aboutsummaryrefslogtreecommitdiff
path: root/drivers/md/dm-snap.c
blob: 89f8dd1bfaa01684cd569374686e6fb3ef109a36 (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
/*
 * dm-snapshot.c
 *
 * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
 *
 * This file is released under the GPL.
 */

#include <linux/blkdev.h>
#include <linux/ctype.h>
#include <linux/device-mapper.h>
#include <linux/fs.h>
#include <linux/init.h>
#include <linux/kdev_t.h>
#include <linux/list.h>
#include <linux/mempool.h>
#include <linux/module.h>
#include <linux/slab.h>
#include <linux/vmalloc.h>

#include "dm-snap.h"
#include "dm-bio-list.h"
#include "kcopyd.h"

#define DM_MSG_PREFIX "snapshots"

/*
 * The percentage increment we will wake up users at
 */
#define WAKE_UP_PERCENT 5

/*
 * kcopyd priority of snapshot operations
 */
#define SNAPSHOT_COPY_PRIORITY 2

/*
 * Each snapshot reserves this many pages for io
 */
#define SNAPSHOT_PAGES 256

struct pending_exception {
	struct exception e;

	/*
	 * Origin buffers waiting for this to complete are held
	 * in a bio list
	 */
	struct bio_list origin_bios;
	struct bio_list snapshot_bios;

	/*
	 * Short-term queue of pending exceptions prior to submission.
	 */
	struct list_head list;

	/*
	 * The primary pending_exception is the one that holds
	 * the sibling_count and the list of origin_bios for a
	 * group of pending_exceptions.  It is always last to get freed.
	 * These fields get set up when writing to the origin.
	 */
	struct pending_exception *primary_pe;

	/*
	 * Number of pending_exceptions processing this chunk.
	 * When this drops to zero we must complete the origin bios.
	 * If incrementing or decrementing this, hold pe->snap->lock for
	 * the sibling concerned and not pe->primary_pe->snap->lock unless
	 * they are the same.
	 */
	atomic_t sibling_count;

	/* Pointer back to snapshot context */
	struct dm_snapshot *snap;

	/*
	 * 1 indicates the exception has already been sent to
	 * kcopyd.
	 */
	int started;
};

/*
 * Hash table mapping origin volumes to lists of snapshots and
 * a lock to protect it
 */
static kmem_cache_t *exception_cache;
static kmem_cache_t *pending_cache;
static mempool_t *pending_pool;

/*
 * One of these per registered origin, held in the snapshot_origins hash
 */
struct origin {
	/* The origin device */
	struct block_device *bdev;

	struct list_head hash_list;

	/* List of snapshots for this origin */
	struct list_head snapshots;
};

/*
 * Size of the hash table for origin volumes. If we make this
 * the size of the minors list then it should be nearly perfect
 */
#define ORIGIN_HASH_SIZE 256
#define ORIGIN_MASK      0xFF
static struct list_head *_origins;
static struct rw_semaphore _origins_lock;

static int init_origin_hash(void)
{
	int i;

	_origins = kmalloc(ORIGIN_HASH_SIZE * sizeof(struct list_head),
			   GFP_KERNEL);
	if (!_origins) {
		DMERR("unable to allocate memory");
		return -ENOMEM;
	}

	for (i = 0; i < ORIGIN_HASH_SIZE; i++)
		INIT_LIST_HEAD(_origins + i);
	init_rwsem(&_origins_lock);

	return 0;
}

static void exit_origin_hash(void)
{
	kfree(_origins);
}

static inline unsigned int origin_hash(struct block_device *bdev)
{
	return bdev->bd_dev & ORIGIN_MASK;
}

static struct origin *__lookup_origin(struct block_device *origin)
{
	struct list_head *ol;
	struct origin *o;

	ol = &_origins[origin_hash(origin)];
	list_for_each_entry (o, ol, hash_list)
		if (bdev_equal(o->bdev, origin))
			return o;

	return NULL;
}

static void __insert_origin(struct origin *o)
{
	struct list_head *sl = &_origins[origin_hash(o->bdev)];
	list_add_tail(&o->hash_list, sl);
}

/*
 * Make a note of the snapshot and its origin so we can look it
 * up when the origin has a write on it.
 */
static int register_snapshot(struct dm_snapshot *snap)
{
	struct origin *o;
	struct block_device *bdev = snap->origin->bdev;

	down_write(&_origins_lock);
	o = __lookup_origin(bdev);

	if (!o) {
		/* New origin */
		o = kmalloc(sizeof(*o), GFP_KERNEL);
		if (!o) {
			up_write(&_origins_lock);
			return -ENOMEM;
		}

		/* Initialise the struct */
		INIT_LIST_HEAD(&o->snapshots);
		o->bdev = bdev;

		__insert_origin(o);
	}

	list_add_tail(&snap->list, &o->snapshots);

	up_write(&_origins_lock);
	return 0;
}

static void unregister_snapshot(struct dm_snapshot *s)
{
	struct origin *o;

	down_write(&_origins_lock);
	o = __lookup_origin(s->origin->bdev);

	list_del(&s->list);
	if (list_empty(&o->snapshots)) {
		list_del(&o->hash_list);
		kfree(o);
	}

	up_write(&_origins_lock);
}

/*
 * Implementation of the exception hash tables.
 */
static int init_exception_table(struct exception_table *et, uint32_t size)
{
	unsigned int i;

	et->hash_mask = size - 1;
	et->table = dm_vcalloc(size, sizeof(struct list_head));
	if (!et->table)
		return -ENOMEM;

	for (i = 0; i < size; i++)
		INIT_LIST_HEAD(et->table + i);

	return 0;
}

static void exit_exception_table(struct exception_table *et, kmem_cache_t *mem)
{
	struct list_head *slot;
	struct exception *ex, *next;
	int i, size;

	size = et->hash_mask + 1;
	for (i = 0; i < size; i++) {
		slot = et->table + i;

		list_for_each_entry_safe (ex, next, slot, hash_list)
			kmem_cache_free(mem, ex);
	}

	vfree(et->table);
}

static inline uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
{
	return chunk & et->hash_mask;
}

static void insert_exception(struct exception_table *eh, struct exception *e)
{
	struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)];
	list_add(&e->hash_list, l);
}

static inline void remove_exception(struct exception *e)
{
	list_del(&e->hash_list);
}

/*
 * Return the exception data for a sector, or NULL if not
 * remapped.
 */
static struct exception *lookup_exception(struct exception_table *et,
					  chunk_t chunk)
{
	struct list_head *slot;
	struct exception *e;

	slot = &et->table[exception_hash(et, chunk)];
	list_for_each_entry (e, slot, hash_list)
		if (e->old_chunk == chunk)
			return e;

	return NULL;
}

static inline struct exception *alloc_exception(void)
{
	struct exception *e;

	e = kmem_cache_alloc(exception_cache, GFP_NOIO);
	if (!e)
		e = kmem_cache_alloc(exception_cache, GFP_ATOMIC);

	return e;
}

static inline void free_exception(struct exception *e)
{
	kmem_cache_free(exception_cache, e);
}

static inline struct pending_exception *alloc_pending_exception(void)
{
	return mempool_alloc(pending_pool, GFP_NOIO);
}

static inline void free_pending_exception(struct pending_exception *pe)
{
	mempool_free(pe, pending_pool);
}

int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
{
	struct exception *e;

	e = alloc_exception();
	if (!e)
		return -ENOMEM;

	e->old_chunk = old;
	e->new_chunk = new;
	insert_exception(&s->complete, e);
	return 0;
}

/*
 * Hard coded magic.
 */
static int calc_max_buckets(void)
{
	/* use a fixed size of 2MB */
	unsigned long mem = 2 * 1024 * 1024;
	mem /= sizeof(struct list_head);

	return mem;
}

/*
 * Rounds a number down to a power of 2.
 */
static inline uint32_t round_down(uint32_t n)
{
	while (n & (n - 1))
		n &= (n - 1);
	return n;
}

/*
 * Allocate room for a suitable hash table.
 */
static int init_hash_tables(struct dm_snapshot *s)
{
	sector_t hash_size, cow_dev_size, origin_dev_size, max_buckets;

	/*
	 * Calculate based on the size of the original volume or
	 * the COW volume...
	 */
	cow_dev_size = get_dev_size(s->cow->bdev);
	origin_dev_size = get_dev_size(s->origin->bdev);
	max_buckets = calc_max_buckets();

	hash_size = min(origin_dev_size, cow_dev_size) >> s->chunk_shift;
	hash_size = min(hash_size, max_buckets);

	/* Round it down to a power of 2 */
	hash_size = round_down(hash_size);
	if (init_exception_table(&s->complete, hash_size))
		return -ENOMEM;

	/*
	 * Allocate hash table for in-flight exceptions
	 * Make this smaller than the real hash table
	 */
	hash_size >>= 3;
	if (hash_size < 64)
		hash_size = 64;

	if (init_exception_table(&s->pending, hash_size)) {
		exit_exception_table(&s->complete, exception_cache);
		return -ENOMEM;
	}

	return 0;
}

/*
 * Round a number up to the nearest 'size' boundary.  size must
 * be a power of 2.
 */
static inline ulong round_up(ulong n, ulong size)
{
	size--;
	return (n + size) & ~size;
}

static void read_snapshot_metadata(struct dm_snapshot *s)
{
	if (s->store.read_metadata(&s->store)) {
		down_write(&s->lock);
		s->valid = 0;
		up_write(&s->lock);

		dm_table_event(s->table);
	}
}

static int set_chunk_size(struct dm_snapshot *s, const char *chunk_size_arg,
			  char **error)
{
	unsigned long chunk_size;
	char *value;

	chunk_size = simple_strtoul(chunk_size_arg, &value, 10);
	if (*chunk_size_arg == '\0' || *value != '\0') {
		*error = "Invalid chunk size";
		return -EINVAL;
	}

	if (!chunk_size) {
		s->chunk_size = s->chunk_mask = s->chunk_shift = 0;
		return 0;
	}

	/*
	 * Chunk size must be multiple of page size.  Silently
	 * round up if it's not.
	 */
	chunk_size = round_up(chunk_size, PAGE_SIZE >> 9);

	/* Check chunk_size is a power of 2 */
	if (chunk_size & (chunk_size - 1)) {
		*error = "Chunk size is not a power of 2";
		return -EINVAL;
	}

	/* Validate the chunk size against the device block size */
	if (chunk_size % (bdev_hardsect_size(s->cow->bdev) >> 9)) {
		*error = "Chunk size is not a multiple of device blocksize";
		return -EINVAL;
	}

	s->chunk_size = chunk_size;
	s->chunk_mask = chunk_size - 1;
	s->chunk_shift = ffs(chunk_size) - 1;

	return 0;
}

/*
 * Construct a snapshot mapping: <origin_dev> <COW-dev> <p/n> <chunk-size>
 */
static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
	struct dm_snapshot *s;
	int r = -EINVAL;
	char persistent;
	char *origin_path;
	char *cow_path;

	if (argc != 4) {
		ti->error = "requires exactly 4 arguments";
		r = -EINVAL;
		goto bad1;
	}

	origin_path = argv[0];
	cow_path = argv[1];
	persistent = toupper(*argv[2]);

	if (persistent != 'P' && persistent != 'N') {
		ti->error = "Persistent flag is not P or N";
		r = -EINVAL;
		goto bad1;
	}

	s = kmalloc(sizeof(*s), GFP_KERNEL);
	if (s == NULL) {
		ti->error = "Cannot allocate snapshot context private "
		    "structure";
		r = -ENOMEM;
		goto bad1;
	}

	r = dm_get_device(ti, origin_path, 0, ti->len, FMODE_READ, &s->origin);
	if (r) {
		ti->error = "Cannot get origin device";
		goto bad2;
	}

	r = dm_get_device(ti, cow_path, 0, 0,
			  FMODE_READ | FMODE_WRITE, &s->cow);
	if (r) {
		dm_put_device(ti, s->origin);
		ti->error = "Cannot get COW device";
		goto bad2;
	}

	r = set_chunk_size(s, argv[3], &ti->error);
	if (r)
		goto bad3;

	s->type = persistent;

	s->valid = 1;
	s->active = 0;
	s->last_percent = 0;
	init_rwsem(&s->lock);
	s->table = ti->table;

	/* Allocate hash table for COW data */
	if (init_hash_tables(s)) {
		ti->error = "Unable to allocate hash table space";
		r = -ENOMEM;
		goto bad3;
	}

	s->store.snap = s;

	if (persistent == 'P')
		r = dm_create_persistent(&s->store);
	else
		r = dm_create_transient(&s->store);

	if (r) {
		ti->error = "Couldn't create exception store";
		r = -EINVAL;
		goto bad4;
	}

	r = kcopyd_client_create(SNAPSHOT_PAGES, &s->kcopyd_client);
	if (r) {
		ti->error = "Could not create kcopyd client";
		goto bad5;
	}

	/* Metadata must only be loaded into one table at once */
	read_snapshot_metadata(s);

	/* Add snapshot to the list of snapshots for this origin */
	/* Exceptions aren't triggered till snapshot_resume() is called */
	if (register_snapshot(s)) {
		r = -EINVAL;
		ti->error = "Cannot register snapshot origin";
		goto bad6;
	}

	ti->private = s;
	ti->split_io = s->chunk_size;

	return 0;

 bad6:
	kcopyd_client_destroy(s->kcopyd_client);

 bad5:
	s->store.destroy(&s->store);

 bad4:
	exit_exception_table(&s->pending, pending_cache);
	exit_exception_table(&s->complete, exception_cache);

 bad3:
	dm_put_device(ti, s->cow);
	dm_put_device(ti, s->origin);

 bad2:
	kfree(s);

 bad1:
	return r;
}

static void snapshot_dtr(struct dm_target *ti)
{
	struct dm_snapshot *s = (struct dm_snapshot *) ti->private;

	/* Prevent further origin writes from using this snapshot. */
	/* After this returns there can be no new kcopyd jobs. */
	unregister_snapshot(s);

	kcopyd_client_destroy(s->kcopyd_client);

	exit_exception_table(&s->pending, pending_cache);
	exit_exception_table(&s->complete, exception_cache);

	/* Deallocate memory used */
	s->store.destroy(&s->store);

	dm_put_device(ti, s->origin);
	dm_put_device(ti, s->cow);

	kfree(s);
}

/*
 * Flush a list of buffers.
 */
static void flush_bios(struct bio *bio)
{
	struct bio *n;

	while (bio) {
		n = bio->bi_next;
		bio->bi_next = NULL;
		generic_make_request(bio);
		bio = n;
	}
}

/*
 * Error a list of buffers.
 */
static void error_bios(struct bio *bio)
{
	struct bio *n;

	while (bio) {
		n = bio->bi_next;
		bio->bi_next = NULL;
		bio_io_error(bio, bio->bi_size);
		bio = n;
	}
}

static inline void error_snapshot_bios(struct pending_exception *pe)
{
	error_bios(bio_list_get(&pe->snapshot_bios));
}

static struct bio *__flush_bios(struct pending_exception *pe)
{
	/*
	 * If this pe is involved in a write to the origin and
	 * it is the last sibling to complete then release
	 * the bios for the original write to the origin.
	 */

	if (pe->primary_pe &&
	    atomic_dec_and_test(&pe->primary_pe->sibling_count))
		return bio_list_get(&pe->primary_pe->origin_bios);

	return NULL;
}

static void __invalidate_snapshot(struct dm_snapshot *s,
				struct pending_exception *pe, int err)
{
	if (!s->valid)
		return;

	if (err == -EIO)
		DMERR("Invalidating snapshot: Error reading/writing.");
	else if (err == -ENOMEM)
		DMERR("Invalidating snapshot: Unable to allocate exception.");

	if (pe)
		remove_exception(&pe->e);

	if (s->store.drop_snapshot)
		s->store.drop_snapshot(&s->store);

	s->valid = 0;

	dm_table_event(s->table);
}

static void pending_complete(struct pending_exception *pe, int success)
{
	struct exception *e;
	struct pending_exception *primary_pe;
	struct dm_snapshot *s = pe->snap;
	struct bio *flush = NULL;

	if (!success) {
		/* Read/write error - snapshot is unusable */
		down_write(&s->lock);
		__invalidate_snapshot(s, pe, -EIO);
		flush = __flush_bios(pe);
		up_write(&s->lock);

		error_snapshot_bios(pe);
		goto out;
	}

	e = alloc_exception();
	if (!e) {
		down_write(&s->lock);
		__invalidate_snapshot(s, pe, -ENOMEM);
		flush = __flush_bios(pe);
		up_write(&s->lock);

		error_snapshot_bios(pe);
		goto out;
	}
	*e = pe->e;

	/*
	 * Add a proper exception, and remove the
	 * in-flight exception from the list.
	 */
	down_write(&s->lock);
	if (!s->valid) {
		flush = __flush_bios(pe);
		up_write(&s->lock);

		free_exception(e);

		error_snapshot_bios(pe);
		goto out;
	}

	insert_exception(&s->complete, e);
	remove_exception(&pe->e);
	flush = __flush_bios(pe);

	up_write(&s->lock);

	/* Submit any pending write bios */
	flush_bios(bio_list_get(&pe->snapshot_bios));

 out:
	primary_pe = pe->primary_pe;

	/*
	 * Free the pe if it's not linked to an origin write or if
	 * it's not itself a primary pe.
	 */
	if (!primary_pe || primary_pe != pe)
		free_pending_exception(pe);

	/*
	 * Free the primary pe if nothing references it.
	 */
	if (primary_pe && !atomic_read(&primary_pe->sibling_count))
		free_pending_exception(primary_pe);

	if (flush)
		flush_bios(flush);
}

static void commit_callback(void *context, int success)
{
	struct pending_exception *pe = (struct pending_exception *) context;
	pending_complete(pe, success);
}

/*
 * Called when the copy I/O has finished.  kcopyd actually runs
 * this code so don't block.
 */
static void copy_callback(int read_err, unsigned int write_err, void *context)
{
	struct pending_exception *pe = (struct pending_exception *) context;
	struct dm_snapshot *s = pe->snap;

	if (read_err || write_err)
		pending_complete(pe, 0);

	else
		/* Update the metadata if we are persistent */
		s->store.commit_exception(&s->store, &pe->e, commit_callback,
					  pe);
}

/*
 * Dispatches the copy operation to kcopyd.
 */
static void start_copy(struct pending_exception *pe)
{
	struct dm_snapshot *s = pe->snap;
	struct io_region src, dest;
	struct block_device *bdev = s->origin->bdev;
	sector_t dev_size;

	dev_size = get_dev_size(bdev);

	src.bdev = bdev;
	src.sector = chunk_to_sector(s, pe->e.old_chunk);
	src.count = min(s->chunk_size, dev_size - src.sector);

	dest.bdev = s->cow->bdev;
	dest.sector = chunk_to_sector(s, pe->e.new_chunk);
	dest.count = src.count;

	/* Hand over to kcopyd */
	kcopyd_copy(s->kcopyd_client,
		    &src, 1, &dest, 0, copy_callback, pe);
}

/*
 * Looks to see if this snapshot already has a pending exception
 * for this chunk, otherwise it allocates a new one and inserts
 * it into the pending table.
 *
 * NOTE: a write lock must be held on snap->lock before calling
 * this.
 */
static struct pending_exception *
__find_pending_exception(struct dm_snapshot *s, struct bio *bio)
{
	struct exception *e;
	struct pending_exception *pe;
	chunk_t chunk = sector_to_chunk(s, bio->bi_sector);

	/*
	 * Is there a pending exception for this already ?
	 */
	e = lookup_exception(&s->pending, chunk);
	if (e) {
		/* cast the exception to a pending exception */
		pe = container_of(e, struct pending_exception, e);
		goto out;
	}

	/*
	 * Create a new pending exception, we don't want
	 * to hold the lock while we do this.
	 */
	up_write(&s->lock);
	pe = alloc_pending_exception();
	down_write(&s->lock);

	if (!s->valid) {
		free_pending_exception(pe);
		return NULL;
	}

	e = lookup_exception(&s->pending, chunk);
	if (e) {
		free_pending_exception(pe);
		pe = container_of(e, struct pending_exception, e);
		goto out;
	}

	pe->e.old_chunk = chunk;
	bio_list_init(&pe->origin_bios);
	bio_list_init(&pe->snapshot_bios);
	pe->primary_pe = NULL;
	atomic_set(&pe->sibling_count, 1);
	pe->snap = s;
	pe->started = 0;

	if (s->store.prepare_exception(&s->store, &pe->e)) {
		free_pending_exception(pe);
		return NULL;
	}

	insert_exception(&s->pending, &pe->e);

 out:
	return pe;
}

static inline void remap_exception(struct dm_snapshot *s, struct exception *e,
				   struct bio *bio)
{
	bio->bi_bdev = s->cow->bdev;
	bio->bi_sector = chunk_to_sector(s, e->new_chunk) +
		(bio->bi_sector & s->chunk_mask);
}

static int snapshot_map(struct dm_target *ti, struct bio *bio,
			union map_info *map_context)
{
	struct exception *e;
	struct dm_snapshot *s = (struct dm_snapshot *) ti->private;
	int copy_needed = 0;
	int r = 1;
	chunk_t chunk;
	struct pending_exception *pe = NULL;

	chunk = sector_to_chunk(s, bio->bi_sector);

	/* Full snapshots are not usable */
	/* To get here the table must be live so s->active is always set. */
	if (!s->valid)
		return -EIO;

	if (unlikely(bio_barrier(bio)))
		return -EOPNOTSUPP;

	/*
	 * Write to snapshot - higher level takes care of RW/RO
	 * flags so we should only get this if we are
	 * writeable.
	 */
	if (bio_rw(bio) == WRITE) {

		/* FIXME: should only take write lock if we need
		 * to copy an exception */
		down_write(&s->lock);

		if (!s->valid) {
			r = -EIO;
			goto out_unlock;
		}

		/* If the block is already remapped - use that, else remap it */
		e = lookup_exception(&s->complete, chunk);
		if (e) {
			remap_exception(s, e, bio);
			goto out_unlock;
		}

		pe = __find_pending_exception(s, bio);
		if (!pe) {
			__invalidate_snapshot(s, pe, -ENOMEM);
			r = -EIO;
			goto out_unlock;
		}

		remap_exception(s, &pe->e, bio);
		bio_list_add(&pe->snapshot_bios, bio);

		if (!pe->started) {
			/* this is protected by snap->lock */
			pe->started = 1;
			copy_needed = 1;
		}

		r = 0;

 out_unlock:
		up_write(&s->lock);

		if (copy_needed)
			start_copy(pe);
	} else {
		/*
		 * FIXME: this read path scares me because we
		 * always use the origin when we have a pending
		 * exception.  However I can't think of a
		 * situation where this is wrong - ejt.
		 */

		/* Do reads */
		down_read(&s->lock);

		if (!s->valid) {
			up_read(&s->lock);
			return -EIO;
		}

		/* See if it it has been remapped */
		e = lookup_exception(&s->complete, chunk);
		if (e)
			remap_exception(s, e, bio);
		else
			bio->bi_bdev = s->origin->bdev;

		up_read(&s->lock);
	}

	return r;
}

static void snapshot_resume(struct dm_target *ti)
{
	struct dm_snapshot *s = (struct dm_snapshot *) ti->private;

	down_write(&s->lock);
	s->active = 1;
	up_write(&s->lock);
}

static int snapshot_status(struct dm_target *ti, status_type_t type,
			   char *result, unsigned int maxlen)
{
	struct dm_snapshot *snap = (struct dm_snapshot *) ti->private;

	switch (type) {
	case STATUSTYPE_INFO:
		if (!snap->valid)
			snprintf(result, maxlen, "Invalid");
		else {
			if (snap->store.fraction_full) {
				sector_t numerator, denominator;
				snap->store.fraction_full(&snap->store,
							  &numerator,
							  &denominator);
				snprintf(result, maxlen, "%llu/%llu",
					(unsigned long long)numerator,
					(unsigned long long)denominator);
			}
			else
				snprintf(result, maxlen, "Unknown");
		}
		break;

	case STATUSTYPE_TABLE:
		/*
		 * kdevname returns a static pointer so we need
		 * to make private copies if the output is to
		 * make sense.
		 */
		snprintf(result, maxlen, "%s %s %c %llu",
			 snap->origin->name, snap->cow->name,
			 snap->type,
			 (unsigned long long)snap->chunk_size);
		break;
	}

	return 0;
}

/*-----------------------------------------------------------------
 * Origin methods
 *---------------------------------------------------------------*/
static int __origin_write(struct list_head *snapshots, struct bio *bio)
{
	int r = 1, first = 0;
	struct dm_snapshot *snap;
	struct exception *e;
	struct pending_exception *pe, *next_pe, *primary_pe = NULL;
	chunk_t chunk;
	LIST_HEAD(pe_queue);

	/* Do all the snapshots on this origin */
	list_for_each_entry (snap, snapshots, list) {

		down_write(&snap->lock);

		/* Only deal with valid and active snapshots */
		if (!snap->valid || !snap->active)
			goto next_snapshot;

		/* Nothing to do if writing beyond end of snapshot */
		if (bio->bi_sector >= dm_table_get_size(snap->table))
			goto next_snapshot;

		/*
		 * Remember, different snapshots can have
		 * different chunk sizes.
		 */
		chunk = sector_to_chunk(snap, bio->bi_sector);

		/*
		 * Check exception table to see if block
		 * is already remapped in this snapshot
		 * and trigger an exception if not.
		 *
		 * sibling_count is initialised to 1 so pending_complete()
		 * won't destroy the primary_pe while we're inside this loop.
		 */
		e = lookup_exception(&snap->complete, chunk);
		if (e)
			goto next_snapshot;

		pe = __find_pending_exception(snap, bio);
		if (!pe) {
			__invalidate_snapshot(snap, pe, -ENOMEM);
			goto next_snapshot;
		}

		if (!primary_pe) {
			/*
			 * Either every pe here has same
			 * primary_pe or none has one yet.
			 */
			if (pe->primary_pe)
				primary_pe = pe->primary_pe;
			else {
				primary_pe = pe;
				first = 1;
			}

			bio_list_add(&primary_pe->origin_bios, bio);

			r = 0;
		}

		if (!pe->primary_pe) {
			atomic_inc(&primary_pe->sibling_count);
			pe->primary_pe = primary_pe;
		}

		if (!pe->started) {
			pe->started = 1;
			list_add_tail(&pe->list, &pe_queue);
		}

 next_snapshot:
		up_write(&snap->lock);
	}

	if (!primary_pe)
		goto out;

	/*
	 * If this is the first time we're processing this chunk and
	 * sibling_count is now 1 it means all the pending exceptions
	 * got completed while we were in the loop above, so it falls to
	 * us here to remove the primary_pe and submit any origin_bios.
	 */

	if (first && atomic_dec_and_test(&primary_pe->sibling_count)) {
		flush_bios(bio_list_get(&primary_pe->origin_bios));
		free_pending_exception(primary_pe);
		/* If we got here, pe_queue is necessarily empty. */
		goto out;
	}

	/*
	 * Now that we have a complete pe list we can start the copying.
	 */
	list_for_each_entry_safe(pe, next_pe, &pe_queue, list)
		start_copy(pe);

 out:
	return r;
}

/*
 * Called on a write from the origin driver.
 */
static int do_origin(struct dm_dev *origin, struct bio *bio)
{
	struct origin *o;
	int r = 1;

	down_read(&_origins_lock);
	o = __lookup_origin(origin->bdev);
	if (o)
		r = __origin_write(&o->snapshots, bio);
	up_read(&_origins_lock);

	return r;
}

/*
 * Origin: maps a linear range of a device, with hooks for snapshotting.
 */

/*
 * Construct an origin mapping: <dev_path>
 * The context for an origin is merely a 'struct dm_dev *'
 * pointing to the real device.
 */
static int origin_ctr(struct dm_target *ti, unsigned int argc, char **argv)
{
	int r;
	struct dm_dev *dev;

	if (argc != 1) {
		ti->error = "origin: incorrect number of arguments";
		return -EINVAL;
	}

	r = dm_get_device(ti, argv[0], 0, ti->len,
			  dm_table_get_mode(ti->table), &dev);
	if (r) {
		ti->error = "Cannot get target device";
		return r;
	}

	ti->private = dev;
	return 0;
}

static void origin_dtr(struct dm_target *ti)
{
	struct dm_dev *dev = (struct dm_dev *) ti->private;
	dm_put_device(ti, dev);
}

static int origin_map(struct dm_target *ti, struct bio *bio,
		      union map_info *map_context)
{
	struct dm_dev *dev = (struct dm_dev *) ti->private;
	bio->bi_bdev = dev->bdev;

	if (unlikely(bio_barrier(bio)))
		return -EOPNOTSUPP;

	/* Only tell snapshots if this is a write */
	return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : 1;
}

#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))

/*
 * Set the target "split_io" field to the minimum of all the snapshots'
 * chunk sizes.
 */
static void origin_resume(struct dm_target *ti)
{
	struct dm_dev *dev = (struct dm_dev *) ti->private;
	struct dm_snapshot *snap;
	struct origin *o;
	chunk_t chunk_size = 0;

	down_read(&_origins_lock);
	o = __lookup_origin(dev->bdev);
	if (o)
		list_for_each_entry (snap, &o->snapshots, list)
			chunk_size = min_not_zero(chunk_size, snap->chunk_size);
	up_read(&_origins_lock);

	ti->split_io = chunk_size;
}

static int origin_status(struct dm_target *ti, status_type_t type, char *result,
			 unsigned int maxlen)
{
	struct dm_dev *dev = (struct dm_dev *) ti->private;

	switch (type) {
	case STATUSTYPE_INFO:
		result[0] = '\0';
		break;

	case STATUSTYPE_TABLE:
		snprintf(result, maxlen, "%s", dev->name);
		break;
	}

	return 0;
}

static struct target_type origin_target = {
	.name    = "snapshot-origin",
	.version = {1, 5, 0},
	.module  = THIS_MODULE,
	.ctr     = origin_ctr,
	.dtr     = origin_dtr,
	.map     = origin_map,
	.resume  = origin_resume,
	.status  = origin_status,
};

static struct target_type snapshot_target = {
	.name    = "snapshot",
	.version = {1, 5, 0},
	.module  = THIS_MODULE,
	.ctr     = snapshot_ctr,
	.dtr     = snapshot_dtr,
	.map     = snapshot_map,
	.resume  = snapshot_resume,
	.status  = snapshot_status,
};

static int __init dm_snapshot_init(void)
{
	int r;

	r = dm_register_target(&snapshot_target);
	if (r) {
		DMERR("snapshot target register failed %d", r);
		return r;
	}

	r = dm_register_target(&origin_target);
	if (r < 0) {
		DMERR("Origin target register failed %d", r);
		goto bad1;
	}

	r = init_origin_hash();
	if (r) {
		DMERR("init_origin_hash failed.");
		goto bad2;
	}

	exception_cache = kmem_cache_create("dm-snapshot-ex",
					    sizeof(struct exception),
					    __alignof__(struct exception),
					    0, NULL, NULL);
	if (!exception_cache) {
		DMERR("Couldn't create exception cache.");
		r = -ENOMEM;
		goto bad3;
	}

	pending_cache =
	    kmem_cache_create("dm-snapshot-in",
			      sizeof(struct pending_exception),
			      __alignof__(struct pending_exception),
			      0, NULL, NULL);
	if (!pending_cache) {
		DMERR("Couldn't create pending cache.");
		r = -ENOMEM;
		goto bad4;
	}

	pending_pool = mempool_create_slab_pool(128, pending_cache);
	if (!pending_pool) {
		DMERR("Couldn't create pending pool.");
		r = -ENOMEM;
		goto bad5;
	}

	return 0;

      bad5:
	kmem_cache_destroy(pending_cache);
      bad4:
	kmem_cache_destroy(exception_cache);
      bad3:
	exit_origin_hash();
      bad2:
	dm_unregister_target(&origin_target);
      bad1:
	dm_unregister_target(&snapshot_target);
	return r;
}

static void __exit dm_snapshot_exit(void)
{
	int r;

	r = dm_unregister_target(&snapshot_target);
	if (r)
		DMERR("snapshot unregister failed %d", r);

	r = dm_unregister_target(&origin_target);
	if (r)
		DMERR("origin unregister failed %d", r);

	exit_origin_hash();
	mempool_destroy(pending_pool);
	kmem_cache_destroy(pending_cache);
	kmem_cache_destroy(exception_cache);
}

/* Module hooks */
module_init(dm_snapshot_init);
module_exit(dm_snapshot_exit);

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