1 /*
2 * Copyright (c) 2001, 2013, 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/symbolTable.hpp"
27 #include "gc_implementation/g1/concurrentMark.inline.hpp"
28 #include "gc_implementation/g1/concurrentMarkThread.inline.hpp"
29 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
30 #include "gc_implementation/g1/g1CollectorPolicy.hpp"
31 #include "gc_implementation/g1/g1ErgoVerbose.hpp"
32 #include "gc_implementation/g1/g1Log.hpp"
33 #include "gc_implementation/g1/g1OopClosures.inline.hpp"
34 #include "gc_implementation/g1/g1RemSet.hpp"
35 #include "gc_implementation/g1/heapRegion.inline.hpp"
36 #include "gc_implementation/g1/heapRegionRemSet.hpp"
37 #include "gc_implementation/g1/heapRegionSeq.inline.hpp"
38 #include "gc_implementation/shared/vmGCOperations.hpp"
39 #include "gc_implementation/shared/gcTimer.hpp"
40 #include "gc_implementation/shared/gcTrace.hpp"
41 #include "gc_implementation/shared/gcTraceTime.hpp"
42 #include "memory/genOopClosures.inline.hpp"
43 #include "memory/referencePolicy.hpp"
44 #include "memory/resourceArea.hpp"
45 #include "oops/oop.inline.hpp"
46 #include "runtime/handles.inline.hpp"
47 #include "runtime/java.hpp"
48 #include "services/memTracker.hpp"
49
50 // Concurrent marking bit map wrapper
51
52 CMBitMapRO::CMBitMapRO(ReservedSpace rs, int shifter) :
53 _bm((uintptr_t*)NULL,0),
54 _shifter(shifter) {
55 _bmStartWord = (HeapWord*)(rs.base());
56 _bmWordSize = rs.size()/HeapWordSize; // rs.size() is in bytes
57 ReservedSpace brs(ReservedSpace::allocation_align_size_up(
58 (_bmWordSize >> (_shifter + LogBitsPerByte)) + 1));
59
60 MemTracker::record_virtual_memory_type((address)brs.base(), mtGC);
61
62 guarantee(brs.is_reserved(), "couldn't allocate concurrent marking bit map");
63 // For now we'll just commit all of the bit map up fromt.
64 // Later on we'll try to be more parsimonious with swap.
65 guarantee(_virtual_space.initialize(brs, brs.size()),
66 "couldn't reseve backing store for concurrent marking bit map");
67 assert(_virtual_space.committed_size() == brs.size(),
68 "didn't reserve backing store for all of concurrent marking bit map?");
69 _bm.set_map((uintptr_t*)_virtual_space.low());
70 assert(_virtual_space.committed_size() << (_shifter + LogBitsPerByte) >=
71 _bmWordSize, "inconsistency in bit map sizing");
72 _bm.set_size(_bmWordSize >> _shifter);
73 }
74
75 HeapWord* CMBitMapRO::getNextMarkedWordAddress(HeapWord* addr,
76 HeapWord* limit) const {
77 // First we must round addr *up* to a possible object boundary.
78 addr = (HeapWord*)align_size_up((intptr_t)addr,
79 HeapWordSize << _shifter);
80 size_t addrOffset = heapWordToOffset(addr);
81 if (limit == NULL) {
82 limit = _bmStartWord + _bmWordSize;
83 }
84 size_t limitOffset = heapWordToOffset(limit);
85 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
86 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
87 assert(nextAddr >= addr, "get_next_one postcondition");
88 assert(nextAddr == limit || isMarked(nextAddr),
89 "get_next_one postcondition");
90 return nextAddr;
91 }
92
93 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(HeapWord* addr,
94 HeapWord* limit) const {
95 size_t addrOffset = heapWordToOffset(addr);
96 if (limit == NULL) {
97 limit = _bmStartWord + _bmWordSize;
98 }
99 size_t limitOffset = heapWordToOffset(limit);
100 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset);
101 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
102 assert(nextAddr >= addr, "get_next_one postcondition");
103 assert(nextAddr == limit || !isMarked(nextAddr),
104 "get_next_one postcondition");
105 return nextAddr;
106 }
107
108 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const {
109 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check");
110 return (int) (diff >> _shifter);
111 }
112
113 #ifndef PRODUCT
114 bool CMBitMapRO::covers(ReservedSpace rs) const {
115 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
116 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
117 "size inconsistency");
118 return _bmStartWord == (HeapWord*)(rs.base()) &&
119 _bmWordSize == rs.size()>>LogHeapWordSize;
120 }
121 #endif
122
123 void CMBitMap::clearAll() {
124 _bm.clear();
125 return;
126 }
127
128 void CMBitMap::markRange(MemRegion mr) {
129 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
130 assert(!mr.is_empty(), "unexpected empty region");
131 assert((offsetToHeapWord(heapWordToOffset(mr.end())) ==
132 ((HeapWord *) mr.end())),
133 "markRange memory region end is not card aligned");
134 // convert address range into offset range
135 _bm.at_put_range(heapWordToOffset(mr.start()),
136 heapWordToOffset(mr.end()), true);
137 }
138
139 void CMBitMap::clearRange(MemRegion mr) {
140 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
141 assert(!mr.is_empty(), "unexpected empty region");
142 // convert address range into offset range
143 _bm.at_put_range(heapWordToOffset(mr.start()),
144 heapWordToOffset(mr.end()), false);
145 }
146
147 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr,
148 HeapWord* end_addr) {
149 HeapWord* start = getNextMarkedWordAddress(addr);
150 start = MIN2(start, end_addr);
151 HeapWord* end = getNextUnmarkedWordAddress(start);
152 end = MIN2(end, end_addr);
153 assert(start <= end, "Consistency check");
154 MemRegion mr(start, end);
155 if (!mr.is_empty()) {
156 clearRange(mr);
157 }
158 return mr;
159 }
160
161 CMMarkStack::CMMarkStack(ConcurrentMark* cm) :
162 _base(NULL), _cm(cm)
163 #ifdef ASSERT
164 , _drain_in_progress(false)
165 , _drain_in_progress_yields(false)
166 #endif
167 {}
168
169 void CMMarkStack::allocate(size_t size) {
170 _base = NEW_C_HEAP_ARRAY(oop, size, mtGC);
171 if (_base == NULL) {
172 vm_exit_during_initialization("Failed to allocate CM region mark stack");
173 }
174 _index = 0;
175 _capacity = (jint) size;
176 _saved_index = -1;
177 NOT_PRODUCT(_max_depth = 0);
178 }
179
180 CMMarkStack::~CMMarkStack() {
181 if (_base != NULL) {
182 FREE_C_HEAP_ARRAY(oop, _base, mtGC);
183 }
184 }
185
186 void CMMarkStack::par_push(oop ptr) {
187 while (true) {
188 if (isFull()) {
189 _overflow = true;
190 return;
191 }
192 // Otherwise...
193 jint index = _index;
194 jint next_index = index+1;
195 jint res = Atomic::cmpxchg(next_index, &_index, index);
196 if (res == index) {
197 _base[index] = ptr;
198 // Note that we don't maintain this atomically. We could, but it
199 // doesn't seem necessary.
200 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
201 return;
202 }
203 // Otherwise, we need to try again.
204 }
205 }
206
207 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) {
208 while (true) {
209 if (isFull()) {
210 _overflow = true;
211 return;
212 }
213 // Otherwise...
214 jint index = _index;
215 jint next_index = index + n;
216 if (next_index > _capacity) {
217 _overflow = true;
218 return;
219 }
220 jint res = Atomic::cmpxchg(next_index, &_index, index);
221 if (res == index) {
222 for (int i = 0; i < n; i++) {
223 int ind = index + i;
224 assert(ind < _capacity, "By overflow test above.");
225 _base[ind] = ptr_arr[i];
226 }
227 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
228 return;
229 }
230 // Otherwise, we need to try again.
231 }
232 }
233
234
235 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
236 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
237 jint start = _index;
238 jint next_index = start + n;
239 if (next_index > _capacity) {
240 _overflow = true;
241 return;
242 }
243 // Otherwise.
244 _index = next_index;
245 for (int i = 0; i < n; i++) {
246 int ind = start + i;
247 assert(ind < _capacity, "By overflow test above.");
248 _base[ind] = ptr_arr[i];
249 }
250 }
251
252
253 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
254 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
255 jint index = _index;
256 if (index == 0) {
257 *n = 0;
258 return false;
259 } else {
260 int k = MIN2(max, index);
261 jint new_ind = index - k;
262 for (int j = 0; j < k; j++) {
263 ptr_arr[j] = _base[new_ind + j];
264 }
265 _index = new_ind;
266 *n = k;
267 return true;
268 }
269 }
270
271 template<class OopClosureClass>
272 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) {
273 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after
274 || SafepointSynchronize::is_at_safepoint(),
275 "Drain recursion must be yield-safe.");
276 bool res = true;
277 debug_only(_drain_in_progress = true);
278 debug_only(_drain_in_progress_yields = yield_after);
279 while (!isEmpty()) {
280 oop newOop = pop();
281 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop");
282 assert(newOop->is_oop(), "Expected an oop");
283 assert(bm == NULL || bm->isMarked((HeapWord*)newOop),
284 "only grey objects on this stack");
285 newOop->oop_iterate(cl);
286 if (yield_after && _cm->do_yield_check()) {
287 res = false;
288 break;
289 }
290 }
291 debug_only(_drain_in_progress = false);
292 return res;
293 }
294
295 void CMMarkStack::note_start_of_gc() {
296 assert(_saved_index == -1,
297 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
298 _saved_index = _index;
299 }
300
301 void CMMarkStack::note_end_of_gc() {
302 // This is intentionally a guarantee, instead of an assert. If we
303 // accidentally add something to the mark stack during GC, it
304 // will be a correctness issue so it's better if we crash. we'll
305 // only check this once per GC anyway, so it won't be a performance
306 // issue in any way.
307 guarantee(_saved_index == _index,
308 err_msg("saved index: %d index: %d", _saved_index, _index));
309 _saved_index = -1;
310 }
311
312 void CMMarkStack::oops_do(OopClosure* f) {
313 assert(_saved_index == _index,
314 err_msg("saved index: %d index: %d", _saved_index, _index));
315 for (int i = 0; i < _index; i += 1) {
316 f->do_oop(&_base[i]);
317 }
318 }
319
320 bool ConcurrentMark::not_yet_marked(oop obj) const {
321 return (_g1h->is_obj_ill(obj)
322 || (_g1h->is_in_permanent(obj)
323 && !nextMarkBitMap()->isMarked((HeapWord*)obj)));
324 }
325
326 CMRootRegions::CMRootRegions() :
327 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
328 _should_abort(false), _next_survivor(NULL) { }
329
330 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) {
331 _young_list = g1h->young_list();
332 _cm = cm;
333 }
334
335 void CMRootRegions::prepare_for_scan() {
336 assert(!scan_in_progress(), "pre-condition");
337
338 // Currently, only survivors can be root regions.
339 assert(_next_survivor == NULL, "pre-condition");
340 _next_survivor = _young_list->first_survivor_region();
341 _scan_in_progress = (_next_survivor != NULL);
342 _should_abort = false;
343 }
344
345 HeapRegion* CMRootRegions::claim_next() {
346 if (_should_abort) {
347 // If someone has set the should_abort flag, we return NULL to
348 // force the caller to bail out of their loop.
349 return NULL;
350 }
351
352 // Currently, only survivors can be root regions.
353 HeapRegion* res = _next_survivor;
354 if (res != NULL) {
355 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
356 // Read it again in case it changed while we were waiting for the lock.
357 res = _next_survivor;
358 if (res != NULL) {
359 if (res == _young_list->last_survivor_region()) {
360 // We just claimed the last survivor so store NULL to indicate
361 // that we're done.
362 _next_survivor = NULL;
363 } else {
364 _next_survivor = res->get_next_young_region();
365 }
366 } else {
367 // Someone else claimed the last survivor while we were trying
368 // to take the lock so nothing else to do.
369 }
370 }
371 assert(res == NULL || res->is_survivor(), "post-condition");
372
373 return res;
374 }
375
376 void CMRootRegions::scan_finished() {
377 assert(scan_in_progress(), "pre-condition");
378
379 // Currently, only survivors can be root regions.
380 if (!_should_abort) {
381 assert(_next_survivor == NULL, "we should have claimed all survivors");
382 }
383 _next_survivor = NULL;
384
385 {
386 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
387 _scan_in_progress = false;
388 RootRegionScan_lock->notify_all();
389 }
390 }
391
392 bool CMRootRegions::wait_until_scan_finished() {
393 if (!scan_in_progress()) return false;
394
395 {
396 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
397 while (scan_in_progress()) {
398 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
399 }
400 }
401 return true;
402 }
403
404 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away
405 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
406 #endif // _MSC_VER
407
408 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
409 return MAX2((n_par_threads + 2) / 4, 1U);
410 }
411
412 ConcurrentMark::ConcurrentMark(ReservedSpace rs, uint max_regions) :
413 _markBitMap1(rs, MinObjAlignment - 1),
414 _markBitMap2(rs, MinObjAlignment - 1),
415
416 _parallel_marking_threads(0),
417 _max_parallel_marking_threads(0),
418 _sleep_factor(0.0),
419 _marking_task_overhead(1.0),
420 _cleanup_sleep_factor(0.0),
421 _cleanup_task_overhead(1.0),
422 _cleanup_list("Cleanup List"),
423 _region_bm((BitMap::idx_t) max_regions, false /* in_resource_area*/),
424 _card_bm((rs.size() + CardTableModRefBS::card_size - 1) >>
425 CardTableModRefBS::card_shift,
426 false /* in_resource_area*/),
427
428 _prevMarkBitMap(&_markBitMap1),
429 _nextMarkBitMap(&_markBitMap2),
430
431 _markStack(this),
432 // _finger set in set_non_marking_state
433
434 _max_task_num(MAX2((uint)ParallelGCThreads, 1U)),
435 // _active_tasks set in set_non_marking_state
436 // _tasks set inside the constructor
437 _task_queues(new CMTaskQueueSet((int) _max_task_num)),
438 _terminator(ParallelTaskTerminator((int) _max_task_num, _task_queues)),
439
440 _has_overflown(false),
441 _concurrent(false),
442 _has_aborted(false),
443 _restart_for_overflow(false),
444 _concurrent_marking_in_progress(false),
445
446 // _verbose_level set below
447
448 _init_times(),
449 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
450 _cleanup_times(),
451 _total_counting_time(0.0),
452 _total_rs_scrub_time(0.0),
453
454 _parallel_workers(NULL),
455
456 _count_card_bitmaps(NULL),
457 _count_marked_bytes(NULL) {
458 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel;
459 if (verbose_level < no_verbose) {
460 verbose_level = no_verbose;
461 }
462 if (verbose_level > high_verbose) {
463 verbose_level = high_verbose;
464 }
465 _verbose_level = verbose_level;
466
467 if (verbose_low()) {
468 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", "
469 "heap end = "PTR_FORMAT, _heap_start, _heap_end);
470 }
471
472 _markStack.allocate(MarkStackSize);
473
474 // Create & start a ConcurrentMark thread.
475 _cmThread = new ConcurrentMarkThread(this);
476 assert(cmThread() != NULL, "CM Thread should have been created");
477 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
478
479 _g1h = G1CollectedHeap::heap();
480 assert(CGC_lock != NULL, "Where's the CGC_lock?");
481 assert(_markBitMap1.covers(rs), "_markBitMap1 inconsistency");
482 assert(_markBitMap2.covers(rs), "_markBitMap2 inconsistency");
483
484 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
485 satb_qs.set_buffer_size(G1SATBBufferSize);
486
487 _root_regions.init(_g1h, this);
488
489 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_task_num, mtGC);
490 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_task_num, mtGC);
491
492 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_task_num, mtGC);
493 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_task_num, mtGC);
494
495 BitMap::idx_t card_bm_size = _card_bm.size();
496
497 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
498 _active_tasks = _max_task_num;
499 for (int i = 0; i < (int) _max_task_num; ++i) {
500 CMTaskQueue* task_queue = new CMTaskQueue();
501 task_queue->initialize();
502 _task_queues->register_queue(i, task_queue);
503
504 _count_card_bitmaps[i] = BitMap(card_bm_size, false);
505 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, (size_t) max_regions, mtGC);
506
507 _tasks[i] = new CMTask(i, this,
508 _count_marked_bytes[i],
509 &_count_card_bitmaps[i],
510 task_queue, _task_queues);
511
512 _accum_task_vtime[i] = 0.0;
513 }
514
515 // Calculate the card number for the bottom of the heap. Used
516 // in biasing indexes into the accounting card bitmaps.
517 _heap_bottom_card_num =
518 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
519 CardTableModRefBS::card_shift);
520
521 // Clear all the liveness counting data
522 clear_all_count_data();
523
524 if (ConcGCThreads > ParallelGCThreads) {
525 vm_exit_during_initialization("Can't have more ConcGCThreads "
526 "than ParallelGCThreads.");
527 }
528 if (ParallelGCThreads == 0) {
529 // if we are not running with any parallel GC threads we will not
530 // spawn any marking threads either
531 _parallel_marking_threads = 0;
532 _max_parallel_marking_threads = 0;
533 _sleep_factor = 0.0;
534 _marking_task_overhead = 1.0;
535 } else {
536 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
537 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
538 // if both are set
539 _sleep_factor = 0.0;
540 _marking_task_overhead = 1.0;
541 } else if (G1MarkingOverheadPercent > 0) {
542 // We will calculate the number of parallel marking threads based
543 // on a target overhead with respect to the soft real-time goal
544 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
545 double overall_cm_overhead =
546 (double) MaxGCPauseMillis * marking_overhead /
547 (double) GCPauseIntervalMillis;
548 double cpu_ratio = 1.0 / (double) os::processor_count();
549 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
550 double marking_task_overhead =
551 overall_cm_overhead / marking_thread_num *
552 (double) os::processor_count();
553 double sleep_factor =
554 (1.0 - marking_task_overhead) / marking_task_overhead;
555
556 FLAG_SET_ERGO(uintx, ConcGCThreads, (uint) marking_thread_num);
557 _sleep_factor = sleep_factor;
558 _marking_task_overhead = marking_task_overhead;
559 } else {
560 // Calculate the number of parallel marking threads by scaling
561 // the number of parallel GC threads.
562 uint marking_thread_num = scale_parallel_threads((uint) ParallelGCThreads);
563 FLAG_SET_ERGO(uintx, ConcGCThreads, marking_thread_num);
564 _sleep_factor = 0.0;
565 _marking_task_overhead = 1.0;
566 }
567
568 assert(ConcGCThreads > 0, "Should have been set");
569 _parallel_marking_threads = (uint) ConcGCThreads;
570 _max_parallel_marking_threads = _parallel_marking_threads;
571
572 if (parallel_marking_threads() > 1) {
573 _cleanup_task_overhead = 1.0;
574 } else {
575 _cleanup_task_overhead = marking_task_overhead();
576 }
577 _cleanup_sleep_factor =
578 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead();
579
580 #if 0
581 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads());
582 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead());
583 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor());
584 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead());
585 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor());
586 #endif
587
588 guarantee(parallel_marking_threads() > 0, "peace of mind");
589 _parallel_workers = new FlexibleWorkGang("G1 Parallel Marking Threads",
590 _max_parallel_marking_threads, false, true);
591 if (_parallel_workers == NULL) {
592 vm_exit_during_initialization("Failed necessary allocation.");
593 } else {
594 _parallel_workers->initialize_workers();
595 }
596 }
597
598 // so that the call below can read a sensible value
599 _heap_start = (HeapWord*) rs.base();
600 set_non_marking_state();
601 }
602
603 void ConcurrentMark::update_g1_committed(bool force) {
604 // If concurrent marking is not in progress, then we do not need to
605 // update _heap_end.
606 if (!concurrent_marking_in_progress() && !force) return;
607
608 MemRegion committed = _g1h->g1_committed();
609 assert(committed.start() == _heap_start, "start shouldn't change");
610 HeapWord* new_end = committed.end();
611 if (new_end > _heap_end) {
612 // The heap has been expanded.
613
614 _heap_end = new_end;
615 }
616 // Notice that the heap can also shrink. However, this only happens
617 // during a Full GC (at least currently) and the entire marking
618 // phase will bail out and the task will not be restarted. So, let's
619 // do nothing.
620 }
621
622 void ConcurrentMark::reset() {
623 // Starting values for these two. This should be called in a STW
624 // phase. CM will be notified of any future g1_committed expansions
625 // will be at the end of evacuation pauses, when tasks are
626 // inactive.
627 MemRegion committed = _g1h->g1_committed();
628 _heap_start = committed.start();
629 _heap_end = committed.end();
630
631 // Separated the asserts so that we know which one fires.
632 assert(_heap_start != NULL, "heap bounds should look ok");
633 assert(_heap_end != NULL, "heap bounds should look ok");
634 assert(_heap_start < _heap_end, "heap bounds should look ok");
635
636 // Reset all the marking data structures and any necessary flags
637 reset_marking_state();
638
639 if (verbose_low()) {
640 gclog_or_tty->print_cr("[global] resetting");
641 }
642
643 // We do reset all of them, since different phases will use
644 // different number of active threads. So, it's easiest to have all
645 // of them ready.
646 for (int i = 0; i < (int) _max_task_num; ++i) {
647 _tasks[i]->reset(_nextMarkBitMap);
648 }
649
650 // we need this to make sure that the flag is on during the evac
651 // pause with initial mark piggy-backed
652 set_concurrent_marking_in_progress();
653 }
654
655
656 void ConcurrentMark::reset_marking_state(bool clear_overflow) {
657 _markStack.setEmpty();
658 _markStack.clear_overflow();
659 if (clear_overflow) {
660 clear_has_overflown();
661 } else {
662 assert(has_overflown(), "pre-condition");
663 }
664 _finger = _heap_start;
665
666 for (uint i = 0; i < _max_task_num; ++i) {
667 CMTaskQueue* queue = _task_queues->queue(i);
668 queue->set_empty();
669 }
670 }
671
672 void ConcurrentMark::set_concurrency(uint active_tasks) {
673 assert(active_tasks <= _max_task_num, "we should not have more");
674
675 _active_tasks = active_tasks;
676 // Need to update the three data structures below according to the
677 // number of active threads for this phase.
678 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
679 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
680 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
681 }
682
683 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
684 set_concurrency(active_tasks);
685
686 _concurrent = concurrent;
687 // We propagate this to all tasks, not just the active ones.
688 for (int i = 0; i < (int) _max_task_num; ++i)
689 _tasks[i]->set_concurrent(concurrent);
690
691 if (concurrent) {
692 set_concurrent_marking_in_progress();
693 } else {
694 // We currently assume that the concurrent flag has been set to
695 // false before we start remark. At this point we should also be
696 // in a STW phase.
697 assert(!concurrent_marking_in_progress(), "invariant");
698 assert(_finger == _heap_end,
699 err_msg("only way to get here: _finger: "PTR_FORMAT", _heap_end: "PTR_FORMAT,
700 _finger, _heap_end));
701 update_g1_committed(true);
702 }
703 }
704
705 void ConcurrentMark::set_non_marking_state() {
706 // We set the global marking state to some default values when we're
707 // not doing marking.
708 reset_marking_state();
709 _active_tasks = 0;
710 clear_concurrent_marking_in_progress();
711 }
712
713 ConcurrentMark::~ConcurrentMark() {
714 // The ConcurrentMark instance is never freed.
715 ShouldNotReachHere();
716 }
717
718 void ConcurrentMark::clearNextBitmap() {
719 G1CollectedHeap* g1h = G1CollectedHeap::heap();
720 G1CollectorPolicy* g1p = g1h->g1_policy();
721
722 // Make sure that the concurrent mark thread looks to still be in
723 // the current cycle.
724 guarantee(cmThread()->during_cycle(), "invariant");
725
726 // We are finishing up the current cycle by clearing the next
727 // marking bitmap and getting it ready for the next cycle. During
728 // this time no other cycle can start. So, let's make sure that this
729 // is the case.
730 guarantee(!g1h->mark_in_progress(), "invariant");
731
732 // clear the mark bitmap (no grey objects to start with).
733 // We need to do this in chunks and offer to yield in between
734 // each chunk.
735 HeapWord* start = _nextMarkBitMap->startWord();
736 HeapWord* end = _nextMarkBitMap->endWord();
737 HeapWord* cur = start;
738 size_t chunkSize = M;
739 while (cur < end) {
740 HeapWord* next = cur + chunkSize;
741 if (next > end) {
742 next = end;
743 }
744 MemRegion mr(cur,next);
745 _nextMarkBitMap->clearRange(mr);
746 cur = next;
747 do_yield_check();
748
749 // Repeat the asserts from above. We'll do them as asserts here to
750 // minimize their overhead on the product. However, we'll have
751 // them as guarantees at the beginning / end of the bitmap
752 // clearing to get some checking in the product.
753 assert(cmThread()->during_cycle(), "invariant");
754 assert(!g1h->mark_in_progress(), "invariant");
755 }
756
757 // Clear the liveness counting data
758 clear_all_count_data();
759
760 // Repeat the asserts from above.
761 guarantee(cmThread()->during_cycle(), "invariant");
762 guarantee(!g1h->mark_in_progress(), "invariant");
763 }
764
765 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
766 public:
767 bool doHeapRegion(HeapRegion* r) {
768 if (!r->continuesHumongous()) {
769 r->note_start_of_marking();
770 }
771 return false;
772 }
773 };
774
775 void ConcurrentMark::checkpointRootsInitialPre() {
776 G1CollectedHeap* g1h = G1CollectedHeap::heap();
777 G1CollectorPolicy* g1p = g1h->g1_policy();
778
779 _has_aborted = false;
780
781 #ifndef PRODUCT
782 if (G1PrintReachableAtInitialMark) {
783 print_reachable("at-cycle-start",
784 VerifyOption_G1UsePrevMarking, true /* all */);
785 }
786 #endif
787
788 // Initialise marking structures. This has to be done in a STW phase.
789 reset();
790
791 // For each region note start of marking.
792 NoteStartOfMarkHRClosure startcl;
793 g1h->heap_region_iterate(&startcl);
794 }
795
796
797 void ConcurrentMark::checkpointRootsInitialPost() {
798 G1CollectedHeap* g1h = G1CollectedHeap::heap();
799
800 // If we force an overflow during remark, the remark operation will
801 // actually abort and we'll restart concurrent marking. If we always
802 // force an oveflow during remark we'll never actually complete the
803 // marking phase. So, we initilize this here, at the start of the
804 // cycle, so that at the remaining overflow number will decrease at
805 // every remark and we'll eventually not need to cause one.
806 force_overflow_stw()->init();
807
808 // Start Concurrent Marking weak-reference discovery.
809 ReferenceProcessor* rp = g1h->ref_processor_cm();
810 // enable ("weak") refs discovery
811 rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
812 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
813
814 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
815 // This is the start of the marking cycle, we're expected all
816 // threads to have SATB queues with active set to false.
817 satb_mq_set.set_active_all_threads(true, /* new active value */
818 false /* expected_active */);
819
820 _root_regions.prepare_for_scan();
821
822 // update_g1_committed() will be called at the end of an evac pause
823 // when marking is on. So, it's also called at the end of the
824 // initial-mark pause to update the heap end, if the heap expands
825 // during it. No need to call it here.
826 }
827
828 /*
829 * Notice that in the next two methods, we actually leave the STS
830 * during the barrier sync and join it immediately afterwards. If we
831 * do not do this, the following deadlock can occur: one thread could
832 * be in the barrier sync code, waiting for the other thread to also
833 * sync up, whereas another one could be trying to yield, while also
834 * waiting for the other threads to sync up too.
835 *
836 * Note, however, that this code is also used during remark and in
837 * this case we should not attempt to leave / enter the STS, otherwise
838 * we'll either hit an asseert (debug / fastdebug) or deadlock
839 * (product). So we should only leave / enter the STS if we are
840 * operating concurrently.
841 *
842 * Because the thread that does the sync barrier has left the STS, it
843 * is possible to be suspended for a Full GC or an evacuation pause
844 * could occur. This is actually safe, since the entering the sync
845 * barrier is one of the last things do_marking_step() does, and it
846 * doesn't manipulate any data structures afterwards.
847 */
848
849 void ConcurrentMark::enter_first_sync_barrier(int task_num) {
850 if (verbose_low()) {
851 gclog_or_tty->print_cr("[%d] entering first barrier", task_num);
852 }
853
854 if (concurrent()) {
855 ConcurrentGCThread::stsLeave();
856 }
857 _first_overflow_barrier_sync.enter();
858 if (concurrent()) {
859 ConcurrentGCThread::stsJoin();
860 }
861 // at this point everyone should have synced up and not be doing any
862 // more work
863
864 if (verbose_low()) {
865 gclog_or_tty->print_cr("[%d] leaving first barrier", task_num);
866 }
867
868 // If we're executing the concurrent phase of marking, reset the marking
869 // state; otherwise the marking state is reset after reference processing,
870 // during the remark pause.
871 // If we reset here as a result of an overflow during the remark we will
872 // see assertion failures from any subsequent set_concurrency_and_phase()
873 // calls.
874 if (concurrent()) {
875 // let the task 0 do this
876 if (task_num == 0) {
877 // task 0 is responsible for clearing the global data structures
878 // We should be here because of an overflow. During STW we should
879 // not clear the overflow flag since we rely on it being true when
880 // we exit this method to abort the pause and restart concurent
881 // marking.
882 reset_marking_state(true /* clear_overflow */);
883 force_overflow()->update();
884
885 if (G1Log::fine()) {
886 gclog_or_tty->date_stamp(PrintGCDateStamps);
887 gclog_or_tty->stamp(PrintGCTimeStamps);
888 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]");
889 }
890 }
891 }
892
893 // after this, each task should reset its own data structures then
894 // then go into the second barrier
895 }
896
897 void ConcurrentMark::enter_second_sync_barrier(int task_num) {
898 if (verbose_low()) {
899 gclog_or_tty->print_cr("[%d] entering second barrier", task_num);
900 }
901
902 if (concurrent()) {
903 ConcurrentGCThread::stsLeave();
904 }
905 _second_overflow_barrier_sync.enter();
906 if (concurrent()) {
907 ConcurrentGCThread::stsJoin();
908 }
909 // at this point everything should be re-initialized and ready to go
910
911 if (verbose_low()) {
912 gclog_or_tty->print_cr("[%d] leaving second barrier", task_num);
913 }
914 }
915
916 #ifndef PRODUCT
917 void ForceOverflowSettings::init() {
918 _num_remaining = G1ConcMarkForceOverflow;
919 _force = false;
920 update();
921 }
922
923 void ForceOverflowSettings::update() {
924 if (_num_remaining > 0) {
925 _num_remaining -= 1;
926 _force = true;
927 } else {
928 _force = false;
929 }
930 }
931
932 bool ForceOverflowSettings::should_force() {
933 if (_force) {
934 _force = false;
935 return true;
936 } else {
937 return false;
938 }
939 }
940 #endif // !PRODUCT
941
942 class CMConcurrentMarkingTask: public AbstractGangTask {
943 private:
944 ConcurrentMark* _cm;
945 ConcurrentMarkThread* _cmt;
946
947 public:
948 void work(uint worker_id) {
949 assert(Thread::current()->is_ConcurrentGC_thread(),
950 "this should only be done by a conc GC thread");
951 ResourceMark rm;
952
953 double start_vtime = os::elapsedVTime();
954
955 ConcurrentGCThread::stsJoin();
956
957 assert(worker_id < _cm->active_tasks(), "invariant");
958 CMTask* the_task = _cm->task(worker_id);
959 the_task->record_start_time();
960 if (!_cm->has_aborted()) {
961 do {
962 double start_vtime_sec = os::elapsedVTime();
963 double start_time_sec = os::elapsedTime();
964 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
965
966 the_task->do_marking_step(mark_step_duration_ms,
967 true /* do_termination */,
968 false /* is_serial*/);
969
970 double end_time_sec = os::elapsedTime();
971 double end_vtime_sec = os::elapsedVTime();
972 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
973 double elapsed_time_sec = end_time_sec - start_time_sec;
974 _cm->clear_has_overflown();
975
976 bool ret = _cm->do_yield_check(worker_id);
977
978 jlong sleep_time_ms;
979 if (!_cm->has_aborted() && the_task->has_aborted()) {
980 sleep_time_ms =
981 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
982 ConcurrentGCThread::stsLeave();
983 os::sleep(Thread::current(), sleep_time_ms, false);
984 ConcurrentGCThread::stsJoin();
985 }
986 double end_time2_sec = os::elapsedTime();
987 double elapsed_time2_sec = end_time2_sec - start_time_sec;
988
989 #if 0
990 gclog_or_tty->print_cr("CM: elapsed %1.4lf ms, sleep %1.4lf ms, "
991 "overhead %1.4lf",
992 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms,
993 the_task->conc_overhead(os::elapsedTime()) * 8.0);
994 gclog_or_tty->print_cr("elapsed time %1.4lf ms, time 2: %1.4lf ms",
995 elapsed_time_sec * 1000.0, elapsed_time2_sec * 1000.0);
996 #endif
997 } while (!_cm->has_aborted() && the_task->has_aborted());
998 }
999 the_task->record_end_time();
1000 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
1001
1002 ConcurrentGCThread::stsLeave();
1003
1004 double end_vtime = os::elapsedVTime();
1005 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
1006 }
1007
1008 CMConcurrentMarkingTask(ConcurrentMark* cm,
1009 ConcurrentMarkThread* cmt) :
1010 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
1011
1012 ~CMConcurrentMarkingTask() { }
1013 };
1014
1015 // Calculates the number of active workers for a concurrent
1016 // phase.
1017 uint ConcurrentMark::calc_parallel_marking_threads() {
1018 if (G1CollectedHeap::use_parallel_gc_threads()) {
1019 uint n_conc_workers = 0;
1020 if (!UseDynamicNumberOfGCThreads ||
1021 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
1022 !ForceDynamicNumberOfGCThreads)) {
1023 n_conc_workers = max_parallel_marking_threads();
1024 } else {
1025 n_conc_workers =
1026 AdaptiveSizePolicy::calc_default_active_workers(
1027 max_parallel_marking_threads(),
1028 1, /* Minimum workers */
1029 parallel_marking_threads(),
1030 Threads::number_of_non_daemon_threads());
1031 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
1032 // that scaling has already gone into "_max_parallel_marking_threads".
1033 }
1034 assert(n_conc_workers > 0, "Always need at least 1");
1035 return n_conc_workers;
1036 }
1037 // If we are not running with any parallel GC threads we will not
1038 // have spawned any marking threads either. Hence the number of
1039 // concurrent workers should be 0.
1040 return 0;
1041 }
1042
1043 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
1044 // Currently, only survivors can be root regions.
1045 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
1046 G1RootRegionScanClosure cl(_g1h, this, worker_id);
1047
1048 const uintx interval = PrefetchScanIntervalInBytes;
1049 HeapWord* curr = hr->bottom();
1050 const HeapWord* end = hr->top();
1051 while (curr < end) {
1052 Prefetch::read(curr, interval);
1053 oop obj = oop(curr);
1054 int size = obj->oop_iterate(&cl);
1055 assert(size == obj->size(), "sanity");
1056 curr += size;
1057 }
1058 }
1059
1060 class CMRootRegionScanTask : public AbstractGangTask {
1061 private:
1062 ConcurrentMark* _cm;
1063
1064 public:
1065 CMRootRegionScanTask(ConcurrentMark* cm) :
1066 AbstractGangTask("Root Region Scan"), _cm(cm) { }
1067
1068 void work(uint worker_id) {
1069 assert(Thread::current()->is_ConcurrentGC_thread(),
1070 "this should only be done by a conc GC thread");
1071
1072 CMRootRegions* root_regions = _cm->root_regions();
1073 HeapRegion* hr = root_regions->claim_next();
1074 while (hr != NULL) {
1075 _cm->scanRootRegion(hr, worker_id);
1076 hr = root_regions->claim_next();
1077 }
1078 }
1079 };
1080
1081 void ConcurrentMark::scanRootRegions() {
1082 // scan_in_progress() will have been set to true only if there was
1083 // at least one root region to scan. So, if it's false, we
1084 // should not attempt to do any further work.
1085 if (root_regions()->scan_in_progress()) {
1086 _parallel_marking_threads = calc_parallel_marking_threads();
1087 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1088 "Maximum number of marking threads exceeded");
1089 uint active_workers = MAX2(1U, parallel_marking_threads());
1090
1091 CMRootRegionScanTask task(this);
1092 if (use_parallel_marking_threads()) {
1093 _parallel_workers->set_active_workers((int) active_workers);
1094 _parallel_workers->run_task(&task);
1095 } else {
1096 task.work(0);
1097 }
1098
1099 // It's possible that has_aborted() is true here without actually
1100 // aborting the survivor scan earlier. This is OK as it's
1101 // mainly used for sanity checking.
1102 root_regions()->scan_finished();
1103 }
1104 }
1105
1106 void ConcurrentMark::markFromRoots() {
1107 // we might be tempted to assert that:
1108 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1109 // "inconsistent argument?");
1110 // However that wouldn't be right, because it's possible that
1111 // a safepoint is indeed in progress as a younger generation
1112 // stop-the-world GC happens even as we mark in this generation.
1113
1114 _restart_for_overflow = false;
1115 force_overflow_conc()->init();
1116
1117 // _g1h has _n_par_threads
1118 _parallel_marking_threads = calc_parallel_marking_threads();
1119 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1120 "Maximum number of marking threads exceeded");
1121
1122 uint active_workers = MAX2(1U, parallel_marking_threads());
1123
1124 // Parallel task terminator is set in "set_concurrency_and_phase()"
1125 set_concurrency_and_phase(active_workers, true /* concurrent */);
1126
1127 CMConcurrentMarkingTask markingTask(this, cmThread());
1128 if (use_parallel_marking_threads()) {
1129 _parallel_workers->set_active_workers((int)active_workers);
1130 // Don't set _n_par_threads because it affects MT in proceess_strong_roots()
1131 // and the decisions on that MT processing is made elsewhere.
1132 assert(_parallel_workers->active_workers() > 0, "Should have been set");
1133 _parallel_workers->run_task(&markingTask);
1134 } else {
1135 markingTask.work(0);
1136 }
1137 print_stats();
1138 }
1139
1140 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1141 // world is stopped at this checkpoint
1142 assert(SafepointSynchronize::is_at_safepoint(),
1143 "world should be stopped");
1144
1145 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1146
1147 // If a full collection has happened, we shouldn't do this.
1148 if (has_aborted()) {
1149 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1150 return;
1151 }
1152
1153 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1154
1155 if (VerifyDuringGC) {
1156 HandleMark hm; // handle scope
1157 gclog_or_tty->print(" VerifyDuringGC:(before)");
1158 Universe::heap()->prepare_for_verify();
1159 Universe::verify(/* silent */ false,
1160 /* option */ VerifyOption_G1UsePrevMarking);
1161 }
1162
1163 G1CollectorPolicy* g1p = g1h->g1_policy();
1164 g1p->record_concurrent_mark_remark_start();
1165
1166 double start = os::elapsedTime();
1167
1168 checkpointRootsFinalWork();
1169
1170 double mark_work_end = os::elapsedTime();
1171
1172 weakRefsWork(clear_all_soft_refs);
1173
1174 if (has_overflown()) {
1175 // Oops. We overflowed. Restart concurrent marking.
1176 _restart_for_overflow = true;
1177 if (G1TraceMarkStackOverflow) {
1178 gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
1179 }
1180
1181 // Verify the heap w.r.t. the previous marking bitmap.
1182 if (VerifyDuringGC) {
1183 HandleMark hm; // handle scope
1184 gclog_or_tty->print(" VerifyDuringGC:(overflow)");
1185 Universe::heap()->prepare_for_verify();
1186 Universe::verify(/* silent */ false,
1187 /* option */ VerifyOption_G1UsePrevMarking);
1188 }
1189
1190 // Clear the marking state because we will be restarting
1191 // marking due to overflowing the global mark stack.
1192 reset_marking_state();
1193 } else {
1194 // Aggregate the per-task counting data that we have accumulated
1195 // while marking.
1196 aggregate_count_data();
1197
1198 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1199 // We're done with marking.
1200 // This is the end of the marking cycle, we're expected all
1201 // threads to have SATB queues with active set to true.
1202 satb_mq_set.set_active_all_threads(false, /* new active value */
1203 true /* expected_active */);
1204
1205 if (VerifyDuringGC) {
1206 HandleMark hm; // handle scope
1207 gclog_or_tty->print(" VerifyDuringGC:(after)");
1208 Universe::heap()->prepare_for_verify();
1209 Universe::verify(/* silent */ false,
1210 /* option */ VerifyOption_G1UseNextMarking);
1211 }
1212 assert(!restart_for_overflow(), "sanity");
1213 // Completely reset the marking state since marking completed
1214 set_non_marking_state();
1215 }
1216
1217 #if VERIFY_OBJS_PROCESSED
1218 _scan_obj_cl.objs_processed = 0;
1219 ThreadLocalObjQueue::objs_enqueued = 0;
1220 #endif
1221
1222 // Statistics
1223 double now = os::elapsedTime();
1224 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1225 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1226 _remark_times.add((now - start) * 1000.0);
1227
1228 g1p->record_concurrent_mark_remark_end();
1229
1230 G1CMIsAliveClosure is_alive(g1h);
1231 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive);
1232 }
1233
1234 // Base class of the closures that finalize and verify the
1235 // liveness counting data.
1236 class CMCountDataClosureBase: public HeapRegionClosure {
1237 protected:
1238 G1CollectedHeap* _g1h;
1239 ConcurrentMark* _cm;
1240 CardTableModRefBS* _ct_bs;
1241
1242 BitMap* _region_bm;
1243 BitMap* _card_bm;
1244
1245 // Takes a region that's not empty (i.e., it has at least one
1246 // live object in it and sets its corresponding bit on the region
1247 // bitmap to 1. If the region is "starts humongous" it will also set
1248 // to 1 the bits on the region bitmap that correspond to its
1249 // associated "continues humongous" regions.
1250 void set_bit_for_region(HeapRegion* hr) {
1251 assert(!hr->continuesHumongous(), "should have filtered those out");
1252
1253 BitMap::idx_t index = (BitMap::idx_t) hr->hrs_index();
1254 if (!hr->startsHumongous()) {
1255 // Normal (non-humongous) case: just set the bit.
1256 _region_bm->par_at_put(index, true);
1257 } else {
1258 // Starts humongous case: calculate how many regions are part of
1259 // this humongous region and then set the bit range.
1260 BitMap::idx_t end_index = (BitMap::idx_t) hr->last_hc_index();
1261 _region_bm->par_at_put_range(index, end_index, true);
1262 }
1263 }
1264
1265 public:
1266 CMCountDataClosureBase(G1CollectedHeap* g1h,
1267 BitMap* region_bm, BitMap* card_bm):
1268 _g1h(g1h), _cm(g1h->concurrent_mark()),
1269 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())),
1270 _region_bm(region_bm), _card_bm(card_bm) { }
1271 };
1272
1273 // Closure that calculates the # live objects per region. Used
1274 // for verification purposes during the cleanup pause.
1275 class CalcLiveObjectsClosure: public CMCountDataClosureBase {
1276 CMBitMapRO* _bm;
1277 size_t _region_marked_bytes;
1278
1279 public:
1280 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h,
1281 BitMap* region_bm, BitMap* card_bm) :
1282 CMCountDataClosureBase(g1h, region_bm, card_bm),
1283 _bm(bm), _region_marked_bytes(0) { }
1284
1285 bool doHeapRegion(HeapRegion* hr) {
1286
1287 if (hr->continuesHumongous()) {
1288 // We will ignore these here and process them when their
1289 // associated "starts humongous" region is processed (see
1290 // set_bit_for_heap_region()). Note that we cannot rely on their
1291 // associated "starts humongous" region to have their bit set to
1292 // 1 since, due to the region chunking in the parallel region
1293 // iteration, a "continues humongous" region might be visited
1294 // before its associated "starts humongous".
1295 return false;
1296 }
1297
1298 HeapWord* ntams = hr->next_top_at_mark_start();
1299 HeapWord* start = hr->bottom();
1300
1301 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1302 err_msg("Preconditions not met - "
1303 "start: "PTR_FORMAT", ntams: "PTR_FORMAT", end: "PTR_FORMAT,
1304 start, ntams, hr->end()));
1305
1306 // Find the first marked object at or after "start".
1307 start = _bm->getNextMarkedWordAddress(start, ntams);
1308
1309 size_t marked_bytes = 0;
1310
1311 while (start < ntams) {
1312 oop obj = oop(start);
1313 int obj_sz = obj->size();
1314 HeapWord* obj_end = start + obj_sz;
1315
1316 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
1317 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end);
1318
1319 // Note: if we're looking at the last region in heap - obj_end
1320 // could be actually just beyond the end of the heap; end_idx
1321 // will then correspond to a (non-existent) card that is also
1322 // just beyond the heap.
1323 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) {
1324 // end of object is not card aligned - increment to cover
1325 // all the cards spanned by the object
1326 end_idx += 1;
1327 }
1328
1329 // Set the bits in the card BM for the cards spanned by this object.
1330 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1331
1332 // Add the size of this object to the number of marked bytes.
1333 marked_bytes += (size_t)obj_sz * HeapWordSize;
1334
1335 // Find the next marked object after this one.
1336 start = _bm->getNextMarkedWordAddress(obj_end, ntams);
1337 }
1338
1339 // Mark the allocated-since-marking portion...
1340 HeapWord* top = hr->top();
1341 if (ntams < top) {
1342 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1343 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1344
1345 // Note: if we're looking at the last region in heap - top
1346 // could be actually just beyond the end of the heap; end_idx
1347 // will then correspond to a (non-existent) card that is also
1348 // just beyond the heap.
1349 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1350 // end of object is not card aligned - increment to cover
1351 // all the cards spanned by the object
1352 end_idx += 1;
1353 }
1354 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1355
1356 // This definitely means the region has live objects.
1357 set_bit_for_region(hr);
1358 }
1359
1360 // Update the live region bitmap.
1361 if (marked_bytes > 0) {
1362 set_bit_for_region(hr);
1363 }
1364
1365 // Set the marked bytes for the current region so that
1366 // it can be queried by a calling verificiation routine
1367 _region_marked_bytes = marked_bytes;
1368
1369 return false;
1370 }
1371
1372 size_t region_marked_bytes() const { return _region_marked_bytes; }
1373 };
1374
1375 // Heap region closure used for verifying the counting data
1376 // that was accumulated concurrently and aggregated during
1377 // the remark pause. This closure is applied to the heap
1378 // regions during the STW cleanup pause.
1379
1380 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure {
1381 G1CollectedHeap* _g1h;
1382 ConcurrentMark* _cm;
1383 CalcLiveObjectsClosure _calc_cl;
1384 BitMap* _region_bm; // Region BM to be verified
1385 BitMap* _card_bm; // Card BM to be verified
1386 bool _verbose; // verbose output?
1387
1388 BitMap* _exp_region_bm; // Expected Region BM values
1389 BitMap* _exp_card_bm; // Expected card BM values
1390
1391 int _failures;
1392
1393 public:
1394 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h,
1395 BitMap* region_bm,
1396 BitMap* card_bm,
1397 BitMap* exp_region_bm,
1398 BitMap* exp_card_bm,
1399 bool verbose) :
1400 _g1h(g1h), _cm(g1h->concurrent_mark()),
1401 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm),
1402 _region_bm(region_bm), _card_bm(card_bm), _verbose(verbose),
1403 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm),
1404 _failures(0) { }
1405
1406 int failures() const { return _failures; }
1407
1408 bool doHeapRegion(HeapRegion* hr) {
1409 if (hr->continuesHumongous()) {
1410 // We will ignore these here and process them when their
1411 // associated "starts humongous" region is processed (see
1412 // set_bit_for_heap_region()). Note that we cannot rely on their
1413 // associated "starts humongous" region to have their bit set to
1414 // 1 since, due to the region chunking in the parallel region
1415 // iteration, a "continues humongous" region might be visited
1416 // before its associated "starts humongous".
1417 return false;
1418 }
1419
1420 int failures = 0;
1421
1422 // Call the CalcLiveObjectsClosure to walk the marking bitmap for
1423 // this region and set the corresponding bits in the expected region
1424 // and card bitmaps.
1425 bool res = _calc_cl.doHeapRegion(hr);
1426 assert(res == false, "should be continuing");
1427
1428 MutexLockerEx x((_verbose ? ParGCRareEvent_lock : NULL),
1429 Mutex::_no_safepoint_check_flag);
1430
1431 // Verify the marked bytes for this region.
1432 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1433 size_t act_marked_bytes = hr->next_marked_bytes();
1434
1435 // We're not OK if expected marked bytes > actual marked bytes. It means
1436 // we have missed accounting some objects during the actual marking.
1437 if (exp_marked_bytes > act_marked_bytes) {
1438 if (_verbose) {
1439 gclog_or_tty->print_cr("Region %u: marked bytes mismatch: "
1440 "expected: " SIZE_FORMAT ", actual: " SIZE_FORMAT,
1441 hr->hrs_index(), exp_marked_bytes, act_marked_bytes);
1442 }
1443 failures += 1;
1444 }
1445
1446 // Verify the bit, for this region, in the actual and expected
1447 // (which was just calculated) region bit maps.
1448 // We're not OK if the bit in the calculated expected region
1449 // bitmap is set and the bit in the actual region bitmap is not.
1450 BitMap::idx_t index = (BitMap::idx_t) hr->hrs_index();
1451
1452 bool expected = _exp_region_bm->at(index);
1453 bool actual = _region_bm->at(index);
1454 if (expected && !actual) {
1455 if (_verbose) {
1456 gclog_or_tty->print_cr("Region %u: region bitmap mismatch: "
1457 "expected: %s, actual: %s",
1458 hr->hrs_index(),
1459 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1460 }
1461 failures += 1;
1462 }
1463
1464 // Verify that the card bit maps for the cards spanned by the current
1465 // region match. We have an error if we have a set bit in the expected
1466 // bit map and the corresponding bit in the actual bitmap is not set.
1467
1468 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom());
1469 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top());
1470
1471 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1472 expected = _exp_card_bm->at(i);
1473 actual = _card_bm->at(i);
1474
1475 if (expected && !actual) {
1476 if (_verbose) {
1477 gclog_or_tty->print_cr("Region %u: card bitmap mismatch at " SIZE_FORMAT ": "
1478 "expected: %s, actual: %s",
1479 hr->hrs_index(), i,
1480 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1481 }
1482 failures += 1;
1483 }
1484 }
1485
1486 if (failures > 0 && _verbose) {
1487 gclog_or_tty->print_cr("Region " HR_FORMAT ", ntams: " PTR_FORMAT ", "
1488 "marked_bytes: calc/actual " SIZE_FORMAT "/" SIZE_FORMAT,
1489 HR_FORMAT_PARAMS(hr), hr->next_top_at_mark_start(),
1490 _calc_cl.region_marked_bytes(), hr->next_marked_bytes());
1491 }
1492
1493 _failures += failures;
1494
1495 // We could stop iteration over the heap when we
1496 // find the first violating region by returning true.
1497 return false;
1498 }
1499 };
1500
1501
1502 class G1ParVerifyFinalCountTask: public AbstractGangTask {
1503 protected:
1504 G1CollectedHeap* _g1h;
1505 ConcurrentMark* _cm;
1506 BitMap* _actual_region_bm;
1507 BitMap* _actual_card_bm;
1508
1509 uint _n_workers;
1510
1511 BitMap* _expected_region_bm;
1512 BitMap* _expected_card_bm;
1513
1514 int _failures;
1515 bool _verbose;
1516
1517 public:
1518 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h,
1519 BitMap* region_bm, BitMap* card_bm,
1520 BitMap* expected_region_bm, BitMap* expected_card_bm)
1521 : AbstractGangTask("G1 verify final counting"),
1522 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1523 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1524 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm),
1525 _failures(0), _verbose(false),
1526 _n_workers(0) {
1527 assert(VerifyDuringGC, "don't call this otherwise");
1528
1529 // Use the value already set as the number of active threads
1530 // in the call to run_task().
1531 if (G1CollectedHeap::use_parallel_gc_threads()) {
1532 assert( _g1h->workers()->active_workers() > 0,
1533 "Should have been previously set");
1534 _n_workers = _g1h->workers()->active_workers();
1535 } else {
1536 _n_workers = 1;
1537 }
1538
1539 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity");
1540 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity");
1541
1542 _verbose = _cm->verbose_medium();
1543 }
1544
1545 void work(uint worker_id) {
1546 assert(worker_id < _n_workers, "invariant");
1547
1548 VerifyLiveObjectDataHRClosure verify_cl(_g1h,
1549 _actual_region_bm, _actual_card_bm,
1550 _expected_region_bm,
1551 _expected_card_bm,
1552 _verbose);
1553
1554 if (G1CollectedHeap::use_parallel_gc_threads()) {
1555 _g1h->heap_region_par_iterate_chunked(&verify_cl,
1556 worker_id,
1557 _n_workers,
1558 HeapRegion::VerifyCountClaimValue);
1559 } else {
1560 _g1h->heap_region_iterate(&verify_cl);
1561 }
1562
1563 Atomic::add(verify_cl.failures(), &_failures);
1564 }
1565
1566 int failures() const { return _failures; }
1567 };
1568
1569 // Closure that finalizes the liveness counting data.
1570 // Used during the cleanup pause.
1571 // Sets the bits corresponding to the interval [NTAMS, top]
1572 // (which contains the implicitly live objects) in the
1573 // card liveness bitmap. Also sets the bit for each region,
1574 // containing live data, in the region liveness bitmap.
1575
1576 class FinalCountDataUpdateClosure: public CMCountDataClosureBase {
1577 public:
1578 FinalCountDataUpdateClosure(G1CollectedHeap* g1h,
1579 BitMap* region_bm,
1580 BitMap* card_bm) :
1581 CMCountDataClosureBase(g1h, region_bm, card_bm) { }
1582
1583 bool doHeapRegion(HeapRegion* hr) {
1584
1585 if (hr->continuesHumongous()) {
1586 // We will ignore these here and process them when their
1587 // associated "starts humongous" region is processed (see
1588 // set_bit_for_heap_region()). Note that we cannot rely on their
1589 // associated "starts humongous" region to have their bit set to
1590 // 1 since, due to the region chunking in the parallel region
1591 // iteration, a "continues humongous" region might be visited
1592 // before its associated "starts humongous".
1593 return false;
1594 }
1595
1596 HeapWord* ntams = hr->next_top_at_mark_start();
1597 HeapWord* top = hr->top();
1598
1599 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1600
1601 // Mark the allocated-since-marking portion...
1602 if (ntams < top) {
1603 // This definitely means the region has live objects.
1604 set_bit_for_region(hr);
1605
1606 // Now set the bits in the card bitmap for [ntams, top)
1607 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1608 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1609
1610 // Note: if we're looking at the last region in heap - top
1611 // could be actually just beyond the end of the heap; end_idx
1612 // will then correspond to a (non-existent) card that is also
1613 // just beyond the heap.
1614 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1615 // end of object is not card aligned - increment to cover
1616 // all the cards spanned by the object
1617 end_idx += 1;
1618 }
1619
1620 assert(end_idx <= _card_bm->size(),
1621 err_msg("oob: end_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1622 end_idx, _card_bm->size()));
1623 assert(start_idx < _card_bm->size(),
1624 err_msg("oob: start_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1625 start_idx, _card_bm->size()));
1626
1627 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1628 }
1629
1630 // Set the bit for the region if it contains live data
1631 if (hr->next_marked_bytes() > 0) {
1632 set_bit_for_region(hr);
1633 }
1634
1635 return false;
1636 }
1637 };
1638
1639 class G1ParFinalCountTask: public AbstractGangTask {
1640 protected:
1641 G1CollectedHeap* _g1h;
1642 ConcurrentMark* _cm;
1643 BitMap* _actual_region_bm;
1644 BitMap* _actual_card_bm;
1645
1646 uint _n_workers;
1647
1648 public:
1649 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm)
1650 : AbstractGangTask("G1 final counting"),
1651 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1652 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1653 _n_workers(0) {
1654 // Use the value already set as the number of active threads
1655 // in the call to run_task().
1656 if (G1CollectedHeap::use_parallel_gc_threads()) {
1657 assert( _g1h->workers()->active_workers() > 0,
1658 "Should have been previously set");
1659 _n_workers = _g1h->workers()->active_workers();
1660 } else {
1661 _n_workers = 1;
1662 }
1663 }
1664
1665 void work(uint worker_id) {
1666 assert(worker_id < _n_workers, "invariant");
1667
1668 FinalCountDataUpdateClosure final_update_cl(_g1h,
1669 _actual_region_bm,
1670 _actual_card_bm);
1671
1672 if (G1CollectedHeap::use_parallel_gc_threads()) {
1673 _g1h->heap_region_par_iterate_chunked(&final_update_cl,
1674 worker_id,
1675 _n_workers,
1676 HeapRegion::FinalCountClaimValue);
1677 } else {
1678 _g1h->heap_region_iterate(&final_update_cl);
1679 }
1680 }
1681 };
1682
1683 class G1ParNoteEndTask;
1684
1685 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1686 G1CollectedHeap* _g1;
1687 int _worker_num;
1688 size_t _max_live_bytes;
1689 uint _regions_claimed;
1690 size_t _freed_bytes;
1691 FreeRegionList* _local_cleanup_list;
1692 OldRegionSet* _old_proxy_set;
1693 HumongousRegionSet* _humongous_proxy_set;
1694 HRRSCleanupTask* _hrrs_cleanup_task;
1695 double _claimed_region_time;
1696 double _max_region_time;
1697
1698 public:
1699 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1700 int worker_num,
1701 FreeRegionList* local_cleanup_list,
1702 OldRegionSet* old_proxy_set,
1703 HumongousRegionSet* humongous_proxy_set,
1704 HRRSCleanupTask* hrrs_cleanup_task) :
1705 _g1(g1), _worker_num(worker_num),
1706 _max_live_bytes(0), _regions_claimed(0),
1707 _freed_bytes(0),
1708 _claimed_region_time(0.0), _max_region_time(0.0),
1709 _local_cleanup_list(local_cleanup_list),
1710 _old_proxy_set(old_proxy_set),
1711 _humongous_proxy_set(humongous_proxy_set),
1712 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1713
1714 size_t freed_bytes() { return _freed_bytes; }
1715
1716 bool doHeapRegion(HeapRegion *hr) {
1717 if (hr->continuesHumongous()) {
1718 return false;
1719 }
1720 // We use a claim value of zero here because all regions
1721 // were claimed with value 1 in the FinalCount task.
1722 _g1->reset_gc_time_stamps(hr);
1723 double start = os::elapsedTime();
1724 _regions_claimed++;
1725 hr->note_end_of_marking();
1726 _max_live_bytes += hr->max_live_bytes();
1727 _g1->free_region_if_empty(hr,
1728 &_freed_bytes,
1729 _local_cleanup_list,
1730 _old_proxy_set,
1731 _humongous_proxy_set,
1732 _hrrs_cleanup_task,
1733 true /* par */);
1734 double region_time = (os::elapsedTime() - start);
1735 _claimed_region_time += region_time;
1736 if (region_time > _max_region_time) {
1737 _max_region_time = region_time;
1738 }
1739 return false;
1740 }
1741
1742 size_t max_live_bytes() { return _max_live_bytes; }
1743 uint regions_claimed() { return _regions_claimed; }
1744 double claimed_region_time_sec() { return _claimed_region_time; }
1745 double max_region_time_sec() { return _max_region_time; }
1746 };
1747
1748 class G1ParNoteEndTask: public AbstractGangTask {
1749 friend class G1NoteEndOfConcMarkClosure;
1750
1751 protected:
1752 G1CollectedHeap* _g1h;
1753 size_t _max_live_bytes;
1754 size_t _freed_bytes;
1755 FreeRegionList* _cleanup_list;
1756
1757 public:
1758 G1ParNoteEndTask(G1CollectedHeap* g1h,
1759 FreeRegionList* cleanup_list) :
1760 AbstractGangTask("G1 note end"), _g1h(g1h),
1761 _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list) { }
1762
1763 void work(uint worker_id) {
1764 double start = os::elapsedTime();
1765 FreeRegionList local_cleanup_list("Local Cleanup List");
1766 OldRegionSet old_proxy_set("Local Cleanup Old Proxy Set");
1767 HumongousRegionSet humongous_proxy_set("Local Cleanup Humongous Proxy Set");
1768 HRRSCleanupTask hrrs_cleanup_task;
1769 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, worker_id, &local_cleanup_list,
1770 &old_proxy_set,
1771 &humongous_proxy_set,
1772 &hrrs_cleanup_task);
1773 if (G1CollectedHeap::use_parallel_gc_threads()) {
1774 _g1h->heap_region_par_iterate_chunked(&g1_note_end, worker_id,
1775 _g1h->workers()->active_workers(),
1776 HeapRegion::NoteEndClaimValue);
1777 } else {
1778 _g1h->heap_region_iterate(&g1_note_end);
1779 }
1780 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1781
1782 // Now update the lists
1783 _g1h->update_sets_after_freeing_regions(g1_note_end.freed_bytes(),
1784 NULL /* free_list */,
1785 &old_proxy_set,
1786 &humongous_proxy_set,
1787 true /* par */);
1788 {
1789 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1790 _max_live_bytes += g1_note_end.max_live_bytes();
1791 _freed_bytes += g1_note_end.freed_bytes();
1792
1793 // If we iterate over the global cleanup list at the end of
1794 // cleanup to do this printing we will not guarantee to only
1795 // generate output for the newly-reclaimed regions (the list
1796 // might not be empty at the beginning of cleanup; we might
1797 // still be working on its previous contents). So we do the
1798 // printing here, before we append the new regions to the global
1799 // cleanup list.
1800
1801 G1HRPrinter* hr_printer = _g1h->hr_printer();
1802 if (hr_printer->is_active()) {
1803 HeapRegionLinkedListIterator iter(&local_cleanup_list);
1804 while (iter.more_available()) {
1805 HeapRegion* hr = iter.get_next();
1806 hr_printer->cleanup(hr);
1807 }
1808 }
1809
1810 _cleanup_list->add_as_tail(&local_cleanup_list);
1811 assert(local_cleanup_list.is_empty(), "post-condition");
1812
1813 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1814 }
1815 }
1816 size_t max_live_bytes() { return _max_live_bytes; }
1817 size_t freed_bytes() { return _freed_bytes; }
1818 };
1819
1820 class G1ParScrubRemSetTask: public AbstractGangTask {
1821 protected:
1822 G1RemSet* _g1rs;
1823 BitMap* _region_bm;
1824 BitMap* _card_bm;
1825 public:
1826 G1ParScrubRemSetTask(G1CollectedHeap* g1h,
1827 BitMap* region_bm, BitMap* card_bm) :
1828 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()),
1829 _region_bm(region_bm), _card_bm(card_bm) { }
1830
1831 void work(uint worker_id) {
1832 if (G1CollectedHeap::use_parallel_gc_threads()) {
1833 _g1rs->scrub_par(_region_bm, _card_bm, worker_id,
1834 HeapRegion::ScrubRemSetClaimValue);
1835 } else {
1836 _g1rs->scrub(_region_bm, _card_bm);
1837 }
1838 }
1839
1840 };
1841
1842 void ConcurrentMark::cleanup() {
1843 // world is stopped at this checkpoint
1844 assert(SafepointSynchronize::is_at_safepoint(),
1845 "world should be stopped");
1846 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1847
1848 // If a full collection has happened, we shouldn't do this.
1849 if (has_aborted()) {
1850 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1851 return;
1852 }
1853
1854 HRSPhaseSetter x(HRSPhaseCleanup);
1855 g1h->verify_region_sets_optional();
1856
1857 if (VerifyDuringGC) {
1858 HandleMark hm; // handle scope
1859 gclog_or_tty->print(" VerifyDuringGC:(before)");
1860 Universe::heap()->prepare_for_verify();
1861 Universe::verify(/* silent */ false,
1862 /* option */ VerifyOption_G1UsePrevMarking);
1863 }
1864
1865 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy();
1866 g1p->record_concurrent_mark_cleanup_start();
1867
1868 double start = os::elapsedTime();
1869
1870 HeapRegionRemSet::reset_for_cleanup_tasks();
1871
1872 uint n_workers;
1873
1874 // Do counting once more with the world stopped for good measure.
1875 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
1876
1877 if (G1CollectedHeap::use_parallel_gc_threads()) {
1878 assert(g1h->check_heap_region_claim_values(HeapRegion::InitialClaimValue),
1879 "sanity check");
1880
1881 g1h->set_par_threads();
1882 n_workers = g1h->n_par_threads();
1883 assert(g1h->n_par_threads() == n_workers,
1884 "Should not have been reset");
1885 g1h->workers()->run_task(&g1_par_count_task);
1886 // Done with the parallel phase so reset to 0.
1887 g1h->set_par_threads(0);
1888
1889 assert(g1h->check_heap_region_claim_values(HeapRegion::FinalCountClaimValue),
1890 "sanity check");
1891 } else {
1892 n_workers = 1;
1893 g1_par_count_task.work(0);
1894 }
1895
1896 if (VerifyDuringGC) {
1897 // Verify that the counting data accumulated during marking matches
1898 // that calculated by walking the marking bitmap.
1899
1900 // Bitmaps to hold expected values
1901 BitMap expected_region_bm(_region_bm.size(), false);
1902 BitMap expected_card_bm(_card_bm.size(), false);
1903
1904 G1ParVerifyFinalCountTask g1_par_verify_task(g1h,
1905 &_region_bm,
1906 &_card_bm,
1907 &expected_region_bm,
1908 &expected_card_bm);
1909
1910 if (G1CollectedHeap::use_parallel_gc_threads()) {
1911 g1h->set_par_threads((int)n_workers);
1912 g1h->workers()->run_task(&g1_par_verify_task);
1913 // Done with the parallel phase so reset to 0.
1914 g1h->set_par_threads(0);
1915
1916 assert(g1h->check_heap_region_claim_values(HeapRegion::VerifyCountClaimValue),
1917 "sanity check");
1918 } else {
1919 g1_par_verify_task.work(0);
1920 }
1921
1922 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures");
1923 }
1924
1925 size_t start_used_bytes = g1h->used();
1926 g1h->set_marking_complete();
1927
1928 double count_end = os::elapsedTime();
1929 double this_final_counting_time = (count_end - start);
1930 _total_counting_time += this_final_counting_time;
1931
1932 if (G1PrintRegionLivenessInfo) {
1933 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking");
1934 _g1h->heap_region_iterate(&cl);
1935 }
1936
1937 // Install newly created mark bitMap as "prev".
1938 swapMarkBitMaps();
1939
1940 g1h->reset_gc_time_stamp();
1941
1942 // Note end of marking in all heap regions.
1943 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list);
1944 if (G1CollectedHeap::use_parallel_gc_threads()) {
1945 g1h->set_par_threads((int)n_workers);
1946 g1h->workers()->run_task(&g1_par_note_end_task);
1947 g1h->set_par_threads(0);
1948
1949 assert(g1h->check_heap_region_claim_values(HeapRegion::NoteEndClaimValue),
1950 "sanity check");
1951 } else {
1952 g1_par_note_end_task.work(0);
1953 }
1954 g1h->check_gc_time_stamps();
1955
1956 if (!cleanup_list_is_empty()) {
1957 // The cleanup list is not empty, so we'll have to process it
1958 // concurrently. Notify anyone else that might be wanting free
1959 // regions that there will be more free regions coming soon.
1960 g1h->set_free_regions_coming();
1961 }
1962
1963 // call below, since it affects the metric by which we sort the heap
1964 // regions.
1965 if (G1ScrubRemSets) {
1966 double rs_scrub_start = os::elapsedTime();
1967 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm);
1968 if (G1CollectedHeap::use_parallel_gc_threads()) {
1969 g1h->set_par_threads((int)n_workers);
1970 g1h->workers()->run_task(&g1_par_scrub_rs_task);
1971 g1h->set_par_threads(0);
1972
1973 assert(g1h->check_heap_region_claim_values(
1974 HeapRegion::ScrubRemSetClaimValue),
1975 "sanity check");
1976 } else {
1977 g1_par_scrub_rs_task.work(0);
1978 }
1979
1980 double rs_scrub_end = os::elapsedTime();
1981 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start);
1982 _total_rs_scrub_time += this_rs_scrub_time;
1983 }
1984
1985 // this will also free any regions totally full of garbage objects,
1986 // and sort the regions.
1987 g1h->g1_policy()->record_concurrent_mark_cleanup_end((int)n_workers);
1988
1989 // Statistics.
1990 double end = os::elapsedTime();
1991 _cleanup_times.add((end - start) * 1000.0);
1992
1993 if (G1Log::fine()) {
1994 g1h->print_size_transition(gclog_or_tty,
1995 start_used_bytes,
1996 g1h->used(),
1997 g1h->capacity());
1998 }
1999
2000 // Clean up will have freed any regions completely full of garbage.
2001 // Update the soft reference policy with the new heap occupancy.
2002 Universe::update_heap_info_at_gc();
2003
2004 // We need to make this be a "collection" so any collection pause that
2005 // races with it goes around and waits for completeCleanup to finish.
2006 g1h->increment_total_collections();
2007
2008 // We reclaimed old regions so we should calculate the sizes to make
2009 // sure we update the old gen/space data.
2010 g1h->g1mm()->update_sizes();
2011
2012 if (VerifyDuringGC) {
2013 HandleMark hm; // handle scope
2014 gclog_or_tty->print(" VerifyDuringGC:(after)");
2015 Universe::heap()->prepare_for_verify();
2016 Universe::verify(/* silent */ false,
2017 /* option */ VerifyOption_G1UsePrevMarking);
2018 }
2019
2020 g1h->verify_region_sets_optional();
2021 g1h->trace_heap_after_concurrent_cycle();
2022 }
2023
2024 void ConcurrentMark::completeCleanup() {
2025 if (has_aborted()) return;
2026
2027 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2028
2029 _cleanup_list.verify_optional();
2030 FreeRegionList tmp_free_list("Tmp Free List");
2031
2032 if (G1ConcRegionFreeingVerbose) {
2033 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2034 "cleanup list has %u entries",
2035 _cleanup_list.length());
2036 }
2037
2038 // Noone else should be accessing the _cleanup_list at this point,
2039 // so it's not necessary to take any locks
2040 while (!_cleanup_list.is_empty()) {
2041 HeapRegion* hr = _cleanup_list.remove_head();
2042 assert(hr != NULL, "the list was not empty");
2043 hr->par_clear();
2044 tmp_free_list.add_as_tail(hr);
2045
2046 // Instead of adding one region at a time to the secondary_free_list,
2047 // we accumulate them in the local list and move them a few at a
2048 // time. This also cuts down on the number of notify_all() calls
2049 // we do during this process. We'll also append the local list when
2050 // _cleanup_list is empty (which means we just removed the last
2051 // region from the _cleanup_list).
2052 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
2053 _cleanup_list.is_empty()) {
2054 if (G1ConcRegionFreeingVerbose) {
2055 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2056 "appending %u entries to the secondary_free_list, "
2057 "cleanup list still has %u entries",
2058 tmp_free_list.length(),
2059 _cleanup_list.length());
2060 }
2061
2062 {
2063 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
2064 g1h->secondary_free_list_add_as_tail(&tmp_free_list);
2065 SecondaryFreeList_lock->notify_all();
2066 }
2067
2068 if (G1StressConcRegionFreeing) {
2069 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
2070 os::sleep(Thread::current(), (jlong) 1, false);
2071 }
2072 }
2073 }
2074 }
2075 assert(tmp_free_list.is_empty(), "post-condition");
2076 }
2077
2078 // Supporting Object and Oop closures for reference discovery
2079 // and processing in during marking
2080
2081 bool G1CMIsAliveClosure::do_object_b(oop obj) {
2082 HeapWord* addr = (HeapWord*)obj;
2083 return addr != NULL &&
2084 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
2085 }
2086
2087 // 'Keep Alive' oop closure used by both serial parallel reference processing.
2088 // Uses the CMTask associated with a worker thread (for serial reference
2089 // processing the CMTask for worker 0 is used) to preserve (mark) and
2090 // trace referent objects.
2091 //
2092 // Using the CMTask and embedded local queues avoids having the worker
2093 // threads operating on the global mark stack. This reduces the risk
2094 // of overflowing the stack - which we would rather avoid at this late
2095 // state. Also using the tasks' local queues removes the potential
2096 // of the workers interfering with each other that could occur if
2097 // operating on the global stack.
2098
2099 class G1CMKeepAliveAndDrainClosure: public OopClosure {
2100 ConcurrentMark* _cm;
2101 CMTask* _task;
2102 int _ref_counter_limit;
2103 int _ref_counter;
2104 bool _is_serial;
2105 public:
2106 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2107 _cm(cm), _task(task), _is_serial(is_serial),
2108 _ref_counter_limit(G1RefProcDrainInterval) {
2109 assert(_ref_counter_limit > 0, "sanity");
2110 assert(!_is_serial || _task->task_id() == 0, "only task 0 for serial code");
2111 _ref_counter = _ref_counter_limit;
2112 }
2113
2114 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
2115 virtual void do_oop( oop* p) { do_oop_work(p); }
2116
2117 template <class T> void do_oop_work(T* p) {
2118 if (!_cm->has_overflown()) {
2119 oop obj = oopDesc::load_decode_heap_oop(p);
2120 if (_cm->verbose_high()) {
2121 gclog_or_tty->print_cr("\t[%d] we're looking at location "
2122 "*"PTR_FORMAT" = "PTR_FORMAT,
2123 _task->task_id(), p, (void*) obj);
2124 }
2125
2126 _task->deal_with_reference(obj);
2127 _ref_counter--;
2128
2129 if (_ref_counter == 0) {
2130 // We have dealt with _ref_counter_limit references, pushing them
2131 // and objects reachable from them on to the local stack (and
2132 // possibly the global stack). Call CMTask::do_marking_step() to
2133 // process these entries.
2134 //
2135 // We call CMTask::do_marking_step() in a loop, which we'll exit if
2136 // there's nothing more to do (i.e. we're done with the entries that
2137 // were pushed as a result of the CMTask::deal_with_reference() calls
2138 // above) or we overflow.
2139 //
2140 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2141 // flag while there may still be some work to do. (See the comment at
2142 // the beginning of CMTask::do_marking_step() for those conditions -
2143 // one of which is reaching the specified time target.) It is only
2144 // when CMTask::do_marking_step() returns without setting the
2145 // has_aborted() flag that the marking step has completed.
2146 do {
2147 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
2148 _task->do_marking_step(mark_step_duration_ms,
2149 false /* do_termination */,
2150 _is_serial);
2151 } while (_task->has_aborted() && !_cm->has_overflown());
2152 _ref_counter = _ref_counter_limit;
2153 }
2154 } else {
2155 if (_cm->verbose_high()) {
2156 gclog_or_tty->print_cr("\t[%d] CM Overflow", _task->task_id());
2157 }
2158 }
2159 }
2160 };
2161
2162 // 'Drain' oop closure used by both serial and parallel reference processing.
2163 // Uses the CMTask associated with a given worker thread (for serial
2164 // reference processing the CMtask for worker 0 is used). Calls the
2165 // do_marking_step routine, with an unbelievably large timeout value,
2166 // to drain the marking data structures of the remaining entries
2167 // added by the 'keep alive' oop closure above.
2168
2169 class G1CMDrainMarkingStackClosure: public VoidClosure {
2170 ConcurrentMark* _cm;
2171 CMTask* _task;
2172 bool _is_serial;
2173 public:
2174 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2175 _cm(cm), _task(task), _is_serial(is_serial) {
2176 assert(!_is_serial || _task->task_id() == 0, "only task 0 for serial code");
2177 }
2178
2179 void do_void() {
2180 do {
2181 if (_cm->verbose_high()) {
2182 gclog_or_tty->print_cr("\t[%d] Drain: Calling do_marking_step - serial: %s",
2183 _task->task_id(), BOOL_TO_STR(_is_serial));
2184 }
2185
2186 // We call CMTask::do_marking_step() to completely drain the local
2187 // and global marking stacks of entries pushed by the 'keep alive'
2188 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
2189 //
2190 // CMTask::do_marking_step() is called in a loop, which we'll exit
2191 // if there's nothing more to do (i.e. we'completely drained the
2192 // entries that were pushed as a a result of applying the 'keep alive'
2193 // closure to the entries on the discovered ref lists) or we overflow
2194 // the global marking stack.
2195 //
2196 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2197 // flag while there may still be some work to do. (See the comment at
2198 // the beginning of CMTask::do_marking_step() for those conditions -
2199 // one of which is reaching the specified time target.) It is only
2200 // when CMTask::do_marking_step() returns without setting the
2201 // has_aborted() flag that the marking step has completed.
2202
2203 _task->do_marking_step(1000000000.0 /* something very large */,
2204 true /* do_termination */,
2205 _is_serial);
2206 } while (_task->has_aborted() && !_cm->has_overflown());
2207 }
2208 };
2209
2210 // Implementation of AbstractRefProcTaskExecutor for parallel
2211 // reference processing at the end of G1 concurrent marking
2212
2213 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
2214 private:
2215 G1CollectedHeap* _g1h;
2216 ConcurrentMark* _cm;
2217 WorkGang* _workers;
2218 int _active_workers;
2219
2220 public:
2221 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
2222 ConcurrentMark* cm,
2223 WorkGang* workers,
2224 int n_workers) :
2225 _g1h(g1h), _cm(cm),
2226 _workers(workers), _active_workers(n_workers) { }
2227
2228 // Executes the given task using concurrent marking worker threads.
2229 virtual void execute(ProcessTask& task);
2230 virtual void execute(EnqueueTask& task);
2231 };
2232
2233 class G1CMRefProcTaskProxy: public AbstractGangTask {
2234 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
2235 ProcessTask& _proc_task;
2236 G1CollectedHeap* _g1h;
2237 ConcurrentMark* _cm;
2238
2239 public:
2240 G1CMRefProcTaskProxy(ProcessTask& proc_task,
2241 G1CollectedHeap* g1h,
2242 ConcurrentMark* cm) :
2243 AbstractGangTask("Process reference objects in parallel"),
2244 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
2245 ReferenceProcessor* rp = _g1h->ref_processor_cm();
2246 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
2247 }
2248
2249 virtual void work(uint worker_id) {
2250 CMTask* task = _cm->task(worker_id);
2251 G1CMIsAliveClosure g1_is_alive(_g1h);
2252 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
2253 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
2254
2255 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
2256 }
2257 };
2258
2259 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
2260 assert(_workers != NULL, "Need parallel worker threads.");
2261 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2262
2263 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
2264
2265 // We need to reset the concurrency level before each
2266 // proxy task execution, so that the termination protocol
2267 // and overflow handling in CMTask::do_marking_step() knows
2268 // how many workers to wait for.
2269 _cm->set_concurrency(_active_workers);
2270 _g1h->set_par_threads(_active_workers);
2271 _workers->run_task(&proc_task_proxy);
2272 _g1h->set_par_threads(0);
2273 }
2274
2275 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
2276 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
2277 EnqueueTask& _enq_task;
2278
2279 public:
2280 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
2281 AbstractGangTask("Enqueue reference objects in parallel"),
2282 _enq_task(enq_task) { }
2283
2284 virtual void work(uint worker_id) {
2285 _enq_task.work(worker_id);
2286 }
2287 };
2288
2289 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2290 assert(_workers != NULL, "Need parallel worker threads.");
2291 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2292
2293 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2294
2295 // Not strictly necessary but...
2296 //
2297 // We need to reset the concurrency level before each
2298 // proxy task execution, so that the termination protocol
2299 // and overflow handling in CMTask::do_marking_step() knows
2300 // how many workers to wait for.
2301 _cm->set_concurrency(_active_workers);
2302 _g1h->set_par_threads(_active_workers);
2303 _workers->run_task(&enq_task_proxy);
2304 _g1h->set_par_threads(0);
2305 }
2306
2307 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2308 if (has_overflown()) {
2309 // Skip processing the discovered references if we have
2310 // overflown the global marking stack. Reference objects
2311 // only get discovered once so it is OK to not
2312 // de-populate the discovered reference lists. We could have,
2313 // but the only benefit would be that, when marking restarts,
2314 // less reference objects are discovered.
2315 return;
2316 }
2317
2318 ResourceMark rm;
2319 HandleMark hm;
2320
2321 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2322
2323 // Is alive closure.
2324 G1CMIsAliveClosure g1_is_alive(g1h);
2325
2326 // Inner scope to exclude the cleaning of the string and symbol
2327 // tables from the displayed time.
2328 {
2329 if (G1Log::finer()) {
2330 gclog_or_tty->put(' ');
2331 }
2332 GCTraceTime t("GC ref-proc", G1Log::finer(), false, g1h->gc_timer_cm());
2333
2334 ReferenceProcessor* rp = g1h->ref_processor_cm();
2335
2336 // See the comment in G1CollectedHeap::ref_processing_init()
2337 // about how reference processing currently works in G1.
2338
2339 // Set the soft reference policy
2340 rp->setup_policy(clear_all_soft_refs);
2341 assert(_markStack.isEmpty(), "mark stack should be empty");
2342
2343 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2344 // in serial reference processing. Note these closures are also
2345 // used for serially processing (by the the current thread) the
2346 // JNI references during parallel reference processing.
2347 //
2348 // These closures do not need to synchronize with the worker
2349 // threads involved in parallel reference processing as these
2350 // instances are executed serially by the current thread (e.g.
2351 // reference processing is not multi-threaded and is thus
2352 // performed by the current thread instead of a gang worker).
2353 //
2354 // The gang tasks involved in parallel reference procssing create
2355 // their own instances of these closures, which do their own
2356 // synchronization among themselves.
2357 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2358 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2359
2360 // We need at least one active thread. If reference processing
2361 // is not multi-threaded we use the current (VMThread) thread,
2362 // otherwise we use the work gang from the G1CollectedHeap and
2363 // we utilize all the worker threads we can.
2364 bool processing_is_mt = rp->processing_is_mt() && g1h->workers() != NULL;
2365 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2366 active_workers = MAX2(MIN2(active_workers, _max_task_num), 1U);
2367
2368 // Parallel processing task executor.
2369 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2370 g1h->workers(), active_workers);
2371 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2372
2373 ReferenceProcessorStats stats;
2374
2375 // Set the concurrency level. The phase was already set prior to
2376 // executing the remark task.
2377 set_concurrency(active_workers);
2378
2379 // Set the degree of MT processing here. If the discovery was done MT,
2380 // the number of threads involved during discovery could differ from
2381 // the number of active workers. This is OK as long as the discovered
2382 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2383 rp->set_active_mt_degree(active_workers);
2384
2385 // Process the weak references.
2386 stats = rp->process_discovered_references(&g1_is_alive,
2387 &g1_keep_alive,
2388 &g1_drain_mark_stack,
2389 executor,
2390 g1h->gc_timer_cm());
2391
2392 // The do_oop work routines of the keep_alive and drain_marking_stack
2393 // oop closures will set the has_overflown flag if we overflow the
2394 // global marking stack.
2395
2396 g1h->gc_tracer_cm()->report_gc_reference_stats(stats);
2397
2398 assert(_markStack.overflow() || _markStack.isEmpty(),
2399 "mark stack should be empty (unless it overflowed)");
2400
2401 if (_markStack.overflow()) {
2402 // This should have been done already when we tried to push an
2403 // entry on to the global mark stack. But let's do it again.
2404 set_has_overflown();
2405 }
2406
2407 assert(rp->num_q() == active_workers, "why not");
2408
2409 rp->enqueue_discovered_references(executor);
2410
2411 rp->verify_no_references_recorded();
2412 assert(!rp->discovery_enabled(), "Post condition");
2413 }
2414
2415 // Now clean up stale oops in StringTable
2416 StringTable::unlink(&g1_is_alive);
2417 // Clean up unreferenced symbols in symbol table.
2418 SymbolTable::unlink();
2419 }
2420
2421 void ConcurrentMark::swapMarkBitMaps() {
2422 CMBitMapRO* temp = _prevMarkBitMap;
2423 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap;
2424 _nextMarkBitMap = (CMBitMap*) temp;
2425 }
2426
2427 class CMRemarkTask: public AbstractGangTask {
2428 private:
2429 ConcurrentMark* _cm;
2430 bool _is_serial;
2431 public:
2432 void work(uint worker_id) {
2433 // Since all available tasks are actually started, we should
2434 // only proceed if we're supposed to be actived.
2435 if (worker_id < _cm->active_tasks()) {
2436 CMTask* task = _cm->task(worker_id);
2437 task->record_start_time();
2438 do {
2439 task->do_marking_step(1000000000.0 /* something very large */,
2440 true /* do_termination */,
2441 _is_serial);
2442 } while (task->has_aborted() && !_cm->has_overflown());
2443 // If we overflow, then we do not want to restart. We instead
2444 // want to abort remark and do concurrent marking again.
2445 task->record_end_time();
2446 }
2447 }
2448
2449 CMRemarkTask(ConcurrentMark* cm, int active_workers, bool is_serial) :
2450 AbstractGangTask("Par Remark"), _cm(cm), _is_serial(is_serial) {
2451 _cm->terminator()->reset_for_reuse(active_workers);
2452 }
2453 };
2454
2455 void ConcurrentMark::checkpointRootsFinalWork() {
2456 ResourceMark rm;
2457 HandleMark hm;
2458 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2459
2460 g1h->ensure_parsability(false);
2461
2462 if (G1CollectedHeap::use_parallel_gc_threads()) {
2463 G1CollectedHeap::StrongRootsScope srs(g1h);
2464 // this is remark, so we'll use up all active threads
2465 uint active_workers = g1h->workers()->active_workers();
2466 if (active_workers == 0) {
2467 assert(active_workers > 0, "Should have been set earlier");
2468 active_workers = (uint) ParallelGCThreads;
2469 g1h->workers()->set_active_workers(active_workers);
2470 }
2471 set_concurrency_and_phase(active_workers, false /* concurrent */);
2472 // Leave _parallel_marking_threads at it's
2473 // value originally calculated in the ConcurrentMark
2474 // constructor and pass values of the active workers
2475 // through the gang in the task.
2476
2477 CMRemarkTask remarkTask(this, active_workers, false /* is_serial */);
2478 // We will start all available threads, even if we decide that the
2479 // active_workers will be fewer. The extra ones will just bail out
2480 // immediately.
2481 g1h->set_par_threads(active_workers);
2482 g1h->workers()->run_task(&remarkTask);
2483 g1h->set_par_threads(0);
2484 } else {
2485 G1CollectedHeap::StrongRootsScope srs(g1h);
2486 uint active_workers = 1;
2487 set_concurrency_and_phase(active_workers, false /* concurrent */);
2488
2489 // Note - if there's no work gang then the VMThread will be
2490 // the thread to execute the remark - serially. We have
2491 // to pass true for the is_serial parameter so that
2492 // CMTask::do_marking_step() doesn't enter the sync
2493 // barriers in the event of an overflow. Doing so will
2494 // cause an assert that the current thread is not a
2495 // concurrent GC thread.
2496 CMRemarkTask remarkTask(this, active_workers, true /* is_serial*/);
2497 remarkTask.work(0);
2498 }
2499 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2500 guarantee(has_overflown() ||
2501 satb_mq_set.completed_buffers_num() == 0,
2502 err_msg("Invariant: has_overflown = %s, num buffers = %d",
2503 BOOL_TO_STR(has_overflown()),
2504 satb_mq_set.completed_buffers_num()));
2505
2506 print_stats();
2507
2508 #if VERIFY_OBJS_PROCESSED
2509 if (_scan_obj_cl.objs_processed != ThreadLocalObjQueue::objs_enqueued) {
2510 gclog_or_tty->print_cr("Processed = %d, enqueued = %d.",
2511 _scan_obj_cl.objs_processed,
2512 ThreadLocalObjQueue::objs_enqueued);
2513 guarantee(_scan_obj_cl.objs_processed ==
2514 ThreadLocalObjQueue::objs_enqueued,
2515 "Different number of objs processed and enqueued.");
2516 }
2517 #endif
2518 }
2519
2520 #ifndef PRODUCT
2521
2522 class PrintReachableOopClosure: public OopClosure {
2523 private:
2524 G1CollectedHeap* _g1h;
2525 outputStream* _out;
2526 VerifyOption _vo;
2527 bool _all;
2528
2529 public:
2530 PrintReachableOopClosure(outputStream* out,
2531 VerifyOption vo,
2532 bool all) :
2533 _g1h(G1CollectedHeap::heap()),
2534 _out(out), _vo(vo), _all(all) { }
2535
2536 void do_oop(narrowOop* p) { do_oop_work(p); }
2537 void do_oop( oop* p) { do_oop_work(p); }
2538
2539 template <class T> void do_oop_work(T* p) {
2540 oop obj = oopDesc::load_decode_heap_oop(p);
2541 const char* str = NULL;
2542 const char* str2 = "";
2543
2544 if (obj == NULL) {
2545 str = "";
2546 } else if (!_g1h->is_in_g1_reserved(obj)) {
2547 str = " O";
2548 } else {
2549 HeapRegion* hr = _g1h->heap_region_containing(obj);
2550 guarantee(hr != NULL, "invariant");
2551 bool over_tams = _g1h->allocated_since_marking(obj, hr, _vo);
2552 bool marked = _g1h->is_marked(obj, _vo);
2553
2554 if (over_tams) {
2555 str = " >";
2556 if (marked) {
2557 str2 = " AND MARKED";
2558 }
2559 } else if (marked) {
2560 str = " M";
2561 } else {
2562 str = " NOT";
2563 }
2564 }
2565
2566 _out->print_cr(" "PTR_FORMAT": "PTR_FORMAT"%s%s",
2567 p, (void*) obj, str, str2);
2568 }
2569 };
2570
2571 class PrintReachableObjectClosure : public ObjectClosure {
2572 private:
2573 G1CollectedHeap* _g1h;
2574 outputStream* _out;
2575 VerifyOption _vo;
2576 bool _all;
2577 HeapRegion* _hr;
2578
2579 public:
2580 PrintReachableObjectClosure(outputStream* out,
2581 VerifyOption vo,
2582 bool all,
2583 HeapRegion* hr) :
2584 _g1h(G1CollectedHeap::heap()),
2585 _out(out), _vo(vo), _all(all), _hr(hr) { }
2586
2587 void do_object(oop o) {
2588 bool over_tams = _g1h->allocated_since_marking(o, _hr, _vo);
2589 bool marked = _g1h->is_marked(o, _vo);
2590 bool print_it = _all || over_tams || marked;
2591
2592 if (print_it) {
2593 _out->print_cr(" "PTR_FORMAT"%s",
2594 o, (over_tams) ? " >" : (marked) ? " M" : "");
2595 PrintReachableOopClosure oopCl(_out, _vo, _all);
2596 o->oop_iterate(&oopCl);
2597 }
2598 }
2599 };
2600
2601 class PrintReachableRegionClosure : public HeapRegionClosure {
2602 private:
2603 G1CollectedHeap* _g1h;
2604 outputStream* _out;
2605 VerifyOption _vo;
2606 bool _all;
2607
2608 public:
2609 bool doHeapRegion(HeapRegion* hr) {
2610 HeapWord* b = hr->bottom();
2611 HeapWord* e = hr->end();
2612 HeapWord* t = hr->top();
2613 HeapWord* p = _g1h->top_at_mark_start(hr, _vo);
2614 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" "
2615 "TAMS: "PTR_FORMAT, b, e, t, p);
2616 _out->cr();
2617
2618 HeapWord* from = b;
2619 HeapWord* to = t;
2620
2621 if (to > from) {
2622 _out->print_cr("Objects in ["PTR_FORMAT", "PTR_FORMAT"]", from, to);
2623 _out->cr();
2624 PrintReachableObjectClosure ocl(_out, _vo, _all, hr);
2625 hr->object_iterate_mem_careful(MemRegion(from, to), &ocl);
2626 _out->cr();
2627 }
2628
2629 return false;
2630 }
2631
2632 PrintReachableRegionClosure(outputStream* out,
2633 VerifyOption vo,
2634 bool all) :
2635 _g1h(G1CollectedHeap::heap()), _out(out), _vo(vo), _all(all) { }
2636 };
2637
2638 void ConcurrentMark::print_reachable(const char* str,
2639 VerifyOption vo,
2640 bool all) {
2641 gclog_or_tty->cr();
2642 gclog_or_tty->print_cr("== Doing heap dump... ");
2643
2644 if (G1PrintReachableBaseFile == NULL) {
2645 gclog_or_tty->print_cr(" #### error: no base file defined");
2646 return;
2647 }
2648
2649 if (strlen(G1PrintReachableBaseFile) + 1 + strlen(str) >
2650 (JVM_MAXPATHLEN - 1)) {
2651 gclog_or_tty->print_cr(" #### error: file name too long");
2652 return;
2653 }
2654
2655 char file_name[JVM_MAXPATHLEN];
2656 sprintf(file_name, "%s.%s", G1PrintReachableBaseFile, str);
2657 gclog_or_tty->print_cr(" dumping to file %s", file_name);
2658
2659 fileStream fout(file_name);
2660 if (!fout.is_open()) {
2661 gclog_or_tty->print_cr(" #### error: could not open file");
2662 return;
2663 }
2664
2665 outputStream* out = &fout;
2666 out->print_cr("-- USING %s", _g1h->top_at_mark_start_str(vo));
2667 out->cr();
2668
2669 out->print_cr("--- ITERATING OVER REGIONS");
2670 out->cr();
2671 PrintReachableRegionClosure rcl(out, vo, all);
2672 _g1h->heap_region_iterate(&rcl);
2673 out->cr();
2674
2675 gclog_or_tty->print_cr(" done");
2676 gclog_or_tty->flush();
2677 }
2678
2679 #endif // PRODUCT
2680
2681 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2682 // Note we are overriding the read-only view of the prev map here, via
2683 // the cast.
2684 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2685 }
2686
2687 void ConcurrentMark::clearRangeNextBitmap(MemRegion mr) {
2688 _nextMarkBitMap->clearRange(mr);
2689 }
2690
2691 void ConcurrentMark::clearRangeBothBitmaps(MemRegion mr) {
2692 clearRangePrevBitmap(mr);
2693 clearRangeNextBitmap(mr);
2694 }
2695
2696 HeapRegion*
2697 ConcurrentMark::claim_region(int task_num) {
2698 // "checkpoint" the finger
2699 HeapWord* finger = _finger;
2700
2701 // _heap_end will not change underneath our feet; it only changes at
2702 // yield points.
2703 while (finger < _heap_end) {
2704 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2705
2706 // Note on how this code handles humongous regions. In the
2707 // normal case the finger will reach the start of a "starts
2708 // humongous" (SH) region. Its end will either be the end of the
2709 // last "continues humongous" (CH) region in the sequence, or the
2710 // standard end of the SH region (if the SH is the only region in
2711 // the sequence). That way claim_region() will skip over the CH
2712 // regions. However, there is a subtle race between a CM thread
2713 // executing this method and a mutator thread doing a humongous
2714 // object allocation. The two are not mutually exclusive as the CM
2715 // thread does not need to hold the Heap_lock when it gets
2716 // here. So there is a chance that claim_region() will come across
2717 // a free region that's in the progress of becoming a SH or a CH
2718 // region. In the former case, it will either
2719 // a) Miss the update to the region's end, in which case it will
2720 // visit every subsequent CH region, will find their bitmaps
2721 // empty, and do nothing, or
2722 // b) Will observe the update of the region's end (in which case
2723 // it will skip the subsequent CH regions).
2724 // If it comes across a region that suddenly becomes CH, the
2725 // scenario will be similar to b). So, the race between
2726 // claim_region() and a humongous object allocation might force us
2727 // to do a bit of unnecessary work (due to some unnecessary bitmap
2728 // iterations) but it should not introduce and correctness issues.
2729 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger);
2730 HeapWord* bottom = curr_region->bottom();
2731 HeapWord* end = curr_region->end();
2732 HeapWord* limit = curr_region->next_top_at_mark_start();
2733
2734 if (verbose_low()) {
2735 gclog_or_tty->print_cr("[%d] curr_region = "PTR_FORMAT" "
2736 "["PTR_FORMAT", "PTR_FORMAT"), "
2737 "limit = "PTR_FORMAT,
2738 task_num, curr_region, bottom, end, limit);
2739 }
2740
2741 // Is the gap between reading the finger and doing the CAS too long?
2742 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2743 if (res == finger) {
2744 // we succeeded
2745
2746 // notice that _finger == end cannot be guaranteed here since,
2747 // someone else might have moved the finger even further
2748 assert(_finger >= end, "the finger should have moved forward");
2749
2750 if (verbose_low()) {
2751 gclog_or_tty->print_cr("[%d] we were successful with region = "
2752 PTR_FORMAT, task_num, curr_region);
2753 }
2754
2755 if (limit > bottom) {
2756 if (verbose_low()) {
2757 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is not empty, "
2758 "returning it ", task_num, curr_region);
2759 }
2760 return curr_region;
2761 } else {
2762 assert(limit == bottom,
2763 "the region limit should be at bottom");
2764 if (verbose_low()) {
2765 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is empty, "
2766 "returning NULL", task_num, curr_region);
2767 }
2768 // we return NULL and the caller should try calling
2769 // claim_region() again.
2770 return NULL;
2771 }
2772 } else {
2773 assert(_finger > finger, "the finger should have moved forward");
2774 if (verbose_low()) {
2775 gclog_or_tty->print_cr("[%d] somebody else moved the finger, "
2776 "global finger = "PTR_FORMAT", "
2777 "our finger = "PTR_FORMAT,
2778 task_num, _finger, finger);
2779 }
2780
2781 // read it again
2782 finger = _finger;
2783 }
2784 }
2785
2786 return NULL;
2787 }
2788
2789 #ifndef PRODUCT
2790 enum VerifyNoCSetOopsPhase {
2791 VerifyNoCSetOopsStack,
2792 VerifyNoCSetOopsQueues,
2793 VerifyNoCSetOopsSATBCompleted,
2794 VerifyNoCSetOopsSATBThread
2795 };
2796
2797 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure {
2798 private:
2799 G1CollectedHeap* _g1h;
2800 VerifyNoCSetOopsPhase _phase;
2801 int _info;
2802
2803 const char* phase_str() {
2804 switch (_phase) {
2805 case VerifyNoCSetOopsStack: return "Stack";
2806 case VerifyNoCSetOopsQueues: return "Queue";
2807 case VerifyNoCSetOopsSATBCompleted: return "Completed SATB Buffers";
2808 case VerifyNoCSetOopsSATBThread: return "Thread SATB Buffers";
2809 default: ShouldNotReachHere();
2810 }
2811 return NULL;
2812 }
2813
2814 void do_object_work(oop obj) {
2815 guarantee(!_g1h->obj_in_cs(obj),
2816 err_msg("obj: "PTR_FORMAT" in CSet, phase: %s, info: %d",
2817 (void*) obj, phase_str(), _info));
2818 }
2819
2820 public:
2821 VerifyNoCSetOopsClosure() : _g1h(G1CollectedHeap::heap()) { }
2822
2823 void set_phase(VerifyNoCSetOopsPhase phase, int info = -1) {
2824 _phase = phase;
2825 _info = info;
2826 }
2827
2828 virtual void do_oop(oop* p) {
2829 oop obj = oopDesc::load_decode_heap_oop(p);
2830 do_object_work(obj);
2831 }
2832
2833 virtual void do_oop(narrowOop* p) {
2834 // We should not come across narrow oops while scanning marking
2835 // stacks and SATB buffers.
2836 ShouldNotReachHere();
2837 }
2838
2839 virtual void do_object(oop obj) {
2840 do_object_work(obj);
2841 }
2842 };
2843
2844 void ConcurrentMark::verify_no_cset_oops(bool verify_stacks,
2845 bool verify_enqueued_buffers,
2846 bool verify_thread_buffers,
2847 bool verify_fingers) {
2848 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2849 if (!G1CollectedHeap::heap()->mark_in_progress()) {
2850 return;
2851 }
2852
2853 VerifyNoCSetOopsClosure cl;
2854
2855 if (verify_stacks) {
2856 // Verify entries on the global mark stack
2857 cl.set_phase(VerifyNoCSetOopsStack);
2858 _markStack.oops_do(&cl);
2859
2860 // Verify entries on the task queues
2861 for (int i = 0; i < (int) _max_task_num; i += 1) {
2862 cl.set_phase(VerifyNoCSetOopsQueues, i);
2863 OopTaskQueue* queue = _task_queues->queue(i);
2864 queue->oops_do(&cl);
2865 }
2866 }
2867
2868 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
2869
2870 // Verify entries on the enqueued SATB buffers
2871 if (verify_enqueued_buffers) {
2872 cl.set_phase(VerifyNoCSetOopsSATBCompleted);
2873 satb_qs.iterate_completed_buffers_read_only(&cl);
2874 }
2875
2876 // Verify entries on the per-thread SATB buffers
2877 if (verify_thread_buffers) {
2878 cl.set_phase(VerifyNoCSetOopsSATBThread);
2879 satb_qs.iterate_thread_buffers_read_only(&cl);
2880 }
2881
2882 if (verify_fingers) {
2883 // Verify the global finger
2884 HeapWord* global_finger = finger();
2885 if (global_finger != NULL && global_finger < _heap_end) {
2886 // The global finger always points to a heap region boundary. We
2887 // use heap_region_containing_raw() to get the containing region
2888 // given that the global finger could be pointing to a free region
2889 // which subsequently becomes continues humongous. If that
2890 // happens, heap_region_containing() will return the bottom of the
2891 // corresponding starts humongous region and the check below will
2892 // not hold any more.
2893 HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger);
2894 guarantee(global_finger == global_hr->bottom(),
2895 err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT,
2896 global_finger, HR_FORMAT_PARAMS(global_hr)));
2897 }
2898
2899 // Verify the task fingers
2900 assert(parallel_marking_threads() <= _max_task_num, "sanity");
2901 for (int i = 0; i < (int) parallel_marking_threads(); i += 1) {
2902 CMTask* task = _tasks[i];
2903 HeapWord* task_finger = task->finger();
2904 if (task_finger != NULL && task_finger < _heap_end) {
2905 // See above note on the global finger verification.
2906 HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger);
2907 guarantee(task_finger == task_hr->bottom() ||
2908 !task_hr->in_collection_set(),
2909 err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT,
2910 task_finger, HR_FORMAT_PARAMS(task_hr)));
2911 }
2912 }
2913 }
2914 }
2915 #endif // PRODUCT
2916
2917 // Aggregate the counting data that was constructed concurrently
2918 // with marking.
2919 class AggregateCountDataHRClosure: public HeapRegionClosure {
2920 G1CollectedHeap* _g1h;
2921 ConcurrentMark* _cm;
2922 CardTableModRefBS* _ct_bs;
2923 BitMap* _cm_card_bm;
2924 size_t _max_task_num;
2925
2926 public:
2927 AggregateCountDataHRClosure(G1CollectedHeap* g1h,
2928 BitMap* cm_card_bm,
2929 size_t max_task_num) :
2930 _g1h(g1h), _cm(g1h->concurrent_mark()),
2931 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())),
2932 _cm_card_bm(cm_card_bm), _max_task_num(max_task_num) { }
2933
2934 bool doHeapRegion(HeapRegion* hr) {
2935 if (hr->continuesHumongous()) {
2936 // We will ignore these here and process them when their
2937 // associated "starts humongous" region is processed.
2938 // Note that we cannot rely on their associated
2939 // "starts humongous" region to have their bit set to 1
2940 // since, due to the region chunking in the parallel region
2941 // iteration, a "continues humongous" region might be visited
2942 // before its associated "starts humongous".
2943 return false;
2944 }
2945
2946 HeapWord* start = hr->bottom();
2947 HeapWord* limit = hr->next_top_at_mark_start();
2948 HeapWord* end = hr->end();
2949
2950 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(),
2951 err_msg("Preconditions not met - "
2952 "start: "PTR_FORMAT", limit: "PTR_FORMAT", "
2953 "top: "PTR_FORMAT", end: "PTR_FORMAT,
2954 start, limit, hr->top(), hr->end()));
2955
2956 assert(hr->next_marked_bytes() == 0, "Precondition");
2957
2958 if (start == limit) {
2959 // NTAMS of this region has not been set so nothing to do.
2960 return false;
2961 }
2962
2963 // 'start' should be in the heap.
2964 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity");
2965 // 'end' *may* be just beyone the end of the heap (if hr is the last region)
2966 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity");
2967
2968 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
2969 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit);
2970 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end);
2971
2972 // If ntams is not card aligned then we bump card bitmap index
2973 // for limit so that we get the all the cards spanned by
2974 // the object ending at ntams.
2975 // Note: if this is the last region in the heap then ntams
2976 // could be actually just beyond the end of the the heap;
2977 // limit_idx will then correspond to a (non-existent) card
2978 // that is also outside the heap.
2979 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) {
2980 limit_idx += 1;
2981 }
2982
2983 assert(limit_idx <= end_idx, "or else use atomics");
2984
2985 // Aggregate the "stripe" in the count data associated with hr.
2986 uint hrs_index = hr->hrs_index();
2987 size_t marked_bytes = 0;
2988
2989 for (int i = 0; (size_t)i < _max_task_num; i += 1) {
2990 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i);
2991 BitMap* task_card_bm = _cm->count_card_bitmap_for(i);
2992
2993 // Fetch the marked_bytes in this region for task i and
2994 // add it to the running total for this region.
2995 marked_bytes += marked_bytes_array[hrs_index];
2996
2997 // Now union the bitmaps[0,max_task_num)[start_idx..limit_idx)
2998 // into the global card bitmap.
2999 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx);
3000
3001 while (scan_idx < limit_idx) {
3002 assert(task_card_bm->at(scan_idx) == true, "should be");
3003 _cm_card_bm->set_bit(scan_idx);
3004 assert(_cm_card_bm->at(scan_idx) == true, "should be");
3005
3006 // BitMap::get_next_one_offset() can handle the case when
3007 // its left_offset parameter is greater than its right_offset
3008 // parameter. It does, however, have an early exit if
3009 // left_offset == right_offset. So let's limit the value
3010 // passed in for left offset here.
3011 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx);
3012 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx);
3013 }
3014 }
3015
3016 // Update the marked bytes for this region.
3017 hr->add_to_marked_bytes(marked_bytes);
3018
3019 // Next heap region
3020 return false;
3021 }
3022 };
3023
3024 class G1AggregateCountDataTask: public AbstractGangTask {
3025 protected:
3026 G1CollectedHeap* _g1h;
3027 ConcurrentMark* _cm;
3028 BitMap* _cm_card_bm;
3029 size_t _max_task_num;
3030 int _active_workers;
3031
3032 public:
3033 G1AggregateCountDataTask(G1CollectedHeap* g1h,
3034 ConcurrentMark* cm,
3035 BitMap* cm_card_bm,
3036 size_t max_task_num,
3037 int n_workers) :
3038 AbstractGangTask("Count Aggregation"),
3039 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm),
3040 _max_task_num(max_task_num),
3041 _active_workers(n_workers) { }
3042
3043 void work(uint worker_id) {
3044 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_task_num);
3045
3046 if (G1CollectedHeap::use_parallel_gc_threads()) {
3047 _g1h->heap_region_par_iterate_chunked(&cl, worker_id,
3048 _active_workers,
3049 HeapRegion::AggregateCountClaimValue);
3050 } else {
3051 _g1h->heap_region_iterate(&cl);
3052 }
3053 }
3054 };
3055
3056
3057 void ConcurrentMark::aggregate_count_data() {
3058 int n_workers = (G1CollectedHeap::use_parallel_gc_threads() ?
3059 _g1h->workers()->active_workers() :
3060 1);
3061
3062 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm,
3063 _max_task_num, n_workers);
3064
3065 if (G1CollectedHeap::use_parallel_gc_threads()) {
3066 assert(_g1h->check_heap_region_claim_values(HeapRegion::InitialClaimValue),
3067 "sanity check");
3068 _g1h->set_par_threads(n_workers);
3069 _g1h->workers()->run_task(&g1_par_agg_task);
3070 _g1h->set_par_threads(0);
3071
3072 assert(_g1h->check_heap_region_claim_values(HeapRegion::AggregateCountClaimValue),
3073 "sanity check");
3074 _g1h->reset_heap_region_claim_values();
3075 } else {
3076 g1_par_agg_task.work(0);
3077 }
3078 }
3079
3080 // Clear the per-worker arrays used to store the per-region counting data
3081 void ConcurrentMark::clear_all_count_data() {
3082 // Clear the global card bitmap - it will be filled during
3083 // liveness count aggregation (during remark) and the
3084 // final counting task.
3085 _card_bm.clear();
3086
3087 // Clear the global region bitmap - it will be filled as part
3088 // of the final counting task.
3089 _region_bm.clear();
3090
3091 uint max_regions = _g1h->max_regions();
3092 assert(_max_task_num != 0, "unitialized");
3093
3094 for (int i = 0; (size_t) i < _max_task_num; i += 1) {
3095 BitMap* task_card_bm = count_card_bitmap_for(i);
3096 size_t* marked_bytes_array = count_marked_bytes_array_for(i);
3097
3098 assert(task_card_bm->size() == _card_bm.size(), "size mismatch");
3099 assert(marked_bytes_array != NULL, "uninitialized");
3100
3101 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t));
3102 task_card_bm->clear();
3103 }
3104 }
3105
3106 void ConcurrentMark::print_stats() {
3107 if (verbose_stats()) {
3108 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3109 for (size_t i = 0; i < _active_tasks; ++i) {
3110 _tasks[i]->print_stats();
3111 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3112 }
3113 }
3114 }
3115
3116 // abandon current marking iteration due to a Full GC
3117 void ConcurrentMark::abort() {
3118 // Clear all marks to force marking thread to do nothing
3119 _nextMarkBitMap->clearAll();
3120 // Clear the liveness counting data
3121 clear_all_count_data();
3122 // Empty mark stack
3123 reset_marking_state();
3124 for (int i = 0; i < (int)_max_task_num; ++i) {
3125 _tasks[i]->clear_region_fields();
3126 }
3127 _has_aborted = true;
3128
3129 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3130 satb_mq_set.abandon_partial_marking();
3131 // This can be called either during or outside marking, we'll read
3132 // the expected_active value from the SATB queue set.
3133 satb_mq_set.set_active_all_threads(
3134 false, /* new active value */
3135 satb_mq_set.is_active() /* expected_active */);
3136
3137 _g1h->trace_heap_after_concurrent_cycle();
3138 _g1h->register_concurrent_cycle_end();
3139 }
3140
3141 static void print_ms_time_info(const char* prefix, const char* name,
3142 NumberSeq& ns) {
3143 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
3144 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
3145 if (ns.num() > 0) {
3146 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]",
3147 prefix, ns.sd(), ns.maximum());
3148 }
3149 }
3150
3151 void ConcurrentMark::print_summary_info() {
3152 gclog_or_tty->print_cr(" Concurrent marking:");
3153 print_ms_time_info(" ", "init marks", _init_times);
3154 print_ms_time_info(" ", "remarks", _remark_times);
3155 {
3156 print_ms_time_info(" ", "final marks", _remark_mark_times);
3157 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
3158
3159 }
3160 print_ms_time_info(" ", "cleanups", _cleanup_times);
3161 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).",
3162 _total_counting_time,
3163 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 /
3164 (double)_cleanup_times.num()
3165 : 0.0));
3166 if (G1ScrubRemSets) {
3167 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
3168 _total_rs_scrub_time,
3169 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 /
3170 (double)_cleanup_times.num()
3171 : 0.0));
3172 }
3173 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.",
3174 (_init_times.sum() + _remark_times.sum() +
3175 _cleanup_times.sum())/1000.0);
3176 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s "
3177 "(%8.2f s marking).",
3178 cmThread()->vtime_accum(),
3179 cmThread()->vtime_mark_accum());
3180 }
3181
3182 void ConcurrentMark::print_worker_threads_on(outputStream* st) const {
3183 if (use_parallel_marking_threads()) {
3184 _parallel_workers->print_worker_threads_on(st);
3185 }
3186 }
3187
3188 // We take a break if someone is trying to stop the world.
3189 bool ConcurrentMark::do_yield_check(uint worker_id) {
3190 if (should_yield()) {
3191 if (worker_id == 0) {
3192 _g1h->g1_policy()->record_concurrent_pause();
3193 }
3194 cmThread()->yield();
3195 return true;
3196 } else {
3197 return false;
3198 }
3199 }
3200
3201 bool ConcurrentMark::should_yield() {
3202 return cmThread()->should_yield();
3203 }
3204
3205 bool ConcurrentMark::containing_card_is_marked(void* p) {
3206 size_t offset = pointer_delta(p, _g1h->reserved_region().start(), 1);
3207 return _card_bm.at(offset >> CardTableModRefBS::card_shift);
3208 }
3209
3210 bool ConcurrentMark::containing_cards_are_marked(void* start,
3211 void* last) {
3212 return containing_card_is_marked(start) &&
3213 containing_card_is_marked(last);
3214 }
3215
3216 #ifndef PRODUCT
3217 // for debugging purposes
3218 void ConcurrentMark::print_finger() {
3219 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT,
3220 _heap_start, _heap_end, _finger);
3221 for (int i = 0; i < (int) _max_task_num; ++i) {
3222 gclog_or_tty->print(" %d: "PTR_FORMAT, i, _tasks[i]->finger());
3223 }
3224 gclog_or_tty->print_cr("");
3225 }
3226 #endif
3227
3228 void CMTask::scan_object(oop obj) {
3229 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
3230
3231 if (_cm->verbose_high()) {
3232 gclog_or_tty->print_cr("[%d] we're scanning object "PTR_FORMAT,
3233 _task_id, (void*) obj);
3234 }
3235
3236 size_t obj_size = obj->size();
3237 _words_scanned += obj_size;
3238
3239 obj->oop_iterate(_cm_oop_closure);
3240 statsOnly( ++_objs_scanned );
3241 check_limits();
3242 }
3243
3244 // Closure for iteration over bitmaps
3245 class CMBitMapClosure : public BitMapClosure {
3246 private:
3247 // the bitmap that is being iterated over
3248 CMBitMap* _nextMarkBitMap;
3249 ConcurrentMark* _cm;
3250 CMTask* _task;
3251
3252 public:
3253 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) :
3254 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
3255
3256 bool do_bit(size_t offset) {
3257 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
3258 assert(_nextMarkBitMap->isMarked(addr), "invariant");
3259 assert( addr < _cm->finger(), "invariant");
3260
3261 statsOnly( _task->increase_objs_found_on_bitmap() );
3262 assert(addr >= _task->finger(), "invariant");
3263
3264 // We move that task's local finger along.
3265 _task->move_finger_to(addr);
3266
3267 _task->scan_object(oop(addr));
3268 // we only partially drain the local queue and global stack
3269 _task->drain_local_queue(true);
3270 _task->drain_global_stack(true);
3271
3272 // if the has_aborted flag has been raised, we need to bail out of
3273 // the iteration
3274 return !_task->has_aborted();
3275 }
3276 };
3277
3278 // Closure for iterating over objects, currently only used for
3279 // processing SATB buffers.
3280 class CMObjectClosure : public ObjectClosure {
3281 private:
3282 CMTask* _task;
3283
3284 public:
3285 void do_object(oop obj) {
3286 _task->deal_with_reference(obj);
3287 }
3288
3289 CMObjectClosure(CMTask* task) : _task(task) { }
3290 };
3291
3292 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
3293 ConcurrentMark* cm,
3294 CMTask* task)
3295 : _g1h(g1h), _cm(cm), _task(task) {
3296 assert(_ref_processor == NULL, "should be initialized to NULL");
3297
3298 if (G1UseConcMarkReferenceProcessing) {
3299 _ref_processor = g1h->ref_processor_cm();
3300 assert(_ref_processor != NULL, "should not be NULL");
3301 }
3302 }
3303
3304 void CMTask::setup_for_region(HeapRegion* hr) {
3305 // Separated the asserts so that we know which one fires.
3306 assert(hr != NULL,
3307 "claim_region() should have filtered out continues humongous regions");
3308 assert(!hr->continuesHumongous(),
3309 "claim_region() should have filtered out continues humongous regions");
3310
3311 if (_cm->verbose_low()) {
3312 gclog_or_tty->print_cr("[%d] setting up for region "PTR_FORMAT,
3313 _task_id, hr);
3314 }
3315
3316 _curr_region = hr;
3317 _finger = hr->bottom();
3318 update_region_limit();
3319 }
3320
3321 void CMTask::update_region_limit() {
3322 HeapRegion* hr = _curr_region;
3323 HeapWord* bottom = hr->bottom();
3324 HeapWord* limit = hr->next_top_at_mark_start();
3325
3326 if (limit == bottom) {
3327 if (_cm->verbose_low()) {
3328 gclog_or_tty->print_cr("[%d] found an empty region "
3329 "["PTR_FORMAT", "PTR_FORMAT")",
3330 _task_id, bottom, limit);
3331 }
3332 // The region was collected underneath our feet.
3333 // We set the finger to bottom to ensure that the bitmap
3334 // iteration that will follow this will not do anything.
3335 // (this is not a condition that holds when we set the region up,
3336 // as the region is not supposed to be empty in the first place)
3337 _finger = bottom;
3338 } else if (limit >= _region_limit) {
3339 assert(limit >= _finger, "peace of mind");
3340 } else {
3341 assert(limit < _region_limit, "only way to get here");
3342 // This can happen under some pretty unusual circumstances. An
3343 // evacuation pause empties the region underneath our feet (NTAMS
3344 // at bottom). We then do some allocation in the region (NTAMS
3345 // stays at bottom), followed by the region being used as a GC
3346 // alloc region (NTAMS will move to top() and the objects
3347 // originally below it will be grayed). All objects now marked in
3348 // the region are explicitly grayed, if below the global finger,
3349 // and we do not need in fact to scan anything else. So, we simply
3350 // set _finger to be limit to ensure that the bitmap iteration
3351 // doesn't do anything.
3352 _finger = limit;
3353 }
3354
3355 _region_limit = limit;
3356 }
3357
3358 void CMTask::giveup_current_region() {
3359 assert(_curr_region != NULL, "invariant");
3360 if (_cm->verbose_low()) {
3361 gclog_or_tty->print_cr("[%d] giving up region "PTR_FORMAT,
3362 _task_id, _curr_region);
3363 }
3364 clear_region_fields();
3365 }
3366
3367 void CMTask::clear_region_fields() {
3368 // Values for these three fields that indicate that we're not
3369 // holding on to a region.
3370 _curr_region = NULL;
3371 _finger = NULL;
3372 _region_limit = NULL;
3373 }
3374
3375 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
3376 if (cm_oop_closure == NULL) {
3377 assert(_cm_oop_closure != NULL, "invariant");
3378 } else {
3379 assert(_cm_oop_closure == NULL, "invariant");
3380 }
3381 _cm_oop_closure = cm_oop_closure;
3382 }
3383
3384 void CMTask::reset(CMBitMap* nextMarkBitMap) {
3385 guarantee(nextMarkBitMap != NULL, "invariant");
3386
3387 if (_cm->verbose_low()) {
3388 gclog_or_tty->print_cr("[%d] resetting", _task_id);
3389 }
3390
3391 _nextMarkBitMap = nextMarkBitMap;
3392 clear_region_fields();
3393
3394 _calls = 0;
3395 _elapsed_time_ms = 0.0;
3396 _termination_time_ms = 0.0;
3397 _termination_start_time_ms = 0.0;
3398
3399 #if _MARKING_STATS_
3400 _local_pushes = 0;
3401 _local_pops = 0;
3402 _local_max_size = 0;
3403 _objs_scanned = 0;
3404 _global_pushes = 0;
3405 _global_pops = 0;
3406 _global_max_size = 0;
3407 _global_transfers_to = 0;
3408 _global_transfers_from = 0;
3409 _regions_claimed = 0;
3410 _objs_found_on_bitmap = 0;
3411 _satb_buffers_processed = 0;
3412 _steal_attempts = 0;
3413 _steals = 0;
3414 _aborted = 0;
3415 _aborted_overflow = 0;
3416 _aborted_cm_aborted = 0;
3417 _aborted_yield = 0;
3418 _aborted_timed_out = 0;
3419 _aborted_satb = 0;
3420 _aborted_termination = 0;
3421 #endif // _MARKING_STATS_
3422 }
3423
3424 bool CMTask::should_exit_termination() {
3425 regular_clock_call();
3426 // This is called when we are in the termination protocol. We should
3427 // quit if, for some reason, this task wants to abort or the global
3428 // stack is not empty (this means that we can get work from it).
3429 return !_cm->mark_stack_empty() || has_aborted();
3430 }
3431
3432 void CMTask::reached_limit() {
3433 assert(_words_scanned >= _words_scanned_limit ||
3434 _refs_reached >= _refs_reached_limit ,
3435 "shouldn't have been called otherwise");
3436 regular_clock_call();
3437 }
3438
3439 void CMTask::regular_clock_call() {
3440 if (has_aborted()) return;
3441
3442 // First, we need to recalculate the words scanned and refs reached
3443 // limits for the next clock call.
3444 recalculate_limits();
3445
3446 // During the regular clock call we do the following
3447
3448 // (1) If an overflow has been flagged, then we abort.
3449 if (_cm->has_overflown()) {
3450 set_has_aborted();
3451 return;
3452 }
3453
3454 // If we are not concurrent (i.e. we're doing remark) we don't need
3455 // to check anything else. The other steps are only needed during
3456 // the concurrent marking phase.
3457 if (!concurrent()) return;
3458
3459 // (2) If marking has been aborted for Full GC, then we also abort.
3460 if (_cm->has_aborted()) {
3461 set_has_aborted();
3462 statsOnly( ++_aborted_cm_aborted );
3463 return;
3464 }
3465
3466 double curr_time_ms = os::elapsedVTime() * 1000.0;
3467
3468 // (3) If marking stats are enabled, then we update the step history.
3469 #if _MARKING_STATS_
3470 if (_words_scanned >= _words_scanned_limit) {
3471 ++_clock_due_to_scanning;
3472 }
3473 if (_refs_reached >= _refs_reached_limit) {
3474 ++_clock_due_to_marking;
3475 }
3476
3477 double last_interval_ms = curr_time_ms - _interval_start_time_ms;
3478 _interval_start_time_ms = curr_time_ms;
3479 _all_clock_intervals_ms.add(last_interval_ms);
3480
3481 if (_cm->verbose_medium()) {
3482 gclog_or_tty->print_cr("[%d] regular clock, interval = %1.2lfms, "
3483 "scanned = %d%s, refs reached = %d%s",
3484 _task_id, last_interval_ms,
3485 _words_scanned,
3486 (_words_scanned >= _words_scanned_limit) ? " (*)" : "",
3487 _refs_reached,
3488 (_refs_reached >= _refs_reached_limit) ? " (*)" : "");
3489 }
3490 #endif // _MARKING_STATS_
3491
3492 // (4) We check whether we should yield. If we have to, then we abort.
3493 if (_cm->should_yield()) {
3494 // We should yield. To do this we abort the task. The caller is
3495 // responsible for yielding.
3496 set_has_aborted();
3497 statsOnly( ++_aborted_yield );
3498 return;
3499 }
3500
3501 // (5) We check whether we've reached our time quota. If we have,
3502 // then we abort.
3503 double elapsed_time_ms = curr_time_ms - _start_time_ms;
3504 if (elapsed_time_ms > _time_target_ms) {
3505 set_has_aborted();
3506 _has_timed_out = true;
3507 statsOnly( ++_aborted_timed_out );
3508 return;
3509 }
3510
3511 // (6) Finally, we check whether there are enough completed STAB
3512 // buffers available for processing. If there are, we abort.
3513 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3514 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
3515 if (_cm->verbose_low()) {
3516 gclog_or_tty->print_cr("[%d] aborting to deal with pending SATB buffers",
3517 _task_id);
3518 }
3519 // we do need to process SATB buffers, we'll abort and restart
3520 // the marking task to do so
3521 set_has_aborted();
3522 statsOnly( ++_aborted_satb );
3523 return;
3524 }
3525 }
3526
3527 void CMTask::recalculate_limits() {
3528 _real_words_scanned_limit = _words_scanned + words_scanned_period;
3529 _words_scanned_limit = _real_words_scanned_limit;
3530
3531 _real_refs_reached_limit = _refs_reached + refs_reached_period;
3532 _refs_reached_limit = _real_refs_reached_limit;
3533 }
3534
3535 void CMTask::decrease_limits() {
3536 // This is called when we believe that we're going to do an infrequent
3537 // operation which will increase the per byte scanned cost (i.e. move
3538 // entries to/from the global stack). It basically tries to decrease the
3539 // scanning limit so that the clock is called earlier.
3540
3541 if (_cm->verbose_medium()) {
3542 gclog_or_tty->print_cr("[%d] decreasing limits", _task_id);
3543 }
3544
3545 _words_scanned_limit = _real_words_scanned_limit -
3546 3 * words_scanned_period / 4;
3547 _refs_reached_limit = _real_refs_reached_limit -
3548 3 * refs_reached_period / 4;
3549 }
3550
3551 void CMTask::move_entries_to_global_stack() {
3552 // local array where we'll store the entries that will be popped
3553 // from the local queue
3554 oop buffer[global_stack_transfer_size];
3555
3556 int n = 0;
3557 oop obj;
3558 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
3559 buffer[n] = obj;
3560 ++n;
3561 }
3562
3563 if (n > 0) {
3564 // we popped at least one entry from the local queue
3565
3566 statsOnly( ++_global_transfers_to; _local_pops += n );
3567
3568 if (!_cm->mark_stack_push(buffer, n)) {
3569 if (_cm->verbose_low()) {
3570 gclog_or_tty->print_cr("[%d] aborting due to global stack overflow",
3571 _task_id);
3572 }
3573 set_has_aborted();
3574 } else {
3575 // the transfer was successful
3576
3577 if (_cm->verbose_medium()) {
3578 gclog_or_tty->print_cr("[%d] pushed %d entries to the global stack",
3579 _task_id, n);
3580 }
3581 statsOnly( int tmp_size = _cm->mark_stack_size();
3582 if (tmp_size > _global_max_size) {
3583 _global_max_size = tmp_size;
3584 }
3585 _global_pushes += n );
3586 }
3587 }
3588
3589 // this operation was quite expensive, so decrease the limits
3590 decrease_limits();
3591 }
3592
3593 void CMTask::get_entries_from_global_stack() {
3594 // local array where we'll store the entries that will be popped
3595 // from the global stack.
3596 oop buffer[global_stack_transfer_size];
3597 int n;
3598 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
3599 assert(n <= global_stack_transfer_size,
3600 "we should not pop more than the given limit");
3601 if (n > 0) {
3602 // yes, we did actually pop at least one entry
3603
3604 statsOnly( ++_global_transfers_from; _global_pops += n );
3605 if (_cm->verbose_medium()) {
3606 gclog_or_tty->print_cr("[%d] popped %d entries from the global stack",
3607 _task_id, n);
3608 }
3609 for (int i = 0; i < n; ++i) {
3610 bool success = _task_queue->push(buffer[i]);
3611 // We only call this when the local queue is empty or under a
3612 // given target limit. So, we do not expect this push to fail.
3613 assert(success, "invariant");
3614 }
3615
3616 statsOnly( int tmp_size = _task_queue->size();
3617 if (tmp_size > _local_max_size) {
3618 _local_max_size = tmp_size;
3619 }
3620 _local_pushes += n );
3621 }
3622
3623 // this operation was quite expensive, so decrease the limits
3624 decrease_limits();
3625 }
3626
3627 void CMTask::drain_local_queue(bool partially) {
3628 if (has_aborted()) return;
3629
3630 // Decide what the target size is, depending whether we're going to
3631 // drain it partially (so that other tasks can steal if they run out
3632 // of things to do) or totally (at the very end).
3633 size_t target_size;
3634 if (partially) {
3635 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
3636 } else {
3637 target_size = 0;
3638 }
3639
3640 if (_task_queue->size() > target_size) {
3641 if (_cm->verbose_high()) {
3642 gclog_or_tty->print_cr("[%d] draining local queue, target size = %d",
3643 _task_id, target_size);
3644 }
3645
3646 oop obj;
3647 bool ret = _task_queue->pop_local(obj);
3648 while (ret) {
3649 statsOnly( ++_local_pops );
3650
3651 if (_cm->verbose_high()) {
3652 gclog_or_tty->print_cr("[%d] popped "PTR_FORMAT, _task_id,
3653 (void*) obj);
3654 }
3655
3656 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
3657 assert(!_g1h->is_on_master_free_list(
3658 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
3659
3660 scan_object(obj);
3661
3662 if (_task_queue->size() <= target_size || has_aborted()) {
3663 ret = false;
3664 } else {
3665 ret = _task_queue->pop_local(obj);
3666 }
3667 }
3668
3669 if (_cm->verbose_high()) {
3670 gclog_or_tty->print_cr("[%d] drained local queue, size = %d",
3671 _task_id, _task_queue->size());
3672 }
3673 }
3674 }
3675
3676 void CMTask::drain_global_stack(bool partially) {
3677 if (has_aborted()) return;
3678
3679 // We have a policy to drain the local queue before we attempt to
3680 // drain the global stack.
3681 assert(partially || _task_queue->size() == 0, "invariant");
3682
3683 // Decide what the target size is, depending whether we're going to
3684 // drain it partially (so that other tasks can steal if they run out
3685 // of things to do) or totally (at the very end). Notice that,
3686 // because we move entries from the global stack in chunks or
3687 // because another task might be doing the same, we might in fact
3688 // drop below the target. But, this is not a problem.
3689 size_t target_size;
3690 if (partially) {
3691 target_size = _cm->partial_mark_stack_size_target();
3692 } else {
3693 target_size = 0;
3694 }
3695
3696 if (_cm->mark_stack_size() > target_size) {
3697 if (_cm->verbose_low()) {
3698 gclog_or_tty->print_cr("[%d] draining global_stack, target size %d",
3699 _task_id, target_size);
3700 }
3701
3702 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3703 get_entries_from_global_stack();
3704 drain_local_queue(partially);
3705 }
3706
3707 if (_cm->verbose_low()) {
3708 gclog_or_tty->print_cr("[%d] drained global stack, size = %d",
3709 _task_id, _cm->mark_stack_size());
3710 }
3711 }
3712 }
3713
3714 // SATB Queue has several assumptions on whether to call the par or
3715 // non-par versions of the methods. this is why some of the code is
3716 // replicated. We should really get rid of the single-threaded version
3717 // of the code to simplify things.
3718 void CMTask::drain_satb_buffers() {
3719 if (has_aborted()) return;
3720
3721 // We set this so that the regular clock knows that we're in the
3722 // middle of draining buffers and doesn't set the abort flag when it
3723 // notices that SATB buffers are available for draining. It'd be
3724 // very counter productive if it did that. :-)
3725 _draining_satb_buffers = true;
3726
3727 CMObjectClosure oc(this);
3728 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3729 if (G1CollectedHeap::use_parallel_gc_threads()) {
3730 satb_mq_set.set_par_closure(_task_id, &oc);
3731 } else {
3732 satb_mq_set.set_closure(&oc);
3733 }
3734
3735 // This keeps claiming and applying the closure to completed buffers
3736 // until we run out of buffers or we need to abort.
3737 if (G1CollectedHeap::use_parallel_gc_threads()) {
3738 while (!has_aborted() &&
3739 satb_mq_set.par_apply_closure_to_completed_buffer(_task_id)) {
3740 if (_cm->verbose_medium()) {
3741 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id);
3742 }
3743 statsOnly( ++_satb_buffers_processed );
3744 regular_clock_call();
3745 }
3746 } else {
3747 while (!has_aborted() &&
3748 satb_mq_set.apply_closure_to_completed_buffer()) {
3749 if (_cm->verbose_medium()) {
3750 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id);
3751 }
3752 statsOnly( ++_satb_buffers_processed );
3753 regular_clock_call();
3754 }
3755 }
3756
3757 if (!concurrent() && !has_aborted()) {
3758 // We should only do this during remark.
3759 if (G1CollectedHeap::use_parallel_gc_threads()) {
3760 satb_mq_set.par_iterate_closure_all_threads(_task_id);
3761 } else {
3762 satb_mq_set.iterate_closure_all_threads();
3763 }
3764 }
3765
3766 _draining_satb_buffers = false;
3767
3768 assert(has_aborted() ||
3769 concurrent() ||
3770 satb_mq_set.completed_buffers_num() == 0, "invariant");
3771
3772 if (G1CollectedHeap::use_parallel_gc_threads()) {
3773 satb_mq_set.set_par_closure(_task_id, NULL);
3774 } else {
3775 satb_mq_set.set_closure(NULL);
3776 }
3777
3778 // again, this was a potentially expensive operation, decrease the
3779 // limits to get the regular clock call early
3780 decrease_limits();
3781 }
3782
3783 void CMTask::print_stats() {
3784 gclog_or_tty->print_cr("Marking Stats, task = %d, calls = %d",
3785 _task_id, _calls);
3786 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
3787 _elapsed_time_ms, _termination_time_ms);
3788 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3789 _step_times_ms.num(), _step_times_ms.avg(),
3790 _step_times_ms.sd());
3791 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3792 _step_times_ms.maximum(), _step_times_ms.sum());
3793
3794 #if _MARKING_STATS_
3795 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3796 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(),
3797 _all_clock_intervals_ms.sd());
3798 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3799 _all_clock_intervals_ms.maximum(),
3800 _all_clock_intervals_ms.sum());
3801 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d",
3802 _clock_due_to_scanning, _clock_due_to_marking);
3803 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d",
3804 _objs_scanned, _objs_found_on_bitmap);
3805 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d",
3806 _local_pushes, _local_pops, _local_max_size);
3807 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d",
3808 _global_pushes, _global_pops, _global_max_size);
3809 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d",
3810 _global_transfers_to,_global_transfers_from);
3811 gclog_or_tty->print_cr(" Regions: claimed = %d", _regions_claimed);
3812 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed);
3813 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d",
3814 _steal_attempts, _steals);
3815 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted);
3816 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d",
3817 _aborted_overflow, _aborted_cm_aborted, _aborted_yield);
3818 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d",
3819 _aborted_timed_out, _aborted_satb, _aborted_termination);
3820 #endif // _MARKING_STATS_
3821 }
3822
3823 /*****************************************************************************
3824
3825 The do_marking_step(time_target_ms, ...) method is the building
3826 block of the parallel marking framework. It can be called in parallel
3827 with other invocations of do_marking_step() on different tasks
3828 (but only one per task, obviously) and concurrently with the
3829 mutator threads, or during remark, hence it eliminates the need
3830 for two versions of the code. When called during remark, it will
3831 pick up from where the task left off during the concurrent marking
3832 phase. Interestingly, tasks are also claimable during evacuation
3833 pauses too, since do_marking_step() ensures that it aborts before
3834 it needs to yield.
3835
3836 The data structures that it uses to do marking work are the
3837 following:
3838
3839 (1) Marking Bitmap. If there are gray objects that appear only
3840 on the bitmap (this happens either when dealing with an overflow
3841 or when the initial marking phase has simply marked the roots
3842 and didn't push them on the stack), then tasks claim heap
3843 regions whose bitmap they then scan to find gray objects. A
3844 global finger indicates where the end of the last claimed region
3845 is. A local finger indicates how far into the region a task has
3846 scanned. The two fingers are used to determine how to gray an
3847 object (i.e. whether simply marking it is OK, as it will be
3848 visited by a task in the future, or whether it needs to be also
3849 pushed on a stack).
3850
3851 (2) Local Queue. The local queue of the task which is accessed
3852 reasonably efficiently by the task. Other tasks can steal from
3853 it when they run out of work. Throughout the marking phase, a
3854 task attempts to keep its local queue short but not totally
3855 empty, so that entries are available for stealing by other
3856 tasks. Only when there is no more work, a task will totally
3857 drain its local queue.
3858
3859 (3) Global Mark Stack. This handles local queue overflow. During
3860 marking only sets of entries are moved between it and the local
3861 queues, as access to it requires a mutex and more fine-grain
3862 interaction with it which might cause contention. If it
3863 overflows, then the marking phase should restart and iterate
3864 over the bitmap to identify gray objects. Throughout the marking
3865 phase, tasks attempt to keep the global mark stack at a small
3866 length but not totally empty, so that entries are available for
3867 popping by other tasks. Only when there is no more work, tasks
3868 will totally drain the global mark stack.
3869
3870 (4) SATB Buffer Queue. This is where completed SATB buffers are
3871 made available. Buffers are regularly removed from this queue
3872 and scanned for roots, so that the queue doesn't get too
3873 long. During remark, all completed buffers are processed, as
3874 well as the filled in parts of any uncompleted buffers.
3875
3876 The do_marking_step() method tries to abort when the time target
3877 has been reached. There are a few other cases when the
3878 do_marking_step() method also aborts:
3879
3880 (1) When the marking phase has been aborted (after a Full GC).
3881
3882 (2) When a global overflow (on the global stack) has been
3883 triggered. Before the task aborts, it will actually sync up with
3884 the other tasks to ensure that all the marking data structures
3885 (local queues, stacks, fingers etc.) are re-initialized so that
3886 when do_marking_step() completes, the marking phase can
3887 immediately restart.
3888
3889 (3) When enough completed SATB buffers are available. The
3890 do_marking_step() method only tries to drain SATB buffers right
3891 at the beginning. So, if enough buffers are available, the
3892 marking step aborts and the SATB buffers are processed at
3893 the beginning of the next invocation.
3894
3895 (4) To yield. when we have to yield then we abort and yield
3896 right at the end of do_marking_step(). This saves us from a lot
3897 of hassle as, by yielding we might allow a Full GC. If this
3898 happens then objects will be compacted underneath our feet, the
3899 heap might shrink, etc. We save checking for this by just
3900 aborting and doing the yield right at the end.
3901
3902 From the above it follows that the do_marking_step() method should
3903 be called in a loop (or, otherwise, regularly) until it completes.
3904
3905 If a marking step completes without its has_aborted() flag being
3906 true, it means it has completed the current marking phase (and
3907 also all other marking tasks have done so and have all synced up).
3908
3909 A method called regular_clock_call() is invoked "regularly" (in
3910 sub ms intervals) throughout marking. It is this clock method that
3911 checks all the abort conditions which were mentioned above and
3912 decides when the task should abort. A work-based scheme is used to
3913 trigger this clock method: when the number of object words the
3914 marking phase has scanned or the number of references the marking
3915 phase has visited reach a given limit. Additional invocations to
3916 the method clock have been planted in a few other strategic places
3917 too. The initial reason for the clock method was to avoid calling
3918 vtime too regularly, as it is quite expensive. So, once it was in
3919 place, it was natural to piggy-back all the other conditions on it
3920 too and not constantly check them throughout the code.
3921
3922 If do_termination is true then do_marking_step will enter its
3923 termination protocol.
3924
3925 The value of is_serial must be true when do_marking_step is being
3926 called serially (i.e. by the VMThread) and do_marking_step should
3927 skip any synchronization in the termination and overflow code.
3928 Examples include the serial remark code and the serial reference
3929 processing closures.
3930
3931 The value of is_serial must be false when do_marking_step is
3932 being called by any of the worker threads in a work gang.
3933 Examples include the concurrent marking code (CMMarkingTask),
3934 the MT remark code, and the MT reference processing closures.
3935
3936 *****************************************************************************/
3937
3938 void CMTask::do_marking_step(double time_target_ms,
3939 bool do_termination,
3940 bool is_serial) {
3941 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
3942 assert(concurrent() == _cm->concurrent(), "they should be the same");
3943
3944 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
3945 assert(_task_queues != NULL, "invariant");
3946 assert(_task_queue != NULL, "invariant");
3947 assert(_task_queues->queue(_task_id) == _task_queue, "invariant");
3948
3949 assert(!_claimed,
3950 "only one thread should claim this task at any one time");
3951
3952 // OK, this doesn't safeguard again all possible scenarios, as it is
3953 // possible for two threads to set the _claimed flag at the same
3954 // time. But it is only for debugging purposes anyway and it will
3955 // catch most problems.
3956 _claimed = true;
3957
3958 _start_time_ms = os::elapsedVTime() * 1000.0;
3959 statsOnly( _interval_start_time_ms = _start_time_ms );
3960
3961 // If do_stealing is true then do_marking_step will attempt to
3962 // steal work from the other CMTasks. It only makes sense to
3963 // enable stealing when the termination protocol is enabled
3964 // and do_marking_step() is not being called serially.
3965 bool do_stealing = do_termination && !is_serial;
3966
3967 double diff_prediction_ms =
3968 g1_policy->get_new_prediction(&_marking_step_diffs_ms);
3969 _time_target_ms = time_target_ms - diff_prediction_ms;
3970
3971 // set up the variables that are used in the work-based scheme to
3972 // call the regular clock method
3973 _words_scanned = 0;
3974 _refs_reached = 0;
3975 recalculate_limits();
3976
3977 // clear all flags
3978 clear_has_aborted();
3979 _has_timed_out = false;
3980 _draining_satb_buffers = false;
3981
3982 ++_calls;
3983
3984 if (_cm->verbose_low()) {
3985 gclog_or_tty->print_cr("[%d] >>>>>>>>>> START, call = %d, "
3986 "target = %1.2lfms >>>>>>>>>>",
3987 _task_id, _calls, _time_target_ms);
3988 }
3989
3990 // Set up the bitmap and oop closures. Anything that uses them is
3991 // eventually called from this method, so it is OK to allocate these
3992 // statically.
3993 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
3994 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
3995 set_cm_oop_closure(&cm_oop_closure);
3996
3997 if (_cm->has_overflown()) {
3998 // This can happen if the mark stack overflows during a GC pause
3999 // and this task, after a yield point, restarts. We have to abort
4000 // as we need to get into the overflow protocol which happens
4001 // right at the end of this task.
4002 set_has_aborted();
4003 }
4004
4005 // First drain any available SATB buffers. After this, we will not
4006 // look at SATB buffers before the next invocation of this method.
4007 // If enough completed SATB buffers are queued up, the regular clock
4008 // will abort this task so that it restarts.
4009 drain_satb_buffers();
4010 // ...then partially drain the local queue and the global stack
4011 drain_local_queue(true);
4012 drain_global_stack(true);
4013
4014 do {
4015 if (!has_aborted() && _curr_region != NULL) {
4016 // This means that we're already holding on to a region.
4017 assert(_finger != NULL, "if region is not NULL, then the finger "
4018 "should not be NULL either");
4019
4020 // We might have restarted this task after an evacuation pause
4021 // which might have evacuated the region we're holding on to
4022 // underneath our feet. Let's read its limit again to make sure
4023 // that we do not iterate over a region of the heap that
4024 // contains garbage (update_region_limit() will also move
4025 // _finger to the start of the region if it is found empty).
4026 update_region_limit();
4027 // We will start from _finger not from the start of the region,
4028 // as we might be restarting this task after aborting half-way
4029 // through scanning this region. In this case, _finger points to
4030 // the address where we last found a marked object. If this is a
4031 // fresh region, _finger points to start().
4032 MemRegion mr = MemRegion(_finger, _region_limit);
4033
4034 if (_cm->verbose_low()) {
4035 gclog_or_tty->print_cr("[%d] we're scanning part "
4036 "["PTR_FORMAT", "PTR_FORMAT") "
4037 "of region "PTR_FORMAT,
4038 _task_id, _finger, _region_limit, _curr_region);
4039 }
4040
4041 // Let's iterate over the bitmap of the part of the
4042 // region that is left.
4043 if (mr.is_empty() || _nextMarkBitMap->iterate(&bitmap_closure, mr)) {
4044 // We successfully completed iterating over the region. Now,
4045 // let's give up the region.
4046 giveup_current_region();
4047 regular_clock_call();
4048 } else {
4049 assert(has_aborted(), "currently the only way to do so");
4050 // The only way to abort the bitmap iteration is to return
4051 // false from the do_bit() method. However, inside the
4052 // do_bit() method we move the _finger to point to the
4053 // object currently being looked at. So, if we bail out, we
4054 // have definitely set _finger to something non-null.
4055 assert(_finger != NULL, "invariant");
4056
4057 // Region iteration was actually aborted. So now _finger
4058 // points to the address of the object we last scanned. If we
4059 // leave it there, when we restart this task, we will rescan
4060 // the object. It is easy to avoid this. We move the finger by
4061 // enough to point to the next possible object header (the
4062 // bitmap knows by how much we need to move it as it knows its
4063 // granularity).
4064 assert(_finger < _region_limit, "invariant");
4065 HeapWord* new_finger = _nextMarkBitMap->nextWord(_finger);
4066 // Check if bitmap iteration was aborted while scanning the last object
4067 if (new_finger >= _region_limit) {
4068 giveup_current_region();
4069 } else {
4070 move_finger_to(new_finger);
4071 }
4072 }
4073 }
4074 // At this point we have either completed iterating over the
4075 // region we were holding on to, or we have aborted.
4076
4077 // We then partially drain the local queue and the global stack.
4078 // (Do we really need this?)
4079 drain_local_queue(true);
4080 drain_global_stack(true);
4081
4082 // Read the note on the claim_region() method on why it might
4083 // return NULL with potentially more regions available for
4084 // claiming and why we have to check out_of_regions() to determine
4085 // whether we're done or not.
4086 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
4087 // We are going to try to claim a new region. We should have
4088 // given up on the previous one.
4089 // Separated the asserts so that we know which one fires.
4090 assert(_curr_region == NULL, "invariant");
4091 assert(_finger == NULL, "invariant");
4092 assert(_region_limit == NULL, "invariant");
4093 if (_cm->verbose_low()) {
4094 gclog_or_tty->print_cr("[%d] trying to claim a new region", _task_id);
4095 }
4096 HeapRegion* claimed_region = _cm->claim_region(_task_id);
4097 if (claimed_region != NULL) {
4098 // Yes, we managed to claim one
4099 statsOnly( ++_regions_claimed );
4100
4101 if (_cm->verbose_low()) {
4102 gclog_or_tty->print_cr("[%d] we successfully claimed "
4103 "region "PTR_FORMAT,
4104 _task_id, claimed_region);
4105 }
4106
4107 setup_for_region(claimed_region);
4108 assert(_curr_region == claimed_region, "invariant");
4109 }
4110 // It is important to call the regular clock here. It might take
4111 // a while to claim a region if, for example, we hit a large
4112 // block of empty regions. So we need to call the regular clock
4113 // method once round the loop to make sure it's called
4114 // frequently enough.
4115 regular_clock_call();
4116 }
4117
4118 if (!has_aborted() && _curr_region == NULL) {
4119 assert(_cm->out_of_regions(),
4120 "at this point we should be out of regions");
4121 }
4122 } while ( _curr_region != NULL && !has_aborted());
4123
4124 if (!has_aborted()) {
4125 // We cannot check whether the global stack is empty, since other
4126 // tasks might be pushing objects to it concurrently.
4127 assert(_cm->out_of_regions(),
4128 "at this point we should be out of regions");
4129
4130 if (_cm->verbose_low()) {
4131 gclog_or_tty->print_cr("[%d] all regions claimed", _task_id);
4132 }
4133
4134 // Try to reduce the number of available SATB buffers so that
4135 // remark has less work to do.
4136 drain_satb_buffers();
4137 }
4138
4139 // Since we've done everything else, we can now totally drain the
4140 // local queue and global stack.
4141 drain_local_queue(false);
4142 drain_global_stack(false);
4143
4144 // Attempt at work stealing from other task's queues.
4145 if (do_stealing && !has_aborted()) {
4146 // We have not aborted. This means that we have finished all that
4147 // we could. Let's try to do some stealing...
4148
4149 // We cannot check whether the global stack is empty, since other
4150 // tasks might be pushing objects to it concurrently.
4151 assert(_cm->out_of_regions() && _task_queue->size() == 0,
4152 "only way to reach here");
4153
4154 if (_cm->verbose_low()) {
4155 gclog_or_tty->print_cr("[%d] starting to steal", _task_id);
4156 }
4157
4158 while (!has_aborted()) {
4159 oop obj;
4160 statsOnly( ++_steal_attempts );
4161
4162 if (_cm->try_stealing(_task_id, &_hash_seed, obj)) {
4163 if (_cm->verbose_medium()) {
4164 gclog_or_tty->print_cr("[%d] stolen "PTR_FORMAT" successfully",
4165 _task_id, (void*) obj);
4166 }
4167
4168 statsOnly( ++_steals );
4169
4170 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
4171 "any stolen object should be marked");
4172 scan_object(obj);
4173
4174 // And since we're towards the end, let's totally drain the
4175 // local queue and global stack.
4176 drain_local_queue(false);
4177 drain_global_stack(false);
4178 } else {
4179 break;
4180 }
4181 }
4182 }
4183
4184 // If we are about to wrap up and go into termination, check if we
4185 // should raise the overflow flag.
4186 if (do_termination && !has_aborted()) {
4187 if (_cm->force_overflow()->should_force()) {
4188 _cm->set_has_overflown();
4189 regular_clock_call();
4190 }
4191 }
4192
4193 // We still haven't aborted. Now, let's try to get into the
4194 // termination protocol.
4195 if (do_termination && !has_aborted()) {
4196 // We cannot check whether the global stack is empty, since other
4197 // tasks might be concurrently pushing objects on it.
4198 // Separated the asserts so that we know which one fires.
4199 assert(_cm->out_of_regions(), "only way to reach here");
4200 assert(_task_queue->size() == 0, "only way to reach here");
4201
4202 if (_cm->verbose_low()) {
4203 gclog_or_tty->print_cr("[%d] starting termination protocol", _task_id);
4204 }
4205
4206 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
4207
4208 // The CMTask class also extends the TerminatorTerminator class,
4209 // hence its should_exit_termination() method will also decide
4210 // whether to exit the termination protocol or not.
4211 bool finished = (is_serial ||
4212 _cm->terminator()->offer_termination(this));
4213 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
4214 _termination_time_ms +=
4215 termination_end_time_ms - _termination_start_time_ms;
4216
4217 if (finished) {
4218 // We're all done.
4219
4220 if (_task_id == 0) {
4221 // let's allow task 0 to do this
4222 if (concurrent()) {
4223 assert(_cm->concurrent_marking_in_progress(), "invariant");
4224 // we need to set this to false before the next
4225 // safepoint. This way we ensure that the marking phase
4226 // doesn't observe any more heap expansions.
4227 _cm->clear_concurrent_marking_in_progress();
4228 }
4229 }
4230
4231 // We can now guarantee that the global stack is empty, since
4232 // all other tasks have finished. We separated the guarantees so
4233 // that, if a condition is false, we can immediately find out
4234 // which one.
4235 guarantee(_cm->out_of_regions(), "only way to reach here");
4236 guarantee(_cm->mark_stack_empty(), "only way to reach here");
4237 guarantee(_task_queue->size() == 0, "only way to reach here");
4238 guarantee(!_cm->has_overflown(), "only way to reach here");
4239 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
4240
4241 if (_cm->verbose_low()) {
4242 gclog_or_tty->print_cr("[%d] all tasks terminated", _task_id);
4243 }
4244 } else {
4245 // Apparently there's more work to do. Let's abort this task. It
4246 // will restart it and we can hopefully find more things to do.
4247
4248 if (_cm->verbose_low()) {
4249 gclog_or_tty->print_cr("[%d] apparently there is more work to do",
4250 _task_id);
4251 }
4252
4253 set_has_aborted();
4254 statsOnly( ++_aborted_termination );
4255 }
4256 }
4257
4258 // Mainly for debugging purposes to make sure that a pointer to the
4259 // closure which was statically allocated in this frame doesn't
4260 // escape it by accident.
4261 set_cm_oop_closure(NULL);
4262 double end_time_ms = os::elapsedVTime() * 1000.0;
4263 double elapsed_time_ms = end_time_ms - _start_time_ms;
4264 // Update the step history.
4265 _step_times_ms.add(elapsed_time_ms);
4266
4267 if (has_aborted()) {
4268 // The task was aborted for some reason.
4269
4270 statsOnly( ++_aborted );
4271
4272 if (_has_timed_out) {
4273 double diff_ms = elapsed_time_ms - _time_target_ms;
4274 // Keep statistics of how well we did with respect to hitting
4275 // our target only if we actually timed out (if we aborted for
4276 // other reasons, then the results might get skewed).
4277 _marking_step_diffs_ms.add(diff_ms);
4278 }
4279
4280 if (_cm->has_overflown()) {
4281 // This is the interesting one. We aborted because a global
4282 // overflow was raised. This means we have to restart the
4283 // marking phase and start iterating over regions. However, in
4284 // order to do this we have to make sure that all tasks stop
4285 // what they are doing and re-initialise in a safe manner. We
4286 // will achieve this with the use of two barrier sync points.
4287
4288 if (_cm->verbose_low()) {
4289 gclog_or_tty->print_cr("[%d] detected overflow", _task_id);
4290 }
4291
4292 if (!is_serial) {
4293 // We only need to enter the sync barrier if being called
4294 // from a parallel context
4295 _cm->enter_first_sync_barrier(_task_id);
4296
4297 // When we exit this sync barrier we know that all tasks have
4298 // stopped doing marking work. So, it's now safe to
4299 // re-initialise our data structures. At the end of this method,
4300 // task 0 will clear the global data structures.
4301 }
4302
4303 statsOnly( ++_aborted_overflow );
4304
4305 // We clear the local state of this task...
4306 clear_region_fields();
4307
4308 if (!is_serial) {
4309 // ...and enter the second barrier.
4310 _cm->enter_second_sync_barrier(_task_id);
4311 }
4312 // At this point, if we're during the concurrent phase of
4313 // marking, everything has been re-initialized and we're
4314 // ready to restart.
4315 }
4316
4317 if (_cm->verbose_low()) {
4318 gclog_or_tty->print_cr("[%d] <<<<<<<<<< ABORTING, target = %1.2lfms, "
4319 "elapsed = %1.2lfms <<<<<<<<<<",
4320 _task_id, _time_target_ms, elapsed_time_ms);
4321 if (_cm->has_aborted()) {
4322 gclog_or_tty->print_cr("[%d] ========== MARKING ABORTED ==========",
4323 _task_id);
4324 }
4325 }
4326 } else {
4327 if (_cm->verbose_low()) {
4328 gclog_or_tty->print_cr("[%d] <<<<<<<<<< FINISHED, target = %1.2lfms, "
4329 "elapsed = %1.2lfms <<<<<<<<<<",
4330 _task_id, _time_target_ms, elapsed_time_ms);
4331 }
4332 }
4333
4334 _claimed = false;
4335 }
4336
4337 CMTask::CMTask(int task_id,
4338 ConcurrentMark* cm,
4339 size_t* marked_bytes,
4340 BitMap* card_bm,
4341 CMTaskQueue* task_queue,
4342 CMTaskQueueSet* task_queues)
4343 : _g1h(G1CollectedHeap::heap()),
4344 _task_id(task_id), _cm(cm),
4345 _claimed(false),
4346 _nextMarkBitMap(NULL), _hash_seed(17),
4347 _task_queue(task_queue),
4348 _task_queues(task_queues),
4349 _cm_oop_closure(NULL),
4350 _marked_bytes_array(marked_bytes),
4351 _card_bm(card_bm) {
4352 guarantee(task_queue != NULL, "invariant");
4353 guarantee(task_queues != NULL, "invariant");
4354
4355 statsOnly( _clock_due_to_scanning = 0;
4356 _clock_due_to_marking = 0 );
4357
4358 _marking_step_diffs_ms.add(0.5);
4359 }
4360
4361 // These are formatting macros that are used below to ensure
4362 // consistent formatting. The *_H_* versions are used to format the
4363 // header for a particular value and they should be kept consistent
4364 // with the corresponding macro. Also note that most of the macros add
4365 // the necessary white space (as a prefix) which makes them a bit
4366 // easier to compose.
4367
4368 // All the output lines are prefixed with this string to be able to
4369 // identify them easily in a large log file.
4370 #define G1PPRL_LINE_PREFIX "###"
4371
4372 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT
4373 #ifdef _LP64
4374 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
4375 #else // _LP64
4376 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
4377 #endif // _LP64
4378
4379 // For per-region info
4380 #define G1PPRL_TYPE_FORMAT " %-4s"
4381 #define G1PPRL_TYPE_H_FORMAT " %4s"
4382 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9)
4383 #define G1PPRL_BYTE_H_FORMAT " %9s"
4384 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
4385 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
4386
4387 // For summary info
4388 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT
4389 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT
4390 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB"
4391 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%"
4392
4393 G1PrintRegionLivenessInfoClosure::
4394 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name)
4395 : _out(out),
4396 _total_used_bytes(0), _total_capacity_bytes(0),
4397 _total_prev_live_bytes(0), _total_next_live_bytes(0),
4398 _hum_used_bytes(0), _hum_capacity_bytes(0),
4399 _hum_prev_live_bytes(0), _hum_next_live_bytes(0) {
4400 G1CollectedHeap* g1h = G1CollectedHeap::heap();
4401 MemRegion g1_committed = g1h->g1_committed();
4402 MemRegion g1_reserved = g1h->g1_reserved();
4403 double now = os::elapsedTime();
4404
4405 // Print the header of the output.
4406 _out->cr();
4407 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
4408 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP"
4409 G1PPRL_SUM_ADDR_FORMAT("committed")
4410 G1PPRL_SUM_ADDR_FORMAT("reserved")
4411 G1PPRL_SUM_BYTE_FORMAT("region-size"),
4412 g1_committed.start(), g1_committed.end(),
4413 g1_reserved.start(), g1_reserved.end(),
4414 HeapRegion::GrainBytes);
4415 _out->print_cr(G1PPRL_LINE_PREFIX);
4416 _out->print_cr(G1PPRL_LINE_PREFIX
4417 G1PPRL_TYPE_H_FORMAT
4418 G1PPRL_ADDR_BASE_H_FORMAT
4419 G1PPRL_BYTE_H_FORMAT
4420 G1PPRL_BYTE_H_FORMAT
4421 G1PPRL_BYTE_H_FORMAT
4422 G1PPRL_DOUBLE_H_FORMAT,
4423 "type", "address-range",
4424 "used", "prev-live", "next-live", "gc-eff");
4425 _out->print_cr(G1PPRL_LINE_PREFIX
4426 G1PPRL_TYPE_H_FORMAT
4427 G1PPRL_ADDR_BASE_H_FORMAT
4428 G1PPRL_BYTE_H_FORMAT
4429 G1PPRL_BYTE_H_FORMAT
4430 G1PPRL_BYTE_H_FORMAT
4431 G1PPRL_DOUBLE_H_FORMAT,
4432 "", "",
4433 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)");
4434 }
4435
4436 // It takes as a parameter a reference to one of the _hum_* fields, it
4437 // deduces the corresponding value for a region in a humongous region
4438 // series (either the region size, or what's left if the _hum_* field
4439 // is < the region size), and updates the _hum_* field accordingly.
4440 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
4441 size_t bytes = 0;
4442 // The > 0 check is to deal with the prev and next live bytes which
4443 // could be 0.
4444 if (*hum_bytes > 0) {
4445 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
4446 *hum_bytes -= bytes;
4447 }
4448 return bytes;
4449 }
4450
4451 // It deduces the values for a region in a humongous region series
4452 // from the _hum_* fields and updates those accordingly. It assumes
4453 // that that _hum_* fields have already been set up from the "starts
4454 // humongous" region and we visit the regions in address order.
4455 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
4456 size_t* capacity_bytes,
4457 size_t* prev_live_bytes,
4458 size_t* next_live_bytes) {
4459 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
4460 *used_bytes = get_hum_bytes(&_hum_used_bytes);
4461 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
4462 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
4463 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
4464 }
4465
4466 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
4467 const char* type = "";
4468 HeapWord* bottom = r->bottom();
4469 HeapWord* end = r->end();
4470 size_t capacity_bytes = r->capacity();
4471 size_t used_bytes = r->used();
4472 size_t prev_live_bytes = r->live_bytes();
4473 size_t next_live_bytes = r->next_live_bytes();
4474 double gc_eff = r->gc_efficiency();
4475 if (r->used() == 0) {
4476 type = "FREE";
4477 } else if (r->is_survivor()) {
4478 type = "SURV";
4479 } else if (r->is_young()) {
4480 type = "EDEN";
4481 } else if (r->startsHumongous()) {
4482 type = "HUMS";
4483
4484 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
4485 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
4486 "they should have been zeroed after the last time we used them");
4487 // Set up the _hum_* fields.
4488 _hum_capacity_bytes = capacity_bytes;
4489 _hum_used_bytes = used_bytes;
4490 _hum_prev_live_bytes = prev_live_bytes;
4491 _hum_next_live_bytes = next_live_bytes;
4492 get_hum_bytes(&used_bytes, &capacity_bytes,
4493 &prev_live_bytes, &next_live_bytes);
4494 end = bottom + HeapRegion::GrainWords;
4495 } else if (r->continuesHumongous()) {
4496 type = "HUMC";
4497 get_hum_bytes(&used_bytes, &capacity_bytes,
4498 &prev_live_bytes, &next_live_bytes);
4499 assert(end == bottom + HeapRegion::GrainWords, "invariant");
4500 } else {
4501 type = "OLD";
4502 }
4503
4504 _total_used_bytes += used_bytes;
4505 _total_capacity_bytes += capacity_bytes;
4506 _total_prev_live_bytes += prev_live_bytes;
4507 _total_next_live_bytes += next_live_bytes;
4508
4509 // Print a line for this particular region.
4510 _out->print_cr(G1PPRL_LINE_PREFIX
4511 G1PPRL_TYPE_FORMAT
4512 G1PPRL_ADDR_BASE_FORMAT
4513 G1PPRL_BYTE_FORMAT
4514 G1PPRL_BYTE_FORMAT
4515 G1PPRL_BYTE_FORMAT
4516 G1PPRL_DOUBLE_FORMAT,
4517 type, bottom, end,
4518 used_bytes, prev_live_bytes, next_live_bytes, gc_eff);
4519
4520 return false;
4521 }
4522
4523 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
4524 // Print the footer of the output.
4525 _out->print_cr(G1PPRL_LINE_PREFIX);
4526 _out->print_cr(G1PPRL_LINE_PREFIX
4527 " SUMMARY"
4528 G1PPRL_SUM_MB_FORMAT("capacity")
4529 G1PPRL_SUM_MB_PERC_FORMAT("used")
4530 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
4531 G1PPRL_SUM_MB_PERC_FORMAT("next-live"),
4532 bytes_to_mb(_total_capacity_bytes),
4533 bytes_to_mb(_total_used_bytes),
4534 perc(_total_used_bytes, _total_capacity_bytes),
4535 bytes_to_mb(_total_prev_live_bytes),
4536 perc(_total_prev_live_bytes, _total_capacity_bytes),
4537 bytes_to_mb(_total_next_live_bytes),
4538 perc(_total_next_live_bytes, _total_capacity_bytes));
4539 _out->cr();
4540 }