dm-snap-persistent.c 20.5 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12
/*
 * 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/mm.h>
#include <linux/pagemap.h>
#include <linux/vmalloc.h>
13
#include <linux/export.h>
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
#include <linux/slab.h>
#include <linux/dm-io.h>

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

/*-----------------------------------------------------------------
 * 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

59 60
#define NUM_SNAPSHOT_HDR_CHUNKS 1

61
struct disk_header {
62
	__le32 magic;
63 64 65 66 67

	/*
	 * Is this snapshot valid.  There is no way of recovering
	 * an invalid snapshot.
	 */
68
	__le32 valid;
69 70 71 72 73

	/*
	 * Simple, incrementing version. no backward
	 * compatibility.
	 */
74
	__le32 version;
75 76

	/* In sectors */
77 78
	__le32 chunk_size;
} __packed;
79 80

struct disk_exception {
81 82 83 84 85
	__le64 old_chunk;
	__le64 new_chunk;
} __packed;

struct core_exception {
86 87 88 89 90 91 92 93 94 95 96 97 98
	uint64_t old_chunk;
	uint64_t new_chunk;
};

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

/*
 * The top level structure for a persistent exception store.
 */
struct pstore {
99
	struct dm_exception_store *store;
100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115
	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;

116 117 118 119 120 121 122
	/*
	 * 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;

123 124 125 126 127 128 129 130
	/*
	 * 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.
131 132 133 134 135 136 137 138 139 140 141 142 143 144
	 *
	 * 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.
145
	 */
146

147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167
	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;

168
	len = ps->store->chunk_size << SECTOR_SHIFT;
169 170 171 172 173 174 175

	/*
	 * Allocate the chunk_size block of memory that will hold
	 * a single metadata area.
	 */
	ps->area = vmalloc(len);
	if (!ps->area)
176
		goto err_area;
177

Joe Perches's avatar
Joe Perches committed
178
	ps->zero_area = vzalloc(len);
179 180
	if (!ps->zero_area)
		goto err_zero_area;
181

182 183 184 185
	ps->header_area = vmalloc(len);
	if (!ps->header_area)
		goto err_header_area;

186
	return 0;
187 188 189 190 191 192 193 194 195

err_header_area:
	vfree(ps->zero_area);

err_zero_area:
	vfree(ps->area);

err_area:
	return r;
196 197 198 199
}

static void free_area(struct pstore *ps)
{
200 201
	if (ps->area)
		vfree(ps->area);
202
	ps->area = NULL;
203 204 205

	if (ps->zero_area)
		vfree(ps->zero_area);
206
	ps->zero_area = NULL;
207 208 209 210

	if (ps->header_area)
		vfree(ps->header_area);
	ps->header_area = NULL;
211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
}

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.
 */
230 231
static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, int rw,
		    int metadata)
232 233
{
	struct dm_io_region where = {
234
		.bdev = dm_snap_cow(ps->store->snap)->bdev,
235 236
		.sector = ps->store->chunk_size * chunk,
		.count = ps->store->chunk_size,
237 238 239 240
	};
	struct dm_io_request io_req = {
		.bi_rw = rw,
		.mem.type = DM_IO_VMA,
241
		.mem.ptr.vma = area,
242 243 244 245 246 247 248 249 250 251 252 253 254 255 256
		.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 generic_make_request recursion.
	 */
257
	INIT_WORK_ONSTACK(&req.work, do_metadata);
258
	queue_work(ps->metadata_wq, &req.work);
259
	flush_work(&req.work);
260 261 262 263 264 265 266 267 268

	return req.result;
}

/*
 * Convert a metadata area index to a chunk index.
 */
static chunk_t area_location(struct pstore *ps, chunk_t area)
{
269
	return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
270 271 272 273 274 275 276 277 278 279 280 281 282
}

/*
 * Read or write a metadata area.  Remembering to skip the first
 * chunk which holds the header.
 */
static int area_io(struct pstore *ps, int rw)
{
	int r;
	chunk_t chunk;

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

283
	r = chunk_io(ps, ps->area, chunk, rw, 0);
284 285 286 287 288 289 290 291
	if (r)
		return r;

	return 0;
}

static void zero_memory_area(struct pstore *ps)
{
292
	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
293 294 295 296
}

static int zero_disk_area(struct pstore *ps, chunk_t area)
{
297
	return chunk_io(ps, ps->zero_area, area_location(ps, area), WRITE, 0);
298 299 300 301 302 303
}

static int read_header(struct pstore *ps, int *new_snapshot)
{
	int r;
	struct disk_header *dh;
304
	unsigned chunk_size;
305
	int chunk_size_supplied = 1;
306
	char *chunk_err;
307 308

	/*
309 310
	 * Use default chunk size (or logical_block_size, if larger)
	 * if none supplied
311
	 */
312 313
	if (!ps->store->chunk_size) {
		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
314 315
		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
					    bdev) >> 9);
316 317
		ps->store->chunk_mask = ps->store->chunk_size - 1;
		ps->store->chunk_shift = ffs(ps->store->chunk_size) - 1;
318 319 320
		chunk_size_supplied = 0;
	}

321
	ps->io_client = dm_io_client_create();
322 323 324 325 326 327 328
	if (IS_ERR(ps->io_client))
		return PTR_ERR(ps->io_client);

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

329
	r = chunk_io(ps, ps->header_area, 0, READ, 1);
330 331 332
	if (r)
		goto bad;

333
	dh = ps->header_area;
334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350

	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);

351
	if (ps->store->chunk_size == chunk_size)
352 353
		return 0;

354
	if (chunk_size_supplied)
355 356 357
		DMWARN("chunk size %u in device metadata overrides "
		       "table chunk size of %u.",
		       chunk_size, ps->store->chunk_size);
358 359 360 361

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

362 363 364
	r = dm_exception_store_set_chunk_size(ps->store, chunk_size,
					      &chunk_err);
	if (r) {
365 366
		DMERR("invalid on-disk chunk size %u: %s.",
		      chunk_size, chunk_err);
367 368
		return r;
	}
369 370 371 372 373 374 375 376 377 378 379 380 381

	r = alloc_area(ps);
	return r;

bad:
	free_area(ps);
	return r;
}

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

382
	memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT);
383

384
	dh = ps->header_area;
385 386 387
	dh->magic = cpu_to_le32(SNAP_MAGIC);
	dh->valid = cpu_to_le32(ps->valid);
	dh->version = cpu_to_le32(ps->version);
388
	dh->chunk_size = cpu_to_le32(ps->store->chunk_size);
389

390
	return chunk_io(ps, ps->header_area, 0, WRITE, 1);
391 392 393 394 395 396 397 398 399 400 401 402 403
}

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

	return ((struct disk_exception *) ps->area) + index;
}

static void read_exception(struct pstore *ps,
404
			   uint32_t index, struct core_exception *result)
405
{
406
	struct disk_exception *de = get_exception(ps, index);
407 408

	/* copy it */
409 410
	result->old_chunk = le64_to_cpu(de->old_chunk);
	result->new_chunk = le64_to_cpu(de->new_chunk);
411 412 413
}

static void write_exception(struct pstore *ps,
414
			    uint32_t index, struct core_exception *e)
415
{
416
	struct disk_exception *de = get_exception(ps, index);
417 418

	/* copy it */
419 420
	de->old_chunk = cpu_to_le64(e->old_chunk);
	de->new_chunk = cpu_to_le64(e->new_chunk);
421 422
}

423 424
static void clear_exception(struct pstore *ps, uint32_t index)
{
425
	struct disk_exception *de = get_exception(ps, index);
426 427

	/* clear it */
428 429
	de->old_chunk = 0;
	de->new_chunk = 0;
430 431
}

432 433 434 435 436
/*
 * Registers the exceptions that are present in the current area.
 * 'full' is filled in to indicate if the area has been
 * filled.
 */
437 438 439 440 441
static int insert_exceptions(struct pstore *ps,
			     int (*callback)(void *callback_context,
					     chunk_t old, chunk_t new),
			     void *callback_context,
			     int *full)
442 443 444
{
	int r;
	unsigned int i;
445
	struct core_exception e;
446 447 448 449 450

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

	for (i = 0; i < ps->exceptions_per_area; i++) {
451
		read_exception(ps, i, &e);
452 453 454 455 456 457 458

		/*
		 * 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.
		 */
459
		if (e.new_chunk == 0LL) {
460 461 462 463 464 465 466 467
			ps->current_committed = i;
			*full = 0;
			break;
		}

		/*
		 * Keep track of the start of the free chunks.
		 */
468 469
		if (ps->next_free <= e.new_chunk)
			ps->next_free = e.new_chunk + 1;
470 471 472 473

		/*
		 * Otherwise we add the exception to the snapshot.
		 */
474
		r = callback(callback_context, e.old_chunk, e.new_chunk);
475 476 477 478 479 480 481
		if (r)
			return r;
	}

	return 0;
}

482 483 484 485
static int read_exceptions(struct pstore *ps,
			   int (*callback)(void *callback_context, chunk_t old,
					   chunk_t new),
			   void *callback_context)
486 487 488 489 490 491 492 493 494 495 496 497
{
	int r, full = 1;

	/*
	 * Keeping reading chunks and inserting exceptions until
	 * we find a partially full area.
	 */
	for (ps->current_area = 0; full; ps->current_area++) {
		r = area_io(ps, READ);
		if (r)
			return r;

498
		r = insert_exceptions(ps, callback, callback_context, &full);
499 500 501 502 503 504 505 506 507 508 509 510 511 512
		if (r)
			return r;
	}

	ps->current_area--;

	return 0;
}

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

513 514 515 516
static void persistent_usage(struct dm_exception_store *store,
			     sector_t *total_sectors,
			     sector_t *sectors_allocated,
			     sector_t *metadata_sectors)
517
{
518 519 520
	struct pstore *ps = get_info(store);

	*sectors_allocated = ps->next_free * store->chunk_size;
521
	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
522 523 524 525 526 527

	/*
	 * 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.
	 */
528 529
	*metadata_sectors = (ps->current_area + 1 + NUM_SNAPSHOT_HDR_CHUNKS) *
			    store->chunk_size;
530 531
}

532
static void persistent_dtr(struct dm_exception_store *store)
533 534 535 536
{
	struct pstore *ps = get_info(store);

	destroy_workqueue(ps->metadata_wq);
537 538 539 540

	/* Created in read_header */
	if (ps->io_client)
		dm_io_client_destroy(ps->io_client);
541
	free_area(ps);
542 543 544 545 546

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

547 548 549
	kfree(ps);
}

550 551 552 553
static int persistent_read_metadata(struct dm_exception_store *store,
				    int (*callback)(void *callback_context,
						    chunk_t old, chunk_t new),
				    void *callback_context)
554 555 556 557 558 559 560 561 562 563 564 565 566 567
{
	int r, uninitialized_var(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.
	 */
568 569
	ps->exceptions_per_area = (ps->store->chunk_size << SECTOR_SHIFT) /
				  sizeof(struct disk_exception);
570
	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
571
				   sizeof(*ps->callbacks));
572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587
	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);
588
		if (r)
589
			DMWARN("zero_disk_area(0) failed");
590 591 592 593 594 595 596 597 598 599
		return r;
	}
	/*
	 * Sanity checks.
	 */
	if (ps->version != SNAPSHOT_DISK_VERSION) {
		DMWARN("unable to handle snapshot disk version %d",
		       ps->version);
		return -EINVAL;
	}
600

601 602 603 604 605
	/*
	 * Metadata are valid, but snapshot is invalidated
	 */
	if (!ps->valid)
		return 1;
606

607 608 609 610
	/*
	 * Read the metadata.
	 */
	r = read_exceptions(ps, callback, callback_context);
611

612
	return r;
613 614
}

615
static int persistent_prepare_exception(struct dm_exception_store *store,
616
					struct dm_exception *e)
617 618 619 620
{
	struct pstore *ps = get_info(store);
	uint32_t stride;
	chunk_t next_free;
621
	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
622 623

	/* Is there enough room ? */
624
	if (size < ((ps->next_free + 1) * store->chunk_size))
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641
		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.
	 */
	stride = (ps->exceptions_per_area + 1);
	next_free = ++ps->next_free;
	if (sector_div(next_free, stride) == 1)
		ps->next_free++;

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

642
static void persistent_commit_exception(struct dm_exception_store *store,
643
					struct dm_exception *e,
644 645
					void (*callback) (void *, int success),
					void *callback_context)
646 647 648
{
	unsigned int i;
	struct pstore *ps = get_info(store);
649
	struct core_exception ce;
650 651
	struct commit_callback *cb;

652 653 654
	ce.old_chunk = e->old_chunk;
	ce.new_chunk = e->new_chunk;
	write_exception(ps, ps->current_committed++, &ce);
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677

	/*
	 * 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) &&
678
	    zero_disk_area(ps, ps->current_area + 1))
679 680 681 682 683
		ps->valid = 0;

	/*
	 * Commit exceptions to disk.
	 */
684
	if (ps->valid && area_io(ps, WRITE_FLUSH_FUA))
685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703
		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;
}

704 705 706 707 708
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);
709
	struct core_exception ce;
710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729
	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, READ);
		if (r < 0)
			return r;
		ps->current_committed = ps->exceptions_per_area;
	}

730 731 732
	read_exception(ps, ps->current_committed - 1, &ce);
	*last_old_chunk = ce.old_chunk;
	*last_new_chunk = ce.new_chunk;
733 734 735 736 737 738 739 740

	/*
	 * 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->current_committed - 1 - nr_consecutive,
741 742 743
			       &ce);
		if (ce.old_chunk != *last_old_chunk - nr_consecutive ||
		    ce.new_chunk != *last_new_chunk - nr_consecutive)
744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760
			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);

761
	r = area_io(ps, WRITE_FLUSH_FUA);
762 763 764 765 766 767 768 769 770 771 772 773 774 775 776
	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.
	 */
777 778
	ps->next_free = area_location(ps, ps->current_area) +
			ps->current_committed + 1;
779 780 781 782

	return 0;
}

783
static void persistent_drop_snapshot(struct dm_exception_store *store)
784 785 786 787 788 789 790 791
{
	struct pstore *ps = get_info(store);

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

792 793
static int persistent_ctr(struct dm_exception_store *store,
			  unsigned argc, char **argv)
794 795 796 797
{
	struct pstore *ps;

	/* allocate the pstore */
798
	ps = kzalloc(sizeof(*ps), GFP_KERNEL);
799 800 801
	if (!ps)
		return -ENOMEM;

802
	ps->store = store;
803 804 805
	ps->valid = 1;
	ps->version = SNAPSHOT_DISK_VERSION;
	ps->area = NULL;
806 807
	ps->zero_area = NULL;
	ps->header_area = NULL;
808
	ps->next_free = NUM_SNAPSHOT_HDR_CHUNKS + 1; /* header and 1st area */
809 810 811 812 813 814
	ps->current_committed = 0;

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

815
	ps->metadata_wq = alloc_workqueue("ksnaphd", WQ_MEM_RECLAIM, 0);
816 817 818 819 820 821 822 823 824 825 826
	if (!ps->metadata_wq) {
		kfree(ps);
		DMERR("couldn't start header metadata update thread");
		return -ENOMEM;
	}

	store->context = ps;

	return 0;
}

827 828 829
static unsigned persistent_status(struct dm_exception_store *store,
				  status_type_t status, char *result,
				  unsigned maxlen)
830
{
831 832 833 834 835 836
	unsigned sz = 0;

	switch (status) {
	case STATUSTYPE_INFO:
		break;
	case STATUSTYPE_TABLE:
837
		DMEMIT(" P %llu", (unsigned long long)store->chunk_size);
838
	}
839 840 841 842 843 844 845 846 847 848 849 850

	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,
851 852
	.prepare_merge = persistent_prepare_merge,
	.commit_merge = persistent_commit_merge,
853
	.drop_snapshot = persistent_drop_snapshot,
854
	.usage = persistent_usage,
855 856 857 858 859 860 861 862 863 864 865
	.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,
866 867
	.prepare_merge = persistent_prepare_merge,
	.commit_merge = persistent_commit_merge,
868
	.drop_snapshot = persistent_drop_snapshot,
869
	.usage = persistent_usage,
870 871 872
	.status = persistent_status,
};

873 874
int dm_persistent_snapshot_init(void)
{
875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891
	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;
892 893 894 895
}

void dm_persistent_snapshot_exit(void)
{
896 897
	dm_exception_store_type_unregister(&_persistent_type);
	dm_exception_store_type_unregister(&_persistent_compat_type);
898
}