1 /*
2 * Copyright (c) 2001, 2011, 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/g1OopClosures.inline.hpp"
33 #include "gc_implementation/g1/g1RemSet.hpp"
34 #include "gc_implementation/g1/heapRegionRemSet.hpp"
35 #include "gc_implementation/g1/heapRegionSeq.inline.hpp"
36 #include "gc_implementation/shared/vmGCOperations.hpp"
37 #include "memory/genOopClosures.inline.hpp"
38 #include "memory/referencePolicy.hpp"
39 #include "memory/resourceArea.hpp"
40 #include "oops/oop.inline.hpp"
41 #include "runtime/handles.inline.hpp"
42 #include "runtime/java.hpp"
43
44 //
45 // CMS Bit Map Wrapper
46
47 CMBitMapRO::CMBitMapRO(ReservedSpace rs, int shifter) :
48 _bm((uintptr_t*)NULL,0),
49 _shifter(shifter) {
50 _bmStartWord = (HeapWord*)(rs.base());
51 _bmWordSize = rs.size()/HeapWordSize; // rs.size() is in bytes
52 ReservedSpace brs(ReservedSpace::allocation_align_size_up(
53 (_bmWordSize >> (_shifter + LogBitsPerByte)) + 1));
54
55 guarantee(brs.is_reserved(), "couldn't allocate CMS bit map");
56 // For now we'll just commit all of the bit map up fromt.
57 // Later on we'll try to be more parsimonious with swap.
58 guarantee(_virtual_space.initialize(brs, brs.size()),
59 "couldn't reseve backing store for CMS bit map");
60 assert(_virtual_space.committed_size() == brs.size(),
61 "didn't reserve backing store for all of CMS bit map?");
62 _bm.set_map((uintptr_t*)_virtual_space.low());
63 assert(_virtual_space.committed_size() << (_shifter + LogBitsPerByte) >=
64 _bmWordSize, "inconsistency in bit map sizing");
65 _bm.set_size(_bmWordSize >> _shifter);
66 }
67
68 HeapWord* CMBitMapRO::getNextMarkedWordAddress(HeapWord* addr,
69 HeapWord* limit) const {
70 // First we must round addr *up* to a possible object boundary.
71 addr = (HeapWord*)align_size_up((intptr_t)addr,
72 HeapWordSize << _shifter);
73 size_t addrOffset = heapWordToOffset(addr);
74 if (limit == NULL) {
75 limit = _bmStartWord + _bmWordSize;
76 }
77 size_t limitOffset = heapWordToOffset(limit);
78 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
79 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
80 assert(nextAddr >= addr, "get_next_one postcondition");
81 assert(nextAddr == limit || isMarked(nextAddr),
82 "get_next_one postcondition");
83 return nextAddr;
84 }
85
86 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(HeapWord* addr,
87 HeapWord* limit) const {
88 size_t addrOffset = heapWordToOffset(addr);
89 if (limit == NULL) {
90 limit = _bmStartWord + _bmWordSize;
91 }
92 size_t limitOffset = heapWordToOffset(limit);
93 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset);
94 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
95 assert(nextAddr >= addr, "get_next_one postcondition");
96 assert(nextAddr == limit || !isMarked(nextAddr),
97 "get_next_one postcondition");
98 return nextAddr;
99 }
100
101 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const {
102 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check");
103 return (int) (diff >> _shifter);
104 }
105
106 bool CMBitMapRO::iterate(BitMapClosure* cl, MemRegion mr) {
107 HeapWord* left = MAX2(_bmStartWord, mr.start());
108 HeapWord* right = MIN2(_bmStartWord + _bmWordSize, mr.end());
109 if (right > left) {
110 // Right-open interval [leftOffset, rightOffset).
111 return _bm.iterate(cl, heapWordToOffset(left), heapWordToOffset(right));
112 } else {
113 return true;
114 }
115 }
116
117 void CMBitMapRO::mostly_disjoint_range_union(BitMap* from_bitmap,
118 size_t from_start_index,
119 HeapWord* to_start_word,
120 size_t word_num) {
121 _bm.mostly_disjoint_range_union(from_bitmap,
122 from_start_index,
123 heapWordToOffset(to_start_word),
124 word_num);
125 }
126
127 #ifndef PRODUCT
128 bool CMBitMapRO::covers(ReservedSpace rs) const {
129 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
130 assert(((size_t)_bm.size() * (size_t)(1 << _shifter)) == _bmWordSize,
131 "size inconsistency");
132 return _bmStartWord == (HeapWord*)(rs.base()) &&
133 _bmWordSize == rs.size()>>LogHeapWordSize;
134 }
135 #endif
136
137 void CMBitMap::clearAll() {
138 _bm.clear();
139 return;
140 }
141
142 void CMBitMap::markRange(MemRegion mr) {
143 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
144 assert(!mr.is_empty(), "unexpected empty region");
145 assert((offsetToHeapWord(heapWordToOffset(mr.end())) ==
146 ((HeapWord *) mr.end())),
147 "markRange memory region end is not card aligned");
148 // convert address range into offset range
149 _bm.at_put_range(heapWordToOffset(mr.start()),
150 heapWordToOffset(mr.end()), true);
151 }
152
153 void CMBitMap::clearRange(MemRegion mr) {
154 mr.intersection(MemRegion(_bmStartWord, _bmWordSize));
155 assert(!mr.is_empty(), "unexpected empty region");
156 // convert address range into offset range
157 _bm.at_put_range(heapWordToOffset(mr.start()),
158 heapWordToOffset(mr.end()), false);
159 }
160
161 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr,
162 HeapWord* end_addr) {
163 HeapWord* start = getNextMarkedWordAddress(addr);
164 start = MIN2(start, end_addr);
165 HeapWord* end = getNextUnmarkedWordAddress(start);
166 end = MIN2(end, end_addr);
167 assert(start <= end, "Consistency check");
168 MemRegion mr(start, end);
169 if (!mr.is_empty()) {
170 clearRange(mr);
171 }
172 return mr;
173 }
174
175 CMMarkStack::CMMarkStack(ConcurrentMark* cm) :
176 _base(NULL), _cm(cm)
177 #ifdef ASSERT
178 , _drain_in_progress(false)
179 , _drain_in_progress_yields(false)
180 #endif
181 {}
182
183 void CMMarkStack::allocate(size_t size) {
184 _base = NEW_C_HEAP_ARRAY(oop, size);
185 if (_base == NULL) {
186 vm_exit_during_initialization("Failed to allocate "
187 "CM region mark stack");
188 }
189 _index = 0;
190 _capacity = (jint) size;
191 _oops_do_bound = -1;
192 NOT_PRODUCT(_max_depth = 0);
193 }
194
195 CMMarkStack::~CMMarkStack() {
196 if (_base != NULL) {
197 FREE_C_HEAP_ARRAY(oop, _base);
198 }
199 }
200
201 void CMMarkStack::par_push(oop ptr) {
202 while (true) {
203 if (isFull()) {
204 _overflow = true;
205 return;
206 }
207 // Otherwise...
208 jint index = _index;
209 jint next_index = index+1;
210 jint res = Atomic::cmpxchg(next_index, &_index, index);
211 if (res == index) {
212 _base[index] = ptr;
213 // Note that we don't maintain this atomically. We could, but it
214 // doesn't seem necessary.
215 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
216 return;
217 }
218 // Otherwise, we need to try again.
219 }
220 }
221
222 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) {
223 while (true) {
224 if (isFull()) {
225 _overflow = true;
226 return;
227 }
228 // Otherwise...
229 jint index = _index;
230 jint next_index = index + n;
231 if (next_index > _capacity) {
232 _overflow = true;
233 return;
234 }
235 jint res = Atomic::cmpxchg(next_index, &_index, index);
236 if (res == index) {
237 for (int i = 0; i < n; i++) {
238 int ind = index + i;
239 assert(ind < _capacity, "By overflow test above.");
240 _base[ind] = ptr_arr[i];
241 }
242 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
243 return;
244 }
245 // Otherwise, we need to try again.
246 }
247 }
248
249
250 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
251 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
252 jint start = _index;
253 jint next_index = start + n;
254 if (next_index > _capacity) {
255 _overflow = true;
256 return;
257 }
258 // Otherwise.
259 _index = next_index;
260 for (int i = 0; i < n; i++) {
261 int ind = start + i;
262 assert(ind < _capacity, "By overflow test above.");
263 _base[ind] = ptr_arr[i];
264 }
265 }
266
267
268 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
269 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
270 jint index = _index;
271 if (index == 0) {
272 *n = 0;
273 return false;
274 } else {
275 int k = MIN2(max, index);
276 jint new_ind = index - k;
277 for (int j = 0; j < k; j++) {
278 ptr_arr[j] = _base[new_ind + j];
279 }
280 _index = new_ind;
281 *n = k;
282 return true;
283 }
284 }
285
286
287 CMRegionStack::CMRegionStack() : _base(NULL) {}
288
289 void CMRegionStack::allocate(size_t size) {
290 _base = NEW_C_HEAP_ARRAY(MemRegion, size);
291 if (_base == NULL) {
292 vm_exit_during_initialization("Failed to allocate CM region mark stack");
293 }
294 _index = 0;
295 _capacity = (jint) size;
296 }
297
298 CMRegionStack::~CMRegionStack() {
299 if (_base != NULL) {
300 FREE_C_HEAP_ARRAY(oop, _base);
301 }
302 }
303
304 void CMRegionStack::push_lock_free(MemRegion mr) {
305 assert(mr.word_size() > 0, "Precondition");
306 while (true) {
307 jint index = _index;
308
309 if (index >= _capacity) {
310 _overflow = true;
311 return;
312 }
313 // Otherwise...
314 jint next_index = index+1;
315 jint res = Atomic::cmpxchg(next_index, &_index, index);
316 if (res == index) {
317 _base[index] = mr;
318 return;
319 }
320 // Otherwise, we need to try again.
321 }
322 }
323
324 // Lock-free pop of the region stack. Called during the concurrent
325 // marking / remark phases. Should only be called in tandem with
326 // other lock-free pops.
327 MemRegion CMRegionStack::pop_lock_free() {
328 while (true) {
329 jint index = _index;
330
331 if (index == 0) {
332 return MemRegion();
333 }
334 // Otherwise...
335 jint next_index = index-1;
336 jint res = Atomic::cmpxchg(next_index, &_index, index);
337 if (res == index) {
338 MemRegion mr = _base[next_index];
339 if (mr.start() != NULL) {
340 assert(mr.end() != NULL, "invariant");
341 assert(mr.word_size() > 0, "invariant");
342 return mr;
343 } else {
344 // that entry was invalidated... let's skip it
345 assert(mr.end() == NULL, "invariant");
346 }
347 }
348 // Otherwise, we need to try again.
349 }
350 }
351
352 #if 0
353 // The routines that manipulate the region stack with a lock are
354 // not currently used. They should be retained, however, as a
355 // diagnostic aid.
356
357 void CMRegionStack::push_with_lock(MemRegion mr) {
358 assert(mr.word_size() > 0, "Precondition");
359 MutexLockerEx x(CMRegionStack_lock, Mutex::_no_safepoint_check_flag);
360
361 if (isFull()) {
362 _overflow = true;
363 return;
364 }
365
366 _base[_index] = mr;
367 _index += 1;
368 }
369
370 MemRegion CMRegionStack::pop_with_lock() {
371 MutexLockerEx x(CMRegionStack_lock, Mutex::_no_safepoint_check_flag);
372
373 while (true) {
374 if (_index == 0) {
375 return MemRegion();
376 }
377 _index -= 1;
378
379 MemRegion mr = _base[_index];
380 if (mr.start() != NULL) {
381 assert(mr.end() != NULL, "invariant");
382 assert(mr.word_size() > 0, "invariant");
383 return mr;
384 } else {
385 // that entry was invalidated... let's skip it
386 assert(mr.end() == NULL, "invariant");
387 }
388 }
389 }
390 #endif
391
392 bool CMRegionStack::invalidate_entries_into_cset() {
393 bool result = false;
394 G1CollectedHeap* g1h = G1CollectedHeap::heap();
395 for (int i = 0; i < _oops_do_bound; ++i) {
396 MemRegion mr = _base[i];
397 if (mr.start() != NULL) {
398 assert(mr.end() != NULL, "invariant");
399 assert(mr.word_size() > 0, "invariant");
400 HeapRegion* hr = g1h->heap_region_containing(mr.start());
401 assert(hr != NULL, "invariant");
402 if (hr->in_collection_set()) {
403 // The region points into the collection set
404 _base[i] = MemRegion();
405 result = true;
406 }
407 } else {
408 // that entry was invalidated... let's skip it
409 assert(mr.end() == NULL, "invariant");
410 }
411 }
412 return result;
413 }
414
415 template<class OopClosureClass>
416 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) {
417 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after
418 || SafepointSynchronize::is_at_safepoint(),
419 "Drain recursion must be yield-safe.");
420 bool res = true;
421 debug_only(_drain_in_progress = true);
422 debug_only(_drain_in_progress_yields = yield_after);
423 while (!isEmpty()) {
424 oop newOop = pop();
425 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop");
426 assert(newOop->is_oop(), "Expected an oop");
427 assert(bm == NULL || bm->isMarked((HeapWord*)newOop),
428 "only grey objects on this stack");
429 // iterate over the oops in this oop, marking and pushing
430 // the ones in CMS generation.
431 newOop->oop_iterate(cl);
432 if (yield_after && _cm->do_yield_check()) {
433 res = false;
434 break;
435 }
436 }
437 debug_only(_drain_in_progress = false);
438 return res;
439 }
440
441 void CMMarkStack::oops_do(OopClosure* f) {
442 if (_index == 0) return;
443 assert(_oops_do_bound != -1 && _oops_do_bound <= _index,
444 "Bound must be set.");
445 for (int i = 0; i < _oops_do_bound; i++) {
446 f->do_oop(&_base[i]);
447 }
448 _oops_do_bound = -1;
449 }
450
451 bool ConcurrentMark::not_yet_marked(oop obj) const {
452 return (_g1h->is_obj_ill(obj)
453 || (_g1h->is_in_permanent(obj)
454 && !nextMarkBitMap()->isMarked((HeapWord*)obj)));
455 }
456
457 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away
458 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
459 #endif // _MSC_VER
460
461 size_t ConcurrentMark::scale_parallel_threads(size_t n_par_threads) {
462 return MAX2((n_par_threads + 2) / 4, (size_t)1);
463 }
464
465 ConcurrentMark::ConcurrentMark(ReservedSpace rs,
466 int max_regions) :
467 _markBitMap1(rs, MinObjAlignment - 1),
468 _markBitMap2(rs, MinObjAlignment - 1),
469
470 _parallel_marking_threads(0),
471 _max_parallel_marking_threads(0),
472 _sleep_factor(0.0),
473 _marking_task_overhead(1.0),
474 _cleanup_sleep_factor(0.0),
475 _cleanup_task_overhead(1.0),
476 _cleanup_list("Cleanup List"),
477 _region_bm(max_regions, false /* in_resource_area*/),
478 _card_bm((rs.size() + CardTableModRefBS::card_size - 1) >>
479 CardTableModRefBS::card_shift,
480 false /* in_resource_area*/),
481 _prevMarkBitMap(&_markBitMap1),
482 _nextMarkBitMap(&_markBitMap2),
483 _at_least_one_mark_complete(false),
484
485 _markStack(this),
486 _regionStack(),
487 // _finger set in set_non_marking_state
488
489 _max_task_num(MAX2(ParallelGCThreads, (size_t)1)),
490 // _active_tasks set in set_non_marking_state
491 // _tasks set inside the constructor
492 _task_queues(new CMTaskQueueSet((int) _max_task_num)),
493 _terminator(ParallelTaskTerminator((int) _max_task_num, _task_queues)),
494
495 _has_overflown(false),
496 _concurrent(false),
497 _has_aborted(false),
498 _restart_for_overflow(false),
499 _concurrent_marking_in_progress(false),
500 _should_gray_objects(false),
501
502 // _verbose_level set below
503
504 _init_times(),
505 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
506 _cleanup_times(),
507 _total_counting_time(0.0),
508 _total_rs_scrub_time(0.0),
509
510 _parallel_workers(NULL) {
511 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel;
512 if (verbose_level < no_verbose) {
513 verbose_level = no_verbose;
514 }
515 if (verbose_level > high_verbose) {
516 verbose_level = high_verbose;
517 }
518 _verbose_level = verbose_level;
519
520 if (verbose_low()) {
521 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", "
522 "heap end = "PTR_FORMAT, _heap_start, _heap_end);
523 }
524
525 _markStack.allocate(MarkStackSize);
526 _regionStack.allocate(G1MarkRegionStackSize);
527
528 // Create & start a ConcurrentMark thread.
529 _cmThread = new ConcurrentMarkThread(this);
530 assert(cmThread() != NULL, "CM Thread should have been created");
531 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
532
533 _g1h = G1CollectedHeap::heap();
534 assert(CGC_lock != NULL, "Where's the CGC_lock?");
535 assert(_markBitMap1.covers(rs), "_markBitMap1 inconsistency");
536 assert(_markBitMap2.covers(rs), "_markBitMap2 inconsistency");
537
538 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
539 satb_qs.set_buffer_size(G1SATBBufferSize);
540
541 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_task_num);
542 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_task_num);
543
544 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
545 _active_tasks = _max_task_num;
546 for (int i = 0; i < (int) _max_task_num; ++i) {
547 CMTaskQueue* task_queue = new CMTaskQueue();
548 task_queue->initialize();
549 _task_queues->register_queue(i, task_queue);
550
551 _tasks[i] = new CMTask(i, this, task_queue, _task_queues);
552 _accum_task_vtime[i] = 0.0;
553 }
554
555 if (ConcGCThreads > ParallelGCThreads) {
556 vm_exit_during_initialization("Can't have more ConcGCThreads "
557 "than ParallelGCThreads.");
558 }
559 if (ParallelGCThreads == 0) {
560 // if we are not running with any parallel GC threads we will not
561 // spawn any marking threads either
562 _parallel_marking_threads = 0;
563 _max_parallel_marking_threads = 0;
564 _sleep_factor = 0.0;
565 _marking_task_overhead = 1.0;
566 } else {
567 if (ConcGCThreads > 0) {
568 // notice that ConcGCThreads overwrites G1MarkingOverheadPercent
569 // if both are set
570
571 _parallel_marking_threads = ConcGCThreads;
572 _max_parallel_marking_threads = _parallel_marking_threads;
573 _sleep_factor = 0.0;
574 _marking_task_overhead = 1.0;
575 } else if (G1MarkingOverheadPercent > 0) {
576 // we will calculate the number of parallel marking threads
577 // based on a target overhead with respect to the soft real-time
578 // goal
579
580 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
581 double overall_cm_overhead =
582 (double) MaxGCPauseMillis * marking_overhead /
583 (double) GCPauseIntervalMillis;
584 double cpu_ratio = 1.0 / (double) os::processor_count();
585 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
586 double marking_task_overhead =
587 overall_cm_overhead / marking_thread_num *
588 (double) os::processor_count();
589 double sleep_factor =
590 (1.0 - marking_task_overhead) / marking_task_overhead;
591
592 _parallel_marking_threads = (size_t) marking_thread_num;
593 _max_parallel_marking_threads = _parallel_marking_threads;
594 _sleep_factor = sleep_factor;
595 _marking_task_overhead = marking_task_overhead;
596 } else {
597 _parallel_marking_threads = scale_parallel_threads(ParallelGCThreads);
598 _max_parallel_marking_threads = _parallel_marking_threads;
599 _sleep_factor = 0.0;
600 _marking_task_overhead = 1.0;
601 }
602
603 if (parallel_marking_threads() > 1) {
604 _cleanup_task_overhead = 1.0;
605 } else {
606 _cleanup_task_overhead = marking_task_overhead();
607 }
608 _cleanup_sleep_factor =
609 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead();
610
611 #if 0
612 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads());
613 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead());
614 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor());
615 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead());
616 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor());
617 #endif
618
619 guarantee(parallel_marking_threads() > 0, "peace of mind");
620 _parallel_workers = new FlexibleWorkGang("G1 Parallel Marking Threads",
621 (int) _max_parallel_marking_threads, false, true);
622 if (_parallel_workers == NULL) {
623 vm_exit_during_initialization("Failed necessary allocation.");
624 } else {
625 _parallel_workers->initialize_workers();
626 }
627 }
628
629 // so that the call below can read a sensible value
630 _heap_start = (HeapWord*) rs.base();
631 set_non_marking_state();
632 }
633
634 void ConcurrentMark::update_g1_committed(bool force) {
635 // If concurrent marking is not in progress, then we do not need to
636 // update _heap_end. This has a subtle and important
637 // side-effect. Imagine that two evacuation pauses happen between
638 // marking completion and remark. The first one can grow the
639 // heap (hence now the finger is below the heap end). Then, the
640 // second one could unnecessarily push regions on the region
641 // stack. This causes the invariant that the region stack is empty
642 // at the beginning of remark to be false. By ensuring that we do
643 // not observe heap expansions after marking is complete, then we do
644 // not have this problem.
645 if (!concurrent_marking_in_progress() && !force) return;
646
647 MemRegion committed = _g1h->g1_committed();
648 assert(committed.start() == _heap_start, "start shouldn't change");
649 HeapWord* new_end = committed.end();
650 if (new_end > _heap_end) {
651 // The heap has been expanded.
652
653 _heap_end = new_end;
654 }
655 // Notice that the heap can also shrink. However, this only happens
656 // during a Full GC (at least currently) and the entire marking
657 // phase will bail out and the task will not be restarted. So, let's
658 // do nothing.
659 }
660
661 void ConcurrentMark::reset() {
662 // Starting values for these two. This should be called in a STW
663 // phase. CM will be notified of any future g1_committed expansions
664 // will be at the end of evacuation pauses, when tasks are
665 // inactive.
666 MemRegion committed = _g1h->g1_committed();
667 _heap_start = committed.start();
668 _heap_end = committed.end();
669
670 // Separated the asserts so that we know which one fires.
671 assert(_heap_start != NULL, "heap bounds should look ok");
672 assert(_heap_end != NULL, "heap bounds should look ok");
673 assert(_heap_start < _heap_end, "heap bounds should look ok");
674
675 // reset all the marking data structures and any necessary flags
676 clear_marking_state();
677
678 if (verbose_low()) {
679 gclog_or_tty->print_cr("[global] resetting");
680 }
681
682 // We do reset all of them, since different phases will use
683 // different number of active threads. So, it's easiest to have all
684 // of them ready.
685 for (int i = 0; i < (int) _max_task_num; ++i) {
686 _tasks[i]->reset(_nextMarkBitMap);
687 }
688
689 // we need this to make sure that the flag is on during the evac
690 // pause with initial mark piggy-backed
691 set_concurrent_marking_in_progress();
692 }
693
694 void ConcurrentMark::set_phase(size_t active_tasks, bool concurrent) {
695 assert(active_tasks <= _max_task_num, "we should not have more");
696
697 _active_tasks = active_tasks;
698 // Need to update the three data structures below according to the
699 // number of active threads for this phase.
700 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
701 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
702 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
703
704 _concurrent = concurrent;
705 // We propagate this to all tasks, not just the active ones.
706 for (int i = 0; i < (int) _max_task_num; ++i)
707 _tasks[i]->set_concurrent(concurrent);
708
709 if (concurrent) {
710 set_concurrent_marking_in_progress();
711 } else {
712 // We currently assume that the concurrent flag has been set to
713 // false before we start remark. At this point we should also be
714 // in a STW phase.
715 assert(!concurrent_marking_in_progress(), "invariant");
716 assert(_finger == _heap_end, "only way to get here");
717 update_g1_committed(true);
718 }
719 }
720
721 void ConcurrentMark::set_non_marking_state() {
722 // We set the global marking state to some default values when we're
723 // not doing marking.
724 clear_marking_state();
725 _active_tasks = 0;
726 clear_concurrent_marking_in_progress();
727 }
728
729 ConcurrentMark::~ConcurrentMark() {
730 for (int i = 0; i < (int) _max_task_num; ++i) {
731 delete _task_queues->queue(i);
732 delete _tasks[i];
733 }
734 delete _task_queues;
735 FREE_C_HEAP_ARRAY(CMTask*, _max_task_num);
736 }
737
738 // This closure is used to mark refs into the g1 generation
739 // from external roots in the CMS bit map.
740 // Called at the first checkpoint.
741 //
742
743 void ConcurrentMark::clearNextBitmap() {
744 G1CollectedHeap* g1h = G1CollectedHeap::heap();
745 G1CollectorPolicy* g1p = g1h->g1_policy();
746
747 // Make sure that the concurrent mark thread looks to still be in
748 // the current cycle.
749 guarantee(cmThread()->during_cycle(), "invariant");
750
751 // We are finishing up the current cycle by clearing the next
752 // marking bitmap and getting it ready for the next cycle. During
753 // this time no other cycle can start. So, let's make sure that this
754 // is the case.
755 guarantee(!g1h->mark_in_progress(), "invariant");
756
757 // clear the mark bitmap (no grey objects to start with).
758 // We need to do this in chunks and offer to yield in between
759 // each chunk.
760 HeapWord* start = _nextMarkBitMap->startWord();
761 HeapWord* end = _nextMarkBitMap->endWord();
762 HeapWord* cur = start;
763 size_t chunkSize = M;
764 while (cur < end) {
765 HeapWord* next = cur + chunkSize;
766 if (next > end) {
767 next = end;
768 }
769 MemRegion mr(cur,next);
770 _nextMarkBitMap->clearRange(mr);
771 cur = next;
772 do_yield_check();
773
774 // Repeat the asserts from above. We'll do them as asserts here to
775 // minimize their overhead on the product. However, we'll have
776 // them as guarantees at the beginning / end of the bitmap
777 // clearing to get some checking in the product.
778 assert(cmThread()->during_cycle(), "invariant");
779 assert(!g1h->mark_in_progress(), "invariant");
780 }
781
782 // Repeat the asserts from above.
783 guarantee(cmThread()->during_cycle(), "invariant");
784 guarantee(!g1h->mark_in_progress(), "invariant");
785 }
786
787 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
788 public:
789 bool doHeapRegion(HeapRegion* r) {
790 if (!r->continuesHumongous()) {
791 r->note_start_of_marking(true);
792 }
793 return false;
794 }
795 };
796
797 void ConcurrentMark::checkpointRootsInitialPre() {
798 G1CollectedHeap* g1h = G1CollectedHeap::heap();
799 G1CollectorPolicy* g1p = g1h->g1_policy();
800
801 _has_aborted = false;
802
803 #ifndef PRODUCT
804 if (G1PrintReachableAtInitialMark) {
805 print_reachable("at-cycle-start",
806 VerifyOption_G1UsePrevMarking, true /* all */);
807 }
808 #endif
809
810 // Initialise marking structures. This has to be done in a STW phase.
811 reset();
812 }
813
814
815 void ConcurrentMark::checkpointRootsInitialPost() {
816 G1CollectedHeap* g1h = G1CollectedHeap::heap();
817
818 // If we force an overflow during remark, the remark operation will
819 // actually abort and we'll restart concurrent marking. If we always
820 // force an oveflow during remark we'll never actually complete the
821 // marking phase. So, we initilize this here, at the start of the
822 // cycle, so that at the remaining overflow number will decrease at
823 // every remark and we'll eventually not need to cause one.
824 force_overflow_stw()->init();
825
826 // For each region note start of marking.
827 NoteStartOfMarkHRClosure startcl;
828 g1h->heap_region_iterate(&startcl);
829
830 // Start Concurrent Marking weak-reference discovery.
831 ReferenceProcessor* rp = g1h->ref_processor_cm();
832 // enable ("weak") refs discovery
833 rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
834 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
835
836 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
837 // This is the start of the marking cycle, we're expected all
838 // threads to have SATB queues with active set to false.
839 satb_mq_set.set_active_all_threads(true, /* new active value */
840 false /* expected_active */);
841
842 // update_g1_committed() will be called at the end of an evac pause
843 // when marking is on. So, it's also called at the end of the
844 // initial-mark pause to update the heap end, if the heap expands
845 // during it. No need to call it here.
846 }
847
848 /*
849 * Notice that in the next two methods, we actually leave the STS
850 * during the barrier sync and join it immediately afterwards. If we
851 * do not do this, the following deadlock can occur: one thread could
852 * be in the barrier sync code, waiting for the other thread to also
853 * sync up, whereas another one could be trying to yield, while also
854 * waiting for the other threads to sync up too.
855 *
856 * Note, however, that this code is also used during remark and in
857 * this case we should not attempt to leave / enter the STS, otherwise
858 * we'll either hit an asseert (debug / fastdebug) or deadlock
859 * (product). So we should only leave / enter the STS if we are
860 * operating concurrently.
861 *
862 * Because the thread that does the sync barrier has left the STS, it
863 * is possible to be suspended for a Full GC or an evacuation pause
864 * could occur. This is actually safe, since the entering the sync
865 * barrier is one of the last things do_marking_step() does, and it
866 * doesn't manipulate any data structures afterwards.
867 */
868
869 void ConcurrentMark::enter_first_sync_barrier(int task_num) {
870 if (verbose_low()) {
871 gclog_or_tty->print_cr("[%d] entering first barrier", task_num);
872 }
873
874 if (concurrent()) {
875 ConcurrentGCThread::stsLeave();
876 }
877 _first_overflow_barrier_sync.enter();
878 if (concurrent()) {
879 ConcurrentGCThread::stsJoin();
880 }
881 // at this point everyone should have synced up and not be doing any
882 // more work
883
884 if (verbose_low()) {
885 gclog_or_tty->print_cr("[%d] leaving first barrier", task_num);
886 }
887
888 // let task 0 do this
889 if (task_num == 0) {
890 // task 0 is responsible for clearing the global data structures
891 // We should be here because of an overflow. During STW we should
892 // not clear the overflow flag since we rely on it being true when
893 // we exit this method to abort the pause and restart concurent
894 // marking.
895 clear_marking_state(concurrent() /* clear_overflow */);
896 force_overflow()->update();
897
898 if (PrintGC) {
899 gclog_or_tty->date_stamp(PrintGCDateStamps);
900 gclog_or_tty->stamp(PrintGCTimeStamps);
901 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]");
902 }
903 }
904
905 // after this, each task should reset its own data structures then
906 // then go into the second barrier
907 }
908
909 void ConcurrentMark::enter_second_sync_barrier(int task_num) {
910 if (verbose_low()) {
911 gclog_or_tty->print_cr("[%d] entering second barrier", task_num);
912 }
913
914 if (concurrent()) {
915 ConcurrentGCThread::stsLeave();
916 }
917 _second_overflow_barrier_sync.enter();
918 if (concurrent()) {
919 ConcurrentGCThread::stsJoin();
920 }
921 // at this point everything should be re-initialised and ready to go
922
923 if (verbose_low()) {
924 gclog_or_tty->print_cr("[%d] leaving second barrier", task_num);
925 }
926 }
927
928 #ifndef PRODUCT
929 void ForceOverflowSettings::init() {
930 _num_remaining = G1ConcMarkForceOverflow;
931 _force = false;
932 update();
933 }
934
935 void ForceOverflowSettings::update() {
936 if (_num_remaining > 0) {
937 _num_remaining -= 1;
938 _force = true;
939 } else {
940 _force = false;
941 }
942 }
943
944 bool ForceOverflowSettings::should_force() {
945 if (_force) {
946 _force = false;
947 return true;
948 } else {
949 return false;
950 }
951 }
952 #endif // !PRODUCT
953
954 void ConcurrentMark::grayRoot(oop p) {
955 HeapWord* addr = (HeapWord*) p;
956 // We can't really check against _heap_start and _heap_end, since it
957 // is possible during an evacuation pause with piggy-backed
958 // initial-mark that the committed space is expanded during the
959 // pause without CM observing this change. So the assertions below
960 // is a bit conservative; but better than nothing.
961 assert(_g1h->g1_committed().contains(addr),
962 "address should be within the heap bounds");
963
964 if (!_nextMarkBitMap->isMarked(addr)) {
965 _nextMarkBitMap->parMark(addr);
966 }
967 }
968
969 void ConcurrentMark::grayRegionIfNecessary(MemRegion mr) {
970 // The objects on the region have already been marked "in bulk" by
971 // the caller. We only need to decide whether to push the region on
972 // the region stack or not.
973
974 if (!concurrent_marking_in_progress() || !_should_gray_objects) {
975 // We're done with marking and waiting for remark. We do not need to
976 // push anything else on the region stack.
977 return;
978 }
979
980 HeapWord* finger = _finger;
981
982 if (verbose_low()) {
983 gclog_or_tty->print_cr("[global] attempting to push "
984 "region ["PTR_FORMAT", "PTR_FORMAT"), finger is at "
985 PTR_FORMAT, mr.start(), mr.end(), finger);
986 }
987
988 if (mr.start() < finger) {
989 // The finger is always heap region aligned and it is not possible
990 // for mr to span heap regions.
991 assert(mr.end() <= finger, "invariant");
992
993 // Separated the asserts so that we know which one fires.
994 assert(mr.start() <= mr.end(),
995 "region boundaries should fall within the committed space");
996 assert(_heap_start <= mr.start(),
997 "region boundaries should fall within the committed space");
998 assert(mr.end() <= _heap_end,
999 "region boundaries should fall within the committed space");
1000 if (verbose_low()) {
1001 gclog_or_tty->print_cr("[global] region ["PTR_FORMAT", "PTR_FORMAT") "
1002 "below the finger, pushing it",
1003 mr.start(), mr.end());
1004 }
1005
1006 if (!region_stack_push_lock_free(mr)) {
1007 if (verbose_low()) {
1008 gclog_or_tty->print_cr("[global] region stack has overflown.");
1009 }
1010 }
1011 }
1012 }
1013
1014 void ConcurrentMark::markAndGrayObjectIfNecessary(oop p) {
1015 // The object is not marked by the caller. We need to at least mark
1016 // it and maybe push in on the stack.
1017
1018 HeapWord* addr = (HeapWord*)p;
1019 if (!_nextMarkBitMap->isMarked(addr)) {
1020 // We definitely need to mark it, irrespective whether we bail out
1021 // because we're done with marking.
1022 if (_nextMarkBitMap->parMark(addr)) {
1023 if (!concurrent_marking_in_progress() || !_should_gray_objects) {
1024 // If we're done with concurrent marking and we're waiting for
1025 // remark, then we're not pushing anything on the stack.
1026 return;
1027 }
1028
1029 // No OrderAccess:store_load() is needed. It is implicit in the
1030 // CAS done in parMark(addr) above
1031 HeapWord* finger = _finger;
1032
1033 if (addr < finger) {
1034 if (!mark_stack_push(oop(addr))) {
1035 if (verbose_low()) {
1036 gclog_or_tty->print_cr("[global] global stack overflow "
1037 "during parMark");
1038 }
1039 }
1040 }
1041 }
1042 }
1043 }
1044
1045 class CMConcurrentMarkingTask: public AbstractGangTask {
1046 private:
1047 ConcurrentMark* _cm;
1048 ConcurrentMarkThread* _cmt;
1049
1050 public:
1051 void work(int worker_i) {
1052 assert(Thread::current()->is_ConcurrentGC_thread(),
1053 "this should only be done by a conc GC thread");
1054 ResourceMark rm;
1055
1056 double start_vtime = os::elapsedVTime();
1057
1058 ConcurrentGCThread::stsJoin();
1059
1060 assert((size_t) worker_i < _cm->active_tasks(), "invariant");
1061 CMTask* the_task = _cm->task(worker_i);
1062 the_task->record_start_time();
1063 if (!_cm->has_aborted()) {
1064 do {
1065 double start_vtime_sec = os::elapsedVTime();
1066 double start_time_sec = os::elapsedTime();
1067 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1068
1069 the_task->do_marking_step(mark_step_duration_ms,
1070 true /* do_stealing */,
1071 true /* do_termination */);
1072
1073 double end_time_sec = os::elapsedTime();
1074 double end_vtime_sec = os::elapsedVTime();
1075 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
1076 double elapsed_time_sec = end_time_sec - start_time_sec;
1077 _cm->clear_has_overflown();
1078
1079 bool ret = _cm->do_yield_check(worker_i);
1080
1081 jlong sleep_time_ms;
1082 if (!_cm->has_aborted() && the_task->has_aborted()) {
1083 sleep_time_ms =
1084 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
1085 ConcurrentGCThread::stsLeave();
1086 os::sleep(Thread::current(), sleep_time_ms, false);
1087 ConcurrentGCThread::stsJoin();
1088 }
1089 double end_time2_sec = os::elapsedTime();
1090 double elapsed_time2_sec = end_time2_sec - start_time_sec;
1091
1092 #if 0
1093 gclog_or_tty->print_cr("CM: elapsed %1.4lf ms, sleep %1.4lf ms, "
1094 "overhead %1.4lf",
1095 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms,
1096 the_task->conc_overhead(os::elapsedTime()) * 8.0);
1097 gclog_or_tty->print_cr("elapsed time %1.4lf ms, time 2: %1.4lf ms",
1098 elapsed_time_sec * 1000.0, elapsed_time2_sec * 1000.0);
1099 #endif
1100 } while (!_cm->has_aborted() && the_task->has_aborted());
1101 }
1102 the_task->record_end_time();
1103 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
1104
1105 ConcurrentGCThread::stsLeave();
1106
1107 double end_vtime = os::elapsedVTime();
1108 _cm->update_accum_task_vtime(worker_i, end_vtime - start_vtime);
1109 }
1110
1111 CMConcurrentMarkingTask(ConcurrentMark* cm,
1112 ConcurrentMarkThread* cmt) :
1113 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
1114
1115 ~CMConcurrentMarkingTask() { }
1116 };
1117
1118 // Calculates the number of active workers for a concurrent
1119 // phase.
1120 size_t ConcurrentMark::calc_parallel_marking_threads() {
1121 if (G1CollectedHeap::use_parallel_gc_threads()) {
1122 size_t n_conc_workers = 0;
1123 if (!UseDynamicNumberOfGCThreads ||
1124 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
1125 !ForceDynamicNumberOfGCThreads)) {
1126 n_conc_workers = max_parallel_marking_threads();
1127 } else {
1128 n_conc_workers =
1129 AdaptiveSizePolicy::calc_default_active_workers(
1130 max_parallel_marking_threads(),
1131 1, /* Minimum workers */
1132 parallel_marking_threads(),
1133 Threads::number_of_non_daemon_threads());
1134 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
1135 // that scaling has already gone into "_max_parallel_marking_threads".
1136 }
1137 assert(n_conc_workers > 0, "Always need at least 1");
1138 return n_conc_workers;
1139 }
1140 // If we are not running with any parallel GC threads we will not
1141 // have spawned any marking threads either. Hence the number of
1142 // concurrent workers should be 0.
1143 return 0;
1144 }
1145
1146 void ConcurrentMark::markFromRoots() {
1147 // we might be tempted to assert that:
1148 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1149 // "inconsistent argument?");
1150 // However that wouldn't be right, because it's possible that
1151 // a safepoint is indeed in progress as a younger generation
1152 // stop-the-world GC happens even as we mark in this generation.
1153
1154 _restart_for_overflow = false;
1155 force_overflow_conc()->init();
1156
1157 // _g1h has _n_par_threads
1158 _parallel_marking_threads = calc_parallel_marking_threads();
1159 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1160 "Maximum number of marking threads exceeded");
1161
1162 size_t active_workers = MAX2((size_t) 1, parallel_marking_threads());
1163
1164 // Parallel task terminator is set in "set_phase()"
1165 set_phase(active_workers, true /* concurrent */);
1166
1167 CMConcurrentMarkingTask markingTask(this, cmThread());
1168 if (parallel_marking_threads() > 0) {
1169 _parallel_workers->set_active_workers((int)active_workers);
1170 // Don't set _n_par_threads because it affects MT in proceess_strong_roots()
1171 // and the decisions on that MT processing is made elsewhere.
1172 assert(_parallel_workers->active_workers() > 0, "Should have been set");
1173 _parallel_workers->run_task(&markingTask);
1174 } else {
1175 markingTask.work(0);
1176 }
1177 print_stats();
1178 }
1179
1180 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1181 // world is stopped at this checkpoint
1182 assert(SafepointSynchronize::is_at_safepoint(),
1183 "world should be stopped");
1184
1185 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1186
1187 // If a full collection has happened, we shouldn't do this.
1188 if (has_aborted()) {
1189 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1190 return;
1191 }
1192
1193 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1194
1195 if (VerifyDuringGC) {
1196 HandleMark hm; // handle scope
1197 gclog_or_tty->print(" VerifyDuringGC:(before)");
1198 Universe::heap()->prepare_for_verify();
1199 Universe::verify(/* allow dirty */ true,
1200 /* silent */ false,
1201 /* option */ VerifyOption_G1UsePrevMarking);
1202 }
1203
1204 G1CollectorPolicy* g1p = g1h->g1_policy();
1205 g1p->record_concurrent_mark_remark_start();
1206
1207 double start = os::elapsedTime();
1208
1209 checkpointRootsFinalWork();
1210
1211 double mark_work_end = os::elapsedTime();
1212
1213 weakRefsWork(clear_all_soft_refs);
1214
1215 if (has_overflown()) {
1216 // Oops. We overflowed. Restart concurrent marking.
1217 _restart_for_overflow = true;
1218 // Clear the flag. We do not need it any more.
1219 clear_has_overflown();
1220 if (G1TraceMarkStackOverflow) {
1221 gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
1222 }
1223 } else {
1224 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1225 // We're done with marking.
1226 // This is the end of the marking cycle, we're expected all
1227 // threads to have SATB queues with active set to true.
1228 satb_mq_set.set_active_all_threads(false, /* new active value */
1229 true /* expected_active */);
1230
1231 if (VerifyDuringGC) {
1232
1233 HandleMark hm; // handle scope
1234 gclog_or_tty->print(" VerifyDuringGC:(after)");
1235 Universe::heap()->prepare_for_verify();
1236 Universe::verify(/* allow dirty */ true,
1237 /* silent */ false,
1238 /* option */ VerifyOption_G1UseNextMarking);
1239 }
1240 assert(!restart_for_overflow(), "sanity");
1241 }
1242
1243 // Reset the marking state if marking completed
1244 if (!restart_for_overflow()) {
1245 set_non_marking_state();
1246 }
1247
1248 #if VERIFY_OBJS_PROCESSED
1249 _scan_obj_cl.objs_processed = 0;
1250 ThreadLocalObjQueue::objs_enqueued = 0;
1251 #endif
1252
1253 // Statistics
1254 double now = os::elapsedTime();
1255 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1256 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1257 _remark_times.add((now - start) * 1000.0);
1258
1259 g1p->record_concurrent_mark_remark_end();
1260 }
1261
1262 #define CARD_BM_TEST_MODE 0
1263
1264 class CalcLiveObjectsClosure: public HeapRegionClosure {
1265
1266 CMBitMapRO* _bm;
1267 ConcurrentMark* _cm;
1268 bool _changed;
1269 bool _yield;
1270 size_t _words_done;
1271 size_t _tot_live;
1272 size_t _tot_used;
1273 size_t _regions_done;
1274 double _start_vtime_sec;
1275
1276 BitMap* _region_bm;
1277 BitMap* _card_bm;
1278 intptr_t _bottom_card_num;
1279 bool _final;
1280
1281 void mark_card_num_range(intptr_t start_card_num, intptr_t last_card_num) {
1282 for (intptr_t i = start_card_num; i <= last_card_num; i++) {
1283 #if CARD_BM_TEST_MODE
1284 guarantee(_card_bm->at(i - _bottom_card_num), "Should already be set.");
1285 #else
1286 _card_bm->par_at_put(i - _bottom_card_num, 1);
1287 #endif
1288 }
1289 }
1290
1291 public:
1292 CalcLiveObjectsClosure(bool final,
1293 CMBitMapRO *bm, ConcurrentMark *cm,
1294 BitMap* region_bm, BitMap* card_bm) :
1295 _bm(bm), _cm(cm), _changed(false), _yield(true),
1296 _words_done(0), _tot_live(0), _tot_used(0),
1297 _region_bm(region_bm), _card_bm(card_bm),_final(final),
1298 _regions_done(0), _start_vtime_sec(0.0)
1299 {
1300 _bottom_card_num =
1301 intptr_t(uintptr_t(G1CollectedHeap::heap()->reserved_region().start()) >>
1302 CardTableModRefBS::card_shift);
1303 }
1304
1305 // It takes a region that's not empty (i.e., it has at least one
1306 // live object in it and sets its corresponding bit on the region
1307 // bitmap to 1. If the region is "starts humongous" it will also set
1308 // to 1 the bits on the region bitmap that correspond to its
1309 // associated "continues humongous" regions.
1310 void set_bit_for_region(HeapRegion* hr) {
1311 assert(!hr->continuesHumongous(), "should have filtered those out");
1312
1313 size_t index = hr->hrs_index();
1314 if (!hr->startsHumongous()) {
1315 // Normal (non-humongous) case: just set the bit.
1316 _region_bm->par_at_put((BitMap::idx_t) index, true);
1317 } else {
1318 // Starts humongous case: calculate how many regions are part of
1319 // this humongous region and then set the bit range. It might
1320 // have been a bit more efficient to look at the object that
1321 // spans these humongous regions to calculate their number from
1322 // the object's size. However, it's a good idea to calculate
1323 // this based on the metadata itself, and not the region
1324 // contents, so that this code is not aware of what goes into
1325 // the humongous regions (in case this changes in the future).
1326 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1327 size_t end_index = index + 1;
1328 while (end_index < g1h->n_regions()) {
1329 HeapRegion* chr = g1h->region_at(end_index);
1330 if (!chr->continuesHumongous()) break;
1331 end_index += 1;
1332 }
1333 _region_bm->par_at_put_range((BitMap::idx_t) index,
1334 (BitMap::idx_t) end_index, true);
1335 }
1336 }
1337
1338 bool doHeapRegion(HeapRegion* hr) {
1339 if (!_final && _regions_done == 0) {
1340 _start_vtime_sec = os::elapsedVTime();
1341 }
1342
1343 if (hr->continuesHumongous()) {
1344 // We will ignore these here and process them when their
1345 // associated "starts humongous" region is processed (see
1346 // set_bit_for_heap_region()). Note that we cannot rely on their
1347 // associated "starts humongous" region to have their bit set to
1348 // 1 since, due to the region chunking in the parallel region
1349 // iteration, a "continues humongous" region might be visited
1350 // before its associated "starts humongous".
1351 return false;
1352 }
1353
1354 HeapWord* nextTop = hr->next_top_at_mark_start();
1355 HeapWord* start = hr->top_at_conc_mark_count();
1356 assert(hr->bottom() <= start && start <= hr->end() &&
1357 hr->bottom() <= nextTop && nextTop <= hr->end() &&
1358 start <= nextTop,
1359 "Preconditions.");
1360 // Otherwise, record the number of word's we'll examine.
1361 size_t words_done = (nextTop - start);
1362 // Find the first marked object at or after "start".
1363 start = _bm->getNextMarkedWordAddress(start, nextTop);
1364 size_t marked_bytes = 0;
1365
1366 // Below, the term "card num" means the result of shifting an address
1367 // by the card shift -- address 0 corresponds to card number 0. One
1368 // must subtract the card num of the bottom of the heap to obtain a
1369 // card table index.
1370 // The first card num of the sequence of live cards currently being
1371 // constructed. -1 ==> no sequence.
1372 intptr_t start_card_num = -1;
1373 // The last card num of the sequence of live cards currently being
1374 // constructed. -1 ==> no sequence.
1375 intptr_t last_card_num = -1;
1376
1377 while (start < nextTop) {
1378 if (_yield && _cm->do_yield_check()) {
1379 // We yielded. It might be for a full collection, in which case
1380 // all bets are off; terminate the traversal.
1381 if (_cm->has_aborted()) {
1382 _changed = false;
1383 return true;
1384 } else {
1385 // Otherwise, it might be a collection pause, and the region
1386 // we're looking at might be in the collection set. We'll
1387 // abandon this region.
1388 return false;
1389 }
1390 }
1391 oop obj = oop(start);
1392 int obj_sz = obj->size();
1393 // The card num of the start of the current object.
1394 intptr_t obj_card_num =
1395 intptr_t(uintptr_t(start) >> CardTableModRefBS::card_shift);
1396
1397 HeapWord* obj_last = start + obj_sz - 1;
1398 intptr_t obj_last_card_num =
1399 intptr_t(uintptr_t(obj_last) >> CardTableModRefBS::card_shift);
1400
1401 if (obj_card_num != last_card_num) {
1402 if (start_card_num == -1) {
1403 assert(last_card_num == -1, "Both or neither.");
1404 start_card_num = obj_card_num;
1405 } else {
1406 assert(last_card_num != -1, "Both or neither.");
1407 assert(obj_card_num >= last_card_num, "Inv");
1408 if ((obj_card_num - last_card_num) > 1) {
1409 // Mark the last run, and start a new one.
1410 mark_card_num_range(start_card_num, last_card_num);
1411 start_card_num = obj_card_num;
1412 }
1413 }
1414 #if CARD_BM_TEST_MODE
1415 /*
1416 gclog_or_tty->print_cr("Setting bits from %d/%d.",
1417 obj_card_num - _bottom_card_num,
1418 obj_last_card_num - _bottom_card_num);
1419 */
1420 for (intptr_t j = obj_card_num; j <= obj_last_card_num; j++) {
1421 _card_bm->par_at_put(j - _bottom_card_num, 1);
1422 }
1423 #endif
1424 }
1425 // In any case, we set the last card num.
1426 last_card_num = obj_last_card_num;
1427
1428 marked_bytes += (size_t)obj_sz * HeapWordSize;
1429 // Find the next marked object after this one.
1430 start = _bm->getNextMarkedWordAddress(start + 1, nextTop);
1431 _changed = true;
1432 }
1433 // Handle the last range, if any.
1434 if (start_card_num != -1) {
1435 mark_card_num_range(start_card_num, last_card_num);
1436 }
1437 if (_final) {
1438 // Mark the allocated-since-marking portion...
1439 HeapWord* tp = hr->top();
1440 if (nextTop < tp) {
1441 start_card_num =
1442 intptr_t(uintptr_t(nextTop) >> CardTableModRefBS::card_shift);
1443 last_card_num =
1444 intptr_t(uintptr_t(tp) >> CardTableModRefBS::card_shift);
1445 mark_card_num_range(start_card_num, last_card_num);
1446 // This definitely means the region has live objects.
1447 set_bit_for_region(hr);
1448 }
1449 }
1450
1451 hr->add_to_marked_bytes(marked_bytes);
1452 // Update the live region bitmap.
1453 if (marked_bytes > 0) {
1454 set_bit_for_region(hr);
1455 }
1456 hr->set_top_at_conc_mark_count(nextTop);
1457 _tot_live += hr->next_live_bytes();
1458 _tot_used += hr->used();
1459 _words_done = words_done;
1460
1461 if (!_final) {
1462 ++_regions_done;
1463 if (_regions_done % 10 == 0) {
1464 double end_vtime_sec = os::elapsedVTime();
1465 double elapsed_vtime_sec = end_vtime_sec - _start_vtime_sec;
1466 if (elapsed_vtime_sec > (10.0 / 1000.0)) {
1467 jlong sleep_time_ms =
1468 (jlong) (elapsed_vtime_sec * _cm->cleanup_sleep_factor() * 1000.0);
1469 os::sleep(Thread::current(), sleep_time_ms, false);
1470 _start_vtime_sec = end_vtime_sec;
1471 }
1472 }
1473 }
1474
1475 return false;
1476 }
1477
1478 bool changed() { return _changed; }
1479 void reset() { _changed = false; _words_done = 0; }
1480 void no_yield() { _yield = false; }
1481 size_t words_done() { return _words_done; }
1482 size_t tot_live() { return _tot_live; }
1483 size_t tot_used() { return _tot_used; }
1484 };
1485
1486
1487 void ConcurrentMark::calcDesiredRegions() {
1488 _region_bm.clear();
1489 _card_bm.clear();
1490 CalcLiveObjectsClosure calccl(false /*final*/,
1491 nextMarkBitMap(), this,
1492 &_region_bm, &_card_bm);
1493 G1CollectedHeap *g1h = G1CollectedHeap::heap();
1494 g1h->heap_region_iterate(&calccl);
1495
1496 do {
1497 calccl.reset();
1498 g1h->heap_region_iterate(&calccl);
1499 } while (calccl.changed());
1500 }
1501
1502 class G1ParFinalCountTask: public AbstractGangTask {
1503 protected:
1504 G1CollectedHeap* _g1h;
1505 CMBitMap* _bm;
1506 size_t _n_workers;
1507 size_t *_live_bytes;
1508 size_t *_used_bytes;
1509 BitMap* _region_bm;
1510 BitMap* _card_bm;
1511 public:
1512 G1ParFinalCountTask(G1CollectedHeap* g1h, CMBitMap* bm,
1513 BitMap* region_bm, BitMap* card_bm)
1514 : AbstractGangTask("G1 final counting"), _g1h(g1h),
1515 _bm(bm), _region_bm(region_bm), _card_bm(card_bm),
1516 _n_workers(0)
1517 {
1518 // Use the value already set as the number of active threads
1519 // in the call to run_task(). Needed for the allocation of
1520 // _live_bytes and _used_bytes.
1521 if (G1CollectedHeap::use_parallel_gc_threads()) {
1522 assert( _g1h->workers()->active_workers() > 0,
1523 "Should have been previously set");
1524 _n_workers = _g1h->workers()->active_workers();
1525 } else {
1526 _n_workers = 1;
1527 }
1528
1529 _live_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers);
1530 _used_bytes = NEW_C_HEAP_ARRAY(size_t, _n_workers);
1531 }
1532
1533 ~G1ParFinalCountTask() {
1534 FREE_C_HEAP_ARRAY(size_t, _live_bytes);
1535 FREE_C_HEAP_ARRAY(size_t, _used_bytes);
1536 }
1537
1538 void work(int i) {
1539 CalcLiveObjectsClosure calccl(true /*final*/,
1540 _bm, _g1h->concurrent_mark(),
1541 _region_bm, _card_bm);
1542 calccl.no_yield();
1543 if (G1CollectedHeap::use_parallel_gc_threads()) {
1544 _g1h->heap_region_par_iterate_chunked(&calccl, i,
1545 (int) _n_workers,
1546 HeapRegion::FinalCountClaimValue);
1547 } else {
1548 _g1h->heap_region_iterate(&calccl);
1549 }
1550 assert(calccl.complete(), "Shouldn't have yielded!");
1551
1552 assert((size_t) i < _n_workers, "invariant");
1553 _live_bytes[i] = calccl.tot_live();
1554 _used_bytes[i] = calccl.tot_used();
1555 }
1556 size_t live_bytes() {
1557 size_t live_bytes = 0;
1558 for (size_t i = 0; i < _n_workers; ++i)
1559 live_bytes += _live_bytes[i];
1560 return live_bytes;
1561 }
1562 size_t used_bytes() {
1563 size_t used_bytes = 0;
1564 for (size_t i = 0; i < _n_workers; ++i)
1565 used_bytes += _used_bytes[i];
1566 return used_bytes;
1567 }
1568 };
1569
1570 class G1ParNoteEndTask;
1571
1572 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1573 G1CollectedHeap* _g1;
1574 int _worker_num;
1575 size_t _max_live_bytes;
1576 size_t _regions_claimed;
1577 size_t _freed_bytes;
1578 FreeRegionList* _local_cleanup_list;
1579 OldRegionSet* _old_proxy_set;
1580 HumongousRegionSet* _humongous_proxy_set;
1581 HRRSCleanupTask* _hrrs_cleanup_task;
1582 double _claimed_region_time;
1583 double _max_region_time;
1584
1585 public:
1586 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1587 int worker_num,
1588 FreeRegionList* local_cleanup_list,
1589 OldRegionSet* old_proxy_set,
1590 HumongousRegionSet* humongous_proxy_set,
1591 HRRSCleanupTask* hrrs_cleanup_task) :
1592 _g1(g1), _worker_num(worker_num),
1593 _max_live_bytes(0), _regions_claimed(0),
1594 _freed_bytes(0),
1595 _claimed_region_time(0.0), _max_region_time(0.0),
1596 _local_cleanup_list(local_cleanup_list),
1597 _old_proxy_set(old_proxy_set),
1598 _humongous_proxy_set(humongous_proxy_set),
1599 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1600
1601 size_t freed_bytes() { return _freed_bytes; }
1602
1603 bool doHeapRegion(HeapRegion *hr) {
1604 // We use a claim value of zero here because all regions
1605 // were claimed with value 1 in the FinalCount task.
1606 hr->reset_gc_time_stamp();
1607 if (!hr->continuesHumongous()) {
1608 double start = os::elapsedTime();
1609 _regions_claimed++;
1610 hr->note_end_of_marking();
1611 _max_live_bytes += hr->max_live_bytes();
1612 _g1->free_region_if_empty(hr,
1613 &_freed_bytes,
1614 _local_cleanup_list,
1615 _old_proxy_set,
1616 _humongous_proxy_set,
1617 _hrrs_cleanup_task,
1618 true /* par */);
1619 double region_time = (os::elapsedTime() - start);
1620 _claimed_region_time += region_time;
1621 if (region_time > _max_region_time) {
1622 _max_region_time = region_time;
1623 }
1624 }
1625 return false;
1626 }
1627
1628 size_t max_live_bytes() { return _max_live_bytes; }
1629 size_t regions_claimed() { return _regions_claimed; }
1630 double claimed_region_time_sec() { return _claimed_region_time; }
1631 double max_region_time_sec() { return _max_region_time; }
1632 };
1633
1634 class G1ParNoteEndTask: public AbstractGangTask {
1635 friend class G1NoteEndOfConcMarkClosure;
1636
1637 protected:
1638 G1CollectedHeap* _g1h;
1639 size_t _max_live_bytes;
1640 size_t _freed_bytes;
1641 FreeRegionList* _cleanup_list;
1642
1643 public:
1644 G1ParNoteEndTask(G1CollectedHeap* g1h,
1645 FreeRegionList* cleanup_list) :
1646 AbstractGangTask("G1 note end"), _g1h(g1h),
1647 _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list) { }
1648
1649 void work(int i) {
1650 double start = os::elapsedTime();
1651 FreeRegionList local_cleanup_list("Local Cleanup List");
1652 OldRegionSet old_proxy_set("Local Cleanup Old Proxy Set");
1653 HumongousRegionSet humongous_proxy_set("Local Cleanup Humongous Proxy Set");
1654 HRRSCleanupTask hrrs_cleanup_task;
1655 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, i, &local_cleanup_list,
1656 &old_proxy_set,
1657 &humongous_proxy_set,
1658 &hrrs_cleanup_task);
1659 if (G1CollectedHeap::use_parallel_gc_threads()) {
1660 _g1h->heap_region_par_iterate_chunked(&g1_note_end, i,
1661 _g1h->workers()->active_workers(),
1662 HeapRegion::NoteEndClaimValue);
1663 } else {
1664 _g1h->heap_region_iterate(&g1_note_end);
1665 }
1666 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1667
1668 // Now update the lists
1669 _g1h->update_sets_after_freeing_regions(g1_note_end.freed_bytes(),
1670 NULL /* free_list */,
1671 &old_proxy_set,
1672 &humongous_proxy_set,
1673 true /* par */);
1674 {
1675 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1676 _max_live_bytes += g1_note_end.max_live_bytes();
1677 _freed_bytes += g1_note_end.freed_bytes();
1678
1679 // If we iterate over the global cleanup list at the end of
1680 // cleanup to do this printing we will not guarantee to only
1681 // generate output for the newly-reclaimed regions (the list
1682 // might not be empty at the beginning of cleanup; we might
1683 // still be working on its previous contents). So we do the
1684 // printing here, before we append the new regions to the global
1685 // cleanup list.
1686
1687 G1HRPrinter* hr_printer = _g1h->hr_printer();
1688 if (hr_printer->is_active()) {
1689 HeapRegionLinkedListIterator iter(&local_cleanup_list);
1690 while (iter.more_available()) {
1691 HeapRegion* hr = iter.get_next();
1692 hr_printer->cleanup(hr);
1693 }
1694 }
1695
1696 _cleanup_list->add_as_tail(&local_cleanup_list);
1697 assert(local_cleanup_list.is_empty(), "post-condition");
1698
1699 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1700 }
1701 double end = os::elapsedTime();
1702 if (G1PrintParCleanupStats) {
1703 gclog_or_tty->print(" Worker thread %d [%8.3f..%8.3f = %8.3f ms] "
1704 "claimed %d regions (tot = %8.3f ms, max = %8.3f ms).\n",
1705 i, start, end, (end-start)*1000.0,
1706 g1_note_end.regions_claimed(),
1707 g1_note_end.claimed_region_time_sec()*1000.0,
1708 g1_note_end.max_region_time_sec()*1000.0);
1709 }
1710 }
1711 size_t max_live_bytes() { return _max_live_bytes; }
1712 size_t freed_bytes() { return _freed_bytes; }
1713 };
1714
1715 class G1ParScrubRemSetTask: public AbstractGangTask {
1716 protected:
1717 G1RemSet* _g1rs;
1718 BitMap* _region_bm;
1719 BitMap* _card_bm;
1720 public:
1721 G1ParScrubRemSetTask(G1CollectedHeap* g1h,
1722 BitMap* region_bm, BitMap* card_bm) :
1723 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()),
1724 _region_bm(region_bm), _card_bm(card_bm)
1725 {}
1726
1727 void work(int i) {
1728 if (G1CollectedHeap::use_parallel_gc_threads()) {
1729 _g1rs->scrub_par(_region_bm, _card_bm, i,
1730 HeapRegion::ScrubRemSetClaimValue);
1731 } else {
1732 _g1rs->scrub(_region_bm, _card_bm);
1733 }
1734 }
1735
1736 };
1737
1738 void ConcurrentMark::cleanup() {
1739 // world is stopped at this checkpoint
1740 assert(SafepointSynchronize::is_at_safepoint(),
1741 "world should be stopped");
1742 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1743
1744 // If a full collection has happened, we shouldn't do this.
1745 if (has_aborted()) {
1746 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1747 return;
1748 }
1749
1750 HRSPhaseSetter x(HRSPhaseCleanup);
1751 g1h->verify_region_sets_optional();
1752
1753 if (VerifyDuringGC) {
1754 HandleMark hm; // handle scope
1755 gclog_or_tty->print(" VerifyDuringGC:(before)");
1756 Universe::heap()->prepare_for_verify();
1757 Universe::verify(/* allow dirty */ true,
1758 /* silent */ false,
1759 /* option */ VerifyOption_G1UsePrevMarking);
1760 }
1761
1762 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy();
1763 g1p->record_concurrent_mark_cleanup_start();
1764
1765 double start = os::elapsedTime();
1766
1767 HeapRegionRemSet::reset_for_cleanup_tasks();
1768
1769 size_t n_workers;
1770
1771 // Do counting once more with the world stopped for good measure.
1772 G1ParFinalCountTask g1_par_count_task(g1h, nextMarkBitMap(),
1773 &_region_bm, &_card_bm);
1774 if (G1CollectedHeap::use_parallel_gc_threads()) {
1775 assert(g1h->check_heap_region_claim_values(
1776 HeapRegion::InitialClaimValue),
1777 "sanity check");
1778
1779 g1h->set_par_threads();
1780 n_workers = g1h->n_par_threads();
1781 assert(g1h->n_par_threads() == (int) n_workers,
1782 "Should not have been reset");
1783 g1h->workers()->run_task(&g1_par_count_task);
1784 // Done with the parallel phase so reset to 0.
1785 g1h->set_par_threads(0);
1786
1787 assert(g1h->check_heap_region_claim_values(
1788 HeapRegion::FinalCountClaimValue),
1789 "sanity check");
1790 } else {
1791 n_workers = 1;
1792 g1_par_count_task.work(0);
1793 }
1794
1795 size_t known_garbage_bytes =
1796 g1_par_count_task.used_bytes() - g1_par_count_task.live_bytes();
1797 g1p->set_known_garbage_bytes(known_garbage_bytes);
1798
1799 size_t start_used_bytes = g1h->used();
1800 _at_least_one_mark_complete = true;
1801 g1h->set_marking_complete();
1802
1803 ergo_verbose4(ErgoConcCycles,
1804 "finish cleanup",
1805 ergo_format_byte("occupancy")
1806 ergo_format_byte("capacity")
1807 ergo_format_byte_perc("known garbage"),
1808 start_used_bytes, g1h->capacity(),
1809 known_garbage_bytes,
1810 ((double) known_garbage_bytes / (double) g1h->capacity()) * 100.0);
1811
1812 double count_end = os::elapsedTime();
1813 double this_final_counting_time = (count_end - start);
1814 if (G1PrintParCleanupStats) {
1815 gclog_or_tty->print_cr("Cleanup:");
1816 gclog_or_tty->print_cr(" Finalize counting: %8.3f ms",
1817 this_final_counting_time*1000.0);
1818 }
1819 _total_counting_time += this_final_counting_time;
1820
1821 if (G1PrintRegionLivenessInfo) {
1822 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking");
1823 _g1h->heap_region_iterate(&cl);
1824 }
1825
1826 // Install newly created mark bitMap as "prev".
1827 swapMarkBitMaps();
1828
1829 g1h->reset_gc_time_stamp();
1830
1831 // Note end of marking in all heap regions.
1832 double note_end_start = os::elapsedTime();
1833 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list);
1834 if (G1CollectedHeap::use_parallel_gc_threads()) {
1835 g1h->set_par_threads((int)n_workers);
1836 g1h->workers()->run_task(&g1_par_note_end_task);
1837 g1h->set_par_threads(0);
1838
1839 assert(g1h->check_heap_region_claim_values(HeapRegion::NoteEndClaimValue),
1840 "sanity check");
1841 } else {
1842 g1_par_note_end_task.work(0);
1843 }
1844
1845 if (!cleanup_list_is_empty()) {
1846 // The cleanup list is not empty, so we'll have to process it
1847 // concurrently. Notify anyone else that might be wanting free
1848 // regions that there will be more free regions coming soon.
1849 g1h->set_free_regions_coming();
1850 }
1851 double note_end_end = os::elapsedTime();
1852 if (G1PrintParCleanupStats) {
1853 gclog_or_tty->print_cr(" note end of marking: %8.3f ms.",
1854 (note_end_end - note_end_start)*1000.0);
1855 }
1856
1857 // call below, since it affects the metric by which we sort the heap
1858 // regions.
1859 if (G1ScrubRemSets) {
1860 double rs_scrub_start = os::elapsedTime();
1861 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm);
1862 if (G1CollectedHeap::use_parallel_gc_threads()) {
1863 g1h->set_par_threads((int)n_workers);
1864 g1h->workers()->run_task(&g1_par_scrub_rs_task);
1865 g1h->set_par_threads(0);
1866
1867 assert(g1h->check_heap_region_claim_values(
1868 HeapRegion::ScrubRemSetClaimValue),
1869 "sanity check");
1870 } else {
1871 g1_par_scrub_rs_task.work(0);
1872 }
1873
1874 double rs_scrub_end = os::elapsedTime();
1875 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start);
1876 _total_rs_scrub_time += this_rs_scrub_time;
1877 }
1878
1879 // this will also free any regions totally full of garbage objects,
1880 // and sort the regions.
1881 g1h->g1_policy()->record_concurrent_mark_cleanup_end((int)n_workers);
1882
1883 // Statistics.
1884 double end = os::elapsedTime();
1885 _cleanup_times.add((end - start) * 1000.0);
1886
1887 // G1CollectedHeap::heap()->print();
1888 // gclog_or_tty->print_cr("HEAP GC TIME STAMP : %d",
1889 // G1CollectedHeap::heap()->get_gc_time_stamp());
1890
1891 if (PrintGC || PrintGCDetails) {
1892 g1h->print_size_transition(gclog_or_tty,
1893 start_used_bytes,
1894 g1h->used(),
1895 g1h->capacity());
1896 }
1897
1898 size_t cleaned_up_bytes = start_used_bytes - g1h->used();
1899 g1p->decrease_known_garbage_bytes(cleaned_up_bytes);
1900
1901 // Clean up will have freed any regions completely full of garbage.
1902 // Update the soft reference policy with the new heap occupancy.
1903 Universe::update_heap_info_at_gc();
1904
1905 // We need to make this be a "collection" so any collection pause that
1906 // races with it goes around and waits for completeCleanup to finish.
1907 g1h->increment_total_collections();
1908
1909 if (VerifyDuringGC) {
1910 HandleMark hm; // handle scope
1911 gclog_or_tty->print(" VerifyDuringGC:(after)");
1912 Universe::heap()->prepare_for_verify();
1913 Universe::verify(/* allow dirty */ true,
1914 /* silent */ false,
1915 /* option */ VerifyOption_G1UsePrevMarking);
1916 }
1917
1918 g1h->verify_region_sets_optional();
1919 }
1920
1921 void ConcurrentMark::completeCleanup() {
1922 if (has_aborted()) return;
1923
1924 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1925
1926 _cleanup_list.verify_optional();
1927 FreeRegionList tmp_free_list("Tmp Free List");
1928
1929 if (G1ConcRegionFreeingVerbose) {
1930 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
1931 "cleanup list has "SIZE_FORMAT" entries",
1932 _cleanup_list.length());
1933 }
1934
1935 // Noone else should be accessing the _cleanup_list at this point,
1936 // so it's not necessary to take any locks
1937 while (!_cleanup_list.is_empty()) {
1938 HeapRegion* hr = _cleanup_list.remove_head();
1939 assert(hr != NULL, "the list was not empty");
1940 hr->par_clear();
1941 tmp_free_list.add_as_tail(hr);
1942
1943 // Instead of adding one region at a time to the secondary_free_list,
1944 // we accumulate them in the local list and move them a few at a
1945 // time. This also cuts down on the number of notify_all() calls
1946 // we do during this process. We'll also append the local list when
1947 // _cleanup_list is empty (which means we just removed the last
1948 // region from the _cleanup_list).
1949 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
1950 _cleanup_list.is_empty()) {
1951 if (G1ConcRegionFreeingVerbose) {
1952 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
1953 "appending "SIZE_FORMAT" entries to the "
1954 "secondary_free_list, clean list still has "
1955 SIZE_FORMAT" entries",
1956 tmp_free_list.length(),
1957 _cleanup_list.length());
1958 }
1959
1960 {
1961 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
1962 g1h->secondary_free_list_add_as_tail(&tmp_free_list);
1963 SecondaryFreeList_lock->notify_all();
1964 }
1965
1966 if (G1StressConcRegionFreeing) {
1967 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
1968 os::sleep(Thread::current(), (jlong) 1, false);
1969 }
1970 }
1971 }
1972 }
1973 assert(tmp_free_list.is_empty(), "post-condition");
1974 }
1975
1976 // Support closures for reference procssing in G1
1977
1978 bool G1CMIsAliveClosure::do_object_b(oop obj) {
1979 HeapWord* addr = (HeapWord*)obj;
1980 return addr != NULL &&
1981 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
1982 }
1983
1984 class G1CMKeepAliveClosure: public OopClosure {
1985 G1CollectedHeap* _g1;
1986 ConcurrentMark* _cm;
1987 CMBitMap* _bitMap;
1988 public:
1989 G1CMKeepAliveClosure(G1CollectedHeap* g1, ConcurrentMark* cm,
1990 CMBitMap* bitMap) :
1991 _g1(g1), _cm(cm),
1992 _bitMap(bitMap) {}
1993
1994 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
1995 virtual void do_oop( oop* p) { do_oop_work(p); }
1996
1997 template <class T> void do_oop_work(T* p) {
1998 oop obj = oopDesc::load_decode_heap_oop(p);
1999 HeapWord* addr = (HeapWord*)obj;
2000
2001 if (_cm->verbose_high()) {
2002 gclog_or_tty->print_cr("\t[0] we're looking at location "
2003 "*"PTR_FORMAT" = "PTR_FORMAT,
2004 p, (void*) obj);
2005 }
2006
2007 if (_g1->is_in_g1_reserved(addr) && _g1->is_obj_ill(obj)) {
2008 _bitMap->mark(addr);
2009 _cm->mark_stack_push(obj);
2010 }
2011 }
2012 };
2013
2014 class G1CMDrainMarkingStackClosure: public VoidClosure {
2015 CMMarkStack* _markStack;
2016 CMBitMap* _bitMap;
2017 G1CMKeepAliveClosure* _oopClosure;
2018 public:
2019 G1CMDrainMarkingStackClosure(CMBitMap* bitMap, CMMarkStack* markStack,
2020 G1CMKeepAliveClosure* oopClosure) :
2021 _bitMap(bitMap),
2022 _markStack(markStack),
2023 _oopClosure(oopClosure)
2024 {}
2025
2026 void do_void() {
2027 _markStack->drain((OopClosure*)_oopClosure, _bitMap, false);
2028 }
2029 };
2030
2031 // 'Keep Alive' closure used by parallel reference processing.
2032 // An instance of this closure is used in the parallel reference processing
2033 // code rather than an instance of G1CMKeepAliveClosure. We could have used
2034 // the G1CMKeepAliveClosure as it is MT-safe. Also reference objects are
2035 // placed on to discovered ref lists once so we can mark and push with no
2036 // need to check whether the object has already been marked. Using the
2037 // G1CMKeepAliveClosure would mean, however, having all the worker threads
2038 // operating on the global mark stack. This means that an individual
2039 // worker would be doing lock-free pushes while it processes its own
2040 // discovered ref list followed by drain call. If the discovered ref lists
2041 // are unbalanced then this could cause interference with the other
2042 // workers. Using a CMTask (and its embedded local data structures)
2043 // avoids that potential interference.
2044 class G1CMParKeepAliveAndDrainClosure: public OopClosure {
2045 ConcurrentMark* _cm;
2046 CMTask* _task;
2047 int _ref_counter_limit;
2048 int _ref_counter;
2049 public:
2050 G1CMParKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task) :
2051 _cm(cm), _task(task),
2052 _ref_counter_limit(G1RefProcDrainInterval) {
2053 assert(_ref_counter_limit > 0, "sanity");
2054 _ref_counter = _ref_counter_limit;
2055 }
2056
2057 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
2058 virtual void do_oop( oop* p) { do_oop_work(p); }
2059
2060 template <class T> void do_oop_work(T* p) {
2061 if (!_cm->has_overflown()) {
2062 oop obj = oopDesc::load_decode_heap_oop(p);
2063 if (_cm->verbose_high()) {
2064 gclog_or_tty->print_cr("\t[%d] we're looking at location "
2065 "*"PTR_FORMAT" = "PTR_FORMAT,
2066 _task->task_id(), p, (void*) obj);
2067 }
2068
2069 _task->deal_with_reference(obj);
2070 _ref_counter--;
2071
2072 if (_ref_counter == 0) {
2073 // We have dealt with _ref_counter_limit references, pushing them and objects
2074 // reachable from them on to the local stack (and possibly the global stack).
2075 // Call do_marking_step() to process these entries. We call the routine in a
2076 // loop, which we'll exit if there's nothing more to do (i.e. we're done
2077 // with the entries that we've pushed as a result of the deal_with_reference
2078 // calls above) or we overflow.
2079 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() flag
2080 // while there may still be some work to do. (See the comment at the
2081 // beginning of CMTask::do_marking_step() for those conditions - one of which
2082 // is reaching the specified time target.) It is only when
2083 // CMTask::do_marking_step() returns without setting the has_aborted() flag
2084 // that the marking has completed.
2085 do {
2086 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
2087 _task->do_marking_step(mark_step_duration_ms,
2088 false /* do_stealing */,
2089 false /* do_termination */);
2090 } while (_task->has_aborted() && !_cm->has_overflown());
2091 _ref_counter = _ref_counter_limit;
2092 }
2093 } else {
2094 if (_cm->verbose_high()) {
2095 gclog_or_tty->print_cr("\t[%d] CM Overflow", _task->task_id());
2096 }
2097 }
2098 }
2099 };
2100
2101 class G1CMParDrainMarkingStackClosure: public VoidClosure {
2102 ConcurrentMark* _cm;
2103 CMTask* _task;
2104 public:
2105 G1CMParDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task) :
2106 _cm(cm), _task(task)
2107 {}
2108
2109 void do_void() {
2110 do {
2111 if (_cm->verbose_high()) {
2112 gclog_or_tty->print_cr("\t[%d] Drain: Calling do marking_step",
2113 _task->task_id());
2114 }
2115
2116 // We call CMTask::do_marking_step() to completely drain the local and
2117 // global marking stacks. The routine is called in a loop, which we'll
2118 // exit if there's nothing more to do (i.e. we'completely drained the
2119 // entries that were pushed as a result of applying the
2120 // G1CMParKeepAliveAndDrainClosure to the entries on the discovered ref
2121 // lists above) or we overflow the global marking stack.
2122 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() flag
2123 // while there may still be some work to do. (See the comment at the
2124 // beginning of CMTask::do_marking_step() for those conditions - one of which
2125 // is reaching the specified time target.) It is only when
2126 // CMTask::do_marking_step() returns without setting the has_aborted() flag
2127 // that the marking has completed.
2128
2129 _task->do_marking_step(1000000000.0 /* something very large */,
2130 true /* do_stealing */,
2131 true /* do_termination */);
2132 } while (_task->has_aborted() && !_cm->has_overflown());
2133 }
2134 };
2135
2136 // Implementation of AbstractRefProcTaskExecutor for parallel
2137 // reference processing at the end of G1 concurrent marking
2138
2139 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
2140 private:
2141 G1CollectedHeap* _g1h;
2142 ConcurrentMark* _cm;
2143 WorkGang* _workers;
2144 int _active_workers;
2145
2146 public:
2147 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
2148 ConcurrentMark* cm,
2149 WorkGang* workers,
2150 int n_workers) :
2151 _g1h(g1h), _cm(cm),
2152 _workers(workers), _active_workers(n_workers) { }
2153
2154 // Executes the given task using concurrent marking worker threads.
2155 virtual void execute(ProcessTask& task);
2156 virtual void execute(EnqueueTask& task);
2157 };
2158
2159 class G1CMRefProcTaskProxy: public AbstractGangTask {
2160 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
2161 ProcessTask& _proc_task;
2162 G1CollectedHeap* _g1h;
2163 ConcurrentMark* _cm;
2164
2165 public:
2166 G1CMRefProcTaskProxy(ProcessTask& proc_task,
2167 G1CollectedHeap* g1h,
2168 ConcurrentMark* cm) :
2169 AbstractGangTask("Process reference objects in parallel"),
2170 _proc_task(proc_task), _g1h(g1h), _cm(cm) { }
2171
2172 virtual void work(int i) {
2173 CMTask* marking_task = _cm->task(i);
2174 G1CMIsAliveClosure g1_is_alive(_g1h);
2175 G1CMParKeepAliveAndDrainClosure g1_par_keep_alive(_cm, marking_task);
2176 G1CMParDrainMarkingStackClosure g1_par_drain(_cm, marking_task);
2177
2178 _proc_task.work(i, g1_is_alive, g1_par_keep_alive, g1_par_drain);
2179 }
2180 };
2181
2182 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
2183 assert(_workers != NULL, "Need parallel worker threads.");
2184
2185 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
2186
2187 // We need to reset the phase for each task execution so that
2188 // the termination protocol of CMTask::do_marking_step works.
2189 _cm->set_phase(_active_workers, false /* concurrent */);
2190 _g1h->set_par_threads(_active_workers);
2191 _workers->run_task(&proc_task_proxy);
2192 _g1h->set_par_threads(0);
2193 }
2194
2195 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
2196 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
2197 EnqueueTask& _enq_task;
2198
2199 public:
2200 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
2201 AbstractGangTask("Enqueue reference objects in parallel"),
2202 _enq_task(enq_task) { }
2203
2204 virtual void work(int i) {
2205 _enq_task.work(i);
2206 }
2207 };
2208
2209 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2210 assert(_workers != NULL, "Need parallel worker threads.");
2211
2212 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2213
2214 _g1h->set_par_threads(_active_workers);
2215 _workers->run_task(&enq_task_proxy);
2216 _g1h->set_par_threads(0);
2217 }
2218
2219 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2220 ResourceMark rm;
2221 HandleMark hm;
2222
2223 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2224
2225 // Is alive closure.
2226 G1CMIsAliveClosure g1_is_alive(g1h);
2227
2228 // Inner scope to exclude the cleaning of the string and symbol
2229 // tables from the displayed time.
2230 {
2231 bool verbose = PrintGC && PrintGCDetails;
2232 if (verbose) {
2233 gclog_or_tty->put(' ');
2234 }
2235 TraceTime t("GC ref-proc", verbose, false, gclog_or_tty);
2236
2237 ReferenceProcessor* rp = g1h->ref_processor_cm();
2238
2239 // See the comment in G1CollectedHeap::ref_processing_init()
2240 // about how reference processing currently works in G1.
2241
2242 // Process weak references.
2243 rp->setup_policy(clear_all_soft_refs);
2244 assert(_markStack.isEmpty(), "mark stack should be empty");
2245
2246 G1CMKeepAliveClosure g1_keep_alive(g1h, this, nextMarkBitMap());
2247 G1CMDrainMarkingStackClosure
2248 g1_drain_mark_stack(nextMarkBitMap(), &_markStack, &g1_keep_alive);
2249
2250 // We use the work gang from the G1CollectedHeap and we utilize all
2251 // the worker threads.
2252 int active_workers = g1h->workers() ? g1h->workers()->active_workers() : 1;
2253 active_workers = MAX2(MIN2(active_workers, (int)_max_task_num), 1);
2254
2255 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2256 g1h->workers(), active_workers);
2257
2258 if (rp->processing_is_mt()) {
2259 // Set the degree of MT here. If the discovery is done MT, there
2260 // may have been a different number of threads doing the discovery
2261 // and a different number of discovered lists may have Ref objects.
2262 // That is OK as long as the Reference lists are balanced (see
2263 // balance_all_queues() and balance_queues()).
2264 rp->set_active_mt_degree(active_workers);
2265
2266 rp->process_discovered_references(&g1_is_alive,
2267 &g1_keep_alive,
2268 &g1_drain_mark_stack,
2269 &par_task_executor);
2270
2271 // The work routines of the parallel keep_alive and drain_marking_stack
2272 // will set the has_overflown flag if we overflow the global marking
2273 // stack.
2274 } else {
2275 rp->process_discovered_references(&g1_is_alive,
2276 &g1_keep_alive,
2277 &g1_drain_mark_stack,
2278 NULL);
2279 }
2280
2281 assert(_markStack.overflow() || _markStack.isEmpty(),
2282 "mark stack should be empty (unless it overflowed)");
2283 if (_markStack.overflow()) {
2284 // Should have been done already when we tried to push an
2285 // entry on to the global mark stack. But let's do it again.
2286 set_has_overflown();
2287 }
2288
2289 if (rp->processing_is_mt()) {
2290 assert(rp->num_q() == active_workers, "why not");
2291 rp->enqueue_discovered_references(&par_task_executor);
2292 } else {
2293 rp->enqueue_discovered_references();
2294 }
2295
2296 rp->verify_no_references_recorded();
2297 assert(!rp->discovery_enabled(), "Post condition");
2298 }
2299
2300 // Now clean up stale oops in StringTable
2301 StringTable::unlink(&g1_is_alive);
2302 // Clean up unreferenced symbols in symbol table.
2303 SymbolTable::unlink();
2304 }
2305
2306 void ConcurrentMark::swapMarkBitMaps() {
2307 CMBitMapRO* temp = _prevMarkBitMap;
2308 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap;
2309 _nextMarkBitMap = (CMBitMap*) temp;
2310 }
2311
2312 class CMRemarkTask: public AbstractGangTask {
2313 private:
2314 ConcurrentMark *_cm;
2315
2316 public:
2317 void work(int worker_i) {
2318 // Since all available tasks are actually started, we should
2319 // only proceed if we're supposed to be actived.
2320 if ((size_t)worker_i < _cm->active_tasks()) {
2321 CMTask* task = _cm->task(worker_i);
2322 task->record_start_time();
2323 do {
2324 task->do_marking_step(1000000000.0 /* something very large */,
2325 true /* do_stealing */,
2326 true /* do_termination */);
2327 } while (task->has_aborted() && !_cm->has_overflown());
2328 // If we overflow, then we do not want to restart. We instead
2329 // want to abort remark and do concurrent marking again.
2330 task->record_end_time();
2331 }
2332 }
2333
2334 CMRemarkTask(ConcurrentMark* cm, int active_workers) :
2335 AbstractGangTask("Par Remark"), _cm(cm) {
2336 _cm->terminator()->reset_for_reuse(active_workers);
2337 }
2338 };
2339
2340 void ConcurrentMark::checkpointRootsFinalWork() {
2341 ResourceMark rm;
2342 HandleMark hm;
2343 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2344
2345 g1h->ensure_parsability(false);
2346
2347 if (G1CollectedHeap::use_parallel_gc_threads()) {
2348 G1CollectedHeap::StrongRootsScope srs(g1h);
2349 // this is remark, so we'll use up all active threads
2350 int active_workers = g1h->workers()->active_workers();
2351 if (active_workers == 0) {
2352 assert(active_workers > 0, "Should have been set earlier");
2353 active_workers = ParallelGCThreads;
2354 g1h->workers()->set_active_workers(active_workers);
2355 }
2356 set_phase(active_workers, false /* concurrent */);
2357 // Leave _parallel_marking_threads at it's
2358 // value originally calculated in the ConcurrentMark
2359 // constructor and pass values of the active workers
2360 // through the gang in the task.
2361
2362 CMRemarkTask remarkTask(this, active_workers);
2363 g1h->set_par_threads(active_workers);
2364 g1h->workers()->run_task(&remarkTask);
2365 g1h->set_par_threads(0);
2366 } else {
2367 G1CollectedHeap::StrongRootsScope srs(g1h);
2368 // this is remark, so we'll use up all available threads
2369 int active_workers = 1;
2370 set_phase(active_workers, false /* concurrent */);
2371
2372 CMRemarkTask remarkTask(this, active_workers);
2373 // We will start all available threads, even if we decide that the
2374 // active_workers will be fewer. The extra ones will just bail out
2375 // immediately.
2376 remarkTask.work(0);
2377 }
2378 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2379 guarantee(satb_mq_set.completed_buffers_num() == 0, "invariant");
2380
2381 print_stats();
2382
2383 #if VERIFY_OBJS_PROCESSED
2384 if (_scan_obj_cl.objs_processed != ThreadLocalObjQueue::objs_enqueued) {
2385 gclog_or_tty->print_cr("Processed = %d, enqueued = %d.",
2386 _scan_obj_cl.objs_processed,
2387 ThreadLocalObjQueue::objs_enqueued);
2388 guarantee(_scan_obj_cl.objs_processed ==
2389 ThreadLocalObjQueue::objs_enqueued,
2390 "Different number of objs processed and enqueued.");
2391 }
2392 #endif
2393 }
2394
2395 #ifndef PRODUCT
2396
2397 class PrintReachableOopClosure: public OopClosure {
2398 private:
2399 G1CollectedHeap* _g1h;
2400 outputStream* _out;
2401 VerifyOption _vo;
2402 bool _all;
2403
2404 public:
2405 PrintReachableOopClosure(outputStream* out,
2406 VerifyOption vo,
2407 bool all) :
2408 _g1h(G1CollectedHeap::heap()),
2409 _out(out), _vo(vo), _all(all) { }
2410
2411 void do_oop(narrowOop* p) { do_oop_work(p); }
2412 void do_oop( oop* p) { do_oop_work(p); }
2413
2414 template <class T> void do_oop_work(T* p) {
2415 oop obj = oopDesc::load_decode_heap_oop(p);
2416 const char* str = NULL;
2417 const char* str2 = "";
2418
2419 if (obj == NULL) {
2420 str = "";
2421 } else if (!_g1h->is_in_g1_reserved(obj)) {
2422 str = " O";
2423 } else {
2424 HeapRegion* hr = _g1h->heap_region_containing(obj);
2425 guarantee(hr != NULL, "invariant");
2426 bool over_tams = false;
2427 bool marked = false;
2428
2429 switch (_vo) {
2430 case VerifyOption_G1UsePrevMarking:
2431 over_tams = hr->obj_allocated_since_prev_marking(obj);
2432 marked = _g1h->isMarkedPrev(obj);
2433 break;
2434 case VerifyOption_G1UseNextMarking:
2435 over_tams = hr->obj_allocated_since_next_marking(obj);
2436 marked = _g1h->isMarkedNext(obj);
2437 break;
2438 case VerifyOption_G1UseMarkWord:
2439 marked = obj->is_gc_marked();
2440 break;
2441 default:
2442 ShouldNotReachHere();
2443 }
2444
2445 if (over_tams) {
2446 str = " >";
2447 if (marked) {
2448 str2 = " AND MARKED";
2449 }
2450 } else if (marked) {
2451 str = " M";
2452 } else {
2453 str = " NOT";
2454 }
2455 }
2456
2457 _out->print_cr(" "PTR_FORMAT": "PTR_FORMAT"%s%s",
2458 p, (void*) obj, str, str2);
2459 }
2460 };
2461
2462 class PrintReachableObjectClosure : public ObjectClosure {
2463 private:
2464 G1CollectedHeap* _g1h;
2465 outputStream* _out;
2466 VerifyOption _vo;
2467 bool _all;
2468 HeapRegion* _hr;
2469
2470 public:
2471 PrintReachableObjectClosure(outputStream* out,
2472 VerifyOption vo,
2473 bool all,
2474 HeapRegion* hr) :
2475 _g1h(G1CollectedHeap::heap()),
2476 _out(out), _vo(vo), _all(all), _hr(hr) { }
2477
2478 void do_object(oop o) {
2479 bool over_tams = false;
2480 bool marked = false;
2481
2482 switch (_vo) {
2483 case VerifyOption_G1UsePrevMarking:
2484 over_tams = _hr->obj_allocated_since_prev_marking(o);
2485 marked = _g1h->isMarkedPrev(o);
2486 break;
2487 case VerifyOption_G1UseNextMarking:
2488 over_tams = _hr->obj_allocated_since_next_marking(o);
2489 marked = _g1h->isMarkedNext(o);
2490 break;
2491 case VerifyOption_G1UseMarkWord:
2492 marked = o->is_gc_marked();
2493 break;
2494 default:
2495 ShouldNotReachHere();
2496 }
2497 bool print_it = _all || over_tams || marked;
2498
2499 if (print_it) {
2500 _out->print_cr(" "PTR_FORMAT"%s",
2501 o, (over_tams) ? " >" : (marked) ? " M" : "");
2502 PrintReachableOopClosure oopCl(_out, _vo, _all);
2503 o->oop_iterate(&oopCl);
2504 }
2505 }
2506 };
2507
2508 class PrintReachableRegionClosure : public HeapRegionClosure {
2509 private:
2510 outputStream* _out;
2511 VerifyOption _vo;
2512 bool _all;
2513
2514 public:
2515 bool doHeapRegion(HeapRegion* hr) {
2516 HeapWord* b = hr->bottom();
2517 HeapWord* e = hr->end();
2518 HeapWord* t = hr->top();
2519 HeapWord* p = NULL;
2520
2521 switch (_vo) {
2522 case VerifyOption_G1UsePrevMarking:
2523 p = hr->prev_top_at_mark_start();
2524 break;
2525 case VerifyOption_G1UseNextMarking:
2526 p = hr->next_top_at_mark_start();
2527 break;
2528 case VerifyOption_G1UseMarkWord:
2529 // When we are verifying marking using the mark word
2530 // TAMS has no relevance.
2531 assert(p == NULL, "post-condition");
2532 break;
2533 default:
2534 ShouldNotReachHere();
2535 }
2536 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" "
2537 "TAMS: "PTR_FORMAT, b, e, t, p);
2538 _out->cr();
2539
2540 HeapWord* from = b;
2541 HeapWord* to = t;
2542
2543 if (to > from) {
2544 _out->print_cr("Objects in ["PTR_FORMAT", "PTR_FORMAT"]", from, to);
2545 _out->cr();
2546 PrintReachableObjectClosure ocl(_out, _vo, _all, hr);
2547 hr->object_iterate_mem_careful(MemRegion(from, to), &ocl);
2548 _out->cr();
2549 }
2550
2551 return false;
2552 }
2553
2554 PrintReachableRegionClosure(outputStream* out,
2555 VerifyOption vo,
2556 bool all) :
2557 _out(out), _vo(vo), _all(all) { }
2558 };
2559
2560 static const char* verify_option_to_tams(VerifyOption vo) {
2561 switch (vo) {
2562 case VerifyOption_G1UsePrevMarking:
2563 return "PTAMS";
2564 case VerifyOption_G1UseNextMarking:
2565 return "NTAMS";
2566 default:
2567 return "NONE";
2568 }
2569 }
2570
2571 void ConcurrentMark::print_reachable(const char* str,
2572 VerifyOption vo,
2573 bool all) {
2574 gclog_or_tty->cr();
2575 gclog_or_tty->print_cr("== Doing heap dump... ");
2576
2577 if (G1PrintReachableBaseFile == NULL) {
2578 gclog_or_tty->print_cr(" #### error: no base file defined");
2579 return;
2580 }
2581
2582 if (strlen(G1PrintReachableBaseFile) + 1 + strlen(str) >
2583 (JVM_MAXPATHLEN - 1)) {
2584 gclog_or_tty->print_cr(" #### error: file name too long");
2585 return;
2586 }
2587
2588 char file_name[JVM_MAXPATHLEN];
2589 sprintf(file_name, "%s.%s", G1PrintReachableBaseFile, str);
2590 gclog_or_tty->print_cr(" dumping to file %s", file_name);
2591
2592 fileStream fout(file_name);
2593 if (!fout.is_open()) {
2594 gclog_or_tty->print_cr(" #### error: could not open file");
2595 return;
2596 }
2597
2598 outputStream* out = &fout;
2599 out->print_cr("-- USING %s", verify_option_to_tams(vo));
2600 out->cr();
2601
2602 out->print_cr("--- ITERATING OVER REGIONS");
2603 out->cr();
2604 PrintReachableRegionClosure rcl(out, vo, all);
2605 _g1h->heap_region_iterate(&rcl);
2606 out->cr();
2607
2608 gclog_or_tty->print_cr(" done");
2609 gclog_or_tty->flush();
2610 }
2611
2612 #endif // PRODUCT
2613
2614 // This note is for drainAllSATBBuffers and the code in between.
2615 // In the future we could reuse a task to do this work during an
2616 // evacuation pause (since now tasks are not active and can be claimed
2617 // during an evacuation pause). This was a late change to the code and
2618 // is currently not being taken advantage of.
2619
2620 class CMGlobalObjectClosure : public ObjectClosure {
2621 private:
2622 ConcurrentMark* _cm;
2623
2624 public:
2625 void do_object(oop obj) {
2626 _cm->deal_with_reference(obj);
2627 }
2628
2629 CMGlobalObjectClosure(ConcurrentMark* cm) : _cm(cm) { }
2630 };
2631
2632 void ConcurrentMark::deal_with_reference(oop obj) {
2633 if (verbose_high()) {
2634 gclog_or_tty->print_cr("[global] we're dealing with reference "PTR_FORMAT,
2635 (void*) obj);
2636 }
2637
2638 HeapWord* objAddr = (HeapWord*) obj;
2639 assert(obj->is_oop_or_null(true /* ignore mark word */), "Error");
2640 if (_g1h->is_in_g1_reserved(objAddr)) {
2641 assert(obj != NULL, "null check is implicit");
2642 if (!_nextMarkBitMap->isMarked(objAddr)) {
2643 // Only get the containing region if the object is not marked on the
2644 // bitmap (otherwise, it's a waste of time since we won't do
2645 // anything with it).
2646 HeapRegion* hr = _g1h->heap_region_containing_raw(obj);
2647 if (!hr->obj_allocated_since_next_marking(obj)) {
2648 if (verbose_high()) {
2649 gclog_or_tty->print_cr("[global] "PTR_FORMAT" is not considered "
2650 "marked", (void*) obj);
2651 }
2652
2653 // we need to mark it first
2654 if (_nextMarkBitMap->parMark(objAddr)) {
2655 // No OrderAccess:store_load() is needed. It is implicit in the
2656 // CAS done in parMark(objAddr) above
2657 HeapWord* finger = _finger;
2658 if (objAddr < finger) {
2659 if (verbose_high()) {
2660 gclog_or_tty->print_cr("[global] below the global finger "
2661 "("PTR_FORMAT"), pushing it", finger);
2662 }
2663 if (!mark_stack_push(obj)) {
2664 if (verbose_low()) {
2665 gclog_or_tty->print_cr("[global] global stack overflow during "
2666 "deal_with_reference");
2667 }
2668 }
2669 }
2670 }
2671 }
2672 }
2673 }
2674 }
2675
2676 void ConcurrentMark::drainAllSATBBuffers() {
2677 CMGlobalObjectClosure oc(this);
2678 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2679 satb_mq_set.set_closure(&oc);
2680
2681 while (satb_mq_set.apply_closure_to_completed_buffer()) {
2682 if (verbose_medium()) {
2683 gclog_or_tty->print_cr("[global] processed an SATB buffer");
2684 }
2685 }
2686
2687 // no need to check whether we should do this, as this is only
2688 // called during an evacuation pause
2689 satb_mq_set.iterate_closure_all_threads();
2690
2691 satb_mq_set.set_closure(NULL);
2692 assert(satb_mq_set.completed_buffers_num() == 0, "invariant");
2693 }
2694
2695 void ConcurrentMark::markPrev(oop p) {
2696 // Note we are overriding the read-only view of the prev map here, via
2697 // the cast.
2698 ((CMBitMap*)_prevMarkBitMap)->mark((HeapWord*)p);
2699 }
2700
2701 void ConcurrentMark::clear(oop p) {
2702 assert(p != NULL && p->is_oop(), "expected an oop");
2703 HeapWord* addr = (HeapWord*)p;
2704 assert(addr >= _nextMarkBitMap->startWord() ||
2705 addr < _nextMarkBitMap->endWord(), "in a region");
2706
2707 _nextMarkBitMap->clear(addr);
2708 }
2709
2710 void ConcurrentMark::clearRangeBothMaps(MemRegion mr) {
2711 // Note we are overriding the read-only view of the prev map here, via
2712 // the cast.
2713 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2714 _nextMarkBitMap->clearRange(mr);
2715 }
2716
2717 HeapRegion*
2718 ConcurrentMark::claim_region(int task_num) {
2719 // "checkpoint" the finger
2720 HeapWord* finger = _finger;
2721
2722 // _heap_end will not change underneath our feet; it only changes at
2723 // yield points.
2724 while (finger < _heap_end) {
2725 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2726
2727 // Note on how this code handles humongous regions. In the
2728 // normal case the finger will reach the start of a "starts
2729 // humongous" (SH) region. Its end will either be the end of the
2730 // last "continues humongous" (CH) region in the sequence, or the
2731 // standard end of the SH region (if the SH is the only region in
2732 // the sequence). That way claim_region() will skip over the CH
2733 // regions. However, there is a subtle race between a CM thread
2734 // executing this method and a mutator thread doing a humongous
2735 // object allocation. The two are not mutually exclusive as the CM
2736 // thread does not need to hold the Heap_lock when it gets
2737 // here. So there is a chance that claim_region() will come across
2738 // a free region that's in the progress of becoming a SH or a CH
2739 // region. In the former case, it will either
2740 // a) Miss the update to the region's end, in which case it will
2741 // visit every subsequent CH region, will find their bitmaps
2742 // empty, and do nothing, or
2743 // b) Will observe the update of the region's end (in which case
2744 // it will skip the subsequent CH regions).
2745 // If it comes across a region that suddenly becomes CH, the
2746 // scenario will be similar to b). So, the race between
2747 // claim_region() and a humongous object allocation might force us
2748 // to do a bit of unnecessary work (due to some unnecessary bitmap
2749 // iterations) but it should not introduce and correctness issues.
2750 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger);
2751 HeapWord* bottom = curr_region->bottom();
2752 HeapWord* end = curr_region->end();
2753 HeapWord* limit = curr_region->next_top_at_mark_start();
2754
2755 if (verbose_low()) {
2756 gclog_or_tty->print_cr("[%d] curr_region = "PTR_FORMAT" "
2757 "["PTR_FORMAT", "PTR_FORMAT"), "
2758 "limit = "PTR_FORMAT,
2759 task_num, curr_region, bottom, end, limit);
2760 }
2761
2762 // Is the gap between reading the finger and doing the CAS too long?
2763 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2764 if (res == finger) {
2765 // we succeeded
2766
2767 // notice that _finger == end cannot be guaranteed here since,
2768 // someone else might have moved the finger even further
2769 assert(_finger >= end, "the finger should have moved forward");
2770
2771 if (verbose_low()) {
2772 gclog_or_tty->print_cr("[%d] we were successful with region = "
2773 PTR_FORMAT, task_num, curr_region);
2774 }
2775
2776 if (limit > bottom) {
2777 if (verbose_low()) {
2778 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is not empty, "
2779 "returning it ", task_num, curr_region);
2780 }
2781 return curr_region;
2782 } else {
2783 assert(limit == bottom,
2784 "the region limit should be at bottom");
2785 if (verbose_low()) {
2786 gclog_or_tty->print_cr("[%d] region "PTR_FORMAT" is empty, "
2787 "returning NULL", task_num, curr_region);
2788 }
2789 // we return NULL and the caller should try calling
2790 // claim_region() again.
2791 return NULL;
2792 }
2793 } else {
2794 assert(_finger > finger, "the finger should have moved forward");
2795 if (verbose_low()) {
2796 gclog_or_tty->print_cr("[%d] somebody else moved the finger, "
2797 "global finger = "PTR_FORMAT", "
2798 "our finger = "PTR_FORMAT,
2799 task_num, _finger, finger);
2800 }
2801
2802 // read it again
2803 finger = _finger;
2804 }
2805 }
2806
2807 return NULL;
2808 }
2809
2810 bool ConcurrentMark::invalidate_aborted_regions_in_cset() {
2811 bool result = false;
2812 for (int i = 0; i < (int)_max_task_num; ++i) {
2813 CMTask* the_task = _tasks[i];
2814 MemRegion mr = the_task->aborted_region();
2815 if (mr.start() != NULL) {
2816 assert(mr.end() != NULL, "invariant");
2817 assert(mr.word_size() > 0, "invariant");
2818 HeapRegion* hr = _g1h->heap_region_containing(mr.start());
2819 assert(hr != NULL, "invariant");
2820 if (hr->in_collection_set()) {
2821 // The region points into the collection set
2822 the_task->set_aborted_region(MemRegion());
2823 result = true;
2824 }
2825 }
2826 }
2827 return result;
2828 }
2829
2830 bool ConcurrentMark::has_aborted_regions() {
2831 for (int i = 0; i < (int)_max_task_num; ++i) {
2832 CMTask* the_task = _tasks[i];
2833 MemRegion mr = the_task->aborted_region();
2834 if (mr.start() != NULL) {
2835 assert(mr.end() != NULL, "invariant");
2836 assert(mr.word_size() > 0, "invariant");
2837 return true;
2838 }
2839 }
2840 return false;
2841 }
2842
2843 void ConcurrentMark::oops_do(OopClosure* cl) {
2844 if (_markStack.size() > 0 && verbose_low()) {
2845 gclog_or_tty->print_cr("[global] scanning the global marking stack, "
2846 "size = %d", _markStack.size());
2847 }
2848 // we first iterate over the contents of the mark stack...
2849 _markStack.oops_do(cl);
2850
2851 for (int i = 0; i < (int)_max_task_num; ++i) {
2852 OopTaskQueue* queue = _task_queues->queue((int)i);
2853
2854 if (queue->size() > 0 && verbose_low()) {
2855 gclog_or_tty->print_cr("[global] scanning task queue of task %d, "
2856 "size = %d", i, queue->size());
2857 }
2858
2859 // ...then over the contents of the all the task queues.
2860 queue->oops_do(cl);
2861 }
2862
2863 // Invalidate any entries, that are in the region stack, that
2864 // point into the collection set
2865 if (_regionStack.invalidate_entries_into_cset()) {
2866 // otherwise, any gray objects copied during the evacuation pause
2867 // might not be visited.
2868 assert(_should_gray_objects, "invariant");
2869 }
2870
2871 // Invalidate any aborted regions, recorded in the individual CM
2872 // tasks, that point into the collection set.
2873 if (invalidate_aborted_regions_in_cset()) {
2874 // otherwise, any gray objects copied during the evacuation pause
2875 // might not be visited.
2876 assert(_should_gray_objects, "invariant");
2877 }
2878
2879 }
2880
2881 void ConcurrentMark::clear_marking_state(bool clear_overflow) {
2882 _markStack.setEmpty();
2883 _markStack.clear_overflow();
2884 _regionStack.setEmpty();
2885 _regionStack.clear_overflow();
2886 if (clear_overflow) {
2887 clear_has_overflown();
2888 } else {
2889 assert(has_overflown(), "pre-condition");
2890 }
2891 _finger = _heap_start;
2892
2893 for (int i = 0; i < (int)_max_task_num; ++i) {
2894 OopTaskQueue* queue = _task_queues->queue(i);
2895 queue->set_empty();
2896 // Clear any partial regions from the CMTasks
2897 _tasks[i]->clear_aborted_region();
2898 }
2899 }
2900
2901 void ConcurrentMark::print_stats() {
2902 if (verbose_stats()) {
2903 gclog_or_tty->print_cr("---------------------------------------------------------------------");
2904 for (size_t i = 0; i < _active_tasks; ++i) {
2905 _tasks[i]->print_stats();
2906 gclog_or_tty->print_cr("---------------------------------------------------------------------");
2907 }
2908 }
2909 }
2910
2911 // Closures used by ConcurrentMark::complete_marking_in_collection_set().
2912
2913 class CSetMarkOopClosure: public OopClosure {
2914 friend class CSetMarkBitMapClosure;
2915
2916 G1CollectedHeap* _g1h;
2917 CMBitMap* _bm;
2918 ConcurrentMark* _cm;
2919 oop* _ms;
2920 jint* _array_ind_stack;
2921 int _ms_size;
2922 int _ms_ind;
2923 int _array_increment;
2924 int _worker_i;
2925
2926 bool push(oop obj, int arr_ind = 0) {
2927 if (_ms_ind == _ms_size) {
2928 gclog_or_tty->print_cr("Mark stack is full.");
2929 return false;
2930 }
2931 _ms[_ms_ind] = obj;
2932 if (obj->is_objArray()) {
2933 _array_ind_stack[_ms_ind] = arr_ind;
2934 }
2935 _ms_ind++;
2936 return true;
2937 }
2938
2939 oop pop() {
2940 if (_ms_ind == 0) {
2941 return NULL;
2942 } else {
2943 _ms_ind--;
2944 return _ms[_ms_ind];
2945 }
2946 }
2947
2948 template <class T> bool drain() {
2949 while (_ms_ind > 0) {
2950 oop obj = pop();
2951 assert(obj != NULL, "Since index was non-zero.");
2952 if (obj->is_objArray()) {
2953 jint arr_ind = _array_ind_stack[_ms_ind];
2954 objArrayOop aobj = objArrayOop(obj);
2955 jint len = aobj->length();
2956 jint next_arr_ind = arr_ind + _array_increment;
2957 if (next_arr_ind < len) {
2958 push(obj, next_arr_ind);
2959 }
2960 // Now process this portion of this one.
2961 int lim = MIN2(next_arr_ind, len);
2962 for (int j = arr_ind; j < lim; j++) {
2963 do_oop(aobj->objArrayOopDesc::obj_at_addr<T>(j));
2964 }
2965 } else {
2966 obj->oop_iterate(this);
2967 }
2968 if (abort()) return false;
2969 }
2970 return true;
2971 }
2972
2973 public:
2974 CSetMarkOopClosure(ConcurrentMark* cm, int ms_size, int worker_i) :
2975 _g1h(G1CollectedHeap::heap()),
2976 _cm(cm),
2977 _bm(cm->nextMarkBitMap()),
2978 _ms_size(ms_size), _ms_ind(0),
2979 _ms(NEW_C_HEAP_ARRAY(oop, ms_size)),
2980 _array_ind_stack(NEW_C_HEAP_ARRAY(jint, ms_size)),
2981 _array_increment(MAX2(ms_size/8, 16)),
2982 _worker_i(worker_i) { }
2983
2984 ~CSetMarkOopClosure() {
2985 FREE_C_HEAP_ARRAY(oop, _ms);
2986 FREE_C_HEAP_ARRAY(jint, _array_ind_stack);
2987 }
2988
2989 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
2990 virtual void do_oop( oop* p) { do_oop_work(p); }
2991
2992 template <class T> void do_oop_work(T* p) {
2993 T heap_oop = oopDesc::load_heap_oop(p);
2994 if (oopDesc::is_null(heap_oop)) return;
2995 oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
2996 if (obj->is_forwarded()) {
2997 // If the object has already been forwarded, we have to make sure
2998 // that it's marked. So follow the forwarding pointer. Note that
2999 // this does the right thing for self-forwarding pointers in the
3000 // evacuation failure case.
3001 obj = obj->forwardee();
3002 }
3003 HeapRegion* hr = _g1h->heap_region_containing(obj);
3004 if (hr != NULL) {
3005 if (hr->in_collection_set()) {
3006 if (_g1h->is_obj_ill(obj)) {
3007 if (_bm->parMark((HeapWord*)obj)) {
3008 if (!push(obj)) {
3009 gclog_or_tty->print_cr("Setting abort in CSetMarkOopClosure because push failed.");
3010 set_abort();
3011 }
3012 }
3013 }
3014 } else {
3015 // Outside the collection set; we need to gray it
3016 _cm->deal_with_reference(obj);
3017 }
3018 }
3019 }
3020 };
3021
3022 class CSetMarkBitMapClosure: public BitMapClosure {
3023 G1CollectedHeap* _g1h;
3024 CMBitMap* _bitMap;
3025 ConcurrentMark* _cm;
3026 CSetMarkOopClosure _oop_cl;
3027 int _worker_i;
3028
3029 public:
3030 CSetMarkBitMapClosure(ConcurrentMark* cm, int ms_size, int worker_i) :
3031 _g1h(G1CollectedHeap::heap()),
3032 _bitMap(cm->nextMarkBitMap()),
3033 _oop_cl(cm, ms_size, worker_i),
3034 _worker_i(worker_i) { }
3035
3036 bool do_bit(size_t offset) {
3037 // convert offset into a HeapWord*
3038 HeapWord* addr = _bitMap->offsetToHeapWord(offset);
3039 assert(_bitMap->endWord() && addr < _bitMap->endWord(),
3040 "address out of range");
3041 assert(_bitMap->isMarked(addr), "tautology");
3042 oop obj = oop(addr);
3043 if (!obj->is_forwarded()) {
3044 if (!_oop_cl.push(obj)) return false;
3045 if (UseCompressedOops) {
3046 if (!_oop_cl.drain<narrowOop>()) return false;
3047 } else {
3048 if (!_oop_cl.drain<oop>()) return false;
3049 }
3050 }
3051 // Otherwise...
3052 return true;
3053 }
3054 };
3055
3056 class CompleteMarkingInCSetHRClosure: public HeapRegionClosure {
3057 CMBitMap* _bm;
3058 CSetMarkBitMapClosure _bit_cl;
3059 int _worker_i;
3060
3061 enum SomePrivateConstants {
3062 MSSize = 1000
3063 };
3064
3065 public:
3066 CompleteMarkingInCSetHRClosure(ConcurrentMark* cm, int worker_i) :
3067 _bm(cm->nextMarkBitMap()),
3068 _bit_cl(cm, MSSize, worker_i),
3069 _worker_i(worker_i) { }
3070
3071 bool doHeapRegion(HeapRegion* hr) {
3072 if (hr->claimHeapRegion(HeapRegion::CompleteMarkCSetClaimValue)) {
3073 // The current worker has successfully claimed the region.
3074 if (!hr->evacuation_failed()) {
3075 MemRegion mr = MemRegion(hr->bottom(), hr->next_top_at_mark_start());
3076 if (!mr.is_empty()) {
3077 bool done = false;
3078 while (!done) {
3079 done = _bm->iterate(&_bit_cl, mr);
3080 }
3081 }
3082 }
3083 }
3084 return false;
3085 }
3086 };
3087
3088 class SetClaimValuesInCSetHRClosure: public HeapRegionClosure {
3089 jint _claim_value;
3090
3091 public:
3092 SetClaimValuesInCSetHRClosure(jint claim_value) :
3093 _claim_value(claim_value) { }
3094
3095 bool doHeapRegion(HeapRegion* hr) {
3096 hr->set_claim_value(_claim_value);
3097 return false;
3098 }
3099 };
3100
3101 class G1ParCompleteMarkInCSetTask: public AbstractGangTask {
3102 protected:
3103 G1CollectedHeap* _g1h;
3104 ConcurrentMark* _cm;
3105
3106 public:
3107 G1ParCompleteMarkInCSetTask(G1CollectedHeap* g1h,
3108 ConcurrentMark* cm) :
3109 AbstractGangTask("Complete Mark in CSet"),
3110 _g1h(g1h), _cm(cm) { }
3111
3112 void work(int worker_i) {
3113 CompleteMarkingInCSetHRClosure cmplt(_cm, worker_i);
3114 HeapRegion* hr = _g1h->start_cset_region_for_worker(worker_i);
3115 _g1h->collection_set_iterate_from(hr, &cmplt);
3116 }
3117 };
3118
3119 void ConcurrentMark::complete_marking_in_collection_set() {
3120 G1CollectedHeap* g1h = G1CollectedHeap::heap();
3121
3122 if (!g1h->mark_in_progress()) {
3123 g1h->g1_policy()->record_mark_closure_time(0.0);
3124 return;
3125 }
3126
3127 double start = os::elapsedTime();
3128 G1ParCompleteMarkInCSetTask complete_mark_task(g1h, this);
3129
3130 assert(g1h->check_cset_heap_region_claim_values(HeapRegion::InitialClaimValue), "sanity");
3131
3132 if (G1CollectedHeap::use_parallel_gc_threads()) {
3133 int n_workers = g1h->workers()->active_workers();
3134 g1h->set_par_threads(n_workers);
3135 g1h->workers()->run_task(&complete_mark_task);
3136 g1h->set_par_threads(0);
3137 } else {
3138 complete_mark_task.work(0);
3139 }
3140
3141 assert(g1h->check_cset_heap_region_claim_values(HeapRegion::CompleteMarkCSetClaimValue), "sanity");
3142
3143 // Now reset the claim values in the regions in the collection set.
3144 SetClaimValuesInCSetHRClosure set_cv_cl(HeapRegion::InitialClaimValue);
3145 g1h->collection_set_iterate(&set_cv_cl);
3146
3147 assert(g1h->check_cset_heap_region_claim_values(HeapRegion::InitialClaimValue), "sanity");
3148
3149 double end_time = os::elapsedTime();
3150 double elapsed_time_ms = (end_time - start) * 1000.0;
3151 g1h->g1_policy()->record_mark_closure_time(elapsed_time_ms);
3152 }
3153
3154 // The next two methods deal with the following optimisation. Some
3155 // objects are gray by being marked and located above the finger. If
3156 // they are copied, during an evacuation pause, below the finger then
3157 // the need to be pushed on the stack. The observation is that, if
3158 // there are no regions in the collection set located above the
3159 // finger, then the above cannot happen, hence we do not need to
3160 // explicitly gray any objects when copying them to below the
3161 // finger. The global stack will be scanned to ensure that, if it
3162 // points to objects being copied, it will update their
3163 // location. There is a tricky situation with the gray objects in
3164 // region stack that are being coped, however. See the comment in
3165 // newCSet().
3166
3167 void ConcurrentMark::newCSet() {
3168 if (!concurrent_marking_in_progress()) {
3169 // nothing to do if marking is not in progress
3170 return;
3171 }
3172
3173 // find what the lowest finger is among the global and local fingers
3174 _min_finger = _finger;
3175 for (int i = 0; i < (int)_max_task_num; ++i) {
3176 CMTask* task = _tasks[i];
3177 HeapWord* task_finger = task->finger();
3178 if (task_finger != NULL && task_finger < _min_finger) {
3179 _min_finger = task_finger;
3180 }
3181 }
3182
3183 _should_gray_objects = false;
3184
3185 // This fixes a very subtle and fustrating bug. It might be the case
3186 // that, during en evacuation pause, heap regions that contain
3187 // objects that are gray (by being in regions contained in the
3188 // region stack) are included in the collection set. Since such gray
3189 // objects will be moved, and because it's not easy to redirect
3190 // region stack entries to point to a new location (because objects
3191 // in one region might be scattered to multiple regions after they
3192 // are copied), one option is to ensure that all marked objects
3193 // copied during a pause are pushed on the stack. Notice, however,
3194 // that this problem can only happen when the region stack is not
3195 // empty during an evacuation pause. So, we make the fix a bit less
3196 // conservative and ensure that regions are pushed on the stack,
3197 // irrespective whether all collection set regions are below the
3198 // finger, if the region stack is not empty. This is expected to be
3199 // a rare case, so I don't think it's necessary to be smarted about it.
3200 if (!region_stack_empty() || has_aborted_regions()) {
3201 _should_gray_objects = true;
3202 }
3203 }
3204
3205 void ConcurrentMark::registerCSetRegion(HeapRegion* hr) {
3206 if (!concurrent_marking_in_progress()) return;
3207
3208 HeapWord* region_end = hr->end();
3209 if (region_end > _min_finger) {
3210 _should_gray_objects = true;
3211 }
3212 }
3213
3214 // Resets the region fields of active CMTasks whose values point
3215 // into the collection set.
3216 void ConcurrentMark::reset_active_task_region_fields_in_cset() {
3217 assert(SafepointSynchronize::is_at_safepoint(), "should be in STW");
3218 assert(parallel_marking_threads() <= _max_task_num, "sanity");
3219
3220 for (int i = 0; i < (int)parallel_marking_threads(); i += 1) {
3221 CMTask* task = _tasks[i];
3222 HeapWord* task_finger = task->finger();
3223 if (task_finger != NULL) {
3224 assert(_g1h->is_in_g1_reserved(task_finger), "not in heap");
3225 HeapRegion* finger_region = _g1h->heap_region_containing(task_finger);
3226 if (finger_region->in_collection_set()) {
3227 // The task's current region is in the collection set.
3228 // This region will be evacuated in the current GC and
3229 // the region fields in the task will be stale.
3230 task->giveup_current_region();
3231 }
3232 }
3233 }
3234 }
3235
3236 // abandon current marking iteration due to a Full GC
3237 void ConcurrentMark::abort() {
3238 // Clear all marks to force marking thread to do nothing
3239 _nextMarkBitMap->clearAll();
3240 // Empty mark stack
3241 clear_marking_state();
3242 for (int i = 0; i < (int)_max_task_num; ++i) {
3243 _tasks[i]->clear_region_fields();
3244 }
3245 _has_aborted = true;
3246
3247 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3248 satb_mq_set.abandon_partial_marking();
3249 // This can be called either during or outside marking, we'll read
3250 // the expected_active value from the SATB queue set.
3251 satb_mq_set.set_active_all_threads(
3252 false, /* new active value */
3253 satb_mq_set.is_active() /* expected_active */);
3254 }
3255
3256 static void print_ms_time_info(const char* prefix, const char* name,
3257 NumberSeq& ns) {
3258 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
3259 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
3260 if (ns.num() > 0) {
3261 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]",
3262 prefix, ns.sd(), ns.maximum());
3263 }
3264 }
3265
3266 void ConcurrentMark::print_summary_info() {
3267 gclog_or_tty->print_cr(" Concurrent marking:");
3268 print_ms_time_info(" ", "init marks", _init_times);
3269 print_ms_time_info(" ", "remarks", _remark_times);
3270 {
3271 print_ms_time_info(" ", "final marks", _remark_mark_times);
3272 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
3273
3274 }
3275 print_ms_time_info(" ", "cleanups", _cleanup_times);
3276 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).",
3277 _total_counting_time,
3278 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 /
3279 (double)_cleanup_times.num()
3280 : 0.0));
3281 if (G1ScrubRemSets) {
3282 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
3283 _total_rs_scrub_time,
3284 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 /
3285 (double)_cleanup_times.num()
3286 : 0.0));
3287 }
3288 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.",
3289 (_init_times.sum() + _remark_times.sum() +
3290 _cleanup_times.sum())/1000.0);
3291 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s "
3292 "(%8.2f s marking, %8.2f s counting).",
3293 cmThread()->vtime_accum(),
3294 cmThread()->vtime_mark_accum(),
3295 cmThread()->vtime_count_accum());
3296 }
3297
3298 void ConcurrentMark::print_worker_threads_on(outputStream* st) const {
3299 _parallel_workers->print_worker_threads_on(st);
3300 }
3301
3302 // Closures
3303 // XXX: there seems to be a lot of code duplication here;
3304 // should refactor and consolidate the shared code.
3305
3306 // This closure is used to mark refs into the CMS generation in
3307 // the CMS bit map. Called at the first checkpoint.
3308
3309 // We take a break if someone is trying to stop the world.
3310 bool ConcurrentMark::do_yield_check(int worker_i) {
3311 if (should_yield()) {
3312 if (worker_i == 0) {
3313 _g1h->g1_policy()->record_concurrent_pause();
3314 }
3315 cmThread()->yield();
3316 if (worker_i == 0) {
3317 _g1h->g1_policy()->record_concurrent_pause_end();
3318 }
3319 return true;
3320 } else {
3321 return false;
3322 }
3323 }
3324
3325 bool ConcurrentMark::should_yield() {
3326 return cmThread()->should_yield();
3327 }
3328
3329 bool ConcurrentMark::containing_card_is_marked(void* p) {
3330 size_t offset = pointer_delta(p, _g1h->reserved_region().start(), 1);
3331 return _card_bm.at(offset >> CardTableModRefBS::card_shift);
3332 }
3333
3334 bool ConcurrentMark::containing_cards_are_marked(void* start,
3335 void* last) {
3336 return containing_card_is_marked(start) &&
3337 containing_card_is_marked(last);
3338 }
3339
3340 #ifndef PRODUCT
3341 // for debugging purposes
3342 void ConcurrentMark::print_finger() {
3343 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT,
3344 _heap_start, _heap_end, _finger);
3345 for (int i = 0; i < (int) _max_task_num; ++i) {
3346 gclog_or_tty->print(" %d: "PTR_FORMAT, i, _tasks[i]->finger());
3347 }
3348 gclog_or_tty->print_cr("");
3349 }
3350 #endif
3351
3352 void CMTask::scan_object(oop obj) {
3353 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
3354
3355 if (_cm->verbose_high()) {
3356 gclog_or_tty->print_cr("[%d] we're scanning object "PTR_FORMAT,
3357 _task_id, (void*) obj);
3358 }
3359
3360 size_t obj_size = obj->size();
3361 _words_scanned += obj_size;
3362
3363 obj->oop_iterate(_cm_oop_closure);
3364 statsOnly( ++_objs_scanned );
3365 check_limits();
3366 }
3367
3368 // Closure for iteration over bitmaps
3369 class CMBitMapClosure : public BitMapClosure {
3370 private:
3371 // the bitmap that is being iterated over
3372 CMBitMap* _nextMarkBitMap;
3373 ConcurrentMark* _cm;
3374 CMTask* _task;
3375 // true if we're scanning a heap region claimed by the task (so that
3376 // we move the finger along), false if we're not, i.e. currently when
3377 // scanning a heap region popped from the region stack (so that we
3378 // do not move the task finger along; it'd be a mistake if we did so).
3379 bool _scanning_heap_region;
3380
3381 public:
3382 CMBitMapClosure(CMTask *task,
3383 ConcurrentMark* cm,
3384 CMBitMap* nextMarkBitMap)
3385 : _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
3386
3387 void set_scanning_heap_region(bool scanning_heap_region) {
3388 _scanning_heap_region = scanning_heap_region;
3389 }
3390
3391 bool do_bit(size_t offset) {
3392 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
3393 assert(_nextMarkBitMap->isMarked(addr), "invariant");
3394 assert( addr < _cm->finger(), "invariant");
3395
3396 if (_scanning_heap_region) {
3397 statsOnly( _task->increase_objs_found_on_bitmap() );
3398 assert(addr >= _task->finger(), "invariant");
3399 // We move that task's local finger along.
3400 _task->move_finger_to(addr);
3401 } else {
3402 // We move the task's region finger along.
3403 _task->move_region_finger_to(addr);
3404 }
3405
3406 _task->scan_object(oop(addr));
3407 // we only partially drain the local queue and global stack
3408 _task->drain_local_queue(true);
3409 _task->drain_global_stack(true);
3410
3411 // if the has_aborted flag has been raised, we need to bail out of
3412 // the iteration
3413 return !_task->has_aborted();
3414 }
3415 };
3416
3417 // Closure for iterating over objects, currently only used for
3418 // processing SATB buffers.
3419 class CMObjectClosure : public ObjectClosure {
3420 private:
3421 CMTask* _task;
3422
3423 public:
3424 void do_object(oop obj) {
3425 _task->deal_with_reference(obj);
3426 }
3427
3428 CMObjectClosure(CMTask* task) : _task(task) { }
3429 };
3430
3431 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
3432 ConcurrentMark* cm,
3433 CMTask* task)
3434 : _g1h(g1h), _cm(cm), _task(task) {
3435 assert(_ref_processor == NULL, "should be initialized to NULL");
3436
3437 if (G1UseConcMarkReferenceProcessing) {
3438 _ref_processor = g1h->ref_processor_cm();
3439 assert(_ref_processor != NULL, "should not be NULL");
3440 }
3441 }
3442
3443 void CMTask::setup_for_region(HeapRegion* hr) {
3444 // Separated the asserts so that we know which one fires.
3445 assert(hr != NULL,
3446 "claim_region() should have filtered out continues humongous regions");
3447 assert(!hr->continuesHumongous(),
3448 "claim_region() should have filtered out continues humongous regions");
3449
3450 if (_cm->verbose_low()) {
3451 gclog_or_tty->print_cr("[%d] setting up for region "PTR_FORMAT,
3452 _task_id, hr);
3453 }
3454
3455 _curr_region = hr;
3456 _finger = hr->bottom();
3457 update_region_limit();
3458 }
3459
3460 void CMTask::update_region_limit() {
3461 HeapRegion* hr = _curr_region;
3462 HeapWord* bottom = hr->bottom();
3463 HeapWord* limit = hr->next_top_at_mark_start();
3464
3465 if (limit == bottom) {
3466 if (_cm->verbose_low()) {
3467 gclog_or_tty->print_cr("[%d] found an empty region "
3468 "["PTR_FORMAT", "PTR_FORMAT")",
3469 _task_id, bottom, limit);
3470 }
3471 // The region was collected underneath our feet.
3472 // We set the finger to bottom to ensure that the bitmap
3473 // iteration that will follow this will not do anything.
3474 // (this is not a condition that holds when we set the region up,
3475 // as the region is not supposed to be empty in the first place)
3476 _finger = bottom;
3477 } else if (limit >= _region_limit) {
3478 assert(limit >= _finger, "peace of mind");
3479 } else {
3480 assert(limit < _region_limit, "only way to get here");
3481 // This can happen under some pretty unusual circumstances. An
3482 // evacuation pause empties the region underneath our feet (NTAMS
3483 // at bottom). We then do some allocation in the region (NTAMS
3484 // stays at bottom), followed by the region being used as a GC
3485 // alloc region (NTAMS will move to top() and the objects
3486 // originally below it will be grayed). All objects now marked in
3487 // the region are explicitly grayed, if below the global finger,
3488 // and we do not need in fact to scan anything else. So, we simply
3489 // set _finger to be limit to ensure that the bitmap iteration
3490 // doesn't do anything.
3491 _finger = limit;
3492 }
3493
3494 _region_limit = limit;
3495 }
3496
3497 void CMTask::giveup_current_region() {
3498 assert(_curr_region != NULL, "invariant");
3499 if (_cm->verbose_low()) {
3500 gclog_or_tty->print_cr("[%d] giving up region "PTR_FORMAT,
3501 _task_id, _curr_region);
3502 }
3503 clear_region_fields();
3504 }
3505
3506 void CMTask::clear_region_fields() {
3507 // Values for these three fields that indicate that we're not
3508 // holding on to a region.
3509 _curr_region = NULL;
3510 _finger = NULL;
3511 _region_limit = NULL;
3512
3513 _region_finger = NULL;
3514 }
3515
3516 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
3517 if (cm_oop_closure == NULL) {
3518 assert(_cm_oop_closure != NULL, "invariant");
3519 } else {
3520 assert(_cm_oop_closure == NULL, "invariant");
3521 }
3522 _cm_oop_closure = cm_oop_closure;
3523 }
3524
3525 void CMTask::reset(CMBitMap* nextMarkBitMap) {
3526 guarantee(nextMarkBitMap != NULL, "invariant");
3527
3528 if (_cm->verbose_low()) {
3529 gclog_or_tty->print_cr("[%d] resetting", _task_id);
3530 }
3531
3532 _nextMarkBitMap = nextMarkBitMap;
3533 clear_region_fields();
3534 assert(_aborted_region.is_empty(), "should have been cleared");
3535
3536 _calls = 0;
3537 _elapsed_time_ms = 0.0;
3538 _termination_time_ms = 0.0;
3539 _termination_start_time_ms = 0.0;
3540
3541 #if _MARKING_STATS_
3542 _local_pushes = 0;
3543 _local_pops = 0;
3544 _local_max_size = 0;
3545 _objs_scanned = 0;
3546 _global_pushes = 0;
3547 _global_pops = 0;
3548 _global_max_size = 0;
3549 _global_transfers_to = 0;
3550 _global_transfers_from = 0;
3551 _region_stack_pops = 0;
3552 _regions_claimed = 0;
3553 _objs_found_on_bitmap = 0;
3554 _satb_buffers_processed = 0;
3555 _steal_attempts = 0;
3556 _steals = 0;
3557 _aborted = 0;
3558 _aborted_overflow = 0;
3559 _aborted_cm_aborted = 0;
3560 _aborted_yield = 0;
3561 _aborted_timed_out = 0;
3562 _aborted_satb = 0;
3563 _aborted_termination = 0;
3564 #endif // _MARKING_STATS_
3565 }
3566
3567 bool CMTask::should_exit_termination() {
3568 regular_clock_call();
3569 // This is called when we are in the termination protocol. We should
3570 // quit if, for some reason, this task wants to abort or the global
3571 // stack is not empty (this means that we can get work from it).
3572 return !_cm->mark_stack_empty() || has_aborted();
3573 }
3574
3575 void CMTask::reached_limit() {
3576 assert(_words_scanned >= _words_scanned_limit ||
3577 _refs_reached >= _refs_reached_limit ,
3578 "shouldn't have been called otherwise");
3579 regular_clock_call();
3580 }
3581
3582 void CMTask::regular_clock_call() {
3583 if (has_aborted()) return;
3584
3585 // First, we need to recalculate the words scanned and refs reached
3586 // limits for the next clock call.
3587 recalculate_limits();
3588
3589 // During the regular clock call we do the following
3590
3591 // (1) If an overflow has been flagged, then we abort.
3592 if (_cm->has_overflown()) {
3593 set_has_aborted();
3594 return;
3595 }
3596
3597 // If we are not concurrent (i.e. we're doing remark) we don't need
3598 // to check anything else. The other steps are only needed during
3599 // the concurrent marking phase.
3600 if (!concurrent()) return;
3601
3602 // (2) If marking has been aborted for Full GC, then we also abort.
3603 if (_cm->has_aborted()) {
3604 set_has_aborted();
3605 statsOnly( ++_aborted_cm_aborted );
3606 return;
3607 }
3608
3609 double curr_time_ms = os::elapsedVTime() * 1000.0;
3610
3611 // (3) If marking stats are enabled, then we update the step history.
3612 #if _MARKING_STATS_
3613 if (_words_scanned >= _words_scanned_limit) {
3614 ++_clock_due_to_scanning;
3615 }
3616 if (_refs_reached >= _refs_reached_limit) {
3617 ++_clock_due_to_marking;
3618 }
3619
3620 double last_interval_ms = curr_time_ms - _interval_start_time_ms;
3621 _interval_start_time_ms = curr_time_ms;
3622 _all_clock_intervals_ms.add(last_interval_ms);
3623
3624 if (_cm->verbose_medium()) {
3625 gclog_or_tty->print_cr("[%d] regular clock, interval = %1.2lfms, "
3626 "scanned = %d%s, refs reached = %d%s",
3627 _task_id, last_interval_ms,
3628 _words_scanned,
3629 (_words_scanned >= _words_scanned_limit) ? " (*)" : "",
3630 _refs_reached,
3631 (_refs_reached >= _refs_reached_limit) ? " (*)" : "");
3632 }
3633 #endif // _MARKING_STATS_
3634
3635 // (4) We check whether we should yield. If we have to, then we abort.
3636 if (_cm->should_yield()) {
3637 // We should yield. To do this we abort the task. The caller is
3638 // responsible for yielding.
3639 set_has_aborted();
3640 statsOnly( ++_aborted_yield );
3641 return;
3642 }
3643
3644 // (5) We check whether we've reached our time quota. If we have,
3645 // then we abort.
3646 double elapsed_time_ms = curr_time_ms - _start_time_ms;
3647 if (elapsed_time_ms > _time_target_ms) {
3648 set_has_aborted();
3649 _has_timed_out = true;
3650 statsOnly( ++_aborted_timed_out );
3651 return;
3652 }
3653
3654 // (6) Finally, we check whether there are enough completed STAB
3655 // buffers available for processing. If there are, we abort.
3656 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3657 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
3658 if (_cm->verbose_low()) {
3659 gclog_or_tty->print_cr("[%d] aborting to deal with pending SATB buffers",
3660 _task_id);
3661 }
3662 // we do need to process SATB buffers, we'll abort and restart
3663 // the marking task to do so
3664 set_has_aborted();
3665 statsOnly( ++_aborted_satb );
3666 return;
3667 }
3668 }
3669
3670 void CMTask::recalculate_limits() {
3671 _real_words_scanned_limit = _words_scanned + words_scanned_period;
3672 _words_scanned_limit = _real_words_scanned_limit;
3673
3674 _real_refs_reached_limit = _refs_reached + refs_reached_period;
3675 _refs_reached_limit = _real_refs_reached_limit;
3676 }
3677
3678 void CMTask::decrease_limits() {
3679 // This is called when we believe that we're going to do an infrequent
3680 // operation which will increase the per byte scanned cost (i.e. move
3681 // entries to/from the global stack). It basically tries to decrease the
3682 // scanning limit so that the clock is called earlier.
3683
3684 if (_cm->verbose_medium()) {
3685 gclog_or_tty->print_cr("[%d] decreasing limits", _task_id);
3686 }
3687
3688 _words_scanned_limit = _real_words_scanned_limit -
3689 3 * words_scanned_period / 4;
3690 _refs_reached_limit = _real_refs_reached_limit -
3691 3 * refs_reached_period / 4;
3692 }
3693
3694 void CMTask::move_entries_to_global_stack() {
3695 // local array where we'll store the entries that will be popped
3696 // from the local queue
3697 oop buffer[global_stack_transfer_size];
3698
3699 int n = 0;
3700 oop obj;
3701 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
3702 buffer[n] = obj;
3703 ++n;
3704 }
3705
3706 if (n > 0) {
3707 // we popped at least one entry from the local queue
3708
3709 statsOnly( ++_global_transfers_to; _local_pops += n );
3710
3711 if (!_cm->mark_stack_push(buffer, n)) {
3712 if (_cm->verbose_low()) {
3713 gclog_or_tty->print_cr("[%d] aborting due to global stack overflow",
3714 _task_id);
3715 }
3716 set_has_aborted();
3717 } else {
3718 // the transfer was successful
3719
3720 if (_cm->verbose_medium()) {
3721 gclog_or_tty->print_cr("[%d] pushed %d entries to the global stack",
3722 _task_id, n);
3723 }
3724 statsOnly( int tmp_size = _cm->mark_stack_size();
3725 if (tmp_size > _global_max_size) {
3726 _global_max_size = tmp_size;
3727 }
3728 _global_pushes += n );
3729 }
3730 }
3731
3732 // this operation was quite expensive, so decrease the limits
3733 decrease_limits();
3734 }
3735
3736 void CMTask::get_entries_from_global_stack() {
3737 // local array where we'll store the entries that will be popped
3738 // from the global stack.
3739 oop buffer[global_stack_transfer_size];
3740 int n;
3741 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
3742 assert(n <= global_stack_transfer_size,
3743 "we should not pop more than the given limit");
3744 if (n > 0) {
3745 // yes, we did actually pop at least one entry
3746
3747 statsOnly( ++_global_transfers_from; _global_pops += n );
3748 if (_cm->verbose_medium()) {
3749 gclog_or_tty->print_cr("[%d] popped %d entries from the global stack",
3750 _task_id, n);
3751 }
3752 for (int i = 0; i < n; ++i) {
3753 bool success = _task_queue->push(buffer[i]);
3754 // We only call this when the local queue is empty or under a
3755 // given target limit. So, we do not expect this push to fail.
3756 assert(success, "invariant");
3757 }
3758
3759 statsOnly( int tmp_size = _task_queue->size();
3760 if (tmp_size > _local_max_size) {
3761 _local_max_size = tmp_size;
3762 }
3763 _local_pushes += n );
3764 }
3765
3766 // this operation was quite expensive, so decrease the limits
3767 decrease_limits();
3768 }
3769
3770 void CMTask::drain_local_queue(bool partially) {
3771 if (has_aborted()) return;
3772
3773 // Decide what the target size is, depending whether we're going to
3774 // drain it partially (so that other tasks can steal if they run out
3775 // of things to do) or totally (at the very end).
3776 size_t target_size;
3777 if (partially) {
3778 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
3779 } else {
3780 target_size = 0;
3781 }
3782
3783 if (_task_queue->size() > target_size) {
3784 if (_cm->verbose_high()) {
3785 gclog_or_tty->print_cr("[%d] draining local queue, target size = %d",
3786 _task_id, target_size);
3787 }
3788
3789 oop obj;
3790 bool ret = _task_queue->pop_local(obj);
3791 while (ret) {
3792 statsOnly( ++_local_pops );
3793
3794 if (_cm->verbose_high()) {
3795 gclog_or_tty->print_cr("[%d] popped "PTR_FORMAT, _task_id,
3796 (void*) obj);
3797 }
3798
3799 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
3800 assert(!_g1h->is_on_master_free_list(
3801 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
3802
3803 scan_object(obj);
3804
3805 if (_task_queue->size() <= target_size || has_aborted()) {
3806 ret = false;
3807 } else {
3808 ret = _task_queue->pop_local(obj);
3809 }
3810 }
3811
3812 if (_cm->verbose_high()) {
3813 gclog_or_tty->print_cr("[%d] drained local queue, size = %d",
3814 _task_id, _task_queue->size());
3815 }
3816 }
3817 }
3818
3819 void CMTask::drain_global_stack(bool partially) {
3820 if (has_aborted()) return;
3821
3822 // We have a policy to drain the local queue before we attempt to
3823 // drain the global stack.
3824 assert(partially || _task_queue->size() == 0, "invariant");
3825
3826 // Decide what the target size is, depending whether we're going to
3827 // drain it partially (so that other tasks can steal if they run out
3828 // of things to do) or totally (at the very end). Notice that,
3829 // because we move entries from the global stack in chunks or
3830 // because another task might be doing the same, we might in fact
3831 // drop below the target. But, this is not a problem.
3832 size_t target_size;
3833 if (partially) {
3834 target_size = _cm->partial_mark_stack_size_target();
3835 } else {
3836 target_size = 0;
3837 }
3838
3839 if (_cm->mark_stack_size() > target_size) {
3840 if (_cm->verbose_low()) {
3841 gclog_or_tty->print_cr("[%d] draining global_stack, target size %d",
3842 _task_id, target_size);
3843 }
3844
3845 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3846 get_entries_from_global_stack();
3847 drain_local_queue(partially);
3848 }
3849
3850 if (_cm->verbose_low()) {
3851 gclog_or_tty->print_cr("[%d] drained global stack, size = %d",
3852 _task_id, _cm->mark_stack_size());
3853 }
3854 }
3855 }
3856
3857 // SATB Queue has several assumptions on whether to call the par or
3858 // non-par versions of the methods. this is why some of the code is
3859 // replicated. We should really get rid of the single-threaded version
3860 // of the code to simplify things.
3861 void CMTask::drain_satb_buffers() {
3862 if (has_aborted()) return;
3863
3864 // We set this so that the regular clock knows that we're in the
3865 // middle of draining buffers and doesn't set the abort flag when it
3866 // notices that SATB buffers are available for draining. It'd be
3867 // very counter productive if it did that. :-)
3868 _draining_satb_buffers = true;
3869
3870 CMObjectClosure oc(this);
3871 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3872 if (G1CollectedHeap::use_parallel_gc_threads()) {
3873 satb_mq_set.set_par_closure(_task_id, &oc);
3874 } else {
3875 satb_mq_set.set_closure(&oc);
3876 }
3877
3878 // This keeps claiming and applying the closure to completed buffers
3879 // until we run out of buffers or we need to abort.
3880 if (G1CollectedHeap::use_parallel_gc_threads()) {
3881 while (!has_aborted() &&
3882 satb_mq_set.par_apply_closure_to_completed_buffer(_task_id)) {
3883 if (_cm->verbose_medium()) {
3884 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id);
3885 }
3886 statsOnly( ++_satb_buffers_processed );
3887 regular_clock_call();
3888 }
3889 } else {
3890 while (!has_aborted() &&
3891 satb_mq_set.apply_closure_to_completed_buffer()) {
3892 if (_cm->verbose_medium()) {
3893 gclog_or_tty->print_cr("[%d] processed an SATB buffer", _task_id);
3894 }
3895 statsOnly( ++_satb_buffers_processed );
3896 regular_clock_call();
3897 }
3898 }
3899
3900 if (!concurrent() && !has_aborted()) {
3901 // We should only do this during remark.
3902 if (G1CollectedHeap::use_parallel_gc_threads()) {
3903 satb_mq_set.par_iterate_closure_all_threads(_task_id);
3904 } else {
3905 satb_mq_set.iterate_closure_all_threads();
3906 }
3907 }
3908
3909 _draining_satb_buffers = false;
3910
3911 assert(has_aborted() ||
3912 concurrent() ||
3913 satb_mq_set.completed_buffers_num() == 0, "invariant");
3914
3915 if (G1CollectedHeap::use_parallel_gc_threads()) {
3916 satb_mq_set.set_par_closure(_task_id, NULL);
3917 } else {
3918 satb_mq_set.set_closure(NULL);
3919 }
3920
3921 // again, this was a potentially expensive operation, decrease the
3922 // limits to get the regular clock call early
3923 decrease_limits();
3924 }
3925
3926 void CMTask::drain_region_stack(BitMapClosure* bc) {
3927 if (has_aborted()) return;
3928
3929 assert(_region_finger == NULL,
3930 "it should be NULL when we're not scanning a region");
3931
3932 if (!_cm->region_stack_empty() || !_aborted_region.is_empty()) {
3933 if (_cm->verbose_low()) {
3934 gclog_or_tty->print_cr("[%d] draining region stack, size = %d",
3935 _task_id, _cm->region_stack_size());
3936 }
3937
3938 MemRegion mr;
3939
3940 if (!_aborted_region.is_empty()) {
3941 mr = _aborted_region;
3942 _aborted_region = MemRegion();
3943
3944 if (_cm->verbose_low()) {
3945 gclog_or_tty->print_cr("[%d] scanning aborted region "
3946 "[ " PTR_FORMAT ", " PTR_FORMAT " )",
3947 _task_id, mr.start(), mr.end());
3948 }
3949 } else {
3950 mr = _cm->region_stack_pop_lock_free();
3951 // it returns MemRegion() if the pop fails
3952 statsOnly(if (mr.start() != NULL) ++_region_stack_pops );
3953 }
3954
3955 while (mr.start() != NULL) {
3956 if (_cm->verbose_medium()) {
3957 gclog_or_tty->print_cr("[%d] we are scanning region "
3958 "["PTR_FORMAT", "PTR_FORMAT")",
3959 _task_id, mr.start(), mr.end());
3960 }
3961
3962 assert(mr.end() <= _cm->finger(),
3963 "otherwise the region shouldn't be on the stack");
3964 assert(!mr.is_empty(), "Only non-empty regions live on the region stack");
3965 if (_nextMarkBitMap->iterate(bc, mr)) {
3966 assert(!has_aborted(),
3967 "cannot abort the task without aborting the bitmap iteration");
3968
3969 // We finished iterating over the region without aborting.
3970 regular_clock_call();
3971 if (has_aborted()) {
3972 mr = MemRegion();
3973 } else {
3974 mr = _cm->region_stack_pop_lock_free();
3975 // it returns MemRegion() if the pop fails
3976 statsOnly(if (mr.start() != NULL) ++_region_stack_pops );
3977 }
3978 } else {
3979 assert(has_aborted(), "currently the only way to do so");
3980
3981 // The only way to abort the bitmap iteration is to return
3982 // false from the do_bit() method. However, inside the
3983 // do_bit() method we move the _region_finger to point to the
3984 // object currently being looked at. So, if we bail out, we
3985 // have definitely set _region_finger to something non-null.
3986 assert(_region_finger != NULL, "invariant");
3987
3988 // Make sure that any previously aborted region has been
3989 // cleared.
3990 assert(_aborted_region.is_empty(), "aborted region not cleared");
3991
3992 // The iteration was actually aborted. So now _region_finger
3993 // points to the address of the object we last scanned. If we
3994 // leave it there, when we restart this task, we will rescan
3995 // the object. It is easy to avoid this. We move the finger by
3996 // enough to point to the next possible object header (the
3997 // bitmap knows by how much we need to move it as it knows its
3998 // granularity).
3999 MemRegion newRegion =
4000 MemRegion(_nextMarkBitMap->nextWord(_region_finger), mr.end());
4001
4002 if (!newRegion.is_empty()) {
4003 if (_cm->verbose_low()) {
4004 gclog_or_tty->print_cr("[%d] recording unscanned region"
4005 "[" PTR_FORMAT "," PTR_FORMAT ") in CMTask",
4006 _task_id,
4007 newRegion.start(), newRegion.end());
4008 }
4009 // Now record the part of the region we didn't scan to
4010 // make sure this task scans it later.
4011 _aborted_region = newRegion;
4012 }
4013 // break from while
4014 mr = MemRegion();
4015 }
4016 _region_finger = NULL;
4017 }
4018
4019 if (_cm->verbose_low()) {
4020 gclog_or_tty->print_cr("[%d] drained region stack, size = %d",
4021 _task_id, _cm->region_stack_size());
4022 }
4023 }
4024 }
4025
4026 void CMTask::print_stats() {
4027 gclog_or_tty->print_cr("Marking Stats, task = %d, calls = %d",
4028 _task_id, _calls);
4029 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
4030 _elapsed_time_ms, _termination_time_ms);
4031 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
4032 _step_times_ms.num(), _step_times_ms.avg(),
4033 _step_times_ms.sd());
4034 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
4035 _step_times_ms.maximum(), _step_times_ms.sum());
4036
4037 #if _MARKING_STATS_
4038 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
4039 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(),
4040 _all_clock_intervals_ms.sd());
4041 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
4042 _all_clock_intervals_ms.maximum(),
4043 _all_clock_intervals_ms.sum());
4044 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d",
4045 _clock_due_to_scanning, _clock_due_to_marking);
4046 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d",
4047 _objs_scanned, _objs_found_on_bitmap);
4048 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d",
4049 _local_pushes, _local_pops, _local_max_size);
4050 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d",
4051 _global_pushes, _global_pops, _global_max_size);
4052 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d",
4053 _global_transfers_to,_global_transfers_from);
4054 gclog_or_tty->print_cr(" Regions: claimed = %d, Region Stack: pops = %d",
4055 _regions_claimed, _region_stack_pops);
4056 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed);
4057 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d",
4058 _steal_attempts, _steals);
4059 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted);
4060 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d",
4061 _aborted_overflow, _aborted_cm_aborted, _aborted_yield);
4062 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d",
4063 _aborted_timed_out, _aborted_satb, _aborted_termination);
4064 #endif // _MARKING_STATS_
4065 }
4066
4067 /*****************************************************************************
4068
4069 The do_marking_step(time_target_ms) method is the building block
4070 of the parallel marking framework. It can be called in parallel
4071 with other invocations of do_marking_step() on different tasks
4072 (but only one per task, obviously) and concurrently with the
4073 mutator threads, or during remark, hence it eliminates the need
4074 for two versions of the code. When called during remark, it will
4075 pick up from where the task left off during the concurrent marking
4076 phase. Interestingly, tasks are also claimable during evacuation
4077 pauses too, since do_marking_step() ensures that it aborts before
4078 it needs to yield.
4079
4080 The data structures that is uses to do marking work are the
4081 following:
4082
4083 (1) Marking Bitmap. If there are gray objects that appear only
4084 on the bitmap (this happens either when dealing with an overflow
4085 or when the initial marking phase has simply marked the roots
4086 and didn't push them on the stack), then tasks claim heap
4087 regions whose bitmap they then scan to find gray objects. A
4088 global finger indicates where the end of the last claimed region
4089 is. A local finger indicates how far into the region a task has
4090 scanned. The two fingers are used to determine how to gray an
4091 object (i.e. whether simply marking it is OK, as it will be
4092 visited by a task in the future, or whether it needs to be also
4093 pushed on a stack).
4094
4095 (2) Local Queue. The local queue of the task which is accessed
4096 reasonably efficiently by the task. Other tasks can steal from
4097 it when they run out of work. Throughout the marking phase, a
4098 task attempts to keep its local queue short but not totally
4099 empty, so that entries are available for stealing by other
4100 tasks. Only when there is no more work, a task will totally
4101 drain its local queue.
4102
4103 (3) Global Mark Stack. This handles local queue overflow. During
4104 marking only sets of entries are moved between it and the local
4105 queues, as access to it requires a mutex and more fine-grain
4106 interaction with it which might cause contention. If it
4107 overflows, then the marking phase should restart and iterate
4108 over the bitmap to identify gray objects. Throughout the marking
4109 phase, tasks attempt to keep the global mark stack at a small
4110 length but not totally empty, so that entries are available for
4111 popping by other tasks. Only when there is no more work, tasks
4112 will totally drain the global mark stack.
4113
4114 (4) Global Region Stack. Entries on it correspond to areas of
4115 the bitmap that need to be scanned since they contain gray
4116 objects. Pushes on the region stack only happen during
4117 evacuation pauses and typically correspond to areas covered by
4118 GC LABS. If it overflows, then the marking phase should restart
4119 and iterate over the bitmap to identify gray objects. Tasks will
4120 try to totally drain the region stack as soon as possible.
4121
4122 (5) SATB Buffer Queue. This is where completed SATB buffers are
4123 made available. Buffers are regularly removed from this queue
4124 and scanned for roots, so that the queue doesn't get too
4125 long. During remark, all completed buffers are processed, as
4126 well as the filled in parts of any uncompleted buffers.
4127
4128 The do_marking_step() method tries to abort when the time target
4129 has been reached. There are a few other cases when the
4130 do_marking_step() method also aborts:
4131
4132 (1) When the marking phase has been aborted (after a Full GC).
4133
4134 (2) When a global overflow (either on the global stack or the
4135 region stack) has been triggered. Before the task aborts, it
4136 will actually sync up with the other tasks to ensure that all
4137 the marking data structures (local queues, stacks, fingers etc.)
4138 are re-initialised so that when do_marking_step() completes,
4139 the marking phase can immediately restart.
4140
4141 (3) When enough completed SATB buffers are available. The
4142 do_marking_step() method only tries to drain SATB buffers right
4143 at the beginning. So, if enough buffers are available, the
4144 marking step aborts and the SATB buffers are processed at
4145 the beginning of the next invocation.
4146
4147 (4) To yield. when we have to yield then we abort and yield
4148 right at the end of do_marking_step(). This saves us from a lot
4149 of hassle as, by yielding we might allow a Full GC. If this
4150 happens then objects will be compacted underneath our feet, the
4151 heap might shrink, etc. We save checking for this by just
4152 aborting and doing the yield right at the end.
4153
4154 From the above it follows that the do_marking_step() method should
4155 be called in a loop (or, otherwise, regularly) until it completes.
4156
4157 If a marking step completes without its has_aborted() flag being
4158 true, it means it has completed the current marking phase (and
4159 also all other marking tasks have done so and have all synced up).
4160
4161 A method called regular_clock_call() is invoked "regularly" (in
4162 sub ms intervals) throughout marking. It is this clock method that
4163 checks all the abort conditions which were mentioned above and
4164 decides when the task should abort. A work-based scheme is used to
4165 trigger this clock method: when the number of object words the
4166 marking phase has scanned or the number of references the marking
4167 phase has visited reach a given limit. Additional invocations to
4168 the method clock have been planted in a few other strategic places
4169 too. The initial reason for the clock method was to avoid calling
4170 vtime too regularly, as it is quite expensive. So, once it was in
4171 place, it was natural to piggy-back all the other conditions on it
4172 too and not constantly check them throughout the code.
4173
4174 *****************************************************************************/
4175
4176 void CMTask::do_marking_step(double time_target_ms,
4177 bool do_stealing,
4178 bool do_termination) {
4179 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
4180 assert(concurrent() == _cm->concurrent(), "they should be the same");
4181
4182 assert(concurrent() || _cm->region_stack_empty(),
4183 "the region stack should have been cleared before remark");
4184 assert(concurrent() || !_cm->has_aborted_regions(),
4185 "aborted regions should have been cleared before remark");
4186 assert(_region_finger == NULL,
4187 "this should be non-null only when a region is being scanned");
4188
4189 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
4190 assert(_task_queues != NULL, "invariant");
4191 assert(_task_queue != NULL, "invariant");
4192 assert(_task_queues->queue(_task_id) == _task_queue, "invariant");
4193
4194 assert(!_claimed,
4195 "only one thread should claim this task at any one time");
4196
4197 // OK, this doesn't safeguard again all possible scenarios, as it is
4198 // possible for two threads to set the _claimed flag at the same
4199 // time. But it is only for debugging purposes anyway and it will
4200 // catch most problems.
4201 _claimed = true;
4202
4203 _start_time_ms = os::elapsedVTime() * 1000.0;
4204 statsOnly( _interval_start_time_ms = _start_time_ms );
4205
4206 double diff_prediction_ms =
4207 g1_policy->get_new_prediction(&_marking_step_diffs_ms);
4208 _time_target_ms = time_target_ms - diff_prediction_ms;
4209
4210 // set up the variables that are used in the work-based scheme to
4211 // call the regular clock method
4212 _words_scanned = 0;
4213 _refs_reached = 0;
4214 recalculate_limits();
4215
4216 // clear all flags
4217 clear_has_aborted();
4218 _has_timed_out = false;
4219 _draining_satb_buffers = false;
4220
4221 ++_calls;
4222
4223 if (_cm->verbose_low()) {
4224 gclog_or_tty->print_cr("[%d] >>>>>>>>>> START, call = %d, "
4225 "target = %1.2lfms >>>>>>>>>>",
4226 _task_id, _calls, _time_target_ms);
4227 }
4228
4229 // Set up the bitmap and oop closures. Anything that uses them is
4230 // eventually called from this method, so it is OK to allocate these
4231 // statically.
4232 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
4233 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
4234 set_cm_oop_closure(&cm_oop_closure);
4235
4236 if (_cm->has_overflown()) {
4237 // This can happen if the region stack or the mark stack overflows
4238 // during a GC pause and this task, after a yield point,
4239 // restarts. We have to abort as we need to get into the overflow
4240 // protocol which happens right at the end of this task.
4241 set_has_aborted();
4242 }
4243
4244 // First drain any available SATB buffers. After this, we will not
4245 // look at SATB buffers before the next invocation of this method.
4246 // If enough completed SATB buffers are queued up, the regular clock
4247 // will abort this task so that it restarts.
4248 drain_satb_buffers();
4249 // ...then partially drain the local queue and the global stack
4250 drain_local_queue(true);
4251 drain_global_stack(true);
4252
4253 // Then totally drain the region stack. We will not look at
4254 // it again before the next invocation of this method. Entries on
4255 // the region stack are only added during evacuation pauses, for
4256 // which we have to yield. When we do, we abort the task anyway so
4257 // it will look at the region stack again when it restarts.
4258 bitmap_closure.set_scanning_heap_region(false);
4259 drain_region_stack(&bitmap_closure);
4260 // ...then partially drain the local queue and the global stack
4261 drain_local_queue(true);
4262 drain_global_stack(true);
4263
4264 do {
4265 if (!has_aborted() && _curr_region != NULL) {
4266 // This means that we're already holding on to a region.
4267 assert(_finger != NULL, "if region is not NULL, then the finger "
4268 "should not be NULL either");
4269
4270 // We might have restarted this task after an evacuation pause
4271 // which might have evacuated the region we're holding on to
4272 // underneath our feet. Let's read its limit again to make sure
4273 // that we do not iterate over a region of the heap that
4274 // contains garbage (update_region_limit() will also move
4275 // _finger to the start of the region if it is found empty).
4276 update_region_limit();
4277 // We will start from _finger not from the start of the region,
4278 // as we might be restarting this task after aborting half-way
4279 // through scanning this region. In this case, _finger points to
4280 // the address where we last found a marked object. If this is a
4281 // fresh region, _finger points to start().
4282 MemRegion mr = MemRegion(_finger, _region_limit);
4283
4284 if (_cm->verbose_low()) {
4285 gclog_or_tty->print_cr("[%d] we're scanning part "
4286 "["PTR_FORMAT", "PTR_FORMAT") "
4287 "of region "PTR_FORMAT,
4288 _task_id, _finger, _region_limit, _curr_region);
4289 }
4290
4291 // Let's iterate over the bitmap of the part of the
4292 // region that is left.
4293 bitmap_closure.set_scanning_heap_region(true);
4294 if (mr.is_empty() ||
4295 _nextMarkBitMap->iterate(&bitmap_closure, mr)) {
4296 // We successfully completed iterating over the region. Now,
4297 // let's give up the region.
4298 giveup_current_region();
4299 regular_clock_call();
4300 } else {
4301 assert(has_aborted(), "currently the only way to do so");
4302 // The only way to abort the bitmap iteration is to return
4303 // false from the do_bit() method. However, inside the
4304 // do_bit() method we move the _finger to point to the
4305 // object currently being looked at. So, if we bail out, we
4306 // have definitely set _finger to something non-null.
4307 assert(_finger != NULL, "invariant");
4308
4309 // Region iteration was actually aborted. So now _finger
4310 // points to the address of the object we last scanned. If we
4311 // leave it there, when we restart this task, we will rescan
4312 // the object. It is easy to avoid this. We move the finger by
4313 // enough to point to the next possible object header (the
4314 // bitmap knows by how much we need to move it as it knows its
4315 // granularity).
4316 assert(_finger < _region_limit, "invariant");
4317 HeapWord* new_finger = _nextMarkBitMap->nextWord(_finger);
4318 // Check if bitmap iteration was aborted while scanning the last object
4319 if (new_finger >= _region_limit) {
4320 giveup_current_region();
4321 } else {
4322 move_finger_to(new_finger);
4323 }
4324 }
4325 }
4326 // At this point we have either completed iterating over the
4327 // region we were holding on to, or we have aborted.
4328
4329 // We then partially drain the local queue and the global stack.
4330 // (Do we really need this?)
4331 drain_local_queue(true);
4332 drain_global_stack(true);
4333
4334 // Read the note on the claim_region() method on why it might
4335 // return NULL with potentially more regions available for
4336 // claiming and why we have to check out_of_regions() to determine
4337 // whether we're done or not.
4338 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
4339 // We are going to try to claim a new region. We should have
4340 // given up on the previous one.
4341 // Separated the asserts so that we know which one fires.
4342 assert(_curr_region == NULL, "invariant");
4343 assert(_finger == NULL, "invariant");
4344 assert(_region_limit == NULL, "invariant");
4345 if (_cm->verbose_low()) {
4346 gclog_or_tty->print_cr("[%d] trying to claim a new region", _task_id);
4347 }
4348 HeapRegion* claimed_region = _cm->claim_region(_task_id);
4349 if (claimed_region != NULL) {
4350 // Yes, we managed to claim one
4351 statsOnly( ++_regions_claimed );
4352
4353 if (_cm->verbose_low()) {
4354 gclog_or_tty->print_cr("[%d] we successfully claimed "
4355 "region "PTR_FORMAT,
4356 _task_id, claimed_region);
4357 }
4358
4359 setup_for_region(claimed_region);
4360 assert(_curr_region == claimed_region, "invariant");
4361 }
4362 // It is important to call the regular clock here. It might take
4363 // a while to claim a region if, for example, we hit a large
4364 // block of empty regions. So we need to call the regular clock
4365 // method once round the loop to make sure it's called
4366 // frequently enough.
4367 regular_clock_call();
4368 }
4369
4370 if (!has_aborted() && _curr_region == NULL) {
4371 assert(_cm->out_of_regions(),
4372 "at this point we should be out of regions");
4373 }
4374 } while ( _curr_region != NULL && !has_aborted());
4375
4376 if (!has_aborted()) {
4377 // We cannot check whether the global stack is empty, since other
4378 // tasks might be pushing objects to it concurrently. We also cannot
4379 // check if the region stack is empty because if a thread is aborting
4380 // it can push a partially done region back.
4381 assert(_cm->out_of_regions(),
4382 "at this point we should be out of regions");
4383
4384 if (_cm->verbose_low()) {
4385 gclog_or_tty->print_cr("[%d] all regions claimed", _task_id);
4386 }
4387
4388 // Try to reduce the number of available SATB buffers so that
4389 // remark has less work to do.
4390 drain_satb_buffers();
4391 }
4392
4393 // Since we've done everything else, we can now totally drain the
4394 // local queue and global stack.
4395 drain_local_queue(false);
4396 drain_global_stack(false);
4397
4398 // Attempt at work stealing from other task's queues.
4399 if (do_stealing && !has_aborted()) {
4400 // We have not aborted. This means that we have finished all that
4401 // we could. Let's try to do some stealing...
4402
4403 // We cannot check whether the global stack is empty, since other
4404 // tasks might be pushing objects to it concurrently. We also cannot
4405 // check if the region stack is empty because if a thread is aborting
4406 // it can push a partially done region back.
4407 assert(_cm->out_of_regions() && _task_queue->size() == 0,
4408 "only way to reach here");
4409
4410 if (_cm->verbose_low()) {
4411 gclog_or_tty->print_cr("[%d] starting to steal", _task_id);
4412 }
4413
4414 while (!has_aborted()) {
4415 oop obj;
4416 statsOnly( ++_steal_attempts );
4417
4418 if (_cm->try_stealing(_task_id, &_hash_seed, obj)) {
4419 if (_cm->verbose_medium()) {
4420 gclog_or_tty->print_cr("[%d] stolen "PTR_FORMAT" successfully",
4421 _task_id, (void*) obj);
4422 }
4423
4424 statsOnly( ++_steals );
4425
4426 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
4427 "any stolen object should be marked");
4428 scan_object(obj);
4429
4430 // And since we're towards the end, let's totally drain the
4431 // local queue and global stack.
4432 drain_local_queue(false);
4433 drain_global_stack(false);
4434 } else {
4435 break;
4436 }
4437 }
4438 }
4439
4440 // If we are about to wrap up and go into termination, check if we
4441 // should raise the overflow flag.
4442 if (do_termination && !has_aborted()) {
4443 if (_cm->force_overflow()->should_force()) {
4444 _cm->set_has_overflown();
4445 regular_clock_call();
4446 }
4447 }
4448
4449 // We still haven't aborted. Now, let's try to get into the
4450 // termination protocol.
4451 if (do_termination && !has_aborted()) {
4452 // We cannot check whether the global stack is empty, since other
4453 // tasks might be concurrently pushing objects on it. We also cannot
4454 // check if the region stack is empty because if a thread is aborting
4455 // it can push a partially done region back.
4456 // Separated the asserts so that we know which one fires.
4457 assert(_cm->out_of_regions(), "only way to reach here");
4458 assert(_task_queue->size() == 0, "only way to reach here");
4459
4460 if (_cm->verbose_low()) {
4461 gclog_or_tty->print_cr("[%d] starting termination protocol", _task_id);
4462 }
4463
4464 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
4465 // The CMTask class also extends the TerminatorTerminator class,
4466 // hence its should_exit_termination() method will also decide
4467 // whether to exit the termination protocol or not.
4468 bool finished = _cm->terminator()->offer_termination(this);
4469 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
4470 _termination_time_ms +=
4471 termination_end_time_ms - _termination_start_time_ms;
4472
4473 if (finished) {
4474 // We're all done.
4475
4476 if (_task_id == 0) {
4477 // let's allow task 0 to do this
4478 if (concurrent()) {
4479 assert(_cm->concurrent_marking_in_progress(), "invariant");
4480 // we need to set this to false before the next
4481 // safepoint. This way we ensure that the marking phase
4482 // doesn't observe any more heap expansions.
4483 _cm->clear_concurrent_marking_in_progress();
4484 }
4485 }
4486
4487 // We can now guarantee that the global stack is empty, since
4488 // all other tasks have finished. We separated the guarantees so
4489 // that, if a condition is false, we can immediately find out
4490 // which one.
4491 guarantee(_cm->out_of_regions(), "only way to reach here");
4492 guarantee(_aborted_region.is_empty(), "only way to reach here");
4493 guarantee(_cm->region_stack_empty(), "only way to reach here");
4494 guarantee(_cm->mark_stack_empty(), "only way to reach here");
4495 guarantee(_task_queue->size() == 0, "only way to reach here");
4496 guarantee(!_cm->has_overflown(), "only way to reach here");
4497 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
4498 guarantee(!_cm->region_stack_overflow(), "only way to reach here");
4499
4500 if (_cm->verbose_low()) {
4501 gclog_or_tty->print_cr("[%d] all tasks terminated", _task_id);
4502 }
4503 } else {
4504 // Apparently there's more work to do. Let's abort this task. It
4505 // will restart it and we can hopefully find more things to do.
4506
4507 if (_cm->verbose_low()) {
4508 gclog_or_tty->print_cr("[%d] apparently there is more work to do",
4509 _task_id);
4510 }
4511
4512 set_has_aborted();
4513 statsOnly( ++_aborted_termination );
4514 }
4515 }
4516
4517 // Mainly for debugging purposes to make sure that a pointer to the
4518 // closure which was statically allocated in this frame doesn't
4519 // escape it by accident.
4520 set_cm_oop_closure(NULL);
4521 double end_time_ms = os::elapsedVTime() * 1000.0;
4522 double elapsed_time_ms = end_time_ms - _start_time_ms;
4523 // Update the step history.
4524 _step_times_ms.add(elapsed_time_ms);
4525
4526 if (has_aborted()) {
4527 // The task was aborted for some reason.
4528
4529 statsOnly( ++_aborted );
4530
4531 if (_has_timed_out) {
4532 double diff_ms = elapsed_time_ms - _time_target_ms;
4533 // Keep statistics of how well we did with respect to hitting
4534 // our target only if we actually timed out (if we aborted for
4535 // other reasons, then the results might get skewed).
4536 _marking_step_diffs_ms.add(diff_ms);
4537 }
4538
4539 if (_cm->has_overflown()) {
4540 // This is the interesting one. We aborted because a global
4541 // overflow was raised. This means we have to restart the
4542 // marking phase and start iterating over regions. However, in
4543 // order to do this we have to make sure that all tasks stop
4544 // what they are doing and re-initialise in a safe manner. We
4545 // will achieve this with the use of two barrier sync points.
4546
4547 if (_cm->verbose_low()) {
4548 gclog_or_tty->print_cr("[%d] detected overflow", _task_id);
4549 }
4550
4551 _cm->enter_first_sync_barrier(_task_id);
4552 // When we exit this sync barrier we know that all tasks have
4553 // stopped doing marking work. So, it's now safe to
4554 // re-initialise our data structures. At the end of this method,
4555 // task 0 will clear the global data structures.
4556
4557 statsOnly( ++_aborted_overflow );
4558
4559 // We clear the local state of this task...
4560 clear_region_fields();
4561
4562 // ...and enter the second barrier.
4563 _cm->enter_second_sync_barrier(_task_id);
4564 // At this point everything has bee re-initialised and we're
4565 // ready to restart.
4566 }
4567
4568 if (_cm->verbose_low()) {
4569 gclog_or_tty->print_cr("[%d] <<<<<<<<<< ABORTING, target = %1.2lfms, "
4570 "elapsed = %1.2lfms <<<<<<<<<<",
4571 _task_id, _time_target_ms, elapsed_time_ms);
4572 if (_cm->has_aborted()) {
4573 gclog_or_tty->print_cr("[%d] ========== MARKING ABORTED ==========",
4574 _task_id);
4575 }
4576 }
4577 } else {
4578 if (_cm->verbose_low()) {
4579 gclog_or_tty->print_cr("[%d] <<<<<<<<<< FINISHED, target = %1.2lfms, "
4580 "elapsed = %1.2lfms <<<<<<<<<<",
4581 _task_id, _time_target_ms, elapsed_time_ms);
4582 }
4583 }
4584
4585 _claimed = false;
4586 }
4587
4588 CMTask::CMTask(int task_id,
4589 ConcurrentMark* cm,
4590 CMTaskQueue* task_queue,
4591 CMTaskQueueSet* task_queues)
4592 : _g1h(G1CollectedHeap::heap()),
4593 _task_id(task_id), _cm(cm),
4594 _claimed(false),
4595 _nextMarkBitMap(NULL), _hash_seed(17),
4596 _task_queue(task_queue),
4597 _task_queues(task_queues),
4598 _cm_oop_closure(NULL),
4599 _aborted_region(MemRegion()) {
4600 guarantee(task_queue != NULL, "invariant");
4601 guarantee(task_queues != NULL, "invariant");
4602
4603 statsOnly( _clock_due_to_scanning = 0;
4604 _clock_due_to_marking = 0 );
4605
4606 _marking_step_diffs_ms.add(0.5);
4607 }
4608
4609 // These are formatting macros that are used below to ensure
4610 // consistent formatting. The *_H_* versions are used to format the
4611 // header for a particular value and they should be kept consistent
4612 // with the corresponding macro. Also note that most of the macros add
4613 // the necessary white space (as a prefix) which makes them a bit
4614 // easier to compose.
4615
4616 // All the output lines are prefixed with this string to be able to
4617 // identify them easily in a large log file.
4618 #define G1PPRL_LINE_PREFIX "###"
4619
4620 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT
4621 #ifdef _LP64
4622 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
4623 #else // _LP64
4624 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
4625 #endif // _LP64
4626
4627 // For per-region info
4628 #define G1PPRL_TYPE_FORMAT " %-4s"
4629 #define G1PPRL_TYPE_H_FORMAT " %4s"
4630 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9)
4631 #define G1PPRL_BYTE_H_FORMAT " %9s"
4632 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
4633 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
4634
4635 // For summary info
4636 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT
4637 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT
4638 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB"
4639 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%"
4640
4641 G1PrintRegionLivenessInfoClosure::
4642 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name)
4643 : _out(out),
4644 _total_used_bytes(0), _total_capacity_bytes(0),
4645 _total_prev_live_bytes(0), _total_next_live_bytes(0),
4646 _hum_used_bytes(0), _hum_capacity_bytes(0),
4647 _hum_prev_live_bytes(0), _hum_next_live_bytes(0) {
4648 G1CollectedHeap* g1h = G1CollectedHeap::heap();
4649 MemRegion g1_committed = g1h->g1_committed();
4650 MemRegion g1_reserved = g1h->g1_reserved();
4651 double now = os::elapsedTime();
4652
4653 // Print the header of the output.
4654 _out->cr();
4655 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
4656 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP"
4657 G1PPRL_SUM_ADDR_FORMAT("committed")
4658 G1PPRL_SUM_ADDR_FORMAT("reserved")
4659 G1PPRL_SUM_BYTE_FORMAT("region-size"),
4660 g1_committed.start(), g1_committed.end(),
4661 g1_reserved.start(), g1_reserved.end(),
4662 HeapRegion::GrainBytes);
4663 _out->print_cr(G1PPRL_LINE_PREFIX);
4664 _out->print_cr(G1PPRL_LINE_PREFIX
4665 G1PPRL_TYPE_H_FORMAT
4666 G1PPRL_ADDR_BASE_H_FORMAT
4667 G1PPRL_BYTE_H_FORMAT
4668 G1PPRL_BYTE_H_FORMAT
4669 G1PPRL_BYTE_H_FORMAT
4670 G1PPRL_DOUBLE_H_FORMAT,
4671 "type", "address-range",
4672 "used", "prev-live", "next-live", "gc-eff");
4673 _out->print_cr(G1PPRL_LINE_PREFIX
4674 G1PPRL_TYPE_H_FORMAT
4675 G1PPRL_ADDR_BASE_H_FORMAT
4676 G1PPRL_BYTE_H_FORMAT
4677 G1PPRL_BYTE_H_FORMAT
4678 G1PPRL_BYTE_H_FORMAT
4679 G1PPRL_DOUBLE_H_FORMAT,
4680 "", "",
4681 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)");
4682 }
4683
4684 // It takes as a parameter a reference to one of the _hum_* fields, it
4685 // deduces the corresponding value for a region in a humongous region
4686 // series (either the region size, or what's left if the _hum_* field
4687 // is < the region size), and updates the _hum_* field accordingly.
4688 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
4689 size_t bytes = 0;
4690 // The > 0 check is to deal with the prev and next live bytes which
4691 // could be 0.
4692 if (*hum_bytes > 0) {
4693 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
4694 *hum_bytes -= bytes;
4695 }
4696 return bytes;
4697 }
4698
4699 // It deduces the values for a region in a humongous region series
4700 // from the _hum_* fields and updates those accordingly. It assumes
4701 // that that _hum_* fields have already been set up from the "starts
4702 // humongous" region and we visit the regions in address order.
4703 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
4704 size_t* capacity_bytes,
4705 size_t* prev_live_bytes,
4706 size_t* next_live_bytes) {
4707 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
4708 *used_bytes = get_hum_bytes(&_hum_used_bytes);
4709 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
4710 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
4711 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
4712 }
4713
4714 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
4715 const char* type = "";
4716 HeapWord* bottom = r->bottom();
4717 HeapWord* end = r->end();
4718 size_t capacity_bytes = r->capacity();
4719 size_t used_bytes = r->used();
4720 size_t prev_live_bytes = r->live_bytes();
4721 size_t next_live_bytes = r->next_live_bytes();
4722 double gc_eff = r->gc_efficiency();
4723 if (r->used() == 0) {
4724 type = "FREE";
4725 } else if (r->is_survivor()) {
4726 type = "SURV";
4727 } else if (r->is_young()) {
4728 type = "EDEN";
4729 } else if (r->startsHumongous()) {
4730 type = "HUMS";
4731
4732 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
4733 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
4734 "they should have been zeroed after the last time we used them");
4735 // Set up the _hum_* fields.
4736 _hum_capacity_bytes = capacity_bytes;
4737 _hum_used_bytes = used_bytes;
4738 _hum_prev_live_bytes = prev_live_bytes;
4739 _hum_next_live_bytes = next_live_bytes;
4740 get_hum_bytes(&used_bytes, &capacity_bytes,
4741 &prev_live_bytes, &next_live_bytes);
4742 end = bottom + HeapRegion::GrainWords;
4743 } else if (r->continuesHumongous()) {
4744 type = "HUMC";
4745 get_hum_bytes(&used_bytes, &capacity_bytes,
4746 &prev_live_bytes, &next_live_bytes);
4747 assert(end == bottom + HeapRegion::GrainWords, "invariant");
4748 } else {
4749 type = "OLD";
4750 }
4751
4752 _total_used_bytes += used_bytes;
4753 _total_capacity_bytes += capacity_bytes;
4754 _total_prev_live_bytes += prev_live_bytes;
4755 _total_next_live_bytes += next_live_bytes;
4756
4757 // Print a line for this particular region.
4758 _out->print_cr(G1PPRL_LINE_PREFIX
4759 G1PPRL_TYPE_FORMAT
4760 G1PPRL_ADDR_BASE_FORMAT
4761 G1PPRL_BYTE_FORMAT
4762 G1PPRL_BYTE_FORMAT
4763 G1PPRL_BYTE_FORMAT
4764 G1PPRL_DOUBLE_FORMAT,
4765 type, bottom, end,
4766 used_bytes, prev_live_bytes, next_live_bytes, gc_eff);
4767
4768 return false;
4769 }
4770
4771 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
4772 // Print the footer of the output.
4773 _out->print_cr(G1PPRL_LINE_PREFIX);
4774 _out->print_cr(G1PPRL_LINE_PREFIX
4775 " SUMMARY"
4776 G1PPRL_SUM_MB_FORMAT("capacity")
4777 G1PPRL_SUM_MB_PERC_FORMAT("used")
4778 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
4779 G1PPRL_SUM_MB_PERC_FORMAT("next-live"),
4780 bytes_to_mb(_total_capacity_bytes),
4781 bytes_to_mb(_total_used_bytes),
4782 perc(_total_used_bytes, _total_capacity_bytes),
4783 bytes_to_mb(_total_prev_live_bytes),
4784 perc(_total_prev_live_bytes, _total_capacity_bytes),
4785 bytes_to_mb(_total_next_live_bytes),
4786 perc(_total_next_live_bytes, _total_capacity_bytes));
4787 _out->cr();
4788 }