1 /*
2 * Copyright (c) 2001, 2012, 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 "gc_implementation/concurrentMarkSweep/cmsLockVerifier.hpp"
27 #include "gc_implementation/concurrentMarkSweep/compactibleFreeListSpace.hpp"
28 #include "gc_implementation/concurrentMarkSweep/concurrentMarkSweepGeneration.inline.hpp"
29 #include "gc_implementation/concurrentMarkSweep/concurrentMarkSweepThread.hpp"
30 #include "gc_implementation/shared/liveRange.hpp"
31 #include "gc_implementation/shared/spaceDecorator.hpp"
32 #include "gc_interface/collectedHeap.hpp"
33 #include "memory/allocation.inline.hpp"
34 #include "memory/blockOffsetTable.inline.hpp"
35 #include "memory/resourceArea.hpp"
36 #include "memory/universe.inline.hpp"
37 #include "oops/oop.inline.hpp"
38 #include "runtime/globals.hpp"
39 #include "runtime/handles.inline.hpp"
40 #include "runtime/init.hpp"
41 #include "runtime/java.hpp"
42 #include "runtime/vmThread.hpp"
43 #include "utilities/copy.hpp"
44
45 /////////////////////////////////////////////////////////////////////////
46 //// CompactibleFreeListSpace
47 /////////////////////////////////////////////////////////////////////////
48
49 // highest ranked free list lock rank
50 int CompactibleFreeListSpace::_lockRank = Mutex::leaf + 3;
51
52 // Defaults are 0 so things will break badly if incorrectly initialized.
53 size_t CompactibleFreeListSpace::IndexSetStart = 0;
54 size_t CompactibleFreeListSpace::IndexSetStride = 0;
55
56 size_t MinChunkSize = 0;
57
58 void CompactibleFreeListSpace::set_cms_values() {
59 // Set CMS global values
60 assert(MinChunkSize == 0, "already set");
61 #define numQuanta(x,y) ((x+y-1)/y)
62 MinChunkSize = numQuanta(sizeof(FreeChunk), MinObjAlignmentInBytes) * MinObjAlignment;
63
64 assert(IndexSetStart == 0 && IndexSetStride == 0, "already set");
65 IndexSetStart = MinChunkSize;
66 IndexSetStride = MinObjAlignment;
67 }
68
69 // Constructor
70 CompactibleFreeListSpace::CompactibleFreeListSpace(BlockOffsetSharedArray* bs,
71 MemRegion mr, bool use_adaptive_freelists,
72 FreeBlockDictionary<FreeChunk>::DictionaryChoice dictionaryChoice) :
73 _dictionaryChoice(dictionaryChoice),
74 _adaptive_freelists(use_adaptive_freelists),
75 _bt(bs, mr),
76 // free list locks are in the range of values taken by _lockRank
77 // This range currently is [_leaf+2, _leaf+3]
78 // Note: this requires that CFLspace c'tors
79 // are called serially in the order in which the locks are
80 // are acquired in the program text. This is true today.
81 _freelistLock(_lockRank--, "CompactibleFreeListSpace._lock", true),
82 _parDictionaryAllocLock(Mutex::leaf - 1, // == rank(ExpandHeap_lock) - 1
83 "CompactibleFreeListSpace._dict_par_lock", true),
84 _rescan_task_size(CardTableModRefBS::card_size_in_words * BitsPerWord *
85 CMSRescanMultiple),
86 _marking_task_size(CardTableModRefBS::card_size_in_words * BitsPerWord *
87 CMSConcMarkMultiple),
88 _collector(NULL)
89 {
90 assert(sizeof(FreeChunk) / BytesPerWord <= MinChunkSize,
91 "FreeChunk is larger than expected");
92 _bt.set_space(this);
93 initialize(mr, SpaceDecorator::Clear, SpaceDecorator::Mangle);
94 // We have all of "mr", all of which we place in the dictionary
95 // as one big chunk. We'll need to decide here which of several
96 // possible alternative dictionary implementations to use. For
97 // now the choice is easy, since we have only one working
98 // implementation, namely, the simple binary tree (splaying
99 // temporarily disabled).
100 switch (dictionaryChoice) {
101 case FreeBlockDictionary<FreeChunk>::dictionarySplayTree:
102 case FreeBlockDictionary<FreeChunk>::dictionarySkipList:
103 default:
104 warning("dictionaryChoice: selected option not understood; using"
105 " default BinaryTreeDictionary implementation instead.");
106 case FreeBlockDictionary<FreeChunk>::dictionaryBinaryTree:
107 _dictionary = new BinaryTreeDictionary<FreeChunk>(mr, use_adaptive_freelists);
108 break;
109 }
110 assert(_dictionary != NULL, "CMS dictionary initialization");
111 // The indexed free lists are initially all empty and are lazily
112 // filled in on demand. Initialize the array elements to NULL.
113 initializeIndexedFreeListArray();
114
115 // Not using adaptive free lists assumes that allocation is first
116 // from the linAB's. Also a cms perm gen which can be compacted
117 // has to have the klass's klassKlass allocated at a lower
118 // address in the heap than the klass so that the klassKlass is
119 // moved to its new location before the klass is moved.
120 // Set the _refillSize for the linear allocation blocks
121 if (!use_adaptive_freelists) {
122 FreeChunk* fc = _dictionary->get_chunk(mr.word_size());
123 // The small linAB initially has all the space and will allocate
124 // a chunk of any size.
125 HeapWord* addr = (HeapWord*) fc;
126 _smallLinearAllocBlock.set(addr, fc->size() ,
127 1024*SmallForLinearAlloc, fc->size());
128 // Note that _unallocated_block is not updated here.
129 // Allocations from the linear allocation block should
130 // update it.
131 } else {
132 _smallLinearAllocBlock.set(0, 0, 1024*SmallForLinearAlloc,
133 SmallForLinearAlloc);
134 }
135 // CMSIndexedFreeListReplenish should be at least 1
136 CMSIndexedFreeListReplenish = MAX2((uintx)1, CMSIndexedFreeListReplenish);
137 _promoInfo.setSpace(this);
138 if (UseCMSBestFit) {
139 _fitStrategy = FreeBlockBestFitFirst;
140 } else {
141 _fitStrategy = FreeBlockStrategyNone;
142 }
143 check_free_list_consistency();
144
145 // Initialize locks for parallel case.
146
147 if (CollectedHeap::use_parallel_gc_threads()) {
148 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
149 _indexedFreeListParLocks[i] = new Mutex(Mutex::leaf - 1, // == ExpandHeap_lock - 1
150 "a freelist par lock",
151 true);
152 if (_indexedFreeListParLocks[i] == NULL)
153 vm_exit_during_initialization("Could not allocate a par lock");
154 DEBUG_ONLY(
155 _indexedFreeList[i].set_protecting_lock(_indexedFreeListParLocks[i]);
156 )
157 }
158 _dictionary->set_par_lock(&_parDictionaryAllocLock);
159 }
160 }
161
162 // Like CompactibleSpace forward() but always calls cross_threshold() to
163 // update the block offset table. Removed initialize_threshold call because
164 // CFLS does not use a block offset array for contiguous spaces.
165 HeapWord* CompactibleFreeListSpace::forward(oop q, size_t size,
166 CompactPoint* cp, HeapWord* compact_top) {
167 // q is alive
168 // First check if we should switch compaction space
169 assert(this == cp->space, "'this' should be current compaction space.");
170 size_t compaction_max_size = pointer_delta(end(), compact_top);
171 assert(adjustObjectSize(size) == cp->space->adjust_object_size_v(size),
172 "virtual adjustObjectSize_v() method is not correct");
173 size_t adjusted_size = adjustObjectSize(size);
174 assert(compaction_max_size >= MinChunkSize || compaction_max_size == 0,
175 "no small fragments allowed");
176 assert(minimum_free_block_size() == MinChunkSize,
177 "for de-virtualized reference below");
178 // Can't leave a nonzero size, residual fragment smaller than MinChunkSize
179 if (adjusted_size + MinChunkSize > compaction_max_size &&
180 adjusted_size != compaction_max_size) {
181 do {
182 // switch to next compaction space
183 cp->space->set_compaction_top(compact_top);
184 cp->space = cp->space->next_compaction_space();
185 if (cp->space == NULL) {
186 cp->gen = GenCollectedHeap::heap()->prev_gen(cp->gen);
187 assert(cp->gen != NULL, "compaction must succeed");
188 cp->space = cp->gen->first_compaction_space();
189 assert(cp->space != NULL, "generation must have a first compaction space");
190 }
191 compact_top = cp->space->bottom();
192 cp->space->set_compaction_top(compact_top);
193 // The correct adjusted_size may not be the same as that for this method
194 // (i.e., cp->space may no longer be "this" so adjust the size again.
195 // Use the virtual method which is not used above to save the virtual
196 // dispatch.
197 adjusted_size = cp->space->adjust_object_size_v(size);
198 compaction_max_size = pointer_delta(cp->space->end(), compact_top);
199 assert(cp->space->minimum_free_block_size() == 0, "just checking");
200 } while (adjusted_size > compaction_max_size);
201 }
202
203 // store the forwarding pointer into the mark word
204 if ((HeapWord*)q != compact_top) {
205 q->forward_to(oop(compact_top));
206 assert(q->is_gc_marked(), "encoding the pointer should preserve the mark");
207 } else {
208 // if the object isn't moving we can just set the mark to the default
209 // mark and handle it specially later on.
210 q->init_mark();
211 assert(q->forwardee() == NULL, "should be forwarded to NULL");
212 }
213
214 VALIDATE_MARK_SWEEP_ONLY(MarkSweep::register_live_oop(q, adjusted_size));
215 compact_top += adjusted_size;
216
217 // we need to update the offset table so that the beginnings of objects can be
218 // found during scavenge. Note that we are updating the offset table based on
219 // where the object will be once the compaction phase finishes.
220
221 // Always call cross_threshold(). A contiguous space can only call it when
222 // the compaction_top exceeds the current threshold but not for an
223 // non-contiguous space.
224 cp->threshold =
225 cp->space->cross_threshold(compact_top - adjusted_size, compact_top);
226 return compact_top;
227 }
228
229 // A modified copy of OffsetTableContigSpace::cross_threshold() with _offsets -> _bt
230 // and use of single_block instead of alloc_block. The name here is not really
231 // appropriate - maybe a more general name could be invented for both the
232 // contiguous and noncontiguous spaces.
233
234 HeapWord* CompactibleFreeListSpace::cross_threshold(HeapWord* start, HeapWord* the_end) {
235 _bt.single_block(start, the_end);
236 return end();
237 }
238
239 // Initialize them to NULL.
240 void CompactibleFreeListSpace::initializeIndexedFreeListArray() {
241 for (size_t i = 0; i < IndexSetSize; i++) {
242 // Note that on platforms where objects are double word aligned,
243 // the odd array elements are not used. It is convenient, however,
244 // to map directly from the object size to the array element.
245 _indexedFreeList[i].reset(IndexSetSize);
246 _indexedFreeList[i].set_size(i);
247 assert(_indexedFreeList[i].count() == 0, "reset check failed");
248 assert(_indexedFreeList[i].head() == NULL, "reset check failed");
249 assert(_indexedFreeList[i].tail() == NULL, "reset check failed");
250 assert(_indexedFreeList[i].hint() == IndexSetSize, "reset check failed");
251 }
252 }
253
254 void CompactibleFreeListSpace::resetIndexedFreeListArray() {
255 for (size_t i = 1; i < IndexSetSize; i++) {
256 assert(_indexedFreeList[i].size() == (size_t) i,
257 "Indexed free list sizes are incorrect");
258 _indexedFreeList[i].reset(IndexSetSize);
259 assert(_indexedFreeList[i].count() == 0, "reset check failed");
260 assert(_indexedFreeList[i].head() == NULL, "reset check failed");
261 assert(_indexedFreeList[i].tail() == NULL, "reset check failed");
262 assert(_indexedFreeList[i].hint() == IndexSetSize, "reset check failed");
263 }
264 }
265
266 void CompactibleFreeListSpace::reset(MemRegion mr) {
267 resetIndexedFreeListArray();
268 dictionary()->reset();
269 if (BlockOffsetArrayUseUnallocatedBlock) {
270 assert(end() == mr.end(), "We are compacting to the bottom of CMS gen");
271 // Everything's allocated until proven otherwise.
272 _bt.set_unallocated_block(end());
273 }
274 if (!mr.is_empty()) {
275 assert(mr.word_size() >= MinChunkSize, "Chunk size is too small");
276 _bt.single_block(mr.start(), mr.word_size());
277 FreeChunk* fc = (FreeChunk*) mr.start();
278 fc->set_size(mr.word_size());
279 if (mr.word_size() >= IndexSetSize ) {
280 returnChunkToDictionary(fc);
281 } else {
282 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
283 _indexedFreeList[mr.word_size()].return_chunk_at_head(fc);
284 }
285 }
286 _promoInfo.reset();
287 _smallLinearAllocBlock._ptr = NULL;
288 _smallLinearAllocBlock._word_size = 0;
289 }
290
291 void CompactibleFreeListSpace::reset_after_compaction() {
292 // Reset the space to the new reality - one free chunk.
293 MemRegion mr(compaction_top(), end());
294 reset(mr);
295 // Now refill the linear allocation block(s) if possible.
296 if (_adaptive_freelists) {
297 refillLinearAllocBlocksIfNeeded();
298 } else {
299 // Place as much of mr in the linAB as we can get,
300 // provided it was big enough to go into the dictionary.
301 FreeChunk* fc = dictionary()->find_largest_dict();
302 if (fc != NULL) {
303 assert(fc->size() == mr.word_size(),
304 "Why was the chunk broken up?");
305 removeChunkFromDictionary(fc);
306 HeapWord* addr = (HeapWord*) fc;
307 _smallLinearAllocBlock.set(addr, fc->size() ,
308 1024*SmallForLinearAlloc, fc->size());
309 // Note that _unallocated_block is not updated here.
310 }
311 }
312 }
313
314 // Walks the entire dictionary, returning a coterminal
315 // chunk, if it exists. Use with caution since it involves
316 // a potentially complete walk of a potentially large tree.
317 FreeChunk* CompactibleFreeListSpace::find_chunk_at_end() {
318
319 assert_lock_strong(&_freelistLock);
320
321 return dictionary()->find_chunk_ends_at(end());
322 }
323
324
325 #ifndef PRODUCT
326 void CompactibleFreeListSpace::initializeIndexedFreeListArrayReturnedBytes() {
327 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
328 _indexedFreeList[i].allocation_stats()->set_returned_bytes(0);
329 }
330 }
331
332 size_t CompactibleFreeListSpace::sumIndexedFreeListArrayReturnedBytes() {
333 size_t sum = 0;
334 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
335 sum += _indexedFreeList[i].allocation_stats()->returned_bytes();
336 }
337 return sum;
338 }
339
340 size_t CompactibleFreeListSpace::totalCountInIndexedFreeLists() const {
341 size_t count = 0;
342 for (size_t i = IndexSetStart; i < IndexSetSize; i++) {
343 debug_only(
344 ssize_t total_list_count = 0;
345 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
346 fc = fc->next()) {
347 total_list_count++;
348 }
349 assert(total_list_count == _indexedFreeList[i].count(),
350 "Count in list is incorrect");
351 )
352 count += _indexedFreeList[i].count();
353 }
354 return count;
355 }
356
357 size_t CompactibleFreeListSpace::totalCount() {
358 size_t num = totalCountInIndexedFreeLists();
359 num += dictionary()->total_count();
360 if (_smallLinearAllocBlock._word_size != 0) {
361 num++;
362 }
363 return num;
364 }
365 #endif
366
367 bool CompactibleFreeListSpace::is_free_block(const HeapWord* p) const {
368 FreeChunk* fc = (FreeChunk*) p;
369 return fc->is_free();
370 }
371
372 size_t CompactibleFreeListSpace::used() const {
373 return capacity() - free();
374 }
375
376 size_t CompactibleFreeListSpace::free() const {
377 // "MT-safe, but not MT-precise"(TM), if you will: i.e.
378 // if you do this while the structures are in flux you
379 // may get an approximate answer only; for instance
380 // because there is concurrent allocation either
381 // directly by mutators or for promotion during a GC.
382 // It's "MT-safe", however, in the sense that you are guaranteed
383 // not to crash and burn, for instance, because of walking
384 // pointers that could disappear as you were walking them.
385 // The approximation is because the various components
386 // that are read below are not read atomically (and
387 // further the computation of totalSizeInIndexedFreeLists()
388 // is itself a non-atomic computation. The normal use of
389 // this is during a resize operation at the end of GC
390 // and at that time you are guaranteed to get the
391 // correct actual value. However, for instance, this is
392 // also read completely asynchronously by the "perf-sampler"
393 // that supports jvmstat, and you are apt to see the values
394 // flicker in such cases.
395 assert(_dictionary != NULL, "No _dictionary?");
396 return (_dictionary->total_chunk_size(DEBUG_ONLY(freelistLock())) +
397 totalSizeInIndexedFreeLists() +
398 _smallLinearAllocBlock._word_size) * HeapWordSize;
399 }
400
401 size_t CompactibleFreeListSpace::max_alloc_in_words() const {
402 assert(_dictionary != NULL, "No _dictionary?");
403 assert_locked();
404 size_t res = _dictionary->max_chunk_size();
405 res = MAX2(res, MIN2(_smallLinearAllocBlock._word_size,
406 (size_t) SmallForLinearAlloc - 1));
407 // XXX the following could potentially be pretty slow;
408 // should one, pesimally for the rare cases when res
409 // caclulated above is less than IndexSetSize,
410 // just return res calculated above? My reasoning was that
411 // those cases will be so rare that the extra time spent doesn't
412 // really matter....
413 // Note: do not change the loop test i >= res + IndexSetStride
414 // to i > res below, because i is unsigned and res may be zero.
415 for (size_t i = IndexSetSize - 1; i >= res + IndexSetStride;
416 i -= IndexSetStride) {
417 if (_indexedFreeList[i].head() != NULL) {
418 assert(_indexedFreeList[i].count() != 0, "Inconsistent FreeList");
419 return i;
420 }
421 }
422 return res;
423 }
424
425 void LinearAllocBlock::print_on(outputStream* st) const {
426 st->print_cr(" LinearAllocBlock: ptr = " PTR_FORMAT ", word_size = " SIZE_FORMAT
427 ", refillsize = " SIZE_FORMAT ", allocation_size_limit = " SIZE_FORMAT,
428 _ptr, _word_size, _refillSize, _allocation_size_limit);
429 }
430
431 void CompactibleFreeListSpace::print_on(outputStream* st) const {
432 st->print_cr("COMPACTIBLE FREELIST SPACE");
433 st->print_cr(" Space:");
434 Space::print_on(st);
435
436 st->print_cr("promoInfo:");
437 _promoInfo.print_on(st);
438
439 st->print_cr("_smallLinearAllocBlock");
440 _smallLinearAllocBlock.print_on(st);
441
442 // dump_memory_block(_smallLinearAllocBlock->_ptr, 128);
443
444 st->print_cr(" _fitStrategy = %s, _adaptive_freelists = %s",
445 _fitStrategy?"true":"false", _adaptive_freelists?"true":"false");
446 }
447
448 void CompactibleFreeListSpace::print_indexed_free_lists(outputStream* st)
449 const {
450 reportIndexedFreeListStatistics();
451 gclog_or_tty->print_cr("Layout of Indexed Freelists");
452 gclog_or_tty->print_cr("---------------------------");
453 FreeList<FreeChunk>::print_labels_on(st, "size");
454 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
455 _indexedFreeList[i].print_on(gclog_or_tty);
456 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
457 fc = fc->next()) {
458 gclog_or_tty->print_cr("\t[" PTR_FORMAT "," PTR_FORMAT ") %s",
459 fc, (HeapWord*)fc + i,
460 fc->cantCoalesce() ? "\t CC" : "");
461 }
462 }
463 }
464
465 void CompactibleFreeListSpace::print_promo_info_blocks(outputStream* st)
466 const {
467 _promoInfo.print_on(st);
468 }
469
470 void CompactibleFreeListSpace::print_dictionary_free_lists(outputStream* st)
471 const {
472 _dictionary->report_statistics();
473 st->print_cr("Layout of Freelists in Tree");
474 st->print_cr("---------------------------");
475 _dictionary->print_free_lists(st);
476 }
477
478 class BlkPrintingClosure: public BlkClosure {
479 const CMSCollector* _collector;
480 const CompactibleFreeListSpace* _sp;
481 const CMSBitMap* _live_bit_map;
482 const bool _post_remark;
483 outputStream* _st;
484 public:
485 BlkPrintingClosure(const CMSCollector* collector,
486 const CompactibleFreeListSpace* sp,
487 const CMSBitMap* live_bit_map,
488 outputStream* st):
489 _collector(collector),
490 _sp(sp),
491 _live_bit_map(live_bit_map),
492 _post_remark(collector->abstract_state() > CMSCollector::FinalMarking),
493 _st(st) { }
494 size_t do_blk(HeapWord* addr);
495 };
496
497 size_t BlkPrintingClosure::do_blk(HeapWord* addr) {
498 size_t sz = _sp->block_size_no_stall(addr, _collector);
499 assert(sz != 0, "Should always be able to compute a size");
500 if (_sp->block_is_obj(addr)) {
501 const bool dead = _post_remark && !_live_bit_map->isMarked(addr);
502 _st->print_cr(PTR_FORMAT ": %s object of size " SIZE_FORMAT "%s",
503 addr,
504 dead ? "dead" : "live",
505 sz,
506 (!dead && CMSPrintObjectsInDump) ? ":" : ".");
507 if (CMSPrintObjectsInDump && !dead) {
508 oop(addr)->print_on(_st);
509 _st->print_cr("--------------------------------------");
510 }
511 } else { // free block
512 _st->print_cr(PTR_FORMAT ": free block of size " SIZE_FORMAT "%s",
513 addr, sz, CMSPrintChunksInDump ? ":" : ".");
514 if (CMSPrintChunksInDump) {
515 ((FreeChunk*)addr)->print_on(_st);
516 _st->print_cr("--------------------------------------");
517 }
518 }
519 return sz;
520 }
521
522 void CompactibleFreeListSpace::dump_at_safepoint_with_locks(CMSCollector* c,
523 outputStream* st) {
524 st->print_cr("\n=========================");
525 st->print_cr("Block layout in CMS Heap:");
526 st->print_cr("=========================");
527 BlkPrintingClosure bpcl(c, this, c->markBitMap(), st);
528 blk_iterate(&bpcl);
529
530 st->print_cr("\n=======================================");
531 st->print_cr("Order & Layout of Promotion Info Blocks");
532 st->print_cr("=======================================");
533 print_promo_info_blocks(st);
534
535 st->print_cr("\n===========================");
536 st->print_cr("Order of Indexed Free Lists");
537 st->print_cr("=========================");
538 print_indexed_free_lists(st);
539
540 st->print_cr("\n=================================");
541 st->print_cr("Order of Free Lists in Dictionary");
542 st->print_cr("=================================");
543 print_dictionary_free_lists(st);
544 }
545
546
547 void CompactibleFreeListSpace::reportFreeListStatistics() const {
548 assert_lock_strong(&_freelistLock);
549 assert(PrintFLSStatistics != 0, "Reporting error");
550 _dictionary->report_statistics();
551 if (PrintFLSStatistics > 1) {
552 reportIndexedFreeListStatistics();
553 size_t total_size = totalSizeInIndexedFreeLists() +
554 _dictionary->total_chunk_size(DEBUG_ONLY(freelistLock()));
555 gclog_or_tty->print(" free=%ld frag=%1.4f\n", total_size, flsFrag());
556 }
557 }
558
559 void CompactibleFreeListSpace::reportIndexedFreeListStatistics() const {
560 assert_lock_strong(&_freelistLock);
561 gclog_or_tty->print("Statistics for IndexedFreeLists:\n"
562 "--------------------------------\n");
563 size_t total_size = totalSizeInIndexedFreeLists();
564 size_t free_blocks = numFreeBlocksInIndexedFreeLists();
565 gclog_or_tty->print("Total Free Space: %d\n", total_size);
566 gclog_or_tty->print("Max Chunk Size: %d\n", maxChunkSizeInIndexedFreeLists());
567 gclog_or_tty->print("Number of Blocks: %d\n", free_blocks);
568 if (free_blocks != 0) {
569 gclog_or_tty->print("Av. Block Size: %d\n", total_size/free_blocks);
570 }
571 }
572
573 size_t CompactibleFreeListSpace::numFreeBlocksInIndexedFreeLists() const {
574 size_t res = 0;
575 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
576 debug_only(
577 ssize_t recount = 0;
578 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
579 fc = fc->next()) {
580 recount += 1;
581 }
582 assert(recount == _indexedFreeList[i].count(),
583 "Incorrect count in list");
584 )
585 res += _indexedFreeList[i].count();
586 }
587 return res;
588 }
589
590 size_t CompactibleFreeListSpace::maxChunkSizeInIndexedFreeLists() const {
591 for (size_t i = IndexSetSize - 1; i != 0; i -= IndexSetStride) {
592 if (_indexedFreeList[i].head() != NULL) {
593 assert(_indexedFreeList[i].count() != 0, "Inconsistent FreeList");
594 return (size_t)i;
595 }
596 }
597 return 0;
598 }
599
600 void CompactibleFreeListSpace::set_end(HeapWord* value) {
601 HeapWord* prevEnd = end();
602 assert(prevEnd != value, "unnecessary set_end call");
603 assert(prevEnd == NULL || !BlockOffsetArrayUseUnallocatedBlock || value >= unallocated_block(),
604 "New end is below unallocated block");
605 _end = value;
606 if (prevEnd != NULL) {
607 // Resize the underlying block offset table.
608 _bt.resize(pointer_delta(value, bottom()));
609 if (value <= prevEnd) {
610 assert(!BlockOffsetArrayUseUnallocatedBlock || value >= unallocated_block(),
611 "New end is below unallocated block");
612 } else {
613 // Now, take this new chunk and add it to the free blocks.
614 // Note that the BOT has not yet been updated for this block.
615 size_t newFcSize = pointer_delta(value, prevEnd);
616 // XXX This is REALLY UGLY and should be fixed up. XXX
617 if (!_adaptive_freelists && _smallLinearAllocBlock._ptr == NULL) {
618 // Mark the boundary of the new block in BOT
619 _bt.mark_block(prevEnd, value);
620 // put it all in the linAB
621 if (ParallelGCThreads == 0) {
622 _smallLinearAllocBlock._ptr = prevEnd;
623 _smallLinearAllocBlock._word_size = newFcSize;
624 repairLinearAllocBlock(&_smallLinearAllocBlock);
625 } else { // ParallelGCThreads > 0
626 MutexLockerEx x(parDictionaryAllocLock(),
627 Mutex::_no_safepoint_check_flag);
628 _smallLinearAllocBlock._ptr = prevEnd;
629 _smallLinearAllocBlock._word_size = newFcSize;
630 repairLinearAllocBlock(&_smallLinearAllocBlock);
631 }
632 // Births of chunks put into a LinAB are not recorded. Births
633 // of chunks as they are allocated out of a LinAB are.
634 } else {
635 // Add the block to the free lists, if possible coalescing it
636 // with the last free block, and update the BOT and census data.
637 addChunkToFreeListsAtEndRecordingStats(prevEnd, newFcSize);
638 }
639 }
640 }
641 }
642
643 class FreeListSpace_DCTOC : public Filtering_DCTOC {
644 CompactibleFreeListSpace* _cfls;
645 CMSCollector* _collector;
646 protected:
647 // Override.
648 #define walk_mem_region_with_cl_DECL(ClosureType) \
649 virtual void walk_mem_region_with_cl(MemRegion mr, \
650 HeapWord* bottom, HeapWord* top, \
651 ClosureType* cl); \
652 void walk_mem_region_with_cl_par(MemRegion mr, \
653 HeapWord* bottom, HeapWord* top, \
654 ClosureType* cl); \
655 void walk_mem_region_with_cl_nopar(MemRegion mr, \
656 HeapWord* bottom, HeapWord* top, \
657 ClosureType* cl)
658 walk_mem_region_with_cl_DECL(OopClosure);
659 walk_mem_region_with_cl_DECL(FilteringClosure);
660
661 public:
662 FreeListSpace_DCTOC(CompactibleFreeListSpace* sp,
663 CMSCollector* collector,
664 OopClosure* cl,
665 CardTableModRefBS::PrecisionStyle precision,
666 HeapWord* boundary) :
667 Filtering_DCTOC(sp, cl, precision, boundary),
668 _cfls(sp), _collector(collector) {}
669 };
670
671 // We de-virtualize the block-related calls below, since we know that our
672 // space is a CompactibleFreeListSpace.
673
674 #define FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(ClosureType) \
675 void FreeListSpace_DCTOC::walk_mem_region_with_cl(MemRegion mr, \
676 HeapWord* bottom, \
677 HeapWord* top, \
678 ClosureType* cl) { \
679 bool is_par = SharedHeap::heap()->n_par_threads() > 0; \
680 if (is_par) { \
681 assert(SharedHeap::heap()->n_par_threads() == \
682 SharedHeap::heap()->workers()->active_workers(), "Mismatch"); \
683 walk_mem_region_with_cl_par(mr, bottom, top, cl); \
684 } else { \
685 walk_mem_region_with_cl_nopar(mr, bottom, top, cl); \
686 } \
687 } \
688 void FreeListSpace_DCTOC::walk_mem_region_with_cl_par(MemRegion mr, \
689 HeapWord* bottom, \
690 HeapWord* top, \
691 ClosureType* cl) { \
692 /* Skip parts that are before "mr", in case "block_start" sent us \
693 back too far. */ \
694 HeapWord* mr_start = mr.start(); \
695 size_t bot_size = _cfls->CompactibleFreeListSpace::block_size(bottom); \
696 HeapWord* next = bottom + bot_size; \
697 while (next < mr_start) { \
698 bottom = next; \
699 bot_size = _cfls->CompactibleFreeListSpace::block_size(bottom); \
700 next = bottom + bot_size; \
701 } \
702 \
703 while (bottom < top) { \
704 if (_cfls->CompactibleFreeListSpace::block_is_obj(bottom) && \
705 !_cfls->CompactibleFreeListSpace::obj_allocated_since_save_marks( \
706 oop(bottom)) && \
707 !_collector->CMSCollector::is_dead_obj(oop(bottom))) { \
708 size_t word_sz = oop(bottom)->oop_iterate(cl, mr); \
709 bottom += _cfls->adjustObjectSize(word_sz); \
710 } else { \
711 bottom += _cfls->CompactibleFreeListSpace::block_size(bottom); \
712 } \
713 } \
714 } \
715 void FreeListSpace_DCTOC::walk_mem_region_with_cl_nopar(MemRegion mr, \
716 HeapWord* bottom, \
717 HeapWord* top, \
718 ClosureType* cl) { \
719 /* Skip parts that are before "mr", in case "block_start" sent us \
720 back too far. */ \
721 HeapWord* mr_start = mr.start(); \
722 size_t bot_size = _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
723 HeapWord* next = bottom + bot_size; \
724 while (next < mr_start) { \
725 bottom = next; \
726 bot_size = _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
727 next = bottom + bot_size; \
728 } \
729 \
730 while (bottom < top) { \
731 if (_cfls->CompactibleFreeListSpace::block_is_obj_nopar(bottom) && \
732 !_cfls->CompactibleFreeListSpace::obj_allocated_since_save_marks( \
733 oop(bottom)) && \
734 !_collector->CMSCollector::is_dead_obj(oop(bottom))) { \
735 size_t word_sz = oop(bottom)->oop_iterate(cl, mr); \
736 bottom += _cfls->adjustObjectSize(word_sz); \
737 } else { \
738 bottom += _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
739 } \
740 } \
741 }
742
743 // (There are only two of these, rather than N, because the split is due
744 // only to the introduction of the FilteringClosure, a local part of the
745 // impl of this abstraction.)
746 FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(OopClosure)
747 FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(FilteringClosure)
748
749 DirtyCardToOopClosure*
750 CompactibleFreeListSpace::new_dcto_cl(OopClosure* cl,
751 CardTableModRefBS::PrecisionStyle precision,
752 HeapWord* boundary) {
753 return new FreeListSpace_DCTOC(this, _collector, cl, precision, boundary);
754 }
755
756
757 // Note on locking for the space iteration functions:
758 // since the collector's iteration activities are concurrent with
759 // allocation activities by mutators, absent a suitable mutual exclusion
760 // mechanism the iterators may go awry. For instace a block being iterated
761 // may suddenly be allocated or divided up and part of it allocated and
762 // so on.
763
764 // Apply the given closure to each block in the space.
765 void CompactibleFreeListSpace::blk_iterate_careful(BlkClosureCareful* cl) {
766 assert_lock_strong(freelistLock());
767 HeapWord *cur, *limit;
768 for (cur = bottom(), limit = end(); cur < limit;
769 cur += cl->do_blk_careful(cur));
770 }
771
772 // Apply the given closure to each block in the space.
773 void CompactibleFreeListSpace::blk_iterate(BlkClosure* cl) {
774 assert_lock_strong(freelistLock());
775 HeapWord *cur, *limit;
776 for (cur = bottom(), limit = end(); cur < limit;
777 cur += cl->do_blk(cur));
778 }
779
780 // Apply the given closure to each oop in the space.
781 void CompactibleFreeListSpace::oop_iterate(OopClosure* cl) {
782 assert_lock_strong(freelistLock());
783 HeapWord *cur, *limit;
784 size_t curSize;
785 for (cur = bottom(), limit = end(); cur < limit;
786 cur += curSize) {
787 curSize = block_size(cur);
788 if (block_is_obj(cur)) {
789 oop(cur)->oop_iterate(cl);
790 }
791 }
792 }
793
794 // Apply the given closure to each oop in the space \intersect memory region.
795 void CompactibleFreeListSpace::oop_iterate(MemRegion mr, OopClosure* cl) {
796 assert_lock_strong(freelistLock());
797 if (is_empty()) {
798 return;
799 }
800 MemRegion cur = MemRegion(bottom(), end());
801 mr = mr.intersection(cur);
802 if (mr.is_empty()) {
803 return;
804 }
805 if (mr.equals(cur)) {
806 oop_iterate(cl);
807 return;
808 }
809 assert(mr.end() <= end(), "just took an intersection above");
810 HeapWord* obj_addr = block_start(mr.start());
811 HeapWord* t = mr.end();
812
813 SpaceMemRegionOopsIterClosure smr_blk(cl, mr);
814 if (block_is_obj(obj_addr)) {
815 // Handle first object specially.
816 oop obj = oop(obj_addr);
817 obj_addr += adjustObjectSize(obj->oop_iterate(&smr_blk));
818 } else {
819 FreeChunk* fc = (FreeChunk*)obj_addr;
820 obj_addr += fc->size();
821 }
822 while (obj_addr < t) {
823 HeapWord* obj = obj_addr;
824 obj_addr += block_size(obj_addr);
825 // If "obj_addr" is not greater than top, then the
826 // entire object "obj" is within the region.
827 if (obj_addr <= t) {
828 if (block_is_obj(obj)) {
829 oop(obj)->oop_iterate(cl);
830 }
831 } else {
832 // "obj" extends beyond end of region
833 if (block_is_obj(obj)) {
834 oop(obj)->oop_iterate(&smr_blk);
835 }
836 break;
837 }
838 }
839 }
840
841 // NOTE: In the following methods, in order to safely be able to
842 // apply the closure to an object, we need to be sure that the
843 // object has been initialized. We are guaranteed that an object
844 // is initialized if we are holding the Heap_lock with the
845 // world stopped.
846 void CompactibleFreeListSpace::verify_objects_initialized() const {
847 if (is_init_completed()) {
848 assert_locked_or_safepoint(Heap_lock);
849 if (Universe::is_fully_initialized()) {
850 guarantee(SafepointSynchronize::is_at_safepoint(),
851 "Required for objects to be initialized");
852 }
853 } // else make a concession at vm start-up
854 }
855
856 // Apply the given closure to each object in the space
857 void CompactibleFreeListSpace::object_iterate(ObjectClosure* blk) {
858 assert_lock_strong(freelistLock());
859 NOT_PRODUCT(verify_objects_initialized());
860 HeapWord *cur, *limit;
861 size_t curSize;
862 for (cur = bottom(), limit = end(); cur < limit;
863 cur += curSize) {
864 curSize = block_size(cur);
865 if (block_is_obj(cur)) {
866 blk->do_object(oop(cur));
867 }
868 }
869 }
870
871 // Apply the given closure to each live object in the space
872 // The usage of CompactibleFreeListSpace
873 // by the ConcurrentMarkSweepGeneration for concurrent GC's allows
874 // objects in the space with references to objects that are no longer
875 // valid. For example, an object may reference another object
876 // that has already been sweep up (collected). This method uses
877 // obj_is_alive() to determine whether it is safe to apply the closure to
878 // an object. See obj_is_alive() for details on how liveness of an
879 // object is decided.
880
881 void CompactibleFreeListSpace::safe_object_iterate(ObjectClosure* blk) {
882 assert_lock_strong(freelistLock());
883 NOT_PRODUCT(verify_objects_initialized());
884 HeapWord *cur, *limit;
885 size_t curSize;
886 for (cur = bottom(), limit = end(); cur < limit;
887 cur += curSize) {
888 curSize = block_size(cur);
889 if (block_is_obj(cur) && obj_is_alive(cur)) {
890 blk->do_object(oop(cur));
891 }
892 }
893 }
894
895 void CompactibleFreeListSpace::object_iterate_mem(MemRegion mr,
896 UpwardsObjectClosure* cl) {
897 assert_locked(freelistLock());
898 NOT_PRODUCT(verify_objects_initialized());
899 Space::object_iterate_mem(mr, cl);
900 }
901
902 // Callers of this iterator beware: The closure application should
903 // be robust in the face of uninitialized objects and should (always)
904 // return a correct size so that the next addr + size below gives us a
905 // valid block boundary. [See for instance,
906 // ScanMarkedObjectsAgainCarefullyClosure::do_object_careful()
907 // in ConcurrentMarkSweepGeneration.cpp.]
908 HeapWord*
909 CompactibleFreeListSpace::object_iterate_careful(ObjectClosureCareful* cl) {
910 assert_lock_strong(freelistLock());
911 HeapWord *addr, *last;
912 size_t size;
913 for (addr = bottom(), last = end();
914 addr < last; addr += size) {
915 FreeChunk* fc = (FreeChunk*)addr;
916 if (fc->is_free()) {
917 // Since we hold the free list lock, which protects direct
918 // allocation in this generation by mutators, a free object
919 // will remain free throughout this iteration code.
920 size = fc->size();
921 } else {
922 // Note that the object need not necessarily be initialized,
923 // because (for instance) the free list lock does NOT protect
924 // object initialization. The closure application below must
925 // therefore be correct in the face of uninitialized objects.
926 size = cl->do_object_careful(oop(addr));
927 if (size == 0) {
928 // An unparsable object found. Signal early termination.
929 return addr;
930 }
931 }
932 }
933 return NULL;
934 }
935
936 // Callers of this iterator beware: The closure application should
937 // be robust in the face of uninitialized objects and should (always)
938 // return a correct size so that the next addr + size below gives us a
939 // valid block boundary. [See for instance,
940 // ScanMarkedObjectsAgainCarefullyClosure::do_object_careful()
941 // in ConcurrentMarkSweepGeneration.cpp.]
942 HeapWord*
943 CompactibleFreeListSpace::object_iterate_careful_m(MemRegion mr,
944 ObjectClosureCareful* cl) {
945 assert_lock_strong(freelistLock());
946 // Can't use used_region() below because it may not necessarily
947 // be the same as [bottom(),end()); although we could
948 // use [used_region().start(),round_to(used_region().end(),CardSize)),
949 // that appears too cumbersome, so we just do the simpler check
950 // in the assertion below.
951 assert(!mr.is_empty() && MemRegion(bottom(),end()).contains(mr),
952 "mr should be non-empty and within used space");
953 HeapWord *addr, *end;
954 size_t size;
955 for (addr = block_start_careful(mr.start()), end = mr.end();
956 addr < end; addr += size) {
957 FreeChunk* fc = (FreeChunk*)addr;
958 if (fc->is_free()) {
959 // Since we hold the free list lock, which protects direct
960 // allocation in this generation by mutators, a free object
961 // will remain free throughout this iteration code.
962 size = fc->size();
963 } else {
964 // Note that the object need not necessarily be initialized,
965 // because (for instance) the free list lock does NOT protect
966 // object initialization. The closure application below must
967 // therefore be correct in the face of uninitialized objects.
968 size = cl->do_object_careful_m(oop(addr), mr);
969 if (size == 0) {
970 // An unparsable object found. Signal early termination.
971 return addr;
972 }
973 }
974 }
975 return NULL;
976 }
977
978
979 HeapWord* CompactibleFreeListSpace::block_start_const(const void* p) const {
980 NOT_PRODUCT(verify_objects_initialized());
981 return _bt.block_start(p);
982 }
983
984 HeapWord* CompactibleFreeListSpace::block_start_careful(const void* p) const {
985 return _bt.block_start_careful(p);
986 }
987
988 size_t CompactibleFreeListSpace::block_size(const HeapWord* p) const {
989 NOT_PRODUCT(verify_objects_initialized());
990 // This must be volatile, or else there is a danger that the compiler
991 // will compile the code below into a sometimes-infinite loop, by keeping
992 // the value read the first time in a register.
993 while (true) {
994 // We must do this until we get a consistent view of the object.
995 if (FreeChunk::indicatesFreeChunk(p)) {
996 volatile FreeChunk* fc = (volatile FreeChunk*)p;
997 size_t res = fc->size();
998 // If the object is still a free chunk, return the size, else it
999 // has been allocated so try again.
1000 if (FreeChunk::indicatesFreeChunk(p)) {
1001 assert(res != 0, "Block size should not be 0");
1002 return res;
1003 }
1004 } else {
1005 // must read from what 'p' points to in each loop.
1006 klassOop k = ((volatile oopDesc*)p)->klass_or_null();
1007 if (k != NULL) {
1008 assert(k->is_oop(true /* ignore mark word */), "Should be klass oop");
1009 oop o = (oop)p;
1010 assert(o->is_parsable(), "Should be parsable");
1011 assert(o->is_oop(true /* ignore mark word */), "Should be an oop.");
1012 size_t res = o->size_given_klass(k->klass_part());
1013 res = adjustObjectSize(res);
1014 assert(res != 0, "Block size should not be 0");
1015 return res;
1016 }
1017 }
1018 }
1019 }
1020
1021 // A variant of the above that uses the Printezis bits for
1022 // unparsable but allocated objects. This avoids any possible
1023 // stalls waiting for mutators to initialize objects, and is
1024 // thus potentially faster than the variant above. However,
1025 // this variant may return a zero size for a block that is
1026 // under mutation and for which a consistent size cannot be
1027 // inferred without stalling; see CMSCollector::block_size_if_printezis_bits().
1028 size_t CompactibleFreeListSpace::block_size_no_stall(HeapWord* p,
1029 const CMSCollector* c)
1030 const {
1031 assert(MemRegion(bottom(), end()).contains(p), "p not in space");
1032 // This must be volatile, or else there is a danger that the compiler
1033 // will compile the code below into a sometimes-infinite loop, by keeping
1034 // the value read the first time in a register.
1035 DEBUG_ONLY(uint loops = 0;)
1036 while (true) {
1037 // We must do this until we get a consistent view of the object.
1038 if (FreeChunk::indicatesFreeChunk(p)) {
1039 volatile FreeChunk* fc = (volatile FreeChunk*)p;
1040 size_t res = fc->size();
1041 if (FreeChunk::indicatesFreeChunk(p)) {
1042 assert(res != 0, "Block size should not be 0");
1043 assert(loops == 0, "Should be 0");
1044 return res;
1045 }
1046 } else {
1047 // must read from what 'p' points to in each loop.
1048 klassOop k = ((volatile oopDesc*)p)->klass_or_null();
1049 // We trust the size of any object that has a non-NULL
1050 // klass and (for those in the perm gen) is parsable
1051 // -- irrespective of its conc_safe-ty.
1052 if (k != NULL && ((oopDesc*)p)->is_parsable()) {
1053 assert(k->is_oop(), "Should really be klass oop.");
1054 oop o = (oop)p;
1055 assert(o->is_oop(), "Should be an oop");
1056 size_t res = o->size_given_klass(k->klass_part());
1057 res = adjustObjectSize(res);
1058 assert(res != 0, "Block size should not be 0");
1059 return res;
1060 } else {
1061 // May return 0 if P-bits not present.
1062 return c->block_size_if_printezis_bits(p);
1063 }
1064 }
1065 assert(loops == 0, "Can loop at most once");
1066 DEBUG_ONLY(loops++;)
1067 }
1068 }
1069
1070 size_t CompactibleFreeListSpace::block_size_nopar(const HeapWord* p) const {
1071 NOT_PRODUCT(verify_objects_initialized());
1072 assert(MemRegion(bottom(), end()).contains(p), "p not in space");
1073 FreeChunk* fc = (FreeChunk*)p;
1074 if (fc->is_free()) {
1075 return fc->size();
1076 } else {
1077 // Ignore mark word because this may be a recently promoted
1078 // object whose mark word is used to chain together grey
1079 // objects (the last one would have a null value).
1080 assert(oop(p)->is_oop(true), "Should be an oop");
1081 return adjustObjectSize(oop(p)->size());
1082 }
1083 }
1084
1085 // This implementation assumes that the property of "being an object" is
1086 // stable. But being a free chunk may not be (because of parallel
1087 // promotion.)
1088 bool CompactibleFreeListSpace::block_is_obj(const HeapWord* p) const {
1089 FreeChunk* fc = (FreeChunk*)p;
1090 assert(is_in_reserved(p), "Should be in space");
1091 // When doing a mark-sweep-compact of the CMS generation, this
1092 // assertion may fail because prepare_for_compaction() uses
1093 // space that is garbage to maintain information on ranges of
1094 // live objects so that these live ranges can be moved as a whole.
1095 // Comment out this assertion until that problem can be solved
1096 // (i.e., that the block start calculation may look at objects
1097 // at address below "p" in finding the object that contains "p"
1098 // and those objects (if garbage) may have been modified to hold
1099 // live range information.
1100 // assert(CollectedHeap::use_parallel_gc_threads() || _bt.block_start(p) == p,
1101 // "Should be a block boundary");
1102 if (FreeChunk::indicatesFreeChunk(p)) return false;
1103 klassOop k = oop(p)->klass_or_null();
1104 if (k != NULL) {
1105 // Ignore mark word because it may have been used to
1106 // chain together promoted objects (the last one
1107 // would have a null value).
1108 assert(oop(p)->is_oop(true), "Should be an oop");
1109 return true;
1110 } else {
1111 return false; // Was not an object at the start of collection.
1112 }
1113 }
1114
1115 // Check if the object is alive. This fact is checked either by consulting
1116 // the main marking bitmap in the sweeping phase or, if it's a permanent
1117 // generation and we're not in the sweeping phase, by checking the
1118 // perm_gen_verify_bit_map where we store the "deadness" information if
1119 // we did not sweep the perm gen in the most recent previous GC cycle.
1120 bool CompactibleFreeListSpace::obj_is_alive(const HeapWord* p) const {
1121 assert(SafepointSynchronize::is_at_safepoint() || !is_init_completed(),
1122 "Else races are possible");
1123 assert(block_is_obj(p), "The address should point to an object");
1124
1125 // If we're sweeping, we use object liveness information from the main bit map
1126 // for both perm gen and old gen.
1127 // We don't need to lock the bitmap (live_map or dead_map below), because
1128 // EITHER we are in the middle of the sweeping phase, and the
1129 // main marking bit map (live_map below) is locked,
1130 // OR we're in other phases and perm_gen_verify_bit_map (dead_map below)
1131 // is stable, because it's mutated only in the sweeping phase.
1132 // NOTE: This method is also used by jmap where, if class unloading is
1133 // off, the results can return "false" for legitimate perm objects,
1134 // when we are not in the midst of a sweeping phase, which can result
1135 // in jmap not reporting certain perm gen objects. This will be moot
1136 // if/when the perm gen goes away in the future.
1137 if (_collector->abstract_state() == CMSCollector::Sweeping) {
1138 CMSBitMap* live_map = _collector->markBitMap();
1139 return live_map->par_isMarked((HeapWord*) p);
1140 } else {
1141 // If we're not currently sweeping and we haven't swept the perm gen in
1142 // the previous concurrent cycle then we may have dead but unswept objects
1143 // in the perm gen. In this case, we use the "deadness" information
1144 // that we had saved in perm_gen_verify_bit_map at the last sweep.
1145 if (!CMSClassUnloadingEnabled && _collector->_permGen->reserved().contains(p)) {
1146 if (_collector->verifying()) {
1147 CMSBitMap* dead_map = _collector->perm_gen_verify_bit_map();
1148 // Object is marked in the dead_map bitmap at the previous sweep
1149 // when we know that it's dead; if the bitmap is not allocated then
1150 // the object is alive.
1151 return (dead_map->sizeInBits() == 0) // bit_map has been allocated
1152 || !dead_map->par_isMarked((HeapWord*) p);
1153 } else {
1154 return false; // We can't say for sure if it's live, so we say that it's dead.
1155 }
1156 }
1157 }
1158 return true;
1159 }
1160
1161 bool CompactibleFreeListSpace::block_is_obj_nopar(const HeapWord* p) const {
1162 FreeChunk* fc = (FreeChunk*)p;
1163 assert(is_in_reserved(p), "Should be in space");
1164 assert(_bt.block_start(p) == p, "Should be a block boundary");
1165 if (!fc->is_free()) {
1166 // Ignore mark word because it may have been used to
1167 // chain together promoted objects (the last one
1168 // would have a null value).
1169 assert(oop(p)->is_oop(true), "Should be an oop");
1170 return true;
1171 }
1172 return false;
1173 }
1174
1175 // "MT-safe but not guaranteed MT-precise" (TM); you may get an
1176 // approximate answer if you don't hold the freelistlock when you call this.
1177 size_t CompactibleFreeListSpace::totalSizeInIndexedFreeLists() const {
1178 size_t size = 0;
1179 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
1180 debug_only(
1181 // We may be calling here without the lock in which case we
1182 // won't do this modest sanity check.
1183 if (freelistLock()->owned_by_self()) {
1184 size_t total_list_size = 0;
1185 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
1186 fc = fc->next()) {
1187 total_list_size += i;
1188 }
1189 assert(total_list_size == i * _indexedFreeList[i].count(),
1190 "Count in list is incorrect");
1191 }
1192 )
1193 size += i * _indexedFreeList[i].count();
1194 }
1195 return size;
1196 }
1197
1198 HeapWord* CompactibleFreeListSpace::par_allocate(size_t size) {
1199 MutexLockerEx x(freelistLock(), Mutex::_no_safepoint_check_flag);
1200 return allocate(size);
1201 }
1202
1203 HeapWord*
1204 CompactibleFreeListSpace::getChunkFromSmallLinearAllocBlockRemainder(size_t size) {
1205 return getChunkFromLinearAllocBlockRemainder(&_smallLinearAllocBlock, size);
1206 }
1207
1208 HeapWord* CompactibleFreeListSpace::allocate(size_t size) {
1209 assert_lock_strong(freelistLock());
1210 HeapWord* res = NULL;
1211 assert(size == adjustObjectSize(size),
1212 "use adjustObjectSize() before calling into allocate()");
1213
1214 if (_adaptive_freelists) {
1215 res = allocate_adaptive_freelists(size);
1216 } else { // non-adaptive free lists
1217 res = allocate_non_adaptive_freelists(size);
1218 }
1219
1220 if (res != NULL) {
1221 // check that res does lie in this space!
1222 assert(is_in_reserved(res), "Not in this space!");
1223 assert(is_aligned((void*)res), "alignment check");
1224
1225 FreeChunk* fc = (FreeChunk*)res;
1226 fc->markNotFree();
1227 assert(!fc->is_free(), "shouldn't be marked free");
1228 assert(oop(fc)->klass_or_null() == NULL, "should look uninitialized");
1229 // Verify that the block offset table shows this to
1230 // be a single block, but not one which is unallocated.
1231 _bt.verify_single_block(res, size);
1232 _bt.verify_not_unallocated(res, size);
1233 // mangle a just allocated object with a distinct pattern.
1234 debug_only(fc->mangleAllocated(size));
1235 }
1236
1237 return res;
1238 }
1239
1240 HeapWord* CompactibleFreeListSpace::allocate_non_adaptive_freelists(size_t size) {
1241 HeapWord* res = NULL;
1242 // try and use linear allocation for smaller blocks
1243 if (size < _smallLinearAllocBlock._allocation_size_limit) {
1244 // if successful, the following also adjusts block offset table
1245 res = getChunkFromSmallLinearAllocBlock(size);
1246 }
1247 // Else triage to indexed lists for smaller sizes
1248 if (res == NULL) {
1249 if (size < SmallForDictionary) {
1250 res = (HeapWord*) getChunkFromIndexedFreeList(size);
1251 } else {
1252 // else get it from the big dictionary; if even this doesn't
1253 // work we are out of luck.
1254 res = (HeapWord*)getChunkFromDictionaryExact(size);
1255 }
1256 }
1257
1258 return res;
1259 }
1260
1261 HeapWord* CompactibleFreeListSpace::allocate_adaptive_freelists(size_t size) {
1262 assert_lock_strong(freelistLock());
1263 HeapWord* res = NULL;
1264 assert(size == adjustObjectSize(size),
1265 "use adjustObjectSize() before calling into allocate()");
1266
1267 // Strategy
1268 // if small
1269 // exact size from small object indexed list if small
1270 // small or large linear allocation block (linAB) as appropriate
1271 // take from lists of greater sized chunks
1272 // else
1273 // dictionary
1274 // small or large linear allocation block if it has the space
1275 // Try allocating exact size from indexTable first
1276 if (size < IndexSetSize) {
1277 res = (HeapWord*) getChunkFromIndexedFreeList(size);
1278 if(res != NULL) {
1279 assert(res != (HeapWord*)_indexedFreeList[size].head(),
1280 "Not removed from free list");
1281 // no block offset table adjustment is necessary on blocks in
1282 // the indexed lists.
1283
1284 // Try allocating from the small LinAB
1285 } else if (size < _smallLinearAllocBlock._allocation_size_limit &&
1286 (res = getChunkFromSmallLinearAllocBlock(size)) != NULL) {
1287 // if successful, the above also adjusts block offset table
1288 // Note that this call will refill the LinAB to
1289 // satisfy the request. This is different that
1290 // evm.
1291 // Don't record chunk off a LinAB? smallSplitBirth(size);
1292 } else {
1293 // Raid the exact free lists larger than size, even if they are not
1294 // overpopulated.
1295 res = (HeapWord*) getChunkFromGreater(size);
1296 }
1297 } else {
1298 // Big objects get allocated directly from the dictionary.
1299 res = (HeapWord*) getChunkFromDictionaryExact(size);
1300 if (res == NULL) {
1301 // Try hard not to fail since an allocation failure will likely
1302 // trigger a synchronous GC. Try to get the space from the
1303 // allocation blocks.
1304 res = getChunkFromSmallLinearAllocBlockRemainder(size);
1305 }
1306 }
1307
1308 return res;
1309 }
1310
1311 // A worst-case estimate of the space required (in HeapWords) to expand the heap
1312 // when promoting obj.
1313 size_t CompactibleFreeListSpace::expansionSpaceRequired(size_t obj_size) const {
1314 // Depending on the object size, expansion may require refilling either a
1315 // bigLAB or a smallLAB plus refilling a PromotionInfo object. MinChunkSize
1316 // is added because the dictionary may over-allocate to avoid fragmentation.
1317 size_t space = obj_size;
1318 if (!_adaptive_freelists) {
1319 space = MAX2(space, _smallLinearAllocBlock._refillSize);
1320 }
1321 space += _promoInfo.refillSize() + 2 * MinChunkSize;
1322 return space;
1323 }
1324
1325 FreeChunk* CompactibleFreeListSpace::getChunkFromGreater(size_t numWords) {
1326 FreeChunk* ret;
1327
1328 assert(numWords >= MinChunkSize, "Size is less than minimum");
1329 assert(linearAllocationWouldFail() || bestFitFirst(),
1330 "Should not be here");
1331
1332 size_t i;
1333 size_t currSize = numWords + MinChunkSize;
1334 assert(currSize % MinObjAlignment == 0, "currSize should be aligned");
1335 for (i = currSize; i < IndexSetSize; i += IndexSetStride) {
1336 FreeList<FreeChunk>* fl = &_indexedFreeList[i];
1337 if (fl->head()) {
1338 ret = getFromListGreater(fl, numWords);
1339 assert(ret == NULL || ret->is_free(), "Should be returning a free chunk");
1340 return ret;
1341 }
1342 }
1343
1344 currSize = MAX2((size_t)SmallForDictionary,
1345 (size_t)(numWords + MinChunkSize));
1346
1347 /* Try to get a chunk that satisfies request, while avoiding
1348 fragmentation that can't be handled. */
1349 {
1350 ret = dictionary()->get_chunk(currSize);
1351 if (ret != NULL) {
1352 assert(ret->size() - numWords >= MinChunkSize,
1353 "Chunk is too small");
1354 _bt.allocated((HeapWord*)ret, ret->size());
1355 /* Carve returned chunk. */
1356 (void) splitChunkAndReturnRemainder(ret, numWords);
1357 /* Label this as no longer a free chunk. */
1358 assert(ret->is_free(), "This chunk should be free");
1359 ret->link_prev(NULL);
1360 }
1361 assert(ret == NULL || ret->is_free(), "Should be returning a free chunk");
1362 return ret;
1363 }
1364 ShouldNotReachHere();
1365 }
1366
1367 bool CompactibleFreeListSpace::verifyChunkInIndexedFreeLists(FreeChunk* fc) const {
1368 assert(fc->size() < IndexSetSize, "Size of chunk is too large");
1369 return _indexedFreeList[fc->size()].verify_chunk_in_free_list(fc);
1370 }
1371
1372 bool CompactibleFreeListSpace::verify_chunk_is_linear_alloc_block(FreeChunk* fc) const {
1373 assert((_smallLinearAllocBlock._ptr != (HeapWord*)fc) ||
1374 (_smallLinearAllocBlock._word_size == fc->size()),
1375 "Linear allocation block shows incorrect size");
1376 return ((_smallLinearAllocBlock._ptr == (HeapWord*)fc) &&
1377 (_smallLinearAllocBlock._word_size == fc->size()));
1378 }
1379
1380 // Check if the purported free chunk is present either as a linear
1381 // allocation block, the size-indexed table of (smaller) free blocks,
1382 // or the larger free blocks kept in the binary tree dictionary.
1383 bool CompactibleFreeListSpace::verify_chunk_in_free_list(FreeChunk* fc) const {
1384 if (verify_chunk_is_linear_alloc_block(fc)) {
1385 return true;
1386 } else if (fc->size() < IndexSetSize) {
1387 return verifyChunkInIndexedFreeLists(fc);
1388 } else {
1389 return dictionary()->verify_chunk_in_free_list(fc);
1390 }
1391 }
1392
1393 #ifndef PRODUCT
1394 void CompactibleFreeListSpace::assert_locked() const {
1395 CMSLockVerifier::assert_locked(freelistLock(), parDictionaryAllocLock());
1396 }
1397
1398 void CompactibleFreeListSpace::assert_locked(const Mutex* lock) const {
1399 CMSLockVerifier::assert_locked(lock);
1400 }
1401 #endif
1402
1403 FreeChunk* CompactibleFreeListSpace::allocateScratch(size_t size) {
1404 // In the parallel case, the main thread holds the free list lock
1405 // on behalf the parallel threads.
1406 FreeChunk* fc;
1407 {
1408 // If GC is parallel, this might be called by several threads.
1409 // This should be rare enough that the locking overhead won't affect
1410 // the sequential code.
1411 MutexLockerEx x(parDictionaryAllocLock(),
1412 Mutex::_no_safepoint_check_flag);
1413 fc = getChunkFromDictionary(size);
1414 }
1415 if (fc != NULL) {
1416 fc->dontCoalesce();
1417 assert(fc->is_free(), "Should be free, but not coalescable");
1418 // Verify that the block offset table shows this to
1419 // be a single block, but not one which is unallocated.
1420 _bt.verify_single_block((HeapWord*)fc, fc->size());
1421 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
1422 }
1423 return fc;
1424 }
1425
1426 oop CompactibleFreeListSpace::promote(oop obj, size_t obj_size) {
1427 assert(obj_size == (size_t)obj->size(), "bad obj_size passed in");
1428 assert_locked();
1429
1430 // if we are tracking promotions, then first ensure space for
1431 // promotion (including spooling space for saving header if necessary).
1432 // then allocate and copy, then track promoted info if needed.
1433 // When tracking (see PromotionInfo::track()), the mark word may
1434 // be displaced and in this case restoration of the mark word
1435 // occurs in the (oop_since_save_marks_)iterate phase.
1436 if (_promoInfo.tracking() && !_promoInfo.ensure_spooling_space()) {
1437 return NULL;
1438 }
1439 // Call the allocate(size_t, bool) form directly to avoid the
1440 // additional call through the allocate(size_t) form. Having
1441 // the compile inline the call is problematic because allocate(size_t)
1442 // is a virtual method.
1443 HeapWord* res = allocate(adjustObjectSize(obj_size));
1444 if (res != NULL) {
1445 Copy::aligned_disjoint_words((HeapWord*)obj, res, obj_size);
1446 // if we should be tracking promotions, do so.
1447 if (_promoInfo.tracking()) {
1448 _promoInfo.track((PromotedObject*)res);
1449 }
1450 }
1451 return oop(res);
1452 }
1453
1454 HeapWord*
1455 CompactibleFreeListSpace::getChunkFromSmallLinearAllocBlock(size_t size) {
1456 assert_locked();
1457 assert(size >= MinChunkSize, "minimum chunk size");
1458 assert(size < _smallLinearAllocBlock._allocation_size_limit,
1459 "maximum from smallLinearAllocBlock");
1460 return getChunkFromLinearAllocBlock(&_smallLinearAllocBlock, size);
1461 }
1462
1463 HeapWord*
1464 CompactibleFreeListSpace::getChunkFromLinearAllocBlock(LinearAllocBlock *blk,
1465 size_t size) {
1466 assert_locked();
1467 assert(size >= MinChunkSize, "too small");
1468 HeapWord* res = NULL;
1469 // Try to do linear allocation from blk, making sure that
1470 if (blk->_word_size == 0) {
1471 // We have probably been unable to fill this either in the prologue or
1472 // when it was exhausted at the last linear allocation. Bail out until
1473 // next time.
1474 assert(blk->_ptr == NULL, "consistency check");
1475 return NULL;
1476 }
1477 assert(blk->_word_size != 0 && blk->_ptr != NULL, "consistency check");
1478 res = getChunkFromLinearAllocBlockRemainder(blk, size);
1479 if (res != NULL) return res;
1480
1481 // about to exhaust this linear allocation block
1482 if (blk->_word_size == size) { // exactly satisfied
1483 res = blk->_ptr;
1484 _bt.allocated(res, blk->_word_size);
1485 } else if (size + MinChunkSize <= blk->_refillSize) {
1486 size_t sz = blk->_word_size;
1487 // Update _unallocated_block if the size is such that chunk would be
1488 // returned to the indexed free list. All other chunks in the indexed
1489 // free lists are allocated from the dictionary so that _unallocated_block
1490 // has already been adjusted for them. Do it here so that the cost
1491 // for all chunks added back to the indexed free lists.
1492 if (sz < SmallForDictionary) {
1493 _bt.allocated(blk->_ptr, sz);
1494 }
1495 // Return the chunk that isn't big enough, and then refill below.
1496 addChunkToFreeLists(blk->_ptr, sz);
1497 split_birth(sz);
1498 // Don't keep statistics on adding back chunk from a LinAB.
1499 } else {
1500 // A refilled block would not satisfy the request.
1501 return NULL;
1502 }
1503
1504 blk->_ptr = NULL; blk->_word_size = 0;
1505 refillLinearAllocBlock(blk);
1506 assert(blk->_ptr == NULL || blk->_word_size >= size + MinChunkSize,
1507 "block was replenished");
1508 if (res != NULL) {
1509 split_birth(size);
1510 repairLinearAllocBlock(blk);
1511 } else if (blk->_ptr != NULL) {
1512 res = blk->_ptr;
1513 size_t blk_size = blk->_word_size;
1514 blk->_word_size -= size;
1515 blk->_ptr += size;
1516 split_birth(size);
1517 repairLinearAllocBlock(blk);
1518 // Update BOT last so that other (parallel) GC threads see a consistent
1519 // view of the BOT and free blocks.
1520 // Above must occur before BOT is updated below.
1521 OrderAccess::storestore();
1522 _bt.split_block(res, blk_size, size); // adjust block offset table
1523 }
1524 return res;
1525 }
1526
1527 HeapWord* CompactibleFreeListSpace::getChunkFromLinearAllocBlockRemainder(
1528 LinearAllocBlock* blk,
1529 size_t size) {
1530 assert_locked();
1531 assert(size >= MinChunkSize, "too small");
1532
1533 HeapWord* res = NULL;
1534 // This is the common case. Keep it simple.
1535 if (blk->_word_size >= size + MinChunkSize) {
1536 assert(blk->_ptr != NULL, "consistency check");
1537 res = blk->_ptr;
1538 // Note that the BOT is up-to-date for the linAB before allocation. It
1539 // indicates the start of the linAB. The split_block() updates the
1540 // BOT for the linAB after the allocation (indicates the start of the
1541 // next chunk to be allocated).
1542 size_t blk_size = blk->_word_size;
1543 blk->_word_size -= size;
1544 blk->_ptr += size;
1545 split_birth(size);
1546 repairLinearAllocBlock(blk);
1547 // Update BOT last so that other (parallel) GC threads see a consistent
1548 // view of the BOT and free blocks.
1549 // Above must occur before BOT is updated below.
1550 OrderAccess::storestore();
1551 _bt.split_block(res, blk_size, size); // adjust block offset table
1552 _bt.allocated(res, size);
1553 }
1554 return res;
1555 }
1556
1557 FreeChunk*
1558 CompactibleFreeListSpace::getChunkFromIndexedFreeList(size_t size) {
1559 assert_locked();
1560 assert(size < SmallForDictionary, "just checking");
1561 FreeChunk* res;
1562 res = _indexedFreeList[size].get_chunk_at_head();
1563 if (res == NULL) {
1564 res = getChunkFromIndexedFreeListHelper(size);
1565 }
1566 _bt.verify_not_unallocated((HeapWord*) res, size);
1567 assert(res == NULL || res->size() == size, "Incorrect block size");
1568 return res;
1569 }
1570
1571 FreeChunk*
1572 CompactibleFreeListSpace::getChunkFromIndexedFreeListHelper(size_t size,
1573 bool replenish) {
1574 assert_locked();
1575 FreeChunk* fc = NULL;
1576 if (size < SmallForDictionary) {
1577 assert(_indexedFreeList[size].head() == NULL ||
1578 _indexedFreeList[size].surplus() <= 0,
1579 "List for this size should be empty or under populated");
1580 // Try best fit in exact lists before replenishing the list
1581 if (!bestFitFirst() || (fc = bestFitSmall(size)) == NULL) {
1582 // Replenish list.
1583 //
1584 // Things tried that failed.
1585 // Tried allocating out of the two LinAB's first before
1586 // replenishing lists.
1587 // Tried small linAB of size 256 (size in indexed list)
1588 // and replenishing indexed lists from the small linAB.
1589 //
1590 FreeChunk* newFc = NULL;
1591 const size_t replenish_size = CMSIndexedFreeListReplenish * size;
1592 if (replenish_size < SmallForDictionary) {
1593 // Do not replenish from an underpopulated size.
1594 if (_indexedFreeList[replenish_size].surplus() > 0 &&
1595 _indexedFreeList[replenish_size].head() != NULL) {
1596 newFc = _indexedFreeList[replenish_size].get_chunk_at_head();
1597 } else if (bestFitFirst()) {
1598 newFc = bestFitSmall(replenish_size);
1599 }
1600 }
1601 if (newFc == NULL && replenish_size > size) {
1602 assert(CMSIndexedFreeListReplenish > 1, "ctl pt invariant");
1603 newFc = getChunkFromIndexedFreeListHelper(replenish_size, false);
1604 }
1605 // Note: The stats update re split-death of block obtained above
1606 // will be recorded below precisely when we know we are going to
1607 // be actually splitting it into more than one pieces below.
1608 if (newFc != NULL) {
1609 if (replenish || CMSReplenishIntermediate) {
1610 // Replenish this list and return one block to caller.
1611 size_t i;
1612 FreeChunk *curFc, *nextFc;
1613 size_t num_blk = newFc->size() / size;
1614 assert(num_blk >= 1, "Smaller than requested?");
1615 assert(newFc->size() % size == 0, "Should be integral multiple of request");
1616 if (num_blk > 1) {
1617 // we are sure we will be splitting the block just obtained
1618 // into multiple pieces; record the split-death of the original
1619 splitDeath(replenish_size);
1620 }
1621 // carve up and link blocks 0, ..., num_blk - 2
1622 // The last chunk is not added to the lists but is returned as the
1623 // free chunk.
1624 for (curFc = newFc, nextFc = (FreeChunk*)((HeapWord*)curFc + size),
1625 i = 0;
1626 i < (num_blk - 1);
1627 curFc = nextFc, nextFc = (FreeChunk*)((HeapWord*)nextFc + size),
1628 i++) {
1629 curFc->set_size(size);
1630 // Don't record this as a return in order to try and
1631 // determine the "returns" from a GC.
1632 _bt.verify_not_unallocated((HeapWord*) fc, size);
1633 _indexedFreeList[size].return_chunk_at_tail(curFc, false);
1634 _bt.mark_block((HeapWord*)curFc, size);
1635 split_birth(size);
1636 // Don't record the initial population of the indexed list
1637 // as a split birth.
1638 }
1639
1640 // check that the arithmetic was OK above
1641 assert((HeapWord*)nextFc == (HeapWord*)newFc + num_blk*size,
1642 "inconsistency in carving newFc");
1643 curFc->set_size(size);
1644 _bt.mark_block((HeapWord*)curFc, size);
1645 split_birth(size);
1646 fc = curFc;
1647 } else {
1648 // Return entire block to caller
1649 fc = newFc;
1650 }
1651 }
1652 }
1653 } else {
1654 // Get a free chunk from the free chunk dictionary to be returned to
1655 // replenish the indexed free list.
1656 fc = getChunkFromDictionaryExact(size);
1657 }
1658 // assert(fc == NULL || fc->is_free(), "Should be returning a free chunk");
1659 return fc;
1660 }
1661
1662 FreeChunk*
1663 CompactibleFreeListSpace::getChunkFromDictionary(size_t size) {
1664 assert_locked();
1665 FreeChunk* fc = _dictionary->get_chunk(size);
1666 if (fc == NULL) {
1667 return NULL;
1668 }
1669 _bt.allocated((HeapWord*)fc, fc->size());
1670 if (fc->size() >= size + MinChunkSize) {
1671 fc = splitChunkAndReturnRemainder(fc, size);
1672 }
1673 assert(fc->size() >= size, "chunk too small");
1674 assert(fc->size() < size + MinChunkSize, "chunk too big");
1675 _bt.verify_single_block((HeapWord*)fc, fc->size());
1676 return fc;
1677 }
1678
1679 FreeChunk*
1680 CompactibleFreeListSpace::getChunkFromDictionaryExact(size_t size) {
1681 assert_locked();
1682 FreeChunk* fc = _dictionary->get_chunk(size);
1683 if (fc == NULL) {
1684 return fc;
1685 }
1686 _bt.allocated((HeapWord*)fc, fc->size());
1687 if (fc->size() == size) {
1688 _bt.verify_single_block((HeapWord*)fc, size);
1689 return fc;
1690 }
1691 assert(fc->size() > size, "get_chunk() guarantee");
1692 if (fc->size() < size + MinChunkSize) {
1693 // Return the chunk to the dictionary and go get a bigger one.
1694 returnChunkToDictionary(fc);
1695 fc = _dictionary->get_chunk(size + MinChunkSize);
1696 if (fc == NULL) {
1697 return NULL;
1698 }
1699 _bt.allocated((HeapWord*)fc, fc->size());
1700 }
1701 assert(fc->size() >= size + MinChunkSize, "tautology");
1702 fc = splitChunkAndReturnRemainder(fc, size);
1703 assert(fc->size() == size, "chunk is wrong size");
1704 _bt.verify_single_block((HeapWord*)fc, size);
1705 return fc;
1706 }
1707
1708 void
1709 CompactibleFreeListSpace::returnChunkToDictionary(FreeChunk* chunk) {
1710 assert_locked();
1711
1712 size_t size = chunk->size();
1713 _bt.verify_single_block((HeapWord*)chunk, size);
1714 // adjust _unallocated_block downward, as necessary
1715 _bt.freed((HeapWord*)chunk, size);
1716 _dictionary->return_chunk(chunk);
1717 #ifndef PRODUCT
1718 if (CMSCollector::abstract_state() != CMSCollector::Sweeping) {
1719 TreeChunk<FreeChunk>::as_TreeChunk(chunk)->list()->verify_stats();
1720 }
1721 #endif // PRODUCT
1722 }
1723
1724 void
1725 CompactibleFreeListSpace::returnChunkToFreeList(FreeChunk* fc) {
1726 assert_locked();
1727 size_t size = fc->size();
1728 _bt.verify_single_block((HeapWord*) fc, size);
1729 _bt.verify_not_unallocated((HeapWord*) fc, size);
1730 if (_adaptive_freelists) {
1731 _indexedFreeList[size].return_chunk_at_tail(fc);
1732 } else {
1733 _indexedFreeList[size].return_chunk_at_head(fc);
1734 }
1735 #ifndef PRODUCT
1736 if (CMSCollector::abstract_state() != CMSCollector::Sweeping) {
1737 _indexedFreeList[size].verify_stats();
1738 }
1739 #endif // PRODUCT
1740 }
1741
1742 // Add chunk to end of last block -- if it's the largest
1743 // block -- and update BOT and census data. We would
1744 // of course have preferred to coalesce it with the
1745 // last block, but it's currently less expensive to find the
1746 // largest block than it is to find the last.
1747 void
1748 CompactibleFreeListSpace::addChunkToFreeListsAtEndRecordingStats(
1749 HeapWord* chunk, size_t size) {
1750 // check that the chunk does lie in this space!
1751 assert(chunk != NULL && is_in_reserved(chunk), "Not in this space!");
1752 // One of the parallel gc task threads may be here
1753 // whilst others are allocating.
1754 Mutex* lock = NULL;
1755 if (ParallelGCThreads != 0) {
1756 lock = &_parDictionaryAllocLock;
1757 }
1758 FreeChunk* ec;
1759 {
1760 MutexLockerEx x(lock, Mutex::_no_safepoint_check_flag);
1761 ec = dictionary()->find_largest_dict(); // get largest block
1762 if (ec != NULL && ec->end() == chunk) {
1763 // It's a coterminal block - we can coalesce.
1764 size_t old_size = ec->size();
1765 coalDeath(old_size);
1766 removeChunkFromDictionary(ec);
1767 size += old_size;
1768 } else {
1769 ec = (FreeChunk*)chunk;
1770 }
1771 }
1772 ec->set_size(size);
1773 debug_only(ec->mangleFreed(size));
1774 if (size < SmallForDictionary) {
1775 lock = _indexedFreeListParLocks[size];
1776 }
1777 MutexLockerEx x(lock, Mutex::_no_safepoint_check_flag);
1778 addChunkAndRepairOffsetTable((HeapWord*)ec, size, true);
1779 // record the birth under the lock since the recording involves
1780 // manipulation of the list on which the chunk lives and
1781 // if the chunk is allocated and is the last on the list,
1782 // the list can go away.
1783 coalBirth(size);
1784 }
1785
1786 void
1787 CompactibleFreeListSpace::addChunkToFreeLists(HeapWord* chunk,
1788 size_t size) {
1789 // check that the chunk does lie in this space!
1790 assert(chunk != NULL && is_in_reserved(chunk), "Not in this space!");
1791 assert_locked();
1792 _bt.verify_single_block(chunk, size);
1793
1794 FreeChunk* fc = (FreeChunk*) chunk;
1795 fc->set_size(size);
1796 debug_only(fc->mangleFreed(size));
1797 if (size < SmallForDictionary) {
1798 returnChunkToFreeList(fc);
1799 } else {
1800 returnChunkToDictionary(fc);
1801 }
1802 }
1803
1804 void
1805 CompactibleFreeListSpace::addChunkAndRepairOffsetTable(HeapWord* chunk,
1806 size_t size, bool coalesced) {
1807 assert_locked();
1808 assert(chunk != NULL, "null chunk");
1809 if (coalesced) {
1810 // repair BOT
1811 _bt.single_block(chunk, size);
1812 }
1813 addChunkToFreeLists(chunk, size);
1814 }
1815
1816 // We _must_ find the purported chunk on our free lists;
1817 // we assert if we don't.
1818 void
1819 CompactibleFreeListSpace::removeFreeChunkFromFreeLists(FreeChunk* fc) {
1820 size_t size = fc->size();
1821 assert_locked();
1822 debug_only(verifyFreeLists());
1823 if (size < SmallForDictionary) {
1824 removeChunkFromIndexedFreeList(fc);
1825 } else {
1826 removeChunkFromDictionary(fc);
1827 }
1828 _bt.verify_single_block((HeapWord*)fc, size);
1829 debug_only(verifyFreeLists());
1830 }
1831
1832 void
1833 CompactibleFreeListSpace::removeChunkFromDictionary(FreeChunk* fc) {
1834 size_t size = fc->size();
1835 assert_locked();
1836 assert(fc != NULL, "null chunk");
1837 _bt.verify_single_block((HeapWord*)fc, size);
1838 _dictionary->remove_chunk(fc);
1839 // adjust _unallocated_block upward, as necessary
1840 _bt.allocated((HeapWord*)fc, size);
1841 }
1842
1843 void
1844 CompactibleFreeListSpace::removeChunkFromIndexedFreeList(FreeChunk* fc) {
1845 assert_locked();
1846 size_t size = fc->size();
1847 _bt.verify_single_block((HeapWord*)fc, size);
1848 NOT_PRODUCT(
1849 if (FLSVerifyIndexTable) {
1850 verifyIndexedFreeList(size);
1851 }
1852 )
1853 _indexedFreeList[size].remove_chunk(fc);
1854 NOT_PRODUCT(
1855 if (FLSVerifyIndexTable) {
1856 verifyIndexedFreeList(size);
1857 }
1858 )
1859 }
1860
1861 FreeChunk* CompactibleFreeListSpace::bestFitSmall(size_t numWords) {
1862 /* A hint is the next larger size that has a surplus.
1863 Start search at a size large enough to guarantee that
1864 the excess is >= MIN_CHUNK. */
1865 size_t start = align_object_size(numWords + MinChunkSize);
1866 if (start < IndexSetSize) {
1867 FreeList<FreeChunk>* it = _indexedFreeList;
1868 size_t hint = _indexedFreeList[start].hint();
1869 while (hint < IndexSetSize) {
1870 assert(hint % MinObjAlignment == 0, "hint should be aligned");
1871 FreeList<FreeChunk> *fl = &_indexedFreeList[hint];
1872 if (fl->surplus() > 0 && fl->head() != NULL) {
1873 // Found a list with surplus, reset original hint
1874 // and split out a free chunk which is returned.
1875 _indexedFreeList[start].set_hint(hint);
1876 FreeChunk* res = getFromListGreater(fl, numWords);
1877 assert(res == NULL || res->is_free(),
1878 "Should be returning a free chunk");
1879 return res;
1880 }
1881 hint = fl->hint(); /* keep looking */
1882 }
1883 /* None found. */
1884 it[start].set_hint(IndexSetSize);
1885 }
1886 return NULL;
1887 }
1888
1889 /* Requires fl->size >= numWords + MinChunkSize */
1890 FreeChunk* CompactibleFreeListSpace::getFromListGreater(FreeList<FreeChunk>* fl,
1891 size_t numWords) {
1892 FreeChunk *curr = fl->head();
1893 size_t oldNumWords = curr->size();
1894 assert(numWords >= MinChunkSize, "Word size is too small");
1895 assert(curr != NULL, "List is empty");
1896 assert(oldNumWords >= numWords + MinChunkSize,
1897 "Size of chunks in the list is too small");
1898
1899 fl->remove_chunk(curr);
1900 // recorded indirectly by splitChunkAndReturnRemainder -
1901 // smallSplit(oldNumWords, numWords);
1902 FreeChunk* new_chunk = splitChunkAndReturnRemainder(curr, numWords);
1903 // Does anything have to be done for the remainder in terms of
1904 // fixing the card table?
1905 assert(new_chunk == NULL || new_chunk->is_free(),
1906 "Should be returning a free chunk");
1907 return new_chunk;
1908 }
1909
1910 FreeChunk*
1911 CompactibleFreeListSpace::splitChunkAndReturnRemainder(FreeChunk* chunk,
1912 size_t new_size) {
1913 assert_locked();
1914 size_t size = chunk->size();
1915 assert(size > new_size, "Split from a smaller block?");
1916 assert(is_aligned(chunk), "alignment problem");
1917 assert(size == adjustObjectSize(size), "alignment problem");
1918 size_t rem_size = size - new_size;
1919 assert(rem_size == adjustObjectSize(rem_size), "alignment problem");
1920 assert(rem_size >= MinChunkSize, "Free chunk smaller than minimum");
1921 FreeChunk* ffc = (FreeChunk*)((HeapWord*)chunk + new_size);
1922 assert(is_aligned(ffc), "alignment problem");
1923 ffc->set_size(rem_size);
1924 ffc->link_next(NULL);
1925 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
1926 // Above must occur before BOT is updated below.
1927 // adjust block offset table
1928 OrderAccess::storestore();
1929 assert(chunk->is_free() && ffc->is_free(), "Error");
1930 _bt.split_block((HeapWord*)chunk, chunk->size(), new_size);
1931 if (rem_size < SmallForDictionary) {
1932 bool is_par = (SharedHeap::heap()->n_par_threads() > 0);
1933 if (is_par) _indexedFreeListParLocks[rem_size]->lock();
1934 assert(!is_par ||
1935 (SharedHeap::heap()->n_par_threads() ==
1936 SharedHeap::heap()->workers()->active_workers()), "Mismatch");
1937 returnChunkToFreeList(ffc);
1938 split(size, rem_size);
1939 if (is_par) _indexedFreeListParLocks[rem_size]->unlock();
1940 } else {
1941 returnChunkToDictionary(ffc);
1942 split(size ,rem_size);
1943 }
1944 chunk->set_size(new_size);
1945 return chunk;
1946 }
1947
1948 void
1949 CompactibleFreeListSpace::sweep_completed() {
1950 // Now that space is probably plentiful, refill linear
1951 // allocation blocks as needed.
1952 refillLinearAllocBlocksIfNeeded();
1953 }
1954
1955 void
1956 CompactibleFreeListSpace::gc_prologue() {
1957 assert_locked();
1958 if (PrintFLSStatistics != 0) {
1959 gclog_or_tty->print("Before GC:\n");
1960 reportFreeListStatistics();
1961 }
1962 refillLinearAllocBlocksIfNeeded();
1963 }
1964
1965 void
1966 CompactibleFreeListSpace::gc_epilogue() {
1967 assert_locked();
1968 if (PrintGCDetails && Verbose && !_adaptive_freelists) {
1969 if (_smallLinearAllocBlock._word_size == 0)
1970 warning("CompactibleFreeListSpace(epilogue):: Linear allocation failure");
1971 }
1972 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency");
1973 _promoInfo.stopTrackingPromotions();
1974 repairLinearAllocationBlocks();
1975 // Print Space's stats
1976 if (PrintFLSStatistics != 0) {
1977 gclog_or_tty->print("After GC:\n");
1978 reportFreeListStatistics();
1979 }
1980 }
1981
1982 // Iteration support, mostly delegated from a CMS generation
1983
1984 void CompactibleFreeListSpace::save_marks() {
1985 assert(Thread::current()->is_VM_thread(),
1986 "Global variable should only be set when single-threaded");
1987 // Mark the "end" of the used space at the time of this call;
1988 // note, however, that promoted objects from this point
1989 // on are tracked in the _promoInfo below.
1990 set_saved_mark_word(unallocated_block());
1991 #ifdef ASSERT
1992 // Check the sanity of save_marks() etc.
1993 MemRegion ur = used_region();
1994 MemRegion urasm = used_region_at_save_marks();
1995 assert(ur.contains(urasm),
1996 err_msg(" Error at save_marks(): [" PTR_FORMAT "," PTR_FORMAT ")"
1997 " should contain [" PTR_FORMAT "," PTR_FORMAT ")",
1998 ur.start(), ur.end(), urasm.start(), urasm.end()));
1999 #endif
2000 // inform allocator that promotions should be tracked.
2001 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency");
2002 _promoInfo.startTrackingPromotions();
2003 }
2004
2005 bool CompactibleFreeListSpace::no_allocs_since_save_marks() {
2006 assert(_promoInfo.tracking(), "No preceding save_marks?");
2007 assert(SharedHeap::heap()->n_par_threads() == 0,
2008 "Shouldn't be called if using parallel gc.");
2009 return _promoInfo.noPromotions();
2010 }
2011
2012 #define CFLS_OOP_SINCE_SAVE_MARKS_DEFN(OopClosureType, nv_suffix) \
2013 \
2014 void CompactibleFreeListSpace:: \
2015 oop_since_save_marks_iterate##nv_suffix(OopClosureType* blk) { \
2016 assert(SharedHeap::heap()->n_par_threads() == 0, \
2017 "Shouldn't be called (yet) during parallel part of gc."); \
2018 _promoInfo.promoted_oops_iterate##nv_suffix(blk); \
2019 /* \
2020 * This also restores any displaced headers and removes the elements from \
2021 * the iteration set as they are processed, so that we have a clean slate \
2022 * at the end of the iteration. Note, thus, that if new objects are \
2023 * promoted as a result of the iteration they are iterated over as well. \
2024 */ \
2025 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency"); \
2026 }
2027
2028 ALL_SINCE_SAVE_MARKS_CLOSURES(CFLS_OOP_SINCE_SAVE_MARKS_DEFN)
2029
2030
2031 void CompactibleFreeListSpace::object_iterate_since_last_GC(ObjectClosure* cl) {
2032 // ugghh... how would one do this efficiently for a non-contiguous space?
2033 guarantee(false, "NYI");
2034 }
2035
2036 bool CompactibleFreeListSpace::linearAllocationWouldFail() const {
2037 return _smallLinearAllocBlock._word_size == 0;
2038 }
2039
2040 void CompactibleFreeListSpace::repairLinearAllocationBlocks() {
2041 // Fix up linear allocation blocks to look like free blocks
2042 repairLinearAllocBlock(&_smallLinearAllocBlock);
2043 }
2044
2045 void CompactibleFreeListSpace::repairLinearAllocBlock(LinearAllocBlock* blk) {
2046 assert_locked();
2047 if (blk->_ptr != NULL) {
2048 assert(blk->_word_size != 0 && blk->_word_size >= MinChunkSize,
2049 "Minimum block size requirement");
2050 FreeChunk* fc = (FreeChunk*)(blk->_ptr);
2051 fc->set_size(blk->_word_size);
2052 fc->link_prev(NULL); // mark as free
2053 fc->dontCoalesce();
2054 assert(fc->is_free(), "just marked it free");
2055 assert(fc->cantCoalesce(), "just marked it uncoalescable");
2056 }
2057 }
2058
2059 void CompactibleFreeListSpace::refillLinearAllocBlocksIfNeeded() {
2060 assert_locked();
2061 if (_smallLinearAllocBlock._ptr == NULL) {
2062 assert(_smallLinearAllocBlock._word_size == 0,
2063 "Size of linAB should be zero if the ptr is NULL");
2064 // Reset the linAB refill and allocation size limit.
2065 _smallLinearAllocBlock.set(0, 0, 1024*SmallForLinearAlloc, SmallForLinearAlloc);
2066 }
2067 refillLinearAllocBlockIfNeeded(&_smallLinearAllocBlock);
2068 }
2069
2070 void
2071 CompactibleFreeListSpace::refillLinearAllocBlockIfNeeded(LinearAllocBlock* blk) {
2072 assert_locked();
2073 assert((blk->_ptr == NULL && blk->_word_size == 0) ||
2074 (blk->_ptr != NULL && blk->_word_size >= MinChunkSize),
2075 "blk invariant");
2076 if (blk->_ptr == NULL) {
2077 refillLinearAllocBlock(blk);
2078 }
2079 if (PrintMiscellaneous && Verbose) {
2080 if (blk->_word_size == 0) {
2081 warning("CompactibleFreeListSpace(prologue):: Linear allocation failure");
2082 }
2083 }
2084 }
2085
2086 void
2087 CompactibleFreeListSpace::refillLinearAllocBlock(LinearAllocBlock* blk) {
2088 assert_locked();
2089 assert(blk->_word_size == 0 && blk->_ptr == NULL,
2090 "linear allocation block should be empty");
2091 FreeChunk* fc;
2092 if (blk->_refillSize < SmallForDictionary &&
2093 (fc = getChunkFromIndexedFreeList(blk->_refillSize)) != NULL) {
2094 // A linAB's strategy might be to use small sizes to reduce
2095 // fragmentation but still get the benefits of allocation from a
2096 // linAB.
2097 } else {
2098 fc = getChunkFromDictionary(blk->_refillSize);
2099 }
2100 if (fc != NULL) {
2101 blk->_ptr = (HeapWord*)fc;
2102 blk->_word_size = fc->size();
2103 fc->dontCoalesce(); // to prevent sweeper from sweeping us up
2104 }
2105 }
2106
2107 // Support for concurrent collection policy decisions.
2108 bool CompactibleFreeListSpace::should_concurrent_collect() const {
2109 // In the future we might want to add in frgamentation stats --
2110 // including erosion of the "mountain" into this decision as well.
2111 return !adaptive_freelists() && linearAllocationWouldFail();
2112 }
2113
2114 // Support for compaction
2115
2116 void CompactibleFreeListSpace::prepare_for_compaction(CompactPoint* cp) {
2117 SCAN_AND_FORWARD(cp,end,block_is_obj,block_size);
2118 // prepare_for_compaction() uses the space between live objects
2119 // so that later phase can skip dead space quickly. So verification
2120 // of the free lists doesn't work after.
2121 }
2122
2123 #define obj_size(q) adjustObjectSize(oop(q)->size())
2124 #define adjust_obj_size(s) adjustObjectSize(s)
2125
2126 void CompactibleFreeListSpace::adjust_pointers() {
2127 // In other versions of adjust_pointers(), a bail out
2128 // based on the amount of live data in the generation
2129 // (i.e., if 0, bail out) may be used.
2130 // Cannot test used() == 0 here because the free lists have already
2131 // been mangled by the compaction.
2132
2133 SCAN_AND_ADJUST_POINTERS(adjust_obj_size);
2134 // See note about verification in prepare_for_compaction().
2135 }
2136
2137 void CompactibleFreeListSpace::compact() {
2138 SCAN_AND_COMPACT(obj_size);
2139 }
2140
2141 // fragmentation_metric = 1 - [sum of (fbs**2) / (sum of fbs)**2]
2142 // where fbs is free block sizes
2143 double CompactibleFreeListSpace::flsFrag() const {
2144 size_t itabFree = totalSizeInIndexedFreeLists();
2145 double frag = 0.0;
2146 size_t i;
2147
2148 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2149 double sz = i;
2150 frag += _indexedFreeList[i].count() * (sz * sz);
2151 }
2152
2153 double totFree = itabFree +
2154 _dictionary->total_chunk_size(DEBUG_ONLY(freelistLock()));
2155 if (totFree > 0) {
2156 frag = ((frag + _dictionary->sum_of_squared_block_sizes()) /
2157 (totFree * totFree));
2158 frag = (double)1.0 - frag;
2159 } else {
2160 assert(frag == 0.0, "Follows from totFree == 0");
2161 }
2162 return frag;
2163 }
2164
2165 void CompactibleFreeListSpace::beginSweepFLCensus(
2166 float inter_sweep_current,
2167 float inter_sweep_estimate,
2168 float intra_sweep_estimate) {
2169 assert_locked();
2170 size_t i;
2171 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2172 FreeList<FreeChunk>* fl = &_indexedFreeList[i];
2173 if (PrintFLSStatistics > 1) {
2174 gclog_or_tty->print("size[%d] : ", i);
2175 }
2176 fl->compute_desired(inter_sweep_current, inter_sweep_estimate, intra_sweep_estimate);
2177 fl->set_coal_desired((ssize_t)((double)fl->desired() * CMSSmallCoalSurplusPercent));
2178 fl->set_before_sweep(fl->count());
2179 fl->set_bfr_surp(fl->surplus());
2180 }
2181 _dictionary->begin_sweep_dict_census(CMSLargeCoalSurplusPercent,
2182 inter_sweep_current,
2183 inter_sweep_estimate,
2184 intra_sweep_estimate);
2185 }
2186
2187 void CompactibleFreeListSpace::setFLSurplus() {
2188 assert_locked();
2189 size_t i;
2190 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2191 FreeList<FreeChunk> *fl = &_indexedFreeList[i];
2192 fl->set_surplus(fl->count() -
2193 (ssize_t)((double)fl->desired() * CMSSmallSplitSurplusPercent));
2194 }
2195 }
2196
2197 void CompactibleFreeListSpace::setFLHints() {
2198 assert_locked();
2199 size_t i;
2200 size_t h = IndexSetSize;
2201 for (i = IndexSetSize - 1; i != 0; i -= IndexSetStride) {
2202 FreeList<FreeChunk> *fl = &_indexedFreeList[i];
2203 fl->set_hint(h);
2204 if (fl->surplus() > 0) {
2205 h = i;
2206 }
2207 }
2208 }
2209
2210 void CompactibleFreeListSpace::clearFLCensus() {
2211 assert_locked();
2212 size_t i;
2213 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2214 FreeList<FreeChunk> *fl = &_indexedFreeList[i];
2215 fl->set_prev_sweep(fl->count());
2216 fl->set_coal_births(0);
2217 fl->set_coal_deaths(0);
2218 fl->set_split_births(0);
2219 fl->set_split_deaths(0);
2220 }
2221 }
2222
2223 void CompactibleFreeListSpace::endSweepFLCensus(size_t sweep_count) {
2224 if (PrintFLSStatistics > 0) {
2225 HeapWord* largestAddr = (HeapWord*) dictionary()->find_largest_dict();
2226 gclog_or_tty->print_cr("CMS: Large block " PTR_FORMAT,
2227 largestAddr);
2228 }
2229 setFLSurplus();
2230 setFLHints();
2231 if (PrintGC && PrintFLSCensus > 0) {
2232 printFLCensus(sweep_count);
2233 }
2234 clearFLCensus();
2235 assert_locked();
2236 _dictionary->end_sweep_dict_census(CMSLargeSplitSurplusPercent);
2237 }
2238
2239 bool CompactibleFreeListSpace::coalOverPopulated(size_t size) {
2240 if (size < SmallForDictionary) {
2241 FreeList<FreeChunk> *fl = &_indexedFreeList[size];
2242 return (fl->coal_desired() < 0) ||
2243 ((int)fl->count() > fl->coal_desired());
2244 } else {
2245 return dictionary()->coal_dict_over_populated(size);
2246 }
2247 }
2248
2249 void CompactibleFreeListSpace::smallCoalBirth(size_t size) {
2250 assert(size < SmallForDictionary, "Size too large for indexed list");
2251 FreeList<FreeChunk> *fl = &_indexedFreeList[size];
2252 fl->increment_coal_births();
2253 fl->increment_surplus();
2254 }
2255
2256 void CompactibleFreeListSpace::smallCoalDeath(size_t size) {
2257 assert(size < SmallForDictionary, "Size too large for indexed list");
2258 FreeList<FreeChunk> *fl = &_indexedFreeList[size];
2259 fl->increment_coal_deaths();
2260 fl->decrement_surplus();
2261 }
2262
2263 void CompactibleFreeListSpace::coalBirth(size_t size) {
2264 if (size < SmallForDictionary) {
2265 smallCoalBirth(size);
2266 } else {
2267 dictionary()->dict_census_udpate(size,
2268 false /* split */,
2269 true /* birth */);
2270 }
2271 }
2272
2273 void CompactibleFreeListSpace::coalDeath(size_t size) {
2274 if(size < SmallForDictionary) {
2275 smallCoalDeath(size);
2276 } else {
2277 dictionary()->dict_census_udpate(size,
2278 false /* split */,
2279 false /* birth */);
2280 }
2281 }
2282
2283 void CompactibleFreeListSpace::smallSplitBirth(size_t size) {
2284 assert(size < SmallForDictionary, "Size too large for indexed list");
2285 FreeList<FreeChunk> *fl = &_indexedFreeList[size];
2286 fl->increment_split_births();
2287 fl->increment_surplus();
2288 }
2289
2290 void CompactibleFreeListSpace::smallSplitDeath(size_t size) {
2291 assert(size < SmallForDictionary, "Size too large for indexed list");
2292 FreeList<FreeChunk> *fl = &_indexedFreeList[size];
2293 fl->increment_split_deaths();
2294 fl->decrement_surplus();
2295 }
2296
2297 void CompactibleFreeListSpace::split_birth(size_t size) {
2298 if (size < SmallForDictionary) {
2299 smallSplitBirth(size);
2300 } else {
2301 dictionary()->dict_census_udpate(size,
2302 true /* split */,
2303 true /* birth */);
2304 }
2305 }
2306
2307 void CompactibleFreeListSpace::splitDeath(size_t size) {
2308 if (size < SmallForDictionary) {
2309 smallSplitDeath(size);
2310 } else {
2311 dictionary()->dict_census_udpate(size,
2312 true /* split */,
2313 false /* birth */);
2314 }
2315 }
2316
2317 void CompactibleFreeListSpace::split(size_t from, size_t to1) {
2318 size_t to2 = from - to1;
2319 splitDeath(from);
2320 split_birth(to1);
2321 split_birth(to2);
2322 }
2323
2324 void CompactibleFreeListSpace::print() const {
2325 print_on(tty);
2326 }
2327
2328 void CompactibleFreeListSpace::prepare_for_verify() {
2329 assert_locked();
2330 repairLinearAllocationBlocks();
2331 // Verify that the SpoolBlocks look like free blocks of
2332 // appropriate sizes... To be done ...
2333 }
2334
2335 class VerifyAllBlksClosure: public BlkClosure {
2336 private:
2337 const CompactibleFreeListSpace* _sp;
2338 const MemRegion _span;
2339 HeapWord* _last_addr;
2340 size_t _last_size;
2341 bool _last_was_obj;
2342 bool _last_was_live;
2343
2344 public:
2345 VerifyAllBlksClosure(const CompactibleFreeListSpace* sp,
2346 MemRegion span) : _sp(sp), _span(span),
2347 _last_addr(NULL), _last_size(0),
2348 _last_was_obj(false), _last_was_live(false) { }
2349
2350 virtual size_t do_blk(HeapWord* addr) {
2351 size_t res;
2352 bool was_obj = false;
2353 bool was_live = false;
2354 if (_sp->block_is_obj(addr)) {
2355 was_obj = true;
2356 oop p = oop(addr);
2357 guarantee(p->is_oop(), "Should be an oop");
2358 res = _sp->adjustObjectSize(p->size());
2359 if (_sp->obj_is_alive(addr)) {
2360 was_live = true;
2361 p->verify();
2362 }
2363 } else {
2364 FreeChunk* fc = (FreeChunk*)addr;
2365 res = fc->size();
2366 if (FLSVerifyLists && !fc->cantCoalesce()) {
2367 guarantee(_sp->verify_chunk_in_free_list(fc),
2368 "Chunk should be on a free list");
2369 }
2370 }
2371 if (res == 0) {
2372 gclog_or_tty->print_cr("Livelock: no rank reduction!");
2373 gclog_or_tty->print_cr(
2374 " Current: addr = " PTR_FORMAT ", size = " SIZE_FORMAT ", obj = %s, live = %s \n"
2375 " Previous: addr = " PTR_FORMAT ", size = " SIZE_FORMAT ", obj = %s, live = %s \n",
2376 addr, res, was_obj ?"true":"false", was_live ?"true":"false",
2377 _last_addr, _last_size, _last_was_obj?"true":"false", _last_was_live?"true":"false");
2378 _sp->print_on(gclog_or_tty);
2379 guarantee(false, "Seppuku!");
2380 }
2381 _last_addr = addr;
2382 _last_size = res;
2383 _last_was_obj = was_obj;
2384 _last_was_live = was_live;
2385 return res;
2386 }
2387 };
2388
2389 class VerifyAllOopsClosure: public OopClosure {
2390 private:
2391 const CMSCollector* _collector;
2392 const CompactibleFreeListSpace* _sp;
2393 const MemRegion _span;
2394 const bool _past_remark;
2395 const CMSBitMap* _bit_map;
2396
2397 protected:
2398 void do_oop(void* p, oop obj) {
2399 if (_span.contains(obj)) { // the interior oop points into CMS heap
2400 if (!_span.contains(p)) { // reference from outside CMS heap
2401 // Should be a valid object; the first disjunct below allows
2402 // us to sidestep an assertion in block_is_obj() that insists
2403 // that p be in _sp. Note that several generations (and spaces)
2404 // are spanned by _span (CMS heap) above.
2405 guarantee(!_sp->is_in_reserved(obj) ||
2406 _sp->block_is_obj((HeapWord*)obj),
2407 "Should be an object");
2408 guarantee(obj->is_oop(), "Should be an oop");
2409 obj->verify();
2410 if (_past_remark) {
2411 // Remark has been completed, the object should be marked
2412 _bit_map->isMarked((HeapWord*)obj);
2413 }
2414 } else { // reference within CMS heap
2415 if (_past_remark) {
2416 // Remark has been completed -- so the referent should have
2417 // been marked, if referring object is.
2418 if (_bit_map->isMarked(_collector->block_start(p))) {
2419 guarantee(_bit_map->isMarked((HeapWord*)obj), "Marking error?");
2420 }
2421 }
2422 }
2423 } else if (_sp->is_in_reserved(p)) {
2424 // the reference is from FLS, and points out of FLS
2425 guarantee(obj->is_oop(), "Should be an oop");
2426 obj->verify();
2427 }
2428 }
2429
2430 template <class T> void do_oop_work(T* p) {
2431 T heap_oop = oopDesc::load_heap_oop(p);
2432 if (!oopDesc::is_null(heap_oop)) {
2433 oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
2434 do_oop(p, obj);
2435 }
2436 }
2437
2438 public:
2439 VerifyAllOopsClosure(const CMSCollector* collector,
2440 const CompactibleFreeListSpace* sp, MemRegion span,
2441 bool past_remark, CMSBitMap* bit_map) :
2442 OopClosure(), _collector(collector), _sp(sp), _span(span),
2443 _past_remark(past_remark), _bit_map(bit_map) { }
2444
2445 virtual void do_oop(oop* p) { VerifyAllOopsClosure::do_oop_work(p); }
2446 virtual void do_oop(narrowOop* p) { VerifyAllOopsClosure::do_oop_work(p); }
2447 };
2448
2449 void CompactibleFreeListSpace::verify() const {
2450 assert_lock_strong(&_freelistLock);
2451 verify_objects_initialized();
2452 MemRegion span = _collector->_span;
2453 bool past_remark = (_collector->abstract_state() ==
2454 CMSCollector::Sweeping);
2455
2456 ResourceMark rm;
2457 HandleMark hm;
2458
2459 // Check integrity of CFL data structures
2460 _promoInfo.verify();
2461 _dictionary->verify();
2462 if (FLSVerifyIndexTable) {
2463 verifyIndexedFreeLists();
2464 }
2465 // Check integrity of all objects and free blocks in space
2466 {
2467 VerifyAllBlksClosure cl(this, span);
2468 ((CompactibleFreeListSpace*)this)->blk_iterate(&cl); // cast off const
2469 }
2470 // Check that all references in the heap to FLS
2471 // are to valid objects in FLS or that references in
2472 // FLS are to valid objects elsewhere in the heap
2473 if (FLSVerifyAllHeapReferences)
2474 {
2475 VerifyAllOopsClosure cl(_collector, this, span, past_remark,
2476 _collector->markBitMap());
2477 CollectedHeap* ch = Universe::heap();
2478 ch->oop_iterate(&cl); // all oops in generations
2479 ch->permanent_oop_iterate(&cl); // all oops in perm gen
2480 }
2481
2482 if (VerifyObjectStartArray) {
2483 // Verify the block offset table
2484 _bt.verify();
2485 }
2486 }
2487
2488 #ifndef PRODUCT
2489 void CompactibleFreeListSpace::verifyFreeLists() const {
2490 if (FLSVerifyLists) {
2491 _dictionary->verify();
2492 verifyIndexedFreeLists();
2493 } else {
2494 if (FLSVerifyDictionary) {
2495 _dictionary->verify();
2496 }
2497 if (FLSVerifyIndexTable) {
2498 verifyIndexedFreeLists();
2499 }
2500 }
2501 }
2502 #endif
2503
2504 void CompactibleFreeListSpace::verifyIndexedFreeLists() const {
2505 size_t i = 0;
2506 for (; i < IndexSetStart; i++) {
2507 guarantee(_indexedFreeList[i].head() == NULL, "should be NULL");
2508 }
2509 for (; i < IndexSetSize; i++) {
2510 verifyIndexedFreeList(i);
2511 }
2512 }
2513
2514 void CompactibleFreeListSpace::verifyIndexedFreeList(size_t size) const {
2515 FreeChunk* fc = _indexedFreeList[size].head();
2516 FreeChunk* tail = _indexedFreeList[size].tail();
2517 size_t num = _indexedFreeList[size].count();
2518 size_t n = 0;
2519 guarantee(((size >= IndexSetStart) && (size % IndexSetStride == 0)) || fc == NULL,
2520 "Slot should have been empty");
2521 for (; fc != NULL; fc = fc->next(), n++) {
2522 guarantee(fc->size() == size, "Size inconsistency");
2523 guarantee(fc->is_free(), "!free?");
2524 guarantee(fc->next() == NULL || fc->next()->prev() == fc, "Broken list");
2525 guarantee((fc->next() == NULL) == (fc == tail), "Incorrect tail");
2526 }
2527 guarantee(n == num, "Incorrect count");
2528 }
2529
2530 #ifndef PRODUCT
2531 void CompactibleFreeListSpace::check_free_list_consistency() const {
2532 assert(_dictionary->min_size() <= IndexSetSize,
2533 "Some sizes can't be allocated without recourse to"
2534 " linear allocation buffers");
2535 assert(BinaryTreeDictionary<FreeChunk>::min_tree_chunk_size*HeapWordSize == sizeof(TreeChunk<FreeChunk>),
2536 "else MIN_TREE_CHUNK_SIZE is wrong");
2537 assert((IndexSetStride == 2 && IndexSetStart == 4) || // 32-bit
2538 (IndexSetStride == 1 && IndexSetStart == 3), "just checking"); // 64-bit
2539 assert((IndexSetStride != 2) || (IndexSetStart % 2 == 0),
2540 "Some for-loops may be incorrectly initialized");
2541 assert((IndexSetStride != 2) || (IndexSetSize % 2 == 1),
2542 "For-loops that iterate over IndexSet with stride 2 may be wrong");
2543 }
2544 #endif
2545
2546 void CompactibleFreeListSpace::printFLCensus(size_t sweep_count) const {
2547 assert_lock_strong(&_freelistLock);
2548 FreeList<FreeChunk> total;
2549 gclog_or_tty->print("end sweep# " SIZE_FORMAT "\n", sweep_count);
2550 FreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
2551 size_t total_free = 0;
2552 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2553 const FreeList<FreeChunk> *fl = &_indexedFreeList[i];
2554 total_free += fl->count() * fl->size();
2555 if (i % (40*IndexSetStride) == 0) {
2556 FreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
2557 }
2558 fl->print_on(gclog_or_tty);
2559 total.set_bfr_surp( total.bfr_surp() + fl->bfr_surp() );
2560 total.set_surplus( total.surplus() + fl->surplus() );
2561 total.set_desired( total.desired() + fl->desired() );
2562 total.set_prev_sweep( total.prev_sweep() + fl->prev_sweep() );
2563 total.set_before_sweep(total.before_sweep() + fl->before_sweep());
2564 total.set_count( total.count() + fl->count() );
2565 total.set_coal_births( total.coal_births() + fl->coal_births() );
2566 total.set_coal_deaths( total.coal_deaths() + fl->coal_deaths() );
2567 total.set_split_births(total.split_births() + fl->split_births());
2568 total.set_split_deaths(total.split_deaths() + fl->split_deaths());
2569 }
2570 total.print_on(gclog_or_tty, "TOTAL");
2571 gclog_or_tty->print_cr("Total free in indexed lists "
2572 SIZE_FORMAT " words", total_free);
2573 gclog_or_tty->print("growth: %8.5f deficit: %8.5f\n",
2574 (double)(total.split_births()+total.coal_births()-total.split_deaths()-total.coal_deaths())/
2575 (total.prev_sweep() != 0 ? (double)total.prev_sweep() : 1.0),
2576 (double)(total.desired() - total.count())/(total.desired() != 0 ? (double)total.desired() : 1.0));
2577 _dictionary->print_dict_census();
2578 }
2579
2580 ///////////////////////////////////////////////////////////////////////////
2581 // CFLS_LAB
2582 ///////////////////////////////////////////////////////////////////////////
2583
2584 #define VECTOR_257(x) \
2585 /* 1 2 3 4 5 6 7 8 9 1x 11 12 13 14 15 16 17 18 19 2x 21 22 23 24 25 26 27 28 29 3x 31 32 */ \
2586 { x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2587 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2588 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2589 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2590 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2591 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2592 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2593 x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, x, \
2594 x }
2595
2596 // Initialize with default setting of CMSParPromoteBlocksToClaim, _not_
2597 // OldPLABSize, whose static default is different; if overridden at the
2598 // command-line, this will get reinitialized via a call to
2599 // modify_initialization() below.
2600 AdaptiveWeightedAverage CFLS_LAB::_blocks_to_claim[] =
2601 VECTOR_257(AdaptiveWeightedAverage(OldPLABWeight, (float)CMSParPromoteBlocksToClaim));
2602 size_t CFLS_LAB::_global_num_blocks[] = VECTOR_257(0);
2603 uint CFLS_LAB::_global_num_workers[] = VECTOR_257(0);
2604
2605 CFLS_LAB::CFLS_LAB(CompactibleFreeListSpace* cfls) :
2606 _cfls(cfls)
2607 {
2608 assert(CompactibleFreeListSpace::IndexSetSize == 257, "Modify VECTOR_257() macro above");
2609 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2610 i < CompactibleFreeListSpace::IndexSetSize;
2611 i += CompactibleFreeListSpace::IndexSetStride) {
2612 _indexedFreeList[i].set_size(i);
2613 _num_blocks[i] = 0;
2614 }
2615 }
2616
2617 static bool _CFLS_LAB_modified = false;
2618
2619 void CFLS_LAB::modify_initialization(size_t n, unsigned wt) {
2620 assert(!_CFLS_LAB_modified, "Call only once");
2621 _CFLS_LAB_modified = true;
2622 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2623 i < CompactibleFreeListSpace::IndexSetSize;
2624 i += CompactibleFreeListSpace::IndexSetStride) {
2625 _blocks_to_claim[i].modify(n, wt, true /* force */);
2626 }
2627 }
2628
2629 HeapWord* CFLS_LAB::alloc(size_t word_sz) {
2630 FreeChunk* res;
2631 assert(word_sz == _cfls->adjustObjectSize(word_sz), "Error");
2632 if (word_sz >= CompactibleFreeListSpace::IndexSetSize) {
2633 // This locking manages sync with other large object allocations.
2634 MutexLockerEx x(_cfls->parDictionaryAllocLock(),
2635 Mutex::_no_safepoint_check_flag);
2636 res = _cfls->getChunkFromDictionaryExact(word_sz);
2637 if (res == NULL) return NULL;
2638 } else {
2639 FreeList<FreeChunk>* fl = &_indexedFreeList[word_sz];
2640 if (fl->count() == 0) {
2641 // Attempt to refill this local free list.
2642 get_from_global_pool(word_sz, fl);
2643 // If it didn't work, give up.
2644 if (fl->count() == 0) return NULL;
2645 }
2646 res = fl->get_chunk_at_head();
2647 assert(res != NULL, "Why was count non-zero?");
2648 }
2649 res->markNotFree();
2650 assert(!res->is_free(), "shouldn't be marked free");
2651 assert(oop(res)->klass_or_null() == NULL, "should look uninitialized");
2652 // mangle a just allocated object with a distinct pattern.
2653 debug_only(res->mangleAllocated(word_sz));
2654 return (HeapWord*)res;
2655 }
2656
2657 // Get a chunk of blocks of the right size and update related
2658 // book-keeping stats
2659 void CFLS_LAB::get_from_global_pool(size_t word_sz, FreeList<FreeChunk>* fl) {
2660 // Get the #blocks we want to claim
2661 size_t n_blks = (size_t)_blocks_to_claim[word_sz].average();
2662 assert(n_blks > 0, "Error");
2663 assert(ResizePLAB || n_blks == OldPLABSize, "Error");
2664 // In some cases, when the application has a phase change,
2665 // there may be a sudden and sharp shift in the object survival
2666 // profile, and updating the counts at the end of a scavenge
2667 // may not be quick enough, giving rise to large scavenge pauses
2668 // during these phase changes. It is beneficial to detect such
2669 // changes on-the-fly during a scavenge and avoid such a phase-change
2670 // pothole. The following code is a heuristic attempt to do that.
2671 // It is protected by a product flag until we have gained
2672 // enough experience with this heuristic and fine-tuned its behaviour.
2673 // WARNING: This might increase fragmentation if we overreact to
2674 // small spikes, so some kind of historical smoothing based on
2675 // previous experience with the greater reactivity might be useful.
2676 // Lacking sufficient experience, CMSOldPLABResizeQuicker is disabled by
2677 // default.
2678 if (ResizeOldPLAB && CMSOldPLABResizeQuicker) {
2679 size_t multiple = _num_blocks[word_sz]/(CMSOldPLABToleranceFactor*CMSOldPLABNumRefills*n_blks);
2680 n_blks += CMSOldPLABReactivityFactor*multiple*n_blks;
2681 n_blks = MIN2(n_blks, CMSOldPLABMax);
2682 }
2683 assert(n_blks > 0, "Error");
2684 _cfls->par_get_chunk_of_blocks(word_sz, n_blks, fl);
2685 // Update stats table entry for this block size
2686 _num_blocks[word_sz] += fl->count();
2687 }
2688
2689 void CFLS_LAB::compute_desired_plab_size() {
2690 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2691 i < CompactibleFreeListSpace::IndexSetSize;
2692 i += CompactibleFreeListSpace::IndexSetStride) {
2693 assert((_global_num_workers[i] == 0) == (_global_num_blocks[i] == 0),
2694 "Counter inconsistency");
2695 if (_global_num_workers[i] > 0) {
2696 // Need to smooth wrt historical average
2697 if (ResizeOldPLAB) {
2698 _blocks_to_claim[i].sample(
2699 MAX2((size_t)CMSOldPLABMin,
2700 MIN2((size_t)CMSOldPLABMax,
2701 _global_num_blocks[i]/(_global_num_workers[i]*CMSOldPLABNumRefills))));
2702 }
2703 // Reset counters for next round
2704 _global_num_workers[i] = 0;
2705 _global_num_blocks[i] = 0;
2706 if (PrintOldPLAB) {
2707 gclog_or_tty->print_cr("[%d]: %d", i, (size_t)_blocks_to_claim[i].average());
2708 }
2709 }
2710 }
2711 }
2712
2713 // If this is changed in the future to allow parallel
2714 // access, one would need to take the FL locks and,
2715 // depending on how it is used, stagger access from
2716 // parallel threads to reduce contention.
2717 void CFLS_LAB::retire(int tid) {
2718 // We run this single threaded with the world stopped;
2719 // so no need for locks and such.
2720 NOT_PRODUCT(Thread* t = Thread::current();)
2721 assert(Thread::current()->is_VM_thread(), "Error");
2722 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2723 i < CompactibleFreeListSpace::IndexSetSize;
2724 i += CompactibleFreeListSpace::IndexSetStride) {
2725 assert(_num_blocks[i] >= (size_t)_indexedFreeList[i].count(),
2726 "Can't retire more than what we obtained");
2727 if (_num_blocks[i] > 0) {
2728 size_t num_retire = _indexedFreeList[i].count();
2729 assert(_num_blocks[i] > num_retire, "Should have used at least one");
2730 {
2731 // MutexLockerEx x(_cfls->_indexedFreeListParLocks[i],
2732 // Mutex::_no_safepoint_check_flag);
2733
2734 // Update globals stats for num_blocks used
2735 _global_num_blocks[i] += (_num_blocks[i] - num_retire);
2736 _global_num_workers[i]++;
2737 assert(_global_num_workers[i] <= ParallelGCThreads, "Too big");
2738 if (num_retire > 0) {
2739 _cfls->_indexedFreeList[i].prepend(&_indexedFreeList[i]);
2740 // Reset this list.
2741 _indexedFreeList[i] = FreeList<FreeChunk>();
2742 _indexedFreeList[i].set_size(i);
2743 }
2744 }
2745 if (PrintOldPLAB) {
2746 gclog_or_tty->print_cr("%d[%d]: %d/%d/%d",
2747 tid, i, num_retire, _num_blocks[i], (size_t)_blocks_to_claim[i].average());
2748 }
2749 // Reset stats for next round
2750 _num_blocks[i] = 0;
2751 }
2752 }
2753 }
2754
2755 void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n, FreeList<FreeChunk>* fl) {
2756 assert(fl->count() == 0, "Precondition.");
2757 assert(word_sz < CompactibleFreeListSpace::IndexSetSize,
2758 "Precondition");
2759
2760 // We'll try all multiples of word_sz in the indexed set, starting with
2761 // word_sz itself and, if CMSSplitIndexedFreeListBlocks, try larger multiples,
2762 // then try getting a big chunk and splitting it.
2763 {
2764 bool found;
2765 int k;
2766 size_t cur_sz;
2767 for (k = 1, cur_sz = k * word_sz, found = false;
2768 (cur_sz < CompactibleFreeListSpace::IndexSetSize) &&
2769 (CMSSplitIndexedFreeListBlocks || k <= 1);
2770 k++, cur_sz = k * word_sz) {
2771 FreeList<FreeChunk> fl_for_cur_sz; // Empty.
2772 fl_for_cur_sz.set_size(cur_sz);
2773 {
2774 MutexLockerEx x(_indexedFreeListParLocks[cur_sz],
2775 Mutex::_no_safepoint_check_flag);
2776 FreeList<FreeChunk>* gfl = &_indexedFreeList[cur_sz];
2777 if (gfl->count() != 0) {
2778 // nn is the number of chunks of size cur_sz that
2779 // we'd need to split k-ways each, in order to create
2780 // "n" chunks of size word_sz each.
2781 const size_t nn = MAX2(n/k, (size_t)1);
2782 gfl->getFirstNChunksFromList(nn, &fl_for_cur_sz);
2783 found = true;
2784 if (k > 1) {
2785 // Update split death stats for the cur_sz-size blocks list:
2786 // we increment the split death count by the number of blocks
2787 // we just took from the cur_sz-size blocks list and which
2788 // we will be splitting below.
2789 ssize_t deaths = gfl->split_deaths() +
2790 fl_for_cur_sz.count();
2791 gfl->set_split_deaths(deaths);
2792 }
2793 }
2794 }
2795 // Now transfer fl_for_cur_sz to fl. Common case, we hope, is k = 1.
2796 if (found) {
2797 if (k == 1) {
2798 fl->prepend(&fl_for_cur_sz);
2799 } else {
2800 // Divide each block on fl_for_cur_sz up k ways.
2801 FreeChunk* fc;
2802 while ((fc = fl_for_cur_sz.get_chunk_at_head()) != NULL) {
2803 // Must do this in reverse order, so that anybody attempting to
2804 // access the main chunk sees it as a single free block until we
2805 // change it.
2806 size_t fc_size = fc->size();
2807 assert(fc->is_free(), "Error");
2808 for (int i = k-1; i >= 0; i--) {
2809 FreeChunk* ffc = (FreeChunk*)((HeapWord*)fc + i * word_sz);
2810 assert((i != 0) ||
2811 ((fc == ffc) && ffc->is_free() &&
2812 (ffc->size() == k*word_sz) && (fc_size == word_sz)),
2813 "Counting error");
2814 ffc->set_size(word_sz);
2815 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2816 ffc->link_next(NULL);
2817 // Above must occur before BOT is updated below.
2818 OrderAccess::storestore();
2819 // splitting from the right, fc_size == i * word_sz
2820 _bt.mark_block((HeapWord*)ffc, word_sz, true /* reducing */);
2821 fc_size -= word_sz;
2822 assert(fc_size == i*word_sz, "Error");
2823 _bt.verify_not_unallocated((HeapWord*)ffc, word_sz);
2824 _bt.verify_single_block((HeapWord*)fc, fc_size);
2825 _bt.verify_single_block((HeapWord*)ffc, word_sz);
2826 // Push this on "fl".
2827 fl->return_chunk_at_head(ffc);
2828 }
2829 // TRAP
2830 assert(fl->tail()->next() == NULL, "List invariant.");
2831 }
2832 }
2833 // Update birth stats for this block size.
2834 size_t num = fl->count();
2835 MutexLockerEx x(_indexedFreeListParLocks[word_sz],
2836 Mutex::_no_safepoint_check_flag);
2837 ssize_t births = _indexedFreeList[word_sz].split_births() + num;
2838 _indexedFreeList[word_sz].set_split_births(births);
2839 return;
2840 }
2841 }
2842 }
2843 // Otherwise, we'll split a block from the dictionary.
2844 FreeChunk* fc = NULL;
2845 FreeChunk* rem_fc = NULL;
2846 size_t rem;
2847 {
2848 MutexLockerEx x(parDictionaryAllocLock(),
2849 Mutex::_no_safepoint_check_flag);
2850 while (n > 0) {
2851 fc = dictionary()->get_chunk(MAX2(n * word_sz,
2852 _dictionary->min_size()),
2853 FreeBlockDictionary<FreeChunk>::atLeast);
2854 if (fc != NULL) {
2855 _bt.allocated((HeapWord*)fc, fc->size(), true /* reducing */); // update _unallocated_blk
2856 dictionary()->dict_census_udpate(fc->size(),
2857 true /*split*/,
2858 false /*birth*/);
2859 break;
2860 } else {
2861 n--;
2862 }
2863 }
2864 if (fc == NULL) return;
2865 // Otherwise, split up that block.
2866 assert((ssize_t)n >= 1, "Control point invariant");
2867 assert(fc->is_free(), "Error: should be a free block");
2868 _bt.verify_single_block((HeapWord*)fc, fc->size());
2869 const size_t nn = fc->size() / word_sz;
2870 n = MIN2(nn, n);
2871 assert((ssize_t)n >= 1, "Control point invariant");
2872 rem = fc->size() - n * word_sz;
2873 // If there is a remainder, and it's too small, allocate one fewer.
2874 if (rem > 0 && rem < MinChunkSize) {
2875 n--; rem += word_sz;
2876 }
2877 // Note that at this point we may have n == 0.
2878 assert((ssize_t)n >= 0, "Control point invariant");
2879
2880 // If n is 0, the chunk fc that was found is not large
2881 // enough to leave a viable remainder. We are unable to
2882 // allocate even one block. Return fc to the
2883 // dictionary and return, leaving "fl" empty.
2884 if (n == 0) {
2885 returnChunkToDictionary(fc);
2886 assert(fl->count() == 0, "We never allocated any blocks");
2887 return;
2888 }
2889
2890 // First return the remainder, if any.
2891 // Note that we hold the lock until we decide if we're going to give
2892 // back the remainder to the dictionary, since a concurrent allocation
2893 // may otherwise see the heap as empty. (We're willing to take that
2894 // hit if the block is a small block.)
2895 if (rem > 0) {
2896 size_t prefix_size = n * word_sz;
2897 rem_fc = (FreeChunk*)((HeapWord*)fc + prefix_size);
2898 rem_fc->set_size(rem);
2899 rem_fc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2900 rem_fc->link_next(NULL);
2901 // Above must occur before BOT is updated below.
2902 assert((ssize_t)n > 0 && prefix_size > 0 && rem_fc > fc, "Error");
2903 OrderAccess::storestore();
2904 _bt.split_block((HeapWord*)fc, fc->size(), prefix_size);
2905 assert(fc->is_free(), "Error");
2906 fc->set_size(prefix_size);
2907 if (rem >= IndexSetSize) {
2908 returnChunkToDictionary(rem_fc);
2909 dictionary()->dict_census_udpate(rem, true /*split*/, true /*birth*/);
2910 rem_fc = NULL;
2911 }
2912 // Otherwise, return it to the small list below.
2913 }
2914 }
2915 if (rem_fc != NULL) {
2916 MutexLockerEx x(_indexedFreeListParLocks[rem],
2917 Mutex::_no_safepoint_check_flag);
2918 _bt.verify_not_unallocated((HeapWord*)rem_fc, rem_fc->size());
2919 _indexedFreeList[rem].return_chunk_at_head(rem_fc);
2920 smallSplitBirth(rem);
2921 }
2922 assert((ssize_t)n > 0 && fc != NULL, "Consistency");
2923 // Now do the splitting up.
2924 // Must do this in reverse order, so that anybody attempting to
2925 // access the main chunk sees it as a single free block until we
2926 // change it.
2927 size_t fc_size = n * word_sz;
2928 // All but first chunk in this loop
2929 for (ssize_t i = n-1; i > 0; i--) {
2930 FreeChunk* ffc = (FreeChunk*)((HeapWord*)fc + i * word_sz);
2931 ffc->set_size(word_sz);
2932 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2933 ffc->link_next(NULL);
2934 // Above must occur before BOT is updated below.
2935 OrderAccess::storestore();
2936 // splitting from the right, fc_size == (n - i + 1) * wordsize
2937 _bt.mark_block((HeapWord*)ffc, word_sz, true /* reducing */);
2938 fc_size -= word_sz;
2939 _bt.verify_not_unallocated((HeapWord*)ffc, ffc->size());
2940 _bt.verify_single_block((HeapWord*)ffc, ffc->size());
2941 _bt.verify_single_block((HeapWord*)fc, fc_size);
2942 // Push this on "fl".
2943 fl->return_chunk_at_head(ffc);
2944 }
2945 // First chunk
2946 assert(fc->is_free() && fc->size() == n*word_sz, "Error: should still be a free block");
2947 // The blocks above should show their new sizes before the first block below
2948 fc->set_size(word_sz);
2949 fc->link_prev(NULL); // idempotent wrt free-ness, see assert above
2950 fc->link_next(NULL);
2951 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
2952 _bt.verify_single_block((HeapWord*)fc, fc->size());
2953 fl->return_chunk_at_head(fc);
2954
2955 assert((ssize_t)n > 0 && (ssize_t)n == fl->count(), "Incorrect number of blocks");
2956 {
2957 // Update the stats for this block size.
2958 MutexLockerEx x(_indexedFreeListParLocks[word_sz],
2959 Mutex::_no_safepoint_check_flag);
2960 const ssize_t births = _indexedFreeList[word_sz].split_births() + n;
2961 _indexedFreeList[word_sz].set_split_births(births);
2962 // ssize_t new_surplus = _indexedFreeList[word_sz].surplus() + n;
2963 // _indexedFreeList[word_sz].set_surplus(new_surplus);
2964 }
2965
2966 // TRAP
2967 assert(fl->tail()->next() == NULL, "List invariant.");
2968 }
2969
2970 // Set up the space's par_seq_tasks structure for work claiming
2971 // for parallel rescan. See CMSParRemarkTask where this is currently used.
2972 // XXX Need to suitably abstract and generalize this and the next
2973 // method into one.
2974 void
2975 CompactibleFreeListSpace::
2976 initialize_sequential_subtasks_for_rescan(int n_threads) {
2977 // The "size" of each task is fixed according to rescan_task_size.
2978 assert(n_threads > 0, "Unexpected n_threads argument");
2979 const size_t task_size = rescan_task_size();
2980 size_t n_tasks = (used_region().word_size() + task_size - 1)/task_size;
2981 assert((n_tasks == 0) == used_region().is_empty(), "n_tasks incorrect");
2982 assert(n_tasks == 0 ||
2983 ((used_region().start() + (n_tasks - 1)*task_size < used_region().end()) &&
2984 (used_region().start() + n_tasks*task_size >= used_region().end())),
2985 "n_tasks calculation incorrect");
2986 SequentialSubTasksDone* pst = conc_par_seq_tasks();
2987 assert(!pst->valid(), "Clobbering existing data?");
2988 // Sets the condition for completion of the subtask (how many threads
2989 // need to finish in order to be done).
2990 pst->set_n_threads(n_threads);
2991 pst->set_n_tasks((int)n_tasks);
2992 }
2993
2994 // Set up the space's par_seq_tasks structure for work claiming
2995 // for parallel concurrent marking. See CMSConcMarkTask where this is currently used.
2996 void
2997 CompactibleFreeListSpace::
2998 initialize_sequential_subtasks_for_marking(int n_threads,
2999 HeapWord* low) {
3000 // The "size" of each task is fixed according to rescan_task_size.
3001 assert(n_threads > 0, "Unexpected n_threads argument");
3002 const size_t task_size = marking_task_size();
3003 assert(task_size > CardTableModRefBS::card_size_in_words &&
3004 (task_size % CardTableModRefBS::card_size_in_words == 0),
3005 "Otherwise arithmetic below would be incorrect");
3006 MemRegion span = _gen->reserved();
3007 if (low != NULL) {
3008 if (span.contains(low)) {
3009 // Align low down to a card boundary so that
3010 // we can use block_offset_careful() on span boundaries.
3011 HeapWord* aligned_low = (HeapWord*)align_size_down((uintptr_t)low,
3012 CardTableModRefBS::card_size);
3013 // Clip span prefix at aligned_low
3014 span = span.intersection(MemRegion(aligned_low, span.end()));
3015 } else if (low > span.end()) {
3016 span = MemRegion(low, low); // Null region
3017 } // else use entire span
3018 }
3019 assert(span.is_empty() ||
3020 ((uintptr_t)span.start() % CardTableModRefBS::card_size == 0),
3021 "span should start at a card boundary");
3022 size_t n_tasks = (span.word_size() + task_size - 1)/task_size;
3023 assert((n_tasks == 0) == span.is_empty(), "Inconsistency");
3024 assert(n_tasks == 0 ||
3025 ((span.start() + (n_tasks - 1)*task_size < span.end()) &&
3026 (span.start() + n_tasks*task_size >= span.end())),
3027 "n_tasks calculation incorrect");
3028 SequentialSubTasksDone* pst = conc_par_seq_tasks();
3029 assert(!pst->valid(), "Clobbering existing data?");
3030 // Sets the condition for completion of the subtask (how many threads
3031 // need to finish in order to be done).
3032 pst->set_n_threads(n_threads);
3033 pst->set_n_tasks((int)n_tasks);
3034 }