1 /* 2 * Copyright (c) 2001, 2017, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25 #include "precompiled.hpp" 26 #include "classfile/metadataOnStackMark.hpp" 27 #include "classfile/symbolTable.hpp" 28 #include "code/codeCache.hpp" 29 #include "gc/g1/concurrentMarkThread.inline.hpp" 30 #include "gc/g1/g1CollectedHeap.inline.hpp" 31 #include "gc/g1/g1CollectorState.hpp" 32 #include "gc/g1/g1ConcurrentMark.inline.hpp" 33 #include "gc/g1/g1HeapVerifier.hpp" 34 #include "gc/g1/g1OopClosures.inline.hpp" 35 #include "gc/g1/g1CardLiveData.inline.hpp" 36 #include "gc/g1/g1Policy.hpp" 37 #include "gc/g1/g1StringDedup.hpp" 38 #include "gc/g1/heapRegion.inline.hpp" 39 #include "gc/g1/heapRegionRemSet.hpp" 40 #include "gc/g1/heapRegionSet.inline.hpp" 41 #include "gc/g1/suspendibleThreadSet.hpp" 42 #include "gc/shared/gcId.hpp" 43 #include "gc/shared/gcTimer.hpp" 44 #include "gc/shared/gcTrace.hpp" 45 #include "gc/shared/gcTraceTime.inline.hpp" 46 #include "gc/shared/genOopClosures.inline.hpp" 47 #include "gc/shared/referencePolicy.hpp" 48 #include "gc/shared/strongRootsScope.hpp" 49 #include "gc/shared/taskqueue.inline.hpp" 50 #include "gc/shared/vmGCOperations.hpp" 51 #include "logging/log.hpp" 52 #include "memory/allocation.hpp" 53 #include "memory/resourceArea.hpp" 54 #include "oops/oop.inline.hpp" 55 #include "runtime/atomic.hpp" 56 #include "runtime/handles.inline.hpp" 57 #include "runtime/java.hpp" 58 #include "runtime/prefetch.inline.hpp" 59 #include "services/memTracker.hpp" 60 #include "utilities/align.hpp" 61 #include "utilities/growableArray.hpp" 62 63 // Concurrent marking bit map wrapper 64 65 G1CMBitMapRO::G1CMBitMapRO(int shifter) : 66 _bm(), 67 _shifter(shifter) { 68 _bmStartWord = 0; 69 _bmWordSize = 0; 70 } 71 72 HeapWord* G1CMBitMapRO::getNextMarkedWordAddress(const HeapWord* addr, 73 const HeapWord* limit) const { 74 // First we must round addr *up* to a possible object boundary. 75 addr = align_up(addr, HeapWordSize << _shifter); 76 size_t addrOffset = heapWordToOffset(addr); 77 assert(limit != NULL, "limit must not be NULL"); 78 size_t limitOffset = heapWordToOffset(limit); 79 size_t nextOffset = _bm.get_next_one_offset(addrOffset, limitOffset); 80 HeapWord* nextAddr = offsetToHeapWord(nextOffset); 81 assert(nextAddr >= addr, "get_next_one postcondition"); 82 assert(nextAddr == limit || isMarked(nextAddr), 83 "get_next_one postcondition"); 84 return nextAddr; 85 } 86 87 #ifndef PRODUCT 88 bool G1CMBitMapRO::covers(MemRegion heap_rs) const { 89 // assert(_bm.map() == _virtual_space.low(), "map inconsistency"); 90 assert(((size_t)_bm.size() * ((size_t)1 << _shifter)) == _bmWordSize, 91 "size inconsistency"); 92 return _bmStartWord == (HeapWord*)(heap_rs.start()) && 93 _bmWordSize == heap_rs.word_size(); 94 } 95 #endif 96 97 void G1CMBitMapRO::print_on_error(outputStream* st, const char* prefix) const { 98 _bm.print_on_error(st, prefix); 99 } 100 101 size_t G1CMBitMap::compute_size(size_t heap_size) { 102 return ReservedSpace::allocation_align_size_up(heap_size / mark_distance()); 103 } 104 105 size_t G1CMBitMap::mark_distance() { 106 return MinObjAlignmentInBytes * BitsPerByte; 107 } 108 109 void G1CMBitMap::initialize(MemRegion heap, G1RegionToSpaceMapper* storage) { 110 _bmStartWord = heap.start(); 111 _bmWordSize = heap.word_size(); 112 113 _bm = BitMapView((BitMap::bm_word_t*) storage->reserved().start(), _bmWordSize >> _shifter); 114 115 storage->set_mapping_changed_listener(&_listener); 116 } 117 118 void G1CMBitMapMappingChangedListener::on_commit(uint start_region, size_t num_regions, bool zero_filled) { 119 if (zero_filled) { 120 return; 121 } 122 // We need to clear the bitmap on commit, removing any existing information. 123 MemRegion mr(G1CollectedHeap::heap()->bottom_addr_for_region(start_region), num_regions * HeapRegion::GrainWords); 124 _bm->clear_range(mr); 125 } 126 127 void G1CMBitMap::clear_range(MemRegion mr) { 128 mr.intersection(MemRegion(_bmStartWord, _bmWordSize)); 129 assert(!mr.is_empty(), "unexpected empty region"); 130 // convert address range into offset range 131 _bm.at_put_range(heapWordToOffset(mr.start()), 132 heapWordToOffset(mr.end()), false); 133 } 134 135 G1CMMarkStack::G1CMMarkStack() : 136 _max_chunk_capacity(0), 137 _base(NULL), 138 _chunk_capacity(0) { 139 set_empty(); 140 } 141 142 bool G1CMMarkStack::resize(size_t new_capacity) { 143 assert(is_empty(), "Only resize when stack is empty."); 144 assert(new_capacity <= _max_chunk_capacity, 145 "Trying to resize stack to " SIZE_FORMAT " chunks when the maximum is " SIZE_FORMAT, new_capacity, _max_chunk_capacity); 146 147 TaskQueueEntryChunk* new_base = MmapArrayAllocator<TaskQueueEntryChunk>::allocate_or_null(new_capacity, mtGC); 148 149 if (new_base == NULL) { 150 log_warning(gc)("Failed to reserve memory for new overflow mark stack with " SIZE_FORMAT " chunks and size " SIZE_FORMAT "B.", new_capacity, new_capacity * sizeof(TaskQueueEntryChunk)); 151 return false; 152 } 153 // Release old mapping. 154 if (_base != NULL) { 155 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity); 156 } 157 158 _base = new_base; 159 _chunk_capacity = new_capacity; 160 set_empty(); 161 162 return true; 163 } 164 165 size_t G1CMMarkStack::capacity_alignment() { 166 return (size_t)lcm(os::vm_allocation_granularity(), sizeof(TaskQueueEntryChunk)) / sizeof(G1TaskQueueEntry); 167 } 168 169 bool G1CMMarkStack::initialize(size_t initial_capacity, size_t max_capacity) { 170 guarantee(_max_chunk_capacity == 0, "G1CMMarkStack already initialized."); 171 172 size_t const TaskEntryChunkSizeInVoidStar = sizeof(TaskQueueEntryChunk) / sizeof(G1TaskQueueEntry); 173 174 _max_chunk_capacity = align_up(max_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar; 175 size_t initial_chunk_capacity = align_up(initial_capacity, capacity_alignment()) / TaskEntryChunkSizeInVoidStar; 176 177 guarantee(initial_chunk_capacity <= _max_chunk_capacity, 178 "Maximum chunk capacity " SIZE_FORMAT " smaller than initial capacity " SIZE_FORMAT, 179 _max_chunk_capacity, 180 initial_chunk_capacity); 181 182 log_debug(gc)("Initialize mark stack with " SIZE_FORMAT " chunks, maximum " SIZE_FORMAT, 183 initial_chunk_capacity, _max_chunk_capacity); 184 185 return resize(initial_chunk_capacity); 186 } 187 188 void G1CMMarkStack::expand() { 189 if (_chunk_capacity == _max_chunk_capacity) { 190 log_debug(gc)("Can not expand overflow mark stack further, already at maximum capacity of " SIZE_FORMAT " chunks.", _chunk_capacity); 191 return; 192 } 193 size_t old_capacity = _chunk_capacity; 194 // Double capacity if possible 195 size_t new_capacity = MIN2(old_capacity * 2, _max_chunk_capacity); 196 197 if (resize(new_capacity)) { 198 log_debug(gc)("Expanded mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks", 199 old_capacity, new_capacity); 200 } else { 201 log_warning(gc)("Failed to expand mark stack capacity from " SIZE_FORMAT " to " SIZE_FORMAT " chunks", 202 old_capacity, new_capacity); 203 } 204 } 205 206 G1CMMarkStack::~G1CMMarkStack() { 207 if (_base != NULL) { 208 MmapArrayAllocator<TaskQueueEntryChunk>::free(_base, _chunk_capacity); 209 } 210 } 211 212 void G1CMMarkStack::add_chunk_to_list(TaskQueueEntryChunk* volatile* list, TaskQueueEntryChunk* elem) { 213 elem->next = *list; 214 *list = elem; 215 } 216 217 void G1CMMarkStack::add_chunk_to_chunk_list(TaskQueueEntryChunk* elem) { 218 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag); 219 add_chunk_to_list(&_chunk_list, elem); 220 _chunks_in_chunk_list++; 221 } 222 223 void G1CMMarkStack::add_chunk_to_free_list(TaskQueueEntryChunk* elem) { 224 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag); 225 add_chunk_to_list(&_free_list, elem); 226 } 227 228 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_list(TaskQueueEntryChunk* volatile* list) { 229 TaskQueueEntryChunk* result = *list; 230 if (result != NULL) { 231 *list = (*list)->next; 232 } 233 return result; 234 } 235 236 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_chunk_list() { 237 MutexLockerEx x(MarkStackChunkList_lock, Mutex::_no_safepoint_check_flag); 238 TaskQueueEntryChunk* result = remove_chunk_from_list(&_chunk_list); 239 if (result != NULL) { 240 _chunks_in_chunk_list--; 241 } 242 return result; 243 } 244 245 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::remove_chunk_from_free_list() { 246 MutexLockerEx x(MarkStackFreeList_lock, Mutex::_no_safepoint_check_flag); 247 return remove_chunk_from_list(&_free_list); 248 } 249 250 G1CMMarkStack::TaskQueueEntryChunk* G1CMMarkStack::allocate_new_chunk() { 251 // This dirty read of _hwm is okay because we only ever increase the _hwm in parallel code. 252 // Further this limits _hwm to a value of _chunk_capacity + #threads, avoiding 253 // wraparound of _hwm. 254 if (_hwm >= _chunk_capacity) { 255 return NULL; 256 } 257 258 size_t cur_idx = Atomic::add(1, &_hwm) - 1; 259 if (cur_idx >= _chunk_capacity) { 260 return NULL; 261 } 262 263 TaskQueueEntryChunk* result = ::new (&_base[cur_idx]) TaskQueueEntryChunk; 264 result->next = NULL; 265 return result; 266 } 267 268 bool G1CMMarkStack::par_push_chunk(G1TaskQueueEntry* ptr_arr) { 269 // Get a new chunk. 270 TaskQueueEntryChunk* new_chunk = remove_chunk_from_free_list(); 271 272 if (new_chunk == NULL) { 273 // Did not get a chunk from the free list. Allocate from backing memory. 274 new_chunk = allocate_new_chunk(); 275 276 if (new_chunk == NULL) { 277 return false; 278 } 279 } 280 281 Copy::conjoint_memory_atomic(ptr_arr, new_chunk->data, EntriesPerChunk * sizeof(G1TaskQueueEntry)); 282 283 add_chunk_to_chunk_list(new_chunk); 284 285 return true; 286 } 287 288 bool G1CMMarkStack::par_pop_chunk(G1TaskQueueEntry* ptr_arr) { 289 TaskQueueEntryChunk* cur = remove_chunk_from_chunk_list(); 290 291 if (cur == NULL) { 292 return false; 293 } 294 295 Copy::conjoint_memory_atomic(cur->data, ptr_arr, EntriesPerChunk * sizeof(G1TaskQueueEntry)); 296 297 add_chunk_to_free_list(cur); 298 return true; 299 } 300 301 void G1CMMarkStack::set_empty() { 302 _chunks_in_chunk_list = 0; 303 _hwm = 0; 304 _chunk_list = NULL; 305 _free_list = NULL; 306 } 307 308 G1CMRootRegions::G1CMRootRegions() : 309 _cm(NULL), _scan_in_progress(false), 310 _should_abort(false), _claimed_survivor_index(0) { } 311 312 void G1CMRootRegions::init(const G1SurvivorRegions* survivors, G1ConcurrentMark* cm) { 313 _survivors = survivors; 314 _cm = cm; 315 } 316 317 void G1CMRootRegions::prepare_for_scan() { 318 assert(!scan_in_progress(), "pre-condition"); 319 320 // Currently, only survivors can be root regions. 321 _claimed_survivor_index = 0; 322 _scan_in_progress = _survivors->regions()->is_nonempty(); 323 _should_abort = false; 324 } 325 326 HeapRegion* G1CMRootRegions::claim_next() { 327 if (_should_abort) { 328 // If someone has set the should_abort flag, we return NULL to 329 // force the caller to bail out of their loop. 330 return NULL; 331 } 332 333 // Currently, only survivors can be root regions. 334 const GrowableArray<HeapRegion*>* survivor_regions = _survivors->regions(); 335 336 int claimed_index = Atomic::add(1, &_claimed_survivor_index) - 1; 337 if (claimed_index < survivor_regions->length()) { 338 return survivor_regions->at(claimed_index); 339 } 340 return NULL; 341 } 342 343 uint G1CMRootRegions::num_root_regions() const { 344 return (uint)_survivors->regions()->length(); 345 } 346 347 void G1CMRootRegions::notify_scan_done() { 348 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 349 _scan_in_progress = false; 350 RootRegionScan_lock->notify_all(); 351 } 352 353 void G1CMRootRegions::cancel_scan() { 354 notify_scan_done(); 355 } 356 357 void G1CMRootRegions::scan_finished() { 358 assert(scan_in_progress(), "pre-condition"); 359 360 // Currently, only survivors can be root regions. 361 if (!_should_abort) { 362 assert(_claimed_survivor_index >= 0, "otherwise comparison is invalid: %d", _claimed_survivor_index); 363 assert((uint)_claimed_survivor_index >= _survivors->length(), 364 "we should have claimed all survivors, claimed index = %u, length = %u", 365 (uint)_claimed_survivor_index, _survivors->length()); 366 } 367 368 notify_scan_done(); 369 } 370 371 bool G1CMRootRegions::wait_until_scan_finished() { 372 if (!scan_in_progress()) return false; 373 374 { 375 MutexLockerEx x(RootRegionScan_lock, Mutex::_no_safepoint_check_flag); 376 while (scan_in_progress()) { 377 RootRegionScan_lock->wait(Mutex::_no_safepoint_check_flag); 378 } 379 } 380 return true; 381 } 382 383 uint G1ConcurrentMark::scale_parallel_threads(uint n_par_threads) { 384 return MAX2((n_par_threads + 2) / 4, 1U); 385 } 386 387 G1ConcurrentMark::G1ConcurrentMark(G1CollectedHeap* g1h, G1RegionToSpaceMapper* prev_bitmap_storage, G1RegionToSpaceMapper* next_bitmap_storage) : 388 _g1h(g1h), 389 _markBitMap1(), 390 _markBitMap2(), 391 _parallel_marking_threads(0), 392 _max_parallel_marking_threads(0), 393 _sleep_factor(0.0), 394 _marking_task_overhead(1.0), 395 _cleanup_list("Cleanup List"), 396 397 _prevMarkBitMap(&_markBitMap1), 398 _nextMarkBitMap(&_markBitMap2), 399 400 _global_mark_stack(), 401 // _finger set in set_non_marking_state 402 403 _max_worker_id(ParallelGCThreads), 404 // _active_tasks set in set_non_marking_state 405 // _tasks set inside the constructor 406 _task_queues(new G1CMTaskQueueSet((int) _max_worker_id)), 407 _terminator(ParallelTaskTerminator((int) _max_worker_id, _task_queues)), 408 409 _has_overflown(false), 410 _concurrent(false), 411 _has_aborted(false), 412 _restart_for_overflow(false), 413 _concurrent_marking_in_progress(false), 414 _gc_timer_cm(new (ResourceObj::C_HEAP, mtGC) ConcurrentGCTimer()), 415 _gc_tracer_cm(new (ResourceObj::C_HEAP, mtGC) G1OldTracer()), 416 417 // _verbose_level set below 418 419 _init_times(), 420 _remark_times(), _remark_mark_times(), _remark_weak_ref_times(), 421 _cleanup_times(), 422 _total_counting_time(0.0), 423 _total_rs_scrub_time(0.0), 424 425 _parallel_workers(NULL), 426 427 _completed_initialization(false) { 428 429 _markBitMap1.initialize(g1h->reserved_region(), prev_bitmap_storage); 430 _markBitMap2.initialize(g1h->reserved_region(), next_bitmap_storage); 431 432 // Create & start a ConcurrentMark thread. 433 _cmThread = new ConcurrentMarkThread(this); 434 assert(cmThread() != NULL, "CM Thread should have been created"); 435 assert(cmThread()->cm() != NULL, "CM Thread should refer to this cm"); 436 if (_cmThread->osthread() == NULL) { 437 vm_shutdown_during_initialization("Could not create ConcurrentMarkThread"); 438 } 439 440 assert(CGC_lock != NULL, "Where's the CGC_lock?"); 441 assert(_markBitMap1.covers(g1h->reserved_region()), "_markBitMap1 inconsistency"); 442 assert(_markBitMap2.covers(g1h->reserved_region()), "_markBitMap2 inconsistency"); 443 444 SATBMarkQueueSet& satb_qs = JavaThread::satb_mark_queue_set(); 445 satb_qs.set_buffer_size(G1SATBBufferSize); 446 447 _root_regions.init(_g1h->survivor(), this); 448 449 if (ConcGCThreads > ParallelGCThreads) { 450 log_warning(gc)("Can't have more ConcGCThreads (%u) than ParallelGCThreads (%u).", 451 ConcGCThreads, ParallelGCThreads); 452 return; 453 } 454 if (!FLAG_IS_DEFAULT(ConcGCThreads) && ConcGCThreads > 0) { 455 // Note: ConcGCThreads has precedence over G1MarkingOverheadPercent 456 // if both are set 457 _sleep_factor = 0.0; 458 _marking_task_overhead = 1.0; 459 } else if (G1MarkingOverheadPercent > 0) { 460 // We will calculate the number of parallel marking threads based 461 // on a target overhead with respect to the soft real-time goal 462 double marking_overhead = (double) G1MarkingOverheadPercent / 100.0; 463 double overall_cm_overhead = 464 (double) MaxGCPauseMillis * marking_overhead / 465 (double) GCPauseIntervalMillis; 466 double cpu_ratio = 1.0 / os::initial_active_processor_count(); 467 double marking_thread_num = ceil(overall_cm_overhead / cpu_ratio); 468 double marking_task_overhead = 469 overall_cm_overhead / marking_thread_num * os::initial_active_processor_count(); 470 double sleep_factor = 471 (1.0 - marking_task_overhead) / marking_task_overhead; 472 473 FLAG_SET_ERGO(uint, ConcGCThreads, (uint) marking_thread_num); 474 _sleep_factor = sleep_factor; 475 _marking_task_overhead = marking_task_overhead; 476 } else { 477 // Calculate the number of parallel marking threads by scaling 478 // the number of parallel GC threads. 479 uint marking_thread_num = scale_parallel_threads(ParallelGCThreads); 480 FLAG_SET_ERGO(uint, ConcGCThreads, marking_thread_num); 481 _sleep_factor = 0.0; 482 _marking_task_overhead = 1.0; 483 } 484 485 assert(ConcGCThreads > 0, "Should have been set"); 486 log_debug(gc)("ConcGCThreads: %u", ConcGCThreads); 487 log_debug(gc)("ParallelGCThreads: %u", ParallelGCThreads); 488 _parallel_marking_threads = ConcGCThreads; 489 _max_parallel_marking_threads = _parallel_marking_threads; 490 491 _parallel_workers = new WorkGang("G1 Marker", 492 _max_parallel_marking_threads, false, true); 493 if (_parallel_workers == NULL) { 494 vm_exit_during_initialization("Failed necessary allocation."); 495 } else { 496 _parallel_workers->initialize_workers(); 497 } 498 499 if (FLAG_IS_DEFAULT(MarkStackSize)) { 500 size_t mark_stack_size = 501 MIN2(MarkStackSizeMax, 502 MAX2(MarkStackSize, (size_t) (parallel_marking_threads() * TASKQUEUE_SIZE))); 503 // Verify that the calculated value for MarkStackSize is in range. 504 // It would be nice to use the private utility routine from Arguments. 505 if (!(mark_stack_size >= 1 && mark_stack_size <= MarkStackSizeMax)) { 506 log_warning(gc)("Invalid value calculated for MarkStackSize (" SIZE_FORMAT "): " 507 "must be between 1 and " SIZE_FORMAT, 508 mark_stack_size, MarkStackSizeMax); 509 return; 510 } 511 FLAG_SET_ERGO(size_t, MarkStackSize, mark_stack_size); 512 } else { 513 // Verify MarkStackSize is in range. 514 if (FLAG_IS_CMDLINE(MarkStackSize)) { 515 if (FLAG_IS_DEFAULT(MarkStackSizeMax)) { 516 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 517 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT "): " 518 "must be between 1 and " SIZE_FORMAT, 519 MarkStackSize, MarkStackSizeMax); 520 return; 521 } 522 } else if (FLAG_IS_CMDLINE(MarkStackSizeMax)) { 523 if (!(MarkStackSize >= 1 && MarkStackSize <= MarkStackSizeMax)) { 524 log_warning(gc)("Invalid value specified for MarkStackSize (" SIZE_FORMAT ")" 525 " or for MarkStackSizeMax (" SIZE_FORMAT ")", 526 MarkStackSize, MarkStackSizeMax); 527 return; 528 } 529 } 530 } 531 } 532 533 if (!_global_mark_stack.initialize(MarkStackSize, MarkStackSizeMax)) { 534 vm_exit_during_initialization("Failed to allocate initial concurrent mark overflow mark stack."); 535 } 536 537 _tasks = NEW_C_HEAP_ARRAY(G1CMTask*, _max_worker_id, mtGC); 538 _accum_task_vtime = NEW_C_HEAP_ARRAY(double, _max_worker_id, mtGC); 539 540 // so that the assertion in MarkingTaskQueue::task_queue doesn't fail 541 _active_tasks = _max_worker_id; 542 543 for (uint i = 0; i < _max_worker_id; ++i) { 544 G1CMTaskQueue* task_queue = new G1CMTaskQueue(); 545 task_queue->initialize(); 546 _task_queues->register_queue(i, task_queue); 547 548 _tasks[i] = new G1CMTask(i, this, task_queue, _task_queues); 549 550 _accum_task_vtime[i] = 0.0; 551 } 552 553 // so that the call below can read a sensible value 554 _heap_start = g1h->reserved_region().start(); 555 set_non_marking_state(); 556 _completed_initialization = true; 557 } 558 559 void G1ConcurrentMark::reset() { 560 // Starting values for these two. This should be called in a STW 561 // phase. 562 MemRegion reserved = _g1h->g1_reserved(); 563 _heap_start = reserved.start(); 564 _heap_end = reserved.end(); 565 566 // Separated the asserts so that we know which one fires. 567 assert(_heap_start != NULL, "heap bounds should look ok"); 568 assert(_heap_end != NULL, "heap bounds should look ok"); 569 assert(_heap_start < _heap_end, "heap bounds should look ok"); 570 571 // Reset all the marking data structures and any necessary flags 572 reset_marking_state(); 573 574 // We do reset all of them, since different phases will use 575 // different number of active threads. So, it's easiest to have all 576 // of them ready. 577 for (uint i = 0; i < _max_worker_id; ++i) { 578 _tasks[i]->reset(_nextMarkBitMap); 579 } 580 581 // we need this to make sure that the flag is on during the evac 582 // pause with initial mark piggy-backed 583 set_concurrent_marking_in_progress(); 584 } 585 586 587 void G1ConcurrentMark::reset_marking_state() { 588 _global_mark_stack.set_empty(); 589 590 // Expand the marking stack, if we have to and if we can. 591 if (has_overflown()) { 592 _global_mark_stack.expand(); 593 } 594 595 clear_has_overflown(); 596 _finger = _heap_start; 597 598 for (uint i = 0; i < _max_worker_id; ++i) { 599 G1CMTaskQueue* queue = _task_queues->queue(i); 600 queue->set_empty(); 601 } 602 } 603 604 void G1ConcurrentMark::set_concurrency(uint active_tasks) { 605 assert(active_tasks <= _max_worker_id, "we should not have more"); 606 607 _active_tasks = active_tasks; 608 // Need to update the three data structures below according to the 609 // number of active threads for this phase. 610 _terminator = ParallelTaskTerminator((int) active_tasks, _task_queues); 611 _first_overflow_barrier_sync.set_n_workers((int) active_tasks); 612 _second_overflow_barrier_sync.set_n_workers((int) active_tasks); 613 } 614 615 void G1ConcurrentMark::set_concurrency_and_phase(uint active_tasks, bool concurrent) { 616 set_concurrency(active_tasks); 617 618 _concurrent = concurrent; 619 // We propagate this to all tasks, not just the active ones. 620 for (uint i = 0; i < _max_worker_id; ++i) 621 _tasks[i]->set_concurrent(concurrent); 622 623 if (concurrent) { 624 set_concurrent_marking_in_progress(); 625 } else { 626 // We currently assume that the concurrent flag has been set to 627 // false before we start remark. At this point we should also be 628 // in a STW phase. 629 assert(!concurrent_marking_in_progress(), "invariant"); 630 assert(out_of_regions(), 631 "only way to get here: _finger: " PTR_FORMAT ", _heap_end: " PTR_FORMAT, 632 p2i(_finger), p2i(_heap_end)); 633 } 634 } 635 636 void G1ConcurrentMark::set_non_marking_state() { 637 // We set the global marking state to some default values when we're 638 // not doing marking. 639 reset_marking_state(); 640 _active_tasks = 0; 641 clear_concurrent_marking_in_progress(); 642 } 643 644 G1ConcurrentMark::~G1ConcurrentMark() { 645 // The G1ConcurrentMark instance is never freed. 646 ShouldNotReachHere(); 647 } 648 649 class G1ClearBitMapTask : public AbstractGangTask { 650 public: 651 static size_t chunk_size() { return M; } 652 653 private: 654 // Heap region closure used for clearing the given mark bitmap. 655 class G1ClearBitmapHRClosure : public HeapRegionClosure { 656 private: 657 G1CMBitMap* _bitmap; 658 G1ConcurrentMark* _cm; 659 public: 660 G1ClearBitmapHRClosure(G1CMBitMap* bitmap, G1ConcurrentMark* cm) : HeapRegionClosure(), _cm(cm), _bitmap(bitmap) { 661 } 662 663 virtual bool doHeapRegion(HeapRegion* r) { 664 size_t const chunk_size_in_words = G1ClearBitMapTask::chunk_size() / HeapWordSize; 665 666 HeapWord* cur = r->bottom(); 667 HeapWord* const end = r->end(); 668 669 while (cur < end) { 670 MemRegion mr(cur, MIN2(cur + chunk_size_in_words, end)); 671 _bitmap->clear_range(mr); 672 673 cur += chunk_size_in_words; 674 675 // Abort iteration if after yielding the marking has been aborted. 676 if (_cm != NULL && _cm->do_yield_check() && _cm->has_aborted()) { 677 return true; 678 } 679 // Repeat the asserts from before the start of the closure. We will do them 680 // as asserts here to minimize their overhead on the product. However, we 681 // will have them as guarantees at the beginning / end of the bitmap 682 // clearing to get some checking in the product. 683 assert(_cm == NULL || _cm->cmThread()->during_cycle(), "invariant"); 684 assert(_cm == NULL || !G1CollectedHeap::heap()->collector_state()->mark_in_progress(), "invariant"); 685 } 686 assert(cur == end, "Must have completed iteration over the bitmap for region %u.", r->hrm_index()); 687 688 return false; 689 } 690 }; 691 692 G1ClearBitmapHRClosure _cl; 693 HeapRegionClaimer _hr_claimer; 694 bool _suspendible; // If the task is suspendible, workers must join the STS. 695 696 public: 697 G1ClearBitMapTask(G1CMBitMap* bitmap, G1ConcurrentMark* cm, uint n_workers, bool suspendible) : 698 AbstractGangTask("G1 Clear Bitmap"), 699 _cl(bitmap, suspendible ? cm : NULL), 700 _hr_claimer(n_workers), 701 _suspendible(suspendible) 702 { } 703 704 void work(uint worker_id) { 705 SuspendibleThreadSetJoiner sts_join(_suspendible); 706 G1CollectedHeap::heap()->heap_region_par_iterate(&_cl, worker_id, &_hr_claimer); 707 } 708 709 bool is_complete() { 710 return _cl.complete(); 711 } 712 }; 713 714 void G1ConcurrentMark::clear_bitmap(G1CMBitMap* bitmap, WorkGang* workers, bool may_yield) { 715 assert(may_yield || SafepointSynchronize::is_at_safepoint(), "Non-yielding bitmap clear only allowed at safepoint."); 716 717 size_t const num_bytes_to_clear = (HeapRegion::GrainBytes * _g1h->num_regions()) / G1CMBitMap::heap_map_factor(); 718 size_t const num_chunks = align_up(num_bytes_to_clear, G1ClearBitMapTask::chunk_size()) / G1ClearBitMapTask::chunk_size(); 719 720 uint const num_workers = (uint)MIN2(num_chunks, (size_t)workers->active_workers()); 721 722 G1ClearBitMapTask cl(bitmap, this, num_workers, may_yield); 723 724 log_debug(gc, ergo)("Running %s with %u workers for " SIZE_FORMAT " work units.", cl.name(), num_workers, num_chunks); 725 workers->run_task(&cl, num_workers); 726 guarantee(!may_yield || cl.is_complete(), "Must have completed iteration when not yielding."); 727 } 728 729 void G1ConcurrentMark::cleanup_for_next_mark() { 730 // Make sure that the concurrent mark thread looks to still be in 731 // the current cycle. 732 guarantee(cmThread()->during_cycle(), "invariant"); 733 734 // We are finishing up the current cycle by clearing the next 735 // marking bitmap and getting it ready for the next cycle. During 736 // this time no other cycle can start. So, let's make sure that this 737 // is the case. 738 guarantee(!_g1h->collector_state()->mark_in_progress(), "invariant"); 739 740 clear_bitmap(_nextMarkBitMap, _parallel_workers, true); 741 742 // Clear the live count data. If the marking has been aborted, the abort() 743 // call already did that. 744 if (!has_aborted()) { 745 clear_live_data(_parallel_workers); 746 DEBUG_ONLY(verify_live_data_clear()); 747 } 748 749 // Repeat the asserts from above. 750 guarantee(cmThread()->during_cycle(), "invariant"); 751 guarantee(!_g1h->collector_state()->mark_in_progress(), "invariant"); 752 } 753 754 void G1ConcurrentMark::clear_prev_bitmap(WorkGang* workers) { 755 assert(SafepointSynchronize::is_at_safepoint(), "Should only clear the entire prev bitmap at a safepoint."); 756 clear_bitmap((G1CMBitMap*)_prevMarkBitMap, workers, false); 757 } 758 759 class CheckBitmapClearHRClosure : public HeapRegionClosure { 760 G1CMBitMap* _bitmap; 761 bool _error; 762 public: 763 CheckBitmapClearHRClosure(G1CMBitMap* bitmap) : _bitmap(bitmap) { 764 } 765 766 virtual bool doHeapRegion(HeapRegion* r) { 767 // This closure can be called concurrently to the mutator, so we must make sure 768 // that the result of the getNextMarkedWordAddress() call is compared to the 769 // value passed to it as limit to detect any found bits. 770 // end never changes in G1. 771 HeapWord* end = r->end(); 772 return _bitmap->getNextMarkedWordAddress(r->bottom(), end) != end; 773 } 774 }; 775 776 bool G1ConcurrentMark::nextMarkBitmapIsClear() { 777 CheckBitmapClearHRClosure cl(_nextMarkBitMap); 778 _g1h->heap_region_iterate(&cl); 779 return cl.complete(); 780 } 781 782 class NoteStartOfMarkHRClosure: public HeapRegionClosure { 783 public: 784 bool doHeapRegion(HeapRegion* r) { 785 r->note_start_of_marking(); 786 return false; 787 } 788 }; 789 790 void G1ConcurrentMark::checkpointRootsInitialPre() { 791 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 792 G1Policy* g1p = g1h->g1_policy(); 793 794 _has_aborted = false; 795 796 // Initialize marking structures. This has to be done in a STW phase. 797 reset(); 798 799 // For each region note start of marking. 800 NoteStartOfMarkHRClosure startcl; 801 g1h->heap_region_iterate(&startcl); 802 } 803 804 805 void G1ConcurrentMark::checkpointRootsInitialPost() { 806 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 807 808 // Start Concurrent Marking weak-reference discovery. 809 ReferenceProcessor* rp = g1h->ref_processor_cm(); 810 // enable ("weak") refs discovery 811 rp->enable_discovery(); 812 rp->setup_policy(false); // snapshot the soft ref policy to be used in this cycle 813 814 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 815 // This is the start of the marking cycle, we're expected all 816 // threads to have SATB queues with active set to false. 817 satb_mq_set.set_active_all_threads(true, /* new active value */ 818 false /* expected_active */); 819 820 _root_regions.prepare_for_scan(); 821 822 // update_g1_committed() will be called at the end of an evac pause 823 // when marking is on. So, it's also called at the end of the 824 // initial-mark pause to update the heap end, if the heap expands 825 // during it. No need to call it here. 826 } 827 828 /* 829 * Notice that in the next two methods, we actually leave the STS 830 * during the barrier sync and join it immediately afterwards. If we 831 * do not do this, the following deadlock can occur: one thread could 832 * be in the barrier sync code, waiting for the other thread to also 833 * sync up, whereas another one could be trying to yield, while also 834 * waiting for the other threads to sync up too. 835 * 836 * Note, however, that this code is also used during remark and in 837 * this case we should not attempt to leave / enter the STS, otherwise 838 * we'll either hit an assert (debug / fastdebug) or deadlock 839 * (product). So we should only leave / enter the STS if we are 840 * operating concurrently. 841 * 842 * Because the thread that does the sync barrier has left the STS, it 843 * is possible to be suspended for a Full GC or an evacuation pause 844 * could occur. This is actually safe, since the entering the sync 845 * barrier is one of the last things do_marking_step() does, and it 846 * doesn't manipulate any data structures afterwards. 847 */ 848 849 void G1ConcurrentMark::enter_first_sync_barrier(uint worker_id) { 850 bool barrier_aborted; 851 { 852 SuspendibleThreadSetLeaver sts_leave(concurrent()); 853 barrier_aborted = !_first_overflow_barrier_sync.enter(); 854 } 855 856 // at this point everyone should have synced up and not be doing any 857 // more work 858 859 if (barrier_aborted) { 860 // If the barrier aborted we ignore the overflow condition and 861 // just abort the whole marking phase as quickly as possible. 862 return; 863 } 864 865 // If we're executing the concurrent phase of marking, reset the marking 866 // state; otherwise the marking state is reset after reference processing, 867 // during the remark pause. 868 // If we reset here as a result of an overflow during the remark we will 869 // see assertion failures from any subsequent set_concurrency_and_phase() 870 // calls. 871 if (concurrent()) { 872 // let the task associated with with worker 0 do this 873 if (worker_id == 0) { 874 // task 0 is responsible for clearing the global data structures 875 // We should be here because of an overflow. During STW we should 876 // not clear the overflow flag since we rely on it being true when 877 // we exit this method to abort the pause and restart concurrent 878 // marking. 879 reset_marking_state(); 880 881 log_info(gc, marking)("Concurrent Mark reset for overflow"); 882 } 883 } 884 885 // after this, each task should reset its own data structures then 886 // then go into the second barrier 887 } 888 889 void G1ConcurrentMark::enter_second_sync_barrier(uint worker_id) { 890 SuspendibleThreadSetLeaver sts_leave(concurrent()); 891 _second_overflow_barrier_sync.enter(); 892 893 // at this point everything should be re-initialized and ready to go 894 } 895 896 class G1CMConcurrentMarkingTask: public AbstractGangTask { 897 private: 898 G1ConcurrentMark* _cm; 899 ConcurrentMarkThread* _cmt; 900 901 public: 902 void work(uint worker_id) { 903 assert(Thread::current()->is_ConcurrentGC_thread(), 904 "this should only be done by a conc GC thread"); 905 ResourceMark rm; 906 907 double start_vtime = os::elapsedVTime(); 908 909 { 910 SuspendibleThreadSetJoiner sts_join; 911 912 assert(worker_id < _cm->active_tasks(), "invariant"); 913 G1CMTask* the_task = _cm->task(worker_id); 914 the_task->record_start_time(); 915 if (!_cm->has_aborted()) { 916 do { 917 double start_vtime_sec = os::elapsedVTime(); 918 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 919 920 the_task->do_marking_step(mark_step_duration_ms, 921 true /* do_termination */, 922 false /* is_serial*/); 923 924 double end_vtime_sec = os::elapsedVTime(); 925 double elapsed_vtime_sec = end_vtime_sec - start_vtime_sec; 926 _cm->do_yield_check(); 927 928 jlong sleep_time_ms; 929 if (!_cm->has_aborted() && the_task->has_aborted()) { 930 sleep_time_ms = 931 (jlong) (elapsed_vtime_sec * _cm->sleep_factor() * 1000.0); 932 { 933 SuspendibleThreadSetLeaver sts_leave; 934 os::sleep(Thread::current(), sleep_time_ms, false); 935 } 936 } 937 } while (!_cm->has_aborted() && the_task->has_aborted()); 938 } 939 the_task->record_end_time(); 940 guarantee(!the_task->has_aborted() || _cm->has_aborted(), "invariant"); 941 } 942 943 double end_vtime = os::elapsedVTime(); 944 _cm->update_accum_task_vtime(worker_id, end_vtime - start_vtime); 945 } 946 947 G1CMConcurrentMarkingTask(G1ConcurrentMark* cm, 948 ConcurrentMarkThread* cmt) : 949 AbstractGangTask("Concurrent Mark"), _cm(cm), _cmt(cmt) { } 950 951 ~G1CMConcurrentMarkingTask() { } 952 }; 953 954 // Calculates the number of active workers for a concurrent 955 // phase. 956 uint G1ConcurrentMark::calc_parallel_marking_threads() { 957 uint n_conc_workers = 0; 958 if (!UseDynamicNumberOfGCThreads || 959 (!FLAG_IS_DEFAULT(ConcGCThreads) && 960 !ForceDynamicNumberOfGCThreads)) { 961 n_conc_workers = max_parallel_marking_threads(); 962 } else { 963 n_conc_workers = 964 AdaptiveSizePolicy::calc_default_active_workers(max_parallel_marking_threads(), 965 1, /* Minimum workers */ 966 parallel_marking_threads(), 967 Threads::number_of_non_daemon_threads()); 968 // Don't scale down "n_conc_workers" by scale_parallel_threads() because 969 // that scaling has already gone into "_max_parallel_marking_threads". 970 } 971 assert(n_conc_workers > 0 && n_conc_workers <= max_parallel_marking_threads(), 972 "Calculated number of workers must be larger than zero and at most the maximum %u, but is %u", 973 max_parallel_marking_threads(), n_conc_workers); 974 return n_conc_workers; 975 } 976 977 void G1ConcurrentMark::scanRootRegion(HeapRegion* hr) { 978 // Currently, only survivors can be root regions. 979 assert(hr->next_top_at_mark_start() == hr->bottom(), "invariant"); 980 G1RootRegionScanClosure cl(_g1h, this); 981 982 const uintx interval = PrefetchScanIntervalInBytes; 983 HeapWord* curr = hr->bottom(); 984 const HeapWord* end = hr->top(); 985 while (curr < end) { 986 Prefetch::read(curr, interval); 987 oop obj = oop(curr); 988 int size = obj->oop_iterate_size(&cl); 989 assert(size == obj->size(), "sanity"); 990 curr += size; 991 } 992 } 993 994 class G1CMRootRegionScanTask : public AbstractGangTask { 995 private: 996 G1ConcurrentMark* _cm; 997 998 public: 999 G1CMRootRegionScanTask(G1ConcurrentMark* cm) : 1000 AbstractGangTask("G1 Root Region Scan"), _cm(cm) { } 1001 1002 void work(uint worker_id) { 1003 assert(Thread::current()->is_ConcurrentGC_thread(), 1004 "this should only be done by a conc GC thread"); 1005 1006 G1CMRootRegions* root_regions = _cm->root_regions(); 1007 HeapRegion* hr = root_regions->claim_next(); 1008 while (hr != NULL) { 1009 _cm->scanRootRegion(hr); 1010 hr = root_regions->claim_next(); 1011 } 1012 } 1013 }; 1014 1015 void G1ConcurrentMark::scan_root_regions() { 1016 // scan_in_progress() will have been set to true only if there was 1017 // at least one root region to scan. So, if it's false, we 1018 // should not attempt to do any further work. 1019 if (root_regions()->scan_in_progress()) { 1020 assert(!has_aborted(), "Aborting before root region scanning is finished not supported."); 1021 1022 _parallel_marking_threads = MIN2(calc_parallel_marking_threads(), 1023 // We distribute work on a per-region basis, so starting 1024 // more threads than that is useless. 1025 root_regions()->num_root_regions()); 1026 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1027 "Maximum number of marking threads exceeded"); 1028 1029 G1CMRootRegionScanTask task(this); 1030 log_debug(gc, ergo)("Running %s using %u workers for %u work units.", 1031 task.name(), _parallel_marking_threads, root_regions()->num_root_regions()); 1032 _parallel_workers->run_task(&task, _parallel_marking_threads); 1033 1034 // It's possible that has_aborted() is true here without actually 1035 // aborting the survivor scan earlier. This is OK as it's 1036 // mainly used for sanity checking. 1037 root_regions()->scan_finished(); 1038 } 1039 } 1040 1041 void G1ConcurrentMark::concurrent_cycle_start() { 1042 _gc_timer_cm->register_gc_start(); 1043 1044 _gc_tracer_cm->report_gc_start(GCCause::_no_gc /* first parameter is not used */, _gc_timer_cm->gc_start()); 1045 1046 _g1h->trace_heap_before_gc(_gc_tracer_cm); 1047 } 1048 1049 void G1ConcurrentMark::concurrent_cycle_end() { 1050 _g1h->trace_heap_after_gc(_gc_tracer_cm); 1051 1052 if (has_aborted()) { 1053 _gc_tracer_cm->report_concurrent_mode_failure(); 1054 } 1055 1056 _gc_timer_cm->register_gc_end(); 1057 1058 _gc_tracer_cm->report_gc_end(_gc_timer_cm->gc_end(), _gc_timer_cm->time_partitions()); 1059 } 1060 1061 void G1ConcurrentMark::mark_from_roots() { 1062 // we might be tempted to assert that: 1063 // assert(asynch == !SafepointSynchronize::is_at_safepoint(), 1064 // "inconsistent argument?"); 1065 // However that wouldn't be right, because it's possible that 1066 // a safepoint is indeed in progress as a younger generation 1067 // stop-the-world GC happens even as we mark in this generation. 1068 1069 _restart_for_overflow = false; 1070 1071 // _g1h has _n_par_threads 1072 _parallel_marking_threads = calc_parallel_marking_threads(); 1073 assert(parallel_marking_threads() <= max_parallel_marking_threads(), 1074 "Maximum number of marking threads exceeded"); 1075 1076 uint active_workers = MAX2(1U, parallel_marking_threads()); 1077 assert(active_workers > 0, "Should have been set"); 1078 1079 // Setting active workers is not guaranteed since fewer 1080 // worker threads may currently exist and more may not be 1081 // available. 1082 active_workers = _parallel_workers->update_active_workers(active_workers); 1083 log_info(gc, task)("Using %u workers of %u for marking", active_workers, _parallel_workers->total_workers()); 1084 1085 // Parallel task terminator is set in "set_concurrency_and_phase()" 1086 set_concurrency_and_phase(active_workers, true /* concurrent */); 1087 1088 G1CMConcurrentMarkingTask markingTask(this, cmThread()); 1089 _parallel_workers->run_task(&markingTask); 1090 print_stats(); 1091 } 1092 1093 void G1ConcurrentMark::checkpointRootsFinal(bool clear_all_soft_refs) { 1094 // world is stopped at this checkpoint 1095 assert(SafepointSynchronize::is_at_safepoint(), 1096 "world should be stopped"); 1097 1098 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1099 1100 // If a full collection has happened, we shouldn't do this. 1101 if (has_aborted()) { 1102 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused 1103 return; 1104 } 1105 1106 SvcGCMarker sgcm(SvcGCMarker::OTHER); 1107 1108 if (VerifyDuringGC) { 1109 HandleMark hm; // handle scope 1110 g1h->prepare_for_verify(); 1111 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)"); 1112 } 1113 g1h->verifier()->check_bitmaps("Remark Start"); 1114 1115 G1Policy* g1p = g1h->g1_policy(); 1116 g1p->record_concurrent_mark_remark_start(); 1117 1118 double start = os::elapsedTime(); 1119 1120 checkpointRootsFinalWork(); 1121 1122 double mark_work_end = os::elapsedTime(); 1123 1124 weakRefsWork(clear_all_soft_refs); 1125 1126 if (has_overflown()) { 1127 // We overflowed. Restart concurrent marking. 1128 _restart_for_overflow = true; 1129 1130 // Verify the heap w.r.t. the previous marking bitmap. 1131 if (VerifyDuringGC) { 1132 HandleMark hm; // handle scope 1133 g1h->prepare_for_verify(); 1134 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (overflow)"); 1135 } 1136 1137 // Clear the marking state because we will be restarting 1138 // marking due to overflowing the global mark stack. 1139 reset_marking_state(); 1140 } else { 1141 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 1142 // We're done with marking. 1143 // This is the end of the marking cycle, we're expected all 1144 // threads to have SATB queues with active set to true. 1145 satb_mq_set.set_active_all_threads(false, /* new active value */ 1146 true /* expected_active */); 1147 1148 if (VerifyDuringGC) { 1149 HandleMark hm; // handle scope 1150 g1h->prepare_for_verify(); 1151 Universe::verify(VerifyOption_G1UseNextMarking, "During GC (after)"); 1152 } 1153 g1h->verifier()->check_bitmaps("Remark End"); 1154 assert(!restart_for_overflow(), "sanity"); 1155 // Completely reset the marking state since marking completed 1156 set_non_marking_state(); 1157 } 1158 1159 // Statistics 1160 double now = os::elapsedTime(); 1161 _remark_mark_times.add((mark_work_end - start) * 1000.0); 1162 _remark_weak_ref_times.add((now - mark_work_end) * 1000.0); 1163 _remark_times.add((now - start) * 1000.0); 1164 1165 g1p->record_concurrent_mark_remark_end(); 1166 1167 G1CMIsAliveClosure is_alive(g1h); 1168 _gc_tracer_cm->report_object_count_after_gc(&is_alive); 1169 } 1170 1171 class G1NoteEndOfConcMarkClosure : public HeapRegionClosure { 1172 G1CollectedHeap* _g1; 1173 size_t _freed_bytes; 1174 FreeRegionList* _local_cleanup_list; 1175 uint _old_regions_removed; 1176 uint _humongous_regions_removed; 1177 HRRSCleanupTask* _hrrs_cleanup_task; 1178 1179 public: 1180 G1NoteEndOfConcMarkClosure(G1CollectedHeap* g1, 1181 FreeRegionList* local_cleanup_list, 1182 HRRSCleanupTask* hrrs_cleanup_task) : 1183 _g1(g1), 1184 _freed_bytes(0), 1185 _local_cleanup_list(local_cleanup_list), 1186 _old_regions_removed(0), 1187 _humongous_regions_removed(0), 1188 _hrrs_cleanup_task(hrrs_cleanup_task) { } 1189 1190 size_t freed_bytes() { return _freed_bytes; } 1191 const uint old_regions_removed() { return _old_regions_removed; } 1192 const uint humongous_regions_removed() { return _humongous_regions_removed; } 1193 1194 bool doHeapRegion(HeapRegion *hr) { 1195 if (hr->is_archive()) { 1196 return false; 1197 } 1198 _g1->reset_gc_time_stamps(hr); 1199 hr->note_end_of_marking(); 1200 1201 if (hr->used() > 0 && hr->max_live_bytes() == 0 && !hr->is_young()) { 1202 _freed_bytes += hr->used(); 1203 hr->set_containing_set(NULL); 1204 if (hr->is_humongous()) { 1205 _humongous_regions_removed++; 1206 _g1->free_humongous_region(hr, _local_cleanup_list, true /* skip_remset */); 1207 } else { 1208 _old_regions_removed++; 1209 _g1->free_region(hr, _local_cleanup_list, true /* skip_remset */); 1210 } 1211 } else { 1212 hr->rem_set()->do_cleanup_work(_hrrs_cleanup_task); 1213 } 1214 1215 return false; 1216 } 1217 }; 1218 1219 class G1ParNoteEndTask: public AbstractGangTask { 1220 friend class G1NoteEndOfConcMarkClosure; 1221 1222 protected: 1223 G1CollectedHeap* _g1h; 1224 FreeRegionList* _cleanup_list; 1225 HeapRegionClaimer _hrclaimer; 1226 1227 public: 1228 G1ParNoteEndTask(G1CollectedHeap* g1h, FreeRegionList* cleanup_list, uint n_workers) : 1229 AbstractGangTask("G1 note end"), _g1h(g1h), _cleanup_list(cleanup_list), _hrclaimer(n_workers) { 1230 } 1231 1232 void work(uint worker_id) { 1233 FreeRegionList local_cleanup_list("Local Cleanup List"); 1234 HRRSCleanupTask hrrs_cleanup_task; 1235 G1NoteEndOfConcMarkClosure g1_note_end(_g1h, &local_cleanup_list, 1236 &hrrs_cleanup_task); 1237 _g1h->heap_region_par_iterate(&g1_note_end, worker_id, &_hrclaimer); 1238 assert(g1_note_end.complete(), "Shouldn't have yielded!"); 1239 1240 // Now update the lists 1241 _g1h->remove_from_old_sets(g1_note_end.old_regions_removed(), g1_note_end.humongous_regions_removed()); 1242 { 1243 MutexLockerEx x(ParGCRareEvent_lock, Mutex::_no_safepoint_check_flag); 1244 _g1h->decrement_summary_bytes(g1_note_end.freed_bytes()); 1245 1246 // If we iterate over the global cleanup list at the end of 1247 // cleanup to do this printing we will not guarantee to only 1248 // generate output for the newly-reclaimed regions (the list 1249 // might not be empty at the beginning of cleanup; we might 1250 // still be working on its previous contents). So we do the 1251 // printing here, before we append the new regions to the global 1252 // cleanup list. 1253 1254 G1HRPrinter* hr_printer = _g1h->hr_printer(); 1255 if (hr_printer->is_active()) { 1256 FreeRegionListIterator iter(&local_cleanup_list); 1257 while (iter.more_available()) { 1258 HeapRegion* hr = iter.get_next(); 1259 hr_printer->cleanup(hr); 1260 } 1261 } 1262 1263 _cleanup_list->add_ordered(&local_cleanup_list); 1264 assert(local_cleanup_list.is_empty(), "post-condition"); 1265 1266 HeapRegionRemSet::finish_cleanup_task(&hrrs_cleanup_task); 1267 } 1268 } 1269 }; 1270 1271 void G1ConcurrentMark::cleanup() { 1272 // world is stopped at this checkpoint 1273 assert(SafepointSynchronize::is_at_safepoint(), 1274 "world should be stopped"); 1275 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1276 1277 // If a full collection has happened, we shouldn't do this. 1278 if (has_aborted()) { 1279 g1h->collector_state()->set_mark_in_progress(false); // So bitmap clearing isn't confused 1280 return; 1281 } 1282 1283 g1h->verifier()->verify_region_sets_optional(); 1284 1285 if (VerifyDuringGC) { 1286 HandleMark hm; // handle scope 1287 g1h->prepare_for_verify(); 1288 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (before)"); 1289 } 1290 g1h->verifier()->check_bitmaps("Cleanup Start"); 1291 1292 G1Policy* g1p = g1h->g1_policy(); 1293 g1p->record_concurrent_mark_cleanup_start(); 1294 1295 double start = os::elapsedTime(); 1296 1297 HeapRegionRemSet::reset_for_cleanup_tasks(); 1298 1299 { 1300 GCTraceTime(Debug, gc)("Finalize Live Data"); 1301 finalize_live_data(); 1302 } 1303 1304 if (VerifyDuringGC) { 1305 GCTraceTime(Debug, gc)("Verify Live Data"); 1306 verify_live_data(); 1307 } 1308 1309 g1h->collector_state()->set_mark_in_progress(false); 1310 1311 double count_end = os::elapsedTime(); 1312 double this_final_counting_time = (count_end - start); 1313 _total_counting_time += this_final_counting_time; 1314 1315 if (log_is_enabled(Trace, gc, liveness)) { 1316 G1PrintRegionLivenessInfoClosure cl("Post-Marking"); 1317 _g1h->heap_region_iterate(&cl); 1318 } 1319 1320 // Install newly created mark bitMap as "prev". 1321 swapMarkBitMaps(); 1322 1323 g1h->reset_gc_time_stamp(); 1324 1325 uint n_workers = _g1h->workers()->active_workers(); 1326 1327 // Note end of marking in all heap regions. 1328 G1ParNoteEndTask g1_par_note_end_task(g1h, &_cleanup_list, n_workers); 1329 g1h->workers()->run_task(&g1_par_note_end_task); 1330 g1h->check_gc_time_stamps(); 1331 1332 if (!cleanup_list_is_empty()) { 1333 // The cleanup list is not empty, so we'll have to process it 1334 // concurrently. Notify anyone else that might be wanting free 1335 // regions that there will be more free regions coming soon. 1336 g1h->set_free_regions_coming(); 1337 } 1338 1339 // call below, since it affects the metric by which we sort the heap 1340 // regions. 1341 if (G1ScrubRemSets) { 1342 double rs_scrub_start = os::elapsedTime(); 1343 g1h->scrub_rem_set(); 1344 _total_rs_scrub_time += (os::elapsedTime() - rs_scrub_start); 1345 } 1346 1347 // this will also free any regions totally full of garbage objects, 1348 // and sort the regions. 1349 g1h->g1_policy()->record_concurrent_mark_cleanup_end(); 1350 1351 // Statistics. 1352 double end = os::elapsedTime(); 1353 _cleanup_times.add((end - start) * 1000.0); 1354 1355 // Clean up will have freed any regions completely full of garbage. 1356 // Update the soft reference policy with the new heap occupancy. 1357 Universe::update_heap_info_at_gc(); 1358 1359 if (VerifyDuringGC) { 1360 HandleMark hm; // handle scope 1361 g1h->prepare_for_verify(); 1362 Universe::verify(VerifyOption_G1UsePrevMarking, "During GC (after)"); 1363 } 1364 1365 g1h->verifier()->check_bitmaps("Cleanup End"); 1366 1367 g1h->verifier()->verify_region_sets_optional(); 1368 1369 // We need to make this be a "collection" so any collection pause that 1370 // races with it goes around and waits for completeCleanup to finish. 1371 g1h->increment_total_collections(); 1372 1373 // Clean out dead classes and update Metaspace sizes. 1374 if (ClassUnloadingWithConcurrentMark) { 1375 ClassLoaderDataGraph::purge(); 1376 } 1377 MetaspaceGC::compute_new_size(); 1378 1379 // We reclaimed old regions so we should calculate the sizes to make 1380 // sure we update the old gen/space data. 1381 g1h->g1mm()->update_sizes(); 1382 g1h->allocation_context_stats().update_after_mark(); 1383 } 1384 1385 void G1ConcurrentMark::complete_cleanup() { 1386 if (has_aborted()) return; 1387 1388 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1389 1390 _cleanup_list.verify_optional(); 1391 FreeRegionList tmp_free_list("Tmp Free List"); 1392 1393 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : " 1394 "cleanup list has %u entries", 1395 _cleanup_list.length()); 1396 1397 // No one else should be accessing the _cleanup_list at this point, 1398 // so it is not necessary to take any locks 1399 while (!_cleanup_list.is_empty()) { 1400 HeapRegion* hr = _cleanup_list.remove_region(true /* from_head */); 1401 assert(hr != NULL, "Got NULL from a non-empty list"); 1402 hr->par_clear(); 1403 tmp_free_list.add_ordered(hr); 1404 1405 // Instead of adding one region at a time to the secondary_free_list, 1406 // we accumulate them in the local list and move them a few at a 1407 // time. This also cuts down on the number of notify_all() calls 1408 // we do during this process. We'll also append the local list when 1409 // _cleanup_list is empty (which means we just removed the last 1410 // region from the _cleanup_list). 1411 if ((tmp_free_list.length() % G1SecondaryFreeListAppendLength == 0) || 1412 _cleanup_list.is_empty()) { 1413 log_develop_trace(gc, freelist)("G1ConcRegionFreeing [complete cleanup] : " 1414 "appending %u entries to the secondary_free_list, " 1415 "cleanup list still has %u entries", 1416 tmp_free_list.length(), 1417 _cleanup_list.length()); 1418 1419 { 1420 MutexLockerEx x(SecondaryFreeList_lock, Mutex::_no_safepoint_check_flag); 1421 g1h->secondary_free_list_add(&tmp_free_list); 1422 SecondaryFreeList_lock->notify_all(); 1423 } 1424 #ifndef PRODUCT 1425 if (G1StressConcRegionFreeing) { 1426 for (uintx i = 0; i < G1StressConcRegionFreeingDelayMillis; ++i) { 1427 os::sleep(Thread::current(), (jlong) 1, false); 1428 } 1429 } 1430 #endif 1431 } 1432 } 1433 assert(tmp_free_list.is_empty(), "post-condition"); 1434 } 1435 1436 // Supporting Object and Oop closures for reference discovery 1437 // and processing in during marking 1438 1439 bool G1CMIsAliveClosure::do_object_b(oop obj) { 1440 HeapWord* addr = (HeapWord*)obj; 1441 return addr != NULL && 1442 (!_g1->is_in_g1_reserved(addr) || !_g1->is_obj_ill(obj)); 1443 } 1444 1445 // 'Keep Alive' oop closure used by both serial parallel reference processing. 1446 // Uses the G1CMTask associated with a worker thread (for serial reference 1447 // processing the G1CMTask for worker 0 is used) to preserve (mark) and 1448 // trace referent objects. 1449 // 1450 // Using the G1CMTask and embedded local queues avoids having the worker 1451 // threads operating on the global mark stack. This reduces the risk 1452 // of overflowing the stack - which we would rather avoid at this late 1453 // state. Also using the tasks' local queues removes the potential 1454 // of the workers interfering with each other that could occur if 1455 // operating on the global stack. 1456 1457 class G1CMKeepAliveAndDrainClosure: public OopClosure { 1458 G1ConcurrentMark* _cm; 1459 G1CMTask* _task; 1460 int _ref_counter_limit; 1461 int _ref_counter; 1462 bool _is_serial; 1463 public: 1464 G1CMKeepAliveAndDrainClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) : 1465 _cm(cm), _task(task), _is_serial(is_serial), 1466 _ref_counter_limit(G1RefProcDrainInterval) { 1467 assert(_ref_counter_limit > 0, "sanity"); 1468 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 1469 _ref_counter = _ref_counter_limit; 1470 } 1471 1472 virtual void do_oop(narrowOop* p) { do_oop_work(p); } 1473 virtual void do_oop( oop* p) { do_oop_work(p); } 1474 1475 template <class T> void do_oop_work(T* p) { 1476 if (!_cm->has_overflown()) { 1477 oop obj = oopDesc::load_decode_heap_oop(p); 1478 _task->deal_with_reference(obj); 1479 _ref_counter--; 1480 1481 if (_ref_counter == 0) { 1482 // We have dealt with _ref_counter_limit references, pushing them 1483 // and objects reachable from them on to the local stack (and 1484 // possibly the global stack). Call G1CMTask::do_marking_step() to 1485 // process these entries. 1486 // 1487 // We call G1CMTask::do_marking_step() in a loop, which we'll exit if 1488 // there's nothing more to do (i.e. we're done with the entries that 1489 // were pushed as a result of the G1CMTask::deal_with_reference() calls 1490 // above) or we overflow. 1491 // 1492 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted() 1493 // flag while there may still be some work to do. (See the comment at 1494 // the beginning of G1CMTask::do_marking_step() for those conditions - 1495 // one of which is reaching the specified time target.) It is only 1496 // when G1CMTask::do_marking_step() returns without setting the 1497 // has_aborted() flag that the marking step has completed. 1498 do { 1499 double mark_step_duration_ms = G1ConcMarkStepDurationMillis; 1500 _task->do_marking_step(mark_step_duration_ms, 1501 false /* do_termination */, 1502 _is_serial); 1503 } while (_task->has_aborted() && !_cm->has_overflown()); 1504 _ref_counter = _ref_counter_limit; 1505 } 1506 } 1507 } 1508 }; 1509 1510 // 'Drain' oop closure used by both serial and parallel reference processing. 1511 // Uses the G1CMTask associated with a given worker thread (for serial 1512 // reference processing the G1CMtask for worker 0 is used). Calls the 1513 // do_marking_step routine, with an unbelievably large timeout value, 1514 // to drain the marking data structures of the remaining entries 1515 // added by the 'keep alive' oop closure above. 1516 1517 class G1CMDrainMarkingStackClosure: public VoidClosure { 1518 G1ConcurrentMark* _cm; 1519 G1CMTask* _task; 1520 bool _is_serial; 1521 public: 1522 G1CMDrainMarkingStackClosure(G1ConcurrentMark* cm, G1CMTask* task, bool is_serial) : 1523 _cm(cm), _task(task), _is_serial(is_serial) { 1524 assert(!_is_serial || _task->worker_id() == 0, "only task 0 for serial code"); 1525 } 1526 1527 void do_void() { 1528 do { 1529 // We call G1CMTask::do_marking_step() to completely drain the local 1530 // and global marking stacks of entries pushed by the 'keep alive' 1531 // oop closure (an instance of G1CMKeepAliveAndDrainClosure above). 1532 // 1533 // G1CMTask::do_marking_step() is called in a loop, which we'll exit 1534 // if there's nothing more to do (i.e. we've completely drained the 1535 // entries that were pushed as a a result of applying the 'keep alive' 1536 // closure to the entries on the discovered ref lists) or we overflow 1537 // the global marking stack. 1538 // 1539 // Note: G1CMTask::do_marking_step() can set the G1CMTask::has_aborted() 1540 // flag while there may still be some work to do. (See the comment at 1541 // the beginning of G1CMTask::do_marking_step() for those conditions - 1542 // one of which is reaching the specified time target.) It is only 1543 // when G1CMTask::do_marking_step() returns without setting the 1544 // has_aborted() flag that the marking step has completed. 1545 1546 _task->do_marking_step(1000000000.0 /* something very large */, 1547 true /* do_termination */, 1548 _is_serial); 1549 } while (_task->has_aborted() && !_cm->has_overflown()); 1550 } 1551 }; 1552 1553 // Implementation of AbstractRefProcTaskExecutor for parallel 1554 // reference processing at the end of G1 concurrent marking 1555 1556 class G1CMRefProcTaskExecutor: public AbstractRefProcTaskExecutor { 1557 private: 1558 G1CollectedHeap* _g1h; 1559 G1ConcurrentMark* _cm; 1560 WorkGang* _workers; 1561 uint _active_workers; 1562 1563 public: 1564 G1CMRefProcTaskExecutor(G1CollectedHeap* g1h, 1565 G1ConcurrentMark* cm, 1566 WorkGang* workers, 1567 uint n_workers) : 1568 _g1h(g1h), _cm(cm), 1569 _workers(workers), _active_workers(n_workers) { } 1570 1571 // Executes the given task using concurrent marking worker threads. 1572 virtual void execute(ProcessTask& task); 1573 virtual void execute(EnqueueTask& task); 1574 }; 1575 1576 class G1CMRefProcTaskProxy: public AbstractGangTask { 1577 typedef AbstractRefProcTaskExecutor::ProcessTask ProcessTask; 1578 ProcessTask& _proc_task; 1579 G1CollectedHeap* _g1h; 1580 G1ConcurrentMark* _cm; 1581 1582 public: 1583 G1CMRefProcTaskProxy(ProcessTask& proc_task, 1584 G1CollectedHeap* g1h, 1585 G1ConcurrentMark* cm) : 1586 AbstractGangTask("Process reference objects in parallel"), 1587 _proc_task(proc_task), _g1h(g1h), _cm(cm) { 1588 ReferenceProcessor* rp = _g1h->ref_processor_cm(); 1589 assert(rp->processing_is_mt(), "shouldn't be here otherwise"); 1590 } 1591 1592 virtual void work(uint worker_id) { 1593 ResourceMark rm; 1594 HandleMark hm; 1595 G1CMTask* task = _cm->task(worker_id); 1596 G1CMIsAliveClosure g1_is_alive(_g1h); 1597 G1CMKeepAliveAndDrainClosure g1_par_keep_alive(_cm, task, false /* is_serial */); 1598 G1CMDrainMarkingStackClosure g1_par_drain(_cm, task, false /* is_serial */); 1599 1600 _proc_task.work(worker_id, g1_is_alive, g1_par_keep_alive, g1_par_drain); 1601 } 1602 }; 1603 1604 void G1CMRefProcTaskExecutor::execute(ProcessTask& proc_task) { 1605 assert(_workers != NULL, "Need parallel worker threads."); 1606 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 1607 1608 G1CMRefProcTaskProxy proc_task_proxy(proc_task, _g1h, _cm); 1609 1610 // We need to reset the concurrency level before each 1611 // proxy task execution, so that the termination protocol 1612 // and overflow handling in G1CMTask::do_marking_step() knows 1613 // how many workers to wait for. 1614 _cm->set_concurrency(_active_workers); 1615 _workers->run_task(&proc_task_proxy); 1616 } 1617 1618 class G1CMRefEnqueueTaskProxy: public AbstractGangTask { 1619 typedef AbstractRefProcTaskExecutor::EnqueueTask EnqueueTask; 1620 EnqueueTask& _enq_task; 1621 1622 public: 1623 G1CMRefEnqueueTaskProxy(EnqueueTask& enq_task) : 1624 AbstractGangTask("Enqueue reference objects in parallel"), 1625 _enq_task(enq_task) { } 1626 1627 virtual void work(uint worker_id) { 1628 _enq_task.work(worker_id); 1629 } 1630 }; 1631 1632 void G1CMRefProcTaskExecutor::execute(EnqueueTask& enq_task) { 1633 assert(_workers != NULL, "Need parallel worker threads."); 1634 assert(_g1h->ref_processor_cm()->processing_is_mt(), "processing is not MT"); 1635 1636 G1CMRefEnqueueTaskProxy enq_task_proxy(enq_task); 1637 1638 // Not strictly necessary but... 1639 // 1640 // We need to reset the concurrency level before each 1641 // proxy task execution, so that the termination protocol 1642 // and overflow handling in G1CMTask::do_marking_step() knows 1643 // how many workers to wait for. 1644 _cm->set_concurrency(_active_workers); 1645 _workers->run_task(&enq_task_proxy); 1646 } 1647 1648 void G1ConcurrentMark::weakRefsWork(bool clear_all_soft_refs) { 1649 if (has_overflown()) { 1650 // Skip processing the discovered references if we have 1651 // overflown the global marking stack. Reference objects 1652 // only get discovered once so it is OK to not 1653 // de-populate the discovered reference lists. We could have, 1654 // but the only benefit would be that, when marking restarts, 1655 // less reference objects are discovered. 1656 return; 1657 } 1658 1659 ResourceMark rm; 1660 HandleMark hm; 1661 1662 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1663 1664 // Is alive closure. 1665 G1CMIsAliveClosure g1_is_alive(g1h); 1666 1667 // Inner scope to exclude the cleaning of the string and symbol 1668 // tables from the displayed time. 1669 { 1670 GCTraceTime(Debug, gc, phases) trace("Reference Processing", _gc_timer_cm); 1671 1672 ReferenceProcessor* rp = g1h->ref_processor_cm(); 1673 1674 // See the comment in G1CollectedHeap::ref_processing_init() 1675 // about how reference processing currently works in G1. 1676 1677 // Set the soft reference policy 1678 rp->setup_policy(clear_all_soft_refs); 1679 assert(_global_mark_stack.is_empty(), "mark stack should be empty"); 1680 1681 // Instances of the 'Keep Alive' and 'Complete GC' closures used 1682 // in serial reference processing. Note these closures are also 1683 // used for serially processing (by the the current thread) the 1684 // JNI references during parallel reference processing. 1685 // 1686 // These closures do not need to synchronize with the worker 1687 // threads involved in parallel reference processing as these 1688 // instances are executed serially by the current thread (e.g. 1689 // reference processing is not multi-threaded and is thus 1690 // performed by the current thread instead of a gang worker). 1691 // 1692 // The gang tasks involved in parallel reference processing create 1693 // their own instances of these closures, which do their own 1694 // synchronization among themselves. 1695 G1CMKeepAliveAndDrainClosure g1_keep_alive(this, task(0), true /* is_serial */); 1696 G1CMDrainMarkingStackClosure g1_drain_mark_stack(this, task(0), true /* is_serial */); 1697 1698 // We need at least one active thread. If reference processing 1699 // is not multi-threaded we use the current (VMThread) thread, 1700 // otherwise we use the work gang from the G1CollectedHeap and 1701 // we utilize all the worker threads we can. 1702 bool processing_is_mt = rp->processing_is_mt(); 1703 uint active_workers = (processing_is_mt ? g1h->workers()->active_workers() : 1U); 1704 active_workers = MAX2(MIN2(active_workers, _max_worker_id), 1U); 1705 1706 // Parallel processing task executor. 1707 G1CMRefProcTaskExecutor par_task_executor(g1h, this, 1708 g1h->workers(), active_workers); 1709 AbstractRefProcTaskExecutor* executor = (processing_is_mt ? &par_task_executor : NULL); 1710 1711 // Set the concurrency level. The phase was already set prior to 1712 // executing the remark task. 1713 set_concurrency(active_workers); 1714 1715 // Set the degree of MT processing here. If the discovery was done MT, 1716 // the number of threads involved during discovery could differ from 1717 // the number of active workers. This is OK as long as the discovered 1718 // Reference lists are balanced (see balance_all_queues() and balance_queues()). 1719 rp->set_active_mt_degree(active_workers); 1720 1721 ReferenceProcessorPhaseTimes pt(_gc_timer_cm, rp->num_q()); 1722 1723 // Process the weak references. 1724 const ReferenceProcessorStats& stats = 1725 rp->process_discovered_references(&g1_is_alive, 1726 &g1_keep_alive, 1727 &g1_drain_mark_stack, 1728 executor, 1729 &pt); 1730 _gc_tracer_cm->report_gc_reference_stats(stats); 1731 pt.print_all_references(); 1732 1733 // The do_oop work routines of the keep_alive and drain_marking_stack 1734 // oop closures will set the has_overflown flag if we overflow the 1735 // global marking stack. 1736 1737 assert(has_overflown() || _global_mark_stack.is_empty(), 1738 "Mark stack should be empty (unless it has overflown)"); 1739 1740 assert(rp->num_q() == active_workers, "why not"); 1741 1742 rp->enqueue_discovered_references(executor, &pt); 1743 1744 rp->verify_no_references_recorded(); 1745 1746 pt.print_enqueue_phase(); 1747 1748 assert(!rp->discovery_enabled(), "Post condition"); 1749 } 1750 1751 if (has_overflown()) { 1752 // We can not trust g1_is_alive if the marking stack overflowed 1753 return; 1754 } 1755 1756 assert(_global_mark_stack.is_empty(), "Marking should have completed"); 1757 1758 // Unload Klasses, String, Symbols, Code Cache, etc. 1759 if (ClassUnloadingWithConcurrentMark) { 1760 GCTraceTime(Debug, gc, phases) debug("Class Unloading", _gc_timer_cm); 1761 bool purged_classes = SystemDictionary::do_unloading(&g1_is_alive, _gc_timer_cm, false /* Defer cleaning */); 1762 g1h->complete_cleaning(&g1_is_alive, purged_classes); 1763 } else { 1764 GCTraceTime(Debug, gc, phases) debug("Cleanup", _gc_timer_cm); 1765 // No need to clean string table and symbol table as they are treated as strong roots when 1766 // class unloading is disabled. 1767 g1h->partial_cleaning(&g1_is_alive, false, false, G1StringDedup::is_enabled()); 1768 1769 } 1770 } 1771 1772 void G1ConcurrentMark::swapMarkBitMaps() { 1773 G1CMBitMapRO* temp = _prevMarkBitMap; 1774 _prevMarkBitMap = (G1CMBitMapRO*)_nextMarkBitMap; 1775 _nextMarkBitMap = (G1CMBitMap*) temp; 1776 } 1777 1778 // Closure for marking entries in SATB buffers. 1779 class G1CMSATBBufferClosure : public SATBBufferClosure { 1780 private: 1781 G1CMTask* _task; 1782 G1CollectedHeap* _g1h; 1783 1784 // This is very similar to G1CMTask::deal_with_reference, but with 1785 // more relaxed requirements for the argument, so this must be more 1786 // circumspect about treating the argument as an object. 1787 void do_entry(void* entry) const { 1788 _task->increment_refs_reached(); 1789 HeapRegion* hr = _g1h->heap_region_containing(entry); 1790 if (entry < hr->next_top_at_mark_start()) { 1791 // Until we get here, we don't know whether entry refers to a valid 1792 // object; it could instead have been a stale reference. 1793 oop obj = static_cast<oop>(entry); 1794 assert(obj->is_oop(true /* ignore mark word */), 1795 "Invalid oop in SATB buffer: " PTR_FORMAT, p2i(obj)); 1796 _task->make_reference_grey(obj); 1797 } 1798 } 1799 1800 public: 1801 G1CMSATBBufferClosure(G1CMTask* task, G1CollectedHeap* g1h) 1802 : _task(task), _g1h(g1h) { } 1803 1804 virtual void do_buffer(void** buffer, size_t size) { 1805 for (size_t i = 0; i < size; ++i) { 1806 do_entry(buffer[i]); 1807 } 1808 } 1809 }; 1810 1811 class G1RemarkThreadsClosure : public ThreadClosure { 1812 G1CMSATBBufferClosure _cm_satb_cl; 1813 G1CMOopClosure _cm_cl; 1814 MarkingCodeBlobClosure _code_cl; 1815 int _thread_parity; 1816 1817 public: 1818 G1RemarkThreadsClosure(G1CollectedHeap* g1h, G1CMTask* task) : 1819 _cm_satb_cl(task, g1h), 1820 _cm_cl(g1h, g1h->concurrent_mark(), task), 1821 _code_cl(&_cm_cl, !CodeBlobToOopClosure::FixRelocations), 1822 _thread_parity(Threads::thread_claim_parity()) {} 1823 1824 void do_thread(Thread* thread) { 1825 if (thread->is_Java_thread()) { 1826 if (thread->claim_oops_do(true, _thread_parity)) { 1827 JavaThread* jt = (JavaThread*)thread; 1828 1829 // In theory it should not be neccessary to explicitly walk the nmethods to find roots for concurrent marking 1830 // however the liveness of oops reachable from nmethods have very complex lifecycles: 1831 // * Alive if on the stack of an executing method 1832 // * Weakly reachable otherwise 1833 // Some objects reachable from nmethods, such as the class loader (or klass_holder) of the receiver should be 1834 // live by the SATB invariant but other oops recorded in nmethods may behave differently. 1835 jt->nmethods_do(&_code_cl); 1836 1837 jt->satb_mark_queue().apply_closure_and_empty(&_cm_satb_cl); 1838 } 1839 } else if (thread->is_VM_thread()) { 1840 if (thread->claim_oops_do(true, _thread_parity)) { 1841 JavaThread::satb_mark_queue_set().shared_satb_queue()->apply_closure_and_empty(&_cm_satb_cl); 1842 } 1843 } 1844 } 1845 }; 1846 1847 class G1CMRemarkTask: public AbstractGangTask { 1848 private: 1849 G1ConcurrentMark* _cm; 1850 public: 1851 void work(uint worker_id) { 1852 // Since all available tasks are actually started, we should 1853 // only proceed if we're supposed to be active. 1854 if (worker_id < _cm->active_tasks()) { 1855 G1CMTask* task = _cm->task(worker_id); 1856 task->record_start_time(); 1857 { 1858 ResourceMark rm; 1859 HandleMark hm; 1860 1861 G1RemarkThreadsClosure threads_f(G1CollectedHeap::heap(), task); 1862 Threads::threads_do(&threads_f); 1863 } 1864 1865 do { 1866 task->do_marking_step(1000000000.0 /* something very large */, 1867 true /* do_termination */, 1868 false /* is_serial */); 1869 } while (task->has_aborted() && !_cm->has_overflown()); 1870 // If we overflow, then we do not want to restart. We instead 1871 // want to abort remark and do concurrent marking again. 1872 task->record_end_time(); 1873 } 1874 } 1875 1876 G1CMRemarkTask(G1ConcurrentMark* cm, uint active_workers) : 1877 AbstractGangTask("Par Remark"), _cm(cm) { 1878 _cm->terminator()->reset_for_reuse(active_workers); 1879 } 1880 }; 1881 1882 void G1ConcurrentMark::checkpointRootsFinalWork() { 1883 ResourceMark rm; 1884 HandleMark hm; 1885 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 1886 1887 GCTraceTime(Debug, gc, phases) trace("Finalize Marking", _gc_timer_cm); 1888 1889 g1h->ensure_parsability(false); 1890 1891 // this is remark, so we'll use up all active threads 1892 uint active_workers = g1h->workers()->active_workers(); 1893 set_concurrency_and_phase(active_workers, false /* concurrent */); 1894 // Leave _parallel_marking_threads at it's 1895 // value originally calculated in the G1ConcurrentMark 1896 // constructor and pass values of the active workers 1897 // through the gang in the task. 1898 1899 { 1900 StrongRootsScope srs(active_workers); 1901 1902 G1CMRemarkTask remarkTask(this, active_workers); 1903 // We will start all available threads, even if we decide that the 1904 // active_workers will be fewer. The extra ones will just bail out 1905 // immediately. 1906 g1h->workers()->run_task(&remarkTask); 1907 } 1908 1909 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 1910 guarantee(has_overflown() || 1911 satb_mq_set.completed_buffers_num() == 0, 1912 "Invariant: has_overflown = %s, num buffers = " SIZE_FORMAT, 1913 BOOL_TO_STR(has_overflown()), 1914 satb_mq_set.completed_buffers_num()); 1915 1916 print_stats(); 1917 } 1918 1919 void G1ConcurrentMark::clearRangePrevBitmap(MemRegion mr) { 1920 // Note we are overriding the read-only view of the prev map here, via 1921 // the cast. 1922 ((G1CMBitMap*)_prevMarkBitMap)->clear_range(mr); 1923 } 1924 1925 HeapRegion* 1926 G1ConcurrentMark::claim_region(uint worker_id) { 1927 // "checkpoint" the finger 1928 HeapWord* finger = _finger; 1929 1930 // _heap_end will not change underneath our feet; it only changes at 1931 // yield points. 1932 while (finger < _heap_end) { 1933 assert(_g1h->is_in_g1_reserved(finger), "invariant"); 1934 1935 HeapRegion* curr_region = _g1h->heap_region_containing(finger); 1936 // Make sure that the reads below do not float before loading curr_region. 1937 OrderAccess::loadload(); 1938 // Above heap_region_containing may return NULL as we always scan claim 1939 // until the end of the heap. In this case, just jump to the next region. 1940 HeapWord* end = curr_region != NULL ? curr_region->end() : finger + HeapRegion::GrainWords; 1941 1942 // Is the gap between reading the finger and doing the CAS too long? 1943 HeapWord* res = (HeapWord*) Atomic::cmpxchg_ptr(end, &_finger, finger); 1944 if (res == finger && curr_region != NULL) { 1945 // we succeeded 1946 HeapWord* bottom = curr_region->bottom(); 1947 HeapWord* limit = curr_region->next_top_at_mark_start(); 1948 1949 // notice that _finger == end cannot be guaranteed here since, 1950 // someone else might have moved the finger even further 1951 assert(_finger >= end, "the finger should have moved forward"); 1952 1953 if (limit > bottom) { 1954 return curr_region; 1955 } else { 1956 assert(limit == bottom, 1957 "the region limit should be at bottom"); 1958 // we return NULL and the caller should try calling 1959 // claim_region() again. 1960 return NULL; 1961 } 1962 } else { 1963 assert(_finger > finger, "the finger should have moved forward"); 1964 // read it again 1965 finger = _finger; 1966 } 1967 } 1968 1969 return NULL; 1970 } 1971 1972 #ifndef PRODUCT 1973 class VerifyNoCSetOops VALUE_OBJ_CLASS_SPEC { 1974 private: 1975 G1CollectedHeap* _g1h; 1976 const char* _phase; 1977 int _info; 1978 1979 public: 1980 VerifyNoCSetOops(const char* phase, int info = -1) : 1981 _g1h(G1CollectedHeap::heap()), 1982 _phase(phase), 1983 _info(info) 1984 { } 1985 1986 void operator()(G1TaskQueueEntry task_entry) const { 1987 if (task_entry.is_array_slice()) { 1988 guarantee(_g1h->is_in_reserved(task_entry.slice()), "Slice " PTR_FORMAT " must be in heap.", p2i(task_entry.slice())); 1989 return; 1990 } 1991 guarantee(task_entry.obj()->is_oop(), 1992 "Non-oop " PTR_FORMAT ", phase: %s, info: %d", 1993 p2i(task_entry.obj()), _phase, _info); 1994 guarantee(!_g1h->is_in_cset(task_entry.obj()), 1995 "obj: " PTR_FORMAT " in CSet, phase: %s, info: %d", 1996 p2i(task_entry.obj()), _phase, _info); 1997 } 1998 }; 1999 2000 void G1ConcurrentMark::verify_no_cset_oops() { 2001 assert(SafepointSynchronize::is_at_safepoint(), "should be at a safepoint"); 2002 if (!G1CollectedHeap::heap()->collector_state()->mark_in_progress()) { 2003 return; 2004 } 2005 2006 // Verify entries on the global mark stack 2007 _global_mark_stack.iterate(VerifyNoCSetOops("Stack")); 2008 2009 // Verify entries on the task queues 2010 for (uint i = 0; i < _max_worker_id; ++i) { 2011 G1CMTaskQueue* queue = _task_queues->queue(i); 2012 queue->iterate(VerifyNoCSetOops("Queue", i)); 2013 } 2014 2015 // Verify the global finger 2016 HeapWord* global_finger = finger(); 2017 if (global_finger != NULL && global_finger < _heap_end) { 2018 // Since we always iterate over all regions, we might get a NULL HeapRegion 2019 // here. 2020 HeapRegion* global_hr = _g1h->heap_region_containing(global_finger); 2021 guarantee(global_hr == NULL || global_finger == global_hr->bottom(), 2022 "global finger: " PTR_FORMAT " region: " HR_FORMAT, 2023 p2i(global_finger), HR_FORMAT_PARAMS(global_hr)); 2024 } 2025 2026 // Verify the task fingers 2027 assert(parallel_marking_threads() <= _max_worker_id, "sanity"); 2028 for (uint i = 0; i < parallel_marking_threads(); ++i) { 2029 G1CMTask* task = _tasks[i]; 2030 HeapWord* task_finger = task->finger(); 2031 if (task_finger != NULL && task_finger < _heap_end) { 2032 // See above note on the global finger verification. 2033 HeapRegion* task_hr = _g1h->heap_region_containing(task_finger); 2034 guarantee(task_hr == NULL || task_finger == task_hr->bottom() || 2035 !task_hr->in_collection_set(), 2036 "task finger: " PTR_FORMAT " region: " HR_FORMAT, 2037 p2i(task_finger), HR_FORMAT_PARAMS(task_hr)); 2038 } 2039 } 2040 } 2041 #endif // PRODUCT 2042 void G1ConcurrentMark::create_live_data() { 2043 _g1h->g1_rem_set()->create_card_live_data(_parallel_workers, _nextMarkBitMap); 2044 } 2045 2046 void G1ConcurrentMark::finalize_live_data() { 2047 _g1h->g1_rem_set()->finalize_card_live_data(_g1h->workers(), _nextMarkBitMap); 2048 } 2049 2050 void G1ConcurrentMark::verify_live_data() { 2051 _g1h->g1_rem_set()->verify_card_live_data(_g1h->workers(), _nextMarkBitMap); 2052 } 2053 2054 void G1ConcurrentMark::clear_live_data(WorkGang* workers) { 2055 _g1h->g1_rem_set()->clear_card_live_data(workers); 2056 } 2057 2058 #ifdef ASSERT 2059 void G1ConcurrentMark::verify_live_data_clear() { 2060 _g1h->g1_rem_set()->verify_card_live_data_is_clear(); 2061 } 2062 #endif 2063 2064 void G1ConcurrentMark::print_stats() { 2065 if (!log_is_enabled(Debug, gc, stats)) { 2066 return; 2067 } 2068 log_debug(gc, stats)("---------------------------------------------------------------------"); 2069 for (size_t i = 0; i < _active_tasks; ++i) { 2070 _tasks[i]->print_stats(); 2071 log_debug(gc, stats)("---------------------------------------------------------------------"); 2072 } 2073 } 2074 2075 void G1ConcurrentMark::abort() { 2076 if (!cmThread()->during_cycle() || _has_aborted) { 2077 // We haven't started a concurrent cycle or we have already aborted it. No need to do anything. 2078 return; 2079 } 2080 2081 // Clear all marks in the next bitmap for the next marking cycle. This will allow us to skip the next 2082 // concurrent bitmap clearing. 2083 { 2084 GCTraceTime(Debug, gc)("Clear Next Bitmap"); 2085 clear_bitmap(_nextMarkBitMap, _g1h->workers(), false); 2086 } 2087 // Note we cannot clear the previous marking bitmap here 2088 // since VerifyDuringGC verifies the objects marked during 2089 // a full GC against the previous bitmap. 2090 2091 { 2092 GCTraceTime(Debug, gc)("Clear Live Data"); 2093 clear_live_data(_g1h->workers()); 2094 } 2095 DEBUG_ONLY({ 2096 GCTraceTime(Debug, gc)("Verify Live Data Clear"); 2097 verify_live_data_clear(); 2098 }) 2099 // Empty mark stack 2100 reset_marking_state(); 2101 for (uint i = 0; i < _max_worker_id; ++i) { 2102 _tasks[i]->clear_region_fields(); 2103 } 2104 _first_overflow_barrier_sync.abort(); 2105 _second_overflow_barrier_sync.abort(); 2106 _has_aborted = true; 2107 2108 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 2109 satb_mq_set.abandon_partial_marking(); 2110 // This can be called either during or outside marking, we'll read 2111 // the expected_active value from the SATB queue set. 2112 satb_mq_set.set_active_all_threads( 2113 false, /* new active value */ 2114 satb_mq_set.is_active() /* expected_active */); 2115 } 2116 2117 static void print_ms_time_info(const char* prefix, const char* name, 2118 NumberSeq& ns) { 2119 log_trace(gc, marking)("%s%5d %12s: total time = %8.2f s (avg = %8.2f ms).", 2120 prefix, ns.num(), name, ns.sum()/1000.0, ns.avg()); 2121 if (ns.num() > 0) { 2122 log_trace(gc, marking)("%s [std. dev = %8.2f ms, max = %8.2f ms]", 2123 prefix, ns.sd(), ns.maximum()); 2124 } 2125 } 2126 2127 void G1ConcurrentMark::print_summary_info() { 2128 Log(gc, marking) log; 2129 if (!log.is_trace()) { 2130 return; 2131 } 2132 2133 log.trace(" Concurrent marking:"); 2134 print_ms_time_info(" ", "init marks", _init_times); 2135 print_ms_time_info(" ", "remarks", _remark_times); 2136 { 2137 print_ms_time_info(" ", "final marks", _remark_mark_times); 2138 print_ms_time_info(" ", "weak refs", _remark_weak_ref_times); 2139 2140 } 2141 print_ms_time_info(" ", "cleanups", _cleanup_times); 2142 log.trace(" Finalize live data total time = %8.2f s (avg = %8.2f ms).", 2143 _total_counting_time, (_cleanup_times.num() > 0 ? _total_counting_time * 1000.0 / (double)_cleanup_times.num() : 0.0)); 2144 if (G1ScrubRemSets) { 2145 log.trace(" RS scrub total time = %8.2f s (avg = %8.2f ms).", 2146 _total_rs_scrub_time, (_cleanup_times.num() > 0 ? _total_rs_scrub_time * 1000.0 / (double)_cleanup_times.num() : 0.0)); 2147 } 2148 log.trace(" Total stop_world time = %8.2f s.", 2149 (_init_times.sum() + _remark_times.sum() + _cleanup_times.sum())/1000.0); 2150 log.trace(" Total concurrent time = %8.2f s (%8.2f s marking).", 2151 cmThread()->vtime_accum(), cmThread()->vtime_mark_accum()); 2152 } 2153 2154 void G1ConcurrentMark::print_worker_threads_on(outputStream* st) const { 2155 _parallel_workers->print_worker_threads_on(st); 2156 } 2157 2158 void G1ConcurrentMark::threads_do(ThreadClosure* tc) const { 2159 _parallel_workers->threads_do(tc); 2160 } 2161 2162 void G1ConcurrentMark::print_on_error(outputStream* st) const { 2163 st->print_cr("Marking Bits (Prev, Next): (CMBitMap*) " PTR_FORMAT ", (CMBitMap*) " PTR_FORMAT, 2164 p2i(_prevMarkBitMap), p2i(_nextMarkBitMap)); 2165 _prevMarkBitMap->print_on_error(st, " Prev Bits: "); 2166 _nextMarkBitMap->print_on_error(st, " Next Bits: "); 2167 } 2168 2169 // Closure for iteration over bitmaps 2170 class G1CMBitMapClosure : public BitMapClosure { 2171 private: 2172 // the bitmap that is being iterated over 2173 G1CMBitMap* _nextMarkBitMap; 2174 G1ConcurrentMark* _cm; 2175 G1CMTask* _task; 2176 2177 public: 2178 G1CMBitMapClosure(G1CMTask *task, G1ConcurrentMark* cm, G1CMBitMap* nextMarkBitMap) : 2179 _task(task), _cm(cm), _nextMarkBitMap(nextMarkBitMap) { } 2180 2181 bool do_bit(size_t offset) { 2182 HeapWord* addr = _nextMarkBitMap->offsetToHeapWord(offset); 2183 assert(_nextMarkBitMap->isMarked(addr), "invariant"); 2184 assert( addr < _cm->finger(), "invariant"); 2185 assert(addr >= _task->finger(), "invariant"); 2186 2187 // We move that task's local finger along. 2188 _task->move_finger_to(addr); 2189 2190 _task->scan_task_entry(G1TaskQueueEntry::from_oop(oop(addr))); 2191 // we only partially drain the local queue and global stack 2192 _task->drain_local_queue(true); 2193 _task->drain_global_stack(true); 2194 2195 // if the has_aborted flag has been raised, we need to bail out of 2196 // the iteration 2197 return !_task->has_aborted(); 2198 } 2199 }; 2200 2201 static ReferenceProcessor* get_cm_oop_closure_ref_processor(G1CollectedHeap* g1h) { 2202 ReferenceProcessor* result = g1h->ref_processor_cm(); 2203 assert(result != NULL, "CM reference processor should not be NULL"); 2204 return result; 2205 } 2206 2207 G1CMOopClosure::G1CMOopClosure(G1CollectedHeap* g1h, 2208 G1ConcurrentMark* cm, 2209 G1CMTask* task) 2210 : MetadataAwareOopClosure(get_cm_oop_closure_ref_processor(g1h)), 2211 _g1h(g1h), _cm(cm), _task(task) 2212 { } 2213 2214 void G1CMTask::setup_for_region(HeapRegion* hr) { 2215 assert(hr != NULL, 2216 "claim_region() should have filtered out NULL regions"); 2217 _curr_region = hr; 2218 _finger = hr->bottom(); 2219 update_region_limit(); 2220 } 2221 2222 void G1CMTask::update_region_limit() { 2223 HeapRegion* hr = _curr_region; 2224 HeapWord* bottom = hr->bottom(); 2225 HeapWord* limit = hr->next_top_at_mark_start(); 2226 2227 if (limit == bottom) { 2228 // The region was collected underneath our feet. 2229 // We set the finger to bottom to ensure that the bitmap 2230 // iteration that will follow this will not do anything. 2231 // (this is not a condition that holds when we set the region up, 2232 // as the region is not supposed to be empty in the first place) 2233 _finger = bottom; 2234 } else if (limit >= _region_limit) { 2235 assert(limit >= _finger, "peace of mind"); 2236 } else { 2237 assert(limit < _region_limit, "only way to get here"); 2238 // This can happen under some pretty unusual circumstances. An 2239 // evacuation pause empties the region underneath our feet (NTAMS 2240 // at bottom). We then do some allocation in the region (NTAMS 2241 // stays at bottom), followed by the region being used as a GC 2242 // alloc region (NTAMS will move to top() and the objects 2243 // originally below it will be grayed). All objects now marked in 2244 // the region are explicitly grayed, if below the global finger, 2245 // and we do not need in fact to scan anything else. So, we simply 2246 // set _finger to be limit to ensure that the bitmap iteration 2247 // doesn't do anything. 2248 _finger = limit; 2249 } 2250 2251 _region_limit = limit; 2252 } 2253 2254 void G1CMTask::giveup_current_region() { 2255 assert(_curr_region != NULL, "invariant"); 2256 clear_region_fields(); 2257 } 2258 2259 void G1CMTask::clear_region_fields() { 2260 // Values for these three fields that indicate that we're not 2261 // holding on to a region. 2262 _curr_region = NULL; 2263 _finger = NULL; 2264 _region_limit = NULL; 2265 } 2266 2267 void G1CMTask::set_cm_oop_closure(G1CMOopClosure* cm_oop_closure) { 2268 if (cm_oop_closure == NULL) { 2269 assert(_cm_oop_closure != NULL, "invariant"); 2270 } else { 2271 assert(_cm_oop_closure == NULL, "invariant"); 2272 } 2273 _cm_oop_closure = cm_oop_closure; 2274 } 2275 2276 void G1CMTask::reset(G1CMBitMap* nextMarkBitMap) { 2277 guarantee(nextMarkBitMap != NULL, "invariant"); 2278 _nextMarkBitMap = nextMarkBitMap; 2279 clear_region_fields(); 2280 2281 _calls = 0; 2282 _elapsed_time_ms = 0.0; 2283 _termination_time_ms = 0.0; 2284 _termination_start_time_ms = 0.0; 2285 } 2286 2287 bool G1CMTask::should_exit_termination() { 2288 regular_clock_call(); 2289 // This is called when we are in the termination protocol. We should 2290 // quit if, for some reason, this task wants to abort or the global 2291 // stack is not empty (this means that we can get work from it). 2292 return !_cm->mark_stack_empty() || has_aborted(); 2293 } 2294 2295 void G1CMTask::reached_limit() { 2296 assert(_words_scanned >= _words_scanned_limit || 2297 _refs_reached >= _refs_reached_limit , 2298 "shouldn't have been called otherwise"); 2299 regular_clock_call(); 2300 } 2301 2302 void G1CMTask::regular_clock_call() { 2303 if (has_aborted()) return; 2304 2305 // First, we need to recalculate the words scanned and refs reached 2306 // limits for the next clock call. 2307 recalculate_limits(); 2308 2309 // During the regular clock call we do the following 2310 2311 // (1) If an overflow has been flagged, then we abort. 2312 if (_cm->has_overflown()) { 2313 set_has_aborted(); 2314 return; 2315 } 2316 2317 // If we are not concurrent (i.e. we're doing remark) we don't need 2318 // to check anything else. The other steps are only needed during 2319 // the concurrent marking phase. 2320 if (!concurrent()) return; 2321 2322 // (2) If marking has been aborted for Full GC, then we also abort. 2323 if (_cm->has_aborted()) { 2324 set_has_aborted(); 2325 return; 2326 } 2327 2328 double curr_time_ms = os::elapsedVTime() * 1000.0; 2329 2330 // (4) We check whether we should yield. If we have to, then we abort. 2331 if (SuspendibleThreadSet::should_yield()) { 2332 // We should yield. To do this we abort the task. The caller is 2333 // responsible for yielding. 2334 set_has_aborted(); 2335 return; 2336 } 2337 2338 // (5) We check whether we've reached our time quota. If we have, 2339 // then we abort. 2340 double elapsed_time_ms = curr_time_ms - _start_time_ms; 2341 if (elapsed_time_ms > _time_target_ms) { 2342 set_has_aborted(); 2343 _has_timed_out = true; 2344 return; 2345 } 2346 2347 // (6) Finally, we check whether there are enough completed STAB 2348 // buffers available for processing. If there are, we abort. 2349 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 2350 if (!_draining_satb_buffers && satb_mq_set.process_completed_buffers()) { 2351 // we do need to process SATB buffers, we'll abort and restart 2352 // the marking task to do so 2353 set_has_aborted(); 2354 return; 2355 } 2356 } 2357 2358 void G1CMTask::recalculate_limits() { 2359 _real_words_scanned_limit = _words_scanned + words_scanned_period; 2360 _words_scanned_limit = _real_words_scanned_limit; 2361 2362 _real_refs_reached_limit = _refs_reached + refs_reached_period; 2363 _refs_reached_limit = _real_refs_reached_limit; 2364 } 2365 2366 void G1CMTask::decrease_limits() { 2367 // This is called when we believe that we're going to do an infrequent 2368 // operation which will increase the per byte scanned cost (i.e. move 2369 // entries to/from the global stack). It basically tries to decrease the 2370 // scanning limit so that the clock is called earlier. 2371 2372 _words_scanned_limit = _real_words_scanned_limit - 2373 3 * words_scanned_period / 4; 2374 _refs_reached_limit = _real_refs_reached_limit - 2375 3 * refs_reached_period / 4; 2376 } 2377 2378 void G1CMTask::move_entries_to_global_stack() { 2379 // Local array where we'll store the entries that will be popped 2380 // from the local queue. 2381 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk]; 2382 2383 size_t n = 0; 2384 G1TaskQueueEntry task_entry; 2385 while (n < G1CMMarkStack::EntriesPerChunk && _task_queue->pop_local(task_entry)) { 2386 buffer[n] = task_entry; 2387 ++n; 2388 } 2389 if (n < G1CMMarkStack::EntriesPerChunk) { 2390 buffer[n] = G1TaskQueueEntry(); 2391 } 2392 2393 if (n > 0) { 2394 if (!_cm->mark_stack_push(buffer)) { 2395 set_has_aborted(); 2396 } 2397 } 2398 2399 // This operation was quite expensive, so decrease the limits. 2400 decrease_limits(); 2401 } 2402 2403 bool G1CMTask::get_entries_from_global_stack() { 2404 // Local array where we'll store the entries that will be popped 2405 // from the global stack. 2406 G1TaskQueueEntry buffer[G1CMMarkStack::EntriesPerChunk]; 2407 2408 if (!_cm->mark_stack_pop(buffer)) { 2409 return false; 2410 } 2411 2412 // We did actually pop at least one entry. 2413 for (size_t i = 0; i < G1CMMarkStack::EntriesPerChunk; ++i) { 2414 G1TaskQueueEntry task_entry = buffer[i]; 2415 if (task_entry.is_null()) { 2416 break; 2417 } 2418 assert(task_entry.is_array_slice() || task_entry.obj()->is_oop(), "Element " PTR_FORMAT " must be an array slice or oop", p2i(task_entry.obj())); 2419 bool success = _task_queue->push(task_entry); 2420 // We only call this when the local queue is empty or under a 2421 // given target limit. So, we do not expect this push to fail. 2422 assert(success, "invariant"); 2423 } 2424 2425 // This operation was quite expensive, so decrease the limits 2426 decrease_limits(); 2427 return true; 2428 } 2429 2430 void G1CMTask::drain_local_queue(bool partially) { 2431 if (has_aborted()) { 2432 return; 2433 } 2434 2435 // Decide what the target size is, depending whether we're going to 2436 // drain it partially (so that other tasks can steal if they run out 2437 // of things to do) or totally (at the very end). 2438 size_t target_size; 2439 if (partially) { 2440 target_size = MIN2((size_t)_task_queue->max_elems()/3, GCDrainStackTargetSize); 2441 } else { 2442 target_size = 0; 2443 } 2444 2445 if (_task_queue->size() > target_size) { 2446 G1TaskQueueEntry entry; 2447 bool ret = _task_queue->pop_local(entry); 2448 while (ret) { 2449 scan_task_entry(entry); 2450 if (_task_queue->size() <= target_size || has_aborted()) { 2451 ret = false; 2452 } else { 2453 ret = _task_queue->pop_local(entry); 2454 } 2455 } 2456 } 2457 } 2458 2459 void G1CMTask::drain_global_stack(bool partially) { 2460 if (has_aborted()) return; 2461 2462 // We have a policy to drain the local queue before we attempt to 2463 // drain the global stack. 2464 assert(partially || _task_queue->size() == 0, "invariant"); 2465 2466 // Decide what the target size is, depending whether we're going to 2467 // drain it partially (so that other tasks can steal if they run out 2468 // of things to do) or totally (at the very end). 2469 // Notice that when draining the global mark stack partially, due to the racyness 2470 // of the mark stack size update we might in fact drop below the target. But, 2471 // this is not a problem. 2472 // In case of total draining, we simply process until the global mark stack is 2473 // totally empty, disregarding the size counter. 2474 if (partially) { 2475 size_t const target_size = _cm->partial_mark_stack_size_target(); 2476 while (!has_aborted() && _cm->mark_stack_size() > target_size) { 2477 if (get_entries_from_global_stack()) { 2478 drain_local_queue(partially); 2479 } 2480 } 2481 } else { 2482 while (!has_aborted() && get_entries_from_global_stack()) { 2483 drain_local_queue(partially); 2484 } 2485 } 2486 } 2487 2488 // SATB Queue has several assumptions on whether to call the par or 2489 // non-par versions of the methods. this is why some of the code is 2490 // replicated. We should really get rid of the single-threaded version 2491 // of the code to simplify things. 2492 void G1CMTask::drain_satb_buffers() { 2493 if (has_aborted()) return; 2494 2495 // We set this so that the regular clock knows that we're in the 2496 // middle of draining buffers and doesn't set the abort flag when it 2497 // notices that SATB buffers are available for draining. It'd be 2498 // very counter productive if it did that. :-) 2499 _draining_satb_buffers = true; 2500 2501 G1CMSATBBufferClosure satb_cl(this, _g1h); 2502 SATBMarkQueueSet& satb_mq_set = JavaThread::satb_mark_queue_set(); 2503 2504 // This keeps claiming and applying the closure to completed buffers 2505 // until we run out of buffers or we need to abort. 2506 while (!has_aborted() && 2507 satb_mq_set.apply_closure_to_completed_buffer(&satb_cl)) { 2508 regular_clock_call(); 2509 } 2510 2511 _draining_satb_buffers = false; 2512 2513 assert(has_aborted() || 2514 concurrent() || 2515 satb_mq_set.completed_buffers_num() == 0, "invariant"); 2516 2517 // again, this was a potentially expensive operation, decrease the 2518 // limits to get the regular clock call early 2519 decrease_limits(); 2520 } 2521 2522 void G1CMTask::print_stats() { 2523 log_debug(gc, stats)("Marking Stats, task = %u, calls = %d", 2524 _worker_id, _calls); 2525 log_debug(gc, stats)(" Elapsed time = %1.2lfms, Termination time = %1.2lfms", 2526 _elapsed_time_ms, _termination_time_ms); 2527 log_debug(gc, stats)(" Step Times (cum): num = %d, avg = %1.2lfms, sd = %1.2lfms", 2528 _step_times_ms.num(), _step_times_ms.avg(), 2529 _step_times_ms.sd()); 2530 log_debug(gc, stats)(" max = %1.2lfms, total = %1.2lfms", 2531 _step_times_ms.maximum(), _step_times_ms.sum()); 2532 } 2533 2534 bool G1ConcurrentMark::try_stealing(uint worker_id, int* hash_seed, G1TaskQueueEntry& task_entry) { 2535 return _task_queues->steal(worker_id, hash_seed, task_entry); 2536 } 2537 2538 /***************************************************************************** 2539 2540 The do_marking_step(time_target_ms, ...) method is the building 2541 block of the parallel marking framework. It can be called in parallel 2542 with other invocations of do_marking_step() on different tasks 2543 (but only one per task, obviously) and concurrently with the 2544 mutator threads, or during remark, hence it eliminates the need 2545 for two versions of the code. When called during remark, it will 2546 pick up from where the task left off during the concurrent marking 2547 phase. Interestingly, tasks are also claimable during evacuation 2548 pauses too, since do_marking_step() ensures that it aborts before 2549 it needs to yield. 2550 2551 The data structures that it uses to do marking work are the 2552 following: 2553 2554 (1) Marking Bitmap. If there are gray objects that appear only 2555 on the bitmap (this happens either when dealing with an overflow 2556 or when the initial marking phase has simply marked the roots 2557 and didn't push them on the stack), then tasks claim heap 2558 regions whose bitmap they then scan to find gray objects. A 2559 global finger indicates where the end of the last claimed region 2560 is. A local finger indicates how far into the region a task has 2561 scanned. The two fingers are used to determine how to gray an 2562 object (i.e. whether simply marking it is OK, as it will be 2563 visited by a task in the future, or whether it needs to be also 2564 pushed on a stack). 2565 2566 (2) Local Queue. The local queue of the task which is accessed 2567 reasonably efficiently by the task. Other tasks can steal from 2568 it when they run out of work. Throughout the marking phase, a 2569 task attempts to keep its local queue short but not totally 2570 empty, so that entries are available for stealing by other 2571 tasks. Only when there is no more work, a task will totally 2572 drain its local queue. 2573 2574 (3) Global Mark Stack. This handles local queue overflow. During 2575 marking only sets of entries are moved between it and the local 2576 queues, as access to it requires a mutex and more fine-grain 2577 interaction with it which might cause contention. If it 2578 overflows, then the marking phase should restart and iterate 2579 over the bitmap to identify gray objects. Throughout the marking 2580 phase, tasks attempt to keep the global mark stack at a small 2581 length but not totally empty, so that entries are available for 2582 popping by other tasks. Only when there is no more work, tasks 2583 will totally drain the global mark stack. 2584 2585 (4) SATB Buffer Queue. This is where completed SATB buffers are 2586 made available. Buffers are regularly removed from this queue 2587 and scanned for roots, so that the queue doesn't get too 2588 long. During remark, all completed buffers are processed, as 2589 well as the filled in parts of any uncompleted buffers. 2590 2591 The do_marking_step() method tries to abort when the time target 2592 has been reached. There are a few other cases when the 2593 do_marking_step() method also aborts: 2594 2595 (1) When the marking phase has been aborted (after a Full GC). 2596 2597 (2) When a global overflow (on the global stack) has been 2598 triggered. Before the task aborts, it will actually sync up with 2599 the other tasks to ensure that all the marking data structures 2600 (local queues, stacks, fingers etc.) are re-initialized so that 2601 when do_marking_step() completes, the marking phase can 2602 immediately restart. 2603 2604 (3) When enough completed SATB buffers are available. The 2605 do_marking_step() method only tries to drain SATB buffers right 2606 at the beginning. So, if enough buffers are available, the 2607 marking step aborts and the SATB buffers are processed at 2608 the beginning of the next invocation. 2609 2610 (4) To yield. when we have to yield then we abort and yield 2611 right at the end of do_marking_step(). This saves us from a lot 2612 of hassle as, by yielding we might allow a Full GC. If this 2613 happens then objects will be compacted underneath our feet, the 2614 heap might shrink, etc. We save checking for this by just 2615 aborting and doing the yield right at the end. 2616 2617 From the above it follows that the do_marking_step() method should 2618 be called in a loop (or, otherwise, regularly) until it completes. 2619 2620 If a marking step completes without its has_aborted() flag being 2621 true, it means it has completed the current marking phase (and 2622 also all other marking tasks have done so and have all synced up). 2623 2624 A method called regular_clock_call() is invoked "regularly" (in 2625 sub ms intervals) throughout marking. It is this clock method that 2626 checks all the abort conditions which were mentioned above and 2627 decides when the task should abort. A work-based scheme is used to 2628 trigger this clock method: when the number of object words the 2629 marking phase has scanned or the number of references the marking 2630 phase has visited reach a given limit. Additional invocations to 2631 the method clock have been planted in a few other strategic places 2632 too. The initial reason for the clock method was to avoid calling 2633 vtime too regularly, as it is quite expensive. So, once it was in 2634 place, it was natural to piggy-back all the other conditions on it 2635 too and not constantly check them throughout the code. 2636 2637 If do_termination is true then do_marking_step will enter its 2638 termination protocol. 2639 2640 The value of is_serial must be true when do_marking_step is being 2641 called serially (i.e. by the VMThread) and do_marking_step should 2642 skip any synchronization in the termination and overflow code. 2643 Examples include the serial remark code and the serial reference 2644 processing closures. 2645 2646 The value of is_serial must be false when do_marking_step is 2647 being called by any of the worker threads in a work gang. 2648 Examples include the concurrent marking code (CMMarkingTask), 2649 the MT remark code, and the MT reference processing closures. 2650 2651 *****************************************************************************/ 2652 2653 void G1CMTask::do_marking_step(double time_target_ms, 2654 bool do_termination, 2655 bool is_serial) { 2656 assert(time_target_ms >= 1.0, "minimum granularity is 1ms"); 2657 assert(concurrent() == _cm->concurrent(), "they should be the same"); 2658 2659 G1Policy* g1_policy = _g1h->g1_policy(); 2660 assert(_task_queues != NULL, "invariant"); 2661 assert(_task_queue != NULL, "invariant"); 2662 assert(_task_queues->queue(_worker_id) == _task_queue, "invariant"); 2663 2664 assert(!_claimed, 2665 "only one thread should claim this task at any one time"); 2666 2667 // OK, this doesn't safeguard again all possible scenarios, as it is 2668 // possible for two threads to set the _claimed flag at the same 2669 // time. But it is only for debugging purposes anyway and it will 2670 // catch most problems. 2671 _claimed = true; 2672 2673 _start_time_ms = os::elapsedVTime() * 1000.0; 2674 2675 // If do_stealing is true then do_marking_step will attempt to 2676 // steal work from the other G1CMTasks. It only makes sense to 2677 // enable stealing when the termination protocol is enabled 2678 // and do_marking_step() is not being called serially. 2679 bool do_stealing = do_termination && !is_serial; 2680 2681 double diff_prediction_ms = _g1h->g1_policy()->predictor().get_new_prediction(&_marking_step_diffs_ms); 2682 _time_target_ms = time_target_ms - diff_prediction_ms; 2683 2684 // set up the variables that are used in the work-based scheme to 2685 // call the regular clock method 2686 _words_scanned = 0; 2687 _refs_reached = 0; 2688 recalculate_limits(); 2689 2690 // clear all flags 2691 clear_has_aborted(); 2692 _has_timed_out = false; 2693 _draining_satb_buffers = false; 2694 2695 ++_calls; 2696 2697 // Set up the bitmap and oop closures. Anything that uses them is 2698 // eventually called from this method, so it is OK to allocate these 2699 // statically. 2700 G1CMBitMapClosure bitmap_closure(this, _cm, _nextMarkBitMap); 2701 G1CMOopClosure cm_oop_closure(_g1h, _cm, this); 2702 set_cm_oop_closure(&cm_oop_closure); 2703 2704 if (_cm->has_overflown()) { 2705 // This can happen if the mark stack overflows during a GC pause 2706 // and this task, after a yield point, restarts. We have to abort 2707 // as we need to get into the overflow protocol which happens 2708 // right at the end of this task. 2709 set_has_aborted(); 2710 } 2711 2712 // First drain any available SATB buffers. After this, we will not 2713 // look at SATB buffers before the next invocation of this method. 2714 // If enough completed SATB buffers are queued up, the regular clock 2715 // will abort this task so that it restarts. 2716 drain_satb_buffers(); 2717 // ...then partially drain the local queue and the global stack 2718 drain_local_queue(true); 2719 drain_global_stack(true); 2720 2721 do { 2722 if (!has_aborted() && _curr_region != NULL) { 2723 // This means that we're already holding on to a region. 2724 assert(_finger != NULL, "if region is not NULL, then the finger " 2725 "should not be NULL either"); 2726 2727 // We might have restarted this task after an evacuation pause 2728 // which might have evacuated the region we're holding on to 2729 // underneath our feet. Let's read its limit again to make sure 2730 // that we do not iterate over a region of the heap that 2731 // contains garbage (update_region_limit() will also move 2732 // _finger to the start of the region if it is found empty). 2733 update_region_limit(); 2734 // We will start from _finger not from the start of the region, 2735 // as we might be restarting this task after aborting half-way 2736 // through scanning this region. In this case, _finger points to 2737 // the address where we last found a marked object. If this is a 2738 // fresh region, _finger points to start(). 2739 MemRegion mr = MemRegion(_finger, _region_limit); 2740 2741 assert(!_curr_region->is_humongous() || mr.start() == _curr_region->bottom(), 2742 "humongous regions should go around loop once only"); 2743 2744 // Some special cases: 2745 // If the memory region is empty, we can just give up the region. 2746 // If the current region is humongous then we only need to check 2747 // the bitmap for the bit associated with the start of the object, 2748 // scan the object if it's live, and give up the region. 2749 // Otherwise, let's iterate over the bitmap of the part of the region 2750 // that is left. 2751 // If the iteration is successful, give up the region. 2752 if (mr.is_empty()) { 2753 giveup_current_region(); 2754 regular_clock_call(); 2755 } else if (_curr_region->is_humongous() && mr.start() == _curr_region->bottom()) { 2756 if (_nextMarkBitMap->isMarked(mr.start())) { 2757 // The object is marked - apply the closure 2758 BitMap::idx_t offset = _nextMarkBitMap->heapWordToOffset(mr.start()); 2759 bitmap_closure.do_bit(offset); 2760 } 2761 // Even if this task aborted while scanning the humongous object 2762 // we can (and should) give up the current region. 2763 giveup_current_region(); 2764 regular_clock_call(); 2765 } else if (_nextMarkBitMap->iterate(&bitmap_closure, mr)) { 2766 giveup_current_region(); 2767 regular_clock_call(); 2768 } else { 2769 assert(has_aborted(), "currently the only way to do so"); 2770 // The only way to abort the bitmap iteration is to return 2771 // false from the do_bit() method. However, inside the 2772 // do_bit() method we move the _finger to point to the 2773 // object currently being looked at. So, if we bail out, we 2774 // have definitely set _finger to something non-null. 2775 assert(_finger != NULL, "invariant"); 2776 2777 // Region iteration was actually aborted. So now _finger 2778 // points to the address of the object we last scanned. If we 2779 // leave it there, when we restart this task, we will rescan 2780 // the object. It is easy to avoid this. We move the finger by 2781 // enough to point to the next possible object header (the 2782 // bitmap knows by how much we need to move it as it knows its 2783 // granularity). 2784 assert(_finger < _region_limit, "invariant"); 2785 HeapWord* new_finger = _nextMarkBitMap->nextObject(_finger); 2786 // Check if bitmap iteration was aborted while scanning the last object 2787 if (new_finger >= _region_limit) { 2788 giveup_current_region(); 2789 } else { 2790 move_finger_to(new_finger); 2791 } 2792 } 2793 } 2794 // At this point we have either completed iterating over the 2795 // region we were holding on to, or we have aborted. 2796 2797 // We then partially drain the local queue and the global stack. 2798 // (Do we really need this?) 2799 drain_local_queue(true); 2800 drain_global_stack(true); 2801 2802 // Read the note on the claim_region() method on why it might 2803 // return NULL with potentially more regions available for 2804 // claiming and why we have to check out_of_regions() to determine 2805 // whether we're done or not. 2806 while (!has_aborted() && _curr_region == NULL && !_cm->out_of_regions()) { 2807 // We are going to try to claim a new region. We should have 2808 // given up on the previous one. 2809 // Separated the asserts so that we know which one fires. 2810 assert(_curr_region == NULL, "invariant"); 2811 assert(_finger == NULL, "invariant"); 2812 assert(_region_limit == NULL, "invariant"); 2813 HeapRegion* claimed_region = _cm->claim_region(_worker_id); 2814 if (claimed_region != NULL) { 2815 // Yes, we managed to claim one 2816 setup_for_region(claimed_region); 2817 assert(_curr_region == claimed_region, "invariant"); 2818 } 2819 // It is important to call the regular clock here. It might take 2820 // a while to claim a region if, for example, we hit a large 2821 // block of empty regions. So we need to call the regular clock 2822 // method once round the loop to make sure it's called 2823 // frequently enough. 2824 regular_clock_call(); 2825 } 2826 2827 if (!has_aborted() && _curr_region == NULL) { 2828 assert(_cm->out_of_regions(), 2829 "at this point we should be out of regions"); 2830 } 2831 } while ( _curr_region != NULL && !has_aborted()); 2832 2833 if (!has_aborted()) { 2834 // We cannot check whether the global stack is empty, since other 2835 // tasks might be pushing objects to it concurrently. 2836 assert(_cm->out_of_regions(), 2837 "at this point we should be out of regions"); 2838 // Try to reduce the number of available SATB buffers so that 2839 // remark has less work to do. 2840 drain_satb_buffers(); 2841 } 2842 2843 // Since we've done everything else, we can now totally drain the 2844 // local queue and global stack. 2845 drain_local_queue(false); 2846 drain_global_stack(false); 2847 2848 // Attempt at work stealing from other task's queues. 2849 if (do_stealing && !has_aborted()) { 2850 // We have not aborted. This means that we have finished all that 2851 // we could. Let's try to do some stealing... 2852 2853 // We cannot check whether the global stack is empty, since other 2854 // tasks might be pushing objects to it concurrently. 2855 assert(_cm->out_of_regions() && _task_queue->size() == 0, 2856 "only way to reach here"); 2857 while (!has_aborted()) { 2858 G1TaskQueueEntry entry; 2859 if (_cm->try_stealing(_worker_id, &_hash_seed, entry)) { 2860 scan_task_entry(entry); 2861 2862 // And since we're towards the end, let's totally drain the 2863 // local queue and global stack. 2864 drain_local_queue(false); 2865 drain_global_stack(false); 2866 } else { 2867 break; 2868 } 2869 } 2870 } 2871 2872 // We still haven't aborted. Now, let's try to get into the 2873 // termination protocol. 2874 if (do_termination && !has_aborted()) { 2875 // We cannot check whether the global stack is empty, since other 2876 // tasks might be concurrently pushing objects on it. 2877 // Separated the asserts so that we know which one fires. 2878 assert(_cm->out_of_regions(), "only way to reach here"); 2879 assert(_task_queue->size() == 0, "only way to reach here"); 2880 _termination_start_time_ms = os::elapsedVTime() * 1000.0; 2881 2882 // The G1CMTask class also extends the TerminatorTerminator class, 2883 // hence its should_exit_termination() method will also decide 2884 // whether to exit the termination protocol or not. 2885 bool finished = (is_serial || 2886 _cm->terminator()->offer_termination(this)); 2887 double termination_end_time_ms = os::elapsedVTime() * 1000.0; 2888 _termination_time_ms += 2889 termination_end_time_ms - _termination_start_time_ms; 2890 2891 if (finished) { 2892 // We're all done. 2893 2894 if (_worker_id == 0) { 2895 // let's allow task 0 to do this 2896 if (concurrent()) { 2897 assert(_cm->concurrent_marking_in_progress(), "invariant"); 2898 // we need to set this to false before the next 2899 // safepoint. This way we ensure that the marking phase 2900 // doesn't observe any more heap expansions. 2901 _cm->clear_concurrent_marking_in_progress(); 2902 } 2903 } 2904 2905 // We can now guarantee that the global stack is empty, since 2906 // all other tasks have finished. We separated the guarantees so 2907 // that, if a condition is false, we can immediately find out 2908 // which one. 2909 guarantee(_cm->out_of_regions(), "only way to reach here"); 2910 guarantee(_cm->mark_stack_empty(), "only way to reach here"); 2911 guarantee(_task_queue->size() == 0, "only way to reach here"); 2912 guarantee(!_cm->has_overflown(), "only way to reach here"); 2913 } else { 2914 // Apparently there's more work to do. Let's abort this task. It 2915 // will restart it and we can hopefully find more things to do. 2916 set_has_aborted(); 2917 } 2918 } 2919 2920 // Mainly for debugging purposes to make sure that a pointer to the 2921 // closure which was statically allocated in this frame doesn't 2922 // escape it by accident. 2923 set_cm_oop_closure(NULL); 2924 double end_time_ms = os::elapsedVTime() * 1000.0; 2925 double elapsed_time_ms = end_time_ms - _start_time_ms; 2926 // Update the step history. 2927 _step_times_ms.add(elapsed_time_ms); 2928 2929 if (has_aborted()) { 2930 // The task was aborted for some reason. 2931 if (_has_timed_out) { 2932 double diff_ms = elapsed_time_ms - _time_target_ms; 2933 // Keep statistics of how well we did with respect to hitting 2934 // our target only if we actually timed out (if we aborted for 2935 // other reasons, then the results might get skewed). 2936 _marking_step_diffs_ms.add(diff_ms); 2937 } 2938 2939 if (_cm->has_overflown()) { 2940 // This is the interesting one. We aborted because a global 2941 // overflow was raised. This means we have to restart the 2942 // marking phase and start iterating over regions. However, in 2943 // order to do this we have to make sure that all tasks stop 2944 // what they are doing and re-initialize in a safe manner. We 2945 // will achieve this with the use of two barrier sync points. 2946 2947 if (!is_serial) { 2948 // We only need to enter the sync barrier if being called 2949 // from a parallel context 2950 _cm->enter_first_sync_barrier(_worker_id); 2951 2952 // When we exit this sync barrier we know that all tasks have 2953 // stopped doing marking work. So, it's now safe to 2954 // re-initialize our data structures. At the end of this method, 2955 // task 0 will clear the global data structures. 2956 } 2957 2958 // We clear the local state of this task... 2959 clear_region_fields(); 2960 2961 if (!is_serial) { 2962 // ...and enter the second barrier. 2963 _cm->enter_second_sync_barrier(_worker_id); 2964 } 2965 // At this point, if we're during the concurrent phase of 2966 // marking, everything has been re-initialized and we're 2967 // ready to restart. 2968 } 2969 } 2970 2971 _claimed = false; 2972 } 2973 2974 G1CMTask::G1CMTask(uint worker_id, 2975 G1ConcurrentMark* cm, 2976 G1CMTaskQueue* task_queue, 2977 G1CMTaskQueueSet* task_queues) 2978 : _g1h(G1CollectedHeap::heap()), 2979 _worker_id(worker_id), _cm(cm), 2980 _objArray_processor(this), 2981 _claimed(false), 2982 _nextMarkBitMap(NULL), _hash_seed(17), 2983 _task_queue(task_queue), 2984 _task_queues(task_queues), 2985 _cm_oop_closure(NULL) { 2986 guarantee(task_queue != NULL, "invariant"); 2987 guarantee(task_queues != NULL, "invariant"); 2988 2989 _marking_step_diffs_ms.add(0.5); 2990 } 2991 2992 // These are formatting macros that are used below to ensure 2993 // consistent formatting. The *_H_* versions are used to format the 2994 // header for a particular value and they should be kept consistent 2995 // with the corresponding macro. Also note that most of the macros add 2996 // the necessary white space (as a prefix) which makes them a bit 2997 // easier to compose. 2998 2999 // All the output lines are prefixed with this string to be able to 3000 // identify them easily in a large log file. 3001 #define G1PPRL_LINE_PREFIX "###" 3002 3003 #define G1PPRL_ADDR_BASE_FORMAT " " PTR_FORMAT "-" PTR_FORMAT 3004 #ifdef _LP64 3005 #define G1PPRL_ADDR_BASE_H_FORMAT " %37s" 3006 #else // _LP64 3007 #define G1PPRL_ADDR_BASE_H_FORMAT " %21s" 3008 #endif // _LP64 3009 3010 // For per-region info 3011 #define G1PPRL_TYPE_FORMAT " %-4s" 3012 #define G1PPRL_TYPE_H_FORMAT " %4s" 3013 #define G1PPRL_BYTE_FORMAT " " SIZE_FORMAT_W(9) 3014 #define G1PPRL_BYTE_H_FORMAT " %9s" 3015 #define G1PPRL_DOUBLE_FORMAT " %14.1f" 3016 #define G1PPRL_DOUBLE_H_FORMAT " %14s" 3017 3018 // For summary info 3019 #define G1PPRL_SUM_ADDR_FORMAT(tag) " " tag ":" G1PPRL_ADDR_BASE_FORMAT 3020 #define G1PPRL_SUM_BYTE_FORMAT(tag) " " tag ": " SIZE_FORMAT 3021 #define G1PPRL_SUM_MB_FORMAT(tag) " " tag ": %1.2f MB" 3022 #define G1PPRL_SUM_MB_PERC_FORMAT(tag) G1PPRL_SUM_MB_FORMAT(tag) " / %1.2f %%" 3023 3024 G1PrintRegionLivenessInfoClosure:: 3025 G1PrintRegionLivenessInfoClosure(const char* phase_name) 3026 : _total_used_bytes(0), _total_capacity_bytes(0), 3027 _total_prev_live_bytes(0), _total_next_live_bytes(0), 3028 _total_remset_bytes(0), _total_strong_code_roots_bytes(0) { 3029 G1CollectedHeap* g1h = G1CollectedHeap::heap(); 3030 MemRegion g1_reserved = g1h->g1_reserved(); 3031 double now = os::elapsedTime(); 3032 3033 // Print the header of the output. 3034 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" PHASE %s @ %1.3f", phase_name, now); 3035 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX" HEAP" 3036 G1PPRL_SUM_ADDR_FORMAT("reserved") 3037 G1PPRL_SUM_BYTE_FORMAT("region-size"), 3038 p2i(g1_reserved.start()), p2i(g1_reserved.end()), 3039 HeapRegion::GrainBytes); 3040 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX); 3041 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX 3042 G1PPRL_TYPE_H_FORMAT 3043 G1PPRL_ADDR_BASE_H_FORMAT 3044 G1PPRL_BYTE_H_FORMAT 3045 G1PPRL_BYTE_H_FORMAT 3046 G1PPRL_BYTE_H_FORMAT 3047 G1PPRL_DOUBLE_H_FORMAT 3048 G1PPRL_BYTE_H_FORMAT 3049 G1PPRL_BYTE_H_FORMAT, 3050 "type", "address-range", 3051 "used", "prev-live", "next-live", "gc-eff", 3052 "remset", "code-roots"); 3053 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX 3054 G1PPRL_TYPE_H_FORMAT 3055 G1PPRL_ADDR_BASE_H_FORMAT 3056 G1PPRL_BYTE_H_FORMAT 3057 G1PPRL_BYTE_H_FORMAT 3058 G1PPRL_BYTE_H_FORMAT 3059 G1PPRL_DOUBLE_H_FORMAT 3060 G1PPRL_BYTE_H_FORMAT 3061 G1PPRL_BYTE_H_FORMAT, 3062 "", "", 3063 "(bytes)", "(bytes)", "(bytes)", "(bytes/ms)", 3064 "(bytes)", "(bytes)"); 3065 } 3066 3067 bool G1PrintRegionLivenessInfoClosure::doHeapRegion(HeapRegion* r) { 3068 const char* type = r->get_type_str(); 3069 HeapWord* bottom = r->bottom(); 3070 HeapWord* end = r->end(); 3071 size_t capacity_bytes = r->capacity(); 3072 size_t used_bytes = r->used(); 3073 size_t prev_live_bytes = r->live_bytes(); 3074 size_t next_live_bytes = r->next_live_bytes(); 3075 double gc_eff = r->gc_efficiency(); 3076 size_t remset_bytes = r->rem_set()->mem_size(); 3077 size_t strong_code_roots_bytes = r->rem_set()->strong_code_roots_mem_size(); 3078 3079 _total_used_bytes += used_bytes; 3080 _total_capacity_bytes += capacity_bytes; 3081 _total_prev_live_bytes += prev_live_bytes; 3082 _total_next_live_bytes += next_live_bytes; 3083 _total_remset_bytes += remset_bytes; 3084 _total_strong_code_roots_bytes += strong_code_roots_bytes; 3085 3086 // Print a line for this particular region. 3087 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX 3088 G1PPRL_TYPE_FORMAT 3089 G1PPRL_ADDR_BASE_FORMAT 3090 G1PPRL_BYTE_FORMAT 3091 G1PPRL_BYTE_FORMAT 3092 G1PPRL_BYTE_FORMAT 3093 G1PPRL_DOUBLE_FORMAT 3094 G1PPRL_BYTE_FORMAT 3095 G1PPRL_BYTE_FORMAT, 3096 type, p2i(bottom), p2i(end), 3097 used_bytes, prev_live_bytes, next_live_bytes, gc_eff, 3098 remset_bytes, strong_code_roots_bytes); 3099 3100 return false; 3101 } 3102 3103 G1PrintRegionLivenessInfoClosure::~G1PrintRegionLivenessInfoClosure() { 3104 // add static memory usages to remembered set sizes 3105 _total_remset_bytes += HeapRegionRemSet::fl_mem_size() + HeapRegionRemSet::static_mem_size(); 3106 // Print the footer of the output. 3107 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX); 3108 log_trace(gc, liveness)(G1PPRL_LINE_PREFIX 3109 " SUMMARY" 3110 G1PPRL_SUM_MB_FORMAT("capacity") 3111 G1PPRL_SUM_MB_PERC_FORMAT("used") 3112 G1PPRL_SUM_MB_PERC_FORMAT("prev-live") 3113 G1PPRL_SUM_MB_PERC_FORMAT("next-live") 3114 G1PPRL_SUM_MB_FORMAT("remset") 3115 G1PPRL_SUM_MB_FORMAT("code-roots"), 3116 bytes_to_mb(_total_capacity_bytes), 3117 bytes_to_mb(_total_used_bytes), 3118 perc(_total_used_bytes, _total_capacity_bytes), 3119 bytes_to_mb(_total_prev_live_bytes), 3120 perc(_total_prev_live_bytes, _total_capacity_bytes), 3121 bytes_to_mb(_total_next_live_bytes), 3122 perc(_total_next_live_bytes, _total_capacity_bytes), 3123 bytes_to_mb(_total_remset_bytes), 3124 bytes_to_mb(_total_strong_code_roots_bytes)); 3125 }