dm.c 62.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2
/*
 * Copyright (C) 2001, 2002 Sistina Software (UK) Limited.
M
Milan Broz 已提交
3
 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
L
Linus Torvalds 已提交
4 5 6 7 8
 *
 * This file is released under the GPL.
 */

#include "dm.h"
M
Mike Anderson 已提交
9
#include "dm-uevent.h"
L
Linus Torvalds 已提交
10 11 12

#include <linux/init.h>
#include <linux/module.h>
A
Arjan van de Ven 已提交
13
#include <linux/mutex.h>
L
Linus Torvalds 已提交
14 15 16 17 18 19
#include <linux/moduleparam.h>
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/mempool.h>
#include <linux/slab.h>
#include <linux/idr.h>
D
Darrick J. Wong 已提交
20
#include <linux/hdreg.h>
21
#include <linux/delay.h>
22 23

#include <trace/events/block.h>
L
Linus Torvalds 已提交
24

25 26
#define DM_MSG_PREFIX "core"

N
Namhyung Kim 已提交
27 28 29 30 31 32 33 34 35 36
#ifdef CONFIG_PRINTK
/*
 * ratelimit state to be used in DMXXX_LIMIT().
 */
DEFINE_RATELIMIT_STATE(dm_ratelimit_state,
		       DEFAULT_RATELIMIT_INTERVAL,
		       DEFAULT_RATELIMIT_BURST);
EXPORT_SYMBOL(dm_ratelimit_state);
#endif

M
Milan Broz 已提交
37 38 39 40 41 42 43
/*
 * Cookies are numeric values sent with CHANGE and REMOVE
 * uevents while resuming, removing or renaming the device.
 */
#define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE"
#define DM_COOKIE_LENGTH 24

L
Linus Torvalds 已提交
44 45 46 47 48
static const char *_name = DM_NAME;

static unsigned int major = 0;
static unsigned int _major = 0;

49 50
static DEFINE_IDR(_minor_idr);

51
static DEFINE_SPINLOCK(_minor_lock);
L
Linus Torvalds 已提交
52
/*
K
Kiyoshi Ueda 已提交
53
 * For bio-based dm.
L
Linus Torvalds 已提交
54 55 56 57 58 59
 * One of these is allocated per bio.
 */
struct dm_io {
	struct mapped_device *md;
	int error;
	atomic_t io_count;
R
Richard Kennedy 已提交
60
	struct bio *bio;
61
	unsigned long start_time;
62
	spinlock_t endio_lock;
L
Linus Torvalds 已提交
63 64
};

K
Kiyoshi Ueda 已提交
65 66 67 68 69 70 71 72 73 74 75 76 77
/*
 * For request-based dm.
 * One of these is allocated per request.
 */
struct dm_rq_target_io {
	struct mapped_device *md;
	struct dm_target *ti;
	struct request *orig, clone;
	int error;
	union map_info info;
};

/*
78 79 80 81 82 83
 * For request-based dm - the bio clones we allocate are embedded in these
 * structs.
 *
 * We allocate these with bio_alloc_bioset, using the front_pad parameter when
 * the bioset is created - this means the bio has to come at the end of the
 * struct.
K
Kiyoshi Ueda 已提交
84 85 86
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
87
	struct dm_rq_target_io *tio;
88
	struct bio clone;
K
Kiyoshi Ueda 已提交
89 90
};

L
Linus Torvalds 已提交
91 92
union map_info *dm_get_mapinfo(struct bio *bio)
{
A
Alasdair G Kergon 已提交
93
	if (bio && bio->bi_private)
A
Alasdair G Kergon 已提交
94
		return &((struct dm_target_io *)bio->bi_private)->info;
A
Alasdair G Kergon 已提交
95
	return NULL;
L
Linus Torvalds 已提交
96 97
}

98 99 100 101 102 103 104 105
union map_info *dm_get_rq_mapinfo(struct request *rq)
{
	if (rq && rq->end_io_data)
		return &((struct dm_rq_target_io *)rq->end_io_data)->info;
	return NULL;
}
EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo);

106 107
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
108 109 110
/*
 * Bits for the md->flags field.
 */
111
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
112
#define DMF_SUSPENDED 1
113
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
114
#define DMF_FREEING 3
115
#define DMF_DELETING 4
116
#define DMF_NOFLUSH_SUSPENDING 5
117
#define DMF_MERGE_IS_OPTIONAL 6
L
Linus Torvalds 已提交
118

119 120 121
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
122
struct mapped_device {
123
	struct rw_semaphore io_lock;
124
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
125 126
	rwlock_t map_lock;
	atomic_t holders;
127
	atomic_t open_count;
L
Linus Torvalds 已提交
128 129 130

	unsigned long flags;

131
	struct request_queue *queue;
132
	unsigned type;
133
	/* Protect queue and type against concurrent access. */
134 135
	struct mutex type_lock;

136 137
	struct target_type *immutable_target_type;

L
Linus Torvalds 已提交
138
	struct gendisk *disk;
M
Mike Anderson 已提交
139
	char name[16];
L
Linus Torvalds 已提交
140 141 142 143 144 145

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
146
	atomic_t pending[2];
L
Linus Torvalds 已提交
147
	wait_queue_head_t wait;
148
	struct work_struct work;
K
Kiyoshi Ueda 已提交
149
	struct bio_list deferred;
150
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
151

152
	/*
153
	 * Processing queue (flush)
154 155 156
	 */
	struct workqueue_struct *wq;

L
Linus Torvalds 已提交
157 158 159 160 161 162 163 164 165 166
	/*
	 * The current mapping.
	 */
	struct dm_table *map;

	/*
	 * io objects are allocated from here.
	 */
	mempool_t *io_pool;

S
Stefan Bader 已提交
167 168
	struct bio_set *bs;

L
Linus Torvalds 已提交
169 170 171 172 173
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
174 175 176
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
177 178 179 180 181

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
182
	struct block_device *bdev;
D
Darrick J. Wong 已提交
183 184 185

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
186 187 188

	/* sysfs handle */
	struct kobject kobj;
189

190 191
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
L
Linus Torvalds 已提交
192 193
};

K
Kiyoshi Ueda 已提交
194 195 196 197 198 199 200 201
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
	struct bio_set *bs;
};

L
Linus Torvalds 已提交
202
#define MIN_IOS 256
203
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
204
static struct kmem_cache *_rq_tio_cache;
205

L
Linus Torvalds 已提交
206 207
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
208
	int r = -ENOMEM;
L
Linus Torvalds 已提交
209 210

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
211
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
212
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
213
		return r;
L
Linus Torvalds 已提交
214

K
Kiyoshi Ueda 已提交
215 216
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
217
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
218

M
Mike Anderson 已提交
219
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
220
	if (r)
221
		goto out_free_rq_tio_cache;
M
Mike Anderson 已提交
222

L
Linus Torvalds 已提交
223 224
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
225 226
	if (r < 0)
		goto out_uevent_exit;
L
Linus Torvalds 已提交
227 228 229 230 231

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
232 233 234

out_uevent_exit:
	dm_uevent_exit();
K
Kiyoshi Ueda 已提交
235 236
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
237 238 239 240
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
241 242 243 244
}

static void local_exit(void)
{
K
Kiyoshi Ueda 已提交
245
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
246
	kmem_cache_destroy(_io_cache);
247
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
248
	dm_uevent_exit();
L
Linus Torvalds 已提交
249 250 251 252 253 254

	_major = 0;

	DMINFO("cleaned up");
}

255
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
256 257 258 259
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
260
	dm_io_init,
261
	dm_kcopyd_init,
L
Linus Torvalds 已提交
262 263 264
	dm_interface_init,
};

265
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
266 267 268 269
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
270
	dm_io_exit,
271
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301
	dm_interface_exit,
};

static int __init dm_init(void)
{
	const int count = ARRAY_SIZE(_inits);

	int r, i;

	for (i = 0; i < count; i++) {
		r = _inits[i]();
		if (r)
			goto bad;
	}

	return 0;

      bad:
	while (i--)
		_exits[i]();

	return r;
}

static void __exit dm_exit(void)
{
	int i = ARRAY_SIZE(_exits);

	while (i--)
		_exits[i]();
302 303 304 305 306

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
307 308 309 310 311
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
312 313 314 315 316
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
317
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
318 319 320
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
321 322
	spin_lock(&_minor_lock);

A
Al Viro 已提交
323
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
324 325 326
	if (!md)
		goto out;

327
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
328
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
329 330 331 332
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
333
	dm_get(md);
334
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
335 336 337 338 339

out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
340 341
}

A
Al Viro 已提交
342
static int dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
343
{
A
Al Viro 已提交
344
	struct mapped_device *md = disk->private_data;
345

346 347
	spin_lock(&_minor_lock);

348
	atomic_dec(&md->open_count);
L
Linus Torvalds 已提交
349
	dm_put(md);
350 351

	spin_unlock(&_minor_lock);
352

L
Linus Torvalds 已提交
353 354 355
	return 0;
}

356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379
int dm_open_count(struct mapped_device *md)
{
	return atomic_read(&md->open_count);
}

/*
 * Guarantees nothing is using the device before it's deleted.
 */
int dm_lock_for_deletion(struct mapped_device *md)
{
	int r = 0;

	spin_lock(&_minor_lock);

	if (dm_open_count(md))
		r = -EBUSY;
	else
		set_bit(DMF_DELETING, &md->flags);

	spin_unlock(&_minor_lock);

	return r;
}

D
Darrick J. Wong 已提交
380 381 382 383 384 385 386
static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
{
	struct mapped_device *md = bdev->bd_disk->private_data;

	return dm_get_geometry(md, geo);
}

A
Al Viro 已提交
387
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
388 389
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
390
	struct mapped_device *md = bdev->bd_disk->private_data;
391
	struct dm_table *map = dm_get_live_table(md);
392 393 394 395 396 397 398 399 400 401 402 403
	struct dm_target *tgt;
	int r = -ENOTTY;

	if (!map || !dm_table_get_size(map))
		goto out;

	/* We only support devices that have a single target */
	if (dm_table_get_num_targets(map) != 1)
		goto out;

	tgt = dm_table_get_target(map, 0);

404
	if (dm_suspended_md(md)) {
405 406 407 408 409
		r = -EAGAIN;
		goto out;
	}

	if (tgt->type->ioctl)
410
		r = tgt->type->ioctl(tgt, cmd, arg);
411 412 413 414 415 416 417

out:
	dm_table_put(map);

	return r;
}

A
Alasdair G Kergon 已提交
418
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
419 420 421 422
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
423
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
424 425 426 427
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
428
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
429
{
430
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
431 432
}

K
Kiyoshi Ueda 已提交
433 434
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
435
{
J
Jun'ichi Nomura 已提交
436
	return mempool_alloc(md->io_pool, gfp_mask);
437 438 439 440
}

static void free_rq_tio(struct dm_rq_target_io *tio)
{
J
Jun'ichi Nomura 已提交
441
	mempool_free(tio, tio->md->io_pool);
442 443
}

K
Kiyoshi Ueda 已提交
444 445 446 447 448 449
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

450 451 452
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
T
Tejun Heo 已提交
453
	int cpu;
454
	int rw = bio_data_dir(io->bio);
455 456 457

	io->start_time = jiffies;

T
Tejun Heo 已提交
458 459 460
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
461 462
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
463 464
}

465
static void end_io_acct(struct dm_io *io)
466 467 468 469
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
T
Tejun Heo 已提交
470
	int pending, cpu;
471 472
	int rw = bio_data_dir(bio);

T
Tejun Heo 已提交
473 474 475 476
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration);
	part_stat_unlock();
477

478 479
	/*
	 * After this is decremented the bio must not be touched if it is
480
	 * a flush.
481
	 */
482 483
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
484
	pending += atomic_read(&md->pending[rw^0x1]);
485

486 487 488
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
489 490
}

L
Linus Torvalds 已提交
491 492 493
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
494
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
495
{
496
	unsigned long flags;
L
Linus Torvalds 已提交
497

498
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
499
	bio_list_add(&md->deferred, bio);
500
	spin_unlock_irqrestore(&md->deferred_lock, flags);
501
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
502 503 504 505 506 507 508
}

/*
 * Everyone (including functions in this file), should use this
 * function to access the md->map field, and make sure they call
 * dm_table_put() when finished.
 */
509
struct dm_table *dm_get_live_table(struct mapped_device *md)
L
Linus Torvalds 已提交
510 511
{
	struct dm_table *t;
512
	unsigned long flags;
L
Linus Torvalds 已提交
513

514
	read_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
515 516 517
	t = md->map;
	if (t)
		dm_table_get(t);
518
	read_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
519 520 521 522

	return t;
}

D
Darrick J. Wong 已提交
523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
/*
 * Get the geometry associated with a dm device
 */
int dm_get_geometry(struct mapped_device *md, struct hd_geometry *geo)
{
	*geo = md->geometry;

	return 0;
}

/*
 * Set the geometry of a device.
 */
int dm_set_geometry(struct mapped_device *md, struct hd_geometry *geo)
{
	sector_t sz = (sector_t)geo->cylinders * geo->heads * geo->sectors;

	if (geo->start > sz) {
		DMWARN("Start sector is beyond the geometry limits.");
		return -EINVAL;
	}

	md->geometry = *geo;

	return 0;
}

L
Linus Torvalds 已提交
550 551 552 553 554 555 556 557 558
/*-----------------------------------------------------------------
 * CRUD START:
 *   A more elegant soln is in the works that uses the queue
 *   merge fn, unfortunately there are a couple of changes to
 *   the block layer that I want to make for this.  So in the
 *   interests of getting something for people to use I give
 *   you this clearly demarcated crap.
 *---------------------------------------------------------------*/

559 560 561 562 563
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
564 565 566 567
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
568
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
569
{
570
	unsigned long flags;
571 572 573
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
574 575

	/* Push-back supersedes any I/O errors */
576 577 578 579 580 581
	if (unlikely(error)) {
		spin_lock_irqsave(&io->endio_lock, flags);
		if (!(io->error > 0 && __noflush_suspending(md)))
			io->error = error;
		spin_unlock_irqrestore(&io->endio_lock, flags);
	}
L
Linus Torvalds 已提交
582 583

	if (atomic_dec_and_test(&io->io_count)) {
584 585 586 587
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
588
			spin_lock_irqsave(&md->deferred_lock, flags);
589 590 591
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
592 593
				/* noflush suspend was interrupted. */
				io->error = -EIO;
594
			spin_unlock_irqrestore(&md->deferred_lock, flags);
595 596
		}

597 598
		io_error = io->error;
		bio = io->bio;
599 600 601 602 603
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
604

605
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_size) {
606
			/*
607 608
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
609
			 */
610 611
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
612
		} else {
613 614
			/* done with normal IO or empty flush */
			bio_endio(bio, io_error);
615
		}
L
Linus Torvalds 已提交
616 617 618
	}
}

619
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
620 621
{
	int r = 0;
A
Alasdair G Kergon 已提交
622
	struct dm_target_io *tio = bio->bi_private;
623
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
624
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
625 626 627 628 629 630
	dm_endio_fn endio = tio->ti->type->end_io;

	if (!bio_flagged(bio, BIO_UPTODATE) && !error)
		error = -EIO;

	if (endio) {
M
Mikulas Patocka 已提交
631
		r = endio(tio->ti, bio, error);
632 633 634 635 636
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
637
			error = r;
638 639
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
640
			return;
641 642 643 644
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
645 646
	}

S
Stefan Bader 已提交
647
	free_tio(md, tio);
648
	dec_pending(io, error);
L
Linus Torvalds 已提交
649 650
}

651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705
/*
 * Partial completion handling for request-based dm
 */
static void end_clone_bio(struct bio *clone, int error)
{
	struct dm_rq_clone_bio_info *info = clone->bi_private;
	struct dm_rq_target_io *tio = info->tio;
	struct bio *bio = info->orig;
	unsigned int nr_bytes = info->orig->bi_size;

	bio_put(clone);

	if (tio->error)
		/*
		 * An error has already been detected on the request.
		 * Once error occurred, just let clone->end_io() handle
		 * the remainder.
		 */
		return;
	else if (error) {
		/*
		 * Don't notice the error to the upper layer yet.
		 * The error handling decision is made by the target driver,
		 * when the request is completed.
		 */
		tio->error = error;
		return;
	}

	/*
	 * I/O for the bio successfully completed.
	 * Notice the data completion to the upper layer.
	 */

	/*
	 * bios are processed from the head of the list.
	 * So the completing bio should always be rq->bio.
	 * If it's not, something wrong is happening.
	 */
	if (tio->orig->bio != bio)
		DMERR("bio completion is going in the middle of the request");

	/*
	 * Update the original request.
	 * Do not use blk_end_request() here, because it may complete
	 * the original request before the clone, and break the ordering.
	 */
	blk_update_request(tio->orig, 0, nr_bytes);
}

/*
 * Don't touch any member of the md after calling this function because
 * the md may be freed in dm_put() at the end of this function.
 * Or do dm_get() before calling this function and dm_put() later.
 */
706
static void rq_completed(struct mapped_device *md, int rw, int run_queue)
707
{
708
	atomic_dec(&md->pending[rw]);
709 710

	/* nudge anyone waiting on suspend queue */
711
	if (!md_in_flight(md))
712 713
		wake_up(&md->wait);

714 715 716 717 718 719
	/*
	 * Run this off this callpath, as drivers could invoke end_io while
	 * inside their request_fn (and holding the queue lock). Calling
	 * back into ->request_fn() could deadlock attempting to grab the
	 * queue lock again.
	 */
720
	if (run_queue)
721
		blk_run_queue_async(md->queue);
722 723 724 725 726 727 728

	/*
	 * dm_put() must be at the end of this function. See the comment above
	 */
	dm_put(md);
}

729 730 731 732 733 734 735 736
static void free_rq_clone(struct request *clone)
{
	struct dm_rq_target_io *tio = clone->end_io_data;

	blk_rq_unprep_clone(clone);
	free_rq_tio(tio);
}

K
Kiyoshi Ueda 已提交
737 738 739 740 741 742 743 744 745 746 747
/*
 * Complete the clone and the original request.
 * Must be called without queue lock.
 */
static void dm_end_request(struct request *clone, int error)
{
	int rw = rq_data_dir(clone);
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;

748
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
749 750 751 752 753 754 755 756 757 758 759 760 761
		rq->errors = clone->errors;
		rq->resid_len = clone->resid_len;

		if (rq->sense)
			/*
			 * We are using the sense buffer of the original
			 * request.
			 * So setting the length of the sense data is enough.
			 */
			rq->sense_len = clone->sense_len;
	}

	free_rq_clone(clone);
762 763
	blk_end_request_all(rq, error);
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
764 765
}

766 767 768 769 770 771 772
static void dm_unprep_request(struct request *rq)
{
	struct request *clone = rq->special;

	rq->special = NULL;
	rq->cmd_flags &= ~REQ_DONTPREP;

773
	free_rq_clone(clone);
774 775 776 777 778 779 780
}

/*
 * Requeue the original request of a clone.
 */
void dm_requeue_unmapped_request(struct request *clone)
{
781
	int rw = rq_data_dir(clone);
782 783 784 785 786 787 788 789 790 791 792 793
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;
	struct request_queue *q = rq->q;
	unsigned long flags;

	dm_unprep_request(rq);

	spin_lock_irqsave(q->queue_lock, flags);
	blk_requeue_request(q, rq);
	spin_unlock_irqrestore(q->queue_lock, flags);

794
	rq_completed(md, rw, 0);
795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826
}
EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request);

static void __stop_queue(struct request_queue *q)
{
	blk_stop_queue(q);
}

static void stop_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	__stop_queue(q);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

static void __start_queue(struct request_queue *q)
{
	if (blk_queue_stopped(q))
		blk_start_queue(q);
}

static void start_queue(struct request_queue *q)
{
	unsigned long flags;

	spin_lock_irqsave(q->queue_lock, flags);
	__start_queue(q);
	spin_unlock_irqrestore(q->queue_lock, flags);
}

827
static void dm_done(struct request *clone, int error, bool mapped)
828
{
829
	int r = error;
830
	struct dm_rq_target_io *tio = clone->end_io_data;
831
	dm_request_endio_fn rq_end_io = NULL;
832

833 834 835 836 837 838
	if (tio->ti) {
		rq_end_io = tio->ti->type->rq_end_io;

		if (mapped && rq_end_io)
			r = rq_end_io(tio->ti, clone, error, &tio->info);
	}
839

840
	if (r <= 0)
841
		/* The target wants to complete the I/O */
842 843
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
844 845
		/* The target will handle the I/O */
		return;
846
	else if (r == DM_ENDIO_REQUEUE)
847 848 849
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
850
		DMWARN("unimplemented target endio return value: %d", r);
851 852 853 854
		BUG();
	}
}

855 856 857 858 859 860 861 862 863 864 865 866 867 868 869
/*
 * Request completion handler for request-based dm
 */
static void dm_softirq_done(struct request *rq)
{
	bool mapped = true;
	struct request *clone = rq->completion_data;
	struct dm_rq_target_io *tio = clone->end_io_data;

	if (rq->cmd_flags & REQ_FAILED)
		mapped = false;

	dm_done(clone, tio->error, mapped);
}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923
/*
 * Complete the clone and the original request with the error status
 * through softirq context.
 */
static void dm_complete_request(struct request *clone, int error)
{
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct request *rq = tio->orig;

	tio->error = error;
	rq->completion_data = clone;
	blk_complete_request(rq);
}

/*
 * Complete the not-mapped clone and the original request with the error status
 * through softirq context.
 * Target's rq_end_io() function isn't called.
 * This may be used when the target's map_rq() function fails.
 */
void dm_kill_unmapped_request(struct request *clone, int error)
{
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct request *rq = tio->orig;

	rq->cmd_flags |= REQ_FAILED;
	dm_complete_request(clone, error);
}
EXPORT_SYMBOL_GPL(dm_kill_unmapped_request);

/*
 * Called with the queue lock held
 */
static void end_clone_request(struct request *clone, int error)
{
	/*
	 * For just cleaning up the information of the queue in which
	 * the clone was dispatched.
	 * The clone is *NOT* freed actually here because it is alloced from
	 * dm own mempool and REQ_ALLOCED isn't set in clone->cmd_flags.
	 */
	__blk_put_request(clone->q, clone);

	/*
	 * Actual request completion is done in a softirq context which doesn't
	 * hold the queue lock.  Otherwise, deadlock could occur because:
	 *     - another request may be submitted by the upper level driver
	 *       of the stacking during the completion
	 *     - the submission which requires queue lock may be done
	 *       against this queue
	 */
	dm_complete_request(clone, error);
}

924 925 926 927 928 929 930 931 932 933 934 935
/*
 * Return maximum size of I/O possible at the supplied sector up to the current
 * target boundary.
 */
static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti)
{
	sector_t target_offset = dm_target_offset(ti, sector);

	return ti->len - target_offset;
}

static sector_t max_io_len(sector_t sector, struct dm_target *ti)
L
Linus Torvalds 已提交
936
{
937
	sector_t len = max_io_len_target_boundary(sector, ti);
938
	sector_t offset, max_len;
L
Linus Torvalds 已提交
939 940

	/*
941
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
942
	 */
943 944 945 946 947 948 949 950 951 952
	if (ti->max_io_len) {
		offset = dm_target_offset(ti, sector);
		if (unlikely(ti->max_io_len & (ti->max_io_len - 1)))
			max_len = sector_div(offset, ti->max_io_len);
		else
			max_len = offset & (ti->max_io_len - 1);
		max_len = ti->max_io_len - max_len;

		if (len > max_len)
			len = max_len;
L
Linus Torvalds 已提交
953 954 955 956 957
	}

	return len;
}

958 959 960 961 962 963 964 965 966 967 968 969 970 971 972
int dm_set_target_max_io_len(struct dm_target *ti, sector_t len)
{
	if (len > UINT_MAX) {
		DMERR("Specified maximum size of target IO (%llu) exceeds limit (%u)",
		      (unsigned long long)len, UINT_MAX);
		ti->error = "Maximum size of target IO is too large";
		return -EINVAL;
	}

	ti->max_io_len = (uint32_t) len;

	return 0;
}
EXPORT_SYMBOL_GPL(dm_set_target_max_io_len);

A
Alasdair G Kergon 已提交
973
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
974 975
{
	int r;
976
	sector_t sector;
S
Stefan Bader 已提交
977
	struct mapped_device *md;
978
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
979
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
980 981 982 983 984 985 986 987 988 989

	clone->bi_end_io = clone_endio;
	clone->bi_private = tio;

	/*
	 * Map the clone.  If r == 0 we don't need to do
	 * anything, the target has assumed ownership of
	 * this io.
	 */
	atomic_inc(&tio->io->io_count);
990
	sector = clone->bi_sector;
M
Mikulas Patocka 已提交
991
	r = ti->type->map(ti, clone);
992
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
993
		/* the bio has been remapped so dispatch it */
994

995 996
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
997

L
Linus Torvalds 已提交
998
		generic_make_request(clone);
999 1000
	} else if (r < 0 || r == DM_MAPIO_REQUEUE) {
		/* error the io and bail out, or requeue it if needed */
S
Stefan Bader 已提交
1001 1002 1003
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1004 1005 1006
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019
	}
}

struct clone_info {
	struct mapped_device *md;
	struct dm_table *map;
	struct bio *bio;
	struct dm_io *io;
	sector_t sector;
	sector_t sector_count;
	unsigned short idx;
};

A
Alasdair G Kergon 已提交
1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045
static void bio_setup_sector(struct bio *bio, sector_t sector, sector_t len)
{
	bio->bi_sector = sector;
	bio->bi_size = to_bytes(len);
}

static void bio_setup_bv(struct bio *bio, unsigned short idx, unsigned short bv_count)
{
	bio->bi_idx = idx;
	bio->bi_vcnt = idx + bv_count;
	bio->bi_flags &= ~(1 << BIO_SEG_VALID);
}

static void clone_bio_integrity(struct bio *bio, struct bio *clone,
				unsigned short idx, unsigned len, unsigned offset,
				unsigned trim)
{
	if (!bio_integrity(bio))
		return;

	bio_integrity_clone(clone, bio, GFP_NOIO);

	if (trim)
		bio_integrity_trim(clone, bio_sector_offset(bio, idx, offset), len);
}

L
Linus Torvalds 已提交
1046
/*
1047
 * Creates a little bio that just does part of a bvec.
L
Linus Torvalds 已提交
1048
 */
1049 1050 1051
static void clone_split_bio(struct dm_target_io *tio, struct bio *bio,
			    sector_t sector, unsigned short idx,
			    unsigned offset, unsigned len)
L
Linus Torvalds 已提交
1052
{
1053
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1054 1055 1056 1057
	struct bio_vec *bv = bio->bi_io_vec + idx;

	*clone->bi_io_vec = *bv;

A
Alasdair G Kergon 已提交
1058 1059
	bio_setup_sector(clone, sector, len);

L
Linus Torvalds 已提交
1060
	clone->bi_bdev = bio->bi_bdev;
1061
	clone->bi_rw = bio->bi_rw;
L
Linus Torvalds 已提交
1062 1063 1064
	clone->bi_vcnt = 1;
	clone->bi_io_vec->bv_offset = offset;
	clone->bi_io_vec->bv_len = clone->bi_size;
M
Martin K. Petersen 已提交
1065
	clone->bi_flags |= 1 << BIO_CLONED;
L
Linus Torvalds 已提交
1066

A
Alasdair G Kergon 已提交
1067
	clone_bio_integrity(bio, clone, idx, len, offset, 1);
L
Linus Torvalds 已提交
1068 1069 1070 1071 1072
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1073 1074
static void clone_bio(struct dm_target_io *tio, struct bio *bio,
		      sector_t sector, unsigned short idx,
A
Alasdair G Kergon 已提交
1075
		      unsigned short bv_count, unsigned len)
L
Linus Torvalds 已提交
1076
{
1077
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1078
	unsigned trim = 0;
L
Linus Torvalds 已提交
1079

S
Stefan Bader 已提交
1080
	__bio_clone(clone, bio);
A
Alasdair G Kergon 已提交
1081 1082 1083 1084 1085 1086
	bio_setup_sector(clone, sector, len);
	bio_setup_bv(clone, idx, bv_count);

	if (idx != bio->bi_idx || clone->bi_size < bio->bi_size)
		trim = 1;
	clone_bio_integrity(bio, clone, idx, len, 0, trim);
L
Linus Torvalds 已提交
1087 1088
}

1089
static struct dm_target_io *alloc_tio(struct clone_info *ci,
A
Alasdair G Kergon 已提交
1090
				      struct dm_target *ti, int nr_iovecs,
1091
				      unsigned target_bio_nr)
1092
{
1093 1094 1095 1096 1097
	struct dm_target_io *tio;
	struct bio *clone;

	clone = bio_alloc_bioset(GFP_NOIO, nr_iovecs, ci->md->bs);
	tio = container_of(clone, struct dm_target_io, clone);
1098 1099 1100 1101

	tio->io = ci->io;
	tio->ti = ti;
	memset(&tio->info, 0, sizeof(tio->info));
1102
	tio->target_bio_nr = target_bio_nr;
1103 1104 1105 1106

	return tio;
}

1107 1108 1109
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
				       unsigned target_bio_nr, sector_t len)
1110
{
1111
	struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs, target_bio_nr);
1112
	struct bio *clone = &tio->clone;
1113

1114 1115 1116 1117 1118
	/*
	 * Discard requests require the bio's inline iovecs be initialized.
	 * ci->bio->bi_max_vecs is BIO_INLINE_VECS anyway, for both flush
	 * and discard, so no need for concern about wasted bvec allocations.
	 */
1119
	 __bio_clone(clone, ci->bio);
A
Alasdair G Kergon 已提交
1120 1121
	if (len)
		bio_setup_sector(clone, ci->sector, len);
1122

A
Alasdair G Kergon 已提交
1123
	__map_bio(tio);
1124 1125
}

1126 1127
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
				  unsigned num_bios, sector_t len)
1128
{
1129
	unsigned target_bio_nr;
1130

1131
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1132
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1133 1134
}

1135
static int __send_empty_flush(struct clone_info *ci)
1136
{
1137
	unsigned target_nr = 0;
1138 1139
	struct dm_target *ti;

1140
	BUG_ON(bio_has_data(ci->bio));
1141
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1142
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, 0);
1143 1144 1145 1146

	return 0;
}

A
Alasdair G Kergon 已提交
1147 1148 1149 1150 1151
static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
				     sector_t sector, int nr_iovecs,
				     unsigned short idx, unsigned short bv_count,
				     unsigned offset, unsigned len,
				     unsigned split_bvec)
M
Mike Snitzer 已提交
1152
{
1153
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1154
	struct dm_target_io *tio;
1155 1156
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
M
Mike Snitzer 已提交
1157

1158 1159 1160 1161 1162
	/*
	 * Does the target want to receive duplicate copies of the bio?
	 */
	if (bio_data_dir(bio) == WRITE && ti->num_write_bios)
		num_target_bios = ti->num_write_bios(ti, bio);
A
Alasdair G Kergon 已提交
1163

1164 1165 1166 1167 1168 1169 1170 1171
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
		tio = alloc_tio(ci, ti, nr_iovecs, target_bio_nr);
		if (split_bvec)
			clone_split_bio(tio, bio, sector, idx, offset, len);
		else
			clone_bio(tio, bio, sector, idx, bv_count, len);
		__map_bio(tio);
	}
M
Mike Snitzer 已提交
1172 1173
}

1174
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1175

1176
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1177
{
1178
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1179 1180
}

1181
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1182
{
1183
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1184 1185 1186 1187 1188 1189
}

typedef bool (*is_split_required_fn)(struct dm_target *ti);

static bool is_split_required_for_discard(struct dm_target *ti)
{
1190
	return ti->split_discard_bios;
M
Mike Snitzer 已提交
1191 1192
}

1193 1194 1195
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
M
Mike Snitzer 已提交
1196 1197
{
	struct dm_target *ti;
1198
	sector_t len;
1199
	unsigned num_bios;
M
Mike Snitzer 已提交
1200

1201 1202 1203 1204
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
M
Mike Snitzer 已提交
1205 1206

		/*
M
Mike Snitzer 已提交
1207 1208
		 * Even though the device advertised support for this type of
		 * request, that does not mean every target supports it, and
M
Mike Snitzer 已提交
1209
		 * reconfiguration might also have changed that since the
1210
		 * check was performed.
M
Mike Snitzer 已提交
1211
		 */
1212 1213
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1214
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1215

M
Mike Snitzer 已提交
1216
		if (is_split_required && !is_split_required(ti))
1217 1218 1219
			len = min(ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
		else
			len = min(ci->sector_count, max_io_len(ci->sector, ti));
1220

1221
		__send_duplicate_bios(ci, ti, num_bios, len);
1222 1223 1224

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1225 1226 1227 1228

	return 0;
}

1229
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1230
{
1231 1232
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1233 1234
}

1235
static int __send_write_same(struct clone_info *ci)
M
Mike Snitzer 已提交
1236
{
1237
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
M
Mike Snitzer 已提交
1238 1239
}

A
Alasdair G Kergon 已提交
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296
/*
 * Find maximum number of sectors / bvecs we can process with a single bio.
 */
static sector_t __len_within_target(struct clone_info *ci, sector_t max, int *idx)
{
	struct bio *bio = ci->bio;
	sector_t bv_len, total_len = 0;

	for (*idx = ci->idx; max && (*idx < bio->bi_vcnt); (*idx)++) {
		bv_len = to_sector(bio->bi_io_vec[*idx].bv_len);

		if (bv_len > max)
			break;

		max -= bv_len;
		total_len += bv_len;
	}

	return total_len;
}

static int __split_bvec_across_targets(struct clone_info *ci,
				       struct dm_target *ti, sector_t max)
{
	struct bio *bio = ci->bio;
	struct bio_vec *bv = bio->bi_io_vec + ci->idx;
	sector_t remaining = to_sector(bv->bv_len);
	unsigned offset = 0;
	sector_t len;

	do {
		if (offset) {
			ti = dm_table_find_target(ci->map, ci->sector);
			if (!dm_target_is_valid(ti))
				return -EIO;

			max = max_io_len(ci->sector, ti);
		}

		len = min(remaining, max);

		__clone_and_map_data_bio(ci, ti, ci->sector, 1, ci->idx, 0,
					 bv->bv_offset + offset, len, 1);

		ci->sector += len;
		ci->sector_count -= len;
		offset += to_bytes(len);
	} while (remaining -= len);

	ci->idx++;

	return 0;
}

/*
 * Select the correct strategy for processing a non-flush bio.
 */
1297
static int __split_and_process_non_flush(struct clone_info *ci)
L
Linus Torvalds 已提交
1298
{
1299
	struct bio *bio = ci->bio;
1300
	struct dm_target *ti;
A
Alasdair G Kergon 已提交
1301 1302
	sector_t len, max;
	int idx;
L
Linus Torvalds 已提交
1303

M
Mike Snitzer 已提交
1304
	if (unlikely(bio->bi_rw & REQ_DISCARD))
1305
		return __send_discard(ci);
M
Mike Snitzer 已提交
1306
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
1307
		return __send_write_same(ci);
M
Mike Snitzer 已提交
1308

1309 1310 1311 1312
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1313
	max = max_io_len(ci->sector, ti);
1314

A
Alasdair G Kergon 已提交
1315 1316 1317 1318
	/*
	 * Optimise for the simple case where we can do all of
	 * the remaining io with a single clone.
	 */
L
Linus Torvalds 已提交
1319
	if (ci->sector_count <= max) {
A
Alasdair G Kergon 已提交
1320 1321 1322 1323 1324 1325
		__clone_and_map_data_bio(ci, ti, ci->sector, bio->bi_max_vecs,
					 ci->idx, bio->bi_vcnt - ci->idx, 0,
					 ci->sector_count, 0);
		ci->sector_count = 0;
		return 0;
	}
L
Linus Torvalds 已提交
1326

A
Alasdair G Kergon 已提交
1327 1328 1329 1330 1331 1332
	/*
	 * There are some bvecs that don't span targets.
	 * Do as many of these as possible.
	 */
	if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) {
		len = __len_within_target(ci, max, &idx);
L
Linus Torvalds 已提交
1333

A
Alasdair G Kergon 已提交
1334 1335
		__clone_and_map_data_bio(ci, ti, ci->sector, bio->bi_max_vecs,
					 ci->idx, idx - ci->idx, 0, len, 0);
L
Linus Torvalds 已提交
1336 1337 1338

		ci->sector += len;
		ci->sector_count -= len;
A
Alasdair G Kergon 已提交
1339
		ci->idx = idx;
1340

A
Alasdair G Kergon 已提交
1341
		return 0;
L
Linus Torvalds 已提交
1342
	}
1343

A
Alasdair G Kergon 已提交
1344 1345 1346 1347
	/*
	 * Handle a bvec that must be split between two or more targets.
	 */
	return __split_bvec_across_targets(ci, ti, max);
L
Linus Torvalds 已提交
1348 1349 1350
}

/*
1351
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1352
 */
1353
static void __split_and_process_bio(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
1354 1355
{
	struct clone_info ci;
1356
	int error = 0;
L
Linus Torvalds 已提交
1357

1358
	ci.map = dm_get_live_table(md);
1359
	if (unlikely(!ci.map)) {
1360
		bio_io_error(bio);
1361 1362
		return;
	}
1363

L
Linus Torvalds 已提交
1364 1365 1366 1367 1368 1369
	ci.md = md;
	ci.io = alloc_io(md);
	ci.io->error = 0;
	atomic_set(&ci.io->io_count, 1);
	ci.io->bio = bio;
	ci.io->md = md;
1370
	spin_lock_init(&ci.io->endio_lock);
L
Linus Torvalds 已提交
1371 1372 1373
	ci.sector = bio->bi_sector;
	ci.idx = bio->bi_idx;

1374
	start_io_acct(ci.io);
A
Alasdair G Kergon 已提交
1375

1376 1377 1378
	if (bio->bi_rw & REQ_FLUSH) {
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1379
		error = __send_empty_flush(&ci);
1380 1381
		/* dec_pending submits any data associated with flush */
	} else {
1382
		ci.bio = bio;
1383
		ci.sector_count = bio_sectors(bio);
1384
		while (ci.sector_count && !error)
1385
			error = __split_and_process_non_flush(&ci);
1386
	}
L
Linus Torvalds 已提交
1387 1388

	/* drop the extra reference count */
1389
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1390 1391 1392 1393 1394 1395
	dm_table_put(ci.map);
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1396 1397 1398 1399 1400
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
1401
	struct dm_table *map = dm_get_live_table(md);
M
Milan Broz 已提交
1402 1403
	struct dm_target *ti;
	sector_t max_sectors;
1404
	int max_size = 0;
M
Milan Broz 已提交
1405 1406

	if (unlikely(!map))
1407
		goto out;
M
Milan Broz 已提交
1408 1409

	ti = dm_table_find_target(map, bvm->bi_sector);
1410 1411
	if (!dm_target_is_valid(ti))
		goto out_table;
M
Milan Broz 已提交
1412 1413 1414 1415

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1416
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
M
Milan Broz 已提交
1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428
			  (sector_t) BIO_MAX_SECTORS);
	max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size;
	if (max_size < 0)
		max_size = 0;

	/*
	 * merge_bvec_fn() returns number of bytes
	 * it can accept at this offset
	 * max is precomputed maximal io size
	 */
	if (max_size && ti->type->merge)
		max_size = ti->type->merge(ti, bvm, biovec, max_size);
1429 1430 1431 1432 1433 1434 1435 1436 1437 1438
	/*
	 * If the target doesn't support merge method and some of the devices
	 * provided their merge_bvec method (we know this by looking at
	 * queue_max_hw_sectors), then we can't allow bios with multiple vector
	 * entries.  So always set max_size to 0, and the code below allows
	 * just one page.
	 */
	else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9)

		max_size = 0;
M
Milan Broz 已提交
1439

1440
out_table:
1441 1442 1443
	dm_table_put(map);

out:
M
Milan Broz 已提交
1444 1445 1446 1447 1448 1449 1450 1451 1452
	/*
	 * Always allow an entire first page
	 */
	if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT))
		max_size = biovec->bv_len;

	return max_size;
}

L
Linus Torvalds 已提交
1453 1454 1455 1456
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1457
static void _dm_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1458
{
1459
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1460
	struct mapped_device *md = q->queuedata;
T
Tejun Heo 已提交
1461
	int cpu;
L
Linus Torvalds 已提交
1462

1463
	down_read(&md->io_lock);
L
Linus Torvalds 已提交
1464

T
Tejun Heo 已提交
1465 1466 1467 1468
	cpu = part_stat_lock();
	part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]);
	part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio));
	part_stat_unlock();
1469

1470 1471
	/* if we're suspended, we have to queue this io for later */
	if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
1472
		up_read(&md->io_lock);
L
Linus Torvalds 已提交
1473

1474 1475 1476
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1477
			bio_io_error(bio);
1478
		return;
L
Linus Torvalds 已提交
1479 1480
	}

1481
	__split_and_process_bio(md, bio);
1482
	up_read(&md->io_lock);
1483
	return;
1484 1485 1486 1487 1488 1489 1490
}

static int dm_request_based(struct mapped_device *md)
{
	return blk_queue_stackable(md->queue);
}

1491
static void dm_request(struct request_queue *q, struct bio *bio)
1492 1493 1494 1495
{
	struct mapped_device *md = q->queuedata;

	if (dm_request_based(md))
1496 1497 1498
		blk_queue_bio(q, bio);
	else
		_dm_request(q, bio);
1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
}

void dm_dispatch_request(struct request *rq)
{
	int r;

	if (blk_queue_io_stat(rq->q))
		rq->cmd_flags |= REQ_IO_STAT;

	rq->start_time = jiffies;
	r = blk_insert_cloned_request(rq->q, rq);
	if (r)
		dm_complete_request(rq, r);
}
EXPORT_SYMBOL_GPL(dm_dispatch_request);

static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
				 void *data)
{
	struct dm_rq_target_io *tio = data;
1519 1520
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);
1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532

	info->orig = bio_orig;
	info->tio = tio;
	bio->bi_end_io = end_clone_bio;
	bio->bi_private = info;

	return 0;
}

static int setup_clone(struct request *clone, struct request *rq,
		       struct dm_rq_target_io *tio)
{
1533
	int r;
1534

1535 1536 1537 1538
	r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;
1539

1540 1541 1542 1543
	clone->cmd = rq->cmd;
	clone->cmd_len = rq->cmd_len;
	clone->sense = rq->sense;
	clone->buffer = rq->buffer;
1544 1545 1546 1547 1548 1549
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	return 0;
}

K
Kiyoshi Ueda 已提交
1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575
static struct request *clone_rq(struct request *rq, struct mapped_device *md,
				gfp_t gfp_mask)
{
	struct request *clone;
	struct dm_rq_target_io *tio;

	tio = alloc_rq_tio(md, gfp_mask);
	if (!tio)
		return NULL;

	tio->md = md;
	tio->ti = NULL;
	tio->orig = rq;
	tio->error = 0;
	memset(&tio->info, 0, sizeof(tio->info));

	clone = &tio->clone;
	if (setup_clone(clone, rq, tio)) {
		/* -ENOMEM */
		free_rq_tio(tio);
		return NULL;
	}

	return clone;
}

1576 1577 1578 1579 1580 1581 1582 1583 1584 1585 1586 1587 1588
/*
 * Called with the queue lock held.
 */
static int dm_prep_fn(struct request_queue *q, struct request *rq)
{
	struct mapped_device *md = q->queuedata;
	struct request *clone;

	if (unlikely(rq->special)) {
		DMWARN("Already has something in rq->special.");
		return BLKPREP_KILL;
	}

K
Kiyoshi Ueda 已提交
1589 1590
	clone = clone_rq(rq, md, GFP_ATOMIC);
	if (!clone)
1591 1592 1593 1594 1595 1596 1597 1598
		return BLKPREP_DEFER;

	rq->special = clone;
	rq->cmd_flags |= REQ_DONTPREP;

	return BLKPREP_OK;
}

1599 1600 1601 1602 1603 1604 1605
/*
 * Returns:
 * 0  : the request has been processed (not requeued)
 * !0 : the request has been requeued
 */
static int map_request(struct dm_target *ti, struct request *clone,
		       struct mapped_device *md)
1606
{
1607
	int r, requeued = 0;
1608 1609 1610 1611 1612 1613 1614 1615 1616 1617
	struct dm_rq_target_io *tio = clone->end_io_data;

	tio->ti = ti;
	r = ti->type->map_rq(ti, clone, &tio->info);
	switch (r) {
	case DM_MAPIO_SUBMITTED:
		/* The target has taken the I/O to submit by itself later */
		break;
	case DM_MAPIO_REMAPPED:
		/* The target has remapped the I/O so dispatch it */
1618 1619
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(tio->orig));
1620 1621 1622 1623 1624
		dm_dispatch_request(clone);
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
1625
		requeued = 1;
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636
		break;
	default:
		if (r > 0) {
			DMWARN("unimplemented target map return value: %d", r);
			BUG();
		}

		/* The target wants to complete the I/O */
		dm_kill_unmapped_request(clone, r);
		break;
	}
1637 1638

	return requeued;
1639 1640
}

1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 1654 1655 1656 1657 1658 1659 1660
static struct request *dm_start_request(struct mapped_device *md, struct request *orig)
{
	struct request *clone;

	blk_start_request(orig);
	clone = orig->special;
	atomic_inc(&md->pending[rq_data_dir(clone)]);

	/*
	 * Hold the md reference here for the in-flight I/O.
	 * We can't rely on the reference count by device opener,
	 * because the device may be closed during the request completion
	 * when all bios are completed.
	 * See the comment in rq_completed() too.
	 */
	dm_get(md);

	return clone;
}

1661 1662 1663 1664 1665 1666 1667
/*
 * q->request_fn for request-based dm.
 * Called with the queue lock held.
 */
static void dm_request_fn(struct request_queue *q)
{
	struct mapped_device *md = q->queuedata;
1668
	struct dm_table *map = dm_get_live_table(md);
1669
	struct dm_target *ti;
1670
	struct request *rq, *clone;
1671
	sector_t pos;
1672 1673

	/*
1674 1675 1676 1677
	 * For suspend, check blk_queue_stopped() and increment
	 * ->pending within a single queue_lock not to increment the
	 * number of in-flight I/Os after the queue is stopped in
	 * dm_suspend().
1678
	 */
J
Jens Axboe 已提交
1679
	while (!blk_queue_stopped(q)) {
1680 1681
		rq = blk_peek_request(q);
		if (!rq)
J
Jens Axboe 已提交
1682
			goto delay_and_out;
1683

1684 1685 1686 1687 1688 1689
		/* always use block 0 to find the target for flushes for now */
		pos = 0;
		if (!(rq->cmd_flags & REQ_FLUSH))
			pos = blk_rq_pos(rq);

		ti = dm_table_find_target(map, pos);
1690 1691 1692 1693 1694 1695 1696 1697 1698 1699
		if (!dm_target_is_valid(ti)) {
			/*
			 * Must perform setup, that dm_done() requires,
			 * before calling dm_kill_unmapped_request
			 */
			DMERR_LIMIT("request attempted access beyond the end of device");
			clone = dm_start_request(md, rq);
			dm_kill_unmapped_request(clone, -EIO);
			continue;
		}
1700

1701
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
1702
			goto delay_and_out;
1703

1704
		clone = dm_start_request(md, rq);
1705

1706
		spin_unlock(q->queue_lock);
1707 1708 1709
		if (map_request(ti, clone, md))
			goto requeued;

1710 1711
		BUG_ON(!irqs_disabled());
		spin_lock(q->queue_lock);
1712 1713 1714 1715
	}

	goto out;

1716
requeued:
1717 1718
	BUG_ON(!irqs_disabled());
	spin_lock(q->queue_lock);
1719

J
Jens Axboe 已提交
1720 1721
delay_and_out:
	blk_delay_queue(q, HZ / 10);
1722 1723 1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735
out:
	dm_table_put(map);
}

int dm_underlying_device_busy(struct request_queue *q)
{
	return blk_lld_busy(q);
}
EXPORT_SYMBOL_GPL(dm_underlying_device_busy);

static int dm_lld_busy(struct request_queue *q)
{
	int r;
	struct mapped_device *md = q->queuedata;
1736
	struct dm_table *map = dm_get_live_table(md);
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746 1747

	if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))
		r = 1;
	else
		r = dm_table_any_busy_target(map);

	dm_table_put(map);

	return r;
}

L
Linus Torvalds 已提交
1748 1749
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1750 1751 1752
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1753

1754
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
1755
		map = dm_get_live_table(md);
1756
		if (map) {
1757 1758 1759 1760 1761 1762 1763 1764 1765 1766
			/*
			 * Request-based dm cares about only own queue for
			 * the query about congestion status of request_queue
			 */
			if (dm_request_based(md))
				r = md->queue->backing_dev_info.state &
				    bdi_bits;
			else
				r = dm_table_any_congested(map, bdi_bits);

1767 1768 1769 1770
			dm_table_put(map);
		}
	}

L
Linus Torvalds 已提交
1771 1772 1773 1774 1775 1776
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1777
static void free_minor(int minor)
L
Linus Torvalds 已提交
1778
{
1779
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1780
	idr_remove(&_minor_idr, minor);
1781
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1782 1783 1784 1785 1786
}

/*
 * See if the device with a specific minor # is free.
 */
1787
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1788
{
T
Tejun Heo 已提交
1789
	int r;
L
Linus Torvalds 已提交
1790 1791 1792 1793

	if (minor >= (1 << MINORBITS))
		return -EINVAL;

T
Tejun Heo 已提交
1794
	idr_preload(GFP_KERNEL);
1795
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1796

T
Tejun Heo 已提交
1797
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1798

1799
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1800 1801 1802 1803
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1804 1805
}

1806
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1807
{
T
Tejun Heo 已提交
1808
	int r;
J
Jeff Mahoney 已提交
1809

T
Tejun Heo 已提交
1810
	idr_preload(GFP_KERNEL);
1811
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1812

T
Tejun Heo 已提交
1813
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1814

1815
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1816 1817 1818 1819 1820
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1821 1822
}

1823
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1824

1825 1826
static void dm_wq_work(struct work_struct *work);

1827 1828 1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847
static void dm_init_md_queue(struct mapped_device *md)
{
	/*
	 * Request-based dm devices cannot be stacked on top of bio-based dm
	 * devices.  The type of this dm device has not been decided yet.
	 * The type is decided at the first table loading time.
	 * To prevent problematic device stacking, clear the queue flag
	 * for request stacking support until then.
	 *
	 * This queue is new, so no concurrency on the queue_flags.
	 */
	queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue);

	md->queue->queuedata = md;
	md->queue->backing_dev_info.congested_fn = dm_any_congested;
	md->queue->backing_dev_info.congested_data = md;
	blk_queue_make_request(md->queue, dm_request);
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
	blk_queue_merge_bvec(md->queue, dm_merge_bvec);
}

L
Linus Torvalds 已提交
1848 1849 1850
/*
 * Allocate and initialise a blank device with a given minor.
 */
1851
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1852 1853
{
	int r;
1854
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
1855
	void *old_md;
L
Linus Torvalds 已提交
1856 1857 1858 1859 1860 1861

	if (!md) {
		DMWARN("unable to allocate device, out of memory.");
		return NULL;
	}

1862
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1863
		goto bad_module_get;
1864

L
Linus Torvalds 已提交
1865
	/* get a minor number for the dev */
1866
	if (minor == DM_ANY_MINOR)
1867
		r = next_free_minor(&minor);
1868
	else
1869
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1870
	if (r < 0)
M
Milan Broz 已提交
1871
		goto bad_minor;
L
Linus Torvalds 已提交
1872

1873
	md->type = DM_TYPE_NONE;
1874
	init_rwsem(&md->io_lock);
1875
	mutex_init(&md->suspend_lock);
1876
	mutex_init(&md->type_lock);
1877
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1878 1879
	rwlock_init(&md->map_lock);
	atomic_set(&md->holders, 1);
1880
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1881
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1882 1883 1884
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1885

1886
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
1887
	if (!md->queue)
M
Milan Broz 已提交
1888
		goto bad_queue;
L
Linus Torvalds 已提交
1889

1890
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1891

L
Linus Torvalds 已提交
1892 1893
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
1894
		goto bad_disk;
L
Linus Torvalds 已提交
1895

1896 1897
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1898
	init_waitqueue_head(&md->wait);
1899
	INIT_WORK(&md->work, dm_wq_work);
1900 1901
	init_waitqueue_head(&md->eventq);

L
Linus Torvalds 已提交
1902 1903 1904 1905 1906 1907 1908
	md->disk->major = _major;
	md->disk->first_minor = minor;
	md->disk->fops = &dm_blk_dops;
	md->disk->queue = md->queue;
	md->disk->private_data = md;
	sprintf(md->disk->disk_name, "dm-%d", minor);
	add_disk(md->disk);
M
Mike Anderson 已提交
1909
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1910

1911 1912
	md->wq = alloc_workqueue("kdmflush",
				 WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0);
1913 1914 1915
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
1916 1917 1918 1919
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

1920 1921 1922 1923
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
	md->flush_bio.bi_rw = WRITE_FLUSH;

1924
	/* Populate the mapping, nobody knows we exist yet */
1925
	spin_lock(&_minor_lock);
1926
	old_md = idr_replace(&_minor_idr, md, minor);
1927
	spin_unlock(&_minor_lock);
1928 1929 1930

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1931 1932
	return md;

M
Mikulas Patocka 已提交
1933 1934
bad_bdev:
	destroy_workqueue(md->wq);
1935
bad_thread:
1936
	del_gendisk(md->disk);
1937
	put_disk(md->disk);
M
Milan Broz 已提交
1938
bad_disk:
1939
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
1940
bad_queue:
L
Linus Torvalds 已提交
1941
	free_minor(minor);
M
Milan Broz 已提交
1942
bad_minor:
1943
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1944
bad_module_get:
L
Linus Torvalds 已提交
1945 1946 1947 1948
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1949 1950
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1951 1952
static void free_dev(struct mapped_device *md)
{
1953
	int minor = MINOR(disk_devt(md->disk));
1954

M
Mikulas Patocka 已提交
1955 1956
	unlock_fs(md);
	bdput(md->bdev);
1957
	destroy_workqueue(md->wq);
K
Kiyoshi Ueda 已提交
1958 1959 1960 1961
	if (md->io_pool)
		mempool_destroy(md->io_pool);
	if (md->bs)
		bioset_free(md->bs);
M
Martin K. Petersen 已提交
1962
	blk_integrity_unregister(md->disk);
L
Linus Torvalds 已提交
1963
	del_gendisk(md->disk);
1964
	free_minor(minor);
J
Jeff Mahoney 已提交
1965 1966 1967 1968 1969

	spin_lock(&_minor_lock);
	md->disk->private_data = NULL;
	spin_unlock(&_minor_lock);

L
Linus Torvalds 已提交
1970
	put_disk(md->disk);
1971
	blk_cleanup_queue(md->queue);
1972
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1973 1974 1975
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1976 1977
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1978
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1979

J
Jun'ichi Nomura 已提交
1980
	if (md->io_pool && md->bs) {
1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999
		/* The md already has necessary mempools. */
		if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) {
			/*
			 * Reload bioset because front_pad may have changed
			 * because a different table was loaded.
			 */
			bioset_free(md->bs);
			md->bs = p->bs;
			p->bs = NULL;
		} else if (dm_table_get_type(t) == DM_TYPE_REQUEST_BASED) {
			/*
			 * There's no need to reload with request-based dm
			 * because the size of front_pad doesn't change.
			 * Note for future: If you are to reload bioset,
			 * prep-ed requests in the queue may refer
			 * to bio from the old bioset, so you must walk
			 * through the queue to unprep.
			 */
		}
K
Kiyoshi Ueda 已提交
2000
		goto out;
M
Mikulas Patocka 已提交
2001
	}
K
Kiyoshi Ueda 已提交
2002

J
Jun'ichi Nomura 已提交
2003
	BUG_ON(!p || md->io_pool || md->bs);
K
Kiyoshi Ueda 已提交
2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014

	md->io_pool = p->io_pool;
	p->io_pool = NULL;
	md->bs = p->bs;
	p->bs = NULL;

out:
	/* mempool bind completed, now no need any mempools in the table */
	dm_table_free_md_mempools(t);
}

L
Linus Torvalds 已提交
2015 2016 2017 2018 2019
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2020 2021
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2022 2023
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2024 2025 2026 2027
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2028
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2029

L
Linus Torvalds 已提交
2030 2031 2032 2033
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2034 2035 2036
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2037
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2038
{
2039
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2040

2041
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2042 2043
}

2044 2045 2046 2047 2048 2049 2050 2051 2052 2053 2054 2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
/*
 * Return 1 if the queue has a compulsory merge_bvec_fn function.
 *
 * If this function returns 0, then the device is either a non-dm
 * device without a merge_bvec_fn, or it is a dm device that is
 * able to split any bios it receives that are too big.
 */
int dm_queue_merge_is_compulsory(struct request_queue *q)
{
	struct mapped_device *dev_md;

	if (!q->merge_bvec_fn)
		return 0;

	if (q->make_request_fn == dm_request) {
		dev_md = q->queuedata;
		if (test_bit(DMF_MERGE_IS_OPTIONAL, &dev_md->flags))
			return 0;
	}

	return 1;
}

static int dm_device_merge_is_compulsory(struct dm_target *ti,
					 struct dm_dev *dev, sector_t start,
					 sector_t len, void *data)
{
	struct block_device *bdev = dev->bdev;
	struct request_queue *q = bdev_get_queue(bdev);

	return dm_queue_merge_is_compulsory(q);
}

/*
 * Return 1 if it is acceptable to ignore merge_bvec_fn based
 * on the properties of the underlying devices.
 */
static int dm_table_merge_is_optional(struct dm_table *table)
{
	unsigned i = 0;
	struct dm_target *ti;

	while (i < dm_table_get_num_targets(table)) {
		ti = dm_table_get_target(table, i++);

		if (ti->type->iterate_devices &&
		    ti->type->iterate_devices(ti, dm_device_merge_is_compulsory, NULL))
			return 0;
	}

	return 1;
}

2097 2098 2099 2100 2101
/*
 * Returns old map, which caller must destroy.
 */
static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
			       struct queue_limits *limits)
L
Linus Torvalds 已提交
2102
{
2103
	struct dm_table *old_map;
2104
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2105
	sector_t size;
2106
	unsigned long flags;
2107
	int merge_is_optional;
L
Linus Torvalds 已提交
2108 2109

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2110 2111 2112 2113 2114 2115 2116

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
	if (size != get_capacity(md->disk))
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
2117
	__set_size(md, size);
2118

2119 2120
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132
	/*
	 * The queue hasn't been stopped yet, if the old table type wasn't
	 * for request-based during suspension.  So stop it to prevent
	 * I/O mapping before resume.
	 * This must be done before setting the queue restrictions,
	 * because request-based dm may be run just after the setting.
	 */
	if (dm_table_request_based(t) && !blk_queue_stopped(q))
		stop_queue(q);

	__bind_mempools(md, t);

2133 2134
	merge_is_optional = dm_table_merge_is_optional(t);

2135
	write_lock_irqsave(&md->map_lock, flags);
2136
	old_map = md->map;
L
Linus Torvalds 已提交
2137
	md->map = t;
2138 2139
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2140
	dm_table_set_restrictions(t, q, limits);
2141 2142 2143 2144
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
2145
	write_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
2146

2147
	return old_map;
L
Linus Torvalds 已提交
2148 2149
}

2150 2151 2152 2153
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2154 2155
{
	struct dm_table *map = md->map;
2156
	unsigned long flags;
L
Linus Torvalds 已提交
2157 2158

	if (!map)
2159
		return NULL;
L
Linus Torvalds 已提交
2160 2161

	dm_table_event_callback(map, NULL, NULL);
2162
	write_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
2163
	md->map = NULL;
2164
	write_unlock_irqrestore(&md->map_lock, flags);
2165 2166

	return map;
L
Linus Torvalds 已提交
2167 2168 2169 2170 2171
}

/*
 * Constructor for a new device.
 */
2172
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2173 2174 2175
{
	struct mapped_device *md;

2176
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2177 2178 2179
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2180 2181
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2182 2183 2184 2185
	*result = md;
	return 0;
}

2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209
/*
 * Functions to manage md->type.
 * All are required to hold md->type_lock.
 */
void dm_lock_md_type(struct mapped_device *md)
{
	mutex_lock(&md->type_lock);
}

void dm_unlock_md_type(struct mapped_device *md)
{
	mutex_unlock(&md->type_lock);
}

void dm_set_md_type(struct mapped_device *md, unsigned type)
{
	md->type = type;
}

unsigned dm_get_md_type(struct mapped_device *md)
{
	return md->type;
}

2210 2211 2212 2213 2214
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
/*
 * Fully initialize a request-based queue (->elevator, ->request_fn, etc).
 */
static int dm_init_request_based_queue(struct mapped_device *md)
{
	struct request_queue *q = NULL;

	if (md->queue->elevator)
		return 1;

	/* Fully initialize the queue */
	q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL);
	if (!q)
		return 0;

	md->queue = q;
	dm_init_md_queue(md);
	blk_queue_softirq_done(md->queue, dm_softirq_done);
	blk_queue_prep_rq(md->queue, dm_prep_fn);
	blk_queue_lld_busy(md->queue, dm_lld_busy);

	elv_register_queue(md->queue);

	return 1;
}

/*
 * Setup the DM device's queue based on md's type
 */
int dm_setup_md_queue(struct mapped_device *md)
{
	if ((dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) &&
	    !dm_init_request_based_queue(md)) {
		DMWARN("Cannot initialize queue for request-based mapped device");
		return -EINVAL;
	}

	return 0;
}

2255
static struct mapped_device *dm_find_md(dev_t dev)
L
Linus Torvalds 已提交
2256 2257 2258 2259 2260 2261 2262
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

	if (MAJOR(dev) != _major || minor >= (1 << MINORBITS))
		return NULL;

2263
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2264 2265

	md = idr_find(&_minor_idr, minor);
J
Jeff Mahoney 已提交
2266
	if (md && (md == MINOR_ALLOCED ||
2267
		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
2268
		   dm_deleting_md(md) ||
A
Alasdair G Kergon 已提交
2269
		   test_bit(DMF_FREEING, &md->flags))) {
2270
		md = NULL;
J
Jeff Mahoney 已提交
2271 2272
		goto out;
	}
L
Linus Torvalds 已提交
2273

J
Jeff Mahoney 已提交
2274
out:
2275
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2276

2277 2278 2279
	return md;
}

2280 2281 2282 2283 2284 2285 2286 2287 2288
struct mapped_device *dm_get_md(dev_t dev)
{
	struct mapped_device *md = dm_find_md(dev);

	if (md)
		dm_get(md);

	return md;
}
A
Alasdair G Kergon 已提交
2289
EXPORT_SYMBOL_GPL(dm_get_md);
2290

A
Alasdair G Kergon 已提交
2291
void *dm_get_mdptr(struct mapped_device *md)
2292
{
A
Alasdair G Kergon 已提交
2293
	return md->interface_ptr;
L
Linus Torvalds 已提交
2294 2295 2296 2297 2298 2299 2300 2301 2302 2303
}

void dm_set_mdptr(struct mapped_device *md, void *ptr)
{
	md->interface_ptr = ptr;
}

void dm_get(struct mapped_device *md)
{
	atomic_inc(&md->holders);
2304
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2305 2306
}

2307 2308 2309 2310 2311 2312
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2313
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2314
{
M
Mike Anderson 已提交
2315
	struct dm_table *map;
L
Linus Torvalds 已提交
2316

2317
	might_sleep();
J
Jeff Mahoney 已提交
2318

2319 2320 2321 2322 2323 2324 2325 2326 2327
	spin_lock(&_minor_lock);
	map = dm_get_live_table(md);
	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
	set_bit(DMF_FREEING, &md->flags);
	spin_unlock(&_minor_lock);

	if (!dm_suspended_md(md)) {
		dm_table_presuspend_targets(map);
		dm_table_postsuspend_targets(map);
L
Linus Torvalds 已提交
2328
	}
2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361

	/*
	 * Rare, but there may be I/O requests still going to complete,
	 * for example.  Wait for all references to disappear.
	 * No one should increment the reference count of the mapped_device,
	 * after the mapped_device state becomes DMF_FREEING.
	 */
	if (wait)
		while (atomic_read(&md->holders))
			msleep(1);
	else if (atomic_read(&md->holders))
		DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)",
		       dm_device_name(md), atomic_read(&md->holders));

	dm_sysfs_exit(md);
	dm_table_put(map);
	dm_table_destroy(__unbind(md));
	free_dev(md);
}

void dm_destroy(struct mapped_device *md)
{
	__dm_destroy(md, true);
}

void dm_destroy_immediate(struct mapped_device *md)
{
	__dm_destroy(md, false);
}

void dm_put(struct mapped_device *md)
{
	atomic_dec(&md->holders);
L
Linus Torvalds 已提交
2362
}
E
Edward Goggin 已提交
2363
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2364

2365
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2366 2367
{
	int r = 0;
2368 2369 2370
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2371 2372

	while (1) {
2373
		set_current_state(interruptible);
2374

2375
		if (!md_in_flight(md))
2376 2377
			break;

2378 2379
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2380 2381 2382 2383 2384 2385 2386 2387
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2388 2389
	remove_wait_queue(&md->wait, &wait);

2390 2391 2392
	return r;
}

L
Linus Torvalds 已提交
2393 2394 2395
/*
 * Process the deferred bios
 */
2396
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2397
{
2398 2399
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2400
	struct bio *c;
L
Linus Torvalds 已提交
2401

2402
	down_read(&md->io_lock);
2403

2404
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2405 2406 2407 2408
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2409
		if (!c)
A
Alasdair G Kergon 已提交
2410
			break;
2411

2412
		up_read(&md->io_lock);
2413

K
Kiyoshi Ueda 已提交
2414 2415
		if (dm_request_based(md))
			generic_make_request(c);
2416 2417
		else
			__split_and_process_bio(md, c);
2418

2419
		down_read(&md->io_lock);
2420
	}
M
Milan Broz 已提交
2421

2422
	up_read(&md->io_lock);
L
Linus Torvalds 已提交
2423 2424
}

2425
static void dm_queue_flush(struct mapped_device *md)
2426
{
2427 2428
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	smp_mb__after_clear_bit();
2429
	queue_work(md->wq, &md->work);
2430 2431
}

L
Linus Torvalds 已提交
2432
/*
2433
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2434
 */
2435
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2436
{
2437
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2438
	struct queue_limits limits;
2439
	int r;
L
Linus Torvalds 已提交
2440

2441
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2442 2443

	/* device must be suspended */
2444
	if (!dm_suspended_md(md))
2445
		goto out;
L
Linus Torvalds 已提交
2446

2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459
	/*
	 * If the new table has no data devices, retain the existing limits.
	 * This helps multipath with queue_if_no_path if all paths disappear,
	 * then new I/O is queued based on these limits, and then some paths
	 * reappear.
	 */
	if (dm_table_has_no_data_devices(table)) {
		live_map = dm_get_live_table(md);
		if (live_map)
			limits = md->queue->limits;
		dm_table_put(live_map);
	}

2460 2461 2462 2463 2464 2465
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2466
	}
2467

2468
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2469

2470
out:
2471
	mutex_unlock(&md->suspend_lock);
2472
	return map;
L
Linus Torvalds 已提交
2473 2474 2475 2476 2477 2478
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2479
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2480
{
2481
	int r;
L
Linus Torvalds 已提交
2482 2483

	WARN_ON(md->frozen_sb);
2484

2485
	md->frozen_sb = freeze_bdev(md->bdev);
2486
	if (IS_ERR(md->frozen_sb)) {
2487
		r = PTR_ERR(md->frozen_sb);
2488 2489
		md->frozen_sb = NULL;
		return r;
2490 2491
	}

2492 2493
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2494 2495 2496
	return 0;
}

2497
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2498
{
2499 2500 2501
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2502
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2503
	md->frozen_sb = NULL;
2504
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2505 2506 2507 2508 2509 2510 2511 2512 2513
}

/*
 * We need to be able to change a mapping table under a mounted
 * filesystem.  For example we might want to move some data in
 * the background.  Before the table can be swapped with
 * dm_bind_table, dm_suspend must be called to flush any in
 * flight bios and ensure that any further io gets deferred.
 */
2514 2515 2516
/*
 * Suspend mechanism in request-based dm.
 *
K
Kiyoshi Ueda 已提交
2517 2518 2519
 * 1. Flush all I/Os by lock_fs() if needed.
 * 2. Stop dispatching any I/O by stopping the request_queue.
 * 3. Wait for all in-flight I/Os to be completed or requeued.
2520
 *
K
Kiyoshi Ueda 已提交
2521
 * To abort suspend, start the request_queue.
2522
 */
2523
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
L
Linus Torvalds 已提交
2524
{
2525
	struct dm_table *map = NULL;
2526
	int r = 0;
2527
	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
2528
	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
L
Linus Torvalds 已提交
2529

2530
	mutex_lock(&md->suspend_lock);
2531

2532
	if (dm_suspended_md(md)) {
M
Milan Broz 已提交
2533
		r = -EINVAL;
2534
		goto out_unlock;
M
Milan Broz 已提交
2535
	}
L
Linus Torvalds 已提交
2536

2537
	map = dm_get_live_table(md);
L
Linus Torvalds 已提交
2538

2539 2540 2541 2542 2543 2544 2545
	/*
	 * DMF_NOFLUSH_SUSPENDING must be set before presuspend.
	 * This flag is cleared before dm_suspend returns.
	 */
	if (noflush)
		set_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);

2546 2547 2548
	/* This does not get reverted if there's an error later. */
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2549
	/*
K
Kiyoshi Ueda 已提交
2550 2551 2552 2553
	 * Flush I/O to the device.
	 * Any I/O submitted after lock_fs() may not be flushed.
	 * noflush takes precedence over do_lockfs.
	 * (lock_fs() flushes I/Os and waits for them to complete.)
M
Mikulas Patocka 已提交
2554 2555 2556 2557
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
		if (r)
K
Kiyoshi Ueda 已提交
2558
			goto out;
2559
	}
L
Linus Torvalds 已提交
2560 2561

	/*
2562 2563 2564 2565 2566 2567 2568
	 * Here we must make sure that no processes are submitting requests
	 * to target drivers i.e. no one may be executing
	 * __split_and_process_bio. This is called from dm_request and
	 * dm_wq_work.
	 *
	 * To get all processes out of __split_and_process_bio in dm_request,
	 * we take the write lock. To prevent any process from reentering
2569 2570 2571
	 * __split_and_process_bio from dm_request and quiesce the thread
	 * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call
	 * flush_workqueue(md->wq).
L
Linus Torvalds 已提交
2572
	 */
2573
	down_write(&md->io_lock);
2574
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2575
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
2576

2577
	/*
2578 2579
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2580
	 */
2581
	if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2582
		stop_queue(md->queue);
2583

2584 2585
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2586
	/*
2587 2588 2589
	 * At this point no more requests are entering target request routines.
	 * We call dm_wait_for_completion to wait for all existing requests
	 * to finish.
L
Linus Torvalds 已提交
2590
	 */
2591
	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
L
Linus Torvalds 已提交
2592

2593
	down_write(&md->io_lock);
2594
	if (noflush)
2595
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2596
	up_write(&md->io_lock);
2597

L
Linus Torvalds 已提交
2598
	/* were we interrupted ? */
2599
	if (r < 0) {
2600
		dm_queue_flush(md);
M
Milan Broz 已提交
2601

2602
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2603
			start_queue(md->queue);
2604

2605
		unlock_fs(md);
2606
		goto out; /* pushback list is already flushed, so skip flush */
2607
	}
L
Linus Torvalds 已提交
2608

2609 2610 2611 2612 2613 2614
	/*
	 * If dm_wait_for_completion returned 0, the device is completely
	 * quiescent now. There is no request-processing activity. All new
	 * requests are being added to md->deferred list.
	 */

2615
	set_bit(DMF_SUSPENDED, &md->flags);
2616

2617 2618
	dm_table_postsuspend_targets(map);

2619 2620
out:
	dm_table_put(map);
2621 2622

out_unlock:
2623
	mutex_unlock(&md->suspend_lock);
2624
	return r;
L
Linus Torvalds 已提交
2625 2626 2627 2628
}

int dm_resume(struct mapped_device *md)
{
2629 2630
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2631

2632
	mutex_lock(&md->suspend_lock);
2633
	if (!dm_suspended_md(md))
2634 2635
		goto out;

2636
	map = dm_get_live_table(md);
2637
	if (!map || !dm_table_get_size(map))
2638
		goto out;
L
Linus Torvalds 已提交
2639

2640 2641 2642
	r = dm_table_resume_targets(map);
	if (r)
		goto out;
2643

2644
	dm_queue_flush(md);
2645

2646 2647 2648 2649 2650 2651 2652 2653
	/*
	 * Flushing deferred I/Os must be done after targets are resumed
	 * so that mapping of targets can work correctly.
	 * Request-based dm is queueing the deferred I/Os in its request_queue.
	 */
	if (dm_request_based(md))
		start_queue(md->queue);

2654 2655 2656 2657
	unlock_fs(md);

	clear_bit(DMF_SUSPENDED, &md->flags);

2658 2659 2660
	r = 0;
out:
	dm_table_put(map);
2661
	mutex_unlock(&md->suspend_lock);
2662

2663
	return r;
L
Linus Torvalds 已提交
2664 2665 2666 2667 2668
}

/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2669
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2670
		       unsigned cookie)
2671
{
M
Milan Broz 已提交
2672 2673 2674 2675
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2676
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2677 2678 2679
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2680 2681
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2682
	}
2683 2684
}

M
Mike Anderson 已提交
2685 2686 2687 2688 2689
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2690 2691 2692 2693 2694 2695 2696 2697 2698 2699 2700
uint32_t dm_get_event_nr(struct mapped_device *md)
{
	return atomic_read(&md->event_nr);
}

int dm_wait_event(struct mapped_device *md, int event_nr)
{
	return wait_event_interruptible(md->eventq,
			(event_nr != atomic_read(&md->event_nr)));
}

M
Mike Anderson 已提交
2701 2702 2703 2704 2705 2706 2707 2708 2709
void dm_uevent_add(struct mapped_device *md, struct list_head *elist)
{
	unsigned long flags;

	spin_lock_irqsave(&md->uevent_lock, flags);
	list_add(elist, &md->uevent_list);
	spin_unlock_irqrestore(&md->uevent_lock, flags);
}

L
Linus Torvalds 已提交
2710 2711 2712 2713 2714 2715 2716 2717 2718
/*
 * The gendisk is only valid as long as you have a reference
 * count on 'md'.
 */
struct gendisk *dm_disk(struct mapped_device *md)
{
	return md->disk;
}

M
Milan Broz 已提交
2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735
struct kobject *dm_kobject(struct mapped_device *md)
{
	return &md->kobj;
}

/*
 * struct mapped_device should not be exported outside of dm.c
 * so use this check to verify that kobj is part of md structure
 */
struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
{
	struct mapped_device *md;

	md = container_of(kobj, struct mapped_device, kobj);
	if (&md->kobj != kobj)
		return NULL;

2736
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2737
	    dm_deleting_md(md))
2738 2739
		return NULL;

M
Milan Broz 已提交
2740 2741 2742 2743
	dm_get(md);
	return md;
}

2744
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2745 2746 2747 2748
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2749 2750
int dm_suspended(struct dm_target *ti)
{
2751
	return dm_suspended_md(dm_table_get_md(ti->table));
2752 2753 2754
}
EXPORT_SYMBOL_GPL(dm_suspended);

2755 2756
int dm_noflush_suspending(struct dm_target *ti)
{
2757
	return __noflush_suspending(dm_table_get_md(ti->table));
2758 2759 2760
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

M
Mikulas Patocka 已提交
2761
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size)
K
Kiyoshi Ueda 已提交
2762
{
J
Jun'ichi Nomura 已提交
2763 2764 2765 2766
	struct dm_md_mempools *pools = kzalloc(sizeof(*pools), GFP_KERNEL);
	struct kmem_cache *cachep;
	unsigned int pool_size;
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2767 2768 2769 2770

	if (!pools)
		return NULL;

2771
	if (type == DM_TYPE_BIO_BASED) {
J
Jun'ichi Nomura 已提交
2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782
		cachep = _io_cache;
		pool_size = 16;
		front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
	} else if (type == DM_TYPE_REQUEST_BASED) {
		cachep = _rq_tio_cache;
		pool_size = MIN_IOS;
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
		/* per_bio_data_size is not used. See __bind_mempools(). */
		WARN_ON(per_bio_data_size != 0);
	} else
		goto out;
K
Kiyoshi Ueda 已提交
2783

J
Jun'ichi Nomura 已提交
2784 2785 2786
	pools->io_pool = mempool_create_slab_pool(MIN_IOS, cachep);
	if (!pools->io_pool)
		goto out;
K
Kiyoshi Ueda 已提交
2787

J
Jun'ichi Nomura 已提交
2788
	pools->bs = bioset_create(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
2789
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2790
		goto out;
K
Kiyoshi Ueda 已提交
2791

2792
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2793
		goto out;
2794

K
Kiyoshi Ueda 已提交
2795 2796
	return pools;

J
Jun'ichi Nomura 已提交
2797 2798
out:
	dm_free_md_mempools(pools);
K
Kiyoshi Ueda 已提交
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816

	return NULL;
}

void dm_free_md_mempools(struct dm_md_mempools *pools)
{
	if (!pools)
		return;

	if (pools->io_pool)
		mempool_destroy(pools->io_pool);

	if (pools->bs)
		bioset_free(pools->bs);

	kfree(pools);
}

2817
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2818 2819
	.open = dm_blk_open,
	.release = dm_blk_close,
2820
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2821
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837
	.owner = THIS_MODULE
};

EXPORT_SYMBOL(dm_get_mapinfo);

/*
 * module hooks
 */
module_init(dm_init);
module_exit(dm_exit);

module_param(major, uint, 0);
MODULE_PARM_DESC(major, "The major number of the device mapper");
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");