1 /*
2 * Copyright (c) 2001, 2014, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation.
8 *
9 * This code is distributed in the hope that it will be useful, but WITHOUT
10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
12 * version 2 for more details (a copy is included in the LICENSE file that
13 * accompanied this code).
14 *
15 * You should have received a copy of the GNU General Public License version
16 * 2 along with this work; if not, write to the Free Software Foundation,
17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
18 *
19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
20 * or visit www.oracle.com if you need additional information or have any
21 * questions.
22 *
23 */
24
25 #include "precompiled.hpp"
26 #include "classfile/symbolTable.hpp"
27 #include "gc_implementation/g1/concurrentMark.inline.hpp"
28 #include "gc_implementation/g1/concurrentMarkThread.inline.hpp"
29 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp"
30 #include "gc_implementation/g1/g1CollectorPolicy.hpp"
31 #include "gc_implementation/g1/g1ErgoVerbose.hpp"
32 #include "gc_implementation/g1/g1Log.hpp"
33 #include "gc_implementation/g1/g1OopClosures.inline.hpp"
34 #include "gc_implementation/g1/g1RemSet.hpp"
35 #include "gc_implementation/g1/heapRegion.inline.hpp"
36 #include "gc_implementation/g1/heapRegionRemSet.hpp"
37 #include "gc_implementation/g1/heapRegionSeq.inline.hpp"
38 #include "gc_implementation/shared/vmGCOperations.hpp"
39 #include "gc_implementation/shared/gcTimer.hpp"
40 #include "gc_implementation/shared/gcTrace.hpp"
41 #include "gc_implementation/shared/gcTraceTime.hpp"
42 #include "memory/genOopClosures.inline.hpp"
43 #include "memory/referencePolicy.hpp"
44 #include "memory/resourceArea.hpp"
45 #include "oops/oop.inline.hpp"
46 #include "runtime/handles.inline.hpp"
47 #include "runtime/java.hpp"
48 #include "runtime/prefetch.inline.hpp"
49 #include "services/memTracker.hpp"
50
51 // Concurrent marking bit map wrapper
52
53 CMBitMapRO::CMBitMapRO(int shifter) :
54 _bm(),
55 _shifter(shifter) {
56 _bmStartWord = 0;
57 _bmWordSize = 0;
58 }
59
60 HeapWord* CMBitMapRO::getNextMarkedWordAddress(HeapWord* addr,
61 HeapWord* limit) const {
62 // First we must round addr *up* to a possible object boundary.
63 addr = (HeapWord*)align_size_up((intptr_t)addr,
64 HeapWordSize << _shifter);
65 size_t addrOffset = heapWordToOffset(addr);
66 if (limit == NULL) {
67 limit = _bmStartWord + _bmWordSize;
68 }
69 size_t limitOffset = heapWordToOffset(limit);
70 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset);
71 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
72 assert(nextAddr >= addr, "get_next_one postcondition");
73 assert(nextAddr == limit || isMarked(nextAddr),
74 "get_next_one postcondition");
75 return nextAddr;
76 }
77
78 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(HeapWord* addr,
79 HeapWord* limit) const {
80 size_t addrOffset = heapWordToOffset(addr);
81 if (limit == NULL) {
82 limit = _bmStartWord + _bmWordSize;
83 }
84 size_t limitOffset = heapWordToOffset(limit);
85 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset);
86 HeapWord* nextAddr = offsetToHeapWord(nextOffset);
87 assert(nextAddr >= addr, "get_next_one postcondition");
88 assert(nextAddr == limit || !isMarked(nextAddr),
89 "get_next_one postcondition");
90 return nextAddr;
91 }
92
93 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const {
94 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check");
95 return (int) (diff >> _shifter);
96 }
97
98 #ifndef PRODUCT
99 bool CMBitMapRO::covers(ReservedSpace heap_rs) const {
100 // assert(_bm.map() == _virtual_space.low(), "map inconsistency");
101 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize,
102 "size inconsistency");
103 return _bmStartWord == (HeapWord*)(heap_rs.base()) &&
104 _bmWordSize == heap_rs.size()>>LogHeapWordSize;
105 }
106 #endif
107
108 void CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const {
109 _bm.print_on_error(st, prefix);
110 }
111
112 bool CMBitMap::allocate(ReservedSpace heap_rs) {
113 _bmStartWord = (HeapWord*)(heap_rs.base());
114 _bmWordSize = heap_rs.size()/HeapWordSize; // heap_rs.size() is in bytes
115 ReservedSpace brs(ReservedSpace::allocation_align_size_up(
116 (_bmWordSize >> (_shifter + LogBitsPerByte)) + 1));
117 if (!brs.is_reserved()) {
118 warning("ConcurrentMark marking bit map allocation failure");
119 return false;
120 }
121 MemTracker::record_virtual_memory_type((address)brs.base(), mtGC);
122 // For now we'll just commit all of the bit map up front.
123 // Later on we'll try to be more parsimonious with swap.
124 if (!_virtual_space.initialize(brs, brs.size())) {
125 warning("ConcurrentMark marking bit map backing store failure");
126 return false;
127 }
128 assert(_virtual_space.committed_size() == brs.size(),
129 "didn't reserve backing store for all of concurrent marking bit map?");
130 _bm.set_map((BitMap::bm_word_t*)_virtual_space.low());
131 assert(_virtual_space.committed_size() << (_shifter + LogBitsPerByte) >=
132 _bmWordSize, "inconsistency in bit map sizing");
133 _bm.set_size(_bmWordSize >> _shifter);
134 return true;
135 }
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 bool CMMarkStack::allocate(size_t capacity) {
184 // allocate a stack of the requisite depth
185 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop)));
186 if (!rs.is_reserved()) {
187 warning("ConcurrentMark MarkStack allocation failure");
188 return false;
189 }
190 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC);
191 if (!_virtual_space.initialize(rs, rs.size())) {
192 warning("ConcurrentMark MarkStack backing store failure");
193 // Release the virtual memory reserved for the marking stack
194 rs.release();
195 return false;
196 }
197 assert(_virtual_space.committed_size() == rs.size(),
198 "Didn't reserve backing store for all of ConcurrentMark stack?");
199 _base = (oop*) _virtual_space.low();
200 setEmpty();
201 _capacity = (jint) capacity;
202 _saved_index = -1;
203 _should_expand = false;
204 NOT_PRODUCT(_max_depth = 0);
205 return true;
206 }
207
208 void CMMarkStack::expand() {
209 // Called, during remark, if we've overflown the marking stack during marking.
210 assert(isEmpty(), "stack should been emptied while handling overflow");
211 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted");
212 // Clear expansion flag
213 _should_expand = false;
214 if (_capacity == (jint) MarkStackSizeMax) {
215 if (PrintGCDetails && Verbose) {
216 gclog_or_tty->print_cr(" (benign) Can't expand marking stack capacity, at max size limit");
217 }
218 return;
219 }
220 // Double capacity if possible
221 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax);
222 // Do not give up existing stack until we have managed to
223 // get the double capacity that we desired.
224 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity *
225 sizeof(oop)));
226 if (rs.is_reserved()) {
227 // Release the backing store associated with old stack
228 _virtual_space.release();
229 // Reinitialize virtual space for new stack
230 if (!_virtual_space.initialize(rs, rs.size())) {
231 fatal("Not enough swap for expanded marking stack capacity");
232 }
233 _base = (oop*)(_virtual_space.low());
234 _index = 0;
235 _capacity = new_capacity;
236 } else {
237 if (PrintGCDetails && Verbose) {
238 // Failed to double capacity, continue;
239 gclog_or_tty->print(" (benign) Failed to expand marking stack capacity from "
240 SIZE_FORMAT"K to " SIZE_FORMAT"K",
241 _capacity / K, new_capacity / K);
242 }
243 }
244 }
245
246 void CMMarkStack::set_should_expand() {
247 // If we're resetting the marking state because of an
248 // marking stack overflow, record that we should, if
249 // possible, expand the stack.
250 _should_expand = _cm->has_overflown();
251 }
252
253 CMMarkStack::~CMMarkStack() {
254 if (_base != NULL) {
255 _base = NULL;
256 _virtual_space.release();
257 }
258 }
259
260 void CMMarkStack::par_push(oop ptr) {
261 while (true) {
262 if (isFull()) {
263 _overflow = true;
264 return;
265 }
266 // Otherwise...
267 jint index = _index;
268 jint next_index = index+1;
269 jint res = Atomic::cmpxchg(next_index, &_index, index);
270 if (res == index) {
271 _base[index] = ptr;
272 // Note that we don't maintain this atomically. We could, but it
273 // doesn't seem necessary.
274 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
275 return;
276 }
277 // Otherwise, we need to try again.
278 }
279 }
280
281 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) {
282 while (true) {
283 if (isFull()) {
284 _overflow = true;
285 return;
286 }
287 // Otherwise...
288 jint index = _index;
289 jint next_index = index + n;
290 if (next_index > _capacity) {
291 _overflow = true;
292 return;
293 }
294 jint res = Atomic::cmpxchg(next_index, &_index, index);
295 if (res == index) {
296 for (int i = 0; i < n; i++) {
297 int ind = index + i;
298 assert(ind < _capacity, "By overflow test above.");
299 _base[ind] = ptr_arr[i];
300 }
301 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
302 return;
303 }
304 // Otherwise, we need to try again.
305 }
306 }
307
308 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) {
309 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
310 jint start = _index;
311 jint next_index = start + n;
312 if (next_index > _capacity) {
313 _overflow = true;
314 return;
315 }
316 // Otherwise.
317 _index = next_index;
318 for (int i = 0; i < n; i++) {
319 int ind = start + i;
320 assert(ind < _capacity, "By overflow test above.");
321 _base[ind] = ptr_arr[i];
322 }
323 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index));
324 }
325
326 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) {
327 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
328 jint index = _index;
329 if (index == 0) {
330 *n = 0;
331 return false;
332 } else {
333 int k = MIN2(max, index);
334 jint new_ind = index - k;
335 for (int j = 0; j < k; j++) {
336 ptr_arr[j] = _base[new_ind + j];
337 }
338 _index = new_ind;
339 *n = k;
340 return true;
341 }
342 }
343
344 template<class OopClosureClass>
345 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) {
346 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after
347 || SafepointSynchronize::is_at_safepoint(),
348 "Drain recursion must be yield-safe.");
349 bool res = true;
350 debug_only(_drain_in_progress = true);
351 debug_only(_drain_in_progress_yields = yield_after);
352 while (!isEmpty()) {
353 oop newOop = pop();
354 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop");
355 assert(newOop->is_oop(), "Expected an oop");
356 assert(bm == NULL || bm->isMarked((HeapWord*)newOop),
357 "only grey objects on this stack");
358 newOop->oop_iterate(cl);
359 if (yield_after && _cm->do_yield_check()) {
360 res = false;
361 break;
362 }
363 }
364 debug_only(_drain_in_progress = false);
365 return res;
366 }
367
368 void CMMarkStack::note_start_of_gc() {
369 assert(_saved_index == -1,
370 "note_start_of_gc()/end_of_gc() bracketed incorrectly");
371 _saved_index = _index;
372 }
373
374 void CMMarkStack::note_end_of_gc() {
375 // This is intentionally a guarantee, instead of an assert. If we
376 // accidentally add something to the mark stack during GC, it
377 // will be a correctness issue so it's better if we crash. we'll
378 // only check this once per GC anyway, so it won't be a performance
379 // issue in any way.
380 guarantee(_saved_index == _index,
381 err_msg("saved index: %d index: %d", _saved_index, _index));
382 _saved_index = -1;
383 }
384
385 void CMMarkStack::oops_do(OopClosure* f) {
386 assert(_saved_index == _index,
387 err_msg("saved index: %d index: %d", _saved_index, _index));
388 for (int i = 0; i < _index; i += 1) {
389 f->do_oop(&_base[i]);
390 }
391 }
392
393 bool ConcurrentMark::not_yet_marked(oop obj) const {
394 return _g1h->is_obj_ill(obj);
395 }
396
397 CMRootRegions::CMRootRegions() :
398 _young_list(NULL), _cm(NULL), _scan_in_progress(false),
399 _should_abort(false), _next_survivor(NULL) { }
400
401 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) {
402 _young_list = g1h->young_list();
403 _cm = cm;
404 }
405
406 void CMRootRegions::prepare_for_scan() {
407 assert(!scan_in_progress(), "pre-condition");
408
409 // Currently, only survivors can be root regions.
410 assert(_next_survivor == NULL, "pre-condition");
411 _next_survivor = _young_list->first_survivor_region();
412 _scan_in_progress = (_next_survivor != NULL);
413 _should_abort = false;
414 }
415
416 HeapRegion* CMRootRegions::claim_next() {
417 if (_should_abort) {
418 // If someone has set the should_abort flag, we return NULL to
419 // force the caller to bail out of their loop.
420 return NULL;
421 }
422
423 // Currently, only survivors can be root regions.
424 HeapRegion* res = _next_survivor;
425 if (res != NULL) {
426 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
427 // Read it again in case it changed while we were waiting for the lock.
428 res = _next_survivor;
429 if (res != NULL) {
430 if (res == _young_list->last_survivor_region()) {
431 // We just claimed the last survivor so store NULL to indicate
432 // that we're done.
433 _next_survivor = NULL;
434 } else {
435 _next_survivor = res->get_next_young_region();
436 }
437 } else {
438 // Someone else claimed the last survivor while we were trying
439 // to take the lock so nothing else to do.
440 }
441 }
442 assert(res == NULL || res->is_survivor(), "post-condition");
443
444 return res;
445 }
446
447 void CMRootRegions::scan_finished() {
448 assert(scan_in_progress(), "pre-condition");
449
450 // Currently, only survivors can be root regions.
451 if (!_should_abort) {
452 assert(_next_survivor == NULL, "we should have claimed all survivors");
453 }
454 _next_survivor = NULL;
455
456 {
457 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
458 _scan_in_progress = false;
459 RootRegionScan_lock->notify_all();
460 }
461 }
462
463 bool CMRootRegions::wait_until_scan_finished() {
464 if (!scan_in_progress()) return false;
465
466 {
467 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag);
468 while (scan_in_progress()) {
469 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag);
470 }
471 }
472 return true;
473 }
474
475 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away
476 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list
477 #endif // _MSC_VER
478
479 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) {
480 return MAX2((n_par_threads + 2) / 4, 1U);
481 }
482
483 ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, ReservedSpace heap_rs) :
484 _g1h(g1h),
485 _markBitMap1(log2_intptr(MinObjAlignment)),
486 _markBitMap2(log2_intptr(MinObjAlignment)),
487 _parallel_marking_threads(0),
488 _max_parallel_marking_threads(0),
489 _sleep_factor(0.0),
490 _marking_task_overhead(1.0),
491 _cleanup_sleep_factor(0.0),
492 _cleanup_task_overhead(1.0),
493 _cleanup_list("Cleanup List"),
494 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/),
495 _card_bm((heap_rs.size() + CardTableModRefBS::card_size - 1) >>
496 CardTableModRefBS::card_shift,
497 false /* in_resource_area*/),
498
499 _prevMarkBitMap(&_markBitMap1),
500 _nextMarkBitMap(&_markBitMap2),
501
502 _markStack(this),
503 // _finger set in set_non_marking_state
504
505 _max_worker_id(MAX2((uint)ParallelGCThreads, 1U)),
506 // _active_tasks set in set_non_marking_state
507 // _tasks set inside the constructor
508 _task_queues(new CMTaskQueueSet((int) _max_worker_id)),
509 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)),
510
511 _has_overflown(false),
512 _concurrent(false),
513 _has_aborted(false),
514 _aborted_gc_id(GCId::undefined()),
515 _restart_for_overflow(false),
516 _concurrent_marking_in_progress(false),
517
518 // _verbose_level set below
519
520 _init_times(),
521 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(),
522 _cleanup_times(),
523 _total_counting_time(0.0),
524 _total_rs_scrub_time(0.0),
525
526 _parallel_workers(NULL),
527
528 _count_card_bitmaps(NULL),
529 _count_marked_bytes(NULL),
530 _completed_initialization(false) {
531 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel;
532 if (verbose_level < no_verbose) {
533 verbose_level = no_verbose;
534 }
535 if (verbose_level > high_verbose) {
536 verbose_level = high_verbose;
537 }
538 _verbose_level = verbose_level;
539
540 if (verbose_low()) {
541 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", "
542 "heap end = " INTPTR_FORMAT, p2i(_heap_start), p2i(_heap_end));
543 }
544
545 if (!_markBitMap1.allocate(heap_rs)) {
546 warning("Failed to allocate first CM bit map");
547 return;
548 }
549 if (!_markBitMap2.allocate(heap_rs)) {
550 warning("Failed to allocate second CM bit map");
551 return;
552 }
553
554 // Create & start a ConcurrentMark thread.
555 _cmThread = new ConcurrentMarkThread(this);
556 assert(cmThread() != NULL, "CM Thread should have been created");
557 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm");
558 if (_cmThread->osthread() == NULL) {
559 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread");
560 }
561
562 assert(CGC_lock != NULL, "Where's the CGC_lock?");
563 assert(_markBitMap1.covers(heap_rs), "_markBitMap1 inconsistency");
564 assert(_markBitMap2.covers(heap_rs), "_markBitMap2 inconsistency");
565
566 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
567 satb_qs.set_buffer_size(G1SATBBufferSize);
568
569 _root_regions.init(_g1h, this);
570
571 if (ConcGCThreads > ParallelGCThreads) {
572 warning("Can't have more ConcGCThreads (" UINTX_FORMAT ") "
573 "than ParallelGCThreads (" UINTX_FORMAT ").",
574 ConcGCThreads, ParallelGCThreads);
575 return;
576 }
577 if (ParallelGCThreads == 0) {
578 // if we are not running with any parallel GC threads we will not
579 // spawn any marking threads either
580 _parallel_marking_threads = 0;
581 _max_parallel_marking_threads = 0;
582 _sleep_factor = 0.0;
583 _marking_task_overhead = 1.0;
584 } else {
585 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) {
586 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent
587 // if both are set
588 _sleep_factor = 0.0;
589 _marking_task_overhead = 1.0;
590 } else if (G1MarkingOverheadPercent > 0) {
591 // We will calculate the number of parallel marking threads based
592 // on a target overhead with respect to the soft real-time goal
593 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0;
594 double overall_cm_overhead =
595 (double) MaxGCPauseMillis * marking_overhead /
596 (double) GCPauseIntervalMillis;
597 double cpu_ratio = 1.0 / (double) os::processor_count();
598 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio);
599 double marking_task_overhead =
600 overall_cm_overhead / marking_thread_num *
601 (double) os::processor_count();
602 double sleep_factor =
603 (1.0 - marking_task_overhead) / marking_task_overhead;
604
605 FLAG_SET_ERGO(uintx, ConcGCThreads, (uint) marking_thread_num);
606 _sleep_factor = sleep_factor;
607 _marking_task_overhead = marking_task_overhead;
608 } else {
609 // Calculate the number of parallel marking threads by scaling
610 // the number of parallel GC threads.
611 uint marking_thread_num = scale_parallel_threads((uint) ParallelGCThreads);
612 FLAG_SET_ERGO(uintx, ConcGCThreads, marking_thread_num);
613 _sleep_factor = 0.0;
614 _marking_task_overhead = 1.0;
615 }
616
617 assert(ConcGCThreads > 0, "Should have been set");
618 _parallel_marking_threads = (uint) ConcGCThreads;
619 _max_parallel_marking_threads = _parallel_marking_threads;
620
621 if (parallel_marking_threads() > 1) {
622 _cleanup_task_overhead = 1.0;
623 } else {
624 _cleanup_task_overhead = marking_task_overhead();
625 }
626 _cleanup_sleep_factor =
627 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead();
628
629 #if 0
630 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads());
631 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead());
632 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor());
633 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead());
634 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor());
635 #endif
636
637 guarantee(parallel_marking_threads() > 0, "peace of mind");
638 _parallel_workers = new FlexibleWorkGang("G1 Parallel Marking Threads",
639 _max_parallel_marking_threads, false, true);
640 if (_parallel_workers == NULL) {
641 vm_exit_during_initialization("Failed necessary allocation.");
642 } else {
643 _parallel_workers->initialize_workers();
644 }
645 }
646
647 if (FLAG_IS_DEFAULT(MarkStackSize)) {
648 uintx mark_stack_size =
649 MIN2(MarkStackSizeMax,
650 MAX2(MarkStackSize, (uintx) (parallel_marking_threads() * TASKQUEUE_SIZE)));
651 // Verify that the calculated value for MarkStackSize is in range.
652 // It would be nice to use the private utility routine from Arguments.
653 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) {
654 warning("Invalid value calculated for MarkStackSize (" UINTX_FORMAT "): "
655 "must be between " UINTX_FORMAT " and " UINTX_FORMAT,
656 mark_stack_size, (uintx) 1, MarkStackSizeMax);
657 return;
658 }
659 FLAG_SET_ERGO(uintx, MarkStackSize, mark_stack_size);
660 } else {
661 // Verify MarkStackSize is in range.
662 if (FLAG_IS_CMDLINE(MarkStackSize)) {
663 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) {
664 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
665 warning("Invalid value specified for MarkStackSize (" UINTX_FORMAT "): "
666 "must be between " UINTX_FORMAT " and " UINTX_FORMAT,
667 MarkStackSize, (uintx) 1, MarkStackSizeMax);
668 return;
669 }
670 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) {
671 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) {
672 warning("Invalid value specified for MarkStackSize (" UINTX_FORMAT ")"
673 " or for MarkStackSizeMax (" UINTX_FORMAT ")",
674 MarkStackSize, MarkStackSizeMax);
675 return;
676 }
677 }
678 }
679 }
680
681 if (!_markStack.allocate(MarkStackSize)) {
682 warning("Failed to allocate CM marking stack");
683 return;
684 }
685
686 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_worker_id, mtGC);
687 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC);
688
689 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC);
690 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC);
691
692 BitMap::idx_t card_bm_size = _card_bm.size();
693
694 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail
695 _active_tasks = _max_worker_id;
696
697 size_t max_regions = (size_t) _g1h->max_regions();
698 for (uint i = 0; i < _max_worker_id; ++i) {
699 CMTaskQueue* task_queue = new CMTaskQueue();
700 task_queue->initialize();
701 _task_queues->register_queue(i, task_queue);
702
703 _count_card_bitmaps[i] = BitMap(card_bm_size, false);
704 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC);
705
706 _tasks[i] = new CMTask(i, this,
707 _count_marked_bytes[i],
708 &_count_card_bitmaps[i],
709 task_queue, _task_queues);
710
711 _accum_task_vtime[i] = 0.0;
712 }
713
714 // Calculate the card number for the bottom of the heap. Used
715 // in biasing indexes into the accounting card bitmaps.
716 _heap_bottom_card_num =
717 intptr_t(uintptr_t(_g1h->reserved_region().start()) >>
718 CardTableModRefBS::card_shift);
719
720 // Clear all the liveness counting data
721 clear_all_count_data();
722
723 // so that the call below can read a sensible value
724 _heap_start = (HeapWord*) heap_rs.base();
725 set_non_marking_state();
726 _completed_initialization = true;
727 }
728
729 void ConcurrentMark::update_g1_committed(bool force) {
730 // If concurrent marking is not in progress, then we do not need to
731 // update _heap_end.
732 if (!concurrent_marking_in_progress() && !force) return;
733
734 MemRegion committed = _g1h->g1_committed();
735 assert(committed.start() == _heap_start, "start shouldn't change");
736 HeapWord* new_end = committed.end();
737 if (new_end > _heap_end) {
738 // The heap has been expanded.
739
740 _heap_end = new_end;
741 }
742 // Notice that the heap can also shrink. However, this only happens
743 // during a Full GC (at least currently) and the entire marking
744 // phase will bail out and the task will not be restarted. So, let's
745 // do nothing.
746 }
747
748 void ConcurrentMark::reset() {
749 // Starting values for these two. This should be called in a STW
750 // phase. CM will be notified of any future g1_committed expansions
751 // will be at the end of evacuation pauses, when tasks are
752 // inactive.
753 MemRegion committed = _g1h->g1_committed();
754 _heap_start = committed.start();
755 _heap_end = committed.end();
756
757 // Separated the asserts so that we know which one fires.
758 assert(_heap_start != NULL, "heap bounds should look ok");
759 assert(_heap_end != NULL, "heap bounds should look ok");
760 assert(_heap_start < _heap_end, "heap bounds should look ok");
761
762 // Reset all the marking data structures and any necessary flags
763 reset_marking_state();
764
765 if (verbose_low()) {
766 gclog_or_tty->print_cr("[global] resetting");
767 }
768
769 // We do reset all of them, since different phases will use
770 // different number of active threads. So, it's easiest to have all
771 // of them ready.
772 for (uint i = 0; i < _max_worker_id; ++i) {
773 _tasks[i]->reset(_nextMarkBitMap);
774 }
775
776 // we need this to make sure that the flag is on during the evac
777 // pause with initial mark piggy-backed
778 set_concurrent_marking_in_progress();
779 }
780
781
782 void ConcurrentMark::reset_marking_state(bool clear_overflow) {
783 _markStack.set_should_expand();
784 _markStack.setEmpty(); // Also clears the _markStack overflow flag
785 if (clear_overflow) {
786 clear_has_overflown();
787 } else {
788 assert(has_overflown(), "pre-condition");
789 }
790 _finger = _heap_start;
791
792 for (uint i = 0; i < _max_worker_id; ++i) {
793 CMTaskQueue* queue = _task_queues->queue(i);
794 queue->set_empty();
795 }
796 }
797
798 void ConcurrentMark::set_concurrency(uint active_tasks) {
799 assert(active_tasks <= _max_worker_id, "we should not have more");
800
801 _active_tasks = active_tasks;
802 // Need to update the three data structures below according to the
803 // number of active threads for this phase.
804 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues);
805 _first_overflow_barrier_sync.set_n_workers((int) active_tasks);
806 _second_overflow_barrier_sync.set_n_workers((int) active_tasks);
807 }
808
809 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) {
810 set_concurrency(active_tasks);
811
812 _concurrent = concurrent;
813 // We propagate this to all tasks, not just the active ones.
814 for (uint i = 0; i < _max_worker_id; ++i)
815 _tasks[i]->set_concurrent(concurrent);
816
817 if (concurrent) {
818 set_concurrent_marking_in_progress();
819 } else {
820 // We currently assume that the concurrent flag has been set to
821 // false before we start remark. At this point we should also be
822 // in a STW phase.
823 assert(!concurrent_marking_in_progress(), "invariant");
824 assert(out_of_regions(),
825 err_msg("only way to get here: _finger: "PTR_FORMAT", _heap_end: "PTR_FORMAT,
826 p2i(_finger), p2i(_heap_end)));
827 update_g1_committed(true);
828 }
829 }
830
831 void ConcurrentMark::set_non_marking_state() {
832 // We set the global marking state to some default values when we're
833 // not doing marking.
834 reset_marking_state();
835 _active_tasks = 0;
836 clear_concurrent_marking_in_progress();
837 }
838
839 ConcurrentMark::~ConcurrentMark() {
840 // The ConcurrentMark instance is never freed.
841 ShouldNotReachHere();
842 }
843
844 void ConcurrentMark::clearNextBitmap() {
845 G1CollectedHeap* g1h = G1CollectedHeap::heap();
846 G1CollectorPolicy* g1p = g1h->g1_policy();
847
848 // Make sure that the concurrent mark thread looks to still be in
849 // the current cycle.
850 guarantee(cmThread()->during_cycle(), "invariant");
851
852 // We are finishing up the current cycle by clearing the next
853 // marking bitmap and getting it ready for the next cycle. During
854 // this time no other cycle can start. So, let's make sure that this
855 // is the case.
856 guarantee(!g1h->mark_in_progress(), "invariant");
857
858 // clear the mark bitmap (no grey objects to start with).
859 // We need to do this in chunks and offer to yield in between
860 // each chunk.
861 HeapWord* start = _nextMarkBitMap->startWord();
862 HeapWord* end = _nextMarkBitMap->endWord();
863 HeapWord* cur = start;
864 size_t chunkSize = M;
865 while (cur < end) {
866 HeapWord* next = cur + chunkSize;
867 if (next > end) {
868 next = end;
869 }
870 MemRegion mr(cur,next);
871 _nextMarkBitMap->clearRange(mr);
872 cur = next;
873 do_yield_check();
874
875 // Repeat the asserts from above. We'll do them as asserts here to
876 // minimize their overhead on the product. However, we'll have
877 // them as guarantees at the beginning / end of the bitmap
878 // clearing to get some checking in the product.
879 assert(cmThread()->during_cycle(), "invariant");
880 assert(!g1h->mark_in_progress(), "invariant");
881 }
882
883 // Clear the liveness counting data
884 clear_all_count_data();
885
886 // Repeat the asserts from above.
887 guarantee(cmThread()->during_cycle(), "invariant");
888 guarantee(!g1h->mark_in_progress(), "invariant");
889 }
890
891 class NoteStartOfMarkHRClosure: public HeapRegionClosure {
892 public:
893 bool doHeapRegion(HeapRegion* r) {
894 if (!r->continuesHumongous()) {
895 r->note_start_of_marking();
896 }
897 return false;
898 }
899 };
900
901 void ConcurrentMark::checkpointRootsInitialPre() {
902 G1CollectedHeap* g1h = G1CollectedHeap::heap();
903 G1CollectorPolicy* g1p = g1h->g1_policy();
904
905 _has_aborted = false;
906
907 #ifndef PRODUCT
908 if (G1PrintReachableAtInitialMark) {
909 print_reachable("at-cycle-start",
910 VerifyOption_G1UsePrevMarking, true /* all */);
911 }
912 #endif
913
914 // Initialise marking structures. This has to be done in a STW phase.
915 reset();
916
917 // For each region note start of marking.
918 NoteStartOfMarkHRClosure startcl;
919 g1h->heap_region_iterate(&startcl);
920 }
921
922
923 void ConcurrentMark::checkpointRootsInitialPost() {
924 G1CollectedHeap* g1h = G1CollectedHeap::heap();
925
926 // If we force an overflow during remark, the remark operation will
927 // actually abort and we'll restart concurrent marking. If we always
928 // force an oveflow during remark we'll never actually complete the
929 // marking phase. So, we initilize this here, at the start of the
930 // cycle, so that at the remaining overflow number will decrease at
931 // every remark and we'll eventually not need to cause one.
932 force_overflow_stw()->init();
933
934 // Start Concurrent Marking weak-reference discovery.
935 ReferenceProcessor* rp = g1h->ref_processor_cm();
936 // enable ("weak") refs discovery
937 rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/);
938 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle
939
940 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
941 // This is the start of the marking cycle, we're expected all
942 // threads to have SATB queues with active set to false.
943 satb_mq_set.set_active_all_threads(true, /* new active value */
944 false /* expected_active */);
945
946 _root_regions.prepare_for_scan();
947
948 // update_g1_committed() will be called at the end of an evac pause
949 // when marking is on. So, it's also called at the end of the
950 // initial-mark pause to update the heap end, if the heap expands
951 // during it. No need to call it here.
952 }
953
954 /*
955 * Notice that in the next two methods, we actually leave the STS
956 * during the barrier sync and join it immediately afterwards. If we
957 * do not do this, the following deadlock can occur: one thread could
958 * be in the barrier sync code, waiting for the other thread to also
959 * sync up, whereas another one could be trying to yield, while also
960 * waiting for the other threads to sync up too.
961 *
962 * Note, however, that this code is also used during remark and in
963 * this case we should not attempt to leave / enter the STS, otherwise
964 * we'll either hit an asseert (debug / fastdebug) or deadlock
965 * (product). So we should only leave / enter the STS if we are
966 * operating concurrently.
967 *
968 * Because the thread that does the sync barrier has left the STS, it
969 * is possible to be suspended for a Full GC or an evacuation pause
970 * could occur. This is actually safe, since the entering the sync
971 * barrier is one of the last things do_marking_step() does, and it
972 * doesn't manipulate any data structures afterwards.
973 */
974
975 void ConcurrentMark::enter_first_sync_barrier(uint worker_id) {
976 if (verbose_low()) {
977 gclog_or_tty->print_cr("[%u] entering first barrier", worker_id);
978 }
979
980 if (concurrent()) {
981 SuspendibleThreadSet::leave();
982 }
983
984 bool barrier_aborted = !_first_overflow_barrier_sync.enter();
985
986 if (concurrent()) {
987 SuspendibleThreadSet::join();
988 }
989 // at this point everyone should have synced up and not be doing any
990 // more work
991
992 if (verbose_low()) {
993 if (barrier_aborted) {
994 gclog_or_tty->print_cr("[%u] aborted first barrier", worker_id);
995 } else {
996 gclog_or_tty->print_cr("[%u] leaving first barrier", worker_id);
997 }
998 }
999
1000 if (barrier_aborted) {
1001 // If the barrier aborted we ignore the overflow condition and
1002 // just abort the whole marking phase as quickly as possible.
1003 return;
1004 }
1005
1006 // If we're executing the concurrent phase of marking, reset the marking
1007 // state; otherwise the marking state is reset after reference processing,
1008 // during the remark pause.
1009 // If we reset here as a result of an overflow during the remark we will
1010 // see assertion failures from any subsequent set_concurrency_and_phase()
1011 // calls.
1012 if (concurrent()) {
1013 // let the task associated with with worker 0 do this
1014 if (worker_id == 0) {
1015 // task 0 is responsible for clearing the global data structures
1016 // We should be here because of an overflow. During STW we should
1017 // not clear the overflow flag since we rely on it being true when
1018 // we exit this method to abort the pause and restart concurent
1019 // marking.
1020 reset_marking_state(true /* clear_overflow */);
1021 force_overflow()->update();
1022
1023 if (G1Log::fine()) {
1024 gclog_or_tty->gclog_stamp(concurrent_gc_id());
1025 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]");
1026 }
1027 }
1028 }
1029
1030 // after this, each task should reset its own data structures then
1031 // then go into the second barrier
1032 }
1033
1034 void ConcurrentMark::enter_second_sync_barrier(uint worker_id) {
1035 if (verbose_low()) {
1036 gclog_or_tty->print_cr("[%u] entering second barrier", worker_id);
1037 }
1038
1039 if (concurrent()) {
1040 SuspendibleThreadSet::leave();
1041 }
1042
1043 bool barrier_aborted = !_second_overflow_barrier_sync.enter();
1044
1045 if (concurrent()) {
1046 SuspendibleThreadSet::join();
1047 }
1048 // at this point everything should be re-initialized and ready to go
1049
1050 if (verbose_low()) {
1051 if (barrier_aborted) {
1052 gclog_or_tty->print_cr("[%u] aborted second barrier", worker_id);
1053 } else {
1054 gclog_or_tty->print_cr("[%u] leaving second barrier", worker_id);
1055 }
1056 }
1057 }
1058
1059 #ifndef PRODUCT
1060 void ForceOverflowSettings::init() {
1061 _num_remaining = G1ConcMarkForceOverflow;
1062 _force = false;
1063 update();
1064 }
1065
1066 void ForceOverflowSettings::update() {
1067 if (_num_remaining > 0) {
1068 _num_remaining -= 1;
1069 _force = true;
1070 } else {
1071 _force = false;
1072 }
1073 }
1074
1075 bool ForceOverflowSettings::should_force() {
1076 if (_force) {
1077 _force = false;
1078 return true;
1079 } else {
1080 return false;
1081 }
1082 }
1083 #endif // !PRODUCT
1084
1085 class CMConcurrentMarkingTask: public AbstractGangTask {
1086 private:
1087 ConcurrentMark* _cm;
1088 ConcurrentMarkThread* _cmt;
1089
1090 public:
1091 void work(uint worker_id) {
1092 assert(Thread::current()->is_ConcurrentGC_thread(),
1093 "this should only be done by a conc GC thread");
1094 ResourceMark rm;
1095
1096 double start_vtime = os::elapsedVTime();
1097
1098 SuspendibleThreadSet::join();
1099
1100 assert(worker_id < _cm->active_tasks(), "invariant");
1101 CMTask* the_task = _cm->task(worker_id);
1102 the_task->record_start_time();
1103 if (!_cm->has_aborted()) {
1104 do {
1105 double start_vtime_sec = os::elapsedVTime();
1106 double start_time_sec = os::elapsedTime();
1107 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
1108
1109 the_task->do_marking_step(mark_step_duration_ms,
1110 true /* do_termination */,
1111 false /* is_serial*/);
1112
1113 double end_time_sec = os::elapsedTime();
1114 double end_vtime_sec = os::elapsedVTime();
1115 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec;
1116 double elapsed_time_sec = end_time_sec - start_time_sec;
1117 _cm->clear_has_overflown();
1118
1119 bool ret = _cm->do_yield_check(worker_id);
1120
1121 jlong sleep_time_ms;
1122 if (!_cm->has_aborted() && the_task->has_aborted()) {
1123 sleep_time_ms =
1124 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0);
1125 SuspendibleThreadSet::leave();
1126 os::sleep(Thread::current(), sleep_time_ms, false);
1127 SuspendibleThreadSet::join();
1128 }
1129 double end_time2_sec = os::elapsedTime();
1130 double elapsed_time2_sec = end_time2_sec - start_time_sec;
1131
1132 #if 0
1133 gclog_or_tty->print_cr("CM: elapsed %1.4lf ms, sleep %1.4lf ms, "
1134 "overhead %1.4lf",
1135 elapsed_vtime_sec * 1000.0, (double) sleep_time_ms,
1136 the_task->conc_overhead(os::elapsedTime()) * 8.0);
1137 gclog_or_tty->print_cr("elapsed time %1.4lf ms, time 2: %1.4lf ms",
1138 elapsed_time_sec * 1000.0, elapsed_time2_sec * 1000.0);
1139 #endif
1140 } while (!_cm->has_aborted() && the_task->has_aborted());
1141 }
1142 the_task->record_end_time();
1143 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant");
1144
1145 SuspendibleThreadSet::leave();
1146
1147 double end_vtime = os::elapsedVTime();
1148 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime);
1149 }
1150
1151 CMConcurrentMarkingTask(ConcurrentMark* cm,
1152 ConcurrentMarkThread* cmt) :
1153 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { }
1154
1155 ~CMConcurrentMarkingTask() { }
1156 };
1157
1158 // Calculates the number of active workers for a concurrent
1159 // phase.
1160 uint ConcurrentMark::calc_parallel_marking_threads() {
1161 if (G1CollectedHeap::use_parallel_gc_threads()) {
1162 uint n_conc_workers = 0;
1163 if (!UseDynamicNumberOfGCThreads ||
1164 (!FLAG_IS_DEFAULT(ConcGCThreads) &&
1165 !ForceDynamicNumberOfGCThreads)) {
1166 n_conc_workers = max_parallel_marking_threads();
1167 } else {
1168 n_conc_workers =
1169 AdaptiveSizePolicy::calc_default_active_workers(
1170 max_parallel_marking_threads(),
1171 1, /* Minimum workers */
1172 parallel_marking_threads(),
1173 Threads::number_of_non_daemon_threads());
1174 // Don't scale down "n_conc_workers" by scale_parallel_threads() because
1175 // that scaling has already gone into "_max_parallel_marking_threads".
1176 }
1177 assert(n_conc_workers > 0, "Always need at least 1");
1178 return n_conc_workers;
1179 }
1180 // If we are not running with any parallel GC threads we will not
1181 // have spawned any marking threads either. Hence the number of
1182 // concurrent workers should be 0.
1183 return 0;
1184 }
1185
1186 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) {
1187 // Currently, only survivors can be root regions.
1188 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant");
1189 G1RootRegionScanClosure cl(_g1h, this, worker_id);
1190
1191 const uintx interval = PrefetchScanIntervalInBytes;
1192 HeapWord* curr = hr->bottom();
1193 const HeapWord* end = hr->top();
1194 while (curr < end) {
1195 Prefetch::read(curr, interval);
1196 oop obj = oop(curr);
1197 int size = obj->oop_iterate(&cl);
1198 assert(size == obj->size(), "sanity");
1199 curr += size;
1200 }
1201 }
1202
1203 class CMRootRegionScanTask : public AbstractGangTask {
1204 private:
1205 ConcurrentMark* _cm;
1206
1207 public:
1208 CMRootRegionScanTask(ConcurrentMark* cm) :
1209 AbstractGangTask("Root Region Scan"), _cm(cm) { }
1210
1211 void work(uint worker_id) {
1212 assert(Thread::current()->is_ConcurrentGC_thread(),
1213 "this should only be done by a conc GC thread");
1214
1215 CMRootRegions* root_regions = _cm->root_regions();
1216 HeapRegion* hr = root_regions->claim_next();
1217 while (hr != NULL) {
1218 _cm->scanRootRegion(hr, worker_id);
1219 hr = root_regions->claim_next();
1220 }
1221 }
1222 };
1223
1224 void ConcurrentMark::scanRootRegions() {
1225 // scan_in_progress() will have been set to true only if there was
1226 // at least one root region to scan. So, if it's false, we
1227 // should not attempt to do any further work.
1228 if (root_regions()->scan_in_progress()) {
1229 _parallel_marking_threads = calc_parallel_marking_threads();
1230 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1231 "Maximum number of marking threads exceeded");
1232 uint active_workers = MAX2(1U, parallel_marking_threads());
1233
1234 CMRootRegionScanTask task(this);
1235 if (use_parallel_marking_threads()) {
1236 _parallel_workers->set_active_workers((int) active_workers);
1237 _parallel_workers->run_task(&task);
1238 } else {
1239 task.work(0);
1240 }
1241
1242 // It's possible that has_aborted() is true here without actually
1243 // aborting the survivor scan earlier. This is OK as it's
1244 // mainly used for sanity checking.
1245 root_regions()->scan_finished();
1246 }
1247 }
1248
1249 void ConcurrentMark::markFromRoots() {
1250 // we might be tempted to assert that:
1251 // assert(asynch == !SafepointSynchronize::is_at_safepoint(),
1252 // "inconsistent argument?");
1253 // However that wouldn't be right, because it's possible that
1254 // a safepoint is indeed in progress as a younger generation
1255 // stop-the-world GC happens even as we mark in this generation.
1256
1257 _restart_for_overflow = false;
1258 force_overflow_conc()->init();
1259
1260 // _g1h has _n_par_threads
1261 _parallel_marking_threads = calc_parallel_marking_threads();
1262 assert(parallel_marking_threads() <= max_parallel_marking_threads(),
1263 "Maximum number of marking threads exceeded");
1264
1265 uint active_workers = MAX2(1U, parallel_marking_threads());
1266
1267 // Parallel task terminator is set in "set_concurrency_and_phase()"
1268 set_concurrency_and_phase(active_workers, true /* concurrent */);
1269
1270 CMConcurrentMarkingTask markingTask(this, cmThread());
1271 if (use_parallel_marking_threads()) {
1272 _parallel_workers->set_active_workers((int)active_workers);
1273 // Don't set _n_par_threads because it affects MT in proceess_strong_roots()
1274 // and the decisions on that MT processing is made elsewhere.
1275 assert(_parallel_workers->active_workers() > 0, "Should have been set");
1276 _parallel_workers->run_task(&markingTask);
1277 } else {
1278 markingTask.work(0);
1279 }
1280 print_stats();
1281 }
1282
1283 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) {
1284 // world is stopped at this checkpoint
1285 assert(SafepointSynchronize::is_at_safepoint(),
1286 "world should be stopped");
1287
1288 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1289
1290 // If a full collection has happened, we shouldn't do this.
1291 if (has_aborted()) {
1292 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1293 return;
1294 }
1295
1296 SvcGCMarker sgcm(SvcGCMarker::OTHER);
1297
1298 if (VerifyDuringGC) {
1299 HandleMark hm; // handle scope
1300 Universe::heap()->prepare_for_verify();
1301 Universe::verify(VerifyOption_G1UsePrevMarking,
1302 " VerifyDuringGC:(before)");
1303 }
1304
1305 G1CollectorPolicy* g1p = g1h->g1_policy();
1306 g1p->record_concurrent_mark_remark_start();
1307
1308 double start = os::elapsedTime();
1309
1310 checkpointRootsFinalWork();
1311
1312 double mark_work_end = os::elapsedTime();
1313
1314 weakRefsWork(clear_all_soft_refs);
1315
1316 if (has_overflown()) {
1317 // Oops. We overflowed. Restart concurrent marking.
1318 _restart_for_overflow = true;
1319 if (G1TraceMarkStackOverflow) {
1320 gclog_or_tty->print_cr("\nRemark led to restart for overflow.");
1321 }
1322
1323 // Verify the heap w.r.t. the previous marking bitmap.
1324 if (VerifyDuringGC) {
1325 HandleMark hm; // handle scope
1326 Universe::heap()->prepare_for_verify();
1327 Universe::verify(VerifyOption_G1UsePrevMarking,
1328 " VerifyDuringGC:(overflow)");
1329 }
1330
1331 // Clear the marking state because we will be restarting
1332 // marking due to overflowing the global mark stack.
1333 reset_marking_state();
1334 } else {
1335 // Aggregate the per-task counting data that we have accumulated
1336 // while marking.
1337 aggregate_count_data();
1338
1339 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
1340 // We're done with marking.
1341 // This is the end of the marking cycle, we're expected all
1342 // threads to have SATB queues with active set to true.
1343 satb_mq_set.set_active_all_threads(false, /* new active value */
1344 true /* expected_active */);
1345
1346 if (VerifyDuringGC) {
1347 HandleMark hm; // handle scope
1348 Universe::heap()->prepare_for_verify();
1349 Universe::verify(VerifyOption_G1UseNextMarking,
1350 " VerifyDuringGC:(after)");
1351 }
1352 assert(!restart_for_overflow(), "sanity");
1353 // Completely reset the marking state since marking completed
1354 set_non_marking_state();
1355 }
1356
1357 // Expand the marking stack, if we have to and if we can.
1358 if (_markStack.should_expand()) {
1359 _markStack.expand();
1360 }
1361
1362 // Statistics
1363 double now = os::elapsedTime();
1364 _remark_mark_times.add((mark_work_end - start) * 1000.0);
1365 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0);
1366 _remark_times.add((now - start) * 1000.0);
1367
1368 g1p->record_concurrent_mark_remark_end();
1369
1370 G1CMIsAliveClosure is_alive(g1h);
1371 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive);
1372 }
1373
1374 // Base class of the closures that finalize and verify the
1375 // liveness counting data.
1376 class CMCountDataClosureBase: public HeapRegionClosure {
1377 protected:
1378 G1CollectedHeap* _g1h;
1379 ConcurrentMark* _cm;
1380 CardTableModRefBS* _ct_bs;
1381
1382 BitMap* _region_bm;
1383 BitMap* _card_bm;
1384
1385 // Takes a region that's not empty (i.e., it has at least one
1386 // live object in it and sets its corresponding bit on the region
1387 // bitmap to 1. If the region is "starts humongous" it will also set
1388 // to 1 the bits on the region bitmap that correspond to its
1389 // associated "continues humongous" regions.
1390 void set_bit_for_region(HeapRegion* hr) {
1391 assert(!hr->continuesHumongous(), "should have filtered those out");
1392
1393 BitMap::idx_t index = (BitMap::idx_t) hr->hrs_index();
1394 if (!hr->startsHumongous()) {
1395 // Normal (non-humongous) case: just set the bit.
1396 _region_bm->par_at_put(index, true);
1397 } else {
1398 // Starts humongous case: calculate how many regions are part of
1399 // this humongous region and then set the bit range.
1400 BitMap::idx_t end_index = (BitMap::idx_t) hr->last_hc_index();
1401 _region_bm->par_at_put_range(index, end_index, true);
1402 }
1403 }
1404
1405 public:
1406 CMCountDataClosureBase(G1CollectedHeap* g1h,
1407 BitMap* region_bm, BitMap* card_bm):
1408 _g1h(g1h), _cm(g1h->concurrent_mark()),
1409 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())),
1410 _region_bm(region_bm), _card_bm(card_bm) { }
1411 };
1412
1413 // Closure that calculates the # live objects per region. Used
1414 // for verification purposes during the cleanup pause.
1415 class CalcLiveObjectsClosure: public CMCountDataClosureBase {
1416 CMBitMapRO* _bm;
1417 size_t _region_marked_bytes;
1418
1419 public:
1420 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h,
1421 BitMap* region_bm, BitMap* card_bm) :
1422 CMCountDataClosureBase(g1h, region_bm, card_bm),
1423 _bm(bm), _region_marked_bytes(0) { }
1424
1425 bool doHeapRegion(HeapRegion* hr) {
1426
1427 if (hr->continuesHumongous()) {
1428 // We will ignore these here and process them when their
1429 // associated "starts humongous" region is processed (see
1430 // set_bit_for_heap_region()). Note that we cannot rely on their
1431 // associated "starts humongous" region to have their bit set to
1432 // 1 since, due to the region chunking in the parallel region
1433 // iteration, a "continues humongous" region might be visited
1434 // before its associated "starts humongous".
1435 return false;
1436 }
1437
1438 HeapWord* ntams = hr->next_top_at_mark_start();
1439 HeapWord* start = hr->bottom();
1440
1441 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(),
1442 err_msg("Preconditions not met - "
1443 "start: "PTR_FORMAT", ntams: "PTR_FORMAT", end: "PTR_FORMAT,
1444 p2i(start), p2i(ntams), p2i(hr->end())));
1445
1446 // Find the first marked object at or after "start".
1447 start = _bm->getNextMarkedWordAddress(start, ntams);
1448
1449 size_t marked_bytes = 0;
1450
1451 while (start < ntams) {
1452 oop obj = oop(start);
1453 int obj_sz = obj->size();
1454 HeapWord* obj_end = start + obj_sz;
1455
1456 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
1457 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end);
1458
1459 // Note: if we're looking at the last region in heap - obj_end
1460 // could be actually just beyond the end of the heap; end_idx
1461 // will then correspond to a (non-existent) card that is also
1462 // just beyond the heap.
1463 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) {
1464 // end of object is not card aligned - increment to cover
1465 // all the cards spanned by the object
1466 end_idx += 1;
1467 }
1468
1469 // Set the bits in the card BM for the cards spanned by this object.
1470 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1471
1472 // Add the size of this object to the number of marked bytes.
1473 marked_bytes += (size_t)obj_sz * HeapWordSize;
1474
1475 // Find the next marked object after this one.
1476 start = _bm->getNextMarkedWordAddress(obj_end, ntams);
1477 }
1478
1479 // Mark the allocated-since-marking portion...
1480 HeapWord* top = hr->top();
1481 if (ntams < top) {
1482 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1483 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1484
1485 // Note: if we're looking at the last region in heap - top
1486 // could be actually just beyond the end of the heap; end_idx
1487 // will then correspond to a (non-existent) card that is also
1488 // just beyond the heap.
1489 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1490 // end of object is not card aligned - increment to cover
1491 // all the cards spanned by the object
1492 end_idx += 1;
1493 }
1494 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1495
1496 // This definitely means the region has live objects.
1497 set_bit_for_region(hr);
1498 }
1499
1500 // Update the live region bitmap.
1501 if (marked_bytes > 0) {
1502 set_bit_for_region(hr);
1503 }
1504
1505 // Set the marked bytes for the current region so that
1506 // it can be queried by a calling verificiation routine
1507 _region_marked_bytes = marked_bytes;
1508
1509 return false;
1510 }
1511
1512 size_t region_marked_bytes() const { return _region_marked_bytes; }
1513 };
1514
1515 // Heap region closure used for verifying the counting data
1516 // that was accumulated concurrently and aggregated during
1517 // the remark pause. This closure is applied to the heap
1518 // regions during the STW cleanup pause.
1519
1520 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure {
1521 G1CollectedHeap* _g1h;
1522 ConcurrentMark* _cm;
1523 CalcLiveObjectsClosure _calc_cl;
1524 BitMap* _region_bm; // Region BM to be verified
1525 BitMap* _card_bm; // Card BM to be verified
1526 bool _verbose; // verbose output?
1527
1528 BitMap* _exp_region_bm; // Expected Region BM values
1529 BitMap* _exp_card_bm; // Expected card BM values
1530
1531 int _failures;
1532
1533 public:
1534 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h,
1535 BitMap* region_bm,
1536 BitMap* card_bm,
1537 BitMap* exp_region_bm,
1538 BitMap* exp_card_bm,
1539 bool verbose) :
1540 _g1h(g1h), _cm(g1h->concurrent_mark()),
1541 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm),
1542 _region_bm(region_bm), _card_bm(card_bm), _verbose(verbose),
1543 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm),
1544 _failures(0) { }
1545
1546 int failures() const { return _failures; }
1547
1548 bool doHeapRegion(HeapRegion* hr) {
1549 if (hr->continuesHumongous()) {
1550 // We will ignore these here and process them when their
1551 // associated "starts humongous" region is processed (see
1552 // set_bit_for_heap_region()). Note that we cannot rely on their
1553 // associated "starts humongous" region to have their bit set to
1554 // 1 since, due to the region chunking in the parallel region
1555 // iteration, a "continues humongous" region might be visited
1556 // before its associated "starts humongous".
1557 return false;
1558 }
1559
1560 int failures = 0;
1561
1562 // Call the CalcLiveObjectsClosure to walk the marking bitmap for
1563 // this region and set the corresponding bits in the expected region
1564 // and card bitmaps.
1565 bool res = _calc_cl.doHeapRegion(hr);
1566 assert(res == false, "should be continuing");
1567
1568 MutexLockerEx x((_verbose ? ParGCRareEvent_lock : NULL),
1569 Mutex::_no_safepoint_check_flag);
1570
1571 // Verify the marked bytes for this region.
1572 size_t exp_marked_bytes = _calc_cl.region_marked_bytes();
1573 size_t act_marked_bytes = hr->next_marked_bytes();
1574
1575 // We're not OK if expected marked bytes > actual marked bytes. It means
1576 // we have missed accounting some objects during the actual marking.
1577 if (exp_marked_bytes > act_marked_bytes) {
1578 if (_verbose) {
1579 gclog_or_tty->print_cr("Region %u: marked bytes mismatch: "
1580 "expected: " SIZE_FORMAT ", actual: " SIZE_FORMAT,
1581 hr->hrs_index(), exp_marked_bytes, act_marked_bytes);
1582 }
1583 failures += 1;
1584 }
1585
1586 // Verify the bit, for this region, in the actual and expected
1587 // (which was just calculated) region bit maps.
1588 // We're not OK if the bit in the calculated expected region
1589 // bitmap is set and the bit in the actual region bitmap is not.
1590 BitMap::idx_t index = (BitMap::idx_t) hr->hrs_index();
1591
1592 bool expected = _exp_region_bm->at(index);
1593 bool actual = _region_bm->at(index);
1594 if (expected && !actual) {
1595 if (_verbose) {
1596 gclog_or_tty->print_cr("Region %u: region bitmap mismatch: "
1597 "expected: %s, actual: %s",
1598 hr->hrs_index(),
1599 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1600 }
1601 failures += 1;
1602 }
1603
1604 // Verify that the card bit maps for the cards spanned by the current
1605 // region match. We have an error if we have a set bit in the expected
1606 // bit map and the corresponding bit in the actual bitmap is not set.
1607
1608 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom());
1609 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top());
1610
1611 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) {
1612 expected = _exp_card_bm->at(i);
1613 actual = _card_bm->at(i);
1614
1615 if (expected && !actual) {
1616 if (_verbose) {
1617 gclog_or_tty->print_cr("Region %u: card bitmap mismatch at " SIZE_FORMAT ": "
1618 "expected: %s, actual: %s",
1619 hr->hrs_index(), i,
1620 BOOL_TO_STR(expected), BOOL_TO_STR(actual));
1621 }
1622 failures += 1;
1623 }
1624 }
1625
1626 if (failures > 0 && _verbose) {
1627 gclog_or_tty->print_cr("Region " HR_FORMAT ", ntams: " PTR_FORMAT ", "
1628 "marked_bytes: calc/actual " SIZE_FORMAT "/" SIZE_FORMAT,
1629 HR_FORMAT_PARAMS(hr), p2i(hr->next_top_at_mark_start()),
1630 _calc_cl.region_marked_bytes(), hr->next_marked_bytes());
1631 }
1632
1633 _failures += failures;
1634
1635 // We could stop iteration over the heap when we
1636 // find the first violating region by returning true.
1637 return false;
1638 }
1639 };
1640
1641 class G1ParVerifyFinalCountTask: public AbstractGangTask {
1642 protected:
1643 G1CollectedHeap* _g1h;
1644 ConcurrentMark* _cm;
1645 BitMap* _actual_region_bm;
1646 BitMap* _actual_card_bm;
1647
1648 uint _n_workers;
1649
1650 BitMap* _expected_region_bm;
1651 BitMap* _expected_card_bm;
1652
1653 int _failures;
1654 bool _verbose;
1655
1656 public:
1657 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h,
1658 BitMap* region_bm, BitMap* card_bm,
1659 BitMap* expected_region_bm, BitMap* expected_card_bm)
1660 : AbstractGangTask("G1 verify final counting"),
1661 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1662 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1663 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm),
1664 _failures(0), _verbose(false),
1665 _n_workers(0) {
1666 assert(VerifyDuringGC, "don't call this otherwise");
1667
1668 // Use the value already set as the number of active threads
1669 // in the call to run_task().
1670 if (G1CollectedHeap::use_parallel_gc_threads()) {
1671 assert( _g1h->workers()->active_workers() > 0,
1672 "Should have been previously set");
1673 _n_workers = _g1h->workers()->active_workers();
1674 } else {
1675 _n_workers = 1;
1676 }
1677
1678 assert(_expected_card_bm->size() == _actual_card_bm->size(), "sanity");
1679 assert(_expected_region_bm->size() == _actual_region_bm->size(), "sanity");
1680
1681 _verbose = _cm->verbose_medium();
1682 }
1683
1684 void work(uint worker_id) {
1685 assert(worker_id < _n_workers, "invariant");
1686
1687 VerifyLiveObjectDataHRClosure verify_cl(_g1h,
1688 _actual_region_bm, _actual_card_bm,
1689 _expected_region_bm,
1690 _expected_card_bm,
1691 _verbose);
1692
1693 if (G1CollectedHeap::use_parallel_gc_threads()) {
1694 _g1h->heap_region_par_iterate_chunked(&verify_cl,
1695 worker_id,
1696 _n_workers,
1697 HeapRegion::VerifyCountClaimValue);
1698 } else {
1699 _g1h->heap_region_iterate(&verify_cl);
1700 }
1701
1702 Atomic::add(verify_cl.failures(), &_failures);
1703 }
1704
1705 int failures() const { return _failures; }
1706 };
1707
1708 // Closure that finalizes the liveness counting data.
1709 // Used during the cleanup pause.
1710 // Sets the bits corresponding to the interval [NTAMS, top]
1711 // (which contains the implicitly live objects) in the
1712 // card liveness bitmap. Also sets the bit for each region,
1713 // containing live data, in the region liveness bitmap.
1714
1715 class FinalCountDataUpdateClosure: public CMCountDataClosureBase {
1716 public:
1717 FinalCountDataUpdateClosure(G1CollectedHeap* g1h,
1718 BitMap* region_bm,
1719 BitMap* card_bm) :
1720 CMCountDataClosureBase(g1h, region_bm, card_bm) { }
1721
1722 bool doHeapRegion(HeapRegion* hr) {
1723
1724 if (hr->continuesHumongous()) {
1725 // We will ignore these here and process them when their
1726 // associated "starts humongous" region is processed (see
1727 // set_bit_for_heap_region()). Note that we cannot rely on their
1728 // associated "starts humongous" region to have their bit set to
1729 // 1 since, due to the region chunking in the parallel region
1730 // iteration, a "continues humongous" region might be visited
1731 // before its associated "starts humongous".
1732 return false;
1733 }
1734
1735 HeapWord* ntams = hr->next_top_at_mark_start();
1736 HeapWord* top = hr->top();
1737
1738 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions.");
1739
1740 // Mark the allocated-since-marking portion...
1741 if (ntams < top) {
1742 // This definitely means the region has live objects.
1743 set_bit_for_region(hr);
1744
1745 // Now set the bits in the card bitmap for [ntams, top)
1746 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams);
1747 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top);
1748
1749 // Note: if we're looking at the last region in heap - top
1750 // could be actually just beyond the end of the heap; end_idx
1751 // will then correspond to a (non-existent) card that is also
1752 // just beyond the heap.
1753 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) {
1754 // end of object is not card aligned - increment to cover
1755 // all the cards spanned by the object
1756 end_idx += 1;
1757 }
1758
1759 assert(end_idx <= _card_bm->size(),
1760 err_msg("oob: end_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1761 end_idx, _card_bm->size()));
1762 assert(start_idx < _card_bm->size(),
1763 err_msg("oob: start_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT,
1764 start_idx, _card_bm->size()));
1765
1766 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */);
1767 }
1768
1769 // Set the bit for the region if it contains live data
1770 if (hr->next_marked_bytes() > 0) {
1771 set_bit_for_region(hr);
1772 }
1773
1774 return false;
1775 }
1776 };
1777
1778 class G1ParFinalCountTask: public AbstractGangTask {
1779 protected:
1780 G1CollectedHeap* _g1h;
1781 ConcurrentMark* _cm;
1782 BitMap* _actual_region_bm;
1783 BitMap* _actual_card_bm;
1784
1785 uint _n_workers;
1786
1787 public:
1788 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm)
1789 : AbstractGangTask("G1 final counting"),
1790 _g1h(g1h), _cm(_g1h->concurrent_mark()),
1791 _actual_region_bm(region_bm), _actual_card_bm(card_bm),
1792 _n_workers(0) {
1793 // Use the value already set as the number of active threads
1794 // in the call to run_task().
1795 if (G1CollectedHeap::use_parallel_gc_threads()) {
1796 assert( _g1h->workers()->active_workers() > 0,
1797 "Should have been previously set");
1798 _n_workers = _g1h->workers()->active_workers();
1799 } else {
1800 _n_workers = 1;
1801 }
1802 }
1803
1804 void work(uint worker_id) {
1805 assert(worker_id < _n_workers, "invariant");
1806
1807 FinalCountDataUpdateClosure final_update_cl(_g1h,
1808 _actual_region_bm,
1809 _actual_card_bm);
1810
1811 if (G1CollectedHeap::use_parallel_gc_threads()) {
1812 _g1h->heap_region_par_iterate_chunked(&final_update_cl,
1813 worker_id,
1814 _n_workers,
1815 HeapRegion::FinalCountClaimValue);
1816 } else {
1817 _g1h->heap_region_iterate(&final_update_cl);
1818 }
1819 }
1820 };
1821
1822 class G1ParNoteEndTask;
1823
1824 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure {
1825 G1CollectedHeap* _g1;
1826 size_t _max_live_bytes;
1827 uint _regions_claimed;
1828 size_t _freed_bytes;
1829 FreeRegionList* _local_cleanup_list;
1830 HeapRegionSetCount _old_regions_removed;
1831 HeapRegionSetCount _humongous_regions_removed;
1832 HRRSCleanupTask* _hrrs_cleanup_task;
1833 double _claimed_region_time;
1834 double _max_region_time;
1835
1836 public:
1837 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1,
1838 FreeRegionList* local_cleanup_list,
1839 HRRSCleanupTask* hrrs_cleanup_task) :
1840 _g1(g1),
1841 _max_live_bytes(0), _regions_claimed(0),
1842 _freed_bytes(0),
1843 _claimed_region_time(0.0), _max_region_time(0.0),
1844 _local_cleanup_list(local_cleanup_list),
1845 _old_regions_removed(),
1846 _humongous_regions_removed(),
1847 _hrrs_cleanup_task(hrrs_cleanup_task) { }
1848
1849 size_t freed_bytes() { return _freed_bytes; }
1850 const HeapRegionSetCount& old_regions_removed() { return _old_regions_removed; }
1851 const HeapRegionSetCount& humongous_regions_removed() { return _humongous_regions_removed; }
1852
1853 bool doHeapRegion(HeapRegion *hr) {
1854 if (hr->continuesHumongous()) {
1855 return false;
1856 }
1857 // We use a claim value of zero here because all regions
1858 // were claimed with value 1 in the FinalCount task.
1859 _g1->reset_gc_time_stamps(hr);
1860 double start = os::elapsedTime();
1861 _regions_claimed++;
1862 hr->note_end_of_marking();
1863 _max_live_bytes += hr->max_live_bytes();
1864
1865 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) {
1866 _freed_bytes += hr->used();
1867 hr->set_containing_set(NULL);
1868 if (hr->isHumongous()) {
1869 assert(hr->startsHumongous(), "we should only see starts humongous");
1870 _humongous_regions_removed.increment(1u, hr->capacity());
1871 _g1->free_humongous_region(hr, _local_cleanup_list, true);
1872 } else {
1873 _old_regions_removed.increment(1u, hr->capacity());
1874 _g1->free_region(hr, _local_cleanup_list, true);
1875 }
1876 } else {
1877 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task);
1878 }
1879
1880 double region_time = (os::elapsedTime() - start);
1881 _claimed_region_time += region_time;
1882 if (region_time > _max_region_time) {
1883 _max_region_time = region_time;
1884 }
1885 return false;
1886 }
1887
1888 size_t max_live_bytes() { return _max_live_bytes; }
1889 uint regions_claimed() { return _regions_claimed; }
1890 double claimed_region_time_sec() { return _claimed_region_time; }
1891 double max_region_time_sec() { return _max_region_time; }
1892 };
1893
1894 class G1ParNoteEndTask: public AbstractGangTask {
1895 friend class G1NoteEndOfConcMarkClosure;
1896
1897 protected:
1898 G1CollectedHeap* _g1h;
1899 size_t _max_live_bytes;
1900 size_t _freed_bytes;
1901 FreeRegionList* _cleanup_list;
1902
1903 public:
1904 G1ParNoteEndTask(G1CollectedHeap* g1h,
1905 FreeRegionList* cleanup_list) :
1906 AbstractGangTask("G1 note end"), _g1h(g1h),
1907 _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list) { }
1908
1909 void work(uint worker_id) {
1910 double start = os::elapsedTime();
1911 FreeRegionList local_cleanup_list("Local Cleanup List");
1912 HRRSCleanupTask hrrs_cleanup_task;
1913 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list,
1914 &hrrs_cleanup_task);
1915 if (G1CollectedHeap::use_parallel_gc_threads()) {
1916 _g1h->heap_region_par_iterate_chunked(&g1_note_end, worker_id,
1917 _g1h->workers()->active_workers(),
1918 HeapRegion::NoteEndClaimValue);
1919 } else {
1920 _g1h->heap_region_iterate(&g1_note_end);
1921 }
1922 assert(g1_note_end.complete(), "Shouldn't have yielded!");
1923
1924 // Now update the lists
1925 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed());
1926 {
1927 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag);
1928 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes());
1929 _max_live_bytes += g1_note_end.max_live_bytes();
1930 _freed_bytes += g1_note_end.freed_bytes();
1931
1932 // If we iterate over the global cleanup list at the end of
1933 // cleanup to do this printing we will not guarantee to only
1934 // generate output for the newly-reclaimed regions (the list
1935 // might not be empty at the beginning of cleanup; we might
1936 // still be working on its previous contents). So we do the
1937 // printing here, before we append the new regions to the global
1938 // cleanup list.
1939
1940 G1HRPrinter* hr_printer = _g1h->hr_printer();
1941 if (hr_printer->is_active()) {
1942 FreeRegionListIterator iter(&local_cleanup_list);
1943 while (iter.more_available()) {
1944 HeapRegion* hr = iter.get_next();
1945 hr_printer->cleanup(hr);
1946 }
1947 }
1948
1949 _cleanup_list->add_ordered(&local_cleanup_list);
1950 assert(local_cleanup_list.is_empty(), "post-condition");
1951
1952 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task);
1953 }
1954 }
1955 size_t max_live_bytes() { return _max_live_bytes; }
1956 size_t freed_bytes() { return _freed_bytes; }
1957 };
1958
1959 class G1ParScrubRemSetTask: public AbstractGangTask {
1960 protected:
1961 G1RemSet* _g1rs;
1962 BitMap* _region_bm;
1963 BitMap* _card_bm;
1964 public:
1965 G1ParScrubRemSetTask(G1CollectedHeap* g1h,
1966 BitMap* region_bm, BitMap* card_bm) :
1967 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()),
1968 _region_bm(region_bm), _card_bm(card_bm) { }
1969
1970 void work(uint worker_id) {
1971 if (G1CollectedHeap::use_parallel_gc_threads()) {
1972 _g1rs->scrub_par(_region_bm, _card_bm, worker_id,
1973 HeapRegion::ScrubRemSetClaimValue);
1974 } else {
1975 _g1rs->scrub(_region_bm, _card_bm);
1976 }
1977 }
1978
1979 };
1980
1981 void ConcurrentMark::cleanup() {
1982 // world is stopped at this checkpoint
1983 assert(SafepointSynchronize::is_at_safepoint(),
1984 "world should be stopped");
1985 G1CollectedHeap* g1h = G1CollectedHeap::heap();
1986
1987 // If a full collection has happened, we shouldn't do this.
1988 if (has_aborted()) {
1989 g1h->set_marking_complete(); // So bitmap clearing isn't confused
1990 return;
1991 }
1992
1993 g1h->verify_region_sets_optional();
1994
1995 if (VerifyDuringGC) {
1996 HandleMark hm; // handle scope
1997 Universe::heap()->prepare_for_verify();
1998 Universe::verify(VerifyOption_G1UsePrevMarking,
1999 " VerifyDuringGC:(before)");
2000 }
2001
2002 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy();
2003 g1p->record_concurrent_mark_cleanup_start();
2004
2005 double start = os::elapsedTime();
2006
2007 HeapRegionRemSet::reset_for_cleanup_tasks();
2008
2009 uint n_workers;
2010
2011 // Do counting once more with the world stopped for good measure.
2012 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm);
2013
2014 if (G1CollectedHeap::use_parallel_gc_threads()) {
2015 assert(g1h->check_heap_region_claim_values(HeapRegion::InitialClaimValue),
2016 "sanity check");
2017
2018 g1h->set_par_threads();
2019 n_workers = g1h->n_par_threads();
2020 assert(g1h->n_par_threads() == n_workers,
2021 "Should not have been reset");
2022 g1h->workers()->run_task(&g1_par_count_task);
2023 // Done with the parallel phase so reset to 0.
2024 g1h->set_par_threads(0);
2025
2026 assert(g1h->check_heap_region_claim_values(HeapRegion::FinalCountClaimValue),
2027 "sanity check");
2028 } else {
2029 n_workers = 1;
2030 g1_par_count_task.work(0);
2031 }
2032
2033 if (VerifyDuringGC) {
2034 // Verify that the counting data accumulated during marking matches
2035 // that calculated by walking the marking bitmap.
2036
2037 // Bitmaps to hold expected values
2038 BitMap expected_region_bm(_region_bm.size(), false);
2039 BitMap expected_card_bm(_card_bm.size(), false);
2040
2041 G1ParVerifyFinalCountTask g1_par_verify_task(g1h,
2042 &_region_bm,
2043 &_card_bm,
2044 &expected_region_bm,
2045 &expected_card_bm);
2046
2047 if (G1CollectedHeap::use_parallel_gc_threads()) {
2048 g1h->set_par_threads((int)n_workers);
2049 g1h->workers()->run_task(&g1_par_verify_task);
2050 // Done with the parallel phase so reset to 0.
2051 g1h->set_par_threads(0);
2052
2053 assert(g1h->check_heap_region_claim_values(HeapRegion::VerifyCountClaimValue),
2054 "sanity check");
2055 } else {
2056 g1_par_verify_task.work(0);
2057 }
2058
2059 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures");
2060 }
2061
2062 size_t start_used_bytes = g1h->used();
2063 g1h->set_marking_complete();
2064
2065 double count_end = os::elapsedTime();
2066 double this_final_counting_time = (count_end - start);
2067 _total_counting_time += this_final_counting_time;
2068
2069 if (G1PrintRegionLivenessInfo) {
2070 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking");
2071 _g1h->heap_region_iterate(&cl);
2072 }
2073
2074 // Install newly created mark bitMap as "prev".
2075 swapMarkBitMaps();
2076
2077 g1h->reset_gc_time_stamp();
2078
2079 // Note end of marking in all heap regions.
2080 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list);
2081 if (G1CollectedHeap::use_parallel_gc_threads()) {
2082 g1h->set_par_threads((int)n_workers);
2083 g1h->workers()->run_task(&g1_par_note_end_task);
2084 g1h->set_par_threads(0);
2085
2086 assert(g1h->check_heap_region_claim_values(HeapRegion::NoteEndClaimValue),
2087 "sanity check");
2088 } else {
2089 g1_par_note_end_task.work(0);
2090 }
2091 g1h->check_gc_time_stamps();
2092
2093 if (!cleanup_list_is_empty()) {
2094 // The cleanup list is not empty, so we'll have to process it
2095 // concurrently. Notify anyone else that might be wanting free
2096 // regions that there will be more free regions coming soon.
2097 g1h->set_free_regions_coming();
2098 }
2099
2100 // call below, since it affects the metric by which we sort the heap
2101 // regions.
2102 if (G1ScrubRemSets) {
2103 double rs_scrub_start = os::elapsedTime();
2104 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm);
2105 if (G1CollectedHeap::use_parallel_gc_threads()) {
2106 g1h->set_par_threads((int)n_workers);
2107 g1h->workers()->run_task(&g1_par_scrub_rs_task);
2108 g1h->set_par_threads(0);
2109
2110 assert(g1h->check_heap_region_claim_values(
2111 HeapRegion::ScrubRemSetClaimValue),
2112 "sanity check");
2113 } else {
2114 g1_par_scrub_rs_task.work(0);
2115 }
2116
2117 double rs_scrub_end = os::elapsedTime();
2118 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start);
2119 _total_rs_scrub_time += this_rs_scrub_time;
2120 }
2121
2122 // this will also free any regions totally full of garbage objects,
2123 // and sort the regions.
2124 g1h->g1_policy()->record_concurrent_mark_cleanup_end((int)n_workers);
2125
2126 // Statistics.
2127 double end = os::elapsedTime();
2128 _cleanup_times.add((end - start) * 1000.0);
2129
2130 if (G1Log::fine()) {
2131 g1h->print_size_transition(gclog_or_tty,
2132 start_used_bytes,
2133 g1h->used(),
2134 g1h->capacity());
2135 }
2136
2137 // Clean up will have freed any regions completely full of garbage.
2138 // Update the soft reference policy with the new heap occupancy.
2139 Universe::update_heap_info_at_gc();
2140
2141 // We need to make this be a "collection" so any collection pause that
2142 // races with it goes around and waits for completeCleanup to finish.
2143 g1h->increment_total_collections();
2144
2145 // We reclaimed old regions so we should calculate the sizes to make
2146 // sure we update the old gen/space data.
2147 g1h->g1mm()->update_sizes();
2148
2149 if (VerifyDuringGC) {
2150 HandleMark hm; // handle scope
2151 Universe::heap()->prepare_for_verify();
2152 Universe::verify(VerifyOption_G1UsePrevMarking,
2153 " VerifyDuringGC:(after)");
2154 }
2155
2156 g1h->verify_region_sets_optional();
2157 g1h->trace_heap_after_concurrent_cycle();
2158 }
2159
2160 void ConcurrentMark::completeCleanup() {
2161 if (has_aborted()) return;
2162
2163 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2164
2165 _cleanup_list.verify_optional();
2166 FreeRegionList tmp_free_list("Tmp Free List");
2167
2168 if (G1ConcRegionFreeingVerbose) {
2169 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2170 "cleanup list has %u entries",
2171 _cleanup_list.length());
2172 }
2173
2174 // Noone else should be accessing the _cleanup_list at this point,
2175 // so it's not necessary to take any locks
2176 while (!_cleanup_list.is_empty()) {
2177 HeapRegion* hr = _cleanup_list.remove_head();
2178 assert(hr != NULL, "Got NULL from a non-empty list");
2179 hr->par_clear();
2180 tmp_free_list.add_ordered(hr);
2181
2182 // Instead of adding one region at a time to the secondary_free_list,
2183 // we accumulate them in the local list and move them a few at a
2184 // time. This also cuts down on the number of notify_all() calls
2185 // we do during this process. We'll also append the local list when
2186 // _cleanup_list is empty (which means we just removed the last
2187 // region from the _cleanup_list).
2188 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) ||
2189 _cleanup_list.is_empty()) {
2190 if (G1ConcRegionFreeingVerbose) {
2191 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : "
2192 "appending %u entries to the secondary_free_list, "
2193 "cleanup list still has %u entries",
2194 tmp_free_list.length(),
2195 _cleanup_list.length());
2196 }
2197
2198 {
2199 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag);
2200 g1h->secondary_free_list_add(&tmp_free_list);
2201 SecondaryFreeList_lock->notify_all();
2202 }
2203
2204 if (G1StressConcRegionFreeing) {
2205 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) {
2206 os::sleep(Thread::current(), (jlong) 1, false);
2207 }
2208 }
2209 }
2210 }
2211 assert(tmp_free_list.is_empty(), "post-condition");
2212 }
2213
2214 // Supporting Object and Oop closures for reference discovery
2215 // and processing in during marking
2216
2217 bool G1CMIsAliveClosure::do_object_b(oop obj) {
2218 HeapWord* addr = (HeapWord*)obj;
2219 return addr != NULL &&
2220 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj));
2221 }
2222
2223 // 'Keep Alive' oop closure used by both serial parallel reference processing.
2224 // Uses the CMTask associated with a worker thread (for serial reference
2225 // processing the CMTask for worker 0 is used) to preserve (mark) and
2226 // trace referent objects.
2227 //
2228 // Using the CMTask and embedded local queues avoids having the worker
2229 // threads operating on the global mark stack. This reduces the risk
2230 // of overflowing the stack - which we would rather avoid at this late
2231 // state. Also using the tasks' local queues removes the potential
2232 // of the workers interfering with each other that could occur if
2233 // operating on the global stack.
2234
2235 class G1CMKeepAliveAndDrainClosure: public OopClosure {
2236 ConcurrentMark* _cm;
2237 CMTask* _task;
2238 int _ref_counter_limit;
2239 int _ref_counter;
2240 bool _is_serial;
2241 public:
2242 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2243 _cm(cm), _task(task), _is_serial(is_serial),
2244 _ref_counter_limit(G1RefProcDrainInterval) {
2245 assert(_ref_counter_limit > 0, "sanity");
2246 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
2247 _ref_counter = _ref_counter_limit;
2248 }
2249
2250 virtual void do_oop(narrowOop* p) { do_oop_work(p); }
2251 virtual void do_oop( oop* p) { do_oop_work(p); }
2252
2253 template <class T> void do_oop_work(T* p) {
2254 if (!_cm->has_overflown()) {
2255 oop obj = oopDesc::load_decode_heap_oop(p);
2256 if (_cm->verbose_high()) {
2257 gclog_or_tty->print_cr("\t[%u] we're looking at location "
2258 "*"PTR_FORMAT" = "PTR_FORMAT,
2259 _task->worker_id(), p2i(p), p2i((void*) obj));
2260 }
2261
2262 _task->deal_with_reference(obj);
2263 _ref_counter--;
2264
2265 if (_ref_counter == 0) {
2266 // We have dealt with _ref_counter_limit references, pushing them
2267 // and objects reachable from them on to the local stack (and
2268 // possibly the global stack). Call CMTask::do_marking_step() to
2269 // process these entries.
2270 //
2271 // We call CMTask::do_marking_step() in a loop, which we'll exit if
2272 // there's nothing more to do (i.e. we're done with the entries that
2273 // were pushed as a result of the CMTask::deal_with_reference() calls
2274 // above) or we overflow.
2275 //
2276 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2277 // flag while there may still be some work to do. (See the comment at
2278 // the beginning of CMTask::do_marking_step() for those conditions -
2279 // one of which is reaching the specified time target.) It is only
2280 // when CMTask::do_marking_step() returns without setting the
2281 // has_aborted() flag that the marking step has completed.
2282 do {
2283 double mark_step_duration_ms = G1ConcMarkStepDurationMillis;
2284 _task->do_marking_step(mark_step_duration_ms,
2285 false /* do_termination */,
2286 _is_serial);
2287 } while (_task->has_aborted() && !_cm->has_overflown());
2288 _ref_counter = _ref_counter_limit;
2289 }
2290 } else {
2291 if (_cm->verbose_high()) {
2292 gclog_or_tty->print_cr("\t[%u] CM Overflow", _task->worker_id());
2293 }
2294 }
2295 }
2296 };
2297
2298 // 'Drain' oop closure used by both serial and parallel reference processing.
2299 // Uses the CMTask associated with a given worker thread (for serial
2300 // reference processing the CMtask for worker 0 is used). Calls the
2301 // do_marking_step routine, with an unbelievably large timeout value,
2302 // to drain the marking data structures of the remaining entries
2303 // added by the 'keep alive' oop closure above.
2304
2305 class G1CMDrainMarkingStackClosure: public VoidClosure {
2306 ConcurrentMark* _cm;
2307 CMTask* _task;
2308 bool _is_serial;
2309 public:
2310 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) :
2311 _cm(cm), _task(task), _is_serial(is_serial) {
2312 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code");
2313 }
2314
2315 void do_void() {
2316 do {
2317 if (_cm->verbose_high()) {
2318 gclog_or_tty->print_cr("\t[%u] Drain: Calling do_marking_step - serial: %s",
2319 _task->worker_id(), BOOL_TO_STR(_is_serial));
2320 }
2321
2322 // We call CMTask::do_marking_step() to completely drain the local
2323 // and global marking stacks of entries pushed by the 'keep alive'
2324 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above).
2325 //
2326 // CMTask::do_marking_step() is called in a loop, which we'll exit
2327 // if there's nothing more to do (i.e. we'completely drained the
2328 // entries that were pushed as a a result of applying the 'keep alive'
2329 // closure to the entries on the discovered ref lists) or we overflow
2330 // the global marking stack.
2331 //
2332 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted()
2333 // flag while there may still be some work to do. (See the comment at
2334 // the beginning of CMTask::do_marking_step() for those conditions -
2335 // one of which is reaching the specified time target.) It is only
2336 // when CMTask::do_marking_step() returns without setting the
2337 // has_aborted() flag that the marking step has completed.
2338
2339 _task->do_marking_step(1000000000.0 /* something very large */,
2340 true /* do_termination */,
2341 _is_serial);
2342 } while (_task->has_aborted() && !_cm->has_overflown());
2343 }
2344 };
2345
2346 // Implementation of AbstractRefProcTaskExecutor for parallel
2347 // reference processing at the end of G1 concurrent marking
2348
2349 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor {
2350 private:
2351 G1CollectedHeap* _g1h;
2352 ConcurrentMark* _cm;
2353 WorkGang* _workers;
2354 int _active_workers;
2355
2356 public:
2357 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h,
2358 ConcurrentMark* cm,
2359 WorkGang* workers,
2360 int n_workers) :
2361 _g1h(g1h), _cm(cm),
2362 _workers(workers), _active_workers(n_workers) { }
2363
2364 // Executes the given task using concurrent marking worker threads.
2365 virtual void execute(ProcessTask& task);
2366 virtual void execute(EnqueueTask& task);
2367 };
2368
2369 class G1CMRefProcTaskProxy: public AbstractGangTask {
2370 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask;
2371 ProcessTask& _proc_task;
2372 G1CollectedHeap* _g1h;
2373 ConcurrentMark* _cm;
2374
2375 public:
2376 G1CMRefProcTaskProxy(ProcessTask& proc_task,
2377 G1CollectedHeap* g1h,
2378 ConcurrentMark* cm) :
2379 AbstractGangTask("Process reference objects in parallel"),
2380 _proc_task(proc_task), _g1h(g1h), _cm(cm) {
2381 ReferenceProcessor* rp = _g1h->ref_processor_cm();
2382 assert(rp->processing_is_mt(), "shouldn't be here otherwise");
2383 }
2384
2385 virtual void work(uint worker_id) {
2386 CMTask* task = _cm->task(worker_id);
2387 G1CMIsAliveClosure g1_is_alive(_g1h);
2388 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */);
2389 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */);
2390
2391 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain);
2392 }
2393 };
2394
2395 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) {
2396 assert(_workers != NULL, "Need parallel worker threads.");
2397 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2398
2399 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm);
2400
2401 // We need to reset the concurrency level before each
2402 // proxy task execution, so that the termination protocol
2403 // and overflow handling in CMTask::do_marking_step() knows
2404 // how many workers to wait for.
2405 _cm->set_concurrency(_active_workers);
2406 _g1h->set_par_threads(_active_workers);
2407 _workers->run_task(&proc_task_proxy);
2408 _g1h->set_par_threads(0);
2409 }
2410
2411 class G1CMRefEnqueueTaskProxy: public AbstractGangTask {
2412 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask;
2413 EnqueueTask& _enq_task;
2414
2415 public:
2416 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) :
2417 AbstractGangTask("Enqueue reference objects in parallel"),
2418 _enq_task(enq_task) { }
2419
2420 virtual void work(uint worker_id) {
2421 _enq_task.work(worker_id);
2422 }
2423 };
2424
2425 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) {
2426 assert(_workers != NULL, "Need parallel worker threads.");
2427 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT");
2428
2429 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task);
2430
2431 // Not strictly necessary but...
2432 //
2433 // We need to reset the concurrency level before each
2434 // proxy task execution, so that the termination protocol
2435 // and overflow handling in CMTask::do_marking_step() knows
2436 // how many workers to wait for.
2437 _cm->set_concurrency(_active_workers);
2438 _g1h->set_par_threads(_active_workers);
2439 _workers->run_task(&enq_task_proxy);
2440 _g1h->set_par_threads(0);
2441 }
2442
2443 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) {
2444 if (has_overflown()) {
2445 // Skip processing the discovered references if we have
2446 // overflown the global marking stack. Reference objects
2447 // only get discovered once so it is OK to not
2448 // de-populate the discovered reference lists. We could have,
2449 // but the only benefit would be that, when marking restarts,
2450 // less reference objects are discovered.
2451 return;
2452 }
2453
2454 ResourceMark rm;
2455 HandleMark hm;
2456
2457 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2458
2459 // Is alive closure.
2460 G1CMIsAliveClosure g1_is_alive(g1h);
2461
2462 // Inner scope to exclude the cleaning of the string and symbol
2463 // tables from the displayed time.
2464 {
2465 if (G1Log::finer()) {
2466 gclog_or_tty->put(' ');
2467 }
2468 GCTraceTime t("GC ref-proc", G1Log::finer(), false, g1h->gc_timer_cm(), concurrent_gc_id());
2469
2470 ReferenceProcessor* rp = g1h->ref_processor_cm();
2471
2472 // See the comment in G1CollectedHeap::ref_processing_init()
2473 // about how reference processing currently works in G1.
2474
2475 // Set the soft reference policy
2476 rp->setup_policy(clear_all_soft_refs);
2477 assert(_markStack.isEmpty(), "mark stack should be empty");
2478
2479 // Instances of the 'Keep Alive' and 'Complete GC' closures used
2480 // in serial reference processing. Note these closures are also
2481 // used for serially processing (by the the current thread) the
2482 // JNI references during parallel reference processing.
2483 //
2484 // These closures do not need to synchronize with the worker
2485 // threads involved in parallel reference processing as these
2486 // instances are executed serially by the current thread (e.g.
2487 // reference processing is not multi-threaded and is thus
2488 // performed by the current thread instead of a gang worker).
2489 //
2490 // The gang tasks involved in parallel reference procssing create
2491 // their own instances of these closures, which do their own
2492 // synchronization among themselves.
2493 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */);
2494 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */);
2495
2496 // We need at least one active thread. If reference processing
2497 // is not multi-threaded we use the current (VMThread) thread,
2498 // otherwise we use the work gang from the G1CollectedHeap and
2499 // we utilize all the worker threads we can.
2500 bool processing_is_mt = rp->processing_is_mt() && g1h->workers() != NULL;
2501 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U);
2502 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U);
2503
2504 // Parallel processing task executor.
2505 G1CMRefProcTaskExecutor par_task_executor(g1h, this,
2506 g1h->workers(), active_workers);
2507 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL);
2508
2509 // Set the concurrency level. The phase was already set prior to
2510 // executing the remark task.
2511 set_concurrency(active_workers);
2512
2513 // Set the degree of MT processing here. If the discovery was done MT,
2514 // the number of threads involved during discovery could differ from
2515 // the number of active workers. This is OK as long as the discovered
2516 // Reference lists are balanced (see balance_all_queues() and balance_queues()).
2517 rp->set_active_mt_degree(active_workers);
2518
2519 // Process the weak references.
2520 const ReferenceProcessorStats& stats =
2521 rp->process_discovered_references(&g1_is_alive,
2522 &g1_keep_alive,
2523 &g1_drain_mark_stack,
2524 executor,
2525 g1h->gc_timer_cm(),
2526 concurrent_gc_id());
2527 g1h->gc_tracer_cm()->report_gc_reference_stats(stats);
2528
2529 // The do_oop work routines of the keep_alive and drain_marking_stack
2530 // oop closures will set the has_overflown flag if we overflow the
2531 // global marking stack.
2532
2533 assert(_markStack.overflow() || _markStack.isEmpty(),
2534 "mark stack should be empty (unless it overflowed)");
2535
2536 if (_markStack.overflow()) {
2537 // This should have been done already when we tried to push an
2538 // entry on to the global mark stack. But let's do it again.
2539 set_has_overflown();
2540 }
2541
2542 assert(rp->num_q() == active_workers, "why not");
2543
2544 rp->enqueue_discovered_references(executor);
2545
2546 rp->verify_no_references_recorded();
2547 assert(!rp->discovery_enabled(), "Post condition");
2548 }
2549
2550 if (has_overflown()) {
2551 // We can not trust g1_is_alive if the marking stack overflowed
2552 return;
2553 }
2554
2555 g1h->unlink_string_and_symbol_table(&g1_is_alive,
2556 /* process_strings */ false, // currently strings are always roots
2557 /* process_symbols */ true);
2558 }
2559
2560 void ConcurrentMark::swapMarkBitMaps() {
2561 CMBitMapRO* temp = _prevMarkBitMap;
2562 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap;
2563 _nextMarkBitMap = (CMBitMap*) temp;
2564 }
2565
2566 class CMRemarkTask: public AbstractGangTask {
2567 private:
2568 ConcurrentMark* _cm;
2569 bool _is_serial;
2570 public:
2571 void work(uint worker_id) {
2572 // Since all available tasks are actually started, we should
2573 // only proceed if we're supposed to be actived.
2574 if (worker_id < _cm->active_tasks()) {
2575 CMTask* task = _cm->task(worker_id);
2576 task->record_start_time();
2577 do {
2578 task->do_marking_step(1000000000.0 /* something very large */,
2579 true /* do_termination */,
2580 _is_serial);
2581 } while (task->has_aborted() && !_cm->has_overflown());
2582 // If we overflow, then we do not want to restart. We instead
2583 // want to abort remark and do concurrent marking again.
2584 task->record_end_time();
2585 }
2586 }
2587
2588 CMRemarkTask(ConcurrentMark* cm, int active_workers, bool is_serial) :
2589 AbstractGangTask("Par Remark"), _cm(cm), _is_serial(is_serial) {
2590 _cm->terminator()->reset_for_reuse(active_workers);
2591 }
2592 };
2593
2594 void ConcurrentMark::checkpointRootsFinalWork() {
2595 ResourceMark rm;
2596 HandleMark hm;
2597 G1CollectedHeap* g1h = G1CollectedHeap::heap();
2598
2599 g1h->ensure_parsability(false);
2600
2601 if (G1CollectedHeap::use_parallel_gc_threads()) {
2602 G1CollectedHeap::StrongRootsScope srs(g1h);
2603 // this is remark, so we'll use up all active threads
2604 uint active_workers = g1h->workers()->active_workers();
2605 if (active_workers == 0) {
2606 assert(active_workers > 0, "Should have been set earlier");
2607 active_workers = (uint) ParallelGCThreads;
2608 g1h->workers()->set_active_workers(active_workers);
2609 }
2610 set_concurrency_and_phase(active_workers, false /* concurrent */);
2611 // Leave _parallel_marking_threads at it's
2612 // value originally calculated in the ConcurrentMark
2613 // constructor and pass values of the active workers
2614 // through the gang in the task.
2615
2616 CMRemarkTask remarkTask(this, active_workers, false /* is_serial */);
2617 // We will start all available threads, even if we decide that the
2618 // active_workers will be fewer. The extra ones will just bail out
2619 // immediately.
2620 g1h->set_par_threads(active_workers);
2621 g1h->workers()->run_task(&remarkTask);
2622 g1h->set_par_threads(0);
2623 } else {
2624 G1CollectedHeap::StrongRootsScope srs(g1h);
2625 uint active_workers = 1;
2626 set_concurrency_and_phase(active_workers, false /* concurrent */);
2627
2628 // Note - if there's no work gang then the VMThread will be
2629 // the thread to execute the remark - serially. We have
2630 // to pass true for the is_serial parameter so that
2631 // CMTask::do_marking_step() doesn't enter the sync
2632 // barriers in the event of an overflow. Doing so will
2633 // cause an assert that the current thread is not a
2634 // concurrent GC thread.
2635 CMRemarkTask remarkTask(this, active_workers, true /* is_serial*/);
2636 remarkTask.work(0);
2637 }
2638 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
2639 guarantee(has_overflown() ||
2640 satb_mq_set.completed_buffers_num() == 0,
2641 err_msg("Invariant: has_overflown = %s, num buffers = %d",
2642 BOOL_TO_STR(has_overflown()),
2643 satb_mq_set.completed_buffers_num()));
2644
2645 print_stats();
2646 }
2647
2648 #ifndef PRODUCT
2649
2650 class PrintReachableOopClosure: public OopClosure {
2651 private:
2652 G1CollectedHeap* _g1h;
2653 outputStream* _out;
2654 VerifyOption _vo;
2655 bool _all;
2656
2657 public:
2658 PrintReachableOopClosure(outputStream* out,
2659 VerifyOption vo,
2660 bool all) :
2661 _g1h(G1CollectedHeap::heap()),
2662 _out(out), _vo(vo), _all(all) { }
2663
2664 void do_oop(narrowOop* p) { do_oop_work(p); }
2665 void do_oop( oop* p) { do_oop_work(p); }
2666
2667 template <class T> void do_oop_work(T* p) {
2668 oop obj = oopDesc::load_decode_heap_oop(p);
2669 const char* str = NULL;
2670 const char* str2 = "";
2671
2672 if (obj == NULL) {
2673 str = "";
2674 } else if (!_g1h->is_in_g1_reserved(obj)) {
2675 str = " O";
2676 } else {
2677 HeapRegion* hr = _g1h->heap_region_containing(obj);
2678 guarantee(hr != NULL, "invariant");
2679 bool over_tams = _g1h->allocated_since_marking(obj, hr, _vo);
2680 bool marked = _g1h->is_marked(obj, _vo);
2681
2682 if (over_tams) {
2683 str = " >";
2684 if (marked) {
2685 str2 = " AND MARKED";
2686 }
2687 } else if (marked) {
2688 str = " M";
2689 } else {
2690 str = " NOT";
2691 }
2692 }
2693
2694 _out->print_cr(" "PTR_FORMAT": "PTR_FORMAT"%s%s",
2695 p2i(p), p2i((void*) obj), str, str2);
2696 }
2697 };
2698
2699 class PrintReachableObjectClosure : public ObjectClosure {
2700 private:
2701 G1CollectedHeap* _g1h;
2702 outputStream* _out;
2703 VerifyOption _vo;
2704 bool _all;
2705 HeapRegion* _hr;
2706
2707 public:
2708 PrintReachableObjectClosure(outputStream* out,
2709 VerifyOption vo,
2710 bool all,
2711 HeapRegion* hr) :
2712 _g1h(G1CollectedHeap::heap()),
2713 _out(out), _vo(vo), _all(all), _hr(hr) { }
2714
2715 void do_object(oop o) {
2716 bool over_tams = _g1h->allocated_since_marking(o, _hr, _vo);
2717 bool marked = _g1h->is_marked(o, _vo);
2718 bool print_it = _all || over_tams || marked;
2719
2720 if (print_it) {
2721 _out->print_cr(" "PTR_FORMAT"%s",
2722 p2i((void *)o), (over_tams) ? " >" : (marked) ? " M" : "");
2723 PrintReachableOopClosure oopCl(_out, _vo, _all);
2724 o->oop_iterate_no_header(&oopCl);
2725 }
2726 }
2727 };
2728
2729 class PrintReachableRegionClosure : public HeapRegionClosure {
2730 private:
2731 G1CollectedHeap* _g1h;
2732 outputStream* _out;
2733 VerifyOption _vo;
2734 bool _all;
2735
2736 public:
2737 bool doHeapRegion(HeapRegion* hr) {
2738 HeapWord* b = hr->bottom();
2739 HeapWord* e = hr->end();
2740 HeapWord* t = hr->top();
2741 HeapWord* p = _g1h->top_at_mark_start(hr, _vo);
2742 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" "
2743 "TAMS: " PTR_FORMAT, p2i(b), p2i(e), p2i(t), p2i(p));
2744 _out->cr();
2745
2746 HeapWord* from = b;
2747 HeapWord* to = t;
2748
2749 if (to > from) {
2750 _out->print_cr("Objects in [" PTR_FORMAT ", " PTR_FORMAT "]", p2i(from), p2i(to));
2751 _out->cr();
2752 PrintReachableObjectClosure ocl(_out, _vo, _all, hr);
2753 hr->object_iterate_mem_careful(MemRegion(from, to), &ocl);
2754 _out->cr();
2755 }
2756
2757 return false;
2758 }
2759
2760 PrintReachableRegionClosure(outputStream* out,
2761 VerifyOption vo,
2762 bool all) :
2763 _g1h(G1CollectedHeap::heap()), _out(out), _vo(vo), _all(all) { }
2764 };
2765
2766 void ConcurrentMark::print_reachable(const char* str,
2767 VerifyOption vo,
2768 bool all) {
2769 gclog_or_tty->cr();
2770 gclog_or_tty->print_cr("== Doing heap dump... ");
2771
2772 if (G1PrintReachableBaseFile == NULL) {
2773 gclog_or_tty->print_cr(" #### error: no base file defined");
2774 return;
2775 }
2776
2777 if (strlen(G1PrintReachableBaseFile) + 1 + strlen(str) >
2778 (JVM_MAXPATHLEN - 1)) {
2779 gclog_or_tty->print_cr(" #### error: file name too long");
2780 return;
2781 }
2782
2783 char file_name[JVM_MAXPATHLEN];
2784 sprintf(file_name, "%s.%s", G1PrintReachableBaseFile, str);
2785 gclog_or_tty->print_cr(" dumping to file %s", file_name);
2786
2787 fileStream fout(file_name);
2788 if (!fout.is_open()) {
2789 gclog_or_tty->print_cr(" #### error: could not open file");
2790 return;
2791 }
2792
2793 outputStream* out = &fout;
2794 out->print_cr("-- USING %s", _g1h->top_at_mark_start_str(vo));
2795 out->cr();
2796
2797 out->print_cr("--- ITERATING OVER REGIONS");
2798 out->cr();
2799 PrintReachableRegionClosure rcl(out, vo, all);
2800 _g1h->heap_region_iterate(&rcl);
2801 out->cr();
2802
2803 gclog_or_tty->print_cr(" done");
2804 gclog_or_tty->flush();
2805 }
2806
2807 #endif // PRODUCT
2808
2809 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) {
2810 // Note we are overriding the read-only view of the prev map here, via
2811 // the cast.
2812 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr);
2813 }
2814
2815 void ConcurrentMark::clearRangeNextBitmap(MemRegion mr) {
2816 _nextMarkBitMap->clearRange(mr);
2817 }
2818
2819 void ConcurrentMark::clearRangeBothBitmaps(MemRegion mr) {
2820 clearRangePrevBitmap(mr);
2821 clearRangeNextBitmap(mr);
2822 }
2823
2824 HeapRegion*
2825 ConcurrentMark::claim_region(uint worker_id) {
2826 // "checkpoint" the finger
2827 HeapWord* finger = _finger;
2828
2829 // _heap_end will not change underneath our feet; it only changes at
2830 // yield points.
2831 while (finger < _heap_end) {
2832 assert(_g1h->is_in_g1_reserved(finger), "invariant");
2833
2834 // Note on how this code handles humongous regions. In the
2835 // normal case the finger will reach the start of a "starts
2836 // humongous" (SH) region. Its end will either be the end of the
2837 // last "continues humongous" (CH) region in the sequence, or the
2838 // standard end of the SH region (if the SH is the only region in
2839 // the sequence). That way claim_region() will skip over the CH
2840 // regions. However, there is a subtle race between a CM thread
2841 // executing this method and a mutator thread doing a humongous
2842 // object allocation. The two are not mutually exclusive as the CM
2843 // thread does not need to hold the Heap_lock when it gets
2844 // here. So there is a chance that claim_region() will come across
2845 // a free region that's in the progress of becoming a SH or a CH
2846 // region. In the former case, it will either
2847 // a) Miss the update to the region's end, in which case it will
2848 // visit every subsequent CH region, will find their bitmaps
2849 // empty, and do nothing, or
2850 // b) Will observe the update of the region's end (in which case
2851 // it will skip the subsequent CH regions).
2852 // If it comes across a region that suddenly becomes CH, the
2853 // scenario will be similar to b). So, the race between
2854 // claim_region() and a humongous object allocation might force us
2855 // to do a bit of unnecessary work (due to some unnecessary bitmap
2856 // iterations) but it should not introduce and correctness issues.
2857 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger);
2858 HeapWord* bottom = curr_region->bottom();
2859 HeapWord* end = curr_region->end();
2860 HeapWord* limit = curr_region->next_top_at_mark_start();
2861
2862 if (verbose_low()) {
2863 gclog_or_tty->print_cr("[%u] curr_region = "PTR_FORMAT" "
2864 "["PTR_FORMAT", "PTR_FORMAT"), "
2865 "limit = "PTR_FORMAT,
2866 worker_id, p2i(curr_region), p2i(bottom), p2i(end), p2i(limit));
2867 }
2868
2869 // Is the gap between reading the finger and doing the CAS too long?
2870 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger);
2871 if (res == finger) {
2872 // we succeeded
2873
2874 // notice that _finger == end cannot be guaranteed here since,
2875 // someone else might have moved the finger even further
2876 assert(_finger >= end, "the finger should have moved forward");
2877
2878 if (verbose_low()) {
2879 gclog_or_tty->print_cr("[%u] we were successful with region = "
2880 PTR_FORMAT, worker_id, p2i(curr_region));
2881 }
2882
2883 if (limit > bottom) {
2884 if (verbose_low()) {
2885 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is not empty, "
2886 "returning it ", worker_id, p2i(curr_region));
2887 }
2888 return curr_region;
2889 } else {
2890 assert(limit == bottom,
2891 "the region limit should be at bottom");
2892 if (verbose_low()) {
2893 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is empty, "
2894 "returning NULL", worker_id, p2i(curr_region));
2895 }
2896 // we return NULL and the caller should try calling
2897 // claim_region() again.
2898 return NULL;
2899 }
2900 } else {
2901 assert(_finger > finger, "the finger should have moved forward");
2902 if (verbose_low()) {
2903 gclog_or_tty->print_cr("[%u] somebody else moved the finger, "
2904 "global finger = "PTR_FORMAT", "
2905 "our finger = "PTR_FORMAT,
2906 worker_id, p2i(_finger), p2i(finger));
2907 }
2908
2909 // read it again
2910 finger = _finger;
2911 }
2912 }
2913
2914 return NULL;
2915 }
2916
2917 #ifndef PRODUCT
2918 enum VerifyNoCSetOopsPhase {
2919 VerifyNoCSetOopsStack,
2920 VerifyNoCSetOopsQueues,
2921 VerifyNoCSetOopsSATBCompleted,
2922 VerifyNoCSetOopsSATBThread
2923 };
2924
2925 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure {
2926 private:
2927 G1CollectedHeap* _g1h;
2928 VerifyNoCSetOopsPhase _phase;
2929 int _info;
2930
2931 const char* phase_str() {
2932 switch (_phase) {
2933 case VerifyNoCSetOopsStack: return "Stack";
2934 case VerifyNoCSetOopsQueues: return "Queue";
2935 case VerifyNoCSetOopsSATBCompleted: return "Completed SATB Buffers";
2936 case VerifyNoCSetOopsSATBThread: return "Thread SATB Buffers";
2937 default: ShouldNotReachHere();
2938 }
2939 return NULL;
2940 }
2941
2942 void do_object_work(oop obj) {
2943 guarantee(!_g1h->obj_in_cs(obj),
2944 err_msg("obj: "PTR_FORMAT" in CSet, phase: %s, info: %d",
2945 p2i((void*) obj), phase_str(), _info));
2946 }
2947
2948 public:
2949 VerifyNoCSetOopsClosure() : _g1h(G1CollectedHeap::heap()) { }
2950
2951 void set_phase(VerifyNoCSetOopsPhase phase, int info = -1) {
2952 _phase = phase;
2953 _info = info;
2954 }
2955
2956 virtual void do_oop(oop* p) {
2957 oop obj = oopDesc::load_decode_heap_oop(p);
2958 do_object_work(obj);
2959 }
2960
2961 virtual void do_oop(narrowOop* p) {
2962 // We should not come across narrow oops while scanning marking
2963 // stacks and SATB buffers.
2964 ShouldNotReachHere();
2965 }
2966
2967 virtual void do_object(oop obj) {
2968 do_object_work(obj);
2969 }
2970 };
2971
2972 void ConcurrentMark::verify_no_cset_oops(bool verify_stacks,
2973 bool verify_enqueued_buffers,
2974 bool verify_thread_buffers,
2975 bool verify_fingers) {
2976 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint");
2977 if (!G1CollectedHeap::heap()->mark_in_progress()) {
2978 return;
2979 }
2980
2981 VerifyNoCSetOopsClosure cl;
2982
2983 if (verify_stacks) {
2984 // Verify entries on the global mark stack
2985 cl.set_phase(VerifyNoCSetOopsStack);
2986 _markStack.oops_do(&cl);
2987
2988 // Verify entries on the task queues
2989 for (uint i = 0; i < _max_worker_id; i += 1) {
2990 cl.set_phase(VerifyNoCSetOopsQueues, i);
2991 CMTaskQueue* queue = _task_queues->queue(i);
2992 queue->oops_do(&cl);
2993 }
2994 }
2995
2996 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set();
2997
2998 // Verify entries on the enqueued SATB buffers
2999 if (verify_enqueued_buffers) {
3000 cl.set_phase(VerifyNoCSetOopsSATBCompleted);
3001 satb_qs.iterate_completed_buffers_read_only(&cl);
3002 }
3003
3004 // Verify entries on the per-thread SATB buffers
3005 if (verify_thread_buffers) {
3006 cl.set_phase(VerifyNoCSetOopsSATBThread);
3007 satb_qs.iterate_thread_buffers_read_only(&cl);
3008 }
3009
3010 if (verify_fingers) {
3011 // Verify the global finger
3012 HeapWord* global_finger = finger();
3013 if (global_finger != NULL && global_finger < _heap_end) {
3014 // The global finger always points to a heap region boundary. We
3015 // use heap_region_containing_raw() to get the containing region
3016 // given that the global finger could be pointing to a free region
3017 // which subsequently becomes continues humongous. If that
3018 // happens, heap_region_containing() will return the bottom of the
3019 // corresponding starts humongous region and the check below will
3020 // not hold any more.
3021 HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger);
3022 guarantee(global_finger == global_hr->bottom(),
3023 err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT,
3024 p2i(global_finger), HR_FORMAT_PARAMS(global_hr)));
3025 }
3026
3027 // Verify the task fingers
3028 assert(parallel_marking_threads() <= _max_worker_id, "sanity");
3029 for (int i = 0; i < (int) parallel_marking_threads(); i += 1) {
3030 CMTask* task = _tasks[i];
3031 HeapWord* task_finger = task->finger();
3032 if (task_finger != NULL && task_finger < _heap_end) {
3033 // See above note on the global finger verification.
3034 HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger);
3035 guarantee(task_finger == task_hr->bottom() ||
3036 !task_hr->in_collection_set(),
3037 err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT,
3038 p2i(task_finger), HR_FORMAT_PARAMS(task_hr)));
3039 }
3040 }
3041 }
3042 }
3043 #endif // PRODUCT
3044
3045 // Aggregate the counting data that was constructed concurrently
3046 // with marking.
3047 class AggregateCountDataHRClosure: public HeapRegionClosure {
3048 G1CollectedHeap* _g1h;
3049 ConcurrentMark* _cm;
3050 CardTableModRefBS* _ct_bs;
3051 BitMap* _cm_card_bm;
3052 uint _max_worker_id;
3053
3054 public:
3055 AggregateCountDataHRClosure(G1CollectedHeap* g1h,
3056 BitMap* cm_card_bm,
3057 uint max_worker_id) :
3058 _g1h(g1h), _cm(g1h->concurrent_mark()),
3059 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())),
3060 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { }
3061
3062 bool doHeapRegion(HeapRegion* hr) {
3063 if (hr->continuesHumongous()) {
3064 // We will ignore these here and process them when their
3065 // associated "starts humongous" region is processed.
3066 // Note that we cannot rely on their associated
3067 // "starts humongous" region to have their bit set to 1
3068 // since, due to the region chunking in the parallel region
3069 // iteration, a "continues humongous" region might be visited
3070 // before its associated "starts humongous".
3071 return false;
3072 }
3073
3074 HeapWord* start = hr->bottom();
3075 HeapWord* limit = hr->next_top_at_mark_start();
3076 HeapWord* end = hr->end();
3077
3078 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(),
3079 err_msg("Preconditions not met - "
3080 "start: "PTR_FORMAT", limit: "PTR_FORMAT", "
3081 "top: "PTR_FORMAT", end: "PTR_FORMAT,
3082 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end())));
3083
3084 assert(hr->next_marked_bytes() == 0, "Precondition");
3085
3086 if (start == limit) {
3087 // NTAMS of this region has not been set so nothing to do.
3088 return false;
3089 }
3090
3091 // 'start' should be in the heap.
3092 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity");
3093 // 'end' *may* be just beyone the end of the heap (if hr is the last region)
3094 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity");
3095
3096 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start);
3097 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit);
3098 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end);
3099
3100 // If ntams is not card aligned then we bump card bitmap index
3101 // for limit so that we get the all the cards spanned by
3102 // the object ending at ntams.
3103 // Note: if this is the last region in the heap then ntams
3104 // could be actually just beyond the end of the the heap;
3105 // limit_idx will then correspond to a (non-existent) card
3106 // that is also outside the heap.
3107 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) {
3108 limit_idx += 1;
3109 }
3110
3111 assert(limit_idx <= end_idx, "or else use atomics");
3112
3113 // Aggregate the "stripe" in the count data associated with hr.
3114 uint hrs_index = hr->hrs_index();
3115 size_t marked_bytes = 0;
3116
3117 for (uint i = 0; i < _max_worker_id; i += 1) {
3118 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i);
3119 BitMap* task_card_bm = _cm->count_card_bitmap_for(i);
3120
3121 // Fetch the marked_bytes in this region for task i and
3122 // add it to the running total for this region.
3123 marked_bytes += marked_bytes_array[hrs_index];
3124
3125 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx)
3126 // into the global card bitmap.
3127 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx);
3128
3129 while (scan_idx < limit_idx) {
3130 assert(task_card_bm->at(scan_idx) == true, "should be");
3131 _cm_card_bm->set_bit(scan_idx);
3132 assert(_cm_card_bm->at(scan_idx) == true, "should be");
3133
3134 // BitMap::get_next_one_offset() can handle the case when
3135 // its left_offset parameter is greater than its right_offset
3136 // parameter. It does, however, have an early exit if
3137 // left_offset == right_offset. So let's limit the value
3138 // passed in for left offset here.
3139 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx);
3140 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx);
3141 }
3142 }
3143
3144 // Update the marked bytes for this region.
3145 hr->add_to_marked_bytes(marked_bytes);
3146
3147 // Next heap region
3148 return false;
3149 }
3150 };
3151
3152 class G1AggregateCountDataTask: public AbstractGangTask {
3153 protected:
3154 G1CollectedHeap* _g1h;
3155 ConcurrentMark* _cm;
3156 BitMap* _cm_card_bm;
3157 uint _max_worker_id;
3158 int _active_workers;
3159
3160 public:
3161 G1AggregateCountDataTask(G1CollectedHeap* g1h,
3162 ConcurrentMark* cm,
3163 BitMap* cm_card_bm,
3164 uint max_worker_id,
3165 int n_workers) :
3166 AbstractGangTask("Count Aggregation"),
3167 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm),
3168 _max_worker_id(max_worker_id),
3169 _active_workers(n_workers) { }
3170
3171 void work(uint worker_id) {
3172 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id);
3173
3174 if (G1CollectedHeap::use_parallel_gc_threads()) {
3175 _g1h->heap_region_par_iterate_chunked(&cl, worker_id,
3176 _active_workers,
3177 HeapRegion::AggregateCountClaimValue);
3178 } else {
3179 _g1h->heap_region_iterate(&cl);
3180 }
3181 }
3182 };
3183
3184
3185 void ConcurrentMark::aggregate_count_data() {
3186 int n_workers = (G1CollectedHeap::use_parallel_gc_threads() ?
3187 _g1h->workers()->active_workers() :
3188 1);
3189
3190 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm,
3191 _max_worker_id, n_workers);
3192
3193 if (G1CollectedHeap::use_parallel_gc_threads()) {
3194 assert(_g1h->check_heap_region_claim_values(HeapRegion::InitialClaimValue),
3195 "sanity check");
3196 _g1h->set_par_threads(n_workers);
3197 _g1h->workers()->run_task(&g1_par_agg_task);
3198 _g1h->set_par_threads(0);
3199
3200 assert(_g1h->check_heap_region_claim_values(HeapRegion::AggregateCountClaimValue),
3201 "sanity check");
3202 _g1h->reset_heap_region_claim_values();
3203 } else {
3204 g1_par_agg_task.work(0);
3205 }
3206 }
3207
3208 // Clear the per-worker arrays used to store the per-region counting data
3209 void ConcurrentMark::clear_all_count_data() {
3210 // Clear the global card bitmap - it will be filled during
3211 // liveness count aggregation (during remark) and the
3212 // final counting task.
3213 _card_bm.clear();
3214
3215 // Clear the global region bitmap - it will be filled as part
3216 // of the final counting task.
3217 _region_bm.clear();
3218
3219 uint max_regions = _g1h->max_regions();
3220 assert(_max_worker_id > 0, "uninitialized");
3221
3222 for (uint i = 0; i < _max_worker_id; i += 1) {
3223 BitMap* task_card_bm = count_card_bitmap_for(i);
3224 size_t* marked_bytes_array = count_marked_bytes_array_for(i);
3225
3226 assert(task_card_bm->size() == _card_bm.size(), "size mismatch");
3227 assert(marked_bytes_array != NULL, "uninitialized");
3228
3229 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t));
3230 task_card_bm->clear();
3231 }
3232 }
3233
3234 void ConcurrentMark::print_stats() {
3235 if (verbose_stats()) {
3236 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3237 for (size_t i = 0; i < _active_tasks; ++i) {
3238 _tasks[i]->print_stats();
3239 gclog_or_tty->print_cr("---------------------------------------------------------------------");
3240 }
3241 }
3242 }
3243
3244 // abandon current marking iteration due to a Full GC
3245 void ConcurrentMark::abort() {
3246 // Clear all marks to force marking thread to do nothing
3247 _nextMarkBitMap->clearAll();
3248 // Clear the liveness counting data
3249 clear_all_count_data();
3250 // Empty mark stack
3251 reset_marking_state();
3252 for (uint i = 0; i < _max_worker_id; ++i) {
3253 _tasks[i]->clear_region_fields();
3254 }
3255 _first_overflow_barrier_sync.abort();
3256 _second_overflow_barrier_sync.abort();
3257 const GCId& gc_id = _g1h->gc_tracer_cm()->gc_id();
3258 if (!gc_id.is_undefined()) {
3259 // We can do multiple full GCs before ConcurrentMarkThread::run() gets a chance
3260 // to detect that it was aborted. Only keep track of the first GC id that we aborted.
3261 _aborted_gc_id = gc_id;
3262 }
3263 _has_aborted = true;
3264
3265 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3266 satb_mq_set.abandon_partial_marking();
3267 // This can be called either during or outside marking, we'll read
3268 // the expected_active value from the SATB queue set.
3269 satb_mq_set.set_active_all_threads(
3270 false, /* new active value */
3271 satb_mq_set.is_active() /* expected_active */);
3272
3273 _g1h->trace_heap_after_concurrent_cycle();
3274 _g1h->register_concurrent_cycle_end();
3275 }
3276
3277 const GCId& ConcurrentMark::concurrent_gc_id() {
3278 if (has_aborted()) {
3279 return _aborted_gc_id;
3280 }
3281 return _g1h->gc_tracer_cm()->gc_id();
3282 }
3283
3284 static void print_ms_time_info(const char* prefix, const char* name,
3285 NumberSeq& ns) {
3286 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).",
3287 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg());
3288 if (ns.num() > 0) {
3289 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]",
3290 prefix, ns.sd(), ns.maximum());
3291 }
3292 }
3293
3294 void ConcurrentMark::print_summary_info() {
3295 gclog_or_tty->print_cr(" Concurrent marking:");
3296 print_ms_time_info(" ", "init marks", _init_times);
3297 print_ms_time_info(" ", "remarks", _remark_times);
3298 {
3299 print_ms_time_info(" ", "final marks", _remark_mark_times);
3300 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times);
3301
3302 }
3303 print_ms_time_info(" ", "cleanups", _cleanup_times);
3304 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).",
3305 _total_counting_time,
3306 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 /
3307 (double)_cleanup_times.num()
3308 : 0.0));
3309 if (G1ScrubRemSets) {
3310 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).",
3311 _total_rs_scrub_time,
3312 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 /
3313 (double)_cleanup_times.num()
3314 : 0.0));
3315 }
3316 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.",
3317 (_init_times.sum() + _remark_times.sum() +
3318 _cleanup_times.sum())/1000.0);
3319 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s "
3320 "(%8.2f s marking).",
3321 cmThread()->vtime_accum(),
3322 cmThread()->vtime_mark_accum());
3323 }
3324
3325 void ConcurrentMark::print_worker_threads_on(outputStream* st) const {
3326 if (use_parallel_marking_threads()) {
3327 _parallel_workers->print_worker_threads_on(st);
3328 }
3329 }
3330
3331 void ConcurrentMark::print_on_error(outputStream* st) const {
3332 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT,
3333 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap));
3334 _prevMarkBitMap->print_on_error(st, " Prev Bits: ");
3335 _nextMarkBitMap->print_on_error(st, " Next Bits: ");
3336 }
3337
3338 // We take a break if someone is trying to stop the world.
3339 bool ConcurrentMark::do_yield_check(uint worker_id) {
3340 if (SuspendibleThreadSet::should_yield()) {
3341 if (worker_id == 0) {
3342 _g1h->g1_policy()->record_concurrent_pause();
3343 }
3344 SuspendibleThreadSet::yield();
3345 return true;
3346 } else {
3347 return false;
3348 }
3349 }
3350
3351 bool ConcurrentMark::containing_card_is_marked(void* p) {
3352 size_t offset = pointer_delta(p, _g1h->reserved_region().start(), 1);
3353 return _card_bm.at(offset >> CardTableModRefBS::card_shift);
3354 }
3355
3356 bool ConcurrentMark::containing_cards_are_marked(void* start,
3357 void* last) {
3358 return containing_card_is_marked(start) &&
3359 containing_card_is_marked(last);
3360 }
3361
3362 #ifndef PRODUCT
3363 // for debugging purposes
3364 void ConcurrentMark::print_finger() {
3365 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT,
3366 p2i(_heap_start), p2i(_heap_end), p2i(_finger));
3367 for (uint i = 0; i < _max_worker_id; ++i) {
3368 gclog_or_tty->print(" %u: " PTR_FORMAT, i, p2i(_tasks[i]->finger()));
3369 }
3370 gclog_or_tty->cr();
3371 }
3372 #endif
3373
3374 void CMTask::scan_object(oop obj) {
3375 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant");
3376
3377 if (_cm->verbose_high()) {
3378 gclog_or_tty->print_cr("[%u] we're scanning object "PTR_FORMAT,
3379 _worker_id, p2i((void*) obj));
3380 }
3381
3382 size_t obj_size = obj->size();
3383 _words_scanned += obj_size;
3384
3385 obj->oop_iterate(_cm_oop_closure);
3386 statsOnly( ++_objs_scanned );
3387 check_limits();
3388 }
3389
3390 // Closure for iteration over bitmaps
3391 class CMBitMapClosure : public BitMapClosure {
3392 private:
3393 // the bitmap that is being iterated over
3394 CMBitMap* _nextMarkBitMap;
3395 ConcurrentMark* _cm;
3396 CMTask* _task;
3397
3398 public:
3399 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) :
3400 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { }
3401
3402 bool do_bit(size_t offset) {
3403 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset);
3404 assert(_nextMarkBitMap->isMarked(addr), "invariant");
3405 assert( addr < _cm->finger(), "invariant");
3406
3407 statsOnly( _task->increase_objs_found_on_bitmap() );
3408 assert(addr >= _task->finger(), "invariant");
3409
3410 // We move that task's local finger along.
3411 _task->move_finger_to(addr);
3412
3413 _task->scan_object(oop(addr));
3414 // we only partially drain the local queue and global stack
3415 _task->drain_local_queue(true);
3416 _task->drain_global_stack(true);
3417
3418 // if the has_aborted flag has been raised, we need to bail out of
3419 // the iteration
3420 return !_task->has_aborted();
3421 }
3422 };
3423
3424 // Closure for iterating over objects, currently only used for
3425 // processing SATB buffers.
3426 class CMObjectClosure : public ObjectClosure {
3427 private:
3428 CMTask* _task;
3429
3430 public:
3431 void do_object(oop obj) {
3432 _task->deal_with_reference(obj);
3433 }
3434
3435 CMObjectClosure(CMTask* task) : _task(task) { }
3436 };
3437
3438 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h,
3439 ConcurrentMark* cm,
3440 CMTask* task)
3441 : _g1h(g1h), _cm(cm), _task(task) {
3442 assert(_ref_processor == NULL, "should be initialized to NULL");
3443
3444 if (G1UseConcMarkReferenceProcessing) {
3445 _ref_processor = g1h->ref_processor_cm();
3446 assert(_ref_processor != NULL, "should not be NULL");
3447 }
3448 }
3449
3450 void CMTask::setup_for_region(HeapRegion* hr) {
3451 // Separated the asserts so that we know which one fires.
3452 assert(hr != NULL,
3453 "claim_region() should have filtered out continues humongous regions");
3454 assert(!hr->continuesHumongous(),
3455 "claim_region() should have filtered out continues humongous regions");
3456
3457 if (_cm->verbose_low()) {
3458 gclog_or_tty->print_cr("[%u] setting up for region "PTR_FORMAT,
3459 _worker_id, p2i(hr));
3460 }
3461
3462 _curr_region = hr;
3463 _finger = hr->bottom();
3464 update_region_limit();
3465 }
3466
3467 void CMTask::update_region_limit() {
3468 HeapRegion* hr = _curr_region;
3469 HeapWord* bottom = hr->bottom();
3470 HeapWord* limit = hr->next_top_at_mark_start();
3471
3472 if (limit == bottom) {
3473 if (_cm->verbose_low()) {
3474 gclog_or_tty->print_cr("[%u] found an empty region "
3475 "["PTR_FORMAT", "PTR_FORMAT")",
3476 _worker_id, p2i(bottom), p2i(limit));
3477 }
3478 // The region was collected underneath our feet.
3479 // We set the finger to bottom to ensure that the bitmap
3480 // iteration that will follow this will not do anything.
3481 // (this is not a condition that holds when we set the region up,
3482 // as the region is not supposed to be empty in the first place)
3483 _finger = bottom;
3484 } else if (limit >= _region_limit) {
3485 assert(limit >= _finger, "peace of mind");
3486 } else {
3487 assert(limit < _region_limit, "only way to get here");
3488 // This can happen under some pretty unusual circumstances. An
3489 // evacuation pause empties the region underneath our feet (NTAMS
3490 // at bottom). We then do some allocation in the region (NTAMS
3491 // stays at bottom), followed by the region being used as a GC
3492 // alloc region (NTAMS will move to top() and the objects
3493 // originally below it will be grayed). All objects now marked in
3494 // the region are explicitly grayed, if below the global finger,
3495 // and we do not need in fact to scan anything else. So, we simply
3496 // set _finger to be limit to ensure that the bitmap iteration
3497 // doesn't do anything.
3498 _finger = limit;
3499 }
3500
3501 _region_limit = limit;
3502 }
3503
3504 void CMTask::giveup_current_region() {
3505 assert(_curr_region != NULL, "invariant");
3506 if (_cm->verbose_low()) {
3507 gclog_or_tty->print_cr("[%u] giving up region "PTR_FORMAT,
3508 _worker_id, p2i(_curr_region));
3509 }
3510 clear_region_fields();
3511 }
3512
3513 void CMTask::clear_region_fields() {
3514 // Values for these three fields that indicate that we're not
3515 // holding on to a region.
3516 _curr_region = NULL;
3517 _finger = NULL;
3518 _region_limit = NULL;
3519 }
3520
3521 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) {
3522 if (cm_oop_closure == NULL) {
3523 assert(_cm_oop_closure != NULL, "invariant");
3524 } else {
3525 assert(_cm_oop_closure == NULL, "invariant");
3526 }
3527 _cm_oop_closure = cm_oop_closure;
3528 }
3529
3530 void CMTask::reset(CMBitMap* nextMarkBitMap) {
3531 guarantee(nextMarkBitMap != NULL, "invariant");
3532
3533 if (_cm->verbose_low()) {
3534 gclog_or_tty->print_cr("[%u] resetting", _worker_id);
3535 }
3536
3537 _nextMarkBitMap = nextMarkBitMap;
3538 clear_region_fields();
3539
3540 _calls = 0;
3541 _elapsed_time_ms = 0.0;
3542 _termination_time_ms = 0.0;
3543 _termination_start_time_ms = 0.0;
3544
3545 #if _MARKING_STATS_
3546 _local_pushes = 0;
3547 _local_pops = 0;
3548 _local_max_size = 0;
3549 _objs_scanned = 0;
3550 _global_pushes = 0;
3551 _global_pops = 0;
3552 _global_max_size = 0;
3553 _global_transfers_to = 0;
3554 _global_transfers_from = 0;
3555 _regions_claimed = 0;
3556 _objs_found_on_bitmap = 0;
3557 _satb_buffers_processed = 0;
3558 _steal_attempts = 0;
3559 _steals = 0;
3560 _aborted = 0;
3561 _aborted_overflow = 0;
3562 _aborted_cm_aborted = 0;
3563 _aborted_yield = 0;
3564 _aborted_timed_out = 0;
3565 _aborted_satb = 0;
3566 _aborted_termination = 0;
3567 #endif // _MARKING_STATS_
3568 }
3569
3570 bool CMTask::should_exit_termination() {
3571 regular_clock_call();
3572 // This is called when we are in the termination protocol. We should
3573 // quit if, for some reason, this task wants to abort or the global
3574 // stack is not empty (this means that we can get work from it).
3575 return !_cm->mark_stack_empty() || has_aborted();
3576 }
3577
3578 void CMTask::reached_limit() {
3579 assert(_words_scanned >= _words_scanned_limit ||
3580 _refs_reached >= _refs_reached_limit ,
3581 "shouldn't have been called otherwise");
3582 regular_clock_call();
3583 }
3584
3585 void CMTask::regular_clock_call() {
3586 if (has_aborted()) return;
3587
3588 // First, we need to recalculate the words scanned and refs reached
3589 // limits for the next clock call.
3590 recalculate_limits();
3591
3592 // During the regular clock call we do the following
3593
3594 // (1) If an overflow has been flagged, then we abort.
3595 if (_cm->has_overflown()) {
3596 set_has_aborted();
3597 return;
3598 }
3599
3600 // If we are not concurrent (i.e. we're doing remark) we don't need
3601 // to check anything else. The other steps are only needed during
3602 // the concurrent marking phase.
3603 if (!concurrent()) return;
3604
3605 // (2) If marking has been aborted for Full GC, then we also abort.
3606 if (_cm->has_aborted()) {
3607 set_has_aborted();
3608 statsOnly( ++_aborted_cm_aborted );
3609 return;
3610 }
3611
3612 double curr_time_ms = os::elapsedVTime() * 1000.0;
3613
3614 // (3) If marking stats are enabled, then we update the step history.
3615 #if _MARKING_STATS_
3616 if (_words_scanned >= _words_scanned_limit) {
3617 ++_clock_due_to_scanning;
3618 }
3619 if (_refs_reached >= _refs_reached_limit) {
3620 ++_clock_due_to_marking;
3621 }
3622
3623 double last_interval_ms = curr_time_ms - _interval_start_time_ms;
3624 _interval_start_time_ms = curr_time_ms;
3625 _all_clock_intervals_ms.add(last_interval_ms);
3626
3627 if (_cm->verbose_medium()) {
3628 gclog_or_tty->print_cr("[%u] regular clock, interval = %1.2lfms, "
3629 "scanned = %d%s, refs reached = %d%s",
3630 _worker_id, last_interval_ms,
3631 _words_scanned,
3632 (_words_scanned >= _words_scanned_limit) ? " (*)" : "",
3633 _refs_reached,
3634 (_refs_reached >= _refs_reached_limit) ? " (*)" : "");
3635 }
3636 #endif // _MARKING_STATS_
3637
3638 // (4) We check whether we should yield. If we have to, then we abort.
3639 if (SuspendibleThreadSet::should_yield()) {
3640 // We should yield. To do this we abort the task. The caller is
3641 // responsible for yielding.
3642 set_has_aborted();
3643 statsOnly( ++_aborted_yield );
3644 return;
3645 }
3646
3647 // (5) We check whether we've reached our time quota. If we have,
3648 // then we abort.
3649 double elapsed_time_ms = curr_time_ms - _start_time_ms;
3650 if (elapsed_time_ms > _time_target_ms) {
3651 set_has_aborted();
3652 _has_timed_out = true;
3653 statsOnly( ++_aborted_timed_out );
3654 return;
3655 }
3656
3657 // (6) Finally, we check whether there are enough completed STAB
3658 // buffers available for processing. If there are, we abort.
3659 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3660 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) {
3661 if (_cm->verbose_low()) {
3662 gclog_or_tty->print_cr("[%u] aborting to deal with pending SATB buffers",
3663 _worker_id);
3664 }
3665 // we do need to process SATB buffers, we'll abort and restart
3666 // the marking task to do so
3667 set_has_aborted();
3668 statsOnly( ++_aborted_satb );
3669 return;
3670 }
3671 }
3672
3673 void CMTask::recalculate_limits() {
3674 _real_words_scanned_limit = _words_scanned + words_scanned_period;
3675 _words_scanned_limit = _real_words_scanned_limit;
3676
3677 _real_refs_reached_limit = _refs_reached + refs_reached_period;
3678 _refs_reached_limit = _real_refs_reached_limit;
3679 }
3680
3681 void CMTask::decrease_limits() {
3682 // This is called when we believe that we're going to do an infrequent
3683 // operation which will increase the per byte scanned cost (i.e. move
3684 // entries to/from the global stack). It basically tries to decrease the
3685 // scanning limit so that the clock is called earlier.
3686
3687 if (_cm->verbose_medium()) {
3688 gclog_or_tty->print_cr("[%u] decreasing limits", _worker_id);
3689 }
3690
3691 _words_scanned_limit = _real_words_scanned_limit -
3692 3 * words_scanned_period / 4;
3693 _refs_reached_limit = _real_refs_reached_limit -
3694 3 * refs_reached_period / 4;
3695 }
3696
3697 void CMTask::move_entries_to_global_stack() {
3698 // local array where we'll store the entries that will be popped
3699 // from the local queue
3700 oop buffer[global_stack_transfer_size];
3701
3702 int n = 0;
3703 oop obj;
3704 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) {
3705 buffer[n] = obj;
3706 ++n;
3707 }
3708
3709 if (n > 0) {
3710 // we popped at least one entry from the local queue
3711
3712 statsOnly( ++_global_transfers_to; _local_pops += n );
3713
3714 if (!_cm->mark_stack_push(buffer, n)) {
3715 if (_cm->verbose_low()) {
3716 gclog_or_tty->print_cr("[%u] aborting due to global stack overflow",
3717 _worker_id);
3718 }
3719 set_has_aborted();
3720 } else {
3721 // the transfer was successful
3722
3723 if (_cm->verbose_medium()) {
3724 gclog_or_tty->print_cr("[%u] pushed %d entries to the global stack",
3725 _worker_id, n);
3726 }
3727 statsOnly( int tmp_size = _cm->mark_stack_size();
3728 if (tmp_size > _global_max_size) {
3729 _global_max_size = tmp_size;
3730 }
3731 _global_pushes += n );
3732 }
3733 }
3734
3735 // this operation was quite expensive, so decrease the limits
3736 decrease_limits();
3737 }
3738
3739 void CMTask::get_entries_from_global_stack() {
3740 // local array where we'll store the entries that will be popped
3741 // from the global stack.
3742 oop buffer[global_stack_transfer_size];
3743 int n;
3744 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n);
3745 assert(n <= global_stack_transfer_size,
3746 "we should not pop more than the given limit");
3747 if (n > 0) {
3748 // yes, we did actually pop at least one entry
3749
3750 statsOnly( ++_global_transfers_from; _global_pops += n );
3751 if (_cm->verbose_medium()) {
3752 gclog_or_tty->print_cr("[%u] popped %d entries from the global stack",
3753 _worker_id, n);
3754 }
3755 for (int i = 0; i < n; ++i) {
3756 bool success = _task_queue->push(buffer[i]);
3757 // We only call this when the local queue is empty or under a
3758 // given target limit. So, we do not expect this push to fail.
3759 assert(success, "invariant");
3760 }
3761
3762 statsOnly( int tmp_size = _task_queue->size();
3763 if (tmp_size > _local_max_size) {
3764 _local_max_size = tmp_size;
3765 }
3766 _local_pushes += n );
3767 }
3768
3769 // this operation was quite expensive, so decrease the limits
3770 decrease_limits();
3771 }
3772
3773 void CMTask::drain_local_queue(bool partially) {
3774 if (has_aborted()) return;
3775
3776 // Decide what the target size is, depending whether we're going to
3777 // drain it partially (so that other tasks can steal if they run out
3778 // of things to do) or totally (at the very end).
3779 size_t target_size;
3780 if (partially) {
3781 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize);
3782 } else {
3783 target_size = 0;
3784 }
3785
3786 if (_task_queue->size() > target_size) {
3787 if (_cm->verbose_high()) {
3788 gclog_or_tty->print_cr("[%u] draining local queue, target size = " SIZE_FORMAT,
3789 _worker_id, target_size);
3790 }
3791
3792 oop obj;
3793 bool ret = _task_queue->pop_local(obj);
3794 while (ret) {
3795 statsOnly( ++_local_pops );
3796
3797 if (_cm->verbose_high()) {
3798 gclog_or_tty->print_cr("[%u] popped "PTR_FORMAT, _worker_id,
3799 p2i((void*) obj));
3800 }
3801
3802 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" );
3803 assert(!_g1h->is_on_master_free_list(
3804 _g1h->heap_region_containing((HeapWord*) obj)), "invariant");
3805
3806 scan_object(obj);
3807
3808 if (_task_queue->size() <= target_size || has_aborted()) {
3809 ret = false;
3810 } else {
3811 ret = _task_queue->pop_local(obj);
3812 }
3813 }
3814
3815 if (_cm->verbose_high()) {
3816 gclog_or_tty->print_cr("[%u] drained local queue, size = %d",
3817 _worker_id, _task_queue->size());
3818 }
3819 }
3820 }
3821
3822 void CMTask::drain_global_stack(bool partially) {
3823 if (has_aborted()) return;
3824
3825 // We have a policy to drain the local queue before we attempt to
3826 // drain the global stack.
3827 assert(partially || _task_queue->size() == 0, "invariant");
3828
3829 // Decide what the target size is, depending whether we're going to
3830 // drain it partially (so that other tasks can steal if they run out
3831 // of things to do) or totally (at the very end). Notice that,
3832 // because we move entries from the global stack in chunks or
3833 // because another task might be doing the same, we might in fact
3834 // drop below the target. But, this is not a problem.
3835 size_t target_size;
3836 if (partially) {
3837 target_size = _cm->partial_mark_stack_size_target();
3838 } else {
3839 target_size = 0;
3840 }
3841
3842 if (_cm->mark_stack_size() > target_size) {
3843 if (_cm->verbose_low()) {
3844 gclog_or_tty->print_cr("[%u] draining global_stack, target size " SIZE_FORMAT,
3845 _worker_id, target_size);
3846 }
3847
3848 while (!has_aborted() && _cm->mark_stack_size() > target_size) {
3849 get_entries_from_global_stack();
3850 drain_local_queue(partially);
3851 }
3852
3853 if (_cm->verbose_low()) {
3854 gclog_or_tty->print_cr("[%u] drained global stack, size = " SIZE_FORMAT,
3855 _worker_id, _cm->mark_stack_size());
3856 }
3857 }
3858 }
3859
3860 // SATB Queue has several assumptions on whether to call the par or
3861 // non-par versions of the methods. this is why some of the code is
3862 // replicated. We should really get rid of the single-threaded version
3863 // of the code to simplify things.
3864 void CMTask::drain_satb_buffers() {
3865 if (has_aborted()) return;
3866
3867 // We set this so that the regular clock knows that we're in the
3868 // middle of draining buffers and doesn't set the abort flag when it
3869 // notices that SATB buffers are available for draining. It'd be
3870 // very counter productive if it did that. :-)
3871 _draining_satb_buffers = true;
3872
3873 CMObjectClosure oc(this);
3874 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set();
3875 if (G1CollectedHeap::use_parallel_gc_threads()) {
3876 satb_mq_set.set_par_closure(_worker_id, &oc);
3877 } else {
3878 satb_mq_set.set_closure(&oc);
3879 }
3880
3881 // This keeps claiming and applying the closure to completed buffers
3882 // until we run out of buffers or we need to abort.
3883 if (G1CollectedHeap::use_parallel_gc_threads()) {
3884 while (!has_aborted() &&
3885 satb_mq_set.par_apply_closure_to_completed_buffer(_worker_id)) {
3886 if (_cm->verbose_medium()) {
3887 gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
3888 }
3889 statsOnly( ++_satb_buffers_processed );
3890 regular_clock_call();
3891 }
3892 } else {
3893 while (!has_aborted() &&
3894 satb_mq_set.apply_closure_to_completed_buffer()) {
3895 if (_cm->verbose_medium()) {
3896 gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id);
3897 }
3898 statsOnly( ++_satb_buffers_processed );
3899 regular_clock_call();
3900 }
3901 }
3902
3903 if (!concurrent() && !has_aborted()) {
3904 // We should only do this during remark.
3905 if (G1CollectedHeap::use_parallel_gc_threads()) {
3906 satb_mq_set.par_iterate_closure_all_threads(_worker_id);
3907 } else {
3908 satb_mq_set.iterate_closure_all_threads();
3909 }
3910 }
3911
3912 _draining_satb_buffers = false;
3913
3914 assert(has_aborted() ||
3915 concurrent() ||
3916 satb_mq_set.completed_buffers_num() == 0, "invariant");
3917
3918 if (G1CollectedHeap::use_parallel_gc_threads()) {
3919 satb_mq_set.set_par_closure(_worker_id, NULL);
3920 } else {
3921 satb_mq_set.set_closure(NULL);
3922 }
3923
3924 // again, this was a potentially expensive operation, decrease the
3925 // limits to get the regular clock call early
3926 decrease_limits();
3927 }
3928
3929 void CMTask::print_stats() {
3930 gclog_or_tty->print_cr("Marking Stats, task = %u, calls = %d",
3931 _worker_id, _calls);
3932 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms",
3933 _elapsed_time_ms, _termination_time_ms);
3934 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3935 _step_times_ms.num(), _step_times_ms.avg(),
3936 _step_times_ms.sd());
3937 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3938 _step_times_ms.maximum(), _step_times_ms.sum());
3939
3940 #if _MARKING_STATS_
3941 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms",
3942 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(),
3943 _all_clock_intervals_ms.sd());
3944 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms",
3945 _all_clock_intervals_ms.maximum(),
3946 _all_clock_intervals_ms.sum());
3947 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d",
3948 _clock_due_to_scanning, _clock_due_to_marking);
3949 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d",
3950 _objs_scanned, _objs_found_on_bitmap);
3951 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d",
3952 _local_pushes, _local_pops, _local_max_size);
3953 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d",
3954 _global_pushes, _global_pops, _global_max_size);
3955 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d",
3956 _global_transfers_to,_global_transfers_from);
3957 gclog_or_tty->print_cr(" Regions: claimed = %d", _regions_claimed);
3958 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed);
3959 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d",
3960 _steal_attempts, _steals);
3961 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted);
3962 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d",
3963 _aborted_overflow, _aborted_cm_aborted, _aborted_yield);
3964 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d",
3965 _aborted_timed_out, _aborted_satb, _aborted_termination);
3966 #endif // _MARKING_STATS_
3967 }
3968
3969 /*****************************************************************************
3970
3971 The do_marking_step(time_target_ms, ...) method is the building
3972 block of the parallel marking framework. It can be called in parallel
3973 with other invocations of do_marking_step() on different tasks
3974 (but only one per task, obviously) and concurrently with the
3975 mutator threads, or during remark, hence it eliminates the need
3976 for two versions of the code. When called during remark, it will
3977 pick up from where the task left off during the concurrent marking
3978 phase. Interestingly, tasks are also claimable during evacuation
3979 pauses too, since do_marking_step() ensures that it aborts before
3980 it needs to yield.
3981
3982 The data structures that it uses to do marking work are the
3983 following:
3984
3985 (1) Marking Bitmap. If there are gray objects that appear only
3986 on the bitmap (this happens either when dealing with an overflow
3987 or when the initial marking phase has simply marked the roots
3988 and didn't push them on the stack), then tasks claim heap
3989 regions whose bitmap they then scan to find gray objects. A
3990 global finger indicates where the end of the last claimed region
3991 is. A local finger indicates how far into the region a task has
3992 scanned. The two fingers are used to determine how to gray an
3993 object (i.e. whether simply marking it is OK, as it will be
3994 visited by a task in the future, or whether it needs to be also
3995 pushed on a stack).
3996
3997 (2) Local Queue. The local queue of the task which is accessed
3998 reasonably efficiently by the task. Other tasks can steal from
3999 it when they run out of work. Throughout the marking phase, a
4000 task attempts to keep its local queue short but not totally
4001 empty, so that entries are available for stealing by other
4002 tasks. Only when there is no more work, a task will totally
4003 drain its local queue.
4004
4005 (3) Global Mark Stack. This handles local queue overflow. During
4006 marking only sets of entries are moved between it and the local
4007 queues, as access to it requires a mutex and more fine-grain
4008 interaction with it which might cause contention. If it
4009 overflows, then the marking phase should restart and iterate
4010 over the bitmap to identify gray objects. Throughout the marking
4011 phase, tasks attempt to keep the global mark stack at a small
4012 length but not totally empty, so that entries are available for
4013 popping by other tasks. Only when there is no more work, tasks
4014 will totally drain the global mark stack.
4015
4016 (4) SATB Buffer Queue. This is where completed SATB buffers are
4017 made available. Buffers are regularly removed from this queue
4018 and scanned for roots, so that the queue doesn't get too
4019 long. During remark, all completed buffers are processed, as
4020 well as the filled in parts of any uncompleted buffers.
4021
4022 The do_marking_step() method tries to abort when the time target
4023 has been reached. There are a few other cases when the
4024 do_marking_step() method also aborts:
4025
4026 (1) When the marking phase has been aborted (after a Full GC).
4027
4028 (2) When a global overflow (on the global stack) has been
4029 triggered. Before the task aborts, it will actually sync up with
4030 the other tasks to ensure that all the marking data structures
4031 (local queues, stacks, fingers etc.) are re-initialized so that
4032 when do_marking_step() completes, the marking phase can
4033 immediately restart.
4034
4035 (3) When enough completed SATB buffers are available. The
4036 do_marking_step() method only tries to drain SATB buffers right
4037 at the beginning. So, if enough buffers are available, the
4038 marking step aborts and the SATB buffers are processed at
4039 the beginning of the next invocation.
4040
4041 (4) To yield. when we have to yield then we abort and yield
4042 right at the end of do_marking_step(). This saves us from a lot
4043 of hassle as, by yielding we might allow a Full GC. If this
4044 happens then objects will be compacted underneath our feet, the
4045 heap might shrink, etc. We save checking for this by just
4046 aborting and doing the yield right at the end.
4047
4048 From the above it follows that the do_marking_step() method should
4049 be called in a loop (or, otherwise, regularly) until it completes.
4050
4051 If a marking step completes without its has_aborted() flag being
4052 true, it means it has completed the current marking phase (and
4053 also all other marking tasks have done so and have all synced up).
4054
4055 A method called regular_clock_call() is invoked "regularly" (in
4056 sub ms intervals) throughout marking. It is this clock method that
4057 checks all the abort conditions which were mentioned above and
4058 decides when the task should abort. A work-based scheme is used to
4059 trigger this clock method: when the number of object words the
4060 marking phase has scanned or the number of references the marking
4061 phase has visited reach a given limit. Additional invocations to
4062 the method clock have been planted in a few other strategic places
4063 too. The initial reason for the clock method was to avoid calling
4064 vtime too regularly, as it is quite expensive. So, once it was in
4065 place, it was natural to piggy-back all the other conditions on it
4066 too and not constantly check them throughout the code.
4067
4068 If do_termination is true then do_marking_step will enter its
4069 termination protocol.
4070
4071 The value of is_serial must be true when do_marking_step is being
4072 called serially (i.e. by the VMThread) and do_marking_step should
4073 skip any synchronization in the termination and overflow code.
4074 Examples include the serial remark code and the serial reference
4075 processing closures.
4076
4077 The value of is_serial must be false when do_marking_step is
4078 being called by any of the worker threads in a work gang.
4079 Examples include the concurrent marking code (CMMarkingTask),
4080 the MT remark code, and the MT reference processing closures.
4081
4082 *****************************************************************************/
4083
4084 void CMTask::do_marking_step(double time_target_ms,
4085 bool do_termination,
4086 bool is_serial) {
4087 assert(time_target_ms >= 1.0, "minimum granularity is 1ms");
4088 assert(concurrent() == _cm->concurrent(), "they should be the same");
4089
4090 G1CollectorPolicy* g1_policy = _g1h->g1_policy();
4091 assert(_task_queues != NULL, "invariant");
4092 assert(_task_queue != NULL, "invariant");
4093 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant");
4094
4095 assert(!_claimed,
4096 "only one thread should claim this task at any one time");
4097
4098 // OK, this doesn't safeguard again all possible scenarios, as it is
4099 // possible for two threads to set the _claimed flag at the same
4100 // time. But it is only for debugging purposes anyway and it will
4101 // catch most problems.
4102 _claimed = true;
4103
4104 _start_time_ms = os::elapsedVTime() * 1000.0;
4105 statsOnly( _interval_start_time_ms = _start_time_ms );
4106
4107 // If do_stealing is true then do_marking_step will attempt to
4108 // steal work from the other CMTasks. It only makes sense to
4109 // enable stealing when the termination protocol is enabled
4110 // and do_marking_step() is not being called serially.
4111 bool do_stealing = do_termination && !is_serial;
4112
4113 double diff_prediction_ms =
4114 g1_policy->get_new_prediction(&_marking_step_diffs_ms);
4115 _time_target_ms = time_target_ms - diff_prediction_ms;
4116
4117 // set up the variables that are used in the work-based scheme to
4118 // call the regular clock method
4119 _words_scanned = 0;
4120 _refs_reached = 0;
4121 recalculate_limits();
4122
4123 // clear all flags
4124 clear_has_aborted();
4125 _has_timed_out = false;
4126 _draining_satb_buffers = false;
4127
4128 ++_calls;
4129
4130 if (_cm->verbose_low()) {
4131 gclog_or_tty->print_cr("[%u] >>>>>>>>>> START, call = %d, "
4132 "target = %1.2lfms >>>>>>>>>>",
4133 _worker_id, _calls, _time_target_ms);
4134 }
4135
4136 // Set up the bitmap and oop closures. Anything that uses them is
4137 // eventually called from this method, so it is OK to allocate these
4138 // statically.
4139 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap);
4140 G1CMOopClosure cm_oop_closure(_g1h, _cm, this);
4141 set_cm_oop_closure(&cm_oop_closure);
4142
4143 if (_cm->has_overflown()) {
4144 // This can happen if the mark stack overflows during a GC pause
4145 // and this task, after a yield point, restarts. We have to abort
4146 // as we need to get into the overflow protocol which happens
4147 // right at the end of this task.
4148 set_has_aborted();
4149 }
4150
4151 // First drain any available SATB buffers. After this, we will not
4152 // look at SATB buffers before the next invocation of this method.
4153 // If enough completed SATB buffers are queued up, the regular clock
4154 // will abort this task so that it restarts.
4155 drain_satb_buffers();
4156 // ...then partially drain the local queue and the global stack
4157 drain_local_queue(true);
4158 drain_global_stack(true);
4159
4160 do {
4161 if (!has_aborted() && _curr_region != NULL) {
4162 // This means that we're already holding on to a region.
4163 assert(_finger != NULL, "if region is not NULL, then the finger "
4164 "should not be NULL either");
4165
4166 // We might have restarted this task after an evacuation pause
4167 // which might have evacuated the region we're holding on to
4168 // underneath our feet. Let's read its limit again to make sure
4169 // that we do not iterate over a region of the heap that
4170 // contains garbage (update_region_limit() will also move
4171 // _finger to the start of the region if it is found empty).
4172 update_region_limit();
4173 // We will start from _finger not from the start of the region,
4174 // as we might be restarting this task after aborting half-way
4175 // through scanning this region. In this case, _finger points to
4176 // the address where we last found a marked object. If this is a
4177 // fresh region, _finger points to start().
4178 MemRegion mr = MemRegion(_finger, _region_limit);
4179
4180 if (_cm->verbose_low()) {
4181 gclog_or_tty->print_cr("[%u] we're scanning part "
4182 "["PTR_FORMAT", "PTR_FORMAT") "
4183 "of region "HR_FORMAT,
4184 _worker_id, p2i(_finger), p2i(_region_limit),
4185 HR_FORMAT_PARAMS(_curr_region));
4186 }
4187
4188 assert(!_curr_region->isHumongous() || mr.start() == _curr_region->bottom(),
4189 "humongous regions should go around loop once only");
4190
4191 // Some special cases:
4192 // If the memory region is empty, we can just give up the region.
4193 // If the current region is humongous then we only need to check
4194 // the bitmap for the bit associated with the start of the object,
4195 // scan the object if it's live, and give up the region.
4196 // Otherwise, let's iterate over the bitmap of the part of the region
4197 // that is left.
4198 // If the iteration is successful, give up the region.
4199 if (mr.is_empty()) {
4200 giveup_current_region();
4201 regular_clock_call();
4202 } else if (_curr_region->isHumongous() && mr.start() == _curr_region->bottom()) {
4203 if (_nextMarkBitMap->isMarked(mr.start())) {
4204 // The object is marked - apply the closure
4205 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start());
4206 bitmap_closure.do_bit(offset);
4207 }
4208 // Even if this task aborted while scanning the humongous object
4209 // we can (and should) give up the current region.
4210 giveup_current_region();
4211 regular_clock_call();
4212 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) {
4213 giveup_current_region();
4214 regular_clock_call();
4215 } else {
4216 assert(has_aborted(), "currently the only way to do so");
4217 // The only way to abort the bitmap iteration is to return
4218 // false from the do_bit() method. However, inside the
4219 // do_bit() method we move the _finger to point to the
4220 // object currently being looked at. So, if we bail out, we
4221 // have definitely set _finger to something non-null.
4222 assert(_finger != NULL, "invariant");
4223
4224 // Region iteration was actually aborted. So now _finger
4225 // points to the address of the object we last scanned. If we
4226 // leave it there, when we restart this task, we will rescan
4227 // the object. It is easy to avoid this. We move the finger by
4228 // enough to point to the next possible object header (the
4229 // bitmap knows by how much we need to move it as it knows its
4230 // granularity).
4231 assert(_finger < _region_limit, "invariant");
4232 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger);
4233 // Check if bitmap iteration was aborted while scanning the last object
4234 if (new_finger >= _region_limit) {
4235 giveup_current_region();
4236 } else {
4237 move_finger_to(new_finger);
4238 }
4239 }
4240 }
4241 // At this point we have either completed iterating over the
4242 // region we were holding on to, or we have aborted.
4243
4244 // We then partially drain the local queue and the global stack.
4245 // (Do we really need this?)
4246 drain_local_queue(true);
4247 drain_global_stack(true);
4248
4249 // Read the note on the claim_region() method on why it might
4250 // return NULL with potentially more regions available for
4251 // claiming and why we have to check out_of_regions() to determine
4252 // whether we're done or not.
4253 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) {
4254 // We are going to try to claim a new region. We should have
4255 // given up on the previous one.
4256 // Separated the asserts so that we know which one fires.
4257 assert(_curr_region == NULL, "invariant");
4258 assert(_finger == NULL, "invariant");
4259 assert(_region_limit == NULL, "invariant");
4260 if (_cm->verbose_low()) {
4261 gclog_or_tty->print_cr("[%u] trying to claim a new region", _worker_id);
4262 }
4263 HeapRegion* claimed_region = _cm->claim_region(_worker_id);
4264 if (claimed_region != NULL) {
4265 // Yes, we managed to claim one
4266 statsOnly( ++_regions_claimed );
4267
4268 if (_cm->verbose_low()) {
4269 gclog_or_tty->print_cr("[%u] we successfully claimed "
4270 "region "PTR_FORMAT,
4271 _worker_id, p2i(claimed_region));
4272 }
4273
4274 setup_for_region(claimed_region);
4275 assert(_curr_region == claimed_region, "invariant");
4276 }
4277 // It is important to call the regular clock here. It might take
4278 // a while to claim a region if, for example, we hit a large
4279 // block of empty regions. So we need to call the regular clock
4280 // method once round the loop to make sure it's called
4281 // frequently enough.
4282 regular_clock_call();
4283 }
4284
4285 if (!has_aborted() && _curr_region == NULL) {
4286 assert(_cm->out_of_regions(),
4287 "at this point we should be out of regions");
4288 }
4289 } while ( _curr_region != NULL && !has_aborted());
4290
4291 if (!has_aborted()) {
4292 // We cannot check whether the global stack is empty, since other
4293 // tasks might be pushing objects to it concurrently.
4294 assert(_cm->out_of_regions(),
4295 "at this point we should be out of regions");
4296
4297 if (_cm->verbose_low()) {
4298 gclog_or_tty->print_cr("[%u] all regions claimed", _worker_id);
4299 }
4300
4301 // Try to reduce the number of available SATB buffers so that
4302 // remark has less work to do.
4303 drain_satb_buffers();
4304 }
4305
4306 // Since we've done everything else, we can now totally drain the
4307 // local queue and global stack.
4308 drain_local_queue(false);
4309 drain_global_stack(false);
4310
4311 // Attempt at work stealing from other task's queues.
4312 if (do_stealing && !has_aborted()) {
4313 // We have not aborted. This means that we have finished all that
4314 // we could. Let's try to do some stealing...
4315
4316 // We cannot check whether the global stack is empty, since other
4317 // tasks might be pushing objects to it concurrently.
4318 assert(_cm->out_of_regions() && _task_queue->size() == 0,
4319 "only way to reach here");
4320
4321 if (_cm->verbose_low()) {
4322 gclog_or_tty->print_cr("[%u] starting to steal", _worker_id);
4323 }
4324
4325 while (!has_aborted()) {
4326 oop obj;
4327 statsOnly( ++_steal_attempts );
4328
4329 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) {
4330 if (_cm->verbose_medium()) {
4331 gclog_or_tty->print_cr("[%u] stolen "PTR_FORMAT" successfully",
4332 _worker_id, p2i((void*) obj));
4333 }
4334
4335 statsOnly( ++_steals );
4336
4337 assert(_nextMarkBitMap->isMarked((HeapWord*) obj),
4338 "any stolen object should be marked");
4339 scan_object(obj);
4340
4341 // And since we're towards the end, let's totally drain the
4342 // local queue and global stack.
4343 drain_local_queue(false);
4344 drain_global_stack(false);
4345 } else {
4346 break;
4347 }
4348 }
4349 }
4350
4351 // If we are about to wrap up and go into termination, check if we
4352 // should raise the overflow flag.
4353 if (do_termination && !has_aborted()) {
4354 if (_cm->force_overflow()->should_force()) {
4355 _cm->set_has_overflown();
4356 regular_clock_call();
4357 }
4358 }
4359
4360 // We still haven't aborted. Now, let's try to get into the
4361 // termination protocol.
4362 if (do_termination && !has_aborted()) {
4363 // We cannot check whether the global stack is empty, since other
4364 // tasks might be concurrently pushing objects on it.
4365 // Separated the asserts so that we know which one fires.
4366 assert(_cm->out_of_regions(), "only way to reach here");
4367 assert(_task_queue->size() == 0, "only way to reach here");
4368
4369 if (_cm->verbose_low()) {
4370 gclog_or_tty->print_cr("[%u] starting termination protocol", _worker_id);
4371 }
4372
4373 _termination_start_time_ms = os::elapsedVTime() * 1000.0;
4374
4375 // The CMTask class also extends the TerminatorTerminator class,
4376 // hence its should_exit_termination() method will also decide
4377 // whether to exit the termination protocol or not.
4378 bool finished = (is_serial ||
4379 _cm->terminator()->offer_termination(this));
4380 double termination_end_time_ms = os::elapsedVTime() * 1000.0;
4381 _termination_time_ms +=
4382 termination_end_time_ms - _termination_start_time_ms;
4383
4384 if (finished) {
4385 // We're all done.
4386
4387 if (_worker_id == 0) {
4388 // let's allow task 0 to do this
4389 if (concurrent()) {
4390 assert(_cm->concurrent_marking_in_progress(), "invariant");
4391 // we need to set this to false before the next
4392 // safepoint. This way we ensure that the marking phase
4393 // doesn't observe any more heap expansions.
4394 _cm->clear_concurrent_marking_in_progress();
4395 }
4396 }
4397
4398 // We can now guarantee that the global stack is empty, since
4399 // all other tasks have finished. We separated the guarantees so
4400 // that, if a condition is false, we can immediately find out
4401 // which one.
4402 guarantee(_cm->out_of_regions(), "only way to reach here");
4403 guarantee(_cm->mark_stack_empty(), "only way to reach here");
4404 guarantee(_task_queue->size() == 0, "only way to reach here");
4405 guarantee(!_cm->has_overflown(), "only way to reach here");
4406 guarantee(!_cm->mark_stack_overflow(), "only way to reach here");
4407
4408 if (_cm->verbose_low()) {
4409 gclog_or_tty->print_cr("[%u] all tasks terminated", _worker_id);
4410 }
4411 } else {
4412 // Apparently there's more work to do. Let's abort this task. It
4413 // will restart it and we can hopefully find more things to do.
4414
4415 if (_cm->verbose_low()) {
4416 gclog_or_tty->print_cr("[%u] apparently there is more work to do",
4417 _worker_id);
4418 }
4419
4420 set_has_aborted();
4421 statsOnly( ++_aborted_termination );
4422 }
4423 }
4424
4425 // Mainly for debugging purposes to make sure that a pointer to the
4426 // closure which was statically allocated in this frame doesn't
4427 // escape it by accident.
4428 set_cm_oop_closure(NULL);
4429 double end_time_ms = os::elapsedVTime() * 1000.0;
4430 double elapsed_time_ms = end_time_ms - _start_time_ms;
4431 // Update the step history.
4432 _step_times_ms.add(elapsed_time_ms);
4433
4434 if (has_aborted()) {
4435 // The task was aborted for some reason.
4436
4437 statsOnly( ++_aborted );
4438
4439 if (_has_timed_out) {
4440 double diff_ms = elapsed_time_ms - _time_target_ms;
4441 // Keep statistics of how well we did with respect to hitting
4442 // our target only if we actually timed out (if we aborted for
4443 // other reasons, then the results might get skewed).
4444 _marking_step_diffs_ms.add(diff_ms);
4445 }
4446
4447 if (_cm->has_overflown()) {
4448 // This is the interesting one. We aborted because a global
4449 // overflow was raised. This means we have to restart the
4450 // marking phase and start iterating over regions. However, in
4451 // order to do this we have to make sure that all tasks stop
4452 // what they are doing and re-initialise in a safe manner. We
4453 // will achieve this with the use of two barrier sync points.
4454
4455 if (_cm->verbose_low()) {
4456 gclog_or_tty->print_cr("[%u] detected overflow", _worker_id);
4457 }
4458
4459 if (!is_serial) {
4460 // We only need to enter the sync barrier if being called
4461 // from a parallel context
4462 _cm->enter_first_sync_barrier(_worker_id);
4463
4464 // When we exit this sync barrier we know that all tasks have
4465 // stopped doing marking work. So, it's now safe to
4466 // re-initialise our data structures. At the end of this method,
4467 // task 0 will clear the global data structures.
4468 }
4469
4470 statsOnly( ++_aborted_overflow );
4471
4472 // We clear the local state of this task...
4473 clear_region_fields();
4474
4475 if (!is_serial) {
4476 // ...and enter the second barrier.
4477 _cm->enter_second_sync_barrier(_worker_id);
4478 }
4479 // At this point, if we're during the concurrent phase of
4480 // marking, everything has been re-initialized and we're
4481 // ready to restart.
4482 }
4483
4484 if (_cm->verbose_low()) {
4485 gclog_or_tty->print_cr("[%u] <<<<<<<<<< ABORTING, target = %1.2lfms, "
4486 "elapsed = %1.2lfms <<<<<<<<<<",
4487 _worker_id, _time_target_ms, elapsed_time_ms);
4488 if (_cm->has_aborted()) {
4489 gclog_or_tty->print_cr("[%u] ========== MARKING ABORTED ==========",
4490 _worker_id);
4491 }
4492 }
4493 } else {
4494 if (_cm->verbose_low()) {
4495 gclog_or_tty->print_cr("[%u] <<<<<<<<<< FINISHED, target = %1.2lfms, "
4496 "elapsed = %1.2lfms <<<<<<<<<<",
4497 _worker_id, _time_target_ms, elapsed_time_ms);
4498 }
4499 }
4500
4501 _claimed = false;
4502 }
4503
4504 CMTask::CMTask(uint worker_id,
4505 ConcurrentMark* cm,
4506 size_t* marked_bytes,
4507 BitMap* card_bm,
4508 CMTaskQueue* task_queue,
4509 CMTaskQueueSet* task_queues)
4510 : _g1h(G1CollectedHeap::heap()),
4511 _worker_id(worker_id), _cm(cm),
4512 _claimed(false),
4513 _nextMarkBitMap(NULL), _hash_seed(17),
4514 _task_queue(task_queue),
4515 _task_queues(task_queues),
4516 _cm_oop_closure(NULL),
4517 _marked_bytes_array(marked_bytes),
4518 _card_bm(card_bm) {
4519 guarantee(task_queue != NULL, "invariant");
4520 guarantee(task_queues != NULL, "invariant");
4521
4522 statsOnly( _clock_due_to_scanning = 0;
4523 _clock_due_to_marking = 0 );
4524
4525 _marking_step_diffs_ms.add(0.5);
4526 }
4527
4528 // These are formatting macros that are used below to ensure
4529 // consistent formatting. The *_H_* versions are used to format the
4530 // header for a particular value and they should be kept consistent
4531 // with the corresponding macro. Also note that most of the macros add
4532 // the necessary white space (as a prefix) which makes them a bit
4533 // easier to compose.
4534
4535 // All the output lines are prefixed with this string to be able to
4536 // identify them easily in a large log file.
4537 #define G1PPRL_LINE_PREFIX "###"
4538
4539 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT
4540 #ifdef _LP64
4541 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s"
4542 #else // _LP64
4543 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s"
4544 #endif // _LP64
4545
4546 // For per-region info
4547 #define G1PPRL_TYPE_FORMAT " %-4s"
4548 #define G1PPRL_TYPE_H_FORMAT " %4s"
4549 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9)
4550 #define G1PPRL_BYTE_H_FORMAT " %9s"
4551 #define G1PPRL_DOUBLE_FORMAT " %14.1f"
4552 #define G1PPRL_DOUBLE_H_FORMAT " %14s"
4553
4554 // For summary info
4555 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT
4556 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT
4557 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB"
4558 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%"
4559
4560 G1PrintRegionLivenessInfoClosure::
4561 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name)
4562 : _out(out),
4563 _total_used_bytes(0), _total_capacity_bytes(0),
4564 _total_prev_live_bytes(0), _total_next_live_bytes(0),
4565 _hum_used_bytes(0), _hum_capacity_bytes(0),
4566 _hum_prev_live_bytes(0), _hum_next_live_bytes(0),
4567 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) {
4568 G1CollectedHeap* g1h = G1CollectedHeap::heap();
4569 MemRegion g1_committed = g1h->g1_committed();
4570 MemRegion g1_reserved = g1h->g1_reserved();
4571 double now = os::elapsedTime();
4572
4573 // Print the header of the output.
4574 _out->cr();
4575 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now);
4576 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP"
4577 G1PPRL_SUM_ADDR_FORMAT("committed")
4578 G1PPRL_SUM_ADDR_FORMAT("reserved")
4579 G1PPRL_SUM_BYTE_FORMAT("region-size"),
4580 p2i(g1_committed.start()), p2i(g1_committed.end()),
4581 p2i(g1_reserved.start()), p2i(g1_reserved.end()),
4582 HeapRegion::GrainBytes);
4583 _out->print_cr(G1PPRL_LINE_PREFIX);
4584 _out->print_cr(G1PPRL_LINE_PREFIX
4585 G1PPRL_TYPE_H_FORMAT
4586 G1PPRL_ADDR_BASE_H_FORMAT
4587 G1PPRL_BYTE_H_FORMAT
4588 G1PPRL_BYTE_H_FORMAT
4589 G1PPRL_BYTE_H_FORMAT
4590 G1PPRL_DOUBLE_H_FORMAT
4591 G1PPRL_BYTE_H_FORMAT
4592 G1PPRL_BYTE_H_FORMAT,
4593 "type", "address-range",
4594 "used", "prev-live", "next-live", "gc-eff",
4595 "remset", "code-roots");
4596 _out->print_cr(G1PPRL_LINE_PREFIX
4597 G1PPRL_TYPE_H_FORMAT
4598 G1PPRL_ADDR_BASE_H_FORMAT
4599 G1PPRL_BYTE_H_FORMAT
4600 G1PPRL_BYTE_H_FORMAT
4601 G1PPRL_BYTE_H_FORMAT
4602 G1PPRL_DOUBLE_H_FORMAT
4603 G1PPRL_BYTE_H_FORMAT
4604 G1PPRL_BYTE_H_FORMAT,
4605 "", "",
4606 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)",
4607 "(bytes)", "(bytes)");
4608 }
4609
4610 // It takes as a parameter a reference to one of the _hum_* fields, it
4611 // deduces the corresponding value for a region in a humongous region
4612 // series (either the region size, or what's left if the _hum_* field
4613 // is < the region size), and updates the _hum_* field accordingly.
4614 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) {
4615 size_t bytes = 0;
4616 // The > 0 check is to deal with the prev and next live bytes which
4617 // could be 0.
4618 if (*hum_bytes > 0) {
4619 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes);
4620 *hum_bytes -= bytes;
4621 }
4622 return bytes;
4623 }
4624
4625 // It deduces the values for a region in a humongous region series
4626 // from the _hum_* fields and updates those accordingly. It assumes
4627 // that that _hum_* fields have already been set up from the "starts
4628 // humongous" region and we visit the regions in address order.
4629 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes,
4630 size_t* capacity_bytes,
4631 size_t* prev_live_bytes,
4632 size_t* next_live_bytes) {
4633 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition");
4634 *used_bytes = get_hum_bytes(&_hum_used_bytes);
4635 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes);
4636 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes);
4637 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes);
4638 }
4639
4640 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) {
4641 const char* type = "";
4642 HeapWord* bottom = r->bottom();
4643 HeapWord* end = r->end();
4644 size_t capacity_bytes = r->capacity();
4645 size_t used_bytes = r->used();
4646 size_t prev_live_bytes = r->live_bytes();
4647 size_t next_live_bytes = r->next_live_bytes();
4648 double gc_eff = r->gc_efficiency();
4649 size_t remset_bytes = r->rem_set()->mem_size();
4650 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size();
4651
4652 if (r->used() == 0) {
4653 type = "FREE";
4654 } else if (r->is_survivor()) {
4655 type = "SURV";
4656 } else if (r->is_young()) {
4657 type = "EDEN";
4658 } else if (r->startsHumongous()) {
4659 type = "HUMS";
4660
4661 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 &&
4662 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0,
4663 "they should have been zeroed after the last time we used them");
4664 // Set up the _hum_* fields.
4665 _hum_capacity_bytes = capacity_bytes;
4666 _hum_used_bytes = used_bytes;
4667 _hum_prev_live_bytes = prev_live_bytes;
4668 _hum_next_live_bytes = next_live_bytes;
4669 get_hum_bytes(&used_bytes, &capacity_bytes,
4670 &prev_live_bytes, &next_live_bytes);
4671 end = bottom + HeapRegion::GrainWords;
4672 } else if (r->continuesHumongous()) {
4673 type = "HUMC";
4674 get_hum_bytes(&used_bytes, &capacity_bytes,
4675 &prev_live_bytes, &next_live_bytes);
4676 assert(end == bottom + HeapRegion::GrainWords, "invariant");
4677 } else {
4678 type = "OLD";
4679 }
4680
4681 _total_used_bytes += used_bytes;
4682 _total_capacity_bytes += capacity_bytes;
4683 _total_prev_live_bytes += prev_live_bytes;
4684 _total_next_live_bytes += next_live_bytes;
4685 _total_remset_bytes += remset_bytes;
4686 _total_strong_code_roots_bytes += strong_code_roots_bytes;
4687
4688 // Print a line for this particular region.
4689 _out->print_cr(G1PPRL_LINE_PREFIX
4690 G1PPRL_TYPE_FORMAT
4691 G1PPRL_ADDR_BASE_FORMAT
4692 G1PPRL_BYTE_FORMAT
4693 G1PPRL_BYTE_FORMAT
4694 G1PPRL_BYTE_FORMAT
4695 G1PPRL_DOUBLE_FORMAT
4696 G1PPRL_BYTE_FORMAT
4697 G1PPRL_BYTE_FORMAT,
4698 type, p2i(bottom), p2i(end),
4699 used_bytes, prev_live_bytes, next_live_bytes, gc_eff,
4700 remset_bytes, strong_code_roots_bytes);
4701
4702 return false;
4703 }
4704
4705 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() {
4706 // add static memory usages to remembered set sizes
4707 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size();
4708 // Print the footer of the output.
4709 _out->print_cr(G1PPRL_LINE_PREFIX);
4710 _out->print_cr(G1PPRL_LINE_PREFIX
4711 " SUMMARY"
4712 G1PPRL_SUM_MB_FORMAT("capacity")
4713 G1PPRL_SUM_MB_PERC_FORMAT("used")
4714 G1PPRL_SUM_MB_PERC_FORMAT("prev-live")
4715 G1PPRL_SUM_MB_PERC_FORMAT("next-live")
4716 G1PPRL_SUM_MB_FORMAT("remset")
4717 G1PPRL_SUM_MB_FORMAT("code-roots"),
4718 bytes_to_mb(_total_capacity_bytes),
4719 bytes_to_mb(_total_used_bytes),
4720 perc(_total_used_bytes, _total_capacity_bytes),
4721 bytes_to_mb(_total_prev_live_bytes),
4722 perc(_total_prev_live_bytes, _total_capacity_bytes),
4723 bytes_to_mb(_total_next_live_bytes),
4724 perc(_total_next_live_bytes, _total_capacity_bytes),
4725 bytes_to_mb(_total_remset_bytes),
4726 bytes_to_mb(_total_strong_code_roots_bytes));
4727 _out->cr();
4728 }