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 "code/codeCache.hpp" 28 #include "gc_implementation/g1/concurrentMark.inline.hpp" 29 #include "gc_implementation/g1/concurrentMarkThread.inline.hpp" 30 #include "gc_implementation/g1/g1CollectedHeap.inline.hpp" 31 #include "gc_implementation/g1/g1CollectorPolicy.hpp" 32 #include "gc_implementation/g1/g1ErgoVerbose.hpp" 33 #include "gc_implementation/g1/g1Log.hpp" 34 #include "gc_implementation/g1/g1OopClosures.inline.hpp" 35 #include "gc_implementation/g1/g1RemSet.hpp" 36 #include "gc_implementation/g1/heapRegion.inline.hpp" 37 #include "gc_implementation/g1/heapRegionManager.inline.hpp" 38 #include "gc_implementation/g1/heapRegionRemSet.hpp" 39 #include "gc_implementation/g1/heapRegionSet.inline.hpp" 40 #include "gc_implementation/shared/vmGCOperations.hpp" 41 #include "gc_implementation/shared/gcTimer.hpp" 42 #include "gc_implementation/shared/gcTrace.hpp" 43 #include "gc_implementation/shared/gcTraceTime.hpp" 44 #include "memory/allocation.hpp" 45 #include "memory/genOopClosures.inline.hpp" 46 #include "memory/referencePolicy.hpp" 47 #include "memory/resourceArea.hpp" 48 #include "oops/oop.inline.hpp" 49 #include "runtime/handles.inline.hpp" 50 #include "runtime/java.hpp" 51 #include "runtime/atomic.inline.hpp" 52 #include "runtime/prefetch.inline.hpp" 53 #include "services/memTracker.hpp" 54 55 // Concurrent marking bit map wrapper 56 57 CMBitMapRO::CMBitMapRO(int shifter) : 58 _bm(), 59 _shifter(shifter) { 60 _bmStartWord = 0; 61 _bmWordSize = 0; 62 } 63 64 HeapWord* CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr, 65 const HeapWord* limit) const { 66 // First we must round addr *up* to a possible object boundary. 67 addr = (HeapWord*)align_size_up((intptr_t)addr, 68 HeapWordSize << _shifter); 69 size_t addrOffset = heapWordToOffset(addr); 70 if (limit == NULL) { 71 limit = _bmStartWord + _bmWordSize; 72 } 73 size_t limitOffset = heapWordToOffset(limit); 74 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset); 75 HeapWord* nextAddr = offsetToHeapWord(nextOffset); 76 assert(nextAddr >= addr, "get_next_one postcondition"); 77 assert(nextAddr == limit || isMarked(nextAddr), 78 "get_next_one postcondition"); 79 return nextAddr; 80 } 81 82 HeapWord* CMBitMapRO::getNextUnmarkedWordAddress(const HeapWord* addr, 83 const HeapWord* limit) const { 84 size_t addrOffset = heapWordToOffset(addr); 85 if (limit == NULL) { 86 limit = _bmStartWord + _bmWordSize; 87 } 88 size_t limitOffset = heapWordToOffset(limit); 89 size_t nextOffset = _bm.get_next_zero_offset(addrOffset, limitOffset); 90 HeapWord* nextAddr = offsetToHeapWord(nextOffset); 91 assert(nextAddr >= addr, "get_next_one postcondition"); 92 assert(nextAddr == limit || !isMarked(nextAddr), 93 "get_next_one postcondition"); 94 return nextAddr; 95 } 96 97 int CMBitMapRO::heapWordDiffToOffsetDiff(size_t diff) const { 98 assert((diff & ((1 << _shifter) - 1)) == 0, "argument check"); 99 return (int) (diff >> _shifter); 100 } 101 102 #ifndef PRODUCT 103 bool CMBitMapRO::covers(MemRegion heap_rs) const { 104 // assert(_bm.map() == _virtual_space.low(), "map inconsistency"); 105 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize, 106 "size inconsistency"); 107 return _bmStartWord == (HeapWord*)(heap_rs.start()) && 108 _bmWordSize == heap_rs.word_size(); 109 } 110 #endif 111 112 void CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const { 113 _bm.print_on_error(st, prefix); 114 } 115 116 size_t CMBitMap::compute_size(size_t heap_size) { 117 return heap_size / mark_distance(); 118 } 119 120 size_t CMBitMap::mark_distance() { 121 return MinObjAlignmentInBytes * BitsPerByte; 122 } 123 124 void CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) { 125 _bmStartWord = heap.start(); 126 _bmWordSize = heap.word_size(); 127 128 _bm.set_map((BitMap::bm_word_t*) storage->reserved().start()); 129 _bm.set_size(_bmWordSize >> _shifter); 130 131 storage->set_mapping_changed_listener(&_listener); 132 } 133 134 void CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) { 135 if (zero_filled) { 136 return; 137 } 138 // We need to clear the bitmap on commit, removing any existing information. 139 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords); 140 _bm->clearRange(mr); 141 } 142 143 // Closure used for clearing the given mark bitmap. 144 class ClearBitmapHRClosure : public HeapRegionClosure { 145 private: 146 ConcurrentMark* _cm; 147 CMBitMap* _bitmap; 148 bool _may_yield; // The closure may yield during iteration. If yielded, abort the iteration. 149 public: 150 ClearBitmapHRClosure(ConcurrentMark* cm, CMBitMap* bitmap, bool may_yield) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap), _may_yield(may_yield) { 151 assert(!may_yield || cm != NULL, "CM must be non-NULL if this closure is expected to yield."); 152 } 153 154 virtual bool doHeapRegion(HeapRegion* r) { 155 size_t const chunk_size_in_words = M / HeapWordSize; 156 157 HeapWord* cur = r->bottom(); 158 HeapWord* const end = r->end(); 159 160 while (cur < end) { 161 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end)); 162 _bitmap->clearRange(mr); 163 164 cur += chunk_size_in_words; 165 166 // Abort iteration if after yielding the marking has been aborted. 167 if (_may_yield && _cm->do_yield_check() && _cm->has_aborted()) { 168 return true; 169 } 170 // Repeat the asserts from before the start of the closure. We will do them 171 // as asserts here to minimize their overhead on the product. However, we 172 // will have them as guarantees at the beginning / end of the bitmap 173 // clearing to get some checking in the product. 174 assert(!_may_yield || _cm->cmThread()->during_cycle(), "invariant"); 175 assert(!_may_yield || !G1CollectedHeap::heap()->mark_in_progress(), "invariant"); 176 } 177 178 return false; 179 } 180 }; 181 182 void CMBitMap::clearAll() { 183 ClearBitmapHRClosure cl(NULL, this, false /* may_yield */); 184 G1CollectedHeap::heap()->heap_region_iterate(&cl); 185 guarantee(cl.complete(), "Must have completed iteration."); 186 return; 187 } 188 189 void CMBitMap::markRange(MemRegion mr) { 190 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); 191 assert(!mr.is_empty(), "unexpected empty region"); 192 assert((offsetToHeapWord(heapWordToOffset(mr.end())) == 193 ((HeapWord *) mr.end())), 194 "markRange memory region end is not card aligned"); 195 // convert address range into offset range 196 _bm.at_put_range(heapWordToOffset(mr.start()), 197 heapWordToOffset(mr.end()), true); 198 } 199 200 void CMBitMap::clearRange(MemRegion mr) { 201 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); 202 assert(!mr.is_empty(), "unexpected empty region"); 203 // convert address range into offset range 204 _bm.at_put_range(heapWordToOffset(mr.start()), 205 heapWordToOffset(mr.end()), false); 206 } 207 208 MemRegion CMBitMap::getAndClearMarkedRegion(HeapWord* addr, 209 HeapWord* end_addr) { 210 HeapWord* start = getNextMarkedWordAddress(addr); 211 start = MIN2(start, end_addr); 212 HeapWord* end = getNextUnmarkedWordAddress(start); 213 end = MIN2(end, end_addr); 214 assert(start <= end, "Consistency check"); 215 MemRegion mr(start, end); 216 if (!mr.is_empty()) { 217 clearRange(mr); 218 } 219 return mr; 220 } 221 222 CMMarkStack::CMMarkStack(ConcurrentMark* cm) : 223 _base(NULL), _cm(cm) 224 #ifdef ASSERT 225 , _drain_in_progress(false) 226 , _drain_in_progress_yields(false) 227 #endif 228 {} 229 230 bool CMMarkStack::allocate(size_t capacity) { 231 // allocate a stack of the requisite depth 232 ReservedSpace rs(ReservedSpace::allocation_align_size_up(capacity * sizeof(oop))); 233 if (!rs.is_reserved()) { 234 warning("ConcurrentMark MarkStack allocation failure"); 235 return false; 236 } 237 MemTracker::record_virtual_memory_type((address)rs.base(), mtGC); 238 if (!_virtual_space.initialize(rs, rs.size())) { 239 warning("ConcurrentMark MarkStack backing store failure"); 240 // Release the virtual memory reserved for the marking stack 241 rs.release(); 242 return false; 243 } 244 assert(_virtual_space.committed_size() == rs.size(), 245 "Didn't reserve backing store for all of ConcurrentMark stack?"); 246 _base = (oop*) _virtual_space.low(); 247 setEmpty(); 248 _capacity = (jint) capacity; 249 _saved_index = -1; 250 _should_expand = false; 251 NOT_PRODUCT(_max_depth = 0); 252 return true; 253 } 254 255 void CMMarkStack::expand() { 256 // Called, during remark, if we've overflown the marking stack during marking. 257 assert(isEmpty(), "stack should been emptied while handling overflow"); 258 assert(_capacity <= (jint) MarkStackSizeMax, "stack bigger than permitted"); 259 // Clear expansion flag 260 _should_expand = false; 261 if (_capacity == (jint) MarkStackSizeMax) { 262 if (PrintGCDetails && Verbose) { 263 gclog_or_tty->print_cr(" (benign) Can't expand marking stack capacity, at max size limit"); 264 } 265 return; 266 } 267 // Double capacity if possible 268 jint new_capacity = MIN2(_capacity*2, (jint) MarkStackSizeMax); 269 // Do not give up existing stack until we have managed to 270 // get the double capacity that we desired. 271 ReservedSpace rs(ReservedSpace::allocation_align_size_up(new_capacity * 272 sizeof(oop))); 273 if (rs.is_reserved()) { 274 // Release the backing store associated with old stack 275 _virtual_space.release(); 276 // Reinitialize virtual space for new stack 277 if (!_virtual_space.initialize(rs, rs.size())) { 278 fatal("Not enough swap for expanded marking stack capacity"); 279 } 280 _base = (oop*)(_virtual_space.low()); 281 _index = 0; 282 _capacity = new_capacity; 283 } else { 284 if (PrintGCDetails && Verbose) { 285 // Failed to double capacity, continue; 286 gclog_or_tty->print(" (benign) Failed to expand marking stack capacity from " 287 SIZE_FORMAT"K to " SIZE_FORMAT"K", 288 _capacity / K, new_capacity / K); 289 } 290 } 291 } 292 293 void CMMarkStack::set_should_expand() { 294 // If we're resetting the marking state because of an 295 // marking stack overflow, record that we should, if 296 // possible, expand the stack. 297 _should_expand = _cm->has_overflown(); 298 } 299 300 CMMarkStack::~CMMarkStack() { 301 if (_base != NULL) { 302 _base = NULL; 303 _virtual_space.release(); 304 } 305 } 306 307 void CMMarkStack::par_push(oop ptr) { 308 while (true) { 309 if (isFull()) { 310 _overflow = true; 311 return; 312 } 313 // Otherwise... 314 jint index = _index; 315 jint next_index = index+1; 316 jint res = Atomic::cmpxchg(next_index, &_index, index); 317 if (res == index) { 318 _base[index] = ptr; 319 // Note that we don't maintain this atomically. We could, but it 320 // doesn't seem necessary. 321 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 322 return; 323 } 324 // Otherwise, we need to try again. 325 } 326 } 327 328 void CMMarkStack::par_adjoin_arr(oop* ptr_arr, int n) { 329 while (true) { 330 if (isFull()) { 331 _overflow = true; 332 return; 333 } 334 // Otherwise... 335 jint index = _index; 336 jint next_index = index + n; 337 if (next_index > _capacity) { 338 _overflow = true; 339 return; 340 } 341 jint res = Atomic::cmpxchg(next_index, &_index, index); 342 if (res == index) { 343 for (int i = 0; i < n; i++) { 344 int ind = index + i; 345 assert(ind < _capacity, "By overflow test above."); 346 _base[ind] = ptr_arr[i]; 347 } 348 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 349 return; 350 } 351 // Otherwise, we need to try again. 352 } 353 } 354 355 void CMMarkStack::par_push_arr(oop* ptr_arr, int n) { 356 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 357 jint start = _index; 358 jint next_index = start + n; 359 if (next_index > _capacity) { 360 _overflow = true; 361 return; 362 } 363 // Otherwise. 364 _index = next_index; 365 for (int i = 0; i < n; i++) { 366 int ind = start + i; 367 assert(ind < _capacity, "By overflow test above."); 368 _base[ind] = ptr_arr[i]; 369 } 370 NOT_PRODUCT(_max_depth = MAX2(_max_depth, next_index)); 371 } 372 373 bool CMMarkStack::par_pop_arr(oop* ptr_arr, int max, int* n) { 374 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 375 jint index = _index; 376 if (index == 0) { 377 *n = 0; 378 return false; 379 } else { 380 int k = MIN2(max, index); 381 jint new_ind = index - k; 382 for (int j = 0; j < k; j++) { 383 ptr_arr[j] = _base[new_ind + j]; 384 } 385 _index = new_ind; 386 *n = k; 387 return true; 388 } 389 } 390 391 template<class OopClosureClass> 392 bool CMMarkStack::drain(OopClosureClass* cl, CMBitMap* bm, bool yield_after) { 393 assert(!_drain_in_progress || !_drain_in_progress_yields || yield_after 394 || SafepointSynchronize::is_at_safepoint(), 395 "Drain recursion must be yield-safe."); 396 bool res = true; 397 debug_only(_drain_in_progress = true); 398 debug_only(_drain_in_progress_yields = yield_after); 399 while (!isEmpty()) { 400 oop newOop = pop(); 401 assert(G1CollectedHeap::heap()->is_in_reserved(newOop), "Bad pop"); 402 assert(newOop->is_oop(), "Expected an oop"); 403 assert(bm == NULL || bm->isMarked((HeapWord*)newOop), 404 "only grey objects on this stack"); 405 newOop->oop_iterate(cl); 406 if (yield_after && _cm->do_yield_check()) { 407 res = false; 408 break; 409 } 410 } 411 debug_only(_drain_in_progress = false); 412 return res; 413 } 414 415 void CMMarkStack::note_start_of_gc() { 416 assert(_saved_index == -1, 417 "note_start_of_gc()/end_of_gc() bracketed incorrectly"); 418 _saved_index = _index; 419 } 420 421 void CMMarkStack::note_end_of_gc() { 422 // This is intentionally a guarantee, instead of an assert. If we 423 // accidentally add something to the mark stack during GC, it 424 // will be a correctness issue so it's better if we crash. we'll 425 // only check this once per GC anyway, so it won't be a performance 426 // issue in any way. 427 guarantee(_saved_index == _index, 428 err_msg("saved index: %d index: %d", _saved_index, _index)); 429 _saved_index = -1; 430 } 431 432 void CMMarkStack::oops_do(OopClosure* f) { 433 assert(_saved_index == _index, 434 err_msg("saved index: %d index: %d", _saved_index, _index)); 435 for (int i = 0; i < _index; i += 1) { 436 f->do_oop(&_base[i]); 437 } 438 } 439 440 CMRootRegions::CMRootRegions() : 441 _young_list(NULL), _cm(NULL), _scan_in_progress(false), 442 _should_abort(false), _next_survivor(NULL) { } 443 444 void CMRootRegions::init(G1CollectedHeap* g1h, ConcurrentMark* cm) { 445 _young_list = g1h->young_list(); 446 _cm = cm; 447 } 448 449 void CMRootRegions::prepare_for_scan() { 450 assert(!scan_in_progress(), "pre-condition"); 451 452 // Currently, only survivors can be root regions. 453 assert(_next_survivor == NULL, "pre-condition"); 454 _next_survivor = _young_list->first_survivor_region(); 455 _scan_in_progress = (_next_survivor != NULL); 456 _should_abort = false; 457 } 458 459 HeapRegion* CMRootRegions::claim_next() { 460 if (_should_abort) { 461 // If someone has set the should_abort flag, we return NULL to 462 // force the caller to bail out of their loop. 463 return NULL; 464 } 465 466 // Currently, only survivors can be root regions. 467 HeapRegion* res = _next_survivor; 468 if (res != NULL) { 469 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 470 // Read it again in case it changed while we were waiting for the lock. 471 res = _next_survivor; 472 if (res != NULL) { 473 if (res == _young_list->last_survivor_region()) { 474 // We just claimed the last survivor so store NULL to indicate 475 // that we're done. 476 _next_survivor = NULL; 477 } else { 478 _next_survivor = res->get_next_young_region(); 479 } 480 } else { 481 // Someone else claimed the last survivor while we were trying 482 // to take the lock so nothing else to do. 483 } 484 } 485 assert(res == NULL || res->is_survivor(), "post-condition"); 486 487 return res; 488 } 489 490 void CMRootRegions::scan_finished() { 491 assert(scan_in_progress(), "pre-condition"); 492 493 // Currently, only survivors can be root regions. 494 if (!_should_abort) { 495 assert(_next_survivor == NULL, "we should have claimed all survivors"); 496 } 497 _next_survivor = NULL; 498 499 { 500 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 501 _scan_in_progress = false; 502 RootRegionScan_lock->notify_all(); 503 } 504 } 505 506 bool CMRootRegions::wait_until_scan_finished() { 507 if (!scan_in_progress()) return false; 508 509 { 510 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 511 while (scan_in_progress()) { 512 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag); 513 } 514 } 515 return true; 516 } 517 518 #ifdef _MSC_VER // the use of 'this' below gets a warning, make it go away 519 #pragma warning( disable:4355 ) // 'this' : used in base member initializer list 520 #endif // _MSC_VER 521 522 uint ConcurrentMark::scale_parallel_threads(uint n_par_threads) { 523 return MAX2((n_par_threads + 2) / 4, 1U); 524 } 525 526 ConcurrentMark::ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) : 527 _g1h(g1h), 528 _markBitMap1(), 529 _markBitMap2(), 530 _parallel_marking_threads(0), 531 _max_parallel_marking_threads(0), 532 _sleep_factor(0.0), 533 _marking_task_overhead(1.0), 534 _cleanup_sleep_factor(0.0), 535 _cleanup_task_overhead(1.0), 536 _cleanup_list("Cleanup List"), 537 _region_bm((BitMap::idx_t)(g1h->max_regions()), false /* in_resource_area*/), 538 _card_bm((g1h->reserved_region().byte_size() + CardTableModRefBS::card_size - 1) >> 539 CardTableModRefBS::card_shift, 540 false /* in_resource_area*/), 541 542 _prevMarkBitMap(&_markBitMap1), 543 _nextMarkBitMap(&_markBitMap2), 544 545 _markStack(this), 546 // _finger set in set_non_marking_state 547 548 _max_worker_id(MAX2((uint)ParallelGCThreads, 1U)), 549 // _active_tasks set in set_non_marking_state 550 // _tasks set inside the constructor 551 _task_queues(new CMTaskQueueSet((int) _max_worker_id)), 552 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)), 553 554 _has_overflown(false), 555 _concurrent(false), 556 _has_aborted(false), 557 _aborted_gc_id(GCId::undefined()), 558 _restart_for_overflow(false), 559 _concurrent_marking_in_progress(false), 560 561 // _verbose_level set below 562 563 _init_times(), 564 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(), 565 _cleanup_times(), 566 _total_counting_time(0.0), 567 _total_rs_scrub_time(0.0), 568 569 _parallel_workers(NULL), 570 571 _count_card_bitmaps(NULL), 572 _count_marked_bytes(NULL), 573 _completed_initialization(false) { 574 CMVerboseLevel verbose_level = (CMVerboseLevel) G1MarkingVerboseLevel; 575 if (verbose_level < no_verbose) { 576 verbose_level = no_verbose; 577 } 578 if (verbose_level > high_verbose) { 579 verbose_level = high_verbose; 580 } 581 _verbose_level = verbose_level; 582 583 if (verbose_low()) { 584 gclog_or_tty->print_cr("[global] init, heap start = "PTR_FORMAT", " 585 "heap end = " PTR_FORMAT, p2i(_heap_start), p2i(_heap_end)); 586 } 587 588 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage); 589 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage); 590 591 // Create & start a ConcurrentMark thread. 592 _cmThread = new ConcurrentMarkThread(this); 593 assert(cmThread() != NULL, "CM Thread should have been created"); 594 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm"); 595 if (_cmThread->osthread() == NULL) { 596 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread"); 597 } 598 599 assert(CGC_lock != NULL, "Where's the CGC_lock?"); 600 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency"); 601 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency"); 602 603 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); 604 satb_qs.set_buffer_size(G1SATBBufferSize); 605 606 _root_regions.init(_g1h, this); 607 608 if (ConcGCThreads > ParallelGCThreads) { 609 warning("Can't have more ConcGCThreads (" UINTX_FORMAT ") " 610 "than ParallelGCThreads (" UINTX_FORMAT ").", 611 ConcGCThreads, ParallelGCThreads); 612 return; 613 } 614 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) { 615 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent 616 // if both are set 617 _sleep_factor = 0.0; 618 _marking_task_overhead = 1.0; 619 } else if (G1MarkingOverheadPercent > 0) { 620 // We will calculate the number of parallel marking threads based 621 // on a target overhead with respect to the soft real-time goal 622 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0; 623 double overall_cm_overhead = 624 (double) MaxGCPauseMillis * marking_overhead / 625 (double) GCPauseIntervalMillis; 626 double cpu_ratio = 1.0 / (double) os::processor_count(); 627 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio); 628 double marking_task_overhead = 629 overall_cm_overhead / marking_thread_num * 630 (double) os::processor_count(); 631 double sleep_factor = 632 (1.0 - marking_task_overhead) / marking_task_overhead; 633 634 FLAG_SET_ERGO(uintx, ConcGCThreads, (uint) marking_thread_num); 635 _sleep_factor = sleep_factor; 636 _marking_task_overhead = marking_task_overhead; 637 } else { 638 // Calculate the number of parallel marking threads by scaling 639 // the number of parallel GC threads. 640 uint marking_thread_num = scale_parallel_threads((uint) ParallelGCThreads); 641 FLAG_SET_ERGO(uintx, ConcGCThreads, marking_thread_num); 642 _sleep_factor = 0.0; 643 _marking_task_overhead = 1.0; 644 } 645 646 assert(ConcGCThreads > 0, "Should have been set"); 647 _parallel_marking_threads = (uint) ConcGCThreads; 648 _max_parallel_marking_threads = _parallel_marking_threads; 649 650 if (parallel_marking_threads() > 1) { 651 _cleanup_task_overhead = 1.0; 652 } else { 653 _cleanup_task_overhead = marking_task_overhead(); 654 } 655 _cleanup_sleep_factor = 656 (1.0 - cleanup_task_overhead()) / cleanup_task_overhead(); 657 658 #if 0 659 gclog_or_tty->print_cr("Marking Threads %d", parallel_marking_threads()); 660 gclog_or_tty->print_cr("CM Marking Task Overhead %1.4lf", marking_task_overhead()); 661 gclog_or_tty->print_cr("CM Sleep Factor %1.4lf", sleep_factor()); 662 gclog_or_tty->print_cr("CL Marking Task Overhead %1.4lf", cleanup_task_overhead()); 663 gclog_or_tty->print_cr("CL Sleep Factor %1.4lf", cleanup_sleep_factor()); 664 #endif 665 666 _parallel_workers = new FlexibleWorkGang("G1 Parallel Marking Threads", 667 _max_parallel_marking_threads, false, true); 668 if (_parallel_workers == NULL) { 669 vm_exit_during_initialization("Failed necessary allocation."); 670 } else { 671 _parallel_workers->initialize_workers(); 672 } 673 674 if (FLAG_IS_DEFAULT(MarkStackSize)) { 675 uintx mark_stack_size = 676 MIN2(MarkStackSizeMax, 677 MAX2(MarkStackSize, (uintx) (parallel_marking_threads() * TASKQUEUE_SIZE))); 678 // Verify that the calculated value for MarkStackSize is in range. 679 // It would be nice to use the private utility routine from Arguments. 680 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) { 681 warning("Invalid value calculated for MarkStackSize (" UINTX_FORMAT "): " 682 "must be between " UINTX_FORMAT " and " UINTX_FORMAT, 683 mark_stack_size, (uintx) 1, MarkStackSizeMax); 684 return; 685 } 686 FLAG_SET_ERGO(uintx, MarkStackSize, mark_stack_size); 687 } else { 688 // Verify MarkStackSize is in range. 689 if (FLAG_IS_CMDLINE(MarkStackSize)) { 690 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) { 691 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 692 warning("Invalid value specified for MarkStackSize (" UINTX_FORMAT "): " 693 "must be between " UINTX_FORMAT " and " UINTX_FORMAT, 694 MarkStackSize, (uintx) 1, MarkStackSizeMax); 695 return; 696 } 697 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) { 698 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 699 warning("Invalid value specified for MarkStackSize (" UINTX_FORMAT ")" 700 " or for MarkStackSizeMax (" UINTX_FORMAT ")", 701 MarkStackSize, MarkStackSizeMax); 702 return; 703 } 704 } 705 } 706 } 707 708 if (!_markStack.allocate(MarkStackSize)) { 709 warning("Failed to allocate CM marking stack"); 710 return; 711 } 712 713 _tasks = NEW_C_HEAP_ARRAY(CMTask*, _max_worker_id, mtGC); 714 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC); 715 716 _count_card_bitmaps = NEW_C_HEAP_ARRAY(BitMap, _max_worker_id, mtGC); 717 _count_marked_bytes = NEW_C_HEAP_ARRAY(size_t*, _max_worker_id, mtGC); 718 719 BitMap::idx_t card_bm_size = _card_bm.size(); 720 721 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail 722 _active_tasks = _max_worker_id; 723 724 size_t max_regions = (size_t) _g1h->max_regions(); 725 for (uint i = 0; i < _max_worker_id; ++i) { 726 CMTaskQueue* task_queue = new CMTaskQueue(); 727 task_queue->initialize(); 728 _task_queues->register_queue(i, task_queue); 729 730 _count_card_bitmaps[i] = BitMap(card_bm_size, false); 731 _count_marked_bytes[i] = NEW_C_HEAP_ARRAY(size_t, max_regions, mtGC); 732 733 _tasks[i] = new CMTask(i, this, 734 _count_marked_bytes[i], 735 &_count_card_bitmaps[i], 736 task_queue, _task_queues); 737 738 _accum_task_vtime[i] = 0.0; 739 } 740 741 // Calculate the card number for the bottom of the heap. Used 742 // in biasing indexes into the accounting card bitmaps. 743 _heap_bottom_card_num = 744 intptr_t(uintptr_t(_g1h->reserved_region().start()) >> 745 CardTableModRefBS::card_shift); 746 747 // Clear all the liveness counting data 748 clear_all_count_data(); 749 750 // so that the call below can read a sensible value 751 _heap_start = g1h->reserved_region().start(); 752 set_non_marking_state(); 753 _completed_initialization = true; 754 } 755 756 void ConcurrentMark::reset() { 757 // Starting values for these two. This should be called in a STW 758 // phase. 759 MemRegion reserved = _g1h->g1_reserved(); 760 _heap_start = reserved.start(); 761 _heap_end = reserved.end(); 762 763 // Separated the asserts so that we know which one fires. 764 assert(_heap_start != NULL, "heap bounds should look ok"); 765 assert(_heap_end != NULL, "heap bounds should look ok"); 766 assert(_heap_start < _heap_end, "heap bounds should look ok"); 767 768 // Reset all the marking data structures and any necessary flags 769 reset_marking_state(); 770 771 if (verbose_low()) { 772 gclog_or_tty->print_cr("[global] resetting"); 773 } 774 775 // We do reset all of them, since different phases will use 776 // different number of active threads. So, it's easiest to have all 777 // of them ready. 778 for (uint i = 0; i < _max_worker_id; ++i) { 779 _tasks[i]->reset(_nextMarkBitMap); 780 } 781 782 // we need this to make sure that the flag is on during the evac 783 // pause with initial mark piggy-backed 784 set_concurrent_marking_in_progress(); 785 } 786 787 788 void ConcurrentMark::reset_marking_state(bool clear_overflow) { 789 _markStack.set_should_expand(); 790 _markStack.setEmpty(); // Also clears the _markStack overflow flag 791 if (clear_overflow) { 792 clear_has_overflown(); 793 } else { 794 assert(has_overflown(), "pre-condition"); 795 } 796 _finger = _heap_start; 797 798 for (uint i = 0; i < _max_worker_id; ++i) { 799 CMTaskQueue* queue = _task_queues->queue(i); 800 queue->set_empty(); 801 } 802 } 803 804 void ConcurrentMark::set_concurrency(uint active_tasks) { 805 assert(active_tasks <= _max_worker_id, "we should not have more"); 806 807 _active_tasks = active_tasks; 808 // Need to update the three data structures below according to the 809 // number of active threads for this phase. 810 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues); 811 _first_overflow_barrier_sync.set_n_workers((int) active_tasks); 812 _second_overflow_barrier_sync.set_n_workers((int) active_tasks); 813 } 814 815 void ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) { 816 set_concurrency(active_tasks); 817 818 _concurrent = concurrent; 819 // We propagate this to all tasks, not just the active ones. 820 for (uint i = 0; i < _max_worker_id; ++i) 821 _tasks[i]->set_concurrent(concurrent); 822 823 if (concurrent) { 824 set_concurrent_marking_in_progress(); 825 } else { 826 // We currently assume that the concurrent flag has been set to 827 // false before we start remark. At this point we should also be 828 // in a STW phase. 829 assert(!concurrent_marking_in_progress(), "invariant"); 830 assert(out_of_regions(), 831 err_msg("only way to get here: _finger: "PTR_FORMAT", _heap_end: "PTR_FORMAT, 832 p2i(_finger), p2i(_heap_end))); 833 } 834 } 835 836 void ConcurrentMark::set_non_marking_state() { 837 // We set the global marking state to some default values when we're 838 // not doing marking. 839 reset_marking_state(); 840 _active_tasks = 0; 841 clear_concurrent_marking_in_progress(); 842 } 843 844 ConcurrentMark::~ConcurrentMark() { 845 // The ConcurrentMark instance is never freed. 846 ShouldNotReachHere(); 847 } 848 849 void ConcurrentMark::clearNextBitmap() { 850 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 851 852 // Make sure that the concurrent mark thread looks to still be in 853 // the current cycle. 854 guarantee(cmThread()->during_cycle(), "invariant"); 855 856 // We are finishing up the current cycle by clearing the next 857 // marking bitmap and getting it ready for the next cycle. During 858 // this time no other cycle can start. So, let's make sure that this 859 // is the case. 860 guarantee(!g1h->mark_in_progress(), "invariant"); 861 862 ClearBitmapHRClosure cl(this, _nextMarkBitMap, true /* may_yield */); 863 g1h->heap_region_iterate(&cl); 864 865 // Clear the liveness counting data. If the marking has been aborted, the abort() 866 // call already did that. 867 if (cl.complete()) { 868 clear_all_count_data(); 869 } 870 871 // Repeat the asserts from above. 872 guarantee(cmThread()->during_cycle(), "invariant"); 873 guarantee(!g1h->mark_in_progress(), "invariant"); 874 } 875 876 class CheckBitmapClearHRClosure : public HeapRegionClosure { 877 CMBitMap* _bitmap; 878 bool _error; 879 public: 880 CheckBitmapClearHRClosure(CMBitMap* bitmap) : _bitmap(bitmap) { 881 } 882 883 virtual bool doHeapRegion(HeapRegion* r) { 884 // This closure can be called concurrently to the mutator, so we must make sure 885 // that the result of the getNextMarkedWordAddress() call is compared to the 886 // value passed to it as limit to detect any found bits. 887 // We can use the region's orig_end() for the limit and the comparison value 888 // as it always contains the "real" end of the region that never changes and 889 // has no side effects. 890 // Due to the latter, there can also be no problem with the compiler generating 891 // reloads of the orig_end() call. 892 HeapWord* end = r->orig_end(); 893 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end; 894 } 895 }; 896 897 bool ConcurrentMark::nextMarkBitmapIsClear() { 898 CheckBitmapClearHRClosure cl(_nextMarkBitMap); 899 _g1h->heap_region_iterate(&cl); 900 return cl.complete(); 901 } 902 903 class NoteStartOfMarkHRClosure: public HeapRegionClosure { 904 public: 905 bool doHeapRegion(HeapRegion* r) { 906 if (!r->is_continues_humongous()) { 907 r->note_start_of_marking(); 908 } 909 return false; 910 } 911 }; 912 913 void ConcurrentMark::checkpointRootsInitialPre() { 914 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 915 G1CollectorPolicy* g1p = g1h->g1_policy(); 916 917 _has_aborted = false; 918 919 #ifndef PRODUCT 920 if (G1PrintReachableAtInitialMark) { 921 print_reachable("at-cycle-start", 922 VerifyOption_G1UsePrevMarking, true /* all */); 923 } 924 #endif 925 926 // Initialize marking structures. This has to be done in a STW phase. 927 reset(); 928 929 // For each region note start of marking. 930 NoteStartOfMarkHRClosure startcl; 931 g1h->heap_region_iterate(&startcl); 932 } 933 934 935 void ConcurrentMark::checkpointRootsInitialPost() { 936 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 937 938 // If we force an overflow during remark, the remark operation will 939 // actually abort and we'll restart concurrent marking. If we always 940 // force an overflow during remark we'll never actually complete the 941 // marking phase. So, we initialize this here, at the start of the 942 // cycle, so that at the remaining overflow number will decrease at 943 // every remark and we'll eventually not need to cause one. 944 force_overflow_stw()->init(); 945 946 // Start Concurrent Marking weak-reference discovery. 947 ReferenceProcessor* rp = g1h->ref_processor_cm(); 948 // enable ("weak") refs discovery 949 rp->enable_discovery(true /*verify_disabled*/, true /*verify_no_refs*/); 950 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle 951 952 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 953 // This is the start of the marking cycle, we're expected all 954 // threads to have SATB queues with active set to false. 955 satb_mq_set.set_active_all_threads(true, /* new active value */ 956 false /* expected_active */); 957 958 _root_regions.prepare_for_scan(); 959 960 // update_g1_committed() will be called at the end of an evac pause 961 // when marking is on. So, it's also called at the end of the 962 // initial-mark pause to update the heap end, if the heap expands 963 // during it. No need to call it here. 964 } 965 966 /* 967 * Notice that in the next two methods, we actually leave the STS 968 * during the barrier sync and join it immediately afterwards. If we 969 * do not do this, the following deadlock can occur: one thread could 970 * be in the barrier sync code, waiting for the other thread to also 971 * sync up, whereas another one could be trying to yield, while also 972 * waiting for the other threads to sync up too. 973 * 974 * Note, however, that this code is also used during remark and in 975 * this case we should not attempt to leave / enter the STS, otherwise 976 * we'll either hit an assert (debug / fastdebug) or deadlock 977 * (product). So we should only leave / enter the STS if we are 978 * operating concurrently. 979 * 980 * Because the thread that does the sync barrier has left the STS, it 981 * is possible to be suspended for a Full GC or an evacuation pause 982 * could occur. This is actually safe, since the entering the sync 983 * barrier is one of the last things do_marking_step() does, and it 984 * doesn't manipulate any data structures afterwards. 985 */ 986 987 void ConcurrentMark::enter_first_sync_barrier(uint worker_id) { 988 if (verbose_low()) { 989 gclog_or_tty->print_cr("[%u] entering first barrier", worker_id); 990 } 991 992 if (concurrent()) { 993 SuspendibleThreadSet::leave(); 994 } 995 996 bool barrier_aborted = !_first_overflow_barrier_sync.enter(); 997 998 if (concurrent()) { 999 SuspendibleThreadSet::join(); 1000 } 1001 // at this point everyone should have synced up and not be doing any 1002 // more work 1003 1004 if (verbose_low()) { 1005 if (barrier_aborted) { 1006 gclog_or_tty->print_cr("[%u] aborted first barrier", worker_id); 1007 } else { 1008 gclog_or_tty->print_cr("[%u] leaving first barrier", worker_id); 1009 } 1010 } 1011 1012 if (barrier_aborted) { 1013 // If the barrier aborted we ignore the overflow condition and 1014 // just abort the whole marking phase as quickly as possible. 1015 return; 1016 } 1017 1018 // If we're executing the concurrent phase of marking, reset the marking 1019 // state; otherwise the marking state is reset after reference processing, 1020 // during the remark pause. 1021 // If we reset here as a result of an overflow during the remark we will 1022 // see assertion failures from any subsequent set_concurrency_and_phase() 1023 // calls. 1024 if (concurrent()) { 1025 // let the task associated with with worker 0 do this 1026 if (worker_id == 0) { 1027 // task 0 is responsible for clearing the global data structures 1028 // We should be here because of an overflow. During STW we should 1029 // not clear the overflow flag since we rely on it being true when 1030 // we exit this method to abort the pause and restart concurrent 1031 // marking. 1032 reset_marking_state(true /* clear_overflow */); 1033 force_overflow()->update(); 1034 1035 if (G1Log::fine()) { 1036 gclog_or_tty->gclog_stamp(concurrent_gc_id()); 1037 gclog_or_tty->print_cr("[GC concurrent-mark-reset-for-overflow]"); 1038 } 1039 } 1040 } 1041 1042 // after this, each task should reset its own data structures then 1043 // then go into the second barrier 1044 } 1045 1046 void ConcurrentMark::enter_second_sync_barrier(uint worker_id) { 1047 if (verbose_low()) { 1048 gclog_or_tty->print_cr("[%u] entering second barrier", worker_id); 1049 } 1050 1051 if (concurrent()) { 1052 SuspendibleThreadSet::leave(); 1053 } 1054 1055 bool barrier_aborted = !_second_overflow_barrier_sync.enter(); 1056 1057 if (concurrent()) { 1058 SuspendibleThreadSet::join(); 1059 } 1060 // at this point everything should be re-initialized and ready to go 1061 1062 if (verbose_low()) { 1063 if (barrier_aborted) { 1064 gclog_or_tty->print_cr("[%u] aborted second barrier", worker_id); 1065 } else { 1066 gclog_or_tty->print_cr("[%u] leaving second barrier", worker_id); 1067 } 1068 } 1069 } 1070 1071 #ifndef PRODUCT 1072 void ForceOverflowSettings::init() { 1073 _num_remaining = G1ConcMarkForceOverflow; 1074 _force = false; 1075 update(); 1076 } 1077 1078 void ForceOverflowSettings::update() { 1079 if (_num_remaining > 0) { 1080 _num_remaining -= 1; 1081 _force = true; 1082 } else { 1083 _force = false; 1084 } 1085 } 1086 1087 bool ForceOverflowSettings::should_force() { 1088 if (_force) { 1089 _force = false; 1090 return true; 1091 } else { 1092 return false; 1093 } 1094 } 1095 #endif // !PRODUCT 1096 1097 class CMConcurrentMarkingTask: public AbstractGangTask { 1098 private: 1099 ConcurrentMark* _cm; 1100 ConcurrentMarkThread* _cmt; 1101 1102 public: 1103 void work(uint worker_id) { 1104 assert(Thread::current()->is_ConcurrentGC_thread(), 1105 "this should only be done by a conc GC thread"); 1106 ResourceMark rm; 1107 1108 double start_vtime = os::elapsedVTime(); 1109 1110 SuspendibleThreadSet::join(); 1111 1112 assert(worker_id < _cm->active_tasks(), "invariant"); 1113 CMTask* the_task = _cm->task(worker_id); 1114 the_task->record_start_time(); 1115 if (!_cm->has_aborted()) { 1116 do { 1117 double start_vtime_sec = os::elapsedVTime(); 1118 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 1119 1120 the_task->do_marking_step(mark_step_duration_ms, 1121 true /* do_termination */, 1122 false /* is_serial*/); 1123 1124 double end_vtime_sec = os::elapsedVTime(); 1125 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; 1126 _cm->clear_has_overflown(); 1127 1128 _cm->do_yield_check(worker_id); 1129 1130 jlong sleep_time_ms; 1131 if (!_cm->has_aborted() && the_task->has_aborted()) { 1132 sleep_time_ms = 1133 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0); 1134 SuspendibleThreadSet::leave(); 1135 os::sleep(Thread::current(), sleep_time_ms, false); 1136 SuspendibleThreadSet::join(); 1137 } 1138 } while (!_cm->has_aborted() && the_task->has_aborted()); 1139 } 1140 the_task->record_end_time(); 1141 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant"); 1142 1143 SuspendibleThreadSet::leave(); 1144 1145 double end_vtime = os::elapsedVTime(); 1146 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime); 1147 } 1148 1149 CMConcurrentMarkingTask(ConcurrentMark* cm, 1150 ConcurrentMarkThread* cmt) : 1151 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { } 1152 1153 ~CMConcurrentMarkingTask() { } 1154 }; 1155 1156 // Calculates the number of active workers for a concurrent 1157 // phase. 1158 uint ConcurrentMark::calc_parallel_marking_threads() { 1159 uint n_conc_workers = 0; 1160 if (!UseDynamicNumberOfGCThreads || 1161 (!FLAG_IS_DEFAULT(ConcGCThreads) && 1162 !ForceDynamicNumberOfGCThreads)) { 1163 n_conc_workers = max_parallel_marking_threads(); 1164 } else { 1165 n_conc_workers = 1166 AdaptiveSizePolicy::calc_default_active_workers( 1167 max_parallel_marking_threads(), 1168 1, /* Minimum workers */ 1169 parallel_marking_threads(), 1170 Threads::number_of_non_daemon_threads()); 1171 // Don't scale down "n_conc_workers" by scale_parallel_threads() because 1172 // that scaling has already gone into "_max_parallel_marking_threads". 1173 } 1174 assert(n_conc_workers > 0, "Always need at least 1"); 1175 return n_conc_workers; 1176 } 1177 1178 void ConcurrentMark::scanRootRegion(HeapRegion* hr, uint worker_id) { 1179 // Currently, only survivors can be root regions. 1180 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant"); 1181 G1RootRegionScanClosure cl(_g1h, this, worker_id); 1182 1183 const uintx interval = PrefetchScanIntervalInBytes; 1184 HeapWord* curr = hr->bottom(); 1185 const HeapWord* end = hr->top(); 1186 while (curr < end) { 1187 Prefetch::read(curr, interval); 1188 oop obj = oop(curr); 1189 int size = obj->oop_iterate(&cl); 1190 assert(size == obj->size(), "sanity"); 1191 curr += size; 1192 } 1193 } 1194 1195 class CMRootRegionScanTask : public AbstractGangTask { 1196 private: 1197 ConcurrentMark* _cm; 1198 1199 public: 1200 CMRootRegionScanTask(ConcurrentMark* cm) : 1201 AbstractGangTask("Root Region Scan"), _cm(cm) { } 1202 1203 void work(uint worker_id) { 1204 assert(Thread::current()->is_ConcurrentGC_thread(), 1205 "this should only be done by a conc GC thread"); 1206 1207 CMRootRegions* root_regions = _cm->root_regions(); 1208 HeapRegion* hr = root_regions->claim_next(); 1209 while (hr != NULL) { 1210 _cm->scanRootRegion(hr, worker_id); 1211 hr = root_regions->claim_next(); 1212 } 1213 } 1214 }; 1215 1216 void ConcurrentMark::scanRootRegions() { 1217 // Start of concurrent marking. 1218 ClassLoaderDataGraph::clear_claimed_marks(); 1219 1220 // scan_in_progress() will have been set to true only if there was 1221 // at least one root region to scan. So, if it's false, we 1222 // should not attempt to do any further work. 1223 if (root_regions()->scan_in_progress()) { 1224 _parallel_marking_threads = calc_parallel_marking_threads(); 1225 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1226 "Maximum number of marking threads exceeded"); 1227 uint active_workers = MAX2(1U, parallel_marking_threads()); 1228 1229 CMRootRegionScanTask task(this); 1230 _parallel_workers->set_active_workers(active_workers); 1231 _parallel_workers->run_task(&task); 1232 1233 // It's possible that has_aborted() is true here without actually 1234 // aborting the survivor scan earlier. This is OK as it's 1235 // mainly used for sanity checking. 1236 root_regions()->scan_finished(); 1237 } 1238 } 1239 1240 void ConcurrentMark::markFromRoots() { 1241 // we might be tempted to assert that: 1242 // assert(asynch == !SafepointSynchronize::is_at_safepoint(), 1243 // "inconsistent argument?"); 1244 // However that wouldn't be right, because it's possible that 1245 // a safepoint is indeed in progress as a younger generation 1246 // stop-the-world GC happens even as we mark in this generation. 1247 1248 _restart_for_overflow = false; 1249 force_overflow_conc()->init(); 1250 1251 // _g1h has _n_par_threads 1252 _parallel_marking_threads = calc_parallel_marking_threads(); 1253 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1254 "Maximum number of marking threads exceeded"); 1255 1256 uint active_workers = MAX2(1U, parallel_marking_threads()); 1257 1258 // Parallel task terminator is set in "set_concurrency_and_phase()" 1259 set_concurrency_and_phase(active_workers, true /* concurrent */); 1260 1261 CMConcurrentMarkingTask markingTask(this, cmThread()); 1262 _parallel_workers->set_active_workers(active_workers); 1263 // Don't set _n_par_threads because it affects MT in process_roots() 1264 // and the decisions on that MT processing is made elsewhere. 1265 assert(_parallel_workers->active_workers() > 0, "Should have been set"); 1266 _parallel_workers->run_task(&markingTask); 1267 print_stats(); 1268 } 1269 1270 // Helper class to get rid of some boilerplate code. 1271 class G1CMTraceTime : public GCTraceTime { 1272 static bool doit_and_prepend(bool doit) { 1273 if (doit) { 1274 gclog_or_tty->put(' '); 1275 } 1276 return doit; 1277 } 1278 1279 public: 1280 G1CMTraceTime(const char* title, bool doit) 1281 : GCTraceTime(title, doit_and_prepend(doit), false, G1CollectedHeap::heap()->gc_timer_cm(), 1282 G1CollectedHeap::heap()->concurrent_mark()->concurrent_gc_id()) { 1283 } 1284 }; 1285 1286 void ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) { 1287 // world is stopped at this checkpoint 1288 assert(SafepointSynchronize::is_at_safepoint(), 1289 "world should be stopped"); 1290 1291 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1292 1293 // If a full collection has happened, we shouldn't do this. 1294 if (has_aborted()) { 1295 g1h->set_marking_complete(); // So bitmap clearing isn't confused 1296 return; 1297 } 1298 1299 SvcGCMarker sgcm(SvcGCMarker::OTHER); 1300 1301 if (VerifyDuringGC) { 1302 HandleMark hm; // handle scope 1303 Universe::heap()->prepare_for_verify(); 1304 Universe::verify(VerifyOption_G1UsePrevMarking, 1305 " VerifyDuringGC:(before)"); 1306 } 1307 g1h->check_bitmaps("Remark Start"); 1308 1309 G1CollectorPolicy* g1p = g1h->g1_policy(); 1310 g1p->record_concurrent_mark_remark_start(); 1311 1312 double start = os::elapsedTime(); 1313 1314 checkpointRootsFinalWork(); 1315 1316 double mark_work_end = os::elapsedTime(); 1317 1318 weakRefsWork(clear_all_soft_refs); 1319 1320 if (has_overflown()) { 1321 // Oops. We overflowed. Restart concurrent marking. 1322 _restart_for_overflow = true; 1323 if (G1TraceMarkStackOverflow) { 1324 gclog_or_tty->print_cr("\nRemark led to restart for overflow."); 1325 } 1326 1327 // Verify the heap w.r.t. the previous marking bitmap. 1328 if (VerifyDuringGC) { 1329 HandleMark hm; // handle scope 1330 Universe::heap()->prepare_for_verify(); 1331 Universe::verify(VerifyOption_G1UsePrevMarking, 1332 " VerifyDuringGC:(overflow)"); 1333 } 1334 1335 // Clear the marking state because we will be restarting 1336 // marking due to overflowing the global mark stack. 1337 reset_marking_state(); 1338 } else { 1339 { 1340 G1CMTraceTime trace("GC aggregate-data", G1Log::finer()); 1341 1342 // Aggregate the per-task counting data that we have accumulated 1343 // while marking. 1344 aggregate_count_data(); 1345 } 1346 1347 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 1348 // We're done with marking. 1349 // This is the end of the marking cycle, we're expected all 1350 // threads to have SATB queues with active set to true. 1351 satb_mq_set.set_active_all_threads(false, /* new active value */ 1352 true /* expected_active */); 1353 1354 if (VerifyDuringGC) { 1355 HandleMark hm; // handle scope 1356 Universe::heap()->prepare_for_verify(); 1357 Universe::verify(VerifyOption_G1UseNextMarking, 1358 " VerifyDuringGC:(after)"); 1359 } 1360 g1h->check_bitmaps("Remark End"); 1361 assert(!restart_for_overflow(), "sanity"); 1362 // Completely reset the marking state since marking completed 1363 set_non_marking_state(); 1364 } 1365 1366 // Expand the marking stack, if we have to and if we can. 1367 if (_markStack.should_expand()) { 1368 _markStack.expand(); 1369 } 1370 1371 // Statistics 1372 double now = os::elapsedTime(); 1373 _remark_mark_times.add((mark_work_end - start) * 1000.0); 1374 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0); 1375 _remark_times.add((now - start) * 1000.0); 1376 1377 g1p->record_concurrent_mark_remark_end(); 1378 1379 G1CMIsAliveClosure is_alive(g1h); 1380 g1h->gc_tracer_cm()->report_object_count_after_gc(&is_alive); 1381 } 1382 1383 // Base class of the closures that finalize and verify the 1384 // liveness counting data. 1385 class CMCountDataClosureBase: public HeapRegionClosure { 1386 protected: 1387 G1CollectedHeap* _g1h; 1388 ConcurrentMark* _cm; 1389 CardTableModRefBS* _ct_bs; 1390 1391 BitMap* _region_bm; 1392 BitMap* _card_bm; 1393 1394 // Takes a region that's not empty (i.e., it has at least one 1395 // live object in it and sets its corresponding bit on the region 1396 // bitmap to 1. If the region is "starts humongous" it will also set 1397 // to 1 the bits on the region bitmap that correspond to its 1398 // associated "continues humongous" regions. 1399 void set_bit_for_region(HeapRegion* hr) { 1400 assert(!hr->is_continues_humongous(), "should have filtered those out"); 1401 1402 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index(); 1403 if (!hr->is_starts_humongous()) { 1404 // Normal (non-humongous) case: just set the bit. 1405 _region_bm->par_at_put(index, true); 1406 } else { 1407 // Starts humongous case: calculate how many regions are part of 1408 // this humongous region and then set the bit range. 1409 BitMap::idx_t end_index = (BitMap::idx_t) hr->last_hc_index(); 1410 _region_bm->par_at_put_range(index, end_index, true); 1411 } 1412 } 1413 1414 public: 1415 CMCountDataClosureBase(G1CollectedHeap* g1h, 1416 BitMap* region_bm, BitMap* card_bm): 1417 _g1h(g1h), _cm(g1h->concurrent_mark()), 1418 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())), 1419 _region_bm(region_bm), _card_bm(card_bm) { } 1420 }; 1421 1422 // Closure that calculates the # live objects per region. Used 1423 // for verification purposes during the cleanup pause. 1424 class CalcLiveObjectsClosure: public CMCountDataClosureBase { 1425 CMBitMapRO* _bm; 1426 size_t _region_marked_bytes; 1427 1428 public: 1429 CalcLiveObjectsClosure(CMBitMapRO *bm, G1CollectedHeap* g1h, 1430 BitMap* region_bm, BitMap* card_bm) : 1431 CMCountDataClosureBase(g1h, region_bm, card_bm), 1432 _bm(bm), _region_marked_bytes(0) { } 1433 1434 bool doHeapRegion(HeapRegion* hr) { 1435 1436 if (hr->is_continues_humongous()) { 1437 // We will ignore these here and process them when their 1438 // associated "starts humongous" region is processed (see 1439 // set_bit_for_heap_region()). Note that we cannot rely on their 1440 // associated "starts humongous" region to have their bit set to 1441 // 1 since, due to the region chunking in the parallel region 1442 // iteration, a "continues humongous" region might be visited 1443 // before its associated "starts humongous". 1444 return false; 1445 } 1446 1447 HeapWord* ntams = hr->next_top_at_mark_start(); 1448 HeapWord* start = hr->bottom(); 1449 1450 assert(start <= hr->end() && start <= ntams && ntams <= hr->end(), 1451 err_msg("Preconditions not met - " 1452 "start: "PTR_FORMAT", ntams: "PTR_FORMAT", end: "PTR_FORMAT, 1453 p2i(start), p2i(ntams), p2i(hr->end()))); 1454 1455 // Find the first marked object at or after "start". 1456 start = _bm->getNextMarkedWordAddress(start, ntams); 1457 1458 size_t marked_bytes = 0; 1459 1460 while (start < ntams) { 1461 oop obj = oop(start); 1462 int obj_sz = obj->size(); 1463 HeapWord* obj_end = start + obj_sz; 1464 1465 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start); 1466 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(obj_end); 1467 1468 // Note: if we're looking at the last region in heap - obj_end 1469 // could be actually just beyond the end of the heap; end_idx 1470 // will then correspond to a (non-existent) card that is also 1471 // just beyond the heap. 1472 if (_g1h->is_in_g1_reserved(obj_end) && !_ct_bs->is_card_aligned(obj_end)) { 1473 // end of object is not card aligned - increment to cover 1474 // all the cards spanned by the object 1475 end_idx += 1; 1476 } 1477 1478 // Set the bits in the card BM for the cards spanned by this object. 1479 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1480 1481 // Add the size of this object to the number of marked bytes. 1482 marked_bytes += (size_t)obj_sz * HeapWordSize; 1483 1484 // Find the next marked object after this one. 1485 start = _bm->getNextMarkedWordAddress(obj_end, ntams); 1486 } 1487 1488 // Mark the allocated-since-marking portion... 1489 HeapWord* top = hr->top(); 1490 if (ntams < top) { 1491 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams); 1492 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top); 1493 1494 // Note: if we're looking at the last region in heap - top 1495 // could be actually just beyond the end of the heap; end_idx 1496 // will then correspond to a (non-existent) card that is also 1497 // just beyond the heap. 1498 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) { 1499 // end of object is not card aligned - increment to cover 1500 // all the cards spanned by the object 1501 end_idx += 1; 1502 } 1503 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1504 1505 // This definitely means the region has live objects. 1506 set_bit_for_region(hr); 1507 } 1508 1509 // Update the live region bitmap. 1510 if (marked_bytes > 0) { 1511 set_bit_for_region(hr); 1512 } 1513 1514 // Set the marked bytes for the current region so that 1515 // it can be queried by a calling verification routine 1516 _region_marked_bytes = marked_bytes; 1517 1518 return false; 1519 } 1520 1521 size_t region_marked_bytes() const { return _region_marked_bytes; } 1522 }; 1523 1524 // Heap region closure used for verifying the counting data 1525 // that was accumulated concurrently and aggregated during 1526 // the remark pause. This closure is applied to the heap 1527 // regions during the STW cleanup pause. 1528 1529 class VerifyLiveObjectDataHRClosure: public HeapRegionClosure { 1530 G1CollectedHeap* _g1h; 1531 ConcurrentMark* _cm; 1532 CalcLiveObjectsClosure _calc_cl; 1533 BitMap* _region_bm; // Region BM to be verified 1534 BitMap* _card_bm; // Card BM to be verified 1535 bool _verbose; // verbose output? 1536 1537 BitMap* _exp_region_bm; // Expected Region BM values 1538 BitMap* _exp_card_bm; // Expected card BM values 1539 1540 int _failures; 1541 1542 public: 1543 VerifyLiveObjectDataHRClosure(G1CollectedHeap* g1h, 1544 BitMap* region_bm, 1545 BitMap* card_bm, 1546 BitMap* exp_region_bm, 1547 BitMap* exp_card_bm, 1548 bool verbose) : 1549 _g1h(g1h), _cm(g1h->concurrent_mark()), 1550 _calc_cl(_cm->nextMarkBitMap(), g1h, exp_region_bm, exp_card_bm), 1551 _region_bm(region_bm), _card_bm(card_bm), _verbose(verbose), 1552 _exp_region_bm(exp_region_bm), _exp_card_bm(exp_card_bm), 1553 _failures(0) { } 1554 1555 int failures() const { return _failures; } 1556 1557 bool doHeapRegion(HeapRegion* hr) { 1558 if (hr->is_continues_humongous()) { 1559 // We will ignore these here and process them when their 1560 // associated "starts humongous" region is processed (see 1561 // set_bit_for_heap_region()). Note that we cannot rely on their 1562 // associated "starts humongous" region to have their bit set to 1563 // 1 since, due to the region chunking in the parallel region 1564 // iteration, a "continues humongous" region might be visited 1565 // before its associated "starts humongous". 1566 return false; 1567 } 1568 1569 int failures = 0; 1570 1571 // Call the CalcLiveObjectsClosure to walk the marking bitmap for 1572 // this region and set the corresponding bits in the expected region 1573 // and card bitmaps. 1574 bool res = _calc_cl.doHeapRegion(hr); 1575 assert(res == false, "should be continuing"); 1576 1577 MutexLockerEx x((_verbose ? ParGCRareEvent_lock : NULL), 1578 Mutex::_no_safepoint_check_flag); 1579 1580 // Verify the marked bytes for this region. 1581 size_t exp_marked_bytes = _calc_cl.region_marked_bytes(); 1582 size_t act_marked_bytes = hr->next_marked_bytes(); 1583 1584 // We're not OK if expected marked bytes > actual marked bytes. It means 1585 // we have missed accounting some objects during the actual marking. 1586 if (exp_marked_bytes > act_marked_bytes) { 1587 if (_verbose) { 1588 gclog_or_tty->print_cr("Region %u: marked bytes mismatch: " 1589 "expected: " SIZE_FORMAT ", actual: " SIZE_FORMAT, 1590 hr->hrm_index(), exp_marked_bytes, act_marked_bytes); 1591 } 1592 failures += 1; 1593 } 1594 1595 // Verify the bit, for this region, in the actual and expected 1596 // (which was just calculated) region bit maps. 1597 // We're not OK if the bit in the calculated expected region 1598 // bitmap is set and the bit in the actual region bitmap is not. 1599 BitMap::idx_t index = (BitMap::idx_t) hr->hrm_index(); 1600 1601 bool expected = _exp_region_bm->at(index); 1602 bool actual = _region_bm->at(index); 1603 if (expected && !actual) { 1604 if (_verbose) { 1605 gclog_or_tty->print_cr("Region %u: region bitmap mismatch: " 1606 "expected: %s, actual: %s", 1607 hr->hrm_index(), 1608 BOOL_TO_STR(expected), BOOL_TO_STR(actual)); 1609 } 1610 failures += 1; 1611 } 1612 1613 // Verify that the card bit maps for the cards spanned by the current 1614 // region match. We have an error if we have a set bit in the expected 1615 // bit map and the corresponding bit in the actual bitmap is not set. 1616 1617 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(hr->bottom()); 1618 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(hr->top()); 1619 1620 for (BitMap::idx_t i = start_idx; i < end_idx; i+=1) { 1621 expected = _exp_card_bm->at(i); 1622 actual = _card_bm->at(i); 1623 1624 if (expected && !actual) { 1625 if (_verbose) { 1626 gclog_or_tty->print_cr("Region %u: card bitmap mismatch at " SIZE_FORMAT ": " 1627 "expected: %s, actual: %s", 1628 hr->hrm_index(), i, 1629 BOOL_TO_STR(expected), BOOL_TO_STR(actual)); 1630 } 1631 failures += 1; 1632 } 1633 } 1634 1635 if (failures > 0 && _verbose) { 1636 gclog_or_tty->print_cr("Region " HR_FORMAT ", ntams: " PTR_FORMAT ", " 1637 "marked_bytes: calc/actual " SIZE_FORMAT "/" SIZE_FORMAT, 1638 HR_FORMAT_PARAMS(hr), p2i(hr->next_top_at_mark_start()), 1639 _calc_cl.region_marked_bytes(), hr->next_marked_bytes()); 1640 } 1641 1642 _failures += failures; 1643 1644 // We could stop iteration over the heap when we 1645 // find the first violating region by returning true. 1646 return false; 1647 } 1648 }; 1649 1650 class G1ParVerifyFinalCountTask: public AbstractGangTask { 1651 protected: 1652 G1CollectedHeap* _g1h; 1653 ConcurrentMark* _cm; 1654 BitMap* _actual_region_bm; 1655 BitMap* _actual_card_bm; 1656 1657 uint _n_workers; 1658 1659 BitMap* _expected_region_bm; 1660 BitMap* _expected_card_bm; 1661 1662 int _failures; 1663 bool _verbose; 1664 1665 HeapRegionClaimer _hrclaimer; 1666 1667 public: 1668 G1ParVerifyFinalCountTask(G1CollectedHeap* g1h, 1669 BitMap* region_bm, BitMap* card_bm, 1670 BitMap* expected_region_bm, BitMap* expected_card_bm) 1671 : AbstractGangTask("G1 verify final counting"), 1672 _g1h(g1h), _cm(_g1h->concurrent_mark()), 1673 _actual_region_bm(region_bm), _actual_card_bm(card_bm), 1674 _expected_region_bm(expected_region_bm), _expected_card_bm(expected_card_bm), 1675 _failures(0), _verbose(false), 1676 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) { 1677 assert(VerifyDuringGC, "don't call this otherwise"); 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 _g1h->heap_region_par_iterate(&verify_cl, worker_id, &_hrclaimer); 1694 1695 Atomic::add(verify_cl.failures(), &_failures); 1696 } 1697 1698 int failures() const { return _failures; } 1699 }; 1700 1701 // Closure that finalizes the liveness counting data. 1702 // Used during the cleanup pause. 1703 // Sets the bits corresponding to the interval [NTAMS, top] 1704 // (which contains the implicitly live objects) in the 1705 // card liveness bitmap. Also sets the bit for each region, 1706 // containing live data, in the region liveness bitmap. 1707 1708 class FinalCountDataUpdateClosure: public CMCountDataClosureBase { 1709 public: 1710 FinalCountDataUpdateClosure(G1CollectedHeap* g1h, 1711 BitMap* region_bm, 1712 BitMap* card_bm) : 1713 CMCountDataClosureBase(g1h, region_bm, card_bm) { } 1714 1715 bool doHeapRegion(HeapRegion* hr) { 1716 1717 if (hr->is_continues_humongous()) { 1718 // We will ignore these here and process them when their 1719 // associated "starts humongous" region is processed (see 1720 // set_bit_for_heap_region()). Note that we cannot rely on their 1721 // associated "starts humongous" region to have their bit set to 1722 // 1 since, due to the region chunking in the parallel region 1723 // iteration, a "continues humongous" region might be visited 1724 // before its associated "starts humongous". 1725 return false; 1726 } 1727 1728 HeapWord* ntams = hr->next_top_at_mark_start(); 1729 HeapWord* top = hr->top(); 1730 1731 assert(hr->bottom() <= ntams && ntams <= hr->end(), "Preconditions."); 1732 1733 // Mark the allocated-since-marking portion... 1734 if (ntams < top) { 1735 // This definitely means the region has live objects. 1736 set_bit_for_region(hr); 1737 1738 // Now set the bits in the card bitmap for [ntams, top) 1739 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(ntams); 1740 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(top); 1741 1742 // Note: if we're looking at the last region in heap - top 1743 // could be actually just beyond the end of the heap; end_idx 1744 // will then correspond to a (non-existent) card that is also 1745 // just beyond the heap. 1746 if (_g1h->is_in_g1_reserved(top) && !_ct_bs->is_card_aligned(top)) { 1747 // end of object is not card aligned - increment to cover 1748 // all the cards spanned by the object 1749 end_idx += 1; 1750 } 1751 1752 assert(end_idx <= _card_bm->size(), 1753 err_msg("oob: end_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT, 1754 end_idx, _card_bm->size())); 1755 assert(start_idx < _card_bm->size(), 1756 err_msg("oob: start_idx= "SIZE_FORMAT", bitmap size= "SIZE_FORMAT, 1757 start_idx, _card_bm->size())); 1758 1759 _cm->set_card_bitmap_range(_card_bm, start_idx, end_idx, true /* is_par */); 1760 } 1761 1762 // Set the bit for the region if it contains live data 1763 if (hr->next_marked_bytes() > 0) { 1764 set_bit_for_region(hr); 1765 } 1766 1767 return false; 1768 } 1769 }; 1770 1771 class G1ParFinalCountTask: public AbstractGangTask { 1772 protected: 1773 G1CollectedHeap* _g1h; 1774 ConcurrentMark* _cm; 1775 BitMap* _actual_region_bm; 1776 BitMap* _actual_card_bm; 1777 1778 uint _n_workers; 1779 HeapRegionClaimer _hrclaimer; 1780 1781 public: 1782 G1ParFinalCountTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm) 1783 : AbstractGangTask("G1 final counting"), 1784 _g1h(g1h), _cm(_g1h->concurrent_mark()), 1785 _actual_region_bm(region_bm), _actual_card_bm(card_bm), 1786 _n_workers(_g1h->workers()->active_workers()), _hrclaimer(_n_workers) { 1787 } 1788 1789 void work(uint worker_id) { 1790 assert(worker_id < _n_workers, "invariant"); 1791 1792 FinalCountDataUpdateClosure final_update_cl(_g1h, 1793 _actual_region_bm, 1794 _actual_card_bm); 1795 1796 _g1h->heap_region_par_iterate(&final_update_cl, worker_id, &_hrclaimer); 1797 } 1798 }; 1799 1800 class G1ParNoteEndTask; 1801 1802 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure { 1803 G1CollectedHeap* _g1; 1804 size_t _max_live_bytes; 1805 uint _regions_claimed; 1806 size_t _freed_bytes; 1807 FreeRegionList* _local_cleanup_list; 1808 HeapRegionSetCount _old_regions_removed; 1809 HeapRegionSetCount _humongous_regions_removed; 1810 HRRSCleanupTask* _hrrs_cleanup_task; 1811 double _claimed_region_time; 1812 double _max_region_time; 1813 1814 public: 1815 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, 1816 FreeRegionList* local_cleanup_list, 1817 HRRSCleanupTask* hrrs_cleanup_task) : 1818 _g1(g1), 1819 _max_live_bytes(0), _regions_claimed(0), 1820 _freed_bytes(0), 1821 _claimed_region_time(0.0), _max_region_time(0.0), 1822 _local_cleanup_list(local_cleanup_list), 1823 _old_regions_removed(), 1824 _humongous_regions_removed(), 1825 _hrrs_cleanup_task(hrrs_cleanup_task) { } 1826 1827 size_t freed_bytes() { return _freed_bytes; } 1828 const HeapRegionSetCount& old_regions_removed() { return _old_regions_removed; } 1829 const HeapRegionSetCount& humongous_regions_removed() { return _humongous_regions_removed; } 1830 1831 bool doHeapRegion(HeapRegion *hr) { 1832 if (hr->is_continues_humongous()) { 1833 return false; 1834 } 1835 // We use a claim value of zero here because all regions 1836 // were claimed with value 1 in the FinalCount task. 1837 _g1->reset_gc_time_stamps(hr); 1838 double start = os::elapsedTime(); 1839 _regions_claimed++; 1840 hr->note_end_of_marking(); 1841 _max_live_bytes += hr->max_live_bytes(); 1842 1843 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) { 1844 _freed_bytes += hr->used(); 1845 hr->set_containing_set(NULL); 1846 if (hr->is_humongous()) { 1847 assert(hr->is_starts_humongous(), "we should only see starts humongous"); 1848 _humongous_regions_removed.increment(1u, hr->capacity()); 1849 _g1->free_humongous_region(hr, _local_cleanup_list, true); 1850 } else { 1851 _old_regions_removed.increment(1u, hr->capacity()); 1852 _g1->free_region(hr, _local_cleanup_list, true); 1853 } 1854 } else { 1855 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task); 1856 } 1857 1858 double region_time = (os::elapsedTime() - start); 1859 _claimed_region_time += region_time; 1860 if (region_time > _max_region_time) { 1861 _max_region_time = region_time; 1862 } 1863 return false; 1864 } 1865 1866 size_t max_live_bytes() { return _max_live_bytes; } 1867 uint regions_claimed() { return _regions_claimed; } 1868 double claimed_region_time_sec() { return _claimed_region_time; } 1869 double max_region_time_sec() { return _max_region_time; } 1870 }; 1871 1872 class G1ParNoteEndTask: public AbstractGangTask { 1873 friend class G1NoteEndOfConcMarkClosure; 1874 1875 protected: 1876 G1CollectedHeap* _g1h; 1877 size_t _max_live_bytes; 1878 size_t _freed_bytes; 1879 FreeRegionList* _cleanup_list; 1880 HeapRegionClaimer _hrclaimer; 1881 1882 public: 1883 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) : 1884 AbstractGangTask("G1 note end"), _g1h(g1h), _max_live_bytes(0), _freed_bytes(0), _cleanup_list(cleanup_list), _hrclaimer(n_workers) { 1885 } 1886 1887 void work(uint worker_id) { 1888 double start = os::elapsedTime(); 1889 FreeRegionList local_cleanup_list("Local Cleanup List"); 1890 HRRSCleanupTask hrrs_cleanup_task; 1891 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list, 1892 &hrrs_cleanup_task); 1893 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer); 1894 assert(g1_note_end.complete(), "Shouldn't have yielded!"); 1895 1896 // Now update the lists 1897 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed()); 1898 { 1899 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 1900 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes()); 1901 _max_live_bytes += g1_note_end.max_live_bytes(); 1902 _freed_bytes += g1_note_end.freed_bytes(); 1903 1904 // If we iterate over the global cleanup list at the end of 1905 // cleanup to do this printing we will not guarantee to only 1906 // generate output for the newly-reclaimed regions (the list 1907 // might not be empty at the beginning of cleanup; we might 1908 // still be working on its previous contents). So we do the 1909 // printing here, before we append the new regions to the global 1910 // cleanup list. 1911 1912 G1HRPrinter* hr_printer = _g1h->hr_printer(); 1913 if (hr_printer->is_active()) { 1914 FreeRegionListIterator iter(&local_cleanup_list); 1915 while (iter.more_available()) { 1916 HeapRegion* hr = iter.get_next(); 1917 hr_printer->cleanup(hr); 1918 } 1919 } 1920 1921 _cleanup_list->add_ordered(&local_cleanup_list); 1922 assert(local_cleanup_list.is_empty(), "post-condition"); 1923 1924 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task); 1925 } 1926 } 1927 size_t max_live_bytes() { return _max_live_bytes; } 1928 size_t freed_bytes() { return _freed_bytes; } 1929 }; 1930 1931 class G1ParScrubRemSetTask: public AbstractGangTask { 1932 protected: 1933 G1RemSet* _g1rs; 1934 BitMap* _region_bm; 1935 BitMap* _card_bm; 1936 HeapRegionClaimer _hrclaimer; 1937 1938 public: 1939 G1ParScrubRemSetTask(G1CollectedHeap* g1h, BitMap* region_bm, BitMap* card_bm, uint n_workers) : 1940 AbstractGangTask("G1 ScrubRS"), _g1rs(g1h->g1_rem_set()), _region_bm(region_bm), _card_bm(card_bm), _hrclaimer(n_workers) { 1941 } 1942 1943 void work(uint worker_id) { 1944 _g1rs->scrub(_region_bm, _card_bm, worker_id, &_hrclaimer); 1945 } 1946 1947 }; 1948 1949 void ConcurrentMark::cleanup() { 1950 // world is stopped at this checkpoint 1951 assert(SafepointSynchronize::is_at_safepoint(), 1952 "world should be stopped"); 1953 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1954 1955 // If a full collection has happened, we shouldn't do this. 1956 if (has_aborted()) { 1957 g1h->set_marking_complete(); // So bitmap clearing isn't confused 1958 return; 1959 } 1960 1961 g1h->verify_region_sets_optional(); 1962 1963 if (VerifyDuringGC) { 1964 HandleMark hm; // handle scope 1965 Universe::heap()->prepare_for_verify(); 1966 Universe::verify(VerifyOption_G1UsePrevMarking, 1967 " VerifyDuringGC:(before)"); 1968 } 1969 g1h->check_bitmaps("Cleanup Start"); 1970 1971 G1CollectorPolicy* g1p = G1CollectedHeap::heap()->g1_policy(); 1972 g1p->record_concurrent_mark_cleanup_start(); 1973 1974 double start = os::elapsedTime(); 1975 1976 HeapRegionRemSet::reset_for_cleanup_tasks(); 1977 1978 uint n_workers; 1979 1980 // Do counting once more with the world stopped for good measure. 1981 G1ParFinalCountTask g1_par_count_task(g1h, &_region_bm, &_card_bm); 1982 1983 g1h->set_par_threads(); 1984 n_workers = g1h->n_par_threads(); 1985 assert(g1h->n_par_threads() == n_workers, 1986 "Should not have been reset"); 1987 g1h->workers()->run_task(&g1_par_count_task); 1988 // Done with the parallel phase so reset to 0. 1989 g1h->set_par_threads(0); 1990 1991 if (VerifyDuringGC) { 1992 // Verify that the counting data accumulated during marking matches 1993 // that calculated by walking the marking bitmap. 1994 1995 // Bitmaps to hold expected values 1996 BitMap expected_region_bm(_region_bm.size(), true); 1997 BitMap expected_card_bm(_card_bm.size(), true); 1998 1999 G1ParVerifyFinalCountTask g1_par_verify_task(g1h, 2000 &_region_bm, 2001 &_card_bm, 2002 &expected_region_bm, 2003 &expected_card_bm); 2004 2005 g1h->set_par_threads((int)n_workers); 2006 g1h->workers()->run_task(&g1_par_verify_task); 2007 // Done with the parallel phase so reset to 0. 2008 g1h->set_par_threads(0); 2009 2010 guarantee(g1_par_verify_task.failures() == 0, "Unexpected accounting failures"); 2011 } 2012 2013 size_t start_used_bytes = g1h->used(); 2014 g1h->set_marking_complete(); 2015 2016 double count_end = os::elapsedTime(); 2017 double this_final_counting_time = (count_end - start); 2018 _total_counting_time += this_final_counting_time; 2019 2020 if (G1PrintRegionLivenessInfo) { 2021 G1PrintRegionLivenessInfoClosure cl(gclog_or_tty, "Post-Marking"); 2022 _g1h->heap_region_iterate(&cl); 2023 } 2024 2025 // Install newly created mark bitMap as "prev". 2026 swapMarkBitMaps(); 2027 2028 g1h->reset_gc_time_stamp(); 2029 2030 // Note end of marking in all heap regions. 2031 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers); 2032 g1h->set_par_threads((int)n_workers); 2033 g1h->workers()->run_task(&g1_par_note_end_task); 2034 g1h->set_par_threads(0); 2035 g1h->check_gc_time_stamps(); 2036 2037 if (!cleanup_list_is_empty()) { 2038 // The cleanup list is not empty, so we'll have to process it 2039 // concurrently. Notify anyone else that might be wanting free 2040 // regions that there will be more free regions coming soon. 2041 g1h->set_free_regions_coming(); 2042 } 2043 2044 // call below, since it affects the metric by which we sort the heap 2045 // regions. 2046 if (G1ScrubRemSets) { 2047 double rs_scrub_start = os::elapsedTime(); 2048 G1ParScrubRemSetTask g1_par_scrub_rs_task(g1h, &_region_bm, &_card_bm, n_workers); 2049 g1h->set_par_threads((int)n_workers); 2050 g1h->workers()->run_task(&g1_par_scrub_rs_task); 2051 g1h->set_par_threads(0); 2052 2053 double rs_scrub_end = os::elapsedTime(); 2054 double this_rs_scrub_time = (rs_scrub_end - rs_scrub_start); 2055 _total_rs_scrub_time += this_rs_scrub_time; 2056 } 2057 2058 // this will also free any regions totally full of garbage objects, 2059 // and sort the regions. 2060 g1h->g1_policy()->record_concurrent_mark_cleanup_end((int)n_workers); 2061 2062 // Statistics. 2063 double end = os::elapsedTime(); 2064 _cleanup_times.add((end - start) * 1000.0); 2065 2066 if (G1Log::fine()) { 2067 g1h->print_size_transition(gclog_or_tty, 2068 start_used_bytes, 2069 g1h->used(), 2070 g1h->capacity()); 2071 } 2072 2073 // Clean up will have freed any regions completely full of garbage. 2074 // Update the soft reference policy with the new heap occupancy. 2075 Universe::update_heap_info_at_gc(); 2076 2077 if (VerifyDuringGC) { 2078 HandleMark hm; // handle scope 2079 Universe::heap()->prepare_for_verify(); 2080 Universe::verify(VerifyOption_G1UsePrevMarking, 2081 " VerifyDuringGC:(after)"); 2082 } 2083 2084 g1h->check_bitmaps("Cleanup End"); 2085 2086 g1h->verify_region_sets_optional(); 2087 2088 // We need to make this be a "collection" so any collection pause that 2089 // races with it goes around and waits for completeCleanup to finish. 2090 g1h->increment_total_collections(); 2091 2092 // Clean out dead classes and update Metaspace sizes. 2093 if (ClassUnloadingWithConcurrentMark) { 2094 ClassLoaderDataGraph::purge(); 2095 } 2096 MetaspaceGC::compute_new_size(); 2097 2098 // We reclaimed old regions so we should calculate the sizes to make 2099 // sure we update the old gen/space data. 2100 g1h->g1mm()->update_sizes(); 2101 2102 g1h->trace_heap_after_concurrent_cycle(); 2103 } 2104 2105 void ConcurrentMark::completeCleanup() { 2106 if (has_aborted()) return; 2107 2108 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2109 2110 _cleanup_list.verify_optional(); 2111 FreeRegionList tmp_free_list("Tmp Free List"); 2112 2113 if (G1ConcRegionFreeingVerbose) { 2114 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : " 2115 "cleanup list has %u entries", 2116 _cleanup_list.length()); 2117 } 2118 2119 // No one else should be accessing the _cleanup_list at this point, 2120 // so it is not necessary to take any locks 2121 while (!_cleanup_list.is_empty()) { 2122 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */); 2123 assert(hr != NULL, "Got NULL from a non-empty list"); 2124 hr->par_clear(); 2125 tmp_free_list.add_ordered(hr); 2126 2127 // Instead of adding one region at a time to the secondary_free_list, 2128 // we accumulate them in the local list and move them a few at a 2129 // time. This also cuts down on the number of notify_all() calls 2130 // we do during this process. We'll also append the local list when 2131 // _cleanup_list is empty (which means we just removed the last 2132 // region from the _cleanup_list). 2133 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) || 2134 _cleanup_list.is_empty()) { 2135 if (G1ConcRegionFreeingVerbose) { 2136 gclog_or_tty->print_cr("G1ConcRegionFreeing [complete cleanup] : " 2137 "appending %u entries to the secondary_free_list, " 2138 "cleanup list still has %u entries", 2139 tmp_free_list.length(), 2140 _cleanup_list.length()); 2141 } 2142 2143 { 2144 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag); 2145 g1h->secondary_free_list_add(&tmp_free_list); 2146 SecondaryFreeList_lock->notify_all(); 2147 } 2148 2149 if (G1StressConcRegionFreeing) { 2150 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) { 2151 os::sleep(Thread::current(), (jlong) 1, false); 2152 } 2153 } 2154 } 2155 } 2156 assert(tmp_free_list.is_empty(), "post-condition"); 2157 } 2158 2159 // Supporting Object and Oop closures for reference discovery 2160 // and processing in during marking 2161 2162 bool G1CMIsAliveClosure::do_object_b(oop obj) { 2163 HeapWord* addr = (HeapWord*)obj; 2164 return addr != NULL && 2165 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj)); 2166 } 2167 2168 // 'Keep Alive' oop closure used by both serial parallel reference processing. 2169 // Uses the CMTask associated with a worker thread (for serial reference 2170 // processing the CMTask for worker 0 is used) to preserve (mark) and 2171 // trace referent objects. 2172 // 2173 // Using the CMTask and embedded local queues avoids having the worker 2174 // threads operating on the global mark stack. This reduces the risk 2175 // of overflowing the stack - which we would rather avoid at this late 2176 // state. Also using the tasks' local queues removes the potential 2177 // of the workers interfering with each other that could occur if 2178 // operating on the global stack. 2179 2180 class G1CMKeepAliveAndDrainClosure: public OopClosure { 2181 ConcurrentMark* _cm; 2182 CMTask* _task; 2183 int _ref_counter_limit; 2184 int _ref_counter; 2185 bool _is_serial; 2186 public: 2187 G1CMKeepAliveAndDrainClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) : 2188 _cm(cm), _task(task), _is_serial(is_serial), 2189 _ref_counter_limit(G1RefProcDrainInterval) { 2190 assert(_ref_counter_limit > 0, "sanity"); 2191 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 2192 _ref_counter = _ref_counter_limit; 2193 } 2194 2195 virtual void do_oop(narrowOop* p) { do_oop_work(p); } 2196 virtual void do_oop( oop* p) { do_oop_work(p); } 2197 2198 template <class T> void do_oop_work(T* p) { 2199 if (!_cm->has_overflown()) { 2200 oop obj = oopDesc::load_decode_heap_oop(p); 2201 if (_cm->verbose_high()) { 2202 gclog_or_tty->print_cr("\t[%u] we're looking at location " 2203 "*"PTR_FORMAT" = "PTR_FORMAT, 2204 _task->worker_id(), p2i(p), p2i((void*) obj)); 2205 } 2206 2207 _task->deal_with_reference(obj); 2208 _ref_counter--; 2209 2210 if (_ref_counter == 0) { 2211 // We have dealt with _ref_counter_limit references, pushing them 2212 // and objects reachable from them on to the local stack (and 2213 // possibly the global stack). Call CMTask::do_marking_step() to 2214 // process these entries. 2215 // 2216 // We call CMTask::do_marking_step() in a loop, which we'll exit if 2217 // there's nothing more to do (i.e. we're done with the entries that 2218 // were pushed as a result of the CMTask::deal_with_reference() calls 2219 // above) or we overflow. 2220 // 2221 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() 2222 // flag while there may still be some work to do. (See the comment at 2223 // the beginning of CMTask::do_marking_step() for those conditions - 2224 // one of which is reaching the specified time target.) It is only 2225 // when CMTask::do_marking_step() returns without setting the 2226 // has_aborted() flag that the marking step has completed. 2227 do { 2228 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 2229 _task->do_marking_step(mark_step_duration_ms, 2230 false /* do_termination */, 2231 _is_serial); 2232 } while (_task->has_aborted() && !_cm->has_overflown()); 2233 _ref_counter = _ref_counter_limit; 2234 } 2235 } else { 2236 if (_cm->verbose_high()) { 2237 gclog_or_tty->print_cr("\t[%u] CM Overflow", _task->worker_id()); 2238 } 2239 } 2240 } 2241 }; 2242 2243 // 'Drain' oop closure used by both serial and parallel reference processing. 2244 // Uses the CMTask associated with a given worker thread (for serial 2245 // reference processing the CMtask for worker 0 is used). Calls the 2246 // do_marking_step routine, with an unbelievably large timeout value, 2247 // to drain the marking data structures of the remaining entries 2248 // added by the 'keep alive' oop closure above. 2249 2250 class G1CMDrainMarkingStackClosure: public VoidClosure { 2251 ConcurrentMark* _cm; 2252 CMTask* _task; 2253 bool _is_serial; 2254 public: 2255 G1CMDrainMarkingStackClosure(ConcurrentMark* cm, CMTask* task, bool is_serial) : 2256 _cm(cm), _task(task), _is_serial(is_serial) { 2257 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 2258 } 2259 2260 void do_void() { 2261 do { 2262 if (_cm->verbose_high()) { 2263 gclog_or_tty->print_cr("\t[%u] Drain: Calling do_marking_step - serial: %s", 2264 _task->worker_id(), BOOL_TO_STR(_is_serial)); 2265 } 2266 2267 // We call CMTask::do_marking_step() to completely drain the local 2268 // and global marking stacks of entries pushed by the 'keep alive' 2269 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above). 2270 // 2271 // CMTask::do_marking_step() is called in a loop, which we'll exit 2272 // if there's nothing more to do (i.e. we've completely drained the 2273 // entries that were pushed as a a result of applying the 'keep alive' 2274 // closure to the entries on the discovered ref lists) or we overflow 2275 // the global marking stack. 2276 // 2277 // Note: CMTask::do_marking_step() can set the CMTask::has_aborted() 2278 // flag while there may still be some work to do. (See the comment at 2279 // the beginning of CMTask::do_marking_step() for those conditions - 2280 // one of which is reaching the specified time target.) It is only 2281 // when CMTask::do_marking_step() returns without setting the 2282 // has_aborted() flag that the marking step has completed. 2283 2284 _task->do_marking_step(1000000000.0 /* something very large */, 2285 true /* do_termination */, 2286 _is_serial); 2287 } while (_task->has_aborted() && !_cm->has_overflown()); 2288 } 2289 }; 2290 2291 // Implementation of AbstractRefProcTaskExecutor for parallel 2292 // reference processing at the end of G1 concurrent marking 2293 2294 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor { 2295 private: 2296 G1CollectedHeap* _g1h; 2297 ConcurrentMark* _cm; 2298 WorkGang* _workers; 2299 int _active_workers; 2300 2301 public: 2302 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h, 2303 ConcurrentMark* cm, 2304 WorkGang* workers, 2305 int n_workers) : 2306 _g1h(g1h), _cm(cm), 2307 _workers(workers), _active_workers(n_workers) { } 2308 2309 // Executes the given task using concurrent marking worker threads. 2310 virtual void execute(ProcessTask& task); 2311 virtual void execute(EnqueueTask& task); 2312 }; 2313 2314 class G1CMRefProcTaskProxy: public AbstractGangTask { 2315 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask; 2316 ProcessTask& _proc_task; 2317 G1CollectedHeap* _g1h; 2318 ConcurrentMark* _cm; 2319 2320 public: 2321 G1CMRefProcTaskProxy(ProcessTask& proc_task, 2322 G1CollectedHeap* g1h, 2323 ConcurrentMark* cm) : 2324 AbstractGangTask("Process reference objects in parallel"), 2325 _proc_task(proc_task), _g1h(g1h), _cm(cm) { 2326 ReferenceProcessor* rp = _g1h->ref_processor_cm(); 2327 assert(rp->processing_is_mt(), "shouldn't be here otherwise"); 2328 } 2329 2330 virtual void work(uint worker_id) { 2331 ResourceMark rm; 2332 HandleMark hm; 2333 CMTask* task = _cm->task(worker_id); 2334 G1CMIsAliveClosure g1_is_alive(_g1h); 2335 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */); 2336 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */); 2337 2338 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain); 2339 } 2340 }; 2341 2342 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) { 2343 assert(_workers != NULL, "Need parallel worker threads."); 2344 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 2345 2346 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm); 2347 2348 // We need to reset the concurrency level before each 2349 // proxy task execution, so that the termination protocol 2350 // and overflow handling in CMTask::do_marking_step() knows 2351 // how many workers to wait for. 2352 _cm->set_concurrency(_active_workers); 2353 _g1h->set_par_threads(_active_workers); 2354 _workers->run_task(&proc_task_proxy); 2355 _g1h->set_par_threads(0); 2356 } 2357 2358 class G1CMRefEnqueueTaskProxy: public AbstractGangTask { 2359 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask; 2360 EnqueueTask& _enq_task; 2361 2362 public: 2363 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) : 2364 AbstractGangTask("Enqueue reference objects in parallel"), 2365 _enq_task(enq_task) { } 2366 2367 virtual void work(uint worker_id) { 2368 _enq_task.work(worker_id); 2369 } 2370 }; 2371 2372 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) { 2373 assert(_workers != NULL, "Need parallel worker threads."); 2374 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 2375 2376 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task); 2377 2378 // Not strictly necessary but... 2379 // 2380 // We need to reset the concurrency level before each 2381 // proxy task execution, so that the termination protocol 2382 // and overflow handling in CMTask::do_marking_step() knows 2383 // how many workers to wait for. 2384 _cm->set_concurrency(_active_workers); 2385 _g1h->set_par_threads(_active_workers); 2386 _workers->run_task(&enq_task_proxy); 2387 _g1h->set_par_threads(0); 2388 } 2389 2390 void ConcurrentMark::weakRefsWorkParallelPart(BoolObjectClosure* is_alive, bool purged_classes) { 2391 G1CollectedHeap::heap()->parallel_cleaning(is_alive, true, true, purged_classes); 2392 } 2393 2394 void ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) { 2395 if (has_overflown()) { 2396 // Skip processing the discovered references if we have 2397 // overflown the global marking stack. Reference objects 2398 // only get discovered once so it is OK to not 2399 // de-populate the discovered reference lists. We could have, 2400 // but the only benefit would be that, when marking restarts, 2401 // less reference objects are discovered. 2402 return; 2403 } 2404 2405 ResourceMark rm; 2406 HandleMark hm; 2407 2408 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2409 2410 // Is alive closure. 2411 G1CMIsAliveClosure g1_is_alive(g1h); 2412 2413 // Inner scope to exclude the cleaning of the string and symbol 2414 // tables from the displayed time. 2415 { 2416 G1CMTraceTime t("GC ref-proc", G1Log::finer()); 2417 2418 ReferenceProcessor* rp = g1h->ref_processor_cm(); 2419 2420 // See the comment in G1CollectedHeap::ref_processing_init() 2421 // about how reference processing currently works in G1. 2422 2423 // Set the soft reference policy 2424 rp->setup_policy(clear_all_soft_refs); 2425 assert(_markStack.isEmpty(), "mark stack should be empty"); 2426 2427 // Instances of the 'Keep Alive' and 'Complete GC' closures used 2428 // in serial reference processing. Note these closures are also 2429 // used for serially processing (by the the current thread) the 2430 // JNI references during parallel reference processing. 2431 // 2432 // These closures do not need to synchronize with the worker 2433 // threads involved in parallel reference processing as these 2434 // instances are executed serially by the current thread (e.g. 2435 // reference processing is not multi-threaded and is thus 2436 // performed by the current thread instead of a gang worker). 2437 // 2438 // The gang tasks involved in parallel reference processing create 2439 // their own instances of these closures, which do their own 2440 // synchronization among themselves. 2441 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */); 2442 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */); 2443 2444 // We need at least one active thread. If reference processing 2445 // is not multi-threaded we use the current (VMThread) thread, 2446 // otherwise we use the work gang from the G1CollectedHeap and 2447 // we utilize all the worker threads we can. 2448 bool processing_is_mt = rp->processing_is_mt(); 2449 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U); 2450 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U); 2451 2452 // Parallel processing task executor. 2453 G1CMRefProcTaskExecutor par_task_executor(g1h, this, 2454 g1h->workers(), active_workers); 2455 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL); 2456 2457 // Set the concurrency level. The phase was already set prior to 2458 // executing the remark task. 2459 set_concurrency(active_workers); 2460 2461 // Set the degree of MT processing here. If the discovery was done MT, 2462 // the number of threads involved during discovery could differ from 2463 // the number of active workers. This is OK as long as the discovered 2464 // Reference lists are balanced (see balance_all_queues() and balance_queues()). 2465 rp->set_active_mt_degree(active_workers); 2466 2467 // Process the weak references. 2468 const ReferenceProcessorStats& stats = 2469 rp->process_discovered_references(&g1_is_alive, 2470 &g1_keep_alive, 2471 &g1_drain_mark_stack, 2472 executor, 2473 g1h->gc_timer_cm(), 2474 concurrent_gc_id()); 2475 g1h->gc_tracer_cm()->report_gc_reference_stats(stats); 2476 2477 // The do_oop work routines of the keep_alive and drain_marking_stack 2478 // oop closures will set the has_overflown flag if we overflow the 2479 // global marking stack. 2480 2481 assert(_markStack.overflow() || _markStack.isEmpty(), 2482 "mark stack should be empty (unless it overflowed)"); 2483 2484 if (_markStack.overflow()) { 2485 // This should have been done already when we tried to push an 2486 // entry on to the global mark stack. But let's do it again. 2487 set_has_overflown(); 2488 } 2489 2490 assert(rp->num_q() == active_workers, "why not"); 2491 2492 rp->enqueue_discovered_references(executor); 2493 2494 rp->verify_no_references_recorded(); 2495 assert(!rp->discovery_enabled(), "Post condition"); 2496 } 2497 2498 if (has_overflown()) { 2499 // We can not trust g1_is_alive if the marking stack overflowed 2500 return; 2501 } 2502 2503 assert(_markStack.isEmpty(), "Marking should have completed"); 2504 2505 // Unload Klasses, String, Symbols, Code Cache, etc. 2506 { 2507 G1CMTraceTime trace("Unloading", G1Log::finer()); 2508 2509 if (ClassUnloadingWithConcurrentMark) { 2510 bool purged_classes; 2511 2512 { 2513 G1CMTraceTime trace("System Dictionary Unloading", G1Log::finest()); 2514 purged_classes = SystemDictionary::do_unloading(&g1_is_alive); 2515 } 2516 2517 { 2518 G1CMTraceTime trace("Parallel Unloading", G1Log::finest()); 2519 weakRefsWorkParallelPart(&g1_is_alive, purged_classes); 2520 } 2521 } 2522 2523 if (G1StringDedup::is_enabled()) { 2524 G1CMTraceTime trace("String Deduplication Unlink", G1Log::finest()); 2525 G1StringDedup::unlink(&g1_is_alive); 2526 } 2527 } 2528 } 2529 2530 void ConcurrentMark::swapMarkBitMaps() { 2531 CMBitMapRO* temp = _prevMarkBitMap; 2532 _prevMarkBitMap = (CMBitMapRO*)_nextMarkBitMap; 2533 _nextMarkBitMap = (CMBitMap*) temp; 2534 } 2535 2536 class CMObjectClosure; 2537 2538 // Closure for iterating over objects, currently only used for 2539 // processing SATB buffers. 2540 class CMObjectClosure : public ObjectClosure { 2541 private: 2542 CMTask* _task; 2543 2544 public: 2545 void do_object(oop obj) { 2546 _task->deal_with_reference(obj); 2547 } 2548 2549 CMObjectClosure(CMTask* task) : _task(task) { } 2550 }; 2551 2552 class G1RemarkThreadsClosure : public ThreadClosure { 2553 CMObjectClosure _cm_obj; 2554 G1CMOopClosure _cm_cl; 2555 MarkingCodeBlobClosure _code_cl; 2556 int _thread_parity; 2557 2558 public: 2559 G1RemarkThreadsClosure(G1CollectedHeap* g1h, CMTask* task) : 2560 _cm_obj(task), _cm_cl(g1h, g1h->concurrent_mark(), task), _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations), 2561 _thread_parity(SharedHeap::heap()->strong_roots_parity()) {} 2562 2563 void do_thread(Thread* thread) { 2564 if (thread->is_Java_thread()) { 2565 if (thread->claim_oops_do(true, _thread_parity)) { 2566 JavaThread* jt = (JavaThread*)thread; 2567 2568 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking 2569 // however the liveness of oops reachable from nmethods have very complex lifecycles: 2570 // * Alive if on the stack of an executing method 2571 // * Weakly reachable otherwise 2572 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be 2573 // live by the SATB invariant but other oops recorded in nmethods may behave differently. 2574 jt->nmethods_do(&_code_cl); 2575 2576 jt->satb_mark_queue().apply_closure_and_empty(&_cm_obj); 2577 } 2578 } else if (thread->is_VM_thread()) { 2579 if (thread->claim_oops_do(true, _thread_parity)) { 2580 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_obj); 2581 } 2582 } 2583 } 2584 }; 2585 2586 class CMRemarkTask: public AbstractGangTask { 2587 private: 2588 ConcurrentMark* _cm; 2589 public: 2590 void work(uint worker_id) { 2591 // Since all available tasks are actually started, we should 2592 // only proceed if we're supposed to be active. 2593 if (worker_id < _cm->active_tasks()) { 2594 CMTask* task = _cm->task(worker_id); 2595 task->record_start_time(); 2596 { 2597 ResourceMark rm; 2598 HandleMark hm; 2599 2600 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task); 2601 Threads::threads_do(&threads_f); 2602 } 2603 2604 do { 2605 task->do_marking_step(1000000000.0 /* something very large */, 2606 true /* do_termination */, 2607 false /* is_serial */); 2608 } while (task->has_aborted() && !_cm->has_overflown()); 2609 // If we overflow, then we do not want to restart. We instead 2610 // want to abort remark and do concurrent marking again. 2611 task->record_end_time(); 2612 } 2613 } 2614 2615 CMRemarkTask(ConcurrentMark* cm, int active_workers) : 2616 AbstractGangTask("Par Remark"), _cm(cm) { 2617 _cm->terminator()->reset_for_reuse(active_workers); 2618 } 2619 }; 2620 2621 void ConcurrentMark::checkpointRootsFinalWork() { 2622 ResourceMark rm; 2623 HandleMark hm; 2624 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 2625 2626 G1CMTraceTime trace("Finalize Marking", G1Log::finer()); 2627 2628 g1h->ensure_parsability(false); 2629 2630 G1CollectedHeap::StrongRootsScope srs(g1h); 2631 // this is remark, so we'll use up all active threads 2632 uint active_workers = g1h->workers()->active_workers(); 2633 if (active_workers == 0) { 2634 assert(active_workers > 0, "Should have been set earlier"); 2635 active_workers = (uint) ParallelGCThreads; 2636 g1h->workers()->set_active_workers(active_workers); 2637 } 2638 set_concurrency_and_phase(active_workers, false /* concurrent */); 2639 // Leave _parallel_marking_threads at it's 2640 // value originally calculated in the ConcurrentMark 2641 // constructor and pass values of the active workers 2642 // through the gang in the task. 2643 2644 CMRemarkTask remarkTask(this, active_workers); 2645 // We will start all available threads, even if we decide that the 2646 // active_workers will be fewer. The extra ones will just bail out 2647 // immediately. 2648 g1h->set_par_threads(active_workers); 2649 g1h->workers()->run_task(&remarkTask); 2650 g1h->set_par_threads(0); 2651 2652 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 2653 guarantee(has_overflown() || 2654 satb_mq_set.completed_buffers_num() == 0, 2655 err_msg("Invariant: has_overflown = %s, num buffers = %d", 2656 BOOL_TO_STR(has_overflown()), 2657 satb_mq_set.completed_buffers_num())); 2658 2659 print_stats(); 2660 } 2661 2662 #ifndef PRODUCT 2663 2664 class PrintReachableOopClosure: public OopClosure { 2665 private: 2666 G1CollectedHeap* _g1h; 2667 outputStream* _out; 2668 VerifyOption _vo; 2669 bool _all; 2670 2671 public: 2672 PrintReachableOopClosure(outputStream* out, 2673 VerifyOption vo, 2674 bool all) : 2675 _g1h(G1CollectedHeap::heap()), 2676 _out(out), _vo(vo), _all(all) { } 2677 2678 void do_oop(narrowOop* p) { do_oop_work(p); } 2679 void do_oop( oop* p) { do_oop_work(p); } 2680 2681 template <class T> void do_oop_work(T* p) { 2682 oop obj = oopDesc::load_decode_heap_oop(p); 2683 const char* str = NULL; 2684 const char* str2 = ""; 2685 2686 if (obj == NULL) { 2687 str = ""; 2688 } else if (!_g1h->is_in_g1_reserved(obj)) { 2689 str = " O"; 2690 } else { 2691 HeapRegion* hr = _g1h->heap_region_containing(obj); 2692 bool over_tams = _g1h->allocated_since_marking(obj, hr, _vo); 2693 bool marked = _g1h->is_marked(obj, _vo); 2694 2695 if (over_tams) { 2696 str = " >"; 2697 if (marked) { 2698 str2 = " AND MARKED"; 2699 } 2700 } else if (marked) { 2701 str = " M"; 2702 } else { 2703 str = " NOT"; 2704 } 2705 } 2706 2707 _out->print_cr(" "PTR_FORMAT": "PTR_FORMAT"%s%s", 2708 p2i(p), p2i((void*) obj), str, str2); 2709 } 2710 }; 2711 2712 class PrintReachableObjectClosure : public ObjectClosure { 2713 private: 2714 G1CollectedHeap* _g1h; 2715 outputStream* _out; 2716 VerifyOption _vo; 2717 bool _all; 2718 HeapRegion* _hr; 2719 2720 public: 2721 PrintReachableObjectClosure(outputStream* out, 2722 VerifyOption vo, 2723 bool all, 2724 HeapRegion* hr) : 2725 _g1h(G1CollectedHeap::heap()), 2726 _out(out), _vo(vo), _all(all), _hr(hr) { } 2727 2728 void do_object(oop o) { 2729 bool over_tams = _g1h->allocated_since_marking(o, _hr, _vo); 2730 bool marked = _g1h->is_marked(o, _vo); 2731 bool print_it = _all || over_tams || marked; 2732 2733 if (print_it) { 2734 _out->print_cr(" "PTR_FORMAT"%s", 2735 p2i((void *)o), (over_tams) ? " >" : (marked) ? " M" : ""); 2736 PrintReachableOopClosure oopCl(_out, _vo, _all); 2737 o->oop_iterate_no_header(&oopCl); 2738 } 2739 } 2740 }; 2741 2742 class PrintReachableRegionClosure : public HeapRegionClosure { 2743 private: 2744 G1CollectedHeap* _g1h; 2745 outputStream* _out; 2746 VerifyOption _vo; 2747 bool _all; 2748 2749 public: 2750 bool doHeapRegion(HeapRegion* hr) { 2751 HeapWord* b = hr->bottom(); 2752 HeapWord* e = hr->end(); 2753 HeapWord* t = hr->top(); 2754 HeapWord* p = _g1h->top_at_mark_start(hr, _vo); 2755 _out->print_cr("** ["PTR_FORMAT", "PTR_FORMAT"] top: "PTR_FORMAT" " 2756 "TAMS: " PTR_FORMAT, p2i(b), p2i(e), p2i(t), p2i(p)); 2757 _out->cr(); 2758 2759 HeapWord* from = b; 2760 HeapWord* to = t; 2761 2762 if (to > from) { 2763 _out->print_cr("Objects in [" PTR_FORMAT ", " PTR_FORMAT "]", p2i(from), p2i(to)); 2764 _out->cr(); 2765 PrintReachableObjectClosure ocl(_out, _vo, _all, hr); 2766 hr->object_iterate_mem_careful(MemRegion(from, to), &ocl); 2767 _out->cr(); 2768 } 2769 2770 return false; 2771 } 2772 2773 PrintReachableRegionClosure(outputStream* out, 2774 VerifyOption vo, 2775 bool all) : 2776 _g1h(G1CollectedHeap::heap()), _out(out), _vo(vo), _all(all) { } 2777 }; 2778 2779 void ConcurrentMark::print_reachable(const char* str, 2780 VerifyOption vo, 2781 bool all) { 2782 gclog_or_tty->cr(); 2783 gclog_or_tty->print_cr("== Doing heap dump... "); 2784 2785 if (G1PrintReachableBaseFile == NULL) { 2786 gclog_or_tty->print_cr(" #### error: no base file defined"); 2787 return; 2788 } 2789 2790 if (strlen(G1PrintReachableBaseFile) + 1 + strlen(str) > 2791 (JVM_MAXPATHLEN - 1)) { 2792 gclog_or_tty->print_cr(" #### error: file name too long"); 2793 return; 2794 } 2795 2796 char file_name[JVM_MAXPATHLEN]; 2797 sprintf(file_name, "%s.%s", G1PrintReachableBaseFile, str); 2798 gclog_or_tty->print_cr(" dumping to file %s", file_name); 2799 2800 fileStream fout(file_name); 2801 if (!fout.is_open()) { 2802 gclog_or_tty->print_cr(" #### error: could not open file"); 2803 return; 2804 } 2805 2806 outputStream* out = &fout; 2807 out->print_cr("-- USING %s", _g1h->top_at_mark_start_str(vo)); 2808 out->cr(); 2809 2810 out->print_cr("--- ITERATING OVER REGIONS"); 2811 out->cr(); 2812 PrintReachableRegionClosure rcl(out, vo, all); 2813 _g1h->heap_region_iterate(&rcl); 2814 out->cr(); 2815 2816 gclog_or_tty->print_cr(" done"); 2817 gclog_or_tty->flush(); 2818 } 2819 2820 #endif // PRODUCT 2821 2822 void ConcurrentMark::clearRangePrevBitmap(MemRegion mr) { 2823 // Note we are overriding the read-only view of the prev map here, via 2824 // the cast. 2825 ((CMBitMap*)_prevMarkBitMap)->clearRange(mr); 2826 } 2827 2828 void ConcurrentMark::clearRangeNextBitmap(MemRegion mr) { 2829 _nextMarkBitMap->clearRange(mr); 2830 } 2831 2832 HeapRegion* 2833 ConcurrentMark::claim_region(uint worker_id) { 2834 // "checkpoint" the finger 2835 HeapWord* finger = _finger; 2836 2837 // _heap_end will not change underneath our feet; it only changes at 2838 // yield points. 2839 while (finger < _heap_end) { 2840 assert(_g1h->is_in_g1_reserved(finger), "invariant"); 2841 2842 // Note on how this code handles humongous regions. In the 2843 // normal case the finger will reach the start of a "starts 2844 // humongous" (SH) region. Its end will either be the end of the 2845 // last "continues humongous" (CH) region in the sequence, or the 2846 // standard end of the SH region (if the SH is the only region in 2847 // the sequence). That way claim_region() will skip over the CH 2848 // regions. However, there is a subtle race between a CM thread 2849 // executing this method and a mutator thread doing a humongous 2850 // object allocation. The two are not mutually exclusive as the CM 2851 // thread does not need to hold the Heap_lock when it gets 2852 // here. So there is a chance that claim_region() will come across 2853 // a free region that's in the progress of becoming a SH or a CH 2854 // region. In the former case, it will either 2855 // a) Miss the update to the region's end, in which case it will 2856 // visit every subsequent CH region, will find their bitmaps 2857 // empty, and do nothing, or 2858 // b) Will observe the update of the region's end (in which case 2859 // it will skip the subsequent CH regions). 2860 // If it comes across a region that suddenly becomes CH, the 2861 // scenario will be similar to b). So, the race between 2862 // claim_region() and a humongous object allocation might force us 2863 // to do a bit of unnecessary work (due to some unnecessary bitmap 2864 // iterations) but it should not introduce and correctness issues. 2865 HeapRegion* curr_region = _g1h->heap_region_containing_raw(finger); 2866 2867 // Above heap_region_containing_raw may return NULL as we always scan claim 2868 // until the end of the heap. In this case, just jump to the next region. 2869 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords; 2870 2871 // Is the gap between reading the finger and doing the CAS too long? 2872 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger); 2873 if (res == finger && curr_region != NULL) { 2874 // we succeeded 2875 HeapWord* bottom = curr_region->bottom(); 2876 HeapWord* limit = curr_region->next_top_at_mark_start(); 2877 2878 if (verbose_low()) { 2879 gclog_or_tty->print_cr("[%u] curr_region = "PTR_FORMAT" " 2880 "["PTR_FORMAT", "PTR_FORMAT"), " 2881 "limit = "PTR_FORMAT, 2882 worker_id, p2i(curr_region), p2i(bottom), p2i(end), p2i(limit)); 2883 } 2884 2885 // notice that _finger == end cannot be guaranteed here since, 2886 // someone else might have moved the finger even further 2887 assert(_finger >= end, "the finger should have moved forward"); 2888 2889 if (verbose_low()) { 2890 gclog_or_tty->print_cr("[%u] we were successful with region = " 2891 PTR_FORMAT, worker_id, p2i(curr_region)); 2892 } 2893 2894 if (limit > bottom) { 2895 if (verbose_low()) { 2896 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is not empty, " 2897 "returning it ", worker_id, p2i(curr_region)); 2898 } 2899 return curr_region; 2900 } else { 2901 assert(limit == bottom, 2902 "the region limit should be at bottom"); 2903 if (verbose_low()) { 2904 gclog_or_tty->print_cr("[%u] region "PTR_FORMAT" is empty, " 2905 "returning NULL", worker_id, p2i(curr_region)); 2906 } 2907 // we return NULL and the caller should try calling 2908 // claim_region() again. 2909 return NULL; 2910 } 2911 } else { 2912 assert(_finger > finger, "the finger should have moved forward"); 2913 if (verbose_low()) { 2914 if (curr_region == NULL) { 2915 gclog_or_tty->print_cr("[%u] found uncommitted region, moving finger, " 2916 "global finger = "PTR_FORMAT", " 2917 "our finger = "PTR_FORMAT, 2918 worker_id, p2i(_finger), p2i(finger)); 2919 } else { 2920 gclog_or_tty->print_cr("[%u] somebody else moved the finger, " 2921 "global finger = "PTR_FORMAT", " 2922 "our finger = "PTR_FORMAT, 2923 worker_id, p2i(_finger), p2i(finger)); 2924 } 2925 } 2926 2927 // read it again 2928 finger = _finger; 2929 } 2930 } 2931 2932 return NULL; 2933 } 2934 2935 #ifndef PRODUCT 2936 enum VerifyNoCSetOopsPhase { 2937 VerifyNoCSetOopsStack, 2938 VerifyNoCSetOopsQueues, 2939 VerifyNoCSetOopsSATBCompleted, 2940 VerifyNoCSetOopsSATBThread 2941 }; 2942 2943 class VerifyNoCSetOopsClosure : public OopClosure, public ObjectClosure { 2944 private: 2945 G1CollectedHeap* _g1h; 2946 VerifyNoCSetOopsPhase _phase; 2947 int _info; 2948 2949 const char* phase_str() { 2950 switch (_phase) { 2951 case VerifyNoCSetOopsStack: return "Stack"; 2952 case VerifyNoCSetOopsQueues: return "Queue"; 2953 case VerifyNoCSetOopsSATBCompleted: return "Completed SATB Buffers"; 2954 case VerifyNoCSetOopsSATBThread: return "Thread SATB Buffers"; 2955 default: ShouldNotReachHere(); 2956 } 2957 return NULL; 2958 } 2959 2960 void do_object_work(oop obj) { 2961 guarantee(!_g1h->obj_in_cs(obj), 2962 err_msg("obj: "PTR_FORMAT" in CSet, phase: %s, info: %d", 2963 p2i((void*) obj), phase_str(), _info)); 2964 } 2965 2966 public: 2967 VerifyNoCSetOopsClosure() : _g1h(G1CollectedHeap::heap()) { } 2968 2969 void set_phase(VerifyNoCSetOopsPhase phase, int info = -1) { 2970 _phase = phase; 2971 _info = info; 2972 } 2973 2974 virtual void do_oop(oop* p) { 2975 oop obj = oopDesc::load_decode_heap_oop(p); 2976 do_object_work(obj); 2977 } 2978 2979 virtual void do_oop(narrowOop* p) { 2980 // We should not come across narrow oops while scanning marking 2981 // stacks and SATB buffers. 2982 ShouldNotReachHere(); 2983 } 2984 2985 virtual void do_object(oop obj) { 2986 do_object_work(obj); 2987 } 2988 }; 2989 2990 void ConcurrentMark::verify_no_cset_oops(bool verify_stacks, 2991 bool verify_enqueued_buffers, 2992 bool verify_thread_buffers, 2993 bool verify_fingers) { 2994 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint"); 2995 if (!G1CollectedHeap::heap()->mark_in_progress()) { 2996 return; 2997 } 2998 2999 VerifyNoCSetOopsClosure cl; 3000 3001 if (verify_stacks) { 3002 // Verify entries on the global mark stack 3003 cl.set_phase(VerifyNoCSetOopsStack); 3004 _markStack.oops_do(&cl); 3005 3006 // Verify entries on the task queues 3007 for (uint i = 0; i < _max_worker_id; i += 1) { 3008 cl.set_phase(VerifyNoCSetOopsQueues, i); 3009 CMTaskQueue* queue = _task_queues->queue(i); 3010 queue->oops_do(&cl); 3011 } 3012 } 3013 3014 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); 3015 3016 // Verify entries on the enqueued SATB buffers 3017 if (verify_enqueued_buffers) { 3018 cl.set_phase(VerifyNoCSetOopsSATBCompleted); 3019 satb_qs.iterate_completed_buffers_read_only(&cl); 3020 } 3021 3022 // Verify entries on the per-thread SATB buffers 3023 if (verify_thread_buffers) { 3024 cl.set_phase(VerifyNoCSetOopsSATBThread); 3025 satb_qs.iterate_thread_buffers_read_only(&cl); 3026 } 3027 3028 if (verify_fingers) { 3029 // Verify the global finger 3030 HeapWord* global_finger = finger(); 3031 if (global_finger != NULL && global_finger < _heap_end) { 3032 // The global finger always points to a heap region boundary. We 3033 // use heap_region_containing_raw() to get the containing region 3034 // given that the global finger could be pointing to a free region 3035 // which subsequently becomes continues humongous. If that 3036 // happens, heap_region_containing() will return the bottom of the 3037 // corresponding starts humongous region and the check below will 3038 // not hold any more. 3039 // Since we always iterate over all regions, we might get a NULL HeapRegion 3040 // here. 3041 HeapRegion* global_hr = _g1h->heap_region_containing_raw(global_finger); 3042 guarantee(global_hr == NULL || global_finger == global_hr->bottom(), 3043 err_msg("global finger: "PTR_FORMAT" region: "HR_FORMAT, 3044 p2i(global_finger), HR_FORMAT_PARAMS(global_hr))); 3045 } 3046 3047 // Verify the task fingers 3048 assert(parallel_marking_threads() <= _max_worker_id, "sanity"); 3049 for (int i = 0; i < (int) parallel_marking_threads(); i += 1) { 3050 CMTask* task = _tasks[i]; 3051 HeapWord* task_finger = task->finger(); 3052 if (task_finger != NULL && task_finger < _heap_end) { 3053 // See above note on the global finger verification. 3054 HeapRegion* task_hr = _g1h->heap_region_containing_raw(task_finger); 3055 guarantee(task_hr == NULL || task_finger == task_hr->bottom() || 3056 !task_hr->in_collection_set(), 3057 err_msg("task finger: "PTR_FORMAT" region: "HR_FORMAT, 3058 p2i(task_finger), HR_FORMAT_PARAMS(task_hr))); 3059 } 3060 } 3061 } 3062 } 3063 #endif // PRODUCT 3064 3065 // Aggregate the counting data that was constructed concurrently 3066 // with marking. 3067 class AggregateCountDataHRClosure: public HeapRegionClosure { 3068 G1CollectedHeap* _g1h; 3069 ConcurrentMark* _cm; 3070 CardTableModRefBS* _ct_bs; 3071 BitMap* _cm_card_bm; 3072 uint _max_worker_id; 3073 3074 public: 3075 AggregateCountDataHRClosure(G1CollectedHeap* g1h, 3076 BitMap* cm_card_bm, 3077 uint max_worker_id) : 3078 _g1h(g1h), _cm(g1h->concurrent_mark()), 3079 _ct_bs((CardTableModRefBS*) (g1h->barrier_set())), 3080 _cm_card_bm(cm_card_bm), _max_worker_id(max_worker_id) { } 3081 3082 bool doHeapRegion(HeapRegion* hr) { 3083 if (hr->is_continues_humongous()) { 3084 // We will ignore these here and process them when their 3085 // associated "starts humongous" region is processed. 3086 // Note that we cannot rely on their associated 3087 // "starts humongous" region to have their bit set to 1 3088 // since, due to the region chunking in the parallel region 3089 // iteration, a "continues humongous" region might be visited 3090 // before its associated "starts humongous". 3091 return false; 3092 } 3093 3094 HeapWord* start = hr->bottom(); 3095 HeapWord* limit = hr->next_top_at_mark_start(); 3096 HeapWord* end = hr->end(); 3097 3098 assert(start <= limit && limit <= hr->top() && hr->top() <= hr->end(), 3099 err_msg("Preconditions not met - " 3100 "start: "PTR_FORMAT", limit: "PTR_FORMAT", " 3101 "top: "PTR_FORMAT", end: "PTR_FORMAT, 3102 p2i(start), p2i(limit), p2i(hr->top()), p2i(hr->end()))); 3103 3104 assert(hr->next_marked_bytes() == 0, "Precondition"); 3105 3106 if (start == limit) { 3107 // NTAMS of this region has not been set so nothing to do. 3108 return false; 3109 } 3110 3111 // 'start' should be in the heap. 3112 assert(_g1h->is_in_g1_reserved(start) && _ct_bs->is_card_aligned(start), "sanity"); 3113 // 'end' *may* be just beyond the end of the heap (if hr is the last region) 3114 assert(!_g1h->is_in_g1_reserved(end) || _ct_bs->is_card_aligned(end), "sanity"); 3115 3116 BitMap::idx_t start_idx = _cm->card_bitmap_index_for(start); 3117 BitMap::idx_t limit_idx = _cm->card_bitmap_index_for(limit); 3118 BitMap::idx_t end_idx = _cm->card_bitmap_index_for(end); 3119 3120 // If ntams is not card aligned then we bump card bitmap index 3121 // for limit so that we get the all the cards spanned by 3122 // the object ending at ntams. 3123 // Note: if this is the last region in the heap then ntams 3124 // could be actually just beyond the end of the the heap; 3125 // limit_idx will then correspond to a (non-existent) card 3126 // that is also outside the heap. 3127 if (_g1h->is_in_g1_reserved(limit) && !_ct_bs->is_card_aligned(limit)) { 3128 limit_idx += 1; 3129 } 3130 3131 assert(limit_idx <= end_idx, "or else use atomics"); 3132 3133 // Aggregate the "stripe" in the count data associated with hr. 3134 uint hrm_index = hr->hrm_index(); 3135 size_t marked_bytes = 0; 3136 3137 for (uint i = 0; i < _max_worker_id; i += 1) { 3138 size_t* marked_bytes_array = _cm->count_marked_bytes_array_for(i); 3139 BitMap* task_card_bm = _cm->count_card_bitmap_for(i); 3140 3141 // Fetch the marked_bytes in this region for task i and 3142 // add it to the running total for this region. 3143 marked_bytes += marked_bytes_array[hrm_index]; 3144 3145 // Now union the bitmaps[0,max_worker_id)[start_idx..limit_idx) 3146 // into the global card bitmap. 3147 BitMap::idx_t scan_idx = task_card_bm->get_next_one_offset(start_idx, limit_idx); 3148 3149 while (scan_idx < limit_idx) { 3150 assert(task_card_bm->at(scan_idx) == true, "should be"); 3151 _cm_card_bm->set_bit(scan_idx); 3152 assert(_cm_card_bm->at(scan_idx) == true, "should be"); 3153 3154 // BitMap::get_next_one_offset() can handle the case when 3155 // its left_offset parameter is greater than its right_offset 3156 // parameter. It does, however, have an early exit if 3157 // left_offset == right_offset. So let's limit the value 3158 // passed in for left offset here. 3159 BitMap::idx_t next_idx = MIN2(scan_idx + 1, limit_idx); 3160 scan_idx = task_card_bm->get_next_one_offset(next_idx, limit_idx); 3161 } 3162 } 3163 3164 // Update the marked bytes for this region. 3165 hr->add_to_marked_bytes(marked_bytes); 3166 3167 // Next heap region 3168 return false; 3169 } 3170 }; 3171 3172 class G1AggregateCountDataTask: public AbstractGangTask { 3173 protected: 3174 G1CollectedHeap* _g1h; 3175 ConcurrentMark* _cm; 3176 BitMap* _cm_card_bm; 3177 uint _max_worker_id; 3178 int _active_workers; 3179 HeapRegionClaimer _hrclaimer; 3180 3181 public: 3182 G1AggregateCountDataTask(G1CollectedHeap* g1h, 3183 ConcurrentMark* cm, 3184 BitMap* cm_card_bm, 3185 uint max_worker_id, 3186 int n_workers) : 3187 AbstractGangTask("Count Aggregation"), 3188 _g1h(g1h), _cm(cm), _cm_card_bm(cm_card_bm), 3189 _max_worker_id(max_worker_id), 3190 _active_workers(n_workers), 3191 _hrclaimer(_active_workers) { 3192 } 3193 3194 void work(uint worker_id) { 3195 AggregateCountDataHRClosure cl(_g1h, _cm_card_bm, _max_worker_id); 3196 3197 _g1h->heap_region_par_iterate(&cl, worker_id, &_hrclaimer); 3198 } 3199 }; 3200 3201 3202 void ConcurrentMark::aggregate_count_data() { 3203 int n_workers = _g1h->workers()->active_workers(); 3204 3205 G1AggregateCountDataTask g1_par_agg_task(_g1h, this, &_card_bm, 3206 _max_worker_id, n_workers); 3207 3208 _g1h->set_par_threads(n_workers); 3209 _g1h->workers()->run_task(&g1_par_agg_task); 3210 _g1h->set_par_threads(0); 3211 _g1h->allocation_context_stats().update_at_remark(); 3212 } 3213 3214 // Clear the per-worker arrays used to store the per-region counting data 3215 void ConcurrentMark::clear_all_count_data() { 3216 // Clear the global card bitmap - it will be filled during 3217 // liveness count aggregation (during remark) and the 3218 // final counting task. 3219 _card_bm.clear(); 3220 3221 // Clear the global region bitmap - it will be filled as part 3222 // of the final counting task. 3223 _region_bm.clear(); 3224 3225 uint max_regions = _g1h->max_regions(); 3226 assert(_max_worker_id > 0, "uninitialized"); 3227 3228 for (uint i = 0; i < _max_worker_id; i += 1) { 3229 BitMap* task_card_bm = count_card_bitmap_for(i); 3230 size_t* marked_bytes_array = count_marked_bytes_array_for(i); 3231 3232 assert(task_card_bm->size() == _card_bm.size(), "size mismatch"); 3233 assert(marked_bytes_array != NULL, "uninitialized"); 3234 3235 memset(marked_bytes_array, 0, (size_t) max_regions * sizeof(size_t)); 3236 task_card_bm->clear(); 3237 } 3238 } 3239 3240 void ConcurrentMark::print_stats() { 3241 if (verbose_stats()) { 3242 gclog_or_tty->print_cr("---------------------------------------------------------------------"); 3243 for (size_t i = 0; i < _active_tasks; ++i) { 3244 _tasks[i]->print_stats(); 3245 gclog_or_tty->print_cr("---------------------------------------------------------------------"); 3246 } 3247 } 3248 } 3249 3250 // abandon current marking iteration due to a Full GC 3251 void ConcurrentMark::abort() { 3252 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next 3253 // concurrent bitmap clearing. 3254 _nextMarkBitMap->clearAll(); 3255 3256 // Note we cannot clear the previous marking bitmap here 3257 // since VerifyDuringGC verifies the objects marked during 3258 // a full GC against the previous bitmap. 3259 3260 // Clear the liveness counting data 3261 clear_all_count_data(); 3262 // Empty mark stack 3263 reset_marking_state(); 3264 for (uint i = 0; i < _max_worker_id; ++i) { 3265 _tasks[i]->clear_region_fields(); 3266 } 3267 _first_overflow_barrier_sync.abort(); 3268 _second_overflow_barrier_sync.abort(); 3269 const GCId& gc_id = _g1h->gc_tracer_cm()->gc_id(); 3270 if (!gc_id.is_undefined()) { 3271 // We can do multiple full GCs before ConcurrentMarkThread::run() gets a chance 3272 // to detect that it was aborted. Only keep track of the first GC id that we aborted. 3273 _aborted_gc_id = gc_id; 3274 } 3275 _has_aborted = true; 3276 3277 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3278 satb_mq_set.abandon_partial_marking(); 3279 // This can be called either during or outside marking, we'll read 3280 // the expected_active value from the SATB queue set. 3281 satb_mq_set.set_active_all_threads( 3282 false, /* new active value */ 3283 satb_mq_set.is_active() /* expected_active */); 3284 3285 _g1h->trace_heap_after_concurrent_cycle(); 3286 _g1h->register_concurrent_cycle_end(); 3287 } 3288 3289 const GCId& ConcurrentMark::concurrent_gc_id() { 3290 if (has_aborted()) { 3291 return _aborted_gc_id; 3292 } 3293 return _g1h->gc_tracer_cm()->gc_id(); 3294 } 3295 3296 static void print_ms_time_info(const char* prefix, const char* name, 3297 NumberSeq& ns) { 3298 gclog_or_tty->print_cr("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).", 3299 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg()); 3300 if (ns.num() > 0) { 3301 gclog_or_tty->print_cr("%s [std. dev = %8.2f ms, max = %8.2f ms]", 3302 prefix, ns.sd(), ns.maximum()); 3303 } 3304 } 3305 3306 void ConcurrentMark::print_summary_info() { 3307 gclog_or_tty->print_cr(" Concurrent marking:"); 3308 print_ms_time_info(" ", "init marks", _init_times); 3309 print_ms_time_info(" ", "remarks", _remark_times); 3310 { 3311 print_ms_time_info(" ", "final marks", _remark_mark_times); 3312 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times); 3313 3314 } 3315 print_ms_time_info(" ", "cleanups", _cleanup_times); 3316 gclog_or_tty->print_cr(" Final counting total time = %8.2f s (avg = %8.2f ms).", 3317 _total_counting_time, 3318 (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / 3319 (double)_cleanup_times.num() 3320 : 0.0)); 3321 if (G1ScrubRemSets) { 3322 gclog_or_tty->print_cr(" RS scrub total time = %8.2f s (avg = %8.2f ms).", 3323 _total_rs_scrub_time, 3324 (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / 3325 (double)_cleanup_times.num() 3326 : 0.0)); 3327 } 3328 gclog_or_tty->print_cr(" Total stop_world time = %8.2f s.", 3329 (_init_times.sum() + _remark_times.sum() + 3330 _cleanup_times.sum())/1000.0); 3331 gclog_or_tty->print_cr(" Total concurrent time = %8.2f s " 3332 "(%8.2f s marking).", 3333 cmThread()->vtime_accum(), 3334 cmThread()->vtime_mark_accum()); 3335 } 3336 3337 void ConcurrentMark::print_worker_threads_on(outputStream* st) const { 3338 _parallel_workers->print_worker_threads_on(st); 3339 } 3340 3341 void ConcurrentMark::print_on_error(outputStream* st) const { 3342 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT, 3343 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap)); 3344 _prevMarkBitMap->print_on_error(st, " Prev Bits: "); 3345 _nextMarkBitMap->print_on_error(st, " Next Bits: "); 3346 } 3347 3348 // We take a break if someone is trying to stop the world. 3349 bool ConcurrentMark::do_yield_check(uint worker_id) { 3350 if (SuspendibleThreadSet::should_yield()) { 3351 if (worker_id == 0) { 3352 _g1h->g1_policy()->record_concurrent_pause(); 3353 } 3354 SuspendibleThreadSet::yield(); 3355 return true; 3356 } else { 3357 return false; 3358 } 3359 } 3360 3361 #ifndef PRODUCT 3362 // for debugging purposes 3363 void ConcurrentMark::print_finger() { 3364 gclog_or_tty->print_cr("heap ["PTR_FORMAT", "PTR_FORMAT"), global finger = "PTR_FORMAT, 3365 p2i(_heap_start), p2i(_heap_end), p2i(_finger)); 3366 for (uint i = 0; i < _max_worker_id; ++i) { 3367 gclog_or_tty->print(" %u: " PTR_FORMAT, i, p2i(_tasks[i]->finger())); 3368 } 3369 gclog_or_tty->cr(); 3370 } 3371 #endif 3372 3373 void CMTask::scan_object(oop obj) { 3374 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), "invariant"); 3375 3376 if (_cm->verbose_high()) { 3377 gclog_or_tty->print_cr("[%u] we're scanning object "PTR_FORMAT, 3378 _worker_id, p2i((void*) obj)); 3379 } 3380 3381 size_t obj_size = obj->size(); 3382 _words_scanned += obj_size; 3383 3384 obj->oop_iterate(_cm_oop_closure); 3385 statsOnly( ++_objs_scanned ); 3386 check_limits(); 3387 } 3388 3389 // Closure for iteration over bitmaps 3390 class CMBitMapClosure : public BitMapClosure { 3391 private: 3392 // the bitmap that is being iterated over 3393 CMBitMap* _nextMarkBitMap; 3394 ConcurrentMark* _cm; 3395 CMTask* _task; 3396 3397 public: 3398 CMBitMapClosure(CMTask *task, ConcurrentMark* cm, CMBitMap* nextMarkBitMap) : 3399 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { } 3400 3401 bool do_bit(size_t offset) { 3402 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset); 3403 assert(_nextMarkBitMap->isMarked(addr), "invariant"); 3404 assert( addr < _cm->finger(), "invariant"); 3405 3406 statsOnly( _task->increase_objs_found_on_bitmap() ); 3407 assert(addr >= _task->finger(), "invariant"); 3408 3409 // We move that task's local finger along. 3410 _task->move_finger_to(addr); 3411 3412 _task->scan_object(oop(addr)); 3413 // we only partially drain the local queue and global stack 3414 _task->drain_local_queue(true); 3415 _task->drain_global_stack(true); 3416 3417 // if the has_aborted flag has been raised, we need to bail out of 3418 // the iteration 3419 return !_task->has_aborted(); 3420 } 3421 }; 3422 3423 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h, 3424 ConcurrentMark* cm, 3425 CMTask* task) 3426 : _g1h(g1h), _cm(cm), _task(task) { 3427 assert(_ref_processor == NULL, "should be initialized to NULL"); 3428 3429 if (G1UseConcMarkReferenceProcessing) { 3430 _ref_processor = g1h->ref_processor_cm(); 3431 assert(_ref_processor != NULL, "should not be NULL"); 3432 } 3433 } 3434 3435 void CMTask::setup_for_region(HeapRegion* hr) { 3436 assert(hr != NULL, 3437 "claim_region() should have filtered out NULL regions"); 3438 assert(!hr->is_continues_humongous(), 3439 "claim_region() should have filtered out continues humongous regions"); 3440 3441 if (_cm->verbose_low()) { 3442 gclog_or_tty->print_cr("[%u] setting up for region "PTR_FORMAT, 3443 _worker_id, p2i(hr)); 3444 } 3445 3446 _curr_region = hr; 3447 _finger = hr->bottom(); 3448 update_region_limit(); 3449 } 3450 3451 void CMTask::update_region_limit() { 3452 HeapRegion* hr = _curr_region; 3453 HeapWord* bottom = hr->bottom(); 3454 HeapWord* limit = hr->next_top_at_mark_start(); 3455 3456 if (limit == bottom) { 3457 if (_cm->verbose_low()) { 3458 gclog_or_tty->print_cr("[%u] found an empty region " 3459 "["PTR_FORMAT", "PTR_FORMAT")", 3460 _worker_id, p2i(bottom), p2i(limit)); 3461 } 3462 // The region was collected underneath our feet. 3463 // We set the finger to bottom to ensure that the bitmap 3464 // iteration that will follow this will not do anything. 3465 // (this is not a condition that holds when we set the region up, 3466 // as the region is not supposed to be empty in the first place) 3467 _finger = bottom; 3468 } else if (limit >= _region_limit) { 3469 assert(limit >= _finger, "peace of mind"); 3470 } else { 3471 assert(limit < _region_limit, "only way to get here"); 3472 // This can happen under some pretty unusual circumstances. An 3473 // evacuation pause empties the region underneath our feet (NTAMS 3474 // at bottom). We then do some allocation in the region (NTAMS 3475 // stays at bottom), followed by the region being used as a GC 3476 // alloc region (NTAMS will move to top() and the objects 3477 // originally below it will be grayed). All objects now marked in 3478 // the region are explicitly grayed, if below the global finger, 3479 // and we do not need in fact to scan anything else. So, we simply 3480 // set _finger to be limit to ensure that the bitmap iteration 3481 // doesn't do anything. 3482 _finger = limit; 3483 } 3484 3485 _region_limit = limit; 3486 } 3487 3488 void CMTask::giveup_current_region() { 3489 assert(_curr_region != NULL, "invariant"); 3490 if (_cm->verbose_low()) { 3491 gclog_or_tty->print_cr("[%u] giving up region "PTR_FORMAT, 3492 _worker_id, p2i(_curr_region)); 3493 } 3494 clear_region_fields(); 3495 } 3496 3497 void CMTask::clear_region_fields() { 3498 // Values for these three fields that indicate that we're not 3499 // holding on to a region. 3500 _curr_region = NULL; 3501 _finger = NULL; 3502 _region_limit = NULL; 3503 } 3504 3505 void CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) { 3506 if (cm_oop_closure == NULL) { 3507 assert(_cm_oop_closure != NULL, "invariant"); 3508 } else { 3509 assert(_cm_oop_closure == NULL, "invariant"); 3510 } 3511 _cm_oop_closure = cm_oop_closure; 3512 } 3513 3514 void CMTask::reset(CMBitMap* nextMarkBitMap) { 3515 guarantee(nextMarkBitMap != NULL, "invariant"); 3516 3517 if (_cm->verbose_low()) { 3518 gclog_or_tty->print_cr("[%u] resetting", _worker_id); 3519 } 3520 3521 _nextMarkBitMap = nextMarkBitMap; 3522 clear_region_fields(); 3523 3524 _calls = 0; 3525 _elapsed_time_ms = 0.0; 3526 _termination_time_ms = 0.0; 3527 _termination_start_time_ms = 0.0; 3528 3529 #if _MARKING_STATS_ 3530 _local_pushes = 0; 3531 _local_pops = 0; 3532 _local_max_size = 0; 3533 _objs_scanned = 0; 3534 _global_pushes = 0; 3535 _global_pops = 0; 3536 _global_max_size = 0; 3537 _global_transfers_to = 0; 3538 _global_transfers_from = 0; 3539 _regions_claimed = 0; 3540 _objs_found_on_bitmap = 0; 3541 _satb_buffers_processed = 0; 3542 _steal_attempts = 0; 3543 _steals = 0; 3544 _aborted = 0; 3545 _aborted_overflow = 0; 3546 _aborted_cm_aborted = 0; 3547 _aborted_yield = 0; 3548 _aborted_timed_out = 0; 3549 _aborted_satb = 0; 3550 _aborted_termination = 0; 3551 #endif // _MARKING_STATS_ 3552 } 3553 3554 bool CMTask::should_exit_termination() { 3555 regular_clock_call(); 3556 // This is called when we are in the termination protocol. We should 3557 // quit if, for some reason, this task wants to abort or the global 3558 // stack is not empty (this means that we can get work from it). 3559 return !_cm->mark_stack_empty() || has_aborted(); 3560 } 3561 3562 void CMTask::reached_limit() { 3563 assert(_words_scanned >= _words_scanned_limit || 3564 _refs_reached >= _refs_reached_limit , 3565 "shouldn't have been called otherwise"); 3566 regular_clock_call(); 3567 } 3568 3569 void CMTask::regular_clock_call() { 3570 if (has_aborted()) return; 3571 3572 // First, we need to recalculate the words scanned and refs reached 3573 // limits for the next clock call. 3574 recalculate_limits(); 3575 3576 // During the regular clock call we do the following 3577 3578 // (1) If an overflow has been flagged, then we abort. 3579 if (_cm->has_overflown()) { 3580 set_has_aborted(); 3581 return; 3582 } 3583 3584 // If we are not concurrent (i.e. we're doing remark) we don't need 3585 // to check anything else. The other steps are only needed during 3586 // the concurrent marking phase. 3587 if (!concurrent()) return; 3588 3589 // (2) If marking has been aborted for Full GC, then we also abort. 3590 if (_cm->has_aborted()) { 3591 set_has_aborted(); 3592 statsOnly( ++_aborted_cm_aborted ); 3593 return; 3594 } 3595 3596 double curr_time_ms = os::elapsedVTime() * 1000.0; 3597 3598 // (3) If marking stats are enabled, then we update the step history. 3599 #if _MARKING_STATS_ 3600 if (_words_scanned >= _words_scanned_limit) { 3601 ++_clock_due_to_scanning; 3602 } 3603 if (_refs_reached >= _refs_reached_limit) { 3604 ++_clock_due_to_marking; 3605 } 3606 3607 double last_interval_ms = curr_time_ms - _interval_start_time_ms; 3608 _interval_start_time_ms = curr_time_ms; 3609 _all_clock_intervals_ms.add(last_interval_ms); 3610 3611 if (_cm->verbose_medium()) { 3612 gclog_or_tty->print_cr("[%u] regular clock, interval = %1.2lfms, " 3613 "scanned = "SIZE_FORMAT"%s, refs reached = "SIZE_FORMAT"%s", 3614 _worker_id, last_interval_ms, 3615 _words_scanned, 3616 (_words_scanned >= _words_scanned_limit) ? " (*)" : "", 3617 _refs_reached, 3618 (_refs_reached >= _refs_reached_limit) ? " (*)" : ""); 3619 } 3620 #endif // _MARKING_STATS_ 3621 3622 // (4) We check whether we should yield. If we have to, then we abort. 3623 if (SuspendibleThreadSet::should_yield()) { 3624 // We should yield. To do this we abort the task. The caller is 3625 // responsible for yielding. 3626 set_has_aborted(); 3627 statsOnly( ++_aborted_yield ); 3628 return; 3629 } 3630 3631 // (5) We check whether we've reached our time quota. If we have, 3632 // then we abort. 3633 double elapsed_time_ms = curr_time_ms - _start_time_ms; 3634 if (elapsed_time_ms > _time_target_ms) { 3635 set_has_aborted(); 3636 _has_timed_out = true; 3637 statsOnly( ++_aborted_timed_out ); 3638 return; 3639 } 3640 3641 // (6) Finally, we check whether there are enough completed STAB 3642 // buffers available for processing. If there are, we abort. 3643 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3644 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) { 3645 if (_cm->verbose_low()) { 3646 gclog_or_tty->print_cr("[%u] aborting to deal with pending SATB buffers", 3647 _worker_id); 3648 } 3649 // we do need to process SATB buffers, we'll abort and restart 3650 // the marking task to do so 3651 set_has_aborted(); 3652 statsOnly( ++_aborted_satb ); 3653 return; 3654 } 3655 } 3656 3657 void CMTask::recalculate_limits() { 3658 _real_words_scanned_limit = _words_scanned + words_scanned_period; 3659 _words_scanned_limit = _real_words_scanned_limit; 3660 3661 _real_refs_reached_limit = _refs_reached + refs_reached_period; 3662 _refs_reached_limit = _real_refs_reached_limit; 3663 } 3664 3665 void CMTask::decrease_limits() { 3666 // This is called when we believe that we're going to do an infrequent 3667 // operation which will increase the per byte scanned cost (i.e. move 3668 // entries to/from the global stack). It basically tries to decrease the 3669 // scanning limit so that the clock is called earlier. 3670 3671 if (_cm->verbose_medium()) { 3672 gclog_or_tty->print_cr("[%u] decreasing limits", _worker_id); 3673 } 3674 3675 _words_scanned_limit = _real_words_scanned_limit - 3676 3 * words_scanned_period / 4; 3677 _refs_reached_limit = _real_refs_reached_limit - 3678 3 * refs_reached_period / 4; 3679 } 3680 3681 void CMTask::move_entries_to_global_stack() { 3682 // local array where we'll store the entries that will be popped 3683 // from the local queue 3684 oop buffer[global_stack_transfer_size]; 3685 3686 int n = 0; 3687 oop obj; 3688 while (n < global_stack_transfer_size && _task_queue->pop_local(obj)) { 3689 buffer[n] = obj; 3690 ++n; 3691 } 3692 3693 if (n > 0) { 3694 // we popped at least one entry from the local queue 3695 3696 statsOnly( ++_global_transfers_to; _local_pops += n ); 3697 3698 if (!_cm->mark_stack_push(buffer, n)) { 3699 if (_cm->verbose_low()) { 3700 gclog_or_tty->print_cr("[%u] aborting due to global stack overflow", 3701 _worker_id); 3702 } 3703 set_has_aborted(); 3704 } else { 3705 // the transfer was successful 3706 3707 if (_cm->verbose_medium()) { 3708 gclog_or_tty->print_cr("[%u] pushed %d entries to the global stack", 3709 _worker_id, n); 3710 } 3711 statsOnly( int tmp_size = _cm->mark_stack_size(); 3712 if (tmp_size > _global_max_size) { 3713 _global_max_size = tmp_size; 3714 } 3715 _global_pushes += n ); 3716 } 3717 } 3718 3719 // this operation was quite expensive, so decrease the limits 3720 decrease_limits(); 3721 } 3722 3723 void CMTask::get_entries_from_global_stack() { 3724 // local array where we'll store the entries that will be popped 3725 // from the global stack. 3726 oop buffer[global_stack_transfer_size]; 3727 int n; 3728 _cm->mark_stack_pop(buffer, global_stack_transfer_size, &n); 3729 assert(n <= global_stack_transfer_size, 3730 "we should not pop more than the given limit"); 3731 if (n > 0) { 3732 // yes, we did actually pop at least one entry 3733 3734 statsOnly( ++_global_transfers_from; _global_pops += n ); 3735 if (_cm->verbose_medium()) { 3736 gclog_or_tty->print_cr("[%u] popped %d entries from the global stack", 3737 _worker_id, n); 3738 } 3739 for (int i = 0; i < n; ++i) { 3740 bool success = _task_queue->push(buffer[i]); 3741 // We only call this when the local queue is empty or under a 3742 // given target limit. So, we do not expect this push to fail. 3743 assert(success, "invariant"); 3744 } 3745 3746 statsOnly( int tmp_size = _task_queue->size(); 3747 if (tmp_size > _local_max_size) { 3748 _local_max_size = tmp_size; 3749 } 3750 _local_pushes += n ); 3751 } 3752 3753 // this operation was quite expensive, so decrease the limits 3754 decrease_limits(); 3755 } 3756 3757 void CMTask::drain_local_queue(bool partially) { 3758 if (has_aborted()) return; 3759 3760 // Decide what the target size is, depending whether we're going to 3761 // drain it partially (so that other tasks can steal if they run out 3762 // of things to do) or totally (at the very end). 3763 size_t target_size; 3764 if (partially) { 3765 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize); 3766 } else { 3767 target_size = 0; 3768 } 3769 3770 if (_task_queue->size() > target_size) { 3771 if (_cm->verbose_high()) { 3772 gclog_or_tty->print_cr("[%u] draining local queue, target size = " SIZE_FORMAT, 3773 _worker_id, target_size); 3774 } 3775 3776 oop obj; 3777 bool ret = _task_queue->pop_local(obj); 3778 while (ret) { 3779 statsOnly( ++_local_pops ); 3780 3781 if (_cm->verbose_high()) { 3782 gclog_or_tty->print_cr("[%u] popped "PTR_FORMAT, _worker_id, 3783 p2i((void*) obj)); 3784 } 3785 3786 assert(_g1h->is_in_g1_reserved((HeapWord*) obj), "invariant" ); 3787 assert(!_g1h->is_on_master_free_list( 3788 _g1h->heap_region_containing((HeapWord*) obj)), "invariant"); 3789 3790 scan_object(obj); 3791 3792 if (_task_queue->size() <= target_size || has_aborted()) { 3793 ret = false; 3794 } else { 3795 ret = _task_queue->pop_local(obj); 3796 } 3797 } 3798 3799 if (_cm->verbose_high()) { 3800 gclog_or_tty->print_cr("[%u] drained local queue, size = %u", 3801 _worker_id, _task_queue->size()); 3802 } 3803 } 3804 } 3805 3806 void CMTask::drain_global_stack(bool partially) { 3807 if (has_aborted()) return; 3808 3809 // We have a policy to drain the local queue before we attempt to 3810 // drain the global stack. 3811 assert(partially || _task_queue->size() == 0, "invariant"); 3812 3813 // Decide what the target size is, depending whether we're going to 3814 // drain it partially (so that other tasks can steal if they run out 3815 // of things to do) or totally (at the very end). Notice that, 3816 // because we move entries from the global stack in chunks or 3817 // because another task might be doing the same, we might in fact 3818 // drop below the target. But, this is not a problem. 3819 size_t target_size; 3820 if (partially) { 3821 target_size = _cm->partial_mark_stack_size_target(); 3822 } else { 3823 target_size = 0; 3824 } 3825 3826 if (_cm->mark_stack_size() > target_size) { 3827 if (_cm->verbose_low()) { 3828 gclog_or_tty->print_cr("[%u] draining global_stack, target size " SIZE_FORMAT, 3829 _worker_id, target_size); 3830 } 3831 3832 while (!has_aborted() && _cm->mark_stack_size() > target_size) { 3833 get_entries_from_global_stack(); 3834 drain_local_queue(partially); 3835 } 3836 3837 if (_cm->verbose_low()) { 3838 gclog_or_tty->print_cr("[%u] drained global stack, size = " SIZE_FORMAT, 3839 _worker_id, _cm->mark_stack_size()); 3840 } 3841 } 3842 } 3843 3844 // SATB Queue has several assumptions on whether to call the par or 3845 // non-par versions of the methods. this is why some of the code is 3846 // replicated. We should really get rid of the single-threaded version 3847 // of the code to simplify things. 3848 void CMTask::drain_satb_buffers() { 3849 if (has_aborted()) return; 3850 3851 // We set this so that the regular clock knows that we're in the 3852 // middle of draining buffers and doesn't set the abort flag when it 3853 // notices that SATB buffers are available for draining. It'd be 3854 // very counter productive if it did that. :-) 3855 _draining_satb_buffers = true; 3856 3857 CMObjectClosure oc(this); 3858 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 3859 satb_mq_set.set_closure(_worker_id, &oc); 3860 3861 // This keeps claiming and applying the closure to completed buffers 3862 // until we run out of buffers or we need to abort. 3863 while (!has_aborted() && 3864 satb_mq_set.apply_closure_to_completed_buffer(_worker_id)) { 3865 if (_cm->verbose_medium()) { 3866 gclog_or_tty->print_cr("[%u] processed an SATB buffer", _worker_id); 3867 } 3868 statsOnly( ++_satb_buffers_processed ); 3869 regular_clock_call(); 3870 } 3871 3872 _draining_satb_buffers = false; 3873 3874 assert(has_aborted() || 3875 concurrent() || 3876 satb_mq_set.completed_buffers_num() == 0, "invariant"); 3877 3878 satb_mq_set.set_closure(_worker_id, NULL); 3879 3880 // again, this was a potentially expensive operation, decrease the 3881 // limits to get the regular clock call early 3882 decrease_limits(); 3883 } 3884 3885 void CMTask::print_stats() { 3886 gclog_or_tty->print_cr("Marking Stats, task = %u, calls = %d", 3887 _worker_id, _calls); 3888 gclog_or_tty->print_cr(" Elapsed time = %1.2lfms, Termination time = %1.2lfms", 3889 _elapsed_time_ms, _termination_time_ms); 3890 gclog_or_tty->print_cr(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", 3891 _step_times_ms.num(), _step_times_ms.avg(), 3892 _step_times_ms.sd()); 3893 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", 3894 _step_times_ms.maximum(), _step_times_ms.sum()); 3895 3896 #if _MARKING_STATS_ 3897 gclog_or_tty->print_cr(" Clock Intervals (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", 3898 _all_clock_intervals_ms.num(), _all_clock_intervals_ms.avg(), 3899 _all_clock_intervals_ms.sd()); 3900 gclog_or_tty->print_cr(" max = %1.2lfms, total = %1.2lfms", 3901 _all_clock_intervals_ms.maximum(), 3902 _all_clock_intervals_ms.sum()); 3903 gclog_or_tty->print_cr(" Clock Causes (cum): scanning = %d, marking = %d", 3904 _clock_due_to_scanning, _clock_due_to_marking); 3905 gclog_or_tty->print_cr(" Objects: scanned = %d, found on the bitmap = %d", 3906 _objs_scanned, _objs_found_on_bitmap); 3907 gclog_or_tty->print_cr(" Local Queue: pushes = %d, pops = %d, max size = %d", 3908 _local_pushes, _local_pops, _local_max_size); 3909 gclog_or_tty->print_cr(" Global Stack: pushes = %d, pops = %d, max size = %d", 3910 _global_pushes, _global_pops, _global_max_size); 3911 gclog_or_tty->print_cr(" transfers to = %d, transfers from = %d", 3912 _global_transfers_to,_global_transfers_from); 3913 gclog_or_tty->print_cr(" Regions: claimed = %d", _regions_claimed); 3914 gclog_or_tty->print_cr(" SATB buffers: processed = %d", _satb_buffers_processed); 3915 gclog_or_tty->print_cr(" Steals: attempts = %d, successes = %d", 3916 _steal_attempts, _steals); 3917 gclog_or_tty->print_cr(" Aborted: %d, due to", _aborted); 3918 gclog_or_tty->print_cr(" overflow: %d, global abort: %d, yield: %d", 3919 _aborted_overflow, _aborted_cm_aborted, _aborted_yield); 3920 gclog_or_tty->print_cr(" time out: %d, SATB: %d, termination: %d", 3921 _aborted_timed_out, _aborted_satb, _aborted_termination); 3922 #endif // _MARKING_STATS_ 3923 } 3924 3925 /***************************************************************************** 3926 3927 The do_marking_step(time_target_ms, ...) method is the building 3928 block of the parallel marking framework. It can be called in parallel 3929 with other invocations of do_marking_step() on different tasks 3930 (but only one per task, obviously) and concurrently with the 3931 mutator threads, or during remark, hence it eliminates the need 3932 for two versions of the code. When called during remark, it will 3933 pick up from where the task left off during the concurrent marking 3934 phase. Interestingly, tasks are also claimable during evacuation 3935 pauses too, since do_marking_step() ensures that it aborts before 3936 it needs to yield. 3937 3938 The data structures that it uses to do marking work are the 3939 following: 3940 3941 (1) Marking Bitmap. If there are gray objects that appear only 3942 on the bitmap (this happens either when dealing with an overflow 3943 or when the initial marking phase has simply marked the roots 3944 and didn't push them on the stack), then tasks claim heap 3945 regions whose bitmap they then scan to find gray objects. A 3946 global finger indicates where the end of the last claimed region 3947 is. A local finger indicates how far into the region a task has 3948 scanned. The two fingers are used to determine how to gray an 3949 object (i.e. whether simply marking it is OK, as it will be 3950 visited by a task in the future, or whether it needs to be also 3951 pushed on a stack). 3952 3953 (2) Local Queue. The local queue of the task which is accessed 3954 reasonably efficiently by the task. Other tasks can steal from 3955 it when they run out of work. Throughout the marking phase, a 3956 task attempts to keep its local queue short but not totally 3957 empty, so that entries are available for stealing by other 3958 tasks. Only when there is no more work, a task will totally 3959 drain its local queue. 3960 3961 (3) Global Mark Stack. This handles local queue overflow. During 3962 marking only sets of entries are moved between it and the local 3963 queues, as access to it requires a mutex and more fine-grain 3964 interaction with it which might cause contention. If it 3965 overflows, then the marking phase should restart and iterate 3966 over the bitmap to identify gray objects. Throughout the marking 3967 phase, tasks attempt to keep the global mark stack at a small 3968 length but not totally empty, so that entries are available for 3969 popping by other tasks. Only when there is no more work, tasks 3970 will totally drain the global mark stack. 3971 3972 (4) SATB Buffer Queue. This is where completed SATB buffers are 3973 made available. Buffers are regularly removed from this queue 3974 and scanned for roots, so that the queue doesn't get too 3975 long. During remark, all completed buffers are processed, as 3976 well as the filled in parts of any uncompleted buffers. 3977 3978 The do_marking_step() method tries to abort when the time target 3979 has been reached. There are a few other cases when the 3980 do_marking_step() method also aborts: 3981 3982 (1) When the marking phase has been aborted (after a Full GC). 3983 3984 (2) When a global overflow (on the global stack) has been 3985 triggered. Before the task aborts, it will actually sync up with 3986 the other tasks to ensure that all the marking data structures 3987 (local queues, stacks, fingers etc.) are re-initialized so that 3988 when do_marking_step() completes, the marking phase can 3989 immediately restart. 3990 3991 (3) When enough completed SATB buffers are available. The 3992 do_marking_step() method only tries to drain SATB buffers right 3993 at the beginning. So, if enough buffers are available, the 3994 marking step aborts and the SATB buffers are processed at 3995 the beginning of the next invocation. 3996 3997 (4) To yield. when we have to yield then we abort and yield 3998 right at the end of do_marking_step(). This saves us from a lot 3999 of hassle as, by yielding we might allow a Full GC. If this 4000 happens then objects will be compacted underneath our feet, the 4001 heap might shrink, etc. We save checking for this by just 4002 aborting and doing the yield right at the end. 4003 4004 From the above it follows that the do_marking_step() method should 4005 be called in a loop (or, otherwise, regularly) until it completes. 4006 4007 If a marking step completes without its has_aborted() flag being 4008 true, it means it has completed the current marking phase (and 4009 also all other marking tasks have done so and have all synced up). 4010 4011 A method called regular_clock_call() is invoked "regularly" (in 4012 sub ms intervals) throughout marking. It is this clock method that 4013 checks all the abort conditions which were mentioned above and 4014 decides when the task should abort. A work-based scheme is used to 4015 trigger this clock method: when the number of object words the 4016 marking phase has scanned or the number of references the marking 4017 phase has visited reach a given limit. Additional invocations to 4018 the method clock have been planted in a few other strategic places 4019 too. The initial reason for the clock method was to avoid calling 4020 vtime too regularly, as it is quite expensive. So, once it was in 4021 place, it was natural to piggy-back all the other conditions on it 4022 too and not constantly check them throughout the code. 4023 4024 If do_termination is true then do_marking_step will enter its 4025 termination protocol. 4026 4027 The value of is_serial must be true when do_marking_step is being 4028 called serially (i.e. by the VMThread) and do_marking_step should 4029 skip any synchronization in the termination and overflow code. 4030 Examples include the serial remark code and the serial reference 4031 processing closures. 4032 4033 The value of is_serial must be false when do_marking_step is 4034 being called by any of the worker threads in a work gang. 4035 Examples include the concurrent marking code (CMMarkingTask), 4036 the MT remark code, and the MT reference processing closures. 4037 4038 *****************************************************************************/ 4039 4040 void CMTask::do_marking_step(double time_target_ms, 4041 bool do_termination, 4042 bool is_serial) { 4043 assert(time_target_ms >= 1.0, "minimum granularity is 1ms"); 4044 assert(concurrent() == _cm->concurrent(), "they should be the same"); 4045 4046 G1CollectorPolicy* g1_policy = _g1h->g1_policy(); 4047 assert(_task_queues != NULL, "invariant"); 4048 assert(_task_queue != NULL, "invariant"); 4049 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant"); 4050 4051 assert(!_claimed, 4052 "only one thread should claim this task at any one time"); 4053 4054 // OK, this doesn't safeguard again all possible scenarios, as it is 4055 // possible for two threads to set the _claimed flag at the same 4056 // time. But it is only for debugging purposes anyway and it will 4057 // catch most problems. 4058 _claimed = true; 4059 4060 _start_time_ms = os::elapsedVTime() * 1000.0; 4061 statsOnly( _interval_start_time_ms = _start_time_ms ); 4062 4063 // If do_stealing is true then do_marking_step will attempt to 4064 // steal work from the other CMTasks. It only makes sense to 4065 // enable stealing when the termination protocol is enabled 4066 // and do_marking_step() is not being called serially. 4067 bool do_stealing = do_termination && !is_serial; 4068 4069 double diff_prediction_ms = 4070 g1_policy->get_new_prediction(&_marking_step_diffs_ms); 4071 _time_target_ms = time_target_ms - diff_prediction_ms; 4072 4073 // set up the variables that are used in the work-based scheme to 4074 // call the regular clock method 4075 _words_scanned = 0; 4076 _refs_reached = 0; 4077 recalculate_limits(); 4078 4079 // clear all flags 4080 clear_has_aborted(); 4081 _has_timed_out = false; 4082 _draining_satb_buffers = false; 4083 4084 ++_calls; 4085 4086 if (_cm->verbose_low()) { 4087 gclog_or_tty->print_cr("[%u] >>>>>>>>>> START, call = %d, " 4088 "target = %1.2lfms >>>>>>>>>>", 4089 _worker_id, _calls, _time_target_ms); 4090 } 4091 4092 // Set up the bitmap and oop closures. Anything that uses them is 4093 // eventually called from this method, so it is OK to allocate these 4094 // statically. 4095 CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap); 4096 G1CMOopClosure cm_oop_closure(_g1h, _cm, this); 4097 set_cm_oop_closure(&cm_oop_closure); 4098 4099 if (_cm->has_overflown()) { 4100 // This can happen if the mark stack overflows during a GC pause 4101 // and this task, after a yield point, restarts. We have to abort 4102 // as we need to get into the overflow protocol which happens 4103 // right at the end of this task. 4104 set_has_aborted(); 4105 } 4106 4107 // First drain any available SATB buffers. After this, we will not 4108 // look at SATB buffers before the next invocation of this method. 4109 // If enough completed SATB buffers are queued up, the regular clock 4110 // will abort this task so that it restarts. 4111 drain_satb_buffers(); 4112 // ...then partially drain the local queue and the global stack 4113 drain_local_queue(true); 4114 drain_global_stack(true); 4115 4116 do { 4117 if (!has_aborted() && _curr_region != NULL) { 4118 // This means that we're already holding on to a region. 4119 assert(_finger != NULL, "if region is not NULL, then the finger " 4120 "should not be NULL either"); 4121 4122 // We might have restarted this task after an evacuation pause 4123 // which might have evacuated the region we're holding on to 4124 // underneath our feet. Let's read its limit again to make sure 4125 // that we do not iterate over a region of the heap that 4126 // contains garbage (update_region_limit() will also move 4127 // _finger to the start of the region if it is found empty). 4128 update_region_limit(); 4129 // We will start from _finger not from the start of the region, 4130 // as we might be restarting this task after aborting half-way 4131 // through scanning this region. In this case, _finger points to 4132 // the address where we last found a marked object. If this is a 4133 // fresh region, _finger points to start(). 4134 MemRegion mr = MemRegion(_finger, _region_limit); 4135 4136 if (_cm->verbose_low()) { 4137 gclog_or_tty->print_cr("[%u] we're scanning part " 4138 "["PTR_FORMAT", "PTR_FORMAT") " 4139 "of region "HR_FORMAT, 4140 _worker_id, p2i(_finger), p2i(_region_limit), 4141 HR_FORMAT_PARAMS(_curr_region)); 4142 } 4143 4144 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(), 4145 "humongous regions should go around loop once only"); 4146 4147 // Some special cases: 4148 // If the memory region is empty, we can just give up the region. 4149 // If the current region is humongous then we only need to check 4150 // the bitmap for the bit associated with the start of the object, 4151 // scan the object if it's live, and give up the region. 4152 // Otherwise, let's iterate over the bitmap of the part of the region 4153 // that is left. 4154 // If the iteration is successful, give up the region. 4155 if (mr.is_empty()) { 4156 giveup_current_region(); 4157 regular_clock_call(); 4158 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) { 4159 if (_nextMarkBitMap->isMarked(mr.start())) { 4160 // The object is marked - apply the closure 4161 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start()); 4162 bitmap_closure.do_bit(offset); 4163 } 4164 // Even if this task aborted while scanning the humongous object 4165 // we can (and should) give up the current region. 4166 giveup_current_region(); 4167 regular_clock_call(); 4168 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) { 4169 giveup_current_region(); 4170 regular_clock_call(); 4171 } else { 4172 assert(has_aborted(), "currently the only way to do so"); 4173 // The only way to abort the bitmap iteration is to return 4174 // false from the do_bit() method. However, inside the 4175 // do_bit() method we move the _finger to point to the 4176 // object currently being looked at. So, if we bail out, we 4177 // have definitely set _finger to something non-null. 4178 assert(_finger != NULL, "invariant"); 4179 4180 // Region iteration was actually aborted. So now _finger 4181 // points to the address of the object we last scanned. If we 4182 // leave it there, when we restart this task, we will rescan 4183 // the object. It is easy to avoid this. We move the finger by 4184 // enough to point to the next possible object header (the 4185 // bitmap knows by how much we need to move it as it knows its 4186 // granularity). 4187 assert(_finger < _region_limit, "invariant"); 4188 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger); 4189 // Check if bitmap iteration was aborted while scanning the last object 4190 if (new_finger >= _region_limit) { 4191 giveup_current_region(); 4192 } else { 4193 move_finger_to(new_finger); 4194 } 4195 } 4196 } 4197 // At this point we have either completed iterating over the 4198 // region we were holding on to, or we have aborted. 4199 4200 // We then partially drain the local queue and the global stack. 4201 // (Do we really need this?) 4202 drain_local_queue(true); 4203 drain_global_stack(true); 4204 4205 // Read the note on the claim_region() method on why it might 4206 // return NULL with potentially more regions available for 4207 // claiming and why we have to check out_of_regions() to determine 4208 // whether we're done or not. 4209 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) { 4210 // We are going to try to claim a new region. We should have 4211 // given up on the previous one. 4212 // Separated the asserts so that we know which one fires. 4213 assert(_curr_region == NULL, "invariant"); 4214 assert(_finger == NULL, "invariant"); 4215 assert(_region_limit == NULL, "invariant"); 4216 if (_cm->verbose_low()) { 4217 gclog_or_tty->print_cr("[%u] trying to claim a new region", _worker_id); 4218 } 4219 HeapRegion* claimed_region = _cm->claim_region(_worker_id); 4220 if (claimed_region != NULL) { 4221 // Yes, we managed to claim one 4222 statsOnly( ++_regions_claimed ); 4223 4224 if (_cm->verbose_low()) { 4225 gclog_or_tty->print_cr("[%u] we successfully claimed " 4226 "region "PTR_FORMAT, 4227 _worker_id, p2i(claimed_region)); 4228 } 4229 4230 setup_for_region(claimed_region); 4231 assert(_curr_region == claimed_region, "invariant"); 4232 } 4233 // It is important to call the regular clock here. It might take 4234 // a while to claim a region if, for example, we hit a large 4235 // block of empty regions. So we need to call the regular clock 4236 // method once round the loop to make sure it's called 4237 // frequently enough. 4238 regular_clock_call(); 4239 } 4240 4241 if (!has_aborted() && _curr_region == NULL) { 4242 assert(_cm->out_of_regions(), 4243 "at this point we should be out of regions"); 4244 } 4245 } while ( _curr_region != NULL && !has_aborted()); 4246 4247 if (!has_aborted()) { 4248 // We cannot check whether the global stack is empty, since other 4249 // tasks might be pushing objects to it concurrently. 4250 assert(_cm->out_of_regions(), 4251 "at this point we should be out of regions"); 4252 4253 if (_cm->verbose_low()) { 4254 gclog_or_tty->print_cr("[%u] all regions claimed", _worker_id); 4255 } 4256 4257 // Try to reduce the number of available SATB buffers so that 4258 // remark has less work to do. 4259 drain_satb_buffers(); 4260 } 4261 4262 // Since we've done everything else, we can now totally drain the 4263 // local queue and global stack. 4264 drain_local_queue(false); 4265 drain_global_stack(false); 4266 4267 // Attempt at work stealing from other task's queues. 4268 if (do_stealing && !has_aborted()) { 4269 // We have not aborted. This means that we have finished all that 4270 // we could. Let's try to do some stealing... 4271 4272 // We cannot check whether the global stack is empty, since other 4273 // tasks might be pushing objects to it concurrently. 4274 assert(_cm->out_of_regions() && _task_queue->size() == 0, 4275 "only way to reach here"); 4276 4277 if (_cm->verbose_low()) { 4278 gclog_or_tty->print_cr("[%u] starting to steal", _worker_id); 4279 } 4280 4281 while (!has_aborted()) { 4282 oop obj; 4283 statsOnly( ++_steal_attempts ); 4284 4285 if (_cm->try_stealing(_worker_id, &_hash_seed, obj)) { 4286 if (_cm->verbose_medium()) { 4287 gclog_or_tty->print_cr("[%u] stolen "PTR_FORMAT" successfully", 4288 _worker_id, p2i((void*) obj)); 4289 } 4290 4291 statsOnly( ++_steals ); 4292 4293 assert(_nextMarkBitMap->isMarked((HeapWord*) obj), 4294 "any stolen object should be marked"); 4295 scan_object(obj); 4296 4297 // And since we're towards the end, let's totally drain the 4298 // local queue and global stack. 4299 drain_local_queue(false); 4300 drain_global_stack(false); 4301 } else { 4302 break; 4303 } 4304 } 4305 } 4306 4307 // If we are about to wrap up and go into termination, check if we 4308 // should raise the overflow flag. 4309 if (do_termination && !has_aborted()) { 4310 if (_cm->force_overflow()->should_force()) { 4311 _cm->set_has_overflown(); 4312 regular_clock_call(); 4313 } 4314 } 4315 4316 // We still haven't aborted. Now, let's try to get into the 4317 // termination protocol. 4318 if (do_termination && !has_aborted()) { 4319 // We cannot check whether the global stack is empty, since other 4320 // tasks might be concurrently pushing objects on it. 4321 // Separated the asserts so that we know which one fires. 4322 assert(_cm->out_of_regions(), "only way to reach here"); 4323 assert(_task_queue->size() == 0, "only way to reach here"); 4324 4325 if (_cm->verbose_low()) { 4326 gclog_or_tty->print_cr("[%u] starting termination protocol", _worker_id); 4327 } 4328 4329 _termination_start_time_ms = os::elapsedVTime() * 1000.0; 4330 4331 // The CMTask class also extends the TerminatorTerminator class, 4332 // hence its should_exit_termination() method will also decide 4333 // whether to exit the termination protocol or not. 4334 bool finished = (is_serial || 4335 _cm->terminator()->offer_termination(this)); 4336 double termination_end_time_ms = os::elapsedVTime() * 1000.0; 4337 _termination_time_ms += 4338 termination_end_time_ms - _termination_start_time_ms; 4339 4340 if (finished) { 4341 // We're all done. 4342 4343 if (_worker_id == 0) { 4344 // let's allow task 0 to do this 4345 if (concurrent()) { 4346 assert(_cm->concurrent_marking_in_progress(), "invariant"); 4347 // we need to set this to false before the next 4348 // safepoint. This way we ensure that the marking phase 4349 // doesn't observe any more heap expansions. 4350 _cm->clear_concurrent_marking_in_progress(); 4351 } 4352 } 4353 4354 // We can now guarantee that the global stack is empty, since 4355 // all other tasks have finished. We separated the guarantees so 4356 // that, if a condition is false, we can immediately find out 4357 // which one. 4358 guarantee(_cm->out_of_regions(), "only way to reach here"); 4359 guarantee(_cm->mark_stack_empty(), "only way to reach here"); 4360 guarantee(_task_queue->size() == 0, "only way to reach here"); 4361 guarantee(!_cm->has_overflown(), "only way to reach here"); 4362 guarantee(!_cm->mark_stack_overflow(), "only way to reach here"); 4363 4364 if (_cm->verbose_low()) { 4365 gclog_or_tty->print_cr("[%u] all tasks terminated", _worker_id); 4366 } 4367 } else { 4368 // Apparently there's more work to do. Let's abort this task. It 4369 // will restart it and we can hopefully find more things to do. 4370 4371 if (_cm->verbose_low()) { 4372 gclog_or_tty->print_cr("[%u] apparently there is more work to do", 4373 _worker_id); 4374 } 4375 4376 set_has_aborted(); 4377 statsOnly( ++_aborted_termination ); 4378 } 4379 } 4380 4381 // Mainly for debugging purposes to make sure that a pointer to the 4382 // closure which was statically allocated in this frame doesn't 4383 // escape it by accident. 4384 set_cm_oop_closure(NULL); 4385 double end_time_ms = os::elapsedVTime() * 1000.0; 4386 double elapsed_time_ms = end_time_ms - _start_time_ms; 4387 // Update the step history. 4388 _step_times_ms.add(elapsed_time_ms); 4389 4390 if (has_aborted()) { 4391 // The task was aborted for some reason. 4392 4393 statsOnly( ++_aborted ); 4394 4395 if (_has_timed_out) { 4396 double diff_ms = elapsed_time_ms - _time_target_ms; 4397 // Keep statistics of how well we did with respect to hitting 4398 // our target only if we actually timed out (if we aborted for 4399 // other reasons, then the results might get skewed). 4400 _marking_step_diffs_ms.add(diff_ms); 4401 } 4402 4403 if (_cm->has_overflown()) { 4404 // This is the interesting one. We aborted because a global 4405 // overflow was raised. This means we have to restart the 4406 // marking phase and start iterating over regions. However, in 4407 // order to do this we have to make sure that all tasks stop 4408 // what they are doing and re-initialize in a safe manner. We 4409 // will achieve this with the use of two barrier sync points. 4410 4411 if (_cm->verbose_low()) { 4412 gclog_or_tty->print_cr("[%u] detected overflow", _worker_id); 4413 } 4414 4415 if (!is_serial) { 4416 // We only need to enter the sync barrier if being called 4417 // from a parallel context 4418 _cm->enter_first_sync_barrier(_worker_id); 4419 4420 // When we exit this sync barrier we know that all tasks have 4421 // stopped doing marking work. So, it's now safe to 4422 // re-initialize our data structures. At the end of this method, 4423 // task 0 will clear the global data structures. 4424 } 4425 4426 statsOnly( ++_aborted_overflow ); 4427 4428 // We clear the local state of this task... 4429 clear_region_fields(); 4430 4431 if (!is_serial) { 4432 // ...and enter the second barrier. 4433 _cm->enter_second_sync_barrier(_worker_id); 4434 } 4435 // At this point, if we're during the concurrent phase of 4436 // marking, everything has been re-initialized and we're 4437 // ready to restart. 4438 } 4439 4440 if (_cm->verbose_low()) { 4441 gclog_or_tty->print_cr("[%u] <<<<<<<<<< ABORTING, target = %1.2lfms, " 4442 "elapsed = %1.2lfms <<<<<<<<<<", 4443 _worker_id, _time_target_ms, elapsed_time_ms); 4444 if (_cm->has_aborted()) { 4445 gclog_or_tty->print_cr("[%u] ========== MARKING ABORTED ==========", 4446 _worker_id); 4447 } 4448 } 4449 } else { 4450 if (_cm->verbose_low()) { 4451 gclog_or_tty->print_cr("[%u] <<<<<<<<<< FINISHED, target = %1.2lfms, " 4452 "elapsed = %1.2lfms <<<<<<<<<<", 4453 _worker_id, _time_target_ms, elapsed_time_ms); 4454 } 4455 } 4456 4457 _claimed = false; 4458 } 4459 4460 CMTask::CMTask(uint worker_id, 4461 ConcurrentMark* cm, 4462 size_t* marked_bytes, 4463 BitMap* card_bm, 4464 CMTaskQueue* task_queue, 4465 CMTaskQueueSet* task_queues) 4466 : _g1h(G1CollectedHeap::heap()), 4467 _worker_id(worker_id), _cm(cm), 4468 _claimed(false), 4469 _nextMarkBitMap(NULL), _hash_seed(17), 4470 _task_queue(task_queue), 4471 _task_queues(task_queues), 4472 _cm_oop_closure(NULL), 4473 _marked_bytes_array(marked_bytes), 4474 _card_bm(card_bm) { 4475 guarantee(task_queue != NULL, "invariant"); 4476 guarantee(task_queues != NULL, "invariant"); 4477 4478 statsOnly( _clock_due_to_scanning = 0; 4479 _clock_due_to_marking = 0 ); 4480 4481 _marking_step_diffs_ms.add(0.5); 4482 } 4483 4484 // These are formatting macros that are used below to ensure 4485 // consistent formatting. The *_H_* versions are used to format the 4486 // header for a particular value and they should be kept consistent 4487 // with the corresponding macro. Also note that most of the macros add 4488 // the necessary white space (as a prefix) which makes them a bit 4489 // easier to compose. 4490 4491 // All the output lines are prefixed with this string to be able to 4492 // identify them easily in a large log file. 4493 #define G1PPRL_LINE_PREFIX "###" 4494 4495 #define G1PPRL_ADDR_BASE_FORMAT " "PTR_FORMAT"-"PTR_FORMAT 4496 #ifdef _LP64 4497 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s" 4498 #else // _LP64 4499 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s" 4500 #endif // _LP64 4501 4502 // For per-region info 4503 #define G1PPRL_TYPE_FORMAT " %-4s" 4504 #define G1PPRL_TYPE_H_FORMAT " %4s" 4505 #define G1PPRL_BYTE_FORMAT " "SIZE_FORMAT_W(9) 4506 #define G1PPRL_BYTE_H_FORMAT " %9s" 4507 #define G1PPRL_DOUBLE_FORMAT " %14.1f" 4508 #define G1PPRL_DOUBLE_H_FORMAT " %14s" 4509 4510 // For summary info 4511 #define G1PPRL_SUM_ADDR_FORMAT(tag) " "tag":"G1PPRL_ADDR_BASE_FORMAT 4512 #define G1PPRL_SUM_BYTE_FORMAT(tag) " "tag": "SIZE_FORMAT 4513 #define G1PPRL_SUM_MB_FORMAT(tag) " "tag": %1.2f MB" 4514 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag)" / %1.2f %%" 4515 4516 G1PrintRegionLivenessInfoClosure:: 4517 G1PrintRegionLivenessInfoClosure(outputStream* out, const char* phase_name) 4518 : _out(out), 4519 _total_used_bytes(0), _total_capacity_bytes(0), 4520 _total_prev_live_bytes(0), _total_next_live_bytes(0), 4521 _hum_used_bytes(0), _hum_capacity_bytes(0), 4522 _hum_prev_live_bytes(0), _hum_next_live_bytes(0), 4523 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) { 4524 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 4525 MemRegion g1_reserved = g1h->g1_reserved(); 4526 double now = os::elapsedTime(); 4527 4528 // Print the header of the output. 4529 _out->cr(); 4530 _out->print_cr(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now); 4531 _out->print_cr(G1PPRL_LINE_PREFIX" HEAP" 4532 G1PPRL_SUM_ADDR_FORMAT("reserved") 4533 G1PPRL_SUM_BYTE_FORMAT("region-size"), 4534 p2i(g1_reserved.start()), p2i(g1_reserved.end()), 4535 HeapRegion::GrainBytes); 4536 _out->print_cr(G1PPRL_LINE_PREFIX); 4537 _out->print_cr(G1PPRL_LINE_PREFIX 4538 G1PPRL_TYPE_H_FORMAT 4539 G1PPRL_ADDR_BASE_H_FORMAT 4540 G1PPRL_BYTE_H_FORMAT 4541 G1PPRL_BYTE_H_FORMAT 4542 G1PPRL_BYTE_H_FORMAT 4543 G1PPRL_DOUBLE_H_FORMAT 4544 G1PPRL_BYTE_H_FORMAT 4545 G1PPRL_BYTE_H_FORMAT, 4546 "type", "address-range", 4547 "used", "prev-live", "next-live", "gc-eff", 4548 "remset", "code-roots"); 4549 _out->print_cr(G1PPRL_LINE_PREFIX 4550 G1PPRL_TYPE_H_FORMAT 4551 G1PPRL_ADDR_BASE_H_FORMAT 4552 G1PPRL_BYTE_H_FORMAT 4553 G1PPRL_BYTE_H_FORMAT 4554 G1PPRL_BYTE_H_FORMAT 4555 G1PPRL_DOUBLE_H_FORMAT 4556 G1PPRL_BYTE_H_FORMAT 4557 G1PPRL_BYTE_H_FORMAT, 4558 "", "", 4559 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)", 4560 "(bytes)", "(bytes)"); 4561 } 4562 4563 // It takes as a parameter a reference to one of the _hum_* fields, it 4564 // deduces the corresponding value for a region in a humongous region 4565 // series (either the region size, or what's left if the _hum_* field 4566 // is < the region size), and updates the _hum_* field accordingly. 4567 size_t G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* hum_bytes) { 4568 size_t bytes = 0; 4569 // The > 0 check is to deal with the prev and next live bytes which 4570 // could be 0. 4571 if (*hum_bytes > 0) { 4572 bytes = MIN2(HeapRegion::GrainBytes, *hum_bytes); 4573 *hum_bytes -= bytes; 4574 } 4575 return bytes; 4576 } 4577 4578 // It deduces the values for a region in a humongous region series 4579 // from the _hum_* fields and updates those accordingly. It assumes 4580 // that that _hum_* fields have already been set up from the "starts 4581 // humongous" region and we visit the regions in address order. 4582 void G1PrintRegionLivenessInfoClosure::get_hum_bytes(size_t* used_bytes, 4583 size_t* capacity_bytes, 4584 size_t* prev_live_bytes, 4585 size_t* next_live_bytes) { 4586 assert(_hum_used_bytes > 0 && _hum_capacity_bytes > 0, "pre-condition"); 4587 *used_bytes = get_hum_bytes(&_hum_used_bytes); 4588 *capacity_bytes = get_hum_bytes(&_hum_capacity_bytes); 4589 *prev_live_bytes = get_hum_bytes(&_hum_prev_live_bytes); 4590 *next_live_bytes = get_hum_bytes(&_hum_next_live_bytes); 4591 } 4592 4593 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) { 4594 const char* type = r->get_type_str(); 4595 HeapWord* bottom = r->bottom(); 4596 HeapWord* end = r->end(); 4597 size_t capacity_bytes = r->capacity(); 4598 size_t used_bytes = r->used(); 4599 size_t prev_live_bytes = r->live_bytes(); 4600 size_t next_live_bytes = r->next_live_bytes(); 4601 double gc_eff = r->gc_efficiency(); 4602 size_t remset_bytes = r->rem_set()->mem_size(); 4603 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size(); 4604 4605 if (r->is_starts_humongous()) { 4606 assert(_hum_used_bytes == 0 && _hum_capacity_bytes == 0 && 4607 _hum_prev_live_bytes == 0 && _hum_next_live_bytes == 0, 4608 "they should have been zeroed after the last time we used them"); 4609 // Set up the _hum_* fields. 4610 _hum_capacity_bytes = capacity_bytes; 4611 _hum_used_bytes = used_bytes; 4612 _hum_prev_live_bytes = prev_live_bytes; 4613 _hum_next_live_bytes = next_live_bytes; 4614 get_hum_bytes(&used_bytes, &capacity_bytes, 4615 &prev_live_bytes, &next_live_bytes); 4616 end = bottom + HeapRegion::GrainWords; 4617 } else if (r->is_continues_humongous()) { 4618 get_hum_bytes(&used_bytes, &capacity_bytes, 4619 &prev_live_bytes, &next_live_bytes); 4620 assert(end == bottom + HeapRegion::GrainWords, "invariant"); 4621 } 4622 4623 _total_used_bytes += used_bytes; 4624 _total_capacity_bytes += capacity_bytes; 4625 _total_prev_live_bytes += prev_live_bytes; 4626 _total_next_live_bytes += next_live_bytes; 4627 _total_remset_bytes += remset_bytes; 4628 _total_strong_code_roots_bytes += strong_code_roots_bytes; 4629 4630 // Print a line for this particular region. 4631 _out->print_cr(G1PPRL_LINE_PREFIX 4632 G1PPRL_TYPE_FORMAT 4633 G1PPRL_ADDR_BASE_FORMAT 4634 G1PPRL_BYTE_FORMAT 4635 G1PPRL_BYTE_FORMAT 4636 G1PPRL_BYTE_FORMAT 4637 G1PPRL_DOUBLE_FORMAT 4638 G1PPRL_BYTE_FORMAT 4639 G1PPRL_BYTE_FORMAT, 4640 type, p2i(bottom), p2i(end), 4641 used_bytes, prev_live_bytes, next_live_bytes, gc_eff, 4642 remset_bytes, strong_code_roots_bytes); 4643 4644 return false; 4645 } 4646 4647 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() { 4648 // add static memory usages to remembered set sizes 4649 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size(); 4650 // Print the footer of the output. 4651 _out->print_cr(G1PPRL_LINE_PREFIX); 4652 _out->print_cr(G1PPRL_LINE_PREFIX 4653 " SUMMARY" 4654 G1PPRL_SUM_MB_FORMAT("capacity") 4655 G1PPRL_SUM_MB_PERC_FORMAT("used") 4656 G1PPRL_SUM_MB_PERC_FORMAT("prev-live") 4657 G1PPRL_SUM_MB_PERC_FORMAT("next-live") 4658 G1PPRL_SUM_MB_FORMAT("remset") 4659 G1PPRL_SUM_MB_FORMAT("code-roots"), 4660 bytes_to_mb(_total_capacity_bytes), 4661 bytes_to_mb(_total_used_bytes), 4662 perc(_total_used_bytes, _total_capacity_bytes), 4663 bytes_to_mb(_total_prev_live_bytes), 4664 perc(_total_prev_live_bytes, _total_capacity_bytes), 4665 bytes_to_mb(_total_next_live_bytes), 4666 perc(_total_next_live_bytes, _total_capacity_bytes), 4667 bytes_to_mb(_total_remset_bytes), 4668 bytes_to_mb(_total_strong_code_roots_bytes)); 4669 _out->cr(); 4670 }