forked from torvalds/linux
-
Notifications
You must be signed in to change notification settings - Fork 1
/
dm-raid1.c
1897 lines (1573 loc) · 44.3 KB
/
dm-raid1.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) 2003 Sistina Software Limited.
*
* This file is released under the GPL.
*/
#include "dm.h"
#include "dm-bio-list.h"
#include "dm-bio-record.h"
#include <linux/ctype.h>
#include <linux/init.h>
#include <linux/mempool.h>
#include <linux/module.h>
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/time.h>
#include <linux/vmalloc.h>
#include <linux/workqueue.h>
#include <linux/log2.h>
#include <linux/hardirq.h>
#include <linux/dm-io.h>
#include <linux/dm-dirty-log.h>
#include <linux/dm-kcopyd.h>
#define DM_MSG_PREFIX "raid1"
#define DM_IO_PAGES 64
#define DM_RAID1_HANDLE_ERRORS 0x01
#define errors_handled(p) ((p)->features & DM_RAID1_HANDLE_ERRORS)
static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
/*-----------------------------------------------------------------
* Region hash
*
* The mirror splits itself up into discrete regions. Each
* region can be in one of three states: clean, dirty,
* nosync. There is no need to put clean regions in the hash.
*
* In addition to being present in the hash table a region _may_
* be present on one of three lists.
*
* clean_regions: Regions on this list have no io pending to
* them, they are in sync, we are no longer interested in them,
* they are dull. rh_update_states() will remove them from the
* hash table.
*
* quiesced_regions: These regions have been spun down, ready
* for recovery. rh_recovery_start() will remove regions from
* this list and hand them to kmirrord, which will schedule the
* recovery io with kcopyd.
*
* recovered_regions: Regions that kcopyd has successfully
* recovered. rh_update_states() will now schedule any delayed
* io, up the recovery_count, and remove the region from the
* hash.
*
* There are 2 locks:
* A rw spin lock 'hash_lock' protects just the hash table,
* this is never held in write mode from interrupt context,
* which I believe means that we only have to disable irqs when
* doing a write lock.
*
* An ordinary spin lock 'region_lock' that protects the three
* lists in the region_hash, with the 'state', 'list' and
* 'bhs_delayed' fields of the regions. This is used from irq
* context, so all other uses will have to suspend local irqs.
*---------------------------------------------------------------*/
struct mirror_set;
struct region_hash {
struct mirror_set *ms;
uint32_t region_size;
unsigned region_shift;
/* holds persistent region state */
struct dm_dirty_log *log;
/* hash table */
rwlock_t hash_lock;
mempool_t *region_pool;
unsigned int mask;
unsigned int nr_buckets;
struct list_head *buckets;
spinlock_t region_lock;
atomic_t recovery_in_flight;
struct semaphore recovery_count;
struct list_head clean_regions;
struct list_head quiesced_regions;
struct list_head recovered_regions;
struct list_head failed_recovered_regions;
};
enum {
RH_CLEAN,
RH_DIRTY,
RH_NOSYNC,
RH_RECOVERING
};
struct region {
struct region_hash *rh; /* FIXME: can we get rid of this ? */
region_t key;
int state;
struct list_head hash_list;
struct list_head list;
atomic_t pending;
struct bio_list delayed_bios;
};
/*-----------------------------------------------------------------
* Mirror set structures.
*---------------------------------------------------------------*/
enum dm_raid1_error {
DM_RAID1_WRITE_ERROR,
DM_RAID1_SYNC_ERROR,
DM_RAID1_READ_ERROR
};
struct mirror {
struct mirror_set *ms;
atomic_t error_count;
unsigned long error_type;
struct dm_dev *dev;
sector_t offset;
};
struct mirror_set {
struct dm_target *ti;
struct list_head list;
struct region_hash rh;
struct dm_kcopyd_client *kcopyd_client;
uint64_t features;
spinlock_t lock; /* protects the lists */
struct bio_list reads;
struct bio_list writes;
struct bio_list failures;
struct dm_io_client *io_client;
mempool_t *read_record_pool;
/* recovery */
region_t nr_regions;
int in_sync;
int log_failure;
atomic_t suspend;
atomic_t default_mirror; /* Default mirror */
struct workqueue_struct *kmirrord_wq;
struct work_struct kmirrord_work;
struct timer_list timer;
unsigned long timer_pending;
struct work_struct trigger_event;
unsigned int nr_mirrors;
struct mirror mirror[0];
};
/*
* Conversion fns
*/
static inline region_t bio_to_region(struct region_hash *rh, struct bio *bio)
{
return (bio->bi_sector - rh->ms->ti->begin) >> rh->region_shift;
}
static inline sector_t region_to_sector(struct region_hash *rh, region_t region)
{
return region << rh->region_shift;
}
static void wake(struct mirror_set *ms)
{
queue_work(ms->kmirrord_wq, &ms->kmirrord_work);
}
static void delayed_wake_fn(unsigned long data)
{
struct mirror_set *ms = (struct mirror_set *) data;
clear_bit(0, &ms->timer_pending);
wake(ms);
}
static void delayed_wake(struct mirror_set *ms)
{
if (test_and_set_bit(0, &ms->timer_pending))
return;
ms->timer.expires = jiffies + HZ / 5;
ms->timer.data = (unsigned long) ms;
ms->timer.function = delayed_wake_fn;
add_timer(&ms->timer);
}
/* FIXME move this */
static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw);
#define MIN_REGIONS 64
#define MAX_RECOVERY 1
static int rh_init(struct region_hash *rh, struct mirror_set *ms,
struct dm_dirty_log *log, uint32_t region_size,
region_t nr_regions)
{
unsigned int nr_buckets, max_buckets;
size_t i;
/*
* Calculate a suitable number of buckets for our hash
* table.
*/
max_buckets = nr_regions >> 6;
for (nr_buckets = 128u; nr_buckets < max_buckets; nr_buckets <<= 1)
;
nr_buckets >>= 1;
rh->ms = ms;
rh->log = log;
rh->region_size = region_size;
rh->region_shift = ffs(region_size) - 1;
rwlock_init(&rh->hash_lock);
rh->mask = nr_buckets - 1;
rh->nr_buckets = nr_buckets;
rh->buckets = vmalloc(nr_buckets * sizeof(*rh->buckets));
if (!rh->buckets) {
DMERR("unable to allocate region hash memory");
return -ENOMEM;
}
for (i = 0; i < nr_buckets; i++)
INIT_LIST_HEAD(rh->buckets + i);
spin_lock_init(&rh->region_lock);
sema_init(&rh->recovery_count, 0);
atomic_set(&rh->recovery_in_flight, 0);
INIT_LIST_HEAD(&rh->clean_regions);
INIT_LIST_HEAD(&rh->quiesced_regions);
INIT_LIST_HEAD(&rh->recovered_regions);
INIT_LIST_HEAD(&rh->failed_recovered_regions);
rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
sizeof(struct region));
if (!rh->region_pool) {
vfree(rh->buckets);
rh->buckets = NULL;
return -ENOMEM;
}
return 0;
}
static void rh_exit(struct region_hash *rh)
{
unsigned int h;
struct region *reg, *nreg;
BUG_ON(!list_empty(&rh->quiesced_regions));
for (h = 0; h < rh->nr_buckets; h++) {
list_for_each_entry_safe(reg, nreg, rh->buckets + h, hash_list) {
BUG_ON(atomic_read(®->pending));
mempool_free(reg, rh->region_pool);
}
}
if (rh->log)
dm_dirty_log_destroy(rh->log);
if (rh->region_pool)
mempool_destroy(rh->region_pool);
vfree(rh->buckets);
}
#define RH_HASH_MULT 2654435387U
static inline unsigned int rh_hash(struct region_hash *rh, region_t region)
{
return (unsigned int) ((region * RH_HASH_MULT) >> 12) & rh->mask;
}
static struct region *__rh_lookup(struct region_hash *rh, region_t region)
{
struct region *reg;
list_for_each_entry (reg, rh->buckets + rh_hash(rh, region), hash_list)
if (reg->key == region)
return reg;
return NULL;
}
static void __rh_insert(struct region_hash *rh, struct region *reg)
{
unsigned int h = rh_hash(rh, reg->key);
list_add(®->hash_list, rh->buckets + h);
}
static struct region *__rh_alloc(struct region_hash *rh, region_t region)
{
struct region *reg, *nreg;
read_unlock(&rh->hash_lock);
nreg = mempool_alloc(rh->region_pool, GFP_ATOMIC);
if (unlikely(!nreg))
nreg = kmalloc(sizeof(struct region), GFP_NOIO);
nreg->state = rh->log->type->in_sync(rh->log, region, 1) ?
RH_CLEAN : RH_NOSYNC;
nreg->rh = rh;
nreg->key = region;
INIT_LIST_HEAD(&nreg->list);
atomic_set(&nreg->pending, 0);
bio_list_init(&nreg->delayed_bios);
write_lock_irq(&rh->hash_lock);
reg = __rh_lookup(rh, region);
if (reg)
/* we lost the race */
mempool_free(nreg, rh->region_pool);
else {
__rh_insert(rh, nreg);
if (nreg->state == RH_CLEAN) {
spin_lock(&rh->region_lock);
list_add(&nreg->list, &rh->clean_regions);
spin_unlock(&rh->region_lock);
}
reg = nreg;
}
write_unlock_irq(&rh->hash_lock);
read_lock(&rh->hash_lock);
return reg;
}
static inline struct region *__rh_find(struct region_hash *rh, region_t region)
{
struct region *reg;
reg = __rh_lookup(rh, region);
if (!reg)
reg = __rh_alloc(rh, region);
return reg;
}
static int rh_state(struct region_hash *rh, region_t region, int may_block)
{
int r;
struct region *reg;
read_lock(&rh->hash_lock);
reg = __rh_lookup(rh, region);
read_unlock(&rh->hash_lock);
if (reg)
return reg->state;
/*
* The region wasn't in the hash, so we fall back to the
* dirty log.
*/
r = rh->log->type->in_sync(rh->log, region, may_block);
/*
* Any error from the dirty log (eg. -EWOULDBLOCK) gets
* taken as a RH_NOSYNC
*/
return r == 1 ? RH_CLEAN : RH_NOSYNC;
}
static inline int rh_in_sync(struct region_hash *rh,
region_t region, int may_block)
{
int state = rh_state(rh, region, may_block);
return state == RH_CLEAN || state == RH_DIRTY;
}
static void dispatch_bios(struct mirror_set *ms, struct bio_list *bio_list)
{
struct bio *bio;
while ((bio = bio_list_pop(bio_list))) {
queue_bio(ms, bio, WRITE);
}
}
static void complete_resync_work(struct region *reg, int success)
{
struct region_hash *rh = reg->rh;
rh->log->type->set_region_sync(rh->log, reg->key, success);
/*
* Dispatch the bios before we call 'wake_up_all'.
* This is important because if we are suspending,
* we want to know that recovery is complete and
* the work queue is flushed. If we wake_up_all
* before we dispatch_bios (queue bios and call wake()),
* then we risk suspending before the work queue
* has been properly flushed.
*/
dispatch_bios(rh->ms, ®->delayed_bios);
if (atomic_dec_and_test(&rh->recovery_in_flight))
wake_up_all(&_kmirrord_recovery_stopped);
up(&rh->recovery_count);
}
static void rh_update_states(struct region_hash *rh)
{
struct region *reg, *next;
LIST_HEAD(clean);
LIST_HEAD(recovered);
LIST_HEAD(failed_recovered);
/*
* Quickly grab the lists.
*/
write_lock_irq(&rh->hash_lock);
spin_lock(&rh->region_lock);
if (!list_empty(&rh->clean_regions)) {
list_splice_init(&rh->clean_regions, &clean);
list_for_each_entry(reg, &clean, list)
list_del(®->hash_list);
}
if (!list_empty(&rh->recovered_regions)) {
list_splice_init(&rh->recovered_regions, &recovered);
list_for_each_entry (reg, &recovered, list)
list_del(®->hash_list);
}
if (!list_empty(&rh->failed_recovered_regions)) {
list_splice_init(&rh->failed_recovered_regions,
&failed_recovered);
list_for_each_entry(reg, &failed_recovered, list)
list_del(®->hash_list);
}
spin_unlock(&rh->region_lock);
write_unlock_irq(&rh->hash_lock);
/*
* All the regions on the recovered and clean lists have
* now been pulled out of the system, so no need to do
* any more locking.
*/
list_for_each_entry_safe (reg, next, &recovered, list) {
rh->log->type->clear_region(rh->log, reg->key);
complete_resync_work(reg, 1);
mempool_free(reg, rh->region_pool);
}
list_for_each_entry_safe(reg, next, &failed_recovered, list) {
complete_resync_work(reg, errors_handled(rh->ms) ? 0 : 1);
mempool_free(reg, rh->region_pool);
}
list_for_each_entry_safe(reg, next, &clean, list) {
rh->log->type->clear_region(rh->log, reg->key);
mempool_free(reg, rh->region_pool);
}
rh->log->type->flush(rh->log);
}
static void rh_inc(struct region_hash *rh, region_t region)
{
struct region *reg;
read_lock(&rh->hash_lock);
reg = __rh_find(rh, region);
spin_lock_irq(&rh->region_lock);
atomic_inc(®->pending);
if (reg->state == RH_CLEAN) {
reg->state = RH_DIRTY;
list_del_init(®->list); /* take off the clean list */
spin_unlock_irq(&rh->region_lock);
rh->log->type->mark_region(rh->log, reg->key);
} else
spin_unlock_irq(&rh->region_lock);
read_unlock(&rh->hash_lock);
}
static void rh_inc_pending(struct region_hash *rh, struct bio_list *bios)
{
struct bio *bio;
for (bio = bios->head; bio; bio = bio->bi_next)
rh_inc(rh, bio_to_region(rh, bio));
}
static void rh_dec(struct region_hash *rh, region_t region)
{
unsigned long flags;
struct region *reg;
int should_wake = 0;
read_lock(&rh->hash_lock);
reg = __rh_lookup(rh, region);
read_unlock(&rh->hash_lock);
spin_lock_irqsave(&rh->region_lock, flags);
if (atomic_dec_and_test(®->pending)) {
/*
* There is no pending I/O for this region.
* We can move the region to corresponding list for next action.
* At this point, the region is not yet connected to any list.
*
* If the state is RH_NOSYNC, the region should be kept off
* from clean list.
* The hash entry for RH_NOSYNC will remain in memory
* until the region is recovered or the map is reloaded.
*/
/* do nothing for RH_NOSYNC */
if (reg->state == RH_RECOVERING) {
list_add_tail(®->list, &rh->quiesced_regions);
} else if (reg->state == RH_DIRTY) {
reg->state = RH_CLEAN;
list_add(®->list, &rh->clean_regions);
}
should_wake = 1;
}
spin_unlock_irqrestore(&rh->region_lock, flags);
if (should_wake)
wake(rh->ms);
}
/*
* Starts quiescing a region in preparation for recovery.
*/
static int __rh_recovery_prepare(struct region_hash *rh)
{
int r;
struct region *reg;
region_t region;
/*
* Ask the dirty log what's next.
*/
r = rh->log->type->get_resync_work(rh->log, ®ion);
if (r <= 0)
return r;
/*
* Get this region, and start it quiescing by setting the
* recovering flag.
*/
read_lock(&rh->hash_lock);
reg = __rh_find(rh, region);
read_unlock(&rh->hash_lock);
spin_lock_irq(&rh->region_lock);
reg->state = RH_RECOVERING;
/* Already quiesced ? */
if (atomic_read(®->pending))
list_del_init(®->list);
else
list_move(®->list, &rh->quiesced_regions);
spin_unlock_irq(&rh->region_lock);
return 1;
}
static void rh_recovery_prepare(struct region_hash *rh)
{
/* Extra reference to avoid race with rh_stop_recovery */
atomic_inc(&rh->recovery_in_flight);
while (!down_trylock(&rh->recovery_count)) {
atomic_inc(&rh->recovery_in_flight);
if (__rh_recovery_prepare(rh) <= 0) {
atomic_dec(&rh->recovery_in_flight);
up(&rh->recovery_count);
break;
}
}
/* Drop the extra reference */
if (atomic_dec_and_test(&rh->recovery_in_flight))
wake_up_all(&_kmirrord_recovery_stopped);
}
/*
* Returns any quiesced regions.
*/
static struct region *rh_recovery_start(struct region_hash *rh)
{
struct region *reg = NULL;
spin_lock_irq(&rh->region_lock);
if (!list_empty(&rh->quiesced_regions)) {
reg = list_entry(rh->quiesced_regions.next,
struct region, list);
list_del_init(®->list); /* remove from the quiesced list */
}
spin_unlock_irq(&rh->region_lock);
return reg;
}
static void rh_recovery_end(struct region *reg, int success)
{
struct region_hash *rh = reg->rh;
spin_lock_irq(&rh->region_lock);
if (success)
list_add(®->list, ®->rh->recovered_regions);
else {
reg->state = RH_NOSYNC;
list_add(®->list, ®->rh->failed_recovered_regions);
}
spin_unlock_irq(&rh->region_lock);
wake(rh->ms);
}
static int rh_flush(struct region_hash *rh)
{
return rh->log->type->flush(rh->log);
}
static void rh_delay(struct region_hash *rh, struct bio *bio)
{
struct region *reg;
read_lock(&rh->hash_lock);
reg = __rh_find(rh, bio_to_region(rh, bio));
bio_list_add(®->delayed_bios, bio);
read_unlock(&rh->hash_lock);
}
static void rh_stop_recovery(struct region_hash *rh)
{
int i;
/* wait for any recovering regions */
for (i = 0; i < MAX_RECOVERY; i++)
down(&rh->recovery_count);
}
static void rh_start_recovery(struct region_hash *rh)
{
int i;
for (i = 0; i < MAX_RECOVERY; i++)
up(&rh->recovery_count);
wake(rh->ms);
}
#define MIN_READ_RECORDS 20
struct dm_raid1_read_record {
struct mirror *m;
struct dm_bio_details details;
};
/*
* Every mirror should look like this one.
*/
#define DEFAULT_MIRROR 0
/*
* This is yucky. We squirrel the mirror struct away inside
* bi_next for read/write buffers. This is safe since the bh
* doesn't get submitted to the lower levels of block layer.
*/
static struct mirror *bio_get_m(struct bio *bio)
{
return (struct mirror *) bio->bi_next;
}
static void bio_set_m(struct bio *bio, struct mirror *m)
{
bio->bi_next = (struct bio *) m;
}
static struct mirror *get_default_mirror(struct mirror_set *ms)
{
return &ms->mirror[atomic_read(&ms->default_mirror)];
}
static void set_default_mirror(struct mirror *m)
{
struct mirror_set *ms = m->ms;
struct mirror *m0 = &(ms->mirror[0]);
atomic_set(&ms->default_mirror, m - m0);
}
/* fail_mirror
* @m: mirror device to fail
* @error_type: one of the enum's, DM_RAID1_*_ERROR
*
* If errors are being handled, record the type of
* error encountered for this device. If this type
* of error has already been recorded, we can return;
* otherwise, we must signal userspace by triggering
* an event. Additionally, if the device is the
* primary device, we must choose a new primary, but
* only if the mirror is in-sync.
*
* This function must not block.
*/
static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
{
struct mirror_set *ms = m->ms;
struct mirror *new;
if (!errors_handled(ms))
return;
/*
* error_count is used for nothing more than a
* simple way to tell if a device has encountered
* errors.
*/
atomic_inc(&m->error_count);
if (test_and_set_bit(error_type, &m->error_type))
return;
if (m != get_default_mirror(ms))
goto out;
if (!ms->in_sync) {
/*
* Better to issue requests to same failing device
* than to risk returning corrupt data.
*/
DMERR("Primary mirror (%s) failed while out-of-sync: "
"Reads may fail.", m->dev->name);
goto out;
}
for (new = ms->mirror; new < ms->mirror + ms->nr_mirrors; new++)
if (!atomic_read(&new->error_count)) {
set_default_mirror(new);
break;
}
if (unlikely(new == ms->mirror + ms->nr_mirrors))
DMWARN("All sides of mirror have failed.");
out:
schedule_work(&ms->trigger_event);
}
/*-----------------------------------------------------------------
* Recovery.
*
* When a mirror is first activated we may find that some regions
* are in the no-sync state. We have to recover these by
* recopying from the default mirror to all the others.
*---------------------------------------------------------------*/
static void recovery_complete(int read_err, unsigned long write_err,
void *context)
{
struct region *reg = (struct region *)context;
struct mirror_set *ms = reg->rh->ms;
int m, bit = 0;
if (read_err) {
/* Read error means the failure of default mirror. */
DMERR_LIMIT("Unable to read primary mirror during recovery");
fail_mirror(get_default_mirror(ms), DM_RAID1_SYNC_ERROR);
}
if (write_err) {
DMERR_LIMIT("Write error during recovery (error = 0x%lx)",
write_err);
/*
* Bits correspond to devices (excluding default mirror).
* The default mirror cannot change during recovery.
*/
for (m = 0; m < ms->nr_mirrors; m++) {
if (&ms->mirror[m] == get_default_mirror(ms))
continue;
if (test_bit(bit, &write_err))
fail_mirror(ms->mirror + m,
DM_RAID1_SYNC_ERROR);
bit++;
}
}
rh_recovery_end(reg, !(read_err || write_err));
}
static int recover(struct mirror_set *ms, struct region *reg)
{
int r;
unsigned int i;
struct dm_io_region from, to[DM_KCOPYD_MAX_REGIONS], *dest;
struct mirror *m;
unsigned long flags = 0;
/* fill in the source */
m = get_default_mirror(ms);
from.bdev = m->dev->bdev;
from.sector = m->offset + region_to_sector(reg->rh, reg->key);
if (reg->key == (ms->nr_regions - 1)) {
/*
* The final region may be smaller than
* region_size.
*/
from.count = ms->ti->len & (reg->rh->region_size - 1);
if (!from.count)
from.count = reg->rh->region_size;
} else
from.count = reg->rh->region_size;
/* fill in the destinations */
for (i = 0, dest = to; i < ms->nr_mirrors; i++) {
if (&ms->mirror[i] == get_default_mirror(ms))
continue;
m = ms->mirror + i;
dest->bdev = m->dev->bdev;
dest->sector = m->offset + region_to_sector(reg->rh, reg->key);
dest->count = from.count;
dest++;
}
/* hand to kcopyd */
set_bit(DM_KCOPYD_IGNORE_ERROR, &flags);
r = dm_kcopyd_copy(ms->kcopyd_client, &from, ms->nr_mirrors - 1, to,
flags, recovery_complete, reg);
return r;
}
static void do_recovery(struct mirror_set *ms)
{
int r;
struct region *reg;
struct dm_dirty_log *log = ms->rh.log;
/*
* Start quiescing some regions.
*/
rh_recovery_prepare(&ms->rh);
/*
* Copy any already quiesced regions.
*/
while ((reg = rh_recovery_start(&ms->rh))) {
r = recover(ms, reg);
if (r)
rh_recovery_end(reg, 0);
}
/*
* Update the in sync flag.
*/
if (!ms->in_sync &&
(log->type->get_sync_count(log) == ms->nr_regions)) {
/* the sync is complete */
dm_table_event(ms->ti->table);
ms->in_sync = 1;
}
}
/*-----------------------------------------------------------------
* Reads
*---------------------------------------------------------------*/
static struct mirror *choose_mirror(struct mirror_set *ms, sector_t sector)
{
struct mirror *m = get_default_mirror(ms);
do {
if (likely(!atomic_read(&m->error_count)))
return m;
if (m-- == ms->mirror)
m += ms->nr_mirrors;
} while (m != get_default_mirror(ms));
return NULL;
}
static int default_ok(struct mirror *m)
{
struct mirror *default_mirror = get_default_mirror(m->ms);
return !atomic_read(&default_mirror->error_count);
}
static int mirror_available(struct mirror_set *ms, struct bio *bio)
{
region_t region = bio_to_region(&ms->rh, bio);
if (ms->rh.log->type->in_sync(ms->rh.log, region, 0))
return choose_mirror(ms, bio->bi_sector) ? 1 : 0;
return 0;
}
/*
* remap a buffer to a particular mirror.
*/
static sector_t map_sector(struct mirror *m, struct bio *bio)
{
return m->offset + (bio->bi_sector - m->ms->ti->begin);
}
static void map_bio(struct mirror *m, struct bio *bio)
{
bio->bi_bdev = m->dev->bdev;
bio->bi_sector = map_sector(m, bio);
}
static void map_region(struct dm_io_region *io, struct mirror *m,
struct bio *bio)
{
io->bdev = m->dev->bdev;
io->sector = map_sector(m, bio);
io->count = bio->bi_size >> 9;
}
/*-----------------------------------------------------------------
* Reads
*---------------------------------------------------------------*/
static void read_callback(unsigned long error, void *context)
{
struct bio *bio = context;
struct mirror *m;
m = bio_get_m(bio);
bio_set_m(bio, NULL);
if (likely(!error)) {
bio_endio(bio, 0);
return;
}
fail_mirror(m, DM_RAID1_READ_ERROR);
if (likely(default_ok(m)) || mirror_available(m->ms, bio)) {
DMWARN_LIMIT("Read failure on mirror device %s. "
"Trying alternative device.",
m->dev->name);
queue_bio(m->ms, bio, bio_rw(bio));
return;
}
DMERR_LIMIT("Read failure on mirror device %s. Failing I/O.",
m->dev->name);
bio_endio(bio, -EIO);
}
/* Asynchronous read. */
static void read_async_bio(struct mirror *m, struct bio *bio)
{
struct dm_io_region io;
struct dm_io_request io_req = {
.bi_rw = READ,
.mem.type = DM_IO_BVEC,
.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
.notify.fn = read_callback,
.notify.context = bio,
.client = m->ms->io_client,
};
map_region(&io, m, bio);
bio_set_m(bio, m);
(void) dm_io(&io_req, 1, &io, NULL);
}
static void do_reads(struct mirror_set *ms, struct bio_list *reads)
{
region_t region;
struct bio *bio;
struct mirror *m;
while ((bio = bio_list_pop(reads))) {
region = bio_to_region(&ms->rh, bio);
m = get_default_mirror(ms);
/*
* We can only read balance if the region is in sync.