1 /*
2 * Copyright (c) 2001, 2015, 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.inline.hpp"
33 #include "memory/allocation.inline.hpp"
34 #include "memory/blockOffsetTable.inline.hpp"
35 #include "memory/resourceArea.hpp"
36 #include "memory/space.inline.hpp"
37 #include "memory/universe.inline.hpp"
38 #include "oops/oop.inline.hpp"
39 #include "runtime/globals.hpp"
40 #include "runtime/handles.inline.hpp"
41 #include "runtime/init.hpp"
42 #include "runtime/java.hpp"
43 #include "runtime/orderAccess.inline.hpp"
44 #include "runtime/vmThread.hpp"
45 #include "utilities/copy.hpp"
46
47 /////////////////////////////////////////////////////////////////////////
48 //// CompactibleFreeListSpace
49 /////////////////////////////////////////////////////////////////////////
50
51 // highest ranked free list lock rank
52 int CompactibleFreeListSpace::_lockRank = Mutex::leaf + 3;
53
54 // Defaults are 0 so things will break badly if incorrectly initialized.
55 size_t CompactibleFreeListSpace::IndexSetStart = 0;
56 size_t CompactibleFreeListSpace::IndexSetStride = 0;
57
58 size_t MinChunkSize = 0;
59
60 void CompactibleFreeListSpace::set_cms_values() {
61 // Set CMS global values
62 assert(MinChunkSize == 0, "already set");
63
64 // MinChunkSize should be a multiple of MinObjAlignment and be large enough
65 // for chunks to contain a FreeChunk.
66 size_t min_chunk_size_in_bytes = align_size_up(sizeof(FreeChunk), MinObjAlignmentInBytes);
67 MinChunkSize = min_chunk_size_in_bytes / BytesPerWord;
68
69 assert(IndexSetStart == 0 && IndexSetStride == 0, "already set");
70 IndexSetStart = MinChunkSize;
71 IndexSetStride = MinObjAlignment;
72 }
73
74 // Constructor
75 CompactibleFreeListSpace::CompactibleFreeListSpace(BlockOffsetSharedArray* bs,
76 MemRegion mr, bool use_adaptive_freelists,
77 FreeBlockDictionary<FreeChunk>::DictionaryChoice dictionaryChoice) :
78 _dictionaryChoice(dictionaryChoice),
79 _adaptive_freelists(use_adaptive_freelists),
80 _bt(bs, mr),
81 // free list locks are in the range of values taken by _lockRank
82 // This range currently is [_leaf+2, _leaf+3]
83 // Note: this requires that CFLspace c'tors
84 // are called serially in the order in which the locks are
85 // are acquired in the program text. This is true today.
86 _freelistLock(_lockRank--, "CompactibleFreeListSpace._lock", true,
87 Monitor::_safepoint_check_sometimes),
88 _parDictionaryAllocLock(Mutex::leaf - 1, // == rank(ExpandHeap_lock) - 1
89 "CompactibleFreeListSpace._dict_par_lock", true,
90 Monitor::_safepoint_check_never),
91 _rescan_task_size(CardTableModRefBS::card_size_in_words * BitsPerWord *
92 CMSRescanMultiple),
93 _marking_task_size(CardTableModRefBS::card_size_in_words * BitsPerWord *
94 CMSConcMarkMultiple),
95 _collector(NULL),
96 _preconsumptionDirtyCardClosure(NULL)
97 {
98 assert(sizeof(FreeChunk) / BytesPerWord <= MinChunkSize,
99 "FreeChunk is larger than expected");
100 _bt.set_space(this);
101 initialize(mr, SpaceDecorator::Clear, SpaceDecorator::Mangle);
102 // We have all of "mr", all of which we place in the dictionary
103 // as one big chunk. We'll need to decide here which of several
104 // possible alternative dictionary implementations to use. For
105 // now the choice is easy, since we have only one working
106 // implementation, namely, the simple binary tree (splaying
107 // temporarily disabled).
108 switch (dictionaryChoice) {
109 case FreeBlockDictionary<FreeChunk>::dictionaryBinaryTree:
110 _dictionary = new AFLBinaryTreeDictionary(mr);
111 break;
112 case FreeBlockDictionary<FreeChunk>::dictionarySplayTree:
113 case FreeBlockDictionary<FreeChunk>::dictionarySkipList:
114 default:
115 warning("dictionaryChoice: selected option not understood; using"
116 " default BinaryTreeDictionary implementation instead.");
117 }
118 assert(_dictionary != NULL, "CMS dictionary initialization");
119 // The indexed free lists are initially all empty and are lazily
120 // filled in on demand. Initialize the array elements to NULL.
121 initializeIndexedFreeListArray();
122
123 // Not using adaptive free lists assumes that allocation is first
124 // from the linAB's. Also a cms perm gen which can be compacted
125 // has to have the klass's klassKlass allocated at a lower
126 // address in the heap than the klass so that the klassKlass is
127 // moved to its new location before the klass is moved.
128 // Set the _refillSize for the linear allocation blocks
129 if (!use_adaptive_freelists) {
130 FreeChunk* fc = _dictionary->get_chunk(mr.word_size(),
131 FreeBlockDictionary<FreeChunk>::atLeast);
132 // The small linAB initially has all the space and will allocate
133 // a chunk of any size.
134 HeapWord* addr = (HeapWord*) fc;
135 _smallLinearAllocBlock.set(addr, fc->size() ,
136 1024*SmallForLinearAlloc, fc->size());
137 // Note that _unallocated_block is not updated here.
138 // Allocations from the linear allocation block should
139 // update it.
140 } else {
141 _smallLinearAllocBlock.set(0, 0, 1024*SmallForLinearAlloc,
142 SmallForLinearAlloc);
143 }
144 // CMSIndexedFreeListReplenish should be at least 1
145 CMSIndexedFreeListReplenish = MAX2((uintx)1, CMSIndexedFreeListReplenish);
146 _promoInfo.setSpace(this);
147 if (UseCMSBestFit) {
148 _fitStrategy = FreeBlockBestFitFirst;
149 } else {
150 _fitStrategy = FreeBlockStrategyNone;
151 }
152 check_free_list_consistency();
153
154 // Initialize locks for parallel case.
155 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
156 _indexedFreeListParLocks[i] = new Mutex(Mutex::leaf - 1, // == ExpandHeap_lock - 1
157 "a freelist par lock", true, Mutex::_safepoint_check_sometimes);
158 DEBUG_ONLY(
159 _indexedFreeList[i].set_protecting_lock(_indexedFreeListParLocks[i]);
160 )
161 }
162 _dictionary->set_par_lock(&_parDictionaryAllocLock);
163 }
164
165 // Like CompactibleSpace forward() but always calls cross_threshold() to
166 // update the block offset table. Removed initialize_threshold call because
167 // CFLS does not use a block offset array for contiguous spaces.
168 HeapWord* CompactibleFreeListSpace::forward(oop q, size_t size,
169 CompactPoint* cp, HeapWord* compact_top) {
170 // q is alive
171 // First check if we should switch compaction space
172 assert(this == cp->space, "'this' should be current compaction space.");
173 size_t compaction_max_size = pointer_delta(end(), compact_top);
174 assert(adjustObjectSize(size) == cp->space->adjust_object_size_v(size),
175 "virtual adjustObjectSize_v() method is not correct");
176 size_t adjusted_size = adjustObjectSize(size);
177 assert(compaction_max_size >= MinChunkSize || compaction_max_size == 0,
178 "no small fragments allowed");
179 assert(minimum_free_block_size() == MinChunkSize,
180 "for de-virtualized reference below");
181 // Can't leave a nonzero size, residual fragment smaller than MinChunkSize
182 if (adjusted_size + MinChunkSize > compaction_max_size &&
183 adjusted_size != compaction_max_size) {
184 do {
185 // switch to next compaction space
186 cp->space->set_compaction_top(compact_top);
187 cp->space = cp->space->next_compaction_space();
188 if (cp->space == NULL) {
189 cp->gen = GenCollectedHeap::heap()->prev_gen(cp->gen);
190 assert(cp->gen != NULL, "compaction must succeed");
191 cp->space = cp->gen->first_compaction_space();
192 assert(cp->space != NULL, "generation must have a first compaction space");
193 }
194 compact_top = cp->space->bottom();
195 cp->space->set_compaction_top(compact_top);
196 // The correct adjusted_size may not be the same as that for this method
197 // (i.e., cp->space may no longer be "this" so adjust the size again.
198 // Use the virtual method which is not used above to save the virtual
199 // dispatch.
200 adjusted_size = cp->space->adjust_object_size_v(size);
201 compaction_max_size = pointer_delta(cp->space->end(), compact_top);
202 assert(cp->space->minimum_free_block_size() == 0, "just checking");
203 } while (adjusted_size > compaction_max_size);
204 }
205
206 // store the forwarding pointer into the mark word
207 if ((HeapWord*)q != compact_top) {
208 q->forward_to(oop(compact_top));
209 assert(q->is_gc_marked(), "encoding the pointer should preserve the mark");
210 } else {
211 // if the object isn't moving we can just set the mark to the default
212 // mark and handle it specially later on.
213 q->init_mark();
214 assert(q->forwardee() == NULL, "should be forwarded to NULL");
215 }
216
217 compact_top += adjusted_size;
218
219 // we need to update the offset table so that the beginnings of objects can be
220 // found during scavenge. Note that we are updating the offset table based on
221 // where the object will be once the compaction phase finishes.
222
223 // Always call cross_threshold(). A contiguous space can only call it when
224 // the compaction_top exceeds the current threshold but not for an
225 // non-contiguous space.
226 cp->threshold =
227 cp->space->cross_threshold(compact_top - adjusted_size, compact_top);
228 return compact_top;
229 }
230
231 // A modified copy of OffsetTableContigSpace::cross_threshold() with _offsets -> _bt
232 // and use of single_block instead of alloc_block. The name here is not really
233 // appropriate - maybe a more general name could be invented for both the
234 // contiguous and noncontiguous spaces.
235
236 HeapWord* CompactibleFreeListSpace::cross_threshold(HeapWord* start, HeapWord* the_end) {
237 _bt.single_block(start, the_end);
238 return end();
239 }
240
241 // Initialize them to NULL.
242 void CompactibleFreeListSpace::initializeIndexedFreeListArray() {
243 for (size_t i = 0; i < IndexSetSize; i++) {
244 // Note that on platforms where objects are double word aligned,
245 // the odd array elements are not used. It is convenient, however,
246 // to map directly from the object size to the array element.
247 _indexedFreeList[i].reset(IndexSetSize);
248 _indexedFreeList[i].set_size(i);
249 assert(_indexedFreeList[i].count() == 0, "reset check failed");
250 assert(_indexedFreeList[i].head() == NULL, "reset check failed");
251 assert(_indexedFreeList[i].tail() == NULL, "reset check failed");
252 assert(_indexedFreeList[i].hint() == IndexSetSize, "reset check failed");
253 }
254 }
255
256 void CompactibleFreeListSpace::resetIndexedFreeListArray() {
257 for (size_t i = 1; i < IndexSetSize; i++) {
258 assert(_indexedFreeList[i].size() == (size_t) i,
259 "Indexed free list sizes are incorrect");
260 _indexedFreeList[i].reset(IndexSetSize);
261 assert(_indexedFreeList[i].count() == 0, "reset check failed");
262 assert(_indexedFreeList[i].head() == NULL, "reset check failed");
263 assert(_indexedFreeList[i].tail() == NULL, "reset check failed");
264 assert(_indexedFreeList[i].hint() == IndexSetSize, "reset check failed");
265 }
266 }
267
268 void CompactibleFreeListSpace::reset(MemRegion mr) {
269 resetIndexedFreeListArray();
270 dictionary()->reset();
271 if (BlockOffsetArrayUseUnallocatedBlock) {
272 assert(end() == mr.end(), "We are compacting to the bottom of CMS gen");
273 // Everything's allocated until proven otherwise.
274 _bt.set_unallocated_block(end());
275 }
276 if (!mr.is_empty()) {
277 assert(mr.word_size() >= MinChunkSize, "Chunk size is too small");
278 _bt.single_block(mr.start(), mr.word_size());
279 FreeChunk* fc = (FreeChunk*) mr.start();
280 fc->set_size(mr.word_size());
281 if (mr.word_size() >= IndexSetSize ) {
282 returnChunkToDictionary(fc);
283 } else {
284 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
285 _indexedFreeList[mr.word_size()].return_chunk_at_head(fc);
286 }
287 coalBirth(mr.word_size());
288 }
289 _promoInfo.reset();
290 _smallLinearAllocBlock._ptr = NULL;
291 _smallLinearAllocBlock._word_size = 0;
292 }
293
294 void CompactibleFreeListSpace::reset_after_compaction() {
295 // Reset the space to the new reality - one free chunk.
296 MemRegion mr(compaction_top(), end());
297 reset(mr);
298 // Now refill the linear allocation block(s) if possible.
299 if (_adaptive_freelists) {
300 refillLinearAllocBlocksIfNeeded();
301 } else {
302 // Place as much of mr in the linAB as we can get,
303 // provided it was big enough to go into the dictionary.
304 FreeChunk* fc = dictionary()->find_largest_dict();
305 if (fc != NULL) {
306 assert(fc->size() == mr.word_size(),
307 "Why was the chunk broken up?");
308 removeChunkFromDictionary(fc);
309 HeapWord* addr = (HeapWord*) fc;
310 _smallLinearAllocBlock.set(addr, fc->size() ,
311 1024*SmallForLinearAlloc, fc->size());
312 // Note that _unallocated_block is not updated here.
313 }
314 }
315 }
316
317 // Walks the entire dictionary, returning a coterminal
318 // chunk, if it exists. Use with caution since it involves
319 // a potentially complete walk of a potentially large tree.
320 FreeChunk* CompactibleFreeListSpace::find_chunk_at_end() {
321
322 assert_lock_strong(&_freelistLock);
323
324 return dictionary()->find_chunk_ends_at(end());
325 }
326
327
328 #ifndef PRODUCT
329 void CompactibleFreeListSpace::initializeIndexedFreeListArrayReturnedBytes() {
330 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
331 _indexedFreeList[i].allocation_stats()->set_returned_bytes(0);
332 }
333 }
334
335 size_t CompactibleFreeListSpace::sumIndexedFreeListArrayReturnedBytes() {
336 size_t sum = 0;
337 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
338 sum += _indexedFreeList[i].allocation_stats()->returned_bytes();
339 }
340 return sum;
341 }
342
343 size_t CompactibleFreeListSpace::totalCountInIndexedFreeLists() const {
344 size_t count = 0;
345 for (size_t i = IndexSetStart; i < IndexSetSize; i++) {
346 debug_only(
347 ssize_t total_list_count = 0;
348 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
349 fc = fc->next()) {
350 total_list_count++;
351 }
352 assert(total_list_count == _indexedFreeList[i].count(),
353 "Count in list is incorrect");
354 )
355 count += _indexedFreeList[i].count();
356 }
357 return count;
358 }
359
360 size_t CompactibleFreeListSpace::totalCount() {
361 size_t num = totalCountInIndexedFreeLists();
362 num += dictionary()->total_count();
363 if (_smallLinearAllocBlock._word_size != 0) {
364 num++;
365 }
366 return num;
367 }
368 #endif
369
370 bool CompactibleFreeListSpace::is_free_block(const HeapWord* p) const {
371 FreeChunk* fc = (FreeChunk*) p;
372 return fc->is_free();
373 }
374
375 size_t CompactibleFreeListSpace::used() const {
376 return capacity() - free();
377 }
378
379 size_t CompactibleFreeListSpace::free() const {
380 // "MT-safe, but not MT-precise"(TM), if you will: i.e.
381 // if you do this while the structures are in flux you
382 // may get an approximate answer only; for instance
383 // because there is concurrent allocation either
384 // directly by mutators or for promotion during a GC.
385 // It's "MT-safe", however, in the sense that you are guaranteed
386 // not to crash and burn, for instance, because of walking
387 // pointers that could disappear as you were walking them.
388 // The approximation is because the various components
389 // that are read below are not read atomically (and
390 // further the computation of totalSizeInIndexedFreeLists()
391 // is itself a non-atomic computation. The normal use of
392 // this is during a resize operation at the end of GC
393 // and at that time you are guaranteed to get the
394 // correct actual value. However, for instance, this is
395 // also read completely asynchronously by the "perf-sampler"
396 // that supports jvmstat, and you are apt to see the values
397 // flicker in such cases.
398 assert(_dictionary != NULL, "No _dictionary?");
399 return (_dictionary->total_chunk_size(DEBUG_ONLY(freelistLock())) +
400 totalSizeInIndexedFreeLists() +
401 _smallLinearAllocBlock._word_size) * HeapWordSize;
402 }
403
404 size_t CompactibleFreeListSpace::max_alloc_in_words() const {
405 assert(_dictionary != NULL, "No _dictionary?");
406 assert_locked();
407 size_t res = _dictionary->max_chunk_size();
408 res = MAX2(res, MIN2(_smallLinearAllocBlock._word_size,
409 (size_t) SmallForLinearAlloc - 1));
410 // XXX the following could potentially be pretty slow;
411 // should one, pessimistically for the rare cases when res
412 // calculated above is less than IndexSetSize,
413 // just return res calculated above? My reasoning was that
414 // those cases will be so rare that the extra time spent doesn't
415 // really matter....
416 // Note: do not change the loop test i >= res + IndexSetStride
417 // to i > res below, because i is unsigned and res may be zero.
418 for (size_t i = IndexSetSize - 1; i >= res + IndexSetStride;
419 i -= IndexSetStride) {
420 if (_indexedFreeList[i].head() != NULL) {
421 assert(_indexedFreeList[i].count() != 0, "Inconsistent FreeList");
422 return i;
423 }
424 }
425 return res;
426 }
427
428 void LinearAllocBlock::print_on(outputStream* st) const {
429 st->print_cr(" LinearAllocBlock: ptr = " PTR_FORMAT ", word_size = " SIZE_FORMAT
430 ", refillsize = " SIZE_FORMAT ", allocation_size_limit = " SIZE_FORMAT,
431 p2i(_ptr), _word_size, _refillSize, _allocation_size_limit);
432 }
433
434 void CompactibleFreeListSpace::print_on(outputStream* st) const {
435 st->print_cr("COMPACTIBLE FREELIST SPACE");
436 st->print_cr(" Space:");
437 Space::print_on(st);
438
439 st->print_cr("promoInfo:");
440 _promoInfo.print_on(st);
441
442 st->print_cr("_smallLinearAllocBlock");
443 _smallLinearAllocBlock.print_on(st);
444
445 // dump_memory_block(_smallLinearAllocBlock->_ptr, 128);
446
447 st->print_cr(" _fitStrategy = %s, _adaptive_freelists = %s",
448 _fitStrategy?"true":"false", _adaptive_freelists?"true":"false");
449 }
450
451 void CompactibleFreeListSpace::print_indexed_free_lists(outputStream* st)
452 const {
453 reportIndexedFreeListStatistics();
454 gclog_or_tty->print_cr("Layout of Indexed Freelists");
455 gclog_or_tty->print_cr("---------------------------");
456 AdaptiveFreeList<FreeChunk>::print_labels_on(st, "size");
457 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
458 _indexedFreeList[i].print_on(gclog_or_tty);
459 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
460 fc = fc->next()) {
461 gclog_or_tty->print_cr("\t[" PTR_FORMAT "," PTR_FORMAT ") %s",
462 p2i(fc), p2i((HeapWord*)fc + i),
463 fc->cantCoalesce() ? "\t CC" : "");
464 }
465 }
466 }
467
468 void CompactibleFreeListSpace::print_promo_info_blocks(outputStream* st)
469 const {
470 _promoInfo.print_on(st);
471 }
472
473 void CompactibleFreeListSpace::print_dictionary_free_lists(outputStream* st)
474 const {
475 _dictionary->report_statistics();
476 st->print_cr("Layout of Freelists in Tree");
477 st->print_cr("---------------------------");
478 _dictionary->print_free_lists(st);
479 }
480
481 class BlkPrintingClosure: public BlkClosure {
482 const CMSCollector* _collector;
483 const CompactibleFreeListSpace* _sp;
484 const CMSBitMap* _live_bit_map;
485 const bool _post_remark;
486 outputStream* _st;
487 public:
488 BlkPrintingClosure(const CMSCollector* collector,
489 const CompactibleFreeListSpace* sp,
490 const CMSBitMap* live_bit_map,
491 outputStream* st):
492 _collector(collector),
493 _sp(sp),
494 _live_bit_map(live_bit_map),
495 _post_remark(collector->abstract_state() > CMSCollector::FinalMarking),
496 _st(st) { }
497 size_t do_blk(HeapWord* addr);
498 };
499
500 size_t BlkPrintingClosure::do_blk(HeapWord* addr) {
501 size_t sz = _sp->block_size_no_stall(addr, _collector);
502 assert(sz != 0, "Should always be able to compute a size");
503 if (_sp->block_is_obj(addr)) {
504 const bool dead = _post_remark && !_live_bit_map->isMarked(addr);
505 _st->print_cr(PTR_FORMAT ": %s object of size " SIZE_FORMAT "%s",
506 p2i(addr),
507 dead ? "dead" : "live",
508 sz,
509 (!dead && CMSPrintObjectsInDump) ? ":" : ".");
510 if (CMSPrintObjectsInDump && !dead) {
511 oop(addr)->print_on(_st);
512 _st->print_cr("--------------------------------------");
513 }
514 } else { // free block
515 _st->print_cr(PTR_FORMAT ": free block of size " SIZE_FORMAT "%s",
516 p2i(addr), sz, CMSPrintChunksInDump ? ":" : ".");
517 if (CMSPrintChunksInDump) {
518 ((FreeChunk*)addr)->print_on(_st);
519 _st->print_cr("--------------------------------------");
520 }
521 }
522 return sz;
523 }
524
525 void CompactibleFreeListSpace::dump_at_safepoint_with_locks(CMSCollector* c,
526 outputStream* st) {
527 st->print_cr("\n=========================");
528 st->print_cr("Block layout in CMS Heap:");
529 st->print_cr("=========================");
530 BlkPrintingClosure bpcl(c, this, c->markBitMap(), st);
531 blk_iterate(&bpcl);
532
533 st->print_cr("\n=======================================");
534 st->print_cr("Order & Layout of Promotion Info Blocks");
535 st->print_cr("=======================================");
536 print_promo_info_blocks(st);
537
538 st->print_cr("\n===========================");
539 st->print_cr("Order of Indexed Free Lists");
540 st->print_cr("=========================");
541 print_indexed_free_lists(st);
542
543 st->print_cr("\n=================================");
544 st->print_cr("Order of Free Lists in Dictionary");
545 st->print_cr("=================================");
546 print_dictionary_free_lists(st);
547 }
548
549
550 void CompactibleFreeListSpace::reportFreeListStatistics() const {
551 assert_lock_strong(&_freelistLock);
552 assert(PrintFLSStatistics != 0, "Reporting error");
553 _dictionary->report_statistics();
554 if (PrintFLSStatistics > 1) {
555 reportIndexedFreeListStatistics();
556 size_t total_size = totalSizeInIndexedFreeLists() +
557 _dictionary->total_chunk_size(DEBUG_ONLY(freelistLock()));
558 gclog_or_tty->print(" free=" SIZE_FORMAT " frag=%1.4f\n", total_size, flsFrag());
559 }
560 }
561
562 void CompactibleFreeListSpace::reportIndexedFreeListStatistics() const {
563 assert_lock_strong(&_freelistLock);
564 gclog_or_tty->print("Statistics for IndexedFreeLists:\n"
565 "--------------------------------\n");
566 size_t total_size = totalSizeInIndexedFreeLists();
567 size_t free_blocks = numFreeBlocksInIndexedFreeLists();
568 gclog_or_tty->print("Total Free Space: " SIZE_FORMAT "\n", total_size);
569 gclog_or_tty->print("Max Chunk Size: " SIZE_FORMAT "\n", maxChunkSizeInIndexedFreeLists());
570 gclog_or_tty->print("Number of Blocks: " SIZE_FORMAT "\n", free_blocks);
571 if (free_blocks != 0) {
572 gclog_or_tty->print("Av. Block Size: " SIZE_FORMAT "\n", total_size/free_blocks);
573 }
574 }
575
576 size_t CompactibleFreeListSpace::numFreeBlocksInIndexedFreeLists() const {
577 size_t res = 0;
578 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
579 debug_only(
580 ssize_t recount = 0;
581 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
582 fc = fc->next()) {
583 recount += 1;
584 }
585 assert(recount == _indexedFreeList[i].count(),
586 "Incorrect count in list");
587 )
588 res += _indexedFreeList[i].count();
589 }
590 return res;
591 }
592
593 size_t CompactibleFreeListSpace::maxChunkSizeInIndexedFreeLists() const {
594 for (size_t i = IndexSetSize - 1; i != 0; i -= IndexSetStride) {
595 if (_indexedFreeList[i].head() != NULL) {
596 assert(_indexedFreeList[i].count() != 0, "Inconsistent FreeList");
597 return (size_t)i;
598 }
599 }
600 return 0;
601 }
602
603 void CompactibleFreeListSpace::set_end(HeapWord* value) {
604 HeapWord* prevEnd = end();
605 assert(prevEnd != value, "unnecessary set_end call");
606 assert(prevEnd == NULL || !BlockOffsetArrayUseUnallocatedBlock || value >= unallocated_block(),
607 "New end is below unallocated block");
608 _end = value;
609 if (prevEnd != NULL) {
610 // Resize the underlying block offset table.
611 _bt.resize(pointer_delta(value, bottom()));
612 if (value <= prevEnd) {
613 assert(!BlockOffsetArrayUseUnallocatedBlock || value >= unallocated_block(),
614 "New end is below unallocated block");
615 } else {
616 // Now, take this new chunk and add it to the free blocks.
617 // Note that the BOT has not yet been updated for this block.
618 size_t newFcSize = pointer_delta(value, prevEnd);
619 // XXX This is REALLY UGLY and should be fixed up. XXX
620 if (!_adaptive_freelists && _smallLinearAllocBlock._ptr == NULL) {
621 // Mark the boundary of the new block in BOT
622 _bt.mark_block(prevEnd, value);
623 // put it all in the linAB
624 MutexLockerEx x(parDictionaryAllocLock(),
625 Mutex::_no_safepoint_check_flag);
626 _smallLinearAllocBlock._ptr = prevEnd;
627 _smallLinearAllocBlock._word_size = newFcSize;
628 repairLinearAllocBlock(&_smallLinearAllocBlock);
629 // Births of chunks put into a LinAB are not recorded. Births
630 // of chunks as they are allocated out of a LinAB are.
631 } else {
632 // Add the block to the free lists, if possible coalescing it
633 // with the last free block, and update the BOT and census data.
634 addChunkToFreeListsAtEndRecordingStats(prevEnd, newFcSize);
635 }
636 }
637 }
638 }
639
640 class FreeListSpace_DCTOC : public Filtering_DCTOC {
641 CompactibleFreeListSpace* _cfls;
642 CMSCollector* _collector;
643 protected:
644 // Override.
645 #define walk_mem_region_with_cl_DECL(ClosureType) \
646 virtual void walk_mem_region_with_cl(MemRegion mr, \
647 HeapWord* bottom, HeapWord* top, \
648 ClosureType* cl); \
649 void walk_mem_region_with_cl_par(MemRegion mr, \
650 HeapWord* bottom, HeapWord* top, \
651 ClosureType* cl); \
652 void walk_mem_region_with_cl_nopar(MemRegion mr, \
653 HeapWord* bottom, HeapWord* top, \
654 ClosureType* cl)
655 walk_mem_region_with_cl_DECL(ExtendedOopClosure);
656 walk_mem_region_with_cl_DECL(FilteringClosure);
657
658 public:
659 FreeListSpace_DCTOC(CompactibleFreeListSpace* sp,
660 CMSCollector* collector,
661 ExtendedOopClosure* cl,
662 CardTableModRefBS::PrecisionStyle precision,
663 HeapWord* boundary) :
664 Filtering_DCTOC(sp, cl, precision, boundary),
665 _cfls(sp), _collector(collector) {}
666 };
667
668 // We de-virtualize the block-related calls below, since we know that our
669 // space is a CompactibleFreeListSpace.
670
671 #define FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(ClosureType) \
672 void FreeListSpace_DCTOC::walk_mem_region_with_cl(MemRegion mr, \
673 HeapWord* bottom, \
674 HeapWord* top, \
675 ClosureType* cl) { \
676 bool is_par = SharedHeap::heap()->n_par_threads() > 0; \
677 if (is_par) { \
678 assert(SharedHeap::heap()->n_par_threads() == \
679 SharedHeap::heap()->workers()->active_workers(), "Mismatch"); \
680 walk_mem_region_with_cl_par(mr, bottom, top, cl); \
681 } else { \
682 walk_mem_region_with_cl_nopar(mr, bottom, top, cl); \
683 } \
684 } \
685 void FreeListSpace_DCTOC::walk_mem_region_with_cl_par(MemRegion mr, \
686 HeapWord* bottom, \
687 HeapWord* top, \
688 ClosureType* cl) { \
689 /* Skip parts that are before "mr", in case "block_start" sent us \
690 back too far. */ \
691 HeapWord* mr_start = mr.start(); \
692 size_t bot_size = _cfls->CompactibleFreeListSpace::block_size(bottom); \
693 HeapWord* next = bottom + bot_size; \
694 while (next < mr_start) { \
695 bottom = next; \
696 bot_size = _cfls->CompactibleFreeListSpace::block_size(bottom); \
697 next = bottom + bot_size; \
698 } \
699 \
700 while (bottom < top) { \
701 if (_cfls->CompactibleFreeListSpace::block_is_obj(bottom) && \
702 !_cfls->CompactibleFreeListSpace::obj_allocated_since_save_marks( \
703 oop(bottom)) && \
704 !_collector->CMSCollector::is_dead_obj(oop(bottom))) { \
705 size_t word_sz = oop(bottom)->oop_iterate(cl, mr); \
706 bottom += _cfls->adjustObjectSize(word_sz); \
707 } else { \
708 bottom += _cfls->CompactibleFreeListSpace::block_size(bottom); \
709 } \
710 } \
711 } \
712 void FreeListSpace_DCTOC::walk_mem_region_with_cl_nopar(MemRegion mr, \
713 HeapWord* bottom, \
714 HeapWord* top, \
715 ClosureType* cl) { \
716 /* Skip parts that are before "mr", in case "block_start" sent us \
717 back too far. */ \
718 HeapWord* mr_start = mr.start(); \
719 size_t bot_size = _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
720 HeapWord* next = bottom + bot_size; \
721 while (next < mr_start) { \
722 bottom = next; \
723 bot_size = _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
724 next = bottom + bot_size; \
725 } \
726 \
727 while (bottom < top) { \
728 if (_cfls->CompactibleFreeListSpace::block_is_obj_nopar(bottom) && \
729 !_cfls->CompactibleFreeListSpace::obj_allocated_since_save_marks( \
730 oop(bottom)) && \
731 !_collector->CMSCollector::is_dead_obj(oop(bottom))) { \
732 size_t word_sz = oop(bottom)->oop_iterate(cl, mr); \
733 bottom += _cfls->adjustObjectSize(word_sz); \
734 } else { \
735 bottom += _cfls->CompactibleFreeListSpace::block_size_nopar(bottom); \
736 } \
737 } \
738 }
739
740 // (There are only two of these, rather than N, because the split is due
741 // only to the introduction of the FilteringClosure, a local part of the
742 // impl of this abstraction.)
743 FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(ExtendedOopClosure)
744 FreeListSpace_DCTOC__walk_mem_region_with_cl_DEFN(FilteringClosure)
745
746 DirtyCardToOopClosure*
747 CompactibleFreeListSpace::new_dcto_cl(ExtendedOopClosure* cl,
748 CardTableModRefBS::PrecisionStyle precision,
749 HeapWord* boundary) {
750 return new FreeListSpace_DCTOC(this, _collector, cl, precision, boundary);
751 }
752
753
754 // Note on locking for the space iteration functions:
755 // since the collector's iteration activities are concurrent with
756 // allocation activities by mutators, absent a suitable mutual exclusion
757 // mechanism the iterators may go awry. For instance a block being iterated
758 // may suddenly be allocated or divided up and part of it allocated and
759 // so on.
760
761 // Apply the given closure to each block in the space.
762 void CompactibleFreeListSpace::blk_iterate_careful(BlkClosureCareful* cl) {
763 assert_lock_strong(freelistLock());
764 HeapWord *cur, *limit;
765 size_t res;
766 for (cur = bottom(), limit = end(); cur < limit; cur += res) {
767 res = cl->do_blk_careful(cur);
768 assert(cur + res > cur, "Not monotonically increasing ?");
769 }
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(ExtendedOopClosure* 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 // NOTE: In the following methods, in order to safely be able to
795 // apply the closure to an object, we need to be sure that the
796 // object has been initialized. We are guaranteed that an object
797 // is initialized if we are holding the Heap_lock with the
798 // world stopped.
799 void CompactibleFreeListSpace::verify_objects_initialized() const {
800 if (is_init_completed()) {
801 assert_locked_or_safepoint(Heap_lock);
802 if (Universe::is_fully_initialized()) {
803 guarantee(SafepointSynchronize::is_at_safepoint(),
804 "Required for objects to be initialized");
805 }
806 } // else make a concession at vm start-up
807 }
808
809 // Apply the given closure to each object in the space
810 void CompactibleFreeListSpace::object_iterate(ObjectClosure* blk) {
811 assert_lock_strong(freelistLock());
812 NOT_PRODUCT(verify_objects_initialized());
813 HeapWord *cur, *limit;
814 size_t curSize;
815 for (cur = bottom(), limit = end(); cur < limit;
816 cur += curSize) {
817 curSize = block_size(cur);
818 if (block_is_obj(cur)) {
819 blk->do_object(oop(cur));
820 }
821 }
822 }
823
824 // Apply the given closure to each live object in the space
825 // The usage of CompactibleFreeListSpace
826 // by the ConcurrentMarkSweepGeneration for concurrent GC's allows
827 // objects in the space with references to objects that are no longer
828 // valid. For example, an object may reference another object
829 // that has already been sweep up (collected). This method uses
830 // obj_is_alive() to determine whether it is safe to apply the closure to
831 // an object. See obj_is_alive() for details on how liveness of an
832 // object is decided.
833
834 void CompactibleFreeListSpace::safe_object_iterate(ObjectClosure* blk) {
835 assert_lock_strong(freelistLock());
836 NOT_PRODUCT(verify_objects_initialized());
837 HeapWord *cur, *limit;
838 size_t curSize;
839 for (cur = bottom(), limit = end(); cur < limit;
840 cur += curSize) {
841 curSize = block_size(cur);
842 if (block_is_obj(cur) && obj_is_alive(cur)) {
843 blk->do_object(oop(cur));
844 }
845 }
846 }
847
848 void CompactibleFreeListSpace::object_iterate_mem(MemRegion mr,
849 UpwardsObjectClosure* cl) {
850 assert_locked(freelistLock());
851 NOT_PRODUCT(verify_objects_initialized());
852 assert(!mr.is_empty(), "Should be non-empty");
853 // We use MemRegion(bottom(), end()) rather than used_region() below
854 // because the two are not necessarily equal for some kinds of
855 // spaces, in particular, certain kinds of free list spaces.
856 // We could use the more complicated but more precise:
857 // MemRegion(used_region().start(), round_to(used_region().end(), CardSize))
858 // but the slight imprecision seems acceptable in the assertion check.
859 assert(MemRegion(bottom(), end()).contains(mr),
860 "Should be within used space");
861 HeapWord* prev = cl->previous(); // max address from last time
862 if (prev >= mr.end()) { // nothing to do
863 return;
864 }
865 // This assert will not work when we go from cms space to perm
866 // space, and use same closure. Easy fix deferred for later. XXX YSR
867 // assert(prev == NULL || contains(prev), "Should be within space");
868
869 bool last_was_obj_array = false;
870 HeapWord *blk_start_addr, *region_start_addr;
871 if (prev > mr.start()) {
872 region_start_addr = prev;
873 blk_start_addr = prev;
874 // The previous invocation may have pushed "prev" beyond the
875 // last allocated block yet there may be still be blocks
876 // in this region due to a particular coalescing policy.
877 // Relax the assertion so that the case where the unallocated
878 // block is maintained and "prev" is beyond the unallocated
879 // block does not cause the assertion to fire.
880 assert((BlockOffsetArrayUseUnallocatedBlock &&
881 (!is_in(prev))) ||
882 (blk_start_addr == block_start(region_start_addr)), "invariant");
883 } else {
884 region_start_addr = mr.start();
885 blk_start_addr = block_start(region_start_addr);
886 }
887 HeapWord* region_end_addr = mr.end();
888 MemRegion derived_mr(region_start_addr, region_end_addr);
889 while (blk_start_addr < region_end_addr) {
890 const size_t size = block_size(blk_start_addr);
891 if (block_is_obj(blk_start_addr)) {
892 last_was_obj_array = cl->do_object_bm(oop(blk_start_addr), derived_mr);
893 } else {
894 last_was_obj_array = false;
895 }
896 blk_start_addr += size;
897 }
898 if (!last_was_obj_array) {
899 assert((bottom() <= blk_start_addr) && (blk_start_addr <= end()),
900 "Should be within (closed) used space");
901 assert(blk_start_addr > prev, "Invariant");
902 cl->set_previous(blk_start_addr); // min address for next time
903 }
904 }
905
906
907 // Callers of this iterator beware: The closure application should
908 // be robust in the face of uninitialized objects and should (always)
909 // return a correct size so that the next addr + size below gives us a
910 // valid block boundary. [See for instance,
911 // ScanMarkedObjectsAgainCarefullyClosure::do_object_careful()
912 // in ConcurrentMarkSweepGeneration.cpp.]
913 HeapWord*
914 CompactibleFreeListSpace::object_iterate_careful_m(MemRegion mr,
915 ObjectClosureCareful* cl) {
916 assert_lock_strong(freelistLock());
917 // Can't use used_region() below because it may not necessarily
918 // be the same as [bottom(),end()); although we could
919 // use [used_region().start(),round_to(used_region().end(),CardSize)),
920 // that appears too cumbersome, so we just do the simpler check
921 // in the assertion below.
922 assert(!mr.is_empty() && MemRegion(bottom(),end()).contains(mr),
923 "mr should be non-empty and within used space");
924 HeapWord *addr, *end;
925 size_t size;
926 for (addr = block_start_careful(mr.start()), end = mr.end();
927 addr < end; addr += size) {
928 FreeChunk* fc = (FreeChunk*)addr;
929 if (fc->is_free()) {
930 // Since we hold the free list lock, which protects direct
931 // allocation in this generation by mutators, a free object
932 // will remain free throughout this iteration code.
933 size = fc->size();
934 } else {
935 // Note that the object need not necessarily be initialized,
936 // because (for instance) the free list lock does NOT protect
937 // object initialization. The closure application below must
938 // therefore be correct in the face of uninitialized objects.
939 size = cl->do_object_careful_m(oop(addr), mr);
940 if (size == 0) {
941 // An unparsable object found. Signal early termination.
942 return addr;
943 }
944 }
945 }
946 return NULL;
947 }
948
949
950 HeapWord* CompactibleFreeListSpace::block_start_const(const void* p) const {
951 NOT_PRODUCT(verify_objects_initialized());
952 return _bt.block_start(p);
953 }
954
955 HeapWord* CompactibleFreeListSpace::block_start_careful(const void* p) const {
956 return _bt.block_start_careful(p);
957 }
958
959 size_t CompactibleFreeListSpace::block_size(const HeapWord* p) const {
960 NOT_PRODUCT(verify_objects_initialized());
961 // This must be volatile, or else there is a danger that the compiler
962 // will compile the code below into a sometimes-infinite loop, by keeping
963 // the value read the first time in a register.
964 while (true) {
965 // We must do this until we get a consistent view of the object.
966 if (FreeChunk::indicatesFreeChunk(p)) {
967 volatile FreeChunk* fc = (volatile FreeChunk*)p;
968 size_t res = fc->size();
969
970 // Bugfix for systems with weak memory model (PPC64/IA64). The
971 // block's free bit was set and we have read the size of the
972 // block. Acquire and check the free bit again. If the block is
973 // still free, the read size is correct.
974 OrderAccess::acquire();
975
976 // If the object is still a free chunk, return the size, else it
977 // has been allocated so try again.
978 if (FreeChunk::indicatesFreeChunk(p)) {
979 assert(res != 0, "Block size should not be 0");
980 return res;
981 }
982 } else {
983 // must read from what 'p' points to in each loop.
984 Klass* k = ((volatile oopDesc*)p)->klass_or_null();
985 if (k != NULL) {
986 assert(k->is_klass(), "Should really be klass oop.");
987 oop o = (oop)p;
988 assert(o->is_oop(true /* ignore mark word */), "Should be an oop.");
989
990 // Bugfix for systems with weak memory model (PPC64/IA64).
991 // The object o may be an array. Acquire to make sure that the array
992 // size (third word) is consistent.
993 OrderAccess::acquire();
994
995 size_t res = o->size_given_klass(k);
996 res = adjustObjectSize(res);
997 assert(res != 0, "Block size should not be 0");
998 return res;
999 }
1000 }
1001 }
1002 }
1003
1004 // TODO: Now that is_parsable is gone, we should combine these two functions.
1005 // A variant of the above that uses the Printezis bits for
1006 // unparsable but allocated objects. This avoids any possible
1007 // stalls waiting for mutators to initialize objects, and is
1008 // thus potentially faster than the variant above. However,
1009 // this variant may return a zero size for a block that is
1010 // under mutation and for which a consistent size cannot be
1011 // inferred without stalling; see CMSCollector::block_size_if_printezis_bits().
1012 size_t CompactibleFreeListSpace::block_size_no_stall(HeapWord* p,
1013 const CMSCollector* c)
1014 const {
1015 assert(MemRegion(bottom(), end()).contains(p), "p not in space");
1016 // This must be volatile, or else there is a danger that the compiler
1017 // will compile the code below into a sometimes-infinite loop, by keeping
1018 // the value read the first time in a register.
1019 DEBUG_ONLY(uint loops = 0;)
1020 while (true) {
1021 // We must do this until we get a consistent view of the object.
1022 if (FreeChunk::indicatesFreeChunk(p)) {
1023 volatile FreeChunk* fc = (volatile FreeChunk*)p;
1024 size_t res = fc->size();
1025
1026 // Bugfix for systems with weak memory model (PPC64/IA64). The
1027 // free bit of the block was set and we have read the size of
1028 // the block. Acquire and check the free bit again. If the
1029 // block is still free, the read size is correct.
1030 OrderAccess::acquire();
1031
1032 if (FreeChunk::indicatesFreeChunk(p)) {
1033 assert(res != 0, "Block size should not be 0");
1034 assert(loops == 0, "Should be 0");
1035 return res;
1036 }
1037 } else {
1038 // must read from what 'p' points to in each loop.
1039 Klass* k = ((volatile oopDesc*)p)->klass_or_null();
1040 // We trust the size of any object that has a non-NULL
1041 // klass and (for those in the perm gen) is parsable
1042 // -- irrespective of its conc_safe-ty.
1043 if (k != NULL) {
1044 assert(k->is_klass(), "Should really be klass oop.");
1045 oop o = (oop)p;
1046 assert(o->is_oop(), "Should be an oop");
1047
1048 // Bugfix for systems with weak memory model (PPC64/IA64).
1049 // The object o may be an array. Acquire to make sure that the array
1050 // size (third word) is consistent.
1051 OrderAccess::acquire();
1052
1053 size_t res = o->size_given_klass(k);
1054 res = adjustObjectSize(res);
1055 assert(res != 0, "Block size should not be 0");
1056 return res;
1057 } else {
1058 // May return 0 if P-bits not present.
1059 return c->block_size_if_printezis_bits(p);
1060 }
1061 }
1062 assert(loops == 0, "Can loop at most once");
1063 DEBUG_ONLY(loops++;)
1064 }
1065 }
1066
1067 size_t CompactibleFreeListSpace::block_size_nopar(const HeapWord* p) const {
1068 NOT_PRODUCT(verify_objects_initialized());
1069 assert(MemRegion(bottom(), end()).contains(p), "p not in space");
1070 FreeChunk* fc = (FreeChunk*)p;
1071 if (fc->is_free()) {
1072 return fc->size();
1073 } else {
1074 // Ignore mark word because this may be a recently promoted
1075 // object whose mark word is used to chain together grey
1076 // objects (the last one would have a null value).
1077 assert(oop(p)->is_oop(true), "Should be an oop");
1078 return adjustObjectSize(oop(p)->size());
1079 }
1080 }
1081
1082 // This implementation assumes that the property of "being an object" is
1083 // stable. But being a free chunk may not be (because of parallel
1084 // promotion.)
1085 bool CompactibleFreeListSpace::block_is_obj(const HeapWord* p) const {
1086 FreeChunk* fc = (FreeChunk*)p;
1087 assert(is_in_reserved(p), "Should be in space");
1088 // When doing a mark-sweep-compact of the CMS generation, this
1089 // assertion may fail because prepare_for_compaction() uses
1090 // space that is garbage to maintain information on ranges of
1091 // live objects so that these live ranges can be moved as a whole.
1092 // Comment out this assertion until that problem can be solved
1093 // (i.e., that the block start calculation may look at objects
1094 // at address below "p" in finding the object that contains "p"
1095 // and those objects (if garbage) may have been modified to hold
1096 // live range information.
1097 // assert(CollectedHeap::use_parallel_gc_threads() || _bt.block_start(p) == p,
1098 // "Should be a block boundary");
1099 if (FreeChunk::indicatesFreeChunk(p)) return false;
1100 Klass* k = oop(p)->klass_or_null();
1101 if (k != NULL) {
1102 // Ignore mark word because it may have been used to
1103 // chain together promoted objects (the last one
1104 // would have a null value).
1105 assert(oop(p)->is_oop(true), "Should be an oop");
1106 return true;
1107 } else {
1108 return false; // Was not an object at the start of collection.
1109 }
1110 }
1111
1112 // Check if the object is alive. This fact is checked either by consulting
1113 // the main marking bitmap in the sweeping phase or, if it's a permanent
1114 // generation and we're not in the sweeping phase, by checking the
1115 // perm_gen_verify_bit_map where we store the "deadness" information if
1116 // we did not sweep the perm gen in the most recent previous GC cycle.
1117 bool CompactibleFreeListSpace::obj_is_alive(const HeapWord* p) const {
1118 assert(SafepointSynchronize::is_at_safepoint() || !is_init_completed(),
1119 "Else races are possible");
1120 assert(block_is_obj(p), "The address should point to an object");
1121
1122 // If we're sweeping, we use object liveness information from the main bit map
1123 // for both perm gen and old gen.
1124 // We don't need to lock the bitmap (live_map or dead_map below), because
1125 // EITHER we are in the middle of the sweeping phase, and the
1126 // main marking bit map (live_map below) is locked,
1127 // OR we're in other phases and perm_gen_verify_bit_map (dead_map below)
1128 // is stable, because it's mutated only in the sweeping phase.
1129 // NOTE: This method is also used by jmap where, if class unloading is
1130 // off, the results can return "false" for legitimate perm objects,
1131 // when we are not in the midst of a sweeping phase, which can result
1132 // in jmap not reporting certain perm gen objects. This will be moot
1133 // if/when the perm gen goes away in the future.
1134 if (_collector->abstract_state() == CMSCollector::Sweeping) {
1135 CMSBitMap* live_map = _collector->markBitMap();
1136 return live_map->par_isMarked((HeapWord*) p);
1137 }
1138 return true;
1139 }
1140
1141 bool CompactibleFreeListSpace::block_is_obj_nopar(const HeapWord* p) const {
1142 FreeChunk* fc = (FreeChunk*)p;
1143 assert(is_in_reserved(p), "Should be in space");
1144 assert(_bt.block_start(p) == p, "Should be a block boundary");
1145 if (!fc->is_free()) {
1146 // Ignore mark word because it may have been used to
1147 // chain together promoted objects (the last one
1148 // would have a null value).
1149 assert(oop(p)->is_oop(true), "Should be an oop");
1150 return true;
1151 }
1152 return false;
1153 }
1154
1155 // "MT-safe but not guaranteed MT-precise" (TM); you may get an
1156 // approximate answer if you don't hold the freelistlock when you call this.
1157 size_t CompactibleFreeListSpace::totalSizeInIndexedFreeLists() const {
1158 size_t size = 0;
1159 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
1160 debug_only(
1161 // We may be calling here without the lock in which case we
1162 // won't do this modest sanity check.
1163 if (freelistLock()->owned_by_self()) {
1164 size_t total_list_size = 0;
1165 for (FreeChunk* fc = _indexedFreeList[i].head(); fc != NULL;
1166 fc = fc->next()) {
1167 total_list_size += i;
1168 }
1169 assert(total_list_size == i * _indexedFreeList[i].count(),
1170 "Count in list is incorrect");
1171 }
1172 )
1173 size += i * _indexedFreeList[i].count();
1174 }
1175 return size;
1176 }
1177
1178 HeapWord* CompactibleFreeListSpace::par_allocate(size_t size) {
1179 MutexLockerEx x(freelistLock(), Mutex::_no_safepoint_check_flag);
1180 return allocate(size);
1181 }
1182
1183 HeapWord*
1184 CompactibleFreeListSpace::getChunkFromSmallLinearAllocBlockRemainder(size_t size) {
1185 return getChunkFromLinearAllocBlockRemainder(&_smallLinearAllocBlock, size);
1186 }
1187
1188 HeapWord* CompactibleFreeListSpace::allocate(size_t size) {
1189 assert_lock_strong(freelistLock());
1190 HeapWord* res = NULL;
1191 assert(size == adjustObjectSize(size),
1192 "use adjustObjectSize() before calling into allocate()");
1193
1194 if (_adaptive_freelists) {
1195 res = allocate_adaptive_freelists(size);
1196 } else { // non-adaptive free lists
1197 res = allocate_non_adaptive_freelists(size);
1198 }
1199
1200 if (res != NULL) {
1201 // check that res does lie in this space!
1202 assert(is_in_reserved(res), "Not in this space!");
1203 assert(is_aligned((void*)res), "alignment check");
1204
1205 FreeChunk* fc = (FreeChunk*)res;
1206 fc->markNotFree();
1207 assert(!fc->is_free(), "shouldn't be marked free");
1208 assert(oop(fc)->klass_or_null() == NULL, "should look uninitialized");
1209 // Verify that the block offset table shows this to
1210 // be a single block, but not one which is unallocated.
1211 _bt.verify_single_block(res, size);
1212 _bt.verify_not_unallocated(res, size);
1213 // mangle a just allocated object with a distinct pattern.
1214 debug_only(fc->mangleAllocated(size));
1215 }
1216
1217 return res;
1218 }
1219
1220 HeapWord* CompactibleFreeListSpace::allocate_non_adaptive_freelists(size_t size) {
1221 HeapWord* res = NULL;
1222 // try and use linear allocation for smaller blocks
1223 if (size < _smallLinearAllocBlock._allocation_size_limit) {
1224 // if successful, the following also adjusts block offset table
1225 res = getChunkFromSmallLinearAllocBlock(size);
1226 }
1227 // Else triage to indexed lists for smaller sizes
1228 if (res == NULL) {
1229 if (size < SmallForDictionary) {
1230 res = (HeapWord*) getChunkFromIndexedFreeList(size);
1231 } else {
1232 // else get it from the big dictionary; if even this doesn't
1233 // work we are out of luck.
1234 res = (HeapWord*)getChunkFromDictionaryExact(size);
1235 }
1236 }
1237
1238 return res;
1239 }
1240
1241 HeapWord* CompactibleFreeListSpace::allocate_adaptive_freelists(size_t size) {
1242 assert_lock_strong(freelistLock());
1243 HeapWord* res = NULL;
1244 assert(size == adjustObjectSize(size),
1245 "use adjustObjectSize() before calling into allocate()");
1246
1247 // Strategy
1248 // if small
1249 // exact size from small object indexed list if small
1250 // small or large linear allocation block (linAB) as appropriate
1251 // take from lists of greater sized chunks
1252 // else
1253 // dictionary
1254 // small or large linear allocation block if it has the space
1255 // Try allocating exact size from indexTable first
1256 if (size < IndexSetSize) {
1257 res = (HeapWord*) getChunkFromIndexedFreeList(size);
1258 if(res != NULL) {
1259 assert(res != (HeapWord*)_indexedFreeList[size].head(),
1260 "Not removed from free list");
1261 // no block offset table adjustment is necessary on blocks in
1262 // the indexed lists.
1263
1264 // Try allocating from the small LinAB
1265 } else if (size < _smallLinearAllocBlock._allocation_size_limit &&
1266 (res = getChunkFromSmallLinearAllocBlock(size)) != NULL) {
1267 // if successful, the above also adjusts block offset table
1268 // Note that this call will refill the LinAB to
1269 // satisfy the request. This is different that
1270 // evm.
1271 // Don't record chunk off a LinAB? smallSplitBirth(size);
1272 } else {
1273 // Raid the exact free lists larger than size, even if they are not
1274 // overpopulated.
1275 res = (HeapWord*) getChunkFromGreater(size);
1276 }
1277 } else {
1278 // Big objects get allocated directly from the dictionary.
1279 res = (HeapWord*) getChunkFromDictionaryExact(size);
1280 if (res == NULL) {
1281 // Try hard not to fail since an allocation failure will likely
1282 // trigger a synchronous GC. Try to get the space from the
1283 // allocation blocks.
1284 res = getChunkFromSmallLinearAllocBlockRemainder(size);
1285 }
1286 }
1287
1288 return res;
1289 }
1290
1291 // A worst-case estimate of the space required (in HeapWords) to expand the heap
1292 // when promoting obj.
1293 size_t CompactibleFreeListSpace::expansionSpaceRequired(size_t obj_size) const {
1294 // Depending on the object size, expansion may require refilling either a
1295 // bigLAB or a smallLAB plus refilling a PromotionInfo object. MinChunkSize
1296 // is added because the dictionary may over-allocate to avoid fragmentation.
1297 size_t space = obj_size;
1298 if (!_adaptive_freelists) {
1299 space = MAX2(space, _smallLinearAllocBlock._refillSize);
1300 }
1301 space += _promoInfo.refillSize() + 2 * MinChunkSize;
1302 return space;
1303 }
1304
1305 FreeChunk* CompactibleFreeListSpace::getChunkFromGreater(size_t numWords) {
1306 FreeChunk* ret;
1307
1308 assert(numWords >= MinChunkSize, "Size is less than minimum");
1309 assert(linearAllocationWouldFail() || bestFitFirst(),
1310 "Should not be here");
1311
1312 size_t i;
1313 size_t currSize = numWords + MinChunkSize;
1314 assert(currSize % MinObjAlignment == 0, "currSize should be aligned");
1315 for (i = currSize; i < IndexSetSize; i += IndexSetStride) {
1316 AdaptiveFreeList<FreeChunk>* fl = &_indexedFreeList[i];
1317 if (fl->head()) {
1318 ret = getFromListGreater(fl, numWords);
1319 assert(ret == NULL || ret->is_free(), "Should be returning a free chunk");
1320 return ret;
1321 }
1322 }
1323
1324 currSize = MAX2((size_t)SmallForDictionary,
1325 (size_t)(numWords + MinChunkSize));
1326
1327 /* Try to get a chunk that satisfies request, while avoiding
1328 fragmentation that can't be handled. */
1329 {
1330 ret = dictionary()->get_chunk(currSize);
1331 if (ret != NULL) {
1332 assert(ret->size() - numWords >= MinChunkSize,
1333 "Chunk is too small");
1334 _bt.allocated((HeapWord*)ret, ret->size());
1335 /* Carve returned chunk. */
1336 (void) splitChunkAndReturnRemainder(ret, numWords);
1337 /* Label this as no longer a free chunk. */
1338 assert(ret->is_free(), "This chunk should be free");
1339 ret->link_prev(NULL);
1340 }
1341 assert(ret == NULL || ret->is_free(), "Should be returning a free chunk");
1342 return ret;
1343 }
1344 ShouldNotReachHere();
1345 }
1346
1347 bool CompactibleFreeListSpace::verifyChunkInIndexedFreeLists(FreeChunk* fc) const {
1348 assert(fc->size() < IndexSetSize, "Size of chunk is too large");
1349 return _indexedFreeList[fc->size()].verify_chunk_in_free_list(fc);
1350 }
1351
1352 bool CompactibleFreeListSpace::verify_chunk_is_linear_alloc_block(FreeChunk* fc) const {
1353 assert((_smallLinearAllocBlock._ptr != (HeapWord*)fc) ||
1354 (_smallLinearAllocBlock._word_size == fc->size()),
1355 "Linear allocation block shows incorrect size");
1356 return ((_smallLinearAllocBlock._ptr == (HeapWord*)fc) &&
1357 (_smallLinearAllocBlock._word_size == fc->size()));
1358 }
1359
1360 // Check if the purported free chunk is present either as a linear
1361 // allocation block, the size-indexed table of (smaller) free blocks,
1362 // or the larger free blocks kept in the binary tree dictionary.
1363 bool CompactibleFreeListSpace::verify_chunk_in_free_list(FreeChunk* fc) const {
1364 if (verify_chunk_is_linear_alloc_block(fc)) {
1365 return true;
1366 } else if (fc->size() < IndexSetSize) {
1367 return verifyChunkInIndexedFreeLists(fc);
1368 } else {
1369 return dictionary()->verify_chunk_in_free_list(fc);
1370 }
1371 }
1372
1373 #ifndef PRODUCT
1374 void CompactibleFreeListSpace::assert_locked() const {
1375 CMSLockVerifier::assert_locked(freelistLock(), parDictionaryAllocLock());
1376 }
1377
1378 void CompactibleFreeListSpace::assert_locked(const Mutex* lock) const {
1379 CMSLockVerifier::assert_locked(lock);
1380 }
1381 #endif
1382
1383 FreeChunk* CompactibleFreeListSpace::allocateScratch(size_t size) {
1384 // In the parallel case, the main thread holds the free list lock
1385 // on behalf the parallel threads.
1386 FreeChunk* fc;
1387 {
1388 // If GC is parallel, this might be called by several threads.
1389 // This should be rare enough that the locking overhead won't affect
1390 // the sequential code.
1391 MutexLockerEx x(parDictionaryAllocLock(),
1392 Mutex::_no_safepoint_check_flag);
1393 fc = getChunkFromDictionary(size);
1394 }
1395 if (fc != NULL) {
1396 fc->dontCoalesce();
1397 assert(fc->is_free(), "Should be free, but not coalescable");
1398 // Verify that the block offset table shows this to
1399 // be a single block, but not one which is unallocated.
1400 _bt.verify_single_block((HeapWord*)fc, fc->size());
1401 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
1402 }
1403 return fc;
1404 }
1405
1406 oop CompactibleFreeListSpace::promote(oop obj, size_t obj_size) {
1407 assert(obj_size == (size_t)obj->size(), "bad obj_size passed in");
1408 assert_locked();
1409
1410 // if we are tracking promotions, then first ensure space for
1411 // promotion (including spooling space for saving header if necessary).
1412 // then allocate and copy, then track promoted info if needed.
1413 // When tracking (see PromotionInfo::track()), the mark word may
1414 // be displaced and in this case restoration of the mark word
1415 // occurs in the (oop_since_save_marks_)iterate phase.
1416 if (_promoInfo.tracking() && !_promoInfo.ensure_spooling_space()) {
1417 return NULL;
1418 }
1419 // Call the allocate(size_t, bool) form directly to avoid the
1420 // additional call through the allocate(size_t) form. Having
1421 // the compile inline the call is problematic because allocate(size_t)
1422 // is a virtual method.
1423 HeapWord* res = allocate(adjustObjectSize(obj_size));
1424 if (res != NULL) {
1425 Copy::aligned_disjoint_words((HeapWord*)obj, res, obj_size);
1426 // if we should be tracking promotions, do so.
1427 if (_promoInfo.tracking()) {
1428 _promoInfo.track((PromotedObject*)res);
1429 }
1430 }
1431 return oop(res);
1432 }
1433
1434 HeapWord*
1435 CompactibleFreeListSpace::getChunkFromSmallLinearAllocBlock(size_t size) {
1436 assert_locked();
1437 assert(size >= MinChunkSize, "minimum chunk size");
1438 assert(size < _smallLinearAllocBlock._allocation_size_limit,
1439 "maximum from smallLinearAllocBlock");
1440 return getChunkFromLinearAllocBlock(&_smallLinearAllocBlock, size);
1441 }
1442
1443 HeapWord*
1444 CompactibleFreeListSpace::getChunkFromLinearAllocBlock(LinearAllocBlock *blk,
1445 size_t size) {
1446 assert_locked();
1447 assert(size >= MinChunkSize, "too small");
1448 HeapWord* res = NULL;
1449 // Try to do linear allocation from blk, making sure that
1450 if (blk->_word_size == 0) {
1451 // We have probably been unable to fill this either in the prologue or
1452 // when it was exhausted at the last linear allocation. Bail out until
1453 // next time.
1454 assert(blk->_ptr == NULL, "consistency check");
1455 return NULL;
1456 }
1457 assert(blk->_word_size != 0 && blk->_ptr != NULL, "consistency check");
1458 res = getChunkFromLinearAllocBlockRemainder(blk, size);
1459 if (res != NULL) return res;
1460
1461 // about to exhaust this linear allocation block
1462 if (blk->_word_size == size) { // exactly satisfied
1463 res = blk->_ptr;
1464 _bt.allocated(res, blk->_word_size);
1465 } else if (size + MinChunkSize <= blk->_refillSize) {
1466 size_t sz = blk->_word_size;
1467 // Update _unallocated_block if the size is such that chunk would be
1468 // returned to the indexed free list. All other chunks in the indexed
1469 // free lists are allocated from the dictionary so that _unallocated_block
1470 // has already been adjusted for them. Do it here so that the cost
1471 // for all chunks added back to the indexed free lists.
1472 if (sz < SmallForDictionary) {
1473 _bt.allocated(blk->_ptr, sz);
1474 }
1475 // Return the chunk that isn't big enough, and then refill below.
1476 addChunkToFreeLists(blk->_ptr, sz);
1477 split_birth(sz);
1478 // Don't keep statistics on adding back chunk from a LinAB.
1479 } else {
1480 // A refilled block would not satisfy the request.
1481 return NULL;
1482 }
1483
1484 blk->_ptr = NULL; blk->_word_size = 0;
1485 refillLinearAllocBlock(blk);
1486 assert(blk->_ptr == NULL || blk->_word_size >= size + MinChunkSize,
1487 "block was replenished");
1488 if (res != NULL) {
1489 split_birth(size);
1490 repairLinearAllocBlock(blk);
1491 } else if (blk->_ptr != NULL) {
1492 res = blk->_ptr;
1493 size_t blk_size = blk->_word_size;
1494 blk->_word_size -= size;
1495 blk->_ptr += size;
1496 split_birth(size);
1497 repairLinearAllocBlock(blk);
1498 // Update BOT last so that other (parallel) GC threads see a consistent
1499 // view of the BOT and free blocks.
1500 // Above must occur before BOT is updated below.
1501 OrderAccess::storestore();
1502 _bt.split_block(res, blk_size, size); // adjust block offset table
1503 }
1504 return res;
1505 }
1506
1507 HeapWord* CompactibleFreeListSpace::getChunkFromLinearAllocBlockRemainder(
1508 LinearAllocBlock* blk,
1509 size_t size) {
1510 assert_locked();
1511 assert(size >= MinChunkSize, "too small");
1512
1513 HeapWord* res = NULL;
1514 // This is the common case. Keep it simple.
1515 if (blk->_word_size >= size + MinChunkSize) {
1516 assert(blk->_ptr != NULL, "consistency check");
1517 res = blk->_ptr;
1518 // Note that the BOT is up-to-date for the linAB before allocation. It
1519 // indicates the start of the linAB. The split_block() updates the
1520 // BOT for the linAB after the allocation (indicates the start of the
1521 // next chunk to be allocated).
1522 size_t blk_size = blk->_word_size;
1523 blk->_word_size -= size;
1524 blk->_ptr += size;
1525 split_birth(size);
1526 repairLinearAllocBlock(blk);
1527 // Update BOT last so that other (parallel) GC threads see a consistent
1528 // view of the BOT and free blocks.
1529 // Above must occur before BOT is updated below.
1530 OrderAccess::storestore();
1531 _bt.split_block(res, blk_size, size); // adjust block offset table
1532 _bt.allocated(res, size);
1533 }
1534 return res;
1535 }
1536
1537 FreeChunk*
1538 CompactibleFreeListSpace::getChunkFromIndexedFreeList(size_t size) {
1539 assert_locked();
1540 assert(size < SmallForDictionary, "just checking");
1541 FreeChunk* res;
1542 res = _indexedFreeList[size].get_chunk_at_head();
1543 if (res == NULL) {
1544 res = getChunkFromIndexedFreeListHelper(size);
1545 }
1546 _bt.verify_not_unallocated((HeapWord*) res, size);
1547 assert(res == NULL || res->size() == size, "Incorrect block size");
1548 return res;
1549 }
1550
1551 FreeChunk*
1552 CompactibleFreeListSpace::getChunkFromIndexedFreeListHelper(size_t size,
1553 bool replenish) {
1554 assert_locked();
1555 FreeChunk* fc = NULL;
1556 if (size < SmallForDictionary) {
1557 assert(_indexedFreeList[size].head() == NULL ||
1558 _indexedFreeList[size].surplus() <= 0,
1559 "List for this size should be empty or under populated");
1560 // Try best fit in exact lists before replenishing the list
1561 if (!bestFitFirst() || (fc = bestFitSmall(size)) == NULL) {
1562 // Replenish list.
1563 //
1564 // Things tried that failed.
1565 // Tried allocating out of the two LinAB's first before
1566 // replenishing lists.
1567 // Tried small linAB of size 256 (size in indexed list)
1568 // and replenishing indexed lists from the small linAB.
1569 //
1570 FreeChunk* newFc = NULL;
1571 const size_t replenish_size = CMSIndexedFreeListReplenish * size;
1572 if (replenish_size < SmallForDictionary) {
1573 // Do not replenish from an underpopulated size.
1574 if (_indexedFreeList[replenish_size].surplus() > 0 &&
1575 _indexedFreeList[replenish_size].head() != NULL) {
1576 newFc = _indexedFreeList[replenish_size].get_chunk_at_head();
1577 } else if (bestFitFirst()) {
1578 newFc = bestFitSmall(replenish_size);
1579 }
1580 }
1581 if (newFc == NULL && replenish_size > size) {
1582 assert(CMSIndexedFreeListReplenish > 1, "ctl pt invariant");
1583 newFc = getChunkFromIndexedFreeListHelper(replenish_size, false);
1584 }
1585 // Note: The stats update re split-death of block obtained above
1586 // will be recorded below precisely when we know we are going to
1587 // be actually splitting it into more than one pieces below.
1588 if (newFc != NULL) {
1589 if (replenish || CMSReplenishIntermediate) {
1590 // Replenish this list and return one block to caller.
1591 size_t i;
1592 FreeChunk *curFc, *nextFc;
1593 size_t num_blk = newFc->size() / size;
1594 assert(num_blk >= 1, "Smaller than requested?");
1595 assert(newFc->size() % size == 0, "Should be integral multiple of request");
1596 if (num_blk > 1) {
1597 // we are sure we will be splitting the block just obtained
1598 // into multiple pieces; record the split-death of the original
1599 splitDeath(replenish_size);
1600 }
1601 // carve up and link blocks 0, ..., num_blk - 2
1602 // The last chunk is not added to the lists but is returned as the
1603 // free chunk.
1604 for (curFc = newFc, nextFc = (FreeChunk*)((HeapWord*)curFc + size),
1605 i = 0;
1606 i < (num_blk - 1);
1607 curFc = nextFc, nextFc = (FreeChunk*)((HeapWord*)nextFc + size),
1608 i++) {
1609 curFc->set_size(size);
1610 // Don't record this as a return in order to try and
1611 // determine the "returns" from a GC.
1612 _bt.verify_not_unallocated((HeapWord*) fc, size);
1613 _indexedFreeList[size].return_chunk_at_tail(curFc, false);
1614 _bt.mark_block((HeapWord*)curFc, size);
1615 split_birth(size);
1616 // Don't record the initial population of the indexed list
1617 // as a split birth.
1618 }
1619
1620 // check that the arithmetic was OK above
1621 assert((HeapWord*)nextFc == (HeapWord*)newFc + num_blk*size,
1622 "inconsistency in carving newFc");
1623 curFc->set_size(size);
1624 _bt.mark_block((HeapWord*)curFc, size);
1625 split_birth(size);
1626 fc = curFc;
1627 } else {
1628 // Return entire block to caller
1629 fc = newFc;
1630 }
1631 }
1632 }
1633 } else {
1634 // Get a free chunk from the free chunk dictionary to be returned to
1635 // replenish the indexed free list.
1636 fc = getChunkFromDictionaryExact(size);
1637 }
1638 // assert(fc == NULL || fc->is_free(), "Should be returning a free chunk");
1639 return fc;
1640 }
1641
1642 FreeChunk*
1643 CompactibleFreeListSpace::getChunkFromDictionary(size_t size) {
1644 assert_locked();
1645 FreeChunk* fc = _dictionary->get_chunk(size,
1646 FreeBlockDictionary<FreeChunk>::atLeast);
1647 if (fc == NULL) {
1648 return NULL;
1649 }
1650 _bt.allocated((HeapWord*)fc, fc->size());
1651 if (fc->size() >= size + MinChunkSize) {
1652 fc = splitChunkAndReturnRemainder(fc, size);
1653 }
1654 assert(fc->size() >= size, "chunk too small");
1655 assert(fc->size() < size + MinChunkSize, "chunk too big");
1656 _bt.verify_single_block((HeapWord*)fc, fc->size());
1657 return fc;
1658 }
1659
1660 FreeChunk*
1661 CompactibleFreeListSpace::getChunkFromDictionaryExact(size_t size) {
1662 assert_locked();
1663 FreeChunk* fc = _dictionary->get_chunk(size,
1664 FreeBlockDictionary<FreeChunk>::atLeast);
1665 if (fc == NULL) {
1666 return fc;
1667 }
1668 _bt.allocated((HeapWord*)fc, fc->size());
1669 if (fc->size() == size) {
1670 _bt.verify_single_block((HeapWord*)fc, size);
1671 return fc;
1672 }
1673 assert(fc->size() > size, "get_chunk() guarantee");
1674 if (fc->size() < size + MinChunkSize) {
1675 // Return the chunk to the dictionary and go get a bigger one.
1676 returnChunkToDictionary(fc);
1677 fc = _dictionary->get_chunk(size + MinChunkSize,
1678 FreeBlockDictionary<FreeChunk>::atLeast);
1679 if (fc == NULL) {
1680 return NULL;
1681 }
1682 _bt.allocated((HeapWord*)fc, fc->size());
1683 }
1684 assert(fc->size() >= size + MinChunkSize, "tautology");
1685 fc = splitChunkAndReturnRemainder(fc, size);
1686 assert(fc->size() == size, "chunk is wrong size");
1687 _bt.verify_single_block((HeapWord*)fc, size);
1688 return fc;
1689 }
1690
1691 void
1692 CompactibleFreeListSpace::returnChunkToDictionary(FreeChunk* chunk) {
1693 assert_locked();
1694
1695 size_t size = chunk->size();
1696 _bt.verify_single_block((HeapWord*)chunk, size);
1697 // adjust _unallocated_block downward, as necessary
1698 _bt.freed((HeapWord*)chunk, size);
1699 _dictionary->return_chunk(chunk);
1700 #ifndef PRODUCT
1701 if (CMSCollector::abstract_state() != CMSCollector::Sweeping) {
1702 TreeChunk<FreeChunk, AdaptiveFreeList<FreeChunk> >* tc = TreeChunk<FreeChunk, AdaptiveFreeList<FreeChunk> >::as_TreeChunk(chunk);
1703 TreeList<FreeChunk, AdaptiveFreeList<FreeChunk> >* tl = tc->list();
1704 tl->verify_stats();
1705 }
1706 #endif // PRODUCT
1707 }
1708
1709 void
1710 CompactibleFreeListSpace::returnChunkToFreeList(FreeChunk* fc) {
1711 assert_locked();
1712 size_t size = fc->size();
1713 _bt.verify_single_block((HeapWord*) fc, size);
1714 _bt.verify_not_unallocated((HeapWord*) fc, size);
1715 if (_adaptive_freelists) {
1716 _indexedFreeList[size].return_chunk_at_tail(fc);
1717 } else {
1718 _indexedFreeList[size].return_chunk_at_head(fc);
1719 }
1720 #ifndef PRODUCT
1721 if (CMSCollector::abstract_state() != CMSCollector::Sweeping) {
1722 _indexedFreeList[size].verify_stats();
1723 }
1724 #endif // PRODUCT
1725 }
1726
1727 // Add chunk to end of last block -- if it's the largest
1728 // block -- and update BOT and census data. We would
1729 // of course have preferred to coalesce it with the
1730 // last block, but it's currently less expensive to find the
1731 // largest block than it is to find the last.
1732 void
1733 CompactibleFreeListSpace::addChunkToFreeListsAtEndRecordingStats(
1734 HeapWord* chunk, size_t size) {
1735 // check that the chunk does lie in this space!
1736 assert(chunk != NULL && is_in_reserved(chunk), "Not in this space!");
1737 // One of the parallel gc task threads may be here
1738 // whilst others are allocating.
1739 Mutex* lock = &_parDictionaryAllocLock;
1740 FreeChunk* ec;
1741 {
1742 MutexLockerEx x(lock, Mutex::_no_safepoint_check_flag);
1743 ec = dictionary()->find_largest_dict(); // get largest block
1744 if (ec != NULL && ec->end() == (uintptr_t*) chunk) {
1745 // It's a coterminal block - we can coalesce.
1746 size_t old_size = ec->size();
1747 coalDeath(old_size);
1748 removeChunkFromDictionary(ec);
1749 size += old_size;
1750 } else {
1751 ec = (FreeChunk*)chunk;
1752 }
1753 }
1754 ec->set_size(size);
1755 debug_only(ec->mangleFreed(size));
1756 if (size < SmallForDictionary) {
1757 lock = _indexedFreeListParLocks[size];
1758 }
1759 MutexLockerEx x(lock, Mutex::_no_safepoint_check_flag);
1760 addChunkAndRepairOffsetTable((HeapWord*)ec, size, true);
1761 // record the birth under the lock since the recording involves
1762 // manipulation of the list on which the chunk lives and
1763 // if the chunk is allocated and is the last on the list,
1764 // the list can go away.
1765 coalBirth(size);
1766 }
1767
1768 void
1769 CompactibleFreeListSpace::addChunkToFreeLists(HeapWord* chunk,
1770 size_t size) {
1771 // check that the chunk does lie in this space!
1772 assert(chunk != NULL && is_in_reserved(chunk), "Not in this space!");
1773 assert_locked();
1774 _bt.verify_single_block(chunk, size);
1775
1776 FreeChunk* fc = (FreeChunk*) chunk;
1777 fc->set_size(size);
1778 debug_only(fc->mangleFreed(size));
1779 if (size < SmallForDictionary) {
1780 returnChunkToFreeList(fc);
1781 } else {
1782 returnChunkToDictionary(fc);
1783 }
1784 }
1785
1786 void
1787 CompactibleFreeListSpace::addChunkAndRepairOffsetTable(HeapWord* chunk,
1788 size_t size, bool coalesced) {
1789 assert_locked();
1790 assert(chunk != NULL, "null chunk");
1791 if (coalesced) {
1792 // repair BOT
1793 _bt.single_block(chunk, size);
1794 }
1795 addChunkToFreeLists(chunk, size);
1796 }
1797
1798 // We _must_ find the purported chunk on our free lists;
1799 // we assert if we don't.
1800 void
1801 CompactibleFreeListSpace::removeFreeChunkFromFreeLists(FreeChunk* fc) {
1802 size_t size = fc->size();
1803 assert_locked();
1804 debug_only(verifyFreeLists());
1805 if (size < SmallForDictionary) {
1806 removeChunkFromIndexedFreeList(fc);
1807 } else {
1808 removeChunkFromDictionary(fc);
1809 }
1810 _bt.verify_single_block((HeapWord*)fc, size);
1811 debug_only(verifyFreeLists());
1812 }
1813
1814 void
1815 CompactibleFreeListSpace::removeChunkFromDictionary(FreeChunk* fc) {
1816 size_t size = fc->size();
1817 assert_locked();
1818 assert(fc != NULL, "null chunk");
1819 _bt.verify_single_block((HeapWord*)fc, size);
1820 _dictionary->remove_chunk(fc);
1821 // adjust _unallocated_block upward, as necessary
1822 _bt.allocated((HeapWord*)fc, size);
1823 }
1824
1825 void
1826 CompactibleFreeListSpace::removeChunkFromIndexedFreeList(FreeChunk* fc) {
1827 assert_locked();
1828 size_t size = fc->size();
1829 _bt.verify_single_block((HeapWord*)fc, size);
1830 NOT_PRODUCT(
1831 if (FLSVerifyIndexTable) {
1832 verifyIndexedFreeList(size);
1833 }
1834 )
1835 _indexedFreeList[size].remove_chunk(fc);
1836 NOT_PRODUCT(
1837 if (FLSVerifyIndexTable) {
1838 verifyIndexedFreeList(size);
1839 }
1840 )
1841 }
1842
1843 FreeChunk* CompactibleFreeListSpace::bestFitSmall(size_t numWords) {
1844 /* A hint is the next larger size that has a surplus.
1845 Start search at a size large enough to guarantee that
1846 the excess is >= MIN_CHUNK. */
1847 size_t start = align_object_size(numWords + MinChunkSize);
1848 if (start < IndexSetSize) {
1849 AdaptiveFreeList<FreeChunk>* it = _indexedFreeList;
1850 size_t hint = _indexedFreeList[start].hint();
1851 while (hint < IndexSetSize) {
1852 assert(hint % MinObjAlignment == 0, "hint should be aligned");
1853 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[hint];
1854 if (fl->surplus() > 0 && fl->head() != NULL) {
1855 // Found a list with surplus, reset original hint
1856 // and split out a free chunk which is returned.
1857 _indexedFreeList[start].set_hint(hint);
1858 FreeChunk* res = getFromListGreater(fl, numWords);
1859 assert(res == NULL || res->is_free(),
1860 "Should be returning a free chunk");
1861 return res;
1862 }
1863 hint = fl->hint(); /* keep looking */
1864 }
1865 /* None found. */
1866 it[start].set_hint(IndexSetSize);
1867 }
1868 return NULL;
1869 }
1870
1871 /* Requires fl->size >= numWords + MinChunkSize */
1872 FreeChunk* CompactibleFreeListSpace::getFromListGreater(AdaptiveFreeList<FreeChunk>* fl,
1873 size_t numWords) {
1874 FreeChunk *curr = fl->head();
1875 size_t oldNumWords = curr->size();
1876 assert(numWords >= MinChunkSize, "Word size is too small");
1877 assert(curr != NULL, "List is empty");
1878 assert(oldNumWords >= numWords + MinChunkSize,
1879 "Size of chunks in the list is too small");
1880
1881 fl->remove_chunk(curr);
1882 // recorded indirectly by splitChunkAndReturnRemainder -
1883 // smallSplit(oldNumWords, numWords);
1884 FreeChunk* new_chunk = splitChunkAndReturnRemainder(curr, numWords);
1885 // Does anything have to be done for the remainder in terms of
1886 // fixing the card table?
1887 assert(new_chunk == NULL || new_chunk->is_free(),
1888 "Should be returning a free chunk");
1889 return new_chunk;
1890 }
1891
1892 FreeChunk*
1893 CompactibleFreeListSpace::splitChunkAndReturnRemainder(FreeChunk* chunk,
1894 size_t new_size) {
1895 assert_locked();
1896 size_t size = chunk->size();
1897 assert(size > new_size, "Split from a smaller block?");
1898 assert(is_aligned(chunk), "alignment problem");
1899 assert(size == adjustObjectSize(size), "alignment problem");
1900 size_t rem_sz = size - new_size;
1901 assert(rem_sz == adjustObjectSize(rem_sz), "alignment problem");
1902 assert(rem_sz >= MinChunkSize, "Free chunk smaller than minimum");
1903 FreeChunk* ffc = (FreeChunk*)((HeapWord*)chunk + new_size);
1904 assert(is_aligned(ffc), "alignment problem");
1905 ffc->set_size(rem_sz);
1906 ffc->link_next(NULL);
1907 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
1908 // Above must occur before BOT is updated below.
1909 // adjust block offset table
1910 OrderAccess::storestore();
1911 assert(chunk->is_free() && ffc->is_free(), "Error");
1912 _bt.split_block((HeapWord*)chunk, chunk->size(), new_size);
1913 if (rem_sz < SmallForDictionary) {
1914 bool is_par = (SharedHeap::heap()->n_par_threads() > 0);
1915 if (is_par) _indexedFreeListParLocks[rem_sz]->lock();
1916 assert(!is_par ||
1917 (SharedHeap::heap()->n_par_threads() ==
1918 SharedHeap::heap()->workers()->active_workers()), "Mismatch");
1919 returnChunkToFreeList(ffc);
1920 split(size, rem_sz);
1921 if (is_par) _indexedFreeListParLocks[rem_sz]->unlock();
1922 } else {
1923 returnChunkToDictionary(ffc);
1924 split(size, rem_sz);
1925 }
1926 chunk->set_size(new_size);
1927 return chunk;
1928 }
1929
1930 void
1931 CompactibleFreeListSpace::sweep_completed() {
1932 // Now that space is probably plentiful, refill linear
1933 // allocation blocks as needed.
1934 refillLinearAllocBlocksIfNeeded();
1935 }
1936
1937 void
1938 CompactibleFreeListSpace::gc_prologue() {
1939 assert_locked();
1940 if (PrintFLSStatistics != 0) {
1941 gclog_or_tty->print("Before GC:\n");
1942 reportFreeListStatistics();
1943 }
1944 refillLinearAllocBlocksIfNeeded();
1945 }
1946
1947 void
1948 CompactibleFreeListSpace::gc_epilogue() {
1949 assert_locked();
1950 if (PrintGCDetails && Verbose && !_adaptive_freelists) {
1951 if (_smallLinearAllocBlock._word_size == 0)
1952 warning("CompactibleFreeListSpace(epilogue):: Linear allocation failure");
1953 }
1954 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency");
1955 _promoInfo.stopTrackingPromotions();
1956 repairLinearAllocationBlocks();
1957 // Print Space's stats
1958 if (PrintFLSStatistics != 0) {
1959 gclog_or_tty->print("After GC:\n");
1960 reportFreeListStatistics();
1961 }
1962 }
1963
1964 // Iteration support, mostly delegated from a CMS generation
1965
1966 void CompactibleFreeListSpace::save_marks() {
1967 assert(Thread::current()->is_VM_thread(),
1968 "Global variable should only be set when single-threaded");
1969 // Mark the "end" of the used space at the time of this call;
1970 // note, however, that promoted objects from this point
1971 // on are tracked in the _promoInfo below.
1972 set_saved_mark_word(unallocated_block());
1973 #ifdef ASSERT
1974 // Check the sanity of save_marks() etc.
1975 MemRegion ur = used_region();
1976 MemRegion urasm = used_region_at_save_marks();
1977 assert(ur.contains(urasm),
1978 err_msg(" Error at save_marks(): [" PTR_FORMAT "," PTR_FORMAT ")"
1979 " should contain [" PTR_FORMAT "," PTR_FORMAT ")",
1980 p2i(ur.start()), p2i(ur.end()), p2i(urasm.start()), p2i(urasm.end())));
1981 #endif
1982 // inform allocator that promotions should be tracked.
1983 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency");
1984 _promoInfo.startTrackingPromotions();
1985 }
1986
1987 bool CompactibleFreeListSpace::no_allocs_since_save_marks() {
1988 assert(_promoInfo.tracking(), "No preceding save_marks?");
1989 assert(SharedHeap::heap()->n_par_threads() == 0,
1990 "Shouldn't be called if using parallel gc.");
1991 return _promoInfo.noPromotions();
1992 }
1993
1994 #define CFLS_OOP_SINCE_SAVE_MARKS_DEFN(OopClosureType, nv_suffix) \
1995 \
1996 void CompactibleFreeListSpace:: \
1997 oop_since_save_marks_iterate##nv_suffix(OopClosureType* blk) { \
1998 assert(SharedHeap::heap()->n_par_threads() == 0, \
1999 "Shouldn't be called (yet) during parallel part of gc."); \
2000 _promoInfo.promoted_oops_iterate##nv_suffix(blk); \
2001 /* \
2002 * This also restores any displaced headers and removes the elements from \
2003 * the iteration set as they are processed, so that we have a clean slate \
2004 * at the end of the iteration. Note, thus, that if new objects are \
2005 * promoted as a result of the iteration they are iterated over as well. \
2006 */ \
2007 assert(_promoInfo.noPromotions(), "_promoInfo inconsistency"); \
2008 }
2009
2010 ALL_SINCE_SAVE_MARKS_CLOSURES(CFLS_OOP_SINCE_SAVE_MARKS_DEFN)
2011
2012 bool CompactibleFreeListSpace::linearAllocationWouldFail() const {
2013 return _smallLinearAllocBlock._word_size == 0;
2014 }
2015
2016 void CompactibleFreeListSpace::repairLinearAllocationBlocks() {
2017 // Fix up linear allocation blocks to look like free blocks
2018 repairLinearAllocBlock(&_smallLinearAllocBlock);
2019 }
2020
2021 void CompactibleFreeListSpace::repairLinearAllocBlock(LinearAllocBlock* blk) {
2022 assert_locked();
2023 if (blk->_ptr != NULL) {
2024 assert(blk->_word_size != 0 && blk->_word_size >= MinChunkSize,
2025 "Minimum block size requirement");
2026 FreeChunk* fc = (FreeChunk*)(blk->_ptr);
2027 fc->set_size(blk->_word_size);
2028 fc->link_prev(NULL); // mark as free
2029 fc->dontCoalesce();
2030 assert(fc->is_free(), "just marked it free");
2031 assert(fc->cantCoalesce(), "just marked it uncoalescable");
2032 }
2033 }
2034
2035 void CompactibleFreeListSpace::refillLinearAllocBlocksIfNeeded() {
2036 assert_locked();
2037 if (_smallLinearAllocBlock._ptr == NULL) {
2038 assert(_smallLinearAllocBlock._word_size == 0,
2039 "Size of linAB should be zero if the ptr is NULL");
2040 // Reset the linAB refill and allocation size limit.
2041 _smallLinearAllocBlock.set(0, 0, 1024*SmallForLinearAlloc, SmallForLinearAlloc);
2042 }
2043 refillLinearAllocBlockIfNeeded(&_smallLinearAllocBlock);
2044 }
2045
2046 void
2047 CompactibleFreeListSpace::refillLinearAllocBlockIfNeeded(LinearAllocBlock* blk) {
2048 assert_locked();
2049 assert((blk->_ptr == NULL && blk->_word_size == 0) ||
2050 (blk->_ptr != NULL && blk->_word_size >= MinChunkSize),
2051 "blk invariant");
2052 if (blk->_ptr == NULL) {
2053 refillLinearAllocBlock(blk);
2054 }
2055 if (PrintMiscellaneous && Verbose) {
2056 if (blk->_word_size == 0) {
2057 warning("CompactibleFreeListSpace(prologue):: Linear allocation failure");
2058 }
2059 }
2060 }
2061
2062 void
2063 CompactibleFreeListSpace::refillLinearAllocBlock(LinearAllocBlock* blk) {
2064 assert_locked();
2065 assert(blk->_word_size == 0 && blk->_ptr == NULL,
2066 "linear allocation block should be empty");
2067 FreeChunk* fc;
2068 if (blk->_refillSize < SmallForDictionary &&
2069 (fc = getChunkFromIndexedFreeList(blk->_refillSize)) != NULL) {
2070 // A linAB's strategy might be to use small sizes to reduce
2071 // fragmentation but still get the benefits of allocation from a
2072 // linAB.
2073 } else {
2074 fc = getChunkFromDictionary(blk->_refillSize);
2075 }
2076 if (fc != NULL) {
2077 blk->_ptr = (HeapWord*)fc;
2078 blk->_word_size = fc->size();
2079 fc->dontCoalesce(); // to prevent sweeper from sweeping us up
2080 }
2081 }
2082
2083 // Support for concurrent collection policy decisions.
2084 bool CompactibleFreeListSpace::should_concurrent_collect() const {
2085 // In the future we might want to add in fragmentation stats --
2086 // including erosion of the "mountain" into this decision as well.
2087 return !adaptive_freelists() && linearAllocationWouldFail();
2088 }
2089
2090 // Support for compaction
2091 void CompactibleFreeListSpace::prepare_for_compaction(CompactPoint* cp) {
2092 scan_and_forward(this, cp);
2093 // Prepare_for_compaction() uses the space between live objects
2094 // so that later phase can skip dead space quickly. So verification
2095 // of the free lists doesn't work after.
2096 }
2097
2098 void CompactibleFreeListSpace::adjust_pointers() {
2099 // In other versions of adjust_pointers(), a bail out
2100 // based on the amount of live data in the generation
2101 // (i.e., if 0, bail out) may be used.
2102 // Cannot test used() == 0 here because the free lists have already
2103 // been mangled by the compaction.
2104
2105 scan_and_adjust_pointers(this);
2106 // See note about verification in prepare_for_compaction().
2107 }
2108
2109 void CompactibleFreeListSpace::compact() {
2110 scan_and_compact(this);
2111 }
2112
2113 // Fragmentation metric = 1 - [sum of (fbs**2) / (sum of fbs)**2]
2114 // where fbs is free block sizes
2115 double CompactibleFreeListSpace::flsFrag() const {
2116 size_t itabFree = totalSizeInIndexedFreeLists();
2117 double frag = 0.0;
2118 size_t i;
2119
2120 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2121 double sz = i;
2122 frag += _indexedFreeList[i].count() * (sz * sz);
2123 }
2124
2125 double totFree = itabFree +
2126 _dictionary->total_chunk_size(DEBUG_ONLY(freelistLock()));
2127 if (totFree > 0) {
2128 frag = ((frag + _dictionary->sum_of_squared_block_sizes()) /
2129 (totFree * totFree));
2130 frag = (double)1.0 - frag;
2131 } else {
2132 assert(frag == 0.0, "Follows from totFree == 0");
2133 }
2134 return frag;
2135 }
2136
2137 void CompactibleFreeListSpace::beginSweepFLCensus(
2138 float inter_sweep_current,
2139 float inter_sweep_estimate,
2140 float intra_sweep_estimate) {
2141 assert_locked();
2142 size_t i;
2143 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2144 AdaptiveFreeList<FreeChunk>* fl = &_indexedFreeList[i];
2145 if (PrintFLSStatistics > 1) {
2146 gclog_or_tty->print("size[" SIZE_FORMAT "] : ", i);
2147 }
2148 fl->compute_desired(inter_sweep_current, inter_sweep_estimate, intra_sweep_estimate);
2149 fl->set_coal_desired((ssize_t)((double)fl->desired() * CMSSmallCoalSurplusPercent));
2150 fl->set_before_sweep(fl->count());
2151 fl->set_bfr_surp(fl->surplus());
2152 }
2153 _dictionary->begin_sweep_dict_census(CMSLargeCoalSurplusPercent,
2154 inter_sweep_current,
2155 inter_sweep_estimate,
2156 intra_sweep_estimate);
2157 }
2158
2159 void CompactibleFreeListSpace::setFLSurplus() {
2160 assert_locked();
2161 size_t i;
2162 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2163 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
2164 fl->set_surplus(fl->count() -
2165 (ssize_t)((double)fl->desired() * CMSSmallSplitSurplusPercent));
2166 }
2167 }
2168
2169 void CompactibleFreeListSpace::setFLHints() {
2170 assert_locked();
2171 size_t i;
2172 size_t h = IndexSetSize;
2173 for (i = IndexSetSize - 1; i != 0; i -= IndexSetStride) {
2174 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
2175 fl->set_hint(h);
2176 if (fl->surplus() > 0) {
2177 h = i;
2178 }
2179 }
2180 }
2181
2182 void CompactibleFreeListSpace::clearFLCensus() {
2183 assert_locked();
2184 size_t i;
2185 for (i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2186 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
2187 fl->set_prev_sweep(fl->count());
2188 fl->set_coal_births(0);
2189 fl->set_coal_deaths(0);
2190 fl->set_split_births(0);
2191 fl->set_split_deaths(0);
2192 }
2193 }
2194
2195 void CompactibleFreeListSpace::endSweepFLCensus(size_t sweep_count) {
2196 if (PrintFLSStatistics > 0) {
2197 HeapWord* largestAddr = (HeapWord*) dictionary()->find_largest_dict();
2198 gclog_or_tty->print_cr("CMS: Large block " PTR_FORMAT,
2199 p2i(largestAddr));
2200 }
2201 setFLSurplus();
2202 setFLHints();
2203 if (PrintGC && PrintFLSCensus > 0) {
2204 printFLCensus(sweep_count);
2205 }
2206 clearFLCensus();
2207 assert_locked();
2208 _dictionary->end_sweep_dict_census(CMSLargeSplitSurplusPercent);
2209 }
2210
2211 bool CompactibleFreeListSpace::coalOverPopulated(size_t size) {
2212 if (size < SmallForDictionary) {
2213 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
2214 return (fl->coal_desired() < 0) ||
2215 ((int)fl->count() > fl->coal_desired());
2216 } else {
2217 return dictionary()->coal_dict_over_populated(size);
2218 }
2219 }
2220
2221 void CompactibleFreeListSpace::smallCoalBirth(size_t size) {
2222 assert(size < SmallForDictionary, "Size too large for indexed list");
2223 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
2224 fl->increment_coal_births();
2225 fl->increment_surplus();
2226 }
2227
2228 void CompactibleFreeListSpace::smallCoalDeath(size_t size) {
2229 assert(size < SmallForDictionary, "Size too large for indexed list");
2230 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
2231 fl->increment_coal_deaths();
2232 fl->decrement_surplus();
2233 }
2234
2235 void CompactibleFreeListSpace::coalBirth(size_t size) {
2236 if (size < SmallForDictionary) {
2237 smallCoalBirth(size);
2238 } else {
2239 dictionary()->dict_census_update(size,
2240 false /* split */,
2241 true /* birth */);
2242 }
2243 }
2244
2245 void CompactibleFreeListSpace::coalDeath(size_t size) {
2246 if(size < SmallForDictionary) {
2247 smallCoalDeath(size);
2248 } else {
2249 dictionary()->dict_census_update(size,
2250 false /* split */,
2251 false /* birth */);
2252 }
2253 }
2254
2255 void CompactibleFreeListSpace::smallSplitBirth(size_t size) {
2256 assert(size < SmallForDictionary, "Size too large for indexed list");
2257 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
2258 fl->increment_split_births();
2259 fl->increment_surplus();
2260 }
2261
2262 void CompactibleFreeListSpace::smallSplitDeath(size_t size) {
2263 assert(size < SmallForDictionary, "Size too large for indexed list");
2264 AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[size];
2265 fl->increment_split_deaths();
2266 fl->decrement_surplus();
2267 }
2268
2269 void CompactibleFreeListSpace::split_birth(size_t size) {
2270 if (size < SmallForDictionary) {
2271 smallSplitBirth(size);
2272 } else {
2273 dictionary()->dict_census_update(size,
2274 true /* split */,
2275 true /* birth */);
2276 }
2277 }
2278
2279 void CompactibleFreeListSpace::splitDeath(size_t size) {
2280 if (size < SmallForDictionary) {
2281 smallSplitDeath(size);
2282 } else {
2283 dictionary()->dict_census_update(size,
2284 true /* split */,
2285 false /* birth */);
2286 }
2287 }
2288
2289 void CompactibleFreeListSpace::split(size_t from, size_t to1) {
2290 size_t to2 = from - to1;
2291 splitDeath(from);
2292 split_birth(to1);
2293 split_birth(to2);
2294 }
2295
2296 void CompactibleFreeListSpace::print() const {
2297 print_on(tty);
2298 }
2299
2300 void CompactibleFreeListSpace::prepare_for_verify() {
2301 assert_locked();
2302 repairLinearAllocationBlocks();
2303 // Verify that the SpoolBlocks look like free blocks of
2304 // appropriate sizes... To be done ...
2305 }
2306
2307 class VerifyAllBlksClosure: public BlkClosure {
2308 private:
2309 const CompactibleFreeListSpace* _sp;
2310 const MemRegion _span;
2311 HeapWord* _last_addr;
2312 size_t _last_size;
2313 bool _last_was_obj;
2314 bool _last_was_live;
2315
2316 public:
2317 VerifyAllBlksClosure(const CompactibleFreeListSpace* sp,
2318 MemRegion span) : _sp(sp), _span(span),
2319 _last_addr(NULL), _last_size(0),
2320 _last_was_obj(false), _last_was_live(false) { }
2321
2322 virtual size_t do_blk(HeapWord* addr) {
2323 size_t res;
2324 bool was_obj = false;
2325 bool was_live = false;
2326 if (_sp->block_is_obj(addr)) {
2327 was_obj = true;
2328 oop p = oop(addr);
2329 guarantee(p->is_oop(), "Should be an oop");
2330 res = _sp->adjustObjectSize(p->size());
2331 if (_sp->obj_is_alive(addr)) {
2332 was_live = true;
2333 p->verify();
2334 }
2335 } else {
2336 FreeChunk* fc = (FreeChunk*)addr;
2337 res = fc->size();
2338 if (FLSVerifyLists && !fc->cantCoalesce()) {
2339 guarantee(_sp->verify_chunk_in_free_list(fc),
2340 "Chunk should be on a free list");
2341 }
2342 }
2343 if (res == 0) {
2344 gclog_or_tty->print_cr("Livelock: no rank reduction!");
2345 gclog_or_tty->print_cr(
2346 " Current: addr = " PTR_FORMAT ", size = " SIZE_FORMAT ", obj = %s, live = %s \n"
2347 " Previous: addr = " PTR_FORMAT ", size = " SIZE_FORMAT ", obj = %s, live = %s \n",
2348 p2i(addr), res, was_obj ?"true":"false", was_live ?"true":"false",
2349 p2i(_last_addr), _last_size, _last_was_obj?"true":"false", _last_was_live?"true":"false");
2350 _sp->print_on(gclog_or_tty);
2351 guarantee(false, "Seppuku!");
2352 }
2353 _last_addr = addr;
2354 _last_size = res;
2355 _last_was_obj = was_obj;
2356 _last_was_live = was_live;
2357 return res;
2358 }
2359 };
2360
2361 class VerifyAllOopsClosure: public OopClosure {
2362 private:
2363 const CMSCollector* _collector;
2364 const CompactibleFreeListSpace* _sp;
2365 const MemRegion _span;
2366 const bool _past_remark;
2367 const CMSBitMap* _bit_map;
2368
2369 protected:
2370 void do_oop(void* p, oop obj) {
2371 if (_span.contains(obj)) { // the interior oop points into CMS heap
2372 if (!_span.contains(p)) { // reference from outside CMS heap
2373 // Should be a valid object; the first disjunct below allows
2374 // us to sidestep an assertion in block_is_obj() that insists
2375 // that p be in _sp. Note that several generations (and spaces)
2376 // are spanned by _span (CMS heap) above.
2377 guarantee(!_sp->is_in_reserved(obj) ||
2378 _sp->block_is_obj((HeapWord*)obj),
2379 "Should be an object");
2380 guarantee(obj->is_oop(), "Should be an oop");
2381 obj->verify();
2382 if (_past_remark) {
2383 // Remark has been completed, the object should be marked
2384 _bit_map->isMarked((HeapWord*)obj);
2385 }
2386 } else { // reference within CMS heap
2387 if (_past_remark) {
2388 // Remark has been completed -- so the referent should have
2389 // been marked, if referring object is.
2390 if (_bit_map->isMarked(_collector->block_start(p))) {
2391 guarantee(_bit_map->isMarked((HeapWord*)obj), "Marking error?");
2392 }
2393 }
2394 }
2395 } else if (_sp->is_in_reserved(p)) {
2396 // the reference is from FLS, and points out of FLS
2397 guarantee(obj->is_oop(), "Should be an oop");
2398 obj->verify();
2399 }
2400 }
2401
2402 template <class T> void do_oop_work(T* p) {
2403 T heap_oop = oopDesc::load_heap_oop(p);
2404 if (!oopDesc::is_null(heap_oop)) {
2405 oop obj = oopDesc::decode_heap_oop_not_null(heap_oop);
2406 do_oop(p, obj);
2407 }
2408 }
2409
2410 public:
2411 VerifyAllOopsClosure(const CMSCollector* collector,
2412 const CompactibleFreeListSpace* sp, MemRegion span,
2413 bool past_remark, CMSBitMap* bit_map) :
2414 _collector(collector), _sp(sp), _span(span),
2415 _past_remark(past_remark), _bit_map(bit_map) { }
2416
2417 virtual void do_oop(oop* p) { VerifyAllOopsClosure::do_oop_work(p); }
2418 virtual void do_oop(narrowOop* p) { VerifyAllOopsClosure::do_oop_work(p); }
2419 };
2420
2421 void CompactibleFreeListSpace::verify() const {
2422 assert_lock_strong(&_freelistLock);
2423 verify_objects_initialized();
2424 MemRegion span = _collector->_span;
2425 bool past_remark = (_collector->abstract_state() ==
2426 CMSCollector::Sweeping);
2427
2428 ResourceMark rm;
2429 HandleMark hm;
2430
2431 // Check integrity of CFL data structures
2432 _promoInfo.verify();
2433 _dictionary->verify();
2434 if (FLSVerifyIndexTable) {
2435 verifyIndexedFreeLists();
2436 }
2437 // Check integrity of all objects and free blocks in space
2438 {
2439 VerifyAllBlksClosure cl(this, span);
2440 ((CompactibleFreeListSpace*)this)->blk_iterate(&cl); // cast off const
2441 }
2442 // Check that all references in the heap to FLS
2443 // are to valid objects in FLS or that references in
2444 // FLS are to valid objects elsewhere in the heap
2445 if (FLSVerifyAllHeapReferences)
2446 {
2447 VerifyAllOopsClosure cl(_collector, this, span, past_remark,
2448 _collector->markBitMap());
2449 CollectedHeap* ch = Universe::heap();
2450
2451 // Iterate over all oops in the heap. Uses the _no_header version
2452 // since we are not interested in following the klass pointers.
2453 ch->oop_iterate_no_header(&cl);
2454 }
2455
2456 if (VerifyObjectStartArray) {
2457 // Verify the block offset table
2458 _bt.verify();
2459 }
2460 }
2461
2462 #ifndef PRODUCT
2463 void CompactibleFreeListSpace::verifyFreeLists() const {
2464 if (FLSVerifyLists) {
2465 _dictionary->verify();
2466 verifyIndexedFreeLists();
2467 } else {
2468 if (FLSVerifyDictionary) {
2469 _dictionary->verify();
2470 }
2471 if (FLSVerifyIndexTable) {
2472 verifyIndexedFreeLists();
2473 }
2474 }
2475 }
2476 #endif
2477
2478 void CompactibleFreeListSpace::verifyIndexedFreeLists() const {
2479 size_t i = 0;
2480 for (; i < IndexSetStart; i++) {
2481 guarantee(_indexedFreeList[i].head() == NULL, "should be NULL");
2482 }
2483 for (; i < IndexSetSize; i++) {
2484 verifyIndexedFreeList(i);
2485 }
2486 }
2487
2488 void CompactibleFreeListSpace::verifyIndexedFreeList(size_t size) const {
2489 FreeChunk* fc = _indexedFreeList[size].head();
2490 FreeChunk* tail = _indexedFreeList[size].tail();
2491 size_t num = _indexedFreeList[size].count();
2492 size_t n = 0;
2493 guarantee(((size >= IndexSetStart) && (size % IndexSetStride == 0)) || fc == NULL,
2494 "Slot should have been empty");
2495 for (; fc != NULL; fc = fc->next(), n++) {
2496 guarantee(fc->size() == size, "Size inconsistency");
2497 guarantee(fc->is_free(), "!free?");
2498 guarantee(fc->next() == NULL || fc->next()->prev() == fc, "Broken list");
2499 guarantee((fc->next() == NULL) == (fc == tail), "Incorrect tail");
2500 }
2501 guarantee(n == num, "Incorrect count");
2502 }
2503
2504 #ifndef PRODUCT
2505 void CompactibleFreeListSpace::check_free_list_consistency() const {
2506 assert((TreeChunk<FreeChunk, AdaptiveFreeList<FreeChunk> >::min_size() <= IndexSetSize),
2507 "Some sizes can't be allocated without recourse to"
2508 " linear allocation buffers");
2509 assert((TreeChunk<FreeChunk, AdaptiveFreeList<FreeChunk> >::min_size()*HeapWordSize == sizeof(TreeChunk<FreeChunk, AdaptiveFreeList<FreeChunk> >)),
2510 "else MIN_TREE_CHUNK_SIZE is wrong");
2511 assert(IndexSetStart != 0, "IndexSetStart not initialized");
2512 assert(IndexSetStride != 0, "IndexSetStride not initialized");
2513 }
2514 #endif
2515
2516 void CompactibleFreeListSpace::printFLCensus(size_t sweep_count) const {
2517 assert_lock_strong(&_freelistLock);
2518 AdaptiveFreeList<FreeChunk> total;
2519 gclog_or_tty->print("end sweep# " SIZE_FORMAT "\n", sweep_count);
2520 AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
2521 size_t total_free = 0;
2522 for (size_t i = IndexSetStart; i < IndexSetSize; i += IndexSetStride) {
2523 const AdaptiveFreeList<FreeChunk> *fl = &_indexedFreeList[i];
2524 total_free += fl->count() * fl->size();
2525 if (i % (40*IndexSetStride) == 0) {
2526 AdaptiveFreeList<FreeChunk>::print_labels_on(gclog_or_tty, "size");
2527 }
2528 fl->print_on(gclog_or_tty);
2529 total.set_bfr_surp( total.bfr_surp() + fl->bfr_surp() );
2530 total.set_surplus( total.surplus() + fl->surplus() );
2531 total.set_desired( total.desired() + fl->desired() );
2532 total.set_prev_sweep( total.prev_sweep() + fl->prev_sweep() );
2533 total.set_before_sweep(total.before_sweep() + fl->before_sweep());
2534 total.set_count( total.count() + fl->count() );
2535 total.set_coal_births( total.coal_births() + fl->coal_births() );
2536 total.set_coal_deaths( total.coal_deaths() + fl->coal_deaths() );
2537 total.set_split_births(total.split_births() + fl->split_births());
2538 total.set_split_deaths(total.split_deaths() + fl->split_deaths());
2539 }
2540 total.print_on(gclog_or_tty, "TOTAL");
2541 gclog_or_tty->print_cr("Total free in indexed lists "
2542 SIZE_FORMAT " words", total_free);
2543 gclog_or_tty->print("growth: %8.5f deficit: %8.5f\n",
2544 (double)(total.split_births()+total.coal_births()-total.split_deaths()-total.coal_deaths())/
2545 (total.prev_sweep() != 0 ? (double)total.prev_sweep() : 1.0),
2546 (double)(total.desired() - total.count())/(total.desired() != 0 ? (double)total.desired() : 1.0));
2547 _dictionary->print_dict_census();
2548 }
2549
2550 ///////////////////////////////////////////////////////////////////////////
2551 // CFLS_LAB
2552 ///////////////////////////////////////////////////////////////////////////
2553
2554 #define VECTOR_257(x) \
2555 /* 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 */ \
2556 { 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, \
2557 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, \
2558 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, \
2559 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, \
2560 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, \
2561 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, \
2562 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, \
2563 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, \
2564 x }
2565
2566 // Initialize with default setting for CMS, _not_
2567 // generic OldPLABSize, whose static default is different; if overridden at the
2568 // command-line, this will get reinitialized via a call to
2569 // modify_initialization() below.
2570 AdaptiveWeightedAverage CFLS_LAB::_blocks_to_claim[] =
2571 VECTOR_257(AdaptiveWeightedAverage(OldPLABWeight, (float)CFLS_LAB::_default_dynamic_old_plab_size));
2572 size_t CFLS_LAB::_global_num_blocks[] = VECTOR_257(0);
2573 uint CFLS_LAB::_global_num_workers[] = VECTOR_257(0);
2574
2575 CFLS_LAB::CFLS_LAB(CompactibleFreeListSpace* cfls) :
2576 _cfls(cfls)
2577 {
2578 assert(CompactibleFreeListSpace::IndexSetSize == 257, "Modify VECTOR_257() macro above");
2579 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2580 i < CompactibleFreeListSpace::IndexSetSize;
2581 i += CompactibleFreeListSpace::IndexSetStride) {
2582 _indexedFreeList[i].set_size(i);
2583 _num_blocks[i] = 0;
2584 }
2585 }
2586
2587 static bool _CFLS_LAB_modified = false;
2588
2589 void CFLS_LAB::modify_initialization(size_t n, unsigned wt) {
2590 assert(!_CFLS_LAB_modified, "Call only once");
2591 _CFLS_LAB_modified = true;
2592 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2593 i < CompactibleFreeListSpace::IndexSetSize;
2594 i += CompactibleFreeListSpace::IndexSetStride) {
2595 _blocks_to_claim[i].modify(n, wt, true /* force */);
2596 }
2597 }
2598
2599 HeapWord* CFLS_LAB::alloc(size_t word_sz) {
2600 FreeChunk* res;
2601 assert(word_sz == _cfls->adjustObjectSize(word_sz), "Error");
2602 if (word_sz >= CompactibleFreeListSpace::IndexSetSize) {
2603 // This locking manages sync with other large object allocations.
2604 MutexLockerEx x(_cfls->parDictionaryAllocLock(),
2605 Mutex::_no_safepoint_check_flag);
2606 res = _cfls->getChunkFromDictionaryExact(word_sz);
2607 if (res == NULL) return NULL;
2608 } else {
2609 AdaptiveFreeList<FreeChunk>* fl = &_indexedFreeList[word_sz];
2610 if (fl->count() == 0) {
2611 // Attempt to refill this local free list.
2612 get_from_global_pool(word_sz, fl);
2613 // If it didn't work, give up.
2614 if (fl->count() == 0) return NULL;
2615 }
2616 res = fl->get_chunk_at_head();
2617 assert(res != NULL, "Why was count non-zero?");
2618 }
2619 res->markNotFree();
2620 assert(!res->is_free(), "shouldn't be marked free");
2621 assert(oop(res)->klass_or_null() == NULL, "should look uninitialized");
2622 // mangle a just allocated object with a distinct pattern.
2623 debug_only(res->mangleAllocated(word_sz));
2624 return (HeapWord*)res;
2625 }
2626
2627 // Get a chunk of blocks of the right size and update related
2628 // book-keeping stats
2629 void CFLS_LAB::get_from_global_pool(size_t word_sz, AdaptiveFreeList<FreeChunk>* fl) {
2630 // Get the #blocks we want to claim
2631 size_t n_blks = (size_t)_blocks_to_claim[word_sz].average();
2632 assert(n_blks > 0, "Error");
2633 assert(ResizeOldPLAB || n_blks == OldPLABSize, "Error");
2634 // In some cases, when the application has a phase change,
2635 // there may be a sudden and sharp shift in the object survival
2636 // profile, and updating the counts at the end of a scavenge
2637 // may not be quick enough, giving rise to large scavenge pauses
2638 // during these phase changes. It is beneficial to detect such
2639 // changes on-the-fly during a scavenge and avoid such a phase-change
2640 // pothole. The following code is a heuristic attempt to do that.
2641 // It is protected by a product flag until we have gained
2642 // enough experience with this heuristic and fine-tuned its behavior.
2643 // WARNING: This might increase fragmentation if we overreact to
2644 // small spikes, so some kind of historical smoothing based on
2645 // previous experience with the greater reactivity might be useful.
2646 // Lacking sufficient experience, CMSOldPLABResizeQuicker is disabled by
2647 // default.
2648 if (ResizeOldPLAB && CMSOldPLABResizeQuicker) {
2649 size_t multiple = _num_blocks[word_sz]/(CMSOldPLABToleranceFactor*CMSOldPLABNumRefills*n_blks);
2650 n_blks += CMSOldPLABReactivityFactor*multiple*n_blks;
2651 n_blks = MIN2(n_blks, CMSOldPLABMax);
2652 }
2653 assert(n_blks > 0, "Error");
2654 _cfls->par_get_chunk_of_blocks(word_sz, n_blks, fl);
2655 // Update stats table entry for this block size
2656 _num_blocks[word_sz] += fl->count();
2657 }
2658
2659 void CFLS_LAB::compute_desired_plab_size() {
2660 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2661 i < CompactibleFreeListSpace::IndexSetSize;
2662 i += CompactibleFreeListSpace::IndexSetStride) {
2663 assert((_global_num_workers[i] == 0) == (_global_num_blocks[i] == 0),
2664 "Counter inconsistency");
2665 if (_global_num_workers[i] > 0) {
2666 // Need to smooth wrt historical average
2667 if (ResizeOldPLAB) {
2668 _blocks_to_claim[i].sample(
2669 MAX2((size_t)CMSOldPLABMin,
2670 MIN2((size_t)CMSOldPLABMax,
2671 _global_num_blocks[i]/(_global_num_workers[i]*CMSOldPLABNumRefills))));
2672 }
2673 // Reset counters for next round
2674 _global_num_workers[i] = 0;
2675 _global_num_blocks[i] = 0;
2676 if (PrintOldPLAB) {
2677 gclog_or_tty->print_cr("[" SIZE_FORMAT "]: " SIZE_FORMAT,
2678 i, (size_t)_blocks_to_claim[i].average());
2679 }
2680 }
2681 }
2682 }
2683
2684 // If this is changed in the future to allow parallel
2685 // access, one would need to take the FL locks and,
2686 // depending on how it is used, stagger access from
2687 // parallel threads to reduce contention.
2688 void CFLS_LAB::retire(int tid) {
2689 // We run this single threaded with the world stopped;
2690 // so no need for locks and such.
2691 NOT_PRODUCT(Thread* t = Thread::current();)
2692 assert(Thread::current()->is_VM_thread(), "Error");
2693 for (size_t i = CompactibleFreeListSpace::IndexSetStart;
2694 i < CompactibleFreeListSpace::IndexSetSize;
2695 i += CompactibleFreeListSpace::IndexSetStride) {
2696 assert(_num_blocks[i] >= (size_t)_indexedFreeList[i].count(),
2697 "Can't retire more than what we obtained");
2698 if (_num_blocks[i] > 0) {
2699 size_t num_retire = _indexedFreeList[i].count();
2700 assert(_num_blocks[i] > num_retire, "Should have used at least one");
2701 {
2702 // MutexLockerEx x(_cfls->_indexedFreeListParLocks[i],
2703 // Mutex::_no_safepoint_check_flag);
2704
2705 // Update globals stats for num_blocks used
2706 _global_num_blocks[i] += (_num_blocks[i] - num_retire);
2707 _global_num_workers[i]++;
2708 assert(_global_num_workers[i] <= ParallelGCThreads, "Too big");
2709 if (num_retire > 0) {
2710 _cfls->_indexedFreeList[i].prepend(&_indexedFreeList[i]);
2711 // Reset this list.
2712 _indexedFreeList[i] = AdaptiveFreeList<FreeChunk>();
2713 _indexedFreeList[i].set_size(i);
2714 }
2715 }
2716 if (PrintOldPLAB) {
2717 gclog_or_tty->print_cr("%d[" SIZE_FORMAT "]: " SIZE_FORMAT "/" SIZE_FORMAT "/" SIZE_FORMAT,
2718 tid, i, num_retire, _num_blocks[i], (size_t)_blocks_to_claim[i].average());
2719 }
2720 // Reset stats for next round
2721 _num_blocks[i] = 0;
2722 }
2723 }
2724 }
2725
2726 // Used by par_get_chunk_of_blocks() for the chunks from the
2727 // indexed_free_lists. Looks for a chunk with size that is a multiple
2728 // of "word_sz" and if found, splits it into "word_sz" chunks and add
2729 // to the free list "fl". "n" is the maximum number of chunks to
2730 // be added to "fl".
2731 bool CompactibleFreeListSpace:: par_get_chunk_of_blocks_IFL(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl) {
2732
2733 // We'll try all multiples of word_sz in the indexed set, starting with
2734 // word_sz itself and, if CMSSplitIndexedFreeListBlocks, try larger multiples,
2735 // then try getting a big chunk and splitting it.
2736 {
2737 bool found;
2738 int k;
2739 size_t cur_sz;
2740 for (k = 1, cur_sz = k * word_sz, found = false;
2741 (cur_sz < CompactibleFreeListSpace::IndexSetSize) &&
2742 (CMSSplitIndexedFreeListBlocks || k <= 1);
2743 k++, cur_sz = k * word_sz) {
2744 AdaptiveFreeList<FreeChunk> fl_for_cur_sz; // Empty.
2745 fl_for_cur_sz.set_size(cur_sz);
2746 {
2747 MutexLockerEx x(_indexedFreeListParLocks[cur_sz],
2748 Mutex::_no_safepoint_check_flag);
2749 AdaptiveFreeList<FreeChunk>* gfl = &_indexedFreeList[cur_sz];
2750 if (gfl->count() != 0) {
2751 // nn is the number of chunks of size cur_sz that
2752 // we'd need to split k-ways each, in order to create
2753 // "n" chunks of size word_sz each.
2754 const size_t nn = MAX2(n/k, (size_t)1);
2755 gfl->getFirstNChunksFromList(nn, &fl_for_cur_sz);
2756 found = true;
2757 if (k > 1) {
2758 // Update split death stats for the cur_sz-size blocks list:
2759 // we increment the split death count by the number of blocks
2760 // we just took from the cur_sz-size blocks list and which
2761 // we will be splitting below.
2762 ssize_t deaths = gfl->split_deaths() +
2763 fl_for_cur_sz.count();
2764 gfl->set_split_deaths(deaths);
2765 }
2766 }
2767 }
2768 // Now transfer fl_for_cur_sz to fl. Common case, we hope, is k = 1.
2769 if (found) {
2770 if (k == 1) {
2771 fl->prepend(&fl_for_cur_sz);
2772 } else {
2773 // Divide each block on fl_for_cur_sz up k ways.
2774 FreeChunk* fc;
2775 while ((fc = fl_for_cur_sz.get_chunk_at_head()) != NULL) {
2776 // Must do this in reverse order, so that anybody attempting to
2777 // access the main chunk sees it as a single free block until we
2778 // change it.
2779 size_t fc_size = fc->size();
2780 assert(fc->is_free(), "Error");
2781 for (int i = k-1; i >= 0; i--) {
2782 FreeChunk* ffc = (FreeChunk*)((HeapWord*)fc + i * word_sz);
2783 assert((i != 0) ||
2784 ((fc == ffc) && ffc->is_free() &&
2785 (ffc->size() == k*word_sz) && (fc_size == word_sz)),
2786 "Counting error");
2787 ffc->set_size(word_sz);
2788 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2789 ffc->link_next(NULL);
2790 // Above must occur before BOT is updated below.
2791 OrderAccess::storestore();
2792 // splitting from the right, fc_size == i * word_sz
2793 _bt.mark_block((HeapWord*)ffc, word_sz, true /* reducing */);
2794 fc_size -= word_sz;
2795 assert(fc_size == i*word_sz, "Error");
2796 _bt.verify_not_unallocated((HeapWord*)ffc, word_sz);
2797 _bt.verify_single_block((HeapWord*)fc, fc_size);
2798 _bt.verify_single_block((HeapWord*)ffc, word_sz);
2799 // Push this on "fl".
2800 fl->return_chunk_at_head(ffc);
2801 }
2802 // TRAP
2803 assert(fl->tail()->next() == NULL, "List invariant.");
2804 }
2805 }
2806 // Update birth stats for this block size.
2807 size_t num = fl->count();
2808 MutexLockerEx x(_indexedFreeListParLocks[word_sz],
2809 Mutex::_no_safepoint_check_flag);
2810 ssize_t births = _indexedFreeList[word_sz].split_births() + num;
2811 _indexedFreeList[word_sz].set_split_births(births);
2812 return true;
2813 }
2814 }
2815 return found;
2816 }
2817 }
2818
2819 FreeChunk* CompactibleFreeListSpace::get_n_way_chunk_to_split(size_t word_sz, size_t n) {
2820
2821 FreeChunk* fc = NULL;
2822 FreeChunk* rem_fc = NULL;
2823 size_t rem;
2824 {
2825 MutexLockerEx x(parDictionaryAllocLock(),
2826 Mutex::_no_safepoint_check_flag);
2827 while (n > 0) {
2828 fc = dictionary()->get_chunk(MAX2(n * word_sz, _dictionary->min_size()),
2829 FreeBlockDictionary<FreeChunk>::atLeast);
2830 if (fc != NULL) {
2831 break;
2832 } else {
2833 n--;
2834 }
2835 }
2836 if (fc == NULL) return NULL;
2837 // Otherwise, split up that block.
2838 assert((ssize_t)n >= 1, "Control point invariant");
2839 assert(fc->is_free(), "Error: should be a free block");
2840 _bt.verify_single_block((HeapWord*)fc, fc->size());
2841 const size_t nn = fc->size() / word_sz;
2842 n = MIN2(nn, n);
2843 assert((ssize_t)n >= 1, "Control point invariant");
2844 rem = fc->size() - n * word_sz;
2845 // If there is a remainder, and it's too small, allocate one fewer.
2846 if (rem > 0 && rem < MinChunkSize) {
2847 n--; rem += word_sz;
2848 }
2849 // Note that at this point we may have n == 0.
2850 assert((ssize_t)n >= 0, "Control point invariant");
2851
2852 // If n is 0, the chunk fc that was found is not large
2853 // enough to leave a viable remainder. We are unable to
2854 // allocate even one block. Return fc to the
2855 // dictionary and return, leaving "fl" empty.
2856 if (n == 0) {
2857 returnChunkToDictionary(fc);
2858 return NULL;
2859 }
2860
2861 _bt.allocated((HeapWord*)fc, fc->size(), true /* reducing */); // update _unallocated_blk
2862 dictionary()->dict_census_update(fc->size(),
2863 true /*split*/,
2864 false /*birth*/);
2865
2866 // First return the remainder, if any.
2867 // Note that we hold the lock until we decide if we're going to give
2868 // back the remainder to the dictionary, since a concurrent allocation
2869 // may otherwise see the heap as empty. (We're willing to take that
2870 // hit if the block is a small block.)
2871 if (rem > 0) {
2872 size_t prefix_size = n * word_sz;
2873 rem_fc = (FreeChunk*)((HeapWord*)fc + prefix_size);
2874 rem_fc->set_size(rem);
2875 rem_fc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2876 rem_fc->link_next(NULL);
2877 // Above must occur before BOT is updated below.
2878 assert((ssize_t)n > 0 && prefix_size > 0 && rem_fc > fc, "Error");
2879 OrderAccess::storestore();
2880 _bt.split_block((HeapWord*)fc, fc->size(), prefix_size);
2881 assert(fc->is_free(), "Error");
2882 fc->set_size(prefix_size);
2883 if (rem >= IndexSetSize) {
2884 returnChunkToDictionary(rem_fc);
2885 dictionary()->dict_census_update(rem, true /*split*/, true /*birth*/);
2886 rem_fc = NULL;
2887 }
2888 // Otherwise, return it to the small list below.
2889 }
2890 }
2891 if (rem_fc != NULL) {
2892 MutexLockerEx x(_indexedFreeListParLocks[rem],
2893 Mutex::_no_safepoint_check_flag);
2894 _bt.verify_not_unallocated((HeapWord*)rem_fc, rem_fc->size());
2895 _indexedFreeList[rem].return_chunk_at_head(rem_fc);
2896 smallSplitBirth(rem);
2897 }
2898 assert(n * word_sz == fc->size(),
2899 err_msg("Chunk size " SIZE_FORMAT " is not exactly splittable by "
2900 SIZE_FORMAT " sized chunks of size " SIZE_FORMAT,
2901 fc->size(), n, word_sz));
2902 return fc;
2903 }
2904
2905 void CompactibleFreeListSpace:: par_get_chunk_of_blocks_dictionary(size_t word_sz, size_t targetted_number_of_chunks, AdaptiveFreeList<FreeChunk>* fl) {
2906
2907 FreeChunk* fc = get_n_way_chunk_to_split(word_sz, targetted_number_of_chunks);
2908
2909 if (fc == NULL) {
2910 return;
2911 }
2912
2913 size_t n = fc->size() / word_sz;
2914
2915 assert((ssize_t)n > 0, "Consistency");
2916 // Now do the splitting up.
2917 // Must do this in reverse order, so that anybody attempting to
2918 // access the main chunk sees it as a single free block until we
2919 // change it.
2920 size_t fc_size = n * word_sz;
2921 // All but first chunk in this loop
2922 for (ssize_t i = n-1; i > 0; i--) {
2923 FreeChunk* ffc = (FreeChunk*)((HeapWord*)fc + i * word_sz);
2924 ffc->set_size(word_sz);
2925 ffc->link_prev(NULL); // Mark as a free block for other (parallel) GC threads.
2926 ffc->link_next(NULL);
2927 // Above must occur before BOT is updated below.
2928 OrderAccess::storestore();
2929 // splitting from the right, fc_size == (n - i + 1) * wordsize
2930 _bt.mark_block((HeapWord*)ffc, word_sz, true /* reducing */);
2931 fc_size -= word_sz;
2932 _bt.verify_not_unallocated((HeapWord*)ffc, ffc->size());
2933 _bt.verify_single_block((HeapWord*)ffc, ffc->size());
2934 _bt.verify_single_block((HeapWord*)fc, fc_size);
2935 // Push this on "fl".
2936 fl->return_chunk_at_head(ffc);
2937 }
2938 // First chunk
2939 assert(fc->is_free() && fc->size() == n*word_sz, "Error: should still be a free block");
2940 // The blocks above should show their new sizes before the first block below
2941 fc->set_size(word_sz);
2942 fc->link_prev(NULL); // idempotent wrt free-ness, see assert above
2943 fc->link_next(NULL);
2944 _bt.verify_not_unallocated((HeapWord*)fc, fc->size());
2945 _bt.verify_single_block((HeapWord*)fc, fc->size());
2946 fl->return_chunk_at_head(fc);
2947
2948 assert((ssize_t)n > 0 && (ssize_t)n == fl->count(), "Incorrect number of blocks");
2949 {
2950 // Update the stats for this block size.
2951 MutexLockerEx x(_indexedFreeListParLocks[word_sz],
2952 Mutex::_no_safepoint_check_flag);
2953 const ssize_t births = _indexedFreeList[word_sz].split_births() + n;
2954 _indexedFreeList[word_sz].set_split_births(births);
2955 // ssize_t new_surplus = _indexedFreeList[word_sz].surplus() + n;
2956 // _indexedFreeList[word_sz].set_surplus(new_surplus);
2957 }
2958
2959 // TRAP
2960 assert(fl->tail()->next() == NULL, "List invariant.");
2961 }
2962
2963 void CompactibleFreeListSpace:: par_get_chunk_of_blocks(size_t word_sz, size_t n, AdaptiveFreeList<FreeChunk>* fl) {
2964 assert(fl->count() == 0, "Precondition.");
2965 assert(word_sz < CompactibleFreeListSpace::IndexSetSize,
2966 "Precondition");
2967
2968 if (par_get_chunk_of_blocks_IFL(word_sz, n, fl)) {
2969 // Got it
2970 return;
2971 }
2972
2973 // Otherwise, we'll split a block from the dictionary.
2974 par_get_chunk_of_blocks_dictionary(word_sz, n, fl);
2975 }
2976
2977 // Set up the space's par_seq_tasks structure for work claiming
2978 // for parallel rescan. See CMSParRemarkTask where this is currently used.
2979 // XXX Need to suitably abstract and generalize this and the next
2980 // method into one.
2981 void
2982 CompactibleFreeListSpace::
2983 initialize_sequential_subtasks_for_rescan(int n_threads) {
2984 // The "size" of each task is fixed according to rescan_task_size.
2985 assert(n_threads > 0, "Unexpected n_threads argument");
2986 const size_t task_size = rescan_task_size();
2987 size_t n_tasks = (used_region().word_size() + task_size - 1)/task_size;
2988 assert((n_tasks == 0) == used_region().is_empty(), "n_tasks incorrect");
2989 assert(n_tasks == 0 ||
2990 ((used_region().start() + (n_tasks - 1)*task_size < used_region().end()) &&
2991 (used_region().start() + n_tasks*task_size >= used_region().end())),
2992 "n_tasks calculation incorrect");
2993 SequentialSubTasksDone* pst = conc_par_seq_tasks();
2994 assert(!pst->valid(), "Clobbering existing data?");
2995 // Sets the condition for completion of the subtask (how many threads
2996 // need to finish in order to be done).
2997 pst->set_n_threads(n_threads);
2998 pst->set_n_tasks((int)n_tasks);
2999 }
3000
3001 // Set up the space's par_seq_tasks structure for work claiming
3002 // for parallel concurrent marking. See CMSConcMarkTask where this is currently used.
3003 void
3004 CompactibleFreeListSpace::
3005 initialize_sequential_subtasks_for_marking(int n_threads,
3006 HeapWord* low) {
3007 // The "size" of each task is fixed according to rescan_task_size.
3008 assert(n_threads > 0, "Unexpected n_threads argument");
3009 const size_t task_size = marking_task_size();
3010 assert(task_size > CardTableModRefBS::card_size_in_words &&
3011 (task_size % CardTableModRefBS::card_size_in_words == 0),
3012 "Otherwise arithmetic below would be incorrect");
3013 MemRegion span = _gen->reserved();
3014 if (low != NULL) {
3015 if (span.contains(low)) {
3016 // Align low down to a card boundary so that
3017 // we can use block_offset_careful() on span boundaries.
3018 HeapWord* aligned_low = (HeapWord*)align_size_down((uintptr_t)low,
3019 CardTableModRefBS::card_size);
3020 // Clip span prefix at aligned_low
3021 span = span.intersection(MemRegion(aligned_low, span.end()));
3022 } else if (low > span.end()) {
3023 span = MemRegion(low, low); // Null region
3024 } // else use entire span
3025 }
3026 assert(span.is_empty() ||
3027 ((uintptr_t)span.start() % CardTableModRefBS::card_size == 0),
3028 "span should start at a card boundary");
3029 size_t n_tasks = (span.word_size() + task_size - 1)/task_size;
3030 assert((n_tasks == 0) == span.is_empty(), "Inconsistency");
3031 assert(n_tasks == 0 ||
3032 ((span.start() + (n_tasks - 1)*task_size < span.end()) &&
3033 (span.start() + n_tasks*task_size >= span.end())),
3034 "n_tasks calculation incorrect");
3035 SequentialSubTasksDone* pst = conc_par_seq_tasks();
3036 assert(!pst->valid(), "Clobbering existing data?");
3037 // Sets the condition for completion of the subtask (how many threads
3038 // need to finish in order to be done).
3039 pst->set_n_threads(n_threads);
3040 pst->set_n_tasks((int)n_tasks);
3041 }