Blame view
drivers/md/dm-snap.c
52.9 KB
1da177e4c
|
1 2 3 4 5 6 7 8 9 |
/* * dm-snapshot.c * * Copyright (C) 2001-2002 Sistina Software (UK) Limited. * * This file is released under the GPL. */ #include <linux/blkdev.h> |
1da177e4c
|
10 |
#include <linux/device-mapper.h> |
90fa1527b
|
11 |
#include <linux/delay.h> |
1da177e4c
|
12 13 14 15 16 17 18 19 |
#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> |
6f3c3f0af
|
20 |
#include <linux/log2.h> |
a765e20ee
|
21 |
#include <linux/dm-kcopyd.h> |
1da177e4c
|
22 |
|
aea53d92f
|
23 |
#include "dm-exception-store.h" |
1da177e4c
|
24 |
|
72d948616
|
25 |
#define DM_MSG_PREFIX "snapshots" |
d698aa450
|
26 27 28 29 |
static const char dm_snapshot_merge_target_name[] = "snapshot-merge"; #define dm_target_is_snapshot_merge(ti) \ ((ti)->type->name == dm_snapshot_merge_target_name) |
1da177e4c
|
30 |
/* |
cd45daffd
|
31 32 33 |
* The size of the mempool used to track chunks in use. */ #define MIN_IOS 256 |
ccc45ea8a
|
34 35 36 |
#define DM_TRACKED_CHUNK_HASH_SIZE 16 #define DM_TRACKED_CHUNK_HASH(x) ((unsigned long)(x) & \ (DM_TRACKED_CHUNK_HASH_SIZE - 1)) |
191437a53
|
37 |
struct dm_exception_table { |
ccc45ea8a
|
38 39 40 41 42 43 44 45 46 |
uint32_t hash_mask; unsigned hash_shift; struct list_head *table; }; struct dm_snapshot { struct rw_semaphore lock; struct dm_dev *origin; |
fc56f6fbc
|
47 48 49 |
struct dm_dev *cow; struct dm_target *ti; |
ccc45ea8a
|
50 51 52 |
/* List of snapshots per Origin */ struct list_head list; |
d8ddb1cff
|
53 54 55 56 |
/* * You can't use a snapshot if this is 0 (e.g. if full). * A snapshot-merge target never clears this. */ |
ccc45ea8a
|
57 58 59 60 |
int valid; /* Origin writes don't trigger exceptions until this is set */ int active; |
ccc45ea8a
|
61 |
atomic_t pending_exceptions_count; |
924e600d4
|
62 |
mempool_t *pending_pool; |
191437a53
|
63 64 |
struct dm_exception_table pending; struct dm_exception_table complete; |
ccc45ea8a
|
65 66 67 68 69 70 |
/* * pe_lock protects all pending_exception operations and access * as well as the snapshot_bios list. */ spinlock_t pe_lock; |
924e600d4
|
71 72 73 74 |
/* Chunks with outstanding reads */ spinlock_t tracked_chunk_lock; mempool_t *tracked_chunk_pool; struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE]; |
ccc45ea8a
|
75 76 77 78 |
/* The on disk metadata handler */ struct dm_exception_store *store; struct dm_kcopyd_client *kcopyd_client; |
924e600d4
|
79 80 81 82 83 84 |
/* Wait for events based on state_bits */ unsigned long state_bits; /* Range of chunks currently being merged. */ chunk_t first_merging_chunk; int num_merging_chunks; |
1e03f97e4
|
85 |
|
d8ddb1cff
|
86 87 88 89 90 91 92 93 94 95 96 97 98 99 |
/* * The merge operation failed if this flag is set. * Failure modes are handled as follows: * - I/O error reading the header * => don't load the target; abort. * - Header does not have "valid" flag set * => use the origin; forget about the snapshot. * - I/O error when reading exceptions * => don't load the target; abort. * (We can't use the intermediate origin state.) * - I/O error while merging * => stop merging; set merge_failed; process I/O normally. */ int merge_failed; |
9fe862548
|
100 101 102 103 104 |
/* * Incoming bios that overlap with chunks being merged must wait * for them to be committed. */ struct bio_list bios_queued_during_merge; |
ccc45ea8a
|
105 |
}; |
1e03f97e4
|
106 107 108 109 110 111 112 113 |
/* * state_bits: * RUNNING_MERGE - Merge operation is in progress. * SHUTDOWN_MERGE - Set to signal that merge needs to be stopped; * cleared afterwards. */ #define RUNNING_MERGE 0 #define SHUTDOWN_MERGE 1 |
c24110450
|
114 115 116 117 118 |
struct dm_dev *dm_snap_origin(struct dm_snapshot *s) { return s->origin; } EXPORT_SYMBOL(dm_snap_origin); |
fc56f6fbc
|
119 120 121 122 123 |
struct dm_dev *dm_snap_cow(struct dm_snapshot *s) { return s->cow; } EXPORT_SYMBOL(dm_snap_cow); |
ccc45ea8a
|
124 125 126 127 128 129 130 131 132 133 134 135 136 137 |
static sector_t chunk_to_sector(struct dm_exception_store *store, chunk_t chunk) { return chunk << store->chunk_shift; } static int bdev_equal(struct block_device *lhs, struct block_device *rhs) { /* * There is only ever one instance of a particular block * device so we can compare pointers safely. */ return lhs == rhs; } |
028867ac2
|
138 |
struct dm_snap_pending_exception { |
1d4989c85
|
139 |
struct dm_exception e; |
1da177e4c
|
140 141 142 143 144 145 146 |
/* * Origin buffers waiting for this to complete are held * in a bio list */ struct bio_list origin_bios; struct bio_list snapshot_bios; |
1da177e4c
|
147 148 149 150 151 152 153 154 |
/* Pointer back to snapshot context */ struct dm_snapshot *snap; /* * 1 indicates the exception has already been sent to * kcopyd. */ int started; |
a6e50b409
|
155 156 157 158 159 160 161 |
/* * For writing a complete chunk, bypassing the copy. */ struct bio *full_bio; bio_end_io_t *full_bio_end_io; void *full_bio_private; |
1da177e4c
|
162 163 164 165 166 167 |
}; /* * Hash table mapping origin volumes to lists of snapshots and * a lock to protect it */ |
e18b890bb
|
168 169 |
static struct kmem_cache *exception_cache; static struct kmem_cache *pending_cache; |
1da177e4c
|
170 |
|
cd45daffd
|
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 |
struct dm_snap_tracked_chunk { struct hlist_node node; chunk_t chunk; }; static struct kmem_cache *tracked_chunk_cache; static struct dm_snap_tracked_chunk *track_chunk(struct dm_snapshot *s, chunk_t chunk) { struct dm_snap_tracked_chunk *c = mempool_alloc(s->tracked_chunk_pool, GFP_NOIO); unsigned long flags; c->chunk = chunk; spin_lock_irqsave(&s->tracked_chunk_lock, flags); hlist_add_head(&c->node, &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)]); spin_unlock_irqrestore(&s->tracked_chunk_lock, flags); return c; } static void stop_tracking_chunk(struct dm_snapshot *s, struct dm_snap_tracked_chunk *c) { unsigned long flags; spin_lock_irqsave(&s->tracked_chunk_lock, flags); hlist_del(&c->node); spin_unlock_irqrestore(&s->tracked_chunk_lock, flags); mempool_free(c, s->tracked_chunk_pool); } |
a8d41b59f
|
206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 |
static int __chunk_is_tracked(struct dm_snapshot *s, chunk_t chunk) { struct dm_snap_tracked_chunk *c; struct hlist_node *hn; int found = 0; spin_lock_irq(&s->tracked_chunk_lock); hlist_for_each_entry(c, hn, &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)], node) { if (c->chunk == chunk) { found = 1; break; } } spin_unlock_irq(&s->tracked_chunk_lock); return found; } |
1da177e4c
|
226 |
/* |
615d1eb9c
|
227 228 229 230 231 232 233 234 235 236 |
* This conflicting I/O is extremely improbable in the caller, * so msleep(1) is sufficient and there is no need for a wait queue. */ static void __check_for_conflicting_io(struct dm_snapshot *s, chunk_t chunk) { while (__chunk_is_tracked(s, chunk)) msleep(1); } /* |
1da177e4c
|
237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 |
* 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; |
73dfd078c
|
257 258 259 |
static DECLARE_WAIT_QUEUE_HEAD(_pending_exceptions_done); static DEFINE_SPINLOCK(_pending_exceptions_done_spinlock); static uint64_t _pending_exceptions_done_count; |
1da177e4c
|
260 261 262 263 264 265 266 |
static int init_origin_hash(void) { int i; _origins = kmalloc(ORIGIN_HASH_SIZE * sizeof(struct list_head), GFP_KERNEL); if (!_origins) { |
72d948616
|
267 |
DMERR("unable to allocate memory"); |
1da177e4c
|
268 269 270 271 272 273 274 275 276 277 278 279 280 281 |
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); } |
028867ac2
|
282 |
static unsigned origin_hash(struct block_device *bdev) |
1da177e4c
|
283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 |
{ 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); } /* |
c1f0c183f
|
307 308 309 310 |
* _origins_lock must be held when calling this function. * Returns number of snapshots registered using the supplied cow device, plus: * snap_src - a snapshot suitable for use as a source of exception handover * snap_dest - a snapshot capable of receiving exception handover. |
9d3b15c4c
|
311 312 |
* snap_merge - an existing snapshot-merge target linked to the same origin. * There can be at most one snapshot-merge target. The parameter is optional. |
c1f0c183f
|
313 |
* |
9d3b15c4c
|
314 |
* Possible return values and states of snap_src and snap_dest. |
c1f0c183f
|
315 316 317 318 319 320 321 322 |
* 0: NULL, NULL - first new snapshot * 1: snap_src, NULL - normal snapshot * 2: snap_src, snap_dest - waiting for handover * 2: snap_src, NULL - handed over, waiting for old to be deleted * 1: NULL, snap_dest - source got destroyed without handover */ static int __find_snapshots_sharing_cow(struct dm_snapshot *snap, struct dm_snapshot **snap_src, |
9d3b15c4c
|
323 324 |
struct dm_snapshot **snap_dest, struct dm_snapshot **snap_merge) |
c1f0c183f
|
325 326 327 328 329 330 331 332 333 334 335 |
{ struct dm_snapshot *s; struct origin *o; int count = 0; int active; o = __lookup_origin(snap->origin->bdev); if (!o) goto out; list_for_each_entry(s, &o->snapshots, list) { |
9d3b15c4c
|
336 337 |
if (dm_target_is_snapshot_merge(s->ti) && snap_merge) *snap_merge = s; |
c1f0c183f
|
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 |
if (!bdev_equal(s->cow->bdev, snap->cow->bdev)) continue; down_read(&s->lock); active = s->active; up_read(&s->lock); if (active) { if (snap_src) *snap_src = s; } else if (snap_dest) *snap_dest = s; count++; } out: return count; } /* * On success, returns 1 if this snapshot is a handover destination, * otherwise returns 0. */ static int __validate_exception_handover(struct dm_snapshot *snap) { struct dm_snapshot *snap_src = NULL, *snap_dest = NULL; |
9d3b15c4c
|
365 |
struct dm_snapshot *snap_merge = NULL; |
c1f0c183f
|
366 367 |
/* Does snapshot need exceptions handed over to it? */ |
9d3b15c4c
|
368 369 |
if ((__find_snapshots_sharing_cow(snap, &snap_src, &snap_dest, &snap_merge) == 2) || |
c1f0c183f
|
370 371 372 373 374 375 376 377 378 379 380 381 |
snap_dest) { snap->ti->error = "Snapshot cow pairing for exception " "table handover failed"; return -EINVAL; } /* * If no snap_src was found, snap cannot become a handover * destination. */ if (!snap_src) return 0; |
9d3b15c4c
|
382 383 384 385 386 387 388 389 390 391 392 393 394 |
/* * Non-snapshot-merge handover? */ if (!dm_target_is_snapshot_merge(snap->ti)) return 1; /* * Do not allow more than one merging snapshot. */ if (snap_merge) { snap->ti->error = "A snapshot is already merging."; return -EINVAL; } |
1e03f97e4
|
395 396 397 398 399 400 |
if (!snap_src->store->type->prepare_merge || !snap_src->store->type->commit_merge) { snap->ti->error = "Snapshot exception store does not " "support snapshot-merge."; return -EINVAL; } |
c1f0c183f
|
401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 |
return 1; } static void __insert_snapshot(struct origin *o, struct dm_snapshot *s) { struct dm_snapshot *l; /* Sort the list according to chunk size, largest-first smallest-last */ list_for_each_entry(l, &o->snapshots, list) if (l->store->chunk_size < s->store->chunk_size) break; list_add_tail(&s->list, &l->list); } /* |
1da177e4c
|
416 417 |
* Make a note of the snapshot and its origin so we can look it * up when the origin has a write on it. |
c1f0c183f
|
418 419 420 421 |
* * Also validate snapshot exception store handovers. * On success, returns 1 if this registration is a handover destination, * otherwise returns 0. |
1da177e4c
|
422 423 424 |
*/ static int register_snapshot(struct dm_snapshot *snap) { |
c1f0c183f
|
425 |
struct origin *o, *new_o = NULL; |
1da177e4c
|
426 |
struct block_device *bdev = snap->origin->bdev; |
c1f0c183f
|
427 |
int r = 0; |
1da177e4c
|
428 |
|
60c856c8e
|
429 430 431 |
new_o = kmalloc(sizeof(*new_o), GFP_KERNEL); if (!new_o) return -ENOMEM; |
1da177e4c
|
432 |
down_write(&_origins_lock); |
1da177e4c
|
433 |
|
c1f0c183f
|
434 435 436 437 438 439 440 |
r = __validate_exception_handover(snap); if (r < 0) { kfree(new_o); goto out; } o = __lookup_origin(bdev); |
60c856c8e
|
441 442 443 |
if (o) kfree(new_o); else { |
1da177e4c
|
444 |
/* New origin */ |
60c856c8e
|
445 |
o = new_o; |
1da177e4c
|
446 447 448 449 450 451 452 |
/* Initialise the struct */ INIT_LIST_HEAD(&o->snapshots); o->bdev = bdev; __insert_origin(o); } |
c1f0c183f
|
453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 |
__insert_snapshot(o, snap); out: up_write(&_origins_lock); return r; } /* * Move snapshot to correct place in list according to chunk size. */ static void reregister_snapshot(struct dm_snapshot *s) { struct block_device *bdev = s->origin->bdev; down_write(&_origins_lock); list_del(&s->list); __insert_snapshot(__lookup_origin(bdev), s); |
1da177e4c
|
472 473 |
up_write(&_origins_lock); |
1da177e4c
|
474 475 476 477 478 479 480 481 482 483 |
} 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); |
c1f0c183f
|
484 |
if (o && list_empty(&o->snapshots)) { |
1da177e4c
|
485 486 487 488 489 490 491 492 493 |
list_del(&o->hash_list); kfree(o); } up_write(&_origins_lock); } /* * Implementation of the exception hash tables. |
d74f81f8a
|
494 495 |
* The lowest hash_shift bits of the chunk number are ignored, allowing * some consecutive chunks to be grouped together. |
1da177e4c
|
496 |
*/ |
3510cb94f
|
497 498 |
static int dm_exception_table_init(struct dm_exception_table *et, uint32_t size, unsigned hash_shift) |
1da177e4c
|
499 500 |
{ unsigned int i; |
d74f81f8a
|
501 |
et->hash_shift = hash_shift; |
1da177e4c
|
502 503 504 505 506 507 508 509 510 511 |
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; } |
3510cb94f
|
512 513 |
static void dm_exception_table_exit(struct dm_exception_table *et, struct kmem_cache *mem) |
1da177e4c
|
514 515 |
{ struct list_head *slot; |
1d4989c85
|
516 |
struct dm_exception *ex, *next; |
1da177e4c
|
517 518 519 520 521 522 523 524 525 526 527 528 |
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); } |
191437a53
|
529 |
static uint32_t exception_hash(struct dm_exception_table *et, chunk_t chunk) |
1da177e4c
|
530 |
{ |
d74f81f8a
|
531 |
return (chunk >> et->hash_shift) & et->hash_mask; |
1da177e4c
|
532 |
} |
3510cb94f
|
533 |
static void dm_remove_exception(struct dm_exception *e) |
1da177e4c
|
534 535 536 537 538 539 540 541 |
{ list_del(&e->hash_list); } /* * Return the exception data for a sector, or NULL if not * remapped. */ |
3510cb94f
|
542 543 |
static struct dm_exception *dm_lookup_exception(struct dm_exception_table *et, chunk_t chunk) |
1da177e4c
|
544 545 |
{ struct list_head *slot; |
1d4989c85
|
546 |
struct dm_exception *e; |
1da177e4c
|
547 548 549 |
slot = &et->table[exception_hash(et, chunk)]; list_for_each_entry (e, slot, hash_list) |
d74f81f8a
|
550 551 |
if (chunk >= e->old_chunk && chunk <= e->old_chunk + dm_consecutive_chunk_count(e)) |
1da177e4c
|
552 553 554 555 |
return e; return NULL; } |
3510cb94f
|
556 |
static struct dm_exception *alloc_completed_exception(void) |
1da177e4c
|
557 |
{ |
1d4989c85
|
558 |
struct dm_exception *e; |
1da177e4c
|
559 560 561 562 563 564 565 |
e = kmem_cache_alloc(exception_cache, GFP_NOIO); if (!e) e = kmem_cache_alloc(exception_cache, GFP_ATOMIC); return e; } |
3510cb94f
|
566 |
static void free_completed_exception(struct dm_exception *e) |
1da177e4c
|
567 568 569 |
{ kmem_cache_free(exception_cache, e); } |
92e868122
|
570 |
static struct dm_snap_pending_exception *alloc_pending_exception(struct dm_snapshot *s) |
1da177e4c
|
571 |
{ |
92e868122
|
572 573 |
struct dm_snap_pending_exception *pe = mempool_alloc(s->pending_pool, GFP_NOIO); |
879129d20
|
574 |
atomic_inc(&s->pending_exceptions_count); |
92e868122
|
575 576 577 |
pe->snap = s; return pe; |
1da177e4c
|
578 |
} |
028867ac2
|
579 |
static void free_pending_exception(struct dm_snap_pending_exception *pe) |
1da177e4c
|
580 |
{ |
879129d20
|
581 582 583 584 585 |
struct dm_snapshot *s = pe->snap; mempool_free(pe, s->pending_pool); smp_mb__before_atomic_dec(); atomic_dec(&s->pending_exceptions_count); |
1da177e4c
|
586 |
} |
3510cb94f
|
587 588 |
static void dm_insert_exception(struct dm_exception_table *eh, struct dm_exception *new_e) |
d74f81f8a
|
589 |
{ |
d74f81f8a
|
590 |
struct list_head *l; |
1d4989c85
|
591 |
struct dm_exception *e = NULL; |
d74f81f8a
|
592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 |
l = &eh->table[exception_hash(eh, new_e->old_chunk)]; /* Add immediately if this table doesn't support consecutive chunks */ if (!eh->hash_shift) goto out; /* List is ordered by old_chunk */ list_for_each_entry_reverse(e, l, hash_list) { /* Insert after an existing chunk? */ if (new_e->old_chunk == (e->old_chunk + dm_consecutive_chunk_count(e) + 1) && new_e->new_chunk == (dm_chunk_number(e->new_chunk) + dm_consecutive_chunk_count(e) + 1)) { dm_consecutive_chunk_count_inc(e); |
3510cb94f
|
607 |
free_completed_exception(new_e); |
d74f81f8a
|
608 609 610 611 612 613 614 615 616 |
return; } /* Insert before an existing chunk? */ if (new_e->old_chunk == (e->old_chunk - 1) && new_e->new_chunk == (dm_chunk_number(e->new_chunk) - 1)) { dm_consecutive_chunk_count_inc(e); e->old_chunk--; e->new_chunk--; |
3510cb94f
|
617 |
free_completed_exception(new_e); |
d74f81f8a
|
618 619 620 621 622 623 624 625 626 627 |
return; } if (new_e->old_chunk > e->old_chunk) break; } out: list_add(&new_e->hash_list, e ? &e->hash_list : l); } |
a159c1ac5
|
628 629 630 631 632 |
/* * Callback used by the exception stores to load exceptions when * initialising. */ static int dm_add_exception(void *context, chunk_t old, chunk_t new) |
1da177e4c
|
633 |
{ |
a159c1ac5
|
634 |
struct dm_snapshot *s = context; |
1d4989c85
|
635 |
struct dm_exception *e; |
1da177e4c
|
636 |
|
3510cb94f
|
637 |
e = alloc_completed_exception(); |
1da177e4c
|
638 639 640 641 |
if (!e) return -ENOMEM; e->old_chunk = old; |
d74f81f8a
|
642 643 |
/* Consecutive_count is implicitly initialised to zero */ |
1da177e4c
|
644 |
e->new_chunk = new; |
d74f81f8a
|
645 |
|
3510cb94f
|
646 |
dm_insert_exception(&s->complete, e); |
d74f81f8a
|
647 |
|
1da177e4c
|
648 649 |
return 0; } |
7e201b351
|
650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 |
/* * Return a minimum chunk size of all snapshots that have the specified origin. * Return zero if the origin has no snapshots. */ static sector_t __minimum_chunk_size(struct origin *o) { struct dm_snapshot *snap; unsigned chunk_size = 0; if (o) list_for_each_entry(snap, &o->snapshots, list) chunk_size = min_not_zero(chunk_size, snap->store->chunk_size); return chunk_size; } |
1da177e4c
|
666 667 668 669 670 671 672 673 674 675 676 677 678 |
/* * 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; } /* |
1da177e4c
|
679 680 |
* Allocate room for a suitable hash table. */ |
fee1998e9
|
681 |
static int init_hash_tables(struct dm_snapshot *s) |
1da177e4c
|
682 683 684 685 686 687 688 |
{ 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... */ |
fc56f6fbc
|
689 |
cow_dev_size = get_dev_size(s->cow->bdev); |
1da177e4c
|
690 691 |
origin_dev_size = get_dev_size(s->origin->bdev); max_buckets = calc_max_buckets(); |
fee1998e9
|
692 |
hash_size = min(origin_dev_size, cow_dev_size) >> s->store->chunk_shift; |
1da177e4c
|
693 |
hash_size = min(hash_size, max_buckets); |
8e87b9b81
|
694 695 |
if (hash_size < 64) hash_size = 64; |
8defd8308
|
696 |
hash_size = rounddown_pow_of_two(hash_size); |
3510cb94f
|
697 698 |
if (dm_exception_table_init(&s->complete, hash_size, DM_CHUNK_CONSECUTIVE_BITS)) |
1da177e4c
|
699 700 701 702 703 704 705 706 707 |
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; |
3510cb94f
|
708 709 |
if (dm_exception_table_init(&s->pending, hash_size, 0)) { dm_exception_table_exit(&s->complete, exception_cache); |
1da177e4c
|
710 711 712 713 714 |
return -ENOMEM; } return 0; } |
1e03f97e4
|
715 716 717 718 719 720 |
static void merge_shutdown(struct dm_snapshot *s) { clear_bit_unlock(RUNNING_MERGE, &s->state_bits); smp_mb__after_clear_bit(); wake_up_bit(&s->state_bits, RUNNING_MERGE); } |
9fe862548
|
721 722 723 724 725 726 727 |
static struct bio *__release_queued_bios_after_merge(struct dm_snapshot *s) { s->first_merging_chunk = 0; s->num_merging_chunks = 0; return bio_list_get(&s->bios_queued_during_merge); } |
1e03f97e4
|
728 729 730 731 732 733 734 |
/* * Remove one chunk from the index of completed exceptions. */ static int __remove_single_exception_chunk(struct dm_snapshot *s, chunk_t old_chunk) { struct dm_exception *e; |
1e03f97e4
|
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 |
e = dm_lookup_exception(&s->complete, old_chunk); if (!e) { DMERR("Corruption detected: exception for block %llu is " "on disk but not in memory", (unsigned long long)old_chunk); return -EINVAL; } /* * If this is the only chunk using this exception, remove exception. */ if (!dm_consecutive_chunk_count(e)) { dm_remove_exception(e); free_completed_exception(e); return 0; } /* * The chunk may be either at the beginning or the end of a * group of consecutive chunks - never in the middle. We are * removing chunks in the opposite order to that in which they * were added, so this should always be true. * Decrement the consecutive chunk counter and adjust the * starting point if necessary. */ if (old_chunk == e->old_chunk) { e->old_chunk++; e->new_chunk++; } else if (old_chunk != e->old_chunk + dm_consecutive_chunk_count(e)) { DMERR("Attempt to merge block %llu from the " "middle of a chunk range [%llu - %llu]", (unsigned long long)old_chunk, (unsigned long long)e->old_chunk, (unsigned long long) e->old_chunk + dm_consecutive_chunk_count(e)); return -EINVAL; } dm_consecutive_chunk_count_dec(e); return 0; } |
9fe862548
|
778 779 780 |
static void flush_bios(struct bio *bio); static int remove_single_exception_chunk(struct dm_snapshot *s) |
1e03f97e4
|
781 |
{ |
9fe862548
|
782 783 784 |
struct bio *b = NULL; int r; chunk_t old_chunk = s->first_merging_chunk + s->num_merging_chunks - 1; |
1e03f97e4
|
785 786 |
down_write(&s->lock); |
9fe862548
|
787 788 789 790 791 792 793 794 795 796 797 798 799 800 |
/* * Process chunks (and associated exceptions) in reverse order * so that dm_consecutive_chunk_count_dec() accounting works. */ do { r = __remove_single_exception_chunk(s, old_chunk); if (r) goto out; } while (old_chunk-- > s->first_merging_chunk); b = __release_queued_bios_after_merge(s); out: |
1e03f97e4
|
801 |
up_write(&s->lock); |
9fe862548
|
802 803 |
if (b) flush_bios(b); |
1e03f97e4
|
804 805 806 |
return r; } |
73dfd078c
|
807 808 |
static int origin_write_extent(struct dm_snapshot *merging_snap, sector_t sector, unsigned chunk_size); |
1e03f97e4
|
809 810 |
static void merge_callback(int read_err, unsigned long write_err, void *context); |
73dfd078c
|
811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 |
static uint64_t read_pending_exceptions_done_count(void) { uint64_t pending_exceptions_done; spin_lock(&_pending_exceptions_done_spinlock); pending_exceptions_done = _pending_exceptions_done_count; spin_unlock(&_pending_exceptions_done_spinlock); return pending_exceptions_done; } static void increment_pending_exceptions_done_count(void) { spin_lock(&_pending_exceptions_done_spinlock); _pending_exceptions_done_count++; spin_unlock(&_pending_exceptions_done_spinlock); wake_up_all(&_pending_exceptions_done); } |
1e03f97e4
|
830 831 |
static void snapshot_merge_next_chunks(struct dm_snapshot *s) { |
8a2d52862
|
832 |
int i, linear_chunks; |
1e03f97e4
|
833 834 |
chunk_t old_chunk, new_chunk; struct dm_io_region src, dest; |
8a2d52862
|
835 |
sector_t io_size; |
73dfd078c
|
836 |
uint64_t previous_count; |
1e03f97e4
|
837 838 839 840 841 842 843 844 845 846 847 848 |
BUG_ON(!test_bit(RUNNING_MERGE, &s->state_bits)); if (unlikely(test_bit(SHUTDOWN_MERGE, &s->state_bits))) goto shut; /* * valid flag never changes during merge, so no lock required. */ if (!s->valid) { DMERR("Snapshot is invalid: can't merge"); goto shut; } |
8a2d52862
|
849 850 851 |
linear_chunks = s->store->type->prepare_merge(s->store, &old_chunk, &new_chunk); if (linear_chunks <= 0) { |
d8ddb1cff
|
852 |
if (linear_chunks < 0) { |
1e03f97e4
|
853 854 |
DMERR("Read error in exception store: " "shutting down merge"); |
d8ddb1cff
|
855 856 857 858 |
down_write(&s->lock); s->merge_failed = 1; up_write(&s->lock); } |
1e03f97e4
|
859 860 |
goto shut; } |
8a2d52862
|
861 862 863 864 865 866 867 868 869 |
/* Adjust old_chunk and new_chunk to reflect start of linear region */ old_chunk = old_chunk + 1 - linear_chunks; new_chunk = new_chunk + 1 - linear_chunks; /* * Use one (potentially large) I/O to copy all 'linear_chunks' * from the exception store to the origin */ io_size = linear_chunks * s->store->chunk_size; |
1e03f97e4
|
870 |
|
1e03f97e4
|
871 872 |
dest.bdev = s->origin->bdev; dest.sector = chunk_to_sector(s->store, old_chunk); |
8a2d52862
|
873 |
dest.count = min(io_size, get_dev_size(dest.bdev) - dest.sector); |
1e03f97e4
|
874 875 876 877 |
src.bdev = s->cow->bdev; src.sector = chunk_to_sector(s->store, new_chunk); src.count = dest.count; |
73dfd078c
|
878 879 880 881 882 883 884 885 886 887 |
/* * Reallocate any exceptions needed in other snapshots then * wait for the pending exceptions to complete. * Each time any pending exception (globally on the system) * completes we are woken and repeat the process to find out * if we can proceed. While this may not seem a particularly * efficient algorithm, it is not expected to have any * significant impact on performance. */ previous_count = read_pending_exceptions_done_count(); |
8a2d52862
|
888 |
while (origin_write_extent(s, dest.sector, io_size)) { |
73dfd078c
|
889 890 891 892 893 894 |
wait_event(_pending_exceptions_done, (read_pending_exceptions_done_count() != previous_count)); /* Retry after the wait, until all exceptions are done. */ previous_count = read_pending_exceptions_done_count(); } |
9fe862548
|
895 896 |
down_write(&s->lock); s->first_merging_chunk = old_chunk; |
8a2d52862
|
897 |
s->num_merging_chunks = linear_chunks; |
9fe862548
|
898 |
up_write(&s->lock); |
8a2d52862
|
899 900 901 |
/* Wait until writes to all 'linear_chunks' drain */ for (i = 0; i < linear_chunks; i++) __check_for_conflicting_io(s, old_chunk + i); |
9fe862548
|
902 |
|
1e03f97e4
|
903 904 905 906 907 908 |
dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, merge_callback, s); return; shut: merge_shutdown(s); } |
9fe862548
|
909 |
static void error_bios(struct bio *bio); |
1e03f97e4
|
910 911 912 |
static void merge_callback(int read_err, unsigned long write_err, void *context) { struct dm_snapshot *s = context; |
9fe862548
|
913 |
struct bio *b = NULL; |
1e03f97e4
|
914 915 916 917 918 919 920 921 |
if (read_err || write_err) { if (read_err) DMERR("Read error: shutting down merge."); else DMERR("Write error: shutting down merge."); goto shut; } |
9fe862548
|
922 923 |
if (s->store->type->commit_merge(s->store, s->num_merging_chunks) < 0) { |
1e03f97e4
|
924 925 926 |
DMERR("Write error in exception store: shutting down merge"); goto shut; } |
9fe862548
|
927 928 |
if (remove_single_exception_chunk(s) < 0) goto shut; |
1e03f97e4
|
929 930 931 932 933 |
snapshot_merge_next_chunks(s); return; shut: |
9fe862548
|
934 |
down_write(&s->lock); |
d8ddb1cff
|
935 |
s->merge_failed = 1; |
9fe862548
|
936 937 938 |
b = __release_queued_bios_after_merge(s); up_write(&s->lock); error_bios(b); |
1e03f97e4
|
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 |
merge_shutdown(s); } static void start_merge(struct dm_snapshot *s) { if (!test_and_set_bit(RUNNING_MERGE, &s->state_bits)) snapshot_merge_next_chunks(s); } static int wait_schedule(void *ptr) { schedule(); return 0; } /* * Stop the merging process and wait until it finishes. */ static void stop_merge(struct dm_snapshot *s) { set_bit(SHUTDOWN_MERGE, &s->state_bits); wait_on_bit(&s->state_bits, RUNNING_MERGE, wait_schedule, TASK_UNINTERRUPTIBLE); clear_bit(SHUTDOWN_MERGE, &s->state_bits); } |
1da177e4c
|
965 |
/* |
1da177e4c
|
966 967 968 969 970 |
* 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; |
cd45daffd
|
971 |
int i; |
1da177e4c
|
972 |
int r = -EINVAL; |
fc56f6fbc
|
973 |
char *origin_path, *cow_path; |
10b8106a7
|
974 975 |
unsigned args_used, num_flush_requests = 1; fmode_t origin_mode = FMODE_READ; |
1da177e4c
|
976 |
|
4c7e3bf44
|
977 |
if (argc != 4) { |
72d948616
|
978 |
ti->error = "requires exactly 4 arguments"; |
1da177e4c
|
979 |
r = -EINVAL; |
fc56f6fbc
|
980 |
goto bad; |
1da177e4c
|
981 |
} |
10b8106a7
|
982 983 984 985 |
if (dm_target_is_snapshot_merge(ti)) { num_flush_requests = 2; origin_mode = FMODE_WRITE; } |
fc56f6fbc
|
986 987 |
s = kmalloc(sizeof(*s), GFP_KERNEL); if (!s) { |
a2d2b0345
|
988 |
ti->error = "Cannot allocate private snapshot structure"; |
fc56f6fbc
|
989 990 991 |
r = -ENOMEM; goto bad; } |
c24110450
|
992 993 994 995 996 997 998 999 1000 |
origin_path = argv[0]; argv++; argc--; r = dm_get_device(ti, origin_path, origin_mode, &s->origin); if (r) { ti->error = "Cannot get origin device"; goto bad_origin; } |
fc56f6fbc
|
1001 1002 1003 |
cow_path = argv[0]; argv++; argc--; |
024d37e95
|
1004 |
r = dm_get_device(ti, cow_path, dm_table_get_mode(ti->table), &s->cow); |
fc56f6fbc
|
1005 1006 1007 1008 1009 1010 |
if (r) { ti->error = "Cannot get COW device"; goto bad_cow; } r = dm_exception_store_create(ti, argc, argv, s, &args_used, &s->store); |
fee1998e9
|
1011 1012 |
if (r) { ti->error = "Couldn't create exception store"; |
1da177e4c
|
1013 |
r = -EINVAL; |
fc56f6fbc
|
1014 |
goto bad_store; |
1da177e4c
|
1015 |
} |
fee1998e9
|
1016 1017 |
argv += args_used; argc -= args_used; |
fc56f6fbc
|
1018 |
s->ti = ti; |
1da177e4c
|
1019 |
s->valid = 1; |
aa14edeb9
|
1020 |
s->active = 0; |
879129d20
|
1021 |
atomic_set(&s->pending_exceptions_count, 0); |
1da177e4c
|
1022 |
init_rwsem(&s->lock); |
c1f0c183f
|
1023 |
INIT_LIST_HEAD(&s->list); |
ca3a931fd
|
1024 |
spin_lock_init(&s->pe_lock); |
1e03f97e4
|
1025 |
s->state_bits = 0; |
d8ddb1cff
|
1026 |
s->merge_failed = 0; |
9fe862548
|
1027 1028 1029 |
s->first_merging_chunk = 0; s->num_merging_chunks = 0; bio_list_init(&s->bios_queued_during_merge); |
1da177e4c
|
1030 1031 |
/* Allocate hash table for COW data */ |
fee1998e9
|
1032 |
if (init_hash_tables(s)) { |
1da177e4c
|
1033 1034 |
ti->error = "Unable to allocate hash table space"; r = -ENOMEM; |
fee1998e9
|
1035 |
goto bad_hash_tables; |
1da177e4c
|
1036 |
} |
fa34ce730
|
1037 1038 1039 |
s->kcopyd_client = dm_kcopyd_client_create(); if (IS_ERR(s->kcopyd_client)) { r = PTR_ERR(s->kcopyd_client); |
1da177e4c
|
1040 |
ti->error = "Could not create kcopyd client"; |
fee1998e9
|
1041 |
goto bad_kcopyd; |
1da177e4c
|
1042 |
} |
92e868122
|
1043 1044 1045 |
s->pending_pool = mempool_create_slab_pool(MIN_IOS, pending_cache); if (!s->pending_pool) { ti->error = "Could not allocate mempool for pending exceptions"; |
fee1998e9
|
1046 |
goto bad_pending_pool; |
92e868122
|
1047 |
} |
cd45daffd
|
1048 1049 1050 1051 1052 |
s->tracked_chunk_pool = mempool_create_slab_pool(MIN_IOS, tracked_chunk_cache); if (!s->tracked_chunk_pool) { ti->error = "Could not allocate tracked_chunk mempool for " "tracking reads"; |
92e868122
|
1053 |
goto bad_tracked_chunk_pool; |
cd45daffd
|
1054 1055 1056 1057 1058 1059 |
} for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++) INIT_HLIST_HEAD(&s->tracked_chunk_hash[i]); spin_lock_init(&s->tracked_chunk_lock); |
c1f0c183f
|
1060 |
ti->private = s; |
10b8106a7
|
1061 |
ti->num_flush_requests = num_flush_requests; |
c1f0c183f
|
1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 |
/* Add snapshot to the list of snapshots for this origin */ /* Exceptions aren't triggered till snapshot_resume() is called */ r = register_snapshot(s); if (r == -ENOMEM) { ti->error = "Snapshot origin struct allocation failed"; goto bad_load_and_register; } else if (r < 0) { /* invalid handover, register_snapshot has set ti->error */ goto bad_load_and_register; } /* * Metadata must only be loaded into one table at once, so skip this * if metadata will be handed over during resume. * Chunk size will be set during the handover - set it to zero to * ensure it's ignored. */ if (r > 0) { s->store->chunk_size = 0; return 0; } |
493df71c6
|
1084 1085 |
r = s->store->type->read_metadata(s->store, dm_add_exception, (void *)s); |
0764147b1
|
1086 |
if (r < 0) { |
f9cea4f70
|
1087 |
ti->error = "Failed to read snapshot metadata"; |
c1f0c183f
|
1088 |
goto bad_read_metadata; |
0764147b1
|
1089 1090 1091 |
} else if (r > 0) { s->valid = 0; DMWARN("Snapshot is marked invalid."); |
f9cea4f70
|
1092 |
} |
aa14edeb9
|
1093 |
|
3f2412dc8
|
1094 1095 |
if (!s->store->chunk_size) { ti->error = "Chunk size not set"; |
c1f0c183f
|
1096 |
goto bad_read_metadata; |
1da177e4c
|
1097 |
} |
d02168495
|
1098 |
ti->split_io = s->store->chunk_size; |
1da177e4c
|
1099 1100 |
return 0; |
c1f0c183f
|
1101 1102 |
bad_read_metadata: unregister_snapshot(s); |
fee1998e9
|
1103 |
bad_load_and_register: |
cd45daffd
|
1104 |
mempool_destroy(s->tracked_chunk_pool); |
fee1998e9
|
1105 |
bad_tracked_chunk_pool: |
92e868122
|
1106 |
mempool_destroy(s->pending_pool); |
fee1998e9
|
1107 |
bad_pending_pool: |
eb69aca5d
|
1108 |
dm_kcopyd_client_destroy(s->kcopyd_client); |
1da177e4c
|
1109 |
|
fee1998e9
|
1110 |
bad_kcopyd: |
3510cb94f
|
1111 1112 |
dm_exception_table_exit(&s->pending, pending_cache); dm_exception_table_exit(&s->complete, exception_cache); |
1da177e4c
|
1113 |
|
fee1998e9
|
1114 |
bad_hash_tables: |
fc56f6fbc
|
1115 |
dm_exception_store_destroy(s->store); |
1da177e4c
|
1116 |
|
fc56f6fbc
|
1117 1118 |
bad_store: dm_put_device(ti, s->cow); |
fee1998e9
|
1119 |
|
fc56f6fbc
|
1120 |
bad_cow: |
c24110450
|
1121 1122 1123 |
dm_put_device(ti, s->origin); bad_origin: |
fc56f6fbc
|
1124 1125 1126 |
kfree(s); bad: |
1da177e4c
|
1127 1128 |
return r; } |
31c93a0c2
|
1129 1130 |
static void __free_exceptions(struct dm_snapshot *s) { |
eb69aca5d
|
1131 |
dm_kcopyd_client_destroy(s->kcopyd_client); |
31c93a0c2
|
1132 |
s->kcopyd_client = NULL; |
3510cb94f
|
1133 1134 |
dm_exception_table_exit(&s->pending, pending_cache); dm_exception_table_exit(&s->complete, exception_cache); |
31c93a0c2
|
1135 |
} |
c1f0c183f
|
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 |
static void __handover_exceptions(struct dm_snapshot *snap_src, struct dm_snapshot *snap_dest) { union { struct dm_exception_table table_swap; struct dm_exception_store *store_swap; } u; /* * Swap all snapshot context information between the two instances. */ u.table_swap = snap_dest->complete; snap_dest->complete = snap_src->complete; snap_src->complete = u.table_swap; u.store_swap = snap_dest->store; snap_dest->store = snap_src->store; snap_src->store = u.store_swap; snap_dest->store->snap = snap_dest; snap_src->store->snap = snap_src; snap_dest->ti->split_io = snap_dest->store->chunk_size; snap_dest->valid = snap_src->valid; /* * Set source invalid to ensure it receives no further I/O. */ snap_src->valid = 0; } |
1da177e4c
|
1166 1167 |
static void snapshot_dtr(struct dm_target *ti) { |
cd45daffd
|
1168 1169 1170 |
#ifdef CONFIG_DM_DEBUG int i; #endif |
028867ac2
|
1171 |
struct dm_snapshot *s = ti->private; |
c1f0c183f
|
1172 |
struct dm_snapshot *snap_src = NULL, *snap_dest = NULL; |
1da177e4c
|
1173 |
|
c1f0c183f
|
1174 1175 |
down_read(&_origins_lock); /* Check whether exception handover must be cancelled */ |
9d3b15c4c
|
1176 |
(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); |
c1f0c183f
|
1177 1178 1179 1180 1181 1182 1183 |
if (snap_src && snap_dest && (s == snap_src)) { down_write(&snap_dest->lock); snap_dest->valid = 0; up_write(&snap_dest->lock); DMERR("Cancelling snapshot handover."); } up_read(&_origins_lock); |
1e03f97e4
|
1184 1185 |
if (dm_target_is_snapshot_merge(ti)) stop_merge(s); |
138728dc9
|
1186 1187 |
/* Prevent further origin writes from using this snapshot. */ /* After this returns there can be no new kcopyd jobs. */ |
1da177e4c
|
1188 |
unregister_snapshot(s); |
879129d20
|
1189 |
while (atomic_read(&s->pending_exceptions_count)) |
90fa1527b
|
1190 |
msleep(1); |
879129d20
|
1191 1192 1193 1194 1195 |
/* * Ensure instructions in mempool_destroy aren't reordered * before atomic_read. */ smp_mb(); |
cd45daffd
|
1196 1197 1198 1199 1200 1201 |
#ifdef CONFIG_DM_DEBUG for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++) BUG_ON(!hlist_empty(&s->tracked_chunk_hash[i])); #endif mempool_destroy(s->tracked_chunk_pool); |
31c93a0c2
|
1202 |
__free_exceptions(s); |
1da177e4c
|
1203 |
|
92e868122
|
1204 |
mempool_destroy(s->pending_pool); |
fee1998e9
|
1205 |
dm_exception_store_destroy(s->store); |
138728dc9
|
1206 |
|
fc56f6fbc
|
1207 |
dm_put_device(ti, s->cow); |
c24110450
|
1208 |
dm_put_device(ti, s->origin); |
1da177e4c
|
1209 1210 1211 1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 |
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; } } |
515ad66cc
|
1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 |
static int do_origin(struct dm_dev *origin, struct bio *bio); /* * Flush a list of buffers. */ static void retry_origin_bios(struct dm_snapshot *s, struct bio *bio) { struct bio *n; int r; while (bio) { n = bio->bi_next; bio->bi_next = NULL; r = do_origin(s->origin, bio); if (r == DM_MAPIO_REMAPPED) generic_make_request(bio); bio = n; } } |
1da177e4c
|
1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 |
/* * 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; |
6712ecf8f
|
1255 |
bio_io_error(bio); |
1da177e4c
|
1256 1257 1258 |
bio = n; } } |
695368ac3
|
1259 |
static void __invalidate_snapshot(struct dm_snapshot *s, int err) |
76df1c651
|
1260 1261 1262 1263 1264 1265 1266 1267 |
{ if (!s->valid) return; if (err == -EIO) DMERR("Invalidating snapshot: Error reading/writing."); else if (err == -ENOMEM) DMERR("Invalidating snapshot: Unable to allocate exception."); |
493df71c6
|
1268 1269 |
if (s->store->type->drop_snapshot) s->store->type->drop_snapshot(s->store); |
76df1c651
|
1270 1271 |
s->valid = 0; |
fc56f6fbc
|
1272 |
dm_table_event(s->ti->table); |
76df1c651
|
1273 |
} |
028867ac2
|
1274 |
static void pending_complete(struct dm_snap_pending_exception *pe, int success) |
1da177e4c
|
1275 |
{ |
1d4989c85
|
1276 |
struct dm_exception *e; |
1da177e4c
|
1277 |
struct dm_snapshot *s = pe->snap; |
9d493fa8c
|
1278 1279 |
struct bio *origin_bios = NULL; struct bio *snapshot_bios = NULL; |
a6e50b409
|
1280 |
struct bio *full_bio = NULL; |
9d493fa8c
|
1281 |
int error = 0; |
1da177e4c
|
1282 |
|
76df1c651
|
1283 1284 |
if (!success) { /* Read/write error - snapshot is unusable */ |
1da177e4c
|
1285 |
down_write(&s->lock); |
695368ac3
|
1286 |
__invalidate_snapshot(s, -EIO); |
9d493fa8c
|
1287 |
error = 1; |
76df1c651
|
1288 1289 |
goto out; } |
3510cb94f
|
1290 |
e = alloc_completed_exception(); |
76df1c651
|
1291 |
if (!e) { |
1da177e4c
|
1292 |
down_write(&s->lock); |
695368ac3
|
1293 |
__invalidate_snapshot(s, -ENOMEM); |
9d493fa8c
|
1294 |
error = 1; |
76df1c651
|
1295 1296 1297 |
goto out; } *e = pe->e; |
1da177e4c
|
1298 |
|
76df1c651
|
1299 1300 |
down_write(&s->lock); if (!s->valid) { |
3510cb94f
|
1301 |
free_completed_exception(e); |
9d493fa8c
|
1302 |
error = 1; |
76df1c651
|
1303 |
goto out; |
1da177e4c
|
1304 |
} |
615d1eb9c
|
1305 1306 |
/* Check for conflicting reads */ __check_for_conflicting_io(s, pe->e.old_chunk); |
a8d41b59f
|
1307 1308 |
/* |
9d493fa8c
|
1309 1310 1311 |
* Add a proper exception, and remove the * in-flight exception from the list. */ |
3510cb94f
|
1312 |
dm_insert_exception(&s->complete, e); |
76df1c651
|
1313 |
|
a2d2b0345
|
1314 |
out: |
3510cb94f
|
1315 |
dm_remove_exception(&pe->e); |
9d493fa8c
|
1316 |
snapshot_bios = bio_list_get(&pe->snapshot_bios); |
515ad66cc
|
1317 |
origin_bios = bio_list_get(&pe->origin_bios); |
a6e50b409
|
1318 1319 1320 1321 1322 |
full_bio = pe->full_bio; if (full_bio) { full_bio->bi_end_io = pe->full_bio_end_io; full_bio->bi_private = pe->full_bio_private; } |
515ad66cc
|
1323 |
free_pending_exception(pe); |
1da177e4c
|
1324 |
|
73dfd078c
|
1325 |
increment_pending_exceptions_done_count(); |
9d493fa8c
|
1326 1327 1328 |
up_write(&s->lock); /* Submit any pending write bios */ |
a6e50b409
|
1329 1330 1331 |
if (error) { if (full_bio) bio_io_error(full_bio); |
9d493fa8c
|
1332 |
error_bios(snapshot_bios); |
a6e50b409
|
1333 1334 1335 |
} else { if (full_bio) bio_endio(full_bio, 0); |
9d493fa8c
|
1336 |
flush_bios(snapshot_bios); |
a6e50b409
|
1337 |
} |
9d493fa8c
|
1338 |
|
515ad66cc
|
1339 |
retry_origin_bios(s, origin_bios); |
1da177e4c
|
1340 1341 1342 1343 |
} static void commit_callback(void *context, int success) { |
028867ac2
|
1344 |
struct dm_snap_pending_exception *pe = context; |
1da177e4c
|
1345 1346 1347 1348 1349 1350 1351 |
pending_complete(pe, success); } /* * Called when the copy I/O has finished. kcopyd actually runs * this code so don't block. */ |
4cdc1d1fa
|
1352 |
static void copy_callback(int read_err, unsigned long write_err, void *context) |
1da177e4c
|
1353 |
{ |
028867ac2
|
1354 |
struct dm_snap_pending_exception *pe = context; |
1da177e4c
|
1355 1356 1357 1358 1359 1360 1361 |
struct dm_snapshot *s = pe->snap; if (read_err || write_err) pending_complete(pe, 0); else /* Update the metadata if we are persistent */ |
493df71c6
|
1362 1363 |
s->store->type->commit_exception(s->store, &pe->e, commit_callback, pe); |
1da177e4c
|
1364 1365 1366 1367 1368 |
} /* * Dispatches the copy operation to kcopyd. */ |
028867ac2
|
1369 |
static void start_copy(struct dm_snap_pending_exception *pe) |
1da177e4c
|
1370 1371 |
{ struct dm_snapshot *s = pe->snap; |
22a1ceb1e
|
1372 |
struct dm_io_region src, dest; |
1da177e4c
|
1373 1374 1375 1376 1377 1378 |
struct block_device *bdev = s->origin->bdev; sector_t dev_size; dev_size = get_dev_size(bdev); src.bdev = bdev; |
71fab00a6
|
1379 |
src.sector = chunk_to_sector(s->store, pe->e.old_chunk); |
df96eee67
|
1380 |
src.count = min((sector_t)s->store->chunk_size, dev_size - src.sector); |
1da177e4c
|
1381 |
|
fc56f6fbc
|
1382 |
dest.bdev = s->cow->bdev; |
71fab00a6
|
1383 |
dest.sector = chunk_to_sector(s->store, pe->e.new_chunk); |
1da177e4c
|
1384 1385 1386 |
dest.count = src.count; /* Hand over to kcopyd */ |
a2d2b0345
|
1387 |
dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, copy_callback, pe); |
1da177e4c
|
1388 |
} |
a6e50b409
|
1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 |
static void full_bio_end_io(struct bio *bio, int error) { void *callback_data = bio->bi_private; dm_kcopyd_do_callback(callback_data, 0, error ? 1 : 0); } static void start_full_bio(struct dm_snap_pending_exception *pe, struct bio *bio) { struct dm_snapshot *s = pe->snap; void *callback_data; pe->full_bio = bio; pe->full_bio_end_io = bio->bi_end_io; pe->full_bio_private = bio->bi_private; callback_data = dm_kcopyd_prepare_callback(s->kcopyd_client, copy_callback, pe); bio->bi_end_io = full_bio_end_io; bio->bi_private = callback_data; generic_make_request(bio); } |
2913808eb
|
1414 1415 1416 |
static struct dm_snap_pending_exception * __lookup_pending_exception(struct dm_snapshot *s, chunk_t chunk) { |
3510cb94f
|
1417 |
struct dm_exception *e = dm_lookup_exception(&s->pending, chunk); |
2913808eb
|
1418 1419 1420 1421 1422 1423 |
if (!e) return NULL; return container_of(e, struct dm_snap_pending_exception, e); } |
1da177e4c
|
1424 1425 1426 1427 1428 1429 1430 1431 |
/* * 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. */ |
028867ac2
|
1432 |
static struct dm_snap_pending_exception * |
c66213921
|
1433 1434 |
__find_pending_exception(struct dm_snapshot *s, struct dm_snap_pending_exception *pe, chunk_t chunk) |
1da177e4c
|
1435 |
{ |
c66213921
|
1436 |
struct dm_snap_pending_exception *pe2; |
1da177e4c
|
1437 |
|
2913808eb
|
1438 1439 |
pe2 = __lookup_pending_exception(s, chunk); if (pe2) { |
76df1c651
|
1440 |
free_pending_exception(pe); |
2913808eb
|
1441 |
return pe2; |
1da177e4c
|
1442 |
} |
76df1c651
|
1443 1444 1445 |
pe->e.old_chunk = chunk; bio_list_init(&pe->origin_bios); bio_list_init(&pe->snapshot_bios); |
76df1c651
|
1446 |
pe->started = 0; |
a6e50b409
|
1447 |
pe->full_bio = NULL; |
76df1c651
|
1448 |
|
493df71c6
|
1449 |
if (s->store->type->prepare_exception(s->store, &pe->e)) { |
76df1c651
|
1450 1451 1452 |
free_pending_exception(pe); return NULL; } |
3510cb94f
|
1453 |
dm_insert_exception(&s->pending, &pe->e); |
76df1c651
|
1454 |
|
1da177e4c
|
1455 1456 |
return pe; } |
1d4989c85
|
1457 |
static void remap_exception(struct dm_snapshot *s, struct dm_exception *e, |
d74f81f8a
|
1458 |
struct bio *bio, chunk_t chunk) |
1da177e4c
|
1459 |
{ |
fc56f6fbc
|
1460 |
bio->bi_bdev = s->cow->bdev; |
71fab00a6
|
1461 1462 1463 1464 1465 |
bio->bi_sector = chunk_to_sector(s->store, dm_chunk_number(e->new_chunk) + (chunk - e->old_chunk)) + (bio->bi_sector & s->store->chunk_mask); |
1da177e4c
|
1466 1467 1468 1469 1470 |
} static int snapshot_map(struct dm_target *ti, struct bio *bio, union map_info *map_context) { |
1d4989c85
|
1471 |
struct dm_exception *e; |
028867ac2
|
1472 |
struct dm_snapshot *s = ti->private; |
d2a7ad29a
|
1473 |
int r = DM_MAPIO_REMAPPED; |
1da177e4c
|
1474 |
chunk_t chunk; |
028867ac2
|
1475 |
struct dm_snap_pending_exception *pe = NULL; |
1da177e4c
|
1476 |
|
d87f4c14f
|
1477 |
if (bio->bi_rw & REQ_FLUSH) { |
fc56f6fbc
|
1478 |
bio->bi_bdev = s->cow->bdev; |
494b3ee7d
|
1479 1480 |
return DM_MAPIO_REMAPPED; } |
71fab00a6
|
1481 |
chunk = sector_to_chunk(s->store, bio->bi_sector); |
1da177e4c
|
1482 1483 |
/* Full snapshots are not usable */ |
76df1c651
|
1484 |
/* To get here the table must be live so s->active is always set. */ |
1da177e4c
|
1485 |
if (!s->valid) |
f6a80ea8e
|
1486 |
return -EIO; |
1da177e4c
|
1487 |
|
ba40a2aa6
|
1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 |
/* 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 */ |
3510cb94f
|
1498 |
e = dm_lookup_exception(&s->complete, chunk); |
ba40a2aa6
|
1499 |
if (e) { |
d74f81f8a
|
1500 |
remap_exception(s, e, bio, chunk); |
ba40a2aa6
|
1501 1502 |
goto out_unlock; } |
1da177e4c
|
1503 1504 1505 1506 1507 1508 |
/* * 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) { |
2913808eb
|
1509 |
pe = __lookup_pending_exception(s, chunk); |
76df1c651
|
1510 |
if (!pe) { |
c66213921
|
1511 1512 1513 1514 1515 1516 1517 1518 1519 |
up_write(&s->lock); pe = alloc_pending_exception(s); down_write(&s->lock); if (!s->valid) { free_pending_exception(pe); r = -EIO; goto out_unlock; } |
3510cb94f
|
1520 |
e = dm_lookup_exception(&s->complete, chunk); |
35bf659b0
|
1521 1522 1523 1524 1525 |
if (e) { free_pending_exception(pe); remap_exception(s, e, bio, chunk); goto out_unlock; } |
c66213921
|
1526 |
pe = __find_pending_exception(s, pe, chunk); |
2913808eb
|
1527 1528 1529 1530 1531 |
if (!pe) { __invalidate_snapshot(s, -ENOMEM); r = -EIO; goto out_unlock; } |
1da177e4c
|
1532 |
} |
d74f81f8a
|
1533 |
remap_exception(s, &pe->e, bio, chunk); |
76df1c651
|
1534 |
|
d2a7ad29a
|
1535 |
r = DM_MAPIO_SUBMITTED; |
ba40a2aa6
|
1536 |
|
a6e50b409
|
1537 1538 1539 1540 1541 1542 1543 1544 1545 |
if (!pe->started && bio->bi_size == (s->store->chunk_size << SECTOR_SHIFT)) { pe->started = 1; up_write(&s->lock); start_full_bio(pe, bio); goto out; } bio_list_add(&pe->snapshot_bios, bio); |
76df1c651
|
1546 1547 1548 |
if (!pe->started) { /* this is protected by snap->lock */ pe->started = 1; |
ba40a2aa6
|
1549 |
up_write(&s->lock); |
76df1c651
|
1550 |
start_copy(pe); |
ba40a2aa6
|
1551 1552 |
goto out; } |
cd45daffd
|
1553 |
} else { |
ba40a2aa6
|
1554 |
bio->bi_bdev = s->origin->bdev; |
cd45daffd
|
1555 1556 |
map_context->ptr = track_chunk(s, chunk); } |
1da177e4c
|
1557 |
|
a2d2b0345
|
1558 |
out_unlock: |
ba40a2aa6
|
1559 |
up_write(&s->lock); |
a2d2b0345
|
1560 |
out: |
1da177e4c
|
1561 1562 |
return r; } |
3452c2a1e
|
1563 1564 1565 1566 1567 1568 1569 1570 1571 |
/* * A snapshot-merge target behaves like a combination of a snapshot * target and a snapshot-origin target. It only generates new * exceptions in other snapshots and not in the one that is being * merged. * * For each chunk, if there is an existing exception, it is used to * redirect I/O to the cow device. Otherwise I/O is sent to the origin, * which in turn might generate exceptions in other snapshots. |
9fe862548
|
1572 1573 |
* If merging is currently taking place on the chunk in question, the * I/O is deferred by adding it to s->bios_queued_during_merge. |
3452c2a1e
|
1574 1575 1576 1577 1578 1579 1580 1581 |
*/ static int snapshot_merge_map(struct dm_target *ti, struct bio *bio, union map_info *map_context) { struct dm_exception *e; struct dm_snapshot *s = ti->private; int r = DM_MAPIO_REMAPPED; chunk_t chunk; |
d87f4c14f
|
1582 |
if (bio->bi_rw & REQ_FLUSH) { |
57cba5d36
|
1583 |
if (!map_context->target_request_nr) |
10b8106a7
|
1584 1585 1586 1587 1588 1589 |
bio->bi_bdev = s->origin->bdev; else bio->bi_bdev = s->cow->bdev; map_context->ptr = NULL; return DM_MAPIO_REMAPPED; } |
3452c2a1e
|
1590 |
chunk = sector_to_chunk(s->store, bio->bi_sector); |
9fe862548
|
1591 |
down_write(&s->lock); |
3452c2a1e
|
1592 |
|
d2fdb776e
|
1593 1594 1595 |
/* Full merging snapshots are redirected to the origin */ if (!s->valid) goto redirect_to_origin; |
3452c2a1e
|
1596 1597 1598 1599 |
/* If the block is already remapped - use that */ e = dm_lookup_exception(&s->complete, chunk); if (e) { |
9fe862548
|
1600 1601 1602 1603 1604 1605 1606 1607 1608 1609 |
/* Queue writes overlapping with chunks being merged */ if (bio_rw(bio) == WRITE && chunk >= s->first_merging_chunk && chunk < (s->first_merging_chunk + s->num_merging_chunks)) { bio->bi_bdev = s->origin->bdev; bio_list_add(&s->bios_queued_during_merge, bio); r = DM_MAPIO_SUBMITTED; goto out_unlock; } |
17aa03326
|
1610 |
|
3452c2a1e
|
1611 |
remap_exception(s, e, bio, chunk); |
17aa03326
|
1612 1613 1614 |
if (bio_rw(bio) == WRITE) map_context->ptr = track_chunk(s, chunk); |
3452c2a1e
|
1615 1616 |
goto out_unlock; } |
d2fdb776e
|
1617 |
redirect_to_origin: |
3452c2a1e
|
1618 1619 1620 |
bio->bi_bdev = s->origin->bdev; if (bio_rw(bio) == WRITE) { |
9fe862548
|
1621 |
up_write(&s->lock); |
3452c2a1e
|
1622 1623 1624 1625 |
return do_origin(s->origin, bio); } out_unlock: |
9fe862548
|
1626 |
up_write(&s->lock); |
3452c2a1e
|
1627 1628 1629 |
return r; } |
cd45daffd
|
1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 |
static int snapshot_end_io(struct dm_target *ti, struct bio *bio, int error, union map_info *map_context) { struct dm_snapshot *s = ti->private; struct dm_snap_tracked_chunk *c = map_context->ptr; if (c) stop_tracking_chunk(s, c); return 0; } |
1e03f97e4
|
1641 1642 1643 1644 1645 1646 |
static void snapshot_merge_presuspend(struct dm_target *ti) { struct dm_snapshot *s = ti->private; stop_merge(s); } |
c1f0c183f
|
1647 1648 1649 1650 1651 1652 1653 |
static int snapshot_preresume(struct dm_target *ti) { int r = 0; struct dm_snapshot *s = ti->private; struct dm_snapshot *snap_src = NULL, *snap_dest = NULL; down_read(&_origins_lock); |
9d3b15c4c
|
1654 |
(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); |
c1f0c183f
|
1655 1656 1657 1658 1659 1660 |
if (snap_src && snap_dest) { down_read(&snap_src->lock); if (s == snap_src) { DMERR("Unable to resume snapshot source until " "handover completes."); r = -EINVAL; |
b83b2f295
|
1661 |
} else if (!dm_suspended(snap_src->ti)) { |
c1f0c183f
|
1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 |
DMERR("Unable to perform snapshot handover until " "source is suspended."); r = -EINVAL; } up_read(&snap_src->lock); } up_read(&_origins_lock); return r; } |
1da177e4c
|
1672 1673 |
static void snapshot_resume(struct dm_target *ti) { |
028867ac2
|
1674 |
struct dm_snapshot *s = ti->private; |
c1f0c183f
|
1675 1676 1677 |
struct dm_snapshot *snap_src = NULL, *snap_dest = NULL; down_read(&_origins_lock); |
9d3b15c4c
|
1678 |
(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL); |
c1f0c183f
|
1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 |
if (snap_src && snap_dest) { down_write(&snap_src->lock); down_write_nested(&snap_dest->lock, SINGLE_DEPTH_NESTING); __handover_exceptions(snap_src, snap_dest); up_write(&snap_dest->lock); up_write(&snap_src->lock); } up_read(&_origins_lock); /* Now we have correct chunk size, reregister */ reregister_snapshot(s); |
1da177e4c
|
1690 |
|
aa14edeb9
|
1691 1692 1693 |
down_write(&s->lock); s->active = 1; up_write(&s->lock); |
1da177e4c
|
1694 |
} |
1e03f97e4
|
1695 1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716 1717 1718 1719 1720 1721 |
static sector_t get_origin_minimum_chunksize(struct block_device *bdev) { sector_t min_chunksize; down_read(&_origins_lock); min_chunksize = __minimum_chunk_size(__lookup_origin(bdev)); up_read(&_origins_lock); return min_chunksize; } static void snapshot_merge_resume(struct dm_target *ti) { struct dm_snapshot *s = ti->private; /* * Handover exceptions from existing snapshot. */ snapshot_resume(ti); /* * snapshot-merge acts as an origin, so set ti->split_io */ ti->split_io = get_origin_minimum_chunksize(s->origin->bdev); start_merge(s); } |
1da177e4c
|
1722 1723 1724 |
static int snapshot_status(struct dm_target *ti, status_type_t type, char *result, unsigned int maxlen) { |
2e4a31df2
|
1725 |
unsigned sz = 0; |
028867ac2
|
1726 |
struct dm_snapshot *snap = ti->private; |
1da177e4c
|
1727 1728 1729 |
switch (type) { case STATUSTYPE_INFO: |
94e76572b
|
1730 1731 |
down_write(&snap->lock); |
1da177e4c
|
1732 |
if (!snap->valid) |
2e4a31df2
|
1733 |
DMEMIT("Invalid"); |
d8ddb1cff
|
1734 1735 |
else if (snap->merge_failed) DMEMIT("Merge failed"); |
1da177e4c
|
1736 |
else { |
985903bb3
|
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747 |
if (snap->store->type->usage) { sector_t total_sectors, sectors_allocated, metadata_sectors; snap->store->type->usage(snap->store, &total_sectors, §ors_allocated, &metadata_sectors); DMEMIT("%llu/%llu %llu", (unsigned long long)sectors_allocated, (unsigned long long)total_sectors, (unsigned long long)metadata_sectors); |
1da177e4c
|
1748 1749 |
} else |
2e4a31df2
|
1750 |
DMEMIT("Unknown"); |
1da177e4c
|
1751 |
} |
94e76572b
|
1752 1753 |
up_write(&snap->lock); |
1da177e4c
|
1754 1755 1756 1757 1758 1759 1760 1761 |
break; case STATUSTYPE_TABLE: /* * kdevname returns a static pointer so we need * to make private copies if the output is to * make sense. */ |
fc56f6fbc
|
1762 |
DMEMIT("%s %s", snap->origin->name, snap->cow->name); |
1e302a929
|
1763 1764 |
snap->store->type->status(snap->store, type, result + sz, maxlen - sz); |
1da177e4c
|
1765 1766 1767 1768 1769 |
break; } return 0; } |
8811f46c1
|
1770 1771 1772 1773 |
static int snapshot_iterate_devices(struct dm_target *ti, iterate_devices_callout_fn fn, void *data) { struct dm_snapshot *snap = ti->private; |
1e5554c84
|
1774 1775 1776 |
int r; r = fn(ti, snap->origin, 0, ti->len, data); |
8811f46c1
|
1777 |
|
1e5554c84
|
1778 1779 1780 1781 |
if (!r) r = fn(ti, snap->cow, 0, get_dev_size(snap->cow->bdev), data); return r; |
8811f46c1
|
1782 |
} |
1da177e4c
|
1783 1784 1785 |
/*----------------------------------------------------------------- * Origin methods *---------------------------------------------------------------*/ |
9eaae8ffb
|
1786 1787 1788 1789 1790 1791 1792 1793 1794 1795 1796 1797 1798 |
/* * If no exceptions need creating, DM_MAPIO_REMAPPED is returned and any * supplied bio was ignored. The caller may submit it immediately. * (No remapping actually occurs as the origin is always a direct linear * map.) * * If further exceptions are required, DM_MAPIO_SUBMITTED is returned * and any supplied bio is added to a list to be submitted once all * the necessary exceptions exist. */ static int __origin_write(struct list_head *snapshots, sector_t sector, struct bio *bio) |
1da177e4c
|
1799 |
{ |
515ad66cc
|
1800 |
int r = DM_MAPIO_REMAPPED; |
1da177e4c
|
1801 |
struct dm_snapshot *snap; |
1d4989c85
|
1802 |
struct dm_exception *e; |
515ad66cc
|
1803 1804 1805 |
struct dm_snap_pending_exception *pe; struct dm_snap_pending_exception *pe_to_start_now = NULL; struct dm_snap_pending_exception *pe_to_start_last = NULL; |
1da177e4c
|
1806 1807 1808 1809 |
chunk_t chunk; /* Do all the snapshots on this origin */ list_for_each_entry (snap, snapshots, list) { |
3452c2a1e
|
1810 1811 1812 1813 1814 1815 |
/* * Don't make new exceptions in a merging snapshot * because it has effectively been deleted */ if (dm_target_is_snapshot_merge(snap->ti)) continue; |
76df1c651
|
1816 |
down_write(&snap->lock); |
aa14edeb9
|
1817 1818 |
/* Only deal with valid and active snapshots */ if (!snap->valid || !snap->active) |
76df1c651
|
1819 |
goto next_snapshot; |
1da177e4c
|
1820 |
|
d5e404c10
|
1821 |
/* Nothing to do if writing beyond end of snapshot */ |
9eaae8ffb
|
1822 |
if (sector >= dm_table_get_size(snap->ti->table)) |
76df1c651
|
1823 |
goto next_snapshot; |
1da177e4c
|
1824 1825 1826 1827 1828 |
/* * Remember, different snapshots can have * different chunk sizes. */ |
9eaae8ffb
|
1829 |
chunk = sector_to_chunk(snap->store, sector); |
1da177e4c
|
1830 1831 1832 1833 1834 1835 |
/* * Check exception table to see if block * is already remapped in this snapshot * and trigger an exception if not. */ |
3510cb94f
|
1836 |
e = dm_lookup_exception(&snap->complete, chunk); |
76df1c651
|
1837 1838 |
if (e) goto next_snapshot; |
2913808eb
|
1839 |
pe = __lookup_pending_exception(snap, chunk); |
76df1c651
|
1840 |
if (!pe) { |
c66213921
|
1841 1842 1843 1844 1845 1846 1847 1848 |
up_write(&snap->lock); pe = alloc_pending_exception(snap); down_write(&snap->lock); if (!snap->valid) { free_pending_exception(pe); goto next_snapshot; } |
3510cb94f
|
1849 |
e = dm_lookup_exception(&snap->complete, chunk); |
35bf659b0
|
1850 1851 1852 1853 |
if (e) { free_pending_exception(pe); goto next_snapshot; } |
c66213921
|
1854 |
pe = __find_pending_exception(snap, pe, chunk); |
2913808eb
|
1855 1856 1857 1858 |
if (!pe) { __invalidate_snapshot(snap, -ENOMEM); goto next_snapshot; } |
76df1c651
|
1859 |
} |
515ad66cc
|
1860 |
r = DM_MAPIO_SUBMITTED; |
76df1c651
|
1861 |
|
515ad66cc
|
1862 1863 1864 1865 1866 1867 1868 1869 |
/* * If an origin bio was supplied, queue it to wait for the * completion of this exception, and start this one last, * at the end of the function. */ if (bio) { bio_list_add(&pe->origin_bios, bio); bio = NULL; |
76df1c651
|
1870 |
|
515ad66cc
|
1871 1872 1873 1874 |
if (!pe->started) { pe->started = 1; pe_to_start_last = pe; } |
76df1c651
|
1875 1876 1877 1878 |
} if (!pe->started) { pe->started = 1; |
515ad66cc
|
1879 |
pe_to_start_now = pe; |
1da177e4c
|
1880 |
} |
a2d2b0345
|
1881 |
next_snapshot: |
1da177e4c
|
1882 |
up_write(&snap->lock); |
1da177e4c
|
1883 |
|
515ad66cc
|
1884 1885 1886 1887 |
if (pe_to_start_now) { start_copy(pe_to_start_now); pe_to_start_now = NULL; } |
b4b610f68
|
1888 |
} |
1da177e4c
|
1889 |
/* |
515ad66cc
|
1890 1891 |
* Submit the exception against which the bio is queued last, * to give the other exceptions a head start. |
1da177e4c
|
1892 |
*/ |
515ad66cc
|
1893 1894 |
if (pe_to_start_last) start_copy(pe_to_start_last); |
1da177e4c
|
1895 1896 1897 1898 1899 1900 1901 1902 1903 1904 |
return r; } /* * Called on a write from the origin driver. */ static int do_origin(struct dm_dev *origin, struct bio *bio) { struct origin *o; |
d2a7ad29a
|
1905 |
int r = DM_MAPIO_REMAPPED; |
1da177e4c
|
1906 1907 1908 1909 |
down_read(&_origins_lock); o = __lookup_origin(origin->bdev); if (o) |
9eaae8ffb
|
1910 |
r = __origin_write(&o->snapshots, bio->bi_sector, bio); |
1da177e4c
|
1911 1912 1913 1914 1915 1916 |
up_read(&_origins_lock); return r; } /* |
73dfd078c
|
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 |
* Trigger exceptions in all non-merging snapshots. * * The chunk size of the merging snapshot may be larger than the chunk * size of some other snapshot so we may need to reallocate multiple * chunks in other snapshots. * * We scan all the overlapping exceptions in the other snapshots. * Returns 1 if anything was reallocated and must be waited for, * otherwise returns 0. * * size must be a multiple of merging_snap's chunk_size. */ static int origin_write_extent(struct dm_snapshot *merging_snap, sector_t sector, unsigned size) { int must_wait = 0; sector_t n; struct origin *o; /* * The origin's __minimum_chunk_size() got stored in split_io * by snapshot_merge_resume(). */ down_read(&_origins_lock); o = __lookup_origin(merging_snap->origin->bdev); for (n = 0; n < size; n += merging_snap->ti->split_io) if (__origin_write(&o->snapshots, sector + n, NULL) == DM_MAPIO_SUBMITTED) must_wait = 1; up_read(&_origins_lock); return must_wait; } /* |
1da177e4c
|
1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 |
* 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) { |
72d948616
|
1966 |
ti->error = "origin: incorrect number of arguments"; |
1da177e4c
|
1967 1968 |
return -EINVAL; } |
8215d6ec5
|
1969 |
r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &dev); |
1da177e4c
|
1970 1971 1972 1973 1974 1975 |
if (r) { ti->error = "Cannot get target device"; return r; } ti->private = dev; |
494b3ee7d
|
1976 |
ti->num_flush_requests = 1; |
1da177e4c
|
1977 1978 1979 1980 1981 |
return 0; } static void origin_dtr(struct dm_target *ti) { |
028867ac2
|
1982 |
struct dm_dev *dev = ti->private; |
1da177e4c
|
1983 1984 1985 1986 1987 1988 |
dm_put_device(ti, dev); } static int origin_map(struct dm_target *ti, struct bio *bio, union map_info *map_context) { |
028867ac2
|
1989 |
struct dm_dev *dev = ti->private; |
1da177e4c
|
1990 |
bio->bi_bdev = dev->bdev; |
d87f4c14f
|
1991 |
if (bio->bi_rw & REQ_FLUSH) |
494b3ee7d
|
1992 |
return DM_MAPIO_REMAPPED; |
1da177e4c
|
1993 |
/* Only tell snapshots if this is a write */ |
d2a7ad29a
|
1994 |
return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED; |
1da177e4c
|
1995 |
} |
1da177e4c
|
1996 1997 1998 1999 2000 2001 |
/* * Set the target "split_io" field to the minimum of all the snapshots' * chunk sizes. */ static void origin_resume(struct dm_target *ti) { |
028867ac2
|
2002 |
struct dm_dev *dev = ti->private; |
1da177e4c
|
2003 |
|
1e03f97e4
|
2004 |
ti->split_io = get_origin_minimum_chunksize(dev->bdev); |
1da177e4c
|
2005 2006 2007 2008 2009 |
} static int origin_status(struct dm_target *ti, status_type_t type, char *result, unsigned int maxlen) { |
028867ac2
|
2010 |
struct dm_dev *dev = ti->private; |
1da177e4c
|
2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 |
switch (type) { case STATUSTYPE_INFO: result[0] = '\0'; break; case STATUSTYPE_TABLE: snprintf(result, maxlen, "%s", dev->name); break; } return 0; } |
b1d555283
|
2024 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 2037 |
static int origin_merge(struct dm_target *ti, struct bvec_merge_data *bvm, struct bio_vec *biovec, int max_size) { struct dm_dev *dev = ti->private; struct request_queue *q = bdev_get_queue(dev->bdev); if (!q->merge_bvec_fn) return max_size; bvm->bi_bdev = dev->bdev; bvm->bi_sector = bvm->bi_sector; return min(max_size, q->merge_bvec_fn(q, bvm, biovec)); } |
8811f46c1
|
2038 2039 2040 2041 2042 2043 2044 |
static int origin_iterate_devices(struct dm_target *ti, iterate_devices_callout_fn fn, void *data) { struct dm_dev *dev = ti->private; return fn(ti, dev, 0, ti->len, data); } |
1da177e4c
|
2045 2046 |
static struct target_type origin_target = { .name = "snapshot-origin", |
b83b2f295
|
2047 |
.version = {1, 7, 1}, |
1da177e4c
|
2048 2049 2050 2051 2052 2053 |
.module = THIS_MODULE, .ctr = origin_ctr, .dtr = origin_dtr, .map = origin_map, .resume = origin_resume, .status = origin_status, |
b1d555283
|
2054 |
.merge = origin_merge, |
8811f46c1
|
2055 |
.iterate_devices = origin_iterate_devices, |
1da177e4c
|
2056 2057 2058 2059 |
}; static struct target_type snapshot_target = { .name = "snapshot", |
b83b2f295
|
2060 |
.version = {1, 10, 0}, |
1da177e4c
|
2061 2062 2063 2064 |
.module = THIS_MODULE, .ctr = snapshot_ctr, .dtr = snapshot_dtr, .map = snapshot_map, |
cd45daffd
|
2065 |
.end_io = snapshot_end_io, |
c1f0c183f
|
2066 |
.preresume = snapshot_preresume, |
1da177e4c
|
2067 2068 |
.resume = snapshot_resume, .status = snapshot_status, |
8811f46c1
|
2069 |
.iterate_devices = snapshot_iterate_devices, |
1da177e4c
|
2070 |
}; |
d698aa450
|
2071 2072 |
static struct target_type merge_target = { .name = dm_snapshot_merge_target_name, |
b83b2f295
|
2073 |
.version = {1, 1, 0}, |
d698aa450
|
2074 2075 2076 |
.module = THIS_MODULE, .ctr = snapshot_ctr, .dtr = snapshot_dtr, |
3452c2a1e
|
2077 |
.map = snapshot_merge_map, |
d698aa450
|
2078 |
.end_io = snapshot_end_io, |
1e03f97e4
|
2079 |
.presuspend = snapshot_merge_presuspend, |
d698aa450
|
2080 |
.preresume = snapshot_preresume, |
1e03f97e4
|
2081 |
.resume = snapshot_merge_resume, |
d698aa450
|
2082 2083 2084 |
.status = snapshot_status, .iterate_devices = snapshot_iterate_devices, }; |
1da177e4c
|
2085 2086 2087 |
static int __init dm_snapshot_init(void) { int r; |
4db6bfe02
|
2088 2089 2090 2091 2092 |
r = dm_exception_store_init(); if (r) { DMERR("Failed to initialize exception stores"); return r; } |
1da177e4c
|
2093 |
r = dm_register_target(&snapshot_target); |
d698aa450
|
2094 |
if (r < 0) { |
1da177e4c
|
2095 |
DMERR("snapshot target register failed %d", r); |
034a186d2
|
2096 |
goto bad_register_snapshot_target; |
1da177e4c
|
2097 2098 2099 2100 |
} r = dm_register_target(&origin_target); if (r < 0) { |
72d948616
|
2101 |
DMERR("Origin target register failed %d", r); |
d698aa450
|
2102 2103 2104 2105 2106 2107 2108 |
goto bad_register_origin_target; } r = dm_register_target(&merge_target); if (r < 0) { DMERR("Merge target register failed %d", r); goto bad_register_merge_target; |
1da177e4c
|
2109 2110 2111 2112 2113 |
} r = init_origin_hash(); if (r) { DMERR("init_origin_hash failed."); |
d698aa450
|
2114 |
goto bad_origin_hash; |
1da177e4c
|
2115 |
} |
1d4989c85
|
2116 |
exception_cache = KMEM_CACHE(dm_exception, 0); |
1da177e4c
|
2117 2118 2119 |
if (!exception_cache) { DMERR("Couldn't create exception cache."); r = -ENOMEM; |
d698aa450
|
2120 |
goto bad_exception_cache; |
1da177e4c
|
2121 |
} |
028867ac2
|
2122 |
pending_cache = KMEM_CACHE(dm_snap_pending_exception, 0); |
1da177e4c
|
2123 2124 2125 |
if (!pending_cache) { DMERR("Couldn't create pending cache."); r = -ENOMEM; |
d698aa450
|
2126 |
goto bad_pending_cache; |
1da177e4c
|
2127 |
} |
cd45daffd
|
2128 2129 2130 2131 |
tracked_chunk_cache = KMEM_CACHE(dm_snap_tracked_chunk, 0); if (!tracked_chunk_cache) { DMERR("Couldn't create cache to track chunks in use."); r = -ENOMEM; |
d698aa450
|
2132 |
goto bad_tracked_chunk_cache; |
cd45daffd
|
2133 |
} |
1da177e4c
|
2134 |
return 0; |
d698aa450
|
2135 |
bad_tracked_chunk_cache: |
1da177e4c
|
2136 |
kmem_cache_destroy(pending_cache); |
d698aa450
|
2137 |
bad_pending_cache: |
1da177e4c
|
2138 |
kmem_cache_destroy(exception_cache); |
d698aa450
|
2139 |
bad_exception_cache: |
1da177e4c
|
2140 |
exit_origin_hash(); |
d698aa450
|
2141 2142 2143 |
bad_origin_hash: dm_unregister_target(&merge_target); bad_register_merge_target: |
1da177e4c
|
2144 |
dm_unregister_target(&origin_target); |
d698aa450
|
2145 |
bad_register_origin_target: |
1da177e4c
|
2146 |
dm_unregister_target(&snapshot_target); |
034a186d2
|
2147 2148 |
bad_register_snapshot_target: dm_exception_store_exit(); |
d698aa450
|
2149 |
|
1da177e4c
|
2150 2151 2152 2153 2154 |
return r; } static void __exit dm_snapshot_exit(void) { |
10d3bd09a
|
2155 2156 |
dm_unregister_target(&snapshot_target); dm_unregister_target(&origin_target); |
d698aa450
|
2157 |
dm_unregister_target(&merge_target); |
1da177e4c
|
2158 2159 |
exit_origin_hash(); |
1da177e4c
|
2160 2161 |
kmem_cache_destroy(pending_cache); kmem_cache_destroy(exception_cache); |
cd45daffd
|
2162 |
kmem_cache_destroy(tracked_chunk_cache); |
4db6bfe02
|
2163 2164 |
dm_exception_store_exit(); |
1da177e4c
|
2165 2166 2167 2168 2169 2170 2171 2172 2173 |
} /* 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"); |