr);
241 }
242 return mr;
243 }
244
245 CMMarkStack::CMMarkStack(ConcurrentMark* cm) :
246 _base(NULL), _cm(cm)
247 #ifdef ASSERT
248 , _drain_in_progress(false)
249 , _drain_in_progress_yields(false)
250 #endif
251 {}
252
253 bool CMMarkStack::allocate(size_t capacity) {
254 // allocate a stack of the requisite depth
255 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop)));
256 if (!rs.is_reserved()) {
257 warning("ConcurrentMark MarkStack allocation failure");
258 return false;
259 }
260 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC);
261 if (!_virtual_space.initialize(rs, rs.size())) {
262 warning("ConcurrentMark MarkStack backing store failure");
263 // Release the virtual memory reserved for the marking stack
264 rs.release();
265 return false;
266 }
267 assert(_virtual_space.committed_size() == rs.size(),
268 "Didn't reserve backing store for all of ConcurrentMark stack?");
269 _base = (oop*) _virtual_space.low();
270 setEmpty();
271 _capacity = (jint) capacity;
272 _saved_index = -1;
273 _should_expand = false;
274 return true;
275 }
276
277 void CMMarkStack::expand() {
278 // Called, during remark, if we've overflown the marking stack during marking.
279 assert(isEmpty(), "stack should been emptied while handling overflow");
280 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted");
281 // Clear expansion flag
282 _should_expand = false;
283 if (_capacity == (jint) MarkStackSizeMax) {
284 if (PrintGCDetails && Verbose) {
285 gclog_or_tty->print_cr(" (benign) Can't expand marking stack capacity, at max size limit");
286 }
287 return;
288 }
289 // Double capacity if possible
290 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax);
291 // Do not give up existing stack until we have managed to
292 // get the double capacity that we desired.
293 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity *
294 sizeof(oop)));
295 if (rs.is_reserved()) {
296 // Release the backing store associated with old stack
297 _virtual_space.release();
298 // Reinitialize virtual space for new stack
299 if (!_virtual_space.initialize(rs, rs.size())) {
300 fatal("Not enough swap for expanded marking stack capacity");
301 }
302 _base = (oop*)(_virtual_space.low());
303 _index = 0;
304 _capacity = new_capacity;
305 } else {
306 if (PrintGCDetails && Verbose) {
307 // Failed to double capacity, continue;
308 gclog_or_tty->print(" (benign) Failed to expand marking stack capacity from "
309 SIZE_FORMAT"K to " SIZE_FORMAT"K",
310 _capacity / K, new_capacity / K);
311 }
312 }
313 }
314
315 void CMMarkStack::set_should_expand() {
316 // If we're resetting the marking state because of an
317 // marking stack overflow, record that we should, if
318 // possible, expand the stack.
319 _should_expand = _cm->has_overflown();
320 }
321
322 CMMarkStack::~CMMarkStack() {
323 if (_base != NULL) {
324 _base = NULL;
325 _virtual_space.release();
326 }
327 }
328
329 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
330 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
331 jint start = _index;
332 jint next_index = start + n;
333 if (next_index > _capacity) {
334 _overflow = true;
335 return;
336 }
337 // Otherwise.
338 _index = next_index;
339 for (int i = 0; i < n; i++) {
340 int ind = start + i;
341 assert(ind < _capacity, "By overflow test above.");
342 _base[ind] = ptr_arr[i];
343 }
344 }
345
346 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
347 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
348 jint index = _index;
349 if (index == 0) {
350 *n = 0;
351 return false;
352 } else {
353 int k = MIN2(max, index);
354 jint new_ind = index - k;
355 for (int j = 0; j < k; j++) {
356 ptr_arr[j] = _base[new_ind + j];
357 }
358 _index = new_ind;
359 *n = k;
360 return true;
361 }
362 }
363
364 template<class OopClosureClass>
365 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) {
366 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after
367 || SafepointSynchronize::is_at_safepoint(),
368 "Drain recursion must be yield-safe.");
369 bool res = true;
370 debug_only(_drain_in_progress = true);
371 debug_only(_drain_in_progress_yields = yield_after);
372 while (!isEmpty()) {
373 oop newOop = pop();
374 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop");
375 assert(newOop->is_oop(), "Expected an oop");
376 assert(bm == NULL || bm->isMarked((HeapWord*)newOop),
377 "only grey objects on this stack");
378 newOop->oop_iterate(cl);
379 if (yield_after && _cm->do_yield_check()) {
380 res = false;
381 break;
382 }
383 }
384 debug_only(_drain_in_progress = false);
385 return res;
386 }
387
388 void CMMarkStack::note_start_of_gc() {
389 assert(_saved_index == -1,
390 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
391 _saved_index = _index;
392 }
393
394 void CMMarkStack::note_end_of_gc() {
395 // This is intentionally a guarantee, instead of an assert. If we
396 // accidentally add something to the mark stack during GC, it
397 // will be a correctness issue so it's better if we crash. we'll
398 // only check this once per GC anyway, so it won't be a performance
399 // issue in any way.
400 guarantee(_saved_index == _index,
401 err_msg("saved index: %d index: %d", _saved_index, _index));
402 _saved_index = -1;
403 }
404
405 void CMMarkStack::oops_do(OopClosure* f) {
406 assert(_saved_index == _index,
407 err_msg("saved index: %d index: %d", _saved_index, _index));
408 for (int i = 0; i < _index; i += 1) {
409 f->do_oop(&_base[i]);
410 }
411 }
412
413 CMRootRegions::CMRootRegions() :
414 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
415 _should_abort(false), _next_survivor(NULL) { }
416
417 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) {
418 _young_list = g1h->young_list();
419 _cm = cm;
420 }
421
422 void CMRootRegions::prepare_for_scan() {
423 assert(!scan_in_progress(), "pre-condition");
424
425 // Currently, only survivors can be root regions.
426 assert(_next_survivor == NULL, "pre-condition");
427 _next_survivor = _young_list->first_survivor_region();
428 _scan_in_progress = (_next_survivor != NULL);
429 _should_abort = false;
430 }
431
432 HeapRegion* CMRootRegions::claim_next() {
433 if (_should_abort) {
434 // If someone has set the should_abort flag, we return NULL to
435 // force the caller to bail out of their loop.
436 return NULL;
437 }
438
439 // Currently, only survivors can be root regions.
440 HeapRegion* res = _next_survivor;
441 if (res != NULL) {
442 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
443 // Read it again in case it changed while we were waiting for the lock.
444 res = _next_survivor;
445 if (res != NULL) {
446 if (res == _young_list->last_survivor_region()) {
447 // We just claimed the last survivor so store NULL to indicate
448 // that we're done.
449 _next_survivor = NULL;
450 } else {
451 _next_survivor = res->get_next_young_region();
452 }
453 } else {
454 // Someone else claimed the last survivor while we were trying
455 // to take the lock so nothing else to do.
456 }
457 }
458 assert(res == NULL || res->is_survivor(), "post-condition");
459
460 return res;
461 }
462
463 void CMRootRegions::scan_finished() {
464 assert(scan_in_progress(), "pre-condition");
465
466 // Currently, only survivors can be root regions.
467 if (!_should_abort) {
468 assert(_next_survivor == NULL, "we should have claimed all survivors");
469 }
470 _next_survivor = NULL;
471
472 {
473 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
474 _scan_in_progress = false;
475 RootRegionScan_lock->notify_all();
476 }
477 }
478
479 bool CMRootRegions::wait_until_scan_finished() {
480 if (!scan_in_progress()) return false;
481
482 {
483 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
484 while (scan_in_progress()) {
485 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
486 }
487 }
488 return true;
489 }
490
491 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away
492 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
493 #endif // _MSC_VER
494
495 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
496 return MAX2((n_par_threads + 2) / 4, 1U);
497 }
498
499 ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) :
500 _g1h(g1h),
501 _markBitMap1(),
502 _markBitMap2(),
503 _parallel_marking_threads(0),
504 _max_parallel_marking_threads(0),
505 _sleep_factor(0.0),
506 _marking_task_overhead(1.0),
507 _cleanup_sleep_factor(0.0),
508 _cleanup_task_overhead(1.0),
509 _cleanup_list("Cleanup List"),
510 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/),
511 _card_bm((g1h->reserved_region().byte_size() + CardTableModRefBS::card_size - 1) >>
512 CardTableModRefBS::card_shift,
513 false /* in_resource_area*/),
514
515 _prevMarkBitMap(&_markBitMap1),
516 _nextMarkBitMap(&_markBitMap2),
517
518 _markStack(this),
519 // _finger set in set_non_marking_state
520
521 _max_worker_id((uint)ParallelGCThreads),
522 // _active_tasks set in set_non_marking_state
523 // _tasks set inside the constructor
524 _task_queues(new CMTaskQueueSet((int) _max_worker_id)),
525 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)),
526
527 _has_overflown(false),
528 _concurrent(false),
529 _has_aborted(false),
530 _aborted_gc_id(GCId::undefined()),
531 _restart_for_overflow(false),
532 _concurrent_marking_in_progress(false),
533
534 // _verbose_level set below
535
536 _init_times(),
537 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
538 _cleanup_times(),
539 _total_counting_time(0.0),
540 _total_rs_scrub_time(0.0),
541
542 _parallel_workers(NULL),
543
544 _count_card_bitmaps(NULL),
545 _count_marked_bytes(NULL),
546 _completed_initialization(false) {
547 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel;
548 if (verbose_level < no_verbose) {
549 verbose_level = no_verbose;
550 }
551 if (verbose_level > high_verbose) {
552 verbose_level = high_verbose;
553 }
554 _verbose_level = verbose_level;
555
556 if (verbose_low()) {
557 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", "
558 "heap end = " PTR_FORMAT, p2i(_heap_start), p2i(_heap_end));
559 }
560
561 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage);
562 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage);
563
564 // Create & start a ConcurrentMark thread.
565 _cmThread = new ConcurrentMarkThread(this);
566 assert(cmThread() != NULL, "CM Thread should have been created");
567 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
568 if (_cmThread->osthread() == NULL) {
569 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
570 }
571
572 assert(CGC_lock != NULL, "Where's the CGC_lock?");
573 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency");
574 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency");
575
576 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
577 satb_qs.set_buffer_size(G1SATBBufferSize);
578
579 _root_regions.init(_g1h, this);
580
581 if (ConcGCThreads > ParallelGCThreads) {
582 warning("Can't have more ConcGCThreads (" UINTX_FORMAT ") "
583 "than ParallelGCThreads (" UINTX_FORMAT ").",
584 ConcGCThreads, ParallelGCThreads);
585 return;
586 }
587 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
588 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
589 // if both are set
590 _sleep_factor = 0.0;
591 _marking_task_overhead = 1.0;
592 } else if (G1MarkingOverheadPercent > 0) {
593 // We will calculate the number of parallel marking threads based
594 // on a target overhead with respect to the soft real-time goal
595 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
596 double overall_cm_overhead =
597 (double) MaxGCPauseMillis * marking_overhead /
598 (double) GCPauseIntervalMillis;
599 double cpu_ratio = 1.0 / (double) os::processor_count();
600 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
601 double marking_task_overhead =
602 overall_cm_overhead / marking_thread_num *
603 (double) os::processor_count();
604 double sleep_factor =
605 (1.0 - marking_task_overhead) / marking_task_overhead;
606
607 FLAG_SET_ERGO(uintx, ConcGCThreads, (uint) marking_thread_num);
608 _sleep_factor = sleep_factor;
609 _marking_task_overhead = marking_task_overhead;
610 } else {
611 // Calculate the number of parallel marking threads by scaling
612 // the number of parallel GC threads.
613 uint marking_thread_num = scale_parallel_threads((uint) ParallelGCThreads);
614 FLAG_SET_ERGO(uintx, ConcGCThreads, marking_thread_num);
615 _sleep_factor = 0.0;
616 _marking_task_overhead = 1.0;
617 }
618
619 assert(ConcGCThreads > 0, "Should have been set");
620 _parallel_marking_threads = (uint) ConcGCThreads;
621 _max_parallel_marking_threads = _parallel_marking_threads;
622
623 if (parallel_marking_threads() > 1) {
624 _cleanup_task_overhead = 1.0;
625 } else {
626 _cleanup_task_overhead = marking_task_overhead();
627 }
628 _cleanup_sleep_factor =
629 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead();
630
631 #if 0
632 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads());
633 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead());
634 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor());
635 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead());
636 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor());
637 #endif
638
639 _parallel_workers = new FlexibleWorkGang("G1 Marker",
640 _max_parallel_marking_threads, false, true);
641 if (_parallel_workers == NULL) {
642 vm_exit_during_initialization("Failed necessary allocation.");
643 } else {
644 _parallel_workers->initialize_workers();
645 }
646
647 if (FLAG_IS_DEFAULT(MarkStackSize)) {
648 size_t mark_stack_size =
649 MIN2(MarkStackSizeMax,
650 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE)));
651 // Verify that the calculated value for MarkStackSize is in range.
652 // It would be nice to use the private utility routine from Arguments.
653 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
654 warning("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): "
655 "must be between 1 and " SIZE_FORMAT,
656 mark_stack_size, MarkStackSizeMax);
657 return;
658 }
659 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size);
660 } else {
661 // Verify MarkStackSize is in range.
662 if (FLAG_IS_CMDLINE(MarkStackSize)) {
663 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
664 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
665 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): "
666 "must be between 1 and " SIZE_FORMAT,
667 MarkStackSize, MarkStackSizeMax);
668 return;
669 }
670 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
671 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
672 warning("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")"
673 " or for MarkStackSizeMax (" SIZE_FORMAT ")",
674 MarkStackSize, MarkStackSizeMax);
675 return;
676 }
677 }
678 }
679 }
680
681 if (!_markStack.allocate(MarkStackSize)) {
682 warning("Failed to allocate CM marking stack");
683 return;
684 }
685
686 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_worker_id, mtGC);
687 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC);
688
689 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC);
690 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC);
691
692 BitMap::idx_t card_bm_size = _card_bm.size();
693
694 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
695 _active_tasks = _max_worker_id;
696
697 uint max_regions = _g1h->max_regions();
698 for (uint i = 0; i < _max_worker_id; ++i) {
699 CMTaskQueue* task_queue = new CMTaskQueue();
700 task_queue->initialize();
701 _task_queues->register_queue(i, task_queue);
702
703 _count_card_bitmaps[i] = BitMap(card_bm_size, false);
704 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC);
705
706 _tasks[i] = new CMTask(i, this,
707 _count_marked_bytes[i],
708 &_count_card_bitmaps[i],
709 task_queue, _task_queues);
710
711 _accum_task_vtime[i] = 0.0;
712 }
713
714 // Calculate the card number for the bottom of the heap. Used
715 // in biasing indexes into the accounting card bitmaps.
716 _heap_bottom_card_num =
717 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
718 CardTableModRefBS::card_shift);
719
720 // Clear all the liveness counting data
721 clear_all_count_data();
722
723 // so that the call below can read a sensible value
724 _heap_start = g1h->reserved_region().start();
725 set_non_marking_state();
726 _completed_initialization = true;
727 }
728
729 void ConcurrentMark::reset() {
730 // Starting values for these two. This should be called in a STW
731 // phase.
732 MemRegion reserved = _g1h->g1_reserved();
733 _heap_start = reserved.start();
734 _heap_end = reserved.end();
735
736 // Separated the asserts so that we know which one fires.
737 assert(_heap_start != NULL, "heap bounds should look ok");
738 assert(_heap_end != NULL, "heap bounds should look ok");
739 assert(_heap_start < _heap_end, "heap bounds should look ok");
740
741 // Reset all the marking data structures and any necessary flags
742 reset_marking_state();
743
744 if (verbose_low()) {
745 gclog_or_tty->print_cr("[global] resetting");
746 }
747
748 // We do reset all of them, since different phases will use
749 // different number of active threads. So, it's easiest to have all
750 // of them ready.
751 for (uint i = 0; i < _max_worker_id; ++i) {
752 _tasks[i]->reset(_nextMarkBitMap);
753 }
754
755 // we need this to make sure that the flag is on during the evac
756 // pause with initial mark piggy-backed
757 set_concurrent_marking_in_progress();
758 }
759
760
761 void ConcurrentMark::reset_marking_state(bool clear_overflow) {
762 _markStack.set_should_expand();
763 _markStack.setEmpty(); // Also clears the _markStack overflow flag
764 if (clear_overflow) {
765 clear_has_overflown();
766 } else {
767 assert(has_overflown(), "pre-condition");
768 }
769 _finger = _heap_start;
770
771 for (uint i = 0; i < _max_worker_id; ++i) {
772 CMTaskQueue* queue = _task_queues->queue(i);
773 queue->set_empty();
774 }
775 }
776
777 void ConcurrentMark::set_concurrency(uint active_tasks) {
778 assert(active_tasks <= _max_worker_id, "we should not have more");
779
780 _active_tasks = active_tasks;
781 // Need to update the three data structures below according to the
782 // number of active threads for this phase.
783 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
784 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
785 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
786 }
787
788 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
789 set_concurrency(active_tasks);
790
791 _concurrent = concurrent;
792 // We propagate this to all tasks, not just the active ones.
793 for (uint i = 0; i < _max_worker_id; ++i)
794 _tasks[i]->set_concurrent(concurrent);
795
796 if (concurrent) {
797 set_concurrent_marking_in_progress();
798 } else {
799 // We currently assume that the concurrent flag has been set to
800 // false before we start remark. At this point we should also be
801 // in a STW phase.
802 assert(!concurrent_marking_in_progress(), "invariant");
803 assert(out_of_regions(),
804 err_msg("only way to get here: _finger: "PTR_FORMAT", _heap_end: "PTR_FORMAT,
805 p2i(_finger), p2i(_heap_end)));
806 }
807 }
808
809 void ConcurrentMark::set_non_marking_state() {
810 // We set the global marking state to some default values when we're
811 // not doing marking.
812 reset_marking_state();
813 _active_tasks = 0;
814 clear_concurrent_marking_in_progress();
815 }
816
817 ConcurrentMark::~ConcurrentMark() {
818 // The ConcurrentMark instance is never freed.
819 ShouldNotReachHere();
820 }
821
822 void ConcurrentMark::clearNextBitmap() {
823 G1CollectedHeap* g1h = G1CollectedHeap::heap();
824
825 // Make sure that the concurrent mark thread looks to still be in
826 // the current cycle.
827 guarantee(cmThread()->during_cycle(), "invariant");
828
829 // We are finishing up the current cycle by clearing the next
830 // marking bitmap and getting it ready for the next cycle. During
831 // this time no other cycle can start. So, let's make sure that this
832 // is the case.
833 guarantee(!g1h->mark_in_progress(), "invariant");
834
835 ClearBitmapHRClosure cl(this, _nextMarkBitMap, true /* may_yield */);
836 ParClearNextMarkBitmapTask task(&cl, parallel_marking_threads(), true);
837 _parallel_workers->run_task(&task);
838
839 // Clear the liveness counting data. If the marking has been aborted, the abort()
840 // call already did that.
841 if (cl.complete()) {
842 clear_all_count_data();
843 }
844
845 // Repeat the asserts from above.
846 guarantee(cmThread()->during_cycle(), "invariant");
847 guarantee(!g1h->mark_in_progress(), "invariant");
848 }
849
850 class CheckBitmapClearHRClosure : public HeapRegionClosure {
851 CMBitMap* _bitmap;
852 bool _error;
853 public:
854 CheckBitmapClearHRClosure(CMBitMap* bitmap) : _bitmap(bitmap) {
855 }
856
857 virtual bool doHeapRegion(HeapRegion* r) {
858 // This closure can be called concurrently to the mutator, so we must make sure
859 // that the result of the getNextMarkedWordAddress() call is compared to the
860 // value passed to it as limit to detect any found bits.
861 // We can use the region's orig_end() for the limit and the comparison value
862 // as it always contains the "real" end of the region that never changes and
863 // has no side effects.
864 // Due to the latter, there can also be no problem with the compiler generating
865 // reloads of the orig_end() call.
866 HeapWord* end = r->orig_end();
867 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end;
868 }
869 };
870
871 bool ConcurrentMark::nextMarkBitmapIsClear() {
872 CheckBitmapClearHRClosure cl(_nextMarkBitMap);
873 _g1h->heap_region_iterate(&cl);
874 return cl.complete();
875 }
876
877 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
878 public:
879 bool doHeapRegion(HeapRegion* r) {
880 if (!r->is_continues_humongous()) {
881 r->note_start_of_marking();
882 }
883 return false;
884 }
885 };
886
887 void ConcurrentMark::checkpointRootsInitialPre() {
888 G1CollectedHeap* g1h = G1CollectedHeap::heap();
889 G1CollectorPolicy* g1p = g1h->g1_policy();
890
891 _has_aborted = false;
892
893 // Initialize marking structures. This has to be done in a STW phase.
894 reset();
895
896 // For each region note start of marking.
897 NoteStartOfMarkHRClosure startcl;
898 g1h->heap_region_iterate(&startcl);
899 }
900
901
902 void ConcurrentMark::checkpointRootsInitialPost() {
903 G1CollectedHeap* g1h = G1CollectedHeap::heap();
904
905 // If we force an overflow during remark, the remark operation will
906 // actually abort and we'll restart concurrent marking. If we always
907 // force an overflow during remark we'll never actually complete the
908 // marking phase. So, we initialize this here, at the start of the
909 // cycle, so that at the remaining overflow number will decrease at
910 // every remark and we'll eventually not need to cause one.
911 force_overflow_stw()->init();
912
913 // Start Concurrent Marking weak-reference discovery.
914 ReferenceProcessor* rp = g1h->ref_processor_cm();
915 // enable ("weak") refs discovery
916 rp->enable_discovery();
917 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
918
919 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
920 // This is the start of the marking cycle, we're expected all
921 // threads to have SATB queues with active set to false.
922 satb_mq_set.set_active_all_threads(true, /* new active value */
923 false /* expected_active */);
924
925 _root_regions.prepare_for_scan();
926
927 // update_g1_committed() will be called at the end of an evac pause
928 // when marking is on. So, it's also called at the end of the
929 // initial-mark pause to update the heap end, if the heap expands
930 // during it. No need to call it here.
931 }
932
933 /*
934 * Notice that in the next two methods, we actually leave the STS
935 * during the barrier sync and join it immediately afterwards. If we
936 * do not do this, the following deadlock can occur: one thread could
937 * be in the barrier sync code, waiting for the other thread to also
938 * sync up, whereas another one could be trying to yield, while also
939 * waiting for the other threads to sync up too.
940 *
941 * Note, however, that this code is also used during remark and in
942 * this case we should not attempt to leave / enter the STS, otherwise
943 * we'll either hit an assert (debug / fastdebug) or deadlock
944 * (product). So we should only leave / enter the STS if we are
945 * operating concurrently.
946 *
947 * Because the thread that does the sync barrier has left the STS, it
948 * is possible to be suspended for a Full GC or an evacuation pause
949 * could occur. This is actually safe, since the entering the sync
950 * barrier is one of the last things do_marking_step() does, and it
951 * doesn't manipulate any data structures afterwards.
952 */
953
954 void ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
955 bool barrier_aborted;
956
957 if (verbose_low()) {
958 gclog_or_tty->print_cr("[%u] entering first barrier", worker_id);
959 }
960
961 {
962 SuspendibleThreadSetLeaver sts_leave(concurrent());
963 barrier_aborted = !_first_overflow_barrier_sync.enter();
964 }
965
966 // at this point everyone should have synced up and not be doing any
967 // more work
968
969 if (verbose_low()) {
970 if (barrier_aborted) {
971 gclog_or_tty->print_cr("[%u] aborted first barrier", worker_id);
972 } else {
973 gclog_or_tty->print_cr("[%u] leaving first barrier", worker_id);
974 }
975 }
976
977 if (barrier_aborted) {
978 // If the barrier aborted we ignore the overflow condition and
979 // just abort the whole marking phase as quickly as possible.
980 return;
981 }
982
983 // If we're executing the concurrent phase of marking, reset the marking
984 // state; otherwise the marking state is reset after reference processing,
985 // during the remark pause.
986 // If we reset here as a result of an overflow during the remark we will
987 // see assertion failures from any subsequent set_concurrency_and_phase()
988 // calls.
989 if (concurrent()) {
990 // let the task associated with with worker 0 do this
991 if (worker_id == 0) {
992 // task 0 is responsible for clearing the global data structures
993 // We should be here because of an overflow. During STW we should
994 // not clear the overflow flag since we rely on it being true when
995 // we exit this method to abort the pause and restart concurrent
996 // marking.
997 reset_marking_state(true /* clear_overflow */);
998 force_overflow()->update();
999
1000 if (G1Log::fine()) {
1001 gclog_or_tty->gclog_stamp(concurrent_gc_id());
1002 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]");
1003 }
1004 }
1005 }
1006
1007 // after this, each task should reset its own data structures then
1008 // then go into the second barrier
1009 }
1010
1011 void ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
1012 bool barrier_aborted;
1013
1014 if (verbose_low()) {
1015 gclog_or_tty->print_cr("[%u] entering second barrier", worker_id);
1016 }
1017
1018 {
1019 SuspendibleThreadSetLeaver sts_leave(concurrent());
1020 barrier_aborted = !_second_overflow_barrier_sync.enter();
1021 }
1022
1023 // at this point everything should be re-initialized and ready to go
1024
1025 if (verbose_low()) {
1026 if (barrier_aborted) {
1027 gclog_or_tty->print_cr("[%u] aborted second barrier", worker_id);
1028 } else {
1029 gclog_or_tty->print_cr("[%u] leaving second barrier", worker_id);
1030 }
1031 }
1032 }
1033
1034 #ifndef PRODUCT
1035 void ForceOverflowSettings::init() {
1036 _num_remaining = G1ConcMarkForceOverflow;
1037 _force = false;
1038 update();
1039 }
1040
1041 void ForceOverflowSettings::update() {
1042 if (_num_remaining > 0) {
1043 _num_remaining -= 1;
1044 _force = true;
1045 } else {
1046 _force = false;
1047 }
1048 }
1049
1050 bool ForceOverflowSettings::should_force() {
1051 if (_force) {
1052 _force = false;
1053 return true;
1054 } else {
1055 return false;
1056 }
1057 }
1058 #endif // !PRODUCT
1059
1060 class CMConcurrentMarkingTask: public AbstractGangTask {
1061 private:
1062 ConcurrentMark* _cm;
1063 ConcurrentMarkThread* _cmt;
1064
1065 public:
1066 void work(uint worker_id) {
1067 assert(Thread::current()->is_ConcurrentGC_thread(),
1068 "this should only be done by a conc GC thread");
1069 ResourceMark rm;
1070
1071 double start_vtime = os::elapsedVTime();
1072
1073 {
1074 SuspendibleThreadSetJoiner sts_join;
1075
1076 assert(worker_id < _cm->active_tasks(), "invariant");
1077 CMTask* the_task = _cm->task(worker_id);
1078 the_task->record_start_time();
1079 if (!_cm->has_aborted()) {
1080 do {
1081 double start_vtime_sec = os::elapsedVTime();
1082 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1083
1084 the_task->do_marking_step(mark_step_duration_ms,
1085 true /* do_termination */,
1086 false /* is_serial*/);
1087
1088 double end_vtime_sec = os::elapsedVTime();
1089 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
1090 _cm->clear_has_overflown();
1091
1092 _cm->do_yield_check(worker_id);
1093
1094 jlong sleep_time_ms;
1095 if (!_cm->has_aborted() && the_task->has_aborted()) {
1096 sleep_time_ms =
1097 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
1098 {
1099 SuspendibleThreadSetLeaver sts_leave;
1100 os::sleep(Thread::current(), sleep_time_ms, false);
1101 }
1102 }
1103 } while (!_cm->has_aborted() && the_task->has_aborted());
1104 }
1105 the_task->record_end_time();
1106 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
1107 }
1108
1109 double end_vtime = os::elapsedVTime();
1110 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
1111 }
1112
1113 CMConcurrentMarkingTask(ConcurrentMark* cm,
1114 ConcurrentMarkThread* cmt) :
1115 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
1116
1117 ~CMConcurrentMarkingTask() { }
1118 };
1119
1120 // Calculates the number of active workers for a concurrent
1121 // phase.
1122 uint ConcurrentMark::calc_parallel_marking_threads() {
1123 uint n_conc_workers = 0;
1124 if (!UseDynamicNumberOfGCThreads ||
1125 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
1126 !ForceDynamicNumberOfGCThreads)) {
1127 n_conc_workers = max_parallel_marking_threads();
1128 } else {
1129 n_conc_workers =
1130 AdaptiveSizePolicy::calc_default_active_workers(
1131 max_parallel_marking_threads(),
1132 1, /* Minimum workers */
1133 parallel_marking_threads(),
1134 Threads::number_of_non_daemon_threads());
1135 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
1136 // that scaling has already gone into "_max_parallel_marking_threads".
1137 }
1138 assert(n_conc_workers > 0, "Always need at least 1");
1139 return n_conc_workers;
1140 }
1141
1142 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
1143 // Currently, only survivors can be root regions.
1144 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
1145 G1RootRegionScanClosure cl(_g1h, this, worker_id);
1146
1147 const uintx interval = PrefetchScanIntervalInBytes;
1148 HeapWord* curr = hr->bottom();
1149 const HeapWord* end = hr->top();
1150 while (curr < end) {
1151 Prefetch::read(curr, interval);
1152 oop obj = oop(curr);
1153 int size = obj->oop_iterate(&cl);
1154 assert(size == obj->size(), "sanity");
1155 curr += size;
1156 }
1157 }
1158
1159 class CMRootRegionScanTask : public AbstractGangTask {
1160 private:
1161 ConcurrentMark* _cm;
1162
1163 public:
1164 CMRootRegionScanTask(ConcurrentMark* cm) :
1165 AbstractGangTask("Root Region Scan"), _cm(cm) { }
1166
1167 void work(uint worker_id) {
1168 assert(Thread::current()->is_ConcurrentGC_thread(),
1169 "this should only be done by a conc GC thread");
1170
1171 CMRootRegions* root_regions = _cm->root_regions();
1172 HeapRegion* hr = root_regions->claim_next();
1173 while (hr != NULL) {
1174 _cm->scanRootRegion(hr, worker_id);
1175 hr = root_regions->claim_next();
1176 }
1177 }
1178 };
1179
1180 void ConcurrentMark::scanRootRegions() {
1181 // Start of concurrent marking.
1182 ClassLoaderDataGraph::clear_claimed_marks();
1183
1184 // scan_in_progress() will have been set to true only if there was
1185 // at least one root region to scan. So, if it's false, we
1186 // should not attempt to do any further work.
1187 if (root_regions()->scan_in_progress()) {
1188 _parallel_marking_threads = calc_parallel_marking_threads();
1189 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1190 "Maximum number of marking threads exceeded");
1191 uint active_workers = MAX2(1U, parallel_marking_threads());
1192
1193 CMRootRegionScanTask task(this);
1194 _parallel_workers->set_active_workers(active_workers);
1195 _parallel_workers->run_task(&task);
1196
1197 // It's possible that has_aborted() is true here without actually
1198 // aborting the survivor scan earlier. This is OK as it's
1199 // mainly used for sanity checking.
1200 root_regions()->scan_finished();
1201 }
1202 }
1203
1204 void ConcurrentMark::markFromRoots() {
1205 // we might be tempted to assert that:
1206 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1207 // "inconsistent argument?");
1208 // However that wouldn't be right, because it's possible that
1209 // a safepoint is indeed in progress as a younger generation
1210 // stop-the-world GC happens even as we mark in this generation.
1211
1212 _restart_for_overflow = false;
1213 force_overflow_conc()->init();
1214
1215 // _g1h has _n_par_threads
1216 _parallel_marking_threads = calc_parallel_marking_threads();
1217 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1218 "Maximum number of marking threads exceeded");
1219
1220 uint active_workers = MAX2(1U, parallel_marking_threads());
1221 assert(active_workers > 0, "Should have been set");
1222
1223 // Parallel task terminator is set in "set_concurrency_and_phase()"
1224 set_concurrency_and_phase(active_workers, true /* concurrent */);
1225
1226 CMConcurrentMarkingTask markingTask(this, cmThread());
1227 _parallel_workers->set_active_workers(active_workers);
1228 _parallel_workers->run_task(&markingTask);
1229 print_stats();
1230 }
1231
1232 // Helper class to get rid of some boilerplate code.
1233 class G1CMTraceTime : public GCTraceTime {
1234 static bool doit_and_prepend(bool doit) {
1235 if (doit) {
1236 gclog_or_tty->put(' ');
1237 }
1238 return doit;
1239 }
1240
1241 public:
1242 G1CMTraceTime(const char* title, bool doit)
1243 : GCTraceTime(title, doit_and_prepend(doit), false, G1CollectedHeap::heap()->gc_timer_cm(),
1244 G1CollectedHeap::heap()->concurrent_mark()->concurrent_gc_id()) {
1245 }
1246 };
1247
1248 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1249 // world is stopped at this checkpoint
1250 assert(SafepointSynchronize::is_at_safepoint(),
1251 "world should be stopped");
1252
1253 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1254
1255 // If a full collection has happened, we shouldn't do this.
1256 if (has_aborted()) {
1257 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1258 return;
1259 }
1260
1261 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1262
1263 if (VerifyDuringGC) {
1264 HandleMark hm; // handle scope
1265 g1h->prepare_for_verify();
1266 Universe::verify(VerifyOption_G1UsePrevMarking,
1267 " VerifyDuringGC:(before)");
1268 }
1269 g1h->check_bitmaps("Remark Start");
1270
1271 G1CollectorPolicy* g1p = g1h->g1_policy();
1272 g1p->record_concurrent_mark_remark_start();
1273
1274 double start = os::elapsedTime();
1275
1276 checkpointRootsFinalWork();
1277
1278 double mark_work_end = os::elapsedTime();
1279
1280 weakRefsWork(clear_all_soft_refs);
1281
1282 if (has_overflown()) {
1283 // Oops. We overflowed. Restart concurrent marking.
1284 _restart_for_overflow = true;
1285 if (G1TraceMarkStackOverflow) {
1286 gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
1287 }
1288
1289 // Verify the heap w.r.t. the previous marking bitmap.
1290 if (VerifyDuringGC) {
1291 HandleMark hm; // handle scope
1292 g1h->prepare_for_verify();
1293 Universe::verify(VerifyOption_G1UsePrevMarking,
1294 " VerifyDuringGC:(overflow)");
1295 }
1296
1297 // Clear the marking state because we will be restarting
1298 // marking due to overflowing the global mark stack.
1299 reset_marking_state();
1300 } else {
1301 {
1302 G1CMTraceTime trace("GC aggregate-data", G1Log::finer());
1303
1304 // Aggregate the per-task counting data that we have accumulated
1305 // while marking.
1306 aggregate_count_data();
1307 }
1308
1309 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1310 // We're done with marking.
1311 // This is the end of the marking cycle, we're expected all
1312 // threads to have SATB queues with active set to true.
1313 satb_mq_set.set_active_all_threads(false, /* new active value */
1314 true /* expected_active */);
1315
1316 if (VerifyDuringGC) {
1317 HandleMark hm; // handle scope
1318 g1h->prepare_for_verify();
1319 Universe::verify(VerifyOption_G1UseNextMarking,
1320 " VerifyDuringGC:(after)");
1321 }
1322 g1h->check_bitmaps("Remark End");
1323 assert(!restart_for_overflow(), "sanity");
1324 // Completely reset the marking state since marking completed
1325 set_non_marking_state();
1326 }
1327
1328 // Expand the marking stack, if we have to and if we can.
1329 if (_markStack.should_expand()) {
1330 _markStack.expand();
1331 }
1332
1333 // Statistics
1334 double now = os::elapsedTime();
1335 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1336 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1337 _remark_times.add((now - start) * 1000.0);
1338
1339 g1p->record_concurrent_mark_remark_end();
1340
1341 G1CMIsAliveClosure is_alive(g1h);
1342 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive);
1343 }
1344
1345 // Base class of the closures that finalize and verify the
1346 // liveness counting data.
1347 class CMCountDataClosureBase: public HeapRegionClosure {
1348 protected:
1349 G1CollectedHeap* _g1h;
1350 ConcurrentMark* _cm;
1351 CardTableModRefBS* _ct_bs;
1352
1353 BitMap* _region_bm;
1354 BitMap* _card_bm;
1355
1356 // Takes a region that's not empty (i.e., it has at least one
1357 // live object in it and sets its corresponding bit on the region
1358 // bitmap to 1. If the region is "starts humongous" it will also set
1359 // to 1 the bits on the region bitmap that correspond to its
1360 // associated "continues humongous" regions.
1361 void set_bit_for_region(HeapRegion* hr) {
1362 assert(!hr->is_continues_humongous(), "should have filtered those out");
1363
1364 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1365 if (!hr->is_starts_humongous()) {
1366 // Normal (non-humongous) case: just set the bit.
1367 _region_bm->par_at_put(index, true);
1368 } else {
1369 // Starts humongous case: calculate how many regions are part of
1370 // this humongous region and then set the bit range.
1371 BitMap::idx_t end_index = (BitMap::idx_t) hr->last_hc_index();
1372 _region_bm->par_at_put_range(index, end_index, true);
1373 }
1374 }
1375
1376 public:
1377 CMCountDataClosureBase(G1CollectedHeap* g1h,
1378 BitMap* region_bm, BitMap* card_bm):
1379 _g1h(g1h), _cm(g1h->concurrent_mark()),
1380 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
1381 _region_bm(region_bm), _card_bm(card_bm) { }
1382 };
1383
1384 // Closure that calculates the # live objects per region. Used
1385 // for verification purposes during the cleanup pause.
1386 class CalcLiveObjectsClosure: public CMCountDataClosureBase {
1387 CMBitMapRO* _bm;
1388 size_t _region_marked_bytes;
1389
1390 public:
1391 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h,
1392 BitMap* region_bm, BitMap* card_bm) :
1393 CMCountDataClosureBase(g1h, region_bm, card_bm),
1394 _bm(bm), _region_marked_bytes(0) { }
1395
1396 bool doHeapRegion(HeapRegion* hr) {
1397
1398 if (hr->is_continues_humongous()) {
1399 // We will ignore these here and process them when their
1400 // associated "starts humongous" region is processed (see
1401 // set_bit_for_heap_region()). Note that we cannot rely on their
1402 // associated "starts humongous" region to have their bit set to
1403 // 1 since, due to the region chunking in the parallel region
1404 // iteration, a "continues humongous" region might be visited
1405 // before its associated "starts humongous".
1406 return false;
1407 }
1408
1409 HeapWord* ntams = hr->next_top_at_mark_start();
1410 HeapWord* start = hr->bottom();
1411
1412 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1413 err_msg("Preconditions not met - "
1414 "start: "PTR_FORMAT", ntams: "PTR_FORMAT", end: "PTR_FORMAT,
1415 p2i(start), p2i(ntams), p2i(hr->end())));
1416
1417 // Find the first marked object at or after "start".
1418 start = _bm->getNextMarkedWordAddress(start, ntams);
1419
1420 size_t marked_bytes = 0;
1421
1422 while (start < ntams) {
1423 oop obj = oop(start);
1424 int obj_sz = obj->size();
1425 HeapWord* obj_end = start + obj_sz;
1426
1427 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
1428 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end);
1429
1430 // Note: if we're looking at the last region in heap - obj_end
1431 // could be actually just beyond the end of the heap; end_idx
1432 // will then correspond to a (non-existent) card that is also
1433 // just beyond the heap.
1434 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) {
1435 // end of object is not card aligned - increment to cover
1436 // all the cards spanned by the object
1437 end_idx += 1;
1438 }
1439
1440 // Set the bits in the card BM for the cards spanned by this object.
1441 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1442
1443 // Add the size of this object to the number of marked bytes.
1444 marked_bytes += (size_t)obj_sz * HeapWordSize;
1445
1446 // Find the next marked object after this one.
1447 start = _bm->getNextMarkedWordAddress(obj_end, ntams);
1448 }
1449
1450 // Mark the allocated-since-marking portion...
1451 HeapWord* top = hr->top();
1452 if (ntams < top) {
1453 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1454 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1455
1456 // Note: if we're looking at the last region in heap - top
1457 // could be actually just beyond the end of the heap; end_idx
1458 // will then correspond to a (non-existent) card that is also
1459 // just beyond the heap.
1460 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1461 // end of object is not card aligned - increment to cover
1462 // all the cards spanned by the object
1463 end_idx += 1;
1464 }
1465 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1466
1467 // This definitely means the region has live objects.
1468 set_bit_for_region(hr);
1469 }
1470
1471 // Update the live region bitmap.
1472 if (marked_bytes > 0) {
1473 set_bit_for_region(hr);
1474 }
1475
1476 // Set the marked bytes for the current region so that
1477 // it can be queried by a calling verification routine
1478 _region_marked_bytes = marked_bytes;
1479
1480 return false;
1481 }
1482
1483 size_t region_marked_bytes() const { return _region_marked_bytes; }
1484 };
1485
1486 // Heap region closure used for verifying the counting data
1487 // that was accumulated concurrently and aggregated during
1488 // the remark pause. This closure is applied to the heap
1489 // regions during the STW cleanup pause.
1490
1491 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure {
1492 G1CollectedHeap* _g1h;
1493 ConcurrentMark* _cm;
1494 CalcLiveObjectsClosure _calc_cl;
1495 BitMap* _region_bm; // Region BM to be verified
1496 BitMap* _card_bm; // Card BM to be verified
1497 bool _verbose; // verbose output?
1498
1499 BitMap* _exp_region_bm; // Expected Region BM values
1500 BitMap* _exp_card_bm; // Expected card BM values
1501
1502 int _failures;
1503
1504 public:
1505 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h,
1506 BitMap* region_bm,
1507 BitMap* card_bm,
1508 BitMap* exp_region_bm,
1509 BitMap* exp_card_bm,
1510 bool verbose) :
1511 _g1h(g1h), _cm(g1h->concurrent_mark()),
1512 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm),
1513 _region_bm(region_bm), _card_bm(card_bm), _verbose(verbose),
1514 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm),
1515 _failures(0) { }
1516
1517 int failures() const { return _failures; }
1518
1519 bool doHeapRegion(HeapRegion* hr) {
1520 if (hr->is_continues_humongous()) {
1521 // We will ignore these here and process them when their
1522 // associated "starts humongous" region is processed (see
1523 // set_bit_for_heap_region()). Note that we cannot rely on their
1524 // associated "starts humongous" region to have their bit set to
1525 // 1 since, due to the region chunking in the parallel region
1526 // iteration, a "continues humongous" region might be visited
1527 // before its associated "starts humongous".
1528 return false;
1529 }
1530
1531 int failures = 0;
1532
1533 // Call the CalcLiveObjectsClosure to walk the marking bitmap for
1534 // this region and set the corresponding bits in the expected region
1535 // and card bitmaps.
1536 bool res = _calc_cl.doHeapRegion(hr);
1537 assert(res == false, "should be continuing");
1538
1539 MutexLockerEx x((_verbose ? ParGCRareEvent_lock : NULL),
1540 Mutex::_no_safepoint_check_flag);
1541
1542 // Verify the marked bytes for this region.
1543 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1544 size_t act_marked_bytes = hr->next_marked_bytes();
1545
1546 // We're not OK if expected marked bytes > actual marked bytes. It means
1547 // we have missed accounting some objects during the actual marking.
1548 if (exp_marked_bytes > act_marked_bytes) {
1549 if (_verbose) {
1550 gclog_or_tty->print_cr("Region %u: marked bytes mismatch: "
1551 "expected: " SIZE_FORMAT ", actual: " SIZE_FORMAT,
1552 hr->hrm_index(), exp_marked_bytes, act_marked_bytes);
1553 }
1554 failures += 1;
1555 }
1556
1557 // Verify the bit, for this region, in the actual and expected
1558 // (which was just calculated) region bit maps.
1559 // We're not OK if the bit in the calculated expected region
1560 // bitmap is set and the bit in the actual region bitmap is not.
1561 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index();
1562
1563 bool expected = _exp_region_bm->at(index);
1564 bool actual = _region_bm->at(index);
1565 if (expected && !actual) {
1566 if (_verbose) {
1567 gclog_or_tty->print_cr("Region %u: region bitmap mismatch: "
1568 "expected: %s, actual: %s",
1569 hr->hrm_index(),
1570 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1571 }
1572 failures += 1;
1573 }
1574
1575 // Verify that the card bit maps for the cards spanned by the current
1576 // region match. We have an error if we have a set bit in the expected
1577 // bit map and the corresponding bit in the actual bitmap is not set.
1578
1579 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom());
1580 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top());
1581
1582 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1583 expected = _exp_card_bm->at(i);
1584 actual = _card_bm->at(i);
1585
1586 if (expected && !actual) {
1587 if (_verbose) {
1588 gclog_or_tty->print_cr("Region %u: card bitmap mismatch at " SIZE_FORMAT ": "
1589 "expected: %s, actual: %s",
1590 hr->hrm_index(), i,
1591 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1592 }
1593 failures += 1;
1594 }
1595 }
1596
1597 if (failures > 0 && _verbose) {
1598 gclog_or_tty->print_cr("Region " HR_FORMAT ", ntams: " PTR_FORMAT ", "
1599 "marked_bytes: calc/actual " SIZE_FORMAT "/" SIZE_FORMAT,
1600 HR_FORMAT_PARAMS(hr), p2i(hr->next_top_at_mark_start()),
1601 _calc_cl.region_marked_bytes(), hr->next_marked_bytes());
1602 }
1603
1604 _failures += failures;
1605
1606 // We could stop iteration over the heap when we
1607 // find the first violating region by returning true.
1608 return false;
1609 }
1610 };
1611
1612 class G1ParVerifyFinalCountTask: public AbstractGangTask {
1613 protected:
1614 G1CollectedHeap* _g1h;
1615 ConcurrentMark* _cm;
1616 BitMap* _actual_region_bm;
1617 BitMap* _actual_card_bm;
1618
1619 uint _n_workers;
1620
1621 BitMap* _expected_region_bm;
1622 BitMap* _expected_card_bm;
1623
1624 int _failures;
1625 bool _verbose;
1626
1627 HeapRegionClaimer _hrclaimer;
1628
1629 public:
1630 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h,
1631 BitMap* region_bm, BitMap* card_bm,
1632 BitMap* expected_region_bm, BitMap* expected_card_bm)
1633 : AbstractGangTask("G1 verify final counting"),
1634 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1635 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1636 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm),
1637 _failures(0), _verbose(false),
1638 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1639 assert(VerifyDuringGC, "don't call this otherwise");
1640 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity");
1641 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity");
1642
1643 _verbose = _cm->verbose_medium();
1644 }
1645
1646 void work(uint worker_id) {
1647 assert(worker_id < _n_workers, "invariant");
1648
1649 VerifyLiveObjectDataHRClosure verify_cl(_g1h,
1650 _actual_region_bm, _actual_card_bm,
1651 _expected_region_bm,
1652 _expected_card_bm,
1653 _verbose);
1654
1655 _g1h->heap_region_par_iterate(&verify_cl, worker_id, &_hrclaimer);
1656
1657 Atomic::add(verify_cl.failures(), &_failures);
1658 }
1659
1660 int failures() const { return _failures; }
1661 };
1662
1663 // Closure that finalizes the liveness counting data.
1664 // Used during the cleanup pause.
1665 // Sets the bits corresponding to the interval [NTAMS, top]
1666 // (which contains the implicitly live objects) in the
1667 // card liveness bitmap. Also sets the bit for each region,
1668 // containing live data, in the region liveness bitmap.
1669
1670 class FinalCountDataUpdateClosure: public CMCountDataClosureBase {
1671 public:
1672 FinalCountDataUpdateClosure(G1CollectedHeap* g1h,
1673 BitMap* region_bm,
1674 BitMap* card_bm) :
1675 CMCountDataClosureBase(g1h, region_bm, card_bm) { }
1676
1677 bool doHeapRegion(HeapRegion* hr) {
1678
1679 if (hr->is_continues_humongous()) {
1680 // We will ignore these here and process them when their
1681 // associated "starts humongous" region is processed (see
1682 // set_bit_for_heap_region()). Note that we cannot rely on their
1683 // associated "starts humongous" region to have their bit set to
1684 // 1 since, due to the region chunking in the parallel region
1685 // iteration, a "continues humongous" region might be visited
1686 // before its associated "starts humongous".
1687 return false;
1688 }
1689
1690 HeapWord* ntams = hr->next_top_at_mark_start();
1691 HeapWord* top = hr->top();
1692
1693 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1694
1695 // Mark the allocated-since-marking portion...
1696 if (ntams < top) {
1697 // This definitely means the region has live objects.
1698 set_bit_for_region(hr);
1699
1700 // Now set the bits in the card bitmap for [ntams, top)
1701 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1702 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1703
1704 // Note: if we're looking at the last region in heap - top
1705 // could be actually just beyond the end of the heap; end_idx
1706 // will then correspond to a (non-existent) card that is also
1707 // just beyond the heap.
1708 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1709 // end of object is not card aligned - increment to cover
1710 // all the cards spanned by the object
1711 end_idx += 1;
1712 }
1713
1714 assert(end_idx <= _card_bm->size(),
1715 err_msg("oob: end_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1716 end_idx, _card_bm->size()));
1717 assert(start_idx < _card_bm->size(),
1718 err_msg("oob: start_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1719 start_idx, _card_bm->size()));
1720
1721 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1722 }
1723
1724 // Set the bit for the region if it contains live data
1725 if (hr->next_marked_bytes() > 0) {
1726 set_bit_for_region(hr);
1727 }
1728
1729 return false;
1730 }
1731 };
1732
1733 class G1ParFinalCountTask: public AbstractGangTask {
1734 protected:
1735 G1CollectedHeap* _g1h;
1736 ConcurrentMark* _cm;
1737 BitMap* _actual_region_bm;
1738 BitMap* _actual_card_bm;
1739
1740 uint _n_workers;
1741 HeapRegionClaimer _hrclaimer;
1742
1743 public:
1744 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm)
1745 : AbstractGangTask("G1 final counting"),
1746 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1747 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1748 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) {
1749 }
1750
1751 void work(uint worker_id) {
1752 assert(worker_id < _n_workers, "invariant");
1753
1754 FinalCountDataUpdateClosure final_update_cl(_g1h,
1755 _actual_region_bm,
1756 _actual_card_bm);
1757
1758 _g1h->heap_region_par_iterate(&final_update_cl, worker_id, &_hrclaimer);
1759 }
1760 };
1761
1762 class G1ParNoteEndTask;
1763
1764 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1765 G1CollectedHeap* _g1;
1766 size_t _max_live_bytes;
1767 uint _regions_claimed;
1768 size_t _freed_bytes;
1769 FreeRegionList* _local_cleanup_list;
1770 HeapRegionSetCount _old_regions_removed;
1771 HeapRegionSetCount _humongous_regions_removed;
1772 HRRSCleanupTask* _hrrs_cleanup_task;
1773 double _claimed_region_time;
1774 double _max_region_time;
1775
1776 public:
1777 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1778 FreeRegionList* local_cleanup_list,
1779 HRRSCleanupTask* hrrs_cleanup_task) :
1780 _g1(g1),
1781 _max_live_bytes(0), _regions_claimed(0),
1782 _freed_bytes(0),
1783 _claimed_region_time(0.0), _max_region_time(0.0),
1784 _local_cleanup_list(local_cleanup_list),
1785 _old_regions_removed(),
1786 _humongous_regions_removed(),
1787 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1788
1789 size_t freed_bytes() { return _freed_bytes; }
1790 const HeapRegionSetCount& old_regions_removed() { return _old_regions_removed; }
1791 const HeapRegionSetCount& humongous_regions_removed() { return _humongous_regions_removed; }
1792
1793 bool doHeapRegion(HeapRegion *hr) {
1794 if (hr->is_continues_humongous()) {
1795 return false;
1796 }
1797 // We use a claim value of zero here because all regions
1798 // were claimed with value 1 in the FinalCount task.
1799 _g1->reset_gc_time_stamps(hr);
1800 double start = os::elapsedTime();
1801 _regions_claimed++;
1802 hr->note_end_of_marking();
1803 _max_live_bytes += hr->max_live_bytes();
1804
1805 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) {
1806 _freed_bytes += hr->used();
1807 hr->set_containing_set(NULL);
1808 if (hr->is_humongous()) {
1809 assert(hr->is_starts_humongous(), "we should only see starts humongous");
1810 _humongous_regions_removed.increment(1u, hr->capacity());
1811 _g1->free_humongous_region(hr, _local_cleanup_list, true);
1812 } else {
1813 _old_regions_removed.increment(1u, hr->capacity());
1814 _g1->free_region(hr, _local_cleanup_list, true);
1815 }
1816 } else {
1817 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1818 }
1819
1820 double region_time = (os::elapsedTime() - start);
1821 _claimed_region_time += region_time;
1822 if (region_time > _max_region_time) {
1823 _max_region_time = region_time;
1824 }
1825 return false;
1826 }
1827
1828 size_t max_live_bytes() { return _max_live_bytes; }
1829 uint regions_claimed() { return _regions_claimed; }
1830 double claimed_region_time_sec() { return _claimed_region_time; }
1831 double max_region_time_sec() { return _max_region_time; }
1832 };
1833
1834 class G1ParNoteEndTask: public AbstractGangTask {
1835 friend class G1NoteEndOfConcMarkClosure;
1836
1837 protected:
1838 G1CollectedHeap* _g1h;
1839 size_t _max_live_bytes;
1840 size_t _freed_bytes;
1841 FreeRegionList* _cleanup_list;
1842 HeapRegionClaimer _hrclaimer;
1843
1844 public:
1845 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) :
1846 AbstractGangTask("G1 note end"), _g1h(g1h), _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list), _hrclaimer(n_workers) {
1847 }
1848
1849 void work(uint worker_id) {
1850 FreeRegionList local_cleanup_list("Local Cleanup List");
1851 HRRSCleanupTask hrrs_cleanup_task;
1852 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
1853 &hrrs_cleanup_task);
1854 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer);
1855 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1856
1857 // Now update the lists
1858 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed());
1859 {
1860 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1861 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes());
1862 _max_live_bytes += g1_note_end.max_live_bytes();
1863 _freed_bytes += g1_note_end.freed_bytes();
1864
1865 // If we iterate over the global cleanup list at the end of
1866 // cleanup to do this printing we will not guarantee to only
1867 // generate output for the newly-reclaimed regions (the list
1868 // might not be empty at the beginning of cleanup; we might
1869 // still be working on its previous contents). So we do the
1870 // printing here, before we append the new regions to the global
1871 // cleanup list.
1872
1873 G1HRPrinter* hr_printer = _g1h->hr_printer();
1874 if (hr_printer->is_active()) {
1875 FreeRegionListIterator iter(&local_cleanup_list);
1876 while (iter.more_available()) {
1877 HeapRegion* hr = iter.get_next();
1878 hr_printer->cleanup(hr);
1879 }
1880 }
1881
1882 _cleanup_list->add_ordered(&local_cleanup_list);
1883 assert(local_cleanup_list.is_empty(), "post-condition");
1884
1885 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1886 }
1887 }
1888 size_t max_live_bytes() { return _max_live_bytes; }
1889 size_t freed_bytes() { return _freed_bytes; }
1890 };
1891
1892 class G1ParScrubRemSetTask: public AbstractGangTask {
1893 protected:
1894 G1RemSet* _g1rs;
1895 BitMap* _region_bm;
1896 BitMap* _card_bm;
1897 HeapRegionClaimer _hrclaimer;
1898
1899 public:
1900 G1ParScrubRemSetTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm, uint n_workers) :
1901 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), _region_bm(region_bm), _card_bm(card_bm), _hrclaimer(n_workers) {
1902 }
1903
1904 void work(uint worker_id) {
1905 _g1rs->scrub(_region_bm, _card_bm, worker_id, &_hrclaimer);
1906 }
1907
1908 };
1909
1910 void ConcurrentMark::cleanup() {
1911 // world is stopped at this checkpoint
1912 assert(SafepointSynchronize::is_at_safepoint(),
1913 "world should be stopped");
1914 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1915
1916 // If a full collection has happened, we shouldn't do this.
1917 if (has_aborted()) {
1918 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1919 return;
1920 }
1921
1922 g1h->verify_region_sets_optional();
1923
1924 if (VerifyDuringGC) {
1925 HandleMark hm; // handle scope
1926 g1h->prepare_for_verify();
1927 Universe::verify(VerifyOption_G1UsePrevMarking,
1928 " VerifyDuringGC:(before)");
1929 }
1930 g1h->check_bitmaps("Cleanup Start");
1931
1932 G1CollectorPolicy* g1p = g1h->g1_policy();
1933 g1p->record_concurrent_mark_cleanup_start();
1934
1935 double start = os::elapsedTime();
1936
1937 HeapRegionRemSet::reset_for_cleanup_tasks();
1938
1939 // Do counting once more with the world stopped for good measure.
1940 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
1941
1942 g1h->workers()->run_task(&g1_par_count_task);
1943
1944 if (VerifyDuringGC) {
1945 // Verify that the counting data accumulated during marking matches
1946 // that calculated by walking the marking bitmap.
1947
1948 // Bitmaps to hold expected values
1949 BitMap expected_region_bm(_region_bm.size(), true);
1950 BitMap expected_card_bm(_card_bm.size(), true);
1951
1952 G1ParVerifyFinalCountTask g1_par_verify_task(g1h,
1953 &_region_bm,
1954 &_card_bm,
1955 &expected_region_bm,
1956 &expected_card_bm);
1957
1958 g1h->workers()->run_task(&g1_par_verify_task);
1959
1960 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures");
1961 }
1962
1963 size_t start_used_bytes = g1h->used();
1964 g1h->set_marking_complete();
1965
1966 double count_end = os::elapsedTime();
1967 double this_final_counting_time = (count_end - start);
1968 _total_counting_time += this_final_counting_time;
1969
1970 if (G1PrintRegionLivenessInfo) {
1971 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking");
1972 _g1h->heap_region_iterate(&cl);
1973 }
1974
1975 // Install newly created mark bitMap as "prev".
1976 swapMarkBitMaps();
1977
1978 g1h->reset_gc_time_stamp();
1979
1980 uint n_workers = _g1h->workers()->active_workers();
1981
1982 // Note end of marking in all heap regions.
1983 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers);
1984 g1h->workers()->run_task(&g1_par_note_end_task);
1985 g1h->check_gc_time_stamps();
1986
1987 if (!cleanup_list_is_empty()) {
1988 // The cleanup list is not empty, so we'll have to process it
1989 // concurrently. Notify anyone else that might be wanting free
1990 // regions that there will be more free regions coming soon.
1991 g1h->set_free_regions_coming();
1992 }
1993
1994 // call below, since it affects the metric by which we sort the heap
1995 // regions.
1996 if (G1ScrubRemSets) {
1997 double rs_scrub_start = os::elapsedTime();
1998 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm, n_workers);
1999 g1h->workers()->run_task(&g1_par_scrub_rs_task);
2000
2001 double rs_scrub_end = os::elapsedTime();
2002 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start);
2003 _total_rs_scrub_time += this_rs_scrub_time;
2004 }
2005
2006 // this will also free any regions totally full of garbage objects,
2007 // and sort the regions.
2008 g1h->g1_policy()->record_concurrent_mark_cleanup_end();
2009
2010 // Statistics.
2011 double end = os::elapsedTime();
2012 _cleanup_times.add((end - start) * 1000.0);
2013
2014 if (G1Log::fine()) {
2015 g1h->g1_policy()->print_heap_transition(start_used_bytes);
2016 }
2017
2018 // Clean up will have freed any regions completely full of garbage.
2019 // Update the soft reference policy with the new heap occupancy.
2020 Universe::update_heap_info_at_gc();
2021
2022 if (VerifyDuringGC) {
2023 HandleMark hm; // handle scope
2024 g1h->prepare_for_verify();
2025 Universe::verify(VerifyOption_G1UsePrevMarking,
2026 " VerifyDuringGC:(after)");
2027 }
2028
2029 g1h->check_bitmaps("Cleanup End");
2030
2031 g1h->verify_region_sets_optional();
2032
2033 // We need to make this be a "collection" so any collection pause that
2034 // races with it goes around and waits for completeCleanup to finish.
2035 g1h->increment_total_collections();
2036
2037 // Clean out dead classes and update Metaspace sizes.
2038 if (ClassUnloadingWithConcurrentMark) {
2039 ClassLoaderDataGraph::purge();
2040 }
2041 MetaspaceGC::compute_new_size();
2042
2043 // We reclaimed old regions so we should calculate the sizes to make
2044 // sure we update the old gen/space data.
2045 g1h->g1mm()->update_sizes();
2046 g1h->allocation_context_stats().update_after_mark();
2047
2048 g1h->trace_heap_after_concurrent_cycle();
2049 }
2050
2051 void ConcurrentMark::completeCleanup() {
2052 if (has_aborted()) return;
2053
2054 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2055
2056 _cleanup_list.verify_optional();
2057 FreeRegionList tmp_free_list("Tmp Free List");
2058
2059 if (G1ConcRegionFreeingVerbose) {
2060 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2061 "cleanup list has %u entries",
2062 _cleanup_list.length());
2063 }
2064
2065 // No one else should be accessing the _cleanup_list at this point,
2066 // so it is not necessary to take any locks
2067 while (!_cleanup_list.is_empty()) {
2068 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */);
2069 assert(hr != NULL, "Got NULL from a non-empty list");
2070 hr->par_clear();
2071 tmp_free_list.add_ordered(hr);
2072
2073 // Instead of adding one region at a time to the secondary_free_list,
2074 // we accumulate them in the local list and move them a few at a
2075 // time. This also cuts down on the number of notify_all() calls
2076 // we do during this process. We'll also append the local list when
2077 // _cleanup_list is empty (which means we just removed the last
2078 // region from the _cleanup_list).
2079 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
2080 _cleanup_list.is_empty()) {
2081 if (G1ConcRegionFreeingVerbose) {
2082 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2083 "appending %u entries to the secondary_free_list, "
2084 "cleanup list still has %u entries",
2085 tmp_free_list.length(),
2086 _cleanup_list.length());
2087 }
2088
2089 {
2090 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
2091 g1h->secondary_free_list_add(&tmp_free_list);
2092 SecondaryFreeList_lock->notify_all();
2093 }
2094 #ifndef PRODUCT
2095 if (G1StressConcRegionFreeing) {
2096 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
2097 os::sleep(Thread::current(), (jlong) 1, false);
2098 }
2099 }
2100 #endif
2101 }
2102 }
2103 assert(tmp_free_list.is_empty(), "post-condition");
2104 }
2105
2106 // Supporting Object and Oop closures for reference discovery
2107 // and processing in during marking
2108
2109 bool G1CMIsAliveClosure::do_object_b(oop obj) {
2110 HeapWord* addr = (HeapWord*)obj;
2111 return addr != NULL &&
2112 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
2113 }
2114
2115 // 'Keep Alive' oop closure used by both serial parallel reference processing.
2116 // Uses the CMTask associated with a worker thread (for serial reference
2117 // processing the CMTask for worker 0 is used) to preserve (mark) and
2118 // trace referent objects.
2119 //
2120 // Using the CMTask and embedded local queues avoids having the worker
2121 // threads operating on the global mark stack. This reduces the risk
2122 // of overflowing the stack - which we would rather avoid at this late
2123 // state. Also using the tasks' local queues removes the potential
2124 // of the workers interfering with each other that could occur if
2125 // operating on the global stack.
2126
2127 class G1CMKeepAliveAndDrainClosure: public OopClosure {
2128 ConcurrentMark* _cm;
2129 CMTask* _task;
2130 int _ref_counter_limit;
2131 int _ref_counter;
2132 bool _is_serial;
2133 public:
2134 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2135 _cm(cm), _task(task), _is_serial(is_serial),
2136 _ref_counter_limit(G1RefProcDrainInterval) {
2137 assert(_ref_counter_limit > 0, "sanity");
2138 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
2139 _ref_counter = _ref_counter_limit;
2140 }
2141
2142 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
2143 virtual void do_oop( oop* p) { do_oop_work(p); }
2144
2145 template <class T> void do_oop_work(T* p) {
2146 if (!_cm->has_overflown()) {
2147 oop obj = oopDesc::load_decode_heap_oop(p);
2148 if (_cm->verbose_high()) {
2149 gclog_or_tty->print_cr("\t[%u] we're looking at location "
2150 "*"PTR_FORMAT" = "PTR_FORMAT,
2151 _task->worker_id(), p2i(p), p2i((void*) obj));
2152 }
2153
2154 _task->deal_with_reference(obj);
2155 _ref_counter--;
2156
2157 if (_ref_counter == 0) {
2158 // We have dealt with _ref_counter_limit references, pushing them
2159 // and objects reachable from them on to the local stack (and
2160 // possibly the global stack). Call CMTask::do_marking_step() to
2161 // process these entries.
2162 //
2163 // We call CMTask::do_marking_step() in a loop, which we'll exit if
2164 // there's nothing more to do (i.e. we're done with the entries that
2165 // were pushed as a result of the CMTask::deal_with_reference() calls
2166 // above) or we overflow.
2167 //
2168 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2169 // flag while there may still be some work to do. (See the comment at
2170 // the beginning of CMTask::do_marking_step() for those conditions -
2171 // one of which is reaching the specified time target.) It is only
2172 // when CMTask::do_marking_step() returns without setting the
2173 // has_aborted() flag that the marking step has completed.
2174 do {
2175 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
2176 _task->do_marking_step(mark_step_duration_ms,
2177 false /* do_termination */,
2178 _is_serial);
2179 } while (_task->has_aborted() && !_cm->has_overflown());
2180 _ref_counter = _ref_counter_limit;
2181 }
2182 } else {
2183 if (_cm->verbose_high()) {
2184 gclog_or_tty->print_cr("\t[%u] CM Overflow", _task->worker_id());
2185 }
2186 }
2187 }
2188 };
2189
2190 // 'Drain' oop closure used by both serial and parallel reference processing.
2191 // Uses the CMTask associated with a given worker thread (for serial
2192 // reference processing the CMtask for worker 0 is used). Calls the
2193 // do_marking_step routine, with an unbelievably large timeout value,
2194 // to drain the marking data structures of the remaining entries
2195 // added by the 'keep alive' oop closure above.
2196
2197 class G1CMDrainMarkingStackClosure: public VoidClosure {
2198 ConcurrentMark* _cm;
2199 CMTask* _task;
2200 bool _is_serial;
2201 public:
2202 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2203 _cm(cm), _task(task), _is_serial(is_serial) {
2204 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
2205 }
2206
2207 void do_void() {
2208 do {
2209 if (_cm->verbose_high()) {
2210 gclog_or_tty->print_cr("\t[%u] Drain: Calling do_marking_step - serial: %s",
2211 _task->worker_id(), BOOL_TO_STR(_is_serial));
2212 }
2213
2214 // We call CMTask::do_marking_step() to completely drain the local
2215 // and global marking stacks of entries pushed by the 'keep alive'
2216 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
2217 //
2218 // CMTask::do_marking_step() is called in a loop, which we'll exit
2219 // if there's nothing more to do (i.e. we've completely drained the
2220 // entries that were pushed as a a result of applying the 'keep alive'
2221 // closure to the entries on the discovered ref lists) or we overflow
2222 // the global marking stack.
2223 //
2224 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2225 // flag while there may still be some work to do. (See the comment at
2226 // the beginning of CMTask::do_marking_step() for those conditions -
2227 // one of which is reaching the specified time target.) It is only
2228 // when CMTask::do_marking_step() returns without setting the
2229 // has_aborted() flag that the marking step has completed.
2230
2231 _task->do_marking_step(1000000000.0 /* something very large */,
2232 true /* do_termination */,
2233 _is_serial);
2234 } while (_task->has_aborted() && !_cm->has_overflown());
2235 }
2236 };
2237
2238 // Implementation of AbstractRefProcTaskExecutor for parallel
2239 // reference processing at the end of G1 concurrent marking
2240
2241 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
2242 private:
2243 G1CollectedHeap* _g1h;
2244 ConcurrentMark* _cm;
2245 WorkGang* _workers;
2246 uint _active_workers;
2247
2248 public:
2249 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
2250 ConcurrentMark* cm,
2251 WorkGang* workers,
2252 uint n_workers) :
2253 _g1h(g1h), _cm(cm),
2254 _workers(workers), _active_workers(n_workers) { }
2255
2256 // Executes the given task using concurrent marking worker threads.
2257 virtual void execute(ProcessTask& task);
2258 virtual void execute(EnqueueTask& task);
2259 };
2260
2261 class G1CMRefProcTaskProxy: public AbstractGangTask {
2262 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
2263 ProcessTask& _proc_task;
2264 G1CollectedHeap* _g1h;
2265 ConcurrentMark* _cm;
2266
2267 public:
2268 G1CMRefProcTaskProxy(ProcessTask& proc_task,
2269 G1CollectedHeap* g1h,
2270 ConcurrentMark* cm) :
2271 AbstractGangTask("Process reference objects in parallel"),
2272 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
2273 ReferenceProcessor* rp = _g1h->ref_processor_cm();
2274 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
2275 }
2276
2277 virtual void work(uint worker_id) {
2278 ResourceMark rm;
2279 HandleMark hm;
2280 CMTask* task = _cm->task(worker_id);
2281 G1CMIsAliveClosure g1_is_alive(_g1h);
2282 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
2283 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
2284
2285 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
2286 }
2287 };
2288
2289 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
2290 assert(_workers != NULL, "Need parallel worker threads.");
2291 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2292
2293 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
2294
2295 // We need to reset the concurrency level before each
2296 // proxy task execution, so that the termination protocol
2297 // and overflow handling in CMTask::do_marking_step() knows
2298 // how many workers to wait for.
2299 _cm->set_concurrency(_active_workers);
2300 _workers->run_task(&proc_task_proxy);
2301 }
2302
2303 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
2304 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
2305 EnqueueTask& _enq_task;
2306
2307 public:
2308 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
2309 AbstractGangTask("Enqueue reference objects in parallel"),
2310 _enq_task(enq_task) { }
2311
2312 virtual void work(uint worker_id) {
2313 _enq_task.work(worker_id);
2314 }
2315 };
2316
2317 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2318 assert(_workers != NULL, "Need parallel worker threads.");
2319 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2320
2321 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2322
2323 // Not strictly necessary but...
2324 //
2325 // We need to reset the concurrency level before each
2326 // proxy task execution, so that the termination protocol
2327 // and overflow handling in CMTask::do_marking_step() knows
2328 // how many workers to wait for.
2329 _cm->set_concurrency(_active_workers);
2330 _workers->run_task(&enq_task_proxy);
2331 }
2332
2333 void ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) {
2334 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes);
2335 }
2336
2337 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2338 if (has_overflown()) {
2339 // Skip processing the discovered references if we have
2340 // overflown the global marking stack. Reference objects
2341 // only get discovered once so it is OK to not
2342 // de-populate the discovered reference lists. We could have,
2343 // but the only benefit would be that, when marking restarts,
2344 // less reference objects are discovered.
2345 return;
2346 }
2347
2348 ResourceMark rm;
2349 HandleMark hm;
2350
2351 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2352
2353 // Is alive closure.
2354 G1CMIsAliveClosure g1_is_alive(g1h);
2355
2356 // Inner scope to exclude the cleaning of the string and symbol
2357 // tables from the displayed time.
2358 {
2359 G1CMTraceTime t("GC ref-proc", G1Log::finer());
2360
2361 ReferenceProcessor* rp = g1h->ref_processor_cm();
2362
2363 // See the comment in G1CollectedHeap::ref_processing_init()
2364 // about how reference processing currently works in G1.
2365
2366 // Set the soft reference policy
2367 rp->setup_policy(clear_all_soft_refs);
2368 assert(_markStack.isEmpty(), "mark stack should be empty");
2369
2370 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2371 // in serial reference processing. Note these closures are also
2372 // used for serially processing (by the the current thread) the
2373 // JNI references during parallel reference processing.
2374 //
2375 // These closures do not need to synchronize with the worker
2376 // threads involved in parallel reference processing as these
2377 // instances are executed serially by the current thread (e.g.
2378 // reference processing is not multi-threaded and is thus
2379 // performed by the current thread instead of a gang worker).
2380 //
2381 // The gang tasks involved in parallel reference processing create
2382 // their own instances of these closures, which do their own
2383 // synchronization among themselves.
2384 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2385 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2386
2387 // We need at least one active thread. If reference processing
2388 // is not multi-threaded we use the current (VMThread) thread,
2389 // otherwise we use the work gang from the G1CollectedHeap and
2390 // we utilize all the worker threads we can.
2391 bool processing_is_mt = rp->processing_is_mt();
2392 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2393 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U);
2394
2395 // Parallel processing task executor.
2396 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2397 g1h->workers(), active_workers);
2398 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2399
2400 // Set the concurrency level. The phase was already set prior to
2401 // executing the remark task.
2402 set_concurrency(active_workers);
2403
2404 // Set the degree of MT processing here. If the discovery was done MT,
2405 // the number of threads involved during discovery could differ from
2406 // the number of active workers. This is OK as long as the discovered
2407 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2408 rp->set_active_mt_degree(active_workers);
2409
2410 // Process the weak references.
2411 const ReferenceProcessorStats& stats =
2412 rp->process_discovered_references(&g1_is_alive,
2413 &g1_keep_alive,
2414 &g1_drain_mark_stack,
2415 executor,
2416 g1h->gc_timer_cm(),
2417 concurrent_gc_id());
2418 g1h->gc_tracer_cm()->report_gc_reference_stats(stats);
2419
2420 // The do_oop work routines of the keep_alive and drain_marking_stack
2421 // oop closures will set the has_overflown flag if we overflow the
2422 // global marking stack.
2423
2424 assert(_markStack.overflow() || _markStack.isEmpty(),
2425 "mark stack should be empty (unless it overflowed)");
2426
2427 if (_markStack.overflow()) {
2428 // This should have been done already when we tried to push an
2429 // entry on to the global mark stack. But let's do it again.
2430 set_has_overflown();
2431 }
2432
2433 assert(rp->num_q() == active_workers, "why not");
2434
2435 rp->enqueue_discovered_references(executor);
2436
2437 rp->verify_no_references_recorded();
2438 assert(!rp->discovery_enabled(), "Post condition");
2439 }
2440
2441 if (has_overflown()) {
2442 // We can not trust g1_is_alive if the marking stack overflowed
2443 return;
2444 }
2445
2446 assert(_markStack.isEmpty(), "Marking should have completed");
2447
2448 // Unload Klasses, String, Symbols, Code Cache, etc.
2449 {
2450 G1CMTraceTime trace("Unloading", G1Log::finer());
2451
2452 if (ClassUnloadingWithConcurrentMark) {
2453 bool purged_classes;
2454
2455 {
2456 G1CMTraceTime trace("System Dictionary Unloading", G1Log::finest());
2457 purged_classes = SystemDictionary::do_unloading(&g1_is_alive, false /* Defer klass cleaning */);
2458 }
2459
2460 {
2461 G1CMTraceTime trace("Parallel Unloading", G1Log::finest());
2462 weakRefsWorkParallelPart(&g1_is_alive, purged_classes);
2463 }
2464 }
2465
2466 if (G1StringDedup::is_enabled()) {
2467 G1CMTraceTime trace("String Deduplication Unlink", G1Log::finest());
2468 G1StringDedup::unlink(&g1_is_alive);
2469 }
2470 }
2471 }
2472
2473 void ConcurrentMark::swapMarkBitMaps() {
2474 CMBitMapRO* temp = _prevMarkBitMap;
2475 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap;
2476 _nextMarkBitMap = (CMBitMap*) temp;
2477 }
2478
2479 // Closure for marking entries in SATB buffers.
2480 class CMSATBBufferClosure : public SATBBufferClosure {
2481 private:
2482 CMTask* _task;
2483 G1CollectedHeap* _g1h;
2484
2485 // This is very similar to CMTask::deal_with_reference, but with
2486 // more relaxed requirements for the argument, so this must be more
2487 // circumspect about treating the argument as an object.
2488 void do_entry(void* entry) const {
2489 _task->increment_refs_reached();
2490 HeapRegion* hr = _g1h->heap_region_containing_raw(entry);
2491 if (entry < hr->next_top_at_mark_start()) {
2492 // Until we get here, we don't know whether entry refers to a valid
2493 // object; it could instead have been a stale reference.
2494 oop obj = static_cast<oop>(entry);
2495 assert(obj->is_oop(true /* ignore mark word */),
2496 err_msg("Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj)));
2497 _task->make_reference_grey(obj, hr);
2498 }
2499 }
2500
2501 public:
2502 CMSATBBufferClosure(CMTask* task, G1CollectedHeap* g1h)
2503 : _task(task), _g1h(g1h) { }
2504
2505 virtual void do_buffer(void** buffer, size_t size) {
2506 for (size_t i = 0; i < size; ++i) {
2507 do_entry(buffer[i]);
2508 }
2509 }
2510 };
2511
2512 class G1RemarkThreadsClosure : public ThreadClosure {
2513 CMSATBBufferClosure _cm_satb_cl;
2514 G1CMOopClosure _cm_cl;
2515 MarkingCodeBlobClosure _code_cl;
2516 int _thread_parity;
2517
2518 public:
2519 G1RemarkThreadsClosure(G1CollectedHeap* g1h, CMTask* task) :
2520 _cm_satb_cl(task, g1h),
2521 _cm_cl(g1h, g1h->concurrent_mark(), task),
2522 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations),
2523 _thread_parity(Threads::thread_claim_parity()) {}
2524
2525 void do_thread(Thread* thread) {
2526 if (thread->is_Java_thread()) {
2527 if (thread->claim_oops_do(true, _thread_parity)) {
2528 JavaThread* jt = (JavaThread*)thread;
2529
2530 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking
2531 // however the liveness of oops reachable from nmethods have very complex lifecycles:
2532 // * Alive if on the stack of an executing method
2533 // * Weakly reachable otherwise
2534 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be
2535 // live by the SATB invariant but other oops recorded in nmethods may behave differently.
2536 jt->nmethods_do(&_code_cl);
2537
2538 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl);
2539 }
2540 } else if (thread->is_VM_thread()) {
2541 if (thread->claim_oops_do(true, _thread_parity)) {
2542 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl);
2543 }
2544 }
2545 }
2546 };
2547
2548 class CMRemarkTask: public AbstractGangTask {
2549 private:
2550 ConcurrentMark* _cm;
2551 public:
2552 void work(uint worker_id) {
2553 // Since all available tasks are actually started, we should
2554 // only proceed if we're supposed to be active.
2555 if (worker_id < _cm->active_tasks()) {
2556 CMTask* task = _cm->task(worker_id);
2557 task->record_start_time();
2558 {
2559 ResourceMark rm;
2560 HandleMark hm;
2561
2562 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task);
2563 Threads::threads_do(&threads_f);
2564 }
2565
2566 do {
2567 task->do_marking_step(1000000000.0 /* something very large */,
2568 true /* do_termination */,
2569 false /* is_serial */);
2570 } while (task->has_aborted() && !_cm->has_overflown());
2571 // If we overflow, then we do not want to restart. We instead
2572 // want to abort remark and do concurrent marking again.
2573 task->record_end_time();
2574 }
2575 }
2576
2577 CMRemarkTask(ConcurrentMark* cm, uint active_workers) :
2578 AbstractGangTask("Par Remark"), _cm(cm) {
2579 _cm->terminator()->reset_for_reuse(active_workers);
2580 }
2581 };
2582
2583 void ConcurrentMark::checkpointRootsFinalWork() {
2584 ResourceMark rm;
2585 HandleMark hm;
2586 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2587
2588 G1CMTraceTime trace("Finalize Marking", G1Log::finer());
2589
2590 g1h->ensure_parsability(false);
2591
2592 // this is remark, so we'll use up all active threads
2593 uint active_workers = g1h->workers()->active_workers();
2594 set_concurrency_and_phase(active_workers, false /* concurrent */);
2595 // Leave _parallel_marking_threads at it's
2596 // value originally calculated in the ConcurrentMark
2597 // constructor and pass values of the active workers
2598 // through the gang in the task.
2599
2600 {
2601 StrongRootsScope srs(active_workers);
2602
2603 CMRemarkTask remarkTask(this, active_workers);
2604 // We will start all available threads, even if we decide that the
2605 // active_workers will be fewer. The extra ones will just bail out
2606 // immediately.
2607 g1h->workers()->run_task(&remarkTask);
2608 }
2609
2610 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2611 guarantee(has_overflown() ||
2612 satb_mq_set.completed_buffers_num() == 0,
2613 err_msg("Invariant: has_overflown = %s, num buffers = %d",
2614 BOOL_TO_STR(has_overflown()),
2615 satb_mq_set.completed_buffers_num()));
2616
2617 print_stats();
2618 }
2619
2620 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2621 // Note we are overriding the read-only view of the prev map here, via
2622 // the cast.
2623 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2624 }
2625
2626 void ConcurrentMark::clearRangeNextBitmap(MemRegion mr) {
2627 _nextMarkBitMap->clearRange(mr);
2628 }
2629
2630 HeapRegion*
2631 ConcurrentMark::claim_region(uint worker_id) {
2632 // "checkpoint" the finger
2633 HeapWord* finger = _finger;
2634
2635 // _heap_end will not change underneath our feet; it only changes at
2636 // yield points.
2637 while (finger < _heap_end) {
2638 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2639
2640 // Note on how this code handles humongous regions. In the
2641 // normal case the finger will reach the start of a "starts
2642 // humongous" (SH) region. Its end will either be the end of the
2643 // last "continues humongous" (CH) region in the sequence, or the
2644 // standard end of the SH region (if the SH is the only region in
2645 // the sequence). That way claim_region() will skip over the CH
2646 // regions. However, there is a subtle race between a CM thread
2647 // executing this method and a mutator thread doing a humongous
2648 // object allocation. The two are not mutually exclusive as the CM
2649 // thread does not need to hold the Heap_lock when it gets
2650 // here. So there is a chance that claim_region() will come across
2651 // a free region that's in the progress of becoming a SH or a CH
2652 // region. In the former case, it will either
2653 // a) Miss the update to the region's end, in which case it will
2654 // visit every subsequent CH region, will find their bitmaps
2655 // empty, and do nothing, or
2656 // b) Will observe the update of the region's end (in which case
2657 // it will skip the subsequent CH regions).
2658 // If it comes across a region that suddenly becomes CH, the
2659 // scenario will be similar to b). So, the race between
2660 // claim_region() and a humongous object allocation might force us
2661 // to do a bit of unnecessary work (due to some unnecessary bitmap
2662 // iterations) but it should not introduce and correctness issues.
2663 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger);
2664
2665 // Above heap_region_containing_raw may return NULL as we always scan claim
2666 // until the end of the heap. In this case, just jump to the next region.
2667 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords;
2668
2669 // Is the gap between reading the finger and doing the CAS too long?
2670 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2671 if (res == finger && curr_region != NULL) {
2672 // we succeeded
2673 HeapWord* bottom = curr_region->bottom();
2674 HeapWord* limit = curr_region->next_top_at_mark_start();
2675
2676 if (verbose_low()) {
2677 gclog_or_tty->print_cr("[%u] curr_region = "PTR_FORMAT" "
2678 "["PTR_FORMAT", "PTR_FORMAT"), "
2679 "limit = "PTR_FORMAT,
2680 worker_id, p2i(curr_region), p2i(bottom), p2i(end), p2i(limit));
2681 }
2682
2683 // notice that _finger == end cannot be guaranteed here since,
2684 // someone else might have moved the finger even further
2685 assert(_finger >= end, "the finger should have moved forward");
2686
2687 if (verbose_low()) {
2688 gclog_or_tty->print_cr("[%u] we were successful with region = "
2689 PTR_FORMAT, worker_id, p2i(curr_region));
2690 }
2691
2692 if (limit > bottom) {
2693 if (verbose_low()) {
2694 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is not empty, "
2695 "returning it ", worker_id, p2i(curr_region));
2696 }
2697 return curr_region;
2698 } else {
2699 assert(limit == bottom,
2700 "the region limit should be at bottom");
2701 if (verbose_low()) {
2702 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is empty, "
2703 "returning NULL", worker_id, p2i(curr_region));
2704 }
2705 // we return NULL and the caller should try calling
2706 // claim_region() again.
2707 return NULL;
2708 }
2709 } else {
2710 assert(_finger > finger, "the finger should have moved forward");
2711 if (verbose_low()) {
2712 if (curr_region == NULL) {
2713 gclog_or_tty->print_cr("[%u] found uncommitted region, moving finger, "
2714 "global finger = "PTR_FORMAT", "
2715 "our finger = "PTR_FORMAT,
2716 worker_id, p2i(_finger), p2i(finger));
2717 } else {
2718 gclog_or_tty->print_cr("[%u] somebody else moved the finger, "
2719 "global finger = "PTR_FORMAT", "
2720 "our finger = "PTR_FORMAT,
2721 worker_id, p2i(_finger), p2i(finger));
2722 }
2723 }
2724
2725 // read it again
2726 finger = _finger;
2727 }
2728 }
2729
2730 return NULL;
2731 }
2732
2733 #ifndef PRODUCT
2734 enum VerifyNoCSetOopsPhase {
2735 VerifyNoCSetOopsStack,
2736 VerifyNoCSetOopsQueues
2737 };
2738
2739 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure {
2740 private:
2741 G1CollectedHeap* _g1h;
2742 VerifyNoCSetOopsPhase _phase;
2743 int _info;
2744
2745 const char* phase_str() {
2746 switch (_phase) {
2747 case VerifyNoCSetOopsStack: return "Stack";
2748 case VerifyNoCSetOopsQueues: return "Queue";
2749 default: ShouldNotReachHere();
2750 }
2751 return NULL;
2752 }
2753
2754 void do_object_work(oop obj) {
2755 guarantee(!_g1h->obj_in_cs(obj),
2756 err_msg("obj: "PTR_FORMAT" in CSet, phase: %s, info: %d",
2757 p2i((void*) obj), phase_str(), _info));
2758 }
2759
2760 public:
2761 VerifyNoCSetOopsClosure() : _g1h(G1CollectedHeap::heap()) { }
2762
2763 void set_phase(VerifyNoCSetOopsPhase phase, int info = -1) {
2764 _phase = phase;
2765 _info = info;
2766 }
2767
2768 virtual void do_oop(oop* p) {
2769 oop obj = oopDesc::load_decode_heap_oop(p);
2770 do_object_work(obj);
2771 }
2772
2773 virtual void do_oop(narrowOop* p) {
2774 // We should not come across narrow oops while scanning marking
2775 // stacks
2776 ShouldNotReachHere();
2777 }
2778
2779 virtual void do_object(oop obj) {
2780 do_object_work(obj);
2781 }
2782 };
2783
2784 void ConcurrentMark::verify_no_cset_oops() {
2785 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2786 if (!G1CollectedHeap::heap()->mark_in_progress()) {
2787 return;
2788 }
2789
2790 VerifyNoCSetOopsClosure cl;
2791
2792 // Verify entries on the global mark stack
2793 cl.set_phase(VerifyNoCSetOopsStack);
2794 _markStack.oops_do(&cl);
2795
2796 // Verify entries on the task queues
2797 for (uint i = 0; i < _max_worker_id; i += 1) {
2798 cl.set_phase(VerifyNoCSetOopsQueues, i);
2799 CMTaskQueue* queue = _task_queues->queue(i);
2800 queue->oops_do(&cl);
2801 }
2802
2803 // Verify the global finger
2804 HeapWord* global_finger = finger();
2805 if (global_finger != NULL && global_finger < _heap_end) {
2806 // The global finger always points to a heap region boundary. We
2807 // use heap_region_containing_raw() to get the containing region
2808 // given that the global finger could be pointing to a free region
2809 // which subsequently becomes continues humongous. If that
2810 // happens, heap_region_containing() will return the bottom of the
2811 // corresponding starts humongous region and the check below will
2812 // not hold any more.
2813 // Since we always iterate over all regions, we might get a NULL HeapRegion
2814 // here.
2815 HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger);
2816 guarantee(global_hr == NULL || global_finger == global_hr->bottom(),
2817 err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT,
2818 p2i(global_finger), HR_FORMAT_PARAMS(global_hr)));
2819 }
2820
2821 // Verify the task fingers
2822 assert(parallel_marking_threads() <= _max_worker_id, "sanity");
2823 for (int i = 0; i < (int) parallel_marking_threads(); i += 1) {
2824 CMTask* task = _tasks[i];
2825 HeapWord* task_finger = task->finger();
2826 if (task_finger != NULL && task_finger < _heap_end) {
2827 // See above note on the global finger verification.
2828 HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger);
2829 guarantee(task_hr == NULL || task_finger == task_hr->bottom() ||
2830 !task_hr->in_collection_set(),
2831 err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT,
2832 p2i(task_finger), HR_FORMAT_PARAMS(task_hr)));
2833 }
2834 }
2835 }
2836 #endif // PRODUCT
2837
2838 // Aggregate the counting data that was constructed concurrently
2839 // with marking.
2840 class AggregateCountDataHRClosure: public HeapRegionClosure {
2841 G1CollectedHeap* _g1h;
2842 ConcurrentMark* _cm;
2843 CardTableModRefBS* _ct_bs;
2844 BitMap* _cm_card_bm;
2845 uint _max_worker_id;
2846
2847 public:
2848 AggregateCountDataHRClosure(G1CollectedHeap* g1h,
2849 BitMap* cm_card_bm,
2850 uint max_worker_id) :
2851 _g1h(g1h), _cm(g1h->concurrent_mark()),
2852 _ct_bs(barrier_set_cast<CardTableModRefBS>(g1h->barrier_set())),
2853 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { }
2854
2855 bool doHeapRegion(HeapRegion* hr) {
2856 if (hr->is_continues_humongous()) {
2857 // We will ignore these here and process them when their
2858 // associated "starts humongous" region is processed.
2859 // Note that we cannot rely on their associated
2860 // "starts humongous" region to have their bit set to 1
2861 // since, due to the region chunking in the parallel region
2862 // iteration, a "continues humongous" region might be visited
2863 // before its associated "starts humongous".
2864 return false;
2865 }
2866
2867 HeapWord* start = hr->bottom();
2868 HeapWord* limit = hr->next_top_at_mark_start();
2869 HeapWord* end = hr->end();
2870
2871 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(),
2872 err_msg("Preconditions not met - "
2873 "start: "PTR_FORMAT", limit: "PTR_FORMAT", "
2874 "top: "PTR_FORMAT", end: "PTR_FORMAT,
2875 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end())));
2876
2877 assert(hr->next_marked_bytes() == 0, "Precondition");
2878
2879 if (start == limit) {
2880 // NTAMS of this region has not been set so nothing to do.
2881 return false;
2882 }
2883
2884 // 'start' should be in the heap.
2885 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity");
2886 // 'end' *may* be just beyond the end of the heap (if hr is the last region)
2887 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity");
2888
2889 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
2890 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit);
2891 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end);
2892
2893 // If ntams is not card aligned then we bump card bitmap index
2894 // for limit so that we get the all the cards spanned by
2895 // the object ending at ntams.
2896 // Note: if this is the last region in the heap then ntams
2897 // could be actually just beyond the end of the the heap;
2898 // limit_idx will then correspond to a (non-existent) card
2899 // that is also outside the heap.
2900 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) {
2901 limit_idx += 1;
2902 }
2903
2904 assert(limit_idx <= end_idx, "or else use atomics");
2905
2906 // Aggregate the "stripe" in the count data associated with hr.
2907 uint hrm_index = hr->hrm_index();
2908 size_t marked_bytes = 0;
2909
2910 for (uint i = 0; i < _max_worker_id; i += 1) {
2911 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i);
2912 BitMap* task_card_bm = _cm->count_card_bitmap_for(i);
2913
2914 // Fetch the marked_bytes in this region for task i and
2915 // add it to the running total for this region.
2916 marked_bytes += marked_bytes_array[hrm_index];
2917
2918 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx)
2919 // into the global card bitmap.
2920 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx);
2921
2922 while (scan_idx < limit_idx) {
2923 assert(task_card_bm->at(scan_idx) == true, "should be");
2924 _cm_card_bm->set_bit(scan_idx);
2925 assert(_cm_card_bm->at(scan_idx) == true, "should be");
2926
2927 // BitMap::get_next_one_offset() can handle the case when
2928 // its left_offset parameter is greater than its right_offset
2929 // parameter. It does, however, have an early exit if
2930 // left_offset == right_offset. So let's limit the value
2931 // passed in for left offset here.
2932 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx);
2933 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx);
2934 }
2935 }
2936
2937 // Update the marked bytes for this region.
2938 hr->add_to_marked_bytes(marked_bytes);
2939
2940 // Next heap region
2941 return false;
2942 }
2943 };
2944
2945 class G1AggregateCountDataTask: public AbstractGangTask {
2946 protected:
2947 G1CollectedHeap* _g1h;
2948 ConcurrentMark* _cm;
2949 BitMap* _cm_card_bm;
2950 uint _max_worker_id;
2951 uint _active_workers;
2952 HeapRegionClaimer _hrclaimer;
2953
2954 public:
2955 G1AggregateCountDataTask(G1CollectedHeap* g1h,
2956 ConcurrentMark* cm,
2957 BitMap* cm_card_bm,
2958 uint max_worker_id,
2959 uint n_workers) :
2960 AbstractGangTask("Count Aggregation"),
2961 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm),
2962 _max_worker_id(max_worker_id),
2963 _active_workers(n_workers),
2964 _hrclaimer(_active_workers) {
2965 }
2966
2967 void work(uint worker_id) {
2968 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id);
2969
2970 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer);
2971 }
2972 };
2973
2974
2975 void ConcurrentMark::aggregate_count_data() {
2976 uint n_workers = _g1h->workers()->active_workers();
2977
2978 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm,
2979 _max_worker_id, n_workers);
2980
2981 _g1h->workers()->run_task(&g1_par_agg_task);
2982 }
2983
2984 // Clear the per-worker arrays used to store the per-region counting data
2985 void ConcurrentMark::clear_all_count_data() {
2986 // Clear the global card bitmap - it will be filled during
2987 // liveness count aggregation (during remark) and the
2988 // final counting task.
2989 _card_bm.clear();
2990
2991 // Clear the global region bitmap - it will be filled as part
2992 // of the final counting task.
2993 _region_bm.clear();
2994
2995 uint max_regions = _g1h->max_regions();
2996 assert(_max_worker_id > 0, "uninitialized");
2997
2998 for (uint i = 0; i < _max_worker_id; i += 1) {
2999 BitMap* task_card_bm = count_card_bitmap_for(i);
3000 size_t* marked_bytes_array = count_marked_bytes_array_for(i);
3001
3002 assert(task_card_bm->size() == _card_bm.size(), "size mismatch");
3003 assert(marked_bytes_array != NULL, "uninitialized");
3004
3005 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t));
3006 task_card_bm->clear();
3007 }
3008 }
3009
3010 void ConcurrentMark::print_stats() {
3011 if (verbose_stats()) {
3012 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3013 for (size_t i = 0; i < _active_tasks; ++i) {
3014 _tasks[i]->print_stats();
3015 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3016 }
3017 }
3018 }
3019
3020 // abandon current marking iteration due to a Full GC
3021 void ConcurrentMark::abort() {
3022 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next
3023 // concurrent bitmap clearing.
3024 _nextMarkBitMap->clearAll();
3025
3026 // Note we cannot clear the previous marking bitmap here
3027 // since VerifyDuringGC verifies the objects marked during
3028 // a full GC against the previous bitmap.
3029
3030 // Clear the liveness counting data
3031 clear_all_count_data();
3032 // Empty mark stack
3033 reset_marking_state();
3034 for (uint i = 0; i < _max_worker_id; ++i) {
3035 _tasks[i]->clear_region_fields();
3036 }
3037 _first_overflow_barrier_sync.abort();
3038 _second_overflow_barrier_sync.abort();
3039 const GCId& gc_id = _g1h->gc_tracer_cm()->gc_id();
3040 if (!gc_id.is_undefined()) {
3041 // We can do multiple full GCs before ConcurrentMarkThread::run() gets a chance
3042 // to detect that it was aborted. Only keep track of the first GC id that we aborted.
3043 _aborted_gc_id = gc_id;
3044 }
3045 _has_aborted = true;
3046
3047 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3048 satb_mq_set.abandon_partial_marking();
3049 // This can be called either during or outside marking, we'll read
3050 // the expected_active value from the SATB queue set.
3051 satb_mq_set.set_active_all_threads(
3052 false, /* new active value */
3053 satb_mq_set.is_active() /* expected_active */);
3054
3055 _g1h->trace_heap_after_concurrent_cycle();
3056 _g1h->register_concurrent_cycle_end();
3057 }
3058
3059 const GCId& ConcurrentMark::concurrent_gc_id() {
3060 if (has_aborted()) {
3061 return _aborted_gc_id;
3062 }
3063 return _g1h->gc_tracer_cm()->gc_id();
3064 }
3065
3066 static void print_ms_time_info(const char* prefix, const char* name,
3067 NumberSeq& ns) {
3068 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
3069 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
3070 if (ns.num() > 0) {
3071 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]",
3072 prefix, ns.sd(), ns.maximum());
3073 }
3074 }
3075
3076 void ConcurrentMark::print_summary_info() {
3077 gclog_or_tty->print_cr(" Concurrent marking:");
3078 print_ms_time_info(" ", "init marks", _init_times);
3079 print_ms_time_info(" ", "remarks", _remark_times);
3080 {
3081 print_ms_time_info(" ", "final marks", _remark_mark_times);
3082 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
3083
3084 }
3085 print_ms_time_info(" ", "cleanups", _cleanup_times);
3086 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).",
3087 _total_counting_time,
3088 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 /
3089 (double)_cleanup_times.num()
3090 : 0.0));
3091 if (G1ScrubRemSets) {
3092 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
3093 _total_rs_scrub_time,
3094 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 /
3095 (double)_cleanup_times.num()
3096 : 0.0));
3097 }
3098 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.",
3099 (_init_times.sum() + _remark_times.sum() +
3100 _cleanup_times.sum())/1000.0);
3101 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s "
3102 "(%8.2f s marking).",
3103 cmThread()->vtime_accum(),
3104 cmThread()->vtime_mark_accum());
3105 }
3106
3107 void ConcurrentMark::print_worker_threads_on(outputStream* st) const {
3108 _parallel_workers->print_worker_threads_on(st);
3109 }
3110
3111 void ConcurrentMark::print_on_error(outputStream* st) const {
3112 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
3113 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap));
3114 _prevMarkBitMap->print_on_error(st, " Prev Bits: ");
3115 _nextMarkBitMap->print_on_error(st, " Next Bits: ");
3116 }
3117
3118 // We take a break if someone is trying to stop the world.
3119 bool ConcurrentMark::do_yield_check(uint worker_id) {
3120 if (SuspendibleThreadSet::should_yield()) {
3121 if (worker_id == 0) {
3122 _g1h->g1_policy()->record_concurrent_pause();
3123 }
3124 SuspendibleThreadSet::yield();
3125 return true;
3126 } else {
3127 return false;
3128 }
3129 }
3130
3131 #ifndef PRODUCT
3132 // for debugging purposes
3133 void ConcurrentMark::print_finger() {
3134 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT,
3135 p2i(_heap_start), p2i(_heap_end), p2i(_finger));
3136 for (uint i = 0; i < _max_worker_id; ++i) {
3137 gclog_or_tty->print(" %u: " PTR_FORMAT, i, p2i(_tasks[i]->finger()));
3138 }
3139 gclog_or_tty->cr();
3140 }
3141 #endif
3142
3143 template<bool scan>
3144 inline void CMTask::process_grey_object(oop obj) {
3145 assert(scan || obj->is_typeArray(), "Skipping scan of grey non-typeArray");
3146 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
3147
3148 if (_cm->verbose_high()) {
3149 gclog_or_tty->print_cr("[%u] processing grey object " PTR_FORMAT,
3150 _worker_id, p2i((void*) obj));
3151 }
3152
3153 size_t obj_size = obj->size();
3154 _words_scanned += obj_size;
3155
3156 if (scan) {
3157 obj->oop_iterate(_cm_oop_closure);
3158 }
3159 statsOnly( ++_objs_scanned );
3160 check_limits();
3161 }
3162
3163 template void CMTask::process_grey_object<true>(oop);
3164 template void CMTask::process_grey_object<false>(oop);
3165
3166 // Closure for iteration over bitmaps
3167 class CMBitMapClosure : public BitMapClosure {
3168 private:
3169 // the bitmap that is being iterated over
3170 CMBitMap* _nextMarkBitMap;
3171 ConcurrentMark* _cm;
3172 CMTask* _task;
3173
3174 public:
3175 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) :
3176 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
3177
3178 bool do_bit(size_t offset) {
3179 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
3180 assert(_nextMarkBitMap->isMarked(addr), "invariant");
3181 assert( addr < _cm->finger(), "invariant");
3182
3183 statsOnly( _task->increase_objs_found_on_bitmap() );
3184 assert(addr >= _task->finger(), "invariant");
3185
3186 // We move that task's local finger along.
3187 _task->move_finger_to(addr);
3188
3189 _task->scan_object(oop(addr));
3190 // we only partially drain the local queue and global stack
3191 _task->drain_local_queue(true);
3192 _task->drain_global_stack(true);
3193
3194 // if the has_aborted flag has been raised, we need to bail out of
3195 // the iteration
3196 return !_task->has_aborted();
3197 }
3198 };
3199
3200 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
3201 ConcurrentMark* cm,
3202 CMTask* task)
3203 : _g1h(g1h), _cm(cm), _task(task) {
3204 assert(_ref_processor == NULL, "should be initialized to NULL");
3205
3206 if (G1UseConcMarkReferenceProcessing) {
3207 _ref_processor = g1h->ref_processor_cm();
3208 assert(_ref_processor != NULL, "should not be NULL");
3209 }
3210 }
3211
3212 void CMTask::setup_for_region(HeapRegion* hr) {
3213 assert(hr != NULL,
3214 "claim_region() should have filtered out NULL regions");
3215 assert(!hr->is_continues_humongous(),
3216 "claim_region() should have filtered out continues humongous regions");
3217
3218 if (_cm->verbose_low()) {
3219 gclog_or_tty->print_cr("[%u] setting up for region "PTR_FORMAT,
3220 _worker_id, p2i(hr));
3221 }
3222
3223 _curr_region = hr;
3224 _finger = hr->bottom();
3225 update_region_limit();
3226 }
3227
3228 void CMTask::update_region_limit() {
3229 HeapRegion* hr = _curr_region;
3230 HeapWord* bottom = hr->bottom();
3231 HeapWord* limit = hr->next_top_at_mark_start();
3232
3233 if (limit == bottom) {
3234 if (_cm->verbose_low()) {
3235 gclog_or_tty->print_cr("[%u] found an empty region "
3236 "["PTR_FORMAT", "PTR_FORMAT")",
3237 _worker_id, p2i(bottom), p2i(limit));
3238 }
3239 // The region was collected underneath our feet.
3240 // We set the finger to bottom to ensure that the bitmap
3241 // iteration that will follow this will not do anything.
3242 // (this is not a condition that holds when we set the region up,
3243 // as the region is not supposed to be empty in the first place)
3244 _finger = bottom;
3245 } else if (limit >= _region_limit) {
3246 assert(limit >= _finger, "peace of mind");
3247 } else {
3248 assert(limit < _region_limit, "only way to get here");
3249 // This can happen under some pretty unusual circumstances. An
3250 // evacuation pause empties the region underneath our feet (NTAMS
3251 // at bottom). We then do some allocation in the region (NTAMS
3252 // stays at bottom), followed by the region being used as a GC
3253 // alloc region (NTAMS will move to top() and the objects
3254 // originally below it will be grayed). All objects now marked in
3255 // the region are explicitly grayed, if below the global finger,
3256 // and we do not need in fact to scan anything else. So, we simply
3257 // set _finger to be limit to ensure that the bitmap iteration
3258 // doesn't do anything.
3259 _finger = limit;
3260 }
3261
3262 _region_limit = limit;
3263 }
3264
3265 void CMTask::giveup_current_region() {
3266 assert(_curr_region != NULL, "invariant");
3267 if (_cm->verbose_low()) {
3268 gclog_or_tty->print_cr("[%u] giving up region "PTR_FORMAT,
3269 _worker_id, p2i(_curr_region));
3270 }
3271 clear_region_fields();
3272 }
3273
3274 void CMTask::clear_region_fields() {
3275 // Values for these three fields that indicate that we're not
3276 // holding on to a region.
3277 _curr_region = NULL;
3278 _finger = NULL;
3279 _region_limit = NULL;
3280 }
3281
3282 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
3283 if (cm_oop_closure == NULL) {
3284 assert(_cm_oop_closure != NULL, "invariant");
3285 } else {
3286 assert(_cm_oop_closure == NULL, "invariant");
3287 }
3288 _cm_oop_closure = cm_oop_closure;
3289 }
3290
3291 void CMTask::reset(CMBitMap* nextMarkBitMap) {
3292 guarantee(nextMarkBitMap != NULL, "invariant");
3293
3294 if (_cm->verbose_low()) {
3295 gclog_or_tty->print_cr("[%u] resetting", _worker_id);
3296 }
3297
3298 _nextMarkBitMap = nextMarkBitMap;
3299 clear_region_fields();
3300
3301 _calls = 0;
3302 _elapsed_time_ms = 0.0;
3303 _termination_time_ms = 0.0;
3304 _termination_start_time_ms = 0.0;
3305
3306 #if _MARKING_STATS_
3307 _aborted = 0;
3308 _aborted_overflow = 0;
3309 _aborted_cm_aborted = 0;
3310 _aborted_yield = 0;
3311 _aborted_timed_out = 0;
3312 _aborted_satb = 0;
3313 _aborted_termination = 0;
3314 _steal_attempts = 0;
3315 _steals = 0;
3316 _local_pushes = 0;
3317 _local_pops = 0;
3318 _local_max_size = 0;
3319 _objs_scanned = 0;
3320 _global_pushes = 0;
3321 _global_pops = 0;
3322 _global_max_size = 0;
3323 _global_transfers_to = 0;
3324 _global_transfers_from = 0;
3325 _regions_claimed = 0;
3326 _objs_found_on_bitmap = 0;
3327 _satb_buffers_processed = 0;
3328 #endif // _MARKING_STATS_
3329 }
3330
3331 bool CMTask::should_exit_termination() {
3332 regular_clock_call();
3333 // This is called when we are in the termination protocol. We should
3334 // quit if, for some reason, this task wants to abort or the global
3335 // stack is not empty (this means that we can get work from it).
3336 return !_cm->mark_stack_empty() || has_aborted();
3337 }
3338
3339 void CMTask::reached_limit() {
3340 assert(_words_scanned >= _words_scanned_limit ||
3341 _refs_reached >= _refs_reached_limit ,
3342 "shouldn't have been called otherwise");
3343 regular_clock_call();
3344 }
3345
3346 void CMTask::regular_clock_call() {
3347 if (has_aborted()) return;
3348
3349 // First, we need to recalculate the words scanned and refs reached
3350 // limits for the next clock call.
3351 recalculate_limits();
3352
3353 // During the regular clock call we do the following
3354
3355 // (1) If an overflow has been flagged, then we abort.
3356 if (_cm->has_overflown()) {
3357 set_has_aborted();
3358 return;
3359 }
3360
3361 // If we are not concurrent (i.e. we're doing remark) we don't need
3362 // to check anything else. The other steps are only needed during
3363 // the concurrent marking phase.
3364 if (!concurrent()) return;
3365
3366 // (2) If marking has been aborted for Full GC, then we also abort.
3367 if (_cm->has_aborted()) {
3368 set_has_aborted();
3369 statsOnly( ++_aborted_cm_aborted );
3370 return;
3371 }
3372
3373 double curr_time_ms = os::elapsedVTime() * 1000.0;
3374
3375 // (3) If marking stats are enabled, then we update the step history.
3376 #if _MARKING_STATS_
3377 if (_words_scanned >= _words_scanned_limit) {
3378 ++_clock_due_to_scanning;
3379 }
3380 if (_refs_reached >= _refs_reached_limit) {
3381 ++_clock_due_to_marking;
3382 }
3383
3384 double last_interval_ms = curr_time_ms - _interval_start_time_ms;
3385 _interval_start_time_ms = curr_time_ms;
3386 _all_clock_intervals_ms.add(last_interval_ms);
3387
3388 if (_cm->verbose_medium()) {
3389 gclog_or_tty->print_cr("[%u] regular clock, interval = %1.2lfms, "
3390 "scanned = "SIZE_FORMAT"%s, refs reached = "SIZE_FORMAT"%s",
3391 _worker_id, last_interval_ms,
3392 _words_scanned,
3393 (_words_scanned >= _words_scanned_limit) ? " (*)" : "",
3394 _refs_reached,
3395 (_refs_reached >= _refs_reached_limit) ? " (*)" : "");
3396 }
3397 #endif // _MARKING_STATS_
3398
3399 // (4) We check whether we should yield. If we have to, then we abort.
3400 if (SuspendibleThreadSet::should_yield()) {
3401 // We should yield. To do this we abort the task. The caller is
3402 // responsible for yielding.
3403 set_has_aborted();
3404 statsOnly( ++_aborted_yield );
3405 return;
3406 }
3407
3408 // (5) We check whether we've reached our time quota. If we have,
3409 // then we abort.
3410 double elapsed_time_ms = curr_time_ms - _start_time_ms;
3411 if (elapsed_time_ms > _time_target_ms) {
3412 set_has_aborted();
3413 _has_timed_out = true;
3414 statsOnly( ++_aborted_timed_out );
3415 return;
3416 }
3417
3418 // (6) Finally, we check whether there are enough completed STAB
3419 // buffers available for processing. If there are, we abort.
3420 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3421 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
3422 if (_cm->verbose_low()) {
3423 gclog_or_tty->print_cr("[%u] aborting to deal with pending SATB buffers",
3424 _worker_id);
3425 }
3426 // we do need to process SATB buffers, we'll abort and restart
3427 // the marking task to do so
3428 set_has_aborted();
3429 statsOnly( ++_aborted_satb );
3430 return;
3431 }
3432 }
3433
3434 void CMTask::recalculate_limits() {
3435 _real_words_scanned_limit = _words_scanned + words_scanned_period;
3436 _words_scanned_limit = _real_words_scanned_limit;
3437
3438 _real_refs_reached_limit = _refs_reached + refs_reached_period;
3439 _refs_reached_limit = _real_refs_reached_limit;
3440 }
3441
3442 void CMTask::decrease_limits() {
3443 // This is called when we believe that we're going to do an infrequent
3444 // operation which will increase the per byte scanned cost (i.e. move
3445 // entries to/from the global stack). It basically tries to decrease the
3446 // scanning limit so that the clock is called earlier.
3447
3448 if (_cm->verbose_medium()) {
3449 gclog_or_tty->print_cr("[%u] decreasing limits", _worker_id);
3450 }
3451
3452 _words_scanned_limit = _real_words_scanned_limit -
3453 3 * words_scanned_period / 4;
3454 _refs_reached_limit = _real_refs_reached_limit -
3455 3 * refs_reached_period / 4;
3456 }
3457
3458 void CMTask::move_entries_to_global_stack() {
3459 // local array where we'll store the entries that will be popped
3460 // from the local queue
3461 oop buffer[global_stack_transfer_size];
3462
3463 int n = 0;
3464 oop obj;
3465 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
3466 buffer[n] = obj;
3467 ++n;
3468 }
3469
3470 if (n > 0) {
3471 // we popped at least one entry from the local queue
3472
3473 statsOnly( ++_global_transfers_to; _local_pops += n );
3474
3475 if (!_cm->mark_stack_push(buffer, n)) {
3476 if (_cm->verbose_low()) {
3477 gclog_or_tty->print_cr("[%u] aborting due to global stack overflow",
3478 _worker_id);
3479 }
3480 set_has_aborted();
3481 } else {
3482 // the transfer was successful
3483
3484 if (_cm->verbose_medium()) {
3485 gclog_or_tty->print_cr("[%u] pushed %d entries to the global stack",
3486 _worker_id, n);
3487 }
3488 statsOnly( size_t tmp_size = _cm->mark_stack_size();
3489 if (tmp_size > _global_max_size) {
3490 _global_max_size = tmp_size;
3491 }
3492 _global_pushes += n );
3493 }
3494 }
3495
3496 // this operation was quite expensive, so decrease the limits
3497 decrease_limits();
3498 }
3499
3500 void CMTask::get_entries_from_global_stack() {
3501 // local array where we'll store the entries that will be popped
3502 // from the global stack.
3503 oop buffer[global_stack_transfer_size];
3504 int n;
3505 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
3506 assert(n <= global_stack_transfer_size,
3507 "we should not pop more than the given limit");
3508 if (n > 0) {
3509 // yes, we did actually pop at least one entry
3510
3511 statsOnly( ++_global_transfers_from; _global_pops += n );
3512 if (_cm->verbose_medium()) {
3513 gclog_or_tty->print_cr("[%u] popped %d entries from the global stack",
3514 _worker_id, n);
3515 }
3516 for (int i = 0; i < n; ++i) {
3517 bool success = _task_queue->push(buffer[i]);
3518 // We only call this when the local queue is empty or under a
3519 // given target limit. So, we do not expect this push to fail.
3520 assert(success, "invariant");
3521 }
3522
3523 statsOnly( size_t tmp_size = (size_t)_task_queue->size();
3524 if (tmp_size > _local_max_size) {
3525 _local_max_size = tmp_size;
3526 }
3527 _local_pushes += n );
3528 }
3529
3530 // this operation was quite expensive, so decrease the limits
3531 decrease_limits();
3532 }
3533
3534 void CMTask::drain_local_queue(bool partially) {
3535 if (has_aborted()) return;
3536
3537 // Decide what the target size is, depending whether we're going to
3538 // drain it partially (so that other tasks can steal if they run out
3539 // of things to do) or totally (at the very end).
3540 size_t target_size;
3541 if (partially) {
3542 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
3543 } else {
3544 target_size = 0;
3545 }
3546
3547 if (_task_queue->size() > target_size) {
3548 if (_cm->verbose_high()) {
3549 gclog_or_tty->print_cr("[%u] draining local queue, target size = " SIZE_FORMAT,
3550 _worker_id, target_size);
3551 }
3552
3553 oop obj;
3554 bool ret = _task_queue->pop_local(obj);
3555 while (ret) {
3556 statsOnly( ++_local_pops );
3557
3558 if (_cm->verbose_high()) {
3559 gclog_or_tty->print_cr("[%u] popped "PTR_FORMAT, _worker_id,
3560 p2i((void*) obj));
3561 }
3562
3563 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
3564 assert(!_g1h->is_on_master_free_list(
3565 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
3566
3567 scan_object(obj);
3568
3569 if (_task_queue->size() <= target_size || has_aborted()) {
3570 ret = false;
3571 } else {
3572 ret = _task_queue->pop_local(obj);
3573 }
3574 }
3575
3576 if (_cm->verbose_high()) {
3577 gclog_or_tty->print_cr("[%u] drained local queue, size = %u",
3578 _worker_id, _task_queue->size());
3579 }
3580 }
3581 }
3582
3583 void CMTask::drain_global_stack(bool partially) {
3584 if (has_aborted()) return;
3585
3586 // We have a policy to drain the local queue before we attempt to
3587 // drain the global stack.
3588 assert(partially || _task_queue->size() == 0, "invariant");
3589
3590 // Decide what the target size is, depending whether we're going to
3591 // drain it partially (so that other tasks can steal if they run out
3592 // of things to do) or totally (at the very end). Notice that,
3593 // because we move entries from the global stack in chunks or
3594 // because another task might be doing the same, we might in fact
3595 // drop below the target. But, this is not a problem.
3596 size_t target_size;
3597 if (partially) {
3598 target_size = _cm->partial_mark_stack_size_target();
3599 } else {
3600 target_size = 0;
3601 }
3602
3603 if (_cm->mark_stack_size() > target_size) {
3604 if (_cm->verbose_low()) {
3605 gclog_or_tty->print_cr("[%u] draining global_stack, target size " SIZE_FORMAT,
3606 _worker_id, target_size);
3607 }
3608
3609 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3610 get_entries_from_global_stack();
3611 drain_local_queue(partially);
3612 }
3613
3614 if (_cm->verbose_low()) {
3615 gclog_or_tty->print_cr("[%u] drained global stack, size = " SIZE_FORMAT,
3616 _worker_id, _cm->mark_stack_size());
3617 }
3618 }
3619 }
3620
3621 // SATB Queue has several assumptions on whether to call the par or
3622 // non-par versions of the methods. this is why some of the code is
3623 // replicated. We should really get rid of the single-threaded version
3624 // of the code to simplify things.
3625 void CMTask::drain_satb_buffers() {
3626 if (has_aborted()) return;
3627
3628 // We set this so that the regular clock knows that we're in the
3629 // middle of draining buffers and doesn't set the abort flag when it
3630 // notices that SATB buffers are available for draining. It'd be
3631 // very counter productive if it did that. :-)
3632 _draining_satb_buffers = true;
3633
3634 CMSATBBufferClosure satb_cl(this, _g1h);
3635 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3636
3637 // This keeps claiming and applying the closure to completed buffers
3638 // until we run out of buffers or we need to abort.
3639 while (!has_aborted() &&
3640 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) {
3641 if (_cm->verbose_medium()) {
3642 gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
3643 }
3644 statsOnly( ++_satb_buffers_processed );
3645 regular_clock_call();
3646 }
3647
3648 _draining_satb_buffers = false;
3649
3650 assert(has_aborted() ||
3651 concurrent() ||
3652 satb_mq_set.completed_buffers_num() == 0, "invariant");
3653
3654 // again, this was a potentially expensive operation, decrease the
3655 // limits to get the regular clock call early
3656 decrease_limits();
3657 }
3658
3659 void CMTask::print_stats() {
3660 gclog_or_tty->print_cr("Marking Stats, task = %u, calls = %d",
3661 _worker_id, _calls);
3662 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
3663 _elapsed_time_ms, _termination_time_ms);
3664 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3665 _step_times_ms.num(), _step_times_ms.avg(),
3666 _step_times_ms.sd());
3667 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3668 _step_times_ms.maximum(), _step_times_ms.sum());
3669
3670 #if _MARKING_STATS_
3671 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3672 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(),
3673 _all_clock_intervals_ms.sd());
3674 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3675 _all_clock_intervals_ms.maximum(),
3676 _all_clock_intervals_ms.sum());
3677 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = " SIZE_FORMAT ", marking = " SIZE_FORMAT,
3678 _clock_due_to_scanning, _clock_due_to_marking);
3679 gclog_or_tty->print_cr(" Objects: scanned = " SIZE_FORMAT ", found on the bitmap = " SIZE_FORMAT,
3680 _objs_scanned, _objs_found_on_bitmap);
3681 gclog_or_tty->print_cr(" Local Queue: pushes = " SIZE_FORMAT ", pops = " SIZE_FORMAT ", max size = " SIZE_FORMAT,
3682 _local_pushes, _local_pops, _local_max_size);
3683 gclog_or_tty->print_cr(" Global Stack: pushes = " SIZE_FORMAT ", pops = " SIZE_FORMAT ", max size = " SIZE_FORMAT,
3684 _global_pushes, _global_pops, _global_max_size);
3685 gclog_or_tty->print_cr(" transfers to = " SIZE_FORMAT ", transfers from = " SIZE_FORMAT,
3686 _global_transfers_to,_global_transfers_from);
3687 gclog_or_tty->print_cr(" Regions: claimed = " SIZE_FORMAT, _regions_claimed);
3688 gclog_or_tty->print_cr(" SATB buffers: processed = " SIZE_FORMAT, _satb_buffers_processed);
3689 gclog_or_tty->print_cr(" Steals: attempts = " SIZE_FORMAT ", successes = " SIZE_FORMAT,
3690 _steal_attempts, _steals);
3691 gclog_or_tty->print_cr(" Aborted: " SIZE_FORMAT ", due to", _aborted);
3692 gclog_or_tty->print_cr(" overflow: " SIZE_FORMAT ", global abort: " SIZE_FORMAT ", yield: " SIZE_FORMAT,
3693 _aborted_overflow, _aborted_cm_aborted, _aborted_yield);
3694 gclog_or_tty->print_cr(" time out: " SIZE_FORMAT ", SATB: " SIZE_FORMAT ", termination: " SIZE_FORMAT,
3695 _aborted_timed_out, _aborted_satb, _aborted_termination);
3696 #endif // _MARKING_STATS_
3697 }
3698
3699 bool ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, oop& obj) {
3700 return _task_queues->steal(worker_id, hash_seed, obj);
3701 }
3702
3703 /*****************************************************************************
3704
3705 The do_marking_step(time_target_ms, ...) method is the building
3706 block of the parallel marking framework. It can be called in parallel
3707 with other invocations of do_marking_step() on different tasks
3708 (but only one per task, obviously) and concurrently with the
3709 mutator threads, or during remark, hence it eliminates the need
3710 for two versions of the code. When called during remark, it will
3711 pick up from where the task left off during the concurrent marking
3712 phase. Interestingly, tasks are also claimable during evacuation
3713 pauses too, since do_marking_step() ensures that it aborts before
3714 it needs to yield.
3715
3716 The data structures that it uses to do marking work are the
3717 following:
3718
3719 (1) Marking Bitmap. If there are gray objects that appear only
3720 on the bitmap (this happens either when dealing with an overflow
3721 or when the initial marking phase has simply marked the roots
3722 and didn't push them on the stack), then tasks claim heap
3723 regions whose bitmap they then scan to find gray objects. A
3724 global finger indicates where the end of the last claimed region
3725 is. A local finger indicates how far into the region a task has
3726 scanned. The two fingers are used to determine how to gray an
3727 object (i.e. whether simply marking it is OK, as it will be
3728 visited by a task in the future, or whether it needs to be also
3729 pushed on a stack).
3730
3731 (2) Local Queue. The local queue of the task which is accessed
3732 reasonably efficiently by the task. Other tasks can steal from
3733 it when they run out of work. Throughout the marking phase, a
3734 task attempts to keep its local queue short but not totally
3735 empty, so that entries are available for stealing by other
3736 tasks. Only when there is no more work, a task will totally
3737 drain its local queue.
3738
3739 (3) Global Mark Stack. This handles local queue overflow. During
3740 marking only sets of entries are moved between it and the local
3741 queues, as access to it requires a mutex and more fine-grain
3742 interaction with it which might cause contention. If it
3743 overflows, then the marking phase should restart and iterate
3744 over the bitmap to identify gray objects. Throughout the marking
3745 phase, tasks attempt to keep the global mark stack at a small
3746 length but not totally empty, so that entries are available for
3747 popping by other tasks. Only when there is no more work, tasks
3748 will totally drain the global mark stack.
3749
3750 (4) SATB Buffer Queue. This is where completed SATB buffers are
3751 made available. Buffers are regularly removed from this queue
3752 and scanned for roots, so that the queue doesn't get too
3753 long. During remark, all completed buffers are processed, as
3754 well as the filled in parts of any uncompleted buffers.
3755
3756 The do_marking_step() method tries to abort when the time target
3757 has been reached. There are a few other cases when the
3758 do_marking_step() method also aborts:
3759
3760 (1) When the marking phase has been aborted (after a Full GC).
3761
3762 (2) When a global overflow (on the global stack) has been
3763 triggered. Before the task aborts, it will actually sync up with
3764 the other tasks to ensure that all the marking data structures
3765 (local queues, stacks, fingers etc.) are re-initialized so that
3766 when do_marking_step() completes, the marking phase can
3767 immediately restart.
3768
3769 (3) When enough completed SATB buffers are available. The
3770 do_marking_step() method only tries to drain SATB buffers right
3771 at the beginning. So, if enough buffers are available, the
3772 marking step aborts and the SATB buffers are processed at
3773 the beginning of the next invocation.
3774
3775 (4) To yield. when we have to yield then we abort and yield
3776 right at the end of do_marking_step(). This saves us from a lot
3777 of hassle as, by yielding we might allow a Full GC. If this
3778 happens then objects will be compacted underneath our feet, the
3779 heap might shrink, etc. We save checking for this by just
3780 aborting and doing the yield right at the end.
3781
3782 From the above it follows that the do_marking_step() method should
3783 be called in a loop (or, otherwise, regularly) until it completes.
3784
3785 If a marking step completes without its has_aborted() flag being
3786 true, it means it has completed the current marking phase (and
3787 also all other marking tasks have done so and have all synced up).
3788
3789 A method called regular_clock_call() is invoked "regularly" (in
3790 sub ms intervals) throughout marking. It is this clock method that
3791 checks all the abort conditions which were mentioned above and
3792 decides when the task should abort. A work-based scheme is used to
3793 trigger this clock method: when the number of object words the
3794 marking phase has scanned or the number of references the marking
3795 phase has visited reach a given limit. Additional invocations to
3796 the method clock have been planted in a few other strategic places
3797 too. The initial reason for the clock method was to avoid calling
3798 vtime too regularly, as it is quite expensive. So, once it was in
3799 place, it was natural to piggy-back all the other conditions on it
3800 too and not constantly check them throughout the code.
3801
3802 If do_termination is true then do_marking_step will enter its
3803 termination protocol.
3804
3805 The value of is_serial must be true when do_marking_step is being
3806 called serially (i.e. by the VMThread) and do_marking_step should
3807 skip any synchronization in the termination and overflow code.
3808 Examples include the serial remark code and the serial reference
3809 processing closures.
3810
3811 The value of is_serial must be false when do_marking_step is
3812 being called by any of the worker threads in a work gang.
3813 Examples include the concurrent marking code (CMMarkingTask),
3814 the MT remark code, and the MT reference processing closures.
3815
3816 *****************************************************************************/
3817
3818 void CMTask::do_marking_step(double time_target_ms,
3819 bool do_termination,
3820 bool is_serial) {
3821 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
3822 assert(concurrent() == _cm->concurrent(), "they should be the same");
3823
3824 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
3825 assert(_task_queues != NULL, "invariant");
3826 assert(_task_queue != NULL, "invariant");
3827 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant");
3828
3829 assert(!_claimed,
3830 "only one thread should claim this task at any one time");
3831
3832 // OK, this doesn't safeguard again all possible scenarios, as it is
3833 // possible for two threads to set the _claimed flag at the same
3834 // time. But it is only for debugging purposes anyway and it will
3835 // catch most problems.
3836 _claimed = true;
3837
3838 _start_time_ms = os::elapsedVTime() * 1000.0;
3839 statsOnly( _interval_start_time_ms = _start_time_ms );
3840
3841 // If do_stealing is true then do_marking_step will attempt to
3842 // steal work from the other CMTasks. It only makes sense to
3843 // enable stealing when the termination protocol is enabled
3844 // and do_marking_step() is not being called serially.
3845 bool do_stealing = do_termination && !is_serial;
3846
3847 double diff_prediction_ms =
3848 g1_policy->get_new_prediction(&_marking_step_diffs_ms);
3849 _time_target_ms = time_target_ms - diff_prediction_ms;
3850
3851 // set up the variables that are used in the work-based scheme to
3852 // call the regular clock method
3853 _words_scanned = 0;
3854 _refs_reached = 0;
3855 recalculate_limits();
3856
3857 // clear all flags
3858 clear_has_aborted();
3859 _has_timed_out = false;
3860 _draining_satb_buffers = false;
3861
3862 ++_calls;
3863
3864 if (_cm->verbose_low()) {
3865 gclog_or_tty->print_cr("[%u] >>>>>>>>>> START, call = %d, "
3866 "target = %1.2lfms >>>>>>>>>>",
3867 _worker_id, _calls, _time_target_ms);
3868 }
3869
3870 // Set up the bitmap and oop closures. Anything that uses them is
3871 // eventually called from this method, so it is OK to allocate these
3872 // statically.
3873 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
3874 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
3875 set_cm_oop_closure(&cm_oop_closure);
3876
3877 if (_cm->has_overflown()) {
3878 // This can happen if the mark stack overflows during a GC pause
3879 // and this task, after a yield point, restarts. We have to abort
3880 // as we need to get into the overflow protocol which happens
3881 // right at the end of this task.
3882 set_has_aborted();
3883 }
3884
3885 // First drain any available SATB buffers. After this, we will not
3886 // look at SATB buffers before the next invocation of this method.
3887 // If enough completed SATB buffers are queued up, the regular clock
3888 // will abort this task so that it restarts.
3889 drain_satb_buffers();
3890 // ...then partially drain the local queue and the global stack
3891 drain_local_queue(true);
3892 drain_global_stack(true);
3893
3894 do {
3895 if (!has_aborted() && _curr_region != NULL) {
3896 // This means that we're already holding on to a region.
3897 assert(_finger != NULL, "if region is not NULL, then the finger "
3898 "should not be NULL either");
3899
3900 // We might have restarted this task after an evacuation pause
3901 // which might have evacuated the region we're holding on to
3902 // underneath our feet. Let's read its limit again to make sure
3903 // that we do not iterate over a region of the heap that
3904 // contains garbage (update_region_limit() will also move
3905 // _finger to the start of the region if it is found empty).
3906 update_region_limit();
3907 // We will start from _finger not from the start of the region,
3908 // as we might be restarting this task after aborting half-way
3909 // through scanning this region. In this case, _finger points to
3910 // the address where we last found a marked object. If this is a
3911 // fresh region, _finger points to start().
3912 MemRegion mr = MemRegion(_finger, _region_limit);
3913
3914 if (_cm->verbose_low()) {
3915 gclog_or_tty->print_cr("[%u] we're scanning part "
3916 "["PTR_FORMAT", "PTR_FORMAT") "
3917 "of region "HR_FORMAT,
3918 _worker_id, p2i(_finger), p2i(_region_limit),
3919 HR_FORMAT_PARAMS(_curr_region));
3920 }
3921
3922 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(),
3923 "humongous regions should go around loop once only");
3924
3925 // Some special cases:
3926 // If the memory region is empty, we can just give up the region.
3927 // If the current region is humongous then we only need to check
3928 // the bitmap for the bit associated with the start of the object,
3929 // scan the object if it's live, and give up the region.
3930 // Otherwise, let's iterate over the bitmap of the part of the region
3931 // that is left.
3932 // If the iteration is successful, give up the region.
3933 if (mr.is_empty()) {
3934 giveup_current_region();
3935 regular_clock_call();
3936 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) {
3937 if (_nextMarkBitMap->isMarked(mr.start())) {
3938 // The object is marked - apply the closure
3939 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start());
3940 bitmap_closure.do_bit(offset);
3941 }
3942 // Even if this task aborted while scanning the humongous object
3943 // we can (and should) give up the current region.
3944 giveup_current_region();
3945 regular_clock_call();
3946 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) {
3947 giveup_current_region();
3948 regular_clock_call();
3949 } else {
3950 assert(has_aborted(), "currently the only way to do so");
3951 // The only way to abort the bitmap iteration is to return
3952 // false from the do_bit() method. However, inside the
3953 // do_bit() method we move the _finger to point to the
3954 // object currently being looked at. So, if we bail out, we
3955 // have definitely set _finger to something non-null.
3956 assert(_finger != NULL, "invariant");
3957
3958 // Region iteration was actually aborted. So now _finger
3959 // points to the address of the object we last scanned. If we
3960 // leave it there, when we restart this task, we will rescan
3961 // the object. It is easy to avoid this. We move the finger by
3962 // enough to point to the next possible object header (the
3963 // bitmap knows by how much we need to move it as it knows its
3964 // granularity).
3965 assert(_finger < _region_limit, "invariant");
3966 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger);
3967 // Check if bitmap iteration was aborted while scanning the last object
3968 if (new_finger >= _region_limit) {
3969 giveup_current_region();
3970 } else {
3971 move_finger_to(new_finger);
3972 }
3973 }
3974 }
3975 // At this point we have either completed iterating over the
3976 // region we were holding on to, or we have aborted.
3977
3978 // We then partially drain the local queue and the global stack.
3979 // (Do we really need this?)
3980 drain_local_queue(true);
3981 drain_global_stack(true);
3982
3983 // Read the note on the claim_region() method on why it might
3984 // return NULL with potentially more regions available for
3985 // claiming and why we have to check out_of_regions() to determine
3986 // whether we're done or not.
3987 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
3988 // We are going to try to claim a new region. We should have
3989 // given up on the previous one.
3990 // Separated the asserts so that we know which one fires.
3991 assert(_curr_region == NULL, "invariant");
3992 assert(_finger == NULL, "invariant");
3993 assert(_region_limit == NULL, "invariant");
3994 if (_cm->verbose_low()) {
3995 gclog_or_tty->print_cr("[%u] trying to claim a new region", _worker_id);
3996 }
3997 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
3998 if (claimed_region != NULL) {
3999 // Yes, we managed to claim one
4000 statsOnly( ++_regions_claimed );
4001
4002 if (_cm->verbose_low()) {
4003 gclog_or_tty->print_cr("[%u] we successfully claimed "
4004 "region "PTR_FORMAT,
4005 _worker_id, p2i(claimed_region));
4006 }
4007
4008 setup_for_region(claimed_region);
4009 assert(_curr_region == claimed_region, "invariant");
4010 }
4011 // It is important to call the regular clock here. It might take
4012 // a while to claim a region if, for example, we hit a large
4013 // block of empty regions. So we need to call the regular clock
4014 // method once round the loop to make sure it's called
4015 // frequently enough.
4016 regular_clock_call();
4017 }
4018
4019 if (!has_aborted() && _curr_region == NULL) {
4020 assert(_cm->out_of_regions(),
4021 "at this point we should be out of regions");
4022 }
4023 } while ( _curr_region != NULL && !has_aborted());
4024
4025 if (!has_aborted()) {
4026 // We cannot check whether the global stack is empty, since other
4027 // tasks might be pushing objects to it concurrently.
4028 assert(_cm->out_of_regions(),
4029 "at this point we should be out of regions");
4030
4031 if (_cm->verbose_low()) {
4032 gclog_or_tty->print_cr("[%u] all regions claimed", _worker_id);
4033 }
4034
4035 // Try to reduce the number of available SATB buffers so that
4036 // remark has less work to do.
4037 drain_satb_buffers();
4038 }
4039
4040 // Since we've done everything else, we can now totally drain the
4041 // local queue and global stack.
4042 drain_local_queue(false);
4043 drain_global_stack(false);
4044
4045 // Attempt at work stealing from other task's queues.
4046 if (do_stealing && !has_aborted()) {
4047 // We have not aborted. This means that we have finished all that
4048 // we could. Let's try to do some stealing...
4049
4050 // We cannot check whether the global stack is empty, since other
4051 // tasks might be pushing objects to it concurrently.
4052 assert(_cm->out_of_regions() && _task_queue->size() == 0,
4053 "only way to reach here");
4054
4055 if (_cm->verbose_low()) {
4056 gclog_or_tty->print_cr("[%u] starting to steal", _worker_id);
4057 }
4058
4059 while (!has_aborted()) {
4060 oop obj;
4061 statsOnly( ++_steal_attempts );
4062
4063 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) {
4064 if (_cm->verbose_medium()) {
4065 gclog_or_tty->print_cr("[%u] stolen "PTR_FORMAT" successfully",
4066 _worker_id, p2i((void*) obj));
4067 }
4068
4069 statsOnly( ++_steals );
4070
4071 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
4072 "any stolen object should be marked");
4073 scan_object(obj);
4074
4075 // And since we're towards the end, let's totally drain the
4076 // local queue and global stack.
4077 drain_local_queue(false);
4078 drain_global_stack(false);
4079 } else {
4080 break;
4081 }
4082 }
4083 }
4084
4085 // If we are about to wrap up and go into termination, check if we
4086 // should raise the overflow flag.
4087 if (do_termination && !has_aborted()) {
4088 if (_cm->force_overflow()->should_force()) {
4089 _cm->set_has_overflown();
4090 regular_clock_call();
4091 }
4092 }
4093
4094 // We still haven't aborted. Now, let's try to get into the
4095 // termination protocol.
4096 if (do_termination && !has_aborted()) {
4097 // We cannot check whether the global stack is empty, since other
4098 // tasks might be concurrently pushing objects on it.
4099 // Separated the asserts so that we know which one fires.
4100 assert(_cm->out_of_regions(), "only way to reach here");
4101 assert(_task_queue->size() == 0, "only way to reach here");
4102
4103 if (_cm->verbose_low()) {
4104 gclog_or_tty->print_cr("[%u] starting termination protocol", _worker_id);
4105 }
4106
4107 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
4108
4109 // The CMTask class also extends the TerminatorTerminator class,
4110 // hence its should_exit_termination() method will also decide
4111 // whether to exit the termination protocol or not.
4112 bool finished = (is_serial ||
4113 _cm->terminator()->offer_termination(this));
4114 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
4115 _termination_time_ms +=
4116 termination_end_time_ms - _termination_start_time_ms;
4117
4118 if (finished) {
4119 // We're all done.
4120
4121 if (_worker_id == 0) {
4122 // let's allow task 0 to do this
4123 if (concurrent()) {
4124 assert(_cm->concurrent_marking_in_progress(), "invariant");
4125 // we need to set this to false before the next
4126 // safepoint. This way we ensure that the marking phase
4127 // doesn't observe any more heap expansions.
4128 _cm->clear_concurrent_marking_in_progress();
4129 }
4130 }
4131
4132 // We can now guarantee that the global stack is empty, since
4133 // all other tasks have finished. We separated the guarantees so
4134 // that, if a condition is false, we can immediately find out
4135 // which one.
4136 guarantee(_cm->out_of_regions(), "only way to reach here");
4137 guarantee(_cm->mark_stack_empty(), "only way to reach here");
4138 guarantee(_task_queue->size() == 0, "only way to reach here");
4139 guarantee(!_cm->has_overflown(), "only way to reach here");
4140 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
4141
4142 if (_cm->verbose_low()) {
4143 gclog_or_tty->print_cr("[%u] all tasks terminated", _worker_id);
4144 }
4145 } else {
4146 // Apparently there's more work to do. Let's abort this task. It
4147 // will restart it and we can hopefully find more things to do.
4148
4149 if (_cm->verbose_low()) {
4150 gclog_or_tty->print_cr("[%u] apparently there is more work to do",
4151 _worker_id);
4152 }
4153
4154 set_has_aborted();
4155 statsOnly( ++_aborted_termination );
4156 }
4157 }
4158
4159 // Mainly for debugging purposes to make sure that a pointer to the
4160 // closure which was statically allocated in this frame doesn't
4161 // escape it by accident.
4162 set_cm_oop_closure(NULL);
4163 double end_time_ms = os::elapsedVTime() * 1000.0;
4164 double elapsed_time_ms = end_time_ms - _start_time_ms;
4165 // Update the step history.
4166 _step_times_ms.add(elapsed_time_ms);
4167
4168 if (has_aborted()) {
4169 // The task was aborted for some reason.
4170
4171 statsOnly( ++_aborted );
4172
4173 if (_has_timed_out) {
4174 double diff_ms = elapsed_time_ms - _time_target_ms;
4175 // Keep statistics of how well we did with respect to hitting
4176 // our target only if we actually timed out (if we aborted for
4177 // other reasons, then the results might get skewed).
4178 _marking_step_diffs_ms.add(diff_ms);
4179 }
4180
4181 if (_cm->has_overflown()) {
4182 // This is the interesting one. We aborted because a global
4183 // overflow was raised. This means we have to restart the
4184 // marking phase and start iterating over regions. However, in
4185 // order to do this we have to make sure that all tasks stop
4186 // what they are doing and re-initialize in a safe manner. We
4187 // will achieve this with the use of two barrier sync points.
4188
4189 if (_cm->verbose_low()) {
4190 gclog_or_tty->print_cr("[%u] detected overflow", _worker_id);
4191 }
4192
4193 if (!is_serial) {
4194 // We only need to enter the sync barrier if being called
4195 // from a parallel context
4196 _cm->enter_first_sync_barrier(_worker_id);
4197
4198 // When we exit this sync barrier we know that all tasks have
4199 // stopped doing marking work. So, it's now safe to
4200 // re-initialize our data structures. At the end of this method,
4201 // task 0 will clear the global data structures.
4202 }
4203
4204 statsOnly( ++_aborted_overflow );
4205
4206 // We clear the local state of this task...
4207 clear_region_fields();
4208
4209 if (!is_serial) {
4210 // ...and enter the second barrier.
4211 _cm->enter_second_sync_barrier(_worker_id);
4212 }
4213 // At this point, if we're during the concurrent phase of
4214 // marking, everything has been re-initialized and we're
4215 // ready to restart.
4216 }
4217
4218 if (_cm->verbose_low()) {
4219 gclog_or_tty->print_cr("[%u] <<<<<<<<<< ABORTING, target = %1.2lfms, "
4220 "elapsed = %1.2lfms <<<<<<<<<<",
4221 _worker_id, _time_target_ms, elapsed_time_ms);
4222 if (_cm->has_aborted()) {
4223 gclog_or_tty->print_cr("[%u] ========== MARKING ABORTED ==========",
4224 _worker_id);
4225 }
4226 }
4227 } else {
4228 if (_cm->verbose_low()) {
4229 gclog_or_tty->print_cr("[%u] <<<<<<<<<< FINISHED, target = %1.2lfms, "
4230 "elapsed = %1.2lfms <<<<<<<<<<",
4231 _worker_id, _time_target_ms, elapsed_time_ms);
4232 }
4233 }
4234
4235 _claimed = false;
4236 }
4237
4238 CMTask::CMTask(uint worker_id,
4239 ConcurrentMark* cm,
4240 size_t* marked_bytes,
4241 BitMap* card_bm,
4242 CMTaskQueue* task_queue,
4243 CMTaskQueueSet* task_queues)
4244 : _g1h(G1CollectedHeap::heap()),
4245 _worker_id(worker_id), _cm(cm),
4246 _claimed(false),
4247 _nextMarkBitMap(NULL), _hash_seed(17),
4248 _task_queue(task_queue),
4249 _task_queues(task_queues),
4250 _cm_oop_closure(NULL),
4251 _marked_bytes_array(marked_bytes),
4252 _card_bm(card_bm) {
4253 guarantee(task_queue != NULL, "invariant");
4254 guarantee(task_queues != NULL, "invariant");
4255
4256 statsOnly( _clock_due_to_scanning = 0;
4257 _clock_due_to_marking = 0 );
4258
4259 _marking_step_diffs_ms.add(0.5);
4260 }
4261
4262 // These are formatting macros that are used below to ensure
4263 // consistent formatting. The *_H_* versions are used to format the
4264 // header for a particular value and they should be kept consistent
4265 // with the corresponding macro. Also note that most of the macros add
4266 // the necessary white space (as a prefix) which makes them a bit
4267 // easier to compose.
4268
4269 // All the output lines are prefixed with this string to be able to
4270 // identify them easily in a large log file.
4271 #define G1PPRL_LINE_PREFIX "###"
4272
4273 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT
4274 #ifdef _LP64
4275 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
4276 #else // _LP64
4277 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
4278 #endif // _LP64
4279
4280 // For per-region info
4281 #define G1PPRL_TYPE_FORMAT " %-4s"
4282 #define G1PPRL_TYPE_H_FORMAT " %4s"
4283 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9)
4284 #define G1PPRL_BYTE_H_FORMAT " %9s"
4285 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
4286 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
4287
4288 // For summary info
4289 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT
4290 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT
4291 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB"
4292 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%"
4293
4294 G1PrintRegionLivenessInfoClosure::
4295 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name)
4296 : _out(out),
4297 _total_used_bytes(0), _total_capacity_bytes(0),
4298 _total_prev_live_bytes(0), _total_next_live_bytes(0),
4299 _hum_used_bytes(0), _hum_capacity_bytes(0),
4300 _hum_prev_live_bytes(0), _hum_next_live_bytes(0),
4301 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) {
4302 G1CollectedHeap* g1h = G1CollectedHeap::heap();
4303 MemRegion g1_reserved = g1h->g1_reserved();
4304 double now = os::elapsedTime();
4305
4306 // Print the header of the output.
4307 _out->cr();
4308 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
4309 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP"
4310 G1PPRL_SUM_ADDR_FORMAT("reserved")
4311 G1PPRL_SUM_BYTE_FORMAT("region-size"),
4312 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
4313 HeapRegion::GrainBytes);
4314 _out->print_cr(G1PPRL_LINE_PREFIX);
4315 _out->print_cr(G1PPRL_LINE_PREFIX
4316 G1PPRL_TYPE_H_FORMAT
4317 G1PPRL_ADDR_BASE_H_FORMAT
4318 G1PPRL_BYTE_H_FORMAT
4319 G1PPRL_BYTE_H_FORMAT
4320 G1PPRL_BYTE_H_FORMAT
4321 G1PPRL_DOUBLE_H_FORMAT
4322 G1PPRL_BYTE_H_FORMAT
4323 G1PPRL_BYTE_H_FORMAT,
4324 "type", "address-range",
4325 "used", "prev-live", "next-live", "gc-eff",
4326 "remset", "code-roots");
4327 _out->print_cr(G1PPRL_LINE_PREFIX
4328 G1PPRL_TYPE_H_FORMAT
4329 G1PPRL_ADDR_BASE_H_FORMAT
4330 G1PPRL_BYTE_H_FORMAT
4331 G1PPRL_BYTE_H_FORMAT
4332 G1PPRL_BYTE_H_FORMAT
4333 G1PPRL_DOUBLE_H_FORMAT
4334 G1PPRL_BYTE_H_FORMAT
4335 G1PPRL_BYTE_H_FORMAT,
4336 "", "",
4337 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
4338 "(bytes)", "(bytes)");
4339 }
4340
4341 // It takes as a parameter a reference to one of the _hum_* fields, it
4342 // deduces the corresponding value for a region in a humongous region
4343 // series (either the region size, or what's left if the _hum_* field
4344 // is < the region size), and updates the _hum_* field accordingly.
4345 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
4346 size_t bytes = 0;
4347 // The > 0 check is to deal with the prev and next live bytes which
4348 // could be 0.
4349 if (*hum_bytes > 0) {
4350 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
4351 *hum_bytes -= bytes;
4352 }
4353 return bytes;
4354 }
4355
4356 // It deduces the values for a region in a humongous region series
4357 // from the _hum_* fields and updates those accordingly. It assumes
4358 // that that _hum_* fields have already been set up from the "starts
4359 // humongous" region and we visit the regions in address order.
4360 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
4361 size_t* capacity_bytes,
4362 size_t* prev_live_bytes,
4363 size_t* next_live_bytes) {
4364 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
4365 *used_bytes = get_hum_bytes(&_hum_used_bytes);
4366 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
4367 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
4368 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
4369 }
4370
4371 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
4372 const char* type = r->get_type_str();
4373 HeapWord* bottom = r->bottom();
4374 HeapWord* end = r->end();
4375 size_t capacity_bytes = r->capacity();
4376 size_t used_bytes = r->used();
4377 size_t prev_live_bytes = r->live_bytes();
4378 size_t next_live_bytes = r->next_live_bytes();
4379 double gc_eff = r->gc_efficiency();
4380 size_t remset_bytes = r->rem_set()->mem_size();
4381 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
4382
4383 if (r->is_starts_humongous()) {
4384 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
4385 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
4386 "they should have been zeroed after the last time we used them");
4387 // Set up the _hum_* fields.
4388 _hum_capacity_bytes = capacity_bytes;
4389 _hum_used_bytes = used_bytes;
4390 _hum_prev_live_bytes = prev_live_bytes;
4391 _hum_next_live_bytes = next_live_bytes;
4392 get_hum_bytes(&used_bytes, &capacity_bytes,
4393 &prev_live_bytes, &next_live_bytes);
4394 end = bottom + HeapRegion::GrainWords;
4395 } else if (r->is_continues_humongous()) {
4396 get_hum_bytes(&used_bytes, &capacity_bytes,
4397 &prev_live_bytes, &next_live_bytes);
4398 assert(end == bottom + HeapRegion::GrainWords, "invariant");
4399 }
4400
4401 _total_used_bytes += used_bytes;
4402 _total_capacity_bytes += capacity_bytes;
4403 _total_prev_live_bytes += prev_live_bytes;
4404 _total_next_live_bytes += next_live_bytes;
4405 _total_remset_bytes += remset_bytes;
4406 _total_strong_code_roots_bytes += strong_code_roots_bytes;
4407
4408 // Print a line for this particular region.
4409 _out->print_cr(G1PPRL_LINE_PREFIX
4410 G1PPRL_TYPE_FORMAT
4411 G1PPRL_ADDR_BASE_FORMAT
4412 G1PPRL_BYTE_FORMAT
4413 G1PPRL_BYTE_FORMAT
4414 G1PPRL_BYTE_FORMAT
4415 G1PPRL_DOUBLE_FORMAT
4416 G1PPRL_BYTE_FORMAT
4417 G1PPRL_BYTE_FORMAT,
4418 type, p2i(bottom), p2i(end),
4419 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
4420 remset_bytes, strong_code_roots_bytes);
4421
4422 return false;
4423 }
4424
4425 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
4426 // add static memory usages to remembered set sizes
4427 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
4428 // Print the footer of the output.
4429 _out->print_cr(G1PPRL_LINE_PREFIX);
4430 _out->print_cr(G1PPRL_LINE_PREFIX
4431 " SUMMARY"
4432 G1PPRL_SUM_MB_FORMAT("capacity")
4433 G1PPRL_SUM_MB_PERC_FORMAT("used")
4434 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
4435 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
4436 G1PPRL_SUM_MB_FORMAT("remset")
4437 G1PPRL_SUM_MB_FORMAT("code-roots"),
4438 bytes_to_mb(_total_capacity_bytes),
4439 bytes_to_mb(_total_used_bytes),
4440 perc(_total_used_bytes, _total_capacity_bytes),
4441 bytes_to_mb(_total_prev_live_bytes),
4442 perc(_total_prev_live_bytes, _total_capacity_bytes),
4443 bytes_to_mb(_total_next_live_bytes),
4444 perc(_total_next_live_bytes, _total_capacity_bytes),
4445 bytes_to_mb(_total_remset_bytes),
4446 bytes_to_mb(_total_strong_code_roots_bytes));
4447 _out->cr();
4448 }