1 /*
2 * Copyright (c) 2018, 2019, 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/shared/oopStorage.inline.hpp"
27 #include "gc/shared/oopStorageParState.inline.hpp"
28 #include "logging/log.hpp"
29 #include "logging/logStream.hpp"
30 #include "memory/allocation.inline.hpp"
31 #include "runtime/atomic.hpp"
32 #include "runtime/globals.hpp"
33 #include "runtime/handles.inline.hpp"
34 #include "runtime/interfaceSupport.inline.hpp"
35 #include "runtime/mutex.hpp"
36 #include "runtime/mutexLocker.inline.hpp"
37 #include "runtime/orderAccess.hpp"
38 #include "runtime/os.hpp"
39 #include "runtime/safepoint.hpp"
40 #include "runtime/stubRoutines.hpp"
41 #include "runtime/thread.hpp"
42 #include "utilities/align.hpp"
43 #include "utilities/count_trailing_zeros.hpp"
44 #include "utilities/debug.hpp"
45 #include "utilities/globalDefinitions.hpp"
46 #include "utilities/macros.hpp"
47 #include "utilities/ostream.hpp"
48
49 OopStorage::AllocationListEntry::AllocationListEntry() : _prev(NULL), _next(NULL) {}
50
51 OopStorage::AllocationListEntry::~AllocationListEntry() {
52 assert(_prev == NULL, "deleting attached block");
53 assert(_next == NULL, "deleting attached block");
54 }
55
56 OopStorage::AllocationList::AllocationList() : _head(NULL), _tail(NULL) {}
57
58 OopStorage::AllocationList::~AllocationList() {
59 // ~OopStorage() empties its lists before destroying them.
60 assert(_head == NULL, "deleting non-empty block list");
61 assert(_tail == NULL, "deleting non-empty block list");
62 }
63
64 void OopStorage::AllocationList::push_front(const Block& block) {
65 const Block* old = _head;
66 if (old == NULL) {
67 assert(_tail == NULL, "invariant");
68 _head = _tail = █
69 } else {
70 block.allocation_list_entry()._next = old;
71 old->allocation_list_entry()._prev = █
72 _head = █
73 }
74 }
75
76 void OopStorage::AllocationList::push_back(const Block& block) {
77 const Block* old = _tail;
78 if (old == NULL) {
79 assert(_head == NULL, "invariant");
80 _head = _tail = █
81 } else {
82 old->allocation_list_entry()._next = █
83 block.allocation_list_entry()._prev = old;
84 _tail = █
85 }
86 }
87
88 void OopStorage::AllocationList::unlink(const Block& block) {
89 const AllocationListEntry& block_entry = block.allocation_list_entry();
90 const Block* prev_blk = block_entry._prev;
91 const Block* next_blk = block_entry._next;
92 block_entry._prev = NULL;
93 block_entry._next = NULL;
94 if ((prev_blk == NULL) && (next_blk == NULL)) {
95 assert(_head == &block, "invariant");
96 assert(_tail == &block, "invariant");
97 _head = _tail = NULL;
98 } else if (prev_blk == NULL) {
99 assert(_head == &block, "invariant");
100 next_blk->allocation_list_entry()._prev = NULL;
101 _head = next_blk;
102 } else if (next_blk == NULL) {
103 assert(_tail == &block, "invariant");
104 prev_blk->allocation_list_entry()._next = NULL;
105 _tail = prev_blk;
106 } else {
107 next_blk->allocation_list_entry()._prev = prev_blk;
108 prev_blk->allocation_list_entry()._next = next_blk;
109 }
110 }
111
112 OopStorage::ActiveArray::ActiveArray(size_t size) :
113 _size(size),
114 _block_count(0),
115 _refcount(0)
116 {}
117
118 OopStorage::ActiveArray::~ActiveArray() {
119 assert(_refcount == 0, "precondition");
120 }
121
122 OopStorage::ActiveArray* OopStorage::ActiveArray::create(size_t size, AllocFailType alloc_fail) {
123 size_t size_in_bytes = blocks_offset() + sizeof(Block*) * size;
124 void* mem = NEW_C_HEAP_ARRAY3(char, size_in_bytes, mtGC, CURRENT_PC, alloc_fail);
125 if (mem == NULL) return NULL;
126 return new (mem) ActiveArray(size);
127 }
128
129 void OopStorage::ActiveArray::destroy(ActiveArray* ba) {
130 ba->~ActiveArray();
131 FREE_C_HEAP_ARRAY(char, ba);
132 }
133
134 size_t OopStorage::ActiveArray::size() const {
135 return _size;
136 }
137
138 size_t OopStorage::ActiveArray::block_count() const {
139 return _block_count;
140 }
141
142 size_t OopStorage::ActiveArray::block_count_acquire() const {
143 return OrderAccess::load_acquire(&_block_count);
144 }
145
146 void OopStorage::ActiveArray::increment_refcount() const {
147 int new_value = Atomic::add(1, &_refcount);
148 assert(new_value >= 1, "negative refcount %d", new_value - 1);
149 }
150
151 bool OopStorage::ActiveArray::decrement_refcount() const {
152 int new_value = Atomic::sub(1, &_refcount);
153 assert(new_value >= 0, "negative refcount %d", new_value);
154 return new_value == 0;
155 }
156
157 bool OopStorage::ActiveArray::push(Block* block) {
158 size_t index = _block_count;
159 if (index < _size) {
160 block->set_active_index(index);
161 *block_ptr(index) = block;
162 // Use a release_store to ensure all the setup is complete before
163 // making the block visible.
164 OrderAccess::release_store(&_block_count, index + 1);
165 return true;
166 } else {
167 return false;
168 }
169 }
170
171 void OopStorage::ActiveArray::remove(Block* block) {
172 assert(_block_count > 0, "array is empty");
173 size_t index = block->active_index();
174 assert(*block_ptr(index) == block, "block not present");
175 size_t last_index = _block_count - 1;
176 Block* last_block = *block_ptr(last_index);
177 last_block->set_active_index(index);
178 *block_ptr(index) = last_block;
179 _block_count = last_index;
180 }
181
182 void OopStorage::ActiveArray::copy_from(const ActiveArray* from) {
183 assert(_block_count == 0, "array must be empty");
184 size_t count = from->_block_count;
185 assert(count <= _size, "precondition");
186 Block* const* from_ptr = from->block_ptr(0);
187 Block** to_ptr = block_ptr(0);
188 for (size_t i = 0; i < count; ++i) {
189 Block* block = *from_ptr++;
190 assert(block->active_index() == i, "invariant");
191 *to_ptr++ = block;
192 }
193 _block_count = count;
194 }
195
196 // Blocks start with an array of BitsPerWord oop entries. That array
197 // is divided into conceptual BytesPerWord sections of BitsPerByte
198 // entries. Blocks are allocated aligned on section boundaries, for
199 // the convenience of mapping from an entry to the containing block;
200 // see block_for_ptr(). Aligning on section boundary rather than on
201 // the full _data wastes a lot less space, but makes for a bit more
202 // work in block_for_ptr().
203
204 const unsigned section_size = BitsPerByte;
205 const unsigned section_count = BytesPerWord;
206 const unsigned block_alignment = sizeof(oop) * section_size;
207
208 OopStorage::Block::Block(const OopStorage* owner, void* memory) :
209 _data(),
210 _allocated_bitmask(0),
211 _owner_address(reinterpret_cast<intptr_t>(owner)),
212 _memory(memory),
213 _active_index(0),
214 _allocation_list_entry(),
215 _deferred_updates_next(NULL),
216 _release_refcount(0)
217 {
218 STATIC_ASSERT(_data_pos == 0);
219 STATIC_ASSERT(section_size * section_count == ARRAY_SIZE(_data));
220 assert(offset_of(Block, _data) == _data_pos, "invariant");
221 assert(owner != NULL, "NULL owner");
222 assert(is_aligned(this, block_alignment), "misaligned block");
223 }
224
225 OopStorage::Block::~Block() {
226 assert(_release_refcount == 0, "deleting block while releasing");
227 assert(_deferred_updates_next == NULL, "deleting block with deferred update");
228 // Clear fields used by block_for_ptr and entry validation, which
229 // might help catch bugs. Volatile to prevent dead-store elimination.
230 const_cast<uintx volatile&>(_allocated_bitmask) = 0;
231 const_cast<intptr_t volatile&>(_owner_address) = 0;
232 }
233
234 size_t OopStorage::Block::allocation_size() {
235 // _data must be first member, so aligning Block aligns _data.
236 STATIC_ASSERT(_data_pos == 0);
237 return sizeof(Block) + block_alignment - sizeof(void*);
238 }
239
240 size_t OopStorage::Block::allocation_alignment_shift() {
241 return exact_log2(block_alignment);
242 }
243
244 inline bool is_full_bitmask(uintx bitmask) { return ~bitmask == 0; }
245 inline bool is_empty_bitmask(uintx bitmask) { return bitmask == 0; }
246
247 bool OopStorage::Block::is_full() const {
248 return is_full_bitmask(allocated_bitmask());
249 }
250
251 bool OopStorage::Block::is_empty() const {
252 return is_empty_bitmask(allocated_bitmask());
253 }
254
255 uintx OopStorage::Block::bitmask_for_entry(const oop* ptr) const {
256 return bitmask_for_index(get_index(ptr));
257 }
258
259 // An empty block is not yet deletable if either:
260 // (1) There is a release() operation currently operating on it.
261 // (2) It is in the deferred updates list.
262 // For interaction with release(), these must follow the empty check,
263 // and the order of these checks is important.
264 bool OopStorage::Block::is_safe_to_delete() const {
265 assert(is_empty(), "precondition");
266 OrderAccess::loadload();
267 return (OrderAccess::load_acquire(&_release_refcount) == 0) &&
268 (OrderAccess::load_acquire(&_deferred_updates_next) == NULL);
269 }
270
271 OopStorage::Block* OopStorage::Block::deferred_updates_next() const {
272 return _deferred_updates_next;
273 }
274
275 void OopStorage::Block::set_deferred_updates_next(Block* block) {
276 _deferred_updates_next = block;
277 }
278
279 bool OopStorage::Block::contains(const oop* ptr) const {
280 const oop* base = get_pointer(0);
281 return (base <= ptr) && (ptr < (base + ARRAY_SIZE(_data)));
282 }
283
284 size_t OopStorage::Block::active_index() const {
285 return _active_index;
286 }
287
288 void OopStorage::Block::set_active_index(size_t index) {
289 _active_index = index;
290 }
291
292 size_t OopStorage::Block::active_index_safe(const Block* block) {
293 STATIC_ASSERT(sizeof(intptr_t) == sizeof(block->_active_index));
294 assert(CanUseSafeFetchN(), "precondition");
295 return SafeFetchN((intptr_t*)&block->_active_index, 0);
296 }
297
298 unsigned OopStorage::Block::get_index(const oop* ptr) const {
299 assert(contains(ptr), PTR_FORMAT " not in block " PTR_FORMAT, p2i(ptr), p2i(this));
300 return static_cast<unsigned>(ptr - get_pointer(0));
301 }
302
303 oop* OopStorage::Block::allocate() {
304 // Use CAS loop because release may change bitmask outside of lock.
305 uintx allocated = allocated_bitmask();
306 while (true) {
307 assert(!is_full_bitmask(allocated), "attempt to allocate from full block");
308 unsigned index = count_trailing_zeros(~allocated);
309 uintx new_value = allocated | bitmask_for_index(index);
310 uintx fetched = Atomic::cmpxchg(new_value, &_allocated_bitmask, allocated);
311 if (fetched == allocated) {
312 return get_pointer(index); // CAS succeeded; return entry for index.
313 }
314 allocated = fetched; // CAS failed; retry with latest value.
315 }
316 }
317
318 OopStorage::Block* OopStorage::Block::new_block(const OopStorage* owner) {
319 // _data must be first member: aligning block => aligning _data.
320 STATIC_ASSERT(_data_pos == 0);
321 size_t size_needed = allocation_size();
322 void* memory = NEW_C_HEAP_ARRAY_RETURN_NULL(char, size_needed, mtGC);
323 if (memory == NULL) {
324 return NULL;
325 }
326 void* block_mem = align_up(memory, block_alignment);
327 assert(sizeof(Block) + pointer_delta(block_mem, memory, 1) <= size_needed,
328 "allocated insufficient space for aligned block");
329 return ::new (block_mem) Block(owner, memory);
330 }
331
332 void OopStorage::Block::delete_block(const Block& block) {
333 void* memory = block._memory;
334 block.Block::~Block();
335 FREE_C_HEAP_ARRAY(char, memory);
336 }
337
338 // This can return a false positive if ptr is not contained by some
339 // block. For some uses, it is a precondition that ptr is valid,
340 // e.g. contained in some block in owner's _active_array. Other uses
341 // require additional validation of the result.
342 OopStorage::Block*
343 OopStorage::Block::block_for_ptr(const OopStorage* owner, const oop* ptr) {
344 assert(CanUseSafeFetchN(), "precondition");
345 STATIC_ASSERT(_data_pos == 0);
346 // Const-ness of ptr is not related to const-ness of containing block.
347 // Blocks are allocated section-aligned, so get the containing section.
348 oop* section_start = align_down(const_cast<oop*>(ptr), block_alignment);
349 // Start with a guess that the containing section is the last section,
350 // so the block starts section_count-1 sections earlier.
351 oop* section = section_start - (section_size * (section_count - 1));
352 // Walk up through the potential block start positions, looking for
353 // the owner in the expected location. If we're below the actual block
354 // start position, the value at the owner position will be some oop
355 // (possibly NULL), which can never match the owner.
356 intptr_t owner_addr = reinterpret_cast<intptr_t>(owner);
357 for (unsigned i = 0; i < section_count; ++i, section += section_size) {
358 Block* candidate = reinterpret_cast<Block*>(section);
359 if (SafeFetchN(&candidate->_owner_address, 0) == owner_addr) {
360 return candidate;
361 }
362 }
363 return NULL;
364 }
365
366 //////////////////////////////////////////////////////////////////////////////
367 // Allocation
368 //
369 // Allocation involves the _allocation_list, which contains a subset of the
370 // blocks owned by a storage object. This is a doubly-linked list, linked
371 // through dedicated fields in the blocks. Full blocks are removed from this
372 // list, though they are still present in the _active_array. Empty blocks are
373 // kept at the end of the _allocation_list, to make it easy for empty block
374 // deletion to find them.
375 //
376 // allocate(), and delete_empty_blocks() lock the
377 // _allocation_mutex while performing any list and array modifications.
378 //
379 // allocate() and release() update a block's _allocated_bitmask using CAS
380 // loops. This prevents loss of updates even though release() performs
381 // its updates without any locking.
382 //
383 // allocate() obtains the entry from the first block in the _allocation_list,
384 // and updates that block's _allocated_bitmask to indicate the entry is in
385 // use. If this makes the block full (all entries in use), the block is
386 // removed from the _allocation_list so it won't be considered by future
387 // allocations until some entries in it are released.
388 //
389 // release() is performed lock-free. (Note: This means it can't notify the
390 // service thread of pending cleanup work. It must be lock-free because
391 // it is called in all kinds of contexts where even quite low ranked locks
392 // may be held.) release() first looks up the block for
393 // the entry, using address alignment to find the enclosing block (thereby
394 // avoiding iteration over the _active_array). Once the block has been
395 // determined, its _allocated_bitmask needs to be updated, and its position in
396 // the _allocation_list may need to be updated. There are two cases:
397 //
398 // (a) If the block is neither full nor would become empty with the release of
399 // the entry, only its _allocated_bitmask needs to be updated. But if the CAS
400 // update fails, the applicable case may change for the retry.
401 //
402 // (b) Otherwise, the _allocation_list also needs to be modified. This requires
403 // locking the _allocation_mutex. To keep the release() operation lock-free,
404 // rather than updating the _allocation_list itself, it instead performs a
405 // lock-free push of the block onto the _deferred_updates list. Entries on
406 // that list are processed by allocate() and delete_empty_blocks(), while
407 // they already hold the necessary lock. That processing makes the block's
408 // list state consistent with its current _allocated_bitmask. The block is
409 // added to the _allocation_list if not already present and the bitmask is not
410 // full. The block is moved to the end of the _allocation_list if the bitmask
411 // is empty, for ease of empty block deletion processing.
412
413 oop* OopStorage::allocate() {
414 MutexLocker ml(_allocation_mutex, Mutex::_no_safepoint_check_flag);
415
416 Block* block = block_for_allocation();
417 if (block == NULL) return NULL; // Block allocation failed.
418 assert(!block->is_full(), "invariant");
419 if (block->is_empty()) {
420 // Transitioning from empty to not empty.
421 log_trace(oopstorage, blocks)("%s: block not empty " PTR_FORMAT, name(), p2i(block));
422 }
423 oop* result = block->allocate();
424 assert(result != NULL, "allocation failed");
425 assert(!block->is_empty(), "postcondition");
426 Atomic::inc(&_allocation_count); // release updates outside lock.
427 if (block->is_full()) {
428 // Transitioning from not full to full.
429 // Remove full blocks from consideration by future allocates.
430 log_trace(oopstorage, blocks)("%s: block full " PTR_FORMAT, name(), p2i(block));
431 _allocation_list.unlink(*block);
432 }
433 log_trace(oopstorage, ref)("%s: allocated " PTR_FORMAT, name(), p2i(result));
434 return result;
435 }
436
437 bool OopStorage::try_add_block() {
438 assert_lock_strong(_allocation_mutex);
439 Block* block;
440 {
441 MutexUnlocker ul(_allocation_mutex, Mutex::_no_safepoint_check_flag);
442 block = Block::new_block(this);
443 }
444 if (block == NULL) return false;
445
446 // Add new block to the _active_array, growing if needed.
447 if (!_active_array->push(block)) {
448 if (expand_active_array()) {
449 guarantee(_active_array->push(block), "push failed after expansion");
450 } else {
451 log_debug(oopstorage, blocks)("%s: failed active array expand", name());
452 Block::delete_block(*block);
453 return false;
454 }
455 }
456 // Add to end of _allocation_list. The mutex release allowed other
457 // threads to add blocks to the _allocation_list. We prefer to
458 // allocate from non-empty blocks, to allow empty blocks to be
459 // deleted. But we don't bother notifying about the empty block
460 // because we're (probably) about to allocate an entry from it.
461 _allocation_list.push_back(*block);
462 log_debug(oopstorage, blocks)("%s: new block " PTR_FORMAT, name(), p2i(block));
463 return true;
464 }
465
466 OopStorage::Block* OopStorage::block_for_allocation() {
467 assert_lock_strong(_allocation_mutex);
468 while (true) {
469 // Use the first block in _allocation_list for the allocation.
470 Block* block = _allocation_list.head();
471 if (block != NULL) {
472 return block;
473 } else if (reduce_deferred_updates()) {
474 // Might have added a block to the _allocation_list, so retry.
475 } else if (try_add_block()) {
476 // Successfully added a new block to the list, so retry.
477 assert(_allocation_list.chead() != NULL, "invariant");
478 } else if (_allocation_list.chead() != NULL) {
479 // Trying to add a block failed, but some other thread added to the
480 // list while we'd dropped the lock over the new block allocation.
481 } else if (!reduce_deferred_updates()) { // Once more before failure.
482 // Attempt to add a block failed, no other thread added a block,
483 // and no deferred updated added a block, then allocation failed.
484 log_info(oopstorage, blocks)("%s: failed block allocation", name());
485 return NULL;
486 }
487 }
488 }
489
490 // Create a new, larger, active array with the same content as the
491 // current array, and then replace, relinquishing the old array.
492 // Return true if the array was successfully expanded, false to
493 // indicate allocation failure.
494 bool OopStorage::expand_active_array() {
495 assert_lock_strong(_allocation_mutex);
496 ActiveArray* old_array = _active_array;
497 size_t new_size = 2 * old_array->size();
498 log_debug(oopstorage, blocks)("%s: expand active array " SIZE_FORMAT,
499 name(), new_size);
500 ActiveArray* new_array = ActiveArray::create(new_size, AllocFailStrategy::RETURN_NULL);
501 if (new_array == NULL) return false;
502 new_array->copy_from(old_array);
503 replace_active_array(new_array);
504 relinquish_block_array(old_array);
505 return true;
506 }
507
508 // Make new_array the _active_array. Increments new_array's refcount
509 // to account for the new reference. The assignment is atomic wrto
510 // obtain_active_array; once this function returns, it is safe for the
511 // caller to relinquish the old array.
512 void OopStorage::replace_active_array(ActiveArray* new_array) {
513 // Caller has the old array that is the current value of _active_array.
514 // Update new_array refcount to account for the new reference.
515 new_array->increment_refcount();
516 // Install new_array, ensuring its initialization is complete first.
517 OrderAccess::release_store(&_active_array, new_array);
518 // Wait for any readers that could read the old array from _active_array.
519 // Can't use GlobalCounter here, because this is called from allocate(),
520 // which may be called in the scope of a GlobalCounter critical section
521 // when inserting a StringTable entry.
522 _protect_active.synchronize();
523 // All obtain critical sections that could see the old array have
524 // completed, having incremented the refcount of the old array. The
525 // caller can now safely relinquish the old array.
526 }
527
528 // Atomically (wrto replace_active_array) get the active array and
529 // increment its refcount. This provides safe access to the array,
530 // even if an allocate operation expands and replaces the value of
531 // _active_array. The caller must relinquish the array when done
532 // using it.
533 OopStorage::ActiveArray* OopStorage::obtain_active_array() const {
534 SingleWriterSynchronizer::CriticalSection cs(&_protect_active);
535 ActiveArray* result = OrderAccess::load_acquire(&_active_array);
536 result->increment_refcount();
537 return result;
538 }
539
540 // Decrement refcount of array and destroy if refcount is zero.
541 void OopStorage::relinquish_block_array(ActiveArray* array) const {
542 if (array->decrement_refcount()) {
543 assert(array != _active_array, "invariant");
544 ActiveArray::destroy(array);
545 }
546 }
547
548 class OopStorage::WithActiveArray : public StackObj {
549 const OopStorage* _storage;
550 ActiveArray* _active_array;
551
552 public:
553 WithActiveArray(const OopStorage* storage) :
554 _storage(storage),
555 _active_array(storage->obtain_active_array())
556 {}
557
558 ~WithActiveArray() {
559 _storage->relinquish_block_array(_active_array);
560 }
561
562 ActiveArray& active_array() const {
563 return *_active_array;
564 }
565 };
566
567 OopStorage::Block* OopStorage::find_block_or_null(const oop* ptr) const {
568 assert(ptr != NULL, "precondition");
569 return Block::block_for_ptr(this, ptr);
570 }
571
572 static void log_release_transitions(uintx releasing,
573 uintx old_allocated,
574 const OopStorage* owner,
575 const void* block) {
576 LogTarget(Trace, oopstorage, blocks) lt;
577 if (lt.is_enabled()) {
578 LogStream ls(lt);
579 if (is_full_bitmask(old_allocated)) {
580 ls.print_cr("%s: block not full " PTR_FORMAT, owner->name(), p2i(block));
581 }
582 if (releasing == old_allocated) {
583 ls.print_cr("%s: block empty " PTR_FORMAT, owner->name(), p2i(block));
584 }
585 }
586 }
587
588 void OopStorage::Block::release_entries(uintx releasing, OopStorage* owner) {
589 assert(releasing != 0, "preconditon");
590 // Prevent empty block deletion when transitioning to empty.
591 Atomic::inc(&_release_refcount);
592
593 // Atomically update allocated bitmask.
594 uintx old_allocated = _allocated_bitmask;
595 while (true) {
596 assert((releasing & ~old_allocated) == 0, "releasing unallocated entries");
597 uintx new_value = old_allocated ^ releasing;
598 uintx fetched = Atomic::cmpxchg(new_value, &_allocated_bitmask, old_allocated);
599 if (fetched == old_allocated) break; // Successful update.
600 old_allocated = fetched; // Retry with updated bitmask.
601 }
602
603 // Now that the bitmask has been updated, if we have a state transition
604 // (updated bitmask is empty or old bitmask was full), atomically push
605 // this block onto the deferred updates list. Some future call to
606 // reduce_deferred_updates will make any needed changes related to this
607 // block and _allocation_list. This deferral avoids _allocation_list
608 // updates and the associated locking here.
609 if ((releasing == old_allocated) || is_full_bitmask(old_allocated)) {
610 // Log transitions. Both transitions are possible in a single update.
611 log_release_transitions(releasing, old_allocated, owner, this);
612 // Attempt to claim responsibility for adding this block to the deferred
613 // list, by setting the link to non-NULL by self-looping. If this fails,
614 // then someone else has made such a claim and the deferred update has not
615 // yet been processed and will include our change, so we don't need to do
616 // anything further.
617 if (Atomic::replace_if_null(this, &_deferred_updates_next)) {
618 // Successfully claimed. Push, with self-loop for end-of-list.
619 Block* head = owner->_deferred_updates;
620 while (true) {
621 _deferred_updates_next = (head == NULL) ? this : head;
622 Block* fetched = Atomic::cmpxchg(this, &owner->_deferred_updates, head);
623 if (fetched == head) break; // Successful update.
624 head = fetched; // Retry with updated head.
625 }
626 // Only request cleanup for to-empty transitions, not for from-full.
627 // There isn't any rush to process from-full transitions. Allocation
628 // will reduce deferrals before allocating new blocks, so may process
629 // some. And the service thread will drain the entire deferred list
630 // if there are any pending to-empty transitions.
631 if (releasing == old_allocated) {
632 owner->record_needs_cleanup();
633 }
634 log_trace(oopstorage, blocks)("%s: deferred update " PTR_FORMAT,
635 owner->name(), p2i(this));
636 }
637 }
638 // Release hold on empty block deletion.
639 Atomic::dec(&_release_refcount);
640 }
641
642 // Process one available deferred update. Returns true if one was processed.
643 bool OopStorage::reduce_deferred_updates() {
644 assert_lock_strong(_allocation_mutex);
645 // Atomically pop a block off the list, if any available.
646 // No ABA issue because this is only called by one thread at a time.
647 // The atomicity is wrto pushes by release().
648 Block* block = OrderAccess::load_acquire(&_deferred_updates);
649 while (true) {
650 if (block == NULL) return false;
651 // Try atomic pop of block from list.
652 Block* tail = block->deferred_updates_next();
653 if (block == tail) tail = NULL; // Handle self-loop end marker.
654 Block* fetched = Atomic::cmpxchg(tail, &_deferred_updates, block);
655 if (fetched == block) break; // Update successful.
656 block = fetched; // Retry with updated block.
657 }
658 block->set_deferred_updates_next(NULL); // Clear tail after updating head.
659 // Ensure bitmask read after pop is complete, including clearing tail, for
660 // ordering with release(). Without this, we may be processing a stale
661 // bitmask state here while blocking a release() operation from recording
662 // the deferred update needed for its bitmask change.
663 OrderAccess::fence();
664 // Process popped block.
665 uintx allocated = block->allocated_bitmask();
666
667 // Make membership in list consistent with bitmask state.
668 if ((_allocation_list.ctail() != NULL) &&
669 ((_allocation_list.ctail() == block) ||
670 (_allocation_list.next(*block) != NULL))) {
671 // Block is in the _allocation_list.
672 assert(!is_full_bitmask(allocated), "invariant");
673 } else if (!is_full_bitmask(allocated)) {
674 // Block is not in the _allocation_list, but now should be.
675 _allocation_list.push_front(*block);
676 } // Else block is full and not in list, which is correct.
677
678 // Move empty block to end of list, for possible deletion.
679 if (is_empty_bitmask(allocated)) {
680 _allocation_list.unlink(*block);
681 _allocation_list.push_back(*block);
682 }
683
684 log_trace(oopstorage, blocks)("%s: processed deferred update " PTR_FORMAT,
685 name(), p2i(block));
686 return true; // Processed one pending update.
687 }
688
689 inline void check_release_entry(const oop* entry) {
690 assert(entry != NULL, "Releasing NULL");
691 assert(*entry == NULL, "Releasing uncleared entry: " PTR_FORMAT, p2i(entry));
692 }
693
694 void OopStorage::release(const oop* ptr) {
695 check_release_entry(ptr);
696 Block* block = find_block_or_null(ptr);
697 assert(block != NULL, "%s: invalid release " PTR_FORMAT, name(), p2i(ptr));
698 log_trace(oopstorage, ref)("%s: released " PTR_FORMAT, name(), p2i(ptr));
699 block->release_entries(block->bitmask_for_entry(ptr), this);
700 Atomic::dec(&_allocation_count);
701 }
702
703 void OopStorage::release(const oop* const* ptrs, size_t size) {
704 size_t i = 0;
705 while (i < size) {
706 check_release_entry(ptrs[i]);
707 Block* block = find_block_or_null(ptrs[i]);
708 assert(block != NULL, "%s: invalid release " PTR_FORMAT, name(), p2i(ptrs[i]));
709 log_trace(oopstorage, ref)("%s: released " PTR_FORMAT, name(), p2i(ptrs[i]));
710 size_t count = 0;
711 uintx releasing = 0;
712 for ( ; i < size; ++i) {
713 const oop* entry = ptrs[i];
714 check_release_entry(entry);
715 // If entry not in block, finish block and resume outer loop with entry.
716 if (!block->contains(entry)) break;
717 // Add entry to releasing bitmap.
718 log_trace(oopstorage, ref)("%s: released " PTR_FORMAT, name(), p2i(entry));
719 uintx entry_bitmask = block->bitmask_for_entry(entry);
720 assert((releasing & entry_bitmask) == 0,
721 "Duplicate entry: " PTR_FORMAT, p2i(entry));
722 releasing |= entry_bitmask;
723 ++count;
724 }
725 // Release the contiguous entries that are in block.
726 block->release_entries(releasing, this);
727 Atomic::sub(count, &_allocation_count);
728 }
729 }
730
731 const size_t initial_active_array_size = 8;
732
733 OopStorage::OopStorage(const char* name,
734 Mutex* allocation_mutex,
735 Mutex* active_mutex) :
736 _name(os::strdup(name)),
737 _active_array(ActiveArray::create(initial_active_array_size)),
738 _allocation_list(),
739 _deferred_updates(NULL),
740 _allocation_mutex(allocation_mutex),
741 _active_mutex(active_mutex),
742 _allocation_count(0),
743 _concurrent_iteration_count(0),
744 _needs_cleanup(false)
745 {
746 _active_array->increment_refcount();
747 assert(_active_mutex->rank() < _allocation_mutex->rank(),
748 "%s: active_mutex must have lower rank than allocation_mutex", _name);
749 assert(Service_lock->rank() < _active_mutex->rank(),
750 "%s: active_mutex must have higher rank than Service_lock", _name);
751 assert(_active_mutex->_safepoint_check_required == Mutex::_safepoint_check_never,
752 "%s: active mutex requires never safepoint check", _name);
753 assert(_allocation_mutex->_safepoint_check_required == Mutex::_safepoint_check_never,
754 "%s: allocation mutex requires never safepoint check", _name);
755 }
756
757 void OopStorage::delete_empty_block(const Block& block) {
758 assert(block.is_empty(), "discarding non-empty block");
759 log_debug(oopstorage, blocks)("%s: delete empty block " PTR_FORMAT, name(), p2i(&block));
760 Block::delete_block(block);
761 }
762
763 OopStorage::~OopStorage() {
764 Block* block;
765 while ((block = _deferred_updates) != NULL) {
766 _deferred_updates = block->deferred_updates_next();
767 block->set_deferred_updates_next(NULL);
768 }
769 while ((block = _allocation_list.head()) != NULL) {
770 _allocation_list.unlink(*block);
771 }
772 bool unreferenced = _active_array->decrement_refcount();
773 assert(unreferenced, "deleting storage while _active_array is referenced");
774 for (size_t i = _active_array->block_count(); 0 < i; ) {
775 block = _active_array->at(--i);
776 Block::delete_block(*block);
777 }
778 ActiveArray::destroy(_active_array);
779 os::free(const_cast<char*>(_name));
780 }
781
782 // Managing service thread notifications.
783 //
784 // We don't want cleanup work to linger indefinitely, but we also don't want
785 // to run the service thread too often. We're also very limited in what we
786 // can do in a release operation, where cleanup work is created.
787 //
788 // When a release operation changes a block's state to empty, it records the
789 // need for cleanup in both the associated storage object and in the global
790 // request state. A safepoint cleanup task notifies the service thread when
791 // there may be cleanup work for any storage object, based on the global
792 // request state. But that notification is deferred if the service thread
793 // has run recently, and we also avoid duplicate notifications. The service
794 // thread updates the timestamp and resets the state flags on every iteration.
795
796 // Global cleanup request state.
797 static volatile bool needs_cleanup_requested = false;
798
799 // Flag for avoiding duplicate notifications.
800 static bool needs_cleanup_triggered = false;
801
802 // Time after which a notification can be made.
803 static jlong cleanup_trigger_permit_time = 0;
804
805 // Minimum time since last service thread check before notification is
806 // permitted. The value of 500ms was an arbitrary choice; frequent, but not
807 // too frequent.
808 const jlong cleanup_trigger_defer_period = 500 * NANOSECS_PER_MILLISEC;
809
810 void OopStorage::trigger_cleanup_if_needed() {
811 MonitorLocker ml(Service_lock, Monitor::_no_safepoint_check_flag);
812 if (Atomic::load(&needs_cleanup_requested) &&
813 !needs_cleanup_triggered &&
814 (os::javaTimeNanos() > cleanup_trigger_permit_time)) {
815 needs_cleanup_triggered = true;
816 ml.notify_all();
817 }
818 }
819
820 bool OopStorage::has_cleanup_work_and_reset() {
821 assert_lock_strong(Service_lock);
822 cleanup_trigger_permit_time =
823 os::javaTimeNanos() + cleanup_trigger_defer_period;
824 needs_cleanup_triggered = false;
825 // Set the request flag false and return its old value.
826 // Needs to be atomic to avoid dropping a concurrent request.
827 // Can't use Atomic::xchg, which may not support bool.
828 return Atomic::cmpxchg(false, &needs_cleanup_requested, true);
829 }
830
831 // Record that cleanup is needed, without notifying the Service thread.
832 // Used by release(), where we can't lock even Service_lock.
833 void OopStorage::record_needs_cleanup() {
834 // Set local flag first, else service thread could wake up and miss
835 // the request. This order may instead (rarely) unnecessarily notify.
836 OrderAccess::release_store(&_needs_cleanup, true);
837 OrderAccess::release_store_fence(&needs_cleanup_requested, true);
838 }
839
840 bool OopStorage::delete_empty_blocks() {
841 // Service thread might have oopstorage work, but not for this object.
842 // Check for deferred updates even though that's not a service thread
843 // trigger; since we're here, we might as well process them.
844 if (!OrderAccess::load_acquire(&_needs_cleanup) &&
845 (OrderAccess::load_acquire(&_deferred_updates) == NULL)) {
846 return false;
847 }
848
849 MutexLocker ml(_allocation_mutex, Mutex::_no_safepoint_check_flag);
850
851 // Clear the request before processing.
852 OrderAccess::release_store_fence(&_needs_cleanup, false);
853
854 // Other threads could be adding to the empty block count or the
855 // deferred update list while we're working. Set an upper bound on
856 // how many updates we'll process and blocks we'll try to release,
857 // so other threads can't cause an unbounded stay in this function.
858 // We add a bit of slop because the reduce_deferred_updates clause
859 // can cause blocks to be double counted. If there are few blocks
860 // and many of them are deferred and empty, we might hit the limit
861 // and spin the caller without doing very much work. Otherwise,
862 // we don't normally hit the limit anyway, instead running out of
863 // work to do.
864 size_t limit = block_count() + 10;
865
866 for (size_t i = 0; i < limit; ++i) {
867 // Process deferred updates, which might make empty blocks available.
868 // Continue checking once deletion starts, since additional updates
869 // might become available while we're working.
870 if (reduce_deferred_updates()) {
871 // Be safepoint-polite while looping.
872 MutexUnlocker ul(_allocation_mutex, Mutex::_no_safepoint_check_flag);
873 ThreadBlockInVM tbiv(JavaThread::current());
874 } else {
875 Block* block = _allocation_list.tail();
876 if ((block == NULL) || !block->is_empty()) {
877 return false;
878 } else if (!block->is_safe_to_delete()) {
879 // Look for other work while waiting for block to be deletable.
880 break;
881 }
882
883 // Try to delete the block. First, try to remove from _active_array.
884 {
885 MutexLocker aml(_active_mutex, Mutex::_no_safepoint_check_flag);
886 // Don't interfere with an active concurrent iteration.
887 // Instead, give up immediately. There is more work to do,
888 // but don't re-notify, to avoid useless spinning of the
889 // service thread. Instead, iteration completion notifies.
890 if (_concurrent_iteration_count > 0) return true;
891 _active_array->remove(block);
892 }
893 // Remove block from _allocation_list and delete it.
894 _allocation_list.unlink(*block);
895 // Be safepoint-polite while deleting and looping.
896 MutexUnlocker ul(_allocation_mutex, Mutex::_no_safepoint_check_flag);
897 delete_empty_block(*block);
898 ThreadBlockInVM tbiv(JavaThread::current());
899 }
900 }
901 // Exceeded work limit or can't delete last block. This will
902 // cause the service thread to loop, giving other subtasks an
903 // opportunity to run too. There's no need for a notification,
904 // because we are part of the service thread (unless gtesting).
905 record_needs_cleanup();
906 return true;
907 }
908
909 OopStorage::EntryStatus OopStorage::allocation_status(const oop* ptr) const {
910 const Block* block = find_block_or_null(ptr);
911 if (block != NULL) {
912 // Prevent block deletion and _active_array modification.
913 MutexLocker ml(_allocation_mutex, Mutex::_no_safepoint_check_flag);
914 // Block could be a false positive, so get index carefully.
915 size_t index = Block::active_index_safe(block);
916 if ((index < _active_array->block_count()) &&
917 (block == _active_array->at(index)) &&
918 block->contains(ptr)) {
919 if ((block->allocated_bitmask() & block->bitmask_for_entry(ptr)) != 0) {
920 return ALLOCATED_ENTRY;
921 } else {
922 return UNALLOCATED_ENTRY;
923 }
924 }
925 }
926 return INVALID_ENTRY;
927 }
928
929 size_t OopStorage::allocation_count() const {
930 return _allocation_count;
931 }
932
933 size_t OopStorage::block_count() const {
934 WithActiveArray wab(this);
935 // Count access is racy, but don't care.
936 return wab.active_array().block_count();
937 }
938
939 size_t OopStorage::total_memory_usage() const {
940 size_t total_size = sizeof(OopStorage);
941 total_size += strlen(name()) + 1;
942 total_size += sizeof(ActiveArray);
943 WithActiveArray wab(this);
944 const ActiveArray& blocks = wab.active_array();
945 // Count access is racy, but don't care.
946 total_size += blocks.block_count() * Block::allocation_size();
947 total_size += blocks.size() * sizeof(Block*);
948 return total_size;
949 }
950
951 // Parallel iteration support
952
953 uint OopStorage::BasicParState::default_estimated_thread_count(bool concurrent) {
954 uint configured = concurrent ? ConcGCThreads : ParallelGCThreads;
955 return MAX2(1u, configured); // Never estimate zero threads.
956 }
957
958 OopStorage::BasicParState::BasicParState(const OopStorage* storage,
959 uint estimated_thread_count,
960 bool concurrent) :
961 _storage(storage),
962 _active_array(_storage->obtain_active_array()),
963 _block_count(0), // initialized properly below
964 _next_block(0),
965 _estimated_thread_count(estimated_thread_count),
966 _concurrent(concurrent)
967 {
968 assert(estimated_thread_count > 0, "estimated thread count must be positive");
969 update_concurrent_iteration_count(1);
970 // Get the block count *after* iteration state updated, so concurrent
971 // empty block deletion is suppressed and can't reduce the count. But
972 // ensure the count we use was written after the block with that count
973 // was fully initialized; see ActiveArray::push.
974 _block_count = _active_array->block_count_acquire();
975 }
976
977 OopStorage::BasicParState::~BasicParState() {
978 _storage->relinquish_block_array(_active_array);
979 update_concurrent_iteration_count(-1);
980 if (_concurrent) {
981 // We may have deferred some cleanup work.
982 const_cast<OopStorage*>(_storage)->record_needs_cleanup();
983 }
984 }
985
986 void OopStorage::BasicParState::update_concurrent_iteration_count(int value) {
987 if (_concurrent) {
988 MutexLocker ml(_storage->_active_mutex, Mutex::_no_safepoint_check_flag);
989 _storage->_concurrent_iteration_count += value;
990 assert(_storage->_concurrent_iteration_count >= 0, "invariant");
991 }
992 }
993
994 bool OopStorage::BasicParState::claim_next_segment(IterationData* data) {
995 data->_processed += data->_segment_end - data->_segment_start;
996 size_t start = OrderAccess::load_acquire(&_next_block);
997 if (start >= _block_count) {
998 return finish_iteration(data); // No more blocks available.
999 }
1000 // Try to claim several at a time, but not *too* many. We want to
1001 // avoid deciding there are many available and selecting a large
1002 // quantity, get delayed, and then end up claiming most or all of
1003 // the remaining largish amount of work, leaving nothing for other
1004 // threads to do. But too small a step can lead to contention
1005 // over _next_block, esp. when the work per block is small.
1006 size_t max_step = 10;
1007 size_t remaining = _block_count - start;
1008 size_t step = MIN2(max_step, 1 + (remaining / _estimated_thread_count));
1009 // Atomic::add with possible overshoot. This can perform better
1010 // than a CAS loop on some platforms when there is contention.
1011 // We can cope with the uncertainty by recomputing start/end from
1012 // the result of the add, and dealing with potential overshoot.
1013 size_t end = Atomic::add(step, &_next_block);
1014 // _next_block may have changed, so recompute start from result of add.
1015 start = end - step;
1016 // _next_block may have changed so much that end has overshot.
1017 end = MIN2(end, _block_count);
1018 // _next_block may have changed so much that even start has overshot.
1019 if (start < _block_count) {
1020 // Record claimed segment for iteration.
1021 data->_segment_start = start;
1022 data->_segment_end = end;
1023 return true; // Success.
1024 } else {
1025 // No more blocks to claim.
1026 return finish_iteration(data);
1027 }
1028 }
1029
1030 bool OopStorage::BasicParState::finish_iteration(const IterationData* data) const {
1031 log_info(oopstorage, blocks, stats)
1032 ("Parallel iteration on %s: blocks = " SIZE_FORMAT
1033 ", processed = " SIZE_FORMAT " (%2.f%%)",
1034 _storage->name(), _block_count, data->_processed,
1035 percent_of(data->_processed, _block_count));
1036 return false;
1037 }
1038
1039 const char* OopStorage::name() const { return _name; }
1040
1041 #ifndef PRODUCT
1042
1043 void OopStorage::print_on(outputStream* st) const {
1044 size_t allocations = _allocation_count;
1045 size_t blocks = _active_array->block_count();
1046
1047 double data_size = section_size * section_count;
1048 double alloc_percentage = percent_of((double)allocations, blocks * data_size);
1049
1050 st->print("%s: " SIZE_FORMAT " entries in " SIZE_FORMAT " blocks (%.F%%), " SIZE_FORMAT " bytes",
1051 name(), allocations, blocks, alloc_percentage, total_memory_usage());
1052 if (_concurrent_iteration_count > 0) {
1053 st->print(", concurrent iteration active");
1054 }
1055 }
1056
1057 #endif // !PRODUCT