aboutsummaryrefslogtreecommitdiff
path: root/drivers/md/dm-raid1.c
blob: 752a29e1855b327f13ee24fa9fcfb4478fe32b29 (plain)
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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
/*
 * Copyright (C) 2003 Sistina Software Limited.
 * Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved.
 *
 * This file is released under the GPL.
 */

#include "dm-bio-record.h"

#include <linux/init.h>
#include <linux/mempool.h>
#include <linux/module.h>
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/workqueue.h>
#include <linux/device-mapper.h>
#include <linux/dm-io.h>
#include <linux/dm-dirty-log.h>
#include <linux/dm-kcopyd.h>
#include <linux/dm-region-hash.h>

#define DM_MSG_PREFIX "raid1"

#define MAX_RECOVERY 1	/* Maximum number of regions recovered in parallel. */
#define DM_IO_PAGES 64
#define DM_KCOPYD_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);

/*-----------------------------------------------------------------
 * 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;

	uint64_t features;

	spinlock_t lock;	/* protects the lists */
	struct bio_list reads;
	struct bio_list writes;
	struct bio_list failures;

	struct dm_region_hash *rh;
	struct dm_kcopyd_client *kcopyd_client;
	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 nr_mirrors;
	struct mirror mirror[0];
};

static void wakeup_mirrord(void *context)
{
	struct mirror_set *ms = context;

	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);
	wakeup_mirrord(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);
}

static void wakeup_all_recovery_waiters(void *context)
{
	wake_up_all(&_kmirrord_recovery_stopped);
}

static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw)
{
	unsigned long flags;
	int should_wake = 0;
	struct bio_list *bl;

	bl = (rw == WRITE) ? &ms->writes : &ms->reads;
	spin_lock_irqsave(&ms->lock, flags);
	should_wake = !(bl->head);
	bio_list_add(bl, bio);
	spin_unlock_irqrestore(&ms->lock, flags);

	if (should_wake)
		wakeup_mirrord(ms);
}

static void dispatch_bios(void *context, struct bio_list *bio_list)
{
	struct mirror_set *ms = context;
	struct bio *bio;

	while ((bio = bio_list_pop(bio_list)))
		queue_bio(ms, bio, WRITE);
}

#define MIN_READ_RECORDS 20
struct dm_raid1_read_record {
	struct mirror *m;
	struct dm_bio_details details;
};

static struct kmem_cache *_dm_raid1_read_record_cache;

/*
 * 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;

	/*
	 * 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 (!errors_handled(ms))
		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 dm_region *reg = context;
	struct mirror_set *ms = dm_rh_region_context(reg);
	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++;
		}
	}

	dm_rh_recovery_end(reg, !(read_err || write_err));
}

static int recover(struct mirror_set *ms, struct dm_region *reg)
{
	int r;
	unsigned i;
	struct dm_io_region from, to[DM_KCOPYD_MAX_REGIONS], *dest;
	struct mirror *m;
	unsigned long flags = 0;
	region_t key = dm_rh_get_region_key(reg);
	sector_t region_size = dm_rh_get_region_size(ms->rh);

	/* fill in the source */
	m = get_default_mirror(ms);
	from.bdev = m->dev->bdev;
	from.sector = m->offset + dm_rh_region_to_sector(ms->rh, key);
	if (key == (ms->nr_regions - 1)) {
		/*
		 * The final region may be smaller than
		 * region_size.
		 */
		from.count = ms->ti->len & (region_size - 1);
		if (!from.count)
			from.count = region_size;
	} else
		from.count = 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 + dm_rh_region_to_sector(ms->rh, key);
		dest->count = from.count;
		dest++;
	}

	/* hand to kcopyd */
	if (!errors_handled(ms))
		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)
{
	struct dm_region *reg;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
	int r;

	/*
	 * Start quiescing some regions.
	 */
	dm_rh_recovery_prepare(ms->rh);

	/*
	 * Copy any already quiesced regions.
	 */
	while ((reg = dm_rh_recovery_start(ms->rh))) {
		r = recover(ms, reg);
		if (r)
			dm_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)
{
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
	region_t region = dm_rh_bio_to_region(ms->rh, bio);

	if (log->type->in_sync(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)
{
	if (unlikely(!bio->bi_size))
		return 0;
	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);
	BUG_ON(dm_io(&io_req, 1, &io, NULL));
}

static inline int region_in_sync(struct mirror_set *ms, region_t region,
				 int may_block)
{
	int state = dm_rh_get_state(ms->rh, region, may_block);
	return state == DM_RH_CLEAN || state == DM_RH_DIRTY;
}

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 = dm_rh_bio_to_region(ms->rh, bio);
		m = get_default_mirror(ms);

		/*
		 * We can only read balance if the region is in sync.
		 */
		if (likely(region_in_sync(ms, region, 1)))
			m = choose_mirror(ms, bio->bi_sector);
		else if (m && atomic_read(&m->error_count))
			m = NULL;

		if (likely(m))
			read_async_bio(m, bio);
		else
			bio_endio(bio, -EIO);
	}
}

/*-----------------------------------------------------------------
 * Writes.
 *
 * We do different things with the write io depending on the
 * state of the region that it's in:
 *
 * SYNC: 	increment pending, use kcopyd to write to *all* mirrors
 * RECOVERING:	delay the io until recovery completes
 * NOSYNC:	increment pending, just write to the default mirror
 *---------------------------------------------------------------*/


static void write_callback(unsigned long error, void *context)
{
	unsigned i, ret = 0;
	struct bio *bio = (struct bio *) context;
	struct mirror_set *ms;
	int uptodate = 0;
	int should_wake = 0;
	unsigned long flags;

	ms = bio_get_m(bio)->ms;
	bio_set_m(bio, NULL);

	/*
	 * NOTE: We don't decrement the pending count here,
	 * instead it is done by the targets endio function.
	 * This way we handle both writes to SYNC and NOSYNC
	 * regions with the same code.
	 */
	if (likely(!error))
		goto out;

	for (i = 0; i < ms->nr_mirrors; i++)
		if (test_bit(i, &error))
			fail_mirror(ms->mirror + i, DM_RAID1_WRITE_ERROR);
		else
			uptodate = 1;

	if (unlikely(!uptodate)) {
		DMERR("All replicated volumes dead, failing I/O");
		/* None of the writes succeeded, fail the I/O. */
		ret = -EIO;
	} else if (errors_handled(ms)) {
		/*
		 * Need to raise event.  Since raising
		 * events can block, we need to do it in
		 * the main thread.
		 */
		spin_lock_irqsave(&ms->lock, flags);
		if (!ms->failures.head)
			should_wake = 1;
		bio_list_add(&ms->failures, bio);
		spin_unlock_irqrestore(&ms->lock, flags);
		if (should_wake)
			wakeup_mirrord(ms);
		return;
	}
out:
	bio_endio(bio, ret);
}

static void do_write(struct mirror_set *ms, struct bio *bio)
{
	unsigned int i;
	struct dm_io_region io[ms->nr_mirrors], *dest = io;
	struct mirror *m;
	struct dm_io_request io_req = {
		.bi_rw = WRITE | (bio->bi_rw & WRITE_BARRIER),
		.mem.type = DM_IO_BVEC,
		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
		.notify.fn = write_callback,
		.notify.context = bio,
		.client = ms->io_client,
	};

	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++)
		map_region(dest++, m, bio);

	/*
	 * Use default mirror because we only need it to retrieve the reference
	 * to the mirror set in write_callback().
	 */
	bio_set_m(bio, get_default_mirror(ms));

	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
}

static void do_writes(struct mirror_set *ms, struct bio_list *writes)
{
	int state;
	struct bio *bio;
	struct bio_list sync, nosync, recover, *this_list = NULL;
	struct bio_list requeue;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
	region_t region;

	if (!writes->head)
		return;

	/*
	 * Classify each write.
	 */
	bio_list_init(&sync);
	bio_list_init(&nosync);
	bio_list_init(&recover);
	bio_list_init(&requeue);

	while ((bio = bio_list_pop(writes))) {
		if (unlikely(bio_empty_barrier(bio))) {
			bio_list_add(&sync, bio);
			continue;
		}

		region = dm_rh_bio_to_region(ms->rh, bio);

		if (log->type->is_remote_recovering &&
		    log->type->is_remote_recovering(log, region)) {
			bio_list_add(&requeue, bio);
			continue;
		}

		state = dm_rh_get_state(ms->rh, region, 1);
		switch (state) {
		case DM_RH_CLEAN:
		case DM_RH_DIRTY:
			this_list = &sync;
			break;

		case DM_RH_NOSYNC:
			this_list = &nosync;
			break;

		case DM_RH_RECOVERING:
			this_list = &recover;
			break;
		}

		bio_list_add(this_list, bio);
	}

	/*
	 * Add bios that are delayed due to remote recovery
	 * back on to the write queue
	 */
	if (unlikely(requeue.head)) {
		spin_lock_irq(&ms->lock);
		bio_list_merge(&ms->writes, &requeue);
		spin_unlock_irq(&ms->lock);
		delayed_wake(ms);
	}

	/*
	 * Increment the pending counts for any regions that will
	 * be written to (writes to recover regions are going to
	 * be delayed).
	 */
	dm_rh_inc_pending(ms->rh, &sync);
	dm_rh_inc_pending(ms->rh, &nosync);

	/*
	 * If the flush fails on a previous call and succeeds here,
	 * we must not reset the log_failure variable.  We need
	 * userspace interaction to do that.
	 */
	ms->log_failure = dm_rh_flush(ms->rh) ? 1 : ms->log_failure;

	/*
	 * Dispatch io.
	 */
	if (unlikely(ms->log_failure)) {
		spin_lock_irq(&ms->lock);
		bio_list_merge(&ms->failures, &sync);
		spin_unlock_irq(&ms->lock);
		wakeup_mirrord(ms);
	} else
		while ((bio = bio_list_pop(&sync)))
			do_write(ms, bio);

	while ((bio = bio_list_pop(&recover)))
		dm_rh_delay(ms->rh, bio);

	while ((bio = bio_list_pop(&nosync))) {
		map_bio(get_default_mirror(ms), bio);
		generic_make_request(bio);
	}
}

static void do_failures(struct mirror_set *ms, struct bio_list *failures)
{
	struct bio *bio;

	if (!failures->head)
		return;

	if (!ms->log_failure) {
		while ((bio = bio_list_pop(failures))) {
			ms->in_sync = 0;
			dm_rh_mark_nosync(ms->rh, bio, bio->bi_size, 0);
		}
		return;
	}

	/*
	 * If the log has failed, unattempted writes are being
	 * put on the failures list.  We can't issue those writes
	 * until a log has been marked, so we must store them.
	 *
	 * If a 'noflush' suspend is in progress, we can requeue
	 * the I/O's to the core.  This give userspace a chance
	 * to reconfigure the mirror, at which point the core
	 * will reissue the writes.  If the 'noflush' flag is
	 * not set, we have no choice but to return errors.
	 *
	 * Some writes on the failures list may have been
	 * submitted before the log failure and represent a
	 * failure to write to one of the devices.  It is ok
	 * for us to treat them the same and requeue them
	 * as well.
	 */
	if (dm_noflush_suspending(ms->ti)) {
		while ((bio = bio_list_pop(failures)))
			bio_endio(bio, DM_ENDIO_REQUEUE);
		return;
	}

	if (atomic_read(&ms->suspend)) {
		while ((bio = bio_list_pop(failures)))
			bio_endio(bio, -EIO);
		return;
	}

	spin_lock_irq(&ms->lock);
	bio_list_merge(&ms->failures, failures);
	spin_unlock_irq(&ms->lock);

	delayed_wake(ms);
}

static void trigger_event(struct work_struct *work)
{
	struct mirror_set *ms =
		container_of(work, struct mirror_set, trigger_event);

	dm_table_event(ms->ti->table);
}

/*-----------------------------------------------------------------
 * kmirrord
 *---------------------------------------------------------------*/
static void do_mirror(struct work_struct *work)
{
	struct mirror_set *ms = container_of(work, struct mirror_set,
					     kmirrord_work);
	struct bio_list reads, writes, failures;
	unsigned long flags;

	spin_lock_irqsave(&ms->lock, flags);
	reads = ms->reads;
	writes = ms->writes;
	failures = ms->failures;
	bio_list_init(&ms->reads);
	bio_list_init(&ms->writes);
	bio_list_init(&ms->failures);
	spin_unlock_irqrestore(&ms->lock, flags);

	dm_rh_update_states(ms->rh, errors_handled(ms));
	do_recovery(ms);
	do_reads(ms, &reads);
	do_writes(ms, &writes);
	do_failures(ms, &failures);

	dm_table_unplug_all(ms->ti->table);
}

/*-----------------------------------------------------------------
 * Target functions
 *---------------------------------------------------------------*/
static struct mirror_set *alloc_context(unsigned int nr_mirrors,
					uint32_t region_size,
					struct dm_target *ti,
					struct dm_dirty_log *dl)
{
	size_t len;
	struct mirror_set *ms = NULL;

	len = sizeof(*ms) + (sizeof(ms->mirror[0]) * nr_mirrors);

	ms = kzalloc(len, GFP_KERNEL);
	if (!ms) {
		ti->error = "Cannot allocate mirror context";
		return NULL;
	}

	spin_lock_init(&ms->lock);

	ms->ti = ti;
	ms->nr_mirrors = nr_mirrors;
	ms->nr_regions = dm_sector_div_up(ti->len, region_size);
	ms->in_sync = 0;
	ms->log_failure = 0;
	atomic_set(&ms->suspend, 0);
	atomic_set(&ms->default_mirror, DEFAULT_MIRROR);

	ms->read_record_pool = mempool_create_slab_pool(MIN_READ_RECORDS,
						_dm_raid1_read_record_cache);

	if (!ms->read_record_pool) {
		ti->error = "Error creating mirror read_record_pool";
		kfree(ms);
		return NULL;
	}

	ms->io_client = dm_io_client_create(DM_IO_PAGES);
	if (IS_ERR(ms->io_client)) {
		ti->error = "Error creating dm_io client";
		mempool_destroy(ms->read_record_pool);
		kfree(ms);
 		return NULL;
	}

	ms->rh = dm_region_hash_create(ms, dispatch_bios, wakeup_mirrord,
				       wakeup_all_recovery_waiters,
				       ms->ti->begin, MAX_RECOVERY,
				       dl, region_size, ms->nr_regions);
	if (IS_ERR(ms->rh)) {
		ti->error = "Error creating dirty region hash";
		dm_io_client_destroy(ms->io_client);
		mempool_destroy(ms->read_record_pool);
		kfree(ms);
		return NULL;
	}

	return ms;
}

static void free_context(struct mirror_set *ms, struct dm_target *ti,
			 unsigned int m)
{
	while (m--)
		dm_put_device(ti, ms->mirror[m].dev);

	dm_io_client_destroy(ms->io_client);
	dm_region_hash_destroy(ms->rh);
	mempool_destroy(ms->read_record_pool);
	kfree(ms);
}

static int get_mirror(struct mirror_set *ms, struct dm_target *ti,
		      unsigned int mirror, char **argv)
{
	unsigned long long offset;

	if (sscanf(argv[1], "%llu", &offset) != 1) {
		ti->error = "Invalid offset";
		return -EINVAL;
	}

	if (dm_get_device(ti, argv[0], offset, ti->len,
			  dm_table_get_mode(ti->table),
			  &ms->mirror[mirror].dev)) {
		ti->error = "Device lookup failure";
		return -ENXIO;
	}

	ms->mirror[mirror].ms = ms;
	atomic_set(&(ms->mirror[mirror].error_count), 0);
	ms->mirror[mirror].error_type = 0;
	ms->mirror[mirror].offset = offset;

	return 0;
}

/*
 * Create dirty log: log_type #log_params <log_params>
 */
static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
					     unsigned argc, char **argv,
					     unsigned *args_used)
{
	unsigned param_count;
	struct dm_dirty_log *dl;

	if (argc < 2) {
		ti->error = "Insufficient mirror log arguments";
		return NULL;
	}

	if (sscanf(argv[1], "%u", &param_count) != 1) {
		ti->error = "Invalid mirror log argument count";
		return NULL;
	}

	*args_used = 2 + param_count;

	if (argc < *args_used) {
		ti->error = "Insufficient mirror log arguments";
		return NULL;
	}

	dl = dm_dirty_log_create(argv[0], ti, param_count, argv + 2);
	if (!dl) {
		ti->error = "Error creating mirror dirty log";
		return NULL;
	}

	return dl;
}

static int parse_features(struct mirror_set *ms, unsigned argc, char **argv,
			  unsigned *args_used)
{
	unsigned num_features;
	struct dm_target *ti = ms->ti;

	*args_used = 0;

	if (!argc)
		return 0;

	if (sscanf(argv[0], "%u", &num_features) != 1) {
		ti->error = "Invalid number of features";
		return -EINVAL;
	}

	argc--;
	argv++;
	(*args_used)++;

	if (num_features > argc) {
		ti->error = "Not enough arguments to support feature count";
		return -EINVAL;
	}

	if (!strcmp("handle_errors", argv[0]))
		ms->features |= DM_RAID1_HANDLE_ERRORS;
	else {
		ti->error = "Unrecognised feature requested";
		return -EINVAL;
	}

	(*args_used)++;

	return 0;
}

/*
 * Construct a mirror mapping:
 *
 * log_type #log_params <log_params>
 * #mirrors [mirror_path offset]{2,}
 * [#features <features>]
 *
 * log_type is "core" or "disk"
 * #log_params is between 1 and 3
 *
 * If present, features must be "handle_errors".
 */
static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
	int r;
	unsigned int nr_mirrors, m, args_used;
	struct mirror_set *ms;
	struct dm_dirty_log *dl;

	dl = create_dirty_log(ti, argc, argv, &args_used);
	if (!dl)
		return -EINVAL;

	argv += args_used;
	argc -= args_used;

	if (!argc || sscanf(argv[0], "%u", &nr_mirrors) != 1 ||
	    nr_mirrors < 2 || nr_mirrors > DM_KCOPYD_MAX_REGIONS + 1) {
		ti->error = "Invalid number of mirrors";
		dm_dirty_log_destroy(dl);
		return -EINVAL;
	}

	argv++, argc--;

	if (argc < nr_mirrors * 2) {
		ti->error = "Too few mirror arguments";
		dm_dirty_log_destroy(dl);
		return -EINVAL;
	}

	ms = alloc_context(nr_mirrors, dl->type->get_region_size(dl), ti, dl);
	if (!ms) {
		dm_dirty_log_destroy(dl);
		return -ENOMEM;
	}

	/* Get the mirror parameter sets */
	for (m = 0; m < nr_mirrors; m++) {
		r = get_mirror(ms, ti, m, argv);
		if (r) {
			free_context(ms, ti, m);
			return r;
		}
		argv += 2;
		argc -= 2;
	}

	ti->private = ms;
	ti->split_io = dm_rh_get_region_size(ms->rh);
	ti->num_flush_requests = 1;

	ms->kmirrord_wq = create_singlethread_workqueue("kmirrord");
	if (!ms->kmirrord_wq) {
		DMERR("couldn't start kmirrord");
		r = -ENOMEM;
		goto err_free_context;
	}
	INIT_WORK(&ms->kmirrord_work, do_mirror);
	init_timer(&ms->timer);
	ms->timer_pending = 0;
	INIT_WORK(&ms->trigger_event, trigger_event);

	r = parse_features(ms, argc, argv, &args_used);
	if (r)
		goto err_destroy_wq;

	argv += args_used;
	argc -= args_used;

	/*
	 * Any read-balancing addition depends on the
	 * DM_RAID1_HANDLE_ERRORS flag being present.
	 * This is because the decision to balance depends
	 * on the sync state of a region.  If the above
	 * flag is not present, we ignore errors; and
	 * the sync state may be inaccurate.
	 */

	if (argc) {
		ti->error = "Too many mirror arguments";
		r = -EINVAL;
		goto err_destroy_wq;
	}

	r = dm_kcopyd_client_create(DM_KCOPYD_PAGES, &ms->kcopyd_client);
	if (r)
		goto err_destroy_wq;

	wakeup_mirrord(ms);
	return 0;

err_destroy_wq:
	destroy_workqueue(ms->kmirrord_wq);
err_free_context:
	free_context(ms, ti, ms->nr_mirrors);
	return r;
}

static void mirror_dtr(struct dm_target *ti)
{
	struct mirror_set *ms = (struct mirror_set *) ti->private;

	del_timer_sync(&ms->timer);
	flush_workqueue(ms->kmirrord_wq);
	flush_scheduled_work();
	dm_kcopyd_client_destroy(ms->kcopyd_client);
	destroy_workqueue(ms->kmirrord_wq);
	free_context(ms, ti, ms->nr_mirrors);
}

/*
 * Mirror mapping function
 */
static int mirror_map(struct dm_target *ti, struct bio *bio,
		      union map_info *map_context)
{
	int r, rw = bio_rw(bio);
	struct mirror *m;
	struct mirror_set *ms = ti->private;
	struct dm_raid1_read_record *read_record = NULL;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);

	if (rw == WRITE) {
		/* Save region for mirror_end_io() handler */
		map_context->ll = dm_rh_bio_to_region(ms->rh, bio);
		queue_bio(ms, bio, rw);
		return DM_MAPIO_SUBMITTED;
	}

	r = log->type->in_sync(log, dm_rh_bio_to_region(ms->rh, bio), 0);
	if (r < 0 && r != -EWOULDBLOCK)
		return r;

	/*
	 * If region is not in-sync queue the bio.
	 */
	if (!r || (r == -EWOULDBLOCK)) {
		if (rw == READA)
			return -EWOULDBLOCK;

		queue_bio(ms, bio, rw);
		return DM_MAPIO_SUBMITTED;
	}

	/*
	 * The region is in-sync and we can perform reads directly.
	 * Store enough information so we can retry if it fails.
	 */
	m = choose_mirror(ms, bio->bi_sector);
	if (unlikely(!m))
		return -EIO;

	read_record = mempool_alloc(ms->read_record_pool, GFP_NOIO);
	if (likely(read_record)) {
		dm_bio_record(&read_record->details, bio);
		map_context->ptr = read_record;
		read_record->m = m;
	}

	map_bio(m, bio);

	return DM_MAPIO_REMAPPED;
}

static int mirror_end_io(struct dm_target *ti, struct bio *bio,
			 int error, union map_info *map_context)
{
	int rw = bio_rw(bio);
	struct mirror_set *ms = (struct mirror_set *) ti->private;
	struct mirror *m = NULL;
	struct dm_bio_details *bd = NULL;
	struct dm_raid1_read_record *read_record = map_context->ptr;

	/*
	 * We need to dec pending if this was a write.
	 */
	if (rw == WRITE) {
		if (likely(!bio_empty_barrier(bio)))
			dm_rh_dec(ms->rh, map_context->ll);
		return error;
	}

	if (error == -EOPNOTSUPP)
		goto out;

	if ((error == -EWOULDBLOCK) && bio_rw_flagged(bio, BIO_RW_AHEAD))
		goto out;

	if (unlikely(error)) {
		if (!read_record) {
			/*
			 * There wasn't enough memory to record necessary
			 * information for a retry or there was no other
			 * mirror in-sync.
			 */
			DMERR_LIMIT("Mirror read failed.");
			return -EIO;
		}

		m = read_record->m;

		DMERR("Mirror read failed from %s. Trying alternative device.",
		      m->dev->name);

		fail_mirror(m, DM_RAID1_READ_ERROR);

		/*
		 * A failed read is requeued for another attempt using an intact
		 * mirror.
		 */
		if (default_ok(m) || mirror_available(ms, bio)) {
			bd = &read_record->details;

			dm_bio_restore(bd, bio);
			mempool_free(read_record, ms->read_record_pool);
			map_context->ptr = NULL;
			queue_bio(ms, bio, rw);
			return 1;
		}
		DMERR("All replicated volumes dead, failing I/O");
	}

out:
	if (read_record) {
		mempool_free(read_record, ms->read_record_pool);
		map_context->ptr = NULL;
	}

	return error;
}

static void mirror_presuspend(struct dm_target *ti)
{
	struct mirror_set *ms = (struct mirror_set *) ti->private;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);

	atomic_set(&ms->suspend, 1);

	/*
	 * We must finish up all the work that we've
	 * generated (i.e. recovery work).
	 */
	dm_rh_stop_recovery(ms->rh);

	wait_event(_kmirrord_recovery_stopped,
		   !dm_rh_recovery_in_flight(ms->rh));

	if (log->type->presuspend && log->type->presuspend(log))
		/* FIXME: need better error handling */
		DMWARN("log presuspend failed");

	/*
	 * Now that recovery is complete/stopped and the
	 * delayed bios are queued, we need to wait for
	 * the worker thread to complete.  This way,
	 * we know that all of our I/O has been pushed.
	 */
	flush_workqueue(ms->kmirrord_wq);
}

static void mirror_postsuspend(struct dm_target *ti)
{
	struct mirror_set *ms = ti->private;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);

	if (log->type->postsuspend && log->type->postsuspend(log))
		/* FIXME: need better error handling */
		DMWARN("log postsuspend failed");
}

static void mirror_resume(struct dm_target *ti)
{
	struct mirror_set *ms = ti->private;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);

	atomic_set(&ms->suspend, 0);
	if (log->type->resume && log->type->resume(log))
		/* FIXME: need better error handling */
		DMWARN("log resume failed");
	dm_rh_start_recovery(ms->rh);
}

/*
 * device_status_char
 * @m: mirror device/leg we want the status of
 *
 * We return one character representing the most severe error
 * we have encountered.
 *    A => Alive - No failures
 *    D => Dead - A write failure occurred leaving mirror out-of-sync
 *    S => Sync - A sychronization failure occurred, mirror out-of-sync
 *    R => Read - A read failure occurred, mirror data unaffected
 *
 * Returns: <char>
 */
static char device_status_char(struct mirror *m)
{
	if (!atomic_read(&(m->error_count)))
		return 'A';

	return (test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
		(test_bit(DM_RAID1_SYNC_ERROR, &(m->error_type))) ? 'S' :
		(test_bit(DM_RAID1_READ_ERROR, &(m->error_type))) ? 'R' : 'U';
}


static int mirror_status(struct dm_target *ti, status_type_t type,
			 char *result, unsigned int maxlen)
{
	unsigned int m, sz = 0;
	struct mirror_set *ms = (struct mirror_set *) ti->private;
	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
	char buffer[ms->nr_mirrors + 1];

	switch (type) {
	case STATUSTYPE_INFO:
		DMEMIT("%d ", ms->nr_mirrors);
		for (m = 0; m < ms->nr_mirrors; m++) {
			DMEMIT("%s ", ms->mirror[m].dev->name);
			buffer[m] = device_status_char(&(ms->mirror[m]));
		}
		buffer[m] = '\0';

		DMEMIT("%llu/%llu 1 %s ",
		      (unsigned long long)log->type->get_sync_count(log),
		      (unsigned long long)ms->nr_regions, buffer);

		sz += log->type->status(log, type, result+sz, maxlen-sz);

		break;

	case STATUSTYPE_TABLE:
		sz = log->type->status(log, type, result, maxlen);

		DMEMIT("%d", ms->nr_mirrors);
		for (m = 0; m < ms->nr_mirrors; m++)
			DMEMIT(" %s %llu", ms->mirror[m].dev->name,
			       (unsigned long long)ms->mirror[m].offset);

		if (ms->features & DM_RAID1_HANDLE_ERRORS)
			DMEMIT(" 1 handle_errors");
	}

	return 0;
}

static int mirror_iterate_devices(struct dm_target *ti,
				  iterate_devices_callout_fn fn, void *data)
{
	struct mirror_set *ms = ti->private;
	int ret = 0;
	unsigned i;

	for (i = 0; !ret && i < ms->nr_mirrors; i++)
		ret = fn(ti, ms->mirror[i].dev,
			 ms->mirror[i].offset, ti->len, data);

	return ret;
}

static struct target_type mirror_target = {
	.name	 = "mirror",
	.version = {1, 12, 0},
	.module	 = THIS_MODULE,
	.ctr	 = mirror_ctr,
	.dtr	 = mirror_dtr,
	.map	 = mirror_map,
	.end_io	 = mirror_end_io,
	.presuspend = mirror_presuspend,
	.postsuspend = mirror_postsuspend,
	.resume	 = mirror_resume,
	.status	 = mirror_status,
	.iterate_devices = mirror_iterate_devices,
};

static int __init dm_mirror_init(void)
{
	int r;

	_dm_raid1_read_record_cache = KMEM_CACHE(dm_raid1_read_record, 0);
	if (!_dm_raid1_read_record_cache) {
		DMERR("Can't allocate dm_raid1_read_record cache");
		r = -ENOMEM;
		goto bad_cache;
	}

	r = dm_register_target(&mirror_target);
	if (r < 0) {
		DMERR("Failed to register mirror target");
		goto bad_target;
	}

	return 0;

bad_target:
	kmem_cache_destroy(_dm_raid1_read_record_cache);
bad_cache:
	return r;
}

static void __exit dm_mirror_exit(void)
{
	dm_unregister_target(&mirror_target);
	kmem_cache_destroy(_dm_raid1_read_record_cache);
}

/* Module hooks */
module_init(dm_mirror_init);
module_exit(dm_mirror_exit);

MODULE_DESCRIPTION(DM_NAME " mirror target");
MODULE_AUTHOR("Joe Thornber");
MODULE_LICENSE("GPL");