dm.c 61.6 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
#include <linux/moduleparam.h>
#include <linux/blkpg.h>
#include <linux/bio.h>
#include <linux/buffer_head.h>
18
#include <linux/smp_lock.h>
L
Linus Torvalds 已提交
19 20 21
#include <linux/mempool.h>
#include <linux/slab.h>
#include <linux/idr.h>
D
Darrick J. Wong 已提交
22
#include <linux/hdreg.h>
23
#include <linux/delay.h>
24 25

#include <trace/events/block.h>
L
Linus Torvalds 已提交
26

27 28
#define DM_MSG_PREFIX "core"

M
Milan Broz 已提交
29 30 31 32 33 34 35
/*
 * 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
36 37 38 39 40
static const char *_name = DM_NAME;

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

41
static DEFINE_SPINLOCK(_minor_lock);
L
Linus Torvalds 已提交
42
/*
K
Kiyoshi Ueda 已提交
43
 * For bio-based dm.
L
Linus Torvalds 已提交
44 45 46 47 48 49
 * 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
50
	struct bio *bio;
51
	unsigned long start_time;
52
	spinlock_t endio_lock;
L
Linus Torvalds 已提交
53 54 55
};

/*
K
Kiyoshi Ueda 已提交
56
 * For bio-based dm.
L
Linus Torvalds 已提交
57 58 59
 * 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
60
struct dm_target_io {
L
Linus Torvalds 已提交
61 62 63 64 65
	struct dm_io *io;
	struct dm_target *ti;
	union map_info info;
};

K
Kiyoshi Ueda 已提交
66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83
/*
 * 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;
};

/*
 * For request-based dm.
 * One of these is allocated per bio.
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
84
	struct dm_rq_target_io *tio;
K
Kiyoshi Ueda 已提交
85 86
};

L
Linus Torvalds 已提交
87 88
union map_info *dm_get_mapinfo(struct bio *bio)
{
A
Alasdair G Kergon 已提交
89
	if (bio && bio->bi_private)
A
Alasdair G Kergon 已提交
90
		return &((struct dm_target_io *)bio->bi_private)->info;
A
Alasdair G Kergon 已提交
91
	return NULL;
L
Linus Torvalds 已提交
92 93
}

94 95 96 97 98 99 100 101
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);

102 103
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
104 105 106
/*
 * Bits for the md->flags field.
 */
107
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
108
#define DMF_SUSPENDED 1
109
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
110
#define DMF_FREEING 3
111
#define DMF_DELETING 4
112
#define DMF_NOFLUSH_SUSPENDING 5
113
#define DMF_QUEUE_IO_TO_THREAD 6
L
Linus Torvalds 已提交
114

115 116 117
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
118
struct mapped_device {
119
	struct rw_semaphore io_lock;
120
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
121 122
	rwlock_t map_lock;
	atomic_t holders;
123
	atomic_t open_count;
L
Linus Torvalds 已提交
124 125 126

	unsigned long flags;

127
	struct request_queue *queue;
L
Linus Torvalds 已提交
128
	struct gendisk *disk;
M
Mike Anderson 已提交
129
	char name[16];
L
Linus Torvalds 已提交
130 131 132 133 134 135

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
136
	atomic_t pending[2];
L
Linus Torvalds 已提交
137
	wait_queue_head_t wait;
138
	struct work_struct work;
K
Kiyoshi Ueda 已提交
139
	struct bio_list deferred;
140
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
141

142 143 144 145 146
	/*
	 * An error from the barrier request currently being processed.
	 */
	int barrier_error;

147 148 149 150 151 152
	/*
	 * Protect barrier_error from concurrent endio processing
	 * in request-based dm.
	 */
	spinlock_t barrier_error_lock;

153 154 155 156
	/*
	 * Processing queue (flush/barriers)
	 */
	struct workqueue_struct *wq;
157 158 159 160
	struct work_struct barrier_work;

	/* A pointer to the currently processing pre/post flush request */
	struct request *flush_request;
161

L
Linus Torvalds 已提交
162 163 164 165 166 167 168 169 170 171 172
	/*
	 * 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
173 174
	struct bio_set *bs;

L
Linus Torvalds 已提交
175 176 177 178 179
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
180 181 182
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
183 184 185 186 187

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
188
	struct block_device *bdev;
D
Darrick J. Wong 已提交
189 190 191

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
192

193 194 195
	/* For saving the address of __make_request for request based dm */
	make_request_fn *saved_make_request_fn;

M
Milan Broz 已提交
196 197
	/* sysfs handle */
	struct kobject kobj;
198 199 200

	/* zero-length barrier that will be cloned and submitted to targets */
	struct bio barrier_bio;
L
Linus Torvalds 已提交
201 202
};

K
Kiyoshi Ueda 已提交
203 204 205 206 207 208 209 210 211
/*
 * 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
212
#define MIN_IOS 256
213 214
static struct kmem_cache *_io_cache;
static struct kmem_cache *_tio_cache;
K
Kiyoshi Ueda 已提交
215 216
static struct kmem_cache *_rq_tio_cache;
static struct kmem_cache *_rq_bio_info_cache;
L
Linus Torvalds 已提交
217 218 219

static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
220
	int r = -ENOMEM;
L
Linus Torvalds 已提交
221 222

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
223
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
224
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
225
		return r;
L
Linus Torvalds 已提交
226 227

	/* allocate a slab for the target ios */
A
Alasdair G Kergon 已提交
228
	_tio_cache = KMEM_CACHE(dm_target_io, 0);
K
Kiyoshi Ueda 已提交
229 230
	if (!_tio_cache)
		goto out_free_io_cache;
L
Linus Torvalds 已提交
231

K
Kiyoshi Ueda 已提交
232 233 234 235 236 237 238 239
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
		goto out_free_tio_cache;

	_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
240
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
241
	if (r)
K
Kiyoshi Ueda 已提交
242
		goto out_free_rq_bio_info_cache;
M
Mike Anderson 已提交
243

L
Linus Torvalds 已提交
244 245
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
246 247
	if (r < 0)
		goto out_uevent_exit;
L
Linus Torvalds 已提交
248 249 250 251 252

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
253 254 255

out_uevent_exit:
	dm_uevent_exit();
K
Kiyoshi Ueda 已提交
256 257 258 259
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
260 261 262 263 264 265
out_free_tio_cache:
	kmem_cache_destroy(_tio_cache);
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
266 267 268 269
}

static void local_exit(void)
{
K
Kiyoshi Ueda 已提交
270 271
	kmem_cache_destroy(_rq_bio_info_cache);
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
272 273
	kmem_cache_destroy(_tio_cache);
	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
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
	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]();
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
334 335 336 337 338
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
339
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
340 341 342
{
	struct mapped_device *md;

343
	lock_kernel();
J
Jeff Mahoney 已提交
344 345
	spin_lock(&_minor_lock);

A
Al Viro 已提交
346
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
347 348 349
	if (!md)
		goto out;

350
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
351
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
352 353 354 355
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
356
	dm_get(md);
357
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
358 359 360

out:
	spin_unlock(&_minor_lock);
361
	unlock_kernel();
J
Jeff Mahoney 已提交
362 363

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
364 365
}

A
Al Viro 已提交
366
static int dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
367
{
A
Al Viro 已提交
368
	struct mapped_device *md = disk->private_data;
369 370

	lock_kernel();
371
	atomic_dec(&md->open_count);
L
Linus Torvalds 已提交
372
	dm_put(md);
373 374
	unlock_kernel();

L
Linus Torvalds 已提交
375 376 377
	return 0;
}

378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401
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
402 403 404 405 406 407 408
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
409
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
410 411
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
412
	struct mapped_device *md = bdev->bd_disk->private_data;
413
	struct dm_table *map = dm_get_live_table(md);
414 415 416 417 418 419 420 421 422 423 424 425
	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);

426
	if (dm_suspended_md(md)) {
427 428 429 430 431
		r = -EAGAIN;
		goto out;
	}

	if (tgt->type->ioctl)
432
		r = tgt->type->ioctl(tgt, cmd, arg);
433 434 435 436 437 438 439

out:
	dm_table_put(map);

	return r;
}

A
Alasdair G Kergon 已提交
440
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
441 442 443 444
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
445
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
446 447 448 449
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
450
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
451 452 453 454
{
	mempool_free(tio, md->tio_pool);
}

K
Kiyoshi Ueda 已提交
455 456
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
457
{
K
Kiyoshi Ueda 已提交
458
	return mempool_alloc(md->tio_pool, gfp_mask);
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475
}

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

static struct dm_rq_clone_bio_info *alloc_bio_info(struct mapped_device *md)
{
	return mempool_alloc(md->io_pool, GFP_ATOMIC);
}

static void free_bio_info(struct dm_rq_clone_bio_info *info)
{
	mempool_free(info, info->tio->md->io_pool);
}

K
Kiyoshi Ueda 已提交
476 477 478 479 480 481
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

482 483 484
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
T
Tejun Heo 已提交
485
	int cpu;
486
	int rw = bio_data_dir(io->bio);
487 488 489

	io->start_time = jiffies;

T
Tejun Heo 已提交
490 491 492
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
493
	dm_disk(md)->part0.in_flight[rw] = atomic_inc_return(&md->pending[rw]);
494 495
}

496
static void end_io_acct(struct dm_io *io)
497 498 499 500
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
T
Tejun Heo 已提交
501
	int pending, cpu;
502 503
	int rw = bio_data_dir(bio);

T
Tejun Heo 已提交
504 505 506 507
	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();
508

509 510 511 512
	/*
	 * After this is decremented the bio must not be touched if it is
	 * a barrier.
	 */
513 514 515
	dm_disk(md)->part0.in_flight[rw] = pending =
		atomic_dec_return(&md->pending[rw]);
	pending += atomic_read(&md->pending[rw^0x1]);
516

517 518 519
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
520 521
}

L
Linus Torvalds 已提交
522 523 524
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
525
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
526
{
527
	down_write(&md->io_lock);
L
Linus Torvalds 已提交
528

529
	spin_lock_irq(&md->deferred_lock);
L
Linus Torvalds 已提交
530
	bio_list_add(&md->deferred, bio);
531
	spin_unlock_irq(&md->deferred_lock);
L
Linus Torvalds 已提交
532

M
Mikulas Patocka 已提交
533 534 535
	if (!test_and_set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags))
		queue_work(md->wq, &md->work);

536
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
537 538 539 540 541 542 543
}

/*
 * 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.
 */
544
struct dm_table *dm_get_live_table(struct mapped_device *md)
L
Linus Torvalds 已提交
545 546
{
	struct dm_table *t;
547
	unsigned long flags;
L
Linus Torvalds 已提交
548

549
	read_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
550 551 552
	t = md->map;
	if (t)
		dm_table_get(t);
553
	read_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
554 555 556 557

	return t;
}

D
Darrick J. Wong 已提交
558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
/*
 * 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
585 586 587 588 589 590 591 592 593
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

594 595 596 597 598
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
599 600 601 602
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
603
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
604
{
605
	unsigned long flags;
606 607 608
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
609 610

	/* Push-back supersedes any I/O errors */
611 612 613 614 615 616
	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
617 618

	if (atomic_dec_and_test(&io->io_count)) {
619 620 621 622
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
623
			spin_lock_irqsave(&md->deferred_lock, flags);
624
			if (__noflush_suspending(md)) {
625
				if (!(io->bio->bi_rw & REQ_HARDBARRIER))
626 627 628
					bio_list_add_head(&md->deferred,
							  io->bio);
			} else
629 630
				/* noflush suspend was interrupted. */
				io->error = -EIO;
631
			spin_unlock_irqrestore(&md->deferred_lock, flags);
632 633
		}

634 635
		io_error = io->error;
		bio = io->bio;
636

637
		if (bio->bi_rw & REQ_HARDBARRIER) {
638 639 640 641 642
			/*
			 * There can be just one barrier request so we use
			 * a per-device variable for error reporting.
			 * Note that you can't touch the bio after end_io_acct
			 */
643
			if (!md->barrier_error && io_error != -EOPNOTSUPP)
644
				md->barrier_error = io_error;
645
			end_io_acct(io);
646
			free_io(md, io);
647 648
		} else {
			end_io_acct(io);
649
			free_io(md, io);
650

651 652
			if (io_error != DM_ENDIO_REQUEUE) {
				trace_block_bio_complete(md->queue, bio);
653

654 655
				bio_endio(bio, io_error);
			}
656
		}
L
Linus Torvalds 已提交
657 658 659
	}
}

660
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
661 662
{
	int r = 0;
A
Alasdair G Kergon 已提交
663
	struct dm_target_io *tio = bio->bi_private;
664
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
665
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
666 667 668 669 670 671 672
	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);
673 674 675 676 677
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
678
			error = r;
679 680
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
681
			return;
682 683 684 685
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
686 687
	}

S
Stefan Bader 已提交
688 689 690 691 692 693
	/*
	 * Store md for cleanup instead of tio which is about to get freed.
	 */
	bio->bi_private = md->bs;

	free_tio(md, tio);
694 695
	bio_put(bio);
	dec_pending(io, error);
L
Linus Torvalds 已提交
696 697
}

698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747
/*
 * 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);
}

748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764
static void store_barrier_error(struct mapped_device *md, int error)
{
	unsigned long flags;

	spin_lock_irqsave(&md->barrier_error_lock, flags);
	/*
	 * Basically, the first error is taken, but:
	 *   -EOPNOTSUPP supersedes any I/O error.
	 *   Requeue request supersedes any I/O error but -EOPNOTSUPP.
	 */
	if (!md->barrier_error || error == -EOPNOTSUPP ||
	    (md->barrier_error != -EOPNOTSUPP &&
	     error == DM_ENDIO_REQUEUE))
		md->barrier_error = error;
	spin_unlock_irqrestore(&md->barrier_error_lock, flags);
}

765 766 767 768 769
/*
 * 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.
 */
770
static void rq_completed(struct mapped_device *md, int rw, int run_queue)
771
{
772
	atomic_dec(&md->pending[rw]);
773 774

	/* nudge anyone waiting on suspend queue */
775
	if (!md_in_flight(md))
776 777 778
		wake_up(&md->wait);

	if (run_queue)
779
		blk_run_queue(md->queue);
780 781 782 783 784 785 786

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

787 788 789 790 791 792 793 794
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
795 796 797 798 799 800 801
/*
 * 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);
802
	int run_queue = 1;
803
	bool is_barrier = clone->cmd_flags & REQ_HARDBARRIER;
K
Kiyoshi Ueda 已提交
804 805 806 807
	struct dm_rq_target_io *tio = clone->end_io_data;
	struct mapped_device *md = tio->md;
	struct request *rq = tio->orig;

808
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC && !is_barrier) {
K
Kiyoshi Ueda 已提交
809 810 811 812 813 814 815 816 817 818 819 820 821 822
		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);

823 824 825 826 827 828
	if (unlikely(is_barrier)) {
		if (unlikely(error))
			store_barrier_error(md, error);
		run_queue = 0;
	} else
		blk_end_request_all(rq, error);
K
Kiyoshi Ueda 已提交
829

830
	rq_completed(md, rw, run_queue);
K
Kiyoshi Ueda 已提交
831 832
}

833 834 835 836 837 838 839
static void dm_unprep_request(struct request *rq)
{
	struct request *clone = rq->special;

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

840
	free_rq_clone(clone);
841 842 843 844 845 846 847
}

/*
 * Requeue the original request of a clone.
 */
void dm_requeue_unmapped_request(struct request *clone)
{
848
	int rw = rq_data_dir(clone);
849 850 851 852 853 854
	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;

855
	if (unlikely(clone->cmd_flags & REQ_HARDBARRIER)) {
856 857 858 859 860 861 862 863 864
		/*
		 * Barrier clones share an original request.
		 * Leave it to dm_end_request(), which handles this special
		 * case.
		 */
		dm_end_request(clone, DM_ENDIO_REQUEUE);
		return;
	}

865 866 867 868 869 870 871 872
	dm_unprep_request(rq);

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

873
	rq_completed(md, rw, 0);
874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905
}
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);
}

906
static void dm_done(struct request *clone, int error, bool mapped)
907
{
908
	int r = error;
909 910 911
	struct dm_rq_target_io *tio = clone->end_io_data;
	dm_request_endio_fn rq_end_io = tio->ti->type->rq_end_io;

912 913
	if (mapped && rq_end_io)
		r = rq_end_io(tio->ti, clone, error, &tio->info);
914

915
	if (r <= 0)
916
		/* The target wants to complete the I/O */
917 918
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
919 920
		/* The target will handle the I/O */
		return;
921
	else if (r == DM_ENDIO_REQUEUE)
922 923 924
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
925
		DMWARN("unimplemented target endio return value: %d", r);
926 927 928 929
		BUG();
	}
}

930 931 932 933 934 935 936 937 938 939 940 941 942 943 944
/*
 * 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);
}

945 946 947 948 949 950 951 952 953
/*
 * 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;

954
	if (unlikely(clone->cmd_flags & REQ_HARDBARRIER)) {
955 956 957 958 959 960 961 962 963 964 965 966
		/*
		 * Barrier clones share an original request.  So can't use
		 * softirq_done with the original.
		 * Pass the clone to dm_done() directly in this special case.
		 * It is safe (even if clone->q->queue_lock is held here)
		 * because there is no I/O dispatching during the completion
		 * of barrier clone.
		 */
		dm_done(clone, error, true);
		return;
	}

967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982
	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;

983
	if (unlikely(clone->cmd_flags & REQ_HARDBARRIER)) {
984 985 986 987 988 989 990 991 992 993
		/*
		 * Barrier clones share an original request.
		 * Leave it to dm_end_request(), which handles this special
		 * case.
		 */
		BUG_ON(error > 0);
		dm_end_request(clone, error);
		return;
	}

994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
	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);
}

L
Linus Torvalds 已提交
1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043
static sector_t max_io_len(struct mapped_device *md,
			   sector_t sector, struct dm_target *ti)
{
	sector_t offset = sector - ti->begin;
	sector_t len = ti->len - offset;

	/*
	 * Does the target need to split even further ?
	 */
	if (ti->split_io) {
		sector_t boundary;
		boundary = ((offset + ti->split_io) & ~(ti->split_io - 1))
			   - offset;
		if (len > boundary)
			len = boundary;
	}

	return len;
}

static void __map_bio(struct dm_target *ti, struct bio *clone,
A
Alasdair G Kergon 已提交
1044
		      struct dm_target_io *tio)
L
Linus Torvalds 已提交
1045 1046
{
	int r;
1047
	sector_t sector;
S
Stefan Bader 已提交
1048
	struct mapped_device *md;
L
Linus Torvalds 已提交
1049 1050 1051 1052 1053 1054 1055 1056 1057 1058

	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);
1059
	sector = clone->bi_sector;
L
Linus Torvalds 已提交
1060
	r = ti->type->map(ti, clone, &tio->info);
1061
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1062
		/* the bio has been remapped so dispatch it */
1063

1064
		trace_block_remap(bdev_get_queue(clone->bi_bdev), clone,
1065
				    tio->io->bio->bi_bdev->bd_dev, sector);
1066

L
Linus Torvalds 已提交
1067
		generic_make_request(clone);
1068 1069
	} 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
1070 1071 1072 1073 1074 1075
		md = tio->io->md;
		dec_pending(tio->io, r);
		/*
		 * Store bio_set for cleanup.
		 */
		clone->bi_private = md->bs;
L
Linus Torvalds 已提交
1076
		bio_put(clone);
S
Stefan Bader 已提交
1077
		free_tio(md, tio);
1078 1079 1080
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093
	}
}

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;
};

P
Peter Osterlund 已提交
1094 1095
static void dm_bio_destructor(struct bio *bio)
{
S
Stefan Bader 已提交
1096 1097 1098
	struct bio_set *bs = bio->bi_private;

	bio_free(bio, bs);
P
Peter Osterlund 已提交
1099 1100
}

L
Linus Torvalds 已提交
1101 1102 1103 1104 1105
/*
 * Creates a little bio that is just does part of a bvec.
 */
static struct bio *split_bvec(struct bio *bio, sector_t sector,
			      unsigned short idx, unsigned int offset,
S
Stefan Bader 已提交
1106
			      unsigned int len, struct bio_set *bs)
L
Linus Torvalds 已提交
1107 1108 1109 1110
{
	struct bio *clone;
	struct bio_vec *bv = bio->bi_io_vec + idx;

S
Stefan Bader 已提交
1111
	clone = bio_alloc_bioset(GFP_NOIO, 1, bs);
P
Peter Osterlund 已提交
1112
	clone->bi_destructor = dm_bio_destructor;
L
Linus Torvalds 已提交
1113 1114 1115 1116
	*clone->bi_io_vec = *bv;

	clone->bi_sector = sector;
	clone->bi_bdev = bio->bi_bdev;
1117
	clone->bi_rw = bio->bi_rw & ~REQ_HARDBARRIER;
L
Linus Torvalds 已提交
1118 1119 1120 1121
	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
1122
	clone->bi_flags |= 1 << BIO_CLONED;
L
Linus Torvalds 已提交
1123

M
Martin K. Petersen 已提交
1124
	if (bio_integrity(bio)) {
1125
		bio_integrity_clone(clone, bio, GFP_NOIO, bs);
M
Martin K. Petersen 已提交
1126 1127 1128 1129
		bio_integrity_trim(clone,
				   bio_sector_offset(bio, idx, offset), len);
	}

L
Linus Torvalds 已提交
1130 1131 1132 1133 1134 1135 1136 1137
	return clone;
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
static struct bio *clone_bio(struct bio *bio, sector_t sector,
			     unsigned short idx, unsigned short bv_count,
S
Stefan Bader 已提交
1138
			     unsigned int len, struct bio_set *bs)
L
Linus Torvalds 已提交
1139 1140 1141
{
	struct bio *clone;

S
Stefan Bader 已提交
1142 1143
	clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs);
	__bio_clone(clone, bio);
1144
	clone->bi_rw &= ~REQ_HARDBARRIER;
S
Stefan Bader 已提交
1145
	clone->bi_destructor = dm_bio_destructor;
L
Linus Torvalds 已提交
1146 1147 1148 1149 1150 1151
	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
1152
	if (bio_integrity(bio)) {
1153
		bio_integrity_clone(clone, bio, GFP_NOIO, bs);
M
Martin K. Petersen 已提交
1154 1155 1156 1157 1158 1159

		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
1160 1161 1162
	return clone;
}

1163 1164
static struct dm_target_io *alloc_tio(struct clone_info *ci,
				      struct dm_target *ti)
1165
{
1166
	struct dm_target_io *tio = mempool_alloc(ci->md->tio_pool, GFP_NOIO);
1167 1168 1169 1170

	tio->io = ci->io;
	tio->ti = ti;
	memset(&tio->info, 0, sizeof(tio->info));
1171 1172 1173 1174 1175 1176 1177 1178 1179 1180

	return tio;
}

static void __flush_target(struct clone_info *ci, struct dm_target *ti,
			  unsigned flush_nr)
{
	struct dm_target_io *tio = alloc_tio(ci, ti);
	struct bio *clone;

1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
	tio->info.flush_request = flush_nr;

	clone = bio_alloc_bioset(GFP_NOIO, 0, ci->md->bs);
	__bio_clone(clone, ci->bio);
	clone->bi_destructor = dm_bio_destructor;

	__map_bio(ti, clone, tio);
}

static int __clone_and_map_empty_barrier(struct clone_info *ci)
{
	unsigned target_nr = 0, flush_nr;
	struct dm_target *ti;

	while ((ti = dm_table_get_target(ci->map, target_nr++)))
		for (flush_nr = 0; flush_nr < ti->num_flush_requests;
		     flush_nr++)
			__flush_target(ci, ti, flush_nr);

	ci->sector_count = 0;

	return 0;
}

1205
static int __clone_and_map(struct clone_info *ci)
L
Linus Torvalds 已提交
1206 1207
{
	struct bio *clone, *bio = ci->bio;
1208 1209
	struct dm_target *ti;
	sector_t len = 0, max;
A
Alasdair G Kergon 已提交
1210
	struct dm_target_io *tio;
L
Linus Torvalds 已提交
1211

1212 1213 1214
	if (unlikely(bio_empty_barrier(bio)))
		return __clone_and_map_empty_barrier(ci);

1215 1216 1217 1218 1219 1220
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

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

L
Linus Torvalds 已提交
1221 1222 1223
	/*
	 * Allocate a target io object.
	 */
1224
	tio = alloc_tio(ci, ti);
L
Linus Torvalds 已提交
1225 1226 1227 1228 1229 1230 1231

	if (ci->sector_count <= max) {
		/*
		 * Optimise for the simple case where we can do all of
		 * the remaining io with a single clone.
		 */
		clone = clone_bio(bio, ci->sector, ci->idx,
S
Stefan Bader 已提交
1232 1233
				  bio->bi_vcnt - ci->idx, ci->sector_count,
				  ci->md->bs);
L
Linus Torvalds 已提交
1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
		__map_bio(ti, clone, tio);
		ci->sector_count = 0;

	} 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;
		}

S
Stefan Bader 已提交
1256 1257
		clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len,
				  ci->md->bs);
L
Linus Torvalds 已提交
1258 1259 1260 1261 1262 1263 1264 1265
		__map_bio(ti, clone, tio);

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

	} else {
		/*
1266
		 * Handle a bvec that must be split between two or more targets.
L
Linus Torvalds 已提交
1267 1268
		 */
		struct bio_vec *bv = bio->bi_io_vec + ci->idx;
1269 1270
		sector_t remaining = to_sector(bv->bv_len);
		unsigned int offset = 0;
L
Linus Torvalds 已提交
1271

1272 1273 1274
		do {
			if (offset) {
				ti = dm_table_find_target(ci->map, ci->sector);
1275 1276 1277
				if (!dm_target_is_valid(ti))
					return -EIO;

1278
				max = max_io_len(ci->md, ci->sector, ti);
L
Linus Torvalds 已提交
1279

1280
				tio = alloc_tio(ci, ti);
1281 1282 1283 1284 1285
			}

			len = min(remaining, max);

			clone = split_bvec(bio, ci->sector, ci->idx,
S
Stefan Bader 已提交
1286 1287
					   bv->bv_offset + offset, len,
					   ci->md->bs);
1288 1289 1290 1291 1292 1293 1294

			__map_bio(ti, clone, tio);

			ci->sector += len;
			ci->sector_count -= len;
			offset += to_bytes(len);
		} while (remaining -= len);
L
Linus Torvalds 已提交
1295 1296 1297

		ci->idx++;
	}
1298 1299

	return 0;
L
Linus Torvalds 已提交
1300 1301 1302
}

/*
M
Mikulas Patocka 已提交
1303
 * Split the bio into several clones and submit it to targets.
L
Linus Torvalds 已提交
1304
 */
1305
static void __split_and_process_bio(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
1306 1307
{
	struct clone_info ci;
1308
	int error = 0;
L
Linus Torvalds 已提交
1309

1310
	ci.map = dm_get_live_table(md);
1311
	if (unlikely(!ci.map)) {
1312
		if (!(bio->bi_rw & REQ_HARDBARRIER))
1313 1314
			bio_io_error(bio);
		else
1315 1316
			if (!md->barrier_error)
				md->barrier_error = -EIO;
1317 1318
		return;
	}
1319

L
Linus Torvalds 已提交
1320 1321 1322 1323 1324 1325 1326
	ci.md = md;
	ci.bio = bio;
	ci.io = alloc_io(md);
	ci.io->error = 0;
	atomic_set(&ci.io->io_count, 1);
	ci.io->bio = bio;
	ci.io->md = md;
1327
	spin_lock_init(&ci.io->endio_lock);
L
Linus Torvalds 已提交
1328 1329
	ci.sector = bio->bi_sector;
	ci.sector_count = bio_sectors(bio);
1330 1331
	if (unlikely(bio_empty_barrier(bio)))
		ci.sector_count = 1;
L
Linus Torvalds 已提交
1332 1333
	ci.idx = bio->bi_idx;

1334
	start_io_acct(ci.io);
1335 1336
	while (ci.sector_count && !error)
		error = __clone_and_map(&ci);
L
Linus Torvalds 已提交
1337 1338

	/* drop the extra reference count */
1339
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1340 1341 1342 1343 1344 1345
	dm_table_put(ci.map);
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1346 1347 1348 1349 1350
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
1351
	struct dm_table *map = dm_get_live_table(md);
M
Milan Broz 已提交
1352 1353
	struct dm_target *ti;
	sector_t max_sectors;
1354
	int max_size = 0;
M
Milan Broz 已提交
1355 1356

	if (unlikely(!map))
1357
		goto out;
M
Milan Broz 已提交
1358 1359

	ti = dm_table_find_target(map, bvm->bi_sector);
1360 1361
	if (!dm_target_is_valid(ti))
		goto out_table;
M
Milan Broz 已提交
1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
	max_sectors = min(max_io_len(md, bvm->bi_sector, ti),
			  (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);
1379 1380 1381 1382 1383 1384 1385 1386 1387 1388
	/*
	 * 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
1389

1390
out_table:
1391 1392 1393
	dm_table_put(map);

out:
M
Milan Broz 已提交
1394 1395 1396 1397 1398 1399 1400 1401 1402
	/*
	 * 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
1403 1404 1405 1406
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1407
static int _dm_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1408
{
1409
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1410
	struct mapped_device *md = q->queuedata;
T
Tejun Heo 已提交
1411
	int cpu;
L
Linus Torvalds 已提交
1412

1413
	down_read(&md->io_lock);
L
Linus Torvalds 已提交
1414

T
Tejun Heo 已提交
1415 1416 1417 1418
	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();
1419

L
Linus Torvalds 已提交
1420
	/*
1421 1422
	 * If we're suspended or the thread is processing barriers
	 * we have to queue this io for later.
L
Linus Torvalds 已提交
1423
	 */
1424
	if (unlikely(test_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags)) ||
1425
	    unlikely(bio->bi_rw & REQ_HARDBARRIER)) {
1426
		up_read(&md->io_lock);
L
Linus Torvalds 已提交
1427

A
Alasdair G Kergon 已提交
1428 1429 1430 1431 1432
		if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) &&
		    bio_rw(bio) == READA) {
			bio_io_error(bio);
			return 0;
		}
L
Linus Torvalds 已提交
1433

M
Mikulas Patocka 已提交
1434
		queue_io(md, bio);
L
Linus Torvalds 已提交
1435

M
Mikulas Patocka 已提交
1436
		return 0;
L
Linus Torvalds 已提交
1437 1438
	}

1439
	__split_and_process_bio(md, bio);
1440
	up_read(&md->io_lock);
1441
	return 0;
L
Linus Torvalds 已提交
1442 1443
}

1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 1465
static int dm_make_request(struct request_queue *q, struct bio *bio)
{
	struct mapped_device *md = q->queuedata;

	return md->saved_make_request_fn(q, bio); /* call __make_request() */
}

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

static int dm_request(struct request_queue *q, struct bio *bio)
{
	struct mapped_device *md = q->queuedata;

	if (dm_request_based(md))
		return dm_make_request(q, bio);

	return _dm_request(q, bio);
}

1466 1467
static bool dm_rq_is_flush_request(struct request *rq)
{
1468
	if (rq->cmd_flags & REQ_FLUSH)
1469 1470 1471 1472 1473
		return true;
	else
		return false;
}

1474 1475 1476 1477 1478 1479 1480 1481 1482 1483 1484 1485 1486 1487 1488 1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507 1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518
void dm_dispatch_request(struct request *rq)
{
	int r;

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

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

static void dm_rq_bio_destructor(struct bio *bio)
{
	struct dm_rq_clone_bio_info *info = bio->bi_private;
	struct mapped_device *md = info->tio->md;

	free_bio_info(info);
	bio_free(bio, md->bs);
}

static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
				 void *data)
{
	struct dm_rq_target_io *tio = data;
	struct mapped_device *md = tio->md;
	struct dm_rq_clone_bio_info *info = alloc_bio_info(md);

	if (!info)
		return -ENOMEM;

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

	return 0;
}

static int setup_clone(struct request *clone, struct request *rq,
		       struct dm_rq_target_io *tio)
{
1519
	int r;
1520

1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535
	if (dm_rq_is_flush_request(rq)) {
		blk_rq_init(NULL, clone);
		clone->cmd_type = REQ_TYPE_FS;
		clone->cmd_flags |= (REQ_HARDBARRIER | WRITE);
	} else {
		r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
				      dm_rq_bio_constructor, tio);
		if (r)
			return r;

		clone->cmd = rq->cmd;
		clone->cmd_len = rq->cmd_len;
		clone->sense = rq->sense;
		clone->buffer = rq->buffer;
	}
1536 1537 1538 1539 1540 1541 1542

	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	return 0;
}

K
Kiyoshi Ueda 已提交
1543 1544 1545 1546 1547 1548 1549 1550 1551 1552 1553 1554 1555 1556 1557 1558 1559 1560 1561 1562 1563 1564 1565 1566 1567 1568
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;
}

1569 1570 1571 1572 1573 1574 1575 1576
/*
 * 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;

1577 1578 1579
	if (unlikely(dm_rq_is_flush_request(rq)))
		return BLKPREP_OK;

1580 1581 1582 1583 1584
	if (unlikely(rq->special)) {
		DMWARN("Already has something in rq->special.");
		return BLKPREP_KILL;
	}

K
Kiyoshi Ueda 已提交
1585 1586
	clone = clone_rq(rq, md, GFP_ATOMIC);
	if (!clone)
1587 1588 1589 1590 1591 1592 1593 1594
		return BLKPREP_DEFER;

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

	return BLKPREP_OK;
}

1595 1596 1597 1598 1599 1600 1601
/*
 * 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)
1602
{
1603
	int r, requeued = 0;
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622
	struct dm_rq_target_io *tio = clone->end_io_data;

	/*
	 * 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);

	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 */
1623 1624
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(tio->orig));
1625 1626 1627 1628 1629
		dm_dispatch_request(clone);
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
1630
		requeued = 1;
1631 1632 1633 1634 1635 1636 1637 1638 1639 1640 1641
		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;
	}
1642 1643

	return requeued;
1644 1645 1646 1647 1648 1649 1650 1651 1652
}

/*
 * 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;
1653
	struct dm_table *map = dm_get_live_table(md);
1654
	struct dm_target *ti;
1655
	struct request *rq, *clone;
1656 1657

	/*
1658 1659 1660 1661
	 * 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().
1662 1663 1664 1665 1666 1667
	 */
	while (!blk_queue_plugged(q) && !blk_queue_stopped(q)) {
		rq = blk_peek_request(q);
		if (!rq)
			goto plug_and_out;

1668 1669 1670 1671 1672 1673 1674 1675
		if (unlikely(dm_rq_is_flush_request(rq))) {
			BUG_ON(md->flush_request);
			md->flush_request = rq;
			blk_start_request(rq);
			queue_work(md->wq, &md->barrier_work);
			goto out;
		}

1676 1677 1678 1679 1680
		ti = dm_table_find_target(map, blk_rq_pos(rq));
		if (ti->type->busy && ti->type->busy(ti))
			goto plug_and_out;

		blk_start_request(rq);
1681 1682 1683
		clone = rq->special;
		atomic_inc(&md->pending[rq_data_dir(clone)]);

1684
		spin_unlock(q->queue_lock);
1685 1686 1687
		if (map_request(ti, clone, md))
			goto requeued;

1688 1689 1690 1691 1692
		spin_lock_irq(q->queue_lock);
	}

	goto out;

1693 1694 1695
requeued:
	spin_lock_irq(q->queue_lock);

1696 1697 1698 1699 1700 1701 1702 1703 1704 1705 1706 1707 1708 1709 1710 1711 1712 1713 1714 1715 1716
plug_and_out:
	if (!elv_queue_empty(q))
		/* Some requests still remain, retry later */
		blk_plug_device(q);

out:
	dm_table_put(map);

	return;
}

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;
1717
	struct dm_table *map = dm_get_live_table(md);
1718 1719 1720 1721 1722 1723 1724 1725 1726 1727 1728

	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;
}

1729
static void dm_unplug_all(struct request_queue *q)
L
Linus Torvalds 已提交
1730 1731
{
	struct mapped_device *md = q->queuedata;
1732
	struct dm_table *map = dm_get_live_table(md);
L
Linus Torvalds 已提交
1733 1734

	if (map) {
1735 1736 1737
		if (dm_request_based(md))
			generic_unplug_device(q);

L
Linus Torvalds 已提交
1738 1739 1740 1741 1742 1743 1744
		dm_table_unplug_all(map);
		dm_table_put(map);
	}
}

static int dm_any_congested(void *congested_data, int bdi_bits)
{
1745 1746 1747
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1748

1749
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
1750
		map = dm_get_live_table(md);
1751
		if (map) {
1752 1753 1754 1755 1756 1757 1758 1759 1760 1761
			/*
			 * 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);

1762 1763 1764 1765
			dm_table_put(map);
		}
	}

L
Linus Torvalds 已提交
1766 1767 1768 1769 1770 1771 1772 1773
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
static DEFINE_IDR(_minor_idr);

1774
static void free_minor(int minor)
L
Linus Torvalds 已提交
1775
{
1776
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1777
	idr_remove(&_minor_idr, minor);
1778
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1779 1780 1781 1782 1783
}

/*
 * See if the device with a specific minor # is free.
 */
1784
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1785 1786 1787 1788 1789 1790
{
	int r, m;

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

J
Jeff Mahoney 已提交
1791 1792 1793 1794
	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
	if (!r)
		return -ENOMEM;

1795
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1796 1797 1798 1799 1800 1801

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

1802
	r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m);
J
Jeff Mahoney 已提交
1803
	if (r)
L
Linus Torvalds 已提交
1804 1805 1806 1807 1808 1809 1810 1811 1812
		goto out;

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

out:
1813
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1814 1815 1816
	return r;
}

1817
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1818
{
1819
	int r, m;
L
Linus Torvalds 已提交
1820 1821

	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
J
Jeff Mahoney 已提交
1822 1823 1824
	if (!r)
		return -ENOMEM;

1825
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1826

1827
	r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m);
1828
	if (r)
L
Linus Torvalds 已提交
1829 1830 1831 1832 1833 1834 1835 1836 1837 1838 1839
		goto out;

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

	*minor = m;

out:
1840
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1841 1842 1843
	return r;
}

1844
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1845

1846
static void dm_wq_work(struct work_struct *work);
1847
static void dm_rq_barrier_work(struct work_struct *work);
1848

L
Linus Torvalds 已提交
1849 1850 1851
/*
 * Allocate and initialise a blank device with a given minor.
 */
1852
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1853 1854
{
	int r;
1855
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
1856
	void *old_md;
L
Linus Torvalds 已提交
1857 1858 1859 1860 1861 1862

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

1863
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1864
		goto bad_module_get;
1865

L
Linus Torvalds 已提交
1866
	/* get a minor number for the dev */
1867
	if (minor == DM_ANY_MINOR)
1868
		r = next_free_minor(&minor);
1869
	else
1870
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1871
	if (r < 0)
M
Milan Broz 已提交
1872
		goto bad_minor;
L
Linus Torvalds 已提交
1873

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

K
Kiyoshi Ueda 已提交
1886
	md->queue = blk_init_queue(dm_request_fn, NULL);
L
Linus Torvalds 已提交
1887
	if (!md->queue)
M
Milan Broz 已提交
1888
		goto bad_queue;
L
Linus Torvalds 已提交
1889

K
Kiyoshi Ueda 已提交
1890 1891 1892 1893 1894 1895 1896 1897 1898 1899 1900 1901
	/*
	 * 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,
	 * although we initialized the queue using blk_init_queue().
	 * 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->saved_make_request_fn = md->queue->make_request_fn;
L
Linus Torvalds 已提交
1902 1903 1904 1905
	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);
1906
	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
L
Linus Torvalds 已提交
1907
	md->queue->unplug_fn = dm_unplug_all;
M
Milan Broz 已提交
1908
	blk_queue_merge_bvec(md->queue, dm_merge_bvec);
K
Kiyoshi Ueda 已提交
1909 1910 1911
	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);
1912
	blk_queue_ordered(md->queue, QUEUE_ORDERED_DRAIN_FLUSH);
S
Stefan Bader 已提交
1913

L
Linus Torvalds 已提交
1914 1915
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
1916
		goto bad_disk;
L
Linus Torvalds 已提交
1917

1918 1919
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
1920
	init_waitqueue_head(&md->wait);
1921
	INIT_WORK(&md->work, dm_wq_work);
1922
	INIT_WORK(&md->barrier_work, dm_rq_barrier_work);
1923 1924
	init_waitqueue_head(&md->eventq);

L
Linus Torvalds 已提交
1925 1926 1927 1928 1929 1930 1931
	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
1932
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
1933

1934 1935 1936 1937
	md->wq = create_singlethread_workqueue("kdmflush");
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
1938 1939 1940 1941
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

1942
	/* Populate the mapping, nobody knows we exist yet */
1943
	spin_lock(&_minor_lock);
1944
	old_md = idr_replace(&_minor_idr, md, minor);
1945
	spin_unlock(&_minor_lock);
1946 1947 1948

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
1949 1950
	return md;

M
Mikulas Patocka 已提交
1951 1952
bad_bdev:
	destroy_workqueue(md->wq);
1953
bad_thread:
1954
	del_gendisk(md->disk);
1955
	put_disk(md->disk);
M
Milan Broz 已提交
1956
bad_disk:
1957
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
1958
bad_queue:
L
Linus Torvalds 已提交
1959
	free_minor(minor);
M
Milan Broz 已提交
1960
bad_minor:
1961
	module_put(THIS_MODULE);
M
Milan Broz 已提交
1962
bad_module_get:
L
Linus Torvalds 已提交
1963 1964 1965 1966
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
1967 1968
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
1969 1970
static void free_dev(struct mapped_device *md)
{
1971
	int minor = MINOR(disk_devt(md->disk));
1972

M
Mikulas Patocka 已提交
1973 1974
	unlock_fs(md);
	bdput(md->bdev);
1975
	destroy_workqueue(md->wq);
K
Kiyoshi Ueda 已提交
1976 1977 1978 1979 1980 1981
	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
1982
	blk_integrity_unregister(md->disk);
L
Linus Torvalds 已提交
1983
	del_gendisk(md->disk);
1984
	free_minor(minor);
J
Jeff Mahoney 已提交
1985 1986 1987 1988 1989

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

L
Linus Torvalds 已提交
1990
	put_disk(md->disk);
1991
	blk_cleanup_queue(md->queue);
1992
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
1993 1994 1995
	kfree(md);
}

K
Kiyoshi Ueda 已提交
1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
	struct dm_md_mempools *p;

	if (md->io_pool && md->tio_pool && md->bs)
		/* 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
2019 2020 2021 2022 2023
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2024 2025
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2026 2027
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2028 2029 2030 2031
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2032
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2033

L
Linus Torvalds 已提交
2034 2035 2036 2037
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2038
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2039
{
2040
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2041

2042 2043 2044
	mutex_lock(&md->bdev->bd_inode->i_mutex);
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
	mutex_unlock(&md->bdev->bd_inode->i_mutex);
L
Linus Torvalds 已提交
2045 2046
}

2047 2048 2049 2050 2051
/*
 * 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
2052
{
2053
	struct dm_table *old_map;
2054
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2055
	sector_t size;
2056
	unsigned long flags;
L
Linus Torvalds 已提交
2057 2058

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2059 2060 2061 2062 2063 2064 2065

	/*
	 * 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
2066
	__set_size(md, size);
2067

2068 2069
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
	/*
	 * 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);

2082
	write_lock_irqsave(&md->map_lock, flags);
2083
	old_map = md->map;
L
Linus Torvalds 已提交
2084
	md->map = t;
2085
	dm_table_set_restrictions(t, q, limits);
2086
	write_unlock_irqrestore(&md->map_lock, flags);
L
Linus Torvalds 已提交
2087

2088
	return old_map;
L
Linus Torvalds 已提交
2089 2090
}

2091 2092 2093 2094
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2095 2096
{
	struct dm_table *map = md->map;
2097
	unsigned long flags;
L
Linus Torvalds 已提交
2098 2099

	if (!map)
2100
		return NULL;
L
Linus Torvalds 已提交
2101 2102

	dm_table_event_callback(map, NULL, NULL);
2103
	write_lock_irqsave(&md->map_lock, flags);
L
Linus Torvalds 已提交
2104
	md->map = NULL;
2105
	write_unlock_irqrestore(&md->map_lock, flags);
2106 2107

	return map;
L
Linus Torvalds 已提交
2108 2109 2110 2111 2112
}

/*
 * Constructor for a new device.
 */
2113
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2114 2115 2116
{
	struct mapped_device *md;

2117
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2118 2119 2120
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2121 2122
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2123 2124 2125 2126
	*result = md;
	return 0;
}

2127
static struct mapped_device *dm_find_md(dev_t dev)
L
Linus Torvalds 已提交
2128 2129 2130 2131 2132 2133 2134
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2135
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2136 2137

	md = idr_find(&_minor_idr, minor);
J
Jeff Mahoney 已提交
2138
	if (md && (md == MINOR_ALLOCED ||
2139
		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
2140
		   dm_deleting_md(md) ||
A
Alasdair G Kergon 已提交
2141
		   test_bit(DMF_FREEING, &md->flags))) {
2142
		md = NULL;
J
Jeff Mahoney 已提交
2143 2144
		goto out;
	}
L
Linus Torvalds 已提交
2145

J
Jeff Mahoney 已提交
2146
out:
2147
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2148

2149 2150 2151
	return md;
}

2152 2153 2154 2155 2156 2157 2158 2159 2160 2161
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
2162
void *dm_get_mdptr(struct mapped_device *md)
2163
{
A
Alasdair G Kergon 已提交
2164
	return md->interface_ptr;
L
Linus Torvalds 已提交
2165 2166 2167 2168 2169 2170 2171 2172 2173 2174
}

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);
2175
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2176 2177
}

2178 2179 2180 2181 2182 2183
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2184
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2185
{
M
Mike Anderson 已提交
2186
	struct dm_table *map;
L
Linus Torvalds 已提交
2187

2188
	might_sleep();
J
Jeff Mahoney 已提交
2189

2190 2191 2192 2193 2194 2195 2196 2197 2198
	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
2199
	}
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

	/*
	 * 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
2233
}
E
Edward Goggin 已提交
2234
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2235

2236
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2237 2238
{
	int r = 0;
2239 2240 2241 2242 2243
	DECLARE_WAITQUEUE(wait, current);

	dm_unplug_all(md->queue);

	add_wait_queue(&md->wait, &wait);
2244 2245

	while (1) {
2246
		set_current_state(interruptible);
2247 2248

		smp_mb();
2249
		if (!md_in_flight(md))
2250 2251
			break;

2252 2253
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2254 2255 2256 2257 2258 2259 2260 2261
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2262 2263
	remove_wait_queue(&md->wait, &wait);

2264 2265 2266
	return r;
}

M
Mikulas Patocka 已提交
2267
static void dm_flush(struct mapped_device *md)
2268 2269
{
	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
2270 2271 2272 2273 2274 2275 2276

	bio_init(&md->barrier_bio);
	md->barrier_bio.bi_bdev = md->bdev;
	md->barrier_bio.bi_rw = WRITE_BARRIER;
	__split_and_process_bio(md, &md->barrier_bio);

	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
2277 2278 2279 2280
}

static void process_barrier(struct mapped_device *md, struct bio *bio)
{
2281 2282
	md->barrier_error = 0;

M
Mikulas Patocka 已提交
2283
	dm_flush(md);
2284

2285 2286 2287
	if (!bio_empty_barrier(bio)) {
		__split_and_process_bio(md, bio);
		dm_flush(md);
2288 2289 2290
	}

	if (md->barrier_error != DM_ENDIO_REQUEUE)
M
Mikulas Patocka 已提交
2291
		bio_endio(bio, md->barrier_error);
2292 2293 2294 2295 2296
	else {
		spin_lock_irq(&md->deferred_lock);
		bio_list_add_head(&md->deferred, bio);
		spin_unlock_irq(&md->deferred_lock);
	}
2297 2298
}

L
Linus Torvalds 已提交
2299 2300 2301
/*
 * Process the deferred bios
 */
2302
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2303
{
2304 2305
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2306
	struct bio *c;
L
Linus Torvalds 已提交
2307

2308 2309
	down_write(&md->io_lock);

2310
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2311 2312 2313 2314 2315
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

		if (!c) {
2316
			clear_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags);
A
Alasdair G Kergon 已提交
2317 2318
			break;
		}
2319

2320 2321
		up_write(&md->io_lock);

K
Kiyoshi Ueda 已提交
2322 2323 2324
		if (dm_request_based(md))
			generic_make_request(c);
		else {
2325
			if (c->bi_rw & REQ_HARDBARRIER)
K
Kiyoshi Ueda 已提交
2326 2327 2328 2329
				process_barrier(md, c);
			else
				__split_and_process_bio(md, c);
		}
2330 2331

		down_write(&md->io_lock);
2332
	}
M
Milan Broz 已提交
2333

2334
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
2335 2336
}

2337
static void dm_queue_flush(struct mapped_device *md)
2338
{
2339 2340
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	smp_mb__after_clear_bit();
2341
	queue_work(md->wq, &md->work);
2342 2343
}

2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354
static void dm_rq_set_flush_nr(struct request *clone, unsigned flush_nr)
{
	struct dm_rq_target_io *tio = clone->end_io_data;

	tio->info.flush_request = flush_nr;
}

/* Issue barrier requests to targets and wait for their completion. */
static int dm_rq_barrier(struct mapped_device *md)
{
	int i, j;
2355
	struct dm_table *map = dm_get_live_table(md);
2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375 2376 2377 2378 2379 2380 2381 2382 2383 2384 2385 2386 2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408 2409 2410
	unsigned num_targets = dm_table_get_num_targets(map);
	struct dm_target *ti;
	struct request *clone;

	md->barrier_error = 0;

	for (i = 0; i < num_targets; i++) {
		ti = dm_table_get_target(map, i);
		for (j = 0; j < ti->num_flush_requests; j++) {
			clone = clone_rq(md->flush_request, md, GFP_NOIO);
			dm_rq_set_flush_nr(clone, j);
			atomic_inc(&md->pending[rq_data_dir(clone)]);
			map_request(ti, clone, md);
		}
	}

	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
	dm_table_put(map);

	return md->barrier_error;
}

static void dm_rq_barrier_work(struct work_struct *work)
{
	int error;
	struct mapped_device *md = container_of(work, struct mapped_device,
						barrier_work);
	struct request_queue *q = md->queue;
	struct request *rq;
	unsigned long flags;

	/*
	 * Hold the md reference here and leave it at the last part so that
	 * the md can't be deleted by device opener when the barrier request
	 * completes.
	 */
	dm_get(md);

	error = dm_rq_barrier(md);

	rq = md->flush_request;
	md->flush_request = NULL;

	if (error == DM_ENDIO_REQUEUE) {
		spin_lock_irqsave(q->queue_lock, flags);
		blk_requeue_request(q, rq);
		spin_unlock_irqrestore(q->queue_lock, flags);
	} else
		blk_end_request_all(rq, error);

	blk_run_queue(q);

	dm_put(md);
}

L
Linus Torvalds 已提交
2411
/*
2412
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2413
 */
2414
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2415
{
2416
	struct dm_table *map = ERR_PTR(-EINVAL);
2417
	struct queue_limits limits;
2418
	int r;
L
Linus Torvalds 已提交
2419

2420
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2421 2422

	/* device must be suspended */
2423
	if (!dm_suspended_md(md))
2424
		goto out;
L
Linus Torvalds 已提交
2425

2426
	r = dm_calculate_queue_limits(table, &limits);
2427 2428
	if (r) {
		map = ERR_PTR(r);
2429
		goto out;
2430
	}
2431

K
Kiyoshi Ueda 已提交
2432 2433 2434 2435 2436 2437 2438
	/* cannot change the device type, once a table is bound */
	if (md->map &&
	    (dm_table_get_type(md->map) != dm_table_get_type(table))) {
		DMWARN("can't change the device type after a table is bound");
		goto out;
	}

2439
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2440

2441
out:
2442
	mutex_unlock(&md->suspend_lock);
2443
	return map;
L
Linus Torvalds 已提交
2444 2445 2446 2447 2448 2449
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2450
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2451
{
2452
	int r;
L
Linus Torvalds 已提交
2453 2454

	WARN_ON(md->frozen_sb);
2455

2456
	md->frozen_sb = freeze_bdev(md->bdev);
2457
	if (IS_ERR(md->frozen_sb)) {
2458
		r = PTR_ERR(md->frozen_sb);
2459 2460
		md->frozen_sb = NULL;
		return r;
2461 2462
	}

2463 2464
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2465 2466 2467
	return 0;
}

2468
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2469
{
2470 2471 2472
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2473
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2474
	md->frozen_sb = NULL;
2475
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2476 2477 2478 2479 2480 2481 2482 2483 2484
}

/*
 * 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.
 */
2485 2486 2487
/*
 * Suspend mechanism in request-based dm.
 *
K
Kiyoshi Ueda 已提交
2488 2489 2490
 * 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.
2491
 *
K
Kiyoshi Ueda 已提交
2492
 * To abort suspend, start the request_queue.
2493
 */
2494
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
L
Linus Torvalds 已提交
2495
{
2496
	struct dm_table *map = NULL;
2497
	int r = 0;
2498
	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
2499
	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
L
Linus Torvalds 已提交
2500

2501
	mutex_lock(&md->suspend_lock);
2502

2503
	if (dm_suspended_md(md)) {
M
Milan Broz 已提交
2504
		r = -EINVAL;
2505
		goto out_unlock;
M
Milan Broz 已提交
2506
	}
L
Linus Torvalds 已提交
2507

2508
	map = dm_get_live_table(md);
L
Linus Torvalds 已提交
2509

2510 2511 2512 2513 2514 2515 2516
	/*
	 * 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);

2517 2518 2519
	/* This does not get reverted if there's an error later. */
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2520
	/*
K
Kiyoshi Ueda 已提交
2521 2522 2523 2524
	 * 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
2525 2526 2527 2528
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
		if (r)
K
Kiyoshi Ueda 已提交
2529
			goto out;
2530
	}
L
Linus Torvalds 已提交
2531 2532

	/*
2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546
	 * 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
	 * __split_and_process_bio from dm_request, we set
	 * DMF_QUEUE_IO_TO_THREAD.
	 *
	 * To quiesce the thread (dm_wq_work), we set DMF_BLOCK_IO_FOR_SUSPEND
	 * and call flush_workqueue(md->wq). flush_workqueue will wait until
	 * dm_wq_work exits and DMF_BLOCK_IO_FOR_SUSPEND will prevent any
	 * further calls to __split_and_process_bio from dm_wq_work.
L
Linus Torvalds 已提交
2547
	 */
2548
	down_write(&md->io_lock);
2549 2550
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	set_bit(DMF_QUEUE_IO_TO_THREAD, &md->flags);
2551
	up_write(&md->io_lock);
L
Linus Torvalds 已提交
2552

2553 2554 2555 2556 2557
	/*
	 * Request-based dm uses md->wq for barrier (dm_rq_barrier_work) which
	 * can be kicked until md->queue is stopped.  So stop md->queue before
	 * flushing md->wq.
	 */
2558
	if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2559
		stop_queue(md->queue);
2560

2561 2562
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2563
	/*
2564 2565 2566
	 * 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
2567
	 */
2568
	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
L
Linus Torvalds 已提交
2569

2570
	down_write(&md->io_lock);
2571
	if (noflush)
2572
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
2573
	up_write(&md->io_lock);
2574

L
Linus Torvalds 已提交
2575
	/* were we interrupted ? */
2576
	if (r < 0) {
2577
		dm_queue_flush(md);
M
Milan Broz 已提交
2578

2579
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2580
			start_queue(md->queue);
2581

2582
		unlock_fs(md);
2583
		goto out; /* pushback list is already flushed, so skip flush */
2584
	}
L
Linus Torvalds 已提交
2585

2586 2587 2588 2589 2590 2591
	/*
	 * 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.
	 */

2592
	set_bit(DMF_SUSPENDED, &md->flags);
2593

2594 2595
	dm_table_postsuspend_targets(map);

2596 2597
out:
	dm_table_put(map);
2598 2599

out_unlock:
2600
	mutex_unlock(&md->suspend_lock);
2601
	return r;
L
Linus Torvalds 已提交
2602 2603 2604 2605
}

int dm_resume(struct mapped_device *md)
{
2606 2607
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2608

2609
	mutex_lock(&md->suspend_lock);
2610
	if (!dm_suspended_md(md))
2611 2612
		goto out;

2613
	map = dm_get_live_table(md);
2614
	if (!map || !dm_table_get_size(map))
2615
		goto out;
L
Linus Torvalds 已提交
2616

2617 2618 2619
	r = dm_table_resume_targets(map);
	if (r)
		goto out;
2620

2621
	dm_queue_flush(md);
2622

2623 2624 2625 2626 2627 2628 2629 2630
	/*
	 * 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);

2631 2632 2633 2634
	unlock_fs(md);

	clear_bit(DMF_SUSPENDED, &md->flags);

L
Linus Torvalds 已提交
2635
	dm_table_unplug_all(map);
2636 2637 2638
	r = 0;
out:
	dm_table_put(map);
2639
	mutex_unlock(&md->suspend_lock);
2640

2641
	return r;
L
Linus Torvalds 已提交
2642 2643 2644 2645 2646
}

/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2647
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2648
		       unsigned cookie)
2649
{
M
Milan Broz 已提交
2650 2651 2652 2653
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2654
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2655 2656 2657
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2658 2659
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2660
	}
2661 2662
}

M
Mike Anderson 已提交
2663 2664 2665 2666 2667
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2668 2669 2670 2671 2672 2673 2674 2675 2676 2677 2678
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
2679 2680 2681 2682 2683 2684 2685 2686 2687
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
2688 2689 2690 2691 2692 2693 2694 2695 2696
/*
 * 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
2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713
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;

2714
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2715
	    dm_deleting_md(md))
2716 2717
		return NULL;

M
Milan Broz 已提交
2718 2719 2720 2721
	dm_get(md);
	return md;
}

2722
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2723 2724 2725 2726
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2727 2728
int dm_suspended(struct dm_target *ti)
{
2729
	return dm_suspended_md(dm_table_get_md(ti->table));
2730 2731 2732
}
EXPORT_SYMBOL_GPL(dm_suspended);

2733 2734
int dm_noflush_suspending(struct dm_target *ti)
{
2735
	return __noflush_suspending(dm_table_get_md(ti->table));
2736 2737 2738
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

K
Kiyoshi Ueda 已提交
2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783 2784 2785 2786 2787 2788 2789 2790 2791 2792 2793
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type)
{
	struct dm_md_mempools *pools = kmalloc(sizeof(*pools), GFP_KERNEL);

	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;

	pools->tio_pool = (type == DM_TYPE_BIO_BASED) ?
			  mempool_create_slab_pool(MIN_IOS, _tio_cache) :
			  mempool_create_slab_pool(MIN_IOS, _rq_tio_cache);
	if (!pools->tio_pool)
		goto free_io_pool_and_out;

	pools->bs = (type == DM_TYPE_BIO_BASED) ?
		    bioset_create(16, 0) : bioset_create(MIN_IOS, 0);
	if (!pools->bs)
		goto free_tio_pool_and_out;

	return pools;

free_tio_pool_and_out:
	mempool_destroy(pools->tio_pool);

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);
}

2794
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2795 2796
	.open = dm_blk_open,
	.release = dm_blk_close,
2797
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2798
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809 2810 2811 2812 2813 2814
	.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");