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
157 158 159 160 161 162 163 164 165 166 167
	/*
	 * 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
168 169
	struct bio_set *bs;

L
Linus Torvalds 已提交
170 171 172 173 174
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
175 176 177
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
178 179 180 181 182

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
183
	struct block_device *bdev;
D
Darrick J. Wong 已提交
184 185 186

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
187 188 189

	/* sysfs handle */
	struct kobject kobj;
190

191 192
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
L
Linus Torvalds 已提交
193 194
};

K
Kiyoshi Ueda 已提交
195 196 197 198 199 200 201 202 203
/*
 * 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
204
#define MIN_IOS 256
205
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
206
static struct kmem_cache *_rq_tio_cache;
207 208 209 210 211

/*
 * 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
212
static struct kmem_cache *_rq_bio_info_cache;
L
Linus Torvalds 已提交
213 214 215

static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
216
	int r = -ENOMEM;
L
Linus Torvalds 已提交
217 218

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
219
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
220
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
221
		return r;
L
Linus Torvalds 已提交
222

K
Kiyoshi Ueda 已提交
223 224
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
225
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
226 227 228 229 230

	_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
231
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
232
	if (r)
K
Kiyoshi Ueda 已提交
233
		goto out_free_rq_bio_info_cache;
M
Mike Anderson 已提交
234

L
Linus Torvalds 已提交
235 236
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
237 238
	if (r < 0)
		goto out_uevent_exit;
L
Linus Torvalds 已提交
239 240 241 242 243

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
244 245 246

out_uevent_exit:
	dm_uevent_exit();
K
Kiyoshi Ueda 已提交
247 248 249 250
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
251 252 253 254
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
255 256 257 258
}

static void local_exit(void)
{
K
Kiyoshi Ueda 已提交
259 260
	kmem_cache_destroy(_rq_bio_info_cache);
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
261
	kmem_cache_destroy(_io_cache);
262
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
263
	dm_uevent_exit();
L
Linus Torvalds 已提交
264 265 266 267 268 269

	_major = 0;

	DMINFO("cleaned up");
}

270
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
271 272 273 274
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
275
	dm_io_init,
276
	dm_kcopyd_init,
L
Linus Torvalds 已提交
277 278 279
	dm_interface_init,
};

280
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
281 282 283 284
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
285
	dm_io_exit,
286
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316
	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]();
317 318 319 320 321

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
322 323 324 325 326
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
327 328 329 330 331
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
332
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
333 334 335
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
336 337
	spin_lock(&_minor_lock);

A
Al Viro 已提交
338
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
339 340 341
	if (!md)
		goto out;

342
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
343
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
344 345 346 347
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
348
	dm_get(md);
349
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
350 351 352 353 354

out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
355 356
}

A
Al Viro 已提交
357
static int dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
358
{
A
Al Viro 已提交
359
	struct mapped_device *md = disk->private_data;
360

361 362
	spin_lock(&_minor_lock);

363
	atomic_dec(&md->open_count);
L
Linus Torvalds 已提交
364
	dm_put(md);
365 366

	spin_unlock(&_minor_lock);
367

L
Linus Torvalds 已提交
368 369 370
	return 0;
}

371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394
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
395 396 397 398 399 400 401
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
402
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
403 404
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
405
	struct mapped_device *md = bdev->bd_disk->private_data;
406
	struct dm_table *map = dm_get_live_table(md);
407 408 409 410 411 412 413 414 415 416 417 418
	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);

419
	if (dm_suspended_md(md)) {
420 421 422 423 424
		r = -EAGAIN;
		goto out;
	}

	if (tgt->type->ioctl)
425
		r = tgt->type->ioctl(tgt, cmd, arg);
426 427 428 429 430 431 432

out:
	dm_table_put(map);

	return r;
}

A
Alasdair G Kergon 已提交
433
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
434 435 436 437
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
438
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
439 440 441 442
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
443
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
444
{
445
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
446 447
}

K
Kiyoshi Ueda 已提交
448 449
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
450
{
K
Kiyoshi Ueda 已提交
451
	return mempool_alloc(md->tio_pool, gfp_mask);
452 453 454 455 456 457 458
}

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

K
Kiyoshi Ueda 已提交
459 460 461 462 463 464
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

465 466 467
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
T
Tejun Heo 已提交
468
	int cpu;
469
	int rw = bio_data_dir(io->bio);
470 471 472

	io->start_time = jiffies;

T
Tejun Heo 已提交
473 474 475
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
476 477
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
478 479
}

480
static void end_io_acct(struct dm_io *io)
481 482 483 484
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
T
Tejun Heo 已提交
485
	int pending, cpu;
486 487
	int rw = bio_data_dir(bio);

T
Tejun Heo 已提交
488 489 490 491
	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();
492

493 494
	/*
	 * After this is decremented the bio must not be touched if it is
495
	 * a flush.
496
	 */
497 498
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
499
	pending += atomic_read(&md->pending[rw^0x1]);
500

501 502 503
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
504 505
}

L
Linus Torvalds 已提交
506 507 508
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
509
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
510
{
511
	unsigned long flags;
L
Linus Torvalds 已提交
512

513
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
514
	bio_list_add(&md->deferred, bio);
515
	spin_unlock_irqrestore(&md->deferred_lock, flags);
516
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
517 518 519 520 521 522 523
}

/*
 * 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.
 */
524
struct dm_table *dm_get_live_table(struct mapped_device *md)
L
Linus Torvalds 已提交
525 526
{
	struct dm_table *t;
527
	unsigned long flags;
L
Linus Torvalds 已提交
528

529
	read_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
530 531 532
	t = md->map;
	if (t)
		dm_table_get(t);
533
	read_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
534 535 536 537

	return t;
}

D
Darrick J. Wong 已提交
538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564
/*
 * 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
565 566 567 568 569 570 571 572 573
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

574 575 576 577 578
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
579 580 581 582
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
583
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
584
{
585
	unsigned long flags;
586 587 588
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
589 590

	/* Push-back supersedes any I/O errors */
591 592 593 594 595 596
	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
597 598

	if (atomic_dec_and_test(&io->io_count)) {
599 600 601 602
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
603
			spin_lock_irqsave(&md->deferred_lock, flags);
604 605 606
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
607 608
				/* noflush suspend was interrupted. */
				io->error = -EIO;
609
			spin_unlock_irqrestore(&md->deferred_lock, flags);
610 611
		}

612 613
		io_error = io->error;
		bio = io->bio;
614 615 616 617 618
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
619

620
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_size) {
621
			/*
622 623
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
624
			 */
625 626
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
627
		} else {
628 629
			/* done with normal IO or empty flush */
			bio_endio(bio, io_error);
630
		}
L
Linus Torvalds 已提交
631 632 633
	}
}

634
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
635 636
{
	int r = 0;
A
Alasdair G Kergon 已提交
637
	struct dm_target_io *tio = bio->bi_private;
638
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
639
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
640 641 642 643 644 645
	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
646
		r = endio(tio->ti, bio, error);
647 648 649 650 651
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
652
			error = r;
653 654
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
655
			return;
656 657 658 659
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
660 661
	}

S
Stefan Bader 已提交
662
	free_tio(md, tio);
663
	dec_pending(io, error);
L
Linus Torvalds 已提交
664 665
}

666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720
/*
 * 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.
 */
721
static void rq_completed(struct mapped_device *md, int rw, int run_queue)
722
{
723
	atomic_dec(&md->pending[rw]);
724 725

	/* nudge anyone waiting on suspend queue */
726
	if (!md_in_flight(md))
727 728
		wake_up(&md->wait);

729 730 731 732 733 734
	/*
	 * 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.
	 */
735
	if (run_queue)
736
		blk_run_queue_async(md->queue);
737 738 739 740 741 742 743

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

744 745 746 747 748 749 750 751
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
752 753 754 755 756 757 758 759 760 761 762
/*
 * 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;

763
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
764 765 766 767 768 769 770 771 772 773 774 775 776
		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);
777 778
	blk_end_request_all(rq, error);
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
779 780
}

781 782 783 784 785 786 787
static void dm_unprep_request(struct request *rq)
{
	struct request *clone = rq->special;

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

788
	free_rq_clone(clone);
789 790 791 792 793 794 795
}

/*
 * Requeue the original request of a clone.
 */
void dm_requeue_unmapped_request(struct request *clone)
{
796
	int rw = rq_data_dir(clone);
797 798 799 800 801 802 803 804 805 806 807 808
	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);

809
	rq_completed(md, rw, 0);
810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841
}
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);
}

842
static void dm_done(struct request *clone, int error, bool mapped)
843
{
844
	int r = error;
845
	struct dm_rq_target_io *tio = clone->end_io_data;
846
	dm_request_endio_fn rq_end_io = NULL;
847

848 849 850 851 852 853
	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);
	}
854

855
	if (r <= 0)
856
		/* The target wants to complete the I/O */
857 858
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
859 860
		/* The target will handle the I/O */
		return;
861
	else if (r == DM_ENDIO_REQUEUE)
862 863 864
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
865
		DMWARN("unimplemented target endio return value: %d", r);
866 867 868 869
		BUG();
	}
}

870 871 872 873 874 875 876 877 878 879 880 881 882 883 884
/*
 * 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);
}

885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938
/*
 * 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);
}

939 940 941 942 943 944 945 946 947 948 949 950
/*
 * 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
951
{
952
	sector_t len = max_io_len_target_boundary(sector, ti);
953
	sector_t offset, max_len;
L
Linus Torvalds 已提交
954 955

	/*
956
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
957
	 */
958 959 960 961 962 963 964 965 966 967
	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
968 969 970 971 972
	}

	return len;
}

973 974 975 976 977 978 979 980 981 982 983 984 985 986 987
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
988
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
989 990
{
	int r;
991
	sector_t sector;
S
Stefan Bader 已提交
992
	struct mapped_device *md;
993
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
994
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
995 996 997 998 999 1000 1001 1002 1003 1004

	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);
1005
	sector = clone->bi_sector;
M
Mikulas Patocka 已提交
1006
	r = ti->type->map(ti, clone);
1007
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1008
		/* the bio has been remapped so dispatch it */
1009

1010 1011
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
1012

L
Linus Torvalds 已提交
1013
		generic_make_request(clone);
1014 1015
	} 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
1016 1017 1018
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1019 1020 1021
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034
	}
}

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
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060
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
1061
/*
1062
 * Creates a little bio that just does part of a bvec.
L
Linus Torvalds 已提交
1063
 */
1064 1065
static void split_bvec(struct dm_target_io *tio, struct bio *bio,
		       sector_t sector, unsigned short idx, unsigned int offset,
A
Alasdair G Kergon 已提交
1066
		       unsigned int len)
L
Linus Torvalds 已提交
1067
{
1068
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1069 1070 1071 1072
	struct bio_vec *bv = bio->bi_io_vec + idx;

	*clone->bi_io_vec = *bv;

A
Alasdair G Kergon 已提交
1073 1074
	bio_setup_sector(clone, sector, len);

L
Linus Torvalds 已提交
1075
	clone->bi_bdev = bio->bi_bdev;
1076
	clone->bi_rw = bio->bi_rw;
L
Linus Torvalds 已提交
1077 1078 1079
	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
1080
	clone->bi_flags |= 1 << BIO_CLONED;
L
Linus Torvalds 已提交
1081

A
Alasdair G Kergon 已提交
1082
	clone_bio_integrity(bio, clone, idx, len, offset, 1);
L
Linus Torvalds 已提交
1083 1084 1085 1086 1087
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1088 1089
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
1090
		      unsigned short bv_count, unsigned int len)
L
Linus Torvalds 已提交
1091
{
1092
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1093
	unsigned trim = 0;
L
Linus Torvalds 已提交
1094

S
Stefan Bader 已提交
1095
	__bio_clone(clone, bio);
A
Alasdair G Kergon 已提交
1096 1097 1098 1099 1100 1101
	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
1102 1103
}

1104
static struct dm_target_io *alloc_tio(struct clone_info *ci,
A
Alasdair G Kergon 已提交
1105 1106
				      struct dm_target *ti, int nr_iovecs,
				      unsigned target_request_nr)
1107
{
1108 1109 1110 1111 1112
	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);
1113 1114 1115 1116

	tio->io = ci->io;
	tio->ti = ti;
	memset(&tio->info, 0, sizeof(tio->info));
A
Alasdair G Kergon 已提交
1117
	tio->target_request_nr = target_request_nr;
1118 1119 1120 1121

	return tio;
}

1122
static void __issue_target_request(struct clone_info *ci, struct dm_target *ti,
1123
				   unsigned request_nr, sector_t len)
1124
{
A
Alasdair G Kergon 已提交
1125
	struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs, request_nr);
1126
	struct bio *clone = &tio->clone;
1127

1128 1129 1130 1131 1132
	/*
	 * 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.
	 */
1133
	 __bio_clone(clone, ci->bio);
A
Alasdair G Kergon 已提交
1134 1135
	if (len)
		bio_setup_sector(clone, ci->sector, len);
1136

A
Alasdair G Kergon 已提交
1137
	__map_bio(tio);
1138 1139
}

1140
static void __issue_target_requests(struct clone_info *ci, struct dm_target *ti,
1141
				    unsigned num_requests, sector_t len)
1142 1143 1144 1145
{
	unsigned request_nr;

	for (request_nr = 0; request_nr < num_requests; request_nr++)
1146
		__issue_target_request(ci, ti, request_nr, len);
1147 1148
}

1149
static int __clone_and_map_empty_flush(struct clone_info *ci)
1150
{
1151
	unsigned target_nr = 0;
1152 1153
	struct dm_target *ti;

1154
	BUG_ON(bio_has_data(ci->bio));
1155
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1156
		__issue_target_requests(ci, ti, ti->num_flush_requests, 0);
1157 1158 1159 1160

	return 0;
}

M
Mike Snitzer 已提交
1161 1162 1163 1164 1165
/*
 * Perform all io with a single clone.
 */
static void __clone_and_map_simple(struct clone_info *ci, struct dm_target *ti)
{
1166
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1167 1168
	struct dm_target_io *tio;

A
Alasdair G Kergon 已提交
1169
	tio = alloc_tio(ci, ti, bio->bi_max_vecs, 0);
1170
	clone_bio(tio, bio, ci->sector, ci->idx, bio->bi_vcnt - ci->idx,
A
Alasdair G Kergon 已提交
1171 1172
		  ci->sector_count);
	__map_bio(tio);
M
Mike Snitzer 已提交
1173 1174 1175
	ci->sector_count = 0;
}

M
Mike Snitzer 已提交
1176 1177 1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197
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
1198 1199
{
	struct dm_target *ti;
1200
	sector_t len;
1201
	unsigned num_requests;
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
		 */
1214 1215
		num_requests = get_num_requests ? get_num_requests(ti) : 0;
		if (!num_requests)
1216
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1217

M
Mike Snitzer 已提交
1218
		if (is_split_required && !is_split_required(ti))
1219 1220 1221
			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));
1222

1223
		__issue_target_requests(ci, ti, num_requests, len);
1224 1225 1226

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1227 1228 1229 1230

	return 0;
}

M
Mike Snitzer 已提交
1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241
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);
}

1242
static int __clone_and_map(struct clone_info *ci)
L
Linus Torvalds 已提交
1243
{
1244
	struct bio *bio = ci->bio;
1245 1246
	struct dm_target *ti;
	sector_t len = 0, max;
A
Alasdair G Kergon 已提交
1247
	struct dm_target_io *tio;
L
Linus Torvalds 已提交
1248

M
Mike Snitzer 已提交
1249 1250
	if (unlikely(bio->bi_rw & REQ_DISCARD))
		return __clone_and_map_discard(ci);
M
Mike Snitzer 已提交
1251 1252
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
		return __clone_and_map_write_same(ci);
M
Mike Snitzer 已提交
1253

1254 1255 1256 1257
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1258
	max = max_io_len(ci->sector, ti);
1259

L
Linus Torvalds 已提交
1260 1261 1262 1263 1264
	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
1265
		__clone_and_map_simple(ci, ti);
L
Linus Torvalds 已提交
1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284 1285

	} 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;
		}

A
Alasdair G Kergon 已提交
1286 1287 1288
		tio = alloc_tio(ci, ti, bio->bi_max_vecs, 0);
		clone_bio(tio, bio, ci->sector, ci->idx, i - ci->idx, len);
		__map_bio(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);

A
Alasdair G Kergon 已提交
1313
			tio = alloc_tio(ci, ti, 1, 0);
1314
			split_bvec(tio, bio, ci->sector, ci->idx,
A
Alasdair G Kergon 已提交
1315
				   bv->bv_offset + offset, len);
1316

A
Alasdair G Kergon 已提交
1317
			__map_bio(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);
A
Alasdair G Kergon 已提交
1355

1356 1357 1358 1359 1360 1361
	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 {
1362
		ci.bio = bio;
1363
		ci.sector_count = bio_sectors(bio);
1364
		while (ci.sector_count && !error)
1365 1366
			error = __clone_and_map(&ci);
	}
L
Linus Torvalds 已提交
1367 1368

	/* drop the extra reference count */
1369
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1370 1371 1372 1373 1374 1375
	dm_table_put(ci.map);
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1376 1377 1378 1379 1380
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
1381
	struct dm_table *map = dm_get_live_table(md);
M
Milan Broz 已提交
1382 1383
	struct dm_target *ti;
	sector_t max_sectors;
1384
	int max_size = 0;
M
Milan Broz 已提交
1385 1386

	if (unlikely(!map))
1387
		goto out;
M
Milan Broz 已提交
1388 1389

	ti = dm_table_find_target(map, bvm->bi_sector);
1390 1391
	if (!dm_target_is_valid(ti))
		goto out_table;
M
Milan Broz 已提交
1392 1393 1394 1395

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1396
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
M
Milan Broz 已提交
1397 1398 1399 1400 1401 1402 1403 1404 1405 1406 1407 1408
			  (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);
1409 1410 1411 1412 1413 1414 1415 1416 1417 1418
	/*
	 * 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
1419

1420
out_table:
1421 1422 1423
	dm_table_put(map);

out:
M
Milan Broz 已提交
1424 1425 1426 1427 1428 1429 1430 1431 1432
	/*
	 * 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
1433 1434 1435 1436
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1437
static void _dm_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1438
{
1439
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1440
	struct mapped_device *md = q->queuedata;
T
Tejun Heo 已提交
1441
	int cpu;
L
Linus Torvalds 已提交
1442

1443
	down_read(&md->io_lock);
L
Linus Torvalds 已提交
1444

T
Tejun Heo 已提交
1445 1446 1447 1448
	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();
1449

1450 1451
	/* if we're suspended, we have to queue this io for later */
	if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
1452
		up_read(&md->io_lock);
L
Linus Torvalds 已提交
1453

1454 1455 1456
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1457
			bio_io_error(bio);
1458
		return;
L
Linus Torvalds 已提交
1459 1460
	}

1461
	__split_and_process_bio(md, bio);
1462
	up_read(&md->io_lock);
1463
	return;
1464 1465 1466 1467 1468 1469 1470
}

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

1471
static void dm_request(struct request_queue *q, struct bio *bio)
1472 1473 1474 1475
{
	struct mapped_device *md = q->queuedata;

	if (dm_request_based(md))
1476 1477 1478
		blk_queue_bio(q, bio);
	else
		_dm_request(q, bio);
1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498
}

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;
1499 1500
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);
1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512

	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)
{
1513
	int r;
1514

1515 1516 1517 1518
	r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;
1519

1520 1521 1522 1523
	clone->cmd = rq->cmd;
	clone->cmd_len = rq->cmd_len;
	clone->sense = rq->sense;
	clone->buffer = rq->buffer;
1524 1525 1526 1527 1528 1529
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	return 0;
}

K
Kiyoshi Ueda 已提交
1530 1531 1532 1533 1534 1535 1536 1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555
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;
}

1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
/*
 * 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
1569 1570
	clone = clone_rq(rq, md, GFP_ATOMIC);
	if (!clone)
1571 1572 1573 1574 1575 1576 1577 1578
		return BLKPREP_DEFER;

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

	return BLKPREP_OK;
}

1579 1580 1581 1582 1583 1584 1585
/*
 * 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)
1586
{
1587
	int r, requeued = 0;
1588 1589 1590 1591 1592 1593 1594 1595 1596 1597
	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 */
1598 1599
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(tio->orig));
1600 1601 1602 1603 1604
		dm_dispatch_request(clone);
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
1605
		requeued = 1;
1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616
		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;
	}
1617 1618

	return requeued;
1619 1620
}

1621 1622 1623 1624 1625 1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637 1638 1639 1640
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;
}

1641 1642 1643 1644 1645 1646 1647
/*
 * 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;
1648
	struct dm_table *map = dm_get_live_table(md);
1649
	struct dm_target *ti;
1650
	struct request *rq, *clone;
1651
	sector_t pos;
1652 1653

	/*
1654 1655 1656 1657
	 * 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().
1658
	 */
J
Jens Axboe 已提交
1659
	while (!blk_queue_stopped(q)) {
1660 1661
		rq = blk_peek_request(q);
		if (!rq)
J
Jens Axboe 已提交
1662
			goto delay_and_out;
1663

1664 1665 1666 1667 1668 1669
		/* 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);
1670 1671 1672 1673 1674 1675 1676 1677 1678 1679
		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;
		}
1680

1681
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
1682
			goto delay_and_out;
1683

1684
		clone = dm_start_request(md, rq);
1685

1686
		spin_unlock(q->queue_lock);
1687 1688 1689
		if (map_request(ti, clone, md))
			goto requeued;

1690 1691
		BUG_ON(!irqs_disabled());
		spin_lock(q->queue_lock);
1692 1693 1694 1695
	}

	goto out;

1696
requeued:
1697 1698
	BUG_ON(!irqs_disabled());
	spin_lock(q->queue_lock);
1699

J
Jens Axboe 已提交
1700 1701
delay_and_out:
	blk_delay_queue(q, HZ / 10);
1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715
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;
1716
	struct dm_table *map = dm_get_live_table(md);
1717 1718 1719 1720 1721 1722 1723 1724 1725 1726 1727

	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
1728 1729
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1730 1731 1732
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1733

1734
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
1735
		map = dm_get_live_table(md);
1736
		if (map) {
1737 1738 1739 1740 1741 1742 1743 1744 1745 1746
			/*
			 * 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);

1747 1748 1749 1750
			dm_table_put(map);
		}
	}

L
Linus Torvalds 已提交
1751 1752 1753 1754 1755 1756
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1757
static void free_minor(int minor)
L
Linus Torvalds 已提交
1758
{
1759
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1760
	idr_remove(&_minor_idr, minor);
1761
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1762 1763 1764 1765 1766
}

/*
 * See if the device with a specific minor # is free.
 */
1767
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1768
{
T
Tejun Heo 已提交
1769
	int r;
L
Linus Torvalds 已提交
1770 1771 1772 1773

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

T
Tejun Heo 已提交
1774
	idr_preload(GFP_KERNEL);
1775
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1776

T
Tejun Heo 已提交
1777
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1778

1779
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1780 1781 1782 1783
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1784 1785
}

1786
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1787
{
T
Tejun Heo 已提交
1788
	int r;
J
Jeff Mahoney 已提交
1789

T
Tejun Heo 已提交
1790
	idr_preload(GFP_KERNEL);
1791
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1792

T
Tejun Heo 已提交
1793
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1794

1795
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1796 1797 1798 1799 1800
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1801 1802
}

1803
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1804

1805 1806
static void dm_wq_work(struct work_struct *work);

1807 1808 1809 1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 1826 1827
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
1828 1829 1830
/*
 * Allocate and initialise a blank device with a given minor.
 */
1831
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1832 1833
{
	int r;
1834
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
1835
	void *old_md;
L
Linus Torvalds 已提交
1836 1837 1838 1839 1840 1841

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

1842
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1843
		goto bad_module_get;
1844

L
Linus Torvalds 已提交
1845
	/* get a minor number for the dev */
1846
	if (minor == DM_ANY_MINOR)
1847
		r = next_free_minor(&minor);
1848
	else
1849
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1850
	if (r < 0)
M
Milan Broz 已提交
1851
		goto bad_minor;
L
Linus Torvalds 已提交
1852

1853
	md->type = DM_TYPE_NONE;
1854
	init_rwsem(&md->io_lock);
1855
	mutex_init(&md->suspend_lock);
1856
	mutex_init(&md->type_lock);
1857
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1858 1859
	rwlock_init(&md->map_lock);
	atomic_set(&md->holders, 1);
1860
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1861
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1862 1863 1864
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1865

1866
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
1867
	if (!md->queue)
M
Milan Broz 已提交
1868
		goto bad_queue;
L
Linus Torvalds 已提交
1869

1870
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1871

L
Linus Torvalds 已提交
1872 1873
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
1874
		goto bad_disk;
L
Linus Torvalds 已提交
1875

1876 1877
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1878
	init_waitqueue_head(&md->wait);
1879
	INIT_WORK(&md->work, dm_wq_work);
1880 1881
	init_waitqueue_head(&md->eventq);

L
Linus Torvalds 已提交
1882 1883 1884 1885 1886 1887 1888
	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
1889
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1890

1891 1892
	md->wq = alloc_workqueue("kdmflush",
				 WQ_NON_REENTRANT | WQ_MEM_RECLAIM, 0);
1893 1894 1895
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
1896 1897 1898 1899
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

1900 1901 1902 1903
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
	md->flush_bio.bi_rw = WRITE_FLUSH;

1904
	/* Populate the mapping, nobody knows we exist yet */
1905
	spin_lock(&_minor_lock);
1906
	old_md = idr_replace(&_minor_idr, md, minor);
1907
	spin_unlock(&_minor_lock);
1908 1909 1910

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1911 1912
	return md;

M
Mikulas Patocka 已提交
1913 1914
bad_bdev:
	destroy_workqueue(md->wq);
1915
bad_thread:
1916
	del_gendisk(md->disk);
1917
	put_disk(md->disk);
M
Milan Broz 已提交
1918
bad_disk:
1919
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
1920
bad_queue:
L
Linus Torvalds 已提交
1921
	free_minor(minor);
M
Milan Broz 已提交
1922
bad_minor:
1923
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1924
bad_module_get:
L
Linus Torvalds 已提交
1925 1926 1927 1928
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1929 1930
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1931 1932
static void free_dev(struct mapped_device *md)
{
1933
	int minor = MINOR(disk_devt(md->disk));
1934

M
Mikulas Patocka 已提交
1935 1936
	unlock_fs(md);
	bdput(md->bdev);
1937
	destroy_workqueue(md->wq);
K
Kiyoshi Ueda 已提交
1938 1939 1940 1941 1942 1943
	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
1944
	blk_integrity_unregister(md->disk);
L
Linus Torvalds 已提交
1945
	del_gendisk(md->disk);
1946
	free_minor(minor);
J
Jeff Mahoney 已提交
1947 1948 1949 1950 1951

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

L
Linus Torvalds 已提交
1952
	put_disk(md->disk);
1953
	blk_cleanup_queue(md->queue);
1954
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1955 1956 1957
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1958 1959
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
1960
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
1961

1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982
	if (md->io_pool && md->bs) {
		/* 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) {
			BUG_ON(!md->tio_pool);
			/*
			 * 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
1983
		goto out;
M
Mikulas Patocka 已提交
1984
	}
K
Kiyoshi Ueda 已提交
1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999

	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
2000 2001 2002 2003 2004
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2005 2006
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2007 2008
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2009 2010 2011 2012
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2013
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2014

L
Linus Torvalds 已提交
2015 2016 2017 2018
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2019 2020 2021
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2022
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2023
{
2024
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2025

2026
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2027 2028
}

2029 2030 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
/*
 * 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;
}

2082 2083 2084 2085 2086
/*
 * 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
2087
{
2088
	struct dm_table *old_map;
2089
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2090
	sector_t size;
2091
	unsigned long flags;
2092
	int merge_is_optional;
L
Linus Torvalds 已提交
2093 2094

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2095 2096 2097 2098 2099 2100 2101

	/*
	 * 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
2102
	__set_size(md, size);
2103

2104 2105
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117
	/*
	 * 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);

2118 2119
	merge_is_optional = dm_table_merge_is_optional(t);

2120
	write_lock_irqsave(&md->map_lock, flags);
2121
	old_map = md->map;
L
Linus Torvalds 已提交
2122
	md->map = t;
2123 2124
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2125
	dm_table_set_restrictions(t, q, limits);
2126 2127 2128 2129
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
2130
	write_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
2131

2132
	return old_map;
L
Linus Torvalds 已提交
2133 2134
}

2135 2136 2137 2138
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2139 2140
{
	struct dm_table *map = md->map;
2141
	unsigned long flags;
L
Linus Torvalds 已提交
2142 2143

	if (!map)
2144
		return NULL;
L
Linus Torvalds 已提交
2145 2146

	dm_table_event_callback(map, NULL, NULL);
2147
	write_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
2148
	md->map = NULL;
2149
	write_unlock_irqrestore(&md->map_lock, flags);
2150 2151

	return map;
L
Linus Torvalds 已提交
2152 2153 2154 2155 2156
}

/*
 * Constructor for a new device.
 */
2157
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2158 2159 2160
{
	struct mapped_device *md;

2161
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2162 2163 2164
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2165 2166
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2167 2168 2169 2170
	*result = md;
	return 0;
}

2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194
/*
 * 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;
}

2195 2196 2197 2198 2199
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2200 2201 2202 2203 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
/*
 * 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;
}

2240
static struct mapped_device *dm_find_md(dev_t dev)
L
Linus Torvalds 已提交
2241 2242 2243 2244 2245 2246 2247
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2248
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2249 2250

	md = idr_find(&_minor_idr, minor);
J
Jeff Mahoney 已提交
2251
	if (md && (md == MINOR_ALLOCED ||
2252
		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
2253
		   dm_deleting_md(md) ||
A
Alasdair G Kergon 已提交
2254
		   test_bit(DMF_FREEING, &md->flags))) {
2255
		md = NULL;
J
Jeff Mahoney 已提交
2256 2257
		goto out;
	}
L
Linus Torvalds 已提交
2258

J
Jeff Mahoney 已提交
2259
out:
2260
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2261

2262 2263 2264
	return md;
}

2265 2266 2267 2268 2269 2270 2271 2272 2273
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
2274
EXPORT_SYMBOL_GPL(dm_get_md);
2275

A
Alasdair G Kergon 已提交
2276
void *dm_get_mdptr(struct mapped_device *md)
2277
{
A
Alasdair G Kergon 已提交
2278
	return md->interface_ptr;
L
Linus Torvalds 已提交
2279 2280 2281 2282 2283 2284 2285 2286 2287 2288
}

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);
2289
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2290 2291
}

2292 2293 2294 2295 2296 2297
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2298
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2299
{
M
Mike Anderson 已提交
2300
	struct dm_table *map;
L
Linus Torvalds 已提交
2301

2302
	might_sleep();
J
Jeff Mahoney 已提交
2303

2304 2305 2306 2307 2308 2309 2310 2311 2312
	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
2313
	}
2314 2315 2316 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

	/*
	 * 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
2347
}
E
Edward Goggin 已提交
2348
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2349

2350
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2351 2352
{
	int r = 0;
2353 2354 2355
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2356 2357

	while (1) {
2358
		set_current_state(interruptible);
2359

2360
		if (!md_in_flight(md))
2361 2362
			break;

2363 2364
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2365 2366 2367 2368 2369 2370 2371 2372
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2373 2374
	remove_wait_queue(&md->wait, &wait);

2375 2376 2377
	return r;
}

L
Linus Torvalds 已提交
2378 2379 2380
/*
 * Process the deferred bios
 */
2381
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2382
{
2383 2384
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2385
	struct bio *c;
L
Linus Torvalds 已提交
2386

2387
	down_read(&md->io_lock);
2388

2389
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2390 2391 2392 2393
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2394
		if (!c)
A
Alasdair G Kergon 已提交
2395
			break;
2396

2397
		up_read(&md->io_lock);
2398

K
Kiyoshi Ueda 已提交
2399 2400
		if (dm_request_based(md))
			generic_make_request(c);
2401 2402
		else
			__split_and_process_bio(md, c);
2403

2404
		down_read(&md->io_lock);
2405
	}
M
Milan Broz 已提交
2406

2407
	up_read(&md->io_lock);
L
Linus Torvalds 已提交
2408 2409
}

2410
static void dm_queue_flush(struct mapped_device *md)
2411
{
2412 2413
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	smp_mb__after_clear_bit();
2414
	queue_work(md->wq, &md->work);
2415 2416
}

L
Linus Torvalds 已提交
2417
/*
2418
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2419
 */
2420
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2421
{
2422
	struct dm_table *live_map, *map = ERR_PTR(-EINVAL);
2423
	struct queue_limits limits;
2424
	int r;
L
Linus Torvalds 已提交
2425

2426
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2427 2428

	/* device must be suspended */
2429
	if (!dm_suspended_md(md))
2430
		goto out;
L
Linus Torvalds 已提交
2431

2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
	/*
	 * 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);
	}

2445
	r = dm_calculate_queue_limits(table, &limits);
2446 2447
	if (r) {
		map = ERR_PTR(r);
2448
		goto out;
2449
	}
2450

2451
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2452

2453
out:
2454
	mutex_unlock(&md->suspend_lock);
2455
	return map;
L
Linus Torvalds 已提交
2456 2457 2458 2459 2460 2461
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2462
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2463
{
2464
	int r;
L
Linus Torvalds 已提交
2465 2466

	WARN_ON(md->frozen_sb);
2467

2468
	md->frozen_sb = freeze_bdev(md->bdev);
2469
	if (IS_ERR(md->frozen_sb)) {
2470
		r = PTR_ERR(md->frozen_sb);
2471 2472
		md->frozen_sb = NULL;
		return r;
2473 2474
	}

2475 2476
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2477 2478 2479
	return 0;
}

2480
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2481
{
2482 2483 2484
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2485
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2486
	md->frozen_sb = NULL;
2487
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2488 2489 2490 2491 2492 2493 2494 2495 2496
}

/*
 * 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.
 */
2497 2498 2499
/*
 * Suspend mechanism in request-based dm.
 *
K
Kiyoshi Ueda 已提交
2500 2501 2502
 * 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.
2503
 *
K
Kiyoshi Ueda 已提交
2504
 * To abort suspend, start the request_queue.
2505
 */
2506
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
L
Linus Torvalds 已提交
2507
{
2508
	struct dm_table *map = NULL;
2509
	int r = 0;
2510
	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
2511
	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
L
Linus Torvalds 已提交
2512

2513
	mutex_lock(&md->suspend_lock);
2514

2515
	if (dm_suspended_md(md)) {
M
Milan Broz 已提交
2516
		r = -EINVAL;
2517
		goto out_unlock;
M
Milan Broz 已提交
2518
	}
L
Linus Torvalds 已提交
2519

2520
	map = dm_get_live_table(md);
L
Linus Torvalds 已提交
2521

2522 2523 2524 2525 2526 2527 2528
	/*
	 * 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);

2529 2530 2531
	/* This does not get reverted if there's an error later. */
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2532
	/*
K
Kiyoshi Ueda 已提交
2533 2534 2535 2536
	 * 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
2537 2538 2539 2540
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
		if (r)
K
Kiyoshi Ueda 已提交
2541
			goto out;
2542
	}
L
Linus Torvalds 已提交
2543 2544

	/*
2545 2546 2547 2548 2549 2550 2551
	 * 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
2552 2553 2554
	 * __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
2555
	 */
2556
	down_write(&md->io_lock);
2557
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
2558
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
2559

2560
	/*
2561 2562
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2563
	 */
2564
	if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2565
		stop_queue(md->queue);
2566

2567 2568
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2569
	/*
2570 2571 2572
	 * 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
2573
	 */
2574
	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
L
Linus Torvalds 已提交
2575

2576
	down_write(&md->io_lock);
2577
	if (noflush)
2578
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2579
	up_write(&md->io_lock);
2580

L
Linus Torvalds 已提交
2581
	/* were we interrupted ? */
2582
	if (r < 0) {
2583
		dm_queue_flush(md);
M
Milan Broz 已提交
2584

2585
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2586
			start_queue(md->queue);
2587

2588
		unlock_fs(md);
2589
		goto out; /* pushback list is already flushed, so skip flush */
2590
	}
L
Linus Torvalds 已提交
2591

2592 2593 2594 2595 2596 2597
	/*
	 * 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.
	 */

2598
	set_bit(DMF_SUSPENDED, &md->flags);
2599

2600 2601
	dm_table_postsuspend_targets(map);

2602 2603
out:
	dm_table_put(map);
2604 2605

out_unlock:
2606
	mutex_unlock(&md->suspend_lock);
2607
	return r;
L
Linus Torvalds 已提交
2608 2609 2610 2611
}

int dm_resume(struct mapped_device *md)
{
2612 2613
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2614

2615
	mutex_lock(&md->suspend_lock);
2616
	if (!dm_suspended_md(md))
2617 2618
		goto out;

2619
	map = dm_get_live_table(md);
2620
	if (!map || !dm_table_get_size(map))
2621
		goto out;
L
Linus Torvalds 已提交
2622

2623 2624 2625
	r = dm_table_resume_targets(map);
	if (r)
		goto out;
2626

2627
	dm_queue_flush(md);
2628

2629 2630 2631 2632 2633 2634 2635 2636
	/*
	 * 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);

2637 2638 2639 2640
	unlock_fs(md);

	clear_bit(DMF_SUSPENDED, &md->flags);

2641 2642 2643
	r = 0;
out:
	dm_table_put(map);
2644
	mutex_unlock(&md->suspend_lock);
2645

2646
	return r;
L
Linus Torvalds 已提交
2647 2648 2649 2650 2651
}

/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2652
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2653
		       unsigned cookie)
2654
{
M
Milan Broz 已提交
2655 2656 2657 2658
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2659
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2660 2661 2662
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2663 2664
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2665
	}
2666 2667
}

M
Mike Anderson 已提交
2668 2669 2670 2671 2672
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2673 2674 2675 2676 2677 2678 2679 2680 2681 2682 2683
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
2684 2685 2686 2687 2688 2689 2690 2691 2692
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
2693 2694 2695 2696 2697 2698 2699 2700 2701
/*
 * 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
2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718
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;

2719
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2720
	    dm_deleting_md(md))
2721 2722
		return NULL;

M
Milan Broz 已提交
2723 2724 2725 2726
	dm_get(md);
	return md;
}

2727
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2728 2729 2730 2731
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2732 2733
int dm_suspended(struct dm_target *ti)
{
2734
	return dm_suspended_md(dm_table_get_md(ti->table));
2735 2736 2737
}
EXPORT_SYMBOL_GPL(dm_suspended);

2738 2739
int dm_noflush_suspending(struct dm_target *ti)
{
2740
	return __noflush_suspending(dm_table_get_md(ti->table));
2741 2742 2743
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

M
Mikulas Patocka 已提交
2744
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size)
K
Kiyoshi Ueda 已提交
2745 2746
{
	struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL);
2747
	unsigned int pool_size = (type == DM_TYPE_BIO_BASED) ? 16 : MIN_IOS;
K
Kiyoshi Ueda 已提交
2748 2749 2750 2751

	if (!pools)
		return NULL;

M
Mikulas Patocka 已提交
2752 2753
	per_bio_data_size = roundup(per_bio_data_size, __alignof__(struct dm_target_io));

K
Kiyoshi Ueda 已提交
2754 2755 2756 2757 2758 2759
	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;

2760 2761 2762 2763 2764 2765
	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
2766

2767
	pools->bs = (type == DM_TYPE_BIO_BASED) ?
2768
		bioset_create(pool_size,
M
Mikulas Patocka 已提交
2769
			      per_bio_data_size + offsetof(struct dm_target_io, clone)) :
2770 2771
		bioset_create(pool_size,
			      offsetof(struct dm_rq_clone_bio_info, clone));
K
Kiyoshi Ueda 已提交
2772 2773 2774
	if (!pools->bs)
		goto free_tio_pool_and_out;

2775 2776 2777
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
		goto free_bioset_and_out;

K
Kiyoshi Ueda 已提交
2778 2779
	return pools;

2780 2781 2782
free_bioset_and_out:
	bioset_free(pools->bs);

K
Kiyoshi Ueda 已提交
2783
free_tio_pool_and_out:
2784 2785
	if (pools->tio_pool)
		mempool_destroy(pools->tio_pool);
K
Kiyoshi Ueda 已提交
2786 2787 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

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);
}

2813
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2814 2815
	.open = dm_blk_open,
	.release = dm_blk_close,
2816
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2817
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833
	.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");