dm.c 66.7 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;
M
Mikulas Patocka 已提交
63
	struct dm_stats_aux stats_aux;
L
Linus Torvalds 已提交
64 65
};

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

/*
79 80 81 82 83 84
 * 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
85 86 87
 */
struct dm_rq_clone_bio_info {
	struct bio *orig;
88
	struct dm_rq_target_io *tio;
89
	struct bio clone;
K
Kiyoshi Ueda 已提交
90 91
};

L
Linus Torvalds 已提交
92 93
union map_info *dm_get_mapinfo(struct bio *bio)
{
A
Alasdair G Kergon 已提交
94
	if (bio && bio->bi_private)
A
Alasdair G Kergon 已提交
95
		return &((struct dm_target_io *)bio->bi_private)->info;
A
Alasdair G Kergon 已提交
96
	return NULL;
L
Linus Torvalds 已提交
97 98
}

99 100 101 102 103 104 105 106
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);

107 108
#define MINOR_ALLOCED ((void *)-1)

L
Linus Torvalds 已提交
109 110 111
/*
 * Bits for the md->flags field.
 */
112
#define DMF_BLOCK_IO_FOR_SUSPEND 0
L
Linus Torvalds 已提交
113
#define DMF_SUSPENDED 1
114
#define DMF_FROZEN 2
J
Jeff Mahoney 已提交
115
#define DMF_FREEING 3
116
#define DMF_DELETING 4
117
#define DMF_NOFLUSH_SUSPENDING 5
118
#define DMF_MERGE_IS_OPTIONAL 6
L
Linus Torvalds 已提交
119

M
Mikulas Patocka 已提交
120 121 122 123 124 125 126 127
/*
 * A dummy definition to make RCU happy.
 * struct dm_table should never be dereferenced in this file.
 */
struct dm_table {
	int undefined__;
};

128 129 130
/*
 * Work processed by per-device workqueue.
 */
L
Linus Torvalds 已提交
131
struct mapped_device {
M
Mikulas Patocka 已提交
132
	struct srcu_struct io_barrier;
133
	struct mutex suspend_lock;
L
Linus Torvalds 已提交
134
	atomic_t holders;
135
	atomic_t open_count;
L
Linus Torvalds 已提交
136

M
Mikulas Patocka 已提交
137 138 139 140 141 142 143
	/*
	 * The current mapping.
	 * Use dm_get_live_table{_fast} or take suspend_lock for
	 * dereference.
	 */
	struct dm_table *map;

L
Linus Torvalds 已提交
144 145
	unsigned long flags;

146
	struct request_queue *queue;
147
	unsigned type;
148
	/* Protect queue and type against concurrent access. */
149 150
	struct mutex type_lock;

151 152
	struct target_type *immutable_target_type;

L
Linus Torvalds 已提交
153
	struct gendisk *disk;
M
Mike Anderson 已提交
154
	char name[16];
L
Linus Torvalds 已提交
155 156 157 158 159 160

	void *interface_ptr;

	/*
	 * A list of ios that arrived while we were suspended.
	 */
161
	atomic_t pending[2];
L
Linus Torvalds 已提交
162
	wait_queue_head_t wait;
163
	struct work_struct work;
K
Kiyoshi Ueda 已提交
164
	struct bio_list deferred;
165
	spinlock_t deferred_lock;
L
Linus Torvalds 已提交
166

167
	/*
168
	 * Processing queue (flush)
169 170 171
	 */
	struct workqueue_struct *wq;

L
Linus Torvalds 已提交
172 173 174 175 176
	/*
	 * io objects are allocated from here.
	 */
	mempool_t *io_pool;

S
Stefan Bader 已提交
177 178
	struct bio_set *bs;

L
Linus Torvalds 已提交
179 180 181 182 183
	/*
	 * Event handling.
	 */
	atomic_t event_nr;
	wait_queue_head_t eventq;
M
Mike Anderson 已提交
184 185 186
	atomic_t uevent_seq;
	struct list_head uevent_list;
	spinlock_t uevent_lock; /* Protect access to uevent_list */
L
Linus Torvalds 已提交
187 188 189 190 191

	/*
	 * freeze/thaw support require holding onto a super block
	 */
	struct super_block *frozen_sb;
192
	struct block_device *bdev;
D
Darrick J. Wong 已提交
193 194 195

	/* forced geometry settings */
	struct hd_geometry geometry;
M
Milan Broz 已提交
196 197 198

	/* sysfs handle */
	struct kobject kobj;
199

200 201
	/* zero-length flush that will be cloned and submitted to targets */
	struct bio flush_bio;
M
Mikulas Patocka 已提交
202 203

	struct dm_stats stats;
L
Linus Torvalds 已提交
204 205
};

K
Kiyoshi Ueda 已提交
206 207 208 209 210 211 212 213
/*
 * For mempools pre-allocation at the table loading time.
 */
struct dm_md_mempools {
	mempool_t *io_pool;
	struct bio_set *bs;
};

214 215
#define RESERVED_BIO_BASED_IOS		16
#define RESERVED_REQUEST_BASED_IOS	256
216
#define RESERVED_MAX_IOS		1024
217
static struct kmem_cache *_io_cache;
K
Kiyoshi Ueda 已提交
218
static struct kmem_cache *_rq_tio_cache;
219

220 221 222 223 224
/*
 * Bio-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;

225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248
/*
 * Request-based DM's mempools' reserved IOs set by the user.
 */
static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;

static unsigned __dm_get_reserved_ios(unsigned *reserved_ios,
				      unsigned def, unsigned max)
{
	unsigned ios = ACCESS_ONCE(*reserved_ios);
	unsigned modified_ios = 0;

	if (!ios)
		modified_ios = def;
	else if (ios > max)
		modified_ios = max;

	if (modified_ios) {
		(void)cmpxchg(reserved_ios, ios, modified_ios);
		ios = modified_ios;
	}

	return ios;
}

249 250 251 252 253 254 255
unsigned dm_get_reserved_bio_based_ios(void)
{
	return __dm_get_reserved_ios(&reserved_bio_based_ios,
				     RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);

256 257 258 259 260 261 262
unsigned dm_get_reserved_rq_based_ios(void)
{
	return __dm_get_reserved_ios(&reserved_rq_based_ios,
				     RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS);
}
EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);

L
Linus Torvalds 已提交
263 264
static int __init local_init(void)
{
K
Kiyoshi Ueda 已提交
265
	int r = -ENOMEM;
L
Linus Torvalds 已提交
266 267

	/* allocate a slab for the dm_ios */
A
Alasdair G Kergon 已提交
268
	_io_cache = KMEM_CACHE(dm_io, 0);
L
Linus Torvalds 已提交
269
	if (!_io_cache)
K
Kiyoshi Ueda 已提交
270
		return r;
L
Linus Torvalds 已提交
271

K
Kiyoshi Ueda 已提交
272 273
	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
	if (!_rq_tio_cache)
274
		goto out_free_io_cache;
K
Kiyoshi Ueda 已提交
275

M
Mike Anderson 已提交
276
	r = dm_uevent_init();
K
Kiyoshi Ueda 已提交
277
	if (r)
278
		goto out_free_rq_tio_cache;
M
Mike Anderson 已提交
279

L
Linus Torvalds 已提交
280 281
	_major = major;
	r = register_blkdev(_major, _name);
K
Kiyoshi Ueda 已提交
282 283
	if (r < 0)
		goto out_uevent_exit;
L
Linus Torvalds 已提交
284 285 286 287 288

	if (!_major)
		_major = r;

	return 0;
K
Kiyoshi Ueda 已提交
289 290 291

out_uevent_exit:
	dm_uevent_exit();
K
Kiyoshi Ueda 已提交
292 293
out_free_rq_tio_cache:
	kmem_cache_destroy(_rq_tio_cache);
K
Kiyoshi Ueda 已提交
294 295 296 297
out_free_io_cache:
	kmem_cache_destroy(_io_cache);

	return r;
L
Linus Torvalds 已提交
298 299 300 301
}

static void local_exit(void)
{
K
Kiyoshi Ueda 已提交
302
	kmem_cache_destroy(_rq_tio_cache);
L
Linus Torvalds 已提交
303
	kmem_cache_destroy(_io_cache);
304
	unregister_blkdev(_major, _name);
M
Mike Anderson 已提交
305
	dm_uevent_exit();
L
Linus Torvalds 已提交
306 307 308 309 310 311

	_major = 0;

	DMINFO("cleaned up");
}

312
static int (*_inits[])(void) __initdata = {
L
Linus Torvalds 已提交
313 314 315 316
	local_init,
	dm_target_init,
	dm_linear_init,
	dm_stripe_init,
M
Mikulas Patocka 已提交
317
	dm_io_init,
318
	dm_kcopyd_init,
L
Linus Torvalds 已提交
319
	dm_interface_init,
M
Mikulas Patocka 已提交
320
	dm_statistics_init,
L
Linus Torvalds 已提交
321 322
};

323
static void (*_exits[])(void) = {
L
Linus Torvalds 已提交
324 325 326 327
	local_exit,
	dm_target_exit,
	dm_linear_exit,
	dm_stripe_exit,
M
Mikulas Patocka 已提交
328
	dm_io_exit,
329
	dm_kcopyd_exit,
L
Linus Torvalds 已提交
330
	dm_interface_exit,
M
Mikulas Patocka 已提交
331
	dm_statistics_exit,
L
Linus Torvalds 已提交
332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360
};

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]();
361 362 363 364 365

	/*
	 * Should be empty by this point.
	 */
	idr_destroy(&_minor_idr);
L
Linus Torvalds 已提交
366 367 368 369 370
}

/*
 * Block device functions
 */
M
Mike Anderson 已提交
371 372 373 374 375
int dm_deleting_md(struct mapped_device *md)
{
	return test_bit(DMF_DELETING, &md->flags);
}

A
Al Viro 已提交
376
static int dm_blk_open(struct block_device *bdev, fmode_t mode)
L
Linus Torvalds 已提交
377 378 379
{
	struct mapped_device *md;

J
Jeff Mahoney 已提交
380 381
	spin_lock(&_minor_lock);

A
Al Viro 已提交
382
	md = bdev->bd_disk->private_data;
J
Jeff Mahoney 已提交
383 384 385
	if (!md)
		goto out;

386
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
387
	    dm_deleting_md(md)) {
J
Jeff Mahoney 已提交
388 389 390 391
		md = NULL;
		goto out;
	}

L
Linus Torvalds 已提交
392
	dm_get(md);
393
	atomic_inc(&md->open_count);
J
Jeff Mahoney 已提交
394 395 396 397 398

out:
	spin_unlock(&_minor_lock);

	return md ? 0 : -ENXIO;
L
Linus Torvalds 已提交
399 400
}

401
static void dm_blk_close(struct gendisk *disk, fmode_t mode)
L
Linus Torvalds 已提交
402
{
A
Al Viro 已提交
403
	struct mapped_device *md = disk->private_data;
404

405 406
	spin_lock(&_minor_lock);

407
	atomic_dec(&md->open_count);
L
Linus Torvalds 已提交
408
	dm_put(md);
409 410

	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
411 412
}

413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436
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;
}

M
Mikulas Patocka 已提交
437 438 439 440 441 442 443 444 445 446
sector_t dm_get_size(struct mapped_device *md)
{
	return get_capacity(md->disk);
}

struct dm_stats *dm_get_stats(struct mapped_device *md)
{
	return &md->stats;
}

D
Darrick J. Wong 已提交
447 448 449 450 451 452 453
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
454
static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
455 456
			unsigned int cmd, unsigned long arg)
{
A
Al Viro 已提交
457
	struct mapped_device *md = bdev->bd_disk->private_data;
M
Mikulas Patocka 已提交
458
	int srcu_idx;
459
	struct dm_table *map;
460 461 462
	struct dm_target *tgt;
	int r = -ENOTTY;

463
retry:
M
Mikulas Patocka 已提交
464 465
	map = dm_get_live_table(md, &srcu_idx);

466 467 468 469 470 471 472 473 474
	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);

475
	if (dm_suspended_md(md)) {
476 477 478 479 480
		r = -EAGAIN;
		goto out;
	}

	if (tgt->type->ioctl)
481
		r = tgt->type->ioctl(tgt, cmd, arg);
482 483

out:
M
Mikulas Patocka 已提交
484
	dm_put_live_table(md, srcu_idx);
485

486 487 488 489 490
	if (r == -ENOTCONN) {
		msleep(10);
		goto retry;
	}

491 492 493
	return r;
}

A
Alasdair G Kergon 已提交
494
static struct dm_io *alloc_io(struct mapped_device *md)
L
Linus Torvalds 已提交
495 496 497 498
{
	return mempool_alloc(md->io_pool, GFP_NOIO);
}

A
Alasdair G Kergon 已提交
499
static void free_io(struct mapped_device *md, struct dm_io *io)
L
Linus Torvalds 已提交
500 501 502 503
{
	mempool_free(io, md->io_pool);
}

A
Alasdair G Kergon 已提交
504
static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
L
Linus Torvalds 已提交
505
{
506
	bio_put(&tio->clone);
L
Linus Torvalds 已提交
507 508
}

K
Kiyoshi Ueda 已提交
509 510
static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
					    gfp_t gfp_mask)
511
{
J
Jun'ichi Nomura 已提交
512
	return mempool_alloc(md->io_pool, gfp_mask);
513 514 515 516
}

static void free_rq_tio(struct dm_rq_target_io *tio)
{
J
Jun'ichi Nomura 已提交
517
	mempool_free(tio, tio->md->io_pool);
518 519
}

K
Kiyoshi Ueda 已提交
520 521 522 523 524 525
static int md_in_flight(struct mapped_device *md)
{
	return atomic_read(&md->pending[READ]) +
	       atomic_read(&md->pending[WRITE]);
}

526 527 528
static void start_io_acct(struct dm_io *io)
{
	struct mapped_device *md = io->md;
M
Mikulas Patocka 已提交
529
	struct bio *bio = io->bio;
T
Tejun Heo 已提交
530
	int cpu;
M
Mikulas Patocka 已提交
531
	int rw = bio_data_dir(bio);
532 533 534

	io->start_time = jiffies;

T
Tejun Heo 已提交
535 536 537
	cpu = part_stat_lock();
	part_round_stats(cpu, &dm_disk(md)->part0);
	part_stat_unlock();
538 539
	atomic_set(&dm_disk(md)->part0.in_flight[rw],
		atomic_inc_return(&md->pending[rw]));
M
Mikulas Patocka 已提交
540 541 542 543

	if (unlikely(dm_stats_used(&md->stats)))
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_sector,
				    bio_sectors(bio), false, 0, &io->stats_aux);
544 545
}

546
static void end_io_acct(struct dm_io *io)
547 548 549 550
{
	struct mapped_device *md = io->md;
	struct bio *bio = io->bio;
	unsigned long duration = jiffies - io->start_time;
T
Tejun Heo 已提交
551
	int pending, cpu;
552 553
	int rw = bio_data_dir(bio);

T
Tejun Heo 已提交
554 555 556 557
	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();
558

M
Mikulas Patocka 已提交
559 560 561 562
	if (unlikely(dm_stats_used(&md->stats)))
		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_sector,
				    bio_sectors(bio), true, duration, &io->stats_aux);

563 564
	/*
	 * After this is decremented the bio must not be touched if it is
565
	 * a flush.
566
	 */
567 568
	pending = atomic_dec_return(&md->pending[rw]);
	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
569
	pending += atomic_read(&md->pending[rw^0x1]);
570

571 572 573
	/* nudge anyone waiting on suspend queue */
	if (!pending)
		wake_up(&md->wait);
574 575
}

L
Linus Torvalds 已提交
576 577 578
/*
 * Add the bio to the list of deferred io.
 */
M
Mikulas Patocka 已提交
579
static void queue_io(struct mapped_device *md, struct bio *bio)
L
Linus Torvalds 已提交
580
{
581
	unsigned long flags;
L
Linus Torvalds 已提交
582

583
	spin_lock_irqsave(&md->deferred_lock, flags);
L
Linus Torvalds 已提交
584
	bio_list_add(&md->deferred, bio);
585
	spin_unlock_irqrestore(&md->deferred_lock, flags);
586
	queue_work(md->wq, &md->work);
L
Linus Torvalds 已提交
587 588 589 590 591
}

/*
 * Everyone (including functions in this file), should use this
 * function to access the md->map field, and make sure they call
M
Mikulas Patocka 已提交
592
 * dm_put_live_table() when finished.
L
Linus Torvalds 已提交
593
 */
M
Mikulas Patocka 已提交
594
struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
L
Linus Torvalds 已提交
595
{
M
Mikulas Patocka 已提交
596 597 598 599
	*srcu_idx = srcu_read_lock(&md->io_barrier);

	return srcu_dereference(md->map, &md->io_barrier);
}
L
Linus Torvalds 已提交
600

M
Mikulas Patocka 已提交
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620
void dm_put_live_table(struct mapped_device *md, int srcu_idx) __releases(md->io_barrier)
{
	srcu_read_unlock(&md->io_barrier, srcu_idx);
}

void dm_sync_table(struct mapped_device *md)
{
	synchronize_srcu(&md->io_barrier);
	synchronize_rcu_expedited();
}

/*
 * A fast alternative to dm_get_live_table/dm_put_live_table.
 * The caller must not block between these two functions.
 */
static struct dm_table *dm_get_live_table_fast(struct mapped_device *md) __acquires(RCU)
{
	rcu_read_lock();
	return rcu_dereference(md->map);
}
L
Linus Torvalds 已提交
621

M
Mikulas Patocka 已提交
622 623 624
static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
{
	rcu_read_unlock();
L
Linus Torvalds 已提交
625 626
}

D
Darrick J. Wong 已提交
627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653
/*
 * 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
654 655 656 657 658 659 660 661 662
/*-----------------------------------------------------------------
 * 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.
 *---------------------------------------------------------------*/

663 664 665 666 667
static int __noflush_suspending(struct mapped_device *md)
{
	return test_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
}

L
Linus Torvalds 已提交
668 669 670 671
/*
 * Decrements the number of outstanding ios that a bio has been
 * cloned into, completing the original io if necc.
 */
672
static void dec_pending(struct dm_io *io, int error)
L
Linus Torvalds 已提交
673
{
674
	unsigned long flags;
675 676 677
	int io_error;
	struct bio *bio;
	struct mapped_device *md = io->md;
678 679

	/* Push-back supersedes any I/O errors */
680 681 682 683 684 685
	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
686 687

	if (atomic_dec_and_test(&io->io_count)) {
688 689 690 691
		if (io->error == DM_ENDIO_REQUEUE) {
			/*
			 * Target requested pushing back the I/O.
			 */
692
			spin_lock_irqsave(&md->deferred_lock, flags);
693 694 695
			if (__noflush_suspending(md))
				bio_list_add_head(&md->deferred, io->bio);
			else
696 697
				/* noflush suspend was interrupted. */
				io->error = -EIO;
698
			spin_unlock_irqrestore(&md->deferred_lock, flags);
699 700
		}

701 702
		io_error = io->error;
		bio = io->bio;
703 704 705 706 707
		end_io_acct(io);
		free_io(md, io);

		if (io_error == DM_ENDIO_REQUEUE)
			return;
708

709
		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_size) {
710
			/*
711 712
			 * Preflush done for flush with data, reissue
			 * without REQ_FLUSH.
713
			 */
714 715
			bio->bi_rw &= ~REQ_FLUSH;
			queue_io(md, bio);
716
		} else {
717
			/* done with normal IO or empty flush */
718
			trace_block_bio_complete(md->queue, bio, io_error);
719
			bio_endio(bio, io_error);
720
		}
L
Linus Torvalds 已提交
721 722 723
	}
}

724
static void clone_endio(struct bio *bio, int error)
L
Linus Torvalds 已提交
725 726
{
	int r = 0;
A
Alasdair G Kergon 已提交
727
	struct dm_target_io *tio = bio->bi_private;
728
	struct dm_io *io = tio->io;
S
Stefan Bader 已提交
729
	struct mapped_device *md = tio->io->md;
L
Linus Torvalds 已提交
730 731 732 733 734 735
	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
736
		r = endio(tio->ti, bio, error);
737 738 739 740 741
		if (r < 0 || r == DM_ENDIO_REQUEUE)
			/*
			 * error and requeue request are handled
			 * in dec_pending().
			 */
L
Linus Torvalds 已提交
742
			error = r;
743 744
		else if (r == DM_ENDIO_INCOMPLETE)
			/* The target will handle the io */
745
			return;
746 747 748 749
		else if (r) {
			DMWARN("unimplemented target endio return value: %d", r);
			BUG();
		}
L
Linus Torvalds 已提交
750 751
	}

S
Stefan Bader 已提交
752
	free_tio(md, tio);
753
	dec_pending(io, error);
L
Linus Torvalds 已提交
754 755
}

756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810
/*
 * 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.
 */
811
static void rq_completed(struct mapped_device *md, int rw, int run_queue)
812
{
813
	atomic_dec(&md->pending[rw]);
814 815

	/* nudge anyone waiting on suspend queue */
816
	if (!md_in_flight(md))
817 818
		wake_up(&md->wait);

819 820 821 822 823 824
	/*
	 * 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.
	 */
825
	if (run_queue)
826
		blk_run_queue_async(md->queue);
827 828 829 830 831 832 833

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

834 835 836 837 838 839 840 841
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
842 843 844 845 846 847 848 849 850 851 852
/*
 * 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;

853
	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
K
Kiyoshi Ueda 已提交
854 855 856 857 858 859 860 861 862 863 864 865 866
		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);
867 868
	blk_end_request_all(rq, error);
	rq_completed(md, rw, true);
K
Kiyoshi Ueda 已提交
869 870
}

871 872 873 874 875 876 877
static void dm_unprep_request(struct request *rq)
{
	struct request *clone = rq->special;

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

878
	free_rq_clone(clone);
879 880 881 882 883 884 885
}

/*
 * Requeue the original request of a clone.
 */
void dm_requeue_unmapped_request(struct request *clone)
{
886
	int rw = rq_data_dir(clone);
887 888 889 890 891 892 893 894 895 896 897 898
	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);

899
	rq_completed(md, rw, 0);
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
}
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);
}

932
static void dm_done(struct request *clone, int error, bool mapped)
933
{
934
	int r = error;
935
	struct dm_rq_target_io *tio = clone->end_io_data;
936
	dm_request_endio_fn rq_end_io = NULL;
937

938 939 940 941 942 943
	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);
	}
944

945
	if (r <= 0)
946
		/* The target wants to complete the I/O */
947 948
		dm_end_request(clone, r);
	else if (r == DM_ENDIO_INCOMPLETE)
949 950
		/* The target will handle the I/O */
		return;
951
	else if (r == DM_ENDIO_REQUEUE)
952 953 954
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
	else {
955
		DMWARN("unimplemented target endio return value: %d", r);
956 957 958 959
		BUG();
	}
}

960 961 962 963 964 965 966 967 968 969 970 971 972 973 974
/*
 * 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);
}

975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
/*
 * 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);
}

1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
/*
 * 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
1041
{
1042
	sector_t len = max_io_len_target_boundary(sector, ti);
1043
	sector_t offset, max_len;
L
Linus Torvalds 已提交
1044 1045

	/*
1046
	 * Does the target need to split even further?
L
Linus Torvalds 已提交
1047
	 */
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057
	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
1058 1059 1060 1061 1062
	}

	return len;
}

1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 1076 1077
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
1078
static void __map_bio(struct dm_target_io *tio)
L
Linus Torvalds 已提交
1079 1080
{
	int r;
1081
	sector_t sector;
S
Stefan Bader 已提交
1082
	struct mapped_device *md;
1083
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1084
	struct dm_target *ti = tio->ti;
L
Linus Torvalds 已提交
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094

	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);
1095
	sector = clone->bi_sector;
M
Mikulas Patocka 已提交
1096
	r = ti->type->map(ti, clone);
1097
	if (r == DM_MAPIO_REMAPPED) {
L
Linus Torvalds 已提交
1098
		/* the bio has been remapped so dispatch it */
1099

1100 1101
		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
				      tio->io->bio->bi_bdev->bd_dev, sector);
1102

L
Linus Torvalds 已提交
1103
		generic_make_request(clone);
1104 1105
	} 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
1106 1107 1108
		md = tio->io->md;
		dec_pending(tio->io, r);
		free_tio(md, tio);
1109 1110 1111
	} else if (r) {
		DMWARN("unimplemented target map return value: %d", r);
		BUG();
L
Linus Torvalds 已提交
1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124
	}
}

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
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150
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
1151
/*
1152
 * Creates a little bio that just does part of a bvec.
L
Linus Torvalds 已提交
1153
 */
1154 1155 1156
static void clone_split_bio(struct dm_target_io *tio, struct bio *bio,
			    sector_t sector, unsigned short idx,
			    unsigned offset, unsigned len)
L
Linus Torvalds 已提交
1157
{
1158
	struct bio *clone = &tio->clone;
L
Linus Torvalds 已提交
1159 1160 1161 1162
	struct bio_vec *bv = bio->bi_io_vec + idx;

	*clone->bi_io_vec = *bv;

A
Alasdair G Kergon 已提交
1163 1164
	bio_setup_sector(clone, sector, len);

L
Linus Torvalds 已提交
1165
	clone->bi_bdev = bio->bi_bdev;
1166
	clone->bi_rw = bio->bi_rw;
L
Linus Torvalds 已提交
1167 1168 1169
	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
1170
	clone->bi_flags |= 1 << BIO_CLONED;
L
Linus Torvalds 已提交
1171

A
Alasdair G Kergon 已提交
1172
	clone_bio_integrity(bio, clone, idx, len, offset, 1);
L
Linus Torvalds 已提交
1173 1174 1175 1176 1177
}

/*
 * Creates a bio that consists of range of complete bvecs.
 */
1178 1179
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
1180
		      unsigned short bv_count, unsigned len)
L
Linus Torvalds 已提交
1181
{
1182
	struct bio *clone = &tio->clone;
A
Alasdair G Kergon 已提交
1183
	unsigned trim = 0;
L
Linus Torvalds 已提交
1184

S
Stefan Bader 已提交
1185
	__bio_clone(clone, bio);
A
Alasdair G Kergon 已提交
1186 1187 1188 1189 1190 1191
	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
1192 1193
}

1194
static struct dm_target_io *alloc_tio(struct clone_info *ci,
A
Alasdair G Kergon 已提交
1195
				      struct dm_target *ti, int nr_iovecs,
1196
				      unsigned target_bio_nr)
1197
{
1198 1199 1200 1201 1202
	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);
1203 1204 1205 1206

	tio->io = ci->io;
	tio->ti = ti;
	memset(&tio->info, 0, sizeof(tio->info));
1207
	tio->target_bio_nr = target_bio_nr;
1208 1209 1210 1211

	return tio;
}

1212 1213 1214
static void __clone_and_map_simple_bio(struct clone_info *ci,
				       struct dm_target *ti,
				       unsigned target_bio_nr, sector_t len)
1215
{
1216
	struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs, target_bio_nr);
1217
	struct bio *clone = &tio->clone;
1218

1219 1220 1221 1222 1223
	/*
	 * 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.
	 */
1224
	 __bio_clone(clone, ci->bio);
A
Alasdair G Kergon 已提交
1225 1226
	if (len)
		bio_setup_sector(clone, ci->sector, len);
1227

A
Alasdair G Kergon 已提交
1228
	__map_bio(tio);
1229 1230
}

1231 1232
static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
				  unsigned num_bios, sector_t len)
1233
{
1234
	unsigned target_bio_nr;
1235

1236
	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
1237
		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
1238 1239
}

1240
static int __send_empty_flush(struct clone_info *ci)
1241
{
1242
	unsigned target_nr = 0;
1243 1244
	struct dm_target *ti;

1245
	BUG_ON(bio_has_data(ci->bio));
1246
	while ((ti = dm_table_get_target(ci->map, target_nr++)))
1247
		__send_duplicate_bios(ci, ti, ti->num_flush_bios, 0);
1248 1249 1250 1251

	return 0;
}

A
Alasdair G Kergon 已提交
1252 1253 1254 1255 1256
static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
				     sector_t sector, int nr_iovecs,
				     unsigned short idx, unsigned short bv_count,
				     unsigned offset, unsigned len,
				     unsigned split_bvec)
M
Mike Snitzer 已提交
1257
{
1258
	struct bio *bio = ci->bio;
M
Mike Snitzer 已提交
1259
	struct dm_target_io *tio;
1260 1261
	unsigned target_bio_nr;
	unsigned num_target_bios = 1;
M
Mike Snitzer 已提交
1262

1263 1264 1265 1266 1267
	/*
	 * Does the target want to receive duplicate copies of the bio?
	 */
	if (bio_data_dir(bio) == WRITE && ti->num_write_bios)
		num_target_bios = ti->num_write_bios(ti, bio);
A
Alasdair G Kergon 已提交
1268

1269 1270 1271 1272 1273 1274 1275 1276
	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
		tio = alloc_tio(ci, ti, nr_iovecs, target_bio_nr);
		if (split_bvec)
			clone_split_bio(tio, bio, sector, idx, offset, len);
		else
			clone_bio(tio, bio, sector, idx, bv_count, len);
		__map_bio(tio);
	}
M
Mike Snitzer 已提交
1277 1278
}

1279
typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
M
Mike Snitzer 已提交
1280

1281
static unsigned get_num_discard_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1282
{
1283
	return ti->num_discard_bios;
M
Mike Snitzer 已提交
1284 1285
}

1286
static unsigned get_num_write_same_bios(struct dm_target *ti)
M
Mike Snitzer 已提交
1287
{
1288
	return ti->num_write_same_bios;
M
Mike Snitzer 已提交
1289 1290 1291 1292 1293 1294
}

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

static bool is_split_required_for_discard(struct dm_target *ti)
{
1295
	return ti->split_discard_bios;
M
Mike Snitzer 已提交
1296 1297
}

1298 1299 1300
static int __send_changing_extent_only(struct clone_info *ci,
				       get_num_bios_fn get_num_bios,
				       is_split_required_fn is_split_required)
M
Mike Snitzer 已提交
1301 1302
{
	struct dm_target *ti;
1303
	sector_t len;
1304
	unsigned num_bios;
M
Mike Snitzer 已提交
1305

1306 1307 1308 1309
	do {
		ti = dm_table_find_target(ci->map, ci->sector);
		if (!dm_target_is_valid(ti))
			return -EIO;
M
Mike Snitzer 已提交
1310 1311

		/*
M
Mike Snitzer 已提交
1312 1313
		 * 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
1314
		 * reconfiguration might also have changed that since the
1315
		 * check was performed.
M
Mike Snitzer 已提交
1316
		 */
1317 1318
		num_bios = get_num_bios ? get_num_bios(ti) : 0;
		if (!num_bios)
1319
			return -EOPNOTSUPP;
M
Mike Snitzer 已提交
1320

M
Mike Snitzer 已提交
1321
		if (is_split_required && !is_split_required(ti))
1322 1323 1324
			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));
1325

1326
		__send_duplicate_bios(ci, ti, num_bios, len);
1327 1328 1329

		ci->sector += len;
	} while (ci->sector_count -= len);
M
Mike Snitzer 已提交
1330 1331 1332 1333

	return 0;
}

1334
static int __send_discard(struct clone_info *ci)
M
Mike Snitzer 已提交
1335
{
1336 1337
	return __send_changing_extent_only(ci, get_num_discard_bios,
					   is_split_required_for_discard);
M
Mike Snitzer 已提交
1338 1339
}

1340
static int __send_write_same(struct clone_info *ci)
M
Mike Snitzer 已提交
1341
{
1342
	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
M
Mike Snitzer 已提交
1343 1344
}

A
Alasdair G Kergon 已提交
1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398 1399 1400 1401
/*
 * Find maximum number of sectors / bvecs we can process with a single bio.
 */
static sector_t __len_within_target(struct clone_info *ci, sector_t max, int *idx)
{
	struct bio *bio = ci->bio;
	sector_t bv_len, total_len = 0;

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

		if (bv_len > max)
			break;

		max -= bv_len;
		total_len += bv_len;
	}

	return total_len;
}

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

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

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

		len = min(remaining, max);

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

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

	ci->idx++;

	return 0;
}

/*
 * Select the correct strategy for processing a non-flush bio.
 */
1402
static int __split_and_process_non_flush(struct clone_info *ci)
L
Linus Torvalds 已提交
1403
{
1404
	struct bio *bio = ci->bio;
1405
	struct dm_target *ti;
A
Alasdair G Kergon 已提交
1406 1407
	sector_t len, max;
	int idx;
L
Linus Torvalds 已提交
1408

M
Mike Snitzer 已提交
1409
	if (unlikely(bio->bi_rw & REQ_DISCARD))
1410
		return __send_discard(ci);
M
Mike Snitzer 已提交
1411
	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
1412
		return __send_write_same(ci);
M
Mike Snitzer 已提交
1413

1414 1415 1416 1417
	ti = dm_table_find_target(ci->map, ci->sector);
	if (!dm_target_is_valid(ti))
		return -EIO;

1418
	max = max_io_len(ci->sector, ti);
1419

A
Alasdair G Kergon 已提交
1420 1421 1422 1423
	/*
	 * Optimise for the simple case where we can do all of
	 * the remaining io with a single clone.
	 */
L
Linus Torvalds 已提交
1424
	if (ci->sector_count <= max) {
A
Alasdair G Kergon 已提交
1425 1426 1427 1428 1429 1430
		__clone_and_map_data_bio(ci, ti, ci->sector, bio->bi_max_vecs,
					 ci->idx, bio->bi_vcnt - ci->idx, 0,
					 ci->sector_count, 0);
		ci->sector_count = 0;
		return 0;
	}
L
Linus Torvalds 已提交
1431

A
Alasdair G Kergon 已提交
1432 1433 1434 1435 1436 1437
	/*
	 * There are some bvecs that don't span targets.
	 * Do as many of these as possible.
	 */
	if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) {
		len = __len_within_target(ci, max, &idx);
L
Linus Torvalds 已提交
1438

A
Alasdair G Kergon 已提交
1439 1440
		__clone_and_map_data_bio(ci, ti, ci->sector, bio->bi_max_vecs,
					 ci->idx, idx - ci->idx, 0, len, 0);
L
Linus Torvalds 已提交
1441 1442 1443

		ci->sector += len;
		ci->sector_count -= len;
A
Alasdair G Kergon 已提交
1444
		ci->idx = idx;
1445

A
Alasdair G Kergon 已提交
1446
		return 0;
L
Linus Torvalds 已提交
1447
	}
1448

A
Alasdair G Kergon 已提交
1449 1450 1451 1452
	/*
	 * Handle a bvec that must be split between two or more targets.
	 */
	return __split_bvec_across_targets(ci, ti, max);
L
Linus Torvalds 已提交
1453 1454 1455
}

/*
1456
 * Entry point to split a bio into clones and submit them to the targets.
L
Linus Torvalds 已提交
1457
 */
M
Mikulas Patocka 已提交
1458 1459
static void __split_and_process_bio(struct mapped_device *md,
				    struct dm_table *map, struct bio *bio)
L
Linus Torvalds 已提交
1460 1461
{
	struct clone_info ci;
1462
	int error = 0;
L
Linus Torvalds 已提交
1463

M
Mikulas Patocka 已提交
1464
	if (unlikely(!map)) {
1465
		bio_io_error(bio);
1466 1467
		return;
	}
1468

M
Mikulas Patocka 已提交
1469
	ci.map = map;
L
Linus Torvalds 已提交
1470 1471 1472 1473 1474 1475
	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;
1476
	spin_lock_init(&ci.io->endio_lock);
L
Linus Torvalds 已提交
1477 1478 1479
	ci.sector = bio->bi_sector;
	ci.idx = bio->bi_idx;

1480
	start_io_acct(ci.io);
A
Alasdair G Kergon 已提交
1481

1482 1483 1484
	if (bio->bi_rw & REQ_FLUSH) {
		ci.bio = &ci.md->flush_bio;
		ci.sector_count = 0;
1485
		error = __send_empty_flush(&ci);
1486 1487
		/* dec_pending submits any data associated with flush */
	} else {
1488
		ci.bio = bio;
1489
		ci.sector_count = bio_sectors(bio);
1490
		while (ci.sector_count && !error)
1491
			error = __split_and_process_non_flush(&ci);
1492
	}
L
Linus Torvalds 已提交
1493 1494

	/* drop the extra reference count */
1495
	dec_pending(ci.io, error);
L
Linus Torvalds 已提交
1496 1497 1498 1499 1500
}
/*-----------------------------------------------------------------
 * CRUD END
 *---------------------------------------------------------------*/

M
Milan Broz 已提交
1501 1502 1503 1504 1505
static int dm_merge_bvec(struct request_queue *q,
			 struct bvec_merge_data *bvm,
			 struct bio_vec *biovec)
{
	struct mapped_device *md = q->queuedata;
M
Mikulas Patocka 已提交
1506
	struct dm_table *map = dm_get_live_table_fast(md);
M
Milan Broz 已提交
1507 1508
	struct dm_target *ti;
	sector_t max_sectors;
1509
	int max_size = 0;
M
Milan Broz 已提交
1510 1511

	if (unlikely(!map))
1512
		goto out;
M
Milan Broz 已提交
1513 1514

	ti = dm_table_find_target(map, bvm->bi_sector);
1515
	if (!dm_target_is_valid(ti))
M
Mikulas Patocka 已提交
1516
		goto out;
M
Milan Broz 已提交
1517 1518 1519 1520

	/*
	 * Find maximum amount of I/O that won't need splitting
	 */
1521
	max_sectors = min(max_io_len(bvm->bi_sector, ti),
M
Milan Broz 已提交
1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533
			  (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);
1534 1535 1536 1537 1538 1539 1540 1541 1542 1543
	/*
	 * 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
1544

1545
out:
M
Mikulas Patocka 已提交
1546
	dm_put_live_table_fast(md);
M
Milan Broz 已提交
1547 1548 1549 1550 1551 1552 1553 1554 1555
	/*
	 * 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
1556 1557 1558 1559
/*
 * The request function that just remaps the bio built up by
 * dm_merge_bvec.
 */
1560
static void _dm_request(struct request_queue *q, struct bio *bio)
L
Linus Torvalds 已提交
1561
{
1562
	int rw = bio_data_dir(bio);
L
Linus Torvalds 已提交
1563
	struct mapped_device *md = q->queuedata;
T
Tejun Heo 已提交
1564
	int cpu;
M
Mikulas Patocka 已提交
1565 1566
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
1567

M
Mikulas Patocka 已提交
1568
	map = dm_get_live_table(md, &srcu_idx);
L
Linus Torvalds 已提交
1569

T
Tejun Heo 已提交
1570 1571 1572 1573
	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();
1574

1575 1576
	/* if we're suspended, we have to queue this io for later */
	if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
M
Mikulas Patocka 已提交
1577
		dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
1578

1579 1580 1581
		if (bio_rw(bio) != READA)
			queue_io(md, bio);
		else
A
Alasdair G Kergon 已提交
1582
			bio_io_error(bio);
1583
		return;
L
Linus Torvalds 已提交
1584 1585
	}

M
Mikulas Patocka 已提交
1586 1587
	__split_and_process_bio(md, map, bio);
	dm_put_live_table(md, srcu_idx);
1588
	return;
1589 1590
}

M
Mikulas Patocka 已提交
1591
int dm_request_based(struct mapped_device *md)
1592 1593 1594 1595
{
	return blk_queue_stackable(md->queue);
}

1596
static void dm_request(struct request_queue *q, struct bio *bio)
1597 1598 1599 1600
{
	struct mapped_device *md = q->queuedata;

	if (dm_request_based(md))
1601 1602 1603
		blk_queue_bio(q, bio);
	else
		_dm_request(q, bio);
1604 1605 1606 1607 1608 1609 1610 1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623
}

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;
1624 1625
	struct dm_rq_clone_bio_info *info =
		container_of(bio, struct dm_rq_clone_bio_info, clone);
1626 1627 1628 1629 1630 1631 1632 1633 1634 1635 1636 1637

	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)
{
1638
	int r;
1639

1640 1641 1642 1643
	r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
			      dm_rq_bio_constructor, tio);
	if (r)
		return r;
1644

1645 1646 1647 1648
	clone->cmd = rq->cmd;
	clone->cmd_len = rq->cmd_len;
	clone->sense = rq->sense;
	clone->buffer = rq->buffer;
1649 1650 1651 1652 1653 1654
	clone->end_io = end_clone_request;
	clone->end_io_data = tio;

	return 0;
}

K
Kiyoshi Ueda 已提交
1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680
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;
}

1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693
/*
 * 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
1694 1695
	clone = clone_rq(rq, md, GFP_ATOMIC);
	if (!clone)
1696 1697 1698 1699 1700 1701 1702 1703
		return BLKPREP_DEFER;

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

	return BLKPREP_OK;
}

1704 1705 1706 1707 1708 1709 1710
/*
 * 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)
1711
{
1712
	int r, requeued = 0;
1713 1714 1715 1716 1717 1718 1719 1720 1721 1722
	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 */
1723 1724
		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
				     blk_rq_pos(tio->orig));
1725 1726 1727 1728 1729
		dm_dispatch_request(clone);
		break;
	case DM_MAPIO_REQUEUE:
		/* The target wants to requeue the I/O */
		dm_requeue_unmapped_request(clone);
1730
		requeued = 1;
1731 1732 1733 1734 1735 1736 1737 1738 1739 1740 1741
		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;
	}
1742 1743

	return requeued;
1744 1745
}

1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765
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;
}

1766 1767 1768 1769 1770 1771 1772
/*
 * 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;
M
Mikulas Patocka 已提交
1773 1774
	int srcu_idx;
	struct dm_table *map = dm_get_live_table(md, &srcu_idx);
1775
	struct dm_target *ti;
1776
	struct request *rq, *clone;
1777
	sector_t pos;
1778 1779

	/*
1780 1781 1782 1783
	 * 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().
1784
	 */
J
Jens Axboe 已提交
1785
	while (!blk_queue_stopped(q)) {
1786 1787
		rq = blk_peek_request(q);
		if (!rq)
J
Jens Axboe 已提交
1788
			goto delay_and_out;
1789

1790 1791 1792 1793 1794 1795
		/* 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);
1796 1797 1798 1799 1800 1801 1802 1803 1804 1805
		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;
		}
1806

1807
		if (ti->type->busy && ti->type->busy(ti))
J
Jens Axboe 已提交
1808
			goto delay_and_out;
1809

1810
		clone = dm_start_request(md, rq);
1811

1812
		spin_unlock(q->queue_lock);
1813 1814 1815
		if (map_request(ti, clone, md))
			goto requeued;

1816 1817
		BUG_ON(!irqs_disabled());
		spin_lock(q->queue_lock);
1818 1819 1820 1821
	}

	goto out;

1822
requeued:
1823 1824
	BUG_ON(!irqs_disabled());
	spin_lock(q->queue_lock);
1825

J
Jens Axboe 已提交
1826 1827
delay_and_out:
	blk_delay_queue(q, HZ / 10);
1828
out:
M
Mikulas Patocka 已提交
1829
	dm_put_live_table(md, srcu_idx);
1830 1831 1832 1833 1834 1835 1836 1837 1838 1839 1840 1841
}

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;
M
Mikulas Patocka 已提交
1842
	struct dm_table *map = dm_get_live_table_fast(md);
1843 1844 1845 1846 1847 1848

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

M
Mikulas Patocka 已提交
1849
	dm_put_live_table_fast(md);
1850 1851 1852 1853

	return r;
}

L
Linus Torvalds 已提交
1854 1855
static int dm_any_congested(void *congested_data, int bdi_bits)
{
1856 1857 1858
	int r = bdi_bits;
	struct mapped_device *md = congested_data;
	struct dm_table *map;
L
Linus Torvalds 已提交
1859

1860
	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
M
Mikulas Patocka 已提交
1861
		map = dm_get_live_table_fast(md);
1862
		if (map) {
1863 1864 1865 1866 1867 1868 1869 1870 1871
			/*
			 * 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);
1872
		}
M
Mikulas Patocka 已提交
1873
		dm_put_live_table_fast(md);
1874 1875
	}

L
Linus Torvalds 已提交
1876 1877 1878 1879 1880 1881
	return r;
}

/*-----------------------------------------------------------------
 * An IDR is used to keep track of allocated minor numbers.
 *---------------------------------------------------------------*/
1882
static void free_minor(int minor)
L
Linus Torvalds 已提交
1883
{
1884
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1885
	idr_remove(&_minor_idr, minor);
1886
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
1887 1888 1889 1890 1891
}

/*
 * See if the device with a specific minor # is free.
 */
1892
static int specific_minor(int minor)
L
Linus Torvalds 已提交
1893
{
T
Tejun Heo 已提交
1894
	int r;
L
Linus Torvalds 已提交
1895 1896 1897 1898

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

T
Tejun Heo 已提交
1899
	idr_preload(GFP_KERNEL);
1900
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1901

T
Tejun Heo 已提交
1902
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
L
Linus Torvalds 已提交
1903

1904
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1905 1906 1907 1908
	idr_preload_end();
	if (r < 0)
		return r == -ENOSPC ? -EBUSY : r;
	return 0;
L
Linus Torvalds 已提交
1909 1910
}

1911
static int next_free_minor(int *minor)
L
Linus Torvalds 已提交
1912
{
T
Tejun Heo 已提交
1913
	int r;
J
Jeff Mahoney 已提交
1914

T
Tejun Heo 已提交
1915
	idr_preload(GFP_KERNEL);
1916
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
1917

T
Tejun Heo 已提交
1918
	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
L
Linus Torvalds 已提交
1919

1920
	spin_unlock(&_minor_lock);
T
Tejun Heo 已提交
1921 1922 1923 1924 1925
	idr_preload_end();
	if (r < 0)
		return r;
	*minor = r;
	return 0;
L
Linus Torvalds 已提交
1926 1927
}

1928
static const struct block_device_operations dm_blk_dops;
L
Linus Torvalds 已提交
1929

1930 1931
static void dm_wq_work(struct work_struct *work);

1932 1933 1934 1935 1936 1937 1938 1939 1940 1941 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952
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
1953 1954 1955
/*
 * Allocate and initialise a blank device with a given minor.
 */
1956
static struct mapped_device *alloc_dev(int minor)
L
Linus Torvalds 已提交
1957 1958
{
	int r;
1959
	struct mapped_device *md = kzalloc(sizeof(*md), GFP_KERNEL);
1960
	void *old_md;
L
Linus Torvalds 已提交
1961 1962 1963 1964 1965 1966

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

1967
	if (!try_module_get(THIS_MODULE))
M
Milan Broz 已提交
1968
		goto bad_module_get;
1969

L
Linus Torvalds 已提交
1970
	/* get a minor number for the dev */
1971
	if (minor == DM_ANY_MINOR)
1972
		r = next_free_minor(&minor);
1973
	else
1974
		r = specific_minor(minor);
L
Linus Torvalds 已提交
1975
	if (r < 0)
M
Milan Broz 已提交
1976
		goto bad_minor;
L
Linus Torvalds 已提交
1977

M
Mikulas Patocka 已提交
1978 1979 1980 1981
	r = init_srcu_struct(&md->io_barrier);
	if (r < 0)
		goto bad_io_barrier;

1982
	md->type = DM_TYPE_NONE;
1983
	mutex_init(&md->suspend_lock);
1984
	mutex_init(&md->type_lock);
1985
	spin_lock_init(&md->deferred_lock);
L
Linus Torvalds 已提交
1986
	atomic_set(&md->holders, 1);
1987
	atomic_set(&md->open_count, 0);
L
Linus Torvalds 已提交
1988
	atomic_set(&md->event_nr, 0);
M
Mike Anderson 已提交
1989 1990 1991
	atomic_set(&md->uevent_seq, 0);
	INIT_LIST_HEAD(&md->uevent_list);
	spin_lock_init(&md->uevent_lock);
L
Linus Torvalds 已提交
1992

1993
	md->queue = blk_alloc_queue(GFP_KERNEL);
L
Linus Torvalds 已提交
1994
	if (!md->queue)
M
Milan Broz 已提交
1995
		goto bad_queue;
L
Linus Torvalds 已提交
1996

1997
	dm_init_md_queue(md);
S
Stefan Bader 已提交
1998

L
Linus Torvalds 已提交
1999 2000
	md->disk = alloc_disk(1);
	if (!md->disk)
M
Milan Broz 已提交
2001
		goto bad_disk;
L
Linus Torvalds 已提交
2002

2003 2004
	atomic_set(&md->pending[0], 0);
	atomic_set(&md->pending[1], 0);
2005
	init_waitqueue_head(&md->wait);
2006
	INIT_WORK(&md->work, dm_wq_work);
2007 2008
	init_waitqueue_head(&md->eventq);

L
Linus Torvalds 已提交
2009 2010 2011 2012 2013 2014 2015
	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
2016
	format_dev_t(md->name, MKDEV(_major, minor));
L
Linus Torvalds 已提交
2017

T
Tejun Heo 已提交
2018
	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
2019 2020 2021
	if (!md->wq)
		goto bad_thread;

M
Mikulas Patocka 已提交
2022 2023 2024 2025
	md->bdev = bdget_disk(md->disk, 0);
	if (!md->bdev)
		goto bad_bdev;

2026 2027 2028 2029
	bio_init(&md->flush_bio);
	md->flush_bio.bi_bdev = md->bdev;
	md->flush_bio.bi_rw = WRITE_FLUSH;

M
Mikulas Patocka 已提交
2030 2031
	dm_stats_init(&md->stats);

2032
	/* Populate the mapping, nobody knows we exist yet */
2033
	spin_lock(&_minor_lock);
2034
	old_md = idr_replace(&_minor_idr, md, minor);
2035
	spin_unlock(&_minor_lock);
2036 2037 2038

	BUG_ON(old_md != MINOR_ALLOCED);

L
Linus Torvalds 已提交
2039 2040
	return md;

M
Mikulas Patocka 已提交
2041 2042
bad_bdev:
	destroy_workqueue(md->wq);
2043
bad_thread:
2044
	del_gendisk(md->disk);
2045
	put_disk(md->disk);
M
Milan Broz 已提交
2046
bad_disk:
2047
	blk_cleanup_queue(md->queue);
M
Milan Broz 已提交
2048
bad_queue:
M
Mikulas Patocka 已提交
2049 2050
	cleanup_srcu_struct(&md->io_barrier);
bad_io_barrier:
L
Linus Torvalds 已提交
2051
	free_minor(minor);
M
Milan Broz 已提交
2052
bad_minor:
2053
	module_put(THIS_MODULE);
M
Milan Broz 已提交
2054
bad_module_get:
L
Linus Torvalds 已提交
2055 2056 2057 2058
	kfree(md);
	return NULL;
}

J
Jun'ichi Nomura 已提交
2059 2060
static void unlock_fs(struct mapped_device *md);

L
Linus Torvalds 已提交
2061 2062
static void free_dev(struct mapped_device *md)
{
2063
	int minor = MINOR(disk_devt(md->disk));
2064

M
Mikulas Patocka 已提交
2065 2066
	unlock_fs(md);
	bdput(md->bdev);
2067
	destroy_workqueue(md->wq);
K
Kiyoshi Ueda 已提交
2068 2069 2070 2071
	if (md->io_pool)
		mempool_destroy(md->io_pool);
	if (md->bs)
		bioset_free(md->bs);
M
Martin K. Petersen 已提交
2072
	blk_integrity_unregister(md->disk);
L
Linus Torvalds 已提交
2073
	del_gendisk(md->disk);
M
Mikulas Patocka 已提交
2074
	cleanup_srcu_struct(&md->io_barrier);
2075
	free_minor(minor);
J
Jeff Mahoney 已提交
2076 2077 2078 2079 2080

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

L
Linus Torvalds 已提交
2081
	put_disk(md->disk);
2082
	blk_cleanup_queue(md->queue);
M
Mikulas Patocka 已提交
2083
	dm_stats_cleanup(&md->stats);
2084
	module_put(THIS_MODULE);
L
Linus Torvalds 已提交
2085 2086 2087
	kfree(md);
}

K
Kiyoshi Ueda 已提交
2088 2089
static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
{
M
Mikulas Patocka 已提交
2090
	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
K
Kiyoshi Ueda 已提交
2091

J
Jun'ichi Nomura 已提交
2092
	if (md->io_pool && md->bs) {
2093 2094 2095 2096 2097 2098 2099 2100 2101 2102 2103 2104 2105 2106 2107 2108 2109 2110 2111
		/* The md already has necessary mempools. */
		if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) {
			/*
			 * Reload bioset because front_pad may have changed
			 * because a different table was loaded.
			 */
			bioset_free(md->bs);
			md->bs = p->bs;
			p->bs = NULL;
		} else if (dm_table_get_type(t) == DM_TYPE_REQUEST_BASED) {
			/*
			 * There's no need to reload with request-based dm
			 * because the size of front_pad doesn't change.
			 * Note for future: If you are to reload bioset,
			 * prep-ed requests in the queue may refer
			 * to bio from the old bioset, so you must walk
			 * through the queue to unprep.
			 */
		}
K
Kiyoshi Ueda 已提交
2112
		goto out;
M
Mikulas Patocka 已提交
2113
	}
K
Kiyoshi Ueda 已提交
2114

J
Jun'ichi Nomura 已提交
2115
	BUG_ON(!p || md->io_pool || md->bs);
K
Kiyoshi Ueda 已提交
2116 2117 2118 2119 2120 2121 2122 2123 2124 2125 2126

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

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

L
Linus Torvalds 已提交
2127 2128 2129 2130 2131
/*
 * Bind a table to the device.
 */
static void event_callback(void *context)
{
M
Mike Anderson 已提交
2132 2133
	unsigned long flags;
	LIST_HEAD(uevents);
L
Linus Torvalds 已提交
2134 2135
	struct mapped_device *md = (struct mapped_device *) context;

M
Mike Anderson 已提交
2136 2137 2138 2139
	spin_lock_irqsave(&md->uevent_lock, flags);
	list_splice_init(&md->uevent_list, &uevents);
	spin_unlock_irqrestore(&md->uevent_lock, flags);

2140
	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
M
Mike Anderson 已提交
2141

L
Linus Torvalds 已提交
2142 2143 2144 2145
	atomic_inc(&md->event_nr);
	wake_up(&md->eventq);
}

2146 2147 2148
/*
 * Protected by md->suspend_lock obtained by dm_swap_table().
 */
2149
static void __set_size(struct mapped_device *md, sector_t size)
L
Linus Torvalds 已提交
2150
{
2151
	set_capacity(md->disk, size);
L
Linus Torvalds 已提交
2152

2153
	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
L
Linus Torvalds 已提交
2154 2155
}

2156 2157 2158 2159 2160 2161 2162 2163 2164 2165 2166 2167 2168 2169 2170 2171 2172 2173 2174 2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208
/*
 * 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;
}

2209 2210 2211 2212 2213
/*
 * 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
2214
{
2215
	struct dm_table *old_map;
2216
	struct request_queue *q = md->queue;
L
Linus Torvalds 已提交
2217
	sector_t size;
2218
	int merge_is_optional;
L
Linus Torvalds 已提交
2219 2220

	size = dm_table_get_size(t);
D
Darrick J. Wong 已提交
2221 2222 2223 2224

	/*
	 * Wipe any geometry if the size of the table changed.
	 */
M
Mikulas Patocka 已提交
2225
	if (size != dm_get_size(md))
D
Darrick J. Wong 已提交
2226 2227
		memset(&md->geometry, 0, sizeof(md->geometry));

M
Mikulas Patocka 已提交
2228
	__set_size(md, size);
2229

2230 2231
	dm_table_event_callback(t, event_callback, md);

K
Kiyoshi Ueda 已提交
2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243
	/*
	 * 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);

2244 2245
	merge_is_optional = dm_table_merge_is_optional(t);

2246
	old_map = md->map;
M
Mikulas Patocka 已提交
2247
	rcu_assign_pointer(md->map, t);
2248 2249
	md->immutable_target_type = dm_table_get_immutable_target_type(t);

2250
	dm_table_set_restrictions(t, q, limits);
2251 2252 2253 2254
	if (merge_is_optional)
		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
	else
		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
M
Mikulas Patocka 已提交
2255
	dm_sync_table(md);
L
Linus Torvalds 已提交
2256

2257
	return old_map;
L
Linus Torvalds 已提交
2258 2259
}

2260 2261 2262 2263
/*
 * Returns unbound table for the caller to free.
 */
static struct dm_table *__unbind(struct mapped_device *md)
L
Linus Torvalds 已提交
2264 2265 2266 2267
{
	struct dm_table *map = md->map;

	if (!map)
2268
		return NULL;
L
Linus Torvalds 已提交
2269 2270

	dm_table_event_callback(map, NULL, NULL);
M
Mikulas Patocka 已提交
2271 2272
	rcu_assign_pointer(md->map, NULL);
	dm_sync_table(md);
2273 2274

	return map;
L
Linus Torvalds 已提交
2275 2276 2277 2278 2279
}

/*
 * Constructor for a new device.
 */
2280
int dm_create(int minor, struct mapped_device **result)
L
Linus Torvalds 已提交
2281 2282 2283
{
	struct mapped_device *md;

2284
	md = alloc_dev(minor);
L
Linus Torvalds 已提交
2285 2286 2287
	if (!md)
		return -ENXIO;

M
Milan Broz 已提交
2288 2289
	dm_sysfs_init(md);

L
Linus Torvalds 已提交
2290 2291 2292 2293
	*result = md;
	return 0;
}

2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306 2307 2308 2309
/*
 * 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)
{
2310
	BUG_ON(!mutex_is_locked(&md->type_lock));
2311 2312 2313 2314 2315
	md->type = type;
}

unsigned dm_get_md_type(struct mapped_device *md)
{
2316
	BUG_ON(!mutex_is_locked(&md->type_lock));
2317 2318 2319
	return md->type;
}

2320 2321 2322 2323 2324
struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
{
	return md->immutable_target_type;
}

2325 2326 2327 2328 2329 2330 2331 2332 2333 2334 2335
/*
 * The queue_limits are only valid as long as you have a reference
 * count on 'md'.
 */
struct queue_limits *dm_get_queue_limits(struct mapped_device *md)
{
	BUG_ON(!atomic_read(&md->holders));
	return &md->queue->limits;
}
EXPORT_SYMBOL_GPL(dm_get_queue_limits);

2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349 2350 2351 2352 2353 2354 2355 2356 2357 2358 2359 2360 2361 2362 2363 2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
/*
 * 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;
}

2376
static struct mapped_device *dm_find_md(dev_t dev)
L
Linus Torvalds 已提交
2377 2378 2379 2380 2381 2382 2383
{
	struct mapped_device *md;
	unsigned minor = MINOR(dev);

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

2384
	spin_lock(&_minor_lock);
L
Linus Torvalds 已提交
2385 2386

	md = idr_find(&_minor_idr, minor);
J
Jeff Mahoney 已提交
2387
	if (md && (md == MINOR_ALLOCED ||
2388
		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
2389
		   dm_deleting_md(md) ||
A
Alasdair G Kergon 已提交
2390
		   test_bit(DMF_FREEING, &md->flags))) {
2391
		md = NULL;
J
Jeff Mahoney 已提交
2392 2393
		goto out;
	}
L
Linus Torvalds 已提交
2394

J
Jeff Mahoney 已提交
2395
out:
2396
	spin_unlock(&_minor_lock);
L
Linus Torvalds 已提交
2397

2398 2399 2400
	return md;
}

2401 2402 2403 2404 2405 2406 2407 2408 2409
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
2410
EXPORT_SYMBOL_GPL(dm_get_md);
2411

A
Alasdair G Kergon 已提交
2412
void *dm_get_mdptr(struct mapped_device *md)
2413
{
A
Alasdair G Kergon 已提交
2414
	return md->interface_ptr;
L
Linus Torvalds 已提交
2415 2416 2417 2418 2419 2420 2421 2422 2423 2424
}

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);
2425
	BUG_ON(test_bit(DMF_FREEING, &md->flags));
L
Linus Torvalds 已提交
2426 2427
}

2428 2429 2430 2431 2432 2433
const char *dm_device_name(struct mapped_device *md)
{
	return md->name;
}
EXPORT_SYMBOL_GPL(dm_device_name);

2434
static void __dm_destroy(struct mapped_device *md, bool wait)
L
Linus Torvalds 已提交
2435
{
M
Mike Anderson 已提交
2436
	struct dm_table *map;
M
Mikulas Patocka 已提交
2437
	int srcu_idx;
L
Linus Torvalds 已提交
2438

2439
	might_sleep();
J
Jeff Mahoney 已提交
2440

2441
	spin_lock(&_minor_lock);
M
Mikulas Patocka 已提交
2442
	map = dm_get_live_table(md, &srcu_idx);
2443 2444 2445 2446 2447 2448 2449
	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
2450
	}
2451

M
Mikulas Patocka 已提交
2452 2453 2454
	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
	dm_put_live_table(md, srcu_idx);

2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 2484 2485
	/*
	 * 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_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
2486
}
E
Edward Goggin 已提交
2487
EXPORT_SYMBOL_GPL(dm_put);
L
Linus Torvalds 已提交
2488

2489
static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
2490 2491
{
	int r = 0;
2492 2493 2494
	DECLARE_WAITQUEUE(wait, current);

	add_wait_queue(&md->wait, &wait);
2495 2496

	while (1) {
2497
		set_current_state(interruptible);
2498

2499
		if (!md_in_flight(md))
2500 2501
			break;

2502 2503
		if (interruptible == TASK_INTERRUPTIBLE &&
		    signal_pending(current)) {
2504 2505 2506 2507 2508 2509 2510 2511
			r = -EINTR;
			break;
		}

		io_schedule();
	}
	set_current_state(TASK_RUNNING);

2512 2513
	remove_wait_queue(&md->wait, &wait);

2514 2515 2516
	return r;
}

L
Linus Torvalds 已提交
2517 2518 2519
/*
 * Process the deferred bios
 */
2520
static void dm_wq_work(struct work_struct *work)
L
Linus Torvalds 已提交
2521
{
2522 2523
	struct mapped_device *md = container_of(work, struct mapped_device,
						work);
2524
	struct bio *c;
M
Mikulas Patocka 已提交
2525 2526
	int srcu_idx;
	struct dm_table *map;
L
Linus Torvalds 已提交
2527

M
Mikulas Patocka 已提交
2528
	map = dm_get_live_table(md, &srcu_idx);
2529

2530
	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
A
Alasdair G Kergon 已提交
2531 2532 2533 2534
		spin_lock_irq(&md->deferred_lock);
		c = bio_list_pop(&md->deferred);
		spin_unlock_irq(&md->deferred_lock);

2535
		if (!c)
A
Alasdair G Kergon 已提交
2536
			break;
2537

K
Kiyoshi Ueda 已提交
2538 2539
		if (dm_request_based(md))
			generic_make_request(c);
2540
		else
M
Mikulas Patocka 已提交
2541
			__split_and_process_bio(md, map, c);
2542
	}
M
Milan Broz 已提交
2543

M
Mikulas Patocka 已提交
2544
	dm_put_live_table(md, srcu_idx);
L
Linus Torvalds 已提交
2545 2546
}

2547
static void dm_queue_flush(struct mapped_device *md)
2548
{
2549 2550
	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	smp_mb__after_clear_bit();
2551
	queue_work(md->wq, &md->work);
2552 2553
}

L
Linus Torvalds 已提交
2554
/*
2555
 * Swap in a new table, returning the old one for the caller to destroy.
L
Linus Torvalds 已提交
2556
 */
2557
struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
L
Linus Torvalds 已提交
2558
{
2559
	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
2560
	struct queue_limits limits;
2561
	int r;
L
Linus Torvalds 已提交
2562

2563
	mutex_lock(&md->suspend_lock);
L
Linus Torvalds 已提交
2564 2565

	/* device must be suspended */
2566
	if (!dm_suspended_md(md))
2567
		goto out;
L
Linus Torvalds 已提交
2568

2569 2570 2571 2572 2573 2574 2575
	/*
	 * 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)) {
M
Mikulas Patocka 已提交
2576
		live_map = dm_get_live_table_fast(md);
2577 2578
		if (live_map)
			limits = md->queue->limits;
M
Mikulas Patocka 已提交
2579
		dm_put_live_table_fast(md);
2580 2581
	}

2582 2583 2584 2585 2586 2587
	if (!live_map) {
		r = dm_calculate_queue_limits(table, &limits);
		if (r) {
			map = ERR_PTR(r);
			goto out;
		}
2588
	}
2589

2590
	map = __bind(md, table, &limits);
L
Linus Torvalds 已提交
2591

2592
out:
2593
	mutex_unlock(&md->suspend_lock);
2594
	return map;
L
Linus Torvalds 已提交
2595 2596 2597 2598 2599 2600
}

/*
 * Functions to lock and unlock any filesystem running on the
 * device.
 */
2601
static int lock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2602
{
2603
	int r;
L
Linus Torvalds 已提交
2604 2605

	WARN_ON(md->frozen_sb);
2606

2607
	md->frozen_sb = freeze_bdev(md->bdev);
2608
	if (IS_ERR(md->frozen_sb)) {
2609
		r = PTR_ERR(md->frozen_sb);
2610 2611
		md->frozen_sb = NULL;
		return r;
2612 2613
	}

2614 2615
	set_bit(DMF_FROZEN, &md->flags);

L
Linus Torvalds 已提交
2616 2617 2618
	return 0;
}

2619
static void unlock_fs(struct mapped_device *md)
L
Linus Torvalds 已提交
2620
{
2621 2622 2623
	if (!test_bit(DMF_FROZEN, &md->flags))
		return;

2624
	thaw_bdev(md->bdev, md->frozen_sb);
L
Linus Torvalds 已提交
2625
	md->frozen_sb = NULL;
2626
	clear_bit(DMF_FROZEN, &md->flags);
L
Linus Torvalds 已提交
2627 2628 2629 2630 2631 2632 2633 2634 2635
}

/*
 * 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.
 */
2636 2637 2638
/*
 * Suspend mechanism in request-based dm.
 *
K
Kiyoshi Ueda 已提交
2639 2640 2641
 * 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.
2642
 *
K
Kiyoshi Ueda 已提交
2643
 * To abort suspend, start the request_queue.
2644
 */
2645
int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
L
Linus Torvalds 已提交
2646
{
2647
	struct dm_table *map = NULL;
2648
	int r = 0;
2649
	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
2650
	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
L
Linus Torvalds 已提交
2651

2652
	mutex_lock(&md->suspend_lock);
2653

2654
	if (dm_suspended_md(md)) {
M
Milan Broz 已提交
2655
		r = -EINVAL;
2656
		goto out_unlock;
M
Milan Broz 已提交
2657
	}
L
Linus Torvalds 已提交
2658

M
Mikulas Patocka 已提交
2659
	map = md->map;
L
Linus Torvalds 已提交
2660

2661 2662 2663 2664 2665 2666 2667
	/*
	 * 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);

2668 2669 2670
	/* This does not get reverted if there's an error later. */
	dm_table_presuspend_targets(map);

M
Mikulas Patocka 已提交
2671
	/*
K
Kiyoshi Ueda 已提交
2672 2673 2674 2675
	 * 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
2676 2677 2678 2679
	 */
	if (!noflush && do_lockfs) {
		r = lock_fs(md);
		if (r)
M
Mikulas Patocka 已提交
2680
			goto out_unlock;
2681
	}
L
Linus Torvalds 已提交
2682 2683

	/*
2684 2685 2686 2687 2688 2689 2690
	 * 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
2691 2692 2693
	 * __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
2694
	 */
2695
	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
M
Mikulas Patocka 已提交
2696
	synchronize_srcu(&md->io_barrier);
L
Linus Torvalds 已提交
2697

2698
	/*
2699 2700
	 * Stop md->queue before flushing md->wq in case request-based
	 * dm defers requests to md->wq from md->queue.
2701
	 */
2702
	if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2703
		stop_queue(md->queue);
2704

2705 2706
	flush_workqueue(md->wq);

L
Linus Torvalds 已提交
2707
	/*
2708 2709 2710
	 * 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
2711
	 */
2712
	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
L
Linus Torvalds 已提交
2713

2714
	if (noflush)
2715
		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
M
Mikulas Patocka 已提交
2716
	synchronize_srcu(&md->io_barrier);
2717

L
Linus Torvalds 已提交
2718
	/* were we interrupted ? */
2719
	if (r < 0) {
2720
		dm_queue_flush(md);
M
Milan Broz 已提交
2721

2722
		if (dm_request_based(md))
K
Kiyoshi Ueda 已提交
2723
			start_queue(md->queue);
2724

2725
		unlock_fs(md);
M
Mikulas Patocka 已提交
2726
		goto out_unlock; /* pushback list is already flushed, so skip flush */
2727
	}
L
Linus Torvalds 已提交
2728

2729 2730 2731 2732 2733 2734
	/*
	 * 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.
	 */

2735
	set_bit(DMF_SUSPENDED, &md->flags);
2736

2737 2738
	dm_table_postsuspend_targets(map);

2739
out_unlock:
2740
	mutex_unlock(&md->suspend_lock);
2741
	return r;
L
Linus Torvalds 已提交
2742 2743 2744 2745
}

int dm_resume(struct mapped_device *md)
{
2746 2747
	int r = -EINVAL;
	struct dm_table *map = NULL;
L
Linus Torvalds 已提交
2748

2749
	mutex_lock(&md->suspend_lock);
2750
	if (!dm_suspended_md(md))
2751 2752
		goto out;

M
Mikulas Patocka 已提交
2753
	map = md->map;
2754
	if (!map || !dm_table_get_size(map))
2755
		goto out;
L
Linus Torvalds 已提交
2756

2757 2758 2759
	r = dm_table_resume_targets(map);
	if (r)
		goto out;
2760

2761
	dm_queue_flush(md);
2762

2763 2764 2765 2766 2767 2768 2769 2770
	/*
	 * 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);

2771 2772 2773 2774
	unlock_fs(md);

	clear_bit(DMF_SUSPENDED, &md->flags);

2775 2776
	r = 0;
out:
2777
	mutex_unlock(&md->suspend_lock);
2778

2779
	return r;
L
Linus Torvalds 已提交
2780 2781
}

M
Mikulas Patocka 已提交
2782 2783 2784 2785 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 2813
/*
 * Internal suspend/resume works like userspace-driven suspend. It waits
 * until all bios finish and prevents issuing new bios to the target drivers.
 * It may be used only from the kernel.
 *
 * Internal suspend holds md->suspend_lock, which prevents interaction with
 * userspace-driven suspend.
 */

void dm_internal_suspend(struct mapped_device *md)
{
	mutex_lock(&md->suspend_lock);
	if (dm_suspended_md(md))
		return;

	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
	synchronize_srcu(&md->io_barrier);
	flush_workqueue(md->wq);
	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
}

void dm_internal_resume(struct mapped_device *md)
{
	if (dm_suspended_md(md))
		goto done;

	dm_queue_flush(md);

done:
	mutex_unlock(&md->suspend_lock);
}

L
Linus Torvalds 已提交
2814 2815 2816
/*-----------------------------------------------------------------
 * Event notification.
 *---------------------------------------------------------------*/
2817
int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
M
Milan Broz 已提交
2818
		       unsigned cookie)
2819
{
M
Milan Broz 已提交
2820 2821 2822 2823
	char udev_cookie[DM_COOKIE_LENGTH];
	char *envp[] = { udev_cookie, NULL };

	if (!cookie)
2824
		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
M
Milan Broz 已提交
2825 2826 2827
	else {
		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
			 DM_COOKIE_ENV_VAR_NAME, cookie);
2828 2829
		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
					  action, envp);
M
Milan Broz 已提交
2830
	}
2831 2832
}

M
Mike Anderson 已提交
2833 2834 2835 2836 2837
uint32_t dm_next_uevent_seq(struct mapped_device *md)
{
	return atomic_add_return(1, &md->uevent_seq);
}

L
Linus Torvalds 已提交
2838 2839 2840 2841 2842 2843 2844 2845 2846 2847 2848
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
2849 2850 2851 2852 2853 2854 2855 2856 2857
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
2858 2859 2860 2861 2862 2863 2864 2865 2866
/*
 * 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
2867 2868 2869 2870 2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883
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;

2884
	if (test_bit(DMF_FREEING, &md->flags) ||
M
Mike Anderson 已提交
2885
	    dm_deleting_md(md))
2886 2887
		return NULL;

M
Milan Broz 已提交
2888 2889 2890 2891
	dm_get(md);
	return md;
}

2892
int dm_suspended_md(struct mapped_device *md)
L
Linus Torvalds 已提交
2893 2894 2895 2896
{
	return test_bit(DMF_SUSPENDED, &md->flags);
}

2897 2898
int dm_suspended(struct dm_target *ti)
{
2899
	return dm_suspended_md(dm_table_get_md(ti->table));
2900 2901 2902
}
EXPORT_SYMBOL_GPL(dm_suspended);

2903 2904
int dm_noflush_suspending(struct dm_target *ti)
{
2905
	return __noflush_suspending(dm_table_get_md(ti->table));
2906 2907 2908
}
EXPORT_SYMBOL_GPL(dm_noflush_suspending);

M
Mikulas Patocka 已提交
2909
struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size)
K
Kiyoshi Ueda 已提交
2910
{
J
Jun'ichi Nomura 已提交
2911 2912 2913 2914
	struct dm_md_mempools *pools = kzalloc(sizeof(*pools), GFP_KERNEL);
	struct kmem_cache *cachep;
	unsigned int pool_size;
	unsigned int front_pad;
K
Kiyoshi Ueda 已提交
2915 2916 2917 2918

	if (!pools)
		return NULL;

2919
	if (type == DM_TYPE_BIO_BASED) {
J
Jun'ichi Nomura 已提交
2920
		cachep = _io_cache;
2921
		pool_size = dm_get_reserved_bio_based_ios();
J
Jun'ichi Nomura 已提交
2922 2923 2924
		front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
	} else if (type == DM_TYPE_REQUEST_BASED) {
		cachep = _rq_tio_cache;
2925
		pool_size = dm_get_reserved_rq_based_ios();
J
Jun'ichi Nomura 已提交
2926 2927 2928 2929 2930
		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
		/* per_bio_data_size is not used. See __bind_mempools(). */
		WARN_ON(per_bio_data_size != 0);
	} else
		goto out;
K
Kiyoshi Ueda 已提交
2931

2932
	pools->io_pool = mempool_create_slab_pool(pool_size, cachep);
J
Jun'ichi Nomura 已提交
2933 2934
	if (!pools->io_pool)
		goto out;
K
Kiyoshi Ueda 已提交
2935

J
Jun'ichi Nomura 已提交
2936
	pools->bs = bioset_create(pool_size, front_pad);
K
Kiyoshi Ueda 已提交
2937
	if (!pools->bs)
J
Jun'ichi Nomura 已提交
2938
		goto out;
K
Kiyoshi Ueda 已提交
2939

2940
	if (integrity && bioset_integrity_create(pools->bs, pool_size))
J
Jun'ichi Nomura 已提交
2941
		goto out;
2942

K
Kiyoshi Ueda 已提交
2943 2944
	return pools;

J
Jun'ichi Nomura 已提交
2945 2946
out:
	dm_free_md_mempools(pools);
K
Kiyoshi Ueda 已提交
2947 2948 2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963 2964

	return NULL;
}

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

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

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

	kfree(pools);
}

2965
static const struct block_device_operations dm_blk_dops = {
L
Linus Torvalds 已提交
2966 2967
	.open = dm_blk_open,
	.release = dm_blk_close,
2968
	.ioctl = dm_blk_ioctl,
D
Darrick J. Wong 已提交
2969
	.getgeo = dm_blk_getgeo,
L
Linus Torvalds 已提交
2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982
	.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");
2983

2984 2985 2986
module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");

2987 2988 2989
module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");

L
Linus Torvalds 已提交
2990 2991 2992
MODULE_DESCRIPTION(DM_NAME " driver");
MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
MODULE_LICENSE("GPL");