dm-thin.c 87.8 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

#include <linux/device-mapper.h>
#include <linux/dm-io.h>
#include <linux/dm-kcopyd.h>
#include <linux/list.h>
15
#include <linux/rculist.h>
16
17
18
#include <linux/init.h>
#include <linux/module.h>
#include <linux/slab.h>
19
#include <linux/rbtree.h>
20
21
22
23
24
25

#define	DM_MSG_PREFIX	"thin"

/*
 * Tunable constants
 */
26
#define ENDIO_HOOK_POOL_SIZE 1024
27
28
#define MAPPING_POOL_SIZE 1024
#define PRISON_CELLS 1024
29
#define COMMIT_PERIOD HZ
30
31
32
#define NO_SPACE_TIMEOUT_SECS 60

static unsigned no_space_timeout_secs = NO_SPACE_TIMEOUT_SECS;
33

34
35
36
DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(snapshot_copy_throttle,
		"A percentage of time allocated for copy on write");

37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
/*
 * 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
69
 * including all devices that share this block.  (see dm_deferred_set code)
70
71
72
73
74
 *
 * 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
75
 * (process_prepared_mapping).  This act of inserting breaks some
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
 * 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,
113
			   dm_block_t b, struct dm_cell_key *key)
114
115
116
117
118
119
120
{
	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,
121
			      struct dm_cell_key *key)
122
123
124
125
126
127
128
129
130
131
132
133
134
{
	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
135
struct dm_thin_new_mapping;
136

137
/*
138
 * The pool runs in 4 modes.  Ordered in degraded order for comparisons.
139
140
141
 */
enum pool_mode {
	PM_WRITE,		/* metadata may be changed */
142
	PM_OUT_OF_DATA_SPACE,	/* metadata may be changed, though data may not be allocated */
143
144
145
146
	PM_READ_ONLY,		/* metadata may not be changed */
	PM_FAIL,		/* all I/O fails */
};

147
struct pool_features {
148
149
	enum pool_mode mode;

Mike Snitzer's avatar
Mike Snitzer committed
150
151
152
	bool zero_new_blocks:1;
	bool discard_enabled:1;
	bool discard_passdown:1;
153
	bool error_if_no_space:1;
154
155
};

156
157
158
159
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);

160
161
162
163
164
165
166
167
168
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;
169
	uint32_t sectors_per_block;
170
	int sectors_per_block_shift;
171

172
	struct pool_features pf;
173
	bool low_water_triggered:1;	/* A dm event has been sent */
174

175
	struct dm_bio_prison *prison;
176
177
178
179
	struct dm_kcopyd_client *copier;

	struct workqueue_struct *wq;
	struct work_struct worker;
180
	struct delayed_work waker;
181
	struct delayed_work no_space_timeout;
182

183
	unsigned long last_commit_jiffies;
184
	unsigned ref_count;
185
186
187
188

	spinlock_t lock;
	struct bio_list deferred_flush_bios;
	struct list_head prepared_mappings;
Joe Thornber's avatar
Joe Thornber committed
189
	struct list_head prepared_discards;
190
	struct list_head active_thins;
191

192
193
	struct dm_deferred_set *shared_read_ds;
	struct dm_deferred_set *all_io_ds;
194

Mike Snitzer's avatar
Mike Snitzer committed
195
	struct dm_thin_new_mapping *next_mapping;
196
	mempool_t *mapping_pool;
197
198
199
200
201
202

	process_bio_fn process_bio;
	process_bio_fn process_discard;

	process_mapping_fn process_prepared_mapping;
	process_mapping_fn process_prepared_discard;
203
204
};

205
static enum pool_mode get_pool_mode(struct pool *pool);
206
static void metadata_operation_failed(struct pool *pool, const char *op, int r);
207

208
209
210
211
212
213
214
215
216
217
218
/*
 * 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;
219
220
	struct pool_features requested_pf; /* Features requested during table load */
	struct pool_features adjusted_pf;  /* Features used after adjusting for constituent devices */
221
222
223
224
225
226
};

/*
 * Target context for a thin.
 */
struct thin_c {
227
	struct list_head list;
228
	struct dm_dev *pool_dev;
229
	struct dm_dev *origin_dev;
230
231
232
233
	dm_thin_id dev_id;

	struct pool *pool;
	struct dm_thin_device *td;
234
	bool requeue_mode:1;
235
236
237
	spinlock_t lock;
	struct bio_list deferred_bio_list;
	struct bio_list retry_on_resume_list;
238
	struct rb_root sort_bio_list; /* sorted list of deferred bios */
239
240
241
242
243
244
245

	/*
	 * Ensures the thin is not destroyed until the worker has finished
	 * iterating the active_thins list.
	 */
	atomic_t refcount;
	struct completion can_destroy;
246
247
248
249
};

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

250
251
252
253
254
255
256
257
258
259
260
/*
 * 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);
}

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

261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
static int bio_detain(struct pool *pool, struct dm_cell_key *key, struct bio *bio,
		      struct dm_bio_prison_cell **cell_result)
{
	int r;
	struct dm_bio_prison_cell *cell_prealloc;

	/*
	 * Allocate a cell from the prison's mempool.
	 * This might block but it can't fail.
	 */
	cell_prealloc = dm_bio_prison_alloc_cell(pool->prison, GFP_NOIO);

	r = dm_bio_detain(pool->prison, key, bio, cell_prealloc, cell_result);
	if (r)
		/*
		 * We reused an old cell; we can get rid of
		 * the new one.
		 */
		dm_bio_prison_free_cell(pool->prison, cell_prealloc);

	return r;
}

static void cell_release(struct pool *pool,
			 struct dm_bio_prison_cell *cell,
			 struct bio_list *bios)
{
	dm_cell_release(pool->prison, cell, bios);
	dm_bio_prison_free_cell(pool->prison, cell);
}

static void cell_release_no_holder(struct pool *pool,
				   struct dm_bio_prison_cell *cell,
				   struct bio_list *bios)
{
	dm_cell_release_no_holder(pool->prison, cell, bios);
	dm_bio_prison_free_cell(pool->prison, cell);
}

300
301
302
303
304
305
static void cell_defer_no_holder_no_free(struct thin_c *tc,
					 struct dm_bio_prison_cell *cell)
{
	struct pool *pool = tc->pool;
	unsigned long flags;

306
307
308
	spin_lock_irqsave(&tc->lock, flags);
	dm_cell_release_no_holder(pool->prison, cell, &tc->deferred_bio_list);
	spin_unlock_irqrestore(&tc->lock, flags);
309
310
311
312

	wake_worker(pool);
}

313
314
static void cell_error_with_code(struct pool *pool,
				 struct dm_bio_prison_cell *cell, int error_code)
315
{
316
	dm_cell_error(pool->prison, cell, error_code);
317
318
319
	dm_bio_prison_free_cell(pool->prison, cell);
}

320
321
322
323
324
static void cell_error(struct pool *pool, struct dm_bio_prison_cell *cell)
{
	cell_error_with_code(pool, cell, -EIO);
}

325
326
/*----------------------------------------------------------------*/

327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
/*
 * 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
387
struct dm_thin_endio_hook {
388
	struct thin_c *tc;
389
390
	struct dm_deferred_entry *shared_read_entry;
	struct dm_deferred_entry *all_io_entry;
Mike Snitzer's avatar
Mike Snitzer committed
391
	struct dm_thin_new_mapping *overwrite_mapping;
392
	struct rb_node rb_node;
393
394
};

395
static void requeue_bio_list(struct thin_c *tc, struct bio_list *master)
396
397
398
{
	struct bio *bio;
	struct bio_list bios;
399
	unsigned long flags;
400
401

	bio_list_init(&bios);
402

403
	spin_lock_irqsave(&tc->lock, flags);
404
405
	bio_list_merge(&bios, master);
	bio_list_init(master);
406
	spin_unlock_irqrestore(&tc->lock, flags);
407

408
409
	while ((bio = bio_list_pop(&bios)))
		bio_endio(bio, DM_ENDIO_REQUEUE);
410
411
412
413
}

static void requeue_io(struct thin_c *tc)
{
414
415
	requeue_bio_list(tc, &tc->deferred_bio_list);
	requeue_bio_list(tc, &tc->retry_on_resume_list);
416
417
}

418
static void error_thin_retry_list(struct thin_c *tc)
419
420
421
422
423
424
425
{
	struct bio *bio;
	unsigned long flags;
	struct bio_list bios;

	bio_list_init(&bios);

426
427
428
429
	spin_lock_irqsave(&tc->lock, flags);
	bio_list_merge(&bios, &tc->retry_on_resume_list);
	bio_list_init(&tc->retry_on_resume_list);
	spin_unlock_irqrestore(&tc->lock, flags);
430
431
432
433
434

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

435
436
437
438
439
440
441
442
443
444
static void error_retry_list(struct pool *pool)
{
	struct thin_c *tc;

	rcu_read_lock();
	list_for_each_entry_rcu(tc, &pool->active_thins, list)
		error_thin_retry_list(tc);
	rcu_read_unlock();
}

445
446
447
448
449
450
451
/*
 * 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.
 */

452
453
454
455
456
static bool block_size_is_power_of_two(struct pool *pool)
{
	return pool->sectors_per_block_shift >= 0;
}

457
458
static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
{
459
	struct pool *pool = tc->pool;
460
	sector_t block_nr = bio->bi_iter.bi_sector;
461

462
463
	if (block_size_is_power_of_two(pool))
		block_nr >>= pool->sectors_per_block_shift;
464
	else
465
		(void) sector_div(block_nr, pool->sectors_per_block);
466
467

	return block_nr;
468
469
470
471
472
}

static void remap(struct thin_c *tc, struct bio *bio, dm_block_t block)
{
	struct pool *pool = tc->pool;
473
	sector_t bi_sector = bio->bi_iter.bi_sector;
474
475

	bio->bi_bdev = tc->pool_dev->bdev;
476
	if (block_size_is_power_of_two(pool))
477
478
479
		bio->bi_iter.bi_sector =
			(block << pool->sectors_per_block_shift) |
			(bi_sector & (pool->sectors_per_block - 1));
480
	else
481
		bio->bi_iter.bi_sector = (block * pool->sectors_per_block) +
482
				 sector_div(bi_sector, pool->sectors_per_block);
483
484
}

485
486
487
488
489
static void remap_to_origin(struct thin_c *tc, struct bio *bio)
{
	bio->bi_bdev = tc->origin_dev->bdev;
}

490
491
492
493
494
495
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);
}

496
497
498
499
500
501
502
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;

503
	h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
504
505
506
	h->all_io_entry = dm_deferred_entry_inc(pool->all_io_ds);
}

507
static void issue(struct thin_c *tc, struct bio *bio)
508
509
510
511
{
	struct pool *pool = tc->pool;
	unsigned long flags;

512
513
514
515
516
	if (!bio_triggers_commit(tc, bio)) {
		generic_make_request(bio);
		return;
	}

517
	/*
518
519
520
	 * 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.
521
	 */
522
523
524
525
526
527
528
529
530
531
532
533
	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);
534
535
}

536
537
538
539
540
541
542
543
544
545
546
547
548
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);
}

549
550
551
552
553
/*----------------------------------------------------------------*/

/*
 * Bio endio functions.
 */
Mike Snitzer's avatar
Mike Snitzer committed
554
struct dm_thin_new_mapping {
555
556
	struct list_head list;

557
558
	bool pass_discard:1;
	bool definitely_not_shared:1;
559

560
561
562
563
564
565
566
	/*
	 * Track quiescing, copying and zeroing preparation actions.  When this
	 * counter hits zero the block is prepared and can be inserted into the
	 * btree.
	 */
	atomic_t prepare_actions;

567
	int err;
568
569
570
	struct thin_c *tc;
	dm_block_t virt_block;
	dm_block_t data_block;
Mike Snitzer's avatar
Mike Snitzer committed
571
	struct dm_bio_prison_cell *cell, *cell2;
572
573
574
575
576
577
578
579
580
581
582

	/*
	 * 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;
};

583
static void __complete_mapping_preparation(struct dm_thin_new_mapping *m)
584
585
586
{
	struct pool *pool = m->tc->pool;

587
	if (atomic_dec_and_test(&m->prepare_actions)) {
588
		list_add_tail(&m->list, &pool->prepared_mappings);
589
590
591
592
593
594
595
		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
596
	struct dm_thin_new_mapping *m = context;
597
598
599
600
601
	struct pool *pool = m->tc->pool;

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

	spin_lock_irqsave(&pool->lock, flags);
602
	__complete_mapping_preparation(m);
603
604
605
606
607
608
	spin_unlock_irqrestore(&pool->lock, flags);
}

static void overwrite_endio(struct bio *bio, int err)
{
	unsigned long flags;
609
	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
610
	struct dm_thin_new_mapping *m = h->overwrite_mapping;
611
612
613
614
615
	struct pool *pool = m->tc->pool;

	m->err = err;

	spin_lock_irqsave(&pool->lock, flags);
616
	__complete_mapping_preparation(m);
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
	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
633
static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *cell)
634
635
636
637
{
	struct pool *pool = tc->pool;
	unsigned long flags;

638
639
640
	spin_lock_irqsave(&tc->lock, flags);
	cell_release(pool, cell, &tc->deferred_bio_list);
	spin_unlock_irqrestore(&tc->lock, flags);
641
642
643
644
645

	wake_worker(pool);
}

/*
646
 * Same as cell_defer above, except it omits the original holder of the cell.
647
 */
648
static void cell_defer_no_holder(struct thin_c *tc, struct dm_bio_prison_cell *cell)
649
650
651
652
{
	struct pool *pool = tc->pool;
	unsigned long flags;

653
654
655
	spin_lock_irqsave(&tc->lock, flags);
	cell_release_no_holder(pool, cell, &tc->deferred_bio_list);
	spin_unlock_irqrestore(&tc->lock, flags);
656
657
658
659

	wake_worker(pool);
}

660
661
static void process_prepared_mapping_fail(struct dm_thin_new_mapping *m)
{
Kent Overstreet's avatar
Kent Overstreet committed
662
	if (m->bio) {
663
		m->bio->bi_end_io = m->saved_bi_end_io;
Kent Overstreet's avatar
Kent Overstreet committed
664
665
		atomic_inc(&m->bio->bi_remaining);
	}
666
	cell_error(m->tc->pool, m->cell);
667
668
669
	list_del(&m->list);
	mempool_free(m, m->tc->pool->mapping_pool);
}
670

Mike Snitzer's avatar
Mike Snitzer committed
671
static void process_prepared_mapping(struct dm_thin_new_mapping *m)
672
673
{
	struct thin_c *tc = m->tc;
674
	struct pool *pool = tc->pool;
675
676
677
678
	struct bio *bio;
	int r;

	bio = m->bio;
Kent Overstreet's avatar
Kent Overstreet committed
679
	if (bio) {
680
		bio->bi_end_io = m->saved_bi_end_io;
Kent Overstreet's avatar
Kent Overstreet committed
681
682
		atomic_inc(&bio->bi_remaining);
	}
683
684

	if (m->err) {
685
		cell_error(pool, m->cell);
686
		goto out;
687
688
689
690
691
692
693
694
695
	}

	/*
	 * 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) {
696
		metadata_operation_failed(pool, "dm_thin_insert_block", r);
697
		cell_error(pool, m->cell);
698
		goto out;
699
700
701
702
703
704
705
706
707
	}

	/*
	 * 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) {
708
		cell_defer_no_holder(tc, m->cell);
709
710
		bio_endio(bio, 0);
	} else
Joe Thornber's avatar
Joe Thornber committed
711
		cell_defer(tc, m->cell);
712

713
out:
714
	list_del(&m->list);
715
	mempool_free(m, pool->mapping_pool);
716
717
}

718
static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
Joe Thornber's avatar
Joe Thornber committed
719
720
721
{
	struct thin_c *tc = m->tc;

722
	bio_io_error(m->bio);
723
724
	cell_defer_no_holder(tc, m->cell);
	cell_defer_no_holder(tc, m->cell2);
725
726
727
728
729
730
	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
731

732
	inc_all_io_entry(tc->pool, m->bio);
733
734
	cell_defer_no_holder(tc, m->cell);
	cell_defer_no_holder(tc, m->cell2);
735

Joe Thornber's avatar
Joe Thornber committed
736
	if (m->pass_discard)
737
738
739
740
741
742
743
744
745
		if (m->definitely_not_shared)
			remap_and_issue(tc, m->bio, m->data_block);
		else {
			bool used = false;
			if (dm_pool_block_is_used(tc->pool->pmd, m->data_block, &used) || used)
				bio_endio(m->bio, 0);
			else
				remap_and_issue(tc, m->bio, m->data_block);
		}
Joe Thornber's avatar
Joe Thornber committed
746
747
748
749
750
751
	else
		bio_endio(m->bio, 0);

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

752
753
754
755
756
757
758
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)
759
		DMERR_LIMIT("dm_thin_remove_block() failed");
760
761
762
763

	process_prepared_discard_passdown(m);
}

Joe Thornber's avatar
Joe Thornber committed
764
static void process_prepared(struct pool *pool, struct list_head *head,
765
			     process_mapping_fn *fn)
766
767
768
{
	unsigned long flags;
	struct list_head maps;
Mike Snitzer's avatar
Mike Snitzer committed
769
	struct dm_thin_new_mapping *m, *tmp;
770
771
772

	INIT_LIST_HEAD(&maps);
	spin_lock_irqsave(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
773
	list_splice_init(head, &maps);
774
775
776
	spin_unlock_irqrestore(&pool->lock, flags);

	list_for_each_entry_safe(m, tmp, &maps, list)
777
		(*fn)(m);
778
779
780
781
782
}

/*
 * Deferred bio jobs.
 */
Joe Thornber's avatar
Joe Thornber committed
783
static int io_overlaps_block(struct pool *pool, struct bio *bio)
784
{
785
786
	return bio->bi_iter.bi_size ==
		(pool->sectors_per_block << SECTOR_SHIFT);
Joe Thornber's avatar
Joe Thornber committed
787
788
789
790
791
792
}

static int io_overwrites_block(struct pool *pool, struct bio *bio)
{
	return (bio_data_dir(bio) == WRITE) &&
		io_overlaps_block(pool, bio);
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
}

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
812
static struct dm_thin_new_mapping *get_next_mapping(struct pool *pool)
813
{
814
	struct dm_thin_new_mapping *m = pool->next_mapping;
815
816
817

	BUG_ON(!pool->next_mapping);

818
819
820
821
	memset(m, 0, sizeof(struct dm_thin_new_mapping));
	INIT_LIST_HEAD(&m->list);
	m->bio = NULL;

822
823
	pool->next_mapping = NULL;

824
	return m;
825
826
827
}

static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
828
829
			  struct dm_dev *origin, dm_block_t data_origin,
			  dm_block_t data_dest,
Mike Snitzer's avatar
Mike Snitzer committed
830
			  struct dm_bio_prison_cell *cell, struct bio *bio)
831
832
833
{
	int r;
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
834
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
835
836
837
838
839
840

	m->tc = tc;
	m->virt_block = virt_block;
	m->data_block = data_dest;
	m->cell = cell;

841
	if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
842
843
844
		atomic_set(&m->prepare_actions, 1); /* copy only */
	else
		atomic_set(&m->prepare_actions, 2); /* quiesce + copy */
845
846
847
848
849
850
851
852

	/*
	 * 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)) {
853
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
854

855
		h->overwrite_mapping = m;
856
857
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
858
		inc_all_io_entry(pool, bio);
859
860
861
862
		remap_and_issue(tc, bio, data_dest);
	} else {
		struct dm_io_region from, to;

863
		from.bdev = origin->bdev;
864
865
866
867
868
869
870
871
872
873
874
		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);
875
			DMERR_LIMIT("dm_kcopyd_copy() failed");
876
			cell_error(pool, cell);
877
878
879
880
		}
	}
}

881
882
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
883
				   struct dm_bio_prison_cell *cell, struct bio *bio)
884
885
886
887
888
889
890
{
	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
891
				   struct dm_bio_prison_cell *cell, struct bio *bio)
892
893
894
895
896
{
	schedule_copy(tc, virt_block, tc->origin_dev,
		      virt_block, data_dest, cell, bio);
}

897
static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
Mike Snitzer's avatar
Mike Snitzer committed
898
			  dm_block_t data_block, struct dm_bio_prison_cell *cell,
899
900
901
			  struct bio *bio)
{
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
902
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
903

904
	atomic_set(&m->prepare_actions, 1); /* no need to quiesce */
905
906
907
908
909
910
911
912
913
914
	m->tc = tc;
	m->virt_block = virt_block;
	m->data_block = data_block;
	m->cell = cell;

	/*
	 * 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.
	 */
915
	if (!pool->pf.zero_new_blocks)
916
917
918
		process_prepared_mapping(m);

	else if (io_overwrites_block(pool, bio)) {
919
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
Mike Snitzer's avatar
Mike Snitzer committed
920

921
		h->overwrite_mapping = m;
922
923
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
924
		inc_all_io_entry(pool, bio);
925
926
927
928
929
930
931
932
933
934
935
936
		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);
937
			DMERR_LIMIT("dm_kcopyd_zero() failed");
938
			cell_error(pool, cell);
939
940
941
942
		}
	}
}

943
944
945
946
/*
 * A non-zero return indicates read_only or fail_io mode.
 * Many callers don't care about the return value.
 */
947
static int commit(struct pool *pool)
948
949
950
{
	int r;

951
	if (get_pool_mode(pool) >= PM_READ_ONLY)
952
953
		return -EINVAL;

954
	r = dm_pool_commit_metadata(pool->pmd);
955
956
	if (r)
		metadata_operation_failed(pool, "dm_pool_commit_metadata", r);
957
958
959
960

	return r;
}

961
962
963
964
965
966
967
968
969
970
971
972
973
974
static void check_low_water_mark(struct pool *pool, dm_block_t free_blocks)
{
	unsigned long flags;

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

975
976
static void set_pool_mode(struct pool *pool, enum pool_mode new_mode);

977
978
979
980
981
982
static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
{
	int r;
	dm_block_t free_blocks;
	struct pool *pool = tc->pool;

983
	if (WARN_ON(get_pool_mode(pool) != PM_WRITE))
984
985
		return -EINVAL;

986
	r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
987
988
	if (r) {
		metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);
989
		return r;
990
	}
991

992
	check_low_water_mark(pool, free_blocks);
993
994

	if (!free_blocks) {
995
996
997
998
		/*
		 * Try to commit to see if that will free up some
		 * more space.
		 */
999
1000
1001
		r = commit(pool);
		if (r)
			return r;
1002

1003
		r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
1004
1005
		if (r) {
			metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);
1006
			return r;
1007
		}
1008

1009
		if (!free_blocks) {
1010
			set_pool_mode(pool, PM_OUT_OF_DATA_SPACE);
1011
			return -ENOSPC;
1012
1013
1014
1015
		}
	}

	r = dm_pool_alloc_data_block(pool->pmd, result);
1016
	if (r) {
1017
		metadata_operation_failed(pool, "dm_pool_alloc_data_block", r);
1018
		return r;
1019
	}
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029

	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)
{
1030
	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
1031
	struct thin_c *tc = h->tc;
1032
1033
	unsigned long flags;

1034
1035
1036
	spin_lock_irqsave(&tc->lock, flags);
	bio_list_add(&tc->retry_on_resume_list, bio);
	spin_unlock_irqrestore(&tc->lock, flags);
1037
1038
}

1039
static int should_error_unserviceable_bio(struct pool *pool)
1040
{
1041
1042
1043
1044
1045
1046
	enum pool_mode m = get_pool_mode(pool);

	switch (m) {
	case PM_WRITE:
		/* Shouldn't get here */
		DMERR_LIMIT("bio unserviceable, yet pool is in PM_WRITE mode");
1047
		return -EIO;
1048
1049

	case PM_OUT_OF_DATA_SPACE:
1050
		return pool->pf.error_if_no_space ? -ENOSPC : 0;
1051
1052
1053

	case PM_READ_ONLY:
	case PM_FAIL:
1054
		return -EIO;
1055
1056
1057
	default:
		/* Shouldn't get here */
		DMERR_LIMIT("bio unserviceable, yet pool has an unknown mode");
1058
		return -EIO;
1059
1060
	}
}
1061

1062
1063
static void handle_unserviceable_bio(struct pool *pool, struct bio *bio)
{
1064
1065
1066
1067
	int error = should_error_unserviceable_bio(pool);

	if (error)
		bio_endio(bio, error);
1068
1069
	else
		retry_on_resume(bio);
1070
1071
}

1072
static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *cell)
1073
1074
1075
{
	struct bio *bio;
	struct bio_list bios;
1076
	int error;
1077

1078
1079
1080
	error = should_error_unserviceable_bio(pool);
	if (error) {
		cell_error_with_code(pool, cell, error);
1081
1082
1083
		return;
	}

1084
	bio_list_init(&bios);
1085
	cell_release(pool, cell, &bios);
1086

1087
1088
	error = should_error_unserviceable_bio(pool);
	if (error)
1089
		while ((bio = bio_list_pop(&bios)))
1090
			bio_endio(bio, error);
1091
1092
1093
	else
		while ((bio = bio_list_pop(&bios)))
			retry_on_resume(bio);
1094
1095
}

Joe Thornber's avatar
Joe Thornber committed
1096
1097
1098
static void process_discard(struct thin_c *tc, struct bio *bio)
{
	int r;
1099
	unsigned long flags;
Joe Thornber's avatar
Joe Thornber committed
1100
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
1101
	struct dm_bio_prison_cell *cell, *cell2;
1102
	struct dm_cell_key key, key2;
Joe Thornber's avatar
Joe Thornber committed
1103
1104
	dm_block_t block = get_bio_block(tc, bio);
	struct dm_thin_lookup_result lookup_result;
Mike Snitzer's avatar
Mike Snitzer committed
1105
	struct dm_thin_new_mapping *m;
Joe Thornber's avatar
Joe Thornber committed
1106
1107

	build_virtual_key(tc->td, block, &key);
1108
	if (bio_detain(tc->pool, &key, bio, &cell))
Joe Thornber's avatar
Joe Thornber committed
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
		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);
1120
		if (bio_detain(tc->pool, &key2, bio, &cell2)) {
1121
			cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
			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;
1132
1133
			m->pass_discard = pool->pf.discard_passdown;
			m->definitely_not_shared = !lookup_result.shared;
Joe Thornber's avatar
Joe Thornber committed
1134
1135
1136
1137
1138
1139
			m->virt_block = block;
			m->data_block = lookup_result.block;
			m->cell = cell;
			m->cell2 = cell2;
			m->bio = bio;

1140
			if (!dm_deferred_set_add_work(pool->all_io_ds, &m->list)) {
1141
				spin_lock_irqsave(&pool->lock, flags);
1142
				list_add_tail(&m->list, &pool->prepared_discards);
1143
				spin_unlock_irqrestore(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
1144
1145
1146
				wake_worker(pool);
			}
		} else {
1147
			inc_all_io_entry(pool, bio);
1148
1149
			cell_defer_no_holder(tc, cell);
			cell_defer_no_holder(tc, cell2);
1150

Joe Thornber's avatar
Joe Thornber committed
1151
			/*
1152
1153
1154
			 * 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
1155
			 */
1156
1157
1158
1159
			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
1160
1161
1162
1163
1164
1165
1166
		}
		break;

	case -ENODATA:
		/*
		 * It isn't provisioned, just forget it.
		 */
1167
		cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
1168
1169
1170
1171
		bio_endio(bio, 0);
		break;

	default:
1172
1173
		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
			    __func__, r);
1174
		cell_defer_no_holder(tc, cell);
Joe Thornber's avatar
Joe Thornber committed
1175
1176
1177
1178
1179
		bio_io_error(bio);
		break;
	}
}

1180
static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,
1181
			  struct dm_cell_key *key,
1182
			  struct dm_thin_lookup_result *lookup_result,
Mike Snitzer's avatar
Mike Snitzer committed
1183
			  struct dm_bio_prison_cell *cell)
1184
1185
1186
{
	int r;
	dm_block_t data_block;
1187
	struct pool *pool = tc->pool;
1188
1189
1190
1191

	r = alloc_data_block(tc, &data_block);
	switch (r) {
	case 0:
1192
1193
		schedule_internal_copy(tc, block, lookup_result->block,
				       data_block, cell, bio);
1194
1195
1196
		break;

	case -ENOSPC:
1197
		retry_bios_on_resume(pool, cell);
1198
1199
1200
		break;

	default:
1201
1202
		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
			    __func__, r);
1203
		cell_error(pool, cell);
1204
1205
1206
1207
1208
1209
1210
1211
		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
1212
	struct dm_bio_prison_cell *cell;
1213
	struct pool *pool = tc->pool;
1214
	struct dm_cell_key key;
1215
1216
1217
1218
1219
1220

	/*
	 * 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);
1221
	if (bio_detain(pool, &key, bio, &cell))
1222
1223
		return;

1224
	if (bio_data_dir(bio) == WRITE && bio->bi_iter.bi_size)
1225
1226
		break_sharing(tc, bio, block, &key, lookup_result, cell);
	else {
1227
		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));