forked from Sricharanti/sricharan
-
Notifications
You must be signed in to change notification settings - Fork 0
/
jfs_logmgr.c
2543 lines (2171 loc) · 59.3 KB
/
jfs_logmgr.c
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) International Business Machines Corp., 2000-2004
* Portions Copyright (C) Christoph Hellwig, 2001-2002
*
* 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; either version 2 of the License, or
* (at your option) any later version.
*
* 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*/
/*
* jfs_logmgr.c: log manager
*
* for related information, see transaction manager (jfs_txnmgr.c), and
* recovery manager (jfs_logredo.c).
*
* note: for detail, RTFS.
*
* log buffer manager:
* special purpose buffer manager supporting log i/o requirements.
* per log serial pageout of logpage
* queuing i/o requests and redrive i/o at iodone
* maintain current logpage buffer
* no caching since append only
* appropriate jfs buffer cache buffers as needed
*
* group commit:
* transactions which wrote COMMIT records in the same in-memory
* log page during the pageout of previous/current log page(s) are
* committed together by the pageout of the page.
*
* TBD lazy commit:
* transactions are committed asynchronously when the log page
* containing it COMMIT is paged out when it becomes full;
*
* serialization:
* . a per log lock serialize log write.
* . a per log lock serialize group commit.
* . a per log lock serialize log open/close;
*
* TBD log integrity:
* careful-write (ping-pong) of last logpage to recover from crash
* in overwrite.
* detection of split (out-of-order) write of physical sectors
* of last logpage via timestamp at end of each sector
* with its mirror data array at trailer).
*
* alternatives:
* lsn - 64-bit monotonically increasing integer vs
* 32-bit lspn and page eor.
*/
#include <linux/fs.h>
#include <linux/blkdev.h>
#include <linux/interrupt.h>
#include <linux/smp_lock.h>
#include <linux/completion.h>
#include <linux/kthread.h>
#include <linux/buffer_head.h> /* for sync_blockdev() */
#include <linux/bio.h>
#include <linux/suspend.h>
#include <linux/delay.h>
#include "jfs_incore.h"
#include "jfs_filsys.h"
#include "jfs_metapage.h"
#include "jfs_superblock.h"
#include "jfs_txnmgr.h"
#include "jfs_debug.h"
/*
* lbuf's ready to be redriven. Protected by log_redrive_lock (jfsIO thread)
*/
static struct lbuf *log_redrive_list;
static DEFINE_SPINLOCK(log_redrive_lock);
/*
* log read/write serialization (per log)
*/
#define LOG_LOCK_INIT(log) mutex_init(&(log)->loglock)
#define LOG_LOCK(log) mutex_lock(&((log)->loglock))
#define LOG_UNLOCK(log) mutex_unlock(&((log)->loglock))
/*
* log group commit serialization (per log)
*/
#define LOGGC_LOCK_INIT(log) spin_lock_init(&(log)->gclock)
#define LOGGC_LOCK(log) spin_lock_irq(&(log)->gclock)
#define LOGGC_UNLOCK(log) spin_unlock_irq(&(log)->gclock)
#define LOGGC_WAKEUP(tblk) wake_up_all(&(tblk)->gcwait)
/*
* log sync serialization (per log)
*/
#define LOGSYNC_DELTA(logsize) min((logsize)/8, 128*LOGPSIZE)
#define LOGSYNC_BARRIER(logsize) ((logsize)/4)
/*
#define LOGSYNC_DELTA(logsize) min((logsize)/4, 256*LOGPSIZE)
#define LOGSYNC_BARRIER(logsize) ((logsize)/2)
*/
/*
* log buffer cache synchronization
*/
static DEFINE_SPINLOCK(jfsLCacheLock);
#define LCACHE_LOCK(flags) spin_lock_irqsave(&jfsLCacheLock, flags)
#define LCACHE_UNLOCK(flags) spin_unlock_irqrestore(&jfsLCacheLock, flags)
/*
* See __SLEEP_COND in jfs_locks.h
*/
#define LCACHE_SLEEP_COND(wq, cond, flags) \
do { \
if (cond) \
break; \
__SLEEP_COND(wq, cond, LCACHE_LOCK(flags), LCACHE_UNLOCK(flags)); \
} while (0)
#define LCACHE_WAKEUP(event) wake_up(event)
/*
* lbuf buffer cache (lCache) control
*/
/* log buffer manager pageout control (cumulative, inclusive) */
#define lbmREAD 0x0001
#define lbmWRITE 0x0002 /* enqueue at tail of write queue;
* init pageout if at head of queue;
*/
#define lbmRELEASE 0x0004 /* remove from write queue
* at completion of pageout;
* do not free/recycle it yet:
* caller will free it;
*/
#define lbmSYNC 0x0008 /* do not return to freelist
* when removed from write queue;
*/
#define lbmFREE 0x0010 /* return to freelist
* at completion of pageout;
* the buffer may be recycled;
*/
#define lbmDONE 0x0020
#define lbmERROR 0x0040
#define lbmGC 0x0080 /* lbmIODone to perform post-GC processing
* of log page
*/
#define lbmDIRECT 0x0100
/*
* Global list of active external journals
*/
static LIST_HEAD(jfs_external_logs);
static struct jfs_log *dummy_log = NULL;
static DECLARE_MUTEX(jfs_log_sem);
/*
* forward references
*/
static int lmWriteRecord(struct jfs_log * log, struct tblock * tblk,
struct lrd * lrd, struct tlock * tlck);
static int lmNextPage(struct jfs_log * log);
static int lmLogFileSystem(struct jfs_log * log, struct jfs_sb_info *sbi,
int activate);
static int open_inline_log(struct super_block *sb);
static int open_dummy_log(struct super_block *sb);
static int lbmLogInit(struct jfs_log * log);
static void lbmLogShutdown(struct jfs_log * log);
static struct lbuf *lbmAllocate(struct jfs_log * log, int);
static void lbmFree(struct lbuf * bp);
static void lbmfree(struct lbuf * bp);
static int lbmRead(struct jfs_log * log, int pn, struct lbuf ** bpp);
static void lbmWrite(struct jfs_log * log, struct lbuf * bp, int flag, int cant_block);
static void lbmDirectWrite(struct jfs_log * log, struct lbuf * bp, int flag);
static int lbmIOWait(struct lbuf * bp, int flag);
static bio_end_io_t lbmIODone;
static void lbmStartIO(struct lbuf * bp);
static void lmGCwrite(struct jfs_log * log, int cant_block);
static int lmLogSync(struct jfs_log * log, int hard_sync);
/*
* statistics
*/
#ifdef CONFIG_JFS_STATISTICS
static struct lmStat {
uint commit; /* # of commit */
uint pagedone; /* # of page written */
uint submitted; /* # of pages submitted */
uint full_page; /* # of full pages submitted */
uint partial_page; /* # of partial pages submitted */
} lmStat;
#endif
/*
* NAME: lmLog()
*
* FUNCTION: write a log record;
*
* PARAMETER:
*
* RETURN: lsn - offset to the next log record to write (end-of-log);
* -1 - error;
*
* note: todo: log error handler
*/
int lmLog(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
struct tlock * tlck)
{
int lsn;
int diffp, difft;
struct metapage *mp = NULL;
unsigned long flags;
jfs_info("lmLog: log:0x%p tblk:0x%p, lrd:0x%p tlck:0x%p",
log, tblk, lrd, tlck);
LOG_LOCK(log);
/* log by (out-of-transaction) JFS ? */
if (tblk == NULL)
goto writeRecord;
/* log from page ? */
if (tlck == NULL ||
tlck->type & tlckBTROOT || (mp = tlck->mp) == NULL)
goto writeRecord;
/*
* initialize/update page/transaction recovery lsn
*/
lsn = log->lsn;
LOGSYNC_LOCK(log, flags);
/*
* initialize page lsn if first log write of the page
*/
if (mp->lsn == 0) {
mp->log = log;
mp->lsn = lsn;
log->count++;
/* insert page at tail of logsynclist */
list_add_tail(&mp->synclist, &log->synclist);
}
/*
* initialize/update lsn of tblock of the page
*
* transaction inherits oldest lsn of pages associated
* with allocation/deallocation of resources (their
* log records are used to reconstruct allocation map
* at recovery time: inode for inode allocation map,
* B+-tree index of extent descriptors for block
* allocation map);
* allocation map pages inherit transaction lsn at
* commit time to allow forwarding log syncpt past log
* records associated with allocation/deallocation of
* resources only after persistent map of these map pages
* have been updated and propagated to home.
*/
/*
* initialize transaction lsn:
*/
if (tblk->lsn == 0) {
/* inherit lsn of its first page logged */
tblk->lsn = mp->lsn;
log->count++;
/* insert tblock after the page on logsynclist */
list_add(&tblk->synclist, &mp->synclist);
}
/*
* update transaction lsn:
*/
else {
/* inherit oldest/smallest lsn of page */
logdiff(diffp, mp->lsn, log);
logdiff(difft, tblk->lsn, log);
if (diffp < difft) {
/* update tblock lsn with page lsn */
tblk->lsn = mp->lsn;
/* move tblock after page on logsynclist */
list_move(&tblk->synclist, &mp->synclist);
}
}
LOGSYNC_UNLOCK(log, flags);
/*
* write the log record
*/
writeRecord:
lsn = lmWriteRecord(log, tblk, lrd, tlck);
/*
* forward log syncpt if log reached next syncpt trigger
*/
logdiff(diffp, lsn, log);
if (diffp >= log->nextsync)
lsn = lmLogSync(log, 0);
/* update end-of-log lsn */
log->lsn = lsn;
LOG_UNLOCK(log);
/* return end-of-log address */
return lsn;
}
/*
* NAME: lmWriteRecord()
*
* FUNCTION: move the log record to current log page
*
* PARAMETER: cd - commit descriptor
*
* RETURN: end-of-log address
*
* serialization: LOG_LOCK() held on entry/exit
*/
static int
lmWriteRecord(struct jfs_log * log, struct tblock * tblk, struct lrd * lrd,
struct tlock * tlck)
{
int lsn = 0; /* end-of-log address */
struct lbuf *bp; /* dst log page buffer */
struct logpage *lp; /* dst log page */
caddr_t dst; /* destination address in log page */
int dstoffset; /* end-of-log offset in log page */
int freespace; /* free space in log page */
caddr_t p; /* src meta-data page */
caddr_t src;
int srclen;
int nbytes; /* number of bytes to move */
int i;
int len;
struct linelock *linelock;
struct lv *lv;
struct lvd *lvd;
int l2linesize;
len = 0;
/* retrieve destination log page to write */
bp = (struct lbuf *) log->bp;
lp = (struct logpage *) bp->l_ldata;
dstoffset = log->eor;
/* any log data to write ? */
if (tlck == NULL)
goto moveLrd;
/*
* move log record data
*/
/* retrieve source meta-data page to log */
if (tlck->flag & tlckPAGELOCK) {
p = (caddr_t) (tlck->mp->data);
linelock = (struct linelock *) & tlck->lock;
}
/* retrieve source in-memory inode to log */
else if (tlck->flag & tlckINODELOCK) {
if (tlck->type & tlckDTREE)
p = (caddr_t) &JFS_IP(tlck->ip)->i_dtroot;
else
p = (caddr_t) &JFS_IP(tlck->ip)->i_xtroot;
linelock = (struct linelock *) & tlck->lock;
}
#ifdef _JFS_WIP
else if (tlck->flag & tlckINLINELOCK) {
inlinelock = (struct inlinelock *) & tlck;
p = (caddr_t) & inlinelock->pxd;
linelock = (struct linelock *) & tlck;
}
#endif /* _JFS_WIP */
else {
jfs_err("lmWriteRecord: UFO tlck:0x%p", tlck);
return 0; /* Probably should trap */
}
l2linesize = linelock->l2linesize;
moveData:
ASSERT(linelock->index <= linelock->maxcnt);
lv = linelock->lv;
for (i = 0; i < linelock->index; i++, lv++) {
if (lv->length == 0)
continue;
/* is page full ? */
if (dstoffset >= LOGPSIZE - LOGPTLRSIZE) {
/* page become full: move on to next page */
lmNextPage(log);
bp = log->bp;
lp = (struct logpage *) bp->l_ldata;
dstoffset = LOGPHDRSIZE;
}
/*
* move log vector data
*/
src = (u8 *) p + (lv->offset << l2linesize);
srclen = lv->length << l2linesize;
len += srclen;
while (srclen > 0) {
freespace = (LOGPSIZE - LOGPTLRSIZE) - dstoffset;
nbytes = min(freespace, srclen);
dst = (caddr_t) lp + dstoffset;
memcpy(dst, src, nbytes);
dstoffset += nbytes;
/* is page not full ? */
if (dstoffset < LOGPSIZE - LOGPTLRSIZE)
break;
/* page become full: move on to next page */
lmNextPage(log);
bp = (struct lbuf *) log->bp;
lp = (struct logpage *) bp->l_ldata;
dstoffset = LOGPHDRSIZE;
srclen -= nbytes;
src += nbytes;
}
/*
* move log vector descriptor
*/
len += 4;
lvd = (struct lvd *) ((caddr_t) lp + dstoffset);
lvd->offset = cpu_to_le16(lv->offset);
lvd->length = cpu_to_le16(lv->length);
dstoffset += 4;
jfs_info("lmWriteRecord: lv offset:%d length:%d",
lv->offset, lv->length);
}
if ((i = linelock->next)) {
linelock = (struct linelock *) lid_to_tlock(i);
goto moveData;
}
/*
* move log record descriptor
*/
moveLrd:
lrd->length = cpu_to_le16(len);
src = (caddr_t) lrd;
srclen = LOGRDSIZE;
while (srclen > 0) {
freespace = (LOGPSIZE - LOGPTLRSIZE) - dstoffset;
nbytes = min(freespace, srclen);
dst = (caddr_t) lp + dstoffset;
memcpy(dst, src, nbytes);
dstoffset += nbytes;
srclen -= nbytes;
/* are there more to move than freespace of page ? */
if (srclen)
goto pageFull;
/*
* end of log record descriptor
*/
/* update last log record eor */
log->eor = dstoffset;
bp->l_eor = dstoffset;
lsn = (log->page << L2LOGPSIZE) + dstoffset;
if (lrd->type & cpu_to_le16(LOG_COMMIT)) {
tblk->clsn = lsn;
jfs_info("wr: tclsn:0x%x, beor:0x%x", tblk->clsn,
bp->l_eor);
INCREMENT(lmStat.commit); /* # of commit */
/*
* enqueue tblock for group commit:
*
* enqueue tblock of non-trivial/synchronous COMMIT
* at tail of group commit queue
* (trivial/asynchronous COMMITs are ignored by
* group commit.)
*/
LOGGC_LOCK(log);
/* init tblock gc state */
tblk->flag = tblkGC_QUEUE;
tblk->bp = log->bp;
tblk->pn = log->page;
tblk->eor = log->eor;
/* enqueue transaction to commit queue */
list_add_tail(&tblk->cqueue, &log->cqueue);
LOGGC_UNLOCK(log);
}
jfs_info("lmWriteRecord: lrd:0x%04x bp:0x%p pn:%d eor:0x%x",
le16_to_cpu(lrd->type), log->bp, log->page, dstoffset);
/* page not full ? */
if (dstoffset < LOGPSIZE - LOGPTLRSIZE)
return lsn;
pageFull:
/* page become full: move on to next page */
lmNextPage(log);
bp = (struct lbuf *) log->bp;
lp = (struct logpage *) bp->l_ldata;
dstoffset = LOGPHDRSIZE;
src += nbytes;
}
return lsn;
}
/*
* NAME: lmNextPage()
*
* FUNCTION: write current page and allocate next page.
*
* PARAMETER: log
*
* RETURN: 0
*
* serialization: LOG_LOCK() held on entry/exit
*/
static int lmNextPage(struct jfs_log * log)
{
struct logpage *lp;
int lspn; /* log sequence page number */
int pn; /* current page number */
struct lbuf *bp;
struct lbuf *nextbp;
struct tblock *tblk;
/* get current log page number and log sequence page number */
pn = log->page;
bp = log->bp;
lp = (struct logpage *) bp->l_ldata;
lspn = le32_to_cpu(lp->h.page);
LOGGC_LOCK(log);
/*
* write or queue the full page at the tail of write queue
*/
/* get the tail tblk on commit queue */
if (list_empty(&log->cqueue))
tblk = NULL;
else
tblk = list_entry(log->cqueue.prev, struct tblock, cqueue);
/* every tblk who has COMMIT record on the current page,
* and has not been committed, must be on commit queue
* since tblk is queued at commit queueu at the time
* of writing its COMMIT record on the page before
* page becomes full (even though the tblk thread
* who wrote COMMIT record may have been suspended
* currently);
*/
/* is page bound with outstanding tail tblk ? */
if (tblk && tblk->pn == pn) {
/* mark tblk for end-of-page */
tblk->flag |= tblkGC_EOP;
if (log->cflag & logGC_PAGEOUT) {
/* if page is not already on write queue,
* just enqueue (no lbmWRITE to prevent redrive)
* buffer to wqueue to ensure correct serial order
* of the pages since log pages will be added
* continuously
*/
if (bp->l_wqnext == NULL)
lbmWrite(log, bp, 0, 0);
} else {
/*
* No current GC leader, initiate group commit
*/
log->cflag |= logGC_PAGEOUT;
lmGCwrite(log, 0);
}
}
/* page is not bound with outstanding tblk:
* init write or mark it to be redriven (lbmWRITE)
*/
else {
/* finalize the page */
bp->l_ceor = bp->l_eor;
lp->h.eor = lp->t.eor = cpu_to_le16(bp->l_ceor);
lbmWrite(log, bp, lbmWRITE | lbmRELEASE | lbmFREE, 0);
}
LOGGC_UNLOCK(log);
/*
* allocate/initialize next page
*/
/* if log wraps, the first data page of log is 2
* (0 never used, 1 is superblock).
*/
log->page = (pn == log->size - 1) ? 2 : pn + 1;
log->eor = LOGPHDRSIZE; /* ? valid page empty/full at logRedo() */
/* allocate/initialize next log page buffer */
nextbp = lbmAllocate(log, log->page);
nextbp->l_eor = log->eor;
log->bp = nextbp;
/* initialize next log page */
lp = (struct logpage *) nextbp->l_ldata;
lp->h.page = lp->t.page = cpu_to_le32(lspn + 1);
lp->h.eor = lp->t.eor = cpu_to_le16(LOGPHDRSIZE);
return 0;
}
/*
* NAME: lmGroupCommit()
*
* FUNCTION: group commit
* initiate pageout of the pages with COMMIT in the order of
* page number - redrive pageout of the page at the head of
* pageout queue until full page has been written.
*
* RETURN:
*
* NOTE:
* LOGGC_LOCK serializes log group commit queue, and
* transaction blocks on the commit queue.
* N.B. LOG_LOCK is NOT held during lmGroupCommit().
*/
int lmGroupCommit(struct jfs_log * log, struct tblock * tblk)
{
int rc = 0;
LOGGC_LOCK(log);
/* group committed already ? */
if (tblk->flag & tblkGC_COMMITTED) {
if (tblk->flag & tblkGC_ERROR)
rc = -EIO;
LOGGC_UNLOCK(log);
return rc;
}
jfs_info("lmGroup Commit: tblk = 0x%p, gcrtc = %d", tblk, log->gcrtc);
if (tblk->xflag & COMMIT_LAZY)
tblk->flag |= tblkGC_LAZY;
if ((!(log->cflag & logGC_PAGEOUT)) && (!list_empty(&log->cqueue)) &&
(!(tblk->xflag & COMMIT_LAZY) || test_bit(log_FLUSH, &log->flag)
|| jfs_tlocks_low)) {
/*
* No pageout in progress
*
* start group commit as its group leader.
*/
log->cflag |= logGC_PAGEOUT;
lmGCwrite(log, 0);
}
if (tblk->xflag & COMMIT_LAZY) {
/*
* Lazy transactions can leave now
*/
LOGGC_UNLOCK(log);
return 0;
}
/* lmGCwrite gives up LOGGC_LOCK, check again */
if (tblk->flag & tblkGC_COMMITTED) {
if (tblk->flag & tblkGC_ERROR)
rc = -EIO;
LOGGC_UNLOCK(log);
return rc;
}
/* upcount transaction waiting for completion
*/
log->gcrtc++;
tblk->flag |= tblkGC_READY;
__SLEEP_COND(tblk->gcwait, (tblk->flag & tblkGC_COMMITTED),
LOGGC_LOCK(log), LOGGC_UNLOCK(log));
/* removed from commit queue */
if (tblk->flag & tblkGC_ERROR)
rc = -EIO;
LOGGC_UNLOCK(log);
return rc;
}
/*
* NAME: lmGCwrite()
*
* FUNCTION: group commit write
* initiate write of log page, building a group of all transactions
* with commit records on that page.
*
* RETURN: None
*
* NOTE:
* LOGGC_LOCK must be held by caller.
* N.B. LOG_LOCK is NOT held during lmGroupCommit().
*/
static void lmGCwrite(struct jfs_log * log, int cant_write)
{
struct lbuf *bp;
struct logpage *lp;
int gcpn; /* group commit page number */
struct tblock *tblk;
struct tblock *xtblk = NULL;
/*
* build the commit group of a log page
*
* scan commit queue and make a commit group of all
* transactions with COMMIT records on the same log page.
*/
/* get the head tblk on the commit queue */
gcpn = list_entry(log->cqueue.next, struct tblock, cqueue)->pn;
list_for_each_entry(tblk, &log->cqueue, cqueue) {
if (tblk->pn != gcpn)
break;
xtblk = tblk;
/* state transition: (QUEUE, READY) -> COMMIT */
tblk->flag |= tblkGC_COMMIT;
}
tblk = xtblk; /* last tblk of the page */
/*
* pageout to commit transactions on the log page.
*/
bp = (struct lbuf *) tblk->bp;
lp = (struct logpage *) bp->l_ldata;
/* is page already full ? */
if (tblk->flag & tblkGC_EOP) {
/* mark page to free at end of group commit of the page */
tblk->flag &= ~tblkGC_EOP;
tblk->flag |= tblkGC_FREE;
bp->l_ceor = bp->l_eor;
lp->h.eor = lp->t.eor = cpu_to_le16(bp->l_ceor);
lbmWrite(log, bp, lbmWRITE | lbmRELEASE | lbmGC,
cant_write);
INCREMENT(lmStat.full_page);
}
/* page is not yet full */
else {
bp->l_ceor = tblk->eor; /* ? bp->l_ceor = bp->l_eor; */
lp->h.eor = lp->t.eor = cpu_to_le16(bp->l_ceor);
lbmWrite(log, bp, lbmWRITE | lbmGC, cant_write);
INCREMENT(lmStat.partial_page);
}
}
/*
* NAME: lmPostGC()
*
* FUNCTION: group commit post-processing
* Processes transactions after their commit records have been written
* to disk, redriving log I/O if necessary.
*
* RETURN: None
*
* NOTE:
* This routine is called a interrupt time by lbmIODone
*/
static void lmPostGC(struct lbuf * bp)
{
unsigned long flags;
struct jfs_log *log = bp->l_log;
struct logpage *lp;
struct tblock *tblk, *temp;
//LOGGC_LOCK(log);
spin_lock_irqsave(&log->gclock, flags);
/*
* current pageout of group commit completed.
*
* remove/wakeup transactions from commit queue who were
* group committed with the current log page
*/
list_for_each_entry_safe(tblk, temp, &log->cqueue, cqueue) {
if (!(tblk->flag & tblkGC_COMMIT))
break;
/* if transaction was marked GC_COMMIT then
* it has been shipped in the current pageout
* and made it to disk - it is committed.
*/
if (bp->l_flag & lbmERROR)
tblk->flag |= tblkGC_ERROR;
/* remove it from the commit queue */
list_del(&tblk->cqueue);
tblk->flag &= ~tblkGC_QUEUE;
if (tblk == log->flush_tblk) {
/* we can stop flushing the log now */
clear_bit(log_FLUSH, &log->flag);
log->flush_tblk = NULL;
}
jfs_info("lmPostGC: tblk = 0x%p, flag = 0x%x", tblk,
tblk->flag);
if (!(tblk->xflag & COMMIT_FORCE))
/*
* Hand tblk over to lazy commit thread
*/
txLazyUnlock(tblk);
else {
/* state transition: COMMIT -> COMMITTED */
tblk->flag |= tblkGC_COMMITTED;
if (tblk->flag & tblkGC_READY)
log->gcrtc--;
LOGGC_WAKEUP(tblk);
}
/* was page full before pageout ?
* (and this is the last tblk bound with the page)
*/
if (tblk->flag & tblkGC_FREE)
lbmFree(bp);
/* did page become full after pageout ?
* (and this is the last tblk bound with the page)
*/
else if (tblk->flag & tblkGC_EOP) {
/* finalize the page */
lp = (struct logpage *) bp->l_ldata;
bp->l_ceor = bp->l_eor;
lp->h.eor = lp->t.eor = cpu_to_le16(bp->l_eor);
jfs_info("lmPostGC: calling lbmWrite");
lbmWrite(log, bp, lbmWRITE | lbmRELEASE | lbmFREE,
1);
}
}
/* are there any transactions who have entered lnGroupCommit()
* (whose COMMITs are after that of the last log page written.
* They are waiting for new group commit (above at (SLEEP 1))
* or lazy transactions are on a full (queued) log page,
* select the latest ready transaction as new group leader and
* wake her up to lead her group.
*/
if ((!list_empty(&log->cqueue)) &&
((log->gcrtc > 0) || (tblk->bp->l_wqnext != NULL) ||
test_bit(log_FLUSH, &log->flag) || jfs_tlocks_low))
/*
* Call lmGCwrite with new group leader
*/
lmGCwrite(log, 1);
/* no transaction are ready yet (transactions are only just
* queued (GC_QUEUE) and not entered for group commit yet).
* the first transaction entering group commit
* will elect herself as new group leader.
*/
else
log->cflag &= ~logGC_PAGEOUT;
//LOGGC_UNLOCK(log);
spin_unlock_irqrestore(&log->gclock, flags);
return;
}
/*
* NAME: lmLogSync()
*
* FUNCTION: write log SYNCPT record for specified log
* if new sync address is available
* (normally the case if sync() is executed by back-ground
* process).
* calculate new value of i_nextsync which determines when
* this code is called again.
*
* PARAMETERS: log - log structure
* hard_sync - 1 to force all metadata to be written
*
* RETURN: 0
*
* serialization: LOG_LOCK() held on entry/exit
*/
static int lmLogSync(struct jfs_log * log, int hard_sync)
{
int logsize;
int written; /* written since last syncpt */
int free; /* free space left available */
int delta; /* additional delta to write normally */
int more; /* additional write granted */
struct lrd lrd;
int lsn;
struct logsyncblk *lp;
struct jfs_sb_info *sbi;
unsigned long flags;
/* push dirty metapages out to disk */
if (hard_sync)
list_for_each_entry(sbi, &log->sb_list, log_list) {
filemap_fdatawrite(sbi->ipbmap->i_mapping);
filemap_fdatawrite(sbi->ipimap->i_mapping);
filemap_fdatawrite(sbi->direct_inode->i_mapping);
}
else
list_for_each_entry(sbi, &log->sb_list, log_list) {
filemap_flush(sbi->ipbmap->i_mapping);
filemap_flush(sbi->ipimap->i_mapping);
filemap_flush(sbi->direct_inode->i_mapping);
}
/*
* forward syncpt
*/
/* if last sync is same as last syncpt,
* invoke sync point forward processing to update sync.
*/
if (log->sync == log->syncpt) {
LOGSYNC_LOCK(log, flags);
if (list_empty(&log->synclist))
log->sync = log->lsn;
else {
lp = list_entry(log->synclist.next,
struct logsyncblk, synclist);
log->sync = lp->lsn;
}
LOGSYNC_UNLOCK(log, flags);
}
/* if sync is different from last syncpt,
* write a SYNCPT record with syncpt = sync.
* reset syncpt = sync
*/
if (log->sync != log->syncpt) {
lrd.logtid = 0;
lrd.backchain = 0;
lrd.type = cpu_to_le16(LOG_SYNCPT);
lrd.length = 0;
lrd.log.syncpt.sync = cpu_to_le32(log->sync);
lsn = lmWriteRecord(log, NULL, &lrd, NULL);
log->syncpt = log->sync;
} else
lsn = log->lsn;
/*
* setup next syncpt trigger (SWAG)
*/
logsize = log->logsize;
logdiff(written, lsn, log);
free = logsize - written;
delta = LOGSYNC_DELTA(logsize);
more = min(free / 2, delta);
if (more < 2 * LOGPSIZE) {