1 | /* -*- mode: C++; c-basic-offset: 4; indent-tabs-mode: nil -*- */ |
2 | // vim: ft=cpp:expandtab:ts=8:sw=4:softtabstop=4: |
3 | #ident "$Id$" |
4 | /*====== |
5 | This file is part of PerconaFT. |
6 | |
7 | |
8 | Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved. |
9 | |
10 | PerconaFT is free software: you can redistribute it and/or modify |
11 | it under the terms of the GNU General Public License, version 2, |
12 | as published by the Free Software Foundation. |
13 | |
14 | PerconaFT is distributed in the hope that it will be useful, |
15 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
16 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
17 | GNU General Public License for more details. |
18 | |
19 | You should have received a copy of the GNU General Public License |
20 | along with PerconaFT. If not, see <http://www.gnu.org/licenses/>. |
21 | |
22 | ---------------------------------------- |
23 | |
24 | PerconaFT is free software: you can redistribute it and/or modify |
25 | it under the terms of the GNU Affero General Public License, version 3, |
26 | as published by the Free Software Foundation. |
27 | |
28 | PerconaFT is distributed in the hope that it will be useful, |
29 | but WITHOUT ANY WARRANTY; without even the implied warranty of |
30 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
31 | GNU Affero General Public License for more details. |
32 | |
33 | You should have received a copy of the GNU Affero General Public License |
34 | along with PerconaFT. If not, see <http://www.gnu.org/licenses/>. |
35 | ======= */ |
36 | |
37 | #ident "Copyright (c) 2006, 2015, Percona and/or its affiliates. All rights reserved." |
38 | |
39 | #include <my_global.h> |
40 | #include "portability/memory.h" |
41 | #include "portability/toku_assert.h" |
42 | #include "portability/toku_portability.h" |
43 | #include "portability/toku_pthread.h" |
44 | |
45 | // ugly but pragmatic, need access to dirty bits while holding translation lock |
46 | // TODO: Refactor this (possibly with FT-301) |
47 | #include "ft/ft-internal.h" |
48 | |
49 | // TODO: reorganize this dependency (FT-303) |
50 | #include "ft/ft-ops.h" // for toku_maybe_truncate_file |
51 | #include "ft/serialize/block_table.h" |
52 | #include "ft/serialize/rbuf.h" |
53 | #include "ft/serialize/wbuf.h" |
54 | #include "ft/serialize/block_allocator.h" |
55 | #include "util/nb_mutex.h" |
56 | #include "util/scoped_malloc.h" |
57 | |
58 | |
59 | toku_instr_key *block_table_mutex_key; |
60 | toku_instr_key *safe_file_size_lock_mutex_key; |
61 | toku_instr_key *safe_file_size_lock_rwlock_key; |
62 | |
63 | // indicates the end of a freelist |
64 | static const BLOCKNUM freelist_null = {-1}; |
65 | |
66 | // value of block_translation_pair.size if blocknum is unused |
67 | static const DISKOFF size_is_free = (DISKOFF)-1; |
68 | |
69 | // value of block_translation_pair.u.diskoff if blocknum is used but does not |
70 | // yet have a diskblock |
71 | static const DISKOFF diskoff_unused = (DISKOFF)-2; |
72 | |
73 | void block_table::_mutex_lock() { toku_mutex_lock(&_mutex); } |
74 | |
75 | void block_table::_mutex_unlock() { toku_mutex_unlock(&_mutex); } |
76 | |
77 | // TODO: Move lock to FT |
78 | void toku_ft_lock(FT ft) { |
79 | block_table *bt = &ft->blocktable; |
80 | bt->_mutex_lock(); |
81 | } |
82 | |
83 | // TODO: Move lock to FT |
84 | void toku_ft_unlock(FT ft) { |
85 | block_table *bt = &ft->blocktable; |
86 | toku_mutex_assert_locked(&bt->_mutex); |
87 | bt->_mutex_unlock(); |
88 | } |
89 | |
90 | // There are two headers: the reserve must fit them both and be suitably |
91 | // aligned. |
92 | static_assert(BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE % |
93 | BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT == |
94 | 0, |
95 | "Block allocator's header reserve must be suitibly aligned" ); |
96 | static_assert( |
97 | BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE * 2 == |
98 | BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE, |
99 | "Block allocator's total header reserve must exactly fit two headers" ); |
100 | |
101 | // does NOT initialize the block allocator: the caller is responsible |
102 | void block_table::_create_internal() { |
103 | memset(&_current, 0, sizeof(struct translation)); |
104 | memset(&_inprogress, 0, sizeof(struct translation)); |
105 | memset(&_checkpointed, 0, sizeof(struct translation)); |
106 | memset(&_mutex, 0, sizeof(_mutex)); |
107 | _bt_block_allocator = new BlockAllocator(); |
108 | toku_mutex_init(*block_table_mutex_key, &_mutex, nullptr); |
109 | nb_mutex_init(*safe_file_size_lock_mutex_key, |
110 | *safe_file_size_lock_rwlock_key, |
111 | &_safe_file_size_lock); |
112 | } |
113 | |
114 | // Fill in the checkpointed translation from buffer, and copy checkpointed to |
115 | // current. |
116 | // The one read from disk is the last known checkpointed one, so we are keeping |
117 | // it in |
118 | // place and then setting current (which is never stored on disk) for current |
119 | // use. |
120 | // The translation_buffer has translation only, we create the rest of the |
121 | // block_table. |
122 | int block_table::create_from_buffer( |
123 | int fd, |
124 | DISKOFF location_on_disk, // Location of translation_buffer |
125 | DISKOFF size_on_disk, |
126 | unsigned char *translation_buffer) { |
127 | // Does not initialize the block allocator |
128 | _create_internal(); |
129 | |
130 | // Deserialize the translation and copy it to current |
131 | int r = _translation_deserialize_from_buffer( |
132 | &_checkpointed, location_on_disk, size_on_disk, translation_buffer); |
133 | if (r != 0) { |
134 | return r; |
135 | } |
136 | _copy_translation(&_current, &_checkpointed, TRANSLATION_CURRENT); |
137 | |
138 | // Determine the file size |
139 | int64_t file_size = 0; |
140 | r = toku_os_get_file_size(fd, &file_size); |
141 | lazy_assert_zero(r); |
142 | invariant(file_size >= 0); |
143 | _safe_file_size = file_size; |
144 | |
145 | // Gather the non-empty translations and use them to create the block |
146 | // allocator |
147 | toku::scoped_malloc pairs_buf(_checkpointed.smallest_never_used_blocknum.b * |
148 | sizeof(struct BlockAllocator::BlockPair)); |
149 | struct BlockAllocator::BlockPair *CAST_FROM_VOIDP(pairs, pairs_buf.get()); |
150 | uint64_t n_pairs = 0; |
151 | for (int64_t i = 0; i < _checkpointed.smallest_never_used_blocknum.b; i++) { |
152 | struct block_translation_pair pair = _checkpointed.block_translation[i]; |
153 | if (pair.size > 0) { |
154 | invariant(pair.u.diskoff != diskoff_unused); |
155 | pairs[n_pairs++] = |
156 | BlockAllocator::BlockPair(pair.u.diskoff, pair.size); |
157 | } |
158 | } |
159 | |
160 | _bt_block_allocator->CreateFromBlockPairs( |
161 | BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE, |
162 | BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT, |
163 | pairs, |
164 | n_pairs); |
165 | |
166 | return 0; |
167 | } |
168 | |
169 | void block_table::create() { |
170 | // Does not initialize the block allocator |
171 | _create_internal(); |
172 | |
173 | _checkpointed.type = TRANSLATION_CHECKPOINTED; |
174 | _checkpointed.smallest_never_used_blocknum = |
175 | make_blocknum(RESERVED_BLOCKNUMS); |
176 | _checkpointed.length_of_array = |
177 | _checkpointed.smallest_never_used_blocknum.b; |
178 | _checkpointed.blocknum_freelist_head = freelist_null; |
179 | XMALLOC_N(_checkpointed.length_of_array, _checkpointed.block_translation); |
180 | for (int64_t i = 0; i < _checkpointed.length_of_array; i++) { |
181 | _checkpointed.block_translation[i].size = 0; |
182 | _checkpointed.block_translation[i].u.diskoff = diskoff_unused; |
183 | } |
184 | |
185 | // we just created a default checkpointed, now copy it to current. |
186 | _copy_translation(&_current, &_checkpointed, TRANSLATION_CURRENT); |
187 | |
188 | // Create an empty block allocator. |
189 | _bt_block_allocator->Create( |
190 | BlockAllocator::BLOCK_ALLOCATOR_TOTAL_HEADER_RESERVE, |
191 | BlockAllocator::BLOCK_ALLOCATOR_ALIGNMENT); |
192 | } |
193 | |
194 | // TODO: Refactor with FT-303 |
195 | static void ft_set_dirty(FT ft, bool for_checkpoint) { |
196 | invariant(ft->h->type == FT_CURRENT); |
197 | if (for_checkpoint) { |
198 | invariant(ft->checkpoint_header->type == FT_CHECKPOINT_INPROGRESS); |
199 | ft->checkpoint_header->dirty = 1; |
200 | } else { |
201 | ft->h->dirty = 1; |
202 | } |
203 | } |
204 | |
205 | void block_table::_maybe_truncate_file(int fd, uint64_t size_needed_before) { |
206 | toku_mutex_assert_locked(&_mutex); |
207 | uint64_t new_size_needed = _bt_block_allocator->AllocatedLimit(); |
208 | // Save a call to toku_os_get_file_size (kernel call) if unlikely to be |
209 | // useful. |
210 | if (new_size_needed < size_needed_before && |
211 | new_size_needed < _safe_file_size) { |
212 | nb_mutex_lock(&_safe_file_size_lock, &_mutex); |
213 | |
214 | // Must hold _safe_file_size_lock to change _safe_file_size. |
215 | if (new_size_needed < _safe_file_size) { |
216 | int64_t safe_file_size_before = _safe_file_size; |
217 | // Not safe to use the 'to-be-truncated' portion until truncate is |
218 | // done. |
219 | _safe_file_size = new_size_needed; |
220 | _mutex_unlock(); |
221 | |
222 | uint64_t size_after; |
223 | toku_maybe_truncate_file( |
224 | fd, new_size_needed, safe_file_size_before, &size_after); |
225 | _mutex_lock(); |
226 | |
227 | _safe_file_size = size_after; |
228 | } |
229 | nb_mutex_unlock(&_safe_file_size_lock); |
230 | } |
231 | } |
232 | |
233 | void block_table::maybe_truncate_file_on_open(int fd) { |
234 | _mutex_lock(); |
235 | _maybe_truncate_file(fd, _safe_file_size); |
236 | _mutex_unlock(); |
237 | } |
238 | |
239 | void block_table::_copy_translation(struct translation *dst, |
240 | struct translation *src, |
241 | enum translation_type newtype) { |
242 | // We intend to malloc a fresh block, so the incoming translation should be |
243 | // empty |
244 | invariant_null(dst->block_translation); |
245 | |
246 | invariant(src->length_of_array >= src->smallest_never_used_blocknum.b); |
247 | invariant(newtype == TRANSLATION_DEBUG || |
248 | (src->type == TRANSLATION_CURRENT && |
249 | newtype == TRANSLATION_INPROGRESS) || |
250 | (src->type == TRANSLATION_CHECKPOINTED && |
251 | newtype == TRANSLATION_CURRENT)); |
252 | dst->type = newtype; |
253 | dst->smallest_never_used_blocknum = src->smallest_never_used_blocknum; |
254 | dst->blocknum_freelist_head = src->blocknum_freelist_head; |
255 | |
256 | // destination btt is of fixed size. Allocate + memcpy the exact length |
257 | // necessary. |
258 | dst->length_of_array = dst->smallest_never_used_blocknum.b; |
259 | XMALLOC_N(dst->length_of_array, dst->block_translation); |
260 | memcpy(dst->block_translation, |
261 | src->block_translation, |
262 | dst->length_of_array * sizeof(*dst->block_translation)); |
263 | |
264 | // New version of btt is not yet stored on disk. |
265 | dst->block_translation[RESERVED_BLOCKNUM_TRANSLATION].size = 0; |
266 | dst->block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff = |
267 | diskoff_unused; |
268 | } |
269 | |
270 | int64_t block_table::get_blocks_in_use_unlocked() { |
271 | BLOCKNUM b; |
272 | struct translation *t = &_current; |
273 | int64_t num_blocks = 0; |
274 | { |
275 | // Reserved blocknums do not get upgraded; They are part of the header. |
276 | for (b.b = RESERVED_BLOCKNUMS; b.b < t->smallest_never_used_blocknum.b; |
277 | b.b++) { |
278 | if (t->block_translation[b.b].size != size_is_free) { |
279 | num_blocks++; |
280 | } |
281 | } |
282 | } |
283 | return num_blocks; |
284 | } |
285 | |
286 | void block_table::_maybe_optimize_translation(struct translation *t) { |
287 | // Reduce 'smallest_never_used_blocknum.b' (completely free blocknums |
288 | // instead of just |
289 | // on a free list. Doing so requires us to regenerate the free list. |
290 | // This is O(n) work, so do it only if you're already doing that. |
291 | |
292 | BLOCKNUM b; |
293 | paranoid_invariant(t->smallest_never_used_blocknum.b >= RESERVED_BLOCKNUMS); |
294 | // Calculate how large the free suffix is. |
295 | int64_t freed; |
296 | { |
297 | for (b.b = t->smallest_never_used_blocknum.b; b.b > RESERVED_BLOCKNUMS; |
298 | b.b--) { |
299 | if (t->block_translation[b.b - 1].size != size_is_free) { |
300 | break; |
301 | } |
302 | } |
303 | freed = t->smallest_never_used_blocknum.b - b.b; |
304 | } |
305 | if (freed > 0) { |
306 | t->smallest_never_used_blocknum.b = b.b; |
307 | if (t->length_of_array / 4 > t->smallest_never_used_blocknum.b) { |
308 | // We're using more memory than necessary to represent this now. |
309 | // Reduce. |
310 | uint64_t new_length = t->smallest_never_used_blocknum.b * 2; |
311 | XREALLOC_N(new_length, t->block_translation); |
312 | t->length_of_array = new_length; |
313 | // No need to zero anything out. |
314 | } |
315 | |
316 | // Regenerate free list. |
317 | t->blocknum_freelist_head.b = freelist_null.b; |
318 | for (b.b = RESERVED_BLOCKNUMS; b.b < t->smallest_never_used_blocknum.b; |
319 | b.b++) { |
320 | if (t->block_translation[b.b].size == size_is_free) { |
321 | t->block_translation[b.b].u.next_free_blocknum = |
322 | t->blocknum_freelist_head; |
323 | t->blocknum_freelist_head = b; |
324 | } |
325 | } |
326 | } |
327 | } |
328 | |
329 | // block table must be locked by caller of this function |
330 | void block_table::note_start_checkpoint_unlocked() { |
331 | toku_mutex_assert_locked(&_mutex); |
332 | |
333 | // We're going to do O(n) work to copy the translation, so we |
334 | // can afford to do O(n) work by optimizing the translation |
335 | _maybe_optimize_translation(&_current); |
336 | |
337 | // Copy current translation to inprogress translation. |
338 | _copy_translation(&_inprogress, &_current, TRANSLATION_INPROGRESS); |
339 | |
340 | _checkpoint_skipped = false; |
341 | } |
342 | |
343 | void block_table::note_skipped_checkpoint() { |
344 | // Purpose, alert block translation that the checkpoint was skipped, e.x. |
345 | // for a non-dirty header |
346 | _mutex_lock(); |
347 | paranoid_invariant_notnull(_inprogress.block_translation); |
348 | _checkpoint_skipped = true; |
349 | _mutex_unlock(); |
350 | } |
351 | |
352 | // Purpose: free any disk space used by previous checkpoint that isn't in use by |
353 | // either |
354 | // - current state |
355 | // - in-progress checkpoint |
356 | // capture inprogress as new checkpointed. |
357 | // For each entry in checkpointBTT |
358 | // if offset does not match offset in inprogress |
359 | // assert offset does not match offset in current |
360 | // free (offset,len) from checkpoint |
361 | // move inprogress to checkpoint (resetting type) |
362 | // inprogress = NULL |
363 | void block_table::note_end_checkpoint(int fd) { |
364 | // Free unused blocks |
365 | _mutex_lock(); |
366 | uint64_t allocated_limit_at_start = _bt_block_allocator->AllocatedLimit(); |
367 | paranoid_invariant_notnull(_inprogress.block_translation); |
368 | if (_checkpoint_skipped) { |
369 | toku_free(_inprogress.block_translation); |
370 | memset(&_inprogress, 0, sizeof(_inprogress)); |
371 | goto end; |
372 | } |
373 | |
374 | // Make certain inprogress was allocated space on disk |
375 | invariant( |
376 | _inprogress.block_translation[RESERVED_BLOCKNUM_TRANSLATION].size > 0); |
377 | invariant( |
378 | _inprogress.block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff > |
379 | 0); |
380 | |
381 | { |
382 | struct translation *t = &_checkpointed; |
383 | for (int64_t i = 0; i < t->length_of_array; i++) { |
384 | struct block_translation_pair *pair = &t->block_translation[i]; |
385 | if (pair->size > 0 && |
386 | !_translation_prevents_freeing( |
387 | &_inprogress, make_blocknum(i), pair)) { |
388 | invariant(!_translation_prevents_freeing( |
389 | &_current, make_blocknum(i), pair)); |
390 | _bt_block_allocator->FreeBlock(pair->u.diskoff, pair->size); |
391 | } |
392 | } |
393 | toku_free(_checkpointed.block_translation); |
394 | _checkpointed = _inprogress; |
395 | _checkpointed.type = TRANSLATION_CHECKPOINTED; |
396 | memset(&_inprogress, 0, sizeof(_inprogress)); |
397 | _maybe_truncate_file(fd, allocated_limit_at_start); |
398 | } |
399 | end: |
400 | _mutex_unlock(); |
401 | } |
402 | |
403 | bool block_table::_is_valid_blocknum(struct translation *t, BLOCKNUM b) { |
404 | invariant(t->length_of_array >= t->smallest_never_used_blocknum.b); |
405 | return b.b >= 0 && b.b < t->smallest_never_used_blocknum.b; |
406 | } |
407 | |
408 | void block_table::_verify_valid_blocknum(struct translation *UU(t), |
409 | BLOCKNUM UU(b)) { |
410 | invariant(_is_valid_blocknum(t, b)); |
411 | } |
412 | |
413 | bool block_table::_is_valid_freeable_blocknum(struct translation *t, |
414 | BLOCKNUM b) { |
415 | invariant(t->length_of_array >= t->smallest_never_used_blocknum.b); |
416 | return b.b >= RESERVED_BLOCKNUMS && b.b < t->smallest_never_used_blocknum.b; |
417 | } |
418 | |
419 | // should be freeable |
420 | void block_table::_verify_valid_freeable_blocknum(struct translation *UU(t), |
421 | BLOCKNUM UU(b)) { |
422 | invariant(_is_valid_freeable_blocknum(t, b)); |
423 | } |
424 | |
425 | // Also used only in ft-serialize-test. |
426 | void block_table::block_free(uint64_t offset, uint64_t size) { |
427 | _mutex_lock(); |
428 | _bt_block_allocator->FreeBlock(offset, size); |
429 | _mutex_unlock(); |
430 | } |
431 | |
432 | int64_t block_table::_calculate_size_on_disk(struct translation *t) { |
433 | return 8 + // smallest_never_used_blocknum |
434 | 8 + // blocknum_freelist_head |
435 | t->smallest_never_used_blocknum.b * 16 + // Array |
436 | 4; // 4 for checksum |
437 | } |
438 | |
439 | // We cannot free the disk space allocated to this blocknum if it is still in |
440 | // use by the given translation table. |
441 | bool block_table::_translation_prevents_freeing( |
442 | struct translation *t, |
443 | BLOCKNUM b, |
444 | struct block_translation_pair *old_pair) { |
445 | return t->block_translation && b.b < t->smallest_never_used_blocknum.b && |
446 | old_pair->u.diskoff == t->block_translation[b.b].u.diskoff; |
447 | } |
448 | |
449 | void block_table::_realloc_on_disk_internal(BLOCKNUM b, |
450 | DISKOFF size, |
451 | DISKOFF *offset, |
452 | FT ft, |
453 | bool for_checkpoint) { |
454 | toku_mutex_assert_locked(&_mutex); |
455 | ft_set_dirty(ft, for_checkpoint); |
456 | |
457 | struct translation *t = &_current; |
458 | struct block_translation_pair old_pair = t->block_translation[b.b]; |
459 | // Free the old block if it is not still in use by the checkpoint in |
460 | // progress or the previous checkpoint |
461 | bool cannot_free = |
462 | (!for_checkpoint && |
463 | _translation_prevents_freeing(&_inprogress, b, &old_pair)) || |
464 | _translation_prevents_freeing(&_checkpointed, b, &old_pair); |
465 | if (!cannot_free && old_pair.u.diskoff != diskoff_unused) { |
466 | _bt_block_allocator->FreeBlock(old_pair.u.diskoff, old_pair.size); |
467 | } |
468 | |
469 | uint64_t allocator_offset = diskoff_unused; |
470 | t->block_translation[b.b].size = size; |
471 | if (size > 0) { |
472 | // Allocate a new block if the size is greater than 0, |
473 | // if the size is just 0, offset will be set to diskoff_unused |
474 | _bt_block_allocator->AllocBlock(size, &allocator_offset); |
475 | } |
476 | t->block_translation[b.b].u.diskoff = allocator_offset; |
477 | *offset = allocator_offset; |
478 | |
479 | // Update inprogress btt if appropriate (if called because Pending bit is |
480 | // set). |
481 | if (for_checkpoint) { |
482 | paranoid_invariant(b.b < _inprogress.length_of_array); |
483 | _inprogress.block_translation[b.b] = t->block_translation[b.b]; |
484 | } |
485 | } |
486 | |
487 | void block_table::_ensure_safe_write_unlocked(int fd, |
488 | DISKOFF block_size, |
489 | DISKOFF block_offset) { |
490 | // Requires: holding _mutex |
491 | uint64_t size_needed = block_size + block_offset; |
492 | if (size_needed > _safe_file_size) { |
493 | // Must hold _safe_file_size_lock to change _safe_file_size. |
494 | nb_mutex_lock(&_safe_file_size_lock, &_mutex); |
495 | if (size_needed > _safe_file_size) { |
496 | _mutex_unlock(); |
497 | |
498 | int64_t size_after; |
499 | toku_maybe_preallocate_in_file( |
500 | fd, size_needed, _safe_file_size, &size_after); |
501 | |
502 | _mutex_lock(); |
503 | _safe_file_size = size_after; |
504 | } |
505 | nb_mutex_unlock(&_safe_file_size_lock); |
506 | } |
507 | } |
508 | |
509 | void block_table::realloc_on_disk(BLOCKNUM b, |
510 | DISKOFF size, |
511 | DISKOFF *offset, |
512 | FT ft, |
513 | int fd, |
514 | bool for_checkpoint) { |
515 | _mutex_lock(); |
516 | struct translation *t = &_current; |
517 | _verify_valid_freeable_blocknum(t, b); |
518 | _realloc_on_disk_internal(b, size, offset, ft, for_checkpoint); |
519 | |
520 | _ensure_safe_write_unlocked(fd, size, *offset); |
521 | _mutex_unlock(); |
522 | } |
523 | |
524 | bool block_table::_pair_is_unallocated(struct block_translation_pair *pair) { |
525 | return pair->size == 0 && pair->u.diskoff == diskoff_unused; |
526 | } |
527 | |
528 | // Effect: figure out where to put the inprogress btt on disk, allocate space |
529 | // for it there. |
530 | // The space must be 512-byte aligned (both the starting address and the |
531 | // size). |
532 | // As a result, the allcoated space may be a little bit bigger (up to the next |
533 | // 512-byte boundary) than the actual btt. |
534 | void block_table::_alloc_inprogress_translation_on_disk_unlocked() { |
535 | toku_mutex_assert_locked(&_mutex); |
536 | |
537 | struct translation *t = &_inprogress; |
538 | paranoid_invariant_notnull(t->block_translation); |
539 | BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION); |
540 | // Each inprogress is allocated only once |
541 | paranoid_invariant(_pair_is_unallocated(&t->block_translation[b.b])); |
542 | |
543 | // Allocate a new block |
544 | int64_t size = _calculate_size_on_disk(t); |
545 | uint64_t offset; |
546 | _bt_block_allocator->AllocBlock(size, &offset); |
547 | t->block_translation[b.b].u.diskoff = offset; |
548 | t->block_translation[b.b].size = size; |
549 | } |
550 | |
551 | // Effect: Serializes the blocktable to a wbuf (which starts uninitialized) |
552 | // A clean shutdown runs checkpoint start so that current and inprogress are |
553 | // copies. |
554 | // The resulting wbuf buffer is guaranteed to be be 512-byte aligned and the |
555 | // total length is a multiple of 512 (so we pad with zeros at the end if |
556 | // needd) |
557 | // The address is guaranteed to be 512-byte aligned, but the size is not |
558 | // guaranteed. |
559 | // It *is* guaranteed that we can read up to the next 512-byte boundary, |
560 | // however |
561 | void block_table::serialize_translation_to_wbuf(int fd, |
562 | struct wbuf *w, |
563 | int64_t *address, |
564 | int64_t *size) { |
565 | _mutex_lock(); |
566 | struct translation *t = &_inprogress; |
567 | |
568 | BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION); |
569 | _alloc_inprogress_translation_on_disk_unlocked(); // The allocated block |
570 | // must be 512-byte |
571 | // aligned to make |
572 | // O_DIRECT happy. |
573 | uint64_t size_translation = _calculate_size_on_disk(t); |
574 | uint64_t size_aligned = roundup_to_multiple(512, size_translation); |
575 | invariant((int64_t)size_translation == t->block_translation[b.b].size); |
576 | { |
577 | // Init wbuf |
578 | if (0) |
579 | printf( |
580 | "%s:%d writing translation table of size_translation %" PRIu64 |
581 | " at %" PRId64 "\n" , |
582 | __FILE__, |
583 | __LINE__, |
584 | size_translation, |
585 | t->block_translation[b.b].u.diskoff); |
586 | char *XMALLOC_N_ALIGNED(512, size_aligned, buf); |
587 | for (uint64_t i = size_translation; i < size_aligned; i++) |
588 | buf[i] = 0; // fill in the end of the buffer with zeros. |
589 | wbuf_init(w, buf, size_aligned); |
590 | } |
591 | wbuf_BLOCKNUM(w, t->smallest_never_used_blocknum); |
592 | wbuf_BLOCKNUM(w, t->blocknum_freelist_head); |
593 | int64_t i; |
594 | for (i = 0; i < t->smallest_never_used_blocknum.b; i++) { |
595 | if (0) |
596 | printf("%s:%d %" PRId64 ",%" PRId64 "\n" , |
597 | __FILE__, |
598 | __LINE__, |
599 | t->block_translation[i].u.diskoff, |
600 | t->block_translation[i].size); |
601 | wbuf_DISKOFF(w, t->block_translation[i].u.diskoff); |
602 | wbuf_DISKOFF(w, t->block_translation[i].size); |
603 | } |
604 | uint32_t checksum = toku_x1764_finish(&w->checksum); |
605 | wbuf_int(w, checksum); |
606 | *address = t->block_translation[b.b].u.diskoff; |
607 | *size = size_translation; |
608 | invariant((*address) % 512 == 0); |
609 | |
610 | _ensure_safe_write_unlocked(fd, size_aligned, *address); |
611 | _mutex_unlock(); |
612 | } |
613 | |
614 | // Perhaps rename: purpose is get disk address of a block, given its blocknum |
615 | // (blockid?) |
616 | void block_table::_translate_blocknum_to_offset_size_unlocked(BLOCKNUM b, |
617 | DISKOFF *offset, |
618 | DISKOFF *size) { |
619 | struct translation *t = &_current; |
620 | _verify_valid_blocknum(t, b); |
621 | if (offset) { |
622 | *offset = t->block_translation[b.b].u.diskoff; |
623 | } |
624 | if (size) { |
625 | *size = t->block_translation[b.b].size; |
626 | } |
627 | } |
628 | |
629 | // Perhaps rename: purpose is get disk address of a block, given its blocknum |
630 | // (blockid?) |
631 | void block_table::translate_blocknum_to_offset_size(BLOCKNUM b, |
632 | DISKOFF *offset, |
633 | DISKOFF *size) { |
634 | _mutex_lock(); |
635 | _translate_blocknum_to_offset_size_unlocked(b, offset, size); |
636 | _mutex_unlock(); |
637 | } |
638 | |
639 | // Only called by toku_allocate_blocknum |
640 | // Effect: expand the array to maintain size invariant |
641 | // given that one more never-used blocknum will soon be used. |
642 | void block_table::_maybe_expand_translation(struct translation *t) { |
643 | if (t->length_of_array <= t->smallest_never_used_blocknum.b) { |
644 | // expansion is necessary |
645 | uint64_t new_length = t->smallest_never_used_blocknum.b * 2; |
646 | XREALLOC_N(new_length, t->block_translation); |
647 | uint64_t i; |
648 | for (i = t->length_of_array; i < new_length; i++) { |
649 | t->block_translation[i].u.next_free_blocknum = freelist_null; |
650 | t->block_translation[i].size = size_is_free; |
651 | } |
652 | t->length_of_array = new_length; |
653 | } |
654 | } |
655 | |
656 | void block_table::_allocate_blocknum_unlocked(BLOCKNUM *res, FT ft) { |
657 | toku_mutex_assert_locked(&_mutex); |
658 | BLOCKNUM result; |
659 | struct translation *t = &_current; |
660 | if (t->blocknum_freelist_head.b == freelist_null.b) { |
661 | // no previously used blocknums are available |
662 | // use a never used blocknum |
663 | _maybe_expand_translation( |
664 | t); // Ensure a never used blocknums is available |
665 | result = t->smallest_never_used_blocknum; |
666 | t->smallest_never_used_blocknum.b++; |
667 | } else { // reuse a previously used blocknum |
668 | result = t->blocknum_freelist_head; |
669 | BLOCKNUM next = t->block_translation[result.b].u.next_free_blocknum; |
670 | t->blocknum_freelist_head = next; |
671 | } |
672 | // Verify the blocknum is free |
673 | paranoid_invariant(t->block_translation[result.b].size == size_is_free); |
674 | // blocknum is not free anymore |
675 | t->block_translation[result.b].u.diskoff = diskoff_unused; |
676 | t->block_translation[result.b].size = 0; |
677 | _verify_valid_freeable_blocknum(t, result); |
678 | *res = result; |
679 | ft_set_dirty(ft, false); |
680 | } |
681 | |
682 | void block_table::allocate_blocknum(BLOCKNUM *res, FT ft) { |
683 | _mutex_lock(); |
684 | _allocate_blocknum_unlocked(res, ft); |
685 | _mutex_unlock(); |
686 | } |
687 | |
688 | void block_table::_free_blocknum_in_translation(struct translation *t, |
689 | BLOCKNUM b) { |
690 | _verify_valid_freeable_blocknum(t, b); |
691 | paranoid_invariant(t->block_translation[b.b].size != size_is_free); |
692 | |
693 | t->block_translation[b.b].size = size_is_free; |
694 | t->block_translation[b.b].u.next_free_blocknum = t->blocknum_freelist_head; |
695 | t->blocknum_freelist_head = b; |
696 | } |
697 | |
698 | // Effect: Free a blocknum. |
699 | // If the blocknum holds the only reference to a block on disk, free that block |
700 | void block_table::_free_blocknum_unlocked(BLOCKNUM *bp, |
701 | FT ft, |
702 | bool for_checkpoint) { |
703 | toku_mutex_assert_locked(&_mutex); |
704 | BLOCKNUM b = *bp; |
705 | bp->b = 0; // Remove caller's reference. |
706 | |
707 | struct block_translation_pair old_pair = _current.block_translation[b.b]; |
708 | |
709 | _free_blocknum_in_translation(&_current, b); |
710 | if (for_checkpoint) { |
711 | paranoid_invariant(ft->checkpoint_header->type == |
712 | FT_CHECKPOINT_INPROGRESS); |
713 | _free_blocknum_in_translation(&_inprogress, b); |
714 | } |
715 | |
716 | // If the size is 0, no disk block has ever been assigned to this blocknum. |
717 | if (old_pair.size > 0) { |
718 | // Free the old block if it is not still in use by the checkpoint in |
719 | // progress or the previous checkpoint |
720 | bool cannot_free = |
721 | _translation_prevents_freeing(&_inprogress, b, &old_pair) || |
722 | _translation_prevents_freeing(&_checkpointed, b, &old_pair); |
723 | if (!cannot_free) { |
724 | _bt_block_allocator->FreeBlock(old_pair.u.diskoff, old_pair.size); |
725 | } |
726 | } else { |
727 | paranoid_invariant(old_pair.size == 0); |
728 | paranoid_invariant(old_pair.u.diskoff == diskoff_unused); |
729 | } |
730 | ft_set_dirty(ft, for_checkpoint); |
731 | } |
732 | |
733 | void block_table::free_blocknum(BLOCKNUM *bp, FT ft, bool for_checkpoint) { |
734 | _mutex_lock(); |
735 | _free_blocknum_unlocked(bp, ft, for_checkpoint); |
736 | _mutex_unlock(); |
737 | } |
738 | |
739 | // Verify there are no free blocks. |
740 | void block_table::verify_no_free_blocknums() { |
741 | invariant(_current.blocknum_freelist_head.b == freelist_null.b); |
742 | } |
743 | |
744 | // Frees blocknums that have a size of 0 and unused diskoff |
745 | // Currently used for eliminating unused cached rollback log nodes |
746 | void block_table::free_unused_blocknums(BLOCKNUM root) { |
747 | _mutex_lock(); |
748 | int64_t smallest = _current.smallest_never_used_blocknum.b; |
749 | for (int64_t i = RESERVED_BLOCKNUMS; i < smallest; i++) { |
750 | if (i == root.b) { |
751 | continue; |
752 | } |
753 | BLOCKNUM b = make_blocknum(i); |
754 | if (_current.block_translation[b.b].size == 0) { |
755 | invariant(_current.block_translation[b.b].u.diskoff == |
756 | diskoff_unused); |
757 | _free_blocknum_in_translation(&_current, b); |
758 | } |
759 | } |
760 | _mutex_unlock(); |
761 | } |
762 | |
763 | bool block_table::_no_data_blocks_except_root(BLOCKNUM root) { |
764 | bool ok = true; |
765 | _mutex_lock(); |
766 | int64_t smallest = _current.smallest_never_used_blocknum.b; |
767 | if (root.b < RESERVED_BLOCKNUMS) { |
768 | ok = false; |
769 | goto cleanup; |
770 | } |
771 | for (int64_t i = RESERVED_BLOCKNUMS; i < smallest; i++) { |
772 | if (i == root.b) { |
773 | continue; |
774 | } |
775 | BLOCKNUM b = make_blocknum(i); |
776 | if (_current.block_translation[b.b].size != size_is_free) { |
777 | ok = false; |
778 | goto cleanup; |
779 | } |
780 | } |
781 | cleanup: |
782 | _mutex_unlock(); |
783 | return ok; |
784 | } |
785 | |
786 | // Verify there are no data blocks except root. |
787 | // TODO(leif): This actually takes a lock, but I don't want to fix all the |
788 | // callers right now. |
789 | void block_table::verify_no_data_blocks_except_root(BLOCKNUM UU(root)) { |
790 | paranoid_invariant(_no_data_blocks_except_root(root)); |
791 | } |
792 | |
793 | bool block_table::_blocknum_allocated(BLOCKNUM b) { |
794 | _mutex_lock(); |
795 | struct translation *t = &_current; |
796 | _verify_valid_blocknum(t, b); |
797 | bool ok = t->block_translation[b.b].size != size_is_free; |
798 | _mutex_unlock(); |
799 | return ok; |
800 | } |
801 | |
802 | // Verify a blocknum is currently allocated. |
803 | void block_table::verify_blocknum_allocated(BLOCKNUM UU(b)) { |
804 | paranoid_invariant(_blocknum_allocated(b)); |
805 | } |
806 | |
807 | // Only used by toku_dump_translation table (debug info) |
808 | void block_table::_dump_translation_internal(FILE *f, struct translation *t) { |
809 | if (t->block_translation) { |
810 | BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_TRANSLATION); |
811 | fprintf(f, " length_of_array[%" PRId64 "]" , t->length_of_array); |
812 | fprintf(f, |
813 | " smallest_never_used_blocknum[%" PRId64 "]" , |
814 | t->smallest_never_used_blocknum.b); |
815 | fprintf(f, |
816 | " blocknum_free_list_head[%" PRId64 "]" , |
817 | t->blocknum_freelist_head.b); |
818 | fprintf( |
819 | f, " size_on_disk[%" PRId64 "]" , t->block_translation[b.b].size); |
820 | fprintf(f, |
821 | " location_on_disk[%" PRId64 "]\n" , |
822 | t->block_translation[b.b].u.diskoff); |
823 | int64_t i; |
824 | for (i = 0; i < t->length_of_array; i++) { |
825 | fprintf(f, |
826 | " %" PRId64 ": %" PRId64 " %" PRId64 "\n" , |
827 | i, |
828 | t->block_translation[i].u.diskoff, |
829 | t->block_translation[i].size); |
830 | } |
831 | fprintf(f, "\n" ); |
832 | } else { |
833 | fprintf(f, " does not exist\n" ); |
834 | } |
835 | } |
836 | |
837 | // Only used by toku_ft_dump which is only for debugging purposes |
838 | // "pretty" just means we use tabs so we can parse output easier later |
839 | void block_table::dump_translation_table_pretty(FILE *f) { |
840 | _mutex_lock(); |
841 | struct translation *t = &_checkpointed; |
842 | invariant(t->block_translation != nullptr); |
843 | for (int64_t i = 0; i < t->length_of_array; ++i) { |
844 | fprintf(f, |
845 | "%" PRId64 "\t%" PRId64 "\t%" PRId64 "\n" , |
846 | i, |
847 | t->block_translation[i].u.diskoff, |
848 | t->block_translation[i].size); |
849 | } |
850 | _mutex_unlock(); |
851 | } |
852 | |
853 | // Only used by toku_ft_dump which is only for debugging purposes |
854 | void block_table::dump_translation_table(FILE *f) { |
855 | _mutex_lock(); |
856 | fprintf(f, "Current block translation:" ); |
857 | _dump_translation_internal(f, &_current); |
858 | fprintf(f, "Checkpoint in progress block translation:" ); |
859 | _dump_translation_internal(f, &_inprogress); |
860 | fprintf(f, "Checkpointed block translation:" ); |
861 | _dump_translation_internal(f, &_checkpointed); |
862 | _mutex_unlock(); |
863 | } |
864 | |
865 | // Only used by ftdump |
866 | void block_table::blocknum_dump_translation(BLOCKNUM b) { |
867 | _mutex_lock(); |
868 | |
869 | struct translation *t = &_current; |
870 | if (b.b < t->length_of_array) { |
871 | struct block_translation_pair *bx = &t->block_translation[b.b]; |
872 | printf("%" PRId64 ": %" PRId64 " %" PRId64 "\n" , |
873 | b.b, |
874 | bx->u.diskoff, |
875 | bx->size); |
876 | } |
877 | _mutex_unlock(); |
878 | } |
879 | |
880 | // Must not call this function when anything else is using the blocktable. |
881 | // No one may use the blocktable afterwards. |
882 | void block_table::destroy(void) { |
883 | // TODO: translation.destroy(); |
884 | toku_free(_current.block_translation); |
885 | toku_free(_inprogress.block_translation); |
886 | toku_free(_checkpointed.block_translation); |
887 | |
888 | _bt_block_allocator->Destroy(); |
889 | delete _bt_block_allocator; |
890 | toku_mutex_destroy(&_mutex); |
891 | nb_mutex_destroy(&_safe_file_size_lock); |
892 | } |
893 | |
894 | int block_table::_translation_deserialize_from_buffer( |
895 | struct translation *t, |
896 | DISKOFF location_on_disk, |
897 | uint64_t size_on_disk, |
898 | // out: buffer with serialized translation |
899 | unsigned char *translation_buffer) { |
900 | int r = 0; |
901 | invariant(location_on_disk != 0); |
902 | t->type = TRANSLATION_CHECKPOINTED; |
903 | |
904 | // check the checksum |
905 | uint32_t x1764 = toku_x1764_memory(translation_buffer, size_on_disk - 4); |
906 | uint64_t offset = size_on_disk - 4; |
907 | uint32_t stored_x1764 = toku_dtoh32(*(int *)(translation_buffer + offset)); |
908 | if (x1764 != stored_x1764) { |
909 | fprintf(stderr, |
910 | "Translation table checksum failure: calc=0x%08x read=0x%08x\n" , |
911 | x1764, |
912 | stored_x1764); |
913 | r = TOKUDB_BAD_CHECKSUM; |
914 | goto exit; |
915 | } |
916 | |
917 | struct rbuf rb; |
918 | rb.buf = translation_buffer; |
919 | rb.ndone = 0; |
920 | rb.size = size_on_disk - 4; // 4==checksum |
921 | |
922 | t->smallest_never_used_blocknum = rbuf_blocknum(&rb); |
923 | t->length_of_array = t->smallest_never_used_blocknum.b; |
924 | invariant(t->smallest_never_used_blocknum.b >= RESERVED_BLOCKNUMS); |
925 | t->blocknum_freelist_head = rbuf_blocknum(&rb); |
926 | XMALLOC_N(t->length_of_array, t->block_translation); |
927 | for (int64_t i = 0; i < t->length_of_array; i++) { |
928 | t->block_translation[i].u.diskoff = rbuf_DISKOFF(&rb); |
929 | t->block_translation[i].size = rbuf_DISKOFF(&rb); |
930 | } |
931 | invariant(_calculate_size_on_disk(t) == (int64_t)size_on_disk); |
932 | invariant(t->block_translation[RESERVED_BLOCKNUM_TRANSLATION].size == |
933 | (int64_t)size_on_disk); |
934 | invariant(t->block_translation[RESERVED_BLOCKNUM_TRANSLATION].u.diskoff == |
935 | location_on_disk); |
936 | |
937 | exit: |
938 | return r; |
939 | } |
940 | |
941 | int block_table::iterate(enum translation_type type, |
942 | BLOCKTABLE_CALLBACK f, |
943 | void *, |
944 | bool data_only, |
945 | bool used_only) { |
946 | struct translation *src; |
947 | |
948 | int r = 0; |
949 | switch (type) { |
950 | case TRANSLATION_CURRENT: |
951 | src = &_current; |
952 | break; |
953 | case TRANSLATION_INPROGRESS: |
954 | src = &_inprogress; |
955 | break; |
956 | case TRANSLATION_CHECKPOINTED: |
957 | src = &_checkpointed; |
958 | break; |
959 | default: |
960 | r = EINVAL; |
961 | } |
962 | |
963 | struct translation fakecurrent; |
964 | memset(&fakecurrent, 0, sizeof(struct translation)); |
965 | |
966 | struct translation *t = &fakecurrent; |
967 | if (r == 0) { |
968 | _mutex_lock(); |
969 | _copy_translation(t, src, TRANSLATION_DEBUG); |
970 | t->block_translation[RESERVED_BLOCKNUM_TRANSLATION] = |
971 | src->block_translation[RESERVED_BLOCKNUM_TRANSLATION]; |
972 | _mutex_unlock(); |
973 | int64_t i; |
974 | for (i = 0; i < t->smallest_never_used_blocknum.b; i++) { |
975 | struct block_translation_pair pair = t->block_translation[i]; |
976 | if (data_only && i < RESERVED_BLOCKNUMS) |
977 | continue; |
978 | if (used_only && pair.size <= 0) |
979 | continue; |
980 | r = f(make_blocknum(i), pair.size, pair.u.diskoff, extra); |
981 | if (r != 0) |
982 | break; |
983 | } |
984 | toku_free(t->block_translation); |
985 | } |
986 | return r; |
987 | } |
988 | |
989 | typedef struct { |
990 | int64_t used_space; |
991 | int64_t total_space; |
992 | } ; |
993 | |
994 | static int frag_helper(BLOCKNUM UU(b), |
995 | int64_t size, |
996 | int64_t address, |
997 | void *) { |
998 | frag_extra *info = (frag_extra *)extra; |
999 | |
1000 | if (size + address > info->total_space) |
1001 | info->total_space = size + address; |
1002 | info->used_space += size; |
1003 | return 0; |
1004 | } |
1005 | |
1006 | void block_table::internal_fragmentation(int64_t *total_sizep, |
1007 | int64_t *used_sizep) { |
1008 | frag_extra info = {0, 0}; |
1009 | int r = iterate(TRANSLATION_CHECKPOINTED, frag_helper, &info, false, true); |
1010 | invariant_zero(r); |
1011 | |
1012 | if (total_sizep) |
1013 | *total_sizep = info.total_space; |
1014 | if (used_sizep) |
1015 | *used_sizep = info.used_space; |
1016 | } |
1017 | |
1018 | void block_table::_realloc_descriptor_on_disk_unlocked(DISKOFF size, |
1019 | DISKOFF *offset, |
1020 | FT ft) { |
1021 | toku_mutex_assert_locked(&_mutex); |
1022 | BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_DESCRIPTOR); |
1023 | _realloc_on_disk_internal(b, size, offset, ft, false); |
1024 | } |
1025 | |
1026 | void block_table::realloc_descriptor_on_disk(DISKOFF size, |
1027 | DISKOFF *offset, |
1028 | FT ft, |
1029 | int fd) { |
1030 | _mutex_lock(); |
1031 | _realloc_descriptor_on_disk_unlocked(size, offset, ft); |
1032 | _ensure_safe_write_unlocked(fd, size, *offset); |
1033 | _mutex_unlock(); |
1034 | } |
1035 | |
1036 | void block_table::get_descriptor_offset_size(DISKOFF *offset, DISKOFF *size) { |
1037 | _mutex_lock(); |
1038 | BLOCKNUM b = make_blocknum(RESERVED_BLOCKNUM_DESCRIPTOR); |
1039 | _translate_blocknum_to_offset_size_unlocked(b, offset, size); |
1040 | _mutex_unlock(); |
1041 | } |
1042 | |
1043 | void block_table::get_fragmentation_unlocked(TOKU_DB_FRAGMENTATION report) { |
1044 | // Requires: blocktable lock is held. |
1045 | // Requires: report->file_size_bytes is already filled in. |
1046 | |
1047 | // Count the headers. |
1048 | report->data_bytes = BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE; |
1049 | report->data_blocks = 1; |
1050 | report->checkpoint_bytes_additional = |
1051 | BlockAllocator::BLOCK_ALLOCATOR_HEADER_RESERVE; |
1052 | report->checkpoint_blocks_additional = 1; |
1053 | |
1054 | struct translation *current = &_current; |
1055 | for (int64_t i = 0; i < current->length_of_array; i++) { |
1056 | struct block_translation_pair *pair = ¤t->block_translation[i]; |
1057 | if (pair->size > 0) { |
1058 | report->data_bytes += pair->size; |
1059 | report->data_blocks++; |
1060 | } |
1061 | } |
1062 | |
1063 | struct translation *checkpointed = &_checkpointed; |
1064 | for (int64_t i = 0; i < checkpointed->length_of_array; i++) { |
1065 | struct block_translation_pair *pair = |
1066 | &checkpointed->block_translation[i]; |
1067 | if (pair->size > 0 && |
1068 | !(i < current->length_of_array && |
1069 | current->block_translation[i].size > 0 && |
1070 | current->block_translation[i].u.diskoff == pair->u.diskoff)) { |
1071 | report->checkpoint_bytes_additional += pair->size; |
1072 | report->checkpoint_blocks_additional++; |
1073 | } |
1074 | } |
1075 | |
1076 | struct translation *inprogress = &_inprogress; |
1077 | for (int64_t i = 0; i < inprogress->length_of_array; i++) { |
1078 | struct block_translation_pair *pair = &inprogress->block_translation[i]; |
1079 | if (pair->size > 0 && |
1080 | !(i < current->length_of_array && |
1081 | current->block_translation[i].size > 0 && |
1082 | current->block_translation[i].u.diskoff == pair->u.diskoff) && |
1083 | !(i < checkpointed->length_of_array && |
1084 | checkpointed->block_translation[i].size > 0 && |
1085 | checkpointed->block_translation[i].u.diskoff == |
1086 | pair->u.diskoff)) { |
1087 | report->checkpoint_bytes_additional += pair->size; |
1088 | report->checkpoint_blocks_additional++; |
1089 | } |
1090 | } |
1091 | |
1092 | _bt_block_allocator->UnusedStatistics(report); |
1093 | } |
1094 | |
1095 | void block_table::get_info64(struct ftinfo64 *s) { |
1096 | _mutex_lock(); |
1097 | |
1098 | struct translation *current = &_current; |
1099 | s->num_blocks_allocated = current->length_of_array; |
1100 | s->num_blocks_in_use = 0; |
1101 | s->size_allocated = 0; |
1102 | s->size_in_use = 0; |
1103 | |
1104 | for (int64_t i = 0; i < current->length_of_array; ++i) { |
1105 | struct block_translation_pair *block = ¤t->block_translation[i]; |
1106 | if (block->size != size_is_free) { |
1107 | ++s->num_blocks_in_use; |
1108 | s->size_in_use += block->size; |
1109 | if (block->u.diskoff != diskoff_unused) { |
1110 | uint64_t limit = block->u.diskoff + block->size; |
1111 | if (limit > s->size_allocated) { |
1112 | s->size_allocated = limit; |
1113 | } |
1114 | } |
1115 | } |
1116 | } |
1117 | |
1118 | _mutex_unlock(); |
1119 | } |
1120 | |
1121 | int block_table::iterate_translation_tables( |
1122 | uint64_t checkpoint_count, |
1123 | int (*iter)(uint64_t checkpoint_count, |
1124 | int64_t total_num_rows, |
1125 | int64_t blocknum, |
1126 | int64_t diskoff, |
1127 | int64_t size, |
1128 | void *), |
1129 | void *) { |
1130 | int error = 0; |
1131 | _mutex_lock(); |
1132 | |
1133 | int64_t total_num_rows = |
1134 | _current.length_of_array + _checkpointed.length_of_array; |
1135 | for (int64_t i = 0; error == 0 && i < _current.length_of_array; ++i) { |
1136 | struct block_translation_pair *block = &_current.block_translation[i]; |
1137 | error = iter(checkpoint_count, |
1138 | total_num_rows, |
1139 | i, |
1140 | block->u.diskoff, |
1141 | block->size, |
1142 | iter_extra); |
1143 | } |
1144 | for (int64_t i = 0; error == 0 && i < _checkpointed.length_of_array; ++i) { |
1145 | struct block_translation_pair *block = |
1146 | &_checkpointed.block_translation[i]; |
1147 | error = iter(checkpoint_count - 1, |
1148 | total_num_rows, |
1149 | i, |
1150 | block->u.diskoff, |
1151 | block->size, |
1152 | iter_extra); |
1153 | } |
1154 | |
1155 | _mutex_unlock(); |
1156 | return error; |
1157 | } |
1158 | |