forked from meeeejin/rw-rbuf
-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathbuf0dblwr.cc
1171 lines (949 loc) · 32.5 KB
/
buf0dblwr.cc
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*****************************************************************************
Copyright (c) 1995, 2014, Oracle and/or its affiliates. All Rights Reserved.
This program is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free Software
Foundation; version 2 of the License.
This program is distributed in the hope that it will be useful, but WITHOUT
ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
*****************************************************************************/
/**************************************************//**
@file buf/buf0dblwr.cc
Doublwrite buffer module
Created 2011/12/19
*******************************************************/
#include "buf0dblwr.h"
#ifdef UNIV_NONINL
#include "buf0buf.ic"
#endif
#include "buf0buf.h"
#include "buf0checksum.h"
#include "srv0start.h"
#include "srv0srv.h"
#include "page0zip.h"
#include "trx0sys.h"
#ifndef UNIV_HOTBACKUP
#ifdef UNIV_PFS_MUTEX
/* Key to register the mutex with performance schema */
UNIV_INTERN mysql_pfs_key_t buf_dblwr_mutex_key;
#endif /* UNIV_PFS_RWLOCK */
/** The doublewrite buffer */
UNIV_INTERN buf_dblwr_t* buf_dblwr = NULL;
/** Set to TRUE when the doublewrite buffer is being created */
UNIV_INTERN ibool buf_dblwr_being_created = FALSE;
/****************************************************************//**
Determines if a page number is located inside the doublewrite buffer.
@return TRUE if the location is inside the two blocks of the
doublewrite buffer */
UNIV_INTERN
ibool
buf_dblwr_page_inside(
/*==================*/
ulint page_no) /*!< in: page number */
{
if (buf_dblwr == NULL) {
return(FALSE);
}
if (page_no >= buf_dblwr->block1
&& page_no < buf_dblwr->block1
+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
return(TRUE);
}
if (page_no >= buf_dblwr->block2
&& page_no < buf_dblwr->block2
+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
return(TRUE);
}
return(FALSE);
}
/****************************************************************//**
Calls buf_page_get() on the TRX_SYS_PAGE and returns a pointer to the
doublewrite buffer within it.
@return pointer to the doublewrite buffer within the filespace header
page. */
UNIV_INLINE
byte*
buf_dblwr_get(
/*==========*/
mtr_t* mtr) /*!< in/out: MTR to hold the page latch */
{
buf_block_t* block;
block = buf_page_get(TRX_SYS_SPACE, 0, TRX_SYS_PAGE_NO,
RW_X_LATCH, mtr);
buf_block_dbg_add_level(block, SYNC_NO_ORDER_CHECK);
return(buf_block_get_frame(block) + TRX_SYS_DOUBLEWRITE);
}
/********************************************************************//**
Flush a batch of writes to the datafiles that have already been
written to the dblwr buffer on disk. */
UNIV_INLINE
void
buf_dblwr_sync_datafiles()
/*======================*/
{
/* Wake possible simulated aio thread to actually post the
writes to the operating system */
os_aio_simulated_wake_handler_threads();
/* Wait that all async writes to tablespaces have been posted to
the OS */
os_aio_wait_until_no_pending_writes();
/* Now we flush the data to disk (for example, with fsync) */
fil_flush_file_spaces(FIL_TABLESPACE);
}
/****************************************************************//**
Creates or initialializes the doublewrite buffer at a database start. */
static
void
buf_dblwr_init(
/*===========*/
byte* doublewrite) /*!< in: pointer to the doublewrite buf
header on trx sys page */
{
ulint buf_size;
buf_dblwr = static_cast<buf_dblwr_t*>(
mem_zalloc(sizeof(buf_dblwr_t)));
/* There are two blocks of same size in the doublewrite
buffer. */
buf_size = 2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE;
/* There must be atleast one buffer for single page writes
and one buffer for batch writes. */
ut_a(srv_doublewrite_batch_size > 0
&& srv_doublewrite_batch_size < buf_size);
mutex_create(buf_dblwr_mutex_key,
&buf_dblwr->mutex, SYNC_DOUBLEWRITE);
buf_dblwr->b_event = os_event_create();
buf_dblwr->s_event = os_event_create();
buf_dblwr->first_free = 0;
buf_dblwr->s_reserved = 0;
buf_dblwr->b_reserved = 0;
buf_dblwr->block1 = mach_read_from_4(
doublewrite + TRX_SYS_DOUBLEWRITE_BLOCK1);
buf_dblwr->block2 = mach_read_from_4(
doublewrite + TRX_SYS_DOUBLEWRITE_BLOCK2);
buf_dblwr->in_use = static_cast<bool*>(
mem_zalloc(buf_size * sizeof(bool)));
buf_dblwr->write_buf_unaligned = static_cast<byte*>(
ut_malloc((1 + buf_size) * UNIV_PAGE_SIZE));
buf_dblwr->write_buf = static_cast<byte*>(
ut_align(buf_dblwr->write_buf_unaligned,
UNIV_PAGE_SIZE));
buf_dblwr->buf_block_arr = static_cast<buf_page_t**>(
mem_zalloc(buf_size * sizeof(void*)));
}
/****************************************************************//**
Creates the doublewrite buffer to a new InnoDB installation. The header of the
doublewrite buffer is placed on the trx system header page. */
UNIV_INTERN
void
buf_dblwr_create(void)
/*==================*/
{
buf_block_t* block2;
buf_block_t* new_block;
byte* doublewrite;
byte* fseg_header;
ulint page_no;
ulint prev_page_no;
ulint i;
mtr_t mtr;
if (buf_dblwr) {
/* Already inited */
return;
}
start_again:
mtr_start(&mtr);
buf_dblwr_being_created = TRUE;
doublewrite = buf_dblwr_get(&mtr);
if (mach_read_from_4(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC)
== TRX_SYS_DOUBLEWRITE_MAGIC_N) {
/* The doublewrite buffer has already been created:
just read in some numbers */
buf_dblwr_init(doublewrite);
mtr_commit(&mtr);
buf_dblwr_being_created = FALSE;
return;
}
ib_logf(IB_LOG_LEVEL_INFO,
"Doublewrite buffer not found: creating new");
if (buf_pool_get_curr_size()
< ((2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE
+ FSP_EXTENT_SIZE / 2 + 100)
* UNIV_PAGE_SIZE)) {
ib_logf(IB_LOG_LEVEL_ERROR,
"Cannot create doublewrite buffer: you must "
"increase your buffer pool size. Cannot continue "
"operation.");
exit(EXIT_FAILURE);
}
block2 = fseg_create(TRX_SYS_SPACE, TRX_SYS_PAGE_NO,
TRX_SYS_DOUBLEWRITE
+ TRX_SYS_DOUBLEWRITE_FSEG, &mtr);
/* fseg_create acquires a second latch on the page,
therefore we must declare it: */
buf_block_dbg_add_level(block2, SYNC_NO_ORDER_CHECK);
if (block2 == NULL) {
ib_logf(IB_LOG_LEVEL_ERROR,
"Cannot create doublewrite buffer: you must "
"increase your tablespace size. "
"Cannot continue operation.");
/* We exit without committing the mtr to prevent
its modifications to the database getting to disk */
exit(EXIT_FAILURE);
}
fseg_header = doublewrite + TRX_SYS_DOUBLEWRITE_FSEG;
prev_page_no = 0;
for (i = 0; i < 2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE
+ FSP_EXTENT_SIZE / 2; i++) {
new_block = fseg_alloc_free_page(
fseg_header, prev_page_no + 1, FSP_UP, &mtr);
if (new_block == NULL) {
ib_logf(IB_LOG_LEVEL_ERROR,
"Cannot create doublewrite buffer: you must "
"increase your tablespace size. "
"Cannot continue operation.");
exit(EXIT_FAILURE);
}
/* We read the allocated pages to the buffer pool;
when they are written to disk in a flush, the space
id and page number fields are also written to the
pages. When we at database startup read pages
from the doublewrite buffer, we know that if the
space id and page number in them are the same as
the page position in the tablespace, then the page
has not been written to in doublewrite. */
ut_ad(rw_lock_get_x_lock_count(&new_block->lock) == 1);
page_no = buf_block_get_page_no(new_block);
if (i == FSP_EXTENT_SIZE / 2) {
ut_a(page_no == FSP_EXTENT_SIZE);
mlog_write_ulint(doublewrite
+ TRX_SYS_DOUBLEWRITE_BLOCK1,
page_no, MLOG_4BYTES, &mtr);
mlog_write_ulint(doublewrite
+ TRX_SYS_DOUBLEWRITE_REPEAT
+ TRX_SYS_DOUBLEWRITE_BLOCK1,
page_no, MLOG_4BYTES, &mtr);
} else if (i == FSP_EXTENT_SIZE / 2
+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
ut_a(page_no == 2 * FSP_EXTENT_SIZE);
mlog_write_ulint(doublewrite
+ TRX_SYS_DOUBLEWRITE_BLOCK2,
page_no, MLOG_4BYTES, &mtr);
mlog_write_ulint(doublewrite
+ TRX_SYS_DOUBLEWRITE_REPEAT
+ TRX_SYS_DOUBLEWRITE_BLOCK2,
page_no, MLOG_4BYTES, &mtr);
} else if (i > FSP_EXTENT_SIZE / 2) {
ut_a(page_no == prev_page_no + 1);
}
if (((i + 1) & 15) == 0) {
/* rw_locks can only be recursively x-locked
2048 times. (on 32 bit platforms,
(lint) 0 - (X_LOCK_DECR * 2049)
is no longer a negative number, and thus
lock_word becomes like a shared lock).
For 4k page size this loop will
lock the fseg header too many times. Since
this code is not done while any other threads
are active, restart the MTR occasionally. */
mtr_commit(&mtr);
mtr_start(&mtr);
doublewrite = buf_dblwr_get(&mtr);
fseg_header = doublewrite
+ TRX_SYS_DOUBLEWRITE_FSEG;
}
prev_page_no = page_no;
}
mlog_write_ulint(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC,
TRX_SYS_DOUBLEWRITE_MAGIC_N,
MLOG_4BYTES, &mtr);
mlog_write_ulint(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC
+ TRX_SYS_DOUBLEWRITE_REPEAT,
TRX_SYS_DOUBLEWRITE_MAGIC_N,
MLOG_4BYTES, &mtr);
mlog_write_ulint(doublewrite
+ TRX_SYS_DOUBLEWRITE_SPACE_ID_STORED,
TRX_SYS_DOUBLEWRITE_SPACE_ID_STORED_N,
MLOG_4BYTES, &mtr);
mtr_commit(&mtr);
/* Flush the modified pages to disk and make a checkpoint */
log_make_checkpoint_at(LSN_MAX, TRUE);
/* Remove doublewrite pages from LRU */
buf_pool_invalidate();
ib_logf(IB_LOG_LEVEL_INFO, "Doublewrite buffer created");
goto start_again;
}
/****************************************************************//**
At a database startup initializes the doublewrite buffer memory structure if
we already have a doublewrite buffer created in the data files. If we are
upgrading to an InnoDB version which supports multiple tablespaces, then this
function performs the necessary update operations. If we are in a crash
recovery, this function loads the pages from double write buffer into memory. */
void
buf_dblwr_init_or_load_pages(
/*=========================*/
os_file_t file,
char* path,
bool load_corrupt_pages)
{
byte* buf;
byte* read_buf;
byte* unaligned_read_buf;
ulint block1;
ulint block2;
byte* page;
ibool reset_space_ids = FALSE;
byte* doublewrite;
ulint space_id;
ulint i;
ulint block_bytes = 0;
recv_dblwr_t& recv_dblwr = recv_sys->dblwr;
/* We do the file i/o past the buffer pool */
unaligned_read_buf = static_cast<byte*>(ut_malloc(2 * UNIV_PAGE_SIZE));
read_buf = static_cast<byte*>(
ut_align(unaligned_read_buf, UNIV_PAGE_SIZE));
/* Read the trx sys header to check if we are using the doublewrite
buffer */
off_t trx_sys_page = TRX_SYS_PAGE_NO * UNIV_PAGE_SIZE;
os_file_read(file, read_buf, trx_sys_page, UNIV_PAGE_SIZE);
doublewrite = read_buf + TRX_SYS_DOUBLEWRITE;
if (mach_read_from_4(doublewrite + TRX_SYS_DOUBLEWRITE_MAGIC)
== TRX_SYS_DOUBLEWRITE_MAGIC_N) {
/* The doublewrite buffer has been created */
buf_dblwr_init(doublewrite);
block1 = buf_dblwr->block1;
block2 = buf_dblwr->block2;
buf = buf_dblwr->write_buf;
} else {
goto leave_func;
}
if (mach_read_from_4(doublewrite + TRX_SYS_DOUBLEWRITE_SPACE_ID_STORED)
!= TRX_SYS_DOUBLEWRITE_SPACE_ID_STORED_N) {
/* We are upgrading from a version < 4.1.x to a version where
multiple tablespaces are supported. We must reset the space id
field in the pages in the doublewrite buffer because starting
from this version the space id is stored to
FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID. */
reset_space_ids = TRUE;
ib_logf(IB_LOG_LEVEL_INFO,
"Resetting space id's in the doublewrite buffer");
}
/* Read the pages from the doublewrite buffer to memory */
block_bytes = TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE;
os_file_read(file, buf, block1 * UNIV_PAGE_SIZE, block_bytes);
os_file_read(file, buf + block_bytes, block2 * UNIV_PAGE_SIZE,
block_bytes);
/* Check if any of these pages is half-written in data files, in the
intended position */
page = buf;
for (i = 0; i < TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * 2; i++) {
ulint source_page_no;
if (reset_space_ids) {
space_id = 0;
mach_write_to_4(page
+ FIL_PAGE_ARCH_LOG_NO_OR_SPACE_ID, space_id);
/* We do not need to calculate new checksums for the
pages because the field .._SPACE_ID does not affect
them. Write the page back to where we read it from. */
if (i < TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
source_page_no = block1 + i;
} else {
source_page_no = block2
+ i - TRX_SYS_DOUBLEWRITE_BLOCK_SIZE;
}
os_file_write(path, file, page,
source_page_no * UNIV_PAGE_SIZE,
UNIV_PAGE_SIZE);
} else if (load_corrupt_pages) {
recv_dblwr.add(page);
}
page += UNIV_PAGE_SIZE;
}
if (reset_space_ids) {
os_file_flush(file);
}
leave_func:
ut_free(unaligned_read_buf);
}
/****************************************************************//**
Process the double write buffer pages. */
void
buf_dblwr_process()
/*===============*/
{
ulint space_id;
ulint page_no;
ulint page_no_dblwr = 0;
byte* page;
byte* read_buf;
byte* unaligned_read_buf;
recv_dblwr_t& recv_dblwr = recv_sys->dblwr;
unaligned_read_buf = static_cast<byte*>(ut_malloc(2 * UNIV_PAGE_SIZE));
read_buf = static_cast<byte*>(
ut_align(unaligned_read_buf, UNIV_PAGE_SIZE));
for (std::list<byte*>::iterator i = recv_dblwr.pages.begin();
i != recv_dblwr.pages.end(); ++i, ++page_no_dblwr ) {
page = *i;
page_no = mach_read_from_4(page + FIL_PAGE_OFFSET);
space_id = mach_read_from_4(page + FIL_PAGE_SPACE_ID);
if (!fil_tablespace_exists_in_mem(space_id)) {
/* Maybe we have dropped the single-table tablespace
and this page once belonged to it: do nothing */
} else if (!fil_check_adress_in_tablespace(space_id,
page_no)) {
ib_logf(IB_LOG_LEVEL_WARN,
"A page in the doublewrite buffer is not "
"within space bounds; space id %lu "
"page number %lu, page %lu in "
"doublewrite buf.",
(ulong) space_id, (ulong) page_no,
page_no_dblwr);
} else {
ulint zip_size = fil_space_get_zip_size(space_id);
/* Read in the actual page from the file */
fil_io(OS_FILE_READ, true, space_id, zip_size,
page_no, 0,
zip_size ? zip_size : UNIV_PAGE_SIZE,
read_buf, NULL);
/* Check if the page is corrupt */
if (buf_page_is_corrupted(true, read_buf, zip_size)) {
fprintf(stderr,
"InnoDB: Warning: database page"
" corruption or a failed\n"
"InnoDB: file read of"
" space %lu page %lu.\n"
"InnoDB: Trying to recover it from"
" the doublewrite buffer.\n",
(ulong) space_id, (ulong) page_no);
if (buf_page_is_corrupted(true,
page, zip_size)) {
fprintf(stderr,
"InnoDB: Dump of the page:\n");
buf_page_print(
read_buf, zip_size,
BUF_PAGE_PRINT_NO_CRASH);
fprintf(stderr,
"InnoDB: Dump of"
" corresponding page"
" in doublewrite buffer:\n");
buf_page_print(
page, zip_size,
BUF_PAGE_PRINT_NO_CRASH);
fprintf(stderr,
"InnoDB: Also the page in the"
" doublewrite buffer"
" is corrupt.\n"
"InnoDB: Cannot continue"
" operation.\n"
"InnoDB: You can try to"
" recover the database"
" with the my.cnf\n"
"InnoDB: option:\n"
"InnoDB:"
" innodb_force_recovery=6\n");
ut_error;
}
/* Write the good page from the
doublewrite buffer to the intended
position */
fil_io(OS_FILE_WRITE, true, space_id,
zip_size, page_no, 0,
zip_size ? zip_size : UNIV_PAGE_SIZE,
page, NULL);
ib_logf(IB_LOG_LEVEL_INFO,
"Recovered the page from"
" the doublewrite buffer.");
} else if (buf_page_is_zeroes(read_buf, zip_size)) {
if (!buf_page_is_zeroes(page, zip_size)
&& !buf_page_is_corrupted(true, page,
zip_size)) {
/* Database page contained only
zeroes, while a valid copy is
available in dblwr buffer. */
fil_io(OS_FILE_WRITE, true, space_id,
zip_size, page_no, 0,
zip_size ? zip_size
: UNIV_PAGE_SIZE,
page, NULL);
}
}
}
}
fil_flush_file_spaces(FIL_TABLESPACE);
ut_free(unaligned_read_buf);
}
/****************************************************************//**
Frees doublewrite buffer. */
UNIV_INTERN
void
buf_dblwr_free(void)
/*================*/
{
/* Free the double write data structures. */
ut_a(buf_dblwr != NULL);
ut_ad(buf_dblwr->s_reserved == 0);
ut_ad(buf_dblwr->b_reserved == 0);
os_event_free(buf_dblwr->b_event);
os_event_free(buf_dblwr->s_event);
ut_free(buf_dblwr->write_buf_unaligned);
buf_dblwr->write_buf_unaligned = NULL;
mem_free(buf_dblwr->buf_block_arr);
buf_dblwr->buf_block_arr = NULL;
mem_free(buf_dblwr->in_use);
buf_dblwr->in_use = NULL;
mutex_free(&buf_dblwr->mutex);
mem_free(buf_dblwr);
buf_dblwr = NULL;
}
/********************************************************************//**
Updates the doublewrite buffer when an IO request is completed. */
UNIV_INTERN
void
buf_dblwr_update(
/*=============*/
const buf_page_t* bpage, /*!< in: buffer block descriptor */
buf_flush_t flush_type)/*!< in: flush type */
{
if (!srv_use_doublewrite_buf || buf_dblwr == NULL) {
return;
}
switch (flush_type) {
case BUF_FLUSH_LIST:
case BUF_FLUSH_LRU:
mutex_enter(&buf_dblwr->mutex);
ut_ad(buf_dblwr->batch_running);
ut_ad(buf_dblwr->b_reserved > 0);
ut_ad(buf_dblwr->b_reserved <= buf_dblwr->first_free);
buf_dblwr->b_reserved--;
if (buf_dblwr->b_reserved == 0) {
mutex_exit(&buf_dblwr->mutex);
/* This will finish the batch. Sync data files
to the disk. */
fil_flush_file_spaces(FIL_TABLESPACE);
mutex_enter(&buf_dblwr->mutex);
/* We can now reuse the doublewrite memory buffer: */
buf_dblwr->first_free = 0;
buf_dblwr->batch_running = false;
os_event_set(buf_dblwr->b_event);
}
mutex_exit(&buf_dblwr->mutex);
break;
case BUF_FLUSH_SINGLE_PAGE:
{
const ulint size = 2 * TRX_SYS_DOUBLEWRITE_BLOCK_SIZE;
ulint i;
mutex_enter(&buf_dblwr->mutex);
for (i = srv_doublewrite_batch_size; i < size; ++i) {
if (buf_dblwr->buf_block_arr[i] == bpage) {
buf_dblwr->s_reserved--;
buf_dblwr->buf_block_arr[i] = NULL;
buf_dblwr->in_use[i] = false;
break;
}
}
/* The block we are looking for must exist as a
reserved block. */
ut_a(i < size);
}
os_event_set(buf_dblwr->s_event);
mutex_exit(&buf_dblwr->mutex);
break;
case BUF_FLUSH_N_TYPES:
ut_error;
}
}
/********************************************************************//**
Check the LSN values on the page. */
static
void
buf_dblwr_check_page_lsn(
/*=====================*/
const page_t* page) /*!< in: page to check */
{
if (memcmp(page + (FIL_PAGE_LSN + 4),
page + (UNIV_PAGE_SIZE
- FIL_PAGE_END_LSN_OLD_CHKSUM + 4),
4)) {
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: ERROR: The page to be written"
" seems corrupt!\n"
"InnoDB: The low 4 bytes of LSN fields do not match "
"(" ULINTPF " != " ULINTPF ")!"
" Noticed in the buffer pool.\n",
mach_read_from_4(
page + FIL_PAGE_LSN + 4),
mach_read_from_4(
page + UNIV_PAGE_SIZE
- FIL_PAGE_END_LSN_OLD_CHKSUM + 4));
}
}
/********************************************************************//**
Asserts when a corrupt block is find during writing out data to the
disk. */
static
void
buf_dblwr_assert_on_corrupt_block(
/*==============================*/
const buf_block_t* block) /*!< in: block to check */
{
buf_page_print(block->frame, 0, BUF_PAGE_PRINT_NO_CRASH);
ut_print_timestamp(stderr);
fprintf(stderr,
" InnoDB: Apparent corruption of an"
" index page n:o %lu in space %lu\n"
"InnoDB: to be written to data file."
" We intentionally crash server\n"
"InnoDB: to prevent corrupt data"
" from ending up in data\n"
"InnoDB: files.\n",
(ulong) buf_block_get_page_no(block),
(ulong) buf_block_get_space(block));
ut_error;
}
/********************************************************************//**
Check the LSN values on the page with which this block is associated.
Also validate the page if the option is set. */
static
void
buf_dblwr_check_block(
/*==================*/
const buf_block_t* block) /*!< in: block to check */
{
if (buf_block_get_state(block) != BUF_BLOCK_FILE_PAGE
|| block->page.zip.data) {
/* No simple validate for compressed pages exists. */
return;
}
buf_dblwr_check_page_lsn(block->frame);
if (!block->check_index_page_at_flush) {
return;
}
if (page_is_comp(block->frame)) {
if (!page_simple_validate_new(block->frame)) {
buf_dblwr_assert_on_corrupt_block(block);
}
} else if (!page_simple_validate_old(block->frame)) {
buf_dblwr_assert_on_corrupt_block(block);
}
}
/********************************************************************//**
Writes a page that has already been written to the doublewrite buffer
to the datafile. It is the job of the caller to sync the datafile. */
static
void
buf_dblwr_write_block_to_datafile(
/*==============================*/
const buf_page_t* bpage, /*!< in: page to write */
bool sync) /*!< in: true if sync IO
is requested */
{
ut_a(bpage);
ut_a(buf_page_in_file(bpage));
const ulint flags = sync
? OS_FILE_WRITE
: OS_FILE_WRITE | OS_AIO_SIMULATED_WAKE_LATER;
if (bpage->zip.data) {
fil_io(flags, sync, buf_page_get_space(bpage),
buf_page_get_zip_size(bpage),
buf_page_get_page_no(bpage), 0,
buf_page_get_zip_size(bpage),
(void*) bpage->zip.data,
(void*) bpage);
return;
}
const buf_block_t* block = (buf_block_t*) bpage;
ut_a(buf_block_get_state(block) == BUF_BLOCK_FILE_PAGE);
buf_dblwr_check_page_lsn(block->frame);
fil_io(flags, sync, buf_block_get_space(block), 0,
buf_block_get_page_no(block), 0, UNIV_PAGE_SIZE,
(void*) block->frame, (void*) block);
}
/********************************************************************//**
Flushes possible buffered writes from the doublewrite memory buffer to disk,
and also wakes up the aio thread if simulated aio is used. It is very
important to call this function after a batch of writes has been posted,
and also when we may have to wait for a page latch! Otherwise a deadlock
of threads can occur. */
UNIV_INTERN
void
buf_dblwr_flush_buffered_writes(void)
/*=================================*/
{
byte* write_buf;
ulint first_free;
ulint len;
if (!srv_use_doublewrite_buf || buf_dblwr == NULL) {
/* Sync the writes to the disk. */
buf_dblwr_sync_datafiles();
return;
}
try_again:
mutex_enter(&buf_dblwr->mutex);
/* Write first to doublewrite buffer blocks. We use synchronous
aio and thus know that file write has been completed when the
control returns. */
if (buf_dblwr->first_free == 0) {
mutex_exit(&buf_dblwr->mutex);
return;
}
if (buf_dblwr->batch_running) {
/* Another thread is running the batch right now. Wait
for it to finish. */
ib_int64_t sig_count = os_event_reset(buf_dblwr->b_event);
mutex_exit(&buf_dblwr->mutex);
os_event_wait_low(buf_dblwr->b_event, sig_count);
goto try_again;
}
ut_a(!buf_dblwr->batch_running);
ut_ad(buf_dblwr->first_free == buf_dblwr->b_reserved);
/* Disallow anyone else to post to doublewrite buffer or to
start another batch of flushing. */
buf_dblwr->batch_running = true;
first_free = buf_dblwr->first_free;
/* Now safe to release the mutex. Note that though no other
thread is allowed to post to the doublewrite batch flushing
but any threads working on single page flushes are allowed
to proceed. */
mutex_exit(&buf_dblwr->mutex);
write_buf = buf_dblwr->write_buf;
for (ulint len2 = 0, i = 0;
i < buf_dblwr->first_free;
len2 += UNIV_PAGE_SIZE, i++) {
const buf_block_t* block;
block = (buf_block_t*) buf_dblwr->buf_block_arr[i];
if (buf_block_get_state(block) != BUF_BLOCK_FILE_PAGE
|| block->page.zip.data) {
/* No simple validate for compressed
pages exists. */
continue;
}
/* Check that the actual page in the buffer pool is
not corrupt and the LSN values are sane. */
buf_dblwr_check_block(block);
/* Check that the page as written to the doublewrite
buffer has sane LSN values. */
buf_dblwr_check_page_lsn(write_buf + len2);
}
/* Write out the first block of the doublewrite buffer */
len = ut_min(TRX_SYS_DOUBLEWRITE_BLOCK_SIZE,
buf_dblwr->first_free) * UNIV_PAGE_SIZE;
fil_io(OS_FILE_WRITE, true, TRX_SYS_SPACE, 0,
buf_dblwr->block1, 0, len,
(void*) write_buf, NULL);
if (buf_dblwr->first_free <= TRX_SYS_DOUBLEWRITE_BLOCK_SIZE) {
/* No unwritten pages in the second block. */
goto flush;
}
/* Write out the second block of the doublewrite buffer. */
len = (buf_dblwr->first_free - TRX_SYS_DOUBLEWRITE_BLOCK_SIZE)
* UNIV_PAGE_SIZE;
write_buf = buf_dblwr->write_buf
+ TRX_SYS_DOUBLEWRITE_BLOCK_SIZE * UNIV_PAGE_SIZE;
fil_io(OS_FILE_WRITE, true, TRX_SYS_SPACE, 0,
buf_dblwr->block2, 0, len,
(void*) write_buf, NULL);
flush:
/* increment the doublewrite flushed pages counter */
srv_stats.dblwr_pages_written.add(buf_dblwr->first_free);
srv_stats.dblwr_writes.inc();
/* Now flush the doublewrite buffer data to disk */
fil_flush(TRX_SYS_SPACE);
/* We know that the writes have been flushed to disk now
and in recovery we will find them in the doublewrite buffer
blocks. Next do the writes to the intended positions. */
/* Up to this point first_free and buf_dblwr->first_free are
same because we have set the buf_dblwr->batch_running flag
disallowing any other thread to post any request but we
can't safely access buf_dblwr->first_free in the loop below.
This is so because it is possible that after we are done with
the last iteration and before we terminate the loop, the batch
gets finished in the IO helper thread and another thread posts
a new batch setting buf_dblwr->first_free to a higher value.
If this happens and we are using buf_dblwr->first_free in the
loop termination condition then we'll end up dispatching
the same block twice from two different threads. */
ut_ad(first_free == buf_dblwr->first_free);
for (ulint i = 0; i < first_free; i++) {
buf_dblwr_write_block_to_datafile(
buf_dblwr->buf_block_arr[i], false);
}
/* Wake possible simulated aio thread to actually post the
writes to the operating system. We don't flush the files
at this point. We leave it to the IO helper thread to flush
datafiles when the whole batch has been processed. */
os_aio_simulated_wake_handler_threads();
}
/********************************************************************//**
Posts a buffer page for writing. If the doublewrite memory buffer is
full, calls buf_dblwr_flush_buffered_writes and waits for for free
space to appear. */
UNIV_INTERN
void
buf_dblwr_add_to_batch(
/*====================*/
buf_page_t* bpage) /*!< in: buffer block to write */
{
ulint zip_size;
ut_a(buf_page_in_file(bpage));
try_again:
mutex_enter(&buf_dblwr->mutex);
ut_a(buf_dblwr->first_free <= srv_doublewrite_batch_size);
if (buf_dblwr->batch_running) {
/* This not nearly as bad as it looks. There is only
page_cleaner thread which does background flushing
in batches therefore it is unlikely to be a contention
point. The only exception is when a user thread is
forced to do a flush batch because of a sync
checkpoint. */
ib_int64_t sig_count = os_event_reset(buf_dblwr->b_event);
mutex_exit(&buf_dblwr->mutex);
os_event_wait_low(buf_dblwr->b_event, sig_count);
goto try_again;
}
if (buf_dblwr->first_free == srv_doublewrite_batch_size) {
mutex_exit(&(buf_dblwr->mutex));
buf_dblwr_flush_buffered_writes();