dm.c 62.1 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 65
};

/*
K
Kiyoshi Ueda 已提交
66
 * For bio-based dm.
L
Linus Torvalds 已提交
67 68 69
 * One of these is allocated per target within a bio.  Hopefully
 * this will be simplified out one day.
 */
A
Alasdair G Kergon 已提交
70
struct dm_target_io {
L
Linus Torvalds 已提交
71 72 73
	struct dm_io *io;
	struct dm_target *ti;
	union map_info info;
74
	struct bio clone;
L
Linus Torvalds 已提交
75 76
};

K
Kiyoshi Ueda 已提交
77 78 79 80 81 82 83 84 85 86 87 88 89
/*
 * 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;
};

/*
90 91 92 93 94 95
 * 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
96 97 98
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
99
	struct dm_rq_target_io *tio;
100
	struct bio clone;
K
Kiyoshi Ueda 已提交
101 102
};

L
Linus Torvalds 已提交
103 104
union map_info *dm_get_mapinfo(struct bio *bio)
{
A
Alasdair G Kergon 已提交
105
	if (bio && bio->bi_private)
A
Alasdair G Kergon 已提交
106
		return &((struct dm_target_io *)bio->bi_private)->info;
A
Alasdair G Kergon 已提交
107
	return NULL;
L
Linus Torvalds 已提交
108 109
}

110 111 112 113 114 115 116 117
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);

118 119
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
120 121 122
/*
 * Bits for the md->flags field.
 */
123
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
124
#define DMF_SUSPENDED 1
125
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
126
#define DMF_FREEING 3
127
#define DMF_DELETING 4
128
#define DMF_NOFLUSH_SUSPENDING 5
129
#define DMF_MERGE_IS_OPTIONAL 6
L
Linus Torvalds 已提交
130

131 132 133
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
134
struct mapped_device {
135
	struct rw_semaphore io_lock;
136
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
137 138
	rwlock_t map_lock;
	atomic_t holders;
139
	atomic_t open_count;
L
Linus Torvalds 已提交
140 141 142

	unsigned long flags;

143
	struct request_queue *queue;
144
	unsigned type;
145
	/* Protect queue and type against concurrent access. */
146 147
	struct mutex type_lock;

148 149
	struct target_type *immutable_target_type;

L
Linus Torvalds 已提交
150
	struct gendisk *disk;
M
Mike Anderson 已提交
151
	char name[16];
L
Linus Torvalds 已提交
152 153 154 155 156 157

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
158
	atomic_t pending[2];
L
Linus Torvalds 已提交
159
	wait_queue_head_t wait;
160
	struct work_struct work;
K
Kiyoshi Ueda 已提交
161
	struct bio_list deferred;
162
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
163

164
	/*
165
	 * Processing queue (flush)
166 167 168
	 */
	struct workqueue_struct *wq;

L
Linus Torvalds 已提交
169 170 171 172 173 174 175 176 177 178 179
	/*
	 * The current mapping.
	 */
	struct dm_table *map;

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

S
Stefan Bader 已提交
180 181
	struct bio_set *bs;

L
Linus Torvalds 已提交
182 183 184 185 186
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
187 188 189
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
190 191 192 193 194

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
195
	struct block_device *bdev;
D
Darrick J. Wong 已提交
196 197 198

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
199 200 201

	/* sysfs handle */
	struct kobject kobj;
202

203 204
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
L
Linus Torvalds 已提交
205 206
};

K
Kiyoshi Ueda 已提交
207 208 209 210 211 212 213 214 215
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
	mempool_t *tio_pool;
	struct bio_set *bs;
};

L
Linus Torvalds 已提交
216
#define MIN_IOS 256
217
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
218
static struct kmem_cache *_rq_tio_cache;
219 220 221 222 223

/*
 * Unused now, and needs to be deleted. But since io_pool is overloaded and it's
 * still used for _io_cache, I'm leaving this for a later cleanup
 */
K
Kiyoshi Ueda 已提交
224
static struct kmem_cache *_rq_bio_info_cache;
L
Linus Torvalds 已提交
225 226 227

static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
228
	int r = -ENOMEM;
L
Linus Torvalds 已提交
229 230

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
231
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
232
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
233
		return r;
L
Linus Torvalds 已提交
234

K
Kiyoshi Ueda 已提交
235 236
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
237
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
238 239 240 241 242

	_rq_bio_info_cache = KMEM_CACHE(dm_rq_clone_bio_info, 0);
	if (!_rq_bio_info_cache)
		goto out_free_rq_tio_cache;

M
Mike Anderson 已提交
243
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
244
	if (r)
K
Kiyoshi Ueda 已提交
245
		goto out_free_rq_bio_info_cache;
M
Mike Anderson 已提交
246

L
Linus Torvalds 已提交
247 248
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
249 250
	if (r < 0)
		goto out_uevent_exit;
L
Linus Torvalds 已提交
251 252 253 254 255

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
256 257 258

out_uevent_exit:
	dm_uevent_exit();
K
Kiyoshi Ueda 已提交
259 260 261 262
out_free_rq_bio_info_cache:
	kmem_cache_destroy(_rq_bio_info_cache);
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
263 264 265 266
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
267 268 269 270
}

static void local_exit(void)
{
K
Kiyoshi Ueda 已提交
271 272
	kmem_cache_destroy(_rq_bio_info_cache);
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
273
	kmem_cache_destroy(_io_cache);
274
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
275
	dm_uevent_exit();
L
Linus Torvalds 已提交
276 277 278 279 280 281

	_major = 0;

	DMINFO("cleaned up");
}

282
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
283 284 285 286
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
287
	dm_io_init,
288
	dm_kcopyd_init,
L
Linus Torvalds 已提交
289 290 291
	dm_interface_init,
};

292
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
293 294 295 296
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
297
	dm_io_exit,
298
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328
	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]();
329 330 331 332 333 334

	/*
	 * Should be empty by this point.
	 */
	idr_remove_all(&_minor_idr);
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
335 336 337 338 339
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
340 341 342 343 344
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
345
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
346 347 348
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
349 350
	spin_lock(&_minor_lock);

A
Al Viro 已提交
351
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
352 353 354
	if (!md)
		goto out;

355
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
356
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
357 358 359 360
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
361
	dm_get(md);
362
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
363 364 365 366 367

out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
368 369
}

A
Al Viro 已提交
370
static int dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
371
{
A
Al Viro 已提交
372
	struct mapped_device *md = disk->private_data;
373

374 375
	spin_lock(&_minor_lock);

376
	atomic_dec(&md->open_count);
L
Linus Torvalds 已提交
377
	dm_put(md);
378 379

	spin_unlock(&_minor_lock);
380

L
Linus Torvalds 已提交
381 382 383
	return 0;
}

384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407
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
408 409 410 411 412 413 414
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
415
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
416 417
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
418
	struct mapped_device *md = bdev->bd_disk->private_data;
419
	struct dm_table *map = dm_get_live_table(md);
420 421 422 423 424 425 426 427 428 429 430 431
	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);

432
	if (dm_suspended_md(md)) {
433 434 435 436 437
		r = -EAGAIN;
		goto out;
	}

	if (tgt->type->ioctl)
438
		r = tgt->type->ioctl(tgt, cmd, arg);
439 440 441 442 443 444 445

out:
	dm_table_put(map);

	return r;
}

A
Alasdair G Kergon 已提交
446
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
447 448 449 450
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
451
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
452 453 454 455
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
456
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
457
{
458
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
459 460
}

K
Kiyoshi Ueda 已提交
461 462
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
463
{
K
Kiyoshi Ueda 已提交
464
	return mempool_alloc(md->tio_pool, gfp_mask);
465 466 467 468 469 470 471
}

static void free_rq_tio(struct dm_rq_target_io *tio)
{
	mempool_free(tio, tio->md->tio_pool);
}

K
Kiyoshi Ueda 已提交
472 473 474 475 476 477
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

478 479 480
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
T
Tejun Heo 已提交
481
	int cpu;
482
	int rw = bio_data_dir(io->bio);
483 484 485

	io->start_time = jiffies;

T
Tejun Heo 已提交
486 487 488
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
489 490
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
491 492
}

493
static void end_io_acct(struct dm_io *io)
494 495 496 497
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
T
Tejun Heo 已提交
498
	int pending, cpu;
499 500
	int rw = bio_data_dir(bio);

T
Tejun Heo 已提交
501 502 503 504
	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();
505

506 507
	/*
	 * After this is decremented the bio must not be touched if it is
508
	 * a flush.
509
	 */
510 511
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
512
	pending += atomic_read(&md->pending[rw^0x1]);
513

514 515 516
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
517 518
}

L
Linus Torvalds 已提交
519 520 521
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
522
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
523
{
524
	unsigned long flags;
L
Linus Torvalds 已提交
525

526
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
527
	bio_list_add(&md->deferred, bio);
528
	spin_unlock_irqrestore(&md->deferred_lock, flags);
529
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
530 531 532 533 534 535 536
}

/*
 * 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.
 */
537
struct dm_table *dm_get_live_table(struct mapped_device *md)
L
Linus Torvalds 已提交
538 539
{
	struct dm_table *t;
540
	unsigned long flags;
L
Linus Torvalds 已提交
541

542
	read_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
543 544 545
	t = md->map;
	if (t)
		dm_table_get(t);
546
	read_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
547 548 549 550

	return t;
}

D
Darrick J. Wong 已提交
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577
/*
 * 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
578 579 580 581 582 583 584 585 586
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

587 588 589 590 591
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
592 593 594 595
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
596
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
597
{
598
	unsigned long flags;
599 600 601
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
602 603

	/* Push-back supersedes any I/O errors */
604 605 606 607 608 609
	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
610 611

	if (atomic_dec_and_test(&io->io_count)) {
612 613 614 615
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
616
			spin_lock_irqsave(&md->deferred_lock, flags);
617 618 619
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
620 621
				/* noflush suspend was interrupted. */
				io->error = -EIO;
622
			spin_unlock_irqrestore(&md->deferred_lock, flags);
623 624
		}

625 626
		io_error = io->error;
		bio = io->bio;
627 628 629 630 631
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
632

633
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_size) {
634
			/*
635 636
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
637
			 */
638 639
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
640
		} else {
641
			/* done with normal IO or empty flush */
642
			trace_block_bio_complete(md->queue, bio, io_error);
643
			bio_endio(bio, io_error);
644
		}
L
Linus Torvalds 已提交
645 646 647
	}
}

648
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
649 650
{
	int r = 0;
A
Alasdair G Kergon 已提交
651
	struct dm_target_io *tio = bio->bi_private;
652
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
653
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
654 655 656 657 658 659 660
	dm_endio_fn endio = tio->ti->type->end_io;

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

	if (endio) {
		r = endio(tio->ti, bio, error, &tio->info);
661 662 663 664 665
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
666
			error = r;
667 668
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
669
			return;
670 671 672 673
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
674 675
	}

S
Stefan Bader 已提交
676
	free_tio(md, tio);
677
	dec_pending(io, error);
L
Linus Torvalds 已提交
678 679
}

680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734
/*
 * 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.
 */
735
static void rq_completed(struct mapped_device *md, int rw, int run_queue)
736
{
737
	atomic_dec(&md->pending[rw]);
738 739

	/* nudge anyone waiting on suspend queue */
740
	if (!md_in_flight(md))
741 742
		wake_up(&md->wait);

743 744 745 746 747 748
	/*
	 * 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.
	 */
749
	if (run_queue)
750
		blk_run_queue_async(md->queue);
751 752 753 754 755 756 757

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

758 759 760 761 762 763 764 765
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
766 767 768 769 770 771 772 773 774 775 776
/*
 * 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;

777
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
778 779 780 781 782 783 784 785 786 787 788 789 790
		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);
791 792
	blk_end_request_all(rq, error);
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
793 794
}

795 796 797 798 799 800 801
static void dm_unprep_request(struct request *rq)
{
	struct request *clone = rq->special;

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

802
	free_rq_clone(clone);
803 804 805 806 807 808 809
}

/*
 * Requeue the original request of a clone.
 */
void dm_requeue_unmapped_request(struct request *clone)
{
810
	int rw = rq_data_dir(clone);
811 812 813 814 815 816 817 818 819 820 821 822
	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);

823
	rq_completed(md, rw, 0);
824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
}
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);
}

856
static void dm_done(struct request *clone, int error, bool mapped)
857
{
858
	int r = error;
859
	struct dm_rq_target_io *tio = clone->end_io_data;
860
	dm_request_endio_fn rq_end_io = NULL;
861

862 863 864 865 866 867
	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);
	}
868

869
	if (r <= 0)
870
		/* The target wants to complete the I/O */
871 872
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
873 874
		/* The target will handle the I/O */
		return;
875
	else if (r == DM_ENDIO_REQUEUE)
876 877 878
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
879
		DMWARN("unimplemented target endio return value: %d", r);
880 881 882 883
		BUG();
	}
}

884 885 886 887 888 889 890 891 892 893 894 895 896 897 898
/*
 * 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);
}

899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
/*
 * 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);
}

953 954 955 956 957 958 959 960 961 962 963 964
/*
 * 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
965
{
966
	sector_t len = max_io_len_target_boundary(sector, ti);
967
	sector_t offset, max_len;
L
Linus Torvalds 已提交
968 969

	/*
970
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
971
	 */
972 973 974 975 976 977 978 979 980 981
	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
982 983 984 985 986
	}

	return len;
}

987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001
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);

1002
static void __map_bio(struct dm_target *ti, struct dm_target_io *tio)
L
Linus Torvalds 已提交
1003 1004
{
	int r;
1005
	sector_t sector;
S
Stefan Bader 已提交
1006
	struct mapped_device *md;
1007
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017

	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);
1018
	sector = clone->bi_sector;
L
Linus Torvalds 已提交
1019
	r = ti->type->map(ti, clone, &tio->info);
1020
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1021
		/* the bio has been remapped so dispatch it */
1022

1023 1024
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
1025

L
Linus Torvalds 已提交
1026
		generic_make_request(clone);
1027 1028
	} 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
1029 1030 1031
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1032 1033 1034
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048
	}
}

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;
};

/*
1049
 * Creates a little bio that just does part of a bvec.
L
Linus Torvalds 已提交
1050
 */
1051 1052 1053
static void split_bvec(struct dm_target_io *tio, struct bio *bio,
		       sector_t sector, unsigned short idx, unsigned int offset,
		       unsigned int len, struct bio_set *bs)
L
Linus Torvalds 已提交
1054
{
1055
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1056 1057 1058 1059 1060 1061
	struct bio_vec *bv = bio->bi_io_vec + idx;

	*clone->bi_io_vec = *bv;

	clone->bi_sector = sector;
	clone->bi_bdev = bio->bi_bdev;
1062
	clone->bi_rw = bio->bi_rw;
L
Linus Torvalds 已提交
1063 1064 1065 1066
	clone->bi_vcnt = 1;
	clone->bi_size = to_bytes(len);
	clone->bi_io_vec->bv_offset = offset;
	clone->bi_io_vec->bv_len = clone->bi_size;
M
Martin K. Petersen 已提交
1067
	clone->bi_flags |= 1 << BIO_CLONED;
L
Linus Torvalds 已提交
1068

M
Martin K. Petersen 已提交
1069
	if (bio_integrity(bio)) {
1070
		bio_integrity_clone(clone, bio, GFP_NOIO);
M
Martin K. Petersen 已提交
1071 1072 1073
		bio_integrity_trim(clone,
				   bio_sector_offset(bio, idx, offset), len);
	}
L
Linus Torvalds 已提交
1074 1075 1076 1077 1078
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1079 1080 1081 1082
static void clone_bio(struct dm_target_io *tio, struct bio *bio,
		      sector_t sector, unsigned short idx,
		      unsigned short bv_count, unsigned int len,
		      struct bio_set *bs)
L
Linus Torvalds 已提交
1083
{
1084
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1085

S
Stefan Bader 已提交
1086
	__bio_clone(clone, bio);
L
Linus Torvalds 已提交
1087 1088 1089 1090 1091 1092
	clone->bi_sector = sector;
	clone->bi_idx = idx;
	clone->bi_vcnt = idx + bv_count;
	clone->bi_size = to_bytes(len);
	clone->bi_flags &= ~(1 << BIO_SEG_VALID);

M
Martin K. Petersen 已提交
1093
	if (bio_integrity(bio)) {
1094
		bio_integrity_clone(clone, bio, GFP_NOIO);
M
Martin K. Petersen 已提交
1095 1096 1097 1098 1099

		if (idx != bio->bi_idx || clone->bi_size < bio->bi_size)
			bio_integrity_trim(clone,
					   bio_sector_offset(bio, idx, 0), len);
	}
L
Linus Torvalds 已提交
1100 1101
}

1102
static struct dm_target_io *alloc_tio(struct clone_info *ci,
1103
				      struct dm_target *ti, int nr_iovecs)
1104
{
1105 1106 1107 1108 1109
	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);
1110 1111 1112 1113

	tio->io = ci->io;
	tio->ti = ti;
	memset(&tio->info, 0, sizeof(tio->info));
1114 1115 1116 1117

	return tio;
}

1118
static void __issue_target_request(struct clone_info *ci, struct dm_target *ti,
1119
				   unsigned request_nr, sector_t len)
1120
{
1121 1122
	struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs);
	struct bio *clone = &tio->clone;
1123

1124
	tio->info.target_request_nr = request_nr;
1125

1126 1127 1128 1129 1130
	/*
	 * 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.
	 */
1131

1132
	 __bio_clone(clone, ci->bio);
1133 1134 1135 1136
	if (len) {
		clone->bi_sector = ci->sector;
		clone->bi_size = to_bytes(len);
	}
1137

1138
	__map_bio(ti, tio);
1139 1140
}

1141
static void __issue_target_requests(struct clone_info *ci, struct dm_target *ti,
1142
				    unsigned num_requests, sector_t len)
1143 1144 1145 1146
{
	unsigned request_nr;

	for (request_nr = 0; request_nr < num_requests; request_nr++)
1147
		__issue_target_request(ci, ti, request_nr, len);
1148 1149
}

1150
static int __clone_and_map_empty_flush(struct clone_info *ci)
1151
{
1152
	unsigned target_nr = 0;
1153 1154
	struct dm_target *ti;

1155
	BUG_ON(bio_has_data(ci->bio));
1156
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1157
		__issue_target_requests(ci, ti, ti->num_flush_requests, 0);
1158 1159 1160 1161

	return 0;
}

M
Mike Snitzer 已提交
1162 1163 1164 1165 1166
/*
 * Perform all io with a single clone.
 */
static void __clone_and_map_simple(struct clone_info *ci, struct dm_target *ti)
{
1167
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1168 1169
	struct dm_target_io *tio;

1170 1171 1172 1173
	tio = alloc_tio(ci, ti, bio->bi_max_vecs);
	clone_bio(tio, bio, ci->sector, ci->idx, bio->bi_vcnt - ci->idx,
		  ci->sector_count, ci->md->bs);
	__map_bio(ti, tio);
M
Mike Snitzer 已提交
1174 1175 1176
	ci->sector_count = 0;
}

M
Mike Snitzer 已提交
1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198
typedef unsigned (*get_num_requests_fn)(struct dm_target *ti);

static unsigned get_num_discard_requests(struct dm_target *ti)
{
	return ti->num_discard_requests;
}

static unsigned get_num_write_same_requests(struct dm_target *ti)
{
	return ti->num_write_same_requests;
}

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

static bool is_split_required_for_discard(struct dm_target *ti)
{
	return ti->split_discard_requests;
}

static int __clone_and_map_changing_extent_only(struct clone_info *ci,
						get_num_requests_fn get_num_requests,
						is_split_required_fn is_split_required)
M
Mike Snitzer 已提交
1199 1200
{
	struct dm_target *ti;
1201
	sector_t len;
M
Mike Snitzer 已提交
1202

1203 1204 1205 1206
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
M
Mike Snitzer 已提交
1207 1208

		/*
M
Mike Snitzer 已提交
1209 1210
		 * 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
1211
		 * reconfiguration might also have changed that since the
1212
		 * check was performed.
M
Mike Snitzer 已提交
1213
		 */
M
Mike Snitzer 已提交
1214
		if (!get_num_requests || !get_num_requests(ti))
1215
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1216

M
Mike Snitzer 已提交
1217
		if (is_split_required && !is_split_required(ti))
1218 1219 1220
			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));
1221

1222 1223 1224 1225
		__issue_target_requests(ci, ti, ti->num_discard_requests, len);

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1226 1227 1228 1229

	return 0;
}

M
Mike Snitzer 已提交
1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240
static int __clone_and_map_discard(struct clone_info *ci)
{
	return __clone_and_map_changing_extent_only(ci, get_num_discard_requests,
						    is_split_required_for_discard);
}

static int __clone_and_map_write_same(struct clone_info *ci)
{
	return __clone_and_map_changing_extent_only(ci, get_num_write_same_requests, NULL);
}

1241
static int __clone_and_map(struct clone_info *ci)
L
Linus Torvalds 已提交
1242
{
1243
	struct bio *bio = ci->bio;
1244 1245
	struct dm_target *ti;
	sector_t len = 0, max;
A
Alasdair G Kergon 已提交
1246
	struct dm_target_io *tio;
L
Linus Torvalds 已提交
1247

M
Mike Snitzer 已提交
1248 1249
	if (unlikely(bio->bi_rw & REQ_DISCARD))
		return __clone_and_map_discard(ci);
M
Mike Snitzer 已提交
1250 1251
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
		return __clone_and_map_write_same(ci);
M
Mike Snitzer 已提交
1252

1253 1254 1255 1256
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1257
	max = max_io_len(ci->sector, ti);
1258

L
Linus Torvalds 已提交
1259 1260 1261 1262 1263
	if (ci->sector_count <= max) {
		/*
		 * Optimise for the simple case where we can do all of
		 * the remaining io with a single clone.
		 */
M
Mike Snitzer 已提交
1264
		__clone_and_map_simple(ci, ti);
L
Linus Torvalds 已提交
1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284

	} else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) {
		/*
		 * There are some bvecs that don't span targets.
		 * Do as many of these as possible.
		 */
		int i;
		sector_t remaining = max;
		sector_t bv_len;

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

			if (bv_len > remaining)
				break;

			remaining -= bv_len;
			len += bv_len;
		}

1285 1286 1287 1288
		tio = alloc_tio(ci, ti, bio->bi_max_vecs);
		clone_bio(tio, bio, ci->sector, ci->idx, i - ci->idx, len,
			  ci->md->bs);
		__map_bio(ti, tio);
L
Linus Torvalds 已提交
1289 1290 1291 1292 1293 1294 1295

		ci->sector += len;
		ci->sector_count -= len;
		ci->idx = i;

	} else {
		/*
1296
		 * Handle a bvec that must be split between two or more targets.
L
Linus Torvalds 已提交
1297 1298
		 */
		struct bio_vec *bv = bio->bi_io_vec + ci->idx;
1299 1300
		sector_t remaining = to_sector(bv->bv_len);
		unsigned int offset = 0;
L
Linus Torvalds 已提交
1301

1302 1303 1304
		do {
			if (offset) {
				ti = dm_table_find_target(ci->map, ci->sector);
1305 1306 1307
				if (!dm_target_is_valid(ti))
					return -EIO;

1308
				max = max_io_len(ci->sector, ti);
1309 1310 1311 1312
			}

			len = min(remaining, max);

1313 1314 1315
			tio = alloc_tio(ci, ti, 1);
			split_bvec(tio, bio, ci->sector, ci->idx,
				   bv->bv_offset + offset, len, ci->md->bs);
1316

1317
			__map_bio(ti, tio);
1318 1319 1320 1321 1322

			ci->sector += len;
			ci->sector_count -= len;
			offset += to_bytes(len);
		} while (remaining -= len);
L
Linus Torvalds 已提交
1323 1324 1325

		ci->idx++;
	}
1326 1327

	return 0;
L
Linus Torvalds 已提交
1328 1329 1330
}

/*
M
Mikulas Patocka 已提交
1331
 * Split the bio into several clones and submit it to targets.
L
Linus Torvalds 已提交
1332
 */
1333
static void __split_and_process_bio(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
1334 1335
{
	struct clone_info ci;
1336
	int error = 0;
L
Linus Torvalds 已提交
1337

1338
	ci.map = dm_get_live_table(md);
1339
	if (unlikely(!ci.map)) {
1340
		bio_io_error(bio);
1341 1342
		return;
	}
1343

L
Linus Torvalds 已提交
1344 1345 1346 1347 1348 1349
	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;
1350
	spin_lock_init(&ci.io->endio_lock);
L
Linus Torvalds 已提交
1351 1352 1353
	ci.sector = bio->bi_sector;
	ci.idx = bio->bi_idx;

1354
	start_io_acct(ci.io);
1355 1356 1357 1358 1359 1360
	if (bio->bi_rw & REQ_FLUSH) {
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
		error = __clone_and_map_empty_flush(&ci);
		/* dec_pending submits any data associated with flush */
	} else {
1361
		ci.bio = bio;
1362
		ci.sector_count = bio_sectors(bio);
1363
		while (ci.sector_count && !error)
1364 1365
			error = __clone_and_map(&ci);
	}
L
Linus Torvalds 已提交
1366 1367

	/* drop the extra reference count */
1368
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1369 1370 1371 1372 1373 1374
	dm_table_put(ci.map);
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1375 1376 1377 1378 1379
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
1380
	struct dm_table *map = dm_get_live_table(md);
M
Milan Broz 已提交
1381 1382
	struct dm_target *ti;
	sector_t max_sectors;
1383
	int max_size = 0;
M
Milan Broz 已提交
1384 1385

	if (unlikely(!map))
1386
		goto out;
M
Milan Broz 已提交
1387 1388

	ti = dm_table_find_target(map, bvm->bi_sector);
1389 1390
	if (!dm_target_is_valid(ti))
		goto out_table;
M
Milan Broz 已提交
1391 1392 1393 1394

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1395
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
M
Milan Broz 已提交
1396 1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407
			  (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);
1408 1409 1410 1411 1412 1413 1414 1415 1416 1417
	/*
	 * 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
1418

1419
out_table:
1420 1421 1422
	dm_table_put(map);

out:
M
Milan Broz 已提交
1423 1424 1425 1426 1427 1428 1429 1430 1431
	/*
	 * 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
1432 1433 1434 1435
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1436
static void _dm_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1437
{
1438
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1439
	struct mapped_device *md = q->queuedata;
T
Tejun Heo 已提交
1440
	int cpu;
L
Linus Torvalds 已提交
1441

1442
	down_read(&md->io_lock);
L
Linus Torvalds 已提交
1443

T
Tejun Heo 已提交
1444 1445 1446 1447
	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();
1448

1449 1450
	/* if we're suspended, we have to queue this io for later */
	if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
1451
		up_read(&md->io_lock);
L
Linus Torvalds 已提交
1452

1453 1454 1455
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1456
			bio_io_error(bio);
1457
		return;
L
Linus Torvalds 已提交
1458 1459
	}

1460
	__split_and_process_bio(md, bio);
1461
	up_read(&md->io_lock);
1462
	return;
1463 1464 1465 1466 1467 1468 1469
}

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

1470
static void dm_request(struct request_queue *q, struct bio *bio)
1471 1472 1473 1474
{
	struct mapped_device *md = q->queuedata;

	if (dm_request_based(md))
1475 1476 1477
		blk_queue_bio(q, bio);
	else
		_dm_request(q, bio);
1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497
}

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;
1498 1499
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);
1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511

	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)
{
1512
	int r;
1513

1514 1515 1516 1517
	r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;
1518

1519 1520 1521 1522
	clone->cmd = rq->cmd;
	clone->cmd_len = rq->cmd_len;
	clone->sense = rq->sense;
	clone->buffer = rq->buffer;
1523 1524 1525 1526 1527 1528
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	return 0;
}

K
Kiyoshi Ueda 已提交
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
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;
}

1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567
/*
 * 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
1568 1569
	clone = clone_rq(rq, md, GFP_ATOMIC);
	if (!clone)
1570 1571 1572 1573 1574 1575 1576 1577
		return BLKPREP_DEFER;

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

	return BLKPREP_OK;
}

1578 1579 1580 1581 1582 1583 1584
/*
 * 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)
1585
{
1586
	int r, requeued = 0;
1587 1588 1589 1590 1591 1592 1593 1594 1595 1596
	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 */
1597 1598
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(tio->orig));
1599 1600 1601 1602 1603
		dm_dispatch_request(clone);
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
1604
		requeued = 1;
1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615
		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;
	}
1616 1617

	return requeued;
1618 1619
}

1620 1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639
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;
}

1640 1641 1642 1643 1644 1645 1646
/*
 * 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;
1647
	struct dm_table *map = dm_get_live_table(md);
1648
	struct dm_target *ti;
1649
	struct request *rq, *clone;
1650
	sector_t pos;
1651 1652

	/*
1653 1654 1655 1656
	 * 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().
1657
	 */
J
Jens Axboe 已提交
1658
	while (!blk_queue_stopped(q)) {
1659 1660
		rq = blk_peek_request(q);
		if (!rq)
J
Jens Axboe 已提交
1661
			goto delay_and_out;
1662

1663 1664 1665 1666 1667 1668
		/* 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);
1669 1670 1671 1672 1673 1674 1675 1676 1677 1678
		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;
		}
1679

1680
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
1681
			goto delay_and_out;
1682

1683
		clone = dm_start_request(md, rq);
1684

1685
		spin_unlock(q->queue_lock);
1686 1687 1688
		if (map_request(ti, clone, md))
			goto requeued;

1689 1690
		BUG_ON(!irqs_disabled());
		spin_lock(q->queue_lock);
1691 1692 1693 1694
	}

	goto out;

1695
requeued:
1696 1697
	BUG_ON(!irqs_disabled());
	spin_lock(q->queue_lock);
1698

J
Jens Axboe 已提交
1699 1700
delay_and_out:
	blk_delay_queue(q, HZ / 10);
1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714
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;
1715
	struct dm_table *map = dm_get_live_table(md);
1716 1717 1718 1719 1720 1721 1722 1723 1724 1725 1726

	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
1727 1728
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1729 1730 1731
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1732

1733
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
1734
		map = dm_get_live_table(md);
1735
		if (map) {
1736 1737 1738 1739 1740 1741 1742 1743 1744 1745
			/*
			 * 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);

1746 1747 1748 1749
			dm_table_put(map);
		}
	}

L
Linus Torvalds 已提交
1750 1751 1752 1753 1754 1755
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1756
static void free_minor(int minor)
L
Linus Torvalds 已提交
1757
{
1758
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1759
	idr_remove(&_minor_idr, minor);
1760
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1761 1762 1763 1764 1765
}

/*
 * See if the device with a specific minor # is free.
 */
1766
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1767 1768 1769 1770 1771 1772
{
	int r, m;

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

J
Jeff Mahoney 已提交
1773 1774 1775 1776
	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
	if (!r)
		return -ENOMEM;

1777
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1778 1779 1780 1781 1782 1783

	if (idr_find(&_minor_idr, minor)) {
		r = -EBUSY;
		goto out;
	}

1784
	r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m);
J
Jeff Mahoney 已提交
1785
	if (r)
L
Linus Torvalds 已提交
1786 1787 1788 1789 1790 1791 1792 1793 1794
		goto out;

	if (m != minor) {
		idr_remove(&_minor_idr, m);
		r = -EBUSY;
		goto out;
	}

out:
1795
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1796 1797 1798
	return r;
}

1799
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1800
{
1801
	int r, m;
L
Linus Torvalds 已提交
1802 1803

	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
J
Jeff Mahoney 已提交
1804 1805 1806
	if (!r)
		return -ENOMEM;

1807
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1808

1809
	r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m);
1810
	if (r)
L
Linus Torvalds 已提交
1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821
		goto out;

	if (m >= (1 << MINORBITS)) {
		idr_remove(&_minor_idr, m);
		r = -ENOSPC;
		goto out;
	}

	*minor = m;

out:
1822
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1823 1824 1825
	return r;
}

1826
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1827

1828 1829
static void dm_wq_work(struct work_struct *work);

1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850
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
1851 1852 1853
/*
 * Allocate and initialise a blank device with a given minor.
 */
1854
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1855 1856
{
	int r;
1857
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
1858
	void *old_md;
L
Linus Torvalds 已提交
1859 1860 1861 1862 1863 1864

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

1865
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1866
		goto bad_module_get;
1867

L
Linus Torvalds 已提交
1868
	/* get a minor number for the dev */
1869
	if (minor == DM_ANY_MINOR)
1870
		r = next_free_minor(&minor);
1871
	else
1872
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1873
	if (r < 0)
M
Milan Broz 已提交
1874
		goto bad_minor;
L
Linus Torvalds 已提交
1875

1876
	md->type = DM_TYPE_NONE;
1877
	init_rwsem(&md->io_lock);
1878
	mutex_init(&md->suspend_lock);
1879
	mutex_init(&md->type_lock);
1880
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1881 1882
	rwlock_init(&md->map_lock);
	atomic_set(&md->holders, 1);
1883
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1884
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1885 1886 1887
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1888

1889
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
1890
	if (!md->queue)
M
Milan Broz 已提交
1891
		goto bad_queue;
L
Linus Torvalds 已提交
1892

1893
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1894

L
Linus Torvalds 已提交
1895 1896
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
1897
		goto bad_disk;
L
Linus Torvalds 已提交
1898

1899 1900
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1901
	init_waitqueue_head(&md->wait);
1902
	INIT_WORK(&md->work, dm_wq_work);
1903 1904
	init_waitqueue_head(&md->eventq);

L
Linus Torvalds 已提交
1905 1906 1907 1908 1909 1910 1911
	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
1912
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1913

1914 1915
	md->wq = alloc_workqueue("kdmflush",
				 WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0);
1916 1917 1918
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
1919 1920 1921 1922
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

1923 1924 1925 1926
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
	md->flush_bio.bi_rw = WRITE_FLUSH;

1927
	/* Populate the mapping, nobody knows we exist yet */
1928
	spin_lock(&_minor_lock);
1929
	old_md = idr_replace(&_minor_idr, md, minor);
1930
	spin_unlock(&_minor_lock);
1931 1932 1933

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1934 1935
	return md;

M
Mikulas Patocka 已提交
1936 1937
bad_bdev:
	destroy_workqueue(md->wq);
1938
bad_thread:
1939
	del_gendisk(md->disk);
1940
	put_disk(md->disk);
M
Milan Broz 已提交
1941
bad_disk:
1942
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
1943
bad_queue:
L
Linus Torvalds 已提交
1944
	free_minor(minor);
M
Milan Broz 已提交
1945
bad_minor:
1946
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1947
bad_module_get:
L
Linus Torvalds 已提交
1948 1949 1950 1951
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1952 1953
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1954 1955
static void free_dev(struct mapped_device *md)
{
1956
	int minor = MINOR(disk_devt(md->disk));
1957

M
Mikulas Patocka 已提交
1958 1959
	unlock_fs(md);
	bdput(md->bdev);
1960
	destroy_workqueue(md->wq);
K
Kiyoshi Ueda 已提交
1961 1962 1963 1964 1965 1966
	if (md->tio_pool)
		mempool_destroy(md->tio_pool);
	if (md->io_pool)
		mempool_destroy(md->io_pool);
	if (md->bs)
		bioset_free(md->bs);
M
Martin K. Petersen 已提交
1967
	blk_integrity_unregister(md->disk);
L
Linus Torvalds 已提交
1968
	del_gendisk(md->disk);
1969
	free_minor(minor);
J
Jeff Mahoney 已提交
1970 1971 1972 1973 1974

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

L
Linus Torvalds 已提交
1975
	put_disk(md->disk);
1976
	blk_cleanup_queue(md->queue);
1977
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1978 1979 1980
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1981 1982 1983 1984
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
	struct dm_md_mempools *p;

1985
	if (md->io_pool && (md->tio_pool || dm_table_get_type(t) == DM_TYPE_BIO_BASED) && md->bs)
K
Kiyoshi Ueda 已提交
1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003
		/* the md already has necessary mempools */
		goto out;

	p = dm_table_get_md_mempools(t);
	BUG_ON(!p || md->io_pool || md->tio_pool || md->bs);

	md->io_pool = p->io_pool;
	p->io_pool = NULL;
	md->tio_pool = p->tio_pool;
	p->tio_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
2004 2005 2006 2007 2008
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2009 2010
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2011 2012
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2013 2014 2015 2016
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2017
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2018

L
Linus Torvalds 已提交
2019 2020 2021 2022
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2023 2024 2025
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2026
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2027
{
2028
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2029

2030
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2031 2032
}

2033 2034 2035 2036 2037 2038 2039 2040 2041 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
/*
 * 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;
}

2086 2087 2088 2089 2090
/*
 * 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
2091
{
2092
	struct dm_table *old_map;
2093
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2094
	sector_t size;
2095
	unsigned long flags;
2096
	int merge_is_optional;
L
Linus Torvalds 已提交
2097 2098

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2099 2100 2101 2102 2103 2104 2105

	/*
	 * 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
2106
	__set_size(md, size);
2107

2108 2109
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2110 2111 2112 2113 2114 2115 2116 2117 2118 2119 2120 2121
	/*
	 * 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);

2122 2123
	merge_is_optional = dm_table_merge_is_optional(t);

2124
	write_lock_irqsave(&md->map_lock, flags);
2125
	old_map = md->map;
L
Linus Torvalds 已提交
2126
	md->map = t;
2127 2128
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2129
	dm_table_set_restrictions(t, q, limits);
2130 2131 2132 2133
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
2134
	write_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
2135

2136
	return old_map;
L
Linus Torvalds 已提交
2137 2138
}

2139 2140 2141 2142
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2143 2144
{
	struct dm_table *map = md->map;
2145
	unsigned long flags;
L
Linus Torvalds 已提交
2146 2147

	if (!map)
2148
		return NULL;
L
Linus Torvalds 已提交
2149 2150

	dm_table_event_callback(map, NULL, NULL);
2151
	write_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
2152
	md->map = NULL;
2153
	write_unlock_irqrestore(&md->map_lock, flags);
2154 2155

	return map;
L
Linus Torvalds 已提交
2156 2157 2158 2159 2160
}

/*
 * Constructor for a new device.
 */
2161
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2162 2163 2164
{
	struct mapped_device *md;

2165
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2166 2167 2168
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2169 2170
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2171 2172 2173 2174
	*result = md;
	return 0;
}

2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198
/*
 * 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;
}

2199 2200 2201 2202 2203
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 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
/*
 * 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;
}

2244
static struct mapped_device *dm_find_md(dev_t dev)
L
Linus Torvalds 已提交
2245 2246 2247 2248 2249 2250 2251
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2252
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2253 2254

	md = idr_find(&_minor_idr, minor);
J
Jeff Mahoney 已提交
2255
	if (md && (md == MINOR_ALLOCED ||
2256
		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
2257
		   dm_deleting_md(md) ||
A
Alasdair G Kergon 已提交
2258
		   test_bit(DMF_FREEING, &md->flags))) {
2259
		md = NULL;
J
Jeff Mahoney 已提交
2260 2261
		goto out;
	}
L
Linus Torvalds 已提交
2262

J
Jeff Mahoney 已提交
2263
out:
2264
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2265

2266 2267 2268
	return md;
}

2269 2270 2271 2272 2273 2274 2275 2276 2277
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
2278
EXPORT_SYMBOL_GPL(dm_get_md);
2279

A
Alasdair G Kergon 已提交
2280
void *dm_get_mdptr(struct mapped_device *md)
2281
{
A
Alasdair G Kergon 已提交
2282
	return md->interface_ptr;
L
Linus Torvalds 已提交
2283 2284 2285 2286 2287 2288 2289 2290 2291 2292
}

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);
2293
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2294 2295
}

2296 2297 2298 2299 2300 2301
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2302
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2303
{
M
Mike Anderson 已提交
2304
	struct dm_table *map;
L
Linus Torvalds 已提交
2305

2306
	might_sleep();
J
Jeff Mahoney 已提交
2307

2308 2309 2310 2311 2312 2313 2314 2315 2316
	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
2317
	}
2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350

	/*
	 * 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
2351
}
E
Edward Goggin 已提交
2352
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2353

2354
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2355 2356
{
	int r = 0;
2357 2358 2359
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2360 2361

	while (1) {
2362
		set_current_state(interruptible);
2363

2364
		if (!md_in_flight(md))
2365 2366
			break;

2367 2368
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2369 2370 2371 2372 2373 2374 2375 2376
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2377 2378
	remove_wait_queue(&md->wait, &wait);

2379 2380 2381
	return r;
}

L
Linus Torvalds 已提交
2382 2383 2384
/*
 * Process the deferred bios
 */
2385
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2386
{
2387 2388
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2389
	struct bio *c;
L
Linus Torvalds 已提交
2390

2391
	down_read(&md->io_lock);
2392

2393
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2394 2395 2396 2397
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2398
		if (!c)
A
Alasdair G Kergon 已提交
2399
			break;
2400

2401
		up_read(&md->io_lock);
2402

K
Kiyoshi Ueda 已提交
2403 2404
		if (dm_request_based(md))
			generic_make_request(c);
2405 2406
		else
			__split_and_process_bio(md, c);
2407

2408
		down_read(&md->io_lock);
2409
	}
M
Milan Broz 已提交
2410

2411
	up_read(&md->io_lock);
L
Linus Torvalds 已提交
2412 2413
}

2414
static void dm_queue_flush(struct mapped_device *md)
2415
{
2416 2417
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	smp_mb__after_clear_bit();
2418
	queue_work(md->wq, &md->work);
2419 2420
}

L
Linus Torvalds 已提交
2421
/*
2422
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2423
 */
2424
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2425
{
2426
	struct dm_table *live_map, *map = ERR_PTR(-EINVAL);
2427
	struct queue_limits limits;
2428
	int r;
L
Linus Torvalds 已提交
2429

2430
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2431 2432

	/* device must be suspended */
2433
	if (!dm_suspended_md(md))
2434
		goto out;
L
Linus Torvalds 已提交
2435

2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448
	/*
	 * 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);
	}

2449
	r = dm_calculate_queue_limits(table, &limits);
2450 2451
	if (r) {
		map = ERR_PTR(r);
2452
		goto out;
2453
	}
2454

2455
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2456

2457
out:
2458
	mutex_unlock(&md->suspend_lock);
2459
	return map;
L
Linus Torvalds 已提交
2460 2461 2462 2463 2464 2465
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2466
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2467
{
2468
	int r;
L
Linus Torvalds 已提交
2469 2470

	WARN_ON(md->frozen_sb);
2471

2472
	md->frozen_sb = freeze_bdev(md->bdev);
2473
	if (IS_ERR(md->frozen_sb)) {
2474
		r = PTR_ERR(md->frozen_sb);
2475 2476
		md->frozen_sb = NULL;
		return r;
2477 2478
	}

2479 2480
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2481 2482 2483
	return 0;
}

2484
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2485
{
2486 2487 2488
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2489
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2490
	md->frozen_sb = NULL;
2491
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2492 2493 2494 2495 2496 2497 2498 2499 2500
}

/*
 * 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.
 */
2501 2502 2503
/*
 * Suspend mechanism in request-based dm.
 *
K
Kiyoshi Ueda 已提交
2504 2505 2506
 * 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.
2507
 *
K
Kiyoshi Ueda 已提交
2508
 * To abort suspend, start the request_queue.
2509
 */
2510
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
L
Linus Torvalds 已提交
2511
{
2512
	struct dm_table *map = NULL;
2513
	int r = 0;
2514
	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
2515
	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
L
Linus Torvalds 已提交
2516

2517
	mutex_lock(&md->suspend_lock);
2518

2519
	if (dm_suspended_md(md)) {
M
Milan Broz 已提交
2520
		r = -EINVAL;
2521
		goto out_unlock;
M
Milan Broz 已提交
2522
	}
L
Linus Torvalds 已提交
2523

2524
	map = dm_get_live_table(md);
L
Linus Torvalds 已提交
2525

2526 2527 2528 2529 2530 2531 2532
	/*
	 * 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);

2533 2534 2535
	/* This does not get reverted if there's an error later. */
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2536
	/*
K
Kiyoshi Ueda 已提交
2537 2538 2539 2540
	 * 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
2541 2542 2543 2544
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
		if (r)
K
Kiyoshi Ueda 已提交
2545
			goto out;
2546
	}
L
Linus Torvalds 已提交
2547 2548

	/*
2549 2550 2551 2552 2553 2554 2555
	 * 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
2556 2557 2558
	 * __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
2559
	 */
2560
	down_write(&md->io_lock);
2561
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2562
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
2563

2564
	/*
2565 2566
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2567
	 */
2568
	if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2569
		stop_queue(md->queue);
2570

2571 2572
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2573
	/*
2574 2575 2576
	 * 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
2577
	 */
2578
	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
L
Linus Torvalds 已提交
2579

2580
	down_write(&md->io_lock);
2581
	if (noflush)
2582
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2583
	up_write(&md->io_lock);
2584

L
Linus Torvalds 已提交
2585
	/* were we interrupted ? */
2586
	if (r < 0) {
2587
		dm_queue_flush(md);
M
Milan Broz 已提交
2588

2589
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2590
			start_queue(md->queue);
2591

2592
		unlock_fs(md);
2593
		goto out; /* pushback list is already flushed, so skip flush */
2594
	}
L
Linus Torvalds 已提交
2595

2596 2597 2598 2599 2600 2601
	/*
	 * 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.
	 */

2602
	set_bit(DMF_SUSPENDED, &md->flags);
2603

2604 2605
	dm_table_postsuspend_targets(map);

2606 2607
out:
	dm_table_put(map);
2608 2609

out_unlock:
2610
	mutex_unlock(&md->suspend_lock);
2611
	return r;
L
Linus Torvalds 已提交
2612 2613 2614 2615
}

int dm_resume(struct mapped_device *md)
{
2616 2617
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2618

2619
	mutex_lock(&md->suspend_lock);
2620
	if (!dm_suspended_md(md))
2621 2622
		goto out;

2623
	map = dm_get_live_table(md);
2624
	if (!map || !dm_table_get_size(map))
2625
		goto out;
L
Linus Torvalds 已提交
2626

2627 2628 2629
	r = dm_table_resume_targets(map);
	if (r)
		goto out;
2630

2631
	dm_queue_flush(md);
2632

2633 2634 2635 2636 2637 2638 2639 2640
	/*
	 * 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);

2641 2642 2643 2644
	unlock_fs(md);

	clear_bit(DMF_SUSPENDED, &md->flags);

2645 2646 2647
	r = 0;
out:
	dm_table_put(map);
2648
	mutex_unlock(&md->suspend_lock);
2649

2650
	return r;
L
Linus Torvalds 已提交
2651 2652 2653 2654 2655
}

/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2656
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2657
		       unsigned cookie)
2658
{
M
Milan Broz 已提交
2659 2660 2661 2662
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2663
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2664 2665 2666
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2667 2668
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2669
	}
2670 2671
}

M
Mike Anderson 已提交
2672 2673 2674 2675 2676
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687
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
2688 2689 2690 2691 2692 2693 2694 2695 2696
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
2697 2698 2699 2700 2701 2702 2703 2704 2705
/*
 * 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
2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722
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;

2723
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2724
	    dm_deleting_md(md))
2725 2726
		return NULL;

M
Milan Broz 已提交
2727 2728 2729 2730
	dm_get(md);
	return md;
}

2731
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2732 2733 2734 2735
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2736 2737
int dm_suspended(struct dm_target *ti)
{
2738
	return dm_suspended_md(dm_table_get_md(ti->table));
2739 2740 2741
}
EXPORT_SYMBOL_GPL(dm_suspended);

2742 2743
int dm_noflush_suspending(struct dm_target *ti)
{
2744
	return __noflush_suspending(dm_table_get_md(ti->table));
2745 2746 2747
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

2748
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity)
K
Kiyoshi Ueda 已提交
2749 2750
{
	struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL);
2751
	unsigned int pool_size = (type == DM_TYPE_BIO_BASED) ? 16 : MIN_IOS;
K
Kiyoshi Ueda 已提交
2752 2753 2754 2755 2756 2757 2758 2759 2760 2761

	if (!pools)
		return NULL;

	pools->io_pool = (type == DM_TYPE_BIO_BASED) ?
			 mempool_create_slab_pool(MIN_IOS, _io_cache) :
			 mempool_create_slab_pool(MIN_IOS, _rq_bio_info_cache);
	if (!pools->io_pool)
		goto free_pools_and_out;

2762 2763 2764 2765 2766 2767
	pools->tio_pool = NULL;
	if (type == DM_TYPE_REQUEST_BASED) {
		pools->tio_pool = mempool_create_slab_pool(MIN_IOS, _rq_tio_cache);
		if (!pools->tio_pool)
			goto free_io_pool_and_out;
	}
K
Kiyoshi Ueda 已提交
2768

2769
	pools->bs = (type == DM_TYPE_BIO_BASED) ?
2770 2771
		bioset_create(pool_size,
			      offsetof(struct dm_target_io, clone)) :
2772 2773
		bioset_create(pool_size,
			      offsetof(struct dm_rq_clone_bio_info, clone));
K
Kiyoshi Ueda 已提交
2774 2775 2776
	if (!pools->bs)
		goto free_tio_pool_and_out;

2777 2778 2779
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
		goto free_bioset_and_out;

K
Kiyoshi Ueda 已提交
2780 2781
	return pools;

2782 2783 2784
free_bioset_and_out:
	bioset_free(pools->bs);

K
Kiyoshi Ueda 已提交
2785
free_tio_pool_and_out:
2786 2787
	if (pools->tio_pool)
		mempool_destroy(pools->tio_pool);
K
Kiyoshi Ueda 已提交
2788 2789 2790 2791 2792 2793 2794 2795 2796 2797 2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814

free_io_pool_and_out:
	mempool_destroy(pools->io_pool);

free_pools_and_out:
	kfree(pools);

	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->tio_pool)
		mempool_destroy(pools->tio_pool);

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

	kfree(pools);
}

2815
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2816 2817
	.open = dm_blk_open,
	.release = dm_blk_close,
2818
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2819
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835
	.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");