dm-snap.c 27.9 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
/*
 * 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"

25 26
#define DM_MSG_PREFIX "snapshots"

Linus Torvalds's avatar
Linus Torvalds committed
27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
/*
 * 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

42
struct workqueue_struct *ksnapd;
David Howells's avatar
David Howells committed
43
static void flush_queued_bios(struct work_struct *work);
44

Linus Torvalds's avatar
Linus Torvalds committed
45 46 47 48 49 50 51 52 53 54
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;

55 56 57 58 59
	/*
	 * Short-term queue of pending exceptions prior to submission.
	 */
	struct list_head list;

Linus Torvalds's avatar
Linus Torvalds committed
60
	/*
61
	 * The primary pending_exception is the one that holds
62
	 * the ref_count and the list of origin_bios for a
63 64
	 * group of pending_exceptions.  It is always last to get freed.
	 * These fields get set up when writing to the origin.
Linus Torvalds's avatar
Linus Torvalds committed
65
	 */
66 67 68 69 70 71 72 73 74
	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.
	 */
75
	atomic_t ref_count;
Linus Torvalds's avatar
Linus Torvalds committed
76 77 78 79 80 81 82 83 84 85 86 87 88 89 90

	/* 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
 */
91 92
static struct kmem_cache *exception_cache;
static struct kmem_cache *pending_cache;
Linus Torvalds's avatar
Linus Torvalds committed
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
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) {
124
		DMERR("unable to allocate memory");
Linus Torvalds's avatar
Linus Torvalds committed
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
		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;
}

231
static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem)
Linus Torvalds's avatar
Linus Torvalds committed
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
{
	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;
}

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
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;
}

Linus Torvalds's avatar
Linus Torvalds committed
435 436 437 438 439 440 441 442 443 444 445
/*
 * 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;

446
	if (argc != 4) {
447
		ti->error = "requires exactly 4 arguments";
Linus Torvalds's avatar
Linus Torvalds committed
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
		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;
	}

484 485
	r = set_chunk_size(s, argv[3], &ti->error);
	if (r)
Linus Torvalds's avatar
Linus Torvalds committed
486 487 488 489 490
		goto bad3;

	s->type = persistent;

	s->valid = 1;
491
	s->active = 0;
Linus Torvalds's avatar
Linus Torvalds committed
492 493
	s->last_percent = 0;
	init_rwsem(&s->lock);
494
	spin_lock_init(&s->pe_lock);
Linus Torvalds's avatar
Linus Torvalds committed
495 496 497 498 499 500 501 502 503 504 505 506
	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')
507
		r = dm_create_persistent(&s->store);
Linus Torvalds's avatar
Linus Torvalds committed
508
	else
509
		r = dm_create_transient(&s->store);
Linus Torvalds's avatar
Linus Torvalds committed
510 511 512 513 514 515 516 517 518 519 520 521 522

	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;
	}

523
	/* Metadata must only be loaded into one table at once */
524 525 526 527 528
	r = s->store.read_metadata(&s->store);
	if (r) {
		ti->error = "Failed to read snapshot metadata";
		goto bad6;
	}
529

530
	bio_list_init(&s->queued_bios);
David Howells's avatar
David Howells committed
531
	INIT_WORK(&s->queued_bios_work, flush_queued_bios);
532

Linus Torvalds's avatar
Linus Torvalds committed
533
	/* Add snapshot to the list of snapshots for this origin */
534
	/* Exceptions aren't triggered till snapshot_resume() is called */
Linus Torvalds's avatar
Linus Torvalds committed
535 536 537 538 539 540 541
	if (register_snapshot(s)) {
		r = -EINVAL;
		ti->error = "Cannot register snapshot origin";
		goto bad6;
	}

	ti->private = s;
542
	ti->split_io = s->chunk_size;
Linus Torvalds's avatar
Linus Torvalds committed
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

	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;

571 572
	flush_workqueue(ksnapd);

573 574
	/* Prevent further origin writes from using this snapshot. */
	/* After this returns there can be no new kcopyd jobs. */
Linus Torvalds's avatar
Linus Torvalds committed
575 576
	unregister_snapshot(s);

577 578
	kcopyd_client_destroy(s->kcopyd_client);

Linus Torvalds's avatar
Linus Torvalds committed
579 580 581 582 583 584 585 586
	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);
587

Linus Torvalds's avatar
Linus Torvalds committed
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605
	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;
	}
}

David Howells's avatar
David Howells committed
606
static void flush_queued_bios(struct work_struct *work)
607
{
David Howells's avatar
David Howells committed
608 609
	struct dm_snapshot *s =
		container_of(work, struct dm_snapshot, queued_bios_work);
610 611 612 613 614 615 616 617 618 619
	struct bio *queued_bios;
	unsigned long flags;

	spin_lock_irqsave(&s->pe_lock, flags);
	queued_bios = bio_list_get(&s->queued_bios);
	spin_unlock_irqrestore(&s->pe_lock, flags);

	flush_bios(queued_bios);
}

Linus Torvalds's avatar
Linus Torvalds committed
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634
/*
 * 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;
	}
}

635
static void __invalidate_snapshot(struct dm_snapshot *s, int err)
636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652
{
	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 (s->store.drop_snapshot)
		s->store.drop_snapshot(&s->store);

	s->valid = 0;

	dm_table_event(s->table);
}

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
static void get_pending_exception(struct pending_exception *pe)
{
	atomic_inc(&pe->ref_count);
}

static struct bio *put_pending_exception(struct pending_exception *pe)
{
	struct pending_exception *primary_pe;
	struct bio *origin_bios = NULL;

	primary_pe = pe->primary_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 (primary_pe &&
	    atomic_dec_and_test(&primary_pe->ref_count))
		origin_bios = bio_list_get(&primary_pe->origin_bios);

	/*
	 * 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->ref_count))
		free_pending_exception(primary_pe);

	return origin_bios;
}

Linus Torvalds's avatar
Linus Torvalds committed
690 691 692 693
static void pending_complete(struct pending_exception *pe, int success)
{
	struct exception *e;
	struct dm_snapshot *s = pe->snap;
694 695 696
	struct bio *origin_bios = NULL;
	struct bio *snapshot_bios = NULL;
	int error = 0;
Linus Torvalds's avatar
Linus Torvalds committed
697

698 699
	if (!success) {
		/* Read/write error - snapshot is unusable */
Linus Torvalds's avatar
Linus Torvalds committed
700
		down_write(&s->lock);
701
		__invalidate_snapshot(s, -EIO);
702
		error = 1;
703 704 705 706 707
		goto out;
	}

	e = alloc_exception();
	if (!e) {
Linus Torvalds's avatar
Linus Torvalds committed
708
		down_write(&s->lock);
709
		__invalidate_snapshot(s, -ENOMEM);
710
		error = 1;
711 712 713
		goto out;
	}
	*e = pe->e;
Linus Torvalds's avatar
Linus Torvalds committed
714

715 716 717
	down_write(&s->lock);
	if (!s->valid) {
		free_exception(e);
718
		error = 1;
719
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
720 721
	}

722 723 724 725
	/*
	 * Add a proper exception, and remove the
	 * in-flight exception from the list.
	 */
726 727
	insert_exception(&s->complete, e);

Linus Torvalds's avatar
Linus Torvalds committed
728
 out:
729
	remove_exception(&pe->e);
730
	snapshot_bios = bio_list_get(&pe->snapshot_bios);
731
	origin_bios = put_pending_exception(pe);
Linus Torvalds's avatar
Linus Torvalds committed
732

733 734 735 736 737 738 739 740 741
	up_write(&s->lock);

	/* Submit any pending write bios */
	if (error)
		error_bios(snapshot_bios);
	else
		flush_bios(snapshot_bios);

	flush_bios(origin_bios);
Linus Torvalds's avatar
Linus Torvalds committed
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
}

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.
 */
771
static void start_copy(struct pending_exception *pe)
Linus Torvalds's avatar
Linus Torvalds committed
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
{
	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);
815 816
		goto out;
	}
Linus Torvalds's avatar
Linus Torvalds committed
817

818 819 820 821 822 823 824
	/*
	 * 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);
Linus Torvalds's avatar
Linus Torvalds committed
825

826 827 828 829
	if (!s->valid) {
		free_pending_exception(pe);
		return NULL;
	}
Linus Torvalds's avatar
Linus Torvalds committed
830

831 832 833 834 835
	e = lookup_exception(&s->pending, chunk);
	if (e) {
		free_pending_exception(pe);
		pe = container_of(e, struct pending_exception, e);
		goto out;
Linus Torvalds's avatar
Linus Torvalds committed
836 837
	}

838 839 840 841
	pe->e.old_chunk = chunk;
	bio_list_init(&pe->origin_bios);
	bio_list_init(&pe->snapshot_bios);
	pe->primary_pe = NULL;
842
	atomic_set(&pe->ref_count, 0);
843 844 845 846 847 848 849 850
	pe->snap = s;
	pe->started = 0;

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

851
	get_pending_exception(pe);
852 853 854
	insert_exception(&s->pending, &pe->e);

 out:
Linus Torvalds's avatar
Linus Torvalds committed
855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872
	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 r = 1;
	chunk_t chunk;
873
	struct pending_exception *pe = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
874 875 876 877

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

	/* Full snapshots are not usable */
878
	/* To get here the table must be live so s->active is always set. */
Linus Torvalds's avatar
Linus Torvalds committed
879
	if (!s->valid)
880
		return -EIO;
Linus Torvalds's avatar
Linus Torvalds committed
881

882 883 884
	if (unlikely(bio_barrier(bio)))
		return -EOPNOTSUPP;

885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
	/* 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;
	}

Linus Torvalds's avatar
Linus Torvalds committed
901 902 903 904 905 906
	/*
	 * 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) {
907 908
		pe = __find_pending_exception(s, bio);
		if (!pe) {
909
			__invalidate_snapshot(s, -ENOMEM);
910 911
			r = -EIO;
			goto out_unlock;
Linus Torvalds's avatar
Linus Torvalds committed
912 913
		}

914 915 916
		remap_exception(s, &pe->e, bio);
		bio_list_add(&pe->snapshot_bios, bio);

917 918
		r = 0;

919 920 921
		if (!pe->started) {
			/* this is protected by snap->lock */
			pe->started = 1;
922
			up_write(&s->lock);
923
			start_copy(pe);
924 925 926
			goto out;
		}
	} else
Linus Torvalds's avatar
Linus Torvalds committed
927 928 929 930 931 932
		/*
		 * 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.
		 */
933
		bio->bi_bdev = s->origin->bdev;
Linus Torvalds's avatar
Linus Torvalds committed
934

935 936 937
 out_unlock:
	up_write(&s->lock);
 out:
Linus Torvalds's avatar
Linus Torvalds committed
938 939 940 941 942 943 944
	return r;
}

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

945 946 947
	down_write(&s->lock);
	s->active = 1;
	up_write(&s->lock);
Linus Torvalds's avatar
Linus Torvalds committed
948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964
}

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);
965 966 967
				snprintf(result, maxlen, "%llu/%llu",
					(unsigned long long)numerator,
					(unsigned long long)denominator);
Linus Torvalds's avatar
Linus Torvalds committed
968 969 970 971 972 973 974 975 976 977 978 979
			}
			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.
		 */
980
		snprintf(result, maxlen, "%s %s %c %llu",
Linus Torvalds's avatar
Linus Torvalds committed
981
			 snap->origin->name, snap->cow->name,
982 983
			 snap->type,
			 (unsigned long long)snap->chunk_size);
Linus Torvalds's avatar
Linus Torvalds committed
984 985 986 987 988 989 990 991 992 993 994
		break;
	}

	return 0;
}

/*-----------------------------------------------------------------
 * Origin methods
 *---------------------------------------------------------------*/
static int __origin_write(struct list_head *snapshots, struct bio *bio)
{
995
	int r = 1, first = 0;
Linus Torvalds's avatar
Linus Torvalds committed
996 997
	struct dm_snapshot *snap;
	struct exception *e;
998
	struct pending_exception *pe, *next_pe, *primary_pe = NULL;
Linus Torvalds's avatar
Linus Torvalds committed
999
	chunk_t chunk;
1000
	LIST_HEAD(pe_queue);
Linus Torvalds's avatar
Linus Torvalds committed
1001 1002 1003 1004

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

1005 1006
		down_write(&snap->lock);

1007 1008
		/* Only deal with valid and active snapshots */
		if (!snap->valid || !snap->active)
1009
			goto next_snapshot;
Linus Torvalds's avatar
Linus Torvalds committed
1010

1011 1012
		/* Nothing to do if writing beyond end of snapshot */
		if (bio->bi_sector >= dm_table_get_size(snap->table))
1013
			goto next_snapshot;
Linus Torvalds's avatar
Linus Torvalds committed
1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024

		/*
		 * 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.
1025
		 *
1026
		 * ref_count is initialised to 1 so pending_complete()
1027
		 * won't destroy the primary_pe while we're inside this loop.
Linus Torvalds's avatar
Linus Torvalds committed
1028 1029
		 */
		e = lookup_exception(&snap->complete, chunk);
1030 1031 1032 1033 1034
		if (e)
			goto next_snapshot;

		pe = __find_pending_exception(snap, bio);
		if (!pe) {
1035
			__invalidate_snapshot(snap, -ENOMEM);
1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
			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;
Linus Torvalds's avatar
Linus Torvalds committed
1049
			}
1050 1051 1052 1053 1054 1055 1056 1057

			bio_list_add(&primary_pe->origin_bios, bio);

			r = 0;
		}

		if (!pe->primary_pe) {
			pe->primary_pe = primary_pe;
1058
			get_pending_exception(primary_pe);
1059 1060 1061 1062 1063
		}

		if (!pe->started) {
			pe->started = 1;
			list_add_tail(&pe->list, &pe_queue);
Linus Torvalds's avatar
Linus Torvalds committed
1064 1065
		}

1066
 next_snapshot:
Linus Torvalds's avatar
Linus Torvalds committed
1067 1068 1069
		up_write(&snap->lock);
	}

1070
	if (!primary_pe)
1071
		return r;
1072 1073 1074

	/*
	 * If this is the first time we're processing this chunk and
1075
	 * ref_count is now 1 it means all the pending exceptions
1076 1077 1078 1079
	 * 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.
	 */

1080
	if (first && atomic_dec_and_test(&primary_pe->ref_count)) {
1081 1082 1083
		flush_bios(bio_list_get(&primary_pe->origin_bios));
		free_pending_exception(primary_pe);
		/* If we got here, pe_queue is necessarily empty. */
1084
		return r;
1085 1086
	}

Linus Torvalds's avatar
Linus Torvalds committed
1087 1088 1089
	/*
	 * Now that we have a complete pe list we can start the copying.
	 */
1090 1091
	list_for_each_entry_safe(pe, next_pe, &pe_queue, list)
		start_copy(pe);
Linus Torvalds's avatar
Linus Torvalds committed
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

	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) {
1128
		ti->error = "origin: incorrect number of arguments";
Linus Torvalds's avatar
Linus Torvalds committed
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
		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;

1155 1156 1157
	if (unlikely(bio_barrier(bio)))
		return -EOPNOTSUPP;

Linus Torvalds's avatar
Linus Torvalds committed
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
	/* 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",
1205
	.version = {1, 5, 0},
Linus Torvalds's avatar
Linus Torvalds committed
1206 1207 1208 1209 1210 1211 1212 1213 1214 1215
	.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",
1216
	.version = {1, 5, 0},
Linus Torvalds's avatar
Linus Torvalds committed
1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236
	.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) {
1237
		DMERR("Origin target register failed %d", r);
Linus Torvalds's avatar
Linus Torvalds committed
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
		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;
	}

1268
	pending_pool = mempool_create_slab_pool(128, pending_cache);
Linus Torvalds's avatar
Linus Torvalds committed
1269 1270 1271 1272 1273 1274
	if (!pending_pool) {
		DMERR("Couldn't create pending pool.");
		r = -ENOMEM;
		goto bad5;
	}

1275 1276 1277 1278 1279 1280 1281
	ksnapd = create_singlethread_workqueue("ksnapd");
	if (!ksnapd) {
		DMERR("Failed to create ksnapd workqueue.");
		r = -ENOMEM;
		goto bad6;
	}

Linus Torvalds's avatar
Linus Torvalds committed
1282 1283
	return 0;

1284 1285
      bad6:
	mempool_destroy(pending_pool);
Linus Torvalds's avatar
Linus Torvalds committed
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302
      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;

1303 1304
	destroy_workqueue(ksnapd);

Linus Torvalds's avatar
Linus Torvalds committed
1305 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325
	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");