dm-thin.c 68 KB
Newer Older
1
/*
2
 * Copyright (C) 2011-2012 Red Hat UK.
3
4
5
6
7
 *
 * This file is released under the GPL.
 */

#include "dm-thin-metadata.h"
8
#include "dm-bio-prison.h"
9
#include "dm.h"
10
11
12
13
14
15
16
17
18
19
20
21
22
23

#include <linux/device-mapper.h>
#include <linux/dm-io.h>
#include <linux/dm-kcopyd.h>
#include <linux/list.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/slab.h>

#define	DM_MSG_PREFIX	"thin"

/*
 * Tunable constants
 */
24
#define ENDIO_HOOK_POOL_SIZE 1024
25
26
#define MAPPING_POOL_SIZE 1024
#define PRISON_CELLS 1024
27
#define COMMIT_PERIOD HZ
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60

/*
 * The block size of the device holding pool data must be
 * between 64KB and 1GB.
 */
#define DATA_DEV_BLOCK_SIZE_MIN_SECTORS (64 * 1024 >> SECTOR_SHIFT)
#define DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT)

/*
 * Device id is restricted to 24 bits.
 */
#define MAX_DEV_ID ((1 << 24) - 1)

/*
 * How do we handle breaking sharing of data blocks?
 * =================================================
 *
 * We use a standard copy-on-write btree to store the mappings for the
 * devices (note I'm talking about copy-on-write of the metadata here, not
 * the data).  When you take an internal snapshot you clone the root node
 * of the origin btree.  After this there is no concept of an origin or a
 * snapshot.  They are just two device trees that happen to point to the
 * same data blocks.
 *
 * When we get a write in we decide if it's to a shared data block using
 * some timestamp magic.  If it is, we have to break sharing.
 *
 * Let's say we write to a shared block in what was the origin.  The
 * steps are:
 *
 * i) plug io further to this physical block. (see bio_prison code).
 *
 * ii) quiesce any read io to that shared data block.  Obviously
61
 * including all devices that share this block.  (see dm_deferred_set code)
62
63
64
65
66
 *
 * iii) copy the data block to a newly allocate block.  This step can be
 * missed out if the io covers the block. (schedule_copy).
 *
 * iv) insert the new mapping into the origin's btree
Joe Thornber's avatar
Joe Thornber committed
67
 * (process_prepared_mapping).  This act of inserting breaks some
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
 * sharing of btree nodes between the two devices.  Breaking sharing only
 * effects the btree of that specific device.  Btrees for the other
 * devices that share the block never change.  The btree for the origin
 * device as it was after the last commit is untouched, ie. we're using
 * persistent data structures in the functional programming sense.
 *
 * v) unplug io to this physical block, including the io that triggered
 * the breaking of sharing.
 *
 * Steps (ii) and (iii) occur in parallel.
 *
 * The metadata _doesn't_ need to be committed before the io continues.  We
 * get away with this because the io is always written to a _new_ block.
 * If there's a crash, then:
 *
 * - The origin mapping will point to the old origin block (the shared
 * one).  This will contain the data as it was before the io that triggered
 * the breaking of sharing came in.
 *
 * - The snap mapping still points to the old block.  As it would after
 * the commit.
 *
 * The downside of this scheme is the timestamp magic isn't perfect, and
 * will continue to think that data block in the snapshot device is shared
 * even after the write to the origin has broken sharing.  I suspect data
 * blocks will typically be shared by many different devices, so we're
 * breaking sharing n + 1 times, rather than n, where n is the number of
 * devices that reference this data block.  At the moment I think the
 * benefits far, far outweigh the disadvantages.
 */

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

/*
 * Key building.
 */
static void build_data_key(struct dm_thin_device *td,
105
			   dm_block_t b, struct dm_cell_key *key)
106
107
108
109
110
111
112
{
	key->virtual = 0;
	key->dev = dm_thin_dev_id(td);
	key->block = b;
}

static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,
113
			      struct dm_cell_key *key)
114
115
116
117
118
119
120
121
122
123
124
125
126
{
	key->virtual = 1;
	key->dev = dm_thin_dev_id(td);
	key->block = b;
}

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

/*
 * A pool device ties together a metadata device and a data device.  It
 * also provides the interface for creating and destroying internal
 * devices.
 */
Mike Snitzer's avatar
Mike Snitzer committed
127
struct dm_thin_new_mapping;
128

129
130
131
132
133
134
135
136
137
/*
 * The pool runs in 3 modes.  Ordered in degraded order for comparisons.
 */
enum pool_mode {
	PM_WRITE,		/* metadata may be changed */
	PM_READ_ONLY,		/* metadata may not be changed */
	PM_FAIL,		/* all I/O fails */
};

138
struct pool_features {
139
140
	enum pool_mode mode;

Mike Snitzer's avatar
Mike Snitzer committed
141
142
143
	bool zero_new_blocks:1;
	bool discard_enabled:1;
	bool discard_passdown:1;
144
145
};

146
147
148
149
struct thin_c;
typedef void (*process_bio_fn)(struct thin_c *tc, struct bio *bio);
typedef void (*process_mapping_fn)(struct dm_thin_new_mapping *m);

150
151
152
153
154
155
156
157
158
struct pool {
	struct list_head list;
	struct dm_target *ti;	/* Only set if a pool target is bound */

	struct mapped_device *pool_md;
	struct block_device *md_dev;
	struct dm_pool_metadata *pmd;

	dm_block_t low_water_blocks;
159
	uint32_t sectors_per_block;
160
	int sectors_per_block_shift;
161

162
	struct pool_features pf;
163
164
165
	unsigned low_water_triggered:1;	/* A dm event has been sent */
	unsigned no_free_space:1;	/* A -ENOSPC warning has been issued */

166
	struct dm_bio_prison *prison;
167
168
169
170
	struct dm_kcopyd_client *copier;

	struct workqueue_struct *wq;
	struct work_struct worker;
171
	struct delayed_work waker;
172

173
	unsigned long last_commit_jiffies;
174
	unsigned ref_count;
175
176
177
178
179

	spinlock_t lock;
	struct bio_list deferred_bios;
	struct bio_list deferred_flush_bios;
	struct list_head prepared_mappings;
Joe Thornber's avatar
Joe Thornber committed
180
	struct list_head prepared_discards;
181
182
183

	struct bio_list retry_on_resume_list;

184
185
	struct dm_deferred_set *shared_read_ds;
	struct dm_deferred_set *all_io_ds;
186

Mike Snitzer's avatar
Mike Snitzer committed
187
	struct dm_thin_new_mapping *next_mapping;
188
	mempool_t *mapping_pool;
189
190
191
192
193
194

	process_bio_fn process_bio;
	process_bio_fn process_discard;

	process_mapping_fn process_prepared_mapping;
	process_mapping_fn process_prepared_discard;
195
196
};

197
198
199
static enum pool_mode get_pool_mode(struct pool *pool);
static void set_pool_mode(struct pool *pool, enum pool_mode mode);

200
201
202
203
204
205
206
207
208
209
210
/*
 * Target context for a pool.
 */
struct pool_c {
	struct dm_target *ti;
	struct pool *pool;
	struct dm_dev *data_dev;
	struct dm_dev *metadata_dev;
	struct dm_target_callbacks callbacks;

	dm_block_t low_water_blocks;
211
212
	struct pool_features requested_pf; /* Features requested during table load */
	struct pool_features adjusted_pf;  /* Features used after adjusting for constituent devices */
213
214
215
216
217
218
219
};

/*
 * Target context for a thin.
 */
struct thin_c {
	struct dm_dev *pool_dev;
220
	struct dm_dev *origin_dev;
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
	dm_thin_id dev_id;

	struct pool *pool;
	struct dm_thin_device *td;
};

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

/*
 * A global list of pools that uses a struct mapped_device as a key.
 */
static struct dm_thin_pool_table {
	struct mutex mutex;
	struct list_head pools;
} dm_thin_pool_table;

static void pool_table_init(void)
{
	mutex_init(&dm_thin_pool_table.mutex);
	INIT_LIST_HEAD(&dm_thin_pool_table.pools);
}

static void __pool_table_insert(struct pool *pool)
{
	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
	list_add(&pool->list, &dm_thin_pool_table.pools);
}

static void __pool_table_remove(struct pool *pool)
{
	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
	list_del(&pool->list);
}

static struct pool *__pool_table_lookup(struct mapped_device *md)
{
	struct pool *pool = NULL, *tmp;

	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));

	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
		if (tmp->pool_md == md) {
			pool = tmp;
			break;
		}
	}

	return pool;
}

static struct pool *__pool_table_lookup_metadata_dev(struct block_device *md_dev)
{
	struct pool *pool = NULL, *tmp;

	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));

	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
		if (tmp->md_dev == md_dev) {
			pool = tmp;
			break;
		}
	}

	return pool;
}

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

Mike Snitzer's avatar
Mike Snitzer committed
289
struct dm_thin_endio_hook {
290
	struct thin_c *tc;
291
292
	struct dm_deferred_entry *shared_read_entry;
	struct dm_deferred_entry *all_io_entry;
Mike Snitzer's avatar
Mike Snitzer committed
293
	struct dm_thin_new_mapping *overwrite_mapping;
294
295
};

296
297
298
299
300
301
302
303
304
305
static void __requeue_bio_list(struct thin_c *tc, struct bio_list *master)
{
	struct bio *bio;
	struct bio_list bios;

	bio_list_init(&bios);
	bio_list_merge(&bios, master);
	bio_list_init(master);

	while ((bio = bio_list_pop(&bios))) {
306
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
307

308
		if (h->tc == tc)
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
			bio_endio(bio, DM_ENDIO_REQUEUE);
		else
			bio_list_add(master, bio);
	}
}

static void requeue_io(struct thin_c *tc)
{
	struct pool *pool = tc->pool;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
	__requeue_bio_list(tc, &pool->deferred_bios);
	__requeue_bio_list(tc, &pool->retry_on_resume_list);
	spin_unlock_irqrestore(&pool->lock, flags);
}

/*
 * This section of code contains the logic for processing a thin device's IO.
 * Much of the code depends on pool object resources (lists, workqueues, etc)
 * but most is exclusively called from the thin target rather than the thin-pool
 * target.
 */

333
334
335
336
337
static bool block_size_is_power_of_two(struct pool *pool)
{
	return pool->sectors_per_block_shift >= 0;
}

338
339
static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
{
340
	struct pool *pool = tc->pool;
341
342
	sector_t block_nr = bio->bi_sector;

343
344
	if (block_size_is_power_of_two(pool))
		block_nr >>= pool->sectors_per_block_shift;
345
	else
346
		(void) sector_div(block_nr, pool->sectors_per_block);
347
348

	return block_nr;
349
350
351
352
353
}

static void remap(struct thin_c *tc, struct bio *bio, dm_block_t block)
{
	struct pool *pool = tc->pool;
354
	sector_t bi_sector = bio->bi_sector;
355
356

	bio->bi_bdev = tc->pool_dev->bdev;
357
	if (block_size_is_power_of_two(pool))
358
359
		bio->bi_sector = (block << pool->sectors_per_block_shift) |
				(bi_sector & (pool->sectors_per_block - 1));
360
361
362
	else
		bio->bi_sector = (block * pool->sectors_per_block) +
				 sector_div(bi_sector, pool->sectors_per_block);
363
364
}

365
366
367
368
369
static void remap_to_origin(struct thin_c *tc, struct bio *bio)
{
	bio->bi_bdev = tc->origin_dev->bdev;
}

370
371
372
373
374
375
static int bio_triggers_commit(struct thin_c *tc, struct bio *bio)
{
	return (bio->bi_rw & (REQ_FLUSH | REQ_FUA)) &&
		dm_thin_changed_this_transaction(tc->td);
}

376
377
378
379
380
381
382
static void inc_all_io_entry(struct pool *pool, struct bio *bio)
{
	struct dm_thin_endio_hook *h;

	if (bio->bi_rw & REQ_DISCARD)
		return;

383
	h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
384
385
386
	h->all_io_entry = dm_deferred_entry_inc(pool->all_io_ds);
}

387
static void issue(struct thin_c *tc, struct bio *bio)
388
389
390
391
{
	struct pool *pool = tc->pool;
	unsigned long flags;

392
393
394
395
396
	if (!bio_triggers_commit(tc, bio)) {
		generic_make_request(bio);
		return;
	}

397
	/*
398
399
400
	 * Complete bio with an error if earlier I/O caused changes to
	 * the metadata that can't be committed e.g, due to I/O errors
	 * on the metadata device.
401
	 */
402
403
404
405
406
407
408
409
410
411
412
413
	if (dm_thin_aborted_changes(tc->td)) {
		bio_io_error(bio);
		return;
	}

	/*
	 * Batch together any bios that trigger commits and then issue a
	 * single commit for them in process_deferred_bios().
	 */
	spin_lock_irqsave(&pool->lock, flags);
	bio_list_add(&pool->deferred_flush_bios, bio);
	spin_unlock_irqrestore(&pool->lock, flags);
414
415
}

416
417
418
419
420
421
422
423
424
425
426
427
428
static void remap_to_origin_and_issue(struct thin_c *tc, struct bio *bio)
{
	remap_to_origin(tc, bio);
	issue(tc, bio);
}

static void remap_and_issue(struct thin_c *tc, struct bio *bio,
			    dm_block_t block)
{
	remap(tc, bio, block);
	issue(tc, bio);
}

429
430
431
432
433
434
435
436
437
438
439
440
441
442
/*
 * wake_worker() is used when new work is queued and when pool_resume is
 * ready to continue deferred IO processing.
 */
static void wake_worker(struct pool *pool)
{
	queue_work(pool->wq, &pool->worker);
}

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

/*
 * Bio endio functions.
 */
Mike Snitzer's avatar
Mike Snitzer committed
443
struct dm_thin_new_mapping {
444
445
	struct list_head list;

446
447
	unsigned quiesced:1;
	unsigned prepared:1;
Joe Thornber's avatar
Joe Thornber committed
448
	unsigned pass_discard:1;
449
450
451
452

	struct thin_c *tc;
	dm_block_t virt_block;
	dm_block_t data_block;
Mike Snitzer's avatar
Mike Snitzer committed
453
	struct dm_bio_prison_cell *cell, *cell2;
454
455
456
457
458
459
460
461
462
463
464
465
	int err;

	/*
	 * If the bio covers the whole area of a block then we can avoid
	 * zeroing or copying.  Instead this bio is hooked.  The bio will
	 * still be in the cell, so care has to be taken to avoid issuing
	 * the bio twice.
	 */
	struct bio *bio;
	bio_end_io_t *saved_bi_end_io;
};

Mike Snitzer's avatar
Mike Snitzer committed
466
static void __maybe_add_mapping(struct dm_thin_new_mapping *m)
467
468
469
{
	struct pool *pool = m->tc->pool;

470
	if (m->quiesced && m->prepared) {
471
472
473
474
475
476
477
478
		list_add(&m->list, &pool->prepared_mappings);
		wake_worker(pool);
	}
}

static void copy_complete(int read_err, unsigned long write_err, void *context)
{
	unsigned long flags;
Mike Snitzer's avatar
Mike Snitzer committed
479
	struct dm_thin_new_mapping *m = context;
480
481
482
483
484
485
486
487
488
489
490
491
492
	struct pool *pool = m->tc->pool;

	m->err = read_err || write_err ? -EIO : 0;

	spin_lock_irqsave(&pool->lock, flags);
	m->prepared = 1;
	__maybe_add_mapping(m);
	spin_unlock_irqrestore(&pool->lock, flags);
}

static void overwrite_endio(struct bio *bio, int err)
{
	unsigned long flags;
493
	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
494
	struct dm_thin_new_mapping *m = h->overwrite_mapping;
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
	struct pool *pool = m->tc->pool;

	m->err = err;

	spin_lock_irqsave(&pool->lock, flags);
	m->prepared = 1;
	__maybe_add_mapping(m);
	spin_unlock_irqrestore(&pool->lock, flags);
}

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

/*
 * Workqueue.
 */

/*
 * Prepared mapping jobs.
 */

/*
 * This sends the bios in the cell back to the deferred_bios list.
 */
Joe Thornber's avatar
Joe Thornber committed
518
static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *cell)
519
520
521
522
523
{
	struct pool *pool = tc->pool;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
524
	dm_cell_release(cell, &pool->deferred_bios);
525
526
527
528
529
530
	spin_unlock_irqrestore(&tc->pool->lock, flags);

	wake_worker(pool);
}

/*
531
 * Same as cell_defer except it omits the original holder of the cell.
532
 */
533
static void cell_defer_no_holder(struct thin_c *tc, struct dm_bio_prison_cell *cell)
534
535
536
537
538
{
	struct pool *pool = tc->pool;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
539
	dm_cell_release_no_holder(cell, &pool->deferred_bios);
540
541
542
543
544
	spin_unlock_irqrestore(&pool->lock, flags);

	wake_worker(pool);
}

545
546
547
548
static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
{
	if (m->bio)
		m->bio->bi_end_io = m->saved_bi_end_io;
549
	dm_cell_error(m->cell);
550
551
552
	list_del(&m->list);
	mempool_free(m, m->tc->pool->mapping_pool);
}
Mike Snitzer's avatar
Mike Snitzer committed
553
static void process_prepared_mapping(struct dm_thin_new_mapping *m)
554
555
556
557
558
559
560
561
562
563
{
	struct thin_c *tc = m->tc;
	struct bio *bio;
	int r;

	bio = m->bio;
	if (bio)
		bio->bi_end_io = m->saved_bi_end_io;

	if (m->err) {
564
		dm_cell_error(m->cell);
565
		goto out;
566
567
568
569
570
571
572
573
574
	}

	/*
	 * Commit the prepared block into the mapping btree.
	 * Any I/O for this block arriving after this point will get
	 * remapped to it directly.
	 */
	r = dm_thin_insert_block(tc->td, m->virt_block, m->data_block);
	if (r) {
575
		DMERR_LIMIT("dm_thin_insert_block() failed");
576
		dm_cell_error(m->cell);
577
		goto out;
578
579
580
581
582
583
584
585
586
	}

	/*
	 * Release any bios held while the block was being provisioned.
	 * If we are processing a write bio that completely covers the block,
	 * we already processed it so can ignore it now when processing
	 * the bios in the cell.
	 */
	if (bio) {
587
		cell_defer_no_holder(tc, m->cell);
588
589
		bio_endio(bio, 0);
	} else
Joe Thornber's avatar
Joe Thornber committed
590
		cell_defer(tc, m->cell);
591

592
out:
593
594
595
596
	list_del(&m->list);
	mempool_free(m, tc->pool->mapping_pool);
}

597
static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
Joe Thornber's avatar
Joe Thornber committed
598
599
600
{
	struct thin_c *tc = m->tc;

601
	bio_io_error(m->bio);
602
603
	cell_defer_no_holder(tc, m->cell);
	cell_defer_no_holder(tc, m->cell2);
604
605
606
607
608
609
	mempool_free(m, tc->pool->mapping_pool);
}

static void process_prepared_discard_passdown(struct dm_thin_new_mapping *m)
{
	struct thin_c *tc = m->tc;
Joe Thornber's avatar
Joe Thornber committed
610

611
	inc_all_io_entry(tc->pool, m->bio);
612
613
	cell_defer_no_holder(tc, m->cell);
	cell_defer_no_holder(tc, m->cell2);
614

Joe Thornber's avatar
Joe Thornber committed
615
616
617
618
619
620
621
622
	if (m->pass_discard)
		remap_and_issue(tc, m->bio, m->data_block);
	else
		bio_endio(m->bio, 0);

	mempool_free(m, tc->pool->mapping_pool);
}

623
624
625
626
627
628
629
static void process_prepared_discard(struct dm_thin_new_mapping *m)
{
	int r;
	struct thin_c *tc = m->tc;

	r = dm_thin_remove_block(tc->td, m->virt_block);
	if (r)
630
		DMERR_LIMIT("dm_thin_remove_block() failed");
631
632
633
634

	process_prepared_discard_passdown(m);
}

Joe Thornber's avatar
Joe Thornber committed
635
static void process_prepared(struct pool *pool, struct list_head *head,
636
			     process_mapping_fn *fn)
637
638
639
{
	unsigned long flags;
	struct list_head maps;
Mike Snitzer's avatar
Mike Snitzer committed
640
	struct dm_thin_new_mapping *m, *tmp;
641
642
643

	INIT_LIST_HEAD(&maps);
	spin_lock_irqsave(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
644
	list_splice_init(head, &maps);
645
646
647
	spin_unlock_irqrestore(&pool->lock, flags);

	list_for_each_entry_safe(m, tmp, &maps, list)
648
		(*fn)(m);
649
650
651
652
653
}

/*
 * Deferred bio jobs.
 */
Joe Thornber's avatar
Joe Thornber committed
654
static int io_overlaps_block(struct pool *pool, struct bio *bio)
655
{
656
	return bio->bi_size == (pool->sectors_per_block << SECTOR_SHIFT);
Joe Thornber's avatar
Joe Thornber committed
657
658
659
660
661
662
}

static int io_overwrites_block(struct pool *pool, struct bio *bio)
{
	return (bio_data_dir(bio) == WRITE) &&
		io_overlaps_block(pool, bio);
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
}

static void save_and_set_endio(struct bio *bio, bio_end_io_t **save,
			       bio_end_io_t *fn)
{
	*save = bio->bi_end_io;
	bio->bi_end_io = fn;
}

static int ensure_next_mapping(struct pool *pool)
{
	if (pool->next_mapping)
		return 0;

	pool->next_mapping = mempool_alloc(pool->mapping_pool, GFP_ATOMIC);

	return pool->next_mapping ? 0 : -ENOMEM;
}

Mike Snitzer's avatar
Mike Snitzer committed
682
static struct dm_thin_new_mapping *get_next_mapping(struct pool *pool)
683
{
Mike Snitzer's avatar
Mike Snitzer committed
684
	struct dm_thin_new_mapping *r = pool->next_mapping;
685
686
687
688
689
690
691
692
693

	BUG_ON(!pool->next_mapping);

	pool->next_mapping = NULL;

	return r;
}

static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
694
695
			  struct dm_dev *origin, dm_block_t data_origin,
			  dm_block_t data_dest,
Mike Snitzer's avatar
Mike Snitzer committed
696
			  struct dm_bio_prison_cell *cell, struct bio *bio)
697
698
699
{
	int r;
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
700
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
701
702

	INIT_LIST_HEAD(&m->list);
703
	m->quiesced = 0;
704
705
706
707
708
709
710
711
	m->prepared = 0;
	m->tc = tc;
	m->virt_block = virt_block;
	m->data_block = data_dest;
	m->cell = cell;
	m->err = 0;
	m->bio = NULL;

712
	if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
713
		m->quiesced = 1;
714
715
716
717
718
719
720
721

	/*
	 * IO to pool_dev remaps to the pool target's data_dev.
	 *
	 * If the whole block of data is being overwritten, we can issue the
	 * bio immediately. Otherwise we use kcopyd to clone the data first.
	 */
	if (io_overwrites_block(pool, bio)) {
722
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
723

724
		h->overwrite_mapping = m;
725
726
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
727
		inc_all_io_entry(pool, bio);
728
729
730
731
		remap_and_issue(tc, bio, data_dest);
	} else {
		struct dm_io_region from, to;

732
		from.bdev = origin->bdev;
733
734
735
736
737
738
739
740
741
742
743
		from.sector = data_origin * pool->sectors_per_block;
		from.count = pool->sectors_per_block;

		to.bdev = tc->pool_dev->bdev;
		to.sector = data_dest * pool->sectors_per_block;
		to.count = pool->sectors_per_block;

		r = dm_kcopyd_copy(pool->copier, &from, 1, &to,
				   0, copy_complete, m);
		if (r < 0) {
			mempool_free(m, pool->mapping_pool);
744
			DMERR_LIMIT("dm_kcopyd_copy() failed");
745
			dm_cell_error(cell);
746
747
748
749
		}
	}
}

750
751
static void schedule_internal_copy(struct thin_c *tc, dm_block_t virt_block,
				   dm_block_t data_origin, dm_block_t data_dest,
Mike Snitzer's avatar
Mike Snitzer committed
752
				   struct dm_bio_prison_cell *cell, struct bio *bio)
753
754
755
756
757
758
759
{
	schedule_copy(tc, virt_block, tc->pool_dev,
		      data_origin, data_dest, cell, bio);
}

static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
				   dm_block_t data_dest,
Mike Snitzer's avatar
Mike Snitzer committed
760
				   struct dm_bio_prison_cell *cell, struct bio *bio)
761
762
763
764
765
{
	schedule_copy(tc, virt_block, tc->origin_dev,
		      virt_block, data_dest, cell, bio);
}

766
static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
Mike Snitzer's avatar
Mike Snitzer committed
767
			  dm_block_t data_block, struct dm_bio_prison_cell *cell,
768
769
770
			  struct bio *bio)
{
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
771
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
772
773

	INIT_LIST_HEAD(&m->list);
774
	m->quiesced = 1;
775
776
777
778
779
780
781
782
783
784
785
786
787
	m->prepared = 0;
	m->tc = tc;
	m->virt_block = virt_block;
	m->data_block = data_block;
	m->cell = cell;
	m->err = 0;
	m->bio = NULL;

	/*
	 * If the whole block of data is being overwritten or we are not
	 * zeroing pre-existing data, we can issue the bio immediately.
	 * Otherwise we use kcopyd to zero the data first.
	 */
788
	if (!pool->pf.zero_new_blocks)
789
790
791
		process_prepared_mapping(m);

	else if (io_overwrites_block(pool, bio)) {
792
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
793

794
		h->overwrite_mapping = m;
795
796
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
797
		inc_all_io_entry(pool, bio);
798
799
800
801
802
803
804
805
806
807
808
809
		remap_and_issue(tc, bio, data_block);
	} else {
		int r;
		struct dm_io_region to;

		to.bdev = tc->pool_dev->bdev;
		to.sector = data_block * pool->sectors_per_block;
		to.count = pool->sectors_per_block;

		r = dm_kcopyd_zero(pool->copier, 1, &to, 0, copy_complete, m);
		if (r < 0) {
			mempool_free(m, pool->mapping_pool);
810
			DMERR_LIMIT("dm_kcopyd_zero() failed");
811
			dm_cell_error(cell);
812
813
814
815
		}
	}
}

816
817
818
819
820
821
static int commit(struct pool *pool)
{
	int r;

	r = dm_pool_commit_metadata(pool->pmd);
	if (r)
822
		DMERR_LIMIT("commit failed: error = %d", r);
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844

	return r;
}

/*
 * A non-zero return indicates read_only or fail_io mode.
 * Many callers don't care about the return value.
 */
static int commit_or_fallback(struct pool *pool)
{
	int r;

	if (get_pool_mode(pool) != PM_WRITE)
		return -EINVAL;

	r = commit(pool);
	if (r)
		set_pool_mode(pool, PM_READ_ONLY);

	return r;
}

845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
{
	int r;
	dm_block_t free_blocks;
	unsigned long flags;
	struct pool *pool = tc->pool;

	r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
	if (r)
		return r;

	if (free_blocks <= pool->low_water_blocks && !pool->low_water_triggered) {
		DMWARN("%s: reached low water mark, sending event.",
		       dm_device_name(pool->pool_md));
		spin_lock_irqsave(&pool->lock, flags);
		pool->low_water_triggered = 1;
		spin_unlock_irqrestore(&pool->lock, flags);
		dm_table_event(pool->ti->table);
	}

	if (!free_blocks) {
		if (pool->no_free_space)
			return -ENOSPC;
		else {
			/*
			 * Try to commit to see if that will free up some
			 * more space.
			 */
873
			(void) commit_or_fallback(pool);
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906

			r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
			if (r)
				return r;

			/*
			 * If we still have no space we set a flag to avoid
			 * doing all this checking and return -ENOSPC.
			 */
			if (!free_blocks) {
				DMWARN("%s: no free space available.",
				       dm_device_name(pool->pool_md));
				spin_lock_irqsave(&pool->lock, flags);
				pool->no_free_space = 1;
				spin_unlock_irqrestore(&pool->lock, flags);
				return -ENOSPC;
			}
		}
	}

	r = dm_pool_alloc_data_block(pool->pmd, result);
	if (r)
		return r;

	return 0;
}

/*
 * If we have run out of space, queue bios until the device is
 * resumed, presumably after having been reloaded with more space.
 */
static void retry_on_resume(struct bio *bio)
{
907
	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
908
	struct thin_c *tc = h->tc;
909
910
911
912
913
914
915
916
	struct pool *pool = tc->pool;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
	bio_list_add(&pool->retry_on_resume_list, bio);
	spin_unlock_irqrestore(&pool->lock, flags);
}

Mike Snitzer's avatar
Mike Snitzer committed
917
static void no_space(struct dm_bio_prison_cell *cell)
918
919
920
921
922
{
	struct bio *bio;
	struct bio_list bios;

	bio_list_init(&bios);
923
	dm_cell_release(cell, &bios);
924
925
926
927
928

	while ((bio = bio_list_pop(&bios)))
		retry_on_resume(bio);
}

Joe Thornber's avatar
Joe Thornber committed
929
930
931
static void process_discard(struct thin_c *tc, struct bio *bio)
{
	int r;
932
	unsigned long flags;
Joe Thornber's avatar
Joe Thornber committed
933
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
934
	struct dm_bio_prison_cell *cell, *cell2;
935
	struct dm_cell_key key, key2;
Joe Thornber's avatar
Joe Thornber committed
936
937
	dm_block_t block = get_bio_block(tc, bio);
	struct dm_thin_lookup_result lookup_result;
Mike Snitzer's avatar
Mike Snitzer committed
938
	struct dm_thin_new_mapping *m;
Joe Thornber's avatar
Joe Thornber committed
939
940

	build_virtual_key(tc->td, block, &key);
941
	if (dm_bio_detain(tc->pool->prison, &key, bio, &cell))
Joe Thornber's avatar
Joe Thornber committed
942
943
944
945
946
947
948
949
950
951
952
		return;

	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
	switch (r) {
	case 0:
		/*
		 * Check nobody is fiddling with this pool block.  This can
		 * happen if someone's in the process of breaking sharing
		 * on this block.
		 */
		build_data_key(tc->td, lookup_result.block, &key2);
953
		if (dm_bio_detain(tc->pool->prison, &key2, bio, &cell2)) {
954
			cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
955
956
957
958
959
960
961
962
963
964
			break;
		}

		if (io_overlaps_block(pool, bio)) {
			/*
			 * IO may still be going to the destination block.  We must
			 * quiesce before we can do the removal.
			 */
			m = get_next_mapping(pool);
			m->tc = tc;
965
			m->pass_discard = (!lookup_result.shared) && pool->pf.discard_passdown;
Joe Thornber's avatar
Joe Thornber committed
966
967
968
969
970
971
972
			m->virt_block = block;
			m->data_block = lookup_result.block;
			m->cell = cell;
			m->cell2 = cell2;
			m->err = 0;
			m->bio = bio;

973
			if (!dm_deferred_set_add_work(pool->all_io_ds, &m->list)) {
974
				spin_lock_irqsave(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
975
				list_add(&m->list, &pool->prepared_discards);
976
				spin_unlock_irqrestore(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
977
978
979
				wake_worker(pool);
			}
		} else {
980
			inc_all_io_entry(pool, bio);
981
982
			cell_defer_no_holder(tc, cell);
			cell_defer_no_holder(tc, cell2);
983

Joe Thornber's avatar
Joe Thornber committed
984
			/*
985
986
987
			 * The DM core makes sure that the discard doesn't span
			 * a block boundary.  So we submit the discard of a
			 * partial block appropriately.
Joe Thornber's avatar
Joe Thornber committed
988
			 */
989
990
991
992
			if ((!lookup_result.shared) && pool->pf.discard_passdown)
				remap_and_issue(tc, bio, lookup_result.block);
			else
				bio_endio(bio, 0);
Joe Thornber's avatar
Joe Thornber committed
993
994
995
996
997
998
999
		}
		break;

	case -ENODATA:
		/*
		 * It isn't provisioned, just forget it.
		 */
1000
		cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
1001
1002
1003
1004
		bio_endio(bio, 0);
		break;

	default:
1005
1006
		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
			    __func__, r);
1007
		cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
1008
1009
1010
1011
1012
		bio_io_error(bio);
		break;
	}
}

1013
static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,
1014
			  struct dm_cell_key *key,
1015
			  struct dm_thin_lookup_result *lookup_result,
Mike Snitzer's avatar
Mike Snitzer committed
1016
			  struct dm_bio_prison_cell *cell)
1017
1018
1019
1020
1021
1022
1023
{
	int r;
	dm_block_t data_block;

	r = alloc_data_block(tc, &data_block);
	switch (r) {
	case 0:
1024
1025
		schedule_internal_copy(tc, block, lookup_result->block,
				       data_block, cell, bio);
1026
1027
1028
1029
1030
1031
1032
		break;

	case -ENOSPC:
		no_space(cell);
		break;

	default:
1033
1034
		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
			    __func__, r);
1035
		dm_cell_error(cell);
1036
1037
1038
1039
1040
1041
1042
1043
		break;
	}
}

static void process_shared_bio(struct thin_c *tc, struct bio *bio,
			       dm_block_t block,
			       struct dm_thin_lookup_result *lookup_result)
{
Mike Snitzer's avatar
Mike Snitzer committed
1044
	struct dm_bio_prison_cell *cell;
1045
	struct pool *pool = tc->pool;
1046
	struct dm_cell_key key;
1047
1048
1049
1050
1051
1052

	/*
	 * If cell is already occupied, then sharing is already in the process
	 * of being broken so we have nothing further to do here.
	 */
	build_data_key(tc->td, lookup_result->block, &key);
1053
	if (dm_bio_detain(pool->prison, &key, bio, &cell))
1054
1055
		return;

1056
	if (bio_data_dir(bio) == WRITE && bio->bi_size)
1057
1058
		break_sharing(tc, bio, block, &key, lookup_result, cell);
	else {
1059
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
1060

1061
		h->shared_read_entry = dm_deferred_entry_inc(pool->shared_read_ds);
1062
		inc_all_io_entry(pool, bio);
1063
		cell_defer_no_holder(tc, cell);
1064

1065
1066
1067
1068
1069
		remap_and_issue(tc, bio, lookup_result->block);
	}
}

static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block,
Mike Snitzer's avatar
Mike Snitzer committed
1070
			    struct dm_bio_prison_cell *cell)
1071
1072
1073
1074
1075
1076
1077
1078
{
	int r;
	dm_block_t data_block;

	/*
	 * Remap empty bios (flushes) immediately, without provisioning.
	 */
	if (!bio->bi_size) {
1079
		inc_all_io_entry(tc->pool, bio);
1080
		cell_defer_no_holder(tc, cell);
1081

1082
1083
1084
1085
1086
1087
1088
1089
1090
		remap_and_issue(tc, bio, 0);
		return;
	}

	/*
	 * Fill read bios with zeroes and complete them immediately.
	 */
	if (bio_data_dir(bio) == READ) {
		zero_fill_bio(bio);
1091
		cell_defer_no_holder(tc, cell);
1092
1093
1094
1095
1096
1097
1098
		bio_endio(bio, 0);
		return;
	}

	r = alloc_data_block(tc, &data_block);
	switch (r) {
	case 0:
1099
1100
1101
1102
		if (tc->origin_dev)
			schedule_external_copy(tc, block, data_block, cell, bio);
		else
			schedule_zero(tc, block, data_block, cell, bio);
1103
1104
1105
1106
1107
1108
1109
		break;

	case -ENOSPC:
		no_space(cell);
		break;

	default:
1110
1111
		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
			    __func__, r);
1112
		set_pool_mode(tc->pool, PM_READ_ONLY);
1113
		dm_cell_error(cell);
1114
1115
1116
1117
1118
1119
1120
1121
		break;
	}
}

static void process_bio(struct thin_c *tc, struct bio *bio)
{
	int r;
	dm_block_t block = get_bio_block(tc, bio);
Mike Snitzer's avatar
Mike Snitzer committed
1122
	struct dm_bio_prison_cell *cell;
1123
	struct dm_cell_key key;
1124
1125
1126
1127
1128
1129
1130
	struct dm_thin_lookup_result lookup_result;

	/*
	 * If cell is already occupied, then the block is already
	 * being provisioned so we have nothing further to do here.
	 */
	build_virtual_key(tc->td, block, &key);
1131
	if (dm_bio_detain(tc->pool->prison, &key, bio, &cell))
1132
1133
1134
1135
1136
		return;

	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
	switch (r) {
	case 0:
1137
		if (lookup_result.shared) {
1138
			process_shared_bio(tc, bio, block, &lookup_result);
1139
			cell_defer_no_holder(tc, cell);
1140
1141
		} else {
			inc_all_io_entry(tc->pool, bio);
1142
			cell_defer_no_holder(tc, cell);
1143

1144
			remap_and_issue(tc, bio, lookup_result.block);
1145
		}
1146
1147
1148
		break;

	case -ENODATA:
1149
		if (bio_data_dir(bio) == READ && tc->origin_dev) {
1150
			inc_all_io_entry(tc->pool, bio);
1151
			cell_defer_no_holder(tc, cell);
1152

1153
1154
1155
			remap_to_origin_and_issue(tc, bio);
		} else
			provision_block(tc, bio, block, cell);
1156
1157
1158
		break;

	default:
1159
1160
		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
			    __func__, r);
1161
		cell_defer_no_holder(tc, cell);
1162
1163
1164
1165
1166
		bio_io_error(bio);
		break;
	}
}

1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
static void process_bio_read_only(struct thin_c *tc, struct bio *bio)
{
	int r;
	int rw = bio_data_dir(bio);
	dm_block_t block = get_bio_block(tc, bio);
	struct dm_thin_lookup_result lookup_result;

	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
	switch (r) {
	case 0:
		if (lookup_result.shared && (rw == WRITE) && bio->bi_size)
			bio_io_error(bio);
1179
1180
		else {
			inc_all_io_entry(tc->pool, bio);
1181
			remap_and_issue(tc, bio, lookup_result.block);
1182
		}
1183
1184
1185
1186
1187
1188
1189
1190
1191
		break;

	case -ENODATA:
		if (rw != READ) {
			bio_io_error(bio);
			break;
		}

		if (tc->origin_dev) {
1192
			inc_all_io_entry(tc->pool, bio);
1193
1194
1195
1196
1197
1198
1199
1200
1201
			remap_to_origin_and_issue(tc, bio);
			break;
		}

		zero_fill_bio(bio);
		bio_endio(bio, 0);
		break;

	default:
1202
1203
		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
			    __func__, r);
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
		bio_io_error(bio);
		break;
	}
}

static void process_bio_fail(struct thin_c *tc, struct bio *bio)
{
	bio_io_error(bio);
}

1214
1215
1216
1217
1218
1219
static int need_commit_due_to_time(struct pool *pool)
{
	return jiffies < pool->last_commit_jiffies ||
	       jiffies > pool->last_commit_jiffies + COMMIT_PERIOD;
}

1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
static void process_deferred_bios(struct pool *pool)
{
	unsigned long flags;
	struct bio *bio;
	struct bio_list bios;

	bio_list_init(&bios);

	spin_lock_irqsave(&pool->lock, flags);
	bio_list_merge(&bios, &pool->deferred_bios);
	bio_list_init(&pool->deferred_bios);
	spin_unlock_irqrestore(&pool->lock, flags);

	while ((bio = bio_list_pop(&bios))) {
1234
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
1235
1236
		struct thin_c *tc = h->tc;

1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
		/*
		 * If we've got no free new_mapping structs, and processing
		 * this bio might require one, we pause until there are some
		 * prepared mappings to process.
		 */
		if (ensure_next_mapping(pool)) {
			spin_lock_irqsave(&pool->lock, flags);
			bio_list_merge(&pool->deferred_bios, &bios);
			spin_unlock_irqrestore(&pool->lock, flags);

			break;
		}
Joe Thornber's avatar
Joe Thornber committed
1249
1250

		if (bio->bi_rw & REQ_DISCARD)
1251
			pool->process_discard(tc, bio);
Joe Thornber's avatar
Joe Thornber committed
1252
		else
1253
			pool->process_bio(tc, bio);
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
	}

	/*
	 * If there are any deferred flush bios, we must commit
	 * the metadata before issuing them.
	 */
	bio_list_init(&bios);
	spin_lock_irqsave(&pool->lock, flags);
	bio_list_merge(&bios, &pool->deferred_flush_bios);
	bio_list_init(&pool->deferred_flush_bios);
	spin_unlock_irqrestore(&pool->lock, flags);

1266
	if (bio_list_empty(&bios) && !need_commit_due_to_time(pool))
1267
1268
		return;

1269
	if (commit_or_fallback(pool)) {
1270
1271
1272
1273
		while ((bio = bio_list_pop(&bios)))
			bio_io_error(bio);
		return;
	}
1274
	pool->last_commit_jiffies = jiffies;
1275
1276
1277
1278
1279
1280
1281
1282
1283

	while ((bio = bio_list_pop(&bios)))
		generic_make_request(bio);
}

static void do_worker(struct work_struct *ws)
{
	struct pool *pool = container_of(ws, struct pool, worker);

1284
1285
	process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
	process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
1286
1287
1288
	process_deferred_bios(pool);
}

1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
/*
 * We want to commit periodically so that not too much
 * unwritten data builds up.
 */
static void do_waker(struct work_struct *ws)
{
	struct pool *pool = container_of(to_delayed_work(ws), struct pool, waker);
	wake_worker(pool);
	queue_delayed_work(pool->wq, &pool->waker, COMMIT_PERIOD);
}

1300
1301
/*----------------------------------------------------------------*/

1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
static enum pool_mode get_pool_mode(struct pool *pool)
{
	return pool->pf.mode;
}

static void set_pool_mode(struct pool *pool, enum pool_mode mode)
{
	int r;

	pool->pf.mode = mode;

	switch (mode) {
	case PM_FAIL:
		DMERR("switching pool to failure mode");
		pool->process_bio = process_bio_fail;
		pool->process_discard = process_bio_fail;
		pool->process_prepared_mapping = process_prepared_mapping_fail;
		pool->process_prepared_discard = process_prepared_discard_fail;
		break;

	case PM_READ_ONLY:
		DMERR("switching pool to read-only mode");
		r = dm_pool_abort_metadata(pool->pmd);
		if (r) {
			DMERR("aborting transaction failed");
			set_pool_mode(pool, PM_FAIL);
		} else {
			dm_pool_metadata_read_only(pool->pmd);
			pool->process_bio = process_bio_read_only;
			pool->process_discard = process_discard;
			pool->process_prepared_mapping = process_prepared_mapping_fail;
			pool->process_prepared_discard = process_prepared_discard_passdown;
		}
		break;

	case PM_WRITE:
		pool->process_bio = process_bio;
		pool->process_discard = process_discard;
		pool->process_prepared_mapping = process_prepared_mapping;
		pool->process_prepared_discard = process_prepared_discard;
		break;
	}
}

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

1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
/*
 * Mapping functions.
 */

/*
 * Called only while mapping a thin bio to hand it over to the workqueue.
 */
static void thin_defer_bio(struct thin_c *tc, struct bio *bio)
{
	unsigned long flags;
	struct pool *pool = tc->pool;

	spin_lock_irqsave(&pool->lock, flags);
	bio_list_add(&pool->deferred_bios, bio);
	spin_unlock_irqrestore(&pool->lock, flags);

	wake_worker(pool);
}

1367
static void thin_hook_bio(struct thin_c *tc, struct bio *bio)
1368
{
1369
	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
1370
1371
1372

	h->tc = tc;
	h->shared_read_entry = NULL;
1373
	h->all_io_entry = NULL;
1374
1375
1376
	h->overwrite_mapping = NULL;
}

1377
1378
1379
/*
 * Non-blocking function called from the thin target's map function.
 */
Mikulas Patocka's avatar
Mikulas Patocka committed
1380
static int thin_bio_map(struct dm_target *ti, struct bio *bio)
1381
1382
1383
1384
1385
1386
{
	int r;
	struct thin_c *tc = ti->private;
	dm_block_t block = get_bio_block(tc, bio);
	struct dm_thin_device *td = tc->td;
	struct dm_thin_lookup_result result;
1387
1388
	struct dm_bio_prison_cell *cell1, *cell2;
	struct dm_cell_key key;
1389

1390
	thin_hook_bio(tc, bio);
1391
1392
1393
1394
1395
1396

	if (get_pool_mode(tc->pool) == PM_FAIL) {
		bio_io_error(bio);
		return DM_MAPIO_SUBMITTED;
	}

Joe Thornber's avatar
Joe Thornber committed
1397
	if (bio->bi_rw & (REQ_DISCARD | REQ_FLUSH | REQ_FUA)) {
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
		thin_defer_bio(tc, bio);
		return DM_MAPIO_SUBMITTED;
	}

	r = dm_thin_find_block(td, block, 0, &result);

	/*
	 * Note that we defer readahead too.
	 */
	switch (r) {
	case 0:
		if (unlikely(result.shared)) {
			/*
			 * We have a race condition here between the
			 * result.shared value returned by the lookup and
			 * snapshot creation, which may cause new
			 * sharing.
			 *
			 * To avoid this always quiesce the origin before
			 * taking the snap.  You want to do this anyway to
			 * ensure a consistent application view
			 * (i.e. lockfs).
			 *
			 * More distant ancestors are irrelevant. The
			 * shared flag will be set in their case.
			 */
			thin_defer_bio(tc, bio);
1425
			return DM_MAPIO_SUBMITTED;
1426
		}
1427
1428
1429
1430
1431
1432
1433

		build_virtual_key(tc->td, block, &key);
		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1))
			return DM_MAPIO_SUBMITTED;

		build_data_key(tc->td, result.block, &key);
		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell2)) {