dm-thin.c 74.8 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.h"
9
10
11
12
13
14
15
16
17
18
19
20
21
22

#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
 */
23
#define ENDIO_HOOK_POOL_SIZE 1024
24
25
26
#define DEFERRED_SET_SIZE 64
#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
61
62
63
64
65
66

/*
 * 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
 * including all devices that share this block.  (see deferred_set code)
 *
 * 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
105
106
107
108
109
110
111
112
113
114
 * 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.
 */

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

/*
 * Sometimes we can't deal with a bio straight away.  We put them in prison
 * where they can't cause any mischief.  Bios are put in a cell identified
 * by a key, multiple bios can be in the same cell.  When the cell is
 * subsequently unlocked the bios become available.
 */
struct bio_prison;

struct cell_key {
	int virtual;
	dm_thin_id dev;
	dm_block_t block;
};

Mike Snitzer's avatar
Mike Snitzer committed
115
struct dm_bio_prison_cell {
116
117
118
	struct hlist_node list;
	struct bio_prison *prison;
	struct cell_key key;
119
	struct bio *holder;
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
	struct bio_list bios;
};

struct bio_prison {
	spinlock_t lock;
	mempool_t *cell_pool;

	unsigned nr_buckets;
	unsigned hash_mask;
	struct hlist_head *cells;
};

static uint32_t calc_nr_buckets(unsigned nr_cells)
{
	uint32_t n = 128;

	nr_cells /= 4;
	nr_cells = min(nr_cells, 8192u);

	while (n < nr_cells)
		n <<= 1;

	return n;
}

Mike Snitzer's avatar
Mike Snitzer committed
145
146
static struct kmem_cache *_cell_cache;

147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
/*
 * @nr_cells should be the number of cells you want in use _concurrently_.
 * Don't confuse it with the number of distinct keys.
 */
static struct bio_prison *prison_create(unsigned nr_cells)
{
	unsigned i;
	uint32_t nr_buckets = calc_nr_buckets(nr_cells);
	size_t len = sizeof(struct bio_prison) +
		(sizeof(struct hlist_head) * nr_buckets);
	struct bio_prison *prison = kmalloc(len, GFP_KERNEL);

	if (!prison)
		return NULL;

	spin_lock_init(&prison->lock);
Mike Snitzer's avatar
Mike Snitzer committed
163
	prison->cell_pool = mempool_create_slab_pool(nr_cells, _cell_cache);
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
	if (!prison->cell_pool) {
		kfree(prison);
		return NULL;
	}

	prison->nr_buckets = nr_buckets;
	prison->hash_mask = nr_buckets - 1;
	prison->cells = (struct hlist_head *) (prison + 1);
	for (i = 0; i < nr_buckets; i++)
		INIT_HLIST_HEAD(prison->cells + i);

	return prison;
}

static void prison_destroy(struct bio_prison *prison)
{
	mempool_destroy(prison->cell_pool);
	kfree(prison);
}

static uint32_t hash_key(struct bio_prison *prison, struct cell_key *key)
{
	const unsigned long BIG_PRIME = 4294967291UL;
	uint64_t hash = key->block * BIG_PRIME;

	return (uint32_t) (hash & prison->hash_mask);
}

static int keys_equal(struct cell_key *lhs, struct cell_key *rhs)
{
	       return (lhs->virtual == rhs->virtual) &&
		       (lhs->dev == rhs->dev) &&
		       (lhs->block == rhs->block);
}

Mike Snitzer's avatar
Mike Snitzer committed
199
200
static struct dm_bio_prison_cell *__search_bucket(struct hlist_head *bucket,
						  struct cell_key *key)
201
{
Mike Snitzer's avatar
Mike Snitzer committed
202
	struct dm_bio_prison_cell *cell;
203
204
205
206
207
208
209
210
211
212
213
214
215
	struct hlist_node *tmp;

	hlist_for_each_entry(cell, tmp, bucket, list)
		if (keys_equal(&cell->key, key))
			return cell;

	return NULL;
}

/*
 * This may block if a new cell needs allocating.  You must ensure that
 * cells will be unlocked even if the calling thread is blocked.
 *
216
 * Returns 1 if the cell was already held, 0 if @inmate is the new holder.
217
218
 */
static int bio_detain(struct bio_prison *prison, struct cell_key *key,
Mike Snitzer's avatar
Mike Snitzer committed
219
		      struct bio *inmate, struct dm_bio_prison_cell **ref)
220
{
221
	int r = 1;
222
223
	unsigned long flags;
	uint32_t hash = hash_key(prison, key);
Mike Snitzer's avatar
Mike Snitzer committed
224
	struct dm_bio_prison_cell *cell, *cell2;
225
226
227
228
229

	BUG_ON(hash > prison->nr_buckets);

	spin_lock_irqsave(&prison->lock, flags);

230
231
232
233
	cell = __search_bucket(prison->cells + hash, key);
	if (cell) {
		bio_list_add(&cell->bios, inmate);
		goto out;
234
235
	}

236
237
238
	/*
	 * Allocate a new cell
	 */
239
	spin_unlock_irqrestore(&prison->lock, flags);
240
241
	cell2 = mempool_alloc(prison->cell_pool, GFP_NOIO);
	spin_lock_irqsave(&prison->lock, flags);
242

243
244
245
246
247
248
	/*
	 * We've been unlocked, so we have to double check that
	 * nobody else has inserted this cell in the meantime.
	 */
	cell = __search_bucket(prison->cells + hash, key);
	if (cell) {
249
		mempool_free(cell2, prison->cell_pool);
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
		bio_list_add(&cell->bios, inmate);
		goto out;
	}

	/*
	 * Use new cell.
	 */
	cell = cell2;

	cell->prison = prison;
	memcpy(&cell->key, key, sizeof(cell->key));
	cell->holder = inmate;
	bio_list_init(&cell->bios);
	hlist_add_head(&cell->list, prison->cells + hash);

	r = 0;

out:
	spin_unlock_irqrestore(&prison->lock, flags);
269
270
271
272
273
274
275
276
277

	*ref = cell;

	return r;
}

/*
 * @inmates must have been initialised prior to this call
 */
Mike Snitzer's avatar
Mike Snitzer committed
278
static void __cell_release(struct dm_bio_prison_cell *cell, struct bio_list *inmates)
279
280
281
282
283
{
	struct bio_prison *prison = cell->prison;

	hlist_del(&cell->list);

284
285
286
287
	if (inmates) {
		bio_list_add(inmates, cell->holder);
		bio_list_merge(inmates, &cell->bios);
	}
288
289
290
291

	mempool_free(cell, prison->cell_pool);
}

Mike Snitzer's avatar
Mike Snitzer committed
292
static void cell_release(struct dm_bio_prison_cell *cell, struct bio_list *bios)
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
{
	unsigned long flags;
	struct bio_prison *prison = cell->prison;

	spin_lock_irqsave(&prison->lock, flags);
	__cell_release(cell, bios);
	spin_unlock_irqrestore(&prison->lock, flags);
}

/*
 * There are a couple of places where we put a bio into a cell briefly
 * before taking it out again.  In these situations we know that no other
 * bio may be in the cell.  This function releases the cell, and also does
 * a sanity check.
 */
Mike Snitzer's avatar
Mike Snitzer committed
308
static void __cell_release_singleton(struct dm_bio_prison_cell *cell, struct bio *bio)
309
310
311
{
	BUG_ON(cell->holder != bio);
	BUG_ON(!bio_list_empty(&cell->bios));
312
313

	__cell_release(cell, NULL);
314
315
}

Mike Snitzer's avatar
Mike Snitzer committed
316
static void cell_release_singleton(struct dm_bio_prison_cell *cell, struct bio *bio)
317
318
{
	unsigned long flags;
319
	struct bio_prison *prison = cell->prison;
320
321

	spin_lock_irqsave(&prison->lock, flags);
322
	__cell_release_singleton(cell, bio);
323
	spin_unlock_irqrestore(&prison->lock, flags);
324
325
326
327
328
}

/*
 * Sometimes we don't want the holder, just the additional bios.
 */
Mike Snitzer's avatar
Mike Snitzer committed
329
330
static void __cell_release_no_holder(struct dm_bio_prison_cell *cell,
				     struct bio_list *inmates)
331
332
333
334
335
336
337
338
339
{
	struct bio_prison *prison = cell->prison;

	hlist_del(&cell->list);
	bio_list_merge(inmates, &cell->bios);

	mempool_free(cell, prison->cell_pool);
}

Mike Snitzer's avatar
Mike Snitzer committed
340
341
static void cell_release_no_holder(struct dm_bio_prison_cell *cell,
				   struct bio_list *inmates)
342
343
344
{
	unsigned long flags;
	struct bio_prison *prison = cell->prison;
345

346
347
348
	spin_lock_irqsave(&prison->lock, flags);
	__cell_release_no_holder(cell, inmates);
	spin_unlock_irqrestore(&prison->lock, flags);
349
350
}

Mike Snitzer's avatar
Mike Snitzer committed
351
static void cell_error(struct dm_bio_prison_cell *cell)
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
{
	struct bio_prison *prison = cell->prison;
	struct bio_list bios;
	struct bio *bio;
	unsigned long flags;

	bio_list_init(&bios);

	spin_lock_irqsave(&prison->lock, flags);
	__cell_release(cell, &bios);
	spin_unlock_irqrestore(&prison->lock, flags);

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

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

/*
 * We use the deferred set to keep track of pending reads to shared blocks.
 * We do this to ensure the new mapping caused by a write isn't performed
 * until these prior reads have completed.  Otherwise the insertion of the
 * new mapping could free the old block that the read bios are mapped to.
 */

struct deferred_set;
struct deferred_entry {
	struct deferred_set *ds;
	unsigned count;
	struct list_head work_items;
};

struct deferred_set {
	spinlock_t lock;
	unsigned current_entry;
	unsigned sweeper;
	struct deferred_entry entries[DEFERRED_SET_SIZE];
};

static void ds_init(struct deferred_set *ds)
{
	int i;

	spin_lock_init(&ds->lock);
	ds->current_entry = 0;
	ds->sweeper = 0;
	for (i = 0; i < DEFERRED_SET_SIZE; i++) {
		ds->entries[i].ds = ds;
		ds->entries[i].count = 0;
		INIT_LIST_HEAD(&ds->entries[i].work_items);
	}
}

static struct deferred_entry *ds_inc(struct deferred_set *ds)
{
	unsigned long flags;
	struct deferred_entry *entry;

	spin_lock_irqsave(&ds->lock, flags);
	entry = ds->entries + ds->current_entry;
	entry->count++;
	spin_unlock_irqrestore(&ds->lock, flags);

	return entry;
}

static unsigned ds_next(unsigned index)
{
	return (index + 1) % DEFERRED_SET_SIZE;
}

static void __sweep(struct deferred_set *ds, struct list_head *head)
{
	while ((ds->sweeper != ds->current_entry) &&
	       !ds->entries[ds->sweeper].count) {
		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
		ds->sweeper = ds_next(ds->sweeper);
	}

	if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)
		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
}

static void ds_dec(struct deferred_entry *entry, struct list_head *head)
{
	unsigned long flags;

	spin_lock_irqsave(&entry->ds->lock, flags);
	BUG_ON(!entry->count);
	--entry->count;
	__sweep(entry->ds, head);
	spin_unlock_irqrestore(&entry->ds->lock, flags);
}

/*
 * Returns 1 if deferred or 0 if no pending items to delay job.
 */
static int ds_add_work(struct deferred_set *ds, struct list_head *work)
{
	int r = 1;
	unsigned long flags;
	unsigned next_entry;

	spin_lock_irqsave(&ds->lock, flags);
	if ((ds->sweeper == ds->current_entry) &&
	    !ds->entries[ds->current_entry].count)
		r = 0;
	else {
		list_add(work, &ds->entries[ds->current_entry].work_items);
		next_entry = ds_next(ds->current_entry);
		if (!ds->entries[next_entry].count)
			ds->current_entry = next_entry;
	}
	spin_unlock_irqrestore(&ds->lock, flags);

	return r;
}

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

/*
 * Key building.
 */
static void build_data_key(struct dm_thin_device *td,
			   dm_block_t b, struct cell_key *key)
{
	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,
			      struct cell_key *key)
{
	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
498
struct dm_thin_new_mapping;
499

500
501
502
503
504
505
506
507
508
/*
 * 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 */
};

509
struct pool_features {
510
511
	enum pool_mode mode;

Mike Snitzer's avatar
Mike Snitzer committed
512
513
514
	bool zero_new_blocks:1;
	bool discard_enabled:1;
	bool discard_passdown:1;
515
516
};

517
518
519
520
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);

521
522
523
524
525
526
527
528
529
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;
530
	uint32_t sectors_per_block;
531
	int sectors_per_block_shift;
532

533
	struct pool_features pf;
534
535
536
537
538
539
540
541
	unsigned low_water_triggered:1;	/* A dm event has been sent */
	unsigned no_free_space:1;	/* A -ENOSPC warning has been issued */

	struct bio_prison *prison;
	struct dm_kcopyd_client *copier;

	struct workqueue_struct *wq;
	struct work_struct worker;
542
	struct delayed_work waker;
543

544
	unsigned long last_commit_jiffies;
545
	unsigned ref_count;
546
547
548
549
550

	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
551
	struct list_head prepared_discards;
552
553
554

	struct bio_list retry_on_resume_list;

555
	struct deferred_set shared_read_ds;
Joe Thornber's avatar
Joe Thornber committed
556
	struct deferred_set all_io_ds;
557

Mike Snitzer's avatar
Mike Snitzer committed
558
	struct dm_thin_new_mapping *next_mapping;
559
560
	mempool_t *mapping_pool;
	mempool_t *endio_hook_pool;
561
562
563
564
565
566

	process_bio_fn process_bio;
	process_bio_fn process_discard;

	process_mapping_fn process_prepared_mapping;
	process_mapping_fn process_prepared_discard;
567
568
};

569
570
571
static enum pool_mode get_pool_mode(struct pool *pool);
static void set_pool_mode(struct pool *pool, enum pool_mode mode);

572
573
574
575
576
577
578
579
580
581
582
/*
 * 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;
583
	struct pool_features pf;
584
585
586
587
588
589
590
};

/*
 * Target context for a thin.
 */
struct thin_c {
	struct dm_dev *pool_dev;
591
	struct dm_dev *origin_dev;
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
	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
660
struct dm_thin_endio_hook {
661
662
	struct thin_c *tc;
	struct deferred_entry *shared_read_entry;
Joe Thornber's avatar
Joe Thornber committed
663
	struct deferred_entry *all_io_entry;
Mike Snitzer's avatar
Mike Snitzer committed
664
	struct dm_thin_new_mapping *overwrite_mapping;
665
666
};

667
668
669
670
671
672
673
674
675
676
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))) {
Mike Snitzer's avatar
Mike Snitzer committed
677
678
		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;

679
		if (h->tc == tc)
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
			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.
 */

static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
{
706
707
	sector_t block_nr = bio->bi_sector;

708
709
710
711
	if (tc->pool->sectors_per_block_shift < 0)
		(void) sector_div(block_nr, tc->pool->sectors_per_block);
	else
		block_nr >>= tc->pool->sectors_per_block_shift;
712
713

	return block_nr;
714
715
716
717
718
}

static void remap(struct thin_c *tc, struct bio *bio, dm_block_t block)
{
	struct pool *pool = tc->pool;
719
	sector_t bi_sector = bio->bi_sector;
720
721

	bio->bi_bdev = tc->pool_dev->bdev;
722
723
724
725
726
727
	if (tc->pool->sectors_per_block_shift < 0)
		bio->bi_sector = (block * pool->sectors_per_block) +
				 sector_div(bi_sector, pool->sectors_per_block);
	else
		bio->bi_sector = (block << pool->sectors_per_block_shift) |
				(bi_sector & (pool->sectors_per_block - 1));
728
729
}

730
731
732
733
734
static void remap_to_origin(struct thin_c *tc, struct bio *bio)
{
	bio->bi_bdev = tc->origin_dev->bdev;
}

735
736
737
738
739
740
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);
}

741
static void issue(struct thin_c *tc, struct bio *bio)
742
743
744
745
{
	struct pool *pool = tc->pool;
	unsigned long flags;

746
747
748
749
750
	if (!bio_triggers_commit(tc, bio)) {
		generic_make_request(bio);
		return;
	}

751
	/*
752
753
754
	 * 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.
755
	 */
756
757
758
759
760
761
762
763
764
765
766
767
	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);
768
769
}

770
771
772
773
774
775
776
777
778
779
780
781
782
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);
}

783
784
785
786
787
788
789
790
791
792
793
794
795
796
/*
 * 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
797
struct dm_thin_new_mapping {
798
799
	struct list_head list;

800
801
	unsigned quiesced:1;
	unsigned prepared:1;
Joe Thornber's avatar
Joe Thornber committed
802
	unsigned pass_discard:1;
803
804
805
806

	struct thin_c *tc;
	dm_block_t virt_block;
	dm_block_t data_block;
Mike Snitzer's avatar
Mike Snitzer committed
807
	struct dm_bio_prison_cell *cell, *cell2;
808
809
810
811
812
813
814
815
816
817
818
819
	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
820
static void __maybe_add_mapping(struct dm_thin_new_mapping *m)
821
822
823
{
	struct pool *pool = m->tc->pool;

824
	if (m->quiesced && m->prepared) {
825
826
827
828
829
830
831
832
		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
833
	struct dm_thin_new_mapping *m = context;
834
835
836
837
838
839
840
841
842
843
844
845
846
	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;
Mike Snitzer's avatar
Mike Snitzer committed
847
848
	struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;
	struct dm_thin_new_mapping *m = h->overwrite_mapping;
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
	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.
 */
Mike Snitzer's avatar
Mike Snitzer committed
872
static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *cell,
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
		       dm_block_t data_block)
{
	struct pool *pool = tc->pool;
	unsigned long flags;

	spin_lock_irqsave(&pool->lock, flags);
	cell_release(cell, &pool->deferred_bios);
	spin_unlock_irqrestore(&tc->pool->lock, flags);

	wake_worker(pool);
}

/*
 * Same as cell_defer above, except it omits one particular detainee,
 * a write bio that covers the block and has already been processed.
 */
Mike Snitzer's avatar
Mike Snitzer committed
889
static void cell_defer_except(struct thin_c *tc, struct dm_bio_prison_cell *cell)
890
891
892
893
894
895
896
897
{
	struct bio_list bios;
	struct pool *pool = tc->pool;
	unsigned long flags;

	bio_list_init(&bios);

	spin_lock_irqsave(&pool->lock, flags);
898
	cell_release_no_holder(cell, &pool->deferred_bios);
899
900
901
902
903
	spin_unlock_irqrestore(&pool->lock, flags);

	wake_worker(pool);
}

904
905
906
907
908
909
910
911
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;
	cell_error(m->cell);
	list_del(&m->list);
	mempool_free(m, m->tc->pool->mapping_pool);
}
Mike Snitzer's avatar
Mike Snitzer committed
912
static void process_prepared_mapping(struct dm_thin_new_mapping *m)
913
914
915
916
917
918
919
920
921
922
923
{
	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) {
		cell_error(m->cell);
924
		goto out;
925
926
927
928
929
930
931
932
933
934
935
	}

	/*
	 * 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) {
		DMERR("dm_thin_insert_block() failed");
		cell_error(m->cell);
936
		goto out;
937
938
939
940
941
942
943
944
945
	}

	/*
	 * 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) {
946
		cell_defer_except(tc, m->cell);
947
948
949
950
		bio_endio(bio, 0);
	} else
		cell_defer(tc, m->cell, m->data_block);

951
out:
952
953
954
955
	list_del(&m->list);
	mempool_free(m, tc->pool->mapping_pool);
}

956
static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
Joe Thornber's avatar
Joe Thornber committed
957
958
959
{
	struct thin_c *tc = m->tc;

960
961
962
963
964
965
966
967
968
	bio_io_error(m->bio);
	cell_defer_except(tc, m->cell);
	cell_defer_except(tc, m->cell2);
	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
969
970
971
972
973
974
975
976
977
978
979

	if (m->pass_discard)
		remap_and_issue(tc, m->bio, m->data_block);
	else
		bio_endio(m->bio, 0);

	cell_defer_except(tc, m->cell);
	cell_defer_except(tc, m->cell2);
	mempool_free(m, tc->pool->mapping_pool);
}

980
981
982
983
984
985
986
987
988
989
990
991
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)
		DMERR("dm_thin_remove_block() failed");

	process_prepared_discard_passdown(m);
}

Joe Thornber's avatar
Joe Thornber committed
992
static void process_prepared(struct pool *pool, struct list_head *head,
993
			     process_mapping_fn *fn)
994
995
996
{
	unsigned long flags;
	struct list_head maps;
Mike Snitzer's avatar
Mike Snitzer committed
997
	struct dm_thin_new_mapping *m, *tmp;
998
999
1000

	INIT_LIST_HEAD(&maps);
	spin_lock_irqsave(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
1001
	list_splice_init(head, &maps);
1002
1003
1004
	spin_unlock_irqrestore(&pool->lock, flags);

	list_for_each_entry_safe(m, tmp, &maps, list)
1005
		(*fn)(m);
1006
1007
1008
1009
1010
}

/*
 * Deferred bio jobs.
 */
Joe Thornber's avatar
Joe Thornber committed
1011
static int io_overlaps_block(struct pool *pool, struct bio *bio)
1012
{
1013
	return bio->bi_size == (pool->sectors_per_block << SECTOR_SHIFT);
Joe Thornber's avatar
Joe Thornber committed
1014
1015
1016
1017
1018
1019
}

static int io_overwrites_block(struct pool *pool, struct bio *bio)
{
	return (bio_data_dir(bio) == WRITE) &&
		io_overlaps_block(pool, bio);
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
}

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
1039
static struct dm_thin_new_mapping *get_next_mapping(struct pool *pool)
1040
{
Mike Snitzer's avatar
Mike Snitzer committed
1041
	struct dm_thin_new_mapping *r = pool->next_mapping;
1042
1043
1044
1045
1046
1047
1048
1049
1050

	BUG_ON(!pool->next_mapping);

	pool->next_mapping = NULL;

	return r;
}

static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
1051
1052
			  struct dm_dev *origin, dm_block_t data_origin,
			  dm_block_t data_dest,
Mike Snitzer's avatar
Mike Snitzer committed
1053
			  struct dm_bio_prison_cell *cell, struct bio *bio)
1054
1055
1056
{
	int r;
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
1057
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
1058
1059

	INIT_LIST_HEAD(&m->list);
1060
	m->quiesced = 0;
1061
1062
1063
1064
1065
1066
1067
1068
	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;

1069
1070
	if (!ds_add_work(&pool->shared_read_ds, &m->list))
		m->quiesced = 1;
1071
1072
1073
1074
1075
1076
1077
1078

	/*
	 * 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)) {
Mike Snitzer's avatar
Mike Snitzer committed
1079
1080
		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;

1081
		h->overwrite_mapping = m;
1082
1083
1084
1085
1086
1087
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
		remap_and_issue(tc, bio, data_dest);
	} else {
		struct dm_io_region from, to;

1088
		from.bdev = origin->bdev;
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
		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);
			DMERR("dm_kcopyd_copy() failed");
			cell_error(cell);
		}
	}
}

1106
1107
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
1108
				   struct dm_bio_prison_cell *cell, struct bio *bio)
1109
1110
1111
1112
1113
1114
1115
{
	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
1116
				   struct dm_bio_prison_cell *cell, struct bio *bio)
1117
1118
1119
1120
1121
{
	schedule_copy(tc, virt_block, tc->origin_dev,
		      virt_block, data_dest, cell, bio);
}

1122
static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
Mike Snitzer's avatar
Mike Snitzer committed
1123
			  dm_block_t data_block, struct dm_bio_prison_cell *cell,
1124
1125
1126
			  struct bio *bio)
{
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
1127
	struct dm_thin_new_mapping *m = get_next_mapping(pool);
1128
1129

	INIT_LIST_HEAD(&m->list);
1130
	m->quiesced = 1;
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
	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.
	 */
1144
	if (!pool->pf.zero_new_blocks)
1145
1146
1147
		process_prepared_mapping(m);

	else if (io_overwrites_block(pool, bio)) {
Mike Snitzer's avatar
Mike Snitzer committed
1148
1149
		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;

1150
		h->overwrite_mapping = m;
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
		m->bio = bio;
		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
		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);
			DMERR("dm_kcopyd_zero() failed");
			cell_error(cell);
		}
	}
}

1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
static int commit(struct pool *pool)
{
	int r;

	r = dm_pool_commit_metadata(pool->pmd);
	if (r)
		DMERR("commit failed, error = %d", r);

	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;
}

1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
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.
			 */
1228
			(void) commit_or_fallback(pool);
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261

			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)
{
Mike Snitzer's avatar
Mike Snitzer committed
1262
	struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;
1263
	struct thin_c *tc = h->tc;
1264
1265
1266
1267
1268
1269
1270
1271
	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
1272
static void no_space(struct dm_bio_prison_cell *cell)
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
{
	struct bio *bio;
	struct bio_list bios;

	bio_list_init(&bios);
	cell_release(cell, &bios);

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

Joe Thornber's avatar
Joe Thornber committed
1284
1285
1286
static void process_discard(struct thin_c *tc, struct bio *bio)
{
	int r;
1287
	unsigned long flags;
Joe Thornber's avatar
Joe Thornber committed
1288
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
1289
	struct dm_bio_prison_cell *cell, *cell2;
Joe Thornber's avatar
Joe Thornber committed
1290
1291
1292
	struct cell_key key, key2;
	dm_block_t block = get_bio_block(tc, bio);
	struct dm_thin_lookup_result lookup_result;
Mike Snitzer's avatar
Mike Snitzer committed
1293
	struct dm_thin_new_mapping *m;
Joe Thornber's avatar
Joe Thornber committed
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319

	build_virtual_key(tc->td, block, &key);
	if (bio_detain(tc->pool->prison, &key, bio, &cell))
		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);
		if (bio_detain(tc->pool->prison, &key2, bio, &cell2)) {
			cell_release_singleton(cell, bio);
			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;
1320
			m->pass_discard = (!lookup_result.shared) && pool->pf.discard_passdown;
Joe Thornber's avatar
Joe Thornber committed
1321
1322
1323
1324
1325
1326
1327
1328
			m->virt_block = block;
			m->data_block = lookup_result.block;
			m->cell = cell;
			m->cell2 = cell2;
			m->err = 0;
			m->bio = bio;

			if (!ds_add_work(&pool->all_io_ds, &m->list)) {
1329
				spin_lock_irqsave(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
1330
				list_add(&m->list, &pool->prepared_discards);
1331
				spin_unlock_irqrestore(&pool->lock, flags);
Joe Thornber's avatar
Joe Thornber committed
1332
1333
1334
1335
				wake_worker(pool);
			}
		} else {
			/*
1336
1337
1338
			 * 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
1339
1340
1341
			 */
			cell_release_singleton(cell, bio);
			cell_release_singleton(cell2, bio);
1342
1343
1344
1345
			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
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
		}
		break;

	case -ENODATA:
		/*
		 * It isn't provisioned, just forget it.
		 */
		cell_release_singleton(cell, bio);
		bio_endio(bio, 0);
		break;

	default:
		DMERR("discard: find block unexpectedly returned %d", r);
		cell_release_singleton(cell, bio);
		bio_io_error(bio);
		break;
	}
}

1365
1366
1367
static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,
			  struct cell_key *key,
			  struct dm_thin_lookup_result *lookup_result,
Mike Snitzer's avatar
Mike Snitzer committed
1368
			  struct dm_bio_prison_cell *cell)
1369
1370
1371
1372
1373
1374
1375
{
	int r;
	dm_block_t data_block;

	r = alloc_data_block(tc, &data_block);
	switch (r) {
	case 0:
1376
1377
		schedule_internal_copy(tc, block, lookup_result->block,
				       data_block, cell, bio);
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
		break;

	case -ENOSPC:
		no_space(cell);
		break;

	default:
		DMERR("%s: alloc_data_block() failed, error = %d", __func__, r);
		cell_error(cell);
		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
1395
	struct dm_bio_prison_cell *cell;
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
	struct pool *pool = tc->pool;
	struct cell_key key;

	/*
	 * 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);
	if (bio_detain(pool->prison, &key, bio, &cell))
		return;

1407
	if (bio_data_dir(bio) == WRITE && bio->bi_size)
1408
1409
		break_sharing(tc, bio, block, &key, lookup_result, cell);
	else {
Mike Snitzer's avatar
Mike Snitzer committed
1410
		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;
1411

1412
		h->shared_read_entry = ds_inc(&pool->shared_read_ds);
1413
1414
1415
1416
1417
1418
1419

		cell_release_singleton(cell, bio);
		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
1420
			    struct dm_bio_prison_cell *cell)
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
{
	int r;
	dm_block_t data_block;

	/*
	 * Remap empty bios (flushes) immediately, without provisioning.
	 */
	if (!bio->bi_size) {
		cell_release_singleton(cell, bio);
		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);
		cell_release_singleton(cell, bio);
		bio_endio(bio, 0);
		return;
	}

	r = alloc_data_block(tc, &data_block);
	switch (r) {
	case 0:
1447
1448
1449
1450
		if (tc->origin_dev)
			schedule_external_copy(tc, block, data_block, cell, bio);
		else
			schedule_zero(tc, block, data_block, cell, bio);
1451
1452
1453
1454
1455
1456
1457
1458
		break;

	case -ENOSPC:
		no_space(cell);
		break;

	default:
		DMERR("%s: alloc_data_block() failed, error = %d", __func__, r);
1459
		set_pool_mode(tc->pool, PM_READ_ONLY);
1460
1461
1462
1463
1464
1465
1466
1467
1468
		cell_error(cell);
		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
1469
	struct dm_bio_prison_cell *cell;
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
	struct cell_key key;
	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);
	if (bio_detain(tc->pool->prison, &key, bio, &cell))
		return;

	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
	switch (r) {
	case 0:
		/*
		 * We can release this cell now.  This thread is the only
		 * one that puts bios into a cell, and we know there were
		 * no preceding bios.
		 */
		/*
		 * TODO: this will probably have to change when discard goes
		 * back in.
		 */
		cell_release_singleton(cell, bio);

		if (lookup_result.shared)
			process_shared_bio(tc, bio, block, &lookup_result);
		else
			remap_and_issue(tc, bio, lookup_result.block);
		break;

	case -ENODATA:
1502
1503
1504
1505
1506
		if (bio_data_dir(bio) == READ && tc->origin_dev) {
			cell_release_singleton(cell, bio);
			remap_to_origin_and_issue(tc, bio);
		} else
			provision_block(tc, bio, block, cell);
1507
1508
1509
1510
		break;

	default:
		DMERR("dm_thin_find_block() failed, error = %d", r);
Joe Thornber's avatar
Joe Thornber committed
1511
		cell_release_singleton(cell, bio);
1512
1513
1514
1515
1516
		bio_io_error(bio);
		break;
	}
}

1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
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);
		else
			remap_and_issue(tc, bio, lookup_result.block);
		break;

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

		if (tc->origin_dev) {
			remap_to_origin_and_issue(tc, bio);
			break;
		}

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

	default:
		DMERR("dm_thin_find_block() failed, error = %d", r);
		bio_io_error(bio);
		break;
	}
}

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

1560
1561
1562
1563
1564
1565
static int need_commit_due_to_time(struct pool *pool)
{
	return jiffies < pool->last_commit_jiffies ||
	       jiffies > pool->last_commit_jiffies + COMMIT_PERIOD;
}

1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
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))) {
Mike Snitzer's avatar
Mike Snitzer committed
1580
		struct dm_thin_endio_hook *h = dm_get_mapinfo(bio)->ptr;
1581
1582
		struct thin_c *tc = h->tc;

1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
		/*
		 * 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
1595
1596

		if (bio->bi_rw & REQ_DISCARD)
1597
			pool->process_discard(tc, bio);
Joe Thornber's avatar
Joe Thornber committed
1598
		else
1599
			pool->process_bio(tc, bio);
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
	}

	/*
	 * 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);

1612
	if (bio_list_empty(&bios) && !need_commit_due_to_time(pool))
1613
1614
		return;

1615
	if (commit_or_fallback(pool)) {
1616
1617
1618
1619
		while ((bio = bio_list_pop(&bios)))
			bio_io_error(bio);
		return;
	}
1620
	pool->last_commit_jiffies = jiffies;
1621
1622
1623
1624
1625
1626
1627
1628
1629

	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);

1630
1631
	process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
	process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
1632
1633
1634
	process_deferred_bios(pool);
}

1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
/*
 * 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);
}

1646
1647
/*----------------------------------------------------------------*/

1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
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;
	}
}

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

1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
/*
 * 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);
}

Mike Snitzer's avatar
Mike Snitzer committed
1713
static struct dm_thin_endio_hook *thin_hook_bio(struct thin_c *tc, struct bio *bio)
1714
1715
{
	struct pool *pool = tc->pool;
Mike Snitzer's avatar
Mike Snitzer committed
1716
	struct dm_thin_endio_hook *h = mempool_alloc(pool->endio_hook_pool, GFP_NOIO);
1717
1718
1719

	h->tc = tc;
	h->shared_read_entry = NULL;
Joe Thornber's avatar
Joe Thornber committed
1720
	h->all_io_entry = bio->bi_rw & REQ_DISCARD ? NULL : ds_inc(&pool->all_io_ds);
1721
1722
1723
1724
1725
	h->overwrite_mapping = NULL;

	return h;
}

1726
1727
1728
1729
1730
1731
1732
1733
1734
1735
1736
1737
/*
 * Non-blocking function called from the thin target's map function.
 */
static int thin_bio_map(struct dm_target *ti, struct bio *bio,
			union map_info *map_context)
{
	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;

1738
	map_context->ptr = thin_hook_bio(tc, bio);
1739
1740
1741
1742
1743
1744

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