summaryrefslogtreecommitdiffstats
path: root/drivers/md/dm-snap-persistent.c
blob: f14e5df2787467d9fa2c1ec827507f50023d1d7e (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
// SPDX-License-Identifier: GPL-2.0-only
/*
 * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
 * Copyright (C) 2006-2008 Red Hat GmbH
 *
 * This file is released under the GPL.
 */

#include "dm-exception-store.h"

#include <linux/ctype.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/vmalloc.h>
#include <linux/export.h>
#include <linux/slab.h>
#include <linux/dm-io.h>
#include <linux/dm-bufio.h>

#define DM_MSG_PREFIX "persistent snapshot"
#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32U	/* 16KB */

#define DM_PREFETCH_CHUNKS		12

/*
 *---------------------------------------------------------------
 * Persistent snapshots, by persistent we mean that the snapshot
 * will survive a reboot.
 *---------------------------------------------------------------
 */

/*
 * We need to store a record of which parts of the origin have
 * been copied to the snapshot device.  The snapshot code
 * requires that we copy exception chunks to chunk aligned areas
 * of the COW store.  It makes sense therefore, to store the
 * metadata in chunk size blocks.
 *
 * There is no backward or forward compatibility implemented,
 * snapshots with different disk versions than the kernel will
 * not be usable.  It is expected that "lvcreate" will blank out
 * the start of a fresh COW device before calling the snapshot
 * constructor.
 *
 * The first chunk of the COW device just contains the header.
 * After this there is a chunk filled with exception metadata,
 * followed by as many exception chunks as can fit in the
 * metadata areas.
 *
 * All on disk structures are in little-endian format.  The end
 * of the exceptions info is indicated by an exception with a
 * new_chunk of 0, which is invalid since it would point to the
 * header chunk.
 */

/*
 * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
 */
#define SNAP_MAGIC 0x70416e53

/*
 * The on-disk version of the metadata.
 */
#define SNAPSHOT_DISK_VERSION 1

#define NUM_SNAPSHOT_HDR_CHUNKS 1

struct disk_header {
	__le32 magic;

	/*
	 * Is this snapshot valid.  There is no way of recovering
	 * an invalid snapshot.
	 */
	__le32 valid;

	/*
	 * Simple, incrementing version. no backward
	 * compatibility.
	 */
	__le32 version;

	/* In sectors */
	__le32 chunk_size;
} __packed;

struct disk_exception {
	__le64 old_chunk;
	__le64 new_chunk;
} __packed;

struct core_exception {
	uint64_t old_chunk;
	uint64_t new_chunk;
};

struct commit_callback {
	void (*callback)(void *ref, int success);
	void *context;
};

/*
 * The top level structure for a persistent exception store.
 */
struct pstore {
	struct dm_exception_store *store;
	int version;
	int valid;
	uint32_t exceptions_per_area;

	/*
	 * Now that we have an asynchronous kcopyd there is no
	 * need for large chunk sizes, so it wont hurt to have a
	 * whole chunks worth of metadata in memory at once.
	 */
	void *area;

	/*
	 * An area of zeros used to clear the next area.
	 */
	void *zero_area;

	/*
	 * An area used for header. The header can be written
	 * concurrently with metadata (when invalidating the snapshot),
	 * so it needs a separate buffer.
	 */
	void *header_area;

	/*
	 * Used to keep track of which metadata area the data in
	 * 'chunk' refers to.
	 */
	chunk_t current_area;

	/*
	 * The next free chunk for an exception.
	 *
	 * When creating exceptions, all the chunks here and above are
	 * free.  It holds the next chunk to be allocated.  On rare
	 * occasions (e.g. after a system crash) holes can be left in
	 * the exception store because chunks can be committed out of
	 * order.
	 *
	 * When merging exceptions, it does not necessarily mean all the
	 * chunks here and above are free.  It holds the value it would
	 * have held if all chunks had been committed in order of
	 * allocation.  Consequently the value may occasionally be
	 * slightly too low, but since it's only used for 'status' and
	 * it can never reach its minimum value too early this doesn't
	 * matter.
	 */

	chunk_t next_free;

	/*
	 * The index of next free exception in the current
	 * metadata area.
	 */
	uint32_t current_committed;

	atomic_t pending_count;
	uint32_t callback_count;
	struct commit_callback *callbacks;
	struct dm_io_client *io_client;

	struct workqueue_struct *metadata_wq;
};

static int alloc_area(struct pstore *ps)
{
	int r = -ENOMEM;
	size_t len;

	len = ps->store->chunk_size << SECTOR_SHIFT;

	/*
	 * Allocate the chunk_size block of memory that will hold
	 * a single metadata area.
	 */
	ps->area = vmalloc(len);
	if (!ps->area)
		goto err_area;

	ps->zero_area = vzalloc(len);
	if (!ps->zero_area)
		goto err_zero_area;

	ps->header_area = vmalloc(len);
	if (!ps->header_area)
		goto err_header_area;

	return 0;

err_header_area:
	vfree(ps->zero_area);

err_zero_area:
	vfree(ps->area);

err_area:
	return r;
}

static void free_area(struct pstore *ps)
{
	vfree(ps->area);
	ps->area = NULL;
	vfree(ps->zero_area);
	ps->zero_area = NULL;
	vfree(ps->header_area);
	ps->header_area = NULL;
}

struct mdata_req {
	struct dm_io_region *where;
	struct dm_io_request *io_req;
	struct work_struct work;
	int result;
};

static void do_metadata(struct work_struct *work)
{
	struct mdata_req *req = container_of(work, struct mdata_req, work);

	req->result = dm_io(req->io_req, 1, req->where, NULL);
}

/*
 * Read or write a chunk aligned and sized block of data from a device.
 */
static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, blk_opf_t opf,
		    int metadata)
{
	struct dm_io_region where = {
		.bdev = dm_snap_cow(ps->store->snap)->bdev,
		.sector = ps->store->chunk_size * chunk,
		.count = ps->store->chunk_size,
	};
	struct dm_io_request io_req = {
		.bi_opf = opf,
		.mem.type = DM_IO_VMA,
		.mem.ptr.vma = area,
		.client = ps->io_client,
		.notify.fn = NULL,
	};
	struct mdata_req req;

	if (!metadata)
		return dm_io(&io_req, 1, &where, NULL);

	req.where = &where;
	req.io_req = &io_req;

	/*
	 * Issue the synchronous I/O from a different thread
	 * to avoid submit_bio_noacct recursion.
	 */
	INIT_WORK_ONSTACK(&req.work, do_metadata);
	queue_work(ps->metadata_wq, &req.work);
	flush_workqueue(ps->metadata_wq);
	destroy_work_on_stack(&req.work);

	return req.result;
}

/*
 * Convert a metadata area index to a chunk index.
 */
static chunk_t area_location(struct pstore *ps, chunk_t area)
{
	return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
}

static void skip_metadata(struct pstore *ps)
{
	uint32_t stride = ps->exceptions_per_area + 1;
	chunk_t next_free = ps->next_free;

	if (sector_div(next_free, stride) == NUM_SNAPSHOT_HDR_CHUNKS)
		ps->next_free++;
}

/*
 * Read or write a metadata area.  Remembering to skip the first
 * chunk which holds the header.
 */
static int area_io(struct pstore *ps, blk_opf_t opf)
{
	chunk_t chunk = area_location(ps, ps->current_area);

	return chunk_io(ps, ps->area, chunk, opf, 0);
}

static void zero_memory_area(struct pstore *ps)
{
	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
}

static int zero_disk_area(struct pstore *ps, chunk_t area)
{
	return chunk_io(ps, ps->zero_area, area_location(ps, area),
			REQ_OP_WRITE, 0);
}

static int read_header(struct pstore *ps, int *new_snapshot)
{
	int r;
	struct disk_header *dh;
	unsigned int chunk_size;
	int chunk_size_supplied = 1;
	char *chunk_err;

	/*
	 * Use default chunk size (or logical_block_size, if larger)
	 * if none supplied
	 */
	if (!ps->store->chunk_size) {
		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
					    bdev) >> 9);
		ps->store->chunk_mask = ps->store->chunk_size - 1;
		ps->store->chunk_shift = __ffs(ps->store->chunk_size);
		chunk_size_supplied = 0;
	}

	ps->io_client = dm_io_client_create();
	if (IS_ERR(ps->io_client))
		return PTR_ERR(ps->io_client);

	r = alloc_area(ps);
	if (r)
		return r;

	r = chunk_io(ps, ps->header_area, 0, REQ_OP_READ, 1);
	if (r)
		goto bad;

	dh = ps->header_area;

	if (le32_to_cpu(dh->magic) == 0) {
		*new_snapshot = 1;
		return 0;
	}

	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
		DMWARN("Invalid or corrupt snapshot");
		r = -ENXIO;
		goto bad;
	}

	*new_snapshot = 0;
	ps->valid = le32_to_cpu(dh->valid);
	ps->version = le32_to_cpu(dh->version);
	chunk_size = le32_to_cpu(dh->chunk_size);

	if (ps->store->chunk_size == chunk_size)
		return 0;

	if (chunk_size_supplied)
		DMWARN("chunk size %u in device metadata overrides table chunk size of %u.",
		       chunk_size, ps->store->chunk_size);

	/* We had a bogus chunk_size. Fix stuff up. */
	free_area(ps);

	r = dm_exception_store_set_chunk_size(ps->store, chunk_size,
					      &chunk_err);
	if (r) {
		DMERR("invalid on-disk chunk size %u: %s.",
		      chunk_size, chunk_err);
		return r;
	}

	r = alloc_area(ps);
	return r;

bad:
	free_area(ps);
	return r;
}

static int write_header(struct pstore *ps)
{
	struct disk_header *dh;

	memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT);

	dh = ps->header_area;
	dh->magic = cpu_to_le32(SNAP_MAGIC);
	dh->valid = cpu_to_le32(ps->valid);
	dh->version = cpu_to_le32(ps->version);
	dh->chunk_size = cpu_to_le32(ps->store->chunk_size);

	return chunk_io(ps, ps->header_area, 0, REQ_OP_WRITE, 1);
}

/*
 * Access functions for the disk exceptions, these do the endian conversions.
 */
static struct disk_exception *get_exception(struct pstore *ps, void *ps_area,
					    uint32_t index)
{
	BUG_ON(index >= ps->exceptions_per_area);

	return ((struct disk_exception *) ps_area) + index;
}

static void read_exception(struct pstore *ps, void *ps_area,
			   uint32_t index, struct core_exception *result)
{
	struct disk_exception *de = get_exception(ps, ps_area, index);

	/* copy it */
	result->old_chunk = le64_to_cpu(de->old_chunk);
	result->new_chunk = le64_to_cpu(de->new_chunk);
}

static void write_exception(struct pstore *ps,
			    uint32_t index, struct core_exception *e)
{
	struct disk_exception *de = get_exception(ps, ps->area, index);

	/* copy it */
	de->old_chunk = cpu_to_le64(e->old_chunk);
	de->new_chunk = cpu_to_le64(e->new_chunk);
}

static void clear_exception(struct pstore *ps, uint32_t index)
{
	struct disk_exception *de = get_exception(ps, ps->area, index);

	/* clear it */
	de->old_chunk = 0;
	de->new_chunk = 0;
}

/*
 * Registers the exceptions that are present in the current area.
 * 'full' is filled in to indicate if the area has been
 * filled.
 */
static int insert_exceptions(struct pstore *ps, void *ps_area,
			     int (*callback)(void *callback_context,
					     chunk_t old, chunk_t new),
			     void *callback_context,
			     int *full)
{
	int r;
	unsigned int i;
	struct core_exception e;

	/* presume the area is full */
	*full = 1;

	for (i = 0; i < ps->exceptions_per_area; i++) {
		read_exception(ps, ps_area, i, &e);

		/*
		 * If the new_chunk is pointing at the start of
		 * the COW device, where the first metadata area
		 * is we know that we've hit the end of the
		 * exceptions.  Therefore the area is not full.
		 */
		if (e.new_chunk == 0LL) {
			ps->current_committed = i;
			*full = 0;
			break;
		}

		/*
		 * Keep track of the start of the free chunks.
		 */
		if (ps->next_free <= e.new_chunk)
			ps->next_free = e.new_chunk + 1;

		/*
		 * Otherwise we add the exception to the snapshot.
		 */
		r = callback(callback_context, e.old_chunk, e.new_chunk);
		if (r)
			return r;
	}

	return 0;
}

static int read_exceptions(struct pstore *ps,
			   int (*callback)(void *callback_context, chunk_t old,
					   chunk_t new),
			   void *callback_context)
{
	int r, full = 1;
	struct dm_bufio_client *client;
	chunk_t prefetch_area = 0;

	client = dm_bufio_client_create(dm_snap_cow(ps->store->snap)->bdev,
					ps->store->chunk_size << SECTOR_SHIFT,
					1, 0, NULL, NULL, 0);

	if (IS_ERR(client))
		return PTR_ERR(client);

	/*
	 * Setup for one current buffer + desired readahead buffers.
	 */
	dm_bufio_set_minimum_buffers(client, 1 + DM_PREFETCH_CHUNKS);

	/*
	 * Keeping reading chunks and inserting exceptions until
	 * we find a partially full area.
	 */
	for (ps->current_area = 0; full; ps->current_area++) {
		struct dm_buffer *bp;
		void *area;
		chunk_t chunk;

		if (unlikely(prefetch_area < ps->current_area))
			prefetch_area = ps->current_area;

		if (DM_PREFETCH_CHUNKS) {
			do {
				chunk_t pf_chunk = area_location(ps, prefetch_area);

				if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
					break;
				dm_bufio_prefetch(client, pf_chunk, 1);
				prefetch_area++;
				if (unlikely(!prefetch_area))
					break;
			} while (prefetch_area <= ps->current_area + DM_PREFETCH_CHUNKS);
		}

		chunk = area_location(ps, ps->current_area);

		area = dm_bufio_read(client, chunk, &bp);
		if (IS_ERR(area)) {
			r = PTR_ERR(area);
			goto ret_destroy_bufio;
		}

		r = insert_exceptions(ps, area, callback, callback_context,
				      &full);

		if (!full)
			memcpy(ps->area, area, ps->store->chunk_size << SECTOR_SHIFT);

		dm_bufio_release(bp);

		dm_bufio_forget(client, chunk);

		if (unlikely(r))
			goto ret_destroy_bufio;
	}

	ps->current_area--;

	skip_metadata(ps);

	r = 0;

ret_destroy_bufio:
	dm_bufio_client_destroy(client);

	return r;
}

static struct pstore *get_info(struct dm_exception_store *store)
{
	return (struct pstore *) store->context;
}

static void persistent_usage(struct dm_exception_store *store,
			     sector_t *total_sectors,
			     sector_t *sectors_allocated,
			     sector_t *metadata_sectors)
{
	struct pstore *ps = get_info(store);

	*sectors_allocated = ps->next_free * store->chunk_size;
	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);

	/*
	 * First chunk is the fixed header.
	 * Then there are (ps->current_area + 1) metadata chunks, each one
	 * separated from the next by ps->exceptions_per_area data chunks.
	 */
	*metadata_sectors = (ps->current_area + 1 + NUM_SNAPSHOT_HDR_CHUNKS) *
			    store->chunk_size;
}

static void persistent_dtr(struct dm_exception_store *store)
{
	struct pstore *ps = get_info(store);

	destroy_workqueue(ps->metadata_wq);

	/* Created in read_header */
	if (ps->io_client)
		dm_io_client_destroy(ps->io_client);
	free_area(ps);

	/* Allocated in persistent_read_metadata */
	kvfree(ps->callbacks);

	kfree(ps);
}

static int persistent_read_metadata(struct dm_exception_store *store,
				    int (*callback)(void *callback_context,
						    chunk_t old, chunk_t new),
				    void *callback_context)
{
	int r, new_snapshot;
	struct pstore *ps = get_info(store);

	/*
	 * Read the snapshot header.
	 */
	r = read_header(ps, &new_snapshot);
	if (r)
		return r;

	/*
	 * Now we know correct chunk_size, complete the initialisation.
	 */
	ps->exceptions_per_area = (ps->store->chunk_size << SECTOR_SHIFT) /
				  sizeof(struct disk_exception);
	ps->callbacks = kvcalloc(ps->exceptions_per_area,
				 sizeof(*ps->callbacks), GFP_KERNEL);
	if (!ps->callbacks)
		return -ENOMEM;

	/*
	 * Do we need to setup a new snapshot ?
	 */
	if (new_snapshot) {
		r = write_header(ps);
		if (r) {
			DMWARN("write_header failed");
			return r;
		}

		ps->current_area = 0;
		zero_memory_area(ps);
		r = zero_disk_area(ps, 0);
		if (r)
			DMWARN("zero_disk_area(0) failed");
		return r;
	}
	/*
	 * Sanity checks.
	 */
	if (ps->version != SNAPSHOT_DISK_VERSION) {
		DMWARN("unable to handle snapshot disk version %d",
		       ps->version);
		return -EINVAL;
	}

	/*
	 * Metadata are valid, but snapshot is invalidated
	 */
	if (!ps->valid)
		return 1;

	/*
	 * Read the metadata.
	 */
	r = read_exceptions(ps, callback, callback_context);

	return r;
}

static int persistent_prepare_exception(struct dm_exception_store *store,
					struct dm_exception *e)
{
	struct pstore *ps = get_info(store);
	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);

	/* Is there enough room ? */
	if (size < ((ps->next_free + 1) * store->chunk_size))
		return -ENOSPC;

	e->new_chunk = ps->next_free;

	/*
	 * Move onto the next free pending, making sure to take
	 * into account the location of the metadata chunks.
	 */
	ps->next_free++;
	skip_metadata(ps);

	atomic_inc(&ps->pending_count);
	return 0;
}

static void persistent_commit_exception(struct dm_exception_store *store,
					struct dm_exception *e, int valid,
					void (*callback)(void *, int success),
					void *callback_context)
{
	unsigned int i;
	struct pstore *ps = get_info(store);
	struct core_exception ce;
	struct commit_callback *cb;

	if (!valid)
		ps->valid = 0;

	ce.old_chunk = e->old_chunk;
	ce.new_chunk = e->new_chunk;
	write_exception(ps, ps->current_committed++, &ce);

	/*
	 * Add the callback to the back of the array.  This code
	 * is the only place where the callback array is
	 * manipulated, and we know that it will never be called
	 * multiple times concurrently.
	 */
	cb = ps->callbacks + ps->callback_count++;
	cb->callback = callback;
	cb->context = callback_context;

	/*
	 * If there are exceptions in flight and we have not yet
	 * filled this metadata area there's nothing more to do.
	 */
	if (!atomic_dec_and_test(&ps->pending_count) &&
	    (ps->current_committed != ps->exceptions_per_area))
		return;

	/*
	 * If we completely filled the current area, then wipe the next one.
	 */
	if ((ps->current_committed == ps->exceptions_per_area) &&
	    zero_disk_area(ps, ps->current_area + 1))
		ps->valid = 0;

	/*
	 * Commit exceptions to disk.
	 */
	if (ps->valid && area_io(ps, REQ_OP_WRITE | REQ_PREFLUSH | REQ_FUA |
				 REQ_SYNC))
		ps->valid = 0;

	/*
	 * Advance to the next area if this one is full.
	 */
	if (ps->current_committed == ps->exceptions_per_area) {
		ps->current_committed = 0;
		ps->current_area++;
		zero_memory_area(ps);
	}

	for (i = 0; i < ps->callback_count; i++) {
		cb = ps->callbacks + i;
		cb->callback(cb->context, ps->valid);
	}

	ps->callback_count = 0;
}

static int persistent_prepare_merge(struct dm_exception_store *store,
				    chunk_t *last_old_chunk,
				    chunk_t *last_new_chunk)
{
	struct pstore *ps = get_info(store);
	struct core_exception ce;
	int nr_consecutive;
	int r;

	/*
	 * When current area is empty, move back to preceding area.
	 */
	if (!ps->current_committed) {
		/*
		 * Have we finished?
		 */
		if (!ps->current_area)
			return 0;

		ps->current_area--;
		r = area_io(ps, REQ_OP_READ);
		if (r < 0)
			return r;
		ps->current_committed = ps->exceptions_per_area;
	}

	read_exception(ps, ps->area, ps->current_committed - 1, &ce);
	*last_old_chunk = ce.old_chunk;
	*last_new_chunk = ce.new_chunk;

	/*
	 * Find number of consecutive chunks within the current area,
	 * working backwards.
	 */
	for (nr_consecutive = 1; nr_consecutive < ps->current_committed;
	     nr_consecutive++) {
		read_exception(ps, ps->area,
			       ps->current_committed - 1 - nr_consecutive, &ce);
		if (ce.old_chunk != *last_old_chunk - nr_consecutive ||
		    ce.new_chunk != *last_new_chunk - nr_consecutive)
			break;
	}

	return nr_consecutive;
}

static int persistent_commit_merge(struct dm_exception_store *store,
				   int nr_merged)
{
	int r, i;
	struct pstore *ps = get_info(store);

	BUG_ON(nr_merged > ps->current_committed);

	for (i = 0; i < nr_merged; i++)
		clear_exception(ps, ps->current_committed - 1 - i);

	r = area_io(ps, REQ_OP_WRITE | REQ_PREFLUSH | REQ_FUA);
	if (r < 0)
		return r;

	ps->current_committed -= nr_merged;

	/*
	 * At this stage, only persistent_usage() uses ps->next_free, so
	 * we make no attempt to keep ps->next_free strictly accurate
	 * as exceptions may have been committed out-of-order originally.
	 * Once a snapshot has become merging, we set it to the value it
	 * would have held had all the exceptions been committed in order.
	 *
	 * ps->current_area does not get reduced by prepare_merge() until
	 * after commit_merge() has removed the nr_merged previous exceptions.
	 */
	ps->next_free = area_location(ps, ps->current_area) +
			ps->current_committed + 1;

	return 0;
}

static void persistent_drop_snapshot(struct dm_exception_store *store)
{
	struct pstore *ps = get_info(store);

	ps->valid = 0;
	if (write_header(ps))
		DMWARN("write header failed");
}

static int persistent_ctr(struct dm_exception_store *store, char *options)
{
	struct pstore *ps;
	int r;

	/* allocate the pstore */
	ps = kzalloc(sizeof(*ps), GFP_KERNEL);
	if (!ps)
		return -ENOMEM;

	ps->store = store;
	ps->valid = 1;
	ps->version = SNAPSHOT_DISK_VERSION;
	ps->area = NULL;
	ps->zero_area = NULL;
	ps->header_area = NULL;
	ps->next_free = NUM_SNAPSHOT_HDR_CHUNKS + 1; /* header and 1st area */
	ps->current_committed = 0;

	ps->callback_count = 0;
	atomic_set(&ps->pending_count, 0);
	ps->callbacks = NULL;

	ps->metadata_wq = alloc_workqueue("ksnaphd", WQ_MEM_RECLAIM, 0);
	if (!ps->metadata_wq) {
		DMERR("couldn't start header metadata update thread");
		r = -ENOMEM;
		goto err_workqueue;
	}

	if (options) {
		char overflow = toupper(options[0]);

		if (overflow == 'O')
			store->userspace_supports_overflow = true;
		else {
			DMERR("Unsupported persistent store option: %s", options);
			r = -EINVAL;
			goto err_options;
		}
	}

	store->context = ps;

	return 0;

err_options:
	destroy_workqueue(ps->metadata_wq);
err_workqueue:
	kfree(ps);

	return r;
}

static unsigned int persistent_status(struct dm_exception_store *store,
				  status_type_t status, char *result,
				  unsigned int maxlen)
{
	unsigned int sz = 0;

	switch (status) {
	case STATUSTYPE_INFO:
		break;
	case STATUSTYPE_TABLE:
		DMEMIT(" %s %llu", store->userspace_supports_overflow ? "PO" : "P",
		       (unsigned long long)store->chunk_size);
		break;
	case STATUSTYPE_IMA:
		*result = '\0';
		break;
	}

	return sz;
}

static struct dm_exception_store_type _persistent_type = {
	.name = "persistent",
	.module = THIS_MODULE,
	.ctr = persistent_ctr,
	.dtr = persistent_dtr,
	.read_metadata = persistent_read_metadata,
	.prepare_exception = persistent_prepare_exception,
	.commit_exception = persistent_commit_exception,
	.prepare_merge = persistent_prepare_merge,
	.commit_merge = persistent_commit_merge,
	.drop_snapshot = persistent_drop_snapshot,
	.usage = persistent_usage,
	.status = persistent_status,
};

static struct dm_exception_store_type _persistent_compat_type = {
	.name = "P",
	.module = THIS_MODULE,
	.ctr = persistent_ctr,
	.dtr = persistent_dtr,
	.read_metadata = persistent_read_metadata,
	.prepare_exception = persistent_prepare_exception,
	.commit_exception = persistent_commit_exception,
	.prepare_merge = persistent_prepare_merge,
	.commit_merge = persistent_commit_merge,
	.drop_snapshot = persistent_drop_snapshot,
	.usage = persistent_usage,
	.status = persistent_status,
};

int dm_persistent_snapshot_init(void)
{
	int r;

	r = dm_exception_store_type_register(&_persistent_type);
	if (r) {
		DMERR("Unable to register persistent exception store type");
		return r;
	}

	r = dm_exception_store_type_register(&_persistent_compat_type);
	if (r) {
		DMERR("Unable to register old-style persistent exception store type");
		dm_exception_store_type_unregister(&_persistent_type);
		return r;
	}

	return r;
}

void dm_persistent_snapshot_exit(void)
{
	dm_exception_store_type_unregister(&_persistent_type);
	dm_exception_store_type_unregister(&_persistent_compat_type);
}